{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.08333333333333333,
  "eval_steps": 2000,
  "global_step": 10000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 8.333333333333333e-05,
      "grad_norm": 33.25,
      "learning_rate": 0.0001,
      "loss": 7.5723,
      "loss/crossentropy": 2.127359576523304,
      "loss/hidden": 3.5390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22891336753964425,
      "step": 10
    },
    {
      "epoch": 0.00016666666666666666,
      "grad_norm": 30.875,
      "grad_norm_var": 4.769791666666666,
      "learning_rate": 0.0001,
      "loss": 7.3973,
      "loss/crossentropy": 1.9497304677963256,
      "loss/hidden": 3.56875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20572521891444923,
      "step": 20
    },
    {
      "epoch": 0.00025,
      "grad_norm": 31.5,
      "grad_norm_var": 4.0494140625,
      "learning_rate": 0.0001,
      "loss": 7.3789,
      "loss/crossentropy": 1.873328334838152,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17487502712756395,
      "step": 30
    },
    {
      "epoch": 0.0003333333333333333,
      "grad_norm": 33.75,
      "grad_norm_var": 7.137434895833334,
      "learning_rate": 0.0001,
      "loss": 7.4058,
      "loss/crossentropy": 1.9326039656996727,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20465592797845603,
      "step": 40
    },
    {
      "epoch": 0.0004166666666666667,
      "grad_norm": 35.25,
      "grad_norm_var": 3.910872395833333,
      "learning_rate": 0.0001,
      "loss": 7.4246,
      "loss/crossentropy": 1.91383947879076,
      "loss/hidden": 3.4359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20186774488538503,
      "step": 50
    },
    {
      "epoch": 0.0005,
      "grad_norm": 40.75,
      "grad_norm_var": 5.63515625,
      "learning_rate": 0.0001,
      "loss": 7.5044,
      "loss/crossentropy": 1.92522851228714,
      "loss/hidden": 3.497265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19990789387375116,
      "step": 60
    },
    {
      "epoch": 0.0005833333333333334,
      "grad_norm": 35.0,
      "grad_norm_var": 9.306266253378193e+17,
      "learning_rate": 0.0001,
      "loss": 7.357,
      "loss/crossentropy": 1.9717040583491325,
      "loss/hidden": 3.326171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1909018535166979,
      "step": 70
    },
    {
      "epoch": 0.0006666666666666666,
      "grad_norm": 33.75,
      "grad_norm_var": 9.306266259608481e+17,
      "learning_rate": 0.0001,
      "loss": 7.3788,
      "loss/crossentropy": 1.9090200901031493,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17854449450969695,
      "step": 80
    },
    {
      "epoch": 0.00075,
      "grad_norm": 33.75,
      "grad_norm_var": 4.45625,
      "learning_rate": 0.0001,
      "loss": 7.4741,
      "loss/crossentropy": 1.8572393998503685,
      "loss/hidden": 3.571875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19387567192316055,
      "step": 90
    },
    {
      "epoch": 0.0008333333333333334,
      "grad_norm": 33.25,
      "grad_norm_var": 2.949934895833333,
      "learning_rate": 0.0001,
      "loss": 7.4307,
      "loss/crossentropy": 2.076287841796875,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21235600393265486,
      "step": 100
    },
    {
      "epoch": 0.0009166666666666666,
      "grad_norm": 33.0,
      "grad_norm_var": 3.5143229166666665,
      "learning_rate": 0.0001,
      "loss": 7.4063,
      "loss/crossentropy": 2.0186501413583757,
      "loss/hidden": 3.4765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20398833509534597,
      "step": 110
    },
    {
      "epoch": 0.001,
      "grad_norm": 31.875,
      "grad_norm_var": 3.6372395833333333,
      "learning_rate": 0.0001,
      "loss": 7.4431,
      "loss/crossentropy": 1.9501568615436553,
      "loss/hidden": 3.541015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1999204397201538,
      "step": 120
    },
    {
      "epoch": 0.0010833333333333333,
      "grad_norm": 35.5,
      "grad_norm_var": 4.7837890625,
      "learning_rate": 0.0001,
      "loss": 7.3751,
      "loss/crossentropy": 2.0366190388798713,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18896052613854408,
      "step": 130
    },
    {
      "epoch": 0.0011666666666666668,
      "grad_norm": 33.25,
      "grad_norm_var": 2.43125,
      "learning_rate": 0.0001,
      "loss": 7.4741,
      "loss/crossentropy": 1.8939224407076836,
      "loss/hidden": 3.4390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19272235538810492,
      "step": 140
    },
    {
      "epoch": 0.00125,
      "grad_norm": 33.0,
      "grad_norm_var": 27.179622395833334,
      "learning_rate": 0.0001,
      "loss": 7.5153,
      "loss/crossentropy": 2.0108851701021195,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20005665346980095,
      "step": 150
    },
    {
      "epoch": 0.0013333333333333333,
      "grad_norm": 37.75,
      "grad_norm_var": 58.23951822916667,
      "learning_rate": 0.0001,
      "loss": 7.4085,
      "loss/crossentropy": 1.9532957911491393,
      "loss/hidden": 3.505078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2099555365741253,
      "step": 160
    },
    {
      "epoch": 0.0014166666666666668,
      "grad_norm": 32.25,
      "grad_norm_var": 39.82858072916667,
      "learning_rate": 0.0001,
      "loss": 7.4047,
      "loss/crossentropy": 1.9121289029717445,
      "loss/hidden": 3.439453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1977582860738039,
      "step": 170
    },
    {
      "epoch": 0.0015,
      "grad_norm": 41.25,
      "grad_norm_var": 2.3565788464023603e+18,
      "learning_rate": 0.0001,
      "loss": 7.3435,
      "loss/crossentropy": 1.9281877517700194,
      "loss/hidden": 3.50546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19753940496593714,
      "step": 180
    },
    {
      "epoch": 0.0015833333333333333,
      "grad_norm": 35.25,
      "grad_norm_var": 2.3565788455580467e+18,
      "learning_rate": 0.0001,
      "loss": 7.471,
      "loss/crossentropy": 2.032211276888847,
      "loss/hidden": 3.503515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19554367773234843,
      "step": 190
    },
    {
      "epoch": 0.0016666666666666668,
      "grad_norm": 31.5,
      "grad_norm_var": 3.676497395833333,
      "learning_rate": 0.0001,
      "loss": 7.2708,
      "loss/crossentropy": 2.005619902908802,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1833203172311187,
      "step": 200
    },
    {
      "epoch": 0.00175,
      "grad_norm": 30.75,
      "grad_norm_var": 8.633268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4096,
      "loss/crossentropy": 2.043836605548859,
      "loss/hidden": 3.467578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20938539411872625,
      "step": 210
    },
    {
      "epoch": 0.0018333333333333333,
      "grad_norm": 33.25,
      "grad_norm_var": 23.039322916666666,
      "learning_rate": 0.0001,
      "loss": 7.3709,
      "loss/crossentropy": 2.054931232333183,
      "loss/hidden": 3.4609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20825725365430117,
      "step": 220
    },
    {
      "epoch": 0.0019166666666666666,
      "grad_norm": 32.75,
      "grad_norm_var": 1.0732816604826787e+18,
      "learning_rate": 0.0001,
      "loss": 7.3838,
      "loss/crossentropy": 1.9404870495200157,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20280153322964906,
      "step": 230
    },
    {
      "epoch": 0.002,
      "grad_norm": 30.375,
      "grad_norm_var": 12.29765625,
      "learning_rate": 0.0001,
      "loss": 7.3121,
      "loss/crossentropy": 1.8668742343783378,
      "loss/hidden": 3.47578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21291877124458553,
      "step": 240
    },
    {
      "epoch": 0.0020833333333333333,
      "grad_norm": 31.25,
      "grad_norm_var": 56.5400390625,
      "learning_rate": 0.0001,
      "loss": 7.3905,
      "loss/crossentropy": 1.9617563866078853,
      "loss/hidden": 3.51640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1919050351716578,
      "step": 250
    },
    {
      "epoch": 0.0021666666666666666,
      "grad_norm": 32.5,
      "grad_norm_var": 47.96640625,
      "learning_rate": 0.0001,
      "loss": 7.3528,
      "loss/crossentropy": 1.8532208397984504,
      "loss/hidden": 3.551171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19277248289436102,
      "step": 260
    },
    {
      "epoch": 0.00225,
      "grad_norm": 32.0,
      "grad_norm_var": 18.443684895833332,
      "learning_rate": 0.0001,
      "loss": 7.3309,
      "loss/crossentropy": 2.0671744480729104,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19330325871706008,
      "step": 270
    },
    {
      "epoch": 0.0023333333333333335,
      "grad_norm": 41.0,
      "grad_norm_var": 23.93125,
      "learning_rate": 0.0001,
      "loss": 7.2214,
      "loss/crossentropy": 1.9675503507256509,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18025756757706404,
      "step": 280
    },
    {
      "epoch": 0.002416666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 17.74765625,
      "learning_rate": 0.0001,
      "loss": 7.2949,
      "loss/crossentropy": 1.9656010001897812,
      "loss/hidden": 3.479296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20177473574876786,
      "step": 290
    },
    {
      "epoch": 0.0025,
      "grad_norm": 43.5,
      "grad_norm_var": 4.011194233137573e+17,
      "learning_rate": 0.0001,
      "loss": 7.4422,
      "loss/crossentropy": 1.9836331263184548,
      "loss/hidden": 3.30546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1909252552315593,
      "step": 300
    },
    {
      "epoch": 0.0025833333333333333,
      "grad_norm": 29.125,
      "grad_norm_var": 4.0111942329792384e+17,
      "learning_rate": 0.0001,
      "loss": 7.4124,
      "loss/crossentropy": 2.082610194385052,
      "loss/hidden": 3.462890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21536452919244767,
      "step": 310
    },
    {
      "epoch": 0.0026666666666666666,
      "grad_norm": 32.5,
      "grad_norm_var": 5.5056640625,
      "learning_rate": 0.0001,
      "loss": 7.3504,
      "loss/crossentropy": 1.9580433249473572,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1865969082340598,
      "step": 320
    },
    {
      "epoch": 0.00275,
      "grad_norm": 34.25,
      "grad_norm_var": 7.421809895833333,
      "learning_rate": 0.0001,
      "loss": 7.2673,
      "loss/crossentropy": 2.0938573241233827,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21014094576239586,
      "step": 330
    },
    {
      "epoch": 0.0028333333333333335,
      "grad_norm": 30.625,
      "grad_norm_var": 5.9494140625,
      "learning_rate": 0.0001,
      "loss": 7.3136,
      "loss/crossentropy": 1.8406853549182416,
      "loss/hidden": 3.399609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1884180039167404,
      "step": 340
    },
    {
      "epoch": 0.002916666666666667,
      "grad_norm": 29.625,
      "grad_norm_var": 14.85625,
      "learning_rate": 0.0001,
      "loss": 7.3683,
      "loss/crossentropy": 1.972689888626337,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20450354916974903,
      "step": 350
    },
    {
      "epoch": 0.003,
      "grad_norm": 35.0,
      "grad_norm_var": 17.483072916666668,
      "learning_rate": 0.0001,
      "loss": 7.3037,
      "loss/crossentropy": 1.995607402920723,
      "loss/hidden": 3.412890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20541023295372723,
      "step": 360
    },
    {
      "epoch": 0.0030833333333333333,
      "grad_norm": 35.0,
      "grad_norm_var": 6.827351345607127e+17,
      "learning_rate": 0.0001,
      "loss": 7.4571,
      "loss/crossentropy": 1.9991235852241516,
      "loss/hidden": 3.4359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20315398182719946,
      "step": 370
    },
    {
      "epoch": 0.0031666666666666666,
      "grad_norm": 36.0,
      "grad_norm_var": 3.971875,
      "learning_rate": 0.0001,
      "loss": 7.3758,
      "loss/crossentropy": 2.0666207425296306,
      "loss/hidden": 3.555859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20830214060842991,
      "step": 380
    },
    {
      "epoch": 0.00325,
      "grad_norm": 34.25,
      "grad_norm_var": 109.1759765625,
      "learning_rate": 0.0001,
      "loss": 7.4163,
      "loss/crossentropy": 1.942155936360359,
      "loss/hidden": 3.51171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19059792384505272,
      "step": 390
    },
    {
      "epoch": 0.0033333333333333335,
      "grad_norm": 29.0,
      "grad_norm_var": 10.253125,
      "learning_rate": 0.0001,
      "loss": 7.24,
      "loss/crossentropy": 1.8737600088119506,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19837535824626684,
      "step": 400
    },
    {
      "epoch": 0.003416666666666667,
      "grad_norm": 29.625,
      "grad_norm_var": 9.715284576287676e+17,
      "learning_rate": 0.0001,
      "loss": 7.3633,
      "loss/crossentropy": 2.0632002711296082,
      "loss/hidden": 3.73828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22990943398326635,
      "step": 410
    },
    {
      "epoch": 0.0035,
      "grad_norm": 31.125,
      "grad_norm_var": 9.715284568977354e+17,
      "learning_rate": 0.0001,
      "loss": 7.3609,
      "loss/crossentropy": 2.0630046755075453,
      "loss/hidden": 3.31484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19341134056448936,
      "step": 420
    },
    {
      "epoch": 0.0035833333333333333,
      "grad_norm": 28.75,
      "grad_norm_var": 3.5770833333333334,
      "learning_rate": 0.0001,
      "loss": 7.3068,
      "loss/crossentropy": 1.9697899878025056,
      "loss/hidden": 3.48515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20131503157317637,
      "step": 430
    },
    {
      "epoch": 0.0036666666666666666,
      "grad_norm": 32.25,
      "grad_norm_var": 6.013997395833333,
      "learning_rate": 0.0001,
      "loss": 7.4332,
      "loss/crossentropy": 1.7330221965909005,
      "loss/hidden": 3.5515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2054268393665552,
      "step": 440
    },
    {
      "epoch": 0.00375,
      "grad_norm": 34.0,
      "grad_norm_var": 6.847330729166667,
      "learning_rate": 0.0001,
      "loss": 7.3632,
      "loss/crossentropy": 1.9284149013459682,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18791294917464257,
      "step": 450
    },
    {
      "epoch": 0.003833333333333333,
      "grad_norm": 30.25,
      "grad_norm_var": 7.338541666666667,
      "learning_rate": 0.0001,
      "loss": 7.4044,
      "loss/crossentropy": 1.9736344426870347,
      "loss/hidden": 3.544921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20589777827262878,
      "step": 460
    },
    {
      "epoch": 0.003916666666666666,
      "grad_norm": 33.75,
      "grad_norm_var": 6.7884765625,
      "learning_rate": 0.0001,
      "loss": 7.437,
      "loss/crossentropy": 1.9907344430685043,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21705251093953848,
      "step": 470
    },
    {
      "epoch": 0.004,
      "grad_norm": 29.75,
      "grad_norm_var": 20.412434895833332,
      "learning_rate": 0.0001,
      "loss": 7.2693,
      "loss/crossentropy": 1.9349011391401292,
      "loss/hidden": 3.591015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21220431197434664,
      "step": 480
    },
    {
      "epoch": 0.004083333333333333,
      "grad_norm": 43.0,
      "grad_norm_var": 27.826822916666668,
      "learning_rate": 0.0001,
      "loss": 7.3501,
      "loss/crossentropy": 1.980695901811123,
      "loss/hidden": 3.42265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19862887468189,
      "step": 490
    },
    {
      "epoch": 0.004166666666666667,
      "grad_norm": 34.5,
      "grad_norm_var": 18.60390625,
      "learning_rate": 0.0001,
      "loss": 7.3467,
      "loss/crossentropy": 2.033171671628952,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20085344444960357,
      "step": 500
    },
    {
      "epoch": 0.00425,
      "grad_norm": 32.0,
      "grad_norm_var": 21.980208333333334,
      "learning_rate": 0.0001,
      "loss": 7.2657,
      "loss/crossentropy": 1.915445376932621,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20700605846941472,
      "step": 510
    },
    {
      "epoch": 0.004333333333333333,
      "grad_norm": 42.5,
      "grad_norm_var": 22.976822916666666,
      "learning_rate": 0.0001,
      "loss": 7.413,
      "loss/crossentropy": 1.9141934983432294,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19435560163110494,
      "step": 520
    },
    {
      "epoch": 0.004416666666666667,
      "grad_norm": 29.625,
      "grad_norm_var": 24.456184895833335,
      "learning_rate": 0.0001,
      "loss": 7.2444,
      "loss/crossentropy": 2.0416490510106087,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19158751983195543,
      "step": 530
    },
    {
      "epoch": 0.0045,
      "grad_norm": 33.25,
      "grad_norm_var": 18.768489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3152,
      "loss/crossentropy": 2.004468746483326,
      "loss/hidden": 3.362890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1849455814808607,
      "step": 540
    },
    {
      "epoch": 0.004583333333333333,
      "grad_norm": 38.75,
      "grad_norm_var": 22.685416666666665,
      "learning_rate": 0.0001,
      "loss": 7.363,
      "loss/crossentropy": 2.100373297929764,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20310039576143027,
      "step": 550
    },
    {
      "epoch": 0.004666666666666667,
      "grad_norm": 65.5,
      "grad_norm_var": 87.36354166666666,
      "learning_rate": 0.0001,
      "loss": 7.4567,
      "loss/crossentropy": 1.9387385457754136,
      "loss/hidden": 3.570703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21101371981203557,
      "step": 560
    },
    {
      "epoch": 0.00475,
      "grad_norm": 35.0,
      "grad_norm_var": 68.48899739583334,
      "learning_rate": 0.0001,
      "loss": 7.3065,
      "loss/crossentropy": 1.8911512687802314,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19550031386315822,
      "step": 570
    },
    {
      "epoch": 0.004833333333333334,
      "grad_norm": 29.125,
      "grad_norm_var": 10.197916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3044,
      "loss/crossentropy": 2.0782967150211333,
      "loss/hidden": 3.488671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20857007205486297,
      "step": 580
    },
    {
      "epoch": 0.004916666666666666,
      "grad_norm": 31.75,
      "grad_norm_var": 9.111393229166667,
      "learning_rate": 0.0001,
      "loss": 7.3021,
      "loss/crossentropy": 1.9965147748589516,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19327867813408375,
      "step": 590
    },
    {
      "epoch": 0.005,
      "grad_norm": 32.0,
      "grad_norm_var": 12.423893229166667,
      "learning_rate": 0.0001,
      "loss": 7.3111,
      "loss/crossentropy": 1.9416659727692605,
      "loss/hidden": 3.4671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19690530523657798,
      "step": 600
    },
    {
      "epoch": 0.005083333333333333,
      "grad_norm": 35.25,
      "grad_norm_var": 17.788541666666667,
      "learning_rate": 0.0001,
      "loss": 7.2461,
      "loss/crossentropy": 1.8854174569249154,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19389428701251746,
      "step": 610
    },
    {
      "epoch": 0.005166666666666667,
      "grad_norm": 31.625,
      "grad_norm_var": 19.061458333333334,
      "learning_rate": 0.0001,
      "loss": 7.3305,
      "loss/crossentropy": 1.9388531804084779,
      "loss/hidden": 3.534765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1920976996421814,
      "step": 620
    },
    {
      "epoch": 0.00525,
      "grad_norm": 33.25,
      "grad_norm_var": 15.886393229166666,
      "learning_rate": 0.0001,
      "loss": 7.3302,
      "loss/crossentropy": 2.0138145625591277,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1938948707655072,
      "step": 630
    },
    {
      "epoch": 0.005333333333333333,
      "grad_norm": 46.75,
      "grad_norm_var": 27.868489583333332,
      "learning_rate": 0.0001,
      "loss": 7.2544,
      "loss/crossentropy": 2.0624490082263947,
      "loss/hidden": 3.35703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1968511760234833,
      "step": 640
    },
    {
      "epoch": 0.005416666666666667,
      "grad_norm": 31.0,
      "grad_norm_var": 29.6291015625,
      "learning_rate": 0.0001,
      "loss": 7.3249,
      "loss/crossentropy": 1.910299976170063,
      "loss/hidden": 3.54765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20261993119493127,
      "step": 650
    },
    {
      "epoch": 0.0055,
      "grad_norm": 51.5,
      "grad_norm_var": 43.927083333333336,
      "learning_rate": 0.0001,
      "loss": 7.3305,
      "loss/crossentropy": 1.8762669041752815,
      "loss/hidden": 3.3125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17283952739089728,
      "step": 660
    },
    {
      "epoch": 0.005583333333333333,
      "grad_norm": 29.375,
      "grad_norm_var": 43.02233072916667,
      "learning_rate": 0.0001,
      "loss": 7.238,
      "loss/crossentropy": 1.886153695732355,
      "loss/hidden": 3.46328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1988078447058797,
      "step": 670
    },
    {
      "epoch": 0.005666666666666667,
      "grad_norm": 33.25,
      "grad_norm_var": 49.3494140625,
      "learning_rate": 0.0001,
      "loss": 7.3053,
      "loss/crossentropy": 2.0484144926071166,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20717981327325105,
      "step": 680
    },
    {
      "epoch": 0.00575,
      "grad_norm": 39.25,
      "grad_norm_var": 39.08639322916667,
      "learning_rate": 0.0001,
      "loss": 7.299,
      "loss/crossentropy": 1.8476220905780791,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1946946369484067,
      "step": 690
    },
    {
      "epoch": 0.005833333333333334,
      "grad_norm": 31.75,
      "grad_norm_var": 6.162239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3848,
      "loss/crossentropy": 1.9601008757948875,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1941233342513442,
      "step": 700
    },
    {
      "epoch": 0.005916666666666666,
      "grad_norm": 29.5,
      "grad_norm_var": 109.7134765625,
      "learning_rate": 0.0001,
      "loss": 7.2428,
      "loss/crossentropy": 1.945550049096346,
      "loss/hidden": 3.311328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20260647609829902,
      "step": 710
    },
    {
      "epoch": 0.006,
      "grad_norm": 42.75,
      "grad_norm_var": 68.07291666666667,
      "learning_rate": 0.0001,
      "loss": 7.3668,
      "loss/crossentropy": 1.9872938007116319,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2058260628953576,
      "step": 720
    },
    {
      "epoch": 0.006083333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 35.67858072916667,
      "learning_rate": 0.0001,
      "loss": 7.2796,
      "loss/crossentropy": 2.049102196097374,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19005358330905436,
      "step": 730
    },
    {
      "epoch": 0.006166666666666667,
      "grad_norm": 31.875,
      "grad_norm_var": 13.388541666666667,
      "learning_rate": 0.0001,
      "loss": 7.3165,
      "loss/crossentropy": 2.000891861319542,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18923768866807222,
      "step": 740
    },
    {
      "epoch": 0.00625,
      "grad_norm": 31.5,
      "grad_norm_var": 1.78515625,
      "learning_rate": 0.0001,
      "loss": 7.2249,
      "loss/crossentropy": 1.9180716082453728,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19066319558769465,
      "step": 750
    },
    {
      "epoch": 0.006333333333333333,
      "grad_norm": 30.125,
      "grad_norm_var": 6.3947265625,
      "learning_rate": 0.0001,
      "loss": 7.2432,
      "loss/crossentropy": 1.9366105154156685,
      "loss/hidden": 3.309375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1816173456609249,
      "step": 760
    },
    {
      "epoch": 0.006416666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 6.420833333333333,
      "learning_rate": 0.0001,
      "loss": 7.375,
      "loss/crossentropy": 1.9630885288119315,
      "loss/hidden": 3.5234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20541108381003143,
      "step": 770
    },
    {
      "epoch": 0.0065,
      "grad_norm": 30.625,
      "grad_norm_var": 5.01875,
      "learning_rate": 0.0001,
      "loss": 7.2215,
      "loss/crossentropy": 1.8655893057584763,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17294289451092482,
      "step": 780
    },
    {
      "epoch": 0.006583333333333333,
      "grad_norm": 31.5,
      "grad_norm_var": 31.75390625,
      "learning_rate": 0.0001,
      "loss": 7.3612,
      "loss/crossentropy": 2.009425923228264,
      "loss/hidden": 3.426953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19349737111479043,
      "step": 790
    },
    {
      "epoch": 0.006666666666666667,
      "grad_norm": 28.875,
      "grad_norm_var": 256.80358072916664,
      "learning_rate": 0.0001,
      "loss": 7.2728,
      "loss/crossentropy": 1.9748301401734352,
      "loss/hidden": 3.28359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1777269348502159,
      "step": 800
    },
    {
      "epoch": 0.00675,
      "grad_norm": 32.5,
      "grad_norm_var": 13.159309895833333,
      "learning_rate": 0.0001,
      "loss": 7.3129,
      "loss/crossentropy": 2.023061712086201,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18830966539680957,
      "step": 810
    },
    {
      "epoch": 0.006833333333333334,
      "grad_norm": 35.0,
      "grad_norm_var": 13.510872395833333,
      "learning_rate": 0.0001,
      "loss": 7.1579,
      "loss/crossentropy": 2.0136666923761366,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22291016448289155,
      "step": 820
    },
    {
      "epoch": 0.0069166666666666664,
      "grad_norm": 33.0,
      "grad_norm_var": 9.578125,
      "learning_rate": 0.0001,
      "loss": 7.3571,
      "loss/crossentropy": 1.9204814687371254,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20071878507733346,
      "step": 830
    },
    {
      "epoch": 0.007,
      "grad_norm": 30.0,
      "grad_norm_var": 12.862239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3315,
      "loss/crossentropy": 1.894421537220478,
      "loss/hidden": 3.394140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19227557741105555,
      "step": 840
    },
    {
      "epoch": 0.007083333333333333,
      "grad_norm": 30.875,
      "grad_norm_var": 9.105143229166666,
      "learning_rate": 0.0001,
      "loss": 7.1733,
      "loss/crossentropy": 1.917093026638031,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18238474596291782,
      "step": 850
    },
    {
      "epoch": 0.007166666666666667,
      "grad_norm": 38.5,
      "grad_norm_var": 9.84375,
      "learning_rate": 0.0001,
      "loss": 7.2598,
      "loss/crossentropy": 1.971535401046276,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19669556841254235,
      "step": 860
    },
    {
      "epoch": 0.00725,
      "grad_norm": 32.5,
      "grad_norm_var": 11.320572916666666,
      "learning_rate": 0.0001,
      "loss": 7.2687,
      "loss/crossentropy": 2.0588334918022158,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20278499722480775,
      "step": 870
    },
    {
      "epoch": 0.007333333333333333,
      "grad_norm": 36.25,
      "grad_norm_var": 9.912239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2785,
      "loss/crossentropy": 2.085460352897644,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2105460435152054,
      "step": 880
    },
    {
      "epoch": 0.007416666666666667,
      "grad_norm": 37.75,
      "grad_norm_var": 22.637434895833334,
      "learning_rate": 0.0001,
      "loss": 7.3492,
      "loss/crossentropy": 1.9137649670243264,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21051024934276938,
      "step": 890
    },
    {
      "epoch": 0.0075,
      "grad_norm": 33.5,
      "grad_norm_var": 5.320572916666666,
      "learning_rate": 0.0001,
      "loss": 7.2658,
      "loss/crossentropy": 1.9578298330307007,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18825461007654667,
      "step": 900
    },
    {
      "epoch": 0.007583333333333333,
      "grad_norm": 33.75,
      "grad_norm_var": 2.7322265625,
      "learning_rate": 0.0001,
      "loss": 7.283,
      "loss/crossentropy": 1.92996806204319,
      "loss/hidden": 3.50078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2107054302468896,
      "step": 910
    },
    {
      "epoch": 0.007666666666666666,
      "grad_norm": 31.0,
      "grad_norm_var": 5.770572916666667,
      "learning_rate": 0.0001,
      "loss": 7.3248,
      "loss/crossentropy": 2.0229760840535165,
      "loss/hidden": 3.50859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2188296400010586,
      "step": 920
    },
    {
      "epoch": 0.00775,
      "grad_norm": 33.5,
      "grad_norm_var": 5.871875,
      "learning_rate": 0.0001,
      "loss": 7.2833,
      "loss/crossentropy": 1.9864106953144074,
      "loss/hidden": 3.520703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2264138963073492,
      "step": 930
    },
    {
      "epoch": 0.007833333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 6.97265625,
      "learning_rate": 0.0001,
      "loss": 7.2909,
      "loss/crossentropy": 1.871564555168152,
      "loss/hidden": 3.64296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20268587870523333,
      "step": 940
    },
    {
      "epoch": 0.007916666666666667,
      "grad_norm": 36.25,
      "grad_norm_var": 6.064518229166667,
      "learning_rate": 0.0001,
      "loss": 7.2552,
      "loss/crossentropy": 1.8941823691129684,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19341870285570623,
      "step": 950
    },
    {
      "epoch": 0.008,
      "grad_norm": 31.75,
      "grad_norm_var": 4.970768229166667,
      "learning_rate": 0.0001,
      "loss": 7.2977,
      "loss/crossentropy": 1.9714406847953796,
      "loss/hidden": 3.262109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1801678713411093,
      "step": 960
    },
    {
      "epoch": 0.008083333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 49.87083333333333,
      "learning_rate": 0.0001,
      "loss": 7.5097,
      "loss/crossentropy": 1.8991910368204117,
      "loss/hidden": 3.495703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19257726836949587,
      "step": 970
    },
    {
      "epoch": 0.008166666666666666,
      "grad_norm": 32.25,
      "grad_norm_var": 1.0388009758452352e+18,
      "learning_rate": 0.0001,
      "loss": 7.2699,
      "loss/crossentropy": 1.9593139976263045,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18533678650856017,
      "step": 980
    },
    {
      "epoch": 0.00825,
      "grad_norm": 29.25,
      "grad_norm_var": 1.038800976104286e+18,
      "learning_rate": 0.0001,
      "loss": 7.233,
      "loss/crossentropy": 1.9886707931756973,
      "loss/hidden": 3.484765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2002451216802001,
      "step": 990
    },
    {
      "epoch": 0.008333333333333333,
      "grad_norm": 31.25,
      "grad_norm_var": 7.783072916666667,
      "learning_rate": 0.0001,
      "loss": 7.2139,
      "loss/crossentropy": 1.982450208067894,
      "loss/hidden": 3.299609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19080037772655487,
      "step": 1000
    },
    {
      "epoch": 0.008416666666666666,
      "grad_norm": 38.75,
      "grad_norm_var": 7.674739583333333,
      "learning_rate": 0.0001,
      "loss": 7.2816,
      "loss/crossentropy": 1.8640219122171402,
      "loss/hidden": 3.4703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18669026363641023,
      "step": 1010
    },
    {
      "epoch": 0.0085,
      "grad_norm": 36.5,
      "grad_norm_var": 6.824934895833334,
      "learning_rate": 0.0001,
      "loss": 7.2353,
      "loss/crossentropy": 1.8924013078212738,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1948640437796712,
      "step": 1020
    },
    {
      "epoch": 0.008583333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 3.4916666666666667,
      "learning_rate": 0.0001,
      "loss": 7.4108,
      "loss/crossentropy": 1.9207781434059144,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21109477896243334,
      "step": 1030
    },
    {
      "epoch": 0.008666666666666666,
      "grad_norm": 39.75,
      "grad_norm_var": 7.131705729166667,
      "learning_rate": 0.0001,
      "loss": 7.378,
      "loss/crossentropy": 1.9148639217019081,
      "loss/hidden": 3.352734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19126542042940856,
      "step": 1040
    },
    {
      "epoch": 0.00875,
      "grad_norm": 30.875,
      "grad_norm_var": 8.465625,
      "learning_rate": 0.0001,
      "loss": 7.3071,
      "loss/crossentropy": 1.9637104764580726,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1933687962591648,
      "step": 1050
    },
    {
      "epoch": 0.008833333333333334,
      "grad_norm": 31.5,
      "grad_norm_var": 3.6837890625,
      "learning_rate": 0.0001,
      "loss": 7.2891,
      "loss/crossentropy": 1.898938211798668,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20170777514576912,
      "step": 1060
    },
    {
      "epoch": 0.008916666666666666,
      "grad_norm": 36.0,
      "grad_norm_var": 3.20625,
      "learning_rate": 0.0001,
      "loss": 7.3074,
      "loss/crossentropy": 1.875780703127384,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1741238312795758,
      "step": 1070
    },
    {
      "epoch": 0.009,
      "grad_norm": 30.875,
      "grad_norm_var": 4.985872395833334,
      "learning_rate": 0.0001,
      "loss": 7.2715,
      "loss/crossentropy": 1.8494450084865093,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1831109117716551,
      "step": 1080
    },
    {
      "epoch": 0.009083333333333334,
      "grad_norm": 31.625,
      "grad_norm_var": 29.108072916666668,
      "learning_rate": 0.0001,
      "loss": 7.3007,
      "loss/crossentropy": 1.9538781769573688,
      "loss/hidden": 3.47890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19717481937259435,
      "step": 1090
    },
    {
      "epoch": 0.009166666666666667,
      "grad_norm": 28.875,
      "grad_norm_var": 3.83515625,
      "learning_rate": 0.0001,
      "loss": 7.2203,
      "loss/crossentropy": 1.9142641603946686,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19166908524930476,
      "step": 1100
    },
    {
      "epoch": 0.00925,
      "grad_norm": 37.5,
      "grad_norm_var": 6.734830729166666,
      "learning_rate": 0.0001,
      "loss": 7.326,
      "loss/crossentropy": 2.016929093003273,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18157481551170349,
      "step": 1110
    },
    {
      "epoch": 0.009333333333333334,
      "grad_norm": 35.0,
      "grad_norm_var": 4.9652184708252966e+17,
      "learning_rate": 0.0001,
      "loss": 7.4337,
      "loss/crossentropy": 1.982716566324234,
      "loss/hidden": 3.396875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18657856304198503,
      "step": 1120
    },
    {
      "epoch": 0.009416666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 4.965218471471219e+17,
      "learning_rate": 0.0001,
      "loss": 7.1922,
      "loss/crossentropy": 1.8523512050509452,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19456181898713112,
      "step": 1130
    },
    {
      "epoch": 0.0095,
      "grad_norm": 30.375,
      "grad_norm_var": 1.073281661622271e+18,
      "learning_rate": 0.0001,
      "loss": 7.2354,
      "loss/crossentropy": 1.9637323051691056,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18722216449677945,
      "step": 1140
    },
    {
      "epoch": 0.009583333333333333,
      "grad_norm": 33.5,
      "grad_norm_var": 2.786393229166667,
      "learning_rate": 0.0001,
      "loss": 7.2889,
      "loss/crossentropy": 1.9170468106865883,
      "loss/hidden": 3.496875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19212216939777135,
      "step": 1150
    },
    {
      "epoch": 0.009666666666666667,
      "grad_norm": 32.75,
      "grad_norm_var": 2.0218098958333335,
      "learning_rate": 0.0001,
      "loss": 7.2555,
      "loss/crossentropy": 2.1359424561262133,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1815421387553215,
      "step": 1160
    },
    {
      "epoch": 0.00975,
      "grad_norm": 35.0,
      "grad_norm_var": 2.94765625,
      "learning_rate": 0.0001,
      "loss": 7.277,
      "loss/crossentropy": 1.827228417992592,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1859055045992136,
      "step": 1170
    },
    {
      "epoch": 0.009833333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 3.2119140625,
      "learning_rate": 0.0001,
      "loss": 7.2983,
      "loss/crossentropy": 1.9235713258385658,
      "loss/hidden": 3.4140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20665957890450953,
      "step": 1180
    },
    {
      "epoch": 0.009916666666666667,
      "grad_norm": 28.875,
      "grad_norm_var": 1.3067057291666666,
      "learning_rate": 0.0001,
      "loss": 7.2995,
      "loss/crossentropy": 2.06373308300972,
      "loss/hidden": 3.301953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894452802836895,
      "step": 1190
    },
    {
      "epoch": 0.01,
      "grad_norm": 31.875,
      "grad_norm_var": 9.71528458302303e+17,
      "learning_rate": 0.0001,
      "loss": 7.366,
      "loss/crossentropy": 1.9264583349227906,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1899071641266346,
      "step": 1200
    },
    {
      "epoch": 0.010083333333333333,
      "grad_norm": 30.375,
      "grad_norm_var": 9.71528457661623e+17,
      "learning_rate": 0.0001,
      "loss": 7.3836,
      "loss/crossentropy": 1.7987317427992822,
      "loss/hidden": 3.514453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20049565955996512,
      "step": 1210
    },
    {
      "epoch": 0.010166666666666666,
      "grad_norm": 32.75,
      "grad_norm_var": 1.9603515625,
      "learning_rate": 0.0001,
      "loss": 7.2874,
      "loss/crossentropy": 1.7559522911906242,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18231481197290123,
      "step": 1220
    },
    {
      "epoch": 0.01025,
      "grad_norm": 37.75,
      "grad_norm_var": 4.79765625,
      "learning_rate": 0.0001,
      "loss": 7.2641,
      "loss/crossentropy": 1.9591181308031083,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18080000337213278,
      "step": 1230
    },
    {
      "epoch": 0.010333333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 5.617122395833333,
      "learning_rate": 0.0001,
      "loss": 7.2299,
      "loss/crossentropy": 1.996937209367752,
      "loss/hidden": 3.291796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17810146119445563,
      "step": 1240
    },
    {
      "epoch": 0.010416666666666666,
      "grad_norm": 33.25,
      "grad_norm_var": 7.272916666666666,
      "learning_rate": 0.0001,
      "loss": 7.31,
      "loss/crossentropy": 1.8335382163524627,
      "loss/hidden": 3.55078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1968407742679119,
      "step": 1250
    },
    {
      "epoch": 0.0105,
      "grad_norm": 34.25,
      "grad_norm_var": 17.586393229166667,
      "learning_rate": 0.0001,
      "loss": 7.2393,
      "loss/crossentropy": 2.059291285276413,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19821931160986422,
      "step": 1260
    },
    {
      "epoch": 0.010583333333333333,
      "grad_norm": 32.5,
      "grad_norm_var": 2.60390625,
      "learning_rate": 0.0001,
      "loss": 7.2797,
      "loss/crossentropy": 1.9161242499947548,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1930157072842121,
      "step": 1270
    },
    {
      "epoch": 0.010666666666666666,
      "grad_norm": 34.5,
      "grad_norm_var": 5.16015625,
      "learning_rate": 0.0001,
      "loss": 7.2758,
      "loss/crossentropy": 1.946441213786602,
      "loss/hidden": 3.425,
      "loss/jsd": 0.0,
      "loss/logits": 0.2005026711151004,
      "step": 1280
    },
    {
      "epoch": 0.01075,
      "grad_norm": 28.75,
      "grad_norm_var": 3.9330729166666667,
      "learning_rate": 0.0001,
      "loss": 7.2913,
      "loss/crossentropy": 1.981220117211342,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20416617318987845,
      "step": 1290
    },
    {
      "epoch": 0.010833333333333334,
      "grad_norm": 35.0,
      "grad_norm_var": 6.75390625,
      "learning_rate": 0.0001,
      "loss": 7.284,
      "loss/crossentropy": 2.0280979484319688,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17846004143357277,
      "step": 1300
    },
    {
      "epoch": 0.010916666666666667,
      "grad_norm": 31.25,
      "grad_norm_var": 6.831705729166667,
      "learning_rate": 0.0001,
      "loss": 7.2516,
      "loss/crossentropy": 1.9297500923275948,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18526345416903495,
      "step": 1310
    },
    {
      "epoch": 0.011,
      "grad_norm": 36.25,
      "grad_norm_var": 5.120247395833333,
      "learning_rate": 0.0001,
      "loss": 7.278,
      "loss/crossentropy": 1.9518141075968742,
      "loss/hidden": 3.47265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2055202412419021,
      "step": 1320
    },
    {
      "epoch": 0.011083333333333334,
      "grad_norm": 32.75,
      "grad_norm_var": 10.774739583333334,
      "learning_rate": 0.0001,
      "loss": 7.2781,
      "loss/crossentropy": 2.0175456278026105,
      "loss/hidden": 3.473828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20037419367581605,
      "step": 1330
    },
    {
      "epoch": 0.011166666666666667,
      "grad_norm": 32.0,
      "grad_norm_var": 7.321164001851736e+17,
      "learning_rate": 0.0001,
      "loss": 7.4105,
      "loss/crossentropy": 1.9731741398572922,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19264885168522597,
      "step": 1340
    },
    {
      "epoch": 0.01125,
      "grad_norm": 34.5,
      "grad_norm_var": 7.32116400149522e+17,
      "learning_rate": 0.0001,
      "loss": 7.2418,
      "loss/crossentropy": 2.0094937115907667,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1830705888569355,
      "step": 1350
    },
    {
      "epoch": 0.011333333333333334,
      "grad_norm": 30.75,
      "grad_norm_var": 6.511393229166667,
      "learning_rate": 0.0001,
      "loss": 7.2534,
      "loss/crossentropy": 1.8466748759150504,
      "loss/hidden": 3.255859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17569661382585763,
      "step": 1360
    },
    {
      "epoch": 0.011416666666666667,
      "grad_norm": 27.75,
      "grad_norm_var": 6.679622395833333,
      "learning_rate": 0.0001,
      "loss": 7.1384,
      "loss/crossentropy": 1.933292143791914,
      "loss/hidden": 3.423828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18499917536973953,
      "step": 1370
    },
    {
      "epoch": 0.0115,
      "grad_norm": 33.0,
      "grad_norm_var": 17.3541015625,
      "learning_rate": 0.0001,
      "loss": 7.3339,
      "loss/crossentropy": 1.9056864947080612,
      "loss/hidden": 3.523828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22226835750043392,
      "step": 1380
    },
    {
      "epoch": 0.011583333333333333,
      "grad_norm": 30.75,
      "grad_norm_var": 12.066666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3259,
      "loss/crossentropy": 1.971571609377861,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19764023572206496,
      "step": 1390
    },
    {
      "epoch": 0.011666666666666667,
      "grad_norm": 31.0,
      "grad_norm_var": 2.26015625,
      "learning_rate": 0.0001,
      "loss": 7.3098,
      "loss/crossentropy": 1.8674658387899399,
      "loss/hidden": 3.403515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18480249140411614,
      "step": 1400
    },
    {
      "epoch": 0.01175,
      "grad_norm": 42.0,
      "grad_norm_var": 1.3401023452742157e+18,
      "learning_rate": 0.0001,
      "loss": 7.3637,
      "loss/crossentropy": 2.1348548874258997,
      "loss/hidden": 3.63984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23163463342934848,
      "step": 1410
    },
    {
      "epoch": 0.011833333333333333,
      "grad_norm": 31.75,
      "grad_norm_var": 1.3401023450764544e+18,
      "learning_rate": 0.0001,
      "loss": 7.2733,
      "loss/crossentropy": 2.0169719740748406,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19758520759642123,
      "step": 1420
    },
    {
      "epoch": 0.011916666666666667,
      "grad_norm": 35.75,
      "grad_norm_var": 2.9155598958333333,
      "learning_rate": 0.0001,
      "loss": 7.256,
      "loss/crossentropy": 2.0090753018856047,
      "loss/hidden": 3.27578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18631072752177716,
      "step": 1430
    },
    {
      "epoch": 0.012,
      "grad_norm": 30.375,
      "grad_norm_var": 6.63515625,
      "learning_rate": 0.0001,
      "loss": 7.3997,
      "loss/crossentropy": 2.062183755636215,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17819378785789014,
      "step": 1440
    },
    {
      "epoch": 0.012083333333333333,
      "grad_norm": 36.25,
      "grad_norm_var": 7.536458333333333,
      "learning_rate": 0.0001,
      "loss": 7.1829,
      "loss/crossentropy": 1.9344430059194564,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17964769918471574,
      "step": 1450
    },
    {
      "epoch": 0.012166666666666666,
      "grad_norm": 33.5,
      "grad_norm_var": 18.2009765625,
      "learning_rate": 0.0001,
      "loss": 7.3129,
      "loss/crossentropy": 1.9684132590889931,
      "loss/hidden": 3.333203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17724689934402704,
      "step": 1460
    },
    {
      "epoch": 0.01225,
      "grad_norm": 30.875,
      "grad_norm_var": 57.805989583333336,
      "learning_rate": 0.0001,
      "loss": 7.2262,
      "loss/crossentropy": 1.9477881044149399,
      "loss/hidden": 3.336328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18544685766100882,
      "step": 1470
    },
    {
      "epoch": 0.012333333333333333,
      "grad_norm": 29.125,
      "grad_norm_var": 4.303059895833333,
      "learning_rate": 0.0001,
      "loss": 7.2255,
      "loss/crossentropy": 2.0414924576878546,
      "loss/hidden": 3.444140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19923711474984884,
      "step": 1480
    },
    {
      "epoch": 0.012416666666666666,
      "grad_norm": 32.75,
      "grad_norm_var": 37.46555989583333,
      "learning_rate": 0.0001,
      "loss": 7.2586,
      "loss/crossentropy": 1.9072564639151097,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19603695515543224,
      "step": 1490
    },
    {
      "epoch": 0.0125,
      "grad_norm": 31.875,
      "grad_norm_var": 5.642708333333333,
      "learning_rate": 0.0001,
      "loss": 7.3078,
      "loss/crossentropy": 1.9207553781569005,
      "loss/hidden": 3.49609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19159257756546139,
      "step": 1500
    },
    {
      "epoch": 0.012583333333333334,
      "grad_norm": 45.0,
      "grad_norm_var": 12.5,
      "learning_rate": 0.0001,
      "loss": 7.2107,
      "loss/crossentropy": 1.908722074329853,
      "loss/hidden": 3.325390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19376476481556892,
      "step": 1510
    },
    {
      "epoch": 0.012666666666666666,
      "grad_norm": 35.25,
      "grad_norm_var": 65.01295572916666,
      "learning_rate": 0.0001,
      "loss": 7.3764,
      "loss/crossentropy": 1.9983658343553543,
      "loss/hidden": 3.370703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18891664631664754,
      "step": 1520
    },
    {
      "epoch": 0.01275,
      "grad_norm": 33.5,
      "grad_norm_var": 45.13645833333333,
      "learning_rate": 0.0001,
      "loss": 7.2604,
      "loss/crossentropy": 1.9507597625255584,
      "loss/hidden": 3.37109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20500100627541543,
      "step": 1530
    },
    {
      "epoch": 0.012833333333333334,
      "grad_norm": 30.875,
      "grad_norm_var": 12.14140625,
      "learning_rate": 0.0001,
      "loss": 7.3091,
      "loss/crossentropy": 2.0078801713883876,
      "loss/hidden": 3.383203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18963296040892602,
      "step": 1540
    },
    {
      "epoch": 0.012916666666666667,
      "grad_norm": 33.0,
      "grad_norm_var": 9.196809895833333,
      "learning_rate": 0.0001,
      "loss": 7.2892,
      "loss/crossentropy": 1.9844240590929985,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20112392045557498,
      "step": 1550
    },
    {
      "epoch": 0.013,
      "grad_norm": 29.5,
      "grad_norm_var": 4.860416666666667,
      "learning_rate": 0.0001,
      "loss": 7.2492,
      "loss/crossentropy": 1.9080413848161697,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1872928399592638,
      "step": 1560
    },
    {
      "epoch": 0.013083333333333334,
      "grad_norm": 30.375,
      "grad_norm_var": 17.445768229166667,
      "learning_rate": 0.0001,
      "loss": 7.4383,
      "loss/crossentropy": 2.0557202830910684,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1941663671284914,
      "step": 1570
    },
    {
      "epoch": 0.013166666666666667,
      "grad_norm": 33.5,
      "grad_norm_var": 25.912239583333335,
      "learning_rate": 0.0001,
      "loss": 7.3577,
      "loss/crossentropy": 2.019923450052738,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19055009353905916,
      "step": 1580
    },
    {
      "epoch": 0.01325,
      "grad_norm": 32.75,
      "grad_norm_var": 7.486393229166667,
      "learning_rate": 0.0001,
      "loss": 7.3418,
      "loss/crossentropy": 1.9912660360336303,
      "loss/hidden": 3.425,
      "loss/jsd": 0.0,
      "loss/logits": 0.19967138823121786,
      "step": 1590
    },
    {
      "epoch": 0.013333333333333334,
      "grad_norm": 29.5,
      "grad_norm_var": 13.739322916666667,
      "learning_rate": 0.0001,
      "loss": 7.329,
      "loss/crossentropy": 1.976361881196499,
      "loss/hidden": 3.4578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1988378567621112,
      "step": 1600
    },
    {
      "epoch": 0.013416666666666667,
      "grad_norm": 30.0,
      "grad_norm_var": 8.120768229166666,
      "learning_rate": 0.0001,
      "loss": 7.3197,
      "loss/crossentropy": 1.9507541894912719,
      "loss/hidden": 3.448046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1952985545620322,
      "step": 1610
    },
    {
      "epoch": 0.0135,
      "grad_norm": 30.875,
      "grad_norm_var": 6.485183333490755e+17,
      "learning_rate": 0.0001,
      "loss": 7.3644,
      "loss/crossentropy": 2.0464282095432282,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19348717443645,
      "step": 1620
    },
    {
      "epoch": 0.013583333333333333,
      "grad_norm": 31.625,
      "grad_norm_var": 5.680143229166666,
      "learning_rate": 0.0001,
      "loss": 7.233,
      "loss/crossentropy": 1.9385505408048629,
      "loss/hidden": 3.330078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1836165890097618,
      "step": 1630
    },
    {
      "epoch": 0.013666666666666667,
      "grad_norm": 30.625,
      "grad_norm_var": 2.2791015625,
      "learning_rate": 0.0001,
      "loss": 7.1957,
      "loss/crossentropy": 1.939904248714447,
      "loss/hidden": 3.254296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.178390958532691,
      "step": 1640
    },
    {
      "epoch": 0.01375,
      "grad_norm": 30.125,
      "grad_norm_var": 2.909375,
      "learning_rate": 0.0001,
      "loss": 7.1277,
      "loss/crossentropy": 1.8794662043452264,
      "loss/hidden": 3.32421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18380587548017502,
      "step": 1650
    },
    {
      "epoch": 0.013833333333333333,
      "grad_norm": 28.875,
      "grad_norm_var": 2.2900390625,
      "learning_rate": 0.0001,
      "loss": 7.2484,
      "loss/crossentropy": 1.9310500830411912,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18547237031161784,
      "step": 1660
    },
    {
      "epoch": 0.013916666666666667,
      "grad_norm": 33.25,
      "grad_norm_var": 6.414322916666666,
      "learning_rate": 0.0001,
      "loss": 7.2795,
      "loss/crossentropy": 1.9204590037465095,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19697155160829424,
      "step": 1670
    },
    {
      "epoch": 0.014,
      "grad_norm": 36.0,
      "grad_norm_var": 4.1650390625,
      "learning_rate": 0.0001,
      "loss": 7.2973,
      "loss/crossentropy": 2.122091555595398,
      "loss/hidden": 3.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18852671552449465,
      "step": 1680
    },
    {
      "epoch": 0.014083333333333333,
      "grad_norm": 33.25,
      "grad_norm_var": 6.1994140625,
      "learning_rate": 0.0001,
      "loss": 7.3361,
      "loss/crossentropy": 2.156261298060417,
      "loss/hidden": 3.516015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20998501256108285,
      "step": 1690
    },
    {
      "epoch": 0.014166666666666666,
      "grad_norm": 30.125,
      "grad_norm_var": 5.687955729166666,
      "learning_rate": 0.0001,
      "loss": 7.2512,
      "loss/crossentropy": 1.9670209422707559,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20212535988539457,
      "step": 1700
    },
    {
      "epoch": 0.01425,
      "grad_norm": 31.75,
      "grad_norm_var": 3.2947265625,
      "learning_rate": 0.0001,
      "loss": 7.3058,
      "loss/crossentropy": 2.0846331417560577,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20701619572937488,
      "step": 1710
    },
    {
      "epoch": 0.014333333333333333,
      "grad_norm": 30.0,
      "grad_norm_var": 3.5455729166666665,
      "learning_rate": 0.0001,
      "loss": 7.3898,
      "loss/crossentropy": 2.029135060310364,
      "loss/hidden": 3.43125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19423509035259484,
      "step": 1720
    },
    {
      "epoch": 0.014416666666666666,
      "grad_norm": 32.75,
      "grad_norm_var": 25.8306640625,
      "learning_rate": 0.0001,
      "loss": 7.2523,
      "loss/crossentropy": 1.8986969135701657,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18308270815759897,
      "step": 1730
    },
    {
      "epoch": 0.0145,
      "grad_norm": 35.5,
      "grad_norm_var": 8.847330729166666,
      "learning_rate": 0.0001,
      "loss": 7.385,
      "loss/crossentropy": 1.9802353143692017,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20486843958497047,
      "step": 1740
    },
    {
      "epoch": 0.014583333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 4.5166015625,
      "learning_rate": 0.0001,
      "loss": 7.2707,
      "loss/crossentropy": 1.8174491569399833,
      "loss/hidden": 3.43125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18351891580969096,
      "step": 1750
    },
    {
      "epoch": 0.014666666666666666,
      "grad_norm": 30.375,
      "grad_norm_var": 2.098893229166667,
      "learning_rate": 0.0001,
      "loss": 7.279,
      "loss/crossentropy": 2.161423406004906,
      "loss/hidden": 3.280859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18401132225990297,
      "step": 1760
    },
    {
      "epoch": 0.01475,
      "grad_norm": 29.75,
      "grad_norm_var": 5.0384765625,
      "learning_rate": 0.0001,
      "loss": 7.2269,
      "loss/crossentropy": 1.8804115176200866,
      "loss/hidden": 3.401171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18950734920799733,
      "step": 1770
    },
    {
      "epoch": 0.014833333333333334,
      "grad_norm": 34.75,
      "grad_norm_var": 4.3917132166325434e+17,
      "learning_rate": 0.0001,
      "loss": 7.321,
      "loss/crossentropy": 1.9431498855352403,
      "loss/hidden": 3.61484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20938508119434118,
      "step": 1780
    },
    {
      "epoch": 0.014916666666666667,
      "grad_norm": 48.0,
      "grad_norm_var": 4.3917132079346054e+17,
      "learning_rate": 0.0001,
      "loss": 7.2738,
      "loss/crossentropy": 1.991785955429077,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19057564698159696,
      "step": 1790
    },
    {
      "epoch": 0.015,
      "grad_norm": 30.875,
      "grad_norm_var": 32.218684895833334,
      "learning_rate": 0.0001,
      "loss": 7.3331,
      "loss/crossentropy": 1.9663019806146622,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18586504831910133,
      "step": 1800
    },
    {
      "epoch": 0.015083333333333334,
      "grad_norm": 30.0,
      "grad_norm_var": 3.289322916666667,
      "learning_rate": 0.0001,
      "loss": 7.2361,
      "loss/crossentropy": 1.9916753873229027,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20035373903810977,
      "step": 1810
    },
    {
      "epoch": 0.015166666666666667,
      "grad_norm": 35.25,
      "grad_norm_var": 3.5205729166666666,
      "learning_rate": 0.0001,
      "loss": 7.1769,
      "loss/crossentropy": 2.0380251288414,
      "loss/hidden": 3.321875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18559953831136228,
      "step": 1820
    },
    {
      "epoch": 0.01525,
      "grad_norm": 32.0,
      "grad_norm_var": 9.2384765625,
      "learning_rate": 0.0001,
      "loss": 7.318,
      "loss/crossentropy": 1.9261852458119393,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18253942150622607,
      "step": 1830
    },
    {
      "epoch": 0.015333333333333332,
      "grad_norm": 30.125,
      "grad_norm_var": 7.781705729166666,
      "learning_rate": 0.0001,
      "loss": 7.2674,
      "loss/crossentropy": 1.8436972185969354,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18778917007148266,
      "step": 1840
    },
    {
      "epoch": 0.015416666666666667,
      "grad_norm": 31.875,
      "grad_norm_var": 1.9494140625,
      "learning_rate": 0.0001,
      "loss": 7.3212,
      "loss/crossentropy": 2.020959322154522,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19789048228412867,
      "step": 1850
    },
    {
      "epoch": 0.0155,
      "grad_norm": 37.25,
      "grad_norm_var": 31.046809895833334,
      "learning_rate": 0.0001,
      "loss": 7.2136,
      "loss/crossentropy": 1.8541840478777885,
      "loss/hidden": 3.3453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17573605831712485,
      "step": 1860
    },
    {
      "epoch": 0.015583333333333333,
      "grad_norm": 31.875,
      "grad_norm_var": 32.603580729166666,
      "learning_rate": 0.0001,
      "loss": 7.2691,
      "loss/crossentropy": 1.9017258927226066,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17979437448084354,
      "step": 1870
    },
    {
      "epoch": 0.015666666666666666,
      "grad_norm": 31.125,
      "grad_norm_var": 4.289518229166666,
      "learning_rate": 0.0001,
      "loss": 7.2647,
      "loss/crossentropy": 1.9352346003055572,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19231825936585664,
      "step": 1880
    },
    {
      "epoch": 0.01575,
      "grad_norm": 35.75,
      "grad_norm_var": 1.6906090573603364e+18,
      "learning_rate": 0.0001,
      "loss": 7.4439,
      "loss/crossentropy": 1.943103864789009,
      "loss/hidden": 3.44765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1879008635878563,
      "step": 1890
    },
    {
      "epoch": 0.015833333333333335,
      "grad_norm": 32.0,
      "grad_norm_var": 1.6906090571002895e+18,
      "learning_rate": 0.0001,
      "loss": 7.3468,
      "loss/crossentropy": 1.9841052293777466,
      "loss/hidden": 3.466796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19862290378659964,
      "step": 1900
    },
    {
      "epoch": 0.015916666666666666,
      "grad_norm": 32.25,
      "grad_norm_var": 295.4098307291667,
      "learning_rate": 0.0001,
      "loss": 7.4047,
      "loss/crossentropy": 1.9083608210086822,
      "loss/hidden": 3.30625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18380118031054735,
      "step": 1910
    },
    {
      "epoch": 0.016,
      "grad_norm": 30.25,
      "grad_norm_var": 342.01458333333335,
      "learning_rate": 0.0001,
      "loss": 7.3901,
      "loss/crossentropy": 2.0147466957569122,
      "loss/hidden": 3.38671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1957575250416994,
      "step": 1920
    },
    {
      "epoch": 0.016083333333333335,
      "grad_norm": 31.5,
      "grad_norm_var": 7.322916666666667,
      "learning_rate": 0.0001,
      "loss": 7.3132,
      "loss/crossentropy": 1.8240422800183296,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18423110404983162,
      "step": 1930
    },
    {
      "epoch": 0.016166666666666666,
      "grad_norm": 33.25,
      "grad_norm_var": 21.651822916666667,
      "learning_rate": 0.0001,
      "loss": 7.2571,
      "loss/crossentropy": 1.9640603423118592,
      "loss/hidden": 3.316796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19020091257989408,
      "step": 1940
    },
    {
      "epoch": 0.01625,
      "grad_norm": 38.75,
      "grad_norm_var": 23.867122395833334,
      "learning_rate": 0.0001,
      "loss": 7.2404,
      "loss/crossentropy": 1.7818415738642215,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17666226997971535,
      "step": 1950
    },
    {
      "epoch": 0.01633333333333333,
      "grad_norm": 34.0,
      "grad_norm_var": 4.885416666666667,
      "learning_rate": 0.0001,
      "loss": 7.34,
      "loss/crossentropy": 1.9440691128373147,
      "loss/hidden": 3.605859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21677861418575048,
      "step": 1960
    },
    {
      "epoch": 0.016416666666666666,
      "grad_norm": 36.75,
      "grad_norm_var": 5.0841416521866726e+17,
      "learning_rate": 0.0001,
      "loss": 7.3199,
      "loss/crossentropy": 1.976449340581894,
      "loss/hidden": 3.5921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2107225891202688,
      "step": 1970
    },
    {
      "epoch": 0.0165,
      "grad_norm": 30.875,
      "grad_norm_var": 5.084141640391941e+17,
      "learning_rate": 0.0001,
      "loss": 7.389,
      "loss/crossentropy": 2.0012796193361284,
      "loss/hidden": 3.39140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19476055447012186,
      "step": 1980
    },
    {
      "epoch": 0.016583333333333332,
      "grad_norm": 33.25,
      "grad_norm_var": 69.82805989583333,
      "learning_rate": 0.0001,
      "loss": 7.2807,
      "loss/crossentropy": 2.057659167051315,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19637791439890862,
      "step": 1990
    },
    {
      "epoch": 0.016666666666666666,
      "grad_norm": 33.75,
      "grad_norm_var": 2.4427083333333335,
      "learning_rate": 0.0001,
      "loss": 7.2848,
      "loss/crossentropy": 2.1016288846731186,
      "loss/hidden": 3.500390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19531550351530313,
      "step": 2000
    },
    {
      "epoch": 0.01675,
      "grad_norm": 30.0,
      "grad_norm_var": 4.6884765625,
      "learning_rate": 0.0001,
      "loss": 7.153,
      "loss/crossentropy": 1.9537018239498138,
      "loss/hidden": 3.43203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18602741174399853,
      "step": 2010
    },
    {
      "epoch": 0.016833333333333332,
      "grad_norm": 31.375,
      "grad_norm_var": 3.06015625,
      "learning_rate": 0.0001,
      "loss": 7.2727,
      "loss/crossentropy": 1.9659764975309373,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1949648916721344,
      "step": 2020
    },
    {
      "epoch": 0.016916666666666667,
      "grad_norm": 33.0,
      "grad_norm_var": 2.5697265625,
      "learning_rate": 0.0001,
      "loss": 7.2785,
      "loss/crossentropy": 1.8973619878292083,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18199929129332304,
      "step": 2030
    },
    {
      "epoch": 0.017,
      "grad_norm": 44.25,
      "grad_norm_var": 14.552018229166666,
      "learning_rate": 0.0001,
      "loss": 7.2118,
      "loss/crossentropy": 1.9548767402768135,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19267449919134377,
      "step": 2040
    },
    {
      "epoch": 0.017083333333333332,
      "grad_norm": 30.625,
      "grad_norm_var": 12.8181640625,
      "learning_rate": 0.0001,
      "loss": 7.3086,
      "loss/crossentropy": 1.9166896522045136,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18487014006823302,
      "step": 2050
    },
    {
      "epoch": 0.017166666666666667,
      "grad_norm": 37.5,
      "grad_norm_var": 7.750455729166666,
      "learning_rate": 0.0001,
      "loss": 7.2843,
      "loss/crossentropy": 1.8707634150981902,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18515090448781849,
      "step": 2060
    },
    {
      "epoch": 0.01725,
      "grad_norm": 38.75,
      "grad_norm_var": 9.956184895833333,
      "learning_rate": 0.0001,
      "loss": 7.2942,
      "loss/crossentropy": 1.8519902683794498,
      "loss/hidden": 3.48046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1862996906042099,
      "step": 2070
    },
    {
      "epoch": 0.017333333333333333,
      "grad_norm": 30.125,
      "grad_norm_var": 5.755989583333333,
      "learning_rate": 0.0001,
      "loss": 7.246,
      "loss/crossentropy": 1.939876636862755,
      "loss/hidden": 3.439453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1866559438407421,
      "step": 2080
    },
    {
      "epoch": 0.017416666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 11.2634765625,
      "learning_rate": 0.0001,
      "loss": 7.4043,
      "loss/crossentropy": 1.95171078145504,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19473332334309817,
      "step": 2090
    },
    {
      "epoch": 0.0175,
      "grad_norm": 33.25,
      "grad_norm_var": 8.5150390625,
      "learning_rate": 0.0001,
      "loss": 7.1663,
      "loss/crossentropy": 1.997160328924656,
      "loss/hidden": 3.2921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17636931519955396,
      "step": 2100
    },
    {
      "epoch": 0.017583333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 5.155989583333334,
      "learning_rate": 0.0001,
      "loss": 7.2116,
      "loss/crossentropy": 1.9414543956518173,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18860510028898717,
      "step": 2110
    },
    {
      "epoch": 0.017666666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 4.614322916666667,
      "learning_rate": 0.0001,
      "loss": 7.196,
      "loss/crossentropy": 2.0807690426707266,
      "loss/hidden": 3.3625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18818104062229396,
      "step": 2120
    },
    {
      "epoch": 0.01775,
      "grad_norm": 34.25,
      "grad_norm_var": 6.62265625,
      "learning_rate": 0.0001,
      "loss": 7.2894,
      "loss/crossentropy": 1.9585611373186111,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19511018712073563,
      "step": 2130
    },
    {
      "epoch": 0.017833333333333333,
      "grad_norm": 38.5,
      "grad_norm_var": 11.9634765625,
      "learning_rate": 0.0001,
      "loss": 7.276,
      "loss/crossentropy": 1.942204162478447,
      "loss/hidden": 3.32265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18475140817463398,
      "step": 2140
    },
    {
      "epoch": 0.017916666666666668,
      "grad_norm": 30.75,
      "grad_norm_var": 9.11875,
      "learning_rate": 0.0001,
      "loss": 7.2854,
      "loss/crossentropy": 1.9810814321041108,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18507389035075902,
      "step": 2150
    },
    {
      "epoch": 0.018,
      "grad_norm": 33.75,
      "grad_norm_var": 7.000455729166666,
      "learning_rate": 0.0001,
      "loss": 7.2508,
      "loss/crossentropy": 2.141017886996269,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19872815869748592,
      "step": 2160
    },
    {
      "epoch": 0.018083333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 4.131184895833333,
      "learning_rate": 0.0001,
      "loss": 7.3377,
      "loss/crossentropy": 1.9983078092336655,
      "loss/hidden": 3.45390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18931595589965583,
      "step": 2170
    },
    {
      "epoch": 0.018166666666666668,
      "grad_norm": 30.5,
      "grad_norm_var": 5.385416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3045,
      "loss/crossentropy": 1.9222343221306801,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17565734032541513,
      "step": 2180
    },
    {
      "epoch": 0.01825,
      "grad_norm": 31.375,
      "grad_norm_var": 6.27890625,
      "learning_rate": 0.0001,
      "loss": 7.2599,
      "loss/crossentropy": 2.0131379783153536,
      "loss/hidden": 3.276953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18543861098587514,
      "step": 2190
    },
    {
      "epoch": 0.018333333333333333,
      "grad_norm": 31.25,
      "grad_norm_var": 2.69765625,
      "learning_rate": 0.0001,
      "loss": 7.185,
      "loss/crossentropy": 1.8784870684146882,
      "loss/hidden": 3.304296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19702841751277447,
      "step": 2200
    },
    {
      "epoch": 0.018416666666666668,
      "grad_norm": 31.875,
      "grad_norm_var": 6.620995007756109e+17,
      "learning_rate": 0.0001,
      "loss": 7.3567,
      "loss/crossentropy": 1.9045794017612934,
      "loss/hidden": 3.304296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889135385863483,
      "step": 2210
    },
    {
      "epoch": 0.0185,
      "grad_norm": 34.5,
      "grad_norm_var": 6.620995005009888e+17,
      "learning_rate": 0.0001,
      "loss": 7.2957,
      "loss/crossentropy": 1.7774473994970321,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16910541616380215,
      "step": 2220
    },
    {
      "epoch": 0.018583333333333334,
      "grad_norm": 33.0,
      "grad_norm_var": 13.5150390625,
      "learning_rate": 0.0001,
      "loss": 7.3112,
      "loss/crossentropy": 2.0893754065036774,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19306899029761554,
      "step": 2230
    },
    {
      "epoch": 0.018666666666666668,
      "grad_norm": 31.25,
      "grad_norm_var": 2.10390625,
      "learning_rate": 0.0001,
      "loss": 7.3108,
      "loss/crossentropy": 1.9501435413956643,
      "loss/hidden": 3.2921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1907532515004277,
      "step": 2240
    },
    {
      "epoch": 0.01875,
      "grad_norm": 32.0,
      "grad_norm_var": 5.5739081064856954e+17,
      "learning_rate": 0.0001,
      "loss": 7.3127,
      "loss/crossentropy": 1.936707255989313,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19168335273861886,
      "step": 2250
    },
    {
      "epoch": 0.018833333333333334,
      "grad_norm": 30.875,
      "grad_norm_var": 4.310872395833333,
      "learning_rate": 0.0001,
      "loss": 7.2447,
      "loss/crossentropy": 2.0652841836214066,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22945739645510912,
      "step": 2260
    },
    {
      "epoch": 0.018916666666666665,
      "grad_norm": 34.75,
      "grad_norm_var": 4.591666666666667,
      "learning_rate": 0.0001,
      "loss": 7.2527,
      "loss/crossentropy": 1.8847783967852592,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17899461332708597,
      "step": 2270
    },
    {
      "epoch": 0.019,
      "grad_norm": 33.5,
      "grad_norm_var": 6.779622395833333,
      "learning_rate": 0.0001,
      "loss": 7.2593,
      "loss/crossentropy": 1.8318727254867553,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1841424321755767,
      "step": 2280
    },
    {
      "epoch": 0.019083333333333334,
      "grad_norm": 31.5,
      "grad_norm_var": 4.3056640625,
      "learning_rate": 0.0001,
      "loss": 7.2545,
      "loss/crossentropy": 1.9013791278004646,
      "loss/hidden": 3.4734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20086494665592908,
      "step": 2290
    },
    {
      "epoch": 0.019166666666666665,
      "grad_norm": 31.875,
      "grad_norm_var": 2.9077473958333333,
      "learning_rate": 0.0001,
      "loss": 7.3744,
      "loss/crossentropy": 1.7958017572760583,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2134784586727619,
      "step": 2300
    },
    {
      "epoch": 0.01925,
      "grad_norm": 33.0,
      "grad_norm_var": 2.1228515625,
      "learning_rate": 0.0001,
      "loss": 7.3052,
      "loss/crossentropy": 2.032643774151802,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19537682719528676,
      "step": 2310
    },
    {
      "epoch": 0.019333333333333334,
      "grad_norm": 30.25,
      "grad_norm_var": 3.6864583333333334,
      "learning_rate": 0.0001,
      "loss": 7.2865,
      "loss/crossentropy": 1.9596822246909142,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18629487175494433,
      "step": 2320
    },
    {
      "epoch": 0.019416666666666665,
      "grad_norm": 36.25,
      "grad_norm_var": 4.873372395833333,
      "learning_rate": 0.0001,
      "loss": 7.3174,
      "loss/crossentropy": 2.0085081800818445,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19248725334182382,
      "step": 2330
    },
    {
      "epoch": 0.0195,
      "grad_norm": 36.75,
      "grad_norm_var": 6.5744140625,
      "learning_rate": 0.0001,
      "loss": 7.3592,
      "loss/crossentropy": 1.9844809271395207,
      "loss/hidden": 3.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18588697481900454,
      "step": 2340
    },
    {
      "epoch": 0.019583333333333335,
      "grad_norm": 38.5,
      "grad_norm_var": 1.2824703428670262e+18,
      "learning_rate": 0.0001,
      "loss": 7.3201,
      "loss/crossentropy": 1.9626279659569263,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18211031127721072,
      "step": 2350
    },
    {
      "epoch": 0.019666666666666666,
      "grad_norm": 30.0,
      "grad_norm_var": 1.282470342291358e+18,
      "learning_rate": 0.0001,
      "loss": 7.3481,
      "loss/crossentropy": 2.0168631345033647,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20021111145615578,
      "step": 2360
    },
    {
      "epoch": 0.01975,
      "grad_norm": 31.375,
      "grad_norm_var": 6.151811406201619e+17,
      "learning_rate": 0.0001,
      "loss": 7.2642,
      "loss/crossentropy": 2.0320428051054478,
      "loss/hidden": 3.651171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17399079585447907,
      "step": 2370
    },
    {
      "epoch": 0.019833333333333335,
      "grad_norm": 28.125,
      "grad_norm_var": 7.537955729166667,
      "learning_rate": 0.0001,
      "loss": 7.1327,
      "loss/crossentropy": 2.011017915606499,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19383390806615353,
      "step": 2380
    },
    {
      "epoch": 0.019916666666666666,
      "grad_norm": 30.0,
      "grad_norm_var": 8.124934895833333,
      "learning_rate": 0.0001,
      "loss": 7.3095,
      "loss/crossentropy": 1.9657955452799798,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18745789155364037,
      "step": 2390
    },
    {
      "epoch": 0.02,
      "grad_norm": 34.75,
      "grad_norm_var": 6.0822265625,
      "learning_rate": 0.0001,
      "loss": 7.2852,
      "loss/crossentropy": 1.9604591697454452,
      "loss/hidden": 3.470703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20235047750175,
      "step": 2400
    },
    {
      "epoch": 0.020083333333333335,
      "grad_norm": 32.75,
      "grad_norm_var": 28.183072916666667,
      "learning_rate": 0.0001,
      "loss": 7.2632,
      "loss/crossentropy": 1.944566234201193,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18141100630164148,
      "step": 2410
    },
    {
      "epoch": 0.020166666666666666,
      "grad_norm": 31.0,
      "grad_norm_var": 30.720572916666665,
      "learning_rate": 0.0001,
      "loss": 7.2207,
      "loss/crossentropy": 1.992261415719986,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20662249848246575,
      "step": 2420
    },
    {
      "epoch": 0.02025,
      "grad_norm": 31.375,
      "grad_norm_var": 3.2684895833333334,
      "learning_rate": 0.0001,
      "loss": 7.3458,
      "loss/crossentropy": 1.891785178333521,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19321308471262455,
      "step": 2430
    },
    {
      "epoch": 0.02033333333333333,
      "grad_norm": 4664066048.0,
      "grad_norm_var": 1.3595944877801428e+18,
      "learning_rate": 0.0001,
      "loss": 7.1966,
      "loss/crossentropy": 1.934097670018673,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1767245288938284,
      "step": 2440
    },
    {
      "epoch": 0.020416666666666666,
      "grad_norm": 30.625,
      "grad_norm_var": 1.3595944861768701e+18,
      "learning_rate": 0.0001,
      "loss": 7.3028,
      "loss/crossentropy": 1.9819417625665665,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1829627426341176,
      "step": 2450
    },
    {
      "epoch": 0.0205,
      "grad_norm": 31.75,
      "grad_norm_var": 3.872330729166667,
      "learning_rate": 0.0001,
      "loss": 7.1355,
      "loss/crossentropy": 1.9402824997901917,
      "loss/hidden": 3.478125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19555577524006368,
      "step": 2460
    },
    {
      "epoch": 0.020583333333333332,
      "grad_norm": 32.75,
      "grad_norm_var": 2.3962890625,
      "learning_rate": 0.0001,
      "loss": 7.1806,
      "loss/crossentropy": 1.8212218895554542,
      "loss/hidden": 3.303125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18379593417048454,
      "step": 2470
    },
    {
      "epoch": 0.020666666666666667,
      "grad_norm": 33.0,
      "grad_norm_var": 8.39140625,
      "learning_rate": 0.0001,
      "loss": 7.3467,
      "loss/crossentropy": 1.9318528160452844,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22001873292028903,
      "step": 2480
    },
    {
      "epoch": 0.02075,
      "grad_norm": 48.75,
      "grad_norm_var": 23.338541666666668,
      "learning_rate": 0.0001,
      "loss": 7.26,
      "loss/crossentropy": 1.7162697166204453,
      "loss/hidden": 3.55859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2028669610619545,
      "step": 2490
    },
    {
      "epoch": 0.020833333333333332,
      "grad_norm": 29.875,
      "grad_norm_var": 25.046875,
      "learning_rate": 0.0001,
      "loss": 7.0735,
      "loss/crossentropy": 1.9614305034279824,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1880598498508334,
      "step": 2500
    },
    {
      "epoch": 0.020916666666666667,
      "grad_norm": 48.25,
      "grad_norm_var": 7.393116035913265e+17,
      "learning_rate": 0.0001,
      "loss": 7.3996,
      "loss/crossentropy": 2.056662857532501,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19303689561784268,
      "step": 2510
    },
    {
      "epoch": 0.021,
      "grad_norm": 30.125,
      "grad_norm_var": 7.393116032760547e+17,
      "learning_rate": 0.0001,
      "loss": 7.3348,
      "loss/crossentropy": 2.052212342619896,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18852751776576043,
      "step": 2520
    },
    {
      "epoch": 0.021083333333333332,
      "grad_norm": 30.0,
      "grad_norm_var": 9.699739583333333,
      "learning_rate": 0.0001,
      "loss": 7.2244,
      "loss/crossentropy": 1.9503659322857856,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18111072592437266,
      "step": 2530
    },
    {
      "epoch": 0.021166666666666667,
      "grad_norm": 31.0,
      "grad_norm_var": 6.4875,
      "learning_rate": 0.0001,
      "loss": 7.1369,
      "loss/crossentropy": 1.96889967918396,
      "loss/hidden": 3.255859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18848092965781688,
      "step": 2540
    },
    {
      "epoch": 0.02125,
      "grad_norm": 31.0,
      "grad_norm_var": 5.251497395833334,
      "learning_rate": 0.0001,
      "loss": 7.2827,
      "loss/crossentropy": 1.9892879366874694,
      "loss/hidden": 3.458203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1890705717727542,
      "step": 2550
    },
    {
      "epoch": 0.021333333333333333,
      "grad_norm": 30.125,
      "grad_norm_var": 2.1483723958333334,
      "learning_rate": 0.0001,
      "loss": 7.3539,
      "loss/crossentropy": 1.9124898374080659,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17518300358206035,
      "step": 2560
    },
    {
      "epoch": 0.021416666666666667,
      "grad_norm": 36.25,
      "grad_norm_var": 2.5650390625,
      "learning_rate": 0.0001,
      "loss": 7.2363,
      "loss/crossentropy": 1.976254241168499,
      "loss/hidden": 3.612109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2144654095172882,
      "step": 2570
    },
    {
      "epoch": 0.0215,
      "grad_norm": 31.375,
      "grad_norm_var": 5.50625,
      "learning_rate": 0.0001,
      "loss": 7.3128,
      "loss/crossentropy": 1.9184046685695648,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21248191315680742,
      "step": 2580
    },
    {
      "epoch": 0.021583333333333333,
      "grad_norm": 32.5,
      "grad_norm_var": 3331.639583333333,
      "learning_rate": 0.0001,
      "loss": 7.298,
      "loss/crossentropy": 1.9172766536474228,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1916635401546955,
      "step": 2590
    },
    {
      "epoch": 0.021666666666666667,
      "grad_norm": 34.0,
      "grad_norm_var": 3321.0608723958335,
      "learning_rate": 0.0001,
      "loss": 7.2846,
      "loss/crossentropy": 1.9668544963002206,
      "loss/hidden": 3.433984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1907737607136369,
      "step": 2600
    },
    {
      "epoch": 0.02175,
      "grad_norm": 38.5,
      "grad_norm_var": 7.981398732307433e+17,
      "learning_rate": 0.0001,
      "loss": 7.4946,
      "loss/crossentropy": 1.9127167880535125,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1960796787403524,
      "step": 2610
    },
    {
      "epoch": 0.021833333333333333,
      "grad_norm": 30.0,
      "grad_norm_var": 7.98139873078123e+17,
      "learning_rate": 0.0001,
      "loss": 7.3146,
      "loss/crossentropy": 1.996176090836525,
      "loss/hidden": 3.433984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1958073414862156,
      "step": 2620
    },
    {
      "epoch": 0.021916666666666668,
      "grad_norm": 32.25,
      "grad_norm_var": 4.320572916666666,
      "learning_rate": 0.0001,
      "loss": 7.3888,
      "loss/crossentropy": 2.0682441249489782,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1959614047780633,
      "step": 2630
    },
    {
      "epoch": 0.022,
      "grad_norm": 31.25,
      "grad_norm_var": 5.614322916666667,
      "learning_rate": 0.0001,
      "loss": 7.3234,
      "loss/crossentropy": 2.0070124730467795,
      "loss/hidden": 3.4625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18341080471873283,
      "step": 2640
    },
    {
      "epoch": 0.022083333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 2.0249348958333333,
      "learning_rate": 0.0001,
      "loss": 7.2761,
      "loss/crossentropy": 1.967797513306141,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19041062686592342,
      "step": 2650
    },
    {
      "epoch": 0.022166666666666668,
      "grad_norm": 32.0,
      "grad_norm_var": 1.9171223958333334,
      "learning_rate": 0.0001,
      "loss": 7.2174,
      "loss/crossentropy": 1.974215391278267,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17478577867150308,
      "step": 2660
    },
    {
      "epoch": 0.02225,
      "grad_norm": 34.75,
      "grad_norm_var": 2.5660807291666665,
      "learning_rate": 0.0001,
      "loss": 7.2954,
      "loss/crossentropy": 1.7671419337391854,
      "loss/hidden": 3.617578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18350028842687607,
      "step": 2670
    },
    {
      "epoch": 0.022333333333333334,
      "grad_norm": 35.0,
      "grad_norm_var": 3.29765625,
      "learning_rate": 0.0001,
      "loss": 7.352,
      "loss/crossentropy": 1.9952654451131822,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19615703728049994,
      "step": 2680
    },
    {
      "epoch": 0.022416666666666668,
      "grad_norm": 36.25,
      "grad_norm_var": 6.985416666666667,
      "learning_rate": 0.0001,
      "loss": 7.277,
      "loss/crossentropy": 2.0219733342528343,
      "loss/hidden": 3.354296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18562499135732652,
      "step": 2690
    },
    {
      "epoch": 0.0225,
      "grad_norm": 41.0,
      "grad_norm_var": 9.537239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3328,
      "loss/crossentropy": 2.0624974846839903,
      "loss/hidden": 3.502734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.209426661580801,
      "step": 2700
    },
    {
      "epoch": 0.022583333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 9.608268229166667,
      "learning_rate": 0.0001,
      "loss": 7.2719,
      "loss/crossentropy": 1.9747545182704926,
      "loss/hidden": 3.336328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18756564743816853,
      "step": 2710
    },
    {
      "epoch": 0.02266666666666667,
      "grad_norm": 29.0,
      "grad_norm_var": 1.6083333333333334,
      "learning_rate": 0.0001,
      "loss": 7.3397,
      "loss/crossentropy": 2.012811505794525,
      "loss/hidden": 3.483203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20681646689772606,
      "step": 2720
    },
    {
      "epoch": 0.02275,
      "grad_norm": 33.25,
      "grad_norm_var": 2.201041666666667,
      "learning_rate": 0.0001,
      "loss": 7.2686,
      "loss/crossentropy": 1.9331883117556572,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20859357807785273,
      "step": 2730
    },
    {
      "epoch": 0.022833333333333334,
      "grad_norm": 33.5,
      "grad_norm_var": 2.1176432291666667,
      "learning_rate": 0.0001,
      "loss": 7.2957,
      "loss/crossentropy": 1.926119977235794,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20198929514735936,
      "step": 2740
    },
    {
      "epoch": 0.022916666666666665,
      "grad_norm": 30.125,
      "grad_norm_var": 3.3317057291666665,
      "learning_rate": 0.0001,
      "loss": 7.2201,
      "loss/crossentropy": 2.0820239409804344,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20356852617114782,
      "step": 2750
    },
    {
      "epoch": 0.023,
      "grad_norm": 32.5,
      "grad_norm_var": 3.29375,
      "learning_rate": 0.0001,
      "loss": 7.1686,
      "loss/crossentropy": 1.9975199311971665,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18412500955164432,
      "step": 2760
    },
    {
      "epoch": 0.023083333333333334,
      "grad_norm": 34.5,
      "grad_norm_var": 3.133072916666667,
      "learning_rate": 0.0001,
      "loss": 7.2266,
      "loss/crossentropy": 2.1272772580385206,
      "loss/hidden": 3.491015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22087238356471062,
      "step": 2770
    },
    {
      "epoch": 0.023166666666666665,
      "grad_norm": 32.0,
      "grad_norm_var": 2.6455729166666666,
      "learning_rate": 0.0001,
      "loss": 7.2456,
      "loss/crossentropy": 1.8811777606606483,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19566608127206564,
      "step": 2780
    },
    {
      "epoch": 0.02325,
      "grad_norm": 31.625,
      "grad_norm_var": 2.508072916666667,
      "learning_rate": 0.0001,
      "loss": 7.2141,
      "loss/crossentropy": 1.9168774321675301,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17958773095160724,
      "step": 2790
    },
    {
      "epoch": 0.023333333333333334,
      "grad_norm": 31.875,
      "grad_norm_var": 2.005143229166667,
      "learning_rate": 0.0001,
      "loss": 7.2308,
      "loss/crossentropy": 1.9714482218027114,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1880871295928955,
      "step": 2800
    },
    {
      "epoch": 0.023416666666666665,
      "grad_norm": 30.625,
      "grad_norm_var": 4.439583333333333,
      "learning_rate": 0.0001,
      "loss": 7.1113,
      "loss/crossentropy": 1.987887953221798,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19080262612551452,
      "step": 2810
    },
    {
      "epoch": 0.0235,
      "grad_norm": 32.25,
      "grad_norm_var": 5.260416666666667,
      "learning_rate": 0.0001,
      "loss": 7.345,
      "loss/crossentropy": 1.948936927318573,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18754115160554646,
      "step": 2820
    },
    {
      "epoch": 0.023583333333333335,
      "grad_norm": 33.0,
      "grad_norm_var": 2.3934895833333334,
      "learning_rate": 0.0001,
      "loss": 7.2216,
      "loss/crossentropy": 1.8182113721966744,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18257561102509498,
      "step": 2830
    },
    {
      "epoch": 0.023666666666666666,
      "grad_norm": 31.5,
      "grad_norm_var": 1.7332682291666666,
      "learning_rate": 0.0001,
      "loss": 7.3073,
      "loss/crossentropy": 2.11132807135582,
      "loss/hidden": 3.462109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1939231114462018,
      "step": 2840
    },
    {
      "epoch": 0.02375,
      "grad_norm": 32.25,
      "grad_norm_var": 3.3650390625,
      "learning_rate": 0.0001,
      "loss": 7.2723,
      "loss/crossentropy": 1.9388402462005616,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19834725223481656,
      "step": 2850
    },
    {
      "epoch": 0.023833333333333335,
      "grad_norm": 39.75,
      "grad_norm_var": 6.1634765625,
      "learning_rate": 0.0001,
      "loss": 7.3178,
      "loss/crossentropy": 1.9689893007278443,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19809619728475808,
      "step": 2860
    },
    {
      "epoch": 0.023916666666666666,
      "grad_norm": 32.75,
      "grad_norm_var": 7.628059895833333,
      "learning_rate": 0.0001,
      "loss": 7.3359,
      "loss/crossentropy": 2.027157208323479,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1937933772802353,
      "step": 2870
    },
    {
      "epoch": 0.024,
      "grad_norm": 64.0,
      "grad_norm_var": 70.29583333333333,
      "learning_rate": 0.0001,
      "loss": 7.2869,
      "loss/crossentropy": 1.9325517199933528,
      "loss/hidden": 3.315234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18406210988759994,
      "step": 2880
    },
    {
      "epoch": 0.024083333333333335,
      "grad_norm": 31.375,
      "grad_norm_var": 66.80390625,
      "learning_rate": 0.0001,
      "loss": 7.3758,
      "loss/crossentropy": 1.9735887706279756,
      "loss/hidden": 3.495703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1949346676468849,
      "step": 2890
    },
    {
      "epoch": 0.024166666666666666,
      "grad_norm": 33.5,
      "grad_norm_var": 9.49375,
      "learning_rate": 0.0001,
      "loss": 7.3159,
      "loss/crossentropy": 2.036432480812073,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2019389744848013,
      "step": 2900
    },
    {
      "epoch": 0.02425,
      "grad_norm": 44.5,
      "grad_norm_var": 19.467122395833332,
      "learning_rate": 0.0001,
      "loss": 7.1925,
      "loss/crossentropy": 2.054004454612732,
      "loss/hidden": 3.24765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1765751089900732,
      "step": 2910
    },
    {
      "epoch": 0.024333333333333332,
      "grad_norm": 31.0,
      "grad_norm_var": 16.184830729166666,
      "learning_rate": 0.0001,
      "loss": 7.2403,
      "loss/crossentropy": 1.9157620191574096,
      "loss/hidden": 3.461328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18368239123374225,
      "step": 2920
    },
    {
      "epoch": 0.024416666666666666,
      "grad_norm": 29.0,
      "grad_norm_var": 32.4791015625,
      "learning_rate": 0.0001,
      "loss": 7.2583,
      "loss/crossentropy": 2.1513744235038756,
      "loss/hidden": 3.2515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.182504341006279,
      "step": 2930
    },
    {
      "epoch": 0.0245,
      "grad_norm": 31.875,
      "grad_norm_var": 27.1947265625,
      "learning_rate": 0.0001,
      "loss": 7.2578,
      "loss/crossentropy": 1.9272947192192078,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19014938455075026,
      "step": 2940
    },
    {
      "epoch": 0.024583333333333332,
      "grad_norm": 29.875,
      "grad_norm_var": 19.096809895833335,
      "learning_rate": 0.0001,
      "loss": 7.2335,
      "loss/crossentropy": 2.084392386674881,
      "loss/hidden": 3.483984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20528450179845095,
      "step": 2950
    },
    {
      "epoch": 0.024666666666666667,
      "grad_norm": 35.5,
      "grad_norm_var": 20.470833333333335,
      "learning_rate": 0.0001,
      "loss": 7.2383,
      "loss/crossentropy": 1.9980163618922233,
      "loss/hidden": 3.508203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19989981167018414,
      "step": 2960
    },
    {
      "epoch": 0.02475,
      "grad_norm": 40.0,
      "grad_norm_var": 17.977018229166667,
      "learning_rate": 0.0001,
      "loss": 7.3249,
      "loss/crossentropy": 1.9498015969991684,
      "loss/hidden": 3.4703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1932325528934598,
      "step": 2970
    },
    {
      "epoch": 0.024833333333333332,
      "grad_norm": 4764729344.0,
      "grad_norm_var": 1.4189153377602678e+18,
      "learning_rate": 0.0001,
      "loss": 7.3108,
      "loss/crossentropy": 1.9112057030200957,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.198336130566895,
      "step": 2980
    },
    {
      "epoch": 0.024916666666666667,
      "grad_norm": 28.75,
      "grad_norm_var": 1.4189153376361864e+18,
      "learning_rate": 0.0001,
      "loss": 7.2281,
      "loss/crossentropy": 1.911822970956564,
      "loss/hidden": 3.26171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17767341239377857,
      "step": 2990
    },
    {
      "epoch": 0.025,
      "grad_norm": 36.0,
      "grad_norm_var": 7.55,
      "learning_rate": 0.0001,
      "loss": 7.1727,
      "loss/crossentropy": 2.0090511620044706,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18732633236795665,
      "step": 3000
    },
    {
      "epoch": 0.025083333333333332,
      "grad_norm": 28.75,
      "grad_norm_var": 3.983268229166667,
      "learning_rate": 0.0001,
      "loss": 7.2384,
      "loss/crossentropy": 1.8529448568820954,
      "loss/hidden": 3.457421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17940086945891381,
      "step": 3010
    },
    {
      "epoch": 0.025166666666666667,
      "grad_norm": 36.5,
      "grad_norm_var": 8.026822916666667,
      "learning_rate": 0.0001,
      "loss": 7.191,
      "loss/crossentropy": 1.8407732352614403,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18361600376665593,
      "step": 3020
    },
    {
      "epoch": 0.02525,
      "grad_norm": 30.625,
      "grad_norm_var": 6.8150390625,
      "learning_rate": 0.0001,
      "loss": 7.2616,
      "loss/crossentropy": 1.9298807948827743,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18290845435112715,
      "step": 3030
    },
    {
      "epoch": 0.025333333333333333,
      "grad_norm": 30.375,
      "grad_norm_var": 8.673372395833333,
      "learning_rate": 0.0001,
      "loss": 7.3267,
      "loss/crossentropy": 1.8523324035108089,
      "loss/hidden": 3.45234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20208420231938362,
      "step": 3040
    },
    {
      "epoch": 0.025416666666666667,
      "grad_norm": 33.0,
      "grad_norm_var": 6.653125,
      "learning_rate": 0.0001,
      "loss": 7.1421,
      "loss/crossentropy": 1.8332019917666913,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1813520684838295,
      "step": 3050
    },
    {
      "epoch": 0.0255,
      "grad_norm": 29.375,
      "grad_norm_var": 8.258072916666666,
      "learning_rate": 0.0001,
      "loss": 7.3024,
      "loss/crossentropy": 2.017191821336746,
      "loss/hidden": 3.47265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20145763643085957,
      "step": 3060
    },
    {
      "epoch": 0.025583333333333333,
      "grad_norm": 32.5,
      "grad_norm_var": 7.99140625,
      "learning_rate": 0.0001,
      "loss": 7.2919,
      "loss/crossentropy": 2.0802741676568983,
      "loss/hidden": 3.2875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20051774512976409,
      "step": 3070
    },
    {
      "epoch": 0.025666666666666667,
      "grad_norm": 37.25,
      "grad_norm_var": 6.896840476364637e+17,
      "learning_rate": 0.0001,
      "loss": 7.3932,
      "loss/crossentropy": 1.917404356598854,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19901793487370015,
      "step": 3080
    },
    {
      "epoch": 0.02575,
      "grad_norm": 30.5,
      "grad_norm_var": 6.896840478406214e+17,
      "learning_rate": 0.0001,
      "loss": 7.2225,
      "loss/crossentropy": 1.8461115274578332,
      "loss/hidden": 3.4578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18547605257481337,
      "step": 3090
    },
    {
      "epoch": 0.025833333333333333,
      "grad_norm": 30.875,
      "grad_norm_var": 16.2197265625,
      "learning_rate": 0.0001,
      "loss": 7.1225,
      "loss/crossentropy": 1.929587610065937,
      "loss/hidden": 3.309765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1925740350037813,
      "step": 3100
    },
    {
      "epoch": 0.025916666666666668,
      "grad_norm": 37.25,
      "grad_norm_var": 17.442122395833334,
      "learning_rate": 0.0001,
      "loss": 7.1508,
      "loss/crossentropy": 1.9709638714790345,
      "loss/hidden": 3.34140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18460453823208808,
      "step": 3110
    },
    {
      "epoch": 0.026,
      "grad_norm": 29.375,
      "grad_norm_var": 4.7931640625,
      "learning_rate": 0.0001,
      "loss": 7.1906,
      "loss/crossentropy": 2.0181252285838127,
      "loss/hidden": 3.4359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18887849394232034,
      "step": 3120
    },
    {
      "epoch": 0.026083333333333333,
      "grad_norm": 31.375,
      "grad_norm_var": 12.138541666666667,
      "learning_rate": 0.0001,
      "loss": 7.3697,
      "loss/crossentropy": 2.027159887552261,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20400238893926143,
      "step": 3130
    },
    {
      "epoch": 0.026166666666666668,
      "grad_norm": 29.375,
      "grad_norm_var": 10.220768229166667,
      "learning_rate": 0.0001,
      "loss": 7.1989,
      "loss/crossentropy": 2.016098701953888,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17855217922478914,
      "step": 3140
    },
    {
      "epoch": 0.02625,
      "grad_norm": 30.75,
      "grad_norm_var": 9.47890625,
      "learning_rate": 0.0001,
      "loss": 7.3656,
      "loss/crossentropy": 1.9397494405508042,
      "loss/hidden": 3.625390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2223890544846654,
      "step": 3150
    },
    {
      "epoch": 0.026333333333333334,
      "grad_norm": 34.0,
      "grad_norm_var": 3563.3494140625,
      "learning_rate": 0.0001,
      "loss": 7.3418,
      "loss/crossentropy": 1.8907858923077583,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18646195530891418,
      "step": 3160
    },
    {
      "epoch": 0.026416666666666668,
      "grad_norm": 34.0,
      "grad_norm_var": 7.910872395833334,
      "learning_rate": 0.0001,
      "loss": 7.2217,
      "loss/crossentropy": 1.9405520528554916,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.179235702380538,
      "step": 3170
    },
    {
      "epoch": 0.0265,
      "grad_norm": 33.5,
      "grad_norm_var": 5.532291666666667,
      "learning_rate": 0.0001,
      "loss": 7.2206,
      "loss/crossentropy": 2.064801651239395,
      "loss/hidden": 3.234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18341027162969112,
      "step": 3180
    },
    {
      "epoch": 0.026583333333333334,
      "grad_norm": 30.125,
      "grad_norm_var": 7.162434895833333,
      "learning_rate": 0.0001,
      "loss": 7.2847,
      "loss/crossentropy": 2.023468551039696,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19835583344101906,
      "step": 3190
    },
    {
      "epoch": 0.02666666666666667,
      "grad_norm": 31.25,
      "grad_norm_var": 1.8931640625,
      "learning_rate": 0.0001,
      "loss": 7.3574,
      "loss/crossentropy": 1.9950867831707,
      "loss/hidden": 3.32578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19693203810602428,
      "step": 3200
    },
    {
      "epoch": 0.02675,
      "grad_norm": 30.875,
      "grad_norm_var": 2.21015625,
      "learning_rate": 0.0001,
      "loss": 7.1773,
      "loss/crossentropy": 1.859079574048519,
      "loss/hidden": 3.353515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18082537315785885,
      "step": 3210
    },
    {
      "epoch": 0.026833333333333334,
      "grad_norm": 31.0,
      "grad_norm_var": 1.4770833333333333,
      "learning_rate": 0.0001,
      "loss": 7.2378,
      "loss/crossentropy": 2.074732157588005,
      "loss/hidden": 3.315234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19291758965700864,
      "step": 3220
    },
    {
      "epoch": 0.026916666666666665,
      "grad_norm": 31.625,
      "grad_norm_var": 1.3343098958333333,
      "learning_rate": 0.0001,
      "loss": 7.1845,
      "loss/crossentropy": 1.9518444910645485,
      "loss/hidden": 3.4375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21211670562624932,
      "step": 3230
    },
    {
      "epoch": 0.027,
      "grad_norm": 30.25,
      "grad_norm_var": 1.9811848958333333,
      "learning_rate": 0.0001,
      "loss": 7.2353,
      "loss/crossentropy": 1.908726019412279,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1893859247211367,
      "step": 3240
    },
    {
      "epoch": 0.027083333333333334,
      "grad_norm": 30.5,
      "grad_norm_var": 4.86640625,
      "learning_rate": 0.0001,
      "loss": 7.2767,
      "loss/crossentropy": 1.83207483664155,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1776171642355621,
      "step": 3250
    },
    {
      "epoch": 0.027166666666666665,
      "grad_norm": 29.25,
      "grad_norm_var": 15.318489583333333,
      "learning_rate": 0.0001,
      "loss": 7.2133,
      "loss/crossentropy": 2.0540854394435883,
      "loss/hidden": 3.353125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19790230877697468,
      "step": 3260
    },
    {
      "epoch": 0.02725,
      "grad_norm": 29.75,
      "grad_norm_var": 13.468684895833333,
      "learning_rate": 0.0001,
      "loss": 7.2909,
      "loss/crossentropy": 2.1424532026052474,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20493081752210857,
      "step": 3270
    },
    {
      "epoch": 0.027333333333333334,
      "grad_norm": 35.0,
      "grad_norm_var": 8.204166666666667,
      "learning_rate": 0.0001,
      "loss": 7.2174,
      "loss/crossentropy": 1.8585446879267693,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19328337758779526,
      "step": 3280
    },
    {
      "epoch": 0.027416666666666666,
      "grad_norm": 32.0,
      "grad_norm_var": 9.145121829761844e+17,
      "learning_rate": 0.0001,
      "loss": 7.2046,
      "loss/crossentropy": 1.935442891716957,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17625004909932612,
      "step": 3290
    },
    {
      "epoch": 0.0275,
      "grad_norm": 32.0,
      "grad_norm_var": 4.680208333333334,
      "learning_rate": 0.0001,
      "loss": 7.2284,
      "loss/crossentropy": 1.925745696015656,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18133268980309367,
      "step": 3300
    },
    {
      "epoch": 0.027583333333333335,
      "grad_norm": 36.0,
      "grad_norm_var": 33.514322916666664,
      "learning_rate": 0.0001,
      "loss": 7.1611,
      "loss/crossentropy": 1.934358787536621,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1927736137062311,
      "step": 3310
    },
    {
      "epoch": 0.027666666666666666,
      "grad_norm": 31.875,
      "grad_norm_var": 164.93326822916666,
      "learning_rate": 0.0001,
      "loss": 7.2316,
      "loss/crossentropy": 1.8626260861754418,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18418960385024546,
      "step": 3320
    },
    {
      "epoch": 0.02775,
      "grad_norm": 31.875,
      "grad_norm_var": 162.27545572916668,
      "learning_rate": 0.0001,
      "loss": 7.2925,
      "loss/crossentropy": 1.9802043229341506,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1877164736390114,
      "step": 3330
    },
    {
      "epoch": 0.027833333333333335,
      "grad_norm": 30.25,
      "grad_norm_var": 3.8229166666666665,
      "learning_rate": 0.0001,
      "loss": 7.1663,
      "loss/crossentropy": 1.9671151712536812,
      "loss/hidden": 3.40859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19295375384390354,
      "step": 3340
    },
    {
      "epoch": 0.027916666666666666,
      "grad_norm": 29.75,
      "grad_norm_var": 1.1892317587907108e+18,
      "learning_rate": 0.0001,
      "loss": 7.2766,
      "loss/crossentropy": 2.1149969041347503,
      "loss/hidden": 3.4578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20657425597310067,
      "step": 3350
    },
    {
      "epoch": 0.028,
      "grad_norm": 33.0,
      "grad_norm_var": 30.4509765625,
      "learning_rate": 0.0001,
      "loss": 7.2297,
      "loss/crossentropy": 2.0748997569084167,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20927353482693434,
      "step": 3360
    },
    {
      "epoch": 0.02808333333333333,
      "grad_norm": 30.5,
      "grad_norm_var": 4.559309895833334,
      "learning_rate": 0.0001,
      "loss": 7.1992,
      "loss/crossentropy": 1.7197825767099857,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1713511989451945,
      "step": 3370
    },
    {
      "epoch": 0.028166666666666666,
      "grad_norm": 31.125,
      "grad_norm_var": 8.859309895833333,
      "learning_rate": 0.0001,
      "loss": 7.2538,
      "loss/crossentropy": 1.9892003759741783,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1924882870167494,
      "step": 3380
    },
    {
      "epoch": 0.02825,
      "grad_norm": 35.5,
      "grad_norm_var": 20.360416666666666,
      "learning_rate": 0.0001,
      "loss": 7.3701,
      "loss/crossentropy": 2.0029900878667832,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1986021962016821,
      "step": 3390
    },
    {
      "epoch": 0.028333333333333332,
      "grad_norm": 38.0,
      "grad_norm_var": 5.1509765625,
      "learning_rate": 0.0001,
      "loss": 7.3267,
      "loss/crossentropy": 1.9108583837747575,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.18161017037928104,
      "step": 3400
    },
    {
      "epoch": 0.028416666666666666,
      "grad_norm": 30.75,
      "grad_norm_var": 24.479166666666668,
      "learning_rate": 0.0001,
      "loss": 7.2388,
      "loss/crossentropy": 1.936232401430607,
      "loss/hidden": 3.51171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21553935762494802,
      "step": 3410
    },
    {
      "epoch": 0.0285,
      "grad_norm": 37.0,
      "grad_norm_var": 14.5244140625,
      "learning_rate": 0.0001,
      "loss": 7.1867,
      "loss/crossentropy": 1.9755698680877685,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19823556169867515,
      "step": 3420
    },
    {
      "epoch": 0.028583333333333332,
      "grad_norm": 32.25,
      "grad_norm_var": 4.248372395833333,
      "learning_rate": 0.0001,
      "loss": 7.2332,
      "loss/crossentropy": 1.8893498435616494,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18988429009914398,
      "step": 3430
    },
    {
      "epoch": 0.028666666666666667,
      "grad_norm": 38.25,
      "grad_norm_var": 4.315559895833333,
      "learning_rate": 0.0001,
      "loss": 7.3593,
      "loss/crossentropy": 1.9460684165358544,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18759000655263663,
      "step": 3440
    },
    {
      "epoch": 0.02875,
      "grad_norm": 33.0,
      "grad_norm_var": 9.5150390625,
      "learning_rate": 0.0001,
      "loss": 7.3355,
      "loss/crossentropy": 1.843932069838047,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18119751177728177,
      "step": 3450
    },
    {
      "epoch": 0.028833333333333332,
      "grad_norm": 37.0,
      "grad_norm_var": 8.888997395833334,
      "learning_rate": 0.0001,
      "loss": 7.4385,
      "loss/crossentropy": 1.9830153197050095,
      "loss/hidden": 3.5125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1973841156810522,
      "step": 3460
    },
    {
      "epoch": 0.028916666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 4.643489583333333,
      "learning_rate": 0.0001,
      "loss": 7.3091,
      "loss/crossentropy": 1.9742940947413445,
      "loss/hidden": 3.35234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1943078614771366,
      "step": 3470
    },
    {
      "epoch": 0.029,
      "grad_norm": 32.0,
      "grad_norm_var": 3.71875,
      "learning_rate": 0.0001,
      "loss": 7.4103,
      "loss/crossentropy": 2.0165591590106486,
      "loss/hidden": 3.31875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18567599412053823,
      "step": 3480
    },
    {
      "epoch": 0.029083333333333333,
      "grad_norm": 30.75,
      "grad_norm_var": 2.623893229166667,
      "learning_rate": 0.0001,
      "loss": 7.1689,
      "loss/crossentropy": 1.8976772978901864,
      "loss/hidden": 3.273828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17701343428343536,
      "step": 3490
    },
    {
      "epoch": 0.029166666666666667,
      "grad_norm": 32.75,
      "grad_norm_var": 4.560070433854806e+17,
      "learning_rate": 0.0001,
      "loss": 7.4131,
      "loss/crossentropy": 1.8664681144058703,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18947559054940938,
      "step": 3500
    },
    {
      "epoch": 0.02925,
      "grad_norm": 29.375,
      "grad_norm_var": 4.5600704344738157e+17,
      "learning_rate": 0.0001,
      "loss": 7.2189,
      "loss/crossentropy": 1.9561820104718208,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2091151824221015,
      "step": 3510
    },
    {
      "epoch": 0.029333333333333333,
      "grad_norm": 29.625,
      "grad_norm_var": 4.15390625,
      "learning_rate": 0.0001,
      "loss": 7.2372,
      "loss/crossentropy": 1.9695313319563865,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19255149811506272,
      "step": 3520
    },
    {
      "epoch": 0.029416666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 22.8572265625,
      "learning_rate": 0.0001,
      "loss": 7.2377,
      "loss/crossentropy": 2.091267944872379,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2002470575273037,
      "step": 3530
    },
    {
      "epoch": 0.0295,
      "grad_norm": 31.0,
      "grad_norm_var": 12.553059895833334,
      "learning_rate": 0.0001,
      "loss": 7.3079,
      "loss/crossentropy": 2.050358733534813,
      "loss/hidden": 3.278515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18941335808485746,
      "step": 3540
    },
    {
      "epoch": 0.029583333333333333,
      "grad_norm": 31.375,
      "grad_norm_var": 4.202018229166667,
      "learning_rate": 0.0001,
      "loss": 7.2156,
      "loss/crossentropy": 1.9859870471060277,
      "loss/hidden": 3.35,
      "loss/jsd": 0.0,
      "loss/logits": 0.1879461133852601,
      "step": 3550
    },
    {
      "epoch": 0.029666666666666668,
      "grad_norm": 32.25,
      "grad_norm_var": 2.253125,
      "learning_rate": 0.0001,
      "loss": 7.2648,
      "loss/crossentropy": 1.9784325510263443,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18699357379227877,
      "step": 3560
    },
    {
      "epoch": 0.02975,
      "grad_norm": 35.0,
      "grad_norm_var": 19.438541666666666,
      "learning_rate": 0.0001,
      "loss": 7.1386,
      "loss/crossentropy": 1.868053089082241,
      "loss/hidden": 3.45625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19436977664008737,
      "step": 3570
    },
    {
      "epoch": 0.029833333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 21.012239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2663,
      "loss/crossentropy": 2.0065255418419836,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18793275263160467,
      "step": 3580
    },
    {
      "epoch": 0.029916666666666668,
      "grad_norm": 30.75,
      "grad_norm_var": 12.87890625,
      "learning_rate": 0.0001,
      "loss": 7.2809,
      "loss/crossentropy": 1.8890562951564789,
      "loss/hidden": 3.338671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1753269799053669,
      "step": 3590
    },
    {
      "epoch": 0.03,
      "grad_norm": 30.25,
      "grad_norm_var": 10.074934895833334,
      "learning_rate": 0.0001,
      "loss": 7.1971,
      "loss/crossentropy": 1.9849582225084306,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19147213082760572,
      "step": 3600
    },
    {
      "epoch": 0.030083333333333333,
      "grad_norm": 34.25,
      "grad_norm_var": 3.9020182291666665,
      "learning_rate": 0.0001,
      "loss": 7.1947,
      "loss/crossentropy": 1.927957746386528,
      "loss/hidden": 3.256640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1846942937001586,
      "step": 3610
    },
    {
      "epoch": 0.030166666666666668,
      "grad_norm": 38.5,
      "grad_norm_var": 5.361393229166667,
      "learning_rate": 0.0001,
      "loss": 7.169,
      "loss/crossentropy": 1.9098347902297974,
      "loss/hidden": 3.480859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18104900792241096,
      "step": 3620
    },
    {
      "epoch": 0.03025,
      "grad_norm": 33.5,
      "grad_norm_var": 5.13125,
      "learning_rate": 0.0001,
      "loss": 7.144,
      "loss/crossentropy": 1.7323764339089394,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18193134814500808,
      "step": 3630
    },
    {
      "epoch": 0.030333333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 9.798143776342934e+17,
      "learning_rate": 0.0001,
      "loss": 7.289,
      "loss/crossentropy": 1.9735290616750718,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1941153584048152,
      "step": 3640
    },
    {
      "epoch": 0.030416666666666668,
      "grad_norm": 30.625,
      "grad_norm_var": 10.925,
      "learning_rate": 0.0001,
      "loss": 7.2602,
      "loss/crossentropy": 1.8394082367420197,
      "loss/hidden": 3.36875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17473349459469317,
      "step": 3650
    },
    {
      "epoch": 0.0305,
      "grad_norm": 29.375,
      "grad_norm_var": 2.091080729166667,
      "learning_rate": 0.0001,
      "loss": 7.2306,
      "loss/crossentropy": 1.8136562265455722,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17720981892198323,
      "step": 3660
    },
    {
      "epoch": 0.030583333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 2.7747395833333335,
      "learning_rate": 0.0001,
      "loss": 7.2363,
      "loss/crossentropy": 1.947695901989937,
      "loss/hidden": 3.412890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18550261501222848,
      "step": 3670
    },
    {
      "epoch": 0.030666666666666665,
      "grad_norm": 36.5,
      "grad_norm_var": 7.684442640909425e+17,
      "learning_rate": 0.0001,
      "loss": 7.3297,
      "loss/crossentropy": 1.8268452778458595,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1864478962495923,
      "step": 3680
    },
    {
      "epoch": 0.03075,
      "grad_norm": 29.25,
      "grad_norm_var": 7.684442642078237e+17,
      "learning_rate": 0.0001,
      "loss": 7.3815,
      "loss/crossentropy": 2.044998970627785,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19564838595688344,
      "step": 3690
    },
    {
      "epoch": 0.030833333333333334,
      "grad_norm": 34.0,
      "grad_norm_var": 3.412239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2426,
      "loss/crossentropy": 1.9021501779556274,
      "loss/hidden": 3.353125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1848083844408393,
      "step": 3700
    },
    {
      "epoch": 0.030916666666666665,
      "grad_norm": 30.25,
      "grad_norm_var": 3.105208333333333,
      "learning_rate": 0.0001,
      "loss": 7.3239,
      "loss/crossentropy": 2.0004999369382857,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1975437816232443,
      "step": 3710
    },
    {
      "epoch": 0.031,
      "grad_norm": 29.375,
      "grad_norm_var": 4.511458333333334,
      "learning_rate": 0.0001,
      "loss": 7.3091,
      "loss/crossentropy": 1.9500782281160354,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1928380098193884,
      "step": 3720
    },
    {
      "epoch": 0.031083333333333334,
      "grad_norm": 33.5,
      "grad_norm_var": 4.1056640625,
      "learning_rate": 0.0001,
      "loss": 7.2011,
      "loss/crossentropy": 2.0728418275713922,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1800950923934579,
      "step": 3730
    },
    {
      "epoch": 0.031166666666666665,
      "grad_norm": 30.75,
      "grad_norm_var": 2.7327473958333335,
      "learning_rate": 0.0001,
      "loss": 7.2847,
      "loss/crossentropy": 2.007042866945267,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19376245383173227,
      "step": 3740
    },
    {
      "epoch": 0.03125,
      "grad_norm": 39.0,
      "grad_norm_var": 45.46139322916667,
      "learning_rate": 0.0001,
      "loss": 7.0946,
      "loss/crossentropy": 1.9019963197410106,
      "loss/hidden": 3.30703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17656781915575265,
      "step": 3750
    },
    {
      "epoch": 0.03133333333333333,
      "grad_norm": 30.375,
      "grad_norm_var": 3.8014953780616896e+17,
      "learning_rate": 0.0001,
      "loss": 7.2806,
      "loss/crossentropy": 1.9263739496469499,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1908752067014575,
      "step": 3760
    },
    {
      "epoch": 0.03141666666666667,
      "grad_norm": 28.25,
      "grad_norm_var": 3.8014953838162746e+17,
      "learning_rate": 0.0001,
      "loss": 7.2388,
      "loss/crossentropy": 1.965820837020874,
      "loss/hidden": 3.49609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19951398996636271,
      "step": 3770
    },
    {
      "epoch": 0.0315,
      "grad_norm": 29.875,
      "grad_norm_var": 70.99576822916667,
      "learning_rate": 0.0001,
      "loss": 7.346,
      "loss/crossentropy": 1.9226266965270042,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18729710709303618,
      "step": 3780
    },
    {
      "epoch": 0.03158333333333333,
      "grad_norm": 33.0,
      "grad_norm_var": 66.08125,
      "learning_rate": 0.0001,
      "loss": 7.2438,
      "loss/crossentropy": 1.8775743916630745,
      "loss/hidden": 3.50390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19226148668676615,
      "step": 3790
    },
    {
      "epoch": 0.03166666666666667,
      "grad_norm": 31.125,
      "grad_norm_var": 3.207291666666667,
      "learning_rate": 0.0001,
      "loss": 7.2464,
      "loss/crossentropy": 1.960534116625786,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20045664943754674,
      "step": 3800
    },
    {
      "epoch": 0.03175,
      "grad_norm": 30.25,
      "grad_norm_var": 1.6436848958333334,
      "learning_rate": 0.0001,
      "loss": 7.1574,
      "loss/crossentropy": 1.7869322314858436,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1767852820456028,
      "step": 3810
    },
    {
      "epoch": 0.03183333333333333,
      "grad_norm": 34.25,
      "grad_norm_var": 1.9650390625,
      "learning_rate": 0.0001,
      "loss": 7.2089,
      "loss/crossentropy": 2.004594850540161,
      "loss/hidden": 3.271875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17339777592569589,
      "step": 3820
    },
    {
      "epoch": 0.03191666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 8.370572916666667,
      "learning_rate": 0.0001,
      "loss": 7.1548,
      "loss/crossentropy": 1.9374451048672199,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1855450887233019,
      "step": 3830
    },
    {
      "epoch": 0.032,
      "grad_norm": 31.0,
      "grad_norm_var": 3.9514973958333335,
      "learning_rate": 0.0001,
      "loss": 7.2219,
      "loss/crossentropy": 1.9749352604150772,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18502496387809514,
      "step": 3840
    },
    {
      "epoch": 0.03208333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 1.7186848958333334,
      "learning_rate": 0.0001,
      "loss": 7.1275,
      "loss/crossentropy": 2.0351619154214857,
      "loss/hidden": 3.33515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17815996166318654,
      "step": 3850
    },
    {
      "epoch": 0.03216666666666667,
      "grad_norm": 33.25,
      "grad_norm_var": 10.888997395833334,
      "learning_rate": 0.0001,
      "loss": 7.2478,
      "loss/crossentropy": 1.976029747724533,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18774439115077257,
      "step": 3860
    },
    {
      "epoch": 0.03225,
      "grad_norm": 30.875,
      "grad_norm_var": 6.245572916666666,
      "learning_rate": 0.0001,
      "loss": 7.2188,
      "loss/crossentropy": 1.9812984496355057,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19507236313074827,
      "step": 3870
    },
    {
      "epoch": 0.03233333333333333,
      "grad_norm": 31.875,
      "grad_norm_var": 152.28229166666668,
      "learning_rate": 0.0001,
      "loss": 7.2458,
      "loss/crossentropy": 1.984493115544319,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19963345900177956,
      "step": 3880
    },
    {
      "epoch": 0.03241666666666667,
      "grad_norm": 31.625,
      "grad_norm_var": 7.1625,
      "learning_rate": 0.0001,
      "loss": 7.389,
      "loss/crossentropy": 1.9200308933854102,
      "loss/hidden": 3.479296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1876375088468194,
      "step": 3890
    },
    {
      "epoch": 0.0325,
      "grad_norm": 28.625,
      "grad_norm_var": 5.81015625,
      "learning_rate": 0.0001,
      "loss": 7.2937,
      "loss/crossentropy": 1.983275644481182,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18901068139821292,
      "step": 3900
    },
    {
      "epoch": 0.03258333333333333,
      "grad_norm": 42.75,
      "grad_norm_var": 13.860416666666667,
      "learning_rate": 0.0001,
      "loss": 7.2993,
      "loss/crossentropy": 2.002154844999313,
      "loss/hidden": 3.51796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1924261325970292,
      "step": 3910
    },
    {
      "epoch": 0.03266666666666666,
      "grad_norm": 31.125,
      "grad_norm_var": 25.407747395833333,
      "learning_rate": 0.0001,
      "loss": 7.2245,
      "loss/crossentropy": 1.9173000499606132,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19295201431959869,
      "step": 3920
    },
    {
      "epoch": 0.03275,
      "grad_norm": 29.625,
      "grad_norm_var": 18.592643229166665,
      "learning_rate": 0.0001,
      "loss": 7.2637,
      "loss/crossentropy": 2.0171883307397365,
      "loss/hidden": 3.290234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17894033640623092,
      "step": 3930
    },
    {
      "epoch": 0.03283333333333333,
      "grad_norm": 31.875,
      "grad_norm_var": 1.7561848958333333,
      "learning_rate": 0.0001,
      "loss": 7.2559,
      "loss/crossentropy": 1.973395998775959,
      "loss/hidden": 3.53671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20406727930530905,
      "step": 3940
    },
    {
      "epoch": 0.032916666666666664,
      "grad_norm": 32.25,
      "grad_norm_var": 1.5947916666666666,
      "learning_rate": 0.0001,
      "loss": 7.2626,
      "loss/crossentropy": 1.95164203196764,
      "loss/hidden": 3.408203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1913003684952855,
      "step": 3950
    },
    {
      "epoch": 0.033,
      "grad_norm": 29.875,
      "grad_norm_var": 2.794205729166667,
      "learning_rate": 0.0001,
      "loss": 7.2546,
      "loss/crossentropy": 2.006377327442169,
      "loss/hidden": 3.27578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18781272880733013,
      "step": 3960
    },
    {
      "epoch": 0.03308333333333333,
      "grad_norm": 29.875,
      "grad_norm_var": 4.484830729166666,
      "learning_rate": 0.0001,
      "loss": 7.1955,
      "loss/crossentropy": 1.8874030619859696,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17777375876903534,
      "step": 3970
    },
    {
      "epoch": 0.033166666666666664,
      "grad_norm": 30.75,
      "grad_norm_var": 4.337239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3097,
      "loss/crossentropy": 1.9324250996112824,
      "loss/hidden": 3.219140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1771338377147913,
      "step": 3980
    },
    {
      "epoch": 0.03325,
      "grad_norm": 31.75,
      "grad_norm_var": 2.3025390625,
      "learning_rate": 0.0001,
      "loss": 7.1529,
      "loss/crossentropy": 1.8945629693567754,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17384705552831292,
      "step": 3990
    },
    {
      "epoch": 0.03333333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 12.415625,
      "learning_rate": 0.0001,
      "loss": 7.2273,
      "loss/crossentropy": 1.954460334777832,
      "loss/hidden": 3.33046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20084381606429816,
      "step": 4000
    },
    {
      "epoch": 0.033416666666666664,
      "grad_norm": 34.25,
      "grad_norm_var": 8.487434895833333,
      "learning_rate": 0.0001,
      "loss": 7.262,
      "loss/crossentropy": 1.8653246648609638,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17524506486952304,
      "step": 4010
    },
    {
      "epoch": 0.0335,
      "grad_norm": 32.25,
      "grad_norm_var": 3.2333333333333334,
      "learning_rate": 0.0001,
      "loss": 7.2548,
      "loss/crossentropy": 1.9703700937330724,
      "loss/hidden": 3.245703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18107774439267815,
      "step": 4020
    },
    {
      "epoch": 0.03358333333333333,
      "grad_norm": 29.375,
      "grad_norm_var": 5.416666666666667,
      "learning_rate": 0.0001,
      "loss": 7.2275,
      "loss/crossentropy": 2.0328017130494116,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19874867778271438,
      "step": 4030
    },
    {
      "epoch": 0.033666666666666664,
      "grad_norm": 34.25,
      "grad_norm_var": 3.8785807291666665,
      "learning_rate": 0.0001,
      "loss": 7.1966,
      "loss/crossentropy": 2.034155286103487,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1778541506268084,
      "step": 4040
    },
    {
      "epoch": 0.03375,
      "grad_norm": 32.75,
      "grad_norm_var": 9.233333333333333,
      "learning_rate": 0.0001,
      "loss": 7.391,
      "loss/crossentropy": 1.956402349472046,
      "loss/hidden": 3.274609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18872720897197723,
      "step": 4050
    },
    {
      "epoch": 0.03383333333333333,
      "grad_norm": 31.125,
      "grad_norm_var": 5.89144705786341e+17,
      "learning_rate": 0.0001,
      "loss": 7.1737,
      "loss/crossentropy": 1.983364275097847,
      "loss/hidden": 3.54609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20503977742046117,
      "step": 4060
    },
    {
      "epoch": 0.033916666666666664,
      "grad_norm": 47.5,
      "grad_norm_var": 23.84140625,
      "learning_rate": 0.0001,
      "loss": 7.2168,
      "loss/crossentropy": 1.9249817907810212,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1917848229408264,
      "step": 4070
    },
    {
      "epoch": 0.034,
      "grad_norm": 32.25,
      "grad_norm_var": 20.892708333333335,
      "learning_rate": 0.0001,
      "loss": 7.3556,
      "loss/crossentropy": 1.9946053527295589,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20452868696302176,
      "step": 4080
    },
    {
      "epoch": 0.034083333333333334,
      "grad_norm": 32.0,
      "grad_norm_var": 6.035872395833334,
      "learning_rate": 0.0001,
      "loss": 7.2173,
      "loss/crossentropy": 1.9789577759802341,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17965719159692525,
      "step": 4090
    },
    {
      "epoch": 0.034166666666666665,
      "grad_norm": 34.0,
      "grad_norm_var": 3.5254557291666666,
      "learning_rate": 0.0001,
      "loss": 7.1648,
      "loss/crossentropy": 1.95723911896348,
      "loss/hidden": 3.2890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18108865497633814,
      "step": 4100
    },
    {
      "epoch": 0.03425,
      "grad_norm": 30.0,
      "grad_norm_var": 2.8067057291666666,
      "learning_rate": 0.0001,
      "loss": 7.1109,
      "loss/crossentropy": 1.962564404308796,
      "loss/hidden": 3.441796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19903685674071311,
      "step": 4110
    },
    {
      "epoch": 0.034333333333333334,
      "grad_norm": 30.875,
      "grad_norm_var": 2.3650390625,
      "learning_rate": 0.0001,
      "loss": 7.2889,
      "loss/crossentropy": 1.9317152202129364,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894014548510313,
      "step": 4120
    },
    {
      "epoch": 0.034416666666666665,
      "grad_norm": 31.5,
      "grad_norm_var": 1.6869140625,
      "learning_rate": 0.0001,
      "loss": 7.2726,
      "loss/crossentropy": 1.888142066448927,
      "loss/hidden": 3.511328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19417026322335004,
      "step": 4130
    },
    {
      "epoch": 0.0345,
      "grad_norm": 29.0,
      "grad_norm_var": 4.608333333333333,
      "learning_rate": 0.0001,
      "loss": 7.2027,
      "loss/crossentropy": 1.9418023459613323,
      "loss/hidden": 3.26484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18381806444376708,
      "step": 4140
    },
    {
      "epoch": 0.034583333333333334,
      "grad_norm": 32.75,
      "grad_norm_var": 5.323958333333334,
      "learning_rate": 0.0001,
      "loss": 7.2148,
      "loss/crossentropy": 1.9318063765764237,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20140955168753863,
      "step": 4150
    },
    {
      "epoch": 0.034666666666666665,
      "grad_norm": 175.0,
      "grad_norm_var": 1278.2921223958333,
      "learning_rate": 0.0001,
      "loss": 7.1926,
      "loss/crossentropy": 1.8987142145633698,
      "loss/hidden": 3.428515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20260288771241902,
      "step": 4160
    },
    {
      "epoch": 0.03475,
      "grad_norm": 31.25,
      "grad_norm_var": 1286.3910807291666,
      "learning_rate": 0.0001,
      "loss": 7.1593,
      "loss/crossentropy": 2.034174790978432,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18820980936288834,
      "step": 4170
    },
    {
      "epoch": 0.034833333333333334,
      "grad_norm": 30.625,
      "grad_norm_var": 15.941666666666666,
      "learning_rate": 0.0001,
      "loss": 7.2882,
      "loss/crossentropy": 1.9306144535541534,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19824343249201776,
      "step": 4180
    },
    {
      "epoch": 0.034916666666666665,
      "grad_norm": 31.125,
      "grad_norm_var": 85.56770833333333,
      "learning_rate": 0.0001,
      "loss": 7.1008,
      "loss/crossentropy": 1.951157060265541,
      "loss/hidden": 3.349609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17776616513729096,
      "step": 4190
    },
    {
      "epoch": 0.035,
      "grad_norm": 38.25,
      "grad_norm_var": 11.1625,
      "learning_rate": 0.0001,
      "loss": 7.2101,
      "loss/crossentropy": 1.9059597700834274,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1789044144563377,
      "step": 4200
    },
    {
      "epoch": 0.035083333333333334,
      "grad_norm": 33.5,
      "grad_norm_var": 5.891447057895392e+17,
      "learning_rate": 0.0001,
      "loss": 7.4328,
      "loss/crossentropy": 1.9277472093701362,
      "loss/hidden": 3.615234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19346159268170596,
      "step": 4210
    },
    {
      "epoch": 0.035166666666666666,
      "grad_norm": 30.0,
      "grad_norm_var": 6.06875,
      "learning_rate": 0.0001,
      "loss": 7.137,
      "loss/crossentropy": 1.976280900835991,
      "loss/hidden": 3.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18772026468068362,
      "step": 4220
    },
    {
      "epoch": 0.03525,
      "grad_norm": 31.875,
      "grad_norm_var": 5.734830729166666,
      "learning_rate": 0.0001,
      "loss": 7.265,
      "loss/crossentropy": 1.9235099971294403,
      "loss/hidden": 3.493359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18395463675260543,
      "step": 4230
    },
    {
      "epoch": 0.035333333333333335,
      "grad_norm": 35.25,
      "grad_norm_var": 5.07890625,
      "learning_rate": 0.0001,
      "loss": 7.1643,
      "loss/crossentropy": 1.8247085146605968,
      "loss/hidden": 3.302734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17993250191211702,
      "step": 4240
    },
    {
      "epoch": 0.035416666666666666,
      "grad_norm": 31.875,
      "grad_norm_var": 233.09557291666667,
      "learning_rate": 0.0001,
      "loss": 7.2326,
      "loss/crossentropy": 1.900027759373188,
      "loss/hidden": 3.49140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1891545556485653,
      "step": 4250
    },
    {
      "epoch": 0.0355,
      "grad_norm": 33.0,
      "grad_norm_var": 5.8525390625,
      "learning_rate": 0.0001,
      "loss": 7.2668,
      "loss/crossentropy": 2.0083697855472566,
      "loss/hidden": 3.343359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19048811830580234,
      "step": 4260
    },
    {
      "epoch": 0.035583333333333335,
      "grad_norm": 30.125,
      "grad_norm_var": 7.295247395833333,
      "learning_rate": 0.0001,
      "loss": 7.2289,
      "loss/crossentropy": 1.8813670575618744,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20953154526650905,
      "step": 4270
    },
    {
      "epoch": 0.035666666666666666,
      "grad_norm": 4009754624.0,
      "grad_norm_var": 1.0048832424736108e+18,
      "learning_rate": 0.0001,
      "loss": 7.2215,
      "loss/crossentropy": 2.036851942539215,
      "loss/hidden": 3.364453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18767790962010622,
      "step": 4280
    },
    {
      "epoch": 0.03575,
      "grad_norm": 35.5,
      "grad_norm_var": 1.0048832419306232e+18,
      "learning_rate": 0.0001,
      "loss": 7.1731,
      "loss/crossentropy": 1.8813318252563476,
      "loss/hidden": 3.33984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17605223394930364,
      "step": 4290
    },
    {
      "epoch": 0.035833333333333335,
      "grad_norm": 32.0,
      "grad_norm_var": 8.367708333333333,
      "learning_rate": 0.0001,
      "loss": 7.2143,
      "loss/crossentropy": 1.7793952539563178,
      "loss/hidden": 3.450390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17674990901723503,
      "step": 4300
    },
    {
      "epoch": 0.035916666666666666,
      "grad_norm": 31.0,
      "grad_norm_var": 5.8125,
      "learning_rate": 0.0001,
      "loss": 7.1788,
      "loss/crossentropy": 2.114986628293991,
      "loss/hidden": 3.28203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17624464612454177,
      "step": 4310
    },
    {
      "epoch": 0.036,
      "grad_norm": 29.375,
      "grad_norm_var": 9.32265625,
      "learning_rate": 0.0001,
      "loss": 7.1168,
      "loss/crossentropy": 1.965844801068306,
      "loss/hidden": 3.227734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18552668560296298,
      "step": 4320
    },
    {
      "epoch": 0.036083333333333335,
      "grad_norm": 33.5,
      "grad_norm_var": 12.855143229166666,
      "learning_rate": 0.0001,
      "loss": 7.2536,
      "loss/crossentropy": 2.029114265739918,
      "loss/hidden": 3.353125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20089028123766184,
      "step": 4330
    },
    {
      "epoch": 0.036166666666666666,
      "grad_norm": 29.125,
      "grad_norm_var": 2.6889973958333333,
      "learning_rate": 0.0001,
      "loss": 7.1575,
      "loss/crossentropy": 1.9205705314874648,
      "loss/hidden": 3.329296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19222395457327365,
      "step": 4340
    },
    {
      "epoch": 0.03625,
      "grad_norm": 29.875,
      "grad_norm_var": 18.380143229166666,
      "learning_rate": 0.0001,
      "loss": 7.2364,
      "loss/crossentropy": 2.080160215497017,
      "loss/hidden": 3.32265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2045929528772831,
      "step": 4350
    },
    {
      "epoch": 0.036333333333333336,
      "grad_norm": 32.25,
      "grad_norm_var": 1.9692057291666667,
      "learning_rate": 0.0001,
      "loss": 7.1983,
      "loss/crossentropy": 1.9375086218118667,
      "loss/hidden": 3.4015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18691876921802758,
      "step": 4360
    },
    {
      "epoch": 0.03641666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 7.981398734615348e+17,
      "learning_rate": 0.0001,
      "loss": 7.3271,
      "loss/crossentropy": 2.005251409113407,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18281115759164096,
      "step": 4370
    },
    {
      "epoch": 0.0365,
      "grad_norm": 32.5,
      "grad_norm_var": 52.04420572916667,
      "learning_rate": 0.0001,
      "loss": 7.2931,
      "loss/crossentropy": 1.9374827235937118,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19405684284865857,
      "step": 4380
    },
    {
      "epoch": 0.036583333333333336,
      "grad_norm": 34.75,
      "grad_norm_var": 20.840559895833334,
      "learning_rate": 0.0001,
      "loss": 7.2968,
      "loss/crossentropy": 1.9063488006591798,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20247304160147905,
      "step": 4390
    },
    {
      "epoch": 0.03666666666666667,
      "grad_norm": 31.125,
      "grad_norm_var": 23.69140625,
      "learning_rate": 0.0001,
      "loss": 7.2249,
      "loss/crossentropy": 1.9918562591075897,
      "loss/hidden": 3.3015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18467331416904925,
      "step": 4400
    },
    {
      "epoch": 0.03675,
      "grad_norm": 30.625,
      "grad_norm_var": 2.3025390625,
      "learning_rate": 0.0001,
      "loss": 7.117,
      "loss/crossentropy": 1.9270878240466118,
      "loss/hidden": 3.48828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20000088401138783,
      "step": 4410
    },
    {
      "epoch": 0.036833333333333336,
      "grad_norm": 31.375,
      "grad_norm_var": 3.7552083333333335,
      "learning_rate": 0.0001,
      "loss": 7.3101,
      "loss/crossentropy": 2.071250331401825,
      "loss/hidden": 3.58828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19236841816455125,
      "step": 4420
    },
    {
      "epoch": 0.03691666666666667,
      "grad_norm": 62.75,
      "grad_norm_var": 65.09837239583334,
      "learning_rate": 0.0001,
      "loss": 7.1969,
      "loss/crossentropy": 1.966628861427307,
      "loss/hidden": 3.471484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20192189700901508,
      "step": 4430
    },
    {
      "epoch": 0.037,
      "grad_norm": 27.75,
      "grad_norm_var": 66.3728515625,
      "learning_rate": 0.0001,
      "loss": 7.1672,
      "loss/crossentropy": 1.8696994706988335,
      "loss/hidden": 3.386328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1916260415688157,
      "step": 4440
    },
    {
      "epoch": 0.037083333333333336,
      "grad_norm": 33.0,
      "grad_norm_var": 3.7775390625,
      "learning_rate": 0.0001,
      "loss": 7.278,
      "loss/crossentropy": 1.9153100848197937,
      "loss/hidden": 3.465625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20059756319969893,
      "step": 4450
    },
    {
      "epoch": 0.03716666666666667,
      "grad_norm": 37.25,
      "grad_norm_var": 4.753125,
      "learning_rate": 0.0001,
      "loss": 7.2516,
      "loss/crossentropy": 2.06074358522892,
      "loss/hidden": 3.4609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1892743492498994,
      "step": 4460
    },
    {
      "epoch": 0.03725,
      "grad_norm": 29.375,
      "grad_norm_var": 5.079166666666667,
      "learning_rate": 0.0001,
      "loss": 7.1622,
      "loss/crossentropy": 1.954483150690794,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1837238807231188,
      "step": 4470
    },
    {
      "epoch": 0.037333333333333336,
      "grad_norm": 37.0,
      "grad_norm_var": 5.5337890625,
      "learning_rate": 0.0001,
      "loss": 7.241,
      "loss/crossentropy": 1.9478939175605774,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20669157840311528,
      "step": 4480
    },
    {
      "epoch": 0.03741666666666667,
      "grad_norm": 34.75,
      "grad_norm_var": 4.60390625,
      "learning_rate": 0.0001,
      "loss": 7.2972,
      "loss/crossentropy": 1.890691903233528,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19282174929976464,
      "step": 4490
    },
    {
      "epoch": 0.0375,
      "grad_norm": 32.75,
      "grad_norm_var": 3.24140625,
      "learning_rate": 0.0001,
      "loss": 7.2593,
      "loss/crossentropy": 1.8015254758298398,
      "loss/hidden": 3.396484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17511156313121318,
      "step": 4500
    },
    {
      "epoch": 0.03758333333333334,
      "grad_norm": 31.75,
      "grad_norm_var": 3.465625,
      "learning_rate": 0.0001,
      "loss": 7.188,
      "loss/crossentropy": 1.984817424416542,
      "loss/hidden": 3.3078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17105563301593066,
      "step": 4510
    },
    {
      "epoch": 0.03766666666666667,
      "grad_norm": 40.75,
      "grad_norm_var": 21.156705729166667,
      "learning_rate": 0.0001,
      "loss": 7.2147,
      "loss/crossentropy": 1.9874697402119637,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19757193494588138,
      "step": 4520
    },
    {
      "epoch": 0.03775,
      "grad_norm": 37.5,
      "grad_norm_var": 342.69166666666666,
      "learning_rate": 0.0001,
      "loss": 7.2505,
      "loss/crossentropy": 1.845185561478138,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.192365712672472,
      "step": 4530
    },
    {
      "epoch": 0.03783333333333333,
      "grad_norm": 31.5,
      "grad_norm_var": 330.7806640625,
      "learning_rate": 0.0001,
      "loss": 7.203,
      "loss/crossentropy": 2.0427382573485375,
      "loss/hidden": 3.343359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17846857775002717,
      "step": 4540
    },
    {
      "epoch": 0.03791666666666667,
      "grad_norm": 31.25,
      "grad_norm_var": 5.216080729166666,
      "learning_rate": 0.0001,
      "loss": 7.1229,
      "loss/crossentropy": 1.8890719719231128,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1911027053371072,
      "step": 4550
    },
    {
      "epoch": 0.038,
      "grad_norm": 29.875,
      "grad_norm_var": 3.1801432291666667,
      "learning_rate": 0.0001,
      "loss": 7.2172,
      "loss/crossentropy": 2.0080476641654967,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.192045708745718,
      "step": 4560
    },
    {
      "epoch": 0.03808333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 5.338997395833333,
      "learning_rate": 0.0001,
      "loss": 7.1599,
      "loss/crossentropy": 1.9886827737092971,
      "loss/hidden": 3.30546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17904456499964,
      "step": 4570
    },
    {
      "epoch": 0.03816666666666667,
      "grad_norm": 33.75,
      "grad_norm_var": 4.708333333333333,
      "learning_rate": 0.0001,
      "loss": 7.289,
      "loss/crossentropy": 2.07514126971364,
      "loss/hidden": 3.37109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21552205746993422,
      "step": 4580
    },
    {
      "epoch": 0.03825,
      "grad_norm": 34.5,
      "grad_norm_var": 3.268684895833333,
      "learning_rate": 0.0001,
      "loss": 7.3541,
      "loss/crossentropy": 1.8997907042503357,
      "loss/hidden": 3.496484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20105953328311443,
      "step": 4590
    },
    {
      "epoch": 0.03833333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 3.0416015625,
      "learning_rate": 0.0001,
      "loss": 7.3417,
      "loss/crossentropy": 1.9448037043213844,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19288307912647723,
      "step": 4600
    },
    {
      "epoch": 0.03841666666666667,
      "grad_norm": 30.125,
      "grad_norm_var": 1.5434895833333333,
      "learning_rate": 0.0001,
      "loss": 7.1478,
      "loss/crossentropy": 1.9898255258798598,
      "loss/hidden": 3.267578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18608010280877352,
      "step": 4610
    },
    {
      "epoch": 0.0385,
      "grad_norm": 32.75,
      "grad_norm_var": 11.652018229166666,
      "learning_rate": 0.0001,
      "loss": 7.3242,
      "loss/crossentropy": 2.009296101331711,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1843553153797984,
      "step": 4620
    },
    {
      "epoch": 0.03858333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 13.0119140625,
      "learning_rate": 0.0001,
      "loss": 7.1898,
      "loss/crossentropy": 1.9398988649249076,
      "loss/hidden": 3.399609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18743706177920103,
      "step": 4630
    },
    {
      "epoch": 0.03866666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 2.059830729166667,
      "learning_rate": 0.0001,
      "loss": 7.1591,
      "loss/crossentropy": 1.853212970495224,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17872758507728576,
      "step": 4640
    },
    {
      "epoch": 0.03875,
      "grad_norm": 30.125,
      "grad_norm_var": 2.08515625,
      "learning_rate": 0.0001,
      "loss": 7.1896,
      "loss/crossentropy": 2.0782950162887572,
      "loss/hidden": 3.225390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17811987847089766,
      "step": 4650
    },
    {
      "epoch": 0.03883333333333333,
      "grad_norm": 37.5,
      "grad_norm_var": 2.575671875150546e+17,
      "learning_rate": 0.0001,
      "loss": 7.3045,
      "loss/crossentropy": 2.0085189670324324,
      "loss/hidden": 3.325390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20201543867588043,
      "step": 4660
    },
    {
      "epoch": 0.03891666666666667,
      "grad_norm": 30.25,
      "grad_norm_var": 2.5756718751928384e+17,
      "learning_rate": 0.0001,
      "loss": 7.0863,
      "loss/crossentropy": 1.9253064393997192,
      "loss/hidden": 3.444140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21200950276106595,
      "step": 4670
    },
    {
      "epoch": 0.039,
      "grad_norm": 32.0,
      "grad_norm_var": 2.4166015625,
      "learning_rate": 0.0001,
      "loss": 7.2533,
      "loss/crossentropy": 1.9820420145988464,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19638330191373826,
      "step": 4680
    },
    {
      "epoch": 0.03908333333333333,
      "grad_norm": 30.125,
      "grad_norm_var": 2.9374348958333334,
      "learning_rate": 0.0001,
      "loss": 7.1681,
      "loss/crossentropy": 1.8506903916597366,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1865494478493929,
      "step": 4690
    },
    {
      "epoch": 0.03916666666666667,
      "grad_norm": 31.5,
      "grad_norm_var": 2.941666666666667,
      "learning_rate": 0.0001,
      "loss": 7.2402,
      "loss/crossentropy": 2.0131797283887862,
      "loss/hidden": 3.546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20369560234248638,
      "step": 4700
    },
    {
      "epoch": 0.03925,
      "grad_norm": 31.625,
      "grad_norm_var": 3.776497395833333,
      "learning_rate": 0.0001,
      "loss": 7.2278,
      "loss/crossentropy": 1.7893060639500618,
      "loss/hidden": 3.353515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18423357047140598,
      "step": 4710
    },
    {
      "epoch": 0.03933333333333333,
      "grad_norm": 34.5,
      "grad_norm_var": 2.2072265625,
      "learning_rate": 0.0001,
      "loss": 7.2598,
      "loss/crossentropy": 1.9981599599123,
      "loss/hidden": 3.358203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18187536485493183,
      "step": 4720
    },
    {
      "epoch": 0.03941666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 4.0259765625,
      "learning_rate": 0.0001,
      "loss": 7.2328,
      "loss/crossentropy": 1.969702386856079,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20060717556625604,
      "step": 4730
    },
    {
      "epoch": 0.0395,
      "grad_norm": 31.0,
      "grad_norm_var": 3.226822916666667,
      "learning_rate": 0.0001,
      "loss": 7.1703,
      "loss/crossentropy": 1.8951549768447875,
      "loss/hidden": 3.269921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19447669275105,
      "step": 4740
    },
    {
      "epoch": 0.03958333333333333,
      "grad_norm": 33.0,
      "grad_norm_var": 2.737434895833333,
      "learning_rate": 0.0001,
      "loss": 7.2429,
      "loss/crossentropy": 2.080011251568794,
      "loss/hidden": 3.290625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19000240378081798,
      "step": 4750
    },
    {
      "epoch": 0.03966666666666667,
      "grad_norm": 28.125,
      "grad_norm_var": 3.9301432291666667,
      "learning_rate": 0.0001,
      "loss": 7.1889,
      "loss/crossentropy": 1.9444591283798218,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1841985611245036,
      "step": 4760
    },
    {
      "epoch": 0.03975,
      "grad_norm": 29.125,
      "grad_norm_var": 6.077083333333333,
      "learning_rate": 0.0001,
      "loss": 7.2539,
      "loss/crossentropy": 1.9875277951359749,
      "loss/hidden": 3.476171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19982990454882382,
      "step": 4770
    },
    {
      "epoch": 0.03983333333333333,
      "grad_norm": 28.875,
      "grad_norm_var": 3.3684895833333335,
      "learning_rate": 0.0001,
      "loss": 7.2338,
      "loss/crossentropy": 2.018096478283405,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1817854383960366,
      "step": 4780
    },
    {
      "epoch": 0.03991666666666667,
      "grad_norm": 35.0,
      "grad_norm_var": 5.211458333333334,
      "learning_rate": 0.0001,
      "loss": 7.2624,
      "loss/crossentropy": 1.9615140050649642,
      "loss/hidden": 3.316796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18594696801155805,
      "step": 4790
    },
    {
      "epoch": 0.04,
      "grad_norm": 33.0,
      "grad_norm_var": 7.9509765625,
      "learning_rate": 0.0001,
      "loss": 7.3972,
      "loss/crossentropy": 1.9175285905599595,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19690106008201838,
      "step": 4800
    },
    {
      "epoch": 0.04008333333333333,
      "grad_norm": 28.125,
      "grad_norm_var": 61.534375,
      "learning_rate": 0.0001,
      "loss": 7.2732,
      "loss/crossentropy": 1.8362210482358932,
      "loss/hidden": 3.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1917069649323821,
      "step": 4810
    },
    {
      "epoch": 0.04016666666666667,
      "grad_norm": 31.125,
      "grad_norm_var": 57.2712890625,
      "learning_rate": 0.0001,
      "loss": 7.3767,
      "loss/crossentropy": 2.133534160256386,
      "loss/hidden": 3.51015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21782590821385384,
      "step": 4820
    },
    {
      "epoch": 0.04025,
      "grad_norm": 30.125,
      "grad_norm_var": 3.082291666666667,
      "learning_rate": 0.0001,
      "loss": 7.2232,
      "loss/crossentropy": 1.951946584880352,
      "loss/hidden": 3.4390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19556211698800324,
      "step": 4830
    },
    {
      "epoch": 0.04033333333333333,
      "grad_norm": 31.5,
      "grad_norm_var": 7.76015625,
      "learning_rate": 0.0001,
      "loss": 7.2368,
      "loss/crossentropy": 1.9654059708118439,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19498455952852964,
      "step": 4840
    },
    {
      "epoch": 0.04041666666666666,
      "grad_norm": 35.25,
      "grad_norm_var": 6.287239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2589,
      "loss/crossentropy": 1.9640298388898372,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1852282472886145,
      "step": 4850
    },
    {
      "epoch": 0.0405,
      "grad_norm": 30.5,
      "grad_norm_var": 2.3489583333333335,
      "learning_rate": 0.0001,
      "loss": 7.2713,
      "loss/crossentropy": 1.896032267808914,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20271046720445157,
      "step": 4860
    },
    {
      "epoch": 0.04058333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 54.316080729166664,
      "learning_rate": 0.0001,
      "loss": 7.1648,
      "loss/crossentropy": 1.9707864001393318,
      "loss/hidden": 3.50234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.192245776951313,
      "step": 4870
    },
    {
      "epoch": 0.04066666666666666,
      "grad_norm": 32.0,
      "grad_norm_var": 2.3556640625,
      "learning_rate": 0.0001,
      "loss": 7.2257,
      "loss/crossentropy": 1.7333999305963517,
      "loss/hidden": 3.49765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17861571693792938,
      "step": 4880
    },
    {
      "epoch": 0.04075,
      "grad_norm": 31.25,
      "grad_norm_var": 952.7306640625,
      "learning_rate": 0.0001,
      "loss": 7.2747,
      "loss/crossentropy": 1.9761729918420314,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18439243715256454,
      "step": 4890
    },
    {
      "epoch": 0.04083333333333333,
      "grad_norm": 31.5,
      "grad_norm_var": 951.7489583333333,
      "learning_rate": 0.0001,
      "loss": 7.1721,
      "loss/crossentropy": 1.976510974764824,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18401159718632698,
      "step": 4900
    },
    {
      "epoch": 0.040916666666666664,
      "grad_norm": 31.625,
      "grad_norm_var": 7.0353515625,
      "learning_rate": 0.0001,
      "loss": 7.3696,
      "loss/crossentropy": 1.8498829215765,
      "loss/hidden": 3.497265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19358056131750345,
      "step": 4910
    },
    {
      "epoch": 0.041,
      "grad_norm": 32.25,
      "grad_norm_var": 5.943684895833333,
      "learning_rate": 0.0001,
      "loss": 7.3129,
      "loss/crossentropy": 1.993993727862835,
      "loss/hidden": 3.331640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2032605566084385,
      "step": 4920
    },
    {
      "epoch": 0.04108333333333333,
      "grad_norm": 33.25,
      "grad_norm_var": 6.502018229166667,
      "learning_rate": 0.0001,
      "loss": 7.161,
      "loss/crossentropy": 1.8945767790079118,
      "loss/hidden": 3.301953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1947425028309226,
      "step": 4930
    },
    {
      "epoch": 0.041166666666666664,
      "grad_norm": 34.0,
      "grad_norm_var": 2.505989583333333,
      "learning_rate": 0.0001,
      "loss": 7.3919,
      "loss/crossentropy": 1.9350941553711891,
      "loss/hidden": 3.70625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20500725619494914,
      "step": 4940
    },
    {
      "epoch": 0.04125,
      "grad_norm": 29.875,
      "grad_norm_var": 18.84140625,
      "learning_rate": 0.0001,
      "loss": 7.2145,
      "loss/crossentropy": 2.050392009317875,
      "loss/hidden": 3.31015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18438252825289964,
      "step": 4950
    },
    {
      "epoch": 0.04133333333333333,
      "grad_norm": 37.5,
      "grad_norm_var": 21.251041666666666,
      "learning_rate": 0.0001,
      "loss": 7.148,
      "loss/crossentropy": 1.8834857508540153,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20490446221083403,
      "step": 4960
    },
    {
      "epoch": 0.041416666666666664,
      "grad_norm": 29.875,
      "grad_norm_var": 4.446809895833334,
      "learning_rate": 0.0001,
      "loss": 7.1951,
      "loss/crossentropy": 1.9395118832588196,
      "loss/hidden": 3.240625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18307873839512467,
      "step": 4970
    },
    {
      "epoch": 0.0415,
      "grad_norm": 31.5,
      "grad_norm_var": 4.9369140625,
      "learning_rate": 0.0001,
      "loss": 7.3024,
      "loss/crossentropy": 2.0352481067180634,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20577008351683618,
      "step": 4980
    },
    {
      "epoch": 0.04158333333333333,
      "grad_norm": 37.75,
      "grad_norm_var": 7.832217005980997e+17,
      "learning_rate": 0.0001,
      "loss": 7.2377,
      "loss/crossentropy": 1.969672729074955,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20676071532070636,
      "step": 4990
    },
    {
      "epoch": 0.041666666666666664,
      "grad_norm": 31.375,
      "grad_norm_var": 4.487955729166667,
      "learning_rate": 0.0001,
      "loss": 7.1396,
      "loss/crossentropy": 1.9443906158208848,
      "loss/hidden": 3.26953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18803976774215697,
      "step": 5000
    },
    {
      "epoch": 0.04175,
      "grad_norm": 36.25,
      "grad_norm_var": 1.848278976250949e+17,
      "learning_rate": 0.0001,
      "loss": 7.3254,
      "loss/crossentropy": 1.8845553174614906,
      "loss/hidden": 3.31015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17818044014275075,
      "step": 5010
    },
    {
      "epoch": 0.041833333333333333,
      "grad_norm": 31.75,
      "grad_norm_var": 1.848278975319464e+17,
      "learning_rate": 0.0001,
      "loss": 7.2489,
      "loss/crossentropy": 1.882246272265911,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1847611464560032,
      "step": 5020
    },
    {
      "epoch": 0.041916666666666665,
      "grad_norm": 2785017856.0,
      "grad_norm_var": 4.847702676187994e+17,
      "learning_rate": 0.0001,
      "loss": 7.2388,
      "loss/crossentropy": 2.051472933590412,
      "loss/hidden": 3.255859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18801066745072603,
      "step": 5030
    },
    {
      "epoch": 0.042,
      "grad_norm": 30.125,
      "grad_norm_var": 4.847702674592411e+17,
      "learning_rate": 0.0001,
      "loss": 7.1795,
      "loss/crossentropy": 1.8703616850078106,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20192000661045312,
      "step": 5040
    },
    {
      "epoch": 0.042083333333333334,
      "grad_norm": 49.75,
      "grad_norm_var": 22.523372395833334,
      "learning_rate": 0.0001,
      "loss": 7.1898,
      "loss/crossentropy": 1.9267778247594833,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18499485440552235,
      "step": 5050
    },
    {
      "epoch": 0.042166666666666665,
      "grad_norm": 30.75,
      "grad_norm_var": 25.1697265625,
      "learning_rate": 0.0001,
      "loss": 7.1415,
      "loss/crossentropy": 1.8913537606596946,
      "loss/hidden": 3.242578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1697727482765913,
      "step": 5060
    },
    {
      "epoch": 0.04225,
      "grad_norm": 28.75,
      "grad_norm_var": 2.05,
      "learning_rate": 0.0001,
      "loss": 7.1967,
      "loss/crossentropy": 2.001948466897011,
      "loss/hidden": 3.284765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18900279961526395,
      "step": 5070
    },
    {
      "epoch": 0.042333333333333334,
      "grad_norm": 30.25,
      "grad_norm_var": 2.2384765625,
      "learning_rate": 0.0001,
      "loss": 7.1734,
      "loss/crossentropy": 2.0092270568013193,
      "loss/hidden": 3.31796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18744736202061177,
      "step": 5080
    },
    {
      "epoch": 0.042416666666666665,
      "grad_norm": 35.25,
      "grad_norm_var": 4.459830729166667,
      "learning_rate": 0.0001,
      "loss": 7.3271,
      "loss/crossentropy": 1.9950493648648262,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18363405521959067,
      "step": 5090
    },
    {
      "epoch": 0.0425,
      "grad_norm": 33.75,
      "grad_norm_var": 3.466666666666667,
      "learning_rate": 0.0001,
      "loss": 7.2988,
      "loss/crossentropy": 1.96678287088871,
      "loss/hidden": 3.278515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19190824180841445,
      "step": 5100
    },
    {
      "epoch": 0.042583333333333334,
      "grad_norm": 30.625,
      "grad_norm_var": 3.0796223958333333,
      "learning_rate": 0.0001,
      "loss": 7.1883,
      "loss/crossentropy": 2.0521468251943586,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20394890960305928,
      "step": 5110
    },
    {
      "epoch": 0.042666666666666665,
      "grad_norm": 32.0,
      "grad_norm_var": 3.670768229166667,
      "learning_rate": 0.0001,
      "loss": 7.1932,
      "loss/crossentropy": 1.894241599738598,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18954524621367455,
      "step": 5120
    },
    {
      "epoch": 0.04275,
      "grad_norm": 32.75,
      "grad_norm_var": 4.25390625,
      "learning_rate": 0.0001,
      "loss": 7.1971,
      "loss/crossentropy": 1.9609125055372716,
      "loss/hidden": 3.294921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18426925949752332,
      "step": 5130
    },
    {
      "epoch": 0.042833333333333334,
      "grad_norm": 30.875,
      "grad_norm_var": 3.32265625,
      "learning_rate": 0.0001,
      "loss": 7.1633,
      "loss/crossentropy": 1.9538515985012055,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18203786462545396,
      "step": 5140
    },
    {
      "epoch": 0.042916666666666665,
      "grad_norm": 31.375,
      "grad_norm_var": 3.286393229166667,
      "learning_rate": 0.0001,
      "loss": 7.1574,
      "loss/crossentropy": 1.8362636983394622,
      "loss/hidden": 3.331640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1781992632895708,
      "step": 5150
    },
    {
      "epoch": 0.043,
      "grad_norm": 30.0,
      "grad_norm_var": 6.076497395833333,
      "learning_rate": 0.0001,
      "loss": 7.2141,
      "loss/crossentropy": 2.0191104084253313,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19397008065134286,
      "step": 5160
    },
    {
      "epoch": 0.043083333333333335,
      "grad_norm": 29.25,
      "grad_norm_var": 11.00390625,
      "learning_rate": 0.0001,
      "loss": 7.2431,
      "loss/crossentropy": 1.8258389428257942,
      "loss/hidden": 3.272265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18870493108406663,
      "step": 5170
    },
    {
      "epoch": 0.043166666666666666,
      "grad_norm": 34.0,
      "grad_norm_var": 11.309309895833334,
      "learning_rate": 0.0001,
      "loss": 7.2124,
      "loss/crossentropy": 1.9334841251373291,
      "loss/hidden": 3.464453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2005193768069148,
      "step": 5180
    },
    {
      "epoch": 0.04325,
      "grad_norm": 31.125,
      "grad_norm_var": 5.317122395833334,
      "learning_rate": 0.0001,
      "loss": 7.1607,
      "loss/crossentropy": 1.9018930494785309,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1909161601215601,
      "step": 5190
    },
    {
      "epoch": 0.043333333333333335,
      "grad_norm": 29.125,
      "grad_norm_var": 5.542708333333334,
      "learning_rate": 0.0001,
      "loss": 7.1734,
      "loss/crossentropy": 1.9978152081370353,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20025275563821196,
      "step": 5200
    },
    {
      "epoch": 0.043416666666666666,
      "grad_norm": 30.75,
      "grad_norm_var": 24.744791666666668,
      "learning_rate": 0.0001,
      "loss": 7.2824,
      "loss/crossentropy": 1.9716323763132095,
      "loss/hidden": 3.33984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18187592513859271,
      "step": 5210
    },
    {
      "epoch": 0.0435,
      "grad_norm": 30.0,
      "grad_norm_var": 83.13307291666666,
      "learning_rate": 0.0001,
      "loss": 7.1963,
      "loss/crossentropy": 2.118371768295765,
      "loss/hidden": 3.206640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18061571158468723,
      "step": 5220
    },
    {
      "epoch": 0.043583333333333335,
      "grad_norm": 31.375,
      "grad_norm_var": 83.16354166666666,
      "learning_rate": 0.0001,
      "loss": 7.1726,
      "loss/crossentropy": 1.953398549556732,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18984109461307525,
      "step": 5230
    },
    {
      "epoch": 0.043666666666666666,
      "grad_norm": 32.5,
      "grad_norm_var": 4.545572916666667,
      "learning_rate": 0.0001,
      "loss": 7.1202,
      "loss/crossentropy": 1.9805527925491333,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19125975519418717,
      "step": 5240
    },
    {
      "epoch": 0.04375,
      "grad_norm": 31.75,
      "grad_norm_var": 5.461393229166666,
      "learning_rate": 0.0001,
      "loss": 7.2709,
      "loss/crossentropy": 1.9394566029310227,
      "loss/hidden": 3.29140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17967610489577054,
      "step": 5250
    },
    {
      "epoch": 0.043833333333333335,
      "grad_norm": 30.125,
      "grad_norm_var": 12.862239583333333,
      "learning_rate": 0.0001,
      "loss": 7.1176,
      "loss/crossentropy": 1.933772175014019,
      "loss/hidden": 3.349609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18062447737902404,
      "step": 5260
    },
    {
      "epoch": 0.043916666666666666,
      "grad_norm": 31.25,
      "grad_norm_var": 14.672330729166667,
      "learning_rate": 0.0001,
      "loss": 7.2236,
      "loss/crossentropy": 2.000246451795101,
      "loss/hidden": 3.305859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18543695881962777,
      "step": 5270
    },
    {
      "epoch": 0.044,
      "grad_norm": 33.0,
      "grad_norm_var": 2.2853515625,
      "learning_rate": 0.0001,
      "loss": 7.2063,
      "loss/crossentropy": 1.8413677945733071,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18291340861469507,
      "step": 5280
    },
    {
      "epoch": 0.044083333333333335,
      "grad_norm": 31.125,
      "grad_norm_var": 1.5697265625,
      "learning_rate": 0.0001,
      "loss": 7.2327,
      "loss/crossentropy": 1.9906944945454597,
      "loss/hidden": 3.312109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18375325668603182,
      "step": 5290
    },
    {
      "epoch": 0.04416666666666667,
      "grad_norm": 34.75,
      "grad_norm_var": 6.108333333333333,
      "learning_rate": 0.0001,
      "loss": 7.3108,
      "loss/crossentropy": 1.8987505346536637,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1902456210926175,
      "step": 5300
    },
    {
      "epoch": 0.04425,
      "grad_norm": 32.0,
      "grad_norm_var": 5.784309895833333,
      "learning_rate": 0.0001,
      "loss": 7.1342,
      "loss/crossentropy": 2.0367724671959877,
      "loss/hidden": 3.45234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1810287894681096,
      "step": 5310
    },
    {
      "epoch": 0.044333333333333336,
      "grad_norm": 31.0,
      "grad_norm_var": 3.567122395833333,
      "learning_rate": 0.0001,
      "loss": 7.1938,
      "loss/crossentropy": 1.8333687089383601,
      "loss/hidden": 3.273828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1711188904941082,
      "step": 5320
    },
    {
      "epoch": 0.04441666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 1.0830729166666666,
      "learning_rate": 0.0001,
      "loss": 7.1841,
      "loss/crossentropy": 2.008009234815836,
      "loss/hidden": 3.408203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1856338769197464,
      "step": 5330
    },
    {
      "epoch": 0.0445,
      "grad_norm": 31.125,
      "grad_norm_var": 2.0052083333333335,
      "learning_rate": 0.0001,
      "loss": 7.312,
      "loss/crossentropy": 2.0298721551895142,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19467000029981135,
      "step": 5340
    },
    {
      "epoch": 0.044583333333333336,
      "grad_norm": 30.625,
      "grad_norm_var": 1.8518229166666667,
      "learning_rate": 0.0001,
      "loss": 7.2527,
      "loss/crossentropy": 1.946864341199398,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18655043169856073,
      "step": 5350
    },
    {
      "epoch": 0.04466666666666667,
      "grad_norm": 28.625,
      "grad_norm_var": 8.37890625,
      "learning_rate": 0.0001,
      "loss": 7.2448,
      "loss/crossentropy": 1.986108300089836,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1911590563133359,
      "step": 5360
    },
    {
      "epoch": 0.04475,
      "grad_norm": 31.75,
      "grad_norm_var": 8.673893229166667,
      "learning_rate": 0.0001,
      "loss": 7.2376,
      "loss/crossentropy": 1.9386815264821053,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18868138398975134,
      "step": 5370
    },
    {
      "epoch": 0.044833333333333336,
      "grad_norm": 30.875,
      "grad_norm_var": 4.776822916666666,
      "learning_rate": 0.0001,
      "loss": 7.2728,
      "loss/crossentropy": 1.9182999551296234,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1911846835166216,
      "step": 5380
    },
    {
      "epoch": 0.04491666666666667,
      "grad_norm": 31.25,
      "grad_norm_var": 3.9879557291666665,
      "learning_rate": 0.0001,
      "loss": 7.2084,
      "loss/crossentropy": 2.096855753660202,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1850230830721557,
      "step": 5390
    },
    {
      "epoch": 0.045,
      "grad_norm": 30.875,
      "grad_norm_var": 1.78515625,
      "learning_rate": 0.0001,
      "loss": 7.2046,
      "loss/crossentropy": 2.1250849097967146,
      "loss/hidden": 3.326171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18073338493704796,
      "step": 5400
    },
    {
      "epoch": 0.045083333333333336,
      "grad_norm": 33.25,
      "grad_norm_var": 5.0245041383900454e+17,
      "learning_rate": 0.0001,
      "loss": 7.4045,
      "loss/crossentropy": 1.8851775869727134,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1752549959346652,
      "step": 5410
    },
    {
      "epoch": 0.04516666666666667,
      "grad_norm": 35.5,
      "grad_norm_var": 5.024504136293068e+17,
      "learning_rate": 0.0001,
      "loss": 7.3242,
      "loss/crossentropy": 2.0536662988364696,
      "loss/hidden": 3.443359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19238244714215397,
      "step": 5420
    },
    {
      "epoch": 0.04525,
      "grad_norm": 31.0,
      "grad_norm_var": 37.017643229166666,
      "learning_rate": 0.0001,
      "loss": 7.2895,
      "loss/crossentropy": 1.911766418814659,
      "loss/hidden": 3.533203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18743917755782605,
      "step": 5430
    },
    {
      "epoch": 0.04533333333333334,
      "grad_norm": 30.0,
      "grad_norm_var": 17.709375,
      "learning_rate": 0.0001,
      "loss": 7.2069,
      "loss/crossentropy": 2.0367508992552756,
      "loss/hidden": 3.52109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2101318620145321,
      "step": 5440
    },
    {
      "epoch": 0.04541666666666667,
      "grad_norm": 30.625,
      "grad_norm_var": 9.678580729166667,
      "learning_rate": 0.0001,
      "loss": 7.2162,
      "loss/crossentropy": 1.9954749569296837,
      "loss/hidden": 3.311328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17785828560590744,
      "step": 5450
    },
    {
      "epoch": 0.0455,
      "grad_norm": 30.0,
      "grad_norm_var": 9.764102533822118e+16,
      "learning_rate": 0.0001,
      "loss": 7.2735,
      "loss/crossentropy": 2.088523915410042,
      "loss/hidden": 3.2859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.178264039196074,
      "step": 5460
    },
    {
      "epoch": 0.04558333333333333,
      "grad_norm": 38.75,
      "grad_norm_var": 204.03326822916668,
      "learning_rate": 0.0001,
      "loss": 7.1777,
      "loss/crossentropy": 1.9395600646734237,
      "loss/hidden": 3.28359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1719443194568157,
      "step": 5470
    },
    {
      "epoch": 0.04566666666666667,
      "grad_norm": 34.0,
      "grad_norm_var": 197.82890625,
      "learning_rate": 0.0001,
      "loss": 7.3162,
      "loss/crossentropy": 1.8322873935103416,
      "loss/hidden": 3.364453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17720702290534973,
      "step": 5480
    },
    {
      "epoch": 0.04575,
      "grad_norm": 31.25,
      "grad_norm_var": 2.2583333333333333,
      "learning_rate": 0.0001,
      "loss": 7.188,
      "loss/crossentropy": 1.9261434845626355,
      "loss/hidden": 3.338671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18068330511450767,
      "step": 5490
    },
    {
      "epoch": 0.04583333333333333,
      "grad_norm": 29.5,
      "grad_norm_var": 1.4999801304737034e+18,
      "learning_rate": 0.0001,
      "loss": 7.3305,
      "loss/crossentropy": 1.8927360117435454,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1788069920614362,
      "step": 5500
    },
    {
      "epoch": 0.04591666666666667,
      "grad_norm": 29.0,
      "grad_norm_var": 7.321164001103054e+17,
      "learning_rate": 0.0001,
      "loss": 7.3202,
      "loss/crossentropy": 1.9236746475100517,
      "loss/hidden": 3.2640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17970658484846352,
      "step": 5510
    },
    {
      "epoch": 0.046,
      "grad_norm": 31.875,
      "grad_norm_var": 1.2517332698401866e+18,
      "learning_rate": 0.0001,
      "loss": 7.2784,
      "loss/crossentropy": 1.7933879487216473,
      "loss/hidden": 3.293359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16929342793300747,
      "step": 5520
    },
    {
      "epoch": 0.04608333333333333,
      "grad_norm": 29.125,
      "grad_norm_var": 2.5488932291666666,
      "learning_rate": 0.0001,
      "loss": 7.0643,
      "loss/crossentropy": 1.9319801151752471,
      "loss/hidden": 3.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1735060378909111,
      "step": 5530
    },
    {
      "epoch": 0.04616666666666667,
      "grad_norm": 33.5,
      "grad_norm_var": 1.5353515625,
      "learning_rate": 0.0001,
      "loss": 7.2155,
      "loss/crossentropy": 2.0292537212371826,
      "loss/hidden": 3.48203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2000424973666668,
      "step": 5540
    },
    {
      "epoch": 0.04625,
      "grad_norm": 33.0,
      "grad_norm_var": 4.726497395833333,
      "learning_rate": 0.0001,
      "loss": 7.3183,
      "loss/crossentropy": 2.046224331855774,
      "loss/hidden": 3.470703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20262836031615733,
      "step": 5550
    },
    {
      "epoch": 0.04633333333333333,
      "grad_norm": 30.875,
      "grad_norm_var": 4.4666015625,
      "learning_rate": 0.0001,
      "loss": 7.1892,
      "loss/crossentropy": 1.9176844008266927,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1835050239227712,
      "step": 5560
    },
    {
      "epoch": 0.04641666666666667,
      "grad_norm": 30.375,
      "grad_norm_var": 1.359594487167984e+18,
      "learning_rate": 0.0001,
      "loss": 7.25,
      "loss/crossentropy": 1.9457813799381256,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19265421126037835,
      "step": 5570
    },
    {
      "epoch": 0.0465,
      "grad_norm": 29.25,
      "grad_norm_var": 9.5369140625,
      "learning_rate": 0.0001,
      "loss": 7.0833,
      "loss/crossentropy": 1.8953604146838188,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19760203436017038,
      "step": 5580
    },
    {
      "epoch": 0.04658333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 2.1322265625,
      "learning_rate": 0.0001,
      "loss": 7.2633,
      "loss/crossentropy": 2.192387434840202,
      "loss/hidden": 3.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20997173935174943,
      "step": 5590
    },
    {
      "epoch": 0.04666666666666667,
      "grad_norm": 29.875,
      "grad_norm_var": 1.9613932291666667,
      "learning_rate": 0.0001,
      "loss": 7.1848,
      "loss/crossentropy": 1.9025974899530411,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18724445197731257,
      "step": 5600
    },
    {
      "epoch": 0.04675,
      "grad_norm": 30.5,
      "grad_norm_var": 2.4145833333333333,
      "learning_rate": 0.0001,
      "loss": 7.2077,
      "loss/crossentropy": 1.8980001240968705,
      "loss/hidden": 3.425,
      "loss/jsd": 0.0,
      "loss/logits": 0.18669269513338804,
      "step": 5610
    },
    {
      "epoch": 0.04683333333333333,
      "grad_norm": 30.875,
      "grad_norm_var": 157.37682291666667,
      "learning_rate": 0.0001,
      "loss": 7.3404,
      "loss/crossentropy": 2.0061571091413497,
      "loss/hidden": 3.483984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2119721319526434,
      "step": 5620
    },
    {
      "epoch": 0.04691666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 4.900455729166667,
      "learning_rate": 0.0001,
      "loss": 7.307,
      "loss/crossentropy": 1.9688873529434203,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.1799240939319134,
      "step": 5630
    },
    {
      "epoch": 0.047,
      "grad_norm": 36.0,
      "grad_norm_var": 4.467708333333333,
      "learning_rate": 0.0001,
      "loss": 7.1158,
      "loss/crossentropy": 1.9035341456532477,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894345859065652,
      "step": 5640
    },
    {
      "epoch": 0.04708333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 5.3541015625,
      "learning_rate": 0.0001,
      "loss": 7.2714,
      "loss/crossentropy": 1.98752211779356,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18840861935168504,
      "step": 5650
    },
    {
      "epoch": 0.04716666666666667,
      "grad_norm": 30.375,
      "grad_norm_var": 8.378580729166666,
      "learning_rate": 0.0001,
      "loss": 7.124,
      "loss/crossentropy": 1.8570011980831622,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17682648468762635,
      "step": 5660
    },
    {
      "epoch": 0.04725,
      "grad_norm": 30.5,
      "grad_norm_var": 1.8770182291666666,
      "learning_rate": 0.0001,
      "loss": 7.1694,
      "loss/crossentropy": 1.944897037744522,
      "loss/hidden": 3.343359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17851272821426392,
      "step": 5670
    },
    {
      "epoch": 0.04733333333333333,
      "grad_norm": 29.75,
      "grad_norm_var": 6.12890625,
      "learning_rate": 0.0001,
      "loss": 7.2989,
      "loss/crossentropy": 2.0170719534158708,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19368235394358635,
      "step": 5680
    },
    {
      "epoch": 0.04741666666666667,
      "grad_norm": 30.875,
      "grad_norm_var": 5.084141647551967e+17,
      "learning_rate": 0.0001,
      "loss": 7.2023,
      "loss/crossentropy": 1.8989535093307495,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889783263206482,
      "step": 5690
    },
    {
      "epoch": 0.0475,
      "grad_norm": 34.5,
      "grad_norm_var": 3.012955729166667,
      "learning_rate": 0.0001,
      "loss": 7.2679,
      "loss/crossentropy": 1.937063890695572,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2034616844728589,
      "step": 5700
    },
    {
      "epoch": 0.04758333333333333,
      "grad_norm": 30.375,
      "grad_norm_var": 2.137955729166667,
      "learning_rate": 0.0001,
      "loss": 7.1541,
      "loss/crossentropy": 1.8730784460902214,
      "loss/hidden": 3.303125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18296674843877553,
      "step": 5710
    },
    {
      "epoch": 0.04766666666666667,
      "grad_norm": 30.375,
      "grad_norm_var": 50.691080729166664,
      "learning_rate": 0.0001,
      "loss": 7.2326,
      "loss/crossentropy": 1.9757542133331298,
      "loss/hidden": 3.299609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1815430372953415,
      "step": 5720
    },
    {
      "epoch": 0.04775,
      "grad_norm": 27.5,
      "grad_norm_var": 76.8822265625,
      "learning_rate": 0.0001,
      "loss": 7.2095,
      "loss/crossentropy": 1.8371678739786148,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1928836090490222,
      "step": 5730
    },
    {
      "epoch": 0.04783333333333333,
      "grad_norm": 34.0,
      "grad_norm_var": 7.036874271525002e+17,
      "learning_rate": 0.0001,
      "loss": 7.315,
      "loss/crossentropy": 1.9386666312813758,
      "loss/hidden": 3.57734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1786883294582367,
      "step": 5740
    },
    {
      "epoch": 0.04791666666666667,
      "grad_norm": 33.25,
      "grad_norm_var": 4.572330729166667,
      "learning_rate": 0.0001,
      "loss": 7.112,
      "loss/crossentropy": 1.8475066609680653,
      "loss/hidden": 3.22890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1685674911364913,
      "step": 5750
    },
    {
      "epoch": 0.048,
      "grad_norm": 32.0,
      "grad_norm_var": 5.036393229166666,
      "learning_rate": 0.0001,
      "loss": 7.2146,
      "loss/crossentropy": 1.8362929932773113,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18263157112523914,
      "step": 5760
    },
    {
      "epoch": 0.04808333333333333,
      "grad_norm": 41.0,
      "grad_norm_var": 5.0841416496613523e+17,
      "learning_rate": 0.0001,
      "loss": 7.2376,
      "loss/crossentropy": 1.992903184890747,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.18581640850752593,
      "step": 5770
    },
    {
      "epoch": 0.04816666666666667,
      "grad_norm": 34.25,
      "grad_norm_var": 5.0841416490077395e+17,
      "learning_rate": 0.0001,
      "loss": 7.1021,
      "loss/crossentropy": 1.8959848627448082,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1710287369787693,
      "step": 5780
    },
    {
      "epoch": 0.04825,
      "grad_norm": 35.0,
      "grad_norm_var": 289.79993489583336,
      "learning_rate": 0.0001,
      "loss": 7.2732,
      "loss/crossentropy": 1.9366097539663314,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18170985132455825,
      "step": 5790
    },
    {
      "epoch": 0.04833333333333333,
      "grad_norm": 54.5,
      "grad_norm_var": 34.83639322916667,
      "learning_rate": 0.0001,
      "loss": 7.2101,
      "loss/crossentropy": 2.073981301486492,
      "loss/hidden": 3.446875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20776797477155923,
      "step": 5800
    },
    {
      "epoch": 0.04841666666666666,
      "grad_norm": 32.5,
      "grad_norm_var": 35.93515625,
      "learning_rate": 0.0001,
      "loss": 7.2147,
      "loss/crossentropy": 1.8804841123521328,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16808632388710976,
      "step": 5810
    },
    {
      "epoch": 0.0485,
      "grad_norm": 33.5,
      "grad_norm_var": 1.8875,
      "learning_rate": 0.0001,
      "loss": 7.1936,
      "loss/crossentropy": 1.9499920338392258,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18369101714342834,
      "step": 5820
    },
    {
      "epoch": 0.04858333333333333,
      "grad_norm": 32.5,
      "grad_norm_var": 3.4247395833333334,
      "learning_rate": 0.0001,
      "loss": 7.2225,
      "loss/crossentropy": 1.8653924077749253,
      "loss/hidden": 3.526171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19181002490222454,
      "step": 5830
    },
    {
      "epoch": 0.048666666666666664,
      "grad_norm": 33.75,
      "grad_norm_var": 1.1439318802114807e+18,
      "learning_rate": 0.0001,
      "loss": 7.3217,
      "loss/crossentropy": 1.969824068248272,
      "loss/hidden": 3.58828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20675076451152563,
      "step": 5840
    },
    {
      "epoch": 0.04875,
      "grad_norm": 38.25,
      "grad_norm_var": 5.338541666666667,
      "learning_rate": 0.0001,
      "loss": 7.141,
      "loss/crossentropy": 1.9277636885643006,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18821494542062284,
      "step": 5850
    },
    {
      "epoch": 0.04883333333333333,
      "grad_norm": 3019898880.0,
      "grad_norm_var": 5.6998681567140256e+17,
      "learning_rate": 0.0001,
      "loss": 7.2917,
      "loss/crossentropy": 1.8802243813872337,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18886747639626264,
      "step": 5860
    },
    {
      "epoch": 0.048916666666666664,
      "grad_norm": 31.125,
      "grad_norm_var": 5.6998681553613626e+17,
      "learning_rate": 0.0001,
      "loss": 7.2099,
      "loss/crossentropy": 2.0238874942064284,
      "loss/hidden": 3.3203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17390586640685796,
      "step": 5870
    },
    {
      "epoch": 0.049,
      "grad_norm": 37.5,
      "grad_norm_var": 9.881354820405691e+17,
      "learning_rate": 0.0001,
      "loss": 7.2301,
      "loss/crossentropy": 2.0223625183105467,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19623048808425664,
      "step": 5880
    },
    {
      "epoch": 0.04908333333333333,
      "grad_norm": 30.125,
      "grad_norm_var": 9.881354818210497e+17,
      "learning_rate": 0.0001,
      "loss": 7.1501,
      "loss/crossentropy": 1.8737989097833634,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16763302199542524,
      "step": 5890
    },
    {
      "epoch": 0.049166666666666664,
      "grad_norm": 32.0,
      "grad_norm_var": 5.7712890625,
      "learning_rate": 0.0001,
      "loss": 7.2475,
      "loss/crossentropy": 1.8810926377773285,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1831731954589486,
      "step": 5900
    },
    {
      "epoch": 0.04925,
      "grad_norm": 31.125,
      "grad_norm_var": 3.748893229166667,
      "learning_rate": 0.0001,
      "loss": 7.2608,
      "loss/crossentropy": 2.062958613038063,
      "loss/hidden": 3.55390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21889359913766385,
      "step": 5910
    },
    {
      "epoch": 0.04933333333333333,
      "grad_norm": 35.5,
      "grad_norm_var": 1021.2436848958333,
      "learning_rate": 0.0001,
      "loss": 7.2497,
      "loss/crossentropy": 1.898280094563961,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18923573587089776,
      "step": 5920
    },
    {
      "epoch": 0.049416666666666664,
      "grad_norm": 31.875,
      "grad_norm_var": 1012.7455729166667,
      "learning_rate": 0.0001,
      "loss": 7.283,
      "loss/crossentropy": 2.062563756108284,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18664450161159038,
      "step": 5930
    },
    {
      "epoch": 0.0495,
      "grad_norm": 34.25,
      "grad_norm_var": 6.758214050691854e+17,
      "learning_rate": 0.0001,
      "loss": 7.377,
      "loss/crossentropy": 1.9490419641137122,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1910331157967448,
      "step": 5940
    },
    {
      "epoch": 0.04958333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 6.758214053740413e+17,
      "learning_rate": 0.0001,
      "loss": 7.263,
      "loss/crossentropy": 1.9519288882613182,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19702470256015658,
      "step": 5950
    },
    {
      "epoch": 0.049666666666666665,
      "grad_norm": 31.25,
      "grad_norm_var": 1.3080729166666667,
      "learning_rate": 0.0001,
      "loss": 7.3382,
      "loss/crossentropy": 2.08356169462204,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1919506683945656,
      "step": 5960
    },
    {
      "epoch": 0.04975,
      "grad_norm": 33.25,
      "grad_norm_var": 9.986393229166667,
      "learning_rate": 0.0001,
      "loss": 7.2119,
      "loss/crossentropy": 1.8876198440790177,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18486291356384754,
      "step": 5970
    },
    {
      "epoch": 0.049833333333333334,
      "grad_norm": 31.875,
      "grad_norm_var": 3.50390625,
      "learning_rate": 0.0001,
      "loss": 7.162,
      "loss/crossentropy": 1.9698369055986404,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18667118791490794,
      "step": 5980
    },
    {
      "epoch": 0.049916666666666665,
      "grad_norm": 31.0,
      "grad_norm_var": 53.614322916666666,
      "learning_rate": 0.0001,
      "loss": 7.3357,
      "loss/crossentropy": 2.067339900135994,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2117473976686597,
      "step": 5990
    },
    {
      "epoch": 0.05,
      "grad_norm": 31.0,
      "grad_norm_var": 45.57890625,
      "learning_rate": 0.0001,
      "loss": 7.195,
      "loss/crossentropy": 1.9594978839159012,
      "loss/hidden": 3.250390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1724390959367156,
      "step": 6000
    },
    {
      "epoch": 0.050083333333333334,
      "grad_norm": 34.25,
      "grad_norm_var": 2.4427083333333335,
      "learning_rate": 0.0001,
      "loss": 7.1326,
      "loss/crossentropy": 1.8880977675318718,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18454405162483453,
      "step": 6010
    },
    {
      "epoch": 0.050166666666666665,
      "grad_norm": 31.0,
      "grad_norm_var": 4.503599518374517e+17,
      "learning_rate": 0.0001,
      "loss": 7.3398,
      "loss/crossentropy": 2.040569248795509,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19138383716344834,
      "step": 6020
    },
    {
      "epoch": 0.05025,
      "grad_norm": 31.875,
      "grad_norm_var": 1.85390625,
      "learning_rate": 0.0001,
      "loss": 7.1766,
      "loss/crossentropy": 2.0133946374058724,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18018098138272762,
      "step": 6030
    },
    {
      "epoch": 0.050333333333333334,
      "grad_norm": 30.75,
      "grad_norm_var": 4.374739583333334,
      "learning_rate": 0.0001,
      "loss": 7.4233,
      "loss/crossentropy": 2.0365489065647124,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1943318447098136,
      "step": 6040
    },
    {
      "epoch": 0.050416666666666665,
      "grad_norm": 38.0,
      "grad_norm_var": 6.753580729166667,
      "learning_rate": 0.0001,
      "loss": 7.1707,
      "loss/crossentropy": 2.019296956062317,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1895071506500244,
      "step": 6050
    },
    {
      "epoch": 0.0505,
      "grad_norm": 31.0,
      "grad_norm_var": 5.6509765625,
      "learning_rate": 0.0001,
      "loss": 7.1773,
      "loss/crossentropy": 1.9771725565195084,
      "loss/hidden": 3.310546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18213878907263278,
      "step": 6060
    },
    {
      "epoch": 0.050583333333333334,
      "grad_norm": 33.5,
      "grad_norm_var": 1.2442057291666666,
      "learning_rate": 0.0001,
      "loss": 7.0858,
      "loss/crossentropy": 1.9022766441106795,
      "loss/hidden": 3.548828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20166466273367406,
      "step": 6070
    },
    {
      "epoch": 0.050666666666666665,
      "grad_norm": 28.875,
      "grad_norm_var": 1.9160807291666666,
      "learning_rate": 0.0001,
      "loss": 7.1795,
      "loss/crossentropy": 1.961769598722458,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1797276422381401,
      "step": 6080
    },
    {
      "epoch": 0.05075,
      "grad_norm": 30.125,
      "grad_norm_var": 6.016666666666667,
      "learning_rate": 0.0001,
      "loss": 7.1952,
      "loss/crossentropy": 1.9573363393545151,
      "loss/hidden": 3.24765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16141145564615728,
      "step": 6090
    },
    {
      "epoch": 0.050833333333333335,
      "grad_norm": 35.75,
      "grad_norm_var": 140.92545572916666,
      "learning_rate": 0.0001,
      "loss": 7.2033,
      "loss/crossentropy": 1.8590539634227752,
      "loss/hidden": 3.386328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18147912044078113,
      "step": 6100
    },
    {
      "epoch": 0.050916666666666666,
      "grad_norm": 33.0,
      "grad_norm_var": 3.3018772934252755e+17,
      "learning_rate": 0.0001,
      "loss": 7.3202,
      "loss/crossentropy": 2.0075626581907273,
      "loss/hidden": 3.6578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2295171281322837,
      "step": 6110
    },
    {
      "epoch": 0.051,
      "grad_norm": 44.0,
      "grad_norm_var": 18.712434895833333,
      "learning_rate": 0.0001,
      "loss": 7.1691,
      "loss/crossentropy": 1.8493891522288322,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17705651726573707,
      "step": 6120
    },
    {
      "epoch": 0.051083333333333335,
      "grad_norm": 30.5,
      "grad_norm_var": 19.46015625,
      "learning_rate": 0.0001,
      "loss": 7.1577,
      "loss/crossentropy": 2.1470450803637506,
      "loss/hidden": 3.2515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17646269779652357,
      "step": 6130
    },
    {
      "epoch": 0.051166666666666666,
      "grad_norm": 30.5,
      "grad_norm_var": 7.367708333333334,
      "learning_rate": 0.0001,
      "loss": 7.2079,
      "loss/crossentropy": 1.9435604512691498,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1852132784202695,
      "step": 6140
    },
    {
      "epoch": 0.05125,
      "grad_norm": 31.25,
      "grad_norm_var": 4.164322916666666,
      "learning_rate": 0.0001,
      "loss": 7.2506,
      "loss/crossentropy": 1.8571129024028779,
      "loss/hidden": 3.533984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20155182387679815,
      "step": 6150
    },
    {
      "epoch": 0.051333333333333335,
      "grad_norm": 32.75,
      "grad_norm_var": 2.8613932291666666,
      "learning_rate": 0.0001,
      "loss": 7.1611,
      "loss/crossentropy": 1.930855180323124,
      "loss/hidden": 3.4859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17727227360010148,
      "step": 6160
    },
    {
      "epoch": 0.051416666666666666,
      "grad_norm": 35.5,
      "grad_norm_var": 4.572916666666667,
      "learning_rate": 0.0001,
      "loss": 7.2118,
      "loss/crossentropy": 1.973878675699234,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19820086359977723,
      "step": 6170
    },
    {
      "epoch": 0.0515,
      "grad_norm": 29.875,
      "grad_norm_var": 7.744205729166667,
      "learning_rate": 0.0001,
      "loss": 7.3277,
      "loss/crossentropy": 2.0238889157772064,
      "loss/hidden": 3.530078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2186591826379299,
      "step": 6180
    },
    {
      "epoch": 0.051583333333333335,
      "grad_norm": 32.5,
      "grad_norm_var": 2.8692057291666666,
      "learning_rate": 0.0001,
      "loss": 7.1675,
      "loss/crossentropy": 1.9423449069261551,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19517161659896373,
      "step": 6190
    },
    {
      "epoch": 0.051666666666666666,
      "grad_norm": 31.125,
      "grad_norm_var": 11.745572916666667,
      "learning_rate": 0.0001,
      "loss": 7.3246,
      "loss/crossentropy": 2.0157795667648317,
      "loss/hidden": 3.300390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19794415701180695,
      "step": 6200
    },
    {
      "epoch": 0.05175,
      "grad_norm": 31.0,
      "grad_norm_var": 73.48723958333333,
      "learning_rate": 0.0001,
      "loss": 7.1391,
      "loss/crossentropy": 1.8031481310725213,
      "loss/hidden": 3.5703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18407038692384958,
      "step": 6210
    },
    {
      "epoch": 0.051833333333333335,
      "grad_norm": 31.125,
      "grad_norm_var": 3.1051432291666665,
      "learning_rate": 0.0001,
      "loss": 7.1885,
      "loss/crossentropy": 2.1037486374378203,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19484200291335582,
      "step": 6220
    },
    {
      "epoch": 0.051916666666666667,
      "grad_norm": 34.0,
      "grad_norm_var": 3.786393229166667,
      "learning_rate": 0.0001,
      "loss": 7.2227,
      "loss/crossentropy": 1.926634880900383,
      "loss/hidden": 3.423828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19628361649811268,
      "step": 6230
    },
    {
      "epoch": 0.052,
      "grad_norm": 29.75,
      "grad_norm_var": 2.2018229166666665,
      "learning_rate": 0.0001,
      "loss": 7.1722,
      "loss/crossentropy": 2.1934867203235626,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19669417291879654,
      "step": 6240
    },
    {
      "epoch": 0.052083333333333336,
      "grad_norm": 33.5,
      "grad_norm_var": 3.0205729166666666,
      "learning_rate": 0.0001,
      "loss": 7.2297,
      "loss/crossentropy": 1.886684399843216,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18135682679712772,
      "step": 6250
    },
    {
      "epoch": 0.05216666666666667,
      "grad_norm": 30.25,
      "grad_norm_var": 4.359375,
      "learning_rate": 0.0001,
      "loss": 7.2224,
      "loss/crossentropy": 1.9629553347826003,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19749300181865692,
      "step": 6260
    },
    {
      "epoch": 0.05225,
      "grad_norm": 31.125,
      "grad_norm_var": 6.217782107599537e+17,
      "learning_rate": 0.0001,
      "loss": 7.3326,
      "loss/crossentropy": 2.1625852316617964,
      "loss/hidden": 3.262890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18893173113465309,
      "step": 6270
    },
    {
      "epoch": 0.052333333333333336,
      "grad_norm": 29.75,
      "grad_norm_var": 6.217781961524684e+17,
      "learning_rate": 0.0001,
      "loss": 7.2523,
      "loss/crossentropy": 1.9883932605385781,
      "loss/hidden": 3.46796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21924990750849246,
      "step": 6280
    },
    {
      "epoch": 0.05241666666666667,
      "grad_norm": 29.375,
      "grad_norm_var": 23.372916666666665,
      "learning_rate": 0.0001,
      "loss": 7.1854,
      "loss/crossentropy": 2.079185092449188,
      "loss/hidden": 3.28671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18457219656556845,
      "step": 6290
    },
    {
      "epoch": 0.0525,
      "grad_norm": 33.0,
      "grad_norm_var": 19.95390625,
      "learning_rate": 0.0001,
      "loss": 7.2171,
      "loss/crossentropy": 1.9342011585831642,
      "loss/hidden": 3.265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18812895119190215,
      "step": 6300
    },
    {
      "epoch": 0.052583333333333336,
      "grad_norm": 30.25,
      "grad_norm_var": 2.0369140625,
      "learning_rate": 0.0001,
      "loss": 7.1543,
      "loss/crossentropy": 1.9667259231209755,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1891764525324106,
      "step": 6310
    },
    {
      "epoch": 0.05266666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 3.3833333333333333,
      "learning_rate": 0.0001,
      "loss": 7.1719,
      "loss/crossentropy": 1.883527809381485,
      "loss/hidden": 3.3453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18346853349357845,
      "step": 6320
    },
    {
      "epoch": 0.05275,
      "grad_norm": 32.75,
      "grad_norm_var": 1.1529214870959075e+18,
      "learning_rate": 0.0001,
      "loss": 7.277,
      "loss/crossentropy": 1.9553183883428573,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17686478048563004,
      "step": 6330
    },
    {
      "epoch": 0.052833333333333336,
      "grad_norm": 38.75,
      "grad_norm_var": 1.1529214856642518e+18,
      "learning_rate": 0.0001,
      "loss": 7.3041,
      "loss/crossentropy": 1.9471174091100694,
      "loss/hidden": 3.26171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17402494940906763,
      "step": 6340
    },
    {
      "epoch": 0.05291666666666667,
      "grad_norm": 32.0,
      "grad_norm_var": 25.320572916666666,
      "learning_rate": 0.0001,
      "loss": 7.161,
      "loss/crossentropy": 1.9604827284812927,
      "loss/hidden": 3.29140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19619520641863347,
      "step": 6350
    },
    {
      "epoch": 0.053,
      "grad_norm": 29.375,
      "grad_norm_var": 1.1171223958333334,
      "learning_rate": 0.0001,
      "loss": 7.1766,
      "loss/crossentropy": 2.016821599006653,
      "loss/hidden": 3.36875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20081317238509655,
      "step": 6360
    },
    {
      "epoch": 0.05308333333333334,
      "grad_norm": 31.125,
      "grad_norm_var": 2.622330729166667,
      "learning_rate": 0.0001,
      "loss": 7.3007,
      "loss/crossentropy": 2.0598457247018813,
      "loss/hidden": 3.478515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20236392691731453,
      "step": 6370
    },
    {
      "epoch": 0.05316666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 2.2853515625,
      "learning_rate": 0.0001,
      "loss": 7.2402,
      "loss/crossentropy": 1.9487939864397048,
      "loss/hidden": 3.283984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17032433431595564,
      "step": 6380
    },
    {
      "epoch": 0.05325,
      "grad_norm": 31.5,
      "grad_norm_var": 4.0697265625,
      "learning_rate": 0.0001,
      "loss": 7.2519,
      "loss/crossentropy": 1.79762829169631,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17330624535679817,
      "step": 6390
    },
    {
      "epoch": 0.05333333333333334,
      "grad_norm": 33.75,
      "grad_norm_var": 4.148958333333334,
      "learning_rate": 0.0001,
      "loss": 7.2327,
      "loss/crossentropy": 1.8139575704932214,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19837790299206973,
      "step": 6400
    },
    {
      "epoch": 0.05341666666666667,
      "grad_norm": 34.25,
      "grad_norm_var": 2.787955729166667,
      "learning_rate": 0.0001,
      "loss": 7.1422,
      "loss/crossentropy": 1.8317793741822244,
      "loss/hidden": 3.4140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19687960278242828,
      "step": 6410
    },
    {
      "epoch": 0.0535,
      "grad_norm": 30.75,
      "grad_norm_var": 1.9747395833333334,
      "learning_rate": 0.0001,
      "loss": 7.2303,
      "loss/crossentropy": 1.8772493988275527,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18726407140493392,
      "step": 6420
    },
    {
      "epoch": 0.05358333333333333,
      "grad_norm": 29.75,
      "grad_norm_var": 3.876822916666667,
      "learning_rate": 0.0001,
      "loss": 7.2184,
      "loss/crossentropy": 1.924004825949669,
      "loss/hidden": 3.4578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18612550031393765,
      "step": 6430
    },
    {
      "epoch": 0.05366666666666667,
      "grad_norm": 55.0,
      "grad_norm_var": 38.6572265625,
      "learning_rate": 0.0001,
      "loss": 7.2838,
      "loss/crossentropy": 1.971806037425995,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19059366807341577,
      "step": 6440
    },
    {
      "epoch": 0.05375,
      "grad_norm": 33.25,
      "grad_norm_var": 31.5244140625,
      "learning_rate": 0.0001,
      "loss": 7.2897,
      "loss/crossentropy": 2.0371891140937803,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19689792413264512,
      "step": 6450
    },
    {
      "epoch": 0.05383333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 7.057291666666667,
      "learning_rate": 0.0001,
      "loss": 7.1589,
      "loss/crossentropy": 1.815991611778736,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18021370992064475,
      "step": 6460
    },
    {
      "epoch": 0.05391666666666667,
      "grad_norm": 33.25,
      "grad_norm_var": 3.139518229166667,
      "learning_rate": 0.0001,
      "loss": 7.2613,
      "loss/crossentropy": 1.9856169998645783,
      "loss/hidden": 3.4453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19363545887172223,
      "step": 6470
    },
    {
      "epoch": 0.054,
      "grad_norm": 75.5,
      "grad_norm_var": 8467.976041666667,
      "learning_rate": 0.0001,
      "loss": 7.3998,
      "loss/crossentropy": 2.0012280732393264,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21550402007997035,
      "step": 6480
    },
    {
      "epoch": 0.05408333333333333,
      "grad_norm": 33.25,
      "grad_norm_var": 8393.768489583334,
      "learning_rate": 0.0001,
      "loss": 7.1939,
      "loss/crossentropy": 1.9625369042158127,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19765603579580784,
      "step": 6490
    },
    {
      "epoch": 0.05416666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 8.258333333333333,
      "learning_rate": 0.0001,
      "loss": 7.2261,
      "loss/crossentropy": 1.9609766840934753,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17874824199825526,
      "step": 6500
    },
    {
      "epoch": 0.05425,
      "grad_norm": 40.75,
      "grad_norm_var": 39.4853515625,
      "learning_rate": 0.0001,
      "loss": 7.1973,
      "loss/crossentropy": 2.050165778398514,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17521476447582246,
      "step": 6510
    },
    {
      "epoch": 0.05433333333333333,
      "grad_norm": 29.125,
      "grad_norm_var": 59.44583333333333,
      "learning_rate": 0.0001,
      "loss": 7.194,
      "loss/crossentropy": 1.8978053271770476,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1794705156236887,
      "step": 6520
    },
    {
      "epoch": 0.05441666666666667,
      "grad_norm": 32.75,
      "grad_norm_var": 1.263541147958615e+18,
      "learning_rate": 0.0001,
      "loss": 7.3198,
      "loss/crossentropy": 2.025116375088692,
      "loss/hidden": 3.44375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18641012702137233,
      "step": 6530
    },
    {
      "epoch": 0.0545,
      "grad_norm": 70.5,
      "grad_norm_var": 1.2635411437105544e+18,
      "learning_rate": 0.0001,
      "loss": 7.1999,
      "loss/crossentropy": 1.8800811529159547,
      "loss/hidden": 3.541015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20000641271471978,
      "step": 6540
    },
    {
      "epoch": 0.05458333333333333,
      "grad_norm": 37.5,
      "grad_norm_var": 366.4457682291667,
      "learning_rate": 0.0001,
      "loss": 7.2098,
      "loss/crossentropy": 1.88163965344429,
      "loss/hidden": 3.45703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18751533031463624,
      "step": 6550
    },
    {
      "epoch": 0.05466666666666667,
      "grad_norm": 30.0,
      "grad_norm_var": 13.5025390625,
      "learning_rate": 0.0001,
      "loss": 7.1116,
      "loss/crossentropy": 1.8538421645760537,
      "loss/hidden": 3.262890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17072358094155787,
      "step": 6560
    },
    {
      "epoch": 0.05475,
      "grad_norm": 36.25,
      "grad_norm_var": 16.593684895833334,
      "learning_rate": 0.0001,
      "loss": 7.1972,
      "loss/crossentropy": 1.9755418807268144,
      "loss/hidden": 3.25390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18982569519430398,
      "step": 6570
    },
    {
      "epoch": 0.05483333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 15.738997395833334,
      "learning_rate": 0.0001,
      "loss": 7.1361,
      "loss/crossentropy": 1.706727545708418,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18362588211894035,
      "step": 6580
    },
    {
      "epoch": 0.05491666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 13.137955729166666,
      "learning_rate": 0.0001,
      "loss": 7.247,
      "loss/crossentropy": 1.8340010598301888,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1724576648324728,
      "step": 6590
    },
    {
      "epoch": 0.055,
      "grad_norm": 40.25,
      "grad_norm_var": 13.128580729166666,
      "learning_rate": 0.0001,
      "loss": 7.3337,
      "loss/crossentropy": 1.93518505692482,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17210706733167172,
      "step": 6600
    },
    {
      "epoch": 0.05508333333333333,
      "grad_norm": 41.5,
      "grad_norm_var": 1.3990009827900785e+18,
      "learning_rate": 0.0001,
      "loss": 7.276,
      "loss/crossentropy": 1.979572981595993,
      "loss/hidden": 3.3390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19375349935144187,
      "step": 6610
    },
    {
      "epoch": 0.05516666666666667,
      "grad_norm": 31.75,
      "grad_norm_var": 1.3990009825535199e+18,
      "learning_rate": 0.0001,
      "loss": 7.1396,
      "loss/crossentropy": 1.8551251679658889,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1853062679991126,
      "step": 6620
    },
    {
      "epoch": 0.05525,
      "grad_norm": 38.0,
      "grad_norm_var": 27.79765625,
      "learning_rate": 0.0001,
      "loss": 7.4334,
      "loss/crossentropy": 1.956180915236473,
      "loss/hidden": 3.46484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2130817821249366,
      "step": 6630
    },
    {
      "epoch": 0.05533333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 35.608333333333334,
      "learning_rate": 0.0001,
      "loss": 7.1717,
      "loss/crossentropy": 1.8848136879503727,
      "loss/hidden": 3.32890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17505101822316646,
      "step": 6640
    },
    {
      "epoch": 0.05541666666666667,
      "grad_norm": 29.75,
      "grad_norm_var": 24.198958333333334,
      "learning_rate": 0.0001,
      "loss": 7.2696,
      "loss/crossentropy": 2.1142563939094545,
      "loss/hidden": 3.303125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2016379315406084,
      "step": 6650
    },
    {
      "epoch": 0.0555,
      "grad_norm": 32.5,
      "grad_norm_var": 1.18923175838631e+18,
      "learning_rate": 0.0001,
      "loss": 7.2787,
      "loss/crossentropy": 1.877379097044468,
      "loss/hidden": 3.28671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17101834136992694,
      "step": 6660
    },
    {
      "epoch": 0.05558333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 31.5,
      "learning_rate": 0.0001,
      "loss": 7.1617,
      "loss/crossentropy": 1.967728652060032,
      "loss/hidden": 3.2625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16908176708966494,
      "step": 6670
    },
    {
      "epoch": 0.05566666666666667,
      "grad_norm": 30.0,
      "grad_norm_var": 10.295833333333333,
      "learning_rate": 0.0001,
      "loss": 7.1966,
      "loss/crossentropy": 1.9864210113883018,
      "loss/hidden": 3.312890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17814714480191468,
      "step": 6680
    },
    {
      "epoch": 0.05575,
      "grad_norm": 30.25,
      "grad_norm_var": 8.192643229166666,
      "learning_rate": 0.0001,
      "loss": 7.2298,
      "loss/crossentropy": 1.9852227136492728,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18118114713579417,
      "step": 6690
    },
    {
      "epoch": 0.05583333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 4.3362978299559725e+17,
      "learning_rate": 0.0001,
      "loss": 7.2515,
      "loss/crossentropy": 1.9546830110251903,
      "loss/hidden": 3.516796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20408000089228154,
      "step": 6700
    },
    {
      "epoch": 0.05591666666666667,
      "grad_norm": 35.0,
      "grad_norm_var": 1.0178463309959793e+18,
      "learning_rate": 0.0001,
      "loss": 7.311,
      "loss/crossentropy": 1.8227561213076116,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1906059012748301,
      "step": 6710
    },
    {
      "epoch": 0.056,
      "grad_norm": 32.25,
      "grad_norm_var": 12.212434895833333,
      "learning_rate": 0.0001,
      "loss": 7.1884,
      "loss/crossentropy": 1.8481850996613503,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1807496551424265,
      "step": 6720
    },
    {
      "epoch": 0.05608333333333333,
      "grad_norm": 33.5,
      "grad_norm_var": 1.244752697754334e+18,
      "learning_rate": 0.0001,
      "loss": 7.3472,
      "loss/crossentropy": 2.0029425621032715,
      "loss/hidden": 3.3171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18047370873391627,
      "step": 6730
    },
    {
      "epoch": 0.05616666666666666,
      "grad_norm": 29.75,
      "grad_norm_var": 1.2447526969454625e+18,
      "learning_rate": 0.0001,
      "loss": 7.2428,
      "loss/crossentropy": 2.070193290710449,
      "loss/hidden": 3.2890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19471360072493554,
      "step": 6740
    },
    {
      "epoch": 0.05625,
      "grad_norm": 34.75,
      "grad_norm_var": 10.155143229166667,
      "learning_rate": 0.0001,
      "loss": 7.2488,
      "loss/crossentropy": 1.9138814441859722,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1864965006709099,
      "step": 6750
    },
    {
      "epoch": 0.05633333333333333,
      "grad_norm": 57.25,
      "grad_norm_var": 65.1634765625,
      "learning_rate": 0.0001,
      "loss": 7.2888,
      "loss/crossentropy": 1.9506281644105912,
      "loss/hidden": 3.5109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21938262470066547,
      "step": 6760
    },
    {
      "epoch": 0.056416666666666664,
      "grad_norm": 31.75,
      "grad_norm_var": 72.72493489583333,
      "learning_rate": 0.0001,
      "loss": 7.1287,
      "loss/crossentropy": 1.8292874485254287,
      "loss/hidden": 3.433984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17849782444536685,
      "step": 6770
    },
    {
      "epoch": 0.0565,
      "grad_norm": 35.25,
      "grad_norm_var": 1.7566853074176182e+18,
      "learning_rate": 0.0001,
      "loss": 7.2384,
      "loss/crossentropy": 1.8679364308714868,
      "loss/hidden": 3.41015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17026210222393273,
      "step": 6780
    },
    {
      "epoch": 0.05658333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 1.7566852899057772e+18,
      "learning_rate": 0.0001,
      "loss": 7.2511,
      "loss/crossentropy": 1.9325649484992027,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18963379934430122,
      "step": 6790
    },
    {
      "epoch": 0.056666666666666664,
      "grad_norm": 31.125,
      "grad_norm_var": 7281.924934895834,
      "learning_rate": 0.0001,
      "loss": 7.2677,
      "loss/crossentropy": 1.9077018201351166,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19200860038399697,
      "step": 6800
    },
    {
      "epoch": 0.05675,
      "grad_norm": 31.0,
      "grad_norm_var": 2.996875,
      "learning_rate": 0.0001,
      "loss": 7.1384,
      "loss/crossentropy": 1.9394373342394828,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18492530584335326,
      "step": 6810
    },
    {
      "epoch": 0.05683333333333333,
      "grad_norm": 27.75,
      "grad_norm_var": 5.395572916666667,
      "learning_rate": 0.0001,
      "loss": 7.2212,
      "loss/crossentropy": 1.9307297945022583,
      "loss/hidden": 3.49609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20020147040486336,
      "step": 6820
    },
    {
      "epoch": 0.056916666666666664,
      "grad_norm": 30.875,
      "grad_norm_var": 3.728125,
      "learning_rate": 0.0001,
      "loss": 7.1808,
      "loss/crossentropy": 1.8511271178722382,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18137127477675677,
      "step": 6830
    },
    {
      "epoch": 0.057,
      "grad_norm": 28.5,
      "grad_norm_var": 21.429166666666667,
      "learning_rate": 0.0001,
      "loss": 7.2759,
      "loss/crossentropy": 2.0388415962457658,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18986355643719435,
      "step": 6840
    },
    {
      "epoch": 0.05708333333333333,
      "grad_norm": 31.625,
      "grad_norm_var": 6.610872395833334,
      "learning_rate": 0.0001,
      "loss": 7.2423,
      "loss/crossentropy": 2.036486195027828,
      "loss/hidden": 3.365625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19451405573636293,
      "step": 6850
    },
    {
      "epoch": 0.057166666666666664,
      "grad_norm": 33.0,
      "grad_norm_var": 2.421875,
      "learning_rate": 0.0001,
      "loss": 7.3234,
      "loss/crossentropy": 1.9453188359737397,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18416350986808538,
      "step": 6860
    },
    {
      "epoch": 0.05725,
      "grad_norm": 30.875,
      "grad_norm_var": 38.827083333333334,
      "learning_rate": 0.0001,
      "loss": 7.2471,
      "loss/crossentropy": 2.118215595185757,
      "loss/hidden": 3.2765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18926584757864476,
      "step": 6870
    },
    {
      "epoch": 0.05733333333333333,
      "grad_norm": 28.25,
      "grad_norm_var": 42.50462239583333,
      "learning_rate": 0.0001,
      "loss": 7.1489,
      "loss/crossentropy": 2.0976534694433213,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18597040716558694,
      "step": 6880
    },
    {
      "epoch": 0.057416666666666664,
      "grad_norm": 32.25,
      "grad_norm_var": 3.66640625,
      "learning_rate": 0.0001,
      "loss": 7.229,
      "loss/crossentropy": 2.0895703792572022,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18702018074691296,
      "step": 6890
    },
    {
      "epoch": 0.0575,
      "grad_norm": 31.5,
      "grad_norm_var": 2.1205729166666667,
      "learning_rate": 0.0001,
      "loss": 7.1657,
      "loss/crossentropy": 2.014855782687664,
      "loss/hidden": 3.44296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19445941504091024,
      "step": 6900
    },
    {
      "epoch": 0.057583333333333334,
      "grad_norm": 35.25,
      "grad_norm_var": 3.93125,
      "learning_rate": 0.0001,
      "loss": 7.2919,
      "loss/crossentropy": 1.94414019882679,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1919761449098587,
      "step": 6910
    },
    {
      "epoch": 0.057666666666666665,
      "grad_norm": 33.75,
      "grad_norm_var": 4.228580729166667,
      "learning_rate": 0.0001,
      "loss": 7.3015,
      "loss/crossentropy": 1.9129973486065865,
      "loss/hidden": 3.339453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1791930379346013,
      "step": 6920
    },
    {
      "epoch": 0.05775,
      "grad_norm": 29.875,
      "grad_norm_var": 5.5739081030638426e+17,
      "learning_rate": 0.0001,
      "loss": 7.2774,
      "loss/crossentropy": 2.124286249279976,
      "loss/hidden": 3.36484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18690722845494748,
      "step": 6930
    },
    {
      "epoch": 0.057833333333333334,
      "grad_norm": 31.125,
      "grad_norm_var": 5.784375,
      "learning_rate": 0.0001,
      "loss": 7.1534,
      "loss/crossentropy": 1.817179811000824,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18679010756313802,
      "step": 6940
    },
    {
      "epoch": 0.057916666666666665,
      "grad_norm": 30.125,
      "grad_norm_var": 238.24557291666667,
      "learning_rate": 0.0001,
      "loss": 7.2343,
      "loss/crossentropy": 2.0582338005304335,
      "loss/hidden": 3.34140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17898097299039364,
      "step": 6950
    },
    {
      "epoch": 0.058,
      "grad_norm": 30.5,
      "grad_norm_var": 11.303580729166667,
      "learning_rate": 0.0001,
      "loss": 7.2738,
      "loss/crossentropy": 2.0119086511433126,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1978511307388544,
      "step": 6960
    },
    {
      "epoch": 0.058083333333333334,
      "grad_norm": 29.375,
      "grad_norm_var": 2.4184895833333333,
      "learning_rate": 0.0001,
      "loss": 7.2317,
      "loss/crossentropy": 1.9403888657689095,
      "loss/hidden": 3.399609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19264563638716936,
      "step": 6970
    },
    {
      "epoch": 0.058166666666666665,
      "grad_norm": 35.75,
      "grad_norm_var": 2.44140625,
      "learning_rate": 0.0001,
      "loss": 7.2749,
      "loss/crossentropy": 1.9770619094371795,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889339266344905,
      "step": 6980
    },
    {
      "epoch": 0.05825,
      "grad_norm": 28.875,
      "grad_norm_var": 14.030989583333334,
      "learning_rate": 0.0001,
      "loss": 7.2932,
      "loss/crossentropy": 1.8545608639717102,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1864480933174491,
      "step": 6990
    },
    {
      "epoch": 0.058333333333333334,
      "grad_norm": 30.0,
      "grad_norm_var": 3.9218098958333334,
      "learning_rate": 0.0001,
      "loss": 7.0992,
      "loss/crossentropy": 1.830371941626072,
      "loss/hidden": 3.34375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17968840897083282,
      "step": 7000
    },
    {
      "epoch": 0.058416666666666665,
      "grad_norm": 28.75,
      "grad_norm_var": 11.968684895833333,
      "learning_rate": 0.0001,
      "loss": 7.2045,
      "loss/crossentropy": 1.9251724004745483,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17747993357479572,
      "step": 7010
    },
    {
      "epoch": 0.0585,
      "grad_norm": 30.125,
      "grad_norm_var": 12.1087890625,
      "learning_rate": 0.0001,
      "loss": 7.1432,
      "loss/crossentropy": 1.8945713669061661,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1888335794210434,
      "step": 7020
    },
    {
      "epoch": 0.058583333333333334,
      "grad_norm": 31.375,
      "grad_norm_var": 8.276822916666667,
      "learning_rate": 0.0001,
      "loss": 7.1716,
      "loss/crossentropy": 1.9128420159220696,
      "loss/hidden": 3.19453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1727126305922866,
      "step": 7030
    },
    {
      "epoch": 0.058666666666666666,
      "grad_norm": 33.5,
      "grad_norm_var": 17.006705729166665,
      "learning_rate": 0.0001,
      "loss": 7.1884,
      "loss/crossentropy": 1.991888478398323,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19327655341476202,
      "step": 7040
    },
    {
      "epoch": 0.05875,
      "grad_norm": 42.25,
      "grad_norm_var": 12.85390625,
      "learning_rate": 0.0001,
      "loss": 7.2039,
      "loss/crossentropy": 1.845263669639826,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19932820042595267,
      "step": 7050
    },
    {
      "epoch": 0.058833333333333335,
      "grad_norm": 32.0,
      "grad_norm_var": 5221.237434895834,
      "learning_rate": 0.0001,
      "loss": 7.3444,
      "loss/crossentropy": 1.9406698502600193,
      "loss/hidden": 3.4875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18545874748378993,
      "step": 7060
    },
    {
      "epoch": 0.058916666666666666,
      "grad_norm": 30.0,
      "grad_norm_var": 21.25390625,
      "learning_rate": 0.0001,
      "loss": 7.3752,
      "loss/crossentropy": 2.0439330741763113,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20598556995391845,
      "step": 7070
    },
    {
      "epoch": 0.059,
      "grad_norm": 31.25,
      "grad_norm_var": 13.151497395833333,
      "learning_rate": 0.0001,
      "loss": 7.1997,
      "loss/crossentropy": 2.000889018177986,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18128069918602704,
      "step": 7080
    },
    {
      "epoch": 0.059083333333333335,
      "grad_norm": 29.25,
      "grad_norm_var": 13.904166666666667,
      "learning_rate": 0.0001,
      "loss": 7.3386,
      "loss/crossentropy": 1.9829541355371476,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19588420819491148,
      "step": 7090
    },
    {
      "epoch": 0.059166666666666666,
      "grad_norm": 29.5,
      "grad_norm_var": 6.495572916666666,
      "learning_rate": 0.0001,
      "loss": 7.2047,
      "loss/crossentropy": 1.88180633187294,
      "loss/hidden": 3.605078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20174703951925038,
      "step": 7100
    },
    {
      "epoch": 0.05925,
      "grad_norm": 32.5,
      "grad_norm_var": 61.45390625,
      "learning_rate": 0.0001,
      "loss": 7.2763,
      "loss/crossentropy": 1.9542135357856751,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21102935615926982,
      "step": 7110
    },
    {
      "epoch": 0.059333333333333335,
      "grad_norm": 32.0,
      "grad_norm_var": 71.41458333333334,
      "learning_rate": 0.0001,
      "loss": 7.1667,
      "loss/crossentropy": 1.926981085538864,
      "loss/hidden": 3.461328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17952323369681836,
      "step": 7120
    },
    {
      "epoch": 0.059416666666666666,
      "grad_norm": 35.5,
      "grad_norm_var": 38.1884765625,
      "learning_rate": 0.0001,
      "loss": 7.3224,
      "loss/crossentropy": 1.9933403447270392,
      "loss/hidden": 3.466796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21018762681633235,
      "step": 7130
    },
    {
      "epoch": 0.0595,
      "grad_norm": 34.25,
      "grad_norm_var": 23.698893229166668,
      "learning_rate": 0.0001,
      "loss": 7.2229,
      "loss/crossentropy": 1.8195350259542464,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1856760375201702,
      "step": 7140
    },
    {
      "epoch": 0.059583333333333335,
      "grad_norm": 32.5,
      "grad_norm_var": 11.039583333333333,
      "learning_rate": 0.0001,
      "loss": 7.1395,
      "loss/crossentropy": 2.0024550408124924,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889307852834463,
      "step": 7150
    },
    {
      "epoch": 0.059666666666666666,
      "grad_norm": 28.5,
      "grad_norm_var": 7.2103515625,
      "learning_rate": 0.0001,
      "loss": 7.2664,
      "loss/crossentropy": 2.0668145686388018,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19147251602262258,
      "step": 7160
    },
    {
      "epoch": 0.05975,
      "grad_norm": 29.75,
      "grad_norm_var": 18.46015625,
      "learning_rate": 0.0001,
      "loss": 7.1747,
      "loss/crossentropy": 2.0236642003059386,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18423411902040243,
      "step": 7170
    },
    {
      "epoch": 0.059833333333333336,
      "grad_norm": 31.625,
      "grad_norm_var": 15.653580729166666,
      "learning_rate": 0.0001,
      "loss": 7.1739,
      "loss/crossentropy": 2.0101078763604163,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18118085842579604,
      "step": 7180
    },
    {
      "epoch": 0.05991666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 25.6759765625,
      "learning_rate": 0.0001,
      "loss": 7.2501,
      "loss/crossentropy": 2.0987291634082794,
      "loss/hidden": 3.35859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19870839845389127,
      "step": 7190
    },
    {
      "epoch": 0.06,
      "grad_norm": 33.0,
      "grad_norm_var": 26.830143229166666,
      "learning_rate": 0.0001,
      "loss": 7.232,
      "loss/crossentropy": 2.0165331818163397,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20182917425408958,
      "step": 7200
    },
    {
      "epoch": 0.060083333333333336,
      "grad_norm": 30.625,
      "grad_norm_var": 3.4629557291666666,
      "learning_rate": 0.0001,
      "loss": 7.2123,
      "loss/crossentropy": 1.9919808238744736,
      "loss/hidden": 3.303515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1857657887041569,
      "step": 7210
    },
    {
      "epoch": 0.06016666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 3.876822916666667,
      "learning_rate": 0.0001,
      "loss": 7.1908,
      "loss/crossentropy": 2.048959070444107,
      "loss/hidden": 3.290234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18348812833428382,
      "step": 7220
    },
    {
      "epoch": 0.06025,
      "grad_norm": 30.0,
      "grad_norm_var": 2.8577473958333335,
      "learning_rate": 0.0001,
      "loss": 7.2264,
      "loss/crossentropy": 1.9573696821928024,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1977718524634838,
      "step": 7230
    },
    {
      "epoch": 0.060333333333333336,
      "grad_norm": 32.0,
      "grad_norm_var": 3.6363932291666665,
      "learning_rate": 0.0001,
      "loss": 7.1928,
      "loss/crossentropy": 1.8912595480680465,
      "loss/hidden": 3.4921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1922957869246602,
      "step": 7240
    },
    {
      "epoch": 0.06041666666666667,
      "grad_norm": 31.125,
      "grad_norm_var": 3.995833333333333,
      "learning_rate": 0.0001,
      "loss": 7.2126,
      "loss/crossentropy": 1.852707712352276,
      "loss/hidden": 3.3578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17833833750337363,
      "step": 7250
    },
    {
      "epoch": 0.0605,
      "grad_norm": 27.375,
      "grad_norm_var": 36.7791015625,
      "learning_rate": 0.0001,
      "loss": 7.2043,
      "loss/crossentropy": 2.023664565384388,
      "loss/hidden": 3.30078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.195465612411499,
      "step": 7260
    },
    {
      "epoch": 0.060583333333333336,
      "grad_norm": 30.375,
      "grad_norm_var": 42.77708333333333,
      "learning_rate": 0.0001,
      "loss": 7.2333,
      "loss/crossentropy": 1.9334395371377469,
      "loss/hidden": 3.3234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18526536021381618,
      "step": 7270
    },
    {
      "epoch": 0.06066666666666667,
      "grad_norm": 34.25,
      "grad_norm_var": 652.9705729166667,
      "learning_rate": 0.0001,
      "loss": 7.3917,
      "loss/crossentropy": 2.0037595987319947,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1814060116186738,
      "step": 7280
    },
    {
      "epoch": 0.06075,
      "grad_norm": 34.5,
      "grad_norm_var": 5.891447032789868e+17,
      "learning_rate": 0.0001,
      "loss": 7.2725,
      "loss/crossentropy": 2.023529142141342,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1940064739435911,
      "step": 7290
    },
    {
      "epoch": 0.060833333333333336,
      "grad_norm": 33.5,
      "grad_norm_var": 3.2143229166666667,
      "learning_rate": 0.0001,
      "loss": 7.2438,
      "loss/crossentropy": 1.9430765897035598,
      "loss/hidden": 3.51640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1979274285957217,
      "step": 7300
    },
    {
      "epoch": 0.06091666666666667,
      "grad_norm": 32.5,
      "grad_norm_var": 4.655989583333334,
      "learning_rate": 0.0001,
      "loss": 7.1017,
      "loss/crossentropy": 1.8648948967456818,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17332021631300448,
      "step": 7310
    },
    {
      "epoch": 0.061,
      "grad_norm": 29.75,
      "grad_norm_var": 1.9671223958333333,
      "learning_rate": 0.0001,
      "loss": 7.1943,
      "loss/crossentropy": 1.953578931093216,
      "loss/hidden": 3.526953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22345543652772903,
      "step": 7320
    },
    {
      "epoch": 0.06108333333333334,
      "grad_norm": 31.5,
      "grad_norm_var": 2.653059895833333,
      "learning_rate": 0.0001,
      "loss": 7.2887,
      "loss/crossentropy": 1.9954859048128128,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19549989588558675,
      "step": 7330
    },
    {
      "epoch": 0.06116666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 2.7104166666666667,
      "learning_rate": 0.0001,
      "loss": 7.325,
      "loss/crossentropy": 2.1271569937467576,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1951669655740261,
      "step": 7340
    },
    {
      "epoch": 0.06125,
      "grad_norm": 30.75,
      "grad_norm_var": 1.9317057291666666,
      "learning_rate": 0.0001,
      "loss": 7.2158,
      "loss/crossentropy": 1.8807506158947944,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18400093652307986,
      "step": 7350
    },
    {
      "epoch": 0.06133333333333333,
      "grad_norm": 35.75,
      "grad_norm_var": 10.7962890625,
      "learning_rate": 0.0001,
      "loss": 7.3313,
      "loss/crossentropy": 1.98387304097414,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1802327126264572,
      "step": 7360
    },
    {
      "epoch": 0.06141666666666667,
      "grad_norm": 31.375,
      "grad_norm_var": 3.9666015625,
      "learning_rate": 0.0001,
      "loss": 7.2317,
      "loss/crossentropy": 1.8400464594364165,
      "loss/hidden": 3.471875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2012029906734824,
      "step": 7370
    },
    {
      "epoch": 0.0615,
      "grad_norm": 34.0,
      "grad_norm_var": 4.478125,
      "learning_rate": 0.0001,
      "loss": 7.336,
      "loss/crossentropy": 1.9901346012949943,
      "loss/hidden": 3.492578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21220925878733396,
      "step": 7380
    },
    {
      "epoch": 0.06158333333333333,
      "grad_norm": 31.375,
      "grad_norm_var": 4.945833333333334,
      "learning_rate": 0.0001,
      "loss": 7.2569,
      "loss/crossentropy": 2.025828644633293,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1852064810693264,
      "step": 7390
    },
    {
      "epoch": 0.06166666666666667,
      "grad_norm": 29.75,
      "grad_norm_var": 11.0212890625,
      "learning_rate": 0.0001,
      "loss": 7.1639,
      "loss/crossentropy": 1.931120578944683,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19699512235820293,
      "step": 7400
    },
    {
      "epoch": 0.06175,
      "grad_norm": 31.5,
      "grad_norm_var": 1.015625,
      "learning_rate": 0.0001,
      "loss": 7.2259,
      "loss/crossentropy": 1.9323313042521477,
      "loss/hidden": 3.54140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18886004630476236,
      "step": 7410
    },
    {
      "epoch": 0.06183333333333333,
      "grad_norm": 28.625,
      "grad_norm_var": 6.25390625,
      "learning_rate": 0.0001,
      "loss": 7.2594,
      "loss/crossentropy": 1.9984138324856757,
      "loss/hidden": 3.42734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.207680026255548,
      "step": 7420
    },
    {
      "epoch": 0.06191666666666667,
      "grad_norm": 39.0,
      "grad_norm_var": 7.705989583333333,
      "learning_rate": 0.0001,
      "loss": 7.1694,
      "loss/crossentropy": 1.9141117088496684,
      "loss/hidden": 3.396484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18680279217660428,
      "step": 7430
    },
    {
      "epoch": 0.062,
      "grad_norm": 32.25,
      "grad_norm_var": 31.816666666666666,
      "learning_rate": 0.0001,
      "loss": 7.227,
      "loss/crossentropy": 1.915043169260025,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18126301970332862,
      "step": 7440
    },
    {
      "epoch": 0.06208333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 25.651041666666668,
      "learning_rate": 0.0001,
      "loss": 7.2287,
      "loss/crossentropy": 1.9978440970182418,
      "loss/hidden": 3.35703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19587596897035836,
      "step": 7450
    },
    {
      "epoch": 0.06216666666666667,
      "grad_norm": 30.375,
      "grad_norm_var": 6.303580729166667,
      "learning_rate": 0.0001,
      "loss": 7.1764,
      "loss/crossentropy": 2.01900063380599,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18093273509293795,
      "step": 7460
    },
    {
      "epoch": 0.06225,
      "grad_norm": 31.25,
      "grad_norm_var": 5.61640625,
      "learning_rate": 0.0001,
      "loss": 7.1635,
      "loss/crossentropy": 1.9081674709916114,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18367233984172343,
      "step": 7470
    },
    {
      "epoch": 0.06233333333333333,
      "grad_norm": 33.5,
      "grad_norm_var": 2.0056640625,
      "learning_rate": 0.0001,
      "loss": 7.1685,
      "loss/crossentropy": 1.9261738657951355,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18073464091867208,
      "step": 7480
    },
    {
      "epoch": 0.06241666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 99.3087890625,
      "learning_rate": 0.0001,
      "loss": 7.1488,
      "loss/crossentropy": 1.9166977405548096,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18761946000158786,
      "step": 7490
    },
    {
      "epoch": 0.0625,
      "grad_norm": 34.0,
      "grad_norm_var": 3.8020833333333335,
      "learning_rate": 0.0001,
      "loss": 7.2591,
      "loss/crossentropy": 1.902839483320713,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19028478916734456,
      "step": 7500
    },
    {
      "epoch": 0.06258333333333334,
      "grad_norm": 29.25,
      "grad_norm_var": 2.9697916666666666,
      "learning_rate": 0.0001,
      "loss": 7.1728,
      "loss/crossentropy": 1.6974174082279205,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1784145675599575,
      "step": 7510
    },
    {
      "epoch": 0.06266666666666666,
      "grad_norm": 34.75,
      "grad_norm_var": 3.8655598958333335,
      "learning_rate": 0.0001,
      "loss": 7.1342,
      "loss/crossentropy": 1.9428329408168792,
      "loss/hidden": 3.394140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18656713459640742,
      "step": 7520
    },
    {
      "epoch": 0.06275,
      "grad_norm": 31.5,
      "grad_norm_var": 3.423958333333333,
      "learning_rate": 0.0001,
      "loss": 7.2,
      "loss/crossentropy": 1.8696777828037738,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18415667656809093,
      "step": 7530
    },
    {
      "epoch": 0.06283333333333334,
      "grad_norm": 31.625,
      "grad_norm_var": 2.8379557291666666,
      "learning_rate": 0.0001,
      "loss": 7.2881,
      "loss/crossentropy": 1.9623609468340875,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18007479943335056,
      "step": 7540
    },
    {
      "epoch": 0.06291666666666666,
      "grad_norm": 2801795072.0,
      "grad_norm_var": 4.90628464974024e+17,
      "learning_rate": 0.0001,
      "loss": 7.2655,
      "loss/crossentropy": 1.9910573005676269,
      "loss/hidden": 3.6328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20073548033833505,
      "step": 7550
    },
    {
      "epoch": 0.063,
      "grad_norm": 34.5,
      "grad_norm_var": 4.9062846466465914e+17,
      "learning_rate": 0.0001,
      "loss": 7.2488,
      "loss/crossentropy": 1.8616838872432708,
      "loss/hidden": 3.50078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19058813713490963,
      "step": 7560
    },
    {
      "epoch": 0.06308333333333334,
      "grad_norm": 30.5,
      "grad_norm_var": 2.57265625,
      "learning_rate": 0.0001,
      "loss": 7.2364,
      "loss/crossentropy": 1.9160969465970994,
      "loss/hidden": 3.48203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20963120143860578,
      "step": 7570
    },
    {
      "epoch": 0.06316666666666666,
      "grad_norm": 31.875,
      "grad_norm_var": 2.820572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4052,
      "loss/crossentropy": 1.90859245210886,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20211333967745304,
      "step": 7580
    },
    {
      "epoch": 0.06325,
      "grad_norm": 30.5,
      "grad_norm_var": 4.809309895833334,
      "learning_rate": 0.0001,
      "loss": 7.2456,
      "loss/crossentropy": 1.828519704192877,
      "loss/hidden": 3.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19046486187726258,
      "step": 7590
    },
    {
      "epoch": 0.06333333333333334,
      "grad_norm": 32.0,
      "grad_norm_var": 5.226822916666666,
      "learning_rate": 0.0001,
      "loss": 7.1943,
      "loss/crossentropy": 2.030736616253853,
      "loss/hidden": 3.339453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21096080131828784,
      "step": 7600
    },
    {
      "epoch": 0.06341666666666666,
      "grad_norm": 36.0,
      "grad_norm_var": 4.0228515625,
      "learning_rate": 0.0001,
      "loss": 7.2782,
      "loss/crossentropy": 2.016854166984558,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18759209662675858,
      "step": 7610
    },
    {
      "epoch": 0.0635,
      "grad_norm": 31.375,
      "grad_norm_var": 6.439322916666667,
      "learning_rate": 0.0001,
      "loss": 7.1895,
      "loss/crossentropy": 1.918825089931488,
      "loss/hidden": 3.4375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19764372929930688,
      "step": 7620
    },
    {
      "epoch": 0.06358333333333334,
      "grad_norm": 29.375,
      "grad_norm_var": 5.9072265625,
      "learning_rate": 0.0001,
      "loss": 7.2783,
      "loss/crossentropy": 1.9146866127848625,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1974598715081811,
      "step": 7630
    },
    {
      "epoch": 0.06366666666666666,
      "grad_norm": 30.625,
      "grad_norm_var": 5.93125,
      "learning_rate": 0.0001,
      "loss": 7.2713,
      "loss/crossentropy": 1.9532035797834397,
      "loss/hidden": 3.4765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2067047307267785,
      "step": 7640
    },
    {
      "epoch": 0.06375,
      "grad_norm": 32.5,
      "grad_norm_var": 13.1259765625,
      "learning_rate": 0.0001,
      "loss": 7.1361,
      "loss/crossentropy": 1.824920044839382,
      "loss/hidden": 3.440234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19241070449352266,
      "step": 7650
    },
    {
      "epoch": 0.06383333333333334,
      "grad_norm": 33.25,
      "grad_norm_var": 10.99375,
      "learning_rate": 0.0001,
      "loss": 7.2802,
      "loss/crossentropy": 1.7821342311799526,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18938454277813435,
      "step": 7660
    },
    {
      "epoch": 0.06391666666666666,
      "grad_norm": 33.0,
      "grad_norm_var": 52.74557291666667,
      "learning_rate": 0.0001,
      "loss": 7.3821,
      "loss/crossentropy": 2.130902296304703,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2177656836807728,
      "step": 7670
    },
    {
      "epoch": 0.064,
      "grad_norm": 32.0,
      "grad_norm_var": 54.44895833333333,
      "learning_rate": 0.0001,
      "loss": 7.1582,
      "loss/crossentropy": 1.9051862180233001,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17711994536221026,
      "step": 7680
    },
    {
      "epoch": 0.06408333333333334,
      "grad_norm": 32.5,
      "grad_norm_var": 26.953580729166667,
      "learning_rate": 0.0001,
      "loss": 7.3007,
      "loss/crossentropy": 1.997354769706726,
      "loss/hidden": 3.52265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18910952042788268,
      "step": 7690
    },
    {
      "epoch": 0.06416666666666666,
      "grad_norm": 31.5,
      "grad_norm_var": 9.301041666666666,
      "learning_rate": 0.0001,
      "loss": 7.3028,
      "loss/crossentropy": 1.832273904979229,
      "loss/hidden": 3.504296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21006521843373777,
      "step": 7700
    },
    {
      "epoch": 0.06425,
      "grad_norm": 33.25,
      "grad_norm_var": 3.3291666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3769,
      "loss/crossentropy": 2.0843783155083657,
      "loss/hidden": 3.465234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20692231114953757,
      "step": 7710
    },
    {
      "epoch": 0.06433333333333334,
      "grad_norm": 31.75,
      "grad_norm_var": 1.6322265625,
      "learning_rate": 0.0001,
      "loss": 7.2327,
      "loss/crossentropy": 2.079830232262611,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18693204522132872,
      "step": 7720
    },
    {
      "epoch": 0.06441666666666666,
      "grad_norm": 31.625,
      "grad_norm_var": 1.6650390625,
      "learning_rate": 0.0001,
      "loss": 7.2392,
      "loss/crossentropy": 1.9599878072738648,
      "loss/hidden": 3.264453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17181740291416644,
      "step": 7730
    },
    {
      "epoch": 0.0645,
      "grad_norm": 32.5,
      "grad_norm_var": 2.8177083333333335,
      "learning_rate": 0.0001,
      "loss": 7.2603,
      "loss/crossentropy": 1.9764328211545945,
      "loss/hidden": 3.511328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19167867433279753,
      "step": 7740
    },
    {
      "epoch": 0.06458333333333334,
      "grad_norm": 31.25,
      "grad_norm_var": 1.98125,
      "learning_rate": 0.0001,
      "loss": 7.2676,
      "loss/crossentropy": 1.9611605733633042,
      "loss/hidden": 3.50546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19947379268705845,
      "step": 7750
    },
    {
      "epoch": 0.06466666666666666,
      "grad_norm": 32.5,
      "grad_norm_var": 1.4520833333333334,
      "learning_rate": 0.0001,
      "loss": 7.2259,
      "loss/crossentropy": 2.044002892076969,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19694972168654204,
      "step": 7760
    },
    {
      "epoch": 0.06475,
      "grad_norm": 31.125,
      "grad_norm_var": 2.122330729166667,
      "learning_rate": 0.0001,
      "loss": 7.1708,
      "loss/crossentropy": 2.0414538815617562,
      "loss/hidden": 3.237109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17013669721782207,
      "step": 7770
    },
    {
      "epoch": 0.06483333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 3.2395833333333335,
      "learning_rate": 0.0001,
      "loss": 7.3133,
      "loss/crossentropy": 1.910868799686432,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19090145360678434,
      "step": 7780
    },
    {
      "epoch": 0.06491666666666666,
      "grad_norm": 32.25,
      "grad_norm_var": 0.9747395833333333,
      "learning_rate": 0.0001,
      "loss": 7.2692,
      "loss/crossentropy": 2.0562462225556373,
      "loss/hidden": 3.28203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18266476448625327,
      "step": 7790
    },
    {
      "epoch": 0.065,
      "grad_norm": 31.875,
      "grad_norm_var": 2.957291666666667,
      "learning_rate": 0.0001,
      "loss": 7.3863,
      "loss/crossentropy": 1.9470960587263106,
      "loss/hidden": 3.483203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22827607877552508,
      "step": 7800
    },
    {
      "epoch": 0.06508333333333334,
      "grad_norm": 38.5,
      "grad_norm_var": 13.29765625,
      "learning_rate": 0.0001,
      "loss": 7.2335,
      "loss/crossentropy": 1.9175234675407409,
      "loss/hidden": 3.426953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19259056225419044,
      "step": 7810
    },
    {
      "epoch": 0.06516666666666666,
      "grad_norm": 30.625,
      "grad_norm_var": 12.7603515625,
      "learning_rate": 0.0001,
      "loss": 7.2009,
      "loss/crossentropy": 1.9584806606173515,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21137978304177524,
      "step": 7820
    },
    {
      "epoch": 0.06525,
      "grad_norm": 33.0,
      "grad_norm_var": 1.91875,
      "learning_rate": 0.0001,
      "loss": 7.2006,
      "loss/crossentropy": 1.9486504584550857,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18704792261123657,
      "step": 7830
    },
    {
      "epoch": 0.06533333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 2.18125,
      "learning_rate": 0.0001,
      "loss": 7.2527,
      "loss/crossentropy": 1.9827014431357384,
      "loss/hidden": 3.440625,
      "loss/jsd": 0.0,
      "loss/logits": 0.188728235848248,
      "step": 7840
    },
    {
      "epoch": 0.06541666666666666,
      "grad_norm": 35.25,
      "grad_norm_var": 86.52682291666666,
      "learning_rate": 0.0001,
      "loss": 7.39,
      "loss/crossentropy": 2.0009460031986235,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1901438269764185,
      "step": 7850
    },
    {
      "epoch": 0.0655,
      "grad_norm": 33.5,
      "grad_norm_var": 82.5634765625,
      "learning_rate": 0.0001,
      "loss": 7.282,
      "loss/crossentropy": 2.0347262173891068,
      "loss/hidden": 3.28359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1961182001978159,
      "step": 7860
    },
    {
      "epoch": 0.06558333333333333,
      "grad_norm": 31.75,
      "grad_norm_var": 3.896875,
      "learning_rate": 0.0001,
      "loss": 7.1761,
      "loss/crossentropy": 1.9007416464388371,
      "loss/hidden": 3.43203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19023145977407693,
      "step": 7870
    },
    {
      "epoch": 0.06566666666666666,
      "grad_norm": 33.0,
      "grad_norm_var": 4.138541666666667,
      "learning_rate": 0.0001,
      "loss": 7.2198,
      "loss/crossentropy": 2.1901776075363157,
      "loss/hidden": 3.4015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21192686930298804,
      "step": 7880
    },
    {
      "epoch": 0.06575,
      "grad_norm": 31.875,
      "grad_norm_var": 5.09375,
      "learning_rate": 0.0001,
      "loss": 7.1888,
      "loss/crossentropy": 1.9310210451483727,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.176263870857656,
      "step": 7890
    },
    {
      "epoch": 0.06583333333333333,
      "grad_norm": 32.25,
      "grad_norm_var": 9.770833333333334,
      "learning_rate": 0.0001,
      "loss": 7.2021,
      "loss/crossentropy": 1.9472015351057053,
      "loss/hidden": 3.2953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1718252243474126,
      "step": 7900
    },
    {
      "epoch": 0.06591666666666667,
      "grad_norm": 31.0,
      "grad_norm_var": 11.4072265625,
      "learning_rate": 0.0001,
      "loss": 7.1642,
      "loss/crossentropy": 2.060355693101883,
      "loss/hidden": 3.2875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17545824572443963,
      "step": 7910
    },
    {
      "epoch": 0.066,
      "grad_norm": 30.75,
      "grad_norm_var": 1.0473683717269925e+18,
      "learning_rate": 0.0001,
      "loss": 7.1745,
      "loss/crossentropy": 1.7840346470475197,
      "loss/hidden": 3.740625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19948329012840987,
      "step": 7920
    },
    {
      "epoch": 0.06608333333333333,
      "grad_norm": 33.0,
      "grad_norm_var": 1.0473683711683812e+18,
      "learning_rate": 0.0001,
      "loss": 7.3275,
      "loss/crossentropy": 1.9387343034148217,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1954814039170742,
      "step": 7930
    },
    {
      "epoch": 0.06616666666666667,
      "grad_norm": 36.25,
      "grad_norm_var": 9.870572916666667,
      "learning_rate": 0.0001,
      "loss": 7.1739,
      "loss/crossentropy": 1.8542084500193596,
      "loss/hidden": 3.39140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18380195926874876,
      "step": 7940
    },
    {
      "epoch": 0.06625,
      "grad_norm": 31.375,
      "grad_norm_var": 3.309309895833333,
      "learning_rate": 0.0001,
      "loss": 7.2471,
      "loss/crossentropy": 1.891501422226429,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21512426622211933,
      "step": 7950
    },
    {
      "epoch": 0.06633333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 3.7405598958333335,
      "learning_rate": 0.0001,
      "loss": 7.1933,
      "loss/crossentropy": 1.8226936116814614,
      "loss/hidden": 3.441796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1818669240921736,
      "step": 7960
    },
    {
      "epoch": 0.06641666666666667,
      "grad_norm": 37.25,
      "grad_norm_var": 199.25625,
      "learning_rate": 0.0001,
      "loss": 7.2723,
      "loss/crossentropy": 1.8292644947767258,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19608462285250425,
      "step": 7970
    },
    {
      "epoch": 0.0665,
      "grad_norm": 32.75,
      "grad_norm_var": 204.10149739583332,
      "learning_rate": 0.0001,
      "loss": 7.2325,
      "loss/crossentropy": 1.9156912505626678,
      "loss/hidden": 3.544140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19286852981895208,
      "step": 7980
    },
    {
      "epoch": 0.06658333333333333,
      "grad_norm": 31.75,
      "grad_norm_var": 28.8447265625,
      "learning_rate": 0.0001,
      "loss": 7.1774,
      "loss/crossentropy": 1.8666133657097816,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17876000609248877,
      "step": 7990
    },
    {
      "epoch": 0.06666666666666667,
      "grad_norm": 31.125,
      "grad_norm_var": 2.275,
      "learning_rate": 0.0001,
      "loss": 7.2216,
      "loss/crossentropy": 1.841892847418785,
      "loss/hidden": 3.421484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1745407897979021,
      "step": 8000
    },
    {
      "epoch": 0.06675,
      "grad_norm": 28.875,
      "grad_norm_var": 2.4702473958333333,
      "learning_rate": 0.0001,
      "loss": 7.2601,
      "loss/crossentropy": 1.8827014505863189,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20147160943597556,
      "step": 8010
    },
    {
      "epoch": 0.06683333333333333,
      "grad_norm": 34.0,
      "grad_norm_var": 5.549934895833333,
      "learning_rate": 0.0001,
      "loss": 7.3408,
      "loss/crossentropy": 1.9311093546450138,
      "loss/hidden": 3.568359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22386483792215586,
      "step": 8020
    },
    {
      "epoch": 0.06691666666666667,
      "grad_norm": 33.0,
      "grad_norm_var": 2.2541015625,
      "learning_rate": 0.0001,
      "loss": 7.1497,
      "loss/crossentropy": 1.8826589100062847,
      "loss/hidden": 3.29921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17312360629439355,
      "step": 8030
    },
    {
      "epoch": 0.067,
      "grad_norm": 28.875,
      "grad_norm_var": 4.233268229166667,
      "learning_rate": 0.0001,
      "loss": 7.2066,
      "loss/crossentropy": 1.9970761120319367,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19054881893098355,
      "step": 8040
    },
    {
      "epoch": 0.06708333333333333,
      "grad_norm": 30.0,
      "grad_norm_var": 8652.215559895832,
      "learning_rate": 0.0001,
      "loss": 7.2576,
      "loss/crossentropy": 1.9152855359017849,
      "loss/hidden": 3.455078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18708332600072025,
      "step": 8050
    },
    {
      "epoch": 0.06716666666666667,
      "grad_norm": 28.5,
      "grad_norm_var": 5.430208333333334,
      "learning_rate": 0.0001,
      "loss": 7.2357,
      "loss/crossentropy": 2.031227025389671,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.193635768070817,
      "step": 8060
    },
    {
      "epoch": 0.06725,
      "grad_norm": 33.5,
      "grad_norm_var": 4.4931640625,
      "learning_rate": 0.0001,
      "loss": 7.1251,
      "loss/crossentropy": 1.8489420399069787,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17814766112715005,
      "step": 8070
    },
    {
      "epoch": 0.06733333333333333,
      "grad_norm": 31.25,
      "grad_norm_var": 27.78515625,
      "learning_rate": 0.0001,
      "loss": 7.2487,
      "loss/crossentropy": 1.884847517311573,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2042497692629695,
      "step": 8080
    },
    {
      "epoch": 0.06741666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 4.5035995046731245e+17,
      "learning_rate": 0.0001,
      "loss": 7.1985,
      "loss/crossentropy": 1.8161820538342,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17485685739666224,
      "step": 8090
    },
    {
      "epoch": 0.0675,
      "grad_norm": 42.5,
      "grad_norm_var": 19.139583333333334,
      "learning_rate": 0.0001,
      "loss": 7.1604,
      "loss/crossentropy": 1.9721700698137283,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1950763339176774,
      "step": 8100
    },
    {
      "epoch": 0.06758333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 14.6416015625,
      "learning_rate": 0.0001,
      "loss": 7.2636,
      "loss/crossentropy": 2.0180334128439426,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20730137974023818,
      "step": 8110
    },
    {
      "epoch": 0.06766666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 22.159309895833335,
      "learning_rate": 0.0001,
      "loss": 7.2743,
      "loss/crossentropy": 2.1068934723734856,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1977040534839034,
      "step": 8120
    },
    {
      "epoch": 0.06775,
      "grad_norm": 30.875,
      "grad_norm_var": 445.7035807291667,
      "learning_rate": 0.0001,
      "loss": 7.2164,
      "loss/crossentropy": 1.9630880862474442,
      "loss/hidden": 3.396484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19089617524296046,
      "step": 8130
    },
    {
      "epoch": 0.06783333333333333,
      "grad_norm": 31.875,
      "grad_norm_var": 2.3145182291666666,
      "learning_rate": 0.0001,
      "loss": 7.2163,
      "loss/crossentropy": 1.8661000967025756,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19853502176702023,
      "step": 8140
    },
    {
      "epoch": 0.06791666666666667,
      "grad_norm": 31.125,
      "grad_norm_var": 3.31015625,
      "learning_rate": 0.0001,
      "loss": 7.1223,
      "loss/crossentropy": 1.9666982188820838,
      "loss/hidden": 3.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1820853654295206,
      "step": 8150
    },
    {
      "epoch": 0.068,
      "grad_norm": 35.0,
      "grad_norm_var": 29.7681640625,
      "learning_rate": 0.0001,
      "loss": 7.2284,
      "loss/crossentropy": 2.037832336127758,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19564964435994625,
      "step": 8160
    },
    {
      "epoch": 0.06808333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 27.930208333333333,
      "learning_rate": 0.0001,
      "loss": 7.1623,
      "loss/crossentropy": 1.8319560870528222,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18993047419935466,
      "step": 8170
    },
    {
      "epoch": 0.06816666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 2.486458333333333,
      "learning_rate": 0.0001,
      "loss": 7.1666,
      "loss/crossentropy": 1.9121002182364464,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18671723157167436,
      "step": 8180
    },
    {
      "epoch": 0.06825,
      "grad_norm": 30.125,
      "grad_norm_var": 2.5837890625,
      "learning_rate": 0.0001,
      "loss": 7.1998,
      "loss/crossentropy": 1.9544257938861846,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18110576160252095,
      "step": 8190
    },
    {
      "epoch": 0.06833333333333333,
      "grad_norm": 30.125,
      "grad_norm_var": 1.4129557291666666,
      "learning_rate": 0.0001,
      "loss": 7.1602,
      "loss/crossentropy": 1.8966913163661956,
      "loss/hidden": 3.325390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17959749344736337,
      "step": 8200
    },
    {
      "epoch": 0.06841666666666667,
      "grad_norm": 28.875,
      "grad_norm_var": 2.2067057291666665,
      "learning_rate": 0.0001,
      "loss": 7.1619,
      "loss/crossentropy": 1.9131664127111434,
      "loss/hidden": 3.354296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17485106661915778,
      "step": 8210
    },
    {
      "epoch": 0.0685,
      "grad_norm": 30.125,
      "grad_norm_var": 2.4936848958333333,
      "learning_rate": 0.0001,
      "loss": 7.2113,
      "loss/crossentropy": 2.0453411146998404,
      "loss/hidden": 3.2796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18515327256172895,
      "step": 8220
    },
    {
      "epoch": 0.06858333333333333,
      "grad_norm": 29.375,
      "grad_norm_var": 2.59375,
      "learning_rate": 0.0001,
      "loss": 7.2128,
      "loss/crossentropy": 1.9506754860281945,
      "loss/hidden": 3.322265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18393214661628007,
      "step": 8230
    },
    {
      "epoch": 0.06866666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 3.8,
      "learning_rate": 0.0001,
      "loss": 7.2449,
      "loss/crossentropy": 1.9423611342906952,
      "loss/hidden": 3.25078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17854795306921006,
      "step": 8240
    },
    {
      "epoch": 0.06875,
      "grad_norm": 28.875,
      "grad_norm_var": 3.6035807291666666,
      "learning_rate": 0.0001,
      "loss": 7.2422,
      "loss/crossentropy": 1.9720477700233459,
      "loss/hidden": 3.29765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.188589053414762,
      "step": 8250
    },
    {
      "epoch": 0.06883333333333333,
      "grad_norm": 123.0,
      "grad_norm_var": 535.6372395833333,
      "learning_rate": 0.0001,
      "loss": 7.1691,
      "loss/crossentropy": 1.99368005245924,
      "loss/hidden": 3.354296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1896060870960355,
      "step": 8260
    },
    {
      "epoch": 0.06891666666666667,
      "grad_norm": 28.625,
      "grad_norm_var": 521.7205729166667,
      "learning_rate": 0.0001,
      "loss": 7.2283,
      "loss/crossentropy": 1.896087047457695,
      "loss/hidden": 3.341796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1775512419641018,
      "step": 8270
    },
    {
      "epoch": 0.069,
      "grad_norm": 29.125,
      "grad_norm_var": 78.86399739583334,
      "learning_rate": 0.0001,
      "loss": 7.2136,
      "loss/crossentropy": 1.7727789789438249,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19561070688068866,
      "step": 8280
    },
    {
      "epoch": 0.06908333333333333,
      "grad_norm": 31.5,
      "grad_norm_var": 35.99583333333333,
      "learning_rate": 0.0001,
      "loss": 7.32,
      "loss/crossentropy": 2.039402049779892,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1993569979444146,
      "step": 8290
    },
    {
      "epoch": 0.06916666666666667,
      "grad_norm": 28.875,
      "grad_norm_var": 19.953125,
      "learning_rate": 0.0001,
      "loss": 7.1865,
      "loss/crossentropy": 1.8603923156857491,
      "loss/hidden": 3.55234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19390070009976626,
      "step": 8300
    },
    {
      "epoch": 0.06925,
      "grad_norm": 31.625,
      "grad_norm_var": 15.540625,
      "learning_rate": 0.0001,
      "loss": 7.1888,
      "loss/crossentropy": 1.828030915558338,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19069764837622644,
      "step": 8310
    },
    {
      "epoch": 0.06933333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 4.298893229166667,
      "learning_rate": 0.0001,
      "loss": 7.2229,
      "loss/crossentropy": 1.8442706629633903,
      "loss/hidden": 3.43125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1816394252702594,
      "step": 8320
    },
    {
      "epoch": 0.06941666666666667,
      "grad_norm": 32.0,
      "grad_norm_var": 3.35625,
      "learning_rate": 0.0001,
      "loss": 7.2461,
      "loss/crossentropy": 1.895699442923069,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19321332015097142,
      "step": 8330
    },
    {
      "epoch": 0.0695,
      "grad_norm": 40.0,
      "grad_norm_var": 36.78723958333333,
      "learning_rate": 0.0001,
      "loss": 7.1403,
      "loss/crossentropy": 1.9688566237688065,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19061133153736592,
      "step": 8340
    },
    {
      "epoch": 0.06958333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 39.82473958333333,
      "learning_rate": 0.0001,
      "loss": 7.1012,
      "loss/crossentropy": 1.9205253809690475,
      "loss/hidden": 3.401171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1715885553508997,
      "step": 8350
    },
    {
      "epoch": 0.06966666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 2.15625,
      "learning_rate": 0.0001,
      "loss": 7.2205,
      "loss/crossentropy": 1.949224701523781,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1786526907235384,
      "step": 8360
    },
    {
      "epoch": 0.06975,
      "grad_norm": 29.375,
      "grad_norm_var": 4.7025390625,
      "learning_rate": 0.0001,
      "loss": 7.1994,
      "loss/crossentropy": 1.9869706839323045,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1956912238150835,
      "step": 8370
    },
    {
      "epoch": 0.06983333333333333,
      "grad_norm": 34.25,
      "grad_norm_var": 19.805143229166667,
      "learning_rate": 0.0001,
      "loss": 7.2692,
      "loss/crossentropy": 1.9487977296113967,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18634006716310977,
      "step": 8380
    },
    {
      "epoch": 0.06991666666666667,
      "grad_norm": 32.75,
      "grad_norm_var": 2.615625,
      "learning_rate": 0.0001,
      "loss": 7.3005,
      "loss/crossentropy": 1.9617695778608322,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1997200610116124,
      "step": 8390
    },
    {
      "epoch": 0.07,
      "grad_norm": 32.0,
      "grad_norm_var": 2.7393229166666666,
      "learning_rate": 0.0001,
      "loss": 7.3869,
      "loss/crossentropy": 1.9588475212454797,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17612226950004697,
      "step": 8400
    },
    {
      "epoch": 0.07008333333333333,
      "grad_norm": 37.75,
      "grad_norm_var": 5.204472194874912e+17,
      "learning_rate": 0.0001,
      "loss": 7.2926,
      "loss/crossentropy": 2.057014559209347,
      "loss/hidden": 3.5015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19315509274601936,
      "step": 8410
    },
    {
      "epoch": 0.07016666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 5.2044721999248544e+17,
      "learning_rate": 0.0001,
      "loss": 7.2189,
      "loss/crossentropy": 1.9825481489300727,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18724329564720393,
      "step": 8420
    },
    {
      "epoch": 0.07025,
      "grad_norm": 31.25,
      "grad_norm_var": 2.370833333333333,
      "learning_rate": 0.0001,
      "loss": 7.1289,
      "loss/crossentropy": 1.8243992432951928,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1827236920595169,
      "step": 8430
    },
    {
      "epoch": 0.07033333333333333,
      "grad_norm": 30.625,
      "grad_norm_var": 2.078125,
      "learning_rate": 0.0001,
      "loss": 7.1837,
      "loss/crossentropy": 2.044503265619278,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17393332067877054,
      "step": 8440
    },
    {
      "epoch": 0.07041666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 3.126822916666667,
      "learning_rate": 0.0001,
      "loss": 7.2053,
      "loss/crossentropy": 1.915854911506176,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19963504755869507,
      "step": 8450
    },
    {
      "epoch": 0.0705,
      "grad_norm": 31.5,
      "grad_norm_var": 2.7983723958333333,
      "learning_rate": 0.0001,
      "loss": 7.2567,
      "loss/crossentropy": 1.9769734531641006,
      "loss/hidden": 3.295703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.181718045566231,
      "step": 8460
    },
    {
      "epoch": 0.07058333333333333,
      "grad_norm": 32.5,
      "grad_norm_var": 20.070572916666666,
      "learning_rate": 0.0001,
      "loss": 7.1527,
      "loss/crossentropy": 1.8116442456841468,
      "loss/hidden": 3.45546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18748198114335538,
      "step": 8470
    },
    {
      "epoch": 0.07066666666666667,
      "grad_norm": 31.5,
      "grad_norm_var": 28.205143229166666,
      "learning_rate": 0.0001,
      "loss": 7.2199,
      "loss/crossentropy": 1.9818651676177979,
      "loss/hidden": 3.523828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894064551219344,
      "step": 8480
    },
    {
      "epoch": 0.07075,
      "grad_norm": 34.25,
      "grad_norm_var": 1.1083252943282854e+18,
      "learning_rate": 0.0001,
      "loss": 7.2738,
      "loss/crossentropy": 1.9650639399886132,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17558318749070168,
      "step": 8490
    },
    {
      "epoch": 0.07083333333333333,
      "grad_norm": 34.0,
      "grad_norm_var": 9.544205729166666,
      "learning_rate": 0.0001,
      "loss": 7.2025,
      "loss/crossentropy": 1.9761282935738564,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19507634602487087,
      "step": 8500
    },
    {
      "epoch": 0.07091666666666667,
      "grad_norm": 35.25,
      "grad_norm_var": 3.1723307291666667,
      "learning_rate": 0.0001,
      "loss": 7.1301,
      "loss/crossentropy": 1.9544002994894982,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1756620042026043,
      "step": 8510
    },
    {
      "epoch": 0.071,
      "grad_norm": 30.625,
      "grad_norm_var": 2.7962890625,
      "learning_rate": 0.0001,
      "loss": 7.0789,
      "loss/crossentropy": 1.9353080570697785,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18456000369042158,
      "step": 8520
    },
    {
      "epoch": 0.07108333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 3.6405598958333334,
      "learning_rate": 0.0001,
      "loss": 7.1855,
      "loss/crossentropy": 2.0061560861766337,
      "loss/hidden": 3.321875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18838093373924494,
      "step": 8530
    },
    {
      "epoch": 0.07116666666666667,
      "grad_norm": 30.75,
      "grad_norm_var": 12.155989583333334,
      "learning_rate": 0.0001,
      "loss": 7.2221,
      "loss/crossentropy": 1.9317010313272476,
      "loss/hidden": 3.4140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18269629161804915,
      "step": 8540
    },
    {
      "epoch": 0.07125,
      "grad_norm": 30.875,
      "grad_norm_var": 8.755208333333334,
      "learning_rate": 0.0001,
      "loss": 7.1957,
      "loss/crossentropy": 1.9866686046123505,
      "loss/hidden": 3.3171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18688506465405225,
      "step": 8550
    },
    {
      "epoch": 0.07133333333333333,
      "grad_norm": 34.0,
      "grad_norm_var": 4.524739583333333,
      "learning_rate": 0.0001,
      "loss": 7.2597,
      "loss/crossentropy": 1.8756410598754882,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17877673637121916,
      "step": 8560
    },
    {
      "epoch": 0.07141666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 4.362955729166667,
      "learning_rate": 0.0001,
      "loss": 7.2339,
      "loss/crossentropy": 1.9563673168420792,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17372122146189212,
      "step": 8570
    },
    {
      "epoch": 0.0715,
      "grad_norm": 35.0,
      "grad_norm_var": 3.120247395833333,
      "learning_rate": 0.0001,
      "loss": 7.2685,
      "loss/crossentropy": 1.8956820949912072,
      "loss/hidden": 3.33046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17256692852824926,
      "step": 8580
    },
    {
      "epoch": 0.07158333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 12.751497395833333,
      "learning_rate": 0.0001,
      "loss": 7.1566,
      "loss/crossentropy": 1.9644134506583213,
      "loss/hidden": 3.31640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18599690701812505,
      "step": 8590
    },
    {
      "epoch": 0.07166666666666667,
      "grad_norm": 30.375,
      "grad_norm_var": 12.847330729166666,
      "learning_rate": 0.0001,
      "loss": 7.152,
      "loss/crossentropy": 1.9605856448411942,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17473262287676333,
      "step": 8600
    },
    {
      "epoch": 0.07175,
      "grad_norm": 30.625,
      "grad_norm_var": 2.9955729166666667,
      "learning_rate": 0.0001,
      "loss": 7.2577,
      "loss/crossentropy": 2.0416051924228666,
      "loss/hidden": 3.46953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19386923424899577,
      "step": 8610
    },
    {
      "epoch": 0.07183333333333333,
      "grad_norm": 46.25,
      "grad_norm_var": 16312.484309895834,
      "learning_rate": 0.0001,
      "loss": 7.2449,
      "loss/crossentropy": 1.8780235946178436,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18178340941667556,
      "step": 8620
    },
    {
      "epoch": 0.07191666666666667,
      "grad_norm": 30.25,
      "grad_norm_var": 16171.3572265625,
      "learning_rate": 0.0001,
      "loss": 7.2881,
      "loss/crossentropy": 2.017537733912468,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20194268729537726,
      "step": 8630
    },
    {
      "epoch": 0.072,
      "grad_norm": 29.5,
      "grad_norm_var": 1.9639973958333334,
      "learning_rate": 0.0001,
      "loss": 7.0493,
      "loss/crossentropy": 1.7677518472075462,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18404914420098067,
      "step": 8640
    },
    {
      "epoch": 0.07208333333333333,
      "grad_norm": 30.375,
      "grad_norm_var": 3.115625,
      "learning_rate": 0.0001,
      "loss": 7.2341,
      "loss/crossentropy": 1.81800634264946,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17132537085562943,
      "step": 8650
    },
    {
      "epoch": 0.07216666666666667,
      "grad_norm": 35.25,
      "grad_norm_var": 8.748418045451962e+17,
      "learning_rate": 0.0001,
      "loss": 7.2491,
      "loss/crossentropy": 1.9027793630957603,
      "loss/hidden": 3.291015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17584120333194733,
      "step": 8660
    },
    {
      "epoch": 0.07225,
      "grad_norm": 30.75,
      "grad_norm_var": 11.729166666666666,
      "learning_rate": 0.0001,
      "loss": 7.2757,
      "loss/crossentropy": 1.8777679741382598,
      "loss/hidden": 3.534375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19321729261428117,
      "step": 8670
    },
    {
      "epoch": 0.07233333333333333,
      "grad_norm": 39.75,
      "grad_norm_var": 6.401822916666666,
      "learning_rate": 0.0001,
      "loss": 7.1494,
      "loss/crossentropy": 1.9891476452350616,
      "loss/hidden": 3.312890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.188967870734632,
      "step": 8680
    },
    {
      "epoch": 0.07241666666666667,
      "grad_norm": 31.5,
      "grad_norm_var": 7.230208333333334,
      "learning_rate": 0.0001,
      "loss": 7.1312,
      "loss/crossentropy": 1.9339799106121063,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.178991923853755,
      "step": 8690
    },
    {
      "epoch": 0.0725,
      "grad_norm": 30.0,
      "grad_norm_var": 1.9478515625,
      "learning_rate": 0.0001,
      "loss": 7.3105,
      "loss/crossentropy": 2.0435035467147826,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2305173959583044,
      "step": 8700
    },
    {
      "epoch": 0.07258333333333333,
      "grad_norm": 30.875,
      "grad_norm_var": 4.911393229166666,
      "learning_rate": 0.0001,
      "loss": 7.3734,
      "loss/crossentropy": 2.069814404845238,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19525352716445923,
      "step": 8710
    },
    {
      "epoch": 0.07266666666666667,
      "grad_norm": 33.5,
      "grad_norm_var": 5.888997395833333,
      "learning_rate": 0.0001,
      "loss": 7.2247,
      "loss/crossentropy": 1.8153776928782464,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20906406342983247,
      "step": 8720
    },
    {
      "epoch": 0.07275,
      "grad_norm": 32.75,
      "grad_norm_var": 8.466666666666667,
      "learning_rate": 0.0001,
      "loss": 7.2225,
      "loss/crossentropy": 1.986364483833313,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18280372805893422,
      "step": 8730
    },
    {
      "epoch": 0.07283333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 3.3343098958333335,
      "learning_rate": 0.0001,
      "loss": 7.1908,
      "loss/crossentropy": 1.9479957491159439,
      "loss/hidden": 3.314453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17259119898080827,
      "step": 8740
    },
    {
      "epoch": 0.07291666666666667,
      "grad_norm": 33.0,
      "grad_norm_var": 3.1603515625,
      "learning_rate": 0.0001,
      "loss": 7.1168,
      "loss/crossentropy": 1.984895572811365,
      "loss/hidden": 3.251171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18064502384513617,
      "step": 8750
    },
    {
      "epoch": 0.073,
      "grad_norm": 31.625,
      "grad_norm_var": 2.8889973958333335,
      "learning_rate": 0.0001,
      "loss": 7.2268,
      "loss/crossentropy": 1.8017809867858887,
      "loss/hidden": 3.35546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17566005773842336,
      "step": 8760
    },
    {
      "epoch": 0.07308333333333333,
      "grad_norm": 29.5,
      "grad_norm_var": 4.761458333333334,
      "learning_rate": 0.0001,
      "loss": 7.1696,
      "loss/crossentropy": 1.900542539358139,
      "loss/hidden": 3.352734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17970035411417484,
      "step": 8770
    },
    {
      "epoch": 0.07316666666666667,
      "grad_norm": 33.5,
      "grad_norm_var": 4.3166015625,
      "learning_rate": 0.0001,
      "loss": 7.2412,
      "loss/crossentropy": 2.041232463717461,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20939856991171837,
      "step": 8780
    },
    {
      "epoch": 0.07325,
      "grad_norm": 34.0,
      "grad_norm_var": 3.689322916666667,
      "learning_rate": 0.0001,
      "loss": 7.2721,
      "loss/crossentropy": 2.005451811850071,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18856565970927477,
      "step": 8790
    },
    {
      "epoch": 0.07333333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 2.9098307291666665,
      "learning_rate": 0.0001,
      "loss": 7.27,
      "loss/crossentropy": 2.00799767524004,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1781530775129795,
      "step": 8800
    },
    {
      "epoch": 0.07341666666666667,
      "grad_norm": 35.25,
      "grad_norm_var": 2.780143229166667,
      "learning_rate": 0.0001,
      "loss": 7.2018,
      "loss/crossentropy": 2.035827913880348,
      "loss/hidden": 3.473828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20112563883885742,
      "step": 8810
    },
    {
      "epoch": 0.0735,
      "grad_norm": 29.875,
      "grad_norm_var": 4.736393229166667,
      "learning_rate": 0.0001,
      "loss": 7.0714,
      "loss/crossentropy": 1.9053474813699722,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1742972746491432,
      "step": 8820
    },
    {
      "epoch": 0.07358333333333333,
      "grad_norm": 32.0,
      "grad_norm_var": 2.2541015625,
      "learning_rate": 0.0001,
      "loss": 7.2318,
      "loss/crossentropy": 1.8243301972746848,
      "loss/hidden": 3.406640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1931868800893426,
      "step": 8830
    },
    {
      "epoch": 0.07366666666666667,
      "grad_norm": 31.75,
      "grad_norm_var": 3.14140625,
      "learning_rate": 0.0001,
      "loss": 7.2375,
      "loss/crossentropy": 2.120525896549225,
      "loss/hidden": 3.2109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1751797879114747,
      "step": 8840
    },
    {
      "epoch": 0.07375,
      "grad_norm": 29.75,
      "grad_norm_var": 2.7525390625,
      "learning_rate": 0.0001,
      "loss": 7.1501,
      "loss/crossentropy": 1.9116109699010848,
      "loss/hidden": 3.499609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20058850292116404,
      "step": 8850
    },
    {
      "epoch": 0.07383333333333333,
      "grad_norm": 31.375,
      "grad_norm_var": 2.690625,
      "learning_rate": 0.0001,
      "loss": 7.2157,
      "loss/crossentropy": 1.918264901638031,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18105561528354883,
      "step": 8860
    },
    {
      "epoch": 0.07391666666666667,
      "grad_norm": 29.75,
      "grad_norm_var": 2.4238932291666666,
      "learning_rate": 0.0001,
      "loss": 7.1731,
      "loss/crossentropy": 1.997542105615139,
      "loss/hidden": 3.196875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17583039812743664,
      "step": 8870
    },
    {
      "epoch": 0.074,
      "grad_norm": 30.0,
      "grad_norm_var": 1.2075980013739968e+18,
      "learning_rate": 0.0001,
      "loss": 7.2201,
      "loss/crossentropy": 2.035869190096855,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19501792285591363,
      "step": 8880
    },
    {
      "epoch": 0.07408333333333333,
      "grad_norm": 32.75,
      "grad_norm_var": 1.207598000929855e+18,
      "learning_rate": 0.0001,
      "loss": 7.1496,
      "loss/crossentropy": 1.7965922892093658,
      "loss/hidden": 3.341796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16304808165878057,
      "step": 8890
    },
    {
      "epoch": 0.07416666666666667,
      "grad_norm": 3909091328.0,
      "grad_norm_var": 9.550621728791244e+17,
      "learning_rate": 0.0001,
      "loss": 7.2495,
      "loss/crossentropy": 2.001436561346054,
      "loss/hidden": 3.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20285465717315673,
      "step": 8900
    },
    {
      "epoch": 0.07425,
      "grad_norm": 34.25,
      "grad_norm_var": 9.550621713969274e+17,
      "learning_rate": 0.0001,
      "loss": 7.2073,
      "loss/crossentropy": 2.0041428416967393,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20391638875007628,
      "step": 8910
    },
    {
      "epoch": 0.07433333333333333,
      "grad_norm": 31.375,
      "grad_norm_var": 11.8072265625,
      "learning_rate": 0.0001,
      "loss": 7.0817,
      "loss/crossentropy": 1.960794059932232,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17380897589027883,
      "step": 8920
    },
    {
      "epoch": 0.07441666666666667,
      "grad_norm": 32.25,
      "grad_norm_var": 2.9770833333333333,
      "learning_rate": 0.0001,
      "loss": 7.2189,
      "loss/crossentropy": 1.9762230925261974,
      "loss/hidden": 3.339453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17565663568675519,
      "step": 8930
    },
    {
      "epoch": 0.0745,
      "grad_norm": 34.25,
      "grad_norm_var": 3.035872395833333,
      "learning_rate": 0.0001,
      "loss": 7.171,
      "loss/crossentropy": 1.8802208930253983,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18836744036525488,
      "step": 8940
    },
    {
      "epoch": 0.07458333333333333,
      "grad_norm": 31.0,
      "grad_norm_var": 7.203580729166666,
      "learning_rate": 0.0001,
      "loss": 7.1634,
      "loss/crossentropy": 2.038408327102661,
      "loss/hidden": 3.265234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17698741499334575,
      "step": 8950
    },
    {
      "epoch": 0.07466666666666667,
      "grad_norm": 32.0,
      "grad_norm_var": 5.43515625,
      "learning_rate": 0.0001,
      "loss": 7.2006,
      "loss/crossentropy": 1.9461330994963646,
      "loss/hidden": 3.38671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18866215460002422,
      "step": 8960
    },
    {
      "epoch": 0.07475,
      "grad_norm": 33.75,
      "grad_norm_var": 1.9285807291666666,
      "learning_rate": 0.0001,
      "loss": 7.1538,
      "loss/crossentropy": 1.8740751318633557,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18583609201014042,
      "step": 8970
    },
    {
      "epoch": 0.07483333333333334,
      "grad_norm": 33.25,
      "grad_norm_var": 2.3988932291666667,
      "learning_rate": 0.0001,
      "loss": 7.1565,
      "loss/crossentropy": 1.8552331678569316,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.174785933829844,
      "step": 8980
    },
    {
      "epoch": 0.07491666666666667,
      "grad_norm": 40.75,
      "grad_norm_var": 7.595572916666667,
      "learning_rate": 0.0001,
      "loss": 7.1589,
      "loss/crossentropy": 2.018748515844345,
      "loss/hidden": 3.3578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19232797529548407,
      "step": 8990
    },
    {
      "epoch": 0.075,
      "grad_norm": 42.25,
      "grad_norm_var": 37.480143229166664,
      "learning_rate": 0.0001,
      "loss": 7.1619,
      "loss/crossentropy": 1.9784274086356164,
      "loss/hidden": 3.54453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19451803993433714,
      "step": 9000
    },
    {
      "epoch": 0.07508333333333334,
      "grad_norm": 29.5,
      "grad_norm_var": 36.496875,
      "learning_rate": 0.0001,
      "loss": 7.1749,
      "loss/crossentropy": 2.0763546764850616,
      "loss/hidden": 3.287109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18347182385623456,
      "step": 9010
    },
    {
      "epoch": 0.07516666666666667,
      "grad_norm": 47.5,
      "grad_norm_var": 20.393489583333334,
      "learning_rate": 0.0001,
      "loss": 7.1598,
      "loss/crossentropy": 1.9717566907405852,
      "loss/hidden": 3.255859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17117710653692483,
      "step": 9020
    },
    {
      "epoch": 0.07525,
      "grad_norm": 30.375,
      "grad_norm_var": 19.7056640625,
      "learning_rate": 0.0001,
      "loss": 7.234,
      "loss/crossentropy": 1.9948984555900098,
      "loss/hidden": 3.47734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19654347184114157,
      "step": 9030
    },
    {
      "epoch": 0.07533333333333334,
      "grad_norm": 31.75,
      "grad_norm_var": 2.1129557291666665,
      "learning_rate": 0.0001,
      "loss": 7.2049,
      "loss/crossentropy": 1.9566779106855392,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19929395206272602,
      "step": 9040
    },
    {
      "epoch": 0.07541666666666667,
      "grad_norm": 29.625,
      "grad_norm_var": 1.9082682291666666,
      "learning_rate": 0.0001,
      "loss": 7.1134,
      "loss/crossentropy": 1.8074573144316672,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17522313240915538,
      "step": 9050
    },
    {
      "epoch": 0.0755,
      "grad_norm": 28.25,
      "grad_norm_var": 5.104166666666667,
      "learning_rate": 0.0001,
      "loss": 7.0296,
      "loss/crossentropy": 1.8022950239479543,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18761953189969063,
      "step": 9060
    },
    {
      "epoch": 0.07558333333333334,
      "grad_norm": 33.75,
      "grad_norm_var": 5.562239583333334,
      "learning_rate": 0.0001,
      "loss": 7.2742,
      "loss/crossentropy": 2.078881235420704,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19961594343185424,
      "step": 9070
    },
    {
      "epoch": 0.07566666666666666,
      "grad_norm": 31.75,
      "grad_norm_var": 23.4962890625,
      "learning_rate": 0.0001,
      "loss": 7.0822,
      "loss/crossentropy": 1.8851124197244644,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1875220352783799,
      "step": 9080
    },
    {
      "epoch": 0.07575,
      "grad_norm": 32.25,
      "grad_norm_var": 23.117708333333333,
      "learning_rate": 0.0001,
      "loss": 7.2021,
      "loss/crossentropy": 1.8932191953063011,
      "loss/hidden": 3.32890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18913868740200995,
      "step": 9090
    },
    {
      "epoch": 0.07583333333333334,
      "grad_norm": 27.5,
      "grad_norm_var": 12.273893229166667,
      "learning_rate": 0.0001,
      "loss": 7.1001,
      "loss/crossentropy": 2.0837344110012053,
      "loss/hidden": 3.291796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18248570300638675,
      "step": 9100
    },
    {
      "epoch": 0.07591666666666666,
      "grad_norm": 31.25,
      "grad_norm_var": 7.482291666666667,
      "learning_rate": 0.0001,
      "loss": 7.2,
      "loss/crossentropy": 1.8633555322885513,
      "loss/hidden": 3.509375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894706480205059,
      "step": 9110
    },
    {
      "epoch": 0.076,
      "grad_norm": 32.0,
      "grad_norm_var": 5.141666666666667,
      "learning_rate": 0.0001,
      "loss": 7.251,
      "loss/crossentropy": 1.9421419739723205,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.187481820397079,
      "step": 9120
    },
    {
      "epoch": 0.07608333333333334,
      "grad_norm": 39.75,
      "grad_norm_var": 9.668684895833334,
      "learning_rate": 0.0001,
      "loss": 7.2181,
      "loss/crossentropy": 2.014544752240181,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18573549184948207,
      "step": 9130
    },
    {
      "epoch": 0.07616666666666666,
      "grad_norm": 33.25,
      "grad_norm_var": 8.687434895833333,
      "learning_rate": 0.0001,
      "loss": 7.1915,
      "loss/crossentropy": 1.9612121179699897,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18028553631156682,
      "step": 9140
    },
    {
      "epoch": 0.07625,
      "grad_norm": 34.0,
      "grad_norm_var": 5.188997395833334,
      "learning_rate": 0.0001,
      "loss": 7.0426,
      "loss/crossentropy": 1.8344813786447047,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17982687037438155,
      "step": 9150
    },
    {
      "epoch": 0.07633333333333334,
      "grad_norm": 30.625,
      "grad_norm_var": 6.635872395833333,
      "learning_rate": 0.0001,
      "loss": 7.242,
      "loss/crossentropy": 1.9214183151721955,
      "loss/hidden": 3.284765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19156268667429685,
      "step": 9160
    },
    {
      "epoch": 0.07641666666666666,
      "grad_norm": 31.5,
      "grad_norm_var": 4.1572265625,
      "learning_rate": 0.0001,
      "loss": 7.2402,
      "loss/crossentropy": 2.055627739429474,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2214739166200161,
      "step": 9170
    },
    {
      "epoch": 0.0765,
      "grad_norm": 49.75,
      "grad_norm_var": 416.1322916666667,
      "learning_rate": 0.0001,
      "loss": 7.2374,
      "loss/crossentropy": 1.877545380592346,
      "loss/hidden": 3.628515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19467846769839525,
      "step": 9180
    },
    {
      "epoch": 0.07658333333333334,
      "grad_norm": 33.25,
      "grad_norm_var": 454.45774739583334,
      "learning_rate": 0.0001,
      "loss": 7.2325,
      "loss/crossentropy": 2.07303985953331,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1902431882917881,
      "step": 9190
    },
    {
      "epoch": 0.07666666666666666,
      "grad_norm": 33.0,
      "grad_norm_var": 4.3917132086249184e+17,
      "learning_rate": 0.0001,
      "loss": 7.1667,
      "loss/crossentropy": 1.9947309635579586,
      "loss/hidden": 3.2984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17162247989326715,
      "step": 9200
    },
    {
      "epoch": 0.07675,
      "grad_norm": 31.875,
      "grad_norm_var": 17.014583333333334,
      "learning_rate": 0.0001,
      "loss": 7.2712,
      "loss/crossentropy": 1.9334057569503784,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19680443294346334,
      "step": 9210
    },
    {
      "epoch": 0.07683333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 13.737239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2037,
      "loss/crossentropy": 1.9268360614776612,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18807857297360897,
      "step": 9220
    },
    {
      "epoch": 0.07691666666666666,
      "grad_norm": 34.0,
      "grad_norm_var": 8.670132813815612e+17,
      "learning_rate": 0.0001,
      "loss": 7.3321,
      "loss/crossentropy": 1.991458609700203,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18901226110756397,
      "step": 9230
    },
    {
      "epoch": 0.077,
      "grad_norm": 32.5,
      "grad_norm_var": 45.81848958333333,
      "learning_rate": 0.0001,
      "loss": 7.2186,
      "loss/crossentropy": 1.8337467461824417,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18715285304933788,
      "step": 9240
    },
    {
      "epoch": 0.07708333333333334,
      "grad_norm": 31.0,
      "grad_norm_var": 54.86015625,
      "learning_rate": 0.0001,
      "loss": 7.1042,
      "loss/crossentropy": 1.899548515677452,
      "loss/hidden": 3.496875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1904792718589306,
      "step": 9250
    },
    {
      "epoch": 0.07716666666666666,
      "grad_norm": 33.25,
      "grad_norm_var": 3.133072916666667,
      "learning_rate": 0.0001,
      "loss": 7.1672,
      "loss/crossentropy": 2.0225892096757887,
      "loss/hidden": 3.199609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1757547115907073,
      "step": 9260
    },
    {
      "epoch": 0.07725,
      "grad_norm": 30.25,
      "grad_norm_var": 2.4634765625,
      "learning_rate": 0.0001,
      "loss": 7.2112,
      "loss/crossentropy": 1.8952244639396667,
      "loss/hidden": 3.4375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19042186178267,
      "step": 9270
    },
    {
      "epoch": 0.07733333333333334,
      "grad_norm": 32.25,
      "grad_norm_var": 6.151811409763807e+17,
      "learning_rate": 0.0001,
      "loss": 7.1947,
      "loss/crossentropy": 1.92966673001647,
      "loss/hidden": 3.309375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17150711864233018,
      "step": 9280
    },
    {
      "epoch": 0.07741666666666666,
      "grad_norm": 35.0,
      "grad_norm_var": 1.571886422806041e+18,
      "learning_rate": 0.0001,
      "loss": 7.2936,
      "loss/crossentropy": 2.11180402636528,
      "loss/hidden": 3.611328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21517626009881496,
      "step": 9290
    },
    {
      "epoch": 0.0775,
      "grad_norm": 32.25,
      "grad_norm_var": 3.135872395833333,
      "learning_rate": 0.0001,
      "loss": 7.1262,
      "loss/crossentropy": 1.8919033020734788,
      "loss/hidden": 3.291796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16641295626759528,
      "step": 9300
    },
    {
      "epoch": 0.07758333333333334,
      "grad_norm": 33.25,
      "grad_norm_var": 3.58515625,
      "learning_rate": 0.0001,
      "loss": 7.1427,
      "loss/crossentropy": 1.9885196268558503,
      "loss/hidden": 3.467578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20802160874009132,
      "step": 9310
    },
    {
      "epoch": 0.07766666666666666,
      "grad_norm": 30.375,
      "grad_norm_var": 1.1364583333333333,
      "learning_rate": 0.0001,
      "loss": 7.2223,
      "loss/crossentropy": 2.008994552493095,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19474735893309117,
      "step": 9320
    },
    {
      "epoch": 0.07775,
      "grad_norm": 33.75,
      "grad_norm_var": 1.6853515625,
      "learning_rate": 0.0001,
      "loss": 7.0724,
      "loss/crossentropy": 1.9341641962528229,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20453856308013202,
      "step": 9330
    },
    {
      "epoch": 0.07783333333333334,
      "grad_norm": 37.0,
      "grad_norm_var": 3.6552083333333334,
      "learning_rate": 0.0001,
      "loss": 7.2828,
      "loss/crossentropy": 2.112407311797142,
      "loss/hidden": 3.248046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18990872390568256,
      "step": 9340
    },
    {
      "epoch": 0.07791666666666666,
      "grad_norm": 33.75,
      "grad_norm_var": 5.970833333333333,
      "learning_rate": 0.0001,
      "loss": 7.2668,
      "loss/crossentropy": 1.995858334004879,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2229170560836792,
      "step": 9350
    },
    {
      "epoch": 0.078,
      "grad_norm": 30.0,
      "grad_norm_var": 4.68515625,
      "learning_rate": 0.0001,
      "loss": 7.2618,
      "loss/crossentropy": 1.9759315863251685,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.184008826687932,
      "step": 9360
    },
    {
      "epoch": 0.07808333333333334,
      "grad_norm": 32.0,
      "grad_norm_var": 5.64765625,
      "learning_rate": 0.0001,
      "loss": 7.2351,
      "loss/crossentropy": 1.964530172944069,
      "loss/hidden": 3.340234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19536323696374894,
      "step": 9370
    },
    {
      "epoch": 0.07816666666666666,
      "grad_norm": 30.125,
      "grad_norm_var": 2.1129557291666665,
      "learning_rate": 0.0001,
      "loss": 7.2376,
      "loss/crossentropy": 1.9365324467420577,
      "loss/hidden": 3.473828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.193184126727283,
      "step": 9380
    },
    {
      "epoch": 0.07825,
      "grad_norm": 31.5,
      "grad_norm_var": 18.044205729166666,
      "learning_rate": 0.0001,
      "loss": 7.1192,
      "loss/crossentropy": 2.03958273679018,
      "loss/hidden": 3.300390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18845226969569923,
      "step": 9390
    },
    {
      "epoch": 0.07833333333333334,
      "grad_norm": 32.75,
      "grad_norm_var": 20.661458333333332,
      "learning_rate": 0.0001,
      "loss": 7.2336,
      "loss/crossentropy": 1.9258702397346497,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20382192749530076,
      "step": 9400
    },
    {
      "epoch": 0.07841666666666666,
      "grad_norm": 39.5,
      "grad_norm_var": 8.745833333333334,
      "learning_rate": 0.0001,
      "loss": 7.0934,
      "loss/crossentropy": 1.9922194987535478,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19633182231336832,
      "step": 9410
    },
    {
      "epoch": 0.0785,
      "grad_norm": 30.75,
      "grad_norm_var": 5.108072916666667,
      "learning_rate": 0.0001,
      "loss": 7.1712,
      "loss/crossentropy": 1.831391155719757,
      "loss/hidden": 3.465234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18680636398494244,
      "step": 9420
    },
    {
      "epoch": 0.07858333333333334,
      "grad_norm": 31.0,
      "grad_norm_var": 0.5747395833333333,
      "learning_rate": 0.0001,
      "loss": 7.2869,
      "loss/crossentropy": 2.0193733513355254,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1822111329063773,
      "step": 9430
    },
    {
      "epoch": 0.07866666666666666,
      "grad_norm": 31.375,
      "grad_norm_var": 10.3822265625,
      "learning_rate": 0.0001,
      "loss": 7.2302,
      "loss/crossentropy": 1.8986568033695221,
      "loss/hidden": 3.293359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18191856034100057,
      "step": 9440
    },
    {
      "epoch": 0.07875,
      "grad_norm": 30.25,
      "grad_norm_var": 17.498372395833332,
      "learning_rate": 0.0001,
      "loss": 7.1577,
      "loss/crossentropy": 1.8924290344119072,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18720396794378757,
      "step": 9450
    },
    {
      "epoch": 0.07883333333333334,
      "grad_norm": 34.5,
      "grad_norm_var": 8.7625,
      "learning_rate": 0.0001,
      "loss": 7.176,
      "loss/crossentropy": 1.8431319534778594,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1925075102597475,
      "step": 9460
    },
    {
      "epoch": 0.07891666666666666,
      "grad_norm": 30.25,
      "grad_norm_var": 4.309309895833334,
      "learning_rate": 0.0001,
      "loss": 7.1943,
      "loss/crossentropy": 1.9800647720694542,
      "loss/hidden": 3.319921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18900530487298967,
      "step": 9470
    },
    {
      "epoch": 0.079,
      "grad_norm": 31.375,
      "grad_norm_var": 3.1056640625,
      "learning_rate": 0.0001,
      "loss": 7.1064,
      "loss/crossentropy": 1.9323790550231934,
      "loss/hidden": 3.32734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18270213007926941,
      "step": 9480
    },
    {
      "epoch": 0.07908333333333334,
      "grad_norm": 36.25,
      "grad_norm_var": 2.927515591570359e+17,
      "learning_rate": 0.0001,
      "loss": 7.2507,
      "loss/crossentropy": 1.8533887356519698,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1833003519102931,
      "step": 9490
    },
    {
      "epoch": 0.07916666666666666,
      "grad_norm": 32.25,
      "grad_norm_var": 2.92751559080385e+17,
      "learning_rate": 0.0001,
      "loss": 7.2466,
      "loss/crossentropy": 2.010724890232086,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1888010736554861,
      "step": 9500
    },
    {
      "epoch": 0.07925,
      "grad_norm": 29.875,
      "grad_norm_var": 2.25390625,
      "learning_rate": 0.0001,
      "loss": 7.2362,
      "loss/crossentropy": 1.9017160810530185,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20483599258586765,
      "step": 9510
    },
    {
      "epoch": 0.07933333333333334,
      "grad_norm": 28.875,
      "grad_norm_var": 13.957747395833334,
      "learning_rate": 0.0001,
      "loss": 7.3128,
      "loss/crossentropy": 1.917439240217209,
      "loss/hidden": 3.27734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16879426389932634,
      "step": 9520
    },
    {
      "epoch": 0.07941666666666666,
      "grad_norm": 38.75,
      "grad_norm_var": 30.634375,
      "learning_rate": 0.0001,
      "loss": 7.2802,
      "loss/crossentropy": 2.0289409935474394,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19048050362616778,
      "step": 9530
    },
    {
      "epoch": 0.0795,
      "grad_norm": 28.0,
      "grad_norm_var": 25.167122395833335,
      "learning_rate": 0.0001,
      "loss": 7.1468,
      "loss/crossentropy": 1.7849997401237487,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18105144426226616,
      "step": 9540
    },
    {
      "epoch": 0.07958333333333334,
      "grad_norm": 33.25,
      "grad_norm_var": 1.5832967226589184e+18,
      "learning_rate": 0.0001,
      "loss": 7.2816,
      "loss/crossentropy": 1.9449206456542014,
      "loss/hidden": 3.60234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1812568910419941,
      "step": 9550
    },
    {
      "epoch": 0.07966666666666666,
      "grad_norm": 33.0,
      "grad_norm_var": 1.5832967232775782e+18,
      "learning_rate": 0.0001,
      "loss": 7.1754,
      "loss/crossentropy": 1.7890570968389512,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17575022913515567,
      "step": 9560
    },
    {
      "epoch": 0.07975,
      "grad_norm": 30.375,
      "grad_norm_var": 8.748418048102062e+17,
      "learning_rate": 0.0001,
      "loss": 7.2619,
      "loss/crossentropy": 1.9128870755434035,
      "loss/hidden": 3.663671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20060238279402257,
      "step": 9570
    },
    {
      "epoch": 0.07983333333333334,
      "grad_norm": 30.625,
      "grad_norm_var": 1.4955729166666667,
      "learning_rate": 0.0001,
      "loss": 7.2103,
      "loss/crossentropy": 1.8678260400891304,
      "loss/hidden": 3.4265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20040124338120222,
      "step": 9580
    },
    {
      "epoch": 0.07991666666666666,
      "grad_norm": 32.5,
      "grad_norm_var": 16.593489583333334,
      "learning_rate": 0.0001,
      "loss": 7.1735,
      "loss/crossentropy": 1.7469847694039344,
      "loss/hidden": 3.29453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16893960312008857,
      "step": 9590
    },
    {
      "epoch": 0.08,
      "grad_norm": 31.0,
      "grad_norm_var": 14.876497395833333,
      "learning_rate": 0.0001,
      "loss": 7.2358,
      "loss/crossentropy": 2.0145472913980482,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21668562162667512,
      "step": 9600
    },
    {
      "epoch": 0.08008333333333334,
      "grad_norm": 31.25,
      "grad_norm_var": 4.614583333333333,
      "learning_rate": 0.0001,
      "loss": 7.2138,
      "loss/crossentropy": 1.8460545286536216,
      "loss/hidden": 3.547265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19596499875187873,
      "step": 9610
    },
    {
      "epoch": 0.08016666666666666,
      "grad_norm": 30.375,
      "grad_norm_var": 1.2824703444335985e+18,
      "learning_rate": 0.0001,
      "loss": 7.2641,
      "loss/crossentropy": 1.9291812598705291,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1993147611618042,
      "step": 9620
    },
    {
      "epoch": 0.08025,
      "grad_norm": 28.75,
      "grad_norm_var": 2.253059895833333,
      "learning_rate": 0.0001,
      "loss": 7.0381,
      "loss/crossentropy": 1.9004499420523644,
      "loss/hidden": 3.28046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17315188869833947,
      "step": 9630
    },
    {
      "epoch": 0.08033333333333334,
      "grad_norm": 49.75,
      "grad_norm_var": 24.8875,
      "learning_rate": 0.0001,
      "loss": 7.2074,
      "loss/crossentropy": 2.0676659286022185,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19399872440844773,
      "step": 9640
    },
    {
      "epoch": 0.08041666666666666,
      "grad_norm": 28.25,
      "grad_norm_var": 21.844205729166667,
      "learning_rate": 0.0001,
      "loss": 7.1626,
      "loss/crossentropy": 1.8846231751143931,
      "loss/hidden": 3.569921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1813927460461855,
      "step": 9650
    },
    {
      "epoch": 0.0805,
      "grad_norm": 32.75,
      "grad_norm_var": 5.473372395833334,
      "learning_rate": 0.0001,
      "loss": 7.2297,
      "loss/crossentropy": 1.95650455057621,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18862175196409225,
      "step": 9660
    },
    {
      "epoch": 0.08058333333333334,
      "grad_norm": 31.375,
      "grad_norm_var": 3.873893229166667,
      "learning_rate": 0.0001,
      "loss": 7.2405,
      "loss/crossentropy": 2.0532629549503327,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1972910648211837,
      "step": 9670
    },
    {
      "epoch": 0.08066666666666666,
      "grad_norm": 28.625,
      "grad_norm_var": 4.488997395833334,
      "learning_rate": 0.0001,
      "loss": 7.2227,
      "loss/crossentropy": 1.8982819005846978,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18387150373309852,
      "step": 9680
    },
    {
      "epoch": 0.08075,
      "grad_norm": 30.5,
      "grad_norm_var": 23.84765625,
      "learning_rate": 0.0001,
      "loss": 7.2229,
      "loss/crossentropy": 2.042688661813736,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19536025915294886,
      "step": 9690
    },
    {
      "epoch": 0.08083333333333333,
      "grad_norm": 28.5,
      "grad_norm_var": 7.214322916666666,
      "learning_rate": 0.0001,
      "loss": 7.1841,
      "loss/crossentropy": 1.8631189227104188,
      "loss/hidden": 3.24921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16748407240957022,
      "step": 9700
    },
    {
      "epoch": 0.08091666666666666,
      "grad_norm": 38.5,
      "grad_norm_var": 8.705989583333333,
      "learning_rate": 0.0001,
      "loss": 7.0953,
      "loss/crossentropy": 2.096615809202194,
      "loss/hidden": 3.247265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19336612224578859,
      "step": 9710
    },
    {
      "epoch": 0.081,
      "grad_norm": 31.625,
      "grad_norm_var": 8.949739583333333,
      "learning_rate": 0.0001,
      "loss": 7.0645,
      "loss/crossentropy": 1.7769379287958145,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17712265541777014,
      "step": 9720
    },
    {
      "epoch": 0.08108333333333333,
      "grad_norm": 31.875,
      "grad_norm_var": 18.835872395833334,
      "learning_rate": 0.0001,
      "loss": 7.2842,
      "loss/crossentropy": 2.0219400838017463,
      "loss/hidden": 3.3078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19118713159114123,
      "step": 9730
    },
    {
      "epoch": 0.08116666666666666,
      "grad_norm": 41.5,
      "grad_norm_var": 22.01015625,
      "learning_rate": 0.0001,
      "loss": 7.2142,
      "loss/crossentropy": 2.1162992775440217,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19361746832728385,
      "step": 9740
    },
    {
      "epoch": 0.08125,
      "grad_norm": 29.375,
      "grad_norm_var": 10.814583333333333,
      "learning_rate": 0.0001,
      "loss": 7.2149,
      "loss/crossentropy": 1.854822640120983,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17972449399530888,
      "step": 9750
    },
    {
      "epoch": 0.08133333333333333,
      "grad_norm": 31.625,
      "grad_norm_var": 4.9228515625,
      "learning_rate": 0.0001,
      "loss": 7.1515,
      "loss/crossentropy": 2.042907669395208,
      "loss/hidden": 3.523046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20937258377671242,
      "step": 9760
    },
    {
      "epoch": 0.08141666666666666,
      "grad_norm": 33.5,
      "grad_norm_var": 9.533333333333333,
      "learning_rate": 0.0001,
      "loss": 7.1791,
      "loss/crossentropy": 2.0195691674947738,
      "loss/hidden": 3.46796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1953078692778945,
      "step": 9770
    },
    {
      "epoch": 0.0815,
      "grad_norm": 38.0,
      "grad_norm_var": 13.008333333333333,
      "learning_rate": 0.0001,
      "loss": 7.1171,
      "loss/crossentropy": 1.8863514378666877,
      "loss/hidden": 3.316796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18283969424664975,
      "step": 9780
    },
    {
      "epoch": 0.08158333333333333,
      "grad_norm": 40.25,
      "grad_norm_var": 1428.9330729166666,
      "learning_rate": 0.0001,
      "loss": 7.4353,
      "loss/crossentropy": 1.9719516918063165,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2090709051117301,
      "step": 9790
    },
    {
      "epoch": 0.08166666666666667,
      "grad_norm": 33.75,
      "grad_norm_var": 6.641080729166666,
      "learning_rate": 0.0001,
      "loss": 7.181,
      "loss/crossentropy": 2.0369949251413346,
      "loss/hidden": 3.294921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18184055741876365,
      "step": 9800
    },
    {
      "epoch": 0.08175,
      "grad_norm": 31.25,
      "grad_norm_var": 4.873893229166667,
      "learning_rate": 0.0001,
      "loss": 7.0857,
      "loss/crossentropy": 2.026154139637947,
      "loss/hidden": 3.314453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18145214160904288,
      "step": 9810
    },
    {
      "epoch": 0.08183333333333333,
      "grad_norm": 31.875,
      "grad_norm_var": 2.8608723958333333,
      "learning_rate": 0.0001,
      "loss": 7.2217,
      "loss/crossentropy": 1.984486585855484,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2005698978435248,
      "step": 9820
    },
    {
      "epoch": 0.08191666666666667,
      "grad_norm": 32.0,
      "grad_norm_var": 26.6291015625,
      "learning_rate": 0.0001,
      "loss": 7.1779,
      "loss/crossentropy": 2.0041553929448126,
      "loss/hidden": 3.257421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17574419379234313,
      "step": 9830
    },
    {
      "epoch": 0.082,
      "grad_norm": 28.875,
      "grad_norm_var": 3.23125,
      "learning_rate": 0.0001,
      "loss": 7.1577,
      "loss/crossentropy": 1.8619619227945805,
      "loss/hidden": 3.340234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19841333236545325,
      "step": 9840
    },
    {
      "epoch": 0.08208333333333333,
      "grad_norm": 28.625,
      "grad_norm_var": 15.088541666666666,
      "learning_rate": 0.0001,
      "loss": 7.2013,
      "loss/crossentropy": 1.898135930299759,
      "loss/hidden": 3.440625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1854542102664709,
      "step": 9850
    },
    {
      "epoch": 0.08216666666666667,
      "grad_norm": 30.5,
      "grad_norm_var": 12.376822916666667,
      "learning_rate": 0.0001,
      "loss": 7.1536,
      "loss/crossentropy": 1.9185454472899437,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20166656840592623,
      "step": 9860
    },
    {
      "epoch": 0.08225,
      "grad_norm": 32.25,
      "grad_norm_var": 3.834375,
      "learning_rate": 0.0001,
      "loss": 7.2058,
      "loss/crossentropy": 1.919453853368759,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1870553806424141,
      "step": 9870
    },
    {
      "epoch": 0.08233333333333333,
      "grad_norm": 31.125,
      "grad_norm_var": 7.107418945754236e+17,
      "learning_rate": 0.0001,
      "loss": 7.1974,
      "loss/crossentropy": 1.889936888217926,
      "loss/hidden": 3.57265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18729659765958787,
      "step": 9880
    },
    {
      "epoch": 0.08241666666666667,
      "grad_norm": 33.25,
      "grad_norm_var": 7.107418942803543e+17,
      "learning_rate": 0.0001,
      "loss": 7.2175,
      "loss/crossentropy": 2.0169277042150497,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20186357703059912,
      "step": 9890
    },
    {
      "epoch": 0.0825,
      "grad_norm": 34.0,
      "grad_norm_var": 8.82705511779559e+17,
      "learning_rate": 0.0001,
      "loss": 7.3193,
      "loss/crossentropy": 2.0031974256038665,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18207758255302905,
      "step": 9900
    },
    {
      "epoch": 0.08258333333333333,
      "grad_norm": 43.5,
      "grad_norm_var": 8.827055118187059e+17,
      "learning_rate": 0.0001,
      "loss": 7.1468,
      "loss/crossentropy": 1.8897509559988976,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18627819493412973,
      "step": 9910
    },
    {
      "epoch": 0.08266666666666667,
      "grad_norm": 31.625,
      "grad_norm_var": 13.143489583333333,
      "learning_rate": 0.0001,
      "loss": 7.1643,
      "loss/crossentropy": 1.96635133177042,
      "loss/hidden": 3.22109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17648211270570754,
      "step": 9920
    },
    {
      "epoch": 0.08275,
      "grad_norm": 34.75,
      "grad_norm_var": 1.2447526979635249e+18,
      "learning_rate": 0.0001,
      "loss": 7.1337,
      "loss/crossentropy": 1.7700391985476016,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17953332625329493,
      "step": 9930
    },
    {
      "epoch": 0.08283333333333333,
      "grad_norm": 29.0,
      "grad_norm_var": 10.516666666666667,
      "learning_rate": 0.0001,
      "loss": 7.1734,
      "loss/crossentropy": 1.9698360793292522,
      "loss/hidden": 3.2671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18065936379134656,
      "step": 9940
    },
    {
      "epoch": 0.08291666666666667,
      "grad_norm": 30.625,
      "grad_norm_var": 2.0458333333333334,
      "learning_rate": 0.0001,
      "loss": 7.1922,
      "loss/crossentropy": 1.9713639967143535,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1909080557525158,
      "step": 9950
    },
    {
      "epoch": 0.083,
      "grad_norm": 33.25,
      "grad_norm_var": 3.8082682291666665,
      "learning_rate": 0.0001,
      "loss": 7.2286,
      "loss/crossentropy": 1.9554906249046327,
      "loss/hidden": 3.34296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17579805050045252,
      "step": 9960
    },
    {
      "epoch": 0.08308333333333333,
      "grad_norm": 30.75,
      "grad_norm_var": 3.6155598958333335,
      "learning_rate": 0.0001,
      "loss": 7.0742,
      "loss/crossentropy": 1.9254092440009116,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17711936933919786,
      "step": 9970
    },
    {
      "epoch": 0.08316666666666667,
      "grad_norm": 29.875,
      "grad_norm_var": 3.0809895833333334,
      "learning_rate": 0.0001,
      "loss": 7.2025,
      "loss/crossentropy": 2.0318453282117845,
      "loss/hidden": 3.2765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1854577951133251,
      "step": 9980
    },
    {
      "epoch": 0.08325,
      "grad_norm": 30.125,
      "grad_norm_var": 2.5134765625,
      "learning_rate": 0.0001,
      "loss": 7.117,
      "loss/crossentropy": 1.913966417312622,
      "loss/hidden": 3.28984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17688458357006312,
      "step": 9990
    },
    {
      "epoch": 0.08333333333333333,
      "grad_norm": 29.75,
      "grad_norm_var": 2.2705729166666666,
      "learning_rate": 0.0001,
      "loss": 7.1223,
      "loss/crossentropy": 1.8660702288150788,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17869504541158676,
      "step": 10000
    }
  ],
  "logging_steps": 10,
  "max_steps": 120000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 9223372036854775807,
  "save_steps": 10000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 2.8575100320088064e+19,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}