{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.375,
  "eval_steps": 2000,
  "global_step": 15000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.00025,
      "grad_norm": 31.5,
      "learning_rate": 0.0001,
      "loss": 7.633,
      "loss/crossentropy": 2.065455098450184,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20220321230590343,
      "step": 10
    },
    {
      "epoch": 0.0005,
      "grad_norm": 35.0,
      "grad_norm_var": 2.6895182291666666,
      "learning_rate": 0.0001,
      "loss": 7.4618,
      "loss/crossentropy": 1.9399560801684856,
      "loss/hidden": 3.394140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19191570337861777,
      "step": 20
    },
    {
      "epoch": 0.00075,
      "grad_norm": 37.5,
      "grad_norm_var": 6.579622395833334,
      "learning_rate": 0.0001,
      "loss": 7.5972,
      "loss/crossentropy": 2.130601316690445,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20188977513462306,
      "step": 30
    },
    {
      "epoch": 0.001,
      "grad_norm": 33.5,
      "grad_norm_var": 6.253125,
      "learning_rate": 0.0001,
      "loss": 7.5917,
      "loss/crossentropy": 2.2571407079696657,
      "loss/hidden": 3.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19847887996584176,
      "step": 40
    },
    {
      "epoch": 0.00125,
      "grad_norm": 32.25,
      "grad_norm_var": 2.1619140625,
      "learning_rate": 0.0001,
      "loss": 7.6054,
      "loss/crossentropy": 2.1717565625905992,
      "loss/hidden": 3.43359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20264342725276946,
      "step": 50
    },
    {
      "epoch": 0.0015,
      "grad_norm": 35.5,
      "grad_norm_var": 15.786393229166666,
      "learning_rate": 0.0001,
      "loss": 7.5513,
      "loss/crossentropy": 2.070718301087618,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19855907820165158,
      "step": 60
    },
    {
      "epoch": 0.00175,
      "grad_norm": 31.0,
      "grad_norm_var": 12.4625,
      "learning_rate": 0.0001,
      "loss": 7.5447,
      "loss/crossentropy": 2.118075390160084,
      "loss/hidden": 3.473828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20283062420785428,
      "step": 70
    },
    {
      "epoch": 0.002,
      "grad_norm": 32.25,
      "grad_norm_var": 1.2643229166666667,
      "learning_rate": 0.0001,
      "loss": 7.468,
      "loss/crossentropy": 2.0006178975105287,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18958428762853147,
      "step": 80
    },
    {
      "epoch": 0.00225,
      "grad_norm": 30.625,
      "grad_norm_var": 3.470572916666667,
      "learning_rate": 0.0001,
      "loss": 7.5061,
      "loss/crossentropy": 1.9605075903236866,
      "loss/hidden": 3.54375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20559987109154462,
      "step": 90
    },
    {
      "epoch": 0.0025,
      "grad_norm": 31.125,
      "grad_norm_var": 6.763541666666667,
      "learning_rate": 0.0001,
      "loss": 7.4928,
      "loss/crossentropy": 2.1205389350652695,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19496036488562823,
      "step": 100
    },
    {
      "epoch": 0.00275,
      "grad_norm": 31.0,
      "grad_norm_var": 6.1509765625,
      "learning_rate": 0.0001,
      "loss": 7.595,
      "loss/crossentropy": 2.1240097641944886,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19564666803926228,
      "step": 110
    },
    {
      "epoch": 0.003,
      "grad_norm": 31.25,
      "grad_norm_var": 3.348893229166667,
      "learning_rate": 0.0001,
      "loss": 7.5329,
      "loss/crossentropy": 2.175096944719553,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21303062327206135,
      "step": 120
    },
    {
      "epoch": 0.00325,
      "grad_norm": 32.0,
      "grad_norm_var": 2.8541666666666665,
      "learning_rate": 0.0001,
      "loss": 7.5536,
      "loss/crossentropy": 2.1472502022981645,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18929538186639547,
      "step": 130
    },
    {
      "epoch": 0.0035,
      "grad_norm": 29.375,
      "grad_norm_var": 29.683268229166668,
      "learning_rate": 0.0001,
      "loss": 7.5191,
      "loss/crossentropy": 2.015011890232563,
      "loss/hidden": 3.44296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20328481420874595,
      "step": 140
    },
    {
      "epoch": 0.00375,
      "grad_norm": 28.75,
      "grad_norm_var": 28.74765625,
      "learning_rate": 0.0001,
      "loss": 7.4158,
      "loss/crossentropy": 1.9774167470633983,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19464388117194176,
      "step": 150
    },
    {
      "epoch": 0.004,
      "grad_norm": 30.875,
      "grad_norm_var": 1.3635416666666667,
      "learning_rate": 0.0001,
      "loss": 7.6354,
      "loss/crossentropy": 2.320629420876503,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20745602920651435,
      "step": 160
    },
    {
      "epoch": 0.00425,
      "grad_norm": 31.5,
      "grad_norm_var": 1.0270182291666667,
      "learning_rate": 0.0001,
      "loss": 7.4137,
      "loss/crossentropy": 1.900385806709528,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16769229620695114,
      "step": 170
    },
    {
      "epoch": 0.0045,
      "grad_norm": 31.25,
      "grad_norm_var": 0.9833333333333333,
      "learning_rate": 0.0001,
      "loss": 7.5763,
      "loss/crossentropy": 2.129625543951988,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2102549459785223,
      "step": 180
    },
    {
      "epoch": 0.00475,
      "grad_norm": 32.25,
      "grad_norm_var": 3.05390625,
      "learning_rate": 0.0001,
      "loss": 7.6166,
      "loss/crossentropy": 2.1552532628178596,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2250068686902523,
      "step": 190
    },
    {
      "epoch": 0.005,
      "grad_norm": 29.625,
      "grad_norm_var": 3.8375,
      "learning_rate": 0.0001,
      "loss": 7.5745,
      "loss/crossentropy": 1.9441482461988926,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.195942450594157,
      "step": 200
    },
    {
      "epoch": 0.00525,
      "grad_norm": 32.5,
      "grad_norm_var": 18.396875,
      "learning_rate": 0.0001,
      "loss": 7.5292,
      "loss/crossentropy": 1.9941987417638303,
      "loss/hidden": 3.394140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18264975901693106,
      "step": 210
    },
    {
      "epoch": 0.0055,
      "grad_norm": 31.75,
      "grad_norm_var": 20.736393229166666,
      "learning_rate": 0.0001,
      "loss": 7.4899,
      "loss/crossentropy": 2.0191620789468288,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18100650198757648,
      "step": 220
    },
    {
      "epoch": 0.00575,
      "grad_norm": 30.375,
      "grad_norm_var": 2.342643229166667,
      "learning_rate": 0.0001,
      "loss": 7.5199,
      "loss/crossentropy": 2.001779730618,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17959208656102418,
      "step": 230
    },
    {
      "epoch": 0.006,
      "grad_norm": 30.75,
      "grad_norm_var": 1.271875,
      "learning_rate": 0.0001,
      "loss": 7.6842,
      "loss/crossentropy": 2.1846971333026888,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2059234745800495,
      "step": 240
    },
    {
      "epoch": 0.00625,
      "grad_norm": 29.5,
      "grad_norm_var": 5.688541666666667,
      "learning_rate": 0.0001,
      "loss": 7.5196,
      "loss/crossentropy": 2.174124576151371,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20000722594559192,
      "step": 250
    },
    {
      "epoch": 0.0065,
      "grad_norm": 28.75,
      "grad_norm_var": 1.9572265625,
      "learning_rate": 0.0001,
      "loss": 7.3875,
      "loss/crossentropy": 1.9285166233778,
      "loss/hidden": 3.396875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18449910767376423,
      "step": 260
    },
    {
      "epoch": 0.00675,
      "grad_norm": 33.5,
      "grad_norm_var": 2.0999348958333335,
      "learning_rate": 0.0001,
      "loss": 7.5877,
      "loss/crossentropy": 2.0323276594281197,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19395631980150937,
      "step": 270
    },
    {
      "epoch": 0.007,
      "grad_norm": 30.5,
      "grad_norm_var": 2.15390625,
      "learning_rate": 0.0001,
      "loss": 7.5791,
      "loss/crossentropy": 2.126656140387058,
      "loss/hidden": 3.496875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21661139875650406,
      "step": 280
    },
    {
      "epoch": 0.00725,
      "grad_norm": 29.5,
      "grad_norm_var": 3.193489583333333,
      "learning_rate": 0.0001,
      "loss": 7.5587,
      "loss/crossentropy": 2.200097793340683,
      "loss/hidden": 3.529296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21046234332025052,
      "step": 290
    },
    {
      "epoch": 0.0075,
      "grad_norm": 26.75,
      "grad_norm_var": 4.27265625,
      "learning_rate": 0.0001,
      "loss": 7.5404,
      "loss/crossentropy": 2.1184144005179406,
      "loss/hidden": 3.487890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20949590150266886,
      "step": 300
    },
    {
      "epoch": 0.00775,
      "grad_norm": 33.0,
      "grad_norm_var": 3.3643229166666666,
      "learning_rate": 0.0001,
      "loss": 7.5628,
      "loss/crossentropy": 1.9984030593186617,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18789457948878407,
      "step": 310
    },
    {
      "epoch": 0.008,
      "grad_norm": 32.5,
      "grad_norm_var": 2.5645182291666666,
      "learning_rate": 0.0001,
      "loss": 7.5695,
      "loss/crossentropy": 2.143594169616699,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19360470157116652,
      "step": 320
    },
    {
      "epoch": 0.00825,
      "grad_norm": 29.375,
      "grad_norm_var": 1.8749348958333334,
      "learning_rate": 0.0001,
      "loss": 7.3627,
      "loss/crossentropy": 2.1077703177928924,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19771252572536469,
      "step": 330
    },
    {
      "epoch": 0.0085,
      "grad_norm": 29.75,
      "grad_norm_var": 1.5978515625,
      "learning_rate": 0.0001,
      "loss": 7.4192,
      "loss/crossentropy": 2.0583472289144993,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20189273860305548,
      "step": 340
    },
    {
      "epoch": 0.00875,
      "grad_norm": 29.875,
      "grad_norm_var": 1.2872395833333334,
      "learning_rate": 0.0001,
      "loss": 7.5432,
      "loss/crossentropy": 2.0804511278867723,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19735569059848784,
      "step": 350
    },
    {
      "epoch": 0.009,
      "grad_norm": 30.5,
      "grad_norm_var": 18.731184895833334,
      "learning_rate": 0.0001,
      "loss": 7.4948,
      "loss/crossentropy": 2.0466629534959795,
      "loss/hidden": 3.315234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18366040643304588,
      "step": 360
    },
    {
      "epoch": 0.00925,
      "grad_norm": 30.875,
      "grad_norm_var": 25.9916015625,
      "learning_rate": 0.0001,
      "loss": 7.5081,
      "loss/crossentropy": 1.9005662694573402,
      "loss/hidden": 3.501171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1900689721107483,
      "step": 370
    },
    {
      "epoch": 0.0095,
      "grad_norm": 28.75,
      "grad_norm_var": 2.451041666666667,
      "learning_rate": 0.0001,
      "loss": 7.4305,
      "loss/crossentropy": 2.0674299761652946,
      "loss/hidden": 3.517578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21062961965799332,
      "step": 380
    },
    {
      "epoch": 0.00975,
      "grad_norm": 31.25,
      "grad_norm_var": 5.645247395833334,
      "learning_rate": 0.0001,
      "loss": 7.5168,
      "loss/crossentropy": 2.0279919117689134,
      "loss/hidden": 3.503125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20519332773983479,
      "step": 390
    },
    {
      "epoch": 0.01,
      "grad_norm": 31.125,
      "grad_norm_var": 5.928125,
      "learning_rate": 0.0001,
      "loss": 7.4985,
      "loss/crossentropy": 2.0427632443606853,
      "loss/hidden": 3.53125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20287631042301654,
      "step": 400
    },
    {
      "epoch": 0.01025,
      "grad_norm": 38.5,
      "grad_norm_var": 438.43515625,
      "learning_rate": 0.0001,
      "loss": 7.5633,
      "loss/crossentropy": 2.199043881893158,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21130343191325665,
      "step": 410
    },
    {
      "epoch": 0.0105,
      "grad_norm": 30.875,
      "grad_norm_var": 43.14140625,
      "learning_rate": 0.0001,
      "loss": 7.4835,
      "loss/crossentropy": 1.9102243572473525,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1895731385797262,
      "step": 420
    },
    {
      "epoch": 0.01075,
      "grad_norm": 31.75,
      "grad_norm_var": 5.658268229166667,
      "learning_rate": 0.0001,
      "loss": 7.3897,
      "loss/crossentropy": 2.159160128980875,
      "loss/hidden": 3.464453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20280379485338926,
      "step": 430
    },
    {
      "epoch": 0.011,
      "grad_norm": 28.375,
      "grad_norm_var": 16.3375,
      "learning_rate": 0.0001,
      "loss": 7.5463,
      "loss/crossentropy": 2.1217672407627104,
      "loss/hidden": 3.545703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23856931366026402,
      "step": 440
    },
    {
      "epoch": 0.01125,
      "grad_norm": 30.5,
      "grad_norm_var": 17.098372395833334,
      "learning_rate": 0.0001,
      "loss": 7.5225,
      "loss/crossentropy": 1.969854873791337,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19548849146813155,
      "step": 450
    },
    {
      "epoch": 0.0115,
      "grad_norm": 29.875,
      "grad_norm_var": 2.5677083333333335,
      "learning_rate": 0.0001,
      "loss": 7.5046,
      "loss/crossentropy": 2.121321603655815,
      "loss/hidden": 3.476171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19364523217082025,
      "step": 460
    },
    {
      "epoch": 0.01175,
      "grad_norm": 32.25,
      "grad_norm_var": 8.585416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4558,
      "loss/crossentropy": 1.9360710382461548,
      "loss/hidden": 3.382421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1893781816586852,
      "step": 470
    },
    {
      "epoch": 0.012,
      "grad_norm": 29.875,
      "grad_norm_var": 3.417122395833333,
      "learning_rate": 0.0001,
      "loss": 7.531,
      "loss/crossentropy": 2.082458943128586,
      "loss/hidden": 3.471875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2220946006476879,
      "step": 480
    },
    {
      "epoch": 0.01225,
      "grad_norm": 31.0,
      "grad_norm_var": 48.96640625,
      "learning_rate": 0.0001,
      "loss": 7.5651,
      "loss/crossentropy": 2.1382531195878984,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20847559962421655,
      "step": 490
    },
    {
      "epoch": 0.0125,
      "grad_norm": 29.875,
      "grad_norm_var": 49.2666015625,
      "learning_rate": 0.0001,
      "loss": 7.5679,
      "loss/crossentropy": 2.0875915244221686,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1850985599681735,
      "step": 500
    },
    {
      "epoch": 0.01275,
      "grad_norm": 31.875,
      "grad_norm_var": 1.45,
      "learning_rate": 0.0001,
      "loss": 7.5263,
      "loss/crossentropy": 2.182442346215248,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19555890336632728,
      "step": 510
    },
    {
      "epoch": 0.013,
      "grad_norm": 34.0,
      "grad_norm_var": 1.6931640625,
      "learning_rate": 0.0001,
      "loss": 7.5209,
      "loss/crossentropy": 1.9812136888504028,
      "loss/hidden": 3.4921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1965757070109248,
      "step": 520
    },
    {
      "epoch": 0.01325,
      "grad_norm": 31.0,
      "grad_norm_var": 2.101822916666667,
      "learning_rate": 0.0001,
      "loss": 7.6059,
      "loss/crossentropy": 2.0372241511940956,
      "loss/hidden": 3.564453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.204646560549736,
      "step": 530
    },
    {
      "epoch": 0.0135,
      "grad_norm": 29.125,
      "grad_norm_var": 20.071875,
      "learning_rate": 0.0001,
      "loss": 7.5725,
      "loss/crossentropy": 2.155761349201202,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19602423422038556,
      "step": 540
    },
    {
      "epoch": 0.01375,
      "grad_norm": 29.125,
      "grad_norm_var": 20.506705729166665,
      "learning_rate": 0.0001,
      "loss": 7.5842,
      "loss/crossentropy": 1.8869566857814788,
      "loss/hidden": 3.437890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20957522764801978,
      "step": 550
    },
    {
      "epoch": 0.014,
      "grad_norm": 30.625,
      "grad_norm_var": 10.025455729166667,
      "learning_rate": 0.0001,
      "loss": 7.4975,
      "loss/crossentropy": 2.0370677679777147,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19026046600192786,
      "step": 560
    },
    {
      "epoch": 0.01425,
      "grad_norm": 33.0,
      "grad_norm_var": 2.2270833333333333,
      "learning_rate": 0.0001,
      "loss": 7.5688,
      "loss/crossentropy": 2.1931444257497787,
      "loss/hidden": 3.415234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2036376902833581,
      "step": 570
    },
    {
      "epoch": 0.0145,
      "grad_norm": 35.0,
      "grad_norm_var": 3.5681640625,
      "learning_rate": 0.0001,
      "loss": 7.478,
      "loss/crossentropy": 2.061052493005991,
      "loss/hidden": 3.478125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2282864760607481,
      "step": 580
    },
    {
      "epoch": 0.01475,
      "grad_norm": 32.5,
      "grad_norm_var": 2.8705729166666667,
      "learning_rate": 0.0001,
      "loss": 7.5957,
      "loss/crossentropy": 2.0078392371535303,
      "loss/hidden": 3.45,
      "loss/jsd": 0.0,
      "loss/logits": 0.19647251404821872,
      "step": 590
    },
    {
      "epoch": 0.015,
      "grad_norm": 30.25,
      "grad_norm_var": 31.449934895833334,
      "learning_rate": 0.0001,
      "loss": 7.5096,
      "loss/crossentropy": 2.0417068414390087,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19782953998073935,
      "step": 600
    },
    {
      "epoch": 0.01525,
      "grad_norm": 30.5,
      "grad_norm_var": 26.253059895833335,
      "learning_rate": 0.0001,
      "loss": 7.5368,
      "loss/crossentropy": 2.1738049775362014,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1996332859620452,
      "step": 610
    },
    {
      "epoch": 0.0155,
      "grad_norm": 30.125,
      "grad_norm_var": 2.334375,
      "learning_rate": 0.0001,
      "loss": 7.4868,
      "loss/crossentropy": 1.7587297886610032,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18938990794122218,
      "step": 620
    },
    {
      "epoch": 0.01575,
      "grad_norm": 29.25,
      "grad_norm_var": 27.393684895833335,
      "learning_rate": 0.0001,
      "loss": 7.4833,
      "loss/crossentropy": 1.9551145888864994,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20075901364907622,
      "step": 630
    },
    {
      "epoch": 0.016,
      "grad_norm": 29.75,
      "grad_norm_var": 29.6947265625,
      "learning_rate": 0.0001,
      "loss": 7.4608,
      "loss/crossentropy": 2.128718316555023,
      "loss/hidden": 3.3625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19077460393309592,
      "step": 640
    },
    {
      "epoch": 0.01625,
      "grad_norm": 29.75,
      "grad_norm_var": 27.322330729166666,
      "learning_rate": 0.0001,
      "loss": 7.6033,
      "loss/crossentropy": 1.9678708665072917,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18875791020691396,
      "step": 650
    },
    {
      "epoch": 0.0165,
      "grad_norm": 30.375,
      "grad_norm_var": 3.129622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3873,
      "loss/crossentropy": 1.9582339562475681,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18309127148240806,
      "step": 660
    },
    {
      "epoch": 0.01675,
      "grad_norm": 32.75,
      "grad_norm_var": 2.7009765625,
      "learning_rate": 0.0001,
      "loss": 7.4913,
      "loss/crossentropy": 2.0773802563548087,
      "loss/hidden": 3.505078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20910798981785775,
      "step": 670
    },
    {
      "epoch": 0.017,
      "grad_norm": 34.0,
      "grad_norm_var": 3.3854166666666665,
      "learning_rate": 0.0001,
      "loss": 7.4847,
      "loss/crossentropy": 2.12913373708725,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.201920267008245,
      "step": 680
    },
    {
      "epoch": 0.01725,
      "grad_norm": 30.75,
      "grad_norm_var": 1.7176432291666666,
      "learning_rate": 0.0001,
      "loss": 7.5065,
      "loss/crossentropy": 1.9141538538038732,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1841401271522045,
      "step": 690
    },
    {
      "epoch": 0.0175,
      "grad_norm": 31.0,
      "grad_norm_var": 1.6374348958333333,
      "learning_rate": 0.0001,
      "loss": 7.5897,
      "loss/crossentropy": 2.207232800126076,
      "loss/hidden": 3.399609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21376523859798907,
      "step": 700
    },
    {
      "epoch": 0.01775,
      "grad_norm": 32.75,
      "grad_norm_var": 2.3655598958333335,
      "learning_rate": 0.0001,
      "loss": 7.5075,
      "loss/crossentropy": 2.03845998942852,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1920805646572262,
      "step": 710
    },
    {
      "epoch": 0.018,
      "grad_norm": 32.5,
      "grad_norm_var": 1.3893229166666667,
      "learning_rate": 0.0001,
      "loss": 7.4669,
      "loss/crossentropy": 2.054341807588935,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19716067584231495,
      "step": 720
    },
    {
      "epoch": 0.01825,
      "grad_norm": 31.625,
      "grad_norm_var": 3.54140625,
      "learning_rate": 0.0001,
      "loss": 7.517,
      "loss/crossentropy": 2.2111608639359472,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20262118335813284,
      "step": 730
    },
    {
      "epoch": 0.0185,
      "grad_norm": 29.125,
      "grad_norm_var": 4.692122395833334,
      "learning_rate": 0.0001,
      "loss": 7.4784,
      "loss/crossentropy": 2.0551758617162705,
      "loss/hidden": 3.446875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20378697756677866,
      "step": 740
    },
    {
      "epoch": 0.01875,
      "grad_norm": 33.0,
      "grad_norm_var": 4.295572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4016,
      "loss/crossentropy": 2.128055375814438,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19904747987166047,
      "step": 750
    },
    {
      "epoch": 0.019,
      "grad_norm": 6106906624.0,
      "grad_norm_var": 2.3308942582349476e+18,
      "learning_rate": 0.0001,
      "loss": 7.4633,
      "loss/crossentropy": 2.248567137122154,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19723597317934036,
      "step": 760
    },
    {
      "epoch": 0.01925,
      "grad_norm": 28.5,
      "grad_norm_var": 2.330894258158611e+18,
      "learning_rate": 0.0001,
      "loss": 7.4542,
      "loss/crossentropy": 2.132212319970131,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18174959290772677,
      "step": 770
    },
    {
      "epoch": 0.0195,
      "grad_norm": 36.5,
      "grad_norm_var": 4.833333333333333,
      "learning_rate": 0.0001,
      "loss": 7.465,
      "loss/crossentropy": 2.046277052164078,
      "loss/hidden": 3.491015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21161840241402388,
      "step": 780
    },
    {
      "epoch": 0.01975,
      "grad_norm": 32.75,
      "grad_norm_var": 5.137434895833334,
      "learning_rate": 0.0001,
      "loss": 7.4171,
      "loss/crossentropy": 2.058088332414627,
      "loss/hidden": 3.315234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1815673651173711,
      "step": 790
    },
    {
      "epoch": 0.02,
      "grad_norm": 30.125,
      "grad_norm_var": 12.37265625,
      "learning_rate": 0.0001,
      "loss": 7.4153,
      "loss/crossentropy": 2.064726157486439,
      "loss/hidden": 3.515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19402222614735365,
      "step": 800
    },
    {
      "epoch": 0.02025,
      "grad_norm": 32.0,
      "grad_norm_var": 12.240625,
      "learning_rate": 0.0001,
      "loss": 7.3739,
      "loss/crossentropy": 2.0926051691174505,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21017331834882497,
      "step": 810
    },
    {
      "epoch": 0.0205,
      "grad_norm": 31.875,
      "grad_norm_var": 3.6853515625,
      "learning_rate": 0.0001,
      "loss": 7.409,
      "loss/crossentropy": 2.016859006881714,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20363395065069198,
      "step": 820
    },
    {
      "epoch": 0.02075,
      "grad_norm": 34.0,
      "grad_norm_var": 278.1108723958333,
      "learning_rate": 0.0001,
      "loss": 7.6725,
      "loss/crossentropy": 2.03957434669137,
      "loss/hidden": 3.4625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19866096526384353,
      "step": 830
    },
    {
      "epoch": 0.021,
      "grad_norm": 35.75,
      "grad_norm_var": 281.2239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4058,
      "loss/crossentropy": 2.1190530106425287,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19663097113370895,
      "step": 840
    },
    {
      "epoch": 0.02125,
      "grad_norm": 32.25,
      "grad_norm_var": 4.044791666666667,
      "learning_rate": 0.0001,
      "loss": 7.4687,
      "loss/crossentropy": 2.1552326917648315,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19604418501257898,
      "step": 850
    },
    {
      "epoch": 0.0215,
      "grad_norm": 37.25,
      "grad_norm_var": 2.7587362193217157e+18,
      "learning_rate": 0.0001,
      "loss": 7.5552,
      "loss/crossentropy": 2.1164004117250443,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19724889248609542,
      "step": 860
    },
    {
      "epoch": 0.02175,
      "grad_norm": 35.25,
      "grad_norm_var": 2.758736219342478e+18,
      "learning_rate": 0.0001,
      "loss": 7.5021,
      "loss/crossentropy": 2.036998500674963,
      "loss/hidden": 3.298828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18320635841228067,
      "step": 870
    },
    {
      "epoch": 0.022,
      "grad_norm": 37.0,
      "grad_norm_var": 16.9541015625,
      "learning_rate": 0.0001,
      "loss": 7.5059,
      "loss/crossentropy": 1.9707016140222549,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20436920877546072,
      "step": 880
    },
    {
      "epoch": 0.02225,
      "grad_norm": 31.375,
      "grad_norm_var": 30.538541666666667,
      "learning_rate": 0.0001,
      "loss": 7.4935,
      "loss/crossentropy": 2.206394499540329,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20495780408382416,
      "step": 890
    },
    {
      "epoch": 0.0225,
      "grad_norm": 29.875,
      "grad_norm_var": 28.020833333333332,
      "learning_rate": 0.0001,
      "loss": 7.4823,
      "loss/crossentropy": 2.091763325035572,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20592593550682067,
      "step": 900
    },
    {
      "epoch": 0.02275,
      "grad_norm": 31.875,
      "grad_norm_var": 3.5645182291666666,
      "learning_rate": 0.0001,
      "loss": 7.422,
      "loss/crossentropy": 1.9740761511027813,
      "loss/hidden": 3.494921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2015986293554306,
      "step": 910
    },
    {
      "epoch": 0.023,
      "grad_norm": 32.0,
      "grad_norm_var": 56.256184895833336,
      "learning_rate": 0.0001,
      "loss": 7.4528,
      "loss/crossentropy": 2.030415116250515,
      "loss/hidden": 3.205078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1614784031175077,
      "step": 920
    },
    {
      "epoch": 0.02325,
      "grad_norm": 30.0,
      "grad_norm_var": 57.1619140625,
      "learning_rate": 0.0001,
      "loss": 7.3713,
      "loss/crossentropy": 2.0250086903572084,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19023355115205048,
      "step": 930
    },
    {
      "epoch": 0.0235,
      "grad_norm": 30.625,
      "grad_norm_var": 1.3830729166666667,
      "learning_rate": 0.0001,
      "loss": 7.5277,
      "loss/crossentropy": 2.222324788570404,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19078677501529456,
      "step": 940
    },
    {
      "epoch": 0.02375,
      "grad_norm": 31.0,
      "grad_norm_var": 3.1455729166666666,
      "learning_rate": 0.0001,
      "loss": 7.5086,
      "loss/crossentropy": 2.1299516543745995,
      "loss/hidden": 3.49921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21310927756130696,
      "step": 950
    },
    {
      "epoch": 0.024,
      "grad_norm": 29.875,
      "grad_norm_var": 8.883072916666666,
      "learning_rate": 0.0001,
      "loss": 7.5579,
      "loss/crossentropy": 2.0535727672278883,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18507701791822911,
      "step": 960
    },
    {
      "epoch": 0.02425,
      "grad_norm": 32.75,
      "grad_norm_var": 2.5916015625,
      "learning_rate": 0.0001,
      "loss": 7.537,
      "loss/crossentropy": 2.1785535484552385,
      "loss/hidden": 3.309765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1955953363329172,
      "step": 970
    },
    {
      "epoch": 0.0245,
      "grad_norm": 36.5,
      "grad_norm_var": 6.852083333333334,
      "learning_rate": 0.0001,
      "loss": 7.5091,
      "loss/crossentropy": 2.0967498391866686,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2146583067253232,
      "step": 980
    },
    {
      "epoch": 0.02475,
      "grad_norm": 29.625,
      "grad_norm_var": 4.325455729166666,
      "learning_rate": 0.0001,
      "loss": 7.5901,
      "loss/crossentropy": 2.1134474128484726,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19056662563234567,
      "step": 990
    },
    {
      "epoch": 0.025,
      "grad_norm": 42.0,
      "grad_norm_var": 4.1552039405313587e+18,
      "learning_rate": 0.0001,
      "loss": 7.6082,
      "loss/crossentropy": 2.0916516482830048,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19376826155930757,
      "step": 1000
    },
    {
      "epoch": 0.02525,
      "grad_norm": 29.625,
      "grad_norm_var": 4.1552039416015355e+18,
      "learning_rate": 0.0001,
      "loss": 7.4528,
      "loss/crossentropy": 2.003750593960285,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18129821103066207,
      "step": 1010
    },
    {
      "epoch": 0.0255,
      "grad_norm": 35.25,
      "grad_norm_var": 24.095572916666665,
      "learning_rate": 0.0001,
      "loss": 7.5395,
      "loss/crossentropy": 2.0453194856643675,
      "loss/hidden": 3.477734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.199107607267797,
      "step": 1020
    },
    {
      "epoch": 0.02575,
      "grad_norm": 32.25,
      "grad_norm_var": 19.5259765625,
      "learning_rate": 0.0001,
      "loss": 7.31,
      "loss/crossentropy": 2.1016619503498077,
      "loss/hidden": 3.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.184703135676682,
      "step": 1030
    },
    {
      "epoch": 0.026,
      "grad_norm": 30.75,
      "grad_norm_var": 1.87890625,
      "learning_rate": 0.0001,
      "loss": 7.5425,
      "loss/crossentropy": 2.1467826470732687,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20074132941663264,
      "step": 1040
    },
    {
      "epoch": 0.02625,
      "grad_norm": 30.625,
      "grad_norm_var": 0.7452473958333333,
      "learning_rate": 0.0001,
      "loss": 7.4114,
      "loss/crossentropy": 2.049474111199379,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20267941821366547,
      "step": 1050
    },
    {
      "epoch": 0.0265,
      "grad_norm": 31.75,
      "grad_norm_var": 3.124739583333333,
      "learning_rate": 0.0001,
      "loss": 7.4845,
      "loss/crossentropy": 2.036583887040615,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1893632340244949,
      "step": 1060
    },
    {
      "epoch": 0.02675,
      "grad_norm": 40.75,
      "grad_norm_var": 3.405847188209664e+18,
      "learning_rate": 0.0001,
      "loss": 7.3982,
      "loss/crossentropy": 2.124411530792713,
      "loss/hidden": 3.4484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19454579129815103,
      "step": 1070
    },
    {
      "epoch": 0.027,
      "grad_norm": 28.25,
      "grad_norm_var": 3.4058471885941417e+18,
      "learning_rate": 0.0001,
      "loss": 7.3928,
      "loss/crossentropy": 2.0034691862761975,
      "loss/hidden": 3.503515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21349683087319135,
      "step": 1080
    },
    {
      "epoch": 0.02725,
      "grad_norm": 29.875,
      "grad_norm_var": 4.88515625,
      "learning_rate": 0.0001,
      "loss": 7.5095,
      "loss/crossentropy": 1.9183670297265052,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19249978363513948,
      "step": 1090
    },
    {
      "epoch": 0.0275,
      "grad_norm": 30.5,
      "grad_norm_var": 3.2728515625,
      "learning_rate": 0.0001,
      "loss": 7.37,
      "loss/crossentropy": 2.145428071916103,
      "loss/hidden": 3.35703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19729665387421846,
      "step": 1100
    },
    {
      "epoch": 0.02775,
      "grad_norm": 31.25,
      "grad_norm_var": 2.34765625,
      "learning_rate": 0.0001,
      "loss": 7.4772,
      "loss/crossentropy": 2.10652961358428,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19585925145074726,
      "step": 1110
    },
    {
      "epoch": 0.028,
      "grad_norm": 31.25,
      "grad_norm_var": 2.434477049308093e+18,
      "learning_rate": 0.0001,
      "loss": 7.4016,
      "loss/crossentropy": 1.9645449101924897,
      "loss/hidden": 3.44453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19977953620254993,
      "step": 1120
    },
    {
      "epoch": 0.02825,
      "grad_norm": 32.0,
      "grad_norm_var": 2.4344770492950907e+18,
      "learning_rate": 0.0001,
      "loss": 7.4453,
      "loss/crossentropy": 2.131172102689743,
      "loss/hidden": 3.383984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2083016105927527,
      "step": 1130
    },
    {
      "epoch": 0.0285,
      "grad_norm": 32.75,
      "grad_norm_var": 3.7080729166666666,
      "learning_rate": 0.0001,
      "loss": 7.4009,
      "loss/crossentropy": 2.003016713261604,
      "loss/hidden": 3.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18665643623098732,
      "step": 1140
    },
    {
      "epoch": 0.02875,
      "grad_norm": 30.875,
      "grad_norm_var": 1.34765625,
      "learning_rate": 0.0001,
      "loss": 7.5648,
      "loss/crossentropy": 2.0709651306271555,
      "loss/hidden": 3.45703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18793081305921078,
      "step": 1150
    },
    {
      "epoch": 0.029,
      "grad_norm": 32.25,
      "grad_norm_var": 2.1582682291666666,
      "learning_rate": 0.0001,
      "loss": 7.4644,
      "loss/crossentropy": 2.06434089243412,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2109043262898922,
      "step": 1160
    },
    {
      "epoch": 0.02925,
      "grad_norm": 31.375,
      "grad_norm_var": 2.4010416666666665,
      "learning_rate": 0.0001,
      "loss": 7.4403,
      "loss/crossentropy": 2.0107607185840606,
      "loss/hidden": 3.498046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20349722560495137,
      "step": 1170
    },
    {
      "epoch": 0.0295,
      "grad_norm": 33.25,
      "grad_norm_var": 1.2260416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4412,
      "loss/crossentropy": 2.096436749398708,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20087064132094384,
      "step": 1180
    },
    {
      "epoch": 0.02975,
      "grad_norm": 29.75,
      "grad_norm_var": 1.8046223958333334,
      "learning_rate": 0.0001,
      "loss": 7.4458,
      "loss/crossentropy": 1.972258360683918,
      "loss/hidden": 3.583984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20998958311975002,
      "step": 1190
    },
    {
      "epoch": 0.03,
      "grad_norm": 33.75,
      "grad_norm_var": 3.7395833333333335,
      "learning_rate": 0.0001,
      "loss": 7.3931,
      "loss/crossentropy": 1.8556599006056786,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19810242671519518,
      "step": 1200
    },
    {
      "epoch": 0.03025,
      "grad_norm": 29.0,
      "grad_norm_var": 9.394791666666666,
      "learning_rate": 0.0001,
      "loss": 7.5849,
      "loss/crossentropy": 2.0611833460628985,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19216072149574756,
      "step": 1210
    },
    {
      "epoch": 0.0305,
      "grad_norm": 31.75,
      "grad_norm_var": 3.26640625,
      "learning_rate": 0.0001,
      "loss": 7.4844,
      "loss/crossentropy": 2.0546294137835504,
      "loss/hidden": 3.58828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21588555499911308,
      "step": 1220
    },
    {
      "epoch": 0.03075,
      "grad_norm": 31.625,
      "grad_norm_var": 2.3968098958333335,
      "learning_rate": 0.0001,
      "loss": 7.4858,
      "loss/crossentropy": 2.0615282475948336,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.206529095210135,
      "step": 1230
    },
    {
      "epoch": 0.031,
      "grad_norm": 32.0,
      "grad_norm_var": 1.6124348958333334,
      "learning_rate": 0.0001,
      "loss": 7.4647,
      "loss/crossentropy": 1.9786661133170127,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17899234425276517,
      "step": 1240
    },
    {
      "epoch": 0.03125,
      "grad_norm": 5838471168.0,
      "grad_norm_var": 2.1304840753447437e+18,
      "learning_rate": 0.0001,
      "loss": 7.4926,
      "loss/crossentropy": 2.04936410933733,
      "loss/hidden": 3.714453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1995564555749297,
      "step": 1250
    },
    {
      "epoch": 0.0315,
      "grad_norm": 31.25,
      "grad_norm_var": 2.1304840747304878e+18,
      "learning_rate": 0.0001,
      "loss": 7.5078,
      "loss/crossentropy": 2.1189576953649523,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19967459067702292,
      "step": 1260
    },
    {
      "epoch": 0.03175,
      "grad_norm": 30.5,
      "grad_norm_var": 3.178580729166667,
      "learning_rate": 0.0001,
      "loss": 7.4255,
      "loss/crossentropy": 2.163596141338348,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19321363251656293,
      "step": 1270
    },
    {
      "epoch": 0.032,
      "grad_norm": 33.25,
      "grad_norm_var": 2.1639973958333334,
      "learning_rate": 0.0001,
      "loss": 7.4609,
      "loss/crossentropy": 1.9938266813755035,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18334759529680014,
      "step": 1280
    },
    {
      "epoch": 0.03225,
      "grad_norm": 29.375,
      "grad_norm_var": 1.67890625,
      "learning_rate": 0.0001,
      "loss": 7.4652,
      "loss/crossentropy": 2.161333967000246,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19740422032773494,
      "step": 1290
    },
    {
      "epoch": 0.0325,
      "grad_norm": 32.75,
      "grad_norm_var": 3.0385416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3146,
      "loss/crossentropy": 2.0165325723588468,
      "loss/hidden": 3.49921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19117104820907116,
      "step": 1300
    },
    {
      "epoch": 0.03275,
      "grad_norm": 28.25,
      "grad_norm_var": 9.158072916666667,
      "learning_rate": 0.0001,
      "loss": 7.4955,
      "loss/crossentropy": 2.124955786764622,
      "loss/hidden": 3.491015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19802952595055104,
      "step": 1310
    },
    {
      "epoch": 0.033,
      "grad_norm": 30.75,
      "grad_norm_var": 2.4535807291666667,
      "learning_rate": 0.0001,
      "loss": 7.4311,
      "loss/crossentropy": 2.018800371140242,
      "loss/hidden": 3.542578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2196814114227891,
      "step": 1320
    },
    {
      "epoch": 0.03325,
      "grad_norm": 31.375,
      "grad_norm_var": 2.39375,
      "learning_rate": 0.0001,
      "loss": 7.5164,
      "loss/crossentropy": 2.0520452961325644,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2013697015121579,
      "step": 1330
    },
    {
      "epoch": 0.0335,
      "grad_norm": 32.5,
      "grad_norm_var": 1.0431640625,
      "learning_rate": 0.0001,
      "loss": 7.5302,
      "loss/crossentropy": 2.12932348549366,
      "loss/hidden": 3.525,
      "loss/jsd": 0.0,
      "loss/logits": 0.20245677568018436,
      "step": 1340
    },
    {
      "epoch": 0.03375,
      "grad_norm": 30.625,
      "grad_norm_var": 3.3900390625,
      "learning_rate": 0.0001,
      "loss": 7.5292,
      "loss/crossentropy": 2.031618994474411,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19062725063413383,
      "step": 1350
    },
    {
      "epoch": 0.034,
      "grad_norm": 32.0,
      "grad_norm_var": 3.3447265625,
      "learning_rate": 0.0001,
      "loss": 7.5755,
      "loss/crossentropy": 2.2257011234760284,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1979327043518424,
      "step": 1360
    },
    {
      "epoch": 0.03425,
      "grad_norm": 30.625,
      "grad_norm_var": 3.3421223958333335,
      "learning_rate": 0.0001,
      "loss": 7.4219,
      "loss/crossentropy": 2.155778780579567,
      "loss/hidden": 3.31796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19018295016139747,
      "step": 1370
    },
    {
      "epoch": 0.0345,
      "grad_norm": 30.25,
      "grad_norm_var": 2.5872395833333335,
      "learning_rate": 0.0001,
      "loss": 7.4637,
      "loss/crossentropy": 2.058405503630638,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2114524593576789,
      "step": 1380
    },
    {
      "epoch": 0.03475,
      "grad_norm": 32.5,
      "grad_norm_var": 3.2994140625,
      "learning_rate": 0.0001,
      "loss": 7.5834,
      "loss/crossentropy": 2.1654782712459566,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2024593001231551,
      "step": 1390
    },
    {
      "epoch": 0.035,
      "grad_norm": 31.125,
      "grad_norm_var": 12.812239583333334,
      "learning_rate": 0.0001,
      "loss": 7.4442,
      "loss/crossentropy": 2.0921876966953277,
      "loss/hidden": 3.286328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19270132519304753,
      "step": 1400
    },
    {
      "epoch": 0.03525,
      "grad_norm": 29.25,
      "grad_norm_var": 1.5108723958333334,
      "learning_rate": 0.0001,
      "loss": 7.4779,
      "loss/crossentropy": 1.9434148371219635,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17576389852911234,
      "step": 1410
    },
    {
      "epoch": 0.0355,
      "grad_norm": 30.125,
      "grad_norm_var": 2.154166666666667,
      "learning_rate": 0.0001,
      "loss": 7.508,
      "loss/crossentropy": 2.0766889482736586,
      "loss/hidden": 3.485546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20394362770020963,
      "step": 1420
    },
    {
      "epoch": 0.03575,
      "grad_norm": 30.125,
      "grad_norm_var": 17.580208333333335,
      "learning_rate": 0.0001,
      "loss": 7.4612,
      "loss/crossentropy": 2.00380075648427,
      "loss/hidden": 3.4453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18816210143268108,
      "step": 1430
    },
    {
      "epoch": 0.036,
      "grad_norm": 31.375,
      "grad_norm_var": 16.758268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4602,
      "loss/crossentropy": 2.1938020154833793,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2016971528530121,
      "step": 1440
    },
    {
      "epoch": 0.03625,
      "grad_norm": 30.875,
      "grad_norm_var": 1.2556640625,
      "learning_rate": 0.0001,
      "loss": 7.4245,
      "loss/crossentropy": 2.0232372283935547,
      "loss/hidden": 3.40234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19209201391786337,
      "step": 1450
    },
    {
      "epoch": 0.0365,
      "grad_norm": 31.0,
      "grad_norm_var": 1.4041015625,
      "learning_rate": 0.0001,
      "loss": 7.5518,
      "loss/crossentropy": 2.2000616788864136,
      "loss/hidden": 3.473046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22938326951116322,
      "step": 1460
    },
    {
      "epoch": 0.03675,
      "grad_norm": 28.375,
      "grad_norm_var": 2.0322916666666666,
      "learning_rate": 0.0001,
      "loss": 7.4397,
      "loss/crossentropy": 2.0838582158088683,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20685861641541123,
      "step": 1470
    },
    {
      "epoch": 0.037,
      "grad_norm": 32.0,
      "grad_norm_var": 1.5020833333333334,
      "learning_rate": 0.0001,
      "loss": 7.4183,
      "loss/crossentropy": 2.149951633810997,
      "loss/hidden": 3.375390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1984950641170144,
      "step": 1480
    },
    {
      "epoch": 0.03725,
      "grad_norm": 33.75,
      "grad_norm_var": 34.10826822916667,
      "learning_rate": 0.0001,
      "loss": 7.453,
      "loss/crossentropy": 2.128306310623884,
      "loss/hidden": 3.33203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19783397912979125,
      "step": 1490
    },
    {
      "epoch": 0.0375,
      "grad_norm": 29.5,
      "grad_norm_var": 5.008072916666666,
      "learning_rate": 0.0001,
      "loss": 7.469,
      "loss/crossentropy": 2.042660539597273,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19274956732988358,
      "step": 1500
    },
    {
      "epoch": 0.03775,
      "grad_norm": 33.0,
      "grad_norm_var": 19.1775390625,
      "learning_rate": 0.0001,
      "loss": 7.4119,
      "loss/crossentropy": 2.043857058137655,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18266947232186795,
      "step": 1510
    },
    {
      "epoch": 0.038,
      "grad_norm": 29.625,
      "grad_norm_var": 14.303580729166667,
      "learning_rate": 0.0001,
      "loss": 7.4362,
      "loss/crossentropy": 1.9492302805185318,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1754497304558754,
      "step": 1520
    },
    {
      "epoch": 0.03825,
      "grad_norm": 29.75,
      "grad_norm_var": 23.764518229166665,
      "learning_rate": 0.0001,
      "loss": 7.4444,
      "loss/crossentropy": 2.0668226674199106,
      "loss/hidden": 3.473828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1921279976144433,
      "step": 1530
    },
    {
      "epoch": 0.0385,
      "grad_norm": 32.75,
      "grad_norm_var": 3.2226069790467994e+18,
      "learning_rate": 0.0001,
      "loss": 7.5077,
      "loss/crossentropy": 2.1122784771025183,
      "loss/hidden": 3.46953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22245875597000123,
      "step": 1540
    },
    {
      "epoch": 0.03875,
      "grad_norm": 30.25,
      "grad_norm_var": 5.382291666666666,
      "learning_rate": 0.0001,
      "loss": 7.4525,
      "loss/crossentropy": 2.264697426557541,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2075907403603196,
      "step": 1550
    },
    {
      "epoch": 0.039,
      "grad_norm": 30.0,
      "grad_norm_var": 6.353580729166667,
      "learning_rate": 0.0001,
      "loss": 7.5064,
      "loss/crossentropy": 2.1150408178567885,
      "loss/hidden": 3.5203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23207673486322164,
      "step": 1560
    },
    {
      "epoch": 0.03925,
      "grad_norm": 34.25,
      "grad_norm_var": 6.72265625,
      "learning_rate": 0.0001,
      "loss": 7.4578,
      "loss/crossentropy": 2.188142140209675,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20429779235273599,
      "step": 1570
    },
    {
      "epoch": 0.0395,
      "grad_norm": 34.75,
      "grad_norm_var": 897.6666015625,
      "learning_rate": 0.0001,
      "loss": 7.434,
      "loss/crossentropy": 2.0795677445828913,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18706642352044583,
      "step": 1580
    },
    {
      "epoch": 0.03975,
      "grad_norm": 28.0,
      "grad_norm_var": 903.6327473958333,
      "learning_rate": 0.0001,
      "loss": 7.5655,
      "loss/crossentropy": 2.1025844663381577,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1966788914054632,
      "step": 1590
    },
    {
      "epoch": 0.04,
      "grad_norm": 28.625,
      "grad_norm_var": 11.97890625,
      "learning_rate": 0.0001,
      "loss": 7.2578,
      "loss/crossentropy": 2.050418493151665,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20104087069630622,
      "step": 1600
    },
    {
      "epoch": 0.04025,
      "grad_norm": 28.0,
      "grad_norm_var": 2.255989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4393,
      "loss/crossentropy": 2.1767756581306457,
      "loss/hidden": 3.5140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2213939843699336,
      "step": 1610
    },
    {
      "epoch": 0.0405,
      "grad_norm": 29.75,
      "grad_norm_var": 3.80390625,
      "learning_rate": 0.0001,
      "loss": 7.5026,
      "loss/crossentropy": 2.126803469657898,
      "loss/hidden": 3.39375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19106289148330688,
      "step": 1620
    },
    {
      "epoch": 0.04075,
      "grad_norm": 32.0,
      "grad_norm_var": 3.1249348958333334,
      "learning_rate": 0.0001,
      "loss": 7.4274,
      "loss/crossentropy": 2.144256164133549,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21435861438512802,
      "step": 1630
    },
    {
      "epoch": 0.041,
      "grad_norm": 30.25,
      "grad_norm_var": 29.265559895833334,
      "learning_rate": 0.0001,
      "loss": 7.5728,
      "loss/crossentropy": 2.2575725719332693,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20658138059079648,
      "step": 1640
    },
    {
      "epoch": 0.04125,
      "grad_norm": 30.5,
      "grad_norm_var": 48.35390625,
      "learning_rate": 0.0001,
      "loss": 7.5776,
      "loss/crossentropy": 2.096929042041302,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18803389491513373,
      "step": 1650
    },
    {
      "epoch": 0.0415,
      "grad_norm": 30.5,
      "grad_norm_var": 1.1010416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3792,
      "loss/crossentropy": 2.0290944524109364,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19023821037262678,
      "step": 1660
    },
    {
      "epoch": 0.04175,
      "grad_norm": 28.125,
      "grad_norm_var": 33.49270833333333,
      "learning_rate": 0.0001,
      "loss": 7.5018,
      "loss/crossentropy": 2.0678361281752586,
      "loss/hidden": 3.35234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18862500675022603,
      "step": 1670
    },
    {
      "epoch": 0.042,
      "grad_norm": 29.75,
      "grad_norm_var": 2.2955729166666665,
      "learning_rate": 0.0001,
      "loss": 7.4432,
      "loss/crossentropy": 2.0549797296524046,
      "loss/hidden": 3.441796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19089050237089394,
      "step": 1680
    },
    {
      "epoch": 0.04225,
      "grad_norm": 29.75,
      "grad_norm_var": 1.8791666666666667,
      "learning_rate": 0.0001,
      "loss": 7.3842,
      "loss/crossentropy": 2.0077505365014074,
      "loss/hidden": 3.382421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18722779098898173,
      "step": 1690
    },
    {
      "epoch": 0.0425,
      "grad_norm": 29.375,
      "grad_norm_var": 0.9434895833333333,
      "learning_rate": 0.0001,
      "loss": 7.4273,
      "loss/crossentropy": 2.071325332671404,
      "loss/hidden": 3.486328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20270166713744403,
      "step": 1700
    },
    {
      "epoch": 0.04275,
      "grad_norm": 38.25,
      "grad_norm_var": 7.669791666666667,
      "learning_rate": 0.0001,
      "loss": 7.4176,
      "loss/crossentropy": 2.1353142291307448,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19663168713450432,
      "step": 1710
    },
    {
      "epoch": 0.043,
      "grad_norm": 28.25,
      "grad_norm_var": 7.75,
      "learning_rate": 0.0001,
      "loss": 7.3818,
      "loss/crossentropy": 1.9995346069335938,
      "loss/hidden": 3.41015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18310597026720643,
      "step": 1720
    },
    {
      "epoch": 0.04325,
      "grad_norm": 29.5,
      "grad_norm_var": 3.7619140625,
      "learning_rate": 0.0001,
      "loss": 7.4912,
      "loss/crossentropy": 2.1415088951587675,
      "loss/hidden": 3.55078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22313783299177886,
      "step": 1730
    },
    {
      "epoch": 0.0435,
      "grad_norm": 31.625,
      "grad_norm_var": 3.0416666666666665,
      "learning_rate": 0.0001,
      "loss": 7.4999,
      "loss/crossentropy": 2.1686330527067184,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20409150077030064,
      "step": 1740
    },
    {
      "epoch": 0.04375,
      "grad_norm": 31.375,
      "grad_norm_var": 2.724739583333333,
      "learning_rate": 0.0001,
      "loss": 7.438,
      "loss/crossentropy": 1.9411263287067413,
      "loss/hidden": 3.304296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17631518254056572,
      "step": 1750
    },
    {
      "epoch": 0.044,
      "grad_norm": 32.0,
      "grad_norm_var": 1.9145833333333333,
      "learning_rate": 0.0001,
      "loss": 7.679,
      "loss/crossentropy": 2.1614590853452684,
      "loss/hidden": 3.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.194198589771986,
      "step": 1760
    },
    {
      "epoch": 0.04425,
      "grad_norm": 28.5,
      "grad_norm_var": 2.039322916666667,
      "learning_rate": 0.0001,
      "loss": 7.5095,
      "loss/crossentropy": 2.282147654891014,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19978236705064772,
      "step": 1770
    },
    {
      "epoch": 0.0445,
      "grad_norm": 29.625,
      "grad_norm_var": 2.34140625,
      "learning_rate": 0.0001,
      "loss": 7.5296,
      "loss/crossentropy": 2.2078514605760575,
      "loss/hidden": 3.403515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19668537452816964,
      "step": 1780
    },
    {
      "epoch": 0.04475,
      "grad_norm": 30.25,
      "grad_norm_var": 2.70390625,
      "learning_rate": 0.0001,
      "loss": 7.5779,
      "loss/crossentropy": 2.1053253799676894,
      "loss/hidden": 3.433984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20323336366564035,
      "step": 1790
    },
    {
      "epoch": 0.045,
      "grad_norm": 28.5,
      "grad_norm_var": 4.8712890625,
      "learning_rate": 0.0001,
      "loss": 7.4866,
      "loss/crossentropy": 2.060333488881588,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18627767637372017,
      "step": 1800
    },
    {
      "epoch": 0.04525,
      "grad_norm": 28.0,
      "grad_norm_var": 14.480989583333333,
      "learning_rate": 0.0001,
      "loss": 7.5225,
      "loss/crossentropy": 1.9755317773669958,
      "loss/hidden": 3.54375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20334282671101392,
      "step": 1810
    },
    {
      "epoch": 0.0455,
      "grad_norm": 29.875,
      "grad_norm_var": 12.935872395833334,
      "learning_rate": 0.0001,
      "loss": 7.4781,
      "loss/crossentropy": 2.1289859026670457,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1973018018528819,
      "step": 1820
    },
    {
      "epoch": 0.04575,
      "grad_norm": 31.75,
      "grad_norm_var": 2.123893229166667,
      "learning_rate": 0.0001,
      "loss": 7.3915,
      "loss/crossentropy": 1.9609280914068221,
      "loss/hidden": 3.386328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1916458262130618,
      "step": 1830
    },
    {
      "epoch": 0.046,
      "grad_norm": 32.0,
      "grad_norm_var": 1.6332682291666667,
      "learning_rate": 0.0001,
      "loss": 7.5095,
      "loss/crossentropy": 2.0019985377788543,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19768325993791222,
      "step": 1840
    },
    {
      "epoch": 0.04625,
      "grad_norm": 29.875,
      "grad_norm_var": 2.225455729166667,
      "learning_rate": 0.0001,
      "loss": 7.623,
      "loss/crossentropy": 2.0607564479112623,
      "loss/hidden": 3.507421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20858939345926047,
      "step": 1850
    },
    {
      "epoch": 0.0465,
      "grad_norm": 29.5,
      "grad_norm_var": 1.9863932291666666,
      "learning_rate": 0.0001,
      "loss": 7.3836,
      "loss/crossentropy": 2.132562433928251,
      "loss/hidden": 3.40859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1956317812204361,
      "step": 1860
    },
    {
      "epoch": 0.04675,
      "grad_norm": 36.0,
      "grad_norm_var": 3.2171223958333335,
      "learning_rate": 0.0001,
      "loss": 7.4803,
      "loss/crossentropy": 2.0316790327429772,
      "loss/hidden": 3.396875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20630075875669718,
      "step": 1870
    },
    {
      "epoch": 0.047,
      "grad_norm": 33.25,
      "grad_norm_var": 16.304622395833334,
      "learning_rate": 0.0001,
      "loss": 7.576,
      "loss/crossentropy": 2.161964085698128,
      "loss/hidden": 3.513671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21842746511101724,
      "step": 1880
    },
    {
      "epoch": 0.04725,
      "grad_norm": 29.75,
      "grad_norm_var": 2.3541666666666665,
      "learning_rate": 0.0001,
      "loss": 7.5036,
      "loss/crossentropy": 1.8695943117141725,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18793469872325658,
      "step": 1890
    },
    {
      "epoch": 0.0475,
      "grad_norm": 34.25,
      "grad_norm_var": 2.1780598958333335,
      "learning_rate": 0.0001,
      "loss": 7.5623,
      "loss/crossentropy": 2.2376974314451217,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21696731727570295,
      "step": 1900
    },
    {
      "epoch": 0.04775,
      "grad_norm": 30.75,
      "grad_norm_var": 14.924934895833333,
      "learning_rate": 0.0001,
      "loss": 7.388,
      "loss/crossentropy": 1.9403380863368511,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18128401823341847,
      "step": 1910
    },
    {
      "epoch": 0.048,
      "grad_norm": 29.25,
      "grad_norm_var": 25.1916015625,
      "learning_rate": 0.0001,
      "loss": 7.4109,
      "loss/crossentropy": 2.1744547933340073,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20097011709585785,
      "step": 1920
    },
    {
      "epoch": 0.04825,
      "grad_norm": 29.25,
      "grad_norm_var": 14.801822916666667,
      "learning_rate": 0.0001,
      "loss": 7.2893,
      "loss/crossentropy": 2.101319019496441,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1921493023633957,
      "step": 1930
    },
    {
      "epoch": 0.0485,
      "grad_norm": 30.125,
      "grad_norm_var": 14.517708333333333,
      "learning_rate": 0.0001,
      "loss": 7.579,
      "loss/crossentropy": 2.057158224284649,
      "loss/hidden": 3.59140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21765361074358225,
      "step": 1940
    },
    {
      "epoch": 0.04875,
      "grad_norm": 29.625,
      "grad_norm_var": 15.790559895833333,
      "learning_rate": 0.0001,
      "loss": 7.3712,
      "loss/crossentropy": 1.9415803879499436,
      "loss/hidden": 3.3359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18346730088815094,
      "step": 1950
    },
    {
      "epoch": 0.049,
      "grad_norm": 27.625,
      "grad_norm_var": 9.794791666666667,
      "learning_rate": 0.0001,
      "loss": 7.4902,
      "loss/crossentropy": 2.035348242521286,
      "loss/hidden": 3.439453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20268035624176264,
      "step": 1960
    },
    {
      "epoch": 0.04925,
      "grad_norm": 35.25,
      "grad_norm_var": 12.768684895833333,
      "learning_rate": 0.0001,
      "loss": 7.4627,
      "loss/crossentropy": 2.054542076587677,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2003987120464444,
      "step": 1970
    },
    {
      "epoch": 0.0495,
      "grad_norm": 36.0,
      "grad_norm_var": 12.572916666666666,
      "learning_rate": 0.0001,
      "loss": 7.353,
      "loss/crossentropy": 1.9634785205125809,
      "loss/hidden": 3.301171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17985089337453247,
      "step": 1980
    },
    {
      "epoch": 0.04975,
      "grad_norm": 36.25,
      "grad_norm_var": 9.2353515625,
      "learning_rate": 0.0001,
      "loss": 7.4473,
      "loss/crossentropy": 2.059533824026585,
      "loss/hidden": 3.3578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19096513148397207,
      "step": 1990
    },
    {
      "epoch": 0.05,
      "grad_norm": 29.125,
      "grad_norm_var": 13.320572916666666,
      "learning_rate": 0.0001,
      "loss": 7.3914,
      "loss/crossentropy": 2.011685383319855,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19188414234668016,
      "step": 2000
    },
    {
      "epoch": 0.05025,
      "grad_norm": 36.25,
      "grad_norm_var": 14.026822916666667,
      "learning_rate": 0.0001,
      "loss": 7.4213,
      "loss/crossentropy": 2.309766414761543,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20372038893401623,
      "step": 2010
    },
    {
      "epoch": 0.0505,
      "grad_norm": 29.0,
      "grad_norm_var": 9.237239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4145,
      "loss/crossentropy": 2.1240487143397333,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20137840434908866,
      "step": 2020
    },
    {
      "epoch": 0.05075,
      "grad_norm": 38.5,
      "grad_norm_var": 89.21432291666666,
      "learning_rate": 0.0001,
      "loss": 7.3696,
      "loss/crossentropy": 2.112667274475098,
      "loss/hidden": 3.487109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19770587887614965,
      "step": 2030
    },
    {
      "epoch": 0.051,
      "grad_norm": 27.75,
      "grad_norm_var": 94.06015625,
      "learning_rate": 0.0001,
      "loss": 7.2471,
      "loss/crossentropy": 1.9955052442848682,
      "loss/hidden": 3.30546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1880181163549423,
      "step": 2040
    },
    {
      "epoch": 0.05125,
      "grad_norm": 35.25,
      "grad_norm_var": 3.67265625,
      "learning_rate": 0.0001,
      "loss": 7.458,
      "loss/crossentropy": 2.1320972844958304,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18908526431769132,
      "step": 2050
    },
    {
      "epoch": 0.0515,
      "grad_norm": 38.75,
      "grad_norm_var": 10.776822916666667,
      "learning_rate": 0.0001,
      "loss": 7.3769,
      "loss/crossentropy": 2.171598494052887,
      "loss/hidden": 3.29765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18929236195981503,
      "step": 2060
    },
    {
      "epoch": 0.05175,
      "grad_norm": 32.75,
      "grad_norm_var": 10.53515625,
      "learning_rate": 0.0001,
      "loss": 7.5279,
      "loss/crossentropy": 2.0172302186489106,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2013201082125306,
      "step": 2070
    },
    {
      "epoch": 0.052,
      "grad_norm": 32.0,
      "grad_norm_var": 7.678125,
      "learning_rate": 0.0001,
      "loss": 7.3619,
      "loss/crossentropy": 1.982726515084505,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17850281894207,
      "step": 2080
    },
    {
      "epoch": 0.05225,
      "grad_norm": 29.75,
      "grad_norm_var": 63.6681640625,
      "learning_rate": 0.0001,
      "loss": 7.5109,
      "loss/crossentropy": 2.121504098176956,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.240205854550004,
      "step": 2090
    },
    {
      "epoch": 0.0525,
      "grad_norm": 34.5,
      "grad_norm_var": 7.506184895833333,
      "learning_rate": 0.0001,
      "loss": 7.4658,
      "loss/crossentropy": 2.110687591135502,
      "loss/hidden": 3.530078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2039638390764594,
      "step": 2100
    },
    {
      "epoch": 0.05275,
      "grad_norm": 32.5,
      "grad_norm_var": 19.075455729166666,
      "learning_rate": 0.0001,
      "loss": 7.5668,
      "loss/crossentropy": 1.9557841390371322,
      "loss/hidden": 3.462109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18774209143593906,
      "step": 2110
    },
    {
      "epoch": 0.053,
      "grad_norm": 31.125,
      "grad_norm_var": 3.85390625,
      "learning_rate": 0.0001,
      "loss": 7.5735,
      "loss/crossentropy": 2.0219520531594752,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18533632289618254,
      "step": 2120
    },
    {
      "epoch": 0.05325,
      "grad_norm": 32.25,
      "grad_norm_var": 3.8910807291666667,
      "learning_rate": 0.0001,
      "loss": 7.4083,
      "loss/crossentropy": 2.1359280541539194,
      "loss/hidden": 3.412890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897095028311014,
      "step": 2130
    },
    {
      "epoch": 0.0535,
      "grad_norm": 31.25,
      "grad_norm_var": 2.5957682291666666,
      "learning_rate": 0.0001,
      "loss": 7.446,
      "loss/crossentropy": 2.170258317142725,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1826348526403308,
      "step": 2140
    },
    {
      "epoch": 0.05375,
      "grad_norm": 31.25,
      "grad_norm_var": 3.785416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4014,
      "loss/crossentropy": 2.131239393353462,
      "loss/hidden": 3.303515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18656531646847724,
      "step": 2150
    },
    {
      "epoch": 0.054,
      "grad_norm": 31.0,
      "grad_norm_var": 4.8666015625,
      "learning_rate": 0.0001,
      "loss": 7.5478,
      "loss/crossentropy": 2.223896725475788,
      "loss/hidden": 3.383203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1951376979239285,
      "step": 2160
    },
    {
      "epoch": 0.05425,
      "grad_norm": 30.375,
      "grad_norm_var": 8.437955729166667,
      "learning_rate": 0.0001,
      "loss": 7.5562,
      "loss/crossentropy": 2.1203987300395966,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1970507999882102,
      "step": 2170
    },
    {
      "epoch": 0.0545,
      "grad_norm": 32.0,
      "grad_norm_var": 2.9488932291666665,
      "learning_rate": 0.0001,
      "loss": 7.5532,
      "loss/crossentropy": 2.080265050381422,
      "loss/hidden": 3.544140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2216239819303155,
      "step": 2180
    },
    {
      "epoch": 0.05475,
      "grad_norm": 31.125,
      "grad_norm_var": 8.1728515625,
      "learning_rate": 0.0001,
      "loss": 7.382,
      "loss/crossentropy": 2.2114535331726075,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20577374435961246,
      "step": 2190
    },
    {
      "epoch": 0.055,
      "grad_norm": 28.875,
      "grad_norm_var": 14.520833333333334,
      "learning_rate": 0.0001,
      "loss": 7.5766,
      "loss/crossentropy": 2.1003271512687207,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18811229150742292,
      "step": 2200
    },
    {
      "epoch": 0.05525,
      "grad_norm": 33.5,
      "grad_norm_var": 16.099739583333335,
      "learning_rate": 0.0001,
      "loss": 7.5553,
      "loss/crossentropy": 2.1326127350330353,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22006579730659723,
      "step": 2210
    },
    {
      "epoch": 0.0555,
      "grad_norm": 32.25,
      "grad_norm_var": 9.305143229166667,
      "learning_rate": 0.0001,
      "loss": 7.3766,
      "loss/crossentropy": 2.1496046826243402,
      "loss/hidden": 3.476171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1952402491122484,
      "step": 2220
    },
    {
      "epoch": 0.05575,
      "grad_norm": 29.125,
      "grad_norm_var": 6.805143229166666,
      "learning_rate": 0.0001,
      "loss": 7.3648,
      "loss/crossentropy": 2.13938904479146,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19394674636423587,
      "step": 2230
    },
    {
      "epoch": 0.056,
      "grad_norm": 27.625,
      "grad_norm_var": 15.0712890625,
      "learning_rate": 0.0001,
      "loss": 7.4292,
      "loss/crossentropy": 2.0648645758628845,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18520106598734856,
      "step": 2240
    },
    {
      "epoch": 0.05625,
      "grad_norm": 29.25,
      "grad_norm_var": 12.034309895833333,
      "learning_rate": 0.0001,
      "loss": 7.4469,
      "loss/crossentropy": 2.080448921024799,
      "loss/hidden": 3.3109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18507405128329993,
      "step": 2250
    },
    {
      "epoch": 0.0565,
      "grad_norm": 31.375,
      "grad_norm_var": 2.014518229166667,
      "learning_rate": 0.0001,
      "loss": 7.4325,
      "loss/crossentropy": 2.0871294140815735,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20059894528239966,
      "step": 2260
    },
    {
      "epoch": 0.05675,
      "grad_norm": 28.75,
      "grad_norm_var": 1.8103515625,
      "learning_rate": 0.0001,
      "loss": 7.4268,
      "loss/crossentropy": 2.010594163835049,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19413960948586464,
      "step": 2270
    },
    {
      "epoch": 0.057,
      "grad_norm": 32.5,
      "grad_norm_var": 4.0369140625,
      "learning_rate": 0.0001,
      "loss": 7.4346,
      "loss/crossentropy": 2.1129174560308455,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1961110396310687,
      "step": 2280
    },
    {
      "epoch": 0.05725,
      "grad_norm": 39.0,
      "grad_norm_var": 30.42265625,
      "learning_rate": 0.0001,
      "loss": 7.4422,
      "loss/crossentropy": 2.002947611361742,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2081361676566303,
      "step": 2290
    },
    {
      "epoch": 0.0575,
      "grad_norm": 37.25,
      "grad_norm_var": 25.699934895833334,
      "learning_rate": 0.0001,
      "loss": 7.4312,
      "loss/crossentropy": 2.06134437918663,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18918452728539706,
      "step": 2300
    },
    {
      "epoch": 0.05775,
      "grad_norm": 28.875,
      "grad_norm_var": 9.115559895833334,
      "learning_rate": 0.0001,
      "loss": 7.4209,
      "loss/crossentropy": 2.041922479122877,
      "loss/hidden": 3.403515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20907302405685185,
      "step": 2310
    },
    {
      "epoch": 0.058,
      "grad_norm": 30.125,
      "grad_norm_var": 22.248372395833332,
      "learning_rate": 0.0001,
      "loss": 7.6844,
      "loss/crossentropy": 2.0152460247278214,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1905667196959257,
      "step": 2320
    },
    {
      "epoch": 0.05825,
      "grad_norm": 38.25,
      "grad_norm_var": 31.398893229166667,
      "learning_rate": 0.0001,
      "loss": 7.4713,
      "loss/crossentropy": 2.105386929959059,
      "loss/hidden": 3.452734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1982942834496498,
      "step": 2330
    },
    {
      "epoch": 0.0585,
      "grad_norm": 28.375,
      "grad_norm_var": 54.94264322916667,
      "learning_rate": 0.0001,
      "loss": 7.4575,
      "loss/crossentropy": 2.2358868844807147,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19232469592243434,
      "step": 2340
    },
    {
      "epoch": 0.05875,
      "grad_norm": 33.5,
      "grad_norm_var": 165.74583333333334,
      "learning_rate": 0.0001,
      "loss": 7.2987,
      "loss/crossentropy": 1.9657546751201154,
      "loss/hidden": 3.3921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18062973748892547,
      "step": 2350
    },
    {
      "epoch": 0.059,
      "grad_norm": 41.0,
      "grad_norm_var": 15.376822916666667,
      "learning_rate": 0.0001,
      "loss": 7.4431,
      "loss/crossentropy": 2.191007924079895,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2068317520432174,
      "step": 2360
    },
    {
      "epoch": 0.05925,
      "grad_norm": 30.625,
      "grad_norm_var": 12.109375,
      "learning_rate": 0.0001,
      "loss": 7.3325,
      "loss/crossentropy": 2.0140789330005644,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18166892379522323,
      "step": 2370
    },
    {
      "epoch": 0.0595,
      "grad_norm": 31.875,
      "grad_norm_var": 6.941666666666666,
      "learning_rate": 0.0001,
      "loss": 7.4039,
      "loss/crossentropy": 2.0221361994743345,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1934544663876295,
      "step": 2380
    },
    {
      "epoch": 0.05975,
      "grad_norm": 30.125,
      "grad_norm_var": 10.472330729166666,
      "learning_rate": 0.0001,
      "loss": 7.5862,
      "loss/crossentropy": 1.9840030640363693,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19178631734102963,
      "step": 2390
    },
    {
      "epoch": 0.06,
      "grad_norm": 29.875,
      "grad_norm_var": 14.10625,
      "learning_rate": 0.0001,
      "loss": 7.4826,
      "loss/crossentropy": 2.1700179904699324,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1915024297311902,
      "step": 2400
    },
    {
      "epoch": 0.06025,
      "grad_norm": 32.75,
      "grad_norm_var": 7.370768229166667,
      "learning_rate": 0.0001,
      "loss": 7.3889,
      "loss/crossentropy": 2.091843403875828,
      "loss/hidden": 3.358203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18695627991110086,
      "step": 2410
    },
    {
      "epoch": 0.0605,
      "grad_norm": 29.0,
      "grad_norm_var": 9.922330729166667,
      "learning_rate": 0.0001,
      "loss": 7.4655,
      "loss/crossentropy": 2.172381104528904,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20078962799161673,
      "step": 2420
    },
    {
      "epoch": 0.06075,
      "grad_norm": 34.25,
      "grad_norm_var": 8.637239583333333,
      "learning_rate": 0.0001,
      "loss": 7.519,
      "loss/crossentropy": 1.995463601499796,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1993358489125967,
      "step": 2430
    },
    {
      "epoch": 0.061,
      "grad_norm": 31.25,
      "grad_norm_var": 11.9431640625,
      "learning_rate": 0.0001,
      "loss": 7.5169,
      "loss/crossentropy": 2.296917426586151,
      "loss/hidden": 3.513671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23228074796497822,
      "step": 2440
    },
    {
      "epoch": 0.06125,
      "grad_norm": 30.25,
      "grad_norm_var": 3.4368798046573737e+18,
      "learning_rate": 0.0001,
      "loss": 7.5038,
      "loss/crossentropy": 2.1944432735443113,
      "loss/hidden": 3.3921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21073084995150565,
      "step": 2450
    },
    {
      "epoch": 0.0615,
      "grad_norm": 33.5,
      "grad_norm_var": 3.436879805205814e+18,
      "learning_rate": 0.0001,
      "loss": 7.4423,
      "loss/crossentropy": 2.152103579044342,
      "loss/hidden": 3.512109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20929353777319193,
      "step": 2460
    },
    {
      "epoch": 0.06175,
      "grad_norm": 39.0,
      "grad_norm_var": 2.2045823633093297e+18,
      "learning_rate": 0.0001,
      "loss": 7.4382,
      "loss/crossentropy": 2.017627691477537,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19590776292607187,
      "step": 2470
    },
    {
      "epoch": 0.062,
      "grad_norm": 29.375,
      "grad_norm_var": 2.2045823636681523e+18,
      "learning_rate": 0.0001,
      "loss": 7.4072,
      "loss/crossentropy": 2.1076912328600885,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1988623272627592,
      "step": 2480
    },
    {
      "epoch": 0.06225,
      "grad_norm": 30.125,
      "grad_norm_var": 3.2494140625,
      "learning_rate": 0.0001,
      "loss": 7.3192,
      "loss/crossentropy": 1.9777067750692368,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20539684109389783,
      "step": 2490
    },
    {
      "epoch": 0.0625,
      "grad_norm": 29.125,
      "grad_norm_var": 5.580208333333333,
      "learning_rate": 0.0001,
      "loss": 7.3283,
      "loss/crossentropy": 2.061080713570118,
      "loss/hidden": 3.4953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20077812522649766,
      "step": 2500
    },
    {
      "epoch": 0.06275,
      "grad_norm": 28.375,
      "grad_norm_var": 5.618489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4401,
      "loss/crossentropy": 2.2099071338772776,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2055276283994317,
      "step": 2510
    },
    {
      "epoch": 0.063,
      "grad_norm": 28.125,
      "grad_norm_var": 7.118684895833334,
      "learning_rate": 0.0001,
      "loss": 7.3509,
      "loss/crossentropy": 1.962952435016632,
      "loss/hidden": 3.421484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19731322024017572,
      "step": 2520
    },
    {
      "epoch": 0.06325,
      "grad_norm": 31.375,
      "grad_norm_var": 1.9681640625,
      "learning_rate": 0.0001,
      "loss": 7.3695,
      "loss/crossentropy": 1.9843583509325982,
      "loss/hidden": 3.4921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2062232268974185,
      "step": 2530
    },
    {
      "epoch": 0.0635,
      "grad_norm": 31.5,
      "grad_norm_var": 3.7988932291666666,
      "learning_rate": 0.0001,
      "loss": 7.4485,
      "loss/crossentropy": 2.1427679538726805,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2011977185495198,
      "step": 2540
    },
    {
      "epoch": 0.06375,
      "grad_norm": 30.0,
      "grad_norm_var": 2.5885416666666665,
      "learning_rate": 0.0001,
      "loss": 7.4157,
      "loss/crossentropy": 1.9085583783686162,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.17416954301297666,
      "step": 2550
    },
    {
      "epoch": 0.064,
      "grad_norm": 31.25,
      "grad_norm_var": 1.21015625,
      "learning_rate": 0.0001,
      "loss": 7.5141,
      "loss/crossentropy": 1.9622327491641045,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18756412472575903,
      "step": 2560
    },
    {
      "epoch": 0.06425,
      "grad_norm": 30.0,
      "grad_norm_var": 1.7143229166666667,
      "learning_rate": 0.0001,
      "loss": 7.4624,
      "loss/crossentropy": 2.192887546122074,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1984951412305236,
      "step": 2570
    },
    {
      "epoch": 0.0645,
      "grad_norm": 30.125,
      "grad_norm_var": 1.9143229166666667,
      "learning_rate": 0.0001,
      "loss": 7.3947,
      "loss/crossentropy": 2.102549520134926,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1989850653335452,
      "step": 2580
    },
    {
      "epoch": 0.06475,
      "grad_norm": 32.25,
      "grad_norm_var": 9.5322265625,
      "learning_rate": 0.0001,
      "loss": 7.5147,
      "loss/crossentropy": 2.213281115144491,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2027151037938893,
      "step": 2590
    },
    {
      "epoch": 0.065,
      "grad_norm": 30.625,
      "grad_norm_var": 2.3427083333333334,
      "learning_rate": 0.0001,
      "loss": 7.4691,
      "loss/crossentropy": 2.1138279482722284,
      "loss/hidden": 3.379296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20825629755854608,
      "step": 2600
    },
    {
      "epoch": 0.06525,
      "grad_norm": 36.0,
      "grad_norm_var": 3.3395182291666665,
      "learning_rate": 0.0001,
      "loss": 7.4775,
      "loss/crossentropy": 2.107349547743797,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19337845854461194,
      "step": 2610
    },
    {
      "epoch": 0.0655,
      "grad_norm": 29.25,
      "grad_norm_var": 12.757291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5438,
      "loss/crossentropy": 2.0628502368927,
      "loss/hidden": 3.4984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20967572089284658,
      "step": 2620
    },
    {
      "epoch": 0.06575,
      "grad_norm": 28.625,
      "grad_norm_var": 11.805208333333333,
      "learning_rate": 0.0001,
      "loss": 7.3354,
      "loss/crossentropy": 2.1009589530527593,
      "loss/hidden": 3.33828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18132725274190306,
      "step": 2630
    },
    {
      "epoch": 0.066,
      "grad_norm": 32.5,
      "grad_norm_var": 2.730208333333333,
      "learning_rate": 0.0001,
      "loss": 7.4257,
      "loss/crossentropy": 1.983342681080103,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19340286049991845,
      "step": 2640
    },
    {
      "epoch": 0.06625,
      "grad_norm": 30.25,
      "grad_norm_var": 3.7549465282226944e+18,
      "learning_rate": 0.0001,
      "loss": 7.309,
      "loss/crossentropy": 2.0057250812649725,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18936716187745334,
      "step": 2650
    },
    {
      "epoch": 0.0665,
      "grad_norm": 36.25,
      "grad_norm_var": 8.832747395833334,
      "learning_rate": 0.0001,
      "loss": 7.5442,
      "loss/crossentropy": 2.054753464460373,
      "loss/hidden": 3.410546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2035602940246463,
      "step": 2660
    },
    {
      "epoch": 0.06675,
      "grad_norm": 32.5,
      "grad_norm_var": 4.8900390625,
      "learning_rate": 0.0001,
      "loss": 7.4106,
      "loss/crossentropy": 2.0181221179664135,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1878144398331642,
      "step": 2670
    },
    {
      "epoch": 0.067,
      "grad_norm": 30.125,
      "grad_norm_var": 4.280989583333334,
      "learning_rate": 0.0001,
      "loss": 7.4597,
      "loss/crossentropy": 2.200540581345558,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20286752395331858,
      "step": 2680
    },
    {
      "epoch": 0.06725,
      "grad_norm": 31.75,
      "grad_norm_var": 3.8559895833333333,
      "learning_rate": 0.0001,
      "loss": 7.4643,
      "loss/crossentropy": 2.0630861818790436,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20401672925800085,
      "step": 2690
    },
    {
      "epoch": 0.0675,
      "grad_norm": 33.0,
      "grad_norm_var": 7.073958333333334,
      "learning_rate": 0.0001,
      "loss": 7.4001,
      "loss/crossentropy": 1.927167509496212,
      "loss/hidden": 3.31328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17901942003518342,
      "step": 2700
    },
    {
      "epoch": 0.06775,
      "grad_norm": 30.25,
      "grad_norm_var": 8.9009765625,
      "learning_rate": 0.0001,
      "loss": 7.3461,
      "loss/crossentropy": 2.0538916781544687,
      "loss/hidden": 3.35234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1864149821922183,
      "step": 2710
    },
    {
      "epoch": 0.068,
      "grad_norm": 29.5,
      "grad_norm_var": 2.218489583333333,
      "learning_rate": 0.0001,
      "loss": 7.526,
      "loss/crossentropy": 2.211588367819786,
      "loss/hidden": 3.487890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20801848396658898,
      "step": 2720
    },
    {
      "epoch": 0.06825,
      "grad_norm": 31.375,
      "grad_norm_var": 1.0768229166666667,
      "learning_rate": 0.0001,
      "loss": 7.5535,
      "loss/crossentropy": 2.268890543282032,
      "loss/hidden": 3.39921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21352684032171965,
      "step": 2730
    },
    {
      "epoch": 0.0685,
      "grad_norm": 33.25,
      "grad_norm_var": 5.663997395833333,
      "learning_rate": 0.0001,
      "loss": 7.411,
      "loss/crossentropy": 1.902898482978344,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1794701736420393,
      "step": 2740
    },
    {
      "epoch": 0.06875,
      "grad_norm": 32.25,
      "grad_norm_var": 6.167708333333334,
      "learning_rate": 0.0001,
      "loss": 7.3718,
      "loss/crossentropy": 1.9450767874717712,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18759301900863648,
      "step": 2750
    },
    {
      "epoch": 0.069,
      "grad_norm": 31.125,
      "grad_norm_var": 31.185872395833332,
      "learning_rate": 0.0001,
      "loss": 7.4359,
      "loss/crossentropy": 2.0783849939703942,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18503105416893958,
      "step": 2760
    },
    {
      "epoch": 0.06925,
      "grad_norm": 36.5,
      "grad_norm_var": 35.412434895833336,
      "learning_rate": 0.0001,
      "loss": 7.5806,
      "loss/crossentropy": 2.2374701410532,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19615829903632404,
      "step": 2770
    },
    {
      "epoch": 0.0695,
      "grad_norm": 30.25,
      "grad_norm_var": 19.787239583333335,
      "learning_rate": 0.0001,
      "loss": 7.3197,
      "loss/crossentropy": 1.8297001466155052,
      "loss/hidden": 3.3171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16481583826243879,
      "step": 2780
    },
    {
      "epoch": 0.06975,
      "grad_norm": 428.0,
      "grad_norm_var": 9873.31640625,
      "learning_rate": 0.0001,
      "loss": 7.5313,
      "loss/crossentropy": 2.249661484360695,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2018596636131406,
      "step": 2790
    },
    {
      "epoch": 0.07,
      "grad_norm": 31.0,
      "grad_norm_var": 9755.6625,
      "learning_rate": 0.0001,
      "loss": 7.3957,
      "loss/crossentropy": 1.9368772380053998,
      "loss/hidden": 3.48203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18386599626392125,
      "step": 2800
    },
    {
      "epoch": 0.07025,
      "grad_norm": 30.75,
      "grad_norm_var": 1.8317057291666667,
      "learning_rate": 0.0001,
      "loss": 7.4372,
      "loss/crossentropy": 1.98307463824749,
      "loss/hidden": 3.464453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19818334747105837,
      "step": 2810
    },
    {
      "epoch": 0.0705,
      "grad_norm": 29.375,
      "grad_norm_var": 2.589583333333333,
      "learning_rate": 0.0001,
      "loss": 7.5014,
      "loss/crossentropy": 2.1463105253875256,
      "loss/hidden": 3.5046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20105676222592592,
      "step": 2820
    },
    {
      "epoch": 0.07075,
      "grad_norm": 60.5,
      "grad_norm_var": 178.2556640625,
      "learning_rate": 0.0001,
      "loss": 7.4527,
      "loss/crossentropy": 2.0776613369584083,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19452448841184378,
      "step": 2830
    },
    {
      "epoch": 0.071,
      "grad_norm": 29.25,
      "grad_norm_var": 172.31451822916668,
      "learning_rate": 0.0001,
      "loss": 7.4802,
      "loss/crossentropy": 2.1200039610266685,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19831879772245883,
      "step": 2840
    },
    {
      "epoch": 0.07125,
      "grad_norm": 69.0,
      "grad_norm_var": 117.23098958333334,
      "learning_rate": 0.0001,
      "loss": 7.434,
      "loss/crossentropy": 2.024143140017986,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1836528332903981,
      "step": 2850
    },
    {
      "epoch": 0.0715,
      "grad_norm": 31.375,
      "grad_norm_var": 92.53723958333333,
      "learning_rate": 0.0001,
      "loss": 7.4934,
      "loss/crossentropy": 2.2765417456626893,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20736196860671044,
      "step": 2860
    },
    {
      "epoch": 0.07175,
      "grad_norm": 31.625,
      "grad_norm_var": 7.986393229166667,
      "learning_rate": 0.0001,
      "loss": 7.4826,
      "loss/crossentropy": 2.269197002053261,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19869209118187428,
      "step": 2870
    },
    {
      "epoch": 0.072,
      "grad_norm": 31.25,
      "grad_norm_var": 3.1806640625,
      "learning_rate": 0.0001,
      "loss": 7.4018,
      "loss/crossentropy": 2.2985214799642564,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20524807646870613,
      "step": 2880
    },
    {
      "epoch": 0.07225,
      "grad_norm": 30.875,
      "grad_norm_var": 4.801822916666667,
      "learning_rate": 0.0001,
      "loss": 7.5148,
      "loss/crossentropy": 2.2387808740139006,
      "loss/hidden": 3.46015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19951685946434736,
      "step": 2890
    },
    {
      "epoch": 0.0725,
      "grad_norm": 28.875,
      "grad_norm_var": 13.836458333333333,
      "learning_rate": 0.0001,
      "loss": 7.5232,
      "loss/crossentropy": 2.049694790691137,
      "loss/hidden": 3.410546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19052465092390775,
      "step": 2900
    },
    {
      "epoch": 0.07275,
      "grad_norm": 29.625,
      "grad_norm_var": 17.91640625,
      "learning_rate": 0.0001,
      "loss": 7.3227,
      "loss/crossentropy": 2.0360258772969244,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18495636582374572,
      "step": 2910
    },
    {
      "epoch": 0.073,
      "grad_norm": 32.0,
      "grad_norm_var": 1.8926377214767268e+18,
      "learning_rate": 0.0001,
      "loss": 7.4512,
      "loss/crossentropy": 2.13848315179348,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18625867497175932,
      "step": 2920
    },
    {
      "epoch": 0.07325,
      "grad_norm": 29.875,
      "grad_norm_var": 1.8926377199175642e+18,
      "learning_rate": 0.0001,
      "loss": 7.5038,
      "loss/crossentropy": 2.166595605015755,
      "loss/hidden": 3.49375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20948194600641729,
      "step": 2930
    },
    {
      "epoch": 0.0735,
      "grad_norm": 28.5,
      "grad_norm_var": 73.08020833333333,
      "learning_rate": 0.0001,
      "loss": 7.374,
      "loss/crossentropy": 1.9849643550813199,
      "loss/hidden": 3.301171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18302082028239966,
      "step": 2940
    },
    {
      "epoch": 0.07375,
      "grad_norm": 29.125,
      "grad_norm_var": 24.825,
      "learning_rate": 0.0001,
      "loss": 7.3651,
      "loss/crossentropy": 2.057874396443367,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1866615541279316,
      "step": 2950
    },
    {
      "epoch": 0.074,
      "grad_norm": 30.625,
      "grad_norm_var": 883.6354166666666,
      "learning_rate": 0.0001,
      "loss": 7.5415,
      "loss/crossentropy": 2.1631729155778885,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20762786027044058,
      "step": 2960
    },
    {
      "epoch": 0.07425,
      "grad_norm": 32.75,
      "grad_norm_var": 887.2705729166667,
      "learning_rate": 0.0001,
      "loss": 7.4471,
      "loss/crossentropy": 1.9493468508124352,
      "loss/hidden": 3.3578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1884406829252839,
      "step": 2970
    },
    {
      "epoch": 0.0745,
      "grad_norm": 28.875,
      "grad_norm_var": 5.070768229166666,
      "learning_rate": 0.0001,
      "loss": 7.605,
      "loss/crossentropy": 2.122344336658716,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21057356838136912,
      "step": 2980
    },
    {
      "epoch": 0.07475,
      "grad_norm": 37.0,
      "grad_norm_var": 21.535416666666666,
      "learning_rate": 0.0001,
      "loss": 7.469,
      "loss/crossentropy": 2.008989527821541,
      "loss/hidden": 3.54140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2172183733433485,
      "step": 2990
    },
    {
      "epoch": 0.075,
      "grad_norm": 29.375,
      "grad_norm_var": 18.198958333333334,
      "learning_rate": 0.0001,
      "loss": 7.3932,
      "loss/crossentropy": 2.1922819674015046,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20425879992544652,
      "step": 3000
    },
    {
      "epoch": 0.07525,
      "grad_norm": 29.5,
      "grad_norm_var": 2.668684895833333,
      "learning_rate": 0.0001,
      "loss": 7.3505,
      "loss/crossentropy": 2.189265179634094,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20808048862963915,
      "step": 3010
    },
    {
      "epoch": 0.0755,
      "grad_norm": 30.75,
      "grad_norm_var": 14.20625,
      "learning_rate": 0.0001,
      "loss": 7.5013,
      "loss/crossentropy": 2.0573098927736284,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18116160985082388,
      "step": 3020
    },
    {
      "epoch": 0.07575,
      "grad_norm": 31.375,
      "grad_norm_var": 16.983333333333334,
      "learning_rate": 0.0001,
      "loss": 7.4455,
      "loss/crossentropy": 1.9735823571681976,
      "loss/hidden": 3.459765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19495000168681145,
      "step": 3030
    },
    {
      "epoch": 0.076,
      "grad_norm": 7247757312.0,
      "grad_norm_var": 3.2831240991582193e+18,
      "learning_rate": 0.0001,
      "loss": 7.4881,
      "loss/crossentropy": 1.971890377253294,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18015608433634042,
      "step": 3040
    },
    {
      "epoch": 0.07625,
      "grad_norm": 28.25,
      "grad_norm_var": 3.283124098780732e+18,
      "learning_rate": 0.0001,
      "loss": 7.3664,
      "loss/crossentropy": 1.8378953270614147,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1741427879780531,
      "step": 3050
    },
    {
      "epoch": 0.0765,
      "grad_norm": 31.75,
      "grad_norm_var": 1.89140625,
      "learning_rate": 0.0001,
      "loss": 7.5137,
      "loss/crossentropy": 2.141886255145073,
      "loss/hidden": 3.443359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19584037065505983,
      "step": 3060
    },
    {
      "epoch": 0.07675,
      "grad_norm": 27.25,
      "grad_norm_var": 2.4244140625,
      "learning_rate": 0.0001,
      "loss": 7.4296,
      "loss/crossentropy": 2.0373554110527037,
      "loss/hidden": 3.5640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.216986732929945,
      "step": 3070
    },
    {
      "epoch": 0.077,
      "grad_norm": 35.25,
      "grad_norm_var": 3.7322265625,
      "learning_rate": 0.0001,
      "loss": 7.5269,
      "loss/crossentropy": 1.975497831404209,
      "loss/hidden": 3.333984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1780722170136869,
      "step": 3080
    },
    {
      "epoch": 0.07725,
      "grad_norm": 32.75,
      "grad_norm_var": 3.6895182291666666,
      "learning_rate": 0.0001,
      "loss": 7.4938,
      "loss/crossentropy": 2.151789793372154,
      "loss/hidden": 3.502734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21854450944811105,
      "step": 3090
    },
    {
      "epoch": 0.0775,
      "grad_norm": 29.5,
      "grad_norm_var": 6.82265625,
      "learning_rate": 0.0001,
      "loss": 7.4321,
      "loss/crossentropy": 1.9484706297516823,
      "loss/hidden": 3.506640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19896488767117262,
      "step": 3100
    },
    {
      "epoch": 0.07775,
      "grad_norm": 29.75,
      "grad_norm_var": 3.0780598958333334,
      "learning_rate": 0.0001,
      "loss": 7.5471,
      "loss/crossentropy": 2.165594828128815,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20095103643834591,
      "step": 3110
    },
    {
      "epoch": 0.078,
      "grad_norm": 29.0,
      "grad_norm_var": 2.2197916666666666,
      "learning_rate": 0.0001,
      "loss": 7.6334,
      "loss/crossentropy": 2.1854751259088516,
      "loss/hidden": 3.4640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21246263310313224,
      "step": 3120
    },
    {
      "epoch": 0.07825,
      "grad_norm": 29.0,
      "grad_norm_var": 3.71640625,
      "learning_rate": 0.0001,
      "loss": 7.4278,
      "loss/crossentropy": 1.914103902876377,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18373754434287548,
      "step": 3130
    },
    {
      "epoch": 0.0785,
      "grad_norm": 29.0,
      "grad_norm_var": 1.2952473958333333,
      "learning_rate": 0.0001,
      "loss": 7.4487,
      "loss/crossentropy": 1.9421842776238918,
      "loss/hidden": 3.5296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19919300880283117,
      "step": 3140
    },
    {
      "epoch": 0.07875,
      "grad_norm": 29.375,
      "grad_norm_var": 1.8268229166666667,
      "learning_rate": 0.0001,
      "loss": 7.5818,
      "loss/crossentropy": 2.0765694811940194,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19946561977267266,
      "step": 3150
    },
    {
      "epoch": 0.079,
      "grad_norm": 28.125,
      "grad_norm_var": 11.483268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4372,
      "loss/crossentropy": 2.013955050334334,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20109358858317136,
      "step": 3160
    },
    {
      "epoch": 0.07925,
      "grad_norm": 28.875,
      "grad_norm_var": 12.871809895833334,
      "learning_rate": 0.0001,
      "loss": 7.4606,
      "loss/crossentropy": 2.2802242666482924,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21229397617280482,
      "step": 3170
    },
    {
      "epoch": 0.0795,
      "grad_norm": 28.375,
      "grad_norm_var": 1.6301432291666667,
      "learning_rate": 0.0001,
      "loss": 7.4691,
      "loss/crossentropy": 2.134338477253914,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18632632456719875,
      "step": 3180
    },
    {
      "epoch": 0.07975,
      "grad_norm": 30.625,
      "grad_norm_var": 2.6113932291666666,
      "learning_rate": 0.0001,
      "loss": 7.4903,
      "loss/crossentropy": 2.192245528101921,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19276445377618073,
      "step": 3190
    },
    {
      "epoch": 0.08,
      "grad_norm": 27.875,
      "grad_norm_var": 2.6830729166666667,
      "learning_rate": 0.0001,
      "loss": 7.4715,
      "loss/crossentropy": 2.1333388604223726,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1902673264965415,
      "step": 3200
    },
    {
      "epoch": 0.08025,
      "grad_norm": 29.625,
      "grad_norm_var": 2.7072265625,
      "learning_rate": 0.0001,
      "loss": 7.4646,
      "loss/crossentropy": 2.1069626569747926,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18933899328112602,
      "step": 3210
    },
    {
      "epoch": 0.0805,
      "grad_norm": 33.0,
      "grad_norm_var": 1.6457682291666667,
      "learning_rate": 0.0001,
      "loss": 7.3771,
      "loss/crossentropy": 2.143903985619545,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19841080270707606,
      "step": 3220
    },
    {
      "epoch": 0.08075,
      "grad_norm": 29.5,
      "grad_norm_var": 2.405143229166667,
      "learning_rate": 0.0001,
      "loss": 7.4629,
      "loss/crossentropy": 1.9501185864210129,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2003694986924529,
      "step": 3230
    },
    {
      "epoch": 0.081,
      "grad_norm": 35.0,
      "grad_norm_var": 3.4619140625,
      "learning_rate": 0.0001,
      "loss": 7.6085,
      "loss/crossentropy": 2.1099744185805323,
      "loss/hidden": 3.33515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1865939747542143,
      "step": 3240
    },
    {
      "epoch": 0.08125,
      "grad_norm": 38.0,
      "grad_norm_var": 15.54140625,
      "learning_rate": 0.0001,
      "loss": 7.4858,
      "loss/crossentropy": 1.8915734700858593,
      "loss/hidden": 3.550390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20414282865822314,
      "step": 3250
    },
    {
      "epoch": 0.0815,
      "grad_norm": 31.875,
      "grad_norm_var": 15.074934895833334,
      "learning_rate": 0.0001,
      "loss": 7.4995,
      "loss/crossentropy": 2.0746393710374833,
      "loss/hidden": 3.448046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19025763403624296,
      "step": 3260
    },
    {
      "epoch": 0.08175,
      "grad_norm": 29.625,
      "grad_norm_var": 4.532291666666667,
      "learning_rate": 0.0001,
      "loss": 7.4517,
      "loss/crossentropy": 2.201898355782032,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1851862959563732,
      "step": 3270
    },
    {
      "epoch": 0.082,
      "grad_norm": 32.25,
      "grad_norm_var": 9.199739583333333,
      "learning_rate": 0.0001,
      "loss": 7.4085,
      "loss/crossentropy": 1.9774614453315735,
      "loss/hidden": 3.5078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1853517958894372,
      "step": 3280
    },
    {
      "epoch": 0.08225,
      "grad_norm": 31.0,
      "grad_norm_var": 13.801497395833334,
      "learning_rate": 0.0001,
      "loss": 7.4065,
      "loss/crossentropy": 2.1263367265462874,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18529028967022895,
      "step": 3290
    },
    {
      "epoch": 0.0825,
      "grad_norm": 29.5,
      "grad_norm_var": 2.967643229166667,
      "learning_rate": 0.0001,
      "loss": 7.4165,
      "loss/crossentropy": 2.193544697761536,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19897244460880756,
      "step": 3300
    },
    {
      "epoch": 0.08275,
      "grad_norm": 33.75,
      "grad_norm_var": 9.687239583333334,
      "learning_rate": 0.0001,
      "loss": 7.5716,
      "loss/crossentropy": 2.0868531957268717,
      "loss/hidden": 3.616796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21278488002717494,
      "step": 3310
    },
    {
      "epoch": 0.083,
      "grad_norm": 31.0,
      "grad_norm_var": 7.9478515625,
      "learning_rate": 0.0001,
      "loss": 7.5543,
      "loss/crossentropy": 2.1392074063420297,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21113577168434858,
      "step": 3320
    },
    {
      "epoch": 0.08325,
      "grad_norm": 30.0,
      "grad_norm_var": 2.0268229166666667,
      "learning_rate": 0.0001,
      "loss": 7.4454,
      "loss/crossentropy": 2.0691144198179243,
      "loss/hidden": 3.341796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20186964478343725,
      "step": 3330
    },
    {
      "epoch": 0.0835,
      "grad_norm": 31.5,
      "grad_norm_var": 2.6211653265769103e+18,
      "learning_rate": 0.0001,
      "loss": 7.4481,
      "loss/crossentropy": 2.0832756504416468,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19915037509053946,
      "step": 3340
    },
    {
      "epoch": 0.08375,
      "grad_norm": 32.5,
      "grad_norm_var": 2.621165324337292e+18,
      "learning_rate": 0.0001,
      "loss": 7.3606,
      "loss/crossentropy": 2.102260760962963,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19333885367959738,
      "step": 3350
    },
    {
      "epoch": 0.084,
      "grad_norm": 29.25,
      "grad_norm_var": 85.575,
      "learning_rate": 0.0001,
      "loss": 7.4073,
      "loss/crossentropy": 2.149528594315052,
      "loss/hidden": 3.491796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2177526842802763,
      "step": 3360
    },
    {
      "epoch": 0.08425,
      "grad_norm": 30.25,
      "grad_norm_var": 2.8645833333333335,
      "learning_rate": 0.0001,
      "loss": 7.4642,
      "loss/crossentropy": 2.085590344667435,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17804578468203544,
      "step": 3370
    },
    {
      "epoch": 0.0845,
      "grad_norm": 33.25,
      "grad_norm_var": 2.996875,
      "learning_rate": 0.0001,
      "loss": 7.3953,
      "loss/crossentropy": 2.0975965946912765,
      "loss/hidden": 3.3,
      "loss/jsd": 0.0,
      "loss/logits": 0.1847201505675912,
      "step": 3380
    },
    {
      "epoch": 0.08475,
      "grad_norm": 32.0,
      "grad_norm_var": 2.470572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4553,
      "loss/crossentropy": 2.1018140748143197,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18250287007540464,
      "step": 3390
    },
    {
      "epoch": 0.085,
      "grad_norm": 30.25,
      "grad_norm_var": 2.887955729166667,
      "learning_rate": 0.0001,
      "loss": 7.5238,
      "loss/crossentropy": 2.1050665065646172,
      "loss/hidden": 3.501953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2124734738841653,
      "step": 3400
    },
    {
      "epoch": 0.08525,
      "grad_norm": 30.0,
      "grad_norm_var": 1.7143229166666667,
      "learning_rate": 0.0001,
      "loss": 7.2754,
      "loss/crossentropy": 2.0948296964168547,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.178215317055583,
      "step": 3410
    },
    {
      "epoch": 0.0855,
      "grad_norm": 30.875,
      "grad_norm_var": 5.354622395833333,
      "learning_rate": 0.0001,
      "loss": 7.4191,
      "loss/crossentropy": 2.0418393671512605,
      "loss/hidden": 3.4734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18740264605730772,
      "step": 3420
    },
    {
      "epoch": 0.08575,
      "grad_norm": 32.25,
      "grad_norm_var": 6.430989583333333,
      "learning_rate": 0.0001,
      "loss": 7.5642,
      "loss/crossentropy": 2.0279636546969413,
      "loss/hidden": 3.55859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20154636316001415,
      "step": 3430
    },
    {
      "epoch": 0.086,
      "grad_norm": 29.125,
      "grad_norm_var": 53.64791666666667,
      "learning_rate": 0.0001,
      "loss": 7.485,
      "loss/crossentropy": 2.0705729112029077,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22035282999277114,
      "step": 3440
    },
    {
      "epoch": 0.08625,
      "grad_norm": 30.375,
      "grad_norm_var": 5.54765625,
      "learning_rate": 0.0001,
      "loss": 7.428,
      "loss/crossentropy": 1.9830067940056324,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19354272997006775,
      "step": 3450
    },
    {
      "epoch": 0.0865,
      "grad_norm": 28.375,
      "grad_norm_var": 2.758736220726598e+18,
      "learning_rate": 0.0001,
      "loss": 7.4342,
      "loss/crossentropy": 2.1590976014733316,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20231232214719058,
      "step": 3460
    },
    {
      "epoch": 0.08675,
      "grad_norm": 29.125,
      "grad_norm_var": 2.470572916666667,
      "learning_rate": 0.0001,
      "loss": 7.3376,
      "loss/crossentropy": 2.108407254517078,
      "loss/hidden": 3.416796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18425025548785925,
      "step": 3470
    },
    {
      "epoch": 0.087,
      "grad_norm": 32.5,
      "grad_norm_var": 19.315559895833335,
      "learning_rate": 0.0001,
      "loss": 7.391,
      "loss/crossentropy": 2.086346108466387,
      "loss/hidden": 3.380859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19492445401847364,
      "step": 3480
    },
    {
      "epoch": 0.08725,
      "grad_norm": 30.75,
      "grad_norm_var": 3.9009765625,
      "learning_rate": 0.0001,
      "loss": 7.454,
      "loss/crossentropy": 2.0728737086057665,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21246139723807572,
      "step": 3490
    },
    {
      "epoch": 0.0875,
      "grad_norm": 53.25,
      "grad_norm_var": 34.962955729166666,
      "learning_rate": 0.0001,
      "loss": 7.4001,
      "loss/crossentropy": 1.9173476293683052,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18263984741643072,
      "step": 3500
    },
    {
      "epoch": 0.08775,
      "grad_norm": 29.875,
      "grad_norm_var": 36.22389322916667,
      "learning_rate": 0.0001,
      "loss": 7.5855,
      "loss/crossentropy": 1.9761252515017986,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20959299746900797,
      "step": 3510
    },
    {
      "epoch": 0.088,
      "grad_norm": 32.25,
      "grad_norm_var": 17.7337890625,
      "learning_rate": 0.0001,
      "loss": 7.4728,
      "loss/crossentropy": 2.0416554152965545,
      "loss/hidden": 3.4703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19014756735414268,
      "step": 3520
    },
    {
      "epoch": 0.08825,
      "grad_norm": 29.375,
      "grad_norm_var": 14.664322916666666,
      "learning_rate": 0.0001,
      "loss": 7.5608,
      "loss/crossentropy": 2.2834356099367143,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19908843878656626,
      "step": 3530
    },
    {
      "epoch": 0.0885,
      "grad_norm": 31.875,
      "grad_norm_var": 2.6702473958333335,
      "learning_rate": 0.0001,
      "loss": 7.4804,
      "loss/crossentropy": 2.0417330890893934,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20852382443845272,
      "step": 3540
    },
    {
      "epoch": 0.08875,
      "grad_norm": 31.625,
      "grad_norm_var": 2.460724589971584e+18,
      "learning_rate": 0.0001,
      "loss": 7.5559,
      "loss/crossentropy": 2.1676768481731417,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1989177169278264,
      "step": 3550
    },
    {
      "epoch": 0.089,
      "grad_norm": 30.0,
      "grad_norm_var": 6.881705729166667,
      "learning_rate": 0.0001,
      "loss": 7.4678,
      "loss/crossentropy": 2.221273897588253,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.19402988757938147,
      "step": 3560
    },
    {
      "epoch": 0.08925,
      "grad_norm": 31.375,
      "grad_norm_var": 7.732747395833333,
      "learning_rate": 0.0001,
      "loss": 7.4508,
      "loss/crossentropy": 2.1802149415016174,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20121808685362338,
      "step": 3570
    },
    {
      "epoch": 0.0895,
      "grad_norm": 52.5,
      "grad_norm_var": 30.9775390625,
      "learning_rate": 0.0001,
      "loss": 7.3982,
      "loss/crossentropy": 2.085124118626118,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18448642026633025,
      "step": 3580
    },
    {
      "epoch": 0.08975,
      "grad_norm": 30.875,
      "grad_norm_var": 32.91295572916667,
      "learning_rate": 0.0001,
      "loss": 7.4381,
      "loss/crossentropy": 2.1467449337244036,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19393185302615165,
      "step": 3590
    },
    {
      "epoch": 0.09,
      "grad_norm": 29.25,
      "grad_norm_var": 1.4708333333333334,
      "learning_rate": 0.0001,
      "loss": 7.415,
      "loss/crossentropy": 2.0135369554162024,
      "loss/hidden": 3.37109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18443848174065353,
      "step": 3600
    },
    {
      "epoch": 0.09025,
      "grad_norm": 31.375,
      "grad_norm_var": 6.1962890625,
      "learning_rate": 0.0001,
      "loss": 7.4028,
      "loss/crossentropy": 2.1443901300430297,
      "loss/hidden": 3.440234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2054579086601734,
      "step": 3610
    },
    {
      "epoch": 0.0905,
      "grad_norm": 26.5,
      "grad_norm_var": 3.562239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3255,
      "loss/crossentropy": 1.799356396496296,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17441922090947629,
      "step": 3620
    },
    {
      "epoch": 0.09075,
      "grad_norm": 31.375,
      "grad_norm_var": 2.2083333333333335,
      "learning_rate": 0.0001,
      "loss": 7.4272,
      "loss/crossentropy": 1.9925116747617722,
      "loss/hidden": 3.52578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21653544921427964,
      "step": 3630
    },
    {
      "epoch": 0.091,
      "grad_norm": 30.125,
      "grad_norm_var": 0.6125,
      "learning_rate": 0.0001,
      "loss": 7.3649,
      "loss/crossentropy": 2.135761073231697,
      "loss/hidden": 3.4140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18989351522177458,
      "step": 3640
    },
    {
      "epoch": 0.09125,
      "grad_norm": 31.375,
      "grad_norm_var": 1.4330729166666667,
      "learning_rate": 0.0001,
      "loss": 7.4505,
      "loss/crossentropy": 2.0986070543527604,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18352905213832854,
      "step": 3650
    },
    {
      "epoch": 0.0915,
      "grad_norm": 29.625,
      "grad_norm_var": 2.5869140625,
      "learning_rate": 0.0001,
      "loss": 7.4199,
      "loss/crossentropy": 2.1555575743317603,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19261632524430752,
      "step": 3660
    },
    {
      "epoch": 0.09175,
      "grad_norm": 31.5,
      "grad_norm_var": 2.371875,
      "learning_rate": 0.0001,
      "loss": 7.5463,
      "loss/crossentropy": 2.1411691516637803,
      "loss/hidden": 3.446875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2046652188524604,
      "step": 3670
    },
    {
      "epoch": 0.092,
      "grad_norm": 30.625,
      "grad_norm_var": 4.703580729166666,
      "learning_rate": 0.0001,
      "loss": 7.404,
      "loss/crossentropy": 2.142404294013977,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20414466112852098,
      "step": 3680
    },
    {
      "epoch": 0.09225,
      "grad_norm": 30.375,
      "grad_norm_var": 3.25625,
      "learning_rate": 0.0001,
      "loss": 7.4774,
      "loss/crossentropy": 2.187901920080185,
      "loss/hidden": 3.480859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21911972090601922,
      "step": 3690
    },
    {
      "epoch": 0.0925,
      "grad_norm": 31.875,
      "grad_norm_var": 1.2166666666666666,
      "learning_rate": 0.0001,
      "loss": 7.5965,
      "loss/crossentropy": 2.086391404271126,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2020766455680132,
      "step": 3700
    },
    {
      "epoch": 0.09275,
      "grad_norm": 30.625,
      "grad_norm_var": 2.147330729166667,
      "learning_rate": 0.0001,
      "loss": 7.4579,
      "loss/crossentropy": 2.09081457182765,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1868050311692059,
      "step": 3710
    },
    {
      "epoch": 0.093,
      "grad_norm": 34.25,
      "grad_norm_var": 2.467643229166667,
      "learning_rate": 0.0001,
      "loss": 7.522,
      "loss/crossentropy": 2.12264247238636,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18927707765251398,
      "step": 3720
    },
    {
      "epoch": 0.09325,
      "grad_norm": 32.25,
      "grad_norm_var": 3.981184895833333,
      "learning_rate": 0.0001,
      "loss": 7.4155,
      "loss/crossentropy": 2.1118928104639054,
      "loss/hidden": 3.44765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19489197488874196,
      "step": 3730
    },
    {
      "epoch": 0.0935,
      "grad_norm": 34.0,
      "grad_norm_var": 5.312434895833333,
      "learning_rate": 0.0001,
      "loss": 7.5053,
      "loss/crossentropy": 2.1360882744193077,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19313989579677582,
      "step": 3740
    },
    {
      "epoch": 0.09375,
      "grad_norm": 29.125,
      "grad_norm_var": 4.549739583333333,
      "learning_rate": 0.0001,
      "loss": 7.3275,
      "loss/crossentropy": 2.010613538324833,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18421147018671036,
      "step": 3750
    },
    {
      "epoch": 0.094,
      "grad_norm": 31.625,
      "grad_norm_var": 1.5541666666666667,
      "learning_rate": 0.0001,
      "loss": 7.4784,
      "loss/crossentropy": 2.1465295113623144,
      "loss/hidden": 3.323828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18987073097378016,
      "step": 3760
    },
    {
      "epoch": 0.09425,
      "grad_norm": 32.75,
      "grad_norm_var": 1.9018229166666667,
      "learning_rate": 0.0001,
      "loss": 7.3495,
      "loss/crossentropy": 2.17747982442379,
      "loss/hidden": 3.48046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2016214355826378,
      "step": 3770
    },
    {
      "epoch": 0.0945,
      "grad_norm": 30.875,
      "grad_norm_var": 3.088997395833333,
      "learning_rate": 0.0001,
      "loss": 7.5384,
      "loss/crossentropy": 2.179350584745407,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19142594784498215,
      "step": 3780
    },
    {
      "epoch": 0.09475,
      "grad_norm": 29.625,
      "grad_norm_var": 1.1559895833333333,
      "learning_rate": 0.0001,
      "loss": 7.4035,
      "loss/crossentropy": 2.155378046631813,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19720839541405438,
      "step": 3790
    },
    {
      "epoch": 0.095,
      "grad_norm": 30.625,
      "grad_norm_var": 1.1999348958333333,
      "learning_rate": 0.0001,
      "loss": 7.4441,
      "loss/crossentropy": 2.0597486779093743,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19279775265604257,
      "step": 3800
    },
    {
      "epoch": 0.09525,
      "grad_norm": 33.5,
      "grad_norm_var": 2.1666666666666665,
      "learning_rate": 0.0001,
      "loss": 7.5146,
      "loss/crossentropy": 2.1966816753149034,
      "loss/hidden": 3.44375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20174810625612735,
      "step": 3810
    },
    {
      "epoch": 0.0955,
      "grad_norm": 31.5,
      "grad_norm_var": 1.9593098958333333,
      "learning_rate": 0.0001,
      "loss": 7.539,
      "loss/crossentropy": 2.165803623199463,
      "loss/hidden": 3.34375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1953417781740427,
      "step": 3820
    },
    {
      "epoch": 0.09575,
      "grad_norm": 32.0,
      "grad_norm_var": 6.690625,
      "learning_rate": 0.0001,
      "loss": 7.514,
      "loss/crossentropy": 2.0817860513925552,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20838446952402592,
      "step": 3830
    },
    {
      "epoch": 0.096,
      "grad_norm": 32.75,
      "grad_norm_var": 7.6431640625,
      "learning_rate": 0.0001,
      "loss": 7.5472,
      "loss/crossentropy": 2.231910442560911,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21717903479002415,
      "step": 3840
    },
    {
      "epoch": 0.09625,
      "grad_norm": 32.0,
      "grad_norm_var": 16.134375,
      "learning_rate": 0.0001,
      "loss": 7.5807,
      "loss/crossentropy": 2.0746277555823327,
      "loss/hidden": 3.47578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20851925816386938,
      "step": 3850
    },
    {
      "epoch": 0.0965,
      "grad_norm": 30.625,
      "grad_norm_var": 16.132747395833334,
      "learning_rate": 0.0001,
      "loss": 7.3749,
      "loss/crossentropy": 2.1463438466191294,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19305863380432128,
      "step": 3860
    },
    {
      "epoch": 0.09675,
      "grad_norm": 32.5,
      "grad_norm_var": 1.0895182291666667,
      "learning_rate": 0.0001,
      "loss": 7.5499,
      "loss/crossentropy": 2.2108413323760034,
      "loss/hidden": 3.42734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20310868676751853,
      "step": 3870
    },
    {
      "epoch": 0.097,
      "grad_norm": 30.75,
      "grad_norm_var": 1.4559895833333334,
      "learning_rate": 0.0001,
      "loss": 7.4788,
      "loss/crossentropy": 2.0900154620409013,
      "loss/hidden": 3.42734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18780422061681748,
      "step": 3880
    },
    {
      "epoch": 0.09725,
      "grad_norm": 30.625,
      "grad_norm_var": 13.917643229166666,
      "learning_rate": 0.0001,
      "loss": 7.4391,
      "loss/crossentropy": 2.0574848279356956,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19390027467161416,
      "step": 3890
    },
    {
      "epoch": 0.0975,
      "grad_norm": 27.375,
      "grad_norm_var": 13.55,
      "learning_rate": 0.0001,
      "loss": 7.4327,
      "loss/crossentropy": 2.2832688719034193,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20608801003545524,
      "step": 3900
    },
    {
      "epoch": 0.09775,
      "grad_norm": 29.0,
      "grad_norm_var": 3.296875,
      "learning_rate": 0.0001,
      "loss": 7.3691,
      "loss/crossentropy": 1.9183307077735663,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1917601386550814,
      "step": 3910
    },
    {
      "epoch": 0.098,
      "grad_norm": 34.0,
      "grad_norm_var": 3.24765625,
      "learning_rate": 0.0001,
      "loss": 7.4628,
      "loss/crossentropy": 2.0630046002566815,
      "loss/hidden": 3.338671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1871832549571991,
      "step": 3920
    },
    {
      "epoch": 0.09825,
      "grad_norm": 31.75,
      "grad_norm_var": 1.5384765625,
      "learning_rate": 0.0001,
      "loss": 7.4868,
      "loss/crossentropy": 2.061261148750782,
      "loss/hidden": 3.415234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18525551967322826,
      "step": 3930
    },
    {
      "epoch": 0.0985,
      "grad_norm": 29.75,
      "grad_norm_var": 1.584375,
      "learning_rate": 0.0001,
      "loss": 7.5498,
      "loss/crossentropy": 2.0895790114998816,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1932330032810569,
      "step": 3940
    },
    {
      "epoch": 0.09875,
      "grad_norm": 30.625,
      "grad_norm_var": 25.79765625,
      "learning_rate": 0.0001,
      "loss": 7.6502,
      "loss/crossentropy": 2.1616804771125318,
      "loss/hidden": 3.365625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18905209768563508,
      "step": 3950
    },
    {
      "epoch": 0.099,
      "grad_norm": 30.5,
      "grad_norm_var": 28.547916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3334,
      "loss/crossentropy": 2.1435488507151605,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1910943292081356,
      "step": 3960
    },
    {
      "epoch": 0.09925,
      "grad_norm": 32.75,
      "grad_norm_var": 6.3650390625,
      "learning_rate": 0.0001,
      "loss": 7.542,
      "loss/crossentropy": 2.176460310816765,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18821860365569593,
      "step": 3970
    },
    {
      "epoch": 0.0995,
      "grad_norm": 31.625,
      "grad_norm_var": 3.9905598958333335,
      "learning_rate": 0.0001,
      "loss": 7.5231,
      "loss/crossentropy": 2.2077176332473756,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21911400128155947,
      "step": 3980
    },
    {
      "epoch": 0.09975,
      "grad_norm": 31.125,
      "grad_norm_var": 1.75625,
      "learning_rate": 0.0001,
      "loss": 7.4868,
      "loss/crossentropy": 2.105836200714111,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1997914554551244,
      "step": 3990
    },
    {
      "epoch": 0.1,
      "grad_norm": 38.0,
      "grad_norm_var": 4.710416666666666,
      "learning_rate": 0.0001,
      "loss": 7.5675,
      "loss/crossentropy": 2.233233967423439,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20876242108643056,
      "step": 4000
    },
    {
      "epoch": 0.10025,
      "grad_norm": 28.625,
      "grad_norm_var": 7.56640625,
      "learning_rate": 0.0001,
      "loss": 7.4736,
      "loss/crossentropy": 2.103509198874235,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1953927006572485,
      "step": 4010
    },
    {
      "epoch": 0.1005,
      "grad_norm": 28.875,
      "grad_norm_var": 4.119791666666667,
      "learning_rate": 0.0001,
      "loss": 7.4509,
      "loss/crossentropy": 1.9697775058448315,
      "loss/hidden": 3.308984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17186311883851885,
      "step": 4020
    },
    {
      "epoch": 0.10075,
      "grad_norm": 29.5,
      "grad_norm_var": 1.3177083333333333,
      "learning_rate": 0.0001,
      "loss": 7.333,
      "loss/crossentropy": 2.0519870311021804,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1872571600601077,
      "step": 4030
    },
    {
      "epoch": 0.101,
      "grad_norm": 29.5,
      "grad_norm_var": 1.2785807291666667,
      "learning_rate": 0.0001,
      "loss": 7.3466,
      "loss/crossentropy": 2.0663713179528713,
      "loss/hidden": 3.39921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18582073990255593,
      "step": 4040
    },
    {
      "epoch": 0.10125,
      "grad_norm": 30.375,
      "grad_norm_var": 1.9577473958333333,
      "learning_rate": 0.0001,
      "loss": 7.3812,
      "loss/crossentropy": 2.1256399258971213,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19628962082788348,
      "step": 4050
    },
    {
      "epoch": 0.1015,
      "grad_norm": 30.625,
      "grad_norm_var": 0.53125,
      "learning_rate": 0.0001,
      "loss": 7.3726,
      "loss/crossentropy": 2.1235328309237955,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18646292947232723,
      "step": 4060
    },
    {
      "epoch": 0.10175,
      "grad_norm": 29.0,
      "grad_norm_var": 3.19255952647709e+18,
      "learning_rate": 0.0001,
      "loss": 7.4564,
      "loss/crossentropy": 2.0213126331567763,
      "loss/hidden": 3.496875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19607899691909553,
      "step": 4070
    },
    {
      "epoch": 0.102,
      "grad_norm": 28.75,
      "grad_norm_var": 3.48515625,
      "learning_rate": 0.0001,
      "loss": 7.3886,
      "loss/crossentropy": 2.0899658009409903,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1851665174588561,
      "step": 4080
    },
    {
      "epoch": 0.10225,
      "grad_norm": 29.5,
      "grad_norm_var": 1.8692057291666666,
      "learning_rate": 0.0001,
      "loss": 7.4838,
      "loss/crossentropy": 2.027493818849325,
      "loss/hidden": 3.49765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19640162959694862,
      "step": 4090
    },
    {
      "epoch": 0.1025,
      "grad_norm": 29.125,
      "grad_norm_var": 11.762434895833334,
      "learning_rate": 0.0001,
      "loss": 7.5099,
      "loss/crossentropy": 2.056584618985653,
      "loss/hidden": 3.32265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17638762388378382,
      "step": 4100
    },
    {
      "epoch": 0.10275,
      "grad_norm": 30.125,
      "grad_norm_var": 12.459375,
      "learning_rate": 0.0001,
      "loss": 7.5255,
      "loss/crossentropy": 2.0713445380330087,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18587317056953906,
      "step": 4110
    },
    {
      "epoch": 0.103,
      "grad_norm": 33.25,
      "grad_norm_var": 1.9958333333333333,
      "learning_rate": 0.0001,
      "loss": 7.4437,
      "loss/crossentropy": 2.2338072419166566,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18814200926572083,
      "step": 4120
    },
    {
      "epoch": 0.10325,
      "grad_norm": 32.75,
      "grad_norm_var": 3.1259765625,
      "learning_rate": 0.0001,
      "loss": 7.3184,
      "loss/crossentropy": 2.0210259817540646,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18816483654081823,
      "step": 4130
    },
    {
      "epoch": 0.1035,
      "grad_norm": 29.5,
      "grad_norm_var": 2.870247395833333,
      "learning_rate": 0.0001,
      "loss": 7.5124,
      "loss/crossentropy": 2.0151045128703116,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19255878478288652,
      "step": 4140
    },
    {
      "epoch": 0.10375,
      "grad_norm": 30.625,
      "grad_norm_var": 1.3926432291666666,
      "learning_rate": 0.0001,
      "loss": 7.5096,
      "loss/crossentropy": 1.9808883003890514,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19115560222417116,
      "step": 4150
    },
    {
      "epoch": 0.104,
      "grad_norm": 30.75,
      "grad_norm_var": 1.6979166666666667,
      "learning_rate": 0.0001,
      "loss": 7.549,
      "loss/crossentropy": 2.1932784736156465,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20479805655777455,
      "step": 4160
    },
    {
      "epoch": 0.10425,
      "grad_norm": 30.125,
      "grad_norm_var": 2.3333333333333335,
      "learning_rate": 0.0001,
      "loss": 7.3875,
      "loss/crossentropy": 1.8820222720503808,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17310038600116967,
      "step": 4170
    },
    {
      "epoch": 0.1045,
      "grad_norm": 33.0,
      "grad_norm_var": 3.7728515625,
      "learning_rate": 0.0001,
      "loss": 7.4212,
      "loss/crossentropy": 2.082476270198822,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19099258184432982,
      "step": 4180
    },
    {
      "epoch": 0.10475,
      "grad_norm": 30.875,
      "grad_norm_var": 11.408268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4991,
      "loss/crossentropy": 2.287242355942726,
      "loss/hidden": 3.375390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1982285875827074,
      "step": 4190
    },
    {
      "epoch": 0.105,
      "grad_norm": 28.75,
      "grad_norm_var": 2.999739583333333,
      "learning_rate": 0.0001,
      "loss": 7.5959,
      "loss/crossentropy": 2.1783332407474516,
      "loss/hidden": 3.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2117959801107645,
      "step": 4200
    },
    {
      "epoch": 0.10525,
      "grad_norm": 30.0,
      "grad_norm_var": 4.708268229166666,
      "learning_rate": 0.0001,
      "loss": 7.3363,
      "loss/crossentropy": 1.955865352600813,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18177355360239744,
      "step": 4210
    },
    {
      "epoch": 0.1055,
      "grad_norm": 30.625,
      "grad_norm_var": 3.0254557291666666,
      "learning_rate": 0.0001,
      "loss": 7.4673,
      "loss/crossentropy": 1.833389012515545,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1878132861107588,
      "step": 4220
    },
    {
      "epoch": 0.10575,
      "grad_norm": 32.0,
      "grad_norm_var": 3.05,
      "learning_rate": 0.0001,
      "loss": 7.3969,
      "loss/crossentropy": 1.9096243590116502,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17025592969730496,
      "step": 4230
    },
    {
      "epoch": 0.106,
      "grad_norm": 30.875,
      "grad_norm_var": 1.82265625,
      "learning_rate": 0.0001,
      "loss": 7.4638,
      "loss/crossentropy": 2.0454175233840943,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20515710916370153,
      "step": 4240
    },
    {
      "epoch": 0.10625,
      "grad_norm": 30.125,
      "grad_norm_var": 3.1333333333333333,
      "learning_rate": 0.0001,
      "loss": 7.5126,
      "loss/crossentropy": 2.089062933623791,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19156677946448325,
      "step": 4250
    },
    {
      "epoch": 0.1065,
      "grad_norm": 29.0,
      "grad_norm_var": 4.311393229166667,
      "learning_rate": 0.0001,
      "loss": 7.4468,
      "loss/crossentropy": 2.0564094200730323,
      "loss/hidden": 3.433984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19553639348596336,
      "step": 4260
    },
    {
      "epoch": 0.10675,
      "grad_norm": 32.0,
      "grad_norm_var": 3.2587890625,
      "learning_rate": 0.0001,
      "loss": 7.4186,
      "loss/crossentropy": 2.13806764036417,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19822277761995793,
      "step": 4270
    },
    {
      "epoch": 0.107,
      "grad_norm": 28.0,
      "grad_norm_var": 1.6926432291666667,
      "learning_rate": 0.0001,
      "loss": 7.4595,
      "loss/crossentropy": 2.0767486467957497,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1884168043732643,
      "step": 4280
    },
    {
      "epoch": 0.10725,
      "grad_norm": 33.0,
      "grad_norm_var": 2.3059895833333335,
      "learning_rate": 0.0001,
      "loss": 7.4481,
      "loss/crossentropy": 2.033916361629963,
      "loss/hidden": 3.45,
      "loss/jsd": 0.0,
      "loss/logits": 0.20558829829096795,
      "step": 4290
    },
    {
      "epoch": 0.1075,
      "grad_norm": 31.0,
      "grad_norm_var": 2.9375,
      "learning_rate": 0.0001,
      "loss": 7.4871,
      "loss/crossentropy": 2.078028707951307,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.188079852424562,
      "step": 4300
    },
    {
      "epoch": 0.10775,
      "grad_norm": 33.25,
      "grad_norm_var": 2.1020833333333333,
      "learning_rate": 0.0001,
      "loss": 7.5379,
      "loss/crossentropy": 2.003500834107399,
      "loss/hidden": 3.544921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20521650360897184,
      "step": 4310
    },
    {
      "epoch": 0.108,
      "grad_norm": 29.625,
      "grad_norm_var": 2.8447916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3536,
      "loss/crossentropy": 2.043112625181675,
      "loss/hidden": 3.380859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19910661596804857,
      "step": 4320
    },
    {
      "epoch": 0.10825,
      "grad_norm": 28.5,
      "grad_norm_var": 4.000455729166666,
      "learning_rate": 0.0001,
      "loss": 7.3717,
      "loss/crossentropy": 2.1422011658549307,
      "loss/hidden": 3.38671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19375871792435645,
      "step": 4330
    },
    {
      "epoch": 0.1085,
      "grad_norm": 29.0,
      "grad_norm_var": 3.6259765625,
      "learning_rate": 0.0001,
      "loss": 7.5021,
      "loss/crossentropy": 2.131446525454521,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2063008865341544,
      "step": 4340
    },
    {
      "epoch": 0.10875,
      "grad_norm": 32.0,
      "grad_norm_var": 5.9525390625,
      "learning_rate": 0.0001,
      "loss": 7.4749,
      "loss/crossentropy": 2.085691845417023,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889802658930421,
      "step": 4350
    },
    {
      "epoch": 0.109,
      "grad_norm": 30.75,
      "grad_norm_var": 3.154166666666667,
      "learning_rate": 0.0001,
      "loss": 7.3816,
      "loss/crossentropy": 1.8972876839339734,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17174729090183974,
      "step": 4360
    },
    {
      "epoch": 0.10925,
      "grad_norm": 29.875,
      "grad_norm_var": 1.7509765625,
      "learning_rate": 0.0001,
      "loss": 7.4444,
      "loss/crossentropy": 2.127763804793358,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18679574280977249,
      "step": 4370
    },
    {
      "epoch": 0.1095,
      "grad_norm": 29.875,
      "grad_norm_var": 2.16015625,
      "learning_rate": 0.0001,
      "loss": 7.4682,
      "loss/crossentropy": 2.1872297644615175,
      "loss/hidden": 3.31328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.183891461789608,
      "step": 4380
    },
    {
      "epoch": 0.10975,
      "grad_norm": 28.875,
      "grad_norm_var": 3.3692057291666666,
      "learning_rate": 0.0001,
      "loss": 7.429,
      "loss/crossentropy": 2.19267495572567,
      "loss/hidden": 3.394140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20111876968294382,
      "step": 4390
    },
    {
      "epoch": 0.11,
      "grad_norm": 29.375,
      "grad_norm_var": 1.6858723958333333,
      "learning_rate": 0.0001,
      "loss": 7.556,
      "loss/crossentropy": 2.1324411287903784,
      "loss/hidden": 3.4359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2090261412784457,
      "step": 4400
    },
    {
      "epoch": 0.11025,
      "grad_norm": 33.5,
      "grad_norm_var": 3.374739583333333,
      "learning_rate": 0.0001,
      "loss": 7.4081,
      "loss/crossentropy": 1.9800483137369156,
      "loss/hidden": 3.584375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19881114605814218,
      "step": 4410
    },
    {
      "epoch": 0.1105,
      "grad_norm": 31.75,
      "grad_norm_var": 4.13515625,
      "learning_rate": 0.0001,
      "loss": 7.4904,
      "loss/crossentropy": 2.053773292154074,
      "loss/hidden": 3.323828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18270381446927786,
      "step": 4420
    },
    {
      "epoch": 0.11075,
      "grad_norm": 30.75,
      "grad_norm_var": 2.3499348958333335,
      "learning_rate": 0.0001,
      "loss": 7.4509,
      "loss/crossentropy": 2.0641689248383046,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19245190378278493,
      "step": 4430
    },
    {
      "epoch": 0.111,
      "grad_norm": 33.25,
      "grad_norm_var": 3.158333333333333,
      "learning_rate": 0.0001,
      "loss": 7.3576,
      "loss/crossentropy": 2.073286408931017,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1892416624352336,
      "step": 4440
    },
    {
      "epoch": 0.11125,
      "grad_norm": 35.75,
      "grad_norm_var": 6.167122395833333,
      "learning_rate": 0.0001,
      "loss": 7.456,
      "loss/crossentropy": 2.191167525947094,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19327596500515937,
      "step": 4450
    },
    {
      "epoch": 0.1115,
      "grad_norm": 28.0,
      "grad_norm_var": 6.762239583333334,
      "learning_rate": 0.0001,
      "loss": 7.4254,
      "loss/crossentropy": 1.9917161837220192,
      "loss/hidden": 3.48828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18673346154391765,
      "step": 4460
    },
    {
      "epoch": 0.11175,
      "grad_norm": 31.0,
      "grad_norm_var": 2.763541666666667,
      "learning_rate": 0.0001,
      "loss": 7.4458,
      "loss/crossentropy": 2.0167058646678924,
      "loss/hidden": 3.477734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20151916183531285,
      "step": 4470
    },
    {
      "epoch": 0.112,
      "grad_norm": 30.5,
      "grad_norm_var": 7.175455729166667,
      "learning_rate": 0.0001,
      "loss": 7.4057,
      "loss/crossentropy": 2.013149876892567,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1819242848083377,
      "step": 4480
    },
    {
      "epoch": 0.11225,
      "grad_norm": 43.25,
      "grad_norm_var": 13.478580729166667,
      "learning_rate": 0.0001,
      "loss": 7.4416,
      "loss/crossentropy": 2.111778366565704,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.20088088884949684,
      "step": 4490
    },
    {
      "epoch": 0.1125,
      "grad_norm": 30.125,
      "grad_norm_var": 11.905143229166667,
      "learning_rate": 0.0001,
      "loss": 7.4435,
      "loss/crossentropy": 2.0223396182060243,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1967620700597763,
      "step": 4500
    },
    {
      "epoch": 0.11275,
      "grad_norm": 28.375,
      "grad_norm_var": 2.2978515625,
      "learning_rate": 0.0001,
      "loss": 7.3969,
      "loss/crossentropy": 1.9966137878596784,
      "loss/hidden": 3.416796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19062119219452142,
      "step": 4510
    },
    {
      "epoch": 0.113,
      "grad_norm": 29.75,
      "grad_norm_var": 3.1759765625,
      "learning_rate": 0.0001,
      "loss": 7.2845,
      "loss/crossentropy": 1.8878834903240205,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16922880560159684,
      "step": 4520
    },
    {
      "epoch": 0.11325,
      "grad_norm": 33.5,
      "grad_norm_var": 3.78515625,
      "learning_rate": 0.0001,
      "loss": 7.5223,
      "loss/crossentropy": 2.0424712359905244,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18261839263141155,
      "step": 4530
    },
    {
      "epoch": 0.1135,
      "grad_norm": 41.75,
      "grad_norm_var": 13.172330729166667,
      "learning_rate": 0.0001,
      "loss": 7.4917,
      "loss/crossentropy": 2.1800880253314974,
      "loss/hidden": 3.419140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1875661849975586,
      "step": 4540
    },
    {
      "epoch": 0.11375,
      "grad_norm": 29.5,
      "grad_norm_var": 13.737239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4929,
      "loss/crossentropy": 2.1130245834589005,
      "loss/hidden": 3.514453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20742647554725407,
      "step": 4550
    },
    {
      "epoch": 0.114,
      "grad_norm": 31.875,
      "grad_norm_var": 3.1447265625,
      "learning_rate": 0.0001,
      "loss": 7.4885,
      "loss/crossentropy": 2.0878429099917413,
      "loss/hidden": 3.4625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19807947240769863,
      "step": 4560
    },
    {
      "epoch": 0.11425,
      "grad_norm": 32.0,
      "grad_norm_var": 1.9080729166666666,
      "learning_rate": 0.0001,
      "loss": 7.412,
      "loss/crossentropy": 2.045598204433918,
      "loss/hidden": 3.43984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19935160782188177,
      "step": 4570
    },
    {
      "epoch": 0.1145,
      "grad_norm": 31.25,
      "grad_norm_var": 2.703285650940459e+18,
      "learning_rate": 0.0001,
      "loss": 7.4112,
      "loss/crossentropy": 1.9612677067518234,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1939171139150858,
      "step": 4580
    },
    {
      "epoch": 0.11475,
      "grad_norm": 30.125,
      "grad_norm_var": 9.067708333333334,
      "learning_rate": 0.0001,
      "loss": 7.4109,
      "loss/crossentropy": 2.066862888634205,
      "loss/hidden": 3.440625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20057452656328678,
      "step": 4590
    },
    {
      "epoch": 0.115,
      "grad_norm": 29.25,
      "grad_norm_var": 6.670833333333333,
      "learning_rate": 0.0001,
      "loss": 7.3857,
      "loss/crossentropy": 2.0378803849220275,
      "loss/hidden": 3.495703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19217969439923763,
      "step": 4600
    },
    {
      "epoch": 0.11525,
      "grad_norm": 32.0,
      "grad_norm_var": 8.108268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4449,
      "loss/crossentropy": 1.9883966132998467,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1796421378850937,
      "step": 4610
    },
    {
      "epoch": 0.1155,
      "grad_norm": 28.5,
      "grad_norm_var": 2.8853515625,
      "learning_rate": 0.0001,
      "loss": 7.43,
      "loss/crossentropy": 2.2122382700443266,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20737907551229,
      "step": 4620
    },
    {
      "epoch": 0.11575,
      "grad_norm": 30.375,
      "grad_norm_var": 3.7968098958333334,
      "learning_rate": 0.0001,
      "loss": 7.3858,
      "loss/crossentropy": 2.0896764233708383,
      "loss/hidden": 3.540234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20905990786850454,
      "step": 4630
    },
    {
      "epoch": 0.116,
      "grad_norm": 27.5,
      "grad_norm_var": 3.6879557291666667,
      "learning_rate": 0.0001,
      "loss": 7.5145,
      "loss/crossentropy": 2.104724445939064,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19548750538378953,
      "step": 4640
    },
    {
      "epoch": 0.11625,
      "grad_norm": 29.875,
      "grad_norm_var": 8.7056640625,
      "learning_rate": 0.0001,
      "loss": 7.4009,
      "loss/crossentropy": 2.155320603400469,
      "loss/hidden": 3.47578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2002986514940858,
      "step": 4650
    },
    {
      "epoch": 0.1165,
      "grad_norm": 27.0,
      "grad_norm_var": 5.1541015625,
      "learning_rate": 0.0001,
      "loss": 7.3193,
      "loss/crossentropy": 2.085461828112602,
      "loss/hidden": 3.38671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1905359473079443,
      "step": 4660
    },
    {
      "epoch": 0.11675,
      "grad_norm": 30.5,
      "grad_norm_var": 1.5926432291666666,
      "learning_rate": 0.0001,
      "loss": 7.3125,
      "loss/crossentropy": 1.9927285239100456,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17640038076788186,
      "step": 4670
    },
    {
      "epoch": 0.117,
      "grad_norm": 33.75,
      "grad_norm_var": 4.747330729166666,
      "learning_rate": 0.0001,
      "loss": 7.469,
      "loss/crossentropy": 2.1633560836315153,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1862495567649603,
      "step": 4680
    },
    {
      "epoch": 0.11725,
      "grad_norm": 28.25,
      "grad_norm_var": 7.198372395833333,
      "learning_rate": 0.0001,
      "loss": 7.4318,
      "loss/crossentropy": 2.2390024289488792,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2097862558439374,
      "step": 4690
    },
    {
      "epoch": 0.1175,
      "grad_norm": 31.375,
      "grad_norm_var": 5.760872395833333,
      "learning_rate": 0.0001,
      "loss": 7.4669,
      "loss/crossentropy": 2.0608770951628683,
      "loss/hidden": 3.4390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19615320730954408,
      "step": 4700
    },
    {
      "epoch": 0.11775,
      "grad_norm": 34.25,
      "grad_norm_var": 4.1894735190686346e+18,
      "learning_rate": 0.0001,
      "loss": 7.4596,
      "loss/crossentropy": 2.0900899082422257,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17933723451569678,
      "step": 4710
    },
    {
      "epoch": 0.118,
      "grad_norm": 29.625,
      "grad_norm_var": 58.10729166666667,
      "learning_rate": 0.0001,
      "loss": 7.3979,
      "loss/crossentropy": 2.094898019731045,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20720194689929486,
      "step": 4720
    },
    {
      "epoch": 0.11825,
      "grad_norm": 30.25,
      "grad_norm_var": 1.98515625,
      "learning_rate": 0.0001,
      "loss": 7.4519,
      "loss/crossentropy": 2.083225329220295,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20777787994593383,
      "step": 4730
    },
    {
      "epoch": 0.1185,
      "grad_norm": 30.375,
      "grad_norm_var": 4.818684895833333,
      "learning_rate": 0.0001,
      "loss": 7.4795,
      "loss/crossentropy": 2.1974314540624618,
      "loss/hidden": 3.38046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19978385213762523,
      "step": 4740
    },
    {
      "epoch": 0.11875,
      "grad_norm": 32.5,
      "grad_norm_var": 3.439322916666667,
      "learning_rate": 0.0001,
      "loss": 7.3843,
      "loss/crossentropy": 1.9562335655093193,
      "loss/hidden": 3.39140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18924889974296094,
      "step": 4750
    },
    {
      "epoch": 0.119,
      "grad_norm": 30.625,
      "grad_norm_var": 1.3015402743274143e+18,
      "learning_rate": 0.0001,
      "loss": 7.5729,
      "loss/crossentropy": 2.0693807609379293,
      "loss/hidden": 3.339453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18801879994571208,
      "step": 4760
    },
    {
      "epoch": 0.11925,
      "grad_norm": 35.25,
      "grad_norm_var": 258.8791015625,
      "learning_rate": 0.0001,
      "loss": 7.3013,
      "loss/crossentropy": 2.0631250627338886,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18974527437239885,
      "step": 4770
    },
    {
      "epoch": 0.1195,
      "grad_norm": 28.625,
      "grad_norm_var": 301.52233072916664,
      "learning_rate": 0.0001,
      "loss": 7.4639,
      "loss/crossentropy": 2.1473939388990404,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19722200892865657,
      "step": 4780
    },
    {
      "epoch": 0.11975,
      "grad_norm": 31.125,
      "grad_norm_var": 25.472330729166668,
      "learning_rate": 0.0001,
      "loss": 7.3161,
      "loss/crossentropy": 2.1767601929605007,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20041130091995002,
      "step": 4790
    },
    {
      "epoch": 0.12,
      "grad_norm": 29.5,
      "grad_norm_var": 2.8580729166666665,
      "learning_rate": 0.0001,
      "loss": 7.3077,
      "loss/crossentropy": 2.0214909121394156,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19553480856120586,
      "step": 4800
    },
    {
      "epoch": 0.12025,
      "grad_norm": 34.25,
      "grad_norm_var": 2.3666015625,
      "learning_rate": 0.0001,
      "loss": 7.4537,
      "loss/crossentropy": 2.092876334488392,
      "loss/hidden": 3.276171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19079044535756112,
      "step": 4810
    },
    {
      "epoch": 0.1205,
      "grad_norm": 28.75,
      "grad_norm_var": 2.1494140625,
      "learning_rate": 0.0001,
      "loss": 7.3579,
      "loss/crossentropy": 2.159788618981838,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20938555523753166,
      "step": 4820
    },
    {
      "epoch": 0.12075,
      "grad_norm": 31.625,
      "grad_norm_var": 1.2635411529466906e+18,
      "learning_rate": 0.0001,
      "loss": 7.3822,
      "loss/crossentropy": 2.221826246380806,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18899439387023448,
      "step": 4830
    },
    {
      "epoch": 0.121,
      "grad_norm": 29.375,
      "grad_norm_var": 7.171875,
      "learning_rate": 0.0001,
      "loss": 7.3649,
      "loss/crossentropy": 2.2076950490474703,
      "loss/hidden": 3.321875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1911212421953678,
      "step": 4840
    },
    {
      "epoch": 0.12125,
      "grad_norm": 28.875,
      "grad_norm_var": 5.397916666666666,
      "learning_rate": 0.0001,
      "loss": 7.2934,
      "loss/crossentropy": 2.1398009806871414,
      "loss/hidden": 3.276953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18104367554187775,
      "step": 4850
    },
    {
      "epoch": 0.1215,
      "grad_norm": 33.25,
      "grad_norm_var": 2.292122395833333,
      "learning_rate": 0.0001,
      "loss": 7.3944,
      "loss/crossentropy": 2.0568679124116898,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19066975675523282,
      "step": 4860
    },
    {
      "epoch": 0.12175,
      "grad_norm": 31.75,
      "grad_norm_var": 1.5145182291666666,
      "learning_rate": 0.0001,
      "loss": 7.5365,
      "loss/crossentropy": 2.2600763499736787,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20988074019551278,
      "step": 4870
    },
    {
      "epoch": 0.122,
      "grad_norm": 30.125,
      "grad_norm_var": 0.8442057291666667,
      "learning_rate": 0.0001,
      "loss": 7.4425,
      "loss/crossentropy": 2.087808459997177,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20126468148082494,
      "step": 4880
    },
    {
      "epoch": 0.12225,
      "grad_norm": 29.25,
      "grad_norm_var": 1.9455729166666667,
      "learning_rate": 0.0001,
      "loss": 7.4649,
      "loss/crossentropy": 2.089573635160923,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18984669484198094,
      "step": 4890
    },
    {
      "epoch": 0.1225,
      "grad_norm": 29.125,
      "grad_norm_var": 2.7552083333333335,
      "learning_rate": 0.0001,
      "loss": 7.4894,
      "loss/crossentropy": 2.1424145482480528,
      "loss/hidden": 3.47890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20886036530137062,
      "step": 4900
    },
    {
      "epoch": 0.12275,
      "grad_norm": 31.0,
      "grad_norm_var": 4.751497395833334,
      "learning_rate": 0.0001,
      "loss": 7.5033,
      "loss/crossentropy": 2.104494086652994,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1945918256416917,
      "step": 4910
    },
    {
      "epoch": 0.123,
      "grad_norm": 28.125,
      "grad_norm_var": 5.330989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4954,
      "loss/crossentropy": 2.0843611776828768,
      "loss/hidden": 3.358203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1925347488373518,
      "step": 4920
    },
    {
      "epoch": 0.12325,
      "grad_norm": 28.625,
      "grad_norm_var": 3.8166015625,
      "learning_rate": 0.0001,
      "loss": 7.4404,
      "loss/crossentropy": 2.205425333976746,
      "loss/hidden": 3.3359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18580489940941333,
      "step": 4930
    },
    {
      "epoch": 0.1235,
      "grad_norm": 29.375,
      "grad_norm_var": 14.980208333333334,
      "learning_rate": 0.0001,
      "loss": 7.3481,
      "loss/crossentropy": 1.9896500617265702,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1904701752588153,
      "step": 4940
    },
    {
      "epoch": 0.12375,
      "grad_norm": 32.75,
      "grad_norm_var": 19.178580729166665,
      "learning_rate": 0.0001,
      "loss": 7.5252,
      "loss/crossentropy": 2.1207278318703175,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19760717861354352,
      "step": 4950
    },
    {
      "epoch": 0.124,
      "grad_norm": 32.5,
      "grad_norm_var": 17.264583333333334,
      "learning_rate": 0.0001,
      "loss": 7.2678,
      "loss/crossentropy": 1.9271991185843944,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19860625620931388,
      "step": 4960
    },
    {
      "epoch": 0.12425,
      "grad_norm": 28.625,
      "grad_norm_var": 11.196809895833333,
      "learning_rate": 0.0001,
      "loss": 7.3703,
      "loss/crossentropy": 2.0659097760915754,
      "loss/hidden": 3.287109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18224728610366583,
      "step": 4970
    },
    {
      "epoch": 0.1245,
      "grad_norm": 37.75,
      "grad_norm_var": 10.03515625,
      "learning_rate": 0.0001,
      "loss": 7.5041,
      "loss/crossentropy": 1.9809176340699195,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19965030066668987,
      "step": 4980
    },
    {
      "epoch": 0.12475,
      "grad_norm": 27.125,
      "grad_norm_var": 11.567708333333334,
      "learning_rate": 0.0001,
      "loss": 7.327,
      "loss/crossentropy": 2.0197409205138683,
      "loss/hidden": 3.368359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18525638189166785,
      "step": 4990
    },
    {
      "epoch": 0.125,
      "grad_norm": 34.75,
      "grad_norm_var": 8.558268229166666,
      "learning_rate": 0.0001,
      "loss": 7.393,
      "loss/crossentropy": 2.100055608153343,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19607669236138464,
      "step": 5000
    },
    {
      "epoch": 0.12525,
      "grad_norm": 29.75,
      "grad_norm_var": 5.602083333333334,
      "learning_rate": 0.0001,
      "loss": 7.4165,
      "loss/crossentropy": 1.9898378394544125,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18131834492087365,
      "step": 5010
    },
    {
      "epoch": 0.1255,
      "grad_norm": 34.75,
      "grad_norm_var": 5.866666666666666,
      "learning_rate": 0.0001,
      "loss": 7.555,
      "loss/crossentropy": 2.086017055809498,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889553153887391,
      "step": 5020
    },
    {
      "epoch": 0.12575,
      "grad_norm": 33.25,
      "grad_norm_var": 9.083268229166666,
      "learning_rate": 0.0001,
      "loss": 7.4098,
      "loss/crossentropy": 2.1133529357612133,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1881294794380665,
      "step": 5030
    },
    {
      "epoch": 0.126,
      "grad_norm": 37.25,
      "grad_norm_var": 14.268489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4117,
      "loss/crossentropy": 2.1818468660116195,
      "loss/hidden": 3.3390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19351670220494271,
      "step": 5040
    },
    {
      "epoch": 0.12625,
      "grad_norm": 27.75,
      "grad_norm_var": 15.270572916666667,
      "learning_rate": 0.0001,
      "loss": 7.448,
      "loss/crossentropy": 2.133790023624897,
      "loss/hidden": 3.267578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17900315206497908,
      "step": 5050
    },
    {
      "epoch": 0.1265,
      "grad_norm": 27.75,
      "grad_norm_var": 12.469791666666667,
      "learning_rate": 0.0001,
      "loss": 7.4514,
      "loss/crossentropy": 2.013399636745453,
      "loss/hidden": 3.502734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19984339475631713,
      "step": 5060
    },
    {
      "epoch": 0.12675,
      "grad_norm": 28.625,
      "grad_norm_var": 6.479622395833333,
      "learning_rate": 0.0001,
      "loss": 7.4411,
      "loss/crossentropy": 2.1552533119916917,
      "loss/hidden": 3.4265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20999168753623962,
      "step": 5070
    },
    {
      "epoch": 0.127,
      "grad_norm": 30.125,
      "grad_norm_var": 4.280989583333334,
      "learning_rate": 0.0001,
      "loss": 7.4238,
      "loss/crossentropy": 2.1047082796692846,
      "loss/hidden": 3.44375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19757428932935,
      "step": 5080
    },
    {
      "epoch": 0.12725,
      "grad_norm": 29.25,
      "grad_norm_var": 3.971875,
      "learning_rate": 0.0001,
      "loss": 7.547,
      "loss/crossentropy": 2.2064288735389708,
      "loss/hidden": 3.472265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2037733059376478,
      "step": 5090
    },
    {
      "epoch": 0.1275,
      "grad_norm": 30.125,
      "grad_norm_var": 3.809309895833333,
      "learning_rate": 0.0001,
      "loss": 7.5359,
      "loss/crossentropy": 2.307460626959801,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2208320491015911,
      "step": 5100
    },
    {
      "epoch": 0.12775,
      "grad_norm": 30.0,
      "grad_norm_var": 6.887434895833334,
      "learning_rate": 0.0001,
      "loss": 7.3695,
      "loss/crossentropy": 2.1241589702665804,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18603499811142682,
      "step": 5110
    },
    {
      "epoch": 0.128,
      "grad_norm": 30.125,
      "grad_norm_var": 1.8353515625,
      "learning_rate": 0.0001,
      "loss": 7.4045,
      "loss/crossentropy": 2.1248120576143266,
      "loss/hidden": 3.364453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19218573588877916,
      "step": 5120
    },
    {
      "epoch": 0.12825,
      "grad_norm": 30.875,
      "grad_norm_var": 18.843489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3683,
      "loss/crossentropy": 2.0221078641712666,
      "loss/hidden": 3.396484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19441262539476156,
      "step": 5130
    },
    {
      "epoch": 0.1285,
      "grad_norm": 30.25,
      "grad_norm_var": 19.755989583333335,
      "learning_rate": 0.0001,
      "loss": 7.4467,
      "loss/crossentropy": 2.0746863678097727,
      "loss/hidden": 3.41015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20940047055482863,
      "step": 5140
    },
    {
      "epoch": 0.12875,
      "grad_norm": 29.75,
      "grad_norm_var": 7.226497395833333,
      "learning_rate": 0.0001,
      "loss": 7.4125,
      "loss/crossentropy": 2.127023458480835,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19320496991276742,
      "step": 5150
    },
    {
      "epoch": 0.129,
      "grad_norm": 30.875,
      "grad_norm_var": 8.332747395833334,
      "learning_rate": 0.0001,
      "loss": 7.3609,
      "loss/crossentropy": 2.0404578357934953,
      "loss/hidden": 3.271484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18524497244507074,
      "step": 5160
    },
    {
      "epoch": 0.12925,
      "grad_norm": 30.375,
      "grad_norm_var": 5.566080729166667,
      "learning_rate": 0.0001,
      "loss": 7.3806,
      "loss/crossentropy": 2.05174797475338,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20412184661254287,
      "step": 5170
    },
    {
      "epoch": 0.1295,
      "grad_norm": 30.875,
      "grad_norm_var": 72.65201822916667,
      "learning_rate": 0.0001,
      "loss": 7.4596,
      "loss/crossentropy": 2.0945761643350123,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1920377543196082,
      "step": 5180
    },
    {
      "epoch": 0.12975,
      "grad_norm": 34.25,
      "grad_norm_var": 1.8330729166666666,
      "learning_rate": 0.0001,
      "loss": 7.3953,
      "loss/crossentropy": 2.0848742216825484,
      "loss/hidden": 3.548046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22283064387738705,
      "step": 5190
    },
    {
      "epoch": 0.13,
      "grad_norm": 31.25,
      "grad_norm_var": 2.4244140625,
      "learning_rate": 0.0001,
      "loss": 7.5061,
      "loss/crossentropy": 1.997230054438114,
      "loss/hidden": 3.488671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.198976163379848,
      "step": 5200
    },
    {
      "epoch": 0.13025,
      "grad_norm": 42.0,
      "grad_norm_var": 9.762239583333333,
      "learning_rate": 0.0001,
      "loss": 7.523,
      "loss/crossentropy": 2.169138702750206,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2180183682590723,
      "step": 5210
    },
    {
      "epoch": 0.1305,
      "grad_norm": 27.75,
      "grad_norm_var": 13.330989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4174,
      "loss/crossentropy": 2.0436717979609966,
      "loss/hidden": 3.4265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18874377477914095,
      "step": 5220
    },
    {
      "epoch": 0.13075,
      "grad_norm": 33.5,
      "grad_norm_var": 5.186393229166667,
      "learning_rate": 0.0001,
      "loss": 7.4735,
      "loss/crossentropy": 2.1061771392822264,
      "loss/hidden": 3.3125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1841479053720832,
      "step": 5230
    },
    {
      "epoch": 0.131,
      "grad_norm": 28.25,
      "grad_norm_var": 3.480208333333333,
      "learning_rate": 0.0001,
      "loss": 7.4379,
      "loss/crossentropy": 1.9957973182201385,
      "loss/hidden": 3.46171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19976749327033758,
      "step": 5240
    },
    {
      "epoch": 0.13125,
      "grad_norm": 31.0,
      "grad_norm_var": 2.2249348958333335,
      "learning_rate": 0.0001,
      "loss": 7.445,
      "loss/crossentropy": 2.089694794267416,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18123079631477595,
      "step": 5250
    },
    {
      "epoch": 0.1315,
      "grad_norm": 30.625,
      "grad_norm_var": 2.062239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3525,
      "loss/crossentropy": 2.096596322953701,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17838086038827897,
      "step": 5260
    },
    {
      "epoch": 0.13175,
      "grad_norm": 27.75,
      "grad_norm_var": 2.6619140625,
      "learning_rate": 0.0001,
      "loss": 7.4042,
      "loss/crossentropy": 2.086874121427536,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18623477015644313,
      "step": 5270
    },
    {
      "epoch": 0.132,
      "grad_norm": 31.125,
      "grad_norm_var": 1.8416015625,
      "learning_rate": 0.0001,
      "loss": 7.3634,
      "loss/crossentropy": 2.131754931807518,
      "loss/hidden": 3.416796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20106223467737436,
      "step": 5280
    },
    {
      "epoch": 0.13225,
      "grad_norm": 32.5,
      "grad_norm_var": 6.598372395833334,
      "learning_rate": 0.0001,
      "loss": 7.4969,
      "loss/crossentropy": 2.1548122704029082,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20439809635281564,
      "step": 5290
    },
    {
      "epoch": 0.1325,
      "grad_norm": 28.125,
      "grad_norm_var": 5.252083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4373,
      "loss/crossentropy": 2.1770398393273354,
      "loss/hidden": 3.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19656166546046733,
      "step": 5300
    },
    {
      "epoch": 0.13275,
      "grad_norm": 29.875,
      "grad_norm_var": 1.2473307291666667,
      "learning_rate": 0.0001,
      "loss": 7.4308,
      "loss/crossentropy": 2.152033807337284,
      "loss/hidden": 3.3046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18985964702442287,
      "step": 5310
    },
    {
      "epoch": 0.133,
      "grad_norm": 31.625,
      "grad_norm_var": 1.0041666666666667,
      "learning_rate": 0.0001,
      "loss": 7.4327,
      "loss/crossentropy": 2.087932828068733,
      "loss/hidden": 3.3390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.184254783205688,
      "step": 5320
    },
    {
      "epoch": 0.13325,
      "grad_norm": 31.0,
      "grad_norm_var": 2.4344770491390623e+18,
      "learning_rate": 0.0001,
      "loss": 7.3366,
      "loss/crossentropy": 2.034941144287586,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1904723599553108,
      "step": 5330
    },
    {
      "epoch": 0.1335,
      "grad_norm": 31.5,
      "grad_norm_var": 6.255989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4492,
      "loss/crossentropy": 2.152811796963215,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18841406889259815,
      "step": 5340
    },
    {
      "epoch": 0.13375,
      "grad_norm": 31.625,
      "grad_norm_var": 1.6822916666666667,
      "learning_rate": 0.0001,
      "loss": 7.465,
      "loss/crossentropy": 2.1877569228410723,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19606791157275438,
      "step": 5350
    },
    {
      "epoch": 0.134,
      "grad_norm": 30.25,
      "grad_norm_var": 2.17890625,
      "learning_rate": 0.0001,
      "loss": 7.3975,
      "loss/crossentropy": 2.0353255167603495,
      "loss/hidden": 3.340234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1815076546743512,
      "step": 5360
    },
    {
      "epoch": 0.13425,
      "grad_norm": 31.0,
      "grad_norm_var": 12.4525390625,
      "learning_rate": 0.0001,
      "loss": 7.3462,
      "loss/crossentropy": 1.919140312820673,
      "loss/hidden": 3.370703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18229803508147596,
      "step": 5370
    },
    {
      "epoch": 0.1345,
      "grad_norm": 29.625,
      "grad_norm_var": 7.26640625,
      "learning_rate": 0.0001,
      "loss": 7.506,
      "loss/crossentropy": 2.11019846200943,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19277678560465575,
      "step": 5380
    },
    {
      "epoch": 0.13475,
      "grad_norm": 32.0,
      "grad_norm_var": 0.9275390625,
      "learning_rate": 0.0001,
      "loss": 7.3668,
      "loss/crossentropy": 2.1108837127685547,
      "loss/hidden": 3.493359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20013203900307416,
      "step": 5390
    },
    {
      "epoch": 0.135,
      "grad_norm": 28.125,
      "grad_norm_var": 1.7330729166666667,
      "learning_rate": 0.0001,
      "loss": 7.5524,
      "loss/crossentropy": 2.16382010653615,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1924815428443253,
      "step": 5400
    },
    {
      "epoch": 0.13525,
      "grad_norm": 30.25,
      "grad_norm_var": 2.5893229166666667,
      "learning_rate": 0.0001,
      "loss": 7.5393,
      "loss/crossentropy": 2.0622613176703455,
      "loss/hidden": 3.498828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20035731326788664,
      "step": 5410
    },
    {
      "epoch": 0.1355,
      "grad_norm": 29.625,
      "grad_norm_var": 2.8499348958333335,
      "learning_rate": 0.0001,
      "loss": 7.3466,
      "loss/crossentropy": 2.169532992690802,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2085475005209446,
      "step": 5420
    },
    {
      "epoch": 0.13575,
      "grad_norm": 31.0,
      "grad_norm_var": 2.8212890625,
      "learning_rate": 0.0001,
      "loss": 7.4299,
      "loss/crossentropy": 2.056758251786232,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19217969793826342,
      "step": 5430
    },
    {
      "epoch": 0.136,
      "grad_norm": 32.25,
      "grad_norm_var": 3.9749348958333335,
      "learning_rate": 0.0001,
      "loss": 7.4334,
      "loss/crossentropy": 2.1805212616920473,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21980819348245859,
      "step": 5440
    },
    {
      "epoch": 0.13625,
      "grad_norm": 29.5,
      "grad_norm_var": 2.0218098958333335,
      "learning_rate": 0.0001,
      "loss": 7.5382,
      "loss/crossentropy": 2.1516773015260697,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.208776849322021,
      "step": 5450
    },
    {
      "epoch": 0.1365,
      "grad_norm": 31.375,
      "grad_norm_var": 2.082747395833333,
      "learning_rate": 0.0001,
      "loss": 7.4684,
      "loss/crossentropy": 2.1602507561445234,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1842126866802573,
      "step": 5460
    },
    {
      "epoch": 0.13675,
      "grad_norm": 31.25,
      "grad_norm_var": 2.6197265625,
      "learning_rate": 0.0001,
      "loss": 7.3823,
      "loss/crossentropy": 2.081377077102661,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1906685210764408,
      "step": 5470
    },
    {
      "epoch": 0.137,
      "grad_norm": 30.875,
      "grad_norm_var": 2.21875,
      "learning_rate": 0.0001,
      "loss": 7.4128,
      "loss/crossentropy": 2.138934540748596,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18631890565156936,
      "step": 5480
    },
    {
      "epoch": 0.13725,
      "grad_norm": 31.0,
      "grad_norm_var": 3.958268229166667,
      "learning_rate": 0.0001,
      "loss": 7.568,
      "loss/crossentropy": 2.02208868265152,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19863407909870148,
      "step": 5490
    },
    {
      "epoch": 0.1375,
      "grad_norm": 33.0,
      "grad_norm_var": 6.1775390625,
      "learning_rate": 0.0001,
      "loss": 7.3806,
      "loss/crossentropy": 2.0247954726219177,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18538292730227113,
      "step": 5500
    },
    {
      "epoch": 0.13775,
      "grad_norm": 30.625,
      "grad_norm_var": 6.076041666666667,
      "learning_rate": 0.0001,
      "loss": 7.6241,
      "loss/crossentropy": 2.2269895624369385,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.197488261340186,
      "step": 5510
    },
    {
      "epoch": 0.138,
      "grad_norm": 30.25,
      "grad_norm_var": 4.1947265625,
      "learning_rate": 0.0001,
      "loss": 7.4968,
      "loss/crossentropy": 2.077942840754986,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19135653134435415,
      "step": 5520
    },
    {
      "epoch": 0.13825,
      "grad_norm": 27.75,
      "grad_norm_var": 2.701041666666667,
      "learning_rate": 0.0001,
      "loss": 7.3842,
      "loss/crossentropy": 2.104434663057327,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18928063409402968,
      "step": 5530
    },
    {
      "epoch": 0.1385,
      "grad_norm": 31.5,
      "grad_norm_var": 6.4087890625,
      "learning_rate": 0.0001,
      "loss": 7.443,
      "loss/crossentropy": 2.0420807294547556,
      "loss/hidden": 3.362890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18584198467433452,
      "step": 5540
    },
    {
      "epoch": 0.13875,
      "grad_norm": 32.5,
      "grad_norm_var": 4.7556640625,
      "learning_rate": 0.0001,
      "loss": 7.418,
      "loss/crossentropy": 2.0941856279969215,
      "loss/hidden": 3.311328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18290557386353612,
      "step": 5550
    },
    {
      "epoch": 0.139,
      "grad_norm": 34.0,
      "grad_norm_var": 3.0947916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3594,
      "loss/crossentropy": 2.1482032746076585,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19914243686944247,
      "step": 5560
    },
    {
      "epoch": 0.13925,
      "grad_norm": 31.875,
      "grad_norm_var": 4.449739583333334,
      "learning_rate": 0.0001,
      "loss": 7.5652,
      "loss/crossentropy": 2.150991679728031,
      "loss/hidden": 3.4984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20609580241143705,
      "step": 5570
    },
    {
      "epoch": 0.1395,
      "grad_norm": 30.125,
      "grad_norm_var": 8.527083333333334,
      "learning_rate": 0.0001,
      "loss": 7.3112,
      "loss/crossentropy": 2.1712302803993224,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18877983894199132,
      "step": 5580
    },
    {
      "epoch": 0.13975,
      "grad_norm": 28.5,
      "grad_norm_var": 2.1197265625,
      "learning_rate": 0.0001,
      "loss": 7.4033,
      "loss/crossentropy": 2.1502134561538697,
      "loss/hidden": 3.4453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21474836114794016,
      "step": 5590
    },
    {
      "epoch": 0.14,
      "grad_norm": 30.25,
      "grad_norm_var": 3.6809895833333335,
      "learning_rate": 0.0001,
      "loss": 7.5089,
      "loss/crossentropy": 2.230164831876755,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2127472611144185,
      "step": 5600
    },
    {
      "epoch": 0.14025,
      "grad_norm": 30.625,
      "grad_norm_var": 52.13333333333333,
      "learning_rate": 0.0001,
      "loss": 7.5328,
      "loss/crossentropy": 2.1207681491971018,
      "loss/hidden": 3.459765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20428987089544534,
      "step": 5610
    },
    {
      "epoch": 0.1405,
      "grad_norm": 30.625,
      "grad_norm_var": 2.037239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3972,
      "loss/crossentropy": 2.065328547358513,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19245364069938659,
      "step": 5620
    },
    {
      "epoch": 0.14075,
      "grad_norm": 29.375,
      "grad_norm_var": 2.40625,
      "learning_rate": 0.0001,
      "loss": 7.2803,
      "loss/crossentropy": 2.0791175961494446,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1857742078602314,
      "step": 5630
    },
    {
      "epoch": 0.141,
      "grad_norm": 29.0,
      "grad_norm_var": 2.687239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3324,
      "loss/crossentropy": 2.054654690623283,
      "loss/hidden": 3.31328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18879605047404766,
      "step": 5640
    },
    {
      "epoch": 0.14125,
      "grad_norm": 27.625,
      "grad_norm_var": 3.06015625,
      "learning_rate": 0.0001,
      "loss": 7.4804,
      "loss/crossentropy": 2.163857588917017,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2018537001684308,
      "step": 5650
    },
    {
      "epoch": 0.1415,
      "grad_norm": 31.625,
      "grad_norm_var": 3.54140625,
      "learning_rate": 0.0001,
      "loss": 7.4337,
      "loss/crossentropy": 2.1230690620839594,
      "loss/hidden": 3.33046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18610329292714595,
      "step": 5660
    },
    {
      "epoch": 0.14175,
      "grad_norm": 30.0,
      "grad_norm_var": 2.0171223958333333,
      "learning_rate": 0.0001,
      "loss": 7.484,
      "loss/crossentropy": 2.1232656478881835,
      "loss/hidden": 3.306640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18209880087524652,
      "step": 5670
    },
    {
      "epoch": 0.142,
      "grad_norm": 30.625,
      "grad_norm_var": 3.6244140625,
      "learning_rate": 0.0001,
      "loss": 7.3601,
      "loss/crossentropy": 1.9925632011145353,
      "loss/hidden": 3.487109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18384792990982532,
      "step": 5680
    },
    {
      "epoch": 0.14225,
      "grad_norm": 28.875,
      "grad_norm_var": 4.595247395833334,
      "learning_rate": 0.0001,
      "loss": 7.3492,
      "loss/crossentropy": 1.9747695334255695,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18710496351122857,
      "step": 5690
    },
    {
      "epoch": 0.1425,
      "grad_norm": 30.0,
      "grad_norm_var": 5.160872395833334,
      "learning_rate": 0.0001,
      "loss": 7.1869,
      "loss/crossentropy": 1.9229816131293773,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1764959843829274,
      "step": 5700
    },
    {
      "epoch": 0.14275,
      "grad_norm": 31.375,
      "grad_norm_var": 1.1177083333333333,
      "learning_rate": 0.0001,
      "loss": 7.5353,
      "loss/crossentropy": 2.1759460479021073,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2011772884055972,
      "step": 5710
    },
    {
      "epoch": 0.143,
      "grad_norm": 30.875,
      "grad_norm_var": 19.32265625,
      "learning_rate": 0.0001,
      "loss": 7.4364,
      "loss/crossentropy": 1.9983633741736413,
      "loss/hidden": 3.44453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2036839971318841,
      "step": 5720
    },
    {
      "epoch": 0.14325,
      "grad_norm": 31.75,
      "grad_norm_var": 6.330989583333333,
      "learning_rate": 0.0001,
      "loss": 7.3473,
      "loss/crossentropy": 2.2608693316578865,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20173839703202248,
      "step": 5730
    },
    {
      "epoch": 0.1435,
      "grad_norm": 29.0,
      "grad_norm_var": 30.8306640625,
      "learning_rate": 0.0001,
      "loss": 7.4861,
      "loss/crossentropy": 2.191919285058975,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1848000530153513,
      "step": 5740
    },
    {
      "epoch": 0.14375,
      "grad_norm": 32.5,
      "grad_norm_var": 8.167643229166666,
      "learning_rate": 0.0001,
      "loss": 7.5179,
      "loss/crossentropy": 2.098912109434605,
      "loss/hidden": 3.544921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22364525627344847,
      "step": 5750
    },
    {
      "epoch": 0.144,
      "grad_norm": 30.5,
      "grad_norm_var": 2.5479166666666666,
      "learning_rate": 0.0001,
      "loss": 7.4241,
      "loss/crossentropy": 2.089163874089718,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19125983892008663,
      "step": 5760
    },
    {
      "epoch": 0.14425,
      "grad_norm": 29.875,
      "grad_norm_var": 8.223958333333334,
      "learning_rate": 0.0001,
      "loss": 7.45,
      "loss/crossentropy": 2.2600366115570067,
      "loss/hidden": 3.444140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1984367400407791,
      "step": 5770
    },
    {
      "epoch": 0.1445,
      "grad_norm": 32.0,
      "grad_norm_var": 14.806705729166667,
      "learning_rate": 0.0001,
      "loss": 7.3632,
      "loss/crossentropy": 1.9510320864617825,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18547183061018585,
      "step": 5780
    },
    {
      "epoch": 0.14475,
      "grad_norm": 30.0,
      "grad_norm_var": 9.655989583333334,
      "learning_rate": 0.0001,
      "loss": 7.5503,
      "loss/crossentropy": 2.143619356304407,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.20010631643235682,
      "step": 5790
    },
    {
      "epoch": 0.145,
      "grad_norm": 28.375,
      "grad_norm_var": 9.556184895833333,
      "learning_rate": 0.0001,
      "loss": 7.5076,
      "loss/crossentropy": 1.9316529139876366,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18306834027171134,
      "step": 5800
    },
    {
      "epoch": 0.14525,
      "grad_norm": 37.25,
      "grad_norm_var": 10.718489583333334,
      "learning_rate": 0.0001,
      "loss": 7.5004,
      "loss/crossentropy": 2.136544609069824,
      "loss/hidden": 3.3453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20682645812630654,
      "step": 5810
    },
    {
      "epoch": 0.1455,
      "grad_norm": 34.75,
      "grad_norm_var": 9.395572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4109,
      "loss/crossentropy": 2.0811544865369798,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.187607554346323,
      "step": 5820
    },
    {
      "epoch": 0.14575,
      "grad_norm": 30.125,
      "grad_norm_var": 11.476041666666667,
      "learning_rate": 0.0001,
      "loss": 7.4615,
      "loss/crossentropy": 2.2464685887098312,
      "loss/hidden": 3.419140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21834317222237587,
      "step": 5830
    },
    {
      "epoch": 0.146,
      "grad_norm": 31.875,
      "grad_norm_var": 8.106705729166666,
      "learning_rate": 0.0001,
      "loss": 7.379,
      "loss/crossentropy": 2.1494223892688753,
      "loss/hidden": 3.4875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20599585752934219,
      "step": 5840
    },
    {
      "epoch": 0.14625,
      "grad_norm": 32.25,
      "grad_norm_var": 119.0541015625,
      "learning_rate": 0.0001,
      "loss": 7.4223,
      "loss/crossentropy": 2.013238602876663,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17934355642646552,
      "step": 5850
    },
    {
      "epoch": 0.1465,
      "grad_norm": 55.5,
      "grad_norm_var": 40.619205729166666,
      "learning_rate": 0.0001,
      "loss": 7.4766,
      "loss/crossentropy": 2.1309464499354362,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1953899236395955,
      "step": 5860
    },
    {
      "epoch": 0.14675,
      "grad_norm": 31.125,
      "grad_norm_var": 51.764322916666664,
      "learning_rate": 0.0001,
      "loss": 7.5385,
      "loss/crossentropy": 2.203585295379162,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20035494081676006,
      "step": 5870
    },
    {
      "epoch": 0.147,
      "grad_norm": 30.125,
      "grad_norm_var": 8.2541015625,
      "learning_rate": 0.0001,
      "loss": 7.3861,
      "loss/crossentropy": 2.057890709489584,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18680873457342387,
      "step": 5880
    },
    {
      "epoch": 0.14725,
      "grad_norm": 30.25,
      "grad_norm_var": 4.555989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4283,
      "loss/crossentropy": 2.049139867722988,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1856512013822794,
      "step": 5890
    },
    {
      "epoch": 0.1475,
      "grad_norm": 30.375,
      "grad_norm_var": 10.570768229166667,
      "learning_rate": 0.0001,
      "loss": 7.4651,
      "loss/crossentropy": 2.0553019613027574,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18676785845309496,
      "step": 5900
    },
    {
      "epoch": 0.14775,
      "grad_norm": 30.125,
      "grad_norm_var": 14.59140625,
      "learning_rate": 0.0001,
      "loss": 7.4727,
      "loss/crossentropy": 2.0098409935832025,
      "loss/hidden": 3.48125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20080684809945523,
      "step": 5910
    },
    {
      "epoch": 0.148,
      "grad_norm": 28.625,
      "grad_norm_var": 8.489322916666667,
      "learning_rate": 0.0001,
      "loss": 7.4203,
      "loss/crossentropy": 2.2615666806697847,
      "loss/hidden": 3.3,
      "loss/jsd": 0.0,
      "loss/logits": 0.18850413355976342,
      "step": 5920
    },
    {
      "epoch": 0.14825,
      "grad_norm": 30.0,
      "grad_norm_var": 3.6233723958333335,
      "learning_rate": 0.0001,
      "loss": 7.4441,
      "loss/crossentropy": 2.178256964683533,
      "loss/hidden": 3.29609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19248567353934048,
      "step": 5930
    },
    {
      "epoch": 0.1485,
      "grad_norm": 31.0,
      "grad_norm_var": 3.455208333333333,
      "learning_rate": 0.0001,
      "loss": 7.4573,
      "loss/crossentropy": 2.246034747362137,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2096735591068864,
      "step": 5940
    },
    {
      "epoch": 0.14875,
      "grad_norm": 30.625,
      "grad_norm_var": 3.8494140625,
      "learning_rate": 0.0001,
      "loss": 7.4811,
      "loss/crossentropy": 2.180899788439274,
      "loss/hidden": 3.341796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19460927378386259,
      "step": 5950
    },
    {
      "epoch": 0.149,
      "grad_norm": 30.5,
      "grad_norm_var": 3.4385416666666666,
      "learning_rate": 0.0001,
      "loss": 7.3829,
      "loss/crossentropy": 2.258976912498474,
      "loss/hidden": 3.332421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19133044108748437,
      "step": 5960
    },
    {
      "epoch": 0.14925,
      "grad_norm": 29.0,
      "grad_norm_var": 13.80390625,
      "learning_rate": 0.0001,
      "loss": 7.5384,
      "loss/crossentropy": 2.012222741544247,
      "loss/hidden": 3.4640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19505210760980846,
      "step": 5970
    },
    {
      "epoch": 0.1495,
      "grad_norm": 27.875,
      "grad_norm_var": 12.9369140625,
      "learning_rate": 0.0001,
      "loss": 7.4033,
      "loss/crossentropy": 2.0392286255955696,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1930427584797144,
      "step": 5980
    },
    {
      "epoch": 0.14975,
      "grad_norm": 28.25,
      "grad_norm_var": 18.174739583333334,
      "learning_rate": 0.0001,
      "loss": 7.399,
      "loss/crossentropy": 1.9029529005289079,
      "loss/hidden": 3.462109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.198425155505538,
      "step": 5990
    },
    {
      "epoch": 0.15,
      "grad_norm": 31.875,
      "grad_norm_var": 3.0737770860662226e+18,
      "learning_rate": 0.0001,
      "loss": 7.4994,
      "loss/crossentropy": 1.8985859856009484,
      "loss/hidden": 3.4609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1951824951916933,
      "step": 6000
    },
    {
      "epoch": 0.15025,
      "grad_norm": 36.0,
      "grad_norm_var": 3.073777086665239e+18,
      "learning_rate": 0.0001,
      "loss": 7.4659,
      "loss/crossentropy": 2.097201645374298,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18253911342471837,
      "step": 6010
    },
    {
      "epoch": 0.1505,
      "grad_norm": 27.875,
      "grad_norm_var": 6.801041666666666,
      "learning_rate": 0.0001,
      "loss": 7.2415,
      "loss/crossentropy": 2.0210610911250115,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.197306059114635,
      "step": 6020
    },
    {
      "epoch": 0.15075,
      "grad_norm": 31.0,
      "grad_norm_var": 14.46640625,
      "learning_rate": 0.0001,
      "loss": 7.4519,
      "loss/crossentropy": 2.1985476523637772,
      "loss/hidden": 3.475,
      "loss/jsd": 0.0,
      "loss/logits": 0.20262509360909461,
      "step": 6030
    },
    {
      "epoch": 0.151,
      "grad_norm": 30.625,
      "grad_norm_var": 6.254622395833334,
      "learning_rate": 0.0001,
      "loss": 7.3353,
      "loss/crossentropy": 2.0093181416392327,
      "loss/hidden": 3.316015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17620250331237913,
      "step": 6040
    },
    {
      "epoch": 0.15125,
      "grad_norm": 31.625,
      "grad_norm_var": 56.4291015625,
      "learning_rate": 0.0001,
      "loss": 7.4034,
      "loss/crossentropy": 2.177773226797581,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20441538300365208,
      "step": 6050
    },
    {
      "epoch": 0.1515,
      "grad_norm": 26.75,
      "grad_norm_var": 55.889322916666664,
      "learning_rate": 0.0001,
      "loss": 7.3245,
      "loss/crossentropy": 2.1666259437799456,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19311951845884323,
      "step": 6060
    },
    {
      "epoch": 0.15175,
      "grad_norm": 30.25,
      "grad_norm_var": 91.0103515625,
      "learning_rate": 0.0001,
      "loss": 7.368,
      "loss/crossentropy": 2.063462796807289,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1834208857268095,
      "step": 6070
    },
    {
      "epoch": 0.152,
      "grad_norm": 30.5,
      "grad_norm_var": 18.1212890625,
      "learning_rate": 0.0001,
      "loss": 7.4335,
      "loss/crossentropy": 1.9907098844647408,
      "loss/hidden": 3.53671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20707368329167367,
      "step": 6080
    },
    {
      "epoch": 0.15225,
      "grad_norm": 38.25,
      "grad_norm_var": 11.470247395833333,
      "learning_rate": 0.0001,
      "loss": 7.3789,
      "loss/crossentropy": 2.083692157268524,
      "loss/hidden": 3.465625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1846569798886776,
      "step": 6090
    },
    {
      "epoch": 0.1525,
      "grad_norm": 28.25,
      "grad_norm_var": 21.829622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3767,
      "loss/crossentropy": 2.1113929279148578,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1990992769598961,
      "step": 6100
    },
    {
      "epoch": 0.15275,
      "grad_norm": 33.75,
      "grad_norm_var": 20.676497395833334,
      "learning_rate": 0.0001,
      "loss": 7.339,
      "loss/crossentropy": 2.1296695113182067,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1936045665293932,
      "step": 6110
    },
    {
      "epoch": 0.153,
      "grad_norm": 36.25,
      "grad_norm_var": 7.8166015625,
      "learning_rate": 0.0001,
      "loss": 7.5027,
      "loss/crossentropy": 2.1011226207017897,
      "loss/hidden": 3.471875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20695240292698144,
      "step": 6120
    },
    {
      "epoch": 0.15325,
      "grad_norm": 30.625,
      "grad_norm_var": 2.1304840750224113e+18,
      "learning_rate": 0.0001,
      "loss": 7.506,
      "loss/crossentropy": 2.2427969723939896,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1966634625568986,
      "step": 6130
    },
    {
      "epoch": 0.1535,
      "grad_norm": 34.25,
      "grad_norm_var": 36.542643229166664,
      "learning_rate": 0.0001,
      "loss": 7.4413,
      "loss/crossentropy": 2.0855264641344546,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19436217453330756,
      "step": 6140
    },
    {
      "epoch": 0.15375,
      "grad_norm": 32.75,
      "grad_norm_var": 10.153580729166666,
      "learning_rate": 0.0001,
      "loss": 7.3096,
      "loss/crossentropy": 2.0322439685463904,
      "loss/hidden": 3.3453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1753252800554037,
      "step": 6150
    },
    {
      "epoch": 0.154,
      "grad_norm": 31.0,
      "grad_norm_var": 10.216080729166666,
      "learning_rate": 0.0001,
      "loss": 7.2481,
      "loss/crossentropy": 2.074477408081293,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17578690703958272,
      "step": 6160
    },
    {
      "epoch": 0.15425,
      "grad_norm": 33.25,
      "grad_norm_var": 28.79765625,
      "learning_rate": 0.0001,
      "loss": 7.4403,
      "loss/crossentropy": 2.0863804474473,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20699662044644357,
      "step": 6170
    },
    {
      "epoch": 0.1545,
      "grad_norm": 33.5,
      "grad_norm_var": 24.84375,
      "learning_rate": 0.0001,
      "loss": 7.4609,
      "loss/crossentropy": 2.0696858704090118,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2128069180995226,
      "step": 6180
    },
    {
      "epoch": 0.15475,
      "grad_norm": 31.375,
      "grad_norm_var": 5.099739583333333,
      "learning_rate": 0.0001,
      "loss": 7.2575,
      "loss/crossentropy": 2.182169410586357,
      "loss/hidden": 3.31796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.181897877715528,
      "step": 6190
    },
    {
      "epoch": 0.155,
      "grad_norm": 34.25,
      "grad_norm_var": 4.699934895833334,
      "learning_rate": 0.0001,
      "loss": 7.4975,
      "loss/crossentropy": 2.165008749067783,
      "loss/hidden": 3.492578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19800901636481286,
      "step": 6200
    },
    {
      "epoch": 0.15525,
      "grad_norm": 33.75,
      "grad_norm_var": 2.7738932291666667,
      "learning_rate": 0.0001,
      "loss": 7.3992,
      "loss/crossentropy": 2.0653695166110992,
      "loss/hidden": 3.406640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19180236533284187,
      "step": 6210
    },
    {
      "epoch": 0.1555,
      "grad_norm": 35.25,
      "grad_norm_var": 5.31015625,
      "learning_rate": 0.0001,
      "loss": 7.3638,
      "loss/crossentropy": 2.1244568385183813,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18238217020407319,
      "step": 6220
    },
    {
      "epoch": 0.15575,
      "grad_norm": 32.25,
      "grad_norm_var": 3.2817057291666667,
      "learning_rate": 0.0001,
      "loss": 7.5073,
      "loss/crossentropy": 2.1881898671388624,
      "loss/hidden": 3.498828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2045454490929842,
      "step": 6230
    },
    {
      "epoch": 0.156,
      "grad_norm": 28.25,
      "grad_norm_var": 2.6572265625,
      "learning_rate": 0.0001,
      "loss": 7.4609,
      "loss/crossentropy": 2.14600064009428,
      "loss/hidden": 3.4265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18945380430668593,
      "step": 6240
    },
    {
      "epoch": 0.15625,
      "grad_norm": 30.75,
      "grad_norm_var": 35.73118489583333,
      "learning_rate": 0.0001,
      "loss": 7.3786,
      "loss/crossentropy": 2.168429624289274,
      "loss/hidden": 3.296484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18439108245074748,
      "step": 6250
    },
    {
      "epoch": 0.1565,
      "grad_norm": 52.5,
      "grad_norm_var": 64.9962890625,
      "learning_rate": 0.0001,
      "loss": 7.3511,
      "loss/crossentropy": 2.1293379329144955,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1828605517745018,
      "step": 6260
    },
    {
      "epoch": 0.15675,
      "grad_norm": 29.25,
      "grad_norm_var": 59.703125,
      "learning_rate": 0.0001,
      "loss": 7.3978,
      "loss/crossentropy": 1.8641120925545693,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18577109538018705,
      "step": 6270
    },
    {
      "epoch": 0.157,
      "grad_norm": 28.75,
      "grad_norm_var": 32.1994140625,
      "learning_rate": 0.0001,
      "loss": 7.4066,
      "loss/crossentropy": 2.0997040398418902,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18688563201576472,
      "step": 6280
    },
    {
      "epoch": 0.15725,
      "grad_norm": 31.375,
      "grad_norm_var": 17.302018229166666,
      "learning_rate": 0.0001,
      "loss": 7.4043,
      "loss/crossentropy": 1.9712626039981842,
      "loss/hidden": 3.45078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20054549565538765,
      "step": 6290
    },
    {
      "epoch": 0.1575,
      "grad_norm": 31.5,
      "grad_norm_var": 17.8431640625,
      "learning_rate": 0.0001,
      "loss": 7.4502,
      "loss/crossentropy": 2.0252815186977386,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18488127905875446,
      "step": 6300
    },
    {
      "epoch": 0.15775,
      "grad_norm": 30.75,
      "grad_norm_var": 7.995572916666666,
      "learning_rate": 0.0001,
      "loss": 7.3829,
      "loss/crossentropy": 2.030302118510008,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19101340658962726,
      "step": 6310
    },
    {
      "epoch": 0.158,
      "grad_norm": 30.125,
      "grad_norm_var": 5.805143229166666,
      "learning_rate": 0.0001,
      "loss": 7.3852,
      "loss/crossentropy": 1.9795936658978461,
      "loss/hidden": 3.45703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19110800279304385,
      "step": 6320
    },
    {
      "epoch": 0.15825,
      "grad_norm": 34.0,
      "grad_norm_var": 6.91640625,
      "learning_rate": 0.0001,
      "loss": 7.4417,
      "loss/crossentropy": 2.0620448149740698,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.194018579646945,
      "step": 6330
    },
    {
      "epoch": 0.1585,
      "grad_norm": 28.125,
      "grad_norm_var": 31.058268229166668,
      "learning_rate": 0.0001,
      "loss": 7.4142,
      "loss/crossentropy": 2.012200343608856,
      "loss/hidden": 3.4921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21260247621685266,
      "step": 6340
    },
    {
      "epoch": 0.15875,
      "grad_norm": 36.75,
      "grad_norm_var": 35.18118489583333,
      "learning_rate": 0.0001,
      "loss": 7.3776,
      "loss/crossentropy": 1.9757203698158263,
      "loss/hidden": 3.382421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.200297892652452,
      "step": 6350
    },
    {
      "epoch": 0.159,
      "grad_norm": 31.125,
      "grad_norm_var": 17.764583333333334,
      "learning_rate": 0.0001,
      "loss": 7.4762,
      "loss/crossentropy": 2.195678301155567,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1989523505792022,
      "step": 6360
    },
    {
      "epoch": 0.15925,
      "grad_norm": 29.625,
      "grad_norm_var": 12.851041666666667,
      "learning_rate": 0.0001,
      "loss": 7.4661,
      "loss/crossentropy": 2.0537394002079963,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20311654023826123,
      "step": 6370
    },
    {
      "epoch": 0.1595,
      "grad_norm": 30.0,
      "grad_norm_var": 10.0994140625,
      "learning_rate": 0.0001,
      "loss": 7.2759,
      "loss/crossentropy": 2.02838040292263,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19148585237562657,
      "step": 6380
    },
    {
      "epoch": 0.15975,
      "grad_norm": 39.0,
      "grad_norm_var": 2324.6707682291667,
      "learning_rate": 0.0001,
      "loss": 7.3973,
      "loss/crossentropy": 2.0951177358627318,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2158473737537861,
      "step": 6390
    },
    {
      "epoch": 0.16,
      "grad_norm": 40.0,
      "grad_norm_var": 21.121809895833334,
      "learning_rate": 0.0001,
      "loss": 7.2877,
      "loss/crossentropy": 1.878954614698887,
      "loss/hidden": 3.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18514612764120103,
      "step": 6400
    },
    {
      "epoch": 0.16025,
      "grad_norm": 33.5,
      "grad_norm_var": 23.1666015625,
      "learning_rate": 0.0001,
      "loss": 7.3598,
      "loss/crossentropy": 2.123918867111206,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18728599287569522,
      "step": 6410
    },
    {
      "epoch": 0.1605,
      "grad_norm": 29.25,
      "grad_norm_var": 11.230143229166666,
      "learning_rate": 0.0001,
      "loss": 7.476,
      "loss/crossentropy": 2.168968527019024,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20153266489505767,
      "step": 6420
    },
    {
      "epoch": 0.16075,
      "grad_norm": 30.0,
      "grad_norm_var": 90.8056640625,
      "learning_rate": 0.0001,
      "loss": 7.4289,
      "loss/crossentropy": 2.0426762118935584,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19033107869327068,
      "step": 6430
    },
    {
      "epoch": 0.161,
      "grad_norm": 38.25,
      "grad_norm_var": 15.570247395833333,
      "learning_rate": 0.0001,
      "loss": 7.3992,
      "loss/crossentropy": 2.0535445332527162,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19420330366119742,
      "step": 6440
    },
    {
      "epoch": 0.16125,
      "grad_norm": 45.5,
      "grad_norm_var": 30.326822916666668,
      "learning_rate": 0.0001,
      "loss": 7.3881,
      "loss/crossentropy": 1.949498599767685,
      "loss/hidden": 3.337890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18119702748954297,
      "step": 6450
    },
    {
      "epoch": 0.1615,
      "grad_norm": 30.625,
      "grad_norm_var": 96.64837239583333,
      "learning_rate": 0.0001,
      "loss": 7.432,
      "loss/crossentropy": 2.2534308552742006,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20010371711105107,
      "step": 6460
    },
    {
      "epoch": 0.16175,
      "grad_norm": 34.0,
      "grad_norm_var": 82.66015625,
      "learning_rate": 0.0001,
      "loss": 7.3555,
      "loss/crossentropy": 2.114310759305954,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19822147954255342,
      "step": 6470
    },
    {
      "epoch": 0.162,
      "grad_norm": 29.875,
      "grad_norm_var": 8.4541015625,
      "learning_rate": 0.0001,
      "loss": 7.1893,
      "loss/crossentropy": 2.062894639372826,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17771479729562997,
      "step": 6480
    },
    {
      "epoch": 0.16225,
      "grad_norm": 30.375,
      "grad_norm_var": 15.72265625,
      "learning_rate": 0.0001,
      "loss": 7.3665,
      "loss/crossentropy": 2.0109994761645793,
      "loss/hidden": 3.45625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19815738410688938,
      "step": 6490
    },
    {
      "epoch": 0.1625,
      "grad_norm": 33.75,
      "grad_norm_var": 162.74680989583334,
      "learning_rate": 0.0001,
      "loss": 7.4105,
      "loss/crossentropy": 2.100720961391926,
      "loss/hidden": 3.26484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18080853056162596,
      "step": 6500
    },
    {
      "epoch": 0.16275,
      "grad_norm": 36.0,
      "grad_norm_var": 10.530143229166667,
      "learning_rate": 0.0001,
      "loss": 7.4993,
      "loss/crossentropy": 2.208073277771473,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18481182418763636,
      "step": 6510
    },
    {
      "epoch": 0.163,
      "grad_norm": 37.0,
      "grad_norm_var": 8.981705729166666,
      "learning_rate": 0.0001,
      "loss": 7.5196,
      "loss/crossentropy": 2.2666310742497444,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20527655016630889,
      "step": 6520
    },
    {
      "epoch": 0.16325,
      "grad_norm": 29.625,
      "grad_norm_var": 13.959830729166667,
      "learning_rate": 0.0001,
      "loss": 7.4239,
      "loss/crossentropy": 2.2184250116348267,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19118925426155328,
      "step": 6530
    },
    {
      "epoch": 0.1635,
      "grad_norm": 29.375,
      "grad_norm_var": 8.820833333333333,
      "learning_rate": 0.0001,
      "loss": 7.3298,
      "loss/crossentropy": 2.119840921461582,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20046985391527414,
      "step": 6540
    },
    {
      "epoch": 0.16375,
      "grad_norm": 31.375,
      "grad_norm_var": 3.1910807291666665,
      "learning_rate": 0.0001,
      "loss": 7.4507,
      "loss/crossentropy": 2.109931045770645,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18326662238687277,
      "step": 6550
    },
    {
      "epoch": 0.164,
      "grad_norm": 35.25,
      "grad_norm_var": 8.6625,
      "learning_rate": 0.0001,
      "loss": 7.43,
      "loss/crossentropy": 2.090756069123745,
      "loss/hidden": 3.348046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1922204466536641,
      "step": 6560
    },
    {
      "epoch": 0.16425,
      "grad_norm": 30.625,
      "grad_norm_var": 14.567643229166666,
      "learning_rate": 0.0001,
      "loss": 7.1796,
      "loss/crossentropy": 1.9266313910484314,
      "loss/hidden": 3.465234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20681370329111814,
      "step": 6570
    },
    {
      "epoch": 0.1645,
      "grad_norm": 37.5,
      "grad_norm_var": 12.9447265625,
      "learning_rate": 0.0001,
      "loss": 7.5097,
      "loss/crossentropy": 1.875116103887558,
      "loss/hidden": 3.49921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2045787101611495,
      "step": 6580
    },
    {
      "epoch": 0.16475,
      "grad_norm": 29.875,
      "grad_norm_var": 5.706184895833333,
      "learning_rate": 0.0001,
      "loss": 7.329,
      "loss/crossentropy": 2.116366655379534,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18577212654054165,
      "step": 6590
    },
    {
      "epoch": 0.165,
      "grad_norm": 33.75,
      "grad_norm_var": 2.3650390625,
      "learning_rate": 0.0001,
      "loss": 7.3765,
      "loss/crossentropy": 2.0037689693272114,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1971780034713447,
      "step": 6600
    },
    {
      "epoch": 0.16525,
      "grad_norm": 29.625,
      "grad_norm_var": 4.601497395833333,
      "learning_rate": 0.0001,
      "loss": 7.4001,
      "loss/crossentropy": 2.1523181863129137,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.192273567058146,
      "step": 6610
    },
    {
      "epoch": 0.1655,
      "grad_norm": 29.875,
      "grad_norm_var": 7.6525390625,
      "learning_rate": 0.0001,
      "loss": 7.3804,
      "loss/crossentropy": 2.0919234342873096,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20757663380354643,
      "step": 6620
    },
    {
      "epoch": 0.16575,
      "grad_norm": 31.875,
      "grad_norm_var": 6.917708333333334,
      "learning_rate": 0.0001,
      "loss": 7.3889,
      "loss/crossentropy": 2.035097151994705,
      "loss/hidden": 3.31484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19357213731855155,
      "step": 6630
    },
    {
      "epoch": 0.166,
      "grad_norm": 28.125,
      "grad_norm_var": 2.4607245906905574e+18,
      "learning_rate": 0.0001,
      "loss": 7.5149,
      "loss/crossentropy": 2.114539227634668,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19631449952721597,
      "step": 6640
    },
    {
      "epoch": 0.16625,
      "grad_norm": 28.375,
      "grad_norm_var": 2.4607245908931773e+18,
      "learning_rate": 0.0001,
      "loss": 7.3157,
      "loss/crossentropy": 2.0170676171779633,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1893145913258195,
      "step": 6650
    },
    {
      "epoch": 0.1665,
      "grad_norm": 28.0,
      "grad_norm_var": 32.25807291666667,
      "learning_rate": 0.0001,
      "loss": 7.2947,
      "loss/crossentropy": 1.9412188947200775,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18349691890180111,
      "step": 6660
    },
    {
      "epoch": 0.16675,
      "grad_norm": 30.25,
      "grad_norm_var": 48.2375,
      "learning_rate": 0.0001,
      "loss": 7.4086,
      "loss/crossentropy": 2.1157006829977036,
      "loss/hidden": 3.410546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1936411712318659,
      "step": 6670
    },
    {
      "epoch": 0.167,
      "grad_norm": 29.75,
      "grad_norm_var": 36.18333333333333,
      "learning_rate": 0.0001,
      "loss": 7.4096,
      "loss/crossentropy": 1.9384170174598694,
      "loss/hidden": 3.46015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1831628430634737,
      "step": 6680
    },
    {
      "epoch": 0.16725,
      "grad_norm": 30.0,
      "grad_norm_var": 62.6150390625,
      "learning_rate": 0.0001,
      "loss": 7.4604,
      "loss/crossentropy": 2.152873657643795,
      "loss/hidden": 3.421484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20049556214362382,
      "step": 6690
    },
    {
      "epoch": 0.1675,
      "grad_norm": 29.75,
      "grad_norm_var": 28.671875,
      "learning_rate": 0.0001,
      "loss": 7.5739,
      "loss/crossentropy": 2.1935679107904433,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20188184324651956,
      "step": 6700
    },
    {
      "epoch": 0.16775,
      "grad_norm": 28.25,
      "grad_norm_var": 2.278580729166667,
      "learning_rate": 0.0001,
      "loss": 7.4709,
      "loss/crossentropy": 2.062030902504921,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18874304387718438,
      "step": 6710
    },
    {
      "epoch": 0.168,
      "grad_norm": 29.125,
      "grad_norm_var": 3.3150390625,
      "learning_rate": 0.0001,
      "loss": 7.3655,
      "loss/crossentropy": 1.999978879839182,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18526637642644345,
      "step": 6720
    },
    {
      "epoch": 0.16825,
      "grad_norm": 35.25,
      "grad_norm_var": 6.237239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4037,
      "loss/crossentropy": 2.0561595499515533,
      "loss/hidden": 3.41015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20046296287328005,
      "step": 6730
    },
    {
      "epoch": 0.1685,
      "grad_norm": 28.75,
      "grad_norm_var": 5.5619140625,
      "learning_rate": 0.0001,
      "loss": 7.368,
      "loss/crossentropy": 2.0664093092083933,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1885912848636508,
      "step": 6740
    },
    {
      "epoch": 0.16875,
      "grad_norm": 30.875,
      "grad_norm_var": 5.843489583333334,
      "learning_rate": 0.0001,
      "loss": 7.4899,
      "loss/crossentropy": 2.1205774366855623,
      "loss/hidden": 3.491796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21412673257291318,
      "step": 6750
    },
    {
      "epoch": 0.169,
      "grad_norm": 32.0,
      "grad_norm_var": 6.3775390625,
      "learning_rate": 0.0001,
      "loss": 7.3708,
      "loss/crossentropy": 2.0314668610692026,
      "loss/hidden": 3.33203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18550706487149,
      "step": 6760
    },
    {
      "epoch": 0.16925,
      "grad_norm": 32.5,
      "grad_norm_var": 7.378059895833333,
      "learning_rate": 0.0001,
      "loss": 7.4055,
      "loss/crossentropy": 2.1428691864013674,
      "loss/hidden": 3.506640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21024896781891583,
      "step": 6770
    },
    {
      "epoch": 0.1695,
      "grad_norm": 31.375,
      "grad_norm_var": 3.9791666666666665,
      "learning_rate": 0.0001,
      "loss": 7.4444,
      "loss/crossentropy": 1.9500044576823712,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19028451843187213,
      "step": 6780
    },
    {
      "epoch": 0.16975,
      "grad_norm": 29.75,
      "grad_norm_var": 3.2853515625,
      "learning_rate": 0.0001,
      "loss": 7.4369,
      "loss/crossentropy": 2.1563921123743057,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18290520180016756,
      "step": 6790
    },
    {
      "epoch": 0.17,
      "grad_norm": 38.0,
      "grad_norm_var": 8.0947265625,
      "learning_rate": 0.0001,
      "loss": 7.3756,
      "loss/crossentropy": 2.133736363053322,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1851572971791029,
      "step": 6800
    },
    {
      "epoch": 0.17025,
      "grad_norm": 34.75,
      "grad_norm_var": 14.757747395833333,
      "learning_rate": 0.0001,
      "loss": 7.2181,
      "loss/crossentropy": 2.054655596613884,
      "loss/hidden": 3.349609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18349520340561867,
      "step": 6810
    },
    {
      "epoch": 0.1705,
      "grad_norm": 30.375,
      "grad_norm_var": 5.17890625,
      "learning_rate": 0.0001,
      "loss": 7.303,
      "loss/crossentropy": 2.024763736128807,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19044207576662303,
      "step": 6820
    },
    {
      "epoch": 0.17075,
      "grad_norm": 32.25,
      "grad_norm_var": 3.7955729166666665,
      "learning_rate": 0.0001,
      "loss": 7.4884,
      "loss/crossentropy": 1.9924081854522229,
      "loss/hidden": 3.49609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20656490996479987,
      "step": 6830
    },
    {
      "epoch": 0.171,
      "grad_norm": 30.875,
      "grad_norm_var": 3.215559895833333,
      "learning_rate": 0.0001,
      "loss": 7.4845,
      "loss/crossentropy": 2.1104799427092074,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18316805781796575,
      "step": 6840
    },
    {
      "epoch": 0.17125,
      "grad_norm": 31.625,
      "grad_norm_var": 25.610872395833333,
      "learning_rate": 0.0001,
      "loss": 7.3771,
      "loss/crossentropy": 2.02793128117919,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17605492258444427,
      "step": 6850
    },
    {
      "epoch": 0.1715,
      "grad_norm": 29.375,
      "grad_norm_var": 34.992122395833334,
      "learning_rate": 0.0001,
      "loss": 7.5157,
      "loss/crossentropy": 2.1305345237255096,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19113806802779437,
      "step": 6860
    },
    {
      "epoch": 0.17175,
      "grad_norm": 28.375,
      "grad_norm_var": 12.389322916666666,
      "learning_rate": 0.0001,
      "loss": 7.2752,
      "loss/crossentropy": 2.0788576349616052,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19617441901937127,
      "step": 6870
    },
    {
      "epoch": 0.172,
      "grad_norm": 27.625,
      "grad_norm_var": 17.758072916666666,
      "learning_rate": 0.0001,
      "loss": 7.4366,
      "loss/crossentropy": 2.068412736058235,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18466003462672234,
      "step": 6880
    },
    {
      "epoch": 0.17225,
      "grad_norm": 30.375,
      "grad_norm_var": 19.836393229166667,
      "learning_rate": 0.0001,
      "loss": 7.5047,
      "loss/crossentropy": 2.07881121635437,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19641269743442535,
      "step": 6890
    },
    {
      "epoch": 0.1725,
      "grad_norm": 29.625,
      "grad_norm_var": 14.70625,
      "learning_rate": 0.0001,
      "loss": 7.4086,
      "loss/crossentropy": 1.84702168405056,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1932983512058854,
      "step": 6900
    },
    {
      "epoch": 0.17275,
      "grad_norm": 31.875,
      "grad_norm_var": 31.006705729166665,
      "learning_rate": 0.0001,
      "loss": 7.3715,
      "loss/crossentropy": 2.099086304008961,
      "loss/hidden": 3.406640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20250021573156118,
      "step": 6910
    },
    {
      "epoch": 0.173,
      "grad_norm": 29.0,
      "grad_norm_var": 26.07265625,
      "learning_rate": 0.0001,
      "loss": 7.3343,
      "loss/crossentropy": 2.0957569405436516,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2010388659313321,
      "step": 6920
    },
    {
      "epoch": 0.17325,
      "grad_norm": 33.0,
      "grad_norm_var": 3.753059895833333,
      "learning_rate": 0.0001,
      "loss": 7.4937,
      "loss/crossentropy": 2.071177572757006,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19133195597678423,
      "step": 6930
    },
    {
      "epoch": 0.1735,
      "grad_norm": 31.125,
      "grad_norm_var": 5.88515625,
      "learning_rate": 0.0001,
      "loss": 7.4757,
      "loss/crossentropy": 2.0803056344389916,
      "loss/hidden": 3.51875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20995833892375232,
      "step": 6940
    },
    {
      "epoch": 0.17375,
      "grad_norm": 28.5,
      "grad_norm_var": 7.328059895833333,
      "learning_rate": 0.0001,
      "loss": 7.2293,
      "loss/crossentropy": 1.9285863403230905,
      "loss/hidden": 3.283203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1665677004493773,
      "step": 6950
    },
    {
      "epoch": 0.174,
      "grad_norm": 27.375,
      "grad_norm_var": 11.585872395833333,
      "learning_rate": 0.0001,
      "loss": 7.313,
      "loss/crossentropy": 2.0258478805422784,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1821833540685475,
      "step": 6960
    },
    {
      "epoch": 0.17425,
      "grad_norm": 30.5,
      "grad_norm_var": 10.760416666666666,
      "learning_rate": 0.0001,
      "loss": 7.5413,
      "loss/crossentropy": 2.1308654129505156,
      "loss/hidden": 3.43203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21524183861911297,
      "step": 6970
    },
    {
      "epoch": 0.1745,
      "grad_norm": 29.625,
      "grad_norm_var": 4.138541666666667,
      "learning_rate": 0.0001,
      "loss": 7.4692,
      "loss/crossentropy": 2.1182237058877944,
      "loss/hidden": 3.362890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18946228343993426,
      "step": 6980
    },
    {
      "epoch": 0.17475,
      "grad_norm": 31.0,
      "grad_norm_var": 5.499934895833333,
      "learning_rate": 0.0001,
      "loss": 7.4658,
      "loss/crossentropy": 2.0863646306097507,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2054815970361233,
      "step": 6990
    },
    {
      "epoch": 0.175,
      "grad_norm": 31.375,
      "grad_norm_var": 1.7067057291666667,
      "learning_rate": 0.0001,
      "loss": 7.3196,
      "loss/crossentropy": 2.1002516582608224,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19273097421973945,
      "step": 7000
    },
    {
      "epoch": 0.17525,
      "grad_norm": 31.25,
      "grad_norm_var": 1.7999348958333334,
      "learning_rate": 0.0001,
      "loss": 7.3643,
      "loss/crossentropy": 2.015849883854389,
      "loss/hidden": 3.383984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18935495987534523,
      "step": 7010
    },
    {
      "epoch": 0.1755,
      "grad_norm": 28.875,
      "grad_norm_var": 3.3645833333333335,
      "learning_rate": 0.0001,
      "loss": 7.3973,
      "loss/crossentropy": 2.072261115908623,
      "loss/hidden": 3.482421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19815812185406684,
      "step": 7020
    },
    {
      "epoch": 0.17575,
      "grad_norm": 29.375,
      "grad_norm_var": 8.442708333333334,
      "learning_rate": 0.0001,
      "loss": 7.4193,
      "loss/crossentropy": 2.1367180705070496,
      "loss/hidden": 3.31328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1970324844121933,
      "step": 7030
    },
    {
      "epoch": 0.176,
      "grad_norm": 30.375,
      "grad_norm_var": 5.426822916666667,
      "learning_rate": 0.0001,
      "loss": 7.518,
      "loss/crossentropy": 2.210773140192032,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1955376474186778,
      "step": 7040
    },
    {
      "epoch": 0.17625,
      "grad_norm": 30.125,
      "grad_norm_var": 3.1791015625,
      "learning_rate": 0.0001,
      "loss": 7.3883,
      "loss/crossentropy": 2.1343745410442354,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19158909022808074,
      "step": 7050
    },
    {
      "epoch": 0.1765,
      "grad_norm": 29.25,
      "grad_norm_var": 4.112434895833333,
      "learning_rate": 0.0001,
      "loss": 7.4451,
      "loss/crossentropy": 1.9646480686962604,
      "loss/hidden": 3.465234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19925388041883707,
      "step": 7060
    },
    {
      "epoch": 0.17675,
      "grad_norm": 31.5,
      "grad_norm_var": 3.3275390625,
      "learning_rate": 0.0001,
      "loss": 7.4874,
      "loss/crossentropy": 2.1882633604109287,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19836742132902146,
      "step": 7070
    },
    {
      "epoch": 0.177,
      "grad_norm": 34.75,
      "grad_norm_var": 5.7291015625,
      "learning_rate": 0.0001,
      "loss": 7.4907,
      "loss/crossentropy": 2.2362487465143204,
      "loss/hidden": 3.33203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19153916742652655,
      "step": 7080
    },
    {
      "epoch": 0.17725,
      "grad_norm": 29.5,
      "grad_norm_var": 5.002083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4459,
      "loss/crossentropy": 2.164398466050625,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20325577780604362,
      "step": 7090
    },
    {
      "epoch": 0.1775,
      "grad_norm": 32.25,
      "grad_norm_var": 21.282747395833333,
      "learning_rate": 0.0001,
      "loss": 7.4933,
      "loss/crossentropy": 2.1975876331329345,
      "loss/hidden": 3.455078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19638306740671396,
      "step": 7100
    },
    {
      "epoch": 0.17775,
      "grad_norm": 28.5,
      "grad_norm_var": 32.18020833333333,
      "learning_rate": 0.0001,
      "loss": 7.358,
      "loss/crossentropy": 1.9753169894218445,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17774544414132834,
      "step": 7110
    },
    {
      "epoch": 0.178,
      "grad_norm": 30.0,
      "grad_norm_var": 11.808333333333334,
      "learning_rate": 0.0001,
      "loss": 7.4718,
      "loss/crossentropy": 2.0591190218925477,
      "loss/hidden": 3.4578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19963474106043577,
      "step": 7120
    },
    {
      "epoch": 0.17825,
      "grad_norm": 29.25,
      "grad_norm_var": 7.753580729166667,
      "learning_rate": 0.0001,
      "loss": 7.3979,
      "loss/crossentropy": 2.062809920310974,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19096513669937848,
      "step": 7130
    },
    {
      "epoch": 0.1785,
      "grad_norm": 30.125,
      "grad_norm_var": 6.6025390625,
      "learning_rate": 0.0001,
      "loss": 7.3036,
      "loss/crossentropy": 2.0240534149110316,
      "loss/hidden": 3.4453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18496394343674183,
      "step": 7140
    },
    {
      "epoch": 0.17875,
      "grad_norm": 31.375,
      "grad_norm_var": 2.4942057291666666,
      "learning_rate": 0.0001,
      "loss": 7.4778,
      "loss/crossentropy": 2.124583348631859,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2003519142046571,
      "step": 7150
    },
    {
      "epoch": 0.179,
      "grad_norm": 29.0,
      "grad_norm_var": 13.8978515625,
      "learning_rate": 0.0001,
      "loss": 7.3866,
      "loss/crossentropy": 2.035899819433689,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18300745636224747,
      "step": 7160
    },
    {
      "epoch": 0.17925,
      "grad_norm": 28.125,
      "grad_norm_var": 18.375455729166667,
      "learning_rate": 0.0001,
      "loss": 7.3499,
      "loss/crossentropy": 2.086082286387682,
      "loss/hidden": 3.26875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17294995756819845,
      "step": 7170
    },
    {
      "epoch": 0.1795,
      "grad_norm": 28.5,
      "grad_norm_var": 22.834830729166665,
      "learning_rate": 0.0001,
      "loss": 7.4265,
      "loss/crossentropy": 2.0105697728693483,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19119318593293427,
      "step": 7180
    },
    {
      "epoch": 0.17975,
      "grad_norm": 33.0,
      "grad_norm_var": 19.51640625,
      "learning_rate": 0.0001,
      "loss": 7.4722,
      "loss/crossentropy": 2.1506593719124796,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19607089888304471,
      "step": 7190
    },
    {
      "epoch": 0.18,
      "grad_norm": 31.5,
      "grad_norm_var": 9.854622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3205,
      "loss/crossentropy": 2.0767677523195744,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19086614530533552,
      "step": 7200
    },
    {
      "epoch": 0.18025,
      "grad_norm": 27.625,
      "grad_norm_var": 8.885416666666666,
      "learning_rate": 0.0001,
      "loss": 7.2908,
      "loss/crossentropy": 2.230179136991501,
      "loss/hidden": 3.26953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18500201255083085,
      "step": 7210
    },
    {
      "epoch": 0.1805,
      "grad_norm": 38.25,
      "grad_norm_var": 8.60625,
      "learning_rate": 0.0001,
      "loss": 7.4664,
      "loss/crossentropy": 2.192136238515377,
      "loss/hidden": 3.380859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1965922711417079,
      "step": 7220
    },
    {
      "epoch": 0.18075,
      "grad_norm": 33.0,
      "grad_norm_var": 7.258333333333334,
      "learning_rate": 0.0001,
      "loss": 7.3796,
      "loss/crossentropy": 2.010923378914595,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19027266185730696,
      "step": 7230
    },
    {
      "epoch": 0.181,
      "grad_norm": 39.75,
      "grad_norm_var": 11.3353515625,
      "learning_rate": 0.0001,
      "loss": 7.5537,
      "loss/crossentropy": 2.055556283891201,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19257053220644593,
      "step": 7240
    },
    {
      "epoch": 0.18125,
      "grad_norm": 27.5,
      "grad_norm_var": 14.424739583333333,
      "learning_rate": 0.0001,
      "loss": 7.4216,
      "loss/crossentropy": 2.136477355659008,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20195687096565962,
      "step": 7250
    },
    {
      "epoch": 0.1815,
      "grad_norm": 30.125,
      "grad_norm_var": 9.70390625,
      "learning_rate": 0.0001,
      "loss": 7.4656,
      "loss/crossentropy": 2.0134367659687995,
      "loss/hidden": 3.523046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22354123163968326,
      "step": 7260
    },
    {
      "epoch": 0.18175,
      "grad_norm": 30.0,
      "grad_norm_var": 6.324739583333334,
      "learning_rate": 0.0001,
      "loss": 7.3959,
      "loss/crossentropy": 2.1245115220546724,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19102834183722733,
      "step": 7270
    },
    {
      "epoch": 0.182,
      "grad_norm": 29.125,
      "grad_norm_var": 18.903580729166666,
      "learning_rate": 0.0001,
      "loss": 7.3761,
      "loss/crossentropy": 2.035867254436016,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19426564145833253,
      "step": 7280
    },
    {
      "epoch": 0.18225,
      "grad_norm": 30.5,
      "grad_norm_var": 20.406705729166667,
      "learning_rate": 0.0001,
      "loss": 7.3884,
      "loss/crossentropy": 1.9805133253335954,
      "loss/hidden": 3.370703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19129701480269432,
      "step": 7290
    },
    {
      "epoch": 0.1825,
      "grad_norm": 31.375,
      "grad_norm_var": 5.842708333333333,
      "learning_rate": 0.0001,
      "loss": 7.3833,
      "loss/crossentropy": 2.0621849209070207,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19919742476195096,
      "step": 7300
    },
    {
      "epoch": 0.18275,
      "grad_norm": 47.0,
      "grad_norm_var": 2.675771470406222e+18,
      "learning_rate": 0.0001,
      "loss": 7.2357,
      "loss/crossentropy": 2.1282688602805138,
      "loss/hidden": 3.298046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18128441767767073,
      "step": 7310
    },
    {
      "epoch": 0.183,
      "grad_norm": 29.25,
      "grad_norm_var": 28.158333333333335,
      "learning_rate": 0.0001,
      "loss": 7.4531,
      "loss/crossentropy": 2.1078659296035767,
      "loss/hidden": 3.580078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2171280149370432,
      "step": 7320
    },
    {
      "epoch": 0.18325,
      "grad_norm": 35.25,
      "grad_norm_var": 25.365625,
      "learning_rate": 0.0001,
      "loss": 7.3538,
      "loss/crossentropy": 2.200744313001633,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19498275145888327,
      "step": 7330
    },
    {
      "epoch": 0.1835,
      "grad_norm": 28.375,
      "grad_norm_var": 23.859830729166667,
      "learning_rate": 0.0001,
      "loss": 7.3437,
      "loss/crossentropy": 1.9968993581831456,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1875888810493052,
      "step": 7340
    },
    {
      "epoch": 0.18375,
      "grad_norm": 28.0,
      "grad_norm_var": 8.989518229166666,
      "learning_rate": 0.0001,
      "loss": 7.3984,
      "loss/crossentropy": 2.0748091831803324,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1947355069220066,
      "step": 7350
    },
    {
      "epoch": 0.184,
      "grad_norm": 29.875,
      "grad_norm_var": 7.728125,
      "learning_rate": 0.0001,
      "loss": 7.3008,
      "loss/crossentropy": 2.067877373099327,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18336500320583582,
      "step": 7360
    },
    {
      "epoch": 0.18425,
      "grad_norm": 29.25,
      "grad_norm_var": 13.189583333333333,
      "learning_rate": 0.0001,
      "loss": 7.4547,
      "loss/crossentropy": 2.152864509820938,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2102669222280383,
      "step": 7370
    },
    {
      "epoch": 0.1845,
      "grad_norm": 28.5,
      "grad_norm_var": 8.7556640625,
      "learning_rate": 0.0001,
      "loss": 7.286,
      "loss/crossentropy": 1.9991149730980395,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18364950213581324,
      "step": 7380
    },
    {
      "epoch": 0.18475,
      "grad_norm": 31.75,
      "grad_norm_var": 7.770572916666667,
      "learning_rate": 0.0001,
      "loss": 7.3546,
      "loss/crossentropy": 2.0513292245566843,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1970413200557232,
      "step": 7390
    },
    {
      "epoch": 0.185,
      "grad_norm": 30.125,
      "grad_norm_var": 7.620572916666666,
      "learning_rate": 0.0001,
      "loss": 7.5116,
      "loss/crossentropy": 2.102216296643019,
      "loss/hidden": 3.51875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21091360161080958,
      "step": 7400
    },
    {
      "epoch": 0.18525,
      "grad_norm": 32.75,
      "grad_norm_var": 9.191666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3421,
      "loss/crossentropy": 1.9926266744732857,
      "loss/hidden": 3.455078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19832582902163268,
      "step": 7410
    },
    {
      "epoch": 0.1855,
      "grad_norm": 29.5,
      "grad_norm_var": 5.945833333333334,
      "learning_rate": 0.0001,
      "loss": 7.4563,
      "loss/crossentropy": 2.141331580281258,
      "loss/hidden": 3.408203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20020943265408278,
      "step": 7420
    },
    {
      "epoch": 0.18575,
      "grad_norm": 31.0,
      "grad_norm_var": 7.198372395833333,
      "learning_rate": 0.0001,
      "loss": 7.4227,
      "loss/crossentropy": 1.9694693490862847,
      "loss/hidden": 3.421484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18249646089971067,
      "step": 7430
    },
    {
      "epoch": 0.186,
      "grad_norm": 29.375,
      "grad_norm_var": 5.9697265625,
      "learning_rate": 0.0001,
      "loss": 7.4119,
      "loss/crossentropy": 2.1407265037298204,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17720893137156962,
      "step": 7440
    },
    {
      "epoch": 0.18625,
      "grad_norm": 29.875,
      "grad_norm_var": 0.8072916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3441,
      "loss/crossentropy": 2.124198019504547,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1946978410705924,
      "step": 7450
    },
    {
      "epoch": 0.1865,
      "grad_norm": 32.75,
      "grad_norm_var": 1.9634765625,
      "learning_rate": 0.0001,
      "loss": 7.4652,
      "loss/crossentropy": 2.131994958221912,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19461573138833047,
      "step": 7460
    },
    {
      "epoch": 0.18675,
      "grad_norm": 30.875,
      "grad_norm_var": 2.7280598958333333,
      "learning_rate": 0.0001,
      "loss": 7.3278,
      "loss/crossentropy": 2.117748848348856,
      "loss/hidden": 3.2578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18584235943853855,
      "step": 7470
    },
    {
      "epoch": 0.187,
      "grad_norm": 31.625,
      "grad_norm_var": 2.466080729166667,
      "learning_rate": 0.0001,
      "loss": 7.5817,
      "loss/crossentropy": 2.1364282086491584,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21069204956293106,
      "step": 7480
    },
    {
      "epoch": 0.18725,
      "grad_norm": 31.25,
      "grad_norm_var": 3.012434895833333,
      "learning_rate": 0.0001,
      "loss": 7.5197,
      "loss/crossentropy": 2.0523312032222747,
      "loss/hidden": 3.50078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21221144162118435,
      "step": 7490
    },
    {
      "epoch": 0.1875,
      "grad_norm": 33.75,
      "grad_norm_var": 4.995768229166667,
      "learning_rate": 0.0001,
      "loss": 7.4317,
      "loss/crossentropy": 2.0852270901203154,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17908250950276852,
      "step": 7500
    },
    {
      "epoch": 0.18775,
      "grad_norm": 30.875,
      "grad_norm_var": 2.4518229166666665,
      "learning_rate": 0.0001,
      "loss": 7.5444,
      "loss/crossentropy": 2.059383874386549,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19440155941992998,
      "step": 7510
    },
    {
      "epoch": 0.188,
      "grad_norm": 27.625,
      "grad_norm_var": 2.851822916666667,
      "learning_rate": 0.0001,
      "loss": 7.5199,
      "loss/crossentropy": 2.1382876858115196,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19888029601424934,
      "step": 7520
    },
    {
      "epoch": 0.18825,
      "grad_norm": 29.375,
      "grad_norm_var": 3.073372395833333,
      "learning_rate": 0.0001,
      "loss": 7.3012,
      "loss/crossentropy": 2.0625696159899234,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18495072829537093,
      "step": 7530
    },
    {
      "epoch": 0.1885,
      "grad_norm": 32.0,
      "grad_norm_var": 2.2122395833333335,
      "learning_rate": 0.0001,
      "loss": 7.3643,
      "loss/crossentropy": 2.124967637658119,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18375679664313793,
      "step": 7540
    },
    {
      "epoch": 0.18875,
      "grad_norm": 29.125,
      "grad_norm_var": 3.2681640625,
      "learning_rate": 0.0001,
      "loss": 7.351,
      "loss/crossentropy": 2.0680116668343542,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18827605471014977,
      "step": 7550
    },
    {
      "epoch": 0.189,
      "grad_norm": 31.75,
      "grad_norm_var": 1.5337890625,
      "learning_rate": 0.0001,
      "loss": 7.4207,
      "loss/crossentropy": 2.079096484184265,
      "loss/hidden": 3.443359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20160295628011227,
      "step": 7560
    },
    {
      "epoch": 0.18925,
      "grad_norm": 30.625,
      "grad_norm_var": 18.2197265625,
      "learning_rate": 0.0001,
      "loss": 7.4789,
      "loss/crossentropy": 2.058067685365677,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21216327100992202,
      "step": 7570
    },
    {
      "epoch": 0.1895,
      "grad_norm": 34.25,
      "grad_norm_var": 14.415625,
      "learning_rate": 0.0001,
      "loss": 7.4987,
      "loss/crossentropy": 2.0142914205789566,
      "loss/hidden": 3.580078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20794902741909027,
      "step": 7580
    },
    {
      "epoch": 0.18975,
      "grad_norm": 30.5,
      "grad_norm_var": 1.9809895833333333,
      "learning_rate": 0.0001,
      "loss": 7.4585,
      "loss/crossentropy": 2.299562671780586,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19880922697484493,
      "step": 7590
    },
    {
      "epoch": 0.19,
      "grad_norm": 30.875,
      "grad_norm_var": 15.101822916666666,
      "learning_rate": 0.0001,
      "loss": 7.3204,
      "loss/crossentropy": 2.1472302600741386,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18796155080199242,
      "step": 7600
    },
    {
      "epoch": 0.19025,
      "grad_norm": 29.0,
      "grad_norm_var": 2.5940733610451533e+18,
      "learning_rate": 0.0001,
      "loss": 7.5335,
      "loss/crossentropy": 2.1664531916379928,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18874028734862805,
      "step": 7610
    },
    {
      "epoch": 0.1905,
      "grad_norm": 29.125,
      "grad_norm_var": 0.8843098958333333,
      "learning_rate": 0.0001,
      "loss": 7.4846,
      "loss/crossentropy": 2.0765088513493537,
      "loss/hidden": 3.34375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18674521408975125,
      "step": 7620
    },
    {
      "epoch": 0.19075,
      "grad_norm": 27.25,
      "grad_norm_var": 4.381184895833333,
      "learning_rate": 0.0001,
      "loss": 7.3401,
      "loss/crossentropy": 1.7713539503514766,
      "loss/hidden": 3.515234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1749590938910842,
      "step": 7630
    },
    {
      "epoch": 0.191,
      "grad_norm": 33.5,
      "grad_norm_var": 5.26640625,
      "learning_rate": 0.0001,
      "loss": 7.4838,
      "loss/crossentropy": 1.922049730271101,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18183694053441285,
      "step": 7640
    },
    {
      "epoch": 0.19125,
      "grad_norm": 30.5,
      "grad_norm_var": 4.030989583333334,
      "learning_rate": 0.0001,
      "loss": 7.3887,
      "loss/crossentropy": 2.094516658782959,
      "loss/hidden": 3.452734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18757299687713386,
      "step": 7650
    },
    {
      "epoch": 0.1915,
      "grad_norm": 32.0,
      "grad_norm_var": 1.5035807291666667,
      "learning_rate": 0.0001,
      "loss": 7.4445,
      "loss/crossentropy": 1.999229770898819,
      "loss/hidden": 3.516015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19523975029587745,
      "step": 7660
    },
    {
      "epoch": 0.19175,
      "grad_norm": 32.25,
      "grad_norm_var": 3.0275390625,
      "learning_rate": 0.0001,
      "loss": 7.4626,
      "loss/crossentropy": 2.0366951674222946,
      "loss/hidden": 3.42265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18243371956050397,
      "step": 7670
    },
    {
      "epoch": 0.192,
      "grad_norm": 32.75,
      "grad_norm_var": 2.5931640625,
      "learning_rate": 0.0001,
      "loss": 7.5274,
      "loss/crossentropy": 2.0489632681012155,
      "loss/hidden": 3.564453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21230401135981083,
      "step": 7680
    },
    {
      "epoch": 0.19225,
      "grad_norm": 32.75,
      "grad_norm_var": 3.738997395833333,
      "learning_rate": 0.0001,
      "loss": 7.4128,
      "loss/crossentropy": 2.0826203912496566,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18644160348922015,
      "step": 7690
    },
    {
      "epoch": 0.1925,
      "grad_norm": 29.875,
      "grad_norm_var": 4.351041666666666,
      "learning_rate": 0.0001,
      "loss": 7.5195,
      "loss/crossentropy": 2.1674430795013904,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19499621093273162,
      "step": 7700
    },
    {
      "epoch": 0.19275,
      "grad_norm": 28.625,
      "grad_norm_var": 4.939322916666667,
      "learning_rate": 0.0001,
      "loss": 7.4236,
      "loss/crossentropy": 2.092283549904823,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19506504610180855,
      "step": 7710
    },
    {
      "epoch": 0.193,
      "grad_norm": 33.0,
      "grad_norm_var": 9.080208333333333,
      "learning_rate": 0.0001,
      "loss": 7.3501,
      "loss/crossentropy": 2.0526101261377336,
      "loss/hidden": 3.51328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19770189765840768,
      "step": 7720
    },
    {
      "epoch": 0.19325,
      "grad_norm": 29.0,
      "grad_norm_var": 9.731184895833334,
      "learning_rate": 0.0001,
      "loss": 7.4607,
      "loss/crossentropy": 2.05912861302495,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19287334326654673,
      "step": 7730
    },
    {
      "epoch": 0.1935,
      "grad_norm": 30.5,
      "grad_norm_var": 1.8052083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4243,
      "loss/crossentropy": 1.8983285859227181,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18626301139593124,
      "step": 7740
    },
    {
      "epoch": 0.19375,
      "grad_norm": 40.75,
      "grad_norm_var": 9.1541015625,
      "learning_rate": 0.0001,
      "loss": 7.4332,
      "loss/crossentropy": 2.099681233614683,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20556394904851913,
      "step": 7750
    },
    {
      "epoch": 0.194,
      "grad_norm": 37.0,
      "grad_norm_var": 12.584375,
      "learning_rate": 0.0001,
      "loss": 7.516,
      "loss/crossentropy": 2.0436215907335282,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17505232142284513,
      "step": 7760
    },
    {
      "epoch": 0.19425,
      "grad_norm": 30.125,
      "grad_norm_var": 5.2681640625,
      "learning_rate": 0.0001,
      "loss": 7.5486,
      "loss/crossentropy": 2.0449838273227217,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19324074545875192,
      "step": 7770
    },
    {
      "epoch": 0.1945,
      "grad_norm": 30.125,
      "grad_norm_var": 1.4518229166666667,
      "learning_rate": 0.0001,
      "loss": 7.4052,
      "loss/crossentropy": 2.1020638972520826,
      "loss/hidden": 3.3234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1987349819391966,
      "step": 7780
    },
    {
      "epoch": 0.19475,
      "grad_norm": 32.25,
      "grad_norm_var": 1.525,
      "learning_rate": 0.0001,
      "loss": 7.3148,
      "loss/crossentropy": 2.0913542471826077,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1900358498096466,
      "step": 7790
    },
    {
      "epoch": 0.195,
      "grad_norm": 29.875,
      "grad_norm_var": 1.2280598958333333,
      "learning_rate": 0.0001,
      "loss": 7.43,
      "loss/crossentropy": 1.9448820307850838,
      "loss/hidden": 3.423828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18228193083778024,
      "step": 7800
    },
    {
      "epoch": 0.19525,
      "grad_norm": 28.75,
      "grad_norm_var": 2.3791015625,
      "learning_rate": 0.0001,
      "loss": 7.4772,
      "loss/crossentropy": 2.0547610491514208,
      "loss/hidden": 3.459765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19370344914495946,
      "step": 7810
    },
    {
      "epoch": 0.1955,
      "grad_norm": 31.375,
      "grad_norm_var": 6.208072916666667,
      "learning_rate": 0.0001,
      "loss": 7.4448,
      "loss/crossentropy": 2.0798824220895766,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18972196318209172,
      "step": 7820
    },
    {
      "epoch": 0.19575,
      "grad_norm": 29.5,
      "grad_norm_var": 6.248893229166667,
      "learning_rate": 0.0001,
      "loss": 7.5358,
      "loss/crossentropy": 2.2324195951223373,
      "loss/hidden": 3.4375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2047037549316883,
      "step": 7830
    },
    {
      "epoch": 0.196,
      "grad_norm": 29.375,
      "grad_norm_var": 4.453059895833333,
      "learning_rate": 0.0001,
      "loss": 7.3114,
      "loss/crossentropy": 2.1020479179918765,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19328910131007432,
      "step": 7840
    },
    {
      "epoch": 0.19625,
      "grad_norm": 33.5,
      "grad_norm_var": 590.0817057291666,
      "learning_rate": 0.0001,
      "loss": 7.4281,
      "loss/crossentropy": 2.0953447744250298,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1854689259082079,
      "step": 7850
    },
    {
      "epoch": 0.1965,
      "grad_norm": 33.0,
      "grad_norm_var": 625.6192057291667,
      "learning_rate": 0.0001,
      "loss": 7.5283,
      "loss/crossentropy": 2.061315707862377,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18506519980728625,
      "step": 7860
    },
    {
      "epoch": 0.19675,
      "grad_norm": 32.0,
      "grad_norm_var": 69.38430989583334,
      "learning_rate": 0.0001,
      "loss": 7.5314,
      "loss/crossentropy": 2.1605025470256805,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20316088199615479,
      "step": 7870
    },
    {
      "epoch": 0.197,
      "grad_norm": 31.375,
      "grad_norm_var": 1.1067057291666667,
      "learning_rate": 0.0001,
      "loss": 7.4244,
      "loss/crossentropy": 2.1977868393063544,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1992826245725155,
      "step": 7880
    },
    {
      "epoch": 0.19725,
      "grad_norm": 29.25,
      "grad_norm_var": 9.578059895833333,
      "learning_rate": 0.0001,
      "loss": 7.502,
      "loss/crossentropy": 2.0480964958667753,
      "loss/hidden": 3.636328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21483200527727603,
      "step": 7890
    },
    {
      "epoch": 0.1975,
      "grad_norm": 30.875,
      "grad_norm_var": 3.6372395833333333,
      "learning_rate": 0.0001,
      "loss": 7.548,
      "loss/crossentropy": 2.157261362671852,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2084518164396286,
      "step": 7900
    },
    {
      "epoch": 0.19775,
      "grad_norm": 28.5,
      "grad_norm_var": 2.095572916666667,
      "learning_rate": 0.0001,
      "loss": 7.3864,
      "loss/crossentropy": 2.1441849052906035,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20142039898782968,
      "step": 7910
    },
    {
      "epoch": 0.198,
      "grad_norm": 37.5,
      "grad_norm_var": 16.170247395833332,
      "learning_rate": 0.0001,
      "loss": 7.429,
      "loss/crossentropy": 2.001983726769686,
      "loss/hidden": 3.448828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19066998092457652,
      "step": 7920
    },
    {
      "epoch": 0.19825,
      "grad_norm": 29.375,
      "grad_norm_var": 15.561393229166667,
      "learning_rate": 0.0001,
      "loss": 7.4093,
      "loss/crossentropy": 2.225774070620537,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1993710033595562,
      "step": 7930
    },
    {
      "epoch": 0.1985,
      "grad_norm": 34.0,
      "grad_norm_var": 25.343489583333334,
      "learning_rate": 0.0001,
      "loss": 7.4932,
      "loss/crossentropy": 2.0083594918251038,
      "loss/hidden": 3.3453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19412665143609048,
      "step": 7940
    },
    {
      "epoch": 0.19875,
      "grad_norm": 31.625,
      "grad_norm_var": 23.877083333333335,
      "learning_rate": 0.0001,
      "loss": 7.441,
      "loss/crossentropy": 2.08128562271595,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17889103144407273,
      "step": 7950
    },
    {
      "epoch": 0.199,
      "grad_norm": 27.875,
      "grad_norm_var": 1.5921223958333333,
      "learning_rate": 0.0001,
      "loss": 7.4251,
      "loss/crossentropy": 2.1019147261977196,
      "loss/hidden": 3.571484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2211546439677477,
      "step": 7960
    },
    {
      "epoch": 0.19925,
      "grad_norm": 32.0,
      "grad_norm_var": 3.3059895833333335,
      "learning_rate": 0.0001,
      "loss": 7.4213,
      "loss/crossentropy": 2.14247687458992,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18870262056589127,
      "step": 7970
    },
    {
      "epoch": 0.1995,
      "grad_norm": 31.5,
      "grad_norm_var": 2.0947265625,
      "learning_rate": 0.0001,
      "loss": 7.5788,
      "loss/crossentropy": 2.1896591186523438,
      "loss/hidden": 3.479296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20883973222225904,
      "step": 7980
    },
    {
      "epoch": 0.19975,
      "grad_norm": 28.25,
      "grad_norm_var": 2.982291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5217,
      "loss/crossentropy": 2.1884095311164855,
      "loss/hidden": 3.44296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2022842913866043,
      "step": 7990
    },
    {
      "epoch": 0.2,
      "grad_norm": 28.5,
      "grad_norm_var": 2.379622395833333,
      "learning_rate": 0.0001,
      "loss": 7.4975,
      "loss/crossentropy": 2.225523295998573,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2032675376161933,
      "step": 8000
    },
    {
      "epoch": 0.20025,
      "grad_norm": 30.625,
      "grad_norm_var": 4.181705729166667,
      "learning_rate": 0.0001,
      "loss": 7.3874,
      "loss/crossentropy": 1.9566738605499268,
      "loss/hidden": 3.576953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19702311754226684,
      "step": 8010
    },
    {
      "epoch": 0.2005,
      "grad_norm": 32.75,
      "grad_norm_var": 6.3509765625,
      "learning_rate": 0.0001,
      "loss": 7.4528,
      "loss/crossentropy": 2.1517204724252226,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2005010774359107,
      "step": 8020
    },
    {
      "epoch": 0.20075,
      "grad_norm": 30.875,
      "grad_norm_var": 5.44140625,
      "learning_rate": 0.0001,
      "loss": 7.5306,
      "loss/crossentropy": 2.0300184957683087,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18887464031577111,
      "step": 8030
    },
    {
      "epoch": 0.201,
      "grad_norm": 31.75,
      "grad_norm_var": 3.1302083333333335,
      "learning_rate": 0.0001,
      "loss": 7.4025,
      "loss/crossentropy": 2.0889772072434427,
      "loss/hidden": 3.27109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17016669576987625,
      "step": 8040
    },
    {
      "epoch": 0.20125,
      "grad_norm": 27.125,
      "grad_norm_var": 94.78125,
      "learning_rate": 0.0001,
      "loss": 7.4379,
      "loss/crossentropy": 2.158484524488449,
      "loss/hidden": 3.410546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19672231934964657,
      "step": 8050
    },
    {
      "epoch": 0.2015,
      "grad_norm": 38.0,
      "grad_norm_var": 13.220572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4314,
      "loss/crossentropy": 1.9623262777924537,
      "loss/hidden": 3.339453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1833876773715019,
      "step": 8060
    },
    {
      "epoch": 0.20175,
      "grad_norm": 29.875,
      "grad_norm_var": 7.858268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4662,
      "loss/crossentropy": 2.2177498638629913,
      "loss/hidden": 3.383203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1954023003578186,
      "step": 8070
    },
    {
      "epoch": 0.202,
      "grad_norm": 28.875,
      "grad_norm_var": 7.627083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4792,
      "loss/crossentropy": 2.1019868202507497,
      "loss/hidden": 3.44375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1995641984976828,
      "step": 8080
    },
    {
      "epoch": 0.20225,
      "grad_norm": 44.75,
      "grad_norm_var": 20.4306640625,
      "learning_rate": 0.0001,
      "loss": 7.5664,
      "loss/crossentropy": 2.299755599349737,
      "loss/hidden": 3.3015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19075682908296585,
      "step": 8090
    },
    {
      "epoch": 0.2025,
      "grad_norm": 37.75,
      "grad_norm_var": 2.5671221292944763e+18,
      "learning_rate": 0.0001,
      "loss": 7.4526,
      "loss/crossentropy": 2.131952489167452,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1955398641526699,
      "step": 8100
    },
    {
      "epoch": 0.20275,
      "grad_norm": 31.25,
      "grad_norm_var": 20.342643229166665,
      "learning_rate": 0.0001,
      "loss": 7.4687,
      "loss/crossentropy": 1.9825796701014042,
      "loss/hidden": 3.475,
      "loss/jsd": 0.0,
      "loss/logits": 0.20226136669516565,
      "step": 8110
    },
    {
      "epoch": 0.203,
      "grad_norm": 27.125,
      "grad_norm_var": 15.192643229166666,
      "learning_rate": 0.0001,
      "loss": 7.1604,
      "loss/crossentropy": 2.0296560734510423,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1786259189248085,
      "step": 8120
    },
    {
      "epoch": 0.20325,
      "grad_norm": 31.0,
      "grad_norm_var": 7.1775390625,
      "learning_rate": 0.0001,
      "loss": 7.2815,
      "loss/crossentropy": 2.104996609687805,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19307580199092628,
      "step": 8130
    },
    {
      "epoch": 0.2035,
      "grad_norm": 34.75,
      "grad_norm_var": 6.820572916666666,
      "learning_rate": 0.0001,
      "loss": 7.3774,
      "loss/crossentropy": 2.1900447353720667,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20103423558175565,
      "step": 8140
    },
    {
      "epoch": 0.20375,
      "grad_norm": 31.0,
      "grad_norm_var": 5.48125,
      "learning_rate": 0.0001,
      "loss": 7.3628,
      "loss/crossentropy": 2.0671978294849396,
      "loss/hidden": 3.338671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18412660714238882,
      "step": 8150
    },
    {
      "epoch": 0.204,
      "grad_norm": 41.25,
      "grad_norm_var": 14.245572916666667,
      "learning_rate": 0.0001,
      "loss": 7.5547,
      "loss/crossentropy": 2.05537860840559,
      "loss/hidden": 3.48515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1820721985772252,
      "step": 8160
    },
    {
      "epoch": 0.20425,
      "grad_norm": 29.125,
      "grad_norm_var": 13.8625,
      "learning_rate": 0.0001,
      "loss": 7.345,
      "loss/crossentropy": 1.9336151838302613,
      "loss/hidden": 3.564453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20152895338833332,
      "step": 8170
    },
    {
      "epoch": 0.2045,
      "grad_norm": 28.125,
      "grad_norm_var": 7.096875,
      "learning_rate": 0.0001,
      "loss": 7.5005,
      "loss/crossentropy": 2.055295965075493,
      "loss/hidden": 3.35234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1865659100934863,
      "step": 8180
    },
    {
      "epoch": 0.20475,
      "grad_norm": 28.25,
      "grad_norm_var": 8.805989583333334,
      "learning_rate": 0.0001,
      "loss": 7.5324,
      "loss/crossentropy": 1.9976115971803665,
      "loss/hidden": 3.508203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2114253517240286,
      "step": 8190
    },
    {
      "epoch": 0.205,
      "grad_norm": 29.875,
      "grad_norm_var": 12.435416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4232,
      "loss/crossentropy": 2.1122905567288397,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21649520397186278,
      "step": 8200
    },
    {
      "epoch": 0.20525,
      "grad_norm": 28.125,
      "grad_norm_var": 10.055208333333333,
      "learning_rate": 0.0001,
      "loss": 7.4875,
      "loss/crossentropy": 2.1132961876690386,
      "loss/hidden": 3.47734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2026100393384695,
      "step": 8210
    },
    {
      "epoch": 0.2055,
      "grad_norm": 29.125,
      "grad_norm_var": 6.793489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3834,
      "loss/crossentropy": 2.191230720281601,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19885572660714387,
      "step": 8220
    },
    {
      "epoch": 0.20575,
      "grad_norm": 30.75,
      "grad_norm_var": 4.630208333333333,
      "learning_rate": 0.0001,
      "loss": 7.4544,
      "loss/crossentropy": 2.080083931982517,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2282587742432952,
      "step": 8230
    },
    {
      "epoch": 0.206,
      "grad_norm": 31.375,
      "grad_norm_var": 11.458268229166666,
      "learning_rate": 0.0001,
      "loss": 7.449,
      "loss/crossentropy": 2.1384637162089346,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1842938730493188,
      "step": 8240
    },
    {
      "epoch": 0.20625,
      "grad_norm": 31.875,
      "grad_norm_var": 1.7087890625,
      "learning_rate": 0.0001,
      "loss": 7.3946,
      "loss/crossentropy": 2.1355771869421005,
      "loss/hidden": 3.264453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18052869867533444,
      "step": 8250
    },
    {
      "epoch": 0.2065,
      "grad_norm": 31.75,
      "grad_norm_var": 13.0447265625,
      "learning_rate": 0.0001,
      "loss": 7.6316,
      "loss/crossentropy": 2.0203320410102608,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.192235934920609,
      "step": 8260
    },
    {
      "epoch": 0.20675,
      "grad_norm": 29.5,
      "grad_norm_var": 12.270247395833334,
      "learning_rate": 0.0001,
      "loss": 7.4572,
      "loss/crossentropy": 1.9869592547416688,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.202651490829885,
      "step": 8270
    },
    {
      "epoch": 0.207,
      "grad_norm": 30.125,
      "grad_norm_var": 1.4822916666666666,
      "learning_rate": 0.0001,
      "loss": 7.5227,
      "loss/crossentropy": 2.051181730628014,
      "loss/hidden": 3.43984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20994498692452906,
      "step": 8280
    },
    {
      "epoch": 0.20725,
      "grad_norm": 43.75,
      "grad_norm_var": 12.776822916666667,
      "learning_rate": 0.0001,
      "loss": 7.3664,
      "loss/crossentropy": 1.9618318520486355,
      "loss/hidden": 3.558984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19053993374109268,
      "step": 8290
    },
    {
      "epoch": 0.2075,
      "grad_norm": 30.5,
      "grad_norm_var": 14.92890625,
      "learning_rate": 0.0001,
      "loss": 7.318,
      "loss/crossentropy": 2.0114831268787383,
      "loss/hidden": 3.459375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20329152811318635,
      "step": 8300
    },
    {
      "epoch": 0.20775,
      "grad_norm": 34.0,
      "grad_norm_var": 2.8869140625,
      "learning_rate": 0.0001,
      "loss": 7.5489,
      "loss/crossentropy": 2.1187786638736723,
      "loss/hidden": 3.512890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20839224103838205,
      "step": 8310
    },
    {
      "epoch": 0.208,
      "grad_norm": 27.875,
      "grad_norm_var": 2.565559895833333,
      "learning_rate": 0.0001,
      "loss": 7.3443,
      "loss/crossentropy": 2.1263110756874086,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18627664018422366,
      "step": 8320
    },
    {
      "epoch": 0.20825,
      "grad_norm": 30.375,
      "grad_norm_var": 2.755989583333333,
      "learning_rate": 0.0001,
      "loss": 7.3857,
      "loss/crossentropy": 1.944861602783203,
      "loss/hidden": 3.468359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18570939563214778,
      "step": 8330
    },
    {
      "epoch": 0.2085,
      "grad_norm": 31.125,
      "grad_norm_var": 2.340625,
      "learning_rate": 0.0001,
      "loss": 7.5344,
      "loss/crossentropy": 2.1811724051833155,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19298948515206577,
      "step": 8340
    },
    {
      "epoch": 0.20875,
      "grad_norm": 31.5,
      "grad_norm_var": 1.56015625,
      "learning_rate": 0.0001,
      "loss": 7.3826,
      "loss/crossentropy": 2.152976579964161,
      "loss/hidden": 3.305859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.190250195749104,
      "step": 8350
    },
    {
      "epoch": 0.209,
      "grad_norm": 28.875,
      "grad_norm_var": 1.3416015625,
      "learning_rate": 0.0001,
      "loss": 7.323,
      "loss/crossentropy": 2.2053099036216737,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1929330924525857,
      "step": 8360
    },
    {
      "epoch": 0.20925,
      "grad_norm": 29.875,
      "grad_norm_var": 1.65625,
      "learning_rate": 0.0001,
      "loss": 7.4376,
      "loss/crossentropy": 2.011850906908512,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20794765576720237,
      "step": 8370
    },
    {
      "epoch": 0.2095,
      "grad_norm": 29.25,
      "grad_norm_var": 2.381705729166667,
      "learning_rate": 0.0001,
      "loss": 7.4518,
      "loss/crossentropy": 2.284806078672409,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2203810729086399,
      "step": 8380
    },
    {
      "epoch": 0.20975,
      "grad_norm": 43.25,
      "grad_norm_var": 17.695247395833334,
      "learning_rate": 0.0001,
      "loss": 7.418,
      "loss/crossentropy": 2.1161764934659004,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1863908626139164,
      "step": 8390
    },
    {
      "epoch": 0.21,
      "grad_norm": 29.625,
      "grad_norm_var": 19.8697265625,
      "learning_rate": 0.0001,
      "loss": 7.4564,
      "loss/crossentropy": 2.0293585821986198,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1884205201640725,
      "step": 8400
    },
    {
      "epoch": 0.21025,
      "grad_norm": 27.25,
      "grad_norm_var": 11.422330729166667,
      "learning_rate": 0.0001,
      "loss": 7.4317,
      "loss/crossentropy": 2.2351802065968513,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20424611177295446,
      "step": 8410
    },
    {
      "epoch": 0.2105,
      "grad_norm": 30.125,
      "grad_norm_var": 10.734830729166667,
      "learning_rate": 0.0001,
      "loss": 7.3989,
      "loss/crossentropy": 2.1349810734391212,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18979013338685036,
      "step": 8420
    },
    {
      "epoch": 0.21075,
      "grad_norm": 30.875,
      "grad_norm_var": 2.387955729166667,
      "learning_rate": 0.0001,
      "loss": 7.3957,
      "loss/crossentropy": 2.080636392533779,
      "loss/hidden": 3.371875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18893024744465947,
      "step": 8430
    },
    {
      "epoch": 0.211,
      "grad_norm": 28.75,
      "grad_norm_var": 20.089518229166668,
      "learning_rate": 0.0001,
      "loss": 7.4847,
      "loss/crossentropy": 1.9267802774906158,
      "loss/hidden": 3.31484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18156335428357123,
      "step": 8440
    },
    {
      "epoch": 0.21125,
      "grad_norm": 29.125,
      "grad_norm_var": 2.2650390625,
      "learning_rate": 0.0001,
      "loss": 7.3839,
      "loss/crossentropy": 2.14048397988081,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889717074111104,
      "step": 8450
    },
    {
      "epoch": 0.2115,
      "grad_norm": 28.375,
      "grad_norm_var": 3.283124099188418e+18,
      "learning_rate": 0.0001,
      "loss": 7.4661,
      "loss/crossentropy": 2.1406930878758432,
      "loss/hidden": 3.699609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20904620084911585,
      "step": 8460
    },
    {
      "epoch": 0.21175,
      "grad_norm": 30.75,
      "grad_norm_var": 23.509830729166666,
      "learning_rate": 0.0001,
      "loss": 7.2867,
      "loss/crossentropy": 2.1153231114149094,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18998019583523273,
      "step": 8470
    },
    {
      "epoch": 0.212,
      "grad_norm": 31.0,
      "grad_norm_var": 2.3139973958333333,
      "learning_rate": 0.0001,
      "loss": 7.3997,
      "loss/crossentropy": 2.169650764763355,
      "loss/hidden": 3.276953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17556187361478806,
      "step": 8480
    },
    {
      "epoch": 0.21225,
      "grad_norm": 29.625,
      "grad_norm_var": 2.312239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4486,
      "loss/crossentropy": 2.128412726521492,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1975807584822178,
      "step": 8490
    },
    {
      "epoch": 0.2125,
      "grad_norm": 30.75,
      "grad_norm_var": 2.903125,
      "learning_rate": 0.0001,
      "loss": 7.4526,
      "loss/crossentropy": 2.029564914107323,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19610330546274782,
      "step": 8500
    },
    {
      "epoch": 0.21275,
      "grad_norm": 31.375,
      "grad_norm_var": 4.004622395833334,
      "learning_rate": 0.0001,
      "loss": 7.2845,
      "loss/crossentropy": 1.9631854377686977,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1855375848710537,
      "step": 8510
    },
    {
      "epoch": 0.213,
      "grad_norm": 30.0,
      "grad_norm_var": 3.476822916666667,
      "learning_rate": 0.0001,
      "loss": 7.4124,
      "loss/crossentropy": 2.1494273841381073,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19540442936122418,
      "step": 8520
    },
    {
      "epoch": 0.21325,
      "grad_norm": 31.625,
      "grad_norm_var": 0.9452473958333333,
      "learning_rate": 0.0001,
      "loss": 7.4659,
      "loss/crossentropy": 2.0508621491491796,
      "loss/hidden": 3.3078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18023168351501226,
      "step": 8530
    },
    {
      "epoch": 0.2135,
      "grad_norm": 31.25,
      "grad_norm_var": 2.5809895833333334,
      "learning_rate": 0.0001,
      "loss": 7.4522,
      "loss/crossentropy": 2.090715576708317,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18382459450513125,
      "step": 8540
    },
    {
      "epoch": 0.21375,
      "grad_norm": 28.75,
      "grad_norm_var": 2.986168000807314e+18,
      "learning_rate": 0.0001,
      "loss": 7.4917,
      "loss/crossentropy": 2.177123633027077,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.197673611715436,
      "step": 8550
    },
    {
      "epoch": 0.214,
      "grad_norm": 27.75,
      "grad_norm_var": 4.035872395833334,
      "learning_rate": 0.0001,
      "loss": 7.3008,
      "loss/crossentropy": 2.028589369356632,
      "loss/hidden": 3.383984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.196718043461442,
      "step": 8560
    },
    {
      "epoch": 0.21425,
      "grad_norm": 30.5,
      "grad_norm_var": 4.758072916666666,
      "learning_rate": 0.0001,
      "loss": 7.3829,
      "loss/crossentropy": 2.02908306196332,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1757409404963255,
      "step": 8570
    },
    {
      "epoch": 0.2145,
      "grad_norm": 30.875,
      "grad_norm_var": 1.9754557291666666,
      "learning_rate": 0.0001,
      "loss": 7.4107,
      "loss/crossentropy": 2.0411842301487924,
      "loss/hidden": 3.580078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2309743857011199,
      "step": 8580
    },
    {
      "epoch": 0.21475,
      "grad_norm": 30.0,
      "grad_norm_var": 25.7244140625,
      "learning_rate": 0.0001,
      "loss": 7.4038,
      "loss/crossentropy": 2.1026074662804604,
      "loss/hidden": 3.490234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.202506691403687,
      "step": 8590
    },
    {
      "epoch": 0.215,
      "grad_norm": 28.625,
      "grad_norm_var": 3.386458333333333,
      "learning_rate": 0.0001,
      "loss": 7.3559,
      "loss/crossentropy": 2.1690615713596344,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19095612335950135,
      "step": 8600
    },
    {
      "epoch": 0.21525,
      "grad_norm": 29.0,
      "grad_norm_var": 23.880989583333335,
      "learning_rate": 0.0001,
      "loss": 7.4164,
      "loss/crossentropy": 2.099227898567915,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19162636022083462,
      "step": 8610
    },
    {
      "epoch": 0.2155,
      "grad_norm": 34.25,
      "grad_norm_var": 23.880208333333332,
      "learning_rate": 0.0001,
      "loss": 7.3725,
      "loss/crossentropy": 1.9689884655177594,
      "loss/hidden": 3.52265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1830376190133393,
      "step": 8620
    },
    {
      "epoch": 0.21575,
      "grad_norm": 29.875,
      "grad_norm_var": 2.77890625,
      "learning_rate": 0.0001,
      "loss": 7.4511,
      "loss/crossentropy": 2.0263702854514123,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18029460608959197,
      "step": 8630
    },
    {
      "epoch": 0.216,
      "grad_norm": 32.0,
      "grad_norm_var": 23.001822916666665,
      "learning_rate": 0.0001,
      "loss": 7.3863,
      "loss/crossentropy": 1.9046964697539805,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1935629203915596,
      "step": 8640
    },
    {
      "epoch": 0.21625,
      "grad_norm": 33.5,
      "grad_norm_var": 25.009375,
      "learning_rate": 0.0001,
      "loss": 7.3083,
      "loss/crossentropy": 2.129426471889019,
      "loss/hidden": 3.448046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20037918202579022,
      "step": 8650
    },
    {
      "epoch": 0.2165,
      "grad_norm": 28.375,
      "grad_norm_var": 12.242643229166667,
      "learning_rate": 0.0001,
      "loss": 7.4603,
      "loss/crossentropy": 2.2266604125499727,
      "loss/hidden": 3.33203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18952292017638683,
      "step": 8660
    },
    {
      "epoch": 0.21675,
      "grad_norm": 34.75,
      "grad_norm_var": 19.517708333333335,
      "learning_rate": 0.0001,
      "loss": 7.4031,
      "loss/crossentropy": 1.9756697475910188,
      "loss/hidden": 3.529296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20053059812635182,
      "step": 8670
    },
    {
      "epoch": 0.217,
      "grad_norm": 29.25,
      "grad_norm_var": 8.315625,
      "learning_rate": 0.0001,
      "loss": 7.4013,
      "loss/crossentropy": 2.13729098290205,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19440573658794164,
      "step": 8680
    },
    {
      "epoch": 0.21725,
      "grad_norm": 31.5,
      "grad_norm_var": 1.9936848958333333,
      "learning_rate": 0.0001,
      "loss": 7.2982,
      "loss/crossentropy": 2.0935462579131126,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.200434254668653,
      "step": 8690
    },
    {
      "epoch": 0.2175,
      "grad_norm": 31.625,
      "grad_norm_var": 26.66640625,
      "learning_rate": 0.0001,
      "loss": 7.4596,
      "loss/crossentropy": 2.0081637501716614,
      "loss/hidden": 3.42734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.196201959438622,
      "step": 8700
    },
    {
      "epoch": 0.21775,
      "grad_norm": 28.125,
      "grad_norm_var": 1.869627142435242e+18,
      "learning_rate": 0.0001,
      "loss": 7.4726,
      "loss/crossentropy": 2.057238683104515,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18527965154498816,
      "step": 8710
    },
    {
      "epoch": 0.218,
      "grad_norm": 30.25,
      "grad_norm_var": 4.806705729166667,
      "learning_rate": 0.0001,
      "loss": 7.424,
      "loss/crossentropy": 2.225848586857319,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.20839223694056272,
      "step": 8720
    },
    {
      "epoch": 0.21825,
      "grad_norm": 29.5,
      "grad_norm_var": 3.2572265625,
      "learning_rate": 0.0001,
      "loss": 7.3713,
      "loss/crossentropy": 1.9737806752324105,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18166052605956792,
      "step": 8730
    },
    {
      "epoch": 0.2185,
      "grad_norm": 31.25,
      "grad_norm_var": 3.49765625,
      "learning_rate": 0.0001,
      "loss": 7.3192,
      "loss/crossentropy": 2.1625877559185027,
      "loss/hidden": 3.3765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.196309875510633,
      "step": 8740
    },
    {
      "epoch": 0.21875,
      "grad_norm": 31.625,
      "grad_norm_var": 15.733333333333333,
      "learning_rate": 0.0001,
      "loss": 7.3994,
      "loss/crossentropy": 2.1306996777653695,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17798179090023042,
      "step": 8750
    },
    {
      "epoch": 0.219,
      "grad_norm": 31.25,
      "grad_norm_var": 5.465625,
      "learning_rate": 0.0001,
      "loss": 7.4225,
      "loss/crossentropy": 2.168031161278486,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19842574130743743,
      "step": 8760
    },
    {
      "epoch": 0.21925,
      "grad_norm": 29.5,
      "grad_norm_var": 3.457747395833333,
      "learning_rate": 0.0001,
      "loss": 7.4515,
      "loss/crossentropy": 2.081401216983795,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1922367751598358,
      "step": 8770
    },
    {
      "epoch": 0.2195,
      "grad_norm": 28.625,
      "grad_norm_var": 2.9389973958333333,
      "learning_rate": 0.0001,
      "loss": 7.4641,
      "loss/crossentropy": 1.9896476596593857,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18900877684354783,
      "step": 8780
    },
    {
      "epoch": 0.21975,
      "grad_norm": 29.375,
      "grad_norm_var": 3.5931640625,
      "learning_rate": 0.0001,
      "loss": 7.5034,
      "loss/crossentropy": 2.131504286080599,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1981994620524347,
      "step": 8790
    },
    {
      "epoch": 0.22,
      "grad_norm": 29.75,
      "grad_norm_var": 17.612955729166668,
      "learning_rate": 0.0001,
      "loss": 7.3602,
      "loss/crossentropy": 2.367118790745735,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20705808699131012,
      "step": 8800
    },
    {
      "epoch": 0.22025,
      "grad_norm": 30.625,
      "grad_norm_var": 3.588997395833333,
      "learning_rate": 0.0001,
      "loss": 7.3512,
      "loss/crossentropy": 1.915165586769581,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18030493911355733,
      "step": 8810
    },
    {
      "epoch": 0.2205,
      "grad_norm": 31.625,
      "grad_norm_var": 2.732291666666667,
      "learning_rate": 0.0001,
      "loss": 7.3832,
      "loss/crossentropy": 2.1191729307174683,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20246538575738668,
      "step": 8820
    },
    {
      "epoch": 0.22075,
      "grad_norm": 32.0,
      "grad_norm_var": 2.034309895833333,
      "learning_rate": 0.0001,
      "loss": 7.4184,
      "loss/crossentropy": 2.280582541972399,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1913912059739232,
      "step": 8830
    },
    {
      "epoch": 0.221,
      "grad_norm": 28.75,
      "grad_norm_var": 1.4629557291666666,
      "learning_rate": 0.0001,
      "loss": 7.5177,
      "loss/crossentropy": 2.07154730707407,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18417379464954137,
      "step": 8840
    },
    {
      "epoch": 0.22125,
      "grad_norm": 29.25,
      "grad_norm_var": 2.0306640625,
      "learning_rate": 0.0001,
      "loss": 7.4098,
      "loss/crossentropy": 2.0918928742408753,
      "loss/hidden": 3.478515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21339783817529678,
      "step": 8850
    },
    {
      "epoch": 0.2215,
      "grad_norm": 34.75,
      "grad_norm_var": 2.4098307291666665,
      "learning_rate": 0.0001,
      "loss": 7.3907,
      "loss/crossentropy": 2.0262165658175944,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1963033676147461,
      "step": 8860
    },
    {
      "epoch": 0.22175,
      "grad_norm": 29.625,
      "grad_norm_var": 3.1041015625,
      "learning_rate": 0.0001,
      "loss": 7.3643,
      "loss/crossentropy": 2.119324280321598,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19243048634380103,
      "step": 8870
    },
    {
      "epoch": 0.222,
      "grad_norm": 29.25,
      "grad_norm_var": 3.3593098958333334,
      "learning_rate": 0.0001,
      "loss": 7.335,
      "loss/crossentropy": 2.1042064100503923,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2055317424237728,
      "step": 8880
    },
    {
      "epoch": 0.22225,
      "grad_norm": 29.0,
      "grad_norm_var": 4.35390625,
      "learning_rate": 0.0001,
      "loss": 7.3974,
      "loss/crossentropy": 2.110988216102123,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18536690715700388,
      "step": 8890
    },
    {
      "epoch": 0.2225,
      "grad_norm": 30.0,
      "grad_norm_var": 6.014322916666667,
      "learning_rate": 0.0001,
      "loss": 7.4155,
      "loss/crossentropy": 2.033397987484932,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19111265633255242,
      "step": 8900
    },
    {
      "epoch": 0.22275,
      "grad_norm": 32.25,
      "grad_norm_var": 4.747916666666667,
      "learning_rate": 0.0001,
      "loss": 7.4631,
      "loss/crossentropy": 2.090746468305588,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20997797157615422,
      "step": 8910
    },
    {
      "epoch": 0.223,
      "grad_norm": 32.75,
      "grad_norm_var": 2.0247395833333335,
      "learning_rate": 0.0001,
      "loss": 7.3647,
      "loss/crossentropy": 2.107650229334831,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18095682561397552,
      "step": 8920
    },
    {
      "epoch": 0.22325,
      "grad_norm": 33.0,
      "grad_norm_var": 4.983072916666667,
      "learning_rate": 0.0001,
      "loss": 7.4161,
      "loss/crossentropy": 2.1033721581101417,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1837721960619092,
      "step": 8930
    },
    {
      "epoch": 0.2235,
      "grad_norm": 30.0,
      "grad_norm_var": 5.160872395833334,
      "learning_rate": 0.0001,
      "loss": 7.381,
      "loss/crossentropy": 2.2039036631584166,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20040026511996983,
      "step": 8940
    },
    {
      "epoch": 0.22375,
      "grad_norm": 32.75,
      "grad_norm_var": 5.625455729166666,
      "learning_rate": 0.0001,
      "loss": 7.4581,
      "loss/crossentropy": 2.0015091970562935,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1968998895958066,
      "step": 8950
    },
    {
      "epoch": 0.224,
      "grad_norm": 30.125,
      "grad_norm_var": 7.976497395833333,
      "learning_rate": 0.0001,
      "loss": 7.446,
      "loss/crossentropy": 2.1770635031163694,
      "loss/hidden": 3.290234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18164771795272827,
      "step": 8960
    },
    {
      "epoch": 0.22425,
      "grad_norm": 26.875,
      "grad_norm_var": 2.037239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2677,
      "loss/crossentropy": 1.9846746385097505,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19175144601613284,
      "step": 8970
    },
    {
      "epoch": 0.2245,
      "grad_norm": 30.25,
      "grad_norm_var": 2.8353515625,
      "learning_rate": 0.0001,
      "loss": 7.4645,
      "loss/crossentropy": 2.169223573803902,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18991702441126107,
      "step": 8980
    },
    {
      "epoch": 0.22475,
      "grad_norm": 29.875,
      "grad_norm_var": 10.3228515625,
      "learning_rate": 0.0001,
      "loss": 7.4607,
      "loss/crossentropy": 2.1323930069804193,
      "loss/hidden": 3.348046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18631141390651465,
      "step": 8990
    },
    {
      "epoch": 0.225,
      "grad_norm": 29.625,
      "grad_norm_var": 11.431705729166667,
      "learning_rate": 0.0001,
      "loss": 7.4203,
      "loss/crossentropy": 2.1948125064373016,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21075339019298553,
      "step": 9000
    },
    {
      "epoch": 0.22525,
      "grad_norm": 31.0,
      "grad_norm_var": 3.21875,
      "learning_rate": 0.0001,
      "loss": 7.3937,
      "loss/crossentropy": 2.061431697010994,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20084240343421697,
      "step": 9010
    },
    {
      "epoch": 0.2255,
      "grad_norm": 30.0,
      "grad_norm_var": 9.7125,
      "learning_rate": 0.0001,
      "loss": 7.5364,
      "loss/crossentropy": 2.1422473564743996,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20135180205106734,
      "step": 9020
    },
    {
      "epoch": 0.22575,
      "grad_norm": 29.0,
      "grad_norm_var": 10.320247395833333,
      "learning_rate": 0.0001,
      "loss": 7.3638,
      "loss/crossentropy": 2.064805781841278,
      "loss/hidden": 3.509375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19638751186430453,
      "step": 9030
    },
    {
      "epoch": 0.226,
      "grad_norm": 29.0,
      "grad_norm_var": 7.7306640625,
      "learning_rate": 0.0001,
      "loss": 7.367,
      "loss/crossentropy": 2.0203323513269424,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1730576554313302,
      "step": 9040
    },
    {
      "epoch": 0.22625,
      "grad_norm": 28.0,
      "grad_norm_var": 7.5291015625,
      "learning_rate": 0.0001,
      "loss": 7.4083,
      "loss/crossentropy": 2.1449467122554777,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20745128113776445,
      "step": 9050
    },
    {
      "epoch": 0.2265,
      "grad_norm": 29.625,
      "grad_norm_var": 10.059830729166666,
      "learning_rate": 0.0001,
      "loss": 7.3739,
      "loss/crossentropy": 2.277996188402176,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1961175424978137,
      "step": 9060
    },
    {
      "epoch": 0.22675,
      "grad_norm": 29.75,
      "grad_norm_var": 2.9497395833333333,
      "learning_rate": 0.0001,
      "loss": 7.425,
      "loss/crossentropy": 2.128904873877764,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1912422338500619,
      "step": 9070
    },
    {
      "epoch": 0.227,
      "grad_norm": 30.0,
      "grad_norm_var": 2.4291666666666667,
      "learning_rate": 0.0001,
      "loss": 7.3724,
      "loss/crossentropy": 1.8966167330741883,
      "loss/hidden": 3.315234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1772749178111553,
      "step": 9080
    },
    {
      "epoch": 0.22725,
      "grad_norm": 30.625,
      "grad_norm_var": 2.2955729166666665,
      "learning_rate": 0.0001,
      "loss": 7.4341,
      "loss/crossentropy": 1.9993775576353072,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17548400331288577,
      "step": 9090
    },
    {
      "epoch": 0.2275,
      "grad_norm": 30.125,
      "grad_norm_var": 2.3212890625,
      "learning_rate": 0.0001,
      "loss": 7.4512,
      "loss/crossentropy": 2.1553252935409546,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1904754728078842,
      "step": 9100
    },
    {
      "epoch": 0.22775,
      "grad_norm": 31.875,
      "grad_norm_var": 1.8264973958333333,
      "learning_rate": 0.0001,
      "loss": 7.4299,
      "loss/crossentropy": 2.0791067980229854,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18264174591749907,
      "step": 9110
    },
    {
      "epoch": 0.228,
      "grad_norm": 30.625,
      "grad_norm_var": 8.670572916666666,
      "learning_rate": 0.0001,
      "loss": 7.3621,
      "loss/crossentropy": 2.2495081633329392,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19295338317751884,
      "step": 9120
    },
    {
      "epoch": 0.22825,
      "grad_norm": 34.25,
      "grad_norm_var": 9.628580729166666,
      "learning_rate": 0.0001,
      "loss": 7.4069,
      "loss/crossentropy": 2.0221784450113773,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18131311442703008,
      "step": 9130
    },
    {
      "epoch": 0.2285,
      "grad_norm": 31.375,
      "grad_norm_var": 3.8372395833333335,
      "learning_rate": 0.0001,
      "loss": 7.3788,
      "loss/crossentropy": 2.096472094208002,
      "loss/hidden": 3.533984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19181215222924947,
      "step": 9140
    },
    {
      "epoch": 0.22875,
      "grad_norm": 32.75,
      "grad_norm_var": 2.9369140625,
      "learning_rate": 0.0001,
      "loss": 7.4625,
      "loss/crossentropy": 2.161876367032528,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19645992666482925,
      "step": 9150
    },
    {
      "epoch": 0.229,
      "grad_norm": 36.5,
      "grad_norm_var": 13.144791666666666,
      "learning_rate": 0.0001,
      "loss": 7.3883,
      "loss/crossentropy": 1.870260328054428,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19089705124497414,
      "step": 9160
    },
    {
      "epoch": 0.22925,
      "grad_norm": 30.125,
      "grad_norm_var": 11.746809895833334,
      "learning_rate": 0.0001,
      "loss": 7.3778,
      "loss/crossentropy": 2.304895442724228,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19963221047073604,
      "step": 9170
    },
    {
      "epoch": 0.2295,
      "grad_norm": 30.875,
      "grad_norm_var": 2.64839803006287e+18,
      "learning_rate": 0.0001,
      "loss": 7.4899,
      "loss/crossentropy": 2.1439118653535845,
      "loss/hidden": 3.46484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20680125001817942,
      "step": 9180
    },
    {
      "epoch": 0.22975,
      "grad_norm": 29.625,
      "grad_norm_var": 16.839518229166668,
      "learning_rate": 0.0001,
      "loss": 7.3153,
      "loss/crossentropy": 1.9939407154917717,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19113040501251816,
      "step": 9190
    },
    {
      "epoch": 0.23,
      "grad_norm": 30.125,
      "grad_norm_var": 1.8309895833333334,
      "learning_rate": 0.0001,
      "loss": 7.5357,
      "loss/crossentropy": 2.088300883769989,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19538584928959607,
      "step": 9200
    },
    {
      "epoch": 0.23025,
      "grad_norm": 30.75,
      "grad_norm_var": 1.66640625,
      "learning_rate": 0.0001,
      "loss": 7.3255,
      "loss/crossentropy": 2.0176690459251403,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1823873495683074,
      "step": 9210
    },
    {
      "epoch": 0.2305,
      "grad_norm": 32.75,
      "grad_norm_var": 4.837955729166667,
      "learning_rate": 0.0001,
      "loss": 7.3334,
      "loss/crossentropy": 2.061754436790943,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18769590836018324,
      "step": 9220
    },
    {
      "epoch": 0.23075,
      "grad_norm": 31.75,
      "grad_norm_var": 4.683268229166667,
      "learning_rate": 0.0001,
      "loss": 7.3932,
      "loss/crossentropy": 2.0443666532635687,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19146692994982004,
      "step": 9230
    },
    {
      "epoch": 0.231,
      "grad_norm": 33.25,
      "grad_norm_var": 2.594073359575469e+18,
      "learning_rate": 0.0001,
      "loss": 7.5992,
      "loss/crossentropy": 2.189889648556709,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19348510541021824,
      "step": 9240
    },
    {
      "epoch": 0.23125,
      "grad_norm": 29.125,
      "grad_norm_var": 2.594073359702976e+18,
      "learning_rate": 0.0001,
      "loss": 7.4217,
      "loss/crossentropy": 2.0103170931339265,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19076041504740715,
      "step": 9250
    },
    {
      "epoch": 0.2315,
      "grad_norm": 31.375,
      "grad_norm_var": 13.5134765625,
      "learning_rate": 0.0001,
      "loss": 7.3004,
      "loss/crossentropy": 2.0597189858555796,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18948603458702565,
      "step": 9260
    },
    {
      "epoch": 0.23175,
      "grad_norm": 31.625,
      "grad_norm_var": 13.402083333333334,
      "learning_rate": 0.0001,
      "loss": 7.3946,
      "loss/crossentropy": 2.0123729363083838,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1885578565299511,
      "step": 9270
    },
    {
      "epoch": 0.232,
      "grad_norm": 29.875,
      "grad_norm_var": 5.879622395833334,
      "learning_rate": 0.0001,
      "loss": 7.1578,
      "loss/crossentropy": 2.2501363843679427,
      "loss/hidden": 3.286328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19268405642360448,
      "step": 9280
    },
    {
      "epoch": 0.23225,
      "grad_norm": 30.25,
      "grad_norm_var": 16.982291666666665,
      "learning_rate": 0.0001,
      "loss": 7.3481,
      "loss/crossentropy": 2.076053886115551,
      "loss/hidden": 3.571875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21077420487999915,
      "step": 9290
    },
    {
      "epoch": 0.2325,
      "grad_norm": 29.5,
      "grad_norm_var": 21.769791666666666,
      "learning_rate": 0.0001,
      "loss": 7.2711,
      "loss/crossentropy": 2.193411388993263,
      "loss/hidden": 3.33828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18807493168860673,
      "step": 9300
    },
    {
      "epoch": 0.23275,
      "grad_norm": 34.25,
      "grad_norm_var": 12.355143229166666,
      "learning_rate": 0.0001,
      "loss": 7.3819,
      "loss/crossentropy": 2.166665832698345,
      "loss/hidden": 3.3203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18620893750339745,
      "step": 9310
    },
    {
      "epoch": 0.233,
      "grad_norm": 29.875,
      "grad_norm_var": 3.002083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4076,
      "loss/crossentropy": 2.022654353827238,
      "loss/hidden": 3.2984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18289813362061977,
      "step": 9320
    },
    {
      "epoch": 0.23325,
      "grad_norm": 30.0,
      "grad_norm_var": 3.0509765625,
      "learning_rate": 0.0001,
      "loss": 7.3206,
      "loss/crossentropy": 2.020027980953455,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18972874553874136,
      "step": 9330
    },
    {
      "epoch": 0.2335,
      "grad_norm": 31.125,
      "grad_norm_var": 2.098893229166667,
      "learning_rate": 0.0001,
      "loss": 7.4091,
      "loss/crossentropy": 2.029365235567093,
      "loss/hidden": 3.43125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2005317559465766,
      "step": 9340
    },
    {
      "epoch": 0.23375,
      "grad_norm": 29.5,
      "grad_norm_var": 10.520768229166666,
      "learning_rate": 0.0001,
      "loss": 7.3704,
      "loss/crossentropy": 1.95634398534894,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18962019477039577,
      "step": 9350
    },
    {
      "epoch": 0.234,
      "grad_norm": 27.75,
      "grad_norm_var": 13.475,
      "learning_rate": 0.0001,
      "loss": 7.3797,
      "loss/crossentropy": 2.063827896118164,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19280518041923642,
      "step": 9360
    },
    {
      "epoch": 0.23425,
      "grad_norm": 30.5,
      "grad_norm_var": 2.0869140625,
      "learning_rate": 0.0001,
      "loss": 7.4918,
      "loss/crossentropy": 2.068470099568367,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19361184667795897,
      "step": 9370
    },
    {
      "epoch": 0.2345,
      "grad_norm": 28.0,
      "grad_norm_var": 6.534830729166667,
      "learning_rate": 0.0001,
      "loss": 7.4265,
      "loss/crossentropy": 2.125886672735214,
      "loss/hidden": 3.401171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19355848152190447,
      "step": 9380
    },
    {
      "epoch": 0.23475,
      "grad_norm": 30.375,
      "grad_norm_var": 2.2770182291666665,
      "learning_rate": 0.0001,
      "loss": 7.3769,
      "loss/crossentropy": 2.2057121500372885,
      "loss/hidden": 3.28515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1843592157587409,
      "step": 9390
    },
    {
      "epoch": 0.235,
      "grad_norm": 87.5,
      "grad_norm_var": 202.43932291666667,
      "learning_rate": 0.0001,
      "loss": 7.4444,
      "loss/crossentropy": 2.17584248483181,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1986805137246847,
      "step": 9400
    },
    {
      "epoch": 0.23525,
      "grad_norm": 33.0,
      "grad_norm_var": 205.446875,
      "learning_rate": 0.0001,
      "loss": 7.5575,
      "loss/crossentropy": 1.9989879056811333,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21514312122017146,
      "step": 9410
    },
    {
      "epoch": 0.2355,
      "grad_norm": 33.25,
      "grad_norm_var": 3.4775390625,
      "learning_rate": 0.0001,
      "loss": 7.3965,
      "loss/crossentropy": 2.31248200237751,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20192647576332093,
      "step": 9420
    },
    {
      "epoch": 0.23575,
      "grad_norm": 34.0,
      "grad_norm_var": 2.5723307291666666,
      "learning_rate": 0.0001,
      "loss": 7.425,
      "loss/crossentropy": 2.0246524304151534,
      "loss/hidden": 3.43125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1936120893806219,
      "step": 9430
    },
    {
      "epoch": 0.236,
      "grad_norm": 31.5,
      "grad_norm_var": 3.8207682291666667,
      "learning_rate": 0.0001,
      "loss": 7.3458,
      "loss/crossentropy": 2.0714851915836334,
      "loss/hidden": 3.315625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19567677434533834,
      "step": 9440
    },
    {
      "epoch": 0.23625,
      "grad_norm": 32.25,
      "grad_norm_var": 2.0452473958333335,
      "learning_rate": 0.0001,
      "loss": 7.3934,
      "loss/crossentropy": 2.024892423301935,
      "loss/hidden": 3.541796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19850265365093947,
      "step": 9450
    },
    {
      "epoch": 0.2365,
      "grad_norm": 32.5,
      "grad_norm_var": 3.16640625,
      "learning_rate": 0.0001,
      "loss": 7.4192,
      "loss/crossentropy": 2.146814212203026,
      "loss/hidden": 3.43828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20048882197588683,
      "step": 9460
    },
    {
      "epoch": 0.23675,
      "grad_norm": 27.875,
      "grad_norm_var": 12.364322916666667,
      "learning_rate": 0.0001,
      "loss": 7.3745,
      "loss/crossentropy": 2.0004019677639007,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19019459020346402,
      "step": 9470
    },
    {
      "epoch": 0.237,
      "grad_norm": 28.75,
      "grad_norm_var": 12.885416666666666,
      "learning_rate": 0.0001,
      "loss": 7.2858,
      "loss/crossentropy": 2.0364832431077957,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18539349418133497,
      "step": 9480
    },
    {
      "epoch": 0.23725,
      "grad_norm": 26.75,
      "grad_norm_var": 4.488997395833334,
      "learning_rate": 0.0001,
      "loss": 7.2865,
      "loss/crossentropy": 2.257227724790573,
      "loss/hidden": 3.247265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1774066084995866,
      "step": 9490
    },
    {
      "epoch": 0.2375,
      "grad_norm": 28.25,
      "grad_norm_var": 5.15625,
      "learning_rate": 0.0001,
      "loss": 7.3742,
      "loss/crossentropy": 2.192533364892006,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18531391881406306,
      "step": 9500
    },
    {
      "epoch": 0.23775,
      "grad_norm": 28.5,
      "grad_norm_var": 8.7025390625,
      "learning_rate": 0.0001,
      "loss": 7.4007,
      "loss/crossentropy": 1.9954842567443847,
      "loss/hidden": 3.512890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21250668447464705,
      "step": 9510
    },
    {
      "epoch": 0.238,
      "grad_norm": 27.0,
      "grad_norm_var": 3.2018229166666665,
      "learning_rate": 0.0001,
      "loss": 7.3589,
      "loss/crossentropy": 2.0501152604818342,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19774708338081837,
      "step": 9520
    },
    {
      "epoch": 0.23825,
      "grad_norm": 31.125,
      "grad_norm_var": 2.471875,
      "learning_rate": 0.0001,
      "loss": 7.3611,
      "loss/crossentropy": 2.1446721121668815,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19521092902868986,
      "step": 9530
    },
    {
      "epoch": 0.2385,
      "grad_norm": 33.25,
      "grad_norm_var": 2.911393229166667,
      "learning_rate": 0.0001,
      "loss": 7.4037,
      "loss/crossentropy": 2.050874675065279,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18872325737029313,
      "step": 9540
    },
    {
      "epoch": 0.23875,
      "grad_norm": 32.25,
      "grad_norm_var": 1.69375,
      "learning_rate": 0.0001,
      "loss": 7.3851,
      "loss/crossentropy": 2.092629846930504,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1858744696713984,
      "step": 9550
    },
    {
      "epoch": 0.239,
      "grad_norm": 30.25,
      "grad_norm_var": 1.6229166666666666,
      "learning_rate": 0.0001,
      "loss": 7.324,
      "loss/crossentropy": 2.1389416724443437,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18115091007202863,
      "step": 9560
    },
    {
      "epoch": 0.23925,
      "grad_norm": 28.375,
      "grad_norm_var": 0.7893229166666667,
      "learning_rate": 0.0001,
      "loss": 7.3684,
      "loss/crossentropy": 2.1148220866918566,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18993774689733983,
      "step": 9570
    },
    {
      "epoch": 0.2395,
      "grad_norm": 29.125,
      "grad_norm_var": 1.0260416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4711,
      "loss/crossentropy": 1.987958113849163,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19200538750737906,
      "step": 9580
    },
    {
      "epoch": 0.23975,
      "grad_norm": 29.625,
      "grad_norm_var": 0.6018229166666667,
      "learning_rate": 0.0001,
      "loss": 7.3578,
      "loss/crossentropy": 2.076607885956764,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20001101978123187,
      "step": 9590
    },
    {
      "epoch": 0.24,
      "grad_norm": 30.75,
      "grad_norm_var": 1.4124348958333333,
      "learning_rate": 0.0001,
      "loss": 7.4458,
      "loss/crossentropy": 2.1496243715286254,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889566643163562,
      "step": 9600
    },
    {
      "epoch": 0.24025,
      "grad_norm": 29.875,
      "grad_norm_var": 1.04140625,
      "learning_rate": 0.0001,
      "loss": 7.4632,
      "loss/crossentropy": 2.162689308822155,
      "loss/hidden": 3.33828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19204493910074233,
      "step": 9610
    },
    {
      "epoch": 0.2405,
      "grad_norm": 27.375,
      "grad_norm_var": 1.5863932291666667,
      "learning_rate": 0.0001,
      "loss": 7.3897,
      "loss/crossentropy": 2.1323146484792233,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20718522872775794,
      "step": 9620
    },
    {
      "epoch": 0.24075,
      "grad_norm": 32.25,
      "grad_norm_var": 2.09375,
      "learning_rate": 0.0001,
      "loss": 7.2848,
      "loss/crossentropy": 2.0804166465997698,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1966247998178005,
      "step": 9630
    },
    {
      "epoch": 0.241,
      "grad_norm": 31.25,
      "grad_norm_var": 1.3559895833333333,
      "learning_rate": 0.0001,
      "loss": 7.4174,
      "loss/crossentropy": 2.3036757931113243,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18416995517909526,
      "step": 9640
    },
    {
      "epoch": 0.24125,
      "grad_norm": 33.0,
      "grad_norm_var": 2.41015625,
      "learning_rate": 0.0001,
      "loss": 7.2853,
      "loss/crossentropy": 1.9936259984970093,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21977887880057095,
      "step": 9650
    },
    {
      "epoch": 0.2415,
      "grad_norm": 32.5,
      "grad_norm_var": 6.1228515625,
      "learning_rate": 0.0001,
      "loss": 7.4938,
      "loss/crossentropy": 2.2000851720571517,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20301534831523896,
      "step": 9660
    },
    {
      "epoch": 0.24175,
      "grad_norm": 31.0,
      "grad_norm_var": 3.59765625,
      "learning_rate": 0.0001,
      "loss": 7.5133,
      "loss/crossentropy": 2.192278115451336,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19489070847630502,
      "step": 9670
    },
    {
      "epoch": 0.242,
      "grad_norm": 28.875,
      "grad_norm_var": 6.8978515625,
      "learning_rate": 0.0001,
      "loss": 7.4247,
      "loss/crossentropy": 2.194097451120615,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20746590523049235,
      "step": 9680
    },
    {
      "epoch": 0.24225,
      "grad_norm": 29.75,
      "grad_norm_var": 4.0587890625,
      "learning_rate": 0.0001,
      "loss": 7.3413,
      "loss/crossentropy": 2.1902857303619383,
      "loss/hidden": 3.187109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17439354099333287,
      "step": 9690
    },
    {
      "epoch": 0.2425,
      "grad_norm": 41.5,
      "grad_norm_var": 3.952541960104418e+18,
      "learning_rate": 0.0001,
      "loss": 7.5429,
      "loss/crossentropy": 2.046304853260517,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18022917695343493,
      "step": 9700
    },
    {
      "epoch": 0.24275,
      "grad_norm": 30.375,
      "grad_norm_var": 3.952541960261809e+18,
      "learning_rate": 0.0001,
      "loss": 7.2729,
      "loss/crossentropy": 2.082195009291172,
      "loss/hidden": 3.5140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1983122780919075,
      "step": 9710
    },
    {
      "epoch": 0.243,
      "grad_norm": 28.375,
      "grad_norm_var": 2.6056640625,
      "learning_rate": 0.0001,
      "loss": 7.2271,
      "loss/crossentropy": 1.970278625190258,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19305091574788094,
      "step": 9720
    },
    {
      "epoch": 0.24325,
      "grad_norm": 30.375,
      "grad_norm_var": 2.0576524262052987e+18,
      "learning_rate": 0.0001,
      "loss": 7.3979,
      "loss/crossentropy": 2.1140229746699335,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19698369763791562,
      "step": 9730
    },
    {
      "epoch": 0.2435,
      "grad_norm": 29.0,
      "grad_norm_var": 2.057652425978177e+18,
      "learning_rate": 0.0001,
      "loss": 7.2491,
      "loss/crossentropy": 1.9204902969300748,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18290270324796437,
      "step": 9740
    },
    {
      "epoch": 0.24375,
      "grad_norm": 49.25,
      "grad_norm_var": 27.333072916666666,
      "learning_rate": 0.0001,
      "loss": 7.4168,
      "loss/crossentropy": 2.203968660533428,
      "loss/hidden": 3.358203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19210707377642394,
      "step": 9750
    },
    {
      "epoch": 0.244,
      "grad_norm": 27.75,
      "grad_norm_var": 28.5837890625,
      "learning_rate": 0.0001,
      "loss": 7.3667,
      "loss/crossentropy": 2.0645239472389223,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1849998442456126,
      "step": 9760
    },
    {
      "epoch": 0.24425,
      "grad_norm": 30.0,
      "grad_norm_var": 7.441666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3958,
      "loss/crossentropy": 2.0423281893134115,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19458430632948875,
      "step": 9770
    },
    {
      "epoch": 0.2445,
      "grad_norm": 28.25,
      "grad_norm_var": 25.725,
      "learning_rate": 0.0001,
      "loss": 7.371,
      "loss/crossentropy": 2.2162967801094053,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19894264116883278,
      "step": 9780
    },
    {
      "epoch": 0.24475,
      "grad_norm": 31.875,
      "grad_norm_var": 25.701822916666668,
      "learning_rate": 0.0001,
      "loss": 7.4136,
      "loss/crossentropy": 2.052091246843338,
      "loss/hidden": 3.331640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19768061954528093,
      "step": 9790
    },
    {
      "epoch": 0.245,
      "grad_norm": 29.375,
      "grad_norm_var": 2.4247395833333334,
      "learning_rate": 0.0001,
      "loss": 7.3752,
      "loss/crossentropy": 2.0378331199288366,
      "loss/hidden": 3.45703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1888222724199295,
      "step": 9800
    },
    {
      "epoch": 0.24525,
      "grad_norm": 29.75,
      "grad_norm_var": 17.073958333333334,
      "learning_rate": 0.0001,
      "loss": 7.389,
      "loss/crossentropy": 2.0574826121330263,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1817958688363433,
      "step": 9810
    },
    {
      "epoch": 0.2455,
      "grad_norm": 28.125,
      "grad_norm_var": 24.198372395833335,
      "learning_rate": 0.0001,
      "loss": 7.284,
      "loss/crossentropy": 2.068347904086113,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1905125178396702,
      "step": 9820
    },
    {
      "epoch": 0.24575,
      "grad_norm": 28.5,
      "grad_norm_var": 9.8041015625,
      "learning_rate": 0.0001,
      "loss": 7.3867,
      "loss/crossentropy": 2.0113710410892964,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1825895557180047,
      "step": 9830
    },
    {
      "epoch": 0.246,
      "grad_norm": 30.5,
      "grad_norm_var": 1.16015625,
      "learning_rate": 0.0001,
      "loss": 7.3496,
      "loss/crossentropy": 2.0473890252411366,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18930337531492114,
      "step": 9840
    },
    {
      "epoch": 0.24625,
      "grad_norm": 29.875,
      "grad_norm_var": 1.75,
      "learning_rate": 0.0001,
      "loss": 7.4305,
      "loss/crossentropy": 2.167546259611845,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18493952229619026,
      "step": 9850
    },
    {
      "epoch": 0.2465,
      "grad_norm": 28.375,
      "grad_norm_var": 2.7625,
      "learning_rate": 0.0001,
      "loss": 7.444,
      "loss/crossentropy": 2.1046394810080526,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18062518630176783,
      "step": 9860
    },
    {
      "epoch": 0.24675,
      "grad_norm": 28.75,
      "grad_norm_var": 7819.056705729166,
      "learning_rate": 0.0001,
      "loss": 7.4016,
      "loss/crossentropy": 2.0700221791863442,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.190355353243649,
      "step": 9870
    },
    {
      "epoch": 0.247,
      "grad_norm": 29.0,
      "grad_norm_var": 7843.72265625,
      "learning_rate": 0.0001,
      "loss": 7.2875,
      "loss/crossentropy": 2.057565826922655,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18850490506738424,
      "step": 9880
    },
    {
      "epoch": 0.24725,
      "grad_norm": 32.5,
      "grad_norm_var": 3.778059895833333,
      "learning_rate": 0.0001,
      "loss": 7.4329,
      "loss/crossentropy": 2.0141181223094464,
      "loss/hidden": 3.446875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19769613686949014,
      "step": 9890
    },
    {
      "epoch": 0.2475,
      "grad_norm": 30.375,
      "grad_norm_var": 6.312239583333334,
      "learning_rate": 0.0001,
      "loss": 7.336,
      "loss/crossentropy": 2.146551664918661,
      "loss/hidden": 3.34296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18510441221296786,
      "step": 9900
    },
    {
      "epoch": 0.24775,
      "grad_norm": 33.0,
      "grad_norm_var": 6.36640625,
      "learning_rate": 0.0001,
      "loss": 7.306,
      "loss/crossentropy": 2.2865438759326935,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19417236726731063,
      "step": 9910
    },
    {
      "epoch": 0.248,
      "grad_norm": 33.5,
      "grad_norm_var": 2.405989583333333,
      "learning_rate": 0.0001,
      "loss": 7.3855,
      "loss/crossentropy": 2.133404280245304,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19065556656569244,
      "step": 9920
    },
    {
      "epoch": 0.24825,
      "grad_norm": 34.25,
      "grad_norm_var": 3.1626528109486234e+18,
      "learning_rate": 0.0001,
      "loss": 7.4378,
      "loss/crossentropy": 2.0854051023721696,
      "loss/hidden": 3.459765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20351361632347106,
      "step": 9930
    },
    {
      "epoch": 0.2485,
      "grad_norm": 32.25,
      "grad_norm_var": 3.1626528099408717e+18,
      "learning_rate": 0.0001,
      "loss": 7.3826,
      "loss/crossentropy": 2.087091060727835,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19672764679417015,
      "step": 9940
    },
    {
      "epoch": 0.24875,
      "grad_norm": 31.875,
      "grad_norm_var": 3.729622395833333,
      "learning_rate": 0.0001,
      "loss": 7.4303,
      "loss/crossentropy": 1.9529958970844745,
      "loss/hidden": 3.461328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1957404987886548,
      "step": 9950
    },
    {
      "epoch": 0.249,
      "grad_norm": 27.0,
      "grad_norm_var": 3.515625,
      "learning_rate": 0.0001,
      "loss": 7.2963,
      "loss/crossentropy": 2.1291355013847353,
      "loss/hidden": 3.278125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1832482174038887,
      "step": 9960
    },
    {
      "epoch": 0.24925,
      "grad_norm": 31.5,
      "grad_norm_var": 2.403580729166667,
      "learning_rate": 0.0001,
      "loss": 7.3011,
      "loss/crossentropy": 2.073649263381958,
      "loss/hidden": 3.2859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17194083742797375,
      "step": 9970
    },
    {
      "epoch": 0.2495,
      "grad_norm": 30.5,
      "grad_norm_var": 3.7934895833333333,
      "learning_rate": 0.0001,
      "loss": 7.3253,
      "loss/crossentropy": 2.312130460143089,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19424791410565376,
      "step": 9980
    },
    {
      "epoch": 0.24975,
      "grad_norm": 31.875,
      "grad_norm_var": 6.4666015625,
      "learning_rate": 0.0001,
      "loss": 7.3395,
      "loss/crossentropy": 1.9769040577113628,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1865438589360565,
      "step": 9990
    },
    {
      "epoch": 0.25,
      "grad_norm": 37.0,
      "grad_norm_var": 5.488997395833334,
      "learning_rate": 0.0001,
      "loss": 7.3121,
      "loss/crossentropy": 2.0656296610832214,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19510807991027831,
      "step": 10000
    },
    {
      "epoch": 0.25025,
      "grad_norm": 30.5,
      "grad_norm_var": 6.201822916666667,
      "learning_rate": 0.0001,
      "loss": 7.3059,
      "loss/crossentropy": 2.1552528128027917,
      "loss/hidden": 3.338671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1871857862919569,
      "step": 10010
    },
    {
      "epoch": 0.2505,
      "grad_norm": 29.375,
      "grad_norm_var": 3.7697265625,
      "learning_rate": 0.0001,
      "loss": 7.402,
      "loss/crossentropy": 2.1086191907525063,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20956437531858682,
      "step": 10020
    },
    {
      "epoch": 0.25075,
      "grad_norm": 27.0,
      "grad_norm_var": 79.871875,
      "learning_rate": 0.0001,
      "loss": 7.2906,
      "loss/crossentropy": 2.24168366715312,
      "loss/hidden": 3.333984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1914551755413413,
      "step": 10030
    },
    {
      "epoch": 0.251,
      "grad_norm": 28.25,
      "grad_norm_var": 3.5348307291666665,
      "learning_rate": 0.0001,
      "loss": 7.3933,
      "loss/crossentropy": 1.9110309466719628,
      "loss/hidden": 3.52421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1938932467252016,
      "step": 10040
    },
    {
      "epoch": 0.25125,
      "grad_norm": 29.5,
      "grad_norm_var": 3.318489583333333,
      "learning_rate": 0.0001,
      "loss": 7.3066,
      "loss/crossentropy": 2.2151729106903075,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1898823155090213,
      "step": 10050
    },
    {
      "epoch": 0.2515,
      "grad_norm": 27.125,
      "grad_norm_var": 5.114583333333333,
      "learning_rate": 0.0001,
      "loss": 7.3306,
      "loss/crossentropy": 1.9325201705098152,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1914183372631669,
      "step": 10060
    },
    {
      "epoch": 0.25175,
      "grad_norm": 28.375,
      "grad_norm_var": 5.568489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4302,
      "loss/crossentropy": 2.053273378312588,
      "loss/hidden": 3.48828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.193053549900651,
      "step": 10070
    },
    {
      "epoch": 0.252,
      "grad_norm": 32.25,
      "grad_norm_var": 1.5947265625,
      "learning_rate": 0.0001,
      "loss": 7.3679,
      "loss/crossentropy": 2.1425619572401047,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19612068887799977,
      "step": 10080
    },
    {
      "epoch": 0.25225,
      "grad_norm": 28.125,
      "grad_norm_var": 1.2666666666666666,
      "learning_rate": 0.0001,
      "loss": 7.4696,
      "loss/crossentropy": 2.1615766674280166,
      "loss/hidden": 3.437890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21496526505798103,
      "step": 10090
    },
    {
      "epoch": 0.2525,
      "grad_norm": 31.375,
      "grad_norm_var": 87.72389322916666,
      "learning_rate": 0.0001,
      "loss": 7.3933,
      "loss/crossentropy": 2.093507520854473,
      "loss/hidden": 3.306640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18447848707437514,
      "step": 10100
    },
    {
      "epoch": 0.25275,
      "grad_norm": 29.125,
      "grad_norm_var": 87.9962890625,
      "learning_rate": 0.0001,
      "loss": 7.42,
      "loss/crossentropy": 2.2579648420214653,
      "loss/hidden": 3.415625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19929521884769202,
      "step": 10110
    },
    {
      "epoch": 0.253,
      "grad_norm": 31.0,
      "grad_norm_var": 3.0921223958333335,
      "learning_rate": 0.0001,
      "loss": 7.4462,
      "loss/crossentropy": 2.187203352898359,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18776683378964662,
      "step": 10120
    },
    {
      "epoch": 0.25325,
      "grad_norm": 35.0,
      "grad_norm_var": 37.43333333333333,
      "learning_rate": 0.0001,
      "loss": 7.493,
      "loss/crossentropy": 2.156028524041176,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20926487501710653,
      "step": 10130
    },
    {
      "epoch": 0.2535,
      "grad_norm": 30.625,
      "grad_norm_var": 36.87180989583333,
      "learning_rate": 0.0001,
      "loss": 7.4227,
      "loss/crossentropy": 1.9757653154432773,
      "loss/hidden": 3.383984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1927978384308517,
      "step": 10140
    },
    {
      "epoch": 0.25375,
      "grad_norm": 30.75,
      "grad_norm_var": 2.1582682291666666,
      "learning_rate": 0.0001,
      "loss": 7.3403,
      "loss/crossentropy": 2.1322492100298405,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1974259439855814,
      "step": 10150
    },
    {
      "epoch": 0.254,
      "grad_norm": 28.125,
      "grad_norm_var": 2.9259765625,
      "learning_rate": 0.0001,
      "loss": 7.3083,
      "loss/crossentropy": 2.0464363016188143,
      "loss/hidden": 3.27734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17546537732705475,
      "step": 10160
    },
    {
      "epoch": 0.25425,
      "grad_norm": 28.625,
      "grad_norm_var": 3.7514973958333333,
      "learning_rate": 0.0001,
      "loss": 7.3988,
      "loss/crossentropy": 2.059445019811392,
      "loss/hidden": 3.396484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19312301548197866,
      "step": 10170
    },
    {
      "epoch": 0.2545,
      "grad_norm": 29.5,
      "grad_norm_var": 3.36015625,
      "learning_rate": 0.0001,
      "loss": 7.3178,
      "loss/crossentropy": 2.0300123430788517,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18283235086128116,
      "step": 10180
    },
    {
      "epoch": 0.25475,
      "grad_norm": 31.625,
      "grad_norm_var": 2.7192057291666667,
      "learning_rate": 0.0001,
      "loss": 7.4117,
      "loss/crossentropy": 2.052568303793669,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17590052969753742,
      "step": 10190
    },
    {
      "epoch": 0.255,
      "grad_norm": 30.875,
      "grad_norm_var": 3.0660807291666665,
      "learning_rate": 0.0001,
      "loss": 7.4384,
      "loss/crossentropy": 2.1124994076788424,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20564852859824895,
      "step": 10200
    },
    {
      "epoch": 0.25525,
      "grad_norm": 32.25,
      "grad_norm_var": 1.8223307291666666,
      "learning_rate": 0.0001,
      "loss": 7.4006,
      "loss/crossentropy": 1.997454211115837,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.189224443025887,
      "step": 10210
    },
    {
      "epoch": 0.2555,
      "grad_norm": 30.75,
      "grad_norm_var": 12.555989583333334,
      "learning_rate": 0.0001,
      "loss": 7.5754,
      "loss/crossentropy": 2.095755438506603,
      "loss/hidden": 3.470703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1905398152768612,
      "step": 10220
    },
    {
      "epoch": 0.25575,
      "grad_norm": 30.25,
      "grad_norm_var": 2.0666015625,
      "learning_rate": 0.0001,
      "loss": 7.4499,
      "loss/crossentropy": 2.031408229470253,
      "loss/hidden": 3.379296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1826389298774302,
      "step": 10230
    },
    {
      "epoch": 0.256,
      "grad_norm": 30.875,
      "grad_norm_var": 1.1973307291666666,
      "learning_rate": 0.0001,
      "loss": 7.4736,
      "loss/crossentropy": 2.103855460882187,
      "loss/hidden": 3.479296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21460786666721104,
      "step": 10240
    },
    {
      "epoch": 0.25625,
      "grad_norm": 31.125,
      "grad_norm_var": 2.6556640625,
      "learning_rate": 0.0001,
      "loss": 7.5078,
      "loss/crossentropy": 2.0419384971261025,
      "loss/hidden": 3.44609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1855954358354211,
      "step": 10250
    },
    {
      "epoch": 0.2565,
      "grad_norm": 31.0,
      "grad_norm_var": 3.025,
      "learning_rate": 0.0001,
      "loss": 7.3116,
      "loss/crossentropy": 2.2164057731628417,
      "loss/hidden": 3.336328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18515868298709393,
      "step": 10260
    },
    {
      "epoch": 0.25675,
      "grad_norm": 30.75,
      "grad_norm_var": 2.653059895833333,
      "learning_rate": 0.0001,
      "loss": 7.2962,
      "loss/crossentropy": 2.0967153370380402,
      "loss/hidden": 3.5140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2144587781280279,
      "step": 10270
    },
    {
      "epoch": 0.257,
      "grad_norm": 27.625,
      "grad_norm_var": 2874.977018229167,
      "learning_rate": 0.0001,
      "loss": 7.3634,
      "loss/crossentropy": 2.197367396950722,
      "loss/hidden": 3.32421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1903674839064479,
      "step": 10280
    },
    {
      "epoch": 0.25725,
      "grad_norm": 30.25,
      "grad_norm_var": 2830.1666015625,
      "learning_rate": 0.0001,
      "loss": 7.4461,
      "loss/crossentropy": 1.9168385773897172,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18945675920695065,
      "step": 10290
    },
    {
      "epoch": 0.2575,
      "grad_norm": 30.625,
      "grad_norm_var": 10.3634765625,
      "learning_rate": 0.0001,
      "loss": 7.4383,
      "loss/crossentropy": 2.196033439040184,
      "loss/hidden": 3.389453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20106703452765942,
      "step": 10300
    },
    {
      "epoch": 0.25775,
      "grad_norm": 29.75,
      "grad_norm_var": 2.41015625,
      "learning_rate": 0.0001,
      "loss": 7.4,
      "loss/crossentropy": 1.9811425000429153,
      "loss/hidden": 3.2875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18086131382733583,
      "step": 10310
    },
    {
      "epoch": 0.258,
      "grad_norm": 28.625,
      "grad_norm_var": 1.3056640625,
      "learning_rate": 0.0001,
      "loss": 7.3814,
      "loss/crossentropy": 2.141839873790741,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19040859565138818,
      "step": 10320
    },
    {
      "epoch": 0.25825,
      "grad_norm": 29.75,
      "grad_norm_var": 0.9582682291666667,
      "learning_rate": 0.0001,
      "loss": 7.3988,
      "loss/crossentropy": 2.0412203505635262,
      "loss/hidden": 3.471484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19841136487666516,
      "step": 10330
    },
    {
      "epoch": 0.2585,
      "grad_norm": 31.0,
      "grad_norm_var": 1.94765625,
      "learning_rate": 0.0001,
      "loss": 7.464,
      "loss/crossentropy": 2.124895977973938,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19423805810511113,
      "step": 10340
    },
    {
      "epoch": 0.25875,
      "grad_norm": 27.125,
      "grad_norm_var": 20.603125,
      "learning_rate": 0.0001,
      "loss": 7.3612,
      "loss/crossentropy": 2.271273523569107,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19086832217872143,
      "step": 10350
    },
    {
      "epoch": 0.259,
      "grad_norm": 31.75,
      "grad_norm_var": 41.87265625,
      "learning_rate": 0.0001,
      "loss": 7.3645,
      "loss/crossentropy": 2.148430307209492,
      "loss/hidden": 3.33203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18722805399447678,
      "step": 10360
    },
    {
      "epoch": 0.25925,
      "grad_norm": 31.125,
      "grad_norm_var": 28.49140625,
      "learning_rate": 0.0001,
      "loss": 7.2841,
      "loss/crossentropy": 2.088160905241966,
      "loss/hidden": 3.43359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19452540278434755,
      "step": 10370
    },
    {
      "epoch": 0.2595,
      "grad_norm": 30.375,
      "grad_norm_var": 3.8275390625,
      "learning_rate": 0.0001,
      "loss": 7.3269,
      "loss/crossentropy": 2.037956405431032,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19043162725865842,
      "step": 10380
    },
    {
      "epoch": 0.25975,
      "grad_norm": 28.25,
      "grad_norm_var": 8.718489583333334,
      "learning_rate": 0.0001,
      "loss": 7.2968,
      "loss/crossentropy": 1.9823431193828582,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19301451742649078,
      "step": 10390
    },
    {
      "epoch": 0.26,
      "grad_norm": 31.625,
      "grad_norm_var": 8.5625,
      "learning_rate": 0.0001,
      "loss": 7.3314,
      "loss/crossentropy": 2.0179347068071367,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18727120459079744,
      "step": 10400
    },
    {
      "epoch": 0.26025,
      "grad_norm": 29.375,
      "grad_norm_var": 7.3625,
      "learning_rate": 0.0001,
      "loss": 7.3878,
      "loss/crossentropy": 2.0856245614588262,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18157497737556696,
      "step": 10410
    },
    {
      "epoch": 0.2605,
      "grad_norm": 33.0,
      "grad_norm_var": 3.0134765625,
      "learning_rate": 0.0001,
      "loss": 7.3355,
      "loss/crossentropy": 2.0798820555210114,
      "loss/hidden": 3.36875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1937016123905778,
      "step": 10420
    },
    {
      "epoch": 0.26075,
      "grad_norm": 30.625,
      "grad_norm_var": 2.5400390625,
      "learning_rate": 0.0001,
      "loss": 7.4527,
      "loss/crossentropy": 2.0538913279771807,
      "loss/hidden": 3.487890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19370690621435643,
      "step": 10430
    },
    {
      "epoch": 0.261,
      "grad_norm": 30.125,
      "grad_norm_var": 3.42890625,
      "learning_rate": 0.0001,
      "loss": 7.46,
      "loss/crossentropy": 2.181585241854191,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1953628245741129,
      "step": 10440
    },
    {
      "epoch": 0.26125,
      "grad_norm": 27.75,
      "grad_norm_var": 3.783072916666667,
      "learning_rate": 0.0001,
      "loss": 7.3934,
      "loss/crossentropy": 2.183496044576168,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20456707030534743,
      "step": 10450
    },
    {
      "epoch": 0.2615,
      "grad_norm": 30.125,
      "grad_norm_var": 1.4379557291666667,
      "learning_rate": 0.0001,
      "loss": 7.3211,
      "loss/crossentropy": 2.132909268140793,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18731370605528355,
      "step": 10460
    },
    {
      "epoch": 0.26175,
      "grad_norm": 31.25,
      "grad_norm_var": 2.5791666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3587,
      "loss/crossentropy": 2.0115842171013356,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18043918311595916,
      "step": 10470
    },
    {
      "epoch": 0.262,
      "grad_norm": 30.25,
      "grad_norm_var": 3.8268229166666665,
      "learning_rate": 0.0001,
      "loss": 7.2962,
      "loss/crossentropy": 2.1343379452824593,
      "loss/hidden": 3.305859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18047874867916108,
      "step": 10480
    },
    {
      "epoch": 0.26225,
      "grad_norm": 36.0,
      "grad_norm_var": 8.967708333333333,
      "learning_rate": 0.0001,
      "loss": 7.2903,
      "loss/crossentropy": 1.8975451827049254,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17858757209032775,
      "step": 10490
    },
    {
      "epoch": 0.2625,
      "grad_norm": 31.75,
      "grad_norm_var": 10.077083333333333,
      "learning_rate": 0.0001,
      "loss": 7.3074,
      "loss/crossentropy": 2.0936397939920424,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19368096999824047,
      "step": 10500
    },
    {
      "epoch": 0.26275,
      "grad_norm": 29.375,
      "grad_norm_var": 2.020768229166667,
      "learning_rate": 0.0001,
      "loss": 7.3675,
      "loss/crossentropy": 2.06072843298316,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17674694433808327,
      "step": 10510
    },
    {
      "epoch": 0.263,
      "grad_norm": 29.125,
      "grad_norm_var": 3.4686848958333334,
      "learning_rate": 0.0001,
      "loss": 7.3847,
      "loss/crossentropy": 2.057940775156021,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18738706540316344,
      "step": 10520
    },
    {
      "epoch": 0.26325,
      "grad_norm": 31.625,
      "grad_norm_var": 3.225,
      "learning_rate": 0.0001,
      "loss": 7.3449,
      "loss/crossentropy": 2.051366009563208,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18555977791547776,
      "step": 10530
    },
    {
      "epoch": 0.2635,
      "grad_norm": 29.75,
      "grad_norm_var": 32.99108072916667,
      "learning_rate": 0.0001,
      "loss": 7.3795,
      "loss/crossentropy": 2.061921717226505,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1927286960184574,
      "step": 10540
    },
    {
      "epoch": 0.26375,
      "grad_norm": 31.875,
      "grad_norm_var": 34.16451822916667,
      "learning_rate": 0.0001,
      "loss": 7.3305,
      "loss/crossentropy": 2.051614002883434,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20516632981598376,
      "step": 10550
    },
    {
      "epoch": 0.264,
      "grad_norm": 30.25,
      "grad_norm_var": 2.2802083333333334,
      "learning_rate": 0.0001,
      "loss": 7.3251,
      "loss/crossentropy": 2.208609312772751,
      "loss/hidden": 3.302734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1906878400593996,
      "step": 10560
    },
    {
      "epoch": 0.26425,
      "grad_norm": 28.0,
      "grad_norm_var": 2.8785807291666665,
      "learning_rate": 0.0001,
      "loss": 7.4584,
      "loss/crossentropy": 2.2136432066559792,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1983118023723364,
      "step": 10570
    },
    {
      "epoch": 0.2645,
      "grad_norm": 30.5,
      "grad_norm_var": 3.562239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3017,
      "loss/crossentropy": 2.2243916779756545,
      "loss/hidden": 3.276953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19225324541330338,
      "step": 10580
    },
    {
      "epoch": 0.26475,
      "grad_norm": 31.75,
      "grad_norm_var": 2.47265625,
      "learning_rate": 0.0001,
      "loss": 7.2623,
      "loss/crossentropy": 2.085415804386139,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1769318761304021,
      "step": 10590
    },
    {
      "epoch": 0.265,
      "grad_norm": 30.5,
      "grad_norm_var": 1.8895182291666666,
      "learning_rate": 0.0001,
      "loss": 7.4543,
      "loss/crossentropy": 2.0929081469774244,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21267817355692387,
      "step": 10600
    },
    {
      "epoch": 0.26525,
      "grad_norm": 28.75,
      "grad_norm_var": 2.505989583333333,
      "learning_rate": 0.0001,
      "loss": 7.2136,
      "loss/crossentropy": 1.885066507011652,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17078990247100592,
      "step": 10610
    },
    {
      "epoch": 0.2655,
      "grad_norm": 27.25,
      "grad_norm_var": 2.18515625,
      "learning_rate": 0.0001,
      "loss": 7.258,
      "loss/crossentropy": 1.9193078093230724,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18096741195768118,
      "step": 10620
    },
    {
      "epoch": 0.26575,
      "grad_norm": 29.5,
      "grad_norm_var": 2.499739583333333,
      "learning_rate": 0.0001,
      "loss": 7.256,
      "loss/crossentropy": 1.9643688909709454,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16975617725402117,
      "step": 10630
    },
    {
      "epoch": 0.266,
      "grad_norm": 41.25,
      "grad_norm_var": 2.3308942592018744e+18,
      "learning_rate": 0.0001,
      "loss": 7.3436,
      "loss/crossentropy": 1.8120718583464623,
      "loss/hidden": 3.403515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17283489797264337,
      "step": 10640
    },
    {
      "epoch": 0.26625,
      "grad_norm": 30.5,
      "grad_norm_var": 2.3308942571726003e+18,
      "learning_rate": 0.0001,
      "loss": 7.398,
      "loss/crossentropy": 2.2462258487939835,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20116339288651944,
      "step": 10650
    },
    {
      "epoch": 0.2665,
      "grad_norm": 32.0,
      "grad_norm_var": 21.258268229166667,
      "learning_rate": 0.0001,
      "loss": 7.2994,
      "loss/crossentropy": 2.044428373128176,
      "loss/hidden": 3.42265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1989501324482262,
      "step": 10660
    },
    {
      "epoch": 0.26675,
      "grad_norm": 29.875,
      "grad_norm_var": 1.8681640625,
      "learning_rate": 0.0001,
      "loss": 7.3881,
      "loss/crossentropy": 1.9573393151164056,
      "loss/hidden": 3.27421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17021252401173115,
      "step": 10670
    },
    {
      "epoch": 0.267,
      "grad_norm": 33.0,
      "grad_norm_var": 2.6197265625,
      "learning_rate": 0.0001,
      "loss": 7.449,
      "loss/crossentropy": 2.1520077705383303,
      "loss/hidden": 3.498046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20022546350955964,
      "step": 10680
    },
    {
      "epoch": 0.26725,
      "grad_norm": 30.25,
      "grad_norm_var": 69.97473958333333,
      "learning_rate": 0.0001,
      "loss": 7.4428,
      "loss/crossentropy": 2.1147203534841537,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18886133264750243,
      "step": 10690
    },
    {
      "epoch": 0.2675,
      "grad_norm": 33.25,
      "grad_norm_var": 27.978125,
      "learning_rate": 0.0001,
      "loss": 7.3953,
      "loss/crossentropy": 2.1656775265932082,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18777367267757655,
      "step": 10700
    },
    {
      "epoch": 0.26775,
      "grad_norm": 30.75,
      "grad_norm_var": 39.307291666666664,
      "learning_rate": 0.0001,
      "loss": 7.2137,
      "loss/crossentropy": 2.2314133137464522,
      "loss/hidden": 3.280859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18172509353607894,
      "step": 10710
    },
    {
      "epoch": 0.268,
      "grad_norm": 30.5,
      "grad_norm_var": 43.19055989583333,
      "learning_rate": 0.0001,
      "loss": 7.3973,
      "loss/crossentropy": 2.0373408157378434,
      "loss/hidden": 3.500390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19881470818072558,
      "step": 10720
    },
    {
      "epoch": 0.26825,
      "grad_norm": 28.125,
      "grad_norm_var": 7.35625,
      "learning_rate": 0.0001,
      "loss": 7.3064,
      "loss/crossentropy": 2.065886814892292,
      "loss/hidden": 3.328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19001882169395684,
      "step": 10730
    },
    {
      "epoch": 0.2685,
      "grad_norm": 30.0,
      "grad_norm_var": 2.437239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4337,
      "loss/crossentropy": 2.1057795181870462,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19459401965141296,
      "step": 10740
    },
    {
      "epoch": 0.26875,
      "grad_norm": 29.75,
      "grad_norm_var": 3.24765625,
      "learning_rate": 0.0001,
      "loss": 7.4157,
      "loss/crossentropy": 1.7320681288838387,
      "loss/hidden": 3.533984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20864265877753496,
      "step": 10750
    },
    {
      "epoch": 0.269,
      "grad_norm": 31.0,
      "grad_norm_var": 2.9098307291666665,
      "learning_rate": 0.0001,
      "loss": 7.3267,
      "loss/crossentropy": 2.1352997794747353,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19193184426985682,
      "step": 10760
    },
    {
      "epoch": 0.26925,
      "grad_norm": 32.25,
      "grad_norm_var": 3.0444333193999944e+18,
      "learning_rate": 0.0001,
      "loss": 7.3861,
      "loss/crossentropy": 2.070178285241127,
      "loss/hidden": 3.4390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19077087007462978,
      "step": 10770
    },
    {
      "epoch": 0.2695,
      "grad_norm": 28.5,
      "grad_norm_var": 23.91640625,
      "learning_rate": 0.0001,
      "loss": 7.4494,
      "loss/crossentropy": 2.147341425716877,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18359180726110935,
      "step": 10780
    },
    {
      "epoch": 0.26975,
      "grad_norm": 30.875,
      "grad_norm_var": 1.5843098958333333,
      "learning_rate": 0.0001,
      "loss": 7.3566,
      "loss/crossentropy": 2.0286200530827045,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1995188482105732,
      "step": 10790
    },
    {
      "epoch": 0.27,
      "grad_norm": 30.625,
      "grad_norm_var": 1.8247395833333333,
      "learning_rate": 0.0001,
      "loss": 7.3752,
      "loss/crossentropy": 2.1403896272182465,
      "loss/hidden": 3.408203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19066570214927198,
      "step": 10800
    },
    {
      "epoch": 0.27025,
      "grad_norm": 32.25,
      "grad_norm_var": 3.2462890625,
      "learning_rate": 0.0001,
      "loss": 7.3297,
      "loss/crossentropy": 2.109960842132568,
      "loss/hidden": 3.35234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19408894181251526,
      "step": 10810
    },
    {
      "epoch": 0.2705,
      "grad_norm": 31.0,
      "grad_norm_var": 2.125,
      "learning_rate": 0.0001,
      "loss": 7.4115,
      "loss/crossentropy": 2.170359855145216,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1864105872809887,
      "step": 10820
    },
    {
      "epoch": 0.27075,
      "grad_norm": 30.125,
      "grad_norm_var": 2.5025390625,
      "learning_rate": 0.0001,
      "loss": 7.3568,
      "loss/crossentropy": 2.0696767389774324,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18454026989638805,
      "step": 10830
    },
    {
      "epoch": 0.271,
      "grad_norm": 29.625,
      "grad_norm_var": 3.187239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3986,
      "loss/crossentropy": 2.1671934336423875,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19040703494101763,
      "step": 10840
    },
    {
      "epoch": 0.27125,
      "grad_norm": 29.25,
      "grad_norm_var": 3.470833333333333,
      "learning_rate": 0.0001,
      "loss": 7.3105,
      "loss/crossentropy": 1.9061052799224854,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1798303933814168,
      "step": 10850
    },
    {
      "epoch": 0.2715,
      "grad_norm": 33.0,
      "grad_norm_var": 3.2087890625,
      "learning_rate": 0.0001,
      "loss": 7.3692,
      "loss/crossentropy": 2.1218462653458117,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20523111913353204,
      "step": 10860
    },
    {
      "epoch": 0.27175,
      "grad_norm": 29.375,
      "grad_norm_var": 3.544205729166667,
      "learning_rate": 0.0001,
      "loss": 7.3206,
      "loss/crossentropy": 2.006873355805874,
      "loss/hidden": 3.4734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2144127245992422,
      "step": 10870
    },
    {
      "epoch": 0.272,
      "grad_norm": 29.875,
      "grad_norm_var": 1.5832967245463552e+18,
      "learning_rate": 0.0001,
      "loss": 7.4255,
      "loss/crossentropy": 2.1331238821148872,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19302980024367572,
      "step": 10880
    },
    {
      "epoch": 0.27225,
      "grad_norm": 31.875,
      "grad_norm_var": 1.5832967244414976e+18,
      "learning_rate": 0.0001,
      "loss": 7.3477,
      "loss/crossentropy": 2.041455736756325,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.230126572214067,
      "step": 10890
    },
    {
      "epoch": 0.2725,
      "grad_norm": 32.25,
      "grad_norm_var": 16.21015625,
      "learning_rate": 0.0001,
      "loss": 7.3668,
      "loss/crossentropy": 2.11685880869627,
      "loss/hidden": 3.32890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18238812778145075,
      "step": 10900
    },
    {
      "epoch": 0.27275,
      "grad_norm": 29.625,
      "grad_norm_var": 15.566666666666666,
      "learning_rate": 0.0001,
      "loss": 7.349,
      "loss/crossentropy": 2.1108590021729468,
      "loss/hidden": 3.448828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2031965653412044,
      "step": 10910
    },
    {
      "epoch": 0.273,
      "grad_norm": 30.5,
      "grad_norm_var": 6.133072916666666,
      "learning_rate": 0.0001,
      "loss": 7.3685,
      "loss/crossentropy": 1.9768708415329457,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.1786771345883608,
      "step": 10920
    },
    {
      "epoch": 0.27325,
      "grad_norm": 28.75,
      "grad_norm_var": 6.448372395833333,
      "learning_rate": 0.0001,
      "loss": 7.3505,
      "loss/crossentropy": 1.9710132874548436,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18313440950587392,
      "step": 10930
    },
    {
      "epoch": 0.2735,
      "grad_norm": 29.5,
      "grad_norm_var": 2.63515625,
      "learning_rate": 0.0001,
      "loss": 7.4147,
      "loss/crossentropy": 2.1898112446069717,
      "loss/hidden": 3.3546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1952790988609195,
      "step": 10940
    },
    {
      "epoch": 0.27375,
      "grad_norm": 31.25,
      "grad_norm_var": 1.8978515625,
      "learning_rate": 0.0001,
      "loss": 7.3679,
      "loss/crossentropy": 2.135420022904873,
      "loss/hidden": 3.43203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21568923965096473,
      "step": 10950
    },
    {
      "epoch": 0.274,
      "grad_norm": 31.625,
      "grad_norm_var": 3.5184895833333334,
      "learning_rate": 0.0001,
      "loss": 7.2611,
      "loss/crossentropy": 1.9667419284582137,
      "loss/hidden": 3.42265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1834744794294238,
      "step": 10960
    },
    {
      "epoch": 0.27425,
      "grad_norm": 30.625,
      "grad_norm_var": 1.3561848958333333,
      "learning_rate": 0.0001,
      "loss": 7.4273,
      "loss/crossentropy": 2.074997512996197,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18813611660152674,
      "step": 10970
    },
    {
      "epoch": 0.2745,
      "grad_norm": 31.75,
      "grad_norm_var": 2.325455729166667,
      "learning_rate": 0.0001,
      "loss": 7.3688,
      "loss/crossentropy": 2.234239089488983,
      "loss/hidden": 3.254296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18571207812055945,
      "step": 10980
    },
    {
      "epoch": 0.27475,
      "grad_norm": 33.25,
      "grad_norm_var": 5.548958333333333,
      "learning_rate": 0.0001,
      "loss": 7.4528,
      "loss/crossentropy": 2.1355573579669,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20484505780041218,
      "step": 10990
    },
    {
      "epoch": 0.275,
      "grad_norm": 26.625,
      "grad_norm_var": 5.2759765625,
      "learning_rate": 0.0001,
      "loss": 7.4261,
      "loss/crossentropy": 2.2509318992495535,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19250140003859997,
      "step": 11000
    },
    {
      "epoch": 0.27525,
      "grad_norm": 31.0,
      "grad_norm_var": 5.387239583333334,
      "learning_rate": 0.0001,
      "loss": 7.3389,
      "loss/crossentropy": 2.114951176941395,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.198347245156765,
      "step": 11010
    },
    {
      "epoch": 0.2755,
      "grad_norm": 30.375,
      "grad_norm_var": 1.5468098958333334,
      "learning_rate": 0.0001,
      "loss": 7.3199,
      "loss/crossentropy": 2.110514160990715,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1900305664166808,
      "step": 11020
    },
    {
      "epoch": 0.27575,
      "grad_norm": 27.25,
      "grad_norm_var": 2.003059895833333,
      "learning_rate": 0.0001,
      "loss": 7.3304,
      "loss/crossentropy": 2.201597589254379,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20035982932895421,
      "step": 11030
    },
    {
      "epoch": 0.276,
      "grad_norm": 32.25,
      "grad_norm_var": 3.0139973958333335,
      "learning_rate": 0.0001,
      "loss": 7.4256,
      "loss/crossentropy": 2.1428385630249975,
      "loss/hidden": 3.287890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18116684164851904,
      "step": 11040
    },
    {
      "epoch": 0.27625,
      "grad_norm": 31.25,
      "grad_norm_var": 2.4671223958333335,
      "learning_rate": 0.0001,
      "loss": 7.2733,
      "loss/crossentropy": 2.043394061923027,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17936107851564884,
      "step": 11050
    },
    {
      "epoch": 0.2765,
      "grad_norm": 32.75,
      "grad_norm_var": 1.3830729166666667,
      "learning_rate": 0.0001,
      "loss": 7.4904,
      "loss/crossentropy": 2.1975908786058427,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21010352578014135,
      "step": 11060
    },
    {
      "epoch": 0.27675,
      "grad_norm": 30.0,
      "grad_norm_var": 4.671875,
      "learning_rate": 0.0001,
      "loss": 7.3718,
      "loss/crossentropy": 2.1417069509625435,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17585361283272505,
      "step": 11070
    },
    {
      "epoch": 0.277,
      "grad_norm": 29.375,
      "grad_norm_var": 4.309375,
      "learning_rate": 0.0001,
      "loss": 7.2855,
      "loss/crossentropy": 2.041763362288475,
      "loss/hidden": 3.221875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17130352668464183,
      "step": 11080
    },
    {
      "epoch": 0.27725,
      "grad_norm": 28.375,
      "grad_norm_var": 11.15390625,
      "learning_rate": 0.0001,
      "loss": 7.3131,
      "loss/crossentropy": 2.1522015750408174,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18296321779489516,
      "step": 11090
    },
    {
      "epoch": 0.2775,
      "grad_norm": 30.375,
      "grad_norm_var": 11.266080729166667,
      "learning_rate": 0.0001,
      "loss": 7.3499,
      "loss/crossentropy": 2.1265663146972655,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18307223506271839,
      "step": 11100
    },
    {
      "epoch": 0.27775,
      "grad_norm": 30.75,
      "grad_norm_var": 1.6994140625,
      "learning_rate": 0.0001,
      "loss": 7.3764,
      "loss/crossentropy": 2.126886320114136,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1963003095239401,
      "step": 11110
    },
    {
      "epoch": 0.278,
      "grad_norm": 29.75,
      "grad_norm_var": 1.9780598958333333,
      "learning_rate": 0.0001,
      "loss": 7.3686,
      "loss/crossentropy": 2.209390402585268,
      "loss/hidden": 3.284375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18762579131871462,
      "step": 11120
    },
    {
      "epoch": 0.27825,
      "grad_norm": 29.875,
      "grad_norm_var": 2.3806640625,
      "learning_rate": 0.0001,
      "loss": 7.4086,
      "loss/crossentropy": 2.0636716064065697,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18702280572615565,
      "step": 11130
    },
    {
      "epoch": 0.2785,
      "grad_norm": 30.0,
      "grad_norm_var": 11.084375,
      "learning_rate": 0.0001,
      "loss": 7.3611,
      "loss/crossentropy": 1.9728175386786462,
      "loss/hidden": 3.371875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18249935954809188,
      "step": 11140
    },
    {
      "epoch": 0.27875,
      "grad_norm": 30.75,
      "grad_norm_var": 13.183333333333334,
      "learning_rate": 0.0001,
      "loss": 7.2901,
      "loss/crossentropy": 1.9563469380140304,
      "loss/hidden": 3.32421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1720572842285037,
      "step": 11150
    },
    {
      "epoch": 0.279,
      "grad_norm": 40.75,
      "grad_norm_var": 11.42890625,
      "learning_rate": 0.0001,
      "loss": 7.2633,
      "loss/crossentropy": 2.1441919445991515,
      "loss/hidden": 3.30703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1793568328022957,
      "step": 11160
    },
    {
      "epoch": 0.27925,
      "grad_norm": 31.125,
      "grad_norm_var": 9.245833333333334,
      "learning_rate": 0.0001,
      "loss": 7.3546,
      "loss/crossentropy": 2.0893258482217787,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18886512089520693,
      "step": 11170
    },
    {
      "epoch": 0.2795,
      "grad_norm": 29.25,
      "grad_norm_var": 15.84140625,
      "learning_rate": 0.0001,
      "loss": 7.3265,
      "loss/crossentropy": 2.028592649102211,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1916877321898937,
      "step": 11180
    },
    {
      "epoch": 0.27975,
      "grad_norm": 29.0,
      "grad_norm_var": 5.451822916666667,
      "learning_rate": 0.0001,
      "loss": 7.4362,
      "loss/crossentropy": 2.1735318168997764,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.18261966463178397,
      "step": 11190
    },
    {
      "epoch": 0.28,
      "grad_norm": 30.75,
      "grad_norm_var": 3.405143229166667,
      "learning_rate": 0.0001,
      "loss": 7.3648,
      "loss/crossentropy": 2.129372274875641,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19846085608005523,
      "step": 11200
    },
    {
      "epoch": 0.28025,
      "grad_norm": 30.125,
      "grad_norm_var": 3.4462890625,
      "learning_rate": 0.0001,
      "loss": 7.4302,
      "loss/crossentropy": 2.1702652648091316,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20027328319847584,
      "step": 11210
    },
    {
      "epoch": 0.2805,
      "grad_norm": 30.625,
      "grad_norm_var": 2.58515625,
      "learning_rate": 0.0001,
      "loss": 7.3575,
      "loss/crossentropy": 2.134139196574688,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1903523735702038,
      "step": 11220
    },
    {
      "epoch": 0.28075,
      "grad_norm": 32.25,
      "grad_norm_var": 3.1093098958333334,
      "learning_rate": 0.0001,
      "loss": 7.43,
      "loss/crossentropy": 2.026133489608765,
      "loss/hidden": 3.45390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.201400394923985,
      "step": 11230
    },
    {
      "epoch": 0.281,
      "grad_norm": 31.25,
      "grad_norm_var": 12.038997395833333,
      "learning_rate": 0.0001,
      "loss": 7.4162,
      "loss/crossentropy": 2.1442901253700257,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20070651173591614,
      "step": 11240
    },
    {
      "epoch": 0.28125,
      "grad_norm": 28.875,
      "grad_norm_var": 162.22962239583333,
      "learning_rate": 0.0001,
      "loss": 7.4058,
      "loss/crossentropy": 2.1855412632226945,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1976642705500126,
      "step": 11250
    },
    {
      "epoch": 0.2815,
      "grad_norm": 31.75,
      "grad_norm_var": 10.683072916666667,
      "learning_rate": 0.0001,
      "loss": 7.3535,
      "loss/crossentropy": 2.030480499565601,
      "loss/hidden": 3.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20611500162631274,
      "step": 11260
    },
    {
      "epoch": 0.28175,
      "grad_norm": 36.75,
      "grad_norm_var": 11.08515625,
      "learning_rate": 0.0001,
      "loss": 7.287,
      "loss/crossentropy": 2.01583767645061,
      "loss/hidden": 3.312109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18007306316867472,
      "step": 11270
    },
    {
      "epoch": 0.282,
      "grad_norm": 30.625,
      "grad_norm_var": 7.605989583333334,
      "learning_rate": 0.0001,
      "loss": 7.3012,
      "loss/crossentropy": 2.2034949243068693,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18539215214550495,
      "step": 11280
    },
    {
      "epoch": 0.28225,
      "grad_norm": 29.125,
      "grad_norm_var": 4.1791015625,
      "learning_rate": 0.0001,
      "loss": 7.3566,
      "loss/crossentropy": 2.0296434491872786,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18707384672015906,
      "step": 11290
    },
    {
      "epoch": 0.2825,
      "grad_norm": 30.375,
      "grad_norm_var": 7.01015625,
      "learning_rate": 0.0001,
      "loss": 7.2793,
      "loss/crossentropy": 2.0983324527740477,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17865268159657716,
      "step": 11300
    },
    {
      "epoch": 0.28275,
      "grad_norm": 34.0,
      "grad_norm_var": 7.029622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3716,
      "loss/crossentropy": 2.127777361869812,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20420280396938323,
      "step": 11310
    },
    {
      "epoch": 0.283,
      "grad_norm": 32.5,
      "grad_norm_var": 38.66712239583333,
      "learning_rate": 0.0001,
      "loss": 7.3326,
      "loss/crossentropy": 2.1566628091037274,
      "loss/hidden": 3.28828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.180232659727335,
      "step": 11320
    },
    {
      "epoch": 0.28325,
      "grad_norm": 30.5,
      "grad_norm_var": 44.13743489583333,
      "learning_rate": 0.0001,
      "loss": 7.3567,
      "loss/crossentropy": 2.091618612408638,
      "loss/hidden": 3.330078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18182430015876888,
      "step": 11330
    },
    {
      "epoch": 0.2835,
      "grad_norm": 29.625,
      "grad_norm_var": 4.887434895833334,
      "learning_rate": 0.0001,
      "loss": 7.3312,
      "loss/crossentropy": 2.1611782908439636,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1828035417944193,
      "step": 11340
    },
    {
      "epoch": 0.28375,
      "grad_norm": 34.25,
      "grad_norm_var": 6.423893229166667,
      "learning_rate": 0.0001,
      "loss": 7.1759,
      "loss/crossentropy": 1.9736236594617367,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18899882938712836,
      "step": 11350
    },
    {
      "epoch": 0.284,
      "grad_norm": 30.0,
      "grad_norm_var": 9.488541666666666,
      "learning_rate": 0.0001,
      "loss": 7.1771,
      "loss/crossentropy": 2.0265045419335364,
      "loss/hidden": 3.273828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1712389207445085,
      "step": 11360
    },
    {
      "epoch": 0.28425,
      "grad_norm": 33.75,
      "grad_norm_var": 12.092122395833334,
      "learning_rate": 0.0001,
      "loss": 7.361,
      "loss/crossentropy": 2.2598242223262788,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19073396287858485,
      "step": 11370
    },
    {
      "epoch": 0.2845,
      "grad_norm": 29.0,
      "grad_norm_var": 8.455989583333333,
      "learning_rate": 0.0001,
      "loss": 7.3234,
      "loss/crossentropy": 2.144850969314575,
      "loss/hidden": 3.321484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18078193911351265,
      "step": 11380
    },
    {
      "epoch": 0.28475,
      "grad_norm": 29.25,
      "grad_norm_var": 1.7212890625,
      "learning_rate": 0.0001,
      "loss": 7.3,
      "loss/crossentropy": 2.1646609872579576,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18855173885822296,
      "step": 11390
    },
    {
      "epoch": 0.285,
      "grad_norm": 31.875,
      "grad_norm_var": 2.5837890625,
      "learning_rate": 0.0001,
      "loss": 7.2976,
      "loss/crossentropy": 2.120090515911579,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17507072482258082,
      "step": 11400
    },
    {
      "epoch": 0.28525,
      "grad_norm": 29.375,
      "grad_norm_var": 2.0785807291666667,
      "learning_rate": 0.0001,
      "loss": 7.4018,
      "loss/crossentropy": 2.0800730645656587,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20872990731149912,
      "step": 11410
    },
    {
      "epoch": 0.2855,
      "grad_norm": 30.25,
      "grad_norm_var": 13.87890625,
      "learning_rate": 0.0001,
      "loss": 7.1549,
      "loss/crossentropy": 2.1735212251544,
      "loss/hidden": 3.25625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18148906547576188,
      "step": 11420
    },
    {
      "epoch": 0.28575,
      "grad_norm": 30.375,
      "grad_norm_var": 15.132747395833333,
      "learning_rate": 0.0001,
      "loss": 7.206,
      "loss/crossentropy": 1.8314741916954518,
      "loss/hidden": 3.272265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17273464631289243,
      "step": 11430
    },
    {
      "epoch": 0.286,
      "grad_norm": 28.875,
      "grad_norm_var": 2.2330729166666665,
      "learning_rate": 0.0001,
      "loss": 7.3502,
      "loss/crossentropy": 1.9636109337210654,
      "loss/hidden": 3.329296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16999556813389063,
      "step": 11440
    },
    {
      "epoch": 0.28625,
      "grad_norm": 30.5,
      "grad_norm_var": 1.8009765625,
      "learning_rate": 0.0001,
      "loss": 7.2893,
      "loss/crossentropy": 1.9146189287304878,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1754764079116285,
      "step": 11450
    },
    {
      "epoch": 0.2865,
      "grad_norm": 31.75,
      "grad_norm_var": 1.9197265625,
      "learning_rate": 0.0001,
      "loss": 7.3233,
      "loss/crossentropy": 2.0301330149173737,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1742406915873289,
      "step": 11460
    },
    {
      "epoch": 0.28675,
      "grad_norm": 34.25,
      "grad_norm_var": 3.8009765625,
      "learning_rate": 0.0001,
      "loss": 7.4231,
      "loss/crossentropy": 2.1827415734529496,
      "loss/hidden": 3.33515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18332260251045226,
      "step": 11470
    },
    {
      "epoch": 0.287,
      "grad_norm": 30.875,
      "grad_norm_var": 3.479622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3842,
      "loss/crossentropy": 1.9815428338944912,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17839936017990113,
      "step": 11480
    },
    {
      "epoch": 0.28725,
      "grad_norm": 30.125,
      "grad_norm_var": 12.9228515625,
      "learning_rate": 0.0001,
      "loss": 7.3888,
      "loss/crossentropy": 2.10543007850647,
      "loss/hidden": 3.29765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18028822876513004,
      "step": 11490
    },
    {
      "epoch": 0.2875,
      "grad_norm": 30.5,
      "grad_norm_var": 0.9973307291666667,
      "learning_rate": 0.0001,
      "loss": 7.2278,
      "loss/crossentropy": 1.9690312303602695,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19321996718645096,
      "step": 11500
    },
    {
      "epoch": 0.28775,
      "grad_norm": 29.5,
      "grad_norm_var": 7.462239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4485,
      "loss/crossentropy": 2.0997009545564653,
      "loss/hidden": 3.371875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1947010463103652,
      "step": 11510
    },
    {
      "epoch": 0.288,
      "grad_norm": 30.625,
      "grad_norm_var": 6.814583333333333,
      "learning_rate": 0.0001,
      "loss": 7.4014,
      "loss/crossentropy": 2.041436542570591,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17614111984148623,
      "step": 11520
    },
    {
      "epoch": 0.28825,
      "grad_norm": 29.625,
      "grad_norm_var": 1.7247395833333334,
      "learning_rate": 0.0001,
      "loss": 7.3231,
      "loss/crossentropy": 2.056125995516777,
      "loss/hidden": 3.475,
      "loss/jsd": 0.0,
      "loss/logits": 0.20333750136196613,
      "step": 11530
    },
    {
      "epoch": 0.2885,
      "grad_norm": 30.5,
      "grad_norm_var": 3.0014973958333333,
      "learning_rate": 0.0001,
      "loss": 7.2735,
      "loss/crossentropy": 1.9806349158287049,
      "loss/hidden": 3.306640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17432715664617718,
      "step": 11540
    },
    {
      "epoch": 0.28875,
      "grad_norm": 32.75,
      "grad_norm_var": 1.6973307291666666,
      "learning_rate": 0.0001,
      "loss": 7.3842,
      "loss/crossentropy": 2.0468244731426237,
      "loss/hidden": 3.506640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19939990881830455,
      "step": 11550
    },
    {
      "epoch": 0.289,
      "grad_norm": 31.625,
      "grad_norm_var": 3.546875,
      "learning_rate": 0.0001,
      "loss": 7.4212,
      "loss/crossentropy": 2.1170118719339373,
      "loss/hidden": 3.452734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1938545262441039,
      "step": 11560
    },
    {
      "epoch": 0.28925,
      "grad_norm": 29.0,
      "grad_norm_var": 6.987239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3068,
      "loss/crossentropy": 1.9378852248191833,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18525508288294076,
      "step": 11570
    },
    {
      "epoch": 0.2895,
      "grad_norm": 32.25,
      "grad_norm_var": 7.7650390625,
      "learning_rate": 0.0001,
      "loss": 7.3939,
      "loss/crossentropy": 2.094913274049759,
      "loss/hidden": 3.328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19857972972095012,
      "step": 11580
    },
    {
      "epoch": 0.28975,
      "grad_norm": 31.0,
      "grad_norm_var": 2.3934895833333334,
      "learning_rate": 0.0001,
      "loss": 7.3634,
      "loss/crossentropy": 2.112175312638283,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1923027027398348,
      "step": 11590
    },
    {
      "epoch": 0.29,
      "grad_norm": 29.75,
      "grad_norm_var": 2.8676432291666667,
      "learning_rate": 0.0001,
      "loss": 7.4788,
      "loss/crossentropy": 2.347763030230999,
      "loss/hidden": 3.379296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20385736282914876,
      "step": 11600
    },
    {
      "epoch": 0.29025,
      "grad_norm": 36.0,
      "grad_norm_var": 2.1060661210680481e+18,
      "learning_rate": 0.0001,
      "loss": 7.482,
      "loss/crossentropy": 2.1438632771372794,
      "loss/hidden": 3.375390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1947392811998725,
      "step": 11610
    },
    {
      "epoch": 0.2905,
      "grad_norm": 31.5,
      "grad_norm_var": 2.1060661214006216e+18,
      "learning_rate": 0.0001,
      "loss": 7.2853,
      "loss/crossentropy": 2.080791215598583,
      "loss/hidden": 3.2875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19497475158423186,
      "step": 11620
    },
    {
      "epoch": 0.29075,
      "grad_norm": 30.0,
      "grad_norm_var": 1.5931640625,
      "learning_rate": 0.0001,
      "loss": 7.2971,
      "loss/crossentropy": 2.16863374710083,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18860027231276036,
      "step": 11630
    },
    {
      "epoch": 0.291,
      "grad_norm": 31.625,
      "grad_norm_var": 1.5994140625,
      "learning_rate": 0.0001,
      "loss": 7.3835,
      "loss/crossentropy": 2.061967647075653,
      "loss/hidden": 3.366015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18620339650660753,
      "step": 11640
    },
    {
      "epoch": 0.29125,
      "grad_norm": 33.0,
      "grad_norm_var": 3.690207280948122e+18,
      "learning_rate": 0.0001,
      "loss": 7.3738,
      "loss/crossentropy": 2.0998649850487707,
      "loss/hidden": 3.279296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18567133340984582,
      "step": 11650
    },
    {
      "epoch": 0.2915,
      "grad_norm": 29.0,
      "grad_norm_var": 29.6228515625,
      "learning_rate": 0.0001,
      "loss": 7.2405,
      "loss/crossentropy": 2.1259921073913572,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18609032463282346,
      "step": 11660
    },
    {
      "epoch": 0.29175,
      "grad_norm": 31.0,
      "grad_norm_var": 14.421809895833333,
      "learning_rate": 0.0001,
      "loss": 7.4326,
      "loss/crossentropy": 2.066218351200223,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1920524787157774,
      "step": 11670
    },
    {
      "epoch": 0.292,
      "grad_norm": 27.875,
      "grad_norm_var": 49.305989583333336,
      "learning_rate": 0.0001,
      "loss": 7.2881,
      "loss/crossentropy": 2.07202163413167,
      "loss/hidden": 3.35,
      "loss/jsd": 0.0,
      "loss/logits": 0.18048244724050164,
      "step": 11680
    },
    {
      "epoch": 0.29225,
      "grad_norm": 31.875,
      "grad_norm_var": 4.995247395833333,
      "learning_rate": 0.0001,
      "loss": 7.2687,
      "loss/crossentropy": 2.045185898244381,
      "loss/hidden": 3.275,
      "loss/jsd": 0.0,
      "loss/logits": 0.17395208198577167,
      "step": 11690
    },
    {
      "epoch": 0.2925,
      "grad_norm": 30.0,
      "grad_norm_var": 5.696875,
      "learning_rate": 0.0001,
      "loss": 7.3259,
      "loss/crossentropy": 2.3408665537834166,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20136666856706142,
      "step": 11700
    },
    {
      "epoch": 0.29275,
      "grad_norm": 30.0,
      "grad_norm_var": 3.7421223958333334,
      "learning_rate": 0.0001,
      "loss": 7.2137,
      "loss/crossentropy": 1.9074330985546113,
      "loss/hidden": 3.349609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1704694928601384,
      "step": 11710
    },
    {
      "epoch": 0.293,
      "grad_norm": 31.5,
      "grad_norm_var": 5.803125,
      "learning_rate": 0.0001,
      "loss": 7.2817,
      "loss/crossentropy": 2.290114316344261,
      "loss/hidden": 3.36484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20061923637986184,
      "step": 11720
    },
    {
      "epoch": 0.29325,
      "grad_norm": 31.125,
      "grad_norm_var": 2.4936848958333333,
      "learning_rate": 0.0001,
      "loss": 7.4136,
      "loss/crossentropy": 2.2652835667133333,
      "loss/hidden": 3.29375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18269798178225755,
      "step": 11730
    },
    {
      "epoch": 0.2935,
      "grad_norm": 31.875,
      "grad_norm_var": 2.06640625,
      "learning_rate": 0.0001,
      "loss": 7.3219,
      "loss/crossentropy": 2.047122722864151,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18665554728358985,
      "step": 11740
    },
    {
      "epoch": 0.29375,
      "grad_norm": 29.875,
      "grad_norm_var": 11.484309895833333,
      "learning_rate": 0.0001,
      "loss": 7.4668,
      "loss/crossentropy": 2.0893666088581084,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18700389545410873,
      "step": 11750
    },
    {
      "epoch": 0.294,
      "grad_norm": 33.25,
      "grad_norm_var": 2.412239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3389,
      "loss/crossentropy": 2.0646077781915664,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1896628426387906,
      "step": 11760
    },
    {
      "epoch": 0.29425,
      "grad_norm": 36.75,
      "grad_norm_var": 15.843489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3537,
      "loss/crossentropy": 2.14951953291893,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19772466979920864,
      "step": 11770
    },
    {
      "epoch": 0.2945,
      "grad_norm": 34.0,
      "grad_norm_var": 15.808333333333334,
      "learning_rate": 0.0001,
      "loss": 7.2266,
      "loss/crossentropy": 1.9084580048918725,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17457483559846879,
      "step": 11780
    },
    {
      "epoch": 0.29475,
      "grad_norm": 30.5,
      "grad_norm_var": 17.133268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4198,
      "loss/crossentropy": 2.1059816129505635,
      "loss/hidden": 3.2609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19060559011995792,
      "step": 11790
    },
    {
      "epoch": 0.295,
      "grad_norm": 28.875,
      "grad_norm_var": 15.191666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3353,
      "loss/crossentropy": 2.0854270696640014,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18200992513448,
      "step": 11800
    },
    {
      "epoch": 0.29525,
      "grad_norm": 33.0,
      "grad_norm_var": 5.77890625,
      "learning_rate": 0.0001,
      "loss": 7.338,
      "loss/crossentropy": 2.0409404814243315,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18492084443569184,
      "step": 11810
    },
    {
      "epoch": 0.2955,
      "grad_norm": 28.25,
      "grad_norm_var": 4.9556640625,
      "learning_rate": 0.0001,
      "loss": 7.2394,
      "loss/crossentropy": 2.1242421194911003,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17592983674257995,
      "step": 11820
    },
    {
      "epoch": 0.29575,
      "grad_norm": 32.75,
      "grad_norm_var": 12.3994140625,
      "learning_rate": 0.0001,
      "loss": 7.3172,
      "loss/crossentropy": 2.135970714688301,
      "loss/hidden": 3.328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1905512258410454,
      "step": 11830
    },
    {
      "epoch": 0.296,
      "grad_norm": 31.125,
      "grad_norm_var": 13.479166666666666,
      "learning_rate": 0.0001,
      "loss": 7.3928,
      "loss/crossentropy": 2.0788447827100756,
      "loss/hidden": 3.35859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18583104945719242,
      "step": 11840
    },
    {
      "epoch": 0.29625,
      "grad_norm": 29.375,
      "grad_norm_var": 5.16640625,
      "learning_rate": 0.0001,
      "loss": 7.2575,
      "loss/crossentropy": 2.0609687596559523,
      "loss/hidden": 3.36875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1885878125205636,
      "step": 11850
    },
    {
      "epoch": 0.2965,
      "grad_norm": 33.5,
      "grad_norm_var": 4.534309895833333,
      "learning_rate": 0.0001,
      "loss": 7.2397,
      "loss/crossentropy": 2.113059785962105,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19279801957309245,
      "step": 11860
    },
    {
      "epoch": 0.29675,
      "grad_norm": 28.0,
      "grad_norm_var": 2.833268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4813,
      "loss/crossentropy": 2.2016214139759542,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.19246475584805012,
      "step": 11870
    },
    {
      "epoch": 0.297,
      "grad_norm": 35.5,
      "grad_norm_var": 5.829622395833334,
      "learning_rate": 0.0001,
      "loss": 7.4107,
      "loss/crossentropy": 2.1937885224819182,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20546972285956144,
      "step": 11880
    },
    {
      "epoch": 0.29725,
      "grad_norm": 32.25,
      "grad_norm_var": 5.101041666666666,
      "learning_rate": 0.0001,
      "loss": 7.4133,
      "loss/crossentropy": 2.112873890995979,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19512166157364846,
      "step": 11890
    },
    {
      "epoch": 0.2975,
      "grad_norm": 29.0,
      "grad_norm_var": 4.71875,
      "learning_rate": 0.0001,
      "loss": 7.3659,
      "loss/crossentropy": 2.068368895351887,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22229769406840205,
      "step": 11900
    },
    {
      "epoch": 0.29775,
      "grad_norm": 28.25,
      "grad_norm_var": 30.575,
      "learning_rate": 0.0001,
      "loss": 7.3459,
      "loss/crossentropy": 1.9633590430021286,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1919392876327038,
      "step": 11910
    },
    {
      "epoch": 0.298,
      "grad_norm": 31.5,
      "grad_norm_var": 25.169205729166666,
      "learning_rate": 0.0001,
      "loss": 7.2575,
      "loss/crossentropy": 2.0932082675397394,
      "loss/hidden": 3.3484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18578519038856028,
      "step": 11920
    },
    {
      "epoch": 0.29825,
      "grad_norm": 29.375,
      "grad_norm_var": 1.9197916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3554,
      "loss/crossentropy": 2.0443739868700503,
      "loss/hidden": 3.39140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18646586760878564,
      "step": 11930
    },
    {
      "epoch": 0.2985,
      "grad_norm": 29.125,
      "grad_norm_var": 1.1629557291666666,
      "learning_rate": 0.0001,
      "loss": 7.3341,
      "loss/crossentropy": 2.0645432278513907,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20268544293940066,
      "step": 11940
    },
    {
      "epoch": 0.29875,
      "grad_norm": 29.25,
      "grad_norm_var": 4.085416666666666,
      "learning_rate": 0.0001,
      "loss": 7.2596,
      "loss/crossentropy": 1.8821386724710465,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17970443107187747,
      "step": 11950
    },
    {
      "epoch": 0.299,
      "grad_norm": 29.75,
      "grad_norm_var": 2.4072265625,
      "learning_rate": 0.0001,
      "loss": 7.3139,
      "loss/crossentropy": 2.1848059430718423,
      "loss/hidden": 3.311328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19258468672633172,
      "step": 11960
    },
    {
      "epoch": 0.29925,
      "grad_norm": 28.625,
      "grad_norm_var": 1.49140625,
      "learning_rate": 0.0001,
      "loss": 7.3064,
      "loss/crossentropy": 1.9953081101179122,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17766128927469255,
      "step": 11970
    },
    {
      "epoch": 0.2995,
      "grad_norm": 30.625,
      "grad_norm_var": 3.527083333333333,
      "learning_rate": 0.0001,
      "loss": 7.3213,
      "loss/crossentropy": 1.9512401312589644,
      "loss/hidden": 3.257421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16765224877744914,
      "step": 11980
    },
    {
      "epoch": 0.29975,
      "grad_norm": 31.875,
      "grad_norm_var": 2.8322265625,
      "learning_rate": 0.0001,
      "loss": 7.3373,
      "loss/crossentropy": 2.134880567342043,
      "loss/hidden": 3.353515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19845206197351217,
      "step": 11990
    },
    {
      "epoch": 0.3,
      "grad_norm": 29.0,
      "grad_norm_var": 1.88515625,
      "learning_rate": 0.0001,
      "loss": 7.3355,
      "loss/crossentropy": 2.084690621495247,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.18881614562124013,
      "step": 12000
    },
    {
      "epoch": 0.30025,
      "grad_norm": 29.375,
      "grad_norm_var": 1.678125,
      "learning_rate": 0.0001,
      "loss": 7.2797,
      "loss/crossentropy": 2.0142040476202965,
      "loss/hidden": 3.277734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17630351725965737,
      "step": 12010
    },
    {
      "epoch": 0.3005,
      "grad_norm": 30.125,
      "grad_norm_var": 1.0702473958333334,
      "learning_rate": 0.0001,
      "loss": 7.3524,
      "loss/crossentropy": 2.0099969252943994,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17516886200755835,
      "step": 12020
    },
    {
      "epoch": 0.30075,
      "grad_norm": 30.125,
      "grad_norm_var": 2.1509765625,
      "learning_rate": 0.0001,
      "loss": 7.4081,
      "loss/crossentropy": 2.0550965458154677,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18450612868182362,
      "step": 12030
    },
    {
      "epoch": 0.301,
      "grad_norm": 29.375,
      "grad_norm_var": 2.088997395833333,
      "learning_rate": 0.0001,
      "loss": 7.3732,
      "loss/crossentropy": 2.015381334722042,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19866329655051232,
      "step": 12040
    },
    {
      "epoch": 0.30125,
      "grad_norm": 32.5,
      "grad_norm_var": 28.125455729166667,
      "learning_rate": 0.0001,
      "loss": 7.3689,
      "loss/crossentropy": 2.0619105845689774,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19213079251348972,
      "step": 12050
    },
    {
      "epoch": 0.3015,
      "grad_norm": 29.5,
      "grad_norm_var": 2.5869140625,
      "learning_rate": 0.0001,
      "loss": 7.3533,
      "loss/crossentropy": 2.059162912517786,
      "loss/hidden": 3.3625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1831585830077529,
      "step": 12060
    },
    {
      "epoch": 0.30175,
      "grad_norm": 30.375,
      "grad_norm_var": 6.6150390625,
      "learning_rate": 0.0001,
      "loss": 7.339,
      "loss/crossentropy": 2.0344214349985124,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18243852034211158,
      "step": 12070
    },
    {
      "epoch": 0.302,
      "grad_norm": 31.375,
      "grad_norm_var": 10.05390625,
      "learning_rate": 0.0001,
      "loss": 7.3457,
      "loss/crossentropy": 2.222717672586441,
      "loss/hidden": 3.28671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19160165078938007,
      "step": 12080
    },
    {
      "epoch": 0.30225,
      "grad_norm": 30.375,
      "grad_norm_var": 8.060416666666667,
      "learning_rate": 0.0001,
      "loss": 7.2928,
      "loss/crossentropy": 1.9902059100568295,
      "loss/hidden": 3.3921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1817794761620462,
      "step": 12090
    },
    {
      "epoch": 0.3025,
      "grad_norm": 29.0,
      "grad_norm_var": 4.048958333333333,
      "learning_rate": 0.0001,
      "loss": 7.403,
      "loss/crossentropy": 1.9990896947681904,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17872856128960848,
      "step": 12100
    },
    {
      "epoch": 0.30275,
      "grad_norm": 32.0,
      "grad_norm_var": 14.486393229166667,
      "learning_rate": 0.0001,
      "loss": 7.3774,
      "loss/crossentropy": 2.178261125087738,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19976076446473598,
      "step": 12110
    },
    {
      "epoch": 0.303,
      "grad_norm": 29.0,
      "grad_norm_var": 25.053059895833332,
      "learning_rate": 0.0001,
      "loss": 7.3852,
      "loss/crossentropy": 2.1738795921206475,
      "loss/hidden": 3.21171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1802441133186221,
      "step": 12120
    },
    {
      "epoch": 0.30325,
      "grad_norm": 32.5,
      "grad_norm_var": 27.863541666666666,
      "learning_rate": 0.0001,
      "loss": 7.3119,
      "loss/crossentropy": 2.017075891792774,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1819239752367139,
      "step": 12130
    },
    {
      "epoch": 0.3035,
      "grad_norm": 5704253440.0,
      "grad_norm_var": 2.033656683216328e+18,
      "learning_rate": 0.0001,
      "loss": 7.3187,
      "loss/crossentropy": 2.1378798320889474,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18621215373277664,
      "step": 12140
    },
    {
      "epoch": 0.30375,
      "grad_norm": 28.5,
      "grad_norm_var": 2.0336566834599473e+18,
      "learning_rate": 0.0001,
      "loss": 7.3375,
      "loss/crossentropy": 2.0977103441953657,
      "loss/hidden": 3.33046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18937066961079835,
      "step": 12150
    },
    {
      "epoch": 0.304,
      "grad_norm": 30.0,
      "grad_norm_var": 25.7119140625,
      "learning_rate": 0.0001,
      "loss": 7.2813,
      "loss/crossentropy": 2.0422945946455,
      "loss/hidden": 3.29453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17747708465903997,
      "step": 12160
    },
    {
      "epoch": 0.30425,
      "grad_norm": 29.375,
      "grad_norm_var": 1.8973307291666666,
      "learning_rate": 0.0001,
      "loss": 7.2982,
      "loss/crossentropy": 2.1438152998685838,
      "loss/hidden": 3.4,
      "loss/jsd": 0.0,
      "loss/logits": 0.19225314557552337,
      "step": 12170
    },
    {
      "epoch": 0.3045,
      "grad_norm": 32.5,
      "grad_norm_var": 6.688997395833334,
      "learning_rate": 0.0001,
      "loss": 7.2886,
      "loss/crossentropy": 2.033204630762339,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1741345826536417,
      "step": 12180
    },
    {
      "epoch": 0.30475,
      "grad_norm": 28.5,
      "grad_norm_var": 7.262434895833334,
      "learning_rate": 0.0001,
      "loss": 7.3812,
      "loss/crossentropy": 1.9149722829461098,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16993715493008493,
      "step": 12190
    },
    {
      "epoch": 0.305,
      "grad_norm": 30.625,
      "grad_norm_var": 2.315625,
      "learning_rate": 0.0001,
      "loss": 7.355,
      "loss/crossentropy": 2.200256870687008,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1957419654354453,
      "step": 12200
    },
    {
      "epoch": 0.30525,
      "grad_norm": 32.25,
      "grad_norm_var": 1.8291666666666666,
      "learning_rate": 0.0001,
      "loss": 7.2689,
      "loss/crossentropy": 1.9441036701202392,
      "loss/hidden": 3.450390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19158205185085536,
      "step": 12210
    },
    {
      "epoch": 0.3055,
      "grad_norm": 32.5,
      "grad_norm_var": 224.99993489583332,
      "learning_rate": 0.0001,
      "loss": 7.4205,
      "loss/crossentropy": 2.21477717012167,
      "loss/hidden": 3.353515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19972570687532426,
      "step": 12220
    },
    {
      "epoch": 0.30575,
      "grad_norm": 77.0,
      "grad_norm_var": 161.4994140625,
      "learning_rate": 0.0001,
      "loss": 7.3221,
      "loss/crossentropy": 2.243621972203255,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20230026841163634,
      "step": 12230
    },
    {
      "epoch": 0.306,
      "grad_norm": 29.125,
      "grad_norm_var": 136.2509765625,
      "learning_rate": 0.0001,
      "loss": 7.2544,
      "loss/crossentropy": 2.1267666652798654,
      "loss/hidden": 3.30078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18361538834869862,
      "step": 12240
    },
    {
      "epoch": 0.30625,
      "grad_norm": 43.0,
      "grad_norm_var": 15.1384765625,
      "learning_rate": 0.0001,
      "loss": 7.3457,
      "loss/crossentropy": 2.097319718450308,
      "loss/hidden": 3.380859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1941752176731825,
      "step": 12250
    },
    {
      "epoch": 0.3065,
      "grad_norm": 27.125,
      "grad_norm_var": 45.91875,
      "learning_rate": 0.0001,
      "loss": 7.1913,
      "loss/crossentropy": 2.024158325791359,
      "loss/hidden": 3.301171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18036661259829997,
      "step": 12260
    },
    {
      "epoch": 0.30675,
      "grad_norm": 32.5,
      "grad_norm_var": 34.94765625,
      "learning_rate": 0.0001,
      "loss": 7.2604,
      "loss/crossentropy": 2.05818811878562,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19272202141582967,
      "step": 12270
    },
    {
      "epoch": 0.307,
      "grad_norm": 31.625,
      "grad_norm_var": 1.7958333333333334,
      "learning_rate": 0.0001,
      "loss": 7.3478,
      "loss/crossentropy": 2.058871729671955,
      "loss/hidden": 3.316015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1805756026878953,
      "step": 12280
    },
    {
      "epoch": 0.30725,
      "grad_norm": 37.0,
      "grad_norm_var": 4.418489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3202,
      "loss/crossentropy": 1.9400940239429474,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19434008933603764,
      "step": 12290
    },
    {
      "epoch": 0.3075,
      "grad_norm": 30.5,
      "grad_norm_var": 30.1572265625,
      "learning_rate": 0.0001,
      "loss": 7.2939,
      "loss/crossentropy": 2.069047340750694,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20558738969266416,
      "step": 12300
    },
    {
      "epoch": 0.30775,
      "grad_norm": 33.0,
      "grad_norm_var": 11.9666015625,
      "learning_rate": 0.0001,
      "loss": 7.4109,
      "loss/crossentropy": 2.227694994211197,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18980047646909953,
      "step": 12310
    },
    {
      "epoch": 0.308,
      "grad_norm": 29.625,
      "grad_norm_var": 4.90390625,
      "learning_rate": 0.0001,
      "loss": 7.3686,
      "loss/crossentropy": 2.0309864141047003,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1816682495176792,
      "step": 12320
    },
    {
      "epoch": 0.30825,
      "grad_norm": 30.125,
      "grad_norm_var": 1.7249348958333333,
      "learning_rate": 0.0001,
      "loss": 7.4105,
      "loss/crossentropy": 2.2505668699741364,
      "loss/hidden": 3.307421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19684594608843325,
      "step": 12330
    },
    {
      "epoch": 0.3085,
      "grad_norm": 30.75,
      "grad_norm_var": 1.9535807291666667,
      "learning_rate": 0.0001,
      "loss": 7.3679,
      "loss/crossentropy": 2.0422501020133494,
      "loss/hidden": 3.223828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1723391016945243,
      "step": 12340
    },
    {
      "epoch": 0.30875,
      "grad_norm": 30.25,
      "grad_norm_var": 1.9186848958333333,
      "learning_rate": 0.0001,
      "loss": 7.3186,
      "loss/crossentropy": 2.190096604824066,
      "loss/hidden": 3.299609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1794669708237052,
      "step": 12350
    },
    {
      "epoch": 0.309,
      "grad_norm": 30.0,
      "grad_norm_var": 1.7926432291666667,
      "learning_rate": 0.0001,
      "loss": 7.2068,
      "loss/crossentropy": 2.011892383173108,
      "loss/hidden": 3.3046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17626657225191594,
      "step": 12360
    },
    {
      "epoch": 0.30925,
      "grad_norm": 32.25,
      "grad_norm_var": 3.3900390625,
      "learning_rate": 0.0001,
      "loss": 7.3963,
      "loss/crossentropy": 2.1367972552776338,
      "loss/hidden": 3.356640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19975323602557182,
      "step": 12370
    },
    {
      "epoch": 0.3095,
      "grad_norm": 30.875,
      "grad_norm_var": 2.9785807291666666,
      "learning_rate": 0.0001,
      "loss": 7.3586,
      "loss/crossentropy": 2.1065737903118134,
      "loss/hidden": 3.3296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1861647253856063,
      "step": 12380
    },
    {
      "epoch": 0.30975,
      "grad_norm": 34.0,
      "grad_norm_var": 49.94583333333333,
      "learning_rate": 0.0001,
      "loss": 7.4425,
      "loss/crossentropy": 2.047655486315489,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19645441174507142,
      "step": 12390
    },
    {
      "epoch": 0.31,
      "grad_norm": 30.75,
      "grad_norm_var": 55.521809895833336,
      "learning_rate": 0.0001,
      "loss": 7.3384,
      "loss/crossentropy": 2.1421022772789002,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21594423688948156,
      "step": 12400
    },
    {
      "epoch": 0.31025,
      "grad_norm": 28.625,
      "grad_norm_var": 3.630989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4915,
      "loss/crossentropy": 1.9852606005966664,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1970728723332286,
      "step": 12410
    },
    {
      "epoch": 0.3105,
      "grad_norm": 30.375,
      "grad_norm_var": 13.821875,
      "learning_rate": 0.0001,
      "loss": 7.316,
      "loss/crossentropy": 2.0221208080649378,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18263351675122977,
      "step": 12420
    },
    {
      "epoch": 0.31075,
      "grad_norm": 27.75,
      "grad_norm_var": 3.7905598958333333,
      "learning_rate": 0.0001,
      "loss": 7.4569,
      "loss/crossentropy": 2.2214868292212486,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20803299229592084,
      "step": 12430
    },
    {
      "epoch": 0.311,
      "grad_norm": 28.0,
      "grad_norm_var": 5.039322916666666,
      "learning_rate": 0.0001,
      "loss": 7.2246,
      "loss/crossentropy": 1.968916893005371,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1823066620156169,
      "step": 12440
    },
    {
      "epoch": 0.31125,
      "grad_norm": 32.25,
      "grad_norm_var": 3.011458333333333,
      "learning_rate": 0.0001,
      "loss": 7.2986,
      "loss/crossentropy": 2.102560856938362,
      "loss/hidden": 3.462109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19421595316380263,
      "step": 12450
    },
    {
      "epoch": 0.3115,
      "grad_norm": 31.625,
      "grad_norm_var": 1.7122395833333333,
      "learning_rate": 0.0001,
      "loss": 7.4102,
      "loss/crossentropy": 1.8904587842524052,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16765992902219296,
      "step": 12460
    },
    {
      "epoch": 0.31175,
      "grad_norm": 32.75,
      "grad_norm_var": 1.3614583333333334,
      "learning_rate": 0.0001,
      "loss": 7.317,
      "loss/crossentropy": 2.2456528916954994,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18445245549082756,
      "step": 12470
    },
    {
      "epoch": 0.312,
      "grad_norm": 32.5,
      "grad_norm_var": 2.9613932291666667,
      "learning_rate": 0.0001,
      "loss": 7.2527,
      "loss/crossentropy": 1.9254849448800087,
      "loss/hidden": 3.330859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17210696106776596,
      "step": 12480
    },
    {
      "epoch": 0.31225,
      "grad_norm": 29.0,
      "grad_norm_var": 2.503059895833333,
      "learning_rate": 0.0001,
      "loss": 7.3589,
      "loss/crossentropy": 1.963121373206377,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17580699175596237,
      "step": 12490
    },
    {
      "epoch": 0.3125,
      "grad_norm": 27.375,
      "grad_norm_var": 2.6510416666666665,
      "learning_rate": 0.0001,
      "loss": 7.3302,
      "loss/crossentropy": 2.0668979212641716,
      "loss/hidden": 3.33828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1756731817498803,
      "step": 12500
    },
    {
      "epoch": 0.31275,
      "grad_norm": 32.5,
      "grad_norm_var": 9.658072916666667,
      "learning_rate": 0.0001,
      "loss": 7.396,
      "loss/crossentropy": 2.0963846892118454,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19187606330960988,
      "step": 12510
    },
    {
      "epoch": 0.313,
      "grad_norm": 28.125,
      "grad_norm_var": 8.387239583333333,
      "learning_rate": 0.0001,
      "loss": 7.291,
      "loss/crossentropy": 2.1238081738352776,
      "loss/hidden": 3.319921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18899311302229763,
      "step": 12520
    },
    {
      "epoch": 0.31325,
      "grad_norm": 30.625,
      "grad_norm_var": 2.2712890625,
      "learning_rate": 0.0001,
      "loss": 7.2927,
      "loss/crossentropy": 1.8629853092133999,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18793845577165486,
      "step": 12530
    },
    {
      "epoch": 0.3135,
      "grad_norm": 31.25,
      "grad_norm_var": 29.926041666666666,
      "learning_rate": 0.0001,
      "loss": 7.334,
      "loss/crossentropy": 2.1922558069229128,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18994415253400804,
      "step": 12540
    },
    {
      "epoch": 0.31375,
      "grad_norm": 33.0,
      "grad_norm_var": 19.626822916666665,
      "learning_rate": 0.0001,
      "loss": 7.4246,
      "loss/crossentropy": 2.0747738771140574,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1956952316686511,
      "step": 12550
    },
    {
      "epoch": 0.314,
      "grad_norm": 31.0,
      "grad_norm_var": 5.9822265625,
      "learning_rate": 0.0001,
      "loss": 7.3921,
      "loss/crossentropy": 2.168559101223946,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19871533028781413,
      "step": 12560
    },
    {
      "epoch": 0.31425,
      "grad_norm": 29.875,
      "grad_norm_var": 23.478125,
      "learning_rate": 0.0001,
      "loss": 7.381,
      "loss/crossentropy": 2.1687012270092962,
      "loss/hidden": 3.2828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17819978184998037,
      "step": 12570
    },
    {
      "epoch": 0.3145,
      "grad_norm": 30.5,
      "grad_norm_var": 24.369791666666668,
      "learning_rate": 0.0001,
      "loss": 7.2273,
      "loss/crossentropy": 2.0223761543631555,
      "loss/hidden": 3.2953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17995049208402633,
      "step": 12580
    },
    {
      "epoch": 0.31475,
      "grad_norm": 31.125,
      "grad_norm_var": 3.8067057291666666,
      "learning_rate": 0.0001,
      "loss": 7.3672,
      "loss/crossentropy": 2.056264813989401,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19224987253546716,
      "step": 12590
    },
    {
      "epoch": 0.315,
      "grad_norm": 30.125,
      "grad_norm_var": 3.3018229166666666,
      "learning_rate": 0.0001,
      "loss": 7.2404,
      "loss/crossentropy": 2.0127123326063154,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17083216030150652,
      "step": 12600
    },
    {
      "epoch": 0.31525,
      "grad_norm": 31.25,
      "grad_norm_var": 2.94375,
      "learning_rate": 0.0001,
      "loss": 7.3918,
      "loss/crossentropy": 2.0961378000676634,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18820952698588372,
      "step": 12610
    },
    {
      "epoch": 0.3155,
      "grad_norm": 30.75,
      "grad_norm_var": 3.910416666666667,
      "learning_rate": 0.0001,
      "loss": 7.2715,
      "loss/crossentropy": 2.1737264052033423,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1752359176054597,
      "step": 12620
    },
    {
      "epoch": 0.31575,
      "grad_norm": 28.5,
      "grad_norm_var": 12.914322916666666,
      "learning_rate": 0.0001,
      "loss": 7.2866,
      "loss/crossentropy": 2.04836600497365,
      "loss/hidden": 3.308984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1814719710499048,
      "step": 12630
    },
    {
      "epoch": 0.316,
      "grad_norm": 32.5,
      "grad_norm_var": 5.895572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4359,
      "loss/crossentropy": 2.0633833378553392,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1844972724094987,
      "step": 12640
    },
    {
      "epoch": 0.31625,
      "grad_norm": 29.5,
      "grad_norm_var": 4.556705729166667,
      "learning_rate": 0.0001,
      "loss": 7.3404,
      "loss/crossentropy": 2.1031613536179066,
      "loss/hidden": 3.38671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1890261113177985,
      "step": 12650
    },
    {
      "epoch": 0.3165,
      "grad_norm": 33.25,
      "grad_norm_var": 7.43515625,
      "learning_rate": 0.0001,
      "loss": 7.3116,
      "loss/crossentropy": 2.0913070663809776,
      "loss/hidden": 3.48515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19625448603183032,
      "step": 12660
    },
    {
      "epoch": 0.31675,
      "grad_norm": 30.0,
      "grad_norm_var": 24.72265625,
      "learning_rate": 0.0001,
      "loss": 7.309,
      "loss/crossentropy": 2.1766112834215163,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19453570377081633,
      "step": 12670
    },
    {
      "epoch": 0.317,
      "grad_norm": 28.0,
      "grad_norm_var": 3.5518229166666666,
      "learning_rate": 0.0001,
      "loss": 7.2193,
      "loss/crossentropy": 2.1292213678359984,
      "loss/hidden": 3.312890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17923556556925177,
      "step": 12680
    },
    {
      "epoch": 0.31725,
      "grad_norm": 33.75,
      "grad_norm_var": 3.2041666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3568,
      "loss/crossentropy": 2.0966979548335076,
      "loss/hidden": 3.34375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1796438904479146,
      "step": 12690
    },
    {
      "epoch": 0.3175,
      "grad_norm": 30.125,
      "grad_norm_var": 21.9853515625,
      "learning_rate": 0.0001,
      "loss": 7.3281,
      "loss/crossentropy": 2.2050976656377315,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18696724623441696,
      "step": 12700
    },
    {
      "epoch": 0.31775,
      "grad_norm": 31.25,
      "grad_norm_var": 20.936458333333334,
      "learning_rate": 0.0001,
      "loss": 7.2583,
      "loss/crossentropy": 2.0703148849308493,
      "loss/hidden": 3.267578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17689838781952857,
      "step": 12710
    },
    {
      "epoch": 0.318,
      "grad_norm": 31.75,
      "grad_norm_var": 1.9395833333333334,
      "learning_rate": 0.0001,
      "loss": 7.3007,
      "loss/crossentropy": 1.9791635520756246,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17443582694977522,
      "step": 12720
    },
    {
      "epoch": 0.31825,
      "grad_norm": 30.25,
      "grad_norm_var": 2.5155598958333334,
      "learning_rate": 0.0001,
      "loss": 7.2557,
      "loss/crossentropy": 2.0753975957632065,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18456004988402128,
      "step": 12730
    },
    {
      "epoch": 0.3185,
      "grad_norm": 29.125,
      "grad_norm_var": 1.3518229166666667,
      "learning_rate": 0.0001,
      "loss": 7.2658,
      "loss/crossentropy": 2.0314756602048876,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18754468094557525,
      "step": 12740
    },
    {
      "epoch": 0.31875,
      "grad_norm": 29.25,
      "grad_norm_var": 1.6634765625,
      "learning_rate": 0.0001,
      "loss": 7.2545,
      "loss/crossentropy": 2.027373602986336,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18138464018702508,
      "step": 12750
    },
    {
      "epoch": 0.319,
      "grad_norm": 33.0,
      "grad_norm_var": 2.2837890625,
      "learning_rate": 0.0001,
      "loss": 7.2711,
      "loss/crossentropy": 2.0996690608561037,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19384829625487326,
      "step": 12760
    },
    {
      "epoch": 0.31925,
      "grad_norm": 29.5,
      "grad_norm_var": 1.7580729166666667,
      "learning_rate": 0.0001,
      "loss": 7.3448,
      "loss/crossentropy": 2.081255576014519,
      "loss/hidden": 3.48515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2054756512865424,
      "step": 12770
    },
    {
      "epoch": 0.3195,
      "grad_norm": 30.375,
      "grad_norm_var": 1.6337890625,
      "learning_rate": 0.0001,
      "loss": 7.3614,
      "loss/crossentropy": 2.109497997164726,
      "loss/hidden": 3.2984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1826931856572628,
      "step": 12780
    },
    {
      "epoch": 0.31975,
      "grad_norm": 29.125,
      "grad_norm_var": 2.9125,
      "learning_rate": 0.0001,
      "loss": 7.3569,
      "loss/crossentropy": 2.071355660259724,
      "loss/hidden": 3.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18764556515961886,
      "step": 12790
    },
    {
      "epoch": 0.32,
      "grad_norm": 31.0,
      "grad_norm_var": 5.68125,
      "learning_rate": 0.0001,
      "loss": 7.3399,
      "loss/crossentropy": 1.9978457435965538,
      "loss/hidden": 3.365625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17209083335474135,
      "step": 12800
    },
    {
      "epoch": 0.32025,
      "grad_norm": 28.75,
      "grad_norm_var": 20.754166666666666,
      "learning_rate": 0.0001,
      "loss": 7.3016,
      "loss/crossentropy": 1.984939170628786,
      "loss/hidden": 3.50625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19444225933402776,
      "step": 12810
    },
    {
      "epoch": 0.3205,
      "grad_norm": 28.875,
      "grad_norm_var": 3.3942057291666665,
      "learning_rate": 0.0001,
      "loss": 7.3066,
      "loss/crossentropy": 1.8981015786528588,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18696952695026994,
      "step": 12820
    },
    {
      "epoch": 0.32075,
      "grad_norm": 30.25,
      "grad_norm_var": 3.9218098958333334,
      "learning_rate": 0.0001,
      "loss": 7.3974,
      "loss/crossentropy": 2.1415200501680376,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20154954344034196,
      "step": 12830
    },
    {
      "epoch": 0.321,
      "grad_norm": 31.375,
      "grad_norm_var": 35.8400390625,
      "learning_rate": 0.0001,
      "loss": 7.3048,
      "loss/crossentropy": 2.1483262166380883,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19059547781944275,
      "step": 12840
    },
    {
      "epoch": 0.32125,
      "grad_norm": 31.0,
      "grad_norm_var": 2.561393229166667,
      "learning_rate": 0.0001,
      "loss": 7.4233,
      "loss/crossentropy": 2.1605855494737627,
      "loss/hidden": 3.3015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19197470052167773,
      "step": 12850
    },
    {
      "epoch": 0.3215,
      "grad_norm": 29.375,
      "grad_norm_var": 1.98125,
      "learning_rate": 0.0001,
      "loss": 7.3826,
      "loss/crossentropy": 2.1228017687797545,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21111836601048709,
      "step": 12860
    },
    {
      "epoch": 0.32175,
      "grad_norm": 32.75,
      "grad_norm_var": 2.6666666666666665,
      "learning_rate": 0.0001,
      "loss": 7.317,
      "loss/crossentropy": 1.9638956546783448,
      "loss/hidden": 3.3625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18815859649330377,
      "step": 12870
    },
    {
      "epoch": 0.322,
      "grad_norm": 30.875,
      "grad_norm_var": 2.5927083333333334,
      "learning_rate": 0.0001,
      "loss": 7.3412,
      "loss/crossentropy": 2.15369998216629,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19398804232478142,
      "step": 12880
    },
    {
      "epoch": 0.32225,
      "grad_norm": 28.875,
      "grad_norm_var": 4.522916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3785,
      "loss/crossentropy": 2.134490595757961,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19743212331086396,
      "step": 12890
    },
    {
      "epoch": 0.3225,
      "grad_norm": 28.125,
      "grad_norm_var": 4.824739583333334,
      "learning_rate": 0.0001,
      "loss": 7.2795,
      "loss/crossentropy": 2.05883831679821,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1887126075103879,
      "step": 12900
    },
    {
      "epoch": 0.32275,
      "grad_norm": 31.125,
      "grad_norm_var": 1.6257994456428096e+18,
      "learning_rate": 0.0001,
      "loss": 7.3637,
      "loss/crossentropy": 1.865815930068493,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16476531345397233,
      "step": 12910
    },
    {
      "epoch": 0.323,
      "grad_norm": 27.375,
      "grad_norm_var": 1.6257994456056202e+18,
      "learning_rate": 0.0001,
      "loss": 7.3231,
      "loss/crossentropy": 2.1335863292217256,
      "loss/hidden": 3.3234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18054731655865908,
      "step": 12920
    },
    {
      "epoch": 0.32325,
      "grad_norm": 30.5,
      "grad_norm_var": 1.6822265625,
      "learning_rate": 0.0001,
      "loss": 7.2992,
      "loss/crossentropy": 2.029070366919041,
      "loss/hidden": 3.292578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1656917490065098,
      "step": 12930
    },
    {
      "epoch": 0.3235,
      "grad_norm": 29.875,
      "grad_norm_var": 1.9139973958333334,
      "learning_rate": 0.0001,
      "loss": 7.3529,
      "loss/crossentropy": 2.0367226876318454,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18601053059101105,
      "step": 12940
    },
    {
      "epoch": 0.32375,
      "grad_norm": 27.875,
      "grad_norm_var": 2.1228515625,
      "learning_rate": 0.0001,
      "loss": 7.146,
      "loss/crossentropy": 2.008280509710312,
      "loss/hidden": 3.339453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17591428142040968,
      "step": 12950
    },
    {
      "epoch": 0.324,
      "grad_norm": 31.5,
      "grad_norm_var": 1.8893229166666667,
      "learning_rate": 0.0001,
      "loss": 7.3616,
      "loss/crossentropy": 2.1186923176050185,
      "loss/hidden": 3.476171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20130661278963088,
      "step": 12960
    },
    {
      "epoch": 0.32425,
      "grad_norm": 28.0,
      "grad_norm_var": 1.4749348958333333,
      "learning_rate": 0.0001,
      "loss": 7.3933,
      "loss/crossentropy": 2.1367026805877685,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19593327604234217,
      "step": 12970
    },
    {
      "epoch": 0.3245,
      "grad_norm": 33.5,
      "grad_norm_var": 2.5009765625,
      "learning_rate": 0.0001,
      "loss": 7.36,
      "loss/crossentropy": 1.9709487736225129,
      "loss/hidden": 3.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18522804593667389,
      "step": 12980
    },
    {
      "epoch": 0.32475,
      "grad_norm": 32.25,
      "grad_norm_var": 3.037955729166667,
      "learning_rate": 0.0001,
      "loss": 7.3995,
      "loss/crossentropy": 2.182401825487614,
      "loss/hidden": 3.288671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1944254267960787,
      "step": 12990
    },
    {
      "epoch": 0.325,
      "grad_norm": 31.875,
      "grad_norm_var": 1.8259765625,
      "learning_rate": 0.0001,
      "loss": 7.297,
      "loss/crossentropy": 2.144014260172844,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18785431496798993,
      "step": 13000
    },
    {
      "epoch": 0.32525,
      "grad_norm": 29.125,
      "grad_norm_var": 1.5858723958333334,
      "learning_rate": 0.0001,
      "loss": 7.4172,
      "loss/crossentropy": 2.185704696178436,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21230035796761512,
      "step": 13010
    },
    {
      "epoch": 0.3255,
      "grad_norm": 33.75,
      "grad_norm_var": 2.0989583333333335,
      "learning_rate": 0.0001,
      "loss": 7.3542,
      "loss/crossentropy": 2.169140038639307,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20137128187343478,
      "step": 13020
    },
    {
      "epoch": 0.32575,
      "grad_norm": 29.0,
      "grad_norm_var": 1.7613932291666667,
      "learning_rate": 0.0001,
      "loss": 7.3018,
      "loss/crossentropy": 2.0429045438766478,
      "loss/hidden": 3.306640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1801679054275155,
      "step": 13030
    },
    {
      "epoch": 0.326,
      "grad_norm": 28.25,
      "grad_norm_var": 20.3603515625,
      "learning_rate": 0.0001,
      "loss": 7.3076,
      "loss/crossentropy": 2.090936814248562,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18894913028925658,
      "step": 13040
    },
    {
      "epoch": 0.32625,
      "grad_norm": 29.5,
      "grad_norm_var": 21.5759765625,
      "learning_rate": 0.0001,
      "loss": 7.3032,
      "loss/crossentropy": 2.0933708012104035,
      "loss/hidden": 3.306640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18349476121366023,
      "step": 13050
    },
    {
      "epoch": 0.3265,
      "grad_norm": 31.125,
      "grad_norm_var": 23.737434895833335,
      "learning_rate": 0.0001,
      "loss": 7.4007,
      "loss/crossentropy": 1.9636264845728875,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17653203681111335,
      "step": 13060
    },
    {
      "epoch": 0.32675,
      "grad_norm": 30.875,
      "grad_norm_var": 26.170833333333334,
      "learning_rate": 0.0001,
      "loss": 7.2685,
      "loss/crossentropy": 2.0459864370524885,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18335657650604845,
      "step": 13070
    },
    {
      "epoch": 0.327,
      "grad_norm": 32.25,
      "grad_norm_var": 13.033072916666667,
      "learning_rate": 0.0001,
      "loss": 7.3321,
      "loss/crossentropy": 2.2077785804867744,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1968739289790392,
      "step": 13080
    },
    {
      "epoch": 0.32725,
      "grad_norm": 29.625,
      "grad_norm_var": 8.8634765625,
      "learning_rate": 0.0001,
      "loss": 7.2613,
      "loss/crossentropy": 2.039609357714653,
      "loss/hidden": 3.45078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19310947302728892,
      "step": 13090
    },
    {
      "epoch": 0.3275,
      "grad_norm": 29.875,
      "grad_norm_var": 8.334309895833334,
      "learning_rate": 0.0001,
      "loss": 7.3587,
      "loss/crossentropy": 2.0822102136909963,
      "loss/hidden": 3.26640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1863661216571927,
      "step": 13100
    },
    {
      "epoch": 0.32775,
      "grad_norm": 34.25,
      "grad_norm_var": 8.694205729166667,
      "learning_rate": 0.0001,
      "loss": 7.3602,
      "loss/crossentropy": 2.167416235804558,
      "loss/hidden": 3.36484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19521966725587844,
      "step": 13110
    },
    {
      "epoch": 0.328,
      "grad_norm": 29.125,
      "grad_norm_var": 7.191080729166667,
      "learning_rate": 0.0001,
      "loss": 7.3767,
      "loss/crossentropy": 2.07031906619668,
      "loss/hidden": 3.4015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22052818778902292,
      "step": 13120
    },
    {
      "epoch": 0.32825,
      "grad_norm": 35.5,
      "grad_norm_var": 5.821809895833334,
      "learning_rate": 0.0001,
      "loss": 7.5034,
      "loss/crossentropy": 2.1318026900291445,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19695397429168224,
      "step": 13130
    },
    {
      "epoch": 0.3285,
      "grad_norm": 29.5,
      "grad_norm_var": 4.738541666666666,
      "learning_rate": 0.0001,
      "loss": 7.2907,
      "loss/crossentropy": 1.893085064738989,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17454516123980285,
      "step": 13140
    },
    {
      "epoch": 0.32875,
      "grad_norm": 33.25,
      "grad_norm_var": 4.96640625,
      "learning_rate": 0.0001,
      "loss": 7.3911,
      "loss/crossentropy": 2.0709325544536115,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19108830243349076,
      "step": 13150
    },
    {
      "epoch": 0.329,
      "grad_norm": 30.375,
      "grad_norm_var": 6.593489583333334,
      "learning_rate": 0.0001,
      "loss": 7.278,
      "loss/crossentropy": 2.0809618443250657,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1867564545944333,
      "step": 13160
    },
    {
      "epoch": 0.32925,
      "grad_norm": 28.875,
      "grad_norm_var": 5.355143229166667,
      "learning_rate": 0.0001,
      "loss": 7.1865,
      "loss/crossentropy": 1.86175979077816,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18658734802156687,
      "step": 13170
    },
    {
      "epoch": 0.3295,
      "grad_norm": 32.5,
      "grad_norm_var": 7.297916666666667,
      "learning_rate": 0.0001,
      "loss": 7.2993,
      "loss/crossentropy": 2.0688765406608582,
      "loss/hidden": 3.320703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1887798959389329,
      "step": 13180
    },
    {
      "epoch": 0.32975,
      "grad_norm": 32.0,
      "grad_norm_var": 3.1462890625,
      "learning_rate": 0.0001,
      "loss": 7.396,
      "loss/crossentropy": 2.1349074259400367,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1891916124150157,
      "step": 13190
    },
    {
      "epoch": 0.33,
      "grad_norm": 29.375,
      "grad_norm_var": 3.2080729166666666,
      "learning_rate": 0.0001,
      "loss": 7.2194,
      "loss/crossentropy": 2.0701663225889204,
      "loss/hidden": 3.290625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19238610491156577,
      "step": 13200
    },
    {
      "epoch": 0.33025,
      "grad_norm": 29.875,
      "grad_norm_var": 1.3447265625,
      "learning_rate": 0.0001,
      "loss": 7.3476,
      "loss/crossentropy": 2.126928760111332,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18552705701440572,
      "step": 13210
    },
    {
      "epoch": 0.3305,
      "grad_norm": 29.875,
      "grad_norm_var": 2.4916666666666667,
      "learning_rate": 0.0001,
      "loss": 7.3275,
      "loss/crossentropy": 2.0960667990148067,
      "loss/hidden": 3.3546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18840519580990076,
      "step": 13220
    },
    {
      "epoch": 0.33075,
      "grad_norm": 29.0,
      "grad_norm_var": 31.097916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3449,
      "loss/crossentropy": 1.9724566139280797,
      "loss/hidden": 3.401171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17257819082587958,
      "step": 13230
    },
    {
      "epoch": 0.331,
      "grad_norm": 30.25,
      "grad_norm_var": 29.516080729166667,
      "learning_rate": 0.0001,
      "loss": 7.5214,
      "loss/crossentropy": 2.144184100627899,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19250041600316764,
      "step": 13240
    },
    {
      "epoch": 0.33125,
      "grad_norm": 28.25,
      "grad_norm_var": 6.55,
      "learning_rate": 0.0001,
      "loss": 7.3245,
      "loss/crossentropy": 1.9724821582436562,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18454758413136005,
      "step": 13250
    },
    {
      "epoch": 0.3315,
      "grad_norm": 35.75,
      "grad_norm_var": 10.574739583333333,
      "learning_rate": 0.0001,
      "loss": 7.2067,
      "loss/crossentropy": 2.0341859996318816,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19355686828494073,
      "step": 13260
    },
    {
      "epoch": 0.33175,
      "grad_norm": 31.875,
      "grad_norm_var": 6.093489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3386,
      "loss/crossentropy": 2.1593449860811234,
      "loss/hidden": 3.282421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18228702070191502,
      "step": 13270
    },
    {
      "epoch": 0.332,
      "grad_norm": 32.25,
      "grad_norm_var": 4.518489583333333,
      "learning_rate": 0.0001,
      "loss": 7.2584,
      "loss/crossentropy": 2.021310421824455,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.193563433829695,
      "step": 13280
    },
    {
      "epoch": 0.33225,
      "grad_norm": 33.25,
      "grad_norm_var": 3.008072916666667,
      "learning_rate": 0.0001,
      "loss": 7.3654,
      "loss/crossentropy": 2.010246267169714,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18375244587659836,
      "step": 13290
    },
    {
      "epoch": 0.3325,
      "grad_norm": 33.5,
      "grad_norm_var": 1.8309895833333334,
      "learning_rate": 0.0001,
      "loss": 7.4013,
      "loss/crossentropy": 2.1324703454971314,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19651575423777104,
      "step": 13300
    },
    {
      "epoch": 0.33275,
      "grad_norm": 31.625,
      "grad_norm_var": 9.018489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4123,
      "loss/crossentropy": 2.2024870067834854,
      "loss/hidden": 3.36484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19117040373384953,
      "step": 13310
    },
    {
      "epoch": 0.333,
      "grad_norm": 29.5,
      "grad_norm_var": 54.362239583333334,
      "learning_rate": 0.0001,
      "loss": 7.3561,
      "loss/crossentropy": 1.9291939452290534,
      "loss/hidden": 3.33828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18055410776287317,
      "step": 13320
    },
    {
      "epoch": 0.33325,
      "grad_norm": 30.5,
      "grad_norm_var": 53.271875,
      "learning_rate": 0.0001,
      "loss": 7.3417,
      "loss/crossentropy": 1.9812588825821877,
      "loss/hidden": 3.246875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17277830513194203,
      "step": 13330
    },
    {
      "epoch": 0.3335,
      "grad_norm": 30.625,
      "grad_norm_var": 2.0332682291666666,
      "learning_rate": 0.0001,
      "loss": 7.4792,
      "loss/crossentropy": 2.1468162171542646,
      "loss/hidden": 3.29453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1815652133896947,
      "step": 13340
    },
    {
      "epoch": 0.33375,
      "grad_norm": 30.5,
      "grad_norm_var": 1.8942057291666667,
      "learning_rate": 0.0001,
      "loss": 7.3437,
      "loss/crossentropy": 1.9899487346410751,
      "loss/hidden": 3.336328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1736677044071257,
      "step": 13350
    },
    {
      "epoch": 0.334,
      "grad_norm": 32.75,
      "grad_norm_var": 3.374955311060432e+18,
      "learning_rate": 0.0001,
      "loss": 7.4592,
      "loss/crossentropy": 2.0041578873991965,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19686540886759757,
      "step": 13360
    },
    {
      "epoch": 0.33425,
      "grad_norm": 29.375,
      "grad_norm_var": 3.3749553110757407e+18,
      "learning_rate": 0.0001,
      "loss": 7.3521,
      "loss/crossentropy": 2.050678627192974,
      "loss/hidden": 3.35859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1835683614946902,
      "step": 13370
    },
    {
      "epoch": 0.3345,
      "grad_norm": 32.5,
      "grad_norm_var": 1.4561848958333334,
      "learning_rate": 0.0001,
      "loss": 7.3463,
      "loss/crossentropy": 2.255459001660347,
      "loss/hidden": 3.3484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2026721488684416,
      "step": 13380
    },
    {
      "epoch": 0.33475,
      "grad_norm": 55.0,
      "grad_norm_var": 37.805989583333336,
      "learning_rate": 0.0001,
      "loss": 7.432,
      "loss/crossentropy": 2.092339722812176,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20365451760590075,
      "step": 13390
    },
    {
      "epoch": 0.335,
      "grad_norm": 29.5,
      "grad_norm_var": 39.1962890625,
      "learning_rate": 0.0001,
      "loss": 7.3082,
      "loss/crossentropy": 2.060484157502651,
      "loss/hidden": 3.49921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19208470694720745,
      "step": 13400
    },
    {
      "epoch": 0.33525,
      "grad_norm": 30.875,
      "grad_norm_var": 4.119205729166667,
      "learning_rate": 0.0001,
      "loss": 7.3159,
      "loss/crossentropy": 2.0885482341051103,
      "loss/hidden": 3.3484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18701701434329152,
      "step": 13410
    },
    {
      "epoch": 0.3355,
      "grad_norm": 34.75,
      "grad_norm_var": 3.8354166666666667,
      "learning_rate": 0.0001,
      "loss": 7.3792,
      "loss/crossentropy": 2.064309825748205,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20586481597274542,
      "step": 13420
    },
    {
      "epoch": 0.33575,
      "grad_norm": 29.75,
      "grad_norm_var": 2.535416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4193,
      "loss/crossentropy": 1.8737250491976738,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19758351668715476,
      "step": 13430
    },
    {
      "epoch": 0.336,
      "grad_norm": 30.25,
      "grad_norm_var": 3.8686848958333333,
      "learning_rate": 0.0001,
      "loss": 7.4324,
      "loss/crossentropy": 1.9396042831242084,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18089414723217487,
      "step": 13440
    },
    {
      "epoch": 0.33625,
      "grad_norm": 28.875,
      "grad_norm_var": 4.540625,
      "learning_rate": 0.0001,
      "loss": 7.3798,
      "loss/crossentropy": 2.004528859257698,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22120596412569285,
      "step": 13450
    },
    {
      "epoch": 0.3365,
      "grad_norm": 29.125,
      "grad_norm_var": 1.69375,
      "learning_rate": 0.0001,
      "loss": 7.3708,
      "loss/crossentropy": 2.170805335044861,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18870076667517424,
      "step": 13460
    },
    {
      "epoch": 0.33675,
      "grad_norm": 29.375,
      "grad_norm_var": 2.9284656301175997e+18,
      "learning_rate": 0.0001,
      "loss": 7.3594,
      "loss/crossentropy": 2.0726011991500854,
      "loss/hidden": 3.310546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18316982481628657,
      "step": 13470
    },
    {
      "epoch": 0.337,
      "grad_norm": 32.25,
      "grad_norm_var": 3.999739583333333,
      "learning_rate": 0.0001,
      "loss": 7.3072,
      "loss/crossentropy": 2.0701726540923118,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18232779456302523,
      "step": 13480
    },
    {
      "epoch": 0.33725,
      "grad_norm": 31.625,
      "grad_norm_var": 1.01640625,
      "learning_rate": 0.0001,
      "loss": 7.3553,
      "loss/crossentropy": 2.066875821352005,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18299774192273616,
      "step": 13490
    },
    {
      "epoch": 0.3375,
      "grad_norm": 32.0,
      "grad_norm_var": 1.4489583333333333,
      "learning_rate": 0.0001,
      "loss": 7.2708,
      "loss/crossentropy": 2.110422171652317,
      "loss/hidden": 3.353125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19538786429911853,
      "step": 13500
    },
    {
      "epoch": 0.33775,
      "grad_norm": 29.875,
      "grad_norm_var": 1.06015625,
      "learning_rate": 0.0001,
      "loss": 7.3529,
      "loss/crossentropy": 2.1061682522296907,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19031856823712587,
      "step": 13510
    },
    {
      "epoch": 0.338,
      "grad_norm": 32.5,
      "grad_norm_var": 1.40390625,
      "learning_rate": 0.0001,
      "loss": 7.2473,
      "loss/crossentropy": 2.0029176853597166,
      "loss/hidden": 3.284375,
      "loss/jsd": 0.0,
      "loss/logits": 0.182035060133785,
      "step": 13520
    },
    {
      "epoch": 0.33825,
      "grad_norm": 29.375,
      "grad_norm_var": 9.417708333333334,
      "learning_rate": 0.0001,
      "loss": 7.3084,
      "loss/crossentropy": 2.1238187912851574,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.188486148416996,
      "step": 13530
    },
    {
      "epoch": 0.3385,
      "grad_norm": 30.25,
      "grad_norm_var": 1.3983723958333334,
      "learning_rate": 0.0001,
      "loss": 7.2256,
      "loss/crossentropy": 1.9766217768192291,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17294846558943391,
      "step": 13540
    },
    {
      "epoch": 0.33875,
      "grad_norm": 29.375,
      "grad_norm_var": 3.0874348958333333,
      "learning_rate": 0.0001,
      "loss": 7.327,
      "loss/crossentropy": 2.1272383123636245,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18510236088186502,
      "step": 13550
    },
    {
      "epoch": 0.339,
      "grad_norm": 31.0,
      "grad_norm_var": 3.10390625,
      "learning_rate": 0.0001,
      "loss": 7.2885,
      "loss/crossentropy": 2.0515687823295594,
      "loss/hidden": 3.368359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18967761769890784,
      "step": 13560
    },
    {
      "epoch": 0.33925,
      "grad_norm": 29.625,
      "grad_norm_var": 11.933072916666667,
      "learning_rate": 0.0001,
      "loss": 7.3573,
      "loss/crossentropy": 2.0446967758238315,
      "loss/hidden": 3.425,
      "loss/jsd": 0.0,
      "loss/logits": 0.18640230242162942,
      "step": 13570
    },
    {
      "epoch": 0.3395,
      "grad_norm": 28.875,
      "grad_norm_var": 1.84765625,
      "learning_rate": 0.0001,
      "loss": 7.2965,
      "loss/crossentropy": 1.9837520524859429,
      "loss/hidden": 3.3171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1916389312595129,
      "step": 13580
    },
    {
      "epoch": 0.33975,
      "grad_norm": 30.5,
      "grad_norm_var": 3.1791666666666667,
      "learning_rate": 0.0001,
      "loss": 7.3632,
      "loss/crossentropy": 2.1806528866291046,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19690024815499782,
      "step": 13590
    },
    {
      "epoch": 0.34,
      "grad_norm": 30.75,
      "grad_norm_var": 1.3385416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3761,
      "loss/crossentropy": 2.215220719575882,
      "loss/hidden": 3.312109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19034639187157154,
      "step": 13600
    },
    {
      "epoch": 0.34025,
      "grad_norm": 27.375,
      "grad_norm_var": 2.7129557291666666,
      "learning_rate": 0.0001,
      "loss": 7.3129,
      "loss/crossentropy": 2.16262392103672,
      "loss/hidden": 3.503125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20174582321196793,
      "step": 13610
    },
    {
      "epoch": 0.3405,
      "grad_norm": 30.5,
      "grad_norm_var": 3.2363932291666666,
      "learning_rate": 0.0001,
      "loss": 7.3501,
      "loss/crossentropy": 2.126379433274269,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18432336021214724,
      "step": 13620
    },
    {
      "epoch": 0.34075,
      "grad_norm": 30.875,
      "grad_norm_var": 1.8176432291666667,
      "learning_rate": 0.0001,
      "loss": 7.3601,
      "loss/crossentropy": 2.0324837125837805,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18331200983375312,
      "step": 13630
    },
    {
      "epoch": 0.341,
      "grad_norm": 29.0,
      "grad_norm_var": 1.7942057291666667,
      "learning_rate": 0.0001,
      "loss": 7.3882,
      "loss/crossentropy": 2.176556921005249,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19637300558388232,
      "step": 13640
    },
    {
      "epoch": 0.34125,
      "grad_norm": 29.875,
      "grad_norm_var": 3.758072916666667,
      "learning_rate": 0.0001,
      "loss": 7.1992,
      "loss/crossentropy": 1.9680285774171353,
      "loss/hidden": 3.30546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1837764661759138,
      "step": 13650
    },
    {
      "epoch": 0.3415,
      "grad_norm": 29.125,
      "grad_norm_var": 2.6936848958333335,
      "learning_rate": 0.0001,
      "loss": 7.361,
      "loss/crossentropy": 2.1511219844222067,
      "loss/hidden": 3.29140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1746565790846944,
      "step": 13660
    },
    {
      "epoch": 0.34175,
      "grad_norm": 28.375,
      "grad_norm_var": 2.312239583333333,
      "learning_rate": 0.0001,
      "loss": 7.3856,
      "loss/crossentropy": 2.187084162980318,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18925763312727212,
      "step": 13670
    },
    {
      "epoch": 0.342,
      "grad_norm": 29.75,
      "grad_norm_var": 2.4139973958333334,
      "learning_rate": 0.0001,
      "loss": 7.196,
      "loss/crossentropy": 2.0472760550677775,
      "loss/hidden": 3.32109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17813790533691645,
      "step": 13680
    },
    {
      "epoch": 0.34225,
      "grad_norm": 32.25,
      "grad_norm_var": 1.6458333333333333,
      "learning_rate": 0.0001,
      "loss": 7.4184,
      "loss/crossentropy": 2.0482982218265535,
      "loss/hidden": 3.483203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1855379816144705,
      "step": 13690
    },
    {
      "epoch": 0.3425,
      "grad_norm": 27.875,
      "grad_norm_var": 98.41432291666666,
      "learning_rate": 0.0001,
      "loss": 7.2863,
      "loss/crossentropy": 2.0204505778849127,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17318440061062573,
      "step": 13700
    },
    {
      "epoch": 0.34275,
      "grad_norm": 28.375,
      "grad_norm_var": 33.1900390625,
      "learning_rate": 0.0001,
      "loss": 7.3409,
      "loss/crossentropy": 2.0783301174640654,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19091757852584124,
      "step": 13710
    },
    {
      "epoch": 0.343,
      "grad_norm": 32.0,
      "grad_norm_var": 3.7150390625,
      "learning_rate": 0.0001,
      "loss": 7.3412,
      "loss/crossentropy": 2.0553587220609186,
      "loss/hidden": 3.39140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18813807256519793,
      "step": 13720
    },
    {
      "epoch": 0.34325,
      "grad_norm": 34.25,
      "grad_norm_var": 1586.7400390625,
      "learning_rate": 0.0001,
      "loss": 7.4918,
      "loss/crossentropy": 2.14101425036788,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1929340995848179,
      "step": 13730
    },
    {
      "epoch": 0.3435,
      "grad_norm": 32.5,
      "grad_norm_var": 4.633072916666666,
      "learning_rate": 0.0001,
      "loss": 7.3162,
      "loss/crossentropy": 2.0531166955828666,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18961485847830772,
      "step": 13740
    },
    {
      "epoch": 0.34375,
      "grad_norm": 51.5,
      "grad_norm_var": 29.780989583333334,
      "learning_rate": 0.0001,
      "loss": 7.3981,
      "loss/crossentropy": 2.146902695298195,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18639619778841734,
      "step": 13750
    },
    {
      "epoch": 0.344,
      "grad_norm": 31.875,
      "grad_norm_var": 27.141080729166667,
      "learning_rate": 0.0001,
      "loss": 7.3596,
      "loss/crossentropy": 2.2127513885498047,
      "loss/hidden": 3.407421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19791291914880277,
      "step": 13760
    },
    {
      "epoch": 0.34425,
      "grad_norm": 32.25,
      "grad_norm_var": 1.9809895833333333,
      "learning_rate": 0.0001,
      "loss": 7.3245,
      "loss/crossentropy": 2.113487794995308,
      "loss/hidden": 3.396875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18752570655196904,
      "step": 13770
    },
    {
      "epoch": 0.3445,
      "grad_norm": 28.875,
      "grad_norm_var": 2.7666666666666666,
      "learning_rate": 0.0001,
      "loss": 7.3737,
      "loss/crossentropy": 2.192716282606125,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1977760722860694,
      "step": 13780
    },
    {
      "epoch": 0.34475,
      "grad_norm": 30.375,
      "grad_norm_var": 1.4999348958333334,
      "learning_rate": 0.0001,
      "loss": 7.2308,
      "loss/crossentropy": 2.1602986216545106,
      "loss/hidden": 3.2796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1859057329595089,
      "step": 13790
    },
    {
      "epoch": 0.345,
      "grad_norm": 45.25,
      "grad_norm_var": 15.587434895833333,
      "learning_rate": 0.0001,
      "loss": 7.3689,
      "loss/crossentropy": 2.042529730498791,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2055663662031293,
      "step": 13800
    },
    {
      "epoch": 0.34525,
      "grad_norm": 31.375,
      "grad_norm_var": 16.111458333333335,
      "learning_rate": 0.0001,
      "loss": 7.355,
      "loss/crossentropy": 2.004579763114452,
      "loss/hidden": 3.46171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18781258668750525,
      "step": 13810
    },
    {
      "epoch": 0.3455,
      "grad_norm": 27.75,
      "grad_norm_var": 2.9369140625,
      "learning_rate": 0.0001,
      "loss": 7.2946,
      "loss/crossentropy": 1.9647000446915626,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18860411625355483,
      "step": 13820
    },
    {
      "epoch": 0.34575,
      "grad_norm": 28.625,
      "grad_norm_var": 2.4830729166666665,
      "learning_rate": 0.0001,
      "loss": 7.3643,
      "loss/crossentropy": 2.1840324103832245,
      "loss/hidden": 3.321875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18433275502175092,
      "step": 13830
    },
    {
      "epoch": 0.346,
      "grad_norm": 31.0,
      "grad_norm_var": 2.061393229166667,
      "learning_rate": 0.0001,
      "loss": 7.39,
      "loss/crossentropy": 2.1660002395510674,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18589738458395005,
      "step": 13840
    },
    {
      "epoch": 0.34625,
      "grad_norm": 30.625,
      "grad_norm_var": 3.237434895833333,
      "learning_rate": 0.0001,
      "loss": 7.2772,
      "loss/crossentropy": 2.0304790273308755,
      "loss/hidden": 3.383984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19276629090309144,
      "step": 13850
    },
    {
      "epoch": 0.3465,
      "grad_norm": 27.375,
      "grad_norm_var": 4.35390625,
      "learning_rate": 0.0001,
      "loss": 7.2433,
      "loss/crossentropy": 2.0913878247141837,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17961858343333006,
      "step": 13860
    },
    {
      "epoch": 0.34675,
      "grad_norm": 31.625,
      "grad_norm_var": 1.9833333333333334,
      "learning_rate": 0.0001,
      "loss": 7.298,
      "loss/crossentropy": 1.9085475612431764,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1773257221095264,
      "step": 13870
    },
    {
      "epoch": 0.347,
      "grad_norm": 31.75,
      "grad_norm_var": 1.2061848958333334,
      "learning_rate": 0.0001,
      "loss": 7.3026,
      "loss/crossentropy": 2.134579537808895,
      "loss/hidden": 3.29375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18533094711601733,
      "step": 13880
    },
    {
      "epoch": 0.34725,
      "grad_norm": 29.75,
      "grad_norm_var": 1.74140625,
      "learning_rate": 0.0001,
      "loss": 7.2397,
      "loss/crossentropy": 2.0479501873254775,
      "loss/hidden": 3.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1840794005431235,
      "step": 13890
    },
    {
      "epoch": 0.3475,
      "grad_norm": 30.625,
      "grad_norm_var": 2.6400390625,
      "learning_rate": 0.0001,
      "loss": 7.2447,
      "loss/crossentropy": 2.0175841793417932,
      "loss/hidden": 3.29296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17613419592380525,
      "step": 13900
    },
    {
      "epoch": 0.34775,
      "grad_norm": 32.5,
      "grad_norm_var": 5.333072916666667,
      "learning_rate": 0.0001,
      "loss": 7.3573,
      "loss/crossentropy": 2.248972457647324,
      "loss/hidden": 3.35234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19207954667508603,
      "step": 13910
    },
    {
      "epoch": 0.348,
      "grad_norm": 31.0,
      "grad_norm_var": 4.843489583333334,
      "learning_rate": 0.0001,
      "loss": 7.3649,
      "loss/crossentropy": 2.215424671769142,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19513901360332966,
      "step": 13920
    },
    {
      "epoch": 0.34825,
      "grad_norm": 30.25,
      "grad_norm_var": 2.004622395833333,
      "learning_rate": 0.0001,
      "loss": 7.4324,
      "loss/crossentropy": 2.157894307374954,
      "loss/hidden": 3.457421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20247049070894718,
      "step": 13930
    },
    {
      "epoch": 0.3485,
      "grad_norm": 28.5,
      "grad_norm_var": 2.4150390625,
      "learning_rate": 0.0001,
      "loss": 7.3422,
      "loss/crossentropy": 2.14641355201602,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1898975013755262,
      "step": 13940
    },
    {
      "epoch": 0.34875,
      "grad_norm": 28.5,
      "grad_norm_var": 6.0103515625,
      "learning_rate": 0.0001,
      "loss": 7.3309,
      "loss/crossentropy": 2.0432268232107162,
      "loss/hidden": 3.461328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19791582636535168,
      "step": 13950
    },
    {
      "epoch": 0.349,
      "grad_norm": 28.875,
      "grad_norm_var": 4.4744140625,
      "learning_rate": 0.0001,
      "loss": 7.3187,
      "loss/crossentropy": 2.1648118555545808,
      "loss/hidden": 3.301171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1794663654640317,
      "step": 13960
    },
    {
      "epoch": 0.34925,
      "grad_norm": 29.5,
      "grad_norm_var": 3.1332682291666667,
      "learning_rate": 0.0001,
      "loss": 7.3122,
      "loss/crossentropy": 2.0566162675619126,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18033871669322252,
      "step": 13970
    },
    {
      "epoch": 0.3495,
      "grad_norm": 27.875,
      "grad_norm_var": 1.3686848958333333,
      "learning_rate": 0.0001,
      "loss": 7.2538,
      "loss/crossentropy": 2.0309479638934134,
      "loss/hidden": 3.316015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1834420131519437,
      "step": 13980
    },
    {
      "epoch": 0.34975,
      "grad_norm": 30.0,
      "grad_norm_var": 2.3510416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3068,
      "loss/crossentropy": 2.0843526370823384,
      "loss/hidden": 3.35859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18843147568404675,
      "step": 13990
    },
    {
      "epoch": 0.35,
      "grad_norm": 34.25,
      "grad_norm_var": 10.67265625,
      "learning_rate": 0.0001,
      "loss": 7.3839,
      "loss/crossentropy": 2.0708370715379716,
      "loss/hidden": 3.389453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19229107201099396,
      "step": 14000
    },
    {
      "epoch": 0.35025,
      "grad_norm": 28.5,
      "grad_norm_var": 11.178059895833334,
      "learning_rate": 0.0001,
      "loss": 7.3404,
      "loss/crossentropy": 2.016050732135773,
      "loss/hidden": 3.49921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1993856718763709,
      "step": 14010
    },
    {
      "epoch": 0.3505,
      "grad_norm": 31.375,
      "grad_norm_var": 1.55625,
      "learning_rate": 0.0001,
      "loss": 7.4397,
      "loss/crossentropy": 2.0941019743680953,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1941377494484186,
      "step": 14020
    },
    {
      "epoch": 0.35075,
      "grad_norm": 28.125,
      "grad_norm_var": 1.7934895833333333,
      "learning_rate": 0.0001,
      "loss": 7.2787,
      "loss/crossentropy": 2.175595435500145,
      "loss/hidden": 3.36484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1906340267509222,
      "step": 14030
    },
    {
      "epoch": 0.351,
      "grad_norm": 29.625,
      "grad_norm_var": 3.01015625,
      "learning_rate": 0.0001,
      "loss": 7.3322,
      "loss/crossentropy": 1.945583702623844,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17433239622041583,
      "step": 14040
    },
    {
      "epoch": 0.35125,
      "grad_norm": 32.0,
      "grad_norm_var": 11.308072916666667,
      "learning_rate": 0.0001,
      "loss": 7.4361,
      "loss/crossentropy": 2.078453540802002,
      "loss/hidden": 3.271484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1803632376715541,
      "step": 14050
    },
    {
      "epoch": 0.3515,
      "grad_norm": 31.125,
      "grad_norm_var": 11.920833333333333,
      "learning_rate": 0.0001,
      "loss": 7.3917,
      "loss/crossentropy": 2.2025377944111826,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18862273804843427,
      "step": 14060
    },
    {
      "epoch": 0.35175,
      "grad_norm": 26.875,
      "grad_norm_var": 14.82890625,
      "learning_rate": 0.0001,
      "loss": 7.294,
      "loss/crossentropy": 2.0608675971627237,
      "loss/hidden": 3.229296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17895151115953922,
      "step": 14070
    },
    {
      "epoch": 0.352,
      "grad_norm": 31.5,
      "grad_norm_var": 16.512239583333333,
      "learning_rate": 0.0001,
      "loss": 7.2397,
      "loss/crossentropy": 1.951252208650112,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17488121166825293,
      "step": 14080
    },
    {
      "epoch": 0.35225,
      "grad_norm": 29.25,
      "grad_norm_var": 1.6044777159285975e+18,
      "learning_rate": 0.0001,
      "loss": 7.3104,
      "loss/crossentropy": 2.1797886729240417,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19085733741521835,
      "step": 14090
    },
    {
      "epoch": 0.3525,
      "grad_norm": 30.625,
      "grad_norm_var": 1.6044777150524774e+18,
      "learning_rate": 0.0001,
      "loss": 7.269,
      "loss/crossentropy": 2.133794055879116,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18788108974695206,
      "step": 14100
    },
    {
      "epoch": 0.35275,
      "grad_norm": 31.125,
      "grad_norm_var": 57.800455729166664,
      "learning_rate": 0.0001,
      "loss": 7.2859,
      "loss/crossentropy": 2.0392952769994737,
      "loss/hidden": 3.317578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1854009686037898,
      "step": 14110
    },
    {
      "epoch": 0.353,
      "grad_norm": 29.0,
      "grad_norm_var": 2.729622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3152,
      "loss/crossentropy": 2.026089659333229,
      "loss/hidden": 3.36875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18226654808968307,
      "step": 14120
    },
    {
      "epoch": 0.35325,
      "grad_norm": 30.375,
      "grad_norm_var": 2.703125,
      "learning_rate": 0.0001,
      "loss": 7.4705,
      "loss/crossentropy": 2.132002358883619,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2102237056940794,
      "step": 14130
    },
    {
      "epoch": 0.3535,
      "grad_norm": 29.0,
      "grad_norm_var": 2.9497395833333333,
      "learning_rate": 0.0001,
      "loss": 7.3236,
      "loss/crossentropy": 2.0131492763757706,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19223638353869318,
      "step": 14140
    },
    {
      "epoch": 0.35375,
      "grad_norm": 29.5,
      "grad_norm_var": 2.7212890625,
      "learning_rate": 0.0001,
      "loss": 7.271,
      "loss/crossentropy": 2.1130865029990673,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1914577091112733,
      "step": 14150
    },
    {
      "epoch": 0.354,
      "grad_norm": 31.125,
      "grad_norm_var": 1.8254557291666667,
      "learning_rate": 0.0001,
      "loss": 7.4232,
      "loss/crossentropy": 2.12569759786129,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19339151214808226,
      "step": 14160
    },
    {
      "epoch": 0.35425,
      "grad_norm": 29.875,
      "grad_norm_var": 1064.6176432291666,
      "learning_rate": 0.0001,
      "loss": 7.2083,
      "loss/crossentropy": 2.0492417976260184,
      "loss/hidden": 3.37109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18359060864895582,
      "step": 14170
    },
    {
      "epoch": 0.3545,
      "grad_norm": 31.375,
      "grad_norm_var": 9.59765625,
      "learning_rate": 0.0001,
      "loss": 7.3908,
      "loss/crossentropy": 2.099221628904343,
      "loss/hidden": 3.419140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18481487538665534,
      "step": 14180
    },
    {
      "epoch": 0.35475,
      "grad_norm": 44.25,
      "grad_norm_var": 12.892708333333333,
      "learning_rate": 0.0001,
      "loss": 7.3495,
      "loss/crossentropy": 2.04138702750206,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1797080934047699,
      "step": 14190
    },
    {
      "epoch": 0.355,
      "grad_norm": 29.875,
      "grad_norm_var": 15.040559895833333,
      "learning_rate": 0.0001,
      "loss": 7.3724,
      "loss/crossentropy": 2.0216434367001055,
      "loss/hidden": 3.440625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19053793977946043,
      "step": 14200
    },
    {
      "epoch": 0.35525,
      "grad_norm": 28.25,
      "grad_norm_var": 13.096809895833333,
      "learning_rate": 0.0001,
      "loss": 7.299,
      "loss/crossentropy": 2.160589988529682,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1827318999916315,
      "step": 14210
    },
    {
      "epoch": 0.3555,
      "grad_norm": 31.625,
      "grad_norm_var": 9.878059895833333,
      "learning_rate": 0.0001,
      "loss": 7.3381,
      "loss/crossentropy": 1.7803485259413718,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16204789485782384,
      "step": 14220
    },
    {
      "epoch": 0.35575,
      "grad_norm": 28.625,
      "grad_norm_var": 5.543489583333334,
      "learning_rate": 0.0001,
      "loss": 7.2506,
      "loss/crossentropy": 1.8810015477240085,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17873666901141405,
      "step": 14230
    },
    {
      "epoch": 0.356,
      "grad_norm": 30.375,
      "grad_norm_var": 4.208268229166666,
      "learning_rate": 0.0001,
      "loss": 7.4941,
      "loss/crossentropy": 2.016931130737066,
      "loss/hidden": 3.300390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17651305962353944,
      "step": 14240
    },
    {
      "epoch": 0.35625,
      "grad_norm": 29.75,
      "grad_norm_var": 2.067122395833333,
      "learning_rate": 0.0001,
      "loss": 7.3659,
      "loss/crossentropy": 2.1511716455221177,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18431759886443616,
      "step": 14250
    },
    {
      "epoch": 0.3565,
      "grad_norm": 31.5,
      "grad_norm_var": 14.267708333333333,
      "learning_rate": 0.0001,
      "loss": 7.4284,
      "loss/crossentropy": 2.0923829920589925,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20195957068353892,
      "step": 14260
    },
    {
      "epoch": 0.35675,
      "grad_norm": 32.75,
      "grad_norm_var": 2.780989583333333,
      "learning_rate": 0.0001,
      "loss": 7.3516,
      "loss/crossentropy": 2.1890153646469117,
      "loss/hidden": 3.407421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19206568598747253,
      "step": 14270
    },
    {
      "epoch": 0.357,
      "grad_norm": 26.875,
      "grad_norm_var": 3.0858723958333334,
      "learning_rate": 0.0001,
      "loss": 7.2643,
      "loss/crossentropy": 2.109933242201805,
      "loss/hidden": 3.362890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1951975781470537,
      "step": 14280
    },
    {
      "epoch": 0.35725,
      "grad_norm": 28.875,
      "grad_norm_var": 1.7747395833333333,
      "learning_rate": 0.0001,
      "loss": 7.2549,
      "loss/crossentropy": 2.104505704343319,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19365676920861005,
      "step": 14290
    },
    {
      "epoch": 0.3575,
      "grad_norm": 32.25,
      "grad_norm_var": 11.114322916666667,
      "learning_rate": 0.0001,
      "loss": 7.247,
      "loss/crossentropy": 2.108228546380997,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18031150791794062,
      "step": 14300
    },
    {
      "epoch": 0.35775,
      "grad_norm": 33.5,
      "grad_norm_var": 4.084375,
      "learning_rate": 0.0001,
      "loss": 7.2652,
      "loss/crossentropy": 2.11306362003088,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19340378735214472,
      "step": 14310
    },
    {
      "epoch": 0.358,
      "grad_norm": 31.625,
      "grad_norm_var": 2.2712890625,
      "learning_rate": 0.0001,
      "loss": 7.5154,
      "loss/crossentropy": 2.073298954963684,
      "loss/hidden": 3.4765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20225481800734996,
      "step": 14320
    },
    {
      "epoch": 0.35825,
      "grad_norm": 33.0,
      "grad_norm_var": 7.609309895833333,
      "learning_rate": 0.0001,
      "loss": 7.3172,
      "loss/crossentropy": 2.068447032570839,
      "loss/hidden": 3.360546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1848990023136139,
      "step": 14330
    },
    {
      "epoch": 0.3585,
      "grad_norm": 28.625,
      "grad_norm_var": 9.287434895833334,
      "learning_rate": 0.0001,
      "loss": 7.341,
      "loss/crossentropy": 2.0122443050146104,
      "loss/hidden": 3.270703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16750977858901023,
      "step": 14340
    },
    {
      "epoch": 0.35875,
      "grad_norm": 31.625,
      "grad_norm_var": 1.5587890625,
      "learning_rate": 0.0001,
      "loss": 7.4114,
      "loss/crossentropy": 1.9815908901393413,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19162300042808056,
      "step": 14350
    },
    {
      "epoch": 0.359,
      "grad_norm": 34.5,
      "grad_norm_var": 1.75390625,
      "learning_rate": 0.0001,
      "loss": 7.4096,
      "loss/crossentropy": 2.0974107921123504,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18985395636409522,
      "step": 14360
    },
    {
      "epoch": 0.35925,
      "grad_norm": 29.875,
      "grad_norm_var": 27.2291015625,
      "learning_rate": 0.0001,
      "loss": 7.2838,
      "loss/crossentropy": 2.18302740752697,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19919114038348198,
      "step": 14370
    },
    {
      "epoch": 0.3595,
      "grad_norm": 29.0,
      "grad_norm_var": 2.393684895833333,
      "learning_rate": 0.0001,
      "loss": 7.3591,
      "loss/crossentropy": 2.1828642159700395,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1880602626129985,
      "step": 14380
    },
    {
      "epoch": 0.35975,
      "grad_norm": 28.25,
      "grad_norm_var": 1.8875,
      "learning_rate": 0.0001,
      "loss": 7.4149,
      "loss/crossentropy": 2.0250351071357726,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21426690481603144,
      "step": 14390
    },
    {
      "epoch": 0.36,
      "grad_norm": 36.0,
      "grad_norm_var": 4.823893229166667,
      "learning_rate": 0.0001,
      "loss": 7.3843,
      "loss/crossentropy": 1.9927750542759894,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18537398371845484,
      "step": 14400
    },
    {
      "epoch": 0.36025,
      "grad_norm": 48.0,
      "grad_norm_var": 20.270247395833334,
      "learning_rate": 0.0001,
      "loss": 7.3442,
      "loss/crossentropy": 2.2051343381404878,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18966466654092073,
      "step": 14410
    },
    {
      "epoch": 0.3605,
      "grad_norm": 26.75,
      "grad_norm_var": 32.15598958333333,
      "learning_rate": 0.0001,
      "loss": 7.3599,
      "loss/crossentropy": 2.050907912105322,
      "loss/hidden": 3.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18331205267459155,
      "step": 14420
    },
    {
      "epoch": 0.36075,
      "grad_norm": 32.0,
      "grad_norm_var": 15.5244140625,
      "learning_rate": 0.0001,
      "loss": 7.4431,
      "loss/crossentropy": 1.9177762359380721,
      "loss/hidden": 3.507421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18733049537986518,
      "step": 14430
    },
    {
      "epoch": 0.361,
      "grad_norm": 31.625,
      "grad_norm_var": 3.370247395833333,
      "learning_rate": 0.0001,
      "loss": 7.3748,
      "loss/crossentropy": 2.1186861246824265,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1846270103007555,
      "step": 14440
    },
    {
      "epoch": 0.36125,
      "grad_norm": 31.125,
      "grad_norm_var": 2.378059895833333,
      "learning_rate": 0.0001,
      "loss": 7.4001,
      "loss/crossentropy": 2.1338623888790607,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889139147475362,
      "step": 14450
    },
    {
      "epoch": 0.3615,
      "grad_norm": 29.125,
      "grad_norm_var": 5.045572916666667,
      "learning_rate": 0.0001,
      "loss": 7.3355,
      "loss/crossentropy": 1.9586334988474845,
      "loss/hidden": 3.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1830888209864497,
      "step": 14460
    },
    {
      "epoch": 0.36175,
      "grad_norm": 29.375,
      "grad_norm_var": 11.797916666666667,
      "learning_rate": 0.0001,
      "loss": 7.4646,
      "loss/crossentropy": 2.0122861742973326,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2004035959020257,
      "step": 14470
    },
    {
      "epoch": 0.362,
      "grad_norm": 32.0,
      "grad_norm_var": 11.684309895833334,
      "learning_rate": 0.0001,
      "loss": 7.4154,
      "loss/crossentropy": 2.1907495334744453,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2014760635793209,
      "step": 14480
    },
    {
      "epoch": 0.36225,
      "grad_norm": 32.75,
      "grad_norm_var": 14.495247395833333,
      "learning_rate": 0.0001,
      "loss": 7.3255,
      "loss/crossentropy": 2.0414177522063257,
      "loss/hidden": 3.33515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17212325502187015,
      "step": 14490
    },
    {
      "epoch": 0.3625,
      "grad_norm": 28.875,
      "grad_norm_var": 9.85390625,
      "learning_rate": 0.0001,
      "loss": 7.4188,
      "loss/crossentropy": 1.875846453011036,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18002864755690098,
      "step": 14500
    },
    {
      "epoch": 0.36275,
      "grad_norm": 32.5,
      "grad_norm_var": 1.7400390625,
      "learning_rate": 0.0001,
      "loss": 7.3819,
      "loss/crossentropy": 2.2282170712947846,
      "loss/hidden": 3.43359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20617649108171462,
      "step": 14510
    },
    {
      "epoch": 0.363,
      "grad_norm": 29.0,
      "grad_norm_var": 26.622330729166666,
      "learning_rate": 0.0001,
      "loss": 7.3494,
      "loss/crossentropy": 2.1207677230238913,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18537767957895995,
      "step": 14520
    },
    {
      "epoch": 0.36325,
      "grad_norm": 28.5,
      "grad_norm_var": 1.15,
      "learning_rate": 0.0001,
      "loss": 7.2906,
      "loss/crossentropy": 2.1179309889674185,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18408581465482712,
      "step": 14530
    },
    {
      "epoch": 0.3635,
      "grad_norm": 31.125,
      "grad_norm_var": 2.851822916666667,
      "learning_rate": 0.0001,
      "loss": 7.3786,
      "loss/crossentropy": 2.0486052967607975,
      "loss/hidden": 3.35859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17906265445053576,
      "step": 14540
    },
    {
      "epoch": 0.36375,
      "grad_norm": 33.0,
      "grad_norm_var": 3.184830729166667,
      "learning_rate": 0.0001,
      "loss": 7.3423,
      "loss/crossentropy": 2.301698251068592,
      "loss/hidden": 3.3765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19077361291274428,
      "step": 14550
    },
    {
      "epoch": 0.364,
      "grad_norm": 33.25,
      "grad_norm_var": 2.2134765625,
      "learning_rate": 0.0001,
      "loss": 7.3047,
      "loss/crossentropy": 2.074369618296623,
      "loss/hidden": 3.293359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17446096520870924,
      "step": 14560
    },
    {
      "epoch": 0.36425,
      "grad_norm": 32.5,
      "grad_norm_var": 2.6910807291666665,
      "learning_rate": 0.0001,
      "loss": 7.412,
      "loss/crossentropy": 2.070549990236759,
      "loss/hidden": 3.501953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2114204354584217,
      "step": 14570
    },
    {
      "epoch": 0.3645,
      "grad_norm": 31.875,
      "grad_norm_var": 31.843489583333334,
      "learning_rate": 0.0001,
      "loss": 7.4461,
      "loss/crossentropy": 2.2021229028701783,
      "loss/hidden": 3.412890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1946948293596506,
      "step": 14580
    },
    {
      "epoch": 0.36475,
      "grad_norm": 28.875,
      "grad_norm_var": 32.17916666666667,
      "learning_rate": 0.0001,
      "loss": 7.33,
      "loss/crossentropy": 2.07574619948864,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.199923849850893,
      "step": 14590
    },
    {
      "epoch": 0.365,
      "grad_norm": 29.625,
      "grad_norm_var": 1.3020833333333333,
      "learning_rate": 0.0001,
      "loss": 7.3285,
      "loss/crossentropy": 2.236181080341339,
      "loss/hidden": 3.295703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18559221122413874,
      "step": 14600
    },
    {
      "epoch": 0.36525,
      "grad_norm": 27.875,
      "grad_norm_var": 1.6979166666666667,
      "learning_rate": 0.0001,
      "loss": 7.4122,
      "loss/crossentropy": 2.029696011543274,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18932779133319855,
      "step": 14610
    },
    {
      "epoch": 0.3655,
      "grad_norm": 29.375,
      "grad_norm_var": 3.17265625,
      "learning_rate": 0.0001,
      "loss": 7.3296,
      "loss/crossentropy": 2.0676317110657694,
      "loss/hidden": 3.308984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1811735849827528,
      "step": 14620
    },
    {
      "epoch": 0.36575,
      "grad_norm": 31.5,
      "grad_norm_var": 3.8629557291666665,
      "learning_rate": 0.0001,
      "loss": 7.3105,
      "loss/crossentropy": 2.1635128699243067,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19529270604252816,
      "step": 14630
    },
    {
      "epoch": 0.366,
      "grad_norm": 28.25,
      "grad_norm_var": 3.3955729166666666,
      "learning_rate": 0.0001,
      "loss": 7.3913,
      "loss/crossentropy": 2.0359733670949938,
      "loss/hidden": 3.5,
      "loss/jsd": 0.0,
      "loss/logits": 0.20129154790192844,
      "step": 14640
    },
    {
      "epoch": 0.36625,
      "grad_norm": 30.5,
      "grad_norm_var": 3.72265625,
      "learning_rate": 0.0001,
      "loss": 7.3418,
      "loss/crossentropy": 1.895400796085596,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17521814415231346,
      "step": 14650
    },
    {
      "epoch": 0.3665,
      "grad_norm": 30.375,
      "grad_norm_var": 1.7302083333333333,
      "learning_rate": 0.0001,
      "loss": 7.3042,
      "loss/crossentropy": 2.0394834615290165,
      "loss/hidden": 3.21953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17233089366927742,
      "step": 14660
    },
    {
      "epoch": 0.36675,
      "grad_norm": 28.25,
      "grad_norm_var": 36.44479166666667,
      "learning_rate": 0.0001,
      "loss": 7.3335,
      "loss/crossentropy": 1.8397274687886238,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17788148634135723,
      "step": 14670
    },
    {
      "epoch": 0.367,
      "grad_norm": 27.875,
      "grad_norm_var": 38.53639322916667,
      "learning_rate": 0.0001,
      "loss": 7.3786,
      "loss/crossentropy": 2.2247194588184356,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19322688207030297,
      "step": 14680
    },
    {
      "epoch": 0.36725,
      "grad_norm": 40.0,
      "grad_norm_var": 13.45390625,
      "learning_rate": 0.0001,
      "loss": 7.3179,
      "loss/crossentropy": 1.9937927357852459,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17893323805183173,
      "step": 14690
    },
    {
      "epoch": 0.3675,
      "grad_norm": 32.75,
      "grad_norm_var": 11.25,
      "learning_rate": 0.0001,
      "loss": 7.2325,
      "loss/crossentropy": 2.0396489590406417,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1912191865965724,
      "step": 14700
    },
    {
      "epoch": 0.36775,
      "grad_norm": 33.0,
      "grad_norm_var": 7.471875,
      "learning_rate": 0.0001,
      "loss": 7.4509,
      "loss/crossentropy": 2.1277743950486183,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20408784411847591,
      "step": 14710
    },
    {
      "epoch": 0.368,
      "grad_norm": 30.75,
      "grad_norm_var": 8.546875,
      "learning_rate": 0.0001,
      "loss": 7.3855,
      "loss/crossentropy": 2.2228190809488297,
      "loss/hidden": 3.362890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1884877322241664,
      "step": 14720
    },
    {
      "epoch": 0.36825,
      "grad_norm": 28.625,
      "grad_norm_var": 15.142708333333333,
      "learning_rate": 0.0001,
      "loss": 7.4076,
      "loss/crossentropy": 2.1207469016313554,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17780339773744344,
      "step": 14730
    },
    {
      "epoch": 0.3685,
      "grad_norm": 31.125,
      "grad_norm_var": 3.004622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3209,
      "loss/crossentropy": 2.0066810354590414,
      "loss/hidden": 3.519921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20999562088400126,
      "step": 14740
    },
    {
      "epoch": 0.36875,
      "grad_norm": 38.25,
      "grad_norm_var": 15.4259765625,
      "learning_rate": 0.0001,
      "loss": 7.4197,
      "loss/crossentropy": 2.0302488803863525,
      "loss/hidden": 3.491796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1997592320665717,
      "step": 14750
    },
    {
      "epoch": 0.369,
      "grad_norm": 32.5,
      "grad_norm_var": 22.669791666666665,
      "learning_rate": 0.0001,
      "loss": 7.3414,
      "loss/crossentropy": 2.006557123363018,
      "loss/hidden": 3.333203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18148684445768595,
      "step": 14760
    },
    {
      "epoch": 0.36925,
      "grad_norm": 29.5,
      "grad_norm_var": 22.362434895833335,
      "learning_rate": 0.0001,
      "loss": 7.347,
      "loss/crossentropy": 2.07253782749176,
      "loss/hidden": 3.274609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17341973297297955,
      "step": 14770
    },
    {
      "epoch": 0.3695,
      "grad_norm": 30.75,
      "grad_norm_var": 17.227018229166667,
      "learning_rate": 0.0001,
      "loss": 7.3839,
      "loss/crossentropy": 2.056871312856674,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21048452276736498,
      "step": 14780
    },
    {
      "epoch": 0.36975,
      "grad_norm": 27.875,
      "grad_norm_var": 38.5947265625,
      "learning_rate": 0.0001,
      "loss": 7.4979,
      "loss/crossentropy": 2.1910998940467836,
      "loss/hidden": 3.480859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20320441015064716,
      "step": 14790
    },
    {
      "epoch": 0.37,
      "grad_norm": 31.25,
      "grad_norm_var": 2.081788902243793e+18,
      "learning_rate": 0.0001,
      "loss": 7.3396,
      "loss/crossentropy": 2.0854848250746727,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19095470141619444,
      "step": 14800
    },
    {
      "epoch": 0.37025,
      "grad_norm": 34.75,
      "grad_norm_var": 2.0817889032778286e+18,
      "learning_rate": 0.0001,
      "loss": 7.349,
      "loss/crossentropy": 2.024295690655708,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2073504414409399,
      "step": 14810
    },
    {
      "epoch": 0.3705,
      "grad_norm": 30.375,
      "grad_norm_var": 8.116080729166667,
      "learning_rate": 0.0001,
      "loss": 7.1984,
      "loss/crossentropy": 1.9738368421792984,
      "loss/hidden": 3.352734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.175398519821465,
      "step": 14820
    },
    {
      "epoch": 0.37075,
      "grad_norm": 34.0,
      "grad_norm_var": 3.95390625,
      "learning_rate": 0.0001,
      "loss": 7.3269,
      "loss/crossentropy": 2.0109583623707294,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1902273640036583,
      "step": 14830
    },
    {
      "epoch": 0.371,
      "grad_norm": 30.25,
      "grad_norm_var": 3.410416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3309,
      "loss/crossentropy": 2.0899639569222925,
      "loss/hidden": 3.444921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18828139845281838,
      "step": 14840
    },
    {
      "epoch": 0.37125,
      "grad_norm": 31.375,
      "grad_norm_var": 51.5087890625,
      "learning_rate": 0.0001,
      "loss": 7.2761,
      "loss/crossentropy": 2.0561746567487718,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17576684867963194,
      "step": 14850
    },
    {
      "epoch": 0.3715,
      "grad_norm": 32.25,
      "grad_norm_var": 54.873372395833336,
      "learning_rate": 0.0001,
      "loss": 7.2875,
      "loss/crossentropy": 2.1231576301157475,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1990961253643036,
      "step": 14860
    },
    {
      "epoch": 0.37175,
      "grad_norm": 34.25,
      "grad_norm_var": 29.984309895833334,
      "learning_rate": 0.0001,
      "loss": 7.3424,
      "loss/crossentropy": 2.1051357120275496,
      "loss/hidden": 3.29453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17979057859629394,
      "step": 14870
    },
    {
      "epoch": 0.372,
      "grad_norm": 28.25,
      "grad_norm_var": 19850.327018229167,
      "learning_rate": 0.0001,
      "loss": 7.3823,
      "loss/crossentropy": 2.3242007076740263,
      "loss/hidden": 3.49609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2740444682538509,
      "step": 14880
    },
    {
      "epoch": 0.37225,
      "grad_norm": 27.25,
      "grad_norm_var": 35.0625,
      "learning_rate": 0.0001,
      "loss": 7.2273,
      "loss/crossentropy": 2.143503928184509,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19005000293254853,
      "step": 14890
    },
    {
      "epoch": 0.3725,
      "grad_norm": 34.0,
      "grad_norm_var": 38.23125,
      "learning_rate": 0.0001,
      "loss": 7.2655,
      "loss/crossentropy": 2.1128626547753813,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18479138296097516,
      "step": 14900
    },
    {
      "epoch": 0.37275,
      "grad_norm": 31.0,
      "grad_norm_var": 26.7478515625,
      "learning_rate": 0.0001,
      "loss": 7.3153,
      "loss/crossentropy": 1.860854334384203,
      "loss/hidden": 3.255859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1592037882655859,
      "step": 14910
    },
    {
      "epoch": 0.373,
      "grad_norm": 29.625,
      "grad_norm_var": 30.737239583333334,
      "learning_rate": 0.0001,
      "loss": 7.287,
      "loss/crossentropy": 2.067621612548828,
      "loss/hidden": 3.468359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897974604740739,
      "step": 14920
    },
    {
      "epoch": 0.37325,
      "grad_norm": 29.0,
      "grad_norm_var": 51.01223958333333,
      "learning_rate": 0.0001,
      "loss": 7.4038,
      "loss/crossentropy": 2.0471679329872132,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18869173359125851,
      "step": 14930
    },
    {
      "epoch": 0.3735,
      "grad_norm": 40.25,
      "grad_norm_var": 42.828059895833334,
      "learning_rate": 0.0001,
      "loss": 7.2638,
      "loss/crossentropy": 2.0969312518835066,
      "loss/hidden": 3.308984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19103059088811278,
      "step": 14940
    },
    {
      "epoch": 0.37375,
      "grad_norm": 27.625,
      "grad_norm_var": 16.3625,
      "learning_rate": 0.0001,
      "loss": 7.3022,
      "loss/crossentropy": 2.208647185564041,
      "loss/hidden": 3.323046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1936396975070238,
      "step": 14950
    },
    {
      "epoch": 0.374,
      "grad_norm": 28.625,
      "grad_norm_var": 16.837239583333332,
      "learning_rate": 0.0001,
      "loss": 7.2414,
      "loss/crossentropy": 2.1974997609853744,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19469811543822288,
      "step": 14960
    },
    {
      "epoch": 0.37425,
      "grad_norm": 29.625,
      "grad_norm_var": 14.108072916666666,
      "learning_rate": 0.0001,
      "loss": 7.2849,
      "loss/crossentropy": 2.134792809188366,
      "loss/hidden": 3.20625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17098608147352934,
      "step": 14970
    },
    {
      "epoch": 0.3745,
      "grad_norm": 33.25,
      "grad_norm_var": 27.1181640625,
      "learning_rate": 0.0001,
      "loss": 7.3014,
      "loss/crossentropy": 2.1276439666748046,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18990135621279478,
      "step": 14980
    },
    {
      "epoch": 0.37475,
      "grad_norm": 29.75,
      "grad_norm_var": 16.033072916666665,
      "learning_rate": 0.0001,
      "loss": 7.3335,
      "loss/crossentropy": 2.0856949634850026,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18325317203998565,
      "step": 14990
    },
    {
      "epoch": 0.375,
      "grad_norm": 31.625,
      "grad_norm_var": 3.321875,
      "learning_rate": 0.0001,
      "loss": 7.3642,
      "loss/crossentropy": 2.1136090487241743,
      "loss/hidden": 3.523828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20829910095781087,
      "step": 15000
    }
  ],
  "logging_steps": 10,
  "max_steps": 40000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 9223372036854775807,
  "save_steps": 5000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 4.28626504801321e+19,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}