{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.25,
  "eval_steps": 2000,
  "global_step": 10000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.00025,
      "grad_norm": 32.5,
      "learning_rate": 0.0001,
      "loss": 7.9852,
      "loss/crossentropy": 2.2558943748474123,
      "loss/hidden": 3.53671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22032691352069378,
      "step": 10
    },
    {
      "epoch": 0.0005,
      "grad_norm": 39.0,
      "grad_norm_var": 6.1306640625,
      "learning_rate": 0.0001,
      "loss": 8.0827,
      "loss/crossentropy": 2.219619666039944,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20763051956892015,
      "step": 20
    },
    {
      "epoch": 0.00075,
      "grad_norm": 30.375,
      "grad_norm_var": 7.383072916666666,
      "learning_rate": 0.0001,
      "loss": 7.8479,
      "loss/crossentropy": 2.185364603996277,
      "loss/hidden": 3.59296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23391152992844583,
      "step": 30
    },
    {
      "epoch": 0.001,
      "grad_norm": 32.75,
      "grad_norm_var": 187.0322265625,
      "learning_rate": 0.0001,
      "loss": 7.8341,
      "loss/crossentropy": 2.083733668923378,
      "loss/hidden": 3.375390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2040597340092063,
      "step": 40
    },
    {
      "epoch": 0.00125,
      "grad_norm": 32.5,
      "grad_norm_var": 5.111393229166667,
      "learning_rate": 0.0001,
      "loss": 7.6815,
      "loss/crossentropy": 2.182037726044655,
      "loss/hidden": 3.439453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20423058047890663,
      "step": 50
    },
    {
      "epoch": 0.0015,
      "grad_norm": 33.0,
      "grad_norm_var": 1.3853515625,
      "learning_rate": 0.0001,
      "loss": 7.6919,
      "loss/crossentropy": 2.1419573068618774,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19463330563157796,
      "step": 60
    },
    {
      "epoch": 0.00175,
      "grad_norm": 33.25,
      "grad_norm_var": 2.228125,
      "learning_rate": 0.0001,
      "loss": 7.838,
      "loss/crossentropy": 2.242894622683525,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22062199115753173,
      "step": 70
    },
    {
      "epoch": 0.002,
      "grad_norm": 32.5,
      "grad_norm_var": 8.371809895833334,
      "learning_rate": 0.0001,
      "loss": 8.018,
      "loss/crossentropy": 2.0408543169498445,
      "loss/hidden": 3.506640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20080858804285526,
      "step": 80
    },
    {
      "epoch": 0.00225,
      "grad_norm": 32.75,
      "grad_norm_var": 7.322916666666667,
      "learning_rate": 0.0001,
      "loss": 7.8807,
      "loss/crossentropy": 2.0654082030057905,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2109438929706812,
      "step": 90
    },
    {
      "epoch": 0.0025,
      "grad_norm": 34.5,
      "grad_norm_var": 2805.3603515625,
      "learning_rate": 0.0001,
      "loss": 8.0497,
      "loss/crossentropy": 2.0930048365145923,
      "loss/hidden": 3.410546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2103888330049813,
      "step": 100
    },
    {
      "epoch": 0.00275,
      "grad_norm": 33.0,
      "grad_norm_var": 2798.430143229167,
      "learning_rate": 0.0001,
      "loss": 7.8583,
      "loss/crossentropy": 2.3308374524116515,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2045755073428154,
      "step": 110
    },
    {
      "epoch": 0.003,
      "grad_norm": 34.0,
      "grad_norm_var": 1.6999348958333333,
      "learning_rate": 0.0001,
      "loss": 7.7088,
      "loss/crossentropy": 2.120428466796875,
      "loss/hidden": 3.56796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21561774536967276,
      "step": 120
    },
    {
      "epoch": 0.00325,
      "grad_norm": 31.0,
      "grad_norm_var": 40.73899739583333,
      "learning_rate": 0.0001,
      "loss": 7.762,
      "loss/crossentropy": 2.1464689180254934,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2084670951589942,
      "step": 130
    },
    {
      "epoch": 0.0035,
      "grad_norm": 33.0,
      "grad_norm_var": 42.67265625,
      "learning_rate": 0.0001,
      "loss": 7.711,
      "loss/crossentropy": 2.1301105961203577,
      "loss/hidden": 3.512890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21644905991852284,
      "step": 140
    },
    {
      "epoch": 0.00375,
      "grad_norm": 32.75,
      "grad_norm_var": 8.066080729166666,
      "learning_rate": 0.0001,
      "loss": 7.7295,
      "loss/crossentropy": 2.115240353345871,
      "loss/hidden": 3.491796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2064087452366948,
      "step": 150
    },
    {
      "epoch": 0.004,
      "grad_norm": 32.5,
      "grad_norm_var": 422.72057291666664,
      "learning_rate": 0.0001,
      "loss": 7.9367,
      "loss/crossentropy": 2.141779786348343,
      "loss/hidden": 3.457421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21747791785746812,
      "step": 160
    },
    {
      "epoch": 0.00425,
      "grad_norm": 32.0,
      "grad_norm_var": 6.705208333333333,
      "learning_rate": 0.0001,
      "loss": 7.756,
      "loss/crossentropy": 2.103468084335327,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2027540819719434,
      "step": 170
    },
    {
      "epoch": 0.0045,
      "grad_norm": 32.25,
      "grad_norm_var": 17.8556640625,
      "learning_rate": 0.0001,
      "loss": 7.7008,
      "loss/crossentropy": 2.1950977832078933,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2186649737879634,
      "step": 180
    },
    {
      "epoch": 0.00475,
      "grad_norm": 30.5,
      "grad_norm_var": 66.61608072916667,
      "learning_rate": 0.0001,
      "loss": 7.6552,
      "loss/crossentropy": 2.029752139747143,
      "loss/hidden": 3.585546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22110425475984813,
      "step": 190
    },
    {
      "epoch": 0.005,
      "grad_norm": 36.0,
      "grad_norm_var": 13.160872395833334,
      "learning_rate": 0.0001,
      "loss": 7.6964,
      "loss/crossentropy": 2.169684535264969,
      "loss/hidden": 3.52265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22122678495943546,
      "step": 200
    },
    {
      "epoch": 0.00525,
      "grad_norm": 33.25,
      "grad_norm_var": 3.442643229166667,
      "learning_rate": 0.0001,
      "loss": 7.7135,
      "loss/crossentropy": 2.1366169154644012,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19148335698992014,
      "step": 210
    },
    {
      "epoch": 0.0055,
      "grad_norm": 35.25,
      "grad_norm_var": 4.151822916666666,
      "learning_rate": 0.0001,
      "loss": 7.654,
      "loss/crossentropy": 2.196457767486572,
      "loss/hidden": 3.573828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22240160517394542,
      "step": 220
    },
    {
      "epoch": 0.00575,
      "grad_norm": 34.5,
      "grad_norm_var": 61.07057291666667,
      "learning_rate": 0.0001,
      "loss": 7.7286,
      "loss/crossentropy": 2.099469523131847,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19720815271139144,
      "step": 230
    },
    {
      "epoch": 0.006,
      "grad_norm": 33.75,
      "grad_norm_var": 5.479166666666667,
      "learning_rate": 0.0001,
      "loss": 7.6841,
      "loss/crossentropy": 2.0730464071035386,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20290055498480797,
      "step": 240
    },
    {
      "epoch": 0.00625,
      "grad_norm": 29.5,
      "grad_norm_var": 4.9197265625,
      "learning_rate": 0.0001,
      "loss": 7.6999,
      "loss/crossentropy": 2.1514100462198256,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2013280361890793,
      "step": 250
    },
    {
      "epoch": 0.0065,
      "grad_norm": 33.25,
      "grad_norm_var": 150.86712239583332,
      "learning_rate": 0.0001,
      "loss": 7.7617,
      "loss/crossentropy": 2.0071112543344496,
      "loss/hidden": 3.39921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1923872010782361,
      "step": 260
    },
    {
      "epoch": 0.00675,
      "grad_norm": 32.25,
      "grad_norm_var": 41.373958333333334,
      "learning_rate": 0.0001,
      "loss": 7.7968,
      "loss/crossentropy": 2.254330241680145,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20844143405556678,
      "step": 270
    },
    {
      "epoch": 0.007,
      "grad_norm": 30.25,
      "grad_norm_var": 5.9625,
      "learning_rate": 0.0001,
      "loss": 7.7437,
      "loss/crossentropy": 2.2058747708797455,
      "loss/hidden": 3.406640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19603454861789943,
      "step": 280
    },
    {
      "epoch": 0.00725,
      "grad_norm": 32.0,
      "grad_norm_var": 3.4306640625,
      "learning_rate": 0.0001,
      "loss": 7.7308,
      "loss/crossentropy": 2.158563455939293,
      "loss/hidden": 3.485546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20696840062737465,
      "step": 290
    },
    {
      "epoch": 0.0075,
      "grad_norm": 35.25,
      "grad_norm_var": 4.163997395833333,
      "learning_rate": 0.0001,
      "loss": 7.7781,
      "loss/crossentropy": 2.1728759124875068,
      "loss/hidden": 3.6984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2232737574726343,
      "step": 300
    },
    {
      "epoch": 0.00775,
      "grad_norm": 32.25,
      "grad_norm_var": 9.7212890625,
      "learning_rate": 0.0001,
      "loss": 7.7039,
      "loss/crossentropy": 2.1304744452238085,
      "loss/hidden": 3.510546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21460633352398872,
      "step": 310
    },
    {
      "epoch": 0.008,
      "grad_norm": 34.25,
      "grad_norm_var": 102.21979166666667,
      "learning_rate": 0.0001,
      "loss": 7.7386,
      "loss/crossentropy": 2.2150494754314423,
      "loss/hidden": 3.470703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2204894032329321,
      "step": 320
    },
    {
      "epoch": 0.00825,
      "grad_norm": 34.5,
      "grad_norm_var": 103.0806640625,
      "learning_rate": 0.0001,
      "loss": 7.6843,
      "loss/crossentropy": 2.1483607694506643,
      "loss/hidden": 3.525,
      "loss/jsd": 0.0,
      "loss/logits": 0.2168185070157051,
      "step": 330
    },
    {
      "epoch": 0.0085,
      "grad_norm": 29.75,
      "grad_norm_var": 4.245768229166667,
      "learning_rate": 0.0001,
      "loss": 7.6824,
      "loss/crossentropy": 2.2668254554271696,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1978676740080118,
      "step": 340
    },
    {
      "epoch": 0.00875,
      "grad_norm": 30.75,
      "grad_norm_var": 7.785416666666666,
      "learning_rate": 0.0001,
      "loss": 7.6476,
      "loss/crossentropy": 2.247351437807083,
      "loss/hidden": 3.45078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2044550308957696,
      "step": 350
    },
    {
      "epoch": 0.009,
      "grad_norm": 33.0,
      "grad_norm_var": 36.1150390625,
      "learning_rate": 0.0001,
      "loss": 7.6707,
      "loss/crossentropy": 2.2508403569459916,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20586735829710961,
      "step": 360
    },
    {
      "epoch": 0.00925,
      "grad_norm": 27.875,
      "grad_norm_var": 42.02805989583333,
      "learning_rate": 0.0001,
      "loss": 7.787,
      "loss/crossentropy": 2.2524363905191422,
      "loss/hidden": 3.39921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20639074668288232,
      "step": 370
    },
    {
      "epoch": 0.0095,
      "grad_norm": 29.625,
      "grad_norm_var": 19.65625,
      "learning_rate": 0.0001,
      "loss": 7.7258,
      "loss/crossentropy": 2.2901588469743728,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20220000632107257,
      "step": 380
    },
    {
      "epoch": 0.00975,
      "grad_norm": 29.875,
      "grad_norm_var": 15.653059895833334,
      "learning_rate": 0.0001,
      "loss": 7.6445,
      "loss/crossentropy": 2.1256834477186204,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.190180828794837,
      "step": 390
    },
    {
      "epoch": 0.01,
      "grad_norm": 33.5,
      "grad_norm_var": 4.7712890625,
      "learning_rate": 0.0001,
      "loss": 7.7719,
      "loss/crossentropy": 2.2237626880407335,
      "loss/hidden": 3.484765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2264560218900442,
      "step": 400
    },
    {
      "epoch": 0.01025,
      "grad_norm": 33.0,
      "grad_norm_var": 2.9247395833333334,
      "learning_rate": 0.0001,
      "loss": 7.7631,
      "loss/crossentropy": 2.060544753074646,
      "loss/hidden": 3.470703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2202487275004387,
      "step": 410
    },
    {
      "epoch": 0.0105,
      "grad_norm": 32.75,
      "grad_norm_var": 2.6712890625,
      "learning_rate": 0.0001,
      "loss": 7.8193,
      "loss/crossentropy": 2.270045906305313,
      "loss/hidden": 3.49921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22449300419539214,
      "step": 420
    },
    {
      "epoch": 0.01075,
      "grad_norm": 40.0,
      "grad_norm_var": 5.551822916666667,
      "learning_rate": 0.0001,
      "loss": 7.7502,
      "loss/crossentropy": 2.220390594005585,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20708895958960055,
      "step": 430
    },
    {
      "epoch": 0.011,
      "grad_norm": 33.75,
      "grad_norm_var": 6.554622395833333,
      "learning_rate": 0.0001,
      "loss": 7.6386,
      "loss/crossentropy": 2.0412577211856844,
      "loss/hidden": 3.52109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20780573673546315,
      "step": 440
    },
    {
      "epoch": 0.01125,
      "grad_norm": 39.5,
      "grad_norm_var": 9.018684895833333,
      "learning_rate": 0.0001,
      "loss": 7.7522,
      "loss/crossentropy": 2.085190561413765,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22059339918196202,
      "step": 450
    },
    {
      "epoch": 0.0115,
      "grad_norm": 31.75,
      "grad_norm_var": 5.703580729166666,
      "learning_rate": 0.0001,
      "loss": 7.604,
      "loss/crossentropy": 1.9451062515378,
      "loss/hidden": 3.527734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2082566052675247,
      "step": 460
    },
    {
      "epoch": 0.01175,
      "grad_norm": 34.25,
      "grad_norm_var": 1.4416666666666667,
      "learning_rate": 0.0001,
      "loss": 7.7286,
      "loss/crossentropy": 2.23127267062664,
      "loss/hidden": 3.560546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22645943984389305,
      "step": 470
    },
    {
      "epoch": 0.012,
      "grad_norm": 43.5,
      "grad_norm_var": 61.358333333333334,
      "learning_rate": 0.0001,
      "loss": 7.5857,
      "loss/crossentropy": 2.0275631666183473,
      "loss/hidden": 3.341015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19867698503658177,
      "step": 480
    },
    {
      "epoch": 0.01225,
      "grad_norm": 30.5,
      "grad_norm_var": 62.425455729166664,
      "learning_rate": 0.0001,
      "loss": 7.7023,
      "loss/crossentropy": 2.1564531326293945,
      "loss/hidden": 3.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21295207217335702,
      "step": 490
    },
    {
      "epoch": 0.0125,
      "grad_norm": 39.0,
      "grad_norm_var": 7.580208333333333,
      "learning_rate": 0.0001,
      "loss": 7.8035,
      "loss/crossentropy": 2.1716607972979545,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21046865545213223,
      "step": 500
    },
    {
      "epoch": 0.01275,
      "grad_norm": 35.25,
      "grad_norm_var": 5.121809895833334,
      "learning_rate": 0.0001,
      "loss": 7.764,
      "loss/crossentropy": 2.181091034412384,
      "loss/hidden": 3.410546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20339491367340087,
      "step": 510
    },
    {
      "epoch": 0.013,
      "grad_norm": 31.5,
      "grad_norm_var": 3.820572916666667,
      "learning_rate": 0.0001,
      "loss": 7.6112,
      "loss/crossentropy": 2.012446442246437,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20432959645986556,
      "step": 520
    },
    {
      "epoch": 0.01325,
      "grad_norm": 33.0,
      "grad_norm_var": 2.92890625,
      "learning_rate": 0.0001,
      "loss": 7.73,
      "loss/crossentropy": 2.0826023176312445,
      "loss/hidden": 3.51171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21190985422581435,
      "step": 530
    },
    {
      "epoch": 0.0135,
      "grad_norm": 31.375,
      "grad_norm_var": 2.0229166666666667,
      "learning_rate": 0.0001,
      "loss": 7.7527,
      "loss/crossentropy": 2.191486455500126,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1943425141274929,
      "step": 540
    },
    {
      "epoch": 0.01375,
      "grad_norm": 29.75,
      "grad_norm_var": 4.983268229166667,
      "learning_rate": 0.0001,
      "loss": 7.7035,
      "loss/crossentropy": 2.0664332896471023,
      "loss/hidden": 3.475,
      "loss/jsd": 0.0,
      "loss/logits": 0.20404404532164336,
      "step": 550
    },
    {
      "epoch": 0.014,
      "grad_norm": 31.875,
      "grad_norm_var": 3.381705729166667,
      "learning_rate": 0.0001,
      "loss": 7.5679,
      "loss/crossentropy": 2.0470397621393204,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21556914187967777,
      "step": 560
    },
    {
      "epoch": 0.01425,
      "grad_norm": 33.0,
      "grad_norm_var": 2.508072916666667,
      "learning_rate": 0.0001,
      "loss": 7.7909,
      "loss/crossentropy": 2.21784293949604,
      "loss/hidden": 3.4671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20632803943008185,
      "step": 570
    },
    {
      "epoch": 0.0145,
      "grad_norm": 32.25,
      "grad_norm_var": 2.6837890625,
      "learning_rate": 0.0001,
      "loss": 7.6449,
      "loss/crossentropy": 2.291385439038277,
      "loss/hidden": 3.432421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19818231668323277,
      "step": 580
    },
    {
      "epoch": 0.01475,
      "grad_norm": 30.875,
      "grad_norm_var": 106.909375,
      "learning_rate": 0.0001,
      "loss": 7.7009,
      "loss/crossentropy": 2.102944087982178,
      "loss/hidden": 3.532421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20003035496920346,
      "step": 590
    },
    {
      "epoch": 0.015,
      "grad_norm": 33.0,
      "grad_norm_var": 11.816080729166666,
      "learning_rate": 0.0001,
      "loss": 7.5907,
      "loss/crossentropy": 2.239013722538948,
      "loss/hidden": 3.471484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22811597101390363,
      "step": 600
    },
    {
      "epoch": 0.01525,
      "grad_norm": 34.75,
      "grad_norm_var": 3.2811848958333334,
      "learning_rate": 0.0001,
      "loss": 7.6578,
      "loss/crossentropy": 2.1826944231986998,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19526719450950622,
      "step": 610
    },
    {
      "epoch": 0.0155,
      "grad_norm": 32.0,
      "grad_norm_var": 29.54140625,
      "learning_rate": 0.0001,
      "loss": 7.6592,
      "loss/crossentropy": 2.0491475805640222,
      "loss/hidden": 3.47109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19470291025936604,
      "step": 620
    },
    {
      "epoch": 0.01575,
      "grad_norm": 29.625,
      "grad_norm_var": 3.458268229166667,
      "learning_rate": 0.0001,
      "loss": 7.7237,
      "loss/crossentropy": 2.171899539232254,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19583625346422195,
      "step": 630
    },
    {
      "epoch": 0.016,
      "grad_norm": 41.5,
      "grad_norm_var": 835.4395182291667,
      "learning_rate": 0.0001,
      "loss": 7.6493,
      "loss/crossentropy": 2.019927790760994,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889604590833187,
      "step": 640
    },
    {
      "epoch": 0.01625,
      "grad_norm": 33.25,
      "grad_norm_var": 816.1582682291667,
      "learning_rate": 0.0001,
      "loss": 7.7628,
      "loss/crossentropy": 2.146556834876537,
      "loss/hidden": 3.53046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21917179077863694,
      "step": 650
    },
    {
      "epoch": 0.0165,
      "grad_norm": 34.0,
      "grad_norm_var": 13.574739583333333,
      "learning_rate": 0.0001,
      "loss": 7.6199,
      "loss/crossentropy": 2.2131199680268763,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20602547163143753,
      "step": 660
    },
    {
      "epoch": 0.01675,
      "grad_norm": 34.75,
      "grad_norm_var": 28.6125,
      "learning_rate": 0.0001,
      "loss": 7.6113,
      "loss/crossentropy": 2.0343705236911775,
      "loss/hidden": 3.37734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18553176671266555,
      "step": 670
    },
    {
      "epoch": 0.017,
      "grad_norm": 31.625,
      "grad_norm_var": 19.8369140625,
      "learning_rate": 0.0001,
      "loss": 7.6443,
      "loss/crossentropy": 2.1528817296028135,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20593271851539613,
      "step": 680
    },
    {
      "epoch": 0.01725,
      "grad_norm": 36.25,
      "grad_norm_var": 5.97265625,
      "learning_rate": 0.0001,
      "loss": 7.7476,
      "loss/crossentropy": 2.2202903479337692,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2035485502332449,
      "step": 690
    },
    {
      "epoch": 0.0175,
      "grad_norm": 31.5,
      "grad_norm_var": 7.5431640625,
      "learning_rate": 0.0001,
      "loss": 7.621,
      "loss/crossentropy": 2.0744683638215067,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2023961789906025,
      "step": 700
    },
    {
      "epoch": 0.01775,
      "grad_norm": 29.875,
      "grad_norm_var": 26.912239583333335,
      "learning_rate": 0.0001,
      "loss": 7.7058,
      "loss/crossentropy": 2.1098077327013014,
      "loss/hidden": 3.48359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20867060720920563,
      "step": 710
    },
    {
      "epoch": 0.018,
      "grad_norm": 34.0,
      "grad_norm_var": 5.291080729166667,
      "learning_rate": 0.0001,
      "loss": 7.5224,
      "loss/crossentropy": 2.0663078971207143,
      "loss/hidden": 3.439453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1999094202183187,
      "step": 720
    },
    {
      "epoch": 0.01825,
      "grad_norm": 36.75,
      "grad_norm_var": 34.112239583333334,
      "learning_rate": 0.0001,
      "loss": 7.5679,
      "loss/crossentropy": 2.264209559559822,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.213060562312603,
      "step": 730
    },
    {
      "epoch": 0.0185,
      "grad_norm": 31.875,
      "grad_norm_var": 31.756705729166665,
      "learning_rate": 0.0001,
      "loss": 7.6714,
      "loss/crossentropy": 2.128282290697098,
      "loss/hidden": 3.494921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21533375550061465,
      "step": 740
    },
    {
      "epoch": 0.01875,
      "grad_norm": 29.125,
      "grad_norm_var": 4.314518229166667,
      "learning_rate": 0.0001,
      "loss": 7.6825,
      "loss/crossentropy": 2.0806978911161425,
      "loss/hidden": 3.446875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19196727648377418,
      "step": 750
    },
    {
      "epoch": 0.019,
      "grad_norm": 31.625,
      "grad_norm_var": 9.520833333333334,
      "learning_rate": 0.0001,
      "loss": 7.6058,
      "loss/crossentropy": 2.2315777271986006,
      "loss/hidden": 3.484765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21225934717804193,
      "step": 760
    },
    {
      "epoch": 0.01925,
      "grad_norm": 33.25,
      "grad_norm_var": 29.381705729166665,
      "learning_rate": 0.0001,
      "loss": 7.8377,
      "loss/crossentropy": 2.200978134572506,
      "loss/hidden": 3.52890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22721426151692867,
      "step": 770
    },
    {
      "epoch": 0.0195,
      "grad_norm": 32.75,
      "grad_norm_var": 11.095247395833333,
      "learning_rate": 0.0001,
      "loss": 7.7305,
      "loss/crossentropy": 2.2799030035734176,
      "loss/hidden": 3.512890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2418980894610286,
      "step": 780
    },
    {
      "epoch": 0.01975,
      "grad_norm": 46.5,
      "grad_norm_var": 22.12265625,
      "learning_rate": 0.0001,
      "loss": 7.663,
      "loss/crossentropy": 2.0916543275117876,
      "loss/hidden": 3.4765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20638740565627814,
      "step": 790
    },
    {
      "epoch": 0.02,
      "grad_norm": 53.75,
      "grad_norm_var": 2.3053503983968586e+18,
      "learning_rate": 0.0001,
      "loss": 7.6743,
      "loss/crossentropy": 2.2010276943445204,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2011238183826208,
      "step": 800
    },
    {
      "epoch": 0.02025,
      "grad_norm": 34.75,
      "grad_norm_var": 2.3053503962269005e+18,
      "learning_rate": 0.0001,
      "loss": 7.7365,
      "loss/crossentropy": 2.2651585280895232,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22051467839628458,
      "step": 810
    },
    {
      "epoch": 0.0205,
      "grad_norm": 32.25,
      "grad_norm_var": 69.03854166666666,
      "learning_rate": 0.0001,
      "loss": 7.5788,
      "loss/crossentropy": 2.1270318403840065,
      "loss/hidden": 3.47109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20940354652702808,
      "step": 820
    },
    {
      "epoch": 0.02075,
      "grad_norm": 33.0,
      "grad_norm_var": 8.0603515625,
      "learning_rate": 0.0001,
      "loss": 7.6237,
      "loss/crossentropy": 2.1726820170879364,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20644128862768413,
      "step": 830
    },
    {
      "epoch": 0.021,
      "grad_norm": 33.5,
      "grad_norm_var": 9.085416666666667,
      "learning_rate": 0.0001,
      "loss": 7.455,
      "loss/crossentropy": 2.166350546479225,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20701032150536774,
      "step": 840
    },
    {
      "epoch": 0.02125,
      "grad_norm": 37.75,
      "grad_norm_var": 6.362239583333333,
      "learning_rate": 0.0001,
      "loss": 7.5725,
      "loss/crossentropy": 2.095318245887756,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20427223704755307,
      "step": 850
    },
    {
      "epoch": 0.0215,
      "grad_norm": 29.625,
      "grad_norm_var": 15.934830729166666,
      "learning_rate": 0.0001,
      "loss": 7.5728,
      "loss/crossentropy": 2.1999073296785356,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2012161746621132,
      "step": 860
    },
    {
      "epoch": 0.02175,
      "grad_norm": 30.5,
      "grad_norm_var": 5.681184895833334,
      "learning_rate": 0.0001,
      "loss": 7.6168,
      "loss/crossentropy": 2.2957967817783356,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20650502648204566,
      "step": 870
    },
    {
      "epoch": 0.022,
      "grad_norm": 30.0,
      "grad_norm_var": 10.9181640625,
      "learning_rate": 0.0001,
      "loss": 7.6929,
      "loss/crossentropy": 2.2093090921640397,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18699637930840254,
      "step": 880
    },
    {
      "epoch": 0.02225,
      "grad_norm": 35.0,
      "grad_norm_var": 11.192122395833334,
      "learning_rate": 0.0001,
      "loss": 7.5316,
      "loss/crossentropy": 2.0586251467466354,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21464286223053933,
      "step": 890
    },
    {
      "epoch": 0.0225,
      "grad_norm": 48.0,
      "grad_norm_var": 22.92265625,
      "learning_rate": 0.0001,
      "loss": 7.5857,
      "loss/crossentropy": 2.1724458605051042,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19797445200383662,
      "step": 900
    },
    {
      "epoch": 0.02275,
      "grad_norm": 33.0,
      "grad_norm_var": 38.4837890625,
      "learning_rate": 0.0001,
      "loss": 7.6184,
      "loss/crossentropy": 2.15934486836195,
      "loss/hidden": 3.357421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19419073052704333,
      "step": 910
    },
    {
      "epoch": 0.023,
      "grad_norm": 31.0,
      "grad_norm_var": 29.8869140625,
      "learning_rate": 0.0001,
      "loss": 7.6209,
      "loss/crossentropy": 2.1601561695337295,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20290262177586554,
      "step": 920
    },
    {
      "epoch": 0.02325,
      "grad_norm": 33.0,
      "grad_norm_var": 2.5143229166666665,
      "learning_rate": 0.0001,
      "loss": 7.5969,
      "loss/crossentropy": 2.184823766350746,
      "loss/hidden": 3.392578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2016730338335037,
      "step": 930
    },
    {
      "epoch": 0.0235,
      "grad_norm": 30.125,
      "grad_norm_var": 1.5764973958333333,
      "learning_rate": 0.0001,
      "loss": 7.5772,
      "loss/crossentropy": 2.2380657255649568,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19638443663716315,
      "step": 940
    },
    {
      "epoch": 0.02375,
      "grad_norm": 28.875,
      "grad_norm_var": 5.199934895833334,
      "learning_rate": 0.0001,
      "loss": 7.6348,
      "loss/crossentropy": 2.183762513846159,
      "loss/hidden": 3.4171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19700763542205096,
      "step": 950
    },
    {
      "epoch": 0.024,
      "grad_norm": 33.0,
      "grad_norm_var": 20.1666015625,
      "learning_rate": 0.0001,
      "loss": 7.6351,
      "loss/crossentropy": 2.1142914414405825,
      "loss/hidden": 3.46953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2082229983061552,
      "step": 960
    },
    {
      "epoch": 0.02425,
      "grad_norm": 31.5,
      "grad_norm_var": 18.738541666666666,
      "learning_rate": 0.0001,
      "loss": 7.6302,
      "loss/crossentropy": 2.315229868888855,
      "loss/hidden": 3.32734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20684304945170878,
      "step": 970
    },
    {
      "epoch": 0.0245,
      "grad_norm": 33.75,
      "grad_norm_var": 2.501822916666667,
      "learning_rate": 0.0001,
      "loss": 7.6816,
      "loss/crossentropy": 2.094369947910309,
      "loss/hidden": 3.59921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2428071454167366,
      "step": 980
    },
    {
      "epoch": 0.02475,
      "grad_norm": 29.25,
      "grad_norm_var": 4.06015625,
      "learning_rate": 0.0001,
      "loss": 7.6289,
      "loss/crossentropy": 2.183109185099602,
      "loss/hidden": 3.5,
      "loss/jsd": 0.0,
      "loss/logits": 0.21173047311604024,
      "step": 990
    },
    {
      "epoch": 0.025,
      "grad_norm": 33.5,
      "grad_norm_var": 2.9593098958333335,
      "learning_rate": 0.0001,
      "loss": 7.6417,
      "loss/crossentropy": 2.1311034083366396,
      "loss/hidden": 3.44765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1998819222673774,
      "step": 1000
    },
    {
      "epoch": 0.02525,
      "grad_norm": 31.0,
      "grad_norm_var": 4.510872395833333,
      "learning_rate": 0.0001,
      "loss": 7.6299,
      "loss/crossentropy": 2.1297010451555254,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20842864252626897,
      "step": 1010
    },
    {
      "epoch": 0.0255,
      "grad_norm": 30.375,
      "grad_norm_var": 3.9893229166666666,
      "learning_rate": 0.0001,
      "loss": 7.6593,
      "loss/crossentropy": 2.2224902719259263,
      "loss/hidden": 3.509375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21827217563986778,
      "step": 1020
    },
    {
      "epoch": 0.02575,
      "grad_norm": 38.25,
      "grad_norm_var": 6.010416666666667,
      "learning_rate": 0.0001,
      "loss": 7.6712,
      "loss/crossentropy": 2.1976612359285355,
      "loss/hidden": 3.506640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.210753770545125,
      "step": 1030
    },
    {
      "epoch": 0.026,
      "grad_norm": 33.5,
      "grad_norm_var": 9.435872395833334,
      "learning_rate": 0.0001,
      "loss": 7.748,
      "loss/crossentropy": 2.1889317661523817,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2081079863011837,
      "step": 1040
    },
    {
      "epoch": 0.02625,
      "grad_norm": 31.0,
      "grad_norm_var": 6.206184895833333,
      "learning_rate": 0.0001,
      "loss": 7.5803,
      "loss/crossentropy": 2.0802227064967154,
      "loss/hidden": 3.46796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19740718584507705,
      "step": 1050
    },
    {
      "epoch": 0.0265,
      "grad_norm": 31.625,
      "grad_norm_var": 1.3770182291666666,
      "learning_rate": 0.0001,
      "loss": 7.6098,
      "loss/crossentropy": 1.969551184773445,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1954148853197694,
      "step": 1060
    },
    {
      "epoch": 0.02675,
      "grad_norm": 30.75,
      "grad_norm_var": 2.09375,
      "learning_rate": 0.0001,
      "loss": 7.7826,
      "loss/crossentropy": 2.160974936187267,
      "loss/hidden": 3.581640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2183740811422467,
      "step": 1070
    },
    {
      "epoch": 0.027,
      "grad_norm": 30.625,
      "grad_norm_var": 2.6520182291666665,
      "learning_rate": 0.0001,
      "loss": 7.6186,
      "loss/crossentropy": 2.179084411263466,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19020346291363238,
      "step": 1080
    },
    {
      "epoch": 0.02725,
      "grad_norm": 56.5,
      "grad_norm_var": 48.91868489583333,
      "learning_rate": 0.0001,
      "loss": 7.7756,
      "loss/crossentropy": 2.1456103891134264,
      "loss/hidden": 3.5,
      "loss/jsd": 0.0,
      "loss/logits": 0.2178689869120717,
      "step": 1090
    },
    {
      "epoch": 0.0275,
      "grad_norm": 32.0,
      "grad_norm_var": 42.19557291666667,
      "learning_rate": 0.0001,
      "loss": 7.6714,
      "loss/crossentropy": 2.2156156271696092,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2013509316369891,
      "step": 1100
    },
    {
      "epoch": 0.02775,
      "grad_norm": 31.625,
      "grad_norm_var": 29.3775390625,
      "learning_rate": 0.0001,
      "loss": 7.6833,
      "loss/crossentropy": 2.0683740943670275,
      "loss/hidden": 3.5046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21320818569511174,
      "step": 1110
    },
    {
      "epoch": 0.028,
      "grad_norm": 27.5,
      "grad_norm_var": 35.483333333333334,
      "learning_rate": 0.0001,
      "loss": 7.7302,
      "loss/crossentropy": 2.098052313923836,
      "loss/hidden": 3.46484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19890500828623772,
      "step": 1120
    },
    {
      "epoch": 0.02825,
      "grad_norm": 29.0,
      "grad_norm_var": 10.8775390625,
      "learning_rate": 0.0001,
      "loss": 7.6716,
      "loss/crossentropy": 2.0999813921749593,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18992104195058346,
      "step": 1130
    },
    {
      "epoch": 0.0285,
      "grad_norm": 30.875,
      "grad_norm_var": 3.67890625,
      "learning_rate": 0.0001,
      "loss": 7.5401,
      "loss/crossentropy": 2.07411085665226,
      "loss/hidden": 3.433984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2018830729648471,
      "step": 1140
    },
    {
      "epoch": 0.02875,
      "grad_norm": 31.125,
      "grad_norm_var": 18.053580729166665,
      "learning_rate": 0.0001,
      "loss": 7.645,
      "loss/crossentropy": 2.0945233553647995,
      "loss/hidden": 3.4859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21366582233458759,
      "step": 1150
    },
    {
      "epoch": 0.029,
      "grad_norm": 31.25,
      "grad_norm_var": 16.978125,
      "learning_rate": 0.0001,
      "loss": 7.6514,
      "loss/crossentropy": 2.0980678737163543,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19850811325013637,
      "step": 1160
    },
    {
      "epoch": 0.02925,
      "grad_norm": 32.5,
      "grad_norm_var": 30.2556640625,
      "learning_rate": 0.0001,
      "loss": 7.6021,
      "loss/crossentropy": 2.155895306169987,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19983574748039246,
      "step": 1170
    },
    {
      "epoch": 0.0295,
      "grad_norm": 30.5,
      "grad_norm_var": 5.733072916666667,
      "learning_rate": 0.0001,
      "loss": 7.6343,
      "loss/crossentropy": 2.1906268298625946,
      "loss/hidden": 3.348046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19584416709840297,
      "step": 1180
    },
    {
      "epoch": 0.02975,
      "grad_norm": 32.5,
      "grad_norm_var": 3.89765625,
      "learning_rate": 0.0001,
      "loss": 7.7077,
      "loss/crossentropy": 2.163237012922764,
      "loss/hidden": 3.562109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21741114580072463,
      "step": 1190
    },
    {
      "epoch": 0.03,
      "grad_norm": 27.875,
      "grad_norm_var": 3.6639973958333334,
      "learning_rate": 0.0001,
      "loss": 7.6695,
      "loss/crossentropy": 2.1346954315900804,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19178961254656315,
      "step": 1200
    },
    {
      "epoch": 0.03025,
      "grad_norm": 30.5,
      "grad_norm_var": 24.9125,
      "learning_rate": 0.0001,
      "loss": 7.6108,
      "loss/crossentropy": 2.2493597716093063,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.209975734166801,
      "step": 1210
    },
    {
      "epoch": 0.0305,
      "grad_norm": 30.75,
      "grad_norm_var": 5.566080729166667,
      "learning_rate": 0.0001,
      "loss": 7.5902,
      "loss/crossentropy": 2.046968361735344,
      "loss/hidden": 3.451171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1867401722818613,
      "step": 1220
    },
    {
      "epoch": 0.03075,
      "grad_norm": 29.75,
      "grad_norm_var": 5.396875,
      "learning_rate": 0.0001,
      "loss": 7.6659,
      "loss/crossentropy": 2.0429708033800127,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18792454693466426,
      "step": 1230
    },
    {
      "epoch": 0.031,
      "grad_norm": 30.375,
      "grad_norm_var": 2.8827473958333334,
      "learning_rate": 0.0001,
      "loss": 7.5425,
      "loss/crossentropy": 2.124686148762703,
      "loss/hidden": 3.45625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23028194047510625,
      "step": 1240
    },
    {
      "epoch": 0.03125,
      "grad_norm": 33.0,
      "grad_norm_var": 2.220833333333333,
      "learning_rate": 0.0001,
      "loss": 7.7766,
      "loss/crossentropy": 2.1431034594774245,
      "loss/hidden": 3.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1852023523300886,
      "step": 1250
    },
    {
      "epoch": 0.0315,
      "grad_norm": 34.0,
      "grad_norm_var": 3.3080729166666667,
      "learning_rate": 0.0001,
      "loss": 7.6547,
      "loss/crossentropy": 2.208024913072586,
      "loss/hidden": 3.32265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19009452145546674,
      "step": 1260
    },
    {
      "epoch": 0.03175,
      "grad_norm": 31.375,
      "grad_norm_var": 3.7249348958333335,
      "learning_rate": 0.0001,
      "loss": 7.6331,
      "loss/crossentropy": 2.0774734795093535,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19277823474258185,
      "step": 1270
    },
    {
      "epoch": 0.032,
      "grad_norm": 32.75,
      "grad_norm_var": 2.6393229166666665,
      "learning_rate": 0.0001,
      "loss": 7.6147,
      "loss/crossentropy": 2.244540962576866,
      "loss/hidden": 3.444921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20150573179125786,
      "step": 1280
    },
    {
      "epoch": 0.03225,
      "grad_norm": 31.75,
      "grad_norm_var": 3.2280598958333333,
      "learning_rate": 0.0001,
      "loss": 7.6724,
      "loss/crossentropy": 2.1218355029821394,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19727950319647788,
      "step": 1290
    },
    {
      "epoch": 0.0325,
      "grad_norm": 33.25,
      "grad_norm_var": 3.6113932291666666,
      "learning_rate": 0.0001,
      "loss": 7.5881,
      "loss/crossentropy": 2.048227934539318,
      "loss/hidden": 3.352734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19401397118344904,
      "step": 1300
    },
    {
      "epoch": 0.03275,
      "grad_norm": 35.5,
      "grad_norm_var": 2.8384765625,
      "learning_rate": 0.0001,
      "loss": 7.6481,
      "loss/crossentropy": 2.0217724472284315,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19338970091193913,
      "step": 1310
    },
    {
      "epoch": 0.033,
      "grad_norm": 32.75,
      "grad_norm_var": 2.3671223958333334,
      "learning_rate": 0.0001,
      "loss": 7.5734,
      "loss/crossentropy": 2.123840129375458,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19704403057694436,
      "step": 1320
    },
    {
      "epoch": 0.03325,
      "grad_norm": 31.375,
      "grad_norm_var": 1.8207682291666667,
      "learning_rate": 0.0001,
      "loss": 7.5379,
      "loss/crossentropy": 2.1691371381282805,
      "loss/hidden": 3.416796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20379403475672006,
      "step": 1330
    },
    {
      "epoch": 0.0335,
      "grad_norm": 30.5,
      "grad_norm_var": 2.162239583333333,
      "learning_rate": 0.0001,
      "loss": 7.5824,
      "loss/crossentropy": 2.0320975854992867,
      "loss/hidden": 3.490625,
      "loss/jsd": 0.0,
      "loss/logits": 0.204788769595325,
      "step": 1340
    },
    {
      "epoch": 0.03375,
      "grad_norm": 29.875,
      "grad_norm_var": 29.69140625,
      "learning_rate": 0.0001,
      "loss": 7.6835,
      "loss/crossentropy": 2.1799038141965865,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21532316971570253,
      "step": 1350
    },
    {
      "epoch": 0.034,
      "grad_norm": 30.625,
      "grad_norm_var": 2.4837890625,
      "learning_rate": 0.0001,
      "loss": 7.6461,
      "loss/crossentropy": 2.075017270445824,
      "loss/hidden": 3.483984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21185822309926153,
      "step": 1360
    },
    {
      "epoch": 0.03425,
      "grad_norm": 29.375,
      "grad_norm_var": 2.6806640625,
      "learning_rate": 0.0001,
      "loss": 7.6084,
      "loss/crossentropy": 2.2061389327049254,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19864549599587916,
      "step": 1370
    },
    {
      "epoch": 0.0345,
      "grad_norm": 46.25,
      "grad_norm_var": 16.428125,
      "learning_rate": 0.0001,
      "loss": 7.5773,
      "loss/crossentropy": 2.1411263316869737,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18519967906177043,
      "step": 1380
    },
    {
      "epoch": 0.03475,
      "grad_norm": 33.5,
      "grad_norm_var": 312.2525390625,
      "learning_rate": 0.0001,
      "loss": 7.6651,
      "loss/crossentropy": 2.1760765284299852,
      "loss/hidden": 3.457421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19441114887595176,
      "step": 1390
    },
    {
      "epoch": 0.035,
      "grad_norm": 30.375,
      "grad_norm_var": 7.5697265625,
      "learning_rate": 0.0001,
      "loss": 7.6384,
      "loss/crossentropy": 2.148920811712742,
      "loss/hidden": 3.503515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20845398511737584,
      "step": 1400
    },
    {
      "epoch": 0.03525,
      "grad_norm": 31.5,
      "grad_norm_var": 3.2301432291666665,
      "learning_rate": 0.0001,
      "loss": 7.6503,
      "loss/crossentropy": 2.2241507709026336,
      "loss/hidden": 3.478515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2013774536550045,
      "step": 1410
    },
    {
      "epoch": 0.0355,
      "grad_norm": 32.75,
      "grad_norm_var": 3.67265625,
      "learning_rate": 0.0001,
      "loss": 7.6107,
      "loss/crossentropy": 2.151585003733635,
      "loss/hidden": 3.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19790932536125183,
      "step": 1420
    },
    {
      "epoch": 0.03575,
      "grad_norm": 31.0,
      "grad_norm_var": 2.192643229166667,
      "learning_rate": 0.0001,
      "loss": 7.6089,
      "loss/crossentropy": 2.106749549508095,
      "loss/hidden": 3.555078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20702882390469313,
      "step": 1430
    },
    {
      "epoch": 0.036,
      "grad_norm": 34.0,
      "grad_norm_var": 2.8622395833333334,
      "learning_rate": 0.0001,
      "loss": 7.6081,
      "loss/crossentropy": 2.0727885022759436,
      "loss/hidden": 3.425,
      "loss/jsd": 0.0,
      "loss/logits": 0.21057205237448215,
      "step": 1440
    },
    {
      "epoch": 0.03625,
      "grad_norm": 31.25,
      "grad_norm_var": 3.32890625,
      "learning_rate": 0.0001,
      "loss": 7.7005,
      "loss/crossentropy": 2.3014074742794035,
      "loss/hidden": 3.35859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20723759960383176,
      "step": 1450
    },
    {
      "epoch": 0.0365,
      "grad_norm": 28.625,
      "grad_norm_var": 4.240625,
      "learning_rate": 0.0001,
      "loss": 7.5677,
      "loss/crossentropy": 2.12650800794363,
      "loss/hidden": 3.44609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20236929692327976,
      "step": 1460
    },
    {
      "epoch": 0.03675,
      "grad_norm": 31.0,
      "grad_norm_var": 8.195768229166667,
      "learning_rate": 0.0001,
      "loss": 7.7299,
      "loss/crossentropy": 2.187662351131439,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20965678989887238,
      "step": 1470
    },
    {
      "epoch": 0.037,
      "grad_norm": 33.0,
      "grad_norm_var": 6.1666015625,
      "learning_rate": 0.0001,
      "loss": 7.5942,
      "loss/crossentropy": 2.174471014738083,
      "loss/hidden": 3.534765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21107099391520023,
      "step": 1480
    },
    {
      "epoch": 0.03725,
      "grad_norm": 40.5,
      "grad_norm_var": 7.112239583333333,
      "learning_rate": 0.0001,
      "loss": 7.5362,
      "loss/crossentropy": 2.069592148065567,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19253854881972074,
      "step": 1490
    },
    {
      "epoch": 0.0375,
      "grad_norm": 33.25,
      "grad_norm_var": 6.4869140625,
      "learning_rate": 0.0001,
      "loss": 7.5941,
      "loss/crossentropy": 2.0679068714380264,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20371587071567773,
      "step": 1500
    },
    {
      "epoch": 0.03775,
      "grad_norm": 30.875,
      "grad_norm_var": 2.5686848958333335,
      "learning_rate": 0.0001,
      "loss": 7.6557,
      "loss/crossentropy": 2.1364961892366408,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20447015166282653,
      "step": 1510
    },
    {
      "epoch": 0.038,
      "grad_norm": 32.5,
      "grad_norm_var": 2.6738932291666666,
      "learning_rate": 0.0001,
      "loss": 7.622,
      "loss/crossentropy": 2.1275009989738463,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18973923586308955,
      "step": 1520
    },
    {
      "epoch": 0.03825,
      "grad_norm": 32.5,
      "grad_norm_var": 1.3177083333333333,
      "learning_rate": 0.0001,
      "loss": 7.7366,
      "loss/crossentropy": 2.212426933646202,
      "loss/hidden": 3.44296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21735910680145026,
      "step": 1530
    },
    {
      "epoch": 0.0385,
      "grad_norm": 27.75,
      "grad_norm_var": 2.7348307291666667,
      "learning_rate": 0.0001,
      "loss": 7.6356,
      "loss/crossentropy": 2.1074824020266534,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19886015299707652,
      "step": 1540
    },
    {
      "epoch": 0.03875,
      "grad_norm": 31.75,
      "grad_norm_var": 4.234830729166666,
      "learning_rate": 0.0001,
      "loss": 7.7567,
      "loss/crossentropy": 2.177932971715927,
      "loss/hidden": 3.5609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2221821215003729,
      "step": 1550
    },
    {
      "epoch": 0.039,
      "grad_norm": 36.25,
      "grad_norm_var": 2.8791015625,
      "learning_rate": 0.0001,
      "loss": 7.5766,
      "loss/crossentropy": 2.0759232968091963,
      "loss/hidden": 3.386328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19086231291294098,
      "step": 1560
    },
    {
      "epoch": 0.03925,
      "grad_norm": 32.0,
      "grad_norm_var": 3.7046223958333333,
      "learning_rate": 0.0001,
      "loss": 7.537,
      "loss/crossentropy": 2.2487298890948297,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19235554365441204,
      "step": 1570
    },
    {
      "epoch": 0.0395,
      "grad_norm": 33.0,
      "grad_norm_var": 2.5927083333333334,
      "learning_rate": 0.0001,
      "loss": 7.5167,
      "loss/crossentropy": 2.2023983120918276,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20318181458860635,
      "step": 1580
    },
    {
      "epoch": 0.03975,
      "grad_norm": 30.0,
      "grad_norm_var": 3.4436848958333335,
      "learning_rate": 0.0001,
      "loss": 7.5528,
      "loss/crossentropy": 1.9812082558870316,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19151438660919667,
      "step": 1590
    },
    {
      "epoch": 0.04,
      "grad_norm": 30.0,
      "grad_norm_var": 2.6098307291666667,
      "learning_rate": 0.0001,
      "loss": 7.5232,
      "loss/crossentropy": 2.0910235196352005,
      "loss/hidden": 3.4640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20488944984972476,
      "step": 1600
    },
    {
      "epoch": 0.04025,
      "grad_norm": 30.0,
      "grad_norm_var": 1.5854166666666667,
      "learning_rate": 0.0001,
      "loss": 7.7078,
      "loss/crossentropy": 2.079045358300209,
      "loss/hidden": 3.483203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20172932054847478,
      "step": 1610
    },
    {
      "epoch": 0.0405,
      "grad_norm": 33.5,
      "grad_norm_var": 3.4368798010809257e+18,
      "learning_rate": 0.0001,
      "loss": 7.5441,
      "loss/crossentropy": 2.146761792898178,
      "loss/hidden": 3.562890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20237026009708642,
      "step": 1620
    },
    {
      "epoch": 0.04075,
      "grad_norm": 31.625,
      "grad_norm_var": 3.4368798024404393e+18,
      "learning_rate": 0.0001,
      "loss": 7.7162,
      "loss/crossentropy": 2.1575208425521852,
      "loss/hidden": 3.35703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1914537126198411,
      "step": 1630
    },
    {
      "epoch": 0.041,
      "grad_norm": 29.0,
      "grad_norm_var": 37.940625,
      "learning_rate": 0.0001,
      "loss": 7.5673,
      "loss/crossentropy": 2.2059059768915175,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1920375470072031,
      "step": 1640
    },
    {
      "epoch": 0.04125,
      "grad_norm": 30.75,
      "grad_norm_var": 61.631184895833336,
      "learning_rate": 0.0001,
      "loss": 7.5788,
      "loss/crossentropy": 2.1531882882118225,
      "loss/hidden": 3.3765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.192771671153605,
      "step": 1650
    },
    {
      "epoch": 0.0415,
      "grad_norm": 35.25,
      "grad_norm_var": 47.353125,
      "learning_rate": 0.0001,
      "loss": 7.5891,
      "loss/crossentropy": 2.1217811673879625,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19420051630586385,
      "step": 1660
    },
    {
      "epoch": 0.04175,
      "grad_norm": 29.875,
      "grad_norm_var": 21.580989583333334,
      "learning_rate": 0.0001,
      "loss": 7.6361,
      "loss/crossentropy": 2.0970900297164916,
      "loss/hidden": 3.48828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21125762071460485,
      "step": 1670
    },
    {
      "epoch": 0.042,
      "grad_norm": 36.5,
      "grad_norm_var": 72.77916666666667,
      "learning_rate": 0.0001,
      "loss": 7.7045,
      "loss/crossentropy": 2.1662445843219755,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21249269619584082,
      "step": 1680
    },
    {
      "epoch": 0.04225,
      "grad_norm": 32.75,
      "grad_norm_var": 72.42473958333333,
      "learning_rate": 0.0001,
      "loss": 7.5754,
      "loss/crossentropy": 2.1745404630899428,
      "loss/hidden": 3.478125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20168567337095739,
      "step": 1690
    },
    {
      "epoch": 0.0425,
      "grad_norm": 32.5,
      "grad_norm_var": 9.149739583333334,
      "learning_rate": 0.0001,
      "loss": 7.5555,
      "loss/crossentropy": 2.126581160724163,
      "loss/hidden": 3.34296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1933064443990588,
      "step": 1700
    },
    {
      "epoch": 0.04275,
      "grad_norm": 32.0,
      "grad_norm_var": 10.1369140625,
      "learning_rate": 0.0001,
      "loss": 7.5377,
      "loss/crossentropy": 2.1726800590753554,
      "loss/hidden": 3.441796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20911512654274703,
      "step": 1710
    },
    {
      "epoch": 0.043,
      "grad_norm": 31.0,
      "grad_norm_var": 3.314322916666667,
      "learning_rate": 0.0001,
      "loss": 7.6106,
      "loss/crossentropy": 1.9999253153800964,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18493321686983108,
      "step": 1720
    },
    {
      "epoch": 0.04325,
      "grad_norm": 38.0,
      "grad_norm_var": 6.673958333333333,
      "learning_rate": 0.0001,
      "loss": 7.7146,
      "loss/crossentropy": 2.1180618047714233,
      "loss/hidden": 3.46953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2076649811118841,
      "step": 1730
    },
    {
      "epoch": 0.0435,
      "grad_norm": 34.0,
      "grad_norm_var": 7.339322916666666,
      "learning_rate": 0.0001,
      "loss": 7.5464,
      "loss/crossentropy": 2.0583778262138366,
      "loss/hidden": 3.322265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18418019600212573,
      "step": 1740
    },
    {
      "epoch": 0.04375,
      "grad_norm": 31.0,
      "grad_norm_var": 4.749739583333334,
      "learning_rate": 0.0001,
      "loss": 7.5528,
      "loss/crossentropy": 2.2259044647216797,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2031107559800148,
      "step": 1750
    },
    {
      "epoch": 0.044,
      "grad_norm": 30.75,
      "grad_norm_var": 3.1389973958333335,
      "learning_rate": 0.0001,
      "loss": 7.5321,
      "loss/crossentropy": 2.124616578221321,
      "loss/hidden": 3.46484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2149519257247448,
      "step": 1760
    },
    {
      "epoch": 0.04425,
      "grad_norm": 28.625,
      "grad_norm_var": 6.442708333333333,
      "learning_rate": 0.0001,
      "loss": 7.63,
      "loss/crossentropy": 2.1418310686945916,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19125681575387715,
      "step": 1770
    },
    {
      "epoch": 0.0445,
      "grad_norm": 32.25,
      "grad_norm_var": 3.011458333333333,
      "learning_rate": 0.0001,
      "loss": 7.5307,
      "loss/crossentropy": 2.144363935291767,
      "loss/hidden": 3.440625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19334549438208343,
      "step": 1780
    },
    {
      "epoch": 0.04475,
      "grad_norm": 32.75,
      "grad_norm_var": 1.6559895833333333,
      "learning_rate": 0.0001,
      "loss": 7.627,
      "loss/crossentropy": 2.2538854971528055,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2010333575308323,
      "step": 1790
    },
    {
      "epoch": 0.045,
      "grad_norm": 31.5,
      "grad_norm_var": 1.9416015625,
      "learning_rate": 0.0001,
      "loss": 7.6471,
      "loss/crossentropy": 2.091226762533188,
      "loss/hidden": 3.450390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2152847982943058,
      "step": 1800
    },
    {
      "epoch": 0.04525,
      "grad_norm": 28.875,
      "grad_norm_var": 18.03515625,
      "learning_rate": 0.0001,
      "loss": 7.5626,
      "loss/crossentropy": 2.295862782001495,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20952691733837128,
      "step": 1810
    },
    {
      "epoch": 0.0455,
      "grad_norm": 30.75,
      "grad_norm_var": 18.4541015625,
      "learning_rate": 0.0001,
      "loss": 7.6872,
      "loss/crossentropy": 2.0833657890558244,
      "loss/hidden": 3.4875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2070673793554306,
      "step": 1820
    },
    {
      "epoch": 0.04575,
      "grad_norm": 30.25,
      "grad_norm_var": 2.812239583333333,
      "learning_rate": 0.0001,
      "loss": 7.647,
      "loss/crossentropy": 2.1021256439387797,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18996517434716226,
      "step": 1830
    },
    {
      "epoch": 0.046,
      "grad_norm": 33.75,
      "grad_norm_var": 2.7009765625,
      "learning_rate": 0.0001,
      "loss": 7.5203,
      "loss/crossentropy": 2.2004275381565095,
      "loss/hidden": 3.314453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.187726416811347,
      "step": 1840
    },
    {
      "epoch": 0.04625,
      "grad_norm": 30.0,
      "grad_norm_var": 4.067643229166666,
      "learning_rate": 0.0001,
      "loss": 7.6201,
      "loss/crossentropy": 2.1705893486738206,
      "loss/hidden": 3.522265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21116435080766677,
      "step": 1850
    },
    {
      "epoch": 0.0465,
      "grad_norm": 32.5,
      "grad_norm_var": 5.030989583333334,
      "learning_rate": 0.0001,
      "loss": 7.6235,
      "loss/crossentropy": 2.1204057022929192,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19528388790786266,
      "step": 1860
    },
    {
      "epoch": 0.04675,
      "grad_norm": 45.0,
      "grad_norm_var": 15.6603515625,
      "learning_rate": 0.0001,
      "loss": 7.4562,
      "loss/crossentropy": 2.154220977425575,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19214881088119,
      "step": 1870
    },
    {
      "epoch": 0.047,
      "grad_norm": 32.5,
      "grad_norm_var": 15.614518229166666,
      "learning_rate": 0.0001,
      "loss": 7.6258,
      "loss/crossentropy": 2.1000912792980673,
      "loss/hidden": 3.459375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20185065623372794,
      "step": 1880
    },
    {
      "epoch": 0.04725,
      "grad_norm": 33.25,
      "grad_norm_var": 5.931184895833334,
      "learning_rate": 0.0001,
      "loss": 7.4863,
      "loss/crossentropy": 2.2270338363945483,
      "loss/hidden": 3.337890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19286383930593728,
      "step": 1890
    },
    {
      "epoch": 0.0475,
      "grad_norm": 32.5,
      "grad_norm_var": 3.311393229166667,
      "learning_rate": 0.0001,
      "loss": 7.6001,
      "loss/crossentropy": 2.1998794853687285,
      "loss/hidden": 3.382421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19646506551653148,
      "step": 1900
    },
    {
      "epoch": 0.04775,
      "grad_norm": 31.375,
      "grad_norm_var": 3.577018229166667,
      "learning_rate": 0.0001,
      "loss": 7.653,
      "loss/crossentropy": 2.104053999483585,
      "loss/hidden": 3.48203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.201510801166296,
      "step": 1910
    },
    {
      "epoch": 0.048,
      "grad_norm": 30.5,
      "grad_norm_var": 17.68125,
      "learning_rate": 0.0001,
      "loss": 7.5273,
      "loss/crossentropy": 2.1184800997376443,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19248204957693815,
      "step": 1920
    },
    {
      "epoch": 0.04825,
      "grad_norm": 33.5,
      "grad_norm_var": 1.3,
      "learning_rate": 0.0001,
      "loss": 7.655,
      "loss/crossentropy": 2.1734499007463457,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1961166137829423,
      "step": 1930
    },
    {
      "epoch": 0.0485,
      "grad_norm": 30.125,
      "grad_norm_var": 2.5205729166666666,
      "learning_rate": 0.0001,
      "loss": 7.6186,
      "loss/crossentropy": 2.237542712688446,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20616979897022247,
      "step": 1940
    },
    {
      "epoch": 0.04875,
      "grad_norm": 30.75,
      "grad_norm_var": 2.810416666666667,
      "learning_rate": 0.0001,
      "loss": 7.5436,
      "loss/crossentropy": 2.1811093270778654,
      "loss/hidden": 3.40234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1928004425019026,
      "step": 1950
    },
    {
      "epoch": 0.049,
      "grad_norm": 34.25,
      "grad_norm_var": 1.5181640625,
      "learning_rate": 0.0001,
      "loss": 7.6175,
      "loss/crossentropy": 2.045266591012478,
      "loss/hidden": 3.45078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21615136358886958,
      "step": 1960
    },
    {
      "epoch": 0.04925,
      "grad_norm": 33.5,
      "grad_norm_var": 26.225455729166665,
      "learning_rate": 0.0001,
      "loss": 7.7083,
      "loss/crossentropy": 2.086462992429733,
      "loss/hidden": 3.469921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2192224683240056,
      "step": 1970
    },
    {
      "epoch": 0.0495,
      "grad_norm": 29.75,
      "grad_norm_var": 8.340625,
      "learning_rate": 0.0001,
      "loss": 7.5032,
      "loss/crossentropy": 2.0052025958895685,
      "loss/hidden": 3.407421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18473264537751674,
      "step": 1980
    },
    {
      "epoch": 0.04975,
      "grad_norm": 44.5,
      "grad_norm_var": 4635.33515625,
      "learning_rate": 0.0001,
      "loss": 7.5567,
      "loss/crossentropy": 2.04089385792613,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19560968028381467,
      "step": 1990
    },
    {
      "epoch": 0.05,
      "grad_norm": 51.0,
      "grad_norm_var": 4555.746875,
      "learning_rate": 0.0001,
      "loss": 7.6029,
      "loss/crossentropy": 2.2241372987627983,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2209097046405077,
      "step": 2000
    },
    {
      "epoch": 0.05025,
      "grad_norm": 32.5,
      "grad_norm_var": 49.11608072916667,
      "learning_rate": 0.0001,
      "loss": 7.5781,
      "loss/crossentropy": 2.2160742044448853,
      "loss/hidden": 3.4171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19539013858884574,
      "step": 2010
    },
    {
      "epoch": 0.0505,
      "grad_norm": 28.625,
      "grad_norm_var": 73.70201822916667,
      "learning_rate": 0.0001,
      "loss": 7.6375,
      "loss/crossentropy": 2.094863271713257,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20763578601181507,
      "step": 2020
    },
    {
      "epoch": 0.05075,
      "grad_norm": 30.125,
      "grad_norm_var": 22.494791666666668,
      "learning_rate": 0.0001,
      "loss": 7.6164,
      "loss/crossentropy": 2.121490868926048,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19190637897700072,
      "step": 2030
    },
    {
      "epoch": 0.051,
      "grad_norm": 34.5,
      "grad_norm_var": 20.549739583333334,
      "learning_rate": 0.0001,
      "loss": 7.5717,
      "loss/crossentropy": 2.0418698236346247,
      "loss/hidden": 3.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18076165094971658,
      "step": 2040
    },
    {
      "epoch": 0.05125,
      "grad_norm": 29.125,
      "grad_norm_var": 14.407291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5867,
      "loss/crossentropy": 2.1989556729793547,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2011225748807192,
      "step": 2050
    },
    {
      "epoch": 0.0515,
      "grad_norm": 32.75,
      "grad_norm_var": 17.120572916666667,
      "learning_rate": 0.0001,
      "loss": 7.6813,
      "loss/crossentropy": 2.073545518517494,
      "loss/hidden": 3.546484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22323863469064237,
      "step": 2060
    },
    {
      "epoch": 0.05175,
      "grad_norm": 42.25,
      "grad_norm_var": 76.27291666666666,
      "learning_rate": 0.0001,
      "loss": 7.6323,
      "loss/crossentropy": 2.1424515694379807,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19750071745365858,
      "step": 2070
    },
    {
      "epoch": 0.052,
      "grad_norm": 30.625,
      "grad_norm_var": 74.925,
      "learning_rate": 0.0001,
      "loss": 7.6198,
      "loss/crossentropy": 2.006666135787964,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19839615989476442,
      "step": 2080
    },
    {
      "epoch": 0.05225,
      "grad_norm": 41.5,
      "grad_norm_var": 30.869791666666668,
      "learning_rate": 0.0001,
      "loss": 7.5671,
      "loss/crossentropy": 2.2227450221776963,
      "loss/hidden": 3.451171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20443473970517517,
      "step": 2090
    },
    {
      "epoch": 0.0525,
      "grad_norm": 34.75,
      "grad_norm_var": 24.609309895833334,
      "learning_rate": 0.0001,
      "loss": 7.5928,
      "loss/crossentropy": 2.186553081870079,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18928833175450563,
      "step": 2100
    },
    {
      "epoch": 0.05275,
      "grad_norm": 29.125,
      "grad_norm_var": 18.37265625,
      "learning_rate": 0.0001,
      "loss": 7.7389,
      "loss/crossentropy": 2.0680640071630476,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.201955908536911,
      "step": 2110
    },
    {
      "epoch": 0.053,
      "grad_norm": 33.5,
      "grad_norm_var": 9.276497395833333,
      "learning_rate": 0.0001,
      "loss": 7.6148,
      "loss/crossentropy": 2.1981600403785704,
      "loss/hidden": 3.352734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1901057105511427,
      "step": 2120
    },
    {
      "epoch": 0.05325,
      "grad_norm": 31.75,
      "grad_norm_var": 5.308072916666666,
      "learning_rate": 0.0001,
      "loss": 7.48,
      "loss/crossentropy": 2.2574460208415985,
      "loss/hidden": 3.362109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19361322987824678,
      "step": 2130
    },
    {
      "epoch": 0.0535,
      "grad_norm": 29.625,
      "grad_norm_var": 8.5900390625,
      "learning_rate": 0.0001,
      "loss": 7.6644,
      "loss/crossentropy": 2.154197073727846,
      "loss/hidden": 3.50625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2086074635386467,
      "step": 2140
    },
    {
      "epoch": 0.05375,
      "grad_norm": 30.5,
      "grad_norm_var": 8.333072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5279,
      "loss/crossentropy": 2.1245498836040495,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19000006280839443,
      "step": 2150
    },
    {
      "epoch": 0.054,
      "grad_norm": 31.375,
      "grad_norm_var": 10.856184895833334,
      "learning_rate": 0.0001,
      "loss": 7.7098,
      "loss/crossentropy": 2.1968549311161043,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21285793352872134,
      "step": 2160
    },
    {
      "epoch": 0.05425,
      "grad_norm": 34.25,
      "grad_norm_var": 17.820572916666666,
      "learning_rate": 0.0001,
      "loss": 7.7095,
      "loss/crossentropy": 2.2533976465463637,
      "loss/hidden": 3.434375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20727334953844548,
      "step": 2170
    },
    {
      "epoch": 0.0545,
      "grad_norm": 36.0,
      "grad_norm_var": 22.128580729166668,
      "learning_rate": 0.0001,
      "loss": 7.5933,
      "loss/crossentropy": 2.227231651544571,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19693543761968613,
      "step": 2180
    },
    {
      "epoch": 0.05475,
      "grad_norm": 29.125,
      "grad_norm_var": 11.466666666666667,
      "learning_rate": 0.0001,
      "loss": 7.5348,
      "loss/crossentropy": 2.1581582985818386,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19432583590969443,
      "step": 2190
    },
    {
      "epoch": 0.055,
      "grad_norm": 37.25,
      "grad_norm_var": 10.514518229166667,
      "learning_rate": 0.0001,
      "loss": 7.6264,
      "loss/crossentropy": 2.0908095851540565,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20568534098565577,
      "step": 2200
    },
    {
      "epoch": 0.05525,
      "grad_norm": 33.25,
      "grad_norm_var": 5.6494140625,
      "learning_rate": 0.0001,
      "loss": 7.671,
      "loss/crossentropy": 2.0827252097427844,
      "loss/hidden": 3.54921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20747530292719601,
      "step": 2210
    },
    {
      "epoch": 0.0555,
      "grad_norm": 31.875,
      "grad_norm_var": 47.234375,
      "learning_rate": 0.0001,
      "loss": 7.5823,
      "loss/crossentropy": 2.2815075665712357,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20698099359869956,
      "step": 2220
    },
    {
      "epoch": 0.05575,
      "grad_norm": 37.5,
      "grad_norm_var": 30.277018229166668,
      "learning_rate": 0.0001,
      "loss": 7.5812,
      "loss/crossentropy": 2.2282338082790374,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19244133178144693,
      "step": 2230
    },
    {
      "epoch": 0.056,
      "grad_norm": 31.0,
      "grad_norm_var": 27.68515625,
      "learning_rate": 0.0001,
      "loss": 7.6485,
      "loss/crossentropy": 2.184544026851654,
      "loss/hidden": 3.39375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1996760057285428,
      "step": 2240
    },
    {
      "epoch": 0.05625,
      "grad_norm": 30.125,
      "grad_norm_var": 2.223372395833333,
      "learning_rate": 0.0001,
      "loss": 7.6003,
      "loss/crossentropy": 2.1572179198265076,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20082181300967933,
      "step": 2250
    },
    {
      "epoch": 0.0565,
      "grad_norm": 63.5,
      "grad_norm_var": 240.61295572916666,
      "learning_rate": 0.0001,
      "loss": 7.6129,
      "loss/crossentropy": 2.1616927281022074,
      "loss/hidden": 3.494140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21196384327486156,
      "step": 2260
    },
    {
      "epoch": 0.05675,
      "grad_norm": 29.875,
      "grad_norm_var": 276.98932291666665,
      "learning_rate": 0.0001,
      "loss": 7.5683,
      "loss/crossentropy": 2.1882350742816925,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2052389807999134,
      "step": 2270
    },
    {
      "epoch": 0.057,
      "grad_norm": 33.5,
      "grad_norm_var": 11.6681640625,
      "learning_rate": 0.0001,
      "loss": 7.6076,
      "loss/crossentropy": 2.1808354407548904,
      "loss/hidden": 3.540625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21027475781738758,
      "step": 2280
    },
    {
      "epoch": 0.05725,
      "grad_norm": 30.5,
      "grad_norm_var": 1.6872395833333333,
      "learning_rate": 0.0001,
      "loss": 7.5295,
      "loss/crossentropy": 2.1488232225179673,
      "loss/hidden": 3.496875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2111368477344513,
      "step": 2290
    },
    {
      "epoch": 0.0575,
      "grad_norm": 30.125,
      "grad_norm_var": 2.0603515625,
      "learning_rate": 0.0001,
      "loss": 7.6875,
      "loss/crossentropy": 2.1594634115695954,
      "loss/hidden": 3.384765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20478933807462454,
      "step": 2300
    },
    {
      "epoch": 0.05775,
      "grad_norm": 31.625,
      "grad_norm_var": 3.408072916666667,
      "learning_rate": 0.0001,
      "loss": 7.6636,
      "loss/crossentropy": 2.0932443618774412,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21733616031706332,
      "step": 2310
    },
    {
      "epoch": 0.058,
      "grad_norm": 30.5,
      "grad_norm_var": 3.5416015625,
      "learning_rate": 0.0001,
      "loss": 7.5737,
      "loss/crossentropy": 2.1831247925758364,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1989585768431425,
      "step": 2320
    },
    {
      "epoch": 0.05825,
      "grad_norm": 30.25,
      "grad_norm_var": 2.2393229166666666,
      "learning_rate": 0.0001,
      "loss": 7.5009,
      "loss/crossentropy": 2.2847615987062455,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20479252003133297,
      "step": 2330
    },
    {
      "epoch": 0.0585,
      "grad_norm": 33.0,
      "grad_norm_var": 6.374934895833333,
      "learning_rate": 0.0001,
      "loss": 7.7313,
      "loss/crossentropy": 2.1928456306457518,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20380632691085337,
      "step": 2340
    },
    {
      "epoch": 0.05875,
      "grad_norm": 29.125,
      "grad_norm_var": 4.7697265625,
      "learning_rate": 0.0001,
      "loss": 7.5476,
      "loss/crossentropy": 2.1438629984855653,
      "loss/hidden": 3.321875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19849517289549112,
      "step": 2350
    },
    {
      "epoch": 0.059,
      "grad_norm": 33.5,
      "grad_norm_var": 3.8893229166666665,
      "learning_rate": 0.0001,
      "loss": 7.6545,
      "loss/crossentropy": 2.1132273241877555,
      "loss/hidden": 3.419140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21593583207577466,
      "step": 2360
    },
    {
      "epoch": 0.05925,
      "grad_norm": 38.0,
      "grad_norm_var": 3.9934895833333335,
      "learning_rate": 0.0001,
      "loss": 7.6438,
      "loss/crossentropy": 2.1729103833436967,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.204007020406425,
      "step": 2370
    },
    {
      "epoch": 0.0595,
      "grad_norm": 28.875,
      "grad_norm_var": 4.180989583333333,
      "learning_rate": 0.0001,
      "loss": 7.5261,
      "loss/crossentropy": 2.1972236961126326,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21200564429163932,
      "step": 2380
    },
    {
      "epoch": 0.05975,
      "grad_norm": 30.375,
      "grad_norm_var": 2.9854166666666666,
      "learning_rate": 0.0001,
      "loss": 7.6233,
      "loss/crossentropy": 2.130857673287392,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2043182048946619,
      "step": 2390
    },
    {
      "epoch": 0.06,
      "grad_norm": 33.0,
      "grad_norm_var": 2.552018229166667,
      "learning_rate": 0.0001,
      "loss": 7.5626,
      "loss/crossentropy": 2.1780240714550017,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19501971434801818,
      "step": 2400
    },
    {
      "epoch": 0.06025,
      "grad_norm": 33.25,
      "grad_norm_var": 1.4978515625,
      "learning_rate": 0.0001,
      "loss": 7.5495,
      "loss/crossentropy": 2.075847564637661,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19856880297884344,
      "step": 2410
    },
    {
      "epoch": 0.0605,
      "grad_norm": 30.375,
      "grad_norm_var": 8.763541666666667,
      "learning_rate": 0.0001,
      "loss": 7.5161,
      "loss/crossentropy": 1.960635770857334,
      "loss/hidden": 3.521484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2014876109547913,
      "step": 2420
    },
    {
      "epoch": 0.06075,
      "grad_norm": 30.375,
      "grad_norm_var": 5.85390625,
      "learning_rate": 0.0001,
      "loss": 7.6453,
      "loss/crossentropy": 2.235329329967499,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22043778821825982,
      "step": 2430
    },
    {
      "epoch": 0.061,
      "grad_norm": 35.0,
      "grad_norm_var": 2.928465629875169e+18,
      "learning_rate": 0.0001,
      "loss": 7.5153,
      "loss/crossentropy": 2.0534446865320204,
      "loss/hidden": 3.469140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20287865065038205,
      "step": 2440
    },
    {
      "epoch": 0.06125,
      "grad_norm": 31.0,
      "grad_norm_var": 2.928465629810996e+18,
      "learning_rate": 0.0001,
      "loss": 7.5862,
      "loss/crossentropy": 2.1208725392818453,
      "loss/hidden": 3.534765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22640016246587039,
      "step": 2450
    },
    {
      "epoch": 0.0615,
      "grad_norm": 31.125,
      "grad_norm_var": 1.5614583333333334,
      "learning_rate": 0.0001,
      "loss": 7.5751,
      "loss/crossentropy": 2.206133508682251,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19071156904101372,
      "step": 2460
    },
    {
      "epoch": 0.06175,
      "grad_norm": 28.625,
      "grad_norm_var": 7.667643229166667,
      "learning_rate": 0.0001,
      "loss": 7.5604,
      "loss/crossentropy": 2.227986590564251,
      "loss/hidden": 3.30859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19392532519996167,
      "step": 2470
    },
    {
      "epoch": 0.062,
      "grad_norm": 34.0,
      "grad_norm_var": 5.81015625,
      "learning_rate": 0.0001,
      "loss": 7.6077,
      "loss/crossentropy": 2.1948474526405333,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2092638686299324,
      "step": 2480
    },
    {
      "epoch": 0.06225,
      "grad_norm": 34.5,
      "grad_norm_var": 7.656184895833333,
      "learning_rate": 0.0001,
      "loss": 7.4338,
      "loss/crossentropy": 1.935844713449478,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.184383431263268,
      "step": 2490
    },
    {
      "epoch": 0.0625,
      "grad_norm": 42.75,
      "grad_norm_var": 17.27890625,
      "learning_rate": 0.0001,
      "loss": 7.5645,
      "loss/crossentropy": 2.2957077413797378,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2223764518275857,
      "step": 2500
    },
    {
      "epoch": 0.06275,
      "grad_norm": 30.125,
      "grad_norm_var": 31.021809895833332,
      "learning_rate": 0.0001,
      "loss": 7.5539,
      "loss/crossentropy": 2.2163919866085053,
      "loss/hidden": 3.365234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2129704337567091,
      "step": 2510
    },
    {
      "epoch": 0.063,
      "grad_norm": 33.25,
      "grad_norm_var": 40.36764322916667,
      "learning_rate": 0.0001,
      "loss": 7.5664,
      "loss/crossentropy": 2.08260739967227,
      "loss/hidden": 3.4765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20597089193761348,
      "step": 2520
    },
    {
      "epoch": 0.06325,
      "grad_norm": 35.5,
      "grad_norm_var": 34.153125,
      "learning_rate": 0.0001,
      "loss": 7.559,
      "loss/crossentropy": 2.0747475802898405,
      "loss/hidden": 3.383203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1930603832937777,
      "step": 2530
    },
    {
      "epoch": 0.0635,
      "grad_norm": 31.25,
      "grad_norm_var": 12.302018229166666,
      "learning_rate": 0.0001,
      "loss": 7.609,
      "loss/crossentropy": 2.1542711734771727,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19414089974015952,
      "step": 2540
    },
    {
      "epoch": 0.06375,
      "grad_norm": 29.125,
      "grad_norm_var": 34.333072916666666,
      "learning_rate": 0.0001,
      "loss": 7.5103,
      "loss/crossentropy": 2.1697633042931557,
      "loss/hidden": 3.325390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18199986461549997,
      "step": 2550
    },
    {
      "epoch": 0.064,
      "grad_norm": 30.75,
      "grad_norm_var": 21.8322265625,
      "learning_rate": 0.0001,
      "loss": 7.6335,
      "loss/crossentropy": 2.20294488966465,
      "loss/hidden": 3.35703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20988359525799752,
      "step": 2560
    },
    {
      "epoch": 0.06425,
      "grad_norm": 36.5,
      "grad_norm_var": 21.339322916666667,
      "learning_rate": 0.0001,
      "loss": 7.665,
      "loss/crossentropy": 2.1367219746112824,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19768773801624775,
      "step": 2570
    },
    {
      "epoch": 0.0645,
      "grad_norm": 42.5,
      "grad_norm_var": 30.690625,
      "learning_rate": 0.0001,
      "loss": 7.6476,
      "loss/crossentropy": 2.2177338540554046,
      "loss/hidden": 3.605078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22041462864726782,
      "step": 2580
    },
    {
      "epoch": 0.06475,
      "grad_norm": 34.0,
      "grad_norm_var": 31.164583333333333,
      "learning_rate": 0.0001,
      "loss": 7.6158,
      "loss/crossentropy": 2.2423059731721877,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21196699403226377,
      "step": 2590
    },
    {
      "epoch": 0.065,
      "grad_norm": 34.25,
      "grad_norm_var": 23.593684895833334,
      "learning_rate": 0.0001,
      "loss": 7.5107,
      "loss/crossentropy": 2.241020438075066,
      "loss/hidden": 3.25859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18191274981945754,
      "step": 2600
    },
    {
      "epoch": 0.06525,
      "grad_norm": 28.75,
      "grad_norm_var": 30.975455729166665,
      "learning_rate": 0.0001,
      "loss": 7.5492,
      "loss/crossentropy": 2.052956056594849,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2021762602031231,
      "step": 2610
    },
    {
      "epoch": 0.0655,
      "grad_norm": 32.5,
      "grad_norm_var": 31.032291666666666,
      "learning_rate": 0.0001,
      "loss": 7.503,
      "loss/crossentropy": 2.17630957365036,
      "loss/hidden": 3.364453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19023955501616002,
      "step": 2620
    },
    {
      "epoch": 0.06575,
      "grad_norm": 34.5,
      "grad_norm_var": 6.083072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5728,
      "loss/crossentropy": 2.178831994533539,
      "loss/hidden": 3.38046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1946109678596258,
      "step": 2630
    },
    {
      "epoch": 0.066,
      "grad_norm": 33.5,
      "grad_norm_var": 4.01640625,
      "learning_rate": 0.0001,
      "loss": 7.5928,
      "loss/crossentropy": 2.0746659457683565,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19469854161143302,
      "step": 2640
    },
    {
      "epoch": 0.06625,
      "grad_norm": 31.0,
      "grad_norm_var": 21.718489583333334,
      "learning_rate": 0.0001,
      "loss": 7.6008,
      "loss/crossentropy": 2.202178010344505,
      "loss/hidden": 3.44609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2072868559509516,
      "step": 2650
    },
    {
      "epoch": 0.0665,
      "grad_norm": 31.375,
      "grad_norm_var": 24.743489583333332,
      "learning_rate": 0.0001,
      "loss": 7.5532,
      "loss/crossentropy": 2.0579031884670256,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18248077742755414,
      "step": 2660
    },
    {
      "epoch": 0.06675,
      "grad_norm": 29.5,
      "grad_norm_var": 6.5181640625,
      "learning_rate": 0.0001,
      "loss": 7.6173,
      "loss/crossentropy": 2.0323975652456285,
      "loss/hidden": 3.47265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20279558952897786,
      "step": 2670
    },
    {
      "epoch": 0.067,
      "grad_norm": 30.125,
      "grad_norm_var": 9.7931640625,
      "learning_rate": 0.0001,
      "loss": 7.6202,
      "loss/crossentropy": 2.170803511887789,
      "loss/hidden": 3.448828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21887149531394243,
      "step": 2680
    },
    {
      "epoch": 0.06725,
      "grad_norm": 28.75,
      "grad_norm_var": 6.42890625,
      "learning_rate": 0.0001,
      "loss": 7.6042,
      "loss/crossentropy": 2.129136848449707,
      "loss/hidden": 3.401171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21714741103351115,
      "step": 2690
    },
    {
      "epoch": 0.0675,
      "grad_norm": 28.5,
      "grad_norm_var": 5.76640625,
      "learning_rate": 0.0001,
      "loss": 7.5307,
      "loss/crossentropy": 2.2333458453416823,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19759367052465676,
      "step": 2700
    },
    {
      "epoch": 0.06775,
      "grad_norm": 31.25,
      "grad_norm_var": 4.925,
      "learning_rate": 0.0001,
      "loss": 7.5684,
      "loss/crossentropy": 2.1930347234010696,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19194095116108656,
      "step": 2710
    },
    {
      "epoch": 0.068,
      "grad_norm": 29.875,
      "grad_norm_var": 1.0291666666666666,
      "learning_rate": 0.0001,
      "loss": 7.6177,
      "loss/crossentropy": 2.1858380883932114,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18830202352255582,
      "step": 2720
    },
    {
      "epoch": 0.06825,
      "grad_norm": 33.25,
      "grad_norm_var": 5.19765625,
      "learning_rate": 0.0001,
      "loss": 7.6231,
      "loss/crossentropy": 2.1380992412567137,
      "loss/hidden": 3.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19919480197131634,
      "step": 2730
    },
    {
      "epoch": 0.0685,
      "grad_norm": 34.75,
      "grad_norm_var": 2.5872395833333335,
      "learning_rate": 0.0001,
      "loss": 7.7166,
      "loss/crossentropy": 2.299161267280579,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21277474984526634,
      "step": 2740
    },
    {
      "epoch": 0.06875,
      "grad_norm": 28.25,
      "grad_norm_var": 10.6525390625,
      "learning_rate": 0.0001,
      "loss": 7.6385,
      "loss/crossentropy": 2.0595856219530106,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20196862574666738,
      "step": 2750
    },
    {
      "epoch": 0.069,
      "grad_norm": 30.0,
      "grad_norm_var": 3.1885416666666666,
      "learning_rate": 0.0001,
      "loss": 7.5306,
      "loss/crossentropy": 2.1047334015369414,
      "loss/hidden": 3.527734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20494798701256514,
      "step": 2760
    },
    {
      "epoch": 0.06925,
      "grad_norm": 35.5,
      "grad_norm_var": 2.5136418842603423e+18,
      "learning_rate": 0.0001,
      "loss": 7.5758,
      "loss/crossentropy": 1.9750292956829072,
      "loss/hidden": 3.49453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1999529790133238,
      "step": 2770
    },
    {
      "epoch": 0.0695,
      "grad_norm": 30.875,
      "grad_norm_var": 2.513641880435452e+18,
      "learning_rate": 0.0001,
      "loss": 7.5901,
      "loss/crossentropy": 2.1417267471551895,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20058641098439695,
      "step": 2780
    },
    {
      "epoch": 0.06975,
      "grad_norm": 28.25,
      "grad_norm_var": 155.97024739583333,
      "learning_rate": 0.0001,
      "loss": 7.5485,
      "loss/crossentropy": 2.1048891723155974,
      "loss/hidden": 3.33515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19070767909288405,
      "step": 2790
    },
    {
      "epoch": 0.07,
      "grad_norm": 32.25,
      "grad_norm_var": 6.499739583333334,
      "learning_rate": 0.0001,
      "loss": 7.5996,
      "loss/crossentropy": 2.176823168247938,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20276176873594523,
      "step": 2800
    },
    {
      "epoch": 0.07025,
      "grad_norm": 29.5,
      "grad_norm_var": 4.684309895833334,
      "learning_rate": 0.0001,
      "loss": 7.4426,
      "loss/crossentropy": 2.041793665289879,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17560106106102466,
      "step": 2810
    },
    {
      "epoch": 0.0705,
      "grad_norm": 31.75,
      "grad_norm_var": 3.005989583333333,
      "learning_rate": 0.0001,
      "loss": 7.5811,
      "loss/crossentropy": 2.1094699330627917,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19798500649631023,
      "step": 2820
    },
    {
      "epoch": 0.07075,
      "grad_norm": 33.75,
      "grad_norm_var": 4.320572916666666,
      "learning_rate": 0.0001,
      "loss": 7.5475,
      "loss/crossentropy": 2.1311514347791674,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21825175136327743,
      "step": 2830
    },
    {
      "epoch": 0.071,
      "grad_norm": 34.75,
      "grad_norm_var": 3.7549465266724797e+18,
      "learning_rate": 0.0001,
      "loss": 7.5823,
      "loss/crossentropy": 2.120433983206749,
      "loss/hidden": 3.412109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20115374326705932,
      "step": 2840
    },
    {
      "epoch": 0.07125,
      "grad_norm": 29.25,
      "grad_norm_var": 3.7549465268743306e+18,
      "learning_rate": 0.0001,
      "loss": 7.5444,
      "loss/crossentropy": 2.1518867701292037,
      "loss/hidden": 3.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19730570819228888,
      "step": 2850
    },
    {
      "epoch": 0.0715,
      "grad_norm": 31.0,
      "grad_norm_var": 2.317122395833333,
      "learning_rate": 0.0001,
      "loss": 7.6362,
      "loss/crossentropy": 2.137280356884003,
      "loss/hidden": 3.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20695031639188527,
      "step": 2860
    },
    {
      "epoch": 0.07175,
      "grad_norm": 34.0,
      "grad_norm_var": 157.31015625,
      "learning_rate": 0.0001,
      "loss": 7.6643,
      "loss/crossentropy": 2.067914080619812,
      "loss/hidden": 3.5015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22938031535595654,
      "step": 2870
    },
    {
      "epoch": 0.072,
      "grad_norm": 32.5,
      "grad_norm_var": 151.77057291666668,
      "learning_rate": 0.0001,
      "loss": 7.5762,
      "loss/crossentropy": 2.0962916165590286,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19928023852407933,
      "step": 2880
    },
    {
      "epoch": 0.07225,
      "grad_norm": 35.75,
      "grad_norm_var": 104.17962239583333,
      "learning_rate": 0.0001,
      "loss": 7.7066,
      "loss/crossentropy": 2.1184468276798727,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1974171632900834,
      "step": 2890
    },
    {
      "epoch": 0.0725,
      "grad_norm": 32.25,
      "grad_norm_var": 116.02493489583334,
      "learning_rate": 0.0001,
      "loss": 7.6075,
      "loss/crossentropy": 2.1381339877843857,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19655965138226747,
      "step": 2900
    },
    {
      "epoch": 0.07275,
      "grad_norm": 31.5,
      "grad_norm_var": 1.9551432291666666,
      "learning_rate": 0.0001,
      "loss": 7.5247,
      "loss/crossentropy": 2.141963595151901,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19582534320652484,
      "step": 2910
    },
    {
      "epoch": 0.073,
      "grad_norm": 31.875,
      "grad_norm_var": 1.3014973958333333,
      "learning_rate": 0.0001,
      "loss": 7.5588,
      "loss/crossentropy": 2.077046422660351,
      "loss/hidden": 3.465234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20551967658102513,
      "step": 2920
    },
    {
      "epoch": 0.07325,
      "grad_norm": 32.25,
      "grad_norm_var": 8.734375,
      "learning_rate": 0.0001,
      "loss": 7.6195,
      "loss/crossentropy": 2.0110410653054713,
      "loss/hidden": 3.3984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1805876674130559,
      "step": 2930
    },
    {
      "epoch": 0.0735,
      "grad_norm": 32.25,
      "grad_norm_var": 4838.416666666667,
      "learning_rate": 0.0001,
      "loss": 7.6623,
      "loss/crossentropy": 2.1130379527807235,
      "loss/hidden": 3.478125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2008265011012554,
      "step": 2940
    },
    {
      "epoch": 0.07375,
      "grad_norm": 39.0,
      "grad_norm_var": 57.90807291666667,
      "learning_rate": 0.0001,
      "loss": 7.59,
      "loss/crossentropy": 2.1667733818292616,
      "loss/hidden": 3.3109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19806477334350348,
      "step": 2950
    },
    {
      "epoch": 0.074,
      "grad_norm": 31.875,
      "grad_norm_var": 31.228059895833333,
      "learning_rate": 0.0001,
      "loss": 7.5131,
      "loss/crossentropy": 2.2757086992263793,
      "loss/hidden": 3.445703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.197940625064075,
      "step": 2960
    },
    {
      "epoch": 0.07425,
      "grad_norm": 34.25,
      "grad_norm_var": 2.996809895833333,
      "learning_rate": 0.0001,
      "loss": 7.5841,
      "loss/crossentropy": 2.143115535378456,
      "loss/hidden": 3.45625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20177703239023687,
      "step": 2970
    },
    {
      "epoch": 0.0745,
      "grad_norm": 31.0,
      "grad_norm_var": 3.1546223958333335,
      "learning_rate": 0.0001,
      "loss": 7.5363,
      "loss/crossentropy": 2.2850147604942324,
      "loss/hidden": 3.283984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19140432458370923,
      "step": 2980
    },
    {
      "epoch": 0.07475,
      "grad_norm": 32.25,
      "grad_norm_var": 7.310416666666667,
      "learning_rate": 0.0001,
      "loss": 7.5339,
      "loss/crossentropy": 2.104042625427246,
      "loss/hidden": 3.490625,
      "loss/jsd": 0.0,
      "loss/logits": 0.211691821180284,
      "step": 2990
    },
    {
      "epoch": 0.075,
      "grad_norm": 32.5,
      "grad_norm_var": 8.506705729166667,
      "learning_rate": 0.0001,
      "loss": 7.5448,
      "loss/crossentropy": 2.2270253866910936,
      "loss/hidden": 3.4390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20328052509576083,
      "step": 3000
    },
    {
      "epoch": 0.07525,
      "grad_norm": 46.75,
      "grad_norm_var": 17.10390625,
      "learning_rate": 0.0001,
      "loss": 7.6114,
      "loss/crossentropy": 2.2623429775238035,
      "loss/hidden": 3.32265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19479246698319913,
      "step": 3010
    },
    {
      "epoch": 0.0755,
      "grad_norm": 31.75,
      "grad_norm_var": 18.981184895833334,
      "learning_rate": 0.0001,
      "loss": 7.5739,
      "loss/crossentropy": 2.1012531995773314,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20767469964921476,
      "step": 3020
    },
    {
      "epoch": 0.07575,
      "grad_norm": 45.25,
      "grad_norm_var": 15.55,
      "learning_rate": 0.0001,
      "loss": 7.6153,
      "loss/crossentropy": 2.090057593584061,
      "loss/hidden": 3.468359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894347405061126,
      "step": 3030
    },
    {
      "epoch": 0.076,
      "grad_norm": 32.0,
      "grad_norm_var": 29.878125,
      "learning_rate": 0.0001,
      "loss": 7.6585,
      "loss/crossentropy": 2.200800988078117,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19270651414990425,
      "step": 3040
    },
    {
      "epoch": 0.07625,
      "grad_norm": 31.0,
      "grad_norm_var": 10.167708333333334,
      "learning_rate": 0.0001,
      "loss": 7.5696,
      "loss/crossentropy": 2.0458697080612183,
      "loss/hidden": 3.53046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20314501021057368,
      "step": 3050
    },
    {
      "epoch": 0.0765,
      "grad_norm": 30.375,
      "grad_norm_var": 10.0181640625,
      "learning_rate": 0.0001,
      "loss": 7.5747,
      "loss/crossentropy": 2.2118399769067763,
      "loss/hidden": 3.482421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22158339023590087,
      "step": 3060
    },
    {
      "epoch": 0.07675,
      "grad_norm": 32.25,
      "grad_norm_var": 0.9650390625,
      "learning_rate": 0.0001,
      "loss": 7.5631,
      "loss/crossentropy": 2.1812553733587263,
      "loss/hidden": 3.4375,
      "loss/jsd": 0.0,
      "loss/logits": 0.217013025470078,
      "step": 3070
    },
    {
      "epoch": 0.077,
      "grad_norm": 32.75,
      "grad_norm_var": 0.9171223958333333,
      "learning_rate": 0.0001,
      "loss": 7.578,
      "loss/crossentropy": 2.181536224484444,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19952490609139203,
      "step": 3080
    },
    {
      "epoch": 0.07725,
      "grad_norm": 31.5,
      "grad_norm_var": 19.015559895833334,
      "learning_rate": 0.0001,
      "loss": 7.4878,
      "loss/crossentropy": 2.122265163064003,
      "loss/hidden": 3.4171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19191155321896075,
      "step": 3090
    },
    {
      "epoch": 0.0775,
      "grad_norm": 32.75,
      "grad_norm_var": 5.199739583333334,
      "learning_rate": 0.0001,
      "loss": 7.7471,
      "loss/crossentropy": 2.210584083199501,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21269479542970657,
      "step": 3100
    },
    {
      "epoch": 0.07775,
      "grad_norm": 33.25,
      "grad_norm_var": 4.490625,
      "learning_rate": 0.0001,
      "loss": 7.6586,
      "loss/crossentropy": 2.090894425660372,
      "loss/hidden": 3.46328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19853799045085907,
      "step": 3110
    },
    {
      "epoch": 0.078,
      "grad_norm": 31.375,
      "grad_norm_var": 27.370833333333334,
      "learning_rate": 0.0001,
      "loss": 7.5818,
      "loss/crossentropy": 2.044330509006977,
      "loss/hidden": 3.50078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20974230151623488,
      "step": 3120
    },
    {
      "epoch": 0.07825,
      "grad_norm": 31.75,
      "grad_norm_var": 6.688997395833334,
      "learning_rate": 0.0001,
      "loss": 7.654,
      "loss/crossentropy": 2.1380004197359086,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20455153118818997,
      "step": 3130
    },
    {
      "epoch": 0.0785,
      "grad_norm": 29.875,
      "grad_norm_var": 7.715625,
      "learning_rate": 0.0001,
      "loss": 7.6551,
      "loss/crossentropy": 2.1576705113053323,
      "loss/hidden": 3.4484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20728315506130457,
      "step": 3140
    },
    {
      "epoch": 0.07875,
      "grad_norm": 30.875,
      "grad_norm_var": 3.2864583333333335,
      "learning_rate": 0.0001,
      "loss": 7.5478,
      "loss/crossentropy": 2.233546493947506,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20845879297703504,
      "step": 3150
    },
    {
      "epoch": 0.079,
      "grad_norm": 34.25,
      "grad_norm_var": 15.4728515625,
      "learning_rate": 0.0001,
      "loss": 7.6264,
      "loss/crossentropy": 2.0873878210783006,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19576258175075054,
      "step": 3160
    },
    {
      "epoch": 0.07925,
      "grad_norm": 29.625,
      "grad_norm_var": 3.12265625,
      "learning_rate": 0.0001,
      "loss": 7.5192,
      "loss/crossentropy": 2.132347696274519,
      "loss/hidden": 3.345703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1958464809693396,
      "step": 3170
    },
    {
      "epoch": 0.0795,
      "grad_norm": 33.0,
      "grad_norm_var": 3.370247395833333,
      "learning_rate": 0.0001,
      "loss": 7.5169,
      "loss/crossentropy": 2.2280534476041796,
      "loss/hidden": 3.35546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19412651136517525,
      "step": 3180
    },
    {
      "epoch": 0.07975,
      "grad_norm": 28.25,
      "grad_norm_var": 4.989518229166666,
      "learning_rate": 0.0001,
      "loss": 7.6576,
      "loss/crossentropy": 2.138143754005432,
      "loss/hidden": 3.58828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23302078600972892,
      "step": 3190
    },
    {
      "epoch": 0.08,
      "grad_norm": 30.375,
      "grad_norm_var": 6.858072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5707,
      "loss/crossentropy": 2.1512865126132965,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1957532402127981,
      "step": 3200
    },
    {
      "epoch": 0.08025,
      "grad_norm": 39.0,
      "grad_norm_var": 34.83515625,
      "learning_rate": 0.0001,
      "loss": 7.6814,
      "loss/crossentropy": 2.113873428106308,
      "loss/hidden": 3.52890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22486987188458443,
      "step": 3210
    },
    {
      "epoch": 0.0805,
      "grad_norm": 30.625,
      "grad_norm_var": 33.244205729166666,
      "learning_rate": 0.0001,
      "loss": 7.6542,
      "loss/crossentropy": 2.153007471561432,
      "loss/hidden": 3.5078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20893471594899893,
      "step": 3220
    },
    {
      "epoch": 0.08075,
      "grad_norm": 31.25,
      "grad_norm_var": 619.4145182291667,
      "learning_rate": 0.0001,
      "loss": 7.5091,
      "loss/crossentropy": 2.169908273220062,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20535510070621968,
      "step": 3230
    },
    {
      "epoch": 0.081,
      "grad_norm": 30.0,
      "grad_norm_var": 599.9247395833333,
      "learning_rate": 0.0001,
      "loss": 7.5568,
      "loss/crossentropy": 2.048283484578133,
      "loss/hidden": 3.593359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21250182073563337,
      "step": 3240
    },
    {
      "epoch": 0.08125,
      "grad_norm": 30.5,
      "grad_norm_var": 2.982291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5569,
      "loss/crossentropy": 2.0828478574752807,
      "loss/hidden": 3.496484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20137296654284,
      "step": 3250
    },
    {
      "epoch": 0.0815,
      "grad_norm": 28.25,
      "grad_norm_var": 17.9900390625,
      "learning_rate": 0.0001,
      "loss": 7.5869,
      "loss/crossentropy": 2.28429861664772,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21088685300201176,
      "step": 3260
    },
    {
      "epoch": 0.08175,
      "grad_norm": 32.25,
      "grad_norm_var": 2.5587890625,
      "learning_rate": 0.0001,
      "loss": 7.6037,
      "loss/crossentropy": 2.1336950808763504,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20057348478585482,
      "step": 3270
    },
    {
      "epoch": 0.082,
      "grad_norm": 33.75,
      "grad_norm_var": 34.57076822916667,
      "learning_rate": 0.0001,
      "loss": 7.6129,
      "loss/crossentropy": 2.0662791609764097,
      "loss/hidden": 3.408203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22612145096063613,
      "step": 3280
    },
    {
      "epoch": 0.08225,
      "grad_norm": 32.75,
      "grad_norm_var": 18.8181640625,
      "learning_rate": 0.0001,
      "loss": 7.5469,
      "loss/crossentropy": 2.069541847705841,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1844556663185358,
      "step": 3290
    },
    {
      "epoch": 0.0825,
      "grad_norm": 29.375,
      "grad_norm_var": 3.0712890625,
      "learning_rate": 0.0001,
      "loss": 7.6025,
      "loss/crossentropy": 2.2499852567911147,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21941267363727093,
      "step": 3300
    },
    {
      "epoch": 0.08275,
      "grad_norm": 33.0,
      "grad_norm_var": 1.29140625,
      "learning_rate": 0.0001,
      "loss": 7.6334,
      "loss/crossentropy": 2.1277933359146117,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19795072823762894,
      "step": 3310
    },
    {
      "epoch": 0.083,
      "grad_norm": 34.0,
      "grad_norm_var": 3.41015625,
      "learning_rate": 0.0001,
      "loss": 7.6326,
      "loss/crossentropy": 2.214772176742554,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20066177062690257,
      "step": 3320
    },
    {
      "epoch": 0.08325,
      "grad_norm": 29.875,
      "grad_norm_var": 18.768489583333334,
      "learning_rate": 0.0001,
      "loss": 7.569,
      "loss/crossentropy": 2.093920087814331,
      "loss/hidden": 3.4375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2021244278177619,
      "step": 3330
    },
    {
      "epoch": 0.0835,
      "grad_norm": 30.25,
      "grad_norm_var": 18.695833333333333,
      "learning_rate": 0.0001,
      "loss": 7.5041,
      "loss/crossentropy": 2.0556397035717966,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17018966227769852,
      "step": 3340
    },
    {
      "epoch": 0.08375,
      "grad_norm": 28.875,
      "grad_norm_var": 2.854622395833333,
      "learning_rate": 0.0001,
      "loss": 7.5837,
      "loss/crossentropy": 2.0227382972836496,
      "loss/hidden": 3.4890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21123163159936667,
      "step": 3350
    },
    {
      "epoch": 0.084,
      "grad_norm": 33.5,
      "grad_norm_var": 2.894791666666667,
      "learning_rate": 0.0001,
      "loss": 7.5374,
      "loss/crossentropy": 2.146658593416214,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18999559991061687,
      "step": 3360
    },
    {
      "epoch": 0.08425,
      "grad_norm": 32.5,
      "grad_norm_var": 2.5759765625,
      "learning_rate": 0.0001,
      "loss": 7.6197,
      "loss/crossentropy": 1.9798088841140271,
      "loss/hidden": 3.569921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20031734639778734,
      "step": 3370
    },
    {
      "epoch": 0.0845,
      "grad_norm": 30.5,
      "grad_norm_var": 1.6436848958333334,
      "learning_rate": 0.0001,
      "loss": 7.5738,
      "loss/crossentropy": 2.1442878276109694,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20664554908871652,
      "step": 3380
    },
    {
      "epoch": 0.08475,
      "grad_norm": 29.25,
      "grad_norm_var": 1.9875,
      "learning_rate": 0.0001,
      "loss": 7.4965,
      "loss/crossentropy": 2.150037130713463,
      "loss/hidden": 3.3625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19511928483843805,
      "step": 3390
    },
    {
      "epoch": 0.085,
      "grad_norm": 32.25,
      "grad_norm_var": 2.2122395833333335,
      "learning_rate": 0.0001,
      "loss": 7.5151,
      "loss/crossentropy": 2.0501646161079408,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19615819547325372,
      "step": 3400
    },
    {
      "epoch": 0.08525,
      "grad_norm": 33.75,
      "grad_norm_var": 2.720833333333333,
      "learning_rate": 0.0001,
      "loss": 7.6432,
      "loss/crossentropy": 2.2328430742025374,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19816372059285642,
      "step": 3410
    },
    {
      "epoch": 0.0855,
      "grad_norm": 31.375,
      "grad_norm_var": 7.519791666666666,
      "learning_rate": 0.0001,
      "loss": 7.665,
      "loss/crossentropy": 2.1567111521959306,
      "loss/hidden": 3.500390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19839788805693387,
      "step": 3420
    },
    {
      "epoch": 0.08575,
      "grad_norm": 40.25,
      "grad_norm_var": 7.808072916666666,
      "learning_rate": 0.0001,
      "loss": 7.5919,
      "loss/crossentropy": 2.2465982705354692,
      "loss/hidden": 3.46328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21547308284789324,
      "step": 3430
    },
    {
      "epoch": 0.086,
      "grad_norm": 33.0,
      "grad_norm_var": 8.09765625,
      "learning_rate": 0.0001,
      "loss": 7.6239,
      "loss/crossentropy": 2.1680932968854902,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21484404131770135,
      "step": 3440
    },
    {
      "epoch": 0.08625,
      "grad_norm": 28.625,
      "grad_norm_var": 3.070572916666667,
      "learning_rate": 0.0001,
      "loss": 7.4802,
      "loss/crossentropy": 2.050510385632515,
      "loss/hidden": 3.421484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18613745234906673,
      "step": 3450
    },
    {
      "epoch": 0.0865,
      "grad_norm": 33.5,
      "grad_norm_var": 2.2143229166666667,
      "learning_rate": 0.0001,
      "loss": 7.5311,
      "loss/crossentropy": 2.0699805982410906,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1896037317812443,
      "step": 3460
    },
    {
      "epoch": 0.08675,
      "grad_norm": 31.875,
      "grad_norm_var": 5.651822916666666,
      "learning_rate": 0.0001,
      "loss": 7.596,
      "loss/crossentropy": 2.0576356425881386,
      "loss/hidden": 3.522265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20867941789329053,
      "step": 3470
    },
    {
      "epoch": 0.087,
      "grad_norm": 29.625,
      "grad_norm_var": 5.998372395833333,
      "learning_rate": 0.0001,
      "loss": 7.4738,
      "loss/crossentropy": 2.2489717990159988,
      "loss/hidden": 3.3203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18595699593424797,
      "step": 3480
    },
    {
      "epoch": 0.08725,
      "grad_norm": 28.5,
      "grad_norm_var": 5.2056640625,
      "learning_rate": 0.0001,
      "loss": 7.5246,
      "loss/crossentropy": 2.178911143541336,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.204429741948843,
      "step": 3490
    },
    {
      "epoch": 0.0875,
      "grad_norm": 28.25,
      "grad_norm_var": 1.8134765625,
      "learning_rate": 0.0001,
      "loss": 7.4901,
      "loss/crossentropy": 2.1816289871931076,
      "loss/hidden": 3.425390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20369651466608046,
      "step": 3500
    },
    {
      "epoch": 0.08775,
      "grad_norm": 32.25,
      "grad_norm_var": 59.895572916666666,
      "learning_rate": 0.0001,
      "loss": 7.6056,
      "loss/crossentropy": 2.124045217037201,
      "loss/hidden": 3.562890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2361205333843827,
      "step": 3510
    },
    {
      "epoch": 0.088,
      "grad_norm": 31.375,
      "grad_norm_var": 58.15358072916667,
      "learning_rate": 0.0001,
      "loss": 7.6185,
      "loss/crossentropy": 2.0576027542352677,
      "loss/hidden": 3.408984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19163726801052688,
      "step": 3520
    },
    {
      "epoch": 0.08825,
      "grad_norm": 30.875,
      "grad_norm_var": 19.622330729166666,
      "learning_rate": 0.0001,
      "loss": 7.608,
      "loss/crossentropy": 2.2166375398635862,
      "loss/hidden": 3.430078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20556784830987454,
      "step": 3530
    },
    {
      "epoch": 0.0885,
      "grad_norm": 33.25,
      "grad_norm_var": 3.76875,
      "learning_rate": 0.0001,
      "loss": 7.5655,
      "loss/crossentropy": 2.215288892388344,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21560241151601076,
      "step": 3540
    },
    {
      "epoch": 0.08875,
      "grad_norm": 30.0,
      "grad_norm_var": 3.316666666666667,
      "learning_rate": 0.0001,
      "loss": 7.4909,
      "loss/crossentropy": 2.0867604553699493,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1992955395951867,
      "step": 3550
    },
    {
      "epoch": 0.089,
      "grad_norm": 33.25,
      "grad_norm_var": 286.81555989583336,
      "learning_rate": 0.0001,
      "loss": 7.716,
      "loss/crossentropy": 2.2209325939416886,
      "loss/hidden": 3.46484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19783683270215988,
      "step": 3560
    },
    {
      "epoch": 0.08925,
      "grad_norm": 34.75,
      "grad_norm_var": 286.8197916666667,
      "learning_rate": 0.0001,
      "loss": 7.5905,
      "loss/crossentropy": 2.098256954550743,
      "loss/hidden": 3.309765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.184653827175498,
      "step": 3570
    },
    {
      "epoch": 0.0895,
      "grad_norm": 31.125,
      "grad_norm_var": 5.868489583333333,
      "learning_rate": 0.0001,
      "loss": 7.625,
      "loss/crossentropy": 2.0555992782115937,
      "loss/hidden": 3.45546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19005871675908564,
      "step": 3580
    },
    {
      "epoch": 0.08975,
      "grad_norm": 37.0,
      "grad_norm_var": 7.585416666666666,
      "learning_rate": 0.0001,
      "loss": 7.542,
      "loss/crossentropy": 2.0350914053618907,
      "loss/hidden": 3.396484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.192917075753212,
      "step": 3590
    },
    {
      "epoch": 0.09,
      "grad_norm": 34.25,
      "grad_norm_var": 26.8322265625,
      "learning_rate": 0.0001,
      "loss": 7.6028,
      "loss/crossentropy": 2.249291920661926,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20490463990718127,
      "step": 3600
    },
    {
      "epoch": 0.09025,
      "grad_norm": 30.375,
      "grad_norm_var": 2.9457682291666667,
      "learning_rate": 0.0001,
      "loss": 7.5738,
      "loss/crossentropy": 2.201851597428322,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20158183835446836,
      "step": 3610
    },
    {
      "epoch": 0.0905,
      "grad_norm": 30.125,
      "grad_norm_var": 14.446809895833333,
      "learning_rate": 0.0001,
      "loss": 7.5562,
      "loss/crossentropy": 2.188534340262413,
      "loss/hidden": 3.491796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22296689171344042,
      "step": 3620
    },
    {
      "epoch": 0.09075,
      "grad_norm": 32.0,
      "grad_norm_var": 15.792643229166666,
      "learning_rate": 0.0001,
      "loss": 7.5509,
      "loss/crossentropy": 2.1134337186813354,
      "loss/hidden": 3.4078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19732800796627997,
      "step": 3630
    },
    {
      "epoch": 0.091,
      "grad_norm": 29.875,
      "grad_norm_var": 14.366080729166667,
      "learning_rate": 0.0001,
      "loss": 7.6349,
      "loss/crossentropy": 2.0555363953113557,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19681458938866853,
      "step": 3640
    },
    {
      "epoch": 0.09125,
      "grad_norm": 33.25,
      "grad_norm_var": 13.740559895833334,
      "learning_rate": 0.0001,
      "loss": 7.6513,
      "loss/crossentropy": 2.2402344048023224,
      "loss/hidden": 3.46171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20878477580845356,
      "step": 3650
    },
    {
      "epoch": 0.0915,
      "grad_norm": 32.5,
      "grad_norm_var": 30.0150390625,
      "learning_rate": 0.0001,
      "loss": 7.5723,
      "loss/crossentropy": 2.14640394449234,
      "loss/hidden": 3.453515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20020358953624964,
      "step": 3660
    },
    {
      "epoch": 0.09175,
      "grad_norm": 36.0,
      "grad_norm_var": 49.2619140625,
      "learning_rate": 0.0001,
      "loss": 7.5782,
      "loss/crossentropy": 2.2063133299350737,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19907979741692544,
      "step": 3670
    },
    {
      "epoch": 0.092,
      "grad_norm": 32.0,
      "grad_norm_var": 15.4400390625,
      "learning_rate": 0.0001,
      "loss": 7.5501,
      "loss/crossentropy": 2.1933626160025597,
      "loss/hidden": 3.332421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19394716806709766,
      "step": 3680
    },
    {
      "epoch": 0.09225,
      "grad_norm": 28.375,
      "grad_norm_var": 13.159375,
      "learning_rate": 0.0001,
      "loss": 7.5258,
      "loss/crossentropy": 2.0778674989938737,
      "loss/hidden": 3.344921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19343881569802762,
      "step": 3690
    },
    {
      "epoch": 0.0925,
      "grad_norm": 32.5,
      "grad_norm_var": 15.096809895833333,
      "learning_rate": 0.0001,
      "loss": 7.4597,
      "loss/crossentropy": 2.0565507017076015,
      "loss/hidden": 3.389453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19122497290372847,
      "step": 3700
    },
    {
      "epoch": 0.09275,
      "grad_norm": 29.625,
      "grad_norm_var": 2.729622395833333,
      "learning_rate": 0.0001,
      "loss": 7.5714,
      "loss/crossentropy": 2.082234078645706,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2102882768958807,
      "step": 3710
    },
    {
      "epoch": 0.093,
      "grad_norm": 30.0,
      "grad_norm_var": 2.6254557291666667,
      "learning_rate": 0.0001,
      "loss": 7.6155,
      "loss/crossentropy": 2.2312920093536377,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18575988691300155,
      "step": 3720
    },
    {
      "epoch": 0.09325,
      "grad_norm": 33.25,
      "grad_norm_var": 1.6181640625,
      "learning_rate": 0.0001,
      "loss": 7.5344,
      "loss/crossentropy": 2.2397233605384828,
      "loss/hidden": 3.343359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20670964010059834,
      "step": 3730
    },
    {
      "epoch": 0.0935,
      "grad_norm": 33.5,
      "grad_norm_var": 3.3124348958333334,
      "learning_rate": 0.0001,
      "loss": 7.5839,
      "loss/crossentropy": 2.2065307170152666,
      "loss/hidden": 3.348046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20138480551540852,
      "step": 3740
    },
    {
      "epoch": 0.09375,
      "grad_norm": 36.25,
      "grad_norm_var": 9.07265625,
      "learning_rate": 0.0001,
      "loss": 7.6468,
      "loss/crossentropy": 2.192644628882408,
      "loss/hidden": 3.420703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20751077253371478,
      "step": 3750
    },
    {
      "epoch": 0.094,
      "grad_norm": 32.25,
      "grad_norm_var": 18.11015625,
      "learning_rate": 0.0001,
      "loss": 7.731,
      "loss/crossentropy": 2.205154886841774,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20712865255773066,
      "step": 3760
    },
    {
      "epoch": 0.09425,
      "grad_norm": 30.75,
      "grad_norm_var": 30.259309895833333,
      "learning_rate": 0.0001,
      "loss": 7.4267,
      "loss/crossentropy": 2.0174816213548183,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1771852731704712,
      "step": 3770
    },
    {
      "epoch": 0.0945,
      "grad_norm": 32.75,
      "grad_norm_var": 2.687239583333333,
      "learning_rate": 0.0001,
      "loss": 7.5078,
      "loss/crossentropy": 2.142752841114998,
      "loss/hidden": 3.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2145843595266342,
      "step": 3780
    },
    {
      "epoch": 0.09475,
      "grad_norm": 41.5,
      "grad_norm_var": 28.512239583333333,
      "learning_rate": 0.0001,
      "loss": 7.5586,
      "loss/crossentropy": 2.2184203058481216,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2169014524668455,
      "step": 3790
    },
    {
      "epoch": 0.095,
      "grad_norm": 32.5,
      "grad_norm_var": 29.637955729166666,
      "learning_rate": 0.0001,
      "loss": 7.5796,
      "loss/crossentropy": 2.0424424298107624,
      "loss/hidden": 3.440234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19471338465809823,
      "step": 3800
    },
    {
      "epoch": 0.09525,
      "grad_norm": 31.25,
      "grad_norm_var": 2.846809895833333,
      "learning_rate": 0.0001,
      "loss": 7.4697,
      "loss/crossentropy": 2.2178969264030455,
      "loss/hidden": 3.215625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18599100317806005,
      "step": 3810
    },
    {
      "epoch": 0.0955,
      "grad_norm": 35.75,
      "grad_norm_var": 5.571809895833334,
      "learning_rate": 0.0001,
      "loss": 7.5899,
      "loss/crossentropy": 2.147325333952904,
      "loss/hidden": 3.373046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19797458127141,
      "step": 3820
    },
    {
      "epoch": 0.09575,
      "grad_norm": 31.0,
      "grad_norm_var": 19.242708333333333,
      "learning_rate": 0.0001,
      "loss": 7.4524,
      "loss/crossentropy": 2.222689136862755,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20618323031812907,
      "step": 3830
    },
    {
      "epoch": 0.096,
      "grad_norm": 31.125,
      "grad_norm_var": 6.659309895833333,
      "learning_rate": 0.0001,
      "loss": 7.5371,
      "loss/crossentropy": 2.045217160880566,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21105091590434313,
      "step": 3840
    },
    {
      "epoch": 0.09625,
      "grad_norm": 29.5,
      "grad_norm_var": 3.6624348958333335,
      "learning_rate": 0.0001,
      "loss": 7.4832,
      "loss/crossentropy": 2.1246922612190247,
      "loss/hidden": 3.37109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19520843997597695,
      "step": 3850
    },
    {
      "epoch": 0.0965,
      "grad_norm": 32.25,
      "grad_norm_var": 3.840625,
      "learning_rate": 0.0001,
      "loss": 7.5247,
      "loss/crossentropy": 2.1101455599069596,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19655006285756826,
      "step": 3860
    },
    {
      "epoch": 0.09675,
      "grad_norm": 31.375,
      "grad_norm_var": 4.0759765625,
      "learning_rate": 0.0001,
      "loss": 7.7123,
      "loss/crossentropy": 2.1925098091363906,
      "loss/hidden": 3.57265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21303071565926074,
      "step": 3870
    },
    {
      "epoch": 0.097,
      "grad_norm": 32.25,
      "grad_norm_var": 2.940559895833333,
      "learning_rate": 0.0001,
      "loss": 7.6194,
      "loss/crossentropy": 2.0254321210086346,
      "loss/hidden": 3.504296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21549067068845035,
      "step": 3880
    },
    {
      "epoch": 0.09725,
      "grad_norm": 32.75,
      "grad_norm_var": 1.6895833333333334,
      "learning_rate": 0.0001,
      "loss": 7.5341,
      "loss/crossentropy": 2.18448192179203,
      "loss/hidden": 3.331640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18761022239923478,
      "step": 3890
    },
    {
      "epoch": 0.0975,
      "grad_norm": 32.5,
      "grad_norm_var": 14.326822916666666,
      "learning_rate": 0.0001,
      "loss": 7.6378,
      "loss/crossentropy": 2.0941652059555054,
      "loss/hidden": 3.45703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20138136427849532,
      "step": 3900
    },
    {
      "epoch": 0.09775,
      "grad_norm": 31.875,
      "grad_norm_var": 2.1186848958333333,
      "learning_rate": 0.0001,
      "loss": 7.6588,
      "loss/crossentropy": 2.09155390933156,
      "loss/hidden": 3.487109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2071656842716038,
      "step": 3910
    },
    {
      "epoch": 0.098,
      "grad_norm": 32.5,
      "grad_norm_var": 3.729622395833333,
      "learning_rate": 0.0001,
      "loss": 7.6663,
      "loss/crossentropy": 2.1858886659145353,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20006632767617702,
      "step": 3920
    },
    {
      "epoch": 0.09825,
      "grad_norm": 40.25,
      "grad_norm_var": 15.908072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5567,
      "loss/crossentropy": 2.2354005187749864,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20157534964382648,
      "step": 3930
    },
    {
      "epoch": 0.0985,
      "grad_norm": 31.625,
      "grad_norm_var": 17.745768229166668,
      "learning_rate": 0.0001,
      "loss": 7.5189,
      "loss/crossentropy": 2.015377716720104,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1868499366566539,
      "step": 3940
    },
    {
      "epoch": 0.09875,
      "grad_norm": 33.0,
      "grad_norm_var": 7.66640625,
      "learning_rate": 0.0001,
      "loss": 7.533,
      "loss/crossentropy": 2.0591939449310304,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20258171651512386,
      "step": 3950
    },
    {
      "epoch": 0.099,
      "grad_norm": 41.25,
      "grad_norm_var": 3.3442041663272433e+18,
      "learning_rate": 0.0001,
      "loss": 7.4832,
      "loss/crossentropy": 2.1117863088846205,
      "loss/hidden": 3.559375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2428264247253537,
      "step": 3960
    },
    {
      "epoch": 0.09925,
      "grad_norm": 30.875,
      "grad_norm_var": 3.3442041639803904e+18,
      "learning_rate": 0.0001,
      "loss": 7.6253,
      "loss/crossentropy": 2.1176558315753935,
      "loss/hidden": 3.55546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22726768516004087,
      "step": 3970
    },
    {
      "epoch": 0.0995,
      "grad_norm": 31.25,
      "grad_norm_var": 19.5384765625,
      "learning_rate": 0.0001,
      "loss": 7.4413,
      "loss/crossentropy": 2.088257111608982,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19193989606574177,
      "step": 3980
    },
    {
      "epoch": 0.09975,
      "grad_norm": 32.75,
      "grad_norm_var": 2.28125,
      "learning_rate": 0.0001,
      "loss": 7.6017,
      "loss/crossentropy": 2.134918417036533,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19602251183241606,
      "step": 3990
    },
    {
      "epoch": 0.1,
      "grad_norm": 32.25,
      "grad_norm_var": 6.172916666666667,
      "learning_rate": 0.0001,
      "loss": 7.5138,
      "loss/crossentropy": 2.1410045489668845,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20588791109621524,
      "step": 4000
    },
    {
      "epoch": 0.10025,
      "grad_norm": 30.0,
      "grad_norm_var": 11.4306640625,
      "learning_rate": 0.0001,
      "loss": 7.5549,
      "loss/crossentropy": 2.241489386558533,
      "loss/hidden": 3.343359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20121449399739505,
      "step": 4010
    },
    {
      "epoch": 0.1005,
      "grad_norm": 33.75,
      "grad_norm_var": 5.220247395833334,
      "learning_rate": 0.0001,
      "loss": 7.6037,
      "loss/crossentropy": 2.177129751443863,
      "loss/hidden": 3.4484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.195396139472723,
      "step": 4020
    },
    {
      "epoch": 0.10075,
      "grad_norm": 31.375,
      "grad_norm_var": 25.279622395833332,
      "learning_rate": 0.0001,
      "loss": 7.5495,
      "loss/crossentropy": 2.1074128076434135,
      "loss/hidden": 3.40859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18998505976051092,
      "step": 4030
    },
    {
      "epoch": 0.101,
      "grad_norm": 30.375,
      "grad_norm_var": 12.564518229166667,
      "learning_rate": 0.0001,
      "loss": 7.4628,
      "loss/crossentropy": 2.031256873905659,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19107761420309544,
      "step": 4040
    },
    {
      "epoch": 0.10125,
      "grad_norm": 29.25,
      "grad_norm_var": 8.242643229166667,
      "learning_rate": 0.0001,
      "loss": 7.5906,
      "loss/crossentropy": 2.2593255966901777,
      "loss/hidden": 3.33515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20071447864174843,
      "step": 4050
    },
    {
      "epoch": 0.1015,
      "grad_norm": 28.875,
      "grad_norm_var": 7.72890625,
      "learning_rate": 0.0001,
      "loss": 7.5008,
      "loss/crossentropy": 2.1623566120862963,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20114662442356349,
      "step": 4060
    },
    {
      "epoch": 0.10175,
      "grad_norm": 28.75,
      "grad_norm_var": 2.5291015625,
      "learning_rate": 0.0001,
      "loss": 7.4901,
      "loss/crossentropy": 2.1303680926561355,
      "loss/hidden": 3.458984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19380100946873427,
      "step": 4070
    },
    {
      "epoch": 0.102,
      "grad_norm": 33.25,
      "grad_norm_var": 1.9583333333333333,
      "learning_rate": 0.0001,
      "loss": 7.6067,
      "loss/crossentropy": 2.180470046401024,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20080227889120578,
      "step": 4080
    },
    {
      "epoch": 0.10225,
      "grad_norm": 31.0,
      "grad_norm_var": 43.7791015625,
      "learning_rate": 0.0001,
      "loss": 7.3949,
      "loss/crossentropy": 2.04951853454113,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1938589910045266,
      "step": 4090
    },
    {
      "epoch": 0.1025,
      "grad_norm": 31.5,
      "grad_norm_var": 40.233333333333334,
      "learning_rate": 0.0001,
      "loss": 7.5,
      "loss/crossentropy": 1.9436089858412742,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1772780598141253,
      "step": 4100
    },
    {
      "epoch": 0.10275,
      "grad_norm": 30.75,
      "grad_norm_var": 6.14765625,
      "learning_rate": 0.0001,
      "loss": 7.5926,
      "loss/crossentropy": 2.1081591993570328,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20672952029854058,
      "step": 4110
    },
    {
      "epoch": 0.103,
      "grad_norm": 31.875,
      "grad_norm_var": 11.225455729166667,
      "learning_rate": 0.0001,
      "loss": 7.6334,
      "loss/crossentropy": 2.0973087579011915,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20208985283970832,
      "step": 4120
    },
    {
      "epoch": 0.10325,
      "grad_norm": 29.375,
      "grad_norm_var": 25.839583333333334,
      "learning_rate": 0.0001,
      "loss": 7.4644,
      "loss/crossentropy": 2.205477836728096,
      "loss/hidden": 3.3546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19515758529305458,
      "step": 4130
    },
    {
      "epoch": 0.1035,
      "grad_norm": 31.25,
      "grad_norm_var": 3.6567057291666667,
      "learning_rate": 0.0001,
      "loss": 7.4746,
      "loss/crossentropy": 2.1042226657271383,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1850555408746004,
      "step": 4140
    },
    {
      "epoch": 0.10375,
      "grad_norm": 29.875,
      "grad_norm_var": 2.5952473958333333,
      "learning_rate": 0.0001,
      "loss": 7.5805,
      "loss/crossentropy": 2.1080187141895292,
      "loss/hidden": 3.481640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20642356667667627,
      "step": 4150
    },
    {
      "epoch": 0.104,
      "grad_norm": 34.75,
      "grad_norm_var": 12.11875,
      "learning_rate": 0.0001,
      "loss": 7.6864,
      "loss/crossentropy": 2.1868012815713884,
      "loss/hidden": 3.444140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19968394786119462,
      "step": 4160
    },
    {
      "epoch": 0.10425,
      "grad_norm": 32.25,
      "grad_norm_var": 11.406184895833333,
      "learning_rate": 0.0001,
      "loss": 7.6461,
      "loss/crossentropy": 2.0963368862867355,
      "loss/hidden": 3.4609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21419077794998884,
      "step": 4170
    },
    {
      "epoch": 0.1045,
      "grad_norm": 31.0,
      "grad_norm_var": 11.670833333333333,
      "learning_rate": 0.0001,
      "loss": 7.6029,
      "loss/crossentropy": 2.100097879767418,
      "loss/hidden": 3.35390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18958222791552543,
      "step": 4180
    },
    {
      "epoch": 0.10475,
      "grad_norm": 29.375,
      "grad_norm_var": 7.664518229166666,
      "learning_rate": 0.0001,
      "loss": 7.5954,
      "loss/crossentropy": 2.2243005722761153,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19990855641663074,
      "step": 4190
    },
    {
      "epoch": 0.105,
      "grad_norm": 29.75,
      "grad_norm_var": 4.333072916666667,
      "learning_rate": 0.0001,
      "loss": 7.6671,
      "loss/crossentropy": 2.250749832391739,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19082491770386695,
      "step": 4200
    },
    {
      "epoch": 0.10525,
      "grad_norm": 29.25,
      "grad_norm_var": 10.568489583333333,
      "learning_rate": 0.0001,
      "loss": 7.6118,
      "loss/crossentropy": 2.155411234498024,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2127680890262127,
      "step": 4210
    },
    {
      "epoch": 0.1055,
      "grad_norm": 34.25,
      "grad_norm_var": 8.6556640625,
      "learning_rate": 0.0001,
      "loss": 7.4391,
      "loss/crossentropy": 2.037487879395485,
      "loss/hidden": 3.31015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17465929109603168,
      "step": 4220
    },
    {
      "epoch": 0.10575,
      "grad_norm": 30.625,
      "grad_norm_var": 8.183072916666667,
      "learning_rate": 0.0001,
      "loss": 7.6537,
      "loss/crossentropy": 2.067080709338188,
      "loss/hidden": 3.316796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1857584908604622,
      "step": 4230
    },
    {
      "epoch": 0.106,
      "grad_norm": 31.25,
      "grad_norm_var": 1.4171223958333334,
      "learning_rate": 0.0001,
      "loss": 7.608,
      "loss/crossentropy": 2.279679241776466,
      "loss/hidden": 3.33984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.206354571133852,
      "step": 4240
    },
    {
      "epoch": 0.10625,
      "grad_norm": 31.75,
      "grad_norm_var": 31.7681640625,
      "learning_rate": 0.0001,
      "loss": 7.5259,
      "loss/crossentropy": 2.1007855504751207,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1838985349982977,
      "step": 4250
    },
    {
      "epoch": 0.1065,
      "grad_norm": 28.625,
      "grad_norm_var": 2.5104166666666665,
      "learning_rate": 0.0001,
      "loss": 7.5786,
      "loss/crossentropy": 2.1117694169282912,
      "loss/hidden": 3.465625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2134232448413968,
      "step": 4260
    },
    {
      "epoch": 0.10675,
      "grad_norm": 33.25,
      "grad_norm_var": 5.585416666666666,
      "learning_rate": 0.0001,
      "loss": 7.5424,
      "loss/crossentropy": 2.1906253546476364,
      "loss/hidden": 3.380859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19627480674535036,
      "step": 4270
    },
    {
      "epoch": 0.107,
      "grad_norm": 33.75,
      "grad_norm_var": 16.780208333333334,
      "learning_rate": 0.0001,
      "loss": 7.7039,
      "loss/crossentropy": 2.167680537700653,
      "loss/hidden": 3.469921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21632006093859674,
      "step": 4280
    },
    {
      "epoch": 0.10725,
      "grad_norm": 32.75,
      "grad_norm_var": 14.226497395833333,
      "learning_rate": 0.0001,
      "loss": 7.6197,
      "loss/crossentropy": 2.2265933483839033,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20777842812240124,
      "step": 4290
    },
    {
      "epoch": 0.1075,
      "grad_norm": 32.25,
      "grad_norm_var": 5.06640625,
      "learning_rate": 0.0001,
      "loss": 7.5019,
      "loss/crossentropy": 2.0171881064772608,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1733700342476368,
      "step": 4300
    },
    {
      "epoch": 0.10775,
      "grad_norm": 32.75,
      "grad_norm_var": 39.37473958333333,
      "learning_rate": 0.0001,
      "loss": 7.6624,
      "loss/crossentropy": 2.052209459245205,
      "loss/hidden": 3.487109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2088288875296712,
      "step": 4310
    },
    {
      "epoch": 0.108,
      "grad_norm": 29.375,
      "grad_norm_var": 2.460724588873515e+18,
      "learning_rate": 0.0001,
      "loss": 7.5801,
      "loss/crossentropy": 2.096518099308014,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2093063434585929,
      "step": 4320
    },
    {
      "epoch": 0.10825,
      "grad_norm": 28.25,
      "grad_norm_var": 2.4607245888865874e+18,
      "learning_rate": 0.0001,
      "loss": 7.4793,
      "loss/crossentropy": 2.1164163142442702,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19930963944643737,
      "step": 4330
    },
    {
      "epoch": 0.1085,
      "grad_norm": 33.5,
      "grad_norm_var": 46.587239583333336,
      "learning_rate": 0.0001,
      "loss": 7.5915,
      "loss/crossentropy": 2.1831407219171526,
      "loss/hidden": 3.446875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21474861968308687,
      "step": 4340
    },
    {
      "epoch": 0.10875,
      "grad_norm": 31.0,
      "grad_norm_var": 17.534375,
      "learning_rate": 0.0001,
      "loss": 7.4743,
      "loss/crossentropy": 2.1157304018735887,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2028682116419077,
      "step": 4350
    },
    {
      "epoch": 0.109,
      "grad_norm": 28.625,
      "grad_norm_var": 4.282291666666667,
      "learning_rate": 0.0001,
      "loss": 7.6076,
      "loss/crossentropy": 1.980335572361946,
      "loss/hidden": 3.541796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20984734632074833,
      "step": 4360
    },
    {
      "epoch": 0.10925,
      "grad_norm": 31.25,
      "grad_norm_var": 30.812955729166667,
      "learning_rate": 0.0001,
      "loss": 7.5349,
      "loss/crossentropy": 2.107056123018265,
      "loss/hidden": 3.25859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18440892472863196,
      "step": 4370
    },
    {
      "epoch": 0.1095,
      "grad_norm": 30.375,
      "grad_norm_var": 21.462239583333332,
      "learning_rate": 0.0001,
      "loss": 7.5459,
      "loss/crossentropy": 2.2860846698284147,
      "loss/hidden": 3.297265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1911198776215315,
      "step": 4380
    },
    {
      "epoch": 0.10975,
      "grad_norm": 30.0,
      "grad_norm_var": 5.008333333333334,
      "learning_rate": 0.0001,
      "loss": 7.59,
      "loss/crossentropy": 2.150055022537708,
      "loss/hidden": 3.396875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20779079720377922,
      "step": 4390
    },
    {
      "epoch": 0.11,
      "grad_norm": 37.25,
      "grad_norm_var": 24.933268229166668,
      "learning_rate": 0.0001,
      "loss": 7.6256,
      "loss/crossentropy": 2.225931641459465,
      "loss/hidden": 3.569921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23082431070506573,
      "step": 4400
    },
    {
      "epoch": 0.11025,
      "grad_norm": 29.375,
      "grad_norm_var": 26.770572916666666,
      "learning_rate": 0.0001,
      "loss": 7.5585,
      "loss/crossentropy": 2.1318808451294897,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.205277425237,
      "step": 4410
    },
    {
      "epoch": 0.1105,
      "grad_norm": 28.25,
      "grad_norm_var": 6.009830729166667,
      "learning_rate": 0.0001,
      "loss": 7.5037,
      "loss/crossentropy": 2.108688759803772,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19271691460162402,
      "step": 4420
    },
    {
      "epoch": 0.11075,
      "grad_norm": 30.875,
      "grad_norm_var": 2.85390625,
      "learning_rate": 0.0001,
      "loss": 7.5599,
      "loss/crossentropy": 2.1696368783712385,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19752040579915048,
      "step": 4430
    },
    {
      "epoch": 0.111,
      "grad_norm": 30.25,
      "grad_norm_var": 18.476822916666666,
      "learning_rate": 0.0001,
      "loss": 7.5321,
      "loss/crossentropy": 2.167906680703163,
      "loss/hidden": 3.48828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2087454443797469,
      "step": 4440
    },
    {
      "epoch": 0.11125,
      "grad_norm": 30.75,
      "grad_norm_var": 4.6634765625,
      "learning_rate": 0.0001,
      "loss": 7.6168,
      "loss/crossentropy": 2.0555444791913033,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.17809300348162652,
      "step": 4450
    },
    {
      "epoch": 0.1115,
      "grad_norm": 31.25,
      "grad_norm_var": 5.618489583333333,
      "learning_rate": 0.0001,
      "loss": 7.669,
      "loss/crossentropy": 2.2362961381673814,
      "loss/hidden": 3.480078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21373681984841825,
      "step": 4460
    },
    {
      "epoch": 0.11175,
      "grad_norm": 31.625,
      "grad_norm_var": 1.8311848958333334,
      "learning_rate": 0.0001,
      "loss": 7.469,
      "loss/crossentropy": 2.207303923368454,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21020539589226245,
      "step": 4470
    },
    {
      "epoch": 0.112,
      "grad_norm": 66.0,
      "grad_norm_var": 76.79973958333333,
      "learning_rate": 0.0001,
      "loss": 7.5099,
      "loss/crossentropy": 2.170773930847645,
      "loss/hidden": 3.4703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20324019938707352,
      "step": 4480
    },
    {
      "epoch": 0.11225,
      "grad_norm": 35.0,
      "grad_norm_var": 76.34895833333333,
      "learning_rate": 0.0001,
      "loss": 7.5149,
      "loss/crossentropy": 2.047274041175842,
      "loss/hidden": 3.337890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18544426914304496,
      "step": 4490
    },
    {
      "epoch": 0.1125,
      "grad_norm": 28.25,
      "grad_norm_var": 4.081705729166667,
      "learning_rate": 0.0001,
      "loss": 7.5745,
      "loss/crossentropy": 2.1270942091941833,
      "loss/hidden": 3.311328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19201683439314365,
      "step": 4500
    },
    {
      "epoch": 0.11275,
      "grad_norm": 28.25,
      "grad_norm_var": 2.6192057291666666,
      "learning_rate": 0.0001,
      "loss": 7.632,
      "loss/crossentropy": 2.0815075978636743,
      "loss/hidden": 3.438671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20613169986754656,
      "step": 4510
    },
    {
      "epoch": 0.113,
      "grad_norm": 33.25,
      "grad_norm_var": 166.34212239583334,
      "learning_rate": 0.0001,
      "loss": 7.5383,
      "loss/crossentropy": 2.144553080201149,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19253196399658917,
      "step": 4520
    },
    {
      "epoch": 0.11325,
      "grad_norm": 35.75,
      "grad_norm_var": 1.734519148252968e+18,
      "learning_rate": 0.0001,
      "loss": 7.5876,
      "loss/crossentropy": 2.212298333644867,
      "loss/hidden": 3.34765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19719784446060656,
      "step": 4530
    },
    {
      "epoch": 0.1135,
      "grad_norm": 30.625,
      "grad_norm_var": 32.881184895833336,
      "learning_rate": 0.0001,
      "loss": 7.4563,
      "loss/crossentropy": 2.095240616798401,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19202221632003785,
      "step": 4540
    },
    {
      "epoch": 0.11375,
      "grad_norm": 31.625,
      "grad_norm_var": 87.4853515625,
      "learning_rate": 0.0001,
      "loss": 7.5184,
      "loss/crossentropy": 2.134147650748491,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19104634067043663,
      "step": 4550
    },
    {
      "epoch": 0.114,
      "grad_norm": 32.25,
      "grad_norm_var": 78.95618489583333,
      "learning_rate": 0.0001,
      "loss": 7.4668,
      "loss/crossentropy": 2.1294006586074827,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1853051505982876,
      "step": 4560
    },
    {
      "epoch": 0.11425,
      "grad_norm": 30.5,
      "grad_norm_var": 4.706705729166667,
      "learning_rate": 0.0001,
      "loss": 7.5114,
      "loss/crossentropy": 2.0886638939380644,
      "loss/hidden": 3.452734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20859680250287055,
      "step": 4570
    },
    {
      "epoch": 0.1145,
      "grad_norm": 35.5,
      "grad_norm_var": 5.934830729166666,
      "learning_rate": 0.0001,
      "loss": 7.56,
      "loss/crossentropy": 2.1498399868607523,
      "loss/hidden": 3.50703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18966037435457112,
      "step": 4580
    },
    {
      "epoch": 0.11475,
      "grad_norm": 32.75,
      "grad_norm_var": 3.358268229166667,
      "learning_rate": 0.0001,
      "loss": 7.6013,
      "loss/crossentropy": 2.199784816801548,
      "loss/hidden": 3.347265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18477672804147005,
      "step": 4590
    },
    {
      "epoch": 0.115,
      "grad_norm": 31.375,
      "grad_norm_var": 17.598372395833334,
      "learning_rate": 0.0001,
      "loss": 7.5806,
      "loss/crossentropy": 2.165570431947708,
      "loss/hidden": 3.505859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2113142903894186,
      "step": 4600
    },
    {
      "epoch": 0.11525,
      "grad_norm": 31.75,
      "grad_norm_var": 11.908268229166667,
      "learning_rate": 0.0001,
      "loss": 7.6265,
      "loss/crossentropy": 2.07996127307415,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21436248067766428,
      "step": 4610
    },
    {
      "epoch": 0.1155,
      "grad_norm": 28.125,
      "grad_norm_var": 4.520833333333333,
      "learning_rate": 0.0001,
      "loss": 7.57,
      "loss/crossentropy": 2.1484976023435594,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1913412597030401,
      "step": 4620
    },
    {
      "epoch": 0.11575,
      "grad_norm": 30.125,
      "grad_norm_var": 4.1875,
      "learning_rate": 0.0001,
      "loss": 7.4701,
      "loss/crossentropy": 2.0956287920475005,
      "loss/hidden": 3.39140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19558896012604238,
      "step": 4630
    },
    {
      "epoch": 0.116,
      "grad_norm": 29.875,
      "grad_norm_var": 2.106184895833333,
      "learning_rate": 0.0001,
      "loss": 7.4538,
      "loss/crossentropy": 2.1474092990159988,
      "loss/hidden": 3.45234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2039597311988473,
      "step": 4640
    },
    {
      "epoch": 0.11625,
      "grad_norm": 31.75,
      "grad_norm_var": 3.520768229166667,
      "learning_rate": 0.0001,
      "loss": 7.6657,
      "loss/crossentropy": 2.1128339886665346,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21120875477790832,
      "step": 4650
    },
    {
      "epoch": 0.1165,
      "grad_norm": 31.5,
      "grad_norm_var": 2.133072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5441,
      "loss/crossentropy": 2.1643586844205855,
      "loss/hidden": 3.4765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20440610516816377,
      "step": 4660
    },
    {
      "epoch": 0.11675,
      "grad_norm": 33.0,
      "grad_norm_var": 3.504622395833333,
      "learning_rate": 0.0001,
      "loss": 7.6051,
      "loss/crossentropy": 2.195269528031349,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20059011913836003,
      "step": 4670
    },
    {
      "epoch": 0.117,
      "grad_norm": 30.0,
      "grad_norm_var": 31.235416666666666,
      "learning_rate": 0.0001,
      "loss": 7.4526,
      "loss/crossentropy": 2.082709529995918,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19617473538964986,
      "step": 4680
    },
    {
      "epoch": 0.11725,
      "grad_norm": 32.0,
      "grad_norm_var": 3.45625,
      "learning_rate": 0.0001,
      "loss": 7.5593,
      "loss/crossentropy": 2.1473275452852247,
      "loss/hidden": 3.481640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20148523468524218,
      "step": 4690
    },
    {
      "epoch": 0.1175,
      "grad_norm": 31.875,
      "grad_norm_var": 2.145768229166667,
      "learning_rate": 0.0001,
      "loss": 7.4609,
      "loss/crossentropy": 2.14631325006485,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20976583026349543,
      "step": 4700
    },
    {
      "epoch": 0.11775,
      "grad_norm": 32.5,
      "grad_norm_var": 33.20826822916667,
      "learning_rate": 0.0001,
      "loss": 7.5877,
      "loss/crossentropy": 2.1093395471572878,
      "loss/hidden": 3.46953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20207433141767978,
      "step": 4710
    },
    {
      "epoch": 0.118,
      "grad_norm": 30.75,
      "grad_norm_var": 10.86875,
      "learning_rate": 0.0001,
      "loss": 7.5437,
      "loss/crossentropy": 2.1447531282901764,
      "loss/hidden": 3.353515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1963239062577486,
      "step": 4720
    },
    {
      "epoch": 0.11825,
      "grad_norm": 34.5,
      "grad_norm_var": 4.473372395833334,
      "learning_rate": 0.0001,
      "loss": 7.6248,
      "loss/crossentropy": 2.1142423778772352,
      "loss/hidden": 3.283203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19059138614684343,
      "step": 4730
    },
    {
      "epoch": 0.1185,
      "grad_norm": 27.875,
      "grad_norm_var": 3.098372395833333,
      "learning_rate": 0.0001,
      "loss": 7.5631,
      "loss/crossentropy": 2.07417613863945,
      "loss/hidden": 3.455859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19420854579657315,
      "step": 4740
    },
    {
      "epoch": 0.11875,
      "grad_norm": 29.875,
      "grad_norm_var": 2.734375,
      "learning_rate": 0.0001,
      "loss": 7.6499,
      "loss/crossentropy": 2.0995796024799347,
      "loss/hidden": 3.518359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21285411342978477,
      "step": 4750
    },
    {
      "epoch": 0.119,
      "grad_norm": 32.5,
      "grad_norm_var": 2.7228515625,
      "learning_rate": 0.0001,
      "loss": 7.5854,
      "loss/crossentropy": 2.169099047780037,
      "loss/hidden": 3.462109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22147531677037477,
      "step": 4760
    },
    {
      "epoch": 0.11925,
      "grad_norm": 32.5,
      "grad_norm_var": 34.799739583333334,
      "learning_rate": 0.0001,
      "loss": 7.435,
      "loss/crossentropy": 2.0821361050009726,
      "loss/hidden": 3.2953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17594938166439533,
      "step": 4770
    },
    {
      "epoch": 0.1195,
      "grad_norm": 33.75,
      "grad_norm_var": 23.00390625,
      "learning_rate": 0.0001,
      "loss": 7.6152,
      "loss/crossentropy": 2.183753404021263,
      "loss/hidden": 3.364453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19579742290079594,
      "step": 4780
    },
    {
      "epoch": 0.11975,
      "grad_norm": 29.25,
      "grad_norm_var": 23.6541015625,
      "learning_rate": 0.0001,
      "loss": 7.5295,
      "loss/crossentropy": 2.1669270396232605,
      "loss/hidden": 3.505859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21605710163712502,
      "step": 4790
    },
    {
      "epoch": 0.12,
      "grad_norm": 34.25,
      "grad_norm_var": 4.624739583333334,
      "learning_rate": 0.0001,
      "loss": 7.6351,
      "loss/crossentropy": 2.116096779704094,
      "loss/hidden": 3.4625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21523005720227956,
      "step": 4800
    },
    {
      "epoch": 0.12025,
      "grad_norm": 29.5,
      "grad_norm_var": 6.009375,
      "learning_rate": 0.0001,
      "loss": 7.5683,
      "loss/crossentropy": 2.051844981312752,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17574754767119885,
      "step": 4810
    },
    {
      "epoch": 0.1205,
      "grad_norm": 30.5,
      "grad_norm_var": 2.0087890625,
      "learning_rate": 0.0001,
      "loss": 7.5892,
      "loss/crossentropy": 2.2223299980163573,
      "loss/hidden": 3.3765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19802077617496253,
      "step": 4820
    },
    {
      "epoch": 0.12075,
      "grad_norm": 30.125,
      "grad_norm_var": 3.1080729166666665,
      "learning_rate": 0.0001,
      "loss": 7.5951,
      "loss/crossentropy": 2.1982655793428423,
      "loss/hidden": 3.444140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2064062263816595,
      "step": 4830
    },
    {
      "epoch": 0.121,
      "grad_norm": 30.0,
      "grad_norm_var": 12.412434895833334,
      "learning_rate": 0.0001,
      "loss": 7.5254,
      "loss/crossentropy": 1.9753010511398315,
      "loss/hidden": 3.4484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19706314485520124,
      "step": 4840
    },
    {
      "epoch": 0.12125,
      "grad_norm": 30.125,
      "grad_norm_var": 13.1447265625,
      "learning_rate": 0.0001,
      "loss": 7.5351,
      "loss/crossentropy": 2.0562877766788006,
      "loss/hidden": 3.396875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17667807769030333,
      "step": 4850
    },
    {
      "epoch": 0.1215,
      "grad_norm": 33.0,
      "grad_norm_var": 3.465625,
      "learning_rate": 0.0001,
      "loss": 7.6024,
      "loss/crossentropy": 2.1578008987009527,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18234786652028562,
      "step": 4860
    },
    {
      "epoch": 0.12175,
      "grad_norm": 30.375,
      "grad_norm_var": 2.343489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4726,
      "loss/crossentropy": 2.1330083698034286,
      "loss/hidden": 3.333203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18738476932048798,
      "step": 4870
    },
    {
      "epoch": 0.122,
      "grad_norm": 31.625,
      "grad_norm_var": 3.2322265625,
      "learning_rate": 0.0001,
      "loss": 7.605,
      "loss/crossentropy": 2.1732089832425117,
      "loss/hidden": 3.460546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19949225690215827,
      "step": 4880
    },
    {
      "epoch": 0.12225,
      "grad_norm": 33.5,
      "grad_norm_var": 2.238997395833333,
      "learning_rate": 0.0001,
      "loss": 7.5382,
      "loss/crossentropy": 2.188927575945854,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19776681065559387,
      "step": 4890
    },
    {
      "epoch": 0.1225,
      "grad_norm": 29.0,
      "grad_norm_var": 15.819791666666667,
      "learning_rate": 0.0001,
      "loss": 7.5728,
      "loss/crossentropy": 2.202793037891388,
      "loss/hidden": 3.36328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1992744604125619,
      "step": 4900
    },
    {
      "epoch": 0.12275,
      "grad_norm": 30.25,
      "grad_norm_var": 16.1775390625,
      "learning_rate": 0.0001,
      "loss": 7.426,
      "loss/crossentropy": 2.177696394920349,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19968770015984774,
      "step": 4910
    },
    {
      "epoch": 0.123,
      "grad_norm": 35.5,
      "grad_norm_var": 7.707291666666666,
      "learning_rate": 0.0001,
      "loss": 7.7485,
      "loss/crossentropy": 2.040875867009163,
      "loss/hidden": 3.45234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1929738214239478,
      "step": 4920
    },
    {
      "epoch": 0.12325,
      "grad_norm": 32.25,
      "grad_norm_var": 6.771809895833333,
      "learning_rate": 0.0001,
      "loss": 7.5517,
      "loss/crossentropy": 2.199721184372902,
      "loss/hidden": 3.406640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20861553251743317,
      "step": 4930
    },
    {
      "epoch": 0.1235,
      "grad_norm": 28.75,
      "grad_norm_var": 4.235416666666667,
      "learning_rate": 0.0001,
      "loss": 7.5576,
      "loss/crossentropy": 2.133353302627802,
      "loss/hidden": 3.45546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21631606128066777,
      "step": 4940
    },
    {
      "epoch": 0.12375,
      "grad_norm": 40.0,
      "grad_norm_var": 15.1650390625,
      "learning_rate": 0.0001,
      "loss": 7.5129,
      "loss/crossentropy": 2.2380867928266523,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2078275766223669,
      "step": 4950
    },
    {
      "epoch": 0.124,
      "grad_norm": 32.5,
      "grad_norm_var": 9.342643229166667,
      "learning_rate": 0.0001,
      "loss": 7.4421,
      "loss/crossentropy": 2.1563473463058473,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1852986102923751,
      "step": 4960
    },
    {
      "epoch": 0.12425,
      "grad_norm": 32.5,
      "grad_norm_var": 2.3916666666666666,
      "learning_rate": 0.0001,
      "loss": 7.5503,
      "loss/crossentropy": 2.124380439519882,
      "loss/hidden": 3.480859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20270956568419934,
      "step": 4970
    },
    {
      "epoch": 0.1245,
      "grad_norm": 30.875,
      "grad_norm_var": 5.837434895833334,
      "learning_rate": 0.0001,
      "loss": 7.4857,
      "loss/crossentropy": 2.0482941284775733,
      "loss/hidden": 3.373828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18650466352701187,
      "step": 4980
    },
    {
      "epoch": 0.12475,
      "grad_norm": 31.0,
      "grad_norm_var": 3.0858723958333334,
      "learning_rate": 0.0001,
      "loss": 7.5446,
      "loss/crossentropy": 2.160063311457634,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18700905814766883,
      "step": 4990
    },
    {
      "epoch": 0.125,
      "grad_norm": 28.75,
      "grad_norm_var": 5.533333333333333,
      "learning_rate": 0.0001,
      "loss": 7.5333,
      "loss/crossentropy": 2.056824396550655,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18965724110603333,
      "step": 5000
    },
    {
      "epoch": 0.12525,
      "grad_norm": 32.5,
      "grad_norm_var": 32.25807291666667,
      "learning_rate": 0.0001,
      "loss": 7.5864,
      "loss/crossentropy": 2.2407308876514436,
      "loss/hidden": 3.51328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22782632596790792,
      "step": 5010
    },
    {
      "epoch": 0.1255,
      "grad_norm": 31.375,
      "grad_norm_var": 32.7353515625,
      "learning_rate": 0.0001,
      "loss": 7.5671,
      "loss/crossentropy": 2.18543721139431,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19135277662426234,
      "step": 5020
    },
    {
      "epoch": 0.12575,
      "grad_norm": 32.5,
      "grad_norm_var": 2.814322916666667,
      "learning_rate": 0.0001,
      "loss": 7.578,
      "loss/crossentropy": 2.1418938025832177,
      "loss/hidden": 3.362109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20001544915139674,
      "step": 5030
    },
    {
      "epoch": 0.126,
      "grad_norm": 33.0,
      "grad_norm_var": 1.3238932291666667,
      "learning_rate": 0.0001,
      "loss": 7.5594,
      "loss/crossentropy": 2.1838817209005357,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2107716018334031,
      "step": 5040
    },
    {
      "epoch": 0.12625,
      "grad_norm": 31.25,
      "grad_norm_var": 3.47890625,
      "learning_rate": 0.0001,
      "loss": 7.5275,
      "loss/crossentropy": 2.2816754072904586,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.201267159730196,
      "step": 5050
    },
    {
      "epoch": 0.1265,
      "grad_norm": 33.5,
      "grad_norm_var": 4.2087890625,
      "learning_rate": 0.0001,
      "loss": 7.6079,
      "loss/crossentropy": 2.0686806365847588,
      "loss/hidden": 3.603125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2172813605517149,
      "step": 5060
    },
    {
      "epoch": 0.12675,
      "grad_norm": 32.25,
      "grad_norm_var": 3.19765625,
      "learning_rate": 0.0001,
      "loss": 7.5561,
      "loss/crossentropy": 2.1800751775503158,
      "loss/hidden": 3.389453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20199469216167926,
      "step": 5070
    },
    {
      "epoch": 0.127,
      "grad_norm": 30.5,
      "grad_norm_var": 1.8643229166666666,
      "learning_rate": 0.0001,
      "loss": 7.4702,
      "loss/crossentropy": 2.2943209201097488,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1958466824144125,
      "step": 5080
    },
    {
      "epoch": 0.12725,
      "grad_norm": 30.125,
      "grad_norm_var": 1.9030598958333333,
      "learning_rate": 0.0001,
      "loss": 7.4383,
      "loss/crossentropy": 2.291040873527527,
      "loss/hidden": 3.308203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18960105255246162,
      "step": 5090
    },
    {
      "epoch": 0.1275,
      "grad_norm": 32.25,
      "grad_norm_var": 1.4504557291666667,
      "learning_rate": 0.0001,
      "loss": 7.6139,
      "loss/crossentropy": 2.116188834607601,
      "loss/hidden": 3.5015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19662482757121325,
      "step": 5100
    },
    {
      "epoch": 0.12775,
      "grad_norm": 32.25,
      "grad_norm_var": 1.5395182291666667,
      "learning_rate": 0.0001,
      "loss": 7.5074,
      "loss/crossentropy": 2.200083887577057,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19970641760155558,
      "step": 5110
    },
    {
      "epoch": 0.128,
      "grad_norm": 31.125,
      "grad_norm_var": 1.990625,
      "learning_rate": 0.0001,
      "loss": 7.5949,
      "loss/crossentropy": 2.080552561581135,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19426519125699998,
      "step": 5120
    },
    {
      "epoch": 0.12825,
      "grad_norm": 31.125,
      "grad_norm_var": 2.200455729166667,
      "learning_rate": 0.0001,
      "loss": 7.5036,
      "loss/crossentropy": 2.138309660553932,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20698665007948874,
      "step": 5130
    },
    {
      "epoch": 0.1285,
      "grad_norm": 30.125,
      "grad_norm_var": 5.530989583333334,
      "learning_rate": 0.0001,
      "loss": 7.6081,
      "loss/crossentropy": 2.0686957597732545,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22128727175295354,
      "step": 5140
    },
    {
      "epoch": 0.12875,
      "grad_norm": 30.75,
      "grad_norm_var": 2.120833333333333,
      "learning_rate": 0.0001,
      "loss": 7.5236,
      "loss/crossentropy": 2.0967576891183852,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19716181065887212,
      "step": 5150
    },
    {
      "epoch": 0.129,
      "grad_norm": 32.25,
      "grad_norm_var": 1.8468098958333334,
      "learning_rate": 0.0001,
      "loss": 7.5679,
      "loss/crossentropy": 2.172739614546299,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1919636652804911,
      "step": 5160
    },
    {
      "epoch": 0.12925,
      "grad_norm": 43.75,
      "grad_norm_var": 17.77890625,
      "learning_rate": 0.0001,
      "loss": 7.5027,
      "loss/crossentropy": 2.1427868396043777,
      "loss/hidden": 3.51953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20654744990170001,
      "step": 5170
    },
    {
      "epoch": 0.1295,
      "grad_norm": 33.25,
      "grad_norm_var": 61.1962890625,
      "learning_rate": 0.0001,
      "loss": 7.52,
      "loss/crossentropy": 2.1186717480421065,
      "loss/hidden": 3.42421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1938714198768139,
      "step": 5180
    },
    {
      "epoch": 0.12975,
      "grad_norm": 33.0,
      "grad_norm_var": 4.07890625,
      "learning_rate": 0.0001,
      "loss": 7.6231,
      "loss/crossentropy": 2.072993017733097,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18984810579568148,
      "step": 5190
    },
    {
      "epoch": 0.13,
      "grad_norm": 36.25,
      "grad_norm_var": 20.777083333333334,
      "learning_rate": 0.0001,
      "loss": 7.597,
      "loss/crossentropy": 2.206225660443306,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2030201606452465,
      "step": 5200
    },
    {
      "epoch": 0.13025,
      "grad_norm": 43.75,
      "grad_norm_var": 24.43515625,
      "learning_rate": 0.0001,
      "loss": 7.5075,
      "loss/crossentropy": 2.228325179219246,
      "loss/hidden": 3.366796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20732430163770915,
      "step": 5210
    },
    {
      "epoch": 0.1305,
      "grad_norm": 30.5,
      "grad_norm_var": 11.99765625,
      "learning_rate": 0.0001,
      "loss": 7.5295,
      "loss/crossentropy": 2.129982355237007,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19593548215925694,
      "step": 5220
    },
    {
      "epoch": 0.13075,
      "grad_norm": 30.75,
      "grad_norm_var": 4.292708333333334,
      "learning_rate": 0.0001,
      "loss": 7.5261,
      "loss/crossentropy": 2.1533152967691422,
      "loss/hidden": 3.3296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19487107992172242,
      "step": 5230
    },
    {
      "epoch": 0.131,
      "grad_norm": 30.875,
      "grad_norm_var": 12.989583333333334,
      "learning_rate": 0.0001,
      "loss": 7.5566,
      "loss/crossentropy": 2.195969894528389,
      "loss/hidden": 3.48046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2189876638352871,
      "step": 5240
    },
    {
      "epoch": 0.13125,
      "grad_norm": 33.25,
      "grad_norm_var": 22.575,
      "learning_rate": 0.0001,
      "loss": 7.482,
      "loss/crossentropy": 2.074573493748903,
      "loss/hidden": 3.412890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19280508980154992,
      "step": 5250
    },
    {
      "epoch": 0.1315,
      "grad_norm": 29.125,
      "grad_norm_var": 15.599739583333333,
      "learning_rate": 0.0001,
      "loss": 7.5395,
      "loss/crossentropy": 2.227403500676155,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20514250732958317,
      "step": 5260
    },
    {
      "epoch": 0.13175,
      "grad_norm": 31.625,
      "grad_norm_var": 113.15618489583333,
      "learning_rate": 0.0001,
      "loss": 7.5209,
      "loss/crossentropy": 2.055370827019215,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18272479642182587,
      "step": 5270
    },
    {
      "epoch": 0.132,
      "grad_norm": 31.625,
      "grad_norm_var": 1.765625,
      "learning_rate": 0.0001,
      "loss": 7.4798,
      "loss/crossentropy": 2.1356967806816103,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19737351574003698,
      "step": 5280
    },
    {
      "epoch": 0.13225,
      "grad_norm": 31.0,
      "grad_norm_var": 7.879622395833334,
      "learning_rate": 0.0001,
      "loss": 7.4715,
      "loss/crossentropy": 2.177875056862831,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20798433478921652,
      "step": 5290
    },
    {
      "epoch": 0.1325,
      "grad_norm": 29.5,
      "grad_norm_var": 8.760872395833333,
      "learning_rate": 0.0001,
      "loss": 7.5419,
      "loss/crossentropy": 2.182955250144005,
      "loss/hidden": 3.354296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18478553090244532,
      "step": 5300
    },
    {
      "epoch": 0.13275,
      "grad_norm": 29.125,
      "grad_norm_var": 4.260416666666667,
      "learning_rate": 0.0001,
      "loss": 7.5816,
      "loss/crossentropy": 2.0469735309481623,
      "loss/hidden": 3.55859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18724320270121098,
      "step": 5310
    },
    {
      "epoch": 0.133,
      "grad_norm": 33.0,
      "grad_norm_var": 1.8801432291666667,
      "learning_rate": 0.0001,
      "loss": 7.7815,
      "loss/crossentropy": 2.1846155911684035,
      "loss/hidden": 3.499609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21609773077070712,
      "step": 5320
    },
    {
      "epoch": 0.13325,
      "grad_norm": 30.875,
      "grad_norm_var": 5.584375,
      "learning_rate": 0.0001,
      "loss": 7.7067,
      "loss/crossentropy": 2.0075640469789504,
      "loss/hidden": 3.474609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20185804851353167,
      "step": 5330
    },
    {
      "epoch": 0.1335,
      "grad_norm": 27.375,
      "grad_norm_var": 6.120247395833333,
      "learning_rate": 0.0001,
      "loss": 7.5345,
      "loss/crossentropy": 2.058630608022213,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19536744449287652,
      "step": 5340
    },
    {
      "epoch": 0.13375,
      "grad_norm": 52.5,
      "grad_norm_var": 30.692643229166666,
      "learning_rate": 0.0001,
      "loss": 7.6655,
      "loss/crossentropy": 2.050611114501953,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19592730849981307,
      "step": 5350
    },
    {
      "epoch": 0.134,
      "grad_norm": 31.125,
      "grad_norm_var": 28.90390625,
      "learning_rate": 0.0001,
      "loss": 7.5509,
      "loss/crossentropy": 2.1665500849485397,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19354073759168386,
      "step": 5360
    },
    {
      "epoch": 0.13425,
      "grad_norm": 33.5,
      "grad_norm_var": 2.9931640625,
      "learning_rate": 0.0001,
      "loss": 7.6343,
      "loss/crossentropy": 2.1793171644210814,
      "loss/hidden": 3.3328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20293663591146469,
      "step": 5370
    },
    {
      "epoch": 0.1345,
      "grad_norm": 32.5,
      "grad_norm_var": 1.8098307291666667,
      "learning_rate": 0.0001,
      "loss": 7.6,
      "loss/crossentropy": 2.152762657403946,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21694996021687984,
      "step": 5380
    },
    {
      "epoch": 0.13475,
      "grad_norm": 30.125,
      "grad_norm_var": 1.4301432291666667,
      "learning_rate": 0.0001,
      "loss": 7.5121,
      "loss/crossentropy": 2.083320555835962,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18746816255152227,
      "step": 5390
    },
    {
      "epoch": 0.135,
      "grad_norm": 29.625,
      "grad_norm_var": 5.109830729166666,
      "learning_rate": 0.0001,
      "loss": 7.4172,
      "loss/crossentropy": 2.0639937698841093,
      "loss/hidden": 3.3875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18618469405919313,
      "step": 5400
    },
    {
      "epoch": 0.13525,
      "grad_norm": 31.125,
      "grad_norm_var": 12.337434895833333,
      "learning_rate": 0.0001,
      "loss": 7.5749,
      "loss/crossentropy": 2.068472331762314,
      "loss/hidden": 3.524609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18952864613384007,
      "step": 5410
    },
    {
      "epoch": 0.1355,
      "grad_norm": 32.25,
      "grad_norm_var": 7.9541015625,
      "learning_rate": 0.0001,
      "loss": 7.5385,
      "loss/crossentropy": 2.085490897297859,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1854228163138032,
      "step": 5420
    },
    {
      "epoch": 0.13575,
      "grad_norm": 30.0,
      "grad_norm_var": 7.678125,
      "learning_rate": 0.0001,
      "loss": 7.5583,
      "loss/crossentropy": 2.1747709423303605,
      "loss/hidden": 3.444921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20177022367715836,
      "step": 5430
    },
    {
      "epoch": 0.136,
      "grad_norm": 31.875,
      "grad_norm_var": 2.6541015625,
      "learning_rate": 0.0001,
      "loss": 7.4911,
      "loss/crossentropy": 2.167823739349842,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2082101447507739,
      "step": 5440
    },
    {
      "epoch": 0.13625,
      "grad_norm": 30.625,
      "grad_norm_var": 11.2087890625,
      "learning_rate": 0.0001,
      "loss": 7.639,
      "loss/crossentropy": 2.1367167800664904,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20548735409975052,
      "step": 5450
    },
    {
      "epoch": 0.1365,
      "grad_norm": 28.375,
      "grad_norm_var": 15.654166666666667,
      "learning_rate": 0.0001,
      "loss": 7.4951,
      "loss/crossentropy": 2.109811532497406,
      "loss/hidden": 3.4640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1994595667347312,
      "step": 5460
    },
    {
      "epoch": 0.13675,
      "grad_norm": 31.75,
      "grad_norm_var": 4.463997395833333,
      "learning_rate": 0.0001,
      "loss": 7.5263,
      "loss/crossentropy": 2.2288592010736465,
      "loss/hidden": 3.4140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20487434454262257,
      "step": 5470
    },
    {
      "epoch": 0.137,
      "grad_norm": 32.0,
      "grad_norm_var": 3.8400390625,
      "learning_rate": 0.0001,
      "loss": 7.5681,
      "loss/crossentropy": 2.0951829612255097,
      "loss/hidden": 3.455078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20034745894372463,
      "step": 5480
    },
    {
      "epoch": 0.13725,
      "grad_norm": 31.625,
      "grad_norm_var": 2.299934895833333,
      "learning_rate": 0.0001,
      "loss": 7.4772,
      "loss/crossentropy": 2.022165683656931,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18401878620497883,
      "step": 5490
    },
    {
      "epoch": 0.1375,
      "grad_norm": 31.625,
      "grad_norm_var": 4.205989583333333,
      "learning_rate": 0.0001,
      "loss": 7.6869,
      "loss/crossentropy": 2.095045933127403,
      "loss/hidden": 3.609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21982598043978213,
      "step": 5500
    },
    {
      "epoch": 0.13775,
      "grad_norm": 29.875,
      "grad_norm_var": 2.177018229166667,
      "learning_rate": 0.0001,
      "loss": 7.4483,
      "loss/crossentropy": 2.07875951230526,
      "loss/hidden": 3.2890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1898285737261176,
      "step": 5510
    },
    {
      "epoch": 0.138,
      "grad_norm": 30.0,
      "grad_norm_var": 3.4181640625,
      "learning_rate": 0.0001,
      "loss": 7.5258,
      "loss/crossentropy": 2.194947564601898,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21397264376282693,
      "step": 5520
    },
    {
      "epoch": 0.13825,
      "grad_norm": 32.75,
      "grad_norm_var": 5.5541015625,
      "learning_rate": 0.0001,
      "loss": 7.6574,
      "loss/crossentropy": 2.2282156944274902,
      "loss/hidden": 3.375390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19613583907485008,
      "step": 5530
    },
    {
      "epoch": 0.1385,
      "grad_norm": 29.875,
      "grad_norm_var": 4.927018229166666,
      "learning_rate": 0.0001,
      "loss": 7.5361,
      "loss/crossentropy": 2.1092587068676947,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894394876435399,
      "step": 5540
    },
    {
      "epoch": 0.13875,
      "grad_norm": 30.875,
      "grad_norm_var": 6.080989583333333,
      "learning_rate": 0.0001,
      "loss": 7.6223,
      "loss/crossentropy": 2.189313694834709,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.210250511392951,
      "step": 5550
    },
    {
      "epoch": 0.139,
      "grad_norm": 41.25,
      "grad_norm_var": 8.976497395833333,
      "learning_rate": 0.0001,
      "loss": 7.4876,
      "loss/crossentropy": 2.046143325418234,
      "loss/hidden": 3.457421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19476603530347347,
      "step": 5560
    },
    {
      "epoch": 0.13925,
      "grad_norm": 31.25,
      "grad_norm_var": 9.076497395833334,
      "learning_rate": 0.0001,
      "loss": 7.5338,
      "loss/crossentropy": 2.1882025837898254,
      "loss/hidden": 3.45,
      "loss/jsd": 0.0,
      "loss/logits": 0.2058427443727851,
      "step": 5570
    },
    {
      "epoch": 0.1395,
      "grad_norm": 30.25,
      "grad_norm_var": 182.91875,
      "learning_rate": 0.0001,
      "loss": 7.6344,
      "loss/crossentropy": 2.160731779038906,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22695957981050013,
      "step": 5580
    },
    {
      "epoch": 0.13975,
      "grad_norm": 31.0,
      "grad_norm_var": 4.140559895833333,
      "learning_rate": 0.0001,
      "loss": 7.5419,
      "loss/crossentropy": 1.8924851581454276,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1802460763603449,
      "step": 5590
    },
    {
      "epoch": 0.14,
      "grad_norm": 30.25,
      "grad_norm_var": 2.084830729166667,
      "learning_rate": 0.0001,
      "loss": 7.5498,
      "loss/crossentropy": 2.242933538556099,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22144901510328055,
      "step": 5600
    },
    {
      "epoch": 0.14025,
      "grad_norm": 29.75,
      "grad_norm_var": 1.5622395833333333,
      "learning_rate": 0.0001,
      "loss": 7.5133,
      "loss/crossentropy": 2.132666201889515,
      "loss/hidden": 3.492578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19677439089864493,
      "step": 5610
    },
    {
      "epoch": 0.1405,
      "grad_norm": 30.5,
      "grad_norm_var": 14.234375,
      "learning_rate": 0.0001,
      "loss": 7.6226,
      "loss/crossentropy": 2.0671760708093645,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20731164887547493,
      "step": 5620
    },
    {
      "epoch": 0.14075,
      "grad_norm": 35.5,
      "grad_norm_var": 14.44375,
      "learning_rate": 0.0001,
      "loss": 7.5063,
      "loss/crossentropy": 2.060420323908329,
      "loss/hidden": 3.346484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18428542967885733,
      "step": 5630
    },
    {
      "epoch": 0.141,
      "grad_norm": 38.5,
      "grad_norm_var": 19.1509765625,
      "learning_rate": 0.0001,
      "loss": 7.5322,
      "loss/crossentropy": 2.2572733104228973,
      "loss/hidden": 3.291015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19184892270714043,
      "step": 5640
    },
    {
      "epoch": 0.14125,
      "grad_norm": 32.0,
      "grad_norm_var": 20.373893229166665,
      "learning_rate": 0.0001,
      "loss": 7.5473,
      "loss/crossentropy": 2.1559308364987375,
      "loss/hidden": 3.237109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18201812207698823,
      "step": 5650
    },
    {
      "epoch": 0.1415,
      "grad_norm": 28.5,
      "grad_norm_var": 10.033268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4529,
      "loss/crossentropy": 2.1479016572237013,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20532293617725372,
      "step": 5660
    },
    {
      "epoch": 0.14175,
      "grad_norm": 29.5,
      "grad_norm_var": 3.6332682291666667,
      "learning_rate": 0.0001,
      "loss": 7.6001,
      "loss/crossentropy": 2.140682426095009,
      "loss/hidden": 3.321484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18856783863157034,
      "step": 5670
    },
    {
      "epoch": 0.142,
      "grad_norm": 32.75,
      "grad_norm_var": 2.84140625,
      "learning_rate": 0.0001,
      "loss": 7.6348,
      "loss/crossentropy": 2.25071659386158,
      "loss/hidden": 3.312109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2029257183894515,
      "step": 5680
    },
    {
      "epoch": 0.14225,
      "grad_norm": 31.75,
      "grad_norm_var": 2.643684895833333,
      "learning_rate": 0.0001,
      "loss": 7.5323,
      "loss/crossentropy": 2.149606391787529,
      "loss/hidden": 3.436328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2141027996316552,
      "step": 5690
    },
    {
      "epoch": 0.1425,
      "grad_norm": 30.25,
      "grad_norm_var": 2.32890625,
      "learning_rate": 0.0001,
      "loss": 7.5615,
      "loss/crossentropy": 2.2273970007896424,
      "loss/hidden": 3.4703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22348886616528035,
      "step": 5700
    },
    {
      "epoch": 0.14275,
      "grad_norm": 29.875,
      "grad_norm_var": 1.8285807291666667,
      "learning_rate": 0.0001,
      "loss": 7.4366,
      "loss/crossentropy": 2.2010063380002975,
      "loss/hidden": 3.289453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18594364672899247,
      "step": 5710
    },
    {
      "epoch": 0.143,
      "grad_norm": 34.25,
      "grad_norm_var": 2.6020182291666667,
      "learning_rate": 0.0001,
      "loss": 7.508,
      "loss/crossentropy": 2.0483784288167954,
      "loss/hidden": 3.40234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2091974811628461,
      "step": 5720
    },
    {
      "epoch": 0.14325,
      "grad_norm": 33.0,
      "grad_norm_var": 2.6744140625,
      "learning_rate": 0.0001,
      "loss": 7.6646,
      "loss/crossentropy": 2.025853230059147,
      "loss/hidden": 3.589453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2227605242282152,
      "step": 5730
    },
    {
      "epoch": 0.1435,
      "grad_norm": 31.0,
      "grad_norm_var": 1.8327473958333333,
      "learning_rate": 0.0001,
      "loss": 7.6064,
      "loss/crossentropy": 2.075891287624836,
      "loss/hidden": 3.38359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1890827091410756,
      "step": 5740
    },
    {
      "epoch": 0.14375,
      "grad_norm": 31.625,
      "grad_norm_var": 13.943489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4294,
      "loss/crossentropy": 2.0676268830895426,
      "loss/hidden": 3.426953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20394732002168894,
      "step": 5750
    },
    {
      "epoch": 0.144,
      "grad_norm": 30.0,
      "grad_norm_var": 14.658072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5958,
      "loss/crossentropy": 2.1388431686908005,
      "loss/hidden": 3.499609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21454429486766458,
      "step": 5760
    },
    {
      "epoch": 0.14425,
      "grad_norm": 32.0,
      "grad_norm_var": 13.958072916666667,
      "learning_rate": 0.0001,
      "loss": 7.5342,
      "loss/crossentropy": 2.0246446818113326,
      "loss/hidden": 3.3140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18158567249774932,
      "step": 5770
    },
    {
      "epoch": 0.1445,
      "grad_norm": 28.375,
      "grad_norm_var": 14.469205729166667,
      "learning_rate": 0.0001,
      "loss": 7.5141,
      "loss/crossentropy": 2.103305173665285,
      "loss/hidden": 3.3234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1842075905762613,
      "step": 5780
    },
    {
      "epoch": 0.14475,
      "grad_norm": 28.875,
      "grad_norm_var": 3.160872395833333,
      "learning_rate": 0.0001,
      "loss": 7.6213,
      "loss/crossentropy": 2.096319726109505,
      "loss/hidden": 3.378515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19167389422655107,
      "step": 5790
    },
    {
      "epoch": 0.145,
      "grad_norm": 35.25,
      "grad_norm_var": 3.4916666666666667,
      "learning_rate": 0.0001,
      "loss": 7.4461,
      "loss/crossentropy": 2.0800597339868547,
      "loss/hidden": 3.3453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18269639350473882,
      "step": 5800
    },
    {
      "epoch": 0.14525,
      "grad_norm": 45.5,
      "grad_norm_var": 22.3306640625,
      "learning_rate": 0.0001,
      "loss": 7.4855,
      "loss/crossentropy": 2.036512078344822,
      "loss/hidden": 3.462890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1910469362512231,
      "step": 5810
    },
    {
      "epoch": 0.1455,
      "grad_norm": 33.0,
      "grad_norm_var": 23.284309895833335,
      "learning_rate": 0.0001,
      "loss": 7.4838,
      "loss/crossentropy": 2.1922324389219283,
      "loss/hidden": 3.48671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2080824663862586,
      "step": 5820
    },
    {
      "epoch": 0.14575,
      "grad_norm": 31.25,
      "grad_norm_var": 2.7171223958333335,
      "learning_rate": 0.0001,
      "loss": 7.4657,
      "loss/crossentropy": 2.0417690485715867,
      "loss/hidden": 3.4546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19891292043030262,
      "step": 5830
    },
    {
      "epoch": 0.146,
      "grad_norm": 32.75,
      "grad_norm_var": 6.515625,
      "learning_rate": 0.0001,
      "loss": 7.5864,
      "loss/crossentropy": 2.12301287651062,
      "loss/hidden": 3.2953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19781142249703407,
      "step": 5840
    },
    {
      "epoch": 0.14625,
      "grad_norm": 33.0,
      "grad_norm_var": 6.574934895833334,
      "learning_rate": 0.0001,
      "loss": 7.6264,
      "loss/crossentropy": 2.1600560665130617,
      "loss/hidden": 3.448046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22095333114266397,
      "step": 5850
    },
    {
      "epoch": 0.1465,
      "grad_norm": 31.125,
      "grad_norm_var": 1.3822265625,
      "learning_rate": 0.0001,
      "loss": 7.5231,
      "loss/crossentropy": 2.185663253068924,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2029089467599988,
      "step": 5860
    },
    {
      "epoch": 0.14675,
      "grad_norm": 33.0,
      "grad_norm_var": 3.020247395833333,
      "learning_rate": 0.0001,
      "loss": 7.4938,
      "loss/crossentropy": 2.052996274828911,
      "loss/hidden": 3.477734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19343450702726842,
      "step": 5870
    },
    {
      "epoch": 0.147,
      "grad_norm": 31.875,
      "grad_norm_var": 6.112239583333333,
      "learning_rate": 0.0001,
      "loss": 7.5855,
      "loss/crossentropy": 2.2062906324863434,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20249157436192036,
      "step": 5880
    },
    {
      "epoch": 0.14725,
      "grad_norm": 36.25,
      "grad_norm_var": 2.4330729166666667,
      "learning_rate": 0.0001,
      "loss": 7.5644,
      "loss/crossentropy": 2.0493671208620072,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19113321453332902,
      "step": 5890
    },
    {
      "epoch": 0.1475,
      "grad_norm": 31.5,
      "grad_norm_var": 2.6968098958333333,
      "learning_rate": 0.0001,
      "loss": 7.5941,
      "loss/crossentropy": 2.20456420481205,
      "loss/hidden": 3.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20127719175070524,
      "step": 5900
    },
    {
      "epoch": 0.14775,
      "grad_norm": 29.625,
      "grad_norm_var": 1.8268229166666667,
      "learning_rate": 0.0001,
      "loss": 7.4958,
      "loss/crossentropy": 2.2630672723054888,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20422303341329098,
      "step": 5910
    },
    {
      "epoch": 0.148,
      "grad_norm": 33.5,
      "grad_norm_var": 2.45,
      "learning_rate": 0.0001,
      "loss": 7.4893,
      "loss/crossentropy": 2.130161929130554,
      "loss/hidden": 3.442578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2029714561998844,
      "step": 5920
    },
    {
      "epoch": 0.14825,
      "grad_norm": 31.375,
      "grad_norm_var": 0.9728515625,
      "learning_rate": 0.0001,
      "loss": 7.6087,
      "loss/crossentropy": 2.2219777315855027,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20015078466385602,
      "step": 5930
    },
    {
      "epoch": 0.1485,
      "grad_norm": 31.125,
      "grad_norm_var": 2.9514973958333335,
      "learning_rate": 0.0001,
      "loss": 7.5908,
      "loss/crossentropy": 2.2077976912260056,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1996950874105096,
      "step": 5940
    },
    {
      "epoch": 0.14875,
      "grad_norm": 30.0,
      "grad_norm_var": 3.3619140625,
      "learning_rate": 0.0001,
      "loss": 7.5402,
      "loss/crossentropy": 2.1866880118846894,
      "loss/hidden": 3.408203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19317448064684867,
      "step": 5950
    },
    {
      "epoch": 0.149,
      "grad_norm": 31.625,
      "grad_norm_var": 2.6968098958333333,
      "learning_rate": 0.0001,
      "loss": 7.4202,
      "loss/crossentropy": 2.1943173080682756,
      "loss/hidden": 3.2390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18432049825787544,
      "step": 5960
    },
    {
      "epoch": 0.14925,
      "grad_norm": 30.75,
      "grad_norm_var": 16.936393229166665,
      "learning_rate": 0.0001,
      "loss": 7.5056,
      "loss/crossentropy": 2.17142014503479,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1876060863956809,
      "step": 5970
    },
    {
      "epoch": 0.1495,
      "grad_norm": 31.5,
      "grad_norm_var": 4.03515625,
      "learning_rate": 0.0001,
      "loss": 7.5442,
      "loss/crossentropy": 2.2047942698001863,
      "loss/hidden": 3.369140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1988227294757962,
      "step": 5980
    },
    {
      "epoch": 0.14975,
      "grad_norm": 32.0,
      "grad_norm_var": 3.332291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5693,
      "loss/crossentropy": 2.281469625234604,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19730366189032794,
      "step": 5990
    },
    {
      "epoch": 0.15,
      "grad_norm": 30.125,
      "grad_norm_var": 3.7603515625,
      "learning_rate": 0.0001,
      "loss": 7.4556,
      "loss/crossentropy": 2.256136628985405,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1971954844892025,
      "step": 6000
    },
    {
      "epoch": 0.15025,
      "grad_norm": 30.0,
      "grad_norm_var": 2.1285807291666665,
      "learning_rate": 0.0001,
      "loss": 7.5146,
      "loss/crossentropy": 2.105835199356079,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19158641248941422,
      "step": 6010
    },
    {
      "epoch": 0.1505,
      "grad_norm": 40.0,
      "grad_norm_var": 18.542708333333334,
      "learning_rate": 0.0001,
      "loss": 7.4687,
      "loss/crossentropy": 2.095852518081665,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18887621480971575,
      "step": 6020
    },
    {
      "epoch": 0.15075,
      "grad_norm": 30.5,
      "grad_norm_var": 19.169791666666665,
      "learning_rate": 0.0001,
      "loss": 7.6293,
      "loss/crossentropy": 2.0981243371963503,
      "loss/hidden": 3.4875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22096464578062297,
      "step": 6030
    },
    {
      "epoch": 0.151,
      "grad_norm": 30.5,
      "grad_norm_var": 2.716080729166667,
      "learning_rate": 0.0001,
      "loss": 7.5935,
      "loss/crossentropy": 2.212895154953003,
      "loss/hidden": 3.510546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22050899863243104,
      "step": 6040
    },
    {
      "epoch": 0.15125,
      "grad_norm": 30.625,
      "grad_norm_var": 3.042122395833333,
      "learning_rate": 0.0001,
      "loss": 7.5,
      "loss/crossentropy": 2.1280829131603243,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18497859146445988,
      "step": 6050
    },
    {
      "epoch": 0.1515,
      "grad_norm": 29.25,
      "grad_norm_var": 47.15807291666667,
      "learning_rate": 0.0001,
      "loss": 7.5469,
      "loss/crossentropy": 2.1731619790196417,
      "loss/hidden": 3.4578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22194090783596038,
      "step": 6060
    },
    {
      "epoch": 0.15175,
      "grad_norm": 30.5,
      "grad_norm_var": 3.218489583333333,
      "learning_rate": 0.0001,
      "loss": 7.4378,
      "loss/crossentropy": 2.047077566385269,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1876080146059394,
      "step": 6070
    },
    {
      "epoch": 0.152,
      "grad_norm": 33.0,
      "grad_norm_var": 5.45,
      "learning_rate": 0.0001,
      "loss": 7.584,
      "loss/crossentropy": 2.1896773248910906,
      "loss/hidden": 3.341796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19448568094521762,
      "step": 6080
    },
    {
      "epoch": 0.15225,
      "grad_norm": 31.125,
      "grad_norm_var": 3.7634765625,
      "learning_rate": 0.0001,
      "loss": 7.6047,
      "loss/crossentropy": 2.1123378753662108,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1931258851662278,
      "step": 6090
    },
    {
      "epoch": 0.1525,
      "grad_norm": 30.625,
      "grad_norm_var": 2.5488932291666666,
      "learning_rate": 0.0001,
      "loss": 7.5301,
      "loss/crossentropy": 2.2152541011571882,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19427606835961342,
      "step": 6100
    },
    {
      "epoch": 0.15275,
      "grad_norm": 34.25,
      "grad_norm_var": 3.7421223958333334,
      "learning_rate": 0.0001,
      "loss": 7.5936,
      "loss/crossentropy": 2.109275272488594,
      "loss/hidden": 3.531640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2041168488562107,
      "step": 6110
    },
    {
      "epoch": 0.153,
      "grad_norm": 30.0,
      "grad_norm_var": 2.723893229166667,
      "learning_rate": 0.0001,
      "loss": 7.5107,
      "loss/crossentropy": 2.0911407291889192,
      "loss/hidden": 3.403515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19639678560197354,
      "step": 6120
    },
    {
      "epoch": 0.15325,
      "grad_norm": 36.0,
      "grad_norm_var": 3.2561848958333335,
      "learning_rate": 0.0001,
      "loss": 7.4788,
      "loss/crossentropy": 2.1557929456233977,
      "loss/hidden": 3.5140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2058223508298397,
      "step": 6130
    },
    {
      "epoch": 0.1535,
      "grad_norm": 30.0,
      "grad_norm_var": 84.3337890625,
      "learning_rate": 0.0001,
      "loss": 7.5656,
      "loss/crossentropy": 2.0974055036902426,
      "loss/hidden": 3.4140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19299208922311664,
      "step": 6140
    },
    {
      "epoch": 0.15375,
      "grad_norm": 30.625,
      "grad_norm_var": 85.70149739583333,
      "learning_rate": 0.0001,
      "loss": 7.4607,
      "loss/crossentropy": 2.0354842752218247,
      "loss/hidden": 3.418359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2099142510443926,
      "step": 6150
    },
    {
      "epoch": 0.154,
      "grad_norm": 29.75,
      "grad_norm_var": 9.5134765625,
      "learning_rate": 0.0001,
      "loss": 7.4221,
      "loss/crossentropy": 1.9916995614767075,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17939293198287487,
      "step": 6160
    },
    {
      "epoch": 0.15425,
      "grad_norm": 33.5,
      "grad_norm_var": 10.60390625,
      "learning_rate": 0.0001,
      "loss": 7.5836,
      "loss/crossentropy": 2.097960978746414,
      "loss/hidden": 3.353515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1908213123679161,
      "step": 6170
    },
    {
      "epoch": 0.1545,
      "grad_norm": 31.125,
      "grad_norm_var": 3.0942057291666667,
      "learning_rate": 0.0001,
      "loss": 7.5073,
      "loss/crossentropy": 2.253797325491905,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19871626775711776,
      "step": 6180
    },
    {
      "epoch": 0.15475,
      "grad_norm": 30.375,
      "grad_norm_var": 1.3900390625,
      "learning_rate": 0.0001,
      "loss": 7.5189,
      "loss/crossentropy": 2.1756670624017715,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19654417987912892,
      "step": 6190
    },
    {
      "epoch": 0.155,
      "grad_norm": 32.75,
      "grad_norm_var": 1.4181640625,
      "learning_rate": 0.0001,
      "loss": 7.3147,
      "loss/crossentropy": 2.1110276162624357,
      "loss/hidden": 3.294921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1801340376958251,
      "step": 6200
    },
    {
      "epoch": 0.15525,
      "grad_norm": 28.5,
      "grad_norm_var": 1.6223307291666667,
      "learning_rate": 0.0001,
      "loss": 7.6015,
      "loss/crossentropy": 2.1547244489192963,
      "loss/hidden": 3.31171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18071307614445686,
      "step": 6210
    },
    {
      "epoch": 0.1555,
      "grad_norm": 33.0,
      "grad_norm_var": 4.36875,
      "learning_rate": 0.0001,
      "loss": 7.5559,
      "loss/crossentropy": 2.1889162242412565,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.185367326810956,
      "step": 6220
    },
    {
      "epoch": 0.15575,
      "grad_norm": 38.75,
      "grad_norm_var": 25.381184895833332,
      "learning_rate": 0.0001,
      "loss": 7.5323,
      "loss/crossentropy": 2.023681116104126,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19041543006896972,
      "step": 6230
    },
    {
      "epoch": 0.156,
      "grad_norm": 31.5,
      "grad_norm_var": 12.545833333333333,
      "learning_rate": 0.0001,
      "loss": 7.6086,
      "loss/crossentropy": 2.260247975587845,
      "loss/hidden": 3.324609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19489852003753186,
      "step": 6240
    },
    {
      "epoch": 0.15625,
      "grad_norm": 30.5,
      "grad_norm_var": 6.642122395833334,
      "learning_rate": 0.0001,
      "loss": 7.5419,
      "loss/crossentropy": 2.1487904608249666,
      "loss/hidden": 3.37421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1957969294860959,
      "step": 6250
    },
    {
      "epoch": 0.1565,
      "grad_norm": 31.25,
      "grad_norm_var": 5.493489583333333,
      "learning_rate": 0.0001,
      "loss": 7.588,
      "loss/crossentropy": 2.279403430223465,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20682295374572277,
      "step": 6260
    },
    {
      "epoch": 0.15675,
      "grad_norm": 30.125,
      "grad_norm_var": 8.14765625,
      "learning_rate": 0.0001,
      "loss": 7.5894,
      "loss/crossentropy": 2.1407866299152376,
      "loss/hidden": 3.426953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20337907522916793,
      "step": 6270
    },
    {
      "epoch": 0.157,
      "grad_norm": 33.25,
      "grad_norm_var": 1314.8994140625,
      "learning_rate": 0.0001,
      "loss": 7.6872,
      "loss/crossentropy": 2.1802447110414507,
      "loss/hidden": 3.540625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22727628983557224,
      "step": 6280
    },
    {
      "epoch": 0.15725,
      "grad_norm": 30.625,
      "grad_norm_var": 1378.0379557291667,
      "learning_rate": 0.0001,
      "loss": 7.5596,
      "loss/crossentropy": 2.1761491730809213,
      "loss/hidden": 3.335546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18902392424643039,
      "step": 6290
    },
    {
      "epoch": 0.1575,
      "grad_norm": 42.25,
      "grad_norm_var": 146.03020833333332,
      "learning_rate": 0.0001,
      "loss": 7.6132,
      "loss/crossentropy": 2.106969301402569,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2025001836940646,
      "step": 6300
    },
    {
      "epoch": 0.15775,
      "grad_norm": 29.25,
      "grad_norm_var": 14.539322916666666,
      "learning_rate": 0.0001,
      "loss": 7.514,
      "loss/crossentropy": 2.0654825627803803,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19361731074750424,
      "step": 6310
    },
    {
      "epoch": 0.158,
      "grad_norm": 31.75,
      "grad_norm_var": 8.153580729166666,
      "learning_rate": 0.0001,
      "loss": 7.5587,
      "loss/crossentropy": 2.025081543624401,
      "loss/hidden": 3.426953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2247559091076255,
      "step": 6320
    },
    {
      "epoch": 0.15825,
      "grad_norm": 31.875,
      "grad_norm_var": 2.0431640625,
      "learning_rate": 0.0001,
      "loss": 7.4657,
      "loss/crossentropy": 2.136493813991547,
      "loss/hidden": 3.35546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19502629730850457,
      "step": 6330
    },
    {
      "epoch": 0.1585,
      "grad_norm": 33.0,
      "grad_norm_var": 2.189322916666667,
      "learning_rate": 0.0001,
      "loss": 7.5454,
      "loss/crossentropy": 2.176800549030304,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.201618373952806,
      "step": 6340
    },
    {
      "epoch": 0.15875,
      "grad_norm": 28.375,
      "grad_norm_var": 2.97265625,
      "learning_rate": 0.0001,
      "loss": 7.5116,
      "loss/crossentropy": 2.2223973661661147,
      "loss/hidden": 3.43359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2056947773322463,
      "step": 6350
    },
    {
      "epoch": 0.159,
      "grad_norm": 31.5,
      "grad_norm_var": 3.0580729166666667,
      "learning_rate": 0.0001,
      "loss": 7.5133,
      "loss/crossentropy": 2.147826671600342,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19723634477704763,
      "step": 6360
    },
    {
      "epoch": 0.15925,
      "grad_norm": 35.0,
      "grad_norm_var": 18.772916666666667,
      "learning_rate": 0.0001,
      "loss": 7.679,
      "loss/crossentropy": 2.079222206771374,
      "loss/hidden": 3.47890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2045620433986187,
      "step": 6370
    },
    {
      "epoch": 0.1595,
      "grad_norm": 28.75,
      "grad_norm_var": 20.639518229166665,
      "learning_rate": 0.0001,
      "loss": 7.5245,
      "loss/crossentropy": 2.163569709658623,
      "loss/hidden": 3.45546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20100143328309059,
      "step": 6380
    },
    {
      "epoch": 0.15975,
      "grad_norm": 36.75,
      "grad_norm_var": 18.972916666666666,
      "learning_rate": 0.0001,
      "loss": 7.5705,
      "loss/crossentropy": 2.0975339651107787,
      "loss/hidden": 3.435546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19847188219428064,
      "step": 6390
    },
    {
      "epoch": 0.16,
      "grad_norm": 31.375,
      "grad_norm_var": 64.11770833333334,
      "learning_rate": 0.0001,
      "loss": 7.6051,
      "loss/crossentropy": 2.0802886128425597,
      "loss/hidden": 3.381640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20291686560958624,
      "step": 6400
    },
    {
      "epoch": 0.16025,
      "grad_norm": 32.25,
      "grad_norm_var": 74.28958333333334,
      "learning_rate": 0.0001,
      "loss": 7.4922,
      "loss/crossentropy": 2.1224244251847266,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.194674601405859,
      "step": 6410
    },
    {
      "epoch": 0.1605,
      "grad_norm": 62.5,
      "grad_norm_var": 69.86041666666667,
      "learning_rate": 0.0001,
      "loss": 7.6584,
      "loss/crossentropy": 2.1936039954423903,
      "loss/hidden": 3.348828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19804685264825822,
      "step": 6420
    },
    {
      "epoch": 0.16075,
      "grad_norm": 36.5,
      "grad_norm_var": 66.04576822916667,
      "learning_rate": 0.0001,
      "loss": 7.4329,
      "loss/crossentropy": 2.103436988592148,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19700358025729656,
      "step": 6430
    },
    {
      "epoch": 0.161,
      "grad_norm": 28.875,
      "grad_norm_var": 8.187239583333334,
      "learning_rate": 0.0001,
      "loss": 7.5397,
      "loss/crossentropy": 2.107574874162674,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1931281227618456,
      "step": 6440
    },
    {
      "epoch": 0.16125,
      "grad_norm": 32.75,
      "grad_norm_var": 20.245768229166668,
      "learning_rate": 0.0001,
      "loss": 7.5289,
      "loss/crossentropy": 2.251564306020737,
      "loss/hidden": 3.4125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20565700754523278,
      "step": 6450
    },
    {
      "epoch": 0.1615,
      "grad_norm": 29.5,
      "grad_norm_var": 3.187239583333333,
      "learning_rate": 0.0001,
      "loss": 7.4563,
      "loss/crossentropy": 2.0432717867195604,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17879284229129552,
      "step": 6460
    },
    {
      "epoch": 0.16175,
      "grad_norm": 40.0,
      "grad_norm_var": 8.5416015625,
      "learning_rate": 0.0001,
      "loss": 7.5052,
      "loss/crossentropy": 2.130405417084694,
      "loss/hidden": 3.423046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1868636028841138,
      "step": 6470
    },
    {
      "epoch": 0.162,
      "grad_norm": 30.625,
      "grad_norm_var": 8.284375,
      "learning_rate": 0.0001,
      "loss": 7.4387,
      "loss/crossentropy": 2.1113599717617033,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20160141140222548,
      "step": 6480
    },
    {
      "epoch": 0.16225,
      "grad_norm": 30.5,
      "grad_norm_var": 10.103125,
      "learning_rate": 0.0001,
      "loss": 7.4812,
      "loss/crossentropy": 2.177107959985733,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19870908968150616,
      "step": 6490
    },
    {
      "epoch": 0.1625,
      "grad_norm": 29.0,
      "grad_norm_var": 12.816080729166666,
      "learning_rate": 0.0001,
      "loss": 7.5634,
      "loss/crossentropy": 2.2469130218029023,
      "loss/hidden": 3.4421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22473043091595174,
      "step": 6500
    },
    {
      "epoch": 0.16275,
      "grad_norm": 62.75,
      "grad_norm_var": 67.71555989583334,
      "learning_rate": 0.0001,
      "loss": 7.4504,
      "loss/crossentropy": 2.1134796291589737,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20341392308473588,
      "step": 6510
    },
    {
      "epoch": 0.163,
      "grad_norm": 38.5,
      "grad_norm_var": 132.73723958333332,
      "learning_rate": 0.0001,
      "loss": 7.6099,
      "loss/crossentropy": 2.0655704945325852,
      "loss/hidden": 3.45625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2016578745096922,
      "step": 6520
    },
    {
      "epoch": 0.16325,
      "grad_norm": 32.0,
      "grad_norm_var": 117.9541015625,
      "learning_rate": 0.0001,
      "loss": 7.668,
      "loss/crossentropy": 2.171022225916386,
      "loss/hidden": 3.575,
      "loss/jsd": 0.0,
      "loss/logits": 0.23951137959957122,
      "step": 6530
    },
    {
      "epoch": 0.1635,
      "grad_norm": 32.75,
      "grad_norm_var": 6.513541666666667,
      "learning_rate": 0.0001,
      "loss": 7.5344,
      "loss/crossentropy": 2.1631700932979583,
      "loss/hidden": 3.371875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19134688563644886,
      "step": 6540
    },
    {
      "epoch": 0.16375,
      "grad_norm": 31.25,
      "grad_norm_var": 10.718489583333334,
      "learning_rate": 0.0001,
      "loss": 7.5036,
      "loss/crossentropy": 2.154293045401573,
      "loss/hidden": 3.355859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2056902015581727,
      "step": 6550
    },
    {
      "epoch": 0.164,
      "grad_norm": 30.0,
      "grad_norm_var": 4.233268229166667,
      "learning_rate": 0.0001,
      "loss": 7.4959,
      "loss/crossentropy": 2.077804160118103,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18005848359316587,
      "step": 6560
    },
    {
      "epoch": 0.16425,
      "grad_norm": 31.75,
      "grad_norm_var": 3.2343098958333334,
      "learning_rate": 0.0001,
      "loss": 7.3855,
      "loss/crossentropy": 2.238145849108696,
      "loss/hidden": 3.3515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20202569160610437,
      "step": 6570
    },
    {
      "epoch": 0.1645,
      "grad_norm": 29.25,
      "grad_norm_var": 4.284375,
      "learning_rate": 0.0001,
      "loss": 7.5453,
      "loss/crossentropy": 2.1169906362891195,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1980356415733695,
      "step": 6580
    },
    {
      "epoch": 0.16475,
      "grad_norm": 32.5,
      "grad_norm_var": 1.6848307291666667,
      "learning_rate": 0.0001,
      "loss": 7.5151,
      "loss/crossentropy": 2.165386658906937,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19807699434459208,
      "step": 6590
    },
    {
      "epoch": 0.165,
      "grad_norm": 30.125,
      "grad_norm_var": 2.905989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4242,
      "loss/crossentropy": 2.2100414454936983,
      "loss/hidden": 3.452734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21563767176121473,
      "step": 6600
    },
    {
      "epoch": 0.16525,
      "grad_norm": 28.25,
      "grad_norm_var": 21.574934895833334,
      "learning_rate": 0.0001,
      "loss": 7.4757,
      "loss/crossentropy": 2.1352153360843658,
      "loss/hidden": 3.277734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.176172699034214,
      "step": 6610
    },
    {
      "epoch": 0.1655,
      "grad_norm": 32.0,
      "grad_norm_var": 9.960872395833333,
      "learning_rate": 0.0001,
      "loss": 7.5596,
      "loss/crossentropy": 2.1773680597543716,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18952815048396587,
      "step": 6620
    },
    {
      "epoch": 0.16575,
      "grad_norm": 28.625,
      "grad_norm_var": 4.380143229166666,
      "learning_rate": 0.0001,
      "loss": 7.56,
      "loss/crossentropy": 2.1314379185438157,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19869489334523677,
      "step": 6630
    },
    {
      "epoch": 0.166,
      "grad_norm": 32.75,
      "grad_norm_var": 65.24765625,
      "learning_rate": 0.0001,
      "loss": 7.5285,
      "loss/crossentropy": 2.1807857722043993,
      "loss/hidden": 3.44453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20406192895025016,
      "step": 6640
    },
    {
      "epoch": 0.16625,
      "grad_norm": 36.75,
      "grad_norm_var": 61.514322916666664,
      "learning_rate": 0.0001,
      "loss": 7.5774,
      "loss/crossentropy": 2.1761706352233885,
      "loss/hidden": 3.490625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22523548007011412,
      "step": 6650
    },
    {
      "epoch": 0.1665,
      "grad_norm": 40.0,
      "grad_norm_var": 10.858072916666666,
      "learning_rate": 0.0001,
      "loss": 7.4837,
      "loss/crossentropy": 2.0422482162714006,
      "loss/hidden": 3.411328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19435476139187813,
      "step": 6660
    },
    {
      "epoch": 0.16675,
      "grad_norm": 31.125,
      "grad_norm_var": 9.977083333333333,
      "learning_rate": 0.0001,
      "loss": 7.6686,
      "loss/crossentropy": 2.182368849217892,
      "loss/hidden": 3.3578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20404126346111298,
      "step": 6670
    },
    {
      "epoch": 0.167,
      "grad_norm": 32.0,
      "grad_norm_var": 11.715625,
      "learning_rate": 0.0001,
      "loss": 7.5378,
      "loss/crossentropy": 2.246191081404686,
      "loss/hidden": 3.28359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18508785720914603,
      "step": 6680
    },
    {
      "epoch": 0.16725,
      "grad_norm": 32.25,
      "grad_norm_var": 7.314322916666667,
      "learning_rate": 0.0001,
      "loss": 7.5093,
      "loss/crossentropy": 2.1791825108230114,
      "loss/hidden": 3.303515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18074298240244388,
      "step": 6690
    },
    {
      "epoch": 0.1675,
      "grad_norm": 30.75,
      "grad_norm_var": 8.992708333333333,
      "learning_rate": 0.0001,
      "loss": 7.5574,
      "loss/crossentropy": 2.1036971658468246,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1981421067379415,
      "step": 6700
    },
    {
      "epoch": 0.16775,
      "grad_norm": 29.5,
      "grad_norm_var": 14.267643229166667,
      "learning_rate": 0.0001,
      "loss": 7.3344,
      "loss/crossentropy": 2.001809497177601,
      "loss/hidden": 3.39609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1832934280857444,
      "step": 6710
    },
    {
      "epoch": 0.168,
      "grad_norm": 61.25,
      "grad_norm_var": 63.89375,
      "learning_rate": 0.0001,
      "loss": 7.5556,
      "loss/crossentropy": 2.1391955494880674,
      "loss/hidden": 3.49375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20318875089287758,
      "step": 6720
    },
    {
      "epoch": 0.16825,
      "grad_norm": 33.75,
      "grad_norm_var": 56.9791015625,
      "learning_rate": 0.0001,
      "loss": 7.556,
      "loss/crossentropy": 2.155173195898533,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19701551645994186,
      "step": 6730
    },
    {
      "epoch": 0.1685,
      "grad_norm": 30.5,
      "grad_norm_var": 9.07890625,
      "learning_rate": 0.0001,
      "loss": 7.444,
      "loss/crossentropy": 2.1643510669469834,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19851317517459394,
      "step": 6740
    },
    {
      "epoch": 0.16875,
      "grad_norm": 32.5,
      "grad_norm_var": 4.964518229166667,
      "learning_rate": 0.0001,
      "loss": 7.4711,
      "loss/crossentropy": 2.1824121534824372,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20471413098275662,
      "step": 6750
    },
    {
      "epoch": 0.169,
      "grad_norm": 32.75,
      "grad_norm_var": 6.1009765625,
      "learning_rate": 0.0001,
      "loss": 7.4304,
      "loss/crossentropy": 2.0600946068763735,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18951213210821152,
      "step": 6760
    },
    {
      "epoch": 0.16925,
      "grad_norm": 34.25,
      "grad_norm_var": 6.034309895833333,
      "learning_rate": 0.0001,
      "loss": 7.5587,
      "loss/crossentropy": 2.1303545042872427,
      "loss/hidden": 3.32578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19376144148409366,
      "step": 6770
    },
    {
      "epoch": 0.1695,
      "grad_norm": 35.75,
      "grad_norm_var": 8.519205729166666,
      "learning_rate": 0.0001,
      "loss": 7.4825,
      "loss/crossentropy": 2.223961615562439,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1993227696046233,
      "step": 6780
    },
    {
      "epoch": 0.16975,
      "grad_norm": 52.5,
      "grad_norm_var": 31.4837890625,
      "learning_rate": 0.0001,
      "loss": 7.4688,
      "loss/crossentropy": 2.1645863845944406,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19716022843495012,
      "step": 6790
    },
    {
      "epoch": 0.17,
      "grad_norm": 32.75,
      "grad_norm_var": 29.0400390625,
      "learning_rate": 0.0001,
      "loss": 7.5153,
      "loss/crossentropy": 2.251170714199543,
      "loss/hidden": 3.37578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21400584764778613,
      "step": 6800
    },
    {
      "epoch": 0.17025,
      "grad_norm": 32.5,
      "grad_norm_var": 4.945833333333334,
      "learning_rate": 0.0001,
      "loss": 7.4593,
      "loss/crossentropy": 2.1392437756061553,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19613203704357146,
      "step": 6810
    },
    {
      "epoch": 0.1705,
      "grad_norm": 31.5,
      "grad_norm_var": 9.330989583333333,
      "learning_rate": 0.0001,
      "loss": 7.4589,
      "loss/crossentropy": 2.1152432590723036,
      "loss/hidden": 3.330078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1848454039543867,
      "step": 6820
    },
    {
      "epoch": 0.17075,
      "grad_norm": 33.5,
      "grad_norm_var": 15.155208333333333,
      "learning_rate": 0.0001,
      "loss": 7.5984,
      "loss/crossentropy": 2.208776918053627,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20963791087269784,
      "step": 6830
    },
    {
      "epoch": 0.171,
      "grad_norm": 31.0,
      "grad_norm_var": 14.967708333333333,
      "learning_rate": 0.0001,
      "loss": 7.5348,
      "loss/crossentropy": 2.1368553161621096,
      "loss/hidden": 3.47734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2036329936236143,
      "step": 6840
    },
    {
      "epoch": 0.17125,
      "grad_norm": 33.25,
      "grad_norm_var": 11.615625,
      "learning_rate": 0.0001,
      "loss": 7.4543,
      "loss/crossentropy": 2.2001425683498383,
      "loss/hidden": 3.291015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18403711169958115,
      "step": 6850
    },
    {
      "epoch": 0.1715,
      "grad_norm": 31.875,
      "grad_norm_var": 4.045247395833333,
      "learning_rate": 0.0001,
      "loss": 7.4924,
      "loss/crossentropy": 2.0965635985136033,
      "loss/hidden": 3.413671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19117785301059484,
      "step": 6860
    },
    {
      "epoch": 0.17175,
      "grad_norm": 32.0,
      "grad_norm_var": 6.587434895833334,
      "learning_rate": 0.0001,
      "loss": 7.4879,
      "loss/crossentropy": 2.106214761734009,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19950247537344695,
      "step": 6870
    },
    {
      "epoch": 0.172,
      "grad_norm": 30.75,
      "grad_norm_var": 8.289518229166667,
      "learning_rate": 0.0001,
      "loss": 7.4561,
      "loss/crossentropy": 2.1776740878820418,
      "loss/hidden": 3.386328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19133987911045552,
      "step": 6880
    },
    {
      "epoch": 0.17225,
      "grad_norm": 30.25,
      "grad_norm_var": 2.77265625,
      "learning_rate": 0.0001,
      "loss": 7.4422,
      "loss/crossentropy": 2.082959216833115,
      "loss/hidden": 3.544921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2135216325521469,
      "step": 6890
    },
    {
      "epoch": 0.1725,
      "grad_norm": 31.0,
      "grad_norm_var": 4.1556640625,
      "learning_rate": 0.0001,
      "loss": 7.6053,
      "loss/crossentropy": 2.2018980890512467,
      "loss/hidden": 3.495703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21013570427894593,
      "step": 6900
    },
    {
      "epoch": 0.17275,
      "grad_norm": 39.5,
      "grad_norm_var": 6.5322265625,
      "learning_rate": 0.0001,
      "loss": 7.5447,
      "loss/crossentropy": 2.2833516895771027,
      "loss/hidden": 3.3125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19603817090392112,
      "step": 6910
    },
    {
      "epoch": 0.173,
      "grad_norm": 31.125,
      "grad_norm_var": 12.677018229166666,
      "learning_rate": 0.0001,
      "loss": 7.4817,
      "loss/crossentropy": 2.166572627425194,
      "loss/hidden": 3.38671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19017961733043193,
      "step": 6920
    },
    {
      "epoch": 0.17325,
      "grad_norm": 32.0,
      "grad_norm_var": 3.78125,
      "learning_rate": 0.0001,
      "loss": 7.6572,
      "loss/crossentropy": 2.1658833861351012,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22312654759734868,
      "step": 6930
    },
    {
      "epoch": 0.1735,
      "grad_norm": 31.5,
      "grad_norm_var": 3.5747395833333333,
      "learning_rate": 0.0001,
      "loss": 7.4706,
      "loss/crossentropy": 2.1533073887228964,
      "loss/hidden": 3.400390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20075356736779212,
      "step": 6940
    },
    {
      "epoch": 0.17375,
      "grad_norm": 33.0,
      "grad_norm_var": 7.611393229166667,
      "learning_rate": 0.0001,
      "loss": 7.6678,
      "loss/crossentropy": 2.2583969831466675,
      "loss/hidden": 3.406640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2209668504074216,
      "step": 6950
    },
    {
      "epoch": 0.174,
      "grad_norm": 30.0,
      "grad_norm_var": 11.0087890625,
      "learning_rate": 0.0001,
      "loss": 7.3805,
      "loss/crossentropy": 2.1416700780391693,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20209096949547528,
      "step": 6960
    },
    {
      "epoch": 0.17425,
      "grad_norm": 29.0,
      "grad_norm_var": 5.171875,
      "learning_rate": 0.0001,
      "loss": 7.5472,
      "loss/crossentropy": 2.1253055185079575,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19859984703361988,
      "step": 6970
    },
    {
      "epoch": 0.1745,
      "grad_norm": 30.0,
      "grad_norm_var": 6.6603515625,
      "learning_rate": 0.0001,
      "loss": 7.5648,
      "loss/crossentropy": 2.182023701816797,
      "loss/hidden": 3.29296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18740688636898994,
      "step": 6980
    },
    {
      "epoch": 0.17475,
      "grad_norm": 30.75,
      "grad_norm_var": 2.4947916666666665,
      "learning_rate": 0.0001,
      "loss": 7.4586,
      "loss/crossentropy": 2.1650559276342394,
      "loss/hidden": 3.394921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1944490736350417,
      "step": 6990
    },
    {
      "epoch": 0.175,
      "grad_norm": 32.25,
      "grad_norm_var": 5.6587890625,
      "learning_rate": 0.0001,
      "loss": 7.569,
      "loss/crossentropy": 2.177471086382866,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19053987860679628,
      "step": 7000
    },
    {
      "epoch": 0.17525,
      "grad_norm": 32.5,
      "grad_norm_var": 26.516080729166667,
      "learning_rate": 0.0001,
      "loss": 7.4753,
      "loss/crossentropy": 2.154606765508652,
      "loss/hidden": 3.423828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19910989850759506,
      "step": 7010
    },
    {
      "epoch": 0.1755,
      "grad_norm": 34.5,
      "grad_norm_var": 39.670572916666664,
      "learning_rate": 0.0001,
      "loss": 7.631,
      "loss/crossentropy": 2.197327023744583,
      "loss/hidden": 3.541796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21263533756136893,
      "step": 7020
    },
    {
      "epoch": 0.17575,
      "grad_norm": 32.25,
      "grad_norm_var": 19.970833333333335,
      "learning_rate": 0.0001,
      "loss": 7.5712,
      "loss/crossentropy": 2.143507385253906,
      "loss/hidden": 3.41796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20819550901651382,
      "step": 7030
    },
    {
      "epoch": 0.176,
      "grad_norm": 30.25,
      "grad_norm_var": 2.755989583333333,
      "learning_rate": 0.0001,
      "loss": 7.6156,
      "loss/crossentropy": 2.160289117693901,
      "loss/hidden": 3.45859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2082687295973301,
      "step": 7040
    },
    {
      "epoch": 0.17625,
      "grad_norm": 34.0,
      "grad_norm_var": 1.6035807291666666,
      "learning_rate": 0.0001,
      "loss": 7.5106,
      "loss/crossentropy": 2.096612122654915,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2005817864090204,
      "step": 7050
    },
    {
      "epoch": 0.1765,
      "grad_norm": 29.5,
      "grad_norm_var": 1.7921223958333334,
      "learning_rate": 0.0001,
      "loss": 7.4678,
      "loss/crossentropy": 2.150402194261551,
      "loss/hidden": 3.31953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18803946152329445,
      "step": 7060
    },
    {
      "epoch": 0.17675,
      "grad_norm": 30.375,
      "grad_norm_var": 1.7754557291666666,
      "learning_rate": 0.0001,
      "loss": 7.5601,
      "loss/crossentropy": 2.111455664038658,
      "loss/hidden": 3.38125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1900298461318016,
      "step": 7070
    },
    {
      "epoch": 0.177,
      "grad_norm": 30.875,
      "grad_norm_var": 3.0497395833333334,
      "learning_rate": 0.0001,
      "loss": 7.5888,
      "loss/crossentropy": 2.172677582502365,
      "loss/hidden": 3.380859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1932773429900408,
      "step": 7080
    },
    {
      "epoch": 0.17725,
      "grad_norm": 29.25,
      "grad_norm_var": 3.131705729166667,
      "learning_rate": 0.0001,
      "loss": 7.527,
      "loss/crossentropy": 2.2066776901483536,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20891736317425966,
      "step": 7090
    },
    {
      "epoch": 0.1775,
      "grad_norm": 33.0,
      "grad_norm_var": 5.386713240151209e+18,
      "learning_rate": 0.0001,
      "loss": 7.6693,
      "loss/crossentropy": 2.1923015132546424,
      "loss/hidden": 3.3640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2022488683462143,
      "step": 7100
    },
    {
      "epoch": 0.17775,
      "grad_norm": 31.375,
      "grad_norm_var": 2.9284656292120494e+18,
      "learning_rate": 0.0001,
      "loss": 7.557,
      "loss/crossentropy": 2.0847490578889847,
      "loss/hidden": 3.552734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2148862171918154,
      "step": 7110
    },
    {
      "epoch": 0.178,
      "grad_norm": 30.625,
      "grad_norm_var": 1.3369140625,
      "learning_rate": 0.0001,
      "loss": 7.4864,
      "loss/crossentropy": 2.1249582156538964,
      "loss/hidden": 3.296484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19168496485799552,
      "step": 7120
    },
    {
      "epoch": 0.17825,
      "grad_norm": 31.875,
      "grad_norm_var": 2.878125,
      "learning_rate": 0.0001,
      "loss": 7.36,
      "loss/crossentropy": 2.173662793636322,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1978321423754096,
      "step": 7130
    },
    {
      "epoch": 0.1785,
      "grad_norm": 34.5,
      "grad_norm_var": 3.9723307291666665,
      "learning_rate": 0.0001,
      "loss": 7.5663,
      "loss/crossentropy": 2.1832578271627425,
      "loss/hidden": 3.4484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22951248846948147,
      "step": 7140
    },
    {
      "epoch": 0.17875,
      "grad_norm": 32.75,
      "grad_norm_var": 2.022330729166667,
      "learning_rate": 0.0001,
      "loss": 7.5795,
      "loss/crossentropy": 2.1229765862226486,
      "loss/hidden": 3.28203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1857854513451457,
      "step": 7150
    },
    {
      "epoch": 0.179,
      "grad_norm": 30.25,
      "grad_norm_var": 42.5978515625,
      "learning_rate": 0.0001,
      "loss": 7.4524,
      "loss/crossentropy": 2.1213574737310408,
      "loss/hidden": 3.4671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22352912444621326,
      "step": 7160
    },
    {
      "epoch": 0.17925,
      "grad_norm": 29.625,
      "grad_norm_var": 44.90774739583333,
      "learning_rate": 0.0001,
      "loss": 7.5702,
      "loss/crossentropy": 2.2016272962093355,
      "loss/hidden": 3.4109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19035741928964853,
      "step": 7170
    },
    {
      "epoch": 0.1795,
      "grad_norm": 33.25,
      "grad_norm_var": 2.08515625,
      "learning_rate": 0.0001,
      "loss": 7.5485,
      "loss/crossentropy": 2.1106868594884873,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19449420645833015,
      "step": 7180
    },
    {
      "epoch": 0.17975,
      "grad_norm": 33.25,
      "grad_norm_var": 3.414322916666667,
      "learning_rate": 0.0001,
      "loss": 7.4434,
      "loss/crossentropy": 2.1368062049150467,
      "loss/hidden": 3.389453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1976731013506651,
      "step": 7190
    },
    {
      "epoch": 0.18,
      "grad_norm": 33.5,
      "grad_norm_var": 3.5483723958333333,
      "learning_rate": 0.0001,
      "loss": 7.4368,
      "loss/crossentropy": 2.1658909559249877,
      "loss/hidden": 3.476171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897917227819562,
      "step": 7200
    },
    {
      "epoch": 0.18025,
      "grad_norm": 32.5,
      "grad_norm_var": 3.39765625,
      "learning_rate": 0.0001,
      "loss": 7.5235,
      "loss/crossentropy": 2.0003643825650217,
      "loss/hidden": 3.39453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1881322082132101,
      "step": 7210
    },
    {
      "epoch": 0.1805,
      "grad_norm": 32.25,
      "grad_norm_var": 7.2369140625,
      "learning_rate": 0.0001,
      "loss": 7.5351,
      "loss/crossentropy": 2.1550097078084947,
      "loss/hidden": 3.417578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1885005658492446,
      "step": 7220
    },
    {
      "epoch": 0.18075,
      "grad_norm": 29.75,
      "grad_norm_var": 3.6809895833333335,
      "learning_rate": 0.0001,
      "loss": 7.3615,
      "loss/crossentropy": 2.089699313044548,
      "loss/hidden": 3.362109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18596011865884066,
      "step": 7230
    },
    {
      "epoch": 0.181,
      "grad_norm": 29.625,
      "grad_norm_var": 3.780989583333333,
      "learning_rate": 0.0001,
      "loss": 7.521,
      "loss/crossentropy": 2.1315871119499206,
      "loss/hidden": 3.476953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2100736267864704,
      "step": 7240
    },
    {
      "epoch": 0.18125,
      "grad_norm": 34.75,
      "grad_norm_var": 4.068684895833333,
      "learning_rate": 0.0001,
      "loss": 7.4489,
      "loss/crossentropy": 2.0509297475218773,
      "loss/hidden": 3.44296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21798528153449298,
      "step": 7250
    },
    {
      "epoch": 0.1815,
      "grad_norm": 33.0,
      "grad_norm_var": 2.546875,
      "learning_rate": 0.0001,
      "loss": 7.5147,
      "loss/crossentropy": 2.097806680202484,
      "loss/hidden": 3.419921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1954873636364937,
      "step": 7260
    },
    {
      "epoch": 0.18175,
      "grad_norm": 31.375,
      "grad_norm_var": 1.9270833333333333,
      "learning_rate": 0.0001,
      "loss": 7.5335,
      "loss/crossentropy": 2.1867170676589014,
      "loss/hidden": 3.471484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20327441468834878,
      "step": 7270
    },
    {
      "epoch": 0.182,
      "grad_norm": 31.75,
      "grad_norm_var": 2.0434895833333333,
      "learning_rate": 0.0001,
      "loss": 7.5098,
      "loss/crossentropy": 2.059905408322811,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1984263988211751,
      "step": 7280
    },
    {
      "epoch": 0.18225,
      "grad_norm": 30.875,
      "grad_norm_var": 3.0004557291666667,
      "learning_rate": 0.0001,
      "loss": 7.6258,
      "loss/crossentropy": 2.19571368098259,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20168307051062584,
      "step": 7290
    },
    {
      "epoch": 0.1825,
      "grad_norm": 30.25,
      "grad_norm_var": 2.7051432291666666,
      "learning_rate": 0.0001,
      "loss": 7.5923,
      "loss/crossentropy": 2.12765002399683,
      "loss/hidden": 3.518359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19315951522439717,
      "step": 7300
    },
    {
      "epoch": 0.18275,
      "grad_norm": 29.375,
      "grad_norm_var": 1.746875,
      "learning_rate": 0.0001,
      "loss": 7.5356,
      "loss/crossentropy": 2.2317890375852585,
      "loss/hidden": 3.369921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19801790416240692,
      "step": 7310
    },
    {
      "epoch": 0.183,
      "grad_norm": 30.5,
      "grad_norm_var": 3.206184895833333,
      "learning_rate": 0.0001,
      "loss": 7.4805,
      "loss/crossentropy": 2.115889000892639,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20432772096246482,
      "step": 7320
    },
    {
      "epoch": 0.18325,
      "grad_norm": 29.875,
      "grad_norm_var": 7.782291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5582,
      "loss/crossentropy": 2.20442833006382,
      "loss/hidden": 3.488671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20083660595119,
      "step": 7330
    },
    {
      "epoch": 0.1835,
      "grad_norm": 31.5,
      "grad_norm_var": 3.6947916666666667,
      "learning_rate": 0.0001,
      "loss": 7.6294,
      "loss/crossentropy": 2.210834649205208,
      "loss/hidden": 3.363671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19781918153166772,
      "step": 7340
    },
    {
      "epoch": 0.18375,
      "grad_norm": 31.25,
      "grad_norm_var": 3.7791015625,
      "learning_rate": 0.0001,
      "loss": 7.5733,
      "loss/crossentropy": 2.094112278521061,
      "loss/hidden": 3.41171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1902735486626625,
      "step": 7350
    },
    {
      "epoch": 0.184,
      "grad_norm": 31.25,
      "grad_norm_var": 1.3796223958333333,
      "learning_rate": 0.0001,
      "loss": 7.4645,
      "loss/crossentropy": 2.1219489932060243,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1875064203515649,
      "step": 7360
    },
    {
      "epoch": 0.18425,
      "grad_norm": 31.0,
      "grad_norm_var": 5.262955729166666,
      "learning_rate": 0.0001,
      "loss": 7.5375,
      "loss/crossentropy": 2.1768325984478,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20637014620006083,
      "step": 7370
    },
    {
      "epoch": 0.1845,
      "grad_norm": 33.25,
      "grad_norm_var": 4.51015625,
      "learning_rate": 0.0001,
      "loss": 7.5972,
      "loss/crossentropy": 2.088278591632843,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19868529513478278,
      "step": 7380
    },
    {
      "epoch": 0.18475,
      "grad_norm": 28.125,
      "grad_norm_var": 7.376041666666667,
      "learning_rate": 0.0001,
      "loss": 7.4234,
      "loss/crossentropy": 2.162215715646744,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20086740422993898,
      "step": 7390
    },
    {
      "epoch": 0.185,
      "grad_norm": 34.0,
      "grad_norm_var": 141.65201822916666,
      "learning_rate": 0.0001,
      "loss": 7.4958,
      "loss/crossentropy": 2.076205277442932,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1904180521145463,
      "step": 7400
    },
    {
      "epoch": 0.18525,
      "grad_norm": 31.125,
      "grad_norm_var": 138.825,
      "learning_rate": 0.0001,
      "loss": 7.574,
      "loss/crossentropy": 2.1355942092835902,
      "loss/hidden": 3.47734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2197154738008976,
      "step": 7410
    },
    {
      "epoch": 0.1855,
      "grad_norm": 30.75,
      "grad_norm_var": 3.405847189186238e+18,
      "learning_rate": 0.0001,
      "loss": 7.4496,
      "loss/crossentropy": 2.152878683805466,
      "loss/hidden": 3.36953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18649988640099763,
      "step": 7420
    },
    {
      "epoch": 0.18575,
      "grad_norm": 30.25,
      "grad_norm_var": 3.40584718949382e+18,
      "learning_rate": 0.0001,
      "loss": 7.4482,
      "loss/crossentropy": 2.100820633769035,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1907477645203471,
      "step": 7430
    },
    {
      "epoch": 0.186,
      "grad_norm": 30.75,
      "grad_norm_var": 2.299934895833333,
      "learning_rate": 0.0001,
      "loss": 7.436,
      "loss/crossentropy": 2.1752007991075515,
      "loss/hidden": 3.323828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19189766012132167,
      "step": 7440
    },
    {
      "epoch": 0.18625,
      "grad_norm": 30.25,
      "grad_norm_var": 5.689322916666667,
      "learning_rate": 0.0001,
      "loss": 7.5691,
      "loss/crossentropy": 2.242092598974705,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19479812681674957,
      "step": 7450
    },
    {
      "epoch": 0.1865,
      "grad_norm": 30.625,
      "grad_norm_var": 5.6369140625,
      "learning_rate": 0.0001,
      "loss": 7.6412,
      "loss/crossentropy": 2.1058298379182814,
      "loss/hidden": 3.4734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22841914147138595,
      "step": 7460
    },
    {
      "epoch": 0.18675,
      "grad_norm": 29.0,
      "grad_norm_var": 4.435872395833333,
      "learning_rate": 0.0001,
      "loss": 7.3797,
      "loss/crossentropy": 2.0960675440728664,
      "loss/hidden": 3.39375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1850858678109944,
      "step": 7470
    },
    {
      "epoch": 0.187,
      "grad_norm": 30.125,
      "grad_norm_var": 5.34765625,
      "learning_rate": 0.0001,
      "loss": 7.5432,
      "loss/crossentropy": 2.180506870150566,
      "loss/hidden": 3.459765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20674382094293833,
      "step": 7480
    },
    {
      "epoch": 0.18725,
      "grad_norm": 31.25,
      "grad_norm_var": 4.718489583333334,
      "learning_rate": 0.0001,
      "loss": 7.5242,
      "loss/crossentropy": 2.101006045937538,
      "loss/hidden": 3.503125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19774735551327466,
      "step": 7490
    },
    {
      "epoch": 0.1875,
      "grad_norm": 31.375,
      "grad_norm_var": 1.7791666666666666,
      "learning_rate": 0.0001,
      "loss": 7.5166,
      "loss/crossentropy": 2.069665388762951,
      "loss/hidden": 3.41484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19810831565409898,
      "step": 7500
    },
    {
      "epoch": 0.18775,
      "grad_norm": 29.75,
      "grad_norm_var": 2.07890625,
      "learning_rate": 0.0001,
      "loss": 7.4959,
      "loss/crossentropy": 2.0420530915260313,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18399292901158332,
      "step": 7510
    },
    {
      "epoch": 0.188,
      "grad_norm": 30.875,
      "grad_norm_var": 4.0587890625,
      "learning_rate": 0.0001,
      "loss": 7.5325,
      "loss/crossentropy": 2.1312338694930077,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19504649545997382,
      "step": 7520
    },
    {
      "epoch": 0.18825,
      "grad_norm": 43.0,
      "grad_norm_var": 9.308333333333334,
      "learning_rate": 0.0001,
      "loss": 7.6141,
      "loss/crossentropy": 2.2894584849476813,
      "loss/hidden": 3.409375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20183086302131414,
      "step": 7530
    },
    {
      "epoch": 0.1885,
      "grad_norm": 34.0,
      "grad_norm_var": 10.281705729166667,
      "learning_rate": 0.0001,
      "loss": 7.5206,
      "loss/crossentropy": 2.1557983949780466,
      "loss/hidden": 3.327734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19000910818576813,
      "step": 7540
    },
    {
      "epoch": 0.18875,
      "grad_norm": 29.25,
      "grad_norm_var": 4.21015625,
      "learning_rate": 0.0001,
      "loss": 7.5249,
      "loss/crossentropy": 2.124994584918022,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19348742607980968,
      "step": 7550
    },
    {
      "epoch": 0.189,
      "grad_norm": 29.75,
      "grad_norm_var": 4.334830729166667,
      "learning_rate": 0.0001,
      "loss": 7.4858,
      "loss/crossentropy": 2.154696786403656,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19012149404734374,
      "step": 7560
    },
    {
      "epoch": 0.18925,
      "grad_norm": 30.25,
      "grad_norm_var": 3.9872395833333334,
      "learning_rate": 0.0001,
      "loss": 7.487,
      "loss/crossentropy": 2.1938005700707435,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19971045572310686,
      "step": 7570
    },
    {
      "epoch": 0.1895,
      "grad_norm": 30.375,
      "grad_norm_var": 3.5624176423088947e+18,
      "learning_rate": 0.0001,
      "loss": 7.4698,
      "loss/crossentropy": 2.120267179608345,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19198095686733724,
      "step": 7580
    },
    {
      "epoch": 0.18975,
      "grad_norm": 30.75,
      "grad_norm_var": 4.756184895833333,
      "learning_rate": 0.0001,
      "loss": 7.4743,
      "loss/crossentropy": 2.1569736182689665,
      "loss/hidden": 3.53984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21590765863656997,
      "step": 7590
    },
    {
      "epoch": 0.19,
      "grad_norm": 30.125,
      "grad_norm_var": 2.9166015625,
      "learning_rate": 0.0001,
      "loss": 7.4344,
      "loss/crossentropy": 2.1558305487036704,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18447545487433673,
      "step": 7600
    },
    {
      "epoch": 0.19025,
      "grad_norm": 31.375,
      "grad_norm_var": 4.1025390625,
      "learning_rate": 0.0001,
      "loss": 7.4889,
      "loss/crossentropy": 2.0612099058926105,
      "loss/hidden": 3.427734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19961078523192555,
      "step": 7610
    },
    {
      "epoch": 0.1905,
      "grad_norm": 28.625,
      "grad_norm_var": 3.3087890625,
      "learning_rate": 0.0001,
      "loss": 7.6558,
      "loss/crossentropy": 2.04723744392395,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22622217051684856,
      "step": 7620
    },
    {
      "epoch": 0.19075,
      "grad_norm": 29.25,
      "grad_norm_var": 5.770833333333333,
      "learning_rate": 0.0001,
      "loss": 7.4133,
      "loss/crossentropy": 2.1897004157304765,
      "loss/hidden": 3.312109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1826148485764861,
      "step": 7630
    },
    {
      "epoch": 0.191,
      "grad_norm": 31.0,
      "grad_norm_var": 4.260416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4913,
      "loss/crossentropy": 2.157025161385536,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22510031294077634,
      "step": 7640
    },
    {
      "epoch": 0.19125,
      "grad_norm": 29.875,
      "grad_norm_var": 2.2270182291666667,
      "learning_rate": 0.0001,
      "loss": 7.517,
      "loss/crossentropy": 2.090579715371132,
      "loss/hidden": 3.3890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1889673013240099,
      "step": 7650
    },
    {
      "epoch": 0.1915,
      "grad_norm": 31.375,
      "grad_norm_var": 3.0770833333333334,
      "learning_rate": 0.0001,
      "loss": 7.6179,
      "loss/crossentropy": 2.1681595921516417,
      "loss/hidden": 3.555078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2365275712683797,
      "step": 7660
    },
    {
      "epoch": 0.19175,
      "grad_norm": 31.875,
      "grad_norm_var": 3.4280598958333335,
      "learning_rate": 0.0001,
      "loss": 7.5348,
      "loss/crossentropy": 2.207126745581627,
      "loss/hidden": 3.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19211382810026406,
      "step": 7670
    },
    {
      "epoch": 0.192,
      "grad_norm": 28.875,
      "grad_norm_var": 3.796875,
      "learning_rate": 0.0001,
      "loss": 7.553,
      "loss/crossentropy": 2.161148224771023,
      "loss/hidden": 3.52734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22304172217845916,
      "step": 7680
    },
    {
      "epoch": 0.19225,
      "grad_norm": 29.0,
      "grad_norm_var": 3.8889973958333335,
      "learning_rate": 0.0001,
      "loss": 7.5009,
      "loss/crossentropy": 2.0679882526397706,
      "loss/hidden": 3.46015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2086691778153181,
      "step": 7690
    },
    {
      "epoch": 0.1925,
      "grad_norm": 30.75,
      "grad_norm_var": 3.2552083333333335,
      "learning_rate": 0.0001,
      "loss": 7.5372,
      "loss/crossentropy": 2.0694913983345034,
      "loss/hidden": 3.37265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19288846384733915,
      "step": 7700
    },
    {
      "epoch": 0.19275,
      "grad_norm": 32.0,
      "grad_norm_var": 1.8072916666666667,
      "learning_rate": 0.0001,
      "loss": 7.5833,
      "loss/crossentropy": 2.2242089927196504,
      "loss/hidden": 3.306640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1934451697394252,
      "step": 7710
    },
    {
      "epoch": 0.193,
      "grad_norm": 34.5,
      "grad_norm_var": 12.787955729166667,
      "learning_rate": 0.0001,
      "loss": 7.6402,
      "loss/crossentropy": 2.157019394636154,
      "loss/hidden": 3.405078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19785581808537245,
      "step": 7720
    },
    {
      "epoch": 0.19325,
      "grad_norm": 32.0,
      "grad_norm_var": 12.268489583333333,
      "learning_rate": 0.0001,
      "loss": 7.3766,
      "loss/crossentropy": 2.226413035392761,
      "loss/hidden": 3.40625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20680981017649175,
      "step": 7730
    },
    {
      "epoch": 0.1935,
      "grad_norm": 35.5,
      "grad_norm_var": 6.34140625,
      "learning_rate": 0.0001,
      "loss": 7.5761,
      "loss/crossentropy": 2.0320019692182543,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19264598488807677,
      "step": 7740
    },
    {
      "epoch": 0.19375,
      "grad_norm": 34.5,
      "grad_norm_var": 5.897916666666666,
      "learning_rate": 0.0001,
      "loss": 7.453,
      "loss/crossentropy": 2.0598455399274824,
      "loss/hidden": 3.284375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1851517017930746,
      "step": 7750
    },
    {
      "epoch": 0.194,
      "grad_norm": 29.0,
      "grad_norm_var": 6.7025390625,
      "learning_rate": 0.0001,
      "loss": 7.4278,
      "loss/crossentropy": 2.1927490830421448,
      "loss/hidden": 3.414453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20509992372244595,
      "step": 7760
    },
    {
      "epoch": 0.19425,
      "grad_norm": 31.375,
      "grad_norm_var": 5.700455729166666,
      "learning_rate": 0.0001,
      "loss": 7.4874,
      "loss/crossentropy": 2.154856140911579,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20458675045520067,
      "step": 7770
    },
    {
      "epoch": 0.1945,
      "grad_norm": 37.5,
      "grad_norm_var": 6.71015625,
      "learning_rate": 0.0001,
      "loss": 7.5452,
      "loss/crossentropy": 2.1202862530946733,
      "loss/hidden": 3.3078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18985650166869164,
      "step": 7780
    },
    {
      "epoch": 0.19475,
      "grad_norm": 36.25,
      "grad_norm_var": 6.918684895833334,
      "learning_rate": 0.0001,
      "loss": 7.5427,
      "loss/crossentropy": 2.1549450784921644,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18924216069281102,
      "step": 7790
    },
    {
      "epoch": 0.195,
      "grad_norm": 35.25,
      "grad_norm_var": 6.2291015625,
      "learning_rate": 0.0001,
      "loss": 7.4733,
      "loss/crossentropy": 2.0557655200362204,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1919045101851225,
      "step": 7800
    },
    {
      "epoch": 0.19525,
      "grad_norm": 46.75,
      "grad_norm_var": 50.040625,
      "learning_rate": 0.0001,
      "loss": 7.4911,
      "loss/crossentropy": 2.171146012097597,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20546058733016254,
      "step": 7810
    },
    {
      "epoch": 0.1955,
      "grad_norm": 33.25,
      "grad_norm_var": 15.828580729166667,
      "learning_rate": 0.0001,
      "loss": 7.5692,
      "loss/crossentropy": 2.1003697514533997,
      "loss/hidden": 3.489453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20100659225136042,
      "step": 7820
    },
    {
      "epoch": 0.19575,
      "grad_norm": 36.0,
      "grad_norm_var": 29.226497395833334,
      "learning_rate": 0.0001,
      "loss": 7.5267,
      "loss/crossentropy": 2.1091859377920628,
      "loss/hidden": 3.35546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19987453976646066,
      "step": 7830
    },
    {
      "epoch": 0.196,
      "grad_norm": 48.5,
      "grad_norm_var": 42.45201822916667,
      "learning_rate": 0.0001,
      "loss": 7.4625,
      "loss/crossentropy": 2.111352452635765,
      "loss/hidden": 3.31171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18483758307993411,
      "step": 7840
    },
    {
      "epoch": 0.19625,
      "grad_norm": 30.5,
      "grad_norm_var": 25.65390625,
      "learning_rate": 0.0001,
      "loss": 7.5415,
      "loss/crossentropy": 2.0419967979192735,
      "loss/hidden": 3.464453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2173704007640481,
      "step": 7850
    },
    {
      "epoch": 0.1965,
      "grad_norm": 30.875,
      "grad_norm_var": 21.546875,
      "learning_rate": 0.0001,
      "loss": 7.4951,
      "loss/crossentropy": 2.084785957634449,
      "loss/hidden": 3.304296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17884304635226728,
      "step": 7860
    },
    {
      "epoch": 0.19675,
      "grad_norm": 31.75,
      "grad_norm_var": 18.153059895833334,
      "learning_rate": 0.0001,
      "loss": 7.4509,
      "loss/crossentropy": 2.1023138776421546,
      "loss/hidden": 3.479296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20335167981684207,
      "step": 7870
    },
    {
      "epoch": 0.197,
      "grad_norm": 36.25,
      "grad_norm_var": 12.093489583333334,
      "learning_rate": 0.0001,
      "loss": 7.5467,
      "loss/crossentropy": 2.171536546945572,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21221924368292094,
      "step": 7880
    },
    {
      "epoch": 0.19725,
      "grad_norm": 28.75,
      "grad_norm_var": 9.983333333333333,
      "learning_rate": 0.0001,
      "loss": 7.3785,
      "loss/crossentropy": 2.1119479715824125,
      "loss/hidden": 3.39765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18313394635915756,
      "step": 7890
    },
    {
      "epoch": 0.1975,
      "grad_norm": 37.25,
      "grad_norm_var": 11.928580729166667,
      "learning_rate": 0.0001,
      "loss": 7.4945,
      "loss/crossentropy": 1.9945223838090897,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19973532520234585,
      "step": 7900
    },
    {
      "epoch": 0.19775,
      "grad_norm": 32.25,
      "grad_norm_var": 11.839583333333334,
      "learning_rate": 0.0001,
      "loss": 7.4542,
      "loss/crossentropy": 2.1550634860992433,
      "loss/hidden": 3.448046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19187356438487768,
      "step": 7910
    },
    {
      "epoch": 0.198,
      "grad_norm": 31.5,
      "grad_norm_var": 12.584309895833334,
      "learning_rate": 0.0001,
      "loss": 7.4245,
      "loss/crossentropy": 2.092045524716377,
      "loss/hidden": 3.315625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18641792330890894,
      "step": 7920
    },
    {
      "epoch": 0.19825,
      "grad_norm": 28.875,
      "grad_norm_var": 12.959375,
      "learning_rate": 0.0001,
      "loss": 7.4616,
      "loss/crossentropy": 2.1302111998200415,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2105752557516098,
      "step": 7930
    },
    {
      "epoch": 0.1985,
      "grad_norm": 30.75,
      "grad_norm_var": 14.9431640625,
      "learning_rate": 0.0001,
      "loss": 7.4869,
      "loss/crossentropy": 2.1739411771297457,
      "loss/hidden": 3.455078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19759192429482936,
      "step": 7940
    },
    {
      "epoch": 0.19875,
      "grad_norm": 33.25,
      "grad_norm_var": 12.703580729166667,
      "learning_rate": 0.0001,
      "loss": 7.4435,
      "loss/crossentropy": 2.0400329381227493,
      "loss/hidden": 3.3203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18711388055235148,
      "step": 7950
    },
    {
      "epoch": 0.199,
      "grad_norm": 40.0,
      "grad_norm_var": 10.454622395833333,
      "learning_rate": 0.0001,
      "loss": 7.3235,
      "loss/crossentropy": 2.1173346668481825,
      "loss/hidden": 3.383203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.200854654237628,
      "step": 7960
    },
    {
      "epoch": 0.19925,
      "grad_norm": 31.125,
      "grad_norm_var": 18.376822916666665,
      "learning_rate": 0.0001,
      "loss": 7.551,
      "loss/crossentropy": 2.256546127796173,
      "loss/hidden": 3.430859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2098003163933754,
      "step": 7970
    },
    {
      "epoch": 0.1995,
      "grad_norm": 30.625,
      "grad_norm_var": 11.159375,
      "learning_rate": 0.0001,
      "loss": 7.5863,
      "loss/crossentropy": 2.0988403081893923,
      "loss/hidden": 3.372265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2052627831697464,
      "step": 7980
    },
    {
      "epoch": 0.19975,
      "grad_norm": 34.75,
      "grad_norm_var": 3.539322916666667,
      "learning_rate": 0.0001,
      "loss": 7.5501,
      "loss/crossentropy": 2.157486143708229,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19415573356673121,
      "step": 7990
    },
    {
      "epoch": 0.2,
      "grad_norm": 28.75,
      "grad_norm_var": 5.282291666666667,
      "learning_rate": 0.0001,
      "loss": 7.5487,
      "loss/crossentropy": 2.204603946208954,
      "loss/hidden": 3.328515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19015649259090422,
      "step": 8000
    },
    {
      "epoch": 0.20025,
      "grad_norm": 29.5,
      "grad_norm_var": 3.59375,
      "learning_rate": 0.0001,
      "loss": 7.4113,
      "loss/crossentropy": 2.1252332285046576,
      "loss/hidden": 3.287109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1751571564003825,
      "step": 8010
    },
    {
      "epoch": 0.2005,
      "grad_norm": 31.0,
      "grad_norm_var": 3.603125,
      "learning_rate": 0.0001,
      "loss": 7.498,
      "loss/crossentropy": 2.048645743727684,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1930268170312047,
      "step": 8020
    },
    {
      "epoch": 0.20075,
      "grad_norm": 31.5,
      "grad_norm_var": 3.3822265625,
      "learning_rate": 0.0001,
      "loss": 7.4496,
      "loss/crossentropy": 2.1435810789465903,
      "loss/hidden": 3.340625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19256617035716772,
      "step": 8030
    },
    {
      "epoch": 0.201,
      "grad_norm": 31.875,
      "grad_norm_var": 5.8837890625,
      "learning_rate": 0.0001,
      "loss": 7.5991,
      "loss/crossentropy": 2.06428082883358,
      "loss/hidden": 3.397265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18795610759407283,
      "step": 8040
    },
    {
      "epoch": 0.20125,
      "grad_norm": 31.75,
      "grad_norm_var": 4.08515625,
      "learning_rate": 0.0001,
      "loss": 7.5106,
      "loss/crossentropy": 2.2008739590644835,
      "loss/hidden": 3.329296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1890289282426238,
      "step": 8050
    },
    {
      "epoch": 0.2015,
      "grad_norm": 33.25,
      "grad_norm_var": 2.7708333333333335,
      "learning_rate": 0.0001,
      "loss": 7.4302,
      "loss/crossentropy": 2.1280903786420824,
      "loss/hidden": 3.511328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20874056480824948,
      "step": 8060
    },
    {
      "epoch": 0.20175,
      "grad_norm": 29.5,
      "grad_norm_var": 3.2035807291666667,
      "learning_rate": 0.0001,
      "loss": 7.4529,
      "loss/crossentropy": 2.1864554077386855,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19697167426347734,
      "step": 8070
    },
    {
      "epoch": 0.202,
      "grad_norm": 35.0,
      "grad_norm_var": 2.278125,
      "learning_rate": 0.0001,
      "loss": 7.5744,
      "loss/crossentropy": 2.211556833982468,
      "loss/hidden": 3.433203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20347684845328332,
      "step": 8080
    },
    {
      "epoch": 0.20225,
      "grad_norm": 28.875,
      "grad_norm_var": 33.37890625,
      "learning_rate": 0.0001,
      "loss": 7.5018,
      "loss/crossentropy": 2.185184845328331,
      "loss/hidden": 3.358203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2022322118282318,
      "step": 8090
    },
    {
      "epoch": 0.2025,
      "grad_norm": 31.875,
      "grad_norm_var": 3.3931640625,
      "learning_rate": 0.0001,
      "loss": 7.5218,
      "loss/crossentropy": 2.1828758120536804,
      "loss/hidden": 3.337890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1962041450664401,
      "step": 8100
    },
    {
      "epoch": 0.20275,
      "grad_norm": 29.875,
      "grad_norm_var": 1.9488932291666667,
      "learning_rate": 0.0001,
      "loss": 7.4982,
      "loss/crossentropy": 2.159992370009422,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20945403575897217,
      "step": 8110
    },
    {
      "epoch": 0.203,
      "grad_norm": 31.25,
      "grad_norm_var": 4.748958333333333,
      "learning_rate": 0.0001,
      "loss": 7.4657,
      "loss/crossentropy": 2.1400886207818983,
      "loss/hidden": 3.352734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19184730034321545,
      "step": 8120
    },
    {
      "epoch": 0.20325,
      "grad_norm": 32.0,
      "grad_norm_var": 1.0733723958333334,
      "learning_rate": 0.0001,
      "loss": 7.5559,
      "loss/crossentropy": 2.112087991833687,
      "loss/hidden": 3.475390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20091654695570468,
      "step": 8130
    },
    {
      "epoch": 0.2035,
      "grad_norm": 30.625,
      "grad_norm_var": 1.921875,
      "learning_rate": 0.0001,
      "loss": 7.5644,
      "loss/crossentropy": 2.139628532528877,
      "loss/hidden": 3.52265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19385650400072335,
      "step": 8140
    },
    {
      "epoch": 0.20375,
      "grad_norm": 33.25,
      "grad_norm_var": 2.379166666666667,
      "learning_rate": 0.0001,
      "loss": 7.5002,
      "loss/crossentropy": 2.091212958097458,
      "loss/hidden": 3.446484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2040353151038289,
      "step": 8150
    },
    {
      "epoch": 0.204,
      "grad_norm": 32.0,
      "grad_norm_var": 2.2228515625,
      "learning_rate": 0.0001,
      "loss": 7.4494,
      "loss/crossentropy": 2.2567614316940308,
      "loss/hidden": 3.35078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18545747231692075,
      "step": 8160
    },
    {
      "epoch": 0.20425,
      "grad_norm": 31.125,
      "grad_norm_var": 229.56666666666666,
      "learning_rate": 0.0001,
      "loss": 7.4626,
      "loss/crossentropy": 2.1603645354509355,
      "loss/hidden": 3.539453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21300703901797532,
      "step": 8170
    },
    {
      "epoch": 0.2045,
      "grad_norm": 28.625,
      "grad_norm_var": 4.2806640625,
      "learning_rate": 0.0001,
      "loss": 7.5117,
      "loss/crossentropy": 2.1204755783081053,
      "loss/hidden": 3.456640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18535650707781315,
      "step": 8180
    },
    {
      "epoch": 0.20475,
      "grad_norm": 31.25,
      "grad_norm_var": 2.1072265625,
      "learning_rate": 0.0001,
      "loss": 7.5637,
      "loss/crossentropy": 2.1639575958251953,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20935236364603044,
      "step": 8190
    },
    {
      "epoch": 0.205,
      "grad_norm": 28.0,
      "grad_norm_var": 2.2494140625,
      "learning_rate": 0.0001,
      "loss": 7.3019,
      "loss/crossentropy": 2.056389382481575,
      "loss/hidden": 3.3203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1874819153919816,
      "step": 8200
    },
    {
      "epoch": 0.20525,
      "grad_norm": 32.5,
      "grad_norm_var": 4.274934895833334,
      "learning_rate": 0.0001,
      "loss": 7.5162,
      "loss/crossentropy": 2.1393348813056945,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20481376871466636,
      "step": 8210
    },
    {
      "epoch": 0.2055,
      "grad_norm": 29.625,
      "grad_norm_var": 5.553125,
      "learning_rate": 0.0001,
      "loss": 7.5046,
      "loss/crossentropy": 2.1898138865828516,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1965622954070568,
      "step": 8220
    },
    {
      "epoch": 0.20575,
      "grad_norm": 34.5,
      "grad_norm_var": 2.899825551450536e+18,
      "learning_rate": 0.0001,
      "loss": 7.6057,
      "loss/crossentropy": 2.139435464143753,
      "loss/hidden": 3.51328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20182280670851468,
      "step": 8230
    },
    {
      "epoch": 0.206,
      "grad_norm": 37.75,
      "grad_norm_var": 2.899825549761839e+18,
      "learning_rate": 0.0001,
      "loss": 7.5603,
      "loss/crossentropy": 2.248063361644745,
      "loss/hidden": 3.44609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21399107333272696,
      "step": 8240
    },
    {
      "epoch": 0.20625,
      "grad_norm": 33.5,
      "grad_norm_var": 114.23118489583334,
      "learning_rate": 0.0001,
      "loss": 7.4545,
      "loss/crossentropy": 2.140778873860836,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18022785130888225,
      "step": 8250
    },
    {
      "epoch": 0.2065,
      "grad_norm": 30.5,
      "grad_norm_var": 110.5072265625,
      "learning_rate": 0.0001,
      "loss": 7.5364,
      "loss/crossentropy": 2.259766247868538,
      "loss/hidden": 3.312109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19342643208801746,
      "step": 8260
    },
    {
      "epoch": 0.20675,
      "grad_norm": 31.75,
      "grad_norm_var": 0.7666015625,
      "learning_rate": 0.0001,
      "loss": 7.6531,
      "loss/crossentropy": 2.158949288725853,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19808744061738254,
      "step": 8270
    },
    {
      "epoch": 0.207,
      "grad_norm": 32.0,
      "grad_norm_var": 1.4311848958333333,
      "learning_rate": 0.0001,
      "loss": 7.6092,
      "loss/crossentropy": 2.111227548122406,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20227425117045642,
      "step": 8280
    },
    {
      "epoch": 0.20725,
      "grad_norm": 31.5,
      "grad_norm_var": 1.56640625,
      "learning_rate": 0.0001,
      "loss": 7.5372,
      "loss/crossentropy": 2.0839753821492195,
      "loss/hidden": 3.428125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18574036825448276,
      "step": 8290
    },
    {
      "epoch": 0.2075,
      "grad_norm": 31.375,
      "grad_norm_var": 0.8384765625,
      "learning_rate": 0.0001,
      "loss": 7.536,
      "loss/crossentropy": 2.1465844094753264,
      "loss/hidden": 3.407421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2083981443196535,
      "step": 8300
    },
    {
      "epoch": 0.20775,
      "grad_norm": 32.5,
      "grad_norm_var": 20.8416015625,
      "learning_rate": 0.0001,
      "loss": 7.4722,
      "loss/crossentropy": 2.1032681286334993,
      "loss/hidden": 3.439453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19353157542645932,
      "step": 8310
    },
    {
      "epoch": 0.208,
      "grad_norm": 29.75,
      "grad_norm_var": 2.1395833333333334,
      "learning_rate": 0.0001,
      "loss": 7.4453,
      "loss/crossentropy": 2.092882976680994,
      "loss/hidden": 3.41640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20083120074123145,
      "step": 8320
    },
    {
      "epoch": 0.20825,
      "grad_norm": 29.125,
      "grad_norm_var": 2.42265625,
      "learning_rate": 0.0001,
      "loss": 7.4658,
      "loss/crossentropy": 2.047010327875614,
      "loss/hidden": 3.26484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17298451280221344,
      "step": 8330
    },
    {
      "epoch": 0.2085,
      "grad_norm": 31.625,
      "grad_norm_var": 2.8114583333333334,
      "learning_rate": 0.0001,
      "loss": 7.6227,
      "loss/crossentropy": 2.1705314934253694,
      "loss/hidden": 3.44921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2104344118386507,
      "step": 8340
    },
    {
      "epoch": 0.20875,
      "grad_norm": 30.75,
      "grad_norm_var": 15.13515625,
      "learning_rate": 0.0001,
      "loss": 7.5292,
      "loss/crossentropy": 2.251467025279999,
      "loss/hidden": 3.44296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.206053901091218,
      "step": 8350
    },
    {
      "epoch": 0.209,
      "grad_norm": 30.25,
      "grad_norm_var": 9.75,
      "learning_rate": 0.0001,
      "loss": 7.5297,
      "loss/crossentropy": 2.2416324824094773,
      "loss/hidden": 3.38203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18910795077681541,
      "step": 8360
    },
    {
      "epoch": 0.20925,
      "grad_norm": 49.75,
      "grad_norm_var": 177.46145833333333,
      "learning_rate": 0.0001,
      "loss": 7.4604,
      "loss/crossentropy": 1.9895980581641197,
      "loss/hidden": 3.43203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19730695113539695,
      "step": 8370
    },
    {
      "epoch": 0.2095,
      "grad_norm": 56.25,
      "grad_norm_var": 334.9145182291667,
      "learning_rate": 0.0001,
      "loss": 7.5698,
      "loss/crossentropy": 2.2018245279788973,
      "loss/hidden": 3.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21164307594299317,
      "step": 8380
    },
    {
      "epoch": 0.20975,
      "grad_norm": 32.75,
      "grad_norm_var": 72.22890625,
      "learning_rate": 0.0001,
      "loss": 7.5902,
      "loss/crossentropy": 2.1519288808107375,
      "loss/hidden": 3.3796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19217228647321463,
      "step": 8390
    },
    {
      "epoch": 0.21,
      "grad_norm": 30.0,
      "grad_norm_var": 7.153059895833334,
      "learning_rate": 0.0001,
      "loss": 7.451,
      "loss/crossentropy": 2.1779698938131333,
      "loss/hidden": 3.44140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1994494464248419,
      "step": 8400
    },
    {
      "epoch": 0.21025,
      "grad_norm": 32.0,
      "grad_norm_var": 7.124739583333334,
      "learning_rate": 0.0001,
      "loss": 7.5019,
      "loss/crossentropy": 2.140475983917713,
      "loss/hidden": 3.3734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20144703481346368,
      "step": 8410
    },
    {
      "epoch": 0.2105,
      "grad_norm": 31.625,
      "grad_norm_var": 1.2639973958333333,
      "learning_rate": 0.0001,
      "loss": 7.4268,
      "loss/crossentropy": 2.2145755022764204,
      "loss/hidden": 3.328515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19062853194773197,
      "step": 8420
    },
    {
      "epoch": 0.21075,
      "grad_norm": 31.0,
      "grad_norm_var": 2.620247395833333,
      "learning_rate": 0.0001,
      "loss": 7.5149,
      "loss/crossentropy": 2.3047141253948213,
      "loss/hidden": 3.34140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20336541533470154,
      "step": 8430
    },
    {
      "epoch": 0.211,
      "grad_norm": 31.875,
      "grad_norm_var": 2.7514973958333333,
      "learning_rate": 0.0001,
      "loss": 7.5319,
      "loss/crossentropy": 2.232821634411812,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21666326355189086,
      "step": 8440
    },
    {
      "epoch": 0.21125,
      "grad_norm": 33.5,
      "grad_norm_var": 1.49765625,
      "learning_rate": 0.0001,
      "loss": 7.5718,
      "loss/crossentropy": 2.092620450258255,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20793931670486926,
      "step": 8450
    },
    {
      "epoch": 0.2115,
      "grad_norm": 32.5,
      "grad_norm_var": 3.1306640625,
      "learning_rate": 0.0001,
      "loss": 7.4344,
      "loss/crossentropy": 2.151240213960409,
      "loss/hidden": 3.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2046731175854802,
      "step": 8460
    },
    {
      "epoch": 0.21175,
      "grad_norm": 33.75,
      "grad_norm_var": 2.81640625,
      "learning_rate": 0.0001,
      "loss": 7.5581,
      "loss/crossentropy": 2.1520677715539933,
      "loss/hidden": 3.379296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19603268206119537,
      "step": 8470
    },
    {
      "epoch": 0.212,
      "grad_norm": 31.0,
      "grad_norm_var": 4.092643229166667,
      "learning_rate": 0.0001,
      "loss": 7.5382,
      "loss/crossentropy": 2.1510326713323593,
      "loss/hidden": 3.33359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18077819589525462,
      "step": 8480
    },
    {
      "epoch": 0.21225,
      "grad_norm": 32.75,
      "grad_norm_var": 4.753059895833333,
      "learning_rate": 0.0001,
      "loss": 7.4642,
      "loss/crossentropy": 2.0811485938727854,
      "loss/hidden": 3.351171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18939817287027835,
      "step": 8490
    },
    {
      "epoch": 0.2125,
      "grad_norm": 31.25,
      "grad_norm_var": 5.855989583333334,
      "learning_rate": 0.0001,
      "loss": 7.5158,
      "loss/crossentropy": 2.0734725266695024,
      "loss/hidden": 3.316796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18070020200684667,
      "step": 8500
    },
    {
      "epoch": 0.21275,
      "grad_norm": 29.25,
      "grad_norm_var": 5.5462890625,
      "learning_rate": 0.0001,
      "loss": 7.4723,
      "loss/crossentropy": 2.144814722239971,
      "loss/hidden": 3.3828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19649898763746024,
      "step": 8510
    },
    {
      "epoch": 0.213,
      "grad_norm": 28.625,
      "grad_norm_var": 3.2309895833333333,
      "learning_rate": 0.0001,
      "loss": 7.3777,
      "loss/crossentropy": 2.0877407863736153,
      "loss/hidden": 3.42734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19544295389205218,
      "step": 8520
    },
    {
      "epoch": 0.21325,
      "grad_norm": 30.5,
      "grad_norm_var": 1.6583333333333334,
      "learning_rate": 0.0001,
      "loss": 7.5179,
      "loss/crossentropy": 2.1433625385165214,
      "loss/hidden": 3.4734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19359017722308636,
      "step": 8530
    },
    {
      "epoch": 0.2135,
      "grad_norm": 34.25,
      "grad_norm_var": 1.6510416666666667,
      "learning_rate": 0.0001,
      "loss": 7.4658,
      "loss/crossentropy": 2.221453693509102,
      "loss/hidden": 3.43125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20547619834542274,
      "step": 8540
    },
    {
      "epoch": 0.21375,
      "grad_norm": 32.0,
      "grad_norm_var": 2.5791015625,
      "learning_rate": 0.0001,
      "loss": 7.5163,
      "loss/crossentropy": 2.213938871026039,
      "loss/hidden": 3.328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20213278364390136,
      "step": 8550
    },
    {
      "epoch": 0.214,
      "grad_norm": 34.25,
      "grad_norm_var": 2.784375,
      "learning_rate": 0.0001,
      "loss": 7.6361,
      "loss/crossentropy": 2.133354830741882,
      "loss/hidden": 3.512109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.207733928412199,
      "step": 8560
    },
    {
      "epoch": 0.21425,
      "grad_norm": 34.0,
      "grad_norm_var": 7.219205729166666,
      "learning_rate": 0.0001,
      "loss": 7.5661,
      "loss/crossentropy": 2.0587344974279405,
      "loss/hidden": 3.550390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22256299555301667,
      "step": 8570
    },
    {
      "epoch": 0.2145,
      "grad_norm": 31.0,
      "grad_norm_var": 7.805989583333333,
      "learning_rate": 0.0001,
      "loss": 7.5549,
      "loss/crossentropy": 2.0949306935071945,
      "loss/hidden": 3.4796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19931861553341151,
      "step": 8580
    },
    {
      "epoch": 0.21475,
      "grad_norm": 28.5,
      "grad_norm_var": 8.095247395833333,
      "learning_rate": 0.0001,
      "loss": 7.478,
      "loss/crossentropy": 2.1063201159238814,
      "loss/hidden": 3.459375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20397470220923425,
      "step": 8590
    },
    {
      "epoch": 0.215,
      "grad_norm": 33.5,
      "grad_norm_var": 28.859375,
      "learning_rate": 0.0001,
      "loss": 7.5632,
      "loss/crossentropy": 2.177382430434227,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2048982411623001,
      "step": 8600
    },
    {
      "epoch": 0.21525,
      "grad_norm": 31.75,
      "grad_norm_var": 1.7955729166666667,
      "learning_rate": 0.0001,
      "loss": 7.5325,
      "loss/crossentropy": 2.0719266816973687,
      "loss/hidden": 3.395703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19100363925099373,
      "step": 8610
    },
    {
      "epoch": 0.2155,
      "grad_norm": 29.25,
      "grad_norm_var": 1.7372395833333334,
      "learning_rate": 0.0001,
      "loss": 7.5877,
      "loss/crossentropy": 2.147695118188858,
      "loss/hidden": 3.41875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19447963647544383,
      "step": 8620
    },
    {
      "epoch": 0.21575,
      "grad_norm": 31.375,
      "grad_norm_var": 5.406184895833333,
      "learning_rate": 0.0001,
      "loss": 7.5509,
      "loss/crossentropy": 2.1029836043715475,
      "loss/hidden": 3.45,
      "loss/jsd": 0.0,
      "loss/logits": 0.19613635465502738,
      "step": 8630
    },
    {
      "epoch": 0.216,
      "grad_norm": 29.0,
      "grad_norm_var": 10.32265625,
      "learning_rate": 0.0001,
      "loss": 7.555,
      "loss/crossentropy": 2.1027876287698746,
      "loss/hidden": 3.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18709390722215175,
      "step": 8640
    },
    {
      "epoch": 0.21625,
      "grad_norm": 31.75,
      "grad_norm_var": 7.4978515625,
      "learning_rate": 0.0001,
      "loss": 7.481,
      "loss/crossentropy": 2.078503981232643,
      "loss/hidden": 3.3578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19765212927013637,
      "step": 8650
    },
    {
      "epoch": 0.2165,
      "grad_norm": 28.375,
      "grad_norm_var": 6.212434895833334,
      "learning_rate": 0.0001,
      "loss": 7.5951,
      "loss/crossentropy": 2.1425766557455064,
      "loss/hidden": 3.522265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2285369848832488,
      "step": 8660
    },
    {
      "epoch": 0.21675,
      "grad_norm": 37.0,
      "grad_norm_var": 5.02890625,
      "learning_rate": 0.0001,
      "loss": 7.4619,
      "loss/crossentropy": 2.094095268845558,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18905023839324714,
      "step": 8670
    },
    {
      "epoch": 0.217,
      "grad_norm": 32.25,
      "grad_norm_var": 5.205989583333333,
      "learning_rate": 0.0001,
      "loss": 7.6255,
      "loss/crossentropy": 2.0726657152175902,
      "loss/hidden": 3.424609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19401923939585686,
      "step": 8680
    },
    {
      "epoch": 0.21725,
      "grad_norm": 31.125,
      "grad_norm_var": 1.8240281960963725e+18,
      "learning_rate": 0.0001,
      "loss": 7.476,
      "loss/crossentropy": 2.1063093028962614,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18630194589495658,
      "step": 8690
    },
    {
      "epoch": 0.2175,
      "grad_norm": 33.0,
      "grad_norm_var": 9.2947265625,
      "learning_rate": 0.0001,
      "loss": 7.6565,
      "loss/crossentropy": 2.188694643974304,
      "loss/hidden": 3.561328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21225023418664932,
      "step": 8700
    },
    {
      "epoch": 0.21775,
      "grad_norm": 30.0,
      "grad_norm_var": 3.644791666666667,
      "learning_rate": 0.0001,
      "loss": 7.4697,
      "loss/crossentropy": 2.1252513483166693,
      "loss/hidden": 3.334765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19880297109484674,
      "step": 8710
    },
    {
      "epoch": 0.218,
      "grad_norm": 30.875,
      "grad_norm_var": 11.028059895833334,
      "learning_rate": 0.0001,
      "loss": 7.4975,
      "loss/crossentropy": 2.1700349181890486,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20139609836041927,
      "step": 8720
    },
    {
      "epoch": 0.21825,
      "grad_norm": 32.25,
      "grad_norm_var": 1.4189153379885778e+18,
      "learning_rate": 0.0001,
      "loss": 7.616,
      "loss/crossentropy": 2.032346047461033,
      "loss/hidden": 3.425,
      "loss/jsd": 0.0,
      "loss/logits": 0.1754102895502001,
      "step": 8730
    },
    {
      "epoch": 0.2185,
      "grad_norm": 31.25,
      "grad_norm_var": 2.25390625,
      "learning_rate": 0.0001,
      "loss": 7.5088,
      "loss/crossentropy": 2.1633499920368195,
      "loss/hidden": 3.448828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1996068175882101,
      "step": 8740
    },
    {
      "epoch": 0.21875,
      "grad_norm": 29.375,
      "grad_norm_var": 1.6122395833333334,
      "learning_rate": 0.0001,
      "loss": 7.5088,
      "loss/crossentropy": 2.1443722933530807,
      "loss/hidden": 3.501171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2017217763699591,
      "step": 8750
    },
    {
      "epoch": 0.219,
      "grad_norm": 32.25,
      "grad_norm_var": 14.970572916666667,
      "learning_rate": 0.0001,
      "loss": 7.5224,
      "loss/crossentropy": 2.176077055931091,
      "loss/hidden": 3.4203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2048908168449998,
      "step": 8760
    },
    {
      "epoch": 0.21925,
      "grad_norm": 31.0,
      "grad_norm_var": 14.087955729166667,
      "learning_rate": 0.0001,
      "loss": 7.5417,
      "loss/crossentropy": 2.183764386177063,
      "loss/hidden": 3.444140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20730934757739305,
      "step": 8770
    },
    {
      "epoch": 0.2195,
      "grad_norm": 46.5,
      "grad_norm_var": 23.55390625,
      "learning_rate": 0.0001,
      "loss": 7.5464,
      "loss/crossentropy": 2.0760655224323274,
      "loss/hidden": 3.3859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18335200790315867,
      "step": 8780
    },
    {
      "epoch": 0.21975,
      "grad_norm": 31.125,
      "grad_norm_var": 16.1369140625,
      "learning_rate": 0.0001,
      "loss": 7.5001,
      "loss/crossentropy": 2.215297505259514,
      "loss/hidden": 3.33671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19053485784679652,
      "step": 8790
    },
    {
      "epoch": 0.22,
      "grad_norm": 41.75,
      "grad_norm_var": 10.250455729166667,
      "learning_rate": 0.0001,
      "loss": 7.5664,
      "loss/crossentropy": 2.1419076189398765,
      "loss/hidden": 3.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1872857511974871,
      "step": 8800
    },
    {
      "epoch": 0.22025,
      "grad_norm": 31.625,
      "grad_norm_var": 13.099934895833334,
      "learning_rate": 0.0001,
      "loss": 7.5152,
      "loss/crossentropy": 2.136802351474762,
      "loss/hidden": 3.368359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19524938464164734,
      "step": 8810
    },
    {
      "epoch": 0.2205,
      "grad_norm": 29.0,
      "grad_norm_var": 18.0634765625,
      "learning_rate": 0.0001,
      "loss": 7.6215,
      "loss/crossentropy": 2.231374368071556,
      "loss/hidden": 3.314453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18930297438055277,
      "step": 8820
    },
    {
      "epoch": 0.22075,
      "grad_norm": 30.125,
      "grad_norm_var": 18.358333333333334,
      "learning_rate": 0.0001,
      "loss": 7.5354,
      "loss/crossentropy": 2.2225961208343508,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19855661746114492,
      "step": 8830
    },
    {
      "epoch": 0.221,
      "grad_norm": 31.625,
      "grad_norm_var": 18.312239583333334,
      "learning_rate": 0.0001,
      "loss": 7.6216,
      "loss/crossentropy": 2.1328855097293853,
      "loss/hidden": 3.40390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20832530297338964,
      "step": 8840
    },
    {
      "epoch": 0.22125,
      "grad_norm": 28.375,
      "grad_norm_var": 8.085416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3697,
      "loss/crossentropy": 2.0526474595069883,
      "loss/hidden": 3.5015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21284359116107227,
      "step": 8850
    },
    {
      "epoch": 0.2215,
      "grad_norm": 37.25,
      "grad_norm_var": 9.942708333333334,
      "learning_rate": 0.0001,
      "loss": 7.4967,
      "loss/crossentropy": 2.1749773651361464,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20905132126063108,
      "step": 8860
    },
    {
      "epoch": 0.22175,
      "grad_norm": 32.0,
      "grad_norm_var": 6.322330729166667,
      "learning_rate": 0.0001,
      "loss": 7.503,
      "loss/crossentropy": 2.108053085952997,
      "loss/hidden": 3.41953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20318386573344469,
      "step": 8870
    },
    {
      "epoch": 0.222,
      "grad_norm": 34.0,
      "grad_norm_var": 7.685416666666667,
      "learning_rate": 0.0001,
      "loss": 7.3618,
      "loss/crossentropy": 2.1234827637672424,
      "loss/hidden": 3.36796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18484860006719828,
      "step": 8880
    },
    {
      "epoch": 0.22225,
      "grad_norm": 29.75,
      "grad_norm_var": 8.220247395833333,
      "learning_rate": 0.0001,
      "loss": 7.393,
      "loss/crossentropy": 2.099182015657425,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19038616605103015,
      "step": 8890
    },
    {
      "epoch": 0.2225,
      "grad_norm": 31.0,
      "grad_norm_var": 7.569205729166667,
      "learning_rate": 0.0001,
      "loss": 7.4768,
      "loss/crossentropy": 2.148054042458534,
      "loss/hidden": 3.44765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2065868368372321,
      "step": 8900
    },
    {
      "epoch": 0.22275,
      "grad_norm": 28.75,
      "grad_norm_var": 6.009375,
      "learning_rate": 0.0001,
      "loss": 7.5001,
      "loss/crossentropy": 2.126410482823849,
      "loss/hidden": 3.4921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.210067143663764,
      "step": 8910
    },
    {
      "epoch": 0.223,
      "grad_norm": 37.5,
      "grad_norm_var": 7.870572916666666,
      "learning_rate": 0.0001,
      "loss": 7.4678,
      "loss/crossentropy": 2.1258621901273727,
      "loss/hidden": 3.428515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19669083636254073,
      "step": 8920
    },
    {
      "epoch": 0.22325,
      "grad_norm": 30.0,
      "grad_norm_var": 8.908268229166667,
      "learning_rate": 0.0001,
      "loss": 7.5039,
      "loss/crossentropy": 2.1579898312687873,
      "loss/hidden": 3.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18557702358812095,
      "step": 8930
    },
    {
      "epoch": 0.2235,
      "grad_norm": 28.5,
      "grad_norm_var": 8.448372395833333,
      "learning_rate": 0.0001,
      "loss": 7.5477,
      "loss/crossentropy": 2.2186635404825212,
      "loss/hidden": 3.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19810736775398255,
      "step": 8940
    },
    {
      "epoch": 0.22375,
      "grad_norm": 36.25,
      "grad_norm_var": 7.304166666666666,
      "learning_rate": 0.0001,
      "loss": 7.5522,
      "loss/crossentropy": 2.1860512644052505,
      "loss/hidden": 3.33203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18810464218258857,
      "step": 8950
    },
    {
      "epoch": 0.224,
      "grad_norm": 35.75,
      "grad_norm_var": 4.498372395833333,
      "learning_rate": 0.0001,
      "loss": 7.5105,
      "loss/crossentropy": 2.244818753004074,
      "loss/hidden": 3.350390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19993619099259377,
      "step": 8960
    },
    {
      "epoch": 0.22425,
      "grad_norm": 33.25,
      "grad_norm_var": 4.34375,
      "learning_rate": 0.0001,
      "loss": 7.6077,
      "loss/crossentropy": 2.1209320187568665,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2167982429265976,
      "step": 8970
    },
    {
      "epoch": 0.2245,
      "grad_norm": 31.0,
      "grad_norm_var": 18.992122395833334,
      "learning_rate": 0.0001,
      "loss": 7.5306,
      "loss/crossentropy": 2.008339713513851,
      "loss/hidden": 3.43515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19748742282390594,
      "step": 8980
    },
    {
      "epoch": 0.22475,
      "grad_norm": 29.5,
      "grad_norm_var": 20.906705729166667,
      "learning_rate": 0.0001,
      "loss": 7.5837,
      "loss/crossentropy": 2.245701877772808,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19117865581065416,
      "step": 8990
    },
    {
      "epoch": 0.225,
      "grad_norm": 32.75,
      "grad_norm_var": 5.1119140625,
      "learning_rate": 0.0001,
      "loss": 7.5251,
      "loss/crossentropy": 2.050145834684372,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19412143267691134,
      "step": 9000
    },
    {
      "epoch": 0.22525,
      "grad_norm": 32.75,
      "grad_norm_var": 7.176822916666667,
      "learning_rate": 0.0001,
      "loss": 7.5851,
      "loss/crossentropy": 2.190397572517395,
      "loss/hidden": 3.349609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20755842179059983,
      "step": 9010
    },
    {
      "epoch": 0.2255,
      "grad_norm": 30.75,
      "grad_norm_var": 8.977083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4264,
      "loss/crossentropy": 2.284222900867462,
      "loss/hidden": 3.332421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21356079690158367,
      "step": 9020
    },
    {
      "epoch": 0.22575,
      "grad_norm": 31.75,
      "grad_norm_var": 3.0639973958333333,
      "learning_rate": 0.0001,
      "loss": 7.5988,
      "loss/crossentropy": 2.1513551443815233,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19452885985374452,
      "step": 9030
    },
    {
      "epoch": 0.226,
      "grad_norm": 31.25,
      "grad_norm_var": 1.7830729166666666,
      "learning_rate": 0.0001,
      "loss": 7.4915,
      "loss/crossentropy": 2.2022497206926346,
      "loss/hidden": 3.415625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1973773717880249,
      "step": 9040
    },
    {
      "epoch": 0.22625,
      "grad_norm": 28.25,
      "grad_norm_var": 2.584830729166667,
      "learning_rate": 0.0001,
      "loss": 7.5443,
      "loss/crossentropy": 2.1286503240466117,
      "loss/hidden": 3.40546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18610758930444718,
      "step": 9050
    },
    {
      "epoch": 0.2265,
      "grad_norm": 27.75,
      "grad_norm_var": 3.345572916666667,
      "learning_rate": 0.0001,
      "loss": 7.5659,
      "loss/crossentropy": 2.1127908319234847,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19702097605913876,
      "step": 9060
    },
    {
      "epoch": 0.22675,
      "grad_norm": 29.0,
      "grad_norm_var": 2.716080729166667,
      "learning_rate": 0.0001,
      "loss": 7.5575,
      "loss/crossentropy": 2.0748134687542916,
      "loss/hidden": 3.331640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19112884253263474,
      "step": 9070
    },
    {
      "epoch": 0.227,
      "grad_norm": 30.875,
      "grad_norm_var": 1.0764973958333333,
      "learning_rate": 0.0001,
      "loss": 7.6302,
      "loss/crossentropy": 2.1477744698524477,
      "loss/hidden": 3.391796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.198454162850976,
      "step": 9080
    },
    {
      "epoch": 0.22725,
      "grad_norm": 31.0,
      "grad_norm_var": 2.3197265625,
      "learning_rate": 0.0001,
      "loss": 7.4827,
      "loss/crossentropy": 2.1293207883834837,
      "loss/hidden": 3.55625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21182843409478663,
      "step": 9090
    },
    {
      "epoch": 0.2275,
      "grad_norm": 34.25,
      "grad_norm_var": 3.1333333333333333,
      "learning_rate": 0.0001,
      "loss": 7.5027,
      "loss/crossentropy": 2.145271519571543,
      "loss/hidden": 3.391015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18950196234509348,
      "step": 9100
    },
    {
      "epoch": 0.22775,
      "grad_norm": 30.5,
      "grad_norm_var": 2.9337890625,
      "learning_rate": 0.0001,
      "loss": 7.5571,
      "loss/crossentropy": 2.1665233090519904,
      "loss/hidden": 3.48046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20093025229871272,
      "step": 9110
    },
    {
      "epoch": 0.228,
      "grad_norm": 31.5,
      "grad_norm_var": 1.9718098958333334,
      "learning_rate": 0.0001,
      "loss": 7.4992,
      "loss/crossentropy": 2.119904951751232,
      "loss/hidden": 3.37890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19110188409686088,
      "step": 9120
    },
    {
      "epoch": 0.22825,
      "grad_norm": 32.5,
      "grad_norm_var": 1.21875,
      "learning_rate": 0.0001,
      "loss": 7.5861,
      "loss/crossentropy": 2.169039398431778,
      "loss/hidden": 3.344140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18969318978488445,
      "step": 9130
    },
    {
      "epoch": 0.2285,
      "grad_norm": 31.125,
      "grad_norm_var": 2.0134765625,
      "learning_rate": 0.0001,
      "loss": 7.4457,
      "loss/crossentropy": 2.202808904647827,
      "loss/hidden": 3.415625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21587605867534876,
      "step": 9140
    },
    {
      "epoch": 0.22875,
      "grad_norm": 30.125,
      "grad_norm_var": 1.9014973958333334,
      "learning_rate": 0.0001,
      "loss": 7.5869,
      "loss/crossentropy": 2.188230502605438,
      "loss/hidden": 3.32578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20162757411599158,
      "step": 9150
    },
    {
      "epoch": 0.229,
      "grad_norm": 31.125,
      "grad_norm_var": 1.6268229166666666,
      "learning_rate": 0.0001,
      "loss": 7.5041,
      "loss/crossentropy": 2.166210785508156,
      "loss/hidden": 3.346875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19896902665495872,
      "step": 9160
    },
    {
      "epoch": 0.22925,
      "grad_norm": 31.875,
      "grad_norm_var": 1.3582682291666666,
      "learning_rate": 0.0001,
      "loss": 7.4795,
      "loss/crossentropy": 2.1638981848955154,
      "loss/hidden": 3.28046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18999446034431458,
      "step": 9170
    },
    {
      "epoch": 0.2295,
      "grad_norm": 34.0,
      "grad_norm_var": 2.6177083333333333,
      "learning_rate": 0.0001,
      "loss": 7.4529,
      "loss/crossentropy": 2.170040412247181,
      "loss/hidden": 3.399609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22052502054721118,
      "step": 9180
    },
    {
      "epoch": 0.22975,
      "grad_norm": 31.0,
      "grad_norm_var": 3.2400390625,
      "learning_rate": 0.0001,
      "loss": 7.4625,
      "loss/crossentropy": 2.114721930027008,
      "loss/hidden": 3.40859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19851209484040738,
      "step": 9190
    },
    {
      "epoch": 0.23,
      "grad_norm": 31.375,
      "grad_norm_var": 2.2457682291666665,
      "learning_rate": 0.0001,
      "loss": 7.5862,
      "loss/crossentropy": 2.1541184708476067,
      "loss/hidden": 3.374609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1960083631798625,
      "step": 9200
    },
    {
      "epoch": 0.23025,
      "grad_norm": 31.25,
      "grad_norm_var": 2.676041666666667,
      "learning_rate": 0.0001,
      "loss": 7.4881,
      "loss/crossentropy": 2.1309454582631586,
      "loss/hidden": 3.355078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19452987853437662,
      "step": 9210
    },
    {
      "epoch": 0.2305,
      "grad_norm": 31.75,
      "grad_norm_var": 3.671875,
      "learning_rate": 0.0001,
      "loss": 7.4938,
      "loss/crossentropy": 2.18781658411026,
      "loss/hidden": 3.385546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2072685670107603,
      "step": 9220
    },
    {
      "epoch": 0.23075,
      "grad_norm": 33.25,
      "grad_norm_var": 2.2556640625,
      "learning_rate": 0.0001,
      "loss": 7.4355,
      "loss/crossentropy": 2.1591438576579094,
      "loss/hidden": 3.337890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18837961312383414,
      "step": 9230
    },
    {
      "epoch": 0.231,
      "grad_norm": 30.25,
      "grad_norm_var": 2.2087890625,
      "learning_rate": 0.0001,
      "loss": 7.4499,
      "loss/crossentropy": 2.133663722872734,
      "loss/hidden": 3.376953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18666253574192523,
      "step": 9240
    },
    {
      "epoch": 0.23125,
      "grad_norm": 31.875,
      "grad_norm_var": 2.1806640625,
      "learning_rate": 0.0001,
      "loss": 7.5916,
      "loss/crossentropy": 2.12740980386734,
      "loss/hidden": 3.405859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19293057937175034,
      "step": 9250
    },
    {
      "epoch": 0.2315,
      "grad_norm": 32.5,
      "grad_norm_var": 2.1947265625,
      "learning_rate": 0.0001,
      "loss": 7.5307,
      "loss/crossentropy": 2.1379271537065505,
      "loss/hidden": 3.387109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19501187074929477,
      "step": 9260
    },
    {
      "epoch": 0.23175,
      "grad_norm": 30.5,
      "grad_norm_var": 2.8264973958333335,
      "learning_rate": 0.0001,
      "loss": 7.4995,
      "loss/crossentropy": 2.220619598031044,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19177700616419316,
      "step": 9270
    },
    {
      "epoch": 0.232,
      "grad_norm": 31.25,
      "grad_norm_var": 4.855208333333334,
      "learning_rate": 0.0001,
      "loss": 7.461,
      "loss/crossentropy": 2.1386048540472986,
      "loss/hidden": 3.401953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19117407780140638,
      "step": 9280
    },
    {
      "epoch": 0.23225,
      "grad_norm": 33.75,
      "grad_norm_var": 3.1122395833333334,
      "learning_rate": 0.0001,
      "loss": 7.5308,
      "loss/crossentropy": 2.201369822025299,
      "loss/hidden": 3.293359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18732962422072888,
      "step": 9290
    },
    {
      "epoch": 0.2325,
      "grad_norm": 30.75,
      "grad_norm_var": 3.4559895833333334,
      "learning_rate": 0.0001,
      "loss": 7.4868,
      "loss/crossentropy": 2.172528564929962,
      "loss/hidden": 3.3359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19425926376134156,
      "step": 9300
    },
    {
      "epoch": 0.23275,
      "grad_norm": 27.5,
      "grad_norm_var": 2.83515625,
      "learning_rate": 0.0001,
      "loss": 7.4503,
      "loss/crossentropy": 2.021423862874508,
      "loss/hidden": 3.48125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20009233951568603,
      "step": 9310
    },
    {
      "epoch": 0.233,
      "grad_norm": 32.75,
      "grad_norm_var": 5.4125,
      "learning_rate": 0.0001,
      "loss": 7.5898,
      "loss/crossentropy": 2.197957542538643,
      "loss/hidden": 3.426953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21913623586297035,
      "step": 9320
    },
    {
      "epoch": 0.23325,
      "grad_norm": 31.75,
      "grad_norm_var": 3.70390625,
      "learning_rate": 0.0001,
      "loss": 7.5196,
      "loss/crossentropy": 2.00966841802001,
      "loss/hidden": 3.449609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21357689071446656,
      "step": 9330
    },
    {
      "epoch": 0.2335,
      "grad_norm": 32.75,
      "grad_norm_var": 1.5629557291666667,
      "learning_rate": 0.0001,
      "loss": 7.5156,
      "loss/crossentropy": 2.200144296884537,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19718017429113388,
      "step": 9340
    },
    {
      "epoch": 0.23375,
      "grad_norm": 30.75,
      "grad_norm_var": 1.6541666666666666,
      "learning_rate": 0.0001,
      "loss": 7.4899,
      "loss/crossentropy": 1.991941288113594,
      "loss/hidden": 3.426171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1878137281164527,
      "step": 9350
    },
    {
      "epoch": 0.234,
      "grad_norm": 31.125,
      "grad_norm_var": 1.6934895833333334,
      "learning_rate": 0.0001,
      "loss": 7.5461,
      "loss/crossentropy": 2.0645473077893257,
      "loss/hidden": 3.451953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20249587260186672,
      "step": 9360
    },
    {
      "epoch": 0.23425,
      "grad_norm": 30.625,
      "grad_norm_var": 3.3462890625,
      "learning_rate": 0.0001,
      "loss": 7.4449,
      "loss/crossentropy": 2.203160837292671,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20446450710296632,
      "step": 9370
    },
    {
      "epoch": 0.2345,
      "grad_norm": 32.25,
      "grad_norm_var": 4.572330729166667,
      "learning_rate": 0.0001,
      "loss": 7.451,
      "loss/crossentropy": 2.1672566562891005,
      "loss/hidden": 3.419140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22198531460016965,
      "step": 9380
    },
    {
      "epoch": 0.23475,
      "grad_norm": 30.5,
      "grad_norm_var": 3.42265625,
      "learning_rate": 0.0001,
      "loss": 7.6011,
      "loss/crossentropy": 2.0229784891009333,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18394104316830634,
      "step": 9390
    },
    {
      "epoch": 0.235,
      "grad_norm": 30.75,
      "grad_norm_var": 4.330208333333333,
      "learning_rate": 0.0001,
      "loss": 7.4942,
      "loss/crossentropy": 2.1677876338362694,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22315683960914612,
      "step": 9400
    },
    {
      "epoch": 0.23525,
      "grad_norm": 31.375,
      "grad_norm_var": 19.045572916666668,
      "learning_rate": 0.0001,
      "loss": 7.5569,
      "loss/crossentropy": 2.2457904130220414,
      "loss/hidden": 3.35625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20503853298723698,
      "step": 9410
    },
    {
      "epoch": 0.2355,
      "grad_norm": 31.0,
      "grad_norm_var": 18.926822916666666,
      "learning_rate": 0.0001,
      "loss": 7.4461,
      "loss/crossentropy": 2.142396827042103,
      "loss/hidden": 3.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1910883378237486,
      "step": 9420
    },
    {
      "epoch": 0.23575,
      "grad_norm": 31.25,
      "grad_norm_var": 8.741666666666667,
      "learning_rate": 0.0001,
      "loss": 7.475,
      "loss/crossentropy": 2.063981272280216,
      "loss/hidden": 3.4921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20397737752646208,
      "step": 9430
    },
    {
      "epoch": 0.236,
      "grad_norm": 30.0,
      "grad_norm_var": 9.062434895833333,
      "learning_rate": 0.0001,
      "loss": 7.4995,
      "loss/crossentropy": 1.9866131611168385,
      "loss/hidden": 3.516015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19697492588311433,
      "step": 9440
    },
    {
      "epoch": 0.23625,
      "grad_norm": 40.25,
      "grad_norm_var": 12.64140625,
      "learning_rate": 0.0001,
      "loss": 7.5931,
      "loss/crossentropy": 2.168704579770565,
      "loss/hidden": 3.46796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19889446310698985,
      "step": 9450
    },
    {
      "epoch": 0.2365,
      "grad_norm": 29.25,
      "grad_norm_var": 51.16223958333333,
      "learning_rate": 0.0001,
      "loss": 7.4255,
      "loss/crossentropy": 2.241490375995636,
      "loss/hidden": 3.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19278081078082324,
      "step": 9460
    },
    {
      "epoch": 0.23675,
      "grad_norm": 31.125,
      "grad_norm_var": 49.08118489583333,
      "learning_rate": 0.0001,
      "loss": 7.44,
      "loss/crossentropy": 2.0530834168195726,
      "loss/hidden": 3.3125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1845833698287606,
      "step": 9470
    },
    {
      "epoch": 0.237,
      "grad_norm": 35.25,
      "grad_norm_var": 5.190559895833333,
      "learning_rate": 0.0001,
      "loss": 7.4893,
      "loss/crossentropy": 2.1680017322301866,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19126822520047426,
      "step": 9480
    },
    {
      "epoch": 0.23725,
      "grad_norm": 29.75,
      "grad_norm_var": 4.1728515625,
      "learning_rate": 0.0001,
      "loss": 7.5044,
      "loss/crossentropy": 2.1760893553495406,
      "loss/hidden": 3.3609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1902306279167533,
      "step": 9490
    },
    {
      "epoch": 0.2375,
      "grad_norm": 30.5,
      "grad_norm_var": 7.610872395833334,
      "learning_rate": 0.0001,
      "loss": 7.5567,
      "loss/crossentropy": 2.169127979874611,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897792138159275,
      "step": 9500
    },
    {
      "epoch": 0.23775,
      "grad_norm": 34.25,
      "grad_norm_var": 31.978059895833333,
      "learning_rate": 0.0001,
      "loss": 7.5548,
      "loss/crossentropy": 2.0474624037742615,
      "loss/hidden": 3.3921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897742820903659,
      "step": 9510
    },
    {
      "epoch": 0.238,
      "grad_norm": 28.875,
      "grad_norm_var": 33.4025390625,
      "learning_rate": 0.0001,
      "loss": 7.46,
      "loss/crossentropy": 2.167206625640392,
      "loss/hidden": 3.341796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1971887955442071,
      "step": 9520
    },
    {
      "epoch": 0.23825,
      "grad_norm": 38.0,
      "grad_norm_var": 10.677018229166666,
      "learning_rate": 0.0001,
      "loss": 7.4721,
      "loss/crossentropy": 2.160684567689896,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2026047335937619,
      "step": 9530
    },
    {
      "epoch": 0.2385,
      "grad_norm": 33.25,
      "grad_norm_var": 6.915559895833334,
      "learning_rate": 0.0001,
      "loss": 7.4439,
      "loss/crossentropy": 2.101367045938969,
      "loss/hidden": 3.316796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18591827657073737,
      "step": 9540
    },
    {
      "epoch": 0.23875,
      "grad_norm": 29.75,
      "grad_norm_var": 4.285416666666666,
      "learning_rate": 0.0001,
      "loss": 7.6017,
      "loss/crossentropy": 2.144346782565117,
      "loss/hidden": 3.458203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22905603051185608,
      "step": 9550
    },
    {
      "epoch": 0.239,
      "grad_norm": 31.125,
      "grad_norm_var": 5.608333333333333,
      "learning_rate": 0.0001,
      "loss": 7.391,
      "loss/crossentropy": 2.2163674563169478,
      "loss/hidden": 3.491015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20480377711355685,
      "step": 9560
    },
    {
      "epoch": 0.23925,
      "grad_norm": 32.75,
      "grad_norm_var": 5.554166666666666,
      "learning_rate": 0.0001,
      "loss": 7.4754,
      "loss/crossentropy": 2.05806076079607,
      "loss/hidden": 3.43046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1894669458270073,
      "step": 9570
    },
    {
      "epoch": 0.2395,
      "grad_norm": 35.75,
      "grad_norm_var": 6.116666666666666,
      "learning_rate": 0.0001,
      "loss": 7.64,
      "loss/crossentropy": 2.2116071820259093,
      "loss/hidden": 3.38046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20237535443156957,
      "step": 9580
    },
    {
      "epoch": 0.23975,
      "grad_norm": 27.625,
      "grad_norm_var": 6.3166015625,
      "learning_rate": 0.0001,
      "loss": 7.4412,
      "loss/crossentropy": 2.0997579991817474,
      "loss/hidden": 3.375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18707791212946176,
      "step": 9590
    },
    {
      "epoch": 0.24,
      "grad_norm": 31.125,
      "grad_norm_var": 4.391666666666667,
      "learning_rate": 0.0001,
      "loss": 7.5587,
      "loss/crossentropy": 2.1704364523291586,
      "loss/hidden": 3.42109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20182078052312136,
      "step": 9600
    },
    {
      "epoch": 0.24025,
      "grad_norm": 29.75,
      "grad_norm_var": 2.959375,
      "learning_rate": 0.0001,
      "loss": 7.5494,
      "loss/crossentropy": 2.1063621580600738,
      "loss/hidden": 3.4671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2080043438822031,
      "step": 9610
    },
    {
      "epoch": 0.2405,
      "grad_norm": 30.875,
      "grad_norm_var": 2.5872395833333335,
      "learning_rate": 0.0001,
      "loss": 7.4507,
      "loss/crossentropy": 2.1176367297768595,
      "loss/hidden": 3.31015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19296143669635057,
      "step": 9620
    },
    {
      "epoch": 0.24075,
      "grad_norm": 30.125,
      "grad_norm_var": 2.3082682291666665,
      "learning_rate": 0.0001,
      "loss": 7.5035,
      "loss/crossentropy": 2.216022199392319,
      "loss/hidden": 3.486328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2059534139931202,
      "step": 9630
    },
    {
      "epoch": 0.241,
      "grad_norm": 31.5,
      "grad_norm_var": 10.676041666666666,
      "learning_rate": 0.0001,
      "loss": 7.5155,
      "loss/crossentropy": 2.209742319583893,
      "loss/hidden": 3.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19856051169335842,
      "step": 9640
    },
    {
      "epoch": 0.24125,
      "grad_norm": 31.75,
      "grad_norm_var": 1.10390625,
      "learning_rate": 0.0001,
      "loss": 7.4983,
      "loss/crossentropy": 2.040982872247696,
      "loss/hidden": 3.367578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19508841075003147,
      "step": 9650
    },
    {
      "epoch": 0.2415,
      "grad_norm": 29.5,
      "grad_norm_var": 1.1697916666666666,
      "learning_rate": 0.0001,
      "loss": 7.4469,
      "loss/crossentropy": 2.1563887119293215,
      "loss/hidden": 3.2921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18813634980469943,
      "step": 9660
    },
    {
      "epoch": 0.24175,
      "grad_norm": 29.75,
      "grad_norm_var": 2.5768229166666665,
      "learning_rate": 0.0001,
      "loss": 7.5616,
      "loss/crossentropy": 2.056604099273682,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19395184125751258,
      "step": 9670
    },
    {
      "epoch": 0.242,
      "grad_norm": 30.25,
      "grad_norm_var": 2.067643229166667,
      "learning_rate": 0.0001,
      "loss": 7.4244,
      "loss/crossentropy": 2.174482125043869,
      "loss/hidden": 3.484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20453137308359146,
      "step": 9680
    },
    {
      "epoch": 0.24225,
      "grad_norm": 31.0,
      "grad_norm_var": 3.905989583333333,
      "learning_rate": 0.0001,
      "loss": 7.5003,
      "loss/crossentropy": 2.119745084643364,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20908331871032715,
      "step": 9690
    },
    {
      "epoch": 0.2425,
      "grad_norm": 29.25,
      "grad_norm_var": 3.4785807291666666,
      "learning_rate": 0.0001,
      "loss": 7.3362,
      "loss/crossentropy": 2.122311297059059,
      "loss/hidden": 3.3046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18319036178290843,
      "step": 9700
    },
    {
      "epoch": 0.24275,
      "grad_norm": 29.875,
      "grad_norm_var": 1.9541666666666666,
      "learning_rate": 0.0001,
      "loss": 7.4396,
      "loss/crossentropy": 1.9478351414203643,
      "loss/hidden": 3.437109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20451803579926492,
      "step": 9710
    },
    {
      "epoch": 0.243,
      "grad_norm": 32.75,
      "grad_norm_var": 3.3997395833333335,
      "learning_rate": 0.0001,
      "loss": 7.4967,
      "loss/crossentropy": 2.0708641201257705,
      "loss/hidden": 3.393359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19053229298442603,
      "step": 9720
    },
    {
      "epoch": 0.24325,
      "grad_norm": 33.5,
      "grad_norm_var": 3.635872395833333,
      "learning_rate": 0.0001,
      "loss": 7.5016,
      "loss/crossentropy": 2.0913224294781685,
      "loss/hidden": 3.3953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1940496889874339,
      "step": 9730
    },
    {
      "epoch": 0.2435,
      "grad_norm": 30.625,
      "grad_norm_var": 3.3291015625,
      "learning_rate": 0.0001,
      "loss": 7.5077,
      "loss/crossentropy": 2.221079145371914,
      "loss/hidden": 3.315625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1924398820847273,
      "step": 9740
    },
    {
      "epoch": 0.24375,
      "grad_norm": 31.5,
      "grad_norm_var": 1.56875,
      "learning_rate": 0.0001,
      "loss": 7.4463,
      "loss/crossentropy": 2.148006671667099,
      "loss/hidden": 3.3765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20756886284798384,
      "step": 9750
    },
    {
      "epoch": 0.244,
      "grad_norm": 30.25,
      "grad_norm_var": 8.2978515625,
      "learning_rate": 0.0001,
      "loss": 7.4544,
      "loss/crossentropy": 2.2184116363525392,
      "loss/hidden": 3.359765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19135836903005837,
      "step": 9760
    },
    {
      "epoch": 0.24425,
      "grad_norm": 29.75,
      "grad_norm_var": 4.690625,
      "learning_rate": 0.0001,
      "loss": 7.4095,
      "loss/crossentropy": 2.1005363285541536,
      "loss/hidden": 3.36640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19765366949141025,
      "step": 9770
    },
    {
      "epoch": 0.2445,
      "grad_norm": 30.5,
      "grad_norm_var": 1.5327473958333333,
      "learning_rate": 0.0001,
      "loss": 7.5002,
      "loss/crossentropy": 2.167180609703064,
      "loss/hidden": 3.337890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19258719477802516,
      "step": 9780
    },
    {
      "epoch": 0.24475,
      "grad_norm": 29.875,
      "grad_norm_var": 2.4697916666666666,
      "learning_rate": 0.0001,
      "loss": 7.3885,
      "loss/crossentropy": 1.9792610332369804,
      "loss/hidden": 3.3671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18384026363492012,
      "step": 9790
    },
    {
      "epoch": 0.245,
      "grad_norm": 33.75,
      "grad_norm_var": 2.5447265625,
      "learning_rate": 0.0001,
      "loss": 7.5547,
      "loss/crossentropy": 2.0596527442336083,
      "loss/hidden": 3.569921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20697965249419212,
      "step": 9800
    },
    {
      "epoch": 0.24525,
      "grad_norm": 32.75,
      "grad_norm_var": 5.042122395833333,
      "learning_rate": 0.0001,
      "loss": 7.4887,
      "loss/crossentropy": 2.132730546593666,
      "loss/hidden": 3.498828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19533955454826354,
      "step": 9810
    },
    {
      "epoch": 0.2455,
      "grad_norm": 32.25,
      "grad_norm_var": 5.31015625,
      "learning_rate": 0.0001,
      "loss": 7.5278,
      "loss/crossentropy": 2.151031643152237,
      "loss/hidden": 3.450390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19322178810834884,
      "step": 9820
    },
    {
      "epoch": 0.24575,
      "grad_norm": 30.75,
      "grad_norm_var": 1.6059895833333333,
      "learning_rate": 0.0001,
      "loss": 7.5534,
      "loss/crossentropy": 2.143332117795944,
      "loss/hidden": 3.337109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897228866815567,
      "step": 9830
    },
    {
      "epoch": 0.246,
      "grad_norm": 31.75,
      "grad_norm_var": 3.4395182291666666,
      "learning_rate": 0.0001,
      "loss": 7.5657,
      "loss/crossentropy": 2.1097486332058906,
      "loss/hidden": 3.45390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20945656076073646,
      "step": 9840
    },
    {
      "epoch": 0.24625,
      "grad_norm": 32.75,
      "grad_norm_var": 2.9452473958333334,
      "learning_rate": 0.0001,
      "loss": 7.5013,
      "loss/crossentropy": 2.065919445455074,
      "loss/hidden": 3.401171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18999069584533573,
      "step": 9850
    },
    {
      "epoch": 0.2465,
      "grad_norm": 31.875,
      "grad_norm_var": 2.2322265625,
      "learning_rate": 0.0001,
      "loss": 7.5501,
      "loss/crossentropy": 2.183481493592262,
      "loss/hidden": 3.504296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21892244052141904,
      "step": 9860
    },
    {
      "epoch": 0.24675,
      "grad_norm": 32.75,
      "grad_norm_var": 2.4926432291666667,
      "learning_rate": 0.0001,
      "loss": 7.478,
      "loss/crossentropy": 2.1966081470251084,
      "loss/hidden": 3.402734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20018710754811764,
      "step": 9870
    },
    {
      "epoch": 0.247,
      "grad_norm": 31.0,
      "grad_norm_var": 2.2244140625,
      "learning_rate": 0.0001,
      "loss": 7.5127,
      "loss/crossentropy": 1.9914133831858636,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19236233234405517,
      "step": 9880
    },
    {
      "epoch": 0.24725,
      "grad_norm": 32.25,
      "grad_norm_var": 1.8900390625,
      "learning_rate": 0.0001,
      "loss": 7.5007,
      "loss/crossentropy": 2.1605743557214736,
      "loss/hidden": 3.342578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1983661765232682,
      "step": 9890
    },
    {
      "epoch": 0.2475,
      "grad_norm": 32.75,
      "grad_norm_var": 2.0666015625,
      "learning_rate": 0.0001,
      "loss": 7.5307,
      "loss/crossentropy": 2.0920628547668456,
      "loss/hidden": 3.404296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19224398005753757,
      "step": 9900
    },
    {
      "epoch": 0.24775,
      "grad_norm": 29.75,
      "grad_norm_var": 28.910416666666666,
      "learning_rate": 0.0001,
      "loss": 7.4845,
      "loss/crossentropy": 2.13549183011055,
      "loss/hidden": 3.435546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2024377616122365,
      "step": 9910
    },
    {
      "epoch": 0.248,
      "grad_norm": 33.5,
      "grad_norm_var": 3.2122395833333335,
      "learning_rate": 0.0001,
      "loss": 7.5567,
      "loss/crossentropy": 2.077779620885849,
      "loss/hidden": 3.429296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1914847193285823,
      "step": 9920
    },
    {
      "epoch": 0.24825,
      "grad_norm": 32.5,
      "grad_norm_var": 2.3416015625,
      "learning_rate": 0.0001,
      "loss": 7.6034,
      "loss/crossentropy": 2.24170383810997,
      "loss/hidden": 3.416796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21483709737658502,
      "step": 9930
    },
    {
      "epoch": 0.2485,
      "grad_norm": 29.25,
      "grad_norm_var": 1.9525390625,
      "learning_rate": 0.0001,
      "loss": 7.5277,
      "loss/crossentropy": 2.0574432730674745,
      "loss/hidden": 3.403125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19156391881406307,
      "step": 9940
    },
    {
      "epoch": 0.24875,
      "grad_norm": 29.375,
      "grad_norm_var": 1.6181640625,
      "learning_rate": 0.0001,
      "loss": 7.4432,
      "loss/crossentropy": 2.231716367602348,
      "loss/hidden": 3.271484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19992320109158754,
      "step": 9950
    },
    {
      "epoch": 0.249,
      "grad_norm": 30.625,
      "grad_norm_var": 2.496809895833333,
      "learning_rate": 0.0001,
      "loss": 7.5233,
      "loss/crossentropy": 2.2805121034383773,
      "loss/hidden": 3.376171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21222201287746428,
      "step": 9960
    },
    {
      "epoch": 0.24925,
      "grad_norm": 30.25,
      "grad_norm_var": 3.6143229166666666,
      "learning_rate": 0.0001,
      "loss": 7.5022,
      "loss/crossentropy": 2.1643686085939406,
      "loss/hidden": 3.45078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2017618851736188,
      "step": 9970
    },
    {
      "epoch": 0.2495,
      "grad_norm": 33.25,
      "grad_norm_var": 2.9947265625,
      "learning_rate": 0.0001,
      "loss": 7.5351,
      "loss/crossentropy": 2.052297804504633,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19462402295321227,
      "step": 9980
    },
    {
      "epoch": 0.24975,
      "grad_norm": 28.625,
      "grad_norm_var": 1.8958333333333333,
      "learning_rate": 0.0001,
      "loss": 7.4501,
      "loss/crossentropy": 2.1721100926399233,
      "loss/hidden": 3.3234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20133505929261447,
      "step": 9990
    },
    {
      "epoch": 0.25,
      "grad_norm": 28.0,
      "grad_norm_var": 2.2645833333333334,
      "learning_rate": 0.0001,
      "loss": 7.5168,
      "loss/crossentropy": 2.0236786626279355,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18019386120140551,
      "step": 10000
    }
  ],
  "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": 2.8575100320088064e+19,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}