{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 2000,
  "global_step": 10000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.001,
      "grad_norm": 4736.0,
      "learning_rate": 1.9e-05,
      "loss": 132.1055,
      "loss/crossentropy": 12.246079635620116,
      "loss/hidden": 18.7125,
      "loss/jsd": 0.0,
      "loss/logits": 10.372939014434815,
      "step": 10
    },
    {
      "epoch": 0.002,
      "grad_norm": 330.0,
      "grad_norm_var": 91640269.18333334,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 95.9731,
      "loss/crossentropy": 8.862393474578857,
      "loss/hidden": 18.675,
      "loss/jsd": 0.0,
      "loss/logits": 6.677179157733917,
      "step": 20
    },
    {
      "epoch": 0.003,
      "grad_norm": 394.0,
      "grad_norm_var": 237715.45,
      "learning_rate": 3.7e-05,
      "loss": 86.3778,
      "loss/crossentropy": 8.083840227127075,
      "loss/hidden": 18.259375,
      "loss/jsd": 0.0,
      "loss/logits": 6.130921971797943,
      "step": 30
    },
    {
      "epoch": 0.004,
      "grad_norm": 924.0,
      "grad_norm_var": 2.6757682503402172e+16,
      "learning_rate": 4.600000000000001e-05,
      "loss": 82.5914,
      "loss/crossentropy": 7.802511918544769,
      "loss/hidden": 17.440625,
      "loss/jsd": 0.0,
      "loss/logits": 5.772503018379211,
      "step": 40
    },
    {
      "epoch": 0.005,
      "grad_norm": 516.0,
      "grad_norm_var": 38597.583333333336,
      "learning_rate": 5.500000000000001e-05,
      "loss": 75.3397,
      "loss/crossentropy": 7.156700026988983,
      "loss/hidden": 17.253125,
      "loss/jsd": 0.0,
      "loss/logits": 5.156575608253479,
      "step": 50
    },
    {
      "epoch": 0.006,
      "grad_norm": 1232.0,
      "grad_norm_var": 68241.45,
      "learning_rate": 6.400000000000001e-05,
      "loss": 61.2745,
      "loss/crossentropy": 6.0138510942459105,
      "loss/hidden": 15.80625,
      "loss/jsd": 0.0,
      "loss/logits": 3.8037488579750063,
      "step": 60
    },
    {
      "epoch": 0.007,
      "grad_norm": 376.0,
      "grad_norm_var": 626103.4,
      "learning_rate": 7.3e-05,
      "loss": 41.3695,
      "loss/crossentropy": 4.422797441482544,
      "loss/hidden": 13.1125,
      "loss/jsd": 0.0,
      "loss/logits": 2.4006322652101515,
      "step": 70
    },
    {
      "epoch": 0.008,
      "grad_norm": 272.0,
      "grad_norm_var": 674923.45,
      "learning_rate": 8.200000000000001e-05,
      "loss": 27.4755,
      "loss/crossentropy": 3.3576226443052293,
      "loss/hidden": 10.7359375,
      "loss/jsd": 0.0,
      "loss/logits": 1.3968962401151657,
      "step": 80
    },
    {
      "epoch": 0.009,
      "grad_norm": 296.0,
      "grad_norm_var": 15426.383333333333,
      "learning_rate": 9.1e-05,
      "loss": 22.6607,
      "loss/crossentropy": 3.217679074406624,
      "loss/hidden": 9.2140625,
      "loss/jsd": 0.0,
      "loss/logits": 1.055714099109173,
      "step": 90
    },
    {
      "epoch": 0.01,
      "grad_norm": 328.0,
      "grad_norm_var": 9349.666666666666,
      "learning_rate": 0.0001,
      "loss": 20.3108,
      "loss/crossentropy": 2.934060016274452,
      "loss/hidden": 8.40703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.8702833190560341,
      "step": 100
    },
    {
      "epoch": 0.011,
      "grad_norm": 194.0,
      "grad_norm_var": 5992.866666666667,
      "learning_rate": 0.0001,
      "loss": 18.8852,
      "loss/crossentropy": 2.8450062334537507,
      "loss/hidden": 8.221875,
      "loss/jsd": 0.0,
      "loss/logits": 0.8380498677492142,
      "step": 110
    },
    {
      "epoch": 0.012,
      "grad_norm": 244.0,
      "grad_norm_var": 1176.5333333333333,
      "learning_rate": 0.0001,
      "loss": 17.97,
      "loss/crossentropy": 2.612249107658863,
      "loss/hidden": 7.578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.686215291172266,
      "step": 120
    },
    {
      "epoch": 0.013,
      "grad_norm": 242.0,
      "grad_norm_var": 1168.8958333333333,
      "learning_rate": 0.0001,
      "loss": 17.2904,
      "loss/crossentropy": 2.8242316216230394,
      "loss/hidden": 7.7390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.7805894792079926,
      "step": 130
    },
    {
      "epoch": 0.014,
      "grad_norm": 179.0,
      "grad_norm_var": 1465.1333333333334,
      "learning_rate": 0.0001,
      "loss": 16.5581,
      "loss/crossentropy": 2.737143725156784,
      "loss/hidden": 7.3421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.6888546235859394,
      "step": 140
    },
    {
      "epoch": 0.015,
      "grad_norm": 175.0,
      "grad_norm_var": 1119.8625,
      "learning_rate": 0.0001,
      "loss": 16.0501,
      "loss/crossentropy": 2.7599751561880113,
      "loss/hidden": 7.05703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.6640767879784107,
      "step": 150
    },
    {
      "epoch": 0.016,
      "grad_norm": 186.0,
      "grad_norm_var": 1044.5166666666667,
      "learning_rate": 0.0001,
      "loss": 15.4631,
      "loss/crossentropy": 2.6100075274705885,
      "loss/hidden": 6.8203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.5824844464659691,
      "step": 160
    },
    {
      "epoch": 0.017,
      "grad_norm": 179.0,
      "grad_norm_var": 1082.8,
      "learning_rate": 0.0001,
      "loss": 15.2201,
      "loss/crossentropy": 2.4276285111904143,
      "loss/hidden": 6.8203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.5915141828358174,
      "step": 170
    },
    {
      "epoch": 0.018,
      "grad_norm": 153.0,
      "grad_norm_var": 622.6625,
      "learning_rate": 0.0001,
      "loss": 14.9606,
      "loss/crossentropy": 2.630460512638092,
      "loss/hidden": 6.52578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.5396774187684059,
      "step": 180
    },
    {
      "epoch": 0.019,
      "grad_norm": 176.0,
      "grad_norm_var": 1093.2,
      "learning_rate": 0.0001,
      "loss": 14.6255,
      "loss/crossentropy": 2.3158223152160646,
      "loss/hidden": 6.50390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.4905257746577263,
      "step": 190
    },
    {
      "epoch": 0.02,
      "grad_norm": 112.0,
      "grad_norm_var": 695.7291666666666,
      "learning_rate": 0.0001,
      "loss": 14.3647,
      "loss/crossentropy": 2.586851382255554,
      "loss/hidden": 6.42265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.5586091712117195,
      "step": 200
    },
    {
      "epoch": 0.021,
      "grad_norm": 118.5,
      "grad_norm_var": 574.3072916666666,
      "learning_rate": 0.0001,
      "loss": 14.0867,
      "loss/crossentropy": 2.5010055124759676,
      "loss/hidden": 6.34453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.4965482771396637,
      "step": 210
    },
    {
      "epoch": 0.022,
      "grad_norm": 88.5,
      "grad_norm_var": 662.65,
      "learning_rate": 0.0001,
      "loss": 13.6551,
      "loss/crossentropy": 2.573444625735283,
      "loss/hidden": 6.33125,
      "loss/jsd": 0.0,
      "loss/logits": 0.5534068010747433,
      "step": 220
    },
    {
      "epoch": 0.023,
      "grad_norm": 118.0,
      "grad_norm_var": 412.1958333333333,
      "learning_rate": 0.0001,
      "loss": 13.4715,
      "loss/crossentropy": 2.4142292886972427,
      "loss/hidden": 5.96640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.44360905699431896,
      "step": 230
    },
    {
      "epoch": 0.024,
      "grad_norm": 134.0,
      "grad_norm_var": 242.9,
      "learning_rate": 0.0001,
      "loss": 13.3289,
      "loss/crossentropy": 2.4670142769813537,
      "loss/hidden": 5.98671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.47392544001340864,
      "step": 240
    },
    {
      "epoch": 0.025,
      "grad_norm": 137.0,
      "grad_norm_var": 158.4625,
      "learning_rate": 0.0001,
      "loss": 13.0031,
      "loss/crossentropy": 2.416000656783581,
      "loss/hidden": 5.7859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.44607544504106045,
      "step": 250
    },
    {
      "epoch": 0.026,
      "grad_norm": 109.0,
      "grad_norm_var": 279.990625,
      "learning_rate": 0.0001,
      "loss": 13.0076,
      "loss/crossentropy": 2.370332670211792,
      "loss/hidden": 5.9984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.5006627842783928,
      "step": 260
    },
    {
      "epoch": 0.027,
      "grad_norm": 129.0,
      "grad_norm_var": 427.37395833333335,
      "learning_rate": 0.0001,
      "loss": 12.8809,
      "loss/crossentropy": 2.281908763945103,
      "loss/hidden": 5.98671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.45061586182564495,
      "step": 270
    },
    {
      "epoch": 0.028,
      "grad_norm": 98.0,
      "grad_norm_var": 278.1489583333333,
      "learning_rate": 0.0001,
      "loss": 12.8942,
      "loss/crossentropy": 2.3922384053468706,
      "loss/hidden": 5.6984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.44376694336533545,
      "step": 280
    },
    {
      "epoch": 0.029,
      "grad_norm": 99.5,
      "grad_norm_var": 303.55,
      "learning_rate": 0.0001,
      "loss": 12.7122,
      "loss/crossentropy": 2.730095013976097,
      "loss/hidden": 5.49140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.4411045670509338,
      "step": 290
    },
    {
      "epoch": 0.03,
      "grad_norm": 112.5,
      "grad_norm_var": 359.56666666666666,
      "learning_rate": 0.0001,
      "loss": 12.5618,
      "loss/crossentropy": 2.3741705983877184,
      "loss/hidden": 5.43203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.40091707594692705,
      "step": 300
    },
    {
      "epoch": 0.031,
      "grad_norm": 84.5,
      "grad_norm_var": 245.25729166666667,
      "learning_rate": 0.0001,
      "loss": 12.2525,
      "loss/crossentropy": 2.2781229317188263,
      "loss/hidden": 5.53515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.4274128321558237,
      "step": 310
    },
    {
      "epoch": 0.032,
      "grad_norm": 108.5,
      "grad_norm_var": 140.59583333333333,
      "learning_rate": 0.0001,
      "loss": 12.2935,
      "loss/crossentropy": 2.5757294684648513,
      "loss/hidden": 5.4609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.42916890494525434,
      "step": 320
    },
    {
      "epoch": 0.033,
      "grad_norm": 108.0,
      "grad_norm_var": 70.89895833333334,
      "learning_rate": 0.0001,
      "loss": 12.1545,
      "loss/crossentropy": 2.527638339996338,
      "loss/hidden": 5.378125,
      "loss/jsd": 0.0,
      "loss/logits": 0.4032053742557764,
      "step": 330
    },
    {
      "epoch": 0.034,
      "grad_norm": 210.0,
      "grad_norm_var": 1272.465625,
      "learning_rate": 0.0001,
      "loss": 12.2482,
      "loss/crossentropy": 2.5401821002364158,
      "loss/hidden": 5.390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.4444709587842226,
      "step": 340
    },
    {
      "epoch": 0.035,
      "grad_norm": 79.5,
      "grad_norm_var": 1376.5958333333333,
      "learning_rate": 0.0001,
      "loss": 12.08,
      "loss/crossentropy": 2.514840933680534,
      "loss/hidden": 5.2640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.4077944982796907,
      "step": 350
    },
    {
      "epoch": 0.036,
      "grad_norm": 87.0,
      "grad_norm_var": 418.83229166666666,
      "learning_rate": 0.0001,
      "loss": 12.0245,
      "loss/crossentropy": 2.420889538526535,
      "loss/hidden": 5.34921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.44222328886389733,
      "step": 360
    },
    {
      "epoch": 0.037,
      "grad_norm": 76.5,
      "grad_norm_var": 138.5625,
      "learning_rate": 0.0001,
      "loss": 11.7097,
      "loss/crossentropy": 2.2826619133353234,
      "loss/hidden": 5.3296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3849468305706978,
      "step": 370
    },
    {
      "epoch": 0.038,
      "grad_norm": 96.5,
      "grad_norm_var": 184.93229166666666,
      "learning_rate": 0.0001,
      "loss": 11.465,
      "loss/crossentropy": 2.4052042722702027,
      "loss/hidden": 5.16796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.40173302926123144,
      "step": 380
    },
    {
      "epoch": 0.039,
      "grad_norm": 125.5,
      "grad_norm_var": 183.09583333333333,
      "learning_rate": 0.0001,
      "loss": 11.6273,
      "loss/crossentropy": 2.540145033597946,
      "loss/hidden": 5.215625,
      "loss/jsd": 0.0,
      "loss/logits": 0.41224894523620603,
      "step": 390
    },
    {
      "epoch": 0.04,
      "grad_norm": 83.5,
      "grad_norm_var": 258.315625,
      "learning_rate": 0.0001,
      "loss": 11.397,
      "loss/crossentropy": 2.207468980550766,
      "loss/hidden": 5.09296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3590874429792166,
      "step": 400
    },
    {
      "epoch": 0.041,
      "grad_norm": 94.5,
      "grad_norm_var": 184.5625,
      "learning_rate": 0.0001,
      "loss": 11.443,
      "loss/crossentropy": 2.4378984421491623,
      "loss/hidden": 5.21171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.40493359677493573,
      "step": 410
    },
    {
      "epoch": 0.042,
      "grad_norm": 106.5,
      "grad_norm_var": 125.590625,
      "learning_rate": 0.0001,
      "loss": 11.5678,
      "loss/crossentropy": 2.518555220961571,
      "loss/hidden": 5.07265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.4297170080244541,
      "step": 420
    },
    {
      "epoch": 0.043,
      "grad_norm": 87.5,
      "grad_norm_var": 115.765625,
      "learning_rate": 0.0001,
      "loss": 11.3132,
      "loss/crossentropy": 2.490597203373909,
      "loss/hidden": 5.11171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.403754598274827,
      "step": 430
    },
    {
      "epoch": 0.044,
      "grad_norm": 92.5,
      "grad_norm_var": 156.35729166666667,
      "learning_rate": 0.0001,
      "loss": 11.1476,
      "loss/crossentropy": 2.037529316544533,
      "loss/hidden": 5.07421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.35246654506772757,
      "step": 440
    },
    {
      "epoch": 0.045,
      "grad_norm": 80.5,
      "grad_norm_var": 210.66666666666666,
      "learning_rate": 0.0001,
      "loss": 11.3038,
      "loss/crossentropy": 2.3201738983392715,
      "loss/hidden": 5.0828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.38196625709533694,
      "step": 450
    },
    {
      "epoch": 0.046,
      "grad_norm": 107.5,
      "grad_norm_var": 284.1666666666667,
      "learning_rate": 0.0001,
      "loss": 11.3625,
      "loss/crossentropy": 2.4791718110442162,
      "loss/hidden": 4.95546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.36495909169316293,
      "step": 460
    },
    {
      "epoch": 0.047,
      "grad_norm": 91.5,
      "grad_norm_var": 247.39895833333333,
      "learning_rate": 0.0001,
      "loss": 11.0542,
      "loss/crossentropy": 2.3155667960643767,
      "loss/hidden": 4.93828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.362844867631793,
      "step": 470
    },
    {
      "epoch": 0.048,
      "grad_norm": 95.0,
      "grad_norm_var": 194.79895833333333,
      "learning_rate": 0.0001,
      "loss": 11.2413,
      "loss/crossentropy": 2.496318203210831,
      "loss/hidden": 4.840625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3887303464114666,
      "step": 480
    },
    {
      "epoch": 0.049,
      "grad_norm": 74.5,
      "grad_norm_var": 243.840625,
      "learning_rate": 0.0001,
      "loss": 10.9416,
      "loss/crossentropy": 2.385223904252052,
      "loss/hidden": 4.85234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3598880790174007,
      "step": 490
    },
    {
      "epoch": 0.05,
      "grad_norm": 79.0,
      "grad_norm_var": 105.990625,
      "learning_rate": 0.0001,
      "loss": 10.9114,
      "loss/crossentropy": 2.2462552055716514,
      "loss/hidden": 4.80859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3265662036836147,
      "step": 500
    },
    {
      "epoch": 0.051,
      "grad_norm": 96.5,
      "grad_norm_var": 138.43229166666666,
      "learning_rate": 0.0001,
      "loss": 10.8821,
      "loss/crossentropy": 2.297148121893406,
      "loss/hidden": 4.8609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3467547960579395,
      "step": 510
    },
    {
      "epoch": 0.052,
      "grad_norm": 97.5,
      "grad_norm_var": 129.365625,
      "learning_rate": 0.0001,
      "loss": 10.9299,
      "loss/crossentropy": 2.4197026968002318,
      "loss/hidden": 4.7921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3632193084806204,
      "step": 520
    },
    {
      "epoch": 0.053,
      "grad_norm": 81.5,
      "grad_norm_var": 99.47395833333333,
      "learning_rate": 0.0001,
      "loss": 10.787,
      "loss/crossentropy": 2.36982424557209,
      "loss/hidden": 4.825,
      "loss/jsd": 0.0,
      "loss/logits": 0.3405680742114782,
      "step": 530
    },
    {
      "epoch": 0.054,
      "grad_norm": 85.5,
      "grad_norm_var": 48.340625,
      "learning_rate": 0.0001,
      "loss": 10.8675,
      "loss/crossentropy": 2.4611779801547526,
      "loss/hidden": 4.8625,
      "loss/jsd": 0.0,
      "loss/logits": 0.36872007288038733,
      "step": 540
    },
    {
      "epoch": 0.055,
      "grad_norm": 93.5,
      "grad_norm_var": 84.24895833333333,
      "learning_rate": 0.0001,
      "loss": 10.64,
      "loss/crossentropy": 2.1758567959070207,
      "loss/hidden": 4.7484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3336840860545635,
      "step": 550
    },
    {
      "epoch": 0.056,
      "grad_norm": 114.0,
      "grad_norm_var": 129.53098958333334,
      "learning_rate": 0.0001,
      "loss": 10.5615,
      "loss/crossentropy": 2.3970536097884176,
      "loss/hidden": 4.7625,
      "loss/jsd": 0.0,
      "loss/logits": 0.34276723079383375,
      "step": 560
    },
    {
      "epoch": 0.057,
      "grad_norm": 80.0,
      "grad_norm_var": 579.57890625,
      "learning_rate": 0.0001,
      "loss": 10.8999,
      "loss/crossentropy": 2.4695185527205465,
      "loss/hidden": 4.9453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.42829814068973066,
      "step": 570
    },
    {
      "epoch": 0.058,
      "grad_norm": 85.0,
      "grad_norm_var": 596.9572916666667,
      "learning_rate": 0.0001,
      "loss": 10.8802,
      "loss/crossentropy": 2.3520184576511385,
      "loss/hidden": 4.790625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3662864986807108,
      "step": 580
    },
    {
      "epoch": 0.059,
      "grad_norm": 73.0,
      "grad_norm_var": 181.69583333333333,
      "learning_rate": 0.0001,
      "loss": 10.6744,
      "loss/crossentropy": 2.2842736929655074,
      "loss/hidden": 4.71484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3500846643000841,
      "step": 590
    },
    {
      "epoch": 0.06,
      "grad_norm": 97.0,
      "grad_norm_var": 160.58307291666668,
      "learning_rate": 0.0001,
      "loss": 10.6987,
      "loss/crossentropy": 2.29906165599823,
      "loss/hidden": 4.602734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.334361494705081,
      "step": 600
    },
    {
      "epoch": 0.061,
      "grad_norm": 89.0,
      "grad_norm_var": 162.67682291666668,
      "learning_rate": 0.0001,
      "loss": 10.6143,
      "loss/crossentropy": 2.3032930195331573,
      "loss/hidden": 4.6703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3258141163736582,
      "step": 610
    },
    {
      "epoch": 0.062,
      "grad_norm": 77.5,
      "grad_norm_var": 97.12916666666666,
      "learning_rate": 0.0001,
      "loss": 10.5946,
      "loss/crossentropy": 2.452244046330452,
      "loss/hidden": 4.7109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3432691916823387,
      "step": 620
    },
    {
      "epoch": 0.063,
      "grad_norm": 75.5,
      "grad_norm_var": 227.69973958333333,
      "learning_rate": 0.0001,
      "loss": 10.6287,
      "loss/crossentropy": 2.2894835874438284,
      "loss/hidden": 4.74609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.35672005768865345,
      "step": 630
    },
    {
      "epoch": 0.064,
      "grad_norm": 70.0,
      "grad_norm_var": 541.2322916666667,
      "learning_rate": 0.0001,
      "loss": 10.6195,
      "loss/crossentropy": 2.4114772886037827,
      "loss/hidden": 4.70546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.35591375902295114,
      "step": 640
    },
    {
      "epoch": 0.065,
      "grad_norm": 77.0,
      "grad_norm_var": 435.15390625,
      "learning_rate": 0.0001,
      "loss": 10.4142,
      "loss/crossentropy": 2.332440134882927,
      "loss/hidden": 4.634375,
      "loss/jsd": 0.0,
      "loss/logits": 0.339809150993824,
      "step": 650
    },
    {
      "epoch": 0.066,
      "grad_norm": 71.5,
      "grad_norm_var": 118.03307291666667,
      "learning_rate": 0.0001,
      "loss": 10.4602,
      "loss/crossentropy": 2.154422373324633,
      "loss/hidden": 4.54140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3334257358685136,
      "step": 660
    },
    {
      "epoch": 0.067,
      "grad_norm": 73.5,
      "grad_norm_var": 144.94166666666666,
      "learning_rate": 0.0001,
      "loss": 10.5185,
      "loss/crossentropy": 2.3223402693867685,
      "loss/hidden": 4.795703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.37188967503607273,
      "step": 670
    },
    {
      "epoch": 0.068,
      "grad_norm": 61.5,
      "grad_norm_var": 169.65598958333334,
      "learning_rate": 0.0001,
      "loss": 10.5323,
      "loss/crossentropy": 2.332353001832962,
      "loss/hidden": 4.50625,
      "loss/jsd": 0.0,
      "loss/logits": 0.31948004066944125,
      "step": 680
    },
    {
      "epoch": 0.069,
      "grad_norm": 74.0,
      "grad_norm_var": 155.94140625,
      "learning_rate": 0.0001,
      "loss": 10.4359,
      "loss/crossentropy": 2.4077556908130644,
      "loss/hidden": 4.623828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.339173823595047,
      "step": 690
    },
    {
      "epoch": 0.07,
      "grad_norm": 82.5,
      "grad_norm_var": 125.55416666666666,
      "learning_rate": 0.0001,
      "loss": 10.4493,
      "loss/crossentropy": 2.292634981870651,
      "loss/hidden": 4.571875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3477486100047827,
      "step": 700
    },
    {
      "epoch": 0.071,
      "grad_norm": 88.0,
      "grad_norm_var": 155.84166666666667,
      "learning_rate": 0.0001,
      "loss": 10.2041,
      "loss/crossentropy": 2.4034020826220512,
      "loss/hidden": 4.53046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3406600248068571,
      "step": 710
    },
    {
      "epoch": 0.072,
      "grad_norm": 124.0,
      "grad_norm_var": 230.83307291666668,
      "learning_rate": 0.0001,
      "loss": 10.3489,
      "loss/crossentropy": 2.333241228759289,
      "loss/hidden": 4.6015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3285223826766014,
      "step": 720
    },
    {
      "epoch": 0.073,
      "grad_norm": 71.0,
      "grad_norm_var": 278.95390625,
      "learning_rate": 0.0001,
      "loss": 10.1548,
      "loss/crossentropy": 2.4066421508789064,
      "loss/hidden": 4.682421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.338771004602313,
      "step": 730
    },
    {
      "epoch": 0.074,
      "grad_norm": 84.5,
      "grad_norm_var": 166.85729166666667,
      "learning_rate": 0.0001,
      "loss": 10.2647,
      "loss/crossentropy": 2.2724754482507707,
      "loss/hidden": 4.567578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3267147310078144,
      "step": 740
    },
    {
      "epoch": 0.075,
      "grad_norm": 67.5,
      "grad_norm_var": 343.5247395833333,
      "learning_rate": 0.0001,
      "loss": 10.2815,
      "loss/crossentropy": 2.3046080738306047,
      "loss/hidden": 4.473828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.33236319161951544,
      "step": 750
    },
    {
      "epoch": 0.076,
      "grad_norm": 68.5,
      "grad_norm_var": 306.540625,
      "learning_rate": 0.0001,
      "loss": 10.2479,
      "loss/crossentropy": 2.2831736013293265,
      "loss/hidden": 4.62734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3329113606363535,
      "step": 760
    },
    {
      "epoch": 0.077,
      "grad_norm": 88.5,
      "grad_norm_var": 111.57473958333334,
      "learning_rate": 0.0001,
      "loss": 10.2161,
      "loss/crossentropy": 2.3853780582547186,
      "loss/hidden": 4.541015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.31959532871842383,
      "step": 770
    },
    {
      "epoch": 0.078,
      "grad_norm": 80.5,
      "grad_norm_var": 110.65729166666667,
      "learning_rate": 0.0001,
      "loss": 10.2076,
      "loss/crossentropy": 2.3982744574546815,
      "loss/hidden": 4.55859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3542841043323278,
      "step": 780
    },
    {
      "epoch": 0.079,
      "grad_norm": 66.0,
      "grad_norm_var": 275.6322916666667,
      "learning_rate": 0.0001,
      "loss": 10.1697,
      "loss/crossentropy": 2.4292824655771255,
      "loss/hidden": 4.632421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.36711033545434474,
      "step": 790
    },
    {
      "epoch": 0.08,
      "grad_norm": 57.25,
      "grad_norm_var": 290.9291666666667,
      "learning_rate": 0.0001,
      "loss": 10.2176,
      "loss/crossentropy": 2.380542576313019,
      "loss/hidden": 4.509375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3368827097117901,
      "step": 800
    },
    {
      "epoch": 0.081,
      "grad_norm": 60.75,
      "grad_norm_var": 52.67916666666667,
      "learning_rate": 0.0001,
      "loss": 10.2311,
      "loss/crossentropy": 2.4212940514087675,
      "loss/hidden": 4.55546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.355662290379405,
      "step": 810
    },
    {
      "epoch": 0.082,
      "grad_norm": 60.75,
      "grad_norm_var": 65.81666666666666,
      "learning_rate": 0.0001,
      "loss": 10.1866,
      "loss/crossentropy": 2.4809795886278154,
      "loss/hidden": 4.49140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3553234666585922,
      "step": 820
    },
    {
      "epoch": 0.083,
      "grad_norm": 56.5,
      "grad_norm_var": 98.2875,
      "learning_rate": 0.0001,
      "loss": 9.9805,
      "loss/crossentropy": 2.306653854250908,
      "loss/hidden": 4.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3146494958549738,
      "step": 830
    },
    {
      "epoch": 0.084,
      "grad_norm": 68.0,
      "grad_norm_var": 38.51015625,
      "learning_rate": 0.0001,
      "loss": 10.1087,
      "loss/crossentropy": 2.250006601214409,
      "loss/hidden": 4.422265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.30200174674391744,
      "step": 840
    },
    {
      "epoch": 0.085,
      "grad_norm": 70.5,
      "grad_norm_var": 43.483072916666664,
      "learning_rate": 0.0001,
      "loss": 10.0526,
      "loss/crossentropy": 2.211633677780628,
      "loss/hidden": 4.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.31178686060011385,
      "step": 850
    },
    {
      "epoch": 0.086,
      "grad_norm": 61.0,
      "grad_norm_var": 41.545572916666664,
      "learning_rate": 0.0001,
      "loss": 10.1915,
      "loss/crossentropy": 2.5281356513500213,
      "loss/hidden": 4.389453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.34625968635082244,
      "step": 860
    },
    {
      "epoch": 0.087,
      "grad_norm": 72.5,
      "grad_norm_var": 54.475,
      "learning_rate": 0.0001,
      "loss": 10.0007,
      "loss/crossentropy": 2.4020907685160635,
      "loss/hidden": 4.326171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.32252500094473363,
      "step": 870
    },
    {
      "epoch": 0.088,
      "grad_norm": 142.0,
      "grad_norm_var": 499.1375,
      "learning_rate": 0.0001,
      "loss": 9.99,
      "loss/crossentropy": 2.384984764456749,
      "loss/hidden": 4.38515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3189360786229372,
      "step": 880
    },
    {
      "epoch": 0.089,
      "grad_norm": 57.75,
      "grad_norm_var": 527.23515625,
      "learning_rate": 0.0001,
      "loss": 9.9879,
      "loss/crossentropy": 2.3401281625032424,
      "loss/hidden": 4.46328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3382201848551631,
      "step": 890
    },
    {
      "epoch": 0.09,
      "grad_norm": 71.5,
      "grad_norm_var": 95.97265625,
      "learning_rate": 0.0001,
      "loss": 9.9352,
      "loss/crossentropy": 2.3969784706830977,
      "loss/hidden": 4.384375,
      "loss/jsd": 0.0,
      "loss/logits": 0.336395762488246,
      "step": 900
    },
    {
      "epoch": 0.091,
      "grad_norm": 79.0,
      "grad_norm_var": 144.96666666666667,
      "learning_rate": 0.0001,
      "loss": 10.149,
      "loss/crossentropy": 2.4599110893905163,
      "loss/hidden": 4.30703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3240171395242214,
      "step": 910
    },
    {
      "epoch": 0.092,
      "grad_norm": 65.5,
      "grad_norm_var": 119.2375,
      "learning_rate": 0.0001,
      "loss": 9.9634,
      "loss/crossentropy": 2.4210876494646074,
      "loss/hidden": 4.30390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.32166178375482557,
      "step": 920
    },
    {
      "epoch": 0.093,
      "grad_norm": 63.0,
      "grad_norm_var": 41.47083333333333,
      "learning_rate": 0.0001,
      "loss": 9.744,
      "loss/crossentropy": 2.2256636448204516,
      "loss/hidden": 4.284765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.29795306362211704,
      "step": 930
    },
    {
      "epoch": 0.094,
      "grad_norm": 53.5,
      "grad_norm_var": 192.55807291666667,
      "learning_rate": 0.0001,
      "loss": 9.8636,
      "loss/crossentropy": 2.297808923572302,
      "loss/hidden": 4.31640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.30742434673011304,
      "step": 940
    },
    {
      "epoch": 0.095,
      "grad_norm": 61.0,
      "grad_norm_var": 81.95729166666666,
      "learning_rate": 0.0001,
      "loss": 9.798,
      "loss/crossentropy": 2.3219059616327287,
      "loss/hidden": 4.211328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.30037002861499784,
      "step": 950
    },
    {
      "epoch": 0.096,
      "grad_norm": 56.75,
      "grad_norm_var": 61.55807291666667,
      "learning_rate": 0.0001,
      "loss": 9.7449,
      "loss/crossentropy": 2.3104363679885864,
      "loss/hidden": 4.388671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.327311984449625,
      "step": 960
    },
    {
      "epoch": 0.097,
      "grad_norm": 60.0,
      "grad_norm_var": 56.18932291666667,
      "learning_rate": 0.0001,
      "loss": 9.9668,
      "loss/crossentropy": 2.308886554837227,
      "loss/hidden": 4.407421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3183224782347679,
      "step": 970
    },
    {
      "epoch": 0.098,
      "grad_norm": 66.5,
      "grad_norm_var": 42.05416666666667,
      "learning_rate": 0.0001,
      "loss": 9.7807,
      "loss/crossentropy": 2.3363482102751734,
      "loss/hidden": 4.2921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3384779039770365,
      "step": 980
    },
    {
      "epoch": 0.099,
      "grad_norm": 57.25,
      "grad_norm_var": 56.891666666666666,
      "learning_rate": 0.0001,
      "loss": 9.7501,
      "loss/crossentropy": 2.1767295479774473,
      "loss/hidden": 4.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.31410733237862587,
      "step": 990
    },
    {
      "epoch": 0.1,
      "grad_norm": 50.25,
      "grad_norm_var": 75.85598958333334,
      "learning_rate": 0.0001,
      "loss": 9.9273,
      "loss/crossentropy": 2.505411845445633,
      "loss/hidden": 4.36015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.33212706074118614,
      "step": 1000
    },
    {
      "epoch": 0.101,
      "grad_norm": 77.5,
      "grad_norm_var": 196.21848958333334,
      "learning_rate": 0.0001,
      "loss": 9.9237,
      "loss/crossentropy": 2.3281257838010787,
      "loss/hidden": 4.35546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.32293859515339135,
      "step": 1010
    },
    {
      "epoch": 0.102,
      "grad_norm": 63.25,
      "grad_norm_var": 167.42395833333333,
      "learning_rate": 0.0001,
      "loss": 9.7592,
      "loss/crossentropy": 2.3165650010108947,
      "loss/hidden": 4.32890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.31759811006486416,
      "step": 1020
    },
    {
      "epoch": 0.103,
      "grad_norm": 60.0,
      "grad_norm_var": 153.80833333333334,
      "learning_rate": 0.0001,
      "loss": 9.7366,
      "loss/crossentropy": 2.3203016728162766,
      "loss/hidden": 4.28515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.31944827549159527,
      "step": 1030
    },
    {
      "epoch": 0.104,
      "grad_norm": 66.5,
      "grad_norm_var": 3319.3958333333335,
      "learning_rate": 0.0001,
      "loss": 10.0035,
      "loss/crossentropy": 2.4188640087842943,
      "loss/hidden": 4.38828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3581279247999191,
      "step": 1040
    },
    {
      "epoch": 0.105,
      "grad_norm": 60.25,
      "grad_norm_var": 3338.31640625,
      "learning_rate": 0.0001,
      "loss": 9.6837,
      "loss/crossentropy": 2.2860016629099844,
      "loss/hidden": 4.325390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.318701284006238,
      "step": 1050
    },
    {
      "epoch": 0.106,
      "grad_norm": 74.0,
      "grad_norm_var": 112.4,
      "learning_rate": 0.0001,
      "loss": 9.517,
      "loss/crossentropy": 2.4143033266067504,
      "loss/hidden": 4.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3072842717170715,
      "step": 1060
    },
    {
      "epoch": 0.107,
      "grad_norm": 71.5,
      "grad_norm_var": 68.60598958333334,
      "learning_rate": 0.0001,
      "loss": 9.8549,
      "loss/crossentropy": 2.351083371043205,
      "loss/hidden": 4.398046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.33429058492183683,
      "step": 1070
    },
    {
      "epoch": 0.108,
      "grad_norm": 53.25,
      "grad_norm_var": 43.83229166666667,
      "learning_rate": 0.0001,
      "loss": 9.7738,
      "loss/crossentropy": 2.4011227190494537,
      "loss/hidden": 4.29453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3128178097307682,
      "step": 1080
    },
    {
      "epoch": 0.109,
      "grad_norm": 72.0,
      "grad_norm_var": 34.82890625,
      "learning_rate": 0.0001,
      "loss": 9.7432,
      "loss/crossentropy": 2.310031126439571,
      "loss/hidden": 4.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3273486144840717,
      "step": 1090
    },
    {
      "epoch": 0.11,
      "grad_norm": 66.5,
      "grad_norm_var": 111.64895833333334,
      "learning_rate": 0.0001,
      "loss": 9.6743,
      "loss/crossentropy": 2.3055127263069153,
      "loss/hidden": 4.21796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.32233874313533306,
      "step": 1100
    },
    {
      "epoch": 0.111,
      "grad_norm": 51.5,
      "grad_norm_var": 46.70729166666667,
      "learning_rate": 0.0001,
      "loss": 9.8026,
      "loss/crossentropy": 2.314373381435871,
      "loss/hidden": 4.256640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3083756107836962,
      "step": 1110
    },
    {
      "epoch": 0.112,
      "grad_norm": 57.75,
      "grad_norm_var": 7292.4375,
      "learning_rate": 0.0001,
      "loss": 9.7291,
      "loss/crossentropy": 2.5138203650712967,
      "loss/hidden": 4.19921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.30809955932199956,
      "step": 1120
    },
    {
      "epoch": 0.113,
      "grad_norm": 56.5,
      "grad_norm_var": 29.190625,
      "learning_rate": 0.0001,
      "loss": 9.6823,
      "loss/crossentropy": 2.2719234466552733,
      "loss/hidden": 4.294140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3143883816897869,
      "step": 1130
    },
    {
      "epoch": 0.114,
      "grad_norm": 60.5,
      "grad_norm_var": 45.925,
      "learning_rate": 0.0001,
      "loss": 9.7564,
      "loss/crossentropy": 2.4254489660263063,
      "loss/hidden": 4.261328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3154076419770718,
      "step": 1140
    },
    {
      "epoch": 0.115,
      "grad_norm": 56.0,
      "grad_norm_var": 71.74583333333334,
      "learning_rate": 0.0001,
      "loss": 9.7001,
      "loss/crossentropy": 2.28252642005682,
      "loss/hidden": 4.323046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3203336976468563,
      "step": 1150
    },
    {
      "epoch": 0.116,
      "grad_norm": 67.0,
      "grad_norm_var": 46.040625,
      "learning_rate": 0.0001,
      "loss": 9.7436,
      "loss/crossentropy": 2.391976150870323,
      "loss/hidden": 4.225390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.31455044373869895,
      "step": 1160
    },
    {
      "epoch": 0.117,
      "grad_norm": 46.0,
      "grad_norm_var": 47.06640625,
      "learning_rate": 0.0001,
      "loss": 9.5622,
      "loss/crossentropy": 2.3361207604408265,
      "loss/hidden": 4.19296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.30060703232884406,
      "step": 1170
    },
    {
      "epoch": 0.118,
      "grad_norm": 56.25,
      "grad_norm_var": 49.264322916666664,
      "learning_rate": 0.0001,
      "loss": 9.6834,
      "loss/crossentropy": 2.297483670711517,
      "loss/hidden": 4.2890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2907493541017175,
      "step": 1180
    },
    {
      "epoch": 0.119,
      "grad_norm": 52.5,
      "grad_norm_var": 12.27890625,
      "learning_rate": 0.0001,
      "loss": 9.6207,
      "loss/crossentropy": 2.2364058643579483,
      "loss/hidden": 4.277734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3097097765654325,
      "step": 1190
    },
    {
      "epoch": 0.12,
      "grad_norm": 68.5,
      "grad_norm_var": 35.055989583333336,
      "learning_rate": 0.0001,
      "loss": 9.6018,
      "loss/crossentropy": 2.2412969201803206,
      "loss/hidden": 4.287109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.31851550191640854,
      "step": 1200
    },
    {
      "epoch": 0.121,
      "grad_norm": 60.5,
      "grad_norm_var": 25.774739583333332,
      "learning_rate": 0.0001,
      "loss": 9.6979,
      "loss/crossentropy": 2.3062032952904703,
      "loss/hidden": 4.258984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3124631106853485,
      "step": 1210
    },
    {
      "epoch": 0.122,
      "grad_norm": 59.25,
      "grad_norm_var": 20.026822916666667,
      "learning_rate": 0.0001,
      "loss": 9.7129,
      "loss/crossentropy": 2.4036868065595627,
      "loss/hidden": 4.20234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.31101155243813994,
      "step": 1220
    },
    {
      "epoch": 0.123,
      "grad_norm": 53.25,
      "grad_norm_var": 75.30833333333334,
      "learning_rate": 0.0001,
      "loss": 9.7047,
      "loss/crossentropy": 2.3730016142129897,
      "loss/hidden": 4.193359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3105484452098608,
      "step": 1230
    },
    {
      "epoch": 0.124,
      "grad_norm": 62.25,
      "grad_norm_var": 33.27682291666667,
      "learning_rate": 0.0001,
      "loss": 9.6313,
      "loss/crossentropy": 2.2872567594051363,
      "loss/hidden": 4.319140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3244694545865059,
      "step": 1240
    },
    {
      "epoch": 0.125,
      "grad_norm": 61.25,
      "grad_norm_var": 25.673958333333335,
      "learning_rate": 0.0001,
      "loss": 9.6217,
      "loss/crossentropy": 2.3013710603117943,
      "loss/hidden": 4.29140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.32178852558135984,
      "step": 1250
    },
    {
      "epoch": 0.126,
      "grad_norm": 49.5,
      "grad_norm_var": 46.97265625,
      "learning_rate": 0.0001,
      "loss": 9.6151,
      "loss/crossentropy": 2.2743802405893803,
      "loss/hidden": 4.19921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3104738780297339,
      "step": 1260
    },
    {
      "epoch": 0.127,
      "grad_norm": 52.25,
      "grad_norm_var": 318.94557291666666,
      "learning_rate": 0.0001,
      "loss": 9.635,
      "loss/crossentropy": 2.2751111879944803,
      "loss/hidden": 4.17578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2947248375043273,
      "step": 1270
    },
    {
      "epoch": 0.128,
      "grad_norm": 59.5,
      "grad_norm_var": 201.26848958333332,
      "learning_rate": 0.0001,
      "loss": 9.605,
      "loss/crossentropy": 2.3590754181146623,
      "loss/hidden": 4.116015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.29773430675268175,
      "step": 1280
    },
    {
      "epoch": 0.129,
      "grad_norm": 50.0,
      "grad_norm_var": 25.795833333333334,
      "learning_rate": 0.0001,
      "loss": 9.4314,
      "loss/crossentropy": 2.165515697002411,
      "loss/hidden": 4.148046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2729496695101261,
      "step": 1290
    },
    {
      "epoch": 0.13,
      "grad_norm": 51.5,
      "grad_norm_var": 65.69557291666666,
      "learning_rate": 0.0001,
      "loss": 9.4579,
      "loss/crossentropy": 2.425456903874874,
      "loss/hidden": 4.1140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3002984166145325,
      "step": 1300
    },
    {
      "epoch": 0.131,
      "grad_norm": 55.75,
      "grad_norm_var": 74.63515625,
      "learning_rate": 0.0001,
      "loss": 9.562,
      "loss/crossentropy": 2.3212677478790282,
      "loss/hidden": 4.209765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.28645528480410576,
      "step": 1310
    },
    {
      "epoch": 0.132,
      "grad_norm": 44.75,
      "grad_norm_var": 39.139322916666664,
      "learning_rate": 0.0001,
      "loss": 9.305,
      "loss/crossentropy": 2.2911602184176445,
      "loss/hidden": 4.133984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.28404638059437276,
      "step": 1320
    },
    {
      "epoch": 0.133,
      "grad_norm": 52.25,
      "grad_norm_var": 76.19583333333334,
      "learning_rate": 0.0001,
      "loss": 9.3122,
      "loss/crossentropy": 2.3109163105487824,
      "loss/hidden": 4.137109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2864396806806326,
      "step": 1330
    },
    {
      "epoch": 0.134,
      "grad_norm": 47.0,
      "grad_norm_var": 41.66015625,
      "learning_rate": 0.0001,
      "loss": 9.4629,
      "loss/crossentropy": 2.353537403047085,
      "loss/hidden": 4.08203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2971150416880846,
      "step": 1340
    },
    {
      "epoch": 0.135,
      "grad_norm": 46.25,
      "grad_norm_var": 45.31848958333333,
      "learning_rate": 0.0001,
      "loss": 9.365,
      "loss/crossentropy": 2.3774181246757506,
      "loss/hidden": 4.09296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2799839396029711,
      "step": 1350
    },
    {
      "epoch": 0.136,
      "grad_norm": 51.0,
      "grad_norm_var": 17.93515625,
      "learning_rate": 0.0001,
      "loss": 9.3498,
      "loss/crossentropy": 2.246833881735802,
      "loss/hidden": 4.18828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2903384942561388,
      "step": 1360
    },
    {
      "epoch": 0.137,
      "grad_norm": 51.25,
      "grad_norm_var": 12.420833333333333,
      "learning_rate": 0.0001,
      "loss": 9.4976,
      "loss/crossentropy": 2.453240838646889,
      "loss/hidden": 4.173828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3144164770841599,
      "step": 1370
    },
    {
      "epoch": 0.138,
      "grad_norm": 70.0,
      "grad_norm_var": 2011.0322916666667,
      "learning_rate": 0.0001,
      "loss": 9.5884,
      "loss/crossentropy": 2.174116183817387,
      "loss/hidden": 4.24921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2923248626291752,
      "step": 1380
    },
    {
      "epoch": 0.139,
      "grad_norm": 53.75,
      "grad_norm_var": 1988.8833333333334,
      "learning_rate": 0.0001,
      "loss": 9.5249,
      "loss/crossentropy": 2.3638354018330574,
      "loss/hidden": 4.184765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.3066251628100872,
      "step": 1390
    },
    {
      "epoch": 0.14,
      "grad_norm": 55.5,
      "grad_norm_var": 22.779166666666665,
      "learning_rate": 0.0001,
      "loss": 9.3528,
      "loss/crossentropy": 2.4166768550872804,
      "loss/hidden": 4.123828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.29636494982987643,
      "step": 1400
    },
    {
      "epoch": 0.141,
      "grad_norm": 60.5,
      "grad_norm_var": 66.59348958333334,
      "learning_rate": 0.0001,
      "loss": 9.5339,
      "loss/crossentropy": 2.3475931867957116,
      "loss/hidden": 4.1953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.30608872696757317,
      "step": 1410
    },
    {
      "epoch": 0.142,
      "grad_norm": 51.25,
      "grad_norm_var": 62.49140625,
      "learning_rate": 0.0001,
      "loss": 9.3342,
      "loss/crossentropy": 2.1785849004983904,
      "loss/hidden": 4.161328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27641028352081776,
      "step": 1420
    },
    {
      "epoch": 0.143,
      "grad_norm": 54.25,
      "grad_norm_var": 30.154166666666665,
      "learning_rate": 0.0001,
      "loss": 9.3898,
      "loss/crossentropy": 2.3990818440914152,
      "loss/hidden": 4.18359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2944341886788607,
      "step": 1430
    },
    {
      "epoch": 0.144,
      "grad_norm": 51.75,
      "grad_norm_var": 38.93932291666667,
      "learning_rate": 0.0001,
      "loss": 9.4628,
      "loss/crossentropy": 2.4946817860007284,
      "loss/hidden": 4.198828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.31867978498339655,
      "step": 1440
    },
    {
      "epoch": 0.145,
      "grad_norm": 53.75,
      "grad_norm_var": 33.9,
      "learning_rate": 0.0001,
      "loss": 9.3416,
      "loss/crossentropy": 2.2067521095275877,
      "loss/hidden": 4.235546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2976540043950081,
      "step": 1450
    },
    {
      "epoch": 0.146,
      "grad_norm": 60.75,
      "grad_norm_var": 142.08229166666666,
      "learning_rate": 0.0001,
      "loss": 9.4716,
      "loss/crossentropy": 2.4361192852258684,
      "loss/hidden": 4.1140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2877715673297644,
      "step": 1460
    },
    {
      "epoch": 0.147,
      "grad_norm": 58.75,
      "grad_norm_var": 44.35,
      "learning_rate": 0.0001,
      "loss": 9.4006,
      "loss/crossentropy": 2.239429622516036,
      "loss/hidden": 4.026171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27844256814569235,
      "step": 1470
    },
    {
      "epoch": 0.148,
      "grad_norm": 45.0,
      "grad_norm_var": 33.95390625,
      "learning_rate": 0.0001,
      "loss": 9.3993,
      "loss/crossentropy": 2.0759536787867545,
      "loss/hidden": 4.068359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2688772227615118,
      "step": 1480
    },
    {
      "epoch": 0.149,
      "grad_norm": 51.75,
      "grad_norm_var": 25.795833333333334,
      "learning_rate": 0.0001,
      "loss": 9.3786,
      "loss/crossentropy": 2.286362998187542,
      "loss/hidden": 4.15078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2942257083952427,
      "step": 1490
    },
    {
      "epoch": 0.15,
      "grad_norm": 46.75,
      "grad_norm_var": 20.520833333333332,
      "learning_rate": 0.0001,
      "loss": 9.2903,
      "loss/crossentropy": 2.312733788788319,
      "loss/hidden": 3.971484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2691910218447447,
      "step": 1500
    },
    {
      "epoch": 0.151,
      "grad_norm": 50.25,
      "grad_norm_var": 28.290625,
      "learning_rate": 0.0001,
      "loss": 9.3076,
      "loss/crossentropy": 2.2467628076672552,
      "loss/hidden": 4.105078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2887777745723724,
      "step": 1510
    },
    {
      "epoch": 0.152,
      "grad_norm": 63.5,
      "grad_norm_var": 33.73098958333333,
      "learning_rate": 0.0001,
      "loss": 9.4203,
      "loss/crossentropy": 2.372379180788994,
      "loss/hidden": 4.07578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3087839350104332,
      "step": 1520
    },
    {
      "epoch": 0.153,
      "grad_norm": 45.5,
      "grad_norm_var": 40.108333333333334,
      "learning_rate": 0.0001,
      "loss": 9.3215,
      "loss/crossentropy": 2.3452367037534714,
      "loss/hidden": 4.210546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3159611392766237,
      "step": 1530
    },
    {
      "epoch": 0.154,
      "grad_norm": 58.25,
      "grad_norm_var": 27.539322916666666,
      "learning_rate": 0.0001,
      "loss": 9.3755,
      "loss/crossentropy": 2.3029753446578978,
      "loss/hidden": 3.999609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2623455457389355,
      "step": 1540
    },
    {
      "epoch": 0.155,
      "grad_norm": 51.75,
      "grad_norm_var": 26.9,
      "learning_rate": 0.0001,
      "loss": 9.3578,
      "loss/crossentropy": 2.3988554388284684,
      "loss/hidden": 4.08828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2846154376864433,
      "step": 1550
    },
    {
      "epoch": 0.156,
      "grad_norm": 91.0,
      "grad_norm_var": 1307.8372395833333,
      "learning_rate": 0.0001,
      "loss": 9.432,
      "loss/crossentropy": 2.343544365465641,
      "loss/hidden": 4.021875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2907770898193121,
      "step": 1560
    },
    {
      "epoch": 0.157,
      "grad_norm": 52.0,
      "grad_norm_var": 170.62890625,
      "learning_rate": 0.0001,
      "loss": 9.3432,
      "loss/crossentropy": 2.173108433187008,
      "loss/hidden": 4.10859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.28518917988985776,
      "step": 1570
    },
    {
      "epoch": 0.158,
      "grad_norm": 42.0,
      "grad_norm_var": 47.56015625,
      "learning_rate": 0.0001,
      "loss": 9.367,
      "loss/crossentropy": 2.2230691239237785,
      "loss/hidden": 4.23515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.30039387457072736,
      "step": 1580
    },
    {
      "epoch": 0.159,
      "grad_norm": 72.0,
      "grad_norm_var": 1.226104970407838e+18,
      "learning_rate": 0.0001,
      "loss": 9.3564,
      "loss/crossentropy": 2.263391149044037,
      "loss/hidden": 4.10625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2923804897814989,
      "step": 1590
    },
    {
      "epoch": 0.16,
      "grad_norm": 52.5,
      "grad_norm_var": 1.2261049681378806e+18,
      "learning_rate": 0.0001,
      "loss": 9.4959,
      "loss/crossentropy": 2.113241518288851,
      "loss/hidden": 4.087109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2759646028280258,
      "step": 1600
    },
    {
      "epoch": 0.161,
      "grad_norm": 66.0,
      "grad_norm_var": 734.1489583333333,
      "learning_rate": 0.0001,
      "loss": 9.4743,
      "loss/crossentropy": 2.3895165085792542,
      "loss/hidden": 4.059765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2987998936325312,
      "step": 1610
    },
    {
      "epoch": 0.162,
      "grad_norm": 44.75,
      "grad_norm_var": 50.00182291666667,
      "learning_rate": 0.0001,
      "loss": 9.1919,
      "loss/crossentropy": 2.251766300201416,
      "loss/hidden": 4.04375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2781111396849155,
      "step": 1620
    },
    {
      "epoch": 0.163,
      "grad_norm": 52.75,
      "grad_norm_var": 437.49583333333334,
      "learning_rate": 0.0001,
      "loss": 9.4572,
      "loss/crossentropy": 2.382322034239769,
      "loss/hidden": 4.07734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.31318275928497313,
      "step": 1630
    },
    {
      "epoch": 0.164,
      "grad_norm": 61.0,
      "grad_norm_var": 40.301822916666666,
      "learning_rate": 0.0001,
      "loss": 9.2668,
      "loss/crossentropy": 2.1683703124523164,
      "loss/hidden": 4.07578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.283413190767169,
      "step": 1640
    },
    {
      "epoch": 0.165,
      "grad_norm": 42.75,
      "grad_norm_var": 57.307291666666664,
      "learning_rate": 0.0001,
      "loss": 9.339,
      "loss/crossentropy": 2.3430400043725967,
      "loss/hidden": 4.036328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.287694800645113,
      "step": 1650
    },
    {
      "epoch": 0.166,
      "grad_norm": 46.75,
      "grad_norm_var": 65.52395833333334,
      "learning_rate": 0.0001,
      "loss": 9.3768,
      "loss/crossentropy": 2.2867416352033616,
      "loss/hidden": 4.017578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.29683431759476664,
      "step": 1660
    },
    {
      "epoch": 0.167,
      "grad_norm": 52.5,
      "grad_norm_var": 61.18932291666667,
      "learning_rate": 0.0001,
      "loss": 9.2451,
      "loss/crossentropy": 2.3707614041864873,
      "loss/hidden": 4.06015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.29184688804671166,
      "step": 1670
    },
    {
      "epoch": 0.168,
      "grad_norm": 51.75,
      "grad_norm_var": 20.895572916666666,
      "learning_rate": 0.0001,
      "loss": 9.3601,
      "loss/crossentropy": 2.3268392831087112,
      "loss/hidden": 4.12734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.29570323824882505,
      "step": 1680
    },
    {
      "epoch": 0.169,
      "grad_norm": 44.0,
      "grad_norm_var": 10.290625,
      "learning_rate": 0.0001,
      "loss": 9.4214,
      "loss/crossentropy": 2.324131193757057,
      "loss/hidden": 4.1984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3133995305746794,
      "step": 1690
    },
    {
      "epoch": 0.17,
      "grad_norm": 58.25,
      "grad_norm_var": 19.124739583333334,
      "learning_rate": 0.0001,
      "loss": 9.2465,
      "loss/crossentropy": 2.35849623978138,
      "loss/hidden": 4.00703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2762619823217392,
      "step": 1700
    },
    {
      "epoch": 0.171,
      "grad_norm": 45.75,
      "grad_norm_var": 53.89895833333333,
      "learning_rate": 0.0001,
      "loss": 9.1951,
      "loss/crossentropy": 2.3914038598537446,
      "loss/hidden": 3.9984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2871177852153778,
      "step": 1710
    },
    {
      "epoch": 0.172,
      "grad_norm": 43.25,
      "grad_norm_var": 16.479166666666668,
      "learning_rate": 0.0001,
      "loss": 9.1669,
      "loss/crossentropy": 2.152750685811043,
      "loss/hidden": 4.100390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.28708020225167274,
      "step": 1720
    },
    {
      "epoch": 0.173,
      "grad_norm": 49.25,
      "grad_norm_var": 13.45390625,
      "learning_rate": 0.0001,
      "loss": 9.1015,
      "loss/crossentropy": 2.2946193665266037,
      "loss/hidden": 4.085546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.3062314610928297,
      "step": 1730
    },
    {
      "epoch": 0.174,
      "grad_norm": 46.5,
      "grad_norm_var": 22.473958333333332,
      "learning_rate": 0.0001,
      "loss": 9.1287,
      "loss/crossentropy": 2.1538643553853034,
      "loss/hidden": 3.9421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2666194221004844,
      "step": 1740
    },
    {
      "epoch": 0.175,
      "grad_norm": 47.0,
      "grad_norm_var": 32.62057291666667,
      "learning_rate": 0.0001,
      "loss": 9.411,
      "loss/crossentropy": 2.387891933321953,
      "loss/hidden": 4.14921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.29542505368590355,
      "step": 1750
    },
    {
      "epoch": 0.176,
      "grad_norm": 45.25,
      "grad_norm_var": 26.92265625,
      "learning_rate": 0.0001,
      "loss": 9.2833,
      "loss/crossentropy": 2.3024097591638566,
      "loss/hidden": 4.03984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.29066667445003985,
      "step": 1760
    },
    {
      "epoch": 0.177,
      "grad_norm": 53.5,
      "grad_norm_var": 17.832291666666666,
      "learning_rate": 0.0001,
      "loss": 9.2665,
      "loss/crossentropy": 2.4454205125570296,
      "loss/hidden": 3.955859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2910691563040018,
      "step": 1770
    },
    {
      "epoch": 0.178,
      "grad_norm": 42.25,
      "grad_norm_var": 29.865625,
      "learning_rate": 0.0001,
      "loss": 9.1701,
      "loss/crossentropy": 2.2966391056776048,
      "loss/hidden": 4.027734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2789210833609104,
      "step": 1780
    },
    {
      "epoch": 0.179,
      "grad_norm": 48.25,
      "grad_norm_var": 17.548958333333335,
      "learning_rate": 0.0001,
      "loss": 9.1992,
      "loss/crossentropy": 2.395502945780754,
      "loss/hidden": 3.934765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2776679117232561,
      "step": 1790
    },
    {
      "epoch": 0.18,
      "grad_norm": 40.75,
      "grad_norm_var": 13.282291666666667,
      "learning_rate": 0.0001,
      "loss": 9.1046,
      "loss/crossentropy": 2.22285817861557,
      "loss/hidden": 3.9046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.26667180880904195,
      "step": 1800
    },
    {
      "epoch": 0.181,
      "grad_norm": 36.25,
      "grad_norm_var": 34.90807291666667,
      "learning_rate": 0.0001,
      "loss": 9.3204,
      "loss/crossentropy": 2.3842350512743,
      "loss/hidden": 4.009375,
      "loss/jsd": 0.0,
      "loss/logits": 0.300260554254055,
      "step": 1810
    },
    {
      "epoch": 0.182,
      "grad_norm": 46.75,
      "grad_norm_var": 27.77890625,
      "learning_rate": 0.0001,
      "loss": 9.0943,
      "loss/crossentropy": 2.274762773513794,
      "loss/hidden": 4.01328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.28360783979296683,
      "step": 1820
    },
    {
      "epoch": 0.183,
      "grad_norm": 55.5,
      "grad_norm_var": 27.298958333333335,
      "learning_rate": 0.0001,
      "loss": 9.1699,
      "loss/crossentropy": 2.1643219627439976,
      "loss/hidden": 3.9921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.267458438500762,
      "step": 1830
    },
    {
      "epoch": 0.184,
      "grad_norm": 49.75,
      "grad_norm_var": 43.94583333333333,
      "learning_rate": 0.0001,
      "loss": 9.3022,
      "loss/crossentropy": 2.464679929614067,
      "loss/hidden": 3.9546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.29758369028568266,
      "step": 1840
    },
    {
      "epoch": 0.185,
      "grad_norm": 51.0,
      "grad_norm_var": 37.90807291666667,
      "learning_rate": 0.0001,
      "loss": 9.1863,
      "loss/crossentropy": 2.3199010998010636,
      "loss/hidden": 3.99453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27702242247760295,
      "step": 1850
    },
    {
      "epoch": 0.186,
      "grad_norm": 46.5,
      "grad_norm_var": 40.920833333333334,
      "learning_rate": 0.0001,
      "loss": 9.2872,
      "loss/crossentropy": 2.4041683062911035,
      "loss/hidden": 4.09140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.30005627647042277,
      "step": 1860
    },
    {
      "epoch": 0.187,
      "grad_norm": 39.75,
      "grad_norm_var": 40.723958333333336,
      "learning_rate": 0.0001,
      "loss": 9.1081,
      "loss/crossentropy": 2.273802790045738,
      "loss/hidden": 4.175,
      "loss/jsd": 0.0,
      "loss/logits": 0.3045934235677123,
      "step": 1870
    },
    {
      "epoch": 0.188,
      "grad_norm": 43.25,
      "grad_norm_var": 33.35729166666667,
      "learning_rate": 0.0001,
      "loss": 9.106,
      "loss/crossentropy": 2.3607766672968866,
      "loss/hidden": 3.9765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.28193066976964476,
      "step": 1880
    },
    {
      "epoch": 0.189,
      "grad_norm": 48.25,
      "grad_norm_var": 14.915625,
      "learning_rate": 0.0001,
      "loss": 9.1437,
      "loss/crossentropy": 2.2798361241817475,
      "loss/hidden": 3.9890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2721746701747179,
      "step": 1890
    },
    {
      "epoch": 0.19,
      "grad_norm": 44.0,
      "grad_norm_var": 21.223958333333332,
      "learning_rate": 0.0001,
      "loss": 9.0972,
      "loss/crossentropy": 2.21695294380188,
      "loss/hidden": 4.0015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2832322970032692,
      "step": 1900
    },
    {
      "epoch": 0.191,
      "grad_norm": 39.5,
      "grad_norm_var": 27.808333333333334,
      "learning_rate": 0.0001,
      "loss": 9.1587,
      "loss/crossentropy": 2.1728454776108266,
      "loss/hidden": 3.98828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27170457877218723,
      "step": 1910
    },
    {
      "epoch": 0.192,
      "grad_norm": 41.75,
      "grad_norm_var": 13.315625,
      "learning_rate": 0.0001,
      "loss": 9.1326,
      "loss/crossentropy": 2.154237084835768,
      "loss/hidden": 4.063671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27950075305998323,
      "step": 1920
    },
    {
      "epoch": 0.193,
      "grad_norm": 43.0,
      "grad_norm_var": 25.240625,
      "learning_rate": 0.0001,
      "loss": 9.1013,
      "loss/crossentropy": 2.2507698431611063,
      "loss/hidden": 4.01171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2808088269084692,
      "step": 1930
    },
    {
      "epoch": 0.194,
      "grad_norm": 49.25,
      "grad_norm_var": 22.832291666666666,
      "learning_rate": 0.0001,
      "loss": 9.2429,
      "loss/crossentropy": 2.288056728243828,
      "loss/hidden": 4.1546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.31668607220053674,
      "step": 1940
    },
    {
      "epoch": 0.195,
      "grad_norm": 48.5,
      "grad_norm_var": 58.09557291666667,
      "learning_rate": 0.0001,
      "loss": 9.1742,
      "loss/crossentropy": 2.2107961744070055,
      "loss/hidden": 4.05078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2858551822602749,
      "step": 1950
    },
    {
      "epoch": 0.196,
      "grad_norm": 39.25,
      "grad_norm_var": 49.50390625,
      "learning_rate": 0.0001,
      "loss": 9.1293,
      "loss/crossentropy": 2.224529256671667,
      "loss/hidden": 3.9703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27973891496658326,
      "step": 1960
    },
    {
      "epoch": 0.197,
      "grad_norm": 39.25,
      "grad_norm_var": 13.890625,
      "learning_rate": 0.0001,
      "loss": 9.0689,
      "loss/crossentropy": 2.363737019896507,
      "loss/hidden": 4.027734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2919711694121361,
      "step": 1970
    },
    {
      "epoch": 0.198,
      "grad_norm": 55.75,
      "grad_norm_var": 26.655989583333334,
      "learning_rate": 0.0001,
      "loss": 9.2228,
      "loss/crossentropy": 2.3380469545722007,
      "loss/hidden": 4.0171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.28581551983952524,
      "step": 1980
    },
    {
      "epoch": 0.199,
      "grad_norm": 45.0,
      "grad_norm_var": 27.357291666666665,
      "learning_rate": 0.0001,
      "loss": 9.173,
      "loss/crossentropy": 2.43135461807251,
      "loss/hidden": 3.970703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.28222124874591825,
      "step": 1990
    },
    {
      "epoch": 0.2,
      "grad_norm": 44.5,
      "grad_norm_var": 16.04140625,
      "learning_rate": 0.0001,
      "loss": 9.1554,
      "loss/crossentropy": 2.4415812104940415,
      "loss/hidden": 4.064453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2984179027378559,
      "step": 2000
    },
    {
      "epoch": 0.201,
      "grad_norm": 51.75,
      "grad_norm_var": 18.032291666666666,
      "learning_rate": 0.0001,
      "loss": 9.204,
      "loss/crossentropy": 2.571503698825836,
      "loss/hidden": 4.020703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3065837759524584,
      "step": 2010
    },
    {
      "epoch": 0.202,
      "grad_norm": 42.25,
      "grad_norm_var": 16.573958333333334,
      "learning_rate": 0.0001,
      "loss": 9.1575,
      "loss/crossentropy": 2.2947281152009964,
      "loss/hidden": 3.949609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.28560531958937646,
      "step": 2020
    },
    {
      "epoch": 0.203,
      "grad_norm": 43.0,
      "grad_norm_var": 10.068489583333333,
      "learning_rate": 0.0001,
      "loss": 9.1347,
      "loss/crossentropy": 2.432750529050827,
      "loss/hidden": 3.9953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.28870879150927065,
      "step": 2030
    },
    {
      "epoch": 0.204,
      "grad_norm": 45.0,
      "grad_norm_var": 14.683333333333334,
      "learning_rate": 0.0001,
      "loss": 9.0302,
      "loss/crossentropy": 2.291880601644516,
      "loss/hidden": 3.852734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27138952538371086,
      "step": 2040
    },
    {
      "epoch": 0.205,
      "grad_norm": 54.5,
      "grad_norm_var": 24.0625,
      "learning_rate": 0.0001,
      "loss": 9.1623,
      "loss/crossentropy": 2.1033009082078933,
      "loss/hidden": 4.074609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.29145455472171305,
      "step": 2050
    },
    {
      "epoch": 0.206,
      "grad_norm": 42.25,
      "grad_norm_var": 29.780989583333334,
      "learning_rate": 0.0001,
      "loss": 9.1409,
      "loss/crossentropy": 2.3487906470894813,
      "loss/hidden": 3.9140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2789519714191556,
      "step": 2060
    },
    {
      "epoch": 0.207,
      "grad_norm": 42.0,
      "grad_norm_var": 15.395572916666667,
      "learning_rate": 0.0001,
      "loss": 9.0027,
      "loss/crossentropy": 2.210711918771267,
      "loss/hidden": 3.975,
      "loss/jsd": 0.0,
      "loss/logits": 0.2679262701421976,
      "step": 2070
    },
    {
      "epoch": 0.208,
      "grad_norm": 43.5,
      "grad_norm_var": 12.7125,
      "learning_rate": 0.0001,
      "loss": 9.0306,
      "loss/crossentropy": 2.1945893600583077,
      "loss/hidden": 3.948046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.28115708455443383,
      "step": 2080
    },
    {
      "epoch": 0.209,
      "grad_norm": 47.0,
      "grad_norm_var": 28.683072916666667,
      "learning_rate": 0.0001,
      "loss": 9.1344,
      "loss/crossentropy": 2.237662248313427,
      "loss/hidden": 4.035546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2781697390601039,
      "step": 2090
    },
    {
      "epoch": 0.21,
      "grad_norm": 42.0,
      "grad_norm_var": 15.390625,
      "learning_rate": 0.0001,
      "loss": 9.0004,
      "loss/crossentropy": 2.269840542972088,
      "loss/hidden": 3.9765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.29009242728352547,
      "step": 2100
    },
    {
      "epoch": 0.211,
      "grad_norm": 39.75,
      "grad_norm_var": 8.10390625,
      "learning_rate": 0.0001,
      "loss": 9.0135,
      "loss/crossentropy": 2.3315619856119154,
      "loss/hidden": 4.028515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.27684418186545373,
      "step": 2110
    },
    {
      "epoch": 0.212,
      "grad_norm": 45.75,
      "grad_norm_var": 115.615625,
      "learning_rate": 0.0001,
      "loss": 9.2092,
      "loss/crossentropy": 2.4081921339035035,
      "loss/hidden": 4.01171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.29490497298538687,
      "step": 2120
    },
    {
      "epoch": 0.213,
      "grad_norm": 43.0,
      "grad_norm_var": 95.76432291666667,
      "learning_rate": 0.0001,
      "loss": 9.1058,
      "loss/crossentropy": 2.393023744225502,
      "loss/hidden": 3.94921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2818208742886782,
      "step": 2130
    },
    {
      "epoch": 0.214,
      "grad_norm": 39.5,
      "grad_norm_var": 13.873958333333333,
      "learning_rate": 0.0001,
      "loss": 9.0997,
      "loss/crossentropy": 2.1606352396309374,
      "loss/hidden": 3.843359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25883881878107784,
      "step": 2140
    },
    {
      "epoch": 0.215,
      "grad_norm": 44.0,
      "grad_norm_var": 22.845572916666665,
      "learning_rate": 0.0001,
      "loss": 9.1464,
      "loss/crossentropy": 2.179043120145798,
      "loss/hidden": 4.056640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2761132620275021,
      "step": 2150
    },
    {
      "epoch": 0.216,
      "grad_norm": 46.5,
      "grad_norm_var": 28.957291666666666,
      "learning_rate": 0.0001,
      "loss": 9.1057,
      "loss/crossentropy": 2.301611530780792,
      "loss/hidden": 3.959375,
      "loss/jsd": 0.0,
      "loss/logits": 0.29443784058094025,
      "step": 2160
    },
    {
      "epoch": 0.217,
      "grad_norm": 46.25,
      "grad_norm_var": 36.708333333333336,
      "learning_rate": 0.0001,
      "loss": 9.0952,
      "loss/crossentropy": 2.3569841012358665,
      "loss/hidden": 3.994921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.28191804718226193,
      "step": 2170
    },
    {
      "epoch": 0.218,
      "grad_norm": 81.5,
      "grad_norm_var": 117.24895833333333,
      "learning_rate": 0.0001,
      "loss": 9.0595,
      "loss/crossentropy": 2.4225870154798033,
      "loss/hidden": 3.953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2870227605104446,
      "step": 2180
    },
    {
      "epoch": 0.219,
      "grad_norm": 36.75,
      "grad_norm_var": 173.97265625,
      "learning_rate": 0.0001,
      "loss": 9.1612,
      "loss/crossentropy": 2.34947164952755,
      "loss/hidden": 4.0546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2929512483999133,
      "step": 2190
    },
    {
      "epoch": 0.22,
      "grad_norm": 41.75,
      "grad_norm_var": 31.648958333333333,
      "learning_rate": 0.0001,
      "loss": 9.101,
      "loss/crossentropy": 2.218617644906044,
      "loss/hidden": 4.029296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2752823047339916,
      "step": 2200
    },
    {
      "epoch": 0.221,
      "grad_norm": 47.0,
      "grad_norm_var": 10.50390625,
      "learning_rate": 0.0001,
      "loss": 9.0496,
      "loss/crossentropy": 2.4772594451904295,
      "loss/hidden": 3.96953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.28055914528667925,
      "step": 2210
    },
    {
      "epoch": 0.222,
      "grad_norm": 49.5,
      "grad_norm_var": 442.3,
      "learning_rate": 0.0001,
      "loss": 8.9647,
      "loss/crossentropy": 2.44835202395916,
      "loss/hidden": 3.905859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27223448157310487,
      "step": 2220
    },
    {
      "epoch": 0.223,
      "grad_norm": 35.25,
      "grad_norm_var": 577.0458333333333,
      "learning_rate": 0.0001,
      "loss": 9.1287,
      "loss/crossentropy": 2.2784701570868493,
      "loss/hidden": 3.98671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.28111674822866917,
      "step": 2230
    },
    {
      "epoch": 0.224,
      "grad_norm": 43.75,
      "grad_norm_var": 212.74348958333334,
      "learning_rate": 0.0001,
      "loss": 9.1339,
      "loss/crossentropy": 2.2536803498864173,
      "loss/hidden": 3.934375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27366876490414144,
      "step": 2240
    },
    {
      "epoch": 0.225,
      "grad_norm": 42.5,
      "grad_norm_var": 30.143489583333334,
      "learning_rate": 0.0001,
      "loss": 9.0536,
      "loss/crossentropy": 2.300386372208595,
      "loss/hidden": 3.937890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2872367199510336,
      "step": 2250
    },
    {
      "epoch": 0.226,
      "grad_norm": 40.75,
      "grad_norm_var": 19.65390625,
      "learning_rate": 0.0001,
      "loss": 9.0113,
      "loss/crossentropy": 2.318666061758995,
      "loss/hidden": 3.8796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.259370943903923,
      "step": 2260
    },
    {
      "epoch": 0.227,
      "grad_norm": 43.75,
      "grad_norm_var": 12.46015625,
      "learning_rate": 0.0001,
      "loss": 8.9343,
      "loss/crossentropy": 2.169833867251873,
      "loss/hidden": 3.8984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26568643413484094,
      "step": 2270
    },
    {
      "epoch": 0.228,
      "grad_norm": 38.75,
      "grad_norm_var": 11.832291666666666,
      "learning_rate": 0.0001,
      "loss": 9.0042,
      "loss/crossentropy": 2.1532950207591055,
      "loss/hidden": 3.987890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2669289981946349,
      "step": 2280
    },
    {
      "epoch": 0.229,
      "grad_norm": 46.5,
      "grad_norm_var": 13.432291666666666,
      "learning_rate": 0.0001,
      "loss": 9.1282,
      "loss/crossentropy": 2.0724870592355726,
      "loss/hidden": 4.09765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.27939337231218814,
      "step": 2290
    },
    {
      "epoch": 0.23,
      "grad_norm": 35.75,
      "grad_norm_var": 223.3875,
      "learning_rate": 0.0001,
      "loss": 9.1656,
      "loss/crossentropy": 2.4021882474422456,
      "loss/hidden": 3.941796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.28748833425343034,
      "step": 2300
    },
    {
      "epoch": 0.231,
      "grad_norm": 40.5,
      "grad_norm_var": 264.82395833333334,
      "learning_rate": 0.0001,
      "loss": 8.9421,
      "loss/crossentropy": 2.201041653752327,
      "loss/hidden": 4.012109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.29431225806474687,
      "step": 2310
    },
    {
      "epoch": 0.232,
      "grad_norm": 37.25,
      "grad_norm_var": 14.939322916666667,
      "learning_rate": 0.0001,
      "loss": 9.0761,
      "loss/crossentropy": 2.156717260926962,
      "loss/hidden": 4.16484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27824588380753995,
      "step": 2320
    },
    {
      "epoch": 0.233,
      "grad_norm": 44.0,
      "grad_norm_var": 66.07890625,
      "learning_rate": 0.0001,
      "loss": 9.1021,
      "loss/crossentropy": 2.3962995454669,
      "loss/hidden": 3.849609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27611064203083513,
      "step": 2330
    },
    {
      "epoch": 0.234,
      "grad_norm": 37.5,
      "grad_norm_var": 2.5671221192004644e+18,
      "learning_rate": 0.0001,
      "loss": 9.1022,
      "loss/crossentropy": 2.302082321047783,
      "loss/hidden": 3.980859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2941384054720402,
      "step": 2340
    },
    {
      "epoch": 0.235,
      "grad_norm": 42.75,
      "grad_norm_var": 49.62473958333333,
      "learning_rate": 0.0001,
      "loss": 8.7886,
      "loss/crossentropy": 2.239869697391987,
      "loss/hidden": 3.977734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25684802830219267,
      "step": 2350
    },
    {
      "epoch": 0.236,
      "grad_norm": 39.5,
      "grad_norm_var": 10.540625,
      "learning_rate": 0.0001,
      "loss": 8.8748,
      "loss/crossentropy": 2.2847611531615257,
      "loss/hidden": 4.076171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2846489936113358,
      "step": 2360
    },
    {
      "epoch": 0.237,
      "grad_norm": 44.25,
      "grad_norm_var": 33.15416666666667,
      "learning_rate": 0.0001,
      "loss": 8.8811,
      "loss/crossentropy": 2.209371344745159,
      "loss/hidden": 3.907421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2737090703099966,
      "step": 2370
    },
    {
      "epoch": 0.238,
      "grad_norm": 42.0,
      "grad_norm_var": 58.614322916666666,
      "learning_rate": 0.0001,
      "loss": 9.0391,
      "loss/crossentropy": 2.381209687888622,
      "loss/hidden": 3.9296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2849856551736593,
      "step": 2380
    },
    {
      "epoch": 0.239,
      "grad_norm": 39.75,
      "grad_norm_var": 17.541666666666668,
      "learning_rate": 0.0001,
      "loss": 9.1216,
      "loss/crossentropy": 2.2550544410943987,
      "loss/hidden": 4.000390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2924022350460291,
      "step": 2390
    },
    {
      "epoch": 0.24,
      "grad_norm": 46.25,
      "grad_norm_var": 18.491666666666667,
      "learning_rate": 0.0001,
      "loss": 8.9833,
      "loss/crossentropy": 2.2908297032117844,
      "loss/hidden": 3.901953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.26052255779504774,
      "step": 2400
    },
    {
      "epoch": 0.241,
      "grad_norm": 38.25,
      "grad_norm_var": 23.640625,
      "learning_rate": 0.0001,
      "loss": 9.1123,
      "loss/crossentropy": 2.3713207334280013,
      "loss/hidden": 3.988671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2721697688102722,
      "step": 2410
    },
    {
      "epoch": 0.242,
      "grad_norm": 42.5,
      "grad_norm_var": 69.77473958333333,
      "learning_rate": 0.0001,
      "loss": 8.9108,
      "loss/crossentropy": 2.0408870808780195,
      "loss/hidden": 3.873046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24748602956533433,
      "step": 2420
    },
    {
      "epoch": 0.243,
      "grad_norm": 37.25,
      "grad_norm_var": 72.11015625,
      "learning_rate": 0.0001,
      "loss": 9.0111,
      "loss/crossentropy": 2.3027436569333077,
      "loss/hidden": 3.908984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27137077748775484,
      "step": 2430
    },
    {
      "epoch": 0.244,
      "grad_norm": 54.75,
      "grad_norm_var": 55.057291666666664,
      "learning_rate": 0.0001,
      "loss": 9.0763,
      "loss/crossentropy": 2.2573105663061144,
      "loss/hidden": 4.1703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3013453852385283,
      "step": 2440
    },
    {
      "epoch": 0.245,
      "grad_norm": 39.25,
      "grad_norm_var": 50.00416666666667,
      "learning_rate": 0.0001,
      "loss": 9.0658,
      "loss/crossentropy": 2.4534697026014327,
      "loss/hidden": 4.172265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2898527968674898,
      "step": 2450
    },
    {
      "epoch": 0.246,
      "grad_norm": 40.75,
      "grad_norm_var": 18.71640625,
      "learning_rate": 0.0001,
      "loss": 8.9953,
      "loss/crossentropy": 2.357613870501518,
      "loss/hidden": 3.9125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27832051999866964,
      "step": 2460
    },
    {
      "epoch": 0.247,
      "grad_norm": 34.0,
      "grad_norm_var": 14.832291666666666,
      "learning_rate": 0.0001,
      "loss": 8.9576,
      "loss/crossentropy": 2.295011055469513,
      "loss/hidden": 4.029296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.29190085306763647,
      "step": 2470
    },
    {
      "epoch": 0.248,
      "grad_norm": 50.75,
      "grad_norm_var": 24.075,
      "learning_rate": 0.0001,
      "loss": 8.7866,
      "loss/crossentropy": 2.2598410531878472,
      "loss/hidden": 3.84921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2570509884506464,
      "step": 2480
    },
    {
      "epoch": 0.249,
      "grad_norm": 38.0,
      "grad_norm_var": 97.60416666666667,
      "learning_rate": 0.0001,
      "loss": 9.0289,
      "loss/crossentropy": 2.365708181262016,
      "loss/hidden": 3.835546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2712419513612986,
      "step": 2490
    },
    {
      "epoch": 0.25,
      "grad_norm": 44.5,
      "grad_norm_var": 104.10729166666667,
      "learning_rate": 0.0001,
      "loss": 8.8744,
      "loss/crossentropy": 2.063434064388275,
      "loss/hidden": 3.873046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2460779383778572,
      "step": 2500
    },
    {
      "epoch": 0.251,
      "grad_norm": 41.75,
      "grad_norm_var": 28.974739583333335,
      "learning_rate": 0.0001,
      "loss": 8.9272,
      "loss/crossentropy": 2.2377428650856017,
      "loss/hidden": 3.968359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26241020299494267,
      "step": 2510
    },
    {
      "epoch": 0.252,
      "grad_norm": 43.5,
      "grad_norm_var": 20.84140625,
      "learning_rate": 0.0001,
      "loss": 8.9773,
      "loss/crossentropy": 2.2245729833841326,
      "loss/hidden": 3.854296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2788569286465645,
      "step": 2520
    },
    {
      "epoch": 0.253,
      "grad_norm": 37.25,
      "grad_norm_var": 10.36640625,
      "learning_rate": 0.0001,
      "loss": 8.8657,
      "loss/crossentropy": 2.19287933409214,
      "loss/hidden": 3.873046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2571034274995327,
      "step": 2530
    },
    {
      "epoch": 0.254,
      "grad_norm": 42.75,
      "grad_norm_var": 25.641666666666666,
      "learning_rate": 0.0001,
      "loss": 9.0621,
      "loss/crossentropy": 2.344786374270916,
      "loss/hidden": 4.005078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.29999860040843485,
      "step": 2540
    },
    {
      "epoch": 0.255,
      "grad_norm": 45.25,
      "grad_norm_var": 28.983333333333334,
      "learning_rate": 0.0001,
      "loss": 8.8825,
      "loss/crossentropy": 2.165843137353659,
      "loss/hidden": 3.794921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24595264531672,
      "step": 2550
    },
    {
      "epoch": 0.256,
      "grad_norm": 44.0,
      "grad_norm_var": 13.407291666666667,
      "learning_rate": 0.0001,
      "loss": 8.9041,
      "loss/crossentropy": 2.3431010633707046,
      "loss/hidden": 3.880859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27991249822080133,
      "step": 2560
    },
    {
      "epoch": 0.257,
      "grad_norm": 49.25,
      "grad_norm_var": 19.137239583333333,
      "learning_rate": 0.0001,
      "loss": 8.7585,
      "loss/crossentropy": 2.2741693764925004,
      "loss/hidden": 3.848046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2573831077665091,
      "step": 2570
    },
    {
      "epoch": 0.258,
      "grad_norm": 40.75,
      "grad_norm_var": 23.523958333333333,
      "learning_rate": 0.0001,
      "loss": 9.0385,
      "loss/crossentropy": 2.3741003662347793,
      "loss/hidden": 3.9703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.3021115079522133,
      "step": 2580
    },
    {
      "epoch": 0.259,
      "grad_norm": 38.5,
      "grad_norm_var": 20.873958333333334,
      "learning_rate": 0.0001,
      "loss": 8.8554,
      "loss/crossentropy": 2.259954023361206,
      "loss/hidden": 3.92734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2688195243477821,
      "step": 2590
    },
    {
      "epoch": 0.26,
      "grad_norm": 33.25,
      "grad_norm_var": 16.76015625,
      "learning_rate": 0.0001,
      "loss": 8.8802,
      "loss/crossentropy": 2.155090569704771,
      "loss/hidden": 3.91796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.256628708448261,
      "step": 2600
    },
    {
      "epoch": 0.261,
      "grad_norm": 41.5,
      "grad_norm_var": 15.032291666666667,
      "learning_rate": 0.0001,
      "loss": 9.0016,
      "loss/crossentropy": 2.365834577381611,
      "loss/hidden": 3.85859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.27286841757595537,
      "step": 2610
    },
    {
      "epoch": 0.262,
      "grad_norm": 41.75,
      "grad_norm_var": 9.50390625,
      "learning_rate": 0.0001,
      "loss": 8.8993,
      "loss/crossentropy": 2.137889374792576,
      "loss/hidden": 3.96171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27018810212612154,
      "step": 2620
    },
    {
      "epoch": 0.263,
      "grad_norm": 44.75,
      "grad_norm_var": 12.58515625,
      "learning_rate": 0.0001,
      "loss": 8.9333,
      "loss/crossentropy": 2.279120808839798,
      "loss/hidden": 3.8859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26760734505951406,
      "step": 2630
    },
    {
      "epoch": 0.264,
      "grad_norm": 33.75,
      "grad_norm_var": 20.205989583333334,
      "learning_rate": 0.0001,
      "loss": 8.998,
      "loss/crossentropy": 2.133891487121582,
      "loss/hidden": 3.978125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2548325901851058,
      "step": 2640
    },
    {
      "epoch": 0.265,
      "grad_norm": 38.25,
      "grad_norm_var": 24.26640625,
      "learning_rate": 0.0001,
      "loss": 8.8773,
      "loss/crossentropy": 2.0680222399532795,
      "loss/hidden": 3.983984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2616095909848809,
      "step": 2650
    },
    {
      "epoch": 0.266,
      "grad_norm": 37.75,
      "grad_norm_var": 24.65390625,
      "learning_rate": 0.0001,
      "loss": 8.8787,
      "loss/crossentropy": 2.2427713751792906,
      "loss/hidden": 3.976171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2840047996491194,
      "step": 2660
    },
    {
      "epoch": 0.267,
      "grad_norm": 50.0,
      "grad_norm_var": 22.190625,
      "learning_rate": 0.0001,
      "loss": 8.878,
      "loss/crossentropy": 2.3332558259367944,
      "loss/hidden": 3.774609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2527316328138113,
      "step": 2670
    },
    {
      "epoch": 0.268,
      "grad_norm": 58.5,
      "grad_norm_var": 53.95,
      "learning_rate": 0.0001,
      "loss": 8.8498,
      "loss/crossentropy": 2.116853891313076,
      "loss/hidden": 3.810546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24547674022614957,
      "step": 2680
    },
    {
      "epoch": 0.269,
      "grad_norm": 37.5,
      "grad_norm_var": 48.47057291666667,
      "learning_rate": 0.0001,
      "loss": 8.9221,
      "loss/crossentropy": 2.468301197886467,
      "loss/hidden": 3.809765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.26750445999205114,
      "step": 2690
    },
    {
      "epoch": 0.27,
      "grad_norm": 46.5,
      "grad_norm_var": 16.812239583333334,
      "learning_rate": 0.0001,
      "loss": 8.7264,
      "loss/crossentropy": 2.1068901009857655,
      "loss/hidden": 3.970703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2827789710834622,
      "step": 2700
    },
    {
      "epoch": 0.271,
      "grad_norm": 41.25,
      "grad_norm_var": 17.765625,
      "learning_rate": 0.0001,
      "loss": 8.766,
      "loss/crossentropy": 2.4922314494848252,
      "loss/hidden": 3.943359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.28248917534947393,
      "step": 2710
    },
    {
      "epoch": 0.272,
      "grad_norm": 37.25,
      "grad_norm_var": 4.524739583333333,
      "learning_rate": 0.0001,
      "loss": 8.8324,
      "loss/crossentropy": 2.268562327325344,
      "loss/hidden": 3.8296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.250153512135148,
      "step": 2720
    },
    {
      "epoch": 0.273,
      "grad_norm": 38.5,
      "grad_norm_var": 11.832291666666666,
      "learning_rate": 0.0001,
      "loss": 8.8538,
      "loss/crossentropy": 2.171993290632963,
      "loss/hidden": 3.860546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2605165271088481,
      "step": 2730
    },
    {
      "epoch": 0.274,
      "grad_norm": 37.25,
      "grad_norm_var": 17.5625,
      "learning_rate": 0.0001,
      "loss": 8.8625,
      "loss/crossentropy": 2.3496526792645454,
      "loss/hidden": 3.78828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2532901844009757,
      "step": 2740
    },
    {
      "epoch": 0.275,
      "grad_norm": 38.5,
      "grad_norm_var": 18.69140625,
      "learning_rate": 0.0001,
      "loss": 8.8941,
      "loss/crossentropy": 2.377558296918869,
      "loss/hidden": 3.844921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2773334577679634,
      "step": 2750
    },
    {
      "epoch": 0.276,
      "grad_norm": 46.75,
      "grad_norm_var": 22.22890625,
      "learning_rate": 0.0001,
      "loss": 8.7839,
      "loss/crossentropy": 2.248410400748253,
      "loss/hidden": 3.980078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27028046883642676,
      "step": 2760
    },
    {
      "epoch": 0.277,
      "grad_norm": 38.75,
      "grad_norm_var": 23.164322916666666,
      "learning_rate": 0.0001,
      "loss": 8.7791,
      "loss/crossentropy": 2.4241216853260994,
      "loss/hidden": 3.8984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2659046190790832,
      "step": 2770
    },
    {
      "epoch": 0.278,
      "grad_norm": 53.25,
      "grad_norm_var": 59.78098958333333,
      "learning_rate": 0.0001,
      "loss": 8.8019,
      "loss/crossentropy": 2.1859550148248674,
      "loss/hidden": 3.828515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.25679499059915545,
      "step": 2780
    },
    {
      "epoch": 0.279,
      "grad_norm": 41.5,
      "grad_norm_var": 62.925455729166664,
      "learning_rate": 0.0001,
      "loss": 8.8086,
      "loss/crossentropy": 2.312130589783192,
      "loss/hidden": 3.91171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2820352425798774,
      "step": 2790
    },
    {
      "epoch": 0.28,
      "grad_norm": 46.75,
      "grad_norm_var": 18.951822916666668,
      "learning_rate": 0.0001,
      "loss": 8.7941,
      "loss/crossentropy": 2.258910335600376,
      "loss/hidden": 3.902734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2627917256206274,
      "step": 2800
    },
    {
      "epoch": 0.281,
      "grad_norm": 40.25,
      "grad_norm_var": 17.639322916666668,
      "learning_rate": 0.0001,
      "loss": 8.7417,
      "loss/crossentropy": 2.2589244581758976,
      "loss/hidden": 3.753515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2513848140835762,
      "step": 2810
    },
    {
      "epoch": 0.282,
      "grad_norm": 40.5,
      "grad_norm_var": 12.032291666666667,
      "learning_rate": 0.0001,
      "loss": 8.8627,
      "loss/crossentropy": 2.17684805393219,
      "loss/hidden": 3.853125,
      "loss/jsd": 0.0,
      "loss/logits": 0.255695578455925,
      "step": 2820
    },
    {
      "epoch": 0.283,
      "grad_norm": 40.75,
      "grad_norm_var": 114.61432291666667,
      "learning_rate": 0.0001,
      "loss": 8.8412,
      "loss/crossentropy": 2.2954238772392275,
      "loss/hidden": 3.918359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2858715243637562,
      "step": 2830
    },
    {
      "epoch": 0.284,
      "grad_norm": 40.5,
      "grad_norm_var": 5.641666666666667,
      "learning_rate": 0.0001,
      "loss": 8.9391,
      "loss/crossentropy": 2.329847712814808,
      "loss/hidden": 3.7625,
      "loss/jsd": 0.0,
      "loss/logits": 0.25534543097019197,
      "step": 2840
    },
    {
      "epoch": 0.285,
      "grad_norm": 38.5,
      "grad_norm_var": 40.84557291666667,
      "learning_rate": 0.0001,
      "loss": 8.7803,
      "loss/crossentropy": 2.2853938594460486,
      "loss/hidden": 3.866796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24832999743521214,
      "step": 2850
    },
    {
      "epoch": 0.286,
      "grad_norm": 36.75,
      "grad_norm_var": 39.958072916666666,
      "learning_rate": 0.0001,
      "loss": 8.6796,
      "loss/crossentropy": 2.234159553050995,
      "loss/hidden": 3.847265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.26201403168961407,
      "step": 2860
    },
    {
      "epoch": 0.287,
      "grad_norm": 39.0,
      "grad_norm_var": 9.457291666666666,
      "learning_rate": 0.0001,
      "loss": 8.7109,
      "loss/crossentropy": 2.2819659531116487,
      "loss/hidden": 3.894140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.26489345021545885,
      "step": 2870
    },
    {
      "epoch": 0.288,
      "grad_norm": 37.25,
      "grad_norm_var": 9.268489583333333,
      "learning_rate": 0.0001,
      "loss": 8.828,
      "loss/crossentropy": 2.4358034074306487,
      "loss/hidden": 3.81328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27083273865282537,
      "step": 2880
    },
    {
      "epoch": 0.289,
      "grad_norm": 42.0,
      "grad_norm_var": 12.805989583333334,
      "learning_rate": 0.0001,
      "loss": 8.8211,
      "loss/crossentropy": 2.2965500839054585,
      "loss/hidden": 3.869921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.26719480073079466,
      "step": 2890
    },
    {
      "epoch": 0.29,
      "grad_norm": 37.25,
      "grad_norm_var": 9.020572916666667,
      "learning_rate": 0.0001,
      "loss": 8.7927,
      "loss/crossentropy": 2.3898714184761047,
      "loss/hidden": 3.930078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2959909211844206,
      "step": 2900
    },
    {
      "epoch": 0.291,
      "grad_norm": 44.25,
      "grad_norm_var": 10.283333333333333,
      "learning_rate": 0.0001,
      "loss": 8.7893,
      "loss/crossentropy": 2.349339473247528,
      "loss/hidden": 3.7953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25724136754870414,
      "step": 2910
    },
    {
      "epoch": 0.292,
      "grad_norm": 35.5,
      "grad_norm_var": 17.418489583333333,
      "learning_rate": 0.0001,
      "loss": 8.7582,
      "loss/crossentropy": 2.1329104267060757,
      "loss/hidden": 3.83046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2572308249771595,
      "step": 2920
    },
    {
      "epoch": 0.293,
      "grad_norm": 42.0,
      "grad_norm_var": 14.87265625,
      "learning_rate": 0.0001,
      "loss": 8.6388,
      "loss/crossentropy": 2.23364320397377,
      "loss/hidden": 3.82421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2617119399830699,
      "step": 2930
    },
    {
      "epoch": 0.294,
      "grad_norm": 42.75,
      "grad_norm_var": 8.204166666666667,
      "learning_rate": 0.0001,
      "loss": 8.6531,
      "loss/crossentropy": 2.1770762100815775,
      "loss/hidden": 3.833984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2418960839509964,
      "step": 2940
    },
    {
      "epoch": 0.295,
      "grad_norm": 45.5,
      "grad_norm_var": 16.265625,
      "learning_rate": 0.0001,
      "loss": 8.8053,
      "loss/crossentropy": 2.18697277456522,
      "loss/hidden": 3.78125,
      "loss/jsd": 0.0,
      "loss/logits": 0.26141371857374907,
      "step": 2950
    },
    {
      "epoch": 0.296,
      "grad_norm": 32.25,
      "grad_norm_var": 19.145833333333332,
      "learning_rate": 0.0001,
      "loss": 8.7178,
      "loss/crossentropy": 2.303604170680046,
      "loss/hidden": 3.738671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2531938493251801,
      "step": 2960
    },
    {
      "epoch": 0.297,
      "grad_norm": 33.5,
      "grad_norm_var": 10.88515625,
      "learning_rate": 0.0001,
      "loss": 8.5781,
      "loss/crossentropy": 2.112999178469181,
      "loss/hidden": 3.853515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2678428884595633,
      "step": 2970
    },
    {
      "epoch": 0.298,
      "grad_norm": 44.0,
      "grad_norm_var": 32.84973958333333,
      "learning_rate": 0.0001,
      "loss": 8.8415,
      "loss/crossentropy": 2.407758575677872,
      "loss/hidden": 3.925,
      "loss/jsd": 0.0,
      "loss/logits": 0.3035837195813656,
      "step": 2980
    },
    {
      "epoch": 0.299,
      "grad_norm": 36.5,
      "grad_norm_var": 21.0125,
      "learning_rate": 0.0001,
      "loss": 8.6496,
      "loss/crossentropy": 2.1467753663659095,
      "loss/hidden": 3.802734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25200750436633823,
      "step": 2990
    },
    {
      "epoch": 0.3,
      "grad_norm": 37.25,
      "grad_norm_var": 102.01848958333333,
      "learning_rate": 0.0001,
      "loss": 8.5981,
      "loss/crossentropy": 2.22710300385952,
      "loss/hidden": 3.741015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24842255041003228,
      "step": 3000
    },
    {
      "epoch": 0.301,
      "grad_norm": 37.5,
      "grad_norm_var": 20.633072916666666,
      "learning_rate": 0.0001,
      "loss": 8.6443,
      "loss/crossentropy": 2.3478307321667673,
      "loss/hidden": 3.805859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25510224178433416,
      "step": 3010
    },
    {
      "epoch": 0.302,
      "grad_norm": 41.5,
      "grad_norm_var": 8.864322916666667,
      "learning_rate": 0.0001,
      "loss": 8.7802,
      "loss/crossentropy": 2.3206353336572647,
      "loss/hidden": 3.855078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2652175173163414,
      "step": 3020
    },
    {
      "epoch": 0.303,
      "grad_norm": 46.25,
      "grad_norm_var": 1.6257994395224812e+18,
      "learning_rate": 0.0001,
      "loss": 8.8456,
      "loss/crossentropy": 2.3741259276866913,
      "loss/hidden": 3.973046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2950541414320469,
      "step": 3030
    },
    {
      "epoch": 0.304,
      "grad_norm": 38.25,
      "grad_norm_var": 1.6257994392887188e+18,
      "learning_rate": 0.0001,
      "loss": 8.7725,
      "loss/crossentropy": 2.413549691438675,
      "loss/hidden": 3.87734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26055113933980467,
      "step": 3040
    },
    {
      "epoch": 0.305,
      "grad_norm": 46.0,
      "grad_norm_var": 297.95807291666665,
      "learning_rate": 0.0001,
      "loss": 8.66,
      "loss/crossentropy": 2.3086455732584,
      "loss/hidden": 3.956640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2837849177420139,
      "step": 3050
    },
    {
      "epoch": 0.306,
      "grad_norm": 38.75,
      "grad_norm_var": 326.315625,
      "learning_rate": 0.0001,
      "loss": 8.7568,
      "loss/crossentropy": 2.3664418935775755,
      "loss/hidden": 3.912890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2698039198294282,
      "step": 3060
    },
    {
      "epoch": 0.307,
      "grad_norm": 36.5,
      "grad_norm_var": 21.515625,
      "learning_rate": 0.0001,
      "loss": 8.7201,
      "loss/crossentropy": 2.1954890489578247,
      "loss/hidden": 3.8921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2684766609221697,
      "step": 3070
    },
    {
      "epoch": 0.308,
      "grad_norm": 41.0,
      "grad_norm_var": 7.140625,
      "learning_rate": 0.0001,
      "loss": 8.8511,
      "loss/crossentropy": 2.310048124939203,
      "loss/hidden": 3.84609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2778544146567583,
      "step": 3080
    },
    {
      "epoch": 0.309,
      "grad_norm": 35.25,
      "grad_norm_var": 10.1875,
      "learning_rate": 0.0001,
      "loss": 8.7144,
      "loss/crossentropy": 2.3563184320926664,
      "loss/hidden": 3.8625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2816514492034912,
      "step": 3090
    },
    {
      "epoch": 0.31,
      "grad_norm": 36.75,
      "grad_norm_var": 2.279947281849385e+18,
      "learning_rate": 0.0001,
      "loss": 8.7129,
      "loss/crossentropy": 2.439156624674797,
      "loss/hidden": 3.85078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.28767146319150927,
      "step": 3100
    },
    {
      "epoch": 0.311,
      "grad_norm": 39.5,
      "grad_norm_var": 2.2799472807295058e+18,
      "learning_rate": 0.0001,
      "loss": 8.6894,
      "loss/crossentropy": 2.176630274951458,
      "loss/hidden": 3.76484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25743562281131743,
      "step": 3110
    },
    {
      "epoch": 0.312,
      "grad_norm": 36.0,
      "grad_norm_var": 42.69348958333333,
      "learning_rate": 0.0001,
      "loss": 8.6887,
      "loss/crossentropy": 2.2695231288671494,
      "loss/hidden": 3.796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2609828669577837,
      "step": 3120
    },
    {
      "epoch": 0.313,
      "grad_norm": 38.75,
      "grad_norm_var": 12.6875,
      "learning_rate": 0.0001,
      "loss": 8.7316,
      "loss/crossentropy": 2.2914074435830116,
      "loss/hidden": 3.970703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25941712930798533,
      "step": 3130
    },
    {
      "epoch": 0.314,
      "grad_norm": 35.5,
      "grad_norm_var": 9.191666666666666,
      "learning_rate": 0.0001,
      "loss": 8.675,
      "loss/crossentropy": 2.2625655576586725,
      "loss/hidden": 3.950390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.26395085640251637,
      "step": 3140
    },
    {
      "epoch": 0.315,
      "grad_norm": 48.25,
      "grad_norm_var": 64.8125,
      "learning_rate": 0.0001,
      "loss": 8.8068,
      "loss/crossentropy": 2.157360579818487,
      "loss/hidden": 3.814453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2572598461061716,
      "step": 3150
    },
    {
      "epoch": 0.316,
      "grad_norm": 36.0,
      "grad_norm_var": 69.0212890625,
      "learning_rate": 0.0001,
      "loss": 8.656,
      "loss/crossentropy": 2.24196752011776,
      "loss/hidden": 3.821875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27792793661355975,
      "step": 3160
    },
    {
      "epoch": 0.317,
      "grad_norm": 35.0,
      "grad_norm_var": 8.108333333333333,
      "learning_rate": 0.0001,
      "loss": 8.662,
      "loss/crossentropy": 2.2520006895065308,
      "loss/hidden": 3.708203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2597096076235175,
      "step": 3170
    },
    {
      "epoch": 0.318,
      "grad_norm": 34.0,
      "grad_norm_var": 12.848958333333334,
      "learning_rate": 0.0001,
      "loss": 8.5281,
      "loss/crossentropy": 2.2527818381786346,
      "loss/hidden": 3.772265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2495524413883686,
      "step": 3180
    },
    {
      "epoch": 0.319,
      "grad_norm": 41.0,
      "grad_norm_var": 51.365625,
      "learning_rate": 0.0001,
      "loss": 8.7132,
      "loss/crossentropy": 2.2904104314744473,
      "loss/hidden": 3.944140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2817814026027918,
      "step": 3190
    },
    {
      "epoch": 0.32,
      "grad_norm": 36.25,
      "grad_norm_var": 22.22265625,
      "learning_rate": 0.0001,
      "loss": 8.4574,
      "loss/crossentropy": 2.090634661912918,
      "loss/hidden": 3.79921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24919861294329165,
      "step": 3200
    },
    {
      "epoch": 0.321,
      "grad_norm": 38.5,
      "grad_norm_var": 10.030989583333334,
      "learning_rate": 0.0001,
      "loss": 8.5723,
      "loss/crossentropy": 2.117210125923157,
      "loss/hidden": 3.672265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2293582683429122,
      "step": 3210
    },
    {
      "epoch": 0.322,
      "grad_norm": 36.25,
      "grad_norm_var": 17.933072916666667,
      "learning_rate": 0.0001,
      "loss": 8.5557,
      "loss/crossentropy": 2.252604177594185,
      "loss/hidden": 3.71171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2455908928066492,
      "step": 3220
    },
    {
      "epoch": 0.323,
      "grad_norm": 34.0,
      "grad_norm_var": 15.673958333333333,
      "learning_rate": 0.0001,
      "loss": 8.6303,
      "loss/crossentropy": 2.2780065298080445,
      "loss/hidden": 3.701953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25410398468375206,
      "step": 3230
    },
    {
      "epoch": 0.324,
      "grad_norm": 37.75,
      "grad_norm_var": 31.186393229166665,
      "learning_rate": 0.0001,
      "loss": 8.6026,
      "loss/crossentropy": 2.3745017647743225,
      "loss/hidden": 3.819140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2740194508805871,
      "step": 3240
    },
    {
      "epoch": 0.325,
      "grad_norm": 44.75,
      "grad_norm_var": 16.865625,
      "learning_rate": 0.0001,
      "loss": 8.7984,
      "loss/crossentropy": 2.3756151482462884,
      "loss/hidden": 3.894921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2671732004731894,
      "step": 3250
    },
    {
      "epoch": 0.326,
      "grad_norm": 44.0,
      "grad_norm_var": 18.073893229166668,
      "learning_rate": 0.0001,
      "loss": 8.6327,
      "loss/crossentropy": 2.3531317353248595,
      "loss/hidden": 3.717578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25254391208291055,
      "step": 3260
    },
    {
      "epoch": 0.327,
      "grad_norm": 38.5,
      "grad_norm_var": 10.248893229166667,
      "learning_rate": 0.0001,
      "loss": 8.6167,
      "loss/crossentropy": 2.283932936191559,
      "loss/hidden": 3.96171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.28510861806571486,
      "step": 3270
    },
    {
      "epoch": 0.328,
      "grad_norm": 32.25,
      "grad_norm_var": 7.364322916666667,
      "learning_rate": 0.0001,
      "loss": 8.6054,
      "loss/crossentropy": 2.1746396124362946,
      "loss/hidden": 3.709375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24525153413414955,
      "step": 3280
    },
    {
      "epoch": 0.329,
      "grad_norm": 41.25,
      "grad_norm_var": 18.978059895833333,
      "learning_rate": 0.0001,
      "loss": 8.8527,
      "loss/crossentropy": 2.270473413169384,
      "loss/hidden": 3.8609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.274440161883831,
      "step": 3290
    },
    {
      "epoch": 0.33,
      "grad_norm": 48.5,
      "grad_norm_var": 18.778580729166666,
      "learning_rate": 0.0001,
      "loss": 8.6327,
      "loss/crossentropy": 2.3394594848155976,
      "loss/hidden": 3.77109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25994330234825613,
      "step": 3300
    },
    {
      "epoch": 0.331,
      "grad_norm": 34.75,
      "grad_norm_var": 17.190625,
      "learning_rate": 0.0001,
      "loss": 8.7313,
      "loss/crossentropy": 2.347683647274971,
      "loss/hidden": 3.894921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27402915358543395,
      "step": 3310
    },
    {
      "epoch": 0.332,
      "grad_norm": 31.0,
      "grad_norm_var": 27.712239583333332,
      "learning_rate": 0.0001,
      "loss": 8.7637,
      "loss/crossentropy": 2.257617971301079,
      "loss/hidden": 3.88046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2689027152955532,
      "step": 3320
    },
    {
      "epoch": 0.333,
      "grad_norm": 40.0,
      "grad_norm_var": 27.319205729166665,
      "learning_rate": 0.0001,
      "loss": 8.829,
      "loss/crossentropy": 2.2091607600450516,
      "loss/hidden": 3.83515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2670388799160719,
      "step": 3330
    },
    {
      "epoch": 0.334,
      "grad_norm": 37.5,
      "grad_norm_var": 38.63098958333333,
      "learning_rate": 0.0001,
      "loss": 8.759,
      "loss/crossentropy": 2.410393309593201,
      "loss/hidden": 3.73359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2507057674229145,
      "step": 3340
    },
    {
      "epoch": 0.335,
      "grad_norm": 35.25,
      "grad_norm_var": 9.292122395833333,
      "learning_rate": 0.0001,
      "loss": 8.6462,
      "loss/crossentropy": 2.299106788635254,
      "loss/hidden": 3.76328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2520503532141447,
      "step": 3350
    },
    {
      "epoch": 0.336,
      "grad_norm": 44.25,
      "grad_norm_var": 13.2869140625,
      "learning_rate": 0.0001,
      "loss": 8.6432,
      "loss/crossentropy": 2.2736548662185667,
      "loss/hidden": 3.81796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2679216586053371,
      "step": 3360
    },
    {
      "epoch": 0.337,
      "grad_norm": 40.25,
      "grad_norm_var": 16.704622395833333,
      "learning_rate": 0.0001,
      "loss": 8.6969,
      "loss/crossentropy": 2.300890862941742,
      "loss/hidden": 3.85703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27277763597667215,
      "step": 3370
    },
    {
      "epoch": 0.338,
      "grad_norm": 36.25,
      "grad_norm_var": 7.362239583333333,
      "learning_rate": 0.0001,
      "loss": 8.6139,
      "loss/crossentropy": 2.2713751554489137,
      "loss/hidden": 3.739453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2698351971805096,
      "step": 3380
    },
    {
      "epoch": 0.339,
      "grad_norm": 35.5,
      "grad_norm_var": 7.383333333333334,
      "learning_rate": 0.0001,
      "loss": 8.6652,
      "loss/crossentropy": 2.154055279493332,
      "loss/hidden": 3.784375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2516574438661337,
      "step": 3390
    },
    {
      "epoch": 0.34,
      "grad_norm": 46.5,
      "grad_norm_var": 231.42083333333332,
      "learning_rate": 0.0001,
      "loss": 8.7521,
      "loss/crossentropy": 2.1560363829135896,
      "loss/hidden": 3.78828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2552214227616787,
      "step": 3400
    },
    {
      "epoch": 0.341,
      "grad_norm": 34.0,
      "grad_norm_var": 16.01015625,
      "learning_rate": 0.0001,
      "loss": 8.6719,
      "loss/crossentropy": 2.210598033666611,
      "loss/hidden": 3.7078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24775836057960987,
      "step": 3410
    },
    {
      "epoch": 0.342,
      "grad_norm": 46.75,
      "grad_norm_var": 23.555989583333332,
      "learning_rate": 0.0001,
      "loss": 8.732,
      "loss/crossentropy": 2.0089349642395975,
      "loss/hidden": 3.95625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24481147788465024,
      "step": 3420
    },
    {
      "epoch": 0.343,
      "grad_norm": 43.5,
      "grad_norm_var": 24.71015625,
      "learning_rate": 0.0001,
      "loss": 8.5835,
      "loss/crossentropy": 2.167103961110115,
      "loss/hidden": 3.73046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23311931267380714,
      "step": 3430
    },
    {
      "epoch": 0.344,
      "grad_norm": 38.75,
      "grad_norm_var": 16.239322916666666,
      "learning_rate": 0.0001,
      "loss": 8.7327,
      "loss/crossentropy": 2.223423732817173,
      "loss/hidden": 3.916796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27787868976593016,
      "step": 3440
    },
    {
      "epoch": 0.345,
      "grad_norm": 32.5,
      "grad_norm_var": 12.4166015625,
      "learning_rate": 0.0001,
      "loss": 8.6674,
      "loss/crossentropy": 2.190432313084602,
      "loss/hidden": 3.933203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2527661222964525,
      "step": 3450
    },
    {
      "epoch": 0.346,
      "grad_norm": 38.5,
      "grad_norm_var": 239.47337239583334,
      "learning_rate": 0.0001,
      "loss": 8.678,
      "loss/crossentropy": 2.121590741723776,
      "loss/hidden": 3.746875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22831694399937988,
      "step": 3460
    },
    {
      "epoch": 0.347,
      "grad_norm": 32.5,
      "grad_norm_var": 24.29765625,
      "learning_rate": 0.0001,
      "loss": 8.5703,
      "loss/crossentropy": 2.238359749317169,
      "loss/hidden": 3.776171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2545921359211206,
      "step": 3470
    },
    {
      "epoch": 0.348,
      "grad_norm": 31.375,
      "grad_norm_var": 57.77291666666667,
      "learning_rate": 0.0001,
      "loss": 8.667,
      "loss/crossentropy": 2.183069321513176,
      "loss/hidden": 3.723828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.26235801838338374,
      "step": 3480
    },
    {
      "epoch": 0.349,
      "grad_norm": 37.25,
      "grad_norm_var": 47.95104166666667,
      "learning_rate": 0.0001,
      "loss": 8.6657,
      "loss/crossentropy": 2.294294211268425,
      "loss/hidden": 3.7875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2434792961925268,
      "step": 3490
    },
    {
      "epoch": 0.35,
      "grad_norm": 55.75,
      "grad_norm_var": 32.8931640625,
      "learning_rate": 0.0001,
      "loss": 8.6293,
      "loss/crossentropy": 2.2934862852096556,
      "loss/hidden": 3.7671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2526032764464617,
      "step": 3500
    },
    {
      "epoch": 0.351,
      "grad_norm": 41.75,
      "grad_norm_var": 37.820247395833334,
      "learning_rate": 0.0001,
      "loss": 8.6558,
      "loss/crossentropy": 2.240943320095539,
      "loss/hidden": 3.798828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2533996310085058,
      "step": 3510
    },
    {
      "epoch": 0.352,
      "grad_norm": 34.0,
      "grad_norm_var": 19.158333333333335,
      "learning_rate": 0.0001,
      "loss": 8.7265,
      "loss/crossentropy": 2.2778089314699175,
      "loss/hidden": 3.74765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24110115878283978,
      "step": 3520
    },
    {
      "epoch": 0.353,
      "grad_norm": 33.75,
      "grad_norm_var": 14.61640625,
      "learning_rate": 0.0001,
      "loss": 8.6184,
      "loss/crossentropy": 2.2676281452178957,
      "loss/hidden": 3.703515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.25084604155272244,
      "step": 3530
    },
    {
      "epoch": 0.354,
      "grad_norm": 41.25,
      "grad_norm_var": 11.601822916666666,
      "learning_rate": 0.0001,
      "loss": 8.6994,
      "loss/crossentropy": 2.291515235602856,
      "loss/hidden": 3.919140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.29151172675192355,
      "step": 3540
    },
    {
      "epoch": 0.355,
      "grad_norm": 35.75,
      "grad_norm_var": 7.620833333333334,
      "learning_rate": 0.0001,
      "loss": 8.606,
      "loss/crossentropy": 2.2359238654375075,
      "loss/hidden": 3.81484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25299829691648484,
      "step": 3550
    },
    {
      "epoch": 0.356,
      "grad_norm": 35.25,
      "grad_norm_var": 7.1,
      "learning_rate": 0.0001,
      "loss": 8.5051,
      "loss/crossentropy": 2.2826671570539476,
      "loss/hidden": 3.843359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26029736921191216,
      "step": 3560
    },
    {
      "epoch": 0.357,
      "grad_norm": 38.5,
      "grad_norm_var": 7.422330729166666,
      "learning_rate": 0.0001,
      "loss": 8.5601,
      "loss/crossentropy": 2.29564026594162,
      "loss/hidden": 3.790234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24953206069767475,
      "step": 3570
    },
    {
      "epoch": 0.358,
      "grad_norm": 34.25,
      "grad_norm_var": 9.414518229166667,
      "learning_rate": 0.0001,
      "loss": 8.677,
      "loss/crossentropy": 2.360885411500931,
      "loss/hidden": 3.812890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2608415879309177,
      "step": 3580
    },
    {
      "epoch": 0.359,
      "grad_norm": 41.75,
      "grad_norm_var": 7.558333333333334,
      "learning_rate": 0.0001,
      "loss": 8.591,
      "loss/crossentropy": 2.0790357582271097,
      "loss/hidden": 3.759375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23188311588019134,
      "step": 3590
    },
    {
      "epoch": 0.36,
      "grad_norm": 38.0,
      "grad_norm_var": 10.890625,
      "learning_rate": 0.0001,
      "loss": 8.6616,
      "loss/crossentropy": 2.289612150192261,
      "loss/hidden": 3.878125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2747206833213568,
      "step": 3600
    },
    {
      "epoch": 0.361,
      "grad_norm": 42.0,
      "grad_norm_var": 14.145572916666667,
      "learning_rate": 0.0001,
      "loss": 8.589,
      "loss/crossentropy": 2.132586442679167,
      "loss/hidden": 3.811328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25861772149801254,
      "step": 3610
    },
    {
      "epoch": 0.362,
      "grad_norm": 35.0,
      "grad_norm_var": 11.537239583333333,
      "learning_rate": 0.0001,
      "loss": 8.6171,
      "loss/crossentropy": 2.3269239187240602,
      "loss/hidden": 3.8265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2712526451796293,
      "step": 3620
    },
    {
      "epoch": 0.363,
      "grad_norm": 33.75,
      "grad_norm_var": 20.012239583333333,
      "learning_rate": 0.0001,
      "loss": 8.5134,
      "loss/crossentropy": 2.326130175590515,
      "loss/hidden": 3.791796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24642233476042746,
      "step": 3630
    },
    {
      "epoch": 0.364,
      "grad_norm": 31.875,
      "grad_norm_var": 30.459830729166665,
      "learning_rate": 0.0001,
      "loss": 8.5808,
      "loss/crossentropy": 2.30035699903965,
      "loss/hidden": 3.733984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26470062173902986,
      "step": 3640
    },
    {
      "epoch": 0.365,
      "grad_norm": 37.25,
      "grad_norm_var": 8.626041666666667,
      "learning_rate": 0.0001,
      "loss": 8.5614,
      "loss/crossentropy": 2.3303778156638146,
      "loss/hidden": 3.871875,
      "loss/jsd": 0.0,
      "loss/logits": 0.27759894989430905,
      "step": 3650
    },
    {
      "epoch": 0.366,
      "grad_norm": 33.0,
      "grad_norm_var": 5.374739583333334,
      "learning_rate": 0.0001,
      "loss": 8.5104,
      "loss/crossentropy": 2.3886318862438203,
      "loss/hidden": 3.801953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.26098744831979276,
      "step": 3660
    },
    {
      "epoch": 0.367,
      "grad_norm": 28.875,
      "grad_norm_var": 18.685416666666665,
      "learning_rate": 0.0001,
      "loss": 8.5148,
      "loss/crossentropy": 2.3471581265330315,
      "loss/hidden": 3.760546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.258730498701334,
      "step": 3670
    },
    {
      "epoch": 0.368,
      "grad_norm": 48.5,
      "grad_norm_var": 36.4744140625,
      "learning_rate": 0.0001,
      "loss": 8.5484,
      "loss/crossentropy": 2.1698090970516204,
      "loss/hidden": 3.70078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24077225737273694,
      "step": 3680
    },
    {
      "epoch": 0.369,
      "grad_norm": 30.125,
      "grad_norm_var": 25.111393229166666,
      "learning_rate": 0.0001,
      "loss": 8.5671,
      "loss/crossentropy": 2.1866169169545175,
      "loss/hidden": 3.8203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24500060379505156,
      "step": 3690
    },
    {
      "epoch": 0.37,
      "grad_norm": 30.375,
      "grad_norm_var": 2.130484072425508e+18,
      "learning_rate": 0.0001,
      "loss": 8.6635,
      "loss/crossentropy": 2.453831446170807,
      "loss/hidden": 4.06171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2722825076431036,
      "step": 3700
    },
    {
      "epoch": 0.371,
      "grad_norm": 32.5,
      "grad_norm_var": 32.78795572916667,
      "learning_rate": 0.0001,
      "loss": 8.5788,
      "loss/crossentropy": 2.212860561162233,
      "loss/hidden": 3.629296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21873269081115723,
      "step": 3710
    },
    {
      "epoch": 0.372,
      "grad_norm": 29.625,
      "grad_norm_var": 31.475455729166665,
      "learning_rate": 0.0001,
      "loss": 8.5924,
      "loss/crossentropy": 2.3226836264133452,
      "loss/hidden": 3.711328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2526125326752663,
      "step": 3720
    },
    {
      "epoch": 0.373,
      "grad_norm": 40.25,
      "grad_norm_var": 74.74680989583334,
      "learning_rate": 0.0001,
      "loss": 8.7309,
      "loss/crossentropy": 2.211090712249279,
      "loss/hidden": 3.85078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2619786085560918,
      "step": 3730
    },
    {
      "epoch": 0.374,
      "grad_norm": 30.5,
      "grad_norm_var": 49.24108072916667,
      "learning_rate": 0.0001,
      "loss": 8.4683,
      "loss/crossentropy": 2.117314028739929,
      "loss/hidden": 3.787890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24047958478331566,
      "step": 3740
    },
    {
      "epoch": 0.375,
      "grad_norm": 43.0,
      "grad_norm_var": 13.481184895833334,
      "learning_rate": 0.0001,
      "loss": 8.5891,
      "loss/crossentropy": 2.3331554174423217,
      "loss/hidden": 3.737109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.265634342469275,
      "step": 3750
    },
    {
      "epoch": 0.376,
      "grad_norm": 70.0,
      "grad_norm_var": 86.09837239583334,
      "learning_rate": 0.0001,
      "loss": 8.4671,
      "loss/crossentropy": 2.279883709549904,
      "loss/hidden": 3.764453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2557247843593359,
      "step": 3760
    },
    {
      "epoch": 0.377,
      "grad_norm": 37.5,
      "grad_norm_var": 101.2072265625,
      "learning_rate": 0.0001,
      "loss": 8.6263,
      "loss/crossentropy": 2.119837614893913,
      "loss/hidden": 3.7328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24443967882543802,
      "step": 3770
    },
    {
      "epoch": 0.378,
      "grad_norm": 36.0,
      "grad_norm_var": 11.41015625,
      "learning_rate": 0.0001,
      "loss": 8.573,
      "loss/crossentropy": 2.2743908286094667,
      "loss/hidden": 3.7046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24655351527035235,
      "step": 3780
    },
    {
      "epoch": 0.379,
      "grad_norm": 30.5,
      "grad_norm_var": 22.8337890625,
      "learning_rate": 0.0001,
      "loss": 8.5496,
      "loss/crossentropy": 2.298393335938454,
      "loss/hidden": 3.661328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2468328095972538,
      "step": 3790
    },
    {
      "epoch": 0.38,
      "grad_norm": 36.75,
      "grad_norm_var": 21.886393229166668,
      "learning_rate": 0.0001,
      "loss": 8.4728,
      "loss/crossentropy": 2.285892593860626,
      "loss/hidden": 3.696484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22759304326027632,
      "step": 3800
    },
    {
      "epoch": 0.381,
      "grad_norm": 37.0,
      "grad_norm_var": 25.983268229166665,
      "learning_rate": 0.0001,
      "loss": 8.4654,
      "loss/crossentropy": 2.200984264165163,
      "loss/hidden": 3.760546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2451606505550444,
      "step": 3810
    },
    {
      "epoch": 0.382,
      "grad_norm": 37.5,
      "grad_norm_var": 13.956705729166666,
      "learning_rate": 0.0001,
      "loss": 8.5469,
      "loss/crossentropy": 2.2647089801728724,
      "loss/hidden": 3.803125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2533489029854536,
      "step": 3820
    },
    {
      "epoch": 0.383,
      "grad_norm": 34.75,
      "grad_norm_var": 7.221809895833333,
      "learning_rate": 0.0001,
      "loss": 8.579,
      "loss/crossentropy": 2.386853316426277,
      "loss/hidden": 3.795703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2518702711910009,
      "step": 3830
    },
    {
      "epoch": 0.384,
      "grad_norm": 42.0,
      "grad_norm_var": 15.8650390625,
      "learning_rate": 0.0001,
      "loss": 8.4816,
      "loss/crossentropy": 2.3519598811864855,
      "loss/hidden": 3.75625,
      "loss/jsd": 0.0,
      "loss/logits": 0.253318839520216,
      "step": 3840
    },
    {
      "epoch": 0.385,
      "grad_norm": 38.25,
      "grad_norm_var": 14.728580729166667,
      "learning_rate": 0.0001,
      "loss": 8.7292,
      "loss/crossentropy": 2.164312995970249,
      "loss/hidden": 3.8234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24283661209046842,
      "step": 3850
    },
    {
      "epoch": 0.386,
      "grad_norm": 37.75,
      "grad_norm_var": 14.876822916666667,
      "learning_rate": 0.0001,
      "loss": 8.6236,
      "loss/crossentropy": 2.1916888520121574,
      "loss/hidden": 3.69140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23480207994580268,
      "step": 3860
    },
    {
      "epoch": 0.387,
      "grad_norm": 33.25,
      "grad_norm_var": 17.855208333333334,
      "learning_rate": 0.0001,
      "loss": 8.4472,
      "loss/crossentropy": 2.2794968456029894,
      "loss/hidden": 3.759375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2547815594822168,
      "step": 3870
    },
    {
      "epoch": 0.388,
      "grad_norm": 39.75,
      "grad_norm_var": 10.063541666666667,
      "learning_rate": 0.0001,
      "loss": 8.5547,
      "loss/crossentropy": 2.2261194586753845,
      "loss/hidden": 3.73203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24962050542235376,
      "step": 3880
    },
    {
      "epoch": 0.389,
      "grad_norm": 33.75,
      "grad_norm_var": 25.229622395833335,
      "learning_rate": 0.0001,
      "loss": 8.428,
      "loss/crossentropy": 2.2669249922037125,
      "loss/hidden": 3.589453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22769966274499892,
      "step": 3890
    },
    {
      "epoch": 0.39,
      "grad_norm": 34.75,
      "grad_norm_var": 22.89765625,
      "learning_rate": 0.0001,
      "loss": 8.4932,
      "loss/crossentropy": 2.2599128648638724,
      "loss/hidden": 3.744921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24268860407173634,
      "step": 3900
    },
    {
      "epoch": 0.391,
      "grad_norm": 37.0,
      "grad_norm_var": 44.83483072916667,
      "learning_rate": 0.0001,
      "loss": 8.5202,
      "loss/crossentropy": 2.1972982972860335,
      "loss/hidden": 3.812109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25300098545849325,
      "step": 3910
    },
    {
      "epoch": 0.392,
      "grad_norm": 40.25,
      "grad_norm_var": 11.49140625,
      "learning_rate": 0.0001,
      "loss": 8.5002,
      "loss/crossentropy": 2.347226142883301,
      "loss/hidden": 3.658203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24455956518650054,
      "step": 3920
    },
    {
      "epoch": 0.393,
      "grad_norm": 31.125,
      "grad_norm_var": 17.219791666666666,
      "learning_rate": 0.0001,
      "loss": 8.4054,
      "loss/crossentropy": 2.2351839393377304,
      "loss/hidden": 3.59140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22245508581399917,
      "step": 3930
    },
    {
      "epoch": 0.394,
      "grad_norm": 37.0,
      "grad_norm_var": 11.382747395833333,
      "learning_rate": 0.0001,
      "loss": 8.4771,
      "loss/crossentropy": 2.4160192787647246,
      "loss/hidden": 3.700390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2585303969681263,
      "step": 3940
    },
    {
      "epoch": 0.395,
      "grad_norm": 30.875,
      "grad_norm_var": 10.731705729166666,
      "learning_rate": 0.0001,
      "loss": 8.4064,
      "loss/crossentropy": 2.329276342689991,
      "loss/hidden": 3.641796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23809341453015803,
      "step": 3950
    },
    {
      "epoch": 0.396,
      "grad_norm": 31.875,
      "grad_norm_var": 18.456184895833335,
      "learning_rate": 0.0001,
      "loss": 8.4948,
      "loss/crossentropy": 2.397703355550766,
      "loss/hidden": 3.8703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2577156092971563,
      "step": 3960
    },
    {
      "epoch": 0.397,
      "grad_norm": 35.25,
      "grad_norm_var": 16.1291015625,
      "learning_rate": 0.0001,
      "loss": 8.5021,
      "loss/crossentropy": 2.3261855766177177,
      "loss/hidden": 3.694921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2523797513917089,
      "step": 3970
    },
    {
      "epoch": 0.398,
      "grad_norm": 33.0,
      "grad_norm_var": 8.205208333333333,
      "learning_rate": 0.0001,
      "loss": 8.4531,
      "loss/crossentropy": 2.235724928975105,
      "loss/hidden": 3.7296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2461556438356638,
      "step": 3980
    },
    {
      "epoch": 0.399,
      "grad_norm": 35.5,
      "grad_norm_var": 7.25390625,
      "learning_rate": 0.0001,
      "loss": 8.3779,
      "loss/crossentropy": 2.258603112399578,
      "loss/hidden": 3.731640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2523756165057421,
      "step": 3990
    },
    {
      "epoch": 0.4,
      "grad_norm": 32.5,
      "grad_norm_var": 11.34765625,
      "learning_rate": 0.0001,
      "loss": 8.542,
      "loss/crossentropy": 2.2897588342428206,
      "loss/hidden": 3.6453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2516363400965929,
      "step": 4000
    },
    {
      "epoch": 0.401,
      "grad_norm": 32.0,
      "grad_norm_var": 9.889518229166667,
      "learning_rate": 0.0001,
      "loss": 8.4234,
      "loss/crossentropy": 2.1721622362732886,
      "loss/hidden": 3.74609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.25033344645053146,
      "step": 4010
    },
    {
      "epoch": 0.402,
      "grad_norm": 35.75,
      "grad_norm_var": 15.199934895833334,
      "learning_rate": 0.0001,
      "loss": 8.4018,
      "loss/crossentropy": 2.1795839801430703,
      "loss/hidden": 3.78515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24097833968698978,
      "step": 4020
    },
    {
      "epoch": 0.403,
      "grad_norm": 37.75,
      "grad_norm_var": 162.76451822916667,
      "learning_rate": 0.0001,
      "loss": 8.4738,
      "loss/crossentropy": 2.4760424941778183,
      "loss/hidden": 3.86484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.3052004296332598,
      "step": 4030
    },
    {
      "epoch": 0.404,
      "grad_norm": 41.5,
      "grad_norm_var": 2.408370239548424e+18,
      "learning_rate": 0.0001,
      "loss": 8.4302,
      "loss/crossentropy": 2.1855442106723784,
      "loss/hidden": 3.77578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24037305619567634,
      "step": 4040
    },
    {
      "epoch": 0.405,
      "grad_norm": 32.0,
      "grad_norm_var": 2.4083702412167086e+18,
      "learning_rate": 0.0001,
      "loss": 8.3213,
      "loss/crossentropy": 2.22409378439188,
      "loss/hidden": 3.653515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23280739206820728,
      "step": 4050
    },
    {
      "epoch": 0.406,
      "grad_norm": 39.5,
      "grad_norm_var": 9.825,
      "learning_rate": 0.0001,
      "loss": 8.4657,
      "loss/crossentropy": 2.28346493691206,
      "loss/hidden": 3.773828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2542352583259344,
      "step": 4060
    },
    {
      "epoch": 0.407,
      "grad_norm": 39.75,
      "grad_norm_var": 7.640625,
      "learning_rate": 0.0001,
      "loss": 8.3937,
      "loss/crossentropy": 2.2639666229486464,
      "loss/hidden": 3.851953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27007580138742926,
      "step": 4070
    },
    {
      "epoch": 0.408,
      "grad_norm": 37.25,
      "grad_norm_var": 10.493684895833333,
      "learning_rate": 0.0001,
      "loss": 8.468,
      "loss/crossentropy": 2.356726923584938,
      "loss/hidden": 3.716796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.242316972091794,
      "step": 4080
    },
    {
      "epoch": 0.409,
      "grad_norm": 34.5,
      "grad_norm_var": 8.8462890625,
      "learning_rate": 0.0001,
      "loss": 8.3684,
      "loss/crossentropy": 2.151221239566803,
      "loss/hidden": 3.692578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2382324907928705,
      "step": 4090
    },
    {
      "epoch": 0.41,
      "grad_norm": 37.0,
      "grad_norm_var": 13.3775390625,
      "learning_rate": 0.0001,
      "loss": 8.296,
      "loss/crossentropy": 2.1645509719848635,
      "loss/hidden": 3.65703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2371783286333084,
      "step": 4100
    },
    {
      "epoch": 0.411,
      "grad_norm": 32.5,
      "grad_norm_var": 25.270247395833334,
      "learning_rate": 0.0001,
      "loss": 8.3435,
      "loss/crossentropy": 2.1995768398046494,
      "loss/hidden": 3.6640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23499403558671475,
      "step": 4110
    },
    {
      "epoch": 0.412,
      "grad_norm": 38.0,
      "grad_norm_var": 13.875,
      "learning_rate": 0.0001,
      "loss": 8.3746,
      "loss/crossentropy": 2.135378623008728,
      "loss/hidden": 3.612109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24154506418854,
      "step": 4120
    },
    {
      "epoch": 0.413,
      "grad_norm": 40.25,
      "grad_norm_var": 12.233072916666666,
      "learning_rate": 0.0001,
      "loss": 8.3505,
      "loss/crossentropy": 2.356252074241638,
      "loss/hidden": 3.734765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2688302733004093,
      "step": 4130
    },
    {
      "epoch": 0.414,
      "grad_norm": 39.25,
      "grad_norm_var": 14.333072916666667,
      "learning_rate": 0.0001,
      "loss": 8.4228,
      "loss/crossentropy": 2.334370291233063,
      "loss/hidden": 3.648046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23012813031673432,
      "step": 4140
    },
    {
      "epoch": 0.415,
      "grad_norm": 35.5,
      "grad_norm_var": 8.170833333333333,
      "learning_rate": 0.0001,
      "loss": 8.3324,
      "loss/crossentropy": 2.214457754790783,
      "loss/hidden": 3.651171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24744220934808253,
      "step": 4150
    },
    {
      "epoch": 0.416,
      "grad_norm": 37.0,
      "grad_norm_var": 7.603580729166667,
      "learning_rate": 0.0001,
      "loss": 8.3237,
      "loss/crossentropy": 2.3053094416856768,
      "loss/hidden": 3.62578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2290981512516737,
      "step": 4160
    },
    {
      "epoch": 0.417,
      "grad_norm": 50.75,
      "grad_norm_var": 23.6931640625,
      "learning_rate": 0.0001,
      "loss": 8.2895,
      "loss/crossentropy": 2.100640784204006,
      "loss/hidden": 3.72578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24771791882812977,
      "step": 4170
    },
    {
      "epoch": 0.418,
      "grad_norm": 42.25,
      "grad_norm_var": 25.6509765625,
      "learning_rate": 0.0001,
      "loss": 8.4997,
      "loss/crossentropy": 2.4176384449005126,
      "loss/hidden": 3.734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2665034931153059,
      "step": 4180
    },
    {
      "epoch": 0.419,
      "grad_norm": 46.75,
      "grad_norm_var": 16.7041015625,
      "learning_rate": 0.0001,
      "loss": 8.241,
      "loss/crossentropy": 2.0916571110486983,
      "loss/hidden": 3.640234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23109357040375472,
      "step": 4190
    },
    {
      "epoch": 0.42,
      "grad_norm": 34.0,
      "grad_norm_var": 10.265625,
      "learning_rate": 0.0001,
      "loss": 8.3903,
      "loss/crossentropy": 2.4284214213490487,
      "loss/hidden": 3.676171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24520040042698382,
      "step": 4200
    },
    {
      "epoch": 0.421,
      "grad_norm": 48.25,
      "grad_norm_var": 17.694205729166665,
      "learning_rate": 0.0001,
      "loss": 8.4379,
      "loss/crossentropy": 2.342070159316063,
      "loss/hidden": 3.750390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2608378600329161,
      "step": 4210
    },
    {
      "epoch": 0.422,
      "grad_norm": 31.875,
      "grad_norm_var": 21.2806640625,
      "learning_rate": 0.0001,
      "loss": 8.2667,
      "loss/crossentropy": 2.3339773267507553,
      "loss/hidden": 3.68828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24873733669519424,
      "step": 4220
    },
    {
      "epoch": 0.423,
      "grad_norm": 34.0,
      "grad_norm_var": 10.834375,
      "learning_rate": 0.0001,
      "loss": 8.313,
      "loss/crossentropy": 2.2396500378847124,
      "loss/hidden": 3.647265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23453602455556394,
      "step": 4230
    },
    {
      "epoch": 0.424,
      "grad_norm": 31.875,
      "grad_norm_var": 20.712239583333332,
      "learning_rate": 0.0001,
      "loss": 8.3885,
      "loss/crossentropy": 2.1176706120371818,
      "loss/hidden": 3.818359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24183569326996804,
      "step": 4240
    },
    {
      "epoch": 0.425,
      "grad_norm": 46.0,
      "grad_norm_var": 18.212434895833333,
      "learning_rate": 0.0001,
      "loss": 8.4311,
      "loss/crossentropy": 2.161303213238716,
      "loss/hidden": 3.71015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23596356846392155,
      "step": 4250
    },
    {
      "epoch": 0.426,
      "grad_norm": 31.75,
      "grad_norm_var": 18.470833333333335,
      "learning_rate": 0.0001,
      "loss": 8.389,
      "loss/crossentropy": 2.1959901452064514,
      "loss/hidden": 3.683203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23274125456809996,
      "step": 4260
    },
    {
      "epoch": 0.427,
      "grad_norm": 40.0,
      "grad_norm_var": 15.642643229166667,
      "learning_rate": 0.0001,
      "loss": 8.3995,
      "loss/crossentropy": 2.3957006752491,
      "loss/hidden": 3.6546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23920847922563554,
      "step": 4270
    },
    {
      "epoch": 0.428,
      "grad_norm": 32.25,
      "grad_norm_var": 12.506705729166667,
      "learning_rate": 0.0001,
      "loss": 8.3247,
      "loss/crossentropy": 2.140459132194519,
      "loss/hidden": 3.776953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24535099379718303,
      "step": 4280
    },
    {
      "epoch": 0.429,
      "grad_norm": 31.125,
      "grad_norm_var": 14.913997395833333,
      "learning_rate": 0.0001,
      "loss": 8.2296,
      "loss/crossentropy": 2.330910986661911,
      "loss/hidden": 3.65546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.25125612393021585,
      "step": 4290
    },
    {
      "epoch": 0.43,
      "grad_norm": 31.0,
      "grad_norm_var": 9.062955729166667,
      "learning_rate": 0.0001,
      "loss": 8.2789,
      "loss/crossentropy": 2.21281051337719,
      "loss/hidden": 3.717578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24373065643012523,
      "step": 4300
    },
    {
      "epoch": 0.431,
      "grad_norm": 46.25,
      "grad_norm_var": 18.794205729166666,
      "learning_rate": 0.0001,
      "loss": 8.283,
      "loss/crossentropy": 2.182158187031746,
      "loss/hidden": 3.6703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24544784277677537,
      "step": 4310
    },
    {
      "epoch": 0.432,
      "grad_norm": 45.75,
      "grad_norm_var": 38.16458333333333,
      "learning_rate": 0.0001,
      "loss": 8.2852,
      "loss/crossentropy": 2.365989252924919,
      "loss/hidden": 3.706640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24662891514599322,
      "step": 4320
    },
    {
      "epoch": 0.433,
      "grad_norm": 35.25,
      "grad_norm_var": 16.307291666666668,
      "learning_rate": 0.0001,
      "loss": 8.2794,
      "loss/crossentropy": 2.1546150177717207,
      "loss/hidden": 3.620703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2297368910163641,
      "step": 4330
    },
    {
      "epoch": 0.434,
      "grad_norm": 31.625,
      "grad_norm_var": 19.6400390625,
      "learning_rate": 0.0001,
      "loss": 8.3956,
      "loss/crossentropy": 2.385764144361019,
      "loss/hidden": 3.7015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24324760176241397,
      "step": 4340
    },
    {
      "epoch": 0.435,
      "grad_norm": 33.0,
      "grad_norm_var": 9.987239583333333,
      "learning_rate": 0.0001,
      "loss": 8.2695,
      "loss/crossentropy": 2.112358179688454,
      "loss/hidden": 3.788671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24638627246022224,
      "step": 4350
    },
    {
      "epoch": 0.436,
      "grad_norm": 34.5,
      "grad_norm_var": 8.0875,
      "learning_rate": 0.0001,
      "loss": 8.448,
      "loss/crossentropy": 2.182591002434492,
      "loss/hidden": 3.746875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24270438468083738,
      "step": 4360
    },
    {
      "epoch": 0.437,
      "grad_norm": 38.25,
      "grad_norm_var": 5.695833333333334,
      "learning_rate": 0.0001,
      "loss": 8.3975,
      "loss/crossentropy": 2.401635229587555,
      "loss/hidden": 3.65859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24400906264781952,
      "step": 4370
    },
    {
      "epoch": 0.438,
      "grad_norm": 34.75,
      "grad_norm_var": 8.372330729166666,
      "learning_rate": 0.0001,
      "loss": 8.3822,
      "loss/crossentropy": 2.3247624695301057,
      "loss/hidden": 3.591015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24138722717761993,
      "step": 4380
    },
    {
      "epoch": 0.439,
      "grad_norm": 62.75,
      "grad_norm_var": 70.6103515625,
      "learning_rate": 0.0001,
      "loss": 8.3893,
      "loss/crossentropy": 2.1605212301015855,
      "loss/hidden": 3.75,
      "loss/jsd": 0.0,
      "loss/logits": 0.270247707888484,
      "step": 4390
    },
    {
      "epoch": 0.44,
      "grad_norm": 34.5,
      "grad_norm_var": 80.6375,
      "learning_rate": 0.0001,
      "loss": 8.3494,
      "loss/crossentropy": 2.201382315158844,
      "loss/hidden": 3.55390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22101359032094478,
      "step": 4400
    },
    {
      "epoch": 0.441,
      "grad_norm": 37.75,
      "grad_norm_var": 12.06875,
      "learning_rate": 0.0001,
      "loss": 8.3221,
      "loss/crossentropy": 2.1608468025922773,
      "loss/hidden": 3.570703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22636283356696368,
      "step": 4410
    },
    {
      "epoch": 0.442,
      "grad_norm": 32.25,
      "grad_norm_var": 16.13515625,
      "learning_rate": 0.0001,
      "loss": 8.2013,
      "loss/crossentropy": 2.2836680516600607,
      "loss/hidden": 3.6765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24737481120973825,
      "step": 4420
    },
    {
      "epoch": 0.443,
      "grad_norm": 34.5,
      "grad_norm_var": 14.570768229166667,
      "learning_rate": 0.0001,
      "loss": 8.3241,
      "loss/crossentropy": 2.1778072111308573,
      "loss/hidden": 3.685546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23579915445297955,
      "step": 4430
    },
    {
      "epoch": 0.444,
      "grad_norm": 33.5,
      "grad_norm_var": 14.844205729166667,
      "learning_rate": 0.0001,
      "loss": 8.2468,
      "loss/crossentropy": 2.115637184679508,
      "loss/hidden": 3.54921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22633790075778962,
      "step": 4440
    },
    {
      "epoch": 0.445,
      "grad_norm": 40.0,
      "grad_norm_var": 8.757291666666667,
      "learning_rate": 0.0001,
      "loss": 8.3095,
      "loss/crossentropy": 2.2457625687122347,
      "loss/hidden": 3.678125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24750035293400288,
      "step": 4450
    },
    {
      "epoch": 0.446,
      "grad_norm": 30.75,
      "grad_norm_var": 9.359309895833333,
      "learning_rate": 0.0001,
      "loss": 8.3604,
      "loss/crossentropy": 2.2244989693164827,
      "loss/hidden": 3.742578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.27126055024564266,
      "step": 4460
    },
    {
      "epoch": 0.447,
      "grad_norm": 32.75,
      "grad_norm_var": 8.16015625,
      "learning_rate": 0.0001,
      "loss": 8.293,
      "loss/crossentropy": 2.208540087938309,
      "loss/hidden": 3.591796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23848242741078138,
      "step": 4470
    },
    {
      "epoch": 0.448,
      "grad_norm": 31.5,
      "grad_norm_var": 27.215625,
      "learning_rate": 0.0001,
      "loss": 8.4332,
      "loss/crossentropy": 2.182788160443306,
      "loss/hidden": 3.775,
      "loss/jsd": 0.0,
      "loss/logits": 0.26532087065279486,
      "step": 4480
    },
    {
      "epoch": 0.449,
      "grad_norm": 39.5,
      "grad_norm_var": 29.916666666666668,
      "learning_rate": 0.0001,
      "loss": 8.2552,
      "loss/crossentropy": 2.172296644747257,
      "loss/hidden": 3.66875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23516011722385882,
      "step": 4490
    },
    {
      "epoch": 0.45,
      "grad_norm": 33.75,
      "grad_norm_var": 21.775455729166666,
      "learning_rate": 0.0001,
      "loss": 8.3872,
      "loss/crossentropy": 2.304448103904724,
      "loss/hidden": 3.56015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23418739810585976,
      "step": 4500
    },
    {
      "epoch": 0.451,
      "grad_norm": 31.75,
      "grad_norm_var": 23.40390625,
      "learning_rate": 0.0001,
      "loss": 8.4055,
      "loss/crossentropy": 2.2493974685668947,
      "loss/hidden": 3.652734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2410556711256504,
      "step": 4510
    },
    {
      "epoch": 0.452,
      "grad_norm": 32.25,
      "grad_norm_var": 19.04375,
      "learning_rate": 0.0001,
      "loss": 8.2847,
      "loss/crossentropy": 2.0756215125322344,
      "loss/hidden": 3.658984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23735107891261578,
      "step": 4520
    },
    {
      "epoch": 0.453,
      "grad_norm": 38.5,
      "grad_norm_var": 11.048958333333333,
      "learning_rate": 0.0001,
      "loss": 8.4058,
      "loss/crossentropy": 2.2705332577228545,
      "loss/hidden": 3.75703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2552911601960659,
      "step": 4530
    },
    {
      "epoch": 0.454,
      "grad_norm": 32.25,
      "grad_norm_var": 15.964583333333334,
      "learning_rate": 0.0001,
      "loss": 8.3352,
      "loss/crossentropy": 2.2062640622258187,
      "loss/hidden": 3.663671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23420217223465442,
      "step": 4540
    },
    {
      "epoch": 0.455,
      "grad_norm": 35.5,
      "grad_norm_var": 18.8041015625,
      "learning_rate": 0.0001,
      "loss": 8.1398,
      "loss/crossentropy": 2.2153613708913324,
      "loss/hidden": 3.688671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23869724282994867,
      "step": 4550
    },
    {
      "epoch": 0.456,
      "grad_norm": 32.25,
      "grad_norm_var": 15.843489583333334,
      "learning_rate": 0.0001,
      "loss": 8.3482,
      "loss/crossentropy": 2.2614838272333144,
      "loss/hidden": 3.719921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2509554075077176,
      "step": 4560
    },
    {
      "epoch": 0.457,
      "grad_norm": 37.0,
      "grad_norm_var": 17.045572916666668,
      "learning_rate": 0.0001,
      "loss": 8.3724,
      "loss/crossentropy": 2.1315455704927446,
      "loss/hidden": 3.720703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24938025698065758,
      "step": 4570
    },
    {
      "epoch": 0.458,
      "grad_norm": 31.75,
      "grad_norm_var": 18.753580729166668,
      "learning_rate": 0.0001,
      "loss": 8.1789,
      "loss/crossentropy": 2.250936383008957,
      "loss/hidden": 3.601953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23103775745257735,
      "step": 4580
    },
    {
      "epoch": 0.459,
      "grad_norm": 31.75,
      "grad_norm_var": 9.3087890625,
      "learning_rate": 0.0001,
      "loss": 8.2367,
      "loss/crossentropy": 2.0491979137063026,
      "loss/hidden": 3.767578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24958254247903824,
      "step": 4590
    },
    {
      "epoch": 0.46,
      "grad_norm": 32.75,
      "grad_norm_var": 6.034309895833333,
      "learning_rate": 0.0001,
      "loss": 8.2813,
      "loss/crossentropy": 2.1869849786162376,
      "loss/hidden": 3.678125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2445020995102823,
      "step": 4600
    },
    {
      "epoch": 0.461,
      "grad_norm": 33.25,
      "grad_norm_var": 16.960416666666667,
      "learning_rate": 0.0001,
      "loss": 8.2774,
      "loss/crossentropy": 2.3401444420218467,
      "loss/hidden": 3.644140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2346777945756912,
      "step": 4610
    },
    {
      "epoch": 0.462,
      "grad_norm": 31.875,
      "grad_norm_var": 20.257747395833334,
      "learning_rate": 0.0001,
      "loss": 8.261,
      "loss/crossentropy": 2.2197701543569566,
      "loss/hidden": 3.671484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24417277611792088,
      "step": 4620
    },
    {
      "epoch": 0.463,
      "grad_norm": 39.75,
      "grad_norm_var": 21.111458333333335,
      "learning_rate": 0.0001,
      "loss": 8.1879,
      "loss/crossentropy": 2.2979017451405523,
      "loss/hidden": 3.584765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24459463655948638,
      "step": 4630
    },
    {
      "epoch": 0.464,
      "grad_norm": 37.75,
      "grad_norm_var": 18.424739583333334,
      "learning_rate": 0.0001,
      "loss": 8.2222,
      "loss/crossentropy": 2.1899428203701974,
      "loss/hidden": 3.63828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23625375218689443,
      "step": 4640
    },
    {
      "epoch": 0.465,
      "grad_norm": 33.0,
      "grad_norm_var": 5.676822916666667,
      "learning_rate": 0.0001,
      "loss": 8.2134,
      "loss/crossentropy": 2.173795387148857,
      "loss/hidden": 3.653515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21926050689071416,
      "step": 4650
    },
    {
      "epoch": 0.466,
      "grad_norm": 32.75,
      "grad_norm_var": 5.024739583333333,
      "learning_rate": 0.0001,
      "loss": 8.2134,
      "loss/crossentropy": 2.2807129830121995,
      "loss/hidden": 3.5890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22838456649333239,
      "step": 4660
    },
    {
      "epoch": 0.467,
      "grad_norm": 32.75,
      "grad_norm_var": 7.588541666666667,
      "learning_rate": 0.0001,
      "loss": 8.2508,
      "loss/crossentropy": 2.19968124628067,
      "loss/hidden": 3.590625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23003219701349736,
      "step": 4670
    },
    {
      "epoch": 0.468,
      "grad_norm": 32.25,
      "grad_norm_var": 8.268489583333333,
      "learning_rate": 0.0001,
      "loss": 8.2723,
      "loss/crossentropy": 2.3073908984661102,
      "loss/hidden": 3.621484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23195548579096795,
      "step": 4680
    },
    {
      "epoch": 0.469,
      "grad_norm": 34.25,
      "grad_norm_var": 7.398372395833333,
      "learning_rate": 0.0001,
      "loss": 8.3115,
      "loss/crossentropy": 2.251572087407112,
      "loss/hidden": 3.615234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24007561076432465,
      "step": 4690
    },
    {
      "epoch": 0.47,
      "grad_norm": 37.25,
      "grad_norm_var": 12.067708333333334,
      "learning_rate": 0.0001,
      "loss": 8.215,
      "loss/crossentropy": 2.179315264523029,
      "loss/hidden": 3.614453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23048642594367266,
      "step": 4700
    },
    {
      "epoch": 0.471,
      "grad_norm": 33.5,
      "grad_norm_var": 9.384830729166667,
      "learning_rate": 0.0001,
      "loss": 8.4657,
      "loss/crossentropy": 2.304041627049446,
      "loss/hidden": 3.705859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2532807156443596,
      "step": 4710
    },
    {
      "epoch": 0.472,
      "grad_norm": 34.75,
      "grad_norm_var": 2.5403116373864177e+18,
      "learning_rate": 0.0001,
      "loss": 8.3845,
      "loss/crossentropy": 2.28207755535841,
      "loss/hidden": 3.6125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24173217974603176,
      "step": 4720
    },
    {
      "epoch": 0.473,
      "grad_norm": 31.375,
      "grad_norm_var": 2.8580729166666665,
      "learning_rate": 0.0001,
      "loss": 8.1924,
      "loss/crossentropy": 2.317093315720558,
      "loss/hidden": 3.50859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22850329093635083,
      "step": 4730
    },
    {
      "epoch": 0.474,
      "grad_norm": 38.5,
      "grad_norm_var": 55.86354166666667,
      "learning_rate": 0.0001,
      "loss": 8.205,
      "loss/crossentropy": 2.2346624046564103,
      "loss/hidden": 3.76015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2535475058481097,
      "step": 4740
    },
    {
      "epoch": 0.475,
      "grad_norm": 32.75,
      "grad_norm_var": 6.105989583333334,
      "learning_rate": 0.0001,
      "loss": 8.2004,
      "loss/crossentropy": 2.205992843210697,
      "loss/hidden": 3.5,
      "loss/jsd": 0.0,
      "loss/logits": 0.2157002430409193,
      "step": 4750
    },
    {
      "epoch": 0.476,
      "grad_norm": 42.5,
      "grad_norm_var": 7.551041666666666,
      "learning_rate": 0.0001,
      "loss": 8.2493,
      "loss/crossentropy": 2.2944780766963957,
      "loss/hidden": 3.553515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2302501540631056,
      "step": 4760
    },
    {
      "epoch": 0.477,
      "grad_norm": 34.5,
      "grad_norm_var": 54.297916666666666,
      "learning_rate": 0.0001,
      "loss": 8.2422,
      "loss/crossentropy": 2.349091801047325,
      "loss/hidden": 3.726953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25370817482471464,
      "step": 4770
    },
    {
      "epoch": 0.478,
      "grad_norm": 35.25,
      "grad_norm_var": 48.0119140625,
      "learning_rate": 0.0001,
      "loss": 8.1697,
      "loss/crossentropy": 2.419427090883255,
      "loss/hidden": 3.5984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23952311277389526,
      "step": 4780
    },
    {
      "epoch": 0.479,
      "grad_norm": 44.0,
      "grad_norm_var": 34.6509765625,
      "learning_rate": 0.0001,
      "loss": 8.2147,
      "loss/crossentropy": 2.2663041442632674,
      "loss/hidden": 3.512890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22062067724764348,
      "step": 4790
    },
    {
      "epoch": 0.48,
      "grad_norm": 30.375,
      "grad_norm_var": 38.708333333333336,
      "learning_rate": 0.0001,
      "loss": 8.1478,
      "loss/crossentropy": 2.3492169111967085,
      "loss/hidden": 3.63359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23971957936882973,
      "step": 4800
    },
    {
      "epoch": 0.481,
      "grad_norm": 36.75,
      "grad_norm_var": 9.3056640625,
      "learning_rate": 0.0001,
      "loss": 8.165,
      "loss/crossentropy": 2.232720893621445,
      "loss/hidden": 3.580078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22206582501530647,
      "step": 4810
    },
    {
      "epoch": 0.482,
      "grad_norm": 34.25,
      "grad_norm_var": 6.8197265625,
      "learning_rate": 0.0001,
      "loss": 8.1216,
      "loss/crossentropy": 2.1783568069338797,
      "loss/hidden": 3.4515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21557580903172494,
      "step": 4820
    },
    {
      "epoch": 0.483,
      "grad_norm": 31.5,
      "grad_norm_var": 6.039322916666666,
      "learning_rate": 0.0001,
      "loss": 8.2857,
      "loss/crossentropy": 2.3961785644292832,
      "loss/hidden": 3.640234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.257023797929287,
      "step": 4830
    },
    {
      "epoch": 0.484,
      "grad_norm": 33.0,
      "grad_norm_var": 12.29375,
      "learning_rate": 0.0001,
      "loss": 8.1802,
      "loss/crossentropy": 2.310921123623848,
      "loss/hidden": 3.526953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23252013735473157,
      "step": 4840
    },
    {
      "epoch": 0.485,
      "grad_norm": 33.25,
      "grad_norm_var": 23.69140625,
      "learning_rate": 0.0001,
      "loss": 8.2647,
      "loss/crossentropy": 2.152234472334385,
      "loss/hidden": 3.6578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2241989640519023,
      "step": 4850
    },
    {
      "epoch": 0.486,
      "grad_norm": 32.75,
      "grad_norm_var": 26.65390625,
      "learning_rate": 0.0001,
      "loss": 8.2819,
      "loss/crossentropy": 2.1197937928140163,
      "loss/hidden": 3.702734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24686675220727922,
      "step": 4860
    },
    {
      "epoch": 0.487,
      "grad_norm": 37.25,
      "grad_norm_var": 13.0228515625,
      "learning_rate": 0.0001,
      "loss": 8.3156,
      "loss/crossentropy": 2.345510223507881,
      "loss/hidden": 3.571484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22903156131505967,
      "step": 4870
    },
    {
      "epoch": 0.488,
      "grad_norm": 34.5,
      "grad_norm_var": 6.5337890625,
      "learning_rate": 0.0001,
      "loss": 8.1946,
      "loss/crossentropy": 2.1908657550811768,
      "loss/hidden": 3.682421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23578502163290976,
      "step": 4880
    },
    {
      "epoch": 0.489,
      "grad_norm": 33.75,
      "grad_norm_var": 24.545572916666668,
      "learning_rate": 0.0001,
      "loss": 8.1165,
      "loss/crossentropy": 2.180183355510235,
      "loss/hidden": 3.661328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23667961843311786,
      "step": 4890
    },
    {
      "epoch": 0.49,
      "grad_norm": 32.75,
      "grad_norm_var": 58.805989583333336,
      "learning_rate": 0.0001,
      "loss": 8.127,
      "loss/crossentropy": 2.206071509420872,
      "loss/hidden": 3.51796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21433540284633637,
      "step": 4900
    },
    {
      "epoch": 0.491,
      "grad_norm": 38.0,
      "grad_norm_var": 50.01223958333333,
      "learning_rate": 0.0001,
      "loss": 8.1544,
      "loss/crossentropy": 2.1096328511834144,
      "loss/hidden": 3.61484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24193457532674073,
      "step": 4910
    },
    {
      "epoch": 0.492,
      "grad_norm": 39.25,
      "grad_norm_var": 17.08125,
      "learning_rate": 0.0001,
      "loss": 8.202,
      "loss/crossentropy": 2.2372330710291863,
      "loss/hidden": 3.63203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23599924352020024,
      "step": 4920
    },
    {
      "epoch": 0.493,
      "grad_norm": 33.25,
      "grad_norm_var": 12.742643229166667,
      "learning_rate": 0.0001,
      "loss": 8.1759,
      "loss/crossentropy": 2.196950948238373,
      "loss/hidden": 3.638671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24246960394084455,
      "step": 4930
    },
    {
      "epoch": 0.494,
      "grad_norm": 40.5,
      "grad_norm_var": 6.883072916666666,
      "learning_rate": 0.0001,
      "loss": 8.1278,
      "loss/crossentropy": 2.3560511782765388,
      "loss/hidden": 3.633984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2382162045687437,
      "step": 4940
    },
    {
      "epoch": 0.495,
      "grad_norm": 35.0,
      "grad_norm_var": 11.6603515625,
      "learning_rate": 0.0001,
      "loss": 8.2982,
      "loss/crossentropy": 2.3035822331905367,
      "loss/hidden": 3.64140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2485156562179327,
      "step": 4950
    },
    {
      "epoch": 0.496,
      "grad_norm": 34.25,
      "grad_norm_var": 14.192708333333334,
      "learning_rate": 0.0001,
      "loss": 8.1579,
      "loss/crossentropy": 2.184766189754009,
      "loss/hidden": 3.637109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23721186630427837,
      "step": 4960
    },
    {
      "epoch": 0.497,
      "grad_norm": 33.5,
      "grad_norm_var": 13.871809895833334,
      "learning_rate": 0.0001,
      "loss": 8.1975,
      "loss/crossentropy": 2.1983452700078487,
      "loss/hidden": 3.697265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24081590361893176,
      "step": 4970
    },
    {
      "epoch": 0.498,
      "grad_norm": 43.25,
      "grad_norm_var": 13.948372395833333,
      "learning_rate": 0.0001,
      "loss": 8.2262,
      "loss/crossentropy": 2.3249034196138383,
      "loss/hidden": 3.56875,
      "loss/jsd": 0.0,
      "loss/logits": 0.234759721159935,
      "step": 4980
    },
    {
      "epoch": 0.499,
      "grad_norm": 31.5,
      "grad_norm_var": 11.167708333333334,
      "learning_rate": 0.0001,
      "loss": 8.1488,
      "loss/crossentropy": 2.2298896074295045,
      "loss/hidden": 3.6546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2505023546516895,
      "step": 4990
    },
    {
      "epoch": 0.5,
      "grad_norm": 38.75,
      "grad_norm_var": 10.43125,
      "learning_rate": 0.0001,
      "loss": 8.1769,
      "loss/crossentropy": 2.201507803052664,
      "loss/hidden": 3.520703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21926793903112413,
      "step": 5000
    },
    {
      "epoch": 0.501,
      "grad_norm": 44.75,
      "grad_norm_var": 16.2853515625,
      "learning_rate": 0.0001,
      "loss": 8.039,
      "loss/crossentropy": 2.270238833874464,
      "loss/hidden": 3.613671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23682384472340345,
      "step": 5010
    },
    {
      "epoch": 0.502,
      "grad_norm": 31.75,
      "grad_norm_var": 14.38515625,
      "learning_rate": 0.0001,
      "loss": 8.1599,
      "loss/crossentropy": 2.108400362730026,
      "loss/hidden": 3.718359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2417622933164239,
      "step": 5020
    },
    {
      "epoch": 0.503,
      "grad_norm": 33.0,
      "grad_norm_var": 2.8989583333333333,
      "learning_rate": 0.0001,
      "loss": 8.1698,
      "loss/crossentropy": 2.3586475804448126,
      "loss/hidden": 3.571875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23084877729415892,
      "step": 5030
    },
    {
      "epoch": 0.504,
      "grad_norm": 32.5,
      "grad_norm_var": 1.962513882217063e+18,
      "learning_rate": 0.0001,
      "loss": 8.293,
      "loss/crossentropy": 2.308009374141693,
      "loss/hidden": 3.631640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.25030199717730284,
      "step": 5040
    },
    {
      "epoch": 0.505,
      "grad_norm": 34.25,
      "grad_norm_var": 1.962513880139065e+18,
      "learning_rate": 0.0001,
      "loss": 8.2269,
      "loss/crossentropy": 2.2477807879447935,
      "loss/hidden": 3.63046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22722288742661476,
      "step": 5050
    },
    {
      "epoch": 0.506,
      "grad_norm": 31.0,
      "grad_norm_var": 230.7962890625,
      "learning_rate": 0.0001,
      "loss": 8.0995,
      "loss/crossentropy": 2.113031893968582,
      "loss/hidden": 3.625390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2260434988886118,
      "step": 5060
    },
    {
      "epoch": 0.507,
      "grad_norm": 29.5,
      "grad_norm_var": 226.95462239583333,
      "learning_rate": 0.0001,
      "loss": 8.1254,
      "loss/crossentropy": 2.1494341671466826,
      "loss/hidden": 3.628125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2316014662384987,
      "step": 5070
    },
    {
      "epoch": 0.508,
      "grad_norm": 42.25,
      "grad_norm_var": 17.206184895833335,
      "learning_rate": 0.0001,
      "loss": 8.3222,
      "loss/crossentropy": 2.409487584233284,
      "loss/hidden": 3.51953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24138148501515388,
      "step": 5080
    },
    {
      "epoch": 0.509,
      "grad_norm": 33.0,
      "grad_norm_var": 15.231184895833334,
      "learning_rate": 0.0001,
      "loss": 8.0576,
      "loss/crossentropy": 2.285913223773241,
      "loss/hidden": 3.528515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2232502717524767,
      "step": 5090
    },
    {
      "epoch": 0.51,
      "grad_norm": 33.5,
      "grad_norm_var": 15.065559895833333,
      "learning_rate": 0.0001,
      "loss": 8.1747,
      "loss/crossentropy": 2.180917738378048,
      "loss/hidden": 3.56328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23142165634781123,
      "step": 5100
    },
    {
      "epoch": 0.511,
      "grad_norm": 32.25,
      "grad_norm_var": 13.1166015625,
      "learning_rate": 0.0001,
      "loss": 8.1417,
      "loss/crossentropy": 2.226011593639851,
      "loss/hidden": 3.506640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22545368764549495,
      "step": 5110
    },
    {
      "epoch": 0.512,
      "grad_norm": 33.25,
      "grad_norm_var": 4.7572265625,
      "learning_rate": 0.0001,
      "loss": 8.1568,
      "loss/crossentropy": 2.2642487674951552,
      "loss/hidden": 3.626953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24627051521092652,
      "step": 5120
    },
    {
      "epoch": 0.513,
      "grad_norm": 29.375,
      "grad_norm_var": 8.576822916666666,
      "learning_rate": 0.0001,
      "loss": 8.1412,
      "loss/crossentropy": 2.1210575878620146,
      "loss/hidden": 3.603125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22232303582131863,
      "step": 5130
    },
    {
      "epoch": 0.514,
      "grad_norm": 36.25,
      "grad_norm_var": 7.51015625,
      "learning_rate": 0.0001,
      "loss": 8.1105,
      "loss/crossentropy": 2.338705539703369,
      "loss/hidden": 3.61796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2322886861860752,
      "step": 5140
    },
    {
      "epoch": 0.515,
      "grad_norm": 32.25,
      "grad_norm_var": 9.846809895833333,
      "learning_rate": 0.0001,
      "loss": 8.1908,
      "loss/crossentropy": 2.2111464768648146,
      "loss/hidden": 3.72109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.26425624899566175,
      "step": 5150
    },
    {
      "epoch": 0.516,
      "grad_norm": 28.875,
      "grad_norm_var": 9.946809895833333,
      "learning_rate": 0.0001,
      "loss": 8.0375,
      "loss/crossentropy": 2.176833947002888,
      "loss/hidden": 3.54765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22021548971533775,
      "step": 5160
    },
    {
      "epoch": 0.517,
      "grad_norm": 34.0,
      "grad_norm_var": 4.706705729166667,
      "learning_rate": 0.0001,
      "loss": 8.1521,
      "loss/crossentropy": 2.1654283188283445,
      "loss/hidden": 3.48203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2115080550312996,
      "step": 5170
    },
    {
      "epoch": 0.518,
      "grad_norm": 30.25,
      "grad_norm_var": 5.541666666666667,
      "learning_rate": 0.0001,
      "loss": 8.13,
      "loss/crossentropy": 2.2458599150180816,
      "loss/hidden": 3.596484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22666897978633643,
      "step": 5180
    },
    {
      "epoch": 0.519,
      "grad_norm": 44.75,
      "grad_norm_var": 14.142122395833333,
      "learning_rate": 0.0001,
      "loss": 8.145,
      "loss/crossentropy": 2.317767137289047,
      "loss/hidden": 3.5671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23820882234722376,
      "step": 5190
    },
    {
      "epoch": 0.52,
      "grad_norm": 32.5,
      "grad_norm_var": 23.218684895833334,
      "learning_rate": 0.0001,
      "loss": 8.1323,
      "loss/crossentropy": 2.279913380742073,
      "loss/hidden": 3.558984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24836393278092145,
      "step": 5200
    },
    {
      "epoch": 0.521,
      "grad_norm": 31.875,
      "grad_norm_var": 16.846875,
      "learning_rate": 0.0001,
      "loss": 8.2041,
      "loss/crossentropy": 2.201883518695831,
      "loss/hidden": 3.57890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22954922150820495,
      "step": 5210
    },
    {
      "epoch": 0.522,
      "grad_norm": 34.25,
      "grad_norm_var": 221.16223958333333,
      "learning_rate": 0.0001,
      "loss": 8.273,
      "loss/crossentropy": 2.2926857471466064,
      "loss/hidden": 3.541796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22256891019642353,
      "step": 5220
    },
    {
      "epoch": 0.523,
      "grad_norm": 75.0,
      "grad_norm_var": 233.03098958333334,
      "learning_rate": 0.0001,
      "loss": 8.1559,
      "loss/crossentropy": 2.1607275292277337,
      "loss/hidden": 3.65859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24259125851094723,
      "step": 5230
    },
    {
      "epoch": 0.524,
      "grad_norm": 33.5,
      "grad_norm_var": 170.5353515625,
      "learning_rate": 0.0001,
      "loss": 8.1314,
      "loss/crossentropy": 2.2264937654137613,
      "loss/hidden": 3.653125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23244266752153636,
      "step": 5240
    },
    {
      "epoch": 0.525,
      "grad_norm": 34.0,
      "grad_norm_var": 69.98515625,
      "learning_rate": 0.0001,
      "loss": 8.0757,
      "loss/crossentropy": 2.263314816355705,
      "loss/hidden": 3.6921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2497670866549015,
      "step": 5250
    },
    {
      "epoch": 0.526,
      "grad_norm": 32.75,
      "grad_norm_var": 15.91640625,
      "learning_rate": 0.0001,
      "loss": 8.1746,
      "loss/crossentropy": 2.2392116367816923,
      "loss/hidden": 3.659765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24461503997445105,
      "step": 5260
    },
    {
      "epoch": 0.527,
      "grad_norm": 30.875,
      "grad_norm_var": 118.5931640625,
      "learning_rate": 0.0001,
      "loss": 8.145,
      "loss/crossentropy": 2.184428018331528,
      "loss/hidden": 3.6265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24789600986987353,
      "step": 5270
    },
    {
      "epoch": 0.528,
      "grad_norm": 40.5,
      "grad_norm_var": 127.34791666666666,
      "learning_rate": 0.0001,
      "loss": 8.0442,
      "loss/crossentropy": 2.1109130561351774,
      "loss/hidden": 3.528125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21775138471275568,
      "step": 5280
    },
    {
      "epoch": 0.529,
      "grad_norm": 34.5,
      "grad_norm_var": 6.623958333333333,
      "learning_rate": 0.0001,
      "loss": 8.1308,
      "loss/crossentropy": 2.386467677354813,
      "loss/hidden": 3.607421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24021831918507813,
      "step": 5290
    },
    {
      "epoch": 0.53,
      "grad_norm": 34.25,
      "grad_norm_var": 7.5666015625,
      "learning_rate": 0.0001,
      "loss": 8.1256,
      "loss/crossentropy": 2.3109169751405716,
      "loss/hidden": 3.630859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2501339312642813,
      "step": 5300
    },
    {
      "epoch": 0.531,
      "grad_norm": 31.875,
      "grad_norm_var": 12.195572916666666,
      "learning_rate": 0.0001,
      "loss": 8.1229,
      "loss/crossentropy": 2.346800622344017,
      "loss/hidden": 3.608203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2536924373358488,
      "step": 5310
    },
    {
      "epoch": 0.532,
      "grad_norm": 31.875,
      "grad_norm_var": 9.997330729166666,
      "learning_rate": 0.0001,
      "loss": 8.0104,
      "loss/crossentropy": 2.1395985931158066,
      "loss/hidden": 3.471875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21648342311382293,
      "step": 5320
    },
    {
      "epoch": 0.533,
      "grad_norm": 32.25,
      "grad_norm_var": 54.80358072916667,
      "learning_rate": 0.0001,
      "loss": 8.2808,
      "loss/crossentropy": 2.072698312997818,
      "loss/hidden": 3.70625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21971321273595096,
      "step": 5330
    },
    {
      "epoch": 0.534,
      "grad_norm": 37.75,
      "grad_norm_var": 582.04140625,
      "learning_rate": 0.0001,
      "loss": 8.4396,
      "loss/crossentropy": 2.168602865189314,
      "loss/hidden": 3.68828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22807842567563058,
      "step": 5340
    },
    {
      "epoch": 0.535,
      "grad_norm": 36.0,
      "grad_norm_var": 17.0619140625,
      "learning_rate": 0.0001,
      "loss": 8.2354,
      "loss/crossentropy": 2.212322035431862,
      "loss/hidden": 3.579296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21836955063045024,
      "step": 5350
    },
    {
      "epoch": 0.536,
      "grad_norm": 37.75,
      "grad_norm_var": 18.780989583333334,
      "learning_rate": 0.0001,
      "loss": 8.2718,
      "loss/crossentropy": 2.1710034780204297,
      "loss/hidden": 3.656640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2387496206909418,
      "step": 5360
    },
    {
      "epoch": 0.537,
      "grad_norm": 29.125,
      "grad_norm_var": 27.276041666666668,
      "learning_rate": 0.0001,
      "loss": 8.2691,
      "loss/crossentropy": 2.097558119148016,
      "loss/hidden": 3.680078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2329912935383618,
      "step": 5370
    },
    {
      "epoch": 0.538,
      "grad_norm": 37.5,
      "grad_norm_var": 18.0181640625,
      "learning_rate": 0.0001,
      "loss": 8.057,
      "loss/crossentropy": 2.275981144607067,
      "loss/hidden": 3.55703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22314830794930457,
      "step": 5380
    },
    {
      "epoch": 0.539,
      "grad_norm": 34.25,
      "grad_norm_var": 19.930208333333333,
      "learning_rate": 0.0001,
      "loss": 8.182,
      "loss/crossentropy": 2.184271165728569,
      "loss/hidden": 3.6609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23054859917610884,
      "step": 5390
    },
    {
      "epoch": 0.54,
      "grad_norm": 34.0,
      "grad_norm_var": 17.333072916666666,
      "learning_rate": 0.0001,
      "loss": 8.3439,
      "loss/crossentropy": 2.480144701898098,
      "loss/hidden": 3.701171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2607876468449831,
      "step": 5400
    },
    {
      "epoch": 0.541,
      "grad_norm": 33.75,
      "grad_norm_var": 14.249739583333334,
      "learning_rate": 0.0001,
      "loss": 8.2028,
      "loss/crossentropy": 2.2421235501766206,
      "loss/hidden": 3.637890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23026540800929068,
      "step": 5410
    },
    {
      "epoch": 0.542,
      "grad_norm": 33.0,
      "grad_norm_var": 14.408072916666667,
      "learning_rate": 0.0001,
      "loss": 8.122,
      "loss/crossentropy": 2.173307144641876,
      "loss/hidden": 3.64453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22846251353621483,
      "step": 5420
    },
    {
      "epoch": 0.543,
      "grad_norm": 33.0,
      "grad_norm_var": 5.3,
      "learning_rate": 0.0001,
      "loss": 8.2194,
      "loss/crossentropy": 2.3099235713481905,
      "loss/hidden": 3.629296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23289041519165038,
      "step": 5430
    },
    {
      "epoch": 0.544,
      "grad_norm": 36.0,
      "grad_norm_var": 28.040625,
      "learning_rate": 0.0001,
      "loss": 8.1837,
      "loss/crossentropy": 2.2322505958378316,
      "loss/hidden": 3.643359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2281514243222773,
      "step": 5440
    },
    {
      "epoch": 0.545,
      "grad_norm": 30.0,
      "grad_norm_var": 69.69166666666666,
      "learning_rate": 0.0001,
      "loss": 8.1516,
      "loss/crossentropy": 2.199453258514404,
      "loss/hidden": 3.5890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22161313518881798,
      "step": 5450
    },
    {
      "epoch": 0.546,
      "grad_norm": 36.25,
      "grad_norm_var": 88.72890625,
      "learning_rate": 0.0001,
      "loss": 8.305,
      "loss/crossentropy": 2.3247879207134248,
      "loss/hidden": 3.74609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2331052988767624,
      "step": 5460
    },
    {
      "epoch": 0.547,
      "grad_norm": 50.25,
      "grad_norm_var": 49.67265625,
      "learning_rate": 0.0001,
      "loss": 8.1121,
      "loss/crossentropy": 2.2907343961298468,
      "loss/hidden": 3.598828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22796698454767467,
      "step": 5470
    },
    {
      "epoch": 0.548,
      "grad_norm": 34.5,
      "grad_norm_var": 27.73125,
      "learning_rate": 0.0001,
      "loss": 8.1348,
      "loss/crossentropy": 2.4852760285139084,
      "loss/hidden": 3.5296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23372339643537998,
      "step": 5480
    },
    {
      "epoch": 0.549,
      "grad_norm": 33.0,
      "grad_norm_var": 2.703285648974309e+18,
      "learning_rate": 0.0001,
      "loss": 8.0635,
      "loss/crossentropy": 2.3434904247522352,
      "loss/hidden": 3.725390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21989205628633499,
      "step": 5490
    },
    {
      "epoch": 0.55,
      "grad_norm": 30.5,
      "grad_norm_var": 6.481184895833334,
      "learning_rate": 0.0001,
      "loss": 8.1213,
      "loss/crossentropy": 2.048020973801613,
      "loss/hidden": 3.53046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21088404105976225,
      "step": 5500
    },
    {
      "epoch": 0.551,
      "grad_norm": 41.0,
      "grad_norm_var": 16.025,
      "learning_rate": 0.0001,
      "loss": 8.143,
      "loss/crossentropy": 2.246925861388445,
      "loss/hidden": 3.616796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2380803508684039,
      "step": 5510
    },
    {
      "epoch": 0.552,
      "grad_norm": 34.0,
      "grad_norm_var": 19.940625,
      "learning_rate": 0.0001,
      "loss": 8.1266,
      "loss/crossentropy": 2.306339371204376,
      "loss/hidden": 3.707421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2542409796267748,
      "step": 5520
    },
    {
      "epoch": 0.553,
      "grad_norm": 30.25,
      "grad_norm_var": 25.34375,
      "learning_rate": 0.0001,
      "loss": 8.1711,
      "loss/crossentropy": 2.172018714249134,
      "loss/hidden": 3.602734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24370079562067987,
      "step": 5530
    },
    {
      "epoch": 0.554,
      "grad_norm": 42.25,
      "grad_norm_var": 30.013541666666665,
      "learning_rate": 0.0001,
      "loss": 8.1492,
      "loss/crossentropy": 2.22998360991478,
      "loss/hidden": 3.625390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23992773257195948,
      "step": 5540
    },
    {
      "epoch": 0.555,
      "grad_norm": 32.25,
      "grad_norm_var": 28.613541666666666,
      "learning_rate": 0.0001,
      "loss": 8.0695,
      "loss/crossentropy": 2.2180883288383484,
      "loss/hidden": 3.598828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22950777132064104,
      "step": 5550
    },
    {
      "epoch": 0.556,
      "grad_norm": 33.0,
      "grad_norm_var": 34.692708333333336,
      "learning_rate": 0.0001,
      "loss": 8.1162,
      "loss/crossentropy": 2.174109402298927,
      "loss/hidden": 3.702734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24184305276721715,
      "step": 5560
    },
    {
      "epoch": 0.557,
      "grad_norm": 34.0,
      "grad_norm_var": 7.314583333333333,
      "learning_rate": 0.0001,
      "loss": 8.0114,
      "loss/crossentropy": 2.1175956279039383,
      "loss/hidden": 3.53828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21476623937487602,
      "step": 5570
    },
    {
      "epoch": 0.558,
      "grad_norm": 35.25,
      "grad_norm_var": 19.8119140625,
      "learning_rate": 0.0001,
      "loss": 8.1025,
      "loss/crossentropy": 2.2917901635169984,
      "loss/hidden": 3.558984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22450571469962596,
      "step": 5580
    },
    {
      "epoch": 0.559,
      "grad_norm": 38.5,
      "grad_norm_var": 21.97890625,
      "learning_rate": 0.0001,
      "loss": 8.2385,
      "loss/crossentropy": 2.3675080120563505,
      "loss/hidden": 3.63828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23681335002183915,
      "step": 5590
    },
    {
      "epoch": 0.56,
      "grad_norm": 33.0,
      "grad_norm_var": 9.764583333333333,
      "learning_rate": 0.0001,
      "loss": 8.1467,
      "loss/crossentropy": 2.3383423417806624,
      "loss/hidden": 3.623046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23495447412133216,
      "step": 5600
    },
    {
      "epoch": 0.561,
      "grad_norm": 34.25,
      "grad_norm_var": 11.692643229166666,
      "learning_rate": 0.0001,
      "loss": 8.1075,
      "loss/crossentropy": 2.2219532161951063,
      "loss/hidden": 3.581640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21968780737370253,
      "step": 5610
    },
    {
      "epoch": 0.562,
      "grad_norm": 29.875,
      "grad_norm_var": 19.882747395833334,
      "learning_rate": 0.0001,
      "loss": 8.0262,
      "loss/crossentropy": 2.121725457906723,
      "loss/hidden": 3.564453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22616409026086332,
      "step": 5620
    },
    {
      "epoch": 0.563,
      "grad_norm": 31.625,
      "grad_norm_var": 1.869627140463989e+18,
      "learning_rate": 0.0001,
      "loss": 8.2844,
      "loss/crossentropy": 2.2004577577114106,
      "loss/hidden": 3.651171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22547926437109708,
      "step": 5630
    },
    {
      "epoch": 0.564,
      "grad_norm": 32.0,
      "grad_norm_var": 4.611458333333333,
      "learning_rate": 0.0001,
      "loss": 8.1244,
      "loss/crossentropy": 2.1516154944896697,
      "loss/hidden": 3.55078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2172885647043586,
      "step": 5640
    },
    {
      "epoch": 0.565,
      "grad_norm": 30.5,
      "grad_norm_var": 7.317708333333333,
      "learning_rate": 0.0001,
      "loss": 8.0514,
      "loss/crossentropy": 2.2103342309594156,
      "loss/hidden": 3.6140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22790345773100854,
      "step": 5650
    },
    {
      "epoch": 0.566,
      "grad_norm": 34.25,
      "grad_norm_var": 13.6587890625,
      "learning_rate": 0.0001,
      "loss": 8.0864,
      "loss/crossentropy": 2.2035325288772585,
      "loss/hidden": 3.57734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23142583221197127,
      "step": 5660
    },
    {
      "epoch": 0.567,
      "grad_norm": 31.625,
      "grad_norm_var": 9.45625,
      "learning_rate": 0.0001,
      "loss": 8.0716,
      "loss/crossentropy": 2.1144707940518854,
      "loss/hidden": 3.555078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21162977050989867,
      "step": 5670
    },
    {
      "epoch": 0.568,
      "grad_norm": 32.5,
      "grad_norm_var": 31.39140625,
      "learning_rate": 0.0001,
      "loss": 8.096,
      "loss/crossentropy": 2.1229008197784425,
      "loss/hidden": 3.554296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2217079123482108,
      "step": 5680
    },
    {
      "epoch": 0.569,
      "grad_norm": 35.5,
      "grad_norm_var": 30.71875,
      "learning_rate": 0.0001,
      "loss": 8.0824,
      "loss/crossentropy": 2.1256051540374754,
      "loss/hidden": 3.53046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22356193587183953,
      "step": 5690
    },
    {
      "epoch": 0.57,
      "grad_norm": 30.125,
      "grad_norm_var": 6.362239583333333,
      "learning_rate": 0.0001,
      "loss": 8.1334,
      "loss/crossentropy": 2.1438381403684614,
      "loss/hidden": 3.51875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2077854923903942,
      "step": 5700
    },
    {
      "epoch": 0.571,
      "grad_norm": 32.5,
      "grad_norm_var": 14.85390625,
      "learning_rate": 0.0001,
      "loss": 8.0718,
      "loss/crossentropy": 2.2198398813605307,
      "loss/hidden": 3.580859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22509159836918116,
      "step": 5710
    },
    {
      "epoch": 0.572,
      "grad_norm": 34.75,
      "grad_norm_var": 13.4759765625,
      "learning_rate": 0.0001,
      "loss": 8.0583,
      "loss/crossentropy": 2.4060455739498137,
      "loss/hidden": 3.708203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.25263102930039166,
      "step": 5720
    },
    {
      "epoch": 0.573,
      "grad_norm": 28.625,
      "grad_norm_var": 6.946809895833334,
      "learning_rate": 0.0001,
      "loss": 8.0726,
      "loss/crossentropy": 2.1238791063427924,
      "loss/hidden": 3.578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21989983841776847,
      "step": 5730
    },
    {
      "epoch": 0.574,
      "grad_norm": 33.75,
      "grad_norm_var": 9.4884765625,
      "learning_rate": 0.0001,
      "loss": 8.1374,
      "loss/crossentropy": 2.1622555539011956,
      "loss/hidden": 3.583203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21868109367787839,
      "step": 5740
    },
    {
      "epoch": 0.575,
      "grad_norm": 33.0,
      "grad_norm_var": 16.939518229166666,
      "learning_rate": 0.0001,
      "loss": 7.9816,
      "loss/crossentropy": 2.203578273952007,
      "loss/hidden": 3.57109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22708230018615722,
      "step": 5750
    },
    {
      "epoch": 0.576,
      "grad_norm": 33.5,
      "grad_norm_var": 7.49765625,
      "learning_rate": 0.0001,
      "loss": 8.0534,
      "loss/crossentropy": 2.0832700729370117,
      "loss/hidden": 3.5890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2166902620345354,
      "step": 5760
    },
    {
      "epoch": 0.577,
      "grad_norm": 38.5,
      "grad_norm_var": 7.069205729166667,
      "learning_rate": 0.0001,
      "loss": 8.1567,
      "loss/crossentropy": 2.1000006228685377,
      "loss/hidden": 3.6328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2189541209489107,
      "step": 5770
    },
    {
      "epoch": 0.578,
      "grad_norm": 34.0,
      "grad_norm_var": 10.34140625,
      "learning_rate": 0.0001,
      "loss": 7.9902,
      "loss/crossentropy": 2.2755674168467523,
      "loss/hidden": 3.5046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2158666817471385,
      "step": 5780
    },
    {
      "epoch": 0.579,
      "grad_norm": 31.0,
      "grad_norm_var": 7.744791666666667,
      "learning_rate": 0.0001,
      "loss": 8.049,
      "loss/crossentropy": 2.3444782361388206,
      "loss/hidden": 3.565625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22829483300447465,
      "step": 5790
    },
    {
      "epoch": 0.58,
      "grad_norm": 31.75,
      "grad_norm_var": 6.076497395833333,
      "learning_rate": 0.0001,
      "loss": 8.0271,
      "loss/crossentropy": 2.202429711073637,
      "loss/hidden": 3.4640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21492673214524985,
      "step": 5800
    },
    {
      "epoch": 0.581,
      "grad_norm": 31.0,
      "grad_norm_var": 3.6811848958333333,
      "learning_rate": 0.0001,
      "loss": 7.8948,
      "loss/crossentropy": 2.3126792818307877,
      "loss/hidden": 3.46015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22085475884377956,
      "step": 5810
    },
    {
      "epoch": 0.582,
      "grad_norm": 33.0,
      "grad_norm_var": 3.542708333333333,
      "learning_rate": 0.0001,
      "loss": 8.1146,
      "loss/crossentropy": 2.264694780111313,
      "loss/hidden": 3.47421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21074176728725433,
      "step": 5820
    },
    {
      "epoch": 0.583,
      "grad_norm": 31.5,
      "grad_norm_var": 2.3452473958333333,
      "learning_rate": 0.0001,
      "loss": 8.079,
      "loss/crossentropy": 2.2641124561429025,
      "loss/hidden": 3.5125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22780102472752334,
      "step": 5830
    },
    {
      "epoch": 0.584,
      "grad_norm": 32.5,
      "grad_norm_var": 14.5947265625,
      "learning_rate": 0.0001,
      "loss": 8.0337,
      "loss/crossentropy": 2.218563383817673,
      "loss/hidden": 3.62734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24179403018206358,
      "step": 5840
    },
    {
      "epoch": 0.585,
      "grad_norm": 32.75,
      "grad_norm_var": 3.1705729166666665,
      "learning_rate": 0.0001,
      "loss": 8.0352,
      "loss/crossentropy": 2.2127060025930403,
      "loss/hidden": 3.51015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22846792116761208,
      "step": 5850
    },
    {
      "epoch": 0.586,
      "grad_norm": 31.0,
      "grad_norm_var": 5.418489583333334,
      "learning_rate": 0.0001,
      "loss": 8.0163,
      "loss/crossentropy": 2.2808433353900908,
      "loss/hidden": 3.579296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23100998885929586,
      "step": 5860
    },
    {
      "epoch": 0.587,
      "grad_norm": 33.75,
      "grad_norm_var": 7.083333333333333,
      "learning_rate": 0.0001,
      "loss": 8.0979,
      "loss/crossentropy": 2.335478585958481,
      "loss/hidden": 3.541015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2336500260978937,
      "step": 5870
    },
    {
      "epoch": 0.588,
      "grad_norm": 31.5,
      "grad_norm_var": 46.71243489583333,
      "learning_rate": 0.0001,
      "loss": 8.065,
      "loss/crossentropy": 2.1413629055023193,
      "loss/hidden": 3.46171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21667953655123712,
      "step": 5880
    },
    {
      "epoch": 0.589,
      "grad_norm": 33.25,
      "grad_norm_var": 90.58515625,
      "learning_rate": 0.0001,
      "loss": 8.1265,
      "loss/crossentropy": 2.3575997933745385,
      "loss/hidden": 3.5171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23772844970226287,
      "step": 5890
    },
    {
      "epoch": 0.59,
      "grad_norm": 42.25,
      "grad_norm_var": 108.78020833333333,
      "learning_rate": 0.0001,
      "loss": 8.1898,
      "loss/crossentropy": 2.31534286737442,
      "loss/hidden": 3.5765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23516609705984592,
      "step": 5900
    },
    {
      "epoch": 0.591,
      "grad_norm": 30.125,
      "grad_norm_var": 96.6228515625,
      "learning_rate": 0.0001,
      "loss": 8.2098,
      "loss/crossentropy": 2.3408817887306212,
      "loss/hidden": 3.58515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.234616519510746,
      "step": 5910
    },
    {
      "epoch": 0.592,
      "grad_norm": 34.5,
      "grad_norm_var": 5.761458333333334,
      "learning_rate": 0.0001,
      "loss": 8.0412,
      "loss/crossentropy": 2.3171244740486143,
      "loss/hidden": 3.690234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24309027940034866,
      "step": 5920
    },
    {
      "epoch": 0.593,
      "grad_norm": 31.875,
      "grad_norm_var": 18.249934895833334,
      "learning_rate": 0.0001,
      "loss": 8.2448,
      "loss/crossentropy": 2.3113586097955703,
      "loss/hidden": 3.537890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2302385514602065,
      "step": 5930
    },
    {
      "epoch": 0.594,
      "grad_norm": 29.625,
      "grad_norm_var": 16.43125,
      "learning_rate": 0.0001,
      "loss": 7.9336,
      "loss/crossentropy": 2.325835222005844,
      "loss/hidden": 3.434765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21180371306836604,
      "step": 5940
    },
    {
      "epoch": 0.595,
      "grad_norm": 30.5,
      "grad_norm_var": 10.299739583333333,
      "learning_rate": 0.0001,
      "loss": 7.9796,
      "loss/crossentropy": 2.0509180039167405,
      "loss/hidden": 3.604296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2253203097730875,
      "step": 5950
    },
    {
      "epoch": 0.596,
      "grad_norm": 35.5,
      "grad_norm_var": 2.948958333333333,
      "learning_rate": 0.0001,
      "loss": 8.0105,
      "loss/crossentropy": 2.320768731832504,
      "loss/hidden": 3.509375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2229237537831068,
      "step": 5960
    },
    {
      "epoch": 0.597,
      "grad_norm": 28.375,
      "grad_norm_var": 5.462434895833334,
      "learning_rate": 0.0001,
      "loss": 7.9407,
      "loss/crossentropy": 2.3377193987369536,
      "loss/hidden": 3.551171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22489294074475766,
      "step": 5970
    },
    {
      "epoch": 0.598,
      "grad_norm": 37.25,
      "grad_norm_var": 10.767643229166667,
      "learning_rate": 0.0001,
      "loss": 7.9551,
      "loss/crossentropy": 2.26409173309803,
      "loss/hidden": 3.5734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23901313543319702,
      "step": 5980
    },
    {
      "epoch": 0.599,
      "grad_norm": 35.0,
      "grad_norm_var": 14.5375,
      "learning_rate": 0.0001,
      "loss": 7.8592,
      "loss/crossentropy": 2.2190980166196823,
      "loss/hidden": 3.554296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22315293960273266,
      "step": 5990
    },
    {
      "epoch": 0.6,
      "grad_norm": 31.0,
      "grad_norm_var": 11.592643229166667,
      "learning_rate": 0.0001,
      "loss": 8.0295,
      "loss/crossentropy": 2.3040059447288512,
      "loss/hidden": 3.491015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22013801857829093,
      "step": 6000
    },
    {
      "epoch": 0.601,
      "grad_norm": 35.25,
      "grad_norm_var": 3.64375,
      "learning_rate": 9.999861209401554e-05,
      "loss": 8.0943,
      "loss/crossentropy": 1.9778777174651623,
      "loss/hidden": 3.649609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2236777282319963,
      "step": 6010
    },
    {
      "epoch": 0.602,
      "grad_norm": 31.875,
      "grad_norm_var": 15.4681640625,
      "learning_rate": 9.999444846167473e-05,
      "loss": 8.0478,
      "loss/crossentropy": 2.1667948998510838,
      "loss/hidden": 3.52265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21487980112433433,
      "step": 6020
    },
    {
      "epoch": 0.603,
      "grad_norm": 33.75,
      "grad_norm_var": 6.246875,
      "learning_rate": 9.998750935981003e-05,
      "loss": 7.896,
      "loss/crossentropy": 2.338212412595749,
      "loss/hidden": 3.546484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24026264138519765,
      "step": 6030
    },
    {
      "epoch": 0.604,
      "grad_norm": 34.75,
      "grad_norm_var": 19.437434895833334,
      "learning_rate": 9.997779521645793e-05,
      "loss": 8.0332,
      "loss/crossentropy": 2.107576106488705,
      "loss/hidden": 3.59296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22513998029753565,
      "step": 6040
    },
    {
      "epoch": 0.605,
      "grad_norm": 31.625,
      "grad_norm_var": 18.728580729166666,
      "learning_rate": 9.996530663083255e-05,
      "loss": 8.0406,
      "loss/crossentropy": 2.15678728222847,
      "loss/hidden": 3.5578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22172329109162092,
      "step": 6050
    },
    {
      "epoch": 0.606,
      "grad_norm": 35.75,
      "grad_norm_var": 4.276041666666667,
      "learning_rate": 9.995004437328867e-05,
      "loss": 8.0676,
      "loss/crossentropy": 2.1027655113488435,
      "loss/hidden": 3.505078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2081753826700151,
      "step": 6060
    },
    {
      "epoch": 0.607,
      "grad_norm": 33.0,
      "grad_norm_var": 15.7822265625,
      "learning_rate": 9.993200938527422e-05,
      "loss": 7.9934,
      "loss/crossentropy": 2.2116799533367155,
      "loss/hidden": 3.478515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2213600728660822,
      "step": 6070
    },
    {
      "epoch": 0.608,
      "grad_norm": 33.75,
      "grad_norm_var": 16.8009765625,
      "learning_rate": 9.991120277927223e-05,
      "loss": 7.9668,
      "loss/crossentropy": 2.317891779541969,
      "loss/hidden": 3.463671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21628273129463196,
      "step": 6080
    },
    {
      "epoch": 0.609,
      "grad_norm": 35.5,
      "grad_norm_var": 40.41848958333333,
      "learning_rate": 9.988762583873216e-05,
      "loss": 8.1108,
      "loss/crossentropy": 2.2937388941645622,
      "loss/hidden": 3.57734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.24392955824732782,
      "step": 6090
    },
    {
      "epoch": 0.61,
      "grad_norm": 32.75,
      "grad_norm_var": 45.71139322916667,
      "learning_rate": 9.986128001799077e-05,
      "loss": 7.9542,
      "loss/crossentropy": 2.1343540251255035,
      "loss/hidden": 3.569921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21609723400324582,
      "step": 6100
    },
    {
      "epoch": 0.611,
      "grad_norm": 36.5,
      "grad_norm_var": 15.710416666666667,
      "learning_rate": 9.983216694218236e-05,
      "loss": 8.0213,
      "loss/crossentropy": 2.2557659655809403,
      "loss/hidden": 3.57265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23297818191349506,
      "step": 6110
    },
    {
      "epoch": 0.612,
      "grad_norm": 29.75,
      "grad_norm_var": 5.277018229166667,
      "learning_rate": 9.980028840713861e-05,
      "loss": 7.9736,
      "loss/crossentropy": 2.266719642281532,
      "loss/hidden": 3.5078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23359888847917318,
      "step": 6120
    },
    {
      "epoch": 0.613,
      "grad_norm": 32.5,
      "grad_norm_var": 5.777083333333334,
      "learning_rate": 9.97656463792777e-05,
      "loss": 8.0245,
      "loss/crossentropy": 2.160386872291565,
      "loss/hidden": 3.416015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2051708722487092,
      "step": 6130
    },
    {
      "epoch": 0.614,
      "grad_norm": 30.25,
      "grad_norm_var": 3.147330729166667,
      "learning_rate": 9.97282429954831e-05,
      "loss": 7.9544,
      "loss/crossentropy": 2.3059318214654922,
      "loss/hidden": 3.55234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22518980633467436,
      "step": 6140
    },
    {
      "epoch": 0.615,
      "grad_norm": 35.5,
      "grad_norm_var": 4.796875,
      "learning_rate": 9.96880805629717e-05,
      "loss": 8.0717,
      "loss/crossentropy": 2.1113718140870334,
      "loss/hidden": 3.504296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2189197548199445,
      "step": 6150
    },
    {
      "epoch": 0.616,
      "grad_norm": 36.0,
      "grad_norm_var": 5.901822916666666,
      "learning_rate": 9.964516155915151e-05,
      "loss": 8.0206,
      "loss/crossentropy": 2.2431884676218035,
      "loss/hidden": 3.5359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22374145314097404,
      "step": 6160
    },
    {
      "epoch": 0.617,
      "grad_norm": 32.5,
      "grad_norm_var": 7.320247395833333,
      "learning_rate": 9.959948863146887e-05,
      "loss": 8.1478,
      "loss/crossentropy": 2.226063944399357,
      "loss/hidden": 3.57265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.23062228299677373,
      "step": 6170
    },
    {
      "epoch": 0.618,
      "grad_norm": 32.0,
      "grad_norm_var": 4.3009765625,
      "learning_rate": 9.95510645972451e-05,
      "loss": 8.0136,
      "loss/crossentropy": 2.3054941266775133,
      "loss/hidden": 3.554296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2221404740586877,
      "step": 6180
    },
    {
      "epoch": 0.619,
      "grad_norm": 30.375,
      "grad_norm_var": 4.343489583333334,
      "learning_rate": 9.949989244350271e-05,
      "loss": 8.0065,
      "loss/crossentropy": 2.243282663822174,
      "loss/hidden": 3.56328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22724493965506554,
      "step": 6190
    },
    {
      "epoch": 0.62,
      "grad_norm": 32.25,
      "grad_norm_var": 3.2697265625,
      "learning_rate": 9.94459753267812e-05,
      "loss": 7.9606,
      "loss/crossentropy": 2.209024667739868,
      "loss/hidden": 3.48125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2209360484033823,
      "step": 6200
    },
    {
      "epoch": 0.621,
      "grad_norm": 31.75,
      "grad_norm_var": 3.57890625,
      "learning_rate": 9.93893165729423e-05,
      "loss": 8.0263,
      "loss/crossentropy": 2.249543938040733,
      "loss/hidden": 3.55625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21819613687694073,
      "step": 6210
    },
    {
      "epoch": 0.622,
      "grad_norm": 35.5,
      "grad_norm_var": 2.4344770438731136e+18,
      "learning_rate": 9.932991967696483e-05,
      "loss": 8.15,
      "loss/crossentropy": 2.268956796824932,
      "loss/hidden": 3.45,
      "loss/jsd": 0.0,
      "loss/logits": 0.21720595508813859,
      "step": 6220
    },
    {
      "epoch": 0.623,
      "grad_norm": 32.5,
      "grad_norm_var": 2.4344770473252357e+18,
      "learning_rate": 9.926778830272912e-05,
      "loss": 7.8604,
      "loss/crossentropy": 2.275849539041519,
      "loss/hidden": 3.48046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20637927446514368,
      "step": 6230
    },
    {
      "epoch": 0.624,
      "grad_norm": 34.0,
      "grad_norm_var": 7.830208333333333,
      "learning_rate": 9.920292628279099e-05,
      "loss": 8.1704,
      "loss/crossentropy": 2.3654884546995163,
      "loss/hidden": 3.56953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.24397928323596715,
      "step": 6240
    },
    {
      "epoch": 0.625,
      "grad_norm": 33.5,
      "grad_norm_var": 13.568489583333333,
      "learning_rate": 9.913533761814537e-05,
      "loss": 7.9571,
      "loss/crossentropy": 2.158632145822048,
      "loss/hidden": 3.58671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.24105301443487406,
      "step": 6250
    },
    {
      "epoch": 0.626,
      "grad_norm": 30.625,
      "grad_norm_var": 19.718489583333334,
      "learning_rate": 9.906502647797946e-05,
      "loss": 8.014,
      "loss/crossentropy": 2.4216948479413984,
      "loss/hidden": 3.524609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23831300530582666,
      "step": 6260
    },
    {
      "epoch": 0.627,
      "grad_norm": 33.25,
      "grad_norm_var": 5.412434895833333,
      "learning_rate": 9.899199719941559e-05,
      "loss": 7.9918,
      "loss/crossentropy": 2.312904378771782,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2172198722139001,
      "step": 6270
    },
    {
      "epoch": 0.628,
      "grad_norm": 30.75,
      "grad_norm_var": 13.770247395833334,
      "learning_rate": 9.891625428724363e-05,
      "loss": 7.8961,
      "loss/crossentropy": 2.1163319811224937,
      "loss/hidden": 3.453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2137595346197486,
      "step": 6280
    },
    {
      "epoch": 0.629,
      "grad_norm": 40.0,
      "grad_norm_var": 17.7884765625,
      "learning_rate": 9.883780241364324e-05,
      "loss": 7.9233,
      "loss/crossentropy": 2.1940839856863024,
      "loss/hidden": 3.467578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22429651636630296,
      "step": 6290
    },
    {
      "epoch": 0.63,
      "grad_norm": 34.75,
      "grad_norm_var": 10.233333333333333,
      "learning_rate": 9.875664641789545e-05,
      "loss": 7.9085,
      "loss/crossentropy": 2.227154319733381,
      "loss/hidden": 3.551171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20596644701436162,
      "step": 6300
    },
    {
      "epoch": 0.631,
      "grad_norm": 38.0,
      "grad_norm_var": 7.104622395833333,
      "learning_rate": 9.867279130608441e-05,
      "loss": 7.9829,
      "loss/crossentropy": 2.035653126239777,
      "loss/hidden": 3.515234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22415168713778258,
      "step": 6310
    },
    {
      "epoch": 0.632,
      "grad_norm": 35.0,
      "grad_norm_var": 23.770572916666666,
      "learning_rate": 9.858624225078841e-05,
      "loss": 8.0008,
      "loss/crossentropy": 2.1803553849458694,
      "loss/hidden": 3.42890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2177688382565975,
      "step": 6320
    },
    {
      "epoch": 0.633,
      "grad_norm": 31.125,
      "grad_norm_var": 11.841666666666667,
      "learning_rate": 9.849700459076084e-05,
      "loss": 8.0075,
      "loss/crossentropy": 2.176757436990738,
      "loss/hidden": 3.520703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21793113015592097,
      "step": 6330
    },
    {
      "epoch": 0.634,
      "grad_norm": 32.5,
      "grad_norm_var": 22.920572916666668,
      "learning_rate": 9.840508383060093e-05,
      "loss": 7.9591,
      "loss/crossentropy": 2.1699598588049414,
      "loss/hidden": 3.510546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22454311894252896,
      "step": 6340
    },
    {
      "epoch": 0.635,
      "grad_norm": 31.25,
      "grad_norm_var": 13.63515625,
      "learning_rate": 9.831048564041413e-05,
      "loss": 8.0201,
      "loss/crossentropy": 2.1979137234389783,
      "loss/hidden": 3.4484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20997896678745748,
      "step": 6350
    },
    {
      "epoch": 0.636,
      "grad_norm": 29.5,
      "grad_norm_var": 4.823893229166667,
      "learning_rate": 9.821321585546244e-05,
      "loss": 7.999,
      "loss/crossentropy": 2.2549417853355407,
      "loss/hidden": 3.488671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21141785085201265,
      "step": 6360
    },
    {
      "epoch": 0.637,
      "grad_norm": 39.25,
      "grad_norm_var": 12.4650390625,
      "learning_rate": 9.811328047580437e-05,
      "loss": 8.0388,
      "loss/crossentropy": 2.239156422019005,
      "loss/hidden": 3.60625,
      "loss/jsd": 0.0,
      "loss/logits": 0.24648461528122426,
      "step": 6370
    },
    {
      "epoch": 0.638,
      "grad_norm": 47.25,
      "grad_norm_var": 25.014322916666668,
      "learning_rate": 9.801068566592485e-05,
      "loss": 7.9025,
      "loss/crossentropy": 2.1251075878739356,
      "loss/hidden": 3.579296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21702531222254037,
      "step": 6380
    },
    {
      "epoch": 0.639,
      "grad_norm": 30.5,
      "grad_norm_var": 20.876041666666666,
      "learning_rate": 9.790543775435505e-05,
      "loss": 7.8769,
      "loss/crossentropy": 2.0513882406055926,
      "loss/hidden": 3.529296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20966944433748722,
      "step": 6390
    },
    {
      "epoch": 0.64,
      "grad_norm": 31.875,
      "grad_norm_var": 5.6416015625,
      "learning_rate": 9.779754323328192e-05,
      "loss": 7.9332,
      "loss/crossentropy": 2.3918124705553057,
      "loss/hidden": 3.5328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22760416194796562,
      "step": 6400
    },
    {
      "epoch": 0.641,
      "grad_norm": 28.875,
      "grad_norm_var": 7.597330729166667,
      "learning_rate": 9.76870087581477e-05,
      "loss": 7.9651,
      "loss/crossentropy": 2.0368143267929555,
      "loss/hidden": 3.55,
      "loss/jsd": 0.0,
      "loss/logits": 0.21210087668150662,
      "step": 6410
    },
    {
      "epoch": 0.642,
      "grad_norm": 32.25,
      "grad_norm_var": 9.109309895833333,
      "learning_rate": 9.757384114723954e-05,
      "loss": 8.0321,
      "loss/crossentropy": 2.211759842187166,
      "loss/hidden": 3.527734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22823408897966146,
      "step": 6420
    },
    {
      "epoch": 0.643,
      "grad_norm": 28.375,
      "grad_norm_var": 7.3775390625,
      "learning_rate": 9.745804738126871e-05,
      "loss": 7.9925,
      "loss/crossentropy": 2.326525205373764,
      "loss/hidden": 3.47109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22555426377803087,
      "step": 6430
    },
    {
      "epoch": 0.644,
      "grad_norm": 35.75,
      "grad_norm_var": 11.24375,
      "learning_rate": 9.733963460294015e-05,
      "loss": 7.9035,
      "loss/crossentropy": 2.3097167551517486,
      "loss/hidden": 3.581640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22632835302501916,
      "step": 6440
    },
    {
      "epoch": 0.645,
      "grad_norm": 30.5,
      "grad_norm_var": 8.674739583333333,
      "learning_rate": 9.72186101165118e-05,
      "loss": 7.946,
      "loss/crossentropy": 2.008643839508295,
      "loss/hidden": 3.56875,
      "loss/jsd": 0.0,
      "loss/logits": 0.23042700625956059,
      "step": 6450
    },
    {
      "epoch": 0.646,
      "grad_norm": 29.875,
      "grad_norm_var": 4.911393229166666,
      "learning_rate": 9.709498138734405e-05,
      "loss": 8.0037,
      "loss/crossentropy": 2.1890437103807927,
      "loss/hidden": 3.444921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2202771688811481,
      "step": 6460
    },
    {
      "epoch": 0.647,
      "grad_norm": 29.25,
      "grad_norm_var": 3.709830729166667,
      "learning_rate": 9.696875604143924e-05,
      "loss": 7.8795,
      "loss/crossentropy": 2.1602062940597535,
      "loss/hidden": 3.576171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2190501932054758,
      "step": 6470
    },
    {
      "epoch": 0.648,
      "grad_norm": 28.875,
      "grad_norm_var": 11.384375,
      "learning_rate": 9.683994186497132e-05,
      "loss": 8.0117,
      "loss/crossentropy": 2.2572616577148437,
      "loss/hidden": 3.54296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21902738623321055,
      "step": 6480
    },
    {
      "epoch": 0.649,
      "grad_norm": 33.75,
      "grad_norm_var": 14.962239583333334,
      "learning_rate": 9.670854680380544e-05,
      "loss": 7.9967,
      "loss/crossentropy": 2.2480128668248653,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21361452052369714,
      "step": 6490
    },
    {
      "epoch": 0.65,
      "grad_norm": 30.5,
      "grad_norm_var": 13.709309895833334,
      "learning_rate": 9.657457896300791e-05,
      "loss": 7.8606,
      "loss/crossentropy": 2.3608870059251785,
      "loss/hidden": 3.43671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2066267790272832,
      "step": 6500
    },
    {
      "epoch": 0.651,
      "grad_norm": 31.25,
      "grad_norm_var": 6.6869140625,
      "learning_rate": 9.643804660634618e-05,
      "loss": 7.8916,
      "loss/crossentropy": 2.22277874648571,
      "loss/hidden": 3.583984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2291982416063547,
      "step": 6510
    },
    {
      "epoch": 0.652,
      "grad_norm": 33.0,
      "grad_norm_var": 12.510416666666666,
      "learning_rate": 9.629895815577916e-05,
      "loss": 7.8982,
      "loss/crossentropy": 2.181501531600952,
      "loss/hidden": 3.487109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2111635560169816,
      "step": 6520
    },
    {
      "epoch": 0.653,
      "grad_norm": 33.25,
      "grad_norm_var": 6.308072916666666,
      "learning_rate": 9.615732219093762e-05,
      "loss": 7.8706,
      "loss/crossentropy": 2.2949966207146644,
      "loss/hidden": 3.447265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21308735199272633,
      "step": 6530
    },
    {
      "epoch": 0.654,
      "grad_norm": 31.125,
      "grad_norm_var": 7.6837890625,
      "learning_rate": 9.601314744859504e-05,
      "loss": 7.9133,
      "loss/crossentropy": 2.3161712542176245,
      "loss/hidden": 3.4296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21109349709004163,
      "step": 6540
    },
    {
      "epoch": 0.655,
      "grad_norm": 29.375,
      "grad_norm_var": 12.726822916666666,
      "learning_rate": 9.586644282212866e-05,
      "loss": 8.0585,
      "loss/crossentropy": 2.472279852628708,
      "loss/hidden": 3.501171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2305774724110961,
      "step": 6550
    },
    {
      "epoch": 0.656,
      "grad_norm": 31.875,
      "grad_norm_var": 23.931184895833333,
      "learning_rate": 9.571721736097089e-05,
      "loss": 8.1202,
      "loss/crossentropy": 2.19906941652298,
      "loss/hidden": 3.46640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2118291700258851,
      "step": 6560
    },
    {
      "epoch": 0.657,
      "grad_norm": 30.5,
      "grad_norm_var": 16.205989583333334,
      "learning_rate": 9.556548027005106e-05,
      "loss": 7.9494,
      "loss/crossentropy": 2.2343359500169755,
      "loss/hidden": 3.546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22837236057966948,
      "step": 6570
    },
    {
      "epoch": 0.658,
      "grad_norm": 36.75,
      "grad_norm_var": 7.095247395833334,
      "learning_rate": 9.54112409092277e-05,
      "loss": 8.0652,
      "loss/crossentropy": 2.2837061405181887,
      "loss/hidden": 3.6140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22944757621735334,
      "step": 6580
    },
    {
      "epoch": 0.659,
      "grad_norm": 33.0,
      "grad_norm_var": 8.152018229166666,
      "learning_rate": 9.525450879271113e-05,
      "loss": 7.864,
      "loss/crossentropy": 2.1913695335388184,
      "loss/hidden": 3.40859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20043510757386684,
      "step": 6590
    },
    {
      "epoch": 0.66,
      "grad_norm": 31.125,
      "grad_norm_var": 7.634309895833334,
      "learning_rate": 9.509529358847655e-05,
      "loss": 7.8478,
      "loss/crossentropy": 2.260852278769016,
      "loss/hidden": 3.473046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21337826307862998,
      "step": 6600
    },
    {
      "epoch": 0.661,
      "grad_norm": 31.875,
      "grad_norm_var": 2.4364583333333334,
      "learning_rate": 9.493360511766776e-05,
      "loss": 8.002,
      "loss/crossentropy": 2.278784583508968,
      "loss/hidden": 3.47109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21644247248768805,
      "step": 6610
    },
    {
      "epoch": 0.662,
      "grad_norm": 32.0,
      "grad_norm_var": 2.2604166666666665,
      "learning_rate": 9.476945335399122e-05,
      "loss": 7.8403,
      "loss/crossentropy": 2.305820995569229,
      "loss/hidden": 3.46796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2129553822800517,
      "step": 6620
    },
    {
      "epoch": 0.663,
      "grad_norm": 29.875,
      "grad_norm_var": 1.9775390625,
      "learning_rate": 9.460284842310094e-05,
      "loss": 7.8949,
      "loss/crossentropy": 2.3204252138733863,
      "loss/hidden": 3.68359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.23965017758309842,
      "step": 6630
    },
    {
      "epoch": 0.664,
      "grad_norm": 46.75,
      "grad_norm_var": 19.216666666666665,
      "learning_rate": 9.443380060197387e-05,
      "loss": 8.0113,
      "loss/crossentropy": 2.4102306962013245,
      "loss/hidden": 3.605078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23021724484860898,
      "step": 6640
    },
    {
      "epoch": 0.665,
      "grad_norm": 31.875,
      "grad_norm_var": 19.167643229166668,
      "learning_rate": 9.426232031827588e-05,
      "loss": 7.966,
      "loss/crossentropy": 2.0162095427513123,
      "loss/hidden": 3.461328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20972911100834607,
      "step": 6650
    },
    {
      "epoch": 0.666,
      "grad_norm": 31.0,
      "grad_norm_var": 6.491666666666666,
      "learning_rate": 9.408841814971861e-05,
      "loss": 7.842,
      "loss/crossentropy": 2.2311264127492905,
      "loss/hidden": 3.528125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22341707665473223,
      "step": 6660
    },
    {
      "epoch": 0.667,
      "grad_norm": 30.875,
      "grad_norm_var": 20.850455729166665,
      "learning_rate": 9.391210482340699e-05,
      "loss": 8.0171,
      "loss/crossentropy": 2.232379700243473,
      "loss/hidden": 3.530859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22980675138533116,
      "step": 6670
    },
    {
      "epoch": 0.668,
      "grad_norm": 29.75,
      "grad_norm_var": 3.72265625,
      "learning_rate": 9.373339121517747e-05,
      "loss": 7.7597,
      "loss/crossentropy": 2.245617154240608,
      "loss/hidden": 3.379296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20726991835981606,
      "step": 6680
    },
    {
      "epoch": 0.669,
      "grad_norm": 29.25,
      "grad_norm_var": 9.530989583333334,
      "learning_rate": 9.355228834892724e-05,
      "loss": 7.8524,
      "loss/crossentropy": 2.254352422058582,
      "loss/hidden": 3.52890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22110722698271273,
      "step": 6690
    },
    {
      "epoch": 0.67,
      "grad_norm": 30.125,
      "grad_norm_var": 11.1447265625,
      "learning_rate": 9.336880739593416e-05,
      "loss": 7.8409,
      "loss/crossentropy": 2.2463474199175835,
      "loss/hidden": 3.501171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21668503917753695,
      "step": 6700
    },
    {
      "epoch": 0.671,
      "grad_norm": 27.625,
      "grad_norm_var": 16.1822265625,
      "learning_rate": 9.318295967416768e-05,
      "loss": 7.8072,
      "loss/crossentropy": 2.2114156976342203,
      "loss/hidden": 3.390234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2030044339597225,
      "step": 6710
    },
    {
      "epoch": 0.672,
      "grad_norm": 29.875,
      "grad_norm_var": 17.50390625,
      "learning_rate": 9.299475664759069e-05,
      "loss": 7.8587,
      "loss/crossentropy": 2.233850121498108,
      "loss/hidden": 3.415625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21453158278018236,
      "step": 6720
    },
    {
      "epoch": 0.673,
      "grad_norm": 32.0,
      "grad_norm_var": 9.415625,
      "learning_rate": 9.280420992545239e-05,
      "loss": 7.7826,
      "loss/crossentropy": 2.2446775883436203,
      "loss/hidden": 3.443359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2181243021041155,
      "step": 6730
    },
    {
      "epoch": 0.674,
      "grad_norm": 31.5,
      "grad_norm_var": 2.9622395833333335,
      "learning_rate": 9.261133126157218e-05,
      "loss": 7.8484,
      "loss/crossentropy": 2.286695083975792,
      "loss/hidden": 3.55390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21459419634193183,
      "step": 6740
    },
    {
      "epoch": 0.675,
      "grad_norm": 31.375,
      "grad_norm_var": 1.5893229166666667,
      "learning_rate": 9.241613255361455e-05,
      "loss": 7.8822,
      "loss/crossentropy": 2.1801554679870607,
      "loss/hidden": 3.518359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.22029934152960778,
      "step": 6750
    },
    {
      "epoch": 0.676,
      "grad_norm": 29.375,
      "grad_norm_var": 13.489322916666667,
      "learning_rate": 9.221862584235528e-05,
      "loss": 7.8426,
      "loss/crossentropy": 2.1505274340510367,
      "loss/hidden": 3.5609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2274538179859519,
      "step": 6760
    },
    {
      "epoch": 0.677,
      "grad_norm": 39.0,
      "grad_norm_var": 15.030143229166667,
      "learning_rate": 9.20188233109387e-05,
      "loss": 7.837,
      "loss/crossentropy": 2.3796315863728523,
      "loss/hidden": 3.42578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2178290465846658,
      "step": 6770
    },
    {
      "epoch": 0.678,
      "grad_norm": 28.75,
      "grad_norm_var": 19.102083333333333,
      "learning_rate": 9.181673728412605e-05,
      "loss": 7.8801,
      "loss/crossentropy": 2.2974090576171875,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21345944330096245,
      "step": 6780
    },
    {
      "epoch": 0.679,
      "grad_norm": 30.0,
      "grad_norm_var": 1.4692057291666667,
      "learning_rate": 9.161238022753543e-05,
      "loss": 7.8258,
      "loss/crossentropy": 2.314460426568985,
      "loss/hidden": 3.46953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22287285774946214,
      "step": 6790
    },
    {
      "epoch": 0.68,
      "grad_norm": 32.0,
      "grad_norm_var": 2.5254557291666666,
      "learning_rate": 9.140576474687264e-05,
      "loss": 7.8657,
      "loss/crossentropy": 2.1074823416769504,
      "loss/hidden": 3.544140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2290237602777779,
      "step": 6800
    },
    {
      "epoch": 0.681,
      "grad_norm": 27.875,
      "grad_norm_var": 8.753059895833333,
      "learning_rate": 9.11969035871538e-05,
      "loss": 7.7752,
      "loss/crossentropy": 2.0863692626357078,
      "loss/hidden": 3.39296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20366193084046244,
      "step": 6810
    },
    {
      "epoch": 0.682,
      "grad_norm": 29.875,
      "grad_norm_var": 6.998893229166667,
      "learning_rate": 9.098580963191908e-05,
      "loss": 7.8898,
      "loss/crossentropy": 2.2119090080261232,
      "loss/hidden": 3.5390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2230028085410595,
      "step": 6820
    },
    {
      "epoch": 0.683,
      "grad_norm": 28.0,
      "grad_norm_var": 27.265559895833334,
      "learning_rate": 9.077249590243796e-05,
      "loss": 7.8278,
      "loss/crossentropy": 2.314791253209114,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21508586294949056,
      "step": 6830
    },
    {
      "epoch": 0.684,
      "grad_norm": 29.75,
      "grad_norm_var": 2.4322265625,
      "learning_rate": 9.055697555690608e-05,
      "loss": 7.8087,
      "loss/crossentropy": 2.1545565724372864,
      "loss/hidden": 3.504296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2171786168590188,
      "step": 6840
    },
    {
      "epoch": 0.685,
      "grad_norm": 34.5,
      "grad_norm_var": 6.248372395833333,
      "learning_rate": 9.033926188963352e-05,
      "loss": 7.8663,
      "loss/crossentropy": 2.3174977868795397,
      "loss/hidden": 3.441015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21389166973531246,
      "step": 6850
    },
    {
      "epoch": 0.686,
      "grad_norm": 31.625,
      "grad_norm_var": 3.4760416666666667,
      "learning_rate": 9.011936833022484e-05,
      "loss": 7.812,
      "loss/crossentropy": 2.1643209010362625,
      "loss/hidden": 3.61015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2265725104138255,
      "step": 6860
    },
    {
      "epoch": 0.687,
      "grad_norm": 46.0,
      "grad_norm_var": 2.305350403439006e+18,
      "learning_rate": 8.989730844275054e-05,
      "loss": 7.93,
      "loss/crossentropy": 2.2903329521417617,
      "loss/hidden": 3.47890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21145460121333598,
      "step": 6870
    },
    {
      "epoch": 0.688,
      "grad_norm": 29.5,
      "grad_norm_var": 2.3053504035085965e+18,
      "learning_rate": 8.967309592491052e-05,
      "loss": 7.8319,
      "loss/crossentropy": 2.1312880881130694,
      "loss/hidden": 3.5453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.23307712767273187,
      "step": 6880
    },
    {
      "epoch": 0.689,
      "grad_norm": 28.5,
      "grad_norm_var": 1.8455729166666666,
      "learning_rate": 8.944674460718897e-05,
      "loss": 7.7086,
      "loss/crossentropy": 2.212802681326866,
      "loss/hidden": 3.41328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2138791922479868,
      "step": 6890
    },
    {
      "epoch": 0.69,
      "grad_norm": 31.375,
      "grad_norm_var": 36.62473958333333,
      "learning_rate": 8.921826845200139e-05,
      "loss": 7.6737,
      "loss/crossentropy": 2.196607070416212,
      "loss/hidden": 3.46875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21539878509938717,
      "step": 6900
    },
    {
      "epoch": 0.691,
      "grad_norm": 30.5,
      "grad_norm_var": 36.222330729166664,
      "learning_rate": 8.898768155283327e-05,
      "loss": 7.6532,
      "loss/crossentropy": 1.9995386362075807,
      "loss/hidden": 3.491796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19842760860919953,
      "step": 6910
    },
    {
      "epoch": 0.692,
      "grad_norm": 34.0,
      "grad_norm_var": 15.049739583333333,
      "learning_rate": 8.875499813337069e-05,
      "loss": 7.6989,
      "loss/crossentropy": 2.1321763381361962,
      "loss/hidden": 3.3546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20554839987307788,
      "step": 6920
    },
    {
      "epoch": 0.693,
      "grad_norm": 34.5,
      "grad_norm_var": 12.14140625,
      "learning_rate": 8.852023254662299e-05,
      "loss": 7.7572,
      "loss/crossentropy": 2.294306147098541,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20451737083494664,
      "step": 6930
    },
    {
      "epoch": 0.694,
      "grad_norm": 29.125,
      "grad_norm_var": 3.5056640625,
      "learning_rate": 8.828339927403745e-05,
      "loss": 7.9034,
      "loss/crossentropy": 2.4660038471221926,
      "loss/hidden": 3.53828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22238681241869926,
      "step": 6940
    },
    {
      "epoch": 0.695,
      "grad_norm": 29.875,
      "grad_norm_var": 2.9291666666666667,
      "learning_rate": 8.804451292460585e-05,
      "loss": 7.667,
      "loss/crossentropy": 2.2809097945690153,
      "loss/hidden": 3.379296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21668226066976787,
      "step": 6950
    },
    {
      "epoch": 0.696,
      "grad_norm": 43.0,
      "grad_norm_var": 13.6650390625,
      "learning_rate": 8.780358823396352e-05,
      "loss": 7.8654,
      "loss/crossentropy": 2.203066258132458,
      "loss/hidden": 3.431640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20034241620451212,
      "step": 6960
    },
    {
      "epoch": 0.697,
      "grad_norm": 29.375,
      "grad_norm_var": 13.476822916666666,
      "learning_rate": 8.756064006348021e-05,
      "loss": 7.7268,
      "loss/crossentropy": 2.3059852838516237,
      "loss/hidden": 3.362109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20583362504839897,
      "step": 6970
    },
    {
      "epoch": 0.698,
      "grad_norm": 29.75,
      "grad_norm_var": 2.7666666666666666,
      "learning_rate": 8.731568339934349e-05,
      "loss": 7.6686,
      "loss/crossentropy": 2.146751108765602,
      "loss/hidden": 3.36171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20131726199761033,
      "step": 6980
    },
    {
      "epoch": 0.699,
      "grad_norm": 78.0,
      "grad_norm_var": 149.79973958333332,
      "learning_rate": 8.706873335163425e-05,
      "loss": 7.7995,
      "loss/crossentropy": 2.1803457379341125,
      "loss/hidden": 3.45546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20246275123208762,
      "step": 6990
    },
    {
      "epoch": 0.7,
      "grad_norm": 30.875,
      "grad_norm_var": 151.25,
      "learning_rate": 8.681980515339464e-05,
      "loss": 7.7526,
      "loss/crossentropy": 2.2291737273335457,
      "loss/hidden": 3.494921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.22119418531656265,
      "step": 7000
    },
    {
      "epoch": 0.701,
      "grad_norm": 29.25,
      "grad_norm_var": 1.7181640625,
      "learning_rate": 8.656891415968851e-05,
      "loss": 7.746,
      "loss/crossentropy": 2.1491103902459145,
      "loss/hidden": 3.52421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.2262929579243064,
      "step": 7010
    },
    {
      "epoch": 0.702,
      "grad_norm": 29.625,
      "grad_norm_var": 2.3692057291666666,
      "learning_rate": 8.631607584665414e-05,
      "loss": 7.8135,
      "loss/crossentropy": 2.24861903488636,
      "loss/hidden": 3.387890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21290562488138676,
      "step": 7020
    },
    {
      "epoch": 0.703,
      "grad_norm": 28.375,
      "grad_norm_var": 3.7874348958333335,
      "learning_rate": 8.606130581054967e-05,
      "loss": 7.7292,
      "loss/crossentropy": 2.253491559624672,
      "loss/hidden": 3.466015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21812653560191392,
      "step": 7030
    },
    {
      "epoch": 0.704,
      "grad_norm": 28.25,
      "grad_norm_var": 3.6302083333333335,
      "learning_rate": 8.5804619766791e-05,
      "loss": 7.6905,
      "loss/crossentropy": 2.220169448852539,
      "loss/hidden": 3.409765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2117680374532938,
      "step": 7040
    },
    {
      "epoch": 0.705,
      "grad_norm": 28.5,
      "grad_norm_var": 2.9921223958333334,
      "learning_rate": 8.554603354898238e-05,
      "loss": 7.7344,
      "loss/crossentropy": 2.320455204695463,
      "loss/hidden": 3.4328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2154739665798843,
      "step": 7050
    },
    {
      "epoch": 0.706,
      "grad_norm": 29.125,
      "grad_norm_var": 5.070572916666666,
      "learning_rate": 8.52855631079398e-05,
      "loss": 7.7402,
      "loss/crossentropy": 2.145579095184803,
      "loss/hidden": 3.522265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22706612963229417,
      "step": 7060
    },
    {
      "epoch": 0.707,
      "grad_norm": 33.75,
      "grad_norm_var": 6.292122395833333,
      "learning_rate": 8.502322451070698e-05,
      "loss": 7.6167,
      "loss/crossentropy": 2.1853789925575255,
      "loss/hidden": 3.398828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20946646444499492,
      "step": 7070
    },
    {
      "epoch": 0.708,
      "grad_norm": 26.5,
      "grad_norm_var": 2665.2580729166666,
      "learning_rate": 8.475903393956434e-05,
      "loss": 7.7827,
      "loss/crossentropy": 2.153721308708191,
      "loss/hidden": 3.465625,
      "loss/jsd": 0.0,
      "loss/logits": 0.22963123507797717,
      "step": 7080
    },
    {
      "epoch": 0.709,
      "grad_norm": 28.25,
      "grad_norm_var": 2646.9572916666666,
      "learning_rate": 8.449300769103072e-05,
      "loss": 7.7356,
      "loss/crossentropy": 2.3446192115545275,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20571632999926806,
      "step": 7090
    },
    {
      "epoch": 0.71,
      "grad_norm": 28.0,
      "grad_norm_var": 3.1264973958333333,
      "learning_rate": 8.422516217485826e-05,
      "loss": 7.6287,
      "loss/crossentropy": 2.0742784813046455,
      "loss/hidden": 3.4234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19434763304889202,
      "step": 7100
    },
    {
      "epoch": 0.711,
      "grad_norm": 31.875,
      "grad_norm_var": 11.266080729166667,
      "learning_rate": 8.395551391302005e-05,
      "loss": 7.7378,
      "loss/crossentropy": 2.299386578798294,
      "loss/hidden": 3.454296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21604814752936363,
      "step": 7110
    },
    {
      "epoch": 0.712,
      "grad_norm": 26.0,
      "grad_norm_var": 13.764322916666666,
      "learning_rate": 8.368407953869104e-05,
      "loss": 7.7025,
      "loss/crossentropy": 2.3335316255688667,
      "loss/hidden": 3.40859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20577862840145827,
      "step": 7120
    },
    {
      "epoch": 0.713,
      "grad_norm": 31.25,
      "grad_norm_var": 13.109309895833333,
      "learning_rate": 8.3410875795222e-05,
      "loss": 7.6393,
      "loss/crossentropy": 2.1279921546578406,
      "loss/hidden": 3.40078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1942439589649439,
      "step": 7130
    },
    {
      "epoch": 0.714,
      "grad_norm": 29.25,
      "grad_norm_var": 2.7275390625,
      "learning_rate": 8.313591953510675e-05,
      "loss": 7.5204,
      "loss/crossentropy": 2.2679887428879737,
      "loss/hidden": 3.31484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20306757390499114,
      "step": 7140
    },
    {
      "epoch": 0.715,
      "grad_norm": 29.25,
      "grad_norm_var": 1.6723307291666667,
      "learning_rate": 8.285922771894254e-05,
      "loss": 7.6251,
      "loss/crossentropy": 2.279427632689476,
      "loss/hidden": 3.371484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21383309215307236,
      "step": 7150
    },
    {
      "epoch": 0.716,
      "grad_norm": 27.75,
      "grad_norm_var": 1.8122395833333333,
      "learning_rate": 8.258081741438395e-05,
      "loss": 7.5708,
      "loss/crossentropy": 2.2502959579229356,
      "loss/hidden": 3.4046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20335796754807234,
      "step": 7160
    },
    {
      "epoch": 0.717,
      "grad_norm": 28.875,
      "grad_norm_var": 2.6822265625,
      "learning_rate": 8.230070579508998e-05,
      "loss": 7.6801,
      "loss/crossentropy": 2.1685364373028277,
      "loss/hidden": 3.380078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20438444875180722,
      "step": 7170
    },
    {
      "epoch": 0.718,
      "grad_norm": 28.625,
      "grad_norm_var": 3.81640625,
      "learning_rate": 8.201891013966478e-05,
      "loss": 7.5397,
      "loss/crossentropy": 2.1127954378724096,
      "loss/hidden": 3.278125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18592897318303586,
      "step": 7180
    },
    {
      "epoch": 0.719,
      "grad_norm": 26.5,
      "grad_norm_var": 3.56640625,
      "learning_rate": 8.173544783059173e-05,
      "loss": 7.5438,
      "loss/crossentropy": 2.224238908290863,
      "loss/hidden": 3.38984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.2044967941939831,
      "step": 7190
    },
    {
      "epoch": 0.72,
      "grad_norm": 28.375,
      "grad_norm_var": 12.155989583333334,
      "learning_rate": 8.14503363531613e-05,
      "loss": 7.4945,
      "loss/crossentropy": 2.2011847496032715,
      "loss/hidden": 3.234765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18837413098663092,
      "step": 7200
    },
    {
      "epoch": 0.721,
      "grad_norm": 30.25,
      "grad_norm_var": 12.370572916666667,
      "learning_rate": 8.116359329439238e-05,
      "loss": 7.5481,
      "loss/crossentropy": 2.2355149149894715,
      "loss/hidden": 3.361328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19618814308196306,
      "step": 7210
    },
    {
      "epoch": 0.722,
      "grad_norm": 29.25,
      "grad_norm_var": 9.8837890625,
      "learning_rate": 8.087523634194755e-05,
      "loss": 7.691,
      "loss/crossentropy": 2.2584028095006943,
      "loss/hidden": 3.501953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.22446858156472443,
      "step": 7220
    },
    {
      "epoch": 0.723,
      "grad_norm": 28.75,
      "grad_norm_var": 1.8872395833333333,
      "learning_rate": 8.058528328304188e-05,
      "loss": 7.5763,
      "loss/crossentropy": 2.16469197049737,
      "loss/hidden": 3.320703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20104804309085011,
      "step": 7230
    },
    {
      "epoch": 0.724,
      "grad_norm": 30.125,
      "grad_norm_var": 5.76640625,
      "learning_rate": 8.029375200334588e-05,
      "loss": 7.6155,
      "loss/crossentropy": 2.359184819459915,
      "loss/hidden": 3.3703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.21082259565591813,
      "step": 7240
    },
    {
      "epoch": 0.725,
      "grad_norm": 26.375,
      "grad_norm_var": 2.3379557291666666,
      "learning_rate": 8.000066048588211e-05,
      "loss": 7.5525,
      "loss/crossentropy": 2.0703198462724686,
      "loss/hidden": 3.421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19247690346091986,
      "step": 7250
    },
    {
      "epoch": 0.726,
      "grad_norm": 30.0,
      "grad_norm_var": 3.1747395833333334,
      "learning_rate": 7.970602680991594e-05,
      "loss": 7.666,
      "loss/crossentropy": 2.1073908656835556,
      "loss/hidden": 3.36875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20847180765122175,
      "step": 7260
    },
    {
      "epoch": 0.727,
      "grad_norm": 37.25,
      "grad_norm_var": 10.416666666666666,
      "learning_rate": 7.940986914984034e-05,
      "loss": 7.6802,
      "loss/crossentropy": 2.1649453431367873,
      "loss/hidden": 3.4703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20917554441839456,
      "step": 7270
    },
    {
      "epoch": 0.728,
      "grad_norm": 29.25,
      "grad_norm_var": 9.468684895833333,
      "learning_rate": 7.911220577405484e-05,
      "loss": 7.5714,
      "loss/crossentropy": 2.327304023504257,
      "loss/hidden": 3.34609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20701518896967172,
      "step": 7280
    },
    {
      "epoch": 0.729,
      "grad_norm": 27.375,
      "grad_norm_var": 4.26875,
      "learning_rate": 7.881305504383858e-05,
      "loss": 7.5349,
      "loss/crossentropy": 2.3284592747688295,
      "loss/hidden": 3.3546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20663589648902417,
      "step": 7290
    },
    {
      "epoch": 0.73,
      "grad_norm": 26.75,
      "grad_norm_var": 1.3541666666666667,
      "learning_rate": 7.85124354122177e-05,
      "loss": 7.5198,
      "loss/crossentropy": 2.1882129050791264,
      "loss/hidden": 3.255859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18851547054946421,
      "step": 7300
    },
    {
      "epoch": 0.731,
      "grad_norm": 24.125,
      "grad_norm_var": 6.3056640625,
      "learning_rate": 7.821036542282714e-05,
      "loss": 7.4597,
      "loss/crossentropy": 2.2642627701163294,
      "loss/hidden": 3.2890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2040542172268033,
      "step": 7310
    },
    {
      "epoch": 0.732,
      "grad_norm": 27.25,
      "grad_norm_var": 6.921875,
      "learning_rate": 7.790686370876671e-05,
      "loss": 7.6182,
      "loss/crossentropy": 2.323049134016037,
      "loss/hidden": 3.34375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19615541007369758,
      "step": 7320
    },
    {
      "epoch": 0.733,
      "grad_norm": 26.375,
      "grad_norm_var": 2.01640625,
      "learning_rate": 7.760194899145176e-05,
      "loss": 7.6029,
      "loss/crossentropy": 2.192954847216606,
      "loss/hidden": 3.304296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19462318085134028,
      "step": 7330
    },
    {
      "epoch": 0.734,
      "grad_norm": 25.75,
      "grad_norm_var": 3.754946529022024e+18,
      "learning_rate": 7.729564007945835e-05,
      "loss": 7.5679,
      "loss/crossentropy": 2.2880126923322677,
      "loss/hidden": 3.3265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.20459149368107318,
      "step": 7340
    },
    {
      "epoch": 0.735,
      "grad_norm": 27.375,
      "grad_norm_var": 3.754946529441874e+18,
      "learning_rate": 7.698795586736298e-05,
      "loss": 7.4691,
      "loss/crossentropy": 2.2379926055669785,
      "loss/hidden": 3.358984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20208234172314404,
      "step": 7350
    },
    {
      "epoch": 0.736,
      "grad_norm": 25.375,
      "grad_norm_var": 1.1379557291666667,
      "learning_rate": 7.667891533457719e-05,
      "loss": 7.475,
      "loss/crossentropy": 2.191740782558918,
      "loss/hidden": 3.2578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1825101263821125,
      "step": 7360
    },
    {
      "epoch": 0.737,
      "grad_norm": 28.5,
      "grad_norm_var": 6.686458333333333,
      "learning_rate": 7.636853754417676e-05,
      "loss": 7.4922,
      "loss/crossentropy": 2.215168985724449,
      "loss/hidden": 3.278515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1965383216738701,
      "step": 7370
    },
    {
      "epoch": 0.738,
      "grad_norm": 26.0,
      "grad_norm_var": 2.3,
      "learning_rate": 7.60568416417258e-05,
      "loss": 7.3822,
      "loss/crossentropy": 2.033752679824829,
      "loss/hidden": 3.313671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18001100299879907,
      "step": 7380
    },
    {
      "epoch": 0.739,
      "grad_norm": 27.0,
      "grad_norm_var": 0.8629557291666666,
      "learning_rate": 7.574384685409581e-05,
      "loss": 7.5314,
      "loss/crossentropy": 2.2958304405212404,
      "loss/hidden": 3.351953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19766351543366908,
      "step": 7390
    },
    {
      "epoch": 0.74,
      "grad_norm": 30.25,
      "grad_norm_var": 22.751822916666665,
      "learning_rate": 7.542957248827961e-05,
      "loss": 7.5496,
      "loss/crossentropy": 2.1121088288724423,
      "loss/hidden": 3.32578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.2019992781803012,
      "step": 7400
    },
    {
      "epoch": 0.741,
      "grad_norm": 26.375,
      "grad_norm_var": 21.520247395833334,
      "learning_rate": 7.511403793020047e-05,
      "loss": 7.5124,
      "loss/crossentropy": 2.120064043998718,
      "loss/hidden": 3.375390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.21207549944519996,
      "step": 7410
    },
    {
      "epoch": 0.742,
      "grad_norm": 27.875,
      "grad_norm_var": 7.146809895833333,
      "learning_rate": 7.479726264351618e-05,
      "loss": 7.4119,
      "loss/crossentropy": 2.250824949145317,
      "loss/hidden": 3.257421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1866712011396885,
      "step": 7420
    },
    {
      "epoch": 0.743,
      "grad_norm": 25.125,
      "grad_norm_var": 3.443489583333333,
      "learning_rate": 7.447926616841863e-05,
      "loss": 7.4143,
      "loss/crossentropy": 2.309866726398468,
      "loss/hidden": 3.325,
      "loss/jsd": 0.0,
      "loss/logits": 0.1999421713873744,
      "step": 7430
    },
    {
      "epoch": 0.744,
      "grad_norm": 28.75,
      "grad_norm_var": 2.020247395833333,
      "learning_rate": 7.416006812042828e-05,
      "loss": 7.4387,
      "loss/crossentropy": 2.2461145386099814,
      "loss/hidden": 3.257421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18069201447069644,
      "step": 7440
    },
    {
      "epoch": 0.745,
      "grad_norm": 26.125,
      "grad_norm_var": 4.74140625,
      "learning_rate": 7.383968818918426e-05,
      "loss": 7.44,
      "loss/crossentropy": 2.134336383640766,
      "loss/hidden": 3.3109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1897894937545061,
      "step": 7450
    },
    {
      "epoch": 0.746,
      "grad_norm": 27.875,
      "grad_norm_var": 6.017643229166667,
      "learning_rate": 7.35181461372299e-05,
      "loss": 7.4134,
      "loss/crossentropy": 2.169717122614384,
      "loss/hidden": 3.359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19159721322357653,
      "step": 7460
    },
    {
      "epoch": 0.747,
      "grad_norm": 27.75,
      "grad_norm_var": 3.1775390625,
      "learning_rate": 7.319546179879353e-05,
      "loss": 7.5367,
      "loss/crossentropy": 2.3293275028467177,
      "loss/hidden": 3.377734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.21772474870085717,
      "step": 7470
    },
    {
      "epoch": 0.748,
      "grad_norm": 25.625,
      "grad_norm_var": 4.45,
      "learning_rate": 7.287165507856512e-05,
      "loss": 7.4626,
      "loss/crossentropy": 2.397631025314331,
      "loss/hidden": 3.228125,
      "loss/jsd": 0.0,
      "loss/logits": 0.19252193979918958,
      "step": 7480
    },
    {
      "epoch": 0.749,
      "grad_norm": 27.0,
      "grad_norm_var": 0.67265625,
      "learning_rate": 7.254674595046847e-05,
      "loss": 7.4285,
      "loss/crossentropy": 2.141611284017563,
      "loss/hidden": 3.334375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18560682162642478,
      "step": 7490
    },
    {
      "epoch": 0.75,
      "grad_norm": 25.5,
      "grad_norm_var": 4.889518229166667,
      "learning_rate": 7.222075445642904e-05,
      "loss": 7.3644,
      "loss/crossentropy": 2.2735727220773696,
      "loss/hidden": 3.3375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20446046069264412,
      "step": 7500
    },
    {
      "epoch": 0.751,
      "grad_norm": 26.125,
      "grad_norm_var": 2.8181640625,
      "learning_rate": 7.189370070513775e-05,
      "loss": 7.3532,
      "loss/crossentropy": 2.1920345708727837,
      "loss/hidden": 3.27109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18652982264757156,
      "step": 7510
    },
    {
      "epoch": 0.752,
      "grad_norm": 24.5,
      "grad_norm_var": 3.9556640625,
      "learning_rate": 7.156560487081053e-05,
      "loss": 7.4617,
      "loss/crossentropy": 2.2119674131274225,
      "loss/hidden": 3.33828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20092478170990943,
      "step": 7520
    },
    {
      "epoch": 0.753,
      "grad_norm": 24.875,
      "grad_norm_var": 4.035872395833334,
      "learning_rate": 7.123648719194389e-05,
      "loss": 7.4093,
      "loss/crossentropy": 2.0915577203035354,
      "loss/hidden": 3.195703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17126586921513082,
      "step": 7530
    },
    {
      "epoch": 0.754,
      "grad_norm": 25.75,
      "grad_norm_var": 2.4098307291666665,
      "learning_rate": 7.090636797006658e-05,
      "loss": 7.3297,
      "loss/crossentropy": 2.198517936468124,
      "loss/hidden": 3.45078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.20143353156745433,
      "step": 7540
    },
    {
      "epoch": 0.755,
      "grad_norm": 25.0,
      "grad_norm_var": 1.9552083333333334,
      "learning_rate": 7.057526756848719e-05,
      "loss": 7.3531,
      "loss/crossentropy": 2.08135302066803,
      "loss/hidden": 3.2609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1978357734158635,
      "step": 7550
    },
    {
      "epoch": 0.756,
      "grad_norm": 27.0,
      "grad_norm_var": 1.8671223958333334,
      "learning_rate": 7.024320641103812e-05,
      "loss": 7.3367,
      "loss/crossentropy": 2.217959429323673,
      "loss/hidden": 3.180078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1959148071706295,
      "step": 7560
    },
    {
      "epoch": 0.757,
      "grad_norm": 27.0,
      "grad_norm_var": 1.3468098958333334,
      "learning_rate": 6.991020498081569e-05,
      "loss": 7.4209,
      "loss/crossentropy": 2.301048618555069,
      "loss/hidden": 3.284375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19627294316887856,
      "step": 7570
    },
    {
      "epoch": 0.758,
      "grad_norm": 25.875,
      "grad_norm_var": 1.1947916666666667,
      "learning_rate": 6.957628381891673e-05,
      "loss": 7.5366,
      "loss/crossentropy": 2.0502139650285245,
      "loss/hidden": 3.297265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.18287195675075055,
      "step": 7580
    },
    {
      "epoch": 0.759,
      "grad_norm": 26.875,
      "grad_norm_var": 1.6488932291666667,
      "learning_rate": 6.924146352317138e-05,
      "loss": 7.3897,
      "loss/crossentropy": 2.0968832850456236,
      "loss/hidden": 3.34375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20618146620690822,
      "step": 7590
    },
    {
      "epoch": 0.76,
      "grad_norm": 25.75,
      "grad_norm_var": 1.8739583333333334,
      "learning_rate": 6.890576474687263e-05,
      "loss": 7.4515,
      "loss/crossentropy": 2.093619105219841,
      "loss/hidden": 3.216015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17545370263978838,
      "step": 7600
    },
    {
      "epoch": 0.761,
      "grad_norm": 24.375,
      "grad_norm_var": 3.160872395833333,
      "learning_rate": 6.856920819750232e-05,
      "loss": 7.3109,
      "loss/crossentropy": 2.1301353871822357,
      "loss/hidden": 3.273046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1940825032070279,
      "step": 7610
    },
    {
      "epoch": 0.762,
      "grad_norm": 24.625,
      "grad_norm_var": 2.948372395833333,
      "learning_rate": 6.823181463545368e-05,
      "loss": 7.4264,
      "loss/crossentropy": 2.2378656387329103,
      "loss/hidden": 3.28671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19179482720792293,
      "step": 7620
    },
    {
      "epoch": 0.763,
      "grad_norm": 25.375,
      "grad_norm_var": 1.6114583333333334,
      "learning_rate": 6.789360487275092e-05,
      "loss": 7.389,
      "loss/crossentropy": 2.1321434706449507,
      "loss/hidden": 3.32421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.21226093731820583,
      "step": 7630
    },
    {
      "epoch": 0.764,
      "grad_norm": 24.25,
      "grad_norm_var": 3.9145833333333333,
      "learning_rate": 6.755459977176533e-05,
      "loss": 7.3101,
      "loss/crossentropy": 2.1095103308558465,
      "loss/hidden": 3.282421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1865893481299281,
      "step": 7640
    },
    {
      "epoch": 0.765,
      "grad_norm": 24.5,
      "grad_norm_var": 1.1551432291666666,
      "learning_rate": 6.721482024392835e-05,
      "loss": 7.3427,
      "loss/crossentropy": 2.2230965688824655,
      "loss/hidden": 3.233203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1860642520710826,
      "step": 7650
    },
    {
      "epoch": 0.766,
      "grad_norm": 37.75,
      "grad_norm_var": 13.4837890625,
      "learning_rate": 6.687428724844179e-05,
      "loss": 7.3843,
      "loss/crossentropy": 2.2692631423473357,
      "loss/hidden": 3.254296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18289674110710621,
      "step": 7660
    },
    {
      "epoch": 0.767,
      "grad_norm": 29.125,
      "grad_norm_var": 2.7866726146420767e+18,
      "learning_rate": 6.653302179098485e-05,
      "loss": 7.3684,
      "loss/crossentropy": 2.073717230185866,
      "loss/hidden": 3.326953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18691302686929703,
      "step": 7670
    },
    {
      "epoch": 0.768,
      "grad_norm": 25.75,
      "grad_norm_var": 2.786672615963632e+18,
      "learning_rate": 6.619104492241848e-05,
      "loss": 7.2984,
      "loss/crossentropy": 2.2619937866926194,
      "loss/hidden": 3.318359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.20932555850595236,
      "step": 7680
    },
    {
      "epoch": 0.769,
      "grad_norm": 24.625,
      "grad_norm_var": 12.002083333333333,
      "learning_rate": 6.584837773748674e-05,
      "loss": 7.2788,
      "loss/crossentropy": 2.1440966367721557,
      "loss/hidden": 3.273046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.19258161056786777,
      "step": 7690
    },
    {
      "epoch": 0.77,
      "grad_norm": 25.25,
      "grad_norm_var": 9.005143229166666,
      "learning_rate": 6.550504137351576e-05,
      "loss": 7.3098,
      "loss/crossentropy": 2.1658431753516196,
      "loss/hidden": 3.184375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18627576418220998,
      "step": 7700
    },
    {
      "epoch": 0.771,
      "grad_norm": 24.375,
      "grad_norm_var": 1.33125,
      "learning_rate": 6.516105700910967e-05,
      "loss": 7.2005,
      "loss/crossentropy": 2.1858869284391402,
      "loss/hidden": 3.18359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18092084676027298,
      "step": 7710
    },
    {
      "epoch": 0.772,
      "grad_norm": 26.125,
      "grad_norm_var": 1.8452473958333333,
      "learning_rate": 6.481644586284442e-05,
      "loss": 7.2082,
      "loss/crossentropy": 2.295649342238903,
      "loss/hidden": 3.193359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1811120670288801,
      "step": 7720
    },
    {
      "epoch": 0.773,
      "grad_norm": 22.625,
      "grad_norm_var": 5.278580729166666,
      "learning_rate": 6.447122919195875e-05,
      "loss": 7.2226,
      "loss/crossentropy": 2.1884284943342207,
      "loss/hidden": 3.201953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17928373999893665,
      "step": 7730
    },
    {
      "epoch": 0.774,
      "grad_norm": 27.0,
      "grad_norm_var": 4.102018229166666,
      "learning_rate": 6.412542829104307e-05,
      "loss": 7.3724,
      "loss/crossentropy": 2.3521720811724665,
      "loss/hidden": 3.23203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1927757915109396,
      "step": 7740
    },
    {
      "epoch": 0.775,
      "grad_norm": 23.5,
      "grad_norm_var": 4.912239583333333,
      "learning_rate": 6.377906449072578e-05,
      "loss": 7.2959,
      "loss/crossentropy": 2.1575791001319886,
      "loss/hidden": 3.27734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19961551688611506,
      "step": 7750
    },
    {
      "epoch": 0.776,
      "grad_norm": 24.375,
      "grad_norm_var": 0.7666015625,
      "learning_rate": 6.343215915635762e-05,
      "loss": 7.1249,
      "loss/crossentropy": 2.1519226878881454,
      "loss/hidden": 3.190625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17577633038163185,
      "step": 7760
    },
    {
      "epoch": 0.777,
      "grad_norm": 24.375,
      "grad_norm_var": 8.60390625,
      "learning_rate": 6.308473368669367e-05,
      "loss": 7.2177,
      "loss/crossentropy": 2.2416551634669304,
      "loss/hidden": 3.16484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1746464787982404,
      "step": 7770
    },
    {
      "epoch": 0.778,
      "grad_norm": 24.5,
      "grad_norm_var": 8.609830729166667,
      "learning_rate": 6.273680951257342e-05,
      "loss": 7.3032,
      "loss/crossentropy": 2.145657110214233,
      "loss/hidden": 3.16953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18332088328897952,
      "step": 7780
    },
    {
      "epoch": 0.779,
      "grad_norm": 24.5,
      "grad_norm_var": 3.525,
      "learning_rate": 6.238840809559884e-05,
      "loss": 7.3272,
      "loss/crossentropy": 2.1870901376008987,
      "loss/hidden": 3.332421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.20739195179194211,
      "step": 7790
    },
    {
      "epoch": 0.78,
      "grad_norm": 25.75,
      "grad_norm_var": 3.0639973958333333,
      "learning_rate": 6.203955092681039e-05,
      "loss": 7.2885,
      "loss/crossentropy": 2.2870864123106003,
      "loss/hidden": 3.258984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19918184783309698,
      "step": 7800
    },
    {
      "epoch": 0.781,
      "grad_norm": 24.25,
      "grad_norm_var": 3.0737770908802703e+18,
      "learning_rate": 6.169025952536154e-05,
      "loss": 7.2785,
      "loss/crossentropy": 2.115164430439472,
      "loss/hidden": 3.187109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.19251083210110664,
      "step": 7810
    },
    {
      "epoch": 0.782,
      "grad_norm": 24.25,
      "grad_norm_var": 2.634375,
      "learning_rate": 6.134055543719121e-05,
      "loss": 7.1323,
      "loss/crossentropy": 2.1837784573435783,
      "loss/hidden": 3.14765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1888807687908411,
      "step": 7820
    },
    {
      "epoch": 0.783,
      "grad_norm": 23.5,
      "grad_norm_var": 1.3385416666666667,
      "learning_rate": 6.0990460233694854e-05,
      "loss": 7.1168,
      "loss/crossentropy": 2.2223884403705596,
      "loss/hidden": 3.20234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18356790579855442,
      "step": 7830
    },
    {
      "epoch": 0.784,
      "grad_norm": 24.25,
      "grad_norm_var": 0.5754557291666667,
      "learning_rate": 6.06399955103937e-05,
      "loss": 7.0934,
      "loss/crossentropy": 2.0307491227984427,
      "loss/hidden": 3.166015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1779999515041709,
      "step": 7840
    },
    {
      "epoch": 0.785,
      "grad_norm": 23.625,
      "grad_norm_var": 31.23515625,
      "learning_rate": 6.0289182885602704e-05,
      "loss": 7.2069,
      "loss/crossentropy": 2.175394120812416,
      "loss/hidden": 3.194921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18147795423865318,
      "step": 7850
    },
    {
      "epoch": 0.786,
      "grad_norm": 26.625,
      "grad_norm_var": 31.609309895833334,
      "learning_rate": 5.993804399909704e-05,
      "loss": 7.1219,
      "loss/crossentropy": 2.184116542339325,
      "loss/hidden": 3.175,
      "loss/jsd": 0.0,
      "loss/logits": 0.18110901936888696,
      "step": 7860
    },
    {
      "epoch": 0.787,
      "grad_norm": 25.0,
      "grad_norm_var": 1.3416666666666666,
      "learning_rate": 5.9586600510777255e-05,
      "loss": 7.2392,
      "loss/crossentropy": 2.1728118672966956,
      "loss/hidden": 3.187890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17548323906958102,
      "step": 7870
    },
    {
      "epoch": 0.788,
      "grad_norm": 23.625,
      "grad_norm_var": 0.6979166666666666,
      "learning_rate": 5.923487409933316e-05,
      "loss": 7.2005,
      "loss/crossentropy": 2.235177829861641,
      "loss/hidden": 3.097265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17446751594543458,
      "step": 7880
    },
    {
      "epoch": 0.789,
      "grad_norm": 24.625,
      "grad_norm_var": 2.2514973958333333,
      "learning_rate": 5.888288646090656e-05,
      "loss": 7.1577,
      "loss/crossentropy": 1.9824412673711778,
      "loss/hidden": 3.108203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16306445542722942,
      "step": 7890
    },
    {
      "epoch": 0.79,
      "grad_norm": 23.25,
      "grad_norm_var": 1.6604166666666667,
      "learning_rate": 5.8530659307753036e-05,
      "loss": 7.1311,
      "loss/crossentropy": 2.2140675693750382,
      "loss/hidden": 3.209765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.19630825575441121,
      "step": 7900
    },
    {
      "epoch": 0.791,
      "grad_norm": 22.5,
      "grad_norm_var": 1.6166015625,
      "learning_rate": 5.817821436690251e-05,
      "loss": 7.0413,
      "loss/crossentropy": 2.13326867967844,
      "loss/hidden": 3.240234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17188256345689296,
      "step": 7910
    },
    {
      "epoch": 0.792,
      "grad_norm": 23.25,
      "grad_norm_var": 0.8625,
      "learning_rate": 5.782557337881911e-05,
      "loss": 7.0504,
      "loss/crossentropy": 2.22673671990633,
      "loss/hidden": 3.1046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18034335430711507,
      "step": 7920
    },
    {
      "epoch": 0.793,
      "grad_norm": 23.25,
      "grad_norm_var": 0.9280598958333334,
      "learning_rate": 5.747275809606006e-05,
      "loss": 7.0115,
      "loss/crossentropy": 2.233048897981644,
      "loss/hidden": 3.1375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17683454230427742,
      "step": 7930
    },
    {
      "epoch": 0.794,
      "grad_norm": 22.625,
      "grad_norm_var": 0.46139322916666664,
      "learning_rate": 5.7119790281933914e-05,
      "loss": 7.0223,
      "loss/crossentropy": 2.1612079739570618,
      "loss/hidden": 3.196875,
      "loss/jsd": 0.0,
      "loss/logits": 0.18148380517959595,
      "step": 7940
    },
    {
      "epoch": 0.795,
      "grad_norm": 23.875,
      "grad_norm_var": 1.7061848958333334,
      "learning_rate": 5.6766691709158096e-05,
      "loss": 7.0717,
      "loss/crossentropy": 2.174719399213791,
      "loss/hidden": 3.202734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17878109030425549,
      "step": 7950
    },
    {
      "epoch": 0.796,
      "grad_norm": 23.625,
      "grad_norm_var": 2.5629557291666667,
      "learning_rate": 5.641348415851577e-05,
      "loss": 7.0505,
      "loss/crossentropy": 2.113456728309393,
      "loss/hidden": 3.148046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17140786703675986,
      "step": 7960
    },
    {
      "epoch": 0.797,
      "grad_norm": 24.375,
      "grad_norm_var": 2.1333333333333333,
      "learning_rate": 5.6060189417512466e-05,
      "loss": 7.1782,
      "loss/crossentropy": 2.111274632811546,
      "loss/hidden": 3.24453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.18207379188388587,
      "step": 7970
    },
    {
      "epoch": 0.798,
      "grad_norm": 23.5,
      "grad_norm_var": 1.2254557291666666,
      "learning_rate": 5.570682927903194e-05,
      "loss": 7.0835,
      "loss/crossentropy": 2.209978663921356,
      "loss/hidden": 3.193359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18822428286075593,
      "step": 7980
    },
    {
      "epoch": 0.799,
      "grad_norm": 23.25,
      "grad_norm_var": 0.6559895833333333,
      "learning_rate": 5.535342553999201e-05,
      "loss": 7.0226,
      "loss/crossentropy": 2.2532318040728567,
      "loss/hidden": 3.062890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17306482810527085,
      "step": 7990
    },
    {
      "epoch": 0.8,
      "grad_norm": 22.0,
      "grad_norm_var": 1.3056640625,
      "learning_rate": 5.500000000000001e-05,
      "loss": 7.0514,
      "loss/crossentropy": 2.112633790075779,
      "loss/hidden": 3.14453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17193403597921134,
      "step": 8000
    },
    {
      "epoch": 0.801,
      "grad_norm": 23.875,
      "grad_norm_var": 0.9291666666666667,
      "learning_rate": 5.464657446000802e-05,
      "loss": 7.1599,
      "loss/crossentropy": 2.2727265626192095,
      "loss/hidden": 3.120703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1767945010215044,
      "step": 8010
    },
    {
      "epoch": 0.802,
      "grad_norm": 21.75,
      "grad_norm_var": 1.23515625,
      "learning_rate": 5.429317072096808e-05,
      "loss": 6.9704,
      "loss/crossentropy": 2.2333207607269285,
      "loss/hidden": 3.048828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16404416598379612,
      "step": 8020
    },
    {
      "epoch": 0.803,
      "grad_norm": 24.125,
      "grad_norm_var": 1.3072265625,
      "learning_rate": 5.393981058248755e-05,
      "loss": 7.0143,
      "loss/crossentropy": 2.0972195714712143,
      "loss/hidden": 3.14921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1729634340852499,
      "step": 8030
    },
    {
      "epoch": 0.804,
      "grad_norm": 22.25,
      "grad_norm_var": 1.4129557291666666,
      "learning_rate": 5.358651584148423e-05,
      "loss": 7.0155,
      "loss/crossentropy": 1.9586305670440196,
      "loss/hidden": 3.026171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15513296276330948,
      "step": 8040
    },
    {
      "epoch": 0.805,
      "grad_norm": 23.125,
      "grad_norm_var": 0.7759765625,
      "learning_rate": 5.3233308290841935e-05,
      "loss": 7.0464,
      "loss/crossentropy": 2.170274239778519,
      "loss/hidden": 3.231640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.17621967550367118,
      "step": 8050
    },
    {
      "epoch": 0.806,
      "grad_norm": 22.875,
      "grad_norm_var": 1.0160807291666667,
      "learning_rate": 5.288020971806609e-05,
      "loss": 7.0331,
      "loss/crossentropy": 2.2559945076704024,
      "loss/hidden": 3.09375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17008322682231664,
      "step": 8060
    },
    {
      "epoch": 0.807,
      "grad_norm": 22.625,
      "grad_norm_var": 15.751041666666667,
      "learning_rate": 5.252724190393995e-05,
      "loss": 7.0494,
      "loss/crossentropy": 2.177362951636314,
      "loss/hidden": 3.099609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17315508853644132,
      "step": 8070
    },
    {
      "epoch": 0.808,
      "grad_norm": 21.75,
      "grad_norm_var": 1.6014973958333334,
      "learning_rate": 5.2174426621180906e-05,
      "loss": 6.9141,
      "loss/crossentropy": 2.2173873841762544,
      "loss/hidden": 3.11171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17051471378654243,
      "step": 8080
    },
    {
      "epoch": 0.809,
      "grad_norm": 22.375,
      "grad_norm_var": 8.440559895833333,
      "learning_rate": 5.182178563309751e-05,
      "loss": 7.0752,
      "loss/crossentropy": 2.129284977912903,
      "loss/hidden": 3.13203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16524926889687777,
      "step": 8090
    },
    {
      "epoch": 0.81,
      "grad_norm": 22.5,
      "grad_norm_var": 9.026822916666667,
      "learning_rate": 5.1469340692246995e-05,
      "loss": 6.8317,
      "loss/crossentropy": 2.1604677557945253,
      "loss/hidden": 3.07734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16097149150446058,
      "step": 8100
    },
    {
      "epoch": 0.811,
      "grad_norm": 22.375,
      "grad_norm_var": 1.4184895833333333,
      "learning_rate": 5.1117113539093456e-05,
      "loss": 7.0113,
      "loss/crossentropy": 2.253964813053608,
      "loss/hidden": 3.101171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17060668766498566,
      "step": 8110
    },
    {
      "epoch": 0.812,
      "grad_norm": 23.125,
      "grad_norm_var": 4728.34765625,
      "learning_rate": 5.076512590066685e-05,
      "loss": 7.0753,
      "loss/crossentropy": 2.332234174013138,
      "loss/hidden": 2.9859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15933743081986904,
      "step": 8120
    },
    {
      "epoch": 0.813,
      "grad_norm": 23.125,
      "grad_norm_var": 4740.253125,
      "learning_rate": 5.0413399489222755e-05,
      "loss": 6.9925,
      "loss/crossentropy": 2.2431288778781893,
      "loss/hidden": 3.177734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.18101395089179279,
      "step": 8130
    },
    {
      "epoch": 0.814,
      "grad_norm": 21.75,
      "grad_norm_var": 0.7171223958333334,
      "learning_rate": 5.006195600090297e-05,
      "loss": 6.8942,
      "loss/crossentropy": 2.156078750640154,
      "loss/hidden": 3.068359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1705356423743069,
      "step": 8140
    },
    {
      "epoch": 0.815,
      "grad_norm": 22.5,
      "grad_norm_var": 14.7931640625,
      "learning_rate": 4.9710817114397314e-05,
      "loss": 6.9409,
      "loss/crossentropy": 2.2799797251820566,
      "loss/hidden": 3.237890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.2035599894821644,
      "step": 8150
    },
    {
      "epoch": 0.816,
      "grad_norm": 22.0,
      "grad_norm_var": 16.298372395833333,
      "learning_rate": 4.936000448960631e-05,
      "loss": 6.9298,
      "loss/crossentropy": 2.193594951927662,
      "loss/hidden": 3.062890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16013095155358315,
      "step": 8160
    },
    {
      "epoch": 0.817,
      "grad_norm": 22.0,
      "grad_norm_var": 2.030989583333333,
      "learning_rate": 4.9009539766305157e-05,
      "loss": 6.8087,
      "loss/crossentropy": 2.0724834606051443,
      "loss/hidden": 3.053125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16694455780088902,
      "step": 8170
    },
    {
      "epoch": 0.818,
      "grad_norm": 22.0,
      "grad_norm_var": 3.192559534480346e+18,
      "learning_rate": 4.865944456280879e-05,
      "loss": 7.0056,
      "loss/crossentropy": 1.9885679975152015,
      "loss/hidden": 3.039453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15490864738821983,
      "step": 8180
    },
    {
      "epoch": 0.819,
      "grad_norm": 22.0,
      "grad_norm_var": 0.7572916666666667,
      "learning_rate": 4.830974047463847e-05,
      "loss": 6.951,
      "loss/crossentropy": 2.2198346734046934,
      "loss/hidden": 3.1046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1715863322839141,
      "step": 8190
    },
    {
      "epoch": 0.82,
      "grad_norm": 22.375,
      "grad_norm_var": 70.35774739583333,
      "learning_rate": 4.7960449073189606e-05,
      "loss": 6.9039,
      "loss/crossentropy": 2.2345028445124626,
      "loss/hidden": 3.020703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16092772465199232,
      "step": 8200
    },
    {
      "epoch": 0.821,
      "grad_norm": 22.25,
      "grad_norm_var": 0.9499348958333333,
      "learning_rate": 4.761159190440117e-05,
      "loss": 6.8918,
      "loss/crossentropy": 2.25988983809948,
      "loss/hidden": 2.99140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.162513430416584,
      "step": 8210
    },
    {
      "epoch": 0.822,
      "grad_norm": 20.375,
      "grad_norm_var": 2.3824041854824177e+18,
      "learning_rate": 4.7263190487426564e-05,
      "loss": 6.9202,
      "loss/crossentropy": 2.2631596639752387,
      "loss/hidden": 3.311328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17134508192539216,
      "step": 8220
    },
    {
      "epoch": 0.823,
      "grad_norm": 21.75,
      "grad_norm_var": 9.499739583333334,
      "learning_rate": 4.691526631330634e-05,
      "loss": 6.8251,
      "loss/crossentropy": 2.1176006376743315,
      "loss/hidden": 2.975390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15703603606671096,
      "step": 8230
    },
    {
      "epoch": 0.824,
      "grad_norm": 22.25,
      "grad_norm_var": 8.7369140625,
      "learning_rate": 4.6567840843642384e-05,
      "loss": 6.8875,
      "loss/crossentropy": 2.1988213270902635,
      "loss/hidden": 3.0828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15953818447887896,
      "step": 8240
    },
    {
      "epoch": 0.825,
      "grad_norm": 21.625,
      "grad_norm_var": 0.8999348958333333,
      "learning_rate": 4.6220935509274235e-05,
      "loss": 6.954,
      "loss/crossentropy": 2.122755715250969,
      "loss/hidden": 3.181640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1846215335652232,
      "step": 8250
    },
    {
      "epoch": 0.826,
      "grad_norm": 21.375,
      "grad_norm_var": 0.9421223958333333,
      "learning_rate": 4.587457170895696e-05,
      "loss": 6.7388,
      "loss/crossentropy": 2.1177376963198187,
      "loss/hidden": 3.015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16393411150202156,
      "step": 8260
    },
    {
      "epoch": 0.827,
      "grad_norm": 22.25,
      "grad_norm_var": 2.7309405742616806e+18,
      "learning_rate": 4.552877080804126e-05,
      "loss": 6.8162,
      "loss/crossentropy": 2.2066329643130302,
      "loss/hidden": 2.94140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15212294571101664,
      "step": 8270
    },
    {
      "epoch": 0.828,
      "grad_norm": 21.5,
      "grad_norm_var": 12.398958333333333,
      "learning_rate": 4.5183554137155606e-05,
      "loss": 6.9173,
      "loss/crossentropy": 2.264900441467762,
      "loss/hidden": 3.0015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16246109874919057,
      "step": 8280
    },
    {
      "epoch": 0.829,
      "grad_norm": 21.625,
      "grad_norm_var": 12.623893229166667,
      "learning_rate": 4.483894299089034e-05,
      "loss": 6.9313,
      "loss/crossentropy": 2.2778148487210275,
      "loss/hidden": 3.070703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17395032718777656,
      "step": 8290
    },
    {
      "epoch": 0.83,
      "grad_norm": 35.25,
      "grad_norm_var": 24.608333333333334,
      "learning_rate": 4.4494958626484276e-05,
      "loss": 6.6172,
      "loss/crossentropy": 1.8980232715606689,
      "loss/hidden": 3.024609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14263095702044665,
      "step": 8300
    },
    {
      "epoch": 0.831,
      "grad_norm": 21.5,
      "grad_norm_var": 13.6478515625,
      "learning_rate": 4.415162226251327e-05,
      "loss": 6.778,
      "loss/crossentropy": 2.1019492864608766,
      "loss/hidden": 3.0203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16046333201229573,
      "step": 8310
    },
    {
      "epoch": 0.832,
      "grad_norm": 33.75,
      "grad_norm_var": 20.048958333333335,
      "learning_rate": 4.380895507758155e-05,
      "loss": 6.8104,
      "loss/crossentropy": 2.197945240139961,
      "loss/hidden": 3.125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17097027003765106,
      "step": 8320
    },
    {
      "epoch": 0.833,
      "grad_norm": 20.375,
      "grad_norm_var": 10.97265625,
      "learning_rate": 4.346697820901515e-05,
      "loss": 6.8388,
      "loss/crossentropy": 2.1463077016174794,
      "loss/hidden": 3.020703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.14912782991304993,
      "step": 8330
    },
    {
      "epoch": 0.834,
      "grad_norm": 19.125,
      "grad_norm_var": 14.505143229166666,
      "learning_rate": 4.312571275155823e-05,
      "loss": 6.7567,
      "loss/crossentropy": 2.1706424072384833,
      "loss/hidden": 2.925390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15102706514298916,
      "step": 8340
    },
    {
      "epoch": 0.835,
      "grad_norm": 22.5,
      "grad_norm_var": 10.55,
      "learning_rate": 4.278517975607167e-05,
      "loss": 6.6799,
      "loss/crossentropy": 2.2084620505571366,
      "loss/hidden": 2.93203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.14671239536255598,
      "step": 8350
    },
    {
      "epoch": 0.836,
      "grad_norm": 20.25,
      "grad_norm_var": 9.618684895833333,
      "learning_rate": 4.2445400228234686e-05,
      "loss": 6.8098,
      "loss/crossentropy": 2.24121166318655,
      "loss/hidden": 3.064453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17096446882933378,
      "step": 8360
    },
    {
      "epoch": 0.837,
      "grad_norm": 37.75,
      "grad_norm_var": 21.941080729166668,
      "learning_rate": 4.2106395127249076e-05,
      "loss": 6.778,
      "loss/crossentropy": 2.196937990188599,
      "loss/hidden": 2.94296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16012835130095482,
      "step": 8370
    },
    {
      "epoch": 0.838,
      "grad_norm": 21.0,
      "grad_norm_var": 22.3791015625,
      "learning_rate": 4.176818536454633e-05,
      "loss": 6.7274,
      "loss/crossentropy": 2.2157079711556436,
      "loss/hidden": 3.119921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.179295320995152,
      "step": 8380
    },
    {
      "epoch": 0.839,
      "grad_norm": 21.125,
      "grad_norm_var": 11.954622395833333,
      "learning_rate": 4.14307918024977e-05,
      "loss": 6.7335,
      "loss/crossentropy": 1.9891890831291676,
      "loss/hidden": 3.04453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15178165752440692,
      "step": 8390
    },
    {
      "epoch": 0.84,
      "grad_norm": 20.75,
      "grad_norm_var": 12.85390625,
      "learning_rate": 4.109423525312738e-05,
      "loss": 6.8472,
      "loss/crossentropy": 2.2481492303311823,
      "loss/hidden": 3.0734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1720767416059971,
      "step": 8400
    },
    {
      "epoch": 0.841,
      "grad_norm": 20.125,
      "grad_norm_var": 12.489322916666667,
      "learning_rate": 4.0758536476828624e-05,
      "loss": 6.753,
      "loss/crossentropy": 2.162366569042206,
      "loss/hidden": 3.02421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16984781231731177,
      "step": 8410
    },
    {
      "epoch": 0.842,
      "grad_norm": 20.375,
      "grad_norm_var": 13.0087890625,
      "learning_rate": 4.042371618108329e-05,
      "loss": 6.6159,
      "loss/crossentropy": 2.0167182251811027,
      "loss/hidden": 2.815234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.13584452606737613,
      "step": 8420
    },
    {
      "epoch": 0.843,
      "grad_norm": 18.875,
      "grad_norm_var": 10.72890625,
      "learning_rate": 4.0089795019184303e-05,
      "loss": 6.7834,
      "loss/crossentropy": 2.1324578002095222,
      "loss/hidden": 3.08828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1533666567876935,
      "step": 8430
    },
    {
      "epoch": 0.844,
      "grad_norm": 19.25,
      "grad_norm_var": 8.525,
      "learning_rate": 3.9756793588961896e-05,
      "loss": 6.7359,
      "loss/crossentropy": 2.101317623257637,
      "loss/hidden": 2.938671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15315671488642693,
      "step": 8440
    },
    {
      "epoch": 0.845,
      "grad_norm": 21.625,
      "grad_norm_var": 6.635416666666667,
      "learning_rate": 3.942473243151281e-05,
      "loss": 6.7393,
      "loss/crossentropy": 2.094627208262682,
      "loss/hidden": 3.019921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15584880905225873,
      "step": 8450
    },
    {
      "epoch": 0.846,
      "grad_norm": 20.125,
      "grad_norm_var": 4.204622395833334,
      "learning_rate": 3.9093632029933435e-05,
      "loss": 6.721,
      "loss/crossentropy": 2.2385535068809985,
      "loss/hidden": 3.049609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1660825703293085,
      "step": 8460
    },
    {
      "epoch": 0.847,
      "grad_norm": 19.0,
      "grad_norm_var": 10.87890625,
      "learning_rate": 3.8763512808056125e-05,
      "loss": 6.5906,
      "loss/crossentropy": 2.1519340083003042,
      "loss/hidden": 2.949609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15027154460549355,
      "step": 8470
    },
    {
      "epoch": 0.848,
      "grad_norm": 29.625,
      "grad_norm_var": 13.39140625,
      "learning_rate": 3.843439512918949e-05,
      "loss": 6.7149,
      "loss/crossentropy": 2.233719590306282,
      "loss/hidden": 3.010546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1669364469125867,
      "step": 8480
    },
    {
      "epoch": 0.849,
      "grad_norm": 20.5,
      "grad_norm_var": 8.970768229166667,
      "learning_rate": 3.810629929486226e-05,
      "loss": 6.7086,
      "loss/crossentropy": 2.226981985569,
      "loss/hidden": 3.01484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16180209238082172,
      "step": 8490
    },
    {
      "epoch": 0.85,
      "grad_norm": 19.25,
      "grad_norm_var": 8.365559895833334,
      "learning_rate": 3.777924554357096e-05,
      "loss": 6.6027,
      "loss/crossentropy": 2.080954398959875,
      "loss/hidden": 2.93515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1519384733401239,
      "step": 8500
    },
    {
      "epoch": 0.851,
      "grad_norm": 27.625,
      "grad_norm_var": 10.964322916666667,
      "learning_rate": 3.7453254049531536e-05,
      "loss": 6.816,
      "loss/crossentropy": 2.1905113458633423,
      "loss/hidden": 3.0546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16897472254931928,
      "step": 8510
    },
    {
      "epoch": 0.852,
      "grad_norm": 26.625,
      "grad_norm_var": 6.923372395833334,
      "learning_rate": 3.712834492143488e-05,
      "loss": 6.7829,
      "loss/crossentropy": 2.037618358433247,
      "loss/hidden": 2.925,
      "loss/jsd": 0.0,
      "loss/logits": 0.14524593306705355,
      "step": 8520
    },
    {
      "epoch": 0.853,
      "grad_norm": 28.875,
      "grad_norm_var": 8.1009765625,
      "learning_rate": 3.6804538201206483e-05,
      "loss": 6.7153,
      "loss/crossentropy": 2.2185431465506555,
      "loss/hidden": 2.95703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16105205966159702,
      "step": 8530
    },
    {
      "epoch": 0.854,
      "grad_norm": 19.375,
      "grad_norm_var": 9.983072916666666,
      "learning_rate": 3.648185386277011e-05,
      "loss": 6.6926,
      "loss/crossentropy": 1.935778360068798,
      "loss/hidden": 3.0765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15904303397983313,
      "step": 8540
    },
    {
      "epoch": 0.855,
      "grad_norm": 19.625,
      "grad_norm_var": 7.237434895833333,
      "learning_rate": 3.616031181081575e-05,
      "loss": 6.7702,
      "loss/crossentropy": 2.147594654560089,
      "loss/hidden": 3.08125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17960532996803522,
      "step": 8550
    },
    {
      "epoch": 0.856,
      "grad_norm": 20.25,
      "grad_norm_var": 17.895833333333332,
      "learning_rate": 3.583993187957173e-05,
      "loss": 6.6853,
      "loss/crossentropy": 2.1908980406820775,
      "loss/hidden": 2.966796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1526848678011447,
      "step": 8560
    },
    {
      "epoch": 0.857,
      "grad_norm": 19.875,
      "grad_norm_var": 20.8337890625,
      "learning_rate": 3.5520733831581396e-05,
      "loss": 6.6626,
      "loss/crossentropy": 2.1303799264132977,
      "loss/hidden": 3.020703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15846395073458552,
      "step": 8570
    },
    {
      "epoch": 0.858,
      "grad_norm": 28.75,
      "grad_norm_var": 14.406705729166667,
      "learning_rate": 3.520273735648382e-05,
      "loss": 6.7556,
      "loss/crossentropy": 2.3622685760259627,
      "loss/hidden": 2.95625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1620245823636651,
      "step": 8580
    },
    {
      "epoch": 0.859,
      "grad_norm": 19.625,
      "grad_norm_var": 8.196809895833333,
      "learning_rate": 3.4885962069799554e-05,
      "loss": 6.7115,
      "loss/crossentropy": 2.1019712671637536,
      "loss/hidden": 2.92421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14273924147710204,
      "step": 8590
    },
    {
      "epoch": 0.86,
      "grad_norm": 18.5,
      "grad_norm_var": 6.0806640625,
      "learning_rate": 3.45704275117204e-05,
      "loss": 6.6569,
      "loss/crossentropy": 2.1646031074225904,
      "loss/hidden": 3.039453125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1613057630136609,
      "step": 8600
    },
    {
      "epoch": 0.861,
      "grad_norm": 17.625,
      "grad_norm_var": 6.801822916666667,
      "learning_rate": 3.42561531459042e-05,
      "loss": 6.7693,
      "loss/crossentropy": 2.0148328594863414,
      "loss/hidden": 3.09140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16505753956735134,
      "step": 8610
    },
    {
      "epoch": 0.862,
      "grad_norm": 18.75,
      "grad_norm_var": 7.334830729166667,
      "learning_rate": 3.394315835827421e-05,
      "loss": 6.7534,
      "loss/crossentropy": 2.3073144420981406,
      "loss/hidden": 2.981640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15456721372902393,
      "step": 8620
    },
    {
      "epoch": 0.863,
      "grad_norm": 18.625,
      "grad_norm_var": 6.7712890625,
      "learning_rate": 3.363146245582325e-05,
      "loss": 6.6867,
      "loss/crossentropy": 2.1204227074980735,
      "loss/hidden": 2.97578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15875139143317937,
      "step": 8630
    },
    {
      "epoch": 0.864,
      "grad_norm": 27.75,
      "grad_norm_var": 9.9619140625,
      "learning_rate": 3.3321084665422807e-05,
      "loss": 6.7054,
      "loss/crossentropy": 2.1530178070068358,
      "loss/hidden": 3.030078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15609838385134936,
      "step": 8640
    },
    {
      "epoch": 0.865,
      "grad_norm": 19.5,
      "grad_norm_var": 6.264583333333333,
      "learning_rate": 3.301204413263704e-05,
      "loss": 6.7304,
      "loss/crossentropy": 2.134761989861727,
      "loss/hidden": 2.95859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15059877443127334,
      "step": 8650
    },
    {
      "epoch": 0.866,
      "grad_norm": 19.375,
      "grad_norm_var": 61.580989583333334,
      "learning_rate": 3.270435992054166e-05,
      "loss": 6.7185,
      "loss/crossentropy": 2.227277898788452,
      "loss/hidden": 2.878515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15938994903117418,
      "step": 8660
    },
    {
      "epoch": 0.867,
      "grad_norm": 19.5,
      "grad_norm_var": 66.7259765625,
      "learning_rate": 3.239805100854823e-05,
      "loss": 6.7492,
      "loss/crossentropy": 2.2605167113244535,
      "loss/hidden": 2.892578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1458374190144241,
      "step": 8670
    },
    {
      "epoch": 0.868,
      "grad_norm": 20.125,
      "grad_norm_var": 11.395247395833334,
      "learning_rate": 3.2093136291233296e-05,
      "loss": 6.5492,
      "loss/crossentropy": 1.981280042976141,
      "loss/hidden": 2.947265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.14381110519170762,
      "step": 8680
    },
    {
      "epoch": 0.869,
      "grad_norm": 20.125,
      "grad_norm_var": 7.0509765625,
      "learning_rate": 3.178963457717287e-05,
      "loss": 6.7433,
      "loss/crossentropy": 2.047644779831171,
      "loss/hidden": 3.013671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1535273300483823,
      "step": 8690
    },
    {
      "epoch": 0.87,
      "grad_norm": 19.0,
      "grad_norm_var": 4.912239583333333,
      "learning_rate": 3.1487564587782306e-05,
      "loss": 6.6896,
      "loss/crossentropy": 2.097446948289871,
      "loss/hidden": 2.985546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15309033393859864,
      "step": 8700
    },
    {
      "epoch": 0.871,
      "grad_norm": 19.875,
      "grad_norm_var": 4.342643229166667,
      "learning_rate": 3.118694495616143e-05,
      "loss": 6.6671,
      "loss/crossentropy": 2.203226202726364,
      "loss/hidden": 2.9078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1557199139147997,
      "step": 8710
    },
    {
      "epoch": 0.872,
      "grad_norm": 19.25,
      "grad_norm_var": 7.030208333333333,
      "learning_rate": 3.088779422594514e-05,
      "loss": 6.7012,
      "loss/crossentropy": 2.27069154381752,
      "loss/hidden": 3.011328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16882256120443345,
      "step": 8720
    },
    {
      "epoch": 0.873,
      "grad_norm": 18.25,
      "grad_norm_var": 6.295768229166667,
      "learning_rate": 3.059013085015967e-05,
      "loss": 6.7011,
      "loss/crossentropy": 2.1291425973176956,
      "loss/hidden": 3.0296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15904739797115325,
      "step": 8730
    },
    {
      "epoch": 0.874,
      "grad_norm": 18.75,
      "grad_norm_var": 7.843489583333334,
      "learning_rate": 3.0293973190084068e-05,
      "loss": 6.5634,
      "loss/crossentropy": 2.0249536350369453,
      "loss/hidden": 2.866796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14526209011673927,
      "step": 8740
    },
    {
      "epoch": 0.875,
      "grad_norm": 19.25,
      "grad_norm_var": 24.33125,
      "learning_rate": 2.9999339514117912e-05,
      "loss": 6.6656,
      "loss/crossentropy": 2.2232491344213487,
      "loss/hidden": 2.915625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15606734082102774,
      "step": 8750
    },
    {
      "epoch": 0.876,
      "grad_norm": 17.5,
      "grad_norm_var": 25.318684895833332,
      "learning_rate": 2.9706247996654137e-05,
      "loss": 6.6062,
      "loss/crossentropy": 2.239446271210909,
      "loss/hidden": 2.93203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1495282245334238,
      "step": 8760
    },
    {
      "epoch": 0.877,
      "grad_norm": 20.0,
      "grad_norm_var": 10.77265625,
      "learning_rate": 2.9414716716958125e-05,
      "loss": 6.7024,
      "loss/crossentropy": 2.199841946363449,
      "loss/hidden": 2.98984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16247558798640965,
      "step": 8770
    },
    {
      "epoch": 0.878,
      "grad_norm": 19.125,
      "grad_norm_var": 8.998958333333333,
      "learning_rate": 2.9124763658052478e-05,
      "loss": 6.782,
      "loss/crossentropy": 2.1751702331006526,
      "loss/hidden": 2.92890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1545461408793926,
      "step": 8780
    },
    {
      "epoch": 0.879,
      "grad_norm": 19.75,
      "grad_norm_var": 6.75390625,
      "learning_rate": 2.8836406705607627e-05,
      "loss": 6.6353,
      "loss/crossentropy": 2.1896474927663805,
      "loss/hidden": 3.0078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1576914018020034,
      "step": 8790
    },
    {
      "epoch": 0.88,
      "grad_norm": 18.625,
      "grad_norm_var": 6.298372395833334,
      "learning_rate": 2.854966364683872e-05,
      "loss": 6.6865,
      "loss/crossentropy": 2.1086228668689726,
      "loss/hidden": 2.943359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14896233137696982,
      "step": 8800
    },
    {
      "epoch": 0.881,
      "grad_norm": 18.125,
      "grad_norm_var": 8.4791015625,
      "learning_rate": 2.826455216940827e-05,
      "loss": 6.6194,
      "loss/crossentropy": 1.9843620665371418,
      "loss/hidden": 2.90234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.13960447236895562,
      "step": 8810
    },
    {
      "epoch": 0.882,
      "grad_norm": 18.75,
      "grad_norm_var": 8.501822916666667,
      "learning_rate": 2.798108986033523e-05,
      "loss": 6.6478,
      "loss/crossentropy": 2.1602120816707613,
      "loss/hidden": 2.89765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1467864315956831,
      "step": 8820
    },
    {
      "epoch": 0.883,
      "grad_norm": 20.0,
      "grad_norm_var": 14.489518229166666,
      "learning_rate": 2.769929420491002e-05,
      "loss": 6.5831,
      "loss/crossentropy": 2.1451182171702383,
      "loss/hidden": 3.0640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1538067078217864,
      "step": 8830
    },
    {
      "epoch": 0.884,
      "grad_norm": 28.125,
      "grad_norm_var": 11.8650390625,
      "learning_rate": 2.741918258561607e-05,
      "loss": 6.6807,
      "loss/crossentropy": 2.2480786412954332,
      "loss/hidden": 2.9171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15250301491469145,
      "step": 8840
    },
    {
      "epoch": 0.885,
      "grad_norm": 20.125,
      "grad_norm_var": 5.908333333333333,
      "learning_rate": 2.7140772281057468e-05,
      "loss": 6.7234,
      "loss/crossentropy": 2.122820168733597,
      "loss/hidden": 2.941796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15208065062761306,
      "step": 8850
    },
    {
      "epoch": 0.886,
      "grad_norm": 20.5,
      "grad_norm_var": 5.928580729166667,
      "learning_rate": 2.6864080464893282e-05,
      "loss": 6.729,
      "loss/crossentropy": 2.239290738105774,
      "loss/hidden": 2.94921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16029757130891084,
      "step": 8860
    },
    {
      "epoch": 0.887,
      "grad_norm": 17.25,
      "grad_norm_var": 7.351822916666666,
      "learning_rate": 2.6589124204778006e-05,
      "loss": 6.716,
      "loss/crossentropy": 2.0530457444489003,
      "loss/hidden": 3.00390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15680783092975617,
      "step": 8870
    },
    {
      "epoch": 0.888,
      "grad_norm": 20.75,
      "grad_norm_var": 1.4462890625,
      "learning_rate": 2.6315920461308964e-05,
      "loss": 6.6087,
      "loss/crossentropy": 1.8908826246857644,
      "loss/hidden": 2.987109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15084341559559106,
      "step": 8880
    },
    {
      "epoch": 0.889,
      "grad_norm": 18.25,
      "grad_norm_var": 0.871875,
      "learning_rate": 2.6044486086979953e-05,
      "loss": 6.7122,
      "loss/crossentropy": 2.1536234870553015,
      "loss/hidden": 3.026171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15791046461090447,
      "step": 8890
    },
    {
      "epoch": 0.89,
      "grad_norm": 20.125,
      "grad_norm_var": 0.9384765625,
      "learning_rate": 2.577483782514174e-05,
      "loss": 6.7238,
      "loss/crossentropy": 2.2359716653823853,
      "loss/hidden": 3.04921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1683175450190902,
      "step": 8900
    },
    {
      "epoch": 0.891,
      "grad_norm": 20.0,
      "grad_norm_var": 1.2202473958333333,
      "learning_rate": 2.550699230896928e-05,
      "loss": 6.6606,
      "loss/crossentropy": 2.084880957007408,
      "loss/hidden": 3.04296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1634566718712449,
      "step": 8910
    },
    {
      "epoch": 0.892,
      "grad_norm": 19.25,
      "grad_norm_var": 1.1238932291666666,
      "learning_rate": 2.5240966060435677e-05,
      "loss": 6.6584,
      "loss/crossentropy": 2.0426858074963095,
      "loss/hidden": 2.9921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14660524372011424,
      "step": 8920
    },
    {
      "epoch": 0.893,
      "grad_norm": 17.5,
      "grad_norm_var": 0.8619140625,
      "learning_rate": 2.497677548929301e-05,
      "loss": 6.6837,
      "loss/crossentropy": 2.171066698431969,
      "loss/hidden": 3.036328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16031729001551867,
      "step": 8930
    },
    {
      "epoch": 0.894,
      "grad_norm": 19.25,
      "grad_norm_var": 0.73125,
      "learning_rate": 2.4714436892060213e-05,
      "loss": 6.6919,
      "loss/crossentropy": 2.1432586669921876,
      "loss/hidden": 2.93359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14423664398491381,
      "step": 8940
    },
    {
      "epoch": 0.895,
      "grad_norm": 23.5,
      "grad_norm_var": 1.6395833333333334,
      "learning_rate": 2.445396645101762e-05,
      "loss": 6.6747,
      "loss/crossentropy": 2.170257590711117,
      "loss/hidden": 2.941015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15344964815303683,
      "step": 8950
    },
    {
      "epoch": 0.896,
      "grad_norm": 18.75,
      "grad_norm_var": 1.85390625,
      "learning_rate": 2.4195380233209008e-05,
      "loss": 6.6903,
      "loss/crossentropy": 2.2802744664251806,
      "loss/hidden": 3.00078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16574028157629073,
      "step": 8960
    },
    {
      "epoch": 0.897,
      "grad_norm": 18.25,
      "grad_norm_var": 1.1160807291666666,
      "learning_rate": 2.393869418945033e-05,
      "loss": 6.6492,
      "loss/crossentropy": 2.2105372130870817,
      "loss/hidden": 2.98671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15795816536992788,
      "step": 8970
    },
    {
      "epoch": 0.898,
      "grad_norm": 20.0,
      "grad_norm_var": 0.83515625,
      "learning_rate": 2.3683924153345856e-05,
      "loss": 6.6716,
      "loss/crossentropy": 2.190083122253418,
      "loss/hidden": 2.91484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14693394694477319,
      "step": 8980
    },
    {
      "epoch": 0.899,
      "grad_norm": 18.5,
      "grad_norm_var": 1.5457682291666666,
      "learning_rate": 2.3431085840311496e-05,
      "loss": 6.652,
      "loss/crossentropy": 2.168794426321983,
      "loss/hidden": 2.962109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15159044261090457,
      "step": 8990
    },
    {
      "epoch": 0.9,
      "grad_norm": 19.0,
      "grad_norm_var": 0.3184895833333333,
      "learning_rate": 2.3180194846605367e-05,
      "loss": 6.6702,
      "loss/crossentropy": 2.0840190701186656,
      "loss/hidden": 2.9046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14546455312520265,
      "step": 9000
    },
    {
      "epoch": 0.901,
      "grad_norm": 17.125,
      "grad_norm_var": 12.045768229166667,
      "learning_rate": 2.2931266648365763e-05,
      "loss": 6.551,
      "loss/crossentropy": 2.249745097756386,
      "loss/hidden": 2.95234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14781816666945816,
      "step": 9010
    },
    {
      "epoch": 0.902,
      "grad_norm": 18.0,
      "grad_norm_var": 5.9103515625,
      "learning_rate": 2.268431660065651e-05,
      "loss": 6.6396,
      "loss/crossentropy": 2.1970891691744328,
      "loss/hidden": 2.925390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.14704765304923056,
      "step": 9020
    },
    {
      "epoch": 0.903,
      "grad_norm": 19.25,
      "grad_norm_var": 2.7354166666666666,
      "learning_rate": 2.2439359936519788e-05,
      "loss": 6.5875,
      "loss/crossentropy": 2.03671035990119,
      "loss/hidden": 3.018359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15092918761074542,
      "step": 9030
    },
    {
      "epoch": 0.904,
      "grad_norm": 19.625,
      "grad_norm_var": 1.2052083333333334,
      "learning_rate": 2.219641176603649e-05,
      "loss": 6.6755,
      "loss/crossentropy": 2.090250726044178,
      "loss/hidden": 2.977734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15508323018439113,
      "step": 9040
    },
    {
      "epoch": 0.905,
      "grad_norm": 18.75,
      "grad_norm_var": 1.5997395833333334,
      "learning_rate": 2.195548707539416e-05,
      "loss": 6.5437,
      "loss/crossentropy": 2.1223932191729546,
      "loss/hidden": 2.994140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1513491977006197,
      "step": 9050
    },
    {
      "epoch": 0.906,
      "grad_norm": 17.875,
      "grad_norm_var": 1.0489583333333334,
      "learning_rate": 2.1716600725962562e-05,
      "loss": 6.7282,
      "loss/crossentropy": 2.2691319927573206,
      "loss/hidden": 2.9421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15906442627310752,
      "step": 9060
    },
    {
      "epoch": 0.907,
      "grad_norm": 17.375,
      "grad_norm_var": 2.015625,
      "learning_rate": 2.147976745337702e-05,
      "loss": 6.744,
      "loss/crossentropy": 2.1490016683936117,
      "loss/hidden": 3.001953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16720277369022368,
      "step": 9070
    },
    {
      "epoch": 0.908,
      "grad_norm": 18.875,
      "grad_norm_var": 0.653125,
      "learning_rate": 2.1245001866629322e-05,
      "loss": 6.5938,
      "loss/crossentropy": 2.0249759793281554,
      "loss/hidden": 2.986328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15444286894053222,
      "step": 9080
    },
    {
      "epoch": 0.909,
      "grad_norm": 18.75,
      "grad_norm_var": 0.3666015625,
      "learning_rate": 2.1012318447166743e-05,
      "loss": 6.6076,
      "loss/crossentropy": 2.043487349152565,
      "loss/hidden": 3.016796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1601484829559922,
      "step": 9090
    },
    {
      "epoch": 0.91,
      "grad_norm": 19.125,
      "grad_norm_var": 0.36868489583333336,
      "learning_rate": 2.0781731547998614e-05,
      "loss": 6.6279,
      "loss/crossentropy": 2.207579642534256,
      "loss/hidden": 2.980078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15659495070576668,
      "step": 9100
    },
    {
      "epoch": 0.911,
      "grad_norm": 19.0,
      "grad_norm_var": 0.8202473958333333,
      "learning_rate": 2.055325539281104e-05,
      "loss": 6.7914,
      "loss/crossentropy": 2.274830712378025,
      "loss/hidden": 2.973828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1602309113368392,
      "step": 9110
    },
    {
      "epoch": 0.912,
      "grad_norm": 19.5,
      "grad_norm_var": 1.11640625,
      "learning_rate": 2.0326904075089492e-05,
      "loss": 6.7296,
      "loss/crossentropy": 2.2438980847597123,
      "loss/hidden": 3.09921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17445293702185155,
      "step": 9120
    },
    {
      "epoch": 0.913,
      "grad_norm": 19.875,
      "grad_norm_var": 0.7061848958333333,
      "learning_rate": 2.0102691557249456e-05,
      "loss": 6.6545,
      "loss/crossentropy": 2.1133376926183702,
      "loss/hidden": 2.894140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.14513556826859714,
      "step": 9130
    },
    {
      "epoch": 0.914,
      "grad_norm": 17.75,
      "grad_norm_var": 1.8296223958333333,
      "learning_rate": 1.9880631669775164e-05,
      "loss": 6.6221,
      "loss/crossentropy": 2.132697519659996,
      "loss/hidden": 2.955859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16242318861186505,
      "step": 9140
    },
    {
      "epoch": 0.915,
      "grad_norm": 17.625,
      "grad_norm_var": 3.874739583333333,
      "learning_rate": 1.966073811036649e-05,
      "loss": 6.6691,
      "loss/crossentropy": 2.164944548904896,
      "loss/hidden": 2.848046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14622690286487341,
      "step": 9150
    },
    {
      "epoch": 0.916,
      "grad_norm": 22.125,
      "grad_norm_var": 2.6801432291666667,
      "learning_rate": 1.944302444309393e-05,
      "loss": 6.6237,
      "loss/crossentropy": 2.149251754581928,
      "loss/hidden": 3.013671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16157012842595578,
      "step": 9160
    },
    {
      "epoch": 0.917,
      "grad_norm": 18.875,
      "grad_norm_var": 1.50625,
      "learning_rate": 1.922750409756205e-05,
      "loss": 6.6684,
      "loss/crossentropy": 2.1195749223232268,
      "loss/hidden": 2.998828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15725019499659537,
      "step": 9170
    },
    {
      "epoch": 0.918,
      "grad_norm": 19.125,
      "grad_norm_var": 5.559375,
      "learning_rate": 1.9014190368080926e-05,
      "loss": 6.7637,
      "loss/crossentropy": 2.1270042449235915,
      "loss/hidden": 3.12578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.17823693621903658,
      "step": 9180
    },
    {
      "epoch": 0.919,
      "grad_norm": 19.25,
      "grad_norm_var": 5.692122395833334,
      "learning_rate": 1.88030964128462e-05,
      "loss": 6.7194,
      "loss/crossentropy": 2.163611389696598,
      "loss/hidden": 2.858203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.13989691147580743,
      "step": 9190
    },
    {
      "epoch": 0.92,
      "grad_norm": 18.875,
      "grad_norm_var": 0.8552083333333333,
      "learning_rate": 1.8594235253127375e-05,
      "loss": 6.6762,
      "loss/crossentropy": 2.1259265393018723,
      "loss/hidden": 2.963671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15320588331669568,
      "step": 9200
    },
    {
      "epoch": 0.921,
      "grad_norm": 19.75,
      "grad_norm_var": 0.9309895833333334,
      "learning_rate": 1.838761977246459e-05,
      "loss": 6.6246,
      "loss/crossentropy": 2.1278042390942575,
      "loss/hidden": 2.97109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1498261951841414,
      "step": 9210
    },
    {
      "epoch": 0.922,
      "grad_norm": 18.875,
      "grad_norm_var": 4.901822916666666,
      "learning_rate": 1.818326271587394e-05,
      "loss": 6.7278,
      "loss/crossentropy": 2.222877550125122,
      "loss/hidden": 2.95859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1604774661362171,
      "step": 9220
    },
    {
      "epoch": 0.923,
      "grad_norm": 19.5,
      "grad_norm_var": 5.441666666666666,
      "learning_rate": 1.7981176689061317e-05,
      "loss": 6.5921,
      "loss/crossentropy": 2.1431047171354294,
      "loss/hidden": 2.998046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14911785759031773,
      "step": 9230
    },
    {
      "epoch": 0.924,
      "grad_norm": 18.375,
      "grad_norm_var": 1.5634765625,
      "learning_rate": 1.7781374157644715e-05,
      "loss": 6.4714,
      "loss/crossentropy": 2.089540404081345,
      "loss/hidden": 2.984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15576251531019808,
      "step": 9240
    },
    {
      "epoch": 0.925,
      "grad_norm": 18.625,
      "grad_norm_var": 0.5483723958333333,
      "learning_rate": 1.758386744638546e-05,
      "loss": 6.7243,
      "loss/crossentropy": 2.2453290104866026,
      "loss/hidden": 2.94609375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14587981197983027,
      "step": 9250
    },
    {
      "epoch": 0.926,
      "grad_norm": 20.625,
      "grad_norm_var": 3.0152303024252385e+18,
      "learning_rate": 1.738866873842785e-05,
      "loss": 6.7019,
      "loss/crossentropy": 2.1938077017664908,
      "loss/hidden": 2.9220703125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15224464647471905,
      "step": 9260
    },
    {
      "epoch": 0.927,
      "grad_norm": 18.5,
      "grad_norm_var": 3.015230302642294e+18,
      "learning_rate": 1.7195790074547615e-05,
      "loss": 6.6027,
      "loss/crossentropy": 2.1094757497310637,
      "loss/hidden": 2.9421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1458785830065608,
      "step": 9270
    },
    {
      "epoch": 0.928,
      "grad_norm": 18.375,
      "grad_norm_var": 0.6343098958333333,
      "learning_rate": 1.7005243352409334e-05,
      "loss": 6.5892,
      "loss/crossentropy": 2.1868397384881972,
      "loss/hidden": 2.9390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15736450720578432,
      "step": 9280
    },
    {
      "epoch": 0.929,
      "grad_norm": 19.875,
      "grad_norm_var": 0.6218098958333333,
      "learning_rate": 1.681704032583234e-05,
      "loss": 6.6312,
      "loss/crossentropy": 2.102400007843971,
      "loss/hidden": 2.959765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.14601742010563612,
      "step": 9290
    },
    {
      "epoch": 0.93,
      "grad_norm": 18.0,
      "grad_norm_var": 0.90390625,
      "learning_rate": 1.6631192604065855e-05,
      "loss": 6.6723,
      "loss/crossentropy": 2.1556510917842386,
      "loss/hidden": 2.98984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15542004024609923,
      "step": 9300
    },
    {
      "epoch": 0.931,
      "grad_norm": 18.625,
      "grad_norm_var": 0.77265625,
      "learning_rate": 1.644771165107277e-05,
      "loss": 6.6506,
      "loss/crossentropy": 2.025302214920521,
      "loss/hidden": 2.92265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15237770597450434,
      "step": 9310
    },
    {
      "epoch": 0.932,
      "grad_norm": 19.5,
      "grad_norm_var": 0.5552083333333333,
      "learning_rate": 1.6266608784822544e-05,
      "loss": 6.6707,
      "loss/crossentropy": 2.2078075274825095,
      "loss/hidden": 2.873046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15092720007523894,
      "step": 9320
    },
    {
      "epoch": 0.933,
      "grad_norm": 18.75,
      "grad_norm_var": 6.516666666666667,
      "learning_rate": 1.6087895176593026e-05,
      "loss": 6.6366,
      "loss/crossentropy": 2.1986721485853193,
      "loss/hidden": 2.983984375,
      "loss/jsd": 0.0,
      "loss/logits": 0.16585835479199887,
      "step": 9330
    },
    {
      "epoch": 0.934,
      "grad_norm": 19.25,
      "grad_norm_var": 7.455143229166667,
      "learning_rate": 1.5911581850281403e-05,
      "loss": 6.6058,
      "loss/crossentropy": 2.2786722093820573,
      "loss/hidden": 2.87734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14315823893994092,
      "step": 9340
    },
    {
      "epoch": 0.935,
      "grad_norm": 19.0,
      "grad_norm_var": 0.9405598958333333,
      "learning_rate": 1.573767968172413e-05,
      "loss": 6.6968,
      "loss/crossentropy": 2.0444235846400263,
      "loss/hidden": 2.94140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15067112175747752,
      "step": 9350
    },
    {
      "epoch": 0.936,
      "grad_norm": 20.125,
      "grad_norm_var": 0.41848958333333336,
      "learning_rate": 1.556619939802615e-05,
      "loss": 6.6068,
      "loss/crossentropy": 2.088939316570759,
      "loss/hidden": 2.976171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14950306443497538,
      "step": 9360
    },
    {
      "epoch": 0.937,
      "grad_norm": 19.875,
      "grad_norm_var": 0.47890625,
      "learning_rate": 1.5397151576899065e-05,
      "loss": 6.6725,
      "loss/crossentropy": 2.221402445435524,
      "loss/hidden": 3.01015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1524247596040368,
      "step": 9370
    },
    {
      "epoch": 0.938,
      "grad_norm": 19.0,
      "grad_norm_var": 0.8010416666666667,
      "learning_rate": 1.5230546646008795e-05,
      "loss": 6.5985,
      "loss/crossentropy": 2.2344843059778214,
      "loss/hidden": 2.925,
      "loss/jsd": 0.0,
      "loss/logits": 0.15453678853809832,
      "step": 9380
    },
    {
      "epoch": 0.939,
      "grad_norm": 18.25,
      "grad_norm_var": 0.8322916666666667,
      "learning_rate": 1.506639488233226e-05,
      "loss": 6.6656,
      "loss/crossentropy": 2.1397018134593964,
      "loss/hidden": 2.937890625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15283725168555976,
      "step": 9390
    },
    {
      "epoch": 0.94,
      "grad_norm": 20.25,
      "grad_norm_var": 2.08515625,
      "learning_rate": 1.490470641152345e-05,
      "loss": 6.6495,
      "loss/crossentropy": 2.180284807085991,
      "loss/hidden": 3.04765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1779416125267744,
      "step": 9400
    },
    {
      "epoch": 0.941,
      "grad_norm": 18.375,
      "grad_norm_var": 1.56640625,
      "learning_rate": 1.4745491207288874e-05,
      "loss": 6.7046,
      "loss/crossentropy": 1.9674726322293281,
      "loss/hidden": 2.985546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14737507775425912,
      "step": 9410
    },
    {
      "epoch": 0.942,
      "grad_norm": 19.125,
      "grad_norm_var": 0.9546223958333333,
      "learning_rate": 1.4588759090772302e-05,
      "loss": 6.645,
      "loss/crossentropy": 1.932501320540905,
      "loss/hidden": 2.90859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.13745289896614848,
      "step": 9420
    },
    {
      "epoch": 0.943,
      "grad_norm": 19.0,
      "grad_norm_var": 0.4114583333333333,
      "learning_rate": 1.4434519729948942e-05,
      "loss": 6.7417,
      "loss/crossentropy": 2.132400029152632,
      "loss/hidden": 2.90859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1424996647052467,
      "step": 9430
    },
    {
      "epoch": 0.944,
      "grad_norm": 18.625,
      "grad_norm_var": 0.9184895833333333,
      "learning_rate": 1.428278263902913e-05,
      "loss": 6.6339,
      "loss/crossentropy": 2.091708117723465,
      "loss/hidden": 2.963671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14837299454957248,
      "step": 9440
    },
    {
      "epoch": 0.945,
      "grad_norm": 18.25,
      "grad_norm_var": 0.4910807291666667,
      "learning_rate": 1.413355717787134e-05,
      "loss": 6.7095,
      "loss/crossentropy": 2.028689184784889,
      "loss/hidden": 2.925390625,
      "loss/jsd": 0.0,
      "loss/logits": 0.13867047037929298,
      "step": 9450
    },
    {
      "epoch": 0.946,
      "grad_norm": 19.125,
      "grad_norm_var": 0.3458333333333333,
      "learning_rate": 1.3986852551404964e-05,
      "loss": 6.7444,
      "loss/crossentropy": 2.2080354422330855,
      "loss/hidden": 2.956640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1599160810932517,
      "step": 9460
    },
    {
      "epoch": 0.947,
      "grad_norm": 19.375,
      "grad_norm_var": 0.2559895833333333,
      "learning_rate": 1.3842677809062387e-05,
      "loss": 6.6449,
      "loss/crossentropy": 2.1781085431575775,
      "loss/hidden": 2.95625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16033231243491172,
      "step": 9470
    },
    {
      "epoch": 0.948,
      "grad_norm": 18.375,
      "grad_norm_var": 0.34765625,
      "learning_rate": 1.3701041844220849e-05,
      "loss": 6.6578,
      "loss/crossentropy": 2.149868738651276,
      "loss/hidden": 2.9625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1523012975230813,
      "step": 9480
    },
    {
      "epoch": 0.949,
      "grad_norm": 19.875,
      "grad_norm_var": 0.665625,
      "learning_rate": 1.3561953393653824e-05,
      "loss": 6.5907,
      "loss/crossentropy": 2.013017991185188,
      "loss/hidden": 3.104296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16240219958126545,
      "step": 9490
    },
    {
      "epoch": 0.95,
      "grad_norm": 18.375,
      "grad_norm_var": 0.7427083333333333,
      "learning_rate": 1.3425421036992098e-05,
      "loss": 6.5356,
      "loss/crossentropy": 2.1714504957199097,
      "loss/hidden": 2.993359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14401763658970596,
      "step": 9500
    },
    {
      "epoch": 0.951,
      "grad_norm": 18.75,
      "grad_norm_var": 0.8369140625,
      "learning_rate": 1.3291453196194565e-05,
      "loss": 6.6629,
      "loss/crossentropy": 2.0724572435021402,
      "loss/hidden": 3.02578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1599965788424015,
      "step": 9510
    },
    {
      "epoch": 0.952,
      "grad_norm": 18.875,
      "grad_norm_var": 1.1143229166666666,
      "learning_rate": 1.3160058135028691e-05,
      "loss": 6.6944,
      "loss/crossentropy": 2.1555590838193894,
      "loss/hidden": 2.9203125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1390397796407342,
      "step": 9520
    },
    {
      "epoch": 0.953,
      "grad_norm": 19.25,
      "grad_norm_var": 1.3806640625,
      "learning_rate": 1.3031243958560772e-05,
      "loss": 6.5407,
      "loss/crossentropy": 2.054419046640396,
      "loss/hidden": 2.96171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15348356384783984,
      "step": 9530
    },
    {
      "epoch": 0.954,
      "grad_norm": 21.125,
      "grad_norm_var": 0.6129557291666666,
      "learning_rate": 1.2905018612655975e-05,
      "loss": 6.716,
      "loss/crossentropy": 2.2037750400602816,
      "loss/hidden": 2.976171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15483110053464771,
      "step": 9540
    },
    {
      "epoch": 0.955,
      "grad_norm": 18.75,
      "grad_norm_var": 16.742122395833334,
      "learning_rate": 1.2781389883488218e-05,
      "loss": 6.6118,
      "loss/crossentropy": 2.2344287753105165,
      "loss/hidden": 2.926953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1532795075327158,
      "step": 9550
    },
    {
      "epoch": 0.956,
      "grad_norm": 18.5,
      "grad_norm_var": 16.912239583333335,
      "learning_rate": 1.2660365397059856e-05,
      "loss": 6.5768,
      "loss/crossentropy": 2.1113690607249738,
      "loss/hidden": 2.871484375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15772182103246452,
      "step": 9560
    },
    {
      "epoch": 0.957,
      "grad_norm": 18.5,
      "grad_norm_var": 0.4327473958333333,
      "learning_rate": 1.2541952618731295e-05,
      "loss": 6.6605,
      "loss/crossentropy": 2.11531500518322,
      "loss/hidden": 3.0078125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16428461223840712,
      "step": 9570
    },
    {
      "epoch": 0.958,
      "grad_norm": 19.125,
      "grad_norm_var": 1.0853515625,
      "learning_rate": 1.2426158852760462e-05,
      "loss": 6.632,
      "loss/crossentropy": 2.180248150229454,
      "loss/hidden": 3.052734375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1568142266944051,
      "step": 9580
    },
    {
      "epoch": 0.959,
      "grad_norm": 19.625,
      "grad_norm_var": 0.33932291666666664,
      "learning_rate": 1.2312991241852293e-05,
      "loss": 6.7198,
      "loss/crossentropy": 2.157143747806549,
      "loss/hidden": 2.940234375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15367843862622976,
      "step": 9590
    },
    {
      "epoch": 0.96,
      "grad_norm": 18.25,
      "grad_norm_var": 0.4488932291666667,
      "learning_rate": 1.2202456766718093e-05,
      "loss": 6.7218,
      "loss/crossentropy": 2.226324610412121,
      "loss/hidden": 2.95546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1611587280407548,
      "step": 9600
    },
    {
      "epoch": 0.961,
      "grad_norm": 19.125,
      "grad_norm_var": 0.39889322916666664,
      "learning_rate": 1.2094562245644947e-05,
      "loss": 6.5949,
      "loss/crossentropy": 2.206631433963776,
      "loss/hidden": 2.99921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16857987977564334,
      "step": 9610
    },
    {
      "epoch": 0.962,
      "grad_norm": 18.75,
      "grad_norm_var": 0.7893229166666667,
      "learning_rate": 1.1989314334075145e-05,
      "loss": 6.6565,
      "loss/crossentropy": 2.078006048500538,
      "loss/hidden": 2.885546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.13547398131340743,
      "step": 9620
    },
    {
      "epoch": 0.963,
      "grad_norm": 18.375,
      "grad_norm_var": 1.2364583333333334,
      "learning_rate": 1.1886719524195641e-05,
      "loss": 6.6847,
      "loss/crossentropy": 2.1102965086698533,
      "loss/hidden": 3.138671875,
      "loss/jsd": 0.0,
      "loss/logits": 0.17691021095961332,
      "step": 9630
    },
    {
      "epoch": 0.964,
      "grad_norm": 18.75,
      "grad_norm_var": 1.4379557291666667,
      "learning_rate": 1.1786784144537563e-05,
      "loss": 6.6631,
      "loss/crossentropy": 2.0664229825139047,
      "loss/hidden": 2.99921875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1560237266123295,
      "step": 9640
    },
    {
      "epoch": 0.965,
      "grad_norm": 19.125,
      "grad_norm_var": 0.30416666666666664,
      "learning_rate": 1.168951435958588e-05,
      "loss": 6.5308,
      "loss/crossentropy": 2.1858409374952315,
      "loss/hidden": 2.98828125,
      "loss/jsd": 0.0,
      "loss/logits": 0.15965924076735974,
      "step": 9650
    },
    {
      "epoch": 0.966,
      "grad_norm": 22.375,
      "grad_norm_var": 1.6184895833333333,
      "learning_rate": 1.1594916169399088e-05,
      "loss": 6.6978,
      "loss/crossentropy": 2.357753816246986,
      "loss/hidden": 2.981640625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16278257928788661,
      "step": 9660
    },
    {
      "epoch": 0.967,
      "grad_norm": 18.25,
      "grad_norm_var": 1.6067057291666667,
      "learning_rate": 1.1502995409239174e-05,
      "loss": 6.6365,
      "loss/crossentropy": 2.1270371586084367,
      "loss/hidden": 2.872265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1369372447952628,
      "step": 9670
    },
    {
      "epoch": 0.968,
      "grad_norm": 18.125,
      "grad_norm_var": 1.0181640625,
      "learning_rate": 1.1413757749211602e-05,
      "loss": 6.5318,
      "loss/crossentropy": 2.0077289439737798,
      "loss/hidden": 2.90859375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1443242894485593,
      "step": 9680
    },
    {
      "epoch": 0.969,
      "grad_norm": 19.125,
      "grad_norm_var": 1.6197916666666667,
      "learning_rate": 1.132720869391559e-05,
      "loss": 6.6815,
      "loss/crossentropy": 2.147722618281841,
      "loss/hidden": 2.896875,
      "loss/jsd": 0.0,
      "loss/logits": 0.1429056410677731,
      "step": 9690
    },
    {
      "epoch": 0.97,
      "grad_norm": 19.25,
      "grad_norm_var": 1.3723307291666667,
      "learning_rate": 1.1243353582104556e-05,
      "loss": 6.7836,
      "loss/crossentropy": 2.157535497099161,
      "loss/hidden": 3.10546875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16926438817754388,
      "step": 9700
    },
    {
      "epoch": 0.971,
      "grad_norm": 19.625,
      "grad_norm_var": 1.1962890625,
      "learning_rate": 1.116219758635678e-05,
      "loss": 6.5981,
      "loss/crossentropy": 2.0965492375195027,
      "loss/hidden": 2.94140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15332625713199377,
      "step": 9710
    },
    {
      "epoch": 0.972,
      "grad_norm": 18.625,
      "grad_norm_var": 1.2228515625,
      "learning_rate": 1.1083745712756367e-05,
      "loss": 6.741,
      "loss/crossentropy": 2.0914740189909935,
      "loss/hidden": 3.059375,
      "loss/jsd": 0.0,
      "loss/logits": 0.15363588072359563,
      "step": 9720
    },
    {
      "epoch": 0.973,
      "grad_norm": 21.625,
      "grad_norm_var": 1.3393229166666667,
      "learning_rate": 1.1008002800584424e-05,
      "loss": 6.701,
      "loss/crossentropy": 2.141095507889986,
      "loss/hidden": 2.982421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15398450717329978,
      "step": 9730
    },
    {
      "epoch": 0.974,
      "grad_norm": 19.25,
      "grad_norm_var": 1.6205729166666667,
      "learning_rate": 1.0934973522020538e-05,
      "loss": 6.6305,
      "loss/crossentropy": 2.1611881658434866,
      "loss/hidden": 2.96328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.14838593397289515,
      "step": 9740
    },
    {
      "epoch": 0.975,
      "grad_norm": 19.375,
      "grad_norm_var": 2.2900390625,
      "learning_rate": 1.0864662381854632e-05,
      "loss": 6.6726,
      "loss/crossentropy": 2.0327648639678957,
      "loss/hidden": 2.972265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.16227712836116553,
      "step": 9750
    },
    {
      "epoch": 0.976,
      "grad_norm": 20.5,
      "grad_norm_var": 10.545572916666666,
      "learning_rate": 1.0797073717209014e-05,
      "loss": 6.6785,
      "loss/crossentropy": 2.14007937759161,
      "loss/hidden": 2.93515625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15520989168435334,
      "step": 9760
    },
    {
      "epoch": 0.977,
      "grad_norm": 19.5,
      "grad_norm_var": 0.8822265625,
      "learning_rate": 1.0732211697270884e-05,
      "loss": 6.6563,
      "loss/crossentropy": 2.121374186873436,
      "loss/hidden": 2.992578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16025999132543803,
      "step": 9770
    },
    {
      "epoch": 0.978,
      "grad_norm": 18.75,
      "grad_norm_var": 0.56640625,
      "learning_rate": 1.0670080323035176e-05,
      "loss": 6.6324,
      "loss/crossentropy": 2.1174088306725025,
      "loss/hidden": 2.96328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16041831383481622,
      "step": 9780
    },
    {
      "epoch": 0.979,
      "grad_norm": 18.75,
      "grad_norm_var": 0.9497395833333333,
      "learning_rate": 1.0610683427057707e-05,
      "loss": 6.724,
      "loss/crossentropy": 2.211408218741417,
      "loss/hidden": 2.91796875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14411625489592553,
      "step": 9790
    },
    {
      "epoch": 0.98,
      "grad_norm": 19.25,
      "grad_norm_var": 1.04765625,
      "learning_rate": 1.0554024673218807e-05,
      "loss": 6.7129,
      "loss/crossentropy": 2.1691969078034163,
      "loss/hidden": 2.94765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.14351302795112134,
      "step": 9800
    },
    {
      "epoch": 0.981,
      "grad_norm": 20.875,
      "grad_norm_var": 2.7122395833333335,
      "learning_rate": 1.0500107556497298e-05,
      "loss": 6.6663,
      "loss/crossentropy": 2.1762838467955588,
      "loss/hidden": 2.947265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1569036815315485,
      "step": 9810
    },
    {
      "epoch": 0.982,
      "grad_norm": 17.75,
      "grad_norm_var": 2.512955729166667,
      "learning_rate": 1.0448935402754912e-05,
      "loss": 6.6726,
      "loss/crossentropy": 2.022391739487648,
      "loss/hidden": 3.057421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16687135621905327,
      "step": 9820
    },
    {
      "epoch": 0.983,
      "grad_norm": 20.0,
      "grad_norm_var": 0.9244140625,
      "learning_rate": 1.0400511368531138e-05,
      "loss": 6.6903,
      "loss/crossentropy": 2.0364112444221973,
      "loss/hidden": 2.961328125,
      "loss/jsd": 0.0,
      "loss/logits": 0.16487439339980484,
      "step": 9830
    },
    {
      "epoch": 0.984,
      "grad_norm": 19.75,
      "grad_norm_var": 1.1723307291666667,
      "learning_rate": 1.0354838440848503e-05,
      "loss": 6.6904,
      "loss/crossentropy": 2.009164361655712,
      "loss/hidden": 2.976171875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15027653202414512,
      "step": 9840
    },
    {
      "epoch": 0.985,
      "grad_norm": 19.625,
      "grad_norm_var": 1.6759765625,
      "learning_rate": 1.0311919437028318e-05,
      "loss": 6.64,
      "loss/crossentropy": 2.213483475893736,
      "loss/hidden": 2.978125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1542889505624771,
      "step": 9850
    },
    {
      "epoch": 0.986,
      "grad_norm": 18.125,
      "grad_norm_var": 0.7705729166666667,
      "learning_rate": 1.0271757004516918e-05,
      "loss": 6.6129,
      "loss/crossentropy": 2.0778297156095507,
      "loss/hidden": 3.02421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.16476696096360682,
      "step": 9860
    },
    {
      "epoch": 0.987,
      "grad_norm": 18.75,
      "grad_norm_var": 8.672916666666667,
      "learning_rate": 1.023435362072231e-05,
      "loss": 6.6758,
      "loss/crossentropy": 2.1646489948034286,
      "loss/hidden": 2.969140625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15129716228693724,
      "step": 9870
    },
    {
      "epoch": 0.988,
      "grad_norm": 20.75,
      "grad_norm_var": 0.7624348958333333,
      "learning_rate": 1.01997115928614e-05,
      "loss": 6.728,
      "loss/crossentropy": 2.2021071404218673,
      "loss/hidden": 2.879296875,
      "loss/jsd": 0.0,
      "loss/logits": 0.14578848090022803,
      "step": 9880
    },
    {
      "epoch": 0.989,
      "grad_norm": 17.625,
      "grad_norm_var": 8.957291666666666,
      "learning_rate": 1.0167833057817644e-05,
      "loss": 6.6842,
      "loss/crossentropy": 2.0844334810972214,
      "loss/hidden": 3.037109375,
      "loss/jsd": 0.0,
      "loss/logits": 0.17800515107810497,
      "step": 9890
    },
    {
      "epoch": 0.99,
      "grad_norm": 18.75,
      "grad_norm_var": 9.4447265625,
      "learning_rate": 1.0138719982009242e-05,
      "loss": 6.5745,
      "loss/crossentropy": 2.064706768095493,
      "loss/hidden": 3.04765625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15266596768051385,
      "step": 9900
    },
    {
      "epoch": 0.991,
      "grad_norm": 19.875,
      "grad_norm_var": 1.0916666666666666,
      "learning_rate": 1.0112374161267848e-05,
      "loss": 6.7953,
      "loss/crossentropy": 2.100151504948735,
      "loss/hidden": 3.007421875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15070506385527552,
      "step": 9910
    },
    {
      "epoch": 0.992,
      "grad_norm": 18.875,
      "grad_norm_var": 0.6893229166666667,
      "learning_rate": 1.008879722072778e-05,
      "loss": 6.7148,
      "loss/crossentropy": 2.1347645312547683,
      "loss/hidden": 2.993359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.1627992235124111,
      "step": 9920
    },
    {
      "epoch": 0.993,
      "grad_norm": 19.375,
      "grad_norm_var": 0.7395833333333334,
      "learning_rate": 1.0067990614725795e-05,
      "loss": 6.6596,
      "loss/crossentropy": 2.020579759031534,
      "loss/hidden": 2.92578125,
      "loss/jsd": 0.0,
      "loss/logits": 0.14832647573202848,
      "step": 9930
    },
    {
      "epoch": 0.994,
      "grad_norm": 20.5,
      "grad_norm_var": 0.6395833333333333,
      "learning_rate": 1.0049955626711355e-05,
      "loss": 6.6734,
      "loss/crossentropy": 2.1040701150894163,
      "loss/hidden": 3.025,
      "loss/jsd": 0.0,
      "loss/logits": 0.15983649622648954,
      "step": 9940
    },
    {
      "epoch": 0.995,
      "grad_norm": 18.75,
      "grad_norm_var": 0.5171223958333333,
      "learning_rate": 1.003469336916747e-05,
      "loss": 6.764,
      "loss/crossentropy": 2.0992766320705414,
      "loss/hidden": 2.973046875,
      "loss/jsd": 0.0,
      "loss/logits": 0.15146102383732796,
      "step": 9950
    },
    {
      "epoch": 0.996,
      "grad_norm": 19.0,
      "grad_norm_var": 0.6434895833333333,
      "learning_rate": 1.002220478354208e-05,
      "loss": 6.7187,
      "loss/crossentropy": 2.085821607708931,
      "loss/hidden": 3.072265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.15508200749754905,
      "step": 9960
    },
    {
      "epoch": 0.997,
      "grad_norm": 20.625,
      "grad_norm_var": 1.0739583333333333,
      "learning_rate": 1.0012490640189975e-05,
      "loss": 6.6748,
      "loss/crossentropy": 2.0770878665149213,
      "loss/hidden": 2.997265625,
      "loss/jsd": 0.0,
      "loss/logits": 0.14898568596690892,
      "step": 9970
    },
    {
      "epoch": 0.998,
      "grad_norm": 21.5,
      "grad_norm_var": 1.76640625,
      "learning_rate": 1.0005551538325275e-05,
      "loss": 6.6931,
      "loss/crossentropy": 2.3164404213428496,
      "loss/hidden": 2.86015625,
      "loss/jsd": 0.0,
      "loss/logits": 0.1461044853553176,
      "step": 9980
    },
    {
      "epoch": 0.999,
      "grad_norm": 21.5,
      "grad_norm_var": 1.7613932291666667,
      "learning_rate": 1.0001387905984467e-05,
      "loss": 6.7299,
      "loss/crossentropy": 2.1300622135400773,
      "loss/hidden": 2.901953125,
      "loss/jsd": 0.0,
      "loss/logits": 0.1424020084552467,
      "step": 9990
    },
    {
      "epoch": 1.0,
      "grad_norm": 18.875,
      "grad_norm_var": 0.70625,
      "learning_rate": 1e-05,
      "loss": 6.6101,
      "loss/crossentropy": 2.187693312764168,
      "loss/hidden": 2.918359375,
      "loss/jsd": 0.0,
      "loss/logits": 0.14794533271342517,
      "step": 10000
    }
  ],
  "logging_steps": 10,
  "max_steps": 10000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 9223372036854775807,
  "save_steps": 2000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 2.8575100320088064e+19,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}