{
  "best_metric": 0.18269842863082886,
  "best_model_checkpoint": "./ap_train_outputs_new/checkpoint-7725",
  "epoch": 150.0,
  "eval_steps": 500,
  "global_step": 386250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.999948220064725e-05,
      "loss": 2.4306,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.99989644012945e-05,
      "loss": 2.3106,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.999844660194175e-05,
      "loss": 2.1619,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9997928802589e-05,
      "loss": 2.042,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9997411003236248e-05,
      "loss": 2.0323,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9996893203883496e-05,
      "loss": 1.9385,
      "step": 60
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9996375404530747e-05,
      "loss": 1.8729,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9995857605177995e-05,
      "loss": 1.8625,
      "step": 80
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9995339805825243e-05,
      "loss": 1.6289,
      "step": 90
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994822006472494e-05,
      "loss": 1.664,
      "step": 100
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9994304207119742e-05,
      "loss": 1.7535,
      "step": 110
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.9993786407766994e-05,
      "loss": 1.5424,
      "step": 120
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.999326860841424e-05,
      "loss": 1.5944,
      "step": 130
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.999275080906149e-05,
      "loss": 1.4155,
      "step": 140
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999223300970874e-05,
      "loss": 1.3795,
      "step": 150
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.999171521035599e-05,
      "loss": 1.3255,
      "step": 160
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9991197411003237e-05,
      "loss": 1.4067,
      "step": 170
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9990679611650488e-05,
      "loss": 1.423,
      "step": 180
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9990161812297736e-05,
      "loss": 1.2975,
      "step": 190
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9989644012944987e-05,
      "loss": 1.2546,
      "step": 200
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9989126213592235e-05,
      "loss": 1.1929,
      "step": 210
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9988608414239483e-05,
      "loss": 1.2079,
      "step": 220
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.998809061488673e-05,
      "loss": 1.2546,
      "step": 230
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9987572815533982e-05,
      "loss": 1.1213,
      "step": 240
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.998705501618123e-05,
      "loss": 1.0472,
      "step": 250
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.998653721682848e-05,
      "loss": 1.0242,
      "step": 260
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.998601941747573e-05,
      "loss": 1.202,
      "step": 270
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.998550161812298e-05,
      "loss": 1.0183,
      "step": 280
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.998498381877023e-05,
      "loss": 0.9347,
      "step": 290
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9984466019417477e-05,
      "loss": 0.9239,
      "step": 300
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9983948220064725e-05,
      "loss": 1.037,
      "step": 310
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9983430420711976e-05,
      "loss": 1.1471,
      "step": 320
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9982912621359224e-05,
      "loss": 0.9171,
      "step": 330
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9982394822006475e-05,
      "loss": 0.7951,
      "step": 340
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9981877022653723e-05,
      "loss": 0.8798,
      "step": 350
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9981359223300974e-05,
      "loss": 0.9753,
      "step": 360
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9980841423948222e-05,
      "loss": 0.8387,
      "step": 370
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.998032362459547e-05,
      "loss": 0.9324,
      "step": 380
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9979805825242718e-05,
      "loss": 0.8375,
      "step": 390
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997928802588997e-05,
      "loss": 0.7599,
      "step": 400
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9978770226537217e-05,
      "loss": 0.8142,
      "step": 410
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.997825242718447e-05,
      "loss": 0.7867,
      "step": 420
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9977734627831717e-05,
      "loss": 0.7268,
      "step": 430
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9977216828478968e-05,
      "loss": 0.7552,
      "step": 440
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9976699029126216e-05,
      "loss": 0.8379,
      "step": 450
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9976181229773464e-05,
      "loss": 0.7393,
      "step": 460
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.997566343042071e-05,
      "loss": 0.7218,
      "step": 470
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9975145631067963e-05,
      "loss": 0.7362,
      "step": 480
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.997462783171521e-05,
      "loss": 0.7993,
      "step": 490
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9974110032362462e-05,
      "loss": 0.7622,
      "step": 500
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.997359223300971e-05,
      "loss": 0.6441,
      "step": 510
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.997307443365696e-05,
      "loss": 0.7303,
      "step": 520
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9972556634304206e-05,
      "loss": 0.6459,
      "step": 530
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9972038834951457e-05,
      "loss": 0.5734,
      "step": 540
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9971521035598705e-05,
      "loss": 0.6804,
      "step": 550
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9971003236245957e-05,
      "loss": 0.6168,
      "step": 560
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9970485436893204e-05,
      "loss": 0.6384,
      "step": 570
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9969967637540456e-05,
      "loss": 0.6387,
      "step": 580
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9969449838187704e-05,
      "loss": 0.7027,
      "step": 590
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9968932038834955e-05,
      "loss": 0.6637,
      "step": 600
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9968414239482203e-05,
      "loss": 0.5355,
      "step": 610
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.996789644012945e-05,
      "loss": 0.5599,
      "step": 620
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.99673786407767e-05,
      "loss": 0.551,
      "step": 630
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.996686084142395e-05,
      "loss": 0.5663,
      "step": 640
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9966343042071198e-05,
      "loss": 0.6923,
      "step": 650
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.996582524271845e-05,
      "loss": 0.6918,
      "step": 660
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9965307443365697e-05,
      "loss": 0.576,
      "step": 670
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.996478964401295e-05,
      "loss": 0.5371,
      "step": 680
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9964271844660196e-05,
      "loss": 0.5559,
      "step": 690
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9963754045307444e-05,
      "loss": 0.4964,
      "step": 700
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9963236245954692e-05,
      "loss": 0.6167,
      "step": 710
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9962718446601944e-05,
      "loss": 0.6102,
      "step": 720
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.996220064724919e-05,
      "loss": 0.4764,
      "step": 730
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9961682847896443e-05,
      "loss": 0.5488,
      "step": 740
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.996116504854369e-05,
      "loss": 0.544,
      "step": 750
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9960647249190942e-05,
      "loss": 0.3878,
      "step": 760
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.996012944983819e-05,
      "loss": 0.4912,
      "step": 770
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9959611650485438e-05,
      "loss": 0.5108,
      "step": 780
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9959093851132686e-05,
      "loss": 0.3333,
      "step": 790
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9958576051779937e-05,
      "loss": 0.5888,
      "step": 800
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9958058252427185e-05,
      "loss": 0.5131,
      "step": 810
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9957540453074436e-05,
      "loss": 0.4222,
      "step": 820
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9957022653721684e-05,
      "loss": 0.5501,
      "step": 830
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9956504854368936e-05,
      "loss": 0.5809,
      "step": 840
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9955987055016184e-05,
      "loss": 0.4968,
      "step": 850
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.995546925566343e-05,
      "loss": 0.4028,
      "step": 860
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.995495145631068e-05,
      "loss": 0.5314,
      "step": 870
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.995443365695793e-05,
      "loss": 0.5976,
      "step": 880
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.995391585760518e-05,
      "loss": 0.4516,
      "step": 890
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.995339805825243e-05,
      "loss": 0.4242,
      "step": 900
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9952880258899678e-05,
      "loss": 0.5095,
      "step": 910
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.995236245954693e-05,
      "loss": 0.4831,
      "step": 920
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9951844660194177e-05,
      "loss": 0.4483,
      "step": 930
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9951326860841425e-05,
      "loss": 0.5792,
      "step": 940
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9950809061488673e-05,
      "loss": 0.4076,
      "step": 950
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9950291262135924e-05,
      "loss": 0.4534,
      "step": 960
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9949773462783172e-05,
      "loss": 0.4433,
      "step": 970
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9949255663430424e-05,
      "loss": 0.4012,
      "step": 980
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.994873786407767e-05,
      "loss": 0.4392,
      "step": 990
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9948220064724923e-05,
      "loss": 0.346,
      "step": 1000
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.994770226537217e-05,
      "loss": 0.3584,
      "step": 1010
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.994718446601942e-05,
      "loss": 0.466,
      "step": 1020
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9946666666666667e-05,
      "loss": 0.4881,
      "step": 1030
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9946148867313918e-05,
      "loss": 0.4798,
      "step": 1040
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9945631067961166e-05,
      "loss": 0.4721,
      "step": 1050
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9945113268608417e-05,
      "loss": 0.6007,
      "step": 1060
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9944595469255665e-05,
      "loss": 0.3789,
      "step": 1070
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9944077669902916e-05,
      "loss": 0.3315,
      "step": 1080
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9943559870550164e-05,
      "loss": 0.4815,
      "step": 1090
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9943042071197412e-05,
      "loss": 0.3936,
      "step": 1100
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.994252427184466e-05,
      "loss": 0.5256,
      "step": 1110
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.994200647249191e-05,
      "loss": 0.366,
      "step": 1120
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.994148867313916e-05,
      "loss": 0.3964,
      "step": 1130
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.994097087378641e-05,
      "loss": 0.5443,
      "step": 1140
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.994045307443366e-05,
      "loss": 0.3093,
      "step": 1150
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.993993527508091e-05,
      "loss": 0.3967,
      "step": 1160
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9939417475728158e-05,
      "loss": 0.4326,
      "step": 1170
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9938899676375406e-05,
      "loss": 0.2558,
      "step": 1180
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9938381877022654e-05,
      "loss": 0.5055,
      "step": 1190
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9937864077669905e-05,
      "loss": 0.3684,
      "step": 1200
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9937346278317153e-05,
      "loss": 0.2573,
      "step": 1210
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9936828478964404e-05,
      "loss": 0.4943,
      "step": 1220
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9936310679611652e-05,
      "loss": 0.3349,
      "step": 1230
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9935792880258903e-05,
      "loss": 0.4965,
      "step": 1240
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.993527508090615e-05,
      "loss": 0.4059,
      "step": 1250
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.99347572815534e-05,
      "loss": 0.3418,
      "step": 1260
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9934239482200647e-05,
      "loss": 0.3945,
      "step": 1270
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.99337216828479e-05,
      "loss": 0.3294,
      "step": 1280
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9933203883495146e-05,
      "loss": 0.4028,
      "step": 1290
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9932686084142398e-05,
      "loss": 0.3644,
      "step": 1300
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9932168284789646e-05,
      "loss": 0.461,
      "step": 1310
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.9931650485436894e-05,
      "loss": 0.3644,
      "step": 1320
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9931132686084145e-05,
      "loss": 0.426,
      "step": 1330
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.9930614886731393e-05,
      "loss": 0.3641,
      "step": 1340
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.993009708737864e-05,
      "loss": 0.3879,
      "step": 1350
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.9929579288025892e-05,
      "loss": 0.3104,
      "step": 1360
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.992906148867314e-05,
      "loss": 0.415,
      "step": 1370
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.992854368932039e-05,
      "loss": 0.3901,
      "step": 1380
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.992802588996764e-05,
      "loss": 0.3854,
      "step": 1390
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9927508090614887e-05,
      "loss": 0.2604,
      "step": 1400
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.992699029126214e-05,
      "loss": 0.3043,
      "step": 1410
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9926472491909386e-05,
      "loss": 0.3011,
      "step": 1420
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9925954692556634e-05,
      "loss": 0.356,
      "step": 1430
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9925436893203886e-05,
      "loss": 0.3095,
      "step": 1440
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.9924919093851134e-05,
      "loss": 0.5059,
      "step": 1450
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9924401294498385e-05,
      "loss": 0.3261,
      "step": 1460
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9923883495145633e-05,
      "loss": 0.434,
      "step": 1470
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.992336569579288e-05,
      "loss": 0.4395,
      "step": 1480
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.9922847896440132e-05,
      "loss": 0.2863,
      "step": 1490
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.992233009708738e-05,
      "loss": 0.4658,
      "step": 1500
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9921812297734628e-05,
      "loss": 0.3183,
      "step": 1510
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.992129449838188e-05,
      "loss": 0.2679,
      "step": 1520
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9920776699029127e-05,
      "loss": 0.2732,
      "step": 1530
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.992025889967638e-05,
      "loss": 0.2877,
      "step": 1540
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.9919741100323626e-05,
      "loss": 0.3261,
      "step": 1550
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9919223300970874e-05,
      "loss": 0.3296,
      "step": 1560
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9918705501618126e-05,
      "loss": 0.3115,
      "step": 1570
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9918187702265374e-05,
      "loss": 0.3859,
      "step": 1580
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.991766990291262e-05,
      "loss": 0.3414,
      "step": 1590
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9917152103559873e-05,
      "loss": 0.3421,
      "step": 1600
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.991663430420712e-05,
      "loss": 0.3558,
      "step": 1610
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9916116504854372e-05,
      "loss": 0.3691,
      "step": 1620
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.991559870550162e-05,
      "loss": 0.2435,
      "step": 1630
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9915080906148868e-05,
      "loss": 0.283,
      "step": 1640
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.991456310679612e-05,
      "loss": 0.3628,
      "step": 1650
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.9914045307443367e-05,
      "loss": 0.3772,
      "step": 1660
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9913527508090615e-05,
      "loss": 0.3549,
      "step": 1670
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9913009708737866e-05,
      "loss": 0.2832,
      "step": 1680
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9912491909385114e-05,
      "loss": 0.3916,
      "step": 1690
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9911974110032362e-05,
      "loss": 0.4128,
      "step": 1700
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9911456310679613e-05,
      "loss": 0.3781,
      "step": 1710
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.991093851132686e-05,
      "loss": 0.4316,
      "step": 1720
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.9910420711974113e-05,
      "loss": 0.4245,
      "step": 1730
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.990990291262136e-05,
      "loss": 0.2219,
      "step": 1740
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9909385113268612e-05,
      "loss": 0.3784,
      "step": 1750
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.990886731391586e-05,
      "loss": 0.1817,
      "step": 1760
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9908349514563108e-05,
      "loss": 0.3053,
      "step": 1770
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.9907831715210356e-05,
      "loss": 0.3578,
      "step": 1780
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9907313915857607e-05,
      "loss": 0.5767,
      "step": 1790
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9906796116504855e-05,
      "loss": 0.5208,
      "step": 1800
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.9906278317152106e-05,
      "loss": 0.2593,
      "step": 1810
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9905760517799354e-05,
      "loss": 0.2761,
      "step": 1820
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9905242718446606e-05,
      "loss": 0.3884,
      "step": 1830
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.9904724919093853e-05,
      "loss": 0.2963,
      "step": 1840
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.99042071197411e-05,
      "loss": 0.3972,
      "step": 1850
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.990368932038835e-05,
      "loss": 0.407,
      "step": 1860
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.99031715210356e-05,
      "loss": 0.4125,
      "step": 1870
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.990265372168285e-05,
      "loss": 0.3526,
      "step": 1880
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.99021359223301e-05,
      "loss": 0.2346,
      "step": 1890
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9901618122977348e-05,
      "loss": 0.4111,
      "step": 1900
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.99011003236246e-05,
      "loss": 0.3567,
      "step": 1910
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9900582524271847e-05,
      "loss": 0.3123,
      "step": 1920
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9900064724919095e-05,
      "loss": 0.447,
      "step": 1930
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.9899546925566343e-05,
      "loss": 0.2182,
      "step": 1940
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9899029126213594e-05,
      "loss": 0.196,
      "step": 1950
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.9898511326860842e-05,
      "loss": 0.3428,
      "step": 1960
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9897993527508093e-05,
      "loss": 0.4228,
      "step": 1970
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.989747572815534e-05,
      "loss": 0.4352,
      "step": 1980
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.9896957928802593e-05,
      "loss": 0.3245,
      "step": 1990
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9896440129449837e-05,
      "loss": 0.5161,
      "step": 2000
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.989592233009709e-05,
      "loss": 0.3086,
      "step": 2010
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9895404530744336e-05,
      "loss": 0.3461,
      "step": 2020
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9894886731391588e-05,
      "loss": 0.5122,
      "step": 2030
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9894368932038836e-05,
      "loss": 0.339,
      "step": 2040
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9893851132686087e-05,
      "loss": 0.2856,
      "step": 2050
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9893333333333335e-05,
      "loss": 0.3003,
      "step": 2060
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.9892815533980586e-05,
      "loss": 0.374,
      "step": 2070
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.989229773462783e-05,
      "loss": 0.234,
      "step": 2080
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.9891779935275082e-05,
      "loss": 0.2748,
      "step": 2090
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.989126213592233e-05,
      "loss": 0.4099,
      "step": 2100
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.989074433656958e-05,
      "loss": 0.3507,
      "step": 2110
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.989022653721683e-05,
      "loss": 0.3572,
      "step": 2120
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.988970873786408e-05,
      "loss": 0.3164,
      "step": 2130
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.988919093851133e-05,
      "loss": 0.1405,
      "step": 2140
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.988867313915858e-05,
      "loss": 0.3479,
      "step": 2150
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9888155339805824e-05,
      "loss": 0.3915,
      "step": 2160
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9887637540453076e-05,
      "loss": 0.1772,
      "step": 2170
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.9887119741100324e-05,
      "loss": 0.1413,
      "step": 2180
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.9886601941747575e-05,
      "loss": 0.4333,
      "step": 2190
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.9886084142394823e-05,
      "loss": 0.3713,
      "step": 2200
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.9885566343042074e-05,
      "loss": 0.2158,
      "step": 2210
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.9885048543689322e-05,
      "loss": 0.3854,
      "step": 2220
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.9884530744336573e-05,
      "loss": 0.2415,
      "step": 2230
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.9884012944983818e-05,
      "loss": 0.2313,
      "step": 2240
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.988349514563107e-05,
      "loss": 0.1163,
      "step": 2250
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.9882977346278317e-05,
      "loss": 0.4688,
      "step": 2260
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.988245954692557e-05,
      "loss": 0.3287,
      "step": 2270
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9881941747572816e-05,
      "loss": 0.4998,
      "step": 2280
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9881423948220068e-05,
      "loss": 0.3809,
      "step": 2290
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9880906148867316e-05,
      "loss": 0.3248,
      "step": 2300
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9880388349514567e-05,
      "loss": 0.294,
      "step": 2310
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9879870550161815e-05,
      "loss": 0.3172,
      "step": 2320
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9879352750809063e-05,
      "loss": 0.0986,
      "step": 2330
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.987883495145631e-05,
      "loss": 0.3691,
      "step": 2340
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9878317152103562e-05,
      "loss": 0.2916,
      "step": 2350
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.987779935275081e-05,
      "loss": 0.3384,
      "step": 2360
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.987728155339806e-05,
      "loss": 0.2748,
      "step": 2370
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.987676375404531e-05,
      "loss": 0.4302,
      "step": 2380
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.987624595469256e-05,
      "loss": 0.1998,
      "step": 2390
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.987572815533981e-05,
      "loss": 0.1732,
      "step": 2400
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9875210355987056e-05,
      "loss": 0.2847,
      "step": 2410
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9874692556634304e-05,
      "loss": 0.3177,
      "step": 2420
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.9874174757281555e-05,
      "loss": 0.259,
      "step": 2430
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.9873656957928803e-05,
      "loss": 0.3444,
      "step": 2440
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.9873139158576055e-05,
      "loss": 0.3259,
      "step": 2450
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9872621359223303e-05,
      "loss": 0.3542,
      "step": 2460
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9872103559870554e-05,
      "loss": 0.3646,
      "step": 2470
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9871585760517802e-05,
      "loss": 0.3131,
      "step": 2480
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.987106796116505e-05,
      "loss": 0.2609,
      "step": 2490
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.9870550161812298e-05,
      "loss": 0.135,
      "step": 2500
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.987003236245955e-05,
      "loss": 0.3044,
      "step": 2510
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9869514563106797e-05,
      "loss": 0.3589,
      "step": 2520
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9868996763754048e-05,
      "loss": 0.3952,
      "step": 2530
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9868478964401296e-05,
      "loss": 0.2184,
      "step": 2540
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9867961165048548e-05,
      "loss": 0.237,
      "step": 2550
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9867443365695795e-05,
      "loss": 0.2093,
      "step": 2560
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.9866925566343043e-05,
      "loss": 0.2639,
      "step": 2570
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.946079779917469,
      "eval_loss": 0.21915487945079803,
      "eval_runtime": 8.8172,
      "eval_samples_per_second": 412.261,
      "eval_steps_per_second": 51.603,
      "step": 2575
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.986640776699029e-05,
      "loss": 0.268,
      "step": 2580
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.9865889967637543e-05,
      "loss": 0.2991,
      "step": 2590
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.986537216828479e-05,
      "loss": 0.3589,
      "step": 2600
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.9864854368932042e-05,
      "loss": 0.2938,
      "step": 2610
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.986433656957929e-05,
      "loss": 0.3323,
      "step": 2620
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.986381877022654e-05,
      "loss": 0.3702,
      "step": 2630
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.986330097087379e-05,
      "loss": 0.2629,
      "step": 2640
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.9862783171521037e-05,
      "loss": 0.3314,
      "step": 2650
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.9862265372168285e-05,
      "loss": 0.2506,
      "step": 2660
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9861747572815536e-05,
      "loss": 0.2867,
      "step": 2670
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9861229773462784e-05,
      "loss": 0.2409,
      "step": 2680
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9860711974110035e-05,
      "loss": 0.2056,
      "step": 2690
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.9860194174757283e-05,
      "loss": 0.3161,
      "step": 2700
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.9859676375404535e-05,
      "loss": 0.2486,
      "step": 2710
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.9859158576051783e-05,
      "loss": 0.2606,
      "step": 2720
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.985864077669903e-05,
      "loss": 0.3479,
      "step": 2730
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.985812297734628e-05,
      "loss": 0.3214,
      "step": 2740
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.985760517799353e-05,
      "loss": 0.2541,
      "step": 2750
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.9857087378640778e-05,
      "loss": 0.3196,
      "step": 2760
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.985656957928803e-05,
      "loss": 0.229,
      "step": 2770
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.9856051779935277e-05,
      "loss": 0.2583,
      "step": 2780
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.9855533980582525e-05,
      "loss": 0.2678,
      "step": 2790
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.9855016181229776e-05,
      "loss": 0.2552,
      "step": 2800
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.9854498381877024e-05,
      "loss": 0.2432,
      "step": 2810
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.9853980582524272e-05,
      "loss": 0.2503,
      "step": 2820
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.9853462783171523e-05,
      "loss": 0.3042,
      "step": 2830
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.985294498381877e-05,
      "loss": 0.2171,
      "step": 2840
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.9852427184466022e-05,
      "loss": 0.2555,
      "step": 2850
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.985190938511327e-05,
      "loss": 0.2951,
      "step": 2860
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.985139158576052e-05,
      "loss": 0.2772,
      "step": 2870
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.985087378640777e-05,
      "loss": 0.2971,
      "step": 2880
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.9850355987055018e-05,
      "loss": 0.304,
      "step": 2890
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.9849838187702266e-05,
      "loss": 0.3845,
      "step": 2900
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.9849320388349517e-05,
      "loss": 0.24,
      "step": 2910
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.9848802588996765e-05,
      "loss": 0.3067,
      "step": 2920
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.9848284789644016e-05,
      "loss": 0.3467,
      "step": 2930
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.9847766990291264e-05,
      "loss": 0.1481,
      "step": 2940
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.9847249190938512e-05,
      "loss": 0.1497,
      "step": 2950
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.9846731391585763e-05,
      "loss": 0.3703,
      "step": 2960
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.984621359223301e-05,
      "loss": 0.192,
      "step": 2970
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.984569579288026e-05,
      "loss": 0.3424,
      "step": 2980
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.984517799352751e-05,
      "loss": 0.2612,
      "step": 2990
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.9844660194174758e-05,
      "loss": 0.1468,
      "step": 3000
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.984414239482201e-05,
      "loss": 0.176,
      "step": 3010
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.9843624595469258e-05,
      "loss": 0.2704,
      "step": 3020
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.9843106796116505e-05,
      "loss": 0.1322,
      "step": 3030
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.9842588996763757e-05,
      "loss": 0.3755,
      "step": 3040
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.9842071197411005e-05,
      "loss": 0.165,
      "step": 3050
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9841553398058253e-05,
      "loss": 0.298,
      "step": 3060
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9841035598705504e-05,
      "loss": 0.1262,
      "step": 3070
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9840517799352752e-05,
      "loss": 0.2392,
      "step": 3080
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.9840000000000003e-05,
      "loss": 0.3062,
      "step": 3090
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.983948220064725e-05,
      "loss": 0.251,
      "step": 3100
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.98389644012945e-05,
      "loss": 0.2077,
      "step": 3110
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.983844660194175e-05,
      "loss": 0.362,
      "step": 3120
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9837928802588998e-05,
      "loss": 0.2112,
      "step": 3130
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9837411003236246e-05,
      "loss": 0.2425,
      "step": 3140
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.9836893203883497e-05,
      "loss": 0.2709,
      "step": 3150
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9836375404530745e-05,
      "loss": 0.2759,
      "step": 3160
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9835857605177993e-05,
      "loss": 0.2465,
      "step": 3170
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9835339805825245e-05,
      "loss": 0.2674,
      "step": 3180
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9834822006472493e-05,
      "loss": 0.2807,
      "step": 3190
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9834304207119744e-05,
      "loss": 0.1559,
      "step": 3200
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9833786407766992e-05,
      "loss": 0.2219,
      "step": 3210
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.983326860841424e-05,
      "loss": 0.2027,
      "step": 3220
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.983275080906149e-05,
      "loss": 0.1939,
      "step": 3230
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.983223300970874e-05,
      "loss": 0.4326,
      "step": 3240
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9831715210355987e-05,
      "loss": 0.1498,
      "step": 3250
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9831197411003238e-05,
      "loss": 0.5611,
      "step": 3260
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9830679611650486e-05,
      "loss": 0.4854,
      "step": 3270
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9830161812297737e-05,
      "loss": 0.2406,
      "step": 3280
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9829644012944985e-05,
      "loss": 0.1788,
      "step": 3290
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9829126213592233e-05,
      "loss": 0.3951,
      "step": 3300
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.9828608414239485e-05,
      "loss": 0.1912,
      "step": 3310
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.9828090614886733e-05,
      "loss": 0.1423,
      "step": 3320
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.982757281553398e-05,
      "loss": 0.3095,
      "step": 3330
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.9827055016181232e-05,
      "loss": 0.3002,
      "step": 3340
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.982653721682848e-05,
      "loss": 0.2693,
      "step": 3350
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.982601941747573e-05,
      "loss": 0.5431,
      "step": 3360
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.982550161812298e-05,
      "loss": 0.2523,
      "step": 3370
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.982498381877023e-05,
      "loss": 0.3174,
      "step": 3380
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.9824466019417478e-05,
      "loss": 0.2084,
      "step": 3390
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.9823948220064726e-05,
      "loss": 0.2482,
      "step": 3400
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.9823430420711974e-05,
      "loss": 0.1637,
      "step": 3410
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.9822912621359225e-05,
      "loss": 0.4316,
      "step": 3420
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.9822394822006473e-05,
      "loss": 0.1623,
      "step": 3430
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.9821877022653725e-05,
      "loss": 0.2505,
      "step": 3440
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.9821359223300972e-05,
      "loss": 0.282,
      "step": 3450
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.9820841423948224e-05,
      "loss": 0.2962,
      "step": 3460
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.982032362459547e-05,
      "loss": 0.1224,
      "step": 3470
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.981980582524272e-05,
      "loss": 0.3132,
      "step": 3480
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.9819288025889968e-05,
      "loss": 0.4015,
      "step": 3490
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.981877022653722e-05,
      "loss": 0.3483,
      "step": 3500
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.9818252427184467e-05,
      "loss": 0.2043,
      "step": 3510
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.9817734627831718e-05,
      "loss": 0.3625,
      "step": 3520
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.9817216828478966e-05,
      "loss": 0.2839,
      "step": 3530
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.9816699029126217e-05,
      "loss": 0.2373,
      "step": 3540
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.9816181229773462e-05,
      "loss": 0.2288,
      "step": 3550
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.9815663430420713e-05,
      "loss": 0.2,
      "step": 3560
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.981514563106796e-05,
      "loss": 0.3061,
      "step": 3570
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.9814627831715212e-05,
      "loss": 0.2568,
      "step": 3580
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.981411003236246e-05,
      "loss": 0.3587,
      "step": 3590
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.981359223300971e-05,
      "loss": 0.4764,
      "step": 3600
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.981307443365696e-05,
      "loss": 0.2079,
      "step": 3610
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.981255663430421e-05,
      "loss": 0.2546,
      "step": 3620
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.9812038834951455e-05,
      "loss": 0.3153,
      "step": 3630
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.9811521035598707e-05,
      "loss": 0.1584,
      "step": 3640
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.9811003236245955e-05,
      "loss": 0.4659,
      "step": 3650
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.9810485436893206e-05,
      "loss": 0.1902,
      "step": 3660
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.9809967637540454e-05,
      "loss": 0.2489,
      "step": 3670
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.9809449838187705e-05,
      "loss": 0.4111,
      "step": 3680
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.9808932038834953e-05,
      "loss": 0.3518,
      "step": 3690
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.9808414239482204e-05,
      "loss": 0.3978,
      "step": 3700
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.980789644012945e-05,
      "loss": 0.1893,
      "step": 3710
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.98073786407767e-05,
      "loss": 0.2453,
      "step": 3720
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.9806860841423948e-05,
      "loss": 0.2749,
      "step": 3730
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.98063430420712e-05,
      "loss": 0.2735,
      "step": 3740
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.9805825242718447e-05,
      "loss": 0.2187,
      "step": 3750
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.98053074433657e-05,
      "loss": 0.266,
      "step": 3760
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.9804789644012947e-05,
      "loss": 0.2829,
      "step": 3770
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.9804271844660198e-05,
      "loss": 0.1069,
      "step": 3780
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.9803754045307443e-05,
      "loss": 0.4782,
      "step": 3790
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.9803236245954694e-05,
      "loss": 0.3305,
      "step": 3800
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.9802718446601942e-05,
      "loss": 0.2876,
      "step": 3810
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.9802200647249193e-05,
      "loss": 0.2173,
      "step": 3820
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.980168284789644e-05,
      "loss": 0.3115,
      "step": 3830
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.9801165048543692e-05,
      "loss": 0.266,
      "step": 3840
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.980064724919094e-05,
      "loss": 0.4971,
      "step": 3850
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.980012944983819e-05,
      "loss": 0.1597,
      "step": 3860
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.9799611650485436e-05,
      "loss": 0.3279,
      "step": 3870
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.9799093851132687e-05,
      "loss": 0.2791,
      "step": 3880
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.9798576051779935e-05,
      "loss": 0.2035,
      "step": 3890
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.9798058252427187e-05,
      "loss": 0.3005,
      "step": 3900
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.9797540453074435e-05,
      "loss": 0.2286,
      "step": 3910
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.9797022653721686e-05,
      "loss": 0.2433,
      "step": 3920
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.9796504854368934e-05,
      "loss": 0.1744,
      "step": 3930
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.9795987055016185e-05,
      "loss": 0.2935,
      "step": 3940
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.9795469255663433e-05,
      "loss": 0.2748,
      "step": 3950
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.979495145631068e-05,
      "loss": 0.1726,
      "step": 3960
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.979443365695793e-05,
      "loss": 0.3466,
      "step": 3970
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.979391585760518e-05,
      "loss": 0.3159,
      "step": 3980
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.9793398058252428e-05,
      "loss": 0.2478,
      "step": 3990
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.979288025889968e-05,
      "loss": 0.3374,
      "step": 4000
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.9792362459546927e-05,
      "loss": 0.2616,
      "step": 4010
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.979184466019418e-05,
      "loss": 0.43,
      "step": 4020
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.9791326860841427e-05,
      "loss": 0.1363,
      "step": 4030
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.9790809061488675e-05,
      "loss": 0.1238,
      "step": 4040
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.9790291262135922e-05,
      "loss": 0.3047,
      "step": 4050
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9789773462783174e-05,
      "loss": 0.1906,
      "step": 4060
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.978925566343042e-05,
      "loss": 0.3596,
      "step": 4070
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.9788737864077673e-05,
      "loss": 0.3239,
      "step": 4080
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.978822006472492e-05,
      "loss": 0.4306,
      "step": 4090
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.9787702265372172e-05,
      "loss": 0.1729,
      "step": 4100
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.978718446601942e-05,
      "loss": 0.4153,
      "step": 4110
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.9786666666666668e-05,
      "loss": 0.4504,
      "step": 4120
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.9786148867313916e-05,
      "loss": 0.1816,
      "step": 4130
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9785631067961167e-05,
      "loss": 0.2314,
      "step": 4140
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9785113268608415e-05,
      "loss": 0.3013,
      "step": 4150
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.9784595469255667e-05,
      "loss": 0.2564,
      "step": 4160
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.9784077669902914e-05,
      "loss": 0.1081,
      "step": 4170
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.9783559870550166e-05,
      "loss": 0.3437,
      "step": 4180
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.9783042071197414e-05,
      "loss": 0.2626,
      "step": 4190
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.978252427184466e-05,
      "loss": 0.204,
      "step": 4200
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.978200647249191e-05,
      "loss": 0.1679,
      "step": 4210
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.978148867313916e-05,
      "loss": 0.3692,
      "step": 4220
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.978097087378641e-05,
      "loss": 0.2078,
      "step": 4230
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.978045307443366e-05,
      "loss": 0.3114,
      "step": 4240
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.9779935275080908e-05,
      "loss": 0.3261,
      "step": 4250
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.9779417475728156e-05,
      "loss": 0.3261,
      "step": 4260
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.9778899676375407e-05,
      "loss": 0.1267,
      "step": 4270
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.9778381877022655e-05,
      "loss": 0.1723,
      "step": 4280
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.9777864077669903e-05,
      "loss": 0.2227,
      "step": 4290
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.9777346278317154e-05,
      "loss": 0.3133,
      "step": 4300
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.9776828478964402e-05,
      "loss": 0.2955,
      "step": 4310
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.9776310679611654e-05,
      "loss": 0.2596,
      "step": 4320
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.97757928802589e-05,
      "loss": 0.3367,
      "step": 4330
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.977527508090615e-05,
      "loss": 0.3484,
      "step": 4340
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.97747572815534e-05,
      "loss": 0.2865,
      "step": 4350
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.977423948220065e-05,
      "loss": 0.1856,
      "step": 4360
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.9773721682847897e-05,
      "loss": 0.3409,
      "step": 4370
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.9773203883495148e-05,
      "loss": 0.4301,
      "step": 4380
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.9772686084142396e-05,
      "loss": 0.1799,
      "step": 4390
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.9772168284789647e-05,
      "loss": 0.227,
      "step": 4400
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.9771650485436895e-05,
      "loss": 0.2825,
      "step": 4410
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.9771132686084143e-05,
      "loss": 0.1994,
      "step": 4420
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.9770614886731394e-05,
      "loss": 0.2616,
      "step": 4430
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.9770097087378642e-05,
      "loss": 0.2436,
      "step": 4440
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.976957928802589e-05,
      "loss": 0.1199,
      "step": 4450
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.976906148867314e-05,
      "loss": 0.3097,
      "step": 4460
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.976854368932039e-05,
      "loss": 0.3633,
      "step": 4470
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.976802588996764e-05,
      "loss": 0.4737,
      "step": 4480
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.976750809061489e-05,
      "loss": 0.4733,
      "step": 4490
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.9766990291262137e-05,
      "loss": 0.2922,
      "step": 4500
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.9766472491909388e-05,
      "loss": 0.1793,
      "step": 4510
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.9765954692556636e-05,
      "loss": 0.2487,
      "step": 4520
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.9765436893203884e-05,
      "loss": 0.1645,
      "step": 4530
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.9764919093851135e-05,
      "loss": 0.2696,
      "step": 4540
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.9764401294498383e-05,
      "loss": 0.1997,
      "step": 4550
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.9763883495145634e-05,
      "loss": 0.2653,
      "step": 4560
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.9763365695792882e-05,
      "loss": 0.2104,
      "step": 4570
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.976284789644013e-05,
      "loss": 0.1888,
      "step": 4580
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.976233009708738e-05,
      "loss": 0.3716,
      "step": 4590
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.976181229773463e-05,
      "loss": 0.207,
      "step": 4600
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.9761294498381877e-05,
      "loss": 0.3744,
      "step": 4610
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.976077669902913e-05,
      "loss": 0.4255,
      "step": 4620
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.9760258899676377e-05,
      "loss": 0.366,
      "step": 4630
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.9759741100323625e-05,
      "loss": 0.3249,
      "step": 4640
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.9759223300970876e-05,
      "loss": 0.4179,
      "step": 4650
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.9758705501618124e-05,
      "loss": 0.4137,
      "step": 4660
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.9758187702265375e-05,
      "loss": 0.2887,
      "step": 4670
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.9757669902912623e-05,
      "loss": 0.3047,
      "step": 4680
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.975715210355987e-05,
      "loss": 0.2051,
      "step": 4690
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.9756634304207122e-05,
      "loss": 0.3425,
      "step": 4700
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.975611650485437e-05,
      "loss": 0.3201,
      "step": 4710
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.9755598705501618e-05,
      "loss": 0.2463,
      "step": 4720
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.975508090614887e-05,
      "loss": 0.3898,
      "step": 4730
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.9754563106796117e-05,
      "loss": 0.3263,
      "step": 4740
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.975404530744337e-05,
      "loss": 0.0947,
      "step": 4750
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.9753527508090617e-05,
      "loss": 0.1428,
      "step": 4760
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.9753009708737864e-05,
      "loss": 0.1649,
      "step": 4770
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9752491909385116e-05,
      "loss": 0.2952,
      "step": 4780
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9751974110032364e-05,
      "loss": 0.1883,
      "step": 4790
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.975145631067961e-05,
      "loss": 0.3043,
      "step": 4800
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.9750938511326863e-05,
      "loss": 0.262,
      "step": 4810
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.975042071197411e-05,
      "loss": 0.2134,
      "step": 4820
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9749902912621362e-05,
      "loss": 0.3354,
      "step": 4830
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.974938511326861e-05,
      "loss": 0.4463,
      "step": 4840
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.9748867313915858e-05,
      "loss": 0.3004,
      "step": 4850
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.974834951456311e-05,
      "loss": 0.225,
      "step": 4860
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9747831715210357e-05,
      "loss": 0.3776,
      "step": 4870
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9747313915857605e-05,
      "loss": 0.3347,
      "step": 4880
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9746796116504856e-05,
      "loss": 0.1414,
      "step": 4890
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.9746278317152104e-05,
      "loss": 0.1621,
      "step": 4900
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9745760517799356e-05,
      "loss": 0.2268,
      "step": 4910
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9745242718446604e-05,
      "loss": 0.2664,
      "step": 4920
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.974472491909385e-05,
      "loss": 0.347,
      "step": 4930
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.97442071197411e-05,
      "loss": 0.2537,
      "step": 4940
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.974368932038835e-05,
      "loss": 0.3155,
      "step": 4950
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.97431715210356e-05,
      "loss": 0.2735,
      "step": 4960
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.974265372168285e-05,
      "loss": 0.3012,
      "step": 4970
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9742135922330098e-05,
      "loss": 0.2388,
      "step": 4980
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.974161812297735e-05,
      "loss": 0.332,
      "step": 4990
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.9741100323624597e-05,
      "loss": 0.114,
      "step": 5000
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.9740582524271845e-05,
      "loss": 0.171,
      "step": 5010
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.9740064724919093e-05,
      "loss": 0.3387,
      "step": 5020
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.9739546925566344e-05,
      "loss": 0.2106,
      "step": 5030
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9739029126213592e-05,
      "loss": 0.2017,
      "step": 5040
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9738511326860844e-05,
      "loss": 0.3964,
      "step": 5050
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.973799352750809e-05,
      "loss": 0.2325,
      "step": 5060
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.9737475728155343e-05,
      "loss": 0.2173,
      "step": 5070
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.973695792880259e-05,
      "loss": 0.3135,
      "step": 5080
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.9736440129449842e-05,
      "loss": 0.2401,
      "step": 5090
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.9735922330097087e-05,
      "loss": 0.3315,
      "step": 5100
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.9735404530744338e-05,
      "loss": 0.5079,
      "step": 5110
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.9734886731391586e-05,
      "loss": 0.1738,
      "step": 5120
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.9734368932038837e-05,
      "loss": 0.2585,
      "step": 5130
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.9733851132686085e-05,
      "loss": 0.3199,
      "step": 5140
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.9733333333333336e-05,
      "loss": 0.2457,
      "step": 5150
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9463548830811555,
      "eval_loss": 0.20650140941143036,
      "eval_runtime": 8.2354,
      "eval_samples_per_second": 441.386,
      "eval_steps_per_second": 55.249,
      "step": 5150
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.9732815533980584e-05,
      "loss": 0.2043,
      "step": 5160
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.9732297734627836e-05,
      "loss": 0.3017,
      "step": 5170
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.973177993527508e-05,
      "loss": 0.2352,
      "step": 5180
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.973126213592233e-05,
      "loss": 0.25,
      "step": 5190
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.973074433656958e-05,
      "loss": 0.1942,
      "step": 5200
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.973022653721683e-05,
      "loss": 0.1456,
      "step": 5210
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.972970873786408e-05,
      "loss": 0.1798,
      "step": 5220
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.972919093851133e-05,
      "loss": 0.2216,
      "step": 5230
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.9728673139158578e-05,
      "loss": 0.3577,
      "step": 5240
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.972815533980583e-05,
      "loss": 0.267,
      "step": 5250
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.9727637540453074e-05,
      "loss": 0.179,
      "step": 5260
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9727119741100325e-05,
      "loss": 0.1976,
      "step": 5270
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9726601941747573e-05,
      "loss": 0.2285,
      "step": 5280
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9726084142394824e-05,
      "loss": 0.2729,
      "step": 5290
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.9725566343042072e-05,
      "loss": 0.2632,
      "step": 5300
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.9725048543689324e-05,
      "loss": 0.3938,
      "step": 5310
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.972453074433657e-05,
      "loss": 0.251,
      "step": 5320
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.9724012944983823e-05,
      "loss": 0.1824,
      "step": 5330
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.9723495145631067e-05,
      "loss": 0.1875,
      "step": 5340
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.972297734627832e-05,
      "loss": 0.4658,
      "step": 5350
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.9722459546925567e-05,
      "loss": 0.2576,
      "step": 5360
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.9721941747572818e-05,
      "loss": 0.2174,
      "step": 5370
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.9721423948220066e-05,
      "loss": 0.2031,
      "step": 5380
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.9720906148867317e-05,
      "loss": 0.2399,
      "step": 5390
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.9720388349514565e-05,
      "loss": 0.2067,
      "step": 5400
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.9719870550161816e-05,
      "loss": 0.323,
      "step": 5410
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.971935275080906e-05,
      "loss": 0.3193,
      "step": 5420
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.9718834951456312e-05,
      "loss": 0.1714,
      "step": 5430
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.971831715210356e-05,
      "loss": 0.3194,
      "step": 5440
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.971779935275081e-05,
      "loss": 0.3148,
      "step": 5450
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.971728155339806e-05,
      "loss": 0.2355,
      "step": 5460
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.971676375404531e-05,
      "loss": 0.5888,
      "step": 5470
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.971624595469256e-05,
      "loss": 0.1942,
      "step": 5480
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.971572815533981e-05,
      "loss": 0.3442,
      "step": 5490
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.9715210355987054e-05,
      "loss": 0.2042,
      "step": 5500
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.9714692556634306e-05,
      "loss": 0.3018,
      "step": 5510
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.9714174757281554e-05,
      "loss": 0.2446,
      "step": 5520
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.9713656957928805e-05,
      "loss": 0.0684,
      "step": 5530
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.9713139158576053e-05,
      "loss": 0.2362,
      "step": 5540
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.9712621359223304e-05,
      "loss": 0.1807,
      "step": 5550
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.9712103559870552e-05,
      "loss": 0.3736,
      "step": 5560
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.9711585760517803e-05,
      "loss": 0.3552,
      "step": 5570
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.9711067961165048e-05,
      "loss": 0.165,
      "step": 5580
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.97105501618123e-05,
      "loss": 0.3257,
      "step": 5590
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.9710032362459547e-05,
      "loss": 0.2476,
      "step": 5600
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.97095145631068e-05,
      "loss": 0.1588,
      "step": 5610
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.9708996763754046e-05,
      "loss": 0.1726,
      "step": 5620
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.9708478964401298e-05,
      "loss": 0.2349,
      "step": 5630
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.9707961165048546e-05,
      "loss": 0.2167,
      "step": 5640
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.9707443365695797e-05,
      "loss": 0.1561,
      "step": 5650
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.9706925566343045e-05,
      "loss": 0.3793,
      "step": 5660
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.9706407766990293e-05,
      "loss": 0.1965,
      "step": 5670
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.970588996763754e-05,
      "loss": 0.3141,
      "step": 5680
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.9705372168284792e-05,
      "loss": 0.1715,
      "step": 5690
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.970485436893204e-05,
      "loss": 0.2172,
      "step": 5700
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.970433656957929e-05,
      "loss": 0.1682,
      "step": 5710
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.970381877022654e-05,
      "loss": 0.0884,
      "step": 5720
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.9703300970873787e-05,
      "loss": 0.1397,
      "step": 5730
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.970278317152104e-05,
      "loss": 0.2729,
      "step": 5740
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.9702265372168286e-05,
      "loss": 0.4754,
      "step": 5750
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.9701747572815534e-05,
      "loss": 0.1378,
      "step": 5760
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.9701229773462786e-05,
      "loss": 0.2418,
      "step": 5770
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.9700711974110034e-05,
      "loss": 0.1602,
      "step": 5780
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.9700194174757285e-05,
      "loss": 0.2374,
      "step": 5790
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.9699676375404533e-05,
      "loss": 0.3039,
      "step": 5800
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.969915857605178e-05,
      "loss": 0.1992,
      "step": 5810
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.9698640776699032e-05,
      "loss": 0.253,
      "step": 5820
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.969812297734628e-05,
      "loss": 0.2149,
      "step": 5830
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.9697605177993528e-05,
      "loss": 0.2403,
      "step": 5840
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.969708737864078e-05,
      "loss": 0.1518,
      "step": 5850
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.9696569579288027e-05,
      "loss": 0.1981,
      "step": 5860
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.969605177993528e-05,
      "loss": 0.3326,
      "step": 5870
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.9695533980582526e-05,
      "loss": 0.1952,
      "step": 5880
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.9695016181229774e-05,
      "loss": 0.1975,
      "step": 5890
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.9694498381877026e-05,
      "loss": 0.2171,
      "step": 5900
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.9693980582524273e-05,
      "loss": 0.2066,
      "step": 5910
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.969346278317152e-05,
      "loss": 0.2405,
      "step": 5920
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.9692944983818773e-05,
      "loss": 0.0774,
      "step": 5930
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.969242718446602e-05,
      "loss": 0.2246,
      "step": 5940
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.9691909385113272e-05,
      "loss": 0.3062,
      "step": 5950
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.969139158576052e-05,
      "loss": 0.1248,
      "step": 5960
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.9690873786407768e-05,
      "loss": 0.2669,
      "step": 5970
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.969035598705502e-05,
      "loss": 0.2293,
      "step": 5980
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.9689838187702267e-05,
      "loss": 0.1692,
      "step": 5990
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.9689320388349515e-05,
      "loss": 0.2158,
      "step": 6000
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.9688802588996766e-05,
      "loss": 0.2814,
      "step": 6010
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.9688284789644014e-05,
      "loss": 0.3069,
      "step": 6020
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.9687766990291262e-05,
      "loss": 0.2581,
      "step": 6030
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.9687249190938513e-05,
      "loss": 0.2023,
      "step": 6040
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.968673139158576e-05,
      "loss": 0.2735,
      "step": 6050
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.9686213592233013e-05,
      "loss": 0.2229,
      "step": 6060
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.968569579288026e-05,
      "loss": 0.3472,
      "step": 6070
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.968517799352751e-05,
      "loss": 0.2975,
      "step": 6080
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.968466019417476e-05,
      "loss": 0.2075,
      "step": 6090
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.9684142394822008e-05,
      "loss": 0.2801,
      "step": 6100
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.9683624595469256e-05,
      "loss": 0.1996,
      "step": 6110
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.9683106796116507e-05,
      "loss": 0.3453,
      "step": 6120
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.9682588996763755e-05,
      "loss": 0.252,
      "step": 6130
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.9682071197411006e-05,
      "loss": 0.3192,
      "step": 6140
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.9681553398058254e-05,
      "loss": 0.2626,
      "step": 6150
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.9681035598705502e-05,
      "loss": 0.2132,
      "step": 6160
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.9680517799352753e-05,
      "loss": 0.142,
      "step": 6170
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.968e-05,
      "loss": 0.3669,
      "step": 6180
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.967948220064725e-05,
      "loss": 0.1967,
      "step": 6190
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.96789644012945e-05,
      "loss": 0.2668,
      "step": 6200
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.967844660194175e-05,
      "loss": 0.4368,
      "step": 6210
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9677928802589e-05,
      "loss": 0.2825,
      "step": 6220
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9677411003236248e-05,
      "loss": 0.3686,
      "step": 6230
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.9676893203883496e-05,
      "loss": 0.3099,
      "step": 6240
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.9676375404530747e-05,
      "loss": 0.3312,
      "step": 6250
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.9675857605177995e-05,
      "loss": 0.1675,
      "step": 6260
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.9675339805825243e-05,
      "loss": 0.1346,
      "step": 6270
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.9674822006472494e-05,
      "loss": 0.2002,
      "step": 6280
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.9674304207119742e-05,
      "loss": 0.0604,
      "step": 6290
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.9673786407766993e-05,
      "loss": 0.1218,
      "step": 6300
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.967326860841424e-05,
      "loss": 0.2355,
      "step": 6310
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.967275080906149e-05,
      "loss": 0.2469,
      "step": 6320
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.967223300970874e-05,
      "loss": 0.2579,
      "step": 6330
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.967171521035599e-05,
      "loss": 0.2691,
      "step": 6340
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.9671197411003236e-05,
      "loss": 0.2664,
      "step": 6350
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.9670679611650488e-05,
      "loss": 0.4055,
      "step": 6360
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.9670161812297736e-05,
      "loss": 0.2265,
      "step": 6370
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.9669644012944987e-05,
      "loss": 0.2793,
      "step": 6380
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.9669126213592235e-05,
      "loss": 0.1894,
      "step": 6390
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.9668608414239483e-05,
      "loss": 0.1919,
      "step": 6400
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.966809061488673e-05,
      "loss": 0.229,
      "step": 6410
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.9667572815533982e-05,
      "loss": 0.2536,
      "step": 6420
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.966705501618123e-05,
      "loss": 0.4089,
      "step": 6430
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.966653721682848e-05,
      "loss": 0.1393,
      "step": 6440
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.966601941747573e-05,
      "loss": 0.1711,
      "step": 6450
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.966550161812298e-05,
      "loss": 0.2937,
      "step": 6460
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.966498381877023e-05,
      "loss": 0.314,
      "step": 6470
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.9664466019417476e-05,
      "loss": 0.3018,
      "step": 6480
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.9663948220064724e-05,
      "loss": 0.4462,
      "step": 6490
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.9663430420711976e-05,
      "loss": 0.0828,
      "step": 6500
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.9662912621359223e-05,
      "loss": 0.1944,
      "step": 6510
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.9662394822006475e-05,
      "loss": 0.3108,
      "step": 6520
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.9661877022653723e-05,
      "loss": 0.1355,
      "step": 6530
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.9661359223300974e-05,
      "loss": 0.4539,
      "step": 6540
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.9660841423948222e-05,
      "loss": 0.2266,
      "step": 6550
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.966032362459547e-05,
      "loss": 0.1012,
      "step": 6560
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.9659805825242718e-05,
      "loss": 0.2279,
      "step": 6570
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.965928802588997e-05,
      "loss": 0.1487,
      "step": 6580
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.9658770226537217e-05,
      "loss": 0.2495,
      "step": 6590
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.965825242718447e-05,
      "loss": 0.2881,
      "step": 6600
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.9657734627831716e-05,
      "loss": 0.4532,
      "step": 6610
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.9657216828478968e-05,
      "loss": 0.1642,
      "step": 6620
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.9656699029126215e-05,
      "loss": 0.2477,
      "step": 6630
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.9656181229773463e-05,
      "loss": 0.1566,
      "step": 6640
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.965566343042071e-05,
      "loss": 0.3306,
      "step": 6650
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.9655145631067963e-05,
      "loss": 0.232,
      "step": 6660
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.965462783171521e-05,
      "loss": 0.1865,
      "step": 6670
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.9654110032362462e-05,
      "loss": 0.3633,
      "step": 6680
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.965359223300971e-05,
      "loss": 0.3195,
      "step": 6690
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.965307443365696e-05,
      "loss": 0.1437,
      "step": 6700
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.965255663430421e-05,
      "loss": 0.2772,
      "step": 6710
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.9652038834951457e-05,
      "loss": 0.4101,
      "step": 6720
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.9651521035598705e-05,
      "loss": 0.1842,
      "step": 6730
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.9651003236245956e-05,
      "loss": 0.2516,
      "step": 6740
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.9650485436893204e-05,
      "loss": 0.3446,
      "step": 6750
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.9649967637540455e-05,
      "loss": 0.3189,
      "step": 6760
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.9649449838187703e-05,
      "loss": 0.159,
      "step": 6770
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.9648932038834955e-05,
      "loss": 0.1236,
      "step": 6780
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.9648414239482203e-05,
      "loss": 0.2854,
      "step": 6790
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.9647896440129454e-05,
      "loss": 0.1754,
      "step": 6800
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.96473786407767e-05,
      "loss": 0.2756,
      "step": 6810
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.964686084142395e-05,
      "loss": 0.5186,
      "step": 6820
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.9646343042071198e-05,
      "loss": 0.2379,
      "step": 6830
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.964582524271845e-05,
      "loss": 0.3095,
      "step": 6840
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.9645307443365697e-05,
      "loss": 0.1464,
      "step": 6850
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.9644789644012948e-05,
      "loss": 0.213,
      "step": 6860
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.9644271844660196e-05,
      "loss": 0.1876,
      "step": 6870
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.9643754045307447e-05,
      "loss": 0.1391,
      "step": 6880
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.9643236245954692e-05,
      "loss": 0.1769,
      "step": 6890
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.9642718446601943e-05,
      "loss": 0.4605,
      "step": 6900
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.964220064724919e-05,
      "loss": 0.229,
      "step": 6910
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.9641682847896443e-05,
      "loss": 0.2681,
      "step": 6920
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.964116504854369e-05,
      "loss": 0.3891,
      "step": 6930
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.9640647249190942e-05,
      "loss": 0.2151,
      "step": 6940
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.964012944983819e-05,
      "loss": 0.4364,
      "step": 6950
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.963961165048544e-05,
      "loss": 0.3009,
      "step": 6960
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.9639093851132686e-05,
      "loss": 0.0845,
      "step": 6970
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.9638576051779937e-05,
      "loss": 0.1373,
      "step": 6980
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.9638058252427185e-05,
      "loss": 0.1446,
      "step": 6990
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.9637540453074436e-05,
      "loss": 0.1031,
      "step": 7000
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.9637022653721684e-05,
      "loss": 0.2145,
      "step": 7010
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.9636504854368935e-05,
      "loss": 0.1547,
      "step": 7020
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.9635987055016183e-05,
      "loss": 0.2423,
      "step": 7030
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.9635469255663435e-05,
      "loss": 0.4544,
      "step": 7040
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.963495145631068e-05,
      "loss": 0.421,
      "step": 7050
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.963443365695793e-05,
      "loss": 0.3462,
      "step": 7060
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.963391585760518e-05,
      "loss": 0.1668,
      "step": 7070
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.963339805825243e-05,
      "loss": 0.1112,
      "step": 7080
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.9632880258899678e-05,
      "loss": 0.1069,
      "step": 7090
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.963236245954693e-05,
      "loss": 0.1501,
      "step": 7100
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.9631844660194177e-05,
      "loss": 0.2587,
      "step": 7110
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.9631326860841428e-05,
      "loss": 0.163,
      "step": 7120
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.9630809061488673e-05,
      "loss": 0.3376,
      "step": 7130
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.9630291262135924e-05,
      "loss": 0.2132,
      "step": 7140
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.9629773462783172e-05,
      "loss": 0.2756,
      "step": 7150
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.9629255663430423e-05,
      "loss": 0.22,
      "step": 7160
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.962873786407767e-05,
      "loss": 0.1825,
      "step": 7170
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.9628220064724922e-05,
      "loss": 0.204,
      "step": 7180
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.962770226537217e-05,
      "loss": 0.1456,
      "step": 7190
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.962718446601942e-05,
      "loss": 0.0983,
      "step": 7200
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.9626666666666666e-05,
      "loss": 0.252,
      "step": 7210
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.9626148867313918e-05,
      "loss": 0.1811,
      "step": 7220
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9625631067961165e-05,
      "loss": 0.2915,
      "step": 7230
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.9625113268608417e-05,
      "loss": 0.3266,
      "step": 7240
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9624595469255665e-05,
      "loss": 0.3237,
      "step": 7250
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9624077669902916e-05,
      "loss": 0.475,
      "step": 7260
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9623559870550164e-05,
      "loss": 0.4171,
      "step": 7270
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9623042071197412e-05,
      "loss": 0.4969,
      "step": 7280
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.9622524271844663e-05,
      "loss": 0.1132,
      "step": 7290
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.962200647249191e-05,
      "loss": 0.1173,
      "step": 7300
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.962148867313916e-05,
      "loss": 0.2134,
      "step": 7310
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.962097087378641e-05,
      "loss": 0.1437,
      "step": 7320
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.9620453074433658e-05,
      "loss": 0.1454,
      "step": 7330
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.961993527508091e-05,
      "loss": 0.1514,
      "step": 7340
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.9619417475728157e-05,
      "loss": 0.2684,
      "step": 7350
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.9618899676375405e-05,
      "loss": 0.2487,
      "step": 7360
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.9618381877022657e-05,
      "loss": 0.2238,
      "step": 7370
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.9617864077669905e-05,
      "loss": 0.137,
      "step": 7380
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.9617346278317153e-05,
      "loss": 0.3095,
      "step": 7390
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.9616828478964404e-05,
      "loss": 0.3518,
      "step": 7400
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.9616310679611652e-05,
      "loss": 0.1554,
      "step": 7410
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.9615792880258903e-05,
      "loss": 0.2279,
      "step": 7420
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.961527508090615e-05,
      "loss": 0.4632,
      "step": 7430
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.96147572815534e-05,
      "loss": 0.2284,
      "step": 7440
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.961423948220065e-05,
      "loss": 0.1091,
      "step": 7450
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.9613721682847898e-05,
      "loss": 0.2456,
      "step": 7460
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.9613203883495146e-05,
      "loss": 0.1793,
      "step": 7470
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.9612686084142397e-05,
      "loss": 0.3559,
      "step": 7480
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.9612168284789645e-05,
      "loss": 0.2734,
      "step": 7490
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.9611650485436893e-05,
      "loss": 0.3259,
      "step": 7500
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.9611132686084145e-05,
      "loss": 0.157,
      "step": 7510
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.9610614886731393e-05,
      "loss": 0.2073,
      "step": 7520
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.9610097087378644e-05,
      "loss": 0.1661,
      "step": 7530
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.9609579288025892e-05,
      "loss": 0.2655,
      "step": 7540
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.960906148867314e-05,
      "loss": 0.438,
      "step": 7550
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.960854368932039e-05,
      "loss": 0.2627,
      "step": 7560
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.960802588996764e-05,
      "loss": 0.1767,
      "step": 7570
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9607508090614887e-05,
      "loss": 0.1808,
      "step": 7580
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.9606990291262138e-05,
      "loss": 0.1576,
      "step": 7590
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.9606472491909386e-05,
      "loss": 0.1789,
      "step": 7600
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.9605954692556637e-05,
      "loss": 0.0895,
      "step": 7610
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.9605436893203885e-05,
      "loss": 0.2316,
      "step": 7620
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.9604919093851133e-05,
      "loss": 0.1546,
      "step": 7630
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.9604401294498385e-05,
      "loss": 0.1032,
      "step": 7640
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.9603883495145632e-05,
      "loss": 0.2615,
      "step": 7650
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.960336569579288e-05,
      "loss": 0.2983,
      "step": 7660
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.9602847896440132e-05,
      "loss": 0.1713,
      "step": 7670
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.960233009708738e-05,
      "loss": 0.3184,
      "step": 7680
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.960181229773463e-05,
      "loss": 0.2289,
      "step": 7690
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.960129449838188e-05,
      "loss": 0.1238,
      "step": 7700
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.9600776699029127e-05,
      "loss": 0.3203,
      "step": 7710
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.9600258899676378e-05,
      "loss": 0.3157,
      "step": 7720
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.18269842863082886,
      "eval_runtime": 8.2373,
      "eval_samples_per_second": 441.284,
      "eval_steps_per_second": 55.236,
      "step": 7725
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.9599741100323626e-05,
      "loss": 0.0376,
      "step": 7730
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.9599223300970874e-05,
      "loss": 0.4194,
      "step": 7740
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.9598705501618125e-05,
      "loss": 0.1455,
      "step": 7750
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.9598187702265373e-05,
      "loss": 0.4301,
      "step": 7760
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.9597669902912625e-05,
      "loss": 0.2781,
      "step": 7770
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.9597152103559872e-05,
      "loss": 0.3277,
      "step": 7780
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.959663430420712e-05,
      "loss": 0.106,
      "step": 7790
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.959611650485437e-05,
      "loss": 0.2019,
      "step": 7800
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.959559870550162e-05,
      "loss": 0.1731,
      "step": 7810
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9595080906148868e-05,
      "loss": 0.3396,
      "step": 7820
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.959456310679612e-05,
      "loss": 0.4465,
      "step": 7830
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9594045307443367e-05,
      "loss": 0.3698,
      "step": 7840
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9593527508090618e-05,
      "loss": 0.2267,
      "step": 7850
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9593009708737866e-05,
      "loss": 0.3049,
      "step": 7860
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9592491909385114e-05,
      "loss": 0.2227,
      "step": 7870
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9591974110032362e-05,
      "loss": 0.168,
      "step": 7880
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9591456310679613e-05,
      "loss": 0.3125,
      "step": 7890
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.959093851132686e-05,
      "loss": 0.2256,
      "step": 7900
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.9590420711974112e-05,
      "loss": 0.2052,
      "step": 7910
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.958990291262136e-05,
      "loss": 0.2826,
      "step": 7920
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.958938511326861e-05,
      "loss": 0.2107,
      "step": 7930
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.958886731391586e-05,
      "loss": 0.3363,
      "step": 7940
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.9588349514563107e-05,
      "loss": 0.221,
      "step": 7950
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.9587831715210355e-05,
      "loss": 0.0989,
      "step": 7960
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.9587313915857607e-05,
      "loss": 0.1815,
      "step": 7970
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.9586796116504855e-05,
      "loss": 0.1745,
      "step": 7980
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.9586278317152106e-05,
      "loss": 0.1451,
      "step": 7990
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.9585760517799354e-05,
      "loss": 0.1384,
      "step": 8000
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.9585242718446605e-05,
      "loss": 0.1165,
      "step": 8010
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.9584724919093853e-05,
      "loss": 0.137,
      "step": 8020
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.95842071197411e-05,
      "loss": 0.3212,
      "step": 8030
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.958368932038835e-05,
      "loss": 0.2297,
      "step": 8040
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.95831715210356e-05,
      "loss": 0.5524,
      "step": 8050
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.9582653721682848e-05,
      "loss": 0.2841,
      "step": 8060
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.95821359223301e-05,
      "loss": 0.1424,
      "step": 8070
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.9581618122977347e-05,
      "loss": 0.1698,
      "step": 8080
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.95811003236246e-05,
      "loss": 0.2587,
      "step": 8090
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.9580582524271847e-05,
      "loss": 0.1922,
      "step": 8100
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.9580064724919095e-05,
      "loss": 0.1846,
      "step": 8110
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.9579546925566343e-05,
      "loss": 0.2551,
      "step": 8120
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.9579029126213594e-05,
      "loss": 0.3196,
      "step": 8130
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.9578511326860842e-05,
      "loss": 0.1019,
      "step": 8140
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.9577993527508093e-05,
      "loss": 0.1025,
      "step": 8150
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.957747572815534e-05,
      "loss": 0.152,
      "step": 8160
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.9576957928802592e-05,
      "loss": 0.0749,
      "step": 8170
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.957644012944984e-05,
      "loss": 0.1116,
      "step": 8180
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.9575922330097088e-05,
      "loss": 0.2604,
      "step": 8190
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.9575404530744336e-05,
      "loss": 0.4164,
      "step": 8200
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.9574886731391587e-05,
      "loss": 0.1808,
      "step": 8210
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.9574368932038835e-05,
      "loss": 0.2979,
      "step": 8220
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.9573851132686087e-05,
      "loss": 0.2347,
      "step": 8230
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.9573333333333335e-05,
      "loss": 0.2159,
      "step": 8240
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.9572815533980586e-05,
      "loss": 0.4439,
      "step": 8250
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.9572297734627834e-05,
      "loss": 0.2083,
      "step": 8260
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.9571779935275082e-05,
      "loss": 0.0448,
      "step": 8270
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.957126213592233e-05,
      "loss": 0.2388,
      "step": 8280
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.957074433656958e-05,
      "loss": 0.1335,
      "step": 8290
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.957022653721683e-05,
      "loss": 0.4199,
      "step": 8300
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.956970873786408e-05,
      "loss": 0.1894,
      "step": 8310
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.9569190938511328e-05,
      "loss": 0.1142,
      "step": 8320
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.956867313915858e-05,
      "loss": 0.1971,
      "step": 8330
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.9568155339805827e-05,
      "loss": 0.2359,
      "step": 8340
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.9567637540453075e-05,
      "loss": 0.2582,
      "step": 8350
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.9567119741100323e-05,
      "loss": 0.1369,
      "step": 8360
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.9566601941747574e-05,
      "loss": 0.3086,
      "step": 8370
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.9566084142394822e-05,
      "loss": 0.3685,
      "step": 8380
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.9565566343042074e-05,
      "loss": 0.3185,
      "step": 8390
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.956504854368932e-05,
      "loss": 0.2913,
      "step": 8400
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.9564530744336573e-05,
      "loss": 0.139,
      "step": 8410
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.956401294498382e-05,
      "loss": 0.2689,
      "step": 8420
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.9563495145631072e-05,
      "loss": 0.2636,
      "step": 8430
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.9562977346278317e-05,
      "loss": 0.2229,
      "step": 8440
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.9562459546925568e-05,
      "loss": 0.1888,
      "step": 8450
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.9561941747572816e-05,
      "loss": 0.1098,
      "step": 8460
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.9561423948220067e-05,
      "loss": 0.13,
      "step": 8470
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.9560906148867315e-05,
      "loss": 0.1429,
      "step": 8480
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.9560388349514567e-05,
      "loss": 0.2206,
      "step": 8490
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.9559870550161814e-05,
      "loss": 0.1083,
      "step": 8500
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.9559352750809066e-05,
      "loss": 0.0738,
      "step": 8510
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.955883495145631e-05,
      "loss": 0.2183,
      "step": 8520
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.955831715210356e-05,
      "loss": 0.1192,
      "step": 8530
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.955779935275081e-05,
      "loss": 0.3058,
      "step": 8540
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.955728155339806e-05,
      "loss": 0.3926,
      "step": 8550
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.955676375404531e-05,
      "loss": 0.1216,
      "step": 8560
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.955624595469256e-05,
      "loss": 0.2993,
      "step": 8570
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.9555728155339808e-05,
      "loss": 0.2026,
      "step": 8580
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.955521035598706e-05,
      "loss": 0.2474,
      "step": 8590
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.9554692556634304e-05,
      "loss": 0.1775,
      "step": 8600
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.9554174757281555e-05,
      "loss": 0.2943,
      "step": 8610
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.9553656957928803e-05,
      "loss": 0.2321,
      "step": 8620
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.9553139158576054e-05,
      "loss": 0.148,
      "step": 8630
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.9552621359223302e-05,
      "loss": 0.1469,
      "step": 8640
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.9552103559870554e-05,
      "loss": 0.1932,
      "step": 8650
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.95515857605178e-05,
      "loss": 0.1786,
      "step": 8660
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.955106796116505e-05,
      "loss": 0.3065,
      "step": 8670
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.9550550161812297e-05,
      "loss": 0.125,
      "step": 8680
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.955003236245955e-05,
      "loss": 0.1871,
      "step": 8690
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.9549514563106797e-05,
      "loss": 0.41,
      "step": 8700
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.9548996763754048e-05,
      "loss": 0.1822,
      "step": 8710
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.9548478964401296e-05,
      "loss": 0.1299,
      "step": 8720
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.9547961165048547e-05,
      "loss": 0.194,
      "step": 8730
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.9547443365695795e-05,
      "loss": 0.1477,
      "step": 8740
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.9546925566343043e-05,
      "loss": 0.4681,
      "step": 8750
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.954640776699029e-05,
      "loss": 0.4321,
      "step": 8760
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.9545889967637542e-05,
      "loss": 0.2347,
      "step": 8770
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.954537216828479e-05,
      "loss": 0.2222,
      "step": 8780
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.954485436893204e-05,
      "loss": 0.116,
      "step": 8790
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.954433656957929e-05,
      "loss": 0.1101,
      "step": 8800
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.954381877022654e-05,
      "loss": 0.1199,
      "step": 8810
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.954330097087379e-05,
      "loss": 0.2333,
      "step": 8820
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.9542783171521037e-05,
      "loss": 0.291,
      "step": 8830
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.9542265372168285e-05,
      "loss": 0.1248,
      "step": 8840
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.9541747572815536e-05,
      "loss": 0.2109,
      "step": 8850
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.9541229773462784e-05,
      "loss": 0.1833,
      "step": 8860
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.9540711974110035e-05,
      "loss": 0.196,
      "step": 8870
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.9540194174757283e-05,
      "loss": 0.1072,
      "step": 8880
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.9539676375404534e-05,
      "loss": 0.2757,
      "step": 8890
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.9539158576051782e-05,
      "loss": 0.2158,
      "step": 8900
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.953864077669903e-05,
      "loss": 0.3908,
      "step": 8910
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.9538122977346278e-05,
      "loss": 0.4817,
      "step": 8920
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.953760517799353e-05,
      "loss": 0.2603,
      "step": 8930
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.9537087378640777e-05,
      "loss": 0.0576,
      "step": 8940
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.953656957928803e-05,
      "loss": 0.0653,
      "step": 8950
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.9536051779935277e-05,
      "loss": 0.1654,
      "step": 8960
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.9535533980582524e-05,
      "loss": 0.2734,
      "step": 8970
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.9535016181229776e-05,
      "loss": 0.2598,
      "step": 8980
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.9534498381877024e-05,
      "loss": 0.2994,
      "step": 8990
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.9533980582524275e-05,
      "loss": 0.1238,
      "step": 9000
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.9533462783171523e-05,
      "loss": 0.1811,
      "step": 9010
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.953294498381877e-05,
      "loss": 0.3206,
      "step": 9020
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.9532427184466022e-05,
      "loss": 0.2931,
      "step": 9030
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.953190938511327e-05,
      "loss": 0.2408,
      "step": 9040
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.9531391585760518e-05,
      "loss": 0.2155,
      "step": 9050
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.953087378640777e-05,
      "loss": 0.2846,
      "step": 9060
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.9530355987055017e-05,
      "loss": 0.3123,
      "step": 9070
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.952983818770227e-05,
      "loss": 0.0593,
      "step": 9080
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.9529320388349516e-05,
      "loss": 0.2538,
      "step": 9090
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.9528802588996764e-05,
      "loss": 0.3458,
      "step": 9100
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.9528284789644016e-05,
      "loss": 0.1533,
      "step": 9110
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.9527766990291264e-05,
      "loss": 0.2968,
      "step": 9120
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.952724919093851e-05,
      "loss": 0.1229,
      "step": 9130
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.9526731391585763e-05,
      "loss": 0.1086,
      "step": 9140
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.952621359223301e-05,
      "loss": 0.2779,
      "step": 9150
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.9525695792880262e-05,
      "loss": 0.2477,
      "step": 9160
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.952517799352751e-05,
      "loss": 0.1254,
      "step": 9170
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.9524660194174758e-05,
      "loss": 0.3448,
      "step": 9180
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.952414239482201e-05,
      "loss": 0.2008,
      "step": 9190
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.9523624595469257e-05,
      "loss": 0.1633,
      "step": 9200
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.9523106796116505e-05,
      "loss": 0.1776,
      "step": 9210
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.9522588996763756e-05,
      "loss": 0.2706,
      "step": 9220
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.9522071197411004e-05,
      "loss": 0.1618,
      "step": 9230
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.9521553398058256e-05,
      "loss": 0.2364,
      "step": 9240
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.9521035598705504e-05,
      "loss": 0.2281,
      "step": 9250
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.952051779935275e-05,
      "loss": 0.1844,
      "step": 9260
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.9520000000000003e-05,
      "loss": 0.3725,
      "step": 9270
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.951948220064725e-05,
      "loss": 0.208,
      "step": 9280
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.95189644012945e-05,
      "loss": 0.1317,
      "step": 9290
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.951844660194175e-05,
      "loss": 0.2119,
      "step": 9300
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.9517928802588998e-05,
      "loss": 0.1847,
      "step": 9310
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.951741100323625e-05,
      "loss": 0.1502,
      "step": 9320
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.9516893203883497e-05,
      "loss": 0.0884,
      "step": 9330
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.9516375404530745e-05,
      "loss": 0.3843,
      "step": 9340
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.9515857605177993e-05,
      "loss": 0.2725,
      "step": 9350
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.9515339805825244e-05,
      "loss": 0.0753,
      "step": 9360
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.9514822006472492e-05,
      "loss": 0.3185,
      "step": 9370
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.9514304207119744e-05,
      "loss": 0.0921,
      "step": 9380
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.951378640776699e-05,
      "loss": 0.2169,
      "step": 9390
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.9513268608414243e-05,
      "loss": 0.1702,
      "step": 9400
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.951275080906149e-05,
      "loss": 0.2917,
      "step": 9410
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.951223300970874e-05,
      "loss": 0.1971,
      "step": 9420
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.9511715210355987e-05,
      "loss": 0.2863,
      "step": 9430
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.9511197411003238e-05,
      "loss": 0.1765,
      "step": 9440
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.9510679611650486e-05,
      "loss": 0.2719,
      "step": 9450
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.9510161812297737e-05,
      "loss": 0.2328,
      "step": 9460
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.9509644012944985e-05,
      "loss": 0.2412,
      "step": 9470
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.9509126213592236e-05,
      "loss": 0.2867,
      "step": 9480
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.9508608414239484e-05,
      "loss": 0.1914,
      "step": 9490
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.9508090614886732e-05,
      "loss": 0.1623,
      "step": 9500
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.950757281553398e-05,
      "loss": 0.1388,
      "step": 9510
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.950705501618123e-05,
      "loss": 0.1678,
      "step": 9520
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.950653721682848e-05,
      "loss": 0.3427,
      "step": 9530
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.950601941747573e-05,
      "loss": 0.0431,
      "step": 9540
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.950550161812298e-05,
      "loss": 0.2108,
      "step": 9550
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.950498381877023e-05,
      "loss": 0.2641,
      "step": 9560
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.9504466019417478e-05,
      "loss": 0.3133,
      "step": 9570
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.9503948220064726e-05,
      "loss": 0.489,
      "step": 9580
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.9503430420711974e-05,
      "loss": 0.358,
      "step": 9590
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.9502912621359225e-05,
      "loss": 0.1569,
      "step": 9600
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.9502394822006473e-05,
      "loss": 0.3264,
      "step": 9610
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.9501877022653724e-05,
      "loss": 0.0809,
      "step": 9620
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.9501359223300972e-05,
      "loss": 0.4272,
      "step": 9630
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.9500841423948223e-05,
      "loss": 0.2774,
      "step": 9640
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.950032362459547e-05,
      "loss": 0.2113,
      "step": 9650
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.949980582524272e-05,
      "loss": 0.2256,
      "step": 9660
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.9499288025889967e-05,
      "loss": 0.083,
      "step": 9670
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.949877022653722e-05,
      "loss": 0.0949,
      "step": 9680
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.9498252427184466e-05,
      "loss": 0.248,
      "step": 9690
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.9497734627831718e-05,
      "loss": 0.1129,
      "step": 9700
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.9497216828478966e-05,
      "loss": 0.1563,
      "step": 9710
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.9496699029126217e-05,
      "loss": 0.2122,
      "step": 9720
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.9496181229773465e-05,
      "loss": 0.1765,
      "step": 9730
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.9495663430420713e-05,
      "loss": 0.2865,
      "step": 9740
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.949514563106796e-05,
      "loss": 0.274,
      "step": 9750
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.9494627831715212e-05,
      "loss": 0.2661,
      "step": 9760
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.949411003236246e-05,
      "loss": 0.1903,
      "step": 9770
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.949359223300971e-05,
      "loss": 0.1783,
      "step": 9780
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.949307443365696e-05,
      "loss": 0.1957,
      "step": 9790
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.949255663430421e-05,
      "loss": 0.1045,
      "step": 9800
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.949203883495146e-05,
      "loss": 0.1721,
      "step": 9810
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.9491521035598706e-05,
      "loss": 0.2984,
      "step": 9820
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.9491003236245954e-05,
      "loss": 0.1043,
      "step": 9830
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.9490485436893206e-05,
      "loss": 0.2982,
      "step": 9840
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.9489967637540454e-05,
      "loss": 0.0695,
      "step": 9850
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.9489449838187705e-05,
      "loss": 0.1379,
      "step": 9860
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.9488932038834953e-05,
      "loss": 0.6523,
      "step": 9870
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.9488414239482204e-05,
      "loss": 0.1844,
      "step": 9880
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.9487896440129452e-05,
      "loss": 0.1405,
      "step": 9890
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.94873786407767e-05,
      "loss": 0.263,
      "step": 9900
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.9486860841423948e-05,
      "loss": 0.2052,
      "step": 9910
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.94863430420712e-05,
      "loss": 0.3914,
      "step": 9920
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.9485825242718447e-05,
      "loss": 0.2323,
      "step": 9930
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.94853074433657e-05,
      "loss": 0.2894,
      "step": 9940
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.9484789644012946e-05,
      "loss": 0.0943,
      "step": 9950
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.9484271844660198e-05,
      "loss": 0.3892,
      "step": 9960
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.9483754045307446e-05,
      "loss": 0.172,
      "step": 9970
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.9483236245954694e-05,
      "loss": 0.2609,
      "step": 9980
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.948271844660194e-05,
      "loss": 0.2567,
      "step": 9990
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.9482200647249193e-05,
      "loss": 0.0511,
      "step": 10000
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.948168284789644e-05,
      "loss": 0.224,
      "step": 10010
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.9481165048543692e-05,
      "loss": 0.2571,
      "step": 10020
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.948064724919094e-05,
      "loss": 0.1287,
      "step": 10030
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.948012944983819e-05,
      "loss": 0.2959,
      "step": 10040
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.947961165048544e-05,
      "loss": 0.1872,
      "step": 10050
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.9479093851132687e-05,
      "loss": 0.1789,
      "step": 10060
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.9478576051779935e-05,
      "loss": 0.4255,
      "step": 10070
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.9478058252427186e-05,
      "loss": 0.1198,
      "step": 10080
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.9477540453074434e-05,
      "loss": 0.2448,
      "step": 10090
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.9477022653721686e-05,
      "loss": 0.1974,
      "step": 10100
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.9476504854368933e-05,
      "loss": 0.335,
      "step": 10110
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.9475987055016185e-05,
      "loss": 0.2979,
      "step": 10120
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.9475469255663433e-05,
      "loss": 0.1669,
      "step": 10130
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.947495145631068e-05,
      "loss": 0.2886,
      "step": 10140
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.947443365695793e-05,
      "loss": 0.2445,
      "step": 10150
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.947391585760518e-05,
      "loss": 0.3029,
      "step": 10160
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.9473398058252428e-05,
      "loss": 0.156,
      "step": 10170
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.947288025889968e-05,
      "loss": 0.1507,
      "step": 10180
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.9472362459546927e-05,
      "loss": 0.133,
      "step": 10190
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.947184466019418e-05,
      "loss": 0.2364,
      "step": 10200
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.9471326860841426e-05,
      "loss": 0.1888,
      "step": 10210
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.9470809061488674e-05,
      "loss": 0.3673,
      "step": 10220
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.9470291262135922e-05,
      "loss": 0.1687,
      "step": 10230
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.9469773462783173e-05,
      "loss": 0.082,
      "step": 10240
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.946925566343042e-05,
      "loss": 0.2059,
      "step": 10250
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.9468737864077673e-05,
      "loss": 0.3736,
      "step": 10260
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.946822006472492e-05,
      "loss": 0.1831,
      "step": 10270
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.9467702265372172e-05,
      "loss": 0.1977,
      "step": 10280
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.946718446601942e-05,
      "loss": 0.1659,
      "step": 10290
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.9466666666666668e-05,
      "loss": 0.3149,
      "step": 10300
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9488308115543329,
      "eval_loss": 0.1855359673500061,
      "eval_runtime": 8.1547,
      "eval_samples_per_second": 445.755,
      "eval_steps_per_second": 55.796,
      "step": 10300
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.9466148867313916e-05,
      "loss": 0.1827,
      "step": 10310
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.9465631067961167e-05,
      "loss": 0.2903,
      "step": 10320
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.9465113268608415e-05,
      "loss": 0.2063,
      "step": 10330
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.9464595469255666e-05,
      "loss": 0.0959,
      "step": 10340
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.9464077669902914e-05,
      "loss": 0.2297,
      "step": 10350
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.9463559870550165e-05,
      "loss": 0.1273,
      "step": 10360
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.9463042071197413e-05,
      "loss": 0.2147,
      "step": 10370
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.946252427184466e-05,
      "loss": 0.1313,
      "step": 10380
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.946200647249191e-05,
      "loss": 0.3251,
      "step": 10390
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.946148867313916e-05,
      "loss": 0.3569,
      "step": 10400
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.946097087378641e-05,
      "loss": 0.1368,
      "step": 10410
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.946045307443366e-05,
      "loss": 0.116,
      "step": 10420
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.9459935275080908e-05,
      "loss": 0.3314,
      "step": 10430
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.9459417475728156e-05,
      "loss": 0.2843,
      "step": 10440
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.9458899676375407e-05,
      "loss": 0.0952,
      "step": 10450
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.9458381877022655e-05,
      "loss": 0.3193,
      "step": 10460
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.9457864077669903e-05,
      "loss": 0.172,
      "step": 10470
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.9457346278317154e-05,
      "loss": 0.1472,
      "step": 10480
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.9456828478964402e-05,
      "loss": 0.2518,
      "step": 10490
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.9456310679611653e-05,
      "loss": 0.189,
      "step": 10500
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.94557928802589e-05,
      "loss": 0.221,
      "step": 10510
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.945527508090615e-05,
      "loss": 0.1219,
      "step": 10520
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.94547572815534e-05,
      "loss": 0.0485,
      "step": 10530
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.945423948220065e-05,
      "loss": 0.2431,
      "step": 10540
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.9453721682847896e-05,
      "loss": 0.3147,
      "step": 10550
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.9453203883495148e-05,
      "loss": 0.3031,
      "step": 10560
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.9452686084142396e-05,
      "loss": 0.1062,
      "step": 10570
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.9452168284789647e-05,
      "loss": 0.1931,
      "step": 10580
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.9451650485436895e-05,
      "loss": 0.2967,
      "step": 10590
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.9451132686084143e-05,
      "loss": 0.2966,
      "step": 10600
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.9450614886731394e-05,
      "loss": 0.1978,
      "step": 10610
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.9450097087378642e-05,
      "loss": 0.079,
      "step": 10620
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.944957928802589e-05,
      "loss": 0.1591,
      "step": 10630
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.944906148867314e-05,
      "loss": 0.2534,
      "step": 10640
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.944854368932039e-05,
      "loss": 0.1422,
      "step": 10650
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.944802588996764e-05,
      "loss": 0.0285,
      "step": 10660
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.944750809061489e-05,
      "loss": 0.2822,
      "step": 10670
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.9446990291262136e-05,
      "loss": 0.1967,
      "step": 10680
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.9446472491909388e-05,
      "loss": 0.3299,
      "step": 10690
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.9445954692556636e-05,
      "loss": 0.1077,
      "step": 10700
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.9445436893203887e-05,
      "loss": 0.2159,
      "step": 10710
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.9444919093851135e-05,
      "loss": 0.2205,
      "step": 10720
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.9444401294498383e-05,
      "loss": 0.1952,
      "step": 10730
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.9443883495145634e-05,
      "loss": 0.1188,
      "step": 10740
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.9443365695792882e-05,
      "loss": 0.2033,
      "step": 10750
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.944284789644013e-05,
      "loss": 0.107,
      "step": 10760
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.944233009708738e-05,
      "loss": 0.0967,
      "step": 10770
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.944181229773463e-05,
      "loss": 0.2369,
      "step": 10780
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.944129449838188e-05,
      "loss": 0.259,
      "step": 10790
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.944077669902913e-05,
      "loss": 0.4383,
      "step": 10800
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.9440258899676376e-05,
      "loss": 0.3939,
      "step": 10810
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.9439741100323624e-05,
      "loss": 0.1121,
      "step": 10820
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.9439223300970875e-05,
      "loss": 0.189,
      "step": 10830
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.9438705501618123e-05,
      "loss": 0.0679,
      "step": 10840
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.9438187702265375e-05,
      "loss": 0.3046,
      "step": 10850
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.9437669902912623e-05,
      "loss": 0.2156,
      "step": 10860
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.9437152103559874e-05,
      "loss": 0.2912,
      "step": 10870
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.9436634304207122e-05,
      "loss": 0.2968,
      "step": 10880
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.943611650485437e-05,
      "loss": 0.2222,
      "step": 10890
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.9435598705501618e-05,
      "loss": 0.1531,
      "step": 10900
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.943508090614887e-05,
      "loss": 0.0796,
      "step": 10910
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.9434563106796117e-05,
      "loss": 0.0699,
      "step": 10920
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.9434045307443368e-05,
      "loss": 0.2842,
      "step": 10930
    },
    {
      "epoch": 4.25,
      "learning_rate": 1.9433527508090616e-05,
      "loss": 0.0577,
      "step": 10940
    },
    {
      "epoch": 4.25,
      "learning_rate": 1.9433009708737868e-05,
      "loss": 0.1124,
      "step": 10950
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.9432491909385115e-05,
      "loss": 0.2961,
      "step": 10960
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.9431974110032363e-05,
      "loss": 0.2303,
      "step": 10970
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.943145631067961e-05,
      "loss": 0.1608,
      "step": 10980
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.9430938511326863e-05,
      "loss": 0.1517,
      "step": 10990
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.943042071197411e-05,
      "loss": 0.1787,
      "step": 11000
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.9429902912621362e-05,
      "loss": 0.2766,
      "step": 11010
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.942938511326861e-05,
      "loss": 0.1146,
      "step": 11020
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.942886731391586e-05,
      "loss": 0.1747,
      "step": 11030
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.942834951456311e-05,
      "loss": 0.1637,
      "step": 11040
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.9427831715210357e-05,
      "loss": 0.2721,
      "step": 11050
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.9427313915857605e-05,
      "loss": 0.2411,
      "step": 11060
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.9426796116504856e-05,
      "loss": 0.2553,
      "step": 11070
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.9426278317152104e-05,
      "loss": 0.1392,
      "step": 11080
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.9425760517799355e-05,
      "loss": 0.397,
      "step": 11090
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.9425242718446603e-05,
      "loss": 0.1047,
      "step": 11100
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.9424724919093855e-05,
      "loss": 0.2552,
      "step": 11110
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.94242071197411e-05,
      "loss": 0.2713,
      "step": 11120
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.942368932038835e-05,
      "loss": 0.1791,
      "step": 11130
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.94231715210356e-05,
      "loss": 0.1451,
      "step": 11140
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.942265372168285e-05,
      "loss": 0.0645,
      "step": 11150
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.9422135922330098e-05,
      "loss": 0.3552,
      "step": 11160
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.942161812297735e-05,
      "loss": 0.1695,
      "step": 11170
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.9421100323624597e-05,
      "loss": 0.0469,
      "step": 11180
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.9420582524271848e-05,
      "loss": 0.2501,
      "step": 11190
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.9420064724919093e-05,
      "loss": 0.2035,
      "step": 11200
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.9419546925566344e-05,
      "loss": 0.1851,
      "step": 11210
    },
    {
      "epoch": 4.36,
      "learning_rate": 1.9419029126213592e-05,
      "loss": 0.2064,
      "step": 11220
    },
    {
      "epoch": 4.36,
      "learning_rate": 1.9418511326860843e-05,
      "loss": 0.1586,
      "step": 11230
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.941799352750809e-05,
      "loss": 0.2779,
      "step": 11240
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.9417475728155343e-05,
      "loss": 0.1905,
      "step": 11250
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.941695792880259e-05,
      "loss": 0.3921,
      "step": 11260
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.9416440129449842e-05,
      "loss": 0.3075,
      "step": 11270
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.941592233009709e-05,
      "loss": 0.1345,
      "step": 11280
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.9415404530744338e-05,
      "loss": 0.1818,
      "step": 11290
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.9414886731391586e-05,
      "loss": 0.1156,
      "step": 11300
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.9414368932038837e-05,
      "loss": 0.1625,
      "step": 11310
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.9413851132686085e-05,
      "loss": 0.0837,
      "step": 11320
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.9413333333333336e-05,
      "loss": 0.2448,
      "step": 11330
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.9412815533980584e-05,
      "loss": 0.3734,
      "step": 11340
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.9412297734627835e-05,
      "loss": 0.3253,
      "step": 11350
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.9411779935275083e-05,
      "loss": 0.2675,
      "step": 11360
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.941126213592233e-05,
      "loss": 0.1732,
      "step": 11370
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.941074433656958e-05,
      "loss": 0.1855,
      "step": 11380
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.941022653721683e-05,
      "loss": 0.1948,
      "step": 11390
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.940970873786408e-05,
      "loss": 0.2701,
      "step": 11400
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.940919093851133e-05,
      "loss": 0.1931,
      "step": 11410
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.9408673139158578e-05,
      "loss": 0.2496,
      "step": 11420
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.940815533980583e-05,
      "loss": 0.1995,
      "step": 11430
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.9407637540453077e-05,
      "loss": 0.2041,
      "step": 11440
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.9407119741100325e-05,
      "loss": 0.2528,
      "step": 11450
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.9406601941747573e-05,
      "loss": 0.2433,
      "step": 11460
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.9406084142394824e-05,
      "loss": 0.3492,
      "step": 11470
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.9405566343042072e-05,
      "loss": 0.1016,
      "step": 11480
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.9405048543689323e-05,
      "loss": 0.0442,
      "step": 11490
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.940453074433657e-05,
      "loss": 0.1863,
      "step": 11500
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.9404012944983822e-05,
      "loss": 0.2436,
      "step": 11510
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.940349514563107e-05,
      "loss": 0.3174,
      "step": 11520
    },
    {
      "epoch": 4.48,
      "learning_rate": 1.9402977346278318e-05,
      "loss": 0.229,
      "step": 11530
    },
    {
      "epoch": 4.48,
      "learning_rate": 1.9402459546925566e-05,
      "loss": 0.3688,
      "step": 11540
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.9401941747572818e-05,
      "loss": 0.1861,
      "step": 11550
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.9401423948220065e-05,
      "loss": 0.3283,
      "step": 11560
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.9400906148867317e-05,
      "loss": 0.1399,
      "step": 11570
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.9400388349514565e-05,
      "loss": 0.25,
      "step": 11580
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.9399870550161816e-05,
      "loss": 0.147,
      "step": 11590
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.9399352750809064e-05,
      "loss": 0.3092,
      "step": 11600
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.9398834951456312e-05,
      "loss": 0.4029,
      "step": 11610
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.939831715210356e-05,
      "loss": 0.2057,
      "step": 11620
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.939779935275081e-05,
      "loss": 0.0883,
      "step": 11630
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.939728155339806e-05,
      "loss": 0.256,
      "step": 11640
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.939676375404531e-05,
      "loss": 0.149,
      "step": 11650
    },
    {
      "epoch": 4.53,
      "learning_rate": 1.9396245954692558e-05,
      "loss": 0.1123,
      "step": 11660
    },
    {
      "epoch": 4.53,
      "learning_rate": 1.939572815533981e-05,
      "loss": 0.2805,
      "step": 11670
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.9395210355987057e-05,
      "loss": 0.3653,
      "step": 11680
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.9394692556634305e-05,
      "loss": 0.2283,
      "step": 11690
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.9394174757281553e-05,
      "loss": 0.2955,
      "step": 11700
    },
    {
      "epoch": 4.55,
      "learning_rate": 1.9393656957928805e-05,
      "loss": 0.0699,
      "step": 11710
    },
    {
      "epoch": 4.55,
      "learning_rate": 1.9393139158576053e-05,
      "loss": 0.2213,
      "step": 11720
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.9392621359223304e-05,
      "loss": 0.0633,
      "step": 11730
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.9392103559870552e-05,
      "loss": 0.1132,
      "step": 11740
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.9391585760517803e-05,
      "loss": 0.121,
      "step": 11750
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.939106796116505e-05,
      "loss": 0.1701,
      "step": 11760
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.93905501618123e-05,
      "loss": 0.2263,
      "step": 11770
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.9390032362459547e-05,
      "loss": 0.2276,
      "step": 11780
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.9389514563106798e-05,
      "loss": 0.2396,
      "step": 11790
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.9388996763754046e-05,
      "loss": 0.4364,
      "step": 11800
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.9388478964401297e-05,
      "loss": 0.1599,
      "step": 11810
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.9387961165048545e-05,
      "loss": 0.3382,
      "step": 11820
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.9387443365695797e-05,
      "loss": 0.1922,
      "step": 11830
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.9386925566343045e-05,
      "loss": 0.1832,
      "step": 11840
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.9386407766990292e-05,
      "loss": 0.1679,
      "step": 11850
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.938588996763754e-05,
      "loss": 0.1343,
      "step": 11860
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.9385372168284792e-05,
      "loss": 0.2609,
      "step": 11870
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.938485436893204e-05,
      "loss": 0.1784,
      "step": 11880
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.938433656957929e-05,
      "loss": 0.3379,
      "step": 11890
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.938381877022654e-05,
      "loss": 0.284,
      "step": 11900
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.9383300970873787e-05,
      "loss": 0.2585,
      "step": 11910
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.9382783171521038e-05,
      "loss": 0.2068,
      "step": 11920
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.9382265372168286e-05,
      "loss": 0.0549,
      "step": 11930
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.9381747572815534e-05,
      "loss": 0.1902,
      "step": 11940
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.9381229773462785e-05,
      "loss": 0.1435,
      "step": 11950
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.9380711974110033e-05,
      "loss": 0.0556,
      "step": 11960
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.9380194174757285e-05,
      "loss": 0.1662,
      "step": 11970
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.9379676375404532e-05,
      "loss": 0.1787,
      "step": 11980
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.937915857605178e-05,
      "loss": 0.0364,
      "step": 11990
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.937864077669903e-05,
      "loss": 0.2,
      "step": 12000
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.937812297734628e-05,
      "loss": 0.1333,
      "step": 12010
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.9377605177993528e-05,
      "loss": 0.1067,
      "step": 12020
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.937708737864078e-05,
      "loss": 0.2023,
      "step": 12030
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.9376569579288027e-05,
      "loss": 0.2237,
      "step": 12040
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.9376051779935278e-05,
      "loss": 0.062,
      "step": 12050
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.9375533980582526e-05,
      "loss": 0.0995,
      "step": 12060
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.9375016181229774e-05,
      "loss": 0.0906,
      "step": 12070
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.9374498381877025e-05,
      "loss": 0.2716,
      "step": 12080
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.9373980582524273e-05,
      "loss": 0.3635,
      "step": 12090
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.937346278317152e-05,
      "loss": 0.2923,
      "step": 12100
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.9372944983818772e-05,
      "loss": 0.0635,
      "step": 12110
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.937242718446602e-05,
      "loss": 0.1497,
      "step": 12120
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.937190938511327e-05,
      "loss": 0.1464,
      "step": 12130
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.937139158576052e-05,
      "loss": 0.1473,
      "step": 12140
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.9370873786407767e-05,
      "loss": 0.1652,
      "step": 12150
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.937035598705502e-05,
      "loss": 0.2174,
      "step": 12160
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.9369838187702267e-05,
      "loss": 0.2283,
      "step": 12170
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.9369320388349515e-05,
      "loss": 0.2774,
      "step": 12180
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.9368802588996766e-05,
      "loss": 0.1594,
      "step": 12190
    },
    {
      "epoch": 4.74,
      "learning_rate": 1.9368284789644014e-05,
      "loss": 0.1372,
      "step": 12200
    },
    {
      "epoch": 4.74,
      "learning_rate": 1.9367766990291265e-05,
      "loss": 0.1494,
      "step": 12210
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.9367249190938513e-05,
      "loss": 0.3242,
      "step": 12220
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.936673139158576e-05,
      "loss": 0.2053,
      "step": 12230
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.9366213592233012e-05,
      "loss": 0.1955,
      "step": 12240
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.936569579288026e-05,
      "loss": 0.1725,
      "step": 12250
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.9365177993527508e-05,
      "loss": 0.0831,
      "step": 12260
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.936466019417476e-05,
      "loss": 0.2362,
      "step": 12270
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.9364142394822007e-05,
      "loss": 0.0279,
      "step": 12280
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.9363624595469255e-05,
      "loss": 0.3173,
      "step": 12290
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.9363106796116507e-05,
      "loss": 0.1956,
      "step": 12300
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.9362588996763755e-05,
      "loss": 0.2058,
      "step": 12310
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.9362071197411006e-05,
      "loss": 0.3117,
      "step": 12320
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.9361553398058254e-05,
      "loss": 0.3645,
      "step": 12330
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.9361035598705505e-05,
      "loss": 0.2066,
      "step": 12340
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.9360517799352753e-05,
      "loss": 0.1871,
      "step": 12350
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.936e-05,
      "loss": 0.343,
      "step": 12360
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.935948220064725e-05,
      "loss": 0.1753,
      "step": 12370
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.93589644012945e-05,
      "loss": 0.1749,
      "step": 12380
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.9358446601941748e-05,
      "loss": 0.3164,
      "step": 12390
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.9357928802589e-05,
      "loss": 0.3404,
      "step": 12400
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.9357411003236247e-05,
      "loss": 0.113,
      "step": 12410
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.93568932038835e-05,
      "loss": 0.1865,
      "step": 12420
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.9356375404530747e-05,
      "loss": 0.235,
      "step": 12430
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.9355857605177995e-05,
      "loss": 0.3242,
      "step": 12440
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.9355339805825242e-05,
      "loss": 0.0272,
      "step": 12450
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.9354822006472494e-05,
      "loss": 0.0691,
      "step": 12460
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.9354304207119742e-05,
      "loss": 0.241,
      "step": 12470
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.9353786407766993e-05,
      "loss": 0.1624,
      "step": 12480
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.935326860841424e-05,
      "loss": 0.2917,
      "step": 12490
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.9352750809061492e-05,
      "loss": 0.1886,
      "step": 12500
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.935223300970874e-05,
      "loss": 0.1041,
      "step": 12510
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.9351715210355988e-05,
      "loss": 0.4151,
      "step": 12520
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.9351197411003236e-05,
      "loss": 0.2313,
      "step": 12530
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.9350679611650487e-05,
      "loss": 0.1339,
      "step": 12540
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.9350161812297735e-05,
      "loss": 0.2213,
      "step": 12550
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.9349644012944987e-05,
      "loss": 0.2801,
      "step": 12560
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.9349126213592234e-05,
      "loss": 0.1102,
      "step": 12570
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.9348608414239486e-05,
      "loss": 0.1684,
      "step": 12580
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.934809061488673e-05,
      "loss": 0.079,
      "step": 12590
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.934757281553398e-05,
      "loss": 0.0728,
      "step": 12600
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.934705501618123e-05,
      "loss": 0.329,
      "step": 12610
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.934653721682848e-05,
      "loss": 0.3821,
      "step": 12620
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.934601941747573e-05,
      "loss": 0.3502,
      "step": 12630
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.934550161812298e-05,
      "loss": 0.1426,
      "step": 12640
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.9344983818770228e-05,
      "loss": 0.1978,
      "step": 12650
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.934446601941748e-05,
      "loss": 0.2777,
      "step": 12660
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.9343948220064724e-05,
      "loss": 0.2774,
      "step": 12670
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.9343430420711975e-05,
      "loss": 0.1971,
      "step": 12680
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.9342912621359223e-05,
      "loss": 0.1427,
      "step": 12690
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.9342394822006474e-05,
      "loss": 0.3145,
      "step": 12700
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.9341877022653722e-05,
      "loss": 0.2714,
      "step": 12710
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.9341359223300974e-05,
      "loss": 0.2632,
      "step": 12720
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.934084142394822e-05,
      "loss": 0.2628,
      "step": 12730
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.9340323624595473e-05,
      "loss": 0.1228,
      "step": 12740
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.9339805825242717e-05,
      "loss": 0.1835,
      "step": 12750
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.933928802588997e-05,
      "loss": 0.1612,
      "step": 12760
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.9338770226537217e-05,
      "loss": 0.3655,
      "step": 12770
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.9338252427184468e-05,
      "loss": 0.1295,
      "step": 12780
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.9337734627831716e-05,
      "loss": 0.1606,
      "step": 12790
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.9337216828478967e-05,
      "loss": 0.1875,
      "step": 12800
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.9336699029126215e-05,
      "loss": 0.0929,
      "step": 12810
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.9336181229773466e-05,
      "loss": 0.079,
      "step": 12820
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.933566343042071e-05,
      "loss": 0.1728,
      "step": 12830
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.9335145631067962e-05,
      "loss": 0.2651,
      "step": 12840
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.933462783171521e-05,
      "loss": 0.1756,
      "step": 12850
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.933411003236246e-05,
      "loss": 0.1863,
      "step": 12860
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.933359223300971e-05,
      "loss": 0.1212,
      "step": 12870
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9480055020632737,
      "eval_loss": 0.20791418850421906,
      "eval_runtime": 8.2063,
      "eval_samples_per_second": 442.953,
      "eval_steps_per_second": 55.445,
      "step": 12875
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.933307443365696e-05,
      "loss": 0.1102,
      "step": 12880
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.933255663430421e-05,
      "loss": 0.3159,
      "step": 12890
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.933203883495146e-05,
      "loss": 0.0541,
      "step": 12900
    },
    {
      "epoch": 5.01,
      "learning_rate": 1.9331521035598708e-05,
      "loss": 0.1296,
      "step": 12910
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.9331003236245956e-05,
      "loss": 0.1728,
      "step": 12920
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.9330485436893204e-05,
      "loss": 0.212,
      "step": 12930
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.9329967637540455e-05,
      "loss": 0.2518,
      "step": 12940
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.9329449838187703e-05,
      "loss": 0.2406,
      "step": 12950
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.9328932038834954e-05,
      "loss": 0.0417,
      "step": 12960
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.9328414239482202e-05,
      "loss": 0.2155,
      "step": 12970
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.9327896440129454e-05,
      "loss": 0.2082,
      "step": 12980
    },
    {
      "epoch": 5.04,
      "learning_rate": 1.93273786407767e-05,
      "loss": 0.1845,
      "step": 12990
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.932686084142395e-05,
      "loss": 0.1418,
      "step": 13000
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.9326343042071197e-05,
      "loss": 0.2885,
      "step": 13010
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.932582524271845e-05,
      "loss": 0.228,
      "step": 13020
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.9325307443365697e-05,
      "loss": 0.2872,
      "step": 13030
    },
    {
      "epoch": 5.06,
      "learning_rate": 1.9324789644012948e-05,
      "loss": 0.1394,
      "step": 13040
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.9324271844660196e-05,
      "loss": 0.1646,
      "step": 13050
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.9323754045307447e-05,
      "loss": 0.2454,
      "step": 13060
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.9323236245954695e-05,
      "loss": 0.218,
      "step": 13070
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.9322718446601943e-05,
      "loss": 0.115,
      "step": 13080
    },
    {
      "epoch": 5.08,
      "learning_rate": 1.932220064724919e-05,
      "loss": 0.1885,
      "step": 13090
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.9321682847896442e-05,
      "loss": 0.1807,
      "step": 13100
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.932116504854369e-05,
      "loss": 0.1697,
      "step": 13110
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.932064724919094e-05,
      "loss": 0.1327,
      "step": 13120
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.932012944983819e-05,
      "loss": 0.2862,
      "step": 13130
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.931961165048544e-05,
      "loss": 0.1249,
      "step": 13140
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.931909385113269e-05,
      "loss": 0.2647,
      "step": 13150
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.9318576051779937e-05,
      "loss": 0.3135,
      "step": 13160
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.9318058252427184e-05,
      "loss": 0.1216,
      "step": 13170
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.9317540453074436e-05,
      "loss": 0.2142,
      "step": 13180
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.9317022653721684e-05,
      "loss": 0.2801,
      "step": 13190
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.9316504854368935e-05,
      "loss": 0.2111,
      "step": 13200
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.9315987055016183e-05,
      "loss": 0.1448,
      "step": 13210
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.9315469255663434e-05,
      "loss": 0.2453,
      "step": 13220
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.9314951456310682e-05,
      "loss": 0.2063,
      "step": 13230
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.931443365695793e-05,
      "loss": 0.164,
      "step": 13240
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.9313915857605178e-05,
      "loss": 0.4629,
      "step": 13250
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.931339805825243e-05,
      "loss": 0.1818,
      "step": 13260
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.9312880258899677e-05,
      "loss": 0.2935,
      "step": 13270
    },
    {
      "epoch": 5.16,
      "learning_rate": 1.931236245954693e-05,
      "loss": 0.2923,
      "step": 13280
    },
    {
      "epoch": 5.16,
      "learning_rate": 1.9311844660194177e-05,
      "loss": 0.0491,
      "step": 13290
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.9311326860841428e-05,
      "loss": 0.1612,
      "step": 13300
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.9310809061488676e-05,
      "loss": 0.3004,
      "step": 13310
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.9310291262135924e-05,
      "loss": 0.2165,
      "step": 13320
    },
    {
      "epoch": 5.18,
      "learning_rate": 1.930977346278317e-05,
      "loss": 0.2939,
      "step": 13330
    },
    {
      "epoch": 5.18,
      "learning_rate": 1.9309255663430423e-05,
      "loss": 0.1104,
      "step": 13340
    },
    {
      "epoch": 5.18,
      "learning_rate": 1.930873786407767e-05,
      "loss": 0.1774,
      "step": 13350
    },
    {
      "epoch": 5.19,
      "learning_rate": 1.9308220064724922e-05,
      "loss": 0.333,
      "step": 13360
    },
    {
      "epoch": 5.19,
      "learning_rate": 1.930770226537217e-05,
      "loss": 0.1636,
      "step": 13370
    },
    {
      "epoch": 5.2,
      "learning_rate": 1.9307184466019418e-05,
      "loss": 0.1958,
      "step": 13380
    },
    {
      "epoch": 5.2,
      "learning_rate": 1.930666666666667e-05,
      "loss": 0.2305,
      "step": 13390
    },
    {
      "epoch": 5.2,
      "learning_rate": 1.9306148867313917e-05,
      "loss": 0.1662,
      "step": 13400
    },
    {
      "epoch": 5.21,
      "learning_rate": 1.9305631067961165e-05,
      "loss": 0.0824,
      "step": 13410
    },
    {
      "epoch": 5.21,
      "learning_rate": 1.9305113268608416e-05,
      "loss": 0.193,
      "step": 13420
    },
    {
      "epoch": 5.22,
      "learning_rate": 1.9304595469255664e-05,
      "loss": 0.1063,
      "step": 13430
    },
    {
      "epoch": 5.22,
      "learning_rate": 1.9304077669902916e-05,
      "loss": 0.2268,
      "step": 13440
    },
    {
      "epoch": 5.22,
      "learning_rate": 1.9303559870550164e-05,
      "loss": 0.0798,
      "step": 13450
    },
    {
      "epoch": 5.23,
      "learning_rate": 1.930304207119741e-05,
      "loss": 0.1345,
      "step": 13460
    },
    {
      "epoch": 5.23,
      "learning_rate": 1.9302524271844663e-05,
      "loss": 0.1321,
      "step": 13470
    },
    {
      "epoch": 5.23,
      "learning_rate": 1.930200647249191e-05,
      "loss": 0.2028,
      "step": 13480
    },
    {
      "epoch": 5.24,
      "learning_rate": 1.930148867313916e-05,
      "loss": 0.1012,
      "step": 13490
    },
    {
      "epoch": 5.24,
      "learning_rate": 1.930097087378641e-05,
      "loss": 0.1971,
      "step": 13500
    },
    {
      "epoch": 5.25,
      "learning_rate": 1.9300453074433658e-05,
      "loss": 0.1594,
      "step": 13510
    },
    {
      "epoch": 5.25,
      "learning_rate": 1.929993527508091e-05,
      "loss": 0.2138,
      "step": 13520
    },
    {
      "epoch": 5.25,
      "learning_rate": 1.9299417475728157e-05,
      "loss": 0.2599,
      "step": 13530
    },
    {
      "epoch": 5.26,
      "learning_rate": 1.9298899676375405e-05,
      "loss": 0.2212,
      "step": 13540
    },
    {
      "epoch": 5.26,
      "learning_rate": 1.9298381877022656e-05,
      "loss": 0.2012,
      "step": 13550
    },
    {
      "epoch": 5.27,
      "learning_rate": 1.9297864077669904e-05,
      "loss": 0.1309,
      "step": 13560
    },
    {
      "epoch": 5.27,
      "learning_rate": 1.9297346278317152e-05,
      "loss": 0.1861,
      "step": 13570
    },
    {
      "epoch": 5.27,
      "learning_rate": 1.9296828478964404e-05,
      "loss": 0.0331,
      "step": 13580
    },
    {
      "epoch": 5.28,
      "learning_rate": 1.929631067961165e-05,
      "loss": 0.3318,
      "step": 13590
    },
    {
      "epoch": 5.28,
      "learning_rate": 1.9295792880258903e-05,
      "loss": 0.1876,
      "step": 13600
    },
    {
      "epoch": 5.29,
      "learning_rate": 1.929527508090615e-05,
      "loss": 0.2995,
      "step": 13610
    },
    {
      "epoch": 5.29,
      "learning_rate": 1.92947572815534e-05,
      "loss": 0.2846,
      "step": 13620
    },
    {
      "epoch": 5.29,
      "learning_rate": 1.929423948220065e-05,
      "loss": 0.0782,
      "step": 13630
    },
    {
      "epoch": 5.3,
      "learning_rate": 1.9293721682847898e-05,
      "loss": 0.2103,
      "step": 13640
    },
    {
      "epoch": 5.3,
      "learning_rate": 1.9293203883495146e-05,
      "loss": 0.1982,
      "step": 13650
    },
    {
      "epoch": 5.3,
      "learning_rate": 1.9292686084142397e-05,
      "loss": 0.3472,
      "step": 13660
    },
    {
      "epoch": 5.31,
      "learning_rate": 1.9292168284789645e-05,
      "loss": 0.1286,
      "step": 13670
    },
    {
      "epoch": 5.31,
      "learning_rate": 1.9291650485436896e-05,
      "loss": 0.0243,
      "step": 13680
    },
    {
      "epoch": 5.32,
      "learning_rate": 1.9291132686084144e-05,
      "loss": 0.2426,
      "step": 13690
    },
    {
      "epoch": 5.32,
      "learning_rate": 1.9290614886731392e-05,
      "loss": 0.1475,
      "step": 13700
    },
    {
      "epoch": 5.32,
      "learning_rate": 1.9290097087378644e-05,
      "loss": 0.2549,
      "step": 13710
    },
    {
      "epoch": 5.33,
      "learning_rate": 1.928957928802589e-05,
      "loss": 0.1732,
      "step": 13720
    },
    {
      "epoch": 5.33,
      "learning_rate": 1.928906148867314e-05,
      "loss": 0.0929,
      "step": 13730
    },
    {
      "epoch": 5.34,
      "learning_rate": 1.928854368932039e-05,
      "loss": 0.2855,
      "step": 13740
    },
    {
      "epoch": 5.34,
      "learning_rate": 1.928802588996764e-05,
      "loss": 0.1721,
      "step": 13750
    },
    {
      "epoch": 5.34,
      "learning_rate": 1.9287508090614887e-05,
      "loss": 0.0967,
      "step": 13760
    },
    {
      "epoch": 5.35,
      "learning_rate": 1.9286990291262138e-05,
      "loss": 0.3465,
      "step": 13770
    },
    {
      "epoch": 5.35,
      "learning_rate": 1.9286472491909386e-05,
      "loss": 0.2041,
      "step": 13780
    },
    {
      "epoch": 5.36,
      "learning_rate": 1.9285954692556637e-05,
      "loss": 0.2375,
      "step": 13790
    },
    {
      "epoch": 5.36,
      "learning_rate": 1.9285436893203885e-05,
      "loss": 0.2217,
      "step": 13800
    },
    {
      "epoch": 5.36,
      "learning_rate": 1.9284919093851133e-05,
      "loss": 0.2053,
      "step": 13810
    },
    {
      "epoch": 5.37,
      "learning_rate": 1.9284401294498384e-05,
      "loss": 0.1924,
      "step": 13820
    },
    {
      "epoch": 5.37,
      "learning_rate": 1.9283883495145632e-05,
      "loss": 0.1181,
      "step": 13830
    },
    {
      "epoch": 5.37,
      "learning_rate": 1.928336569579288e-05,
      "loss": 0.1699,
      "step": 13840
    },
    {
      "epoch": 5.38,
      "learning_rate": 1.928284789644013e-05,
      "loss": 0.244,
      "step": 13850
    },
    {
      "epoch": 5.38,
      "learning_rate": 1.928233009708738e-05,
      "loss": 0.4094,
      "step": 13860
    },
    {
      "epoch": 5.39,
      "learning_rate": 1.928181229773463e-05,
      "loss": 0.2045,
      "step": 13870
    },
    {
      "epoch": 5.39,
      "learning_rate": 1.928129449838188e-05,
      "loss": 0.1492,
      "step": 13880
    },
    {
      "epoch": 5.39,
      "learning_rate": 1.9280776699029126e-05,
      "loss": 0.1771,
      "step": 13890
    },
    {
      "epoch": 5.4,
      "learning_rate": 1.9280258899676378e-05,
      "loss": 0.1408,
      "step": 13900
    },
    {
      "epoch": 5.4,
      "learning_rate": 1.9279741100323626e-05,
      "loss": 0.1234,
      "step": 13910
    },
    {
      "epoch": 5.41,
      "learning_rate": 1.9279223300970874e-05,
      "loss": 0.1192,
      "step": 13920
    },
    {
      "epoch": 5.41,
      "learning_rate": 1.9278705501618125e-05,
      "loss": 0.1041,
      "step": 13930
    },
    {
      "epoch": 5.41,
      "learning_rate": 1.9278187702265373e-05,
      "loss": 0.1082,
      "step": 13940
    },
    {
      "epoch": 5.42,
      "learning_rate": 1.9277669902912624e-05,
      "loss": 0.0488,
      "step": 13950
    },
    {
      "epoch": 5.42,
      "learning_rate": 1.9277152103559872e-05,
      "loss": 0.0322,
      "step": 13960
    },
    {
      "epoch": 5.43,
      "learning_rate": 1.927663430420712e-05,
      "loss": 0.1699,
      "step": 13970
    },
    {
      "epoch": 5.43,
      "learning_rate": 1.927611650485437e-05,
      "loss": 0.125,
      "step": 13980
    },
    {
      "epoch": 5.43,
      "learning_rate": 1.927559870550162e-05,
      "loss": 0.3023,
      "step": 13990
    },
    {
      "epoch": 5.44,
      "learning_rate": 1.9275080906148867e-05,
      "loss": 0.1151,
      "step": 14000
    },
    {
      "epoch": 5.44,
      "learning_rate": 1.927456310679612e-05,
      "loss": 0.1746,
      "step": 14010
    },
    {
      "epoch": 5.44,
      "learning_rate": 1.9274045307443366e-05,
      "loss": 0.1369,
      "step": 14020
    },
    {
      "epoch": 5.45,
      "learning_rate": 1.9273527508090618e-05,
      "loss": 0.2252,
      "step": 14030
    },
    {
      "epoch": 5.45,
      "learning_rate": 1.9273009708737866e-05,
      "loss": 0.2103,
      "step": 14040
    },
    {
      "epoch": 5.46,
      "learning_rate": 1.9272491909385117e-05,
      "loss": 0.1524,
      "step": 14050
    },
    {
      "epoch": 5.46,
      "learning_rate": 1.927197411003236e-05,
      "loss": 0.3122,
      "step": 14060
    },
    {
      "epoch": 5.46,
      "learning_rate": 1.9271456310679613e-05,
      "loss": 0.132,
      "step": 14070
    },
    {
      "epoch": 5.47,
      "learning_rate": 1.927093851132686e-05,
      "loss": 0.3348,
      "step": 14080
    },
    {
      "epoch": 5.47,
      "learning_rate": 1.9270420711974112e-05,
      "loss": 0.088,
      "step": 14090
    },
    {
      "epoch": 5.48,
      "learning_rate": 1.926990291262136e-05,
      "loss": 0.1786,
      "step": 14100
    },
    {
      "epoch": 5.48,
      "learning_rate": 1.926938511326861e-05,
      "loss": 0.1415,
      "step": 14110
    },
    {
      "epoch": 5.48,
      "learning_rate": 1.926886731391586e-05,
      "loss": 0.4074,
      "step": 14120
    },
    {
      "epoch": 5.49,
      "learning_rate": 1.926834951456311e-05,
      "loss": 0.1725,
      "step": 14130
    },
    {
      "epoch": 5.49,
      "learning_rate": 1.9267831715210355e-05,
      "loss": 0.0728,
      "step": 14140
    },
    {
      "epoch": 5.5,
      "learning_rate": 1.9267313915857606e-05,
      "loss": 0.0673,
      "step": 14150
    },
    {
      "epoch": 5.5,
      "learning_rate": 1.9266796116504854e-05,
      "loss": 0.2062,
      "step": 14160
    },
    {
      "epoch": 5.5,
      "learning_rate": 1.9266278317152106e-05,
      "loss": 0.117,
      "step": 14170
    },
    {
      "epoch": 5.51,
      "learning_rate": 1.9265760517799354e-05,
      "loss": 0.2056,
      "step": 14180
    },
    {
      "epoch": 5.51,
      "learning_rate": 1.9265242718446605e-05,
      "loss": 0.1338,
      "step": 14190
    },
    {
      "epoch": 5.51,
      "learning_rate": 1.9264724919093853e-05,
      "loss": 0.2366,
      "step": 14200
    },
    {
      "epoch": 5.52,
      "learning_rate": 1.9264207119741104e-05,
      "loss": 0.1643,
      "step": 14210
    },
    {
      "epoch": 5.52,
      "learning_rate": 1.926368932038835e-05,
      "loss": 0.2852,
      "step": 14220
    },
    {
      "epoch": 5.53,
      "learning_rate": 1.92631715210356e-05,
      "loss": 0.3748,
      "step": 14230
    },
    {
      "epoch": 5.53,
      "learning_rate": 1.9262653721682848e-05,
      "loss": 0.3026,
      "step": 14240
    },
    {
      "epoch": 5.53,
      "learning_rate": 1.92621359223301e-05,
      "loss": 0.3083,
      "step": 14250
    },
    {
      "epoch": 5.54,
      "learning_rate": 1.9261618122977347e-05,
      "loss": 0.2295,
      "step": 14260
    },
    {
      "epoch": 5.54,
      "learning_rate": 1.92611003236246e-05,
      "loss": 0.146,
      "step": 14270
    },
    {
      "epoch": 5.55,
      "learning_rate": 1.9260582524271846e-05,
      "loss": 0.395,
      "step": 14280
    },
    {
      "epoch": 5.55,
      "learning_rate": 1.9260064724919098e-05,
      "loss": 0.1961,
      "step": 14290
    },
    {
      "epoch": 5.55,
      "learning_rate": 1.9259546925566342e-05,
      "loss": 0.0909,
      "step": 14300
    },
    {
      "epoch": 5.56,
      "learning_rate": 1.9259029126213593e-05,
      "loss": 0.1856,
      "step": 14310
    },
    {
      "epoch": 5.56,
      "learning_rate": 1.925851132686084e-05,
      "loss": 0.0965,
      "step": 14320
    },
    {
      "epoch": 5.57,
      "learning_rate": 1.9257993527508093e-05,
      "loss": 0.2034,
      "step": 14330
    },
    {
      "epoch": 5.57,
      "learning_rate": 1.925747572815534e-05,
      "loss": 0.2148,
      "step": 14340
    },
    {
      "epoch": 5.57,
      "learning_rate": 1.9256957928802592e-05,
      "loss": 0.0668,
      "step": 14350
    },
    {
      "epoch": 5.58,
      "learning_rate": 1.925644012944984e-05,
      "loss": 0.1503,
      "step": 14360
    },
    {
      "epoch": 5.58,
      "learning_rate": 1.925592233009709e-05,
      "loss": 0.109,
      "step": 14370
    },
    {
      "epoch": 5.58,
      "learning_rate": 1.9255404530744336e-05,
      "loss": 0.1229,
      "step": 14380
    },
    {
      "epoch": 5.59,
      "learning_rate": 1.9254886731391587e-05,
      "loss": 0.2157,
      "step": 14390
    },
    {
      "epoch": 5.59,
      "learning_rate": 1.9254368932038835e-05,
      "loss": 0.1816,
      "step": 14400
    },
    {
      "epoch": 5.6,
      "learning_rate": 1.9253851132686086e-05,
      "loss": 0.1208,
      "step": 14410
    },
    {
      "epoch": 5.6,
      "learning_rate": 1.9253333333333334e-05,
      "loss": 0.0814,
      "step": 14420
    },
    {
      "epoch": 5.6,
      "learning_rate": 1.9252815533980586e-05,
      "loss": 0.131,
      "step": 14430
    },
    {
      "epoch": 5.61,
      "learning_rate": 1.9252297734627833e-05,
      "loss": 0.0829,
      "step": 14440
    },
    {
      "epoch": 5.61,
      "learning_rate": 1.9251779935275085e-05,
      "loss": 0.3475,
      "step": 14450
    },
    {
      "epoch": 5.62,
      "learning_rate": 1.925126213592233e-05,
      "loss": 0.2335,
      "step": 14460
    },
    {
      "epoch": 5.62,
      "learning_rate": 1.925074433656958e-05,
      "loss": 0.2357,
      "step": 14470
    },
    {
      "epoch": 5.62,
      "learning_rate": 1.925022653721683e-05,
      "loss": 0.1109,
      "step": 14480
    },
    {
      "epoch": 5.63,
      "learning_rate": 1.924970873786408e-05,
      "loss": 0.1486,
      "step": 14490
    },
    {
      "epoch": 5.63,
      "learning_rate": 1.9249190938511328e-05,
      "loss": 0.155,
      "step": 14500
    },
    {
      "epoch": 5.63,
      "learning_rate": 1.924867313915858e-05,
      "loss": 0.2239,
      "step": 14510
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.9248155339805827e-05,
      "loss": 0.3423,
      "step": 14520
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.924763754045308e-05,
      "loss": 0.0854,
      "step": 14530
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.9247119741100323e-05,
      "loss": 0.3277,
      "step": 14540
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.9246601941747574e-05,
      "loss": 0.2824,
      "step": 14550
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.9246084142394822e-05,
      "loss": 0.2528,
      "step": 14560
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.9245566343042073e-05,
      "loss": 0.1082,
      "step": 14570
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.924504854368932e-05,
      "loss": 0.2426,
      "step": 14580
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.9244530744336573e-05,
      "loss": 0.2329,
      "step": 14590
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.924401294498382e-05,
      "loss": 0.2598,
      "step": 14600
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.9243495145631072e-05,
      "loss": 0.3209,
      "step": 14610
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.924297734627832e-05,
      "loss": 0.2159,
      "step": 14620
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.9242459546925568e-05,
      "loss": 0.1869,
      "step": 14630
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.9241941747572816e-05,
      "loss": 0.349,
      "step": 14640
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.9241423948220067e-05,
      "loss": 0.0921,
      "step": 14650
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.9240906148867315e-05,
      "loss": 0.2549,
      "step": 14660
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.9240388349514566e-05,
      "loss": 0.3339,
      "step": 14670
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.9239870550161814e-05,
      "loss": 0.3539,
      "step": 14680
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.9239352750809065e-05,
      "loss": 0.1854,
      "step": 14690
    },
    {
      "epoch": 5.71,
      "learning_rate": 1.9238834951456313e-05,
      "loss": 0.1628,
      "step": 14700
    },
    {
      "epoch": 5.71,
      "learning_rate": 1.923831715210356e-05,
      "loss": 0.1387,
      "step": 14710
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.923779935275081e-05,
      "loss": 0.2845,
      "step": 14720
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.923728155339806e-05,
      "loss": 0.148,
      "step": 14730
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.923676375404531e-05,
      "loss": 0.2681,
      "step": 14740
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.923624595469256e-05,
      "loss": 0.1988,
      "step": 14750
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.9235728155339808e-05,
      "loss": 0.1291,
      "step": 14760
    },
    {
      "epoch": 5.74,
      "learning_rate": 1.923521035598706e-05,
      "loss": 0.1396,
      "step": 14770
    },
    {
      "epoch": 5.74,
      "learning_rate": 1.9234692556634307e-05,
      "loss": 0.3225,
      "step": 14780
    },
    {
      "epoch": 5.74,
      "learning_rate": 1.9234174757281555e-05,
      "loss": 0.1079,
      "step": 14790
    },
    {
      "epoch": 5.75,
      "learning_rate": 1.9233656957928803e-05,
      "loss": 0.2939,
      "step": 14800
    },
    {
      "epoch": 5.75,
      "learning_rate": 1.9233139158576054e-05,
      "loss": 0.1918,
      "step": 14810
    },
    {
      "epoch": 5.76,
      "learning_rate": 1.9232621359223302e-05,
      "loss": 0.1802,
      "step": 14820
    },
    {
      "epoch": 5.76,
      "learning_rate": 1.9232103559870553e-05,
      "loss": 0.2616,
      "step": 14830
    },
    {
      "epoch": 5.76,
      "learning_rate": 1.92315857605178e-05,
      "loss": 0.1443,
      "step": 14840
    },
    {
      "epoch": 5.77,
      "learning_rate": 1.923106796116505e-05,
      "loss": 0.1951,
      "step": 14850
    },
    {
      "epoch": 5.77,
      "learning_rate": 1.92305501618123e-05,
      "loss": 0.1339,
      "step": 14860
    },
    {
      "epoch": 5.77,
      "learning_rate": 1.923003236245955e-05,
      "loss": 0.2865,
      "step": 14870
    },
    {
      "epoch": 5.78,
      "learning_rate": 1.9229514563106796e-05,
      "loss": 0.4193,
      "step": 14880
    },
    {
      "epoch": 5.78,
      "learning_rate": 1.9228996763754048e-05,
      "loss": 0.0555,
      "step": 14890
    },
    {
      "epoch": 5.79,
      "learning_rate": 1.9228478964401296e-05,
      "loss": 0.3844,
      "step": 14900
    },
    {
      "epoch": 5.79,
      "learning_rate": 1.9227961165048547e-05,
      "loss": 0.1302,
      "step": 14910
    },
    {
      "epoch": 5.79,
      "learning_rate": 1.9227443365695795e-05,
      "loss": 0.1897,
      "step": 14920
    },
    {
      "epoch": 5.8,
      "learning_rate": 1.9226925566343043e-05,
      "loss": 0.0411,
      "step": 14930
    },
    {
      "epoch": 5.8,
      "learning_rate": 1.9226407766990294e-05,
      "loss": 0.1769,
      "step": 14940
    },
    {
      "epoch": 5.81,
      "learning_rate": 1.9225889967637542e-05,
      "loss": 0.3511,
      "step": 14950
    },
    {
      "epoch": 5.81,
      "learning_rate": 1.922537216828479e-05,
      "loss": 0.2955,
      "step": 14960
    },
    {
      "epoch": 5.81,
      "learning_rate": 1.922485436893204e-05,
      "loss": 0.1421,
      "step": 14970
    },
    {
      "epoch": 5.82,
      "learning_rate": 1.922433656957929e-05,
      "loss": 0.1483,
      "step": 14980
    },
    {
      "epoch": 5.82,
      "learning_rate": 1.922381877022654e-05,
      "loss": 0.1852,
      "step": 14990
    },
    {
      "epoch": 5.83,
      "learning_rate": 1.922330097087379e-05,
      "loss": 0.0685,
      "step": 15000
    },
    {
      "epoch": 5.83,
      "learning_rate": 1.9222783171521036e-05,
      "loss": 0.2201,
      "step": 15010
    },
    {
      "epoch": 5.83,
      "learning_rate": 1.9222265372168288e-05,
      "loss": 0.1539,
      "step": 15020
    },
    {
      "epoch": 5.84,
      "learning_rate": 1.9221747572815536e-05,
      "loss": 0.1496,
      "step": 15030
    },
    {
      "epoch": 5.84,
      "learning_rate": 1.9221229773462783e-05,
      "loss": 0.1749,
      "step": 15040
    },
    {
      "epoch": 5.84,
      "learning_rate": 1.9220711974110035e-05,
      "loss": 0.1988,
      "step": 15050
    },
    {
      "epoch": 5.85,
      "learning_rate": 1.9220194174757283e-05,
      "loss": 0.2328,
      "step": 15060
    },
    {
      "epoch": 5.85,
      "learning_rate": 1.9219676375404534e-05,
      "loss": 0.1989,
      "step": 15070
    },
    {
      "epoch": 5.86,
      "learning_rate": 1.9219158576051782e-05,
      "loss": 0.1949,
      "step": 15080
    },
    {
      "epoch": 5.86,
      "learning_rate": 1.921864077669903e-05,
      "loss": 0.1698,
      "step": 15090
    },
    {
      "epoch": 5.86,
      "learning_rate": 1.921812297734628e-05,
      "loss": 0.1237,
      "step": 15100
    },
    {
      "epoch": 5.87,
      "learning_rate": 1.921760517799353e-05,
      "loss": 0.0926,
      "step": 15110
    },
    {
      "epoch": 5.87,
      "learning_rate": 1.9217087378640777e-05,
      "loss": 0.3122,
      "step": 15120
    },
    {
      "epoch": 5.88,
      "learning_rate": 1.9216569579288028e-05,
      "loss": 0.3145,
      "step": 15130
    },
    {
      "epoch": 5.88,
      "learning_rate": 1.9216051779935276e-05,
      "loss": 0.0698,
      "step": 15140
    },
    {
      "epoch": 5.88,
      "learning_rate": 1.9215533980582528e-05,
      "loss": 0.0747,
      "step": 15150
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.9215016181229775e-05,
      "loss": 0.1761,
      "step": 15160
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.9214498381877023e-05,
      "loss": 0.4981,
      "step": 15170
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.9213980582524275e-05,
      "loss": 0.1832,
      "step": 15180
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.9213462783171523e-05,
      "loss": 0.2347,
      "step": 15190
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.921294498381877e-05,
      "loss": 0.0525,
      "step": 15200
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.9212427184466022e-05,
      "loss": 0.2044,
      "step": 15210
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.921190938511327e-05,
      "loss": 0.2739,
      "step": 15220
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.9211391585760518e-05,
      "loss": 0.0824,
      "step": 15230
    },
    {
      "epoch": 5.92,
      "learning_rate": 1.921087378640777e-05,
      "loss": 0.058,
      "step": 15240
    },
    {
      "epoch": 5.92,
      "learning_rate": 1.9210355987055017e-05,
      "loss": 0.0653,
      "step": 15250
    },
    {
      "epoch": 5.93,
      "learning_rate": 1.9209838187702268e-05,
      "loss": 0.1564,
      "step": 15260
    },
    {
      "epoch": 5.93,
      "learning_rate": 1.9209320388349516e-05,
      "loss": 0.4103,
      "step": 15270
    },
    {
      "epoch": 5.93,
      "learning_rate": 1.9208802588996764e-05,
      "loss": 0.2333,
      "step": 15280
    },
    {
      "epoch": 5.94,
      "learning_rate": 1.9208284789644015e-05,
      "loss": 0.0356,
      "step": 15290
    },
    {
      "epoch": 5.94,
      "learning_rate": 1.9207766990291263e-05,
      "loss": 0.1436,
      "step": 15300
    },
    {
      "epoch": 5.95,
      "learning_rate": 1.920724919093851e-05,
      "loss": 0.3055,
      "step": 15310
    },
    {
      "epoch": 5.95,
      "learning_rate": 1.9206731391585763e-05,
      "loss": 0.1501,
      "step": 15320
    },
    {
      "epoch": 5.95,
      "learning_rate": 1.920621359223301e-05,
      "loss": 0.4844,
      "step": 15330
    },
    {
      "epoch": 5.96,
      "learning_rate": 1.9205695792880262e-05,
      "loss": 0.077,
      "step": 15340
    },
    {
      "epoch": 5.96,
      "learning_rate": 1.920517799352751e-05,
      "loss": 0.1104,
      "step": 15350
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.9204660194174758e-05,
      "loss": 0.0874,
      "step": 15360
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.920414239482201e-05,
      "loss": 0.0952,
      "step": 15370
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.9203624595469257e-05,
      "loss": 0.2246,
      "step": 15380
    },
    {
      "epoch": 5.98,
      "learning_rate": 1.9203106796116505e-05,
      "loss": 0.1307,
      "step": 15390
    },
    {
      "epoch": 5.98,
      "learning_rate": 1.9202588996763756e-05,
      "loss": 0.3018,
      "step": 15400
    },
    {
      "epoch": 5.98,
      "learning_rate": 1.9202071197411004e-05,
      "loss": 0.1101,
      "step": 15410
    },
    {
      "epoch": 5.99,
      "learning_rate": 1.9201553398058255e-05,
      "loss": 0.1423,
      "step": 15420
    },
    {
      "epoch": 5.99,
      "learning_rate": 1.9201035598705503e-05,
      "loss": 0.2311,
      "step": 15430
    },
    {
      "epoch": 6.0,
      "learning_rate": 1.920051779935275e-05,
      "loss": 0.1141,
      "step": 15440
    },
    {
      "epoch": 6.0,
      "learning_rate": 1.9200000000000003e-05,
      "loss": 0.078,
      "step": 15450
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.20075452327728271,
      "eval_runtime": 8.2494,
      "eval_samples_per_second": 440.638,
      "eval_steps_per_second": 55.156,
      "step": 15450
    },
    {
      "epoch": 6.0,
      "learning_rate": 1.919948220064725e-05,
      "loss": 0.3557,
      "step": 15460
    },
    {
      "epoch": 6.01,
      "learning_rate": 1.91989644012945e-05,
      "loss": 0.0501,
      "step": 15470
    },
    {
      "epoch": 6.01,
      "learning_rate": 1.919844660194175e-05,
      "loss": 0.1672,
      "step": 15480
    },
    {
      "epoch": 6.02,
      "learning_rate": 1.9197928802588998e-05,
      "loss": 0.1336,
      "step": 15490
    },
    {
      "epoch": 6.02,
      "learning_rate": 1.919741100323625e-05,
      "loss": 0.2093,
      "step": 15500
    },
    {
      "epoch": 6.02,
      "learning_rate": 1.9196893203883497e-05,
      "loss": 0.1891,
      "step": 15510
    },
    {
      "epoch": 6.03,
      "learning_rate": 1.9196375404530745e-05,
      "loss": 0.2799,
      "step": 15520
    },
    {
      "epoch": 6.03,
      "learning_rate": 1.9195857605177993e-05,
      "loss": 0.1978,
      "step": 15530
    },
    {
      "epoch": 6.03,
      "learning_rate": 1.9195339805825244e-05,
      "loss": 0.2252,
      "step": 15540
    },
    {
      "epoch": 6.04,
      "learning_rate": 1.9194822006472492e-05,
      "loss": 0.1,
      "step": 15550
    },
    {
      "epoch": 6.04,
      "learning_rate": 1.9194304207119743e-05,
      "loss": 0.1596,
      "step": 15560
    },
    {
      "epoch": 6.05,
      "learning_rate": 1.919378640776699e-05,
      "loss": 0.1937,
      "step": 15570
    },
    {
      "epoch": 6.05,
      "learning_rate": 1.9193268608414242e-05,
      "loss": 0.1196,
      "step": 15580
    },
    {
      "epoch": 6.05,
      "learning_rate": 1.919275080906149e-05,
      "loss": 0.1603,
      "step": 15590
    },
    {
      "epoch": 6.06,
      "learning_rate": 1.919223300970874e-05,
      "loss": 0.0718,
      "step": 15600
    },
    {
      "epoch": 6.06,
      "learning_rate": 1.9191715210355986e-05,
      "loss": 0.1527,
      "step": 15610
    },
    {
      "epoch": 6.07,
      "learning_rate": 1.9191197411003238e-05,
      "loss": 0.0953,
      "step": 15620
    },
    {
      "epoch": 6.07,
      "learning_rate": 1.9190679611650485e-05,
      "loss": 0.1299,
      "step": 15630
    },
    {
      "epoch": 6.07,
      "learning_rate": 1.9190161812297737e-05,
      "loss": 0.2235,
      "step": 15640
    },
    {
      "epoch": 6.08,
      "learning_rate": 1.9189644012944985e-05,
      "loss": 0.1687,
      "step": 15650
    },
    {
      "epoch": 6.08,
      "learning_rate": 1.9189126213592236e-05,
      "loss": 0.2346,
      "step": 15660
    },
    {
      "epoch": 6.09,
      "learning_rate": 1.9188608414239484e-05,
      "loss": 0.3069,
      "step": 15670
    },
    {
      "epoch": 6.09,
      "learning_rate": 1.9188090614886732e-05,
      "loss": 0.3043,
      "step": 15680
    },
    {
      "epoch": 6.09,
      "learning_rate": 1.918757281553398e-05,
      "loss": 0.2171,
      "step": 15690
    },
    {
      "epoch": 6.1,
      "learning_rate": 1.918705501618123e-05,
      "loss": 0.1069,
      "step": 15700
    },
    {
      "epoch": 6.1,
      "learning_rate": 1.918653721682848e-05,
      "loss": 0.0284,
      "step": 15710
    },
    {
      "epoch": 6.1,
      "learning_rate": 1.918601941747573e-05,
      "loss": 0.162,
      "step": 15720
    },
    {
      "epoch": 6.11,
      "learning_rate": 1.9185501618122978e-05,
      "loss": 0.2666,
      "step": 15730
    },
    {
      "epoch": 6.11,
      "learning_rate": 1.918498381877023e-05,
      "loss": 0.1653,
      "step": 15740
    },
    {
      "epoch": 6.12,
      "learning_rate": 1.9184466019417478e-05,
      "loss": 0.1882,
      "step": 15750
    },
    {
      "epoch": 6.12,
      "learning_rate": 1.918394822006473e-05,
      "loss": 0.1736,
      "step": 15760
    },
    {
      "epoch": 6.12,
      "learning_rate": 1.9183430420711973e-05,
      "loss": 0.1771,
      "step": 15770
    },
    {
      "epoch": 6.13,
      "learning_rate": 1.9182912621359225e-05,
      "loss": 0.194,
      "step": 15780
    },
    {
      "epoch": 6.13,
      "learning_rate": 1.9182394822006473e-05,
      "loss": 0.1675,
      "step": 15790
    },
    {
      "epoch": 6.14,
      "learning_rate": 1.9181877022653724e-05,
      "loss": 0.135,
      "step": 15800
    },
    {
      "epoch": 6.14,
      "learning_rate": 1.9181359223300972e-05,
      "loss": 0.1382,
      "step": 15810
    },
    {
      "epoch": 6.14,
      "learning_rate": 1.9180841423948223e-05,
      "loss": 0.246,
      "step": 15820
    },
    {
      "epoch": 6.15,
      "learning_rate": 1.918032362459547e-05,
      "loss": 0.119,
      "step": 15830
    },
    {
      "epoch": 6.15,
      "learning_rate": 1.9179805825242722e-05,
      "loss": 0.1854,
      "step": 15840
    },
    {
      "epoch": 6.16,
      "learning_rate": 1.9179288025889967e-05,
      "loss": 0.019,
      "step": 15850
    },
    {
      "epoch": 6.16,
      "learning_rate": 1.9178770226537218e-05,
      "loss": 0.2809,
      "step": 15860
    },
    {
      "epoch": 6.16,
      "learning_rate": 1.9178252427184466e-05,
      "loss": 0.1841,
      "step": 15870
    },
    {
      "epoch": 6.17,
      "learning_rate": 1.9177734627831717e-05,
      "loss": 0.4664,
      "step": 15880
    },
    {
      "epoch": 6.17,
      "learning_rate": 1.9177216828478965e-05,
      "loss": 0.2266,
      "step": 15890
    },
    {
      "epoch": 6.17,
      "learning_rate": 1.9176699029126217e-05,
      "loss": 0.1631,
      "step": 15900
    },
    {
      "epoch": 6.18,
      "learning_rate": 1.9176181229773465e-05,
      "loss": 0.2704,
      "step": 15910
    },
    {
      "epoch": 6.18,
      "learning_rate": 1.9175663430420716e-05,
      "loss": 0.2586,
      "step": 15920
    },
    {
      "epoch": 6.19,
      "learning_rate": 1.917514563106796e-05,
      "loss": 0.0879,
      "step": 15930
    },
    {
      "epoch": 6.19,
      "learning_rate": 1.9174627831715212e-05,
      "loss": 0.2178,
      "step": 15940
    },
    {
      "epoch": 6.19,
      "learning_rate": 1.917411003236246e-05,
      "loss": 0.1129,
      "step": 15950
    },
    {
      "epoch": 6.2,
      "learning_rate": 1.917359223300971e-05,
      "loss": 0.161,
      "step": 15960
    },
    {
      "epoch": 6.2,
      "learning_rate": 1.917307443365696e-05,
      "loss": 0.0851,
      "step": 15970
    },
    {
      "epoch": 6.21,
      "learning_rate": 1.917255663430421e-05,
      "loss": 0.1581,
      "step": 15980
    },
    {
      "epoch": 6.21,
      "learning_rate": 1.9172038834951458e-05,
      "loss": 0.201,
      "step": 15990
    },
    {
      "epoch": 6.21,
      "learning_rate": 1.917152103559871e-05,
      "loss": 0.0281,
      "step": 16000
    },
    {
      "epoch": 6.22,
      "learning_rate": 1.9171003236245954e-05,
      "loss": 0.1721,
      "step": 16010
    },
    {
      "epoch": 6.22,
      "learning_rate": 1.9170485436893205e-05,
      "loss": 0.271,
      "step": 16020
    },
    {
      "epoch": 6.23,
      "learning_rate": 1.9169967637540453e-05,
      "loss": 0.2183,
      "step": 16030
    },
    {
      "epoch": 6.23,
      "learning_rate": 1.9169449838187705e-05,
      "loss": 0.1707,
      "step": 16040
    },
    {
      "epoch": 6.23,
      "learning_rate": 1.9168932038834952e-05,
      "loss": 0.1961,
      "step": 16050
    },
    {
      "epoch": 6.24,
      "learning_rate": 1.9168414239482204e-05,
      "loss": 0.1559,
      "step": 16060
    },
    {
      "epoch": 6.24,
      "learning_rate": 1.9167896440129452e-05,
      "loss": 0.1369,
      "step": 16070
    },
    {
      "epoch": 6.24,
      "learning_rate": 1.9167378640776703e-05,
      "loss": 0.2759,
      "step": 16080
    },
    {
      "epoch": 6.25,
      "learning_rate": 1.9166860841423948e-05,
      "loss": 0.2847,
      "step": 16090
    },
    {
      "epoch": 6.25,
      "learning_rate": 1.91663430420712e-05,
      "loss": 0.1215,
      "step": 16100
    },
    {
      "epoch": 6.26,
      "learning_rate": 1.9165825242718447e-05,
      "loss": 0.1622,
      "step": 16110
    },
    {
      "epoch": 6.26,
      "learning_rate": 1.9165307443365698e-05,
      "loss": 0.196,
      "step": 16120
    },
    {
      "epoch": 6.26,
      "learning_rate": 1.9164789644012946e-05,
      "loss": 0.2124,
      "step": 16130
    },
    {
      "epoch": 6.27,
      "learning_rate": 1.9164271844660197e-05,
      "loss": 0.08,
      "step": 16140
    },
    {
      "epoch": 6.27,
      "learning_rate": 1.9163754045307445e-05,
      "loss": 0.0445,
      "step": 16150
    },
    {
      "epoch": 6.28,
      "learning_rate": 1.9163236245954697e-05,
      "loss": 0.1426,
      "step": 16160
    },
    {
      "epoch": 6.28,
      "learning_rate": 1.916271844660194e-05,
      "loss": 0.1171,
      "step": 16170
    },
    {
      "epoch": 6.28,
      "learning_rate": 1.9162200647249192e-05,
      "loss": 0.174,
      "step": 16180
    },
    {
      "epoch": 6.29,
      "learning_rate": 1.916168284789644e-05,
      "loss": 0.2205,
      "step": 16190
    },
    {
      "epoch": 6.29,
      "learning_rate": 1.916116504854369e-05,
      "loss": 0.2687,
      "step": 16200
    },
    {
      "epoch": 6.3,
      "learning_rate": 1.916064724919094e-05,
      "loss": 0.3148,
      "step": 16210
    },
    {
      "epoch": 6.3,
      "learning_rate": 1.916012944983819e-05,
      "loss": 0.3026,
      "step": 16220
    },
    {
      "epoch": 6.3,
      "learning_rate": 1.915961165048544e-05,
      "loss": 0.234,
      "step": 16230
    },
    {
      "epoch": 6.31,
      "learning_rate": 1.915909385113269e-05,
      "loss": 0.2651,
      "step": 16240
    },
    {
      "epoch": 6.31,
      "learning_rate": 1.9158576051779935e-05,
      "loss": 0.0716,
      "step": 16250
    },
    {
      "epoch": 6.31,
      "learning_rate": 1.9158058252427186e-05,
      "loss": 0.3009,
      "step": 16260
    },
    {
      "epoch": 6.32,
      "learning_rate": 1.9157540453074434e-05,
      "loss": 0.1907,
      "step": 16270
    },
    {
      "epoch": 6.32,
      "learning_rate": 1.9157022653721685e-05,
      "loss": 0.0772,
      "step": 16280
    },
    {
      "epoch": 6.33,
      "learning_rate": 1.9156504854368933e-05,
      "loss": 0.1022,
      "step": 16290
    },
    {
      "epoch": 6.33,
      "learning_rate": 1.9155987055016184e-05,
      "loss": 0.0499,
      "step": 16300
    },
    {
      "epoch": 6.33,
      "learning_rate": 1.9155469255663432e-05,
      "loss": 0.0518,
      "step": 16310
    },
    {
      "epoch": 6.34,
      "learning_rate": 1.915495145631068e-05,
      "loss": 0.1924,
      "step": 16320
    },
    {
      "epoch": 6.34,
      "learning_rate": 1.915443365695793e-05,
      "loss": 0.2682,
      "step": 16330
    },
    {
      "epoch": 6.35,
      "learning_rate": 1.915391585760518e-05,
      "loss": 0.0451,
      "step": 16340
    },
    {
      "epoch": 6.35,
      "learning_rate": 1.9153398058252427e-05,
      "loss": 0.2326,
      "step": 16350
    },
    {
      "epoch": 6.35,
      "learning_rate": 1.915288025889968e-05,
      "loss": 0.2737,
      "step": 16360
    },
    {
      "epoch": 6.36,
      "learning_rate": 1.9152362459546927e-05,
      "loss": 0.3229,
      "step": 16370
    },
    {
      "epoch": 6.36,
      "learning_rate": 1.9151844660194178e-05,
      "loss": 0.3361,
      "step": 16380
    },
    {
      "epoch": 6.37,
      "learning_rate": 1.9151326860841426e-05,
      "loss": 0.2134,
      "step": 16390
    },
    {
      "epoch": 6.37,
      "learning_rate": 1.9150809061488674e-05,
      "loss": 0.0914,
      "step": 16400
    },
    {
      "epoch": 6.37,
      "learning_rate": 1.9150291262135925e-05,
      "loss": 0.0745,
      "step": 16410
    },
    {
      "epoch": 6.38,
      "learning_rate": 1.9149773462783173e-05,
      "loss": 0.4557,
      "step": 16420
    },
    {
      "epoch": 6.38,
      "learning_rate": 1.914925566343042e-05,
      "loss": 0.0096,
      "step": 16430
    },
    {
      "epoch": 6.38,
      "learning_rate": 1.9148737864077672e-05,
      "loss": 0.1587,
      "step": 16440
    },
    {
      "epoch": 6.39,
      "learning_rate": 1.914822006472492e-05,
      "loss": 0.2632,
      "step": 16450
    },
    {
      "epoch": 6.39,
      "learning_rate": 1.914770226537217e-05,
      "loss": 0.1912,
      "step": 16460
    },
    {
      "epoch": 6.4,
      "learning_rate": 1.914718446601942e-05,
      "loss": 0.2155,
      "step": 16470
    },
    {
      "epoch": 6.4,
      "learning_rate": 1.9146666666666667e-05,
      "loss": 0.2165,
      "step": 16480
    },
    {
      "epoch": 6.4,
      "learning_rate": 1.914614886731392e-05,
      "loss": 0.1098,
      "step": 16490
    },
    {
      "epoch": 6.41,
      "learning_rate": 1.9145631067961167e-05,
      "loss": 0.151,
      "step": 16500
    },
    {
      "epoch": 6.41,
      "learning_rate": 1.9145113268608415e-05,
      "loss": 0.1824,
      "step": 16510
    },
    {
      "epoch": 6.42,
      "learning_rate": 1.9144595469255666e-05,
      "loss": 0.1634,
      "step": 16520
    },
    {
      "epoch": 6.42,
      "learning_rate": 1.9144077669902914e-05,
      "loss": 0.2347,
      "step": 16530
    },
    {
      "epoch": 6.42,
      "learning_rate": 1.9143559870550165e-05,
      "loss": 0.2851,
      "step": 16540
    },
    {
      "epoch": 6.43,
      "learning_rate": 1.9143042071197413e-05,
      "loss": 0.1948,
      "step": 16550
    },
    {
      "epoch": 6.43,
      "learning_rate": 1.914252427184466e-05,
      "loss": 0.0837,
      "step": 16560
    },
    {
      "epoch": 6.43,
      "learning_rate": 1.9142006472491912e-05,
      "loss": 0.2453,
      "step": 16570
    },
    {
      "epoch": 6.44,
      "learning_rate": 1.914148867313916e-05,
      "loss": 0.137,
      "step": 16580
    },
    {
      "epoch": 6.44,
      "learning_rate": 1.9140970873786408e-05,
      "loss": 0.1914,
      "step": 16590
    },
    {
      "epoch": 6.45,
      "learning_rate": 1.914045307443366e-05,
      "loss": 0.2187,
      "step": 16600
    },
    {
      "epoch": 6.45,
      "learning_rate": 1.9139935275080907e-05,
      "loss": 0.1347,
      "step": 16610
    },
    {
      "epoch": 6.45,
      "learning_rate": 1.9139417475728155e-05,
      "loss": 0.1845,
      "step": 16620
    },
    {
      "epoch": 6.46,
      "learning_rate": 1.9138899676375407e-05,
      "loss": 0.1783,
      "step": 16630
    },
    {
      "epoch": 6.46,
      "learning_rate": 1.9138381877022655e-05,
      "loss": 0.0791,
      "step": 16640
    },
    {
      "epoch": 6.47,
      "learning_rate": 1.9137864077669906e-05,
      "loss": 0.2269,
      "step": 16650
    },
    {
      "epoch": 6.47,
      "learning_rate": 1.9137346278317154e-05,
      "loss": 0.1276,
      "step": 16660
    },
    {
      "epoch": 6.47,
      "learning_rate": 1.9136828478964402e-05,
      "loss": 0.22,
      "step": 16670
    },
    {
      "epoch": 6.48,
      "learning_rate": 1.9136310679611653e-05,
      "loss": 0.1866,
      "step": 16680
    },
    {
      "epoch": 6.48,
      "learning_rate": 1.91357928802589e-05,
      "loss": 0.1774,
      "step": 16690
    },
    {
      "epoch": 6.49,
      "learning_rate": 1.913527508090615e-05,
      "loss": 0.0928,
      "step": 16700
    },
    {
      "epoch": 6.49,
      "learning_rate": 1.91347572815534e-05,
      "loss": 0.1171,
      "step": 16710
    },
    {
      "epoch": 6.49,
      "learning_rate": 1.9134239482200648e-05,
      "loss": 0.1176,
      "step": 16720
    },
    {
      "epoch": 6.5,
      "learning_rate": 1.91337216828479e-05,
      "loss": 0.1518,
      "step": 16730
    },
    {
      "epoch": 6.5,
      "learning_rate": 1.9133203883495147e-05,
      "loss": 0.1251,
      "step": 16740
    },
    {
      "epoch": 6.5,
      "learning_rate": 1.9132686084142395e-05,
      "loss": 0.328,
      "step": 16750
    },
    {
      "epoch": 6.51,
      "learning_rate": 1.9132168284789647e-05,
      "loss": 0.2399,
      "step": 16760
    },
    {
      "epoch": 6.51,
      "learning_rate": 1.9131650485436895e-05,
      "loss": 0.0783,
      "step": 16770
    },
    {
      "epoch": 6.52,
      "learning_rate": 1.9131132686084142e-05,
      "loss": 0.1425,
      "step": 16780
    },
    {
      "epoch": 6.52,
      "learning_rate": 1.9130614886731394e-05,
      "loss": 0.1957,
      "step": 16790
    },
    {
      "epoch": 6.52,
      "learning_rate": 1.913009708737864e-05,
      "loss": 0.3248,
      "step": 16800
    },
    {
      "epoch": 6.53,
      "learning_rate": 1.9129579288025893e-05,
      "loss": 0.3196,
      "step": 16810
    },
    {
      "epoch": 6.53,
      "learning_rate": 1.912906148867314e-05,
      "loss": 0.4002,
      "step": 16820
    },
    {
      "epoch": 6.54,
      "learning_rate": 1.912854368932039e-05,
      "loss": 0.0971,
      "step": 16830
    },
    {
      "epoch": 6.54,
      "learning_rate": 1.912802588996764e-05,
      "loss": 0.203,
      "step": 16840
    },
    {
      "epoch": 6.54,
      "learning_rate": 1.9127508090614888e-05,
      "loss": 0.0624,
      "step": 16850
    },
    {
      "epoch": 6.55,
      "learning_rate": 1.9126990291262136e-05,
      "loss": 0.2052,
      "step": 16860
    },
    {
      "epoch": 6.55,
      "learning_rate": 1.9126472491909387e-05,
      "loss": 0.0696,
      "step": 16870
    },
    {
      "epoch": 6.56,
      "learning_rate": 1.9125954692556635e-05,
      "loss": 0.1223,
      "step": 16880
    },
    {
      "epoch": 6.56,
      "learning_rate": 1.9125436893203887e-05,
      "loss": 0.0699,
      "step": 16890
    },
    {
      "epoch": 6.56,
      "learning_rate": 1.9124919093851134e-05,
      "loss": 0.1825,
      "step": 16900
    },
    {
      "epoch": 6.57,
      "learning_rate": 1.9124401294498382e-05,
      "loss": 0.0294,
      "step": 16910
    },
    {
      "epoch": 6.57,
      "learning_rate": 1.9123883495145634e-05,
      "loss": 0.0947,
      "step": 16920
    },
    {
      "epoch": 6.57,
      "learning_rate": 1.912336569579288e-05,
      "loss": 0.1121,
      "step": 16930
    },
    {
      "epoch": 6.58,
      "learning_rate": 1.912284789644013e-05,
      "loss": 0.127,
      "step": 16940
    },
    {
      "epoch": 6.58,
      "learning_rate": 1.912233009708738e-05,
      "loss": 0.2293,
      "step": 16950
    },
    {
      "epoch": 6.59,
      "learning_rate": 1.912181229773463e-05,
      "loss": 0.1297,
      "step": 16960
    },
    {
      "epoch": 6.59,
      "learning_rate": 1.912129449838188e-05,
      "loss": 0.122,
      "step": 16970
    },
    {
      "epoch": 6.59,
      "learning_rate": 1.9120776699029128e-05,
      "loss": 0.1205,
      "step": 16980
    },
    {
      "epoch": 6.6,
      "learning_rate": 1.9120258899676376e-05,
      "loss": 0.3043,
      "step": 16990
    },
    {
      "epoch": 6.6,
      "learning_rate": 1.9119741100323624e-05,
      "loss": 0.158,
      "step": 17000
    },
    {
      "epoch": 6.61,
      "learning_rate": 1.9119223300970875e-05,
      "loss": 0.0832,
      "step": 17010
    },
    {
      "epoch": 6.61,
      "learning_rate": 1.9118705501618123e-05,
      "loss": 0.1239,
      "step": 17020
    },
    {
      "epoch": 6.61,
      "learning_rate": 1.9118187702265374e-05,
      "loss": 0.1316,
      "step": 17030
    },
    {
      "epoch": 6.62,
      "learning_rate": 1.9117669902912622e-05,
      "loss": 0.2139,
      "step": 17040
    },
    {
      "epoch": 6.62,
      "learning_rate": 1.9117152103559874e-05,
      "loss": 0.2754,
      "step": 17050
    },
    {
      "epoch": 6.63,
      "learning_rate": 1.911663430420712e-05,
      "loss": 0.1551,
      "step": 17060
    },
    {
      "epoch": 6.63,
      "learning_rate": 1.911611650485437e-05,
      "loss": 0.2108,
      "step": 17070
    },
    {
      "epoch": 6.63,
      "learning_rate": 1.9115598705501617e-05,
      "loss": 0.1862,
      "step": 17080
    },
    {
      "epoch": 6.64,
      "learning_rate": 1.911508090614887e-05,
      "loss": 0.1362,
      "step": 17090
    },
    {
      "epoch": 6.64,
      "learning_rate": 1.9114563106796117e-05,
      "loss": 0.3145,
      "step": 17100
    },
    {
      "epoch": 6.64,
      "learning_rate": 1.9114045307443368e-05,
      "loss": 0.3299,
      "step": 17110
    },
    {
      "epoch": 6.65,
      "learning_rate": 1.9113527508090616e-05,
      "loss": 0.1604,
      "step": 17120
    },
    {
      "epoch": 6.65,
      "learning_rate": 1.9113009708737867e-05,
      "loss": 0.2704,
      "step": 17130
    },
    {
      "epoch": 6.66,
      "learning_rate": 1.9112491909385115e-05,
      "loss": 0.0881,
      "step": 17140
    },
    {
      "epoch": 6.66,
      "learning_rate": 1.9111974110032363e-05,
      "loss": 0.2175,
      "step": 17150
    },
    {
      "epoch": 6.66,
      "learning_rate": 1.911145631067961e-05,
      "loss": 0.1405,
      "step": 17160
    },
    {
      "epoch": 6.67,
      "learning_rate": 1.9110938511326862e-05,
      "loss": 0.0751,
      "step": 17170
    },
    {
      "epoch": 6.67,
      "learning_rate": 1.911042071197411e-05,
      "loss": 0.1503,
      "step": 17180
    },
    {
      "epoch": 6.68,
      "learning_rate": 1.910990291262136e-05,
      "loss": 0.2502,
      "step": 17190
    },
    {
      "epoch": 6.68,
      "learning_rate": 1.910938511326861e-05,
      "loss": 0.2273,
      "step": 17200
    },
    {
      "epoch": 6.68,
      "learning_rate": 1.910886731391586e-05,
      "loss": 0.2392,
      "step": 17210
    },
    {
      "epoch": 6.69,
      "learning_rate": 1.910834951456311e-05,
      "loss": 0.1301,
      "step": 17220
    },
    {
      "epoch": 6.69,
      "learning_rate": 1.9107831715210357e-05,
      "loss": 0.1793,
      "step": 17230
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.9107313915857605e-05,
      "loss": 0.1312,
      "step": 17240
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.9106796116504856e-05,
      "loss": 0.1766,
      "step": 17250
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.9106278317152104e-05,
      "loss": 0.0627,
      "step": 17260
    },
    {
      "epoch": 6.71,
      "learning_rate": 1.9105760517799355e-05,
      "loss": 0.2604,
      "step": 17270
    },
    {
      "epoch": 6.71,
      "learning_rate": 1.9105242718446603e-05,
      "loss": 0.3821,
      "step": 17280
    },
    {
      "epoch": 6.71,
      "learning_rate": 1.9104724919093854e-05,
      "loss": 0.2116,
      "step": 17290
    },
    {
      "epoch": 6.72,
      "learning_rate": 1.9104207119741102e-05,
      "loss": 0.1375,
      "step": 17300
    },
    {
      "epoch": 6.72,
      "learning_rate": 1.910368932038835e-05,
      "loss": 0.347,
      "step": 17310
    },
    {
      "epoch": 6.73,
      "learning_rate": 1.9103171521035598e-05,
      "loss": 0.2109,
      "step": 17320
    },
    {
      "epoch": 6.73,
      "learning_rate": 1.910265372168285e-05,
      "loss": 0.2196,
      "step": 17330
    },
    {
      "epoch": 6.73,
      "learning_rate": 1.9102135922330097e-05,
      "loss": 0.178,
      "step": 17340
    },
    {
      "epoch": 6.74,
      "learning_rate": 1.910161812297735e-05,
      "loss": 0.0664,
      "step": 17350
    },
    {
      "epoch": 6.74,
      "learning_rate": 1.9101100323624597e-05,
      "loss": 0.1796,
      "step": 17360
    },
    {
      "epoch": 6.75,
      "learning_rate": 1.9100582524271848e-05,
      "loss": 0.2695,
      "step": 17370
    },
    {
      "epoch": 6.75,
      "learning_rate": 1.9100064724919096e-05,
      "loss": 0.1135,
      "step": 17380
    },
    {
      "epoch": 6.75,
      "learning_rate": 1.9099546925566347e-05,
      "loss": 0.2008,
      "step": 17390
    },
    {
      "epoch": 6.76,
      "learning_rate": 1.909902912621359e-05,
      "loss": 0.214,
      "step": 17400
    },
    {
      "epoch": 6.76,
      "learning_rate": 1.9098511326860843e-05,
      "loss": 0.103,
      "step": 17410
    },
    {
      "epoch": 6.77,
      "learning_rate": 1.909799352750809e-05,
      "loss": 0.2317,
      "step": 17420
    },
    {
      "epoch": 6.77,
      "learning_rate": 1.9097475728155342e-05,
      "loss": 0.2703,
      "step": 17430
    },
    {
      "epoch": 6.77,
      "learning_rate": 1.909695792880259e-05,
      "loss": 0.2738,
      "step": 17440
    },
    {
      "epoch": 6.78,
      "learning_rate": 1.909644012944984e-05,
      "loss": 0.2464,
      "step": 17450
    },
    {
      "epoch": 6.78,
      "learning_rate": 1.909592233009709e-05,
      "loss": 0.1197,
      "step": 17460
    },
    {
      "epoch": 6.78,
      "learning_rate": 1.909540453074434e-05,
      "loss": 0.0769,
      "step": 17470
    },
    {
      "epoch": 6.79,
      "learning_rate": 1.9094886731391585e-05,
      "loss": 0.1328,
      "step": 17480
    },
    {
      "epoch": 6.79,
      "learning_rate": 1.9094368932038837e-05,
      "loss": 0.1637,
      "step": 17490
    },
    {
      "epoch": 6.8,
      "learning_rate": 1.9093851132686084e-05,
      "loss": 0.1053,
      "step": 17500
    },
    {
      "epoch": 6.8,
      "learning_rate": 1.9093333333333336e-05,
      "loss": 0.1495,
      "step": 17510
    },
    {
      "epoch": 6.8,
      "learning_rate": 1.9092815533980584e-05,
      "loss": 0.3424,
      "step": 17520
    },
    {
      "epoch": 6.81,
      "learning_rate": 1.9092297734627835e-05,
      "loss": 0.157,
      "step": 17530
    },
    {
      "epoch": 6.81,
      "learning_rate": 1.9091779935275083e-05,
      "loss": 0.1431,
      "step": 17540
    },
    {
      "epoch": 6.82,
      "learning_rate": 1.9091262135922334e-05,
      "loss": 0.2201,
      "step": 17550
    },
    {
      "epoch": 6.82,
      "learning_rate": 1.909074433656958e-05,
      "loss": 0.14,
      "step": 17560
    },
    {
      "epoch": 6.82,
      "learning_rate": 1.909022653721683e-05,
      "loss": 0.1362,
      "step": 17570
    },
    {
      "epoch": 6.83,
      "learning_rate": 1.9089708737864078e-05,
      "loss": 0.2197,
      "step": 17580
    },
    {
      "epoch": 6.83,
      "learning_rate": 1.908919093851133e-05,
      "loss": 0.4042,
      "step": 17590
    },
    {
      "epoch": 6.83,
      "learning_rate": 1.9088673139158577e-05,
      "loss": 0.1047,
      "step": 17600
    },
    {
      "epoch": 6.84,
      "learning_rate": 1.908815533980583e-05,
      "loss": 0.2355,
      "step": 17610
    },
    {
      "epoch": 6.84,
      "learning_rate": 1.9087637540453076e-05,
      "loss": 0.1239,
      "step": 17620
    },
    {
      "epoch": 6.85,
      "learning_rate": 1.9087119741100328e-05,
      "loss": 0.236,
      "step": 17630
    },
    {
      "epoch": 6.85,
      "learning_rate": 1.9086601941747572e-05,
      "loss": 0.1671,
      "step": 17640
    },
    {
      "epoch": 6.85,
      "learning_rate": 1.9086084142394824e-05,
      "loss": 0.1873,
      "step": 17650
    },
    {
      "epoch": 6.86,
      "learning_rate": 1.908556634304207e-05,
      "loss": 0.3182,
      "step": 17660
    },
    {
      "epoch": 6.86,
      "learning_rate": 1.9085048543689323e-05,
      "loss": 0.1453,
      "step": 17670
    },
    {
      "epoch": 6.87,
      "learning_rate": 1.908453074433657e-05,
      "loss": 0.145,
      "step": 17680
    },
    {
      "epoch": 6.87,
      "learning_rate": 1.9084012944983822e-05,
      "loss": 0.1821,
      "step": 17690
    },
    {
      "epoch": 6.87,
      "learning_rate": 1.908349514563107e-05,
      "loss": 0.1633,
      "step": 17700
    },
    {
      "epoch": 6.88,
      "learning_rate": 1.908297734627832e-05,
      "loss": 0.0998,
      "step": 17710
    },
    {
      "epoch": 6.88,
      "learning_rate": 1.9082459546925566e-05,
      "loss": 0.0711,
      "step": 17720
    },
    {
      "epoch": 6.89,
      "learning_rate": 1.9081941747572817e-05,
      "loss": 0.2573,
      "step": 17730
    },
    {
      "epoch": 6.89,
      "learning_rate": 1.9081423948220065e-05,
      "loss": 0.1048,
      "step": 17740
    },
    {
      "epoch": 6.89,
      "learning_rate": 1.9080906148867316e-05,
      "loss": 0.1529,
      "step": 17750
    },
    {
      "epoch": 6.9,
      "learning_rate": 1.9080388349514564e-05,
      "loss": 0.1716,
      "step": 17760
    },
    {
      "epoch": 6.9,
      "learning_rate": 1.9079870550161816e-05,
      "loss": 0.1658,
      "step": 17770
    },
    {
      "epoch": 6.9,
      "learning_rate": 1.9079352750809064e-05,
      "loss": 0.2003,
      "step": 17780
    },
    {
      "epoch": 6.91,
      "learning_rate": 1.907883495145631e-05,
      "loss": 0.0673,
      "step": 17790
    },
    {
      "epoch": 6.91,
      "learning_rate": 1.907831715210356e-05,
      "loss": 0.1982,
      "step": 17800
    },
    {
      "epoch": 6.92,
      "learning_rate": 1.907779935275081e-05,
      "loss": 0.1791,
      "step": 17810
    },
    {
      "epoch": 6.92,
      "learning_rate": 1.907728155339806e-05,
      "loss": 0.1167,
      "step": 17820
    },
    {
      "epoch": 6.92,
      "learning_rate": 1.907676375404531e-05,
      "loss": 0.0787,
      "step": 17830
    },
    {
      "epoch": 6.93,
      "learning_rate": 1.9076245954692558e-05,
      "loss": 0.201,
      "step": 17840
    },
    {
      "epoch": 6.93,
      "learning_rate": 1.907572815533981e-05,
      "loss": 0.0963,
      "step": 17850
    },
    {
      "epoch": 6.94,
      "learning_rate": 1.9075210355987057e-05,
      "loss": 0.1458,
      "step": 17860
    },
    {
      "epoch": 6.94,
      "learning_rate": 1.9074692556634305e-05,
      "loss": 0.0864,
      "step": 17870
    },
    {
      "epoch": 6.94,
      "learning_rate": 1.9074174757281553e-05,
      "loss": 0.1299,
      "step": 17880
    },
    {
      "epoch": 6.95,
      "learning_rate": 1.9073656957928804e-05,
      "loss": 0.1835,
      "step": 17890
    },
    {
      "epoch": 6.95,
      "learning_rate": 1.9073139158576052e-05,
      "loss": 0.176,
      "step": 17900
    },
    {
      "epoch": 6.96,
      "learning_rate": 1.9072621359223304e-05,
      "loss": 0.2493,
      "step": 17910
    },
    {
      "epoch": 6.96,
      "learning_rate": 1.907210355987055e-05,
      "loss": 0.0915,
      "step": 17920
    },
    {
      "epoch": 6.96,
      "learning_rate": 1.9071585760517803e-05,
      "loss": 0.1241,
      "step": 17930
    },
    {
      "epoch": 6.97,
      "learning_rate": 1.907106796116505e-05,
      "loss": 0.2423,
      "step": 17940
    },
    {
      "epoch": 6.97,
      "learning_rate": 1.90705501618123e-05,
      "loss": 0.1171,
      "step": 17950
    },
    {
      "epoch": 6.97,
      "learning_rate": 1.907003236245955e-05,
      "loss": 0.2988,
      "step": 17960
    },
    {
      "epoch": 6.98,
      "learning_rate": 1.9069514563106798e-05,
      "loss": 0.1784,
      "step": 17970
    },
    {
      "epoch": 6.98,
      "learning_rate": 1.9068996763754046e-05,
      "loss": 0.142,
      "step": 17980
    },
    {
      "epoch": 6.99,
      "learning_rate": 1.9068478964401297e-05,
      "loss": 0.3226,
      "step": 17990
    },
    {
      "epoch": 6.99,
      "learning_rate": 1.9067961165048545e-05,
      "loss": 0.1368,
      "step": 18000
    },
    {
      "epoch": 6.99,
      "learning_rate": 1.9067443365695796e-05,
      "loss": 0.1914,
      "step": 18010
    },
    {
      "epoch": 7.0,
      "learning_rate": 1.9066925566343044e-05,
      "loss": 0.3493,
      "step": 18020
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.949656121045392,
      "eval_loss": 0.20380502939224243,
      "eval_runtime": 8.1755,
      "eval_samples_per_second": 444.622,
      "eval_steps_per_second": 55.654,
      "step": 18025
    },
    {
      "epoch": 7.0,
      "learning_rate": 1.9066407766990292e-05,
      "loss": 0.2773,
      "step": 18030
    },
    {
      "epoch": 7.01,
      "learning_rate": 1.9065889967637543e-05,
      "loss": 0.1871,
      "step": 18040
    },
    {
      "epoch": 7.01,
      "learning_rate": 1.906537216828479e-05,
      "loss": 0.2832,
      "step": 18050
    },
    {
      "epoch": 7.01,
      "learning_rate": 1.906485436893204e-05,
      "loss": 0.1827,
      "step": 18060
    },
    {
      "epoch": 7.02,
      "learning_rate": 1.906433656957929e-05,
      "loss": 0.3218,
      "step": 18070
    },
    {
      "epoch": 7.02,
      "learning_rate": 1.906381877022654e-05,
      "loss": 0.2045,
      "step": 18080
    },
    {
      "epoch": 7.03,
      "learning_rate": 1.9063300970873786e-05,
      "loss": 0.1801,
      "step": 18090
    },
    {
      "epoch": 7.03,
      "learning_rate": 1.9062783171521038e-05,
      "loss": 0.0777,
      "step": 18100
    },
    {
      "epoch": 7.03,
      "learning_rate": 1.9062265372168286e-05,
      "loss": 0.1323,
      "step": 18110
    },
    {
      "epoch": 7.04,
      "learning_rate": 1.9061747572815537e-05,
      "loss": 0.1679,
      "step": 18120
    },
    {
      "epoch": 7.04,
      "learning_rate": 1.9061229773462785e-05,
      "loss": 0.2358,
      "step": 18130
    },
    {
      "epoch": 7.04,
      "learning_rate": 1.9060711974110033e-05,
      "loss": 0.2176,
      "step": 18140
    },
    {
      "epoch": 7.05,
      "learning_rate": 1.9060194174757284e-05,
      "loss": 0.0727,
      "step": 18150
    },
    {
      "epoch": 7.05,
      "learning_rate": 1.9059676375404532e-05,
      "loss": 0.2949,
      "step": 18160
    },
    {
      "epoch": 7.06,
      "learning_rate": 1.905915857605178e-05,
      "loss": 0.2343,
      "step": 18170
    },
    {
      "epoch": 7.06,
      "learning_rate": 1.905864077669903e-05,
      "loss": 0.2669,
      "step": 18180
    },
    {
      "epoch": 7.06,
      "learning_rate": 1.905812297734628e-05,
      "loss": 0.135,
      "step": 18190
    },
    {
      "epoch": 7.07,
      "learning_rate": 1.905760517799353e-05,
      "loss": 0.0075,
      "step": 18200
    },
    {
      "epoch": 7.07,
      "learning_rate": 1.905708737864078e-05,
      "loss": 0.196,
      "step": 18210
    },
    {
      "epoch": 7.08,
      "learning_rate": 1.9056569579288026e-05,
      "loss": 0.1789,
      "step": 18220
    },
    {
      "epoch": 7.08,
      "learning_rate": 1.9056051779935278e-05,
      "loss": 0.1155,
      "step": 18230
    },
    {
      "epoch": 7.08,
      "learning_rate": 1.9055533980582526e-05,
      "loss": 0.2555,
      "step": 18240
    },
    {
      "epoch": 7.09,
      "learning_rate": 1.9055016181229774e-05,
      "loss": 0.1308,
      "step": 18250
    },
    {
      "epoch": 7.09,
      "learning_rate": 1.9054498381877025e-05,
      "loss": 0.2177,
      "step": 18260
    },
    {
      "epoch": 7.1,
      "learning_rate": 1.9053980582524273e-05,
      "loss": 0.1907,
      "step": 18270
    },
    {
      "epoch": 7.1,
      "learning_rate": 1.9053462783171524e-05,
      "loss": 0.176,
      "step": 18280
    },
    {
      "epoch": 7.1,
      "learning_rate": 1.9052944983818772e-05,
      "loss": 0.1049,
      "step": 18290
    },
    {
      "epoch": 7.11,
      "learning_rate": 1.905242718446602e-05,
      "loss": 0.1525,
      "step": 18300
    },
    {
      "epoch": 7.11,
      "learning_rate": 1.905190938511327e-05,
      "loss": 0.1468,
      "step": 18310
    },
    {
      "epoch": 7.11,
      "learning_rate": 1.905139158576052e-05,
      "loss": 0.0204,
      "step": 18320
    },
    {
      "epoch": 7.12,
      "learning_rate": 1.9050873786407767e-05,
      "loss": 0.3421,
      "step": 18330
    },
    {
      "epoch": 7.12,
      "learning_rate": 1.905035598705502e-05,
      "loss": 0.1184,
      "step": 18340
    },
    {
      "epoch": 7.13,
      "learning_rate": 1.9049838187702266e-05,
      "loss": 0.0951,
      "step": 18350
    },
    {
      "epoch": 7.13,
      "learning_rate": 1.9049320388349518e-05,
      "loss": 0.1476,
      "step": 18360
    },
    {
      "epoch": 7.13,
      "learning_rate": 1.9048802588996766e-05,
      "loss": 0.2102,
      "step": 18370
    },
    {
      "epoch": 7.14,
      "learning_rate": 1.9048284789644014e-05,
      "loss": 0.2523,
      "step": 18380
    },
    {
      "epoch": 7.14,
      "learning_rate": 1.9047766990291265e-05,
      "loss": 0.1768,
      "step": 18390
    },
    {
      "epoch": 7.15,
      "learning_rate": 1.9047249190938513e-05,
      "loss": 0.1273,
      "step": 18400
    },
    {
      "epoch": 7.15,
      "learning_rate": 1.904673139158576e-05,
      "loss": 0.19,
      "step": 18410
    },
    {
      "epoch": 7.15,
      "learning_rate": 1.9046213592233012e-05,
      "loss": 0.2322,
      "step": 18420
    },
    {
      "epoch": 7.16,
      "learning_rate": 1.904569579288026e-05,
      "loss": 0.2602,
      "step": 18430
    },
    {
      "epoch": 7.16,
      "learning_rate": 1.904517799352751e-05,
      "loss": 0.2249,
      "step": 18440
    },
    {
      "epoch": 7.17,
      "learning_rate": 1.904466019417476e-05,
      "loss": 0.1441,
      "step": 18450
    },
    {
      "epoch": 7.17,
      "learning_rate": 1.9044142394822007e-05,
      "loss": 0.3593,
      "step": 18460
    },
    {
      "epoch": 7.17,
      "learning_rate": 1.9043624595469255e-05,
      "loss": 0.2256,
      "step": 18470
    },
    {
      "epoch": 7.18,
      "learning_rate": 1.9043106796116506e-05,
      "loss": 0.0688,
      "step": 18480
    },
    {
      "epoch": 7.18,
      "learning_rate": 1.9042588996763754e-05,
      "loss": 0.2086,
      "step": 18490
    },
    {
      "epoch": 7.18,
      "learning_rate": 1.9042071197411006e-05,
      "loss": 0.1778,
      "step": 18500
    },
    {
      "epoch": 7.19,
      "learning_rate": 1.9041553398058254e-05,
      "loss": 0.1509,
      "step": 18510
    },
    {
      "epoch": 7.19,
      "learning_rate": 1.9041035598705505e-05,
      "loss": 0.1653,
      "step": 18520
    },
    {
      "epoch": 7.2,
      "learning_rate": 1.9040517799352753e-05,
      "loss": 0.3335,
      "step": 18530
    },
    {
      "epoch": 7.2,
      "learning_rate": 1.904e-05,
      "loss": 0.1318,
      "step": 18540
    },
    {
      "epoch": 7.2,
      "learning_rate": 1.903948220064725e-05,
      "loss": 0.1013,
      "step": 18550
    },
    {
      "epoch": 7.21,
      "learning_rate": 1.90389644012945e-05,
      "loss": 0.3407,
      "step": 18560
    },
    {
      "epoch": 7.21,
      "learning_rate": 1.9038446601941748e-05,
      "loss": 0.0784,
      "step": 18570
    },
    {
      "epoch": 7.22,
      "learning_rate": 1.9037928802589e-05,
      "loss": 0.1686,
      "step": 18580
    },
    {
      "epoch": 7.22,
      "learning_rate": 1.9037411003236247e-05,
      "loss": 0.0417,
      "step": 18590
    },
    {
      "epoch": 7.22,
      "learning_rate": 1.90368932038835e-05,
      "loss": 0.0899,
      "step": 18600
    },
    {
      "epoch": 7.23,
      "learning_rate": 1.9036375404530746e-05,
      "loss": 0.0554,
      "step": 18610
    },
    {
      "epoch": 7.23,
      "learning_rate": 1.9035857605177994e-05,
      "loss": 0.0539,
      "step": 18620
    },
    {
      "epoch": 7.23,
      "learning_rate": 1.9035339805825242e-05,
      "loss": 0.1866,
      "step": 18630
    },
    {
      "epoch": 7.24,
      "learning_rate": 1.9034822006472493e-05,
      "loss": 0.1733,
      "step": 18640
    },
    {
      "epoch": 7.24,
      "learning_rate": 1.903430420711974e-05,
      "loss": 0.1772,
      "step": 18650
    },
    {
      "epoch": 7.25,
      "learning_rate": 1.9033786407766993e-05,
      "loss": 0.2179,
      "step": 18660
    },
    {
      "epoch": 7.25,
      "learning_rate": 1.903326860841424e-05,
      "loss": 0.1995,
      "step": 18670
    },
    {
      "epoch": 7.25,
      "learning_rate": 1.9032750809061492e-05,
      "loss": 0.0716,
      "step": 18680
    },
    {
      "epoch": 7.26,
      "learning_rate": 1.903223300970874e-05,
      "loss": 0.1851,
      "step": 18690
    },
    {
      "epoch": 7.26,
      "learning_rate": 1.9031715210355988e-05,
      "loss": 0.3621,
      "step": 18700
    },
    {
      "epoch": 7.27,
      "learning_rate": 1.9031197411003236e-05,
      "loss": 0.3943,
      "step": 18710
    },
    {
      "epoch": 7.27,
      "learning_rate": 1.9030679611650487e-05,
      "loss": 0.1051,
      "step": 18720
    },
    {
      "epoch": 7.27,
      "learning_rate": 1.9030161812297735e-05,
      "loss": 0.0623,
      "step": 18730
    },
    {
      "epoch": 7.28,
      "learning_rate": 1.9029644012944986e-05,
      "loss": 0.1224,
      "step": 18740
    },
    {
      "epoch": 7.28,
      "learning_rate": 1.9029126213592234e-05,
      "loss": 0.1794,
      "step": 18750
    },
    {
      "epoch": 7.29,
      "learning_rate": 1.9028608414239485e-05,
      "loss": 0.3187,
      "step": 18760
    },
    {
      "epoch": 7.29,
      "learning_rate": 1.9028090614886733e-05,
      "loss": 0.1153,
      "step": 18770
    },
    {
      "epoch": 7.29,
      "learning_rate": 1.902757281553398e-05,
      "loss": 0.2892,
      "step": 18780
    },
    {
      "epoch": 7.3,
      "learning_rate": 1.902705501618123e-05,
      "loss": 0.24,
      "step": 18790
    },
    {
      "epoch": 7.3,
      "learning_rate": 1.902653721682848e-05,
      "loss": 0.1071,
      "step": 18800
    },
    {
      "epoch": 7.3,
      "learning_rate": 1.902601941747573e-05,
      "loss": 0.1086,
      "step": 18810
    },
    {
      "epoch": 7.31,
      "learning_rate": 1.902550161812298e-05,
      "loss": 0.118,
      "step": 18820
    },
    {
      "epoch": 7.31,
      "learning_rate": 1.9024983818770228e-05,
      "loss": 0.0782,
      "step": 18830
    },
    {
      "epoch": 7.32,
      "learning_rate": 1.902446601941748e-05,
      "loss": 0.2691,
      "step": 18840
    },
    {
      "epoch": 7.32,
      "learning_rate": 1.9023948220064727e-05,
      "loss": 0.1606,
      "step": 18850
    },
    {
      "epoch": 7.32,
      "learning_rate": 1.9023430420711975e-05,
      "loss": 0.2073,
      "step": 18860
    },
    {
      "epoch": 7.33,
      "learning_rate": 1.9022912621359223e-05,
      "loss": 0.1287,
      "step": 18870
    },
    {
      "epoch": 7.33,
      "learning_rate": 1.9022394822006474e-05,
      "loss": 0.1349,
      "step": 18880
    },
    {
      "epoch": 7.34,
      "learning_rate": 1.9021877022653722e-05,
      "loss": 0.0426,
      "step": 18890
    },
    {
      "epoch": 7.34,
      "learning_rate": 1.9021359223300973e-05,
      "loss": 0.1176,
      "step": 18900
    },
    {
      "epoch": 7.34,
      "learning_rate": 1.902084142394822e-05,
      "loss": 0.0702,
      "step": 18910
    },
    {
      "epoch": 7.35,
      "learning_rate": 1.9020323624595473e-05,
      "loss": 0.129,
      "step": 18920
    },
    {
      "epoch": 7.35,
      "learning_rate": 1.901980582524272e-05,
      "loss": 0.1214,
      "step": 18930
    },
    {
      "epoch": 7.36,
      "learning_rate": 1.901928802588997e-05,
      "loss": 0.1423,
      "step": 18940
    },
    {
      "epoch": 7.36,
      "learning_rate": 1.9018770226537216e-05,
      "loss": 0.1851,
      "step": 18950
    },
    {
      "epoch": 7.36,
      "learning_rate": 1.9018252427184468e-05,
      "loss": 0.1689,
      "step": 18960
    },
    {
      "epoch": 7.37,
      "learning_rate": 1.9017734627831716e-05,
      "loss": 0.1754,
      "step": 18970
    },
    {
      "epoch": 7.37,
      "learning_rate": 1.9017216828478967e-05,
      "loss": 0.2236,
      "step": 18980
    },
    {
      "epoch": 7.37,
      "learning_rate": 1.9016699029126215e-05,
      "loss": 0.08,
      "step": 18990
    },
    {
      "epoch": 7.38,
      "learning_rate": 1.9016181229773466e-05,
      "loss": 0.1958,
      "step": 19000
    },
    {
      "epoch": 7.38,
      "learning_rate": 1.9015663430420714e-05,
      "loss": 0.0383,
      "step": 19010
    },
    {
      "epoch": 7.39,
      "learning_rate": 1.9015145631067962e-05,
      "loss": 0.0816,
      "step": 19020
    },
    {
      "epoch": 7.39,
      "learning_rate": 1.901462783171521e-05,
      "loss": 0.3736,
      "step": 19030
    },
    {
      "epoch": 7.39,
      "learning_rate": 1.901411003236246e-05,
      "loss": 0.1936,
      "step": 19040
    },
    {
      "epoch": 7.4,
      "learning_rate": 1.901359223300971e-05,
      "loss": 0.3916,
      "step": 19050
    },
    {
      "epoch": 7.4,
      "learning_rate": 1.901307443365696e-05,
      "loss": 0.2206,
      "step": 19060
    },
    {
      "epoch": 7.41,
      "learning_rate": 1.901255663430421e-05,
      "loss": 0.1495,
      "step": 19070
    },
    {
      "epoch": 7.41,
      "learning_rate": 1.901203883495146e-05,
      "loss": 0.4515,
      "step": 19080
    },
    {
      "epoch": 7.41,
      "learning_rate": 1.9011521035598708e-05,
      "loss": 0.208,
      "step": 19090
    },
    {
      "epoch": 7.42,
      "learning_rate": 1.901100323624596e-05,
      "loss": 0.1397,
      "step": 19100
    },
    {
      "epoch": 7.42,
      "learning_rate": 1.9010485436893203e-05,
      "loss": 0.3406,
      "step": 19110
    },
    {
      "epoch": 7.43,
      "learning_rate": 1.9009967637540455e-05,
      "loss": 0.2142,
      "step": 19120
    },
    {
      "epoch": 7.43,
      "learning_rate": 1.9009449838187703e-05,
      "loss": 0.1653,
      "step": 19130
    },
    {
      "epoch": 7.43,
      "learning_rate": 1.9008932038834954e-05,
      "loss": 0.2037,
      "step": 19140
    },
    {
      "epoch": 7.44,
      "learning_rate": 1.9008414239482202e-05,
      "loss": 0.2726,
      "step": 19150
    },
    {
      "epoch": 7.44,
      "learning_rate": 1.9007896440129453e-05,
      "loss": 0.2485,
      "step": 19160
    },
    {
      "epoch": 7.44,
      "learning_rate": 1.90073786407767e-05,
      "loss": 0.2741,
      "step": 19170
    },
    {
      "epoch": 7.45,
      "learning_rate": 1.9006860841423953e-05,
      "loss": 0.0828,
      "step": 19180
    },
    {
      "epoch": 7.45,
      "learning_rate": 1.9006343042071197e-05,
      "loss": 0.0898,
      "step": 19190
    },
    {
      "epoch": 7.46,
      "learning_rate": 1.900582524271845e-05,
      "loss": 0.1281,
      "step": 19200
    },
    {
      "epoch": 7.46,
      "learning_rate": 1.9005307443365696e-05,
      "loss": 0.1504,
      "step": 19210
    },
    {
      "epoch": 7.46,
      "learning_rate": 1.9004789644012948e-05,
      "loss": 0.1912,
      "step": 19220
    },
    {
      "epoch": 7.47,
      "learning_rate": 1.9004271844660196e-05,
      "loss": 0.1776,
      "step": 19230
    },
    {
      "epoch": 7.47,
      "learning_rate": 1.9003754045307447e-05,
      "loss": 0.214,
      "step": 19240
    },
    {
      "epoch": 7.48,
      "learning_rate": 1.9003236245954695e-05,
      "loss": 0.2167,
      "step": 19250
    },
    {
      "epoch": 7.48,
      "learning_rate": 1.9002718446601943e-05,
      "loss": 0.3702,
      "step": 19260
    },
    {
      "epoch": 7.48,
      "learning_rate": 1.900220064724919e-05,
      "loss": 0.1665,
      "step": 19270
    },
    {
      "epoch": 7.49,
      "learning_rate": 1.9001682847896442e-05,
      "loss": 0.2016,
      "step": 19280
    },
    {
      "epoch": 7.49,
      "learning_rate": 1.900116504854369e-05,
      "loss": 0.1171,
      "step": 19290
    },
    {
      "epoch": 7.5,
      "learning_rate": 1.900064724919094e-05,
      "loss": 0.0423,
      "step": 19300
    },
    {
      "epoch": 7.5,
      "learning_rate": 1.900012944983819e-05,
      "loss": 0.1928,
      "step": 19310
    },
    {
      "epoch": 7.5,
      "learning_rate": 1.899961165048544e-05,
      "loss": 0.2297,
      "step": 19320
    },
    {
      "epoch": 7.51,
      "learning_rate": 1.8999093851132688e-05,
      "loss": 0.1445,
      "step": 19330
    },
    {
      "epoch": 7.51,
      "learning_rate": 1.8998576051779936e-05,
      "loss": 0.1277,
      "step": 19340
    },
    {
      "epoch": 7.51,
      "learning_rate": 1.8998058252427184e-05,
      "loss": 0.1308,
      "step": 19350
    },
    {
      "epoch": 7.52,
      "learning_rate": 1.8997540453074435e-05,
      "loss": 0.1162,
      "step": 19360
    },
    {
      "epoch": 7.52,
      "learning_rate": 1.8997022653721683e-05,
      "loss": 0.23,
      "step": 19370
    },
    {
      "epoch": 7.53,
      "learning_rate": 1.8996504854368935e-05,
      "loss": 0.1473,
      "step": 19380
    },
    {
      "epoch": 7.53,
      "learning_rate": 1.8995987055016183e-05,
      "loss": 0.2774,
      "step": 19390
    },
    {
      "epoch": 7.53,
      "learning_rate": 1.8995469255663434e-05,
      "loss": 0.1961,
      "step": 19400
    },
    {
      "epoch": 7.54,
      "learning_rate": 1.8994951456310682e-05,
      "loss": 0.1145,
      "step": 19410
    },
    {
      "epoch": 7.54,
      "learning_rate": 1.899443365695793e-05,
      "loss": 0.1871,
      "step": 19420
    },
    {
      "epoch": 7.55,
      "learning_rate": 1.8993915857605178e-05,
      "loss": 0.0781,
      "step": 19430
    },
    {
      "epoch": 7.55,
      "learning_rate": 1.899339805825243e-05,
      "loss": 0.2614,
      "step": 19440
    },
    {
      "epoch": 7.55,
      "learning_rate": 1.8992880258899677e-05,
      "loss": 0.3089,
      "step": 19450
    },
    {
      "epoch": 7.56,
      "learning_rate": 1.8992362459546928e-05,
      "loss": 0.2111,
      "step": 19460
    },
    {
      "epoch": 7.56,
      "learning_rate": 1.8991844660194176e-05,
      "loss": 0.1352,
      "step": 19470
    },
    {
      "epoch": 7.57,
      "learning_rate": 1.8991326860841427e-05,
      "loss": 0.0735,
      "step": 19480
    },
    {
      "epoch": 7.57,
      "learning_rate": 1.8990809061488675e-05,
      "loss": 0.2005,
      "step": 19490
    },
    {
      "epoch": 7.57,
      "learning_rate": 1.8990291262135923e-05,
      "loss": 0.1818,
      "step": 19500
    },
    {
      "epoch": 7.58,
      "learning_rate": 1.898977346278317e-05,
      "loss": 0.1991,
      "step": 19510
    },
    {
      "epoch": 7.58,
      "learning_rate": 1.8989255663430423e-05,
      "loss": 0.1019,
      "step": 19520
    },
    {
      "epoch": 7.58,
      "learning_rate": 1.898873786407767e-05,
      "loss": 0.0849,
      "step": 19530
    },
    {
      "epoch": 7.59,
      "learning_rate": 1.8988220064724922e-05,
      "loss": 0.1094,
      "step": 19540
    },
    {
      "epoch": 7.59,
      "learning_rate": 1.898770226537217e-05,
      "loss": 0.1223,
      "step": 19550
    },
    {
      "epoch": 7.6,
      "learning_rate": 1.8987184466019418e-05,
      "loss": 0.3245,
      "step": 19560
    },
    {
      "epoch": 7.6,
      "learning_rate": 1.898666666666667e-05,
      "loss": 0.1273,
      "step": 19570
    },
    {
      "epoch": 7.6,
      "learning_rate": 1.8986148867313917e-05,
      "loss": 0.1751,
      "step": 19580
    },
    {
      "epoch": 7.61,
      "learning_rate": 1.8985631067961165e-05,
      "loss": 0.1757,
      "step": 19590
    },
    {
      "epoch": 7.61,
      "learning_rate": 1.8985113268608416e-05,
      "loss": 0.1356,
      "step": 19600
    },
    {
      "epoch": 7.62,
      "learning_rate": 1.8984595469255664e-05,
      "loss": 0.1659,
      "step": 19610
    },
    {
      "epoch": 7.62,
      "learning_rate": 1.8984077669902915e-05,
      "loss": 0.1526,
      "step": 19620
    },
    {
      "epoch": 7.62,
      "learning_rate": 1.8983559870550163e-05,
      "loss": 0.175,
      "step": 19630
    },
    {
      "epoch": 7.63,
      "learning_rate": 1.898304207119741e-05,
      "loss": 0.1747,
      "step": 19640
    },
    {
      "epoch": 7.63,
      "learning_rate": 1.8982524271844663e-05,
      "loss": 0.1583,
      "step": 19650
    },
    {
      "epoch": 7.63,
      "learning_rate": 1.898200647249191e-05,
      "loss": 0.1331,
      "step": 19660
    },
    {
      "epoch": 7.64,
      "learning_rate": 1.8981488673139162e-05,
      "loss": 0.2647,
      "step": 19670
    },
    {
      "epoch": 7.64,
      "learning_rate": 1.898097087378641e-05,
      "loss": 0.11,
      "step": 19680
    },
    {
      "epoch": 7.65,
      "learning_rate": 1.8980453074433658e-05,
      "loss": 0.0666,
      "step": 19690
    },
    {
      "epoch": 7.65,
      "learning_rate": 1.897993527508091e-05,
      "loss": 0.1649,
      "step": 19700
    },
    {
      "epoch": 7.65,
      "learning_rate": 1.8979417475728157e-05,
      "loss": 0.1123,
      "step": 19710
    },
    {
      "epoch": 7.66,
      "learning_rate": 1.8978899676375405e-05,
      "loss": 0.15,
      "step": 19720
    },
    {
      "epoch": 7.66,
      "learning_rate": 1.8978381877022656e-05,
      "loss": 0.0799,
      "step": 19730
    },
    {
      "epoch": 7.67,
      "learning_rate": 1.8977864077669904e-05,
      "loss": 0.0805,
      "step": 19740
    },
    {
      "epoch": 7.67,
      "learning_rate": 1.8977346278317155e-05,
      "loss": 0.0692,
      "step": 19750
    },
    {
      "epoch": 7.67,
      "learning_rate": 1.8976828478964403e-05,
      "loss": 0.2302,
      "step": 19760
    },
    {
      "epoch": 7.68,
      "learning_rate": 1.897631067961165e-05,
      "loss": 0.2335,
      "step": 19770
    },
    {
      "epoch": 7.68,
      "learning_rate": 1.8975792880258902e-05,
      "loss": 0.1859,
      "step": 19780
    },
    {
      "epoch": 7.69,
      "learning_rate": 1.897527508090615e-05,
      "loss": 0.1434,
      "step": 19790
    },
    {
      "epoch": 7.69,
      "learning_rate": 1.89747572815534e-05,
      "loss": 0.1904,
      "step": 19800
    },
    {
      "epoch": 7.69,
      "learning_rate": 1.897423948220065e-05,
      "loss": 0.2598,
      "step": 19810
    },
    {
      "epoch": 7.7,
      "learning_rate": 1.8973721682847898e-05,
      "loss": 0.2165,
      "step": 19820
    },
    {
      "epoch": 7.7,
      "learning_rate": 1.897320388349515e-05,
      "loss": 0.2428,
      "step": 19830
    },
    {
      "epoch": 7.7,
      "learning_rate": 1.8972686084142397e-05,
      "loss": 0.1131,
      "step": 19840
    },
    {
      "epoch": 7.71,
      "learning_rate": 1.8972168284789645e-05,
      "loss": 0.182,
      "step": 19850
    },
    {
      "epoch": 7.71,
      "learning_rate": 1.8971650485436896e-05,
      "loss": 0.1989,
      "step": 19860
    },
    {
      "epoch": 7.72,
      "learning_rate": 1.8971132686084144e-05,
      "loss": 0.0783,
      "step": 19870
    },
    {
      "epoch": 7.72,
      "learning_rate": 1.8970614886731392e-05,
      "loss": 0.2297,
      "step": 19880
    },
    {
      "epoch": 7.72,
      "learning_rate": 1.8970097087378643e-05,
      "loss": 0.0564,
      "step": 19890
    },
    {
      "epoch": 7.73,
      "learning_rate": 1.896957928802589e-05,
      "loss": 0.086,
      "step": 19900
    },
    {
      "epoch": 7.73,
      "learning_rate": 1.8969061488673142e-05,
      "loss": 0.2131,
      "step": 19910
    },
    {
      "epoch": 7.74,
      "learning_rate": 1.896854368932039e-05,
      "loss": 0.2196,
      "step": 19920
    },
    {
      "epoch": 7.74,
      "learning_rate": 1.8968025889967638e-05,
      "loss": 0.1693,
      "step": 19930
    },
    {
      "epoch": 7.74,
      "learning_rate": 1.8967508090614886e-05,
      "loss": 0.1422,
      "step": 19940
    },
    {
      "epoch": 7.75,
      "learning_rate": 1.8966990291262138e-05,
      "loss": 0.242,
      "step": 19950
    },
    {
      "epoch": 7.75,
      "learning_rate": 1.8966472491909385e-05,
      "loss": 0.1268,
      "step": 19960
    },
    {
      "epoch": 7.76,
      "learning_rate": 1.8965954692556637e-05,
      "loss": 0.2631,
      "step": 19970
    },
    {
      "epoch": 7.76,
      "learning_rate": 1.8965436893203885e-05,
      "loss": 0.1402,
      "step": 19980
    },
    {
      "epoch": 7.76,
      "learning_rate": 1.8964919093851136e-05,
      "loss": 0.152,
      "step": 19990
    },
    {
      "epoch": 7.77,
      "learning_rate": 1.8964401294498384e-05,
      "loss": 0.3736,
      "step": 20000
    },
    {
      "epoch": 7.77,
      "learning_rate": 1.8963883495145632e-05,
      "loss": 0.1815,
      "step": 20010
    },
    {
      "epoch": 7.77,
      "learning_rate": 1.896336569579288e-05,
      "loss": 0.1719,
      "step": 20020
    },
    {
      "epoch": 7.78,
      "learning_rate": 1.896284789644013e-05,
      "loss": 0.0539,
      "step": 20030
    },
    {
      "epoch": 7.78,
      "learning_rate": 1.896233009708738e-05,
      "loss": 0.1113,
      "step": 20040
    },
    {
      "epoch": 7.79,
      "learning_rate": 1.896181229773463e-05,
      "loss": 0.0947,
      "step": 20050
    },
    {
      "epoch": 7.79,
      "learning_rate": 1.8961294498381878e-05,
      "loss": 0.3626,
      "step": 20060
    },
    {
      "epoch": 7.79,
      "learning_rate": 1.896077669902913e-05,
      "loss": 0.2355,
      "step": 20070
    },
    {
      "epoch": 7.8,
      "learning_rate": 1.8960258899676377e-05,
      "loss": 0.1243,
      "step": 20080
    },
    {
      "epoch": 7.8,
      "learning_rate": 1.8959741100323625e-05,
      "loss": 0.1073,
      "step": 20090
    },
    {
      "epoch": 7.81,
      "learning_rate": 1.8959223300970873e-05,
      "loss": 0.0712,
      "step": 20100
    },
    {
      "epoch": 7.81,
      "learning_rate": 1.8958705501618125e-05,
      "loss": 0.129,
      "step": 20110
    },
    {
      "epoch": 7.81,
      "learning_rate": 1.8958187702265373e-05,
      "loss": 0.334,
      "step": 20120
    },
    {
      "epoch": 7.82,
      "learning_rate": 1.8957669902912624e-05,
      "loss": 0.2766,
      "step": 20130
    },
    {
      "epoch": 7.82,
      "learning_rate": 1.8957152103559872e-05,
      "loss": 0.2735,
      "step": 20140
    },
    {
      "epoch": 7.83,
      "learning_rate": 1.8956634304207123e-05,
      "loss": 0.0949,
      "step": 20150
    },
    {
      "epoch": 7.83,
      "learning_rate": 1.895611650485437e-05,
      "loss": 0.0997,
      "step": 20160
    },
    {
      "epoch": 7.83,
      "learning_rate": 1.895559870550162e-05,
      "loss": 0.164,
      "step": 20170
    },
    {
      "epoch": 7.84,
      "learning_rate": 1.8955080906148867e-05,
      "loss": 0.1456,
      "step": 20180
    },
    {
      "epoch": 7.84,
      "learning_rate": 1.8954563106796118e-05,
      "loss": 0.2678,
      "step": 20190
    },
    {
      "epoch": 7.84,
      "learning_rate": 1.8954045307443366e-05,
      "loss": 0.2769,
      "step": 20200
    },
    {
      "epoch": 7.85,
      "learning_rate": 1.8953527508090617e-05,
      "loss": 0.3176,
      "step": 20210
    },
    {
      "epoch": 7.85,
      "learning_rate": 1.8953009708737865e-05,
      "loss": 0.1447,
      "step": 20220
    },
    {
      "epoch": 7.86,
      "learning_rate": 1.8952491909385117e-05,
      "loss": 0.1337,
      "step": 20230
    },
    {
      "epoch": 7.86,
      "learning_rate": 1.8951974110032365e-05,
      "loss": 0.1835,
      "step": 20240
    },
    {
      "epoch": 7.86,
      "learning_rate": 1.8951456310679613e-05,
      "loss": 0.1446,
      "step": 20250
    },
    {
      "epoch": 7.87,
      "learning_rate": 1.895093851132686e-05,
      "loss": 0.1245,
      "step": 20260
    },
    {
      "epoch": 7.87,
      "learning_rate": 1.8950420711974112e-05,
      "loss": 0.1795,
      "step": 20270
    },
    {
      "epoch": 7.88,
      "learning_rate": 1.894990291262136e-05,
      "loss": 0.1811,
      "step": 20280
    },
    {
      "epoch": 7.88,
      "learning_rate": 1.894938511326861e-05,
      "loss": 0.2446,
      "step": 20290
    },
    {
      "epoch": 7.88,
      "learning_rate": 1.894886731391586e-05,
      "loss": 0.1747,
      "step": 20300
    },
    {
      "epoch": 7.89,
      "learning_rate": 1.894834951456311e-05,
      "loss": 0.1293,
      "step": 20310
    },
    {
      "epoch": 7.89,
      "learning_rate": 1.8947831715210358e-05,
      "loss": 0.18,
      "step": 20320
    },
    {
      "epoch": 7.9,
      "learning_rate": 1.8947313915857606e-05,
      "loss": 0.1283,
      "step": 20330
    },
    {
      "epoch": 7.9,
      "learning_rate": 1.8946796116504854e-05,
      "loss": 0.1031,
      "step": 20340
    },
    {
      "epoch": 7.9,
      "learning_rate": 1.8946278317152105e-05,
      "loss": 0.1859,
      "step": 20350
    },
    {
      "epoch": 7.91,
      "learning_rate": 1.8945760517799353e-05,
      "loss": 0.1025,
      "step": 20360
    },
    {
      "epoch": 7.91,
      "learning_rate": 1.8945242718446605e-05,
      "loss": 0.1444,
      "step": 20370
    },
    {
      "epoch": 7.91,
      "learning_rate": 1.8944724919093852e-05,
      "loss": 0.0487,
      "step": 20380
    },
    {
      "epoch": 7.92,
      "learning_rate": 1.8944207119741104e-05,
      "loss": 0.2183,
      "step": 20390
    },
    {
      "epoch": 7.92,
      "learning_rate": 1.894368932038835e-05,
      "loss": 0.0852,
      "step": 20400
    },
    {
      "epoch": 7.93,
      "learning_rate": 1.89431715210356e-05,
      "loss": 0.1409,
      "step": 20410
    },
    {
      "epoch": 7.93,
      "learning_rate": 1.8942653721682848e-05,
      "loss": 0.092,
      "step": 20420
    },
    {
      "epoch": 7.93,
      "learning_rate": 1.89421359223301e-05,
      "loss": 0.0441,
      "step": 20430
    },
    {
      "epoch": 7.94,
      "learning_rate": 1.8941618122977347e-05,
      "loss": 0.028,
      "step": 20440
    },
    {
      "epoch": 7.94,
      "learning_rate": 1.8941100323624598e-05,
      "loss": 0.2525,
      "step": 20450
    },
    {
      "epoch": 7.95,
      "learning_rate": 1.8940582524271846e-05,
      "loss": 0.1649,
      "step": 20460
    },
    {
      "epoch": 7.95,
      "learning_rate": 1.8940064724919097e-05,
      "loss": 0.1598,
      "step": 20470
    },
    {
      "epoch": 7.95,
      "learning_rate": 1.8939546925566345e-05,
      "loss": 0.3265,
      "step": 20480
    },
    {
      "epoch": 7.96,
      "learning_rate": 1.8939029126213593e-05,
      "loss": 0.0942,
      "step": 20490
    },
    {
      "epoch": 7.96,
      "learning_rate": 1.893851132686084e-05,
      "loss": 0.0392,
      "step": 20500
    },
    {
      "epoch": 7.97,
      "learning_rate": 1.8937993527508092e-05,
      "loss": 0.0193,
      "step": 20510
    },
    {
      "epoch": 7.97,
      "learning_rate": 1.893747572815534e-05,
      "loss": 0.1528,
      "step": 20520
    },
    {
      "epoch": 7.97,
      "learning_rate": 1.893695792880259e-05,
      "loss": 0.2312,
      "step": 20530
    },
    {
      "epoch": 7.98,
      "learning_rate": 1.893644012944984e-05,
      "loss": 0.1399,
      "step": 20540
    },
    {
      "epoch": 7.98,
      "learning_rate": 1.893592233009709e-05,
      "loss": 0.1623,
      "step": 20550
    },
    {
      "epoch": 7.98,
      "learning_rate": 1.893540453074434e-05,
      "loss": 0.2021,
      "step": 20560
    },
    {
      "epoch": 7.99,
      "learning_rate": 1.8934886731391587e-05,
      "loss": 0.0852,
      "step": 20570
    },
    {
      "epoch": 7.99,
      "learning_rate": 1.8934368932038835e-05,
      "loss": 0.2163,
      "step": 20580
    },
    {
      "epoch": 8.0,
      "learning_rate": 1.8933851132686086e-05,
      "loss": 0.1724,
      "step": 20590
    },
    {
      "epoch": 8.0,
      "learning_rate": 1.8933333333333334e-05,
      "loss": 0.131,
      "step": 20600
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.20585355162620544,
      "eval_runtime": 8.2357,
      "eval_samples_per_second": 441.37,
      "eval_steps_per_second": 55.247,
      "step": 20600
    },
    {
      "epoch": 8.0,
      "learning_rate": 1.8932815533980585e-05,
      "loss": 0.1869,
      "step": 20610
    },
    {
      "epoch": 8.01,
      "learning_rate": 1.8932297734627833e-05,
      "loss": 0.0795,
      "step": 20620
    },
    {
      "epoch": 8.01,
      "learning_rate": 1.8931779935275084e-05,
      "loss": 0.2046,
      "step": 20630
    },
    {
      "epoch": 8.02,
      "learning_rate": 1.8931262135922332e-05,
      "loss": 0.0538,
      "step": 20640
    },
    {
      "epoch": 8.02,
      "learning_rate": 1.893074433656958e-05,
      "loss": 0.0775,
      "step": 20650
    },
    {
      "epoch": 8.02,
      "learning_rate": 1.8930226537216828e-05,
      "loss": 0.0677,
      "step": 20660
    },
    {
      "epoch": 8.03,
      "learning_rate": 1.892970873786408e-05,
      "loss": 0.3207,
      "step": 20670
    },
    {
      "epoch": 8.03,
      "learning_rate": 1.8929190938511327e-05,
      "loss": 0.1489,
      "step": 20680
    },
    {
      "epoch": 8.03,
      "learning_rate": 1.892867313915858e-05,
      "loss": 0.0776,
      "step": 20690
    },
    {
      "epoch": 8.04,
      "learning_rate": 1.8928155339805827e-05,
      "loss": 0.2284,
      "step": 20700
    },
    {
      "epoch": 8.04,
      "learning_rate": 1.8927637540453078e-05,
      "loss": 0.1624,
      "step": 20710
    },
    {
      "epoch": 8.05,
      "learning_rate": 1.8927119741100326e-05,
      "loss": 0.1928,
      "step": 20720
    },
    {
      "epoch": 8.05,
      "learning_rate": 1.8926601941747574e-05,
      "loss": 0.0365,
      "step": 20730
    },
    {
      "epoch": 8.05,
      "learning_rate": 1.8926084142394822e-05,
      "loss": 0.1466,
      "step": 20740
    },
    {
      "epoch": 8.06,
      "learning_rate": 1.8925566343042073e-05,
      "loss": 0.075,
      "step": 20750
    },
    {
      "epoch": 8.06,
      "learning_rate": 1.892504854368932e-05,
      "loss": 0.3438,
      "step": 20760
    },
    {
      "epoch": 8.07,
      "learning_rate": 1.8924530744336572e-05,
      "loss": 0.1094,
      "step": 20770
    },
    {
      "epoch": 8.07,
      "learning_rate": 1.892401294498382e-05,
      "loss": 0.3573,
      "step": 20780
    },
    {
      "epoch": 8.07,
      "learning_rate": 1.892349514563107e-05,
      "loss": 0.138,
      "step": 20790
    },
    {
      "epoch": 8.08,
      "learning_rate": 1.892297734627832e-05,
      "loss": 0.1114,
      "step": 20800
    },
    {
      "epoch": 8.08,
      "learning_rate": 1.8922459546925567e-05,
      "loss": 0.1337,
      "step": 20810
    },
    {
      "epoch": 8.09,
      "learning_rate": 1.8921941747572815e-05,
      "loss": 0.2441,
      "step": 20820
    },
    {
      "epoch": 8.09,
      "learning_rate": 1.8921423948220067e-05,
      "loss": 0.1587,
      "step": 20830
    },
    {
      "epoch": 8.09,
      "learning_rate": 1.8920906148867315e-05,
      "loss": 0.1904,
      "step": 20840
    },
    {
      "epoch": 8.1,
      "learning_rate": 1.8920388349514566e-05,
      "loss": 0.0232,
      "step": 20850
    },
    {
      "epoch": 8.1,
      "learning_rate": 1.8919870550161814e-05,
      "loss": 0.1243,
      "step": 20860
    },
    {
      "epoch": 8.1,
      "learning_rate": 1.8919352750809065e-05,
      "loss": 0.2785,
      "step": 20870
    },
    {
      "epoch": 8.11,
      "learning_rate": 1.8918834951456313e-05,
      "loss": 0.1649,
      "step": 20880
    },
    {
      "epoch": 8.11,
      "learning_rate": 1.891831715210356e-05,
      "loss": 0.0937,
      "step": 20890
    },
    {
      "epoch": 8.12,
      "learning_rate": 1.891779935275081e-05,
      "loss": 0.1416,
      "step": 20900
    },
    {
      "epoch": 8.12,
      "learning_rate": 1.891728155339806e-05,
      "loss": 0.0647,
      "step": 20910
    },
    {
      "epoch": 8.12,
      "learning_rate": 1.8916763754045308e-05,
      "loss": 0.3121,
      "step": 20920
    },
    {
      "epoch": 8.13,
      "learning_rate": 1.891624595469256e-05,
      "loss": 0.092,
      "step": 20930
    },
    {
      "epoch": 8.13,
      "learning_rate": 1.8915728155339807e-05,
      "loss": 0.0704,
      "step": 20940
    },
    {
      "epoch": 8.14,
      "learning_rate": 1.891521035598706e-05,
      "loss": 0.1044,
      "step": 20950
    },
    {
      "epoch": 8.14,
      "learning_rate": 1.8914692556634307e-05,
      "loss": 0.1408,
      "step": 20960
    },
    {
      "epoch": 8.14,
      "learning_rate": 1.8914174757281555e-05,
      "loss": 0.2793,
      "step": 20970
    },
    {
      "epoch": 8.15,
      "learning_rate": 1.8913656957928802e-05,
      "loss": 0.1684,
      "step": 20980
    },
    {
      "epoch": 8.15,
      "learning_rate": 1.8913139158576054e-05,
      "loss": 0.0694,
      "step": 20990
    },
    {
      "epoch": 8.16,
      "learning_rate": 1.89126213592233e-05,
      "loss": 0.117,
      "step": 21000
    },
    {
      "epoch": 8.16,
      "learning_rate": 1.8912103559870553e-05,
      "loss": 0.2127,
      "step": 21010
    },
    {
      "epoch": 8.16,
      "learning_rate": 1.89115857605178e-05,
      "loss": 0.1166,
      "step": 21020
    },
    {
      "epoch": 8.17,
      "learning_rate": 1.891106796116505e-05,
      "loss": 0.1383,
      "step": 21030
    },
    {
      "epoch": 8.17,
      "learning_rate": 1.89105501618123e-05,
      "loss": 0.1603,
      "step": 21040
    },
    {
      "epoch": 8.17,
      "learning_rate": 1.8910032362459548e-05,
      "loss": 0.2259,
      "step": 21050
    },
    {
      "epoch": 8.18,
      "learning_rate": 1.8909514563106796e-05,
      "loss": 0.2835,
      "step": 21060
    },
    {
      "epoch": 8.18,
      "learning_rate": 1.8908996763754047e-05,
      "loss": 0.1208,
      "step": 21070
    },
    {
      "epoch": 8.19,
      "learning_rate": 1.8908478964401295e-05,
      "loss": 0.0486,
      "step": 21080
    },
    {
      "epoch": 8.19,
      "learning_rate": 1.8907961165048547e-05,
      "loss": 0.0696,
      "step": 21090
    },
    {
      "epoch": 8.19,
      "learning_rate": 1.8907443365695794e-05,
      "loss": 0.2886,
      "step": 21100
    },
    {
      "epoch": 8.2,
      "learning_rate": 1.8906925566343042e-05,
      "loss": 0.2904,
      "step": 21110
    },
    {
      "epoch": 8.2,
      "learning_rate": 1.8906407766990294e-05,
      "loss": 0.1555,
      "step": 21120
    },
    {
      "epoch": 8.21,
      "learning_rate": 1.890588996763754e-05,
      "loss": 0.1413,
      "step": 21130
    },
    {
      "epoch": 8.21,
      "learning_rate": 1.890537216828479e-05,
      "loss": 0.1657,
      "step": 21140
    },
    {
      "epoch": 8.21,
      "learning_rate": 1.890485436893204e-05,
      "loss": 0.0657,
      "step": 21150
    },
    {
      "epoch": 8.22,
      "learning_rate": 1.890433656957929e-05,
      "loss": 0.2767,
      "step": 21160
    },
    {
      "epoch": 8.22,
      "learning_rate": 1.890381877022654e-05,
      "loss": 0.0921,
      "step": 21170
    },
    {
      "epoch": 8.23,
      "learning_rate": 1.8903300970873788e-05,
      "loss": 0.1288,
      "step": 21180
    },
    {
      "epoch": 8.23,
      "learning_rate": 1.8902783171521036e-05,
      "loss": 0.3064,
      "step": 21190
    },
    {
      "epoch": 8.23,
      "learning_rate": 1.8902265372168287e-05,
      "loss": 0.2353,
      "step": 21200
    },
    {
      "epoch": 8.24,
      "learning_rate": 1.8901747572815535e-05,
      "loss": 0.2186,
      "step": 21210
    },
    {
      "epoch": 8.24,
      "learning_rate": 1.8901229773462783e-05,
      "loss": 0.2613,
      "step": 21220
    },
    {
      "epoch": 8.24,
      "learning_rate": 1.8900711974110034e-05,
      "loss": 0.2139,
      "step": 21230
    },
    {
      "epoch": 8.25,
      "learning_rate": 1.8900194174757282e-05,
      "loss": 0.0371,
      "step": 21240
    },
    {
      "epoch": 8.25,
      "learning_rate": 1.8899676375404534e-05,
      "loss": 0.4219,
      "step": 21250
    },
    {
      "epoch": 8.26,
      "learning_rate": 1.889915857605178e-05,
      "loss": 0.2926,
      "step": 21260
    },
    {
      "epoch": 8.26,
      "learning_rate": 1.889864077669903e-05,
      "loss": 0.0565,
      "step": 21270
    },
    {
      "epoch": 8.26,
      "learning_rate": 1.889812297734628e-05,
      "loss": 0.1745,
      "step": 21280
    },
    {
      "epoch": 8.27,
      "learning_rate": 1.889760517799353e-05,
      "loss": 0.1794,
      "step": 21290
    },
    {
      "epoch": 8.27,
      "learning_rate": 1.8897087378640777e-05,
      "loss": 0.1272,
      "step": 21300
    },
    {
      "epoch": 8.28,
      "learning_rate": 1.8896569579288028e-05,
      "loss": 0.0818,
      "step": 21310
    },
    {
      "epoch": 8.28,
      "learning_rate": 1.8896051779935276e-05,
      "loss": 0.11,
      "step": 21320
    },
    {
      "epoch": 8.28,
      "learning_rate": 1.8895533980582527e-05,
      "loss": 0.0895,
      "step": 21330
    },
    {
      "epoch": 8.29,
      "learning_rate": 1.8895016181229775e-05,
      "loss": 0.0777,
      "step": 21340
    },
    {
      "epoch": 8.29,
      "learning_rate": 1.8894498381877023e-05,
      "loss": 0.0032,
      "step": 21350
    },
    {
      "epoch": 8.3,
      "learning_rate": 1.8893980582524274e-05,
      "loss": 0.1753,
      "step": 21360
    },
    {
      "epoch": 8.3,
      "learning_rate": 1.8893462783171522e-05,
      "loss": 0.0212,
      "step": 21370
    },
    {
      "epoch": 8.3,
      "learning_rate": 1.8892944983818774e-05,
      "loss": 0.2488,
      "step": 21380
    },
    {
      "epoch": 8.31,
      "learning_rate": 1.889242718446602e-05,
      "loss": 0.1281,
      "step": 21390
    },
    {
      "epoch": 8.31,
      "learning_rate": 1.889190938511327e-05,
      "loss": 0.1374,
      "step": 21400
    },
    {
      "epoch": 8.31,
      "learning_rate": 1.8891391585760517e-05,
      "loss": 0.2153,
      "step": 21410
    },
    {
      "epoch": 8.32,
      "learning_rate": 1.889087378640777e-05,
      "loss": 0.3827,
      "step": 21420
    },
    {
      "epoch": 8.32,
      "learning_rate": 1.8890355987055017e-05,
      "loss": 0.232,
      "step": 21430
    },
    {
      "epoch": 8.33,
      "learning_rate": 1.8889838187702268e-05,
      "loss": 0.138,
      "step": 21440
    },
    {
      "epoch": 8.33,
      "learning_rate": 1.8889320388349516e-05,
      "loss": 0.3031,
      "step": 21450
    },
    {
      "epoch": 8.33,
      "learning_rate": 1.8888802588996767e-05,
      "loss": 0.3781,
      "step": 21460
    },
    {
      "epoch": 8.34,
      "learning_rate": 1.8888284789644015e-05,
      "loss": 0.0534,
      "step": 21470
    },
    {
      "epoch": 8.34,
      "learning_rate": 1.8887766990291263e-05,
      "loss": 0.1801,
      "step": 21480
    },
    {
      "epoch": 8.35,
      "learning_rate": 1.888724919093851e-05,
      "loss": 0.0915,
      "step": 21490
    },
    {
      "epoch": 8.35,
      "learning_rate": 1.8886731391585762e-05,
      "loss": 0.0893,
      "step": 21500
    },
    {
      "epoch": 8.35,
      "learning_rate": 1.888621359223301e-05,
      "loss": 0.164,
      "step": 21510
    },
    {
      "epoch": 8.36,
      "learning_rate": 1.888569579288026e-05,
      "loss": 0.2233,
      "step": 21520
    },
    {
      "epoch": 8.36,
      "learning_rate": 1.888517799352751e-05,
      "loss": 0.1314,
      "step": 21530
    },
    {
      "epoch": 8.37,
      "learning_rate": 1.888466019417476e-05,
      "loss": 0.0655,
      "step": 21540
    },
    {
      "epoch": 8.37,
      "learning_rate": 1.888414239482201e-05,
      "loss": 0.1256,
      "step": 21550
    },
    {
      "epoch": 8.37,
      "learning_rate": 1.8883624595469257e-05,
      "loss": 0.2023,
      "step": 21560
    },
    {
      "epoch": 8.38,
      "learning_rate": 1.8883106796116504e-05,
      "loss": 0.1501,
      "step": 21570
    },
    {
      "epoch": 8.38,
      "learning_rate": 1.8882588996763756e-05,
      "loss": 0.0956,
      "step": 21580
    },
    {
      "epoch": 8.38,
      "learning_rate": 1.8882071197411004e-05,
      "loss": 0.0981,
      "step": 21590
    },
    {
      "epoch": 8.39,
      "learning_rate": 1.8881553398058255e-05,
      "loss": 0.0134,
      "step": 21600
    },
    {
      "epoch": 8.39,
      "learning_rate": 1.8881035598705503e-05,
      "loss": 0.2811,
      "step": 21610
    },
    {
      "epoch": 8.4,
      "learning_rate": 1.8880517799352754e-05,
      "loss": 0.1102,
      "step": 21620
    },
    {
      "epoch": 8.4,
      "learning_rate": 1.8880000000000002e-05,
      "loss": 0.0439,
      "step": 21630
    },
    {
      "epoch": 8.4,
      "learning_rate": 1.887948220064725e-05,
      "loss": 0.2291,
      "step": 21640
    },
    {
      "epoch": 8.41,
      "learning_rate": 1.8878964401294498e-05,
      "loss": 0.1487,
      "step": 21650
    },
    {
      "epoch": 8.41,
      "learning_rate": 1.887844660194175e-05,
      "loss": 0.2221,
      "step": 21660
    },
    {
      "epoch": 8.42,
      "learning_rate": 1.8877928802588997e-05,
      "loss": 0.1701,
      "step": 21670
    },
    {
      "epoch": 8.42,
      "learning_rate": 1.887741100323625e-05,
      "loss": 0.1539,
      "step": 21680
    },
    {
      "epoch": 8.42,
      "learning_rate": 1.8876893203883497e-05,
      "loss": 0.2316,
      "step": 21690
    },
    {
      "epoch": 8.43,
      "learning_rate": 1.8876375404530748e-05,
      "loss": 0.3173,
      "step": 21700
    },
    {
      "epoch": 8.43,
      "learning_rate": 1.8875857605177992e-05,
      "loss": 0.1826,
      "step": 21710
    },
    {
      "epoch": 8.43,
      "learning_rate": 1.8875339805825244e-05,
      "loss": 0.1129,
      "step": 21720
    },
    {
      "epoch": 8.44,
      "learning_rate": 1.887482200647249e-05,
      "loss": 0.2427,
      "step": 21730
    },
    {
      "epoch": 8.44,
      "learning_rate": 1.8874304207119743e-05,
      "loss": 0.1366,
      "step": 21740
    },
    {
      "epoch": 8.45,
      "learning_rate": 1.887378640776699e-05,
      "loss": 0.0836,
      "step": 21750
    },
    {
      "epoch": 8.45,
      "learning_rate": 1.8873268608414242e-05,
      "loss": 0.1552,
      "step": 21760
    },
    {
      "epoch": 8.45,
      "learning_rate": 1.887275080906149e-05,
      "loss": 0.1704,
      "step": 21770
    },
    {
      "epoch": 8.46,
      "learning_rate": 1.887223300970874e-05,
      "loss": 0.1901,
      "step": 21780
    },
    {
      "epoch": 8.46,
      "learning_rate": 1.8871715210355986e-05,
      "loss": 0.1486,
      "step": 21790
    },
    {
      "epoch": 8.47,
      "learning_rate": 1.8871197411003237e-05,
      "loss": 0.1737,
      "step": 21800
    },
    {
      "epoch": 8.47,
      "learning_rate": 1.8870679611650485e-05,
      "loss": 0.1333,
      "step": 21810
    },
    {
      "epoch": 8.47,
      "learning_rate": 1.8870161812297736e-05,
      "loss": 0.2282,
      "step": 21820
    },
    {
      "epoch": 8.48,
      "learning_rate": 1.8869644012944984e-05,
      "loss": 0.1288,
      "step": 21830
    },
    {
      "epoch": 8.48,
      "learning_rate": 1.8869126213592236e-05,
      "loss": 0.0718,
      "step": 21840
    },
    {
      "epoch": 8.49,
      "learning_rate": 1.8868608414239484e-05,
      "loss": 0.1946,
      "step": 21850
    },
    {
      "epoch": 8.49,
      "learning_rate": 1.8868090614886735e-05,
      "loss": 0.2471,
      "step": 21860
    },
    {
      "epoch": 8.49,
      "learning_rate": 1.8867572815533983e-05,
      "loss": 0.1732,
      "step": 21870
    },
    {
      "epoch": 8.5,
      "learning_rate": 1.886705501618123e-05,
      "loss": 0.045,
      "step": 21880
    },
    {
      "epoch": 8.5,
      "learning_rate": 1.886653721682848e-05,
      "loss": 0.2658,
      "step": 21890
    },
    {
      "epoch": 8.5,
      "learning_rate": 1.886601941747573e-05,
      "loss": 0.007,
      "step": 21900
    },
    {
      "epoch": 8.51,
      "learning_rate": 1.8865501618122978e-05,
      "loss": 0.1167,
      "step": 21910
    },
    {
      "epoch": 8.51,
      "learning_rate": 1.886498381877023e-05,
      "loss": 0.2181,
      "step": 21920
    },
    {
      "epoch": 8.52,
      "learning_rate": 1.8864466019417477e-05,
      "loss": 0.3278,
      "step": 21930
    },
    {
      "epoch": 8.52,
      "learning_rate": 1.886394822006473e-05,
      "loss": 0.0442,
      "step": 21940
    },
    {
      "epoch": 8.52,
      "learning_rate": 1.8863430420711976e-05,
      "loss": 0.2037,
      "step": 21950
    },
    {
      "epoch": 8.53,
      "learning_rate": 1.8862912621359224e-05,
      "loss": 0.1656,
      "step": 21960
    },
    {
      "epoch": 8.53,
      "learning_rate": 1.8862394822006472e-05,
      "loss": 0.2284,
      "step": 21970
    },
    {
      "epoch": 8.54,
      "learning_rate": 1.8861877022653724e-05,
      "loss": 0.1784,
      "step": 21980
    },
    {
      "epoch": 8.54,
      "learning_rate": 1.886135922330097e-05,
      "loss": 0.1356,
      "step": 21990
    },
    {
      "epoch": 8.54,
      "learning_rate": 1.8860841423948223e-05,
      "loss": 0.1266,
      "step": 22000
    },
    {
      "epoch": 8.55,
      "learning_rate": 1.886032362459547e-05,
      "loss": 0.1452,
      "step": 22010
    },
    {
      "epoch": 8.55,
      "learning_rate": 1.8859805825242722e-05,
      "loss": 0.1829,
      "step": 22020
    },
    {
      "epoch": 8.56,
      "learning_rate": 1.885928802588997e-05,
      "loss": 0.0303,
      "step": 22030
    },
    {
      "epoch": 8.56,
      "learning_rate": 1.8858770226537218e-05,
      "loss": 0.342,
      "step": 22040
    },
    {
      "epoch": 8.56,
      "learning_rate": 1.8858252427184466e-05,
      "loss": 0.2479,
      "step": 22050
    },
    {
      "epoch": 8.57,
      "learning_rate": 1.8857734627831717e-05,
      "loss": 0.2139,
      "step": 22060
    },
    {
      "epoch": 8.57,
      "learning_rate": 1.8857216828478965e-05,
      "loss": 0.2386,
      "step": 22070
    },
    {
      "epoch": 8.57,
      "learning_rate": 1.8856699029126216e-05,
      "loss": 0.1465,
      "step": 22080
    },
    {
      "epoch": 8.58,
      "learning_rate": 1.8856181229773464e-05,
      "loss": 0.2001,
      "step": 22090
    },
    {
      "epoch": 8.58,
      "learning_rate": 1.8855663430420716e-05,
      "loss": 0.1362,
      "step": 22100
    },
    {
      "epoch": 8.59,
      "learning_rate": 1.8855145631067964e-05,
      "loss": 0.0865,
      "step": 22110
    },
    {
      "epoch": 8.59,
      "learning_rate": 1.885462783171521e-05,
      "loss": 0.1359,
      "step": 22120
    },
    {
      "epoch": 8.59,
      "learning_rate": 1.885411003236246e-05,
      "loss": 0.1911,
      "step": 22130
    },
    {
      "epoch": 8.6,
      "learning_rate": 1.885359223300971e-05,
      "loss": 0.1852,
      "step": 22140
    },
    {
      "epoch": 8.6,
      "learning_rate": 1.885307443365696e-05,
      "loss": 0.2367,
      "step": 22150
    },
    {
      "epoch": 8.61,
      "learning_rate": 1.885255663430421e-05,
      "loss": 0.1994,
      "step": 22160
    },
    {
      "epoch": 8.61,
      "learning_rate": 1.8852038834951458e-05,
      "loss": 0.2431,
      "step": 22170
    },
    {
      "epoch": 8.61,
      "learning_rate": 1.885152103559871e-05,
      "loss": 0.1813,
      "step": 22180
    },
    {
      "epoch": 8.62,
      "learning_rate": 1.8851003236245957e-05,
      "loss": 0.2346,
      "step": 22190
    },
    {
      "epoch": 8.62,
      "learning_rate": 1.8850485436893205e-05,
      "loss": 0.1245,
      "step": 22200
    },
    {
      "epoch": 8.63,
      "learning_rate": 1.8849967637540453e-05,
      "loss": 0.1945,
      "step": 22210
    },
    {
      "epoch": 8.63,
      "learning_rate": 1.8849449838187704e-05,
      "loss": 0.4594,
      "step": 22220
    },
    {
      "epoch": 8.63,
      "learning_rate": 1.8848932038834952e-05,
      "loss": 0.1934,
      "step": 22230
    },
    {
      "epoch": 8.64,
      "learning_rate": 1.8848414239482203e-05,
      "loss": 0.1258,
      "step": 22240
    },
    {
      "epoch": 8.64,
      "learning_rate": 1.884789644012945e-05,
      "loss": 0.1497,
      "step": 22250
    },
    {
      "epoch": 8.64,
      "learning_rate": 1.8847378640776703e-05,
      "loss": 0.0801,
      "step": 22260
    },
    {
      "epoch": 8.65,
      "learning_rate": 1.884686084142395e-05,
      "loss": 0.0904,
      "step": 22270
    },
    {
      "epoch": 8.65,
      "learning_rate": 1.88463430420712e-05,
      "loss": 0.186,
      "step": 22280
    },
    {
      "epoch": 8.66,
      "learning_rate": 1.8845825242718446e-05,
      "loss": 0.136,
      "step": 22290
    },
    {
      "epoch": 8.66,
      "learning_rate": 1.8845307443365698e-05,
      "loss": 0.0736,
      "step": 22300
    },
    {
      "epoch": 8.66,
      "learning_rate": 1.8844789644012946e-05,
      "loss": 0.1106,
      "step": 22310
    },
    {
      "epoch": 8.67,
      "learning_rate": 1.8844271844660197e-05,
      "loss": 0.1226,
      "step": 22320
    },
    {
      "epoch": 8.67,
      "learning_rate": 1.8843754045307445e-05,
      "loss": 0.1307,
      "step": 22330
    },
    {
      "epoch": 8.68,
      "learning_rate": 1.8843236245954696e-05,
      "loss": 0.2153,
      "step": 22340
    },
    {
      "epoch": 8.68,
      "learning_rate": 1.8842718446601944e-05,
      "loss": 0.2044,
      "step": 22350
    },
    {
      "epoch": 8.68,
      "learning_rate": 1.8842200647249192e-05,
      "loss": 0.2067,
      "step": 22360
    },
    {
      "epoch": 8.69,
      "learning_rate": 1.884168284789644e-05,
      "loss": 0.1372,
      "step": 22370
    },
    {
      "epoch": 8.69,
      "learning_rate": 1.884116504854369e-05,
      "loss": 0.1949,
      "step": 22380
    },
    {
      "epoch": 8.7,
      "learning_rate": 1.884064724919094e-05,
      "loss": 0.1666,
      "step": 22390
    },
    {
      "epoch": 8.7,
      "learning_rate": 1.884012944983819e-05,
      "loss": 0.1552,
      "step": 22400
    },
    {
      "epoch": 8.7,
      "learning_rate": 1.883961165048544e-05,
      "loss": 0.2195,
      "step": 22410
    },
    {
      "epoch": 8.71,
      "learning_rate": 1.883909385113269e-05,
      "loss": 0.1618,
      "step": 22420
    },
    {
      "epoch": 8.71,
      "learning_rate": 1.8838576051779938e-05,
      "loss": 0.1074,
      "step": 22430
    },
    {
      "epoch": 8.71,
      "learning_rate": 1.8838058252427186e-05,
      "loss": 0.1523,
      "step": 22440
    },
    {
      "epoch": 8.72,
      "learning_rate": 1.8837540453074434e-05,
      "loss": 0.2489,
      "step": 22450
    },
    {
      "epoch": 8.72,
      "learning_rate": 1.8837022653721685e-05,
      "loss": 0.0645,
      "step": 22460
    },
    {
      "epoch": 8.73,
      "learning_rate": 1.8836504854368933e-05,
      "loss": 0.08,
      "step": 22470
    },
    {
      "epoch": 8.73,
      "learning_rate": 1.8835987055016184e-05,
      "loss": 0.0803,
      "step": 22480
    },
    {
      "epoch": 8.73,
      "learning_rate": 1.8835469255663432e-05,
      "loss": 0.1131,
      "step": 22490
    },
    {
      "epoch": 8.74,
      "learning_rate": 1.883495145631068e-05,
      "loss": 0.1339,
      "step": 22500
    },
    {
      "epoch": 8.74,
      "learning_rate": 1.883443365695793e-05,
      "loss": 0.179,
      "step": 22510
    },
    {
      "epoch": 8.75,
      "learning_rate": 1.883391585760518e-05,
      "loss": 0.1558,
      "step": 22520
    },
    {
      "epoch": 8.75,
      "learning_rate": 1.8833398058252427e-05,
      "loss": 0.2514,
      "step": 22530
    },
    {
      "epoch": 8.75,
      "learning_rate": 1.883288025889968e-05,
      "loss": 0.2662,
      "step": 22540
    },
    {
      "epoch": 8.76,
      "learning_rate": 1.8832362459546926e-05,
      "loss": 0.0908,
      "step": 22550
    },
    {
      "epoch": 8.76,
      "learning_rate": 1.8831844660194178e-05,
      "loss": 0.191,
      "step": 22560
    },
    {
      "epoch": 8.77,
      "learning_rate": 1.8831326860841426e-05,
      "loss": 0.1862,
      "step": 22570
    },
    {
      "epoch": 8.77,
      "learning_rate": 1.8830809061488674e-05,
      "loss": 0.0145,
      "step": 22580
    },
    {
      "epoch": 8.77,
      "learning_rate": 1.8830291262135925e-05,
      "loss": 0.2788,
      "step": 22590
    },
    {
      "epoch": 8.78,
      "learning_rate": 1.8829773462783173e-05,
      "loss": 0.1295,
      "step": 22600
    },
    {
      "epoch": 8.78,
      "learning_rate": 1.882925566343042e-05,
      "loss": 0.0575,
      "step": 22610
    },
    {
      "epoch": 8.78,
      "learning_rate": 1.8828737864077672e-05,
      "loss": 0.0336,
      "step": 22620
    },
    {
      "epoch": 8.79,
      "learning_rate": 1.882822006472492e-05,
      "loss": 0.1354,
      "step": 22630
    },
    {
      "epoch": 8.79,
      "learning_rate": 1.882770226537217e-05,
      "loss": 0.1404,
      "step": 22640
    },
    {
      "epoch": 8.8,
      "learning_rate": 1.882718446601942e-05,
      "loss": 0.0881,
      "step": 22650
    },
    {
      "epoch": 8.8,
      "learning_rate": 1.8826666666666667e-05,
      "loss": 0.0713,
      "step": 22660
    },
    {
      "epoch": 8.8,
      "learning_rate": 1.882614886731392e-05,
      "loss": 0.2582,
      "step": 22670
    },
    {
      "epoch": 8.81,
      "learning_rate": 1.8825631067961166e-05,
      "loss": 0.2333,
      "step": 22680
    },
    {
      "epoch": 8.81,
      "learning_rate": 1.8825113268608414e-05,
      "loss": 0.279,
      "step": 22690
    },
    {
      "epoch": 8.82,
      "learning_rate": 1.8824595469255666e-05,
      "loss": 0.0331,
      "step": 22700
    },
    {
      "epoch": 8.82,
      "learning_rate": 1.8824077669902914e-05,
      "loss": 0.1121,
      "step": 22710
    },
    {
      "epoch": 8.82,
      "learning_rate": 1.8823559870550165e-05,
      "loss": 0.178,
      "step": 22720
    },
    {
      "epoch": 8.83,
      "learning_rate": 1.8823042071197413e-05,
      "loss": 0.0887,
      "step": 22730
    },
    {
      "epoch": 8.83,
      "learning_rate": 1.882252427184466e-05,
      "loss": 0.1439,
      "step": 22740
    },
    {
      "epoch": 8.83,
      "learning_rate": 1.8822006472491912e-05,
      "loss": 0.1999,
      "step": 22750
    },
    {
      "epoch": 8.84,
      "learning_rate": 1.882148867313916e-05,
      "loss": 0.1251,
      "step": 22760
    },
    {
      "epoch": 8.84,
      "learning_rate": 1.8820970873786408e-05,
      "loss": 0.1234,
      "step": 22770
    },
    {
      "epoch": 8.85,
      "learning_rate": 1.882045307443366e-05,
      "loss": 0.2392,
      "step": 22780
    },
    {
      "epoch": 8.85,
      "learning_rate": 1.8819935275080907e-05,
      "loss": 0.2333,
      "step": 22790
    },
    {
      "epoch": 8.85,
      "learning_rate": 1.881941747572816e-05,
      "loss": 0.0658,
      "step": 22800
    },
    {
      "epoch": 8.86,
      "learning_rate": 1.8818899676375406e-05,
      "loss": 0.2955,
      "step": 22810
    },
    {
      "epoch": 8.86,
      "learning_rate": 1.8818381877022654e-05,
      "loss": 0.1832,
      "step": 22820
    },
    {
      "epoch": 8.87,
      "learning_rate": 1.8817864077669906e-05,
      "loss": 0.1502,
      "step": 22830
    },
    {
      "epoch": 8.87,
      "learning_rate": 1.8817346278317153e-05,
      "loss": 0.0341,
      "step": 22840
    },
    {
      "epoch": 8.87,
      "learning_rate": 1.88168284789644e-05,
      "loss": 0.1151,
      "step": 22850
    },
    {
      "epoch": 8.88,
      "learning_rate": 1.8816310679611653e-05,
      "loss": 0.2136,
      "step": 22860
    },
    {
      "epoch": 8.88,
      "learning_rate": 1.88157928802589e-05,
      "loss": 0.0592,
      "step": 22870
    },
    {
      "epoch": 8.89,
      "learning_rate": 1.881527508090615e-05,
      "loss": 0.1838,
      "step": 22880
    },
    {
      "epoch": 8.89,
      "learning_rate": 1.88147572815534e-05,
      "loss": 0.2911,
      "step": 22890
    },
    {
      "epoch": 8.89,
      "learning_rate": 1.8814239482200648e-05,
      "loss": 0.2536,
      "step": 22900
    },
    {
      "epoch": 8.9,
      "learning_rate": 1.88137216828479e-05,
      "loss": 0.059,
      "step": 22910
    },
    {
      "epoch": 8.9,
      "learning_rate": 1.8813203883495147e-05,
      "loss": 0.0499,
      "step": 22920
    },
    {
      "epoch": 8.9,
      "learning_rate": 1.8812686084142395e-05,
      "loss": 0.0978,
      "step": 22930
    },
    {
      "epoch": 8.91,
      "learning_rate": 1.8812168284789646e-05,
      "loss": 0.1474,
      "step": 22940
    },
    {
      "epoch": 8.91,
      "learning_rate": 1.8811650485436894e-05,
      "loss": 0.2528,
      "step": 22950
    },
    {
      "epoch": 8.92,
      "learning_rate": 1.8811132686084142e-05,
      "loss": 0.2705,
      "step": 22960
    },
    {
      "epoch": 8.92,
      "learning_rate": 1.8810614886731393e-05,
      "loss": 0.177,
      "step": 22970
    },
    {
      "epoch": 8.92,
      "learning_rate": 1.881009708737864e-05,
      "loss": 0.1231,
      "step": 22980
    },
    {
      "epoch": 8.93,
      "learning_rate": 1.8809579288025893e-05,
      "loss": 0.3388,
      "step": 22990
    },
    {
      "epoch": 8.93,
      "learning_rate": 1.880906148867314e-05,
      "loss": 0.1711,
      "step": 23000
    },
    {
      "epoch": 8.94,
      "learning_rate": 1.8808543689320392e-05,
      "loss": 0.1108,
      "step": 23010
    },
    {
      "epoch": 8.94,
      "learning_rate": 1.880802588996764e-05,
      "loss": 0.0579,
      "step": 23020
    },
    {
      "epoch": 8.94,
      "learning_rate": 1.8807508090614888e-05,
      "loss": 0.1931,
      "step": 23030
    },
    {
      "epoch": 8.95,
      "learning_rate": 1.8806990291262136e-05,
      "loss": 0.2855,
      "step": 23040
    },
    {
      "epoch": 8.95,
      "learning_rate": 1.8806472491909387e-05,
      "loss": 0.1447,
      "step": 23050
    },
    {
      "epoch": 8.96,
      "learning_rate": 1.8805954692556635e-05,
      "loss": 0.4596,
      "step": 23060
    },
    {
      "epoch": 8.96,
      "learning_rate": 1.8805436893203886e-05,
      "loss": 0.1517,
      "step": 23070
    },
    {
      "epoch": 8.96,
      "learning_rate": 1.8804919093851134e-05,
      "loss": 0.0564,
      "step": 23080
    },
    {
      "epoch": 8.97,
      "learning_rate": 1.8804401294498385e-05,
      "loss": 0.0409,
      "step": 23090
    },
    {
      "epoch": 8.97,
      "learning_rate": 1.8803883495145633e-05,
      "loss": 0.1732,
      "step": 23100
    },
    {
      "epoch": 8.97,
      "learning_rate": 1.880336569579288e-05,
      "loss": 0.1264,
      "step": 23110
    },
    {
      "epoch": 8.98,
      "learning_rate": 1.880284789644013e-05,
      "loss": 0.0731,
      "step": 23120
    },
    {
      "epoch": 8.98,
      "learning_rate": 1.880233009708738e-05,
      "loss": 0.0486,
      "step": 23130
    },
    {
      "epoch": 8.99,
      "learning_rate": 1.880181229773463e-05,
      "loss": 0.2034,
      "step": 23140
    },
    {
      "epoch": 8.99,
      "learning_rate": 1.880129449838188e-05,
      "loss": 0.2087,
      "step": 23150
    },
    {
      "epoch": 8.99,
      "learning_rate": 1.8800776699029128e-05,
      "loss": 0.0761,
      "step": 23160
    },
    {
      "epoch": 9.0,
      "learning_rate": 1.880025889967638e-05,
      "loss": 0.2658,
      "step": 23170
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.20891419053077698,
      "eval_runtime": 8.1732,
      "eval_samples_per_second": 444.745,
      "eval_steps_per_second": 55.67,
      "step": 23175
    },
    {
      "epoch": 9.0,
      "learning_rate": 1.8799741100323624e-05,
      "loss": 0.1638,
      "step": 23180
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.8799223300970875e-05,
      "loss": 0.1532,
      "step": 23190
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.8798705501618123e-05,
      "loss": 0.1557,
      "step": 23200
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.8798187702265374e-05,
      "loss": 0.0958,
      "step": 23210
    },
    {
      "epoch": 9.02,
      "learning_rate": 1.8797669902912622e-05,
      "loss": 0.1796,
      "step": 23220
    },
    {
      "epoch": 9.02,
      "learning_rate": 1.8797152103559873e-05,
      "loss": 0.1515,
      "step": 23230
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.879663430420712e-05,
      "loss": 0.1639,
      "step": 23240
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.8796116504854373e-05,
      "loss": 0.2391,
      "step": 23250
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.8795598705501617e-05,
      "loss": 0.1436,
      "step": 23260
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.879508090614887e-05,
      "loss": 0.1783,
      "step": 23270
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.8794563106796116e-05,
      "loss": 0.1956,
      "step": 23280
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.8794045307443368e-05,
      "loss": 0.3148,
      "step": 23290
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.8793527508090616e-05,
      "loss": 0.0489,
      "step": 23300
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.8793009708737867e-05,
      "loss": 0.1087,
      "step": 23310
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.8792491909385115e-05,
      "loss": 0.1003,
      "step": 23320
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.8791974110032366e-05,
      "loss": 0.0212,
      "step": 23330
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.879145631067961e-05,
      "loss": 0.2206,
      "step": 23340
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.8790938511326862e-05,
      "loss": 0.1293,
      "step": 23350
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.879042071197411e-05,
      "loss": 0.0781,
      "step": 23360
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.878990291262136e-05,
      "loss": 0.1474,
      "step": 23370
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.878938511326861e-05,
      "loss": 0.1563,
      "step": 23380
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.878886731391586e-05,
      "loss": 0.0669,
      "step": 23390
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.878834951456311e-05,
      "loss": 0.1168,
      "step": 23400
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.878783171521036e-05,
      "loss": 0.3145,
      "step": 23410
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.8787313915857604e-05,
      "loss": 0.4579,
      "step": 23420
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.8786796116504856e-05,
      "loss": 0.1775,
      "step": 23430
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.8786278317152103e-05,
      "loss": 0.042,
      "step": 23440
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.8785760517799355e-05,
      "loss": 0.1732,
      "step": 23450
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.8785242718446603e-05,
      "loss": 0.1292,
      "step": 23460
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.8784724919093854e-05,
      "loss": 0.1745,
      "step": 23470
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.8784207119741102e-05,
      "loss": 0.1205,
      "step": 23480
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.8783689320388353e-05,
      "loss": 0.1973,
      "step": 23490
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.8783171521035598e-05,
      "loss": 0.174,
      "step": 23500
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.878265372168285e-05,
      "loss": 0.1793,
      "step": 23510
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.8782135922330097e-05,
      "loss": 0.0831,
      "step": 23520
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.878161812297735e-05,
      "loss": 0.16,
      "step": 23530
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.8781100323624596e-05,
      "loss": 0.2863,
      "step": 23540
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.8780582524271848e-05,
      "loss": 0.1604,
      "step": 23550
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.8780064724919095e-05,
      "loss": 0.1071,
      "step": 23560
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.8779546925566347e-05,
      "loss": 0.2428,
      "step": 23570
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8779029126213595e-05,
      "loss": 0.2421,
      "step": 23580
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.8778511326860843e-05,
      "loss": 0.2965,
      "step": 23590
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.877799352750809e-05,
      "loss": 0.2266,
      "step": 23600
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.8777475728155342e-05,
      "loss": 0.1337,
      "step": 23610
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.877695792880259e-05,
      "loss": 0.0963,
      "step": 23620
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.877644012944984e-05,
      "loss": 0.1094,
      "step": 23630
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.877592233009709e-05,
      "loss": 0.24,
      "step": 23640
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.877540453074434e-05,
      "loss": 0.1363,
      "step": 23650
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.8774886731391588e-05,
      "loss": 0.1944,
      "step": 23660
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.8774368932038836e-05,
      "loss": 0.207,
      "step": 23670
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.8773851132686084e-05,
      "loss": 0.0788,
      "step": 23680
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.8773333333333335e-05,
      "loss": 0.0214,
      "step": 23690
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.8772815533980583e-05,
      "loss": 0.0217,
      "step": 23700
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.8772297734627835e-05,
      "loss": 0.1607,
      "step": 23710
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.8771779935275083e-05,
      "loss": 0.1618,
      "step": 23720
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.8771262135922334e-05,
      "loss": 0.1063,
      "step": 23730
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.8770744336569582e-05,
      "loss": 0.1788,
      "step": 23740
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.877022653721683e-05,
      "loss": 0.1484,
      "step": 23750
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.8769708737864078e-05,
      "loss": 0.28,
      "step": 23760
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.876919093851133e-05,
      "loss": 0.29,
      "step": 23770
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.8768673139158577e-05,
      "loss": 0.1091,
      "step": 23780
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.8768155339805828e-05,
      "loss": 0.1332,
      "step": 23790
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.8767637540453076e-05,
      "loss": 0.0741,
      "step": 23800
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.8767119741100327e-05,
      "loss": 0.0882,
      "step": 23810
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.8766601941747575e-05,
      "loss": 0.0816,
      "step": 23820
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.8766084142394823e-05,
      "loss": 0.1641,
      "step": 23830
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.876556634304207e-05,
      "loss": 0.1122,
      "step": 23840
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.8765048543689323e-05,
      "loss": 0.3096,
      "step": 23850
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.876453074433657e-05,
      "loss": 0.0827,
      "step": 23860
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.8764012944983822e-05,
      "loss": 0.0879,
      "step": 23870
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.876349514563107e-05,
      "loss": 0.0866,
      "step": 23880
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.876297734627832e-05,
      "loss": 0.1006,
      "step": 23890
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.876245954692557e-05,
      "loss": 0.3453,
      "step": 23900
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.8761941747572817e-05,
      "loss": 0.0474,
      "step": 23910
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.8761423948220065e-05,
      "loss": 0.2194,
      "step": 23920
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.8760906148867316e-05,
      "loss": 0.097,
      "step": 23930
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.8760388349514564e-05,
      "loss": 0.1382,
      "step": 23940
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.8759870550161815e-05,
      "loss": 0.3141,
      "step": 23950
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.8759352750809063e-05,
      "loss": 0.3154,
      "step": 23960
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.875883495145631e-05,
      "loss": 0.168,
      "step": 23970
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.8758317152103562e-05,
      "loss": 0.4143,
      "step": 23980
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.875779935275081e-05,
      "loss": 0.0658,
      "step": 23990
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.875728155339806e-05,
      "loss": 0.0296,
      "step": 24000
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.875676375404531e-05,
      "loss": 0.0106,
      "step": 24010
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.8756245954692558e-05,
      "loss": 0.161,
      "step": 24020
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.875572815533981e-05,
      "loss": 0.1689,
      "step": 24030
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.8755210355987057e-05,
      "loss": 0.2394,
      "step": 24040
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.8754692556634305e-05,
      "loss": 0.2823,
      "step": 24050
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.8754174757281556e-05,
      "loss": 0.207,
      "step": 24060
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.8753656957928804e-05,
      "loss": 0.0422,
      "step": 24070
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.8753139158576052e-05,
      "loss": 0.0894,
      "step": 24080
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.8752621359223303e-05,
      "loss": 0.2712,
      "step": 24090
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.875210355987055e-05,
      "loss": 0.0805,
      "step": 24100
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.8751585760517802e-05,
      "loss": 0.2002,
      "step": 24110
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.875106796116505e-05,
      "loss": 0.1737,
      "step": 24120
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.8750550161812298e-05,
      "loss": 0.2772,
      "step": 24130
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.875003236245955e-05,
      "loss": 0.1467,
      "step": 24140
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.8749514563106798e-05,
      "loss": 0.2014,
      "step": 24150
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.8748996763754045e-05,
      "loss": 0.0934,
      "step": 24160
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.8748478964401297e-05,
      "loss": 0.1686,
      "step": 24170
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.8747961165048545e-05,
      "loss": 0.0807,
      "step": 24180
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.8747443365695796e-05,
      "loss": 0.1102,
      "step": 24190
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.8746925566343044e-05,
      "loss": 0.3011,
      "step": 24200
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.8746407766990292e-05,
      "loss": 0.1743,
      "step": 24210
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.8745889967637543e-05,
      "loss": 0.1716,
      "step": 24220
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.874537216828479e-05,
      "loss": 0.2234,
      "step": 24230
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.874485436893204e-05,
      "loss": 0.1947,
      "step": 24240
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.874433656957929e-05,
      "loss": 0.0559,
      "step": 24250
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.8743818770226538e-05,
      "loss": 0.2059,
      "step": 24260
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.874330097087379e-05,
      "loss": 0.0878,
      "step": 24270
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.8742783171521037e-05,
      "loss": 0.1335,
      "step": 24280
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.8742265372168285e-05,
      "loss": 0.0188,
      "step": 24290
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.8741747572815537e-05,
      "loss": 0.1089,
      "step": 24300
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.8741229773462785e-05,
      "loss": 0.1117,
      "step": 24310
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.8740711974110033e-05,
      "loss": 0.1254,
      "step": 24320
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.8740194174757284e-05,
      "loss": 0.2168,
      "step": 24330
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.8739676375404532e-05,
      "loss": 0.2825,
      "step": 24340
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.873915857605178e-05,
      "loss": 0.1318,
      "step": 24350
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.873864077669903e-05,
      "loss": 0.1697,
      "step": 24360
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.873812297734628e-05,
      "loss": 0.044,
      "step": 24370
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.873760517799353e-05,
      "loss": 0.093,
      "step": 24380
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.8737087378640778e-05,
      "loss": 0.2223,
      "step": 24390
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.8736569579288026e-05,
      "loss": 0.1688,
      "step": 24400
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.8736051779935277e-05,
      "loss": 0.1534,
      "step": 24410
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.8735533980582525e-05,
      "loss": 0.3042,
      "step": 24420
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.8735016181229773e-05,
      "loss": 0.117,
      "step": 24430
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.8734498381877025e-05,
      "loss": 0.3192,
      "step": 24440
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.8733980582524273e-05,
      "loss": 0.1599,
      "step": 24450
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.8733462783171524e-05,
      "loss": 0.1845,
      "step": 24460
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.8732944983818772e-05,
      "loss": 0.3663,
      "step": 24470
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.873242718446602e-05,
      "loss": 0.1378,
      "step": 24480
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.873190938511327e-05,
      "loss": 0.1054,
      "step": 24490
    },
    {
      "epoch": 9.51,
      "learning_rate": 1.873139158576052e-05,
      "loss": 0.2037,
      "step": 24500
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.8730873786407767e-05,
      "loss": 0.1062,
      "step": 24510
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.8730355987055018e-05,
      "loss": 0.1122,
      "step": 24520
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.8729838187702266e-05,
      "loss": 0.2145,
      "step": 24530
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.8729320388349517e-05,
      "loss": 0.1264,
      "step": 24540
    },
    {
      "epoch": 9.53,
      "learning_rate": 1.8728802588996765e-05,
      "loss": 0.1689,
      "step": 24550
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.8728284789644013e-05,
      "loss": 0.524,
      "step": 24560
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.8727766990291265e-05,
      "loss": 0.0244,
      "step": 24570
    },
    {
      "epoch": 9.55,
      "learning_rate": 1.8727249190938512e-05,
      "loss": 0.1799,
      "step": 24580
    },
    {
      "epoch": 9.55,
      "learning_rate": 1.872673139158576e-05,
      "loss": 0.2759,
      "step": 24590
    },
    {
      "epoch": 9.55,
      "learning_rate": 1.8726213592233012e-05,
      "loss": 0.1937,
      "step": 24600
    },
    {
      "epoch": 9.56,
      "learning_rate": 1.872569579288026e-05,
      "loss": 0.146,
      "step": 24610
    },
    {
      "epoch": 9.56,
      "learning_rate": 1.872517799352751e-05,
      "loss": 0.1948,
      "step": 24620
    },
    {
      "epoch": 9.57,
      "learning_rate": 1.872466019417476e-05,
      "loss": 0.2984,
      "step": 24630
    },
    {
      "epoch": 9.57,
      "learning_rate": 1.8724142394822007e-05,
      "loss": 0.0892,
      "step": 24640
    },
    {
      "epoch": 9.57,
      "learning_rate": 1.8723624595469255e-05,
      "loss": 0.0497,
      "step": 24650
    },
    {
      "epoch": 9.58,
      "learning_rate": 1.8723106796116506e-05,
      "loss": 0.3148,
      "step": 24660
    },
    {
      "epoch": 9.58,
      "learning_rate": 1.8722588996763754e-05,
      "loss": 0.1518,
      "step": 24670
    },
    {
      "epoch": 9.58,
      "learning_rate": 1.8722071197411005e-05,
      "loss": 0.0815,
      "step": 24680
    },
    {
      "epoch": 9.59,
      "learning_rate": 1.8721553398058253e-05,
      "loss": 0.1222,
      "step": 24690
    },
    {
      "epoch": 9.59,
      "learning_rate": 1.8721035598705504e-05,
      "loss": 0.0833,
      "step": 24700
    },
    {
      "epoch": 9.6,
      "learning_rate": 1.8720517799352752e-05,
      "loss": 0.0584,
      "step": 24710
    },
    {
      "epoch": 9.6,
      "learning_rate": 1.8720000000000004e-05,
      "loss": 0.1356,
      "step": 24720
    },
    {
      "epoch": 9.6,
      "learning_rate": 1.8719482200647248e-05,
      "loss": 0.1552,
      "step": 24730
    },
    {
      "epoch": 9.61,
      "learning_rate": 1.87189644012945e-05,
      "loss": 0.0614,
      "step": 24740
    },
    {
      "epoch": 9.61,
      "learning_rate": 1.8718446601941747e-05,
      "loss": 0.1161,
      "step": 24750
    },
    {
      "epoch": 9.62,
      "learning_rate": 1.8717928802589e-05,
      "loss": 0.1495,
      "step": 24760
    },
    {
      "epoch": 9.62,
      "learning_rate": 1.8717411003236247e-05,
      "loss": 0.1536,
      "step": 24770
    },
    {
      "epoch": 9.62,
      "learning_rate": 1.8716893203883498e-05,
      "loss": 0.2008,
      "step": 24780
    },
    {
      "epoch": 9.63,
      "learning_rate": 1.8716375404530746e-05,
      "loss": 0.066,
      "step": 24790
    },
    {
      "epoch": 9.63,
      "learning_rate": 1.8715857605177997e-05,
      "loss": 0.1124,
      "step": 24800
    },
    {
      "epoch": 9.63,
      "learning_rate": 1.8715339805825242e-05,
      "loss": 0.383,
      "step": 24810
    },
    {
      "epoch": 9.64,
      "learning_rate": 1.8714822006472493e-05,
      "loss": 0.2125,
      "step": 24820
    },
    {
      "epoch": 9.64,
      "learning_rate": 1.871430420711974e-05,
      "loss": 0.2071,
      "step": 24830
    },
    {
      "epoch": 9.65,
      "learning_rate": 1.8713786407766992e-05,
      "loss": 0.1297,
      "step": 24840
    },
    {
      "epoch": 9.65,
      "learning_rate": 1.871326860841424e-05,
      "loss": 0.1529,
      "step": 24850
    },
    {
      "epoch": 9.65,
      "learning_rate": 1.871275080906149e-05,
      "loss": 0.247,
      "step": 24860
    },
    {
      "epoch": 9.66,
      "learning_rate": 1.871223300970874e-05,
      "loss": 0.2264,
      "step": 24870
    },
    {
      "epoch": 9.66,
      "learning_rate": 1.871171521035599e-05,
      "loss": 0.0495,
      "step": 24880
    },
    {
      "epoch": 9.67,
      "learning_rate": 1.8711197411003235e-05,
      "loss": 0.0692,
      "step": 24890
    },
    {
      "epoch": 9.67,
      "learning_rate": 1.8710679611650487e-05,
      "loss": 0.0613,
      "step": 24900
    },
    {
      "epoch": 9.67,
      "learning_rate": 1.8710161812297735e-05,
      "loss": 0.0759,
      "step": 24910
    },
    {
      "epoch": 9.68,
      "learning_rate": 1.8709644012944986e-05,
      "loss": 0.2253,
      "step": 24920
    },
    {
      "epoch": 9.68,
      "learning_rate": 1.8709126213592234e-05,
      "loss": 0.1108,
      "step": 24930
    },
    {
      "epoch": 9.69,
      "learning_rate": 1.8708608414239485e-05,
      "loss": 0.296,
      "step": 24940
    },
    {
      "epoch": 9.69,
      "learning_rate": 1.8708090614886733e-05,
      "loss": 0.1144,
      "step": 24950
    },
    {
      "epoch": 9.69,
      "learning_rate": 1.8707572815533984e-05,
      "loss": 0.2958,
      "step": 24960
    },
    {
      "epoch": 9.7,
      "learning_rate": 1.870705501618123e-05,
      "loss": 0.0722,
      "step": 24970
    },
    {
      "epoch": 9.7,
      "learning_rate": 1.870653721682848e-05,
      "loss": 0.1813,
      "step": 24980
    },
    {
      "epoch": 9.7,
      "learning_rate": 1.8706019417475728e-05,
      "loss": 0.1417,
      "step": 24990
    },
    {
      "epoch": 9.71,
      "learning_rate": 1.870550161812298e-05,
      "loss": 0.0523,
      "step": 25000
    },
    {
      "epoch": 9.71,
      "learning_rate": 1.8704983818770227e-05,
      "loss": 0.1182,
      "step": 25010
    },
    {
      "epoch": 9.72,
      "learning_rate": 1.870446601941748e-05,
      "loss": 0.2804,
      "step": 25020
    },
    {
      "epoch": 9.72,
      "learning_rate": 1.8703948220064727e-05,
      "loss": 0.1601,
      "step": 25030
    },
    {
      "epoch": 9.72,
      "learning_rate": 1.8703430420711978e-05,
      "loss": 0.1039,
      "step": 25040
    },
    {
      "epoch": 9.73,
      "learning_rate": 1.8702912621359222e-05,
      "loss": 0.1146,
      "step": 25050
    },
    {
      "epoch": 9.73,
      "learning_rate": 1.8702394822006474e-05,
      "loss": 0.4542,
      "step": 25060
    },
    {
      "epoch": 9.74,
      "learning_rate": 1.8701877022653722e-05,
      "loss": 0.1719,
      "step": 25070
    },
    {
      "epoch": 9.74,
      "learning_rate": 1.8701359223300973e-05,
      "loss": 0.0769,
      "step": 25080
    },
    {
      "epoch": 9.74,
      "learning_rate": 1.870084142394822e-05,
      "loss": 0.1786,
      "step": 25090
    },
    {
      "epoch": 9.75,
      "learning_rate": 1.8700323624595472e-05,
      "loss": 0.2288,
      "step": 25100
    },
    {
      "epoch": 9.75,
      "learning_rate": 1.869980582524272e-05,
      "loss": 0.1279,
      "step": 25110
    },
    {
      "epoch": 9.76,
      "learning_rate": 1.869928802588997e-05,
      "loss": 0.1431,
      "step": 25120
    },
    {
      "epoch": 9.76,
      "learning_rate": 1.8698770226537216e-05,
      "loss": 0.0392,
      "step": 25130
    },
    {
      "epoch": 9.76,
      "learning_rate": 1.8698252427184467e-05,
      "loss": 0.3541,
      "step": 25140
    },
    {
      "epoch": 9.77,
      "learning_rate": 1.8697734627831715e-05,
      "loss": 0.2483,
      "step": 25150
    },
    {
      "epoch": 9.77,
      "learning_rate": 1.8697216828478967e-05,
      "loss": 0.119,
      "step": 25160
    },
    {
      "epoch": 9.77,
      "learning_rate": 1.8696699029126215e-05,
      "loss": 0.0419,
      "step": 25170
    },
    {
      "epoch": 9.78,
      "learning_rate": 1.8696181229773466e-05,
      "loss": 0.2045,
      "step": 25180
    },
    {
      "epoch": 9.78,
      "learning_rate": 1.8695663430420714e-05,
      "loss": 0.1985,
      "step": 25190
    },
    {
      "epoch": 9.79,
      "learning_rate": 1.8695145631067965e-05,
      "loss": 0.0111,
      "step": 25200
    },
    {
      "epoch": 9.79,
      "learning_rate": 1.869462783171521e-05,
      "loss": 0.1265,
      "step": 25210
    },
    {
      "epoch": 9.79,
      "learning_rate": 1.869411003236246e-05,
      "loss": 0.1387,
      "step": 25220
    },
    {
      "epoch": 9.8,
      "learning_rate": 1.869359223300971e-05,
      "loss": 0.1604,
      "step": 25230
    },
    {
      "epoch": 9.8,
      "learning_rate": 1.869307443365696e-05,
      "loss": 0.1711,
      "step": 25240
    },
    {
      "epoch": 9.81,
      "learning_rate": 1.8692556634304208e-05,
      "loss": 0.1361,
      "step": 25250
    },
    {
      "epoch": 9.81,
      "learning_rate": 1.869203883495146e-05,
      "loss": 0.2162,
      "step": 25260
    },
    {
      "epoch": 9.81,
      "learning_rate": 1.8691521035598707e-05,
      "loss": 0.161,
      "step": 25270
    },
    {
      "epoch": 9.82,
      "learning_rate": 1.869100323624596e-05,
      "loss": 0.0384,
      "step": 25280
    },
    {
      "epoch": 9.82,
      "learning_rate": 1.8690485436893207e-05,
      "loss": 0.2032,
      "step": 25290
    },
    {
      "epoch": 9.83,
      "learning_rate": 1.8689967637540454e-05,
      "loss": 0.1772,
      "step": 25300
    },
    {
      "epoch": 9.83,
      "learning_rate": 1.8689449838187702e-05,
      "loss": 0.0793,
      "step": 25310
    },
    {
      "epoch": 9.83,
      "learning_rate": 1.8688932038834954e-05,
      "loss": 0.1546,
      "step": 25320
    },
    {
      "epoch": 9.84,
      "learning_rate": 1.86884142394822e-05,
      "loss": 0.1464,
      "step": 25330
    },
    {
      "epoch": 9.84,
      "learning_rate": 1.8687896440129453e-05,
      "loss": 0.2053,
      "step": 25340
    },
    {
      "epoch": 9.84,
      "learning_rate": 1.86873786407767e-05,
      "loss": 0.0882,
      "step": 25350
    },
    {
      "epoch": 9.85,
      "learning_rate": 1.8686860841423952e-05,
      "loss": 0.1026,
      "step": 25360
    },
    {
      "epoch": 9.85,
      "learning_rate": 1.86863430420712e-05,
      "loss": 0.1821,
      "step": 25370
    },
    {
      "epoch": 9.86,
      "learning_rate": 1.8685825242718448e-05,
      "loss": 0.1298,
      "step": 25380
    },
    {
      "epoch": 9.86,
      "learning_rate": 1.8685307443365696e-05,
      "loss": 0.0835,
      "step": 25390
    },
    {
      "epoch": 9.86,
      "learning_rate": 1.8684789644012947e-05,
      "loss": 0.1794,
      "step": 25400
    },
    {
      "epoch": 9.87,
      "learning_rate": 1.8684271844660195e-05,
      "loss": 0.1723,
      "step": 25410
    },
    {
      "epoch": 9.87,
      "learning_rate": 1.8683754045307446e-05,
      "loss": 0.1283,
      "step": 25420
    },
    {
      "epoch": 9.88,
      "learning_rate": 1.8683236245954694e-05,
      "loss": 0.0295,
      "step": 25430
    },
    {
      "epoch": 9.88,
      "learning_rate": 1.8682718446601942e-05,
      "loss": 0.0658,
      "step": 25440
    },
    {
      "epoch": 9.88,
      "learning_rate": 1.8682200647249194e-05,
      "loss": 0.1019,
      "step": 25450
    },
    {
      "epoch": 9.89,
      "learning_rate": 1.868168284789644e-05,
      "loss": 0.0565,
      "step": 25460
    },
    {
      "epoch": 9.89,
      "learning_rate": 1.868116504854369e-05,
      "loss": 0.0662,
      "step": 25470
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.868064724919094e-05,
      "loss": 0.1908,
      "step": 25480
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.868012944983819e-05,
      "loss": 0.2838,
      "step": 25490
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.867961165048544e-05,
      "loss": 0.0283,
      "step": 25500
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8679093851132688e-05,
      "loss": 0.1748,
      "step": 25510
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8678576051779936e-05,
      "loss": 0.1246,
      "step": 25520
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8678058252427187e-05,
      "loss": 0.2015,
      "step": 25530
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.8677540453074435e-05,
      "loss": 0.1296,
      "step": 25540
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.8677022653721683e-05,
      "loss": 0.1104,
      "step": 25550
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.8676504854368934e-05,
      "loss": 0.0931,
      "step": 25560
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.8675987055016182e-05,
      "loss": 0.1437,
      "step": 25570
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.8675469255663434e-05,
      "loss": 0.116,
      "step": 25580
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.867495145631068e-05,
      "loss": 0.1506,
      "step": 25590
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.867443365695793e-05,
      "loss": 0.1208,
      "step": 25600
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.867391585760518e-05,
      "loss": 0.1302,
      "step": 25610
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.867339805825243e-05,
      "loss": 0.1621,
      "step": 25620
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.8672880258899677e-05,
      "loss": 0.0848,
      "step": 25630
    },
    {
      "epoch": 9.96,
      "learning_rate": 1.8672362459546928e-05,
      "loss": 0.1485,
      "step": 25640
    },
    {
      "epoch": 9.96,
      "learning_rate": 1.8671844660194176e-05,
      "loss": 0.118,
      "step": 25650
    },
    {
      "epoch": 9.97,
      "learning_rate": 1.8671326860841427e-05,
      "loss": 0.1084,
      "step": 25660
    },
    {
      "epoch": 9.97,
      "learning_rate": 1.8670809061488675e-05,
      "loss": 0.0492,
      "step": 25670
    },
    {
      "epoch": 9.97,
      "learning_rate": 1.8670291262135923e-05,
      "loss": 0.0966,
      "step": 25680
    },
    {
      "epoch": 9.98,
      "learning_rate": 1.8669773462783174e-05,
      "loss": 0.2145,
      "step": 25690
    },
    {
      "epoch": 9.98,
      "learning_rate": 1.8669255663430422e-05,
      "loss": 0.0569,
      "step": 25700
    },
    {
      "epoch": 9.98,
      "learning_rate": 1.866873786407767e-05,
      "loss": 0.1551,
      "step": 25710
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.866822006472492e-05,
      "loss": 0.1834,
      "step": 25720
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.866770226537217e-05,
      "loss": 0.0876,
      "step": 25730
    },
    {
      "epoch": 10.0,
      "learning_rate": 1.8667184466019417e-05,
      "loss": 0.3497,
      "step": 25740
    },
    {
      "epoch": 10.0,
      "learning_rate": 1.866666666666667e-05,
      "loss": 0.0762,
      "step": 25750
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.9540577716643741,
      "eval_loss": 0.2068237066268921,
      "eval_runtime": 8.2623,
      "eval_samples_per_second": 439.95,
      "eval_steps_per_second": 55.069,
      "step": 25750
    },
    {
      "epoch": 10.0,
      "learning_rate": 1.8666148867313917e-05,
      "loss": 0.1413,
      "step": 25760
    },
    {
      "epoch": 10.01,
      "learning_rate": 1.8665631067961168e-05,
      "loss": 0.0883,
      "step": 25770
    },
    {
      "epoch": 10.01,
      "learning_rate": 1.8665113268608416e-05,
      "loss": 0.0944,
      "step": 25780
    },
    {
      "epoch": 10.02,
      "learning_rate": 1.8664595469255664e-05,
      "loss": 0.1235,
      "step": 25790
    },
    {
      "epoch": 10.02,
      "learning_rate": 1.8664077669902915e-05,
      "loss": 0.1089,
      "step": 25800
    },
    {
      "epoch": 10.02,
      "learning_rate": 1.8663559870550163e-05,
      "loss": 0.0502,
      "step": 25810
    },
    {
      "epoch": 10.03,
      "learning_rate": 1.866304207119741e-05,
      "loss": 0.2277,
      "step": 25820
    },
    {
      "epoch": 10.03,
      "learning_rate": 1.8662524271844662e-05,
      "loss": 0.1843,
      "step": 25830
    },
    {
      "epoch": 10.03,
      "learning_rate": 1.866200647249191e-05,
      "loss": 0.2591,
      "step": 25840
    },
    {
      "epoch": 10.04,
      "learning_rate": 1.866148867313916e-05,
      "loss": 0.1709,
      "step": 25850
    },
    {
      "epoch": 10.04,
      "learning_rate": 1.866097087378641e-05,
      "loss": 0.1836,
      "step": 25860
    },
    {
      "epoch": 10.05,
      "learning_rate": 1.8660453074433657e-05,
      "loss": 0.196,
      "step": 25870
    },
    {
      "epoch": 10.05,
      "learning_rate": 1.865993527508091e-05,
      "loss": 0.2469,
      "step": 25880
    },
    {
      "epoch": 10.05,
      "learning_rate": 1.8659417475728157e-05,
      "loss": 0.1956,
      "step": 25890
    },
    {
      "epoch": 10.06,
      "learning_rate": 1.8658899676375404e-05,
      "loss": 0.2022,
      "step": 25900
    },
    {
      "epoch": 10.06,
      "learning_rate": 1.8658381877022656e-05,
      "loss": 0.1065,
      "step": 25910
    },
    {
      "epoch": 10.07,
      "learning_rate": 1.8657864077669904e-05,
      "loss": 0.2449,
      "step": 25920
    },
    {
      "epoch": 10.07,
      "learning_rate": 1.8657346278317155e-05,
      "loss": 0.1399,
      "step": 25930
    },
    {
      "epoch": 10.07,
      "learning_rate": 1.8656828478964403e-05,
      "loss": 0.2987,
      "step": 25940
    },
    {
      "epoch": 10.08,
      "learning_rate": 1.865631067961165e-05,
      "loss": 0.0369,
      "step": 25950
    },
    {
      "epoch": 10.08,
      "learning_rate": 1.8655792880258902e-05,
      "loss": 0.1711,
      "step": 25960
    },
    {
      "epoch": 10.09,
      "learning_rate": 1.865527508090615e-05,
      "loss": 0.0085,
      "step": 25970
    },
    {
      "epoch": 10.09,
      "learning_rate": 1.8654757281553398e-05,
      "loss": 0.0986,
      "step": 25980
    },
    {
      "epoch": 10.09,
      "learning_rate": 1.865423948220065e-05,
      "loss": 0.0604,
      "step": 25990
    },
    {
      "epoch": 10.1,
      "learning_rate": 1.8653721682847897e-05,
      "loss": 0.219,
      "step": 26000
    },
    {
      "epoch": 10.1,
      "learning_rate": 1.865320388349515e-05,
      "loss": 0.1227,
      "step": 26010
    },
    {
      "epoch": 10.1,
      "learning_rate": 1.8652686084142396e-05,
      "loss": 0.1623,
      "step": 26020
    },
    {
      "epoch": 10.11,
      "learning_rate": 1.8652168284789644e-05,
      "loss": 0.0108,
      "step": 26030
    },
    {
      "epoch": 10.11,
      "learning_rate": 1.8651650485436896e-05,
      "loss": 0.1783,
      "step": 26040
    },
    {
      "epoch": 10.12,
      "learning_rate": 1.8651132686084144e-05,
      "loss": 0.2796,
      "step": 26050
    },
    {
      "epoch": 10.12,
      "learning_rate": 1.865061488673139e-05,
      "loss": 0.1663,
      "step": 26060
    },
    {
      "epoch": 10.12,
      "learning_rate": 1.8650097087378643e-05,
      "loss": 0.0145,
      "step": 26070
    },
    {
      "epoch": 10.13,
      "learning_rate": 1.864957928802589e-05,
      "loss": 0.0374,
      "step": 26080
    },
    {
      "epoch": 10.13,
      "learning_rate": 1.8649061488673142e-05,
      "loss": 0.1968,
      "step": 26090
    },
    {
      "epoch": 10.14,
      "learning_rate": 1.864854368932039e-05,
      "loss": 0.1045,
      "step": 26100
    },
    {
      "epoch": 10.14,
      "learning_rate": 1.8648025889967638e-05,
      "loss": 0.2047,
      "step": 26110
    },
    {
      "epoch": 10.14,
      "learning_rate": 1.8647508090614886e-05,
      "loss": 0.213,
      "step": 26120
    },
    {
      "epoch": 10.15,
      "learning_rate": 1.8646990291262137e-05,
      "loss": 0.0913,
      "step": 26130
    },
    {
      "epoch": 10.15,
      "learning_rate": 1.8646472491909385e-05,
      "loss": 0.0958,
      "step": 26140
    },
    {
      "epoch": 10.16,
      "learning_rate": 1.8645954692556636e-05,
      "loss": 0.3672,
      "step": 26150
    },
    {
      "epoch": 10.16,
      "learning_rate": 1.8645436893203884e-05,
      "loss": 0.1826,
      "step": 26160
    },
    {
      "epoch": 10.16,
      "learning_rate": 1.8644919093851136e-05,
      "loss": 0.142,
      "step": 26170
    },
    {
      "epoch": 10.17,
      "learning_rate": 1.8644401294498384e-05,
      "loss": 0.1153,
      "step": 26180
    },
    {
      "epoch": 10.17,
      "learning_rate": 1.864388349514563e-05,
      "loss": 0.2934,
      "step": 26190
    },
    {
      "epoch": 10.17,
      "learning_rate": 1.864336569579288e-05,
      "loss": 0.1105,
      "step": 26200
    },
    {
      "epoch": 10.18,
      "learning_rate": 1.864284789644013e-05,
      "loss": 0.2273,
      "step": 26210
    },
    {
      "epoch": 10.18,
      "learning_rate": 1.864233009708738e-05,
      "loss": 0.0234,
      "step": 26220
    },
    {
      "epoch": 10.19,
      "learning_rate": 1.864181229773463e-05,
      "loss": 0.173,
      "step": 26230
    },
    {
      "epoch": 10.19,
      "learning_rate": 1.8641294498381878e-05,
      "loss": 0.0606,
      "step": 26240
    },
    {
      "epoch": 10.19,
      "learning_rate": 1.864077669902913e-05,
      "loss": 0.1138,
      "step": 26250
    },
    {
      "epoch": 10.2,
      "learning_rate": 1.8640258899676377e-05,
      "loss": 0.2025,
      "step": 26260
    },
    {
      "epoch": 10.2,
      "learning_rate": 1.8639741100323625e-05,
      "loss": 0.0531,
      "step": 26270
    },
    {
      "epoch": 10.21,
      "learning_rate": 1.8639223300970873e-05,
      "loss": 0.1708,
      "step": 26280
    },
    {
      "epoch": 10.21,
      "learning_rate": 1.8638705501618124e-05,
      "loss": 0.3425,
      "step": 26290
    },
    {
      "epoch": 10.21,
      "learning_rate": 1.8638187702265372e-05,
      "loss": 0.2293,
      "step": 26300
    },
    {
      "epoch": 10.22,
      "learning_rate": 1.8637669902912624e-05,
      "loss": 0.1665,
      "step": 26310
    },
    {
      "epoch": 10.22,
      "learning_rate": 1.863715210355987e-05,
      "loss": 0.1295,
      "step": 26320
    },
    {
      "epoch": 10.23,
      "learning_rate": 1.8636634304207123e-05,
      "loss": 0.1373,
      "step": 26330
    },
    {
      "epoch": 10.23,
      "learning_rate": 1.863611650485437e-05,
      "loss": 0.268,
      "step": 26340
    },
    {
      "epoch": 10.23,
      "learning_rate": 1.8635598705501622e-05,
      "loss": 0.2208,
      "step": 26350
    },
    {
      "epoch": 10.24,
      "learning_rate": 1.8635080906148867e-05,
      "loss": 0.177,
      "step": 26360
    },
    {
      "epoch": 10.24,
      "learning_rate": 1.8634563106796118e-05,
      "loss": 0.0601,
      "step": 26370
    },
    {
      "epoch": 10.24,
      "learning_rate": 1.8634045307443366e-05,
      "loss": 0.1081,
      "step": 26380
    },
    {
      "epoch": 10.25,
      "learning_rate": 1.8633527508090617e-05,
      "loss": 0.1796,
      "step": 26390
    },
    {
      "epoch": 10.25,
      "learning_rate": 1.8633009708737865e-05,
      "loss": 0.1517,
      "step": 26400
    },
    {
      "epoch": 10.26,
      "learning_rate": 1.8632491909385116e-05,
      "loss": 0.1894,
      "step": 26410
    },
    {
      "epoch": 10.26,
      "learning_rate": 1.8631974110032364e-05,
      "loss": 0.1361,
      "step": 26420
    },
    {
      "epoch": 10.26,
      "learning_rate": 1.8631456310679616e-05,
      "loss": 0.1004,
      "step": 26430
    },
    {
      "epoch": 10.27,
      "learning_rate": 1.863093851132686e-05,
      "loss": 0.2438,
      "step": 26440
    },
    {
      "epoch": 10.27,
      "learning_rate": 1.863042071197411e-05,
      "loss": 0.3972,
      "step": 26450
    },
    {
      "epoch": 10.28,
      "learning_rate": 1.862990291262136e-05,
      "loss": 0.2347,
      "step": 26460
    },
    {
      "epoch": 10.28,
      "learning_rate": 1.862938511326861e-05,
      "loss": 0.1886,
      "step": 26470
    },
    {
      "epoch": 10.28,
      "learning_rate": 1.862886731391586e-05,
      "loss": 0.1277,
      "step": 26480
    },
    {
      "epoch": 10.29,
      "learning_rate": 1.862834951456311e-05,
      "loss": 0.1699,
      "step": 26490
    },
    {
      "epoch": 10.29,
      "learning_rate": 1.8627831715210358e-05,
      "loss": 0.1224,
      "step": 26500
    },
    {
      "epoch": 10.3,
      "learning_rate": 1.862731391585761e-05,
      "loss": 0.1342,
      "step": 26510
    },
    {
      "epoch": 10.3,
      "learning_rate": 1.8626796116504854e-05,
      "loss": 0.0604,
      "step": 26520
    },
    {
      "epoch": 10.3,
      "learning_rate": 1.8626278317152105e-05,
      "loss": 0.0914,
      "step": 26530
    },
    {
      "epoch": 10.31,
      "learning_rate": 1.8625760517799353e-05,
      "loss": 0.1004,
      "step": 26540
    },
    {
      "epoch": 10.31,
      "learning_rate": 1.8625242718446604e-05,
      "loss": 0.044,
      "step": 26550
    },
    {
      "epoch": 10.31,
      "learning_rate": 1.8624724919093852e-05,
      "loss": 0.0796,
      "step": 26560
    },
    {
      "epoch": 10.32,
      "learning_rate": 1.8624207119741103e-05,
      "loss": 0.2106,
      "step": 26570
    },
    {
      "epoch": 10.32,
      "learning_rate": 1.862368932038835e-05,
      "loss": 0.145,
      "step": 26580
    },
    {
      "epoch": 10.33,
      "learning_rate": 1.8623171521035603e-05,
      "loss": 0.0645,
      "step": 26590
    },
    {
      "epoch": 10.33,
      "learning_rate": 1.8622653721682847e-05,
      "loss": 0.0597,
      "step": 26600
    },
    {
      "epoch": 10.33,
      "learning_rate": 1.86221359223301e-05,
      "loss": 0.2596,
      "step": 26610
    },
    {
      "epoch": 10.34,
      "learning_rate": 1.8621618122977346e-05,
      "loss": 0.1288,
      "step": 26620
    },
    {
      "epoch": 10.34,
      "learning_rate": 1.8621100323624598e-05,
      "loss": 0.091,
      "step": 26630
    },
    {
      "epoch": 10.35,
      "learning_rate": 1.8620582524271846e-05,
      "loss": 0.0473,
      "step": 26640
    },
    {
      "epoch": 10.35,
      "learning_rate": 1.8620064724919097e-05,
      "loss": 0.2155,
      "step": 26650
    },
    {
      "epoch": 10.35,
      "learning_rate": 1.8619546925566345e-05,
      "loss": 0.1076,
      "step": 26660
    },
    {
      "epoch": 10.36,
      "learning_rate": 1.8619029126213596e-05,
      "loss": 0.1084,
      "step": 26670
    },
    {
      "epoch": 10.36,
      "learning_rate": 1.861851132686084e-05,
      "loss": 0.1246,
      "step": 26680
    },
    {
      "epoch": 10.37,
      "learning_rate": 1.8617993527508092e-05,
      "loss": 0.1036,
      "step": 26690
    },
    {
      "epoch": 10.37,
      "learning_rate": 1.861747572815534e-05,
      "loss": 0.2292,
      "step": 26700
    },
    {
      "epoch": 10.37,
      "learning_rate": 1.861695792880259e-05,
      "loss": 0.2097,
      "step": 26710
    },
    {
      "epoch": 10.38,
      "learning_rate": 1.861644012944984e-05,
      "loss": 0.1032,
      "step": 26720
    },
    {
      "epoch": 10.38,
      "learning_rate": 1.861592233009709e-05,
      "loss": 0.3145,
      "step": 26730
    },
    {
      "epoch": 10.38,
      "learning_rate": 1.861540453074434e-05,
      "loss": 0.0944,
      "step": 26740
    },
    {
      "epoch": 10.39,
      "learning_rate": 1.861488673139159e-05,
      "loss": 0.0892,
      "step": 26750
    },
    {
      "epoch": 10.39,
      "learning_rate": 1.8614368932038834e-05,
      "loss": 0.2271,
      "step": 26760
    },
    {
      "epoch": 10.4,
      "learning_rate": 1.8613851132686086e-05,
      "loss": 0.1669,
      "step": 26770
    },
    {
      "epoch": 10.4,
      "learning_rate": 1.8613333333333334e-05,
      "loss": 0.2078,
      "step": 26780
    },
    {
      "epoch": 10.4,
      "learning_rate": 1.8612815533980585e-05,
      "loss": 0.2071,
      "step": 26790
    },
    {
      "epoch": 10.41,
      "learning_rate": 1.8612297734627833e-05,
      "loss": 0.187,
      "step": 26800
    },
    {
      "epoch": 10.41,
      "learning_rate": 1.8611779935275084e-05,
      "loss": 0.1229,
      "step": 26810
    },
    {
      "epoch": 10.42,
      "learning_rate": 1.8611262135922332e-05,
      "loss": 0.0654,
      "step": 26820
    },
    {
      "epoch": 10.42,
      "learning_rate": 1.8610744336569583e-05,
      "loss": 0.146,
      "step": 26830
    },
    {
      "epoch": 10.42,
      "learning_rate": 1.8610226537216828e-05,
      "loss": 0.0753,
      "step": 26840
    },
    {
      "epoch": 10.43,
      "learning_rate": 1.860970873786408e-05,
      "loss": 0.0933,
      "step": 26850
    },
    {
      "epoch": 10.43,
      "learning_rate": 1.8609190938511327e-05,
      "loss": 0.3146,
      "step": 26860
    },
    {
      "epoch": 10.43,
      "learning_rate": 1.860867313915858e-05,
      "loss": 0.1426,
      "step": 26870
    },
    {
      "epoch": 10.44,
      "learning_rate": 1.8608155339805826e-05,
      "loss": 0.0857,
      "step": 26880
    },
    {
      "epoch": 10.44,
      "learning_rate": 1.8607637540453078e-05,
      "loss": 0.1185,
      "step": 26890
    },
    {
      "epoch": 10.45,
      "learning_rate": 1.8607119741100326e-05,
      "loss": 0.2082,
      "step": 26900
    },
    {
      "epoch": 10.45,
      "learning_rate": 1.8606601941747574e-05,
      "loss": 0.1496,
      "step": 26910
    },
    {
      "epoch": 10.45,
      "learning_rate": 1.8606084142394825e-05,
      "loss": 0.3489,
      "step": 26920
    },
    {
      "epoch": 10.46,
      "learning_rate": 1.8605566343042073e-05,
      "loss": 0.0337,
      "step": 26930
    },
    {
      "epoch": 10.46,
      "learning_rate": 1.860504854368932e-05,
      "loss": 0.1334,
      "step": 26940
    },
    {
      "epoch": 10.47,
      "learning_rate": 1.8604530744336572e-05,
      "loss": 0.033,
      "step": 26950
    },
    {
      "epoch": 10.47,
      "learning_rate": 1.860401294498382e-05,
      "loss": 0.1618,
      "step": 26960
    },
    {
      "epoch": 10.47,
      "learning_rate": 1.860349514563107e-05,
      "loss": 0.0418,
      "step": 26970
    },
    {
      "epoch": 10.48,
      "learning_rate": 1.860297734627832e-05,
      "loss": 0.1975,
      "step": 26980
    },
    {
      "epoch": 10.48,
      "learning_rate": 1.8602459546925567e-05,
      "loss": 0.1968,
      "step": 26990
    },
    {
      "epoch": 10.49,
      "learning_rate": 1.860194174757282e-05,
      "loss": 0.1373,
      "step": 27000
    },
    {
      "epoch": 10.49,
      "learning_rate": 1.8601423948220066e-05,
      "loss": 0.227,
      "step": 27010
    },
    {
      "epoch": 10.49,
      "learning_rate": 1.8600906148867314e-05,
      "loss": 0.0598,
      "step": 27020
    },
    {
      "epoch": 10.5,
      "learning_rate": 1.8600388349514566e-05,
      "loss": 0.0967,
      "step": 27030
    },
    {
      "epoch": 10.5,
      "learning_rate": 1.8599870550161813e-05,
      "loss": 0.1639,
      "step": 27040
    },
    {
      "epoch": 10.5,
      "learning_rate": 1.8599352750809065e-05,
      "loss": 0.2628,
      "step": 27050
    },
    {
      "epoch": 10.51,
      "learning_rate": 1.8598834951456313e-05,
      "loss": 0.2398,
      "step": 27060
    },
    {
      "epoch": 10.51,
      "learning_rate": 1.859831715210356e-05,
      "loss": 0.1593,
      "step": 27070
    },
    {
      "epoch": 10.52,
      "learning_rate": 1.8597799352750812e-05,
      "loss": 0.0469,
      "step": 27080
    },
    {
      "epoch": 10.52,
      "learning_rate": 1.859728155339806e-05,
      "loss": 0.1226,
      "step": 27090
    },
    {
      "epoch": 10.52,
      "learning_rate": 1.8596763754045308e-05,
      "loss": 0.2323,
      "step": 27100
    },
    {
      "epoch": 10.53,
      "learning_rate": 1.859624595469256e-05,
      "loss": 0.1481,
      "step": 27110
    },
    {
      "epoch": 10.53,
      "learning_rate": 1.8595728155339807e-05,
      "loss": 0.2359,
      "step": 27120
    },
    {
      "epoch": 10.54,
      "learning_rate": 1.859521035598706e-05,
      "loss": 0.1212,
      "step": 27130
    },
    {
      "epoch": 10.54,
      "learning_rate": 1.8594692556634306e-05,
      "loss": 0.5169,
      "step": 27140
    },
    {
      "epoch": 10.54,
      "learning_rate": 1.8594174757281554e-05,
      "loss": 0.1984,
      "step": 27150
    },
    {
      "epoch": 10.55,
      "learning_rate": 1.8593656957928805e-05,
      "loss": 0.043,
      "step": 27160
    },
    {
      "epoch": 10.55,
      "learning_rate": 1.8593139158576053e-05,
      "loss": 0.0485,
      "step": 27170
    },
    {
      "epoch": 10.56,
      "learning_rate": 1.85926213592233e-05,
      "loss": 0.108,
      "step": 27180
    },
    {
      "epoch": 10.56,
      "learning_rate": 1.8592103559870553e-05,
      "loss": 0.3133,
      "step": 27190
    },
    {
      "epoch": 10.56,
      "learning_rate": 1.85915857605178e-05,
      "loss": 0.4284,
      "step": 27200
    },
    {
      "epoch": 10.57,
      "learning_rate": 1.859106796116505e-05,
      "loss": 0.094,
      "step": 27210
    },
    {
      "epoch": 10.57,
      "learning_rate": 1.85905501618123e-05,
      "loss": 0.2154,
      "step": 27220
    },
    {
      "epoch": 10.57,
      "learning_rate": 1.8590032362459548e-05,
      "loss": 0.1836,
      "step": 27230
    },
    {
      "epoch": 10.58,
      "learning_rate": 1.85895145631068e-05,
      "loss": 0.2227,
      "step": 27240
    },
    {
      "epoch": 10.58,
      "learning_rate": 1.8588996763754047e-05,
      "loss": 0.0724,
      "step": 27250
    },
    {
      "epoch": 10.59,
      "learning_rate": 1.8588478964401295e-05,
      "loss": 0.0855,
      "step": 27260
    },
    {
      "epoch": 10.59,
      "learning_rate": 1.8587961165048546e-05,
      "loss": 0.105,
      "step": 27270
    },
    {
      "epoch": 10.59,
      "learning_rate": 1.8587443365695794e-05,
      "loss": 0.1321,
      "step": 27280
    },
    {
      "epoch": 10.6,
      "learning_rate": 1.8586925566343042e-05,
      "loss": 0.0996,
      "step": 27290
    },
    {
      "epoch": 10.6,
      "learning_rate": 1.8586407766990293e-05,
      "loss": 0.071,
      "step": 27300
    },
    {
      "epoch": 10.61,
      "learning_rate": 1.858588996763754e-05,
      "loss": 0.291,
      "step": 27310
    },
    {
      "epoch": 10.61,
      "learning_rate": 1.8585372168284793e-05,
      "loss": 0.1029,
      "step": 27320
    },
    {
      "epoch": 10.61,
      "learning_rate": 1.858485436893204e-05,
      "loss": 0.2579,
      "step": 27330
    },
    {
      "epoch": 10.62,
      "learning_rate": 1.858433656957929e-05,
      "loss": 0.0925,
      "step": 27340
    },
    {
      "epoch": 10.62,
      "learning_rate": 1.858381877022654e-05,
      "loss": 0.0832,
      "step": 27350
    },
    {
      "epoch": 10.63,
      "learning_rate": 1.8583300970873788e-05,
      "loss": 0.0713,
      "step": 27360
    },
    {
      "epoch": 10.63,
      "learning_rate": 1.8582783171521036e-05,
      "loss": 0.0871,
      "step": 27370
    },
    {
      "epoch": 10.63,
      "learning_rate": 1.8582265372168287e-05,
      "loss": 0.1462,
      "step": 27380
    },
    {
      "epoch": 10.64,
      "learning_rate": 1.8581747572815535e-05,
      "loss": 0.1943,
      "step": 27390
    },
    {
      "epoch": 10.64,
      "learning_rate": 1.8581229773462786e-05,
      "loss": 0.0648,
      "step": 27400
    },
    {
      "epoch": 10.64,
      "learning_rate": 1.8580711974110034e-05,
      "loss": 0.0124,
      "step": 27410
    },
    {
      "epoch": 10.65,
      "learning_rate": 1.8580194174757282e-05,
      "loss": 0.1284,
      "step": 27420
    },
    {
      "epoch": 10.65,
      "learning_rate": 1.8579676375404533e-05,
      "loss": 0.0942,
      "step": 27430
    },
    {
      "epoch": 10.66,
      "learning_rate": 1.857915857605178e-05,
      "loss": 0.1293,
      "step": 27440
    },
    {
      "epoch": 10.66,
      "learning_rate": 1.857864077669903e-05,
      "loss": 0.1612,
      "step": 27450
    },
    {
      "epoch": 10.66,
      "learning_rate": 1.857812297734628e-05,
      "loss": 0.1978,
      "step": 27460
    },
    {
      "epoch": 10.67,
      "learning_rate": 1.857760517799353e-05,
      "loss": 0.1312,
      "step": 27470
    },
    {
      "epoch": 10.67,
      "learning_rate": 1.857708737864078e-05,
      "loss": 0.0827,
      "step": 27480
    },
    {
      "epoch": 10.68,
      "learning_rate": 1.8576569579288028e-05,
      "loss": 0.2018,
      "step": 27490
    },
    {
      "epoch": 10.68,
      "learning_rate": 1.8576051779935276e-05,
      "loss": 0.0057,
      "step": 27500
    },
    {
      "epoch": 10.68,
      "learning_rate": 1.8575533980582527e-05,
      "loss": 0.466,
      "step": 27510
    },
    {
      "epoch": 10.69,
      "learning_rate": 1.8575016181229775e-05,
      "loss": 0.1091,
      "step": 27520
    },
    {
      "epoch": 10.69,
      "learning_rate": 1.8574498381877023e-05,
      "loss": 0.1364,
      "step": 27530
    },
    {
      "epoch": 10.7,
      "learning_rate": 1.8573980582524274e-05,
      "loss": 0.1108,
      "step": 27540
    },
    {
      "epoch": 10.7,
      "learning_rate": 1.8573462783171522e-05,
      "loss": 0.1916,
      "step": 27550
    },
    {
      "epoch": 10.7,
      "learning_rate": 1.8572944983818773e-05,
      "loss": 0.027,
      "step": 27560
    },
    {
      "epoch": 10.71,
      "learning_rate": 1.857242718446602e-05,
      "loss": 0.1526,
      "step": 27570
    },
    {
      "epoch": 10.71,
      "learning_rate": 1.857190938511327e-05,
      "loss": 0.1889,
      "step": 27580
    },
    {
      "epoch": 10.71,
      "learning_rate": 1.8571391585760517e-05,
      "loss": 0.1239,
      "step": 27590
    },
    {
      "epoch": 10.72,
      "learning_rate": 1.857087378640777e-05,
      "loss": 0.0917,
      "step": 27600
    },
    {
      "epoch": 10.72,
      "learning_rate": 1.8570355987055016e-05,
      "loss": 0.0597,
      "step": 27610
    },
    {
      "epoch": 10.73,
      "learning_rate": 1.8569838187702268e-05,
      "loss": 0.2808,
      "step": 27620
    },
    {
      "epoch": 10.73,
      "learning_rate": 1.8569320388349516e-05,
      "loss": 0.2267,
      "step": 27630
    },
    {
      "epoch": 10.73,
      "learning_rate": 1.8568802588996767e-05,
      "loss": 0.2857,
      "step": 27640
    },
    {
      "epoch": 10.74,
      "learning_rate": 1.8568284789644015e-05,
      "loss": 0.1459,
      "step": 27650
    },
    {
      "epoch": 10.74,
      "learning_rate": 1.8567766990291263e-05,
      "loss": 0.1205,
      "step": 27660
    },
    {
      "epoch": 10.75,
      "learning_rate": 1.856724919093851e-05,
      "loss": 0.1098,
      "step": 27670
    },
    {
      "epoch": 10.75,
      "learning_rate": 1.8566731391585762e-05,
      "loss": 0.1632,
      "step": 27680
    },
    {
      "epoch": 10.75,
      "learning_rate": 1.856621359223301e-05,
      "loss": 0.2184,
      "step": 27690
    },
    {
      "epoch": 10.76,
      "learning_rate": 1.856569579288026e-05,
      "loss": 0.1218,
      "step": 27700
    },
    {
      "epoch": 10.76,
      "learning_rate": 1.856517799352751e-05,
      "loss": 0.1731,
      "step": 27710
    },
    {
      "epoch": 10.77,
      "learning_rate": 1.856466019417476e-05,
      "loss": 0.0595,
      "step": 27720
    },
    {
      "epoch": 10.77,
      "learning_rate": 1.856414239482201e-05,
      "loss": 0.2491,
      "step": 27730
    },
    {
      "epoch": 10.77,
      "learning_rate": 1.8563624595469256e-05,
      "loss": 0.1026,
      "step": 27740
    },
    {
      "epoch": 10.78,
      "learning_rate": 1.8563106796116504e-05,
      "loss": 0.1624,
      "step": 27750
    },
    {
      "epoch": 10.78,
      "learning_rate": 1.8562588996763755e-05,
      "loss": 0.1069,
      "step": 27760
    },
    {
      "epoch": 10.78,
      "learning_rate": 1.8562071197411003e-05,
      "loss": 0.1931,
      "step": 27770
    },
    {
      "epoch": 10.79,
      "learning_rate": 1.8561553398058255e-05,
      "loss": 0.1642,
      "step": 27780
    },
    {
      "epoch": 10.79,
      "learning_rate": 1.8561035598705503e-05,
      "loss": 0.1252,
      "step": 27790
    },
    {
      "epoch": 10.8,
      "learning_rate": 1.8560517799352754e-05,
      "loss": 0.274,
      "step": 27800
    },
    {
      "epoch": 10.8,
      "learning_rate": 1.8560000000000002e-05,
      "loss": 0.0973,
      "step": 27810
    },
    {
      "epoch": 10.8,
      "learning_rate": 1.855948220064725e-05,
      "loss": 0.1297,
      "step": 27820
    },
    {
      "epoch": 10.81,
      "learning_rate": 1.8558964401294498e-05,
      "loss": 0.1829,
      "step": 27830
    },
    {
      "epoch": 10.81,
      "learning_rate": 1.855844660194175e-05,
      "loss": 0.1329,
      "step": 27840
    },
    {
      "epoch": 10.82,
      "learning_rate": 1.8557928802588997e-05,
      "loss": 0.2081,
      "step": 27850
    },
    {
      "epoch": 10.82,
      "learning_rate": 1.8557411003236248e-05,
      "loss": 0.1544,
      "step": 27860
    },
    {
      "epoch": 10.82,
      "learning_rate": 1.8556893203883496e-05,
      "loss": 0.196,
      "step": 27870
    },
    {
      "epoch": 10.83,
      "learning_rate": 1.8556375404530748e-05,
      "loss": 0.0522,
      "step": 27880
    },
    {
      "epoch": 10.83,
      "learning_rate": 1.8555857605177995e-05,
      "loss": 0.1635,
      "step": 27890
    },
    {
      "epoch": 10.83,
      "learning_rate": 1.8555339805825243e-05,
      "loss": 0.126,
      "step": 27900
    },
    {
      "epoch": 10.84,
      "learning_rate": 1.855482200647249e-05,
      "loss": 0.1632,
      "step": 27910
    },
    {
      "epoch": 10.84,
      "learning_rate": 1.8554304207119743e-05,
      "loss": 0.1862,
      "step": 27920
    },
    {
      "epoch": 10.85,
      "learning_rate": 1.855378640776699e-05,
      "loss": 0.0555,
      "step": 27930
    },
    {
      "epoch": 10.85,
      "learning_rate": 1.8553268608414242e-05,
      "loss": 0.0298,
      "step": 27940
    },
    {
      "epoch": 10.85,
      "learning_rate": 1.855275080906149e-05,
      "loss": 0.2461,
      "step": 27950
    },
    {
      "epoch": 10.86,
      "learning_rate": 1.855223300970874e-05,
      "loss": 0.1963,
      "step": 27960
    },
    {
      "epoch": 10.86,
      "learning_rate": 1.855171521035599e-05,
      "loss": 0.1222,
      "step": 27970
    },
    {
      "epoch": 10.87,
      "learning_rate": 1.8551197411003237e-05,
      "loss": 0.0932,
      "step": 27980
    },
    {
      "epoch": 10.87,
      "learning_rate": 1.8550679611650485e-05,
      "loss": 0.0844,
      "step": 27990
    },
    {
      "epoch": 10.87,
      "learning_rate": 1.8550161812297736e-05,
      "loss": 0.0896,
      "step": 28000
    },
    {
      "epoch": 10.88,
      "learning_rate": 1.8549644012944984e-05,
      "loss": 0.1786,
      "step": 28010
    },
    {
      "epoch": 10.88,
      "learning_rate": 1.8549126213592235e-05,
      "loss": 0.3595,
      "step": 28020
    },
    {
      "epoch": 10.89,
      "learning_rate": 1.8548608414239483e-05,
      "loss": 0.2076,
      "step": 28030
    },
    {
      "epoch": 10.89,
      "learning_rate": 1.8548090614886735e-05,
      "loss": 0.094,
      "step": 28040
    },
    {
      "epoch": 10.89,
      "learning_rate": 1.8547572815533983e-05,
      "loss": 0.1093,
      "step": 28050
    },
    {
      "epoch": 10.9,
      "learning_rate": 1.8547055016181234e-05,
      "loss": 0.2313,
      "step": 28060
    },
    {
      "epoch": 10.9,
      "learning_rate": 1.854653721682848e-05,
      "loss": 0.1294,
      "step": 28070
    },
    {
      "epoch": 10.9,
      "learning_rate": 1.854601941747573e-05,
      "loss": 0.1969,
      "step": 28080
    },
    {
      "epoch": 10.91,
      "learning_rate": 1.8545501618122978e-05,
      "loss": 0.0361,
      "step": 28090
    },
    {
      "epoch": 10.91,
      "learning_rate": 1.854498381877023e-05,
      "loss": 0.0569,
      "step": 28100
    },
    {
      "epoch": 10.92,
      "learning_rate": 1.8544466019417477e-05,
      "loss": 0.1107,
      "step": 28110
    },
    {
      "epoch": 10.92,
      "learning_rate": 1.8543948220064728e-05,
      "loss": 0.0683,
      "step": 28120
    },
    {
      "epoch": 10.92,
      "learning_rate": 1.8543430420711976e-05,
      "loss": 0.0947,
      "step": 28130
    },
    {
      "epoch": 10.93,
      "learning_rate": 1.8542912621359227e-05,
      "loss": 0.0736,
      "step": 28140
    },
    {
      "epoch": 10.93,
      "learning_rate": 1.8542394822006472e-05,
      "loss": 0.0343,
      "step": 28150
    },
    {
      "epoch": 10.94,
      "learning_rate": 1.8541877022653723e-05,
      "loss": 0.2855,
      "step": 28160
    },
    {
      "epoch": 10.94,
      "learning_rate": 1.854135922330097e-05,
      "loss": 0.1257,
      "step": 28170
    },
    {
      "epoch": 10.94,
      "learning_rate": 1.8540841423948222e-05,
      "loss": 0.2852,
      "step": 28180
    },
    {
      "epoch": 10.95,
      "learning_rate": 1.854032362459547e-05,
      "loss": 0.2797,
      "step": 28190
    },
    {
      "epoch": 10.95,
      "learning_rate": 1.8539805825242722e-05,
      "loss": 0.1107,
      "step": 28200
    },
    {
      "epoch": 10.96,
      "learning_rate": 1.853928802588997e-05,
      "loss": 0.1122,
      "step": 28210
    },
    {
      "epoch": 10.96,
      "learning_rate": 1.853877022653722e-05,
      "loss": 0.1052,
      "step": 28220
    },
    {
      "epoch": 10.96,
      "learning_rate": 1.8538252427184465e-05,
      "loss": 0.0379,
      "step": 28230
    },
    {
      "epoch": 10.97,
      "learning_rate": 1.8537734627831717e-05,
      "loss": 0.2837,
      "step": 28240
    },
    {
      "epoch": 10.97,
      "learning_rate": 1.8537216828478965e-05,
      "loss": 0.1936,
      "step": 28250
    },
    {
      "epoch": 10.97,
      "learning_rate": 1.8536699029126216e-05,
      "loss": 0.3791,
      "step": 28260
    },
    {
      "epoch": 10.98,
      "learning_rate": 1.8536181229773464e-05,
      "loss": 0.2145,
      "step": 28270
    },
    {
      "epoch": 10.98,
      "learning_rate": 1.8535663430420715e-05,
      "loss": 0.0559,
      "step": 28280
    },
    {
      "epoch": 10.99,
      "learning_rate": 1.8535145631067963e-05,
      "loss": 0.0722,
      "step": 28290
    },
    {
      "epoch": 10.99,
      "learning_rate": 1.8534627831715215e-05,
      "loss": 0.2436,
      "step": 28300
    },
    {
      "epoch": 10.99,
      "learning_rate": 1.853411003236246e-05,
      "loss": 0.1536,
      "step": 28310
    },
    {
      "epoch": 11.0,
      "learning_rate": 1.853359223300971e-05,
      "loss": 0.127,
      "step": 28320
    },
    {
      "epoch": 11.0,
      "eval_accuracy": 0.9543328748280605,
      "eval_loss": 0.19855087995529175,
      "eval_runtime": 8.2681,
      "eval_samples_per_second": 439.642,
      "eval_steps_per_second": 55.031,
      "step": 28325
    },
    {
      "epoch": 11.0,
      "learning_rate": 1.8533074433656958e-05,
      "loss": 0.2633,
      "step": 28330
    },
    {
      "epoch": 11.01,
      "learning_rate": 1.853255663430421e-05,
      "loss": 0.119,
      "step": 28340
    },
    {
      "epoch": 11.01,
      "learning_rate": 1.8532038834951458e-05,
      "loss": 0.2158,
      "step": 28350
    },
    {
      "epoch": 11.01,
      "learning_rate": 1.853152103559871e-05,
      "loss": 0.0274,
      "step": 28360
    },
    {
      "epoch": 11.02,
      "learning_rate": 1.8531003236245957e-05,
      "loss": 0.2392,
      "step": 28370
    },
    {
      "epoch": 11.02,
      "learning_rate": 1.8530485436893205e-05,
      "loss": 0.2463,
      "step": 28380
    },
    {
      "epoch": 11.03,
      "learning_rate": 1.8529967637540453e-05,
      "loss": 0.1769,
      "step": 28390
    },
    {
      "epoch": 11.03,
      "learning_rate": 1.8529449838187704e-05,
      "loss": 0.0163,
      "step": 28400
    },
    {
      "epoch": 11.03,
      "learning_rate": 1.8528932038834952e-05,
      "loss": 0.2622,
      "step": 28410
    },
    {
      "epoch": 11.04,
      "learning_rate": 1.8528414239482203e-05,
      "loss": 0.1987,
      "step": 28420
    },
    {
      "epoch": 11.04,
      "learning_rate": 1.852789644012945e-05,
      "loss": 0.0565,
      "step": 28430
    },
    {
      "epoch": 11.04,
      "learning_rate": 1.8527378640776702e-05,
      "loss": 0.1533,
      "step": 28440
    },
    {
      "epoch": 11.05,
      "learning_rate": 1.852686084142395e-05,
      "loss": 0.0539,
      "step": 28450
    },
    {
      "epoch": 11.05,
      "learning_rate": 1.8526343042071198e-05,
      "loss": 0.1568,
      "step": 28460
    },
    {
      "epoch": 11.06,
      "learning_rate": 1.8525825242718446e-05,
      "loss": 0.1002,
      "step": 28470
    },
    {
      "epoch": 11.06,
      "learning_rate": 1.8525307443365697e-05,
      "loss": 0.0666,
      "step": 28480
    },
    {
      "epoch": 11.06,
      "learning_rate": 1.8524789644012945e-05,
      "loss": 0.2074,
      "step": 28490
    },
    {
      "epoch": 11.07,
      "learning_rate": 1.8524271844660197e-05,
      "loss": 0.1428,
      "step": 28500
    },
    {
      "epoch": 11.07,
      "learning_rate": 1.8523754045307445e-05,
      "loss": 0.1215,
      "step": 28510
    },
    {
      "epoch": 11.08,
      "learning_rate": 1.8523236245954696e-05,
      "loss": 0.1956,
      "step": 28520
    },
    {
      "epoch": 11.08,
      "learning_rate": 1.8522718446601944e-05,
      "loss": 0.0998,
      "step": 28530
    },
    {
      "epoch": 11.08,
      "learning_rate": 1.8522200647249192e-05,
      "loss": 0.083,
      "step": 28540
    },
    {
      "epoch": 11.09,
      "learning_rate": 1.852168284789644e-05,
      "loss": 0.2017,
      "step": 28550
    },
    {
      "epoch": 11.09,
      "learning_rate": 1.852116504854369e-05,
      "loss": 0.131,
      "step": 28560
    },
    {
      "epoch": 11.1,
      "learning_rate": 1.852064724919094e-05,
      "loss": 0.115,
      "step": 28570
    },
    {
      "epoch": 11.1,
      "learning_rate": 1.852012944983819e-05,
      "loss": 0.0231,
      "step": 28580
    },
    {
      "epoch": 11.1,
      "learning_rate": 1.8519611650485438e-05,
      "loss": 0.0994,
      "step": 28590
    },
    {
      "epoch": 11.11,
      "learning_rate": 1.851909385113269e-05,
      "loss": 0.1065,
      "step": 28600
    },
    {
      "epoch": 11.11,
      "learning_rate": 1.8518576051779937e-05,
      "loss": 0.1115,
      "step": 28610
    },
    {
      "epoch": 11.11,
      "learning_rate": 1.8518058252427185e-05,
      "loss": 0.1223,
      "step": 28620
    },
    {
      "epoch": 11.12,
      "learning_rate": 1.8517540453074437e-05,
      "loss": 0.0024,
      "step": 28630
    },
    {
      "epoch": 11.12,
      "learning_rate": 1.8517022653721685e-05,
      "loss": 0.1297,
      "step": 28640
    },
    {
      "epoch": 11.13,
      "learning_rate": 1.8516504854368933e-05,
      "loss": 0.0852,
      "step": 28650
    },
    {
      "epoch": 11.13,
      "learning_rate": 1.8515987055016184e-05,
      "loss": 0.1143,
      "step": 28660
    },
    {
      "epoch": 11.13,
      "learning_rate": 1.8515469255663432e-05,
      "loss": 0.2768,
      "step": 28670
    },
    {
      "epoch": 11.14,
      "learning_rate": 1.851495145631068e-05,
      "loss": 0.1342,
      "step": 28680
    },
    {
      "epoch": 11.14,
      "learning_rate": 1.851443365695793e-05,
      "loss": 0.055,
      "step": 28690
    },
    {
      "epoch": 11.15,
      "learning_rate": 1.851391585760518e-05,
      "loss": 0.1669,
      "step": 28700
    },
    {
      "epoch": 11.15,
      "learning_rate": 1.851339805825243e-05,
      "loss": 0.0961,
      "step": 28710
    },
    {
      "epoch": 11.15,
      "learning_rate": 1.8512880258899678e-05,
      "loss": 0.2785,
      "step": 28720
    },
    {
      "epoch": 11.16,
      "learning_rate": 1.8512362459546926e-05,
      "loss": 0.027,
      "step": 28730
    },
    {
      "epoch": 11.16,
      "learning_rate": 1.8511844660194177e-05,
      "loss": 0.1963,
      "step": 28740
    },
    {
      "epoch": 11.17,
      "learning_rate": 1.8511326860841425e-05,
      "loss": 0.0683,
      "step": 28750
    },
    {
      "epoch": 11.17,
      "learning_rate": 1.8510809061488673e-05,
      "loss": 0.4444,
      "step": 28760
    },
    {
      "epoch": 11.17,
      "learning_rate": 1.8510291262135925e-05,
      "loss": 0.0804,
      "step": 28770
    },
    {
      "epoch": 11.18,
      "learning_rate": 1.8509773462783172e-05,
      "loss": 0.0778,
      "step": 28780
    },
    {
      "epoch": 11.18,
      "learning_rate": 1.8509255663430424e-05,
      "loss": 0.1115,
      "step": 28790
    },
    {
      "epoch": 11.18,
      "learning_rate": 1.8508737864077672e-05,
      "loss": 0.2214,
      "step": 28800
    },
    {
      "epoch": 11.19,
      "learning_rate": 1.850822006472492e-05,
      "loss": 0.0614,
      "step": 28810
    },
    {
      "epoch": 11.19,
      "learning_rate": 1.850770226537217e-05,
      "loss": 0.2345,
      "step": 28820
    },
    {
      "epoch": 11.2,
      "learning_rate": 1.850718446601942e-05,
      "loss": 0.2467,
      "step": 28830
    },
    {
      "epoch": 11.2,
      "learning_rate": 1.8506666666666667e-05,
      "loss": 0.0686,
      "step": 28840
    },
    {
      "epoch": 11.2,
      "learning_rate": 1.8506148867313918e-05,
      "loss": 0.3031,
      "step": 28850
    },
    {
      "epoch": 11.21,
      "learning_rate": 1.8505631067961166e-05,
      "loss": 0.172,
      "step": 28860
    },
    {
      "epoch": 11.21,
      "learning_rate": 1.8505113268608417e-05,
      "loss": 0.4145,
      "step": 28870
    },
    {
      "epoch": 11.22,
      "learning_rate": 1.8504595469255665e-05,
      "loss": 0.0354,
      "step": 28880
    },
    {
      "epoch": 11.22,
      "learning_rate": 1.8504077669902913e-05,
      "loss": 0.0488,
      "step": 28890
    },
    {
      "epoch": 11.22,
      "learning_rate": 1.8503559870550164e-05,
      "loss": 0.173,
      "step": 28900
    },
    {
      "epoch": 11.23,
      "learning_rate": 1.8503042071197412e-05,
      "loss": 0.063,
      "step": 28910
    },
    {
      "epoch": 11.23,
      "learning_rate": 1.850252427184466e-05,
      "loss": 0.0848,
      "step": 28920
    },
    {
      "epoch": 11.23,
      "learning_rate": 1.850200647249191e-05,
      "loss": 0.1574,
      "step": 28930
    },
    {
      "epoch": 11.24,
      "learning_rate": 1.850148867313916e-05,
      "loss": 0.205,
      "step": 28940
    },
    {
      "epoch": 11.24,
      "learning_rate": 1.850097087378641e-05,
      "loss": 0.0258,
      "step": 28950
    },
    {
      "epoch": 11.25,
      "learning_rate": 1.850045307443366e-05,
      "loss": 0.1676,
      "step": 28960
    },
    {
      "epoch": 11.25,
      "learning_rate": 1.8499935275080907e-05,
      "loss": 0.094,
      "step": 28970
    },
    {
      "epoch": 11.25,
      "learning_rate": 1.8499417475728158e-05,
      "loss": 0.2293,
      "step": 28980
    },
    {
      "epoch": 11.26,
      "learning_rate": 1.8498899676375406e-05,
      "loss": 0.2111,
      "step": 28990
    },
    {
      "epoch": 11.26,
      "learning_rate": 1.8498381877022654e-05,
      "loss": 0.1803,
      "step": 29000
    },
    {
      "epoch": 11.27,
      "learning_rate": 1.8497864077669905e-05,
      "loss": 0.2255,
      "step": 29010
    },
    {
      "epoch": 11.27,
      "learning_rate": 1.8497346278317153e-05,
      "loss": 0.1102,
      "step": 29020
    },
    {
      "epoch": 11.27,
      "learning_rate": 1.8496828478964404e-05,
      "loss": 0.1347,
      "step": 29030
    },
    {
      "epoch": 11.28,
      "learning_rate": 1.8496310679611652e-05,
      "loss": 0.0068,
      "step": 29040
    },
    {
      "epoch": 11.28,
      "learning_rate": 1.84957928802589e-05,
      "loss": 0.0684,
      "step": 29050
    },
    {
      "epoch": 11.29,
      "learning_rate": 1.8495275080906148e-05,
      "loss": 0.0341,
      "step": 29060
    },
    {
      "epoch": 11.29,
      "learning_rate": 1.84947572815534e-05,
      "loss": 0.1484,
      "step": 29070
    },
    {
      "epoch": 11.29,
      "learning_rate": 1.8494239482200647e-05,
      "loss": 0.2158,
      "step": 29080
    },
    {
      "epoch": 11.3,
      "learning_rate": 1.84937216828479e-05,
      "loss": 0.1737,
      "step": 29090
    },
    {
      "epoch": 11.3,
      "learning_rate": 1.8493203883495147e-05,
      "loss": 0.1307,
      "step": 29100
    },
    {
      "epoch": 11.3,
      "learning_rate": 1.8492686084142398e-05,
      "loss": 0.0873,
      "step": 29110
    },
    {
      "epoch": 11.31,
      "learning_rate": 1.8492168284789646e-05,
      "loss": 0.268,
      "step": 29120
    },
    {
      "epoch": 11.31,
      "learning_rate": 1.8491650485436894e-05,
      "loss": 0.1517,
      "step": 29130
    },
    {
      "epoch": 11.32,
      "learning_rate": 1.8491132686084142e-05,
      "loss": 0.1677,
      "step": 29140
    },
    {
      "epoch": 11.32,
      "learning_rate": 1.8490614886731393e-05,
      "loss": 0.0447,
      "step": 29150
    },
    {
      "epoch": 11.32,
      "learning_rate": 1.849009708737864e-05,
      "loss": 0.0773,
      "step": 29160
    },
    {
      "epoch": 11.33,
      "learning_rate": 1.8489579288025892e-05,
      "loss": 0.0748,
      "step": 29170
    },
    {
      "epoch": 11.33,
      "learning_rate": 1.848906148867314e-05,
      "loss": 0.157,
      "step": 29180
    },
    {
      "epoch": 11.34,
      "learning_rate": 1.848854368932039e-05,
      "loss": 0.1562,
      "step": 29190
    },
    {
      "epoch": 11.34,
      "learning_rate": 1.848802588996764e-05,
      "loss": 0.2562,
      "step": 29200
    },
    {
      "epoch": 11.34,
      "learning_rate": 1.8487508090614887e-05,
      "loss": 0.2108,
      "step": 29210
    },
    {
      "epoch": 11.35,
      "learning_rate": 1.8486990291262135e-05,
      "loss": 0.0312,
      "step": 29220
    },
    {
      "epoch": 11.35,
      "learning_rate": 1.8486472491909387e-05,
      "loss": 0.5272,
      "step": 29230
    },
    {
      "epoch": 11.36,
      "learning_rate": 1.8485954692556635e-05,
      "loss": 0.2136,
      "step": 29240
    },
    {
      "epoch": 11.36,
      "learning_rate": 1.8485436893203886e-05,
      "loss": 0.1327,
      "step": 29250
    },
    {
      "epoch": 11.36,
      "learning_rate": 1.8484919093851134e-05,
      "loss": 0.0731,
      "step": 29260
    },
    {
      "epoch": 11.37,
      "learning_rate": 1.8484401294498385e-05,
      "loss": 0.1238,
      "step": 29270
    },
    {
      "epoch": 11.37,
      "learning_rate": 1.8483883495145633e-05,
      "loss": 0.0453,
      "step": 29280
    },
    {
      "epoch": 11.37,
      "learning_rate": 1.848336569579288e-05,
      "loss": 0.2308,
      "step": 29290
    },
    {
      "epoch": 11.38,
      "learning_rate": 1.848284789644013e-05,
      "loss": 0.1713,
      "step": 29300
    },
    {
      "epoch": 11.38,
      "learning_rate": 1.848233009708738e-05,
      "loss": 0.1046,
      "step": 29310
    },
    {
      "epoch": 11.39,
      "learning_rate": 1.8481812297734628e-05,
      "loss": 0.1481,
      "step": 29320
    },
    {
      "epoch": 11.39,
      "learning_rate": 1.848129449838188e-05,
      "loss": 0.0161,
      "step": 29330
    },
    {
      "epoch": 11.39,
      "learning_rate": 1.8480776699029127e-05,
      "loss": 0.1971,
      "step": 29340
    },
    {
      "epoch": 11.4,
      "learning_rate": 1.848025889967638e-05,
      "loss": 0.0982,
      "step": 29350
    },
    {
      "epoch": 11.4,
      "learning_rate": 1.8479741100323627e-05,
      "loss": 0.2388,
      "step": 29360
    },
    {
      "epoch": 11.41,
      "learning_rate": 1.8479223300970875e-05,
      "loss": 0.1016,
      "step": 29370
    },
    {
      "epoch": 11.41,
      "learning_rate": 1.8478705501618122e-05,
      "loss": 0.0556,
      "step": 29380
    },
    {
      "epoch": 11.41,
      "learning_rate": 1.8478187702265374e-05,
      "loss": 0.063,
      "step": 29390
    },
    {
      "epoch": 11.42,
      "learning_rate": 1.847766990291262e-05,
      "loss": 0.1998,
      "step": 29400
    },
    {
      "epoch": 11.42,
      "learning_rate": 1.8477152103559873e-05,
      "loss": 0.1542,
      "step": 29410
    },
    {
      "epoch": 11.43,
      "learning_rate": 1.847663430420712e-05,
      "loss": 0.1173,
      "step": 29420
    },
    {
      "epoch": 11.43,
      "learning_rate": 1.8476116504854372e-05,
      "loss": 0.0643,
      "step": 29430
    },
    {
      "epoch": 11.43,
      "learning_rate": 1.847559870550162e-05,
      "loss": 0.0344,
      "step": 29440
    },
    {
      "epoch": 11.44,
      "learning_rate": 1.8475080906148868e-05,
      "loss": 0.1412,
      "step": 29450
    },
    {
      "epoch": 11.44,
      "learning_rate": 1.8474563106796116e-05,
      "loss": 0.1518,
      "step": 29460
    },
    {
      "epoch": 11.44,
      "learning_rate": 1.8474045307443367e-05,
      "loss": 0.037,
      "step": 29470
    },
    {
      "epoch": 11.45,
      "learning_rate": 1.8473527508090615e-05,
      "loss": 0.2464,
      "step": 29480
    },
    {
      "epoch": 11.45,
      "learning_rate": 1.8473009708737867e-05,
      "loss": 0.2612,
      "step": 29490
    },
    {
      "epoch": 11.46,
      "learning_rate": 1.8472491909385114e-05,
      "loss": 0.0337,
      "step": 29500
    },
    {
      "epoch": 11.46,
      "learning_rate": 1.8471974110032366e-05,
      "loss": 0.0956,
      "step": 29510
    },
    {
      "epoch": 11.46,
      "learning_rate": 1.8471456310679614e-05,
      "loss": 0.2567,
      "step": 29520
    },
    {
      "epoch": 11.47,
      "learning_rate": 1.847093851132686e-05,
      "loss": 0.1401,
      "step": 29530
    },
    {
      "epoch": 11.47,
      "learning_rate": 1.847042071197411e-05,
      "loss": 0.1877,
      "step": 29540
    },
    {
      "epoch": 11.48,
      "learning_rate": 1.846990291262136e-05,
      "loss": 0.2209,
      "step": 29550
    },
    {
      "epoch": 11.48,
      "learning_rate": 1.846938511326861e-05,
      "loss": 0.1312,
      "step": 29560
    },
    {
      "epoch": 11.48,
      "learning_rate": 1.846886731391586e-05,
      "loss": 0.1419,
      "step": 29570
    },
    {
      "epoch": 11.49,
      "learning_rate": 1.8468349514563108e-05,
      "loss": 0.1617,
      "step": 29580
    },
    {
      "epoch": 11.49,
      "learning_rate": 1.846783171521036e-05,
      "loss": 0.1461,
      "step": 29590
    },
    {
      "epoch": 11.5,
      "learning_rate": 1.8467313915857607e-05,
      "loss": 0.1026,
      "step": 29600
    },
    {
      "epoch": 11.5,
      "learning_rate": 1.8466796116504855e-05,
      "loss": 0.1873,
      "step": 29610
    },
    {
      "epoch": 11.5,
      "learning_rate": 1.8466278317152103e-05,
      "loss": 0.3198,
      "step": 29620
    },
    {
      "epoch": 11.51,
      "learning_rate": 1.8465760517799354e-05,
      "loss": 0.1514,
      "step": 29630
    },
    {
      "epoch": 11.51,
      "learning_rate": 1.8465242718446602e-05,
      "loss": 0.0979,
      "step": 29640
    },
    {
      "epoch": 11.51,
      "learning_rate": 1.8464724919093854e-05,
      "loss": 0.1201,
      "step": 29650
    },
    {
      "epoch": 11.52,
      "learning_rate": 1.84642071197411e-05,
      "loss": 0.1535,
      "step": 29660
    },
    {
      "epoch": 11.52,
      "learning_rate": 1.8463689320388353e-05,
      "loss": 0.1397,
      "step": 29670
    },
    {
      "epoch": 11.53,
      "learning_rate": 1.84631715210356e-05,
      "loss": 0.0284,
      "step": 29680
    },
    {
      "epoch": 11.53,
      "learning_rate": 1.846265372168285e-05,
      "loss": 0.1282,
      "step": 29690
    },
    {
      "epoch": 11.53,
      "learning_rate": 1.8462135922330097e-05,
      "loss": 0.1669,
      "step": 29700
    },
    {
      "epoch": 11.54,
      "learning_rate": 1.8461618122977348e-05,
      "loss": 0.0772,
      "step": 29710
    },
    {
      "epoch": 11.54,
      "learning_rate": 1.8461100323624596e-05,
      "loss": 0.1666,
      "step": 29720
    },
    {
      "epoch": 11.55,
      "learning_rate": 1.8460582524271847e-05,
      "loss": 0.1348,
      "step": 29730
    },
    {
      "epoch": 11.55,
      "learning_rate": 1.8460064724919095e-05,
      "loss": 0.0316,
      "step": 29740
    },
    {
      "epoch": 11.55,
      "learning_rate": 1.8459546925566346e-05,
      "loss": 0.1813,
      "step": 29750
    },
    {
      "epoch": 11.56,
      "learning_rate": 1.8459029126213594e-05,
      "loss": 0.3339,
      "step": 29760
    },
    {
      "epoch": 11.56,
      "learning_rate": 1.8458511326860846e-05,
      "loss": 0.138,
      "step": 29770
    },
    {
      "epoch": 11.57,
      "learning_rate": 1.845799352750809e-05,
      "loss": 0.0726,
      "step": 29780
    },
    {
      "epoch": 11.57,
      "learning_rate": 1.845747572815534e-05,
      "loss": 0.2006,
      "step": 29790
    },
    {
      "epoch": 11.57,
      "learning_rate": 1.845695792880259e-05,
      "loss": 0.0544,
      "step": 29800
    },
    {
      "epoch": 11.58,
      "learning_rate": 1.845644012944984e-05,
      "loss": 0.152,
      "step": 29810
    },
    {
      "epoch": 11.58,
      "learning_rate": 1.845592233009709e-05,
      "loss": 0.0808,
      "step": 29820
    },
    {
      "epoch": 11.58,
      "learning_rate": 1.845540453074434e-05,
      "loss": 0.024,
      "step": 29830
    },
    {
      "epoch": 11.59,
      "learning_rate": 1.8454886731391588e-05,
      "loss": 0.2796,
      "step": 29840
    },
    {
      "epoch": 11.59,
      "learning_rate": 1.8454368932038836e-05,
      "loss": 0.0544,
      "step": 29850
    },
    {
      "epoch": 11.6,
      "learning_rate": 1.8453851132686084e-05,
      "loss": 0.1197,
      "step": 29860
    },
    {
      "epoch": 11.6,
      "learning_rate": 1.8453333333333335e-05,
      "loss": 0.0596,
      "step": 29870
    },
    {
      "epoch": 11.6,
      "learning_rate": 1.8452815533980583e-05,
      "loss": 0.1816,
      "step": 29880
    },
    {
      "epoch": 11.61,
      "learning_rate": 1.8452297734627834e-05,
      "loss": 0.0503,
      "step": 29890
    },
    {
      "epoch": 11.61,
      "learning_rate": 1.8451779935275082e-05,
      "loss": 0.3043,
      "step": 29900
    },
    {
      "epoch": 11.62,
      "learning_rate": 1.8451262135922334e-05,
      "loss": 0.1636,
      "step": 29910
    },
    {
      "epoch": 11.62,
      "learning_rate": 1.845074433656958e-05,
      "loss": 0.1138,
      "step": 29920
    },
    {
      "epoch": 11.62,
      "learning_rate": 1.845022653721683e-05,
      "loss": 0.1644,
      "step": 29930
    },
    {
      "epoch": 11.63,
      "learning_rate": 1.8449708737864077e-05,
      "loss": 0.2029,
      "step": 29940
    },
    {
      "epoch": 11.63,
      "learning_rate": 1.844919093851133e-05,
      "loss": 0.0987,
      "step": 29950
    },
    {
      "epoch": 11.63,
      "learning_rate": 1.8448673139158577e-05,
      "loss": 0.1597,
      "step": 29960
    },
    {
      "epoch": 11.64,
      "learning_rate": 1.8448155339805828e-05,
      "loss": 0.2333,
      "step": 29970
    },
    {
      "epoch": 11.64,
      "learning_rate": 1.8447637540453076e-05,
      "loss": 0.1462,
      "step": 29980
    },
    {
      "epoch": 11.65,
      "learning_rate": 1.8447119741100327e-05,
      "loss": 0.0382,
      "step": 29990
    },
    {
      "epoch": 11.65,
      "learning_rate": 1.8446601941747575e-05,
      "loss": 0.2675,
      "step": 30000
    },
    {
      "epoch": 11.65,
      "learning_rate": 1.8446084142394823e-05,
      "loss": 0.0828,
      "step": 30010
    },
    {
      "epoch": 11.66,
      "learning_rate": 1.844556634304207e-05,
      "loss": 0.1023,
      "step": 30020
    },
    {
      "epoch": 11.66,
      "learning_rate": 1.8445048543689322e-05,
      "loss": 0.2264,
      "step": 30030
    },
    {
      "epoch": 11.67,
      "learning_rate": 1.844453074433657e-05,
      "loss": 0.1732,
      "step": 30040
    },
    {
      "epoch": 11.67,
      "learning_rate": 1.844401294498382e-05,
      "loss": 0.0812,
      "step": 30050
    },
    {
      "epoch": 11.67,
      "learning_rate": 1.844349514563107e-05,
      "loss": 0.1494,
      "step": 30060
    },
    {
      "epoch": 11.68,
      "learning_rate": 1.844297734627832e-05,
      "loss": 0.3529,
      "step": 30070
    },
    {
      "epoch": 11.68,
      "learning_rate": 1.844245954692557e-05,
      "loss": 0.24,
      "step": 30080
    },
    {
      "epoch": 11.69,
      "learning_rate": 1.8441941747572817e-05,
      "loss": 0.2826,
      "step": 30090
    },
    {
      "epoch": 11.69,
      "learning_rate": 1.8441423948220064e-05,
      "loss": 0.1667,
      "step": 30100
    },
    {
      "epoch": 11.69,
      "learning_rate": 1.8440906148867316e-05,
      "loss": 0.1577,
      "step": 30110
    },
    {
      "epoch": 11.7,
      "learning_rate": 1.8440388349514564e-05,
      "loss": 0.2155,
      "step": 30120
    },
    {
      "epoch": 11.7,
      "learning_rate": 1.8439870550161815e-05,
      "loss": 0.0121,
      "step": 30130
    },
    {
      "epoch": 11.7,
      "learning_rate": 1.8439352750809063e-05,
      "loss": 0.2657,
      "step": 30140
    },
    {
      "epoch": 11.71,
      "learning_rate": 1.843883495145631e-05,
      "loss": 0.0297,
      "step": 30150
    },
    {
      "epoch": 11.71,
      "learning_rate": 1.8438317152103562e-05,
      "loss": 0.175,
      "step": 30160
    },
    {
      "epoch": 11.72,
      "learning_rate": 1.843779935275081e-05,
      "loss": 0.0143,
      "step": 30170
    },
    {
      "epoch": 11.72,
      "learning_rate": 1.8437281553398058e-05,
      "loss": 0.2095,
      "step": 30180
    },
    {
      "epoch": 11.72,
      "learning_rate": 1.843676375404531e-05,
      "loss": 0.0937,
      "step": 30190
    },
    {
      "epoch": 11.73,
      "learning_rate": 1.8436245954692557e-05,
      "loss": 0.1573,
      "step": 30200
    },
    {
      "epoch": 11.73,
      "learning_rate": 1.843572815533981e-05,
      "loss": 0.1188,
      "step": 30210
    },
    {
      "epoch": 11.74,
      "learning_rate": 1.8435210355987056e-05,
      "loss": 0.2221,
      "step": 30220
    },
    {
      "epoch": 11.74,
      "learning_rate": 1.8434692556634304e-05,
      "loss": 0.2503,
      "step": 30230
    },
    {
      "epoch": 11.74,
      "learning_rate": 1.8434174757281556e-05,
      "loss": 0.1174,
      "step": 30240
    },
    {
      "epoch": 11.75,
      "learning_rate": 1.8433656957928804e-05,
      "loss": 0.1356,
      "step": 30250
    },
    {
      "epoch": 11.75,
      "learning_rate": 1.843313915857605e-05,
      "loss": 0.1095,
      "step": 30260
    },
    {
      "epoch": 11.76,
      "learning_rate": 1.8432621359223303e-05,
      "loss": 0.1392,
      "step": 30270
    },
    {
      "epoch": 11.76,
      "learning_rate": 1.843210355987055e-05,
      "loss": 0.2273,
      "step": 30280
    },
    {
      "epoch": 11.76,
      "learning_rate": 1.8431585760517802e-05,
      "loss": 0.2479,
      "step": 30290
    },
    {
      "epoch": 11.77,
      "learning_rate": 1.843106796116505e-05,
      "loss": 0.1376,
      "step": 30300
    },
    {
      "epoch": 11.77,
      "learning_rate": 1.8430550161812298e-05,
      "loss": 0.0605,
      "step": 30310
    },
    {
      "epoch": 11.77,
      "learning_rate": 1.843003236245955e-05,
      "loss": 0.076,
      "step": 30320
    },
    {
      "epoch": 11.78,
      "learning_rate": 1.8429514563106797e-05,
      "loss": 0.1776,
      "step": 30330
    },
    {
      "epoch": 11.78,
      "learning_rate": 1.842899676375405e-05,
      "loss": 0.2124,
      "step": 30340
    },
    {
      "epoch": 11.79,
      "learning_rate": 1.8428478964401296e-05,
      "loss": 0.0746,
      "step": 30350
    },
    {
      "epoch": 11.79,
      "learning_rate": 1.8427961165048544e-05,
      "loss": 0.1378,
      "step": 30360
    },
    {
      "epoch": 11.79,
      "learning_rate": 1.8427443365695796e-05,
      "loss": 0.127,
      "step": 30370
    },
    {
      "epoch": 11.8,
      "learning_rate": 1.8426925566343044e-05,
      "loss": 0.141,
      "step": 30380
    },
    {
      "epoch": 11.8,
      "learning_rate": 1.842640776699029e-05,
      "loss": 0.1375,
      "step": 30390
    },
    {
      "epoch": 11.81,
      "learning_rate": 1.8425889967637543e-05,
      "loss": 0.108,
      "step": 30400
    },
    {
      "epoch": 11.81,
      "learning_rate": 1.842537216828479e-05,
      "loss": 0.0768,
      "step": 30410
    },
    {
      "epoch": 11.81,
      "learning_rate": 1.8424854368932042e-05,
      "loss": 0.1295,
      "step": 30420
    },
    {
      "epoch": 11.82,
      "learning_rate": 1.842433656957929e-05,
      "loss": 0.1432,
      "step": 30430
    },
    {
      "epoch": 11.82,
      "learning_rate": 1.8423818770226538e-05,
      "loss": 0.1294,
      "step": 30440
    },
    {
      "epoch": 11.83,
      "learning_rate": 1.842330097087379e-05,
      "loss": 0.1831,
      "step": 30450
    },
    {
      "epoch": 11.83,
      "learning_rate": 1.8422783171521037e-05,
      "loss": 0.0795,
      "step": 30460
    },
    {
      "epoch": 11.83,
      "learning_rate": 1.8422265372168285e-05,
      "loss": 0.1986,
      "step": 30470
    },
    {
      "epoch": 11.84,
      "learning_rate": 1.8421747572815536e-05,
      "loss": 0.0149,
      "step": 30480
    },
    {
      "epoch": 11.84,
      "learning_rate": 1.8421229773462784e-05,
      "loss": 0.1656,
      "step": 30490
    },
    {
      "epoch": 11.84,
      "learning_rate": 1.8420711974110036e-05,
      "loss": 0.341,
      "step": 30500
    },
    {
      "epoch": 11.85,
      "learning_rate": 1.8420194174757284e-05,
      "loss": 0.2476,
      "step": 30510
    },
    {
      "epoch": 11.85,
      "learning_rate": 1.841967637540453e-05,
      "loss": 0.2287,
      "step": 30520
    },
    {
      "epoch": 11.86,
      "learning_rate": 1.841915857605178e-05,
      "loss": 0.2133,
      "step": 30530
    },
    {
      "epoch": 11.86,
      "learning_rate": 1.841864077669903e-05,
      "loss": 0.1651,
      "step": 30540
    },
    {
      "epoch": 11.86,
      "learning_rate": 1.841812297734628e-05,
      "loss": 0.2385,
      "step": 30550
    },
    {
      "epoch": 11.87,
      "learning_rate": 1.841760517799353e-05,
      "loss": 0.0998,
      "step": 30560
    },
    {
      "epoch": 11.87,
      "learning_rate": 1.8417087378640778e-05,
      "loss": 0.2349,
      "step": 30570
    },
    {
      "epoch": 11.88,
      "learning_rate": 1.841656957928803e-05,
      "loss": 0.108,
      "step": 30580
    },
    {
      "epoch": 11.88,
      "learning_rate": 1.8416051779935277e-05,
      "loss": 0.3248,
      "step": 30590
    },
    {
      "epoch": 11.88,
      "learning_rate": 1.8415533980582525e-05,
      "loss": 0.1185,
      "step": 30600
    },
    {
      "epoch": 11.89,
      "learning_rate": 1.8415016181229773e-05,
      "loss": 0.2152,
      "step": 30610
    },
    {
      "epoch": 11.89,
      "learning_rate": 1.8414498381877024e-05,
      "loss": 0.4354,
      "step": 30620
    },
    {
      "epoch": 11.9,
      "learning_rate": 1.8413980582524272e-05,
      "loss": 0.131,
      "step": 30630
    },
    {
      "epoch": 11.9,
      "learning_rate": 1.8413462783171523e-05,
      "loss": 0.2992,
      "step": 30640
    },
    {
      "epoch": 11.9,
      "learning_rate": 1.841294498381877e-05,
      "loss": 0.0904,
      "step": 30650
    },
    {
      "epoch": 11.91,
      "learning_rate": 1.8412427184466023e-05,
      "loss": 0.1152,
      "step": 30660
    },
    {
      "epoch": 11.91,
      "learning_rate": 1.841190938511327e-05,
      "loss": 0.1348,
      "step": 30670
    },
    {
      "epoch": 11.91,
      "learning_rate": 1.841139158576052e-05,
      "loss": 0.0238,
      "step": 30680
    },
    {
      "epoch": 11.92,
      "learning_rate": 1.8410873786407767e-05,
      "loss": 0.1011,
      "step": 30690
    },
    {
      "epoch": 11.92,
      "learning_rate": 1.8410355987055018e-05,
      "loss": 0.0841,
      "step": 30700
    },
    {
      "epoch": 11.93,
      "learning_rate": 1.8409838187702266e-05,
      "loss": 0.2363,
      "step": 30710
    },
    {
      "epoch": 11.93,
      "learning_rate": 1.8409320388349517e-05,
      "loss": 0.0177,
      "step": 30720
    },
    {
      "epoch": 11.93,
      "learning_rate": 1.8408802588996765e-05,
      "loss": 0.0056,
      "step": 30730
    },
    {
      "epoch": 11.94,
      "learning_rate": 1.8408284789644016e-05,
      "loss": 0.0267,
      "step": 30740
    },
    {
      "epoch": 11.94,
      "learning_rate": 1.8407766990291264e-05,
      "loss": 0.3607,
      "step": 30750
    },
    {
      "epoch": 11.95,
      "learning_rate": 1.8407249190938512e-05,
      "loss": 0.2015,
      "step": 30760
    },
    {
      "epoch": 11.95,
      "learning_rate": 1.840673139158576e-05,
      "loss": 0.1517,
      "step": 30770
    },
    {
      "epoch": 11.95,
      "learning_rate": 1.840621359223301e-05,
      "loss": 0.0252,
      "step": 30780
    },
    {
      "epoch": 11.96,
      "learning_rate": 1.840569579288026e-05,
      "loss": 0.1524,
      "step": 30790
    },
    {
      "epoch": 11.96,
      "learning_rate": 1.840517799352751e-05,
      "loss": 0.223,
      "step": 30800
    },
    {
      "epoch": 11.97,
      "learning_rate": 1.840466019417476e-05,
      "loss": 0.0442,
      "step": 30810
    },
    {
      "epoch": 11.97,
      "learning_rate": 1.840414239482201e-05,
      "loss": 0.2617,
      "step": 30820
    },
    {
      "epoch": 11.97,
      "learning_rate": 1.8403624595469254e-05,
      "loss": 0.0484,
      "step": 30830
    },
    {
      "epoch": 11.98,
      "learning_rate": 1.8403106796116506e-05,
      "loss": 0.1234,
      "step": 30840
    },
    {
      "epoch": 11.98,
      "learning_rate": 1.8402588996763754e-05,
      "loss": 0.1081,
      "step": 30850
    },
    {
      "epoch": 11.98,
      "learning_rate": 1.8402071197411005e-05,
      "loss": 0.1792,
      "step": 30860
    },
    {
      "epoch": 11.99,
      "learning_rate": 1.8401553398058253e-05,
      "loss": 0.1507,
      "step": 30870
    },
    {
      "epoch": 11.99,
      "learning_rate": 1.8401035598705504e-05,
      "loss": 0.3733,
      "step": 30880
    },
    {
      "epoch": 12.0,
      "learning_rate": 1.8400517799352752e-05,
      "loss": 0.0819,
      "step": 30890
    },
    {
      "epoch": 12.0,
      "learning_rate": 1.8400000000000003e-05,
      "loss": 0.181,
      "step": 30900
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.22270826995372772,
      "eval_runtime": 8.2481,
      "eval_samples_per_second": 440.707,
      "eval_steps_per_second": 55.164,
      "step": 30900
    },
    {
      "epoch": 12.0,
      "learning_rate": 1.839948220064725e-05,
      "loss": 0.2422,
      "step": 30910
    },
    {
      "epoch": 12.01,
      "learning_rate": 1.83989644012945e-05,
      "loss": 0.0784,
      "step": 30920
    },
    {
      "epoch": 12.01,
      "learning_rate": 1.8398446601941747e-05,
      "loss": 0.2248,
      "step": 30930
    },
    {
      "epoch": 12.02,
      "learning_rate": 1.8397928802589e-05,
      "loss": 0.1229,
      "step": 30940
    },
    {
      "epoch": 12.02,
      "learning_rate": 1.8397411003236246e-05,
      "loss": 0.1463,
      "step": 30950
    },
    {
      "epoch": 12.02,
      "learning_rate": 1.8396893203883498e-05,
      "loss": 0.0025,
      "step": 30960
    },
    {
      "epoch": 12.03,
      "learning_rate": 1.8396375404530746e-05,
      "loss": 0.1016,
      "step": 30970
    },
    {
      "epoch": 12.03,
      "learning_rate": 1.8395857605177997e-05,
      "loss": 0.2479,
      "step": 30980
    },
    {
      "epoch": 12.03,
      "learning_rate": 1.8395339805825245e-05,
      "loss": 0.114,
      "step": 30990
    },
    {
      "epoch": 12.04,
      "learning_rate": 1.8394822006472493e-05,
      "loss": 0.1417,
      "step": 31000
    },
    {
      "epoch": 12.04,
      "learning_rate": 1.839430420711974e-05,
      "loss": 0.2496,
      "step": 31010
    },
    {
      "epoch": 12.05,
      "learning_rate": 1.8393786407766992e-05,
      "loss": 0.1498,
      "step": 31020
    },
    {
      "epoch": 12.05,
      "learning_rate": 1.839326860841424e-05,
      "loss": 0.1102,
      "step": 31030
    },
    {
      "epoch": 12.05,
      "learning_rate": 1.839275080906149e-05,
      "loss": 0.0756,
      "step": 31040
    },
    {
      "epoch": 12.06,
      "learning_rate": 1.839223300970874e-05,
      "loss": 0.0964,
      "step": 31050
    },
    {
      "epoch": 12.06,
      "learning_rate": 1.839171521035599e-05,
      "loss": 0.0309,
      "step": 31060
    },
    {
      "epoch": 12.07,
      "learning_rate": 1.839119741100324e-05,
      "loss": 0.1237,
      "step": 31070
    },
    {
      "epoch": 12.07,
      "learning_rate": 1.8390679611650486e-05,
      "loss": 0.1544,
      "step": 31080
    },
    {
      "epoch": 12.07,
      "learning_rate": 1.8390161812297734e-05,
      "loss": 0.0469,
      "step": 31090
    },
    {
      "epoch": 12.08,
      "learning_rate": 1.8389644012944986e-05,
      "loss": 0.073,
      "step": 31100
    },
    {
      "epoch": 12.08,
      "learning_rate": 1.8389126213592234e-05,
      "loss": 0.128,
      "step": 31110
    },
    {
      "epoch": 12.09,
      "learning_rate": 1.8388608414239485e-05,
      "loss": 0.0868,
      "step": 31120
    },
    {
      "epoch": 12.09,
      "learning_rate": 1.8388090614886733e-05,
      "loss": 0.0514,
      "step": 31130
    },
    {
      "epoch": 12.09,
      "learning_rate": 1.8387572815533984e-05,
      "loss": 0.0491,
      "step": 31140
    },
    {
      "epoch": 12.1,
      "learning_rate": 1.8387055016181232e-05,
      "loss": 0.0764,
      "step": 31150
    },
    {
      "epoch": 12.1,
      "learning_rate": 1.838653721682848e-05,
      "loss": 0.2001,
      "step": 31160
    },
    {
      "epoch": 12.1,
      "learning_rate": 1.8386019417475728e-05,
      "loss": 0.1513,
      "step": 31170
    },
    {
      "epoch": 12.11,
      "learning_rate": 1.838550161812298e-05,
      "loss": 0.1522,
      "step": 31180
    },
    {
      "epoch": 12.11,
      "learning_rate": 1.8384983818770227e-05,
      "loss": 0.2692,
      "step": 31190
    },
    {
      "epoch": 12.12,
      "learning_rate": 1.838446601941748e-05,
      "loss": 0.026,
      "step": 31200
    },
    {
      "epoch": 12.12,
      "learning_rate": 1.8383948220064726e-05,
      "loss": 0.0297,
      "step": 31210
    },
    {
      "epoch": 12.12,
      "learning_rate": 1.8383430420711978e-05,
      "loss": 0.2263,
      "step": 31220
    },
    {
      "epoch": 12.13,
      "learning_rate": 1.8382912621359226e-05,
      "loss": 0.1093,
      "step": 31230
    },
    {
      "epoch": 12.13,
      "learning_rate": 1.8382394822006473e-05,
      "loss": 0.1256,
      "step": 31240
    },
    {
      "epoch": 12.14,
      "learning_rate": 1.838187702265372e-05,
      "loss": 0.1786,
      "step": 31250
    },
    {
      "epoch": 12.14,
      "learning_rate": 1.8381359223300973e-05,
      "loss": 0.2135,
      "step": 31260
    },
    {
      "epoch": 12.14,
      "learning_rate": 1.838084142394822e-05,
      "loss": 0.1648,
      "step": 31270
    },
    {
      "epoch": 12.15,
      "learning_rate": 1.8380323624595472e-05,
      "loss": 0.3022,
      "step": 31280
    },
    {
      "epoch": 12.15,
      "learning_rate": 1.837980582524272e-05,
      "loss": 0.086,
      "step": 31290
    },
    {
      "epoch": 12.16,
      "learning_rate": 1.837928802588997e-05,
      "loss": 0.0662,
      "step": 31300
    },
    {
      "epoch": 12.16,
      "learning_rate": 1.837877022653722e-05,
      "loss": 0.267,
      "step": 31310
    },
    {
      "epoch": 12.16,
      "learning_rate": 1.8378252427184467e-05,
      "loss": 0.1044,
      "step": 31320
    },
    {
      "epoch": 12.17,
      "learning_rate": 1.8377734627831715e-05,
      "loss": 0.0922,
      "step": 31330
    },
    {
      "epoch": 12.17,
      "learning_rate": 1.8377216828478966e-05,
      "loss": 0.0894,
      "step": 31340
    },
    {
      "epoch": 12.17,
      "learning_rate": 1.8376699029126214e-05,
      "loss": 0.161,
      "step": 31350
    },
    {
      "epoch": 12.18,
      "learning_rate": 1.8376181229773466e-05,
      "loss": 0.0277,
      "step": 31360
    },
    {
      "epoch": 12.18,
      "learning_rate": 1.8375663430420713e-05,
      "loss": 0.3941,
      "step": 31370
    },
    {
      "epoch": 12.19,
      "learning_rate": 1.8375145631067965e-05,
      "loss": 0.0382,
      "step": 31380
    },
    {
      "epoch": 12.19,
      "learning_rate": 1.8374627831715213e-05,
      "loss": 0.1621,
      "step": 31390
    },
    {
      "epoch": 12.19,
      "learning_rate": 1.837411003236246e-05,
      "loss": 0.3261,
      "step": 31400
    },
    {
      "epoch": 12.2,
      "learning_rate": 1.837359223300971e-05,
      "loss": 0.0166,
      "step": 31410
    },
    {
      "epoch": 12.2,
      "learning_rate": 1.837307443365696e-05,
      "loss": 0.0627,
      "step": 31420
    },
    {
      "epoch": 12.21,
      "learning_rate": 1.8372556634304208e-05,
      "loss": 0.0351,
      "step": 31430
    },
    {
      "epoch": 12.21,
      "learning_rate": 1.837203883495146e-05,
      "loss": 0.1453,
      "step": 31440
    },
    {
      "epoch": 12.21,
      "learning_rate": 1.8371521035598707e-05,
      "loss": 0.0169,
      "step": 31450
    },
    {
      "epoch": 12.22,
      "learning_rate": 1.8371003236245958e-05,
      "loss": 0.1542,
      "step": 31460
    },
    {
      "epoch": 12.22,
      "learning_rate": 1.8370485436893206e-05,
      "loss": 0.0705,
      "step": 31470
    },
    {
      "epoch": 12.23,
      "learning_rate": 1.8369967637540454e-05,
      "loss": 0.1581,
      "step": 31480
    },
    {
      "epoch": 12.23,
      "learning_rate": 1.8369449838187702e-05,
      "loss": 0.1376,
      "step": 31490
    },
    {
      "epoch": 12.23,
      "learning_rate": 1.8368932038834953e-05,
      "loss": 0.1243,
      "step": 31500
    },
    {
      "epoch": 12.24,
      "learning_rate": 1.83684142394822e-05,
      "loss": 0.139,
      "step": 31510
    },
    {
      "epoch": 12.24,
      "learning_rate": 1.8367896440129453e-05,
      "loss": 0.1886,
      "step": 31520
    },
    {
      "epoch": 12.24,
      "learning_rate": 1.83673786407767e-05,
      "loss": 0.1449,
      "step": 31530
    },
    {
      "epoch": 12.25,
      "learning_rate": 1.8366860841423952e-05,
      "loss": 0.1177,
      "step": 31540
    },
    {
      "epoch": 12.25,
      "learning_rate": 1.83663430420712e-05,
      "loss": 0.1878,
      "step": 31550
    },
    {
      "epoch": 12.26,
      "learning_rate": 1.8365825242718448e-05,
      "loss": 0.1194,
      "step": 31560
    },
    {
      "epoch": 12.26,
      "learning_rate": 1.8365307443365696e-05,
      "loss": 0.1402,
      "step": 31570
    },
    {
      "epoch": 12.26,
      "learning_rate": 1.8364789644012947e-05,
      "loss": 0.072,
      "step": 31580
    },
    {
      "epoch": 12.27,
      "learning_rate": 1.8364271844660195e-05,
      "loss": 0.2804,
      "step": 31590
    },
    {
      "epoch": 12.27,
      "learning_rate": 1.8363754045307446e-05,
      "loss": 0.1631,
      "step": 31600
    },
    {
      "epoch": 12.28,
      "learning_rate": 1.8363236245954694e-05,
      "loss": 0.2184,
      "step": 31610
    },
    {
      "epoch": 12.28,
      "learning_rate": 1.8362718446601942e-05,
      "loss": 0.2768,
      "step": 31620
    },
    {
      "epoch": 12.28,
      "learning_rate": 1.8362200647249193e-05,
      "loss": 0.1384,
      "step": 31630
    },
    {
      "epoch": 12.29,
      "learning_rate": 1.836168284789644e-05,
      "loss": 0.1128,
      "step": 31640
    },
    {
      "epoch": 12.29,
      "learning_rate": 1.836116504854369e-05,
      "loss": 0.0812,
      "step": 31650
    },
    {
      "epoch": 12.3,
      "learning_rate": 1.836064724919094e-05,
      "loss": 0.2223,
      "step": 31660
    },
    {
      "epoch": 12.3,
      "learning_rate": 1.836012944983819e-05,
      "loss": 0.0691,
      "step": 31670
    },
    {
      "epoch": 12.3,
      "learning_rate": 1.835961165048544e-05,
      "loss": 0.1319,
      "step": 31680
    },
    {
      "epoch": 12.31,
      "learning_rate": 1.8359093851132688e-05,
      "loss": 0.2502,
      "step": 31690
    },
    {
      "epoch": 12.31,
      "learning_rate": 1.8358576051779936e-05,
      "loss": 0.1089,
      "step": 31700
    },
    {
      "epoch": 12.31,
      "learning_rate": 1.8358058252427187e-05,
      "loss": 0.1455,
      "step": 31710
    },
    {
      "epoch": 12.32,
      "learning_rate": 1.8357540453074435e-05,
      "loss": 0.1812,
      "step": 31720
    },
    {
      "epoch": 12.32,
      "learning_rate": 1.8357022653721683e-05,
      "loss": 0.0979,
      "step": 31730
    },
    {
      "epoch": 12.33,
      "learning_rate": 1.8356504854368934e-05,
      "loss": 0.1434,
      "step": 31740
    },
    {
      "epoch": 12.33,
      "learning_rate": 1.8355987055016182e-05,
      "loss": 0.16,
      "step": 31750
    },
    {
      "epoch": 12.33,
      "learning_rate": 1.8355469255663433e-05,
      "loss": 0.1281,
      "step": 31760
    },
    {
      "epoch": 12.34,
      "learning_rate": 1.835495145631068e-05,
      "loss": 0.0247,
      "step": 31770
    },
    {
      "epoch": 12.34,
      "learning_rate": 1.835443365695793e-05,
      "loss": 0.1098,
      "step": 31780
    },
    {
      "epoch": 12.35,
      "learning_rate": 1.835391585760518e-05,
      "loss": 0.1115,
      "step": 31790
    },
    {
      "epoch": 12.35,
      "learning_rate": 1.835339805825243e-05,
      "loss": 0.1328,
      "step": 31800
    },
    {
      "epoch": 12.35,
      "learning_rate": 1.8352880258899676e-05,
      "loss": 0.0744,
      "step": 31810
    },
    {
      "epoch": 12.36,
      "learning_rate": 1.8352362459546928e-05,
      "loss": 0.0702,
      "step": 31820
    },
    {
      "epoch": 12.36,
      "learning_rate": 1.8351844660194176e-05,
      "loss": 0.266,
      "step": 31830
    },
    {
      "epoch": 12.37,
      "learning_rate": 1.8351326860841427e-05,
      "loss": 0.0982,
      "step": 31840
    },
    {
      "epoch": 12.37,
      "learning_rate": 1.8350809061488675e-05,
      "loss": 0.0464,
      "step": 31850
    },
    {
      "epoch": 12.37,
      "learning_rate": 1.8350291262135923e-05,
      "loss": 0.0984,
      "step": 31860
    },
    {
      "epoch": 12.38,
      "learning_rate": 1.8349773462783174e-05,
      "loss": 0.214,
      "step": 31870
    },
    {
      "epoch": 12.38,
      "learning_rate": 1.8349255663430422e-05,
      "loss": 0.204,
      "step": 31880
    },
    {
      "epoch": 12.38,
      "learning_rate": 1.834873786407767e-05,
      "loss": 0.0853,
      "step": 31890
    },
    {
      "epoch": 12.39,
      "learning_rate": 1.834822006472492e-05,
      "loss": 0.0712,
      "step": 31900
    },
    {
      "epoch": 12.39,
      "learning_rate": 1.834770226537217e-05,
      "loss": 0.1779,
      "step": 31910
    },
    {
      "epoch": 12.4,
      "learning_rate": 1.834718446601942e-05,
      "loss": 0.102,
      "step": 31920
    },
    {
      "epoch": 12.4,
      "learning_rate": 1.834666666666667e-05,
      "loss": 0.2528,
      "step": 31930
    },
    {
      "epoch": 12.4,
      "learning_rate": 1.8346148867313916e-05,
      "loss": 0.1474,
      "step": 31940
    },
    {
      "epoch": 12.41,
      "learning_rate": 1.8345631067961168e-05,
      "loss": 0.1474,
      "step": 31950
    },
    {
      "epoch": 12.41,
      "learning_rate": 1.8345113268608415e-05,
      "loss": 0.0646,
      "step": 31960
    },
    {
      "epoch": 12.42,
      "learning_rate": 1.8344595469255667e-05,
      "loss": 0.1524,
      "step": 31970
    },
    {
      "epoch": 12.42,
      "learning_rate": 1.8344077669902915e-05,
      "loss": 0.1509,
      "step": 31980
    },
    {
      "epoch": 12.42,
      "learning_rate": 1.8343559870550163e-05,
      "loss": 0.0598,
      "step": 31990
    },
    {
      "epoch": 12.43,
      "learning_rate": 1.834304207119741e-05,
      "loss": 0.0483,
      "step": 32000
    },
    {
      "epoch": 12.43,
      "learning_rate": 1.8342524271844662e-05,
      "loss": 0.1897,
      "step": 32010
    },
    {
      "epoch": 12.43,
      "learning_rate": 1.834200647249191e-05,
      "loss": 0.2733,
      "step": 32020
    },
    {
      "epoch": 12.44,
      "learning_rate": 1.834148867313916e-05,
      "loss": 0.1624,
      "step": 32030
    },
    {
      "epoch": 12.44,
      "learning_rate": 1.834097087378641e-05,
      "loss": 0.0697,
      "step": 32040
    },
    {
      "epoch": 12.45,
      "learning_rate": 1.834045307443366e-05,
      "loss": 0.1437,
      "step": 32050
    },
    {
      "epoch": 12.45,
      "learning_rate": 1.8339935275080908e-05,
      "loss": 0.3214,
      "step": 32060
    },
    {
      "epoch": 12.45,
      "learning_rate": 1.8339417475728156e-05,
      "loss": 0.0249,
      "step": 32070
    },
    {
      "epoch": 12.46,
      "learning_rate": 1.8338899676375404e-05,
      "loss": 0.177,
      "step": 32080
    },
    {
      "epoch": 12.46,
      "learning_rate": 1.8338381877022655e-05,
      "loss": 0.2614,
      "step": 32090
    },
    {
      "epoch": 12.47,
      "learning_rate": 1.8337864077669903e-05,
      "loss": 0.1922,
      "step": 32100
    },
    {
      "epoch": 12.47,
      "learning_rate": 1.8337346278317155e-05,
      "loss": 0.1546,
      "step": 32110
    },
    {
      "epoch": 12.47,
      "learning_rate": 1.8336828478964403e-05,
      "loss": 0.1133,
      "step": 32120
    },
    {
      "epoch": 12.48,
      "learning_rate": 1.8336310679611654e-05,
      "loss": 0.0516,
      "step": 32130
    },
    {
      "epoch": 12.48,
      "learning_rate": 1.8335792880258902e-05,
      "loss": 0.0929,
      "step": 32140
    },
    {
      "epoch": 12.49,
      "learning_rate": 1.833527508090615e-05,
      "loss": 0.2044,
      "step": 32150
    },
    {
      "epoch": 12.49,
      "learning_rate": 1.8334757281553398e-05,
      "loss": 0.0755,
      "step": 32160
    },
    {
      "epoch": 12.49,
      "learning_rate": 1.833423948220065e-05,
      "loss": 0.0961,
      "step": 32170
    },
    {
      "epoch": 12.5,
      "learning_rate": 1.8333721682847897e-05,
      "loss": 0.1794,
      "step": 32180
    },
    {
      "epoch": 12.5,
      "learning_rate": 1.8333203883495148e-05,
      "loss": 0.0441,
      "step": 32190
    },
    {
      "epoch": 12.5,
      "learning_rate": 1.8332686084142396e-05,
      "loss": 0.212,
      "step": 32200
    },
    {
      "epoch": 12.51,
      "learning_rate": 1.8332168284789647e-05,
      "loss": 0.1339,
      "step": 32210
    },
    {
      "epoch": 12.51,
      "learning_rate": 1.8331650485436895e-05,
      "loss": 0.0652,
      "step": 32220
    },
    {
      "epoch": 12.52,
      "learning_rate": 1.8331132686084143e-05,
      "loss": 0.1072,
      "step": 32230
    },
    {
      "epoch": 12.52,
      "learning_rate": 1.833061488673139e-05,
      "loss": 0.0717,
      "step": 32240
    },
    {
      "epoch": 12.52,
      "learning_rate": 1.8330097087378643e-05,
      "loss": 0.2289,
      "step": 32250
    },
    {
      "epoch": 12.53,
      "learning_rate": 1.832957928802589e-05,
      "loss": 0.0811,
      "step": 32260
    },
    {
      "epoch": 12.53,
      "learning_rate": 1.8329061488673142e-05,
      "loss": 0.1498,
      "step": 32270
    },
    {
      "epoch": 12.54,
      "learning_rate": 1.832854368932039e-05,
      "loss": 0.1239,
      "step": 32280
    },
    {
      "epoch": 12.54,
      "learning_rate": 1.832802588996764e-05,
      "loss": 0.1534,
      "step": 32290
    },
    {
      "epoch": 12.54,
      "learning_rate": 1.8327508090614886e-05,
      "loss": 0.0458,
      "step": 32300
    },
    {
      "epoch": 12.55,
      "learning_rate": 1.8326990291262137e-05,
      "loss": 0.2795,
      "step": 32310
    },
    {
      "epoch": 12.55,
      "learning_rate": 1.8326472491909385e-05,
      "loss": 0.1002,
      "step": 32320
    },
    {
      "epoch": 12.56,
      "learning_rate": 1.8325954692556636e-05,
      "loss": 0.0867,
      "step": 32330
    },
    {
      "epoch": 12.56,
      "learning_rate": 1.8325436893203884e-05,
      "loss": 0.1068,
      "step": 32340
    },
    {
      "epoch": 12.56,
      "learning_rate": 1.8324919093851135e-05,
      "loss": 0.1356,
      "step": 32350
    },
    {
      "epoch": 12.57,
      "learning_rate": 1.8324401294498383e-05,
      "loss": 0.0825,
      "step": 32360
    },
    {
      "epoch": 12.57,
      "learning_rate": 1.8323883495145635e-05,
      "loss": 0.1746,
      "step": 32370
    },
    {
      "epoch": 12.57,
      "learning_rate": 1.832336569579288e-05,
      "loss": 0.1291,
      "step": 32380
    },
    {
      "epoch": 12.58,
      "learning_rate": 1.832284789644013e-05,
      "loss": 0.1322,
      "step": 32390
    },
    {
      "epoch": 12.58,
      "learning_rate": 1.832233009708738e-05,
      "loss": 0.1822,
      "step": 32400
    },
    {
      "epoch": 12.59,
      "learning_rate": 1.832181229773463e-05,
      "loss": 0.0568,
      "step": 32410
    },
    {
      "epoch": 12.59,
      "learning_rate": 1.8321294498381878e-05,
      "loss": 0.1486,
      "step": 32420
    },
    {
      "epoch": 12.59,
      "learning_rate": 1.832077669902913e-05,
      "loss": 0.1412,
      "step": 32430
    },
    {
      "epoch": 12.6,
      "learning_rate": 1.8320258899676377e-05,
      "loss": 0.296,
      "step": 32440
    },
    {
      "epoch": 12.6,
      "learning_rate": 1.8319741100323628e-05,
      "loss": 0.3106,
      "step": 32450
    },
    {
      "epoch": 12.61,
      "learning_rate": 1.8319223300970873e-05,
      "loss": 0.1247,
      "step": 32460
    },
    {
      "epoch": 12.61,
      "learning_rate": 1.8318705501618124e-05,
      "loss": 0.2442,
      "step": 32470
    },
    {
      "epoch": 12.61,
      "learning_rate": 1.8318187702265372e-05,
      "loss": 0.0681,
      "step": 32480
    },
    {
      "epoch": 12.62,
      "learning_rate": 1.8317669902912623e-05,
      "loss": 0.0985,
      "step": 32490
    },
    {
      "epoch": 12.62,
      "learning_rate": 1.831715210355987e-05,
      "loss": 0.1817,
      "step": 32500
    },
    {
      "epoch": 12.63,
      "learning_rate": 1.8316634304207122e-05,
      "loss": 0.175,
      "step": 32510
    },
    {
      "epoch": 12.63,
      "learning_rate": 1.831611650485437e-05,
      "loss": 0.1037,
      "step": 32520
    },
    {
      "epoch": 12.63,
      "learning_rate": 1.831559870550162e-05,
      "loss": 0.0533,
      "step": 32530
    },
    {
      "epoch": 12.64,
      "learning_rate": 1.831508090614887e-05,
      "loss": 0.1052,
      "step": 32540
    },
    {
      "epoch": 12.64,
      "learning_rate": 1.8314563106796118e-05,
      "loss": 0.2132,
      "step": 32550
    },
    {
      "epoch": 12.64,
      "learning_rate": 1.8314045307443365e-05,
      "loss": 0.2087,
      "step": 32560
    },
    {
      "epoch": 12.65,
      "learning_rate": 1.8313527508090617e-05,
      "loss": 0.3662,
      "step": 32570
    },
    {
      "epoch": 12.65,
      "learning_rate": 1.8313009708737865e-05,
      "loss": 0.0445,
      "step": 32580
    },
    {
      "epoch": 12.66,
      "learning_rate": 1.8312491909385116e-05,
      "loss": 0.0943,
      "step": 32590
    },
    {
      "epoch": 12.66,
      "learning_rate": 1.8311974110032364e-05,
      "loss": 0.2825,
      "step": 32600
    },
    {
      "epoch": 12.66,
      "learning_rate": 1.8311456310679615e-05,
      "loss": 0.272,
      "step": 32610
    },
    {
      "epoch": 12.67,
      "learning_rate": 1.8310938511326863e-05,
      "loss": 0.119,
      "step": 32620
    },
    {
      "epoch": 12.67,
      "learning_rate": 1.831042071197411e-05,
      "loss": 0.1634,
      "step": 32630
    },
    {
      "epoch": 12.68,
      "learning_rate": 1.830990291262136e-05,
      "loss": 0.0256,
      "step": 32640
    },
    {
      "epoch": 12.68,
      "learning_rate": 1.830938511326861e-05,
      "loss": 0.3519,
      "step": 32650
    },
    {
      "epoch": 12.68,
      "learning_rate": 1.8308867313915858e-05,
      "loss": 0.1472,
      "step": 32660
    },
    {
      "epoch": 12.69,
      "learning_rate": 1.830834951456311e-05,
      "loss": 0.1875,
      "step": 32670
    },
    {
      "epoch": 12.69,
      "learning_rate": 1.8307831715210357e-05,
      "loss": 0.1831,
      "step": 32680
    },
    {
      "epoch": 12.7,
      "learning_rate": 1.830731391585761e-05,
      "loss": 0.2519,
      "step": 32690
    },
    {
      "epoch": 12.7,
      "learning_rate": 1.8306796116504857e-05,
      "loss": 0.1866,
      "step": 32700
    },
    {
      "epoch": 12.7,
      "learning_rate": 1.8306278317152105e-05,
      "loss": 0.2267,
      "step": 32710
    },
    {
      "epoch": 12.71,
      "learning_rate": 1.8305760517799353e-05,
      "loss": 0.1787,
      "step": 32720
    },
    {
      "epoch": 12.71,
      "learning_rate": 1.8305242718446604e-05,
      "loss": 0.1153,
      "step": 32730
    },
    {
      "epoch": 12.71,
      "learning_rate": 1.8304724919093852e-05,
      "loss": 0.1347,
      "step": 32740
    },
    {
      "epoch": 12.72,
      "learning_rate": 1.8304207119741103e-05,
      "loss": 0.0355,
      "step": 32750
    },
    {
      "epoch": 12.72,
      "learning_rate": 1.830368932038835e-05,
      "loss": 0.1621,
      "step": 32760
    },
    {
      "epoch": 12.73,
      "learning_rate": 1.8303171521035602e-05,
      "loss": 0.0986,
      "step": 32770
    },
    {
      "epoch": 12.73,
      "learning_rate": 1.830265372168285e-05,
      "loss": 0.1976,
      "step": 32780
    },
    {
      "epoch": 12.73,
      "learning_rate": 1.8302135922330098e-05,
      "loss": 0.1079,
      "step": 32790
    },
    {
      "epoch": 12.74,
      "learning_rate": 1.8301618122977346e-05,
      "loss": 0.0607,
      "step": 32800
    },
    {
      "epoch": 12.74,
      "learning_rate": 1.8301100323624597e-05,
      "loss": 0.0796,
      "step": 32810
    },
    {
      "epoch": 12.75,
      "learning_rate": 1.8300582524271845e-05,
      "loss": 0.159,
      "step": 32820
    },
    {
      "epoch": 12.75,
      "learning_rate": 1.8300064724919097e-05,
      "loss": 0.0635,
      "step": 32830
    },
    {
      "epoch": 12.75,
      "learning_rate": 1.8299546925566345e-05,
      "loss": 0.2037,
      "step": 32840
    },
    {
      "epoch": 12.76,
      "learning_rate": 1.8299029126213596e-05,
      "loss": 0.097,
      "step": 32850
    },
    {
      "epoch": 12.76,
      "learning_rate": 1.8298511326860844e-05,
      "loss": 0.2327,
      "step": 32860
    },
    {
      "epoch": 12.77,
      "learning_rate": 1.8297993527508092e-05,
      "loss": 0.2174,
      "step": 32870
    },
    {
      "epoch": 12.77,
      "learning_rate": 1.829747572815534e-05,
      "loss": 0.1167,
      "step": 32880
    },
    {
      "epoch": 12.77,
      "learning_rate": 1.829695792880259e-05,
      "loss": 0.0507,
      "step": 32890
    },
    {
      "epoch": 12.78,
      "learning_rate": 1.829644012944984e-05,
      "loss": 0.1171,
      "step": 32900
    },
    {
      "epoch": 12.78,
      "learning_rate": 1.829592233009709e-05,
      "loss": 0.1201,
      "step": 32910
    },
    {
      "epoch": 12.78,
      "learning_rate": 1.8295404530744338e-05,
      "loss": 0.2656,
      "step": 32920
    },
    {
      "epoch": 12.79,
      "learning_rate": 1.829488673139159e-05,
      "loss": 0.0648,
      "step": 32930
    },
    {
      "epoch": 12.79,
      "learning_rate": 1.8294368932038837e-05,
      "loss": 0.0811,
      "step": 32940
    },
    {
      "epoch": 12.8,
      "learning_rate": 1.8293851132686085e-05,
      "loss": 0.0934,
      "step": 32950
    },
    {
      "epoch": 12.8,
      "learning_rate": 1.8293333333333333e-05,
      "loss": 0.073,
      "step": 32960
    },
    {
      "epoch": 12.8,
      "learning_rate": 1.8292815533980585e-05,
      "loss": 0.0994,
      "step": 32970
    },
    {
      "epoch": 12.81,
      "learning_rate": 1.8292297734627832e-05,
      "loss": 0.1127,
      "step": 32980
    },
    {
      "epoch": 12.81,
      "learning_rate": 1.8291779935275084e-05,
      "loss": 0.1411,
      "step": 32990
    },
    {
      "epoch": 12.82,
      "learning_rate": 1.8291262135922332e-05,
      "loss": 0.124,
      "step": 33000
    },
    {
      "epoch": 12.82,
      "learning_rate": 1.8290744336569583e-05,
      "loss": 0.1083,
      "step": 33010
    },
    {
      "epoch": 12.82,
      "learning_rate": 1.829022653721683e-05,
      "loss": 0.058,
      "step": 33020
    },
    {
      "epoch": 12.83,
      "learning_rate": 1.828970873786408e-05,
      "loss": 0.1674,
      "step": 33030
    },
    {
      "epoch": 12.83,
      "learning_rate": 1.8289190938511327e-05,
      "loss": 0.2379,
      "step": 33040
    },
    {
      "epoch": 12.83,
      "learning_rate": 1.8288673139158578e-05,
      "loss": 0.0548,
      "step": 33050
    },
    {
      "epoch": 12.84,
      "learning_rate": 1.8288155339805826e-05,
      "loss": 0.0388,
      "step": 33060
    },
    {
      "epoch": 12.84,
      "learning_rate": 1.8287637540453077e-05,
      "loss": 0.2646,
      "step": 33070
    },
    {
      "epoch": 12.85,
      "learning_rate": 1.8287119741100325e-05,
      "loss": 0.1471,
      "step": 33080
    },
    {
      "epoch": 12.85,
      "learning_rate": 1.8286601941747573e-05,
      "loss": 0.2605,
      "step": 33090
    },
    {
      "epoch": 12.85,
      "learning_rate": 1.8286084142394825e-05,
      "loss": 0.4131,
      "step": 33100
    },
    {
      "epoch": 12.86,
      "learning_rate": 1.8285566343042072e-05,
      "loss": 0.0939,
      "step": 33110
    },
    {
      "epoch": 12.86,
      "learning_rate": 1.828504854368932e-05,
      "loss": 0.0202,
      "step": 33120
    },
    {
      "epoch": 12.87,
      "learning_rate": 1.828453074433657e-05,
      "loss": 0.1134,
      "step": 33130
    },
    {
      "epoch": 12.87,
      "learning_rate": 1.828401294498382e-05,
      "loss": 0.1518,
      "step": 33140
    },
    {
      "epoch": 12.87,
      "learning_rate": 1.828349514563107e-05,
      "loss": 0.1583,
      "step": 33150
    },
    {
      "epoch": 12.88,
      "learning_rate": 1.828297734627832e-05,
      "loss": 0.2037,
      "step": 33160
    },
    {
      "epoch": 12.88,
      "learning_rate": 1.8282459546925567e-05,
      "loss": 0.1662,
      "step": 33170
    },
    {
      "epoch": 12.89,
      "learning_rate": 1.8281941747572818e-05,
      "loss": 0.1699,
      "step": 33180
    },
    {
      "epoch": 12.89,
      "learning_rate": 1.8281423948220066e-05,
      "loss": 0.0736,
      "step": 33190
    },
    {
      "epoch": 12.89,
      "learning_rate": 1.8280906148867314e-05,
      "loss": 0.1082,
      "step": 33200
    },
    {
      "epoch": 12.9,
      "learning_rate": 1.8280388349514565e-05,
      "loss": 0.1334,
      "step": 33210
    },
    {
      "epoch": 12.9,
      "learning_rate": 1.8279870550161813e-05,
      "loss": 0.2713,
      "step": 33220
    },
    {
      "epoch": 12.9,
      "learning_rate": 1.8279352750809064e-05,
      "loss": 0.0521,
      "step": 33230
    },
    {
      "epoch": 12.91,
      "learning_rate": 1.8278834951456312e-05,
      "loss": 0.167,
      "step": 33240
    },
    {
      "epoch": 12.91,
      "learning_rate": 1.827831715210356e-05,
      "loss": 0.2874,
      "step": 33250
    },
    {
      "epoch": 12.92,
      "learning_rate": 1.827779935275081e-05,
      "loss": 0.1002,
      "step": 33260
    },
    {
      "epoch": 12.92,
      "learning_rate": 1.827728155339806e-05,
      "loss": 0.0337,
      "step": 33270
    },
    {
      "epoch": 12.92,
      "learning_rate": 1.8276763754045307e-05,
      "loss": 0.0689,
      "step": 33280
    },
    {
      "epoch": 12.93,
      "learning_rate": 1.827624595469256e-05,
      "loss": 0.2639,
      "step": 33290
    },
    {
      "epoch": 12.93,
      "learning_rate": 1.8275728155339807e-05,
      "loss": 0.1044,
      "step": 33300
    },
    {
      "epoch": 12.94,
      "learning_rate": 1.8275210355987058e-05,
      "loss": 0.0705,
      "step": 33310
    },
    {
      "epoch": 12.94,
      "learning_rate": 1.8274692556634306e-05,
      "loss": 0.2664,
      "step": 33320
    },
    {
      "epoch": 12.94,
      "learning_rate": 1.8274174757281554e-05,
      "loss": 0.1955,
      "step": 33330
    },
    {
      "epoch": 12.95,
      "learning_rate": 1.8273656957928805e-05,
      "loss": 0.159,
      "step": 33340
    },
    {
      "epoch": 12.95,
      "learning_rate": 1.8273139158576053e-05,
      "loss": 0.1996,
      "step": 33350
    },
    {
      "epoch": 12.96,
      "learning_rate": 1.82726213592233e-05,
      "loss": 0.1577,
      "step": 33360
    },
    {
      "epoch": 12.96,
      "learning_rate": 1.8272103559870552e-05,
      "loss": 0.1965,
      "step": 33370
    },
    {
      "epoch": 12.96,
      "learning_rate": 1.82715857605178e-05,
      "loss": 0.1096,
      "step": 33380
    },
    {
      "epoch": 12.97,
      "learning_rate": 1.827106796116505e-05,
      "loss": 0.2168,
      "step": 33390
    },
    {
      "epoch": 12.97,
      "learning_rate": 1.82705501618123e-05,
      "loss": 0.1023,
      "step": 33400
    },
    {
      "epoch": 12.97,
      "learning_rate": 1.8270032362459547e-05,
      "loss": 0.1975,
      "step": 33410
    },
    {
      "epoch": 12.98,
      "learning_rate": 1.82695145631068e-05,
      "loss": 0.13,
      "step": 33420
    },
    {
      "epoch": 12.98,
      "learning_rate": 1.8268996763754047e-05,
      "loss": 0.2114,
      "step": 33430
    },
    {
      "epoch": 12.99,
      "learning_rate": 1.8268478964401295e-05,
      "loss": 0.2241,
      "step": 33440
    },
    {
      "epoch": 12.99,
      "learning_rate": 1.8267961165048546e-05,
      "loss": 0.2898,
      "step": 33450
    },
    {
      "epoch": 12.99,
      "learning_rate": 1.8267443365695794e-05,
      "loss": 0.2021,
      "step": 33460
    },
    {
      "epoch": 13.0,
      "learning_rate": 1.8266925566343042e-05,
      "loss": 0.1072,
      "step": 33470
    },
    {
      "epoch": 13.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.2302989661693573,
      "eval_runtime": 8.2265,
      "eval_samples_per_second": 441.865,
      "eval_steps_per_second": 55.309,
      "step": 33475
    },
    {
      "epoch": 13.0,
      "learning_rate": 1.8266407766990293e-05,
      "loss": 0.1365,
      "step": 33480
    },
    {
      "epoch": 13.01,
      "learning_rate": 1.826588996763754e-05,
      "loss": 0.2059,
      "step": 33490
    },
    {
      "epoch": 13.01,
      "learning_rate": 1.8265372168284792e-05,
      "loss": 0.3802,
      "step": 33500
    },
    {
      "epoch": 13.01,
      "learning_rate": 1.826485436893204e-05,
      "loss": 0.0427,
      "step": 33510
    },
    {
      "epoch": 13.02,
      "learning_rate": 1.8264336569579288e-05,
      "loss": 0.214,
      "step": 33520
    },
    {
      "epoch": 13.02,
      "learning_rate": 1.826381877022654e-05,
      "loss": 0.1464,
      "step": 33530
    },
    {
      "epoch": 13.03,
      "learning_rate": 1.8263300970873787e-05,
      "loss": 0.102,
      "step": 33540
    },
    {
      "epoch": 13.03,
      "learning_rate": 1.8262783171521035e-05,
      "loss": 0.0757,
      "step": 33550
    },
    {
      "epoch": 13.03,
      "learning_rate": 1.8262265372168287e-05,
      "loss": 0.1547,
      "step": 33560
    },
    {
      "epoch": 13.04,
      "learning_rate": 1.8261747572815535e-05,
      "loss": 0.1187,
      "step": 33570
    },
    {
      "epoch": 13.04,
      "learning_rate": 1.8261229773462786e-05,
      "loss": 0.092,
      "step": 33580
    },
    {
      "epoch": 13.04,
      "learning_rate": 1.8260711974110034e-05,
      "loss": 0.2484,
      "step": 33590
    },
    {
      "epoch": 13.05,
      "learning_rate": 1.826019417475728e-05,
      "loss": 0.0708,
      "step": 33600
    },
    {
      "epoch": 13.05,
      "learning_rate": 1.8259676375404533e-05,
      "loss": 0.1323,
      "step": 33610
    },
    {
      "epoch": 13.06,
      "learning_rate": 1.825915857605178e-05,
      "loss": 0.2608,
      "step": 33620
    },
    {
      "epoch": 13.06,
      "learning_rate": 1.825864077669903e-05,
      "loss": 0.3223,
      "step": 33630
    },
    {
      "epoch": 13.06,
      "learning_rate": 1.825812297734628e-05,
      "loss": 0.0237,
      "step": 33640
    },
    {
      "epoch": 13.07,
      "learning_rate": 1.8257605177993528e-05,
      "loss": 0.2666,
      "step": 33650
    },
    {
      "epoch": 13.07,
      "learning_rate": 1.825708737864078e-05,
      "loss": 0.2934,
      "step": 33660
    },
    {
      "epoch": 13.08,
      "learning_rate": 1.8256569579288027e-05,
      "loss": 0.0447,
      "step": 33670
    },
    {
      "epoch": 13.08,
      "learning_rate": 1.825605177993528e-05,
      "loss": 0.1321,
      "step": 33680
    },
    {
      "epoch": 13.08,
      "learning_rate": 1.8255533980582527e-05,
      "loss": 0.2468,
      "step": 33690
    },
    {
      "epoch": 13.09,
      "learning_rate": 1.8255016181229774e-05,
      "loss": 0.1322,
      "step": 33700
    },
    {
      "epoch": 13.09,
      "learning_rate": 1.8254498381877022e-05,
      "loss": 0.1485,
      "step": 33710
    },
    {
      "epoch": 13.1,
      "learning_rate": 1.8253980582524274e-05,
      "loss": 0.2172,
      "step": 33720
    },
    {
      "epoch": 13.1,
      "learning_rate": 1.825346278317152e-05,
      "loss": 0.0599,
      "step": 33730
    },
    {
      "epoch": 13.1,
      "learning_rate": 1.8252944983818773e-05,
      "loss": 0.1004,
      "step": 33740
    },
    {
      "epoch": 13.11,
      "learning_rate": 1.825242718446602e-05,
      "loss": 0.1175,
      "step": 33750
    },
    {
      "epoch": 13.11,
      "learning_rate": 1.8251909385113272e-05,
      "loss": 0.0409,
      "step": 33760
    },
    {
      "epoch": 13.11,
      "learning_rate": 1.8251391585760517e-05,
      "loss": 0.1746,
      "step": 33770
    },
    {
      "epoch": 13.12,
      "learning_rate": 1.8250873786407768e-05,
      "loss": 0.1425,
      "step": 33780
    },
    {
      "epoch": 13.12,
      "learning_rate": 1.8250355987055016e-05,
      "loss": 0.0375,
      "step": 33790
    },
    {
      "epoch": 13.13,
      "learning_rate": 1.8249838187702267e-05,
      "loss": 0.2546,
      "step": 33800
    },
    {
      "epoch": 13.13,
      "learning_rate": 1.8249320388349515e-05,
      "loss": 0.1485,
      "step": 33810
    },
    {
      "epoch": 13.13,
      "learning_rate": 1.8248802588996767e-05,
      "loss": 0.1268,
      "step": 33820
    },
    {
      "epoch": 13.14,
      "learning_rate": 1.8248284789644014e-05,
      "loss": 0.1427,
      "step": 33830
    },
    {
      "epoch": 13.14,
      "learning_rate": 1.8247766990291266e-05,
      "loss": 0.1468,
      "step": 33840
    },
    {
      "epoch": 13.15,
      "learning_rate": 1.824724919093851e-05,
      "loss": 0.0339,
      "step": 33850
    },
    {
      "epoch": 13.15,
      "learning_rate": 1.824673139158576e-05,
      "loss": 0.1943,
      "step": 33860
    },
    {
      "epoch": 13.15,
      "learning_rate": 1.824621359223301e-05,
      "loss": 0.2452,
      "step": 33870
    },
    {
      "epoch": 13.16,
      "learning_rate": 1.824569579288026e-05,
      "loss": 0.0081,
      "step": 33880
    },
    {
      "epoch": 13.16,
      "learning_rate": 1.824517799352751e-05,
      "loss": 0.3292,
      "step": 33890
    },
    {
      "epoch": 13.17,
      "learning_rate": 1.824466019417476e-05,
      "loss": 0.085,
      "step": 33900
    },
    {
      "epoch": 13.17,
      "learning_rate": 1.8244142394822008e-05,
      "loss": 0.0759,
      "step": 33910
    },
    {
      "epoch": 13.17,
      "learning_rate": 1.824362459546926e-05,
      "loss": 0.1177,
      "step": 33920
    },
    {
      "epoch": 13.18,
      "learning_rate": 1.8243106796116504e-05,
      "loss": 0.0926,
      "step": 33930
    },
    {
      "epoch": 13.18,
      "learning_rate": 1.8242588996763755e-05,
      "loss": 0.138,
      "step": 33940
    },
    {
      "epoch": 13.18,
      "learning_rate": 1.8242071197411003e-05,
      "loss": 0.3009,
      "step": 33950
    },
    {
      "epoch": 13.19,
      "learning_rate": 1.8241553398058254e-05,
      "loss": 0.064,
      "step": 33960
    },
    {
      "epoch": 13.19,
      "learning_rate": 1.8241035598705502e-05,
      "loss": 0.1526,
      "step": 33970
    },
    {
      "epoch": 13.2,
      "learning_rate": 1.8240517799352754e-05,
      "loss": 0.0594,
      "step": 33980
    },
    {
      "epoch": 13.2,
      "learning_rate": 1.824e-05,
      "loss": 0.0694,
      "step": 33990
    },
    {
      "epoch": 13.2,
      "learning_rate": 1.8239482200647253e-05,
      "loss": 0.2333,
      "step": 34000
    },
    {
      "epoch": 13.21,
      "learning_rate": 1.8238964401294497e-05,
      "loss": 0.0691,
      "step": 34010
    },
    {
      "epoch": 13.21,
      "learning_rate": 1.823844660194175e-05,
      "loss": 0.0706,
      "step": 34020
    },
    {
      "epoch": 13.22,
      "learning_rate": 1.8237928802588997e-05,
      "loss": 0.3733,
      "step": 34030
    },
    {
      "epoch": 13.22,
      "learning_rate": 1.8237411003236248e-05,
      "loss": 0.1725,
      "step": 34040
    },
    {
      "epoch": 13.22,
      "learning_rate": 1.8236893203883496e-05,
      "loss": 0.1445,
      "step": 34050
    },
    {
      "epoch": 13.23,
      "learning_rate": 1.8236375404530747e-05,
      "loss": 0.1844,
      "step": 34060
    },
    {
      "epoch": 13.23,
      "learning_rate": 1.8235857605177995e-05,
      "loss": 0.0283,
      "step": 34070
    },
    {
      "epoch": 13.23,
      "learning_rate": 1.8235339805825246e-05,
      "loss": 0.0915,
      "step": 34080
    },
    {
      "epoch": 13.24,
      "learning_rate": 1.823482200647249e-05,
      "loss": 0.1181,
      "step": 34090
    },
    {
      "epoch": 13.24,
      "learning_rate": 1.8234304207119742e-05,
      "loss": 0.1147,
      "step": 34100
    },
    {
      "epoch": 13.25,
      "learning_rate": 1.823378640776699e-05,
      "loss": 0.1686,
      "step": 34110
    },
    {
      "epoch": 13.25,
      "learning_rate": 1.823326860841424e-05,
      "loss": 0.1336,
      "step": 34120
    },
    {
      "epoch": 13.25,
      "learning_rate": 1.823275080906149e-05,
      "loss": 0.1391,
      "step": 34130
    },
    {
      "epoch": 13.26,
      "learning_rate": 1.823223300970874e-05,
      "loss": 0.0655,
      "step": 34140
    },
    {
      "epoch": 13.26,
      "learning_rate": 1.823171521035599e-05,
      "loss": 0.264,
      "step": 34150
    },
    {
      "epoch": 13.27,
      "learning_rate": 1.823119741100324e-05,
      "loss": 0.2483,
      "step": 34160
    },
    {
      "epoch": 13.27,
      "learning_rate": 1.8230679611650485e-05,
      "loss": 0.1062,
      "step": 34170
    },
    {
      "epoch": 13.27,
      "learning_rate": 1.8230161812297736e-05,
      "loss": 0.2293,
      "step": 34180
    },
    {
      "epoch": 13.28,
      "learning_rate": 1.8229644012944984e-05,
      "loss": 0.1499,
      "step": 34190
    },
    {
      "epoch": 13.28,
      "learning_rate": 1.8229126213592235e-05,
      "loss": 0.1994,
      "step": 34200
    },
    {
      "epoch": 13.29,
      "learning_rate": 1.8228608414239483e-05,
      "loss": 0.0719,
      "step": 34210
    },
    {
      "epoch": 13.29,
      "learning_rate": 1.8228090614886734e-05,
      "loss": 0.1006,
      "step": 34220
    },
    {
      "epoch": 13.29,
      "learning_rate": 1.8227572815533982e-05,
      "loss": 0.1666,
      "step": 34230
    },
    {
      "epoch": 13.3,
      "learning_rate": 1.8227055016181234e-05,
      "loss": 0.1722,
      "step": 34240
    },
    {
      "epoch": 13.3,
      "learning_rate": 1.822653721682848e-05,
      "loss": 0.1765,
      "step": 34250
    },
    {
      "epoch": 13.3,
      "learning_rate": 1.822601941747573e-05,
      "loss": 0.1195,
      "step": 34260
    },
    {
      "epoch": 13.31,
      "learning_rate": 1.8225501618122977e-05,
      "loss": 0.2106,
      "step": 34270
    },
    {
      "epoch": 13.31,
      "learning_rate": 1.822498381877023e-05,
      "loss": 0.1151,
      "step": 34280
    },
    {
      "epoch": 13.32,
      "learning_rate": 1.8224466019417477e-05,
      "loss": 0.2099,
      "step": 34290
    },
    {
      "epoch": 13.32,
      "learning_rate": 1.8223948220064728e-05,
      "loss": 0.0759,
      "step": 34300
    },
    {
      "epoch": 13.32,
      "learning_rate": 1.8223430420711976e-05,
      "loss": 0.1435,
      "step": 34310
    },
    {
      "epoch": 13.33,
      "learning_rate": 1.8222912621359227e-05,
      "loss": 0.2181,
      "step": 34320
    },
    {
      "epoch": 13.33,
      "learning_rate": 1.8222394822006475e-05,
      "loss": 0.1993,
      "step": 34330
    },
    {
      "epoch": 13.34,
      "learning_rate": 1.8221877022653723e-05,
      "loss": 0.101,
      "step": 34340
    },
    {
      "epoch": 13.34,
      "learning_rate": 1.822135922330097e-05,
      "loss": 0.134,
      "step": 34350
    },
    {
      "epoch": 13.34,
      "learning_rate": 1.8220841423948222e-05,
      "loss": 0.1501,
      "step": 34360
    },
    {
      "epoch": 13.35,
      "learning_rate": 1.822032362459547e-05,
      "loss": 0.2161,
      "step": 34370
    },
    {
      "epoch": 13.35,
      "learning_rate": 1.821980582524272e-05,
      "loss": 0.1767,
      "step": 34380
    },
    {
      "epoch": 13.36,
      "learning_rate": 1.821928802588997e-05,
      "loss": 0.0727,
      "step": 34390
    },
    {
      "epoch": 13.36,
      "learning_rate": 1.821877022653722e-05,
      "loss": 0.1513,
      "step": 34400
    },
    {
      "epoch": 13.36,
      "learning_rate": 1.821825242718447e-05,
      "loss": 0.1426,
      "step": 34410
    },
    {
      "epoch": 13.37,
      "learning_rate": 1.8217734627831716e-05,
      "loss": 0.1141,
      "step": 34420
    },
    {
      "epoch": 13.37,
      "learning_rate": 1.8217216828478964e-05,
      "loss": 0.073,
      "step": 34430
    },
    {
      "epoch": 13.37,
      "learning_rate": 1.8216699029126216e-05,
      "loss": 0.1668,
      "step": 34440
    },
    {
      "epoch": 13.38,
      "learning_rate": 1.8216181229773464e-05,
      "loss": 0.1802,
      "step": 34450
    },
    {
      "epoch": 13.38,
      "learning_rate": 1.8215663430420715e-05,
      "loss": 0.1654,
      "step": 34460
    },
    {
      "epoch": 13.39,
      "learning_rate": 1.8215145631067963e-05,
      "loss": 0.1481,
      "step": 34470
    },
    {
      "epoch": 13.39,
      "learning_rate": 1.8214627831715214e-05,
      "loss": 0.2609,
      "step": 34480
    },
    {
      "epoch": 13.39,
      "learning_rate": 1.8214110032362462e-05,
      "loss": 0.0737,
      "step": 34490
    },
    {
      "epoch": 13.4,
      "learning_rate": 1.821359223300971e-05,
      "loss": 0.0588,
      "step": 34500
    },
    {
      "epoch": 13.4,
      "learning_rate": 1.8213074433656958e-05,
      "loss": 0.188,
      "step": 34510
    },
    {
      "epoch": 13.41,
      "learning_rate": 1.821255663430421e-05,
      "loss": 0.1259,
      "step": 34520
    },
    {
      "epoch": 13.41,
      "learning_rate": 1.8212038834951457e-05,
      "loss": 0.1118,
      "step": 34530
    },
    {
      "epoch": 13.41,
      "learning_rate": 1.821152103559871e-05,
      "loss": 0.0493,
      "step": 34540
    },
    {
      "epoch": 13.42,
      "learning_rate": 1.8211003236245956e-05,
      "loss": 0.0744,
      "step": 34550
    },
    {
      "epoch": 13.42,
      "learning_rate": 1.8210485436893204e-05,
      "loss": 0.2036,
      "step": 34560
    },
    {
      "epoch": 13.43,
      "learning_rate": 1.8209967637540456e-05,
      "loss": 0.1385,
      "step": 34570
    },
    {
      "epoch": 13.43,
      "learning_rate": 1.8209449838187704e-05,
      "loss": 0.1546,
      "step": 34580
    },
    {
      "epoch": 13.43,
      "learning_rate": 1.820893203883495e-05,
      "loss": 0.1128,
      "step": 34590
    },
    {
      "epoch": 13.44,
      "learning_rate": 1.8208414239482203e-05,
      "loss": 0.0663,
      "step": 34600
    },
    {
      "epoch": 13.44,
      "learning_rate": 1.820789644012945e-05,
      "loss": 0.0962,
      "step": 34610
    },
    {
      "epoch": 13.44,
      "learning_rate": 1.8207378640776702e-05,
      "loss": 0.0853,
      "step": 34620
    },
    {
      "epoch": 13.45,
      "learning_rate": 1.820686084142395e-05,
      "loss": 0.2256,
      "step": 34630
    },
    {
      "epoch": 13.45,
      "learning_rate": 1.8206343042071198e-05,
      "loss": 0.0447,
      "step": 34640
    },
    {
      "epoch": 13.46,
      "learning_rate": 1.820582524271845e-05,
      "loss": 0.1325,
      "step": 34650
    },
    {
      "epoch": 13.46,
      "learning_rate": 1.8205307443365697e-05,
      "loss": 0.0272,
      "step": 34660
    },
    {
      "epoch": 13.46,
      "learning_rate": 1.8204789644012945e-05,
      "loss": 0.016,
      "step": 34670
    },
    {
      "epoch": 13.47,
      "learning_rate": 1.8204271844660196e-05,
      "loss": 0.2529,
      "step": 34680
    },
    {
      "epoch": 13.47,
      "learning_rate": 1.8203754045307444e-05,
      "loss": 0.1998,
      "step": 34690
    },
    {
      "epoch": 13.48,
      "learning_rate": 1.8203236245954696e-05,
      "loss": 0.1606,
      "step": 34700
    },
    {
      "epoch": 13.48,
      "learning_rate": 1.8202718446601944e-05,
      "loss": 0.0831,
      "step": 34710
    },
    {
      "epoch": 13.48,
      "learning_rate": 1.820220064724919e-05,
      "loss": 0.004,
      "step": 34720
    },
    {
      "epoch": 13.49,
      "learning_rate": 1.8201682847896443e-05,
      "loss": 0.085,
      "step": 34730
    },
    {
      "epoch": 13.49,
      "learning_rate": 1.820116504854369e-05,
      "loss": 0.0762,
      "step": 34740
    },
    {
      "epoch": 13.5,
      "learning_rate": 1.820064724919094e-05,
      "loss": 0.166,
      "step": 34750
    },
    {
      "epoch": 13.5,
      "learning_rate": 1.820012944983819e-05,
      "loss": 0.1404,
      "step": 34760
    },
    {
      "epoch": 13.5,
      "learning_rate": 1.8199611650485438e-05,
      "loss": 0.1024,
      "step": 34770
    },
    {
      "epoch": 13.51,
      "learning_rate": 1.819909385113269e-05,
      "loss": 0.0992,
      "step": 34780
    },
    {
      "epoch": 13.51,
      "learning_rate": 1.8198576051779937e-05,
      "loss": 0.1426,
      "step": 34790
    },
    {
      "epoch": 13.51,
      "learning_rate": 1.8198058252427185e-05,
      "loss": 0.3103,
      "step": 34800
    },
    {
      "epoch": 13.52,
      "learning_rate": 1.8197540453074436e-05,
      "loss": 0.12,
      "step": 34810
    },
    {
      "epoch": 13.52,
      "learning_rate": 1.8197022653721684e-05,
      "loss": 0.0674,
      "step": 34820
    },
    {
      "epoch": 13.53,
      "learning_rate": 1.8196504854368932e-05,
      "loss": 0.0767,
      "step": 34830
    },
    {
      "epoch": 13.53,
      "learning_rate": 1.8195987055016184e-05,
      "loss": 0.2512,
      "step": 34840
    },
    {
      "epoch": 13.53,
      "learning_rate": 1.819546925566343e-05,
      "loss": 0.1229,
      "step": 34850
    },
    {
      "epoch": 13.54,
      "learning_rate": 1.8194951456310683e-05,
      "loss": 0.1391,
      "step": 34860
    },
    {
      "epoch": 13.54,
      "learning_rate": 1.819443365695793e-05,
      "loss": 0.1324,
      "step": 34870
    },
    {
      "epoch": 13.55,
      "learning_rate": 1.819391585760518e-05,
      "loss": 0.1865,
      "step": 34880
    },
    {
      "epoch": 13.55,
      "learning_rate": 1.819339805825243e-05,
      "loss": 0.1875,
      "step": 34890
    },
    {
      "epoch": 13.55,
      "learning_rate": 1.8192880258899678e-05,
      "loss": 0.1846,
      "step": 34900
    },
    {
      "epoch": 13.56,
      "learning_rate": 1.8192362459546926e-05,
      "loss": 0.0059,
      "step": 34910
    },
    {
      "epoch": 13.56,
      "learning_rate": 1.8191844660194177e-05,
      "loss": 0.4132,
      "step": 34920
    },
    {
      "epoch": 13.57,
      "learning_rate": 1.8191326860841425e-05,
      "loss": 0.1391,
      "step": 34930
    },
    {
      "epoch": 13.57,
      "learning_rate": 1.8190809061488673e-05,
      "loss": 0.0524,
      "step": 34940
    },
    {
      "epoch": 13.57,
      "learning_rate": 1.8190291262135924e-05,
      "loss": 0.1829,
      "step": 34950
    },
    {
      "epoch": 13.58,
      "learning_rate": 1.8189773462783172e-05,
      "loss": 0.083,
      "step": 34960
    },
    {
      "epoch": 13.58,
      "learning_rate": 1.8189255663430423e-05,
      "loss": 0.2328,
      "step": 34970
    },
    {
      "epoch": 13.58,
      "learning_rate": 1.818873786407767e-05,
      "loss": 0.191,
      "step": 34980
    },
    {
      "epoch": 13.59,
      "learning_rate": 1.818822006472492e-05,
      "loss": 0.1234,
      "step": 34990
    },
    {
      "epoch": 13.59,
      "learning_rate": 1.818770226537217e-05,
      "loss": 0.1355,
      "step": 35000
    },
    {
      "epoch": 13.6,
      "learning_rate": 1.818718446601942e-05,
      "loss": 0.209,
      "step": 35010
    },
    {
      "epoch": 13.6,
      "learning_rate": 1.8186666666666666e-05,
      "loss": 0.0095,
      "step": 35020
    },
    {
      "epoch": 13.6,
      "learning_rate": 1.8186148867313918e-05,
      "loss": 0.1791,
      "step": 35030
    },
    {
      "epoch": 13.61,
      "learning_rate": 1.8185631067961166e-05,
      "loss": 0.1262,
      "step": 35040
    },
    {
      "epoch": 13.61,
      "learning_rate": 1.8185113268608417e-05,
      "loss": 0.1034,
      "step": 35050
    },
    {
      "epoch": 13.62,
      "learning_rate": 1.8184595469255665e-05,
      "loss": 0.0421,
      "step": 35060
    },
    {
      "epoch": 13.62,
      "learning_rate": 1.8184077669902913e-05,
      "loss": 0.2702,
      "step": 35070
    },
    {
      "epoch": 13.62,
      "learning_rate": 1.8183559870550164e-05,
      "loss": 0.071,
      "step": 35080
    },
    {
      "epoch": 13.63,
      "learning_rate": 1.8183042071197412e-05,
      "loss": 0.2968,
      "step": 35090
    },
    {
      "epoch": 13.63,
      "learning_rate": 1.818252427184466e-05,
      "loss": 0.1775,
      "step": 35100
    },
    {
      "epoch": 13.63,
      "learning_rate": 1.818200647249191e-05,
      "loss": 0.1325,
      "step": 35110
    },
    {
      "epoch": 13.64,
      "learning_rate": 1.818148867313916e-05,
      "loss": 0.1273,
      "step": 35120
    },
    {
      "epoch": 13.64,
      "learning_rate": 1.818097087378641e-05,
      "loss": 0.1228,
      "step": 35130
    },
    {
      "epoch": 13.65,
      "learning_rate": 1.818045307443366e-05,
      "loss": 0.1008,
      "step": 35140
    },
    {
      "epoch": 13.65,
      "learning_rate": 1.8179935275080906e-05,
      "loss": 0.2047,
      "step": 35150
    },
    {
      "epoch": 13.65,
      "learning_rate": 1.8179417475728158e-05,
      "loss": 0.1245,
      "step": 35160
    },
    {
      "epoch": 13.66,
      "learning_rate": 1.8178899676375406e-05,
      "loss": 0.1787,
      "step": 35170
    },
    {
      "epoch": 13.66,
      "learning_rate": 1.8178381877022654e-05,
      "loss": 0.0492,
      "step": 35180
    },
    {
      "epoch": 13.67,
      "learning_rate": 1.8177864077669905e-05,
      "loss": 0.176,
      "step": 35190
    },
    {
      "epoch": 13.67,
      "learning_rate": 1.8177346278317153e-05,
      "loss": 0.1849,
      "step": 35200
    },
    {
      "epoch": 13.67,
      "learning_rate": 1.8176828478964404e-05,
      "loss": 0.3279,
      "step": 35210
    },
    {
      "epoch": 13.68,
      "learning_rate": 1.8176310679611652e-05,
      "loss": 0.1023,
      "step": 35220
    },
    {
      "epoch": 13.68,
      "learning_rate": 1.81757928802589e-05,
      "loss": 0.2586,
      "step": 35230
    },
    {
      "epoch": 13.69,
      "learning_rate": 1.8175275080906148e-05,
      "loss": 0.1488,
      "step": 35240
    },
    {
      "epoch": 13.69,
      "learning_rate": 1.81747572815534e-05,
      "loss": 0.0795,
      "step": 35250
    },
    {
      "epoch": 13.69,
      "learning_rate": 1.8174239482200647e-05,
      "loss": 0.0305,
      "step": 35260
    },
    {
      "epoch": 13.7,
      "learning_rate": 1.81737216828479e-05,
      "loss": 0.0984,
      "step": 35270
    },
    {
      "epoch": 13.7,
      "learning_rate": 1.8173203883495146e-05,
      "loss": 0.164,
      "step": 35280
    },
    {
      "epoch": 13.7,
      "learning_rate": 1.8172686084142398e-05,
      "loss": 0.2855,
      "step": 35290
    },
    {
      "epoch": 13.71,
      "learning_rate": 1.8172168284789646e-05,
      "loss": 0.2669,
      "step": 35300
    },
    {
      "epoch": 13.71,
      "learning_rate": 1.8171650485436894e-05,
      "loss": 0.0752,
      "step": 35310
    },
    {
      "epoch": 13.72,
      "learning_rate": 1.817113268608414e-05,
      "loss": 0.1419,
      "step": 35320
    },
    {
      "epoch": 13.72,
      "learning_rate": 1.8170614886731393e-05,
      "loss": 0.0263,
      "step": 35330
    },
    {
      "epoch": 13.72,
      "learning_rate": 1.817009708737864e-05,
      "loss": 0.1472,
      "step": 35340
    },
    {
      "epoch": 13.73,
      "learning_rate": 1.8169579288025892e-05,
      "loss": 0.0431,
      "step": 35350
    },
    {
      "epoch": 13.73,
      "learning_rate": 1.816906148867314e-05,
      "loss": 0.1179,
      "step": 35360
    },
    {
      "epoch": 13.74,
      "learning_rate": 1.816854368932039e-05,
      "loss": 0.0734,
      "step": 35370
    },
    {
      "epoch": 13.74,
      "learning_rate": 1.816802588996764e-05,
      "loss": 0.1569,
      "step": 35380
    },
    {
      "epoch": 13.74,
      "learning_rate": 1.816750809061489e-05,
      "loss": 0.0364,
      "step": 35390
    },
    {
      "epoch": 13.75,
      "learning_rate": 1.8166990291262135e-05,
      "loss": 0.1466,
      "step": 35400
    },
    {
      "epoch": 13.75,
      "learning_rate": 1.8166472491909386e-05,
      "loss": 0.0613,
      "step": 35410
    },
    {
      "epoch": 13.76,
      "learning_rate": 1.8165954692556634e-05,
      "loss": 0.1592,
      "step": 35420
    },
    {
      "epoch": 13.76,
      "learning_rate": 1.8165436893203886e-05,
      "loss": 0.0692,
      "step": 35430
    },
    {
      "epoch": 13.76,
      "learning_rate": 1.8164919093851133e-05,
      "loss": 0.3594,
      "step": 35440
    },
    {
      "epoch": 13.77,
      "learning_rate": 1.8164401294498385e-05,
      "loss": 0.1592,
      "step": 35450
    },
    {
      "epoch": 13.77,
      "learning_rate": 1.8163883495145633e-05,
      "loss": 0.0616,
      "step": 35460
    },
    {
      "epoch": 13.77,
      "learning_rate": 1.8163365695792884e-05,
      "loss": 0.2361,
      "step": 35470
    },
    {
      "epoch": 13.78,
      "learning_rate": 1.816284789644013e-05,
      "loss": 0.0818,
      "step": 35480
    },
    {
      "epoch": 13.78,
      "learning_rate": 1.816233009708738e-05,
      "loss": 0.1398,
      "step": 35490
    },
    {
      "epoch": 13.79,
      "learning_rate": 1.8161812297734628e-05,
      "loss": 0.1174,
      "step": 35500
    },
    {
      "epoch": 13.79,
      "learning_rate": 1.816129449838188e-05,
      "loss": 0.0332,
      "step": 35510
    },
    {
      "epoch": 13.79,
      "learning_rate": 1.8160776699029127e-05,
      "loss": 0.1157,
      "step": 35520
    },
    {
      "epoch": 13.8,
      "learning_rate": 1.816025889967638e-05,
      "loss": 0.1664,
      "step": 35530
    },
    {
      "epoch": 13.8,
      "learning_rate": 1.8159741100323626e-05,
      "loss": 0.2075,
      "step": 35540
    },
    {
      "epoch": 13.81,
      "learning_rate": 1.8159223300970878e-05,
      "loss": 0.0129,
      "step": 35550
    },
    {
      "epoch": 13.81,
      "learning_rate": 1.8158705501618122e-05,
      "loss": 0.0299,
      "step": 35560
    },
    {
      "epoch": 13.81,
      "learning_rate": 1.8158187702265373e-05,
      "loss": 0.0519,
      "step": 35570
    },
    {
      "epoch": 13.82,
      "learning_rate": 1.815766990291262e-05,
      "loss": 0.1967,
      "step": 35580
    },
    {
      "epoch": 13.82,
      "learning_rate": 1.8157152103559873e-05,
      "loss": 0.1316,
      "step": 35590
    },
    {
      "epoch": 13.83,
      "learning_rate": 1.815663430420712e-05,
      "loss": 0.2539,
      "step": 35600
    },
    {
      "epoch": 13.83,
      "learning_rate": 1.8156116504854372e-05,
      "loss": 0.201,
      "step": 35610
    },
    {
      "epoch": 13.83,
      "learning_rate": 1.815559870550162e-05,
      "loss": 0.1136,
      "step": 35620
    },
    {
      "epoch": 13.84,
      "learning_rate": 1.815508090614887e-05,
      "loss": 0.1167,
      "step": 35630
    },
    {
      "epoch": 13.84,
      "learning_rate": 1.8154563106796116e-05,
      "loss": 0.1663,
      "step": 35640
    },
    {
      "epoch": 13.84,
      "learning_rate": 1.8154045307443367e-05,
      "loss": 0.131,
      "step": 35650
    },
    {
      "epoch": 13.85,
      "learning_rate": 1.8153527508090615e-05,
      "loss": 0.135,
      "step": 35660
    },
    {
      "epoch": 13.85,
      "learning_rate": 1.8153009708737866e-05,
      "loss": 0.1261,
      "step": 35670
    },
    {
      "epoch": 13.86,
      "learning_rate": 1.8152491909385114e-05,
      "loss": 0.1435,
      "step": 35680
    },
    {
      "epoch": 13.86,
      "learning_rate": 1.8151974110032365e-05,
      "loss": 0.0684,
      "step": 35690
    },
    {
      "epoch": 13.86,
      "learning_rate": 1.8151456310679613e-05,
      "loss": 0.0289,
      "step": 35700
    },
    {
      "epoch": 13.87,
      "learning_rate": 1.8150938511326865e-05,
      "loss": 0.0978,
      "step": 35710
    },
    {
      "epoch": 13.87,
      "learning_rate": 1.815042071197411e-05,
      "loss": 0.0758,
      "step": 35720
    },
    {
      "epoch": 13.88,
      "learning_rate": 1.814990291262136e-05,
      "loss": 0.2285,
      "step": 35730
    },
    {
      "epoch": 13.88,
      "learning_rate": 1.814938511326861e-05,
      "loss": 0.1023,
      "step": 35740
    },
    {
      "epoch": 13.88,
      "learning_rate": 1.814886731391586e-05,
      "loss": 0.0807,
      "step": 35750
    },
    {
      "epoch": 13.89,
      "learning_rate": 1.8148349514563108e-05,
      "loss": 0.1097,
      "step": 35760
    },
    {
      "epoch": 13.89,
      "learning_rate": 1.814783171521036e-05,
      "loss": 0.341,
      "step": 35770
    },
    {
      "epoch": 13.9,
      "learning_rate": 1.8147313915857607e-05,
      "loss": 0.1208,
      "step": 35780
    },
    {
      "epoch": 13.9,
      "learning_rate": 1.8146796116504858e-05,
      "loss": 0.1371,
      "step": 35790
    },
    {
      "epoch": 13.9,
      "learning_rate": 1.8146278317152103e-05,
      "loss": 0.0318,
      "step": 35800
    },
    {
      "epoch": 13.91,
      "learning_rate": 1.8145760517799354e-05,
      "loss": 0.0884,
      "step": 35810
    },
    {
      "epoch": 13.91,
      "learning_rate": 1.8145242718446602e-05,
      "loss": 0.1914,
      "step": 35820
    },
    {
      "epoch": 13.91,
      "learning_rate": 1.8144724919093853e-05,
      "loss": 0.2362,
      "step": 35830
    },
    {
      "epoch": 13.92,
      "learning_rate": 1.81442071197411e-05,
      "loss": 0.1666,
      "step": 35840
    },
    {
      "epoch": 13.92,
      "learning_rate": 1.8143689320388353e-05,
      "loss": 0.135,
      "step": 35850
    },
    {
      "epoch": 13.93,
      "learning_rate": 1.81431715210356e-05,
      "loss": 0.1548,
      "step": 35860
    },
    {
      "epoch": 13.93,
      "learning_rate": 1.8142653721682852e-05,
      "loss": 0.1692,
      "step": 35870
    },
    {
      "epoch": 13.93,
      "learning_rate": 1.81421359223301e-05,
      "loss": 0.0222,
      "step": 35880
    },
    {
      "epoch": 13.94,
      "learning_rate": 1.8141618122977348e-05,
      "loss": 0.1975,
      "step": 35890
    },
    {
      "epoch": 13.94,
      "learning_rate": 1.8141100323624596e-05,
      "loss": 0.1991,
      "step": 35900
    },
    {
      "epoch": 13.95,
      "learning_rate": 1.8140582524271847e-05,
      "loss": 0.1924,
      "step": 35910
    },
    {
      "epoch": 13.95,
      "learning_rate": 1.8140064724919095e-05,
      "loss": 0.1797,
      "step": 35920
    },
    {
      "epoch": 13.95,
      "learning_rate": 1.8139546925566346e-05,
      "loss": 0.2069,
      "step": 35930
    },
    {
      "epoch": 13.96,
      "learning_rate": 1.8139029126213594e-05,
      "loss": 0.4265,
      "step": 35940
    },
    {
      "epoch": 13.96,
      "learning_rate": 1.8138511326860845e-05,
      "loss": 0.0579,
      "step": 35950
    },
    {
      "epoch": 13.97,
      "learning_rate": 1.8137993527508093e-05,
      "loss": 0.2866,
      "step": 35960
    },
    {
      "epoch": 13.97,
      "learning_rate": 1.813747572815534e-05,
      "loss": 0.2069,
      "step": 35970
    },
    {
      "epoch": 13.97,
      "learning_rate": 1.813695792880259e-05,
      "loss": 0.1737,
      "step": 35980
    },
    {
      "epoch": 13.98,
      "learning_rate": 1.813644012944984e-05,
      "loss": 0.0397,
      "step": 35990
    },
    {
      "epoch": 13.98,
      "learning_rate": 1.813592233009709e-05,
      "loss": 0.133,
      "step": 36000
    },
    {
      "epoch": 13.98,
      "learning_rate": 1.813540453074434e-05,
      "loss": 0.1399,
      "step": 36010
    },
    {
      "epoch": 13.99,
      "learning_rate": 1.8134886731391588e-05,
      "loss": 0.168,
      "step": 36020
    },
    {
      "epoch": 13.99,
      "learning_rate": 1.8134368932038836e-05,
      "loss": 0.1163,
      "step": 36030
    },
    {
      "epoch": 14.0,
      "learning_rate": 1.8133851132686087e-05,
      "loss": 0.1748,
      "step": 36040
    },
    {
      "epoch": 14.0,
      "learning_rate": 1.8133333333333335e-05,
      "loss": 0.0179,
      "step": 36050
    },
    {
      "epoch": 14.0,
      "eval_accuracy": 0.9482806052269601,
      "eval_loss": 0.22397606074810028,
      "eval_runtime": 8.2572,
      "eval_samples_per_second": 440.222,
      "eval_steps_per_second": 55.103,
      "step": 36050
    },
    {
      "epoch": 14.0,
      "learning_rate": 1.8132815533980583e-05,
      "loss": 0.1235,
      "step": 36060
    },
    {
      "epoch": 14.01,
      "learning_rate": 1.8132297734627834e-05,
      "loss": 0.0825,
      "step": 36070
    },
    {
      "epoch": 14.01,
      "learning_rate": 1.8131779935275082e-05,
      "loss": 0.1312,
      "step": 36080
    },
    {
      "epoch": 14.02,
      "learning_rate": 1.8131262135922333e-05,
      "loss": 0.0936,
      "step": 36090
    },
    {
      "epoch": 14.02,
      "learning_rate": 1.813074433656958e-05,
      "loss": 0.0494,
      "step": 36100
    },
    {
      "epoch": 14.02,
      "learning_rate": 1.813022653721683e-05,
      "loss": 0.2018,
      "step": 36110
    },
    {
      "epoch": 14.03,
      "learning_rate": 1.812970873786408e-05,
      "loss": 0.1093,
      "step": 36120
    },
    {
      "epoch": 14.03,
      "learning_rate": 1.812919093851133e-05,
      "loss": 0.2956,
      "step": 36130
    },
    {
      "epoch": 14.03,
      "learning_rate": 1.8128673139158576e-05,
      "loss": 0.0853,
      "step": 36140
    },
    {
      "epoch": 14.04,
      "learning_rate": 1.8128155339805828e-05,
      "loss": 0.0711,
      "step": 36150
    },
    {
      "epoch": 14.04,
      "learning_rate": 1.8127637540453075e-05,
      "loss": 0.098,
      "step": 36160
    },
    {
      "epoch": 14.05,
      "learning_rate": 1.8127119741100327e-05,
      "loss": 0.1364,
      "step": 36170
    },
    {
      "epoch": 14.05,
      "learning_rate": 1.8126601941747575e-05,
      "loss": 0.1451,
      "step": 36180
    },
    {
      "epoch": 14.05,
      "learning_rate": 1.8126084142394823e-05,
      "loss": 0.0825,
      "step": 36190
    },
    {
      "epoch": 14.06,
      "learning_rate": 1.8125566343042074e-05,
      "loss": 0.1643,
      "step": 36200
    },
    {
      "epoch": 14.06,
      "learning_rate": 1.8125048543689322e-05,
      "loss": 0.1,
      "step": 36210
    },
    {
      "epoch": 14.07,
      "learning_rate": 1.812453074433657e-05,
      "loss": 0.2269,
      "step": 36220
    },
    {
      "epoch": 14.07,
      "learning_rate": 1.812401294498382e-05,
      "loss": 0.2601,
      "step": 36230
    },
    {
      "epoch": 14.07,
      "learning_rate": 1.812349514563107e-05,
      "loss": 0.1232,
      "step": 36240
    },
    {
      "epoch": 14.08,
      "learning_rate": 1.812297734627832e-05,
      "loss": 0.0788,
      "step": 36250
    },
    {
      "epoch": 14.08,
      "learning_rate": 1.8122459546925568e-05,
      "loss": 0.159,
      "step": 36260
    },
    {
      "epoch": 14.09,
      "learning_rate": 1.8121941747572816e-05,
      "loss": 0.1311,
      "step": 36270
    },
    {
      "epoch": 14.09,
      "learning_rate": 1.8121423948220068e-05,
      "loss": 0.0923,
      "step": 36280
    },
    {
      "epoch": 14.09,
      "learning_rate": 1.8120906148867315e-05,
      "loss": 0.1791,
      "step": 36290
    },
    {
      "epoch": 14.1,
      "learning_rate": 1.8120388349514563e-05,
      "loss": 0.1007,
      "step": 36300
    },
    {
      "epoch": 14.1,
      "learning_rate": 1.8119870550161815e-05,
      "loss": 0.2406,
      "step": 36310
    },
    {
      "epoch": 14.1,
      "learning_rate": 1.8119352750809063e-05,
      "loss": 0.0147,
      "step": 36320
    },
    {
      "epoch": 14.11,
      "learning_rate": 1.811883495145631e-05,
      "loss": 0.0484,
      "step": 36330
    },
    {
      "epoch": 14.11,
      "learning_rate": 1.8118317152103562e-05,
      "loss": 0.1523,
      "step": 36340
    },
    {
      "epoch": 14.12,
      "learning_rate": 1.811779935275081e-05,
      "loss": 0.0185,
      "step": 36350
    },
    {
      "epoch": 14.12,
      "learning_rate": 1.811728155339806e-05,
      "loss": 0.0667,
      "step": 36360
    },
    {
      "epoch": 14.12,
      "learning_rate": 1.811676375404531e-05,
      "loss": 0.0815,
      "step": 36370
    },
    {
      "epoch": 14.13,
      "learning_rate": 1.8116245954692557e-05,
      "loss": 0.0997,
      "step": 36380
    },
    {
      "epoch": 14.13,
      "learning_rate": 1.8115728155339808e-05,
      "loss": 0.1242,
      "step": 36390
    },
    {
      "epoch": 14.14,
      "learning_rate": 1.8115210355987056e-05,
      "loss": 0.1923,
      "step": 36400
    },
    {
      "epoch": 14.14,
      "learning_rate": 1.8114692556634304e-05,
      "loss": 0.0938,
      "step": 36410
    },
    {
      "epoch": 14.14,
      "learning_rate": 1.8114174757281555e-05,
      "loss": 0.0798,
      "step": 36420
    },
    {
      "epoch": 14.15,
      "learning_rate": 1.8113656957928803e-05,
      "loss": 0.1797,
      "step": 36430
    },
    {
      "epoch": 14.15,
      "learning_rate": 1.8113139158576055e-05,
      "loss": 0.0905,
      "step": 36440
    },
    {
      "epoch": 14.16,
      "learning_rate": 1.8112621359223303e-05,
      "loss": 0.0905,
      "step": 36450
    },
    {
      "epoch": 14.16,
      "learning_rate": 1.811210355987055e-05,
      "loss": 0.2132,
      "step": 36460
    },
    {
      "epoch": 14.16,
      "learning_rate": 1.8111585760517802e-05,
      "loss": 0.1924,
      "step": 36470
    },
    {
      "epoch": 14.17,
      "learning_rate": 1.811106796116505e-05,
      "loss": 0.2039,
      "step": 36480
    },
    {
      "epoch": 14.17,
      "learning_rate": 1.8110550161812298e-05,
      "loss": 0.1392,
      "step": 36490
    },
    {
      "epoch": 14.17,
      "learning_rate": 1.811003236245955e-05,
      "loss": 0.0946,
      "step": 36500
    },
    {
      "epoch": 14.18,
      "learning_rate": 1.8109514563106797e-05,
      "loss": 0.0832,
      "step": 36510
    },
    {
      "epoch": 14.18,
      "learning_rate": 1.8108996763754048e-05,
      "loss": 0.2902,
      "step": 36520
    },
    {
      "epoch": 14.19,
      "learning_rate": 1.8108478964401296e-05,
      "loss": 0.0471,
      "step": 36530
    },
    {
      "epoch": 14.19,
      "learning_rate": 1.8107961165048544e-05,
      "loss": 0.0537,
      "step": 36540
    },
    {
      "epoch": 14.19,
      "learning_rate": 1.8107443365695795e-05,
      "loss": 0.2078,
      "step": 36550
    },
    {
      "epoch": 14.2,
      "learning_rate": 1.8106925566343043e-05,
      "loss": 0.0786,
      "step": 36560
    },
    {
      "epoch": 14.2,
      "learning_rate": 1.810640776699029e-05,
      "loss": 0.163,
      "step": 36570
    },
    {
      "epoch": 14.21,
      "learning_rate": 1.8105889967637543e-05,
      "loss": 0.0875,
      "step": 36580
    },
    {
      "epoch": 14.21,
      "learning_rate": 1.810537216828479e-05,
      "loss": 0.0139,
      "step": 36590
    },
    {
      "epoch": 14.21,
      "learning_rate": 1.8104854368932042e-05,
      "loss": 0.048,
      "step": 36600
    },
    {
      "epoch": 14.22,
      "learning_rate": 1.810433656957929e-05,
      "loss": 0.1637,
      "step": 36610
    },
    {
      "epoch": 14.22,
      "learning_rate": 1.8103818770226538e-05,
      "loss": 0.1923,
      "step": 36620
    },
    {
      "epoch": 14.23,
      "learning_rate": 1.810330097087379e-05,
      "loss": 0.0557,
      "step": 36630
    },
    {
      "epoch": 14.23,
      "learning_rate": 1.8102783171521037e-05,
      "loss": 0.1769,
      "step": 36640
    },
    {
      "epoch": 14.23,
      "learning_rate": 1.8102265372168285e-05,
      "loss": 0.087,
      "step": 36650
    },
    {
      "epoch": 14.24,
      "learning_rate": 1.8101747572815536e-05,
      "loss": 0.1318,
      "step": 36660
    },
    {
      "epoch": 14.24,
      "learning_rate": 1.8101229773462784e-05,
      "loss": 0.1148,
      "step": 36670
    },
    {
      "epoch": 14.24,
      "learning_rate": 1.8100711974110035e-05,
      "loss": 0.1719,
      "step": 36680
    },
    {
      "epoch": 14.25,
      "learning_rate": 1.8100194174757283e-05,
      "loss": 0.2312,
      "step": 36690
    },
    {
      "epoch": 14.25,
      "learning_rate": 1.809967637540453e-05,
      "loss": 0.0628,
      "step": 36700
    },
    {
      "epoch": 14.26,
      "learning_rate": 1.809915857605178e-05,
      "loss": 0.0871,
      "step": 36710
    },
    {
      "epoch": 14.26,
      "learning_rate": 1.809864077669903e-05,
      "loss": 0.0729,
      "step": 36720
    },
    {
      "epoch": 14.26,
      "learning_rate": 1.8098122977346278e-05,
      "loss": 0.0805,
      "step": 36730
    },
    {
      "epoch": 14.27,
      "learning_rate": 1.809760517799353e-05,
      "loss": 0.1184,
      "step": 36740
    },
    {
      "epoch": 14.27,
      "learning_rate": 1.8097087378640778e-05,
      "loss": 0.1463,
      "step": 36750
    },
    {
      "epoch": 14.28,
      "learning_rate": 1.809656957928803e-05,
      "loss": 0.0865,
      "step": 36760
    },
    {
      "epoch": 14.28,
      "learning_rate": 1.8096051779935277e-05,
      "loss": 0.1378,
      "step": 36770
    },
    {
      "epoch": 14.28,
      "learning_rate": 1.8095533980582525e-05,
      "loss": 0.0954,
      "step": 36780
    },
    {
      "epoch": 14.29,
      "learning_rate": 1.8095016181229773e-05,
      "loss": 0.0783,
      "step": 36790
    },
    {
      "epoch": 14.29,
      "learning_rate": 1.8094498381877024e-05,
      "loss": 0.1978,
      "step": 36800
    },
    {
      "epoch": 14.3,
      "learning_rate": 1.8093980582524272e-05,
      "loss": 0.2386,
      "step": 36810
    },
    {
      "epoch": 14.3,
      "learning_rate": 1.8093462783171523e-05,
      "loss": 0.1929,
      "step": 36820
    },
    {
      "epoch": 14.3,
      "learning_rate": 1.809294498381877e-05,
      "loss": 0.0918,
      "step": 36830
    },
    {
      "epoch": 14.31,
      "learning_rate": 1.8092427184466022e-05,
      "loss": 0.0107,
      "step": 36840
    },
    {
      "epoch": 14.31,
      "learning_rate": 1.809190938511327e-05,
      "loss": 0.09,
      "step": 36850
    },
    {
      "epoch": 14.31,
      "learning_rate": 1.8091391585760518e-05,
      "loss": 0.1289,
      "step": 36860
    },
    {
      "epoch": 14.32,
      "learning_rate": 1.8090873786407766e-05,
      "loss": 0.2442,
      "step": 36870
    },
    {
      "epoch": 14.32,
      "learning_rate": 1.8090355987055017e-05,
      "loss": 0.313,
      "step": 36880
    },
    {
      "epoch": 14.33,
      "learning_rate": 1.8089838187702265e-05,
      "loss": 0.1892,
      "step": 36890
    },
    {
      "epoch": 14.33,
      "learning_rate": 1.8089320388349517e-05,
      "loss": 0.1018,
      "step": 36900
    },
    {
      "epoch": 14.33,
      "learning_rate": 1.8088802588996765e-05,
      "loss": 0.0436,
      "step": 36910
    },
    {
      "epoch": 14.34,
      "learning_rate": 1.8088284789644016e-05,
      "loss": 0.1088,
      "step": 36920
    },
    {
      "epoch": 14.34,
      "learning_rate": 1.8087766990291264e-05,
      "loss": 0.1149,
      "step": 36930
    },
    {
      "epoch": 14.35,
      "learning_rate": 1.8087249190938512e-05,
      "loss": 0.1123,
      "step": 36940
    },
    {
      "epoch": 14.35,
      "learning_rate": 1.808673139158576e-05,
      "loss": 0.044,
      "step": 36950
    },
    {
      "epoch": 14.35,
      "learning_rate": 1.808621359223301e-05,
      "loss": 0.0489,
      "step": 36960
    },
    {
      "epoch": 14.36,
      "learning_rate": 1.808569579288026e-05,
      "loss": 0.0751,
      "step": 36970
    },
    {
      "epoch": 14.36,
      "learning_rate": 1.808517799352751e-05,
      "loss": 0.1026,
      "step": 36980
    },
    {
      "epoch": 14.37,
      "learning_rate": 1.8084660194174758e-05,
      "loss": 0.0914,
      "step": 36990
    },
    {
      "epoch": 14.37,
      "learning_rate": 1.808414239482201e-05,
      "loss": 0.0669,
      "step": 37000
    },
    {
      "epoch": 14.37,
      "learning_rate": 1.8083624595469257e-05,
      "loss": 0.1055,
      "step": 37010
    },
    {
      "epoch": 14.38,
      "learning_rate": 1.808310679611651e-05,
      "loss": 0.0614,
      "step": 37020
    },
    {
      "epoch": 14.38,
      "learning_rate": 1.8082588996763753e-05,
      "loss": 0.103,
      "step": 37030
    },
    {
      "epoch": 14.38,
      "learning_rate": 1.8082071197411005e-05,
      "loss": 0.0998,
      "step": 37040
    },
    {
      "epoch": 14.39,
      "learning_rate": 1.8081553398058253e-05,
      "loss": 0.1021,
      "step": 37050
    },
    {
      "epoch": 14.39,
      "learning_rate": 1.8081035598705504e-05,
      "loss": 0.0499,
      "step": 37060
    },
    {
      "epoch": 14.4,
      "learning_rate": 1.8080517799352752e-05,
      "loss": 0.1318,
      "step": 37070
    },
    {
      "epoch": 14.4,
      "learning_rate": 1.8080000000000003e-05,
      "loss": 0.0337,
      "step": 37080
    },
    {
      "epoch": 14.4,
      "learning_rate": 1.807948220064725e-05,
      "loss": 0.1046,
      "step": 37090
    },
    {
      "epoch": 14.41,
      "learning_rate": 1.8078964401294502e-05,
      "loss": 0.0693,
      "step": 37100
    },
    {
      "epoch": 14.41,
      "learning_rate": 1.8078446601941747e-05,
      "loss": 0.0902,
      "step": 37110
    },
    {
      "epoch": 14.42,
      "learning_rate": 1.8077928802588998e-05,
      "loss": 0.1726,
      "step": 37120
    },
    {
      "epoch": 14.42,
      "learning_rate": 1.8077411003236246e-05,
      "loss": 0.1457,
      "step": 37130
    },
    {
      "epoch": 14.42,
      "learning_rate": 1.8076893203883497e-05,
      "loss": 0.316,
      "step": 37140
    },
    {
      "epoch": 14.43,
      "learning_rate": 1.8076375404530745e-05,
      "loss": 0.2566,
      "step": 37150
    },
    {
      "epoch": 14.43,
      "learning_rate": 1.8075857605177997e-05,
      "loss": 0.0387,
      "step": 37160
    },
    {
      "epoch": 14.43,
      "learning_rate": 1.8075339805825245e-05,
      "loss": 0.0984,
      "step": 37170
    },
    {
      "epoch": 14.44,
      "learning_rate": 1.8074822006472496e-05,
      "loss": 0.2197,
      "step": 37180
    },
    {
      "epoch": 14.44,
      "learning_rate": 1.807430420711974e-05,
      "loss": 0.1413,
      "step": 37190
    },
    {
      "epoch": 14.45,
      "learning_rate": 1.8073786407766992e-05,
      "loss": 0.0654,
      "step": 37200
    },
    {
      "epoch": 14.45,
      "learning_rate": 1.807326860841424e-05,
      "loss": 0.1518,
      "step": 37210
    },
    {
      "epoch": 14.45,
      "learning_rate": 1.807275080906149e-05,
      "loss": 0.104,
      "step": 37220
    },
    {
      "epoch": 14.46,
      "learning_rate": 1.807223300970874e-05,
      "loss": 0.0854,
      "step": 37230
    },
    {
      "epoch": 14.46,
      "learning_rate": 1.807171521035599e-05,
      "loss": 0.1246,
      "step": 37240
    },
    {
      "epoch": 14.47,
      "learning_rate": 1.8071197411003238e-05,
      "loss": 0.1575,
      "step": 37250
    },
    {
      "epoch": 14.47,
      "learning_rate": 1.807067961165049e-05,
      "loss": 0.1385,
      "step": 37260
    },
    {
      "epoch": 14.47,
      "learning_rate": 1.8070161812297734e-05,
      "loss": 0.3931,
      "step": 37270
    },
    {
      "epoch": 14.48,
      "learning_rate": 1.8069644012944985e-05,
      "loss": 0.1216,
      "step": 37280
    },
    {
      "epoch": 14.48,
      "learning_rate": 1.8069126213592233e-05,
      "loss": 0.0779,
      "step": 37290
    },
    {
      "epoch": 14.49,
      "learning_rate": 1.8068608414239485e-05,
      "loss": 0.0648,
      "step": 37300
    },
    {
      "epoch": 14.49,
      "learning_rate": 1.8068090614886732e-05,
      "loss": 0.1798,
      "step": 37310
    },
    {
      "epoch": 14.49,
      "learning_rate": 1.8067572815533984e-05,
      "loss": 0.0722,
      "step": 37320
    },
    {
      "epoch": 14.5,
      "learning_rate": 1.806705501618123e-05,
      "loss": 0.1158,
      "step": 37330
    },
    {
      "epoch": 14.5,
      "learning_rate": 1.8066537216828483e-05,
      "loss": 0.1127,
      "step": 37340
    },
    {
      "epoch": 14.5,
      "learning_rate": 1.8066019417475728e-05,
      "loss": 0.0993,
      "step": 37350
    },
    {
      "epoch": 14.51,
      "learning_rate": 1.806550161812298e-05,
      "loss": 0.1285,
      "step": 37360
    },
    {
      "epoch": 14.51,
      "learning_rate": 1.8064983818770227e-05,
      "loss": 0.3027,
      "step": 37370
    },
    {
      "epoch": 14.52,
      "learning_rate": 1.8064466019417478e-05,
      "loss": 0.0685,
      "step": 37380
    },
    {
      "epoch": 14.52,
      "learning_rate": 1.8063948220064726e-05,
      "loss": 0.2246,
      "step": 37390
    },
    {
      "epoch": 14.52,
      "learning_rate": 1.8063430420711977e-05,
      "loss": 0.1423,
      "step": 37400
    },
    {
      "epoch": 14.53,
      "learning_rate": 1.8062912621359225e-05,
      "loss": 0.0529,
      "step": 37410
    },
    {
      "epoch": 14.53,
      "learning_rate": 1.8062394822006477e-05,
      "loss": 0.1809,
      "step": 37420
    },
    {
      "epoch": 14.54,
      "learning_rate": 1.806187702265372e-05,
      "loss": 0.053,
      "step": 37430
    },
    {
      "epoch": 14.54,
      "learning_rate": 1.8061359223300972e-05,
      "loss": 0.0544,
      "step": 37440
    },
    {
      "epoch": 14.54,
      "learning_rate": 1.806084142394822e-05,
      "loss": 0.2064,
      "step": 37450
    },
    {
      "epoch": 14.55,
      "learning_rate": 1.806032362459547e-05,
      "loss": 0.1813,
      "step": 37460
    },
    {
      "epoch": 14.55,
      "learning_rate": 1.805980582524272e-05,
      "loss": 0.1736,
      "step": 37470
    },
    {
      "epoch": 14.56,
      "learning_rate": 1.805928802588997e-05,
      "loss": 0.1937,
      "step": 37480
    },
    {
      "epoch": 14.56,
      "learning_rate": 1.805877022653722e-05,
      "loss": 0.0418,
      "step": 37490
    },
    {
      "epoch": 14.56,
      "learning_rate": 1.8058252427184467e-05,
      "loss": 0.147,
      "step": 37500
    },
    {
      "epoch": 14.57,
      "learning_rate": 1.8057734627831715e-05,
      "loss": 0.1004,
      "step": 37510
    },
    {
      "epoch": 14.57,
      "learning_rate": 1.8057216828478966e-05,
      "loss": 0.0995,
      "step": 37520
    },
    {
      "epoch": 14.57,
      "learning_rate": 1.8056699029126214e-05,
      "loss": 0.074,
      "step": 37530
    },
    {
      "epoch": 14.58,
      "learning_rate": 1.8056181229773465e-05,
      "loss": 0.1279,
      "step": 37540
    },
    {
      "epoch": 14.58,
      "learning_rate": 1.8055663430420713e-05,
      "loss": 0.2038,
      "step": 37550
    },
    {
      "epoch": 14.59,
      "learning_rate": 1.8055145631067964e-05,
      "loss": 0.3478,
      "step": 37560
    },
    {
      "epoch": 14.59,
      "learning_rate": 1.8054627831715212e-05,
      "loss": 0.1836,
      "step": 37570
    },
    {
      "epoch": 14.59,
      "learning_rate": 1.805411003236246e-05,
      "loss": 0.0706,
      "step": 37580
    },
    {
      "epoch": 14.6,
      "learning_rate": 1.805359223300971e-05,
      "loss": 0.2899,
      "step": 37590
    },
    {
      "epoch": 14.6,
      "learning_rate": 1.805307443365696e-05,
      "loss": 0.1112,
      "step": 37600
    },
    {
      "epoch": 14.61,
      "learning_rate": 1.8052556634304207e-05,
      "loss": 0.2469,
      "step": 37610
    },
    {
      "epoch": 14.61,
      "learning_rate": 1.805203883495146e-05,
      "loss": 0.115,
      "step": 37620
    },
    {
      "epoch": 14.61,
      "learning_rate": 1.8051521035598707e-05,
      "loss": 0.1248,
      "step": 37630
    },
    {
      "epoch": 14.62,
      "learning_rate": 1.8051003236245958e-05,
      "loss": 0.0057,
      "step": 37640
    },
    {
      "epoch": 14.62,
      "learning_rate": 1.8050485436893206e-05,
      "loss": 0.0366,
      "step": 37650
    },
    {
      "epoch": 14.63,
      "learning_rate": 1.8049967637540454e-05,
      "loss": 0.2063,
      "step": 37660
    },
    {
      "epoch": 14.63,
      "learning_rate": 1.8049449838187705e-05,
      "loss": 0.2725,
      "step": 37670
    },
    {
      "epoch": 14.63,
      "learning_rate": 1.8048932038834953e-05,
      "loss": 0.1387,
      "step": 37680
    },
    {
      "epoch": 14.64,
      "learning_rate": 1.80484142394822e-05,
      "loss": 0.0462,
      "step": 37690
    },
    {
      "epoch": 14.64,
      "learning_rate": 1.8047896440129452e-05,
      "loss": 0.2134,
      "step": 37700
    },
    {
      "epoch": 14.64,
      "learning_rate": 1.80473786407767e-05,
      "loss": 0.0802,
      "step": 37710
    },
    {
      "epoch": 14.65,
      "learning_rate": 1.804686084142395e-05,
      "loss": 0.276,
      "step": 37720
    },
    {
      "epoch": 14.65,
      "learning_rate": 1.80463430420712e-05,
      "loss": 0.0608,
      "step": 37730
    },
    {
      "epoch": 14.66,
      "learning_rate": 1.8045825242718447e-05,
      "loss": 0.0997,
      "step": 37740
    },
    {
      "epoch": 14.66,
      "learning_rate": 1.80453074433657e-05,
      "loss": 0.1741,
      "step": 37750
    },
    {
      "epoch": 14.66,
      "learning_rate": 1.8044789644012947e-05,
      "loss": 0.0546,
      "step": 37760
    },
    {
      "epoch": 14.67,
      "learning_rate": 1.8044271844660195e-05,
      "loss": 0.0282,
      "step": 37770
    },
    {
      "epoch": 14.67,
      "learning_rate": 1.8043754045307446e-05,
      "loss": 0.1732,
      "step": 37780
    },
    {
      "epoch": 14.68,
      "learning_rate": 1.8043236245954694e-05,
      "loss": 0.1249,
      "step": 37790
    },
    {
      "epoch": 14.68,
      "learning_rate": 1.8042718446601942e-05,
      "loss": 0.1209,
      "step": 37800
    },
    {
      "epoch": 14.68,
      "learning_rate": 1.8042200647249193e-05,
      "loss": 0.0315,
      "step": 37810
    },
    {
      "epoch": 14.69,
      "learning_rate": 1.804168284789644e-05,
      "loss": 0.0918,
      "step": 37820
    },
    {
      "epoch": 14.69,
      "learning_rate": 1.8041165048543692e-05,
      "loss": 0.2115,
      "step": 37830
    },
    {
      "epoch": 14.7,
      "learning_rate": 1.804064724919094e-05,
      "loss": 0.2193,
      "step": 37840
    },
    {
      "epoch": 14.7,
      "learning_rate": 1.8040129449838188e-05,
      "loss": 0.139,
      "step": 37850
    },
    {
      "epoch": 14.7,
      "learning_rate": 1.803961165048544e-05,
      "loss": 0.0947,
      "step": 37860
    },
    {
      "epoch": 14.71,
      "learning_rate": 1.8039093851132687e-05,
      "loss": 0.1034,
      "step": 37870
    },
    {
      "epoch": 14.71,
      "learning_rate": 1.8038576051779935e-05,
      "loss": 0.0765,
      "step": 37880
    },
    {
      "epoch": 14.71,
      "learning_rate": 1.8038058252427187e-05,
      "loss": 0.1034,
      "step": 37890
    },
    {
      "epoch": 14.72,
      "learning_rate": 1.8037540453074434e-05,
      "loss": 0.1092,
      "step": 37900
    },
    {
      "epoch": 14.72,
      "learning_rate": 1.8037022653721686e-05,
      "loss": 0.2558,
      "step": 37910
    },
    {
      "epoch": 14.73,
      "learning_rate": 1.8036504854368934e-05,
      "loss": 0.0578,
      "step": 37920
    },
    {
      "epoch": 14.73,
      "learning_rate": 1.803598705501618e-05,
      "loss": 0.1744,
      "step": 37930
    },
    {
      "epoch": 14.73,
      "learning_rate": 1.8035469255663433e-05,
      "loss": 0.0166,
      "step": 37940
    },
    {
      "epoch": 14.74,
      "learning_rate": 1.803495145631068e-05,
      "loss": 0.2154,
      "step": 37950
    },
    {
      "epoch": 14.74,
      "learning_rate": 1.803443365695793e-05,
      "loss": 0.1728,
      "step": 37960
    },
    {
      "epoch": 14.75,
      "learning_rate": 1.803391585760518e-05,
      "loss": 0.1164,
      "step": 37970
    },
    {
      "epoch": 14.75,
      "learning_rate": 1.8033398058252428e-05,
      "loss": 0.1836,
      "step": 37980
    },
    {
      "epoch": 14.75,
      "learning_rate": 1.803288025889968e-05,
      "loss": 0.3583,
      "step": 37990
    },
    {
      "epoch": 14.76,
      "learning_rate": 1.8032362459546927e-05,
      "loss": 0.0838,
      "step": 38000
    },
    {
      "epoch": 14.76,
      "learning_rate": 1.8031844660194175e-05,
      "loss": 0.2662,
      "step": 38010
    },
    {
      "epoch": 14.77,
      "learning_rate": 1.8031326860841427e-05,
      "loss": 0.0783,
      "step": 38020
    },
    {
      "epoch": 14.77,
      "learning_rate": 1.8030809061488674e-05,
      "loss": 0.1042,
      "step": 38030
    },
    {
      "epoch": 14.77,
      "learning_rate": 1.8030291262135922e-05,
      "loss": 0.1035,
      "step": 38040
    },
    {
      "epoch": 14.78,
      "learning_rate": 1.8029773462783174e-05,
      "loss": 0.0877,
      "step": 38050
    },
    {
      "epoch": 14.78,
      "learning_rate": 1.802925566343042e-05,
      "loss": 0.3283,
      "step": 38060
    },
    {
      "epoch": 14.78,
      "learning_rate": 1.8028737864077673e-05,
      "loss": 0.1205,
      "step": 38070
    },
    {
      "epoch": 14.79,
      "learning_rate": 1.802822006472492e-05,
      "loss": 0.1009,
      "step": 38080
    },
    {
      "epoch": 14.79,
      "learning_rate": 1.802770226537217e-05,
      "loss": 0.0571,
      "step": 38090
    },
    {
      "epoch": 14.8,
      "learning_rate": 1.802718446601942e-05,
      "loss": 0.1193,
      "step": 38100
    },
    {
      "epoch": 14.8,
      "learning_rate": 1.8026666666666668e-05,
      "loss": 0.0988,
      "step": 38110
    },
    {
      "epoch": 14.8,
      "learning_rate": 1.8026148867313916e-05,
      "loss": 0.1389,
      "step": 38120
    },
    {
      "epoch": 14.81,
      "learning_rate": 1.8025631067961167e-05,
      "loss": 0.2079,
      "step": 38130
    },
    {
      "epoch": 14.81,
      "learning_rate": 1.8025113268608415e-05,
      "loss": 0.1587,
      "step": 38140
    },
    {
      "epoch": 14.82,
      "learning_rate": 1.8024595469255666e-05,
      "loss": 0.1053,
      "step": 38150
    },
    {
      "epoch": 14.82,
      "learning_rate": 1.8024077669902914e-05,
      "loss": 0.1324,
      "step": 38160
    },
    {
      "epoch": 14.82,
      "learning_rate": 1.8023559870550162e-05,
      "loss": 0.079,
      "step": 38170
    },
    {
      "epoch": 14.83,
      "learning_rate": 1.802304207119741e-05,
      "loss": 0.1886,
      "step": 38180
    },
    {
      "epoch": 14.83,
      "learning_rate": 1.802252427184466e-05,
      "loss": 0.1384,
      "step": 38190
    },
    {
      "epoch": 14.83,
      "learning_rate": 1.802200647249191e-05,
      "loss": 0.165,
      "step": 38200
    },
    {
      "epoch": 14.84,
      "learning_rate": 1.802148867313916e-05,
      "loss": 0.0638,
      "step": 38210
    },
    {
      "epoch": 14.84,
      "learning_rate": 1.802097087378641e-05,
      "loss": 0.0895,
      "step": 38220
    },
    {
      "epoch": 14.85,
      "learning_rate": 1.802045307443366e-05,
      "loss": 0.1876,
      "step": 38230
    },
    {
      "epoch": 14.85,
      "learning_rate": 1.8019935275080908e-05,
      "loss": 0.0944,
      "step": 38240
    },
    {
      "epoch": 14.85,
      "learning_rate": 1.8019417475728156e-05,
      "loss": 0.0281,
      "step": 38250
    },
    {
      "epoch": 14.86,
      "learning_rate": 1.8018899676375404e-05,
      "loss": 0.2161,
      "step": 38260
    },
    {
      "epoch": 14.86,
      "learning_rate": 1.8018381877022655e-05,
      "loss": 0.1949,
      "step": 38270
    },
    {
      "epoch": 14.87,
      "learning_rate": 1.8017864077669903e-05,
      "loss": 0.0608,
      "step": 38280
    },
    {
      "epoch": 14.87,
      "learning_rate": 1.8017346278317154e-05,
      "loss": 0.2095,
      "step": 38290
    },
    {
      "epoch": 14.87,
      "learning_rate": 1.8016828478964402e-05,
      "loss": 0.1398,
      "step": 38300
    },
    {
      "epoch": 14.88,
      "learning_rate": 1.8016310679611654e-05,
      "loss": 0.2586,
      "step": 38310
    },
    {
      "epoch": 14.88,
      "learning_rate": 1.80157928802589e-05,
      "loss": 0.0606,
      "step": 38320
    },
    {
      "epoch": 14.89,
      "learning_rate": 1.801527508090615e-05,
      "loss": 0.099,
      "step": 38330
    },
    {
      "epoch": 14.89,
      "learning_rate": 1.8014757281553397e-05,
      "loss": 0.2385,
      "step": 38340
    },
    {
      "epoch": 14.89,
      "learning_rate": 1.801423948220065e-05,
      "loss": 0.0977,
      "step": 38350
    },
    {
      "epoch": 14.9,
      "learning_rate": 1.8013721682847897e-05,
      "loss": 0.0959,
      "step": 38360
    },
    {
      "epoch": 14.9,
      "learning_rate": 1.8013203883495148e-05,
      "loss": 0.2248,
      "step": 38370
    },
    {
      "epoch": 14.9,
      "learning_rate": 1.8012686084142396e-05,
      "loss": 0.0647,
      "step": 38380
    },
    {
      "epoch": 14.91,
      "learning_rate": 1.8012168284789647e-05,
      "loss": 0.1855,
      "step": 38390
    },
    {
      "epoch": 14.91,
      "learning_rate": 1.8011650485436895e-05,
      "loss": 0.3621,
      "step": 38400
    },
    {
      "epoch": 14.92,
      "learning_rate": 1.8011132686084143e-05,
      "loss": 0.1729,
      "step": 38410
    },
    {
      "epoch": 14.92,
      "learning_rate": 1.801061488673139e-05,
      "loss": 0.0318,
      "step": 38420
    },
    {
      "epoch": 14.92,
      "learning_rate": 1.8010097087378642e-05,
      "loss": 0.1176,
      "step": 38430
    },
    {
      "epoch": 14.93,
      "learning_rate": 1.800957928802589e-05,
      "loss": 0.2033,
      "step": 38440
    },
    {
      "epoch": 14.93,
      "learning_rate": 1.800906148867314e-05,
      "loss": 0.151,
      "step": 38450
    },
    {
      "epoch": 14.94,
      "learning_rate": 1.800854368932039e-05,
      "loss": 0.1013,
      "step": 38460
    },
    {
      "epoch": 14.94,
      "learning_rate": 1.800802588996764e-05,
      "loss": 0.0302,
      "step": 38470
    },
    {
      "epoch": 14.94,
      "learning_rate": 1.800750809061489e-05,
      "loss": 0.0506,
      "step": 38480
    },
    {
      "epoch": 14.95,
      "learning_rate": 1.8006990291262137e-05,
      "loss": 0.2014,
      "step": 38490
    },
    {
      "epoch": 14.95,
      "learning_rate": 1.8006472491909384e-05,
      "loss": 0.154,
      "step": 38500
    },
    {
      "epoch": 14.96,
      "learning_rate": 1.8005954692556636e-05,
      "loss": 0.0977,
      "step": 38510
    },
    {
      "epoch": 14.96,
      "learning_rate": 1.8005436893203884e-05,
      "loss": 0.2643,
      "step": 38520
    },
    {
      "epoch": 14.96,
      "learning_rate": 1.8004919093851135e-05,
      "loss": 0.093,
      "step": 38530
    },
    {
      "epoch": 14.97,
      "learning_rate": 1.8004401294498383e-05,
      "loss": 0.1036,
      "step": 38540
    },
    {
      "epoch": 14.97,
      "learning_rate": 1.8003883495145634e-05,
      "loss": 0.2291,
      "step": 38550
    },
    {
      "epoch": 14.97,
      "learning_rate": 1.8003365695792882e-05,
      "loss": 0.1159,
      "step": 38560
    },
    {
      "epoch": 14.98,
      "learning_rate": 1.800284789644013e-05,
      "loss": 0.2121,
      "step": 38570
    },
    {
      "epoch": 14.98,
      "learning_rate": 1.8002330097087378e-05,
      "loss": 0.0277,
      "step": 38580
    },
    {
      "epoch": 14.99,
      "learning_rate": 1.800181229773463e-05,
      "loss": 0.3163,
      "step": 38590
    },
    {
      "epoch": 14.99,
      "learning_rate": 1.8001294498381877e-05,
      "loss": 0.0929,
      "step": 38600
    },
    {
      "epoch": 14.99,
      "learning_rate": 1.800077669902913e-05,
      "loss": 0.0888,
      "step": 38610
    },
    {
      "epoch": 15.0,
      "learning_rate": 1.8000258899676376e-05,
      "loss": 0.1447,
      "step": 38620
    },
    {
      "epoch": 15.0,
      "eval_accuracy": 0.9504814305364512,
      "eval_loss": 0.23635753989219666,
      "eval_runtime": 8.2297,
      "eval_samples_per_second": 441.693,
      "eval_steps_per_second": 55.288,
      "step": 38625
    },
    {
      "epoch": 15.0,
      "learning_rate": 1.7999741100323628e-05,
      "loss": 0.0864,
      "step": 38630
    },
    {
      "epoch": 15.01,
      "learning_rate": 1.7999223300970876e-05,
      "loss": 0.0653,
      "step": 38640
    },
    {
      "epoch": 15.01,
      "learning_rate": 1.7998705501618124e-05,
      "loss": 0.05,
      "step": 38650
    },
    {
      "epoch": 15.01,
      "learning_rate": 1.799818770226537e-05,
      "loss": 0.1601,
      "step": 38660
    },
    {
      "epoch": 15.02,
      "learning_rate": 1.7997669902912623e-05,
      "loss": 0.0815,
      "step": 38670
    },
    {
      "epoch": 15.02,
      "learning_rate": 1.799715210355987e-05,
      "loss": 0.0874,
      "step": 38680
    },
    {
      "epoch": 15.03,
      "learning_rate": 1.7996634304207122e-05,
      "loss": 0.0314,
      "step": 38690
    },
    {
      "epoch": 15.03,
      "learning_rate": 1.799611650485437e-05,
      "loss": 0.2666,
      "step": 38700
    },
    {
      "epoch": 15.03,
      "learning_rate": 1.799559870550162e-05,
      "loss": 0.1374,
      "step": 38710
    },
    {
      "epoch": 15.04,
      "learning_rate": 1.799508090614887e-05,
      "loss": 0.2264,
      "step": 38720
    },
    {
      "epoch": 15.04,
      "learning_rate": 1.799456310679612e-05,
      "loss": 0.1946,
      "step": 38730
    },
    {
      "epoch": 15.04,
      "learning_rate": 1.7994045307443365e-05,
      "loss": 0.1186,
      "step": 38740
    },
    {
      "epoch": 15.05,
      "learning_rate": 1.7993527508090616e-05,
      "loss": 0.0412,
      "step": 38750
    },
    {
      "epoch": 15.05,
      "learning_rate": 1.7993009708737864e-05,
      "loss": 0.1545,
      "step": 38760
    },
    {
      "epoch": 15.06,
      "learning_rate": 1.7992491909385116e-05,
      "loss": 0.0305,
      "step": 38770
    },
    {
      "epoch": 15.06,
      "learning_rate": 1.7991974110032364e-05,
      "loss": 0.1615,
      "step": 38780
    },
    {
      "epoch": 15.06,
      "learning_rate": 1.7991456310679615e-05,
      "loss": 0.126,
      "step": 38790
    },
    {
      "epoch": 15.07,
      "learning_rate": 1.7990938511326863e-05,
      "loss": 0.1885,
      "step": 38800
    },
    {
      "epoch": 15.07,
      "learning_rate": 1.7990420711974114e-05,
      "loss": 0.1673,
      "step": 38810
    },
    {
      "epoch": 15.08,
      "learning_rate": 1.798990291262136e-05,
      "loss": 0.1118,
      "step": 38820
    },
    {
      "epoch": 15.08,
      "learning_rate": 1.798938511326861e-05,
      "loss": 0.0604,
      "step": 38830
    },
    {
      "epoch": 15.08,
      "learning_rate": 1.7988867313915858e-05,
      "loss": 0.0213,
      "step": 38840
    },
    {
      "epoch": 15.09,
      "learning_rate": 1.798834951456311e-05,
      "loss": 0.0386,
      "step": 38850
    },
    {
      "epoch": 15.09,
      "learning_rate": 1.7987831715210357e-05,
      "loss": 0.0115,
      "step": 38860
    },
    {
      "epoch": 15.1,
      "learning_rate": 1.798731391585761e-05,
      "loss": 0.0807,
      "step": 38870
    },
    {
      "epoch": 15.1,
      "learning_rate": 1.7986796116504856e-05,
      "loss": 0.0107,
      "step": 38880
    },
    {
      "epoch": 15.1,
      "learning_rate": 1.7986278317152108e-05,
      "loss": 0.241,
      "step": 38890
    },
    {
      "epoch": 15.11,
      "learning_rate": 1.7985760517799352e-05,
      "loss": 0.0841,
      "step": 38900
    },
    {
      "epoch": 15.11,
      "learning_rate": 1.7985242718446604e-05,
      "loss": 0.1765,
      "step": 38910
    },
    {
      "epoch": 15.11,
      "learning_rate": 1.798472491909385e-05,
      "loss": 0.2134,
      "step": 38920
    },
    {
      "epoch": 15.12,
      "learning_rate": 1.7984207119741103e-05,
      "loss": 0.0625,
      "step": 38930
    },
    {
      "epoch": 15.12,
      "learning_rate": 1.798368932038835e-05,
      "loss": 0.2839,
      "step": 38940
    },
    {
      "epoch": 15.13,
      "learning_rate": 1.7983171521035602e-05,
      "loss": 0.0397,
      "step": 38950
    },
    {
      "epoch": 15.13,
      "learning_rate": 1.798265372168285e-05,
      "loss": 0.3216,
      "step": 38960
    },
    {
      "epoch": 15.13,
      "learning_rate": 1.7982135922330098e-05,
      "loss": 0.1513,
      "step": 38970
    },
    {
      "epoch": 15.14,
      "learning_rate": 1.7981618122977346e-05,
      "loss": 0.147,
      "step": 38980
    },
    {
      "epoch": 15.14,
      "learning_rate": 1.7981100323624597e-05,
      "loss": 0.0437,
      "step": 38990
    },
    {
      "epoch": 15.15,
      "learning_rate": 1.7980582524271845e-05,
      "loss": 0.145,
      "step": 39000
    },
    {
      "epoch": 15.15,
      "learning_rate": 1.7980064724919096e-05,
      "loss": 0.1021,
      "step": 39010
    },
    {
      "epoch": 15.15,
      "learning_rate": 1.7979546925566344e-05,
      "loss": 0.0253,
      "step": 39020
    },
    {
      "epoch": 15.16,
      "learning_rate": 1.7979029126213596e-05,
      "loss": 0.1301,
      "step": 39030
    },
    {
      "epoch": 15.16,
      "learning_rate": 1.7978511326860844e-05,
      "loss": 0.115,
      "step": 39040
    },
    {
      "epoch": 15.17,
      "learning_rate": 1.797799352750809e-05,
      "loss": 0.1015,
      "step": 39050
    },
    {
      "epoch": 15.17,
      "learning_rate": 1.797747572815534e-05,
      "loss": 0.1105,
      "step": 39060
    },
    {
      "epoch": 15.17,
      "learning_rate": 1.797695792880259e-05,
      "loss": 0.1061,
      "step": 39070
    },
    {
      "epoch": 15.18,
      "learning_rate": 1.797644012944984e-05,
      "loss": 0.2207,
      "step": 39080
    },
    {
      "epoch": 15.18,
      "learning_rate": 1.797592233009709e-05,
      "loss": 0.1117,
      "step": 39090
    },
    {
      "epoch": 15.18,
      "learning_rate": 1.7975404530744338e-05,
      "loss": 0.0341,
      "step": 39100
    },
    {
      "epoch": 15.19,
      "learning_rate": 1.797488673139159e-05,
      "loss": 0.2546,
      "step": 39110
    },
    {
      "epoch": 15.19,
      "learning_rate": 1.7974368932038837e-05,
      "loss": 0.1726,
      "step": 39120
    },
    {
      "epoch": 15.2,
      "learning_rate": 1.7973851132686085e-05,
      "loss": 0.0683,
      "step": 39130
    },
    {
      "epoch": 15.2,
      "learning_rate": 1.7973333333333333e-05,
      "loss": 0.1419,
      "step": 39140
    },
    {
      "epoch": 15.2,
      "learning_rate": 1.7972815533980584e-05,
      "loss": 0.2322,
      "step": 39150
    },
    {
      "epoch": 15.21,
      "learning_rate": 1.7972297734627832e-05,
      "loss": 0.1566,
      "step": 39160
    },
    {
      "epoch": 15.21,
      "learning_rate": 1.7971779935275083e-05,
      "loss": 0.1279,
      "step": 39170
    },
    {
      "epoch": 15.22,
      "learning_rate": 1.797126213592233e-05,
      "loss": 0.0556,
      "step": 39180
    },
    {
      "epoch": 15.22,
      "learning_rate": 1.7970744336569583e-05,
      "loss": 0.1349,
      "step": 39190
    },
    {
      "epoch": 15.22,
      "learning_rate": 1.797022653721683e-05,
      "loss": 0.0751,
      "step": 39200
    },
    {
      "epoch": 15.23,
      "learning_rate": 1.796970873786408e-05,
      "loss": 0.2012,
      "step": 39210
    },
    {
      "epoch": 15.23,
      "learning_rate": 1.7969190938511326e-05,
      "loss": 0.025,
      "step": 39220
    },
    {
      "epoch": 15.23,
      "learning_rate": 1.7968673139158578e-05,
      "loss": 0.1013,
      "step": 39230
    },
    {
      "epoch": 15.24,
      "learning_rate": 1.7968155339805826e-05,
      "loss": 0.1823,
      "step": 39240
    },
    {
      "epoch": 15.24,
      "learning_rate": 1.7967637540453077e-05,
      "loss": 0.1238,
      "step": 39250
    },
    {
      "epoch": 15.25,
      "learning_rate": 1.7967119741100325e-05,
      "loss": 0.0867,
      "step": 39260
    },
    {
      "epoch": 15.25,
      "learning_rate": 1.7966601941747573e-05,
      "loss": 0.1461,
      "step": 39270
    },
    {
      "epoch": 15.25,
      "learning_rate": 1.7966084142394824e-05,
      "loss": 0.1967,
      "step": 39280
    },
    {
      "epoch": 15.26,
      "learning_rate": 1.7965566343042072e-05,
      "loss": 0.0784,
      "step": 39290
    },
    {
      "epoch": 15.26,
      "learning_rate": 1.7965048543689323e-05,
      "loss": 0.0672,
      "step": 39300
    },
    {
      "epoch": 15.27,
      "learning_rate": 1.796453074433657e-05,
      "loss": 0.1685,
      "step": 39310
    },
    {
      "epoch": 15.27,
      "learning_rate": 1.796401294498382e-05,
      "loss": 0.0564,
      "step": 39320
    },
    {
      "epoch": 15.27,
      "learning_rate": 1.796349514563107e-05,
      "loss": 0.0562,
      "step": 39330
    },
    {
      "epoch": 15.28,
      "learning_rate": 1.796297734627832e-05,
      "loss": 0.0073,
      "step": 39340
    },
    {
      "epoch": 15.28,
      "learning_rate": 1.7962459546925566e-05,
      "loss": 0.1258,
      "step": 39350
    },
    {
      "epoch": 15.29,
      "learning_rate": 1.7961941747572818e-05,
      "loss": 0.1641,
      "step": 39360
    },
    {
      "epoch": 15.29,
      "learning_rate": 1.7961423948220066e-05,
      "loss": 0.1339,
      "step": 39370
    },
    {
      "epoch": 15.29,
      "learning_rate": 1.7960906148867317e-05,
      "loss": 0.0977,
      "step": 39380
    },
    {
      "epoch": 15.3,
      "learning_rate": 1.7960388349514565e-05,
      "loss": 0.2237,
      "step": 39390
    },
    {
      "epoch": 15.3,
      "learning_rate": 1.7959870550161813e-05,
      "loss": 0.2083,
      "step": 39400
    },
    {
      "epoch": 15.3,
      "learning_rate": 1.7959352750809064e-05,
      "loss": 0.0212,
      "step": 39410
    },
    {
      "epoch": 15.31,
      "learning_rate": 1.7958834951456312e-05,
      "loss": 0.0127,
      "step": 39420
    },
    {
      "epoch": 15.31,
      "learning_rate": 1.795831715210356e-05,
      "loss": 0.1612,
      "step": 39430
    },
    {
      "epoch": 15.32,
      "learning_rate": 1.795779935275081e-05,
      "loss": 0.2009,
      "step": 39440
    },
    {
      "epoch": 15.32,
      "learning_rate": 1.795728155339806e-05,
      "loss": 0.0938,
      "step": 39450
    },
    {
      "epoch": 15.32,
      "learning_rate": 1.795676375404531e-05,
      "loss": 0.1605,
      "step": 39460
    },
    {
      "epoch": 15.33,
      "learning_rate": 1.795624595469256e-05,
      "loss": 0.0836,
      "step": 39470
    },
    {
      "epoch": 15.33,
      "learning_rate": 1.7955728155339806e-05,
      "loss": 0.2032,
      "step": 39480
    },
    {
      "epoch": 15.34,
      "learning_rate": 1.7955210355987058e-05,
      "loss": 0.1075,
      "step": 39490
    },
    {
      "epoch": 15.34,
      "learning_rate": 1.7954692556634306e-05,
      "loss": 0.0299,
      "step": 39500
    },
    {
      "epoch": 15.34,
      "learning_rate": 1.7954174757281554e-05,
      "loss": 0.1927,
      "step": 39510
    },
    {
      "epoch": 15.35,
      "learning_rate": 1.7953656957928805e-05,
      "loss": 0.1215,
      "step": 39520
    },
    {
      "epoch": 15.35,
      "learning_rate": 1.7953139158576053e-05,
      "loss": 0.2151,
      "step": 39530
    },
    {
      "epoch": 15.36,
      "learning_rate": 1.7952621359223304e-05,
      "loss": 0.1844,
      "step": 39540
    },
    {
      "epoch": 15.36,
      "learning_rate": 1.7952103559870552e-05,
      "loss": 0.2638,
      "step": 39550
    },
    {
      "epoch": 15.36,
      "learning_rate": 1.79515857605178e-05,
      "loss": 0.0601,
      "step": 39560
    },
    {
      "epoch": 15.37,
      "learning_rate": 1.795106796116505e-05,
      "loss": 0.072,
      "step": 39570
    },
    {
      "epoch": 15.37,
      "learning_rate": 1.79505501618123e-05,
      "loss": 0.196,
      "step": 39580
    },
    {
      "epoch": 15.37,
      "learning_rate": 1.7950032362459547e-05,
      "loss": 0.1307,
      "step": 39590
    },
    {
      "epoch": 15.38,
      "learning_rate": 1.79495145631068e-05,
      "loss": 0.0772,
      "step": 39600
    },
    {
      "epoch": 15.38,
      "learning_rate": 1.7948996763754046e-05,
      "loss": 0.2126,
      "step": 39610
    },
    {
      "epoch": 15.39,
      "learning_rate": 1.7948478964401298e-05,
      "loss": 0.1666,
      "step": 39620
    },
    {
      "epoch": 15.39,
      "learning_rate": 1.7947961165048546e-05,
      "loss": 0.0442,
      "step": 39630
    },
    {
      "epoch": 15.39,
      "learning_rate": 1.7947443365695793e-05,
      "loss": 0.1355,
      "step": 39640
    },
    {
      "epoch": 15.4,
      "learning_rate": 1.794692556634304e-05,
      "loss": 0.1951,
      "step": 39650
    },
    {
      "epoch": 15.4,
      "learning_rate": 1.7946407766990293e-05,
      "loss": 0.1715,
      "step": 39660
    },
    {
      "epoch": 15.41,
      "learning_rate": 1.794588996763754e-05,
      "loss": 0.1028,
      "step": 39670
    },
    {
      "epoch": 15.41,
      "learning_rate": 1.7945372168284792e-05,
      "loss": 0.0578,
      "step": 39680
    },
    {
      "epoch": 15.41,
      "learning_rate": 1.794485436893204e-05,
      "loss": 0.2741,
      "step": 39690
    },
    {
      "epoch": 15.42,
      "learning_rate": 1.794433656957929e-05,
      "loss": 0.2347,
      "step": 39700
    },
    {
      "epoch": 15.42,
      "learning_rate": 1.794381877022654e-05,
      "loss": 0.078,
      "step": 39710
    },
    {
      "epoch": 15.43,
      "learning_rate": 1.7943300970873787e-05,
      "loss": 0.0895,
      "step": 39720
    },
    {
      "epoch": 15.43,
      "learning_rate": 1.7942783171521035e-05,
      "loss": 0.0289,
      "step": 39730
    },
    {
      "epoch": 15.43,
      "learning_rate": 1.7942265372168286e-05,
      "loss": 0.2151,
      "step": 39740
    },
    {
      "epoch": 15.44,
      "learning_rate": 1.7941747572815534e-05,
      "loss": 0.1476,
      "step": 39750
    },
    {
      "epoch": 15.44,
      "learning_rate": 1.7941229773462786e-05,
      "loss": 0.0545,
      "step": 39760
    },
    {
      "epoch": 15.44,
      "learning_rate": 1.7940711974110033e-05,
      "loss": 0.0808,
      "step": 39770
    },
    {
      "epoch": 15.45,
      "learning_rate": 1.7940194174757285e-05,
      "loss": 0.2449,
      "step": 39780
    },
    {
      "epoch": 15.45,
      "learning_rate": 1.7939676375404533e-05,
      "loss": 0.2768,
      "step": 39790
    },
    {
      "epoch": 15.46,
      "learning_rate": 1.793915857605178e-05,
      "loss": 0.346,
      "step": 39800
    },
    {
      "epoch": 15.46,
      "learning_rate": 1.793864077669903e-05,
      "loss": 0.0297,
      "step": 39810
    },
    {
      "epoch": 15.46,
      "learning_rate": 1.793812297734628e-05,
      "loss": 0.0363,
      "step": 39820
    },
    {
      "epoch": 15.47,
      "learning_rate": 1.7937605177993528e-05,
      "loss": 0.234,
      "step": 39830
    },
    {
      "epoch": 15.47,
      "learning_rate": 1.793708737864078e-05,
      "loss": 0.1005,
      "step": 39840
    },
    {
      "epoch": 15.48,
      "learning_rate": 1.7936569579288027e-05,
      "loss": 0.066,
      "step": 39850
    },
    {
      "epoch": 15.48,
      "learning_rate": 1.793605177993528e-05,
      "loss": 0.1582,
      "step": 39860
    },
    {
      "epoch": 15.48,
      "learning_rate": 1.7935533980582526e-05,
      "loss": 0.1144,
      "step": 39870
    },
    {
      "epoch": 15.49,
      "learning_rate": 1.7935016181229774e-05,
      "loss": 0.1548,
      "step": 39880
    },
    {
      "epoch": 15.49,
      "learning_rate": 1.7934498381877022e-05,
      "loss": 0.0878,
      "step": 39890
    },
    {
      "epoch": 15.5,
      "learning_rate": 1.7933980582524273e-05,
      "loss": 0.2833,
      "step": 39900
    },
    {
      "epoch": 15.5,
      "learning_rate": 1.793346278317152e-05,
      "loss": 0.0904,
      "step": 39910
    },
    {
      "epoch": 15.5,
      "learning_rate": 1.7932944983818773e-05,
      "loss": 0.003,
      "step": 39920
    },
    {
      "epoch": 15.51,
      "learning_rate": 1.793242718446602e-05,
      "loss": 0.1377,
      "step": 39930
    },
    {
      "epoch": 15.51,
      "learning_rate": 1.7931909385113272e-05,
      "loss": 0.1616,
      "step": 39940
    },
    {
      "epoch": 15.51,
      "learning_rate": 1.793139158576052e-05,
      "loss": 0.0985,
      "step": 39950
    },
    {
      "epoch": 15.52,
      "learning_rate": 1.7930873786407768e-05,
      "loss": 0.2529,
      "step": 39960
    },
    {
      "epoch": 15.52,
      "learning_rate": 1.7930355987055016e-05,
      "loss": 0.112,
      "step": 39970
    },
    {
      "epoch": 15.53,
      "learning_rate": 1.7929838187702267e-05,
      "loss": 0.086,
      "step": 39980
    },
    {
      "epoch": 15.53,
      "learning_rate": 1.7929320388349515e-05,
      "loss": 0.0861,
      "step": 39990
    },
    {
      "epoch": 15.53,
      "learning_rate": 1.7928802588996766e-05,
      "loss": 0.0478,
      "step": 40000
    },
    {
      "epoch": 15.54,
      "learning_rate": 1.7928284789644014e-05,
      "loss": 0.1289,
      "step": 40010
    },
    {
      "epoch": 15.54,
      "learning_rate": 1.7927766990291265e-05,
      "loss": 0.0726,
      "step": 40020
    },
    {
      "epoch": 15.55,
      "learning_rate": 1.7927249190938513e-05,
      "loss": 0.1647,
      "step": 40030
    },
    {
      "epoch": 15.55,
      "learning_rate": 1.792673139158576e-05,
      "loss": 0.3235,
      "step": 40040
    },
    {
      "epoch": 15.55,
      "learning_rate": 1.792621359223301e-05,
      "loss": 0.0664,
      "step": 40050
    },
    {
      "epoch": 15.56,
      "learning_rate": 1.792569579288026e-05,
      "loss": 0.0341,
      "step": 40060
    },
    {
      "epoch": 15.56,
      "learning_rate": 1.792517799352751e-05,
      "loss": 0.2211,
      "step": 40070
    },
    {
      "epoch": 15.57,
      "learning_rate": 1.792466019417476e-05,
      "loss": 0.0899,
      "step": 40080
    },
    {
      "epoch": 15.57,
      "learning_rate": 1.7924142394822008e-05,
      "loss": 0.0423,
      "step": 40090
    },
    {
      "epoch": 15.57,
      "learning_rate": 1.792362459546926e-05,
      "loss": 0.0569,
      "step": 40100
    },
    {
      "epoch": 15.58,
      "learning_rate": 1.7923106796116507e-05,
      "loss": 0.1058,
      "step": 40110
    },
    {
      "epoch": 15.58,
      "learning_rate": 1.7922588996763755e-05,
      "loss": 0.1233,
      "step": 40120
    },
    {
      "epoch": 15.58,
      "learning_rate": 1.7922071197411003e-05,
      "loss": 0.1034,
      "step": 40130
    },
    {
      "epoch": 15.59,
      "learning_rate": 1.7921553398058254e-05,
      "loss": 0.0297,
      "step": 40140
    },
    {
      "epoch": 15.59,
      "learning_rate": 1.7921035598705502e-05,
      "loss": 0.2903,
      "step": 40150
    },
    {
      "epoch": 15.6,
      "learning_rate": 1.7920517799352753e-05,
      "loss": 0.0203,
      "step": 40160
    },
    {
      "epoch": 15.6,
      "learning_rate": 1.792e-05,
      "loss": 0.0329,
      "step": 40170
    },
    {
      "epoch": 15.6,
      "learning_rate": 1.7919482200647253e-05,
      "loss": 0.1956,
      "step": 40180
    },
    {
      "epoch": 15.61,
      "learning_rate": 1.79189644012945e-05,
      "loss": 0.2007,
      "step": 40190
    },
    {
      "epoch": 15.61,
      "learning_rate": 1.791844660194175e-05,
      "loss": 0.0932,
      "step": 40200
    },
    {
      "epoch": 15.62,
      "learning_rate": 1.7917928802588996e-05,
      "loss": 0.1551,
      "step": 40210
    },
    {
      "epoch": 15.62,
      "learning_rate": 1.7917411003236248e-05,
      "loss": 0.2259,
      "step": 40220
    },
    {
      "epoch": 15.62,
      "learning_rate": 1.7916893203883496e-05,
      "loss": 0.1381,
      "step": 40230
    },
    {
      "epoch": 15.63,
      "learning_rate": 1.7916375404530747e-05,
      "loss": 0.1896,
      "step": 40240
    },
    {
      "epoch": 15.63,
      "learning_rate": 1.7915857605177995e-05,
      "loss": 0.1515,
      "step": 40250
    },
    {
      "epoch": 15.63,
      "learning_rate": 1.7915339805825246e-05,
      "loss": 0.3167,
      "step": 40260
    },
    {
      "epoch": 15.64,
      "learning_rate": 1.7914822006472494e-05,
      "loss": 0.0681,
      "step": 40270
    },
    {
      "epoch": 15.64,
      "learning_rate": 1.7914304207119742e-05,
      "loss": 0.1796,
      "step": 40280
    },
    {
      "epoch": 15.65,
      "learning_rate": 1.791378640776699e-05,
      "loss": 0.1188,
      "step": 40290
    },
    {
      "epoch": 15.65,
      "learning_rate": 1.791326860841424e-05,
      "loss": 0.1347,
      "step": 40300
    },
    {
      "epoch": 15.65,
      "learning_rate": 1.791275080906149e-05,
      "loss": 0.1585,
      "step": 40310
    },
    {
      "epoch": 15.66,
      "learning_rate": 1.791223300970874e-05,
      "loss": 0.2602,
      "step": 40320
    },
    {
      "epoch": 15.66,
      "learning_rate": 1.791171521035599e-05,
      "loss": 0.14,
      "step": 40330
    },
    {
      "epoch": 15.67,
      "learning_rate": 1.791119741100324e-05,
      "loss": 0.0516,
      "step": 40340
    },
    {
      "epoch": 15.67,
      "learning_rate": 1.7910679611650488e-05,
      "loss": 0.1252,
      "step": 40350
    },
    {
      "epoch": 15.67,
      "learning_rate": 1.7910161812297735e-05,
      "loss": 0.0097,
      "step": 40360
    },
    {
      "epoch": 15.68,
      "learning_rate": 1.7909644012944983e-05,
      "loss": 0.2795,
      "step": 40370
    },
    {
      "epoch": 15.68,
      "learning_rate": 1.7909126213592235e-05,
      "loss": 0.0629,
      "step": 40380
    },
    {
      "epoch": 15.69,
      "learning_rate": 1.7908608414239483e-05,
      "loss": 0.1671,
      "step": 40390
    },
    {
      "epoch": 15.69,
      "learning_rate": 1.7908090614886734e-05,
      "loss": 0.1292,
      "step": 40400
    },
    {
      "epoch": 15.69,
      "learning_rate": 1.7907572815533982e-05,
      "loss": 0.0327,
      "step": 40410
    },
    {
      "epoch": 15.7,
      "learning_rate": 1.7907055016181233e-05,
      "loss": 0.422,
      "step": 40420
    },
    {
      "epoch": 15.7,
      "learning_rate": 1.790653721682848e-05,
      "loss": 0.123,
      "step": 40430
    },
    {
      "epoch": 15.7,
      "learning_rate": 1.790601941747573e-05,
      "loss": 0.1824,
      "step": 40440
    },
    {
      "epoch": 15.71,
      "learning_rate": 1.7905501618122977e-05,
      "loss": 0.411,
      "step": 40450
    },
    {
      "epoch": 15.71,
      "learning_rate": 1.7904983818770228e-05,
      "loss": 0.0302,
      "step": 40460
    },
    {
      "epoch": 15.72,
      "learning_rate": 1.7904466019417476e-05,
      "loss": 0.0918,
      "step": 40470
    },
    {
      "epoch": 15.72,
      "learning_rate": 1.7903948220064728e-05,
      "loss": 0.1542,
      "step": 40480
    },
    {
      "epoch": 15.72,
      "learning_rate": 1.7903430420711975e-05,
      "loss": 0.1298,
      "step": 40490
    },
    {
      "epoch": 15.73,
      "learning_rate": 1.7902912621359227e-05,
      "loss": 0.1825,
      "step": 40500
    },
    {
      "epoch": 15.73,
      "learning_rate": 1.7902394822006475e-05,
      "loss": 0.2541,
      "step": 40510
    },
    {
      "epoch": 15.74,
      "learning_rate": 1.7901877022653723e-05,
      "loss": 0.19,
      "step": 40520
    },
    {
      "epoch": 15.74,
      "learning_rate": 1.790135922330097e-05,
      "loss": 0.0482,
      "step": 40530
    },
    {
      "epoch": 15.74,
      "learning_rate": 1.7900841423948222e-05,
      "loss": 0.108,
      "step": 40540
    },
    {
      "epoch": 15.75,
      "learning_rate": 1.790032362459547e-05,
      "loss": 0.0353,
      "step": 40550
    },
    {
      "epoch": 15.75,
      "learning_rate": 1.789980582524272e-05,
      "loss": 0.1436,
      "step": 40560
    },
    {
      "epoch": 15.76,
      "learning_rate": 1.789928802588997e-05,
      "loss": 0.1292,
      "step": 40570
    },
    {
      "epoch": 15.76,
      "learning_rate": 1.789877022653722e-05,
      "loss": 0.1554,
      "step": 40580
    },
    {
      "epoch": 15.76,
      "learning_rate": 1.7898252427184468e-05,
      "loss": 0.2875,
      "step": 40590
    },
    {
      "epoch": 15.77,
      "learning_rate": 1.7897734627831716e-05,
      "loss": 0.1604,
      "step": 40600
    },
    {
      "epoch": 15.77,
      "learning_rate": 1.7897216828478964e-05,
      "loss": 0.1895,
      "step": 40610
    },
    {
      "epoch": 15.77,
      "learning_rate": 1.7896699029126215e-05,
      "loss": 0.1919,
      "step": 40620
    },
    {
      "epoch": 15.78,
      "learning_rate": 1.7896181229773463e-05,
      "loss": 0.0992,
      "step": 40630
    },
    {
      "epoch": 15.78,
      "learning_rate": 1.7895663430420715e-05,
      "loss": 0.0787,
      "step": 40640
    },
    {
      "epoch": 15.79,
      "learning_rate": 1.7895145631067963e-05,
      "loss": 0.1446,
      "step": 40650
    },
    {
      "epoch": 15.79,
      "learning_rate": 1.7894627831715214e-05,
      "loss": 0.1886,
      "step": 40660
    },
    {
      "epoch": 15.79,
      "learning_rate": 1.7894110032362462e-05,
      "loss": 0.1147,
      "step": 40670
    },
    {
      "epoch": 15.8,
      "learning_rate": 1.789359223300971e-05,
      "loss": 0.2694,
      "step": 40680
    },
    {
      "epoch": 15.8,
      "learning_rate": 1.7893074433656958e-05,
      "loss": 0.1065,
      "step": 40690
    },
    {
      "epoch": 15.81,
      "learning_rate": 1.789255663430421e-05,
      "loss": 0.1849,
      "step": 40700
    },
    {
      "epoch": 15.81,
      "learning_rate": 1.7892038834951457e-05,
      "loss": 0.0448,
      "step": 40710
    },
    {
      "epoch": 15.81,
      "learning_rate": 1.7891521035598708e-05,
      "loss": 0.0978,
      "step": 40720
    },
    {
      "epoch": 15.82,
      "learning_rate": 1.7891003236245956e-05,
      "loss": 0.0824,
      "step": 40730
    },
    {
      "epoch": 15.82,
      "learning_rate": 1.7890485436893204e-05,
      "loss": 0.1893,
      "step": 40740
    },
    {
      "epoch": 15.83,
      "learning_rate": 1.7889967637540455e-05,
      "loss": 0.1088,
      "step": 40750
    },
    {
      "epoch": 15.83,
      "learning_rate": 1.7889449838187703e-05,
      "loss": 0.0185,
      "step": 40760
    },
    {
      "epoch": 15.83,
      "learning_rate": 1.788893203883495e-05,
      "loss": 0.1535,
      "step": 40770
    },
    {
      "epoch": 15.84,
      "learning_rate": 1.7888414239482203e-05,
      "loss": 0.2427,
      "step": 40780
    },
    {
      "epoch": 15.84,
      "learning_rate": 1.788789644012945e-05,
      "loss": 0.1099,
      "step": 40790
    },
    {
      "epoch": 15.84,
      "learning_rate": 1.7887378640776702e-05,
      "loss": 0.2341,
      "step": 40800
    },
    {
      "epoch": 15.85,
      "learning_rate": 1.788686084142395e-05,
      "loss": 0.1155,
      "step": 40810
    },
    {
      "epoch": 15.85,
      "learning_rate": 1.7886343042071198e-05,
      "loss": 0.1967,
      "step": 40820
    },
    {
      "epoch": 15.86,
      "learning_rate": 1.788582524271845e-05,
      "loss": 0.1095,
      "step": 40830
    },
    {
      "epoch": 15.86,
      "learning_rate": 1.7885307443365697e-05,
      "loss": 0.0438,
      "step": 40840
    },
    {
      "epoch": 15.86,
      "learning_rate": 1.7884789644012945e-05,
      "loss": 0.0971,
      "step": 40850
    },
    {
      "epoch": 15.87,
      "learning_rate": 1.7884271844660196e-05,
      "loss": 0.1007,
      "step": 40860
    },
    {
      "epoch": 15.87,
      "learning_rate": 1.7883754045307444e-05,
      "loss": 0.2499,
      "step": 40870
    },
    {
      "epoch": 15.88,
      "learning_rate": 1.7883236245954695e-05,
      "loss": 0.1125,
      "step": 40880
    },
    {
      "epoch": 15.88,
      "learning_rate": 1.7882718446601943e-05,
      "loss": 0.1266,
      "step": 40890
    },
    {
      "epoch": 15.88,
      "learning_rate": 1.788220064724919e-05,
      "loss": 0.1306,
      "step": 40900
    },
    {
      "epoch": 15.89,
      "learning_rate": 1.7881682847896442e-05,
      "loss": 0.3246,
      "step": 40910
    },
    {
      "epoch": 15.89,
      "learning_rate": 1.788116504854369e-05,
      "loss": 0.0413,
      "step": 40920
    },
    {
      "epoch": 15.9,
      "learning_rate": 1.7880647249190942e-05,
      "loss": 0.2243,
      "step": 40930
    },
    {
      "epoch": 15.9,
      "learning_rate": 1.788012944983819e-05,
      "loss": 0.1194,
      "step": 40940
    },
    {
      "epoch": 15.9,
      "learning_rate": 1.7879611650485438e-05,
      "loss": 0.2251,
      "step": 40950
    },
    {
      "epoch": 15.91,
      "learning_rate": 1.787909385113269e-05,
      "loss": 0.0519,
      "step": 40960
    },
    {
      "epoch": 15.91,
      "learning_rate": 1.7878576051779937e-05,
      "loss": 0.1821,
      "step": 40970
    },
    {
      "epoch": 15.91,
      "learning_rate": 1.7878058252427185e-05,
      "loss": 0.0971,
      "step": 40980
    },
    {
      "epoch": 15.92,
      "learning_rate": 1.7877540453074436e-05,
      "loss": 0.0712,
      "step": 40990
    },
    {
      "epoch": 15.92,
      "learning_rate": 1.7877022653721684e-05,
      "loss": 0.0794,
      "step": 41000
    },
    {
      "epoch": 15.93,
      "learning_rate": 1.7876504854368935e-05,
      "loss": 0.263,
      "step": 41010
    },
    {
      "epoch": 15.93,
      "learning_rate": 1.7875987055016183e-05,
      "loss": 0.1758,
      "step": 41020
    },
    {
      "epoch": 15.93,
      "learning_rate": 1.787546925566343e-05,
      "loss": 0.32,
      "step": 41030
    },
    {
      "epoch": 15.94,
      "learning_rate": 1.7874951456310682e-05,
      "loss": 0.0714,
      "step": 41040
    },
    {
      "epoch": 15.94,
      "learning_rate": 1.787443365695793e-05,
      "loss": 0.1241,
      "step": 41050
    },
    {
      "epoch": 15.95,
      "learning_rate": 1.7873915857605178e-05,
      "loss": 0.0938,
      "step": 41060
    },
    {
      "epoch": 15.95,
      "learning_rate": 1.787339805825243e-05,
      "loss": 0.0403,
      "step": 41070
    },
    {
      "epoch": 15.95,
      "learning_rate": 1.7872880258899678e-05,
      "loss": 0.1188,
      "step": 41080
    },
    {
      "epoch": 15.96,
      "learning_rate": 1.787236245954693e-05,
      "loss": 0.1944,
      "step": 41090
    },
    {
      "epoch": 15.96,
      "learning_rate": 1.7871844660194177e-05,
      "loss": 0.075,
      "step": 41100
    },
    {
      "epoch": 15.97,
      "learning_rate": 1.7871326860841425e-05,
      "loss": 0.2556,
      "step": 41110
    },
    {
      "epoch": 15.97,
      "learning_rate": 1.7870809061488673e-05,
      "loss": 0.1277,
      "step": 41120
    },
    {
      "epoch": 15.97,
      "learning_rate": 1.7870291262135924e-05,
      "loss": 0.1829,
      "step": 41130
    },
    {
      "epoch": 15.98,
      "learning_rate": 1.7869773462783172e-05,
      "loss": 0.1839,
      "step": 41140
    },
    {
      "epoch": 15.98,
      "learning_rate": 1.7869255663430423e-05,
      "loss": 0.0591,
      "step": 41150
    },
    {
      "epoch": 15.98,
      "learning_rate": 1.786873786407767e-05,
      "loss": 0.0897,
      "step": 41160
    },
    {
      "epoch": 15.99,
      "learning_rate": 1.7868220064724922e-05,
      "loss": 0.1837,
      "step": 41170
    },
    {
      "epoch": 15.99,
      "learning_rate": 1.786770226537217e-05,
      "loss": 0.0294,
      "step": 41180
    },
    {
      "epoch": 16.0,
      "learning_rate": 1.7867184466019418e-05,
      "loss": 0.1413,
      "step": 41190
    },
    {
      "epoch": 16.0,
      "learning_rate": 1.7866666666666666e-05,
      "loss": 0.0933,
      "step": 41200
    },
    {
      "epoch": 16.0,
      "eval_accuracy": 0.953232462173315,
      "eval_loss": 0.2371627688407898,
      "eval_runtime": 8.2508,
      "eval_samples_per_second": 440.563,
      "eval_steps_per_second": 55.146,
      "step": 41200
    },
    {
      "epoch": 16.0,
      "learning_rate": 1.7866148867313917e-05,
      "loss": 0.1805,
      "step": 41210
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.7865631067961165e-05,
      "loss": 0.062,
      "step": 41220
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.7865113268608417e-05,
      "loss": 0.114,
      "step": 41230
    },
    {
      "epoch": 16.02,
      "learning_rate": 1.7864595469255665e-05,
      "loss": 0.0734,
      "step": 41240
    },
    {
      "epoch": 16.02,
      "learning_rate": 1.7864077669902916e-05,
      "loss": 0.3459,
      "step": 41250
    },
    {
      "epoch": 16.02,
      "learning_rate": 1.7863559870550164e-05,
      "loss": 0.0875,
      "step": 41260
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.7863042071197412e-05,
      "loss": 0.0716,
      "step": 41270
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.786252427184466e-05,
      "loss": 0.1456,
      "step": 41280
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.786200647249191e-05,
      "loss": 0.1918,
      "step": 41290
    },
    {
      "epoch": 16.04,
      "learning_rate": 1.786148867313916e-05,
      "loss": 0.1016,
      "step": 41300
    },
    {
      "epoch": 16.04,
      "learning_rate": 1.786097087378641e-05,
      "loss": 0.374,
      "step": 41310
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.7860453074433658e-05,
      "loss": 0.0803,
      "step": 41320
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.785993527508091e-05,
      "loss": 0.1666,
      "step": 41330
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.7859417475728157e-05,
      "loss": 0.121,
      "step": 41340
    },
    {
      "epoch": 16.06,
      "learning_rate": 1.7858899676375405e-05,
      "loss": 0.245,
      "step": 41350
    },
    {
      "epoch": 16.06,
      "learning_rate": 1.7858381877022653e-05,
      "loss": 0.144,
      "step": 41360
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.7857864077669905e-05,
      "loss": 0.0129,
      "step": 41370
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.7857346278317152e-05,
      "loss": 0.0503,
      "step": 41380
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.7856828478964404e-05,
      "loss": 0.0655,
      "step": 41390
    },
    {
      "epoch": 16.08,
      "learning_rate": 1.7856310679611652e-05,
      "loss": 0.1186,
      "step": 41400
    },
    {
      "epoch": 16.08,
      "learning_rate": 1.7855792880258903e-05,
      "loss": 0.0368,
      "step": 41410
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.7855275080906148e-05,
      "loss": 0.1502,
      "step": 41420
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.78547572815534e-05,
      "loss": 0.0478,
      "step": 41430
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.7854239482200647e-05,
      "loss": 0.1422,
      "step": 41440
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.7853721682847898e-05,
      "loss": 0.1043,
      "step": 41450
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.7853203883495146e-05,
      "loss": 0.1982,
      "step": 41460
    },
    {
      "epoch": 16.1,
      "learning_rate": 1.7852686084142397e-05,
      "loss": 0.2513,
      "step": 41470
    },
    {
      "epoch": 16.11,
      "learning_rate": 1.7852168284789645e-05,
      "loss": 0.2204,
      "step": 41480
    },
    {
      "epoch": 16.11,
      "learning_rate": 1.7851650485436897e-05,
      "loss": 0.0062,
      "step": 41490
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.7851132686084145e-05,
      "loss": 0.0779,
      "step": 41500
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.7850614886731392e-05,
      "loss": 0.2577,
      "step": 41510
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.785009708737864e-05,
      "loss": 0.1967,
      "step": 41520
    },
    {
      "epoch": 16.13,
      "learning_rate": 1.784957928802589e-05,
      "loss": 0.03,
      "step": 41530
    },
    {
      "epoch": 16.13,
      "learning_rate": 1.784906148867314e-05,
      "loss": 0.0773,
      "step": 41540
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.784854368932039e-05,
      "loss": 0.0615,
      "step": 41550
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.784802588996764e-05,
      "loss": 0.0592,
      "step": 41560
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.784750809061489e-05,
      "loss": 0.2199,
      "step": 41570
    },
    {
      "epoch": 16.15,
      "learning_rate": 1.7846990291262138e-05,
      "loss": 0.0824,
      "step": 41580
    },
    {
      "epoch": 16.15,
      "learning_rate": 1.7846472491909386e-05,
      "loss": 0.0444,
      "step": 41590
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.7845954692556634e-05,
      "loss": 0.1335,
      "step": 41600
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.7845436893203885e-05,
      "loss": 0.1497,
      "step": 41610
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.7844919093851133e-05,
      "loss": 0.0379,
      "step": 41620
    },
    {
      "epoch": 16.17,
      "learning_rate": 1.7844401294498384e-05,
      "loss": 0.0691,
      "step": 41630
    },
    {
      "epoch": 16.17,
      "learning_rate": 1.7843883495145632e-05,
      "loss": 0.0905,
      "step": 41640
    },
    {
      "epoch": 16.17,
      "learning_rate": 1.7843365695792884e-05,
      "loss": 0.1528,
      "step": 41650
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.784284789644013e-05,
      "loss": 0.0865,
      "step": 41660
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.784233009708738e-05,
      "loss": 0.1851,
      "step": 41670
    },
    {
      "epoch": 16.19,
      "learning_rate": 1.7841812297734627e-05,
      "loss": 0.1702,
      "step": 41680
    },
    {
      "epoch": 16.19,
      "learning_rate": 1.784129449838188e-05,
      "loss": 0.0267,
      "step": 41690
    },
    {
      "epoch": 16.19,
      "learning_rate": 1.7840776699029127e-05,
      "loss": 0.1301,
      "step": 41700
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.7840258899676378e-05,
      "loss": 0.2268,
      "step": 41710
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.7839741100323626e-05,
      "loss": 0.0466,
      "step": 41720
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.7839223300970877e-05,
      "loss": 0.1418,
      "step": 41730
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.7838705501618125e-05,
      "loss": 0.135,
      "step": 41740
    },
    {
      "epoch": 16.21,
      "learning_rate": 1.7838187702265373e-05,
      "loss": 0.0907,
      "step": 41750
    },
    {
      "epoch": 16.22,
      "learning_rate": 1.783766990291262e-05,
      "loss": 0.2559,
      "step": 41760
    },
    {
      "epoch": 16.22,
      "learning_rate": 1.7837152103559872e-05,
      "loss": 0.2166,
      "step": 41770
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.783663430420712e-05,
      "loss": 0.2158,
      "step": 41780
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.783611650485437e-05,
      "loss": 0.1589,
      "step": 41790
    },
    {
      "epoch": 16.23,
      "learning_rate": 1.783559870550162e-05,
      "loss": 0.2711,
      "step": 41800
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.783508090614887e-05,
      "loss": 0.1218,
      "step": 41810
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.783456310679612e-05,
      "loss": 0.0407,
      "step": 41820
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.7834045307443367e-05,
      "loss": 0.0617,
      "step": 41830
    },
    {
      "epoch": 16.25,
      "learning_rate": 1.7833527508090615e-05,
      "loss": 0.2487,
      "step": 41840
    },
    {
      "epoch": 16.25,
      "learning_rate": 1.7833009708737866e-05,
      "loss": 0.0529,
      "step": 41850
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.7832491909385114e-05,
      "loss": 0.115,
      "step": 41860
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.7831974110032365e-05,
      "loss": 0.0215,
      "step": 41870
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.7831456310679613e-05,
      "loss": 0.0421,
      "step": 41880
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.7830938511326864e-05,
      "loss": 0.2363,
      "step": 41890
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.7830420711974112e-05,
      "loss": 0.1887,
      "step": 41900
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.782990291262136e-05,
      "loss": 0.0109,
      "step": 41910
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.7829385113268608e-05,
      "loss": 0.1738,
      "step": 41920
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.782886731391586e-05,
      "loss": 0.1176,
      "step": 41930
    },
    {
      "epoch": 16.29,
      "learning_rate": 1.7828349514563107e-05,
      "loss": 0.0595,
      "step": 41940
    },
    {
      "epoch": 16.29,
      "learning_rate": 1.782783171521036e-05,
      "loss": 0.1036,
      "step": 41950
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.7827313915857607e-05,
      "loss": 0.0619,
      "step": 41960
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.7826796116504858e-05,
      "loss": 0.0761,
      "step": 41970
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.7826278317152106e-05,
      "loss": 0.0891,
      "step": 41980
    },
    {
      "epoch": 16.31,
      "learning_rate": 1.7825760517799354e-05,
      "loss": 0.0929,
      "step": 41990
    },
    {
      "epoch": 16.31,
      "learning_rate": 1.7825242718446602e-05,
      "loss": 0.0498,
      "step": 42000
    },
    {
      "epoch": 16.31,
      "learning_rate": 1.7824724919093853e-05,
      "loss": 0.1508,
      "step": 42010
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.78242071197411e-05,
      "loss": 0.0256,
      "step": 42020
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.7823689320388352e-05,
      "loss": 0.3526,
      "step": 42030
    },
    {
      "epoch": 16.33,
      "learning_rate": 1.78231715210356e-05,
      "loss": 0.2692,
      "step": 42040
    },
    {
      "epoch": 16.33,
      "learning_rate": 1.782265372168285e-05,
      "loss": 0.1235,
      "step": 42050
    },
    {
      "epoch": 16.33,
      "learning_rate": 1.78221359223301e-05,
      "loss": 0.0486,
      "step": 42060
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.7821618122977347e-05,
      "loss": 0.0587,
      "step": 42070
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.7821100323624595e-05,
      "loss": 0.1331,
      "step": 42080
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.7820582524271847e-05,
      "loss": 0.1003,
      "step": 42090
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.7820064724919094e-05,
      "loss": 0.0043,
      "step": 42100
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.7819546925566346e-05,
      "loss": 0.1049,
      "step": 42110
    },
    {
      "epoch": 16.36,
      "learning_rate": 1.7819029126213594e-05,
      "loss": 0.0632,
      "step": 42120
    },
    {
      "epoch": 16.36,
      "learning_rate": 1.7818511326860845e-05,
      "loss": 0.1443,
      "step": 42130
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.7817993527508093e-05,
      "loss": 0.0493,
      "step": 42140
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.781747572815534e-05,
      "loss": 0.1188,
      "step": 42150
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.781695792880259e-05,
      "loss": 0.1276,
      "step": 42160
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.781644012944984e-05,
      "loss": 0.1071,
      "step": 42170
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.7815922330097088e-05,
      "loss": 0.0263,
      "step": 42180
    },
    {
      "epoch": 16.38,
      "learning_rate": 1.781540453074434e-05,
      "loss": 0.0263,
      "step": 42190
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.7814886731391587e-05,
      "loss": 0.1223,
      "step": 42200
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.7814368932038835e-05,
      "loss": 0.1479,
      "step": 42210
    },
    {
      "epoch": 16.4,
      "learning_rate": 1.7813851132686087e-05,
      "loss": 0.1376,
      "step": 42220
    },
    {
      "epoch": 16.4,
      "learning_rate": 1.7813333333333334e-05,
      "loss": 0.0379,
      "step": 42230
    },
    {
      "epoch": 16.4,
      "learning_rate": 1.7812815533980582e-05,
      "loss": 0.1754,
      "step": 42240
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.7812297734627834e-05,
      "loss": 0.0898,
      "step": 42250
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.781177993527508e-05,
      "loss": 0.098,
      "step": 42260
    },
    {
      "epoch": 16.42,
      "learning_rate": 1.7811262135922333e-05,
      "loss": 0.1197,
      "step": 42270
    },
    {
      "epoch": 16.42,
      "learning_rate": 1.781074433656958e-05,
      "loss": 0.1081,
      "step": 42280
    },
    {
      "epoch": 16.42,
      "learning_rate": 1.781022653721683e-05,
      "loss": 0.1308,
      "step": 42290
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.780970873786408e-05,
      "loss": 0.3245,
      "step": 42300
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.7809190938511328e-05,
      "loss": 0.1148,
      "step": 42310
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.7808673139158576e-05,
      "loss": 0.2022,
      "step": 42320
    },
    {
      "epoch": 16.44,
      "learning_rate": 1.7808155339805827e-05,
      "loss": 0.1809,
      "step": 42330
    },
    {
      "epoch": 16.44,
      "learning_rate": 1.7807637540453075e-05,
      "loss": 0.1193,
      "step": 42340
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.7807119741100326e-05,
      "loss": 0.1444,
      "step": 42350
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.7806601941747574e-05,
      "loss": 0.182,
      "step": 42360
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.7806084142394822e-05,
      "loss": 0.272,
      "step": 42370
    },
    {
      "epoch": 16.46,
      "learning_rate": 1.7805566343042074e-05,
      "loss": 0.0614,
      "step": 42380
    },
    {
      "epoch": 16.46,
      "learning_rate": 1.780504854368932e-05,
      "loss": 0.0546,
      "step": 42390
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.780453074433657e-05,
      "loss": 0.1625,
      "step": 42400
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.780401294498382e-05,
      "loss": 0.228,
      "step": 42410
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.780349514563107e-05,
      "loss": 0.1584,
      "step": 42420
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.780297734627832e-05,
      "loss": 0.1176,
      "step": 42430
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.7802459546925568e-05,
      "loss": 0.2057,
      "step": 42440
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.7801941747572816e-05,
      "loss": 0.164,
      "step": 42450
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.7801423948220067e-05,
      "loss": 0.0097,
      "step": 42460
    },
    {
      "epoch": 16.49,
      "learning_rate": 1.7800906148867315e-05,
      "loss": 0.1816,
      "step": 42470
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.7800388349514563e-05,
      "loss": 0.0245,
      "step": 42480
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.7799870550161814e-05,
      "loss": 0.002,
      "step": 42490
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.7799352750809062e-05,
      "loss": 0.0919,
      "step": 42500
    },
    {
      "epoch": 16.51,
      "learning_rate": 1.7798834951456314e-05,
      "loss": 0.0517,
      "step": 42510
    },
    {
      "epoch": 16.51,
      "learning_rate": 1.779831715210356e-05,
      "loss": 0.1919,
      "step": 42520
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.779779935275081e-05,
      "loss": 0.1335,
      "step": 42530
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.779728155339806e-05,
      "loss": 0.0895,
      "step": 42540
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.779676375404531e-05,
      "loss": 0.0179,
      "step": 42550
    },
    {
      "epoch": 16.53,
      "learning_rate": 1.7796245954692557e-05,
      "loss": 0.1697,
      "step": 42560
    },
    {
      "epoch": 16.53,
      "learning_rate": 1.7795728155339808e-05,
      "loss": 0.1091,
      "step": 42570
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.7795210355987056e-05,
      "loss": 0.2665,
      "step": 42580
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.7794692556634304e-05,
      "loss": 0.0633,
      "step": 42590
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.7794174757281555e-05,
      "loss": 0.0405,
      "step": 42600
    },
    {
      "epoch": 16.55,
      "learning_rate": 1.7793656957928803e-05,
      "loss": 0.0191,
      "step": 42610
    },
    {
      "epoch": 16.55,
      "learning_rate": 1.7793139158576054e-05,
      "loss": 0.1624,
      "step": 42620
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.7792621359223302e-05,
      "loss": 0.0654,
      "step": 42630
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.7792103559870554e-05,
      "loss": 0.0721,
      "step": 42640
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.77915857605178e-05,
      "loss": 0.1328,
      "step": 42650
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.779106796116505e-05,
      "loss": 0.0583,
      "step": 42660
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.7790550161812297e-05,
      "loss": 0.0011,
      "step": 42670
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.779003236245955e-05,
      "loss": 0.0286,
      "step": 42680
    },
    {
      "epoch": 16.58,
      "learning_rate": 1.7789514563106797e-05,
      "loss": 0.1218,
      "step": 42690
    },
    {
      "epoch": 16.58,
      "learning_rate": 1.7788996763754048e-05,
      "loss": 0.2277,
      "step": 42700
    },
    {
      "epoch": 16.59,
      "learning_rate": 1.7788478964401296e-05,
      "loss": 0.2786,
      "step": 42710
    },
    {
      "epoch": 16.59,
      "learning_rate": 1.7787961165048547e-05,
      "loss": 0.0296,
      "step": 42720
    },
    {
      "epoch": 16.59,
      "learning_rate": 1.7787443365695795e-05,
      "loss": 0.0994,
      "step": 42730
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.7786925566343043e-05,
      "loss": 0.1604,
      "step": 42740
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.778640776699029e-05,
      "loss": 0.2233,
      "step": 42750
    },
    {
      "epoch": 16.61,
      "learning_rate": 1.7785889967637542e-05,
      "loss": 0.1187,
      "step": 42760
    },
    {
      "epoch": 16.61,
      "learning_rate": 1.778537216828479e-05,
      "loss": 0.112,
      "step": 42770
    },
    {
      "epoch": 16.61,
      "learning_rate": 1.778485436893204e-05,
      "loss": 0.1021,
      "step": 42780
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.778433656957929e-05,
      "loss": 0.1798,
      "step": 42790
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.778381877022654e-05,
      "loss": 0.1738,
      "step": 42800
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.778330097087379e-05,
      "loss": 0.2663,
      "step": 42810
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.7782783171521037e-05,
      "loss": 0.2414,
      "step": 42820
    },
    {
      "epoch": 16.63,
      "learning_rate": 1.7782265372168284e-05,
      "loss": 0.0899,
      "step": 42830
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.7781747572815536e-05,
      "loss": 0.1003,
      "step": 42840
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.7781229773462784e-05,
      "loss": 0.0419,
      "step": 42850
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.7780711974110035e-05,
      "loss": 0.1807,
      "step": 42860
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.7780194174757283e-05,
      "loss": 0.1218,
      "step": 42870
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.7779676375404534e-05,
      "loss": 0.22,
      "step": 42880
    },
    {
      "epoch": 16.66,
      "learning_rate": 1.777915857605178e-05,
      "loss": 0.0312,
      "step": 42890
    },
    {
      "epoch": 16.66,
      "learning_rate": 1.777864077669903e-05,
      "loss": 0.0735,
      "step": 42900
    },
    {
      "epoch": 16.66,
      "learning_rate": 1.7778122977346278e-05,
      "loss": 0.11,
      "step": 42910
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.777760517799353e-05,
      "loss": 0.0098,
      "step": 42920
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.7777087378640777e-05,
      "loss": 0.0133,
      "step": 42930
    },
    {
      "epoch": 16.68,
      "learning_rate": 1.777656957928803e-05,
      "loss": 0.1085,
      "step": 42940
    },
    {
      "epoch": 16.68,
      "learning_rate": 1.7776051779935276e-05,
      "loss": 0.0279,
      "step": 42950
    },
    {
      "epoch": 16.68,
      "learning_rate": 1.7775533980582528e-05,
      "loss": 0.161,
      "step": 42960
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.7775016181229772e-05,
      "loss": 0.0566,
      "step": 42970
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.7774498381877024e-05,
      "loss": 0.0877,
      "step": 42980
    },
    {
      "epoch": 16.7,
      "learning_rate": 1.777398058252427e-05,
      "loss": 0.2516,
      "step": 42990
    },
    {
      "epoch": 16.7,
      "learning_rate": 1.7773462783171523e-05,
      "loss": 0.1728,
      "step": 43000
    },
    {
      "epoch": 16.7,
      "learning_rate": 1.777294498381877e-05,
      "loss": 0.2104,
      "step": 43010
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.7772427184466022e-05,
      "loss": 0.218,
      "step": 43020
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.777190938511327e-05,
      "loss": 0.1869,
      "step": 43030
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.777139158576052e-05,
      "loss": 0.0989,
      "step": 43040
    },
    {
      "epoch": 16.72,
      "learning_rate": 1.7770873786407766e-05,
      "loss": 0.1797,
      "step": 43050
    },
    {
      "epoch": 16.72,
      "learning_rate": 1.7770355987055017e-05,
      "loss": 0.0726,
      "step": 43060
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.7769838187702265e-05,
      "loss": 0.1135,
      "step": 43070
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.7769320388349516e-05,
      "loss": 0.085,
      "step": 43080
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.7768802588996764e-05,
      "loss": 0.2452,
      "step": 43090
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.7768284789644016e-05,
      "loss": 0.0529,
      "step": 43100
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.7767766990291264e-05,
      "loss": 0.248,
      "step": 43110
    },
    {
      "epoch": 16.75,
      "learning_rate": 1.7767249190938515e-05,
      "loss": 0.3458,
      "step": 43120
    },
    {
      "epoch": 16.75,
      "learning_rate": 1.776673139158576e-05,
      "loss": 0.1804,
      "step": 43130
    },
    {
      "epoch": 16.75,
      "learning_rate": 1.776621359223301e-05,
      "loss": 0.1024,
      "step": 43140
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.776569579288026e-05,
      "loss": 0.207,
      "step": 43150
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.776517799352751e-05,
      "loss": 0.0705,
      "step": 43160
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.7764660194174758e-05,
      "loss": 0.1529,
      "step": 43170
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.776414239482201e-05,
      "loss": 0.052,
      "step": 43180
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.7763624595469257e-05,
      "loss": 0.123,
      "step": 43190
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.776310679611651e-05,
      "loss": 0.0278,
      "step": 43200
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.7762588996763756e-05,
      "loss": 0.1315,
      "step": 43210
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.7762071197411004e-05,
      "loss": 0.1767,
      "step": 43220
    },
    {
      "epoch": 16.79,
      "learning_rate": 1.7761553398058252e-05,
      "loss": 0.0305,
      "step": 43230
    },
    {
      "epoch": 16.79,
      "learning_rate": 1.7761035598705504e-05,
      "loss": 0.1318,
      "step": 43240
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.776051779935275e-05,
      "loss": 0.0422,
      "step": 43250
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.7760000000000003e-05,
      "loss": 0.2118,
      "step": 43260
    },
    {
      "epoch": 16.8,
      "learning_rate": 1.775948220064725e-05,
      "loss": 0.1572,
      "step": 43270
    },
    {
      "epoch": 16.81,
      "learning_rate": 1.7758964401294502e-05,
      "loss": 0.268,
      "step": 43280
    },
    {
      "epoch": 16.81,
      "learning_rate": 1.775844660194175e-05,
      "loss": 0.2522,
      "step": 43290
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.7757928802588998e-05,
      "loss": 0.0441,
      "step": 43300
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.7757411003236246e-05,
      "loss": 0.2637,
      "step": 43310
    },
    {
      "epoch": 16.82,
      "learning_rate": 1.7756893203883497e-05,
      "loss": 0.2136,
      "step": 43320
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.7756375404530745e-05,
      "loss": 0.1961,
      "step": 43330
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.7755857605177996e-05,
      "loss": 0.0309,
      "step": 43340
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.7755339805825244e-05,
      "loss": 0.1434,
      "step": 43350
    },
    {
      "epoch": 16.84,
      "learning_rate": 1.7754822006472496e-05,
      "loss": 0.4176,
      "step": 43360
    },
    {
      "epoch": 16.84,
      "learning_rate": 1.7754304207119743e-05,
      "loss": 0.0601,
      "step": 43370
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.775378640776699e-05,
      "loss": 0.0984,
      "step": 43380
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.775326860841424e-05,
      "loss": 0.1693,
      "step": 43390
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.775275080906149e-05,
      "loss": 0.2335,
      "step": 43400
    },
    {
      "epoch": 16.86,
      "learning_rate": 1.775223300970874e-05,
      "loss": 0.2428,
      "step": 43410
    },
    {
      "epoch": 16.86,
      "learning_rate": 1.775171521035599e-05,
      "loss": 0.0497,
      "step": 43420
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.7751197411003238e-05,
      "loss": 0.2228,
      "step": 43430
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.775067961165049e-05,
      "loss": 0.0434,
      "step": 43440
    },
    {
      "epoch": 16.87,
      "learning_rate": 1.7750161812297737e-05,
      "loss": 0.2069,
      "step": 43450
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.7749644012944985e-05,
      "loss": 0.1167,
      "step": 43460
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.7749126213592233e-05,
      "loss": 0.0365,
      "step": 43470
    },
    {
      "epoch": 16.89,
      "learning_rate": 1.7748608414239484e-05,
      "loss": 0.2387,
      "step": 43480
    },
    {
      "epoch": 16.89,
      "learning_rate": 1.7748090614886732e-05,
      "loss": 0.0244,
      "step": 43490
    },
    {
      "epoch": 16.89,
      "learning_rate": 1.7747572815533983e-05,
      "loss": 0.2399,
      "step": 43500
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.774705501618123e-05,
      "loss": 0.1213,
      "step": 43510
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.7746537216828483e-05,
      "loss": 0.0406,
      "step": 43520
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.774601941747573e-05,
      "loss": 0.1046,
      "step": 43530
    },
    {
      "epoch": 16.91,
      "learning_rate": 1.774550161812298e-05,
      "loss": 0.1664,
      "step": 43540
    },
    {
      "epoch": 16.91,
      "learning_rate": 1.7744983818770226e-05,
      "loss": 0.3325,
      "step": 43550
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.7744466019417478e-05,
      "loss": 0.2133,
      "step": 43560
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.7743948220064726e-05,
      "loss": 0.0404,
      "step": 43570
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.7743430420711977e-05,
      "loss": 0.1367,
      "step": 43580
    },
    {
      "epoch": 16.93,
      "learning_rate": 1.7742912621359225e-05,
      "loss": 0.2669,
      "step": 43590
    },
    {
      "epoch": 16.93,
      "learning_rate": 1.7742394822006476e-05,
      "loss": 0.1362,
      "step": 43600
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.7741877022653724e-05,
      "loss": 0.08,
      "step": 43610
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.7741359223300972e-05,
      "loss": 0.073,
      "step": 43620
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.774084142394822e-05,
      "loss": 0.1277,
      "step": 43630
    },
    {
      "epoch": 16.95,
      "learning_rate": 1.774032362459547e-05,
      "loss": 0.0986,
      "step": 43640
    },
    {
      "epoch": 16.95,
      "learning_rate": 1.773980582524272e-05,
      "loss": 0.2773,
      "step": 43650
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.773928802588997e-05,
      "loss": 0.1769,
      "step": 43660
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.773877022653722e-05,
      "loss": 0.1013,
      "step": 43670
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.7738252427184466e-05,
      "loss": 0.1835,
      "step": 43680
    },
    {
      "epoch": 16.97,
      "learning_rate": 1.7737734627831718e-05,
      "loss": 0.0483,
      "step": 43690
    },
    {
      "epoch": 16.97,
      "learning_rate": 1.7737216828478966e-05,
      "loss": 0.0968,
      "step": 43700
    },
    {
      "epoch": 16.97,
      "learning_rate": 1.7736699029126214e-05,
      "loss": 0.1067,
      "step": 43710
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.7736181229773465e-05,
      "loss": 0.1283,
      "step": 43720
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.7735663430420713e-05,
      "loss": 0.1628,
      "step": 43730
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.7735145631067964e-05,
      "loss": 0.2357,
      "step": 43740
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.7734627831715212e-05,
      "loss": 0.0527,
      "step": 43750
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.773411003236246e-05,
      "loss": 0.0019,
      "step": 43760
    },
    {
      "epoch": 17.0,
      "learning_rate": 1.773359223300971e-05,
      "loss": 0.17,
      "step": 43770
    },
    {
      "epoch": 17.0,
      "eval_accuracy": 0.9557083906464925,
      "eval_loss": 0.21660885214805603,
      "eval_runtime": 8.2156,
      "eval_samples_per_second": 442.449,
      "eval_steps_per_second": 55.382,
      "step": 43775
    },
    {
      "epoch": 17.0,
      "learning_rate": 1.773307443365696e-05,
      "loss": 0.179,
      "step": 43780
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.7732556634304207e-05,
      "loss": 0.0662,
      "step": 43790
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.773203883495146e-05,
      "loss": 0.0542,
      "step": 43800
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.7731521035598706e-05,
      "loss": 0.098,
      "step": 43810
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.7731003236245958e-05,
      "loss": 0.0564,
      "step": 43820
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.7730485436893206e-05,
      "loss": 0.1391,
      "step": 43830
    },
    {
      "epoch": 17.03,
      "learning_rate": 1.7729967637540453e-05,
      "loss": 0.0964,
      "step": 43840
    },
    {
      "epoch": 17.03,
      "learning_rate": 1.7729449838187705e-05,
      "loss": 0.0783,
      "step": 43850
    },
    {
      "epoch": 17.03,
      "learning_rate": 1.7728932038834953e-05,
      "loss": 0.0955,
      "step": 43860
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.77284142394822e-05,
      "loss": 0.1657,
      "step": 43870
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.7727896440129452e-05,
      "loss": 0.03,
      "step": 43880
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.77273786407767e-05,
      "loss": 0.0038,
      "step": 43890
    },
    {
      "epoch": 17.05,
      "learning_rate": 1.772686084142395e-05,
      "loss": 0.2355,
      "step": 43900
    },
    {
      "epoch": 17.05,
      "learning_rate": 1.77263430420712e-05,
      "loss": 0.1176,
      "step": 43910
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.7725825242718447e-05,
      "loss": 0.0433,
      "step": 43920
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.77253074433657e-05,
      "loss": 0.1647,
      "step": 43930
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.7724789644012946e-05,
      "loss": 0.2472,
      "step": 43940
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.7724271844660194e-05,
      "loss": 0.2023,
      "step": 43950
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.7723754045307446e-05,
      "loss": 0.0983,
      "step": 43960
    },
    {
      "epoch": 17.08,
      "learning_rate": 1.7723236245954693e-05,
      "loss": 0.0446,
      "step": 43970
    },
    {
      "epoch": 17.08,
      "learning_rate": 1.7722718446601945e-05,
      "loss": 0.1182,
      "step": 43980
    },
    {
      "epoch": 17.08,
      "learning_rate": 1.7722200647249193e-05,
      "loss": 0.0486,
      "step": 43990
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.772168284789644e-05,
      "loss": 0.2129,
      "step": 44000
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.7721165048543692e-05,
      "loss": 0.07,
      "step": 44010
    },
    {
      "epoch": 17.1,
      "learning_rate": 1.772064724919094e-05,
      "loss": 0.0754,
      "step": 44020
    },
    {
      "epoch": 17.1,
      "learning_rate": 1.7720129449838188e-05,
      "loss": 0.2169,
      "step": 44030
    },
    {
      "epoch": 17.1,
      "learning_rate": 1.771961165048544e-05,
      "loss": 0.1001,
      "step": 44040
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.7719093851132687e-05,
      "loss": 0.0647,
      "step": 44050
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.7718576051779935e-05,
      "loss": 0.0436,
      "step": 44060
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.7718058252427186e-05,
      "loss": 0.2746,
      "step": 44070
    },
    {
      "epoch": 17.12,
      "learning_rate": 1.7717540453074434e-05,
      "loss": 0.0876,
      "step": 44080
    },
    {
      "epoch": 17.12,
      "learning_rate": 1.7717022653721685e-05,
      "loss": 0.0782,
      "step": 44090
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.7716504854368933e-05,
      "loss": 0.0963,
      "step": 44100
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.771598705501618e-05,
      "loss": 0.0576,
      "step": 44110
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.7715469255663433e-05,
      "loss": 0.13,
      "step": 44120
    },
    {
      "epoch": 17.14,
      "learning_rate": 1.771495145631068e-05,
      "loss": 0.123,
      "step": 44130
    },
    {
      "epoch": 17.14,
      "learning_rate": 1.771443365695793e-05,
      "loss": 0.0775,
      "step": 44140
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.771391585760518e-05,
      "loss": 0.1593,
      "step": 44150
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.7713398058252428e-05,
      "loss": 0.1236,
      "step": 44160
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.771288025889968e-05,
      "loss": 0.0757,
      "step": 44170
    },
    {
      "epoch": 17.16,
      "learning_rate": 1.7712362459546927e-05,
      "loss": 0.0481,
      "step": 44180
    },
    {
      "epoch": 17.16,
      "learning_rate": 1.7711844660194175e-05,
      "loss": 0.2573,
      "step": 44190
    },
    {
      "epoch": 17.17,
      "learning_rate": 1.7711326860841426e-05,
      "loss": 0.2014,
      "step": 44200
    },
    {
      "epoch": 17.17,
      "learning_rate": 1.7710809061488674e-05,
      "loss": 0.2625,
      "step": 44210
    },
    {
      "epoch": 17.17,
      "learning_rate": 1.7710291262135922e-05,
      "loss": 0.1473,
      "step": 44220
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.7709773462783173e-05,
      "loss": 0.0144,
      "step": 44230
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.770925566343042e-05,
      "loss": 0.1752,
      "step": 44240
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.7708737864077673e-05,
      "loss": 0.1109,
      "step": 44250
    },
    {
      "epoch": 17.19,
      "learning_rate": 1.770822006472492e-05,
      "loss": 0.1503,
      "step": 44260
    },
    {
      "epoch": 17.19,
      "learning_rate": 1.770770226537217e-05,
      "loss": 0.2252,
      "step": 44270
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.770718446601942e-05,
      "loss": 0.107,
      "step": 44280
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.7706666666666668e-05,
      "loss": 0.0922,
      "step": 44290
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.7706148867313916e-05,
      "loss": 0.1514,
      "step": 44300
    },
    {
      "epoch": 17.21,
      "learning_rate": 1.7705631067961167e-05,
      "loss": 0.0022,
      "step": 44310
    },
    {
      "epoch": 17.21,
      "learning_rate": 1.7705113268608415e-05,
      "loss": 0.0029,
      "step": 44320
    },
    {
      "epoch": 17.22,
      "learning_rate": 1.7704595469255666e-05,
      "loss": 0.1095,
      "step": 44330
    },
    {
      "epoch": 17.22,
      "learning_rate": 1.7704077669902914e-05,
      "loss": 0.0231,
      "step": 44340
    },
    {
      "epoch": 17.22,
      "learning_rate": 1.7703559870550165e-05,
      "loss": 0.1088,
      "step": 44350
    },
    {
      "epoch": 17.23,
      "learning_rate": 1.770304207119741e-05,
      "loss": 0.0391,
      "step": 44360
    },
    {
      "epoch": 17.23,
      "learning_rate": 1.770252427184466e-05,
      "loss": 0.226,
      "step": 44370
    },
    {
      "epoch": 17.23,
      "learning_rate": 1.770200647249191e-05,
      "loss": 0.1613,
      "step": 44380
    },
    {
      "epoch": 17.24,
      "learning_rate": 1.770148867313916e-05,
      "loss": 0.0935,
      "step": 44390
    },
    {
      "epoch": 17.24,
      "learning_rate": 1.770097087378641e-05,
      "loss": 0.1589,
      "step": 44400
    },
    {
      "epoch": 17.25,
      "learning_rate": 1.770045307443366e-05,
      "loss": 0.1347,
      "step": 44410
    },
    {
      "epoch": 17.25,
      "learning_rate": 1.7699935275080908e-05,
      "loss": 0.1847,
      "step": 44420
    },
    {
      "epoch": 17.25,
      "learning_rate": 1.769941747572816e-05,
      "loss": 0.1616,
      "step": 44430
    },
    {
      "epoch": 17.26,
      "learning_rate": 1.7698899676375403e-05,
      "loss": 0.1403,
      "step": 44440
    },
    {
      "epoch": 17.26,
      "learning_rate": 1.7698381877022655e-05,
      "loss": 0.0986,
      "step": 44450
    },
    {
      "epoch": 17.27,
      "learning_rate": 1.7697864077669903e-05,
      "loss": 0.0329,
      "step": 44460
    },
    {
      "epoch": 17.27,
      "learning_rate": 1.7697346278317154e-05,
      "loss": 0.2119,
      "step": 44470
    },
    {
      "epoch": 17.27,
      "learning_rate": 1.7696828478964402e-05,
      "loss": 0.0905,
      "step": 44480
    },
    {
      "epoch": 17.28,
      "learning_rate": 1.7696310679611653e-05,
      "loss": 0.0675,
      "step": 44490
    },
    {
      "epoch": 17.28,
      "learning_rate": 1.76957928802589e-05,
      "loss": 0.1257,
      "step": 44500
    },
    {
      "epoch": 17.29,
      "learning_rate": 1.7695275080906152e-05,
      "loss": 0.1056,
      "step": 44510
    },
    {
      "epoch": 17.29,
      "learning_rate": 1.7694757281553397e-05,
      "loss": 0.1767,
      "step": 44520
    },
    {
      "epoch": 17.29,
      "learning_rate": 1.769423948220065e-05,
      "loss": 0.0688,
      "step": 44530
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.7693721682847896e-05,
      "loss": 0.0957,
      "step": 44540
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.7693203883495148e-05,
      "loss": 0.1691,
      "step": 44550
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.7692686084142396e-05,
      "loss": 0.1554,
      "step": 44560
    },
    {
      "epoch": 17.31,
      "learning_rate": 1.7692168284789647e-05,
      "loss": 0.0364,
      "step": 44570
    },
    {
      "epoch": 17.31,
      "learning_rate": 1.7691650485436895e-05,
      "loss": 0.05,
      "step": 44580
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.7691132686084146e-05,
      "loss": 0.0879,
      "step": 44590
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.769061488673139e-05,
      "loss": 0.1228,
      "step": 44600
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.7690097087378642e-05,
      "loss": 0.1043,
      "step": 44610
    },
    {
      "epoch": 17.33,
      "learning_rate": 1.768957928802589e-05,
      "loss": 0.0611,
      "step": 44620
    },
    {
      "epoch": 17.33,
      "learning_rate": 1.768906148867314e-05,
      "loss": 0.0846,
      "step": 44630
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.768854368932039e-05,
      "loss": 0.1771,
      "step": 44640
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.768802588996764e-05,
      "loss": 0.0085,
      "step": 44650
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.7687508090614888e-05,
      "loss": 0.1479,
      "step": 44660
    },
    {
      "epoch": 17.35,
      "learning_rate": 1.768699029126214e-05,
      "loss": 0.1477,
      "step": 44670
    },
    {
      "epoch": 17.35,
      "learning_rate": 1.7686472491909384e-05,
      "loss": 0.0772,
      "step": 44680
    },
    {
      "epoch": 17.36,
      "learning_rate": 1.7685954692556635e-05,
      "loss": 0.0692,
      "step": 44690
    },
    {
      "epoch": 17.36,
      "learning_rate": 1.7685436893203883e-05,
      "loss": 0.0737,
      "step": 44700
    },
    {
      "epoch": 17.36,
      "learning_rate": 1.7684919093851135e-05,
      "loss": 0.1382,
      "step": 44710
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.7684401294498383e-05,
      "loss": 0.1664,
      "step": 44720
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.7683883495145634e-05,
      "loss": 0.1907,
      "step": 44730
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.7683365695792882e-05,
      "loss": 0.0496,
      "step": 44740
    },
    {
      "epoch": 17.38,
      "learning_rate": 1.7682847896440133e-05,
      "loss": 0.1154,
      "step": 44750
    },
    {
      "epoch": 17.38,
      "learning_rate": 1.7682330097087378e-05,
      "loss": 0.1828,
      "step": 44760
    },
    {
      "epoch": 17.39,
      "learning_rate": 1.768181229773463e-05,
      "loss": 0.1881,
      "step": 44770
    },
    {
      "epoch": 17.39,
      "learning_rate": 1.7681294498381877e-05,
      "loss": 0.0786,
      "step": 44780
    },
    {
      "epoch": 17.39,
      "learning_rate": 1.7680776699029128e-05,
      "loss": 0.0438,
      "step": 44790
    },
    {
      "epoch": 17.4,
      "learning_rate": 1.7680258899676376e-05,
      "loss": 0.112,
      "step": 44800
    },
    {
      "epoch": 17.4,
      "learning_rate": 1.7679741100323627e-05,
      "loss": 0.2575,
      "step": 44810
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.7679223300970875e-05,
      "loss": 0.1025,
      "step": 44820
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.7678705501618127e-05,
      "loss": 0.1471,
      "step": 44830
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.767818770226537e-05,
      "loss": 0.1146,
      "step": 44840
    },
    {
      "epoch": 17.42,
      "learning_rate": 1.7677669902912623e-05,
      "loss": 0.0921,
      "step": 44850
    },
    {
      "epoch": 17.42,
      "learning_rate": 1.767715210355987e-05,
      "loss": 0.1402,
      "step": 44860
    },
    {
      "epoch": 17.43,
      "learning_rate": 1.7676634304207122e-05,
      "loss": 0.1369,
      "step": 44870
    },
    {
      "epoch": 17.43,
      "learning_rate": 1.767611650485437e-05,
      "loss": 0.1052,
      "step": 44880
    },
    {
      "epoch": 17.43,
      "learning_rate": 1.767559870550162e-05,
      "loss": 0.0109,
      "step": 44890
    },
    {
      "epoch": 17.44,
      "learning_rate": 1.767508090614887e-05,
      "loss": 0.0268,
      "step": 44900
    },
    {
      "epoch": 17.44,
      "learning_rate": 1.767456310679612e-05,
      "loss": 0.0947,
      "step": 44910
    },
    {
      "epoch": 17.44,
      "learning_rate": 1.7674045307443368e-05,
      "loss": 0.015,
      "step": 44920
    },
    {
      "epoch": 17.45,
      "learning_rate": 1.7673527508090616e-05,
      "loss": 0.0845,
      "step": 44930
    },
    {
      "epoch": 17.45,
      "learning_rate": 1.7673009708737864e-05,
      "loss": 0.2021,
      "step": 44940
    },
    {
      "epoch": 17.46,
      "learning_rate": 1.7672491909385115e-05,
      "loss": 0.2342,
      "step": 44950
    },
    {
      "epoch": 17.46,
      "learning_rate": 1.7671974110032363e-05,
      "loss": 0.0376,
      "step": 44960
    },
    {
      "epoch": 17.46,
      "learning_rate": 1.7671456310679615e-05,
      "loss": 0.0525,
      "step": 44970
    },
    {
      "epoch": 17.47,
      "learning_rate": 1.7670938511326863e-05,
      "loss": 0.4019,
      "step": 44980
    },
    {
      "epoch": 17.47,
      "learning_rate": 1.7670420711974114e-05,
      "loss": 0.1389,
      "step": 44990
    },
    {
      "epoch": 17.48,
      "learning_rate": 1.7669902912621362e-05,
      "loss": 0.0845,
      "step": 45000
    },
    {
      "epoch": 17.48,
      "learning_rate": 1.766938511326861e-05,
      "loss": 0.1614,
      "step": 45010
    },
    {
      "epoch": 17.48,
      "learning_rate": 1.7668867313915858e-05,
      "loss": 0.0573,
      "step": 45020
    },
    {
      "epoch": 17.49,
      "learning_rate": 1.766834951456311e-05,
      "loss": 0.1661,
      "step": 45030
    },
    {
      "epoch": 17.49,
      "learning_rate": 1.7667831715210357e-05,
      "loss": 0.1016,
      "step": 45040
    },
    {
      "epoch": 17.5,
      "learning_rate": 1.7667313915857608e-05,
      "loss": 0.1448,
      "step": 45050
    },
    {
      "epoch": 17.5,
      "learning_rate": 1.7666796116504856e-05,
      "loss": 0.0516,
      "step": 45060
    },
    {
      "epoch": 17.5,
      "learning_rate": 1.7666278317152107e-05,
      "loss": 0.0476,
      "step": 45070
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.7665760517799355e-05,
      "loss": 0.0762,
      "step": 45080
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.7665242718446603e-05,
      "loss": 0.2329,
      "step": 45090
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.766472491909385e-05,
      "loss": 0.1523,
      "step": 45100
    },
    {
      "epoch": 17.52,
      "learning_rate": 1.7664207119741102e-05,
      "loss": 0.1185,
      "step": 45110
    },
    {
      "epoch": 17.52,
      "learning_rate": 1.766368932038835e-05,
      "loss": 0.1227,
      "step": 45120
    },
    {
      "epoch": 17.53,
      "learning_rate": 1.7663171521035602e-05,
      "loss": 0.1114,
      "step": 45130
    },
    {
      "epoch": 17.53,
      "learning_rate": 1.766265372168285e-05,
      "loss": 0.2574,
      "step": 45140
    },
    {
      "epoch": 17.53,
      "learning_rate": 1.7662135922330098e-05,
      "loss": 0.0526,
      "step": 45150
    },
    {
      "epoch": 17.54,
      "learning_rate": 1.766161812297735e-05,
      "loss": 0.1223,
      "step": 45160
    },
    {
      "epoch": 17.54,
      "learning_rate": 1.7661100323624597e-05,
      "loss": 0.0685,
      "step": 45170
    },
    {
      "epoch": 17.55,
      "learning_rate": 1.7660582524271845e-05,
      "loss": 0.2692,
      "step": 45180
    },
    {
      "epoch": 17.55,
      "learning_rate": 1.7660064724919096e-05,
      "loss": 0.0724,
      "step": 45190
    },
    {
      "epoch": 17.55,
      "learning_rate": 1.7659546925566344e-05,
      "loss": 0.077,
      "step": 45200
    },
    {
      "epoch": 17.56,
      "learning_rate": 1.7659029126213595e-05,
      "loss": 0.1528,
      "step": 45210
    },
    {
      "epoch": 17.56,
      "learning_rate": 1.7658511326860843e-05,
      "loss": 0.0472,
      "step": 45220
    },
    {
      "epoch": 17.57,
      "learning_rate": 1.765799352750809e-05,
      "loss": 0.0096,
      "step": 45230
    },
    {
      "epoch": 17.57,
      "learning_rate": 1.7657475728155342e-05,
      "loss": 0.0132,
      "step": 45240
    },
    {
      "epoch": 17.57,
      "learning_rate": 1.765695792880259e-05,
      "loss": 0.1671,
      "step": 45250
    },
    {
      "epoch": 17.58,
      "learning_rate": 1.7656440129449838e-05,
      "loss": 0.036,
      "step": 45260
    },
    {
      "epoch": 17.58,
      "learning_rate": 1.765592233009709e-05,
      "loss": 0.2875,
      "step": 45270
    },
    {
      "epoch": 17.58,
      "learning_rate": 1.7655404530744338e-05,
      "loss": 0.1838,
      "step": 45280
    },
    {
      "epoch": 17.59,
      "learning_rate": 1.765488673139159e-05,
      "loss": 0.0885,
      "step": 45290
    },
    {
      "epoch": 17.59,
      "learning_rate": 1.7654368932038837e-05,
      "loss": 0.1035,
      "step": 45300
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.7653851132686085e-05,
      "loss": 0.2421,
      "step": 45310
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.7653333333333336e-05,
      "loss": 0.2589,
      "step": 45320
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.7652815533980584e-05,
      "loss": 0.0281,
      "step": 45330
    },
    {
      "epoch": 17.61,
      "learning_rate": 1.7652297734627832e-05,
      "loss": 0.132,
      "step": 45340
    },
    {
      "epoch": 17.61,
      "learning_rate": 1.7651779935275083e-05,
      "loss": 0.1197,
      "step": 45350
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.765126213592233e-05,
      "loss": 0.2507,
      "step": 45360
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.7650744336569582e-05,
      "loss": 0.0365,
      "step": 45370
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.765022653721683e-05,
      "loss": 0.0657,
      "step": 45380
    },
    {
      "epoch": 17.63,
      "learning_rate": 1.7649708737864078e-05,
      "loss": 0.2257,
      "step": 45390
    },
    {
      "epoch": 17.63,
      "learning_rate": 1.764919093851133e-05,
      "loss": 0.0685,
      "step": 45400
    },
    {
      "epoch": 17.63,
      "learning_rate": 1.7648673139158577e-05,
      "loss": 0.0575,
      "step": 45410
    },
    {
      "epoch": 17.64,
      "learning_rate": 1.7648155339805825e-05,
      "loss": 0.1274,
      "step": 45420
    },
    {
      "epoch": 17.64,
      "learning_rate": 1.7647637540453077e-05,
      "loss": 0.1426,
      "step": 45430
    },
    {
      "epoch": 17.65,
      "learning_rate": 1.7647119741100325e-05,
      "loss": 0.1887,
      "step": 45440
    },
    {
      "epoch": 17.65,
      "learning_rate": 1.7646601941747576e-05,
      "loss": 0.2834,
      "step": 45450
    },
    {
      "epoch": 17.65,
      "learning_rate": 1.7646084142394824e-05,
      "loss": 0.3783,
      "step": 45460
    },
    {
      "epoch": 17.66,
      "learning_rate": 1.7645566343042072e-05,
      "loss": 0.1254,
      "step": 45470
    },
    {
      "epoch": 17.66,
      "learning_rate": 1.7645048543689323e-05,
      "loss": 0.0853,
      "step": 45480
    },
    {
      "epoch": 17.67,
      "learning_rate": 1.764453074433657e-05,
      "loss": 0.1149,
      "step": 45490
    },
    {
      "epoch": 17.67,
      "learning_rate": 1.764401294498382e-05,
      "loss": 0.1192,
      "step": 45500
    },
    {
      "epoch": 17.67,
      "learning_rate": 1.764349514563107e-05,
      "loss": 0.0112,
      "step": 45510
    },
    {
      "epoch": 17.68,
      "learning_rate": 1.7642977346278318e-05,
      "loss": 0.1331,
      "step": 45520
    },
    {
      "epoch": 17.68,
      "learning_rate": 1.7642459546925566e-05,
      "loss": 0.1607,
      "step": 45530
    },
    {
      "epoch": 17.69,
      "learning_rate": 1.7641941747572817e-05,
      "loss": 0.0372,
      "step": 45540
    },
    {
      "epoch": 17.69,
      "learning_rate": 1.7641423948220065e-05,
      "loss": 0.0667,
      "step": 45550
    },
    {
      "epoch": 17.69,
      "learning_rate": 1.7640906148867317e-05,
      "loss": 0.1425,
      "step": 45560
    },
    {
      "epoch": 17.7,
      "learning_rate": 1.7640388349514565e-05,
      "loss": 0.1035,
      "step": 45570
    },
    {
      "epoch": 17.7,
      "learning_rate": 1.7639870550161812e-05,
      "loss": 0.1475,
      "step": 45580
    },
    {
      "epoch": 17.7,
      "learning_rate": 1.7639352750809064e-05,
      "loss": 0.1162,
      "step": 45590
    },
    {
      "epoch": 17.71,
      "learning_rate": 1.7638834951456312e-05,
      "loss": 0.3411,
      "step": 45600
    },
    {
      "epoch": 17.71,
      "learning_rate": 1.763831715210356e-05,
      "loss": 0.2146,
      "step": 45610
    },
    {
      "epoch": 17.72,
      "learning_rate": 1.763779935275081e-05,
      "loss": 0.1024,
      "step": 45620
    },
    {
      "epoch": 17.72,
      "learning_rate": 1.763728155339806e-05,
      "loss": 0.1062,
      "step": 45630
    },
    {
      "epoch": 17.72,
      "learning_rate": 1.763676375404531e-05,
      "loss": 0.1574,
      "step": 45640
    },
    {
      "epoch": 17.73,
      "learning_rate": 1.7636245954692558e-05,
      "loss": 0.0468,
      "step": 45650
    },
    {
      "epoch": 17.73,
      "learning_rate": 1.7635728155339806e-05,
      "loss": 0.065,
      "step": 45660
    },
    {
      "epoch": 17.74,
      "learning_rate": 1.7635210355987057e-05,
      "loss": 0.0823,
      "step": 45670
    },
    {
      "epoch": 17.74,
      "learning_rate": 1.7634692556634305e-05,
      "loss": 0.0243,
      "step": 45680
    },
    {
      "epoch": 17.74,
      "learning_rate": 1.7634174757281553e-05,
      "loss": 0.1374,
      "step": 45690
    },
    {
      "epoch": 17.75,
      "learning_rate": 1.7633656957928805e-05,
      "loss": 0.1334,
      "step": 45700
    },
    {
      "epoch": 17.75,
      "learning_rate": 1.7633139158576052e-05,
      "loss": 0.0994,
      "step": 45710
    },
    {
      "epoch": 17.76,
      "learning_rate": 1.7632621359223304e-05,
      "loss": 0.2081,
      "step": 45720
    },
    {
      "epoch": 17.76,
      "learning_rate": 1.763210355987055e-05,
      "loss": 0.1685,
      "step": 45730
    },
    {
      "epoch": 17.76,
      "learning_rate": 1.76315857605178e-05,
      "loss": 0.1136,
      "step": 45740
    },
    {
      "epoch": 17.77,
      "learning_rate": 1.763106796116505e-05,
      "loss": 0.1029,
      "step": 45750
    },
    {
      "epoch": 17.77,
      "learning_rate": 1.76305501618123e-05,
      "loss": 0.1598,
      "step": 45760
    },
    {
      "epoch": 17.77,
      "learning_rate": 1.7630032362459547e-05,
      "loss": 0.0964,
      "step": 45770
    },
    {
      "epoch": 17.78,
      "learning_rate": 1.7629514563106798e-05,
      "loss": 0.0707,
      "step": 45780
    },
    {
      "epoch": 17.78,
      "learning_rate": 1.7628996763754046e-05,
      "loss": 0.0628,
      "step": 45790
    },
    {
      "epoch": 17.79,
      "learning_rate": 1.7628478964401297e-05,
      "loss": 0.1575,
      "step": 45800
    },
    {
      "epoch": 17.79,
      "learning_rate": 1.7627961165048545e-05,
      "loss": 0.1441,
      "step": 45810
    },
    {
      "epoch": 17.79,
      "learning_rate": 1.7627443365695793e-05,
      "loss": 0.0448,
      "step": 45820
    },
    {
      "epoch": 17.8,
      "learning_rate": 1.762692556634304e-05,
      "loss": 0.1605,
      "step": 45830
    },
    {
      "epoch": 17.8,
      "learning_rate": 1.7626407766990292e-05,
      "loss": 0.0109,
      "step": 45840
    },
    {
      "epoch": 17.81,
      "learning_rate": 1.762588996763754e-05,
      "loss": 0.0072,
      "step": 45850
    },
    {
      "epoch": 17.81,
      "learning_rate": 1.762537216828479e-05,
      "loss": 0.114,
      "step": 45860
    },
    {
      "epoch": 17.81,
      "learning_rate": 1.762485436893204e-05,
      "loss": 0.0905,
      "step": 45870
    },
    {
      "epoch": 17.82,
      "learning_rate": 1.762433656957929e-05,
      "loss": 0.1212,
      "step": 45880
    },
    {
      "epoch": 17.82,
      "learning_rate": 1.762381877022654e-05,
      "loss": 0.1181,
      "step": 45890
    },
    {
      "epoch": 17.83,
      "learning_rate": 1.7623300970873787e-05,
      "loss": 0.169,
      "step": 45900
    },
    {
      "epoch": 17.83,
      "learning_rate": 1.7622783171521035e-05,
      "loss": 0.0953,
      "step": 45910
    },
    {
      "epoch": 17.83,
      "learning_rate": 1.7622265372168286e-05,
      "loss": 0.0757,
      "step": 45920
    },
    {
      "epoch": 17.84,
      "learning_rate": 1.7621747572815534e-05,
      "loss": 0.0775,
      "step": 45930
    },
    {
      "epoch": 17.84,
      "learning_rate": 1.7621229773462785e-05,
      "loss": 0.1722,
      "step": 45940
    },
    {
      "epoch": 17.84,
      "learning_rate": 1.7620711974110033e-05,
      "loss": 0.0659,
      "step": 45950
    },
    {
      "epoch": 17.85,
      "learning_rate": 1.7620194174757284e-05,
      "loss": 0.1339,
      "step": 45960
    },
    {
      "epoch": 17.85,
      "learning_rate": 1.7619676375404532e-05,
      "loss": 0.155,
      "step": 45970
    },
    {
      "epoch": 17.86,
      "learning_rate": 1.7619158576051784e-05,
      "loss": 0.2083,
      "step": 45980
    },
    {
      "epoch": 17.86,
      "learning_rate": 1.7618640776699028e-05,
      "loss": 0.2501,
      "step": 45990
    },
    {
      "epoch": 17.86,
      "learning_rate": 1.761812297734628e-05,
      "loss": 0.21,
      "step": 46000
    },
    {
      "epoch": 17.87,
      "learning_rate": 1.7617605177993527e-05,
      "loss": 0.1334,
      "step": 46010
    },
    {
      "epoch": 17.87,
      "learning_rate": 1.761708737864078e-05,
      "loss": 0.2606,
      "step": 46020
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.7616569579288027e-05,
      "loss": 0.1457,
      "step": 46030
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.7616051779935278e-05,
      "loss": 0.1713,
      "step": 46040
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.7615533980582526e-05,
      "loss": 0.031,
      "step": 46050
    },
    {
      "epoch": 17.89,
      "learning_rate": 1.7615016181229777e-05,
      "loss": 0.137,
      "step": 46060
    },
    {
      "epoch": 17.89,
      "learning_rate": 1.7614498381877022e-05,
      "loss": 0.2195,
      "step": 46070
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.7613980582524273e-05,
      "loss": 0.1182,
      "step": 46080
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.761346278317152e-05,
      "loss": 0.0461,
      "step": 46090
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.7612944983818772e-05,
      "loss": 0.1202,
      "step": 46100
    },
    {
      "epoch": 17.91,
      "learning_rate": 1.761242718446602e-05,
      "loss": 0.1419,
      "step": 46110
    },
    {
      "epoch": 17.91,
      "learning_rate": 1.761190938511327e-05,
      "loss": 0.2035,
      "step": 46120
    },
    {
      "epoch": 17.91,
      "learning_rate": 1.761139158576052e-05,
      "loss": 0.0719,
      "step": 46130
    },
    {
      "epoch": 17.92,
      "learning_rate": 1.761087378640777e-05,
      "loss": 0.1366,
      "step": 46140
    },
    {
      "epoch": 17.92,
      "learning_rate": 1.7610355987055015e-05,
      "loss": 0.2476,
      "step": 46150
    },
    {
      "epoch": 17.93,
      "learning_rate": 1.7609838187702267e-05,
      "loss": 0.1552,
      "step": 46160
    },
    {
      "epoch": 17.93,
      "learning_rate": 1.7609320388349515e-05,
      "loss": 0.0028,
      "step": 46170
    },
    {
      "epoch": 17.93,
      "learning_rate": 1.7608802588996766e-05,
      "loss": 0.1076,
      "step": 46180
    },
    {
      "epoch": 17.94,
      "learning_rate": 1.7608284789644014e-05,
      "loss": 0.1243,
      "step": 46190
    },
    {
      "epoch": 17.94,
      "learning_rate": 1.7607766990291265e-05,
      "loss": 0.1251,
      "step": 46200
    },
    {
      "epoch": 17.95,
      "learning_rate": 1.7607249190938513e-05,
      "loss": 0.0739,
      "step": 46210
    },
    {
      "epoch": 17.95,
      "learning_rate": 1.7606731391585764e-05,
      "loss": 0.1059,
      "step": 46220
    },
    {
      "epoch": 17.95,
      "learning_rate": 1.760621359223301e-05,
      "loss": 0.1499,
      "step": 46230
    },
    {
      "epoch": 17.96,
      "learning_rate": 1.760569579288026e-05,
      "loss": 0.0061,
      "step": 46240
    },
    {
      "epoch": 17.96,
      "learning_rate": 1.7605177993527508e-05,
      "loss": 0.0413,
      "step": 46250
    },
    {
      "epoch": 17.97,
      "learning_rate": 1.760466019417476e-05,
      "loss": 0.1038,
      "step": 46260
    },
    {
      "epoch": 17.97,
      "learning_rate": 1.7604142394822007e-05,
      "loss": 0.1615,
      "step": 46270
    },
    {
      "epoch": 17.97,
      "learning_rate": 1.760362459546926e-05,
      "loss": 0.1171,
      "step": 46280
    },
    {
      "epoch": 17.98,
      "learning_rate": 1.7603106796116507e-05,
      "loss": 0.0411,
      "step": 46290
    },
    {
      "epoch": 17.98,
      "learning_rate": 1.7602588996763758e-05,
      "loss": 0.0988,
      "step": 46300
    },
    {
      "epoch": 17.98,
      "learning_rate": 1.7602071197411002e-05,
      "loss": 0.2157,
      "step": 46310
    },
    {
      "epoch": 17.99,
      "learning_rate": 1.7601553398058254e-05,
      "loss": 0.0839,
      "step": 46320
    },
    {
      "epoch": 17.99,
      "learning_rate": 1.76010355987055e-05,
      "loss": 0.1509,
      "step": 46330
    },
    {
      "epoch": 18.0,
      "learning_rate": 1.7600517799352753e-05,
      "loss": 0.1595,
      "step": 46340
    },
    {
      "epoch": 18.0,
      "learning_rate": 1.76e-05,
      "loss": 0.0463,
      "step": 46350
    },
    {
      "epoch": 18.0,
      "eval_accuracy": 0.946079779917469,
      "eval_loss": 0.28520387411117554,
      "eval_runtime": 8.2227,
      "eval_samples_per_second": 442.068,
      "eval_steps_per_second": 55.335,
      "step": 46350
    },
    {
      "epoch": 18.0,
      "learning_rate": 1.7599482200647252e-05,
      "loss": 0.1042,
      "step": 46360
    },
    {
      "epoch": 18.01,
      "learning_rate": 1.75989644012945e-05,
      "loss": 0.1458,
      "step": 46370
    },
    {
      "epoch": 18.01,
      "learning_rate": 1.759844660194175e-05,
      "loss": 0.0071,
      "step": 46380
    },
    {
      "epoch": 18.02,
      "learning_rate": 1.7597928802588996e-05,
      "loss": 0.1704,
      "step": 46390
    },
    {
      "epoch": 18.02,
      "learning_rate": 1.7597411003236247e-05,
      "loss": 0.14,
      "step": 46400
    },
    {
      "epoch": 18.02,
      "learning_rate": 1.7596893203883495e-05,
      "loss": 0.2009,
      "step": 46410
    },
    {
      "epoch": 18.03,
      "learning_rate": 1.7596375404530747e-05,
      "loss": 0.1583,
      "step": 46420
    },
    {
      "epoch": 18.03,
      "learning_rate": 1.7595857605177994e-05,
      "loss": 0.1123,
      "step": 46430
    },
    {
      "epoch": 18.03,
      "learning_rate": 1.7595339805825246e-05,
      "loss": 0.0533,
      "step": 46440
    },
    {
      "epoch": 18.04,
      "learning_rate": 1.7594822006472494e-05,
      "loss": 0.1882,
      "step": 46450
    },
    {
      "epoch": 18.04,
      "learning_rate": 1.7594304207119745e-05,
      "loss": 0.2193,
      "step": 46460
    },
    {
      "epoch": 18.05,
      "learning_rate": 1.759378640776699e-05,
      "loss": 0.0626,
      "step": 46470
    },
    {
      "epoch": 18.05,
      "learning_rate": 1.759326860841424e-05,
      "loss": 0.0767,
      "step": 46480
    },
    {
      "epoch": 18.05,
      "learning_rate": 1.759275080906149e-05,
      "loss": 0.1081,
      "step": 46490
    },
    {
      "epoch": 18.06,
      "learning_rate": 1.759223300970874e-05,
      "loss": 0.3299,
      "step": 46500
    },
    {
      "epoch": 18.06,
      "learning_rate": 1.7591715210355988e-05,
      "loss": 0.2457,
      "step": 46510
    },
    {
      "epoch": 18.07,
      "learning_rate": 1.759119741100324e-05,
      "loss": 0.1593,
      "step": 46520
    },
    {
      "epoch": 18.07,
      "learning_rate": 1.7590679611650487e-05,
      "loss": 0.1365,
      "step": 46530
    },
    {
      "epoch": 18.07,
      "learning_rate": 1.759016181229774e-05,
      "loss": 0.0995,
      "step": 46540
    },
    {
      "epoch": 18.08,
      "learning_rate": 1.7589644012944986e-05,
      "loss": 0.2021,
      "step": 46550
    },
    {
      "epoch": 18.08,
      "learning_rate": 1.7589126213592234e-05,
      "loss": 0.1281,
      "step": 46560
    },
    {
      "epoch": 18.09,
      "learning_rate": 1.7588608414239482e-05,
      "loss": 0.0928,
      "step": 46570
    },
    {
      "epoch": 18.09,
      "learning_rate": 1.7588090614886734e-05,
      "loss": 0.1498,
      "step": 46580
    },
    {
      "epoch": 18.09,
      "learning_rate": 1.758757281553398e-05,
      "loss": 0.0905,
      "step": 46590
    },
    {
      "epoch": 18.1,
      "learning_rate": 1.7587055016181233e-05,
      "loss": 0.0382,
      "step": 46600
    },
    {
      "epoch": 18.1,
      "learning_rate": 1.758653721682848e-05,
      "loss": 0.1239,
      "step": 46610
    },
    {
      "epoch": 18.1,
      "learning_rate": 1.758601941747573e-05,
      "loss": 0.1121,
      "step": 46620
    },
    {
      "epoch": 18.11,
      "learning_rate": 1.758550161812298e-05,
      "loss": 0.1685,
      "step": 46630
    },
    {
      "epoch": 18.11,
      "learning_rate": 1.7584983818770228e-05,
      "loss": 0.1674,
      "step": 46640
    },
    {
      "epoch": 18.12,
      "learning_rate": 1.7584466019417476e-05,
      "loss": 0.0595,
      "step": 46650
    },
    {
      "epoch": 18.12,
      "learning_rate": 1.7583948220064727e-05,
      "loss": 0.0589,
      "step": 46660
    },
    {
      "epoch": 18.12,
      "learning_rate": 1.7583430420711975e-05,
      "loss": 0.1053,
      "step": 46670
    },
    {
      "epoch": 18.13,
      "learning_rate": 1.7582912621359226e-05,
      "loss": 0.0976,
      "step": 46680
    },
    {
      "epoch": 18.13,
      "learning_rate": 1.7582394822006474e-05,
      "loss": 0.0779,
      "step": 46690
    },
    {
      "epoch": 18.14,
      "learning_rate": 1.7581877022653722e-05,
      "loss": 0.0884,
      "step": 46700
    },
    {
      "epoch": 18.14,
      "learning_rate": 1.7581359223300974e-05,
      "loss": 0.059,
      "step": 46710
    },
    {
      "epoch": 18.14,
      "learning_rate": 1.758084142394822e-05,
      "loss": 0.1046,
      "step": 46720
    },
    {
      "epoch": 18.15,
      "learning_rate": 1.758032362459547e-05,
      "loss": 0.0854,
      "step": 46730
    },
    {
      "epoch": 18.15,
      "learning_rate": 1.757980582524272e-05,
      "loss": 0.1995,
      "step": 46740
    },
    {
      "epoch": 18.16,
      "learning_rate": 1.757928802588997e-05,
      "loss": 0.2225,
      "step": 46750
    },
    {
      "epoch": 18.16,
      "learning_rate": 1.757877022653722e-05,
      "loss": 0.0064,
      "step": 46760
    },
    {
      "epoch": 18.16,
      "learning_rate": 1.7578252427184468e-05,
      "loss": 0.1476,
      "step": 46770
    },
    {
      "epoch": 18.17,
      "learning_rate": 1.7577734627831716e-05,
      "loss": 0.2512,
      "step": 46780
    },
    {
      "epoch": 18.17,
      "learning_rate": 1.7577216828478967e-05,
      "loss": 0.1123,
      "step": 46790
    },
    {
      "epoch": 18.17,
      "learning_rate": 1.7576699029126215e-05,
      "loss": 0.0425,
      "step": 46800
    },
    {
      "epoch": 18.18,
      "learning_rate": 1.7576181229773463e-05,
      "loss": 0.0631,
      "step": 46810
    },
    {
      "epoch": 18.18,
      "learning_rate": 1.7575663430420714e-05,
      "loss": 0.2232,
      "step": 46820
    },
    {
      "epoch": 18.19,
      "learning_rate": 1.7575145631067962e-05,
      "loss": 0.1489,
      "step": 46830
    },
    {
      "epoch": 18.19,
      "learning_rate": 1.7574627831715214e-05,
      "loss": 0.0832,
      "step": 46840
    },
    {
      "epoch": 18.19,
      "learning_rate": 1.757411003236246e-05,
      "loss": 0.0781,
      "step": 46850
    },
    {
      "epoch": 18.2,
      "learning_rate": 1.757359223300971e-05,
      "loss": 0.1738,
      "step": 46860
    },
    {
      "epoch": 18.2,
      "learning_rate": 1.757307443365696e-05,
      "loss": 0.0749,
      "step": 46870
    },
    {
      "epoch": 18.21,
      "learning_rate": 1.757255663430421e-05,
      "loss": 0.1357,
      "step": 46880
    },
    {
      "epoch": 18.21,
      "learning_rate": 1.7572038834951457e-05,
      "loss": 0.2223,
      "step": 46890
    },
    {
      "epoch": 18.21,
      "learning_rate": 1.7571521035598708e-05,
      "loss": 0.1789,
      "step": 46900
    },
    {
      "epoch": 18.22,
      "learning_rate": 1.7571003236245956e-05,
      "loss": 0.0605,
      "step": 46910
    },
    {
      "epoch": 18.22,
      "learning_rate": 1.7570485436893204e-05,
      "loss": 0.0195,
      "step": 46920
    },
    {
      "epoch": 18.23,
      "learning_rate": 1.7569967637540455e-05,
      "loss": 0.0836,
      "step": 46930
    },
    {
      "epoch": 18.23,
      "learning_rate": 1.7569449838187703e-05,
      "loss": 0.1713,
      "step": 46940
    },
    {
      "epoch": 18.23,
      "learning_rate": 1.7568932038834954e-05,
      "loss": 0.1184,
      "step": 46950
    },
    {
      "epoch": 18.24,
      "learning_rate": 1.7568414239482202e-05,
      "loss": 0.1037,
      "step": 46960
    },
    {
      "epoch": 18.24,
      "learning_rate": 1.756789644012945e-05,
      "loss": 0.1028,
      "step": 46970
    },
    {
      "epoch": 18.24,
      "learning_rate": 1.75673786407767e-05,
      "loss": 0.0194,
      "step": 46980
    },
    {
      "epoch": 18.25,
      "learning_rate": 1.756686084142395e-05,
      "loss": 0.1779,
      "step": 46990
    },
    {
      "epoch": 18.25,
      "learning_rate": 1.7566343042071197e-05,
      "loss": 0.0167,
      "step": 47000
    },
    {
      "epoch": 18.26,
      "learning_rate": 1.756582524271845e-05,
      "loss": 0.0474,
      "step": 47010
    },
    {
      "epoch": 18.26,
      "learning_rate": 1.7565307443365697e-05,
      "loss": 0.0489,
      "step": 47020
    },
    {
      "epoch": 18.26,
      "learning_rate": 1.7564789644012948e-05,
      "loss": 0.0997,
      "step": 47030
    },
    {
      "epoch": 18.27,
      "learning_rate": 1.7564271844660196e-05,
      "loss": 0.1489,
      "step": 47040
    },
    {
      "epoch": 18.27,
      "learning_rate": 1.7563754045307444e-05,
      "loss": 0.0851,
      "step": 47050
    },
    {
      "epoch": 18.28,
      "learning_rate": 1.7563236245954695e-05,
      "loss": 0.1577,
      "step": 47060
    },
    {
      "epoch": 18.28,
      "learning_rate": 1.7562718446601943e-05,
      "loss": 0.0022,
      "step": 47070
    },
    {
      "epoch": 18.28,
      "learning_rate": 1.756220064724919e-05,
      "loss": 0.193,
      "step": 47080
    },
    {
      "epoch": 18.29,
      "learning_rate": 1.7561682847896442e-05,
      "loss": 0.2013,
      "step": 47090
    },
    {
      "epoch": 18.29,
      "learning_rate": 1.756116504854369e-05,
      "loss": 0.123,
      "step": 47100
    },
    {
      "epoch": 18.3,
      "learning_rate": 1.756064724919094e-05,
      "loss": 0.1383,
      "step": 47110
    },
    {
      "epoch": 18.3,
      "learning_rate": 1.756012944983819e-05,
      "loss": 0.0664,
      "step": 47120
    },
    {
      "epoch": 18.3,
      "learning_rate": 1.7559611650485437e-05,
      "loss": 0.0822,
      "step": 47130
    },
    {
      "epoch": 18.31,
      "learning_rate": 1.755909385113269e-05,
      "loss": 0.0078,
      "step": 47140
    },
    {
      "epoch": 18.31,
      "learning_rate": 1.7558576051779936e-05,
      "loss": 0.1105,
      "step": 47150
    },
    {
      "epoch": 18.31,
      "learning_rate": 1.7558058252427184e-05,
      "loss": 0.1319,
      "step": 47160
    },
    {
      "epoch": 18.32,
      "learning_rate": 1.7557540453074436e-05,
      "loss": 0.1639,
      "step": 47170
    },
    {
      "epoch": 18.32,
      "learning_rate": 1.7557022653721684e-05,
      "loss": 0.1652,
      "step": 47180
    },
    {
      "epoch": 18.33,
      "learning_rate": 1.7556504854368935e-05,
      "loss": 0.2102,
      "step": 47190
    },
    {
      "epoch": 18.33,
      "learning_rate": 1.7555987055016183e-05,
      "loss": 0.0331,
      "step": 47200
    },
    {
      "epoch": 18.33,
      "learning_rate": 1.755546925566343e-05,
      "loss": 0.1704,
      "step": 47210
    },
    {
      "epoch": 18.34,
      "learning_rate": 1.7554951456310682e-05,
      "loss": 0.182,
      "step": 47220
    },
    {
      "epoch": 18.34,
      "learning_rate": 1.755443365695793e-05,
      "loss": 0.0544,
      "step": 47230
    },
    {
      "epoch": 18.35,
      "learning_rate": 1.7553915857605178e-05,
      "loss": 0.0668,
      "step": 47240
    },
    {
      "epoch": 18.35,
      "learning_rate": 1.755339805825243e-05,
      "loss": 0.2502,
      "step": 47250
    },
    {
      "epoch": 18.35,
      "learning_rate": 1.7552880258899677e-05,
      "loss": 0.1947,
      "step": 47260
    },
    {
      "epoch": 18.36,
      "learning_rate": 1.755236245954693e-05,
      "loss": 0.0603,
      "step": 47270
    },
    {
      "epoch": 18.36,
      "learning_rate": 1.7551844660194176e-05,
      "loss": 0.1317,
      "step": 47280
    },
    {
      "epoch": 18.37,
      "learning_rate": 1.7551326860841424e-05,
      "loss": 0.0755,
      "step": 47290
    },
    {
      "epoch": 18.37,
      "learning_rate": 1.7550809061488672e-05,
      "loss": 0.179,
      "step": 47300
    },
    {
      "epoch": 18.37,
      "learning_rate": 1.7550291262135924e-05,
      "loss": 0.1046,
      "step": 47310
    },
    {
      "epoch": 18.38,
      "learning_rate": 1.754977346278317e-05,
      "loss": 0.1702,
      "step": 47320
    },
    {
      "epoch": 18.38,
      "learning_rate": 1.7549255663430423e-05,
      "loss": 0.2192,
      "step": 47330
    },
    {
      "epoch": 18.38,
      "learning_rate": 1.754873786407767e-05,
      "loss": 0.0659,
      "step": 47340
    },
    {
      "epoch": 18.39,
      "learning_rate": 1.7548220064724922e-05,
      "loss": 0.1499,
      "step": 47350
    },
    {
      "epoch": 18.39,
      "learning_rate": 1.754770226537217e-05,
      "loss": 0.0945,
      "step": 47360
    },
    {
      "epoch": 18.4,
      "learning_rate": 1.7547184466019418e-05,
      "loss": 0.0475,
      "step": 47370
    },
    {
      "epoch": 18.4,
      "learning_rate": 1.7546666666666666e-05,
      "loss": 0.1128,
      "step": 47380
    },
    {
      "epoch": 18.4,
      "learning_rate": 1.7546148867313917e-05,
      "loss": 0.1344,
      "step": 47390
    },
    {
      "epoch": 18.41,
      "learning_rate": 1.7545631067961165e-05,
      "loss": 0.156,
      "step": 47400
    },
    {
      "epoch": 18.41,
      "learning_rate": 1.7545113268608416e-05,
      "loss": 0.1078,
      "step": 47410
    },
    {
      "epoch": 18.42,
      "learning_rate": 1.7544595469255664e-05,
      "loss": 0.1085,
      "step": 47420
    },
    {
      "epoch": 18.42,
      "learning_rate": 1.7544077669902916e-05,
      "loss": 0.1383,
      "step": 47430
    },
    {
      "epoch": 18.42,
      "learning_rate": 1.7543559870550164e-05,
      "loss": 0.0491,
      "step": 47440
    },
    {
      "epoch": 18.43,
      "learning_rate": 1.754304207119741e-05,
      "loss": 0.2082,
      "step": 47450
    },
    {
      "epoch": 18.43,
      "learning_rate": 1.754252427184466e-05,
      "loss": 0.0286,
      "step": 47460
    },
    {
      "epoch": 18.43,
      "learning_rate": 1.754200647249191e-05,
      "loss": 0.1292,
      "step": 47470
    },
    {
      "epoch": 18.44,
      "learning_rate": 1.754148867313916e-05,
      "loss": 0.1905,
      "step": 47480
    },
    {
      "epoch": 18.44,
      "learning_rate": 1.754097087378641e-05,
      "loss": 0.0992,
      "step": 47490
    },
    {
      "epoch": 18.45,
      "learning_rate": 1.7540453074433658e-05,
      "loss": 0.02,
      "step": 47500
    },
    {
      "epoch": 18.45,
      "learning_rate": 1.753993527508091e-05,
      "loss": 0.0727,
      "step": 47510
    },
    {
      "epoch": 18.45,
      "learning_rate": 1.7539417475728157e-05,
      "loss": 0.1293,
      "step": 47520
    },
    {
      "epoch": 18.46,
      "learning_rate": 1.7538899676375405e-05,
      "loss": 0.3012,
      "step": 47530
    },
    {
      "epoch": 18.46,
      "learning_rate": 1.7538381877022653e-05,
      "loss": 0.1126,
      "step": 47540
    },
    {
      "epoch": 18.47,
      "learning_rate": 1.7537864077669904e-05,
      "loss": 0.2203,
      "step": 47550
    },
    {
      "epoch": 18.47,
      "learning_rate": 1.7537346278317152e-05,
      "loss": 0.1524,
      "step": 47560
    },
    {
      "epoch": 18.47,
      "learning_rate": 1.7536828478964403e-05,
      "loss": 0.0863,
      "step": 47570
    },
    {
      "epoch": 18.48,
      "learning_rate": 1.753631067961165e-05,
      "loss": 0.062,
      "step": 47580
    },
    {
      "epoch": 18.48,
      "learning_rate": 1.7535792880258903e-05,
      "loss": 0.0528,
      "step": 47590
    },
    {
      "epoch": 18.49,
      "learning_rate": 1.753527508090615e-05,
      "loss": 0.1061,
      "step": 47600
    },
    {
      "epoch": 18.49,
      "learning_rate": 1.75347572815534e-05,
      "loss": 0.3054,
      "step": 47610
    },
    {
      "epoch": 18.49,
      "learning_rate": 1.7534239482200646e-05,
      "loss": 0.096,
      "step": 47620
    },
    {
      "epoch": 18.5,
      "learning_rate": 1.7533721682847898e-05,
      "loss": 0.188,
      "step": 47630
    },
    {
      "epoch": 18.5,
      "learning_rate": 1.7533203883495146e-05,
      "loss": 0.142,
      "step": 47640
    },
    {
      "epoch": 18.5,
      "learning_rate": 1.7532686084142397e-05,
      "loss": 0.0408,
      "step": 47650
    },
    {
      "epoch": 18.51,
      "learning_rate": 1.7532168284789645e-05,
      "loss": 0.1064,
      "step": 47660
    },
    {
      "epoch": 18.51,
      "learning_rate": 1.7531650485436896e-05,
      "loss": 0.1284,
      "step": 47670
    },
    {
      "epoch": 18.52,
      "learning_rate": 1.7531132686084144e-05,
      "loss": 0.093,
      "step": 47680
    },
    {
      "epoch": 18.52,
      "learning_rate": 1.7530614886731396e-05,
      "loss": 0.2373,
      "step": 47690
    },
    {
      "epoch": 18.52,
      "learning_rate": 1.753009708737864e-05,
      "loss": 0.1456,
      "step": 47700
    },
    {
      "epoch": 18.53,
      "learning_rate": 1.752957928802589e-05,
      "loss": 0.2189,
      "step": 47710
    },
    {
      "epoch": 18.53,
      "learning_rate": 1.752906148867314e-05,
      "loss": 0.1447,
      "step": 47720
    },
    {
      "epoch": 18.54,
      "learning_rate": 1.752854368932039e-05,
      "loss": 0.2176,
      "step": 47730
    },
    {
      "epoch": 18.54,
      "learning_rate": 1.752802588996764e-05,
      "loss": 0.1511,
      "step": 47740
    },
    {
      "epoch": 18.54,
      "learning_rate": 1.752750809061489e-05,
      "loss": 0.1429,
      "step": 47750
    },
    {
      "epoch": 18.55,
      "learning_rate": 1.7526990291262138e-05,
      "loss": 0.1704,
      "step": 47760
    },
    {
      "epoch": 18.55,
      "learning_rate": 1.752647249190939e-05,
      "loss": 0.2922,
      "step": 47770
    },
    {
      "epoch": 18.56,
      "learning_rate": 1.7525954692556634e-05,
      "loss": 0.0829,
      "step": 47780
    },
    {
      "epoch": 18.56,
      "learning_rate": 1.7525436893203885e-05,
      "loss": 0.084,
      "step": 47790
    },
    {
      "epoch": 18.56,
      "learning_rate": 1.7524919093851133e-05,
      "loss": 0.0331,
      "step": 47800
    },
    {
      "epoch": 18.57,
      "learning_rate": 1.7524401294498384e-05,
      "loss": 0.0745,
      "step": 47810
    },
    {
      "epoch": 18.57,
      "learning_rate": 1.7523883495145632e-05,
      "loss": 0.1303,
      "step": 47820
    },
    {
      "epoch": 18.57,
      "learning_rate": 1.7523365695792883e-05,
      "loss": 0.1263,
      "step": 47830
    },
    {
      "epoch": 18.58,
      "learning_rate": 1.752284789644013e-05,
      "loss": 0.0674,
      "step": 47840
    },
    {
      "epoch": 18.58,
      "learning_rate": 1.7522330097087383e-05,
      "loss": 0.0788,
      "step": 47850
    },
    {
      "epoch": 18.59,
      "learning_rate": 1.7521812297734627e-05,
      "loss": 0.0803,
      "step": 47860
    },
    {
      "epoch": 18.59,
      "learning_rate": 1.752129449838188e-05,
      "loss": 0.1326,
      "step": 47870
    },
    {
      "epoch": 18.59,
      "learning_rate": 1.7520776699029126e-05,
      "loss": 0.0636,
      "step": 47880
    },
    {
      "epoch": 18.6,
      "learning_rate": 1.7520258899676378e-05,
      "loss": 0.1998,
      "step": 47890
    },
    {
      "epoch": 18.6,
      "learning_rate": 1.7519741100323626e-05,
      "loss": 0.0272,
      "step": 47900
    },
    {
      "epoch": 18.61,
      "learning_rate": 1.7519223300970877e-05,
      "loss": 0.1409,
      "step": 47910
    },
    {
      "epoch": 18.61,
      "learning_rate": 1.7518705501618125e-05,
      "loss": 0.1324,
      "step": 47920
    },
    {
      "epoch": 18.61,
      "learning_rate": 1.7518187702265376e-05,
      "loss": 0.2306,
      "step": 47930
    },
    {
      "epoch": 18.62,
      "learning_rate": 1.751766990291262e-05,
      "loss": 0.0745,
      "step": 47940
    },
    {
      "epoch": 18.62,
      "learning_rate": 1.7517152103559872e-05,
      "loss": 0.2658,
      "step": 47950
    },
    {
      "epoch": 18.63,
      "learning_rate": 1.751663430420712e-05,
      "loss": 0.0583,
      "step": 47960
    },
    {
      "epoch": 18.63,
      "learning_rate": 1.751611650485437e-05,
      "loss": 0.0448,
      "step": 47970
    },
    {
      "epoch": 18.63,
      "learning_rate": 1.751559870550162e-05,
      "loss": 0.1412,
      "step": 47980
    },
    {
      "epoch": 18.64,
      "learning_rate": 1.751508090614887e-05,
      "loss": 0.1098,
      "step": 47990
    },
    {
      "epoch": 18.64,
      "learning_rate": 1.751456310679612e-05,
      "loss": 0.1051,
      "step": 48000
    },
    {
      "epoch": 18.64,
      "learning_rate": 1.751404530744337e-05,
      "loss": 0.1473,
      "step": 48010
    },
    {
      "epoch": 18.65,
      "learning_rate": 1.7513527508090614e-05,
      "loss": 0.2206,
      "step": 48020
    },
    {
      "epoch": 18.65,
      "learning_rate": 1.7513009708737866e-05,
      "loss": 0.1427,
      "step": 48030
    },
    {
      "epoch": 18.66,
      "learning_rate": 1.7512491909385114e-05,
      "loss": 0.3028,
      "step": 48040
    },
    {
      "epoch": 18.66,
      "learning_rate": 1.7511974110032365e-05,
      "loss": 0.2534,
      "step": 48050
    },
    {
      "epoch": 18.66,
      "learning_rate": 1.7511456310679613e-05,
      "loss": 0.1399,
      "step": 48060
    },
    {
      "epoch": 18.67,
      "learning_rate": 1.7510938511326864e-05,
      "loss": 0.0056,
      "step": 48070
    },
    {
      "epoch": 18.67,
      "learning_rate": 1.7510420711974112e-05,
      "loss": 0.3489,
      "step": 48080
    },
    {
      "epoch": 18.68,
      "learning_rate": 1.750990291262136e-05,
      "loss": 0.0459,
      "step": 48090
    },
    {
      "epoch": 18.68,
      "learning_rate": 1.7509385113268608e-05,
      "loss": 0.0334,
      "step": 48100
    },
    {
      "epoch": 18.68,
      "learning_rate": 1.750886731391586e-05,
      "loss": 0.2026,
      "step": 48110
    },
    {
      "epoch": 18.69,
      "learning_rate": 1.7508349514563107e-05,
      "loss": 0.0388,
      "step": 48120
    },
    {
      "epoch": 18.69,
      "learning_rate": 1.750783171521036e-05,
      "loss": 0.1521,
      "step": 48130
    },
    {
      "epoch": 18.7,
      "learning_rate": 1.7507313915857606e-05,
      "loss": 0.0793,
      "step": 48140
    },
    {
      "epoch": 18.7,
      "learning_rate": 1.7506796116504858e-05,
      "loss": 0.0712,
      "step": 48150
    },
    {
      "epoch": 18.7,
      "learning_rate": 1.7506278317152106e-05,
      "loss": 0.15,
      "step": 48160
    },
    {
      "epoch": 18.71,
      "learning_rate": 1.7505760517799353e-05,
      "loss": 0.1408,
      "step": 48170
    },
    {
      "epoch": 18.71,
      "learning_rate": 1.75052427184466e-05,
      "loss": 0.2519,
      "step": 48180
    },
    {
      "epoch": 18.71,
      "learning_rate": 1.7504724919093853e-05,
      "loss": 0.1139,
      "step": 48190
    },
    {
      "epoch": 18.72,
      "learning_rate": 1.75042071197411e-05,
      "loss": 0.1582,
      "step": 48200
    },
    {
      "epoch": 18.72,
      "learning_rate": 1.7503689320388352e-05,
      "loss": 0.3006,
      "step": 48210
    },
    {
      "epoch": 18.73,
      "learning_rate": 1.75031715210356e-05,
      "loss": 0.1292,
      "step": 48220
    },
    {
      "epoch": 18.73,
      "learning_rate": 1.750265372168285e-05,
      "loss": 0.1404,
      "step": 48230
    },
    {
      "epoch": 18.73,
      "learning_rate": 1.75021359223301e-05,
      "loss": 0.0875,
      "step": 48240
    },
    {
      "epoch": 18.74,
      "learning_rate": 1.7501618122977347e-05,
      "loss": 0.0996,
      "step": 48250
    },
    {
      "epoch": 18.74,
      "learning_rate": 1.75011003236246e-05,
      "loss": 0.0956,
      "step": 48260
    },
    {
      "epoch": 18.75,
      "learning_rate": 1.7500582524271846e-05,
      "loss": 0.2705,
      "step": 48270
    },
    {
      "epoch": 18.75,
      "learning_rate": 1.7500064724919094e-05,
      "loss": 0.1118,
      "step": 48280
    },
    {
      "epoch": 18.75,
      "learning_rate": 1.7499546925566345e-05,
      "loss": 0.027,
      "step": 48290
    },
    {
      "epoch": 18.76,
      "learning_rate": 1.7499029126213593e-05,
      "loss": 0.0379,
      "step": 48300
    },
    {
      "epoch": 18.76,
      "learning_rate": 1.7498511326860845e-05,
      "loss": 0.1472,
      "step": 48310
    },
    {
      "epoch": 18.77,
      "learning_rate": 1.7497993527508093e-05,
      "loss": 0.0388,
      "step": 48320
    },
    {
      "epoch": 18.77,
      "learning_rate": 1.749747572815534e-05,
      "loss": 0.1289,
      "step": 48330
    },
    {
      "epoch": 18.77,
      "learning_rate": 1.7496957928802592e-05,
      "loss": 0.0486,
      "step": 48340
    },
    {
      "epoch": 18.78,
      "learning_rate": 1.749644012944984e-05,
      "loss": 0.1065,
      "step": 48350
    },
    {
      "epoch": 18.78,
      "learning_rate": 1.7495922330097088e-05,
      "loss": 0.1937,
      "step": 48360
    },
    {
      "epoch": 18.78,
      "learning_rate": 1.749540453074434e-05,
      "loss": 0.1527,
      "step": 48370
    },
    {
      "epoch": 18.79,
      "learning_rate": 1.7494886731391587e-05,
      "loss": 0.1009,
      "step": 48380
    },
    {
      "epoch": 18.79,
      "learning_rate": 1.7494368932038835e-05,
      "loss": 0.0856,
      "step": 48390
    },
    {
      "epoch": 18.8,
      "learning_rate": 1.7493851132686086e-05,
      "loss": 0.0128,
      "step": 48400
    },
    {
      "epoch": 18.8,
      "learning_rate": 1.7493333333333334e-05,
      "loss": 0.0982,
      "step": 48410
    },
    {
      "epoch": 18.8,
      "learning_rate": 1.7492815533980585e-05,
      "loss": 0.1788,
      "step": 48420
    },
    {
      "epoch": 18.81,
      "learning_rate": 1.7492297734627833e-05,
      "loss": 0.0834,
      "step": 48430
    },
    {
      "epoch": 18.81,
      "learning_rate": 1.749177993527508e-05,
      "loss": 0.0765,
      "step": 48440
    },
    {
      "epoch": 18.82,
      "learning_rate": 1.7491262135922333e-05,
      "loss": 0.1937,
      "step": 48450
    },
    {
      "epoch": 18.82,
      "learning_rate": 1.749074433656958e-05,
      "loss": 0.1023,
      "step": 48460
    },
    {
      "epoch": 18.82,
      "learning_rate": 1.749022653721683e-05,
      "loss": 0.1177,
      "step": 48470
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.748970873786408e-05,
      "loss": 0.0062,
      "step": 48480
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.7489190938511328e-05,
      "loss": 0.0574,
      "step": 48490
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.748867313915858e-05,
      "loss": 0.0262,
      "step": 48500
    },
    {
      "epoch": 18.84,
      "learning_rate": 1.7488155339805827e-05,
      "loss": 0.1049,
      "step": 48510
    },
    {
      "epoch": 18.84,
      "learning_rate": 1.7487637540453075e-05,
      "loss": 0.2655,
      "step": 48520
    },
    {
      "epoch": 18.85,
      "learning_rate": 1.7487119741100326e-05,
      "loss": 0.1562,
      "step": 48530
    },
    {
      "epoch": 18.85,
      "learning_rate": 1.7486601941747574e-05,
      "loss": 0.207,
      "step": 48540
    },
    {
      "epoch": 18.85,
      "learning_rate": 1.7486084142394822e-05,
      "loss": 0.0418,
      "step": 48550
    },
    {
      "epoch": 18.86,
      "learning_rate": 1.7485566343042073e-05,
      "loss": 0.1027,
      "step": 48560
    },
    {
      "epoch": 18.86,
      "learning_rate": 1.748504854368932e-05,
      "loss": 0.0225,
      "step": 48570
    },
    {
      "epoch": 18.87,
      "learning_rate": 1.7484530744336573e-05,
      "loss": 0.126,
      "step": 48580
    },
    {
      "epoch": 18.87,
      "learning_rate": 1.748401294498382e-05,
      "loss": 0.135,
      "step": 48590
    },
    {
      "epoch": 18.87,
      "learning_rate": 1.748349514563107e-05,
      "loss": 0.0723,
      "step": 48600
    },
    {
      "epoch": 18.88,
      "learning_rate": 1.748297734627832e-05,
      "loss": 0.0948,
      "step": 48610
    },
    {
      "epoch": 18.88,
      "learning_rate": 1.7482459546925568e-05,
      "loss": 0.0524,
      "step": 48620
    },
    {
      "epoch": 18.89,
      "learning_rate": 1.7481941747572816e-05,
      "loss": 0.0915,
      "step": 48630
    },
    {
      "epoch": 18.89,
      "learning_rate": 1.7481423948220067e-05,
      "loss": 0.1722,
      "step": 48640
    },
    {
      "epoch": 18.89,
      "learning_rate": 1.7480906148867315e-05,
      "loss": 0.042,
      "step": 48650
    },
    {
      "epoch": 18.9,
      "learning_rate": 1.7480388349514566e-05,
      "loss": 0.0231,
      "step": 48660
    },
    {
      "epoch": 18.9,
      "learning_rate": 1.7479870550161814e-05,
      "loss": 0.0652,
      "step": 48670
    },
    {
      "epoch": 18.9,
      "learning_rate": 1.7479352750809062e-05,
      "loss": 0.1463,
      "step": 48680
    },
    {
      "epoch": 18.91,
      "learning_rate": 1.7478834951456313e-05,
      "loss": 0.1106,
      "step": 48690
    },
    {
      "epoch": 18.91,
      "learning_rate": 1.747831715210356e-05,
      "loss": 0.2006,
      "step": 48700
    },
    {
      "epoch": 18.92,
      "learning_rate": 1.747779935275081e-05,
      "loss": 0.3128,
      "step": 48710
    },
    {
      "epoch": 18.92,
      "learning_rate": 1.747728155339806e-05,
      "loss": 0.0172,
      "step": 48720
    },
    {
      "epoch": 18.92,
      "learning_rate": 1.747676375404531e-05,
      "loss": 0.0497,
      "step": 48730
    },
    {
      "epoch": 18.93,
      "learning_rate": 1.747624595469256e-05,
      "loss": 0.1033,
      "step": 48740
    },
    {
      "epoch": 18.93,
      "learning_rate": 1.7475728155339808e-05,
      "loss": 0.0796,
      "step": 48750
    },
    {
      "epoch": 18.94,
      "learning_rate": 1.7475210355987056e-05,
      "loss": 0.1102,
      "step": 48760
    },
    {
      "epoch": 18.94,
      "learning_rate": 1.7474692556634303e-05,
      "loss": 0.0327,
      "step": 48770
    },
    {
      "epoch": 18.94,
      "learning_rate": 1.7474174757281555e-05,
      "loss": 0.1929,
      "step": 48780
    },
    {
      "epoch": 18.95,
      "learning_rate": 1.7473656957928803e-05,
      "loss": 0.2107,
      "step": 48790
    },
    {
      "epoch": 18.95,
      "learning_rate": 1.7473139158576054e-05,
      "loss": 0.1371,
      "step": 48800
    },
    {
      "epoch": 18.96,
      "learning_rate": 1.7472621359223302e-05,
      "loss": 0.0976,
      "step": 48810
    },
    {
      "epoch": 18.96,
      "learning_rate": 1.7472103559870553e-05,
      "loss": 0.0604,
      "step": 48820
    },
    {
      "epoch": 18.96,
      "learning_rate": 1.74715857605178e-05,
      "loss": 0.2363,
      "step": 48830
    },
    {
      "epoch": 18.97,
      "learning_rate": 1.747106796116505e-05,
      "loss": 0.1161,
      "step": 48840
    },
    {
      "epoch": 18.97,
      "learning_rate": 1.7470550161812297e-05,
      "loss": 0.1229,
      "step": 48850
    },
    {
      "epoch": 18.97,
      "learning_rate": 1.7470032362459548e-05,
      "loss": 0.1801,
      "step": 48860
    },
    {
      "epoch": 18.98,
      "learning_rate": 1.7469514563106796e-05,
      "loss": 0.1012,
      "step": 48870
    },
    {
      "epoch": 18.98,
      "learning_rate": 1.7468996763754048e-05,
      "loss": 0.1942,
      "step": 48880
    },
    {
      "epoch": 18.99,
      "learning_rate": 1.7468478964401295e-05,
      "loss": 0.2065,
      "step": 48890
    },
    {
      "epoch": 18.99,
      "learning_rate": 1.7467961165048547e-05,
      "loss": 0.0043,
      "step": 48900
    },
    {
      "epoch": 18.99,
      "learning_rate": 1.7467443365695795e-05,
      "loss": 0.1087,
      "step": 48910
    },
    {
      "epoch": 19.0,
      "learning_rate": 1.7466925566343043e-05,
      "loss": 0.1207,
      "step": 48920
    },
    {
      "epoch": 19.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.2652722895145416,
      "eval_runtime": 8.2313,
      "eval_samples_per_second": 441.609,
      "eval_steps_per_second": 55.277,
      "step": 48925
    },
    {
      "epoch": 19.0,
      "learning_rate": 1.746640776699029e-05,
      "loss": 0.0245,
      "step": 48930
    },
    {
      "epoch": 19.01,
      "learning_rate": 1.7465889967637542e-05,
      "loss": 0.1222,
      "step": 48940
    },
    {
      "epoch": 19.01,
      "learning_rate": 1.746537216828479e-05,
      "loss": 0.159,
      "step": 48950
    },
    {
      "epoch": 19.01,
      "learning_rate": 1.746485436893204e-05,
      "loss": 0.034,
      "step": 48960
    },
    {
      "epoch": 19.02,
      "learning_rate": 1.746433656957929e-05,
      "loss": 0.0806,
      "step": 48970
    },
    {
      "epoch": 19.02,
      "learning_rate": 1.746381877022654e-05,
      "loss": 0.1339,
      "step": 48980
    },
    {
      "epoch": 19.03,
      "learning_rate": 1.7463300970873788e-05,
      "loss": 0.0634,
      "step": 48990
    },
    {
      "epoch": 19.03,
      "learning_rate": 1.7462783171521036e-05,
      "loss": 0.1004,
      "step": 49000
    },
    {
      "epoch": 19.03,
      "learning_rate": 1.7462265372168284e-05,
      "loss": 0.0031,
      "step": 49010
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.7461747572815535e-05,
      "loss": 0.1074,
      "step": 49020
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.7461229773462783e-05,
      "loss": 0.0932,
      "step": 49030
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.7460711974110035e-05,
      "loss": 0.1181,
      "step": 49040
    },
    {
      "epoch": 19.05,
      "learning_rate": 1.7460194174757283e-05,
      "loss": 0.012,
      "step": 49050
    },
    {
      "epoch": 19.05,
      "learning_rate": 1.7459676375404534e-05,
      "loss": 0.1658,
      "step": 49060
    },
    {
      "epoch": 19.06,
      "learning_rate": 1.7459158576051782e-05,
      "loss": 0.0783,
      "step": 49070
    },
    {
      "epoch": 19.06,
      "learning_rate": 1.745864077669903e-05,
      "loss": 0.2154,
      "step": 49080
    },
    {
      "epoch": 19.06,
      "learning_rate": 1.7458122977346278e-05,
      "loss": 0.1293,
      "step": 49090
    },
    {
      "epoch": 19.07,
      "learning_rate": 1.745760517799353e-05,
      "loss": 0.164,
      "step": 49100
    },
    {
      "epoch": 19.07,
      "learning_rate": 1.7457087378640777e-05,
      "loss": 0.2291,
      "step": 49110
    },
    {
      "epoch": 19.08,
      "learning_rate": 1.7456569579288028e-05,
      "loss": 0.2642,
      "step": 49120
    },
    {
      "epoch": 19.08,
      "learning_rate": 1.7456051779935276e-05,
      "loss": 0.2168,
      "step": 49130
    },
    {
      "epoch": 19.08,
      "learning_rate": 1.7455533980582527e-05,
      "loss": 0.169,
      "step": 49140
    },
    {
      "epoch": 19.09,
      "learning_rate": 1.7455016181229775e-05,
      "loss": 0.076,
      "step": 49150
    },
    {
      "epoch": 19.09,
      "learning_rate": 1.7454498381877023e-05,
      "loss": 0.1666,
      "step": 49160
    },
    {
      "epoch": 19.1,
      "learning_rate": 1.745398058252427e-05,
      "loss": 0.1594,
      "step": 49170
    },
    {
      "epoch": 19.1,
      "learning_rate": 1.7453462783171523e-05,
      "loss": 0.1823,
      "step": 49180
    },
    {
      "epoch": 19.1,
      "learning_rate": 1.745294498381877e-05,
      "loss": 0.163,
      "step": 49190
    },
    {
      "epoch": 19.11,
      "learning_rate": 1.7452427184466022e-05,
      "loss": 0.1596,
      "step": 49200
    },
    {
      "epoch": 19.11,
      "learning_rate": 1.745190938511327e-05,
      "loss": 0.2005,
      "step": 49210
    },
    {
      "epoch": 19.11,
      "learning_rate": 1.745139158576052e-05,
      "loss": 0.1091,
      "step": 49220
    },
    {
      "epoch": 19.12,
      "learning_rate": 1.745087378640777e-05,
      "loss": 0.0673,
      "step": 49230
    },
    {
      "epoch": 19.12,
      "learning_rate": 1.7450355987055017e-05,
      "loss": 0.0204,
      "step": 49240
    },
    {
      "epoch": 19.13,
      "learning_rate": 1.7449838187702265e-05,
      "loss": 0.1985,
      "step": 49250
    },
    {
      "epoch": 19.13,
      "learning_rate": 1.7449320388349516e-05,
      "loss": 0.1189,
      "step": 49260
    },
    {
      "epoch": 19.13,
      "learning_rate": 1.7448802588996764e-05,
      "loss": 0.1767,
      "step": 49270
    },
    {
      "epoch": 19.14,
      "learning_rate": 1.7448284789644015e-05,
      "loss": 0.1481,
      "step": 49280
    },
    {
      "epoch": 19.14,
      "learning_rate": 1.7447766990291263e-05,
      "loss": 0.2195,
      "step": 49290
    },
    {
      "epoch": 19.15,
      "learning_rate": 1.7447249190938515e-05,
      "loss": 0.2698,
      "step": 49300
    },
    {
      "epoch": 19.15,
      "learning_rate": 1.7446731391585762e-05,
      "loss": 0.1411,
      "step": 49310
    },
    {
      "epoch": 19.15,
      "learning_rate": 1.744621359223301e-05,
      "loss": 0.1223,
      "step": 49320
    },
    {
      "epoch": 19.16,
      "learning_rate": 1.744569579288026e-05,
      "loss": 0.2185,
      "step": 49330
    },
    {
      "epoch": 19.16,
      "learning_rate": 1.744517799352751e-05,
      "loss": 0.1125,
      "step": 49340
    },
    {
      "epoch": 19.17,
      "learning_rate": 1.7444660194174758e-05,
      "loss": 0.0959,
      "step": 49350
    },
    {
      "epoch": 19.17,
      "learning_rate": 1.744414239482201e-05,
      "loss": 0.0446,
      "step": 49360
    },
    {
      "epoch": 19.17,
      "learning_rate": 1.7443624595469257e-05,
      "loss": 0.3299,
      "step": 49370
    },
    {
      "epoch": 19.18,
      "learning_rate": 1.7443106796116508e-05,
      "loss": 0.1361,
      "step": 49380
    },
    {
      "epoch": 19.18,
      "learning_rate": 1.7442588996763756e-05,
      "loss": 0.0486,
      "step": 49390
    },
    {
      "epoch": 19.18,
      "learning_rate": 1.7442071197411007e-05,
      "loss": 0.1496,
      "step": 49400
    },
    {
      "epoch": 19.19,
      "learning_rate": 1.7441553398058252e-05,
      "loss": 0.3981,
      "step": 49410
    },
    {
      "epoch": 19.19,
      "learning_rate": 1.7441035598705503e-05,
      "loss": 0.1766,
      "step": 49420
    },
    {
      "epoch": 19.2,
      "learning_rate": 1.744051779935275e-05,
      "loss": 0.1676,
      "step": 49430
    },
    {
      "epoch": 19.2,
      "learning_rate": 1.7440000000000002e-05,
      "loss": 0.0564,
      "step": 49440
    },
    {
      "epoch": 19.2,
      "learning_rate": 1.743948220064725e-05,
      "loss": 0.1392,
      "step": 49450
    },
    {
      "epoch": 19.21,
      "learning_rate": 1.74389644012945e-05,
      "loss": 0.1182,
      "step": 49460
    },
    {
      "epoch": 19.21,
      "learning_rate": 1.743844660194175e-05,
      "loss": 0.2032,
      "step": 49470
    },
    {
      "epoch": 19.22,
      "learning_rate": 1.7437928802589e-05,
      "loss": 0.0803,
      "step": 49480
    },
    {
      "epoch": 19.22,
      "learning_rate": 1.7437411003236245e-05,
      "loss": 0.0928,
      "step": 49490
    },
    {
      "epoch": 19.22,
      "learning_rate": 1.7436893203883497e-05,
      "loss": 0.0449,
      "step": 49500
    },
    {
      "epoch": 19.23,
      "learning_rate": 1.7436375404530745e-05,
      "loss": 0.053,
      "step": 49510
    },
    {
      "epoch": 19.23,
      "learning_rate": 1.7435857605177996e-05,
      "loss": 0.0515,
      "step": 49520
    },
    {
      "epoch": 19.23,
      "learning_rate": 1.7435339805825244e-05,
      "loss": 0.001,
      "step": 49530
    },
    {
      "epoch": 19.24,
      "learning_rate": 1.7434822006472495e-05,
      "loss": 0.1151,
      "step": 49540
    },
    {
      "epoch": 19.24,
      "learning_rate": 1.7434304207119743e-05,
      "loss": 0.1922,
      "step": 49550
    },
    {
      "epoch": 19.25,
      "learning_rate": 1.743378640776699e-05,
      "loss": 0.1342,
      "step": 49560
    },
    {
      "epoch": 19.25,
      "learning_rate": 1.743326860841424e-05,
      "loss": 0.0701,
      "step": 49570
    },
    {
      "epoch": 19.25,
      "learning_rate": 1.743275080906149e-05,
      "loss": 0.1211,
      "step": 49580
    },
    {
      "epoch": 19.26,
      "learning_rate": 1.7432233009708738e-05,
      "loss": 0.0897,
      "step": 49590
    },
    {
      "epoch": 19.26,
      "learning_rate": 1.743171521035599e-05,
      "loss": 0.0637,
      "step": 49600
    },
    {
      "epoch": 19.27,
      "learning_rate": 1.7431197411003237e-05,
      "loss": 0.0619,
      "step": 49610
    },
    {
      "epoch": 19.27,
      "learning_rate": 1.743067961165049e-05,
      "loss": 0.1544,
      "step": 49620
    },
    {
      "epoch": 19.27,
      "learning_rate": 1.7430161812297737e-05,
      "loss": 0.0799,
      "step": 49630
    },
    {
      "epoch": 19.28,
      "learning_rate": 1.7429644012944985e-05,
      "loss": 0.1566,
      "step": 49640
    },
    {
      "epoch": 19.28,
      "learning_rate": 1.7429126213592233e-05,
      "loss": 0.1533,
      "step": 49650
    },
    {
      "epoch": 19.29,
      "learning_rate": 1.7428608414239484e-05,
      "loss": 0.2471,
      "step": 49660
    },
    {
      "epoch": 19.29,
      "learning_rate": 1.7428090614886732e-05,
      "loss": 0.1948,
      "step": 49670
    },
    {
      "epoch": 19.29,
      "learning_rate": 1.7427572815533983e-05,
      "loss": 0.1378,
      "step": 49680
    },
    {
      "epoch": 19.3,
      "learning_rate": 1.742705501618123e-05,
      "loss": 0.1995,
      "step": 49690
    },
    {
      "epoch": 19.3,
      "learning_rate": 1.7426537216828482e-05,
      "loss": 0.0676,
      "step": 49700
    },
    {
      "epoch": 19.3,
      "learning_rate": 1.742601941747573e-05,
      "loss": 0.0299,
      "step": 49710
    },
    {
      "epoch": 19.31,
      "learning_rate": 1.7425501618122978e-05,
      "loss": 0.1408,
      "step": 49720
    },
    {
      "epoch": 19.31,
      "learning_rate": 1.7424983818770226e-05,
      "loss": 0.0066,
      "step": 49730
    },
    {
      "epoch": 19.32,
      "learning_rate": 1.7424466019417477e-05,
      "loss": 0.0566,
      "step": 49740
    },
    {
      "epoch": 19.32,
      "learning_rate": 1.7423948220064725e-05,
      "loss": 0.2291,
      "step": 49750
    },
    {
      "epoch": 19.32,
      "learning_rate": 1.7423430420711977e-05,
      "loss": 0.1213,
      "step": 49760
    },
    {
      "epoch": 19.33,
      "learning_rate": 1.7422912621359225e-05,
      "loss": 0.1675,
      "step": 49770
    },
    {
      "epoch": 19.33,
      "learning_rate": 1.7422394822006476e-05,
      "loss": 0.0099,
      "step": 49780
    },
    {
      "epoch": 19.34,
      "learning_rate": 1.7421877022653724e-05,
      "loss": 0.1317,
      "step": 49790
    },
    {
      "epoch": 19.34,
      "learning_rate": 1.7421359223300972e-05,
      "loss": 0.1112,
      "step": 49800
    },
    {
      "epoch": 19.34,
      "learning_rate": 1.742084142394822e-05,
      "loss": 0.1085,
      "step": 49810
    },
    {
      "epoch": 19.35,
      "learning_rate": 1.742032362459547e-05,
      "loss": 0.2475,
      "step": 49820
    },
    {
      "epoch": 19.35,
      "learning_rate": 1.741980582524272e-05,
      "loss": 0.0022,
      "step": 49830
    },
    {
      "epoch": 19.36,
      "learning_rate": 1.741928802588997e-05,
      "loss": 0.0637,
      "step": 49840
    },
    {
      "epoch": 19.36,
      "learning_rate": 1.7418770226537218e-05,
      "loss": 0.1123,
      "step": 49850
    },
    {
      "epoch": 19.36,
      "learning_rate": 1.7418252427184466e-05,
      "loss": 0.008,
      "step": 49860
    },
    {
      "epoch": 19.37,
      "learning_rate": 1.7417734627831717e-05,
      "loss": 0.1519,
      "step": 49870
    },
    {
      "epoch": 19.37,
      "learning_rate": 1.7417216828478965e-05,
      "loss": 0.1478,
      "step": 49880
    },
    {
      "epoch": 19.37,
      "learning_rate": 1.7416699029126213e-05,
      "loss": 0.1335,
      "step": 49890
    },
    {
      "epoch": 19.38,
      "learning_rate": 1.7416181229773465e-05,
      "loss": 0.1403,
      "step": 49900
    },
    {
      "epoch": 19.38,
      "learning_rate": 1.7415663430420712e-05,
      "loss": 0.0549,
      "step": 49910
    },
    {
      "epoch": 19.39,
      "learning_rate": 1.7415145631067964e-05,
      "loss": 0.0783,
      "step": 49920
    },
    {
      "epoch": 19.39,
      "learning_rate": 1.741462783171521e-05,
      "loss": 0.096,
      "step": 49930
    },
    {
      "epoch": 19.39,
      "learning_rate": 1.741411003236246e-05,
      "loss": 0.0044,
      "step": 49940
    },
    {
      "epoch": 19.4,
      "learning_rate": 1.741359223300971e-05,
      "loss": 0.3061,
      "step": 49950
    },
    {
      "epoch": 19.4,
      "learning_rate": 1.741307443365696e-05,
      "loss": 0.1224,
      "step": 49960
    },
    {
      "epoch": 19.41,
      "learning_rate": 1.741255663430421e-05,
      "loss": 0.0699,
      "step": 49970
    },
    {
      "epoch": 19.41,
      "learning_rate": 1.7412038834951458e-05,
      "loss": 0.1114,
      "step": 49980
    },
    {
      "epoch": 19.41,
      "learning_rate": 1.7411521035598706e-05,
      "loss": 0.202,
      "step": 49990
    },
    {
      "epoch": 19.42,
      "learning_rate": 1.7411003236245957e-05,
      "loss": 0.1045,
      "step": 50000
    },
    {
      "epoch": 19.42,
      "learning_rate": 1.7410485436893205e-05,
      "loss": 0.1783,
      "step": 50010
    },
    {
      "epoch": 19.43,
      "learning_rate": 1.7409967637540453e-05,
      "loss": 0.079,
      "step": 50020
    },
    {
      "epoch": 19.43,
      "learning_rate": 1.7409449838187704e-05,
      "loss": 0.161,
      "step": 50030
    },
    {
      "epoch": 19.43,
      "learning_rate": 1.7408932038834952e-05,
      "loss": 0.064,
      "step": 50040
    },
    {
      "epoch": 19.44,
      "learning_rate": 1.7408414239482204e-05,
      "loss": 0.1318,
      "step": 50050
    },
    {
      "epoch": 19.44,
      "learning_rate": 1.740789644012945e-05,
      "loss": 0.0926,
      "step": 50060
    },
    {
      "epoch": 19.44,
      "learning_rate": 1.74073786407767e-05,
      "loss": 0.1274,
      "step": 50070
    },
    {
      "epoch": 19.45,
      "learning_rate": 1.740686084142395e-05,
      "loss": 0.1386,
      "step": 50080
    },
    {
      "epoch": 19.45,
      "learning_rate": 1.74063430420712e-05,
      "loss": 0.217,
      "step": 50090
    },
    {
      "epoch": 19.46,
      "learning_rate": 1.7405825242718447e-05,
      "loss": 0.0667,
      "step": 50100
    },
    {
      "epoch": 19.46,
      "learning_rate": 1.7405307443365698e-05,
      "loss": 0.0133,
      "step": 50110
    },
    {
      "epoch": 19.46,
      "learning_rate": 1.7404789644012946e-05,
      "loss": 0.0245,
      "step": 50120
    },
    {
      "epoch": 19.47,
      "learning_rate": 1.7404271844660197e-05,
      "loss": 0.1623,
      "step": 50130
    },
    {
      "epoch": 19.47,
      "learning_rate": 1.7403754045307445e-05,
      "loss": 0.1507,
      "step": 50140
    },
    {
      "epoch": 19.48,
      "learning_rate": 1.7403236245954693e-05,
      "loss": 0.1245,
      "step": 50150
    },
    {
      "epoch": 19.48,
      "learning_rate": 1.7402718446601944e-05,
      "loss": 0.0739,
      "step": 50160
    },
    {
      "epoch": 19.48,
      "learning_rate": 1.7402200647249192e-05,
      "loss": 0.1465,
      "step": 50170
    },
    {
      "epoch": 19.49,
      "learning_rate": 1.740168284789644e-05,
      "loss": 0.1578,
      "step": 50180
    },
    {
      "epoch": 19.49,
      "learning_rate": 1.740116504854369e-05,
      "loss": 0.0943,
      "step": 50190
    },
    {
      "epoch": 19.5,
      "learning_rate": 1.740064724919094e-05,
      "loss": 0.0279,
      "step": 50200
    },
    {
      "epoch": 19.5,
      "learning_rate": 1.740012944983819e-05,
      "loss": 0.0488,
      "step": 50210
    },
    {
      "epoch": 19.5,
      "learning_rate": 1.739961165048544e-05,
      "loss": 0.0344,
      "step": 50220
    },
    {
      "epoch": 19.51,
      "learning_rate": 1.7399093851132687e-05,
      "loss": 0.1304,
      "step": 50230
    },
    {
      "epoch": 19.51,
      "learning_rate": 1.7398576051779935e-05,
      "loss": 0.1182,
      "step": 50240
    },
    {
      "epoch": 19.51,
      "learning_rate": 1.7398058252427186e-05,
      "loss": 0.1009,
      "step": 50250
    },
    {
      "epoch": 19.52,
      "learning_rate": 1.7397540453074434e-05,
      "loss": 0.0704,
      "step": 50260
    },
    {
      "epoch": 19.52,
      "learning_rate": 1.7397022653721685e-05,
      "loss": 0.2013,
      "step": 50270
    },
    {
      "epoch": 19.53,
      "learning_rate": 1.7396504854368933e-05,
      "loss": 0.0945,
      "step": 50280
    },
    {
      "epoch": 19.53,
      "learning_rate": 1.7395987055016184e-05,
      "loss": 0.0498,
      "step": 50290
    },
    {
      "epoch": 19.53,
      "learning_rate": 1.7395469255663432e-05,
      "loss": 0.0619,
      "step": 50300
    },
    {
      "epoch": 19.54,
      "learning_rate": 1.739495145631068e-05,
      "loss": 0.0972,
      "step": 50310
    },
    {
      "epoch": 19.54,
      "learning_rate": 1.7394433656957928e-05,
      "loss": 0.0718,
      "step": 50320
    },
    {
      "epoch": 19.55,
      "learning_rate": 1.739391585760518e-05,
      "loss": 0.0355,
      "step": 50330
    },
    {
      "epoch": 19.55,
      "learning_rate": 1.7393398058252427e-05,
      "loss": 0.0788,
      "step": 50340
    },
    {
      "epoch": 19.55,
      "learning_rate": 1.739288025889968e-05,
      "loss": 0.2832,
      "step": 50350
    },
    {
      "epoch": 19.56,
      "learning_rate": 1.7392362459546927e-05,
      "loss": 0.0647,
      "step": 50360
    },
    {
      "epoch": 19.56,
      "learning_rate": 1.7391844660194178e-05,
      "loss": 0.1148,
      "step": 50370
    },
    {
      "epoch": 19.57,
      "learning_rate": 1.7391326860841426e-05,
      "loss": 0.1926,
      "step": 50380
    },
    {
      "epoch": 19.57,
      "learning_rate": 1.7390809061488674e-05,
      "loss": 0.0913,
      "step": 50390
    },
    {
      "epoch": 19.57,
      "learning_rate": 1.7390291262135922e-05,
      "loss": 0.0315,
      "step": 50400
    },
    {
      "epoch": 19.58,
      "learning_rate": 1.7389773462783173e-05,
      "loss": 0.1858,
      "step": 50410
    },
    {
      "epoch": 19.58,
      "learning_rate": 1.738925566343042e-05,
      "loss": 0.0953,
      "step": 50420
    },
    {
      "epoch": 19.58,
      "learning_rate": 1.7388737864077672e-05,
      "loss": 0.1604,
      "step": 50430
    },
    {
      "epoch": 19.59,
      "learning_rate": 1.738822006472492e-05,
      "loss": 0.0792,
      "step": 50440
    },
    {
      "epoch": 19.59,
      "learning_rate": 1.738770226537217e-05,
      "loss": 0.1755,
      "step": 50450
    },
    {
      "epoch": 19.6,
      "learning_rate": 1.738718446601942e-05,
      "loss": 0.0299,
      "step": 50460
    },
    {
      "epoch": 19.6,
      "learning_rate": 1.7386666666666667e-05,
      "loss": 0.0274,
      "step": 50470
    },
    {
      "epoch": 19.6,
      "learning_rate": 1.7386148867313915e-05,
      "loss": 0.1064,
      "step": 50480
    },
    {
      "epoch": 19.61,
      "learning_rate": 1.7385631067961167e-05,
      "loss": 0.0189,
      "step": 50490
    },
    {
      "epoch": 19.61,
      "learning_rate": 1.7385113268608415e-05,
      "loss": 0.236,
      "step": 50500
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.7384595469255666e-05,
      "loss": 0.1419,
      "step": 50510
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.7384077669902914e-05,
      "loss": 0.0801,
      "step": 50520
    },
    {
      "epoch": 19.62,
      "learning_rate": 1.7383559870550165e-05,
      "loss": 0.0565,
      "step": 50530
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.7383042071197413e-05,
      "loss": 0.0967,
      "step": 50540
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.738252427184466e-05,
      "loss": 0.1786,
      "step": 50550
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.738200647249191e-05,
      "loss": 0.2562,
      "step": 50560
    },
    {
      "epoch": 19.64,
      "learning_rate": 1.738148867313916e-05,
      "loss": 0.1046,
      "step": 50570
    },
    {
      "epoch": 19.64,
      "learning_rate": 1.7380970873786408e-05,
      "loss": 0.1057,
      "step": 50580
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.738045307443366e-05,
      "loss": 0.2362,
      "step": 50590
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.7379935275080907e-05,
      "loss": 0.1915,
      "step": 50600
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.737941747572816e-05,
      "loss": 0.1377,
      "step": 50610
    },
    {
      "epoch": 19.66,
      "learning_rate": 1.7378899676375407e-05,
      "loss": 0.0067,
      "step": 50620
    },
    {
      "epoch": 19.66,
      "learning_rate": 1.7378381877022654e-05,
      "loss": 0.0672,
      "step": 50630
    },
    {
      "epoch": 19.67,
      "learning_rate": 1.7377864077669902e-05,
      "loss": 0.0935,
      "step": 50640
    },
    {
      "epoch": 19.67,
      "learning_rate": 1.7377346278317154e-05,
      "loss": 0.0526,
      "step": 50650
    },
    {
      "epoch": 19.67,
      "learning_rate": 1.73768284789644e-05,
      "loss": 0.2024,
      "step": 50660
    },
    {
      "epoch": 19.68,
      "learning_rate": 1.7376310679611653e-05,
      "loss": 0.0413,
      "step": 50670
    },
    {
      "epoch": 19.68,
      "learning_rate": 1.73757928802589e-05,
      "loss": 0.1356,
      "step": 50680
    },
    {
      "epoch": 19.69,
      "learning_rate": 1.7375275080906152e-05,
      "loss": 0.1866,
      "step": 50690
    },
    {
      "epoch": 19.69,
      "learning_rate": 1.73747572815534e-05,
      "loss": 0.1994,
      "step": 50700
    },
    {
      "epoch": 19.69,
      "learning_rate": 1.7374239482200648e-05,
      "loss": 0.1442,
      "step": 50710
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.7373721682847896e-05,
      "loss": 0.0445,
      "step": 50720
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.7373203883495147e-05,
      "loss": 0.102,
      "step": 50730
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.7372686084142395e-05,
      "loss": 0.0472,
      "step": 50740
    },
    {
      "epoch": 19.71,
      "learning_rate": 1.7372168284789646e-05,
      "loss": 0.0474,
      "step": 50750
    },
    {
      "epoch": 19.71,
      "learning_rate": 1.7371650485436894e-05,
      "loss": 0.0209,
      "step": 50760
    },
    {
      "epoch": 19.72,
      "learning_rate": 1.7371132686084146e-05,
      "loss": 0.2964,
      "step": 50770
    },
    {
      "epoch": 19.72,
      "learning_rate": 1.7370614886731394e-05,
      "loss": 0.2166,
      "step": 50780
    },
    {
      "epoch": 19.72,
      "learning_rate": 1.737009708737864e-05,
      "loss": 0.227,
      "step": 50790
    },
    {
      "epoch": 19.73,
      "learning_rate": 1.736957928802589e-05,
      "loss": 0.2381,
      "step": 50800
    },
    {
      "epoch": 19.73,
      "learning_rate": 1.736906148867314e-05,
      "loss": 0.057,
      "step": 50810
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.736854368932039e-05,
      "loss": 0.1084,
      "step": 50820
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.736802588996764e-05,
      "loss": 0.2242,
      "step": 50830
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.7367508090614888e-05,
      "loss": 0.1989,
      "step": 50840
    },
    {
      "epoch": 19.75,
      "learning_rate": 1.736699029126214e-05,
      "loss": 0.1409,
      "step": 50850
    },
    {
      "epoch": 19.75,
      "learning_rate": 1.7366472491909387e-05,
      "loss": 0.1323,
      "step": 50860
    },
    {
      "epoch": 19.76,
      "learning_rate": 1.7365954692556635e-05,
      "loss": 0.0579,
      "step": 50870
    },
    {
      "epoch": 19.76,
      "learning_rate": 1.7365436893203883e-05,
      "loss": 0.037,
      "step": 50880
    },
    {
      "epoch": 19.76,
      "learning_rate": 1.7364919093851134e-05,
      "loss": 0.1048,
      "step": 50890
    },
    {
      "epoch": 19.77,
      "learning_rate": 1.7364401294498382e-05,
      "loss": 0.1661,
      "step": 50900
    },
    {
      "epoch": 19.77,
      "learning_rate": 1.7363883495145634e-05,
      "loss": 0.034,
      "step": 50910
    },
    {
      "epoch": 19.77,
      "learning_rate": 1.736336569579288e-05,
      "loss": 0.0959,
      "step": 50920
    },
    {
      "epoch": 19.78,
      "learning_rate": 1.7362847896440133e-05,
      "loss": 0.2041,
      "step": 50930
    },
    {
      "epoch": 19.78,
      "learning_rate": 1.736233009708738e-05,
      "loss": 0.2057,
      "step": 50940
    },
    {
      "epoch": 19.79,
      "learning_rate": 1.736181229773463e-05,
      "loss": 0.0545,
      "step": 50950
    },
    {
      "epoch": 19.79,
      "learning_rate": 1.7361294498381877e-05,
      "loss": 0.2424,
      "step": 50960
    },
    {
      "epoch": 19.79,
      "learning_rate": 1.7360776699029128e-05,
      "loss": 0.0832,
      "step": 50970
    },
    {
      "epoch": 19.8,
      "learning_rate": 1.7360258899676376e-05,
      "loss": 0.1155,
      "step": 50980
    },
    {
      "epoch": 19.8,
      "learning_rate": 1.7359741100323627e-05,
      "loss": 0.1719,
      "step": 50990
    },
    {
      "epoch": 19.81,
      "learning_rate": 1.7359223300970875e-05,
      "loss": 0.2382,
      "step": 51000
    },
    {
      "epoch": 19.81,
      "learning_rate": 1.7358705501618126e-05,
      "loss": 0.0812,
      "step": 51010
    },
    {
      "epoch": 19.81,
      "learning_rate": 1.7358187702265374e-05,
      "loss": 0.1687,
      "step": 51020
    },
    {
      "epoch": 19.82,
      "learning_rate": 1.7357669902912622e-05,
      "loss": 0.3025,
      "step": 51030
    },
    {
      "epoch": 19.82,
      "learning_rate": 1.735715210355987e-05,
      "loss": 0.085,
      "step": 51040
    },
    {
      "epoch": 19.83,
      "learning_rate": 1.735663430420712e-05,
      "loss": 0.0088,
      "step": 51050
    },
    {
      "epoch": 19.83,
      "learning_rate": 1.735611650485437e-05,
      "loss": 0.1703,
      "step": 51060
    },
    {
      "epoch": 19.83,
      "learning_rate": 1.735559870550162e-05,
      "loss": 0.0776,
      "step": 51070
    },
    {
      "epoch": 19.84,
      "learning_rate": 1.735508090614887e-05,
      "loss": 0.1223,
      "step": 51080
    },
    {
      "epoch": 19.84,
      "learning_rate": 1.735456310679612e-05,
      "loss": 0.0896,
      "step": 51090
    },
    {
      "epoch": 19.84,
      "learning_rate": 1.7354045307443368e-05,
      "loss": 0.2157,
      "step": 51100
    },
    {
      "epoch": 19.85,
      "learning_rate": 1.7353527508090616e-05,
      "loss": 0.2527,
      "step": 51110
    },
    {
      "epoch": 19.85,
      "learning_rate": 1.7353009708737864e-05,
      "loss": 0.2172,
      "step": 51120
    },
    {
      "epoch": 19.86,
      "learning_rate": 1.7352491909385115e-05,
      "loss": 0.0325,
      "step": 51130
    },
    {
      "epoch": 19.86,
      "learning_rate": 1.7351974110032363e-05,
      "loss": 0.1322,
      "step": 51140
    },
    {
      "epoch": 19.86,
      "learning_rate": 1.7351456310679614e-05,
      "loss": 0.1467,
      "step": 51150
    },
    {
      "epoch": 19.87,
      "learning_rate": 1.7350938511326862e-05,
      "loss": 0.0802,
      "step": 51160
    },
    {
      "epoch": 19.87,
      "learning_rate": 1.7350420711974114e-05,
      "loss": 0.0035,
      "step": 51170
    },
    {
      "epoch": 19.88,
      "learning_rate": 1.734990291262136e-05,
      "loss": 0.0835,
      "step": 51180
    },
    {
      "epoch": 19.88,
      "learning_rate": 1.734938511326861e-05,
      "loss": 0.0306,
      "step": 51190
    },
    {
      "epoch": 19.88,
      "learning_rate": 1.7348867313915857e-05,
      "loss": 0.1249,
      "step": 51200
    },
    {
      "epoch": 19.89,
      "learning_rate": 1.734834951456311e-05,
      "loss": 0.17,
      "step": 51210
    },
    {
      "epoch": 19.89,
      "learning_rate": 1.7347831715210357e-05,
      "loss": 0.1072,
      "step": 51220
    },
    {
      "epoch": 19.9,
      "learning_rate": 1.7347313915857608e-05,
      "loss": 0.1426,
      "step": 51230
    },
    {
      "epoch": 19.9,
      "learning_rate": 1.7346796116504856e-05,
      "loss": 0.0559,
      "step": 51240
    },
    {
      "epoch": 19.9,
      "learning_rate": 1.7346278317152107e-05,
      "loss": 0.1179,
      "step": 51250
    },
    {
      "epoch": 19.91,
      "learning_rate": 1.7345760517799355e-05,
      "loss": 0.0412,
      "step": 51260
    },
    {
      "epoch": 19.91,
      "learning_rate": 1.7345242718446603e-05,
      "loss": 0.0983,
      "step": 51270
    },
    {
      "epoch": 19.91,
      "learning_rate": 1.734472491909385e-05,
      "loss": 0.1342,
      "step": 51280
    },
    {
      "epoch": 19.92,
      "learning_rate": 1.7344207119741102e-05,
      "loss": 0.0655,
      "step": 51290
    },
    {
      "epoch": 19.92,
      "learning_rate": 1.734368932038835e-05,
      "loss": 0.1522,
      "step": 51300
    },
    {
      "epoch": 19.93,
      "learning_rate": 1.73431715210356e-05,
      "loss": 0.0713,
      "step": 51310
    },
    {
      "epoch": 19.93,
      "learning_rate": 1.734265372168285e-05,
      "loss": 0.1679,
      "step": 51320
    },
    {
      "epoch": 19.93,
      "learning_rate": 1.7342135922330097e-05,
      "loss": 0.2963,
      "step": 51330
    },
    {
      "epoch": 19.94,
      "learning_rate": 1.734161812297735e-05,
      "loss": 0.2025,
      "step": 51340
    },
    {
      "epoch": 19.94,
      "learning_rate": 1.7341100323624596e-05,
      "loss": 0.137,
      "step": 51350
    },
    {
      "epoch": 19.95,
      "learning_rate": 1.7340582524271844e-05,
      "loss": 0.0794,
      "step": 51360
    },
    {
      "epoch": 19.95,
      "learning_rate": 1.7340064724919096e-05,
      "loss": 0.2545,
      "step": 51370
    },
    {
      "epoch": 19.95,
      "learning_rate": 1.7339546925566344e-05,
      "loss": 0.1331,
      "step": 51380
    },
    {
      "epoch": 19.96,
      "learning_rate": 1.7339029126213595e-05,
      "loss": 0.0428,
      "step": 51390
    },
    {
      "epoch": 19.96,
      "learning_rate": 1.7338511326860843e-05,
      "loss": 0.0389,
      "step": 51400
    },
    {
      "epoch": 19.97,
      "learning_rate": 1.733799352750809e-05,
      "loss": 0.0877,
      "step": 51410
    },
    {
      "epoch": 19.97,
      "learning_rate": 1.7337475728155342e-05,
      "loss": 0.1016,
      "step": 51420
    },
    {
      "epoch": 19.97,
      "learning_rate": 1.733695792880259e-05,
      "loss": 0.1743,
      "step": 51430
    },
    {
      "epoch": 19.98,
      "learning_rate": 1.7336440129449838e-05,
      "loss": 0.1156,
      "step": 51440
    },
    {
      "epoch": 19.98,
      "learning_rate": 1.733592233009709e-05,
      "loss": 0.0627,
      "step": 51450
    },
    {
      "epoch": 19.98,
      "learning_rate": 1.7335404530744337e-05,
      "loss": 0.0477,
      "step": 51460
    },
    {
      "epoch": 19.99,
      "learning_rate": 1.733488673139159e-05,
      "loss": 0.099,
      "step": 51470
    },
    {
      "epoch": 19.99,
      "learning_rate": 1.7334368932038836e-05,
      "loss": 0.0362,
      "step": 51480
    },
    {
      "epoch": 20.0,
      "learning_rate": 1.7333851132686084e-05,
      "loss": 0.0542,
      "step": 51490
    },
    {
      "epoch": 20.0,
      "learning_rate": 1.7333333333333336e-05,
      "loss": 0.1761,
      "step": 51500
    },
    {
      "epoch": 20.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.2442636489868164,
      "eval_runtime": 8.2614,
      "eval_samples_per_second": 439.997,
      "eval_steps_per_second": 55.075,
      "step": 51500
    },
    {
      "epoch": 20.0,
      "learning_rate": 1.7332815533980584e-05,
      "loss": 0.163,
      "step": 51510
    },
    {
      "epoch": 20.01,
      "learning_rate": 1.733229773462783e-05,
      "loss": 0.1032,
      "step": 51520
    },
    {
      "epoch": 20.01,
      "learning_rate": 1.7331779935275083e-05,
      "loss": 0.1035,
      "step": 51530
    },
    {
      "epoch": 20.02,
      "learning_rate": 1.733126213592233e-05,
      "loss": 0.0978,
      "step": 51540
    },
    {
      "epoch": 20.02,
      "learning_rate": 1.7330744336569582e-05,
      "loss": 0.1191,
      "step": 51550
    },
    {
      "epoch": 20.02,
      "learning_rate": 1.733022653721683e-05,
      "loss": 0.1474,
      "step": 51560
    },
    {
      "epoch": 20.03,
      "learning_rate": 1.7329708737864078e-05,
      "loss": 0.1916,
      "step": 51570
    },
    {
      "epoch": 20.03,
      "learning_rate": 1.732919093851133e-05,
      "loss": 0.0577,
      "step": 51580
    },
    {
      "epoch": 20.03,
      "learning_rate": 1.7328673139158577e-05,
      "loss": 0.1983,
      "step": 51590
    },
    {
      "epoch": 20.04,
      "learning_rate": 1.732815533980583e-05,
      "loss": 0.1542,
      "step": 51600
    },
    {
      "epoch": 20.04,
      "learning_rate": 1.7327637540453076e-05,
      "loss": 0.0914,
      "step": 51610
    },
    {
      "epoch": 20.05,
      "learning_rate": 1.7327119741100324e-05,
      "loss": 0.0751,
      "step": 51620
    },
    {
      "epoch": 20.05,
      "learning_rate": 1.7326601941747576e-05,
      "loss": 0.1366,
      "step": 51630
    },
    {
      "epoch": 20.05,
      "learning_rate": 1.7326084142394824e-05,
      "loss": 0.1539,
      "step": 51640
    },
    {
      "epoch": 20.06,
      "learning_rate": 1.732556634304207e-05,
      "loss": 0.0435,
      "step": 51650
    },
    {
      "epoch": 20.06,
      "learning_rate": 1.7325048543689323e-05,
      "loss": 0.0372,
      "step": 51660
    },
    {
      "epoch": 20.07,
      "learning_rate": 1.732453074433657e-05,
      "loss": 0.1208,
      "step": 51670
    },
    {
      "epoch": 20.07,
      "learning_rate": 1.7324012944983822e-05,
      "loss": 0.1749,
      "step": 51680
    },
    {
      "epoch": 20.07,
      "learning_rate": 1.732349514563107e-05,
      "loss": 0.0378,
      "step": 51690
    },
    {
      "epoch": 20.08,
      "learning_rate": 1.7322977346278318e-05,
      "loss": 0.0769,
      "step": 51700
    },
    {
      "epoch": 20.08,
      "learning_rate": 1.7322459546925566e-05,
      "loss": 0.1145,
      "step": 51710
    },
    {
      "epoch": 20.09,
      "learning_rate": 1.7321941747572817e-05,
      "loss": 0.1796,
      "step": 51720
    },
    {
      "epoch": 20.09,
      "learning_rate": 1.7321423948220065e-05,
      "loss": 0.0946,
      "step": 51730
    },
    {
      "epoch": 20.09,
      "learning_rate": 1.7320906148867316e-05,
      "loss": 0.0632,
      "step": 51740
    },
    {
      "epoch": 20.1,
      "learning_rate": 1.7320388349514564e-05,
      "loss": 0.1655,
      "step": 51750
    },
    {
      "epoch": 20.1,
      "learning_rate": 1.7319870550161816e-05,
      "loss": 0.0875,
      "step": 51760
    },
    {
      "epoch": 20.1,
      "learning_rate": 1.7319352750809063e-05,
      "loss": 0.1121,
      "step": 51770
    },
    {
      "epoch": 20.11,
      "learning_rate": 1.731883495145631e-05,
      "loss": 0.0812,
      "step": 51780
    },
    {
      "epoch": 20.11,
      "learning_rate": 1.731831715210356e-05,
      "loss": 0.1672,
      "step": 51790
    },
    {
      "epoch": 20.12,
      "learning_rate": 1.731779935275081e-05,
      "loss": 0.1381,
      "step": 51800
    },
    {
      "epoch": 20.12,
      "learning_rate": 1.731728155339806e-05,
      "loss": 0.0979,
      "step": 51810
    },
    {
      "epoch": 20.12,
      "learning_rate": 1.731676375404531e-05,
      "loss": 0.2654,
      "step": 51820
    },
    {
      "epoch": 20.13,
      "learning_rate": 1.7316245954692558e-05,
      "loss": 0.0836,
      "step": 51830
    },
    {
      "epoch": 20.13,
      "learning_rate": 1.731572815533981e-05,
      "loss": 0.0668,
      "step": 51840
    },
    {
      "epoch": 20.14,
      "learning_rate": 1.7315210355987057e-05,
      "loss": 0.0295,
      "step": 51850
    },
    {
      "epoch": 20.14,
      "learning_rate": 1.7314692556634305e-05,
      "loss": 0.0358,
      "step": 51860
    },
    {
      "epoch": 20.14,
      "learning_rate": 1.7314174757281553e-05,
      "loss": 0.0309,
      "step": 51870
    },
    {
      "epoch": 20.15,
      "learning_rate": 1.7313656957928804e-05,
      "loss": 0.0742,
      "step": 51880
    },
    {
      "epoch": 20.15,
      "learning_rate": 1.7313139158576052e-05,
      "loss": 0.0628,
      "step": 51890
    },
    {
      "epoch": 20.16,
      "learning_rate": 1.7312621359223303e-05,
      "loss": 0.021,
      "step": 51900
    },
    {
      "epoch": 20.16,
      "learning_rate": 1.731210355987055e-05,
      "loss": 0.1452,
      "step": 51910
    },
    {
      "epoch": 20.16,
      "learning_rate": 1.7311585760517803e-05,
      "loss": 0.0998,
      "step": 51920
    },
    {
      "epoch": 20.17,
      "learning_rate": 1.731106796116505e-05,
      "loss": 0.1236,
      "step": 51930
    },
    {
      "epoch": 20.17,
      "learning_rate": 1.73105501618123e-05,
      "loss": 0.137,
      "step": 51940
    },
    {
      "epoch": 20.17,
      "learning_rate": 1.7310032362459546e-05,
      "loss": 0.0951,
      "step": 51950
    },
    {
      "epoch": 20.18,
      "learning_rate": 1.7309514563106798e-05,
      "loss": 0.2117,
      "step": 51960
    },
    {
      "epoch": 20.18,
      "learning_rate": 1.7308996763754046e-05,
      "loss": 0.0242,
      "step": 51970
    },
    {
      "epoch": 20.19,
      "learning_rate": 1.7308478964401297e-05,
      "loss": 0.0465,
      "step": 51980
    },
    {
      "epoch": 20.19,
      "learning_rate": 1.7307961165048545e-05,
      "loss": 0.3462,
      "step": 51990
    },
    {
      "epoch": 20.19,
      "learning_rate": 1.7307443365695796e-05,
      "loss": 0.399,
      "step": 52000
    },
    {
      "epoch": 20.2,
      "learning_rate": 1.730692556634304e-05,
      "loss": 0.0979,
      "step": 52010
    },
    {
      "epoch": 20.2,
      "learning_rate": 1.7306407766990292e-05,
      "loss": 0.1911,
      "step": 52020
    },
    {
      "epoch": 20.21,
      "learning_rate": 1.730588996763754e-05,
      "loss": 0.0092,
      "step": 52030
    },
    {
      "epoch": 20.21,
      "learning_rate": 1.730537216828479e-05,
      "loss": 0.235,
      "step": 52040
    },
    {
      "epoch": 20.21,
      "learning_rate": 1.730485436893204e-05,
      "loss": 0.1412,
      "step": 52050
    },
    {
      "epoch": 20.22,
      "learning_rate": 1.730433656957929e-05,
      "loss": 0.1212,
      "step": 52060
    },
    {
      "epoch": 20.22,
      "learning_rate": 1.730381877022654e-05,
      "loss": 0.1283,
      "step": 52070
    },
    {
      "epoch": 20.23,
      "learning_rate": 1.730330097087379e-05,
      "loss": 0.2816,
      "step": 52080
    },
    {
      "epoch": 20.23,
      "learning_rate": 1.7302783171521034e-05,
      "loss": 0.0096,
      "step": 52090
    },
    {
      "epoch": 20.23,
      "learning_rate": 1.7302265372168286e-05,
      "loss": 0.0868,
      "step": 52100
    },
    {
      "epoch": 20.24,
      "learning_rate": 1.7301747572815534e-05,
      "loss": 0.1904,
      "step": 52110
    },
    {
      "epoch": 20.24,
      "learning_rate": 1.7301229773462785e-05,
      "loss": 0.1036,
      "step": 52120
    },
    {
      "epoch": 20.24,
      "learning_rate": 1.7300711974110033e-05,
      "loss": 0.1299,
      "step": 52130
    },
    {
      "epoch": 20.25,
      "learning_rate": 1.7300194174757284e-05,
      "loss": 0.1227,
      "step": 52140
    },
    {
      "epoch": 20.25,
      "learning_rate": 1.7299676375404532e-05,
      "loss": 0.096,
      "step": 52150
    },
    {
      "epoch": 20.26,
      "learning_rate": 1.7299158576051783e-05,
      "loss": 0.1079,
      "step": 52160
    },
    {
      "epoch": 20.26,
      "learning_rate": 1.729864077669903e-05,
      "loss": 0.1389,
      "step": 52170
    },
    {
      "epoch": 20.26,
      "learning_rate": 1.729812297734628e-05,
      "loss": 0.1551,
      "step": 52180
    },
    {
      "epoch": 20.27,
      "learning_rate": 1.7297605177993527e-05,
      "loss": 0.0664,
      "step": 52190
    },
    {
      "epoch": 20.27,
      "learning_rate": 1.729708737864078e-05,
      "loss": 0.0615,
      "step": 52200
    },
    {
      "epoch": 20.28,
      "learning_rate": 1.7296569579288026e-05,
      "loss": 0.3515,
      "step": 52210
    },
    {
      "epoch": 20.28,
      "learning_rate": 1.7296051779935278e-05,
      "loss": 0.2101,
      "step": 52220
    },
    {
      "epoch": 20.28,
      "learning_rate": 1.7295533980582526e-05,
      "loss": 0.1597,
      "step": 52230
    },
    {
      "epoch": 20.29,
      "learning_rate": 1.7295016181229777e-05,
      "loss": 0.1437,
      "step": 52240
    },
    {
      "epoch": 20.29,
      "learning_rate": 1.7294498381877025e-05,
      "loss": 0.1817,
      "step": 52250
    },
    {
      "epoch": 20.3,
      "learning_rate": 1.7293980582524273e-05,
      "loss": 0.0254,
      "step": 52260
    },
    {
      "epoch": 20.3,
      "learning_rate": 1.729346278317152e-05,
      "loss": 0.0866,
      "step": 52270
    },
    {
      "epoch": 20.3,
      "learning_rate": 1.7292944983818772e-05,
      "loss": 0.0428,
      "step": 52280
    },
    {
      "epoch": 20.31,
      "learning_rate": 1.729242718446602e-05,
      "loss": 0.2029,
      "step": 52290
    },
    {
      "epoch": 20.31,
      "learning_rate": 1.729190938511327e-05,
      "loss": 0.0722,
      "step": 52300
    },
    {
      "epoch": 20.31,
      "learning_rate": 1.729139158576052e-05,
      "loss": 0.2068,
      "step": 52310
    },
    {
      "epoch": 20.32,
      "learning_rate": 1.729087378640777e-05,
      "loss": 0.0723,
      "step": 52320
    },
    {
      "epoch": 20.32,
      "learning_rate": 1.729035598705502e-05,
      "loss": 0.1005,
      "step": 52330
    },
    {
      "epoch": 20.33,
      "learning_rate": 1.7289838187702266e-05,
      "loss": 0.0785,
      "step": 52340
    },
    {
      "epoch": 20.33,
      "learning_rate": 1.7289320388349514e-05,
      "loss": 0.0551,
      "step": 52350
    },
    {
      "epoch": 20.33,
      "learning_rate": 1.7288802588996766e-05,
      "loss": 0.1456,
      "step": 52360
    },
    {
      "epoch": 20.34,
      "learning_rate": 1.7288284789644013e-05,
      "loss": 0.0215,
      "step": 52370
    },
    {
      "epoch": 20.34,
      "learning_rate": 1.7287766990291265e-05,
      "loss": 0.2247,
      "step": 52380
    },
    {
      "epoch": 20.35,
      "learning_rate": 1.7287249190938513e-05,
      "loss": 0.0666,
      "step": 52390
    },
    {
      "epoch": 20.35,
      "learning_rate": 1.7286731391585764e-05,
      "loss": 0.2228,
      "step": 52400
    },
    {
      "epoch": 20.35,
      "learning_rate": 1.7286213592233012e-05,
      "loss": 0.0682,
      "step": 52410
    },
    {
      "epoch": 20.36,
      "learning_rate": 1.728569579288026e-05,
      "loss": 0.1255,
      "step": 52420
    },
    {
      "epoch": 20.36,
      "learning_rate": 1.7285177993527508e-05,
      "loss": 0.1495,
      "step": 52430
    },
    {
      "epoch": 20.37,
      "learning_rate": 1.728466019417476e-05,
      "loss": 0.1153,
      "step": 52440
    },
    {
      "epoch": 20.37,
      "learning_rate": 1.7284142394822007e-05,
      "loss": 0.2259,
      "step": 52450
    },
    {
      "epoch": 20.37,
      "learning_rate": 1.728362459546926e-05,
      "loss": 0.1518,
      "step": 52460
    },
    {
      "epoch": 20.38,
      "learning_rate": 1.7283106796116506e-05,
      "loss": 0.1179,
      "step": 52470
    },
    {
      "epoch": 20.38,
      "learning_rate": 1.7282588996763758e-05,
      "loss": 0.0508,
      "step": 52480
    },
    {
      "epoch": 20.38,
      "learning_rate": 1.7282071197411005e-05,
      "loss": 0.073,
      "step": 52490
    },
    {
      "epoch": 20.39,
      "learning_rate": 1.7281553398058253e-05,
      "loss": 0.1103,
      "step": 52500
    },
    {
      "epoch": 20.39,
      "learning_rate": 1.72810355987055e-05,
      "loss": 0.1284,
      "step": 52510
    },
    {
      "epoch": 20.4,
      "learning_rate": 1.7280517799352753e-05,
      "loss": 0.2534,
      "step": 52520
    },
    {
      "epoch": 20.4,
      "learning_rate": 1.728e-05,
      "loss": 0.0707,
      "step": 52530
    },
    {
      "epoch": 20.4,
      "learning_rate": 1.7279482200647252e-05,
      "loss": 0.2195,
      "step": 52540
    },
    {
      "epoch": 20.41,
      "learning_rate": 1.72789644012945e-05,
      "loss": 0.1654,
      "step": 52550
    },
    {
      "epoch": 20.41,
      "learning_rate": 1.727844660194175e-05,
      "loss": 0.0739,
      "step": 52560
    },
    {
      "epoch": 20.42,
      "learning_rate": 1.7277928802589e-05,
      "loss": 0.0841,
      "step": 52570
    },
    {
      "epoch": 20.42,
      "learning_rate": 1.7277411003236247e-05,
      "loss": 0.1261,
      "step": 52580
    },
    {
      "epoch": 20.42,
      "learning_rate": 1.7276893203883495e-05,
      "loss": 0.0644,
      "step": 52590
    },
    {
      "epoch": 20.43,
      "learning_rate": 1.7276375404530746e-05,
      "loss": 0.1813,
      "step": 52600
    },
    {
      "epoch": 20.43,
      "learning_rate": 1.7275857605177994e-05,
      "loss": 0.2338,
      "step": 52610
    },
    {
      "epoch": 20.43,
      "learning_rate": 1.7275339805825245e-05,
      "loss": 0.0836,
      "step": 52620
    },
    {
      "epoch": 20.44,
      "learning_rate": 1.7274822006472493e-05,
      "loss": 0.1102,
      "step": 52630
    },
    {
      "epoch": 20.44,
      "learning_rate": 1.7274304207119745e-05,
      "loss": 0.02,
      "step": 52640
    },
    {
      "epoch": 20.45,
      "learning_rate": 1.7273786407766993e-05,
      "loss": 0.0094,
      "step": 52650
    },
    {
      "epoch": 20.45,
      "learning_rate": 1.727326860841424e-05,
      "loss": 0.2096,
      "step": 52660
    },
    {
      "epoch": 20.45,
      "learning_rate": 1.727275080906149e-05,
      "loss": 0.056,
      "step": 52670
    },
    {
      "epoch": 20.46,
      "learning_rate": 1.727223300970874e-05,
      "loss": 0.194,
      "step": 52680
    },
    {
      "epoch": 20.46,
      "learning_rate": 1.7271715210355988e-05,
      "loss": 0.1976,
      "step": 52690
    },
    {
      "epoch": 20.47,
      "learning_rate": 1.727119741100324e-05,
      "loss": 0.1871,
      "step": 52700
    },
    {
      "epoch": 20.47,
      "learning_rate": 1.7270679611650487e-05,
      "loss": 0.0886,
      "step": 52710
    },
    {
      "epoch": 20.47,
      "learning_rate": 1.7270161812297738e-05,
      "loss": 0.1816,
      "step": 52720
    },
    {
      "epoch": 20.48,
      "learning_rate": 1.7269644012944986e-05,
      "loss": 0.0225,
      "step": 52730
    },
    {
      "epoch": 20.48,
      "learning_rate": 1.7269126213592234e-05,
      "loss": 0.0559,
      "step": 52740
    },
    {
      "epoch": 20.49,
      "learning_rate": 1.7268608414239482e-05,
      "loss": 0.1423,
      "step": 52750
    },
    {
      "epoch": 20.49,
      "learning_rate": 1.7268090614886733e-05,
      "loss": 0.1207,
      "step": 52760
    },
    {
      "epoch": 20.49,
      "learning_rate": 1.726757281553398e-05,
      "loss": 0.1362,
      "step": 52770
    },
    {
      "epoch": 20.5,
      "learning_rate": 1.7267055016181233e-05,
      "loss": 0.1651,
      "step": 52780
    },
    {
      "epoch": 20.5,
      "learning_rate": 1.726653721682848e-05,
      "loss": 0.0464,
      "step": 52790
    },
    {
      "epoch": 20.5,
      "learning_rate": 1.726601941747573e-05,
      "loss": 0.0877,
      "step": 52800
    },
    {
      "epoch": 20.51,
      "learning_rate": 1.726550161812298e-05,
      "loss": 0.0662,
      "step": 52810
    },
    {
      "epoch": 20.51,
      "learning_rate": 1.7264983818770228e-05,
      "loss": 0.079,
      "step": 52820
    },
    {
      "epoch": 20.52,
      "learning_rate": 1.7264466019417476e-05,
      "loss": 0.2121,
      "step": 52830
    },
    {
      "epoch": 20.52,
      "learning_rate": 1.7263948220064727e-05,
      "loss": 0.0305,
      "step": 52840
    },
    {
      "epoch": 20.52,
      "learning_rate": 1.7263430420711975e-05,
      "loss": 0.0381,
      "step": 52850
    },
    {
      "epoch": 20.53,
      "learning_rate": 1.7262912621359226e-05,
      "loss": 0.0165,
      "step": 52860
    },
    {
      "epoch": 20.53,
      "learning_rate": 1.7262394822006474e-05,
      "loss": 0.071,
      "step": 52870
    },
    {
      "epoch": 20.54,
      "learning_rate": 1.7261877022653722e-05,
      "loss": 0.0843,
      "step": 52880
    },
    {
      "epoch": 20.54,
      "learning_rate": 1.7261359223300973e-05,
      "loss": 0.1922,
      "step": 52890
    },
    {
      "epoch": 20.54,
      "learning_rate": 1.726084142394822e-05,
      "loss": 0.3273,
      "step": 52900
    },
    {
      "epoch": 20.55,
      "learning_rate": 1.726032362459547e-05,
      "loss": 0.064,
      "step": 52910
    },
    {
      "epoch": 20.55,
      "learning_rate": 1.725980582524272e-05,
      "loss": 0.0848,
      "step": 52920
    },
    {
      "epoch": 20.56,
      "learning_rate": 1.725928802588997e-05,
      "loss": 0.0974,
      "step": 52930
    },
    {
      "epoch": 20.56,
      "learning_rate": 1.725877022653722e-05,
      "loss": 0.1886,
      "step": 52940
    },
    {
      "epoch": 20.56,
      "learning_rate": 1.7258252427184468e-05,
      "loss": 0.0222,
      "step": 52950
    },
    {
      "epoch": 20.57,
      "learning_rate": 1.7257734627831716e-05,
      "loss": 0.2862,
      "step": 52960
    },
    {
      "epoch": 20.57,
      "learning_rate": 1.7257216828478967e-05,
      "loss": 0.1052,
      "step": 52970
    },
    {
      "epoch": 20.57,
      "learning_rate": 1.7256699029126215e-05,
      "loss": 0.1577,
      "step": 52980
    },
    {
      "epoch": 20.58,
      "learning_rate": 1.7256181229773463e-05,
      "loss": 0.1558,
      "step": 52990
    },
    {
      "epoch": 20.58,
      "learning_rate": 1.7255663430420714e-05,
      "loss": 0.0089,
      "step": 53000
    },
    {
      "epoch": 20.59,
      "learning_rate": 1.7255145631067962e-05,
      "loss": 0.1298,
      "step": 53010
    },
    {
      "epoch": 20.59,
      "learning_rate": 1.7254627831715213e-05,
      "loss": 0.1453,
      "step": 53020
    },
    {
      "epoch": 20.59,
      "learning_rate": 1.725411003236246e-05,
      "loss": 0.1413,
      "step": 53030
    },
    {
      "epoch": 20.6,
      "learning_rate": 1.725359223300971e-05,
      "loss": 0.0606,
      "step": 53040
    },
    {
      "epoch": 20.6,
      "learning_rate": 1.725307443365696e-05,
      "loss": 0.1143,
      "step": 53050
    },
    {
      "epoch": 20.61,
      "learning_rate": 1.725255663430421e-05,
      "loss": 0.0192,
      "step": 53060
    },
    {
      "epoch": 20.61,
      "learning_rate": 1.7252038834951456e-05,
      "loss": 0.1011,
      "step": 53070
    },
    {
      "epoch": 20.61,
      "learning_rate": 1.7251521035598708e-05,
      "loss": 0.0171,
      "step": 53080
    },
    {
      "epoch": 20.62,
      "learning_rate": 1.7251003236245955e-05,
      "loss": 0.3304,
      "step": 53090
    },
    {
      "epoch": 20.62,
      "learning_rate": 1.7250485436893207e-05,
      "loss": 0.0536,
      "step": 53100
    },
    {
      "epoch": 20.63,
      "learning_rate": 1.7249967637540455e-05,
      "loss": 0.0653,
      "step": 53110
    },
    {
      "epoch": 20.63,
      "learning_rate": 1.7249449838187703e-05,
      "loss": 0.0682,
      "step": 53120
    },
    {
      "epoch": 20.63,
      "learning_rate": 1.7248932038834954e-05,
      "loss": 0.1384,
      "step": 53130
    },
    {
      "epoch": 20.64,
      "learning_rate": 1.7248414239482202e-05,
      "loss": 0.0683,
      "step": 53140
    },
    {
      "epoch": 20.64,
      "learning_rate": 1.724789644012945e-05,
      "loss": 0.0719,
      "step": 53150
    },
    {
      "epoch": 20.64,
      "learning_rate": 1.72473786407767e-05,
      "loss": 0.2187,
      "step": 53160
    },
    {
      "epoch": 20.65,
      "learning_rate": 1.724686084142395e-05,
      "loss": 0.1711,
      "step": 53170
    },
    {
      "epoch": 20.65,
      "learning_rate": 1.7246343042071197e-05,
      "loss": 0.1218,
      "step": 53180
    },
    {
      "epoch": 20.66,
      "learning_rate": 1.7245825242718448e-05,
      "loss": 0.1284,
      "step": 53190
    },
    {
      "epoch": 20.66,
      "learning_rate": 1.7245307443365696e-05,
      "loss": 0.0449,
      "step": 53200
    },
    {
      "epoch": 20.66,
      "learning_rate": 1.7244789644012947e-05,
      "loss": 0.0941,
      "step": 53210
    },
    {
      "epoch": 20.67,
      "learning_rate": 1.7244271844660195e-05,
      "loss": 0.089,
      "step": 53220
    },
    {
      "epoch": 20.67,
      "learning_rate": 1.7243754045307443e-05,
      "loss": 0.2326,
      "step": 53230
    },
    {
      "epoch": 20.68,
      "learning_rate": 1.7243236245954695e-05,
      "loss": 0.0099,
      "step": 53240
    },
    {
      "epoch": 20.68,
      "learning_rate": 1.7242718446601943e-05,
      "loss": 0.1499,
      "step": 53250
    },
    {
      "epoch": 20.68,
      "learning_rate": 1.724220064724919e-05,
      "loss": 0.0619,
      "step": 53260
    },
    {
      "epoch": 20.69,
      "learning_rate": 1.7241682847896442e-05,
      "loss": 0.094,
      "step": 53270
    },
    {
      "epoch": 20.69,
      "learning_rate": 1.724116504854369e-05,
      "loss": 0.1198,
      "step": 53280
    },
    {
      "epoch": 20.7,
      "learning_rate": 1.724064724919094e-05,
      "loss": 0.0321,
      "step": 53290
    },
    {
      "epoch": 20.7,
      "learning_rate": 1.724012944983819e-05,
      "loss": 0.1458,
      "step": 53300
    },
    {
      "epoch": 20.7,
      "learning_rate": 1.723961165048544e-05,
      "loss": 0.1425,
      "step": 53310
    },
    {
      "epoch": 20.71,
      "learning_rate": 1.7239093851132688e-05,
      "loss": 0.223,
      "step": 53320
    },
    {
      "epoch": 20.71,
      "learning_rate": 1.7238576051779936e-05,
      "loss": 0.1273,
      "step": 53330
    },
    {
      "epoch": 20.71,
      "learning_rate": 1.7238058252427184e-05,
      "loss": 0.2332,
      "step": 53340
    },
    {
      "epoch": 20.72,
      "learning_rate": 1.7237540453074435e-05,
      "loss": 0.0277,
      "step": 53350
    },
    {
      "epoch": 20.72,
      "learning_rate": 1.7237022653721683e-05,
      "loss": 0.1274,
      "step": 53360
    },
    {
      "epoch": 20.73,
      "learning_rate": 1.7236504854368935e-05,
      "loss": 0.1231,
      "step": 53370
    },
    {
      "epoch": 20.73,
      "learning_rate": 1.7235987055016183e-05,
      "loss": 0.1353,
      "step": 53380
    },
    {
      "epoch": 20.73,
      "learning_rate": 1.7235469255663434e-05,
      "loss": 0.191,
      "step": 53390
    },
    {
      "epoch": 20.74,
      "learning_rate": 1.7234951456310682e-05,
      "loss": 0.1519,
      "step": 53400
    },
    {
      "epoch": 20.74,
      "learning_rate": 1.723443365695793e-05,
      "loss": 0.0325,
      "step": 53410
    },
    {
      "epoch": 20.75,
      "learning_rate": 1.7233915857605178e-05,
      "loss": 0.1875,
      "step": 53420
    },
    {
      "epoch": 20.75,
      "learning_rate": 1.723339805825243e-05,
      "loss": 0.0786,
      "step": 53430
    },
    {
      "epoch": 20.75,
      "learning_rate": 1.7232880258899677e-05,
      "loss": 0.119,
      "step": 53440
    },
    {
      "epoch": 20.76,
      "learning_rate": 1.7232362459546928e-05,
      "loss": 0.1078,
      "step": 53450
    },
    {
      "epoch": 20.76,
      "learning_rate": 1.7231844660194176e-05,
      "loss": 0.1561,
      "step": 53460
    },
    {
      "epoch": 20.77,
      "learning_rate": 1.7231326860841427e-05,
      "loss": 0.2179,
      "step": 53470
    },
    {
      "epoch": 20.77,
      "learning_rate": 1.7230809061488672e-05,
      "loss": 0.0132,
      "step": 53480
    },
    {
      "epoch": 20.77,
      "learning_rate": 1.7230291262135923e-05,
      "loss": 0.0332,
      "step": 53490
    },
    {
      "epoch": 20.78,
      "learning_rate": 1.722977346278317e-05,
      "loss": 0.2199,
      "step": 53500
    },
    {
      "epoch": 20.78,
      "learning_rate": 1.7229255663430422e-05,
      "loss": 0.1592,
      "step": 53510
    },
    {
      "epoch": 20.78,
      "learning_rate": 1.722873786407767e-05,
      "loss": 0.0613,
      "step": 53520
    },
    {
      "epoch": 20.79,
      "learning_rate": 1.7228220064724922e-05,
      "loss": 0.0638,
      "step": 53530
    },
    {
      "epoch": 20.79,
      "learning_rate": 1.722770226537217e-05,
      "loss": 0.0222,
      "step": 53540
    },
    {
      "epoch": 20.8,
      "learning_rate": 1.722718446601942e-05,
      "loss": 0.1946,
      "step": 53550
    },
    {
      "epoch": 20.8,
      "learning_rate": 1.7226666666666665e-05,
      "loss": 0.2062,
      "step": 53560
    },
    {
      "epoch": 20.8,
      "learning_rate": 1.7226148867313917e-05,
      "loss": 0.1673,
      "step": 53570
    },
    {
      "epoch": 20.81,
      "learning_rate": 1.7225631067961165e-05,
      "loss": 0.0688,
      "step": 53580
    },
    {
      "epoch": 20.81,
      "learning_rate": 1.7225113268608416e-05,
      "loss": 0.1346,
      "step": 53590
    },
    {
      "epoch": 20.82,
      "learning_rate": 1.7224595469255664e-05,
      "loss": 0.0634,
      "step": 53600
    },
    {
      "epoch": 20.82,
      "learning_rate": 1.7224077669902915e-05,
      "loss": 0.1253,
      "step": 53610
    },
    {
      "epoch": 20.82,
      "learning_rate": 1.7223559870550163e-05,
      "loss": 0.1076,
      "step": 53620
    },
    {
      "epoch": 20.83,
      "learning_rate": 1.7223042071197415e-05,
      "loss": 0.0792,
      "step": 53630
    },
    {
      "epoch": 20.83,
      "learning_rate": 1.722252427184466e-05,
      "loss": 0.0701,
      "step": 53640
    },
    {
      "epoch": 20.83,
      "learning_rate": 1.722200647249191e-05,
      "loss": 0.0174,
      "step": 53650
    },
    {
      "epoch": 20.84,
      "learning_rate": 1.7221488673139158e-05,
      "loss": 0.2389,
      "step": 53660
    },
    {
      "epoch": 20.84,
      "learning_rate": 1.722097087378641e-05,
      "loss": 0.0127,
      "step": 53670
    },
    {
      "epoch": 20.85,
      "learning_rate": 1.7220453074433658e-05,
      "loss": 0.0954,
      "step": 53680
    },
    {
      "epoch": 20.85,
      "learning_rate": 1.721993527508091e-05,
      "loss": 0.1071,
      "step": 53690
    },
    {
      "epoch": 20.85,
      "learning_rate": 1.7219417475728157e-05,
      "loss": 0.1094,
      "step": 53700
    },
    {
      "epoch": 20.86,
      "learning_rate": 1.7218899676375408e-05,
      "loss": 0.2007,
      "step": 53710
    },
    {
      "epoch": 20.86,
      "learning_rate": 1.7218381877022653e-05,
      "loss": 0.1113,
      "step": 53720
    },
    {
      "epoch": 20.87,
      "learning_rate": 1.7217864077669904e-05,
      "loss": 0.1856,
      "step": 53730
    },
    {
      "epoch": 20.87,
      "learning_rate": 1.7217346278317152e-05,
      "loss": 0.0138,
      "step": 53740
    },
    {
      "epoch": 20.87,
      "learning_rate": 1.7216828478964403e-05,
      "loss": 0.0971,
      "step": 53750
    },
    {
      "epoch": 20.88,
      "learning_rate": 1.721631067961165e-05,
      "loss": 0.1506,
      "step": 53760
    },
    {
      "epoch": 20.88,
      "learning_rate": 1.7215792880258902e-05,
      "loss": 0.0276,
      "step": 53770
    },
    {
      "epoch": 20.89,
      "learning_rate": 1.721527508090615e-05,
      "loss": 0.0198,
      "step": 53780
    },
    {
      "epoch": 20.89,
      "learning_rate": 1.72147572815534e-05,
      "loss": 0.0372,
      "step": 53790
    },
    {
      "epoch": 20.89,
      "learning_rate": 1.7214239482200646e-05,
      "loss": 0.1151,
      "step": 53800
    },
    {
      "epoch": 20.9,
      "learning_rate": 1.7213721682847897e-05,
      "loss": 0.0435,
      "step": 53810
    },
    {
      "epoch": 20.9,
      "learning_rate": 1.7213203883495145e-05,
      "loss": 0.0989,
      "step": 53820
    },
    {
      "epoch": 20.9,
      "learning_rate": 1.7212686084142397e-05,
      "loss": 0.144,
      "step": 53830
    },
    {
      "epoch": 20.91,
      "learning_rate": 1.7212168284789645e-05,
      "loss": 0.1142,
      "step": 53840
    },
    {
      "epoch": 20.91,
      "learning_rate": 1.7211650485436896e-05,
      "loss": 0.0974,
      "step": 53850
    },
    {
      "epoch": 20.92,
      "learning_rate": 1.7211132686084144e-05,
      "loss": 0.1335,
      "step": 53860
    },
    {
      "epoch": 20.92,
      "learning_rate": 1.7210614886731395e-05,
      "loss": 0.0046,
      "step": 53870
    },
    {
      "epoch": 20.92,
      "learning_rate": 1.7210097087378643e-05,
      "loss": 0.13,
      "step": 53880
    },
    {
      "epoch": 20.93,
      "learning_rate": 1.720957928802589e-05,
      "loss": 0.0353,
      "step": 53890
    },
    {
      "epoch": 20.93,
      "learning_rate": 1.720906148867314e-05,
      "loss": 0.0903,
      "step": 53900
    },
    {
      "epoch": 20.94,
      "learning_rate": 1.720854368932039e-05,
      "loss": 0.0495,
      "step": 53910
    },
    {
      "epoch": 20.94,
      "learning_rate": 1.7208025889967638e-05,
      "loss": 0.2099,
      "step": 53920
    },
    {
      "epoch": 20.94,
      "learning_rate": 1.720750809061489e-05,
      "loss": 0.0903,
      "step": 53930
    },
    {
      "epoch": 20.95,
      "learning_rate": 1.7206990291262137e-05,
      "loss": 0.1635,
      "step": 53940
    },
    {
      "epoch": 20.95,
      "learning_rate": 1.720647249190939e-05,
      "loss": 0.1063,
      "step": 53950
    },
    {
      "epoch": 20.96,
      "learning_rate": 1.7205954692556637e-05,
      "loss": 0.0585,
      "step": 53960
    },
    {
      "epoch": 20.96,
      "learning_rate": 1.7205436893203885e-05,
      "loss": 0.0919,
      "step": 53970
    },
    {
      "epoch": 20.96,
      "learning_rate": 1.7204919093851133e-05,
      "loss": 0.1721,
      "step": 53980
    },
    {
      "epoch": 20.97,
      "learning_rate": 1.7204401294498384e-05,
      "loss": 0.0547,
      "step": 53990
    },
    {
      "epoch": 20.97,
      "learning_rate": 1.7203883495145632e-05,
      "loss": 0.1883,
      "step": 54000
    },
    {
      "epoch": 20.97,
      "learning_rate": 1.7203365695792883e-05,
      "loss": 0.2006,
      "step": 54010
    },
    {
      "epoch": 20.98,
      "learning_rate": 1.720284789644013e-05,
      "loss": 0.2289,
      "step": 54020
    },
    {
      "epoch": 20.98,
      "learning_rate": 1.7202330097087382e-05,
      "loss": 0.0398,
      "step": 54030
    },
    {
      "epoch": 20.99,
      "learning_rate": 1.720181229773463e-05,
      "loss": 0.039,
      "step": 54040
    },
    {
      "epoch": 20.99,
      "learning_rate": 1.7201294498381878e-05,
      "loss": 0.0833,
      "step": 54050
    },
    {
      "epoch": 20.99,
      "learning_rate": 1.7200776699029126e-05,
      "loss": 0.0038,
      "step": 54060
    },
    {
      "epoch": 21.0,
      "learning_rate": 1.7200258899676377e-05,
      "loss": 0.1441,
      "step": 54070
    },
    {
      "epoch": 21.0,
      "eval_accuracy": 0.9535075653370014,
      "eval_loss": 0.24637892842292786,
      "eval_runtime": 8.3191,
      "eval_samples_per_second": 436.948,
      "eval_steps_per_second": 54.694,
      "step": 54075
    },
    {
      "epoch": 21.0,
      "learning_rate": 1.7199741100323625e-05,
      "loss": 0.0572,
      "step": 54080
    },
    {
      "epoch": 21.01,
      "learning_rate": 1.7199223300970877e-05,
      "loss": 0.0622,
      "step": 54090
    },
    {
      "epoch": 21.01,
      "learning_rate": 1.7198705501618125e-05,
      "loss": 0.0795,
      "step": 54100
    },
    {
      "epoch": 21.01,
      "learning_rate": 1.7198187702265376e-05,
      "loss": 0.1797,
      "step": 54110
    },
    {
      "epoch": 21.02,
      "learning_rate": 1.7197669902912624e-05,
      "loss": 0.1026,
      "step": 54120
    },
    {
      "epoch": 21.02,
      "learning_rate": 1.7197152103559872e-05,
      "loss": 0.0673,
      "step": 54130
    },
    {
      "epoch": 21.03,
      "learning_rate": 1.719663430420712e-05,
      "loss": 0.0365,
      "step": 54140
    },
    {
      "epoch": 21.03,
      "learning_rate": 1.719611650485437e-05,
      "loss": 0.0678,
      "step": 54150
    },
    {
      "epoch": 21.03,
      "learning_rate": 1.719559870550162e-05,
      "loss": 0.1162,
      "step": 54160
    },
    {
      "epoch": 21.04,
      "learning_rate": 1.719508090614887e-05,
      "loss": 0.106,
      "step": 54170
    },
    {
      "epoch": 21.04,
      "learning_rate": 1.7194563106796118e-05,
      "loss": 0.0695,
      "step": 54180
    },
    {
      "epoch": 21.04,
      "learning_rate": 1.719404530744337e-05,
      "loss": 0.1746,
      "step": 54190
    },
    {
      "epoch": 21.05,
      "learning_rate": 1.7193527508090617e-05,
      "loss": 0.0216,
      "step": 54200
    },
    {
      "epoch": 21.05,
      "learning_rate": 1.7193009708737865e-05,
      "loss": 0.0578,
      "step": 54210
    },
    {
      "epoch": 21.06,
      "learning_rate": 1.7192491909385113e-05,
      "loss": 0.1653,
      "step": 54220
    },
    {
      "epoch": 21.06,
      "learning_rate": 1.7191974110032364e-05,
      "loss": 0.0185,
      "step": 54230
    },
    {
      "epoch": 21.06,
      "learning_rate": 1.7191456310679612e-05,
      "loss": 0.201,
      "step": 54240
    },
    {
      "epoch": 21.07,
      "learning_rate": 1.7190938511326864e-05,
      "loss": 0.0463,
      "step": 54250
    },
    {
      "epoch": 21.07,
      "learning_rate": 1.719042071197411e-05,
      "loss": 0.1246,
      "step": 54260
    },
    {
      "epoch": 21.08,
      "learning_rate": 1.718990291262136e-05,
      "loss": 0.164,
      "step": 54270
    },
    {
      "epoch": 21.08,
      "learning_rate": 1.718938511326861e-05,
      "loss": 0.0791,
      "step": 54280
    },
    {
      "epoch": 21.08,
      "learning_rate": 1.718886731391586e-05,
      "loss": 0.0706,
      "step": 54290
    },
    {
      "epoch": 21.09,
      "learning_rate": 1.7188349514563107e-05,
      "loss": 0.0284,
      "step": 54300
    },
    {
      "epoch": 21.09,
      "learning_rate": 1.7187831715210358e-05,
      "loss": 0.001,
      "step": 54310
    },
    {
      "epoch": 21.1,
      "learning_rate": 1.7187313915857606e-05,
      "loss": 0.096,
      "step": 54320
    },
    {
      "epoch": 21.1,
      "learning_rate": 1.7186796116504857e-05,
      "loss": 0.0586,
      "step": 54330
    },
    {
      "epoch": 21.1,
      "learning_rate": 1.7186278317152105e-05,
      "loss": 0.0814,
      "step": 54340
    },
    {
      "epoch": 21.11,
      "learning_rate": 1.7185760517799353e-05,
      "loss": 0.2206,
      "step": 54350
    },
    {
      "epoch": 21.11,
      "learning_rate": 1.7185242718446604e-05,
      "loss": 0.1949,
      "step": 54360
    },
    {
      "epoch": 21.11,
      "learning_rate": 1.7184724919093852e-05,
      "loss": 0.2117,
      "step": 54370
    },
    {
      "epoch": 21.12,
      "learning_rate": 1.71842071197411e-05,
      "loss": 0.0407,
      "step": 54380
    },
    {
      "epoch": 21.12,
      "learning_rate": 1.718368932038835e-05,
      "loss": 0.1158,
      "step": 54390
    },
    {
      "epoch": 21.13,
      "learning_rate": 1.71831715210356e-05,
      "loss": 0.0969,
      "step": 54400
    },
    {
      "epoch": 21.13,
      "learning_rate": 1.718265372168285e-05,
      "loss": 0.1542,
      "step": 54410
    },
    {
      "epoch": 21.13,
      "learning_rate": 1.71821359223301e-05,
      "loss": 0.1363,
      "step": 54420
    },
    {
      "epoch": 21.14,
      "learning_rate": 1.7181618122977347e-05,
      "loss": 0.0432,
      "step": 54430
    },
    {
      "epoch": 21.14,
      "learning_rate": 1.7181100323624598e-05,
      "loss": 0.0786,
      "step": 54440
    },
    {
      "epoch": 21.15,
      "learning_rate": 1.7180582524271846e-05,
      "loss": 0.0921,
      "step": 54450
    },
    {
      "epoch": 21.15,
      "learning_rate": 1.7180064724919094e-05,
      "loss": 0.0404,
      "step": 54460
    },
    {
      "epoch": 21.15,
      "learning_rate": 1.7179546925566345e-05,
      "loss": 0.0746,
      "step": 54470
    },
    {
      "epoch": 21.16,
      "learning_rate": 1.7179029126213593e-05,
      "loss": 0.075,
      "step": 54480
    },
    {
      "epoch": 21.16,
      "learning_rate": 1.7178511326860844e-05,
      "loss": 0.0497,
      "step": 54490
    },
    {
      "epoch": 21.17,
      "learning_rate": 1.7177993527508092e-05,
      "loss": 0.2104,
      "step": 54500
    },
    {
      "epoch": 21.17,
      "learning_rate": 1.717747572815534e-05,
      "loss": 0.1068,
      "step": 54510
    },
    {
      "epoch": 21.17,
      "learning_rate": 1.717695792880259e-05,
      "loss": 0.0662,
      "step": 54520
    },
    {
      "epoch": 21.18,
      "learning_rate": 1.717644012944984e-05,
      "loss": 0.0811,
      "step": 54530
    },
    {
      "epoch": 21.18,
      "learning_rate": 1.7175922330097087e-05,
      "loss": 0.0535,
      "step": 54540
    },
    {
      "epoch": 21.18,
      "learning_rate": 1.717540453074434e-05,
      "loss": 0.1135,
      "step": 54550
    },
    {
      "epoch": 21.19,
      "learning_rate": 1.7174886731391587e-05,
      "loss": 0.1646,
      "step": 54560
    },
    {
      "epoch": 21.19,
      "learning_rate": 1.7174368932038838e-05,
      "loss": 0.0235,
      "step": 54570
    },
    {
      "epoch": 21.2,
      "learning_rate": 1.7173851132686086e-05,
      "loss": 0.1694,
      "step": 54580
    },
    {
      "epoch": 21.2,
      "learning_rate": 1.7173333333333334e-05,
      "loss": 0.0707,
      "step": 54590
    },
    {
      "epoch": 21.2,
      "learning_rate": 1.7172815533980585e-05,
      "loss": 0.0241,
      "step": 54600
    },
    {
      "epoch": 21.21,
      "learning_rate": 1.7172297734627833e-05,
      "loss": 0.1197,
      "step": 54610
    },
    {
      "epoch": 21.21,
      "learning_rate": 1.717177993527508e-05,
      "loss": 0.155,
      "step": 54620
    },
    {
      "epoch": 21.22,
      "learning_rate": 1.7171262135922332e-05,
      "loss": 0.1072,
      "step": 54630
    },
    {
      "epoch": 21.22,
      "learning_rate": 1.717074433656958e-05,
      "loss": 0.0706,
      "step": 54640
    },
    {
      "epoch": 21.22,
      "learning_rate": 1.7170226537216828e-05,
      "loss": 0.1033,
      "step": 54650
    },
    {
      "epoch": 21.23,
      "learning_rate": 1.716970873786408e-05,
      "loss": 0.1044,
      "step": 54660
    },
    {
      "epoch": 21.23,
      "learning_rate": 1.7169190938511327e-05,
      "loss": 0.0384,
      "step": 54670
    },
    {
      "epoch": 21.23,
      "learning_rate": 1.716867313915858e-05,
      "loss": 0.0771,
      "step": 54680
    },
    {
      "epoch": 21.24,
      "learning_rate": 1.7168155339805827e-05,
      "loss": 0.0795,
      "step": 54690
    },
    {
      "epoch": 21.24,
      "learning_rate": 1.7167637540453075e-05,
      "loss": 0.322,
      "step": 54700
    },
    {
      "epoch": 21.25,
      "learning_rate": 1.7167119741100326e-05,
      "loss": 0.0858,
      "step": 54710
    },
    {
      "epoch": 21.25,
      "learning_rate": 1.7166601941747574e-05,
      "loss": 0.1502,
      "step": 54720
    },
    {
      "epoch": 21.25,
      "learning_rate": 1.716608414239482e-05,
      "loss": 0.0795,
      "step": 54730
    },
    {
      "epoch": 21.26,
      "learning_rate": 1.7165566343042073e-05,
      "loss": 0.0275,
      "step": 54740
    },
    {
      "epoch": 21.26,
      "learning_rate": 1.716504854368932e-05,
      "loss": 0.2038,
      "step": 54750
    },
    {
      "epoch": 21.27,
      "learning_rate": 1.7164530744336572e-05,
      "loss": 0.2753,
      "step": 54760
    },
    {
      "epoch": 21.27,
      "learning_rate": 1.716401294498382e-05,
      "loss": 0.0944,
      "step": 54770
    },
    {
      "epoch": 21.27,
      "learning_rate": 1.7163495145631068e-05,
      "loss": 0.0654,
      "step": 54780
    },
    {
      "epoch": 21.28,
      "learning_rate": 1.716297734627832e-05,
      "loss": 0.126,
      "step": 54790
    },
    {
      "epoch": 21.28,
      "learning_rate": 1.7162459546925567e-05,
      "loss": 0.1474,
      "step": 54800
    },
    {
      "epoch": 21.29,
      "learning_rate": 1.7161941747572815e-05,
      "loss": 0.0384,
      "step": 54810
    },
    {
      "epoch": 21.29,
      "learning_rate": 1.7161423948220067e-05,
      "loss": 0.1351,
      "step": 54820
    },
    {
      "epoch": 21.29,
      "learning_rate": 1.7160906148867314e-05,
      "loss": 0.0447,
      "step": 54830
    },
    {
      "epoch": 21.3,
      "learning_rate": 1.7160388349514566e-05,
      "loss": 0.0885,
      "step": 54840
    },
    {
      "epoch": 21.3,
      "learning_rate": 1.7159870550161814e-05,
      "loss": 0.1213,
      "step": 54850
    },
    {
      "epoch": 21.3,
      "learning_rate": 1.715935275080906e-05,
      "loss": 0.1524,
      "step": 54860
    },
    {
      "epoch": 21.31,
      "learning_rate": 1.7158834951456313e-05,
      "loss": 0.0191,
      "step": 54870
    },
    {
      "epoch": 21.31,
      "learning_rate": 1.715831715210356e-05,
      "loss": 0.1977,
      "step": 54880
    },
    {
      "epoch": 21.32,
      "learning_rate": 1.715779935275081e-05,
      "loss": 0.0336,
      "step": 54890
    },
    {
      "epoch": 21.32,
      "learning_rate": 1.715728155339806e-05,
      "loss": 0.153,
      "step": 54900
    },
    {
      "epoch": 21.32,
      "learning_rate": 1.7156763754045308e-05,
      "loss": 0.1692,
      "step": 54910
    },
    {
      "epoch": 21.33,
      "learning_rate": 1.715624595469256e-05,
      "loss": 0.1391,
      "step": 54920
    },
    {
      "epoch": 21.33,
      "learning_rate": 1.7155728155339807e-05,
      "loss": 0.1667,
      "step": 54930
    },
    {
      "epoch": 21.34,
      "learning_rate": 1.715521035598706e-05,
      "loss": 0.046,
      "step": 54940
    },
    {
      "epoch": 21.34,
      "learning_rate": 1.7154692556634303e-05,
      "loss": 0.1033,
      "step": 54950
    },
    {
      "epoch": 21.34,
      "learning_rate": 1.7154174757281554e-05,
      "loss": 0.2567,
      "step": 54960
    },
    {
      "epoch": 21.35,
      "learning_rate": 1.7153656957928802e-05,
      "loss": 0.2298,
      "step": 54970
    },
    {
      "epoch": 21.35,
      "learning_rate": 1.7153139158576054e-05,
      "loss": 0.3167,
      "step": 54980
    },
    {
      "epoch": 21.36,
      "learning_rate": 1.71526213592233e-05,
      "loss": 0.1198,
      "step": 54990
    },
    {
      "epoch": 21.36,
      "learning_rate": 1.7152103559870553e-05,
      "loss": 0.0241,
      "step": 55000
    },
    {
      "epoch": 21.36,
      "learning_rate": 1.71515857605178e-05,
      "loss": 0.2502,
      "step": 55010
    },
    {
      "epoch": 21.37,
      "learning_rate": 1.7151067961165052e-05,
      "loss": 0.1034,
      "step": 55020
    },
    {
      "epoch": 21.37,
      "learning_rate": 1.7150550161812297e-05,
      "loss": 0.1133,
      "step": 55030
    },
    {
      "epoch": 21.37,
      "learning_rate": 1.7150032362459548e-05,
      "loss": 0.0762,
      "step": 55040
    },
    {
      "epoch": 21.38,
      "learning_rate": 1.7149514563106796e-05,
      "loss": 0.0366,
      "step": 55050
    },
    {
      "epoch": 21.38,
      "learning_rate": 1.7148996763754047e-05,
      "loss": 0.0397,
      "step": 55060
    },
    {
      "epoch": 21.39,
      "learning_rate": 1.7148478964401295e-05,
      "loss": 0.0419,
      "step": 55070
    },
    {
      "epoch": 21.39,
      "learning_rate": 1.7147961165048546e-05,
      "loss": 0.0258,
      "step": 55080
    },
    {
      "epoch": 21.39,
      "learning_rate": 1.7147443365695794e-05,
      "loss": 0.1835,
      "step": 55090
    },
    {
      "epoch": 21.4,
      "learning_rate": 1.7146925566343046e-05,
      "loss": 0.0387,
      "step": 55100
    },
    {
      "epoch": 21.4,
      "learning_rate": 1.714640776699029e-05,
      "loss": 0.0926,
      "step": 55110
    },
    {
      "epoch": 21.41,
      "learning_rate": 1.714588996763754e-05,
      "loss": 0.1444,
      "step": 55120
    },
    {
      "epoch": 21.41,
      "learning_rate": 1.714537216828479e-05,
      "loss": 0.0609,
      "step": 55130
    },
    {
      "epoch": 21.41,
      "learning_rate": 1.714485436893204e-05,
      "loss": 0.1269,
      "step": 55140
    },
    {
      "epoch": 21.42,
      "learning_rate": 1.714433656957929e-05,
      "loss": 0.0858,
      "step": 55150
    },
    {
      "epoch": 21.42,
      "learning_rate": 1.714381877022654e-05,
      "loss": 0.0772,
      "step": 55160
    },
    {
      "epoch": 21.43,
      "learning_rate": 1.7143300970873788e-05,
      "loss": 0.0914,
      "step": 55170
    },
    {
      "epoch": 21.43,
      "learning_rate": 1.714278317152104e-05,
      "loss": 0.1835,
      "step": 55180
    },
    {
      "epoch": 21.43,
      "learning_rate": 1.7142265372168284e-05,
      "loss": 0.1271,
      "step": 55190
    },
    {
      "epoch": 21.44,
      "learning_rate": 1.7141747572815535e-05,
      "loss": 0.0676,
      "step": 55200
    },
    {
      "epoch": 21.44,
      "learning_rate": 1.7141229773462783e-05,
      "loss": 0.0363,
      "step": 55210
    },
    {
      "epoch": 21.44,
      "learning_rate": 1.7140711974110034e-05,
      "loss": 0.0981,
      "step": 55220
    },
    {
      "epoch": 21.45,
      "learning_rate": 1.7140194174757282e-05,
      "loss": 0.1268,
      "step": 55230
    },
    {
      "epoch": 21.45,
      "learning_rate": 1.7139676375404534e-05,
      "loss": 0.1077,
      "step": 55240
    },
    {
      "epoch": 21.46,
      "learning_rate": 1.713915857605178e-05,
      "loss": 0.1698,
      "step": 55250
    },
    {
      "epoch": 21.46,
      "learning_rate": 1.7138640776699033e-05,
      "loss": 0.1275,
      "step": 55260
    },
    {
      "epoch": 21.46,
      "learning_rate": 1.7138122977346277e-05,
      "loss": 0.1187,
      "step": 55270
    },
    {
      "epoch": 21.47,
      "learning_rate": 1.713760517799353e-05,
      "loss": 0.2423,
      "step": 55280
    },
    {
      "epoch": 21.47,
      "learning_rate": 1.7137087378640777e-05,
      "loss": 0.1485,
      "step": 55290
    },
    {
      "epoch": 21.48,
      "learning_rate": 1.7136569579288028e-05,
      "loss": 0.234,
      "step": 55300
    },
    {
      "epoch": 21.48,
      "learning_rate": 1.7136051779935276e-05,
      "loss": 0.1527,
      "step": 55310
    },
    {
      "epoch": 21.48,
      "learning_rate": 1.7135533980582527e-05,
      "loss": 0.1034,
      "step": 55320
    },
    {
      "epoch": 21.49,
      "learning_rate": 1.7135016181229775e-05,
      "loss": 0.0304,
      "step": 55330
    },
    {
      "epoch": 21.49,
      "learning_rate": 1.7134498381877026e-05,
      "loss": 0.1296,
      "step": 55340
    },
    {
      "epoch": 21.5,
      "learning_rate": 1.713398058252427e-05,
      "loss": 0.0273,
      "step": 55350
    },
    {
      "epoch": 21.5,
      "learning_rate": 1.7133462783171522e-05,
      "loss": 0.0918,
      "step": 55360
    },
    {
      "epoch": 21.5,
      "learning_rate": 1.713294498381877e-05,
      "loss": 0.0659,
      "step": 55370
    },
    {
      "epoch": 21.51,
      "learning_rate": 1.713242718446602e-05,
      "loss": 0.0087,
      "step": 55380
    },
    {
      "epoch": 21.51,
      "learning_rate": 1.713190938511327e-05,
      "loss": 0.1006,
      "step": 55390
    },
    {
      "epoch": 21.51,
      "learning_rate": 1.713139158576052e-05,
      "loss": 0.1827,
      "step": 55400
    },
    {
      "epoch": 21.52,
      "learning_rate": 1.713087378640777e-05,
      "loss": 0.0121,
      "step": 55410
    },
    {
      "epoch": 21.52,
      "learning_rate": 1.713035598705502e-05,
      "loss": 0.2817,
      "step": 55420
    },
    {
      "epoch": 21.53,
      "learning_rate": 1.7129838187702264e-05,
      "loss": 0.105,
      "step": 55430
    },
    {
      "epoch": 21.53,
      "learning_rate": 1.7129320388349516e-05,
      "loss": 0.132,
      "step": 55440
    },
    {
      "epoch": 21.53,
      "learning_rate": 1.7128802588996764e-05,
      "loss": 0.1575,
      "step": 55450
    },
    {
      "epoch": 21.54,
      "learning_rate": 1.7128284789644015e-05,
      "loss": 0.3516,
      "step": 55460
    },
    {
      "epoch": 21.54,
      "learning_rate": 1.7127766990291263e-05,
      "loss": 0.0895,
      "step": 55470
    },
    {
      "epoch": 21.55,
      "learning_rate": 1.7127249190938514e-05,
      "loss": 0.0385,
      "step": 55480
    },
    {
      "epoch": 21.55,
      "learning_rate": 1.7126731391585762e-05,
      "loss": 0.0981,
      "step": 55490
    },
    {
      "epoch": 21.55,
      "learning_rate": 1.7126213592233013e-05,
      "loss": 0.0832,
      "step": 55500
    },
    {
      "epoch": 21.56,
      "learning_rate": 1.712569579288026e-05,
      "loss": 0.0416,
      "step": 55510
    },
    {
      "epoch": 21.56,
      "learning_rate": 1.712517799352751e-05,
      "loss": 0.1029,
      "step": 55520
    },
    {
      "epoch": 21.57,
      "learning_rate": 1.7124660194174757e-05,
      "loss": 0.0724,
      "step": 55530
    },
    {
      "epoch": 21.57,
      "learning_rate": 1.712414239482201e-05,
      "loss": 0.2461,
      "step": 55540
    },
    {
      "epoch": 21.57,
      "learning_rate": 1.7123624595469256e-05,
      "loss": 0.1188,
      "step": 55550
    },
    {
      "epoch": 21.58,
      "learning_rate": 1.7123106796116508e-05,
      "loss": 0.0732,
      "step": 55560
    },
    {
      "epoch": 21.58,
      "learning_rate": 1.7122588996763756e-05,
      "loss": 0.1311,
      "step": 55570
    },
    {
      "epoch": 21.58,
      "learning_rate": 1.7122071197411007e-05,
      "loss": 0.0857,
      "step": 55580
    },
    {
      "epoch": 21.59,
      "learning_rate": 1.7121553398058255e-05,
      "loss": 0.1913,
      "step": 55590
    },
    {
      "epoch": 21.59,
      "learning_rate": 1.7121035598705503e-05,
      "loss": 0.1128,
      "step": 55600
    },
    {
      "epoch": 21.6,
      "learning_rate": 1.712051779935275e-05,
      "loss": 0.0664,
      "step": 55610
    },
    {
      "epoch": 21.6,
      "learning_rate": 1.7120000000000002e-05,
      "loss": 0.1148,
      "step": 55620
    },
    {
      "epoch": 21.6,
      "learning_rate": 1.711948220064725e-05,
      "loss": 0.0542,
      "step": 55630
    },
    {
      "epoch": 21.61,
      "learning_rate": 1.71189644012945e-05,
      "loss": 0.1333,
      "step": 55640
    },
    {
      "epoch": 21.61,
      "learning_rate": 1.711844660194175e-05,
      "loss": 0.2394,
      "step": 55650
    },
    {
      "epoch": 21.62,
      "learning_rate": 1.7117928802589e-05,
      "loss": 0.1616,
      "step": 55660
    },
    {
      "epoch": 21.62,
      "learning_rate": 1.711741100323625e-05,
      "loss": 0.1626,
      "step": 55670
    },
    {
      "epoch": 21.62,
      "learning_rate": 1.7116893203883496e-05,
      "loss": 0.1932,
      "step": 55680
    },
    {
      "epoch": 21.63,
      "learning_rate": 1.7116375404530744e-05,
      "loss": 0.0917,
      "step": 55690
    },
    {
      "epoch": 21.63,
      "learning_rate": 1.7115857605177996e-05,
      "loss": 0.1041,
      "step": 55700
    },
    {
      "epoch": 21.63,
      "learning_rate": 1.7115339805825244e-05,
      "loss": 0.0526,
      "step": 55710
    },
    {
      "epoch": 21.64,
      "learning_rate": 1.7114822006472495e-05,
      "loss": 0.0456,
      "step": 55720
    },
    {
      "epoch": 21.64,
      "learning_rate": 1.7114304207119743e-05,
      "loss": 0.0498,
      "step": 55730
    },
    {
      "epoch": 21.65,
      "learning_rate": 1.711378640776699e-05,
      "loss": 0.1055,
      "step": 55740
    },
    {
      "epoch": 21.65,
      "learning_rate": 1.7113268608414242e-05,
      "loss": 0.0494,
      "step": 55750
    },
    {
      "epoch": 21.65,
      "learning_rate": 1.711275080906149e-05,
      "loss": 0.1144,
      "step": 55760
    },
    {
      "epoch": 21.66,
      "learning_rate": 1.7112233009708738e-05,
      "loss": 0.0932,
      "step": 55770
    },
    {
      "epoch": 21.66,
      "learning_rate": 1.711171521035599e-05,
      "loss": 0.1401,
      "step": 55780
    },
    {
      "epoch": 21.67,
      "learning_rate": 1.7111197411003237e-05,
      "loss": 0.0484,
      "step": 55790
    },
    {
      "epoch": 21.67,
      "learning_rate": 1.711067961165049e-05,
      "loss": 0.0679,
      "step": 55800
    },
    {
      "epoch": 21.67,
      "learning_rate": 1.7110161812297736e-05,
      "loss": 0.1739,
      "step": 55810
    },
    {
      "epoch": 21.68,
      "learning_rate": 1.7109644012944984e-05,
      "loss": 0.1994,
      "step": 55820
    },
    {
      "epoch": 21.68,
      "learning_rate": 1.7109126213592236e-05,
      "loss": 0.1304,
      "step": 55830
    },
    {
      "epoch": 21.69,
      "learning_rate": 1.7108608414239484e-05,
      "loss": 0.1328,
      "step": 55840
    },
    {
      "epoch": 21.69,
      "learning_rate": 1.710809061488673e-05,
      "loss": 0.103,
      "step": 55850
    },
    {
      "epoch": 21.69,
      "learning_rate": 1.7107572815533983e-05,
      "loss": 0.1478,
      "step": 55860
    },
    {
      "epoch": 21.7,
      "learning_rate": 1.710705501618123e-05,
      "loss": 0.1,
      "step": 55870
    },
    {
      "epoch": 21.7,
      "learning_rate": 1.7106537216828482e-05,
      "loss": 0.1594,
      "step": 55880
    },
    {
      "epoch": 21.7,
      "learning_rate": 1.710601941747573e-05,
      "loss": 0.2233,
      "step": 55890
    },
    {
      "epoch": 21.71,
      "learning_rate": 1.7105501618122978e-05,
      "loss": 0.1124,
      "step": 55900
    },
    {
      "epoch": 21.71,
      "learning_rate": 1.710498381877023e-05,
      "loss": 0.0616,
      "step": 55910
    },
    {
      "epoch": 21.72,
      "learning_rate": 1.7104466019417477e-05,
      "loss": 0.105,
      "step": 55920
    },
    {
      "epoch": 21.72,
      "learning_rate": 1.7103948220064725e-05,
      "loss": 0.0563,
      "step": 55930
    },
    {
      "epoch": 21.72,
      "learning_rate": 1.7103430420711976e-05,
      "loss": 0.2011,
      "step": 55940
    },
    {
      "epoch": 21.73,
      "learning_rate": 1.7102912621359224e-05,
      "loss": 0.2352,
      "step": 55950
    },
    {
      "epoch": 21.73,
      "learning_rate": 1.7102394822006476e-05,
      "loss": 0.0344,
      "step": 55960
    },
    {
      "epoch": 21.74,
      "learning_rate": 1.7101877022653723e-05,
      "loss": 0.1035,
      "step": 55970
    },
    {
      "epoch": 21.74,
      "learning_rate": 1.710135922330097e-05,
      "loss": 0.3296,
      "step": 55980
    },
    {
      "epoch": 21.74,
      "learning_rate": 1.7100841423948223e-05,
      "loss": 0.0276,
      "step": 55990
    },
    {
      "epoch": 21.75,
      "learning_rate": 1.710032362459547e-05,
      "loss": 0.0698,
      "step": 56000
    },
    {
      "epoch": 21.75,
      "learning_rate": 1.709980582524272e-05,
      "loss": 0.1225,
      "step": 56010
    },
    {
      "epoch": 21.76,
      "learning_rate": 1.709928802588997e-05,
      "loss": 0.0029,
      "step": 56020
    },
    {
      "epoch": 21.76,
      "learning_rate": 1.7098770226537218e-05,
      "loss": 0.2885,
      "step": 56030
    },
    {
      "epoch": 21.76,
      "learning_rate": 1.7098252427184466e-05,
      "loss": 0.2287,
      "step": 56040
    },
    {
      "epoch": 21.77,
      "learning_rate": 1.7097734627831717e-05,
      "loss": 0.0749,
      "step": 56050
    },
    {
      "epoch": 21.77,
      "learning_rate": 1.7097216828478965e-05,
      "loss": 0.0121,
      "step": 56060
    },
    {
      "epoch": 21.77,
      "learning_rate": 1.7096699029126216e-05,
      "loss": 0.0408,
      "step": 56070
    },
    {
      "epoch": 21.78,
      "learning_rate": 1.7096181229773464e-05,
      "loss": 0.0057,
      "step": 56080
    },
    {
      "epoch": 21.78,
      "learning_rate": 1.7095663430420712e-05,
      "loss": 0.0796,
      "step": 56090
    },
    {
      "epoch": 21.79,
      "learning_rate": 1.7095145631067963e-05,
      "loss": 0.1644,
      "step": 56100
    },
    {
      "epoch": 21.79,
      "learning_rate": 1.709462783171521e-05,
      "loss": 0.0027,
      "step": 56110
    },
    {
      "epoch": 21.79,
      "learning_rate": 1.709411003236246e-05,
      "loss": 0.3565,
      "step": 56120
    },
    {
      "epoch": 21.8,
      "learning_rate": 1.709359223300971e-05,
      "loss": 0.0552,
      "step": 56130
    },
    {
      "epoch": 21.8,
      "learning_rate": 1.709307443365696e-05,
      "loss": 0.1265,
      "step": 56140
    },
    {
      "epoch": 21.81,
      "learning_rate": 1.709255663430421e-05,
      "loss": 0.206,
      "step": 56150
    },
    {
      "epoch": 21.81,
      "learning_rate": 1.7092038834951458e-05,
      "loss": 0.259,
      "step": 56160
    },
    {
      "epoch": 21.81,
      "learning_rate": 1.7091521035598706e-05,
      "loss": 0.0806,
      "step": 56170
    },
    {
      "epoch": 21.82,
      "learning_rate": 1.7091003236245957e-05,
      "loss": 0.1418,
      "step": 56180
    },
    {
      "epoch": 21.82,
      "learning_rate": 1.7090485436893205e-05,
      "loss": 0.1497,
      "step": 56190
    },
    {
      "epoch": 21.83,
      "learning_rate": 1.7089967637540453e-05,
      "loss": 0.0129,
      "step": 56200
    },
    {
      "epoch": 21.83,
      "learning_rate": 1.7089449838187704e-05,
      "loss": 0.0373,
      "step": 56210
    },
    {
      "epoch": 21.83,
      "learning_rate": 1.7088932038834952e-05,
      "loss": 0.0755,
      "step": 56220
    },
    {
      "epoch": 21.84,
      "learning_rate": 1.7088414239482203e-05,
      "loss": 0.0171,
      "step": 56230
    },
    {
      "epoch": 21.84,
      "learning_rate": 1.708789644012945e-05,
      "loss": 0.0246,
      "step": 56240
    },
    {
      "epoch": 21.84,
      "learning_rate": 1.70873786407767e-05,
      "loss": 0.2591,
      "step": 56250
    },
    {
      "epoch": 21.85,
      "learning_rate": 1.708686084142395e-05,
      "loss": 0.1271,
      "step": 56260
    },
    {
      "epoch": 21.85,
      "learning_rate": 1.70863430420712e-05,
      "loss": 0.0511,
      "step": 56270
    },
    {
      "epoch": 21.86,
      "learning_rate": 1.7085825242718446e-05,
      "loss": 0.1657,
      "step": 56280
    },
    {
      "epoch": 21.86,
      "learning_rate": 1.7085307443365698e-05,
      "loss": 0.1549,
      "step": 56290
    },
    {
      "epoch": 21.86,
      "learning_rate": 1.7084789644012946e-05,
      "loss": 0.1842,
      "step": 56300
    },
    {
      "epoch": 21.87,
      "learning_rate": 1.7084271844660197e-05,
      "loss": 0.2872,
      "step": 56310
    },
    {
      "epoch": 21.87,
      "learning_rate": 1.7083754045307445e-05,
      "loss": 0.0654,
      "step": 56320
    },
    {
      "epoch": 21.88,
      "learning_rate": 1.7083236245954693e-05,
      "loss": 0.0881,
      "step": 56330
    },
    {
      "epoch": 21.88,
      "learning_rate": 1.7082718446601944e-05,
      "loss": 0.0687,
      "step": 56340
    },
    {
      "epoch": 21.88,
      "learning_rate": 1.7082200647249192e-05,
      "loss": 0.0467,
      "step": 56350
    },
    {
      "epoch": 21.89,
      "learning_rate": 1.708168284789644e-05,
      "loss": 0.0152,
      "step": 56360
    },
    {
      "epoch": 21.89,
      "learning_rate": 1.708116504854369e-05,
      "loss": 0.0566,
      "step": 56370
    },
    {
      "epoch": 21.9,
      "learning_rate": 1.708064724919094e-05,
      "loss": 0.0898,
      "step": 56380
    },
    {
      "epoch": 21.9,
      "learning_rate": 1.708012944983819e-05,
      "loss": 0.0456,
      "step": 56390
    },
    {
      "epoch": 21.9,
      "learning_rate": 1.707961165048544e-05,
      "loss": 0.1176,
      "step": 56400
    },
    {
      "epoch": 21.91,
      "learning_rate": 1.7079093851132686e-05,
      "loss": 0.081,
      "step": 56410
    },
    {
      "epoch": 21.91,
      "learning_rate": 1.7078576051779934e-05,
      "loss": 0.0915,
      "step": 56420
    },
    {
      "epoch": 21.91,
      "learning_rate": 1.7078058252427186e-05,
      "loss": 0.0776,
      "step": 56430
    },
    {
      "epoch": 21.92,
      "learning_rate": 1.7077540453074434e-05,
      "loss": 0.1748,
      "step": 56440
    },
    {
      "epoch": 21.92,
      "learning_rate": 1.7077022653721685e-05,
      "loss": 0.0913,
      "step": 56450
    },
    {
      "epoch": 21.93,
      "learning_rate": 1.7076504854368933e-05,
      "loss": 0.0735,
      "step": 56460
    },
    {
      "epoch": 21.93,
      "learning_rate": 1.7075987055016184e-05,
      "loss": 0.2397,
      "step": 56470
    },
    {
      "epoch": 21.93,
      "learning_rate": 1.7075469255663432e-05,
      "loss": 0.1239,
      "step": 56480
    },
    {
      "epoch": 21.94,
      "learning_rate": 1.707495145631068e-05,
      "loss": 0.1473,
      "step": 56490
    },
    {
      "epoch": 21.94,
      "learning_rate": 1.7074433656957928e-05,
      "loss": 0.0952,
      "step": 56500
    },
    {
      "epoch": 21.95,
      "learning_rate": 1.707391585760518e-05,
      "loss": 0.1722,
      "step": 56510
    },
    {
      "epoch": 21.95,
      "learning_rate": 1.7073398058252427e-05,
      "loss": 0.0992,
      "step": 56520
    },
    {
      "epoch": 21.95,
      "learning_rate": 1.707288025889968e-05,
      "loss": 0.1399,
      "step": 56530
    },
    {
      "epoch": 21.96,
      "learning_rate": 1.7072362459546926e-05,
      "loss": 0.1249,
      "step": 56540
    },
    {
      "epoch": 21.96,
      "learning_rate": 1.7071844660194178e-05,
      "loss": 0.2056,
      "step": 56550
    },
    {
      "epoch": 21.97,
      "learning_rate": 1.7071326860841426e-05,
      "loss": 0.1372,
      "step": 56560
    },
    {
      "epoch": 21.97,
      "learning_rate": 1.7070809061488673e-05,
      "loss": 0.0458,
      "step": 56570
    },
    {
      "epoch": 21.97,
      "learning_rate": 1.707029126213592e-05,
      "loss": 0.1041,
      "step": 56580
    },
    {
      "epoch": 21.98,
      "learning_rate": 1.7069773462783173e-05,
      "loss": 0.1443,
      "step": 56590
    },
    {
      "epoch": 21.98,
      "learning_rate": 1.706925566343042e-05,
      "loss": 0.1043,
      "step": 56600
    },
    {
      "epoch": 21.98,
      "learning_rate": 1.7068737864077672e-05,
      "loss": 0.0388,
      "step": 56610
    },
    {
      "epoch": 21.99,
      "learning_rate": 1.706822006472492e-05,
      "loss": 0.0899,
      "step": 56620
    },
    {
      "epoch": 21.99,
      "learning_rate": 1.706770226537217e-05,
      "loss": 0.0438,
      "step": 56630
    },
    {
      "epoch": 22.0,
      "learning_rate": 1.706718446601942e-05,
      "loss": 0.0551,
      "step": 56640
    },
    {
      "epoch": 22.0,
      "learning_rate": 1.706666666666667e-05,
      "loss": 0.1279,
      "step": 56650
    },
    {
      "epoch": 22.0,
      "eval_accuracy": 0.9499312242090784,
      "eval_loss": 0.26807668805122375,
      "eval_runtime": 8.1793,
      "eval_samples_per_second": 444.417,
      "eval_steps_per_second": 55.628,
      "step": 56650
    },
    {
      "epoch": 22.0,
      "learning_rate": 1.7066148867313915e-05,
      "loss": 0.0868,
      "step": 56660
    },
    {
      "epoch": 22.01,
      "learning_rate": 1.7065631067961166e-05,
      "loss": 0.1483,
      "step": 56670
    },
    {
      "epoch": 22.01,
      "learning_rate": 1.7065113268608414e-05,
      "loss": 0.2797,
      "step": 56680
    },
    {
      "epoch": 22.02,
      "learning_rate": 1.7064595469255665e-05,
      "loss": 0.0626,
      "step": 56690
    },
    {
      "epoch": 22.02,
      "learning_rate": 1.7064077669902913e-05,
      "loss": 0.0286,
      "step": 56700
    },
    {
      "epoch": 22.02,
      "learning_rate": 1.7063559870550165e-05,
      "loss": 0.1776,
      "step": 56710
    },
    {
      "epoch": 22.03,
      "learning_rate": 1.7063042071197413e-05,
      "loss": 0.1385,
      "step": 56720
    },
    {
      "epoch": 22.03,
      "learning_rate": 1.7062524271844664e-05,
      "loss": 0.0497,
      "step": 56730
    },
    {
      "epoch": 22.03,
      "learning_rate": 1.706200647249191e-05,
      "loss": 0.081,
      "step": 56740
    },
    {
      "epoch": 22.04,
      "learning_rate": 1.706148867313916e-05,
      "loss": 0.0546,
      "step": 56750
    },
    {
      "epoch": 22.04,
      "learning_rate": 1.7060970873786408e-05,
      "loss": 0.0368,
      "step": 56760
    },
    {
      "epoch": 22.05,
      "learning_rate": 1.706045307443366e-05,
      "loss": 0.0264,
      "step": 56770
    },
    {
      "epoch": 22.05,
      "learning_rate": 1.7059935275080907e-05,
      "loss": 0.0859,
      "step": 56780
    },
    {
      "epoch": 22.05,
      "learning_rate": 1.7059417475728158e-05,
      "loss": 0.0704,
      "step": 56790
    },
    {
      "epoch": 22.06,
      "learning_rate": 1.7058899676375406e-05,
      "loss": 0.0336,
      "step": 56800
    },
    {
      "epoch": 22.06,
      "learning_rate": 1.7058381877022658e-05,
      "loss": 0.0723,
      "step": 56810
    },
    {
      "epoch": 22.07,
      "learning_rate": 1.7057864077669902e-05,
      "loss": 0.0704,
      "step": 56820
    },
    {
      "epoch": 22.07,
      "learning_rate": 1.7057346278317153e-05,
      "loss": 0.0653,
      "step": 56830
    },
    {
      "epoch": 22.07,
      "learning_rate": 1.70568284789644e-05,
      "loss": 0.0986,
      "step": 56840
    },
    {
      "epoch": 22.08,
      "learning_rate": 1.7056310679611653e-05,
      "loss": 0.1149,
      "step": 56850
    },
    {
      "epoch": 22.08,
      "learning_rate": 1.70557928802589e-05,
      "loss": 0.1184,
      "step": 56860
    },
    {
      "epoch": 22.09,
      "learning_rate": 1.7055275080906152e-05,
      "loss": 0.1502,
      "step": 56870
    },
    {
      "epoch": 22.09,
      "learning_rate": 1.70547572815534e-05,
      "loss": 0.1794,
      "step": 56880
    },
    {
      "epoch": 22.09,
      "learning_rate": 1.705423948220065e-05,
      "loss": 0.1691,
      "step": 56890
    },
    {
      "epoch": 22.1,
      "learning_rate": 1.7053721682847896e-05,
      "loss": 0.0491,
      "step": 56900
    },
    {
      "epoch": 22.1,
      "learning_rate": 1.7053203883495147e-05,
      "loss": 0.1255,
      "step": 56910
    },
    {
      "epoch": 22.1,
      "learning_rate": 1.7052686084142395e-05,
      "loss": 0.1401,
      "step": 56920
    },
    {
      "epoch": 22.11,
      "learning_rate": 1.7052168284789646e-05,
      "loss": 0.0579,
      "step": 56930
    },
    {
      "epoch": 22.11,
      "learning_rate": 1.7051650485436894e-05,
      "loss": 0.0348,
      "step": 56940
    },
    {
      "epoch": 22.12,
      "learning_rate": 1.7051132686084145e-05,
      "loss": 0.1184,
      "step": 56950
    },
    {
      "epoch": 22.12,
      "learning_rate": 1.7050614886731393e-05,
      "loss": 0.0816,
      "step": 56960
    },
    {
      "epoch": 22.12,
      "learning_rate": 1.7050097087378645e-05,
      "loss": 0.0867,
      "step": 56970
    },
    {
      "epoch": 22.13,
      "learning_rate": 1.704957928802589e-05,
      "loss": 0.0812,
      "step": 56980
    },
    {
      "epoch": 22.13,
      "learning_rate": 1.704906148867314e-05,
      "loss": 0.1725,
      "step": 56990
    },
    {
      "epoch": 22.14,
      "learning_rate": 1.704854368932039e-05,
      "loss": 0.0475,
      "step": 57000
    },
    {
      "epoch": 22.14,
      "learning_rate": 1.704802588996764e-05,
      "loss": 0.1612,
      "step": 57010
    },
    {
      "epoch": 22.14,
      "learning_rate": 1.7047508090614888e-05,
      "loss": 0.111,
      "step": 57020
    },
    {
      "epoch": 22.15,
      "learning_rate": 1.704699029126214e-05,
      "loss": 0.0606,
      "step": 57030
    },
    {
      "epoch": 22.15,
      "learning_rate": 1.7046472491909387e-05,
      "loss": 0.2177,
      "step": 57040
    },
    {
      "epoch": 22.16,
      "learning_rate": 1.7045954692556638e-05,
      "loss": 0.1322,
      "step": 57050
    },
    {
      "epoch": 22.16,
      "learning_rate": 1.7045436893203883e-05,
      "loss": 0.1165,
      "step": 57060
    },
    {
      "epoch": 22.16,
      "learning_rate": 1.7044919093851134e-05,
      "loss": 0.0717,
      "step": 57070
    },
    {
      "epoch": 22.17,
      "learning_rate": 1.7044401294498382e-05,
      "loss": 0.0446,
      "step": 57080
    },
    {
      "epoch": 22.17,
      "learning_rate": 1.7043883495145633e-05,
      "loss": 0.1633,
      "step": 57090
    },
    {
      "epoch": 22.17,
      "learning_rate": 1.704336569579288e-05,
      "loss": 0.0606,
      "step": 57100
    },
    {
      "epoch": 22.18,
      "learning_rate": 1.7042847896440133e-05,
      "loss": 0.0775,
      "step": 57110
    },
    {
      "epoch": 22.18,
      "learning_rate": 1.704233009708738e-05,
      "loss": 0.0193,
      "step": 57120
    },
    {
      "epoch": 22.19,
      "learning_rate": 1.7041812297734632e-05,
      "loss": 0.0852,
      "step": 57130
    },
    {
      "epoch": 22.19,
      "learning_rate": 1.7041294498381876e-05,
      "loss": 0.0289,
      "step": 57140
    },
    {
      "epoch": 22.19,
      "learning_rate": 1.7040776699029128e-05,
      "loss": 0.1081,
      "step": 57150
    },
    {
      "epoch": 22.2,
      "learning_rate": 1.7040258899676376e-05,
      "loss": 0.0486,
      "step": 57160
    },
    {
      "epoch": 22.2,
      "learning_rate": 1.7039741100323627e-05,
      "loss": 0.0552,
      "step": 57170
    },
    {
      "epoch": 22.21,
      "learning_rate": 1.7039223300970875e-05,
      "loss": 0.1795,
      "step": 57180
    },
    {
      "epoch": 22.21,
      "learning_rate": 1.7038705501618126e-05,
      "loss": 0.0634,
      "step": 57190
    },
    {
      "epoch": 22.21,
      "learning_rate": 1.7038187702265374e-05,
      "loss": 0.0021,
      "step": 57200
    },
    {
      "epoch": 22.22,
      "learning_rate": 1.7037669902912622e-05,
      "loss": 0.0455,
      "step": 57210
    },
    {
      "epoch": 22.22,
      "learning_rate": 1.7037152103559873e-05,
      "loss": 0.1692,
      "step": 57220
    },
    {
      "epoch": 22.23,
      "learning_rate": 1.703663430420712e-05,
      "loss": 0.1722,
      "step": 57230
    },
    {
      "epoch": 22.23,
      "learning_rate": 1.703611650485437e-05,
      "loss": 0.0644,
      "step": 57240
    },
    {
      "epoch": 22.23,
      "learning_rate": 1.703559870550162e-05,
      "loss": 0.0204,
      "step": 57250
    },
    {
      "epoch": 22.24,
      "learning_rate": 1.703508090614887e-05,
      "loss": 0.0976,
      "step": 57260
    },
    {
      "epoch": 22.24,
      "learning_rate": 1.703456310679612e-05,
      "loss": 0.0794,
      "step": 57270
    },
    {
      "epoch": 22.24,
      "learning_rate": 1.7034045307443368e-05,
      "loss": 0.0689,
      "step": 57280
    },
    {
      "epoch": 22.25,
      "learning_rate": 1.7033527508090615e-05,
      "loss": 0.2256,
      "step": 57290
    },
    {
      "epoch": 22.25,
      "learning_rate": 1.7033009708737867e-05,
      "loss": 0.0108,
      "step": 57300
    },
    {
      "epoch": 22.26,
      "learning_rate": 1.7032491909385115e-05,
      "loss": 0.1096,
      "step": 57310
    },
    {
      "epoch": 22.26,
      "learning_rate": 1.7031974110032363e-05,
      "loss": 0.0758,
      "step": 57320
    },
    {
      "epoch": 22.26,
      "learning_rate": 1.7031456310679614e-05,
      "loss": 0.1962,
      "step": 57330
    },
    {
      "epoch": 22.27,
      "learning_rate": 1.7030938511326862e-05,
      "loss": 0.0962,
      "step": 57340
    },
    {
      "epoch": 22.27,
      "learning_rate": 1.7030420711974113e-05,
      "loss": 0.1346,
      "step": 57350
    },
    {
      "epoch": 22.28,
      "learning_rate": 1.702990291262136e-05,
      "loss": 0.1051,
      "step": 57360
    },
    {
      "epoch": 22.28,
      "learning_rate": 1.702938511326861e-05,
      "loss": 0.089,
      "step": 57370
    },
    {
      "epoch": 22.28,
      "learning_rate": 1.702886731391586e-05,
      "loss": 0.0816,
      "step": 57380
    },
    {
      "epoch": 22.29,
      "learning_rate": 1.7028349514563108e-05,
      "loss": 0.053,
      "step": 57390
    },
    {
      "epoch": 22.29,
      "learning_rate": 1.7027831715210356e-05,
      "loss": 0.1248,
      "step": 57400
    },
    {
      "epoch": 22.3,
      "learning_rate": 1.7027313915857608e-05,
      "loss": 0.0599,
      "step": 57410
    },
    {
      "epoch": 22.3,
      "learning_rate": 1.7026796116504855e-05,
      "loss": 0.2171,
      "step": 57420
    },
    {
      "epoch": 22.3,
      "learning_rate": 1.7026278317152107e-05,
      "loss": 0.1923,
      "step": 57430
    },
    {
      "epoch": 22.31,
      "learning_rate": 1.7025760517799355e-05,
      "loss": 0.1424,
      "step": 57440
    },
    {
      "epoch": 22.31,
      "learning_rate": 1.7025242718446603e-05,
      "loss": 0.2522,
      "step": 57450
    },
    {
      "epoch": 22.31,
      "learning_rate": 1.7024724919093854e-05,
      "loss": 0.096,
      "step": 57460
    },
    {
      "epoch": 22.32,
      "learning_rate": 1.7024207119741102e-05,
      "loss": 0.127,
      "step": 57470
    },
    {
      "epoch": 22.32,
      "learning_rate": 1.702368932038835e-05,
      "loss": 0.1278,
      "step": 57480
    },
    {
      "epoch": 22.33,
      "learning_rate": 1.70231715210356e-05,
      "loss": 0.1319,
      "step": 57490
    },
    {
      "epoch": 22.33,
      "learning_rate": 1.702265372168285e-05,
      "loss": 0.1965,
      "step": 57500
    },
    {
      "epoch": 22.33,
      "learning_rate": 1.7022135922330097e-05,
      "loss": 0.1247,
      "step": 57510
    },
    {
      "epoch": 22.34,
      "learning_rate": 1.7021618122977348e-05,
      "loss": 0.0968,
      "step": 57520
    },
    {
      "epoch": 22.34,
      "learning_rate": 1.7021100323624596e-05,
      "loss": 0.0787,
      "step": 57530
    },
    {
      "epoch": 22.35,
      "learning_rate": 1.7020582524271847e-05,
      "loss": 0.1616,
      "step": 57540
    },
    {
      "epoch": 22.35,
      "learning_rate": 1.7020064724919095e-05,
      "loss": 0.1193,
      "step": 57550
    },
    {
      "epoch": 22.35,
      "learning_rate": 1.7019546925566343e-05,
      "loss": 0.0987,
      "step": 57560
    },
    {
      "epoch": 22.36,
      "learning_rate": 1.7019029126213595e-05,
      "loss": 0.1525,
      "step": 57570
    },
    {
      "epoch": 22.36,
      "learning_rate": 1.7018511326860843e-05,
      "loss": 0.0387,
      "step": 57580
    },
    {
      "epoch": 22.37,
      "learning_rate": 1.701799352750809e-05,
      "loss": 0.2793,
      "step": 57590
    },
    {
      "epoch": 22.37,
      "learning_rate": 1.7017475728155342e-05,
      "loss": 0.0607,
      "step": 57600
    },
    {
      "epoch": 22.37,
      "learning_rate": 1.701695792880259e-05,
      "loss": 0.1299,
      "step": 57610
    },
    {
      "epoch": 22.38,
      "learning_rate": 1.701644012944984e-05,
      "loss": 0.124,
      "step": 57620
    },
    {
      "epoch": 22.38,
      "learning_rate": 1.701592233009709e-05,
      "loss": 0.1442,
      "step": 57630
    },
    {
      "epoch": 22.38,
      "learning_rate": 1.7015404530744337e-05,
      "loss": 0.1942,
      "step": 57640
    },
    {
      "epoch": 22.39,
      "learning_rate": 1.7014886731391588e-05,
      "loss": 0.0763,
      "step": 57650
    },
    {
      "epoch": 22.39,
      "learning_rate": 1.7014368932038836e-05,
      "loss": 0.1281,
      "step": 57660
    },
    {
      "epoch": 22.4,
      "learning_rate": 1.7013851132686084e-05,
      "loss": 0.1476,
      "step": 57670
    },
    {
      "epoch": 22.4,
      "learning_rate": 1.7013333333333335e-05,
      "loss": 0.1007,
      "step": 57680
    },
    {
      "epoch": 22.4,
      "learning_rate": 1.7012815533980583e-05,
      "loss": 0.0673,
      "step": 57690
    },
    {
      "epoch": 22.41,
      "learning_rate": 1.7012297734627835e-05,
      "loss": 0.1719,
      "step": 57700
    },
    {
      "epoch": 22.41,
      "learning_rate": 1.7011779935275082e-05,
      "loss": 0.0525,
      "step": 57710
    },
    {
      "epoch": 22.42,
      "learning_rate": 1.701126213592233e-05,
      "loss": 0.1791,
      "step": 57720
    },
    {
      "epoch": 22.42,
      "learning_rate": 1.7010744336569582e-05,
      "loss": 0.299,
      "step": 57730
    },
    {
      "epoch": 22.42,
      "learning_rate": 1.701022653721683e-05,
      "loss": 0.1779,
      "step": 57740
    },
    {
      "epoch": 22.43,
      "learning_rate": 1.7009708737864078e-05,
      "loss": 0.2476,
      "step": 57750
    },
    {
      "epoch": 22.43,
      "learning_rate": 1.700919093851133e-05,
      "loss": 0.1122,
      "step": 57760
    },
    {
      "epoch": 22.43,
      "learning_rate": 1.7008673139158577e-05,
      "loss": 0.1066,
      "step": 57770
    },
    {
      "epoch": 22.44,
      "learning_rate": 1.7008155339805828e-05,
      "loss": 0.0789,
      "step": 57780
    },
    {
      "epoch": 22.44,
      "learning_rate": 1.7007637540453076e-05,
      "loss": 0.0787,
      "step": 57790
    },
    {
      "epoch": 22.45,
      "learning_rate": 1.7007119741100324e-05,
      "loss": 0.041,
      "step": 57800
    },
    {
      "epoch": 22.45,
      "learning_rate": 1.7006601941747575e-05,
      "loss": 0.1253,
      "step": 57810
    },
    {
      "epoch": 22.45,
      "learning_rate": 1.7006084142394823e-05,
      "loss": 0.0553,
      "step": 57820
    },
    {
      "epoch": 22.46,
      "learning_rate": 1.700556634304207e-05,
      "loss": 0.0448,
      "step": 57830
    },
    {
      "epoch": 22.46,
      "learning_rate": 1.7005048543689322e-05,
      "loss": 0.1449,
      "step": 57840
    },
    {
      "epoch": 22.47,
      "learning_rate": 1.700453074433657e-05,
      "loss": 0.0639,
      "step": 57850
    },
    {
      "epoch": 22.47,
      "learning_rate": 1.700401294498382e-05,
      "loss": 0.147,
      "step": 57860
    },
    {
      "epoch": 22.47,
      "learning_rate": 1.700349514563107e-05,
      "loss": 0.1132,
      "step": 57870
    },
    {
      "epoch": 22.48,
      "learning_rate": 1.7002977346278318e-05,
      "loss": 0.0155,
      "step": 57880
    },
    {
      "epoch": 22.48,
      "learning_rate": 1.7002459546925565e-05,
      "loss": 0.116,
      "step": 57890
    },
    {
      "epoch": 22.49,
      "learning_rate": 1.7001941747572817e-05,
      "loss": 0.1898,
      "step": 57900
    },
    {
      "epoch": 22.49,
      "learning_rate": 1.7001423948220065e-05,
      "loss": 0.0544,
      "step": 57910
    },
    {
      "epoch": 22.49,
      "learning_rate": 1.7000906148867316e-05,
      "loss": 0.1148,
      "step": 57920
    },
    {
      "epoch": 22.5,
      "learning_rate": 1.7000388349514564e-05,
      "loss": 0.1354,
      "step": 57930
    },
    {
      "epoch": 22.5,
      "learning_rate": 1.6999870550161815e-05,
      "loss": 0.2258,
      "step": 57940
    },
    {
      "epoch": 22.5,
      "learning_rate": 1.6999352750809063e-05,
      "loss": 0.0622,
      "step": 57950
    },
    {
      "epoch": 22.51,
      "learning_rate": 1.699883495145631e-05,
      "loss": 0.2274,
      "step": 57960
    },
    {
      "epoch": 22.51,
      "learning_rate": 1.699831715210356e-05,
      "loss": 0.071,
      "step": 57970
    },
    {
      "epoch": 22.52,
      "learning_rate": 1.699779935275081e-05,
      "loss": 0.305,
      "step": 57980
    },
    {
      "epoch": 22.52,
      "learning_rate": 1.6997281553398058e-05,
      "loss": 0.1886,
      "step": 57990
    },
    {
      "epoch": 22.52,
      "learning_rate": 1.699676375404531e-05,
      "loss": 0.0855,
      "step": 58000
    },
    {
      "epoch": 22.53,
      "learning_rate": 1.6996245954692557e-05,
      "loss": 0.0473,
      "step": 58010
    },
    {
      "epoch": 22.53,
      "learning_rate": 1.699572815533981e-05,
      "loss": 0.1049,
      "step": 58020
    },
    {
      "epoch": 22.54,
      "learning_rate": 1.6995210355987057e-05,
      "loss": 0.0841,
      "step": 58030
    },
    {
      "epoch": 22.54,
      "learning_rate": 1.6994692556634305e-05,
      "loss": 0.1048,
      "step": 58040
    },
    {
      "epoch": 22.54,
      "learning_rate": 1.6994174757281553e-05,
      "loss": 0.0617,
      "step": 58050
    },
    {
      "epoch": 22.55,
      "learning_rate": 1.6993656957928804e-05,
      "loss": 0.2287,
      "step": 58060
    },
    {
      "epoch": 22.55,
      "learning_rate": 1.6993139158576052e-05,
      "loss": 0.144,
      "step": 58070
    },
    {
      "epoch": 22.56,
      "learning_rate": 1.6992621359223303e-05,
      "loss": 0.087,
      "step": 58080
    },
    {
      "epoch": 22.56,
      "learning_rate": 1.699210355987055e-05,
      "loss": 0.1708,
      "step": 58090
    },
    {
      "epoch": 22.56,
      "learning_rate": 1.6991585760517802e-05,
      "loss": 0.0515,
      "step": 58100
    },
    {
      "epoch": 22.57,
      "learning_rate": 1.699106796116505e-05,
      "loss": 0.072,
      "step": 58110
    },
    {
      "epoch": 22.57,
      "learning_rate": 1.6990550161812298e-05,
      "loss": 0.2188,
      "step": 58120
    },
    {
      "epoch": 22.57,
      "learning_rate": 1.6990032362459546e-05,
      "loss": 0.2649,
      "step": 58130
    },
    {
      "epoch": 22.58,
      "learning_rate": 1.6989514563106797e-05,
      "loss": 0.0191,
      "step": 58140
    },
    {
      "epoch": 22.58,
      "learning_rate": 1.6988996763754045e-05,
      "loss": 0.1785,
      "step": 58150
    },
    {
      "epoch": 22.59,
      "learning_rate": 1.6988478964401297e-05,
      "loss": 0.0601,
      "step": 58160
    },
    {
      "epoch": 22.59,
      "learning_rate": 1.6987961165048545e-05,
      "loss": 0.0506,
      "step": 58170
    },
    {
      "epoch": 22.59,
      "learning_rate": 1.6987443365695796e-05,
      "loss": 0.0513,
      "step": 58180
    },
    {
      "epoch": 22.6,
      "learning_rate": 1.6986925566343044e-05,
      "loss": 0.0515,
      "step": 58190
    },
    {
      "epoch": 22.6,
      "learning_rate": 1.6986407766990292e-05,
      "loss": 0.0886,
      "step": 58200
    },
    {
      "epoch": 22.61,
      "learning_rate": 1.698588996763754e-05,
      "loss": 0.0661,
      "step": 58210
    },
    {
      "epoch": 22.61,
      "learning_rate": 1.698537216828479e-05,
      "loss": 0.2795,
      "step": 58220
    },
    {
      "epoch": 22.61,
      "learning_rate": 1.698485436893204e-05,
      "loss": 0.0744,
      "step": 58230
    },
    {
      "epoch": 22.62,
      "learning_rate": 1.698433656957929e-05,
      "loss": 0.216,
      "step": 58240
    },
    {
      "epoch": 22.62,
      "learning_rate": 1.6983818770226538e-05,
      "loss": 0.0652,
      "step": 58250
    },
    {
      "epoch": 22.63,
      "learning_rate": 1.698330097087379e-05,
      "loss": 0.1286,
      "step": 58260
    },
    {
      "epoch": 22.63,
      "learning_rate": 1.6982783171521037e-05,
      "loss": 0.0827,
      "step": 58270
    },
    {
      "epoch": 22.63,
      "learning_rate": 1.6982265372168285e-05,
      "loss": 0.0236,
      "step": 58280
    },
    {
      "epoch": 22.64,
      "learning_rate": 1.6981747572815533e-05,
      "loss": 0.126,
      "step": 58290
    },
    {
      "epoch": 22.64,
      "learning_rate": 1.6981229773462785e-05,
      "loss": 0.2212,
      "step": 58300
    },
    {
      "epoch": 22.64,
      "learning_rate": 1.6980711974110032e-05,
      "loss": 0.042,
      "step": 58310
    },
    {
      "epoch": 22.65,
      "learning_rate": 1.6980194174757284e-05,
      "loss": 0.1522,
      "step": 58320
    },
    {
      "epoch": 22.65,
      "learning_rate": 1.6979676375404532e-05,
      "loss": 0.1429,
      "step": 58330
    },
    {
      "epoch": 22.66,
      "learning_rate": 1.6979158576051783e-05,
      "loss": 0.157,
      "step": 58340
    },
    {
      "epoch": 22.66,
      "learning_rate": 1.697864077669903e-05,
      "loss": 0.0994,
      "step": 58350
    },
    {
      "epoch": 22.66,
      "learning_rate": 1.6978122977346282e-05,
      "loss": 0.0834,
      "step": 58360
    },
    {
      "epoch": 22.67,
      "learning_rate": 1.6977605177993527e-05,
      "loss": 0.1411,
      "step": 58370
    },
    {
      "epoch": 22.67,
      "learning_rate": 1.6977087378640778e-05,
      "loss": 0.1666,
      "step": 58380
    },
    {
      "epoch": 22.68,
      "learning_rate": 1.6976569579288026e-05,
      "loss": 0.0351,
      "step": 58390
    },
    {
      "epoch": 22.68,
      "learning_rate": 1.6976051779935277e-05,
      "loss": 0.103,
      "step": 58400
    },
    {
      "epoch": 22.68,
      "learning_rate": 1.6975533980582525e-05,
      "loss": 0.13,
      "step": 58410
    },
    {
      "epoch": 22.69,
      "learning_rate": 1.6975016181229777e-05,
      "loss": 0.0492,
      "step": 58420
    },
    {
      "epoch": 22.69,
      "learning_rate": 1.6974498381877024e-05,
      "loss": 0.0913,
      "step": 58430
    },
    {
      "epoch": 22.7,
      "learning_rate": 1.6973980582524276e-05,
      "loss": 0.0205,
      "step": 58440
    },
    {
      "epoch": 22.7,
      "learning_rate": 1.697346278317152e-05,
      "loss": 0.182,
      "step": 58450
    },
    {
      "epoch": 22.7,
      "learning_rate": 1.697294498381877e-05,
      "loss": 0.068,
      "step": 58460
    },
    {
      "epoch": 22.71,
      "learning_rate": 1.697242718446602e-05,
      "loss": 0.1399,
      "step": 58470
    },
    {
      "epoch": 22.71,
      "learning_rate": 1.697190938511327e-05,
      "loss": 0.1844,
      "step": 58480
    },
    {
      "epoch": 22.71,
      "learning_rate": 1.697139158576052e-05,
      "loss": 0.0053,
      "step": 58490
    },
    {
      "epoch": 22.72,
      "learning_rate": 1.697087378640777e-05,
      "loss": 0.1506,
      "step": 58500
    },
    {
      "epoch": 22.72,
      "learning_rate": 1.6970355987055018e-05,
      "loss": 0.0964,
      "step": 58510
    },
    {
      "epoch": 22.73,
      "learning_rate": 1.696983818770227e-05,
      "loss": 0.2475,
      "step": 58520
    },
    {
      "epoch": 22.73,
      "learning_rate": 1.6969320388349514e-05,
      "loss": 0.1454,
      "step": 58530
    },
    {
      "epoch": 22.73,
      "learning_rate": 1.6968802588996765e-05,
      "loss": 0.2255,
      "step": 58540
    },
    {
      "epoch": 22.74,
      "learning_rate": 1.6968284789644013e-05,
      "loss": 0.1579,
      "step": 58550
    },
    {
      "epoch": 22.74,
      "learning_rate": 1.6967766990291264e-05,
      "loss": 0.0906,
      "step": 58560
    },
    {
      "epoch": 22.75,
      "learning_rate": 1.6967249190938512e-05,
      "loss": 0.1187,
      "step": 58570
    },
    {
      "epoch": 22.75,
      "learning_rate": 1.6966731391585764e-05,
      "loss": 0.0903,
      "step": 58580
    },
    {
      "epoch": 22.75,
      "learning_rate": 1.696621359223301e-05,
      "loss": 0.0859,
      "step": 58590
    },
    {
      "epoch": 22.76,
      "learning_rate": 1.6965695792880263e-05,
      "loss": 0.0535,
      "step": 58600
    },
    {
      "epoch": 22.76,
      "learning_rate": 1.6965177993527507e-05,
      "loss": 0.0656,
      "step": 58610
    },
    {
      "epoch": 22.77,
      "learning_rate": 1.696466019417476e-05,
      "loss": 0.0973,
      "step": 58620
    },
    {
      "epoch": 22.77,
      "learning_rate": 1.6964142394822007e-05,
      "loss": 0.1925,
      "step": 58630
    },
    {
      "epoch": 22.77,
      "learning_rate": 1.6963624595469258e-05,
      "loss": 0.0856,
      "step": 58640
    },
    {
      "epoch": 22.78,
      "learning_rate": 1.6963106796116506e-05,
      "loss": 0.129,
      "step": 58650
    },
    {
      "epoch": 22.78,
      "learning_rate": 1.6962588996763757e-05,
      "loss": 0.1561,
      "step": 58660
    },
    {
      "epoch": 22.78,
      "learning_rate": 1.6962071197411005e-05,
      "loss": 0.2181,
      "step": 58670
    },
    {
      "epoch": 22.79,
      "learning_rate": 1.6961553398058253e-05,
      "loss": 0.0919,
      "step": 58680
    },
    {
      "epoch": 22.79,
      "learning_rate": 1.69610355987055e-05,
      "loss": 0.0479,
      "step": 58690
    },
    {
      "epoch": 22.8,
      "learning_rate": 1.6960517799352752e-05,
      "loss": 0.2773,
      "step": 58700
    },
    {
      "epoch": 22.8,
      "learning_rate": 1.696e-05,
      "loss": 0.2401,
      "step": 58710
    },
    {
      "epoch": 22.8,
      "learning_rate": 1.695948220064725e-05,
      "loss": 0.0825,
      "step": 58720
    },
    {
      "epoch": 22.81,
      "learning_rate": 1.69589644012945e-05,
      "loss": 0.2304,
      "step": 58730
    },
    {
      "epoch": 22.81,
      "learning_rate": 1.695844660194175e-05,
      "loss": 0.2856,
      "step": 58740
    },
    {
      "epoch": 22.82,
      "learning_rate": 1.6957928802589e-05,
      "loss": 0.1123,
      "step": 58750
    },
    {
      "epoch": 22.82,
      "learning_rate": 1.6957411003236247e-05,
      "loss": 0.0891,
      "step": 58760
    },
    {
      "epoch": 22.82,
      "learning_rate": 1.6956893203883495e-05,
      "loss": 0.2267,
      "step": 58770
    },
    {
      "epoch": 22.83,
      "learning_rate": 1.6956375404530746e-05,
      "loss": 0.1192,
      "step": 58780
    },
    {
      "epoch": 22.83,
      "learning_rate": 1.6955857605177994e-05,
      "loss": 0.0996,
      "step": 58790
    },
    {
      "epoch": 22.83,
      "learning_rate": 1.6955339805825245e-05,
      "loss": 0.0248,
      "step": 58800
    },
    {
      "epoch": 22.84,
      "learning_rate": 1.6954822006472493e-05,
      "loss": 0.1564,
      "step": 58810
    },
    {
      "epoch": 22.84,
      "learning_rate": 1.6954304207119744e-05,
      "loss": 0.244,
      "step": 58820
    },
    {
      "epoch": 22.85,
      "learning_rate": 1.6953786407766992e-05,
      "loss": 0.2019,
      "step": 58830
    },
    {
      "epoch": 22.85,
      "learning_rate": 1.695326860841424e-05,
      "loss": 0.3624,
      "step": 58840
    },
    {
      "epoch": 22.85,
      "learning_rate": 1.6952750809061488e-05,
      "loss": 0.1258,
      "step": 58850
    },
    {
      "epoch": 22.86,
      "learning_rate": 1.695223300970874e-05,
      "loss": 0.1295,
      "step": 58860
    },
    {
      "epoch": 22.86,
      "learning_rate": 1.6951715210355987e-05,
      "loss": 0.0489,
      "step": 58870
    },
    {
      "epoch": 22.87,
      "learning_rate": 1.695119741100324e-05,
      "loss": 0.065,
      "step": 58880
    },
    {
      "epoch": 22.87,
      "learning_rate": 1.6950679611650487e-05,
      "loss": 0.0495,
      "step": 58890
    },
    {
      "epoch": 22.87,
      "learning_rate": 1.6950161812297738e-05,
      "loss": 0.0875,
      "step": 58900
    },
    {
      "epoch": 22.88,
      "learning_rate": 1.6949644012944986e-05,
      "loss": 0.1045,
      "step": 58910
    },
    {
      "epoch": 22.88,
      "learning_rate": 1.6949126213592234e-05,
      "loss": 0.0625,
      "step": 58920
    },
    {
      "epoch": 22.89,
      "learning_rate": 1.6948608414239485e-05,
      "loss": 0.0866,
      "step": 58930
    },
    {
      "epoch": 22.89,
      "learning_rate": 1.6948090614886733e-05,
      "loss": 0.1081,
      "step": 58940
    },
    {
      "epoch": 22.89,
      "learning_rate": 1.694757281553398e-05,
      "loss": 0.2141,
      "step": 58950
    },
    {
      "epoch": 22.9,
      "learning_rate": 1.6947055016181232e-05,
      "loss": 0.1194,
      "step": 58960
    },
    {
      "epoch": 22.9,
      "learning_rate": 1.694653721682848e-05,
      "loss": 0.0461,
      "step": 58970
    },
    {
      "epoch": 22.9,
      "learning_rate": 1.6946019417475728e-05,
      "loss": 0.0403,
      "step": 58980
    },
    {
      "epoch": 22.91,
      "learning_rate": 1.694550161812298e-05,
      "loss": 0.2732,
      "step": 58990
    },
    {
      "epoch": 22.91,
      "learning_rate": 1.6944983818770227e-05,
      "loss": 0.034,
      "step": 59000
    },
    {
      "epoch": 22.92,
      "learning_rate": 1.694446601941748e-05,
      "loss": 0.1972,
      "step": 59010
    },
    {
      "epoch": 22.92,
      "learning_rate": 1.6943948220064727e-05,
      "loss": 0.1307,
      "step": 59020
    },
    {
      "epoch": 22.92,
      "learning_rate": 1.6943430420711974e-05,
      "loss": 0.0412,
      "step": 59030
    },
    {
      "epoch": 22.93,
      "learning_rate": 1.6942912621359226e-05,
      "loss": 0.0462,
      "step": 59040
    },
    {
      "epoch": 22.93,
      "learning_rate": 1.6942394822006474e-05,
      "loss": 0.0932,
      "step": 59050
    },
    {
      "epoch": 22.94,
      "learning_rate": 1.694187702265372e-05,
      "loss": 0.0382,
      "step": 59060
    },
    {
      "epoch": 22.94,
      "learning_rate": 1.6941359223300973e-05,
      "loss": 0.1444,
      "step": 59070
    },
    {
      "epoch": 22.94,
      "learning_rate": 1.694084142394822e-05,
      "loss": 0.0536,
      "step": 59080
    },
    {
      "epoch": 22.95,
      "learning_rate": 1.6940323624595472e-05,
      "loss": 0.2126,
      "step": 59090
    },
    {
      "epoch": 22.95,
      "learning_rate": 1.693980582524272e-05,
      "loss": 0.0722,
      "step": 59100
    },
    {
      "epoch": 22.96,
      "learning_rate": 1.6939288025889968e-05,
      "loss": 0.0368,
      "step": 59110
    },
    {
      "epoch": 22.96,
      "learning_rate": 1.693877022653722e-05,
      "loss": 0.0831,
      "step": 59120
    },
    {
      "epoch": 22.96,
      "learning_rate": 1.6938252427184467e-05,
      "loss": 0.0709,
      "step": 59130
    },
    {
      "epoch": 22.97,
      "learning_rate": 1.6937734627831715e-05,
      "loss": 0.1769,
      "step": 59140
    },
    {
      "epoch": 22.97,
      "learning_rate": 1.6937216828478967e-05,
      "loss": 0.0714,
      "step": 59150
    },
    {
      "epoch": 22.97,
      "learning_rate": 1.6936699029126214e-05,
      "loss": 0.0824,
      "step": 59160
    },
    {
      "epoch": 22.98,
      "learning_rate": 1.6936181229773466e-05,
      "loss": 0.0851,
      "step": 59170
    },
    {
      "epoch": 22.98,
      "learning_rate": 1.6935663430420714e-05,
      "loss": 0.1463,
      "step": 59180
    },
    {
      "epoch": 22.99,
      "learning_rate": 1.693514563106796e-05,
      "loss": 0.0013,
      "step": 59190
    },
    {
      "epoch": 22.99,
      "learning_rate": 1.6934627831715213e-05,
      "loss": 0.1208,
      "step": 59200
    },
    {
      "epoch": 22.99,
      "learning_rate": 1.693411003236246e-05,
      "loss": 0.0521,
      "step": 59210
    },
    {
      "epoch": 23.0,
      "learning_rate": 1.693359223300971e-05,
      "loss": 0.1811,
      "step": 59220
    },
    {
      "epoch": 23.0,
      "eval_accuracy": 0.9537826685006877,
      "eval_loss": 0.26258033514022827,
      "eval_runtime": 8.1758,
      "eval_samples_per_second": 444.606,
      "eval_steps_per_second": 55.652,
      "step": 59225
    },
    {
      "epoch": 23.0,
      "learning_rate": 1.693307443365696e-05,
      "loss": 0.0555,
      "step": 59230
    },
    {
      "epoch": 23.01,
      "learning_rate": 1.6932556634304208e-05,
      "loss": 0.0798,
      "step": 59240
    },
    {
      "epoch": 23.01,
      "learning_rate": 1.693203883495146e-05,
      "loss": 0.1749,
      "step": 59250
    },
    {
      "epoch": 23.01,
      "learning_rate": 1.6931521035598707e-05,
      "loss": 0.2087,
      "step": 59260
    },
    {
      "epoch": 23.02,
      "learning_rate": 1.6931003236245955e-05,
      "loss": 0.0888,
      "step": 59270
    },
    {
      "epoch": 23.02,
      "learning_rate": 1.6930485436893206e-05,
      "loss": 0.0434,
      "step": 59280
    },
    {
      "epoch": 23.03,
      "learning_rate": 1.6929967637540454e-05,
      "loss": 0.1861,
      "step": 59290
    },
    {
      "epoch": 23.03,
      "learning_rate": 1.6929449838187702e-05,
      "loss": 0.1173,
      "step": 59300
    },
    {
      "epoch": 23.03,
      "learning_rate": 1.6928932038834954e-05,
      "loss": 0.062,
      "step": 59310
    },
    {
      "epoch": 23.04,
      "learning_rate": 1.69284142394822e-05,
      "loss": 0.0012,
      "step": 59320
    },
    {
      "epoch": 23.04,
      "learning_rate": 1.6927896440129453e-05,
      "loss": 0.1807,
      "step": 59330
    },
    {
      "epoch": 23.04,
      "learning_rate": 1.69273786407767e-05,
      "loss": 0.1428,
      "step": 59340
    },
    {
      "epoch": 23.05,
      "learning_rate": 1.692686084142395e-05,
      "loss": 0.0456,
      "step": 59350
    },
    {
      "epoch": 23.05,
      "learning_rate": 1.6926343042071197e-05,
      "loss": 0.0576,
      "step": 59360
    },
    {
      "epoch": 23.06,
      "learning_rate": 1.6925825242718448e-05,
      "loss": 0.0907,
      "step": 59370
    },
    {
      "epoch": 23.06,
      "learning_rate": 1.6925307443365696e-05,
      "loss": 0.1856,
      "step": 59380
    },
    {
      "epoch": 23.06,
      "learning_rate": 1.6924789644012947e-05,
      "loss": 0.1394,
      "step": 59390
    },
    {
      "epoch": 23.07,
      "learning_rate": 1.6924271844660195e-05,
      "loss": 0.1159,
      "step": 59400
    },
    {
      "epoch": 23.07,
      "learning_rate": 1.6923754045307446e-05,
      "loss": 0.0387,
      "step": 59410
    },
    {
      "epoch": 23.08,
      "learning_rate": 1.6923236245954694e-05,
      "loss": 0.0594,
      "step": 59420
    },
    {
      "epoch": 23.08,
      "learning_rate": 1.6922718446601942e-05,
      "loss": 0.1529,
      "step": 59430
    },
    {
      "epoch": 23.08,
      "learning_rate": 1.692220064724919e-05,
      "loss": 0.1416,
      "step": 59440
    },
    {
      "epoch": 23.09,
      "learning_rate": 1.692168284789644e-05,
      "loss": 0.0862,
      "step": 59450
    },
    {
      "epoch": 23.09,
      "learning_rate": 1.692116504854369e-05,
      "loss": 0.0877,
      "step": 59460
    },
    {
      "epoch": 23.1,
      "learning_rate": 1.692064724919094e-05,
      "loss": 0.0766,
      "step": 59470
    },
    {
      "epoch": 23.1,
      "learning_rate": 1.692012944983819e-05,
      "loss": 0.0084,
      "step": 59480
    },
    {
      "epoch": 23.1,
      "learning_rate": 1.691961165048544e-05,
      "loss": 0.0409,
      "step": 59490
    },
    {
      "epoch": 23.11,
      "learning_rate": 1.6919093851132688e-05,
      "loss": 0.1379,
      "step": 59500
    },
    {
      "epoch": 23.11,
      "learning_rate": 1.6918576051779936e-05,
      "loss": 0.0598,
      "step": 59510
    },
    {
      "epoch": 23.11,
      "learning_rate": 1.6918058252427184e-05,
      "loss": 0.0606,
      "step": 59520
    },
    {
      "epoch": 23.12,
      "learning_rate": 1.6917540453074435e-05,
      "loss": 0.0702,
      "step": 59530
    },
    {
      "epoch": 23.12,
      "learning_rate": 1.6917022653721683e-05,
      "loss": 0.009,
      "step": 59540
    },
    {
      "epoch": 23.13,
      "learning_rate": 1.6916504854368934e-05,
      "loss": 0.0149,
      "step": 59550
    },
    {
      "epoch": 23.13,
      "learning_rate": 1.6915987055016182e-05,
      "loss": 0.055,
      "step": 59560
    },
    {
      "epoch": 23.13,
      "learning_rate": 1.6915469255663434e-05,
      "loss": 0.189,
      "step": 59570
    },
    {
      "epoch": 23.14,
      "learning_rate": 1.691495145631068e-05,
      "loss": 0.2194,
      "step": 59580
    },
    {
      "epoch": 23.14,
      "learning_rate": 1.691443365695793e-05,
      "loss": 0.0497,
      "step": 59590
    },
    {
      "epoch": 23.15,
      "learning_rate": 1.6913915857605177e-05,
      "loss": 0.0698,
      "step": 59600
    },
    {
      "epoch": 23.15,
      "learning_rate": 1.691339805825243e-05,
      "loss": 0.0998,
      "step": 59610
    },
    {
      "epoch": 23.15,
      "learning_rate": 1.6912880258899677e-05,
      "loss": 0.164,
      "step": 59620
    },
    {
      "epoch": 23.16,
      "learning_rate": 1.6912362459546928e-05,
      "loss": 0.0282,
      "step": 59630
    },
    {
      "epoch": 23.16,
      "learning_rate": 1.6911844660194176e-05,
      "loss": 0.224,
      "step": 59640
    },
    {
      "epoch": 23.17,
      "learning_rate": 1.6911326860841427e-05,
      "loss": 0.1048,
      "step": 59650
    },
    {
      "epoch": 23.17,
      "learning_rate": 1.6910809061488675e-05,
      "loss": 0.059,
      "step": 59660
    },
    {
      "epoch": 23.17,
      "learning_rate": 1.6910291262135923e-05,
      "loss": 0.1952,
      "step": 59670
    },
    {
      "epoch": 23.18,
      "learning_rate": 1.690977346278317e-05,
      "loss": 0.263,
      "step": 59680
    },
    {
      "epoch": 23.18,
      "learning_rate": 1.6909255663430422e-05,
      "loss": 0.0432,
      "step": 59690
    },
    {
      "epoch": 23.18,
      "learning_rate": 1.690873786407767e-05,
      "loss": 0.0947,
      "step": 59700
    },
    {
      "epoch": 23.19,
      "learning_rate": 1.690822006472492e-05,
      "loss": 0.1292,
      "step": 59710
    },
    {
      "epoch": 23.19,
      "learning_rate": 1.690770226537217e-05,
      "loss": 0.0824,
      "step": 59720
    },
    {
      "epoch": 23.2,
      "learning_rate": 1.690718446601942e-05,
      "loss": 0.1214,
      "step": 59730
    },
    {
      "epoch": 23.2,
      "learning_rate": 1.690666666666667e-05,
      "loss": 0.1139,
      "step": 59740
    },
    {
      "epoch": 23.2,
      "learning_rate": 1.6906148867313916e-05,
      "loss": 0.0389,
      "step": 59750
    },
    {
      "epoch": 23.21,
      "learning_rate": 1.6905631067961164e-05,
      "loss": 0.0309,
      "step": 59760
    },
    {
      "epoch": 23.21,
      "learning_rate": 1.6905113268608416e-05,
      "loss": 0.168,
      "step": 59770
    },
    {
      "epoch": 23.22,
      "learning_rate": 1.6904595469255664e-05,
      "loss": 0.0307,
      "step": 59780
    },
    {
      "epoch": 23.22,
      "learning_rate": 1.6904077669902915e-05,
      "loss": 0.0525,
      "step": 59790
    },
    {
      "epoch": 23.22,
      "learning_rate": 1.6903559870550163e-05,
      "loss": 0.1572,
      "step": 59800
    },
    {
      "epoch": 23.23,
      "learning_rate": 1.6903042071197414e-05,
      "loss": 0.0713,
      "step": 59810
    },
    {
      "epoch": 23.23,
      "learning_rate": 1.6902524271844662e-05,
      "loss": 0.0648,
      "step": 59820
    },
    {
      "epoch": 23.23,
      "learning_rate": 1.690200647249191e-05,
      "loss": 0.0912,
      "step": 59830
    },
    {
      "epoch": 23.24,
      "learning_rate": 1.6901488673139158e-05,
      "loss": 0.2647,
      "step": 59840
    },
    {
      "epoch": 23.24,
      "learning_rate": 1.690097087378641e-05,
      "loss": 0.1224,
      "step": 59850
    },
    {
      "epoch": 23.25,
      "learning_rate": 1.6900453074433657e-05,
      "loss": 0.0646,
      "step": 59860
    },
    {
      "epoch": 23.25,
      "learning_rate": 1.689993527508091e-05,
      "loss": 0.0245,
      "step": 59870
    },
    {
      "epoch": 23.25,
      "learning_rate": 1.6899417475728156e-05,
      "loss": 0.0702,
      "step": 59880
    },
    {
      "epoch": 23.26,
      "learning_rate": 1.6898899676375408e-05,
      "loss": 0.0674,
      "step": 59890
    },
    {
      "epoch": 23.26,
      "learning_rate": 1.6898381877022656e-05,
      "loss": 0.057,
      "step": 59900
    },
    {
      "epoch": 23.27,
      "learning_rate": 1.6897864077669904e-05,
      "loss": 0.0633,
      "step": 59910
    },
    {
      "epoch": 23.27,
      "learning_rate": 1.689734627831715e-05,
      "loss": 0.1023,
      "step": 59920
    },
    {
      "epoch": 23.27,
      "learning_rate": 1.6896828478964403e-05,
      "loss": 0.079,
      "step": 59930
    },
    {
      "epoch": 23.28,
      "learning_rate": 1.689631067961165e-05,
      "loss": 0.0908,
      "step": 59940
    },
    {
      "epoch": 23.28,
      "learning_rate": 1.6895792880258902e-05,
      "loss": 0.077,
      "step": 59950
    },
    {
      "epoch": 23.29,
      "learning_rate": 1.689527508090615e-05,
      "loss": 0.0877,
      "step": 59960
    },
    {
      "epoch": 23.29,
      "learning_rate": 1.68947572815534e-05,
      "loss": 0.4114,
      "step": 59970
    },
    {
      "epoch": 23.29,
      "learning_rate": 1.689423948220065e-05,
      "loss": 0.0432,
      "step": 59980
    },
    {
      "epoch": 23.3,
      "learning_rate": 1.68937216828479e-05,
      "loss": 0.2354,
      "step": 59990
    },
    {
      "epoch": 23.3,
      "learning_rate": 1.6893203883495145e-05,
      "loss": 0.1698,
      "step": 60000
    },
    {
      "epoch": 23.3,
      "learning_rate": 1.6892686084142396e-05,
      "loss": 0.0811,
      "step": 60010
    },
    {
      "epoch": 23.31,
      "learning_rate": 1.6892168284789644e-05,
      "loss": 0.0182,
      "step": 60020
    },
    {
      "epoch": 23.31,
      "learning_rate": 1.6891650485436896e-05,
      "loss": 0.1474,
      "step": 60030
    },
    {
      "epoch": 23.32,
      "learning_rate": 1.6891132686084144e-05,
      "loss": 0.158,
      "step": 60040
    },
    {
      "epoch": 23.32,
      "learning_rate": 1.6890614886731395e-05,
      "loss": 0.1737,
      "step": 60050
    },
    {
      "epoch": 23.32,
      "learning_rate": 1.6890097087378643e-05,
      "loss": 0.0814,
      "step": 60060
    },
    {
      "epoch": 23.33,
      "learning_rate": 1.6889579288025894e-05,
      "loss": 0.1402,
      "step": 60070
    },
    {
      "epoch": 23.33,
      "learning_rate": 1.688906148867314e-05,
      "loss": 0.079,
      "step": 60080
    },
    {
      "epoch": 23.34,
      "learning_rate": 1.688854368932039e-05,
      "loss": 0.0245,
      "step": 60090
    },
    {
      "epoch": 23.34,
      "learning_rate": 1.6888025889967638e-05,
      "loss": 0.0968,
      "step": 60100
    },
    {
      "epoch": 23.34,
      "learning_rate": 1.688750809061489e-05,
      "loss": 0.2734,
      "step": 60110
    },
    {
      "epoch": 23.35,
      "learning_rate": 1.6886990291262137e-05,
      "loss": 0.1262,
      "step": 60120
    },
    {
      "epoch": 23.35,
      "learning_rate": 1.688647249190939e-05,
      "loss": 0.1816,
      "step": 60130
    },
    {
      "epoch": 23.36,
      "learning_rate": 1.6885954692556636e-05,
      "loss": 0.1202,
      "step": 60140
    },
    {
      "epoch": 23.36,
      "learning_rate": 1.6885436893203884e-05,
      "loss": 0.1726,
      "step": 60150
    },
    {
      "epoch": 23.36,
      "learning_rate": 1.6884919093851132e-05,
      "loss": 0.107,
      "step": 60160
    },
    {
      "epoch": 23.37,
      "learning_rate": 1.6884401294498383e-05,
      "loss": 0.0987,
      "step": 60170
    },
    {
      "epoch": 23.37,
      "learning_rate": 1.688388349514563e-05,
      "loss": 0.1186,
      "step": 60180
    },
    {
      "epoch": 23.37,
      "learning_rate": 1.6883365695792883e-05,
      "loss": 0.0222,
      "step": 60190
    },
    {
      "epoch": 23.38,
      "learning_rate": 1.688284789644013e-05,
      "loss": 0.0677,
      "step": 60200
    },
    {
      "epoch": 23.38,
      "learning_rate": 1.6882330097087382e-05,
      "loss": 0.1066,
      "step": 60210
    },
    {
      "epoch": 23.39,
      "learning_rate": 1.688181229773463e-05,
      "loss": 0.186,
      "step": 60220
    },
    {
      "epoch": 23.39,
      "learning_rate": 1.6881294498381878e-05,
      "loss": 0.0952,
      "step": 60230
    },
    {
      "epoch": 23.39,
      "learning_rate": 1.6880776699029126e-05,
      "loss": 0.0346,
      "step": 60240
    },
    {
      "epoch": 23.4,
      "learning_rate": 1.6880258899676377e-05,
      "loss": 0.0708,
      "step": 60250
    },
    {
      "epoch": 23.4,
      "learning_rate": 1.6879741100323625e-05,
      "loss": 0.0781,
      "step": 60260
    },
    {
      "epoch": 23.41,
      "learning_rate": 1.6879223300970876e-05,
      "loss": 0.0944,
      "step": 60270
    },
    {
      "epoch": 23.41,
      "learning_rate": 1.6878705501618124e-05,
      "loss": 0.1052,
      "step": 60280
    },
    {
      "epoch": 23.41,
      "learning_rate": 1.6878187702265376e-05,
      "loss": 0.1652,
      "step": 60290
    },
    {
      "epoch": 23.42,
      "learning_rate": 1.6877669902912623e-05,
      "loss": 0.0112,
      "step": 60300
    },
    {
      "epoch": 23.42,
      "learning_rate": 1.687715210355987e-05,
      "loss": 0.0763,
      "step": 60310
    },
    {
      "epoch": 23.43,
      "learning_rate": 1.687663430420712e-05,
      "loss": 0.2176,
      "step": 60320
    },
    {
      "epoch": 23.43,
      "learning_rate": 1.687611650485437e-05,
      "loss": 0.1148,
      "step": 60330
    },
    {
      "epoch": 23.43,
      "learning_rate": 1.687559870550162e-05,
      "loss": 0.1124,
      "step": 60340
    },
    {
      "epoch": 23.44,
      "learning_rate": 1.687508090614887e-05,
      "loss": 0.0379,
      "step": 60350
    },
    {
      "epoch": 23.44,
      "learning_rate": 1.6874563106796118e-05,
      "loss": 0.043,
      "step": 60360
    },
    {
      "epoch": 23.44,
      "learning_rate": 1.687404530744337e-05,
      "loss": 0.1121,
      "step": 60370
    },
    {
      "epoch": 23.45,
      "learning_rate": 1.6873527508090617e-05,
      "loss": 0.0255,
      "step": 60380
    },
    {
      "epoch": 23.45,
      "learning_rate": 1.6873009708737865e-05,
      "loss": 0.1132,
      "step": 60390
    },
    {
      "epoch": 23.46,
      "learning_rate": 1.6872491909385113e-05,
      "loss": 0.1593,
      "step": 60400
    },
    {
      "epoch": 23.46,
      "learning_rate": 1.6871974110032364e-05,
      "loss": 0.0571,
      "step": 60410
    },
    {
      "epoch": 23.46,
      "learning_rate": 1.6871456310679612e-05,
      "loss": 0.1191,
      "step": 60420
    },
    {
      "epoch": 23.47,
      "learning_rate": 1.6870938511326863e-05,
      "loss": 0.097,
      "step": 60430
    },
    {
      "epoch": 23.47,
      "learning_rate": 1.687042071197411e-05,
      "loss": 0.1338,
      "step": 60440
    },
    {
      "epoch": 23.48,
      "learning_rate": 1.686990291262136e-05,
      "loss": 0.1444,
      "step": 60450
    },
    {
      "epoch": 23.48,
      "learning_rate": 1.686938511326861e-05,
      "loss": 0.1539,
      "step": 60460
    },
    {
      "epoch": 23.48,
      "learning_rate": 1.686886731391586e-05,
      "loss": 0.0589,
      "step": 60470
    },
    {
      "epoch": 23.49,
      "learning_rate": 1.6868349514563106e-05,
      "loss": 0.1808,
      "step": 60480
    },
    {
      "epoch": 23.49,
      "learning_rate": 1.6867831715210358e-05,
      "loss": 0.1457,
      "step": 60490
    },
    {
      "epoch": 23.5,
      "learning_rate": 1.6867313915857606e-05,
      "loss": 0.1825,
      "step": 60500
    },
    {
      "epoch": 23.5,
      "learning_rate": 1.6866796116504857e-05,
      "loss": 0.0853,
      "step": 60510
    },
    {
      "epoch": 23.5,
      "learning_rate": 1.6866278317152105e-05,
      "loss": 0.0921,
      "step": 60520
    },
    {
      "epoch": 23.51,
      "learning_rate": 1.6865760517799353e-05,
      "loss": 0.1775,
      "step": 60530
    },
    {
      "epoch": 23.51,
      "learning_rate": 1.6865242718446604e-05,
      "loss": 0.1445,
      "step": 60540
    },
    {
      "epoch": 23.51,
      "learning_rate": 1.6864724919093852e-05,
      "loss": 0.0928,
      "step": 60550
    },
    {
      "epoch": 23.52,
      "learning_rate": 1.6864207119741103e-05,
      "loss": 0.2324,
      "step": 60560
    },
    {
      "epoch": 23.52,
      "learning_rate": 1.686368932038835e-05,
      "loss": 0.1454,
      "step": 60570
    },
    {
      "epoch": 23.53,
      "learning_rate": 1.68631715210356e-05,
      "loss": 0.1046,
      "step": 60580
    },
    {
      "epoch": 23.53,
      "learning_rate": 1.686265372168285e-05,
      "loss": 0.0706,
      "step": 60590
    },
    {
      "epoch": 23.53,
      "learning_rate": 1.68621359223301e-05,
      "loss": 0.0777,
      "step": 60600
    },
    {
      "epoch": 23.54,
      "learning_rate": 1.6861618122977346e-05,
      "loss": 0.1355,
      "step": 60610
    },
    {
      "epoch": 23.54,
      "learning_rate": 1.6861100323624598e-05,
      "loss": 0.0708,
      "step": 60620
    },
    {
      "epoch": 23.55,
      "learning_rate": 1.6860582524271846e-05,
      "loss": 0.0663,
      "step": 60630
    },
    {
      "epoch": 23.55,
      "learning_rate": 1.6860064724919097e-05,
      "loss": 0.1126,
      "step": 60640
    },
    {
      "epoch": 23.55,
      "learning_rate": 1.6859546925566345e-05,
      "loss": 0.1843,
      "step": 60650
    },
    {
      "epoch": 23.56,
      "learning_rate": 1.6859029126213593e-05,
      "loss": 0.1467,
      "step": 60660
    },
    {
      "epoch": 23.56,
      "learning_rate": 1.6858511326860844e-05,
      "loss": 0.1487,
      "step": 60670
    },
    {
      "epoch": 23.57,
      "learning_rate": 1.6857993527508092e-05,
      "loss": 0.0759,
      "step": 60680
    },
    {
      "epoch": 23.57,
      "learning_rate": 1.685747572815534e-05,
      "loss": 0.0957,
      "step": 60690
    },
    {
      "epoch": 23.57,
      "learning_rate": 1.685695792880259e-05,
      "loss": 0.0961,
      "step": 60700
    },
    {
      "epoch": 23.58,
      "learning_rate": 1.685644012944984e-05,
      "loss": 0.1028,
      "step": 60710
    },
    {
      "epoch": 23.58,
      "learning_rate": 1.685592233009709e-05,
      "loss": 0.0726,
      "step": 60720
    },
    {
      "epoch": 23.58,
      "learning_rate": 1.685540453074434e-05,
      "loss": 0.1256,
      "step": 60730
    },
    {
      "epoch": 23.59,
      "learning_rate": 1.6854886731391586e-05,
      "loss": 0.0361,
      "step": 60740
    },
    {
      "epoch": 23.59,
      "learning_rate": 1.6854368932038838e-05,
      "loss": 0.1173,
      "step": 60750
    },
    {
      "epoch": 23.6,
      "learning_rate": 1.6853851132686086e-05,
      "loss": 0.1279,
      "step": 60760
    },
    {
      "epoch": 23.6,
      "learning_rate": 1.6853333333333333e-05,
      "loss": 0.095,
      "step": 60770
    },
    {
      "epoch": 23.6,
      "learning_rate": 1.6852815533980585e-05,
      "loss": 0.1192,
      "step": 60780
    },
    {
      "epoch": 23.61,
      "learning_rate": 1.6852297734627833e-05,
      "loss": 0.0155,
      "step": 60790
    },
    {
      "epoch": 23.61,
      "learning_rate": 1.6851779935275084e-05,
      "loss": 0.1017,
      "step": 60800
    },
    {
      "epoch": 23.62,
      "learning_rate": 1.6851262135922332e-05,
      "loss": 0.0914,
      "step": 60810
    },
    {
      "epoch": 23.62,
      "learning_rate": 1.685074433656958e-05,
      "loss": 0.0601,
      "step": 60820
    },
    {
      "epoch": 23.62,
      "learning_rate": 1.6850226537216828e-05,
      "loss": 0.0479,
      "step": 60830
    },
    {
      "epoch": 23.63,
      "learning_rate": 1.684970873786408e-05,
      "loss": 0.0344,
      "step": 60840
    },
    {
      "epoch": 23.63,
      "learning_rate": 1.6849190938511327e-05,
      "loss": 0.1396,
      "step": 60850
    },
    {
      "epoch": 23.63,
      "learning_rate": 1.684867313915858e-05,
      "loss": 0.0618,
      "step": 60860
    },
    {
      "epoch": 23.64,
      "learning_rate": 1.6848155339805826e-05,
      "loss": 0.0494,
      "step": 60870
    },
    {
      "epoch": 23.64,
      "learning_rate": 1.6847637540453078e-05,
      "loss": 0.1461,
      "step": 60880
    },
    {
      "epoch": 23.65,
      "learning_rate": 1.6847119741100326e-05,
      "loss": 0.0476,
      "step": 60890
    },
    {
      "epoch": 23.65,
      "learning_rate": 1.6846601941747573e-05,
      "loss": 0.1925,
      "step": 60900
    },
    {
      "epoch": 23.65,
      "learning_rate": 1.684608414239482e-05,
      "loss": 0.1592,
      "step": 60910
    },
    {
      "epoch": 23.66,
      "learning_rate": 1.6845566343042073e-05,
      "loss": 0.1578,
      "step": 60920
    },
    {
      "epoch": 23.66,
      "learning_rate": 1.684504854368932e-05,
      "loss": 0.0588,
      "step": 60930
    },
    {
      "epoch": 23.67,
      "learning_rate": 1.6844530744336572e-05,
      "loss": 0.2325,
      "step": 60940
    },
    {
      "epoch": 23.67,
      "learning_rate": 1.684401294498382e-05,
      "loss": 0.21,
      "step": 60950
    },
    {
      "epoch": 23.67,
      "learning_rate": 1.684349514563107e-05,
      "loss": 0.1721,
      "step": 60960
    },
    {
      "epoch": 23.68,
      "learning_rate": 1.684297734627832e-05,
      "loss": 0.2849,
      "step": 60970
    },
    {
      "epoch": 23.68,
      "learning_rate": 1.6842459546925567e-05,
      "loss": 0.019,
      "step": 60980
    },
    {
      "epoch": 23.69,
      "learning_rate": 1.6841941747572815e-05,
      "loss": 0.1127,
      "step": 60990
    },
    {
      "epoch": 23.69,
      "learning_rate": 1.6841423948220066e-05,
      "loss": 0.1381,
      "step": 61000
    },
    {
      "epoch": 23.69,
      "learning_rate": 1.6840906148867314e-05,
      "loss": 0.1836,
      "step": 61010
    },
    {
      "epoch": 23.7,
      "learning_rate": 1.6840388349514565e-05,
      "loss": 0.0608,
      "step": 61020
    },
    {
      "epoch": 23.7,
      "learning_rate": 1.6839870550161813e-05,
      "loss": 0.0216,
      "step": 61030
    },
    {
      "epoch": 23.7,
      "learning_rate": 1.6839352750809065e-05,
      "loss": 0.1754,
      "step": 61040
    },
    {
      "epoch": 23.71,
      "learning_rate": 1.6838834951456313e-05,
      "loss": 0.1136,
      "step": 61050
    },
    {
      "epoch": 23.71,
      "learning_rate": 1.683831715210356e-05,
      "loss": 0.1453,
      "step": 61060
    },
    {
      "epoch": 23.72,
      "learning_rate": 1.683779935275081e-05,
      "loss": 0.2038,
      "step": 61070
    },
    {
      "epoch": 23.72,
      "learning_rate": 1.683728155339806e-05,
      "loss": 0.0265,
      "step": 61080
    },
    {
      "epoch": 23.72,
      "learning_rate": 1.6836763754045308e-05,
      "loss": 0.1537,
      "step": 61090
    },
    {
      "epoch": 23.73,
      "learning_rate": 1.683624595469256e-05,
      "loss": 0.2161,
      "step": 61100
    },
    {
      "epoch": 23.73,
      "learning_rate": 1.6835728155339807e-05,
      "loss": 0.1176,
      "step": 61110
    },
    {
      "epoch": 23.74,
      "learning_rate": 1.6835210355987058e-05,
      "loss": 0.0625,
      "step": 61120
    },
    {
      "epoch": 23.74,
      "learning_rate": 1.6834692556634306e-05,
      "loss": 0.0763,
      "step": 61130
    },
    {
      "epoch": 23.74,
      "learning_rate": 1.6834174757281554e-05,
      "loss": 0.1168,
      "step": 61140
    },
    {
      "epoch": 23.75,
      "learning_rate": 1.6833656957928802e-05,
      "loss": 0.0098,
      "step": 61150
    },
    {
      "epoch": 23.75,
      "learning_rate": 1.6833139158576053e-05,
      "loss": 0.1305,
      "step": 61160
    },
    {
      "epoch": 23.76,
      "learning_rate": 1.68326213592233e-05,
      "loss": 0.0783,
      "step": 61170
    },
    {
      "epoch": 23.76,
      "learning_rate": 1.6832103559870553e-05,
      "loss": 0.0913,
      "step": 61180
    },
    {
      "epoch": 23.76,
      "learning_rate": 1.68315857605178e-05,
      "loss": 0.1775,
      "step": 61190
    },
    {
      "epoch": 23.77,
      "learning_rate": 1.6831067961165052e-05,
      "loss": 0.0251,
      "step": 61200
    },
    {
      "epoch": 23.77,
      "learning_rate": 1.68305501618123e-05,
      "loss": 0.1903,
      "step": 61210
    },
    {
      "epoch": 23.77,
      "learning_rate": 1.6830032362459548e-05,
      "loss": 0.1029,
      "step": 61220
    },
    {
      "epoch": 23.78,
      "learning_rate": 1.6829514563106796e-05,
      "loss": 0.0956,
      "step": 61230
    },
    {
      "epoch": 23.78,
      "learning_rate": 1.6828996763754047e-05,
      "loss": 0.1863,
      "step": 61240
    },
    {
      "epoch": 23.79,
      "learning_rate": 1.6828478964401295e-05,
      "loss": 0.1047,
      "step": 61250
    },
    {
      "epoch": 23.79,
      "learning_rate": 1.6827961165048546e-05,
      "loss": 0.3181,
      "step": 61260
    },
    {
      "epoch": 23.79,
      "learning_rate": 1.6827443365695794e-05,
      "loss": 0.0563,
      "step": 61270
    },
    {
      "epoch": 23.8,
      "learning_rate": 1.6826925566343045e-05,
      "loss": 0.097,
      "step": 61280
    },
    {
      "epoch": 23.8,
      "learning_rate": 1.6826407766990293e-05,
      "loss": 0.1726,
      "step": 61290
    },
    {
      "epoch": 23.81,
      "learning_rate": 1.682588996763754e-05,
      "loss": 0.1518,
      "step": 61300
    },
    {
      "epoch": 23.81,
      "learning_rate": 1.682537216828479e-05,
      "loss": 0.1863,
      "step": 61310
    },
    {
      "epoch": 23.81,
      "learning_rate": 1.682485436893204e-05,
      "loss": 0.0856,
      "step": 61320
    },
    {
      "epoch": 23.82,
      "learning_rate": 1.682433656957929e-05,
      "loss": 0.1047,
      "step": 61330
    },
    {
      "epoch": 23.82,
      "learning_rate": 1.682381877022654e-05,
      "loss": 0.3357,
      "step": 61340
    },
    {
      "epoch": 23.83,
      "learning_rate": 1.6823300970873788e-05,
      "loss": 0.0987,
      "step": 61350
    },
    {
      "epoch": 23.83,
      "learning_rate": 1.682278317152104e-05,
      "loss": 0.2132,
      "step": 61360
    },
    {
      "epoch": 23.83,
      "learning_rate": 1.6822265372168287e-05,
      "loss": 0.0929,
      "step": 61370
    },
    {
      "epoch": 23.84,
      "learning_rate": 1.6821747572815535e-05,
      "loss": 0.032,
      "step": 61380
    },
    {
      "epoch": 23.84,
      "learning_rate": 1.6821229773462783e-05,
      "loss": 0.0101,
      "step": 61390
    },
    {
      "epoch": 23.84,
      "learning_rate": 1.6820711974110034e-05,
      "loss": 0.0833,
      "step": 61400
    },
    {
      "epoch": 23.85,
      "learning_rate": 1.6820194174757282e-05,
      "loss": 0.1,
      "step": 61410
    },
    {
      "epoch": 23.85,
      "learning_rate": 1.6819676375404533e-05,
      "loss": 0.0156,
      "step": 61420
    },
    {
      "epoch": 23.86,
      "learning_rate": 1.681915857605178e-05,
      "loss": 0.1022,
      "step": 61430
    },
    {
      "epoch": 23.86,
      "learning_rate": 1.6818640776699032e-05,
      "loss": 0.0226,
      "step": 61440
    },
    {
      "epoch": 23.86,
      "learning_rate": 1.681812297734628e-05,
      "loss": 0.119,
      "step": 61450
    },
    {
      "epoch": 23.87,
      "learning_rate": 1.681760517799353e-05,
      "loss": 0.1717,
      "step": 61460
    },
    {
      "epoch": 23.87,
      "learning_rate": 1.6817087378640776e-05,
      "loss": 0.0109,
      "step": 61470
    },
    {
      "epoch": 23.88,
      "learning_rate": 1.6816569579288028e-05,
      "loss": 0.1251,
      "step": 61480
    },
    {
      "epoch": 23.88,
      "learning_rate": 1.6816051779935275e-05,
      "loss": 0.2281,
      "step": 61490
    },
    {
      "epoch": 23.88,
      "learning_rate": 1.6815533980582527e-05,
      "loss": 0.0369,
      "step": 61500
    },
    {
      "epoch": 23.89,
      "learning_rate": 1.6815016181229775e-05,
      "loss": 0.1574,
      "step": 61510
    },
    {
      "epoch": 23.89,
      "learning_rate": 1.6814498381877026e-05,
      "loss": 0.1556,
      "step": 61520
    },
    {
      "epoch": 23.9,
      "learning_rate": 1.6813980582524274e-05,
      "loss": 0.1218,
      "step": 61530
    },
    {
      "epoch": 23.9,
      "learning_rate": 1.6813462783171522e-05,
      "loss": 0.1425,
      "step": 61540
    },
    {
      "epoch": 23.9,
      "learning_rate": 1.681294498381877e-05,
      "loss": 0.1045,
      "step": 61550
    },
    {
      "epoch": 23.91,
      "learning_rate": 1.681242718446602e-05,
      "loss": 0.1001,
      "step": 61560
    },
    {
      "epoch": 23.91,
      "learning_rate": 1.681190938511327e-05,
      "loss": 0.0289,
      "step": 61570
    },
    {
      "epoch": 23.91,
      "learning_rate": 1.681139158576052e-05,
      "loss": 0.0516,
      "step": 61580
    },
    {
      "epoch": 23.92,
      "learning_rate": 1.6810873786407768e-05,
      "loss": 0.1593,
      "step": 61590
    },
    {
      "epoch": 23.92,
      "learning_rate": 1.681035598705502e-05,
      "loss": 0.1676,
      "step": 61600
    },
    {
      "epoch": 23.93,
      "learning_rate": 1.6809838187702268e-05,
      "loss": 0.1557,
      "step": 61610
    },
    {
      "epoch": 23.93,
      "learning_rate": 1.6809320388349515e-05,
      "loss": 0.0409,
      "step": 61620
    },
    {
      "epoch": 23.93,
      "learning_rate": 1.6808802588996763e-05,
      "loss": 0.1406,
      "step": 61630
    },
    {
      "epoch": 23.94,
      "learning_rate": 1.6808284789644015e-05,
      "loss": 0.1498,
      "step": 61640
    },
    {
      "epoch": 23.94,
      "learning_rate": 1.6807766990291263e-05,
      "loss": 0.115,
      "step": 61650
    },
    {
      "epoch": 23.95,
      "learning_rate": 1.6807249190938514e-05,
      "loss": 0.1659,
      "step": 61660
    },
    {
      "epoch": 23.95,
      "learning_rate": 1.6806731391585762e-05,
      "loss": 0.2063,
      "step": 61670
    },
    {
      "epoch": 23.95,
      "learning_rate": 1.6806213592233013e-05,
      "loss": 0.0548,
      "step": 61680
    },
    {
      "epoch": 23.96,
      "learning_rate": 1.680569579288026e-05,
      "loss": 0.1658,
      "step": 61690
    },
    {
      "epoch": 23.96,
      "learning_rate": 1.680517799352751e-05,
      "loss": 0.1888,
      "step": 61700
    },
    {
      "epoch": 23.97,
      "learning_rate": 1.6804660194174757e-05,
      "loss": 0.0352,
      "step": 61710
    },
    {
      "epoch": 23.97,
      "learning_rate": 1.6804142394822008e-05,
      "loss": 0.0776,
      "step": 61720
    },
    {
      "epoch": 23.97,
      "learning_rate": 1.6803624595469256e-05,
      "loss": 0.0365,
      "step": 61730
    },
    {
      "epoch": 23.98,
      "learning_rate": 1.6803106796116507e-05,
      "loss": 0.0237,
      "step": 61740
    },
    {
      "epoch": 23.98,
      "learning_rate": 1.6802588996763755e-05,
      "loss": 0.1171,
      "step": 61750
    },
    {
      "epoch": 23.98,
      "learning_rate": 1.6802071197411007e-05,
      "loss": 0.0952,
      "step": 61760
    },
    {
      "epoch": 23.99,
      "learning_rate": 1.6801553398058255e-05,
      "loss": 0.2077,
      "step": 61770
    },
    {
      "epoch": 23.99,
      "learning_rate": 1.6801035598705503e-05,
      "loss": 0.0616,
      "step": 61780
    },
    {
      "epoch": 24.0,
      "learning_rate": 1.680051779935275e-05,
      "loss": 0.084,
      "step": 61790
    },
    {
      "epoch": 24.0,
      "learning_rate": 1.6800000000000002e-05,
      "loss": 0.1737,
      "step": 61800
    },
    {
      "epoch": 24.0,
      "eval_accuracy": 0.9540577716643741,
      "eval_loss": 0.2604363262653351,
      "eval_runtime": 8.2035,
      "eval_samples_per_second": 443.105,
      "eval_steps_per_second": 55.464,
      "step": 61800
    },
    {
      "epoch": 24.0,
      "learning_rate": 1.679948220064725e-05,
      "loss": 0.0809,
      "step": 61810
    },
    {
      "epoch": 24.01,
      "learning_rate": 1.67989644012945e-05,
      "loss": 0.0872,
      "step": 61820
    },
    {
      "epoch": 24.01,
      "learning_rate": 1.679844660194175e-05,
      "loss": 0.1132,
      "step": 61830
    },
    {
      "epoch": 24.02,
      "learning_rate": 1.6797928802589e-05,
      "loss": 0.0593,
      "step": 61840
    },
    {
      "epoch": 24.02,
      "learning_rate": 1.6797411003236248e-05,
      "loss": 0.0435,
      "step": 61850
    },
    {
      "epoch": 24.02,
      "learning_rate": 1.6796893203883496e-05,
      "loss": 0.1239,
      "step": 61860
    },
    {
      "epoch": 24.03,
      "learning_rate": 1.6796375404530744e-05,
      "loss": 0.1746,
      "step": 61870
    },
    {
      "epoch": 24.03,
      "learning_rate": 1.6795857605177995e-05,
      "loss": 0.04,
      "step": 61880
    },
    {
      "epoch": 24.03,
      "learning_rate": 1.6795339805825243e-05,
      "loss": 0.0976,
      "step": 61890
    },
    {
      "epoch": 24.04,
      "learning_rate": 1.6794822006472495e-05,
      "loss": 0.0564,
      "step": 61900
    },
    {
      "epoch": 24.04,
      "learning_rate": 1.6794304207119742e-05,
      "loss": 0.0306,
      "step": 61910
    },
    {
      "epoch": 24.05,
      "learning_rate": 1.679378640776699e-05,
      "loss": 0.092,
      "step": 61920
    },
    {
      "epoch": 24.05,
      "learning_rate": 1.6793268608414242e-05,
      "loss": 0.2796,
      "step": 61930
    },
    {
      "epoch": 24.05,
      "learning_rate": 1.679275080906149e-05,
      "loss": 0.024,
      "step": 61940
    },
    {
      "epoch": 24.06,
      "learning_rate": 1.6792233009708738e-05,
      "loss": 0.1511,
      "step": 61950
    },
    {
      "epoch": 24.06,
      "learning_rate": 1.679171521035599e-05,
      "loss": 0.0632,
      "step": 61960
    },
    {
      "epoch": 24.07,
      "learning_rate": 1.6791197411003237e-05,
      "loss": 0.1061,
      "step": 61970
    },
    {
      "epoch": 24.07,
      "learning_rate": 1.6790679611650488e-05,
      "loss": 0.0479,
      "step": 61980
    },
    {
      "epoch": 24.07,
      "learning_rate": 1.6790161812297736e-05,
      "loss": 0.081,
      "step": 61990
    },
    {
      "epoch": 24.08,
      "learning_rate": 1.6789644012944984e-05,
      "loss": 0.0802,
      "step": 62000
    },
    {
      "epoch": 24.08,
      "learning_rate": 1.6789126213592235e-05,
      "loss": 0.126,
      "step": 62010
    },
    {
      "epoch": 24.09,
      "learning_rate": 1.6788608414239483e-05,
      "loss": 0.0588,
      "step": 62020
    },
    {
      "epoch": 24.09,
      "learning_rate": 1.678809061488673e-05,
      "loss": 0.0035,
      "step": 62030
    },
    {
      "epoch": 24.09,
      "learning_rate": 1.6787572815533982e-05,
      "loss": 0.1253,
      "step": 62040
    },
    {
      "epoch": 24.1,
      "learning_rate": 1.678705501618123e-05,
      "loss": 0.1538,
      "step": 62050
    },
    {
      "epoch": 24.1,
      "learning_rate": 1.678653721682848e-05,
      "loss": 0.0946,
      "step": 62060
    },
    {
      "epoch": 24.1,
      "learning_rate": 1.678601941747573e-05,
      "loss": 0.2093,
      "step": 62070
    },
    {
      "epoch": 24.11,
      "learning_rate": 1.6785501618122978e-05,
      "loss": 0.0491,
      "step": 62080
    },
    {
      "epoch": 24.11,
      "learning_rate": 1.678498381877023e-05,
      "loss": 0.1637,
      "step": 62090
    },
    {
      "epoch": 24.12,
      "learning_rate": 1.6784466019417477e-05,
      "loss": 0.1581,
      "step": 62100
    },
    {
      "epoch": 24.12,
      "learning_rate": 1.6783948220064725e-05,
      "loss": 0.082,
      "step": 62110
    },
    {
      "epoch": 24.12,
      "learning_rate": 1.6783430420711976e-05,
      "loss": 0.0302,
      "step": 62120
    },
    {
      "epoch": 24.13,
      "learning_rate": 1.6782912621359224e-05,
      "loss": 0.21,
      "step": 62130
    },
    {
      "epoch": 24.13,
      "learning_rate": 1.6782394822006475e-05,
      "loss": 0.0802,
      "step": 62140
    },
    {
      "epoch": 24.14,
      "learning_rate": 1.6781877022653723e-05,
      "loss": 0.1112,
      "step": 62150
    },
    {
      "epoch": 24.14,
      "learning_rate": 1.678135922330097e-05,
      "loss": 0.1191,
      "step": 62160
    },
    {
      "epoch": 24.14,
      "learning_rate": 1.6780841423948222e-05,
      "loss": 0.1874,
      "step": 62170
    },
    {
      "epoch": 24.15,
      "learning_rate": 1.678032362459547e-05,
      "loss": 0.0778,
      "step": 62180
    },
    {
      "epoch": 24.15,
      "learning_rate": 1.6779805825242718e-05,
      "loss": 0.2343,
      "step": 62190
    },
    {
      "epoch": 24.16,
      "learning_rate": 1.677928802588997e-05,
      "loss": 0.08,
      "step": 62200
    },
    {
      "epoch": 24.16,
      "learning_rate": 1.6778770226537217e-05,
      "loss": 0.351,
      "step": 62210
    },
    {
      "epoch": 24.16,
      "learning_rate": 1.677825242718447e-05,
      "loss": 0.1439,
      "step": 62220
    },
    {
      "epoch": 24.17,
      "learning_rate": 1.6777734627831717e-05,
      "loss": 0.0431,
      "step": 62230
    },
    {
      "epoch": 24.17,
      "learning_rate": 1.6777216828478965e-05,
      "loss": 0.1222,
      "step": 62240
    },
    {
      "epoch": 24.17,
      "learning_rate": 1.6776699029126216e-05,
      "loss": 0.1403,
      "step": 62250
    },
    {
      "epoch": 24.18,
      "learning_rate": 1.6776181229773464e-05,
      "loss": 0.0754,
      "step": 62260
    },
    {
      "epoch": 24.18,
      "learning_rate": 1.6775663430420715e-05,
      "loss": 0.1512,
      "step": 62270
    },
    {
      "epoch": 24.19,
      "learning_rate": 1.6775145631067963e-05,
      "loss": 0.0886,
      "step": 62280
    },
    {
      "epoch": 24.19,
      "learning_rate": 1.677462783171521e-05,
      "loss": 0.1103,
      "step": 62290
    },
    {
      "epoch": 24.19,
      "learning_rate": 1.677411003236246e-05,
      "loss": 0.0549,
      "step": 62300
    },
    {
      "epoch": 24.2,
      "learning_rate": 1.677359223300971e-05,
      "loss": 0.0647,
      "step": 62310
    },
    {
      "epoch": 24.2,
      "learning_rate": 1.6773074433656958e-05,
      "loss": 0.0281,
      "step": 62320
    },
    {
      "epoch": 24.21,
      "learning_rate": 1.677255663430421e-05,
      "loss": 0.1221,
      "step": 62330
    },
    {
      "epoch": 24.21,
      "learning_rate": 1.6772038834951457e-05,
      "loss": 0.0897,
      "step": 62340
    },
    {
      "epoch": 24.21,
      "learning_rate": 1.677152103559871e-05,
      "loss": 0.2709,
      "step": 62350
    },
    {
      "epoch": 24.22,
      "learning_rate": 1.6771003236245957e-05,
      "loss": 0.0977,
      "step": 62360
    },
    {
      "epoch": 24.22,
      "learning_rate": 1.6770485436893205e-05,
      "loss": 0.1195,
      "step": 62370
    },
    {
      "epoch": 24.23,
      "learning_rate": 1.6769967637540453e-05,
      "loss": 0.1038,
      "step": 62380
    },
    {
      "epoch": 24.23,
      "learning_rate": 1.6769449838187704e-05,
      "loss": 0.0849,
      "step": 62390
    },
    {
      "epoch": 24.23,
      "learning_rate": 1.6768932038834952e-05,
      "loss": 0.0767,
      "step": 62400
    },
    {
      "epoch": 24.24,
      "learning_rate": 1.6768414239482203e-05,
      "loss": 0.0058,
      "step": 62410
    },
    {
      "epoch": 24.24,
      "learning_rate": 1.676789644012945e-05,
      "loss": 0.0556,
      "step": 62420
    },
    {
      "epoch": 24.24,
      "learning_rate": 1.6767378640776702e-05,
      "loss": 0.1618,
      "step": 62430
    },
    {
      "epoch": 24.25,
      "learning_rate": 1.676686084142395e-05,
      "loss": 0.0622,
      "step": 62440
    },
    {
      "epoch": 24.25,
      "learning_rate": 1.6766343042071198e-05,
      "loss": 0.3328,
      "step": 62450
    },
    {
      "epoch": 24.26,
      "learning_rate": 1.6765825242718446e-05,
      "loss": 0.1226,
      "step": 62460
    },
    {
      "epoch": 24.26,
      "learning_rate": 1.6765307443365697e-05,
      "loss": 0.1369,
      "step": 62470
    },
    {
      "epoch": 24.26,
      "learning_rate": 1.6764789644012945e-05,
      "loss": 0.0765,
      "step": 62480
    },
    {
      "epoch": 24.27,
      "learning_rate": 1.6764271844660197e-05,
      "loss": 0.0235,
      "step": 62490
    },
    {
      "epoch": 24.27,
      "learning_rate": 1.6763754045307445e-05,
      "loss": 0.1501,
      "step": 62500
    },
    {
      "epoch": 24.28,
      "learning_rate": 1.6763236245954696e-05,
      "loss": 0.0727,
      "step": 62510
    },
    {
      "epoch": 24.28,
      "learning_rate": 1.6762718446601944e-05,
      "loss": 0.1114,
      "step": 62520
    },
    {
      "epoch": 24.28,
      "learning_rate": 1.6762200647249192e-05,
      "loss": 0.1382,
      "step": 62530
    },
    {
      "epoch": 24.29,
      "learning_rate": 1.676168284789644e-05,
      "loss": 0.0102,
      "step": 62540
    },
    {
      "epoch": 24.29,
      "learning_rate": 1.676116504854369e-05,
      "loss": 0.0511,
      "step": 62550
    },
    {
      "epoch": 24.3,
      "learning_rate": 1.676064724919094e-05,
      "loss": 0.0948,
      "step": 62560
    },
    {
      "epoch": 24.3,
      "learning_rate": 1.676012944983819e-05,
      "loss": 0.055,
      "step": 62570
    },
    {
      "epoch": 24.3,
      "learning_rate": 1.6759611650485438e-05,
      "loss": 0.0788,
      "step": 62580
    },
    {
      "epoch": 24.31,
      "learning_rate": 1.675909385113269e-05,
      "loss": 0.0319,
      "step": 62590
    },
    {
      "epoch": 24.31,
      "learning_rate": 1.6758576051779934e-05,
      "loss": 0.1356,
      "step": 62600
    },
    {
      "epoch": 24.31,
      "learning_rate": 1.6758058252427185e-05,
      "loss": 0.0786,
      "step": 62610
    },
    {
      "epoch": 24.32,
      "learning_rate": 1.6757540453074433e-05,
      "loss": 0.2069,
      "step": 62620
    },
    {
      "epoch": 24.32,
      "learning_rate": 1.6757022653721685e-05,
      "loss": 0.0171,
      "step": 62630
    },
    {
      "epoch": 24.33,
      "learning_rate": 1.6756504854368932e-05,
      "loss": 0.0735,
      "step": 62640
    },
    {
      "epoch": 24.33,
      "learning_rate": 1.6755987055016184e-05,
      "loss": 0.0013,
      "step": 62650
    },
    {
      "epoch": 24.33,
      "learning_rate": 1.675546925566343e-05,
      "loss": 0.0037,
      "step": 62660
    },
    {
      "epoch": 24.34,
      "learning_rate": 1.6754951456310683e-05,
      "loss": 0.0525,
      "step": 62670
    },
    {
      "epoch": 24.34,
      "learning_rate": 1.6754433656957928e-05,
      "loss": 0.105,
      "step": 62680
    },
    {
      "epoch": 24.35,
      "learning_rate": 1.675391585760518e-05,
      "loss": 0.0305,
      "step": 62690
    },
    {
      "epoch": 24.35,
      "learning_rate": 1.6753398058252427e-05,
      "loss": 0.1546,
      "step": 62700
    },
    {
      "epoch": 24.35,
      "learning_rate": 1.6752880258899678e-05,
      "loss": 0.1686,
      "step": 62710
    },
    {
      "epoch": 24.36,
      "learning_rate": 1.6752362459546926e-05,
      "loss": 0.1888,
      "step": 62720
    },
    {
      "epoch": 24.36,
      "learning_rate": 1.6751844660194177e-05,
      "loss": 0.0997,
      "step": 62730
    },
    {
      "epoch": 24.37,
      "learning_rate": 1.6751326860841425e-05,
      "loss": 0.1032,
      "step": 62740
    },
    {
      "epoch": 24.37,
      "learning_rate": 1.6750809061488677e-05,
      "loss": 0.0866,
      "step": 62750
    },
    {
      "epoch": 24.37,
      "learning_rate": 1.675029126213592e-05,
      "loss": 0.0618,
      "step": 62760
    },
    {
      "epoch": 24.38,
      "learning_rate": 1.6749773462783172e-05,
      "loss": 0.1807,
      "step": 62770
    },
    {
      "epoch": 24.38,
      "learning_rate": 1.674925566343042e-05,
      "loss": 0.2234,
      "step": 62780
    },
    {
      "epoch": 24.38,
      "learning_rate": 1.674873786407767e-05,
      "loss": 0.0442,
      "step": 62790
    },
    {
      "epoch": 24.39,
      "learning_rate": 1.674822006472492e-05,
      "loss": 0.1047,
      "step": 62800
    },
    {
      "epoch": 24.39,
      "learning_rate": 1.674770226537217e-05,
      "loss": 0.0304,
      "step": 62810
    },
    {
      "epoch": 24.4,
      "learning_rate": 1.674718446601942e-05,
      "loss": 0.1364,
      "step": 62820
    },
    {
      "epoch": 24.4,
      "learning_rate": 1.674666666666667e-05,
      "loss": 0.0706,
      "step": 62830
    },
    {
      "epoch": 24.4,
      "learning_rate": 1.6746148867313918e-05,
      "loss": 0.0764,
      "step": 62840
    },
    {
      "epoch": 24.41,
      "learning_rate": 1.6745631067961166e-05,
      "loss": 0.1014,
      "step": 62850
    },
    {
      "epoch": 24.41,
      "learning_rate": 1.6745113268608414e-05,
      "loss": 0.1057,
      "step": 62860
    },
    {
      "epoch": 24.42,
      "learning_rate": 1.6744595469255665e-05,
      "loss": 0.1007,
      "step": 62870
    },
    {
      "epoch": 24.42,
      "learning_rate": 1.6744077669902913e-05,
      "loss": 0.0548,
      "step": 62880
    },
    {
      "epoch": 24.42,
      "learning_rate": 1.6743559870550164e-05,
      "loss": 0.244,
      "step": 62890
    },
    {
      "epoch": 24.43,
      "learning_rate": 1.6743042071197412e-05,
      "loss": 0.0481,
      "step": 62900
    },
    {
      "epoch": 24.43,
      "learning_rate": 1.6742524271844664e-05,
      "loss": 0.1096,
      "step": 62910
    },
    {
      "epoch": 24.43,
      "learning_rate": 1.674200647249191e-05,
      "loss": 0.1351,
      "step": 62920
    },
    {
      "epoch": 24.44,
      "learning_rate": 1.674148867313916e-05,
      "loss": 0.2504,
      "step": 62930
    },
    {
      "epoch": 24.44,
      "learning_rate": 1.6740970873786407e-05,
      "loss": 0.0355,
      "step": 62940
    },
    {
      "epoch": 24.45,
      "learning_rate": 1.674045307443366e-05,
      "loss": 0.1462,
      "step": 62950
    },
    {
      "epoch": 24.45,
      "learning_rate": 1.6739935275080907e-05,
      "loss": 0.0376,
      "step": 62960
    },
    {
      "epoch": 24.45,
      "learning_rate": 1.6739417475728158e-05,
      "loss": 0.1971,
      "step": 62970
    },
    {
      "epoch": 24.46,
      "learning_rate": 1.6738899676375406e-05,
      "loss": 0.0717,
      "step": 62980
    },
    {
      "epoch": 24.46,
      "learning_rate": 1.6738381877022657e-05,
      "loss": 0.1046,
      "step": 62990
    },
    {
      "epoch": 24.47,
      "learning_rate": 1.6737864077669905e-05,
      "loss": 0.0817,
      "step": 63000
    },
    {
      "epoch": 24.47,
      "learning_rate": 1.6737346278317153e-05,
      "loss": 0.0532,
      "step": 63010
    },
    {
      "epoch": 24.47,
      "learning_rate": 1.67368284789644e-05,
      "loss": 0.0794,
      "step": 63020
    },
    {
      "epoch": 24.48,
      "learning_rate": 1.6736310679611652e-05,
      "loss": 0.0313,
      "step": 63030
    },
    {
      "epoch": 24.48,
      "learning_rate": 1.67357928802589e-05,
      "loss": 0.0028,
      "step": 63040
    },
    {
      "epoch": 24.49,
      "learning_rate": 1.673527508090615e-05,
      "loss": 0.019,
      "step": 63050
    },
    {
      "epoch": 24.49,
      "learning_rate": 1.67347572815534e-05,
      "loss": 0.1981,
      "step": 63060
    },
    {
      "epoch": 24.49,
      "learning_rate": 1.673423948220065e-05,
      "loss": 0.2145,
      "step": 63070
    },
    {
      "epoch": 24.5,
      "learning_rate": 1.67337216828479e-05,
      "loss": 0.0765,
      "step": 63080
    },
    {
      "epoch": 24.5,
      "learning_rate": 1.6733203883495147e-05,
      "loss": 0.0076,
      "step": 63090
    },
    {
      "epoch": 24.5,
      "learning_rate": 1.6732686084142395e-05,
      "loss": 0.1483,
      "step": 63100
    },
    {
      "epoch": 24.51,
      "learning_rate": 1.6732168284789646e-05,
      "loss": 0.2438,
      "step": 63110
    },
    {
      "epoch": 24.51,
      "learning_rate": 1.6731650485436894e-05,
      "loss": 0.047,
      "step": 63120
    },
    {
      "epoch": 24.52,
      "learning_rate": 1.6731132686084145e-05,
      "loss": 0.1352,
      "step": 63130
    },
    {
      "epoch": 24.52,
      "learning_rate": 1.6730614886731393e-05,
      "loss": 0.2002,
      "step": 63140
    },
    {
      "epoch": 24.52,
      "learning_rate": 1.6730097087378644e-05,
      "loss": 0.2107,
      "step": 63150
    },
    {
      "epoch": 24.53,
      "learning_rate": 1.6729579288025892e-05,
      "loss": 0.0021,
      "step": 63160
    },
    {
      "epoch": 24.53,
      "learning_rate": 1.672906148867314e-05,
      "loss": 0.0722,
      "step": 63170
    },
    {
      "epoch": 24.54,
      "learning_rate": 1.6728543689320388e-05,
      "loss": 0.1017,
      "step": 63180
    },
    {
      "epoch": 24.54,
      "learning_rate": 1.672802588996764e-05,
      "loss": 0.064,
      "step": 63190
    },
    {
      "epoch": 24.54,
      "learning_rate": 1.6727508090614887e-05,
      "loss": 0.0644,
      "step": 63200
    },
    {
      "epoch": 24.55,
      "learning_rate": 1.672699029126214e-05,
      "loss": 0.1606,
      "step": 63210
    },
    {
      "epoch": 24.55,
      "learning_rate": 1.6726472491909387e-05,
      "loss": 0.1427,
      "step": 63220
    },
    {
      "epoch": 24.56,
      "learning_rate": 1.6725954692556638e-05,
      "loss": 0.0585,
      "step": 63230
    },
    {
      "epoch": 24.56,
      "learning_rate": 1.6725436893203886e-05,
      "loss": 0.0679,
      "step": 63240
    },
    {
      "epoch": 24.56,
      "learning_rate": 1.6724919093851134e-05,
      "loss": 0.2347,
      "step": 63250
    },
    {
      "epoch": 24.57,
      "learning_rate": 1.672440129449838e-05,
      "loss": 0.1291,
      "step": 63260
    },
    {
      "epoch": 24.57,
      "learning_rate": 1.6723883495145633e-05,
      "loss": 0.0709,
      "step": 63270
    },
    {
      "epoch": 24.57,
      "learning_rate": 1.672336569579288e-05,
      "loss": 0.0474,
      "step": 63280
    },
    {
      "epoch": 24.58,
      "learning_rate": 1.6722847896440132e-05,
      "loss": 0.296,
      "step": 63290
    },
    {
      "epoch": 24.58,
      "learning_rate": 1.672233009708738e-05,
      "loss": 0.1627,
      "step": 63300
    },
    {
      "epoch": 24.59,
      "learning_rate": 1.672181229773463e-05,
      "loss": 0.1216,
      "step": 63310
    },
    {
      "epoch": 24.59,
      "learning_rate": 1.672129449838188e-05,
      "loss": 0.0521,
      "step": 63320
    },
    {
      "epoch": 24.59,
      "learning_rate": 1.6720776699029127e-05,
      "loss": 0.1048,
      "step": 63330
    },
    {
      "epoch": 24.6,
      "learning_rate": 1.6720258899676375e-05,
      "loss": 0.0313,
      "step": 63340
    },
    {
      "epoch": 24.6,
      "learning_rate": 1.6719741100323627e-05,
      "loss": 0.1454,
      "step": 63350
    },
    {
      "epoch": 24.61,
      "learning_rate": 1.6719223300970874e-05,
      "loss": 0.1352,
      "step": 63360
    },
    {
      "epoch": 24.61,
      "learning_rate": 1.6718705501618126e-05,
      "loss": 0.1288,
      "step": 63370
    },
    {
      "epoch": 24.61,
      "learning_rate": 1.6718187702265374e-05,
      "loss": 0.1786,
      "step": 63380
    },
    {
      "epoch": 24.62,
      "learning_rate": 1.671766990291262e-05,
      "loss": 0.2206,
      "step": 63390
    },
    {
      "epoch": 24.62,
      "learning_rate": 1.6717152103559873e-05,
      "loss": 0.1606,
      "step": 63400
    },
    {
      "epoch": 24.63,
      "learning_rate": 1.671663430420712e-05,
      "loss": 0.0036,
      "step": 63410
    },
    {
      "epoch": 24.63,
      "learning_rate": 1.671611650485437e-05,
      "loss": 0.1591,
      "step": 63420
    },
    {
      "epoch": 24.63,
      "learning_rate": 1.671559870550162e-05,
      "loss": 0.0335,
      "step": 63430
    },
    {
      "epoch": 24.64,
      "learning_rate": 1.6715080906148868e-05,
      "loss": 0.0819,
      "step": 63440
    },
    {
      "epoch": 24.64,
      "learning_rate": 1.671456310679612e-05,
      "loss": 0.0852,
      "step": 63450
    },
    {
      "epoch": 24.64,
      "learning_rate": 1.6714045307443367e-05,
      "loss": 0.1022,
      "step": 63460
    },
    {
      "epoch": 24.65,
      "learning_rate": 1.6713527508090615e-05,
      "loss": 0.0267,
      "step": 63470
    },
    {
      "epoch": 24.65,
      "learning_rate": 1.6713009708737866e-05,
      "loss": 0.0603,
      "step": 63480
    },
    {
      "epoch": 24.66,
      "learning_rate": 1.6712491909385114e-05,
      "loss": 0.2353,
      "step": 63490
    },
    {
      "epoch": 24.66,
      "learning_rate": 1.6711974110032362e-05,
      "loss": 0.0822,
      "step": 63500
    },
    {
      "epoch": 24.66,
      "learning_rate": 1.6711456310679614e-05,
      "loss": 0.0855,
      "step": 63510
    },
    {
      "epoch": 24.67,
      "learning_rate": 1.671093851132686e-05,
      "loss": 0.1538,
      "step": 63520
    },
    {
      "epoch": 24.67,
      "learning_rate": 1.6710420711974113e-05,
      "loss": 0.0849,
      "step": 63530
    },
    {
      "epoch": 24.68,
      "learning_rate": 1.670990291262136e-05,
      "loss": 0.0719,
      "step": 63540
    },
    {
      "epoch": 24.68,
      "learning_rate": 1.670938511326861e-05,
      "loss": 0.1341,
      "step": 63550
    },
    {
      "epoch": 24.68,
      "learning_rate": 1.670886731391586e-05,
      "loss": 0.1292,
      "step": 63560
    },
    {
      "epoch": 24.69,
      "learning_rate": 1.6708349514563108e-05,
      "loss": 0.0485,
      "step": 63570
    },
    {
      "epoch": 24.69,
      "learning_rate": 1.6707831715210356e-05,
      "loss": 0.2329,
      "step": 63580
    },
    {
      "epoch": 24.7,
      "learning_rate": 1.6707313915857607e-05,
      "loss": 0.017,
      "step": 63590
    },
    {
      "epoch": 24.7,
      "learning_rate": 1.6706796116504855e-05,
      "loss": 0.0549,
      "step": 63600
    },
    {
      "epoch": 24.7,
      "learning_rate": 1.6706278317152106e-05,
      "loss": 0.0463,
      "step": 63610
    },
    {
      "epoch": 24.71,
      "learning_rate": 1.6705760517799354e-05,
      "loss": 0.0924,
      "step": 63620
    },
    {
      "epoch": 24.71,
      "learning_rate": 1.6705242718446602e-05,
      "loss": 0.0588,
      "step": 63630
    },
    {
      "epoch": 24.71,
      "learning_rate": 1.6704724919093854e-05,
      "loss": 0.0534,
      "step": 63640
    },
    {
      "epoch": 24.72,
      "learning_rate": 1.67042071197411e-05,
      "loss": 0.1262,
      "step": 63650
    },
    {
      "epoch": 24.72,
      "learning_rate": 1.670368932038835e-05,
      "loss": 0.1905,
      "step": 63660
    },
    {
      "epoch": 24.73,
      "learning_rate": 1.67031715210356e-05,
      "loss": 0.0144,
      "step": 63670
    },
    {
      "epoch": 24.73,
      "learning_rate": 1.670265372168285e-05,
      "loss": 0.1623,
      "step": 63680
    },
    {
      "epoch": 24.73,
      "learning_rate": 1.67021359223301e-05,
      "loss": 0.173,
      "step": 63690
    },
    {
      "epoch": 24.74,
      "learning_rate": 1.6701618122977348e-05,
      "loss": 0.1667,
      "step": 63700
    },
    {
      "epoch": 24.74,
      "learning_rate": 1.6701100323624596e-05,
      "loss": 0.206,
      "step": 63710
    },
    {
      "epoch": 24.75,
      "learning_rate": 1.6700582524271847e-05,
      "loss": 0.0932,
      "step": 63720
    },
    {
      "epoch": 24.75,
      "learning_rate": 1.6700064724919095e-05,
      "loss": 0.138,
      "step": 63730
    },
    {
      "epoch": 24.75,
      "learning_rate": 1.6699546925566343e-05,
      "loss": 0.1341,
      "step": 63740
    },
    {
      "epoch": 24.76,
      "learning_rate": 1.6699029126213594e-05,
      "loss": 0.1059,
      "step": 63750
    },
    {
      "epoch": 24.76,
      "learning_rate": 1.6698511326860842e-05,
      "loss": 0.0523,
      "step": 63760
    },
    {
      "epoch": 24.77,
      "learning_rate": 1.669799352750809e-05,
      "loss": 0.081,
      "step": 63770
    },
    {
      "epoch": 24.77,
      "learning_rate": 1.669747572815534e-05,
      "loss": 0.1879,
      "step": 63780
    },
    {
      "epoch": 24.77,
      "learning_rate": 1.669695792880259e-05,
      "loss": 0.0754,
      "step": 63790
    },
    {
      "epoch": 24.78,
      "learning_rate": 1.669644012944984e-05,
      "loss": 0.0109,
      "step": 63800
    },
    {
      "epoch": 24.78,
      "learning_rate": 1.669592233009709e-05,
      "loss": 0.0914,
      "step": 63810
    },
    {
      "epoch": 24.78,
      "learning_rate": 1.6695404530744337e-05,
      "loss": 0.1208,
      "step": 63820
    },
    {
      "epoch": 24.79,
      "learning_rate": 1.6694886731391588e-05,
      "loss": 0.0261,
      "step": 63830
    },
    {
      "epoch": 24.79,
      "learning_rate": 1.6694368932038836e-05,
      "loss": 0.1462,
      "step": 63840
    },
    {
      "epoch": 24.8,
      "learning_rate": 1.6693851132686084e-05,
      "loss": 0.1745,
      "step": 63850
    },
    {
      "epoch": 24.8,
      "learning_rate": 1.6693333333333335e-05,
      "loss": 0.0653,
      "step": 63860
    },
    {
      "epoch": 24.8,
      "learning_rate": 1.6692815533980583e-05,
      "loss": 0.0236,
      "step": 63870
    },
    {
      "epoch": 24.81,
      "learning_rate": 1.6692297734627834e-05,
      "loss": 0.1402,
      "step": 63880
    },
    {
      "epoch": 24.81,
      "learning_rate": 1.6691779935275082e-05,
      "loss": 0.2828,
      "step": 63890
    },
    {
      "epoch": 24.82,
      "learning_rate": 1.669126213592233e-05,
      "loss": 0.1055,
      "step": 63900
    },
    {
      "epoch": 24.82,
      "learning_rate": 1.669074433656958e-05,
      "loss": 0.0644,
      "step": 63910
    },
    {
      "epoch": 24.82,
      "learning_rate": 1.669022653721683e-05,
      "loss": 0.1536,
      "step": 63920
    },
    {
      "epoch": 24.83,
      "learning_rate": 1.6689708737864077e-05,
      "loss": 0.1238,
      "step": 63930
    },
    {
      "epoch": 24.83,
      "learning_rate": 1.668919093851133e-05,
      "loss": 0.1705,
      "step": 63940
    },
    {
      "epoch": 24.83,
      "learning_rate": 1.6688673139158576e-05,
      "loss": 0.1166,
      "step": 63950
    },
    {
      "epoch": 24.84,
      "learning_rate": 1.6688155339805828e-05,
      "loss": 0.1161,
      "step": 63960
    },
    {
      "epoch": 24.84,
      "learning_rate": 1.6687637540453076e-05,
      "loss": 0.2066,
      "step": 63970
    },
    {
      "epoch": 24.85,
      "learning_rate": 1.6687119741100327e-05,
      "loss": 0.1563,
      "step": 63980
    },
    {
      "epoch": 24.85,
      "learning_rate": 1.6686601941747575e-05,
      "loss": 0.0902,
      "step": 63990
    },
    {
      "epoch": 24.85,
      "learning_rate": 1.6686084142394823e-05,
      "loss": 0.053,
      "step": 64000
    },
    {
      "epoch": 24.86,
      "learning_rate": 1.668556634304207e-05,
      "loss": 0.0014,
      "step": 64010
    },
    {
      "epoch": 24.86,
      "learning_rate": 1.6685048543689322e-05,
      "loss": 0.2784,
      "step": 64020
    },
    {
      "epoch": 24.87,
      "learning_rate": 1.668453074433657e-05,
      "loss": 0.1118,
      "step": 64030
    },
    {
      "epoch": 24.87,
      "learning_rate": 1.668401294498382e-05,
      "loss": 0.0675,
      "step": 64040
    },
    {
      "epoch": 24.87,
      "learning_rate": 1.668349514563107e-05,
      "loss": 0.1076,
      "step": 64050
    },
    {
      "epoch": 24.88,
      "learning_rate": 1.668297734627832e-05,
      "loss": 0.1033,
      "step": 64060
    },
    {
      "epoch": 24.88,
      "learning_rate": 1.6682459546925565e-05,
      "loss": 0.1769,
      "step": 64070
    },
    {
      "epoch": 24.89,
      "learning_rate": 1.6681941747572816e-05,
      "loss": 0.0551,
      "step": 64080
    },
    {
      "epoch": 24.89,
      "learning_rate": 1.6681423948220064e-05,
      "loss": 0.1376,
      "step": 64090
    },
    {
      "epoch": 24.89,
      "learning_rate": 1.6680906148867316e-05,
      "loss": 0.1855,
      "step": 64100
    },
    {
      "epoch": 24.9,
      "learning_rate": 1.6680388349514564e-05,
      "loss": 0.0208,
      "step": 64110
    },
    {
      "epoch": 24.9,
      "learning_rate": 1.6679870550161815e-05,
      "loss": 0.1238,
      "step": 64120
    },
    {
      "epoch": 24.9,
      "learning_rate": 1.6679352750809063e-05,
      "loss": 0.0856,
      "step": 64130
    },
    {
      "epoch": 24.91,
      "learning_rate": 1.6678834951456314e-05,
      "loss": 0.05,
      "step": 64140
    },
    {
      "epoch": 24.91,
      "learning_rate": 1.667831715210356e-05,
      "loss": 0.1858,
      "step": 64150
    },
    {
      "epoch": 24.92,
      "learning_rate": 1.667779935275081e-05,
      "loss": 0.0596,
      "step": 64160
    },
    {
      "epoch": 24.92,
      "learning_rate": 1.6677281553398058e-05,
      "loss": 0.06,
      "step": 64170
    },
    {
      "epoch": 24.92,
      "learning_rate": 1.667676375404531e-05,
      "loss": 0.0901,
      "step": 64180
    },
    {
      "epoch": 24.93,
      "learning_rate": 1.6676245954692557e-05,
      "loss": 0.0521,
      "step": 64190
    },
    {
      "epoch": 24.93,
      "learning_rate": 1.667572815533981e-05,
      "loss": 0.19,
      "step": 64200
    },
    {
      "epoch": 24.94,
      "learning_rate": 1.6675210355987056e-05,
      "loss": 0.0783,
      "step": 64210
    },
    {
      "epoch": 24.94,
      "learning_rate": 1.6674692556634308e-05,
      "loss": 0.0445,
      "step": 64220
    },
    {
      "epoch": 24.94,
      "learning_rate": 1.6674174757281552e-05,
      "loss": 0.2725,
      "step": 64230
    },
    {
      "epoch": 24.95,
      "learning_rate": 1.6673656957928804e-05,
      "loss": 0.0178,
      "step": 64240
    },
    {
      "epoch": 24.95,
      "learning_rate": 1.667313915857605e-05,
      "loss": 0.0501,
      "step": 64250
    },
    {
      "epoch": 24.96,
      "learning_rate": 1.6672621359223303e-05,
      "loss": 0.0562,
      "step": 64260
    },
    {
      "epoch": 24.96,
      "learning_rate": 1.667210355987055e-05,
      "loss": 0.1321,
      "step": 64270
    },
    {
      "epoch": 24.96,
      "learning_rate": 1.6671585760517802e-05,
      "loss": 0.0311,
      "step": 64280
    },
    {
      "epoch": 24.97,
      "learning_rate": 1.667106796116505e-05,
      "loss": 0.0921,
      "step": 64290
    },
    {
      "epoch": 24.97,
      "learning_rate": 1.66705501618123e-05,
      "loss": 0.1184,
      "step": 64300
    },
    {
      "epoch": 24.97,
      "learning_rate": 1.6670032362459546e-05,
      "loss": 0.0634,
      "step": 64310
    },
    {
      "epoch": 24.98,
      "learning_rate": 1.6669514563106797e-05,
      "loss": 0.2518,
      "step": 64320
    },
    {
      "epoch": 24.98,
      "learning_rate": 1.6668996763754045e-05,
      "loss": 0.0272,
      "step": 64330
    },
    {
      "epoch": 24.99,
      "learning_rate": 1.6668478964401296e-05,
      "loss": 0.1468,
      "step": 64340
    },
    {
      "epoch": 24.99,
      "learning_rate": 1.6667961165048544e-05,
      "loss": 0.1336,
      "step": 64350
    },
    {
      "epoch": 24.99,
      "learning_rate": 1.6667443365695796e-05,
      "loss": 0.1145,
      "step": 64360
    },
    {
      "epoch": 25.0,
      "learning_rate": 1.6666925566343044e-05,
      "loss": 0.0275,
      "step": 64370
    },
    {
      "epoch": 25.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.26248398423194885,
      "eval_runtime": 8.3281,
      "eval_samples_per_second": 436.473,
      "eval_steps_per_second": 54.634,
      "step": 64375
    },
    {
      "epoch": 25.0,
      "learning_rate": 1.6666407766990295e-05,
      "loss": 0.0285,
      "step": 64380
    },
    {
      "epoch": 25.01,
      "learning_rate": 1.666588996763754e-05,
      "loss": 0.1049,
      "step": 64390
    },
    {
      "epoch": 25.01,
      "learning_rate": 1.666537216828479e-05,
      "loss": 0.1624,
      "step": 64400
    },
    {
      "epoch": 25.01,
      "learning_rate": 1.666485436893204e-05,
      "loss": 0.0456,
      "step": 64410
    },
    {
      "epoch": 25.02,
      "learning_rate": 1.666433656957929e-05,
      "loss": 0.2012,
      "step": 64420
    },
    {
      "epoch": 25.02,
      "learning_rate": 1.6663818770226538e-05,
      "loss": 0.2304,
      "step": 64430
    },
    {
      "epoch": 25.03,
      "learning_rate": 1.666330097087379e-05,
      "loss": 0.2281,
      "step": 64440
    },
    {
      "epoch": 25.03,
      "learning_rate": 1.6662783171521037e-05,
      "loss": 0.0724,
      "step": 64450
    },
    {
      "epoch": 25.03,
      "learning_rate": 1.666226537216829e-05,
      "loss": 0.0483,
      "step": 64460
    },
    {
      "epoch": 25.04,
      "learning_rate": 1.6661747572815536e-05,
      "loss": 0.1143,
      "step": 64470
    },
    {
      "epoch": 25.04,
      "learning_rate": 1.6661229773462784e-05,
      "loss": 0.0603,
      "step": 64480
    },
    {
      "epoch": 25.04,
      "learning_rate": 1.6660711974110032e-05,
      "loss": 0.1586,
      "step": 64490
    },
    {
      "epoch": 25.05,
      "learning_rate": 1.6660194174757283e-05,
      "loss": 0.0503,
      "step": 64500
    },
    {
      "epoch": 25.05,
      "learning_rate": 1.665967637540453e-05,
      "loss": 0.0187,
      "step": 64510
    },
    {
      "epoch": 25.06,
      "learning_rate": 1.6659158576051783e-05,
      "loss": 0.1354,
      "step": 64520
    },
    {
      "epoch": 25.06,
      "learning_rate": 1.665864077669903e-05,
      "loss": 0.1974,
      "step": 64530
    },
    {
      "epoch": 25.06,
      "learning_rate": 1.6658122977346282e-05,
      "loss": 0.1636,
      "step": 64540
    },
    {
      "epoch": 25.07,
      "learning_rate": 1.665760517799353e-05,
      "loss": 0.0756,
      "step": 64550
    },
    {
      "epoch": 25.07,
      "learning_rate": 1.6657087378640778e-05,
      "loss": 0.0205,
      "step": 64560
    },
    {
      "epoch": 25.08,
      "learning_rate": 1.6656569579288026e-05,
      "loss": 0.0723,
      "step": 64570
    },
    {
      "epoch": 25.08,
      "learning_rate": 1.6656051779935277e-05,
      "loss": 0.0924,
      "step": 64580
    },
    {
      "epoch": 25.08,
      "learning_rate": 1.6655533980582525e-05,
      "loss": 0.0205,
      "step": 64590
    },
    {
      "epoch": 25.09,
      "learning_rate": 1.6655016181229776e-05,
      "loss": 0.0645,
      "step": 64600
    },
    {
      "epoch": 25.09,
      "learning_rate": 1.6654498381877024e-05,
      "loss": 0.0473,
      "step": 64610
    },
    {
      "epoch": 25.1,
      "learning_rate": 1.6653980582524275e-05,
      "loss": 0.0309,
      "step": 64620
    },
    {
      "epoch": 25.1,
      "learning_rate": 1.6653462783171523e-05,
      "loss": 0.0681,
      "step": 64630
    },
    {
      "epoch": 25.1,
      "learning_rate": 1.665294498381877e-05,
      "loss": 0.1689,
      "step": 64640
    },
    {
      "epoch": 25.11,
      "learning_rate": 1.665242718446602e-05,
      "loss": 0.14,
      "step": 64650
    },
    {
      "epoch": 25.11,
      "learning_rate": 1.665190938511327e-05,
      "loss": 0.0514,
      "step": 64660
    },
    {
      "epoch": 25.11,
      "learning_rate": 1.665139158576052e-05,
      "loss": 0.0724,
      "step": 64670
    },
    {
      "epoch": 25.12,
      "learning_rate": 1.665087378640777e-05,
      "loss": 0.0684,
      "step": 64680
    },
    {
      "epoch": 25.12,
      "learning_rate": 1.6650355987055018e-05,
      "loss": 0.1625,
      "step": 64690
    },
    {
      "epoch": 25.13,
      "learning_rate": 1.664983818770227e-05,
      "loss": 0.0493,
      "step": 64700
    },
    {
      "epoch": 25.13,
      "learning_rate": 1.6649320388349517e-05,
      "loss": 0.1419,
      "step": 64710
    },
    {
      "epoch": 25.13,
      "learning_rate": 1.6648802588996765e-05,
      "loss": 0.0942,
      "step": 64720
    },
    {
      "epoch": 25.14,
      "learning_rate": 1.6648284789644013e-05,
      "loss": 0.2344,
      "step": 64730
    },
    {
      "epoch": 25.14,
      "learning_rate": 1.6647766990291264e-05,
      "loss": 0.0445,
      "step": 64740
    },
    {
      "epoch": 25.15,
      "learning_rate": 1.6647249190938512e-05,
      "loss": 0.0269,
      "step": 64750
    },
    {
      "epoch": 25.15,
      "learning_rate": 1.6646731391585763e-05,
      "loss": 0.0428,
      "step": 64760
    },
    {
      "epoch": 25.15,
      "learning_rate": 1.664621359223301e-05,
      "loss": 0.1814,
      "step": 64770
    },
    {
      "epoch": 25.16,
      "learning_rate": 1.6645695792880263e-05,
      "loss": 0.151,
      "step": 64780
    },
    {
      "epoch": 25.16,
      "learning_rate": 1.664517799352751e-05,
      "loss": 0.0667,
      "step": 64790
    },
    {
      "epoch": 25.17,
      "learning_rate": 1.664466019417476e-05,
      "loss": 0.1344,
      "step": 64800
    },
    {
      "epoch": 25.17,
      "learning_rate": 1.6644142394822006e-05,
      "loss": 0.1135,
      "step": 64810
    },
    {
      "epoch": 25.17,
      "learning_rate": 1.6643624595469258e-05,
      "loss": 0.0042,
      "step": 64820
    },
    {
      "epoch": 25.18,
      "learning_rate": 1.6643106796116506e-05,
      "loss": 0.0629,
      "step": 64830
    },
    {
      "epoch": 25.18,
      "learning_rate": 1.6642588996763757e-05,
      "loss": 0.2934,
      "step": 64840
    },
    {
      "epoch": 25.18,
      "learning_rate": 1.6642071197411005e-05,
      "loss": 0.112,
      "step": 64850
    },
    {
      "epoch": 25.19,
      "learning_rate": 1.6641553398058253e-05,
      "loss": 0.1309,
      "step": 64860
    },
    {
      "epoch": 25.19,
      "learning_rate": 1.6641035598705504e-05,
      "loss": 0.1322,
      "step": 64870
    },
    {
      "epoch": 25.2,
      "learning_rate": 1.6640517799352752e-05,
      "loss": 0.0448,
      "step": 64880
    },
    {
      "epoch": 25.2,
      "learning_rate": 1.664e-05,
      "loss": 0.0663,
      "step": 64890
    },
    {
      "epoch": 25.2,
      "learning_rate": 1.663948220064725e-05,
      "loss": 0.1879,
      "step": 64900
    },
    {
      "epoch": 25.21,
      "learning_rate": 1.66389644012945e-05,
      "loss": 0.0177,
      "step": 64910
    },
    {
      "epoch": 25.21,
      "learning_rate": 1.663844660194175e-05,
      "loss": 0.1343,
      "step": 64920
    },
    {
      "epoch": 25.22,
      "learning_rate": 1.6637928802589e-05,
      "loss": 0.1194,
      "step": 64930
    },
    {
      "epoch": 25.22,
      "learning_rate": 1.6637411003236246e-05,
      "loss": 0.1405,
      "step": 64940
    },
    {
      "epoch": 25.22,
      "learning_rate": 1.6636893203883498e-05,
      "loss": 0.1592,
      "step": 64950
    },
    {
      "epoch": 25.23,
      "learning_rate": 1.6636375404530746e-05,
      "loss": 0.0773,
      "step": 64960
    },
    {
      "epoch": 25.23,
      "learning_rate": 1.6635857605177993e-05,
      "loss": 0.0662,
      "step": 64970
    },
    {
      "epoch": 25.23,
      "learning_rate": 1.6635339805825245e-05,
      "loss": 0.1106,
      "step": 64980
    },
    {
      "epoch": 25.24,
      "learning_rate": 1.6634822006472493e-05,
      "loss": 0.2341,
      "step": 64990
    },
    {
      "epoch": 25.24,
      "learning_rate": 1.6634304207119744e-05,
      "loss": 0.1738,
      "step": 65000
    },
    {
      "epoch": 25.25,
      "learning_rate": 1.6633786407766992e-05,
      "loss": 0.1843,
      "step": 65010
    },
    {
      "epoch": 25.25,
      "learning_rate": 1.663326860841424e-05,
      "loss": 0.0805,
      "step": 65020
    },
    {
      "epoch": 25.25,
      "learning_rate": 1.663275080906149e-05,
      "loss": 0.0758,
      "step": 65030
    },
    {
      "epoch": 25.26,
      "learning_rate": 1.663223300970874e-05,
      "loss": 0.0712,
      "step": 65040
    },
    {
      "epoch": 25.26,
      "learning_rate": 1.6631715210355987e-05,
      "loss": 0.0525,
      "step": 65050
    },
    {
      "epoch": 25.27,
      "learning_rate": 1.663119741100324e-05,
      "loss": 0.0651,
      "step": 65060
    },
    {
      "epoch": 25.27,
      "learning_rate": 1.6630679611650486e-05,
      "loss": 0.0355,
      "step": 65070
    },
    {
      "epoch": 25.27,
      "learning_rate": 1.6630161812297738e-05,
      "loss": 0.1036,
      "step": 65080
    },
    {
      "epoch": 25.28,
      "learning_rate": 1.6629644012944986e-05,
      "loss": 0.2159,
      "step": 65090
    },
    {
      "epoch": 25.28,
      "learning_rate": 1.6629126213592233e-05,
      "loss": 0.1407,
      "step": 65100
    },
    {
      "epoch": 25.29,
      "learning_rate": 1.6628608414239485e-05,
      "loss": 0.088,
      "step": 65110
    },
    {
      "epoch": 25.29,
      "learning_rate": 1.6628090614886733e-05,
      "loss": 0.0139,
      "step": 65120
    },
    {
      "epoch": 25.29,
      "learning_rate": 1.662757281553398e-05,
      "loss": 0.1104,
      "step": 65130
    },
    {
      "epoch": 25.3,
      "learning_rate": 1.6627055016181232e-05,
      "loss": 0.2004,
      "step": 65140
    },
    {
      "epoch": 25.3,
      "learning_rate": 1.662653721682848e-05,
      "loss": 0.1432,
      "step": 65150
    },
    {
      "epoch": 25.3,
      "learning_rate": 1.662601941747573e-05,
      "loss": 0.0038,
      "step": 65160
    },
    {
      "epoch": 25.31,
      "learning_rate": 1.662550161812298e-05,
      "loss": 0.1546,
      "step": 65170
    },
    {
      "epoch": 25.31,
      "learning_rate": 1.6624983818770227e-05,
      "loss": 0.1645,
      "step": 65180
    },
    {
      "epoch": 25.32,
      "learning_rate": 1.6624466019417478e-05,
      "loss": 0.1445,
      "step": 65190
    },
    {
      "epoch": 25.32,
      "learning_rate": 1.6623948220064726e-05,
      "loss": 0.1014,
      "step": 65200
    },
    {
      "epoch": 25.32,
      "learning_rate": 1.6623430420711974e-05,
      "loss": 0.1674,
      "step": 65210
    },
    {
      "epoch": 25.33,
      "learning_rate": 1.6622912621359225e-05,
      "loss": 0.1391,
      "step": 65220
    },
    {
      "epoch": 25.33,
      "learning_rate": 1.6622394822006473e-05,
      "loss": 0.0973,
      "step": 65230
    },
    {
      "epoch": 25.34,
      "learning_rate": 1.662187702265372e-05,
      "loss": 0.2874,
      "step": 65240
    },
    {
      "epoch": 25.34,
      "learning_rate": 1.6621359223300973e-05,
      "loss": 0.1018,
      "step": 65250
    },
    {
      "epoch": 25.34,
      "learning_rate": 1.662084142394822e-05,
      "loss": 0.244,
      "step": 65260
    },
    {
      "epoch": 25.35,
      "learning_rate": 1.6620323624595472e-05,
      "loss": 0.164,
      "step": 65270
    },
    {
      "epoch": 25.35,
      "learning_rate": 1.661980582524272e-05,
      "loss": 0.082,
      "step": 65280
    },
    {
      "epoch": 25.36,
      "learning_rate": 1.6619288025889968e-05,
      "loss": 0.0965,
      "step": 65290
    },
    {
      "epoch": 25.36,
      "learning_rate": 1.661877022653722e-05,
      "loss": 0.0891,
      "step": 65300
    },
    {
      "epoch": 25.36,
      "learning_rate": 1.6618252427184467e-05,
      "loss": 0.0331,
      "step": 65310
    },
    {
      "epoch": 25.37,
      "learning_rate": 1.6617734627831715e-05,
      "loss": 0.1873,
      "step": 65320
    },
    {
      "epoch": 25.37,
      "learning_rate": 1.6617216828478966e-05,
      "loss": 0.0682,
      "step": 65330
    },
    {
      "epoch": 25.37,
      "learning_rate": 1.6616699029126214e-05,
      "loss": 0.063,
      "step": 65340
    },
    {
      "epoch": 25.38,
      "learning_rate": 1.6616181229773465e-05,
      "loss": 0.0408,
      "step": 65350
    },
    {
      "epoch": 25.38,
      "learning_rate": 1.6615663430420713e-05,
      "loss": 0.1143,
      "step": 65360
    },
    {
      "epoch": 25.39,
      "learning_rate": 1.661514563106796e-05,
      "loss": 0.0638,
      "step": 65370
    },
    {
      "epoch": 25.39,
      "learning_rate": 1.6614627831715213e-05,
      "loss": 0.0042,
      "step": 65380
    },
    {
      "epoch": 25.39,
      "learning_rate": 1.661411003236246e-05,
      "loss": 0.0713,
      "step": 65390
    },
    {
      "epoch": 25.4,
      "learning_rate": 1.661359223300971e-05,
      "loss": 0.0578,
      "step": 65400
    },
    {
      "epoch": 25.4,
      "learning_rate": 1.661307443365696e-05,
      "loss": 0.0989,
      "step": 65410
    },
    {
      "epoch": 25.41,
      "learning_rate": 1.6612556634304208e-05,
      "loss": 0.165,
      "step": 65420
    },
    {
      "epoch": 25.41,
      "learning_rate": 1.661203883495146e-05,
      "loss": 0.1092,
      "step": 65430
    },
    {
      "epoch": 25.41,
      "learning_rate": 1.6611521035598707e-05,
      "loss": 0.0028,
      "step": 65440
    },
    {
      "epoch": 25.42,
      "learning_rate": 1.6611003236245955e-05,
      "loss": 0.077,
      "step": 65450
    },
    {
      "epoch": 25.42,
      "learning_rate": 1.6610485436893206e-05,
      "loss": 0.0232,
      "step": 65460
    },
    {
      "epoch": 25.43,
      "learning_rate": 1.6609967637540454e-05,
      "loss": 0.0792,
      "step": 65470
    },
    {
      "epoch": 25.43,
      "learning_rate": 1.6609449838187702e-05,
      "loss": 0.0069,
      "step": 65480
    },
    {
      "epoch": 25.43,
      "learning_rate": 1.6608932038834953e-05,
      "loss": 0.1734,
      "step": 65490
    },
    {
      "epoch": 25.44,
      "learning_rate": 1.66084142394822e-05,
      "loss": 0.2823,
      "step": 65500
    },
    {
      "epoch": 25.44,
      "learning_rate": 1.6607896440129453e-05,
      "loss": 0.1968,
      "step": 65510
    },
    {
      "epoch": 25.44,
      "learning_rate": 1.66073786407767e-05,
      "loss": 0.1496,
      "step": 65520
    },
    {
      "epoch": 25.45,
      "learning_rate": 1.660686084142395e-05,
      "loss": 0.0885,
      "step": 65530
    },
    {
      "epoch": 25.45,
      "learning_rate": 1.6606343042071196e-05,
      "loss": 0.0965,
      "step": 65540
    },
    {
      "epoch": 25.46,
      "learning_rate": 1.6605825242718448e-05,
      "loss": 0.0724,
      "step": 65550
    },
    {
      "epoch": 25.46,
      "learning_rate": 1.6605307443365696e-05,
      "loss": 0.0954,
      "step": 65560
    },
    {
      "epoch": 25.46,
      "learning_rate": 1.6604789644012947e-05,
      "loss": 0.3889,
      "step": 65570
    },
    {
      "epoch": 25.47,
      "learning_rate": 1.6604271844660195e-05,
      "loss": 0.0621,
      "step": 65580
    },
    {
      "epoch": 25.47,
      "learning_rate": 1.6603754045307446e-05,
      "loss": 0.012,
      "step": 65590
    },
    {
      "epoch": 25.48,
      "learning_rate": 1.6603236245954694e-05,
      "loss": 0.0818,
      "step": 65600
    },
    {
      "epoch": 25.48,
      "learning_rate": 1.6602718446601945e-05,
      "loss": 0.0617,
      "step": 65610
    },
    {
      "epoch": 25.48,
      "learning_rate": 1.660220064724919e-05,
      "loss": 0.0493,
      "step": 65620
    },
    {
      "epoch": 25.49,
      "learning_rate": 1.660168284789644e-05,
      "loss": 0.0038,
      "step": 65630
    },
    {
      "epoch": 25.49,
      "learning_rate": 1.660116504854369e-05,
      "loss": 0.0706,
      "step": 65640
    },
    {
      "epoch": 25.5,
      "learning_rate": 1.660064724919094e-05,
      "loss": 0.0595,
      "step": 65650
    },
    {
      "epoch": 25.5,
      "learning_rate": 1.660012944983819e-05,
      "loss": 0.0454,
      "step": 65660
    },
    {
      "epoch": 25.5,
      "learning_rate": 1.659961165048544e-05,
      "loss": 0.2334,
      "step": 65670
    },
    {
      "epoch": 25.51,
      "learning_rate": 1.6599093851132688e-05,
      "loss": 0.1045,
      "step": 65680
    },
    {
      "epoch": 25.51,
      "learning_rate": 1.659857605177994e-05,
      "loss": 0.1457,
      "step": 65690
    },
    {
      "epoch": 25.51,
      "learning_rate": 1.6598058252427183e-05,
      "loss": 0.0924,
      "step": 65700
    },
    {
      "epoch": 25.52,
      "learning_rate": 1.6597540453074435e-05,
      "loss": 0.0231,
      "step": 65710
    },
    {
      "epoch": 25.52,
      "learning_rate": 1.6597022653721683e-05,
      "loss": 0.071,
      "step": 65720
    },
    {
      "epoch": 25.53,
      "learning_rate": 1.6596504854368934e-05,
      "loss": 0.1649,
      "step": 65730
    },
    {
      "epoch": 25.53,
      "learning_rate": 1.6595987055016182e-05,
      "loss": 0.0943,
      "step": 65740
    },
    {
      "epoch": 25.53,
      "learning_rate": 1.6595469255663433e-05,
      "loss": 0.1265,
      "step": 65750
    },
    {
      "epoch": 25.54,
      "learning_rate": 1.659495145631068e-05,
      "loss": 0.144,
      "step": 65760
    },
    {
      "epoch": 25.54,
      "learning_rate": 1.6594433656957932e-05,
      "loss": 0.0758,
      "step": 65770
    },
    {
      "epoch": 25.55,
      "learning_rate": 1.6593915857605177e-05,
      "loss": 0.0286,
      "step": 65780
    },
    {
      "epoch": 25.55,
      "learning_rate": 1.6593398058252428e-05,
      "loss": 0.1999,
      "step": 65790
    },
    {
      "epoch": 25.55,
      "learning_rate": 1.6592880258899676e-05,
      "loss": 0.0758,
      "step": 65800
    },
    {
      "epoch": 25.56,
      "learning_rate": 1.6592362459546928e-05,
      "loss": 0.0779,
      "step": 65810
    },
    {
      "epoch": 25.56,
      "learning_rate": 1.6591844660194175e-05,
      "loss": 0.0936,
      "step": 65820
    },
    {
      "epoch": 25.57,
      "learning_rate": 1.6591326860841427e-05,
      "loss": 0.1087,
      "step": 65830
    },
    {
      "epoch": 25.57,
      "learning_rate": 1.6590809061488675e-05,
      "loss": 0.1522,
      "step": 65840
    },
    {
      "epoch": 25.57,
      "learning_rate": 1.6590291262135926e-05,
      "loss": 0.0241,
      "step": 65850
    },
    {
      "epoch": 25.58,
      "learning_rate": 1.658977346278317e-05,
      "loss": 0.0383,
      "step": 65860
    },
    {
      "epoch": 25.58,
      "learning_rate": 1.6589255663430422e-05,
      "loss": 0.0951,
      "step": 65870
    },
    {
      "epoch": 25.58,
      "learning_rate": 1.658873786407767e-05,
      "loss": 0.0159,
      "step": 65880
    },
    {
      "epoch": 25.59,
      "learning_rate": 1.658822006472492e-05,
      "loss": 0.1704,
      "step": 65890
    },
    {
      "epoch": 25.59,
      "learning_rate": 1.658770226537217e-05,
      "loss": 0.0196,
      "step": 65900
    },
    {
      "epoch": 25.6,
      "learning_rate": 1.658718446601942e-05,
      "loss": 0.1635,
      "step": 65910
    },
    {
      "epoch": 25.6,
      "learning_rate": 1.6586666666666668e-05,
      "loss": 0.1256,
      "step": 65920
    },
    {
      "epoch": 25.6,
      "learning_rate": 1.658614886731392e-05,
      "loss": 0.0464,
      "step": 65930
    },
    {
      "epoch": 25.61,
      "learning_rate": 1.6585631067961164e-05,
      "loss": 0.0855,
      "step": 65940
    },
    {
      "epoch": 25.61,
      "learning_rate": 1.6585113268608415e-05,
      "loss": 0.0543,
      "step": 65950
    },
    {
      "epoch": 25.62,
      "learning_rate": 1.6584595469255663e-05,
      "loss": 0.0056,
      "step": 65960
    },
    {
      "epoch": 25.62,
      "learning_rate": 1.6584077669902915e-05,
      "loss": 0.0458,
      "step": 65970
    },
    {
      "epoch": 25.62,
      "learning_rate": 1.6583559870550163e-05,
      "loss": 0.1332,
      "step": 65980
    },
    {
      "epoch": 25.63,
      "learning_rate": 1.6583042071197414e-05,
      "loss": 0.0896,
      "step": 65990
    },
    {
      "epoch": 25.63,
      "learning_rate": 1.6582524271844662e-05,
      "loss": 0.2005,
      "step": 66000
    },
    {
      "epoch": 25.63,
      "learning_rate": 1.6582006472491913e-05,
      "loss": 0.1023,
      "step": 66010
    },
    {
      "epoch": 25.64,
      "learning_rate": 1.6581488673139158e-05,
      "loss": 0.1373,
      "step": 66020
    },
    {
      "epoch": 25.64,
      "learning_rate": 1.658097087378641e-05,
      "loss": 0.1066,
      "step": 66030
    },
    {
      "epoch": 25.65,
      "learning_rate": 1.6580453074433657e-05,
      "loss": 0.2469,
      "step": 66040
    },
    {
      "epoch": 25.65,
      "learning_rate": 1.6579935275080908e-05,
      "loss": 0.0675,
      "step": 66050
    },
    {
      "epoch": 25.65,
      "learning_rate": 1.6579417475728156e-05,
      "loss": 0.0914,
      "step": 66060
    },
    {
      "epoch": 25.66,
      "learning_rate": 1.6578899676375407e-05,
      "loss": 0.1389,
      "step": 66070
    },
    {
      "epoch": 25.66,
      "learning_rate": 1.6578381877022655e-05,
      "loss": 0.1254,
      "step": 66080
    },
    {
      "epoch": 25.67,
      "learning_rate": 1.6577864077669907e-05,
      "loss": 0.1682,
      "step": 66090
    },
    {
      "epoch": 25.67,
      "learning_rate": 1.657734627831715e-05,
      "loss": 0.0067,
      "step": 66100
    },
    {
      "epoch": 25.67,
      "learning_rate": 1.6576828478964403e-05,
      "loss": 0.2091,
      "step": 66110
    },
    {
      "epoch": 25.68,
      "learning_rate": 1.657631067961165e-05,
      "loss": 0.1589,
      "step": 66120
    },
    {
      "epoch": 25.68,
      "learning_rate": 1.6575792880258902e-05,
      "loss": 0.0897,
      "step": 66130
    },
    {
      "epoch": 25.69,
      "learning_rate": 1.657527508090615e-05,
      "loss": 0.1822,
      "step": 66140
    },
    {
      "epoch": 25.69,
      "learning_rate": 1.65747572815534e-05,
      "loss": 0.0703,
      "step": 66150
    },
    {
      "epoch": 25.69,
      "learning_rate": 1.657423948220065e-05,
      "loss": 0.2245,
      "step": 66160
    },
    {
      "epoch": 25.7,
      "learning_rate": 1.65737216828479e-05,
      "loss": 0.1029,
      "step": 66170
    },
    {
      "epoch": 25.7,
      "learning_rate": 1.6573203883495148e-05,
      "loss": 0.1517,
      "step": 66180
    },
    {
      "epoch": 25.7,
      "learning_rate": 1.6572686084142396e-05,
      "loss": 0.0774,
      "step": 66190
    },
    {
      "epoch": 25.71,
      "learning_rate": 1.6572168284789644e-05,
      "loss": 0.0308,
      "step": 66200
    },
    {
      "epoch": 25.71,
      "learning_rate": 1.6571650485436895e-05,
      "loss": 0.0496,
      "step": 66210
    },
    {
      "epoch": 25.72,
      "learning_rate": 1.6571132686084143e-05,
      "loss": 0.0789,
      "step": 66220
    },
    {
      "epoch": 25.72,
      "learning_rate": 1.6570614886731395e-05,
      "loss": 0.0966,
      "step": 66230
    },
    {
      "epoch": 25.72,
      "learning_rate": 1.6570097087378642e-05,
      "loss": 0.1551,
      "step": 66240
    },
    {
      "epoch": 25.73,
      "learning_rate": 1.6569579288025894e-05,
      "loss": 0.0209,
      "step": 66250
    },
    {
      "epoch": 25.73,
      "learning_rate": 1.656906148867314e-05,
      "loss": 0.0386,
      "step": 66260
    },
    {
      "epoch": 25.74,
      "learning_rate": 1.656854368932039e-05,
      "loss": 0.1258,
      "step": 66270
    },
    {
      "epoch": 25.74,
      "learning_rate": 1.6568025889967638e-05,
      "loss": 0.1906,
      "step": 66280
    },
    {
      "epoch": 25.74,
      "learning_rate": 1.656750809061489e-05,
      "loss": 0.1585,
      "step": 66290
    },
    {
      "epoch": 25.75,
      "learning_rate": 1.6566990291262137e-05,
      "loss": 0.0538,
      "step": 66300
    },
    {
      "epoch": 25.75,
      "learning_rate": 1.6566472491909388e-05,
      "loss": 0.176,
      "step": 66310
    },
    {
      "epoch": 25.76,
      "learning_rate": 1.6565954692556636e-05,
      "loss": 0.0906,
      "step": 66320
    },
    {
      "epoch": 25.76,
      "learning_rate": 1.6565436893203884e-05,
      "loss": 0.0805,
      "step": 66330
    },
    {
      "epoch": 25.76,
      "learning_rate": 1.6564919093851135e-05,
      "loss": 0.0667,
      "step": 66340
    },
    {
      "epoch": 25.77,
      "learning_rate": 1.6564401294498383e-05,
      "loss": 0.1657,
      "step": 66350
    },
    {
      "epoch": 25.77,
      "learning_rate": 1.656388349514563e-05,
      "loss": 0.0775,
      "step": 66360
    },
    {
      "epoch": 25.77,
      "learning_rate": 1.6563365695792882e-05,
      "loss": 0.0217,
      "step": 66370
    },
    {
      "epoch": 25.78,
      "learning_rate": 1.656284789644013e-05,
      "loss": 0.119,
      "step": 66380
    },
    {
      "epoch": 25.78,
      "learning_rate": 1.656233009708738e-05,
      "loss": 0.0688,
      "step": 66390
    },
    {
      "epoch": 25.79,
      "learning_rate": 1.656181229773463e-05,
      "loss": 0.0591,
      "step": 66400
    },
    {
      "epoch": 25.79,
      "learning_rate": 1.6561294498381877e-05,
      "loss": 0.1574,
      "step": 66410
    },
    {
      "epoch": 25.79,
      "learning_rate": 1.656077669902913e-05,
      "loss": 0.0634,
      "step": 66420
    },
    {
      "epoch": 25.8,
      "learning_rate": 1.6560258899676377e-05,
      "loss": 0.0924,
      "step": 66430
    },
    {
      "epoch": 25.8,
      "learning_rate": 1.6559741100323625e-05,
      "loss": 0.1296,
      "step": 66440
    },
    {
      "epoch": 25.81,
      "learning_rate": 1.6559223300970876e-05,
      "loss": 0.0251,
      "step": 66450
    },
    {
      "epoch": 25.81,
      "learning_rate": 1.6558705501618124e-05,
      "loss": 0.0963,
      "step": 66460
    },
    {
      "epoch": 25.81,
      "learning_rate": 1.6558187702265375e-05,
      "loss": 0.246,
      "step": 66470
    },
    {
      "epoch": 25.82,
      "learning_rate": 1.6557669902912623e-05,
      "loss": 0.0491,
      "step": 66480
    },
    {
      "epoch": 25.82,
      "learning_rate": 1.655715210355987e-05,
      "loss": 0.019,
      "step": 66490
    },
    {
      "epoch": 25.83,
      "learning_rate": 1.6556634304207122e-05,
      "loss": 0.1628,
      "step": 66500
    },
    {
      "epoch": 25.83,
      "learning_rate": 1.655611650485437e-05,
      "loss": 0.0888,
      "step": 66510
    },
    {
      "epoch": 25.83,
      "learning_rate": 1.6555598705501618e-05,
      "loss": 0.231,
      "step": 66520
    },
    {
      "epoch": 25.84,
      "learning_rate": 1.655508090614887e-05,
      "loss": 0.0629,
      "step": 66530
    },
    {
      "epoch": 25.84,
      "learning_rate": 1.6554563106796117e-05,
      "loss": 0.0587,
      "step": 66540
    },
    {
      "epoch": 25.84,
      "learning_rate": 1.655404530744337e-05,
      "loss": 0.0961,
      "step": 66550
    },
    {
      "epoch": 25.85,
      "learning_rate": 1.6553527508090617e-05,
      "loss": 0.2032,
      "step": 66560
    },
    {
      "epoch": 25.85,
      "learning_rate": 1.6553009708737865e-05,
      "loss": 0.0945,
      "step": 66570
    },
    {
      "epoch": 25.86,
      "learning_rate": 1.6552491909385116e-05,
      "loss": 0.0062,
      "step": 66580
    },
    {
      "epoch": 25.86,
      "learning_rate": 1.6551974110032364e-05,
      "loss": 0.0089,
      "step": 66590
    },
    {
      "epoch": 25.86,
      "learning_rate": 1.6551456310679612e-05,
      "loss": 0.3244,
      "step": 66600
    },
    {
      "epoch": 25.87,
      "learning_rate": 1.6550938511326863e-05,
      "loss": 0.1982,
      "step": 66610
    },
    {
      "epoch": 25.87,
      "learning_rate": 1.655042071197411e-05,
      "loss": 0.1438,
      "step": 66620
    },
    {
      "epoch": 25.88,
      "learning_rate": 1.654990291262136e-05,
      "loss": 0.0872,
      "step": 66630
    },
    {
      "epoch": 25.88,
      "learning_rate": 1.654938511326861e-05,
      "loss": 0.1581,
      "step": 66640
    },
    {
      "epoch": 25.88,
      "learning_rate": 1.6548867313915858e-05,
      "loss": 0.2025,
      "step": 66650
    },
    {
      "epoch": 25.89,
      "learning_rate": 1.654834951456311e-05,
      "loss": 0.2043,
      "step": 66660
    },
    {
      "epoch": 25.89,
      "learning_rate": 1.6547831715210357e-05,
      "loss": 0.1725,
      "step": 66670
    },
    {
      "epoch": 25.9,
      "learning_rate": 1.6547313915857605e-05,
      "loss": 0.1061,
      "step": 66680
    },
    {
      "epoch": 25.9,
      "learning_rate": 1.6546796116504857e-05,
      "loss": 0.4017,
      "step": 66690
    },
    {
      "epoch": 25.9,
      "learning_rate": 1.6546278317152105e-05,
      "loss": 0.0259,
      "step": 66700
    },
    {
      "epoch": 25.91,
      "learning_rate": 1.6545760517799352e-05,
      "loss": 0.0706,
      "step": 66710
    },
    {
      "epoch": 25.91,
      "learning_rate": 1.6545242718446604e-05,
      "loss": 0.0385,
      "step": 66720
    },
    {
      "epoch": 25.91,
      "learning_rate": 1.6544724919093852e-05,
      "loss": 0.0597,
      "step": 66730
    },
    {
      "epoch": 25.92,
      "learning_rate": 1.6544207119741103e-05,
      "loss": 0.0756,
      "step": 66740
    },
    {
      "epoch": 25.92,
      "learning_rate": 1.654368932038835e-05,
      "loss": 0.1139,
      "step": 66750
    },
    {
      "epoch": 25.93,
      "learning_rate": 1.65431715210356e-05,
      "loss": 0.0586,
      "step": 66760
    },
    {
      "epoch": 25.93,
      "learning_rate": 1.654265372168285e-05,
      "loss": 0.0811,
      "step": 66770
    },
    {
      "epoch": 25.93,
      "learning_rate": 1.6542135922330098e-05,
      "loss": 0.1537,
      "step": 66780
    },
    {
      "epoch": 25.94,
      "learning_rate": 1.6541618122977346e-05,
      "loss": 0.0024,
      "step": 66790
    },
    {
      "epoch": 25.94,
      "learning_rate": 1.6541100323624597e-05,
      "loss": 0.0245,
      "step": 66800
    },
    {
      "epoch": 25.95,
      "learning_rate": 1.6540582524271845e-05,
      "loss": 0.1005,
      "step": 66810
    },
    {
      "epoch": 25.95,
      "learning_rate": 1.6540064724919097e-05,
      "loss": 0.0486,
      "step": 66820
    },
    {
      "epoch": 25.95,
      "learning_rate": 1.6539546925566345e-05,
      "loss": 0.1163,
      "step": 66830
    },
    {
      "epoch": 25.96,
      "learning_rate": 1.6539029126213592e-05,
      "loss": 0.0559,
      "step": 66840
    },
    {
      "epoch": 25.96,
      "learning_rate": 1.6538511326860844e-05,
      "loss": 0.0251,
      "step": 66850
    },
    {
      "epoch": 25.97,
      "learning_rate": 1.653799352750809e-05,
      "loss": 0.0061,
      "step": 66860
    },
    {
      "epoch": 25.97,
      "learning_rate": 1.653747572815534e-05,
      "loss": 0.0352,
      "step": 66870
    },
    {
      "epoch": 25.97,
      "learning_rate": 1.653695792880259e-05,
      "loss": 0.1817,
      "step": 66880
    },
    {
      "epoch": 25.98,
      "learning_rate": 1.653644012944984e-05,
      "loss": 0.151,
      "step": 66890
    },
    {
      "epoch": 25.98,
      "learning_rate": 1.653592233009709e-05,
      "loss": 0.038,
      "step": 66900
    },
    {
      "epoch": 25.98,
      "learning_rate": 1.6535404530744338e-05,
      "loss": 0.0657,
      "step": 66910
    },
    {
      "epoch": 25.99,
      "learning_rate": 1.6534886731391586e-05,
      "loss": 0.282,
      "step": 66920
    },
    {
      "epoch": 25.99,
      "learning_rate": 1.6534368932038837e-05,
      "loss": 0.1226,
      "step": 66930
    },
    {
      "epoch": 26.0,
      "learning_rate": 1.6533851132686085e-05,
      "loss": 0.1447,
      "step": 66940
    },
    {
      "epoch": 26.0,
      "learning_rate": 1.6533333333333333e-05,
      "loss": 0.1757,
      "step": 66950
    },
    {
      "epoch": 26.0,
      "eval_accuracy": 0.9488308115543329,
      "eval_loss": 0.2819267511367798,
      "eval_runtime": 8.2333,
      "eval_samples_per_second": 441.5,
      "eval_steps_per_second": 55.263,
      "step": 66950
    },
    {
      "epoch": 26.0,
      "learning_rate": 1.6532815533980584e-05,
      "loss": 0.2687,
      "step": 66960
    },
    {
      "epoch": 26.01,
      "learning_rate": 1.6532297734627832e-05,
      "loss": 0.0613,
      "step": 66970
    },
    {
      "epoch": 26.01,
      "learning_rate": 1.6531779935275084e-05,
      "loss": 0.1617,
      "step": 66980
    },
    {
      "epoch": 26.02,
      "learning_rate": 1.653126213592233e-05,
      "loss": 0.1428,
      "step": 66990
    },
    {
      "epoch": 26.02,
      "learning_rate": 1.653074433656958e-05,
      "loss": 0.0238,
      "step": 67000
    },
    {
      "epoch": 26.02,
      "learning_rate": 1.6530226537216827e-05,
      "loss": 0.1096,
      "step": 67010
    },
    {
      "epoch": 26.03,
      "learning_rate": 1.652970873786408e-05,
      "loss": 0.1035,
      "step": 67020
    },
    {
      "epoch": 26.03,
      "learning_rate": 1.6529190938511327e-05,
      "loss": 0.142,
      "step": 67030
    },
    {
      "epoch": 26.03,
      "learning_rate": 1.6528673139158578e-05,
      "loss": 0.0796,
      "step": 67040
    },
    {
      "epoch": 26.04,
      "learning_rate": 1.6528155339805826e-05,
      "loss": 0.1364,
      "step": 67050
    },
    {
      "epoch": 26.04,
      "learning_rate": 1.6527637540453077e-05,
      "loss": 0.0902,
      "step": 67060
    },
    {
      "epoch": 26.05,
      "learning_rate": 1.6527119741100325e-05,
      "loss": 0.1364,
      "step": 67070
    },
    {
      "epoch": 26.05,
      "learning_rate": 1.6526601941747573e-05,
      "loss": 0.0526,
      "step": 67080
    },
    {
      "epoch": 26.05,
      "learning_rate": 1.652608414239482e-05,
      "loss": 0.0587,
      "step": 67090
    },
    {
      "epoch": 26.06,
      "learning_rate": 1.6525566343042072e-05,
      "loss": 0.2204,
      "step": 67100
    },
    {
      "epoch": 26.06,
      "learning_rate": 1.652504854368932e-05,
      "loss": 0.0142,
      "step": 67110
    },
    {
      "epoch": 26.07,
      "learning_rate": 1.652453074433657e-05,
      "loss": 0.1129,
      "step": 67120
    },
    {
      "epoch": 26.07,
      "learning_rate": 1.652401294498382e-05,
      "loss": 0.0086,
      "step": 67130
    },
    {
      "epoch": 26.07,
      "learning_rate": 1.652349514563107e-05,
      "loss": 0.1265,
      "step": 67140
    },
    {
      "epoch": 26.08,
      "learning_rate": 1.652297734627832e-05,
      "loss": 0.0014,
      "step": 67150
    },
    {
      "epoch": 26.08,
      "learning_rate": 1.6522459546925567e-05,
      "loss": 0.0905,
      "step": 67160
    },
    {
      "epoch": 26.09,
      "learning_rate": 1.6521941747572815e-05,
      "loss": 0.0649,
      "step": 67170
    },
    {
      "epoch": 26.09,
      "learning_rate": 1.6521423948220066e-05,
      "loss": 0.1742,
      "step": 67180
    },
    {
      "epoch": 26.09,
      "learning_rate": 1.6520906148867314e-05,
      "loss": 0.0351,
      "step": 67190
    },
    {
      "epoch": 26.1,
      "learning_rate": 1.6520388349514565e-05,
      "loss": 0.1259,
      "step": 67200
    },
    {
      "epoch": 26.1,
      "learning_rate": 1.6519870550161813e-05,
      "loss": 0.0691,
      "step": 67210
    },
    {
      "epoch": 26.1,
      "learning_rate": 1.6519352750809064e-05,
      "loss": 0.1702,
      "step": 67220
    },
    {
      "epoch": 26.11,
      "learning_rate": 1.6518834951456312e-05,
      "loss": 0.1373,
      "step": 67230
    },
    {
      "epoch": 26.11,
      "learning_rate": 1.651831715210356e-05,
      "loss": 0.0456,
      "step": 67240
    },
    {
      "epoch": 26.12,
      "learning_rate": 1.6517799352750808e-05,
      "loss": 0.0823,
      "step": 67250
    },
    {
      "epoch": 26.12,
      "learning_rate": 1.651728155339806e-05,
      "loss": 0.2093,
      "step": 67260
    },
    {
      "epoch": 26.12,
      "learning_rate": 1.6516763754045307e-05,
      "loss": 0.0692,
      "step": 67270
    },
    {
      "epoch": 26.13,
      "learning_rate": 1.651624595469256e-05,
      "loss": 0.0819,
      "step": 67280
    },
    {
      "epoch": 26.13,
      "learning_rate": 1.6515728155339807e-05,
      "loss": 0.1091,
      "step": 67290
    },
    {
      "epoch": 26.14,
      "learning_rate": 1.6515210355987058e-05,
      "loss": 0.0627,
      "step": 67300
    },
    {
      "epoch": 26.14,
      "learning_rate": 1.6514692556634306e-05,
      "loss": 0.1097,
      "step": 67310
    },
    {
      "epoch": 26.14,
      "learning_rate": 1.6514174757281557e-05,
      "loss": 0.1778,
      "step": 67320
    },
    {
      "epoch": 26.15,
      "learning_rate": 1.6513656957928802e-05,
      "loss": 0.0217,
      "step": 67330
    },
    {
      "epoch": 26.15,
      "learning_rate": 1.6513139158576053e-05,
      "loss": 0.1584,
      "step": 67340
    },
    {
      "epoch": 26.16,
      "learning_rate": 1.65126213592233e-05,
      "loss": 0.0014,
      "step": 67350
    },
    {
      "epoch": 26.16,
      "learning_rate": 1.6512103559870552e-05,
      "loss": 0.0829,
      "step": 67360
    },
    {
      "epoch": 26.16,
      "learning_rate": 1.65115857605178e-05,
      "loss": 0.0368,
      "step": 67370
    },
    {
      "epoch": 26.17,
      "learning_rate": 1.651106796116505e-05,
      "loss": 0.0385,
      "step": 67380
    },
    {
      "epoch": 26.17,
      "learning_rate": 1.65105501618123e-05,
      "loss": 0.2557,
      "step": 67390
    },
    {
      "epoch": 26.17,
      "learning_rate": 1.651003236245955e-05,
      "loss": 0.0544,
      "step": 67400
    },
    {
      "epoch": 26.18,
      "learning_rate": 1.6509514563106795e-05,
      "loss": 0.0241,
      "step": 67410
    },
    {
      "epoch": 26.18,
      "learning_rate": 1.6508996763754047e-05,
      "loss": 0.0248,
      "step": 67420
    },
    {
      "epoch": 26.19,
      "learning_rate": 1.6508478964401294e-05,
      "loss": 0.1129,
      "step": 67430
    },
    {
      "epoch": 26.19,
      "learning_rate": 1.6507961165048546e-05,
      "loss": 0.1352,
      "step": 67440
    },
    {
      "epoch": 26.19,
      "learning_rate": 1.6507443365695794e-05,
      "loss": 0.111,
      "step": 67450
    },
    {
      "epoch": 26.2,
      "learning_rate": 1.6506925566343045e-05,
      "loss": 0.1857,
      "step": 67460
    },
    {
      "epoch": 26.2,
      "learning_rate": 1.6506407766990293e-05,
      "loss": 0.134,
      "step": 67470
    },
    {
      "epoch": 26.21,
      "learning_rate": 1.6505889967637544e-05,
      "loss": 0.0525,
      "step": 67480
    },
    {
      "epoch": 26.21,
      "learning_rate": 1.650537216828479e-05,
      "loss": 0.3058,
      "step": 67490
    },
    {
      "epoch": 26.21,
      "learning_rate": 1.650485436893204e-05,
      "loss": 0.0748,
      "step": 67500
    },
    {
      "epoch": 26.22,
      "learning_rate": 1.6504336569579288e-05,
      "loss": 0.1086,
      "step": 67510
    },
    {
      "epoch": 26.22,
      "learning_rate": 1.650381877022654e-05,
      "loss": 0.0246,
      "step": 67520
    },
    {
      "epoch": 26.23,
      "learning_rate": 1.6503300970873787e-05,
      "loss": 0.0554,
      "step": 67530
    },
    {
      "epoch": 26.23,
      "learning_rate": 1.650278317152104e-05,
      "loss": 0.0392,
      "step": 67540
    },
    {
      "epoch": 26.23,
      "learning_rate": 1.6502265372168287e-05,
      "loss": 0.1675,
      "step": 67550
    },
    {
      "epoch": 26.24,
      "learning_rate": 1.6501747572815538e-05,
      "loss": 0.166,
      "step": 67560
    },
    {
      "epoch": 26.24,
      "learning_rate": 1.6501229773462782e-05,
      "loss": 0.0819,
      "step": 67570
    },
    {
      "epoch": 26.24,
      "learning_rate": 1.6500711974110034e-05,
      "loss": 0.1918,
      "step": 67580
    },
    {
      "epoch": 26.25,
      "learning_rate": 1.650019417475728e-05,
      "loss": 0.2159,
      "step": 67590
    },
    {
      "epoch": 26.25,
      "learning_rate": 1.6499676375404533e-05,
      "loss": 0.0261,
      "step": 67600
    },
    {
      "epoch": 26.26,
      "learning_rate": 1.649915857605178e-05,
      "loss": 0.1132,
      "step": 67610
    },
    {
      "epoch": 26.26,
      "learning_rate": 1.6498640776699032e-05,
      "loss": 0.0565,
      "step": 67620
    },
    {
      "epoch": 26.26,
      "learning_rate": 1.649812297734628e-05,
      "loss": 0.0723,
      "step": 67630
    },
    {
      "epoch": 26.27,
      "learning_rate": 1.649760517799353e-05,
      "loss": 0.0923,
      "step": 67640
    },
    {
      "epoch": 26.27,
      "learning_rate": 1.6497087378640776e-05,
      "loss": 0.105,
      "step": 67650
    },
    {
      "epoch": 26.28,
      "learning_rate": 1.6496569579288027e-05,
      "loss": 0.0841,
      "step": 67660
    },
    {
      "epoch": 26.28,
      "learning_rate": 1.6496051779935275e-05,
      "loss": 0.046,
      "step": 67670
    },
    {
      "epoch": 26.28,
      "learning_rate": 1.6495533980582526e-05,
      "loss": 0.1664,
      "step": 67680
    },
    {
      "epoch": 26.29,
      "learning_rate": 1.6495016181229774e-05,
      "loss": 0.0819,
      "step": 67690
    },
    {
      "epoch": 26.29,
      "learning_rate": 1.6494498381877026e-05,
      "loss": 0.1499,
      "step": 67700
    },
    {
      "epoch": 26.3,
      "learning_rate": 1.6493980582524274e-05,
      "loss": 0.0944,
      "step": 67710
    },
    {
      "epoch": 26.3,
      "learning_rate": 1.6493462783171525e-05,
      "loss": 0.1706,
      "step": 67720
    },
    {
      "epoch": 26.3,
      "learning_rate": 1.649294498381877e-05,
      "loss": 0.3778,
      "step": 67730
    },
    {
      "epoch": 26.31,
      "learning_rate": 1.649242718446602e-05,
      "loss": 0.1216,
      "step": 67740
    },
    {
      "epoch": 26.31,
      "learning_rate": 1.649190938511327e-05,
      "loss": 0.0524,
      "step": 67750
    },
    {
      "epoch": 26.31,
      "learning_rate": 1.649139158576052e-05,
      "loss": 0.0661,
      "step": 67760
    },
    {
      "epoch": 26.32,
      "learning_rate": 1.6490873786407768e-05,
      "loss": 0.0861,
      "step": 67770
    },
    {
      "epoch": 26.32,
      "learning_rate": 1.649035598705502e-05,
      "loss": 0.058,
      "step": 67780
    },
    {
      "epoch": 26.33,
      "learning_rate": 1.6489838187702267e-05,
      "loss": 0.0395,
      "step": 67790
    },
    {
      "epoch": 26.33,
      "learning_rate": 1.6489320388349515e-05,
      "loss": 0.0506,
      "step": 67800
    },
    {
      "epoch": 26.33,
      "learning_rate": 1.6488802588996763e-05,
      "loss": 0.0605,
      "step": 67810
    },
    {
      "epoch": 26.34,
      "learning_rate": 1.6488284789644014e-05,
      "loss": 0.2354,
      "step": 67820
    },
    {
      "epoch": 26.34,
      "learning_rate": 1.6487766990291262e-05,
      "loss": 0.0986,
      "step": 67830
    },
    {
      "epoch": 26.35,
      "learning_rate": 1.6487249190938514e-05,
      "loss": 0.2555,
      "step": 67840
    },
    {
      "epoch": 26.35,
      "learning_rate": 1.648673139158576e-05,
      "loss": 0.1615,
      "step": 67850
    },
    {
      "epoch": 26.35,
      "learning_rate": 1.6486213592233013e-05,
      "loss": 0.184,
      "step": 67860
    },
    {
      "epoch": 26.36,
      "learning_rate": 1.648569579288026e-05,
      "loss": 0.0522,
      "step": 67870
    },
    {
      "epoch": 26.36,
      "learning_rate": 1.648517799352751e-05,
      "loss": 0.1792,
      "step": 67880
    },
    {
      "epoch": 26.37,
      "learning_rate": 1.648466019417476e-05,
      "loss": 0.0953,
      "step": 67890
    },
    {
      "epoch": 26.37,
      "learning_rate": 1.6484142394822008e-05,
      "loss": 0.1306,
      "step": 67900
    },
    {
      "epoch": 26.37,
      "learning_rate": 1.6483624595469256e-05,
      "loss": 0.0471,
      "step": 67910
    },
    {
      "epoch": 26.38,
      "learning_rate": 1.6483106796116507e-05,
      "loss": 0.1284,
      "step": 67920
    },
    {
      "epoch": 26.38,
      "learning_rate": 1.6482588996763755e-05,
      "loss": 0.0859,
      "step": 67930
    },
    {
      "epoch": 26.38,
      "learning_rate": 1.6482071197411006e-05,
      "loss": 0.1845,
      "step": 67940
    },
    {
      "epoch": 26.39,
      "learning_rate": 1.6481553398058254e-05,
      "loss": 0.1414,
      "step": 67950
    },
    {
      "epoch": 26.39,
      "learning_rate": 1.6481035598705502e-05,
      "loss": 0.1661,
      "step": 67960
    },
    {
      "epoch": 26.4,
      "learning_rate": 1.6480517799352754e-05,
      "loss": 0.1141,
      "step": 67970
    },
    {
      "epoch": 26.4,
      "learning_rate": 1.648e-05,
      "loss": 0.1002,
      "step": 67980
    },
    {
      "epoch": 26.4,
      "learning_rate": 1.647948220064725e-05,
      "loss": 0.0381,
      "step": 67990
    },
    {
      "epoch": 26.41,
      "learning_rate": 1.64789644012945e-05,
      "loss": 0.1842,
      "step": 68000
    },
    {
      "epoch": 26.41,
      "learning_rate": 1.647844660194175e-05,
      "loss": 0.0699,
      "step": 68010
    },
    {
      "epoch": 26.42,
      "learning_rate": 1.6477928802589e-05,
      "loss": 0.1172,
      "step": 68020
    },
    {
      "epoch": 26.42,
      "learning_rate": 1.6477411003236248e-05,
      "loss": 0.0726,
      "step": 68030
    },
    {
      "epoch": 26.42,
      "learning_rate": 1.6476893203883496e-05,
      "loss": 0.0632,
      "step": 68040
    },
    {
      "epoch": 26.43,
      "learning_rate": 1.6476375404530747e-05,
      "loss": 0.1615,
      "step": 68050
    },
    {
      "epoch": 26.43,
      "learning_rate": 1.6475857605177995e-05,
      "loss": 0.048,
      "step": 68060
    },
    {
      "epoch": 26.43,
      "learning_rate": 1.6475339805825243e-05,
      "loss": 0.1562,
      "step": 68070
    },
    {
      "epoch": 26.44,
      "learning_rate": 1.6474822006472494e-05,
      "loss": 0.1404,
      "step": 68080
    },
    {
      "epoch": 26.44,
      "learning_rate": 1.6474304207119742e-05,
      "loss": 0.1668,
      "step": 68090
    },
    {
      "epoch": 26.45,
      "learning_rate": 1.647378640776699e-05,
      "loss": 0.0352,
      "step": 68100
    },
    {
      "epoch": 26.45,
      "learning_rate": 1.647326860841424e-05,
      "loss": 0.0884,
      "step": 68110
    },
    {
      "epoch": 26.45,
      "learning_rate": 1.647275080906149e-05,
      "loss": 0.1149,
      "step": 68120
    },
    {
      "epoch": 26.46,
      "learning_rate": 1.647223300970874e-05,
      "loss": 0.1411,
      "step": 68130
    },
    {
      "epoch": 26.46,
      "learning_rate": 1.647171521035599e-05,
      "loss": 0.1099,
      "step": 68140
    },
    {
      "epoch": 26.47,
      "learning_rate": 1.6471197411003236e-05,
      "loss": 0.3024,
      "step": 68150
    },
    {
      "epoch": 26.47,
      "learning_rate": 1.6470679611650488e-05,
      "loss": 0.1554,
      "step": 68160
    },
    {
      "epoch": 26.47,
      "learning_rate": 1.6470161812297736e-05,
      "loss": 0.194,
      "step": 68170
    },
    {
      "epoch": 26.48,
      "learning_rate": 1.6469644012944984e-05,
      "loss": 0.0663,
      "step": 68180
    },
    {
      "epoch": 26.48,
      "learning_rate": 1.6469126213592235e-05,
      "loss": 0.2439,
      "step": 68190
    },
    {
      "epoch": 26.49,
      "learning_rate": 1.6468608414239483e-05,
      "loss": 0.0642,
      "step": 68200
    },
    {
      "epoch": 26.49,
      "learning_rate": 1.6468090614886734e-05,
      "loss": 0.1155,
      "step": 68210
    },
    {
      "epoch": 26.49,
      "learning_rate": 1.6467572815533982e-05,
      "loss": 0.048,
      "step": 68220
    },
    {
      "epoch": 26.5,
      "learning_rate": 1.646705501618123e-05,
      "loss": 0.1682,
      "step": 68230
    },
    {
      "epoch": 26.5,
      "learning_rate": 1.646653721682848e-05,
      "loss": 0.0655,
      "step": 68240
    },
    {
      "epoch": 26.5,
      "learning_rate": 1.646601941747573e-05,
      "loss": 0.0573,
      "step": 68250
    },
    {
      "epoch": 26.51,
      "learning_rate": 1.6465501618122977e-05,
      "loss": 0.1785,
      "step": 68260
    },
    {
      "epoch": 26.51,
      "learning_rate": 1.646498381877023e-05,
      "loss": 0.0853,
      "step": 68270
    },
    {
      "epoch": 26.52,
      "learning_rate": 1.6464466019417476e-05,
      "loss": 0.2649,
      "step": 68280
    },
    {
      "epoch": 26.52,
      "learning_rate": 1.6463948220064728e-05,
      "loss": 0.2312,
      "step": 68290
    },
    {
      "epoch": 26.52,
      "learning_rate": 1.6463430420711976e-05,
      "loss": 0.0677,
      "step": 68300
    },
    {
      "epoch": 26.53,
      "learning_rate": 1.6462912621359224e-05,
      "loss": 0.2945,
      "step": 68310
    },
    {
      "epoch": 26.53,
      "learning_rate": 1.6462394822006475e-05,
      "loss": 0.0391,
      "step": 68320
    },
    {
      "epoch": 26.54,
      "learning_rate": 1.6461877022653723e-05,
      "loss": 0.0836,
      "step": 68330
    },
    {
      "epoch": 26.54,
      "learning_rate": 1.646135922330097e-05,
      "loss": 0.013,
      "step": 68340
    },
    {
      "epoch": 26.54,
      "learning_rate": 1.6460841423948222e-05,
      "loss": 0.1046,
      "step": 68350
    },
    {
      "epoch": 26.55,
      "learning_rate": 1.646032362459547e-05,
      "loss": 0.1799,
      "step": 68360
    },
    {
      "epoch": 26.55,
      "learning_rate": 1.645980582524272e-05,
      "loss": 0.0841,
      "step": 68370
    },
    {
      "epoch": 26.56,
      "learning_rate": 1.645928802588997e-05,
      "loss": 0.2196,
      "step": 68380
    },
    {
      "epoch": 26.56,
      "learning_rate": 1.6458770226537217e-05,
      "loss": 0.2903,
      "step": 68390
    },
    {
      "epoch": 26.56,
      "learning_rate": 1.645825242718447e-05,
      "loss": 0.0067,
      "step": 68400
    },
    {
      "epoch": 26.57,
      "learning_rate": 1.6457734627831716e-05,
      "loss": 0.0768,
      "step": 68410
    },
    {
      "epoch": 26.57,
      "learning_rate": 1.6457216828478964e-05,
      "loss": 0.0294,
      "step": 68420
    },
    {
      "epoch": 26.57,
      "learning_rate": 1.6456699029126216e-05,
      "loss": 0.1569,
      "step": 68430
    },
    {
      "epoch": 26.58,
      "learning_rate": 1.6456181229773464e-05,
      "loss": 0.1369,
      "step": 68440
    },
    {
      "epoch": 26.58,
      "learning_rate": 1.6455663430420715e-05,
      "loss": 0.0673,
      "step": 68450
    },
    {
      "epoch": 26.59,
      "learning_rate": 1.6455145631067963e-05,
      "loss": 0.0211,
      "step": 68460
    },
    {
      "epoch": 26.59,
      "learning_rate": 1.645462783171521e-05,
      "loss": 0.1621,
      "step": 68470
    },
    {
      "epoch": 26.59,
      "learning_rate": 1.645411003236246e-05,
      "loss": 0.0463,
      "step": 68480
    },
    {
      "epoch": 26.6,
      "learning_rate": 1.645359223300971e-05,
      "loss": 0.0704,
      "step": 68490
    },
    {
      "epoch": 26.6,
      "learning_rate": 1.6453074433656958e-05,
      "loss": 0.0759,
      "step": 68500
    },
    {
      "epoch": 26.61,
      "learning_rate": 1.645255663430421e-05,
      "loss": 0.1141,
      "step": 68510
    },
    {
      "epoch": 26.61,
      "learning_rate": 1.6452038834951457e-05,
      "loss": 0.1899,
      "step": 68520
    },
    {
      "epoch": 26.61,
      "learning_rate": 1.645152103559871e-05,
      "loss": 0.2431,
      "step": 68530
    },
    {
      "epoch": 26.62,
      "learning_rate": 1.6451003236245956e-05,
      "loss": 0.066,
      "step": 68540
    },
    {
      "epoch": 26.62,
      "learning_rate": 1.6450485436893204e-05,
      "loss": 0.1753,
      "step": 68550
    },
    {
      "epoch": 26.63,
      "learning_rate": 1.6449967637540452e-05,
      "loss": 0.0816,
      "step": 68560
    },
    {
      "epoch": 26.63,
      "learning_rate": 1.6449449838187704e-05,
      "loss": 0.0254,
      "step": 68570
    },
    {
      "epoch": 26.63,
      "learning_rate": 1.644893203883495e-05,
      "loss": 0.1084,
      "step": 68580
    },
    {
      "epoch": 26.64,
      "learning_rate": 1.6448414239482203e-05,
      "loss": 0.0773,
      "step": 68590
    },
    {
      "epoch": 26.64,
      "learning_rate": 1.644789644012945e-05,
      "loss": 0.0597,
      "step": 68600
    },
    {
      "epoch": 26.64,
      "learning_rate": 1.6447378640776702e-05,
      "loss": 0.0197,
      "step": 68610
    },
    {
      "epoch": 26.65,
      "learning_rate": 1.644686084142395e-05,
      "loss": 0.0904,
      "step": 68620
    },
    {
      "epoch": 26.65,
      "learning_rate": 1.6446343042071198e-05,
      "loss": 0.0857,
      "step": 68630
    },
    {
      "epoch": 26.66,
      "learning_rate": 1.6445825242718446e-05,
      "loss": 0.115,
      "step": 68640
    },
    {
      "epoch": 26.66,
      "learning_rate": 1.6445307443365697e-05,
      "loss": 0.0874,
      "step": 68650
    },
    {
      "epoch": 26.66,
      "learning_rate": 1.6444789644012945e-05,
      "loss": 0.0221,
      "step": 68660
    },
    {
      "epoch": 26.67,
      "learning_rate": 1.6444271844660196e-05,
      "loss": 0.1814,
      "step": 68670
    },
    {
      "epoch": 26.67,
      "learning_rate": 1.6443754045307444e-05,
      "loss": 0.1038,
      "step": 68680
    },
    {
      "epoch": 26.68,
      "learning_rate": 1.6443236245954696e-05,
      "loss": 0.1718,
      "step": 68690
    },
    {
      "epoch": 26.68,
      "learning_rate": 1.6442718446601943e-05,
      "loss": 0.1603,
      "step": 68700
    },
    {
      "epoch": 26.68,
      "learning_rate": 1.644220064724919e-05,
      "loss": 0.0919,
      "step": 68710
    },
    {
      "epoch": 26.69,
      "learning_rate": 1.644168284789644e-05,
      "loss": 0.0979,
      "step": 68720
    },
    {
      "epoch": 26.69,
      "learning_rate": 1.644116504854369e-05,
      "loss": 0.0655,
      "step": 68730
    },
    {
      "epoch": 26.7,
      "learning_rate": 1.644064724919094e-05,
      "loss": 0.0955,
      "step": 68740
    },
    {
      "epoch": 26.7,
      "learning_rate": 1.644012944983819e-05,
      "loss": 0.0621,
      "step": 68750
    },
    {
      "epoch": 26.7,
      "learning_rate": 1.6439611650485438e-05,
      "loss": 0.0737,
      "step": 68760
    },
    {
      "epoch": 26.71,
      "learning_rate": 1.643909385113269e-05,
      "loss": 0.0964,
      "step": 68770
    },
    {
      "epoch": 26.71,
      "learning_rate": 1.6438576051779937e-05,
      "loss": 0.0451,
      "step": 68780
    },
    {
      "epoch": 26.71,
      "learning_rate": 1.6438058252427185e-05,
      "loss": 0.0984,
      "step": 68790
    },
    {
      "epoch": 26.72,
      "learning_rate": 1.6437540453074433e-05,
      "loss": 0.0835,
      "step": 68800
    },
    {
      "epoch": 26.72,
      "learning_rate": 1.6437022653721684e-05,
      "loss": 0.0407,
      "step": 68810
    },
    {
      "epoch": 26.73,
      "learning_rate": 1.6436504854368932e-05,
      "loss": 0.182,
      "step": 68820
    },
    {
      "epoch": 26.73,
      "learning_rate": 1.6435987055016183e-05,
      "loss": 0.0139,
      "step": 68830
    },
    {
      "epoch": 26.73,
      "learning_rate": 1.643546925566343e-05,
      "loss": 0.053,
      "step": 68840
    },
    {
      "epoch": 26.74,
      "learning_rate": 1.6434951456310683e-05,
      "loss": 0.1084,
      "step": 68850
    },
    {
      "epoch": 26.74,
      "learning_rate": 1.643443365695793e-05,
      "loss": 0.0142,
      "step": 68860
    },
    {
      "epoch": 26.75,
      "learning_rate": 1.643391585760518e-05,
      "loss": 0.0493,
      "step": 68870
    },
    {
      "epoch": 26.75,
      "learning_rate": 1.6433398058252426e-05,
      "loss": 0.0842,
      "step": 68880
    },
    {
      "epoch": 26.75,
      "learning_rate": 1.6432880258899678e-05,
      "loss": 0.0839,
      "step": 68890
    },
    {
      "epoch": 26.76,
      "learning_rate": 1.6432362459546926e-05,
      "loss": 0.0785,
      "step": 68900
    },
    {
      "epoch": 26.76,
      "learning_rate": 1.6431844660194177e-05,
      "loss": 0.0374,
      "step": 68910
    },
    {
      "epoch": 26.77,
      "learning_rate": 1.6431326860841425e-05,
      "loss": 0.1377,
      "step": 68920
    },
    {
      "epoch": 26.77,
      "learning_rate": 1.6430809061488676e-05,
      "loss": 0.1935,
      "step": 68930
    },
    {
      "epoch": 26.77,
      "learning_rate": 1.6430291262135924e-05,
      "loss": 0.1529,
      "step": 68940
    },
    {
      "epoch": 26.78,
      "learning_rate": 1.6429773462783172e-05,
      "loss": 0.1946,
      "step": 68950
    },
    {
      "epoch": 26.78,
      "learning_rate": 1.642925566343042e-05,
      "loss": 0.0249,
      "step": 68960
    },
    {
      "epoch": 26.78,
      "learning_rate": 1.642873786407767e-05,
      "loss": 0.0871,
      "step": 68970
    },
    {
      "epoch": 26.79,
      "learning_rate": 1.642822006472492e-05,
      "loss": 0.1439,
      "step": 68980
    },
    {
      "epoch": 26.79,
      "learning_rate": 1.642770226537217e-05,
      "loss": 0.077,
      "step": 68990
    },
    {
      "epoch": 26.8,
      "learning_rate": 1.642718446601942e-05,
      "loss": 0.1414,
      "step": 69000
    },
    {
      "epoch": 26.8,
      "learning_rate": 1.642666666666667e-05,
      "loss": 0.1922,
      "step": 69010
    },
    {
      "epoch": 26.8,
      "learning_rate": 1.6426148867313918e-05,
      "loss": 0.0268,
      "step": 69020
    },
    {
      "epoch": 26.81,
      "learning_rate": 1.642563106796117e-05,
      "loss": 0.1738,
      "step": 69030
    },
    {
      "epoch": 26.81,
      "learning_rate": 1.6425113268608414e-05,
      "loss": 0.0785,
      "step": 69040
    },
    {
      "epoch": 26.82,
      "learning_rate": 1.6424595469255665e-05,
      "loss": 0.224,
      "step": 69050
    },
    {
      "epoch": 26.82,
      "learning_rate": 1.6424077669902913e-05,
      "loss": 0.1111,
      "step": 69060
    },
    {
      "epoch": 26.82,
      "learning_rate": 1.6423559870550164e-05,
      "loss": 0.0062,
      "step": 69070
    },
    {
      "epoch": 26.83,
      "learning_rate": 1.6423042071197412e-05,
      "loss": 0.0424,
      "step": 69080
    },
    {
      "epoch": 26.83,
      "learning_rate": 1.6422524271844663e-05,
      "loss": 0.0462,
      "step": 69090
    },
    {
      "epoch": 26.83,
      "learning_rate": 1.642200647249191e-05,
      "loss": 0.1828,
      "step": 69100
    },
    {
      "epoch": 26.84,
      "learning_rate": 1.6421488673139163e-05,
      "loss": 0.2045,
      "step": 69110
    },
    {
      "epoch": 26.84,
      "learning_rate": 1.6420970873786407e-05,
      "loss": 0.0117,
      "step": 69120
    },
    {
      "epoch": 26.85,
      "learning_rate": 1.642045307443366e-05,
      "loss": 0.1222,
      "step": 69130
    },
    {
      "epoch": 26.85,
      "learning_rate": 1.6419935275080906e-05,
      "loss": 0.1172,
      "step": 69140
    },
    {
      "epoch": 26.85,
      "learning_rate": 1.6419417475728158e-05,
      "loss": 0.1005,
      "step": 69150
    },
    {
      "epoch": 26.86,
      "learning_rate": 1.6418899676375406e-05,
      "loss": 0.0821,
      "step": 69160
    },
    {
      "epoch": 26.86,
      "learning_rate": 1.6418381877022657e-05,
      "loss": 0.0643,
      "step": 69170
    },
    {
      "epoch": 26.87,
      "learning_rate": 1.6417864077669905e-05,
      "loss": 0.1347,
      "step": 69180
    },
    {
      "epoch": 26.87,
      "learning_rate": 1.6417346278317156e-05,
      "loss": 0.1221,
      "step": 69190
    },
    {
      "epoch": 26.87,
      "learning_rate": 1.64168284789644e-05,
      "loss": 0.1827,
      "step": 69200
    },
    {
      "epoch": 26.88,
      "learning_rate": 1.6416310679611652e-05,
      "loss": 0.0663,
      "step": 69210
    },
    {
      "epoch": 26.88,
      "learning_rate": 1.64157928802589e-05,
      "loss": 0.0018,
      "step": 69220
    },
    {
      "epoch": 26.89,
      "learning_rate": 1.641527508090615e-05,
      "loss": 0.0528,
      "step": 69230
    },
    {
      "epoch": 26.89,
      "learning_rate": 1.64147572815534e-05,
      "loss": 0.0617,
      "step": 69240
    },
    {
      "epoch": 26.89,
      "learning_rate": 1.641423948220065e-05,
      "loss": 0.164,
      "step": 69250
    },
    {
      "epoch": 26.9,
      "learning_rate": 1.64137216828479e-05,
      "loss": 0.0719,
      "step": 69260
    },
    {
      "epoch": 26.9,
      "learning_rate": 1.6413203883495146e-05,
      "loss": 0.1525,
      "step": 69270
    },
    {
      "epoch": 26.9,
      "learning_rate": 1.6412686084142394e-05,
      "loss": 0.0313,
      "step": 69280
    },
    {
      "epoch": 26.91,
      "learning_rate": 1.6412168284789646e-05,
      "loss": 0.0033,
      "step": 69290
    },
    {
      "epoch": 26.91,
      "learning_rate": 1.6411650485436893e-05,
      "loss": 0.0746,
      "step": 69300
    },
    {
      "epoch": 26.92,
      "learning_rate": 1.6411132686084145e-05,
      "loss": 0.0183,
      "step": 69310
    },
    {
      "epoch": 26.92,
      "learning_rate": 1.6410614886731393e-05,
      "loss": 0.4061,
      "step": 69320
    },
    {
      "epoch": 26.92,
      "learning_rate": 1.6410097087378644e-05,
      "loss": 0.1769,
      "step": 69330
    },
    {
      "epoch": 26.93,
      "learning_rate": 1.6409579288025892e-05,
      "loss": 0.1555,
      "step": 69340
    },
    {
      "epoch": 26.93,
      "learning_rate": 1.640906148867314e-05,
      "loss": 0.2055,
      "step": 69350
    },
    {
      "epoch": 26.94,
      "learning_rate": 1.6408543689320388e-05,
      "loss": 0.0788,
      "step": 69360
    },
    {
      "epoch": 26.94,
      "learning_rate": 1.640802588996764e-05,
      "loss": 0.0964,
      "step": 69370
    },
    {
      "epoch": 26.94,
      "learning_rate": 1.6407508090614887e-05,
      "loss": 0.0662,
      "step": 69380
    },
    {
      "epoch": 26.95,
      "learning_rate": 1.640699029126214e-05,
      "loss": 0.1674,
      "step": 69390
    },
    {
      "epoch": 26.95,
      "learning_rate": 1.6406472491909386e-05,
      "loss": 0.0424,
      "step": 69400
    },
    {
      "epoch": 26.96,
      "learning_rate": 1.6405954692556638e-05,
      "loss": 0.0581,
      "step": 69410
    },
    {
      "epoch": 26.96,
      "learning_rate": 1.6405436893203885e-05,
      "loss": 0.102,
      "step": 69420
    },
    {
      "epoch": 26.96,
      "learning_rate": 1.6404919093851133e-05,
      "loss": 0.138,
      "step": 69430
    },
    {
      "epoch": 26.97,
      "learning_rate": 1.640440129449838e-05,
      "loss": 0.0978,
      "step": 69440
    },
    {
      "epoch": 26.97,
      "learning_rate": 1.6403883495145633e-05,
      "loss": 0.109,
      "step": 69450
    },
    {
      "epoch": 26.97,
      "learning_rate": 1.640336569579288e-05,
      "loss": 0.2634,
      "step": 69460
    },
    {
      "epoch": 26.98,
      "learning_rate": 1.6402847896440132e-05,
      "loss": 0.1017,
      "step": 69470
    },
    {
      "epoch": 26.98,
      "learning_rate": 1.640233009708738e-05,
      "loss": 0.0072,
      "step": 69480
    },
    {
      "epoch": 26.99,
      "learning_rate": 1.640181229773463e-05,
      "loss": 0.1288,
      "step": 69490
    },
    {
      "epoch": 26.99,
      "learning_rate": 1.640129449838188e-05,
      "loss": 0.1487,
      "step": 69500
    },
    {
      "epoch": 26.99,
      "learning_rate": 1.6400776699029127e-05,
      "loss": 0.0333,
      "step": 69510
    },
    {
      "epoch": 27.0,
      "learning_rate": 1.6400258899676378e-05,
      "loss": 0.1257,
      "step": 69520
    },
    {
      "epoch": 27.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.27078500390052795,
      "eval_runtime": 8.1588,
      "eval_samples_per_second": 445.531,
      "eval_steps_per_second": 55.768,
      "step": 69525
    },
    {
      "epoch": 27.0,
      "learning_rate": 1.6399741100323626e-05,
      "loss": 0.0804,
      "step": 69530
    },
    {
      "epoch": 27.01,
      "learning_rate": 1.6399223300970874e-05,
      "loss": 0.007,
      "step": 69540
    },
    {
      "epoch": 27.01,
      "learning_rate": 1.6398705501618125e-05,
      "loss": 0.1715,
      "step": 69550
    },
    {
      "epoch": 27.01,
      "learning_rate": 1.6398187702265373e-05,
      "loss": 0.0659,
      "step": 69560
    },
    {
      "epoch": 27.02,
      "learning_rate": 1.639766990291262e-05,
      "loss": 0.008,
      "step": 69570
    },
    {
      "epoch": 27.02,
      "learning_rate": 1.6397152103559873e-05,
      "loss": 0.0922,
      "step": 69580
    },
    {
      "epoch": 27.03,
      "learning_rate": 1.639663430420712e-05,
      "loss": 0.0303,
      "step": 69590
    },
    {
      "epoch": 27.03,
      "learning_rate": 1.6396116504854372e-05,
      "loss": 0.1504,
      "step": 69600
    },
    {
      "epoch": 27.03,
      "learning_rate": 1.639559870550162e-05,
      "loss": 0.1904,
      "step": 69610
    },
    {
      "epoch": 27.04,
      "learning_rate": 1.6395080906148868e-05,
      "loss": 0.0808,
      "step": 69620
    },
    {
      "epoch": 27.04,
      "learning_rate": 1.639456310679612e-05,
      "loss": 0.1243,
      "step": 69630
    },
    {
      "epoch": 27.04,
      "learning_rate": 1.6394045307443367e-05,
      "loss": 0.0445,
      "step": 69640
    },
    {
      "epoch": 27.05,
      "learning_rate": 1.6393527508090615e-05,
      "loss": 0.2697,
      "step": 69650
    },
    {
      "epoch": 27.05,
      "learning_rate": 1.6393009708737866e-05,
      "loss": 0.2868,
      "step": 69660
    },
    {
      "epoch": 27.06,
      "learning_rate": 1.6392491909385114e-05,
      "loss": 0.0066,
      "step": 69670
    },
    {
      "epoch": 27.06,
      "learning_rate": 1.6391974110032365e-05,
      "loss": 0.2327,
      "step": 69680
    },
    {
      "epoch": 27.06,
      "learning_rate": 1.6391456310679613e-05,
      "loss": 0.1195,
      "step": 69690
    },
    {
      "epoch": 27.07,
      "learning_rate": 1.639093851132686e-05,
      "loss": 0.1175,
      "step": 69700
    },
    {
      "epoch": 27.07,
      "learning_rate": 1.6390420711974113e-05,
      "loss": 0.156,
      "step": 69710
    },
    {
      "epoch": 27.08,
      "learning_rate": 1.638990291262136e-05,
      "loss": 0.0828,
      "step": 69720
    },
    {
      "epoch": 27.08,
      "learning_rate": 1.638938511326861e-05,
      "loss": 0.0129,
      "step": 69730
    },
    {
      "epoch": 27.08,
      "learning_rate": 1.638886731391586e-05,
      "loss": 0.0173,
      "step": 69740
    },
    {
      "epoch": 27.09,
      "learning_rate": 1.6388349514563108e-05,
      "loss": 0.1075,
      "step": 69750
    },
    {
      "epoch": 27.09,
      "learning_rate": 1.638783171521036e-05,
      "loss": 0.1064,
      "step": 69760
    },
    {
      "epoch": 27.1,
      "learning_rate": 1.6387313915857607e-05,
      "loss": 0.0943,
      "step": 69770
    },
    {
      "epoch": 27.1,
      "learning_rate": 1.6386796116504855e-05,
      "loss": 0.0281,
      "step": 69780
    },
    {
      "epoch": 27.1,
      "learning_rate": 1.6386278317152106e-05,
      "loss": 0.1748,
      "step": 69790
    },
    {
      "epoch": 27.11,
      "learning_rate": 1.6385760517799354e-05,
      "loss": 0.0947,
      "step": 69800
    },
    {
      "epoch": 27.11,
      "learning_rate": 1.6385242718446602e-05,
      "loss": 0.1122,
      "step": 69810
    },
    {
      "epoch": 27.11,
      "learning_rate": 1.6384724919093853e-05,
      "loss": 0.0615,
      "step": 69820
    },
    {
      "epoch": 27.12,
      "learning_rate": 1.63842071197411e-05,
      "loss": 0.0361,
      "step": 69830
    },
    {
      "epoch": 27.12,
      "learning_rate": 1.6383689320388352e-05,
      "loss": 0.0861,
      "step": 69840
    },
    {
      "epoch": 27.13,
      "learning_rate": 1.63831715210356e-05,
      "loss": 0.1223,
      "step": 69850
    },
    {
      "epoch": 27.13,
      "learning_rate": 1.638265372168285e-05,
      "loss": 0.1896,
      "step": 69860
    },
    {
      "epoch": 27.13,
      "learning_rate": 1.63821359223301e-05,
      "loss": 0.0317,
      "step": 69870
    },
    {
      "epoch": 27.14,
      "learning_rate": 1.6381618122977348e-05,
      "loss": 0.1458,
      "step": 69880
    },
    {
      "epoch": 27.14,
      "learning_rate": 1.6381100323624595e-05,
      "loss": 0.2369,
      "step": 69890
    },
    {
      "epoch": 27.15,
      "learning_rate": 1.6380582524271847e-05,
      "loss": 0.2179,
      "step": 69900
    },
    {
      "epoch": 27.15,
      "learning_rate": 1.6380064724919095e-05,
      "loss": 0.0202,
      "step": 69910
    },
    {
      "epoch": 27.15,
      "learning_rate": 1.6379546925566346e-05,
      "loss": 0.0373,
      "step": 69920
    },
    {
      "epoch": 27.16,
      "learning_rate": 1.6379029126213594e-05,
      "loss": 0.1711,
      "step": 69930
    },
    {
      "epoch": 27.16,
      "learning_rate": 1.6378511326860842e-05,
      "loss": 0.0783,
      "step": 69940
    },
    {
      "epoch": 27.17,
      "learning_rate": 1.637799352750809e-05,
      "loss": 0.1276,
      "step": 69950
    },
    {
      "epoch": 27.17,
      "learning_rate": 1.637747572815534e-05,
      "loss": 0.1557,
      "step": 69960
    },
    {
      "epoch": 27.17,
      "learning_rate": 1.637695792880259e-05,
      "loss": 0.2191,
      "step": 69970
    },
    {
      "epoch": 27.18,
      "learning_rate": 1.637644012944984e-05,
      "loss": 0.1372,
      "step": 69980
    },
    {
      "epoch": 27.18,
      "learning_rate": 1.6375922330097088e-05,
      "loss": 0.0379,
      "step": 69990
    },
    {
      "epoch": 27.18,
      "learning_rate": 1.637540453074434e-05,
      "loss": 0.0897,
      "step": 70000
    },
    {
      "epoch": 27.19,
      "learning_rate": 1.6374886731391588e-05,
      "loss": 0.0311,
      "step": 70010
    },
    {
      "epoch": 27.19,
      "learning_rate": 1.6374368932038835e-05,
      "loss": 0.0144,
      "step": 70020
    },
    {
      "epoch": 27.2,
      "learning_rate": 1.6373851132686083e-05,
      "loss": 0.1099,
      "step": 70030
    },
    {
      "epoch": 27.2,
      "learning_rate": 1.6373333333333335e-05,
      "loss": 0.1565,
      "step": 70040
    },
    {
      "epoch": 27.2,
      "learning_rate": 1.6372815533980583e-05,
      "loss": 0.035,
      "step": 70050
    },
    {
      "epoch": 27.21,
      "learning_rate": 1.6372297734627834e-05,
      "loss": 0.1513,
      "step": 70060
    },
    {
      "epoch": 27.21,
      "learning_rate": 1.6371779935275082e-05,
      "loss": 0.0494,
      "step": 70070
    },
    {
      "epoch": 27.22,
      "learning_rate": 1.6371262135922333e-05,
      "loss": 0.0836,
      "step": 70080
    },
    {
      "epoch": 27.22,
      "learning_rate": 1.637074433656958e-05,
      "loss": 0.0668,
      "step": 70090
    },
    {
      "epoch": 27.22,
      "learning_rate": 1.637022653721683e-05,
      "loss": 0.1179,
      "step": 70100
    },
    {
      "epoch": 27.23,
      "learning_rate": 1.6369708737864077e-05,
      "loss": 0.0653,
      "step": 70110
    },
    {
      "epoch": 27.23,
      "learning_rate": 1.6369190938511328e-05,
      "loss": 0.1278,
      "step": 70120
    },
    {
      "epoch": 27.23,
      "learning_rate": 1.6368673139158576e-05,
      "loss": 0.2285,
      "step": 70130
    },
    {
      "epoch": 27.24,
      "learning_rate": 1.6368155339805827e-05,
      "loss": 0.0413,
      "step": 70140
    },
    {
      "epoch": 27.24,
      "learning_rate": 1.6367637540453075e-05,
      "loss": 0.1451,
      "step": 70150
    },
    {
      "epoch": 27.25,
      "learning_rate": 1.6367119741100327e-05,
      "loss": 0.0518,
      "step": 70160
    },
    {
      "epoch": 27.25,
      "learning_rate": 1.6366601941747575e-05,
      "loss": 0.0663,
      "step": 70170
    },
    {
      "epoch": 27.25,
      "learning_rate": 1.6366084142394823e-05,
      "loss": 0.2328,
      "step": 70180
    },
    {
      "epoch": 27.26,
      "learning_rate": 1.636556634304207e-05,
      "loss": 0.0438,
      "step": 70190
    },
    {
      "epoch": 27.26,
      "learning_rate": 1.6365048543689322e-05,
      "loss": 0.1098,
      "step": 70200
    },
    {
      "epoch": 27.27,
      "learning_rate": 1.636453074433657e-05,
      "loss": 0.0735,
      "step": 70210
    },
    {
      "epoch": 27.27,
      "learning_rate": 1.636401294498382e-05,
      "loss": 0.1458,
      "step": 70220
    },
    {
      "epoch": 27.27,
      "learning_rate": 1.636349514563107e-05,
      "loss": 0.0174,
      "step": 70230
    },
    {
      "epoch": 27.28,
      "learning_rate": 1.636297734627832e-05,
      "loss": 0.1714,
      "step": 70240
    },
    {
      "epoch": 27.28,
      "learning_rate": 1.6362459546925568e-05,
      "loss": 0.0386,
      "step": 70250
    },
    {
      "epoch": 27.29,
      "learning_rate": 1.6361941747572816e-05,
      "loss": 0.1508,
      "step": 70260
    },
    {
      "epoch": 27.29,
      "learning_rate": 1.6361423948220064e-05,
      "loss": 0.0589,
      "step": 70270
    },
    {
      "epoch": 27.29,
      "learning_rate": 1.6360906148867315e-05,
      "loss": 0.1071,
      "step": 70280
    },
    {
      "epoch": 27.3,
      "learning_rate": 1.6360388349514563e-05,
      "loss": 0.2155,
      "step": 70290
    },
    {
      "epoch": 27.3,
      "learning_rate": 1.6359870550161815e-05,
      "loss": 0.3224,
      "step": 70300
    },
    {
      "epoch": 27.3,
      "learning_rate": 1.6359352750809063e-05,
      "loss": 0.0213,
      "step": 70310
    },
    {
      "epoch": 27.31,
      "learning_rate": 1.6358834951456314e-05,
      "loss": 0.1747,
      "step": 70320
    },
    {
      "epoch": 27.31,
      "learning_rate": 1.6358317152103562e-05,
      "loss": 0.087,
      "step": 70330
    },
    {
      "epoch": 27.32,
      "learning_rate": 1.635779935275081e-05,
      "loss": 0.1314,
      "step": 70340
    },
    {
      "epoch": 27.32,
      "learning_rate": 1.6357281553398058e-05,
      "loss": 0.0077,
      "step": 70350
    },
    {
      "epoch": 27.32,
      "learning_rate": 1.635676375404531e-05,
      "loss": 0.2909,
      "step": 70360
    },
    {
      "epoch": 27.33,
      "learning_rate": 1.6356245954692557e-05,
      "loss": 0.0812,
      "step": 70370
    },
    {
      "epoch": 27.33,
      "learning_rate": 1.6355728155339808e-05,
      "loss": 0.0941,
      "step": 70380
    },
    {
      "epoch": 27.34,
      "learning_rate": 1.6355210355987056e-05,
      "loss": 0.0899,
      "step": 70390
    },
    {
      "epoch": 27.34,
      "learning_rate": 1.6354692556634307e-05,
      "loss": 0.0732,
      "step": 70400
    },
    {
      "epoch": 27.34,
      "learning_rate": 1.6354174757281555e-05,
      "loss": 0.115,
      "step": 70410
    },
    {
      "epoch": 27.35,
      "learning_rate": 1.6353656957928803e-05,
      "loss": 0.0934,
      "step": 70420
    },
    {
      "epoch": 27.35,
      "learning_rate": 1.635313915857605e-05,
      "loss": 0.1289,
      "step": 70430
    },
    {
      "epoch": 27.36,
      "learning_rate": 1.6352621359223302e-05,
      "loss": 0.0564,
      "step": 70440
    },
    {
      "epoch": 27.36,
      "learning_rate": 1.635210355987055e-05,
      "loss": 0.0013,
      "step": 70450
    },
    {
      "epoch": 27.36,
      "learning_rate": 1.6351585760517802e-05,
      "loss": 0.0507,
      "step": 70460
    },
    {
      "epoch": 27.37,
      "learning_rate": 1.635106796116505e-05,
      "loss": 0.097,
      "step": 70470
    },
    {
      "epoch": 27.37,
      "learning_rate": 1.63505501618123e-05,
      "loss": 0.1101,
      "step": 70480
    },
    {
      "epoch": 27.37,
      "learning_rate": 1.635003236245955e-05,
      "loss": 0.2099,
      "step": 70490
    },
    {
      "epoch": 27.38,
      "learning_rate": 1.6349514563106797e-05,
      "loss": 0.1976,
      "step": 70500
    },
    {
      "epoch": 27.38,
      "learning_rate": 1.6348996763754045e-05,
      "loss": 0.0594,
      "step": 70510
    },
    {
      "epoch": 27.39,
      "learning_rate": 1.6348478964401296e-05,
      "loss": 0.1722,
      "step": 70520
    },
    {
      "epoch": 27.39,
      "learning_rate": 1.6347961165048544e-05,
      "loss": 0.1168,
      "step": 70530
    },
    {
      "epoch": 27.39,
      "learning_rate": 1.6347443365695795e-05,
      "loss": 0.1872,
      "step": 70540
    },
    {
      "epoch": 27.4,
      "learning_rate": 1.6346925566343043e-05,
      "loss": 0.0029,
      "step": 70550
    },
    {
      "epoch": 27.4,
      "learning_rate": 1.6346407766990294e-05,
      "loss": 0.0105,
      "step": 70560
    },
    {
      "epoch": 27.41,
      "learning_rate": 1.6345889967637542e-05,
      "loss": 0.2131,
      "step": 70570
    },
    {
      "epoch": 27.41,
      "learning_rate": 1.634537216828479e-05,
      "loss": 0.0265,
      "step": 70580
    },
    {
      "epoch": 27.41,
      "learning_rate": 1.6344854368932038e-05,
      "loss": 0.1323,
      "step": 70590
    },
    {
      "epoch": 27.42,
      "learning_rate": 1.634433656957929e-05,
      "loss": 0.0396,
      "step": 70600
    },
    {
      "epoch": 27.42,
      "learning_rate": 1.6343818770226537e-05,
      "loss": 0.0135,
      "step": 70610
    },
    {
      "epoch": 27.43,
      "learning_rate": 1.634330097087379e-05,
      "loss": 0.0021,
      "step": 70620
    },
    {
      "epoch": 27.43,
      "learning_rate": 1.6342783171521037e-05,
      "loss": 0.2407,
      "step": 70630
    },
    {
      "epoch": 27.43,
      "learning_rate": 1.6342265372168288e-05,
      "loss": 0.0634,
      "step": 70640
    },
    {
      "epoch": 27.44,
      "learning_rate": 1.6341747572815536e-05,
      "loss": 0.0477,
      "step": 70650
    },
    {
      "epoch": 27.44,
      "learning_rate": 1.6341229773462787e-05,
      "loss": 0.0612,
      "step": 70660
    },
    {
      "epoch": 27.44,
      "learning_rate": 1.6340711974110032e-05,
      "loss": 0.0441,
      "step": 70670
    },
    {
      "epoch": 27.45,
      "learning_rate": 1.6340194174757283e-05,
      "loss": 0.071,
      "step": 70680
    },
    {
      "epoch": 27.45,
      "learning_rate": 1.633967637540453e-05,
      "loss": 0.0482,
      "step": 70690
    },
    {
      "epoch": 27.46,
      "learning_rate": 1.6339158576051782e-05,
      "loss": 0.1563,
      "step": 70700
    },
    {
      "epoch": 27.46,
      "learning_rate": 1.633864077669903e-05,
      "loss": 0.2769,
      "step": 70710
    },
    {
      "epoch": 27.46,
      "learning_rate": 1.633812297734628e-05,
      "loss": 0.0251,
      "step": 70720
    },
    {
      "epoch": 27.47,
      "learning_rate": 1.633760517799353e-05,
      "loss": 0.1206,
      "step": 70730
    },
    {
      "epoch": 27.47,
      "learning_rate": 1.6337087378640777e-05,
      "loss": 0.0897,
      "step": 70740
    },
    {
      "epoch": 27.48,
      "learning_rate": 1.6336569579288025e-05,
      "loss": 0.2182,
      "step": 70750
    },
    {
      "epoch": 27.48,
      "learning_rate": 1.6336051779935277e-05,
      "loss": 0.0534,
      "step": 70760
    },
    {
      "epoch": 27.48,
      "learning_rate": 1.6335533980582525e-05,
      "loss": 0.1006,
      "step": 70770
    },
    {
      "epoch": 27.49,
      "learning_rate": 1.6335016181229776e-05,
      "loss": 0.0849,
      "step": 70780
    },
    {
      "epoch": 27.49,
      "learning_rate": 1.6334498381877024e-05,
      "loss": 0.0947,
      "step": 70790
    },
    {
      "epoch": 27.5,
      "learning_rate": 1.6333980582524275e-05,
      "loss": 0.0895,
      "step": 70800
    },
    {
      "epoch": 27.5,
      "learning_rate": 1.6333462783171523e-05,
      "loss": 0.1353,
      "step": 70810
    },
    {
      "epoch": 27.5,
      "learning_rate": 1.633294498381877e-05,
      "loss": 0.0281,
      "step": 70820
    },
    {
      "epoch": 27.51,
      "learning_rate": 1.633242718446602e-05,
      "loss": 0.111,
      "step": 70830
    },
    {
      "epoch": 27.51,
      "learning_rate": 1.633190938511327e-05,
      "loss": 0.1613,
      "step": 70840
    },
    {
      "epoch": 27.51,
      "learning_rate": 1.6331391585760518e-05,
      "loss": 0.0099,
      "step": 70850
    },
    {
      "epoch": 27.52,
      "learning_rate": 1.633087378640777e-05,
      "loss": 0.1011,
      "step": 70860
    },
    {
      "epoch": 27.52,
      "learning_rate": 1.6330355987055017e-05,
      "loss": 0.1146,
      "step": 70870
    },
    {
      "epoch": 27.53,
      "learning_rate": 1.632983818770227e-05,
      "loss": 0.0456,
      "step": 70880
    },
    {
      "epoch": 27.53,
      "learning_rate": 1.6329320388349517e-05,
      "loss": 0.082,
      "step": 70890
    },
    {
      "epoch": 27.53,
      "learning_rate": 1.6328802588996765e-05,
      "loss": 0.0691,
      "step": 70900
    },
    {
      "epoch": 27.54,
      "learning_rate": 1.6328284789644012e-05,
      "loss": 0.0781,
      "step": 70910
    },
    {
      "epoch": 27.54,
      "learning_rate": 1.6327766990291264e-05,
      "loss": 0.2287,
      "step": 70920
    },
    {
      "epoch": 27.55,
      "learning_rate": 1.6327249190938512e-05,
      "loss": 0.1457,
      "step": 70930
    },
    {
      "epoch": 27.55,
      "learning_rate": 1.6326731391585763e-05,
      "loss": 0.0894,
      "step": 70940
    },
    {
      "epoch": 27.55,
      "learning_rate": 1.632621359223301e-05,
      "loss": 0.1618,
      "step": 70950
    },
    {
      "epoch": 27.56,
      "learning_rate": 1.6325695792880262e-05,
      "loss": 0.0891,
      "step": 70960
    },
    {
      "epoch": 27.56,
      "learning_rate": 1.632517799352751e-05,
      "loss": 0.0198,
      "step": 70970
    },
    {
      "epoch": 27.57,
      "learning_rate": 1.6324660194174758e-05,
      "loss": 0.1308,
      "step": 70980
    },
    {
      "epoch": 27.57,
      "learning_rate": 1.6324142394822006e-05,
      "loss": 0.1351,
      "step": 70990
    },
    {
      "epoch": 27.57,
      "learning_rate": 1.6323624595469257e-05,
      "loss": 0.041,
      "step": 71000
    },
    {
      "epoch": 27.58,
      "learning_rate": 1.6323106796116505e-05,
      "loss": 0.0527,
      "step": 71010
    },
    {
      "epoch": 27.58,
      "learning_rate": 1.6322588996763757e-05,
      "loss": 0.083,
      "step": 71020
    },
    {
      "epoch": 27.58,
      "learning_rate": 1.6322071197411005e-05,
      "loss": 0.0965,
      "step": 71030
    },
    {
      "epoch": 27.59,
      "learning_rate": 1.6321553398058252e-05,
      "loss": 0.0661,
      "step": 71040
    },
    {
      "epoch": 27.59,
      "learning_rate": 1.6321035598705504e-05,
      "loss": 0.1696,
      "step": 71050
    },
    {
      "epoch": 27.6,
      "learning_rate": 1.632051779935275e-05,
      "loss": 0.1174,
      "step": 71060
    },
    {
      "epoch": 27.6,
      "learning_rate": 1.632e-05,
      "loss": 0.0841,
      "step": 71070
    },
    {
      "epoch": 27.6,
      "learning_rate": 1.631948220064725e-05,
      "loss": 0.1746,
      "step": 71080
    },
    {
      "epoch": 27.61,
      "learning_rate": 1.63189644012945e-05,
      "loss": 0.0443,
      "step": 71090
    },
    {
      "epoch": 27.61,
      "learning_rate": 1.631844660194175e-05,
      "loss": 0.1345,
      "step": 71100
    },
    {
      "epoch": 27.62,
      "learning_rate": 1.6317928802588998e-05,
      "loss": 0.0301,
      "step": 71110
    },
    {
      "epoch": 27.62,
      "learning_rate": 1.6317411003236246e-05,
      "loss": 0.0681,
      "step": 71120
    },
    {
      "epoch": 27.62,
      "learning_rate": 1.6316893203883497e-05,
      "loss": 0.2174,
      "step": 71130
    },
    {
      "epoch": 27.63,
      "learning_rate": 1.6316375404530745e-05,
      "loss": 0.3897,
      "step": 71140
    },
    {
      "epoch": 27.63,
      "learning_rate": 1.6315857605177993e-05,
      "loss": 0.0434,
      "step": 71150
    },
    {
      "epoch": 27.63,
      "learning_rate": 1.6315339805825244e-05,
      "loss": 0.0844,
      "step": 71160
    },
    {
      "epoch": 27.64,
      "learning_rate": 1.6314822006472492e-05,
      "loss": 0.0983,
      "step": 71170
    },
    {
      "epoch": 27.64,
      "learning_rate": 1.6314304207119744e-05,
      "loss": 0.1079,
      "step": 71180
    },
    {
      "epoch": 27.65,
      "learning_rate": 1.631378640776699e-05,
      "loss": 0.0525,
      "step": 71190
    },
    {
      "epoch": 27.65,
      "learning_rate": 1.631326860841424e-05,
      "loss": 0.1151,
      "step": 71200
    },
    {
      "epoch": 27.65,
      "learning_rate": 1.631275080906149e-05,
      "loss": 0.2254,
      "step": 71210
    },
    {
      "epoch": 27.66,
      "learning_rate": 1.631223300970874e-05,
      "loss": 0.1428,
      "step": 71220
    },
    {
      "epoch": 27.66,
      "learning_rate": 1.631171521035599e-05,
      "loss": 0.1309,
      "step": 71230
    },
    {
      "epoch": 27.67,
      "learning_rate": 1.6311197411003238e-05,
      "loss": 0.0007,
      "step": 71240
    },
    {
      "epoch": 27.67,
      "learning_rate": 1.6310679611650486e-05,
      "loss": 0.0897,
      "step": 71250
    },
    {
      "epoch": 27.67,
      "learning_rate": 1.6310161812297737e-05,
      "loss": 0.1592,
      "step": 71260
    },
    {
      "epoch": 27.68,
      "learning_rate": 1.6309644012944985e-05,
      "loss": 0.1438,
      "step": 71270
    },
    {
      "epoch": 27.68,
      "learning_rate": 1.6309126213592233e-05,
      "loss": 0.1861,
      "step": 71280
    },
    {
      "epoch": 27.69,
      "learning_rate": 1.6308608414239484e-05,
      "loss": 0.0301,
      "step": 71290
    },
    {
      "epoch": 27.69,
      "learning_rate": 1.6308090614886732e-05,
      "loss": 0.1626,
      "step": 71300
    },
    {
      "epoch": 27.69,
      "learning_rate": 1.6307572815533984e-05,
      "loss": 0.0728,
      "step": 71310
    },
    {
      "epoch": 27.7,
      "learning_rate": 1.630705501618123e-05,
      "loss": 0.0821,
      "step": 71320
    },
    {
      "epoch": 27.7,
      "learning_rate": 1.630653721682848e-05,
      "loss": 0.0519,
      "step": 71330
    },
    {
      "epoch": 27.7,
      "learning_rate": 1.630601941747573e-05,
      "loss": 0.0881,
      "step": 71340
    },
    {
      "epoch": 27.71,
      "learning_rate": 1.630550161812298e-05,
      "loss": 0.043,
      "step": 71350
    },
    {
      "epoch": 27.71,
      "learning_rate": 1.6304983818770227e-05,
      "loss": 0.1049,
      "step": 71360
    },
    {
      "epoch": 27.72,
      "learning_rate": 1.6304466019417478e-05,
      "loss": 0.2282,
      "step": 71370
    },
    {
      "epoch": 27.72,
      "learning_rate": 1.6303948220064726e-05,
      "loss": 0.0611,
      "step": 71380
    },
    {
      "epoch": 27.72,
      "learning_rate": 1.6303430420711977e-05,
      "loss": 0.0403,
      "step": 71390
    },
    {
      "epoch": 27.73,
      "learning_rate": 1.6302912621359225e-05,
      "loss": 0.0271,
      "step": 71400
    },
    {
      "epoch": 27.73,
      "learning_rate": 1.6302394822006473e-05,
      "loss": 0.0658,
      "step": 71410
    },
    {
      "epoch": 27.74,
      "learning_rate": 1.630187702265372e-05,
      "loss": 0.0628,
      "step": 71420
    },
    {
      "epoch": 27.74,
      "learning_rate": 1.6301359223300972e-05,
      "loss": 0.2404,
      "step": 71430
    },
    {
      "epoch": 27.74,
      "learning_rate": 1.630084142394822e-05,
      "loss": 0.1614,
      "step": 71440
    },
    {
      "epoch": 27.75,
      "learning_rate": 1.630032362459547e-05,
      "loss": 0.3891,
      "step": 71450
    },
    {
      "epoch": 27.75,
      "learning_rate": 1.629980582524272e-05,
      "loss": 0.1018,
      "step": 71460
    },
    {
      "epoch": 27.76,
      "learning_rate": 1.629928802588997e-05,
      "loss": 0.1949,
      "step": 71470
    },
    {
      "epoch": 27.76,
      "learning_rate": 1.629877022653722e-05,
      "loss": 0.2713,
      "step": 71480
    },
    {
      "epoch": 27.76,
      "learning_rate": 1.6298252427184467e-05,
      "loss": 0.0473,
      "step": 71490
    },
    {
      "epoch": 27.77,
      "learning_rate": 1.6297734627831715e-05,
      "loss": 0.0353,
      "step": 71500
    },
    {
      "epoch": 27.77,
      "learning_rate": 1.6297216828478966e-05,
      "loss": 0.0321,
      "step": 71510
    },
    {
      "epoch": 27.77,
      "learning_rate": 1.6296699029126214e-05,
      "loss": 0.1759,
      "step": 71520
    },
    {
      "epoch": 27.78,
      "learning_rate": 1.6296181229773465e-05,
      "loss": 0.0575,
      "step": 71530
    },
    {
      "epoch": 27.78,
      "learning_rate": 1.6295663430420713e-05,
      "loss": 0.1882,
      "step": 71540
    },
    {
      "epoch": 27.79,
      "learning_rate": 1.6295145631067964e-05,
      "loss": 0.1255,
      "step": 71550
    },
    {
      "epoch": 27.79,
      "learning_rate": 1.6294627831715212e-05,
      "loss": 0.0149,
      "step": 71560
    },
    {
      "epoch": 27.79,
      "learning_rate": 1.629411003236246e-05,
      "loss": 0.1214,
      "step": 71570
    },
    {
      "epoch": 27.8,
      "learning_rate": 1.6293592233009708e-05,
      "loss": 0.1778,
      "step": 71580
    },
    {
      "epoch": 27.8,
      "learning_rate": 1.629307443365696e-05,
      "loss": 0.0918,
      "step": 71590
    },
    {
      "epoch": 27.81,
      "learning_rate": 1.6292556634304207e-05,
      "loss": 0.1516,
      "step": 71600
    },
    {
      "epoch": 27.81,
      "learning_rate": 1.629203883495146e-05,
      "loss": 0.1193,
      "step": 71610
    },
    {
      "epoch": 27.81,
      "learning_rate": 1.6291521035598707e-05,
      "loss": 0.2283,
      "step": 71620
    },
    {
      "epoch": 27.82,
      "learning_rate": 1.6291003236245958e-05,
      "loss": 0.054,
      "step": 71630
    },
    {
      "epoch": 27.82,
      "learning_rate": 1.6290485436893206e-05,
      "loss": 0.079,
      "step": 71640
    },
    {
      "epoch": 27.83,
      "learning_rate": 1.6289967637540454e-05,
      "loss": 0.0882,
      "step": 71650
    },
    {
      "epoch": 27.83,
      "learning_rate": 1.62894498381877e-05,
      "loss": 0.0023,
      "step": 71660
    },
    {
      "epoch": 27.83,
      "learning_rate": 1.6288932038834953e-05,
      "loss": 0.0357,
      "step": 71670
    },
    {
      "epoch": 27.84,
      "learning_rate": 1.62884142394822e-05,
      "loss": 0.0231,
      "step": 71680
    },
    {
      "epoch": 27.84,
      "learning_rate": 1.6287896440129452e-05,
      "loss": 0.2039,
      "step": 71690
    },
    {
      "epoch": 27.84,
      "learning_rate": 1.62873786407767e-05,
      "loss": 0.2393,
      "step": 71700
    },
    {
      "epoch": 27.85,
      "learning_rate": 1.628686084142395e-05,
      "loss": 0.1578,
      "step": 71710
    },
    {
      "epoch": 27.85,
      "learning_rate": 1.6286343042071196e-05,
      "loss": 0.0662,
      "step": 71720
    },
    {
      "epoch": 27.86,
      "learning_rate": 1.6285825242718447e-05,
      "loss": 0.1214,
      "step": 71730
    },
    {
      "epoch": 27.86,
      "learning_rate": 1.6285307443365695e-05,
      "loss": 0.0509,
      "step": 71740
    },
    {
      "epoch": 27.86,
      "learning_rate": 1.6284789644012947e-05,
      "loss": 0.0152,
      "step": 71750
    },
    {
      "epoch": 27.87,
      "learning_rate": 1.6284271844660194e-05,
      "loss": 0.0322,
      "step": 71760
    },
    {
      "epoch": 27.87,
      "learning_rate": 1.6283754045307446e-05,
      "loss": 0.1956,
      "step": 71770
    },
    {
      "epoch": 27.88,
      "learning_rate": 1.6283236245954694e-05,
      "loss": 0.1457,
      "step": 71780
    },
    {
      "epoch": 27.88,
      "learning_rate": 1.6282718446601945e-05,
      "loss": 0.0851,
      "step": 71790
    },
    {
      "epoch": 27.88,
      "learning_rate": 1.6282200647249193e-05,
      "loss": 0.1148,
      "step": 71800
    },
    {
      "epoch": 27.89,
      "learning_rate": 1.628168284789644e-05,
      "loss": 0.0048,
      "step": 71810
    },
    {
      "epoch": 27.89,
      "learning_rate": 1.628116504854369e-05,
      "loss": 0.0442,
      "step": 71820
    },
    {
      "epoch": 27.9,
      "learning_rate": 1.628064724919094e-05,
      "loss": 0.0574,
      "step": 71830
    },
    {
      "epoch": 27.9,
      "learning_rate": 1.6280129449838188e-05,
      "loss": 0.2253,
      "step": 71840
    },
    {
      "epoch": 27.9,
      "learning_rate": 1.627961165048544e-05,
      "loss": 0.1031,
      "step": 71850
    },
    {
      "epoch": 27.91,
      "learning_rate": 1.6279093851132687e-05,
      "loss": 0.0822,
      "step": 71860
    },
    {
      "epoch": 27.91,
      "learning_rate": 1.627857605177994e-05,
      "loss": 0.1899,
      "step": 71870
    },
    {
      "epoch": 27.91,
      "learning_rate": 1.6278058252427186e-05,
      "loss": 0.1379,
      "step": 71880
    },
    {
      "epoch": 27.92,
      "learning_rate": 1.6277540453074434e-05,
      "loss": 0.2689,
      "step": 71890
    },
    {
      "epoch": 27.92,
      "learning_rate": 1.6277022653721682e-05,
      "loss": 0.171,
      "step": 71900
    },
    {
      "epoch": 27.93,
      "learning_rate": 1.6276504854368934e-05,
      "loss": 0.0666,
      "step": 71910
    },
    {
      "epoch": 27.93,
      "learning_rate": 1.627598705501618e-05,
      "loss": 0.067,
      "step": 71920
    },
    {
      "epoch": 27.93,
      "learning_rate": 1.6275469255663433e-05,
      "loss": 0.1305,
      "step": 71930
    },
    {
      "epoch": 27.94,
      "learning_rate": 1.627495145631068e-05,
      "loss": 0.0449,
      "step": 71940
    },
    {
      "epoch": 27.94,
      "learning_rate": 1.6274433656957932e-05,
      "loss": 0.1,
      "step": 71950
    },
    {
      "epoch": 27.95,
      "learning_rate": 1.627391585760518e-05,
      "loss": 0.2153,
      "step": 71960
    },
    {
      "epoch": 27.95,
      "learning_rate": 1.6273398058252428e-05,
      "loss": 0.0631,
      "step": 71970
    },
    {
      "epoch": 27.95,
      "learning_rate": 1.6272880258899676e-05,
      "loss": 0.1186,
      "step": 71980
    },
    {
      "epoch": 27.96,
      "learning_rate": 1.6272362459546927e-05,
      "loss": 0.1032,
      "step": 71990
    },
    {
      "epoch": 27.96,
      "learning_rate": 1.6271844660194175e-05,
      "loss": 0.0692,
      "step": 72000
    },
    {
      "epoch": 27.97,
      "learning_rate": 1.6271326860841426e-05,
      "loss": 0.0326,
      "step": 72010
    },
    {
      "epoch": 27.97,
      "learning_rate": 1.6270809061488674e-05,
      "loss": 0.1689,
      "step": 72020
    },
    {
      "epoch": 27.97,
      "learning_rate": 1.6270291262135926e-05,
      "loss": 0.1152,
      "step": 72030
    },
    {
      "epoch": 27.98,
      "learning_rate": 1.6269773462783174e-05,
      "loss": 0.0746,
      "step": 72040
    },
    {
      "epoch": 27.98,
      "learning_rate": 1.626925566343042e-05,
      "loss": 0.1996,
      "step": 72050
    },
    {
      "epoch": 27.98,
      "learning_rate": 1.626873786407767e-05,
      "loss": 0.199,
      "step": 72060
    },
    {
      "epoch": 27.99,
      "learning_rate": 1.626822006472492e-05,
      "loss": 0.1373,
      "step": 72070
    },
    {
      "epoch": 27.99,
      "learning_rate": 1.626770226537217e-05,
      "loss": 0.1102,
      "step": 72080
    },
    {
      "epoch": 28.0,
      "learning_rate": 1.626718446601942e-05,
      "loss": 0.0525,
      "step": 72090
    },
    {
      "epoch": 28.0,
      "learning_rate": 1.6266666666666668e-05,
      "loss": 0.1097,
      "step": 72100
    },
    {
      "epoch": 28.0,
      "eval_accuracy": 0.951856946354883,
      "eval_loss": 0.2800503671169281,
      "eval_runtime": 8.1466,
      "eval_samples_per_second": 446.199,
      "eval_steps_per_second": 55.852,
      "step": 72100
    },
    {
      "epoch": 28.0,
      "learning_rate": 1.626614886731392e-05,
      "loss": 0.0425,
      "step": 72110
    },
    {
      "epoch": 28.01,
      "learning_rate": 1.6265631067961167e-05,
      "loss": 0.0513,
      "step": 72120
    },
    {
      "epoch": 28.01,
      "learning_rate": 1.6265113268608415e-05,
      "loss": 0.292,
      "step": 72130
    },
    {
      "epoch": 28.02,
      "learning_rate": 1.6264595469255663e-05,
      "loss": 0.0064,
      "step": 72140
    },
    {
      "epoch": 28.02,
      "learning_rate": 1.6264077669902914e-05,
      "loss": 0.119,
      "step": 72150
    },
    {
      "epoch": 28.02,
      "learning_rate": 1.6263559870550162e-05,
      "loss": 0.1892,
      "step": 72160
    },
    {
      "epoch": 28.03,
      "learning_rate": 1.6263042071197414e-05,
      "loss": 0.1641,
      "step": 72170
    },
    {
      "epoch": 28.03,
      "learning_rate": 1.626252427184466e-05,
      "loss": 0.0906,
      "step": 72180
    },
    {
      "epoch": 28.03,
      "learning_rate": 1.6262006472491913e-05,
      "loss": 0.1782,
      "step": 72190
    },
    {
      "epoch": 28.04,
      "learning_rate": 1.626148867313916e-05,
      "loss": 0.137,
      "step": 72200
    },
    {
      "epoch": 28.04,
      "learning_rate": 1.626097087378641e-05,
      "loss": 0.0883,
      "step": 72210
    },
    {
      "epoch": 28.05,
      "learning_rate": 1.6260453074433657e-05,
      "loss": 0.0546,
      "step": 72220
    },
    {
      "epoch": 28.05,
      "learning_rate": 1.6259935275080908e-05,
      "loss": 0.0808,
      "step": 72230
    },
    {
      "epoch": 28.05,
      "learning_rate": 1.6259417475728156e-05,
      "loss": 0.1216,
      "step": 72240
    },
    {
      "epoch": 28.06,
      "learning_rate": 1.6258899676375407e-05,
      "loss": 0.1176,
      "step": 72250
    },
    {
      "epoch": 28.06,
      "learning_rate": 1.6258381877022655e-05,
      "loss": 0.0388,
      "step": 72260
    },
    {
      "epoch": 28.07,
      "learning_rate": 1.6257864077669906e-05,
      "loss": 0.0652,
      "step": 72270
    },
    {
      "epoch": 28.07,
      "learning_rate": 1.6257346278317154e-05,
      "loss": 0.0593,
      "step": 72280
    },
    {
      "epoch": 28.07,
      "learning_rate": 1.6256828478964402e-05,
      "loss": 0.0222,
      "step": 72290
    },
    {
      "epoch": 28.08,
      "learning_rate": 1.625631067961165e-05,
      "loss": 0.0705,
      "step": 72300
    },
    {
      "epoch": 28.08,
      "learning_rate": 1.62557928802589e-05,
      "loss": 0.0409,
      "step": 72310
    },
    {
      "epoch": 28.09,
      "learning_rate": 1.625527508090615e-05,
      "loss": 0.121,
      "step": 72320
    },
    {
      "epoch": 28.09,
      "learning_rate": 1.62547572815534e-05,
      "loss": 0.2273,
      "step": 72330
    },
    {
      "epoch": 28.09,
      "learning_rate": 1.625423948220065e-05,
      "loss": 0.056,
      "step": 72340
    },
    {
      "epoch": 28.1,
      "learning_rate": 1.62537216828479e-05,
      "loss": 0.1972,
      "step": 72350
    },
    {
      "epoch": 28.1,
      "learning_rate": 1.6253203883495148e-05,
      "loss": 0.0378,
      "step": 72360
    },
    {
      "epoch": 28.1,
      "learning_rate": 1.6252686084142396e-05,
      "loss": 0.0375,
      "step": 72370
    },
    {
      "epoch": 28.11,
      "learning_rate": 1.6252168284789644e-05,
      "loss": 0.2175,
      "step": 72380
    },
    {
      "epoch": 28.11,
      "learning_rate": 1.6251650485436895e-05,
      "loss": 0.0769,
      "step": 72390
    },
    {
      "epoch": 28.12,
      "learning_rate": 1.6251132686084143e-05,
      "loss": 0.0579,
      "step": 72400
    },
    {
      "epoch": 28.12,
      "learning_rate": 1.6250614886731394e-05,
      "loss": 0.1902,
      "step": 72410
    },
    {
      "epoch": 28.12,
      "learning_rate": 1.6250097087378642e-05,
      "loss": 0.1552,
      "step": 72420
    },
    {
      "epoch": 28.13,
      "learning_rate": 1.6249579288025893e-05,
      "loss": 0.1726,
      "step": 72430
    },
    {
      "epoch": 28.13,
      "learning_rate": 1.624906148867314e-05,
      "loss": 0.0932,
      "step": 72440
    },
    {
      "epoch": 28.14,
      "learning_rate": 1.624854368932039e-05,
      "loss": 0.0876,
      "step": 72450
    },
    {
      "epoch": 28.14,
      "learning_rate": 1.6248025889967637e-05,
      "loss": 0.2169,
      "step": 72460
    },
    {
      "epoch": 28.14,
      "learning_rate": 1.624750809061489e-05,
      "loss": 0.2056,
      "step": 72470
    },
    {
      "epoch": 28.15,
      "learning_rate": 1.6246990291262136e-05,
      "loss": 0.0035,
      "step": 72480
    },
    {
      "epoch": 28.15,
      "learning_rate": 1.6246472491909388e-05,
      "loss": 0.0736,
      "step": 72490
    },
    {
      "epoch": 28.16,
      "learning_rate": 1.6245954692556636e-05,
      "loss": 0.1212,
      "step": 72500
    },
    {
      "epoch": 28.16,
      "learning_rate": 1.6245436893203884e-05,
      "loss": 0.1098,
      "step": 72510
    },
    {
      "epoch": 28.16,
      "learning_rate": 1.6244919093851135e-05,
      "loss": 0.0605,
      "step": 72520
    },
    {
      "epoch": 28.17,
      "learning_rate": 1.6244401294498383e-05,
      "loss": 0.0569,
      "step": 72530
    },
    {
      "epoch": 28.17,
      "learning_rate": 1.624388349514563e-05,
      "loss": 0.0007,
      "step": 72540
    },
    {
      "epoch": 28.17,
      "learning_rate": 1.6243365695792882e-05,
      "loss": 0.1097,
      "step": 72550
    },
    {
      "epoch": 28.18,
      "learning_rate": 1.624284789644013e-05,
      "loss": 0.0089,
      "step": 72560
    },
    {
      "epoch": 28.18,
      "learning_rate": 1.624233009708738e-05,
      "loss": 0.1139,
      "step": 72570
    },
    {
      "epoch": 28.19,
      "learning_rate": 1.624181229773463e-05,
      "loss": 0.1448,
      "step": 72580
    },
    {
      "epoch": 28.19,
      "learning_rate": 1.6241294498381877e-05,
      "loss": 0.0558,
      "step": 72590
    },
    {
      "epoch": 28.19,
      "learning_rate": 1.624077669902913e-05,
      "loss": 0.1234,
      "step": 72600
    },
    {
      "epoch": 28.2,
      "learning_rate": 1.6240258899676376e-05,
      "loss": 0.2772,
      "step": 72610
    },
    {
      "epoch": 28.2,
      "learning_rate": 1.6239741100323624e-05,
      "loss": 0.1544,
      "step": 72620
    },
    {
      "epoch": 28.21,
      "learning_rate": 1.6239223300970876e-05,
      "loss": 0.002,
      "step": 72630
    },
    {
      "epoch": 28.21,
      "learning_rate": 1.6238705501618124e-05,
      "loss": 0.1048,
      "step": 72640
    },
    {
      "epoch": 28.21,
      "learning_rate": 1.6238187702265375e-05,
      "loss": 0.0262,
      "step": 72650
    },
    {
      "epoch": 28.22,
      "learning_rate": 1.6237669902912623e-05,
      "loss": 0.121,
      "step": 72660
    },
    {
      "epoch": 28.22,
      "learning_rate": 1.623715210355987e-05,
      "loss": 0.1465,
      "step": 72670
    },
    {
      "epoch": 28.23,
      "learning_rate": 1.6236634304207122e-05,
      "loss": 0.2447,
      "step": 72680
    },
    {
      "epoch": 28.23,
      "learning_rate": 1.623611650485437e-05,
      "loss": 0.0763,
      "step": 72690
    },
    {
      "epoch": 28.23,
      "learning_rate": 1.6235598705501618e-05,
      "loss": 0.0157,
      "step": 72700
    },
    {
      "epoch": 28.24,
      "learning_rate": 1.623508090614887e-05,
      "loss": 0.0436,
      "step": 72710
    },
    {
      "epoch": 28.24,
      "learning_rate": 1.6234563106796117e-05,
      "loss": 0.0667,
      "step": 72720
    },
    {
      "epoch": 28.24,
      "learning_rate": 1.623404530744337e-05,
      "loss": 0.0556,
      "step": 72730
    },
    {
      "epoch": 28.25,
      "learning_rate": 1.6233527508090616e-05,
      "loss": 0.0529,
      "step": 72740
    },
    {
      "epoch": 28.25,
      "learning_rate": 1.6233009708737864e-05,
      "loss": 0.1948,
      "step": 72750
    },
    {
      "epoch": 28.26,
      "learning_rate": 1.6232491909385116e-05,
      "loss": 0.0318,
      "step": 72760
    },
    {
      "epoch": 28.26,
      "learning_rate": 1.6231974110032364e-05,
      "loss": 0.0259,
      "step": 72770
    },
    {
      "epoch": 28.26,
      "learning_rate": 1.623145631067961e-05,
      "loss": 0.0374,
      "step": 72780
    },
    {
      "epoch": 28.27,
      "learning_rate": 1.6230938511326863e-05,
      "loss": 0.1323,
      "step": 72790
    },
    {
      "epoch": 28.27,
      "learning_rate": 1.623042071197411e-05,
      "loss": 0.0226,
      "step": 72800
    },
    {
      "epoch": 28.28,
      "learning_rate": 1.6229902912621362e-05,
      "loss": 0.076,
      "step": 72810
    },
    {
      "epoch": 28.28,
      "learning_rate": 1.622938511326861e-05,
      "loss": 0.0097,
      "step": 72820
    },
    {
      "epoch": 28.28,
      "learning_rate": 1.6228867313915858e-05,
      "loss": 0.0789,
      "step": 72830
    },
    {
      "epoch": 28.29,
      "learning_rate": 1.622834951456311e-05,
      "loss": 0.0677,
      "step": 72840
    },
    {
      "epoch": 28.29,
      "learning_rate": 1.6227831715210357e-05,
      "loss": 0.1291,
      "step": 72850
    },
    {
      "epoch": 28.3,
      "learning_rate": 1.6227313915857605e-05,
      "loss": 0.111,
      "step": 72860
    },
    {
      "epoch": 28.3,
      "learning_rate": 1.6226796116504856e-05,
      "loss": 0.0478,
      "step": 72870
    },
    {
      "epoch": 28.3,
      "learning_rate": 1.6226278317152104e-05,
      "loss": 0.0404,
      "step": 72880
    },
    {
      "epoch": 28.31,
      "learning_rate": 1.6225760517799352e-05,
      "loss": 0.0127,
      "step": 72890
    },
    {
      "epoch": 28.31,
      "learning_rate": 1.6225242718446603e-05,
      "loss": 0.2506,
      "step": 72900
    },
    {
      "epoch": 28.31,
      "learning_rate": 1.622472491909385e-05,
      "loss": 0.1219,
      "step": 72910
    },
    {
      "epoch": 28.32,
      "learning_rate": 1.6224207119741103e-05,
      "loss": 0.2397,
      "step": 72920
    },
    {
      "epoch": 28.32,
      "learning_rate": 1.622368932038835e-05,
      "loss": 0.133,
      "step": 72930
    },
    {
      "epoch": 28.33,
      "learning_rate": 1.6223171521035602e-05,
      "loss": 0.1341,
      "step": 72940
    },
    {
      "epoch": 28.33,
      "learning_rate": 1.622265372168285e-05,
      "loss": 0.1546,
      "step": 72950
    },
    {
      "epoch": 28.33,
      "learning_rate": 1.6222135922330098e-05,
      "loss": 0.1232,
      "step": 72960
    },
    {
      "epoch": 28.34,
      "learning_rate": 1.6221618122977346e-05,
      "loss": 0.106,
      "step": 72970
    },
    {
      "epoch": 28.34,
      "learning_rate": 1.6221100323624597e-05,
      "loss": 0.1399,
      "step": 72980
    },
    {
      "epoch": 28.35,
      "learning_rate": 1.6220582524271845e-05,
      "loss": 0.1152,
      "step": 72990
    },
    {
      "epoch": 28.35,
      "learning_rate": 1.6220064724919096e-05,
      "loss": 0.0388,
      "step": 73000
    },
    {
      "epoch": 28.35,
      "learning_rate": 1.6219546925566344e-05,
      "loss": 0.0679,
      "step": 73010
    },
    {
      "epoch": 28.36,
      "learning_rate": 1.6219029126213595e-05,
      "loss": 0.1248,
      "step": 73020
    },
    {
      "epoch": 28.36,
      "learning_rate": 1.6218511326860843e-05,
      "loss": 0.0831,
      "step": 73030
    },
    {
      "epoch": 28.37,
      "learning_rate": 1.621799352750809e-05,
      "loss": 0.1739,
      "step": 73040
    },
    {
      "epoch": 28.37,
      "learning_rate": 1.621747572815534e-05,
      "loss": 0.067,
      "step": 73050
    },
    {
      "epoch": 28.37,
      "learning_rate": 1.621695792880259e-05,
      "loss": 0.036,
      "step": 73060
    },
    {
      "epoch": 28.38,
      "learning_rate": 1.621644012944984e-05,
      "loss": 0.0839,
      "step": 73070
    },
    {
      "epoch": 28.38,
      "learning_rate": 1.621592233009709e-05,
      "loss": 0.0762,
      "step": 73080
    },
    {
      "epoch": 28.38,
      "learning_rate": 1.6215404530744338e-05,
      "loss": 0.1095,
      "step": 73090
    },
    {
      "epoch": 28.39,
      "learning_rate": 1.621488673139159e-05,
      "loss": 0.0824,
      "step": 73100
    },
    {
      "epoch": 28.39,
      "learning_rate": 1.6214368932038837e-05,
      "loss": 0.0446,
      "step": 73110
    },
    {
      "epoch": 28.4,
      "learning_rate": 1.6213851132686085e-05,
      "loss": 0.1084,
      "step": 73120
    },
    {
      "epoch": 28.4,
      "learning_rate": 1.6213333333333333e-05,
      "loss": 0.0489,
      "step": 73130
    },
    {
      "epoch": 28.4,
      "learning_rate": 1.6212815533980584e-05,
      "loss": 0.1798,
      "step": 73140
    },
    {
      "epoch": 28.41,
      "learning_rate": 1.6212297734627832e-05,
      "loss": 0.0151,
      "step": 73150
    },
    {
      "epoch": 28.41,
      "learning_rate": 1.6211779935275083e-05,
      "loss": 0.0184,
      "step": 73160
    },
    {
      "epoch": 28.42,
      "learning_rate": 1.621126213592233e-05,
      "loss": 0.0868,
      "step": 73170
    },
    {
      "epoch": 28.42,
      "learning_rate": 1.6210744336569583e-05,
      "loss": 0.0572,
      "step": 73180
    },
    {
      "epoch": 28.42,
      "learning_rate": 1.6210226537216827e-05,
      "loss": 0.1809,
      "step": 73190
    },
    {
      "epoch": 28.43,
      "learning_rate": 1.620970873786408e-05,
      "loss": 0.0885,
      "step": 73200
    },
    {
      "epoch": 28.43,
      "learning_rate": 1.6209190938511326e-05,
      "loss": 0.0775,
      "step": 73210
    },
    {
      "epoch": 28.43,
      "learning_rate": 1.6208673139158578e-05,
      "loss": 0.1163,
      "step": 73220
    },
    {
      "epoch": 28.44,
      "learning_rate": 1.6208155339805826e-05,
      "loss": 0.1513,
      "step": 73230
    },
    {
      "epoch": 28.44,
      "learning_rate": 1.6207637540453077e-05,
      "loss": 0.1753,
      "step": 73240
    },
    {
      "epoch": 28.45,
      "learning_rate": 1.6207119741100325e-05,
      "loss": 0.0257,
      "step": 73250
    },
    {
      "epoch": 28.45,
      "learning_rate": 1.6206601941747576e-05,
      "loss": 0.0923,
      "step": 73260
    },
    {
      "epoch": 28.45,
      "learning_rate": 1.620608414239482e-05,
      "loss": 0.0113,
      "step": 73270
    },
    {
      "epoch": 28.46,
      "learning_rate": 1.6205566343042072e-05,
      "loss": 0.041,
      "step": 73280
    },
    {
      "epoch": 28.46,
      "learning_rate": 1.620504854368932e-05,
      "loss": 0.1642,
      "step": 73290
    },
    {
      "epoch": 28.47,
      "learning_rate": 1.620453074433657e-05,
      "loss": 0.1876,
      "step": 73300
    },
    {
      "epoch": 28.47,
      "learning_rate": 1.620401294498382e-05,
      "loss": 0.0975,
      "step": 73310
    },
    {
      "epoch": 28.47,
      "learning_rate": 1.620349514563107e-05,
      "loss": 0.1995,
      "step": 73320
    },
    {
      "epoch": 28.48,
      "learning_rate": 1.620297734627832e-05,
      "loss": 0.2974,
      "step": 73330
    },
    {
      "epoch": 28.48,
      "learning_rate": 1.620245954692557e-05,
      "loss": 0.0567,
      "step": 73340
    },
    {
      "epoch": 28.49,
      "learning_rate": 1.6201941747572814e-05,
      "loss": 0.0669,
      "step": 73350
    },
    {
      "epoch": 28.49,
      "learning_rate": 1.6201423948220066e-05,
      "loss": 0.0586,
      "step": 73360
    },
    {
      "epoch": 28.49,
      "learning_rate": 1.6200906148867313e-05,
      "loss": 0.1113,
      "step": 73370
    },
    {
      "epoch": 28.5,
      "learning_rate": 1.6200388349514565e-05,
      "loss": 0.1647,
      "step": 73380
    },
    {
      "epoch": 28.5,
      "learning_rate": 1.6199870550161813e-05,
      "loss": 0.0337,
      "step": 73390
    },
    {
      "epoch": 28.5,
      "learning_rate": 1.6199352750809064e-05,
      "loss": 0.151,
      "step": 73400
    },
    {
      "epoch": 28.51,
      "learning_rate": 1.6198834951456312e-05,
      "loss": 0.1066,
      "step": 73410
    },
    {
      "epoch": 28.51,
      "learning_rate": 1.6198317152103563e-05,
      "loss": 0.1469,
      "step": 73420
    },
    {
      "epoch": 28.52,
      "learning_rate": 1.6197799352750808e-05,
      "loss": 0.1909,
      "step": 73430
    },
    {
      "epoch": 28.52,
      "learning_rate": 1.619728155339806e-05,
      "loss": 0.1372,
      "step": 73440
    },
    {
      "epoch": 28.52,
      "learning_rate": 1.6196763754045307e-05,
      "loss": 0.0419,
      "step": 73450
    },
    {
      "epoch": 28.53,
      "learning_rate": 1.619624595469256e-05,
      "loss": 0.1797,
      "step": 73460
    },
    {
      "epoch": 28.53,
      "learning_rate": 1.6195728155339806e-05,
      "loss": 0.0652,
      "step": 73470
    },
    {
      "epoch": 28.54,
      "learning_rate": 1.6195210355987058e-05,
      "loss": 0.0712,
      "step": 73480
    },
    {
      "epoch": 28.54,
      "learning_rate": 1.6194692556634306e-05,
      "loss": 0.0792,
      "step": 73490
    },
    {
      "epoch": 28.54,
      "learning_rate": 1.6194174757281557e-05,
      "loss": 0.103,
      "step": 73500
    },
    {
      "epoch": 28.55,
      "learning_rate": 1.6193656957928805e-05,
      "loss": 0.1243,
      "step": 73510
    },
    {
      "epoch": 28.55,
      "learning_rate": 1.6193139158576053e-05,
      "loss": 0.1221,
      "step": 73520
    },
    {
      "epoch": 28.56,
      "learning_rate": 1.61926213592233e-05,
      "loss": 0.0211,
      "step": 73530
    },
    {
      "epoch": 28.56,
      "learning_rate": 1.6192103559870552e-05,
      "loss": 0.1466,
      "step": 73540
    },
    {
      "epoch": 28.56,
      "learning_rate": 1.61915857605178e-05,
      "loss": 0.0011,
      "step": 73550
    },
    {
      "epoch": 28.57,
      "learning_rate": 1.619106796116505e-05,
      "loss": 0.1886,
      "step": 73560
    },
    {
      "epoch": 28.57,
      "learning_rate": 1.61905501618123e-05,
      "loss": 0.1091,
      "step": 73570
    },
    {
      "epoch": 28.57,
      "learning_rate": 1.619003236245955e-05,
      "loss": 0.1368,
      "step": 73580
    },
    {
      "epoch": 28.58,
      "learning_rate": 1.61895145631068e-05,
      "loss": 0.3583,
      "step": 73590
    },
    {
      "epoch": 28.58,
      "learning_rate": 1.6188996763754046e-05,
      "loss": 0.0678,
      "step": 73600
    },
    {
      "epoch": 28.59,
      "learning_rate": 1.6188478964401294e-05,
      "loss": 0.0793,
      "step": 73610
    },
    {
      "epoch": 28.59,
      "learning_rate": 1.6187961165048545e-05,
      "loss": 0.135,
      "step": 73620
    },
    {
      "epoch": 28.59,
      "learning_rate": 1.6187443365695793e-05,
      "loss": 0.0352,
      "step": 73630
    },
    {
      "epoch": 28.6,
      "learning_rate": 1.6186925566343045e-05,
      "loss": 0.0855,
      "step": 73640
    },
    {
      "epoch": 28.6,
      "learning_rate": 1.6186407766990293e-05,
      "loss": 0.0677,
      "step": 73650
    },
    {
      "epoch": 28.61,
      "learning_rate": 1.6185889967637544e-05,
      "loss": 0.1002,
      "step": 73660
    },
    {
      "epoch": 28.61,
      "learning_rate": 1.6185372168284792e-05,
      "loss": 0.1231,
      "step": 73670
    },
    {
      "epoch": 28.61,
      "learning_rate": 1.618485436893204e-05,
      "loss": 0.0793,
      "step": 73680
    },
    {
      "epoch": 28.62,
      "learning_rate": 1.6184336569579288e-05,
      "loss": 0.0618,
      "step": 73690
    },
    {
      "epoch": 28.62,
      "learning_rate": 1.618381877022654e-05,
      "loss": 0.1139,
      "step": 73700
    },
    {
      "epoch": 28.63,
      "learning_rate": 1.6183300970873787e-05,
      "loss": 0.0651,
      "step": 73710
    },
    {
      "epoch": 28.63,
      "learning_rate": 1.6182783171521038e-05,
      "loss": 0.0262,
      "step": 73720
    },
    {
      "epoch": 28.63,
      "learning_rate": 1.6182265372168286e-05,
      "loss": 0.0856,
      "step": 73730
    },
    {
      "epoch": 28.64,
      "learning_rate": 1.6181747572815538e-05,
      "loss": 0.0426,
      "step": 73740
    },
    {
      "epoch": 28.64,
      "learning_rate": 1.6181229773462785e-05,
      "loss": 0.1298,
      "step": 73750
    },
    {
      "epoch": 28.64,
      "learning_rate": 1.6180711974110033e-05,
      "loss": 0.2715,
      "step": 73760
    },
    {
      "epoch": 28.65,
      "learning_rate": 1.618019417475728e-05,
      "loss": 0.0271,
      "step": 73770
    },
    {
      "epoch": 28.65,
      "learning_rate": 1.6179676375404533e-05,
      "loss": 0.0959,
      "step": 73780
    },
    {
      "epoch": 28.66,
      "learning_rate": 1.617915857605178e-05,
      "loss": 0.1338,
      "step": 73790
    },
    {
      "epoch": 28.66,
      "learning_rate": 1.6178640776699032e-05,
      "loss": 0.1895,
      "step": 73800
    },
    {
      "epoch": 28.66,
      "learning_rate": 1.617812297734628e-05,
      "loss": 0.0773,
      "step": 73810
    },
    {
      "epoch": 28.67,
      "learning_rate": 1.617760517799353e-05,
      "loss": 0.1972,
      "step": 73820
    },
    {
      "epoch": 28.67,
      "learning_rate": 1.617708737864078e-05,
      "loss": 0.004,
      "step": 73830
    },
    {
      "epoch": 28.68,
      "learning_rate": 1.6176569579288027e-05,
      "loss": 0.0944,
      "step": 73840
    },
    {
      "epoch": 28.68,
      "learning_rate": 1.6176051779935275e-05,
      "loss": 0.0505,
      "step": 73850
    },
    {
      "epoch": 28.68,
      "learning_rate": 1.6175533980582526e-05,
      "loss": 0.0363,
      "step": 73860
    },
    {
      "epoch": 28.69,
      "learning_rate": 1.6175016181229774e-05,
      "loss": 0.0046,
      "step": 73870
    },
    {
      "epoch": 28.69,
      "learning_rate": 1.6174498381877025e-05,
      "loss": 0.0367,
      "step": 73880
    },
    {
      "epoch": 28.7,
      "learning_rate": 1.6173980582524273e-05,
      "loss": 0.0847,
      "step": 73890
    },
    {
      "epoch": 28.7,
      "learning_rate": 1.6173462783171525e-05,
      "loss": 0.0003,
      "step": 73900
    },
    {
      "epoch": 28.7,
      "learning_rate": 1.6172944983818773e-05,
      "loss": 0.1213,
      "step": 73910
    },
    {
      "epoch": 28.71,
      "learning_rate": 1.617242718446602e-05,
      "loss": 0.0536,
      "step": 73920
    },
    {
      "epoch": 28.71,
      "learning_rate": 1.617190938511327e-05,
      "loss": 0.1264,
      "step": 73930
    },
    {
      "epoch": 28.71,
      "learning_rate": 1.617139158576052e-05,
      "loss": 0.1257,
      "step": 73940
    },
    {
      "epoch": 28.72,
      "learning_rate": 1.6170873786407768e-05,
      "loss": 0.06,
      "step": 73950
    },
    {
      "epoch": 28.72,
      "learning_rate": 1.617035598705502e-05,
      "loss": 0.0884,
      "step": 73960
    },
    {
      "epoch": 28.73,
      "learning_rate": 1.6169838187702267e-05,
      "loss": 0.0565,
      "step": 73970
    },
    {
      "epoch": 28.73,
      "learning_rate": 1.6169320388349515e-05,
      "loss": 0.1379,
      "step": 73980
    },
    {
      "epoch": 28.73,
      "learning_rate": 1.6168802588996766e-05,
      "loss": 0.0362,
      "step": 73990
    },
    {
      "epoch": 28.74,
      "learning_rate": 1.6168284789644014e-05,
      "loss": 0.2573,
      "step": 74000
    },
    {
      "epoch": 28.74,
      "learning_rate": 1.6167766990291262e-05,
      "loss": 0.0866,
      "step": 74010
    },
    {
      "epoch": 28.75,
      "learning_rate": 1.6167249190938513e-05,
      "loss": 0.085,
      "step": 74020
    },
    {
      "epoch": 28.75,
      "learning_rate": 1.616673139158576e-05,
      "loss": 0.0483,
      "step": 74030
    },
    {
      "epoch": 28.75,
      "learning_rate": 1.6166213592233012e-05,
      "loss": 0.0751,
      "step": 74040
    },
    {
      "epoch": 28.76,
      "learning_rate": 1.616569579288026e-05,
      "loss": 0.0948,
      "step": 74050
    },
    {
      "epoch": 28.76,
      "learning_rate": 1.616517799352751e-05,
      "loss": 0.0653,
      "step": 74060
    },
    {
      "epoch": 28.77,
      "learning_rate": 1.616466019417476e-05,
      "loss": 0.0103,
      "step": 74070
    },
    {
      "epoch": 28.77,
      "learning_rate": 1.6164142394822008e-05,
      "loss": 0.2064,
      "step": 74080
    },
    {
      "epoch": 28.77,
      "learning_rate": 1.6163624595469255e-05,
      "loss": 0.0406,
      "step": 74090
    },
    {
      "epoch": 28.78,
      "learning_rate": 1.6163106796116507e-05,
      "loss": 0.1267,
      "step": 74100
    },
    {
      "epoch": 28.78,
      "learning_rate": 1.6162588996763755e-05,
      "loss": 0.016,
      "step": 74110
    },
    {
      "epoch": 28.78,
      "learning_rate": 1.6162071197411006e-05,
      "loss": 0.0775,
      "step": 74120
    },
    {
      "epoch": 28.79,
      "learning_rate": 1.6161553398058254e-05,
      "loss": 0.008,
      "step": 74130
    },
    {
      "epoch": 28.79,
      "learning_rate": 1.6161035598705502e-05,
      "loss": 0.2745,
      "step": 74140
    },
    {
      "epoch": 28.8,
      "learning_rate": 1.6160517799352753e-05,
      "loss": 0.0465,
      "step": 74150
    },
    {
      "epoch": 28.8,
      "learning_rate": 1.616e-05,
      "loss": 0.1953,
      "step": 74160
    },
    {
      "epoch": 28.8,
      "learning_rate": 1.615948220064725e-05,
      "loss": 0.0032,
      "step": 74170
    },
    {
      "epoch": 28.81,
      "learning_rate": 1.61589644012945e-05,
      "loss": 0.1072,
      "step": 74180
    },
    {
      "epoch": 28.81,
      "learning_rate": 1.6158446601941748e-05,
      "loss": 0.161,
      "step": 74190
    },
    {
      "epoch": 28.82,
      "learning_rate": 1.6157928802589e-05,
      "loss": 0.2121,
      "step": 74200
    },
    {
      "epoch": 28.82,
      "learning_rate": 1.6157411003236248e-05,
      "loss": 0.2401,
      "step": 74210
    },
    {
      "epoch": 28.82,
      "learning_rate": 1.6156893203883495e-05,
      "loss": 0.0091,
      "step": 74220
    },
    {
      "epoch": 28.83,
      "learning_rate": 1.6156375404530747e-05,
      "loss": 0.0138,
      "step": 74230
    },
    {
      "epoch": 28.83,
      "learning_rate": 1.6155857605177995e-05,
      "loss": 0.0866,
      "step": 74240
    },
    {
      "epoch": 28.83,
      "learning_rate": 1.6155339805825243e-05,
      "loss": 0.2105,
      "step": 74250
    },
    {
      "epoch": 28.84,
      "learning_rate": 1.6154822006472494e-05,
      "loss": 0.1334,
      "step": 74260
    },
    {
      "epoch": 28.84,
      "learning_rate": 1.6154304207119742e-05,
      "loss": 0.053,
      "step": 74270
    },
    {
      "epoch": 28.85,
      "learning_rate": 1.6153786407766993e-05,
      "loss": 0.2218,
      "step": 74280
    },
    {
      "epoch": 28.85,
      "learning_rate": 1.615326860841424e-05,
      "loss": 0.0488,
      "step": 74290
    },
    {
      "epoch": 28.85,
      "learning_rate": 1.615275080906149e-05,
      "loss": 0.0948,
      "step": 74300
    },
    {
      "epoch": 28.86,
      "learning_rate": 1.615223300970874e-05,
      "loss": 0.0305,
      "step": 74310
    },
    {
      "epoch": 28.86,
      "learning_rate": 1.6151715210355988e-05,
      "loss": 0.1675,
      "step": 74320
    },
    {
      "epoch": 28.87,
      "learning_rate": 1.6151197411003236e-05,
      "loss": 0.1703,
      "step": 74330
    },
    {
      "epoch": 28.87,
      "learning_rate": 1.6150679611650487e-05,
      "loss": 0.1081,
      "step": 74340
    },
    {
      "epoch": 28.87,
      "learning_rate": 1.6150161812297735e-05,
      "loss": 0.0641,
      "step": 74350
    },
    {
      "epoch": 28.88,
      "learning_rate": 1.6149644012944983e-05,
      "loss": 0.1166,
      "step": 74360
    },
    {
      "epoch": 28.88,
      "learning_rate": 1.6149126213592235e-05,
      "loss": 0.1269,
      "step": 74370
    },
    {
      "epoch": 28.89,
      "learning_rate": 1.6148608414239483e-05,
      "loss": 0.0622,
      "step": 74380
    },
    {
      "epoch": 28.89,
      "learning_rate": 1.6148090614886734e-05,
      "loss": 0.178,
      "step": 74390
    },
    {
      "epoch": 28.89,
      "learning_rate": 1.6147572815533982e-05,
      "loss": 0.2345,
      "step": 74400
    },
    {
      "epoch": 28.9,
      "learning_rate": 1.614705501618123e-05,
      "loss": 0.0559,
      "step": 74410
    },
    {
      "epoch": 28.9,
      "learning_rate": 1.614653721682848e-05,
      "loss": 0.0899,
      "step": 74420
    },
    {
      "epoch": 28.9,
      "learning_rate": 1.614601941747573e-05,
      "loss": 0.1485,
      "step": 74430
    },
    {
      "epoch": 28.91,
      "learning_rate": 1.6145501618122977e-05,
      "loss": 0.0703,
      "step": 74440
    },
    {
      "epoch": 28.91,
      "learning_rate": 1.6144983818770228e-05,
      "loss": 0.1835,
      "step": 74450
    },
    {
      "epoch": 28.92,
      "learning_rate": 1.6144466019417476e-05,
      "loss": 0.0677,
      "step": 74460
    },
    {
      "epoch": 28.92,
      "learning_rate": 1.6143948220064727e-05,
      "loss": 0.0454,
      "step": 74470
    },
    {
      "epoch": 28.92,
      "learning_rate": 1.6143430420711975e-05,
      "loss": 0.1041,
      "step": 74480
    },
    {
      "epoch": 28.93,
      "learning_rate": 1.6142912621359223e-05,
      "loss": 0.0934,
      "step": 74490
    },
    {
      "epoch": 28.93,
      "learning_rate": 1.6142394822006475e-05,
      "loss": 0.2275,
      "step": 74500
    },
    {
      "epoch": 28.94,
      "learning_rate": 1.6141877022653723e-05,
      "loss": 0.0787,
      "step": 74510
    },
    {
      "epoch": 28.94,
      "learning_rate": 1.614135922330097e-05,
      "loss": 0.1337,
      "step": 74520
    },
    {
      "epoch": 28.94,
      "learning_rate": 1.6140841423948222e-05,
      "loss": 0.1253,
      "step": 74530
    },
    {
      "epoch": 28.95,
      "learning_rate": 1.614032362459547e-05,
      "loss": 0.0463,
      "step": 74540
    },
    {
      "epoch": 28.95,
      "learning_rate": 1.613980582524272e-05,
      "loss": 0.0901,
      "step": 74550
    },
    {
      "epoch": 28.96,
      "learning_rate": 1.613928802588997e-05,
      "loss": 0.0467,
      "step": 74560
    },
    {
      "epoch": 28.96,
      "learning_rate": 1.613877022653722e-05,
      "loss": 0.1027,
      "step": 74570
    },
    {
      "epoch": 28.96,
      "learning_rate": 1.6138252427184468e-05,
      "loss": 0.1134,
      "step": 74580
    },
    {
      "epoch": 28.97,
      "learning_rate": 1.6137734627831716e-05,
      "loss": 0.0116,
      "step": 74590
    },
    {
      "epoch": 28.97,
      "learning_rate": 1.6137216828478964e-05,
      "loss": 0.0777,
      "step": 74600
    },
    {
      "epoch": 28.97,
      "learning_rate": 1.6136699029126215e-05,
      "loss": 0.097,
      "step": 74610
    },
    {
      "epoch": 28.98,
      "learning_rate": 1.6136181229773463e-05,
      "loss": 0.2128,
      "step": 74620
    },
    {
      "epoch": 28.98,
      "learning_rate": 1.6135663430420715e-05,
      "loss": 0.1275,
      "step": 74630
    },
    {
      "epoch": 28.99,
      "learning_rate": 1.6135145631067962e-05,
      "loss": 0.1234,
      "step": 74640
    },
    {
      "epoch": 28.99,
      "learning_rate": 1.6134627831715214e-05,
      "loss": 0.213,
      "step": 74650
    },
    {
      "epoch": 28.99,
      "learning_rate": 1.613411003236246e-05,
      "loss": 0.0163,
      "step": 74660
    },
    {
      "epoch": 29.0,
      "learning_rate": 1.613359223300971e-05,
      "loss": 0.0772,
      "step": 74670
    },
    {
      "epoch": 29.0,
      "eval_accuracy": 0.9499312242090784,
      "eval_loss": 0.28697529435157776,
      "eval_runtime": 8.2691,
      "eval_samples_per_second": 439.587,
      "eval_steps_per_second": 55.024,
      "step": 74675
    },
    {
      "epoch": 29.0,
      "learning_rate": 1.6133074433656958e-05,
      "loss": 0.0545,
      "step": 74680
    },
    {
      "epoch": 29.01,
      "learning_rate": 1.613255663430421e-05,
      "loss": 0.0809,
      "step": 74690
    },
    {
      "epoch": 29.01,
      "learning_rate": 1.6132038834951457e-05,
      "loss": 0.0523,
      "step": 74700
    },
    {
      "epoch": 29.01,
      "learning_rate": 1.6131521035598708e-05,
      "loss": 0.2639,
      "step": 74710
    },
    {
      "epoch": 29.02,
      "learning_rate": 1.6131003236245956e-05,
      "loss": 0.0797,
      "step": 74720
    },
    {
      "epoch": 29.02,
      "learning_rate": 1.6130485436893207e-05,
      "loss": 0.0814,
      "step": 74730
    },
    {
      "epoch": 29.03,
      "learning_rate": 1.6129967637540452e-05,
      "loss": 0.1181,
      "step": 74740
    },
    {
      "epoch": 29.03,
      "learning_rate": 1.6129449838187703e-05,
      "loss": 0.2016,
      "step": 74750
    },
    {
      "epoch": 29.03,
      "learning_rate": 1.612893203883495e-05,
      "loss": 0.0271,
      "step": 74760
    },
    {
      "epoch": 29.04,
      "learning_rate": 1.6128414239482202e-05,
      "loss": 0.2362,
      "step": 74770
    },
    {
      "epoch": 29.04,
      "learning_rate": 1.612789644012945e-05,
      "loss": 0.081,
      "step": 74780
    },
    {
      "epoch": 29.04,
      "learning_rate": 1.61273786407767e-05,
      "loss": 0.1449,
      "step": 74790
    },
    {
      "epoch": 29.05,
      "learning_rate": 1.612686084142395e-05,
      "loss": 0.1793,
      "step": 74800
    },
    {
      "epoch": 29.05,
      "learning_rate": 1.61263430420712e-05,
      "loss": 0.0683,
      "step": 74810
    },
    {
      "epoch": 29.06,
      "learning_rate": 1.6125825242718445e-05,
      "loss": 0.0615,
      "step": 74820
    },
    {
      "epoch": 29.06,
      "learning_rate": 1.6125307443365697e-05,
      "loss": 0.1093,
      "step": 74830
    },
    {
      "epoch": 29.06,
      "learning_rate": 1.6124789644012945e-05,
      "loss": 0.0482,
      "step": 74840
    },
    {
      "epoch": 29.07,
      "learning_rate": 1.6124271844660196e-05,
      "loss": 0.0406,
      "step": 74850
    },
    {
      "epoch": 29.07,
      "learning_rate": 1.6123754045307444e-05,
      "loss": 0.0236,
      "step": 74860
    },
    {
      "epoch": 29.08,
      "learning_rate": 1.6123236245954695e-05,
      "loss": 0.0123,
      "step": 74870
    },
    {
      "epoch": 29.08,
      "learning_rate": 1.6122718446601943e-05,
      "loss": 0.0126,
      "step": 74880
    },
    {
      "epoch": 29.08,
      "learning_rate": 1.6122200647249194e-05,
      "loss": 0.0923,
      "step": 74890
    },
    {
      "epoch": 29.09,
      "learning_rate": 1.612168284789644e-05,
      "loss": 0.1948,
      "step": 74900
    },
    {
      "epoch": 29.09,
      "learning_rate": 1.612116504854369e-05,
      "loss": 0.2001,
      "step": 74910
    },
    {
      "epoch": 29.1,
      "learning_rate": 1.6120647249190938e-05,
      "loss": 0.1796,
      "step": 74920
    },
    {
      "epoch": 29.1,
      "learning_rate": 1.612012944983819e-05,
      "loss": 0.0838,
      "step": 74930
    },
    {
      "epoch": 29.1,
      "learning_rate": 1.6119611650485437e-05,
      "loss": 0.152,
      "step": 74940
    },
    {
      "epoch": 29.11,
      "learning_rate": 1.611909385113269e-05,
      "loss": 0.0312,
      "step": 74950
    },
    {
      "epoch": 29.11,
      "learning_rate": 1.6118576051779937e-05,
      "loss": 0.1139,
      "step": 74960
    },
    {
      "epoch": 29.11,
      "learning_rate": 1.6118058252427188e-05,
      "loss": 0.1211,
      "step": 74970
    },
    {
      "epoch": 29.12,
      "learning_rate": 1.6117540453074433e-05,
      "loss": 0.11,
      "step": 74980
    },
    {
      "epoch": 29.12,
      "learning_rate": 1.6117022653721684e-05,
      "loss": 0.076,
      "step": 74990
    },
    {
      "epoch": 29.13,
      "learning_rate": 1.6116504854368932e-05,
      "loss": 0.0109,
      "step": 75000
    },
    {
      "epoch": 29.13,
      "learning_rate": 1.6115987055016183e-05,
      "loss": 0.1217,
      "step": 75010
    },
    {
      "epoch": 29.13,
      "learning_rate": 1.611546925566343e-05,
      "loss": 0.0553,
      "step": 75020
    },
    {
      "epoch": 29.14,
      "learning_rate": 1.6114951456310682e-05,
      "loss": 0.0318,
      "step": 75030
    },
    {
      "epoch": 29.14,
      "learning_rate": 1.611443365695793e-05,
      "loss": 0.0797,
      "step": 75040
    },
    {
      "epoch": 29.15,
      "learning_rate": 1.611391585760518e-05,
      "loss": 0.0851,
      "step": 75050
    },
    {
      "epoch": 29.15,
      "learning_rate": 1.6113398058252426e-05,
      "loss": 0.076,
      "step": 75060
    },
    {
      "epoch": 29.15,
      "learning_rate": 1.6112880258899677e-05,
      "loss": 0.341,
      "step": 75070
    },
    {
      "epoch": 29.16,
      "learning_rate": 1.6112362459546925e-05,
      "loss": 0.0617,
      "step": 75080
    },
    {
      "epoch": 29.16,
      "learning_rate": 1.6111844660194177e-05,
      "loss": 0.1326,
      "step": 75090
    },
    {
      "epoch": 29.17,
      "learning_rate": 1.6111326860841425e-05,
      "loss": 0.0898,
      "step": 75100
    },
    {
      "epoch": 29.17,
      "learning_rate": 1.6110809061488676e-05,
      "loss": 0.0171,
      "step": 75110
    },
    {
      "epoch": 29.17,
      "learning_rate": 1.6110291262135924e-05,
      "loss": 0.1276,
      "step": 75120
    },
    {
      "epoch": 29.18,
      "learning_rate": 1.6109773462783175e-05,
      "loss": 0.0744,
      "step": 75130
    },
    {
      "epoch": 29.18,
      "learning_rate": 1.6109255663430423e-05,
      "loss": 0.1658,
      "step": 75140
    },
    {
      "epoch": 29.18,
      "learning_rate": 1.610873786407767e-05,
      "loss": 0.0424,
      "step": 75150
    },
    {
      "epoch": 29.19,
      "learning_rate": 1.610822006472492e-05,
      "loss": 0.0401,
      "step": 75160
    },
    {
      "epoch": 29.19,
      "learning_rate": 1.610770226537217e-05,
      "loss": 0.0492,
      "step": 75170
    },
    {
      "epoch": 29.2,
      "learning_rate": 1.6107184466019418e-05,
      "loss": 0.2079,
      "step": 75180
    },
    {
      "epoch": 29.2,
      "learning_rate": 1.610666666666667e-05,
      "loss": 0.1418,
      "step": 75190
    },
    {
      "epoch": 29.2,
      "learning_rate": 1.6106148867313917e-05,
      "loss": 0.0382,
      "step": 75200
    },
    {
      "epoch": 29.21,
      "learning_rate": 1.610563106796117e-05,
      "loss": 0.0121,
      "step": 75210
    },
    {
      "epoch": 29.21,
      "learning_rate": 1.6105113268608417e-05,
      "loss": 0.0222,
      "step": 75220
    },
    {
      "epoch": 29.22,
      "learning_rate": 1.6104595469255665e-05,
      "loss": 0.1503,
      "step": 75230
    },
    {
      "epoch": 29.22,
      "learning_rate": 1.6104077669902912e-05,
      "loss": 0.3979,
      "step": 75240
    },
    {
      "epoch": 29.22,
      "learning_rate": 1.6103559870550164e-05,
      "loss": 0.085,
      "step": 75250
    },
    {
      "epoch": 29.23,
      "learning_rate": 1.610304207119741e-05,
      "loss": 0.2353,
      "step": 75260
    },
    {
      "epoch": 29.23,
      "learning_rate": 1.6102524271844663e-05,
      "loss": 0.0364,
      "step": 75270
    },
    {
      "epoch": 29.23,
      "learning_rate": 1.610200647249191e-05,
      "loss": 0.1273,
      "step": 75280
    },
    {
      "epoch": 29.24,
      "learning_rate": 1.6101488673139162e-05,
      "loss": 0.0048,
      "step": 75290
    },
    {
      "epoch": 29.24,
      "learning_rate": 1.610097087378641e-05,
      "loss": 0.0949,
      "step": 75300
    },
    {
      "epoch": 29.25,
      "learning_rate": 1.6100453074433658e-05,
      "loss": 0.2348,
      "step": 75310
    },
    {
      "epoch": 29.25,
      "learning_rate": 1.6099935275080906e-05,
      "loss": 0.0316,
      "step": 75320
    },
    {
      "epoch": 29.25,
      "learning_rate": 1.6099417475728157e-05,
      "loss": 0.2552,
      "step": 75330
    },
    {
      "epoch": 29.26,
      "learning_rate": 1.6098899676375405e-05,
      "loss": 0.0184,
      "step": 75340
    },
    {
      "epoch": 29.26,
      "learning_rate": 1.6098381877022657e-05,
      "loss": 0.1043,
      "step": 75350
    },
    {
      "epoch": 29.27,
      "learning_rate": 1.6097864077669904e-05,
      "loss": 0.2142,
      "step": 75360
    },
    {
      "epoch": 29.27,
      "learning_rate": 1.6097346278317156e-05,
      "loss": 0.1248,
      "step": 75370
    },
    {
      "epoch": 29.27,
      "learning_rate": 1.6096828478964404e-05,
      "loss": 0.0927,
      "step": 75380
    },
    {
      "epoch": 29.28,
      "learning_rate": 1.609631067961165e-05,
      "loss": 0.0642,
      "step": 75390
    },
    {
      "epoch": 29.28,
      "learning_rate": 1.60957928802589e-05,
      "loss": 0.041,
      "step": 75400
    },
    {
      "epoch": 29.29,
      "learning_rate": 1.609527508090615e-05,
      "loss": 0.0398,
      "step": 75410
    },
    {
      "epoch": 29.29,
      "learning_rate": 1.60947572815534e-05,
      "loss": 0.1773,
      "step": 75420
    },
    {
      "epoch": 29.29,
      "learning_rate": 1.609423948220065e-05,
      "loss": 0.0024,
      "step": 75430
    },
    {
      "epoch": 29.3,
      "learning_rate": 1.6093721682847898e-05,
      "loss": 0.1341,
      "step": 75440
    },
    {
      "epoch": 29.3,
      "learning_rate": 1.6093203883495146e-05,
      "loss": 0.1057,
      "step": 75450
    },
    {
      "epoch": 29.3,
      "learning_rate": 1.6092686084142397e-05,
      "loss": 0.1245,
      "step": 75460
    },
    {
      "epoch": 29.31,
      "learning_rate": 1.6092168284789645e-05,
      "loss": 0.0963,
      "step": 75470
    },
    {
      "epoch": 29.31,
      "learning_rate": 1.6091650485436893e-05,
      "loss": 0.1639,
      "step": 75480
    },
    {
      "epoch": 29.32,
      "learning_rate": 1.6091132686084144e-05,
      "loss": 0.1703,
      "step": 75490
    },
    {
      "epoch": 29.32,
      "learning_rate": 1.6090614886731392e-05,
      "loss": 0.2077,
      "step": 75500
    },
    {
      "epoch": 29.32,
      "learning_rate": 1.6090097087378644e-05,
      "loss": 0.0693,
      "step": 75510
    },
    {
      "epoch": 29.33,
      "learning_rate": 1.608957928802589e-05,
      "loss": 0.0368,
      "step": 75520
    },
    {
      "epoch": 29.33,
      "learning_rate": 1.608906148867314e-05,
      "loss": 0.1151,
      "step": 75530
    },
    {
      "epoch": 29.34,
      "learning_rate": 1.608854368932039e-05,
      "loss": 0.1013,
      "step": 75540
    },
    {
      "epoch": 29.34,
      "learning_rate": 1.608802588996764e-05,
      "loss": 0.1226,
      "step": 75550
    },
    {
      "epoch": 29.34,
      "learning_rate": 1.6087508090614887e-05,
      "loss": 0.0953,
      "step": 75560
    },
    {
      "epoch": 29.35,
      "learning_rate": 1.6086990291262138e-05,
      "loss": 0.1143,
      "step": 75570
    },
    {
      "epoch": 29.35,
      "learning_rate": 1.6086472491909386e-05,
      "loss": 0.1012,
      "step": 75580
    },
    {
      "epoch": 29.36,
      "learning_rate": 1.6085954692556637e-05,
      "loss": 0.2549,
      "step": 75590
    },
    {
      "epoch": 29.36,
      "learning_rate": 1.6085436893203885e-05,
      "loss": 0.09,
      "step": 75600
    },
    {
      "epoch": 29.36,
      "learning_rate": 1.6084919093851133e-05,
      "loss": 0.0375,
      "step": 75610
    },
    {
      "epoch": 29.37,
      "learning_rate": 1.6084401294498384e-05,
      "loss": 0.0546,
      "step": 75620
    },
    {
      "epoch": 29.37,
      "learning_rate": 1.6083883495145632e-05,
      "loss": 0.0864,
      "step": 75630
    },
    {
      "epoch": 29.37,
      "learning_rate": 1.608336569579288e-05,
      "loss": 0.2091,
      "step": 75640
    },
    {
      "epoch": 29.38,
      "learning_rate": 1.608284789644013e-05,
      "loss": 0.1206,
      "step": 75650
    },
    {
      "epoch": 29.38,
      "learning_rate": 1.608233009708738e-05,
      "loss": 0.0514,
      "step": 75660
    },
    {
      "epoch": 29.39,
      "learning_rate": 1.608181229773463e-05,
      "loss": 0.0536,
      "step": 75670
    },
    {
      "epoch": 29.39,
      "learning_rate": 1.608129449838188e-05,
      "loss": 0.1172,
      "step": 75680
    },
    {
      "epoch": 29.39,
      "learning_rate": 1.6080776699029127e-05,
      "loss": 0.055,
      "step": 75690
    },
    {
      "epoch": 29.4,
      "learning_rate": 1.6080258899676378e-05,
      "loss": 0.0694,
      "step": 75700
    },
    {
      "epoch": 29.4,
      "learning_rate": 1.6079741100323626e-05,
      "loss": 0.0793,
      "step": 75710
    },
    {
      "epoch": 29.41,
      "learning_rate": 1.6079223300970874e-05,
      "loss": 0.1669,
      "step": 75720
    },
    {
      "epoch": 29.41,
      "learning_rate": 1.6078705501618125e-05,
      "loss": 0.2075,
      "step": 75730
    },
    {
      "epoch": 29.41,
      "learning_rate": 1.6078187702265373e-05,
      "loss": 0.0885,
      "step": 75740
    },
    {
      "epoch": 29.42,
      "learning_rate": 1.6077669902912624e-05,
      "loss": 0.0482,
      "step": 75750
    },
    {
      "epoch": 29.42,
      "learning_rate": 1.6077152103559872e-05,
      "loss": 0.089,
      "step": 75760
    },
    {
      "epoch": 29.43,
      "learning_rate": 1.607663430420712e-05,
      "loss": 0.2351,
      "step": 75770
    },
    {
      "epoch": 29.43,
      "learning_rate": 1.607611650485437e-05,
      "loss": 0.1133,
      "step": 75780
    },
    {
      "epoch": 29.43,
      "learning_rate": 1.607559870550162e-05,
      "loss": 0.0688,
      "step": 75790
    },
    {
      "epoch": 29.44,
      "learning_rate": 1.6075080906148867e-05,
      "loss": 0.0572,
      "step": 75800
    },
    {
      "epoch": 29.44,
      "learning_rate": 1.607456310679612e-05,
      "loss": 0.129,
      "step": 75810
    },
    {
      "epoch": 29.44,
      "learning_rate": 1.6074045307443367e-05,
      "loss": 0.1221,
      "step": 75820
    },
    {
      "epoch": 29.45,
      "learning_rate": 1.6073527508090615e-05,
      "loss": 0.0193,
      "step": 75830
    },
    {
      "epoch": 29.45,
      "learning_rate": 1.6073009708737866e-05,
      "loss": 0.113,
      "step": 75840
    },
    {
      "epoch": 29.46,
      "learning_rate": 1.6072491909385114e-05,
      "loss": 0.0355,
      "step": 75850
    },
    {
      "epoch": 29.46,
      "learning_rate": 1.6071974110032365e-05,
      "loss": 0.0701,
      "step": 75860
    },
    {
      "epoch": 29.46,
      "learning_rate": 1.6071456310679613e-05,
      "loss": 0.1277,
      "step": 75870
    },
    {
      "epoch": 29.47,
      "learning_rate": 1.607093851132686e-05,
      "loss": 0.1694,
      "step": 75880
    },
    {
      "epoch": 29.47,
      "learning_rate": 1.6070420711974112e-05,
      "loss": 0.299,
      "step": 75890
    },
    {
      "epoch": 29.48,
      "learning_rate": 1.606990291262136e-05,
      "loss": 0.0403,
      "step": 75900
    },
    {
      "epoch": 29.48,
      "learning_rate": 1.6069385113268608e-05,
      "loss": 0.0347,
      "step": 75910
    },
    {
      "epoch": 29.48,
      "learning_rate": 1.606886731391586e-05,
      "loss": 0.0623,
      "step": 75920
    },
    {
      "epoch": 29.49,
      "learning_rate": 1.6068349514563107e-05,
      "loss": 0.0309,
      "step": 75930
    },
    {
      "epoch": 29.49,
      "learning_rate": 1.606783171521036e-05,
      "loss": 0.0269,
      "step": 75940
    },
    {
      "epoch": 29.5,
      "learning_rate": 1.6067313915857607e-05,
      "loss": 0.1389,
      "step": 75950
    },
    {
      "epoch": 29.5,
      "learning_rate": 1.6066796116504854e-05,
      "loss": 0.1986,
      "step": 75960
    },
    {
      "epoch": 29.5,
      "learning_rate": 1.6066278317152106e-05,
      "loss": 0.1505,
      "step": 75970
    },
    {
      "epoch": 29.51,
      "learning_rate": 1.6065760517799354e-05,
      "loss": 0.1877,
      "step": 75980
    },
    {
      "epoch": 29.51,
      "learning_rate": 1.60652427184466e-05,
      "loss": 0.0499,
      "step": 75990
    },
    {
      "epoch": 29.51,
      "learning_rate": 1.6064724919093853e-05,
      "loss": 0.155,
      "step": 76000
    },
    {
      "epoch": 29.52,
      "learning_rate": 1.60642071197411e-05,
      "loss": 0.0553,
      "step": 76010
    },
    {
      "epoch": 29.52,
      "learning_rate": 1.6063689320388352e-05,
      "loss": 0.1507,
      "step": 76020
    },
    {
      "epoch": 29.53,
      "learning_rate": 1.60631715210356e-05,
      "loss": 0.0185,
      "step": 76030
    },
    {
      "epoch": 29.53,
      "learning_rate": 1.6062653721682848e-05,
      "loss": 0.1773,
      "step": 76040
    },
    {
      "epoch": 29.53,
      "learning_rate": 1.60621359223301e-05,
      "loss": 0.1877,
      "step": 76050
    },
    {
      "epoch": 29.54,
      "learning_rate": 1.6061618122977347e-05,
      "loss": 0.006,
      "step": 76060
    },
    {
      "epoch": 29.54,
      "learning_rate": 1.6061100323624595e-05,
      "loss": 0.096,
      "step": 76070
    },
    {
      "epoch": 29.55,
      "learning_rate": 1.6060582524271846e-05,
      "loss": 0.1379,
      "step": 76080
    },
    {
      "epoch": 29.55,
      "learning_rate": 1.6060064724919094e-05,
      "loss": 0.1841,
      "step": 76090
    },
    {
      "epoch": 29.55,
      "learning_rate": 1.6059546925566346e-05,
      "loss": 0.1317,
      "step": 76100
    },
    {
      "epoch": 29.56,
      "learning_rate": 1.6059029126213594e-05,
      "loss": 0.0853,
      "step": 76110
    },
    {
      "epoch": 29.56,
      "learning_rate": 1.605851132686084e-05,
      "loss": 0.0283,
      "step": 76120
    },
    {
      "epoch": 29.57,
      "learning_rate": 1.605799352750809e-05,
      "loss": 0.1316,
      "step": 76130
    },
    {
      "epoch": 29.57,
      "learning_rate": 1.605747572815534e-05,
      "loss": 0.02,
      "step": 76140
    },
    {
      "epoch": 29.57,
      "learning_rate": 1.605695792880259e-05,
      "loss": 0.0106,
      "step": 76150
    },
    {
      "epoch": 29.58,
      "learning_rate": 1.605644012944984e-05,
      "loss": 0.1135,
      "step": 76160
    },
    {
      "epoch": 29.58,
      "learning_rate": 1.6055922330097088e-05,
      "loss": 0.1295,
      "step": 76170
    },
    {
      "epoch": 29.58,
      "learning_rate": 1.605540453074434e-05,
      "loss": 0.0903,
      "step": 76180
    },
    {
      "epoch": 29.59,
      "learning_rate": 1.6054886731391587e-05,
      "loss": 0.133,
      "step": 76190
    },
    {
      "epoch": 29.59,
      "learning_rate": 1.6054368932038835e-05,
      "loss": 0.1343,
      "step": 76200
    },
    {
      "epoch": 29.6,
      "learning_rate": 1.6053851132686083e-05,
      "loss": 0.007,
      "step": 76210
    },
    {
      "epoch": 29.6,
      "learning_rate": 1.6053333333333334e-05,
      "loss": 0.0534,
      "step": 76220
    },
    {
      "epoch": 29.6,
      "learning_rate": 1.6052815533980582e-05,
      "loss": 0.1395,
      "step": 76230
    },
    {
      "epoch": 29.61,
      "learning_rate": 1.6052297734627834e-05,
      "loss": 0.076,
      "step": 76240
    },
    {
      "epoch": 29.61,
      "learning_rate": 1.605177993527508e-05,
      "loss": 0.1048,
      "step": 76250
    },
    {
      "epoch": 29.62,
      "learning_rate": 1.6051262135922333e-05,
      "loss": 0.1096,
      "step": 76260
    },
    {
      "epoch": 29.62,
      "learning_rate": 1.605074433656958e-05,
      "loss": 0.0972,
      "step": 76270
    },
    {
      "epoch": 29.62,
      "learning_rate": 1.6050226537216832e-05,
      "loss": 0.1224,
      "step": 76280
    },
    {
      "epoch": 29.63,
      "learning_rate": 1.6049708737864077e-05,
      "loss": 0.2984,
      "step": 76290
    },
    {
      "epoch": 29.63,
      "learning_rate": 1.6049190938511328e-05,
      "loss": 0.0037,
      "step": 76300
    },
    {
      "epoch": 29.63,
      "learning_rate": 1.6048673139158576e-05,
      "loss": 0.1175,
      "step": 76310
    },
    {
      "epoch": 29.64,
      "learning_rate": 1.6048155339805827e-05,
      "loss": 0.0328,
      "step": 76320
    },
    {
      "epoch": 29.64,
      "learning_rate": 1.6047637540453075e-05,
      "loss": 0.0879,
      "step": 76330
    },
    {
      "epoch": 29.65,
      "learning_rate": 1.6047119741100326e-05,
      "loss": 0.0899,
      "step": 76340
    },
    {
      "epoch": 29.65,
      "learning_rate": 1.6046601941747574e-05,
      "loss": 0.1657,
      "step": 76350
    },
    {
      "epoch": 29.65,
      "learning_rate": 1.6046084142394826e-05,
      "loss": 0.0794,
      "step": 76360
    },
    {
      "epoch": 29.66,
      "learning_rate": 1.604556634304207e-05,
      "loss": 0.1634,
      "step": 76370
    },
    {
      "epoch": 29.66,
      "learning_rate": 1.604504854368932e-05,
      "loss": 0.0171,
      "step": 76380
    },
    {
      "epoch": 29.67,
      "learning_rate": 1.604453074433657e-05,
      "loss": 0.0375,
      "step": 76390
    },
    {
      "epoch": 29.67,
      "learning_rate": 1.604401294498382e-05,
      "loss": 0.022,
      "step": 76400
    },
    {
      "epoch": 29.67,
      "learning_rate": 1.604349514563107e-05,
      "loss": 0.0801,
      "step": 76410
    },
    {
      "epoch": 29.68,
      "learning_rate": 1.604297734627832e-05,
      "loss": 0.1211,
      "step": 76420
    },
    {
      "epoch": 29.68,
      "learning_rate": 1.6042459546925568e-05,
      "loss": 0.1322,
      "step": 76430
    },
    {
      "epoch": 29.69,
      "learning_rate": 1.604194174757282e-05,
      "loss": 0.0265,
      "step": 76440
    },
    {
      "epoch": 29.69,
      "learning_rate": 1.6041423948220064e-05,
      "loss": 0.0865,
      "step": 76450
    },
    {
      "epoch": 29.69,
      "learning_rate": 1.6040906148867315e-05,
      "loss": 0.1182,
      "step": 76460
    },
    {
      "epoch": 29.7,
      "learning_rate": 1.6040388349514563e-05,
      "loss": 0.0817,
      "step": 76470
    },
    {
      "epoch": 29.7,
      "learning_rate": 1.6039870550161814e-05,
      "loss": 0.1714,
      "step": 76480
    },
    {
      "epoch": 29.7,
      "learning_rate": 1.6039352750809062e-05,
      "loss": 0.1174,
      "step": 76490
    },
    {
      "epoch": 29.71,
      "learning_rate": 1.6038834951456313e-05,
      "loss": 0.1867,
      "step": 76500
    },
    {
      "epoch": 29.71,
      "learning_rate": 1.603831715210356e-05,
      "loss": 0.0738,
      "step": 76510
    },
    {
      "epoch": 29.72,
      "learning_rate": 1.6037799352750813e-05,
      "loss": 0.0688,
      "step": 76520
    },
    {
      "epoch": 29.72,
      "learning_rate": 1.6037281553398057e-05,
      "loss": 0.0673,
      "step": 76530
    },
    {
      "epoch": 29.72,
      "learning_rate": 1.603676375404531e-05,
      "loss": 0.0995,
      "step": 76540
    },
    {
      "epoch": 29.73,
      "learning_rate": 1.6036245954692557e-05,
      "loss": 0.076,
      "step": 76550
    },
    {
      "epoch": 29.73,
      "learning_rate": 1.6035728155339808e-05,
      "loss": 0.0972,
      "step": 76560
    },
    {
      "epoch": 29.74,
      "learning_rate": 1.6035210355987056e-05,
      "loss": 0.1128,
      "step": 76570
    },
    {
      "epoch": 29.74,
      "learning_rate": 1.6034692556634307e-05,
      "loss": 0.0687,
      "step": 76580
    },
    {
      "epoch": 29.74,
      "learning_rate": 1.6034174757281555e-05,
      "loss": 0.0484,
      "step": 76590
    },
    {
      "epoch": 29.75,
      "learning_rate": 1.6033656957928806e-05,
      "loss": 0.1293,
      "step": 76600
    },
    {
      "epoch": 29.75,
      "learning_rate": 1.603313915857605e-05,
      "loss": 0.348,
      "step": 76610
    },
    {
      "epoch": 29.76,
      "learning_rate": 1.6032621359223302e-05,
      "loss": 0.1696,
      "step": 76620
    },
    {
      "epoch": 29.76,
      "learning_rate": 1.603210355987055e-05,
      "loss": 0.1494,
      "step": 76630
    },
    {
      "epoch": 29.76,
      "learning_rate": 1.60315857605178e-05,
      "loss": 0.2455,
      "step": 76640
    },
    {
      "epoch": 29.77,
      "learning_rate": 1.603106796116505e-05,
      "loss": 0.23,
      "step": 76650
    },
    {
      "epoch": 29.77,
      "learning_rate": 1.60305501618123e-05,
      "loss": 0.0802,
      "step": 76660
    },
    {
      "epoch": 29.77,
      "learning_rate": 1.603003236245955e-05,
      "loss": 0.0397,
      "step": 76670
    },
    {
      "epoch": 29.78,
      "learning_rate": 1.60295145631068e-05,
      "loss": 0.0308,
      "step": 76680
    },
    {
      "epoch": 29.78,
      "learning_rate": 1.6028996763754044e-05,
      "loss": 0.1286,
      "step": 76690
    },
    {
      "epoch": 29.79,
      "learning_rate": 1.6028478964401296e-05,
      "loss": 0.1032,
      "step": 76700
    },
    {
      "epoch": 29.79,
      "learning_rate": 1.6027961165048544e-05,
      "loss": 0.0076,
      "step": 76710
    },
    {
      "epoch": 29.79,
      "learning_rate": 1.6027443365695795e-05,
      "loss": 0.1466,
      "step": 76720
    },
    {
      "epoch": 29.8,
      "learning_rate": 1.6026925566343043e-05,
      "loss": 0.0929,
      "step": 76730
    },
    {
      "epoch": 29.8,
      "learning_rate": 1.6026407766990294e-05,
      "loss": 0.0549,
      "step": 76740
    },
    {
      "epoch": 29.81,
      "learning_rate": 1.6025889967637542e-05,
      "loss": 0.102,
      "step": 76750
    },
    {
      "epoch": 29.81,
      "learning_rate": 1.6025372168284793e-05,
      "loss": 0.1141,
      "step": 76760
    },
    {
      "epoch": 29.81,
      "learning_rate": 1.6024854368932038e-05,
      "loss": 0.0856,
      "step": 76770
    },
    {
      "epoch": 29.82,
      "learning_rate": 1.602433656957929e-05,
      "loss": 0.1484,
      "step": 76780
    },
    {
      "epoch": 29.82,
      "learning_rate": 1.6023818770226537e-05,
      "loss": 0.3503,
      "step": 76790
    },
    {
      "epoch": 29.83,
      "learning_rate": 1.602330097087379e-05,
      "loss": 0.0135,
      "step": 76800
    },
    {
      "epoch": 29.83,
      "learning_rate": 1.6022783171521036e-05,
      "loss": 0.0364,
      "step": 76810
    },
    {
      "epoch": 29.83,
      "learning_rate": 1.6022265372168288e-05,
      "loss": 0.086,
      "step": 76820
    },
    {
      "epoch": 29.84,
      "learning_rate": 1.6021747572815536e-05,
      "loss": 0.2839,
      "step": 76830
    },
    {
      "epoch": 29.84,
      "learning_rate": 1.6021229773462787e-05,
      "loss": 0.0214,
      "step": 76840
    },
    {
      "epoch": 29.84,
      "learning_rate": 1.6020711974110035e-05,
      "loss": 0.0214,
      "step": 76850
    },
    {
      "epoch": 29.85,
      "learning_rate": 1.6020194174757283e-05,
      "loss": 0.0309,
      "step": 76860
    },
    {
      "epoch": 29.85,
      "learning_rate": 1.601967637540453e-05,
      "loss": 0.3176,
      "step": 76870
    },
    {
      "epoch": 29.86,
      "learning_rate": 1.6019158576051782e-05,
      "loss": 0.0974,
      "step": 76880
    },
    {
      "epoch": 29.86,
      "learning_rate": 1.601864077669903e-05,
      "loss": 0.0458,
      "step": 76890
    },
    {
      "epoch": 29.86,
      "learning_rate": 1.601812297734628e-05,
      "loss": 0.0788,
      "step": 76900
    },
    {
      "epoch": 29.87,
      "learning_rate": 1.601760517799353e-05,
      "loss": 0.0787,
      "step": 76910
    },
    {
      "epoch": 29.87,
      "learning_rate": 1.6017087378640777e-05,
      "loss": 0.199,
      "step": 76920
    },
    {
      "epoch": 29.88,
      "learning_rate": 1.601656957928803e-05,
      "loss": 0.2111,
      "step": 76930
    },
    {
      "epoch": 29.88,
      "learning_rate": 1.6016051779935276e-05,
      "loss": 0.0715,
      "step": 76940
    },
    {
      "epoch": 29.88,
      "learning_rate": 1.6015533980582524e-05,
      "loss": 0.0519,
      "step": 76950
    },
    {
      "epoch": 29.89,
      "learning_rate": 1.6015016181229776e-05,
      "loss": 0.0348,
      "step": 76960
    },
    {
      "epoch": 29.89,
      "learning_rate": 1.6014498381877024e-05,
      "loss": 0.0396,
      "step": 76970
    },
    {
      "epoch": 29.9,
      "learning_rate": 1.6013980582524275e-05,
      "loss": 0.2503,
      "step": 76980
    },
    {
      "epoch": 29.9,
      "learning_rate": 1.6013462783171523e-05,
      "loss": 0.0312,
      "step": 76990
    },
    {
      "epoch": 29.9,
      "learning_rate": 1.601294498381877e-05,
      "loss": 0.0995,
      "step": 77000
    },
    {
      "epoch": 29.91,
      "learning_rate": 1.6012427184466022e-05,
      "loss": 0.0472,
      "step": 77010
    },
    {
      "epoch": 29.91,
      "learning_rate": 1.601190938511327e-05,
      "loss": 0.0589,
      "step": 77020
    },
    {
      "epoch": 29.91,
      "learning_rate": 1.6011391585760518e-05,
      "loss": 0.0698,
      "step": 77030
    },
    {
      "epoch": 29.92,
      "learning_rate": 1.601087378640777e-05,
      "loss": 0.0796,
      "step": 77040
    },
    {
      "epoch": 29.92,
      "learning_rate": 1.6010355987055017e-05,
      "loss": 0.0761,
      "step": 77050
    },
    {
      "epoch": 29.93,
      "learning_rate": 1.600983818770227e-05,
      "loss": 0.1826,
      "step": 77060
    },
    {
      "epoch": 29.93,
      "learning_rate": 1.6009320388349516e-05,
      "loss": 0.0538,
      "step": 77070
    },
    {
      "epoch": 29.93,
      "learning_rate": 1.6008802588996764e-05,
      "loss": 0.1193,
      "step": 77080
    },
    {
      "epoch": 29.94,
      "learning_rate": 1.6008284789644016e-05,
      "loss": 0.2437,
      "step": 77090
    },
    {
      "epoch": 29.94,
      "learning_rate": 1.6007766990291263e-05,
      "loss": 0.106,
      "step": 77100
    },
    {
      "epoch": 29.95,
      "learning_rate": 1.600724919093851e-05,
      "loss": 0.1036,
      "step": 77110
    },
    {
      "epoch": 29.95,
      "learning_rate": 1.6006731391585763e-05,
      "loss": 0.0417,
      "step": 77120
    },
    {
      "epoch": 29.95,
      "learning_rate": 1.600621359223301e-05,
      "loss": 0.1161,
      "step": 77130
    },
    {
      "epoch": 29.96,
      "learning_rate": 1.6005695792880262e-05,
      "loss": 0.002,
      "step": 77140
    },
    {
      "epoch": 29.96,
      "learning_rate": 1.600517799352751e-05,
      "loss": 0.1067,
      "step": 77150
    },
    {
      "epoch": 29.97,
      "learning_rate": 1.6004660194174758e-05,
      "loss": 0.2566,
      "step": 77160
    },
    {
      "epoch": 29.97,
      "learning_rate": 1.600414239482201e-05,
      "loss": 0.1103,
      "step": 77170
    },
    {
      "epoch": 29.97,
      "learning_rate": 1.6003624595469257e-05,
      "loss": 0.0126,
      "step": 77180
    },
    {
      "epoch": 29.98,
      "learning_rate": 1.6003106796116505e-05,
      "loss": 0.08,
      "step": 77190
    },
    {
      "epoch": 29.98,
      "learning_rate": 1.6002588996763756e-05,
      "loss": 0.0198,
      "step": 77200
    },
    {
      "epoch": 29.98,
      "learning_rate": 1.6002071197411004e-05,
      "loss": 0.1228,
      "step": 77210
    },
    {
      "epoch": 29.99,
      "learning_rate": 1.6001553398058252e-05,
      "loss": 0.0993,
      "step": 77220
    },
    {
      "epoch": 29.99,
      "learning_rate": 1.6001035598705503e-05,
      "loss": 0.1189,
      "step": 77230
    },
    {
      "epoch": 30.0,
      "learning_rate": 1.600051779935275e-05,
      "loss": 0.0391,
      "step": 77240
    },
    {
      "epoch": 30.0,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 0.132,
      "step": 77250
    },
    {
      "epoch": 30.0,
      "eval_accuracy": 0.949656121045392,
      "eval_loss": 0.28239795565605164,
      "eval_runtime": 8.2813,
      "eval_samples_per_second": 438.943,
      "eval_steps_per_second": 54.943,
      "step": 77250
    },
    {
      "epoch": 30.0,
      "learning_rate": 1.599948220064725e-05,
      "loss": 0.0245,
      "step": 77260
    },
    {
      "epoch": 30.01,
      "learning_rate": 1.59989644012945e-05,
      "loss": 0.1941,
      "step": 77270
    },
    {
      "epoch": 30.01,
      "learning_rate": 1.599844660194175e-05,
      "loss": 0.0267,
      "step": 77280
    },
    {
      "epoch": 30.02,
      "learning_rate": 1.5997928802588998e-05,
      "loss": 0.0661,
      "step": 77290
    },
    {
      "epoch": 30.02,
      "learning_rate": 1.5997411003236246e-05,
      "loss": 0.1029,
      "step": 77300
    },
    {
      "epoch": 30.02,
      "learning_rate": 1.5996893203883497e-05,
      "loss": 0.0483,
      "step": 77310
    },
    {
      "epoch": 30.03,
      "learning_rate": 1.5996375404530745e-05,
      "loss": 0.1283,
      "step": 77320
    },
    {
      "epoch": 30.03,
      "learning_rate": 1.5995857605177996e-05,
      "loss": 0.1758,
      "step": 77330
    },
    {
      "epoch": 30.03,
      "learning_rate": 1.5995339805825244e-05,
      "loss": 0.1112,
      "step": 77340
    },
    {
      "epoch": 30.04,
      "learning_rate": 1.5994822006472492e-05,
      "loss": 0.0491,
      "step": 77350
    },
    {
      "epoch": 30.04,
      "learning_rate": 1.5994304207119743e-05,
      "loss": 0.1012,
      "step": 77360
    },
    {
      "epoch": 30.05,
      "learning_rate": 1.599378640776699e-05,
      "loss": 0.0417,
      "step": 77370
    },
    {
      "epoch": 30.05,
      "learning_rate": 1.599326860841424e-05,
      "loss": 0.1323,
      "step": 77380
    },
    {
      "epoch": 30.05,
      "learning_rate": 1.599275080906149e-05,
      "loss": 0.0551,
      "step": 77390
    },
    {
      "epoch": 30.06,
      "learning_rate": 1.599223300970874e-05,
      "loss": 0.0913,
      "step": 77400
    },
    {
      "epoch": 30.06,
      "learning_rate": 1.599171521035599e-05,
      "loss": 0.0115,
      "step": 77410
    },
    {
      "epoch": 30.07,
      "learning_rate": 1.5991197411003238e-05,
      "loss": 0.1244,
      "step": 77420
    },
    {
      "epoch": 30.07,
      "learning_rate": 1.5990679611650486e-05,
      "loss": 0.1016,
      "step": 77430
    },
    {
      "epoch": 30.07,
      "learning_rate": 1.5990161812297737e-05,
      "loss": 0.1364,
      "step": 77440
    },
    {
      "epoch": 30.08,
      "learning_rate": 1.5989644012944985e-05,
      "loss": 0.0438,
      "step": 77450
    },
    {
      "epoch": 30.08,
      "learning_rate": 1.5989126213592233e-05,
      "loss": 0.2466,
      "step": 77460
    },
    {
      "epoch": 30.09,
      "learning_rate": 1.5988608414239484e-05,
      "loss": 0.2113,
      "step": 77470
    },
    {
      "epoch": 30.09,
      "learning_rate": 1.5988090614886732e-05,
      "loss": 0.0988,
      "step": 77480
    },
    {
      "epoch": 30.09,
      "learning_rate": 1.5987572815533983e-05,
      "loss": 0.1021,
      "step": 77490
    },
    {
      "epoch": 30.1,
      "learning_rate": 1.598705501618123e-05,
      "loss": 0.0423,
      "step": 77500
    },
    {
      "epoch": 30.1,
      "learning_rate": 1.598653721682848e-05,
      "loss": 0.0808,
      "step": 77510
    },
    {
      "epoch": 30.1,
      "learning_rate": 1.598601941747573e-05,
      "loss": 0.0388,
      "step": 77520
    },
    {
      "epoch": 30.11,
      "learning_rate": 1.598550161812298e-05,
      "loss": 0.0625,
      "step": 77530
    },
    {
      "epoch": 30.11,
      "learning_rate": 1.5984983818770226e-05,
      "loss": 0.0043,
      "step": 77540
    },
    {
      "epoch": 30.12,
      "learning_rate": 1.5984466019417478e-05,
      "loss": 0.0981,
      "step": 77550
    },
    {
      "epoch": 30.12,
      "learning_rate": 1.5983948220064726e-05,
      "loss": 0.0692,
      "step": 77560
    },
    {
      "epoch": 30.12,
      "learning_rate": 1.5983430420711977e-05,
      "loss": 0.0928,
      "step": 77570
    },
    {
      "epoch": 30.13,
      "learning_rate": 1.5982912621359225e-05,
      "loss": 0.0506,
      "step": 77580
    },
    {
      "epoch": 30.13,
      "learning_rate": 1.5982394822006473e-05,
      "loss": 0.2043,
      "step": 77590
    },
    {
      "epoch": 30.14,
      "learning_rate": 1.598187702265372e-05,
      "loss": 0.0303,
      "step": 77600
    },
    {
      "epoch": 30.14,
      "learning_rate": 1.5981359223300972e-05,
      "loss": 0.0751,
      "step": 77610
    },
    {
      "epoch": 30.14,
      "learning_rate": 1.598084142394822e-05,
      "loss": 0.0259,
      "step": 77620
    },
    {
      "epoch": 30.15,
      "learning_rate": 1.598032362459547e-05,
      "loss": 0.0523,
      "step": 77630
    },
    {
      "epoch": 30.15,
      "learning_rate": 1.597980582524272e-05,
      "loss": 0.2317,
      "step": 77640
    },
    {
      "epoch": 30.16,
      "learning_rate": 1.597928802588997e-05,
      "loss": 0.0174,
      "step": 77650
    },
    {
      "epoch": 30.16,
      "learning_rate": 1.597877022653722e-05,
      "loss": 0.1064,
      "step": 77660
    },
    {
      "epoch": 30.16,
      "learning_rate": 1.5978252427184466e-05,
      "loss": 0.067,
      "step": 77670
    },
    {
      "epoch": 30.17,
      "learning_rate": 1.5977734627831714e-05,
      "loss": 0.0523,
      "step": 77680
    },
    {
      "epoch": 30.17,
      "learning_rate": 1.5977216828478966e-05,
      "loss": 0.1837,
      "step": 77690
    },
    {
      "epoch": 30.17,
      "learning_rate": 1.5976699029126213e-05,
      "loss": 0.0691,
      "step": 77700
    },
    {
      "epoch": 30.18,
      "learning_rate": 1.5976181229773465e-05,
      "loss": 0.2029,
      "step": 77710
    },
    {
      "epoch": 30.18,
      "learning_rate": 1.5975663430420713e-05,
      "loss": 0.0012,
      "step": 77720
    },
    {
      "epoch": 30.19,
      "learning_rate": 1.5975145631067964e-05,
      "loss": 0.1693,
      "step": 77730
    },
    {
      "epoch": 30.19,
      "learning_rate": 1.5974627831715212e-05,
      "loss": 0.0533,
      "step": 77740
    },
    {
      "epoch": 30.19,
      "learning_rate": 1.597411003236246e-05,
      "loss": 0.0548,
      "step": 77750
    },
    {
      "epoch": 30.2,
      "learning_rate": 1.5973592233009708e-05,
      "loss": 0.011,
      "step": 77760
    },
    {
      "epoch": 30.2,
      "learning_rate": 1.597307443365696e-05,
      "loss": 0.0256,
      "step": 77770
    },
    {
      "epoch": 30.21,
      "learning_rate": 1.5972556634304207e-05,
      "loss": 0.1355,
      "step": 77780
    },
    {
      "epoch": 30.21,
      "learning_rate": 1.597203883495146e-05,
      "loss": 0.09,
      "step": 77790
    },
    {
      "epoch": 30.21,
      "learning_rate": 1.5971521035598706e-05,
      "loss": 0.1083,
      "step": 77800
    },
    {
      "epoch": 30.22,
      "learning_rate": 1.5971003236245958e-05,
      "loss": 0.1963,
      "step": 77810
    },
    {
      "epoch": 30.22,
      "learning_rate": 1.5970485436893205e-05,
      "loss": 0.0664,
      "step": 77820
    },
    {
      "epoch": 30.23,
      "learning_rate": 1.5969967637540453e-05,
      "loss": 0.0594,
      "step": 77830
    },
    {
      "epoch": 30.23,
      "learning_rate": 1.59694498381877e-05,
      "loss": 0.0344,
      "step": 77840
    },
    {
      "epoch": 30.23,
      "learning_rate": 1.5968932038834953e-05,
      "loss": 0.0562,
      "step": 77850
    },
    {
      "epoch": 30.24,
      "learning_rate": 1.59684142394822e-05,
      "loss": 0.0652,
      "step": 77860
    },
    {
      "epoch": 30.24,
      "learning_rate": 1.5967896440129452e-05,
      "loss": 0.1704,
      "step": 77870
    },
    {
      "epoch": 30.24,
      "learning_rate": 1.59673786407767e-05,
      "loss": 0.1549,
      "step": 77880
    },
    {
      "epoch": 30.25,
      "learning_rate": 1.596686084142395e-05,
      "loss": 0.1344,
      "step": 77890
    },
    {
      "epoch": 30.25,
      "learning_rate": 1.59663430420712e-05,
      "loss": 0.0471,
      "step": 77900
    },
    {
      "epoch": 30.26,
      "learning_rate": 1.5965825242718447e-05,
      "loss": 0.0018,
      "step": 77910
    },
    {
      "epoch": 30.26,
      "learning_rate": 1.5965307443365695e-05,
      "loss": 0.09,
      "step": 77920
    },
    {
      "epoch": 30.26,
      "learning_rate": 1.5964789644012946e-05,
      "loss": 0.1211,
      "step": 77930
    },
    {
      "epoch": 30.27,
      "learning_rate": 1.5964271844660194e-05,
      "loss": 0.0308,
      "step": 77940
    },
    {
      "epoch": 30.27,
      "learning_rate": 1.5963754045307445e-05,
      "loss": 0.1577,
      "step": 77950
    },
    {
      "epoch": 30.28,
      "learning_rate": 1.5963236245954693e-05,
      "loss": 0.1641,
      "step": 77960
    },
    {
      "epoch": 30.28,
      "learning_rate": 1.5962718446601945e-05,
      "loss": 0.0587,
      "step": 77970
    },
    {
      "epoch": 30.28,
      "learning_rate": 1.5962200647249193e-05,
      "loss": 0.1388,
      "step": 77980
    },
    {
      "epoch": 30.29,
      "learning_rate": 1.5961682847896444e-05,
      "loss": 0.0381,
      "step": 77990
    },
    {
      "epoch": 30.29,
      "learning_rate": 1.596116504854369e-05,
      "loss": 0.1661,
      "step": 78000
    },
    {
      "epoch": 30.3,
      "learning_rate": 1.596064724919094e-05,
      "loss": 0.1552,
      "step": 78010
    },
    {
      "epoch": 30.3,
      "learning_rate": 1.5960129449838188e-05,
      "loss": 0.1135,
      "step": 78020
    },
    {
      "epoch": 30.3,
      "learning_rate": 1.595961165048544e-05,
      "loss": 0.2185,
      "step": 78030
    },
    {
      "epoch": 30.31,
      "learning_rate": 1.5959093851132687e-05,
      "loss": 0.0663,
      "step": 78040
    },
    {
      "epoch": 30.31,
      "learning_rate": 1.5958576051779938e-05,
      "loss": 0.2059,
      "step": 78050
    },
    {
      "epoch": 30.31,
      "learning_rate": 1.5958058252427186e-05,
      "loss": 0.1361,
      "step": 78060
    },
    {
      "epoch": 30.32,
      "learning_rate": 1.5957540453074437e-05,
      "loss": 0.005,
      "step": 78070
    },
    {
      "epoch": 30.32,
      "learning_rate": 1.5957022653721682e-05,
      "loss": 0.0644,
      "step": 78080
    },
    {
      "epoch": 30.33,
      "learning_rate": 1.5956504854368933e-05,
      "loss": 0.1107,
      "step": 78090
    },
    {
      "epoch": 30.33,
      "learning_rate": 1.595598705501618e-05,
      "loss": 0.0514,
      "step": 78100
    },
    {
      "epoch": 30.33,
      "learning_rate": 1.5955469255663433e-05,
      "loss": 0.0695,
      "step": 78110
    },
    {
      "epoch": 30.34,
      "learning_rate": 1.595495145631068e-05,
      "loss": 0.0508,
      "step": 78120
    },
    {
      "epoch": 30.34,
      "learning_rate": 1.5954433656957932e-05,
      "loss": 0.0908,
      "step": 78130
    },
    {
      "epoch": 30.35,
      "learning_rate": 1.595391585760518e-05,
      "loss": 0.2458,
      "step": 78140
    },
    {
      "epoch": 30.35,
      "learning_rate": 1.595339805825243e-05,
      "loss": 0.0994,
      "step": 78150
    },
    {
      "epoch": 30.35,
      "learning_rate": 1.5952880258899676e-05,
      "loss": 0.105,
      "step": 78160
    },
    {
      "epoch": 30.36,
      "learning_rate": 1.5952362459546927e-05,
      "loss": 0.1984,
      "step": 78170
    },
    {
      "epoch": 30.36,
      "learning_rate": 1.5951844660194175e-05,
      "loss": 0.073,
      "step": 78180
    },
    {
      "epoch": 30.37,
      "learning_rate": 1.5951326860841426e-05,
      "loss": 0.0595,
      "step": 78190
    },
    {
      "epoch": 30.37,
      "learning_rate": 1.5950809061488674e-05,
      "loss": 0.0456,
      "step": 78200
    },
    {
      "epoch": 30.37,
      "learning_rate": 1.5950291262135925e-05,
      "loss": 0.0888,
      "step": 78210
    },
    {
      "epoch": 30.38,
      "learning_rate": 1.5949773462783173e-05,
      "loss": 0.1325,
      "step": 78220
    },
    {
      "epoch": 30.38,
      "learning_rate": 1.5949255663430425e-05,
      "loss": 0.2159,
      "step": 78230
    },
    {
      "epoch": 30.38,
      "learning_rate": 1.594873786407767e-05,
      "loss": 0.0632,
      "step": 78240
    },
    {
      "epoch": 30.39,
      "learning_rate": 1.594822006472492e-05,
      "loss": 0.1127,
      "step": 78250
    },
    {
      "epoch": 30.39,
      "learning_rate": 1.594770226537217e-05,
      "loss": 0.103,
      "step": 78260
    },
    {
      "epoch": 30.4,
      "learning_rate": 1.594718446601942e-05,
      "loss": 0.0064,
      "step": 78270
    },
    {
      "epoch": 30.4,
      "learning_rate": 1.5946666666666668e-05,
      "loss": 0.2122,
      "step": 78280
    },
    {
      "epoch": 30.4,
      "learning_rate": 1.594614886731392e-05,
      "loss": 0.0691,
      "step": 78290
    },
    {
      "epoch": 30.41,
      "learning_rate": 1.5945631067961167e-05,
      "loss": 0.261,
      "step": 78300
    },
    {
      "epoch": 30.41,
      "learning_rate": 1.5945113268608418e-05,
      "loss": 0.0213,
      "step": 78310
    },
    {
      "epoch": 30.42,
      "learning_rate": 1.5944595469255663e-05,
      "loss": 0.2535,
      "step": 78320
    },
    {
      "epoch": 30.42,
      "learning_rate": 1.5944077669902914e-05,
      "loss": 0.1691,
      "step": 78330
    },
    {
      "epoch": 30.42,
      "learning_rate": 1.5943559870550162e-05,
      "loss": 0.0735,
      "step": 78340
    },
    {
      "epoch": 30.43,
      "learning_rate": 1.5943042071197413e-05,
      "loss": 0.0706,
      "step": 78350
    },
    {
      "epoch": 30.43,
      "learning_rate": 1.594252427184466e-05,
      "loss": 0.0903,
      "step": 78360
    },
    {
      "epoch": 30.43,
      "learning_rate": 1.5942006472491912e-05,
      "loss": 0.0716,
      "step": 78370
    },
    {
      "epoch": 30.44,
      "learning_rate": 1.594148867313916e-05,
      "loss": 0.1718,
      "step": 78380
    },
    {
      "epoch": 30.44,
      "learning_rate": 1.5940970873786408e-05,
      "loss": 0.0486,
      "step": 78390
    },
    {
      "epoch": 30.45,
      "learning_rate": 1.5940453074433656e-05,
      "loss": 0.058,
      "step": 78400
    },
    {
      "epoch": 30.45,
      "learning_rate": 1.5939935275080908e-05,
      "loss": 0.1012,
      "step": 78410
    },
    {
      "epoch": 30.45,
      "learning_rate": 1.5939417475728155e-05,
      "loss": 0.0493,
      "step": 78420
    },
    {
      "epoch": 30.46,
      "learning_rate": 1.5938899676375407e-05,
      "loss": 0.09,
      "step": 78430
    },
    {
      "epoch": 30.46,
      "learning_rate": 1.5938381877022655e-05,
      "loss": 0.0732,
      "step": 78440
    },
    {
      "epoch": 30.47,
      "learning_rate": 1.5937864077669906e-05,
      "loss": 0.1838,
      "step": 78450
    },
    {
      "epoch": 30.47,
      "learning_rate": 1.5937346278317154e-05,
      "loss": 0.3345,
      "step": 78460
    },
    {
      "epoch": 30.47,
      "learning_rate": 1.5936828478964402e-05,
      "loss": 0.0987,
      "step": 78470
    },
    {
      "epoch": 30.48,
      "learning_rate": 1.593631067961165e-05,
      "loss": 0.107,
      "step": 78480
    },
    {
      "epoch": 30.48,
      "learning_rate": 1.59357928802589e-05,
      "loss": 0.1197,
      "step": 78490
    },
    {
      "epoch": 30.49,
      "learning_rate": 1.593527508090615e-05,
      "loss": 0.0753,
      "step": 78500
    },
    {
      "epoch": 30.49,
      "learning_rate": 1.59347572815534e-05,
      "loss": 0.0711,
      "step": 78510
    },
    {
      "epoch": 30.49,
      "learning_rate": 1.5934239482200648e-05,
      "loss": 0.1222,
      "step": 78520
    },
    {
      "epoch": 30.5,
      "learning_rate": 1.59337216828479e-05,
      "loss": 0.2515,
      "step": 78530
    },
    {
      "epoch": 30.5,
      "learning_rate": 1.5933203883495147e-05,
      "loss": 0.1075,
      "step": 78540
    },
    {
      "epoch": 30.5,
      "learning_rate": 1.5932686084142395e-05,
      "loss": 0.1315,
      "step": 78550
    },
    {
      "epoch": 30.51,
      "learning_rate": 1.5932168284789647e-05,
      "loss": 0.035,
      "step": 78560
    },
    {
      "epoch": 30.51,
      "learning_rate": 1.5931650485436895e-05,
      "loss": 0.0501,
      "step": 78570
    },
    {
      "epoch": 30.52,
      "learning_rate": 1.5931132686084143e-05,
      "loss": 0.1169,
      "step": 78580
    },
    {
      "epoch": 30.52,
      "learning_rate": 1.5930614886731394e-05,
      "loss": 0.1462,
      "step": 78590
    },
    {
      "epoch": 30.52,
      "learning_rate": 1.5930097087378642e-05,
      "loss": 0.0369,
      "step": 78600
    },
    {
      "epoch": 30.53,
      "learning_rate": 1.5929579288025893e-05,
      "loss": 0.1698,
      "step": 78610
    },
    {
      "epoch": 30.53,
      "learning_rate": 1.592906148867314e-05,
      "loss": 0.0301,
      "step": 78620
    },
    {
      "epoch": 30.54,
      "learning_rate": 1.592854368932039e-05,
      "loss": 0.1059,
      "step": 78630
    },
    {
      "epoch": 30.54,
      "learning_rate": 1.592802588996764e-05,
      "loss": 0.0142,
      "step": 78640
    },
    {
      "epoch": 30.54,
      "learning_rate": 1.5927508090614888e-05,
      "loss": 0.1174,
      "step": 78650
    },
    {
      "epoch": 30.55,
      "learning_rate": 1.5926990291262136e-05,
      "loss": 0.1363,
      "step": 78660
    },
    {
      "epoch": 30.55,
      "learning_rate": 1.5926472491909387e-05,
      "loss": 0.1413,
      "step": 78670
    },
    {
      "epoch": 30.56,
      "learning_rate": 1.5925954692556635e-05,
      "loss": 0.1624,
      "step": 78680
    },
    {
      "epoch": 30.56,
      "learning_rate": 1.5925436893203883e-05,
      "loss": 0.0621,
      "step": 78690
    },
    {
      "epoch": 30.56,
      "learning_rate": 1.5924919093851135e-05,
      "loss": 0.1311,
      "step": 78700
    },
    {
      "epoch": 30.57,
      "learning_rate": 1.5924401294498383e-05,
      "loss": 0.0233,
      "step": 78710
    },
    {
      "epoch": 30.57,
      "learning_rate": 1.5923883495145634e-05,
      "loss": 0.1499,
      "step": 78720
    },
    {
      "epoch": 30.57,
      "learning_rate": 1.5923365695792882e-05,
      "loss": 0.1287,
      "step": 78730
    },
    {
      "epoch": 30.58,
      "learning_rate": 1.592284789644013e-05,
      "loss": 0.1573,
      "step": 78740
    },
    {
      "epoch": 30.58,
      "learning_rate": 1.592233009708738e-05,
      "loss": 0.058,
      "step": 78750
    },
    {
      "epoch": 30.59,
      "learning_rate": 1.592181229773463e-05,
      "loss": 0.1228,
      "step": 78760
    },
    {
      "epoch": 30.59,
      "learning_rate": 1.5921294498381877e-05,
      "loss": 0.1803,
      "step": 78770
    },
    {
      "epoch": 30.59,
      "learning_rate": 1.5920776699029128e-05,
      "loss": 0.2255,
      "step": 78780
    },
    {
      "epoch": 30.6,
      "learning_rate": 1.5920258899676376e-05,
      "loss": 0.1454,
      "step": 78790
    },
    {
      "epoch": 30.6,
      "learning_rate": 1.5919741100323627e-05,
      "loss": 0.144,
      "step": 78800
    },
    {
      "epoch": 30.61,
      "learning_rate": 1.5919223300970875e-05,
      "loss": 0.0489,
      "step": 78810
    },
    {
      "epoch": 30.61,
      "learning_rate": 1.5918705501618123e-05,
      "loss": 0.0261,
      "step": 78820
    },
    {
      "epoch": 30.61,
      "learning_rate": 1.5918187702265375e-05,
      "loss": 0.0414,
      "step": 78830
    },
    {
      "epoch": 30.62,
      "learning_rate": 1.5917669902912622e-05,
      "loss": 0.1503,
      "step": 78840
    },
    {
      "epoch": 30.62,
      "learning_rate": 1.591715210355987e-05,
      "loss": 0.0788,
      "step": 78850
    },
    {
      "epoch": 30.63,
      "learning_rate": 1.5916634304207122e-05,
      "loss": 0.2331,
      "step": 78860
    },
    {
      "epoch": 30.63,
      "learning_rate": 1.591611650485437e-05,
      "loss": 0.111,
      "step": 78870
    },
    {
      "epoch": 30.63,
      "learning_rate": 1.591559870550162e-05,
      "loss": 0.0049,
      "step": 78880
    },
    {
      "epoch": 30.64,
      "learning_rate": 1.591508090614887e-05,
      "loss": 0.0429,
      "step": 78890
    },
    {
      "epoch": 30.64,
      "learning_rate": 1.5914563106796117e-05,
      "loss": 0.0472,
      "step": 78900
    },
    {
      "epoch": 30.64,
      "learning_rate": 1.5914045307443368e-05,
      "loss": 0.0006,
      "step": 78910
    },
    {
      "epoch": 30.65,
      "learning_rate": 1.5913527508090616e-05,
      "loss": 0.0679,
      "step": 78920
    },
    {
      "epoch": 30.65,
      "learning_rate": 1.5913009708737864e-05,
      "loss": 0.0659,
      "step": 78930
    },
    {
      "epoch": 30.66,
      "learning_rate": 1.5912491909385115e-05,
      "loss": 0.0456,
      "step": 78940
    },
    {
      "epoch": 30.66,
      "learning_rate": 1.5911974110032363e-05,
      "loss": 0.1605,
      "step": 78950
    },
    {
      "epoch": 30.66,
      "learning_rate": 1.5911456310679615e-05,
      "loss": 0.061,
      "step": 78960
    },
    {
      "epoch": 30.67,
      "learning_rate": 1.5910938511326862e-05,
      "loss": 0.0461,
      "step": 78970
    },
    {
      "epoch": 30.67,
      "learning_rate": 1.591042071197411e-05,
      "loss": 0.1427,
      "step": 78980
    },
    {
      "epoch": 30.68,
      "learning_rate": 1.590990291262136e-05,
      "loss": 0.1507,
      "step": 78990
    },
    {
      "epoch": 30.68,
      "learning_rate": 1.590938511326861e-05,
      "loss": 0.1419,
      "step": 79000
    },
    {
      "epoch": 30.68,
      "learning_rate": 1.5908867313915858e-05,
      "loss": 0.0868,
      "step": 79010
    },
    {
      "epoch": 30.69,
      "learning_rate": 1.590834951456311e-05,
      "loss": 0.0648,
      "step": 79020
    },
    {
      "epoch": 30.69,
      "learning_rate": 1.5907831715210357e-05,
      "loss": 0.0555,
      "step": 79030
    },
    {
      "epoch": 30.7,
      "learning_rate": 1.5907313915857608e-05,
      "loss": 0.0425,
      "step": 79040
    },
    {
      "epoch": 30.7,
      "learning_rate": 1.5906796116504856e-05,
      "loss": 0.1427,
      "step": 79050
    },
    {
      "epoch": 30.7,
      "learning_rate": 1.5906278317152104e-05,
      "loss": 0.1153,
      "step": 79060
    },
    {
      "epoch": 30.71,
      "learning_rate": 1.5905760517799352e-05,
      "loss": 0.078,
      "step": 79070
    },
    {
      "epoch": 30.71,
      "learning_rate": 1.5905242718446603e-05,
      "loss": 0.1426,
      "step": 79080
    },
    {
      "epoch": 30.71,
      "learning_rate": 1.590472491909385e-05,
      "loss": 0.0023,
      "step": 79090
    },
    {
      "epoch": 30.72,
      "learning_rate": 1.5904207119741102e-05,
      "loss": 0.012,
      "step": 79100
    },
    {
      "epoch": 30.72,
      "learning_rate": 1.590368932038835e-05,
      "loss": 0.027,
      "step": 79110
    },
    {
      "epoch": 30.73,
      "learning_rate": 1.59031715210356e-05,
      "loss": 0.1149,
      "step": 79120
    },
    {
      "epoch": 30.73,
      "learning_rate": 1.590265372168285e-05,
      "loss": 0.0271,
      "step": 79130
    },
    {
      "epoch": 30.73,
      "learning_rate": 1.5902135922330097e-05,
      "loss": 0.0971,
      "step": 79140
    },
    {
      "epoch": 30.74,
      "learning_rate": 1.5901618122977345e-05,
      "loss": 0.1479,
      "step": 79150
    },
    {
      "epoch": 30.74,
      "learning_rate": 1.5901100323624597e-05,
      "loss": 0.0799,
      "step": 79160
    },
    {
      "epoch": 30.75,
      "learning_rate": 1.5900582524271845e-05,
      "loss": 0.0331,
      "step": 79170
    },
    {
      "epoch": 30.75,
      "learning_rate": 1.5900064724919096e-05,
      "loss": 0.096,
      "step": 79180
    },
    {
      "epoch": 30.75,
      "learning_rate": 1.5899546925566344e-05,
      "loss": 0.0201,
      "step": 79190
    },
    {
      "epoch": 30.76,
      "learning_rate": 1.5899029126213595e-05,
      "loss": 0.1905,
      "step": 79200
    },
    {
      "epoch": 30.76,
      "learning_rate": 1.5898511326860843e-05,
      "loss": 0.2479,
      "step": 79210
    },
    {
      "epoch": 30.77,
      "learning_rate": 1.589799352750809e-05,
      "loss": 0.0261,
      "step": 79220
    },
    {
      "epoch": 30.77,
      "learning_rate": 1.589747572815534e-05,
      "loss": 0.2463,
      "step": 79230
    },
    {
      "epoch": 30.77,
      "learning_rate": 1.589695792880259e-05,
      "loss": 0.042,
      "step": 79240
    },
    {
      "epoch": 30.78,
      "learning_rate": 1.5896440129449838e-05,
      "loss": 0.0816,
      "step": 79250
    },
    {
      "epoch": 30.78,
      "learning_rate": 1.589592233009709e-05,
      "loss": 0.1112,
      "step": 79260
    },
    {
      "epoch": 30.78,
      "learning_rate": 1.5895404530744337e-05,
      "loss": 0.1145,
      "step": 79270
    },
    {
      "epoch": 30.79,
      "learning_rate": 1.589488673139159e-05,
      "loss": 0.1301,
      "step": 79280
    },
    {
      "epoch": 30.79,
      "learning_rate": 1.5894368932038837e-05,
      "loss": 0.2107,
      "step": 79290
    },
    {
      "epoch": 30.8,
      "learning_rate": 1.5893851132686085e-05,
      "loss": 0.0291,
      "step": 79300
    },
    {
      "epoch": 30.8,
      "learning_rate": 1.5893333333333333e-05,
      "loss": 0.0059,
      "step": 79310
    },
    {
      "epoch": 30.8,
      "learning_rate": 1.5892815533980584e-05,
      "loss": 0.0876,
      "step": 79320
    },
    {
      "epoch": 30.81,
      "learning_rate": 1.5892297734627832e-05,
      "loss": 0.1514,
      "step": 79330
    },
    {
      "epoch": 30.81,
      "learning_rate": 1.5891779935275083e-05,
      "loss": 0.0277,
      "step": 79340
    },
    {
      "epoch": 30.82,
      "learning_rate": 1.589126213592233e-05,
      "loss": 0.0741,
      "step": 79350
    },
    {
      "epoch": 30.82,
      "learning_rate": 1.5890744336569582e-05,
      "loss": 0.1573,
      "step": 79360
    },
    {
      "epoch": 30.82,
      "learning_rate": 1.589022653721683e-05,
      "loss": 0.0784,
      "step": 79370
    },
    {
      "epoch": 30.83,
      "learning_rate": 1.5889708737864078e-05,
      "loss": 0.0595,
      "step": 79380
    },
    {
      "epoch": 30.83,
      "learning_rate": 1.5889190938511326e-05,
      "loss": 0.0776,
      "step": 79390
    },
    {
      "epoch": 30.83,
      "learning_rate": 1.5888673139158577e-05,
      "loss": 0.1074,
      "step": 79400
    },
    {
      "epoch": 30.84,
      "learning_rate": 1.5888155339805825e-05,
      "loss": 0.1656,
      "step": 79410
    },
    {
      "epoch": 30.84,
      "learning_rate": 1.5887637540453077e-05,
      "loss": 0.2042,
      "step": 79420
    },
    {
      "epoch": 30.85,
      "learning_rate": 1.5887119741100325e-05,
      "loss": 0.1905,
      "step": 79430
    },
    {
      "epoch": 30.85,
      "learning_rate": 1.5886601941747576e-05,
      "loss": 0.068,
      "step": 79440
    },
    {
      "epoch": 30.85,
      "learning_rate": 1.5886084142394824e-05,
      "loss": 0.032,
      "step": 79450
    },
    {
      "epoch": 30.86,
      "learning_rate": 1.588556634304207e-05,
      "loss": 0.0262,
      "step": 79460
    },
    {
      "epoch": 30.86,
      "learning_rate": 1.588504854368932e-05,
      "loss": 0.031,
      "step": 79470
    },
    {
      "epoch": 30.87,
      "learning_rate": 1.588453074433657e-05,
      "loss": 0.0755,
      "step": 79480
    },
    {
      "epoch": 30.87,
      "learning_rate": 1.588401294498382e-05,
      "loss": 0.067,
      "step": 79490
    },
    {
      "epoch": 30.87,
      "learning_rate": 1.588349514563107e-05,
      "loss": 0.0969,
      "step": 79500
    },
    {
      "epoch": 30.88,
      "learning_rate": 1.5882977346278318e-05,
      "loss": 0.0277,
      "step": 79510
    },
    {
      "epoch": 30.88,
      "learning_rate": 1.588245954692557e-05,
      "loss": 0.0337,
      "step": 79520
    },
    {
      "epoch": 30.89,
      "learning_rate": 1.5881941747572817e-05,
      "loss": 0.095,
      "step": 79530
    },
    {
      "epoch": 30.89,
      "learning_rate": 1.5881423948220065e-05,
      "loss": 0.0844,
      "step": 79540
    },
    {
      "epoch": 30.89,
      "learning_rate": 1.5880906148867313e-05,
      "loss": 0.1799,
      "step": 79550
    },
    {
      "epoch": 30.9,
      "learning_rate": 1.5880388349514564e-05,
      "loss": 0.1793,
      "step": 79560
    },
    {
      "epoch": 30.9,
      "learning_rate": 1.5879870550161812e-05,
      "loss": 0.1063,
      "step": 79570
    },
    {
      "epoch": 30.9,
      "learning_rate": 1.5879352750809064e-05,
      "loss": 0.0079,
      "step": 79580
    },
    {
      "epoch": 30.91,
      "learning_rate": 1.587883495145631e-05,
      "loss": 0.1524,
      "step": 79590
    },
    {
      "epoch": 30.91,
      "learning_rate": 1.5878317152103563e-05,
      "loss": 0.0445,
      "step": 79600
    },
    {
      "epoch": 30.92,
      "learning_rate": 1.587779935275081e-05,
      "loss": 0.0392,
      "step": 79610
    },
    {
      "epoch": 30.92,
      "learning_rate": 1.5877281553398062e-05,
      "loss": 0.0952,
      "step": 79620
    },
    {
      "epoch": 30.92,
      "learning_rate": 1.5876763754045307e-05,
      "loss": 0.235,
      "step": 79630
    },
    {
      "epoch": 30.93,
      "learning_rate": 1.5876245954692558e-05,
      "loss": 0.1323,
      "step": 79640
    },
    {
      "epoch": 30.93,
      "learning_rate": 1.5875728155339806e-05,
      "loss": 0.0152,
      "step": 79650
    },
    {
      "epoch": 30.94,
      "learning_rate": 1.5875210355987057e-05,
      "loss": 0.0946,
      "step": 79660
    },
    {
      "epoch": 30.94,
      "learning_rate": 1.5874692556634305e-05,
      "loss": 0.0599,
      "step": 79670
    },
    {
      "epoch": 30.94,
      "learning_rate": 1.5874174757281557e-05,
      "loss": 0.0856,
      "step": 79680
    },
    {
      "epoch": 30.95,
      "learning_rate": 1.5873656957928804e-05,
      "loss": 0.0525,
      "step": 79690
    },
    {
      "epoch": 30.95,
      "learning_rate": 1.5873139158576056e-05,
      "loss": 0.0371,
      "step": 79700
    },
    {
      "epoch": 30.96,
      "learning_rate": 1.58726213592233e-05,
      "loss": 0.0302,
      "step": 79710
    },
    {
      "epoch": 30.96,
      "learning_rate": 1.587210355987055e-05,
      "loss": 0.1451,
      "step": 79720
    },
    {
      "epoch": 30.96,
      "learning_rate": 1.58715857605178e-05,
      "loss": 0.038,
      "step": 79730
    },
    {
      "epoch": 30.97,
      "learning_rate": 1.587106796116505e-05,
      "loss": 0.0821,
      "step": 79740
    },
    {
      "epoch": 30.97,
      "learning_rate": 1.58705501618123e-05,
      "loss": 0.0432,
      "step": 79750
    },
    {
      "epoch": 30.97,
      "learning_rate": 1.587003236245955e-05,
      "loss": 0.0279,
      "step": 79760
    },
    {
      "epoch": 30.98,
      "learning_rate": 1.5869514563106798e-05,
      "loss": 0.0902,
      "step": 79770
    },
    {
      "epoch": 30.98,
      "learning_rate": 1.586899676375405e-05,
      "loss": 0.1225,
      "step": 79780
    },
    {
      "epoch": 30.99,
      "learning_rate": 1.5868478964401294e-05,
      "loss": 0.1027,
      "step": 79790
    },
    {
      "epoch": 30.99,
      "learning_rate": 1.5867961165048545e-05,
      "loss": 0.0956,
      "step": 79800
    },
    {
      "epoch": 30.99,
      "learning_rate": 1.5867443365695793e-05,
      "loss": 0.1283,
      "step": 79810
    },
    {
      "epoch": 31.0,
      "learning_rate": 1.5866925566343044e-05,
      "loss": 0.0652,
      "step": 79820
    },
    {
      "epoch": 31.0,
      "eval_accuracy": 0.9537826685006877,
      "eval_loss": 0.2627561092376709,
      "eval_runtime": 8.1542,
      "eval_samples_per_second": 445.781,
      "eval_steps_per_second": 55.799,
      "step": 79825
    },
    {
      "epoch": 31.0,
      "learning_rate": 1.5866407766990292e-05,
      "loss": 0.1509,
      "step": 79830
    },
    {
      "epoch": 31.01,
      "learning_rate": 1.5865889967637544e-05,
      "loss": 0.0994,
      "step": 79840
    },
    {
      "epoch": 31.01,
      "learning_rate": 1.586537216828479e-05,
      "loss": 0.2286,
      "step": 79850
    },
    {
      "epoch": 31.01,
      "learning_rate": 1.586485436893204e-05,
      "loss": 0.0751,
      "step": 79860
    },
    {
      "epoch": 31.02,
      "learning_rate": 1.5864336569579287e-05,
      "loss": 0.0423,
      "step": 79870
    },
    {
      "epoch": 31.02,
      "learning_rate": 1.586381877022654e-05,
      "loss": 0.1178,
      "step": 79880
    },
    {
      "epoch": 31.03,
      "learning_rate": 1.5863300970873787e-05,
      "loss": 0.303,
      "step": 79890
    },
    {
      "epoch": 31.03,
      "learning_rate": 1.5862783171521038e-05,
      "loss": 0.0216,
      "step": 79900
    },
    {
      "epoch": 31.03,
      "learning_rate": 1.5862265372168286e-05,
      "loss": 0.0485,
      "step": 79910
    },
    {
      "epoch": 31.04,
      "learning_rate": 1.5861747572815537e-05,
      "loss": 0.0815,
      "step": 79920
    },
    {
      "epoch": 31.04,
      "learning_rate": 1.5861229773462785e-05,
      "loss": 0.1615,
      "step": 79930
    },
    {
      "epoch": 31.04,
      "learning_rate": 1.5860711974110033e-05,
      "loss": 0.2809,
      "step": 79940
    },
    {
      "epoch": 31.05,
      "learning_rate": 1.586019417475728e-05,
      "loss": 0.1023,
      "step": 79950
    },
    {
      "epoch": 31.05,
      "learning_rate": 1.5859676375404532e-05,
      "loss": 0.1653,
      "step": 79960
    },
    {
      "epoch": 31.06,
      "learning_rate": 1.585915857605178e-05,
      "loss": 0.0232,
      "step": 79970
    },
    {
      "epoch": 31.06,
      "learning_rate": 1.585864077669903e-05,
      "loss": 0.0751,
      "step": 79980
    },
    {
      "epoch": 31.06,
      "learning_rate": 1.585812297734628e-05,
      "loss": 0.185,
      "step": 79990
    },
    {
      "epoch": 31.07,
      "learning_rate": 1.585760517799353e-05,
      "loss": 0.0655,
      "step": 80000
    },
    {
      "epoch": 31.07,
      "learning_rate": 1.585708737864078e-05,
      "loss": 0.1214,
      "step": 80010
    },
    {
      "epoch": 31.08,
      "learning_rate": 1.5856569579288027e-05,
      "loss": 0.167,
      "step": 80020
    },
    {
      "epoch": 31.08,
      "learning_rate": 1.5856051779935275e-05,
      "loss": 0.1068,
      "step": 80030
    },
    {
      "epoch": 31.08,
      "learning_rate": 1.5855533980582526e-05,
      "loss": 0.1276,
      "step": 80040
    },
    {
      "epoch": 31.09,
      "learning_rate": 1.5855016181229774e-05,
      "loss": 0.1457,
      "step": 80050
    },
    {
      "epoch": 31.09,
      "learning_rate": 1.5854498381877025e-05,
      "loss": 0.1041,
      "step": 80060
    },
    {
      "epoch": 31.1,
      "learning_rate": 1.5853980582524273e-05,
      "loss": 0.0307,
      "step": 80070
    },
    {
      "epoch": 31.1,
      "learning_rate": 1.5853462783171524e-05,
      "loss": 0.0349,
      "step": 80080
    },
    {
      "epoch": 31.1,
      "learning_rate": 1.5852944983818772e-05,
      "loss": 0.2112,
      "step": 80090
    },
    {
      "epoch": 31.11,
      "learning_rate": 1.585242718446602e-05,
      "loss": 0.0432,
      "step": 80100
    },
    {
      "epoch": 31.11,
      "learning_rate": 1.5851909385113268e-05,
      "loss": 0.0686,
      "step": 80110
    },
    {
      "epoch": 31.11,
      "learning_rate": 1.585139158576052e-05,
      "loss": 0.0527,
      "step": 80120
    },
    {
      "epoch": 31.12,
      "learning_rate": 1.5850873786407767e-05,
      "loss": 0.0291,
      "step": 80130
    },
    {
      "epoch": 31.12,
      "learning_rate": 1.585035598705502e-05,
      "loss": 0.0454,
      "step": 80140
    },
    {
      "epoch": 31.13,
      "learning_rate": 1.5849838187702267e-05,
      "loss": 0.0013,
      "step": 80150
    },
    {
      "epoch": 31.13,
      "learning_rate": 1.5849320388349514e-05,
      "loss": 0.0863,
      "step": 80160
    },
    {
      "epoch": 31.13,
      "learning_rate": 1.5848802588996766e-05,
      "loss": 0.2131,
      "step": 80170
    },
    {
      "epoch": 31.14,
      "learning_rate": 1.5848284789644014e-05,
      "loss": 0.0463,
      "step": 80180
    },
    {
      "epoch": 31.14,
      "learning_rate": 1.5847766990291265e-05,
      "loss": 0.1306,
      "step": 80190
    },
    {
      "epoch": 31.15,
      "learning_rate": 1.5847249190938513e-05,
      "loss": 0.0969,
      "step": 80200
    },
    {
      "epoch": 31.15,
      "learning_rate": 1.584673139158576e-05,
      "loss": 0.1013,
      "step": 80210
    },
    {
      "epoch": 31.15,
      "learning_rate": 1.5846213592233012e-05,
      "loss": 0.1169,
      "step": 80220
    },
    {
      "epoch": 31.16,
      "learning_rate": 1.584569579288026e-05,
      "loss": 0.1242,
      "step": 80230
    },
    {
      "epoch": 31.16,
      "learning_rate": 1.5845177993527508e-05,
      "loss": 0.059,
      "step": 80240
    },
    {
      "epoch": 31.17,
      "learning_rate": 1.584466019417476e-05,
      "loss": 0.1266,
      "step": 80250
    },
    {
      "epoch": 31.17,
      "learning_rate": 1.5844142394822007e-05,
      "loss": 0.0526,
      "step": 80260
    },
    {
      "epoch": 31.17,
      "learning_rate": 1.584362459546926e-05,
      "loss": 0.1307,
      "step": 80270
    },
    {
      "epoch": 31.18,
      "learning_rate": 1.5843106796116506e-05,
      "loss": 0.0904,
      "step": 80280
    },
    {
      "epoch": 31.18,
      "learning_rate": 1.5842588996763754e-05,
      "loss": 0.1415,
      "step": 80290
    },
    {
      "epoch": 31.18,
      "learning_rate": 1.5842071197411006e-05,
      "loss": 0.0838,
      "step": 80300
    },
    {
      "epoch": 31.19,
      "learning_rate": 1.5841553398058254e-05,
      "loss": 0.0904,
      "step": 80310
    },
    {
      "epoch": 31.19,
      "learning_rate": 1.58410355987055e-05,
      "loss": 0.118,
      "step": 80320
    },
    {
      "epoch": 31.2,
      "learning_rate": 1.5840517799352753e-05,
      "loss": 0.0365,
      "step": 80330
    },
    {
      "epoch": 31.2,
      "learning_rate": 1.584e-05,
      "loss": 0.1472,
      "step": 80340
    },
    {
      "epoch": 31.2,
      "learning_rate": 1.5839482200647252e-05,
      "loss": 0.1572,
      "step": 80350
    },
    {
      "epoch": 31.21,
      "learning_rate": 1.58389644012945e-05,
      "loss": 0.0127,
      "step": 80360
    },
    {
      "epoch": 31.21,
      "learning_rate": 1.5838446601941748e-05,
      "loss": 0.1373,
      "step": 80370
    },
    {
      "epoch": 31.22,
      "learning_rate": 1.5837928802589e-05,
      "loss": 0.033,
      "step": 80380
    },
    {
      "epoch": 31.22,
      "learning_rate": 1.5837411003236247e-05,
      "loss": 0.1581,
      "step": 80390
    },
    {
      "epoch": 31.22,
      "learning_rate": 1.5836893203883495e-05,
      "loss": 0.135,
      "step": 80400
    },
    {
      "epoch": 31.23,
      "learning_rate": 1.5836375404530746e-05,
      "loss": 0.0762,
      "step": 80410
    },
    {
      "epoch": 31.23,
      "learning_rate": 1.5835857605177994e-05,
      "loss": 0.1461,
      "step": 80420
    },
    {
      "epoch": 31.23,
      "learning_rate": 1.5835339805825246e-05,
      "loss": 0.118,
      "step": 80430
    },
    {
      "epoch": 31.24,
      "learning_rate": 1.5834822006472494e-05,
      "loss": 0.1128,
      "step": 80440
    },
    {
      "epoch": 31.24,
      "learning_rate": 1.583430420711974e-05,
      "loss": 0.0599,
      "step": 80450
    },
    {
      "epoch": 31.25,
      "learning_rate": 1.5833786407766993e-05,
      "loss": 0.0014,
      "step": 80460
    },
    {
      "epoch": 31.25,
      "learning_rate": 1.583326860841424e-05,
      "loss": 0.2184,
      "step": 80470
    },
    {
      "epoch": 31.25,
      "learning_rate": 1.583275080906149e-05,
      "loss": 0.1088,
      "step": 80480
    },
    {
      "epoch": 31.26,
      "learning_rate": 1.583223300970874e-05,
      "loss": 0.0328,
      "step": 80490
    },
    {
      "epoch": 31.26,
      "learning_rate": 1.5831715210355988e-05,
      "loss": 0.0661,
      "step": 80500
    },
    {
      "epoch": 31.27,
      "learning_rate": 1.583119741100324e-05,
      "loss": 0.1483,
      "step": 80510
    },
    {
      "epoch": 31.27,
      "learning_rate": 1.5830679611650487e-05,
      "loss": 0.0874,
      "step": 80520
    },
    {
      "epoch": 31.27,
      "learning_rate": 1.5830161812297735e-05,
      "loss": 0.0642,
      "step": 80530
    },
    {
      "epoch": 31.28,
      "learning_rate": 1.5829644012944983e-05,
      "loss": 0.1388,
      "step": 80540
    },
    {
      "epoch": 31.28,
      "learning_rate": 1.5829126213592234e-05,
      "loss": 0.029,
      "step": 80550
    },
    {
      "epoch": 31.29,
      "learning_rate": 1.5828608414239482e-05,
      "loss": 0.0226,
      "step": 80560
    },
    {
      "epoch": 31.29,
      "learning_rate": 1.5828090614886734e-05,
      "loss": 0.0293,
      "step": 80570
    },
    {
      "epoch": 31.29,
      "learning_rate": 1.582757281553398e-05,
      "loss": 0.0225,
      "step": 80580
    },
    {
      "epoch": 31.3,
      "learning_rate": 1.5827055016181233e-05,
      "loss": 0.1591,
      "step": 80590
    },
    {
      "epoch": 31.3,
      "learning_rate": 1.582653721682848e-05,
      "loss": 0.134,
      "step": 80600
    },
    {
      "epoch": 31.3,
      "learning_rate": 1.582601941747573e-05,
      "loss": 0.0793,
      "step": 80610
    },
    {
      "epoch": 31.31,
      "learning_rate": 1.5825501618122977e-05,
      "loss": 0.1073,
      "step": 80620
    },
    {
      "epoch": 31.31,
      "learning_rate": 1.5824983818770228e-05,
      "loss": 0.0361,
      "step": 80630
    },
    {
      "epoch": 31.32,
      "learning_rate": 1.5824466019417476e-05,
      "loss": 0.0225,
      "step": 80640
    },
    {
      "epoch": 31.32,
      "learning_rate": 1.5823948220064727e-05,
      "loss": 0.1086,
      "step": 80650
    },
    {
      "epoch": 31.32,
      "learning_rate": 1.5823430420711975e-05,
      "loss": 0.25,
      "step": 80660
    },
    {
      "epoch": 31.33,
      "learning_rate": 1.5822912621359226e-05,
      "loss": 0.0878,
      "step": 80670
    },
    {
      "epoch": 31.33,
      "learning_rate": 1.5822394822006474e-05,
      "loss": 0.0739,
      "step": 80680
    },
    {
      "epoch": 31.34,
      "learning_rate": 1.5821877022653722e-05,
      "loss": 0.0501,
      "step": 80690
    },
    {
      "epoch": 31.34,
      "learning_rate": 1.582135922330097e-05,
      "loss": 0.003,
      "step": 80700
    },
    {
      "epoch": 31.34,
      "learning_rate": 1.582084142394822e-05,
      "loss": 0.0249,
      "step": 80710
    },
    {
      "epoch": 31.35,
      "learning_rate": 1.582032362459547e-05,
      "loss": 0.0447,
      "step": 80720
    },
    {
      "epoch": 31.35,
      "learning_rate": 1.581980582524272e-05,
      "loss": 0.1455,
      "step": 80730
    },
    {
      "epoch": 31.36,
      "learning_rate": 1.581928802588997e-05,
      "loss": 0.0907,
      "step": 80740
    },
    {
      "epoch": 31.36,
      "learning_rate": 1.581877022653722e-05,
      "loss": 0.1561,
      "step": 80750
    },
    {
      "epoch": 31.36,
      "learning_rate": 1.5818252427184468e-05,
      "loss": 0.1493,
      "step": 80760
    },
    {
      "epoch": 31.37,
      "learning_rate": 1.5817734627831716e-05,
      "loss": 0.0899,
      "step": 80770
    },
    {
      "epoch": 31.37,
      "learning_rate": 1.5817216828478964e-05,
      "loss": 0.0159,
      "step": 80780
    },
    {
      "epoch": 31.37,
      "learning_rate": 1.5816699029126215e-05,
      "loss": 0.1836,
      "step": 80790
    },
    {
      "epoch": 31.38,
      "learning_rate": 1.5816181229773463e-05,
      "loss": 0.0444,
      "step": 80800
    },
    {
      "epoch": 31.38,
      "learning_rate": 1.5815663430420714e-05,
      "loss": 0.1191,
      "step": 80810
    },
    {
      "epoch": 31.39,
      "learning_rate": 1.5815145631067962e-05,
      "loss": 0.1113,
      "step": 80820
    },
    {
      "epoch": 31.39,
      "learning_rate": 1.5814627831715213e-05,
      "loss": 0.1062,
      "step": 80830
    },
    {
      "epoch": 31.39,
      "learning_rate": 1.581411003236246e-05,
      "loss": 0.131,
      "step": 80840
    },
    {
      "epoch": 31.4,
      "learning_rate": 1.581359223300971e-05,
      "loss": 0.1514,
      "step": 80850
    },
    {
      "epoch": 31.4,
      "learning_rate": 1.5813074433656957e-05,
      "loss": 0.1239,
      "step": 80860
    },
    {
      "epoch": 31.41,
      "learning_rate": 1.581255663430421e-05,
      "loss": 0.0198,
      "step": 80870
    },
    {
      "epoch": 31.41,
      "learning_rate": 1.5812038834951456e-05,
      "loss": 0.0643,
      "step": 80880
    },
    {
      "epoch": 31.41,
      "learning_rate": 1.5811521035598708e-05,
      "loss": 0.0768,
      "step": 80890
    },
    {
      "epoch": 31.42,
      "learning_rate": 1.5811003236245956e-05,
      "loss": 0.111,
      "step": 80900
    },
    {
      "epoch": 31.42,
      "learning_rate": 1.5810485436893207e-05,
      "loss": 0.0403,
      "step": 80910
    },
    {
      "epoch": 31.43,
      "learning_rate": 1.5809967637540455e-05,
      "loss": 0.0455,
      "step": 80920
    },
    {
      "epoch": 31.43,
      "learning_rate": 1.5809449838187703e-05,
      "loss": 0.1165,
      "step": 80930
    },
    {
      "epoch": 31.43,
      "learning_rate": 1.580893203883495e-05,
      "loss": 0.121,
      "step": 80940
    },
    {
      "epoch": 31.44,
      "learning_rate": 1.5808414239482202e-05,
      "loss": 0.149,
      "step": 80950
    },
    {
      "epoch": 31.44,
      "learning_rate": 1.580789644012945e-05,
      "loss": 0.1459,
      "step": 80960
    },
    {
      "epoch": 31.44,
      "learning_rate": 1.58073786407767e-05,
      "loss": 0.1263,
      "step": 80970
    },
    {
      "epoch": 31.45,
      "learning_rate": 1.580686084142395e-05,
      "loss": 0.1683,
      "step": 80980
    },
    {
      "epoch": 31.45,
      "learning_rate": 1.58063430420712e-05,
      "loss": 0.0472,
      "step": 80990
    },
    {
      "epoch": 31.46,
      "learning_rate": 1.580582524271845e-05,
      "loss": 0.1875,
      "step": 81000
    },
    {
      "epoch": 31.46,
      "learning_rate": 1.5805307443365696e-05,
      "loss": 0.0189,
      "step": 81010
    },
    {
      "epoch": 31.46,
      "learning_rate": 1.5804789644012944e-05,
      "loss": 0.1034,
      "step": 81020
    },
    {
      "epoch": 31.47,
      "learning_rate": 1.5804271844660196e-05,
      "loss": 0.193,
      "step": 81030
    },
    {
      "epoch": 31.47,
      "learning_rate": 1.5803754045307444e-05,
      "loss": 0.0546,
      "step": 81040
    },
    {
      "epoch": 31.48,
      "learning_rate": 1.5803236245954695e-05,
      "loss": 0.0404,
      "step": 81050
    },
    {
      "epoch": 31.48,
      "learning_rate": 1.5802718446601943e-05,
      "loss": 0.2437,
      "step": 81060
    },
    {
      "epoch": 31.48,
      "learning_rate": 1.5802200647249194e-05,
      "loss": 0.0604,
      "step": 81070
    },
    {
      "epoch": 31.49,
      "learning_rate": 1.5801682847896442e-05,
      "loss": 0.1172,
      "step": 81080
    },
    {
      "epoch": 31.49,
      "learning_rate": 1.580116504854369e-05,
      "loss": 0.0731,
      "step": 81090
    },
    {
      "epoch": 31.5,
      "learning_rate": 1.5800647249190938e-05,
      "loss": 0.0267,
      "step": 81100
    },
    {
      "epoch": 31.5,
      "learning_rate": 1.580012944983819e-05,
      "loss": 0.0445,
      "step": 81110
    },
    {
      "epoch": 31.5,
      "learning_rate": 1.5799611650485437e-05,
      "loss": 0.1319,
      "step": 81120
    },
    {
      "epoch": 31.51,
      "learning_rate": 1.579909385113269e-05,
      "loss": 0.0605,
      "step": 81130
    },
    {
      "epoch": 31.51,
      "learning_rate": 1.5798576051779936e-05,
      "loss": 0.0454,
      "step": 81140
    },
    {
      "epoch": 31.51,
      "learning_rate": 1.5798058252427188e-05,
      "loss": 0.0788,
      "step": 81150
    },
    {
      "epoch": 31.52,
      "learning_rate": 1.5797540453074436e-05,
      "loss": 0.0258,
      "step": 81160
    },
    {
      "epoch": 31.52,
      "learning_rate": 1.5797022653721684e-05,
      "loss": 0.0828,
      "step": 81170
    },
    {
      "epoch": 31.53,
      "learning_rate": 1.579650485436893e-05,
      "loss": 0.2522,
      "step": 81180
    },
    {
      "epoch": 31.53,
      "learning_rate": 1.5795987055016183e-05,
      "loss": 0.169,
      "step": 81190
    },
    {
      "epoch": 31.53,
      "learning_rate": 1.579546925566343e-05,
      "loss": 0.0776,
      "step": 81200
    },
    {
      "epoch": 31.54,
      "learning_rate": 1.5794951456310682e-05,
      "loss": 0.1859,
      "step": 81210
    },
    {
      "epoch": 31.54,
      "learning_rate": 1.579443365695793e-05,
      "loss": 0.1544,
      "step": 81220
    },
    {
      "epoch": 31.55,
      "learning_rate": 1.579391585760518e-05,
      "loss": 0.0964,
      "step": 81230
    },
    {
      "epoch": 31.55,
      "learning_rate": 1.579339805825243e-05,
      "loss": 0.0879,
      "step": 81240
    },
    {
      "epoch": 31.55,
      "learning_rate": 1.5792880258899677e-05,
      "loss": 0.1924,
      "step": 81250
    },
    {
      "epoch": 31.56,
      "learning_rate": 1.5792362459546925e-05,
      "loss": 0.1241,
      "step": 81260
    },
    {
      "epoch": 31.56,
      "learning_rate": 1.5791844660194176e-05,
      "loss": 0.096,
      "step": 81270
    },
    {
      "epoch": 31.57,
      "learning_rate": 1.5791326860841424e-05,
      "loss": 0.0443,
      "step": 81280
    },
    {
      "epoch": 31.57,
      "learning_rate": 1.5790809061488676e-05,
      "loss": 0.1975,
      "step": 81290
    },
    {
      "epoch": 31.57,
      "learning_rate": 1.5790291262135923e-05,
      "loss": 0.2285,
      "step": 81300
    },
    {
      "epoch": 31.58,
      "learning_rate": 1.5789773462783175e-05,
      "loss": 0.0578,
      "step": 81310
    },
    {
      "epoch": 31.58,
      "learning_rate": 1.5789255663430423e-05,
      "loss": 0.0368,
      "step": 81320
    },
    {
      "epoch": 31.58,
      "learning_rate": 1.578873786407767e-05,
      "loss": 0.1425,
      "step": 81330
    },
    {
      "epoch": 31.59,
      "learning_rate": 1.578822006472492e-05,
      "loss": 0.0987,
      "step": 81340
    },
    {
      "epoch": 31.59,
      "learning_rate": 1.578770226537217e-05,
      "loss": 0.0983,
      "step": 81350
    },
    {
      "epoch": 31.6,
      "learning_rate": 1.5787184466019418e-05,
      "loss": 0.2737,
      "step": 81360
    },
    {
      "epoch": 31.6,
      "learning_rate": 1.578666666666667e-05,
      "loss": 0.0201,
      "step": 81370
    },
    {
      "epoch": 31.6,
      "learning_rate": 1.5786148867313917e-05,
      "loss": 0.0907,
      "step": 81380
    },
    {
      "epoch": 31.61,
      "learning_rate": 1.578563106796117e-05,
      "loss": 0.0222,
      "step": 81390
    },
    {
      "epoch": 31.61,
      "learning_rate": 1.5785113268608416e-05,
      "loss": 0.0387,
      "step": 81400
    },
    {
      "epoch": 31.62,
      "learning_rate": 1.5784595469255664e-05,
      "loss": 0.0158,
      "step": 81410
    },
    {
      "epoch": 31.62,
      "learning_rate": 1.5784077669902912e-05,
      "loss": 0.0356,
      "step": 81420
    },
    {
      "epoch": 31.62,
      "learning_rate": 1.5783559870550163e-05,
      "loss": 0.151,
      "step": 81430
    },
    {
      "epoch": 31.63,
      "learning_rate": 1.578304207119741e-05,
      "loss": 0.0202,
      "step": 81440
    },
    {
      "epoch": 31.63,
      "learning_rate": 1.5782524271844663e-05,
      "loss": 0.1814,
      "step": 81450
    },
    {
      "epoch": 31.63,
      "learning_rate": 1.578200647249191e-05,
      "loss": 0.0958,
      "step": 81460
    },
    {
      "epoch": 31.64,
      "learning_rate": 1.5781488673139162e-05,
      "loss": 0.1141,
      "step": 81470
    },
    {
      "epoch": 31.64,
      "learning_rate": 1.578097087378641e-05,
      "loss": 0.2331,
      "step": 81480
    },
    {
      "epoch": 31.65,
      "learning_rate": 1.5780453074433658e-05,
      "loss": 0.1246,
      "step": 81490
    },
    {
      "epoch": 31.65,
      "learning_rate": 1.5779935275080906e-05,
      "loss": 0.0844,
      "step": 81500
    },
    {
      "epoch": 31.65,
      "learning_rate": 1.5779417475728157e-05,
      "loss": 0.117,
      "step": 81510
    },
    {
      "epoch": 31.66,
      "learning_rate": 1.5778899676375405e-05,
      "loss": 0.2999,
      "step": 81520
    },
    {
      "epoch": 31.66,
      "learning_rate": 1.5778381877022656e-05,
      "loss": 0.0883,
      "step": 81530
    },
    {
      "epoch": 31.67,
      "learning_rate": 1.5777864077669904e-05,
      "loss": 0.2339,
      "step": 81540
    },
    {
      "epoch": 31.67,
      "learning_rate": 1.5777346278317155e-05,
      "loss": 0.0613,
      "step": 81550
    },
    {
      "epoch": 31.67,
      "learning_rate": 1.5776828478964403e-05,
      "loss": 0.0681,
      "step": 81560
    },
    {
      "epoch": 31.68,
      "learning_rate": 1.577631067961165e-05,
      "loss": 0.1319,
      "step": 81570
    },
    {
      "epoch": 31.68,
      "learning_rate": 1.57757928802589e-05,
      "loss": 0.0603,
      "step": 81580
    },
    {
      "epoch": 31.69,
      "learning_rate": 1.577527508090615e-05,
      "loss": 0.1537,
      "step": 81590
    },
    {
      "epoch": 31.69,
      "learning_rate": 1.57747572815534e-05,
      "loss": 0.0671,
      "step": 81600
    },
    {
      "epoch": 31.69,
      "learning_rate": 1.577423948220065e-05,
      "loss": 0.142,
      "step": 81610
    },
    {
      "epoch": 31.7,
      "learning_rate": 1.5773721682847898e-05,
      "loss": 0.0569,
      "step": 81620
    },
    {
      "epoch": 31.7,
      "learning_rate": 1.5773203883495146e-05,
      "loss": 0.1539,
      "step": 81630
    },
    {
      "epoch": 31.7,
      "learning_rate": 1.5772686084142397e-05,
      "loss": 0.044,
      "step": 81640
    },
    {
      "epoch": 31.71,
      "learning_rate": 1.5772168284789645e-05,
      "loss": 0.1461,
      "step": 81650
    },
    {
      "epoch": 31.71,
      "learning_rate": 1.5771650485436893e-05,
      "loss": 0.0335,
      "step": 81660
    },
    {
      "epoch": 31.72,
      "learning_rate": 1.5771132686084144e-05,
      "loss": 0.1246,
      "step": 81670
    },
    {
      "epoch": 31.72,
      "learning_rate": 1.5770614886731392e-05,
      "loss": 0.1445,
      "step": 81680
    },
    {
      "epoch": 31.72,
      "learning_rate": 1.5770097087378643e-05,
      "loss": 0.2244,
      "step": 81690
    },
    {
      "epoch": 31.73,
      "learning_rate": 1.576957928802589e-05,
      "loss": 0.0412,
      "step": 81700
    },
    {
      "epoch": 31.73,
      "learning_rate": 1.576906148867314e-05,
      "loss": 0.05,
      "step": 81710
    },
    {
      "epoch": 31.74,
      "learning_rate": 1.576854368932039e-05,
      "loss": 0.1203,
      "step": 81720
    },
    {
      "epoch": 31.74,
      "learning_rate": 1.576802588996764e-05,
      "loss": 0.1738,
      "step": 81730
    },
    {
      "epoch": 31.74,
      "learning_rate": 1.5767508090614886e-05,
      "loss": 0.1664,
      "step": 81740
    },
    {
      "epoch": 31.75,
      "learning_rate": 1.5766990291262138e-05,
      "loss": 0.0469,
      "step": 81750
    },
    {
      "epoch": 31.75,
      "learning_rate": 1.5766472491909386e-05,
      "loss": 0.0702,
      "step": 81760
    },
    {
      "epoch": 31.76,
      "learning_rate": 1.5765954692556637e-05,
      "loss": 0.1738,
      "step": 81770
    },
    {
      "epoch": 31.76,
      "learning_rate": 1.5765436893203885e-05,
      "loss": 0.2161,
      "step": 81780
    },
    {
      "epoch": 31.76,
      "learning_rate": 1.5764919093851133e-05,
      "loss": 0.1349,
      "step": 81790
    },
    {
      "epoch": 31.77,
      "learning_rate": 1.5764401294498384e-05,
      "loss": 0.0965,
      "step": 81800
    },
    {
      "epoch": 31.77,
      "learning_rate": 1.5763883495145632e-05,
      "loss": 0.1095,
      "step": 81810
    },
    {
      "epoch": 31.77,
      "learning_rate": 1.576336569579288e-05,
      "loss": 0.0312,
      "step": 81820
    },
    {
      "epoch": 31.78,
      "learning_rate": 1.576284789644013e-05,
      "loss": 0.1628,
      "step": 81830
    },
    {
      "epoch": 31.78,
      "learning_rate": 1.576233009708738e-05,
      "loss": 0.0501,
      "step": 81840
    },
    {
      "epoch": 31.79,
      "learning_rate": 1.576181229773463e-05,
      "loss": 0.0454,
      "step": 81850
    },
    {
      "epoch": 31.79,
      "learning_rate": 1.576129449838188e-05,
      "loss": 0.1038,
      "step": 81860
    },
    {
      "epoch": 31.79,
      "learning_rate": 1.5760776699029126e-05,
      "loss": 0.05,
      "step": 81870
    },
    {
      "epoch": 31.8,
      "learning_rate": 1.5760258899676378e-05,
      "loss": 0.0945,
      "step": 81880
    },
    {
      "epoch": 31.8,
      "learning_rate": 1.5759741100323626e-05,
      "loss": 0.0741,
      "step": 81890
    },
    {
      "epoch": 31.81,
      "learning_rate": 1.5759223300970877e-05,
      "loss": 0.0923,
      "step": 81900
    },
    {
      "epoch": 31.81,
      "learning_rate": 1.5758705501618125e-05,
      "loss": 0.0404,
      "step": 81910
    },
    {
      "epoch": 31.81,
      "learning_rate": 1.5758187702265373e-05,
      "loss": 0.191,
      "step": 81920
    },
    {
      "epoch": 31.82,
      "learning_rate": 1.5757669902912624e-05,
      "loss": 0.139,
      "step": 81930
    },
    {
      "epoch": 31.82,
      "learning_rate": 1.5757152103559872e-05,
      "loss": 0.0312,
      "step": 81940
    },
    {
      "epoch": 31.83,
      "learning_rate": 1.575663430420712e-05,
      "loss": 0.0885,
      "step": 81950
    },
    {
      "epoch": 31.83,
      "learning_rate": 1.575611650485437e-05,
      "loss": 0.0501,
      "step": 81960
    },
    {
      "epoch": 31.83,
      "learning_rate": 1.575559870550162e-05,
      "loss": 0.1062,
      "step": 81970
    },
    {
      "epoch": 31.84,
      "learning_rate": 1.575508090614887e-05,
      "loss": 0.0499,
      "step": 81980
    },
    {
      "epoch": 31.84,
      "learning_rate": 1.575456310679612e-05,
      "loss": 0.0018,
      "step": 81990
    },
    {
      "epoch": 31.84,
      "learning_rate": 1.5754045307443366e-05,
      "loss": 0.0342,
      "step": 82000
    },
    {
      "epoch": 31.85,
      "learning_rate": 1.5753527508090614e-05,
      "loss": 0.0064,
      "step": 82010
    },
    {
      "epoch": 31.85,
      "learning_rate": 1.5753009708737865e-05,
      "loss": 0.0604,
      "step": 82020
    },
    {
      "epoch": 31.86,
      "learning_rate": 1.5752491909385113e-05,
      "loss": 0.1887,
      "step": 82030
    },
    {
      "epoch": 31.86,
      "learning_rate": 1.5751974110032365e-05,
      "loss": 0.0744,
      "step": 82040
    },
    {
      "epoch": 31.86,
      "learning_rate": 1.5751456310679613e-05,
      "loss": 0.0989,
      "step": 82050
    },
    {
      "epoch": 31.87,
      "learning_rate": 1.5750938511326864e-05,
      "loss": 0.1854,
      "step": 82060
    },
    {
      "epoch": 31.87,
      "learning_rate": 1.5750420711974112e-05,
      "loss": 0.1686,
      "step": 82070
    },
    {
      "epoch": 31.88,
      "learning_rate": 1.574990291262136e-05,
      "loss": 0.1062,
      "step": 82080
    },
    {
      "epoch": 31.88,
      "learning_rate": 1.5749385113268608e-05,
      "loss": 0.0241,
      "step": 82090
    },
    {
      "epoch": 31.88,
      "learning_rate": 1.574886731391586e-05,
      "loss": 0.0975,
      "step": 82100
    },
    {
      "epoch": 31.89,
      "learning_rate": 1.5748349514563107e-05,
      "loss": 0.0563,
      "step": 82110
    },
    {
      "epoch": 31.89,
      "learning_rate": 1.5747831715210358e-05,
      "loss": 0.1118,
      "step": 82120
    },
    {
      "epoch": 31.9,
      "learning_rate": 1.5747313915857606e-05,
      "loss": 0.0898,
      "step": 82130
    },
    {
      "epoch": 31.9,
      "learning_rate": 1.5746796116504858e-05,
      "loss": 0.117,
      "step": 82140
    },
    {
      "epoch": 31.9,
      "learning_rate": 1.5746278317152105e-05,
      "loss": 0.2436,
      "step": 82150
    },
    {
      "epoch": 31.91,
      "learning_rate": 1.5745760517799353e-05,
      "loss": 0.0819,
      "step": 82160
    },
    {
      "epoch": 31.91,
      "learning_rate": 1.57452427184466e-05,
      "loss": 0.0787,
      "step": 82170
    },
    {
      "epoch": 31.91,
      "learning_rate": 1.5744724919093853e-05,
      "loss": 0.0105,
      "step": 82180
    },
    {
      "epoch": 31.92,
      "learning_rate": 1.57442071197411e-05,
      "loss": 0.1154,
      "step": 82190
    },
    {
      "epoch": 31.92,
      "learning_rate": 1.5743689320388352e-05,
      "loss": 0.0962,
      "step": 82200
    },
    {
      "epoch": 31.93,
      "learning_rate": 1.57431715210356e-05,
      "loss": 0.1379,
      "step": 82210
    },
    {
      "epoch": 31.93,
      "learning_rate": 1.574265372168285e-05,
      "loss": 0.0586,
      "step": 82220
    },
    {
      "epoch": 31.93,
      "learning_rate": 1.57421359223301e-05,
      "loss": 0.1329,
      "step": 82230
    },
    {
      "epoch": 31.94,
      "learning_rate": 1.5741618122977347e-05,
      "loss": 0.0336,
      "step": 82240
    },
    {
      "epoch": 31.94,
      "learning_rate": 1.5741100323624595e-05,
      "loss": 0.0456,
      "step": 82250
    },
    {
      "epoch": 31.95,
      "learning_rate": 1.5740582524271846e-05,
      "loss": 0.2175,
      "step": 82260
    },
    {
      "epoch": 31.95,
      "learning_rate": 1.5740064724919094e-05,
      "loss": 0.1196,
      "step": 82270
    },
    {
      "epoch": 31.95,
      "learning_rate": 1.5739546925566345e-05,
      "loss": 0.0408,
      "step": 82280
    },
    {
      "epoch": 31.96,
      "learning_rate": 1.5739029126213593e-05,
      "loss": 0.166,
      "step": 82290
    },
    {
      "epoch": 31.96,
      "learning_rate": 1.5738511326860845e-05,
      "loss": 0.1685,
      "step": 82300
    },
    {
      "epoch": 31.97,
      "learning_rate": 1.573799352750809e-05,
      "loss": 0.0751,
      "step": 82310
    },
    {
      "epoch": 31.97,
      "learning_rate": 1.573747572815534e-05,
      "loss": 0.0289,
      "step": 82320
    },
    {
      "epoch": 31.97,
      "learning_rate": 1.573695792880259e-05,
      "loss": 0.0539,
      "step": 82330
    },
    {
      "epoch": 31.98,
      "learning_rate": 1.573644012944984e-05,
      "loss": 0.0254,
      "step": 82340
    },
    {
      "epoch": 31.98,
      "learning_rate": 1.5735922330097088e-05,
      "loss": 0.2161,
      "step": 82350
    },
    {
      "epoch": 31.98,
      "learning_rate": 1.573540453074434e-05,
      "loss": 0.0679,
      "step": 82360
    },
    {
      "epoch": 31.99,
      "learning_rate": 1.5734886731391587e-05,
      "loss": 0.126,
      "step": 82370
    },
    {
      "epoch": 31.99,
      "learning_rate": 1.5734368932038838e-05,
      "loss": 0.1806,
      "step": 82380
    },
    {
      "epoch": 32.0,
      "learning_rate": 1.5733851132686083e-05,
      "loss": 0.3364,
      "step": 82390
    },
    {
      "epoch": 32.0,
      "learning_rate": 1.5733333333333334e-05,
      "loss": 0.0324,
      "step": 82400
    },
    {
      "epoch": 32.0,
      "eval_accuracy": 0.9452544704264099,
      "eval_loss": 0.32230931520462036,
      "eval_runtime": 8.2899,
      "eval_samples_per_second": 438.486,
      "eval_steps_per_second": 54.886,
      "step": 82400
    },
    {
      "epoch": 32.0,
      "learning_rate": 1.5732815533980582e-05,
      "loss": 0.2413,
      "step": 82410
    },
    {
      "epoch": 32.01,
      "learning_rate": 1.5732297734627833e-05,
      "loss": 0.0307,
      "step": 82420
    },
    {
      "epoch": 32.01,
      "learning_rate": 1.573177993527508e-05,
      "loss": 0.1562,
      "step": 82430
    },
    {
      "epoch": 32.02,
      "learning_rate": 1.5731262135922333e-05,
      "loss": 0.1028,
      "step": 82440
    },
    {
      "epoch": 32.02,
      "learning_rate": 1.573074433656958e-05,
      "loss": 0.0045,
      "step": 82450
    },
    {
      "epoch": 32.02,
      "learning_rate": 1.5730226537216832e-05,
      "loss": 0.0736,
      "step": 82460
    },
    {
      "epoch": 32.03,
      "learning_rate": 1.572970873786408e-05,
      "loss": 0.2257,
      "step": 82470
    },
    {
      "epoch": 32.03,
      "learning_rate": 1.5729190938511328e-05,
      "loss": 0.0213,
      "step": 82480
    },
    {
      "epoch": 32.03,
      "learning_rate": 1.5728673139158576e-05,
      "loss": 0.11,
      "step": 82490
    },
    {
      "epoch": 32.04,
      "learning_rate": 1.5728155339805827e-05,
      "loss": 0.1479,
      "step": 82500
    },
    {
      "epoch": 32.04,
      "learning_rate": 1.5727637540453075e-05,
      "loss": 0.1123,
      "step": 82510
    },
    {
      "epoch": 32.05,
      "learning_rate": 1.5727119741100326e-05,
      "loss": 0.1181,
      "step": 82520
    },
    {
      "epoch": 32.05,
      "learning_rate": 1.5726601941747574e-05,
      "loss": 0.0214,
      "step": 82530
    },
    {
      "epoch": 32.05,
      "learning_rate": 1.5726084142394825e-05,
      "loss": 0.0551,
      "step": 82540
    },
    {
      "epoch": 32.06,
      "learning_rate": 1.5725566343042073e-05,
      "loss": 0.0624,
      "step": 82550
    },
    {
      "epoch": 32.06,
      "learning_rate": 1.572504854368932e-05,
      "loss": 0.0674,
      "step": 82560
    },
    {
      "epoch": 32.07,
      "learning_rate": 1.572453074433657e-05,
      "loss": 0.0192,
      "step": 82570
    },
    {
      "epoch": 32.07,
      "learning_rate": 1.572401294498382e-05,
      "loss": 0.14,
      "step": 82580
    },
    {
      "epoch": 32.07,
      "learning_rate": 1.5723495145631068e-05,
      "loss": 0.0624,
      "step": 82590
    },
    {
      "epoch": 32.08,
      "learning_rate": 1.572297734627832e-05,
      "loss": 0.1196,
      "step": 82600
    },
    {
      "epoch": 32.08,
      "learning_rate": 1.5722459546925568e-05,
      "loss": 0.0608,
      "step": 82610
    },
    {
      "epoch": 32.09,
      "learning_rate": 1.572194174757282e-05,
      "loss": 0.193,
      "step": 82620
    },
    {
      "epoch": 32.09,
      "learning_rate": 1.5721423948220067e-05,
      "loss": 0.0403,
      "step": 82630
    },
    {
      "epoch": 32.09,
      "learning_rate": 1.5720906148867315e-05,
      "loss": 0.0047,
      "step": 82640
    },
    {
      "epoch": 32.1,
      "learning_rate": 1.5720388349514563e-05,
      "loss": 0.0988,
      "step": 82650
    },
    {
      "epoch": 32.1,
      "learning_rate": 1.5719870550161814e-05,
      "loss": 0.1564,
      "step": 82660
    },
    {
      "epoch": 32.1,
      "learning_rate": 1.5719352750809062e-05,
      "loss": 0.053,
      "step": 82670
    },
    {
      "epoch": 32.11,
      "learning_rate": 1.5718834951456313e-05,
      "loss": 0.1556,
      "step": 82680
    },
    {
      "epoch": 32.11,
      "learning_rate": 1.571831715210356e-05,
      "loss": 0.0161,
      "step": 82690
    },
    {
      "epoch": 32.12,
      "learning_rate": 1.5717799352750812e-05,
      "loss": 0.101,
      "step": 82700
    },
    {
      "epoch": 32.12,
      "learning_rate": 1.571728155339806e-05,
      "loss": 0.099,
      "step": 82710
    },
    {
      "epoch": 32.12,
      "learning_rate": 1.5716763754045308e-05,
      "loss": 0.0451,
      "step": 82720
    },
    {
      "epoch": 32.13,
      "learning_rate": 1.5716245954692556e-05,
      "loss": 0.1561,
      "step": 82730
    },
    {
      "epoch": 32.13,
      "learning_rate": 1.5715728155339807e-05,
      "loss": 0.0126,
      "step": 82740
    },
    {
      "epoch": 32.14,
      "learning_rate": 1.5715210355987055e-05,
      "loss": 0.029,
      "step": 82750
    },
    {
      "epoch": 32.14,
      "learning_rate": 1.5714692556634307e-05,
      "loss": 0.1009,
      "step": 82760
    },
    {
      "epoch": 32.14,
      "learning_rate": 1.5714174757281555e-05,
      "loss": 0.052,
      "step": 82770
    },
    {
      "epoch": 32.15,
      "learning_rate": 1.5713656957928806e-05,
      "loss": 0.1718,
      "step": 82780
    },
    {
      "epoch": 32.15,
      "learning_rate": 1.5713139158576054e-05,
      "loss": 0.1382,
      "step": 82790
    },
    {
      "epoch": 32.16,
      "learning_rate": 1.5712621359223302e-05,
      "loss": 0.0477,
      "step": 82800
    },
    {
      "epoch": 32.16,
      "learning_rate": 1.571210355987055e-05,
      "loss": 0.0154,
      "step": 82810
    },
    {
      "epoch": 32.16,
      "learning_rate": 1.57115857605178e-05,
      "loss": 0.0992,
      "step": 82820
    },
    {
      "epoch": 32.17,
      "learning_rate": 1.571106796116505e-05,
      "loss": 0.0791,
      "step": 82830
    },
    {
      "epoch": 32.17,
      "learning_rate": 1.57105501618123e-05,
      "loss": 0.1044,
      "step": 82840
    },
    {
      "epoch": 32.17,
      "learning_rate": 1.5710032362459548e-05,
      "loss": 0.077,
      "step": 82850
    },
    {
      "epoch": 32.18,
      "learning_rate": 1.57095145631068e-05,
      "loss": 0.3981,
      "step": 82860
    },
    {
      "epoch": 32.18,
      "learning_rate": 1.5708996763754047e-05,
      "loss": 0.1318,
      "step": 82870
    },
    {
      "epoch": 32.19,
      "learning_rate": 1.5708478964401295e-05,
      "loss": 0.1211,
      "step": 82880
    },
    {
      "epoch": 32.19,
      "learning_rate": 1.5707961165048543e-05,
      "loss": 0.0905,
      "step": 82890
    },
    {
      "epoch": 32.19,
      "learning_rate": 1.5707443365695795e-05,
      "loss": 0.0536,
      "step": 82900
    },
    {
      "epoch": 32.2,
      "learning_rate": 1.5706925566343043e-05,
      "loss": 0.207,
      "step": 82910
    },
    {
      "epoch": 32.2,
      "learning_rate": 1.5706407766990294e-05,
      "loss": 0.0451,
      "step": 82920
    },
    {
      "epoch": 32.21,
      "learning_rate": 1.5705889967637542e-05,
      "loss": 0.2162,
      "step": 82930
    },
    {
      "epoch": 32.21,
      "learning_rate": 1.5705372168284793e-05,
      "loss": 0.2077,
      "step": 82940
    },
    {
      "epoch": 32.21,
      "learning_rate": 1.570485436893204e-05,
      "loss": 0.0194,
      "step": 82950
    },
    {
      "epoch": 32.22,
      "learning_rate": 1.570433656957929e-05,
      "loss": 0.0579,
      "step": 82960
    },
    {
      "epoch": 32.22,
      "learning_rate": 1.5703818770226537e-05,
      "loss": 0.0763,
      "step": 82970
    },
    {
      "epoch": 32.23,
      "learning_rate": 1.5703300970873788e-05,
      "loss": 0.1409,
      "step": 82980
    },
    {
      "epoch": 32.23,
      "learning_rate": 1.5702783171521036e-05,
      "loss": 0.0804,
      "step": 82990
    },
    {
      "epoch": 32.23,
      "learning_rate": 1.5702265372168287e-05,
      "loss": 0.1186,
      "step": 83000
    },
    {
      "epoch": 32.24,
      "learning_rate": 1.5701747572815535e-05,
      "loss": 0.0263,
      "step": 83010
    },
    {
      "epoch": 32.24,
      "learning_rate": 1.5701229773462787e-05,
      "loss": 0.1713,
      "step": 83020
    },
    {
      "epoch": 32.24,
      "learning_rate": 1.5700711974110035e-05,
      "loss": 0.0708,
      "step": 83030
    },
    {
      "epoch": 32.25,
      "learning_rate": 1.5700194174757282e-05,
      "loss": 0.0702,
      "step": 83040
    },
    {
      "epoch": 32.25,
      "learning_rate": 1.569967637540453e-05,
      "loss": 0.085,
      "step": 83050
    },
    {
      "epoch": 32.26,
      "learning_rate": 1.5699158576051782e-05,
      "loss": 0.004,
      "step": 83060
    },
    {
      "epoch": 32.26,
      "learning_rate": 1.569864077669903e-05,
      "loss": 0.0447,
      "step": 83070
    },
    {
      "epoch": 32.26,
      "learning_rate": 1.569812297734628e-05,
      "loss": 0.1735,
      "step": 83080
    },
    {
      "epoch": 32.27,
      "learning_rate": 1.569760517799353e-05,
      "loss": 0.257,
      "step": 83090
    },
    {
      "epoch": 32.27,
      "learning_rate": 1.5697087378640777e-05,
      "loss": 0.0635,
      "step": 83100
    },
    {
      "epoch": 32.28,
      "learning_rate": 1.5696569579288028e-05,
      "loss": 0.0179,
      "step": 83110
    },
    {
      "epoch": 32.28,
      "learning_rate": 1.5696051779935276e-05,
      "loss": 0.039,
      "step": 83120
    },
    {
      "epoch": 32.28,
      "learning_rate": 1.5695533980582524e-05,
      "loss": 0.0313,
      "step": 83130
    },
    {
      "epoch": 32.29,
      "learning_rate": 1.5695016181229775e-05,
      "loss": 0.1883,
      "step": 83140
    },
    {
      "epoch": 32.29,
      "learning_rate": 1.5694498381877023e-05,
      "loss": 0.0029,
      "step": 83150
    },
    {
      "epoch": 32.3,
      "learning_rate": 1.5693980582524275e-05,
      "loss": 0.0904,
      "step": 83160
    },
    {
      "epoch": 32.3,
      "learning_rate": 1.5693462783171522e-05,
      "loss": 0.0563,
      "step": 83170
    },
    {
      "epoch": 32.3,
      "learning_rate": 1.569294498381877e-05,
      "loss": 0.0854,
      "step": 83180
    },
    {
      "epoch": 32.31,
      "learning_rate": 1.569242718446602e-05,
      "loss": 0.0636,
      "step": 83190
    },
    {
      "epoch": 32.31,
      "learning_rate": 1.569190938511327e-05,
      "loss": 0.0248,
      "step": 83200
    },
    {
      "epoch": 32.31,
      "learning_rate": 1.5691391585760518e-05,
      "loss": 0.1648,
      "step": 83210
    },
    {
      "epoch": 32.32,
      "learning_rate": 1.569087378640777e-05,
      "loss": 0.0142,
      "step": 83220
    },
    {
      "epoch": 32.32,
      "learning_rate": 1.5690355987055017e-05,
      "loss": 0.0602,
      "step": 83230
    },
    {
      "epoch": 32.33,
      "learning_rate": 1.5689838187702268e-05,
      "loss": 0.1955,
      "step": 83240
    },
    {
      "epoch": 32.33,
      "learning_rate": 1.5689320388349516e-05,
      "loss": 0.0561,
      "step": 83250
    },
    {
      "epoch": 32.33,
      "learning_rate": 1.5688802588996764e-05,
      "loss": 0.2445,
      "step": 83260
    },
    {
      "epoch": 32.34,
      "learning_rate": 1.5688284789644015e-05,
      "loss": 0.0801,
      "step": 83270
    },
    {
      "epoch": 32.34,
      "learning_rate": 1.5687766990291263e-05,
      "loss": 0.1716,
      "step": 83280
    },
    {
      "epoch": 32.35,
      "learning_rate": 1.568724919093851e-05,
      "loss": 0.0705,
      "step": 83290
    },
    {
      "epoch": 32.35,
      "learning_rate": 1.5686731391585762e-05,
      "loss": 0.1622,
      "step": 83300
    },
    {
      "epoch": 32.35,
      "learning_rate": 1.568621359223301e-05,
      "loss": 0.0651,
      "step": 83310
    },
    {
      "epoch": 32.36,
      "learning_rate": 1.568569579288026e-05,
      "loss": 0.2247,
      "step": 83320
    },
    {
      "epoch": 32.36,
      "learning_rate": 1.568517799352751e-05,
      "loss": 0.0639,
      "step": 83330
    },
    {
      "epoch": 32.37,
      "learning_rate": 1.5684660194174757e-05,
      "loss": 0.0564,
      "step": 83340
    },
    {
      "epoch": 32.37,
      "learning_rate": 1.568414239482201e-05,
      "loss": 0.1711,
      "step": 83350
    },
    {
      "epoch": 32.37,
      "learning_rate": 1.5683624595469257e-05,
      "loss": 0.0619,
      "step": 83360
    },
    {
      "epoch": 32.38,
      "learning_rate": 1.5683106796116505e-05,
      "loss": 0.1027,
      "step": 83370
    },
    {
      "epoch": 32.38,
      "learning_rate": 1.5682588996763756e-05,
      "loss": 0.0578,
      "step": 83380
    },
    {
      "epoch": 32.38,
      "learning_rate": 1.5682071197411004e-05,
      "loss": 0.1397,
      "step": 83390
    },
    {
      "epoch": 32.39,
      "learning_rate": 1.5681553398058255e-05,
      "loss": 0.2056,
      "step": 83400
    },
    {
      "epoch": 32.39,
      "learning_rate": 1.5681035598705503e-05,
      "loss": 0.0224,
      "step": 83410
    },
    {
      "epoch": 32.4,
      "learning_rate": 1.568051779935275e-05,
      "loss": 0.1929,
      "step": 83420
    },
    {
      "epoch": 32.4,
      "learning_rate": 1.5680000000000002e-05,
      "loss": 0.0187,
      "step": 83430
    },
    {
      "epoch": 32.4,
      "learning_rate": 1.567948220064725e-05,
      "loss": 0.0716,
      "step": 83440
    },
    {
      "epoch": 32.41,
      "learning_rate": 1.5678964401294498e-05,
      "loss": 0.0739,
      "step": 83450
    },
    {
      "epoch": 32.41,
      "learning_rate": 1.567844660194175e-05,
      "loss": 0.312,
      "step": 83460
    },
    {
      "epoch": 32.42,
      "learning_rate": 1.5677928802588997e-05,
      "loss": 0.2291,
      "step": 83470
    },
    {
      "epoch": 32.42,
      "learning_rate": 1.5677411003236245e-05,
      "loss": 0.0647,
      "step": 83480
    },
    {
      "epoch": 32.42,
      "learning_rate": 1.5676893203883497e-05,
      "loss": 0.08,
      "step": 83490
    },
    {
      "epoch": 32.43,
      "learning_rate": 1.5676375404530745e-05,
      "loss": 0.0964,
      "step": 83500
    },
    {
      "epoch": 32.43,
      "learning_rate": 1.5675857605177996e-05,
      "loss": 0.0485,
      "step": 83510
    },
    {
      "epoch": 32.43,
      "learning_rate": 1.5675339805825244e-05,
      "loss": 0.0657,
      "step": 83520
    },
    {
      "epoch": 32.44,
      "learning_rate": 1.5674822006472495e-05,
      "loss": 0.2999,
      "step": 83530
    },
    {
      "epoch": 32.44,
      "learning_rate": 1.5674304207119743e-05,
      "loss": 0.0512,
      "step": 83540
    },
    {
      "epoch": 32.45,
      "learning_rate": 1.567378640776699e-05,
      "loss": 0.1279,
      "step": 83550
    },
    {
      "epoch": 32.45,
      "learning_rate": 1.567326860841424e-05,
      "loss": 0.08,
      "step": 83560
    },
    {
      "epoch": 32.45,
      "learning_rate": 1.567275080906149e-05,
      "loss": 0.053,
      "step": 83570
    },
    {
      "epoch": 32.46,
      "learning_rate": 1.5672233009708738e-05,
      "loss": 0.0145,
      "step": 83580
    },
    {
      "epoch": 32.46,
      "learning_rate": 1.567171521035599e-05,
      "loss": 0.036,
      "step": 83590
    },
    {
      "epoch": 32.47,
      "learning_rate": 1.5671197411003237e-05,
      "loss": 0.0758,
      "step": 83600
    },
    {
      "epoch": 32.47,
      "learning_rate": 1.567067961165049e-05,
      "loss": 0.0118,
      "step": 83610
    },
    {
      "epoch": 32.47,
      "learning_rate": 1.5670161812297737e-05,
      "loss": 0.2479,
      "step": 83620
    },
    {
      "epoch": 32.48,
      "learning_rate": 1.5669644012944985e-05,
      "loss": 0.1186,
      "step": 83630
    },
    {
      "epoch": 32.48,
      "learning_rate": 1.5669126213592232e-05,
      "loss": 0.1545,
      "step": 83640
    },
    {
      "epoch": 32.49,
      "learning_rate": 1.5668608414239484e-05,
      "loss": 0.0452,
      "step": 83650
    },
    {
      "epoch": 32.49,
      "learning_rate": 1.5668090614886732e-05,
      "loss": 0.149,
      "step": 83660
    },
    {
      "epoch": 32.49,
      "learning_rate": 1.5667572815533983e-05,
      "loss": 0.2124,
      "step": 83670
    },
    {
      "epoch": 32.5,
      "learning_rate": 1.566705501618123e-05,
      "loss": 0.0628,
      "step": 83680
    },
    {
      "epoch": 32.5,
      "learning_rate": 1.5666537216828482e-05,
      "loss": 0.0468,
      "step": 83690
    },
    {
      "epoch": 32.5,
      "learning_rate": 1.566601941747573e-05,
      "loss": 0.1043,
      "step": 83700
    },
    {
      "epoch": 32.51,
      "learning_rate": 1.5665501618122978e-05,
      "loss": 0.0296,
      "step": 83710
    },
    {
      "epoch": 32.51,
      "learning_rate": 1.5664983818770226e-05,
      "loss": 0.0365,
      "step": 83720
    },
    {
      "epoch": 32.52,
      "learning_rate": 1.5664466019417477e-05,
      "loss": 0.0523,
      "step": 83730
    },
    {
      "epoch": 32.52,
      "learning_rate": 1.5663948220064725e-05,
      "loss": 0.2144,
      "step": 83740
    },
    {
      "epoch": 32.52,
      "learning_rate": 1.5663430420711977e-05,
      "loss": 0.105,
      "step": 83750
    },
    {
      "epoch": 32.53,
      "learning_rate": 1.5662912621359224e-05,
      "loss": 0.0931,
      "step": 83760
    },
    {
      "epoch": 32.53,
      "learning_rate": 1.5662394822006476e-05,
      "loss": 0.2388,
      "step": 83770
    },
    {
      "epoch": 32.54,
      "learning_rate": 1.566187702265372e-05,
      "loss": 0.1017,
      "step": 83780
    },
    {
      "epoch": 32.54,
      "learning_rate": 1.566135922330097e-05,
      "loss": 0.2607,
      "step": 83790
    },
    {
      "epoch": 32.54,
      "learning_rate": 1.566084142394822e-05,
      "loss": 0.1171,
      "step": 83800
    },
    {
      "epoch": 32.55,
      "learning_rate": 1.566032362459547e-05,
      "loss": 0.15,
      "step": 83810
    },
    {
      "epoch": 32.55,
      "learning_rate": 1.565980582524272e-05,
      "loss": 0.0724,
      "step": 83820
    },
    {
      "epoch": 32.56,
      "learning_rate": 1.565928802588997e-05,
      "loss": 0.1723,
      "step": 83830
    },
    {
      "epoch": 32.56,
      "learning_rate": 1.5658770226537218e-05,
      "loss": 0.2702,
      "step": 83840
    },
    {
      "epoch": 32.56,
      "learning_rate": 1.565825242718447e-05,
      "loss": 0.124,
      "step": 83850
    },
    {
      "epoch": 32.57,
      "learning_rate": 1.5657734627831714e-05,
      "loss": 0.0898,
      "step": 83860
    },
    {
      "epoch": 32.57,
      "learning_rate": 1.5657216828478965e-05,
      "loss": 0.0312,
      "step": 83870
    },
    {
      "epoch": 32.57,
      "learning_rate": 1.5656699029126213e-05,
      "loss": 0.1701,
      "step": 83880
    },
    {
      "epoch": 32.58,
      "learning_rate": 1.5656181229773464e-05,
      "loss": 0.193,
      "step": 83890
    },
    {
      "epoch": 32.58,
      "learning_rate": 1.5655663430420712e-05,
      "loss": 0.0892,
      "step": 83900
    },
    {
      "epoch": 32.59,
      "learning_rate": 1.5655145631067964e-05,
      "loss": 0.0333,
      "step": 83910
    },
    {
      "epoch": 32.59,
      "learning_rate": 1.565462783171521e-05,
      "loss": 0.1807,
      "step": 83920
    },
    {
      "epoch": 32.59,
      "learning_rate": 1.5654110032362463e-05,
      "loss": 0.1223,
      "step": 83930
    },
    {
      "epoch": 32.6,
      "learning_rate": 1.5653592233009707e-05,
      "loss": 0.0346,
      "step": 83940
    },
    {
      "epoch": 32.6,
      "learning_rate": 1.565307443365696e-05,
      "loss": 0.1024,
      "step": 83950
    },
    {
      "epoch": 32.61,
      "learning_rate": 1.5652556634304207e-05,
      "loss": 0.0647,
      "step": 83960
    },
    {
      "epoch": 32.61,
      "learning_rate": 1.5652038834951458e-05,
      "loss": 0.0569,
      "step": 83970
    },
    {
      "epoch": 32.61,
      "learning_rate": 1.5651521035598706e-05,
      "loss": 0.1207,
      "step": 83980
    },
    {
      "epoch": 32.62,
      "learning_rate": 1.5651003236245957e-05,
      "loss": 0.1328,
      "step": 83990
    },
    {
      "epoch": 32.62,
      "learning_rate": 1.5650485436893205e-05,
      "loss": 0.077,
      "step": 84000
    },
    {
      "epoch": 32.63,
      "learning_rate": 1.5649967637540456e-05,
      "loss": 0.1139,
      "step": 84010
    },
    {
      "epoch": 32.63,
      "learning_rate": 1.56494498381877e-05,
      "loss": 0.0838,
      "step": 84020
    },
    {
      "epoch": 32.63,
      "learning_rate": 1.5648932038834952e-05,
      "loss": 0.0298,
      "step": 84030
    },
    {
      "epoch": 32.64,
      "learning_rate": 1.56484142394822e-05,
      "loss": 0.0643,
      "step": 84040
    },
    {
      "epoch": 32.64,
      "learning_rate": 1.564789644012945e-05,
      "loss": 0.0666,
      "step": 84050
    },
    {
      "epoch": 32.64,
      "learning_rate": 1.56473786407767e-05,
      "loss": 0.083,
      "step": 84060
    },
    {
      "epoch": 32.65,
      "learning_rate": 1.564686084142395e-05,
      "loss": 0.1468,
      "step": 84070
    },
    {
      "epoch": 32.65,
      "learning_rate": 1.56463430420712e-05,
      "loss": 0.2002,
      "step": 84080
    },
    {
      "epoch": 32.66,
      "learning_rate": 1.564582524271845e-05,
      "loss": 0.1381,
      "step": 84090
    },
    {
      "epoch": 32.66,
      "learning_rate": 1.5645307443365698e-05,
      "loss": 0.1215,
      "step": 84100
    },
    {
      "epoch": 32.66,
      "learning_rate": 1.5644789644012946e-05,
      "loss": 0.2945,
      "step": 84110
    },
    {
      "epoch": 32.67,
      "learning_rate": 1.5644271844660194e-05,
      "loss": 0.0361,
      "step": 84120
    },
    {
      "epoch": 32.67,
      "learning_rate": 1.5643754045307445e-05,
      "loss": 0.0887,
      "step": 84130
    },
    {
      "epoch": 32.68,
      "learning_rate": 1.5643236245954693e-05,
      "loss": 0.1044,
      "step": 84140
    },
    {
      "epoch": 32.68,
      "learning_rate": 1.5642718446601944e-05,
      "loss": 0.118,
      "step": 84150
    },
    {
      "epoch": 32.68,
      "learning_rate": 1.5642200647249192e-05,
      "loss": 0.1307,
      "step": 84160
    },
    {
      "epoch": 32.69,
      "learning_rate": 1.5641682847896444e-05,
      "loss": 0.0094,
      "step": 84170
    },
    {
      "epoch": 32.69,
      "learning_rate": 1.564116504854369e-05,
      "loss": 0.2218,
      "step": 84180
    },
    {
      "epoch": 32.7,
      "learning_rate": 1.564064724919094e-05,
      "loss": 0.2493,
      "step": 84190
    },
    {
      "epoch": 32.7,
      "learning_rate": 1.5640129449838187e-05,
      "loss": 0.0933,
      "step": 84200
    },
    {
      "epoch": 32.7,
      "learning_rate": 1.563961165048544e-05,
      "loss": 0.0463,
      "step": 84210
    },
    {
      "epoch": 32.71,
      "learning_rate": 1.5639093851132687e-05,
      "loss": 0.1694,
      "step": 84220
    },
    {
      "epoch": 32.71,
      "learning_rate": 1.5638576051779938e-05,
      "loss": 0.1059,
      "step": 84230
    },
    {
      "epoch": 32.71,
      "learning_rate": 1.5638058252427186e-05,
      "loss": 0.1331,
      "step": 84240
    },
    {
      "epoch": 32.72,
      "learning_rate": 1.5637540453074437e-05,
      "loss": 0.0403,
      "step": 84250
    },
    {
      "epoch": 32.72,
      "learning_rate": 1.5637022653721685e-05,
      "loss": 0.0669,
      "step": 84260
    },
    {
      "epoch": 32.73,
      "learning_rate": 1.5636504854368933e-05,
      "loss": 0.0785,
      "step": 84270
    },
    {
      "epoch": 32.73,
      "learning_rate": 1.563598705501618e-05,
      "loss": 0.115,
      "step": 84280
    },
    {
      "epoch": 32.73,
      "learning_rate": 1.5635469255663432e-05,
      "loss": 0.0087,
      "step": 84290
    },
    {
      "epoch": 32.74,
      "learning_rate": 1.563495145631068e-05,
      "loss": 0.1416,
      "step": 84300
    },
    {
      "epoch": 32.74,
      "learning_rate": 1.563443365695793e-05,
      "loss": 0.024,
      "step": 84310
    },
    {
      "epoch": 32.75,
      "learning_rate": 1.563391585760518e-05,
      "loss": 0.0388,
      "step": 84320
    },
    {
      "epoch": 32.75,
      "learning_rate": 1.563339805825243e-05,
      "loss": 0.1352,
      "step": 84330
    },
    {
      "epoch": 32.75,
      "learning_rate": 1.563288025889968e-05,
      "loss": 0.1082,
      "step": 84340
    },
    {
      "epoch": 32.76,
      "learning_rate": 1.5632362459546927e-05,
      "loss": 0.1232,
      "step": 84350
    },
    {
      "epoch": 32.76,
      "learning_rate": 1.5631844660194174e-05,
      "loss": 0.1063,
      "step": 84360
    },
    {
      "epoch": 32.77,
      "learning_rate": 1.5631326860841426e-05,
      "loss": 0.0881,
      "step": 84370
    },
    {
      "epoch": 32.77,
      "learning_rate": 1.5630809061488674e-05,
      "loss": 0.0398,
      "step": 84380
    },
    {
      "epoch": 32.77,
      "learning_rate": 1.5630291262135925e-05,
      "loss": 0.1241,
      "step": 84390
    },
    {
      "epoch": 32.78,
      "learning_rate": 1.5629773462783173e-05,
      "loss": 0.1312,
      "step": 84400
    },
    {
      "epoch": 32.78,
      "learning_rate": 1.5629255663430424e-05,
      "loss": 0.0044,
      "step": 84410
    },
    {
      "epoch": 32.78,
      "learning_rate": 1.5628737864077672e-05,
      "loss": 0.0909,
      "step": 84420
    },
    {
      "epoch": 32.79,
      "learning_rate": 1.562822006472492e-05,
      "loss": 0.1683,
      "step": 84430
    },
    {
      "epoch": 32.79,
      "learning_rate": 1.5627702265372168e-05,
      "loss": 0.147,
      "step": 84440
    },
    {
      "epoch": 32.8,
      "learning_rate": 1.562718446601942e-05,
      "loss": 0.0899,
      "step": 84450
    },
    {
      "epoch": 32.8,
      "learning_rate": 1.5626666666666667e-05,
      "loss": 0.1836,
      "step": 84460
    },
    {
      "epoch": 32.8,
      "learning_rate": 1.562614886731392e-05,
      "loss": 0.1717,
      "step": 84470
    },
    {
      "epoch": 32.81,
      "learning_rate": 1.5625631067961166e-05,
      "loss": 0.0735,
      "step": 84480
    },
    {
      "epoch": 32.81,
      "learning_rate": 1.5625113268608418e-05,
      "loss": 0.0494,
      "step": 84490
    },
    {
      "epoch": 32.82,
      "learning_rate": 1.5624595469255666e-05,
      "loss": 0.0856,
      "step": 84500
    },
    {
      "epoch": 32.82,
      "learning_rate": 1.5624077669902914e-05,
      "loss": 0.1731,
      "step": 84510
    },
    {
      "epoch": 32.82,
      "learning_rate": 1.562355987055016e-05,
      "loss": 0.124,
      "step": 84520
    },
    {
      "epoch": 32.83,
      "learning_rate": 1.5623042071197413e-05,
      "loss": 0.1261,
      "step": 84530
    },
    {
      "epoch": 32.83,
      "learning_rate": 1.562252427184466e-05,
      "loss": 0.0194,
      "step": 84540
    },
    {
      "epoch": 32.83,
      "learning_rate": 1.5622006472491912e-05,
      "loss": 0.0407,
      "step": 84550
    },
    {
      "epoch": 32.84,
      "learning_rate": 1.562148867313916e-05,
      "loss": 0.1062,
      "step": 84560
    },
    {
      "epoch": 32.84,
      "learning_rate": 1.5620970873786408e-05,
      "loss": 0.0873,
      "step": 84570
    },
    {
      "epoch": 32.85,
      "learning_rate": 1.562045307443366e-05,
      "loss": 0.0608,
      "step": 84580
    },
    {
      "epoch": 32.85,
      "learning_rate": 1.5619935275080907e-05,
      "loss": 0.0723,
      "step": 84590
    },
    {
      "epoch": 32.85,
      "learning_rate": 1.5619417475728155e-05,
      "loss": 0.0009,
      "step": 84600
    },
    {
      "epoch": 32.86,
      "learning_rate": 1.5618899676375406e-05,
      "loss": 0.0835,
      "step": 84610
    },
    {
      "epoch": 32.86,
      "learning_rate": 1.5618381877022654e-05,
      "loss": 0.207,
      "step": 84620
    },
    {
      "epoch": 32.87,
      "learning_rate": 1.5617864077669906e-05,
      "loss": 0.084,
      "step": 84630
    },
    {
      "epoch": 32.87,
      "learning_rate": 1.5617346278317154e-05,
      "loss": 0.0788,
      "step": 84640
    },
    {
      "epoch": 32.87,
      "learning_rate": 1.56168284789644e-05,
      "loss": 0.1668,
      "step": 84650
    },
    {
      "epoch": 32.88,
      "learning_rate": 1.5616310679611653e-05,
      "loss": 0.2773,
      "step": 84660
    },
    {
      "epoch": 32.88,
      "learning_rate": 1.56157928802589e-05,
      "loss": 0.1947,
      "step": 84670
    },
    {
      "epoch": 32.89,
      "learning_rate": 1.561527508090615e-05,
      "loss": 0.1773,
      "step": 84680
    },
    {
      "epoch": 32.89,
      "learning_rate": 1.56147572815534e-05,
      "loss": 0.0846,
      "step": 84690
    },
    {
      "epoch": 32.89,
      "learning_rate": 1.5614239482200648e-05,
      "loss": 0.0955,
      "step": 84700
    },
    {
      "epoch": 32.9,
      "learning_rate": 1.56137216828479e-05,
      "loss": 0.1581,
      "step": 84710
    },
    {
      "epoch": 32.9,
      "learning_rate": 1.5613203883495147e-05,
      "loss": 0.1885,
      "step": 84720
    },
    {
      "epoch": 32.9,
      "learning_rate": 1.5612686084142395e-05,
      "loss": 0.0497,
      "step": 84730
    },
    {
      "epoch": 32.91,
      "learning_rate": 1.5612168284789646e-05,
      "loss": 0.0427,
      "step": 84740
    },
    {
      "epoch": 32.91,
      "learning_rate": 1.5611650485436894e-05,
      "loss": 0.1476,
      "step": 84750
    },
    {
      "epoch": 32.92,
      "learning_rate": 1.5611132686084142e-05,
      "loss": 0.0444,
      "step": 84760
    },
    {
      "epoch": 32.92,
      "learning_rate": 1.5610614886731394e-05,
      "loss": 0.0544,
      "step": 84770
    },
    {
      "epoch": 32.92,
      "learning_rate": 1.561009708737864e-05,
      "loss": 0.1773,
      "step": 84780
    },
    {
      "epoch": 32.93,
      "learning_rate": 1.5609579288025893e-05,
      "loss": 0.0774,
      "step": 84790
    },
    {
      "epoch": 32.93,
      "learning_rate": 1.560906148867314e-05,
      "loss": 0.0431,
      "step": 84800
    },
    {
      "epoch": 32.94,
      "learning_rate": 1.560854368932039e-05,
      "loss": 0.0304,
      "step": 84810
    },
    {
      "epoch": 32.94,
      "learning_rate": 1.560802588996764e-05,
      "loss": 0.1357,
      "step": 84820
    },
    {
      "epoch": 32.94,
      "learning_rate": 1.5607508090614888e-05,
      "loss": 0.0772,
      "step": 84830
    },
    {
      "epoch": 32.95,
      "learning_rate": 1.5606990291262136e-05,
      "loss": 0.0906,
      "step": 84840
    },
    {
      "epoch": 32.95,
      "learning_rate": 1.5606472491909387e-05,
      "loss": 0.0014,
      "step": 84850
    },
    {
      "epoch": 32.96,
      "learning_rate": 1.5605954692556635e-05,
      "loss": 0.1284,
      "step": 84860
    },
    {
      "epoch": 32.96,
      "learning_rate": 1.5605436893203886e-05,
      "loss": 0.0182,
      "step": 84870
    },
    {
      "epoch": 32.96,
      "learning_rate": 1.5604919093851134e-05,
      "loss": 0.1075,
      "step": 84880
    },
    {
      "epoch": 32.97,
      "learning_rate": 1.5604401294498382e-05,
      "loss": 0.025,
      "step": 84890
    },
    {
      "epoch": 32.97,
      "learning_rate": 1.5603883495145634e-05,
      "loss": 0.0865,
      "step": 84900
    },
    {
      "epoch": 32.97,
      "learning_rate": 1.560336569579288e-05,
      "loss": 0.0605,
      "step": 84910
    },
    {
      "epoch": 32.98,
      "learning_rate": 1.560284789644013e-05,
      "loss": 0.1006,
      "step": 84920
    },
    {
      "epoch": 32.98,
      "learning_rate": 1.560233009708738e-05,
      "loss": 0.0724,
      "step": 84930
    },
    {
      "epoch": 32.99,
      "learning_rate": 1.560181229773463e-05,
      "loss": 0.1276,
      "step": 84940
    },
    {
      "epoch": 32.99,
      "learning_rate": 1.5601294498381877e-05,
      "loss": 0.0629,
      "step": 84950
    },
    {
      "epoch": 32.99,
      "learning_rate": 1.5600776699029128e-05,
      "loss": 0.0592,
      "step": 84960
    },
    {
      "epoch": 33.0,
      "learning_rate": 1.5600258899676376e-05,
      "loss": 0.1774,
      "step": 84970
    },
    {
      "epoch": 33.0,
      "eval_accuracy": 0.9548830811554333,
      "eval_loss": 0.27493834495544434,
      "eval_runtime": 8.2212,
      "eval_samples_per_second": 442.149,
      "eval_steps_per_second": 55.345,
      "step": 84975
    },
    {
      "epoch": 33.0,
      "learning_rate": 1.5599741100323627e-05,
      "loss": 0.0373,
      "step": 84980
    },
    {
      "epoch": 33.01,
      "learning_rate": 1.5599223300970875e-05,
      "loss": 0.0424,
      "step": 84990
    },
    {
      "epoch": 33.01,
      "learning_rate": 1.5598705501618123e-05,
      "loss": 0.1049,
      "step": 85000
    },
    {
      "epoch": 33.01,
      "learning_rate": 1.5598187702265374e-05,
      "loss": 0.0313,
      "step": 85010
    },
    {
      "epoch": 33.02,
      "learning_rate": 1.5597669902912622e-05,
      "loss": 0.1459,
      "step": 85020
    },
    {
      "epoch": 33.02,
      "learning_rate": 1.559715210355987e-05,
      "loss": 0.1136,
      "step": 85030
    },
    {
      "epoch": 33.03,
      "learning_rate": 1.559663430420712e-05,
      "loss": 0.1216,
      "step": 85040
    },
    {
      "epoch": 33.03,
      "learning_rate": 1.559611650485437e-05,
      "loss": 0.0457,
      "step": 85050
    },
    {
      "epoch": 33.03,
      "learning_rate": 1.559559870550162e-05,
      "loss": 0.0329,
      "step": 85060
    },
    {
      "epoch": 33.04,
      "learning_rate": 1.559508090614887e-05,
      "loss": 0.0442,
      "step": 85070
    },
    {
      "epoch": 33.04,
      "learning_rate": 1.5594563106796116e-05,
      "loss": 0.0851,
      "step": 85080
    },
    {
      "epoch": 33.04,
      "learning_rate": 1.5594045307443368e-05,
      "loss": 0.1049,
      "step": 85090
    },
    {
      "epoch": 33.05,
      "learning_rate": 1.5593527508090616e-05,
      "loss": 0.1549,
      "step": 85100
    },
    {
      "epoch": 33.05,
      "learning_rate": 1.5593009708737864e-05,
      "loss": 0.1129,
      "step": 85110
    },
    {
      "epoch": 33.06,
      "learning_rate": 1.5592491909385115e-05,
      "loss": 0.1249,
      "step": 85120
    },
    {
      "epoch": 33.06,
      "learning_rate": 1.5591974110032363e-05,
      "loss": 0.0785,
      "step": 85130
    },
    {
      "epoch": 33.06,
      "learning_rate": 1.5591456310679614e-05,
      "loss": 0.1888,
      "step": 85140
    },
    {
      "epoch": 33.07,
      "learning_rate": 1.5590938511326862e-05,
      "loss": 0.0239,
      "step": 85150
    },
    {
      "epoch": 33.07,
      "learning_rate": 1.559042071197411e-05,
      "loss": 0.1023,
      "step": 85160
    },
    {
      "epoch": 33.08,
      "learning_rate": 1.558990291262136e-05,
      "loss": 0.0238,
      "step": 85170
    },
    {
      "epoch": 33.08,
      "learning_rate": 1.558938511326861e-05,
      "loss": 0.0429,
      "step": 85180
    },
    {
      "epoch": 33.08,
      "learning_rate": 1.5588867313915857e-05,
      "loss": 0.034,
      "step": 85190
    },
    {
      "epoch": 33.09,
      "learning_rate": 1.558834951456311e-05,
      "loss": 0.0933,
      "step": 85200
    },
    {
      "epoch": 33.09,
      "learning_rate": 1.5587831715210356e-05,
      "loss": 0.0275,
      "step": 85210
    },
    {
      "epoch": 33.1,
      "learning_rate": 1.5587313915857608e-05,
      "loss": 0.1023,
      "step": 85220
    },
    {
      "epoch": 33.1,
      "learning_rate": 1.5586796116504856e-05,
      "loss": 0.0137,
      "step": 85230
    },
    {
      "epoch": 33.1,
      "learning_rate": 1.5586278317152107e-05,
      "loss": 0.0795,
      "step": 85240
    },
    {
      "epoch": 33.11,
      "learning_rate": 1.558576051779935e-05,
      "loss": 0.1196,
      "step": 85250
    },
    {
      "epoch": 33.11,
      "learning_rate": 1.5585242718446603e-05,
      "loss": 0.1008,
      "step": 85260
    },
    {
      "epoch": 33.11,
      "learning_rate": 1.558472491909385e-05,
      "loss": 0.0556,
      "step": 85270
    },
    {
      "epoch": 33.12,
      "learning_rate": 1.5584207119741102e-05,
      "loss": 0.026,
      "step": 85280
    },
    {
      "epoch": 33.12,
      "learning_rate": 1.558368932038835e-05,
      "loss": 0.026,
      "step": 85290
    },
    {
      "epoch": 33.13,
      "learning_rate": 1.55831715210356e-05,
      "loss": 0.0679,
      "step": 85300
    },
    {
      "epoch": 33.13,
      "learning_rate": 1.558265372168285e-05,
      "loss": 0.0833,
      "step": 85310
    },
    {
      "epoch": 33.13,
      "learning_rate": 1.55821359223301e-05,
      "loss": 0.0618,
      "step": 85320
    },
    {
      "epoch": 33.14,
      "learning_rate": 1.5581618122977345e-05,
      "loss": 0.1821,
      "step": 85330
    },
    {
      "epoch": 33.14,
      "learning_rate": 1.5581100323624596e-05,
      "loss": 0.0481,
      "step": 85340
    },
    {
      "epoch": 33.15,
      "learning_rate": 1.5580582524271844e-05,
      "loss": 0.0526,
      "step": 85350
    },
    {
      "epoch": 33.15,
      "learning_rate": 1.5580064724919096e-05,
      "loss": 0.2016,
      "step": 85360
    },
    {
      "epoch": 33.15,
      "learning_rate": 1.5579546925566344e-05,
      "loss": 0.0906,
      "step": 85370
    },
    {
      "epoch": 33.16,
      "learning_rate": 1.5579029126213595e-05,
      "loss": 0.1791,
      "step": 85380
    },
    {
      "epoch": 33.16,
      "learning_rate": 1.5578511326860843e-05,
      "loss": 0.1076,
      "step": 85390
    },
    {
      "epoch": 33.17,
      "learning_rate": 1.5577993527508094e-05,
      "loss": 0.1211,
      "step": 85400
    },
    {
      "epoch": 33.17,
      "learning_rate": 1.557747572815534e-05,
      "loss": 0.0944,
      "step": 85410
    },
    {
      "epoch": 33.17,
      "learning_rate": 1.557695792880259e-05,
      "loss": 0.1447,
      "step": 85420
    },
    {
      "epoch": 33.18,
      "learning_rate": 1.5576440129449838e-05,
      "loss": 0.069,
      "step": 85430
    },
    {
      "epoch": 33.18,
      "learning_rate": 1.557592233009709e-05,
      "loss": 0.1043,
      "step": 85440
    },
    {
      "epoch": 33.18,
      "learning_rate": 1.5575404530744337e-05,
      "loss": 0.1669,
      "step": 85450
    },
    {
      "epoch": 33.19,
      "learning_rate": 1.557488673139159e-05,
      "loss": 0.0318,
      "step": 85460
    },
    {
      "epoch": 33.19,
      "learning_rate": 1.5574368932038836e-05,
      "loss": 0.009,
      "step": 85470
    },
    {
      "epoch": 33.2,
      "learning_rate": 1.5573851132686088e-05,
      "loss": 0.0065,
      "step": 85480
    },
    {
      "epoch": 33.2,
      "learning_rate": 1.5573333333333332e-05,
      "loss": 0.1104,
      "step": 85490
    },
    {
      "epoch": 33.2,
      "learning_rate": 1.5572815533980583e-05,
      "loss": 0.1277,
      "step": 85500
    },
    {
      "epoch": 33.21,
      "learning_rate": 1.557229773462783e-05,
      "loss": 0.0874,
      "step": 85510
    },
    {
      "epoch": 33.21,
      "learning_rate": 1.5571779935275083e-05,
      "loss": 0.2378,
      "step": 85520
    },
    {
      "epoch": 33.22,
      "learning_rate": 1.557126213592233e-05,
      "loss": 0.1852,
      "step": 85530
    },
    {
      "epoch": 33.22,
      "learning_rate": 1.5570744336569582e-05,
      "loss": 0.1476,
      "step": 85540
    },
    {
      "epoch": 33.22,
      "learning_rate": 1.557022653721683e-05,
      "loss": 0.1518,
      "step": 85550
    },
    {
      "epoch": 33.23,
      "learning_rate": 1.556970873786408e-05,
      "loss": 0.0617,
      "step": 85560
    },
    {
      "epoch": 33.23,
      "learning_rate": 1.5569190938511326e-05,
      "loss": 0.0284,
      "step": 85570
    },
    {
      "epoch": 33.23,
      "learning_rate": 1.5568673139158577e-05,
      "loss": 0.0453,
      "step": 85580
    },
    {
      "epoch": 33.24,
      "learning_rate": 1.5568155339805825e-05,
      "loss": 0.1235,
      "step": 85590
    },
    {
      "epoch": 33.24,
      "learning_rate": 1.5567637540453076e-05,
      "loss": 0.0505,
      "step": 85600
    },
    {
      "epoch": 33.25,
      "learning_rate": 1.5567119741100324e-05,
      "loss": 0.0432,
      "step": 85610
    },
    {
      "epoch": 33.25,
      "learning_rate": 1.5566601941747576e-05,
      "loss": 0.0729,
      "step": 85620
    },
    {
      "epoch": 33.25,
      "learning_rate": 1.5566084142394823e-05,
      "loss": 0.1585,
      "step": 85630
    },
    {
      "epoch": 33.26,
      "learning_rate": 1.5565566343042075e-05,
      "loss": 0.1454,
      "step": 85640
    },
    {
      "epoch": 33.26,
      "learning_rate": 1.556504854368932e-05,
      "loss": 0.0375,
      "step": 85650
    },
    {
      "epoch": 33.27,
      "learning_rate": 1.556453074433657e-05,
      "loss": 0.1341,
      "step": 85660
    },
    {
      "epoch": 33.27,
      "learning_rate": 1.556401294498382e-05,
      "loss": 0.0855,
      "step": 85670
    },
    {
      "epoch": 33.27,
      "learning_rate": 1.556349514563107e-05,
      "loss": 0.111,
      "step": 85680
    },
    {
      "epoch": 33.28,
      "learning_rate": 1.5562977346278318e-05,
      "loss": 0.176,
      "step": 85690
    },
    {
      "epoch": 33.28,
      "learning_rate": 1.556245954692557e-05,
      "loss": 0.0348,
      "step": 85700
    },
    {
      "epoch": 33.29,
      "learning_rate": 1.5561941747572817e-05,
      "loss": 0.2362,
      "step": 85710
    },
    {
      "epoch": 33.29,
      "learning_rate": 1.556142394822007e-05,
      "loss": 0.0729,
      "step": 85720
    },
    {
      "epoch": 33.29,
      "learning_rate": 1.5560906148867313e-05,
      "loss": 0.0826,
      "step": 85730
    },
    {
      "epoch": 33.3,
      "learning_rate": 1.5560388349514564e-05,
      "loss": 0.0402,
      "step": 85740
    },
    {
      "epoch": 33.3,
      "learning_rate": 1.5559870550161812e-05,
      "loss": 0.1061,
      "step": 85750
    },
    {
      "epoch": 33.3,
      "learning_rate": 1.5559352750809063e-05,
      "loss": 0.1516,
      "step": 85760
    },
    {
      "epoch": 33.31,
      "learning_rate": 1.555883495145631e-05,
      "loss": 0.1515,
      "step": 85770
    },
    {
      "epoch": 33.31,
      "learning_rate": 1.5558317152103563e-05,
      "loss": 0.1499,
      "step": 85780
    },
    {
      "epoch": 33.32,
      "learning_rate": 1.555779935275081e-05,
      "loss": 0.2311,
      "step": 85790
    },
    {
      "epoch": 33.32,
      "learning_rate": 1.5557281553398062e-05,
      "loss": 0.1833,
      "step": 85800
    },
    {
      "epoch": 33.32,
      "learning_rate": 1.555676375404531e-05,
      "loss": 0.1761,
      "step": 85810
    },
    {
      "epoch": 33.33,
      "learning_rate": 1.5556245954692558e-05,
      "loss": 0.1002,
      "step": 85820
    },
    {
      "epoch": 33.33,
      "learning_rate": 1.5555728155339806e-05,
      "loss": 0.0739,
      "step": 85830
    },
    {
      "epoch": 33.34,
      "learning_rate": 1.5555210355987057e-05,
      "loss": 0.0132,
      "step": 85840
    },
    {
      "epoch": 33.34,
      "learning_rate": 1.5554692556634305e-05,
      "loss": 0.1034,
      "step": 85850
    },
    {
      "epoch": 33.34,
      "learning_rate": 1.5554174757281556e-05,
      "loss": 0.1244,
      "step": 85860
    },
    {
      "epoch": 33.35,
      "learning_rate": 1.5553656957928804e-05,
      "loss": 0.0664,
      "step": 85870
    },
    {
      "epoch": 33.35,
      "learning_rate": 1.5553139158576055e-05,
      "loss": 0.0879,
      "step": 85880
    },
    {
      "epoch": 33.36,
      "learning_rate": 1.5552621359223303e-05,
      "loss": 0.0524,
      "step": 85890
    },
    {
      "epoch": 33.36,
      "learning_rate": 1.555210355987055e-05,
      "loss": 0.1575,
      "step": 85900
    },
    {
      "epoch": 33.36,
      "learning_rate": 1.55515857605178e-05,
      "loss": 0.0156,
      "step": 85910
    },
    {
      "epoch": 33.37,
      "learning_rate": 1.555106796116505e-05,
      "loss": 0.1609,
      "step": 85920
    },
    {
      "epoch": 33.37,
      "learning_rate": 1.55505501618123e-05,
      "loss": 0.0701,
      "step": 85930
    },
    {
      "epoch": 33.37,
      "learning_rate": 1.555003236245955e-05,
      "loss": 0.0657,
      "step": 85940
    },
    {
      "epoch": 33.38,
      "learning_rate": 1.5549514563106798e-05,
      "loss": 0.0672,
      "step": 85950
    },
    {
      "epoch": 33.38,
      "learning_rate": 1.554899676375405e-05,
      "loss": 0.0806,
      "step": 85960
    },
    {
      "epoch": 33.39,
      "learning_rate": 1.5548478964401297e-05,
      "loss": 0.0696,
      "step": 85970
    },
    {
      "epoch": 33.39,
      "learning_rate": 1.5547961165048545e-05,
      "loss": 0.0827,
      "step": 85980
    },
    {
      "epoch": 33.39,
      "learning_rate": 1.5547443365695793e-05,
      "loss": 0.1087,
      "step": 85990
    },
    {
      "epoch": 33.4,
      "learning_rate": 1.5546925566343044e-05,
      "loss": 0.0536,
      "step": 86000
    },
    {
      "epoch": 33.4,
      "learning_rate": 1.5546407766990292e-05,
      "loss": 0.0794,
      "step": 86010
    },
    {
      "epoch": 33.41,
      "learning_rate": 1.5545889967637543e-05,
      "loss": 0.0234,
      "step": 86020
    },
    {
      "epoch": 33.41,
      "learning_rate": 1.554537216828479e-05,
      "loss": 0.0668,
      "step": 86030
    },
    {
      "epoch": 33.41,
      "learning_rate": 1.554485436893204e-05,
      "loss": 0.0096,
      "step": 86040
    },
    {
      "epoch": 33.42,
      "learning_rate": 1.554433656957929e-05,
      "loss": 0.0759,
      "step": 86050
    },
    {
      "epoch": 33.42,
      "learning_rate": 1.554381877022654e-05,
      "loss": 0.1724,
      "step": 86060
    },
    {
      "epoch": 33.43,
      "learning_rate": 1.5543300970873786e-05,
      "loss": 0.0943,
      "step": 86070
    },
    {
      "epoch": 33.43,
      "learning_rate": 1.5542783171521038e-05,
      "loss": 0.0636,
      "step": 86080
    },
    {
      "epoch": 33.43,
      "learning_rate": 1.5542265372168286e-05,
      "loss": 0.033,
      "step": 86090
    },
    {
      "epoch": 33.44,
      "learning_rate": 1.5541747572815537e-05,
      "loss": 0.0223,
      "step": 86100
    },
    {
      "epoch": 33.44,
      "learning_rate": 1.5541229773462785e-05,
      "loss": 0.1107,
      "step": 86110
    },
    {
      "epoch": 33.44,
      "learning_rate": 1.5540711974110033e-05,
      "loss": 0.1048,
      "step": 86120
    },
    {
      "epoch": 33.45,
      "learning_rate": 1.5540194174757284e-05,
      "loss": 0.0737,
      "step": 86130
    },
    {
      "epoch": 33.45,
      "learning_rate": 1.5539676375404532e-05,
      "loss": 0.1697,
      "step": 86140
    },
    {
      "epoch": 33.46,
      "learning_rate": 1.553915857605178e-05,
      "loss": 0.0522,
      "step": 86150
    },
    {
      "epoch": 33.46,
      "learning_rate": 1.553864077669903e-05,
      "loss": 0.0388,
      "step": 86160
    },
    {
      "epoch": 33.46,
      "learning_rate": 1.553812297734628e-05,
      "loss": 0.0006,
      "step": 86170
    },
    {
      "epoch": 33.47,
      "learning_rate": 1.553760517799353e-05,
      "loss": 0.0678,
      "step": 86180
    },
    {
      "epoch": 33.47,
      "learning_rate": 1.553708737864078e-05,
      "loss": 0.1789,
      "step": 86190
    },
    {
      "epoch": 33.48,
      "learning_rate": 1.5536569579288026e-05,
      "loss": 0.1003,
      "step": 86200
    },
    {
      "epoch": 33.48,
      "learning_rate": 1.5536051779935278e-05,
      "loss": 0.139,
      "step": 86210
    },
    {
      "epoch": 33.48,
      "learning_rate": 1.5535533980582525e-05,
      "loss": 0.0238,
      "step": 86220
    },
    {
      "epoch": 33.49,
      "learning_rate": 1.5535016181229773e-05,
      "loss": 0.0867,
      "step": 86230
    },
    {
      "epoch": 33.49,
      "learning_rate": 1.5534498381877025e-05,
      "loss": 0.1994,
      "step": 86240
    },
    {
      "epoch": 33.5,
      "learning_rate": 1.5533980582524273e-05,
      "loss": 0.058,
      "step": 86250
    },
    {
      "epoch": 33.5,
      "learning_rate": 1.5533462783171524e-05,
      "loss": 0.0705,
      "step": 86260
    },
    {
      "epoch": 33.5,
      "learning_rate": 1.5532944983818772e-05,
      "loss": 0.05,
      "step": 86270
    },
    {
      "epoch": 33.51,
      "learning_rate": 1.553242718446602e-05,
      "loss": 0.0046,
      "step": 86280
    },
    {
      "epoch": 33.51,
      "learning_rate": 1.553190938511327e-05,
      "loss": 0.0955,
      "step": 86290
    },
    {
      "epoch": 33.51,
      "learning_rate": 1.553139158576052e-05,
      "loss": 0.1053,
      "step": 86300
    },
    {
      "epoch": 33.52,
      "learning_rate": 1.5530873786407767e-05,
      "loss": 0.021,
      "step": 86310
    },
    {
      "epoch": 33.52,
      "learning_rate": 1.5530355987055018e-05,
      "loss": 0.0098,
      "step": 86320
    },
    {
      "epoch": 33.53,
      "learning_rate": 1.5529838187702266e-05,
      "loss": 0.1033,
      "step": 86330
    },
    {
      "epoch": 33.53,
      "learning_rate": 1.5529320388349518e-05,
      "loss": 0.1151,
      "step": 86340
    },
    {
      "epoch": 33.53,
      "learning_rate": 1.5528802588996765e-05,
      "loss": 0.0305,
      "step": 86350
    },
    {
      "epoch": 33.54,
      "learning_rate": 1.5528284789644013e-05,
      "loss": 0.2592,
      "step": 86360
    },
    {
      "epoch": 33.54,
      "learning_rate": 1.5527766990291265e-05,
      "loss": 0.2049,
      "step": 86370
    },
    {
      "epoch": 33.55,
      "learning_rate": 1.5527249190938513e-05,
      "loss": 0.0585,
      "step": 86380
    },
    {
      "epoch": 33.55,
      "learning_rate": 1.552673139158576e-05,
      "loss": 0.0526,
      "step": 86390
    },
    {
      "epoch": 33.55,
      "learning_rate": 1.5526213592233012e-05,
      "loss": 0.0802,
      "step": 86400
    },
    {
      "epoch": 33.56,
      "learning_rate": 1.552569579288026e-05,
      "loss": 0.0706,
      "step": 86410
    },
    {
      "epoch": 33.56,
      "learning_rate": 1.5525177993527508e-05,
      "loss": 0.0709,
      "step": 86420
    },
    {
      "epoch": 33.57,
      "learning_rate": 1.552466019417476e-05,
      "loss": 0.0068,
      "step": 86430
    },
    {
      "epoch": 33.57,
      "learning_rate": 1.5524142394822007e-05,
      "loss": 0.0375,
      "step": 86440
    },
    {
      "epoch": 33.57,
      "learning_rate": 1.5523624595469258e-05,
      "loss": 0.0488,
      "step": 86450
    },
    {
      "epoch": 33.58,
      "learning_rate": 1.5523106796116506e-05,
      "loss": 0.0817,
      "step": 86460
    },
    {
      "epoch": 33.58,
      "learning_rate": 1.5522588996763754e-05,
      "loss": 0.2173,
      "step": 86470
    },
    {
      "epoch": 33.58,
      "learning_rate": 1.5522071197411005e-05,
      "loss": 0.1146,
      "step": 86480
    },
    {
      "epoch": 33.59,
      "learning_rate": 1.5521553398058253e-05,
      "loss": 0.2319,
      "step": 86490
    },
    {
      "epoch": 33.59,
      "learning_rate": 1.55210355987055e-05,
      "loss": 0.0715,
      "step": 86500
    },
    {
      "epoch": 33.6,
      "learning_rate": 1.5520517799352753e-05,
      "loss": 0.2288,
      "step": 86510
    },
    {
      "epoch": 33.6,
      "learning_rate": 1.552e-05,
      "loss": 0.1308,
      "step": 86520
    },
    {
      "epoch": 33.6,
      "learning_rate": 1.5519482200647252e-05,
      "loss": 0.1684,
      "step": 86530
    },
    {
      "epoch": 33.61,
      "learning_rate": 1.55189644012945e-05,
      "loss": 0.1251,
      "step": 86540
    },
    {
      "epoch": 33.61,
      "learning_rate": 1.5518446601941748e-05,
      "loss": 0.0574,
      "step": 86550
    },
    {
      "epoch": 33.62,
      "learning_rate": 1.5517928802589e-05,
      "loss": 0.0678,
      "step": 86560
    },
    {
      "epoch": 33.62,
      "learning_rate": 1.5517411003236247e-05,
      "loss": 0.0517,
      "step": 86570
    },
    {
      "epoch": 33.62,
      "learning_rate": 1.5516893203883495e-05,
      "loss": 0.0281,
      "step": 86580
    },
    {
      "epoch": 33.63,
      "learning_rate": 1.5516375404530746e-05,
      "loss": 0.0736,
      "step": 86590
    },
    {
      "epoch": 33.63,
      "learning_rate": 1.5515857605177994e-05,
      "loss": 0.0903,
      "step": 86600
    },
    {
      "epoch": 33.63,
      "learning_rate": 1.5515339805825245e-05,
      "loss": 0.0599,
      "step": 86610
    },
    {
      "epoch": 33.64,
      "learning_rate": 1.5514822006472493e-05,
      "loss": 0.1121,
      "step": 86620
    },
    {
      "epoch": 33.64,
      "learning_rate": 1.551430420711974e-05,
      "loss": 0.0607,
      "step": 86630
    },
    {
      "epoch": 33.65,
      "learning_rate": 1.5513786407766993e-05,
      "loss": 0.0568,
      "step": 86640
    },
    {
      "epoch": 33.65,
      "learning_rate": 1.551326860841424e-05,
      "loss": 0.0046,
      "step": 86650
    },
    {
      "epoch": 33.65,
      "learning_rate": 1.551275080906149e-05,
      "loss": 0.1076,
      "step": 86660
    },
    {
      "epoch": 33.66,
      "learning_rate": 1.551223300970874e-05,
      "loss": 0.0668,
      "step": 86670
    },
    {
      "epoch": 33.66,
      "learning_rate": 1.5511715210355988e-05,
      "loss": 0.0806,
      "step": 86680
    },
    {
      "epoch": 33.67,
      "learning_rate": 1.551119741100324e-05,
      "loss": 0.1276,
      "step": 86690
    },
    {
      "epoch": 33.67,
      "learning_rate": 1.5510679611650487e-05,
      "loss": 0.1236,
      "step": 86700
    },
    {
      "epoch": 33.67,
      "learning_rate": 1.5510161812297735e-05,
      "loss": 0.051,
      "step": 86710
    },
    {
      "epoch": 33.68,
      "learning_rate": 1.5509644012944983e-05,
      "loss": 0.0186,
      "step": 86720
    },
    {
      "epoch": 33.68,
      "learning_rate": 1.5509126213592234e-05,
      "loss": 0.1289,
      "step": 86730
    },
    {
      "epoch": 33.69,
      "learning_rate": 1.5508608414239482e-05,
      "loss": 0.165,
      "step": 86740
    },
    {
      "epoch": 33.69,
      "learning_rate": 1.5508090614886733e-05,
      "loss": 0.173,
      "step": 86750
    },
    {
      "epoch": 33.69,
      "learning_rate": 1.550757281553398e-05,
      "loss": 0.0113,
      "step": 86760
    },
    {
      "epoch": 33.7,
      "learning_rate": 1.5507055016181232e-05,
      "loss": 0.1486,
      "step": 86770
    },
    {
      "epoch": 33.7,
      "learning_rate": 1.550653721682848e-05,
      "loss": 0.004,
      "step": 86780
    },
    {
      "epoch": 33.7,
      "learning_rate": 1.550601941747573e-05,
      "loss": 0.042,
      "step": 86790
    },
    {
      "epoch": 33.71,
      "learning_rate": 1.5505501618122976e-05,
      "loss": 0.0667,
      "step": 86800
    },
    {
      "epoch": 33.71,
      "learning_rate": 1.5504983818770228e-05,
      "loss": 0.0074,
      "step": 86810
    },
    {
      "epoch": 33.72,
      "learning_rate": 1.5504466019417475e-05,
      "loss": 0.0373,
      "step": 86820
    },
    {
      "epoch": 33.72,
      "learning_rate": 1.5503948220064727e-05,
      "loss": 0.0974,
      "step": 86830
    },
    {
      "epoch": 33.72,
      "learning_rate": 1.5503430420711975e-05,
      "loss": 0.0151,
      "step": 86840
    },
    {
      "epoch": 33.73,
      "learning_rate": 1.5502912621359226e-05,
      "loss": 0.1032,
      "step": 86850
    },
    {
      "epoch": 33.73,
      "learning_rate": 1.5502394822006474e-05,
      "loss": 0.007,
      "step": 86860
    },
    {
      "epoch": 33.74,
      "learning_rate": 1.5501877022653722e-05,
      "loss": 0.0355,
      "step": 86870
    },
    {
      "epoch": 33.74,
      "learning_rate": 1.550135922330097e-05,
      "loss": 0.2135,
      "step": 86880
    },
    {
      "epoch": 33.74,
      "learning_rate": 1.550084142394822e-05,
      "loss": 0.0874,
      "step": 86890
    },
    {
      "epoch": 33.75,
      "learning_rate": 1.550032362459547e-05,
      "loss": 0.1606,
      "step": 86900
    },
    {
      "epoch": 33.75,
      "learning_rate": 1.549980582524272e-05,
      "loss": 0.0406,
      "step": 86910
    },
    {
      "epoch": 33.76,
      "learning_rate": 1.5499288025889968e-05,
      "loss": 0.0932,
      "step": 86920
    },
    {
      "epoch": 33.76,
      "learning_rate": 1.549877022653722e-05,
      "loss": 0.0961,
      "step": 86930
    },
    {
      "epoch": 33.76,
      "learning_rate": 1.5498252427184468e-05,
      "loss": 0.1804,
      "step": 86940
    },
    {
      "epoch": 33.77,
      "learning_rate": 1.549773462783172e-05,
      "loss": 0.0592,
      "step": 86950
    },
    {
      "epoch": 33.77,
      "learning_rate": 1.5497216828478963e-05,
      "loss": 0.0957,
      "step": 86960
    },
    {
      "epoch": 33.77,
      "learning_rate": 1.5496699029126215e-05,
      "loss": 0.0169,
      "step": 86970
    },
    {
      "epoch": 33.78,
      "learning_rate": 1.5496181229773463e-05,
      "loss": 0.0023,
      "step": 86980
    },
    {
      "epoch": 33.78,
      "learning_rate": 1.5495663430420714e-05,
      "loss": 0.1528,
      "step": 86990
    },
    {
      "epoch": 33.79,
      "learning_rate": 1.5495145631067962e-05,
      "loss": 0.1054,
      "step": 87000
    },
    {
      "epoch": 33.79,
      "learning_rate": 1.5494627831715213e-05,
      "loss": 0.2137,
      "step": 87010
    },
    {
      "epoch": 33.79,
      "learning_rate": 1.549411003236246e-05,
      "loss": 0.0683,
      "step": 87020
    },
    {
      "epoch": 33.8,
      "learning_rate": 1.5493592233009712e-05,
      "loss": 0.2045,
      "step": 87030
    },
    {
      "epoch": 33.8,
      "learning_rate": 1.5493074433656957e-05,
      "loss": 0.0981,
      "step": 87040
    },
    {
      "epoch": 33.81,
      "learning_rate": 1.5492556634304208e-05,
      "loss": 0.1077,
      "step": 87050
    },
    {
      "epoch": 33.81,
      "learning_rate": 1.5492038834951456e-05,
      "loss": 0.0214,
      "step": 87060
    },
    {
      "epoch": 33.81,
      "learning_rate": 1.5491521035598707e-05,
      "loss": 0.0241,
      "step": 87070
    },
    {
      "epoch": 33.82,
      "learning_rate": 1.5491003236245955e-05,
      "loss": 0.1419,
      "step": 87080
    },
    {
      "epoch": 33.82,
      "learning_rate": 1.5490485436893207e-05,
      "loss": 0.0352,
      "step": 87090
    },
    {
      "epoch": 33.83,
      "learning_rate": 1.5489967637540455e-05,
      "loss": 0.0456,
      "step": 87100
    },
    {
      "epoch": 33.83,
      "learning_rate": 1.5489449838187706e-05,
      "loss": 0.1057,
      "step": 87110
    },
    {
      "epoch": 33.83,
      "learning_rate": 1.548893203883495e-05,
      "loss": 0.154,
      "step": 87120
    },
    {
      "epoch": 33.84,
      "learning_rate": 1.5488414239482202e-05,
      "loss": 0.1137,
      "step": 87130
    },
    {
      "epoch": 33.84,
      "learning_rate": 1.548789644012945e-05,
      "loss": 0.0839,
      "step": 87140
    },
    {
      "epoch": 33.84,
      "learning_rate": 1.54873786407767e-05,
      "loss": 0.096,
      "step": 87150
    },
    {
      "epoch": 33.85,
      "learning_rate": 1.548686084142395e-05,
      "loss": 0.0841,
      "step": 87160
    },
    {
      "epoch": 33.85,
      "learning_rate": 1.54863430420712e-05,
      "loss": 0.1806,
      "step": 87170
    },
    {
      "epoch": 33.86,
      "learning_rate": 1.5485825242718448e-05,
      "loss": 0.1215,
      "step": 87180
    },
    {
      "epoch": 33.86,
      "learning_rate": 1.54853074433657e-05,
      "loss": 0.2255,
      "step": 87190
    },
    {
      "epoch": 33.86,
      "learning_rate": 1.5484789644012944e-05,
      "loss": 0.0507,
      "step": 87200
    },
    {
      "epoch": 33.87,
      "learning_rate": 1.5484271844660195e-05,
      "loss": 0.1222,
      "step": 87210
    },
    {
      "epoch": 33.87,
      "learning_rate": 1.5483754045307443e-05,
      "loss": 0.2802,
      "step": 87220
    },
    {
      "epoch": 33.88,
      "learning_rate": 1.5483236245954695e-05,
      "loss": 0.0518,
      "step": 87230
    },
    {
      "epoch": 33.88,
      "learning_rate": 1.5482718446601942e-05,
      "loss": 0.109,
      "step": 87240
    },
    {
      "epoch": 33.88,
      "learning_rate": 1.5482200647249194e-05,
      "loss": 0.1683,
      "step": 87250
    },
    {
      "epoch": 33.89,
      "learning_rate": 1.5481682847896442e-05,
      "loss": 0.0134,
      "step": 87260
    },
    {
      "epoch": 33.89,
      "learning_rate": 1.5481165048543693e-05,
      "loss": 0.1729,
      "step": 87270
    },
    {
      "epoch": 33.9,
      "learning_rate": 1.5480647249190938e-05,
      "loss": 0.1158,
      "step": 87280
    },
    {
      "epoch": 33.9,
      "learning_rate": 1.548012944983819e-05,
      "loss": 0.0244,
      "step": 87290
    },
    {
      "epoch": 33.9,
      "learning_rate": 1.5479611650485437e-05,
      "loss": 0.2736,
      "step": 87300
    },
    {
      "epoch": 33.91,
      "learning_rate": 1.5479093851132688e-05,
      "loss": 0.0149,
      "step": 87310
    },
    {
      "epoch": 33.91,
      "learning_rate": 1.5478576051779936e-05,
      "loss": 0.0039,
      "step": 87320
    },
    {
      "epoch": 33.91,
      "learning_rate": 1.5478058252427187e-05,
      "loss": 0.0301,
      "step": 87330
    },
    {
      "epoch": 33.92,
      "learning_rate": 1.5477540453074435e-05,
      "loss": 0.0432,
      "step": 87340
    },
    {
      "epoch": 33.92,
      "learning_rate": 1.5477022653721687e-05,
      "loss": 0.0589,
      "step": 87350
    },
    {
      "epoch": 33.93,
      "learning_rate": 1.547650485436893e-05,
      "loss": 0.016,
      "step": 87360
    },
    {
      "epoch": 33.93,
      "learning_rate": 1.5475987055016182e-05,
      "loss": 0.1356,
      "step": 87370
    },
    {
      "epoch": 33.93,
      "learning_rate": 1.547546925566343e-05,
      "loss": 0.0913,
      "step": 87380
    },
    {
      "epoch": 33.94,
      "learning_rate": 1.547495145631068e-05,
      "loss": 0.0847,
      "step": 87390
    },
    {
      "epoch": 33.94,
      "learning_rate": 1.547443365695793e-05,
      "loss": 0.1327,
      "step": 87400
    },
    {
      "epoch": 33.95,
      "learning_rate": 1.547391585760518e-05,
      "loss": 0.004,
      "step": 87410
    },
    {
      "epoch": 33.95,
      "learning_rate": 1.547339805825243e-05,
      "loss": 0.1022,
      "step": 87420
    },
    {
      "epoch": 33.95,
      "learning_rate": 1.547288025889968e-05,
      "loss": 0.0115,
      "step": 87430
    },
    {
      "epoch": 33.96,
      "learning_rate": 1.5472362459546925e-05,
      "loss": 0.0508,
      "step": 87440
    },
    {
      "epoch": 33.96,
      "learning_rate": 1.5471844660194176e-05,
      "loss": 0.0705,
      "step": 87450
    },
    {
      "epoch": 33.97,
      "learning_rate": 1.5471326860841424e-05,
      "loss": 0.127,
      "step": 87460
    },
    {
      "epoch": 33.97,
      "learning_rate": 1.5470809061488675e-05,
      "loss": 0.0725,
      "step": 87470
    },
    {
      "epoch": 33.97,
      "learning_rate": 1.5470291262135923e-05,
      "loss": 0.0937,
      "step": 87480
    },
    {
      "epoch": 33.98,
      "learning_rate": 1.5469773462783174e-05,
      "loss": 0.1534,
      "step": 87490
    },
    {
      "epoch": 33.98,
      "learning_rate": 1.5469255663430422e-05,
      "loss": 0.2524,
      "step": 87500
    },
    {
      "epoch": 33.98,
      "learning_rate": 1.546873786407767e-05,
      "loss": 0.0724,
      "step": 87510
    },
    {
      "epoch": 33.99,
      "learning_rate": 1.546822006472492e-05,
      "loss": 0.1304,
      "step": 87520
    },
    {
      "epoch": 33.99,
      "learning_rate": 1.546770226537217e-05,
      "loss": 0.082,
      "step": 87530
    },
    {
      "epoch": 34.0,
      "learning_rate": 1.5467184466019417e-05,
      "loss": 0.1633,
      "step": 87540
    },
    {
      "epoch": 34.0,
      "learning_rate": 1.546666666666667e-05,
      "loss": 0.1178,
      "step": 87550
    },
    {
      "epoch": 34.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.2904781699180603,
      "eval_runtime": 8.2577,
      "eval_samples_per_second": 440.194,
      "eval_steps_per_second": 55.1,
      "step": 87550
    },
    {
      "epoch": 34.0,
      "learning_rate": 1.5466148867313917e-05,
      "loss": 0.0713,
      "step": 87560
    },
    {
      "epoch": 34.01,
      "learning_rate": 1.5465631067961168e-05,
      "loss": 0.0771,
      "step": 87570
    },
    {
      "epoch": 34.01,
      "learning_rate": 1.5465113268608416e-05,
      "loss": 0.013,
      "step": 87580
    },
    {
      "epoch": 34.02,
      "learning_rate": 1.5464595469255664e-05,
      "loss": 0.053,
      "step": 87590
    },
    {
      "epoch": 34.02,
      "learning_rate": 1.5464077669902915e-05,
      "loss": 0.0288,
      "step": 87600
    },
    {
      "epoch": 34.02,
      "learning_rate": 1.5463559870550163e-05,
      "loss": 0.0158,
      "step": 87610
    },
    {
      "epoch": 34.03,
      "learning_rate": 1.546304207119741e-05,
      "loss": 0.1144,
      "step": 87620
    },
    {
      "epoch": 34.03,
      "learning_rate": 1.5462524271844662e-05,
      "loss": 0.083,
      "step": 87630
    },
    {
      "epoch": 34.03,
      "learning_rate": 1.546200647249191e-05,
      "loss": 0.0621,
      "step": 87640
    },
    {
      "epoch": 34.04,
      "learning_rate": 1.546148867313916e-05,
      "loss": 0.0595,
      "step": 87650
    },
    {
      "epoch": 34.04,
      "learning_rate": 1.546097087378641e-05,
      "loss": 0.0253,
      "step": 87660
    },
    {
      "epoch": 34.05,
      "learning_rate": 1.5460453074433657e-05,
      "loss": 0.1537,
      "step": 87670
    },
    {
      "epoch": 34.05,
      "learning_rate": 1.545993527508091e-05,
      "loss": 0.0599,
      "step": 87680
    },
    {
      "epoch": 34.05,
      "learning_rate": 1.5459417475728157e-05,
      "loss": 0.1361,
      "step": 87690
    },
    {
      "epoch": 34.06,
      "learning_rate": 1.5458899676375405e-05,
      "loss": 0.1055,
      "step": 87700
    },
    {
      "epoch": 34.06,
      "learning_rate": 1.5458381877022656e-05,
      "loss": 0.1345,
      "step": 87710
    },
    {
      "epoch": 34.07,
      "learning_rate": 1.5457864077669904e-05,
      "loss": 0.1939,
      "step": 87720
    },
    {
      "epoch": 34.07,
      "learning_rate": 1.5457346278317155e-05,
      "loss": 0.1182,
      "step": 87730
    },
    {
      "epoch": 34.07,
      "learning_rate": 1.5456828478964403e-05,
      "loss": 0.0025,
      "step": 87740
    },
    {
      "epoch": 34.08,
      "learning_rate": 1.545631067961165e-05,
      "loss": 0.1201,
      "step": 87750
    },
    {
      "epoch": 34.08,
      "learning_rate": 1.5455792880258902e-05,
      "loss": 0.0724,
      "step": 87760
    },
    {
      "epoch": 34.09,
      "learning_rate": 1.545527508090615e-05,
      "loss": 0.0965,
      "step": 87770
    },
    {
      "epoch": 34.09,
      "learning_rate": 1.5454757281553398e-05,
      "loss": 0.0378,
      "step": 87780
    },
    {
      "epoch": 34.09,
      "learning_rate": 1.545423948220065e-05,
      "loss": 0.0686,
      "step": 87790
    },
    {
      "epoch": 34.1,
      "learning_rate": 1.5453721682847897e-05,
      "loss": 0.0019,
      "step": 87800
    },
    {
      "epoch": 34.1,
      "learning_rate": 1.5453203883495145e-05,
      "loss": 0.0922,
      "step": 87810
    },
    {
      "epoch": 34.1,
      "learning_rate": 1.5452686084142397e-05,
      "loss": 0.1013,
      "step": 87820
    },
    {
      "epoch": 34.11,
      "learning_rate": 1.5452168284789645e-05,
      "loss": 0.0896,
      "step": 87830
    },
    {
      "epoch": 34.11,
      "learning_rate": 1.5451650485436896e-05,
      "loss": 0.0815,
      "step": 87840
    },
    {
      "epoch": 34.12,
      "learning_rate": 1.5451132686084144e-05,
      "loss": 0.0181,
      "step": 87850
    },
    {
      "epoch": 34.12,
      "learning_rate": 1.5450614886731392e-05,
      "loss": 0.1231,
      "step": 87860
    },
    {
      "epoch": 34.12,
      "learning_rate": 1.5450097087378643e-05,
      "loss": 0.1713,
      "step": 87870
    },
    {
      "epoch": 34.13,
      "learning_rate": 1.544957928802589e-05,
      "loss": 0.0516,
      "step": 87880
    },
    {
      "epoch": 34.13,
      "learning_rate": 1.544906148867314e-05,
      "loss": 0.116,
      "step": 87890
    },
    {
      "epoch": 34.14,
      "learning_rate": 1.544854368932039e-05,
      "loss": 0.0335,
      "step": 87900
    },
    {
      "epoch": 34.14,
      "learning_rate": 1.5448025889967638e-05,
      "loss": 0.0935,
      "step": 87910
    },
    {
      "epoch": 34.14,
      "learning_rate": 1.544750809061489e-05,
      "loss": 0.118,
      "step": 87920
    },
    {
      "epoch": 34.15,
      "learning_rate": 1.5446990291262137e-05,
      "loss": 0.0956,
      "step": 87930
    },
    {
      "epoch": 34.15,
      "learning_rate": 1.5446472491909385e-05,
      "loss": 0.1485,
      "step": 87940
    },
    {
      "epoch": 34.16,
      "learning_rate": 1.5445954692556637e-05,
      "loss": 0.1102,
      "step": 87950
    },
    {
      "epoch": 34.16,
      "learning_rate": 1.5445436893203884e-05,
      "loss": 0.1864,
      "step": 87960
    },
    {
      "epoch": 34.16,
      "learning_rate": 1.5444919093851132e-05,
      "loss": 0.0279,
      "step": 87970
    },
    {
      "epoch": 34.17,
      "learning_rate": 1.5444401294498384e-05,
      "loss": 0.0889,
      "step": 87980
    },
    {
      "epoch": 34.17,
      "learning_rate": 1.544388349514563e-05,
      "loss": 0.0304,
      "step": 87990
    },
    {
      "epoch": 34.17,
      "learning_rate": 1.5443365695792883e-05,
      "loss": 0.0026,
      "step": 88000
    },
    {
      "epoch": 34.18,
      "learning_rate": 1.544284789644013e-05,
      "loss": 0.0972,
      "step": 88010
    },
    {
      "epoch": 34.18,
      "learning_rate": 1.544233009708738e-05,
      "loss": 0.0086,
      "step": 88020
    },
    {
      "epoch": 34.19,
      "learning_rate": 1.544181229773463e-05,
      "loss": 0.1022,
      "step": 88030
    },
    {
      "epoch": 34.19,
      "learning_rate": 1.5441294498381878e-05,
      "loss": 0.112,
      "step": 88040
    },
    {
      "epoch": 34.19,
      "learning_rate": 1.5440776699029126e-05,
      "loss": 0.1052,
      "step": 88050
    },
    {
      "epoch": 34.2,
      "learning_rate": 1.5440258899676377e-05,
      "loss": 0.1455,
      "step": 88060
    },
    {
      "epoch": 34.2,
      "learning_rate": 1.5439741100323625e-05,
      "loss": 0.1698,
      "step": 88070
    },
    {
      "epoch": 34.21,
      "learning_rate": 1.5439223300970877e-05,
      "loss": 0.044,
      "step": 88080
    },
    {
      "epoch": 34.21,
      "learning_rate": 1.5438705501618124e-05,
      "loss": 0.0008,
      "step": 88090
    },
    {
      "epoch": 34.21,
      "learning_rate": 1.5438187702265372e-05,
      "loss": 0.0085,
      "step": 88100
    },
    {
      "epoch": 34.22,
      "learning_rate": 1.5437669902912624e-05,
      "loss": 0.072,
      "step": 88110
    },
    {
      "epoch": 34.22,
      "learning_rate": 1.543715210355987e-05,
      "loss": 0.1665,
      "step": 88120
    },
    {
      "epoch": 34.23,
      "learning_rate": 1.543663430420712e-05,
      "loss": 0.0011,
      "step": 88130
    },
    {
      "epoch": 34.23,
      "learning_rate": 1.543611650485437e-05,
      "loss": 0.1055,
      "step": 88140
    },
    {
      "epoch": 34.23,
      "learning_rate": 1.543559870550162e-05,
      "loss": 0.1532,
      "step": 88150
    },
    {
      "epoch": 34.24,
      "learning_rate": 1.543508090614887e-05,
      "loss": 0.0383,
      "step": 88160
    },
    {
      "epoch": 34.24,
      "learning_rate": 1.5434563106796118e-05,
      "loss": 0.041,
      "step": 88170
    },
    {
      "epoch": 34.24,
      "learning_rate": 1.5434045307443366e-05,
      "loss": 0.1241,
      "step": 88180
    },
    {
      "epoch": 34.25,
      "learning_rate": 1.5433527508090614e-05,
      "loss": 0.1371,
      "step": 88190
    },
    {
      "epoch": 34.25,
      "learning_rate": 1.5433009708737865e-05,
      "loss": 0.1814,
      "step": 88200
    },
    {
      "epoch": 34.26,
      "learning_rate": 1.5432491909385113e-05,
      "loss": 0.0859,
      "step": 88210
    },
    {
      "epoch": 34.26,
      "learning_rate": 1.5431974110032364e-05,
      "loss": 0.0144,
      "step": 88220
    },
    {
      "epoch": 34.26,
      "learning_rate": 1.5431456310679612e-05,
      "loss": 0.0596,
      "step": 88230
    },
    {
      "epoch": 34.27,
      "learning_rate": 1.5430938511326864e-05,
      "loss": 0.2371,
      "step": 88240
    },
    {
      "epoch": 34.27,
      "learning_rate": 1.543042071197411e-05,
      "loss": 0.055,
      "step": 88250
    },
    {
      "epoch": 34.28,
      "learning_rate": 1.542990291262136e-05,
      "loss": 0.0389,
      "step": 88260
    },
    {
      "epoch": 34.28,
      "learning_rate": 1.5429385113268607e-05,
      "loss": 0.0362,
      "step": 88270
    },
    {
      "epoch": 34.28,
      "learning_rate": 1.542886731391586e-05,
      "loss": 0.0797,
      "step": 88280
    },
    {
      "epoch": 34.29,
      "learning_rate": 1.5428349514563107e-05,
      "loss": 0.0408,
      "step": 88290
    },
    {
      "epoch": 34.29,
      "learning_rate": 1.5427831715210358e-05,
      "loss": 0.0789,
      "step": 88300
    },
    {
      "epoch": 34.3,
      "learning_rate": 1.5427313915857606e-05,
      "loss": 0.0441,
      "step": 88310
    },
    {
      "epoch": 34.3,
      "learning_rate": 1.5426796116504857e-05,
      "loss": 0.0902,
      "step": 88320
    },
    {
      "epoch": 34.3,
      "learning_rate": 1.5426278317152105e-05,
      "loss": 0.0659,
      "step": 88330
    },
    {
      "epoch": 34.31,
      "learning_rate": 1.5425760517799353e-05,
      "loss": 0.1337,
      "step": 88340
    },
    {
      "epoch": 34.31,
      "learning_rate": 1.54252427184466e-05,
      "loss": 0.1205,
      "step": 88350
    },
    {
      "epoch": 34.31,
      "learning_rate": 1.5424724919093852e-05,
      "loss": 0.1786,
      "step": 88360
    },
    {
      "epoch": 34.32,
      "learning_rate": 1.54242071197411e-05,
      "loss": 0.1009,
      "step": 88370
    },
    {
      "epoch": 34.32,
      "learning_rate": 1.542368932038835e-05,
      "loss": 0.0329,
      "step": 88380
    },
    {
      "epoch": 34.33,
      "learning_rate": 1.54231715210356e-05,
      "loss": 0.0188,
      "step": 88390
    },
    {
      "epoch": 34.33,
      "learning_rate": 1.542265372168285e-05,
      "loss": 0.1515,
      "step": 88400
    },
    {
      "epoch": 34.33,
      "learning_rate": 1.54221359223301e-05,
      "loss": 0.0067,
      "step": 88410
    },
    {
      "epoch": 34.34,
      "learning_rate": 1.5421618122977347e-05,
      "loss": 0.0042,
      "step": 88420
    },
    {
      "epoch": 34.34,
      "learning_rate": 1.5421100323624595e-05,
      "loss": 0.1738,
      "step": 88430
    },
    {
      "epoch": 34.35,
      "learning_rate": 1.5420582524271846e-05,
      "loss": 0.0509,
      "step": 88440
    },
    {
      "epoch": 34.35,
      "learning_rate": 1.5420064724919094e-05,
      "loss": 0.1376,
      "step": 88450
    },
    {
      "epoch": 34.35,
      "learning_rate": 1.5419546925566345e-05,
      "loss": 0.0493,
      "step": 88460
    },
    {
      "epoch": 34.36,
      "learning_rate": 1.5419029126213593e-05,
      "loss": 0.2846,
      "step": 88470
    },
    {
      "epoch": 34.36,
      "learning_rate": 1.5418511326860844e-05,
      "loss": 0.1113,
      "step": 88480
    },
    {
      "epoch": 34.37,
      "learning_rate": 1.5417993527508092e-05,
      "loss": 0.0574,
      "step": 88490
    },
    {
      "epoch": 34.37,
      "learning_rate": 1.541747572815534e-05,
      "loss": 0.0532,
      "step": 88500
    },
    {
      "epoch": 34.37,
      "learning_rate": 1.5416957928802588e-05,
      "loss": 0.1207,
      "step": 88510
    },
    {
      "epoch": 34.38,
      "learning_rate": 1.541644012944984e-05,
      "loss": 0.1223,
      "step": 88520
    },
    {
      "epoch": 34.38,
      "learning_rate": 1.5415922330097087e-05,
      "loss": 0.1484,
      "step": 88530
    },
    {
      "epoch": 34.38,
      "learning_rate": 1.541540453074434e-05,
      "loss": 0.1104,
      "step": 88540
    },
    {
      "epoch": 34.39,
      "learning_rate": 1.5414886731391587e-05,
      "loss": 0.2227,
      "step": 88550
    },
    {
      "epoch": 34.39,
      "learning_rate": 1.5414368932038838e-05,
      "loss": 0.1712,
      "step": 88560
    },
    {
      "epoch": 34.4,
      "learning_rate": 1.5413851132686086e-05,
      "loss": 0.2455,
      "step": 88570
    },
    {
      "epoch": 34.4,
      "learning_rate": 1.5413333333333337e-05,
      "loss": 0.1156,
      "step": 88580
    },
    {
      "epoch": 34.4,
      "learning_rate": 1.541281553398058e-05,
      "loss": 0.3218,
      "step": 88590
    },
    {
      "epoch": 34.41,
      "learning_rate": 1.5412297734627833e-05,
      "loss": 0.209,
      "step": 88600
    },
    {
      "epoch": 34.41,
      "learning_rate": 1.541177993527508e-05,
      "loss": 0.023,
      "step": 88610
    },
    {
      "epoch": 34.42,
      "learning_rate": 1.5411262135922332e-05,
      "loss": 0.0785,
      "step": 88620
    },
    {
      "epoch": 34.42,
      "learning_rate": 1.541074433656958e-05,
      "loss": 0.0031,
      "step": 88630
    },
    {
      "epoch": 34.42,
      "learning_rate": 1.541022653721683e-05,
      "loss": 0.0388,
      "step": 88640
    },
    {
      "epoch": 34.43,
      "learning_rate": 1.540970873786408e-05,
      "loss": 0.0182,
      "step": 88650
    },
    {
      "epoch": 34.43,
      "learning_rate": 1.540919093851133e-05,
      "loss": 0.1296,
      "step": 88660
    },
    {
      "epoch": 34.43,
      "learning_rate": 1.5408673139158575e-05,
      "loss": 0.1659,
      "step": 88670
    },
    {
      "epoch": 34.44,
      "learning_rate": 1.5408155339805827e-05,
      "loss": 0.1408,
      "step": 88680
    },
    {
      "epoch": 34.44,
      "learning_rate": 1.5407637540453074e-05,
      "loss": 0.0599,
      "step": 88690
    },
    {
      "epoch": 34.45,
      "learning_rate": 1.5407119741100326e-05,
      "loss": 0.2001,
      "step": 88700
    },
    {
      "epoch": 34.45,
      "learning_rate": 1.5406601941747574e-05,
      "loss": 0.0837,
      "step": 88710
    },
    {
      "epoch": 34.45,
      "learning_rate": 1.5406084142394825e-05,
      "loss": 0.0851,
      "step": 88720
    },
    {
      "epoch": 34.46,
      "learning_rate": 1.5405566343042073e-05,
      "loss": 0.0694,
      "step": 88730
    },
    {
      "epoch": 34.46,
      "learning_rate": 1.5405048543689324e-05,
      "loss": 0.0596,
      "step": 88740
    },
    {
      "epoch": 34.47,
      "learning_rate": 1.540453074433657e-05,
      "loss": 0.1862,
      "step": 88750
    },
    {
      "epoch": 34.47,
      "learning_rate": 1.540401294498382e-05,
      "loss": 0.0152,
      "step": 88760
    },
    {
      "epoch": 34.47,
      "learning_rate": 1.5403495145631068e-05,
      "loss": 0.0497,
      "step": 88770
    },
    {
      "epoch": 34.48,
      "learning_rate": 1.540297734627832e-05,
      "loss": 0.0979,
      "step": 88780
    },
    {
      "epoch": 34.48,
      "learning_rate": 1.5402459546925567e-05,
      "loss": 0.0946,
      "step": 88790
    },
    {
      "epoch": 34.49,
      "learning_rate": 1.540194174757282e-05,
      "loss": 0.0777,
      "step": 88800
    },
    {
      "epoch": 34.49,
      "learning_rate": 1.5401423948220066e-05,
      "loss": 0.0239,
      "step": 88810
    },
    {
      "epoch": 34.49,
      "learning_rate": 1.5400906148867318e-05,
      "loss": 0.0649,
      "step": 88820
    },
    {
      "epoch": 34.5,
      "learning_rate": 1.5400388349514562e-05,
      "loss": 0.0832,
      "step": 88830
    },
    {
      "epoch": 34.5,
      "learning_rate": 1.5399870550161814e-05,
      "loss": 0.1227,
      "step": 88840
    },
    {
      "epoch": 34.5,
      "learning_rate": 1.539935275080906e-05,
      "loss": 0.038,
      "step": 88850
    },
    {
      "epoch": 34.51,
      "learning_rate": 1.5398834951456313e-05,
      "loss": 0.0202,
      "step": 88860
    },
    {
      "epoch": 34.51,
      "learning_rate": 1.539831715210356e-05,
      "loss": 0.0112,
      "step": 88870
    },
    {
      "epoch": 34.52,
      "learning_rate": 1.5397799352750812e-05,
      "loss": 0.094,
      "step": 88880
    },
    {
      "epoch": 34.52,
      "learning_rate": 1.539728155339806e-05,
      "loss": 0.0699,
      "step": 88890
    },
    {
      "epoch": 34.52,
      "learning_rate": 1.539676375404531e-05,
      "loss": 0.0881,
      "step": 88900
    },
    {
      "epoch": 34.53,
      "learning_rate": 1.5396245954692556e-05,
      "loss": 0.1762,
      "step": 88910
    },
    {
      "epoch": 34.53,
      "learning_rate": 1.5395728155339807e-05,
      "loss": 0.3549,
      "step": 88920
    },
    {
      "epoch": 34.54,
      "learning_rate": 1.5395210355987055e-05,
      "loss": 0.0835,
      "step": 88930
    },
    {
      "epoch": 34.54,
      "learning_rate": 1.5394692556634306e-05,
      "loss": 0.0204,
      "step": 88940
    },
    {
      "epoch": 34.54,
      "learning_rate": 1.5394174757281554e-05,
      "loss": 0.0829,
      "step": 88950
    },
    {
      "epoch": 34.55,
      "learning_rate": 1.5393656957928806e-05,
      "loss": 0.0316,
      "step": 88960
    },
    {
      "epoch": 34.55,
      "learning_rate": 1.5393139158576054e-05,
      "loss": 0.1588,
      "step": 88970
    },
    {
      "epoch": 34.56,
      "learning_rate": 1.53926213592233e-05,
      "loss": 0.0321,
      "step": 88980
    },
    {
      "epoch": 34.56,
      "learning_rate": 1.539210355987055e-05,
      "loss": 0.0475,
      "step": 88990
    },
    {
      "epoch": 34.56,
      "learning_rate": 1.53915857605178e-05,
      "loss": 0.0812,
      "step": 89000
    },
    {
      "epoch": 34.57,
      "learning_rate": 1.539106796116505e-05,
      "loss": 0.0845,
      "step": 89010
    },
    {
      "epoch": 34.57,
      "learning_rate": 1.53905501618123e-05,
      "loss": 0.0147,
      "step": 89020
    },
    {
      "epoch": 34.57,
      "learning_rate": 1.5390032362459548e-05,
      "loss": 0.1045,
      "step": 89030
    },
    {
      "epoch": 34.58,
      "learning_rate": 1.53895145631068e-05,
      "loss": 0.0153,
      "step": 89040
    },
    {
      "epoch": 34.58,
      "learning_rate": 1.5388996763754047e-05,
      "loss": 0.1033,
      "step": 89050
    },
    {
      "epoch": 34.59,
      "learning_rate": 1.5388478964401295e-05,
      "loss": 0.0022,
      "step": 89060
    },
    {
      "epoch": 34.59,
      "learning_rate": 1.5387961165048543e-05,
      "loss": 0.1272,
      "step": 89070
    },
    {
      "epoch": 34.59,
      "learning_rate": 1.5387443365695794e-05,
      "loss": 0.1584,
      "step": 89080
    },
    {
      "epoch": 34.6,
      "learning_rate": 1.5386925566343042e-05,
      "loss": 0.1784,
      "step": 89090
    },
    {
      "epoch": 34.6,
      "learning_rate": 1.5386407766990294e-05,
      "loss": 0.0069,
      "step": 89100
    },
    {
      "epoch": 34.61,
      "learning_rate": 1.538588996763754e-05,
      "loss": 0.1399,
      "step": 89110
    },
    {
      "epoch": 34.61,
      "learning_rate": 1.5385372168284793e-05,
      "loss": 0.1145,
      "step": 89120
    },
    {
      "epoch": 34.61,
      "learning_rate": 1.538485436893204e-05,
      "loss": 0.1506,
      "step": 89130
    },
    {
      "epoch": 34.62,
      "learning_rate": 1.538433656957929e-05,
      "loss": 0.217,
      "step": 89140
    },
    {
      "epoch": 34.62,
      "learning_rate": 1.538381877022654e-05,
      "loss": 0.1038,
      "step": 89150
    },
    {
      "epoch": 34.63,
      "learning_rate": 1.5383300970873788e-05,
      "loss": 0.0834,
      "step": 89160
    },
    {
      "epoch": 34.63,
      "learning_rate": 1.5382783171521036e-05,
      "loss": 0.0603,
      "step": 89170
    },
    {
      "epoch": 34.63,
      "learning_rate": 1.5382265372168287e-05,
      "loss": 0.0496,
      "step": 89180
    },
    {
      "epoch": 34.64,
      "learning_rate": 1.5381747572815535e-05,
      "loss": 0.0495,
      "step": 89190
    },
    {
      "epoch": 34.64,
      "learning_rate": 1.5381229773462786e-05,
      "loss": 0.089,
      "step": 89200
    },
    {
      "epoch": 34.64,
      "learning_rate": 1.5380711974110034e-05,
      "loss": 0.0548,
      "step": 89210
    },
    {
      "epoch": 34.65,
      "learning_rate": 1.5380194174757282e-05,
      "loss": 0.0503,
      "step": 89220
    },
    {
      "epoch": 34.65,
      "learning_rate": 1.5379676375404533e-05,
      "loss": 0.1261,
      "step": 89230
    },
    {
      "epoch": 34.66,
      "learning_rate": 1.537915857605178e-05,
      "loss": 0.1161,
      "step": 89240
    },
    {
      "epoch": 34.66,
      "learning_rate": 1.537864077669903e-05,
      "loss": 0.1373,
      "step": 89250
    },
    {
      "epoch": 34.66,
      "learning_rate": 1.537812297734628e-05,
      "loss": 0.1411,
      "step": 89260
    },
    {
      "epoch": 34.67,
      "learning_rate": 1.537760517799353e-05,
      "loss": 0.2279,
      "step": 89270
    },
    {
      "epoch": 34.67,
      "learning_rate": 1.5377087378640776e-05,
      "loss": 0.0915,
      "step": 89280
    },
    {
      "epoch": 34.68,
      "learning_rate": 1.5376569579288028e-05,
      "loss": 0.2615,
      "step": 89290
    },
    {
      "epoch": 34.68,
      "learning_rate": 1.5376051779935276e-05,
      "loss": 0.0688,
      "step": 89300
    },
    {
      "epoch": 34.68,
      "learning_rate": 1.5375533980582527e-05,
      "loss": 0.0842,
      "step": 89310
    },
    {
      "epoch": 34.69,
      "learning_rate": 1.5375016181229775e-05,
      "loss": 0.1435,
      "step": 89320
    },
    {
      "epoch": 34.69,
      "learning_rate": 1.5374498381877023e-05,
      "loss": 0.1902,
      "step": 89330
    },
    {
      "epoch": 34.7,
      "learning_rate": 1.5373980582524274e-05,
      "loss": 0.0642,
      "step": 89340
    },
    {
      "epoch": 34.7,
      "learning_rate": 1.5373462783171522e-05,
      "loss": 0.034,
      "step": 89350
    },
    {
      "epoch": 34.7,
      "learning_rate": 1.537294498381877e-05,
      "loss": 0.1001,
      "step": 89360
    },
    {
      "epoch": 34.71,
      "learning_rate": 1.537242718446602e-05,
      "loss": 0.2239,
      "step": 89370
    },
    {
      "epoch": 34.71,
      "learning_rate": 1.537190938511327e-05,
      "loss": 0.0725,
      "step": 89380
    },
    {
      "epoch": 34.71,
      "learning_rate": 1.537139158576052e-05,
      "loss": 0.1676,
      "step": 89390
    },
    {
      "epoch": 34.72,
      "learning_rate": 1.537087378640777e-05,
      "loss": 0.0946,
      "step": 89400
    },
    {
      "epoch": 34.72,
      "learning_rate": 1.5370355987055016e-05,
      "loss": 0.0875,
      "step": 89410
    },
    {
      "epoch": 34.73,
      "learning_rate": 1.5369838187702268e-05,
      "loss": 0.1889,
      "step": 89420
    },
    {
      "epoch": 34.73,
      "learning_rate": 1.5369320388349516e-05,
      "loss": 0.0656,
      "step": 89430
    },
    {
      "epoch": 34.73,
      "learning_rate": 1.5368802588996764e-05,
      "loss": 0.0484,
      "step": 89440
    },
    {
      "epoch": 34.74,
      "learning_rate": 1.5368284789644015e-05,
      "loss": 0.1189,
      "step": 89450
    },
    {
      "epoch": 34.74,
      "learning_rate": 1.5367766990291263e-05,
      "loss": 0.0957,
      "step": 89460
    },
    {
      "epoch": 34.75,
      "learning_rate": 1.5367249190938514e-05,
      "loss": 0.1193,
      "step": 89470
    },
    {
      "epoch": 34.75,
      "learning_rate": 1.5366731391585762e-05,
      "loss": 0.0489,
      "step": 89480
    },
    {
      "epoch": 34.75,
      "learning_rate": 1.536621359223301e-05,
      "loss": 0.1087,
      "step": 89490
    },
    {
      "epoch": 34.76,
      "learning_rate": 1.536569579288026e-05,
      "loss": 0.029,
      "step": 89500
    },
    {
      "epoch": 34.76,
      "learning_rate": 1.536517799352751e-05,
      "loss": 0.0134,
      "step": 89510
    },
    {
      "epoch": 34.77,
      "learning_rate": 1.5364660194174757e-05,
      "loss": 0.1008,
      "step": 89520
    },
    {
      "epoch": 34.77,
      "learning_rate": 1.536414239482201e-05,
      "loss": 0.1453,
      "step": 89530
    },
    {
      "epoch": 34.77,
      "learning_rate": 1.5363624595469256e-05,
      "loss": 0.0357,
      "step": 89540
    },
    {
      "epoch": 34.78,
      "learning_rate": 1.5363106796116508e-05,
      "loss": 0.3234,
      "step": 89550
    },
    {
      "epoch": 34.78,
      "learning_rate": 1.5362588996763756e-05,
      "loss": 0.0245,
      "step": 89560
    },
    {
      "epoch": 34.78,
      "learning_rate": 1.5362071197411004e-05,
      "loss": 0.0079,
      "step": 89570
    },
    {
      "epoch": 34.79,
      "learning_rate": 1.5361553398058255e-05,
      "loss": 0.0464,
      "step": 89580
    },
    {
      "epoch": 34.79,
      "learning_rate": 1.5361035598705503e-05,
      "loss": 0.0078,
      "step": 89590
    },
    {
      "epoch": 34.8,
      "learning_rate": 1.536051779935275e-05,
      "loss": 0.1744,
      "step": 89600
    },
    {
      "epoch": 34.8,
      "learning_rate": 1.5360000000000002e-05,
      "loss": 0.0413,
      "step": 89610
    },
    {
      "epoch": 34.8,
      "learning_rate": 1.535948220064725e-05,
      "loss": 0.0185,
      "step": 89620
    },
    {
      "epoch": 34.81,
      "learning_rate": 1.53589644012945e-05,
      "loss": 0.0225,
      "step": 89630
    },
    {
      "epoch": 34.81,
      "learning_rate": 1.535844660194175e-05,
      "loss": 0.2146,
      "step": 89640
    },
    {
      "epoch": 34.82,
      "learning_rate": 1.5357928802588997e-05,
      "loss": 0.3258,
      "step": 89650
    },
    {
      "epoch": 34.82,
      "learning_rate": 1.5357411003236245e-05,
      "loss": 0.0363,
      "step": 89660
    },
    {
      "epoch": 34.82,
      "learning_rate": 1.5356893203883496e-05,
      "loss": 0.2226,
      "step": 89670
    },
    {
      "epoch": 34.83,
      "learning_rate": 1.5356375404530744e-05,
      "loss": 0.0582,
      "step": 89680
    },
    {
      "epoch": 34.83,
      "learning_rate": 1.5355857605177996e-05,
      "loss": 0.0854,
      "step": 89690
    },
    {
      "epoch": 34.83,
      "learning_rate": 1.5355339805825243e-05,
      "loss": 0.172,
      "step": 89700
    },
    {
      "epoch": 34.84,
      "learning_rate": 1.5354822006472495e-05,
      "loss": 0.0775,
      "step": 89710
    },
    {
      "epoch": 34.84,
      "learning_rate": 1.5354304207119743e-05,
      "loss": 0.0212,
      "step": 89720
    },
    {
      "epoch": 34.85,
      "learning_rate": 1.535378640776699e-05,
      "loss": 0.0155,
      "step": 89730
    },
    {
      "epoch": 34.85,
      "learning_rate": 1.535326860841424e-05,
      "loss": 0.0662,
      "step": 89740
    },
    {
      "epoch": 34.85,
      "learning_rate": 1.535275080906149e-05,
      "loss": 0.1445,
      "step": 89750
    },
    {
      "epoch": 34.86,
      "learning_rate": 1.5352233009708738e-05,
      "loss": 0.1408,
      "step": 89760
    },
    {
      "epoch": 34.86,
      "learning_rate": 1.535171521035599e-05,
      "loss": 0.1785,
      "step": 89770
    },
    {
      "epoch": 34.87,
      "learning_rate": 1.5351197411003237e-05,
      "loss": 0.1779,
      "step": 89780
    },
    {
      "epoch": 34.87,
      "learning_rate": 1.535067961165049e-05,
      "loss": 0.0681,
      "step": 89790
    },
    {
      "epoch": 34.87,
      "learning_rate": 1.5350161812297736e-05,
      "loss": 0.0107,
      "step": 89800
    },
    {
      "epoch": 34.88,
      "learning_rate": 1.5349644012944984e-05,
      "loss": 0.2209,
      "step": 89810
    },
    {
      "epoch": 34.88,
      "learning_rate": 1.5349126213592232e-05,
      "loss": 0.028,
      "step": 89820
    },
    {
      "epoch": 34.89,
      "learning_rate": 1.5348608414239483e-05,
      "loss": 0.0759,
      "step": 89830
    },
    {
      "epoch": 34.89,
      "learning_rate": 1.534809061488673e-05,
      "loss": 0.1212,
      "step": 89840
    },
    {
      "epoch": 34.89,
      "learning_rate": 1.5347572815533983e-05,
      "loss": 0.3049,
      "step": 89850
    },
    {
      "epoch": 34.9,
      "learning_rate": 1.534705501618123e-05,
      "loss": 0.1724,
      "step": 89860
    },
    {
      "epoch": 34.9,
      "learning_rate": 1.5346537216828482e-05,
      "loss": 0.0757,
      "step": 89870
    },
    {
      "epoch": 34.9,
      "learning_rate": 1.534601941747573e-05,
      "loss": 0.1087,
      "step": 89880
    },
    {
      "epoch": 34.91,
      "learning_rate": 1.5345501618122978e-05,
      "loss": 0.086,
      "step": 89890
    },
    {
      "epoch": 34.91,
      "learning_rate": 1.5344983818770226e-05,
      "loss": 0.0628,
      "step": 89900
    },
    {
      "epoch": 34.92,
      "learning_rate": 1.5344466019417477e-05,
      "loss": 0.033,
      "step": 89910
    },
    {
      "epoch": 34.92,
      "learning_rate": 1.5343948220064725e-05,
      "loss": 0.2318,
      "step": 89920
    },
    {
      "epoch": 34.92,
      "learning_rate": 1.5343430420711976e-05,
      "loss": 0.0361,
      "step": 89930
    },
    {
      "epoch": 34.93,
      "learning_rate": 1.5342912621359224e-05,
      "loss": 0.143,
      "step": 89940
    },
    {
      "epoch": 34.93,
      "learning_rate": 1.5342394822006475e-05,
      "loss": 0.1206,
      "step": 89950
    },
    {
      "epoch": 34.94,
      "learning_rate": 1.5341877022653723e-05,
      "loss": 0.0727,
      "step": 89960
    },
    {
      "epoch": 34.94,
      "learning_rate": 1.534135922330097e-05,
      "loss": 0.0441,
      "step": 89970
    },
    {
      "epoch": 34.94,
      "learning_rate": 1.534084142394822e-05,
      "loss": 0.0114,
      "step": 89980
    },
    {
      "epoch": 34.95,
      "learning_rate": 1.534032362459547e-05,
      "loss": 0.1423,
      "step": 89990
    },
    {
      "epoch": 34.95,
      "learning_rate": 1.533980582524272e-05,
      "loss": 0.1311,
      "step": 90000
    },
    {
      "epoch": 34.96,
      "learning_rate": 1.533928802588997e-05,
      "loss": 0.127,
      "step": 90010
    },
    {
      "epoch": 34.96,
      "learning_rate": 1.5338770226537218e-05,
      "loss": 0.1807,
      "step": 90020
    },
    {
      "epoch": 34.96,
      "learning_rate": 1.533825242718447e-05,
      "loss": 0.0445,
      "step": 90030
    },
    {
      "epoch": 34.97,
      "learning_rate": 1.5337734627831717e-05,
      "loss": 0.1863,
      "step": 90040
    },
    {
      "epoch": 34.97,
      "learning_rate": 1.5337216828478965e-05,
      "loss": 0.1139,
      "step": 90050
    },
    {
      "epoch": 34.97,
      "learning_rate": 1.5336699029126213e-05,
      "loss": 0.1244,
      "step": 90060
    },
    {
      "epoch": 34.98,
      "learning_rate": 1.5336181229773464e-05,
      "loss": 0.2443,
      "step": 90070
    },
    {
      "epoch": 34.98,
      "learning_rate": 1.5335663430420712e-05,
      "loss": 0.1439,
      "step": 90080
    },
    {
      "epoch": 34.99,
      "learning_rate": 1.5335145631067963e-05,
      "loss": 0.1472,
      "step": 90090
    },
    {
      "epoch": 34.99,
      "learning_rate": 1.533462783171521e-05,
      "loss": 0.0298,
      "step": 90100
    },
    {
      "epoch": 34.99,
      "learning_rate": 1.5334110032362463e-05,
      "loss": 0.0622,
      "step": 90110
    },
    {
      "epoch": 35.0,
      "learning_rate": 1.533359223300971e-05,
      "loss": 0.0804,
      "step": 90120
    },
    {
      "epoch": 35.0,
      "eval_accuracy": 0.9480055020632737,
      "eval_loss": 0.31004270911216736,
      "eval_runtime": 8.2797,
      "eval_samples_per_second": 439.026,
      "eval_steps_per_second": 54.954,
      "step": 90125
    },
    {
      "epoch": 35.0,
      "learning_rate": 1.533307443365696e-05,
      "loss": 0.0607,
      "step": 90130
    },
    {
      "epoch": 35.01,
      "learning_rate": 1.5332556634304206e-05,
      "loss": 0.0386,
      "step": 90140
    },
    {
      "epoch": 35.01,
      "learning_rate": 1.5332038834951458e-05,
      "loss": 0.0873,
      "step": 90150
    },
    {
      "epoch": 35.01,
      "learning_rate": 1.5331521035598706e-05,
      "loss": 0.0027,
      "step": 90160
    },
    {
      "epoch": 35.02,
      "learning_rate": 1.5331003236245957e-05,
      "loss": 0.1629,
      "step": 90170
    },
    {
      "epoch": 35.02,
      "learning_rate": 1.5330485436893205e-05,
      "loss": 0.0639,
      "step": 90180
    },
    {
      "epoch": 35.03,
      "learning_rate": 1.5329967637540456e-05,
      "loss": 0.1503,
      "step": 90190
    },
    {
      "epoch": 35.03,
      "learning_rate": 1.5329449838187704e-05,
      "loss": 0.0294,
      "step": 90200
    },
    {
      "epoch": 35.03,
      "learning_rate": 1.5328932038834952e-05,
      "loss": 0.0771,
      "step": 90210
    },
    {
      "epoch": 35.04,
      "learning_rate": 1.53284142394822e-05,
      "loss": 0.0688,
      "step": 90220
    },
    {
      "epoch": 35.04,
      "learning_rate": 1.532789644012945e-05,
      "loss": 0.0216,
      "step": 90230
    },
    {
      "epoch": 35.04,
      "learning_rate": 1.53273786407767e-05,
      "loss": 0.1791,
      "step": 90240
    },
    {
      "epoch": 35.05,
      "learning_rate": 1.532686084142395e-05,
      "loss": 0.0896,
      "step": 90250
    },
    {
      "epoch": 35.05,
      "learning_rate": 1.53263430420712e-05,
      "loss": 0.0108,
      "step": 90260
    },
    {
      "epoch": 35.06,
      "learning_rate": 1.532582524271845e-05,
      "loss": 0.0367,
      "step": 90270
    },
    {
      "epoch": 35.06,
      "learning_rate": 1.5325307443365698e-05,
      "loss": 0.047,
      "step": 90280
    },
    {
      "epoch": 35.06,
      "learning_rate": 1.532478964401295e-05,
      "loss": 0.015,
      "step": 90290
    },
    {
      "epoch": 35.07,
      "learning_rate": 1.5324271844660193e-05,
      "loss": 0.1414,
      "step": 90300
    },
    {
      "epoch": 35.07,
      "learning_rate": 1.5323754045307445e-05,
      "loss": 0.1428,
      "step": 90310
    },
    {
      "epoch": 35.08,
      "learning_rate": 1.5323236245954693e-05,
      "loss": 0.1067,
      "step": 90320
    },
    {
      "epoch": 35.08,
      "learning_rate": 1.5322718446601944e-05,
      "loss": 0.0126,
      "step": 90330
    },
    {
      "epoch": 35.08,
      "learning_rate": 1.5322200647249192e-05,
      "loss": 0.163,
      "step": 90340
    },
    {
      "epoch": 35.09,
      "learning_rate": 1.5321682847896443e-05,
      "loss": 0.0429,
      "step": 90350
    },
    {
      "epoch": 35.09,
      "learning_rate": 1.532116504854369e-05,
      "loss": 0.0564,
      "step": 90360
    },
    {
      "epoch": 35.1,
      "learning_rate": 1.5320647249190942e-05,
      "loss": 0.0974,
      "step": 90370
    },
    {
      "epoch": 35.1,
      "learning_rate": 1.5320129449838187e-05,
      "loss": 0.0075,
      "step": 90380
    },
    {
      "epoch": 35.1,
      "learning_rate": 1.531961165048544e-05,
      "loss": 0.0841,
      "step": 90390
    },
    {
      "epoch": 35.11,
      "learning_rate": 1.5319093851132686e-05,
      "loss": 0.0059,
      "step": 90400
    },
    {
      "epoch": 35.11,
      "learning_rate": 1.5318576051779938e-05,
      "loss": 0.1126,
      "step": 90410
    },
    {
      "epoch": 35.11,
      "learning_rate": 1.5318058252427186e-05,
      "loss": 0.005,
      "step": 90420
    },
    {
      "epoch": 35.12,
      "learning_rate": 1.5317540453074437e-05,
      "loss": 0.1748,
      "step": 90430
    },
    {
      "epoch": 35.12,
      "learning_rate": 1.5317022653721685e-05,
      "loss": 0.0943,
      "step": 90440
    },
    {
      "epoch": 35.13,
      "learning_rate": 1.5316504854368933e-05,
      "loss": 0.1351,
      "step": 90450
    },
    {
      "epoch": 35.13,
      "learning_rate": 1.531598705501618e-05,
      "loss": 0.0592,
      "step": 90460
    },
    {
      "epoch": 35.13,
      "learning_rate": 1.5315469255663432e-05,
      "loss": 0.125,
      "step": 90470
    },
    {
      "epoch": 35.14,
      "learning_rate": 1.531495145631068e-05,
      "loss": 0.1609,
      "step": 90480
    },
    {
      "epoch": 35.14,
      "learning_rate": 1.531443365695793e-05,
      "loss": 0.0776,
      "step": 90490
    },
    {
      "epoch": 35.15,
      "learning_rate": 1.531391585760518e-05,
      "loss": 0.1172,
      "step": 90500
    },
    {
      "epoch": 35.15,
      "learning_rate": 1.531339805825243e-05,
      "loss": 0.074,
      "step": 90510
    },
    {
      "epoch": 35.15,
      "learning_rate": 1.5312880258899678e-05,
      "loss": 0.0441,
      "step": 90520
    },
    {
      "epoch": 35.16,
      "learning_rate": 1.5312362459546926e-05,
      "loss": 0.1101,
      "step": 90530
    },
    {
      "epoch": 35.16,
      "learning_rate": 1.5311844660194174e-05,
      "loss": 0.0618,
      "step": 90540
    },
    {
      "epoch": 35.17,
      "learning_rate": 1.5311326860841425e-05,
      "loss": 0.0632,
      "step": 90550
    },
    {
      "epoch": 35.17,
      "learning_rate": 1.5310809061488673e-05,
      "loss": 0.1773,
      "step": 90560
    },
    {
      "epoch": 35.17,
      "learning_rate": 1.5310291262135925e-05,
      "loss": 0.0426,
      "step": 90570
    },
    {
      "epoch": 35.18,
      "learning_rate": 1.5309773462783173e-05,
      "loss": 0.1006,
      "step": 90580
    },
    {
      "epoch": 35.18,
      "learning_rate": 1.5309255663430424e-05,
      "loss": 0.0119,
      "step": 90590
    },
    {
      "epoch": 35.18,
      "learning_rate": 1.5308737864077672e-05,
      "loss": 0.1303,
      "step": 90600
    },
    {
      "epoch": 35.19,
      "learning_rate": 1.530822006472492e-05,
      "loss": 0.1378,
      "step": 90610
    },
    {
      "epoch": 35.19,
      "learning_rate": 1.5307702265372168e-05,
      "loss": 0.2001,
      "step": 90620
    },
    {
      "epoch": 35.2,
      "learning_rate": 1.530718446601942e-05,
      "loss": 0.0817,
      "step": 90630
    },
    {
      "epoch": 35.2,
      "learning_rate": 1.5306666666666667e-05,
      "loss": 0.0086,
      "step": 90640
    },
    {
      "epoch": 35.2,
      "learning_rate": 1.5306148867313918e-05,
      "loss": 0.0195,
      "step": 90650
    },
    {
      "epoch": 35.21,
      "learning_rate": 1.5305631067961166e-05,
      "loss": 0.0041,
      "step": 90660
    },
    {
      "epoch": 35.21,
      "learning_rate": 1.5305113268608417e-05,
      "loss": 0.0453,
      "step": 90670
    },
    {
      "epoch": 35.22,
      "learning_rate": 1.5304595469255665e-05,
      "loss": 0.0071,
      "step": 90680
    },
    {
      "epoch": 35.22,
      "learning_rate": 1.5304077669902913e-05,
      "loss": 0.0335,
      "step": 90690
    },
    {
      "epoch": 35.22,
      "learning_rate": 1.530355987055016e-05,
      "loss": 0.0998,
      "step": 90700
    },
    {
      "epoch": 35.23,
      "learning_rate": 1.5303042071197413e-05,
      "loss": 0.1725,
      "step": 90710
    },
    {
      "epoch": 35.23,
      "learning_rate": 1.530252427184466e-05,
      "loss": 0.1592,
      "step": 90720
    },
    {
      "epoch": 35.23,
      "learning_rate": 1.5302006472491912e-05,
      "loss": 0.1254,
      "step": 90730
    },
    {
      "epoch": 35.24,
      "learning_rate": 1.530148867313916e-05,
      "loss": 0.1049,
      "step": 90740
    },
    {
      "epoch": 35.24,
      "learning_rate": 1.5300970873786408e-05,
      "loss": 0.0797,
      "step": 90750
    },
    {
      "epoch": 35.25,
      "learning_rate": 1.530045307443366e-05,
      "loss": 0.0189,
      "step": 90760
    },
    {
      "epoch": 35.25,
      "learning_rate": 1.5299935275080907e-05,
      "loss": 0.1368,
      "step": 90770
    },
    {
      "epoch": 35.25,
      "learning_rate": 1.5299417475728155e-05,
      "loss": 0.1494,
      "step": 90780
    },
    {
      "epoch": 35.26,
      "learning_rate": 1.5298899676375406e-05,
      "loss": 0.0352,
      "step": 90790
    },
    {
      "epoch": 35.26,
      "learning_rate": 1.5298381877022654e-05,
      "loss": 0.0026,
      "step": 90800
    },
    {
      "epoch": 35.27,
      "learning_rate": 1.5297864077669905e-05,
      "loss": 0.07,
      "step": 90810
    },
    {
      "epoch": 35.27,
      "learning_rate": 1.5297346278317153e-05,
      "loss": 0.0979,
      "step": 90820
    },
    {
      "epoch": 35.27,
      "learning_rate": 1.52968284789644e-05,
      "loss": 0.209,
      "step": 90830
    },
    {
      "epoch": 35.28,
      "learning_rate": 1.5296310679611653e-05,
      "loss": 0.1864,
      "step": 90840
    },
    {
      "epoch": 35.28,
      "learning_rate": 1.52957928802589e-05,
      "loss": 0.098,
      "step": 90850
    },
    {
      "epoch": 35.29,
      "learning_rate": 1.5295275080906152e-05,
      "loss": 0.0288,
      "step": 90860
    },
    {
      "epoch": 35.29,
      "learning_rate": 1.52947572815534e-05,
      "loss": 0.143,
      "step": 90870
    },
    {
      "epoch": 35.29,
      "learning_rate": 1.5294239482200648e-05,
      "loss": 0.0423,
      "step": 90880
    },
    {
      "epoch": 35.3,
      "learning_rate": 1.52937216828479e-05,
      "loss": 0.0238,
      "step": 90890
    },
    {
      "epoch": 35.3,
      "learning_rate": 1.5293203883495147e-05,
      "loss": 0.1259,
      "step": 90900
    },
    {
      "epoch": 35.3,
      "learning_rate": 1.5292686084142395e-05,
      "loss": 0.1864,
      "step": 90910
    },
    {
      "epoch": 35.31,
      "learning_rate": 1.5292168284789646e-05,
      "loss": 0.0438,
      "step": 90920
    },
    {
      "epoch": 35.31,
      "learning_rate": 1.5291650485436894e-05,
      "loss": 0.0798,
      "step": 90930
    },
    {
      "epoch": 35.32,
      "learning_rate": 1.5291132686084145e-05,
      "loss": 0.1231,
      "step": 90940
    },
    {
      "epoch": 35.32,
      "learning_rate": 1.5290614886731393e-05,
      "loss": 0.113,
      "step": 90950
    },
    {
      "epoch": 35.32,
      "learning_rate": 1.529009708737864e-05,
      "loss": 0.0624,
      "step": 90960
    },
    {
      "epoch": 35.33,
      "learning_rate": 1.5289579288025892e-05,
      "loss": 0.2431,
      "step": 90970
    },
    {
      "epoch": 35.33,
      "learning_rate": 1.528906148867314e-05,
      "loss": 0.1014,
      "step": 90980
    },
    {
      "epoch": 35.34,
      "learning_rate": 1.528854368932039e-05,
      "loss": 0.128,
      "step": 90990
    },
    {
      "epoch": 35.34,
      "learning_rate": 1.528802588996764e-05,
      "loss": 0.054,
      "step": 91000
    },
    {
      "epoch": 35.34,
      "learning_rate": 1.5287508090614888e-05,
      "loss": 0.0618,
      "step": 91010
    },
    {
      "epoch": 35.35,
      "learning_rate": 1.528699029126214e-05,
      "loss": 0.0382,
      "step": 91020
    },
    {
      "epoch": 35.35,
      "learning_rate": 1.5286472491909387e-05,
      "loss": 0.0492,
      "step": 91030
    },
    {
      "epoch": 35.36,
      "learning_rate": 1.5285954692556635e-05,
      "loss": 0.0525,
      "step": 91040
    },
    {
      "epoch": 35.36,
      "learning_rate": 1.5285436893203886e-05,
      "loss": 0.0844,
      "step": 91050
    },
    {
      "epoch": 35.36,
      "learning_rate": 1.5284919093851134e-05,
      "loss": 0.0363,
      "step": 91060
    },
    {
      "epoch": 35.37,
      "learning_rate": 1.5284401294498382e-05,
      "loss": 0.0989,
      "step": 91070
    },
    {
      "epoch": 35.37,
      "learning_rate": 1.5283883495145633e-05,
      "loss": 0.1451,
      "step": 91080
    },
    {
      "epoch": 35.37,
      "learning_rate": 1.528336569579288e-05,
      "loss": 0.0177,
      "step": 91090
    },
    {
      "epoch": 35.38,
      "learning_rate": 1.5282847896440132e-05,
      "loss": 0.0049,
      "step": 91100
    },
    {
      "epoch": 35.38,
      "learning_rate": 1.528233009708738e-05,
      "loss": 0.0704,
      "step": 91110
    },
    {
      "epoch": 35.39,
      "learning_rate": 1.5281812297734628e-05,
      "loss": 0.1944,
      "step": 91120
    },
    {
      "epoch": 35.39,
      "learning_rate": 1.5281294498381876e-05,
      "loss": 0.0591,
      "step": 91130
    },
    {
      "epoch": 35.39,
      "learning_rate": 1.5280776699029128e-05,
      "loss": 0.0298,
      "step": 91140
    },
    {
      "epoch": 35.4,
      "learning_rate": 1.5280258899676375e-05,
      "loss": 0.0935,
      "step": 91150
    },
    {
      "epoch": 35.4,
      "learning_rate": 1.5279741100323627e-05,
      "loss": 0.0404,
      "step": 91160
    },
    {
      "epoch": 35.41,
      "learning_rate": 1.5279223300970875e-05,
      "loss": 0.047,
      "step": 91170
    },
    {
      "epoch": 35.41,
      "learning_rate": 1.5278705501618126e-05,
      "loss": 0.1165,
      "step": 91180
    },
    {
      "epoch": 35.41,
      "learning_rate": 1.5278187702265374e-05,
      "loss": 0.1042,
      "step": 91190
    },
    {
      "epoch": 35.42,
      "learning_rate": 1.5277669902912622e-05,
      "loss": 0.0404,
      "step": 91200
    },
    {
      "epoch": 35.42,
      "learning_rate": 1.527715210355987e-05,
      "loss": 0.2282,
      "step": 91210
    },
    {
      "epoch": 35.43,
      "learning_rate": 1.527663430420712e-05,
      "loss": 0.1435,
      "step": 91220
    },
    {
      "epoch": 35.43,
      "learning_rate": 1.527611650485437e-05,
      "loss": 0.1177,
      "step": 91230
    },
    {
      "epoch": 35.43,
      "learning_rate": 1.527559870550162e-05,
      "loss": 0.1038,
      "step": 91240
    },
    {
      "epoch": 35.44,
      "learning_rate": 1.5275080906148868e-05,
      "loss": 0.0036,
      "step": 91250
    },
    {
      "epoch": 35.44,
      "learning_rate": 1.527456310679612e-05,
      "loss": 0.04,
      "step": 91260
    },
    {
      "epoch": 35.44,
      "learning_rate": 1.5274045307443367e-05,
      "loss": 0.0638,
      "step": 91270
    },
    {
      "epoch": 35.45,
      "learning_rate": 1.5273527508090615e-05,
      "loss": 0.0416,
      "step": 91280
    },
    {
      "epoch": 35.45,
      "learning_rate": 1.5273009708737863e-05,
      "loss": 0.0266,
      "step": 91290
    },
    {
      "epoch": 35.46,
      "learning_rate": 1.5272491909385115e-05,
      "loss": 0.206,
      "step": 91300
    },
    {
      "epoch": 35.46,
      "learning_rate": 1.5271974110032363e-05,
      "loss": 0.0846,
      "step": 91310
    },
    {
      "epoch": 35.46,
      "learning_rate": 1.5271456310679614e-05,
      "loss": 0.2436,
      "step": 91320
    },
    {
      "epoch": 35.47,
      "learning_rate": 1.5270938511326862e-05,
      "loss": 0.0828,
      "step": 91330
    },
    {
      "epoch": 35.47,
      "learning_rate": 1.5270420711974113e-05,
      "loss": 0.1973,
      "step": 91340
    },
    {
      "epoch": 35.48,
      "learning_rate": 1.526990291262136e-05,
      "loss": 0.2494,
      "step": 91350
    },
    {
      "epoch": 35.48,
      "learning_rate": 1.526938511326861e-05,
      "loss": 0.0325,
      "step": 91360
    },
    {
      "epoch": 35.48,
      "learning_rate": 1.5268867313915857e-05,
      "loss": 0.0661,
      "step": 91370
    },
    {
      "epoch": 35.49,
      "learning_rate": 1.5268349514563108e-05,
      "loss": 0.064,
      "step": 91380
    },
    {
      "epoch": 35.49,
      "learning_rate": 1.5267831715210356e-05,
      "loss": 0.1749,
      "step": 91390
    },
    {
      "epoch": 35.5,
      "learning_rate": 1.5267313915857607e-05,
      "loss": 0.0979,
      "step": 91400
    },
    {
      "epoch": 35.5,
      "learning_rate": 1.5266796116504855e-05,
      "loss": 0.0339,
      "step": 91410
    },
    {
      "epoch": 35.5,
      "learning_rate": 1.5266278317152107e-05,
      "loss": 0.0488,
      "step": 91420
    },
    {
      "epoch": 35.51,
      "learning_rate": 1.5265760517799355e-05,
      "loss": 0.1321,
      "step": 91430
    },
    {
      "epoch": 35.51,
      "learning_rate": 1.5265242718446602e-05,
      "loss": 0.1388,
      "step": 91440
    },
    {
      "epoch": 35.51,
      "learning_rate": 1.526472491909385e-05,
      "loss": 0.1698,
      "step": 91450
    },
    {
      "epoch": 35.52,
      "learning_rate": 1.5264207119741102e-05,
      "loss": 0.024,
      "step": 91460
    },
    {
      "epoch": 35.52,
      "learning_rate": 1.526368932038835e-05,
      "loss": 0.2111,
      "step": 91470
    },
    {
      "epoch": 35.53,
      "learning_rate": 1.52631715210356e-05,
      "loss": 0.0246,
      "step": 91480
    },
    {
      "epoch": 35.53,
      "learning_rate": 1.526265372168285e-05,
      "loss": 0.044,
      "step": 91490
    },
    {
      "epoch": 35.53,
      "learning_rate": 1.52621359223301e-05,
      "loss": 0.1683,
      "step": 91500
    },
    {
      "epoch": 35.54,
      "learning_rate": 1.5261618122977348e-05,
      "loss": 0.1492,
      "step": 91510
    },
    {
      "epoch": 35.54,
      "learning_rate": 1.5261100323624596e-05,
      "loss": 0.0694,
      "step": 91520
    },
    {
      "epoch": 35.55,
      "learning_rate": 1.5260582524271844e-05,
      "loss": 0.1864,
      "step": 91530
    },
    {
      "epoch": 35.55,
      "learning_rate": 1.5260064724919095e-05,
      "loss": 0.1392,
      "step": 91540
    },
    {
      "epoch": 35.55,
      "learning_rate": 1.5259546925566343e-05,
      "loss": 0.0915,
      "step": 91550
    },
    {
      "epoch": 35.56,
      "learning_rate": 1.5259029126213595e-05,
      "loss": 0.1388,
      "step": 91560
    },
    {
      "epoch": 35.56,
      "learning_rate": 1.5258511326860842e-05,
      "loss": 0.2212,
      "step": 91570
    },
    {
      "epoch": 35.57,
      "learning_rate": 1.5257993527508092e-05,
      "loss": 0.0851,
      "step": 91580
    },
    {
      "epoch": 35.57,
      "learning_rate": 1.5257475728155342e-05,
      "loss": 0.0389,
      "step": 91590
    },
    {
      "epoch": 35.57,
      "learning_rate": 1.525695792880259e-05,
      "loss": 0.1544,
      "step": 91600
    },
    {
      "epoch": 35.58,
      "learning_rate": 1.525644012944984e-05,
      "loss": 0.1926,
      "step": 91610
    },
    {
      "epoch": 35.58,
      "learning_rate": 1.5255922330097089e-05,
      "loss": 0.0887,
      "step": 91620
    },
    {
      "epoch": 35.58,
      "learning_rate": 1.5255404530744338e-05,
      "loss": 0.0462,
      "step": 91630
    },
    {
      "epoch": 35.59,
      "learning_rate": 1.5254886731391586e-05,
      "loss": 0.0546,
      "step": 91640
    },
    {
      "epoch": 35.59,
      "learning_rate": 1.5254368932038836e-05,
      "loss": 0.0607,
      "step": 91650
    },
    {
      "epoch": 35.6,
      "learning_rate": 1.5253851132686086e-05,
      "loss": 0.153,
      "step": 91660
    },
    {
      "epoch": 35.6,
      "learning_rate": 1.5253333333333335e-05,
      "loss": 0.042,
      "step": 91670
    },
    {
      "epoch": 35.6,
      "learning_rate": 1.5252815533980583e-05,
      "loss": 0.0588,
      "step": 91680
    },
    {
      "epoch": 35.61,
      "learning_rate": 1.5252297734627833e-05,
      "loss": 0.2418,
      "step": 91690
    },
    {
      "epoch": 35.61,
      "learning_rate": 1.5251779935275082e-05,
      "loss": 0.1063,
      "step": 91700
    },
    {
      "epoch": 35.62,
      "learning_rate": 1.5251262135922332e-05,
      "loss": 0.1134,
      "step": 91710
    },
    {
      "epoch": 35.62,
      "learning_rate": 1.525074433656958e-05,
      "loss": 0.098,
      "step": 91720
    },
    {
      "epoch": 35.62,
      "learning_rate": 1.525022653721683e-05,
      "loss": 0.1643,
      "step": 91730
    },
    {
      "epoch": 35.63,
      "learning_rate": 1.524970873786408e-05,
      "loss": 0.0281,
      "step": 91740
    },
    {
      "epoch": 35.63,
      "learning_rate": 1.5249190938511329e-05,
      "loss": 0.0417,
      "step": 91750
    },
    {
      "epoch": 35.63,
      "learning_rate": 1.5248673139158577e-05,
      "loss": 0.076,
      "step": 91760
    },
    {
      "epoch": 35.64,
      "learning_rate": 1.5248155339805826e-05,
      "loss": 0.2033,
      "step": 91770
    },
    {
      "epoch": 35.64,
      "learning_rate": 1.5247637540453076e-05,
      "loss": 0.2248,
      "step": 91780
    },
    {
      "epoch": 35.65,
      "learning_rate": 1.5247119741100324e-05,
      "loss": 0.0956,
      "step": 91790
    },
    {
      "epoch": 35.65,
      "learning_rate": 1.5246601941747573e-05,
      "loss": 0.1026,
      "step": 91800
    },
    {
      "epoch": 35.65,
      "learning_rate": 1.5246084142394823e-05,
      "loss": 0.1232,
      "step": 91810
    },
    {
      "epoch": 35.66,
      "learning_rate": 1.5245566343042073e-05,
      "loss": 0.0097,
      "step": 91820
    },
    {
      "epoch": 35.66,
      "learning_rate": 1.5245048543689322e-05,
      "loss": 0.0839,
      "step": 91830
    },
    {
      "epoch": 35.67,
      "learning_rate": 1.524453074433657e-05,
      "loss": 0.0494,
      "step": 91840
    },
    {
      "epoch": 35.67,
      "learning_rate": 1.524401294498382e-05,
      "loss": 0.0427,
      "step": 91850
    },
    {
      "epoch": 35.67,
      "learning_rate": 1.524349514563107e-05,
      "loss": 0.1583,
      "step": 91860
    },
    {
      "epoch": 35.68,
      "learning_rate": 1.5242977346278317e-05,
      "loss": 0.0148,
      "step": 91870
    },
    {
      "epoch": 35.68,
      "learning_rate": 1.5242459546925567e-05,
      "loss": 0.1202,
      "step": 91880
    },
    {
      "epoch": 35.69,
      "learning_rate": 1.5241941747572817e-05,
      "loss": 0.134,
      "step": 91890
    },
    {
      "epoch": 35.69,
      "learning_rate": 1.5241423948220066e-05,
      "loss": 0.1154,
      "step": 91900
    },
    {
      "epoch": 35.69,
      "learning_rate": 1.5240906148867316e-05,
      "loss": 0.0893,
      "step": 91910
    },
    {
      "epoch": 35.7,
      "learning_rate": 1.5240388349514564e-05,
      "loss": 0.1528,
      "step": 91920
    },
    {
      "epoch": 35.7,
      "learning_rate": 1.5239870550161813e-05,
      "loss": 0.0425,
      "step": 91930
    },
    {
      "epoch": 35.7,
      "learning_rate": 1.5239352750809061e-05,
      "loss": 0.1177,
      "step": 91940
    },
    {
      "epoch": 35.71,
      "learning_rate": 1.5238834951456311e-05,
      "loss": 0.0695,
      "step": 91950
    },
    {
      "epoch": 35.71,
      "learning_rate": 1.523831715210356e-05,
      "loss": 0.0931,
      "step": 91960
    },
    {
      "epoch": 35.72,
      "learning_rate": 1.523779935275081e-05,
      "loss": 0.0507,
      "step": 91970
    },
    {
      "epoch": 35.72,
      "learning_rate": 1.523728155339806e-05,
      "loss": 0.0183,
      "step": 91980
    },
    {
      "epoch": 35.72,
      "learning_rate": 1.523676375404531e-05,
      "loss": 0.058,
      "step": 91990
    },
    {
      "epoch": 35.73,
      "learning_rate": 1.5236245954692559e-05,
      "loss": 0.2062,
      "step": 92000
    },
    {
      "epoch": 35.73,
      "learning_rate": 1.5235728155339807e-05,
      "loss": 0.0869,
      "step": 92010
    },
    {
      "epoch": 35.74,
      "learning_rate": 1.5235210355987055e-05,
      "loss": 0.0437,
      "step": 92020
    },
    {
      "epoch": 35.74,
      "learning_rate": 1.5234692556634305e-05,
      "loss": 0.1171,
      "step": 92030
    },
    {
      "epoch": 35.74,
      "learning_rate": 1.5234174757281554e-05,
      "loss": 0.0176,
      "step": 92040
    },
    {
      "epoch": 35.75,
      "learning_rate": 1.5233656957928804e-05,
      "loss": 0.0372,
      "step": 92050
    },
    {
      "epoch": 35.75,
      "learning_rate": 1.5233139158576053e-05,
      "loss": 0.0092,
      "step": 92060
    },
    {
      "epoch": 35.76,
      "learning_rate": 1.5232621359223303e-05,
      "loss": 0.0063,
      "step": 92070
    },
    {
      "epoch": 35.76,
      "learning_rate": 1.5232103559870553e-05,
      "loss": 0.1541,
      "step": 92080
    },
    {
      "epoch": 35.76,
      "learning_rate": 1.5231585760517799e-05,
      "loss": 0.1218,
      "step": 92090
    },
    {
      "epoch": 35.77,
      "learning_rate": 1.5231067961165048e-05,
      "loss": 0.0731,
      "step": 92100
    },
    {
      "epoch": 35.77,
      "learning_rate": 1.5230550161812298e-05,
      "loss": 0.1774,
      "step": 92110
    },
    {
      "epoch": 35.77,
      "learning_rate": 1.5230032362459548e-05,
      "loss": 0.0748,
      "step": 92120
    },
    {
      "epoch": 35.78,
      "learning_rate": 1.5229514563106797e-05,
      "loss": 0.1161,
      "step": 92130
    },
    {
      "epoch": 35.78,
      "learning_rate": 1.5228996763754047e-05,
      "loss": 0.195,
      "step": 92140
    },
    {
      "epoch": 35.79,
      "learning_rate": 1.5228478964401297e-05,
      "loss": 0.1248,
      "step": 92150
    },
    {
      "epoch": 35.79,
      "learning_rate": 1.5227961165048546e-05,
      "loss": 0.0818,
      "step": 92160
    },
    {
      "epoch": 35.79,
      "learning_rate": 1.5227443365695792e-05,
      "loss": 0.0044,
      "step": 92170
    },
    {
      "epoch": 35.8,
      "learning_rate": 1.5226925566343042e-05,
      "loss": 0.0683,
      "step": 92180
    },
    {
      "epoch": 35.8,
      "learning_rate": 1.5226407766990292e-05,
      "loss": 0.0857,
      "step": 92190
    },
    {
      "epoch": 35.81,
      "learning_rate": 1.5225889967637541e-05,
      "loss": 0.0591,
      "step": 92200
    },
    {
      "epoch": 35.81,
      "learning_rate": 1.5225372168284791e-05,
      "loss": 0.0524,
      "step": 92210
    },
    {
      "epoch": 35.81,
      "learning_rate": 1.522485436893204e-05,
      "loss": 0.1372,
      "step": 92220
    },
    {
      "epoch": 35.82,
      "learning_rate": 1.522433656957929e-05,
      "loss": 0.0102,
      "step": 92230
    },
    {
      "epoch": 35.82,
      "learning_rate": 1.522381877022654e-05,
      "loss": 0.0712,
      "step": 92240
    },
    {
      "epoch": 35.83,
      "learning_rate": 1.5223300970873786e-05,
      "loss": 0.2411,
      "step": 92250
    },
    {
      "epoch": 35.83,
      "learning_rate": 1.5222783171521036e-05,
      "loss": 0.2539,
      "step": 92260
    },
    {
      "epoch": 35.83,
      "learning_rate": 1.5222265372168285e-05,
      "loss": 0.0092,
      "step": 92270
    },
    {
      "epoch": 35.84,
      "learning_rate": 1.5221747572815535e-05,
      "loss": 0.0935,
      "step": 92280
    },
    {
      "epoch": 35.84,
      "learning_rate": 1.5221229773462784e-05,
      "loss": 0.2267,
      "step": 92290
    },
    {
      "epoch": 35.84,
      "learning_rate": 1.5220711974110034e-05,
      "loss": 0.0235,
      "step": 92300
    },
    {
      "epoch": 35.85,
      "learning_rate": 1.5220194174757284e-05,
      "loss": 0.0207,
      "step": 92310
    },
    {
      "epoch": 35.85,
      "learning_rate": 1.5219676375404533e-05,
      "loss": 0.0079,
      "step": 92320
    },
    {
      "epoch": 35.86,
      "learning_rate": 1.521915857605178e-05,
      "loss": 0.1499,
      "step": 92330
    },
    {
      "epoch": 35.86,
      "learning_rate": 1.521864077669903e-05,
      "loss": 0.009,
      "step": 92340
    },
    {
      "epoch": 35.86,
      "learning_rate": 1.5218122977346279e-05,
      "loss": 0.0953,
      "step": 92350
    },
    {
      "epoch": 35.87,
      "learning_rate": 1.5217605177993528e-05,
      "loss": 0.0739,
      "step": 92360
    },
    {
      "epoch": 35.87,
      "learning_rate": 1.5217087378640778e-05,
      "loss": 0.1097,
      "step": 92370
    },
    {
      "epoch": 35.88,
      "learning_rate": 1.5216569579288028e-05,
      "loss": 0.118,
      "step": 92380
    },
    {
      "epoch": 35.88,
      "learning_rate": 1.5216051779935277e-05,
      "loss": 0.0331,
      "step": 92390
    },
    {
      "epoch": 35.88,
      "learning_rate": 1.5215533980582527e-05,
      "loss": 0.1512,
      "step": 92400
    },
    {
      "epoch": 35.89,
      "learning_rate": 1.5215016181229773e-05,
      "loss": 0.0779,
      "step": 92410
    },
    {
      "epoch": 35.89,
      "learning_rate": 1.5214498381877023e-05,
      "loss": 0.008,
      "step": 92420
    },
    {
      "epoch": 35.9,
      "learning_rate": 1.5213980582524272e-05,
      "loss": 0.1453,
      "step": 92430
    },
    {
      "epoch": 35.9,
      "learning_rate": 1.5213462783171522e-05,
      "loss": 0.0238,
      "step": 92440
    },
    {
      "epoch": 35.9,
      "learning_rate": 1.5212944983818772e-05,
      "loss": 0.1026,
      "step": 92450
    },
    {
      "epoch": 35.91,
      "learning_rate": 1.5212427184466021e-05,
      "loss": 0.0979,
      "step": 92460
    },
    {
      "epoch": 35.91,
      "learning_rate": 1.521190938511327e-05,
      "loss": 0.1086,
      "step": 92470
    },
    {
      "epoch": 35.91,
      "learning_rate": 1.521139158576052e-05,
      "loss": 0.0119,
      "step": 92480
    },
    {
      "epoch": 35.92,
      "learning_rate": 1.5210873786407767e-05,
      "loss": 0.0634,
      "step": 92490
    },
    {
      "epoch": 35.92,
      "learning_rate": 1.5210355987055016e-05,
      "loss": 0.1265,
      "step": 92500
    },
    {
      "epoch": 35.93,
      "learning_rate": 1.5209838187702266e-05,
      "loss": 0.1179,
      "step": 92510
    },
    {
      "epoch": 35.93,
      "learning_rate": 1.5209320388349516e-05,
      "loss": 0.1409,
      "step": 92520
    },
    {
      "epoch": 35.93,
      "learning_rate": 1.5208802588996765e-05,
      "loss": 0.1379,
      "step": 92530
    },
    {
      "epoch": 35.94,
      "learning_rate": 1.5208284789644015e-05,
      "loss": 0.1276,
      "step": 92540
    },
    {
      "epoch": 35.94,
      "learning_rate": 1.5207766990291264e-05,
      "loss": 0.0254,
      "step": 92550
    },
    {
      "epoch": 35.95,
      "learning_rate": 1.5207249190938514e-05,
      "loss": 0.1353,
      "step": 92560
    },
    {
      "epoch": 35.95,
      "learning_rate": 1.5206731391585764e-05,
      "loss": 0.0549,
      "step": 92570
    },
    {
      "epoch": 35.95,
      "learning_rate": 1.520621359223301e-05,
      "loss": 0.0162,
      "step": 92580
    },
    {
      "epoch": 35.96,
      "learning_rate": 1.520569579288026e-05,
      "loss": 0.0698,
      "step": 92590
    },
    {
      "epoch": 35.96,
      "learning_rate": 1.5205177993527509e-05,
      "loss": 0.0192,
      "step": 92600
    },
    {
      "epoch": 35.97,
      "learning_rate": 1.5204660194174759e-05,
      "loss": 0.0527,
      "step": 92610
    },
    {
      "epoch": 35.97,
      "learning_rate": 1.5204142394822008e-05,
      "loss": 0.1091,
      "step": 92620
    },
    {
      "epoch": 35.97,
      "learning_rate": 1.5203624595469258e-05,
      "loss": 0.0307,
      "step": 92630
    },
    {
      "epoch": 35.98,
      "learning_rate": 1.5203106796116508e-05,
      "loss": 0.1475,
      "step": 92640
    },
    {
      "epoch": 35.98,
      "learning_rate": 1.5202588996763757e-05,
      "loss": 0.1601,
      "step": 92650
    },
    {
      "epoch": 35.98,
      "learning_rate": 1.5202071197411003e-05,
      "loss": 0.0768,
      "step": 92660
    },
    {
      "epoch": 35.99,
      "learning_rate": 1.5201553398058253e-05,
      "loss": 0.1309,
      "step": 92670
    },
    {
      "epoch": 35.99,
      "learning_rate": 1.5201035598705503e-05,
      "loss": 0.0855,
      "step": 92680
    },
    {
      "epoch": 36.0,
      "learning_rate": 1.5200517799352752e-05,
      "loss": 0.0709,
      "step": 92690
    },
    {
      "epoch": 36.0,
      "learning_rate": 1.5200000000000002e-05,
      "loss": 0.0617,
      "step": 92700
    },
    {
      "epoch": 36.0,
      "eval_accuracy": 0.9474552957359009,
      "eval_loss": 0.3130703270435333,
      "eval_runtime": 8.2696,
      "eval_samples_per_second": 439.562,
      "eval_steps_per_second": 55.021,
      "step": 92700
    },
    {
      "epoch": 36.0,
      "learning_rate": 1.5199482200647251e-05,
      "loss": 0.0318,
      "step": 92710
    },
    {
      "epoch": 36.01,
      "learning_rate": 1.5198964401294501e-05,
      "loss": 0.031,
      "step": 92720
    },
    {
      "epoch": 36.01,
      "learning_rate": 1.5198446601941749e-05,
      "loss": 0.2634,
      "step": 92730
    },
    {
      "epoch": 36.02,
      "learning_rate": 1.5197928802588997e-05,
      "loss": 0.0569,
      "step": 92740
    },
    {
      "epoch": 36.02,
      "learning_rate": 1.5197411003236247e-05,
      "loss": 0.0431,
      "step": 92750
    },
    {
      "epoch": 36.02,
      "learning_rate": 1.5196893203883496e-05,
      "loss": 0.0897,
      "step": 92760
    },
    {
      "epoch": 36.03,
      "learning_rate": 1.5196375404530746e-05,
      "loss": 0.058,
      "step": 92770
    },
    {
      "epoch": 36.03,
      "learning_rate": 1.5195857605177995e-05,
      "loss": 0.0784,
      "step": 92780
    },
    {
      "epoch": 36.03,
      "learning_rate": 1.5195339805825245e-05,
      "loss": 0.0403,
      "step": 92790
    },
    {
      "epoch": 36.04,
      "learning_rate": 1.5194822006472495e-05,
      "loss": 0.0583,
      "step": 92800
    },
    {
      "epoch": 36.04,
      "learning_rate": 1.5194304207119743e-05,
      "loss": 0.12,
      "step": 92810
    },
    {
      "epoch": 36.05,
      "learning_rate": 1.519378640776699e-05,
      "loss": 0.1237,
      "step": 92820
    },
    {
      "epoch": 36.05,
      "learning_rate": 1.519326860841424e-05,
      "loss": 0.1155,
      "step": 92830
    },
    {
      "epoch": 36.05,
      "learning_rate": 1.519275080906149e-05,
      "loss": 0.0005,
      "step": 92840
    },
    {
      "epoch": 36.06,
      "learning_rate": 1.519223300970874e-05,
      "loss": 0.0209,
      "step": 92850
    },
    {
      "epoch": 36.06,
      "learning_rate": 1.5191715210355989e-05,
      "loss": 0.064,
      "step": 92860
    },
    {
      "epoch": 36.07,
      "learning_rate": 1.5191197411003239e-05,
      "loss": 0.0345,
      "step": 92870
    },
    {
      "epoch": 36.07,
      "learning_rate": 1.5190679611650487e-05,
      "loss": 0.3539,
      "step": 92880
    },
    {
      "epoch": 36.07,
      "learning_rate": 1.5190161812297736e-05,
      "loss": 0.3365,
      "step": 92890
    },
    {
      "epoch": 36.08,
      "learning_rate": 1.5189644012944984e-05,
      "loss": 0.03,
      "step": 92900
    },
    {
      "epoch": 36.08,
      "learning_rate": 1.5189126213592234e-05,
      "loss": 0.0983,
      "step": 92910
    },
    {
      "epoch": 36.09,
      "learning_rate": 1.5188608414239483e-05,
      "loss": 0.0278,
      "step": 92920
    },
    {
      "epoch": 36.09,
      "learning_rate": 1.5188090614886733e-05,
      "loss": 0.0401,
      "step": 92930
    },
    {
      "epoch": 36.09,
      "learning_rate": 1.5187572815533983e-05,
      "loss": 0.0419,
      "step": 92940
    },
    {
      "epoch": 36.1,
      "learning_rate": 1.5187055016181232e-05,
      "loss": 0.0264,
      "step": 92950
    },
    {
      "epoch": 36.1,
      "learning_rate": 1.518653721682848e-05,
      "loss": 0.0326,
      "step": 92960
    },
    {
      "epoch": 36.1,
      "learning_rate": 1.518601941747573e-05,
      "loss": 0.0778,
      "step": 92970
    },
    {
      "epoch": 36.11,
      "learning_rate": 1.5185501618122978e-05,
      "loss": 0.0923,
      "step": 92980
    },
    {
      "epoch": 36.11,
      "learning_rate": 1.5184983818770227e-05,
      "loss": 0.0637,
      "step": 92990
    },
    {
      "epoch": 36.12,
      "learning_rate": 1.5184466019417477e-05,
      "loss": 0.0541,
      "step": 93000
    },
    {
      "epoch": 36.12,
      "learning_rate": 1.5183948220064726e-05,
      "loss": 0.1683,
      "step": 93010
    },
    {
      "epoch": 36.12,
      "learning_rate": 1.5183430420711976e-05,
      "loss": 0.0294,
      "step": 93020
    },
    {
      "epoch": 36.13,
      "learning_rate": 1.5182912621359224e-05,
      "loss": 0.0962,
      "step": 93030
    },
    {
      "epoch": 36.13,
      "learning_rate": 1.5182394822006474e-05,
      "loss": 0.0985,
      "step": 93040
    },
    {
      "epoch": 36.14,
      "learning_rate": 1.5181877022653723e-05,
      "loss": 0.0785,
      "step": 93050
    },
    {
      "epoch": 36.14,
      "learning_rate": 1.5181359223300973e-05,
      "loss": 0.1859,
      "step": 93060
    },
    {
      "epoch": 36.14,
      "learning_rate": 1.518084142394822e-05,
      "loss": 0.2668,
      "step": 93070
    },
    {
      "epoch": 36.15,
      "learning_rate": 1.518032362459547e-05,
      "loss": 0.0589,
      "step": 93080
    },
    {
      "epoch": 36.15,
      "learning_rate": 1.517980582524272e-05,
      "loss": 0.1822,
      "step": 93090
    },
    {
      "epoch": 36.16,
      "learning_rate": 1.517928802588997e-05,
      "loss": 0.0393,
      "step": 93100
    },
    {
      "epoch": 36.16,
      "learning_rate": 1.5178770226537218e-05,
      "loss": 0.0168,
      "step": 93110
    },
    {
      "epoch": 36.16,
      "learning_rate": 1.5178252427184467e-05,
      "loss": 0.0191,
      "step": 93120
    },
    {
      "epoch": 36.17,
      "learning_rate": 1.5177734627831717e-05,
      "loss": 0.1227,
      "step": 93130
    },
    {
      "epoch": 36.17,
      "learning_rate": 1.5177216828478966e-05,
      "loss": 0.1024,
      "step": 93140
    },
    {
      "epoch": 36.17,
      "learning_rate": 1.5176699029126214e-05,
      "loss": 0.0283,
      "step": 93150
    },
    {
      "epoch": 36.18,
      "learning_rate": 1.5176181229773464e-05,
      "loss": 0.0716,
      "step": 93160
    },
    {
      "epoch": 36.18,
      "learning_rate": 1.5175663430420714e-05,
      "loss": 0.0188,
      "step": 93170
    },
    {
      "epoch": 36.19,
      "learning_rate": 1.5175145631067963e-05,
      "loss": 0.0383,
      "step": 93180
    },
    {
      "epoch": 36.19,
      "learning_rate": 1.5174627831715211e-05,
      "loss": 0.1754,
      "step": 93190
    },
    {
      "epoch": 36.19,
      "learning_rate": 1.517411003236246e-05,
      "loss": 0.0824,
      "step": 93200
    },
    {
      "epoch": 36.2,
      "learning_rate": 1.517359223300971e-05,
      "loss": 0.0598,
      "step": 93210
    },
    {
      "epoch": 36.2,
      "learning_rate": 1.517307443365696e-05,
      "loss": 0.1186,
      "step": 93220
    },
    {
      "epoch": 36.21,
      "learning_rate": 1.5172556634304208e-05,
      "loss": 0.0373,
      "step": 93230
    },
    {
      "epoch": 36.21,
      "learning_rate": 1.5172038834951458e-05,
      "loss": 0.0238,
      "step": 93240
    },
    {
      "epoch": 36.21,
      "learning_rate": 1.5171521035598707e-05,
      "loss": 0.0546,
      "step": 93250
    },
    {
      "epoch": 36.22,
      "learning_rate": 1.5171003236245955e-05,
      "loss": 0.0863,
      "step": 93260
    },
    {
      "epoch": 36.22,
      "learning_rate": 1.5170485436893205e-05,
      "loss": 0.0704,
      "step": 93270
    },
    {
      "epoch": 36.23,
      "learning_rate": 1.5169967637540454e-05,
      "loss": 0.1447,
      "step": 93280
    },
    {
      "epoch": 36.23,
      "learning_rate": 1.5169449838187704e-05,
      "loss": 0.017,
      "step": 93290
    },
    {
      "epoch": 36.23,
      "learning_rate": 1.5168932038834954e-05,
      "loss": 0.1501,
      "step": 93300
    },
    {
      "epoch": 36.24,
      "learning_rate": 1.5168414239482201e-05,
      "loss": 0.0822,
      "step": 93310
    },
    {
      "epoch": 36.24,
      "learning_rate": 1.5167896440129451e-05,
      "loss": 0.1364,
      "step": 93320
    },
    {
      "epoch": 36.24,
      "learning_rate": 1.51673786407767e-05,
      "loss": 0.0464,
      "step": 93330
    },
    {
      "epoch": 36.25,
      "learning_rate": 1.5166860841423949e-05,
      "loss": 0.11,
      "step": 93340
    },
    {
      "epoch": 36.25,
      "learning_rate": 1.5166343042071198e-05,
      "loss": 0.1619,
      "step": 93350
    },
    {
      "epoch": 36.26,
      "learning_rate": 1.5165825242718448e-05,
      "loss": 0.078,
      "step": 93360
    },
    {
      "epoch": 36.26,
      "learning_rate": 1.5165307443365697e-05,
      "loss": 0.1987,
      "step": 93370
    },
    {
      "epoch": 36.26,
      "learning_rate": 1.5164789644012947e-05,
      "loss": 0.0572,
      "step": 93380
    },
    {
      "epoch": 36.27,
      "learning_rate": 1.5164271844660195e-05,
      "loss": 0.0573,
      "step": 93390
    },
    {
      "epoch": 36.27,
      "learning_rate": 1.5163754045307445e-05,
      "loss": 0.0801,
      "step": 93400
    },
    {
      "epoch": 36.28,
      "learning_rate": 1.5163236245954693e-05,
      "loss": 0.1024,
      "step": 93410
    },
    {
      "epoch": 36.28,
      "learning_rate": 1.5162718446601942e-05,
      "loss": 0.0838,
      "step": 93420
    },
    {
      "epoch": 36.28,
      "learning_rate": 1.5162200647249192e-05,
      "loss": 0.1235,
      "step": 93430
    },
    {
      "epoch": 36.29,
      "learning_rate": 1.5161682847896441e-05,
      "loss": 0.0799,
      "step": 93440
    },
    {
      "epoch": 36.29,
      "learning_rate": 1.5161165048543691e-05,
      "loss": 0.1574,
      "step": 93450
    },
    {
      "epoch": 36.3,
      "learning_rate": 1.516064724919094e-05,
      "loss": 0.1974,
      "step": 93460
    },
    {
      "epoch": 36.3,
      "learning_rate": 1.5160129449838189e-05,
      "loss": 0.0769,
      "step": 93470
    },
    {
      "epoch": 36.3,
      "learning_rate": 1.5159611650485438e-05,
      "loss": 0.1302,
      "step": 93480
    },
    {
      "epoch": 36.31,
      "learning_rate": 1.5159093851132686e-05,
      "loss": 0.2974,
      "step": 93490
    },
    {
      "epoch": 36.31,
      "learning_rate": 1.5158576051779936e-05,
      "loss": 0.1041,
      "step": 93500
    },
    {
      "epoch": 36.31,
      "learning_rate": 1.5158058252427185e-05,
      "loss": 0.0629,
      "step": 93510
    },
    {
      "epoch": 36.32,
      "learning_rate": 1.5157540453074435e-05,
      "loss": 0.0975,
      "step": 93520
    },
    {
      "epoch": 36.32,
      "learning_rate": 1.5157022653721685e-05,
      "loss": 0.1603,
      "step": 93530
    },
    {
      "epoch": 36.33,
      "learning_rate": 1.5156504854368934e-05,
      "loss": 0.1948,
      "step": 93540
    },
    {
      "epoch": 36.33,
      "learning_rate": 1.5155987055016182e-05,
      "loss": 0.0159,
      "step": 93550
    },
    {
      "epoch": 36.33,
      "learning_rate": 1.515546925566343e-05,
      "loss": 0.0015,
      "step": 93560
    },
    {
      "epoch": 36.34,
      "learning_rate": 1.515495145631068e-05,
      "loss": 0.2686,
      "step": 93570
    },
    {
      "epoch": 36.34,
      "learning_rate": 1.515443365695793e-05,
      "loss": 0.147,
      "step": 93580
    },
    {
      "epoch": 36.35,
      "learning_rate": 1.5153915857605179e-05,
      "loss": 0.054,
      "step": 93590
    },
    {
      "epoch": 36.35,
      "learning_rate": 1.5153398058252429e-05,
      "loss": 0.062,
      "step": 93600
    },
    {
      "epoch": 36.35,
      "learning_rate": 1.5152880258899678e-05,
      "loss": 0.1067,
      "step": 93610
    },
    {
      "epoch": 36.36,
      "learning_rate": 1.5152362459546928e-05,
      "loss": 0.0412,
      "step": 93620
    },
    {
      "epoch": 36.36,
      "learning_rate": 1.5151844660194177e-05,
      "loss": 0.1015,
      "step": 93630
    },
    {
      "epoch": 36.37,
      "learning_rate": 1.5151326860841424e-05,
      "loss": 0.0682,
      "step": 93640
    },
    {
      "epoch": 36.37,
      "learning_rate": 1.5150809061488673e-05,
      "loss": 0.0034,
      "step": 93650
    },
    {
      "epoch": 36.37,
      "learning_rate": 1.5150291262135923e-05,
      "loss": 0.2758,
      "step": 93660
    },
    {
      "epoch": 36.38,
      "learning_rate": 1.5149773462783172e-05,
      "loss": 0.1171,
      "step": 93670
    },
    {
      "epoch": 36.38,
      "learning_rate": 1.5149255663430422e-05,
      "loss": 0.0711,
      "step": 93680
    },
    {
      "epoch": 36.38,
      "learning_rate": 1.5148737864077672e-05,
      "loss": 0.0192,
      "step": 93690
    },
    {
      "epoch": 36.39,
      "learning_rate": 1.5148220064724921e-05,
      "loss": 0.0894,
      "step": 93700
    },
    {
      "epoch": 36.39,
      "learning_rate": 1.5147702265372171e-05,
      "loss": 0.0952,
      "step": 93710
    },
    {
      "epoch": 36.4,
      "learning_rate": 1.5147184466019417e-05,
      "loss": 0.2131,
      "step": 93720
    },
    {
      "epoch": 36.4,
      "learning_rate": 1.5146666666666667e-05,
      "loss": 0.1252,
      "step": 93730
    },
    {
      "epoch": 36.4,
      "learning_rate": 1.5146148867313916e-05,
      "loss": 0.0923,
      "step": 93740
    },
    {
      "epoch": 36.41,
      "learning_rate": 1.5145631067961166e-05,
      "loss": 0.054,
      "step": 93750
    },
    {
      "epoch": 36.41,
      "learning_rate": 1.5145113268608416e-05,
      "loss": 0.0659,
      "step": 93760
    },
    {
      "epoch": 36.42,
      "learning_rate": 1.5144595469255665e-05,
      "loss": 0.236,
      "step": 93770
    },
    {
      "epoch": 36.42,
      "learning_rate": 1.5144077669902915e-05,
      "loss": 0.2097,
      "step": 93780
    },
    {
      "epoch": 36.42,
      "learning_rate": 1.5143559870550164e-05,
      "loss": 0.1374,
      "step": 93790
    },
    {
      "epoch": 36.43,
      "learning_rate": 1.514304207119741e-05,
      "loss": 0.0839,
      "step": 93800
    },
    {
      "epoch": 36.43,
      "learning_rate": 1.514252427184466e-05,
      "loss": 0.0563,
      "step": 93810
    },
    {
      "epoch": 36.43,
      "learning_rate": 1.514200647249191e-05,
      "loss": 0.1055,
      "step": 93820
    },
    {
      "epoch": 36.44,
      "learning_rate": 1.514148867313916e-05,
      "loss": 0.1542,
      "step": 93830
    },
    {
      "epoch": 36.44,
      "learning_rate": 1.514097087378641e-05,
      "loss": 0.4709,
      "step": 93840
    },
    {
      "epoch": 36.45,
      "learning_rate": 1.5140453074433659e-05,
      "loss": 0.0809,
      "step": 93850
    },
    {
      "epoch": 36.45,
      "learning_rate": 1.5139935275080908e-05,
      "loss": 0.0663,
      "step": 93860
    },
    {
      "epoch": 36.45,
      "learning_rate": 1.5139417475728158e-05,
      "loss": 0.0428,
      "step": 93870
    },
    {
      "epoch": 36.46,
      "learning_rate": 1.5138899676375404e-05,
      "loss": 0.0193,
      "step": 93880
    },
    {
      "epoch": 36.46,
      "learning_rate": 1.5138381877022654e-05,
      "loss": 0.1521,
      "step": 93890
    },
    {
      "epoch": 36.47,
      "learning_rate": 1.5137864077669904e-05,
      "loss": 0.1255,
      "step": 93900
    },
    {
      "epoch": 36.47,
      "learning_rate": 1.5137346278317153e-05,
      "loss": 0.0699,
      "step": 93910
    },
    {
      "epoch": 36.47,
      "learning_rate": 1.5136828478964403e-05,
      "loss": 0.0519,
      "step": 93920
    },
    {
      "epoch": 36.48,
      "learning_rate": 1.5136310679611652e-05,
      "loss": 0.0827,
      "step": 93930
    },
    {
      "epoch": 36.48,
      "learning_rate": 1.5135792880258902e-05,
      "loss": 0.0729,
      "step": 93940
    },
    {
      "epoch": 36.49,
      "learning_rate": 1.5135275080906152e-05,
      "loss": 0.1002,
      "step": 93950
    },
    {
      "epoch": 36.49,
      "learning_rate": 1.5134757281553398e-05,
      "loss": 0.0167,
      "step": 93960
    },
    {
      "epoch": 36.49,
      "learning_rate": 1.5134239482200647e-05,
      "loss": 0.0334,
      "step": 93970
    },
    {
      "epoch": 36.5,
      "learning_rate": 1.5133721682847897e-05,
      "loss": 0.133,
      "step": 93980
    },
    {
      "epoch": 36.5,
      "learning_rate": 1.5133203883495147e-05,
      "loss": 0.3595,
      "step": 93990
    },
    {
      "epoch": 36.5,
      "learning_rate": 1.5132686084142396e-05,
      "loss": 0.1328,
      "step": 94000
    },
    {
      "epoch": 36.51,
      "learning_rate": 1.5132168284789646e-05,
      "loss": 0.1484,
      "step": 94010
    },
    {
      "epoch": 36.51,
      "learning_rate": 1.5131650485436896e-05,
      "loss": 0.0356,
      "step": 94020
    },
    {
      "epoch": 36.52,
      "learning_rate": 1.5131132686084145e-05,
      "loss": 0.1089,
      "step": 94030
    },
    {
      "epoch": 36.52,
      "learning_rate": 1.5130614886731391e-05,
      "loss": 0.1307,
      "step": 94040
    },
    {
      "epoch": 36.52,
      "learning_rate": 1.5130097087378641e-05,
      "loss": 0.0854,
      "step": 94050
    },
    {
      "epoch": 36.53,
      "learning_rate": 1.512957928802589e-05,
      "loss": 0.1309,
      "step": 94060
    },
    {
      "epoch": 36.53,
      "learning_rate": 1.512906148867314e-05,
      "loss": 0.1379,
      "step": 94070
    },
    {
      "epoch": 36.54,
      "learning_rate": 1.512854368932039e-05,
      "loss": 0.0739,
      "step": 94080
    },
    {
      "epoch": 36.54,
      "learning_rate": 1.512802588996764e-05,
      "loss": 0.0581,
      "step": 94090
    },
    {
      "epoch": 36.54,
      "learning_rate": 1.5127508090614889e-05,
      "loss": 0.0426,
      "step": 94100
    },
    {
      "epoch": 36.55,
      "learning_rate": 1.5126990291262139e-05,
      "loss": 0.0306,
      "step": 94110
    },
    {
      "epoch": 36.55,
      "learning_rate": 1.5126472491909385e-05,
      "loss": 0.0463,
      "step": 94120
    },
    {
      "epoch": 36.56,
      "learning_rate": 1.5125954692556635e-05,
      "loss": 0.0458,
      "step": 94130
    },
    {
      "epoch": 36.56,
      "learning_rate": 1.5125436893203884e-05,
      "loss": 0.139,
      "step": 94140
    },
    {
      "epoch": 36.56,
      "learning_rate": 1.5124919093851134e-05,
      "loss": 0.1626,
      "step": 94150
    },
    {
      "epoch": 36.57,
      "learning_rate": 1.5124401294498383e-05,
      "loss": 0.1232,
      "step": 94160
    },
    {
      "epoch": 36.57,
      "learning_rate": 1.5123883495145633e-05,
      "loss": 0.0344,
      "step": 94170
    },
    {
      "epoch": 36.57,
      "learning_rate": 1.5123365695792883e-05,
      "loss": 0.0987,
      "step": 94180
    },
    {
      "epoch": 36.58,
      "learning_rate": 1.5122847896440132e-05,
      "loss": 0.0315,
      "step": 94190
    },
    {
      "epoch": 36.58,
      "learning_rate": 1.512233009708738e-05,
      "loss": 0.1932,
      "step": 94200
    },
    {
      "epoch": 36.59,
      "learning_rate": 1.5121812297734628e-05,
      "loss": 0.0714,
      "step": 94210
    },
    {
      "epoch": 36.59,
      "learning_rate": 1.5121294498381878e-05,
      "loss": 0.0703,
      "step": 94220
    },
    {
      "epoch": 36.59,
      "learning_rate": 1.5120776699029127e-05,
      "loss": 0.1712,
      "step": 94230
    },
    {
      "epoch": 36.6,
      "learning_rate": 1.5120258899676377e-05,
      "loss": 0.1755,
      "step": 94240
    },
    {
      "epoch": 36.6,
      "learning_rate": 1.5119741100323627e-05,
      "loss": 0.1147,
      "step": 94250
    },
    {
      "epoch": 36.61,
      "learning_rate": 1.5119223300970876e-05,
      "loss": 0.0209,
      "step": 94260
    },
    {
      "epoch": 36.61,
      "learning_rate": 1.5118705501618126e-05,
      "loss": 0.0383,
      "step": 94270
    },
    {
      "epoch": 36.61,
      "learning_rate": 1.5118187702265374e-05,
      "loss": 0.0615,
      "step": 94280
    },
    {
      "epoch": 36.62,
      "learning_rate": 1.5117669902912622e-05,
      "loss": 0.1083,
      "step": 94290
    },
    {
      "epoch": 36.62,
      "learning_rate": 1.5117152103559871e-05,
      "loss": 0.0946,
      "step": 94300
    },
    {
      "epoch": 36.63,
      "learning_rate": 1.5116634304207121e-05,
      "loss": 0.0993,
      "step": 94310
    },
    {
      "epoch": 36.63,
      "learning_rate": 1.511611650485437e-05,
      "loss": 0.0487,
      "step": 94320
    },
    {
      "epoch": 36.63,
      "learning_rate": 1.511559870550162e-05,
      "loss": 0.092,
      "step": 94330
    },
    {
      "epoch": 36.64,
      "learning_rate": 1.511508090614887e-05,
      "loss": 0.0882,
      "step": 94340
    },
    {
      "epoch": 36.64,
      "learning_rate": 1.5114563106796118e-05,
      "loss": 0.1186,
      "step": 94350
    },
    {
      "epoch": 36.64,
      "learning_rate": 1.5114045307443367e-05,
      "loss": 0.0575,
      "step": 94360
    },
    {
      "epoch": 36.65,
      "learning_rate": 1.5113527508090615e-05,
      "loss": 0.1148,
      "step": 94370
    },
    {
      "epoch": 36.65,
      "learning_rate": 1.5113009708737865e-05,
      "loss": 0.1553,
      "step": 94380
    },
    {
      "epoch": 36.66,
      "learning_rate": 1.5112491909385114e-05,
      "loss": 0.0683,
      "step": 94390
    },
    {
      "epoch": 36.66,
      "learning_rate": 1.5111974110032364e-05,
      "loss": 0.0664,
      "step": 94400
    },
    {
      "epoch": 36.66,
      "learning_rate": 1.5111456310679614e-05,
      "loss": 0.1435,
      "step": 94410
    },
    {
      "epoch": 36.67,
      "learning_rate": 1.5110938511326863e-05,
      "loss": 0.0857,
      "step": 94420
    },
    {
      "epoch": 36.67,
      "learning_rate": 1.5110420711974111e-05,
      "loss": 0.1608,
      "step": 94430
    },
    {
      "epoch": 36.68,
      "learning_rate": 1.5109902912621361e-05,
      "loss": 0.0275,
      "step": 94440
    },
    {
      "epoch": 36.68,
      "learning_rate": 1.5109385113268609e-05,
      "loss": 0.0518,
      "step": 94450
    },
    {
      "epoch": 36.68,
      "learning_rate": 1.5108867313915858e-05,
      "loss": 0.1279,
      "step": 94460
    },
    {
      "epoch": 36.69,
      "learning_rate": 1.5108349514563108e-05,
      "loss": 0.1128,
      "step": 94470
    },
    {
      "epoch": 36.69,
      "learning_rate": 1.5107831715210358e-05,
      "loss": 0.0688,
      "step": 94480
    },
    {
      "epoch": 36.7,
      "learning_rate": 1.5107313915857607e-05,
      "loss": 0.0494,
      "step": 94490
    },
    {
      "epoch": 36.7,
      "learning_rate": 1.5106796116504855e-05,
      "loss": 0.0097,
      "step": 94500
    },
    {
      "epoch": 36.7,
      "learning_rate": 1.5106278317152105e-05,
      "loss": 0.0653,
      "step": 94510
    },
    {
      "epoch": 36.71,
      "learning_rate": 1.5105760517799354e-05,
      "loss": 0.0618,
      "step": 94520
    },
    {
      "epoch": 36.71,
      "learning_rate": 1.5105242718446602e-05,
      "loss": 0.0772,
      "step": 94530
    },
    {
      "epoch": 36.71,
      "learning_rate": 1.5104724919093852e-05,
      "loss": 0.0782,
      "step": 94540
    },
    {
      "epoch": 36.72,
      "learning_rate": 1.5104207119741102e-05,
      "loss": 0.1542,
      "step": 94550
    },
    {
      "epoch": 36.72,
      "learning_rate": 1.5103689320388351e-05,
      "loss": 0.1254,
      "step": 94560
    },
    {
      "epoch": 36.73,
      "learning_rate": 1.51031715210356e-05,
      "loss": 0.028,
      "step": 94570
    },
    {
      "epoch": 36.73,
      "learning_rate": 1.5102653721682849e-05,
      "loss": 0.116,
      "step": 94580
    },
    {
      "epoch": 36.73,
      "learning_rate": 1.5102135922330098e-05,
      "loss": 0.028,
      "step": 94590
    },
    {
      "epoch": 36.74,
      "learning_rate": 1.5101618122977348e-05,
      "loss": 0.2148,
      "step": 94600
    },
    {
      "epoch": 36.74,
      "learning_rate": 1.5101100323624596e-05,
      "loss": 0.1419,
      "step": 94610
    },
    {
      "epoch": 36.75,
      "learning_rate": 1.5100582524271846e-05,
      "loss": 0.1406,
      "step": 94620
    },
    {
      "epoch": 36.75,
      "learning_rate": 1.5100064724919095e-05,
      "loss": 0.0012,
      "step": 94630
    },
    {
      "epoch": 36.75,
      "learning_rate": 1.5099546925566345e-05,
      "loss": 0.1851,
      "step": 94640
    },
    {
      "epoch": 36.76,
      "learning_rate": 1.5099029126213593e-05,
      "loss": 0.0631,
      "step": 94650
    },
    {
      "epoch": 36.76,
      "learning_rate": 1.5098511326860842e-05,
      "loss": 0.1667,
      "step": 94660
    },
    {
      "epoch": 36.77,
      "learning_rate": 1.5097993527508092e-05,
      "loss": 0.0489,
      "step": 94670
    },
    {
      "epoch": 36.77,
      "learning_rate": 1.5097475728155342e-05,
      "loss": 0.1357,
      "step": 94680
    },
    {
      "epoch": 36.77,
      "learning_rate": 1.509695792880259e-05,
      "loss": 0.0155,
      "step": 94690
    },
    {
      "epoch": 36.78,
      "learning_rate": 1.5096440129449839e-05,
      "loss": 0.0062,
      "step": 94700
    },
    {
      "epoch": 36.78,
      "learning_rate": 1.5095922330097089e-05,
      "loss": 0.0452,
      "step": 94710
    },
    {
      "epoch": 36.78,
      "learning_rate": 1.5095404530744338e-05,
      "loss": 0.1386,
      "step": 94720
    },
    {
      "epoch": 36.79,
      "learning_rate": 1.5094886731391586e-05,
      "loss": 0.0236,
      "step": 94730
    },
    {
      "epoch": 36.79,
      "learning_rate": 1.5094368932038836e-05,
      "loss": 0.051,
      "step": 94740
    },
    {
      "epoch": 36.8,
      "learning_rate": 1.5093851132686085e-05,
      "loss": 0.0336,
      "step": 94750
    },
    {
      "epoch": 36.8,
      "learning_rate": 1.5093333333333335e-05,
      "loss": 0.0387,
      "step": 94760
    },
    {
      "epoch": 36.8,
      "learning_rate": 1.5092815533980585e-05,
      "loss": 0.0035,
      "step": 94770
    },
    {
      "epoch": 36.81,
      "learning_rate": 1.5092297734627833e-05,
      "loss": 0.1151,
      "step": 94780
    },
    {
      "epoch": 36.81,
      "learning_rate": 1.5091779935275082e-05,
      "loss": 0.1709,
      "step": 94790
    },
    {
      "epoch": 36.82,
      "learning_rate": 1.5091262135922332e-05,
      "loss": 0.1503,
      "step": 94800
    },
    {
      "epoch": 36.82,
      "learning_rate": 1.509074433656958e-05,
      "loss": 0.2622,
      "step": 94810
    },
    {
      "epoch": 36.82,
      "learning_rate": 1.509022653721683e-05,
      "loss": 0.1608,
      "step": 94820
    },
    {
      "epoch": 36.83,
      "learning_rate": 1.5089708737864079e-05,
      "loss": 0.0409,
      "step": 94830
    },
    {
      "epoch": 36.83,
      "learning_rate": 1.5089190938511329e-05,
      "loss": 0.1265,
      "step": 94840
    },
    {
      "epoch": 36.83,
      "learning_rate": 1.5088673139158578e-05,
      "loss": 0.0602,
      "step": 94850
    },
    {
      "epoch": 36.84,
      "learning_rate": 1.5088155339805826e-05,
      "loss": 0.1065,
      "step": 94860
    },
    {
      "epoch": 36.84,
      "learning_rate": 1.5087637540453076e-05,
      "loss": 0.2009,
      "step": 94870
    },
    {
      "epoch": 36.85,
      "learning_rate": 1.5087119741100324e-05,
      "loss": 0.1632,
      "step": 94880
    },
    {
      "epoch": 36.85,
      "learning_rate": 1.5086601941747573e-05,
      "loss": 0.0638,
      "step": 94890
    },
    {
      "epoch": 36.85,
      "learning_rate": 1.5086084142394823e-05,
      "loss": 0.1538,
      "step": 94900
    },
    {
      "epoch": 36.86,
      "learning_rate": 1.5085566343042073e-05,
      "loss": 0.1409,
      "step": 94910
    },
    {
      "epoch": 36.86,
      "learning_rate": 1.5085048543689322e-05,
      "loss": 0.0334,
      "step": 94920
    },
    {
      "epoch": 36.87,
      "learning_rate": 1.5084530744336572e-05,
      "loss": 0.2208,
      "step": 94930
    },
    {
      "epoch": 36.87,
      "learning_rate": 1.508401294498382e-05,
      "loss": 0.1526,
      "step": 94940
    },
    {
      "epoch": 36.87,
      "learning_rate": 1.508349514563107e-05,
      "loss": 0.108,
      "step": 94950
    },
    {
      "epoch": 36.88,
      "learning_rate": 1.5082977346278317e-05,
      "loss": 0.1055,
      "step": 94960
    },
    {
      "epoch": 36.88,
      "learning_rate": 1.5082459546925567e-05,
      "loss": 0.0666,
      "step": 94970
    },
    {
      "epoch": 36.89,
      "learning_rate": 1.5081941747572817e-05,
      "loss": 0.0109,
      "step": 94980
    },
    {
      "epoch": 36.89,
      "learning_rate": 1.5081423948220066e-05,
      "loss": 0.1345,
      "step": 94990
    },
    {
      "epoch": 36.89,
      "learning_rate": 1.5080906148867316e-05,
      "loss": 0.0656,
      "step": 95000
    },
    {
      "epoch": 36.9,
      "learning_rate": 1.5080388349514565e-05,
      "loss": 0.0369,
      "step": 95010
    },
    {
      "epoch": 36.9,
      "learning_rate": 1.5079870550161813e-05,
      "loss": 0.0238,
      "step": 95020
    },
    {
      "epoch": 36.9,
      "learning_rate": 1.5079352750809061e-05,
      "loss": 0.0998,
      "step": 95030
    },
    {
      "epoch": 36.91,
      "learning_rate": 1.507883495145631e-05,
      "loss": 0.0762,
      "step": 95040
    },
    {
      "epoch": 36.91,
      "learning_rate": 1.507831715210356e-05,
      "loss": 0.1736,
      "step": 95050
    },
    {
      "epoch": 36.92,
      "learning_rate": 1.507779935275081e-05,
      "loss": 0.0631,
      "step": 95060
    },
    {
      "epoch": 36.92,
      "learning_rate": 1.507728155339806e-05,
      "loss": 0.0554,
      "step": 95070
    },
    {
      "epoch": 36.92,
      "learning_rate": 1.507676375404531e-05,
      "loss": 0.0357,
      "step": 95080
    },
    {
      "epoch": 36.93,
      "learning_rate": 1.5076245954692559e-05,
      "loss": 0.0495,
      "step": 95090
    },
    {
      "epoch": 36.93,
      "learning_rate": 1.5075728155339807e-05,
      "loss": 0.1017,
      "step": 95100
    },
    {
      "epoch": 36.94,
      "learning_rate": 1.5075210355987055e-05,
      "loss": 0.1263,
      "step": 95110
    },
    {
      "epoch": 36.94,
      "learning_rate": 1.5074692556634304e-05,
      "loss": 0.0821,
      "step": 95120
    },
    {
      "epoch": 36.94,
      "learning_rate": 1.5074174757281554e-05,
      "loss": 0.1253,
      "step": 95130
    },
    {
      "epoch": 36.95,
      "learning_rate": 1.5073656957928804e-05,
      "loss": 0.0292,
      "step": 95140
    },
    {
      "epoch": 36.95,
      "learning_rate": 1.5073139158576053e-05,
      "loss": 0.0249,
      "step": 95150
    },
    {
      "epoch": 36.96,
      "learning_rate": 1.5072621359223303e-05,
      "loss": 0.1046,
      "step": 95160
    },
    {
      "epoch": 36.96,
      "learning_rate": 1.5072103559870552e-05,
      "loss": 0.1168,
      "step": 95170
    },
    {
      "epoch": 36.96,
      "learning_rate": 1.5071585760517799e-05,
      "loss": 0.0756,
      "step": 95180
    },
    {
      "epoch": 36.97,
      "learning_rate": 1.5071067961165048e-05,
      "loss": 0.0435,
      "step": 95190
    },
    {
      "epoch": 36.97,
      "learning_rate": 1.5070550161812298e-05,
      "loss": 0.1471,
      "step": 95200
    },
    {
      "epoch": 36.97,
      "learning_rate": 1.5070032362459548e-05,
      "loss": 0.0483,
      "step": 95210
    },
    {
      "epoch": 36.98,
      "learning_rate": 1.5069514563106797e-05,
      "loss": 0.1216,
      "step": 95220
    },
    {
      "epoch": 36.98,
      "learning_rate": 1.5068996763754047e-05,
      "loss": 0.1783,
      "step": 95230
    },
    {
      "epoch": 36.99,
      "learning_rate": 1.5068478964401296e-05,
      "loss": 0.1369,
      "step": 95240
    },
    {
      "epoch": 36.99,
      "learning_rate": 1.5067961165048546e-05,
      "loss": 0.2083,
      "step": 95250
    },
    {
      "epoch": 36.99,
      "learning_rate": 1.5067443365695792e-05,
      "loss": 0.1072,
      "step": 95260
    },
    {
      "epoch": 37.0,
      "learning_rate": 1.5066925566343042e-05,
      "loss": 0.0348,
      "step": 95270
    },
    {
      "epoch": 37.0,
      "eval_accuracy": 0.9485557083906465,
      "eval_loss": 0.3341105878353119,
      "eval_runtime": 8.2275,
      "eval_samples_per_second": 441.813,
      "eval_steps_per_second": 55.303,
      "step": 95275
    },
    {
      "epoch": 37.0,
      "learning_rate": 1.5066407766990291e-05,
      "loss": 0.1514,
      "step": 95280
    },
    {
      "epoch": 37.01,
      "learning_rate": 1.5065889967637541e-05,
      "loss": 0.0645,
      "step": 95290
    },
    {
      "epoch": 37.01,
      "learning_rate": 1.506537216828479e-05,
      "loss": 0.0286,
      "step": 95300
    },
    {
      "epoch": 37.01,
      "learning_rate": 1.506485436893204e-05,
      "loss": 0.0382,
      "step": 95310
    },
    {
      "epoch": 37.02,
      "learning_rate": 1.506433656957929e-05,
      "loss": 0.0965,
      "step": 95320
    },
    {
      "epoch": 37.02,
      "learning_rate": 1.506381877022654e-05,
      "loss": 0.0798,
      "step": 95330
    },
    {
      "epoch": 37.03,
      "learning_rate": 1.506330097087379e-05,
      "loss": 0.1452,
      "step": 95340
    },
    {
      "epoch": 37.03,
      "learning_rate": 1.5062783171521035e-05,
      "loss": 0.1878,
      "step": 95350
    },
    {
      "epoch": 37.03,
      "learning_rate": 1.5062265372168285e-05,
      "loss": 0.0231,
      "step": 95360
    },
    {
      "epoch": 37.04,
      "learning_rate": 1.5061747572815535e-05,
      "loss": 0.1118,
      "step": 95370
    },
    {
      "epoch": 37.04,
      "learning_rate": 1.5061229773462784e-05,
      "loss": 0.068,
      "step": 95380
    },
    {
      "epoch": 37.04,
      "learning_rate": 1.5060711974110034e-05,
      "loss": 0.1219,
      "step": 95390
    },
    {
      "epoch": 37.05,
      "learning_rate": 1.5060194174757284e-05,
      "loss": 0.0682,
      "step": 95400
    },
    {
      "epoch": 37.05,
      "learning_rate": 1.5059676375404533e-05,
      "loss": 0.0023,
      "step": 95410
    },
    {
      "epoch": 37.06,
      "learning_rate": 1.5059158576051783e-05,
      "loss": 0.1945,
      "step": 95420
    },
    {
      "epoch": 37.06,
      "learning_rate": 1.5058640776699029e-05,
      "loss": 0.0544,
      "step": 95430
    },
    {
      "epoch": 37.06,
      "learning_rate": 1.5058122977346279e-05,
      "loss": 0.0305,
      "step": 95440
    },
    {
      "epoch": 37.07,
      "learning_rate": 1.5057605177993528e-05,
      "loss": 0.0509,
      "step": 95450
    },
    {
      "epoch": 37.07,
      "learning_rate": 1.5057087378640778e-05,
      "loss": 0.0857,
      "step": 95460
    },
    {
      "epoch": 37.08,
      "learning_rate": 1.5056569579288027e-05,
      "loss": 0.0699,
      "step": 95470
    },
    {
      "epoch": 37.08,
      "learning_rate": 1.5056051779935277e-05,
      "loss": 0.1782,
      "step": 95480
    },
    {
      "epoch": 37.08,
      "learning_rate": 1.5055533980582527e-05,
      "loss": 0.0065,
      "step": 95490
    },
    {
      "epoch": 37.09,
      "learning_rate": 1.5055016181229776e-05,
      "loss": 0.1969,
      "step": 95500
    },
    {
      "epoch": 37.09,
      "learning_rate": 1.5054498381877023e-05,
      "loss": 0.0496,
      "step": 95510
    },
    {
      "epoch": 37.1,
      "learning_rate": 1.5053980582524272e-05,
      "loss": 0.0877,
      "step": 95520
    },
    {
      "epoch": 37.1,
      "learning_rate": 1.5053462783171522e-05,
      "loss": 0.0027,
      "step": 95530
    },
    {
      "epoch": 37.1,
      "learning_rate": 1.5052944983818771e-05,
      "loss": 0.0872,
      "step": 95540
    },
    {
      "epoch": 37.11,
      "learning_rate": 1.5052427184466021e-05,
      "loss": 0.0163,
      "step": 95550
    },
    {
      "epoch": 37.11,
      "learning_rate": 1.505190938511327e-05,
      "loss": 0.1064,
      "step": 95560
    },
    {
      "epoch": 37.11,
      "learning_rate": 1.505139158576052e-05,
      "loss": 0.1265,
      "step": 95570
    },
    {
      "epoch": 37.12,
      "learning_rate": 1.505087378640777e-05,
      "loss": 0.0262,
      "step": 95580
    },
    {
      "epoch": 37.12,
      "learning_rate": 1.5050355987055016e-05,
      "loss": 0.0588,
      "step": 95590
    },
    {
      "epoch": 37.13,
      "learning_rate": 1.5049838187702266e-05,
      "loss": 0.0269,
      "step": 95600
    },
    {
      "epoch": 37.13,
      "learning_rate": 1.5049320388349515e-05,
      "loss": 0.1557,
      "step": 95610
    },
    {
      "epoch": 37.13,
      "learning_rate": 1.5048802588996765e-05,
      "loss": 0.0546,
      "step": 95620
    },
    {
      "epoch": 37.14,
      "learning_rate": 1.5048284789644015e-05,
      "loss": 0.0876,
      "step": 95630
    },
    {
      "epoch": 37.14,
      "learning_rate": 1.5047766990291264e-05,
      "loss": 0.1005,
      "step": 95640
    },
    {
      "epoch": 37.15,
      "learning_rate": 1.5047249190938514e-05,
      "loss": 0.0875,
      "step": 95650
    },
    {
      "epoch": 37.15,
      "learning_rate": 1.5046731391585763e-05,
      "loss": 0.0638,
      "step": 95660
    },
    {
      "epoch": 37.15,
      "learning_rate": 1.504621359223301e-05,
      "loss": 0.1104,
      "step": 95670
    },
    {
      "epoch": 37.16,
      "learning_rate": 1.504569579288026e-05,
      "loss": 0.1343,
      "step": 95680
    },
    {
      "epoch": 37.16,
      "learning_rate": 1.5045177993527509e-05,
      "loss": 0.0659,
      "step": 95690
    },
    {
      "epoch": 37.17,
      "learning_rate": 1.5044660194174759e-05,
      "loss": 0.112,
      "step": 95700
    },
    {
      "epoch": 37.17,
      "learning_rate": 1.5044142394822008e-05,
      "loss": 0.0828,
      "step": 95710
    },
    {
      "epoch": 37.17,
      "learning_rate": 1.5043624595469258e-05,
      "loss": 0.0053,
      "step": 95720
    },
    {
      "epoch": 37.18,
      "learning_rate": 1.5043106796116507e-05,
      "loss": 0.2158,
      "step": 95730
    },
    {
      "epoch": 37.18,
      "learning_rate": 1.5042588996763757e-05,
      "loss": 0.0871,
      "step": 95740
    },
    {
      "epoch": 37.18,
      "learning_rate": 1.5042071197411003e-05,
      "loss": 0.0393,
      "step": 95750
    },
    {
      "epoch": 37.19,
      "learning_rate": 1.5041553398058253e-05,
      "loss": 0.0584,
      "step": 95760
    },
    {
      "epoch": 37.19,
      "learning_rate": 1.5041035598705502e-05,
      "loss": 0.0033,
      "step": 95770
    },
    {
      "epoch": 37.2,
      "learning_rate": 1.5040517799352752e-05,
      "loss": 0.0096,
      "step": 95780
    },
    {
      "epoch": 37.2,
      "learning_rate": 1.5040000000000002e-05,
      "loss": 0.0433,
      "step": 95790
    },
    {
      "epoch": 37.2,
      "learning_rate": 1.5039482200647251e-05,
      "loss": 0.0444,
      "step": 95800
    },
    {
      "epoch": 37.21,
      "learning_rate": 1.5038964401294501e-05,
      "loss": 0.0372,
      "step": 95810
    },
    {
      "epoch": 37.21,
      "learning_rate": 1.5038446601941749e-05,
      "loss": 0.0556,
      "step": 95820
    },
    {
      "epoch": 37.22,
      "learning_rate": 1.5037928802588997e-05,
      "loss": 0.0269,
      "step": 95830
    },
    {
      "epoch": 37.22,
      "learning_rate": 1.5037411003236246e-05,
      "loss": 0.0915,
      "step": 95840
    },
    {
      "epoch": 37.22,
      "learning_rate": 1.5036893203883496e-05,
      "loss": 0.0237,
      "step": 95850
    },
    {
      "epoch": 37.23,
      "learning_rate": 1.5036375404530746e-05,
      "loss": 0.0448,
      "step": 95860
    },
    {
      "epoch": 37.23,
      "learning_rate": 1.5035857605177995e-05,
      "loss": 0.1056,
      "step": 95870
    },
    {
      "epoch": 37.23,
      "learning_rate": 1.5035339805825245e-05,
      "loss": 0.0583,
      "step": 95880
    },
    {
      "epoch": 37.24,
      "learning_rate": 1.5034822006472494e-05,
      "loss": 0.1452,
      "step": 95890
    },
    {
      "epoch": 37.24,
      "learning_rate": 1.5034304207119742e-05,
      "loss": 0.0623,
      "step": 95900
    },
    {
      "epoch": 37.25,
      "learning_rate": 1.5033786407766992e-05,
      "loss": 0.1408,
      "step": 95910
    },
    {
      "epoch": 37.25,
      "learning_rate": 1.503326860841424e-05,
      "loss": 0.0034,
      "step": 95920
    },
    {
      "epoch": 37.25,
      "learning_rate": 1.503275080906149e-05,
      "loss": 0.016,
      "step": 95930
    },
    {
      "epoch": 37.26,
      "learning_rate": 1.503223300970874e-05,
      "loss": 0.0213,
      "step": 95940
    },
    {
      "epoch": 37.26,
      "learning_rate": 1.5031715210355989e-05,
      "loss": 0.0211,
      "step": 95950
    },
    {
      "epoch": 37.27,
      "learning_rate": 1.5031197411003238e-05,
      "loss": 0.0779,
      "step": 95960
    },
    {
      "epoch": 37.27,
      "learning_rate": 1.5030679611650486e-05,
      "loss": 0.0905,
      "step": 95970
    },
    {
      "epoch": 37.27,
      "learning_rate": 1.5030161812297736e-05,
      "loss": 0.013,
      "step": 95980
    },
    {
      "epoch": 37.28,
      "learning_rate": 1.5029644012944986e-05,
      "loss": 0.0651,
      "step": 95990
    },
    {
      "epoch": 37.28,
      "learning_rate": 1.5029126213592234e-05,
      "loss": 0.0211,
      "step": 96000
    },
    {
      "epoch": 37.29,
      "learning_rate": 1.5028608414239483e-05,
      "loss": 0.0303,
      "step": 96010
    },
    {
      "epoch": 37.29,
      "learning_rate": 1.5028090614886733e-05,
      "loss": 0.0275,
      "step": 96020
    },
    {
      "epoch": 37.29,
      "learning_rate": 1.5027572815533982e-05,
      "loss": 0.1318,
      "step": 96030
    },
    {
      "epoch": 37.3,
      "learning_rate": 1.5027055016181232e-05,
      "loss": 0.0548,
      "step": 96040
    },
    {
      "epoch": 37.3,
      "learning_rate": 1.502653721682848e-05,
      "loss": 0.0998,
      "step": 96050
    },
    {
      "epoch": 37.3,
      "learning_rate": 1.502601941747573e-05,
      "loss": 0.0907,
      "step": 96060
    },
    {
      "epoch": 37.31,
      "learning_rate": 1.5025501618122979e-05,
      "loss": 0.0111,
      "step": 96070
    },
    {
      "epoch": 37.31,
      "learning_rate": 1.5024983818770227e-05,
      "loss": 0.1035,
      "step": 96080
    },
    {
      "epoch": 37.32,
      "learning_rate": 1.5024466019417477e-05,
      "loss": 0.0453,
      "step": 96090
    },
    {
      "epoch": 37.32,
      "learning_rate": 1.5023948220064726e-05,
      "loss": 0.0853,
      "step": 96100
    },
    {
      "epoch": 37.32,
      "learning_rate": 1.5023430420711976e-05,
      "loss": 0.0925,
      "step": 96110
    },
    {
      "epoch": 37.33,
      "learning_rate": 1.5022912621359224e-05,
      "loss": 0.1193,
      "step": 96120
    },
    {
      "epoch": 37.33,
      "learning_rate": 1.5022394822006473e-05,
      "loss": 0.1431,
      "step": 96130
    },
    {
      "epoch": 37.34,
      "learning_rate": 1.5021877022653723e-05,
      "loss": 0.2297,
      "step": 96140
    },
    {
      "epoch": 37.34,
      "learning_rate": 1.5021359223300973e-05,
      "loss": 0.0422,
      "step": 96150
    },
    {
      "epoch": 37.34,
      "learning_rate": 1.502084142394822e-05,
      "loss": 0.1476,
      "step": 96160
    },
    {
      "epoch": 37.35,
      "learning_rate": 1.502032362459547e-05,
      "loss": 0.0651,
      "step": 96170
    },
    {
      "epoch": 37.35,
      "learning_rate": 1.501980582524272e-05,
      "loss": 0.0756,
      "step": 96180
    },
    {
      "epoch": 37.36,
      "learning_rate": 1.501928802588997e-05,
      "loss": 0.0229,
      "step": 96190
    },
    {
      "epoch": 37.36,
      "learning_rate": 1.5018770226537217e-05,
      "loss": 0.1384,
      "step": 96200
    },
    {
      "epoch": 37.36,
      "learning_rate": 1.5018252427184467e-05,
      "loss": 0.1253,
      "step": 96210
    },
    {
      "epoch": 37.37,
      "learning_rate": 1.5017734627831717e-05,
      "loss": 0.1963,
      "step": 96220
    },
    {
      "epoch": 37.37,
      "learning_rate": 1.5017216828478966e-05,
      "loss": 0.1905,
      "step": 96230
    },
    {
      "epoch": 37.37,
      "learning_rate": 1.5016699029126214e-05,
      "loss": 0.0245,
      "step": 96240
    },
    {
      "epoch": 37.38,
      "learning_rate": 1.5016181229773464e-05,
      "loss": 0.0322,
      "step": 96250
    },
    {
      "epoch": 37.38,
      "learning_rate": 1.5015663430420713e-05,
      "loss": 0.0751,
      "step": 96260
    },
    {
      "epoch": 37.39,
      "learning_rate": 1.5015145631067963e-05,
      "loss": 0.1146,
      "step": 96270
    },
    {
      "epoch": 37.39,
      "learning_rate": 1.5014627831715211e-05,
      "loss": 0.1243,
      "step": 96280
    },
    {
      "epoch": 37.39,
      "learning_rate": 1.501411003236246e-05,
      "loss": 0.409,
      "step": 96290
    },
    {
      "epoch": 37.4,
      "learning_rate": 1.501359223300971e-05,
      "loss": 0.1193,
      "step": 96300
    },
    {
      "epoch": 37.4,
      "learning_rate": 1.501307443365696e-05,
      "loss": 0.1095,
      "step": 96310
    },
    {
      "epoch": 37.41,
      "learning_rate": 1.5012556634304208e-05,
      "loss": 0.0263,
      "step": 96320
    },
    {
      "epoch": 37.41,
      "learning_rate": 1.5012038834951457e-05,
      "loss": 0.0408,
      "step": 96330
    },
    {
      "epoch": 37.41,
      "learning_rate": 1.5011521035598707e-05,
      "loss": 0.0275,
      "step": 96340
    },
    {
      "epoch": 37.42,
      "learning_rate": 1.5011003236245955e-05,
      "loss": 0.0318,
      "step": 96350
    },
    {
      "epoch": 37.42,
      "learning_rate": 1.5010485436893205e-05,
      "loss": 0.0965,
      "step": 96360
    },
    {
      "epoch": 37.43,
      "learning_rate": 1.5009967637540454e-05,
      "loss": 0.0929,
      "step": 96370
    },
    {
      "epoch": 37.43,
      "learning_rate": 1.5009449838187704e-05,
      "loss": 0.0698,
      "step": 96380
    },
    {
      "epoch": 37.43,
      "learning_rate": 1.5008932038834953e-05,
      "loss": 0.0766,
      "step": 96390
    },
    {
      "epoch": 37.44,
      "learning_rate": 1.5008414239482201e-05,
      "loss": 0.1149,
      "step": 96400
    },
    {
      "epoch": 37.44,
      "learning_rate": 1.5007896440129451e-05,
      "loss": 0.1017,
      "step": 96410
    },
    {
      "epoch": 37.44,
      "learning_rate": 1.50073786407767e-05,
      "loss": 0.1253,
      "step": 96420
    },
    {
      "epoch": 37.45,
      "learning_rate": 1.5006860841423948e-05,
      "loss": 0.0344,
      "step": 96430
    },
    {
      "epoch": 37.45,
      "learning_rate": 1.5006343042071198e-05,
      "loss": 0.0654,
      "step": 96440
    },
    {
      "epoch": 37.46,
      "learning_rate": 1.5005825242718448e-05,
      "loss": 0.0243,
      "step": 96450
    },
    {
      "epoch": 37.46,
      "learning_rate": 1.5005307443365697e-05,
      "loss": 0.1067,
      "step": 96460
    },
    {
      "epoch": 37.46,
      "learning_rate": 1.5004789644012947e-05,
      "loss": 0.1098,
      "step": 96470
    },
    {
      "epoch": 37.47,
      "learning_rate": 1.5004271844660197e-05,
      "loss": 0.1027,
      "step": 96480
    },
    {
      "epoch": 37.47,
      "learning_rate": 1.5003754045307444e-05,
      "loss": 0.1278,
      "step": 96490
    },
    {
      "epoch": 37.48,
      "learning_rate": 1.5003236245954692e-05,
      "loss": 0.0046,
      "step": 96500
    },
    {
      "epoch": 37.48,
      "learning_rate": 1.5002718446601942e-05,
      "loss": 0.2438,
      "step": 96510
    },
    {
      "epoch": 37.48,
      "learning_rate": 1.5002200647249192e-05,
      "loss": 0.0997,
      "step": 96520
    },
    {
      "epoch": 37.49,
      "learning_rate": 1.5001682847896441e-05,
      "loss": 0.0094,
      "step": 96530
    },
    {
      "epoch": 37.49,
      "learning_rate": 1.5001165048543691e-05,
      "loss": 0.0013,
      "step": 96540
    },
    {
      "epoch": 37.5,
      "learning_rate": 1.500064724919094e-05,
      "loss": 0.026,
      "step": 96550
    },
    {
      "epoch": 37.5,
      "learning_rate": 1.500012944983819e-05,
      "loss": 0.1674,
      "step": 96560
    },
    {
      "epoch": 37.5,
      "learning_rate": 1.4999611650485438e-05,
      "loss": 0.0873,
      "step": 96570
    },
    {
      "epoch": 37.51,
      "learning_rate": 1.4999093851132686e-05,
      "loss": 0.0752,
      "step": 96580
    },
    {
      "epoch": 37.51,
      "learning_rate": 1.4998576051779936e-05,
      "loss": 0.0027,
      "step": 96590
    },
    {
      "epoch": 37.51,
      "learning_rate": 1.4998058252427185e-05,
      "loss": 0.036,
      "step": 96600
    },
    {
      "epoch": 37.52,
      "learning_rate": 1.4997540453074435e-05,
      "loss": 0.1156,
      "step": 96610
    },
    {
      "epoch": 37.52,
      "learning_rate": 1.4997022653721684e-05,
      "loss": 0.0375,
      "step": 96620
    },
    {
      "epoch": 37.53,
      "learning_rate": 1.4996504854368934e-05,
      "loss": 0.0554,
      "step": 96630
    },
    {
      "epoch": 37.53,
      "learning_rate": 1.4995987055016184e-05,
      "loss": 0.1024,
      "step": 96640
    },
    {
      "epoch": 37.53,
      "learning_rate": 1.499546925566343e-05,
      "loss": 0.1406,
      "step": 96650
    },
    {
      "epoch": 37.54,
      "learning_rate": 1.499495145631068e-05,
      "loss": 0.1546,
      "step": 96660
    },
    {
      "epoch": 37.54,
      "learning_rate": 1.4994433656957929e-05,
      "loss": 0.0728,
      "step": 96670
    },
    {
      "epoch": 37.55,
      "learning_rate": 1.4993915857605179e-05,
      "loss": 0.0944,
      "step": 96680
    },
    {
      "epoch": 37.55,
      "learning_rate": 1.4993398058252428e-05,
      "loss": 0.1897,
      "step": 96690
    },
    {
      "epoch": 37.55,
      "learning_rate": 1.4992880258899678e-05,
      "loss": 0.1632,
      "step": 96700
    },
    {
      "epoch": 37.56,
      "learning_rate": 1.4992362459546928e-05,
      "loss": 0.0213,
      "step": 96710
    },
    {
      "epoch": 37.56,
      "learning_rate": 1.4991844660194177e-05,
      "loss": 0.0678,
      "step": 96720
    },
    {
      "epoch": 37.57,
      "learning_rate": 1.4991326860841423e-05,
      "loss": 0.0711,
      "step": 96730
    },
    {
      "epoch": 37.57,
      "learning_rate": 1.4990809061488673e-05,
      "loss": 0.1229,
      "step": 96740
    },
    {
      "epoch": 37.57,
      "learning_rate": 1.4990291262135923e-05,
      "loss": 0.2215,
      "step": 96750
    },
    {
      "epoch": 37.58,
      "learning_rate": 1.4989773462783172e-05,
      "loss": 0.1466,
      "step": 96760
    },
    {
      "epoch": 37.58,
      "learning_rate": 1.4989255663430422e-05,
      "loss": 0.0234,
      "step": 96770
    },
    {
      "epoch": 37.58,
      "learning_rate": 1.4988737864077672e-05,
      "loss": 0.1164,
      "step": 96780
    },
    {
      "epoch": 37.59,
      "learning_rate": 1.4988220064724921e-05,
      "loss": 0.1013,
      "step": 96790
    },
    {
      "epoch": 37.59,
      "learning_rate": 1.498770226537217e-05,
      "loss": 0.2159,
      "step": 96800
    },
    {
      "epoch": 37.6,
      "learning_rate": 1.4987184466019417e-05,
      "loss": 0.1013,
      "step": 96810
    },
    {
      "epoch": 37.6,
      "learning_rate": 1.4986666666666667e-05,
      "loss": 0.2663,
      "step": 96820
    },
    {
      "epoch": 37.6,
      "learning_rate": 1.4986148867313916e-05,
      "loss": 0.1339,
      "step": 96830
    },
    {
      "epoch": 37.61,
      "learning_rate": 1.4985631067961166e-05,
      "loss": 0.175,
      "step": 96840
    },
    {
      "epoch": 37.61,
      "learning_rate": 1.4985113268608415e-05,
      "loss": 0.0549,
      "step": 96850
    },
    {
      "epoch": 37.62,
      "learning_rate": 1.4984595469255665e-05,
      "loss": 0.0305,
      "step": 96860
    },
    {
      "epoch": 37.62,
      "learning_rate": 1.4984077669902915e-05,
      "loss": 0.2895,
      "step": 96870
    },
    {
      "epoch": 37.62,
      "learning_rate": 1.4983559870550164e-05,
      "loss": 0.03,
      "step": 96880
    },
    {
      "epoch": 37.63,
      "learning_rate": 1.498304207119741e-05,
      "loss": 0.1723,
      "step": 96890
    },
    {
      "epoch": 37.63,
      "learning_rate": 1.498252427184466e-05,
      "loss": 0.0824,
      "step": 96900
    },
    {
      "epoch": 37.63,
      "learning_rate": 1.498200647249191e-05,
      "loss": 0.0065,
      "step": 96910
    },
    {
      "epoch": 37.64,
      "learning_rate": 1.498148867313916e-05,
      "loss": 0.1941,
      "step": 96920
    },
    {
      "epoch": 37.64,
      "learning_rate": 1.4980970873786409e-05,
      "loss": 0.0297,
      "step": 96930
    },
    {
      "epoch": 37.65,
      "learning_rate": 1.4980453074433659e-05,
      "loss": 0.0185,
      "step": 96940
    },
    {
      "epoch": 37.65,
      "learning_rate": 1.4979935275080908e-05,
      "loss": 0.0917,
      "step": 96950
    },
    {
      "epoch": 37.65,
      "learning_rate": 1.4979417475728158e-05,
      "loss": 0.0583,
      "step": 96960
    },
    {
      "epoch": 37.66,
      "learning_rate": 1.4978899676375404e-05,
      "loss": 0.0536,
      "step": 96970
    },
    {
      "epoch": 37.66,
      "learning_rate": 1.4978381877022654e-05,
      "loss": 0.1402,
      "step": 96980
    },
    {
      "epoch": 37.67,
      "learning_rate": 1.4977864077669903e-05,
      "loss": 0.0915,
      "step": 96990
    },
    {
      "epoch": 37.67,
      "learning_rate": 1.4977346278317153e-05,
      "loss": 0.0288,
      "step": 97000
    },
    {
      "epoch": 37.67,
      "learning_rate": 1.4976828478964403e-05,
      "loss": 0.1571,
      "step": 97010
    },
    {
      "epoch": 37.68,
      "learning_rate": 1.4976310679611652e-05,
      "loss": 0.0853,
      "step": 97020
    },
    {
      "epoch": 37.68,
      "learning_rate": 1.4975792880258902e-05,
      "loss": 0.2014,
      "step": 97030
    },
    {
      "epoch": 37.69,
      "learning_rate": 1.4975275080906151e-05,
      "loss": 0.1132,
      "step": 97040
    },
    {
      "epoch": 37.69,
      "learning_rate": 1.4974757281553401e-05,
      "loss": 0.0319,
      "step": 97050
    },
    {
      "epoch": 37.69,
      "learning_rate": 1.4974239482200647e-05,
      "loss": 0.1447,
      "step": 97060
    },
    {
      "epoch": 37.7,
      "learning_rate": 1.4973721682847897e-05,
      "loss": 0.0286,
      "step": 97070
    },
    {
      "epoch": 37.7,
      "learning_rate": 1.4973203883495147e-05,
      "loss": 0.0375,
      "step": 97080
    },
    {
      "epoch": 37.7,
      "learning_rate": 1.4972686084142396e-05,
      "loss": 0.1596,
      "step": 97090
    },
    {
      "epoch": 37.71,
      "learning_rate": 1.4972168284789646e-05,
      "loss": 0.1598,
      "step": 97100
    },
    {
      "epoch": 37.71,
      "learning_rate": 1.4971650485436895e-05,
      "loss": 0.0608,
      "step": 97110
    },
    {
      "epoch": 37.72,
      "learning_rate": 1.4971132686084145e-05,
      "loss": 0.14,
      "step": 97120
    },
    {
      "epoch": 37.72,
      "learning_rate": 1.4970614886731395e-05,
      "loss": 0.0995,
      "step": 97130
    },
    {
      "epoch": 37.72,
      "learning_rate": 1.497009708737864e-05,
      "loss": 0.2159,
      "step": 97140
    },
    {
      "epoch": 37.73,
      "learning_rate": 1.496957928802589e-05,
      "loss": 0.1292,
      "step": 97150
    },
    {
      "epoch": 37.73,
      "learning_rate": 1.496906148867314e-05,
      "loss": 0.0732,
      "step": 97160
    },
    {
      "epoch": 37.74,
      "learning_rate": 1.496854368932039e-05,
      "loss": 0.0011,
      "step": 97170
    },
    {
      "epoch": 37.74,
      "learning_rate": 1.496802588996764e-05,
      "loss": 0.0329,
      "step": 97180
    },
    {
      "epoch": 37.74,
      "learning_rate": 1.4967508090614889e-05,
      "loss": 0.0605,
      "step": 97190
    },
    {
      "epoch": 37.75,
      "learning_rate": 1.4966990291262139e-05,
      "loss": 0.1259,
      "step": 97200
    },
    {
      "epoch": 37.75,
      "learning_rate": 1.4966472491909388e-05,
      "loss": 0.0552,
      "step": 97210
    },
    {
      "epoch": 37.76,
      "learning_rate": 1.4965954692556634e-05,
      "loss": 0.0562,
      "step": 97220
    },
    {
      "epoch": 37.76,
      "learning_rate": 1.4965436893203884e-05,
      "loss": 0.0922,
      "step": 97230
    },
    {
      "epoch": 37.76,
      "learning_rate": 1.4964919093851134e-05,
      "loss": 0.2566,
      "step": 97240
    },
    {
      "epoch": 37.77,
      "learning_rate": 1.4964401294498383e-05,
      "loss": 0.0814,
      "step": 97250
    },
    {
      "epoch": 37.77,
      "learning_rate": 1.4963883495145633e-05,
      "loss": 0.0759,
      "step": 97260
    },
    {
      "epoch": 37.77,
      "learning_rate": 1.4963365695792882e-05,
      "loss": 0.0873,
      "step": 97270
    },
    {
      "epoch": 37.78,
      "learning_rate": 1.4962847896440132e-05,
      "loss": 0.0482,
      "step": 97280
    },
    {
      "epoch": 37.78,
      "learning_rate": 1.496233009708738e-05,
      "loss": 0.1491,
      "step": 97290
    },
    {
      "epoch": 37.79,
      "learning_rate": 1.4961812297734628e-05,
      "loss": 0.1066,
      "step": 97300
    },
    {
      "epoch": 37.79,
      "learning_rate": 1.4961294498381878e-05,
      "loss": 0.0982,
      "step": 97310
    },
    {
      "epoch": 37.79,
      "learning_rate": 1.4960776699029127e-05,
      "loss": 0.1538,
      "step": 97320
    },
    {
      "epoch": 37.8,
      "learning_rate": 1.4960258899676377e-05,
      "loss": 0.1668,
      "step": 97330
    },
    {
      "epoch": 37.8,
      "learning_rate": 1.4959741100323626e-05,
      "loss": 0.1228,
      "step": 97340
    },
    {
      "epoch": 37.81,
      "learning_rate": 1.4959223300970876e-05,
      "loss": 0.1134,
      "step": 97350
    },
    {
      "epoch": 37.81,
      "learning_rate": 1.4958705501618126e-05,
      "loss": 0.1175,
      "step": 97360
    },
    {
      "epoch": 37.81,
      "learning_rate": 1.4958187702265374e-05,
      "loss": 0.2772,
      "step": 97370
    },
    {
      "epoch": 37.82,
      "learning_rate": 1.4957669902912622e-05,
      "loss": 0.0169,
      "step": 97380
    },
    {
      "epoch": 37.82,
      "learning_rate": 1.4957152103559871e-05,
      "loss": 0.0731,
      "step": 97390
    },
    {
      "epoch": 37.83,
      "learning_rate": 1.495663430420712e-05,
      "loss": 0.0143,
      "step": 97400
    },
    {
      "epoch": 37.83,
      "learning_rate": 1.495611650485437e-05,
      "loss": 0.1522,
      "step": 97410
    },
    {
      "epoch": 37.83,
      "learning_rate": 1.495559870550162e-05,
      "loss": 0.257,
      "step": 97420
    },
    {
      "epoch": 37.84,
      "learning_rate": 1.495508090614887e-05,
      "loss": 0.1106,
      "step": 97430
    },
    {
      "epoch": 37.84,
      "learning_rate": 1.4954563106796118e-05,
      "loss": 0.0584,
      "step": 97440
    },
    {
      "epoch": 37.84,
      "learning_rate": 1.4954045307443367e-05,
      "loss": 0.0513,
      "step": 97450
    },
    {
      "epoch": 37.85,
      "learning_rate": 1.4953527508090615e-05,
      "loss": 0.1277,
      "step": 97460
    },
    {
      "epoch": 37.85,
      "learning_rate": 1.4953009708737865e-05,
      "loss": 0.0379,
      "step": 97470
    },
    {
      "epoch": 37.86,
      "learning_rate": 1.4952491909385114e-05,
      "loss": 0.0425,
      "step": 97480
    },
    {
      "epoch": 37.86,
      "learning_rate": 1.4951974110032364e-05,
      "loss": 0.0308,
      "step": 97490
    },
    {
      "epoch": 37.86,
      "learning_rate": 1.4951456310679614e-05,
      "loss": 0.0262,
      "step": 97500
    },
    {
      "epoch": 37.87,
      "learning_rate": 1.4950938511326863e-05,
      "loss": 0.1337,
      "step": 97510
    },
    {
      "epoch": 37.87,
      "learning_rate": 1.4950420711974111e-05,
      "loss": 0.0584,
      "step": 97520
    },
    {
      "epoch": 37.88,
      "learning_rate": 1.494990291262136e-05,
      "loss": 0.1228,
      "step": 97530
    },
    {
      "epoch": 37.88,
      "learning_rate": 1.4949385113268609e-05,
      "loss": 0.058,
      "step": 97540
    },
    {
      "epoch": 37.88,
      "learning_rate": 1.4948867313915858e-05,
      "loss": 0.1081,
      "step": 97550
    },
    {
      "epoch": 37.89,
      "learning_rate": 1.4948349514563108e-05,
      "loss": 0.1565,
      "step": 97560
    },
    {
      "epoch": 37.89,
      "learning_rate": 1.4947831715210357e-05,
      "loss": 0.0958,
      "step": 97570
    },
    {
      "epoch": 37.9,
      "learning_rate": 1.4947313915857607e-05,
      "loss": 0.1235,
      "step": 97580
    },
    {
      "epoch": 37.9,
      "learning_rate": 1.4946796116504855e-05,
      "loss": 0.0807,
      "step": 97590
    },
    {
      "epoch": 37.9,
      "learning_rate": 1.4946278317152105e-05,
      "loss": 0.1023,
      "step": 97600
    },
    {
      "epoch": 37.91,
      "learning_rate": 1.4945760517799354e-05,
      "loss": 0.1485,
      "step": 97610
    },
    {
      "epoch": 37.91,
      "learning_rate": 1.4945242718446604e-05,
      "loss": 0.1425,
      "step": 97620
    },
    {
      "epoch": 37.91,
      "learning_rate": 1.4944724919093852e-05,
      "loss": 0.0066,
      "step": 97630
    },
    {
      "epoch": 37.92,
      "learning_rate": 1.4944207119741101e-05,
      "loss": 0.0868,
      "step": 97640
    },
    {
      "epoch": 37.92,
      "learning_rate": 1.4943689320388351e-05,
      "loss": 0.0684,
      "step": 97650
    },
    {
      "epoch": 37.93,
      "learning_rate": 1.49431715210356e-05,
      "loss": 0.0031,
      "step": 97660
    },
    {
      "epoch": 37.93,
      "learning_rate": 1.4942653721682849e-05,
      "loss": 0.134,
      "step": 97670
    },
    {
      "epoch": 37.93,
      "learning_rate": 1.4942135922330098e-05,
      "loss": 0.0351,
      "step": 97680
    },
    {
      "epoch": 37.94,
      "learning_rate": 1.4941618122977348e-05,
      "loss": 0.1142,
      "step": 97690
    },
    {
      "epoch": 37.94,
      "learning_rate": 1.4941100323624597e-05,
      "loss": 0.1095,
      "step": 97700
    },
    {
      "epoch": 37.95,
      "learning_rate": 1.4940582524271845e-05,
      "loss": 0.1249,
      "step": 97710
    },
    {
      "epoch": 37.95,
      "learning_rate": 1.4940064724919095e-05,
      "loss": 0.2617,
      "step": 97720
    },
    {
      "epoch": 37.95,
      "learning_rate": 1.4939546925566345e-05,
      "loss": 0.2503,
      "step": 97730
    },
    {
      "epoch": 37.96,
      "learning_rate": 1.4939029126213594e-05,
      "loss": 0.0557,
      "step": 97740
    },
    {
      "epoch": 37.96,
      "learning_rate": 1.4938511326860842e-05,
      "loss": 0.0771,
      "step": 97750
    },
    {
      "epoch": 37.97,
      "learning_rate": 1.4937993527508092e-05,
      "loss": 0.0464,
      "step": 97760
    },
    {
      "epoch": 37.97,
      "learning_rate": 1.4937475728155341e-05,
      "loss": 0.1281,
      "step": 97770
    },
    {
      "epoch": 37.97,
      "learning_rate": 1.4936957928802591e-05,
      "loss": 0.0758,
      "step": 97780
    },
    {
      "epoch": 37.98,
      "learning_rate": 1.4936440129449839e-05,
      "loss": 0.1397,
      "step": 97790
    },
    {
      "epoch": 37.98,
      "learning_rate": 1.4935922330097089e-05,
      "loss": 0.0245,
      "step": 97800
    },
    {
      "epoch": 37.98,
      "learning_rate": 1.4935404530744338e-05,
      "loss": 0.0273,
      "step": 97810
    },
    {
      "epoch": 37.99,
      "learning_rate": 1.4934886731391586e-05,
      "loss": 0.1155,
      "step": 97820
    },
    {
      "epoch": 37.99,
      "learning_rate": 1.4934368932038836e-05,
      "loss": 0.0088,
      "step": 97830
    },
    {
      "epoch": 38.0,
      "learning_rate": 1.4933851132686085e-05,
      "loss": 0.0832,
      "step": 97840
    },
    {
      "epoch": 38.0,
      "learning_rate": 1.4933333333333335e-05,
      "loss": 0.0057,
      "step": 97850
    },
    {
      "epoch": 38.0,
      "eval_accuracy": 0.9466299862448418,
      "eval_loss": 0.3225008249282837,
      "eval_runtime": 8.2867,
      "eval_samples_per_second": 438.657,
      "eval_steps_per_second": 54.908,
      "step": 97850
    },
    {
      "epoch": 38.0,
      "learning_rate": 1.4932815533980585e-05,
      "loss": 0.0213,
      "step": 97860
    },
    {
      "epoch": 38.01,
      "learning_rate": 1.4932297734627832e-05,
      "loss": 0.0784,
      "step": 97870
    },
    {
      "epoch": 38.01,
      "learning_rate": 1.4931779935275082e-05,
      "loss": 0.0805,
      "step": 97880
    },
    {
      "epoch": 38.02,
      "learning_rate": 1.4931262135922332e-05,
      "loss": 0.1741,
      "step": 97890
    },
    {
      "epoch": 38.02,
      "learning_rate": 1.493074433656958e-05,
      "loss": 0.0553,
      "step": 97900
    },
    {
      "epoch": 38.02,
      "learning_rate": 1.493022653721683e-05,
      "loss": 0.0756,
      "step": 97910
    },
    {
      "epoch": 38.03,
      "learning_rate": 1.4929708737864079e-05,
      "loss": 0.0686,
      "step": 97920
    },
    {
      "epoch": 38.03,
      "learning_rate": 1.4929190938511328e-05,
      "loss": 0.1228,
      "step": 97930
    },
    {
      "epoch": 38.03,
      "learning_rate": 1.4928673139158578e-05,
      "loss": 0.0651,
      "step": 97940
    },
    {
      "epoch": 38.04,
      "learning_rate": 1.4928155339805826e-05,
      "loss": 0.0689,
      "step": 97950
    },
    {
      "epoch": 38.04,
      "learning_rate": 1.4927637540453076e-05,
      "loss": 0.149,
      "step": 97960
    },
    {
      "epoch": 38.05,
      "learning_rate": 1.4927119741100324e-05,
      "loss": 0.2265,
      "step": 97970
    },
    {
      "epoch": 38.05,
      "learning_rate": 1.4926601941747573e-05,
      "loss": 0.043,
      "step": 97980
    },
    {
      "epoch": 38.05,
      "learning_rate": 1.4926084142394823e-05,
      "loss": 0.0893,
      "step": 97990
    },
    {
      "epoch": 38.06,
      "learning_rate": 1.4925566343042072e-05,
      "loss": 0.2239,
      "step": 98000
    },
    {
      "epoch": 38.06,
      "learning_rate": 1.4925048543689322e-05,
      "loss": 0.0606,
      "step": 98010
    },
    {
      "epoch": 38.07,
      "learning_rate": 1.4924530744336572e-05,
      "loss": 0.0445,
      "step": 98020
    },
    {
      "epoch": 38.07,
      "learning_rate": 1.492401294498382e-05,
      "loss": 0.2089,
      "step": 98030
    },
    {
      "epoch": 38.07,
      "learning_rate": 1.492349514563107e-05,
      "loss": 0.1229,
      "step": 98040
    },
    {
      "epoch": 38.08,
      "learning_rate": 1.4922977346278317e-05,
      "loss": 0.0649,
      "step": 98050
    },
    {
      "epoch": 38.08,
      "learning_rate": 1.4922459546925567e-05,
      "loss": 0.1671,
      "step": 98060
    },
    {
      "epoch": 38.09,
      "learning_rate": 1.4921941747572816e-05,
      "loss": 0.0612,
      "step": 98070
    },
    {
      "epoch": 38.09,
      "learning_rate": 1.4921423948220066e-05,
      "loss": 0.0411,
      "step": 98080
    },
    {
      "epoch": 38.09,
      "learning_rate": 1.4920906148867316e-05,
      "loss": 0.0514,
      "step": 98090
    },
    {
      "epoch": 38.1,
      "learning_rate": 1.4920388349514565e-05,
      "loss": 0.0103,
      "step": 98100
    },
    {
      "epoch": 38.1,
      "learning_rate": 1.4919870550161815e-05,
      "loss": 0.0283,
      "step": 98110
    },
    {
      "epoch": 38.1,
      "learning_rate": 1.4919352750809061e-05,
      "loss": 0.1448,
      "step": 98120
    },
    {
      "epoch": 38.11,
      "learning_rate": 1.491883495145631e-05,
      "loss": 0.0572,
      "step": 98130
    },
    {
      "epoch": 38.11,
      "learning_rate": 1.491831715210356e-05,
      "loss": 0.0059,
      "step": 98140
    },
    {
      "epoch": 38.12,
      "learning_rate": 1.491779935275081e-05,
      "loss": 0.0229,
      "step": 98150
    },
    {
      "epoch": 38.12,
      "learning_rate": 1.491728155339806e-05,
      "loss": 0.0564,
      "step": 98160
    },
    {
      "epoch": 38.12,
      "learning_rate": 1.4916763754045309e-05,
      "loss": 0.0138,
      "step": 98170
    },
    {
      "epoch": 38.13,
      "learning_rate": 1.4916245954692559e-05,
      "loss": 0.105,
      "step": 98180
    },
    {
      "epoch": 38.13,
      "learning_rate": 1.4915728155339808e-05,
      "loss": 0.0942,
      "step": 98190
    },
    {
      "epoch": 38.14,
      "learning_rate": 1.4915210355987055e-05,
      "loss": 0.0097,
      "step": 98200
    },
    {
      "epoch": 38.14,
      "learning_rate": 1.4914692556634304e-05,
      "loss": 0.2045,
      "step": 98210
    },
    {
      "epoch": 38.14,
      "learning_rate": 1.4914174757281554e-05,
      "loss": 0.0349,
      "step": 98220
    },
    {
      "epoch": 38.15,
      "learning_rate": 1.4913656957928803e-05,
      "loss": 0.0274,
      "step": 98230
    },
    {
      "epoch": 38.15,
      "learning_rate": 1.4913139158576053e-05,
      "loss": 0.014,
      "step": 98240
    },
    {
      "epoch": 38.16,
      "learning_rate": 1.4912621359223303e-05,
      "loss": 0.1243,
      "step": 98250
    },
    {
      "epoch": 38.16,
      "learning_rate": 1.4912103559870552e-05,
      "loss": 0.1181,
      "step": 98260
    },
    {
      "epoch": 38.16,
      "learning_rate": 1.4911585760517802e-05,
      "loss": 0.1167,
      "step": 98270
    },
    {
      "epoch": 38.17,
      "learning_rate": 1.4911067961165048e-05,
      "loss": 0.0165,
      "step": 98280
    },
    {
      "epoch": 38.17,
      "learning_rate": 1.4910550161812298e-05,
      "loss": 0.0169,
      "step": 98290
    },
    {
      "epoch": 38.17,
      "learning_rate": 1.4910032362459547e-05,
      "loss": 0.2017,
      "step": 98300
    },
    {
      "epoch": 38.18,
      "learning_rate": 1.4909514563106797e-05,
      "loss": 0.0331,
      "step": 98310
    },
    {
      "epoch": 38.18,
      "learning_rate": 1.4908996763754047e-05,
      "loss": 0.0805,
      "step": 98320
    },
    {
      "epoch": 38.19,
      "learning_rate": 1.4908478964401296e-05,
      "loss": 0.0559,
      "step": 98330
    },
    {
      "epoch": 38.19,
      "learning_rate": 1.4907961165048546e-05,
      "loss": 0.1904,
      "step": 98340
    },
    {
      "epoch": 38.19,
      "learning_rate": 1.4907443365695795e-05,
      "loss": 0.1672,
      "step": 98350
    },
    {
      "epoch": 38.2,
      "learning_rate": 1.4906925566343042e-05,
      "loss": 0.0821,
      "step": 98360
    },
    {
      "epoch": 38.2,
      "learning_rate": 1.4906407766990291e-05,
      "loss": 0.1725,
      "step": 98370
    },
    {
      "epoch": 38.21,
      "learning_rate": 1.4905889967637541e-05,
      "loss": 0.0242,
      "step": 98380
    },
    {
      "epoch": 38.21,
      "learning_rate": 1.490537216828479e-05,
      "loss": 0.1843,
      "step": 98390
    },
    {
      "epoch": 38.21,
      "learning_rate": 1.490485436893204e-05,
      "loss": 0.0928,
      "step": 98400
    },
    {
      "epoch": 38.22,
      "learning_rate": 1.490433656957929e-05,
      "loss": 0.1337,
      "step": 98410
    },
    {
      "epoch": 38.22,
      "learning_rate": 1.490381877022654e-05,
      "loss": 0.1252,
      "step": 98420
    },
    {
      "epoch": 38.23,
      "learning_rate": 1.4903300970873789e-05,
      "loss": 0.0405,
      "step": 98430
    },
    {
      "epoch": 38.23,
      "learning_rate": 1.4902783171521035e-05,
      "loss": 0.0574,
      "step": 98440
    },
    {
      "epoch": 38.23,
      "learning_rate": 1.4902265372168285e-05,
      "loss": 0.1459,
      "step": 98450
    },
    {
      "epoch": 38.24,
      "learning_rate": 1.4901747572815535e-05,
      "loss": 0.0598,
      "step": 98460
    },
    {
      "epoch": 38.24,
      "learning_rate": 1.4901229773462784e-05,
      "loss": 0.1163,
      "step": 98470
    },
    {
      "epoch": 38.24,
      "learning_rate": 1.4900711974110034e-05,
      "loss": 0.1276,
      "step": 98480
    },
    {
      "epoch": 38.25,
      "learning_rate": 1.4900194174757283e-05,
      "loss": 0.0791,
      "step": 98490
    },
    {
      "epoch": 38.25,
      "learning_rate": 1.4899676375404533e-05,
      "loss": 0.0288,
      "step": 98500
    },
    {
      "epoch": 38.26,
      "learning_rate": 1.4899158576051783e-05,
      "loss": 0.0464,
      "step": 98510
    },
    {
      "epoch": 38.26,
      "learning_rate": 1.4898640776699029e-05,
      "loss": 0.0077,
      "step": 98520
    },
    {
      "epoch": 38.26,
      "learning_rate": 1.4898122977346278e-05,
      "loss": 0.0025,
      "step": 98530
    },
    {
      "epoch": 38.27,
      "learning_rate": 1.4897605177993528e-05,
      "loss": 0.0439,
      "step": 98540
    },
    {
      "epoch": 38.27,
      "learning_rate": 1.4897087378640778e-05,
      "loss": 0.0452,
      "step": 98550
    },
    {
      "epoch": 38.28,
      "learning_rate": 1.4896569579288027e-05,
      "loss": 0.123,
      "step": 98560
    },
    {
      "epoch": 38.28,
      "learning_rate": 1.4896051779935277e-05,
      "loss": 0.0867,
      "step": 98570
    },
    {
      "epoch": 38.28,
      "learning_rate": 1.4895533980582527e-05,
      "loss": 0.0124,
      "step": 98580
    },
    {
      "epoch": 38.29,
      "learning_rate": 1.4895016181229776e-05,
      "loss": 0.0146,
      "step": 98590
    },
    {
      "epoch": 38.29,
      "learning_rate": 1.4894498381877022e-05,
      "loss": 0.2028,
      "step": 98600
    },
    {
      "epoch": 38.3,
      "learning_rate": 1.4893980582524272e-05,
      "loss": 0.0547,
      "step": 98610
    },
    {
      "epoch": 38.3,
      "learning_rate": 1.4893462783171522e-05,
      "loss": 0.0974,
      "step": 98620
    },
    {
      "epoch": 38.3,
      "learning_rate": 1.4892944983818771e-05,
      "loss": 0.1116,
      "step": 98630
    },
    {
      "epoch": 38.31,
      "learning_rate": 1.4892427184466021e-05,
      "loss": 0.0284,
      "step": 98640
    },
    {
      "epoch": 38.31,
      "learning_rate": 1.489190938511327e-05,
      "loss": 0.0312,
      "step": 98650
    },
    {
      "epoch": 38.31,
      "learning_rate": 1.489139158576052e-05,
      "loss": 0.0546,
      "step": 98660
    },
    {
      "epoch": 38.32,
      "learning_rate": 1.489087378640777e-05,
      "loss": 0.1118,
      "step": 98670
    },
    {
      "epoch": 38.32,
      "learning_rate": 1.489035598705502e-05,
      "loss": 0.1236,
      "step": 98680
    },
    {
      "epoch": 38.33,
      "learning_rate": 1.4889838187702266e-05,
      "loss": 0.0501,
      "step": 98690
    },
    {
      "epoch": 38.33,
      "learning_rate": 1.4889320388349515e-05,
      "loss": 0.0733,
      "step": 98700
    },
    {
      "epoch": 38.33,
      "learning_rate": 1.4888802588996765e-05,
      "loss": 0.1451,
      "step": 98710
    },
    {
      "epoch": 38.34,
      "learning_rate": 1.4888284789644014e-05,
      "loss": 0.1026,
      "step": 98720
    },
    {
      "epoch": 38.34,
      "learning_rate": 1.4887766990291264e-05,
      "loss": 0.0682,
      "step": 98730
    },
    {
      "epoch": 38.35,
      "learning_rate": 1.4887249190938514e-05,
      "loss": 0.0998,
      "step": 98740
    },
    {
      "epoch": 38.35,
      "learning_rate": 1.4886731391585763e-05,
      "loss": 0.3023,
      "step": 98750
    },
    {
      "epoch": 38.35,
      "learning_rate": 1.4886213592233011e-05,
      "loss": 0.1911,
      "step": 98760
    },
    {
      "epoch": 38.36,
      "learning_rate": 1.4885695792880259e-05,
      "loss": 0.0133,
      "step": 98770
    },
    {
      "epoch": 38.36,
      "learning_rate": 1.4885177993527509e-05,
      "loss": 0.0247,
      "step": 98780
    },
    {
      "epoch": 38.37,
      "learning_rate": 1.4884660194174758e-05,
      "loss": 0.0336,
      "step": 98790
    },
    {
      "epoch": 38.37,
      "learning_rate": 1.4884142394822008e-05,
      "loss": 0.1235,
      "step": 98800
    },
    {
      "epoch": 38.37,
      "learning_rate": 1.4883624595469258e-05,
      "loss": 0.0945,
      "step": 98810
    },
    {
      "epoch": 38.38,
      "learning_rate": 1.4883106796116507e-05,
      "loss": 0.0054,
      "step": 98820
    },
    {
      "epoch": 38.38,
      "learning_rate": 1.4882588996763757e-05,
      "loss": 0.3205,
      "step": 98830
    },
    {
      "epoch": 38.38,
      "learning_rate": 1.4882071197411005e-05,
      "loss": 0.0298,
      "step": 98840
    },
    {
      "epoch": 38.39,
      "learning_rate": 1.4881553398058253e-05,
      "loss": 0.0541,
      "step": 98850
    },
    {
      "epoch": 38.39,
      "learning_rate": 1.4881035598705502e-05,
      "loss": 0.0412,
      "step": 98860
    },
    {
      "epoch": 38.4,
      "learning_rate": 1.4880517799352752e-05,
      "loss": 0.0576,
      "step": 98870
    },
    {
      "epoch": 38.4,
      "learning_rate": 1.4880000000000002e-05,
      "loss": 0.1107,
      "step": 98880
    },
    {
      "epoch": 38.4,
      "learning_rate": 1.4879482200647251e-05,
      "loss": 0.0133,
      "step": 98890
    },
    {
      "epoch": 38.41,
      "learning_rate": 1.48789644012945e-05,
      "loss": 0.0234,
      "step": 98900
    },
    {
      "epoch": 38.41,
      "learning_rate": 1.4878446601941749e-05,
      "loss": 0.1651,
      "step": 98910
    },
    {
      "epoch": 38.42,
      "learning_rate": 1.4877928802588998e-05,
      "loss": 0.0051,
      "step": 98920
    },
    {
      "epoch": 38.42,
      "learning_rate": 1.4877411003236246e-05,
      "loss": 0.1336,
      "step": 98930
    },
    {
      "epoch": 38.42,
      "learning_rate": 1.4876893203883496e-05,
      "loss": 0.0286,
      "step": 98940
    },
    {
      "epoch": 38.43,
      "learning_rate": 1.4876375404530745e-05,
      "loss": 0.109,
      "step": 98950
    },
    {
      "epoch": 38.43,
      "learning_rate": 1.4875857605177995e-05,
      "loss": 0.0559,
      "step": 98960
    },
    {
      "epoch": 38.43,
      "learning_rate": 1.4875339805825245e-05,
      "loss": 0.2066,
      "step": 98970
    },
    {
      "epoch": 38.44,
      "learning_rate": 1.4874822006472494e-05,
      "loss": 0.1115,
      "step": 98980
    },
    {
      "epoch": 38.44,
      "learning_rate": 1.4874304207119742e-05,
      "loss": 0.0614,
      "step": 98990
    },
    {
      "epoch": 38.45,
      "learning_rate": 1.4873786407766992e-05,
      "loss": 0.1449,
      "step": 99000
    },
    {
      "epoch": 38.45,
      "learning_rate": 1.487326860841424e-05,
      "loss": 0.0305,
      "step": 99010
    },
    {
      "epoch": 38.45,
      "learning_rate": 1.487275080906149e-05,
      "loss": 0.0913,
      "step": 99020
    },
    {
      "epoch": 38.46,
      "learning_rate": 1.4872233009708739e-05,
      "loss": 0.0596,
      "step": 99030
    },
    {
      "epoch": 38.46,
      "learning_rate": 1.4871715210355989e-05,
      "loss": 0.0827,
      "step": 99040
    },
    {
      "epoch": 38.47,
      "learning_rate": 1.4871197411003238e-05,
      "loss": 0.1773,
      "step": 99050
    },
    {
      "epoch": 38.47,
      "learning_rate": 1.4870679611650486e-05,
      "loss": 0.1146,
      "step": 99060
    },
    {
      "epoch": 38.47,
      "learning_rate": 1.4870161812297736e-05,
      "loss": 0.1473,
      "step": 99070
    },
    {
      "epoch": 38.48,
      "learning_rate": 1.4869644012944985e-05,
      "loss": 0.1332,
      "step": 99080
    },
    {
      "epoch": 38.48,
      "learning_rate": 1.4869126213592233e-05,
      "loss": 0.0673,
      "step": 99090
    },
    {
      "epoch": 38.49,
      "learning_rate": 1.4868608414239483e-05,
      "loss": 0.1182,
      "step": 99100
    },
    {
      "epoch": 38.49,
      "learning_rate": 1.4868090614886733e-05,
      "loss": 0.08,
      "step": 99110
    },
    {
      "epoch": 38.49,
      "learning_rate": 1.4867572815533982e-05,
      "loss": 0.0318,
      "step": 99120
    },
    {
      "epoch": 38.5,
      "learning_rate": 1.4867055016181232e-05,
      "loss": 0.1283,
      "step": 99130
    },
    {
      "epoch": 38.5,
      "learning_rate": 1.486653721682848e-05,
      "loss": 0.2015,
      "step": 99140
    },
    {
      "epoch": 38.5,
      "learning_rate": 1.486601941747573e-05,
      "loss": 0.1117,
      "step": 99150
    },
    {
      "epoch": 38.51,
      "learning_rate": 1.4865501618122979e-05,
      "loss": 0.0761,
      "step": 99160
    },
    {
      "epoch": 38.51,
      "learning_rate": 1.4864983818770227e-05,
      "loss": 0.1562,
      "step": 99170
    },
    {
      "epoch": 38.52,
      "learning_rate": 1.4864466019417477e-05,
      "loss": 0.0489,
      "step": 99180
    },
    {
      "epoch": 38.52,
      "learning_rate": 1.4863948220064726e-05,
      "loss": 0.1195,
      "step": 99190
    },
    {
      "epoch": 38.52,
      "learning_rate": 1.4863430420711976e-05,
      "loss": 0.0076,
      "step": 99200
    },
    {
      "epoch": 38.53,
      "learning_rate": 1.4862912621359224e-05,
      "loss": 0.118,
      "step": 99210
    },
    {
      "epoch": 38.53,
      "learning_rate": 1.4862394822006473e-05,
      "loss": 0.0672,
      "step": 99220
    },
    {
      "epoch": 38.54,
      "learning_rate": 1.4861877022653723e-05,
      "loss": 0.1215,
      "step": 99230
    },
    {
      "epoch": 38.54,
      "learning_rate": 1.4861359223300973e-05,
      "loss": 0.1152,
      "step": 99240
    },
    {
      "epoch": 38.54,
      "learning_rate": 1.4860841423948222e-05,
      "loss": 0.2221,
      "step": 99250
    },
    {
      "epoch": 38.55,
      "learning_rate": 1.486032362459547e-05,
      "loss": 0.0165,
      "step": 99260
    },
    {
      "epoch": 38.55,
      "learning_rate": 1.485980582524272e-05,
      "loss": 0.1347,
      "step": 99270
    },
    {
      "epoch": 38.56,
      "learning_rate": 1.485928802588997e-05,
      "loss": 0.004,
      "step": 99280
    },
    {
      "epoch": 38.56,
      "learning_rate": 1.4858770226537217e-05,
      "loss": 0.0853,
      "step": 99290
    },
    {
      "epoch": 38.56,
      "learning_rate": 1.4858252427184467e-05,
      "loss": 0.037,
      "step": 99300
    },
    {
      "epoch": 38.57,
      "learning_rate": 1.4857734627831716e-05,
      "loss": 0.0797,
      "step": 99310
    },
    {
      "epoch": 38.57,
      "learning_rate": 1.4857216828478966e-05,
      "loss": 0.0135,
      "step": 99320
    },
    {
      "epoch": 38.57,
      "learning_rate": 1.4856699029126216e-05,
      "loss": 0.105,
      "step": 99330
    },
    {
      "epoch": 38.58,
      "learning_rate": 1.4856181229773464e-05,
      "loss": 0.0979,
      "step": 99340
    },
    {
      "epoch": 38.58,
      "learning_rate": 1.4855663430420713e-05,
      "loss": 0.0139,
      "step": 99350
    },
    {
      "epoch": 38.59,
      "learning_rate": 1.4855145631067963e-05,
      "loss": 0.1495,
      "step": 99360
    },
    {
      "epoch": 38.59,
      "learning_rate": 1.485462783171521e-05,
      "loss": 0.0074,
      "step": 99370
    },
    {
      "epoch": 38.59,
      "learning_rate": 1.485411003236246e-05,
      "loss": 0.1203,
      "step": 99380
    },
    {
      "epoch": 38.6,
      "learning_rate": 1.485359223300971e-05,
      "loss": 0.1085,
      "step": 99390
    },
    {
      "epoch": 38.6,
      "learning_rate": 1.485307443365696e-05,
      "loss": 0.0567,
      "step": 99400
    },
    {
      "epoch": 38.61,
      "learning_rate": 1.485255663430421e-05,
      "loss": 0.0734,
      "step": 99410
    },
    {
      "epoch": 38.61,
      "learning_rate": 1.4852038834951457e-05,
      "loss": 0.0933,
      "step": 99420
    },
    {
      "epoch": 38.61,
      "learning_rate": 1.4851521035598707e-05,
      "loss": 0.0551,
      "step": 99430
    },
    {
      "epoch": 38.62,
      "learning_rate": 1.4851003236245955e-05,
      "loss": 0.0616,
      "step": 99440
    },
    {
      "epoch": 38.62,
      "learning_rate": 1.4850485436893204e-05,
      "loss": 0.0331,
      "step": 99450
    },
    {
      "epoch": 38.63,
      "learning_rate": 1.4849967637540454e-05,
      "loss": 0.1591,
      "step": 99460
    },
    {
      "epoch": 38.63,
      "learning_rate": 1.4849449838187704e-05,
      "loss": 0.1119,
      "step": 99470
    },
    {
      "epoch": 38.63,
      "learning_rate": 1.4848932038834953e-05,
      "loss": 0.1549,
      "step": 99480
    },
    {
      "epoch": 38.64,
      "learning_rate": 1.4848414239482203e-05,
      "loss": 0.1135,
      "step": 99490
    },
    {
      "epoch": 38.64,
      "learning_rate": 1.484789644012945e-05,
      "loss": 0.0313,
      "step": 99500
    },
    {
      "epoch": 38.64,
      "learning_rate": 1.48473786407767e-05,
      "loss": 0.1744,
      "step": 99510
    },
    {
      "epoch": 38.65,
      "learning_rate": 1.4846860841423948e-05,
      "loss": 0.1557,
      "step": 99520
    },
    {
      "epoch": 38.65,
      "learning_rate": 1.4846343042071198e-05,
      "loss": 0.1287,
      "step": 99530
    },
    {
      "epoch": 38.66,
      "learning_rate": 1.4845825242718448e-05,
      "loss": 0.0355,
      "step": 99540
    },
    {
      "epoch": 38.66,
      "learning_rate": 1.4845307443365697e-05,
      "loss": 0.1067,
      "step": 99550
    },
    {
      "epoch": 38.66,
      "learning_rate": 1.4844789644012947e-05,
      "loss": 0.024,
      "step": 99560
    },
    {
      "epoch": 38.67,
      "learning_rate": 1.4844271844660196e-05,
      "loss": 0.2945,
      "step": 99570
    },
    {
      "epoch": 38.67,
      "learning_rate": 1.4843754045307444e-05,
      "loss": 0.0981,
      "step": 99580
    },
    {
      "epoch": 38.68,
      "learning_rate": 1.4843236245954692e-05,
      "loss": 0.0533,
      "step": 99590
    },
    {
      "epoch": 38.68,
      "learning_rate": 1.4842718446601942e-05,
      "loss": 0.0963,
      "step": 99600
    },
    {
      "epoch": 38.68,
      "learning_rate": 1.4842200647249191e-05,
      "loss": 0.235,
      "step": 99610
    },
    {
      "epoch": 38.69,
      "learning_rate": 1.4841682847896441e-05,
      "loss": 0.1765,
      "step": 99620
    },
    {
      "epoch": 38.69,
      "learning_rate": 1.484116504854369e-05,
      "loss": 0.0639,
      "step": 99630
    },
    {
      "epoch": 38.7,
      "learning_rate": 1.484064724919094e-05,
      "loss": 0.1098,
      "step": 99640
    },
    {
      "epoch": 38.7,
      "learning_rate": 1.484012944983819e-05,
      "loss": 0.0299,
      "step": 99650
    },
    {
      "epoch": 38.7,
      "learning_rate": 1.4839611650485438e-05,
      "loss": 0.1043,
      "step": 99660
    },
    {
      "epoch": 38.71,
      "learning_rate": 1.4839093851132686e-05,
      "loss": 0.2268,
      "step": 99670
    },
    {
      "epoch": 38.71,
      "learning_rate": 1.4838576051779935e-05,
      "loss": 0.2019,
      "step": 99680
    },
    {
      "epoch": 38.71,
      "learning_rate": 1.4838058252427185e-05,
      "loss": 0.0589,
      "step": 99690
    },
    {
      "epoch": 38.72,
      "learning_rate": 1.4837540453074435e-05,
      "loss": 0.0805,
      "step": 99700
    },
    {
      "epoch": 38.72,
      "learning_rate": 1.4837022653721684e-05,
      "loss": 0.0912,
      "step": 99710
    },
    {
      "epoch": 38.73,
      "learning_rate": 1.4836504854368934e-05,
      "loss": 0.0487,
      "step": 99720
    },
    {
      "epoch": 38.73,
      "learning_rate": 1.4835987055016183e-05,
      "loss": 0.1106,
      "step": 99730
    },
    {
      "epoch": 38.73,
      "learning_rate": 1.483546925566343e-05,
      "loss": 0.0887,
      "step": 99740
    },
    {
      "epoch": 38.74,
      "learning_rate": 1.483495145631068e-05,
      "loss": 0.0291,
      "step": 99750
    },
    {
      "epoch": 38.74,
      "learning_rate": 1.4834433656957929e-05,
      "loss": 0.0446,
      "step": 99760
    },
    {
      "epoch": 38.75,
      "learning_rate": 1.4833915857605179e-05,
      "loss": 0.0352,
      "step": 99770
    },
    {
      "epoch": 38.75,
      "learning_rate": 1.4833398058252428e-05,
      "loss": 0.0275,
      "step": 99780
    },
    {
      "epoch": 38.75,
      "learning_rate": 1.4832880258899678e-05,
      "loss": 0.1322,
      "step": 99790
    },
    {
      "epoch": 38.76,
      "learning_rate": 1.4832362459546927e-05,
      "loss": 0.0564,
      "step": 99800
    },
    {
      "epoch": 38.76,
      "learning_rate": 1.4831844660194177e-05,
      "loss": 0.195,
      "step": 99810
    },
    {
      "epoch": 38.77,
      "learning_rate": 1.4831326860841427e-05,
      "loss": 0.0602,
      "step": 99820
    },
    {
      "epoch": 38.77,
      "learning_rate": 1.4830809061488673e-05,
      "loss": 0.0745,
      "step": 99830
    },
    {
      "epoch": 38.77,
      "learning_rate": 1.4830291262135923e-05,
      "loss": 0.1419,
      "step": 99840
    },
    {
      "epoch": 38.78,
      "learning_rate": 1.4829773462783172e-05,
      "loss": 0.2613,
      "step": 99850
    },
    {
      "epoch": 38.78,
      "learning_rate": 1.4829255663430422e-05,
      "loss": 0.2078,
      "step": 99860
    },
    {
      "epoch": 38.78,
      "learning_rate": 1.4828737864077671e-05,
      "loss": 0.1124,
      "step": 99870
    },
    {
      "epoch": 38.79,
      "learning_rate": 1.4828220064724921e-05,
      "loss": 0.1062,
      "step": 99880
    },
    {
      "epoch": 38.79,
      "learning_rate": 1.482770226537217e-05,
      "loss": 0.0164,
      "step": 99890
    },
    {
      "epoch": 38.8,
      "learning_rate": 1.482718446601942e-05,
      "loss": 0.3654,
      "step": 99900
    },
    {
      "epoch": 38.8,
      "learning_rate": 1.4826666666666666e-05,
      "loss": 0.0662,
      "step": 99910
    },
    {
      "epoch": 38.8,
      "learning_rate": 1.4826148867313916e-05,
      "loss": 0.1691,
      "step": 99920
    },
    {
      "epoch": 38.81,
      "learning_rate": 1.4825631067961166e-05,
      "loss": 0.0953,
      "step": 99930
    },
    {
      "epoch": 38.81,
      "learning_rate": 1.4825113268608415e-05,
      "loss": 0.0405,
      "step": 99940
    },
    {
      "epoch": 38.82,
      "learning_rate": 1.4824595469255665e-05,
      "loss": 0.0531,
      "step": 99950
    },
    {
      "epoch": 38.82,
      "learning_rate": 1.4824077669902915e-05,
      "loss": 0.0047,
      "step": 99960
    },
    {
      "epoch": 38.82,
      "learning_rate": 1.4823559870550164e-05,
      "loss": 0.0307,
      "step": 99970
    },
    {
      "epoch": 38.83,
      "learning_rate": 1.4823042071197414e-05,
      "loss": 0.0631,
      "step": 99980
    },
    {
      "epoch": 38.83,
      "learning_rate": 1.482252427184466e-05,
      "loss": 0.0663,
      "step": 99990
    },
    {
      "epoch": 38.83,
      "learning_rate": 1.482200647249191e-05,
      "loss": 0.0995,
      "step": 100000
    },
    {
      "epoch": 38.84,
      "learning_rate": 1.482148867313916e-05,
      "loss": 0.0657,
      "step": 100010
    },
    {
      "epoch": 38.84,
      "learning_rate": 1.4820970873786409e-05,
      "loss": 0.0476,
      "step": 100020
    },
    {
      "epoch": 38.85,
      "learning_rate": 1.4820453074433658e-05,
      "loss": 0.0809,
      "step": 100030
    },
    {
      "epoch": 38.85,
      "learning_rate": 1.4819935275080908e-05,
      "loss": 0.092,
      "step": 100040
    },
    {
      "epoch": 38.85,
      "learning_rate": 1.4819417475728158e-05,
      "loss": 0.025,
      "step": 100050
    },
    {
      "epoch": 38.86,
      "learning_rate": 1.4818899676375407e-05,
      "loss": 0.0307,
      "step": 100060
    },
    {
      "epoch": 38.86,
      "learning_rate": 1.4818381877022654e-05,
      "loss": 0.1003,
      "step": 100070
    },
    {
      "epoch": 38.87,
      "learning_rate": 1.4817864077669903e-05,
      "loss": 0.0164,
      "step": 100080
    },
    {
      "epoch": 38.87,
      "learning_rate": 1.4817346278317153e-05,
      "loss": 0.0425,
      "step": 100090
    },
    {
      "epoch": 38.87,
      "learning_rate": 1.4816828478964402e-05,
      "loss": 0.0542,
      "step": 100100
    },
    {
      "epoch": 38.88,
      "learning_rate": 1.4816310679611652e-05,
      "loss": 0.1034,
      "step": 100110
    },
    {
      "epoch": 38.88,
      "learning_rate": 1.4815792880258902e-05,
      "loss": 0.0907,
      "step": 100120
    },
    {
      "epoch": 38.89,
      "learning_rate": 1.4815275080906151e-05,
      "loss": 0.0755,
      "step": 100130
    },
    {
      "epoch": 38.89,
      "learning_rate": 1.4814757281553401e-05,
      "loss": 0.0016,
      "step": 100140
    },
    {
      "epoch": 38.89,
      "learning_rate": 1.4814239482200647e-05,
      "loss": 0.092,
      "step": 100150
    },
    {
      "epoch": 38.9,
      "learning_rate": 1.4813721682847897e-05,
      "loss": 0.0282,
      "step": 100160
    },
    {
      "epoch": 38.9,
      "learning_rate": 1.4813203883495146e-05,
      "loss": 0.062,
      "step": 100170
    },
    {
      "epoch": 38.9,
      "learning_rate": 1.4812686084142396e-05,
      "loss": 0.1852,
      "step": 100180
    },
    {
      "epoch": 38.91,
      "learning_rate": 1.4812168284789646e-05,
      "loss": 0.1167,
      "step": 100190
    },
    {
      "epoch": 38.91,
      "learning_rate": 1.4811650485436895e-05,
      "loss": 0.1023,
      "step": 100200
    },
    {
      "epoch": 38.92,
      "learning_rate": 1.4811132686084145e-05,
      "loss": 0.1074,
      "step": 100210
    },
    {
      "epoch": 38.92,
      "learning_rate": 1.4810614886731394e-05,
      "loss": 0.128,
      "step": 100220
    },
    {
      "epoch": 38.92,
      "learning_rate": 1.481009708737864e-05,
      "loss": 0.1094,
      "step": 100230
    },
    {
      "epoch": 38.93,
      "learning_rate": 1.480957928802589e-05,
      "loss": 0.153,
      "step": 100240
    },
    {
      "epoch": 38.93,
      "learning_rate": 1.480906148867314e-05,
      "loss": 0.0772,
      "step": 100250
    },
    {
      "epoch": 38.94,
      "learning_rate": 1.480854368932039e-05,
      "loss": 0.178,
      "step": 100260
    },
    {
      "epoch": 38.94,
      "learning_rate": 1.4808025889967639e-05,
      "loss": 0.1571,
      "step": 100270
    },
    {
      "epoch": 38.94,
      "learning_rate": 1.4807508090614889e-05,
      "loss": 0.0535,
      "step": 100280
    },
    {
      "epoch": 38.95,
      "learning_rate": 1.4806990291262138e-05,
      "loss": 0.0066,
      "step": 100290
    },
    {
      "epoch": 38.95,
      "learning_rate": 1.4806472491909388e-05,
      "loss": 0.0617,
      "step": 100300
    },
    {
      "epoch": 38.96,
      "learning_rate": 1.4805954692556634e-05,
      "loss": 0.061,
      "step": 100310
    },
    {
      "epoch": 38.96,
      "learning_rate": 1.4805436893203884e-05,
      "loss": 0.0989,
      "step": 100320
    },
    {
      "epoch": 38.96,
      "learning_rate": 1.4804919093851133e-05,
      "loss": 0.0277,
      "step": 100330
    },
    {
      "epoch": 38.97,
      "learning_rate": 1.4804401294498383e-05,
      "loss": 0.1652,
      "step": 100340
    },
    {
      "epoch": 38.97,
      "learning_rate": 1.4803883495145633e-05,
      "loss": 0.2654,
      "step": 100350
    },
    {
      "epoch": 38.97,
      "learning_rate": 1.4803365695792882e-05,
      "loss": 0.1039,
      "step": 100360
    },
    {
      "epoch": 38.98,
      "learning_rate": 1.4802847896440132e-05,
      "loss": 0.0731,
      "step": 100370
    },
    {
      "epoch": 38.98,
      "learning_rate": 1.480233009708738e-05,
      "loss": 0.1836,
      "step": 100380
    },
    {
      "epoch": 38.99,
      "learning_rate": 1.480181229773463e-05,
      "loss": 0.198,
      "step": 100390
    },
    {
      "epoch": 38.99,
      "learning_rate": 1.4801294498381877e-05,
      "loss": 0.0214,
      "step": 100400
    },
    {
      "epoch": 38.99,
      "learning_rate": 1.4800776699029127e-05,
      "loss": 0.1549,
      "step": 100410
    },
    {
      "epoch": 39.0,
      "learning_rate": 1.4800258899676377e-05,
      "loss": 0.0409,
      "step": 100420
    },
    {
      "epoch": 39.0,
      "eval_accuracy": 0.9482806052269601,
      "eval_loss": 0.3205643892288208,
      "eval_runtime": 8.2805,
      "eval_samples_per_second": 438.983,
      "eval_steps_per_second": 54.948,
      "step": 100425
    },
    {
      "epoch": 39.0,
      "learning_rate": 1.4799741100323626e-05,
      "loss": 0.0326,
      "step": 100430
    },
    {
      "epoch": 39.01,
      "learning_rate": 1.4799223300970876e-05,
      "loss": 0.196,
      "step": 100440
    },
    {
      "epoch": 39.01,
      "learning_rate": 1.4798705501618125e-05,
      "loss": 0.0952,
      "step": 100450
    },
    {
      "epoch": 39.01,
      "learning_rate": 1.4798187702265373e-05,
      "loss": 0.1202,
      "step": 100460
    },
    {
      "epoch": 39.02,
      "learning_rate": 1.4797669902912623e-05,
      "loss": 0.0551,
      "step": 100470
    },
    {
      "epoch": 39.02,
      "learning_rate": 1.4797152103559871e-05,
      "loss": 0.1816,
      "step": 100480
    },
    {
      "epoch": 39.03,
      "learning_rate": 1.479663430420712e-05,
      "loss": 0.0381,
      "step": 100490
    },
    {
      "epoch": 39.03,
      "learning_rate": 1.479611650485437e-05,
      "loss": 0.0738,
      "step": 100500
    },
    {
      "epoch": 39.03,
      "learning_rate": 1.479559870550162e-05,
      "loss": 0.0867,
      "step": 100510
    },
    {
      "epoch": 39.04,
      "learning_rate": 1.479508090614887e-05,
      "loss": 0.0705,
      "step": 100520
    },
    {
      "epoch": 39.04,
      "learning_rate": 1.4794563106796117e-05,
      "loss": 0.0706,
      "step": 100530
    },
    {
      "epoch": 39.04,
      "learning_rate": 1.4794045307443367e-05,
      "loss": 0.0753,
      "step": 100540
    },
    {
      "epoch": 39.05,
      "learning_rate": 1.4793527508090617e-05,
      "loss": 0.0529,
      "step": 100550
    },
    {
      "epoch": 39.05,
      "learning_rate": 1.4793009708737865e-05,
      "loss": 0.0088,
      "step": 100560
    },
    {
      "epoch": 39.06,
      "learning_rate": 1.4792491909385114e-05,
      "loss": 0.0159,
      "step": 100570
    },
    {
      "epoch": 39.06,
      "learning_rate": 1.4791974110032364e-05,
      "loss": 0.0012,
      "step": 100580
    },
    {
      "epoch": 39.06,
      "learning_rate": 1.4791456310679613e-05,
      "loss": 0.0753,
      "step": 100590
    },
    {
      "epoch": 39.07,
      "learning_rate": 1.4790938511326863e-05,
      "loss": 0.1147,
      "step": 100600
    },
    {
      "epoch": 39.07,
      "learning_rate": 1.4790420711974111e-05,
      "loss": 0.0201,
      "step": 100610
    },
    {
      "epoch": 39.08,
      "learning_rate": 1.478990291262136e-05,
      "loss": 0.1459,
      "step": 100620
    },
    {
      "epoch": 39.08,
      "learning_rate": 1.478938511326861e-05,
      "loss": 0.0649,
      "step": 100630
    },
    {
      "epoch": 39.08,
      "learning_rate": 1.4788867313915858e-05,
      "loss": 0.1094,
      "step": 100640
    },
    {
      "epoch": 39.09,
      "learning_rate": 1.4788349514563108e-05,
      "loss": 0.0616,
      "step": 100650
    },
    {
      "epoch": 39.09,
      "learning_rate": 1.4787831715210357e-05,
      "loss": 0.1036,
      "step": 100660
    },
    {
      "epoch": 39.1,
      "learning_rate": 1.4787313915857607e-05,
      "loss": 0.1349,
      "step": 100670
    },
    {
      "epoch": 39.1,
      "learning_rate": 1.4786796116504855e-05,
      "loss": 0.008,
      "step": 100680
    },
    {
      "epoch": 39.1,
      "learning_rate": 1.4786278317152104e-05,
      "loss": 0.0719,
      "step": 100690
    },
    {
      "epoch": 39.11,
      "learning_rate": 1.4785760517799354e-05,
      "loss": 0.0172,
      "step": 100700
    },
    {
      "epoch": 39.11,
      "learning_rate": 1.4785242718446604e-05,
      "loss": 0.0247,
      "step": 100710
    },
    {
      "epoch": 39.11,
      "learning_rate": 1.4784724919093852e-05,
      "loss": 0.1675,
      "step": 100720
    },
    {
      "epoch": 39.12,
      "learning_rate": 1.4784207119741101e-05,
      "loss": 0.0253,
      "step": 100730
    },
    {
      "epoch": 39.12,
      "learning_rate": 1.4783689320388351e-05,
      "loss": 0.0589,
      "step": 100740
    },
    {
      "epoch": 39.13,
      "learning_rate": 1.47831715210356e-05,
      "loss": 0.0433,
      "step": 100750
    },
    {
      "epoch": 39.13,
      "learning_rate": 1.4782653721682848e-05,
      "loss": 0.0155,
      "step": 100760
    },
    {
      "epoch": 39.13,
      "learning_rate": 1.4782135922330098e-05,
      "loss": 0.1465,
      "step": 100770
    },
    {
      "epoch": 39.14,
      "learning_rate": 1.4781618122977348e-05,
      "loss": 0.1849,
      "step": 100780
    },
    {
      "epoch": 39.14,
      "learning_rate": 1.4781100323624597e-05,
      "loss": 0.0778,
      "step": 100790
    },
    {
      "epoch": 39.15,
      "learning_rate": 1.4780582524271845e-05,
      "loss": 0.0338,
      "step": 100800
    },
    {
      "epoch": 39.15,
      "learning_rate": 1.4780064724919095e-05,
      "loss": 0.0299,
      "step": 100810
    },
    {
      "epoch": 39.15,
      "learning_rate": 1.4779546925566344e-05,
      "loss": 0.1034,
      "step": 100820
    },
    {
      "epoch": 39.16,
      "learning_rate": 1.4779029126213594e-05,
      "loss": 0.0614,
      "step": 100830
    },
    {
      "epoch": 39.16,
      "learning_rate": 1.4778511326860842e-05,
      "loss": 0.001,
      "step": 100840
    },
    {
      "epoch": 39.17,
      "learning_rate": 1.4777993527508092e-05,
      "loss": 0.1134,
      "step": 100850
    },
    {
      "epoch": 39.17,
      "learning_rate": 1.4777475728155341e-05,
      "loss": 0.1248,
      "step": 100860
    },
    {
      "epoch": 39.17,
      "learning_rate": 1.477695792880259e-05,
      "loss": 0.1521,
      "step": 100870
    },
    {
      "epoch": 39.18,
      "learning_rate": 1.4776440129449839e-05,
      "loss": 0.0232,
      "step": 100880
    },
    {
      "epoch": 39.18,
      "learning_rate": 1.4775922330097088e-05,
      "loss": 0.0097,
      "step": 100890
    },
    {
      "epoch": 39.18,
      "learning_rate": 1.4775404530744338e-05,
      "loss": 0.1155,
      "step": 100900
    },
    {
      "epoch": 39.19,
      "learning_rate": 1.4774886731391586e-05,
      "loss": 0.0431,
      "step": 100910
    },
    {
      "epoch": 39.19,
      "learning_rate": 1.4774368932038836e-05,
      "loss": 0.1555,
      "step": 100920
    },
    {
      "epoch": 39.2,
      "learning_rate": 1.4773851132686085e-05,
      "loss": 0.0232,
      "step": 100930
    },
    {
      "epoch": 39.2,
      "learning_rate": 1.4773333333333335e-05,
      "loss": 0.1632,
      "step": 100940
    },
    {
      "epoch": 39.2,
      "learning_rate": 1.4772815533980584e-05,
      "loss": 0.152,
      "step": 100950
    },
    {
      "epoch": 39.21,
      "learning_rate": 1.4772297734627834e-05,
      "loss": 0.1385,
      "step": 100960
    },
    {
      "epoch": 39.21,
      "learning_rate": 1.4771779935275082e-05,
      "loss": 0.0568,
      "step": 100970
    },
    {
      "epoch": 39.22,
      "learning_rate": 1.4771262135922332e-05,
      "loss": 0.0036,
      "step": 100980
    },
    {
      "epoch": 39.22,
      "learning_rate": 1.477074433656958e-05,
      "loss": 0.2125,
      "step": 100990
    },
    {
      "epoch": 39.22,
      "learning_rate": 1.4770226537216829e-05,
      "loss": 0.0315,
      "step": 101000
    },
    {
      "epoch": 39.23,
      "learning_rate": 1.4769708737864079e-05,
      "loss": 0.2239,
      "step": 101010
    },
    {
      "epoch": 39.23,
      "learning_rate": 1.4769190938511328e-05,
      "loss": 0.0808,
      "step": 101020
    },
    {
      "epoch": 39.23,
      "learning_rate": 1.4768673139158578e-05,
      "loss": 0.159,
      "step": 101030
    },
    {
      "epoch": 39.24,
      "learning_rate": 1.4768155339805828e-05,
      "loss": 0.1385,
      "step": 101040
    },
    {
      "epoch": 39.24,
      "learning_rate": 1.4767637540453075e-05,
      "loss": 0.0075,
      "step": 101050
    },
    {
      "epoch": 39.25,
      "learning_rate": 1.4767119741100323e-05,
      "loss": 0.0414,
      "step": 101060
    },
    {
      "epoch": 39.25,
      "learning_rate": 1.4766601941747573e-05,
      "loss": 0.0454,
      "step": 101070
    },
    {
      "epoch": 39.25,
      "learning_rate": 1.4766084142394823e-05,
      "loss": 0.1061,
      "step": 101080
    },
    {
      "epoch": 39.26,
      "learning_rate": 1.4765566343042072e-05,
      "loss": 0.1927,
      "step": 101090
    },
    {
      "epoch": 39.26,
      "learning_rate": 1.4765048543689322e-05,
      "loss": 0.131,
      "step": 101100
    },
    {
      "epoch": 39.27,
      "learning_rate": 1.4764530744336571e-05,
      "loss": 0.0615,
      "step": 101110
    },
    {
      "epoch": 39.27,
      "learning_rate": 1.4764012944983821e-05,
      "loss": 0.0475,
      "step": 101120
    },
    {
      "epoch": 39.27,
      "learning_rate": 1.4763495145631069e-05,
      "loss": 0.1368,
      "step": 101130
    },
    {
      "epoch": 39.28,
      "learning_rate": 1.4762977346278317e-05,
      "loss": 0.0901,
      "step": 101140
    },
    {
      "epoch": 39.28,
      "learning_rate": 1.4762459546925567e-05,
      "loss": 0.0224,
      "step": 101150
    },
    {
      "epoch": 39.29,
      "learning_rate": 1.4761941747572816e-05,
      "loss": 0.1132,
      "step": 101160
    },
    {
      "epoch": 39.29,
      "learning_rate": 1.4761423948220066e-05,
      "loss": 0.0308,
      "step": 101170
    },
    {
      "epoch": 39.29,
      "learning_rate": 1.4760906148867315e-05,
      "loss": 0.1596,
      "step": 101180
    },
    {
      "epoch": 39.3,
      "learning_rate": 1.4760388349514565e-05,
      "loss": 0.1732,
      "step": 101190
    },
    {
      "epoch": 39.3,
      "learning_rate": 1.4759870550161815e-05,
      "loss": 0.2038,
      "step": 101200
    },
    {
      "epoch": 39.3,
      "learning_rate": 1.4759352750809061e-05,
      "loss": 0.2404,
      "step": 101210
    },
    {
      "epoch": 39.31,
      "learning_rate": 1.475883495145631e-05,
      "loss": 0.064,
      "step": 101220
    },
    {
      "epoch": 39.31,
      "learning_rate": 1.475831715210356e-05,
      "loss": 0.1412,
      "step": 101230
    },
    {
      "epoch": 39.32,
      "learning_rate": 1.475779935275081e-05,
      "loss": 0.1361,
      "step": 101240
    },
    {
      "epoch": 39.32,
      "learning_rate": 1.475728155339806e-05,
      "loss": 0.0874,
      "step": 101250
    },
    {
      "epoch": 39.32,
      "learning_rate": 1.4756763754045309e-05,
      "loss": 0.1199,
      "step": 101260
    },
    {
      "epoch": 39.33,
      "learning_rate": 1.4756245954692559e-05,
      "loss": 0.0646,
      "step": 101270
    },
    {
      "epoch": 39.33,
      "learning_rate": 1.4755728155339808e-05,
      "loss": 0.0258,
      "step": 101280
    },
    {
      "epoch": 39.34,
      "learning_rate": 1.4755210355987054e-05,
      "loss": 0.0081,
      "step": 101290
    },
    {
      "epoch": 39.34,
      "learning_rate": 1.4754692556634304e-05,
      "loss": 0.1741,
      "step": 101300
    },
    {
      "epoch": 39.34,
      "learning_rate": 1.4754174757281554e-05,
      "loss": 0.1738,
      "step": 101310
    },
    {
      "epoch": 39.35,
      "learning_rate": 1.4753656957928803e-05,
      "loss": 0.1049,
      "step": 101320
    },
    {
      "epoch": 39.35,
      "learning_rate": 1.4753139158576053e-05,
      "loss": 0.0235,
      "step": 101330
    },
    {
      "epoch": 39.36,
      "learning_rate": 1.4752621359223303e-05,
      "loss": 0.0563,
      "step": 101340
    },
    {
      "epoch": 39.36,
      "learning_rate": 1.4752103559870552e-05,
      "loss": 0.1389,
      "step": 101350
    },
    {
      "epoch": 39.36,
      "learning_rate": 1.4751585760517802e-05,
      "loss": 0.1041,
      "step": 101360
    },
    {
      "epoch": 39.37,
      "learning_rate": 1.4751067961165048e-05,
      "loss": 0.0265,
      "step": 101370
    },
    {
      "epoch": 39.37,
      "learning_rate": 1.4750550161812298e-05,
      "loss": 0.1123,
      "step": 101380
    },
    {
      "epoch": 39.37,
      "learning_rate": 1.4750032362459547e-05,
      "loss": 0.0455,
      "step": 101390
    },
    {
      "epoch": 39.38,
      "learning_rate": 1.4749514563106797e-05,
      "loss": 0.0727,
      "step": 101400
    },
    {
      "epoch": 39.38,
      "learning_rate": 1.4748996763754046e-05,
      "loss": 0.1382,
      "step": 101410
    },
    {
      "epoch": 39.39,
      "learning_rate": 1.4748478964401296e-05,
      "loss": 0.0773,
      "step": 101420
    },
    {
      "epoch": 39.39,
      "learning_rate": 1.4747961165048546e-05,
      "loss": 0.0624,
      "step": 101430
    },
    {
      "epoch": 39.39,
      "learning_rate": 1.4747443365695795e-05,
      "loss": 0.1651,
      "step": 101440
    },
    {
      "epoch": 39.4,
      "learning_rate": 1.4746925566343042e-05,
      "loss": 0.0392,
      "step": 101450
    },
    {
      "epoch": 39.4,
      "learning_rate": 1.4746407766990291e-05,
      "loss": 0.0621,
      "step": 101460
    },
    {
      "epoch": 39.41,
      "learning_rate": 1.474588996763754e-05,
      "loss": 0.0484,
      "step": 101470
    },
    {
      "epoch": 39.41,
      "learning_rate": 1.474537216828479e-05,
      "loss": 0.1097,
      "step": 101480
    },
    {
      "epoch": 39.41,
      "learning_rate": 1.474485436893204e-05,
      "loss": 0.0711,
      "step": 101490
    },
    {
      "epoch": 39.42,
      "learning_rate": 1.474433656957929e-05,
      "loss": 0.2309,
      "step": 101500
    },
    {
      "epoch": 39.42,
      "learning_rate": 1.474381877022654e-05,
      "loss": 0.142,
      "step": 101510
    },
    {
      "epoch": 39.43,
      "learning_rate": 1.4743300970873789e-05,
      "loss": 0.0045,
      "step": 101520
    },
    {
      "epoch": 39.43,
      "learning_rate": 1.4742783171521039e-05,
      "loss": 0.0051,
      "step": 101530
    },
    {
      "epoch": 39.43,
      "learning_rate": 1.4742265372168285e-05,
      "loss": 0.0827,
      "step": 101540
    },
    {
      "epoch": 39.44,
      "learning_rate": 1.4741747572815534e-05,
      "loss": 0.0729,
      "step": 101550
    },
    {
      "epoch": 39.44,
      "learning_rate": 1.4741229773462784e-05,
      "loss": 0.1171,
      "step": 101560
    },
    {
      "epoch": 39.44,
      "learning_rate": 1.4740711974110034e-05,
      "loss": 0.0454,
      "step": 101570
    },
    {
      "epoch": 39.45,
      "learning_rate": 1.4740194174757283e-05,
      "loss": 0.0298,
      "step": 101580
    },
    {
      "epoch": 39.45,
      "learning_rate": 1.4739676375404533e-05,
      "loss": 0.193,
      "step": 101590
    },
    {
      "epoch": 39.46,
      "learning_rate": 1.4739158576051782e-05,
      "loss": 0.0743,
      "step": 101600
    },
    {
      "epoch": 39.46,
      "learning_rate": 1.4738640776699032e-05,
      "loss": 0.0672,
      "step": 101610
    },
    {
      "epoch": 39.46,
      "learning_rate": 1.4738122977346278e-05,
      "loss": 0.0298,
      "step": 101620
    },
    {
      "epoch": 39.47,
      "learning_rate": 1.4737605177993528e-05,
      "loss": 0.118,
      "step": 101630
    },
    {
      "epoch": 39.47,
      "learning_rate": 1.4737087378640778e-05,
      "loss": 0.0988,
      "step": 101640
    },
    {
      "epoch": 39.48,
      "learning_rate": 1.4736569579288027e-05,
      "loss": 0.0427,
      "step": 101650
    },
    {
      "epoch": 39.48,
      "learning_rate": 1.4736051779935277e-05,
      "loss": 0.0976,
      "step": 101660
    },
    {
      "epoch": 39.48,
      "learning_rate": 1.4735533980582526e-05,
      "loss": 0.0387,
      "step": 101670
    },
    {
      "epoch": 39.49,
      "learning_rate": 1.4735016181229776e-05,
      "loss": 0.0638,
      "step": 101680
    },
    {
      "epoch": 39.49,
      "learning_rate": 1.4734498381877026e-05,
      "loss": 0.1275,
      "step": 101690
    },
    {
      "epoch": 39.5,
      "learning_rate": 1.4733980582524272e-05,
      "loss": 0.0755,
      "step": 101700
    },
    {
      "epoch": 39.5,
      "learning_rate": 1.4733462783171521e-05,
      "loss": 0.0256,
      "step": 101710
    },
    {
      "epoch": 39.5,
      "learning_rate": 1.4732944983818771e-05,
      "loss": 0.1193,
      "step": 101720
    },
    {
      "epoch": 39.51,
      "learning_rate": 1.473242718446602e-05,
      "loss": 0.1469,
      "step": 101730
    },
    {
      "epoch": 39.51,
      "learning_rate": 1.473190938511327e-05,
      "loss": 0.1694,
      "step": 101740
    },
    {
      "epoch": 39.51,
      "learning_rate": 1.473139158576052e-05,
      "loss": 0.101,
      "step": 101750
    },
    {
      "epoch": 39.52,
      "learning_rate": 1.473087378640777e-05,
      "loss": 0.0822,
      "step": 101760
    },
    {
      "epoch": 39.52,
      "learning_rate": 1.473035598705502e-05,
      "loss": 0.1619,
      "step": 101770
    },
    {
      "epoch": 39.53,
      "learning_rate": 1.4729838187702265e-05,
      "loss": 0.1513,
      "step": 101780
    },
    {
      "epoch": 39.53,
      "learning_rate": 1.4729320388349515e-05,
      "loss": 0.0732,
      "step": 101790
    },
    {
      "epoch": 39.53,
      "learning_rate": 1.4728802588996765e-05,
      "loss": 0.0679,
      "step": 101800
    },
    {
      "epoch": 39.54,
      "learning_rate": 1.4728284789644014e-05,
      "loss": 0.1002,
      "step": 101810
    },
    {
      "epoch": 39.54,
      "learning_rate": 1.4727766990291264e-05,
      "loss": 0.0731,
      "step": 101820
    },
    {
      "epoch": 39.55,
      "learning_rate": 1.4727249190938513e-05,
      "loss": 0.0815,
      "step": 101830
    },
    {
      "epoch": 39.55,
      "learning_rate": 1.4726731391585763e-05,
      "loss": 0.011,
      "step": 101840
    },
    {
      "epoch": 39.55,
      "learning_rate": 1.4726213592233011e-05,
      "loss": 0.0034,
      "step": 101850
    },
    {
      "epoch": 39.56,
      "learning_rate": 1.4725695792880259e-05,
      "loss": 0.1259,
      "step": 101860
    },
    {
      "epoch": 39.56,
      "learning_rate": 1.4725177993527509e-05,
      "loss": 0.1331,
      "step": 101870
    },
    {
      "epoch": 39.57,
      "learning_rate": 1.4724660194174758e-05,
      "loss": 0.1252,
      "step": 101880
    },
    {
      "epoch": 39.57,
      "learning_rate": 1.4724142394822008e-05,
      "loss": 0.12,
      "step": 101890
    },
    {
      "epoch": 39.57,
      "learning_rate": 1.4723624595469257e-05,
      "loss": 0.037,
      "step": 101900
    },
    {
      "epoch": 39.58,
      "learning_rate": 1.4723106796116507e-05,
      "loss": 0.1042,
      "step": 101910
    },
    {
      "epoch": 39.58,
      "learning_rate": 1.4722588996763757e-05,
      "loss": 0.1297,
      "step": 101920
    },
    {
      "epoch": 39.58,
      "learning_rate": 1.4722071197411005e-05,
      "loss": 0.0754,
      "step": 101930
    },
    {
      "epoch": 39.59,
      "learning_rate": 1.4721553398058253e-05,
      "loss": 0.1813,
      "step": 101940
    },
    {
      "epoch": 39.59,
      "learning_rate": 1.4721035598705502e-05,
      "loss": 0.0029,
      "step": 101950
    },
    {
      "epoch": 39.6,
      "learning_rate": 1.4720517799352752e-05,
      "loss": 0.1134,
      "step": 101960
    },
    {
      "epoch": 39.6,
      "learning_rate": 1.4720000000000001e-05,
      "loss": 0.08,
      "step": 101970
    },
    {
      "epoch": 39.6,
      "learning_rate": 1.4719482200647251e-05,
      "loss": 0.0321,
      "step": 101980
    },
    {
      "epoch": 39.61,
      "learning_rate": 1.47189644012945e-05,
      "loss": 0.1706,
      "step": 101990
    },
    {
      "epoch": 39.61,
      "learning_rate": 1.4718446601941749e-05,
      "loss": 0.0078,
      "step": 102000
    },
    {
      "epoch": 39.62,
      "learning_rate": 1.4717928802588998e-05,
      "loss": 0.0844,
      "step": 102010
    },
    {
      "epoch": 39.62,
      "learning_rate": 1.4717411003236246e-05,
      "loss": 0.2054,
      "step": 102020
    },
    {
      "epoch": 39.62,
      "learning_rate": 1.4716893203883496e-05,
      "loss": 0.0832,
      "step": 102030
    },
    {
      "epoch": 39.63,
      "learning_rate": 1.4716375404530745e-05,
      "loss": 0.1389,
      "step": 102040
    },
    {
      "epoch": 39.63,
      "learning_rate": 1.4715857605177995e-05,
      "loss": 0.0742,
      "step": 102050
    },
    {
      "epoch": 39.63,
      "learning_rate": 1.4715339805825245e-05,
      "loss": 0.0016,
      "step": 102060
    },
    {
      "epoch": 39.64,
      "learning_rate": 1.4714822006472494e-05,
      "loss": 0.1111,
      "step": 102070
    },
    {
      "epoch": 39.64,
      "learning_rate": 1.4714304207119742e-05,
      "loss": 0.0486,
      "step": 102080
    },
    {
      "epoch": 39.65,
      "learning_rate": 1.4713786407766992e-05,
      "loss": 0.065,
      "step": 102090
    },
    {
      "epoch": 39.65,
      "learning_rate": 1.4713268608414241e-05,
      "loss": 0.0021,
      "step": 102100
    },
    {
      "epoch": 39.65,
      "learning_rate": 1.471275080906149e-05,
      "loss": 0.0348,
      "step": 102110
    },
    {
      "epoch": 39.66,
      "learning_rate": 1.4712233009708739e-05,
      "loss": 0.1241,
      "step": 102120
    },
    {
      "epoch": 39.66,
      "learning_rate": 1.4711715210355988e-05,
      "loss": 0.1321,
      "step": 102130
    },
    {
      "epoch": 39.67,
      "learning_rate": 1.4711197411003238e-05,
      "loss": 0.1762,
      "step": 102140
    },
    {
      "epoch": 39.67,
      "learning_rate": 1.4710679611650486e-05,
      "loss": 0.2518,
      "step": 102150
    },
    {
      "epoch": 39.67,
      "learning_rate": 1.4710161812297736e-05,
      "loss": 0.0824,
      "step": 102160
    },
    {
      "epoch": 39.68,
      "learning_rate": 1.4709644012944985e-05,
      "loss": 0.0206,
      "step": 102170
    },
    {
      "epoch": 39.68,
      "learning_rate": 1.4709126213592235e-05,
      "loss": 0.1726,
      "step": 102180
    },
    {
      "epoch": 39.69,
      "learning_rate": 1.4708608414239483e-05,
      "loss": 0.0871,
      "step": 102190
    },
    {
      "epoch": 39.69,
      "learning_rate": 1.4708090614886732e-05,
      "loss": 0.0288,
      "step": 102200
    },
    {
      "epoch": 39.69,
      "learning_rate": 1.4707572815533982e-05,
      "loss": 0.05,
      "step": 102210
    },
    {
      "epoch": 39.7,
      "learning_rate": 1.4707055016181232e-05,
      "loss": 0.0791,
      "step": 102220
    },
    {
      "epoch": 39.7,
      "learning_rate": 1.470653721682848e-05,
      "loss": 0.1369,
      "step": 102230
    },
    {
      "epoch": 39.7,
      "learning_rate": 1.470601941747573e-05,
      "loss": 0.0501,
      "step": 102240
    },
    {
      "epoch": 39.71,
      "learning_rate": 1.4705501618122979e-05,
      "loss": 0.0075,
      "step": 102250
    },
    {
      "epoch": 39.71,
      "learning_rate": 1.4704983818770228e-05,
      "loss": 0.0263,
      "step": 102260
    },
    {
      "epoch": 39.72,
      "learning_rate": 1.4704466019417476e-05,
      "loss": 0.1974,
      "step": 102270
    },
    {
      "epoch": 39.72,
      "learning_rate": 1.4703948220064726e-05,
      "loss": 0.0083,
      "step": 102280
    },
    {
      "epoch": 39.72,
      "learning_rate": 1.4703430420711976e-05,
      "loss": 0.1484,
      "step": 102290
    },
    {
      "epoch": 39.73,
      "learning_rate": 1.4702912621359225e-05,
      "loss": 0.0286,
      "step": 102300
    },
    {
      "epoch": 39.73,
      "learning_rate": 1.4702394822006473e-05,
      "loss": 0.1325,
      "step": 102310
    },
    {
      "epoch": 39.74,
      "learning_rate": 1.4701877022653723e-05,
      "loss": 0.0855,
      "step": 102320
    },
    {
      "epoch": 39.74,
      "learning_rate": 1.4701359223300972e-05,
      "loss": 0.1292,
      "step": 102330
    },
    {
      "epoch": 39.74,
      "learning_rate": 1.4700841423948222e-05,
      "loss": 0.201,
      "step": 102340
    },
    {
      "epoch": 39.75,
      "learning_rate": 1.470032362459547e-05,
      "loss": 0.021,
      "step": 102350
    },
    {
      "epoch": 39.75,
      "learning_rate": 1.469980582524272e-05,
      "loss": 0.1761,
      "step": 102360
    },
    {
      "epoch": 39.76,
      "learning_rate": 1.4699288025889969e-05,
      "loss": 0.0509,
      "step": 102370
    },
    {
      "epoch": 39.76,
      "learning_rate": 1.4698770226537217e-05,
      "loss": 0.1158,
      "step": 102380
    },
    {
      "epoch": 39.76,
      "learning_rate": 1.4698252427184467e-05,
      "loss": 0.082,
      "step": 102390
    },
    {
      "epoch": 39.77,
      "learning_rate": 1.4697734627831716e-05,
      "loss": 0.1641,
      "step": 102400
    },
    {
      "epoch": 39.77,
      "learning_rate": 1.4697216828478966e-05,
      "loss": 0.0757,
      "step": 102410
    },
    {
      "epoch": 39.77,
      "learning_rate": 1.4696699029126216e-05,
      "loss": 0.0005,
      "step": 102420
    },
    {
      "epoch": 39.78,
      "learning_rate": 1.4696181229773463e-05,
      "loss": 0.0969,
      "step": 102430
    },
    {
      "epoch": 39.78,
      "learning_rate": 1.4695663430420713e-05,
      "loss": 0.0472,
      "step": 102440
    },
    {
      "epoch": 39.79,
      "learning_rate": 1.4695145631067963e-05,
      "loss": 0.1683,
      "step": 102450
    },
    {
      "epoch": 39.79,
      "learning_rate": 1.469462783171521e-05,
      "loss": 0.1246,
      "step": 102460
    },
    {
      "epoch": 39.79,
      "learning_rate": 1.469411003236246e-05,
      "loss": 0.0176,
      "step": 102470
    },
    {
      "epoch": 39.8,
      "learning_rate": 1.469359223300971e-05,
      "loss": 0.0499,
      "step": 102480
    },
    {
      "epoch": 39.8,
      "learning_rate": 1.469307443365696e-05,
      "loss": 0.0536,
      "step": 102490
    },
    {
      "epoch": 39.81,
      "learning_rate": 1.4692556634304209e-05,
      "loss": 0.1382,
      "step": 102500
    },
    {
      "epoch": 39.81,
      "learning_rate": 1.4692038834951457e-05,
      "loss": 0.0762,
      "step": 102510
    },
    {
      "epoch": 39.81,
      "learning_rate": 1.4691521035598707e-05,
      "loss": 0.0374,
      "step": 102520
    },
    {
      "epoch": 39.82,
      "learning_rate": 1.4691003236245955e-05,
      "loss": 0.0447,
      "step": 102530
    },
    {
      "epoch": 39.82,
      "learning_rate": 1.4690485436893204e-05,
      "loss": 0.0927,
      "step": 102540
    },
    {
      "epoch": 39.83,
      "learning_rate": 1.4689967637540454e-05,
      "loss": 0.2104,
      "step": 102550
    },
    {
      "epoch": 39.83,
      "learning_rate": 1.4689449838187703e-05,
      "loss": 0.0374,
      "step": 102560
    },
    {
      "epoch": 39.83,
      "learning_rate": 1.4688932038834953e-05,
      "loss": 0.104,
      "step": 102570
    },
    {
      "epoch": 39.84,
      "learning_rate": 1.4688414239482203e-05,
      "loss": 0.0678,
      "step": 102580
    },
    {
      "epoch": 39.84,
      "learning_rate": 1.4687896440129452e-05,
      "loss": 0.2001,
      "step": 102590
    },
    {
      "epoch": 39.84,
      "learning_rate": 1.46873786407767e-05,
      "loss": 0.015,
      "step": 102600
    },
    {
      "epoch": 39.85,
      "learning_rate": 1.4686860841423948e-05,
      "loss": 0.0489,
      "step": 102610
    },
    {
      "epoch": 39.85,
      "learning_rate": 1.4686343042071198e-05,
      "loss": 0.0322,
      "step": 102620
    },
    {
      "epoch": 39.86,
      "learning_rate": 1.4685825242718447e-05,
      "loss": 0.0614,
      "step": 102630
    },
    {
      "epoch": 39.86,
      "learning_rate": 1.4685307443365697e-05,
      "loss": 0.0028,
      "step": 102640
    },
    {
      "epoch": 39.86,
      "learning_rate": 1.4684789644012947e-05,
      "loss": 0.093,
      "step": 102650
    },
    {
      "epoch": 39.87,
      "learning_rate": 1.4684271844660196e-05,
      "loss": 0.1306,
      "step": 102660
    },
    {
      "epoch": 39.87,
      "learning_rate": 1.4683754045307446e-05,
      "loss": 0.0899,
      "step": 102670
    },
    {
      "epoch": 39.88,
      "learning_rate": 1.4683236245954692e-05,
      "loss": 0.047,
      "step": 102680
    },
    {
      "epoch": 39.88,
      "learning_rate": 1.4682718446601942e-05,
      "loss": 0.0971,
      "step": 102690
    },
    {
      "epoch": 39.88,
      "learning_rate": 1.4682200647249191e-05,
      "loss": 0.0034,
      "step": 102700
    },
    {
      "epoch": 39.89,
      "learning_rate": 1.4681682847896441e-05,
      "loss": 0.128,
      "step": 102710
    },
    {
      "epoch": 39.89,
      "learning_rate": 1.468116504854369e-05,
      "loss": 0.1238,
      "step": 102720
    },
    {
      "epoch": 39.9,
      "learning_rate": 1.468064724919094e-05,
      "loss": 0.1027,
      "step": 102730
    },
    {
      "epoch": 39.9,
      "learning_rate": 1.468012944983819e-05,
      "loss": 0.0816,
      "step": 102740
    },
    {
      "epoch": 39.9,
      "learning_rate": 1.467961165048544e-05,
      "loss": 0.0714,
      "step": 102750
    },
    {
      "epoch": 39.91,
      "learning_rate": 1.4679093851132686e-05,
      "loss": 0.0042,
      "step": 102760
    },
    {
      "epoch": 39.91,
      "learning_rate": 1.4678576051779935e-05,
      "loss": 0.0157,
      "step": 102770
    },
    {
      "epoch": 39.91,
      "learning_rate": 1.4678058252427185e-05,
      "loss": 0.3695,
      "step": 102780
    },
    {
      "epoch": 39.92,
      "learning_rate": 1.4677540453074434e-05,
      "loss": 0.0301,
      "step": 102790
    },
    {
      "epoch": 39.92,
      "learning_rate": 1.4677022653721684e-05,
      "loss": 0.1392,
      "step": 102800
    },
    {
      "epoch": 39.93,
      "learning_rate": 1.4676504854368934e-05,
      "loss": 0.134,
      "step": 102810
    },
    {
      "epoch": 39.93,
      "learning_rate": 1.4675987055016183e-05,
      "loss": 0.0711,
      "step": 102820
    },
    {
      "epoch": 39.93,
      "learning_rate": 1.4675469255663433e-05,
      "loss": 0.1224,
      "step": 102830
    },
    {
      "epoch": 39.94,
      "learning_rate": 1.467495145631068e-05,
      "loss": 0.0272,
      "step": 102840
    },
    {
      "epoch": 39.94,
      "learning_rate": 1.4674433656957929e-05,
      "loss": 0.0989,
      "step": 102850
    },
    {
      "epoch": 39.95,
      "learning_rate": 1.4673915857605178e-05,
      "loss": 0.156,
      "step": 102860
    },
    {
      "epoch": 39.95,
      "learning_rate": 1.4673398058252428e-05,
      "loss": 0.279,
      "step": 102870
    },
    {
      "epoch": 39.95,
      "learning_rate": 1.4672880258899678e-05,
      "loss": 0.0568,
      "step": 102880
    },
    {
      "epoch": 39.96,
      "learning_rate": 1.4672362459546927e-05,
      "loss": 0.087,
      "step": 102890
    },
    {
      "epoch": 39.96,
      "learning_rate": 1.4671844660194177e-05,
      "loss": 0.0757,
      "step": 102900
    },
    {
      "epoch": 39.97,
      "learning_rate": 1.4671326860841426e-05,
      "loss": 0.0706,
      "step": 102910
    },
    {
      "epoch": 39.97,
      "learning_rate": 1.4670809061488673e-05,
      "loss": 0.3136,
      "step": 102920
    },
    {
      "epoch": 39.97,
      "learning_rate": 1.4670291262135922e-05,
      "loss": 0.1318,
      "step": 102930
    },
    {
      "epoch": 39.98,
      "learning_rate": 1.4669773462783172e-05,
      "loss": 0.0733,
      "step": 102940
    },
    {
      "epoch": 39.98,
      "learning_rate": 1.4669255663430422e-05,
      "loss": 0.1048,
      "step": 102950
    },
    {
      "epoch": 39.98,
      "learning_rate": 1.4668737864077671e-05,
      "loss": 0.128,
      "step": 102960
    },
    {
      "epoch": 39.99,
      "learning_rate": 1.466822006472492e-05,
      "loss": 0.1298,
      "step": 102970
    },
    {
      "epoch": 39.99,
      "learning_rate": 1.466770226537217e-05,
      "loss": 0.122,
      "step": 102980
    },
    {
      "epoch": 40.0,
      "learning_rate": 1.466718446601942e-05,
      "loss": 0.0043,
      "step": 102990
    },
    {
      "epoch": 40.0,
      "learning_rate": 1.4666666666666666e-05,
      "loss": 0.1052,
      "step": 103000
    },
    {
      "epoch": 40.0,
      "eval_accuracy": 0.9493810178817056,
      "eval_loss": 0.32120659947395325,
      "eval_runtime": 8.2083,
      "eval_samples_per_second": 442.844,
      "eval_steps_per_second": 55.432,
      "step": 103000
    },
    {
      "epoch": 40.0,
      "learning_rate": 1.4666148867313916e-05,
      "loss": 0.1314,
      "step": 103010
    },
    {
      "epoch": 40.01,
      "learning_rate": 1.4665631067961166e-05,
      "loss": 0.2195,
      "step": 103020
    },
    {
      "epoch": 40.01,
      "learning_rate": 1.4665113268608415e-05,
      "loss": 0.0419,
      "step": 103030
    },
    {
      "epoch": 40.02,
      "learning_rate": 1.4664595469255665e-05,
      "loss": 0.2252,
      "step": 103040
    },
    {
      "epoch": 40.02,
      "learning_rate": 1.4664077669902914e-05,
      "loss": 0.045,
      "step": 103050
    },
    {
      "epoch": 40.02,
      "learning_rate": 1.4663559870550164e-05,
      "loss": 0.2477,
      "step": 103060
    },
    {
      "epoch": 40.03,
      "learning_rate": 1.4663042071197414e-05,
      "loss": 0.0502,
      "step": 103070
    },
    {
      "epoch": 40.03,
      "learning_rate": 1.466252427184466e-05,
      "loss": 0.1252,
      "step": 103080
    },
    {
      "epoch": 40.03,
      "learning_rate": 1.466200647249191e-05,
      "loss": 0.0679,
      "step": 103090
    },
    {
      "epoch": 40.04,
      "learning_rate": 1.4661488673139159e-05,
      "loss": 0.013,
      "step": 103100
    },
    {
      "epoch": 40.04,
      "learning_rate": 1.4660970873786409e-05,
      "loss": 0.1509,
      "step": 103110
    },
    {
      "epoch": 40.05,
      "learning_rate": 1.4660453074433658e-05,
      "loss": 0.0251,
      "step": 103120
    },
    {
      "epoch": 40.05,
      "learning_rate": 1.4659935275080908e-05,
      "loss": 0.1605,
      "step": 103130
    },
    {
      "epoch": 40.05,
      "learning_rate": 1.4659417475728158e-05,
      "loss": 0.0702,
      "step": 103140
    },
    {
      "epoch": 40.06,
      "learning_rate": 1.4658899676375407e-05,
      "loss": 0.1271,
      "step": 103150
    },
    {
      "epoch": 40.06,
      "learning_rate": 1.4658381877022657e-05,
      "loss": 0.0264,
      "step": 103160
    },
    {
      "epoch": 40.07,
      "learning_rate": 1.4657864077669903e-05,
      "loss": 0.0189,
      "step": 103170
    },
    {
      "epoch": 40.07,
      "learning_rate": 1.4657346278317153e-05,
      "loss": 0.1867,
      "step": 103180
    },
    {
      "epoch": 40.07,
      "learning_rate": 1.4656828478964402e-05,
      "loss": 0.1789,
      "step": 103190
    },
    {
      "epoch": 40.08,
      "learning_rate": 1.4656310679611652e-05,
      "loss": 0.1584,
      "step": 103200
    },
    {
      "epoch": 40.08,
      "learning_rate": 1.4655792880258901e-05,
      "loss": 0.0564,
      "step": 103210
    },
    {
      "epoch": 40.09,
      "learning_rate": 1.4655275080906151e-05,
      "loss": 0.1026,
      "step": 103220
    },
    {
      "epoch": 40.09,
      "learning_rate": 1.46547572815534e-05,
      "loss": 0.0481,
      "step": 103230
    },
    {
      "epoch": 40.09,
      "learning_rate": 1.465423948220065e-05,
      "loss": 0.0857,
      "step": 103240
    },
    {
      "epoch": 40.1,
      "learning_rate": 1.4653721682847897e-05,
      "loss": 0.0057,
      "step": 103250
    },
    {
      "epoch": 40.1,
      "learning_rate": 1.4653203883495146e-05,
      "loss": 0.1944,
      "step": 103260
    },
    {
      "epoch": 40.1,
      "learning_rate": 1.4652686084142396e-05,
      "loss": 0.0327,
      "step": 103270
    },
    {
      "epoch": 40.11,
      "learning_rate": 1.4652168284789645e-05,
      "loss": 0.0838,
      "step": 103280
    },
    {
      "epoch": 40.11,
      "learning_rate": 1.4651650485436895e-05,
      "loss": 0.0295,
      "step": 103290
    },
    {
      "epoch": 40.12,
      "learning_rate": 1.4651132686084145e-05,
      "loss": 0.0373,
      "step": 103300
    },
    {
      "epoch": 40.12,
      "learning_rate": 1.4650614886731394e-05,
      "loss": 0.1162,
      "step": 103310
    },
    {
      "epoch": 40.12,
      "learning_rate": 1.4650097087378642e-05,
      "loss": 0.0627,
      "step": 103320
    },
    {
      "epoch": 40.13,
      "learning_rate": 1.464957928802589e-05,
      "loss": 0.0298,
      "step": 103330
    },
    {
      "epoch": 40.13,
      "learning_rate": 1.464906148867314e-05,
      "loss": 0.0399,
      "step": 103340
    },
    {
      "epoch": 40.14,
      "learning_rate": 1.464854368932039e-05,
      "loss": 0.0534,
      "step": 103350
    },
    {
      "epoch": 40.14,
      "learning_rate": 1.4648025889967639e-05,
      "loss": 0.151,
      "step": 103360
    },
    {
      "epoch": 40.14,
      "learning_rate": 1.4647508090614889e-05,
      "loss": 0.1134,
      "step": 103370
    },
    {
      "epoch": 40.15,
      "learning_rate": 1.4646990291262138e-05,
      "loss": 0.0912,
      "step": 103380
    },
    {
      "epoch": 40.15,
      "learning_rate": 1.4646472491909388e-05,
      "loss": 0.051,
      "step": 103390
    },
    {
      "epoch": 40.16,
      "learning_rate": 1.4645954692556636e-05,
      "loss": 0.0653,
      "step": 103400
    },
    {
      "epoch": 40.16,
      "learning_rate": 1.4645436893203884e-05,
      "loss": 0.113,
      "step": 103410
    },
    {
      "epoch": 40.16,
      "learning_rate": 1.4644919093851133e-05,
      "loss": 0.0629,
      "step": 103420
    },
    {
      "epoch": 40.17,
      "learning_rate": 1.4644401294498383e-05,
      "loss": 0.0813,
      "step": 103430
    },
    {
      "epoch": 40.17,
      "learning_rate": 1.4643883495145633e-05,
      "loss": 0.0229,
      "step": 103440
    },
    {
      "epoch": 40.17,
      "learning_rate": 1.4643365695792882e-05,
      "loss": 0.0783,
      "step": 103450
    },
    {
      "epoch": 40.18,
      "learning_rate": 1.4642847896440132e-05,
      "loss": 0.1043,
      "step": 103460
    },
    {
      "epoch": 40.18,
      "learning_rate": 1.464233009708738e-05,
      "loss": 0.0723,
      "step": 103470
    },
    {
      "epoch": 40.19,
      "learning_rate": 1.464181229773463e-05,
      "loss": 0.1835,
      "step": 103480
    },
    {
      "epoch": 40.19,
      "learning_rate": 1.4641294498381877e-05,
      "loss": 0.0827,
      "step": 103490
    },
    {
      "epoch": 40.19,
      "learning_rate": 1.4640776699029127e-05,
      "loss": 0.1413,
      "step": 103500
    },
    {
      "epoch": 40.2,
      "learning_rate": 1.4640258899676376e-05,
      "loss": 0.2662,
      "step": 103510
    },
    {
      "epoch": 40.2,
      "learning_rate": 1.4639741100323626e-05,
      "loss": 0.0384,
      "step": 103520
    },
    {
      "epoch": 40.21,
      "learning_rate": 1.4639223300970876e-05,
      "loss": 0.1289,
      "step": 103530
    },
    {
      "epoch": 40.21,
      "learning_rate": 1.4638705501618125e-05,
      "loss": 0.1093,
      "step": 103540
    },
    {
      "epoch": 40.21,
      "learning_rate": 1.4638187702265373e-05,
      "loss": 0.0874,
      "step": 103550
    },
    {
      "epoch": 40.22,
      "learning_rate": 1.4637669902912623e-05,
      "loss": 0.2186,
      "step": 103560
    },
    {
      "epoch": 40.22,
      "learning_rate": 1.463715210355987e-05,
      "loss": 0.072,
      "step": 103570
    },
    {
      "epoch": 40.23,
      "learning_rate": 1.463663430420712e-05,
      "loss": 0.0625,
      "step": 103580
    },
    {
      "epoch": 40.23,
      "learning_rate": 1.463611650485437e-05,
      "loss": 0.0409,
      "step": 103590
    },
    {
      "epoch": 40.23,
      "learning_rate": 1.463559870550162e-05,
      "loss": 0.0795,
      "step": 103600
    },
    {
      "epoch": 40.24,
      "learning_rate": 1.463508090614887e-05,
      "loss": 0.095,
      "step": 103610
    },
    {
      "epoch": 40.24,
      "learning_rate": 1.4634563106796117e-05,
      "loss": 0.1445,
      "step": 103620
    },
    {
      "epoch": 40.24,
      "learning_rate": 1.4634045307443367e-05,
      "loss": 0.1215,
      "step": 103630
    },
    {
      "epoch": 40.25,
      "learning_rate": 1.4633527508090616e-05,
      "loss": 0.0006,
      "step": 103640
    },
    {
      "epoch": 40.25,
      "learning_rate": 1.4633009708737864e-05,
      "loss": 0.1537,
      "step": 103650
    },
    {
      "epoch": 40.26,
      "learning_rate": 1.4632491909385114e-05,
      "loss": 0.0679,
      "step": 103660
    },
    {
      "epoch": 40.26,
      "learning_rate": 1.4631974110032364e-05,
      "loss": 0.0255,
      "step": 103670
    },
    {
      "epoch": 40.26,
      "learning_rate": 1.4631456310679613e-05,
      "loss": 0.0167,
      "step": 103680
    },
    {
      "epoch": 40.27,
      "learning_rate": 1.4630938511326863e-05,
      "loss": 0.0367,
      "step": 103690
    },
    {
      "epoch": 40.27,
      "learning_rate": 1.463042071197411e-05,
      "loss": 0.0788,
      "step": 103700
    },
    {
      "epoch": 40.28,
      "learning_rate": 1.462990291262136e-05,
      "loss": 0.0308,
      "step": 103710
    },
    {
      "epoch": 40.28,
      "learning_rate": 1.462938511326861e-05,
      "loss": 0.1272,
      "step": 103720
    },
    {
      "epoch": 40.28,
      "learning_rate": 1.462886731391586e-05,
      "loss": 0.1584,
      "step": 103730
    },
    {
      "epoch": 40.29,
      "learning_rate": 1.4628349514563108e-05,
      "loss": 0.2329,
      "step": 103740
    },
    {
      "epoch": 40.29,
      "learning_rate": 1.4627831715210357e-05,
      "loss": 0.2744,
      "step": 103750
    },
    {
      "epoch": 40.3,
      "learning_rate": 1.4627313915857607e-05,
      "loss": 0.1069,
      "step": 103760
    },
    {
      "epoch": 40.3,
      "learning_rate": 1.4626796116504856e-05,
      "loss": 0.1002,
      "step": 103770
    },
    {
      "epoch": 40.3,
      "learning_rate": 1.4626278317152104e-05,
      "loss": 0.1557,
      "step": 103780
    },
    {
      "epoch": 40.31,
      "learning_rate": 1.4625760517799354e-05,
      "loss": 0.134,
      "step": 103790
    },
    {
      "epoch": 40.31,
      "learning_rate": 1.4625242718446604e-05,
      "loss": 0.11,
      "step": 103800
    },
    {
      "epoch": 40.31,
      "learning_rate": 1.4624724919093853e-05,
      "loss": 0.1116,
      "step": 103810
    },
    {
      "epoch": 40.32,
      "learning_rate": 1.4624207119741101e-05,
      "loss": 0.0648,
      "step": 103820
    },
    {
      "epoch": 40.32,
      "learning_rate": 1.462368932038835e-05,
      "loss": 0.0032,
      "step": 103830
    },
    {
      "epoch": 40.33,
      "learning_rate": 1.46231715210356e-05,
      "loss": 0.0287,
      "step": 103840
    },
    {
      "epoch": 40.33,
      "learning_rate": 1.4622653721682848e-05,
      "loss": 0.0797,
      "step": 103850
    },
    {
      "epoch": 40.33,
      "learning_rate": 1.4622135922330098e-05,
      "loss": 0.0343,
      "step": 103860
    },
    {
      "epoch": 40.34,
      "learning_rate": 1.4621618122977347e-05,
      "loss": 0.1381,
      "step": 103870
    },
    {
      "epoch": 40.34,
      "learning_rate": 1.4621100323624597e-05,
      "loss": 0.0485,
      "step": 103880
    },
    {
      "epoch": 40.35,
      "learning_rate": 1.4620582524271847e-05,
      "loss": 0.1236,
      "step": 103890
    },
    {
      "epoch": 40.35,
      "learning_rate": 1.4620064724919095e-05,
      "loss": 0.0305,
      "step": 103900
    },
    {
      "epoch": 40.35,
      "learning_rate": 1.4619546925566344e-05,
      "loss": 0.0093,
      "step": 103910
    },
    {
      "epoch": 40.36,
      "learning_rate": 1.4619029126213594e-05,
      "loss": 0.0299,
      "step": 103920
    },
    {
      "epoch": 40.36,
      "learning_rate": 1.4618511326860842e-05,
      "loss": 0.0693,
      "step": 103930
    },
    {
      "epoch": 40.37,
      "learning_rate": 1.4617993527508091e-05,
      "loss": 0.0322,
      "step": 103940
    },
    {
      "epoch": 40.37,
      "learning_rate": 1.4617475728155341e-05,
      "loss": 0.0047,
      "step": 103950
    },
    {
      "epoch": 40.37,
      "learning_rate": 1.461695792880259e-05,
      "loss": 0.034,
      "step": 103960
    },
    {
      "epoch": 40.38,
      "learning_rate": 1.461644012944984e-05,
      "loss": 0.0613,
      "step": 103970
    },
    {
      "epoch": 40.38,
      "learning_rate": 1.4615922330097088e-05,
      "loss": 0.0148,
      "step": 103980
    },
    {
      "epoch": 40.38,
      "learning_rate": 1.4615404530744338e-05,
      "loss": 0.0945,
      "step": 103990
    },
    {
      "epoch": 40.39,
      "learning_rate": 1.4614886731391586e-05,
      "loss": 0.0891,
      "step": 104000
    },
    {
      "epoch": 40.39,
      "learning_rate": 1.4614368932038835e-05,
      "loss": 0.0344,
      "step": 104010
    },
    {
      "epoch": 40.4,
      "learning_rate": 1.4613851132686085e-05,
      "loss": 0.1106,
      "step": 104020
    },
    {
      "epoch": 40.4,
      "learning_rate": 1.4613333333333335e-05,
      "loss": 0.0146,
      "step": 104030
    },
    {
      "epoch": 40.4,
      "learning_rate": 1.4612815533980584e-05,
      "loss": 0.1867,
      "step": 104040
    },
    {
      "epoch": 40.41,
      "learning_rate": 1.4612297734627834e-05,
      "loss": 0.0843,
      "step": 104050
    },
    {
      "epoch": 40.41,
      "learning_rate": 1.4611779935275082e-05,
      "loss": 0.1028,
      "step": 104060
    },
    {
      "epoch": 40.42,
      "learning_rate": 1.4611262135922331e-05,
      "loss": 0.2123,
      "step": 104070
    },
    {
      "epoch": 40.42,
      "learning_rate": 1.461074433656958e-05,
      "loss": 0.0612,
      "step": 104080
    },
    {
      "epoch": 40.42,
      "learning_rate": 1.4610226537216829e-05,
      "loss": 0.1338,
      "step": 104090
    },
    {
      "epoch": 40.43,
      "learning_rate": 1.4609708737864079e-05,
      "loss": 0.0421,
      "step": 104100
    },
    {
      "epoch": 40.43,
      "learning_rate": 1.4609190938511328e-05,
      "loss": 0.1733,
      "step": 104110
    },
    {
      "epoch": 40.43,
      "learning_rate": 1.4608673139158578e-05,
      "loss": 0.1029,
      "step": 104120
    },
    {
      "epoch": 40.44,
      "learning_rate": 1.4608155339805827e-05,
      "loss": 0.0488,
      "step": 104130
    },
    {
      "epoch": 40.44,
      "learning_rate": 1.4607637540453075e-05,
      "loss": 0.0673,
      "step": 104140
    },
    {
      "epoch": 40.45,
      "learning_rate": 1.4607119741100323e-05,
      "loss": 0.0781,
      "step": 104150
    },
    {
      "epoch": 40.45,
      "learning_rate": 1.4606601941747573e-05,
      "loss": 0.0287,
      "step": 104160
    },
    {
      "epoch": 40.45,
      "learning_rate": 1.4606084142394822e-05,
      "loss": 0.0422,
      "step": 104170
    },
    {
      "epoch": 40.46,
      "learning_rate": 1.4605566343042072e-05,
      "loss": 0.0202,
      "step": 104180
    },
    {
      "epoch": 40.46,
      "learning_rate": 1.4605048543689322e-05,
      "loss": 0.0862,
      "step": 104190
    },
    {
      "epoch": 40.47,
      "learning_rate": 1.4604530744336571e-05,
      "loss": 0.0402,
      "step": 104200
    },
    {
      "epoch": 40.47,
      "learning_rate": 1.4604012944983821e-05,
      "loss": 0.1079,
      "step": 104210
    },
    {
      "epoch": 40.47,
      "learning_rate": 1.4603495145631069e-05,
      "loss": 0.036,
      "step": 104220
    },
    {
      "epoch": 40.48,
      "learning_rate": 1.4602977346278317e-05,
      "loss": 0.0708,
      "step": 104230
    },
    {
      "epoch": 40.48,
      "learning_rate": 1.4602459546925566e-05,
      "loss": 0.3168,
      "step": 104240
    },
    {
      "epoch": 40.49,
      "learning_rate": 1.4601941747572816e-05,
      "loss": 0.0304,
      "step": 104250
    },
    {
      "epoch": 40.49,
      "learning_rate": 1.4601423948220066e-05,
      "loss": 0.0778,
      "step": 104260
    },
    {
      "epoch": 40.49,
      "learning_rate": 1.4600906148867315e-05,
      "loss": 0.0726,
      "step": 104270
    },
    {
      "epoch": 40.5,
      "learning_rate": 1.4600388349514565e-05,
      "loss": 0.0122,
      "step": 104280
    },
    {
      "epoch": 40.5,
      "learning_rate": 1.4599870550161814e-05,
      "loss": 0.0343,
      "step": 104290
    },
    {
      "epoch": 40.5,
      "learning_rate": 1.4599352750809064e-05,
      "loss": 0.0184,
      "step": 104300
    },
    {
      "epoch": 40.51,
      "learning_rate": 1.459883495145631e-05,
      "loss": 0.1732,
      "step": 104310
    },
    {
      "epoch": 40.51,
      "learning_rate": 1.459831715210356e-05,
      "loss": 0.0248,
      "step": 104320
    },
    {
      "epoch": 40.52,
      "learning_rate": 1.459779935275081e-05,
      "loss": 0.0959,
      "step": 104330
    },
    {
      "epoch": 40.52,
      "learning_rate": 1.459728155339806e-05,
      "loss": 0.2013,
      "step": 104340
    },
    {
      "epoch": 40.52,
      "learning_rate": 1.4596763754045309e-05,
      "loss": 0.1094,
      "step": 104350
    },
    {
      "epoch": 40.53,
      "learning_rate": 1.4596245954692558e-05,
      "loss": 0.0455,
      "step": 104360
    },
    {
      "epoch": 40.53,
      "learning_rate": 1.4595728155339808e-05,
      "loss": 0.1691,
      "step": 104370
    },
    {
      "epoch": 40.54,
      "learning_rate": 1.4595210355987058e-05,
      "loss": 0.1876,
      "step": 104380
    },
    {
      "epoch": 40.54,
      "learning_rate": 1.4594692556634304e-05,
      "loss": 0.0784,
      "step": 104390
    },
    {
      "epoch": 40.54,
      "learning_rate": 1.4594174757281554e-05,
      "loss": 0.0646,
      "step": 104400
    },
    {
      "epoch": 40.55,
      "learning_rate": 1.4593656957928803e-05,
      "loss": 0.1644,
      "step": 104410
    },
    {
      "epoch": 40.55,
      "learning_rate": 1.4593139158576053e-05,
      "loss": 0.0688,
      "step": 104420
    },
    {
      "epoch": 40.56,
      "learning_rate": 1.4592621359223302e-05,
      "loss": 0.1552,
      "step": 104430
    },
    {
      "epoch": 40.56,
      "learning_rate": 1.4592103559870552e-05,
      "loss": 0.0885,
      "step": 104440
    },
    {
      "epoch": 40.56,
      "learning_rate": 1.4591585760517802e-05,
      "loss": 0.0453,
      "step": 104450
    },
    {
      "epoch": 40.57,
      "learning_rate": 1.4591067961165051e-05,
      "loss": 0.0986,
      "step": 104460
    },
    {
      "epoch": 40.57,
      "learning_rate": 1.4590550161812297e-05,
      "loss": 0.013,
      "step": 104470
    },
    {
      "epoch": 40.57,
      "learning_rate": 1.4590032362459547e-05,
      "loss": 0.0926,
      "step": 104480
    },
    {
      "epoch": 40.58,
      "learning_rate": 1.4589514563106797e-05,
      "loss": 0.1861,
      "step": 104490
    },
    {
      "epoch": 40.58,
      "learning_rate": 1.4588996763754046e-05,
      "loss": 0.1103,
      "step": 104500
    },
    {
      "epoch": 40.59,
      "learning_rate": 1.4588478964401296e-05,
      "loss": 0.0343,
      "step": 104510
    },
    {
      "epoch": 40.59,
      "learning_rate": 1.4587961165048546e-05,
      "loss": 0.1347,
      "step": 104520
    },
    {
      "epoch": 40.59,
      "learning_rate": 1.4587443365695795e-05,
      "loss": 0.0597,
      "step": 104530
    },
    {
      "epoch": 40.6,
      "learning_rate": 1.4586925566343045e-05,
      "loss": 0.2743,
      "step": 104540
    },
    {
      "epoch": 40.6,
      "learning_rate": 1.4586407766990291e-05,
      "loss": 0.2749,
      "step": 104550
    },
    {
      "epoch": 40.61,
      "learning_rate": 1.458588996763754e-05,
      "loss": 0.0921,
      "step": 104560
    },
    {
      "epoch": 40.61,
      "learning_rate": 1.458537216828479e-05,
      "loss": 0.0578,
      "step": 104570
    },
    {
      "epoch": 40.61,
      "learning_rate": 1.458485436893204e-05,
      "loss": 0.1066,
      "step": 104580
    },
    {
      "epoch": 40.62,
      "learning_rate": 1.458433656957929e-05,
      "loss": 0.1194,
      "step": 104590
    },
    {
      "epoch": 40.62,
      "learning_rate": 1.4583818770226539e-05,
      "loss": 0.0865,
      "step": 104600
    },
    {
      "epoch": 40.63,
      "learning_rate": 1.4583300970873789e-05,
      "loss": 0.1293,
      "step": 104610
    },
    {
      "epoch": 40.63,
      "learning_rate": 1.4582783171521038e-05,
      "loss": 0.0796,
      "step": 104620
    },
    {
      "epoch": 40.63,
      "learning_rate": 1.4582265372168285e-05,
      "loss": 0.0669,
      "step": 104630
    },
    {
      "epoch": 40.64,
      "learning_rate": 1.4581747572815534e-05,
      "loss": 0.0631,
      "step": 104640
    },
    {
      "epoch": 40.64,
      "learning_rate": 1.4581229773462784e-05,
      "loss": 0.0617,
      "step": 104650
    },
    {
      "epoch": 40.64,
      "learning_rate": 1.4580711974110033e-05,
      "loss": 0.0073,
      "step": 104660
    },
    {
      "epoch": 40.65,
      "learning_rate": 1.4580194174757283e-05,
      "loss": 0.1032,
      "step": 104670
    },
    {
      "epoch": 40.65,
      "learning_rate": 1.4579676375404533e-05,
      "loss": 0.0525,
      "step": 104680
    },
    {
      "epoch": 40.66,
      "learning_rate": 1.4579158576051782e-05,
      "loss": 0.1193,
      "step": 104690
    },
    {
      "epoch": 40.66,
      "learning_rate": 1.4578640776699032e-05,
      "loss": 0.2036,
      "step": 104700
    },
    {
      "epoch": 40.66,
      "learning_rate": 1.4578122977346278e-05,
      "loss": 0.0594,
      "step": 104710
    },
    {
      "epoch": 40.67,
      "learning_rate": 1.4577605177993528e-05,
      "loss": 0.0562,
      "step": 104720
    },
    {
      "epoch": 40.67,
      "learning_rate": 1.4577087378640777e-05,
      "loss": 0.1597,
      "step": 104730
    },
    {
      "epoch": 40.68,
      "learning_rate": 1.4576569579288027e-05,
      "loss": 0.0518,
      "step": 104740
    },
    {
      "epoch": 40.68,
      "learning_rate": 1.4576051779935277e-05,
      "loss": 0.0932,
      "step": 104750
    },
    {
      "epoch": 40.68,
      "learning_rate": 1.4575533980582526e-05,
      "loss": 0.2233,
      "step": 104760
    },
    {
      "epoch": 40.69,
      "learning_rate": 1.4575016181229776e-05,
      "loss": 0.0042,
      "step": 104770
    },
    {
      "epoch": 40.69,
      "learning_rate": 1.4574498381877025e-05,
      "loss": 0.1803,
      "step": 104780
    },
    {
      "epoch": 40.7,
      "learning_rate": 1.4573980582524272e-05,
      "loss": 0.0053,
      "step": 104790
    },
    {
      "epoch": 40.7,
      "learning_rate": 1.4573462783171521e-05,
      "loss": 0.149,
      "step": 104800
    },
    {
      "epoch": 40.7,
      "learning_rate": 1.4572944983818771e-05,
      "loss": 0.0249,
      "step": 104810
    },
    {
      "epoch": 40.71,
      "learning_rate": 1.457242718446602e-05,
      "loss": 0.052,
      "step": 104820
    },
    {
      "epoch": 40.71,
      "learning_rate": 1.457190938511327e-05,
      "loss": 0.004,
      "step": 104830
    },
    {
      "epoch": 40.71,
      "learning_rate": 1.457139158576052e-05,
      "loss": 0.1521,
      "step": 104840
    },
    {
      "epoch": 40.72,
      "learning_rate": 1.457087378640777e-05,
      "loss": 0.0782,
      "step": 104850
    },
    {
      "epoch": 40.72,
      "learning_rate": 1.4570355987055019e-05,
      "loss": 0.01,
      "step": 104860
    },
    {
      "epoch": 40.73,
      "learning_rate": 1.4569838187702267e-05,
      "loss": 0.1625,
      "step": 104870
    },
    {
      "epoch": 40.73,
      "learning_rate": 1.4569320388349515e-05,
      "loss": 0.0531,
      "step": 104880
    },
    {
      "epoch": 40.73,
      "learning_rate": 1.4568802588996764e-05,
      "loss": 0.228,
      "step": 104890
    },
    {
      "epoch": 40.74,
      "learning_rate": 1.4568284789644014e-05,
      "loss": 0.0036,
      "step": 104900
    },
    {
      "epoch": 40.74,
      "learning_rate": 1.4567766990291264e-05,
      "loss": 0.1281,
      "step": 104910
    },
    {
      "epoch": 40.75,
      "learning_rate": 1.4567249190938513e-05,
      "loss": 0.1211,
      "step": 104920
    },
    {
      "epoch": 40.75,
      "learning_rate": 1.4566731391585763e-05,
      "loss": 0.0979,
      "step": 104930
    },
    {
      "epoch": 40.75,
      "learning_rate": 1.4566213592233011e-05,
      "loss": 0.165,
      "step": 104940
    },
    {
      "epoch": 40.76,
      "learning_rate": 1.456569579288026e-05,
      "loss": 0.0326,
      "step": 104950
    },
    {
      "epoch": 40.76,
      "learning_rate": 1.4565177993527508e-05,
      "loss": 0.1103,
      "step": 104960
    },
    {
      "epoch": 40.77,
      "learning_rate": 1.4564660194174758e-05,
      "loss": 0.0319,
      "step": 104970
    },
    {
      "epoch": 40.77,
      "learning_rate": 1.4564142394822008e-05,
      "loss": 0.1743,
      "step": 104980
    },
    {
      "epoch": 40.77,
      "learning_rate": 1.4563624595469257e-05,
      "loss": 0.1302,
      "step": 104990
    },
    {
      "epoch": 40.78,
      "learning_rate": 1.4563106796116507e-05,
      "loss": 0.09,
      "step": 105000
    },
    {
      "epoch": 40.78,
      "learning_rate": 1.4562588996763757e-05,
      "loss": 0.0623,
      "step": 105010
    },
    {
      "epoch": 40.78,
      "learning_rate": 1.4562071197411004e-05,
      "loss": 0.0156,
      "step": 105020
    },
    {
      "epoch": 40.79,
      "learning_rate": 1.4561553398058254e-05,
      "loss": 0.1069,
      "step": 105030
    },
    {
      "epoch": 40.79,
      "learning_rate": 1.4561035598705502e-05,
      "loss": 0.0407,
      "step": 105040
    },
    {
      "epoch": 40.8,
      "learning_rate": 1.4560517799352752e-05,
      "loss": 0.1419,
      "step": 105050
    },
    {
      "epoch": 40.8,
      "learning_rate": 1.4560000000000001e-05,
      "loss": 0.0479,
      "step": 105060
    },
    {
      "epoch": 40.8,
      "learning_rate": 1.455948220064725e-05,
      "loss": 0.1584,
      "step": 105070
    },
    {
      "epoch": 40.81,
      "learning_rate": 1.45589644012945e-05,
      "loss": 0.0117,
      "step": 105080
    },
    {
      "epoch": 40.81,
      "learning_rate": 1.4558446601941748e-05,
      "loss": 0.0621,
      "step": 105090
    },
    {
      "epoch": 40.82,
      "learning_rate": 1.4557928802588998e-05,
      "loss": 0.1259,
      "step": 105100
    },
    {
      "epoch": 40.82,
      "learning_rate": 1.4557411003236248e-05,
      "loss": 0.0587,
      "step": 105110
    },
    {
      "epoch": 40.82,
      "learning_rate": 1.4556893203883496e-05,
      "loss": 0.0933,
      "step": 105120
    },
    {
      "epoch": 40.83,
      "learning_rate": 1.4556375404530745e-05,
      "loss": 0.012,
      "step": 105130
    },
    {
      "epoch": 40.83,
      "learning_rate": 1.4555857605177995e-05,
      "loss": 0.0402,
      "step": 105140
    },
    {
      "epoch": 40.83,
      "learning_rate": 1.4555339805825244e-05,
      "loss": 0.137,
      "step": 105150
    },
    {
      "epoch": 40.84,
      "learning_rate": 1.4554822006472494e-05,
      "loss": 0.2,
      "step": 105160
    },
    {
      "epoch": 40.84,
      "learning_rate": 1.4554304207119742e-05,
      "loss": 0.0897,
      "step": 105170
    },
    {
      "epoch": 40.85,
      "learning_rate": 1.4553786407766992e-05,
      "loss": 0.1763,
      "step": 105180
    },
    {
      "epoch": 40.85,
      "learning_rate": 1.4553268608414241e-05,
      "loss": 0.2383,
      "step": 105190
    },
    {
      "epoch": 40.85,
      "learning_rate": 1.4552750809061489e-05,
      "loss": 0.0314,
      "step": 105200
    },
    {
      "epoch": 40.86,
      "learning_rate": 1.4552233009708739e-05,
      "loss": 0.192,
      "step": 105210
    },
    {
      "epoch": 40.86,
      "learning_rate": 1.4551715210355988e-05,
      "loss": 0.0912,
      "step": 105220
    },
    {
      "epoch": 40.87,
      "learning_rate": 1.4551197411003238e-05,
      "loss": 0.0575,
      "step": 105230
    },
    {
      "epoch": 40.87,
      "learning_rate": 1.4550679611650486e-05,
      "loss": 0.1005,
      "step": 105240
    },
    {
      "epoch": 40.87,
      "learning_rate": 1.4550161812297735e-05,
      "loss": 0.1279,
      "step": 105250
    },
    {
      "epoch": 40.88,
      "learning_rate": 1.4549644012944985e-05,
      "loss": 0.0821,
      "step": 105260
    },
    {
      "epoch": 40.88,
      "learning_rate": 1.4549126213592235e-05,
      "loss": 0.1343,
      "step": 105270
    },
    {
      "epoch": 40.89,
      "learning_rate": 1.4548608414239483e-05,
      "loss": 0.0398,
      "step": 105280
    },
    {
      "epoch": 40.89,
      "learning_rate": 1.4548090614886732e-05,
      "loss": 0.023,
      "step": 105290
    },
    {
      "epoch": 40.89,
      "learning_rate": 1.4547572815533982e-05,
      "loss": 0.1831,
      "step": 105300
    },
    {
      "epoch": 40.9,
      "learning_rate": 1.4547055016181231e-05,
      "loss": 0.0436,
      "step": 105310
    },
    {
      "epoch": 40.9,
      "learning_rate": 1.454653721682848e-05,
      "loss": 0.0251,
      "step": 105320
    },
    {
      "epoch": 40.9,
      "learning_rate": 1.4546019417475729e-05,
      "loss": 0.0694,
      "step": 105330
    },
    {
      "epoch": 40.91,
      "learning_rate": 1.4545501618122979e-05,
      "loss": 0.0694,
      "step": 105340
    },
    {
      "epoch": 40.91,
      "learning_rate": 1.4544983818770228e-05,
      "loss": 0.1078,
      "step": 105350
    },
    {
      "epoch": 40.92,
      "learning_rate": 1.4544466019417476e-05,
      "loss": 0.1371,
      "step": 105360
    },
    {
      "epoch": 40.92,
      "learning_rate": 1.4543948220064726e-05,
      "loss": 0.0754,
      "step": 105370
    },
    {
      "epoch": 40.92,
      "learning_rate": 1.4543430420711975e-05,
      "loss": 0.1372,
      "step": 105380
    },
    {
      "epoch": 40.93,
      "learning_rate": 1.4542912621359225e-05,
      "loss": 0.1088,
      "step": 105390
    },
    {
      "epoch": 40.93,
      "learning_rate": 1.4542394822006473e-05,
      "loss": 0.0711,
      "step": 105400
    },
    {
      "epoch": 40.94,
      "learning_rate": 1.4541877022653723e-05,
      "loss": 0.0614,
      "step": 105410
    },
    {
      "epoch": 40.94,
      "learning_rate": 1.4541359223300972e-05,
      "loss": 0.1192,
      "step": 105420
    },
    {
      "epoch": 40.94,
      "learning_rate": 1.4540841423948222e-05,
      "loss": 0.0211,
      "step": 105430
    },
    {
      "epoch": 40.95,
      "learning_rate": 1.4540323624595471e-05,
      "loss": 0.0042,
      "step": 105440
    },
    {
      "epoch": 40.95,
      "learning_rate": 1.453980582524272e-05,
      "loss": 0.0833,
      "step": 105450
    },
    {
      "epoch": 40.96,
      "learning_rate": 1.4539288025889969e-05,
      "loss": 0.1226,
      "step": 105460
    },
    {
      "epoch": 40.96,
      "learning_rate": 1.4538770226537217e-05,
      "loss": 0.0888,
      "step": 105470
    },
    {
      "epoch": 40.96,
      "learning_rate": 1.4538252427184467e-05,
      "loss": 0.1802,
      "step": 105480
    },
    {
      "epoch": 40.97,
      "learning_rate": 1.4537734627831716e-05,
      "loss": 0.2005,
      "step": 105490
    },
    {
      "epoch": 40.97,
      "learning_rate": 1.4537216828478966e-05,
      "loss": 0.0428,
      "step": 105500
    },
    {
      "epoch": 40.97,
      "learning_rate": 1.4536699029126215e-05,
      "loss": 0.0287,
      "step": 105510
    },
    {
      "epoch": 40.98,
      "learning_rate": 1.4536181229773465e-05,
      "loss": 0.1419,
      "step": 105520
    },
    {
      "epoch": 40.98,
      "learning_rate": 1.4535663430420713e-05,
      "loss": 0.0189,
      "step": 105530
    },
    {
      "epoch": 40.99,
      "learning_rate": 1.4535145631067963e-05,
      "loss": 0.1525,
      "step": 105540
    },
    {
      "epoch": 40.99,
      "learning_rate": 1.453462783171521e-05,
      "loss": 0.1164,
      "step": 105550
    },
    {
      "epoch": 40.99,
      "learning_rate": 1.453411003236246e-05,
      "loss": 0.096,
      "step": 105560
    },
    {
      "epoch": 41.0,
      "learning_rate": 1.453359223300971e-05,
      "loss": 0.0943,
      "step": 105570
    },
    {
      "epoch": 41.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.30752032995224,
      "eval_runtime": 8.282,
      "eval_samples_per_second": 438.906,
      "eval_steps_per_second": 54.939,
      "step": 105575
    },
    {
      "epoch": 41.0,
      "learning_rate": 1.453307443365696e-05,
      "loss": 0.0476,
      "step": 105580
    },
    {
      "epoch": 41.01,
      "learning_rate": 1.4532556634304209e-05,
      "loss": 0.2659,
      "step": 105590
    },
    {
      "epoch": 41.01,
      "learning_rate": 1.4532038834951459e-05,
      "loss": 0.0195,
      "step": 105600
    },
    {
      "epoch": 41.01,
      "learning_rate": 1.4531521035598706e-05,
      "loss": 0.0496,
      "step": 105610
    },
    {
      "epoch": 41.02,
      "learning_rate": 1.4531003236245954e-05,
      "loss": 0.0261,
      "step": 105620
    },
    {
      "epoch": 41.02,
      "learning_rate": 1.4530485436893204e-05,
      "loss": 0.0284,
      "step": 105630
    },
    {
      "epoch": 41.03,
      "learning_rate": 1.4529967637540454e-05,
      "loss": 0.0751,
      "step": 105640
    },
    {
      "epoch": 41.03,
      "learning_rate": 1.4529449838187703e-05,
      "loss": 0.1325,
      "step": 105650
    },
    {
      "epoch": 41.03,
      "learning_rate": 1.4528932038834953e-05,
      "loss": 0.0962,
      "step": 105660
    },
    {
      "epoch": 41.04,
      "learning_rate": 1.4528414239482202e-05,
      "loss": 0.1527,
      "step": 105670
    },
    {
      "epoch": 41.04,
      "learning_rate": 1.4527896440129452e-05,
      "loss": 0.1306,
      "step": 105680
    },
    {
      "epoch": 41.04,
      "learning_rate": 1.45273786407767e-05,
      "loss": 0.0232,
      "step": 105690
    },
    {
      "epoch": 41.05,
      "learning_rate": 1.4526860841423948e-05,
      "loss": 0.055,
      "step": 105700
    },
    {
      "epoch": 41.05,
      "learning_rate": 1.4526343042071198e-05,
      "loss": 0.0105,
      "step": 105710
    },
    {
      "epoch": 41.06,
      "learning_rate": 1.4525825242718447e-05,
      "loss": 0.1397,
      "step": 105720
    },
    {
      "epoch": 41.06,
      "learning_rate": 1.4525307443365697e-05,
      "loss": 0.0572,
      "step": 105730
    },
    {
      "epoch": 41.06,
      "learning_rate": 1.4524789644012946e-05,
      "loss": 0.0725,
      "step": 105740
    },
    {
      "epoch": 41.07,
      "learning_rate": 1.4524271844660196e-05,
      "loss": 0.192,
      "step": 105750
    },
    {
      "epoch": 41.07,
      "learning_rate": 1.4523754045307446e-05,
      "loss": 0.1866,
      "step": 105760
    },
    {
      "epoch": 41.08,
      "learning_rate": 1.4523236245954692e-05,
      "loss": 0.142,
      "step": 105770
    },
    {
      "epoch": 41.08,
      "learning_rate": 1.4522718446601942e-05,
      "loss": 0.0585,
      "step": 105780
    },
    {
      "epoch": 41.08,
      "learning_rate": 1.4522200647249191e-05,
      "loss": 0.0398,
      "step": 105790
    },
    {
      "epoch": 41.09,
      "learning_rate": 1.452168284789644e-05,
      "loss": 0.183,
      "step": 105800
    },
    {
      "epoch": 41.09,
      "learning_rate": 1.452116504854369e-05,
      "loss": 0.0389,
      "step": 105810
    },
    {
      "epoch": 41.1,
      "learning_rate": 1.452064724919094e-05,
      "loss": 0.0698,
      "step": 105820
    },
    {
      "epoch": 41.1,
      "learning_rate": 1.452012944983819e-05,
      "loss": 0.0986,
      "step": 105830
    },
    {
      "epoch": 41.1,
      "learning_rate": 1.451961165048544e-05,
      "loss": 0.2809,
      "step": 105840
    },
    {
      "epoch": 41.11,
      "learning_rate": 1.4519093851132685e-05,
      "loss": 0.0431,
      "step": 105850
    },
    {
      "epoch": 41.11,
      "learning_rate": 1.4518576051779935e-05,
      "loss": 0.1241,
      "step": 105860
    },
    {
      "epoch": 41.11,
      "learning_rate": 1.4518058252427185e-05,
      "loss": 0.0583,
      "step": 105870
    },
    {
      "epoch": 41.12,
      "learning_rate": 1.4517540453074434e-05,
      "loss": 0.0125,
      "step": 105880
    },
    {
      "epoch": 41.12,
      "learning_rate": 1.4517022653721684e-05,
      "loss": 0.0397,
      "step": 105890
    },
    {
      "epoch": 41.13,
      "learning_rate": 1.4516504854368934e-05,
      "loss": 0.1454,
      "step": 105900
    },
    {
      "epoch": 41.13,
      "learning_rate": 1.4515987055016183e-05,
      "loss": 0.031,
      "step": 105910
    },
    {
      "epoch": 41.13,
      "learning_rate": 1.4515469255663433e-05,
      "loss": 0.0266,
      "step": 105920
    },
    {
      "epoch": 41.14,
      "learning_rate": 1.4514951456310679e-05,
      "loss": 0.1638,
      "step": 105930
    },
    {
      "epoch": 41.14,
      "learning_rate": 1.4514433656957929e-05,
      "loss": 0.1444,
      "step": 105940
    },
    {
      "epoch": 41.15,
      "learning_rate": 1.4513915857605178e-05,
      "loss": 0.0429,
      "step": 105950
    },
    {
      "epoch": 41.15,
      "learning_rate": 1.4513398058252428e-05,
      "loss": 0.078,
      "step": 105960
    },
    {
      "epoch": 41.15,
      "learning_rate": 1.4512880258899677e-05,
      "loss": 0.0689,
      "step": 105970
    },
    {
      "epoch": 41.16,
      "learning_rate": 1.4512362459546927e-05,
      "loss": 0.3087,
      "step": 105980
    },
    {
      "epoch": 41.16,
      "learning_rate": 1.4511844660194177e-05,
      "loss": 0.1536,
      "step": 105990
    },
    {
      "epoch": 41.17,
      "learning_rate": 1.4511326860841426e-05,
      "loss": 0.1169,
      "step": 106000
    },
    {
      "epoch": 41.17,
      "learning_rate": 1.4510809061488676e-05,
      "loss": 0.1162,
      "step": 106010
    },
    {
      "epoch": 41.17,
      "learning_rate": 1.4510291262135922e-05,
      "loss": 0.04,
      "step": 106020
    },
    {
      "epoch": 41.18,
      "learning_rate": 1.4509773462783172e-05,
      "loss": 0.1046,
      "step": 106030
    },
    {
      "epoch": 41.18,
      "learning_rate": 1.4509255663430421e-05,
      "loss": 0.0684,
      "step": 106040
    },
    {
      "epoch": 41.18,
      "learning_rate": 1.4508737864077671e-05,
      "loss": 0.2015,
      "step": 106050
    },
    {
      "epoch": 41.19,
      "learning_rate": 1.450822006472492e-05,
      "loss": 0.0765,
      "step": 106060
    },
    {
      "epoch": 41.19,
      "learning_rate": 1.450770226537217e-05,
      "loss": 0.041,
      "step": 106070
    },
    {
      "epoch": 41.2,
      "learning_rate": 1.450718446601942e-05,
      "loss": 0.1708,
      "step": 106080
    },
    {
      "epoch": 41.2,
      "learning_rate": 1.450666666666667e-05,
      "loss": 0.0803,
      "step": 106090
    },
    {
      "epoch": 41.2,
      "learning_rate": 1.4506148867313916e-05,
      "loss": 0.0709,
      "step": 106100
    },
    {
      "epoch": 41.21,
      "learning_rate": 1.4505631067961165e-05,
      "loss": 0.0331,
      "step": 106110
    },
    {
      "epoch": 41.21,
      "learning_rate": 1.4505113268608415e-05,
      "loss": 0.0474,
      "step": 106120
    },
    {
      "epoch": 41.22,
      "learning_rate": 1.4504595469255665e-05,
      "loss": 0.1259,
      "step": 106130
    },
    {
      "epoch": 41.22,
      "learning_rate": 1.4504077669902914e-05,
      "loss": 0.044,
      "step": 106140
    },
    {
      "epoch": 41.22,
      "learning_rate": 1.4503559870550164e-05,
      "loss": 0.1158,
      "step": 106150
    },
    {
      "epoch": 41.23,
      "learning_rate": 1.4503042071197413e-05,
      "loss": 0.0692,
      "step": 106160
    },
    {
      "epoch": 41.23,
      "learning_rate": 1.4502524271844663e-05,
      "loss": 0.0374,
      "step": 106170
    },
    {
      "epoch": 41.23,
      "learning_rate": 1.450200647249191e-05,
      "loss": 0.0912,
      "step": 106180
    },
    {
      "epoch": 41.24,
      "learning_rate": 1.4501488673139159e-05,
      "loss": 0.0327,
      "step": 106190
    },
    {
      "epoch": 41.24,
      "learning_rate": 1.4500970873786409e-05,
      "loss": 0.1957,
      "step": 106200
    },
    {
      "epoch": 41.25,
      "learning_rate": 1.4500453074433658e-05,
      "loss": 0.2219,
      "step": 106210
    },
    {
      "epoch": 41.25,
      "learning_rate": 1.4499935275080908e-05,
      "loss": 0.0375,
      "step": 106220
    },
    {
      "epoch": 41.25,
      "learning_rate": 1.4499417475728157e-05,
      "loss": 0.0592,
      "step": 106230
    },
    {
      "epoch": 41.26,
      "learning_rate": 1.4498899676375407e-05,
      "loss": 0.0302,
      "step": 106240
    },
    {
      "epoch": 41.26,
      "learning_rate": 1.4498381877022657e-05,
      "loss": 0.0888,
      "step": 106250
    },
    {
      "epoch": 41.27,
      "learning_rate": 1.4497864077669903e-05,
      "loss": 0.0585,
      "step": 106260
    },
    {
      "epoch": 41.27,
      "learning_rate": 1.4497346278317152e-05,
      "loss": 0.0684,
      "step": 106270
    },
    {
      "epoch": 41.27,
      "learning_rate": 1.4496828478964402e-05,
      "loss": 0.0634,
      "step": 106280
    },
    {
      "epoch": 41.28,
      "learning_rate": 1.4496310679611652e-05,
      "loss": 0.2387,
      "step": 106290
    },
    {
      "epoch": 41.28,
      "learning_rate": 1.4495792880258901e-05,
      "loss": 0.107,
      "step": 106300
    },
    {
      "epoch": 41.29,
      "learning_rate": 1.4495275080906151e-05,
      "loss": 0.1224,
      "step": 106310
    },
    {
      "epoch": 41.29,
      "learning_rate": 1.44947572815534e-05,
      "loss": 0.0648,
      "step": 106320
    },
    {
      "epoch": 41.29,
      "learning_rate": 1.449423948220065e-05,
      "loss": 0.1275,
      "step": 106330
    },
    {
      "epoch": 41.3,
      "learning_rate": 1.4493721682847896e-05,
      "loss": 0.1111,
      "step": 106340
    },
    {
      "epoch": 41.3,
      "learning_rate": 1.4493203883495146e-05,
      "loss": 0.1144,
      "step": 106350
    },
    {
      "epoch": 41.3,
      "learning_rate": 1.4492686084142396e-05,
      "loss": 0.0088,
      "step": 106360
    },
    {
      "epoch": 41.31,
      "learning_rate": 1.4492168284789645e-05,
      "loss": 0.0863,
      "step": 106370
    },
    {
      "epoch": 41.31,
      "learning_rate": 1.4491650485436895e-05,
      "loss": 0.0435,
      "step": 106380
    },
    {
      "epoch": 41.32,
      "learning_rate": 1.4491132686084145e-05,
      "loss": 0.1411,
      "step": 106390
    },
    {
      "epoch": 41.32,
      "learning_rate": 1.4490614886731394e-05,
      "loss": 0.0611,
      "step": 106400
    },
    {
      "epoch": 41.32,
      "learning_rate": 1.4490097087378642e-05,
      "loss": 0.1254,
      "step": 106410
    },
    {
      "epoch": 41.33,
      "learning_rate": 1.448957928802589e-05,
      "loss": 0.0144,
      "step": 106420
    },
    {
      "epoch": 41.33,
      "learning_rate": 1.448906148867314e-05,
      "loss": 0.0755,
      "step": 106430
    },
    {
      "epoch": 41.34,
      "learning_rate": 1.448854368932039e-05,
      "loss": 0.0007,
      "step": 106440
    },
    {
      "epoch": 41.34,
      "learning_rate": 1.4488025889967639e-05,
      "loss": 0.031,
      "step": 106450
    },
    {
      "epoch": 41.34,
      "learning_rate": 1.4487508090614888e-05,
      "loss": 0.0535,
      "step": 106460
    },
    {
      "epoch": 41.35,
      "learning_rate": 1.4486990291262138e-05,
      "loss": 0.0434,
      "step": 106470
    },
    {
      "epoch": 41.35,
      "learning_rate": 1.4486472491909388e-05,
      "loss": 0.2348,
      "step": 106480
    },
    {
      "epoch": 41.36,
      "learning_rate": 1.4485954692556636e-05,
      "loss": 0.0023,
      "step": 106490
    },
    {
      "epoch": 41.36,
      "learning_rate": 1.4485436893203884e-05,
      "loss": 0.0137,
      "step": 106500
    },
    {
      "epoch": 41.36,
      "learning_rate": 1.4484919093851133e-05,
      "loss": 0.0641,
      "step": 106510
    },
    {
      "epoch": 41.37,
      "learning_rate": 1.4484401294498383e-05,
      "loss": 0.0879,
      "step": 106520
    },
    {
      "epoch": 41.37,
      "learning_rate": 1.4483883495145632e-05,
      "loss": 0.1138,
      "step": 106530
    },
    {
      "epoch": 41.37,
      "learning_rate": 1.4483365695792882e-05,
      "loss": 0.0652,
      "step": 106540
    },
    {
      "epoch": 41.38,
      "learning_rate": 1.4482847896440132e-05,
      "loss": 0.088,
      "step": 106550
    },
    {
      "epoch": 41.38,
      "learning_rate": 1.448233009708738e-05,
      "loss": 0.13,
      "step": 106560
    },
    {
      "epoch": 41.39,
      "learning_rate": 1.448181229773463e-05,
      "loss": 0.0816,
      "step": 106570
    },
    {
      "epoch": 41.39,
      "learning_rate": 1.4481294498381879e-05,
      "loss": 0.0127,
      "step": 106580
    },
    {
      "epoch": 41.39,
      "learning_rate": 1.4480776699029127e-05,
      "loss": 0.0525,
      "step": 106590
    },
    {
      "epoch": 41.4,
      "learning_rate": 1.4480258899676376e-05,
      "loss": 0.0383,
      "step": 106600
    },
    {
      "epoch": 41.4,
      "learning_rate": 1.4479741100323626e-05,
      "loss": 0.0635,
      "step": 106610
    },
    {
      "epoch": 41.41,
      "learning_rate": 1.4479223300970876e-05,
      "loss": 0.1877,
      "step": 106620
    },
    {
      "epoch": 41.41,
      "learning_rate": 1.4478705501618125e-05,
      "loss": 0.2468,
      "step": 106630
    },
    {
      "epoch": 41.41,
      "learning_rate": 1.4478187702265373e-05,
      "loss": 0.0163,
      "step": 106640
    },
    {
      "epoch": 41.42,
      "learning_rate": 1.4477669902912623e-05,
      "loss": 0.1283,
      "step": 106650
    },
    {
      "epoch": 41.42,
      "learning_rate": 1.4477152103559872e-05,
      "loss": 0.0306,
      "step": 106660
    },
    {
      "epoch": 41.43,
      "learning_rate": 1.447663430420712e-05,
      "loss": 0.0272,
      "step": 106670
    },
    {
      "epoch": 41.43,
      "learning_rate": 1.447611650485437e-05,
      "loss": 0.0932,
      "step": 106680
    },
    {
      "epoch": 41.43,
      "learning_rate": 1.447559870550162e-05,
      "loss": 0.0189,
      "step": 106690
    },
    {
      "epoch": 41.44,
      "learning_rate": 1.4475080906148869e-05,
      "loss": 0.0246,
      "step": 106700
    },
    {
      "epoch": 41.44,
      "learning_rate": 1.4474563106796117e-05,
      "loss": 0.0123,
      "step": 106710
    },
    {
      "epoch": 41.44,
      "learning_rate": 1.4474045307443367e-05,
      "loss": 0.0744,
      "step": 106720
    },
    {
      "epoch": 41.45,
      "learning_rate": 1.4473527508090616e-05,
      "loss": 0.0348,
      "step": 106730
    },
    {
      "epoch": 41.45,
      "learning_rate": 1.4473009708737866e-05,
      "loss": 0.0806,
      "step": 106740
    },
    {
      "epoch": 41.46,
      "learning_rate": 1.4472491909385114e-05,
      "loss": 0.0625,
      "step": 106750
    },
    {
      "epoch": 41.46,
      "learning_rate": 1.4471974110032363e-05,
      "loss": 0.0989,
      "step": 106760
    },
    {
      "epoch": 41.46,
      "learning_rate": 1.4471456310679613e-05,
      "loss": 0.0413,
      "step": 106770
    },
    {
      "epoch": 41.47,
      "learning_rate": 1.4470938511326863e-05,
      "loss": 0.1258,
      "step": 106780
    },
    {
      "epoch": 41.47,
      "learning_rate": 1.447042071197411e-05,
      "loss": 0.0746,
      "step": 106790
    },
    {
      "epoch": 41.48,
      "learning_rate": 1.446990291262136e-05,
      "loss": 0.1428,
      "step": 106800
    },
    {
      "epoch": 41.48,
      "learning_rate": 1.446938511326861e-05,
      "loss": 0.0777,
      "step": 106810
    },
    {
      "epoch": 41.48,
      "learning_rate": 1.446886731391586e-05,
      "loss": 0.1865,
      "step": 106820
    },
    {
      "epoch": 41.49,
      "learning_rate": 1.4468349514563107e-05,
      "loss": 0.0564,
      "step": 106830
    },
    {
      "epoch": 41.49,
      "learning_rate": 1.4467831715210357e-05,
      "loss": 0.0956,
      "step": 106840
    },
    {
      "epoch": 41.5,
      "learning_rate": 1.4467313915857607e-05,
      "loss": 0.1574,
      "step": 106850
    },
    {
      "epoch": 41.5,
      "learning_rate": 1.4466796116504856e-05,
      "loss": 0.3024,
      "step": 106860
    },
    {
      "epoch": 41.5,
      "learning_rate": 1.4466278317152104e-05,
      "loss": 0.0117,
      "step": 106870
    },
    {
      "epoch": 41.51,
      "learning_rate": 1.4465760517799354e-05,
      "loss": 0.0215,
      "step": 106880
    },
    {
      "epoch": 41.51,
      "learning_rate": 1.4465242718446603e-05,
      "loss": 0.1202,
      "step": 106890
    },
    {
      "epoch": 41.51,
      "learning_rate": 1.4464724919093853e-05,
      "loss": 0.1238,
      "step": 106900
    },
    {
      "epoch": 41.52,
      "learning_rate": 1.4464207119741101e-05,
      "loss": 0.0007,
      "step": 106910
    },
    {
      "epoch": 41.52,
      "learning_rate": 1.446368932038835e-05,
      "loss": 0.0662,
      "step": 106920
    },
    {
      "epoch": 41.53,
      "learning_rate": 1.44631715210356e-05,
      "loss": 0.1244,
      "step": 106930
    },
    {
      "epoch": 41.53,
      "learning_rate": 1.4462653721682848e-05,
      "loss": 0.2429,
      "step": 106940
    },
    {
      "epoch": 41.53,
      "learning_rate": 1.4462135922330098e-05,
      "loss": 0.1481,
      "step": 106950
    },
    {
      "epoch": 41.54,
      "learning_rate": 1.4461618122977347e-05,
      "loss": 0.0392,
      "step": 106960
    },
    {
      "epoch": 41.54,
      "learning_rate": 1.4461100323624597e-05,
      "loss": 0.137,
      "step": 106970
    },
    {
      "epoch": 41.55,
      "learning_rate": 1.4460582524271847e-05,
      "loss": 0.131,
      "step": 106980
    },
    {
      "epoch": 41.55,
      "learning_rate": 1.4460064724919094e-05,
      "loss": 0.0657,
      "step": 106990
    },
    {
      "epoch": 41.55,
      "learning_rate": 1.4459546925566344e-05,
      "loss": 0.1604,
      "step": 107000
    },
    {
      "epoch": 41.56,
      "learning_rate": 1.4459029126213594e-05,
      "loss": 0.0857,
      "step": 107010
    },
    {
      "epoch": 41.56,
      "learning_rate": 1.4458511326860842e-05,
      "loss": 0.0357,
      "step": 107020
    },
    {
      "epoch": 41.57,
      "learning_rate": 1.4457993527508091e-05,
      "loss": 0.1334,
      "step": 107030
    },
    {
      "epoch": 41.57,
      "learning_rate": 1.4457475728155341e-05,
      "loss": 0.1217,
      "step": 107040
    },
    {
      "epoch": 41.57,
      "learning_rate": 1.445695792880259e-05,
      "loss": 0.0513,
      "step": 107050
    },
    {
      "epoch": 41.58,
      "learning_rate": 1.445644012944984e-05,
      "loss": 0.0502,
      "step": 107060
    },
    {
      "epoch": 41.58,
      "learning_rate": 1.4455922330097088e-05,
      "loss": 0.0889,
      "step": 107070
    },
    {
      "epoch": 41.58,
      "learning_rate": 1.4455404530744338e-05,
      "loss": 0.0425,
      "step": 107080
    },
    {
      "epoch": 41.59,
      "learning_rate": 1.4454886731391586e-05,
      "loss": 0.2441,
      "step": 107090
    },
    {
      "epoch": 41.59,
      "learning_rate": 1.4454368932038835e-05,
      "loss": 0.0927,
      "step": 107100
    },
    {
      "epoch": 41.6,
      "learning_rate": 1.4453851132686085e-05,
      "loss": 0.0172,
      "step": 107110
    },
    {
      "epoch": 41.6,
      "learning_rate": 1.4453333333333334e-05,
      "loss": 0.0007,
      "step": 107120
    },
    {
      "epoch": 41.6,
      "learning_rate": 1.4452815533980584e-05,
      "loss": 0.0533,
      "step": 107130
    },
    {
      "epoch": 41.61,
      "learning_rate": 1.4452297734627834e-05,
      "loss": 0.1533,
      "step": 107140
    },
    {
      "epoch": 41.61,
      "learning_rate": 1.4451779935275083e-05,
      "loss": 0.0994,
      "step": 107150
    },
    {
      "epoch": 41.62,
      "learning_rate": 1.4451262135922331e-05,
      "loss": 0.0784,
      "step": 107160
    },
    {
      "epoch": 41.62,
      "learning_rate": 1.4450744336569579e-05,
      "loss": 0.0782,
      "step": 107170
    },
    {
      "epoch": 41.62,
      "learning_rate": 1.4450226537216829e-05,
      "loss": 0.159,
      "step": 107180
    },
    {
      "epoch": 41.63,
      "learning_rate": 1.4449708737864078e-05,
      "loss": 0.0476,
      "step": 107190
    },
    {
      "epoch": 41.63,
      "learning_rate": 1.4449190938511328e-05,
      "loss": 0.1539,
      "step": 107200
    },
    {
      "epoch": 41.63,
      "learning_rate": 1.4448673139158578e-05,
      "loss": 0.1535,
      "step": 107210
    },
    {
      "epoch": 41.64,
      "learning_rate": 1.4448155339805827e-05,
      "loss": 0.0607,
      "step": 107220
    },
    {
      "epoch": 41.64,
      "learning_rate": 1.4447637540453077e-05,
      "loss": 0.0034,
      "step": 107230
    },
    {
      "epoch": 41.65,
      "learning_rate": 1.4447119741100323e-05,
      "loss": 0.2325,
      "step": 107240
    },
    {
      "epoch": 41.65,
      "learning_rate": 1.4446601941747573e-05,
      "loss": 0.139,
      "step": 107250
    },
    {
      "epoch": 41.65,
      "learning_rate": 1.4446084142394822e-05,
      "loss": 0.059,
      "step": 107260
    },
    {
      "epoch": 41.66,
      "learning_rate": 1.4445566343042072e-05,
      "loss": 0.0452,
      "step": 107270
    },
    {
      "epoch": 41.66,
      "learning_rate": 1.4445048543689322e-05,
      "loss": 0.193,
      "step": 107280
    },
    {
      "epoch": 41.67,
      "learning_rate": 1.4444530744336571e-05,
      "loss": 0.0855,
      "step": 107290
    },
    {
      "epoch": 41.67,
      "learning_rate": 1.444401294498382e-05,
      "loss": 0.0604,
      "step": 107300
    },
    {
      "epoch": 41.67,
      "learning_rate": 1.444349514563107e-05,
      "loss": 0.1636,
      "step": 107310
    },
    {
      "epoch": 41.68,
      "learning_rate": 1.4442977346278317e-05,
      "loss": 0.1498,
      "step": 107320
    },
    {
      "epoch": 41.68,
      "learning_rate": 1.4442459546925566e-05,
      "loss": 0.0537,
      "step": 107330
    },
    {
      "epoch": 41.69,
      "learning_rate": 1.4441941747572816e-05,
      "loss": 0.043,
      "step": 107340
    },
    {
      "epoch": 41.69,
      "learning_rate": 1.4441423948220065e-05,
      "loss": 0.0522,
      "step": 107350
    },
    {
      "epoch": 41.69,
      "learning_rate": 1.4440906148867315e-05,
      "loss": 0.1496,
      "step": 107360
    },
    {
      "epoch": 41.7,
      "learning_rate": 1.4440388349514565e-05,
      "loss": 0.0681,
      "step": 107370
    },
    {
      "epoch": 41.7,
      "learning_rate": 1.4439870550161814e-05,
      "loss": 0.1359,
      "step": 107380
    },
    {
      "epoch": 41.7,
      "learning_rate": 1.4439352750809064e-05,
      "loss": 0.16,
      "step": 107390
    },
    {
      "epoch": 41.71,
      "learning_rate": 1.443883495145631e-05,
      "loss": 0.1084,
      "step": 107400
    },
    {
      "epoch": 41.71,
      "learning_rate": 1.443831715210356e-05,
      "loss": 0.1352,
      "step": 107410
    },
    {
      "epoch": 41.72,
      "learning_rate": 1.443779935275081e-05,
      "loss": 0.1554,
      "step": 107420
    },
    {
      "epoch": 41.72,
      "learning_rate": 1.4437281553398059e-05,
      "loss": 0.0416,
      "step": 107430
    },
    {
      "epoch": 41.72,
      "learning_rate": 1.4436763754045309e-05,
      "loss": 0.0732,
      "step": 107440
    },
    {
      "epoch": 41.73,
      "learning_rate": 1.4436245954692558e-05,
      "loss": 0.1191,
      "step": 107450
    },
    {
      "epoch": 41.73,
      "learning_rate": 1.4435728155339808e-05,
      "loss": 0.0822,
      "step": 107460
    },
    {
      "epoch": 41.74,
      "learning_rate": 1.4435210355987058e-05,
      "loss": 0.0643,
      "step": 107470
    },
    {
      "epoch": 41.74,
      "learning_rate": 1.4434692556634304e-05,
      "loss": 0.042,
      "step": 107480
    },
    {
      "epoch": 41.74,
      "learning_rate": 1.4434174757281553e-05,
      "loss": 0.121,
      "step": 107490
    },
    {
      "epoch": 41.75,
      "learning_rate": 1.4433656957928803e-05,
      "loss": 0.1494,
      "step": 107500
    },
    {
      "epoch": 41.75,
      "learning_rate": 1.4433139158576053e-05,
      "loss": 0.0384,
      "step": 107510
    },
    {
      "epoch": 41.76,
      "learning_rate": 1.4432621359223302e-05,
      "loss": 0.0252,
      "step": 107520
    },
    {
      "epoch": 41.76,
      "learning_rate": 1.4432103559870552e-05,
      "loss": 0.0634,
      "step": 107530
    },
    {
      "epoch": 41.76,
      "learning_rate": 1.4431585760517801e-05,
      "loss": 0.2084,
      "step": 107540
    },
    {
      "epoch": 41.77,
      "learning_rate": 1.4431067961165051e-05,
      "loss": 0.0539,
      "step": 107550
    },
    {
      "epoch": 41.77,
      "learning_rate": 1.4430550161812297e-05,
      "loss": 0.1632,
      "step": 107560
    },
    {
      "epoch": 41.77,
      "learning_rate": 1.4430032362459547e-05,
      "loss": 0.0012,
      "step": 107570
    },
    {
      "epoch": 41.78,
      "learning_rate": 1.4429514563106797e-05,
      "loss": 0.0259,
      "step": 107580
    },
    {
      "epoch": 41.78,
      "learning_rate": 1.4428996763754046e-05,
      "loss": 0.0906,
      "step": 107590
    },
    {
      "epoch": 41.79,
      "learning_rate": 1.4428478964401296e-05,
      "loss": 0.146,
      "step": 107600
    },
    {
      "epoch": 41.79,
      "learning_rate": 1.4427961165048545e-05,
      "loss": 0.0529,
      "step": 107610
    },
    {
      "epoch": 41.79,
      "learning_rate": 1.4427443365695795e-05,
      "loss": 0.0226,
      "step": 107620
    },
    {
      "epoch": 41.8,
      "learning_rate": 1.4426925566343045e-05,
      "loss": 0.0814,
      "step": 107630
    },
    {
      "epoch": 41.8,
      "learning_rate": 1.4426407766990294e-05,
      "loss": 0.0606,
      "step": 107640
    },
    {
      "epoch": 41.81,
      "learning_rate": 1.442588996763754e-05,
      "loss": 0.2571,
      "step": 107650
    },
    {
      "epoch": 41.81,
      "learning_rate": 1.442537216828479e-05,
      "loss": 0.1144,
      "step": 107660
    },
    {
      "epoch": 41.81,
      "learning_rate": 1.442485436893204e-05,
      "loss": 0.0218,
      "step": 107670
    },
    {
      "epoch": 41.82,
      "learning_rate": 1.442433656957929e-05,
      "loss": 0.082,
      "step": 107680
    },
    {
      "epoch": 41.82,
      "learning_rate": 1.4423818770226539e-05,
      "loss": 0.0004,
      "step": 107690
    },
    {
      "epoch": 41.83,
      "learning_rate": 1.4423300970873789e-05,
      "loss": 0.136,
      "step": 107700
    },
    {
      "epoch": 41.83,
      "learning_rate": 1.4422783171521038e-05,
      "loss": 0.1199,
      "step": 107710
    },
    {
      "epoch": 41.83,
      "learning_rate": 1.4422265372168288e-05,
      "loss": 0.077,
      "step": 107720
    },
    {
      "epoch": 41.84,
      "learning_rate": 1.4421747572815534e-05,
      "loss": 0.0015,
      "step": 107730
    },
    {
      "epoch": 41.84,
      "learning_rate": 1.4421229773462784e-05,
      "loss": 0.1094,
      "step": 107740
    },
    {
      "epoch": 41.84,
      "learning_rate": 1.4420711974110033e-05,
      "loss": 0.0005,
      "step": 107750
    },
    {
      "epoch": 41.85,
      "learning_rate": 1.4420194174757283e-05,
      "loss": 0.0749,
      "step": 107760
    },
    {
      "epoch": 41.85,
      "learning_rate": 1.4419676375404532e-05,
      "loss": 0.0443,
      "step": 107770
    },
    {
      "epoch": 41.86,
      "learning_rate": 1.4419158576051782e-05,
      "loss": 0.0195,
      "step": 107780
    },
    {
      "epoch": 41.86,
      "learning_rate": 1.4418640776699032e-05,
      "loss": 0.0594,
      "step": 107790
    },
    {
      "epoch": 41.86,
      "learning_rate": 1.4418122977346281e-05,
      "loss": 0.0118,
      "step": 107800
    },
    {
      "epoch": 41.87,
      "learning_rate": 1.4417605177993528e-05,
      "loss": 0.0457,
      "step": 107810
    },
    {
      "epoch": 41.87,
      "learning_rate": 1.4417087378640777e-05,
      "loss": 0.0327,
      "step": 107820
    },
    {
      "epoch": 41.88,
      "learning_rate": 1.4416569579288027e-05,
      "loss": 0.0659,
      "step": 107830
    },
    {
      "epoch": 41.88,
      "learning_rate": 1.4416051779935276e-05,
      "loss": 0.1326,
      "step": 107840
    },
    {
      "epoch": 41.88,
      "learning_rate": 1.4415533980582526e-05,
      "loss": 0.0028,
      "step": 107850
    },
    {
      "epoch": 41.89,
      "learning_rate": 1.4415016181229776e-05,
      "loss": 0.0004,
      "step": 107860
    },
    {
      "epoch": 41.89,
      "learning_rate": 1.4414498381877025e-05,
      "loss": 0.0308,
      "step": 107870
    },
    {
      "epoch": 41.9,
      "learning_rate": 1.4413980582524273e-05,
      "loss": 0.061,
      "step": 107880
    },
    {
      "epoch": 41.9,
      "learning_rate": 1.4413462783171521e-05,
      "loss": 0.0317,
      "step": 107890
    },
    {
      "epoch": 41.9,
      "learning_rate": 1.441294498381877e-05,
      "loss": 0.3771,
      "step": 107900
    },
    {
      "epoch": 41.91,
      "learning_rate": 1.441242718446602e-05,
      "loss": 0.0993,
      "step": 107910
    },
    {
      "epoch": 41.91,
      "learning_rate": 1.441190938511327e-05,
      "loss": 0.0966,
      "step": 107920
    },
    {
      "epoch": 41.91,
      "learning_rate": 1.441139158576052e-05,
      "loss": 0.0325,
      "step": 107930
    },
    {
      "epoch": 41.92,
      "learning_rate": 1.441087378640777e-05,
      "loss": 0.1256,
      "step": 107940
    },
    {
      "epoch": 41.92,
      "learning_rate": 1.4410355987055019e-05,
      "loss": 0.0293,
      "step": 107950
    },
    {
      "epoch": 41.93,
      "learning_rate": 1.4409838187702267e-05,
      "loss": 0.0121,
      "step": 107960
    },
    {
      "epoch": 41.93,
      "learning_rate": 1.4409320388349515e-05,
      "loss": 0.1379,
      "step": 107970
    },
    {
      "epoch": 41.93,
      "learning_rate": 1.4408802588996764e-05,
      "loss": 0.1667,
      "step": 107980
    },
    {
      "epoch": 41.94,
      "learning_rate": 1.4408284789644014e-05,
      "loss": 0.1153,
      "step": 107990
    },
    {
      "epoch": 41.94,
      "learning_rate": 1.4407766990291264e-05,
      "loss": 0.1259,
      "step": 108000
    },
    {
      "epoch": 41.95,
      "learning_rate": 1.4407249190938513e-05,
      "loss": 0.0505,
      "step": 108010
    },
    {
      "epoch": 41.95,
      "learning_rate": 1.4406731391585763e-05,
      "loss": 0.1428,
      "step": 108020
    },
    {
      "epoch": 41.95,
      "learning_rate": 1.440621359223301e-05,
      "loss": 0.1565,
      "step": 108030
    },
    {
      "epoch": 41.96,
      "learning_rate": 1.440569579288026e-05,
      "loss": 0.1931,
      "step": 108040
    },
    {
      "epoch": 41.96,
      "learning_rate": 1.4405177993527508e-05,
      "loss": 0.0551,
      "step": 108050
    },
    {
      "epoch": 41.97,
      "learning_rate": 1.4404660194174758e-05,
      "loss": 0.1546,
      "step": 108060
    },
    {
      "epoch": 41.97,
      "learning_rate": 1.4404142394822007e-05,
      "loss": 0.0484,
      "step": 108070
    },
    {
      "epoch": 41.97,
      "learning_rate": 1.4403624595469257e-05,
      "loss": 0.2425,
      "step": 108080
    },
    {
      "epoch": 41.98,
      "learning_rate": 1.4403106796116507e-05,
      "loss": 0.0775,
      "step": 108090
    },
    {
      "epoch": 41.98,
      "learning_rate": 1.4402588996763756e-05,
      "loss": 0.1118,
      "step": 108100
    },
    {
      "epoch": 41.98,
      "learning_rate": 1.4402071197411004e-05,
      "loss": 0.0833,
      "step": 108110
    },
    {
      "epoch": 41.99,
      "learning_rate": 1.4401553398058254e-05,
      "loss": 0.0532,
      "step": 108120
    },
    {
      "epoch": 41.99,
      "learning_rate": 1.4401035598705502e-05,
      "loss": 0.2897,
      "step": 108130
    },
    {
      "epoch": 42.0,
      "learning_rate": 1.4400517799352751e-05,
      "loss": 0.0734,
      "step": 108140
    },
    {
      "epoch": 42.0,
      "learning_rate": 1.4400000000000001e-05,
      "loss": 0.0018,
      "step": 108150
    },
    {
      "epoch": 42.0,
      "eval_accuracy": 0.951856946354883,
      "eval_loss": 0.3062381148338318,
      "eval_runtime": 8.1991,
      "eval_samples_per_second": 443.341,
      "eval_steps_per_second": 55.494,
      "step": 108150
    },
    {
      "epoch": 42.0,
      "learning_rate": 1.439948220064725e-05,
      "loss": 0.1145,
      "step": 108160
    },
    {
      "epoch": 42.01,
      "learning_rate": 1.43989644012945e-05,
      "loss": 0.0497,
      "step": 108170
    },
    {
      "epoch": 42.01,
      "learning_rate": 1.4398446601941748e-05,
      "loss": 0.1018,
      "step": 108180
    },
    {
      "epoch": 42.02,
      "learning_rate": 1.4397928802588998e-05,
      "loss": 0.0391,
      "step": 108190
    },
    {
      "epoch": 42.02,
      "learning_rate": 1.4397411003236247e-05,
      "loss": 0.2001,
      "step": 108200
    },
    {
      "epoch": 42.02,
      "learning_rate": 1.4396893203883497e-05,
      "loss": 0.0848,
      "step": 108210
    },
    {
      "epoch": 42.03,
      "learning_rate": 1.4396375404530745e-05,
      "loss": 0.0906,
      "step": 108220
    },
    {
      "epoch": 42.03,
      "learning_rate": 1.4395857605177995e-05,
      "loss": 0.2576,
      "step": 108230
    },
    {
      "epoch": 42.03,
      "learning_rate": 1.4395339805825244e-05,
      "loss": 0.062,
      "step": 108240
    },
    {
      "epoch": 42.04,
      "learning_rate": 1.4394822006472494e-05,
      "loss": 0.1169,
      "step": 108250
    },
    {
      "epoch": 42.04,
      "learning_rate": 1.4394304207119742e-05,
      "loss": 0.0024,
      "step": 108260
    },
    {
      "epoch": 42.05,
      "learning_rate": 1.4393786407766991e-05,
      "loss": 0.0312,
      "step": 108270
    },
    {
      "epoch": 42.05,
      "learning_rate": 1.4393268608414241e-05,
      "loss": 0.116,
      "step": 108280
    },
    {
      "epoch": 42.05,
      "learning_rate": 1.439275080906149e-05,
      "loss": 0.0845,
      "step": 108290
    },
    {
      "epoch": 42.06,
      "learning_rate": 1.4392233009708739e-05,
      "loss": 0.0468,
      "step": 108300
    },
    {
      "epoch": 42.06,
      "learning_rate": 1.4391715210355988e-05,
      "loss": 0.0546,
      "step": 108310
    },
    {
      "epoch": 42.07,
      "learning_rate": 1.4391197411003238e-05,
      "loss": 0.0552,
      "step": 108320
    },
    {
      "epoch": 42.07,
      "learning_rate": 1.4390679611650487e-05,
      "loss": 0.0781,
      "step": 108330
    },
    {
      "epoch": 42.07,
      "learning_rate": 1.4390161812297735e-05,
      "loss": 0.145,
      "step": 108340
    },
    {
      "epoch": 42.08,
      "learning_rate": 1.4389644012944985e-05,
      "loss": 0.0225,
      "step": 108350
    },
    {
      "epoch": 42.08,
      "learning_rate": 1.4389126213592235e-05,
      "loss": 0.142,
      "step": 108360
    },
    {
      "epoch": 42.09,
      "learning_rate": 1.4388608414239484e-05,
      "loss": 0.0222,
      "step": 108370
    },
    {
      "epoch": 42.09,
      "learning_rate": 1.4388090614886732e-05,
      "loss": 0.0384,
      "step": 108380
    },
    {
      "epoch": 42.09,
      "learning_rate": 1.4387572815533982e-05,
      "loss": 0.0186,
      "step": 108390
    },
    {
      "epoch": 42.1,
      "learning_rate": 1.4387055016181231e-05,
      "loss": 0.0383,
      "step": 108400
    },
    {
      "epoch": 42.1,
      "learning_rate": 1.438653721682848e-05,
      "loss": 0.0778,
      "step": 108410
    },
    {
      "epoch": 42.1,
      "learning_rate": 1.4386019417475729e-05,
      "loss": 0.0546,
      "step": 108420
    },
    {
      "epoch": 42.11,
      "learning_rate": 1.4385501618122978e-05,
      "loss": 0.0316,
      "step": 108430
    },
    {
      "epoch": 42.11,
      "learning_rate": 1.4384983818770228e-05,
      "loss": 0.0102,
      "step": 108440
    },
    {
      "epoch": 42.12,
      "learning_rate": 1.4384466019417478e-05,
      "loss": 0.1168,
      "step": 108450
    },
    {
      "epoch": 42.12,
      "learning_rate": 1.4383948220064726e-05,
      "loss": 0.1587,
      "step": 108460
    },
    {
      "epoch": 42.12,
      "learning_rate": 1.4383430420711975e-05,
      "loss": 0.0445,
      "step": 108470
    },
    {
      "epoch": 42.13,
      "learning_rate": 1.4382912621359225e-05,
      "loss": 0.3303,
      "step": 108480
    },
    {
      "epoch": 42.13,
      "learning_rate": 1.4382394822006473e-05,
      "loss": 0.0989,
      "step": 108490
    },
    {
      "epoch": 42.14,
      "learning_rate": 1.4381877022653722e-05,
      "loss": 0.0632,
      "step": 108500
    },
    {
      "epoch": 42.14,
      "learning_rate": 1.4381359223300972e-05,
      "loss": 0.0839,
      "step": 108510
    },
    {
      "epoch": 42.14,
      "learning_rate": 1.4380841423948222e-05,
      "loss": 0.0724,
      "step": 108520
    },
    {
      "epoch": 42.15,
      "learning_rate": 1.4380323624595471e-05,
      "loss": 0.1131,
      "step": 108530
    },
    {
      "epoch": 42.15,
      "learning_rate": 1.437980582524272e-05,
      "loss": 0.0477,
      "step": 108540
    },
    {
      "epoch": 42.16,
      "learning_rate": 1.4379288025889969e-05,
      "loss": 0.1338,
      "step": 108550
    },
    {
      "epoch": 42.16,
      "learning_rate": 1.4378770226537217e-05,
      "loss": 0.0363,
      "step": 108560
    },
    {
      "epoch": 42.16,
      "learning_rate": 1.4378252427184466e-05,
      "loss": 0.0934,
      "step": 108570
    },
    {
      "epoch": 42.17,
      "learning_rate": 1.4377734627831716e-05,
      "loss": 0.0422,
      "step": 108580
    },
    {
      "epoch": 42.17,
      "learning_rate": 1.4377216828478966e-05,
      "loss": 0.0099,
      "step": 108590
    },
    {
      "epoch": 42.17,
      "learning_rate": 1.4376699029126215e-05,
      "loss": 0.0449,
      "step": 108600
    },
    {
      "epoch": 42.18,
      "learning_rate": 1.4376181229773465e-05,
      "loss": 0.0017,
      "step": 108610
    },
    {
      "epoch": 42.18,
      "learning_rate": 1.4375663430420713e-05,
      "loss": 0.0563,
      "step": 108620
    },
    {
      "epoch": 42.19,
      "learning_rate": 1.4375145631067962e-05,
      "loss": 0.1052,
      "step": 108630
    },
    {
      "epoch": 42.19,
      "learning_rate": 1.437462783171521e-05,
      "loss": 0.0569,
      "step": 108640
    },
    {
      "epoch": 42.19,
      "learning_rate": 1.437411003236246e-05,
      "loss": 0.0908,
      "step": 108650
    },
    {
      "epoch": 42.2,
      "learning_rate": 1.437359223300971e-05,
      "loss": 0.0908,
      "step": 108660
    },
    {
      "epoch": 42.2,
      "learning_rate": 1.437307443365696e-05,
      "loss": 0.1022,
      "step": 108670
    },
    {
      "epoch": 42.21,
      "learning_rate": 1.4372556634304209e-05,
      "loss": 0.2435,
      "step": 108680
    },
    {
      "epoch": 42.21,
      "learning_rate": 1.4372038834951458e-05,
      "loss": 0.0425,
      "step": 108690
    },
    {
      "epoch": 42.21,
      "learning_rate": 1.4371521035598706e-05,
      "loss": 0.0185,
      "step": 108700
    },
    {
      "epoch": 42.22,
      "learning_rate": 1.4371003236245954e-05,
      "loss": 0.0525,
      "step": 108710
    },
    {
      "epoch": 42.22,
      "learning_rate": 1.4370485436893204e-05,
      "loss": 0.0953,
      "step": 108720
    },
    {
      "epoch": 42.23,
      "learning_rate": 1.4369967637540453e-05,
      "loss": 0.1887,
      "step": 108730
    },
    {
      "epoch": 42.23,
      "learning_rate": 1.4369449838187703e-05,
      "loss": 0.0085,
      "step": 108740
    },
    {
      "epoch": 42.23,
      "learning_rate": 1.4368932038834953e-05,
      "loss": 0.073,
      "step": 108750
    },
    {
      "epoch": 42.24,
      "learning_rate": 1.4368414239482202e-05,
      "loss": 0.0631,
      "step": 108760
    },
    {
      "epoch": 42.24,
      "learning_rate": 1.4367896440129452e-05,
      "loss": 0.0789,
      "step": 108770
    },
    {
      "epoch": 42.24,
      "learning_rate": 1.4367378640776702e-05,
      "loss": 0.1682,
      "step": 108780
    },
    {
      "epoch": 42.25,
      "learning_rate": 1.4366860841423948e-05,
      "loss": 0.043,
      "step": 108790
    },
    {
      "epoch": 42.25,
      "learning_rate": 1.4366343042071197e-05,
      "loss": 0.0397,
      "step": 108800
    },
    {
      "epoch": 42.26,
      "learning_rate": 1.4365825242718447e-05,
      "loss": 0.1086,
      "step": 108810
    },
    {
      "epoch": 42.26,
      "learning_rate": 1.4365307443365697e-05,
      "loss": 0.0721,
      "step": 108820
    },
    {
      "epoch": 42.26,
      "learning_rate": 1.4364789644012946e-05,
      "loss": 0.1072,
      "step": 108830
    },
    {
      "epoch": 42.27,
      "learning_rate": 1.4364271844660196e-05,
      "loss": 0.0111,
      "step": 108840
    },
    {
      "epoch": 42.27,
      "learning_rate": 1.4363754045307446e-05,
      "loss": 0.0864,
      "step": 108850
    },
    {
      "epoch": 42.28,
      "learning_rate": 1.4363236245954695e-05,
      "loss": 0.0892,
      "step": 108860
    },
    {
      "epoch": 42.28,
      "learning_rate": 1.4362718446601941e-05,
      "loss": 0.0006,
      "step": 108870
    },
    {
      "epoch": 42.28,
      "learning_rate": 1.4362200647249191e-05,
      "loss": 0.0423,
      "step": 108880
    },
    {
      "epoch": 42.29,
      "learning_rate": 1.436168284789644e-05,
      "loss": 0.0112,
      "step": 108890
    },
    {
      "epoch": 42.29,
      "learning_rate": 1.436116504854369e-05,
      "loss": 0.2227,
      "step": 108900
    },
    {
      "epoch": 42.3,
      "learning_rate": 1.436064724919094e-05,
      "loss": 0.0289,
      "step": 108910
    },
    {
      "epoch": 42.3,
      "learning_rate": 1.436012944983819e-05,
      "loss": 0.055,
      "step": 108920
    },
    {
      "epoch": 42.3,
      "learning_rate": 1.4359611650485439e-05,
      "loss": 0.0445,
      "step": 108930
    },
    {
      "epoch": 42.31,
      "learning_rate": 1.4359093851132689e-05,
      "loss": 0.1008,
      "step": 108940
    },
    {
      "epoch": 42.31,
      "learning_rate": 1.4358576051779935e-05,
      "loss": 0.0613,
      "step": 108950
    },
    {
      "epoch": 42.31,
      "learning_rate": 1.4358058252427185e-05,
      "loss": 0.035,
      "step": 108960
    },
    {
      "epoch": 42.32,
      "learning_rate": 1.4357540453074434e-05,
      "loss": 0.044,
      "step": 108970
    },
    {
      "epoch": 42.32,
      "learning_rate": 1.4357022653721684e-05,
      "loss": 0.1074,
      "step": 108980
    },
    {
      "epoch": 42.33,
      "learning_rate": 1.4356504854368933e-05,
      "loss": 0.0143,
      "step": 108990
    },
    {
      "epoch": 42.33,
      "learning_rate": 1.4355987055016183e-05,
      "loss": 0.0813,
      "step": 109000
    },
    {
      "epoch": 42.33,
      "learning_rate": 1.4355469255663433e-05,
      "loss": 0.11,
      "step": 109010
    },
    {
      "epoch": 42.34,
      "learning_rate": 1.4354951456310682e-05,
      "loss": 0.1026,
      "step": 109020
    },
    {
      "epoch": 42.34,
      "learning_rate": 1.4354433656957928e-05,
      "loss": 0.0651,
      "step": 109030
    },
    {
      "epoch": 42.35,
      "learning_rate": 1.4353915857605178e-05,
      "loss": 0.0922,
      "step": 109040
    },
    {
      "epoch": 42.35,
      "learning_rate": 1.4353398058252428e-05,
      "loss": 0.2437,
      "step": 109050
    },
    {
      "epoch": 42.35,
      "learning_rate": 1.4352880258899677e-05,
      "loss": 0.0151,
      "step": 109060
    },
    {
      "epoch": 42.36,
      "learning_rate": 1.4352362459546927e-05,
      "loss": 0.0438,
      "step": 109070
    },
    {
      "epoch": 42.36,
      "learning_rate": 1.4351844660194177e-05,
      "loss": 0.1338,
      "step": 109080
    },
    {
      "epoch": 42.37,
      "learning_rate": 1.4351326860841426e-05,
      "loss": 0.0701,
      "step": 109090
    },
    {
      "epoch": 42.37,
      "learning_rate": 1.4350809061488676e-05,
      "loss": 0.0704,
      "step": 109100
    },
    {
      "epoch": 42.37,
      "learning_rate": 1.4350291262135922e-05,
      "loss": 0.0279,
      "step": 109110
    },
    {
      "epoch": 42.38,
      "learning_rate": 1.4349773462783172e-05,
      "loss": 0.1986,
      "step": 109120
    },
    {
      "epoch": 42.38,
      "learning_rate": 1.4349255663430421e-05,
      "loss": 0.0574,
      "step": 109130
    },
    {
      "epoch": 42.38,
      "learning_rate": 1.4348737864077671e-05,
      "loss": 0.0883,
      "step": 109140
    },
    {
      "epoch": 42.39,
      "learning_rate": 1.434822006472492e-05,
      "loss": 0.0254,
      "step": 109150
    },
    {
      "epoch": 42.39,
      "learning_rate": 1.434770226537217e-05,
      "loss": 0.0012,
      "step": 109160
    },
    {
      "epoch": 42.4,
      "learning_rate": 1.434718446601942e-05,
      "loss": 0.0166,
      "step": 109170
    },
    {
      "epoch": 42.4,
      "learning_rate": 1.434666666666667e-05,
      "loss": 0.0087,
      "step": 109180
    },
    {
      "epoch": 42.4,
      "learning_rate": 1.4346148867313916e-05,
      "loss": 0.0371,
      "step": 109190
    },
    {
      "epoch": 42.41,
      "learning_rate": 1.4345631067961165e-05,
      "loss": 0.092,
      "step": 109200
    },
    {
      "epoch": 42.41,
      "learning_rate": 1.4345113268608415e-05,
      "loss": 0.0889,
      "step": 109210
    },
    {
      "epoch": 42.42,
      "learning_rate": 1.4344595469255664e-05,
      "loss": 0.061,
      "step": 109220
    },
    {
      "epoch": 42.42,
      "learning_rate": 1.4344077669902914e-05,
      "loss": 0.0156,
      "step": 109230
    },
    {
      "epoch": 42.42,
      "learning_rate": 1.4343559870550164e-05,
      "loss": 0.0286,
      "step": 109240
    },
    {
      "epoch": 42.43,
      "learning_rate": 1.4343042071197413e-05,
      "loss": 0.106,
      "step": 109250
    },
    {
      "epoch": 42.43,
      "learning_rate": 1.4342524271844663e-05,
      "loss": 0.0132,
      "step": 109260
    },
    {
      "epoch": 42.43,
      "learning_rate": 1.4342006472491909e-05,
      "loss": 0.0771,
      "step": 109270
    },
    {
      "epoch": 42.44,
      "learning_rate": 1.4341488673139159e-05,
      "loss": 0.0751,
      "step": 109280
    },
    {
      "epoch": 42.44,
      "learning_rate": 1.4340970873786408e-05,
      "loss": 0.1707,
      "step": 109290
    },
    {
      "epoch": 42.45,
      "learning_rate": 1.4340453074433658e-05,
      "loss": 0.0452,
      "step": 109300
    },
    {
      "epoch": 42.45,
      "learning_rate": 1.4339935275080908e-05,
      "loss": 0.0155,
      "step": 109310
    },
    {
      "epoch": 42.45,
      "learning_rate": 1.4339417475728157e-05,
      "loss": 0.0225,
      "step": 109320
    },
    {
      "epoch": 42.46,
      "learning_rate": 1.4338899676375407e-05,
      "loss": 0.0431,
      "step": 109330
    },
    {
      "epoch": 42.46,
      "learning_rate": 1.4338381877022656e-05,
      "loss": 0.2571,
      "step": 109340
    },
    {
      "epoch": 42.47,
      "learning_rate": 1.4337864077669904e-05,
      "loss": 0.2743,
      "step": 109350
    },
    {
      "epoch": 42.47,
      "learning_rate": 1.4337346278317152e-05,
      "loss": 0.0286,
      "step": 109360
    },
    {
      "epoch": 42.47,
      "learning_rate": 1.4336828478964402e-05,
      "loss": 0.0522,
      "step": 109370
    },
    {
      "epoch": 42.48,
      "learning_rate": 1.4336310679611652e-05,
      "loss": 0.0927,
      "step": 109380
    },
    {
      "epoch": 42.48,
      "learning_rate": 1.4335792880258901e-05,
      "loss": 0.1181,
      "step": 109390
    },
    {
      "epoch": 42.49,
      "learning_rate": 1.433527508090615e-05,
      "loss": 0.0794,
      "step": 109400
    },
    {
      "epoch": 42.49,
      "learning_rate": 1.43347572815534e-05,
      "loss": 0.0968,
      "step": 109410
    },
    {
      "epoch": 42.49,
      "learning_rate": 1.433423948220065e-05,
      "loss": 0.3346,
      "step": 109420
    },
    {
      "epoch": 42.5,
      "learning_rate": 1.4333721682847898e-05,
      "loss": 0.1778,
      "step": 109430
    },
    {
      "epoch": 42.5,
      "learning_rate": 1.4333203883495146e-05,
      "loss": 0.1185,
      "step": 109440
    },
    {
      "epoch": 42.5,
      "learning_rate": 1.4332686084142395e-05,
      "loss": 0.128,
      "step": 109450
    },
    {
      "epoch": 42.51,
      "learning_rate": 1.4332168284789645e-05,
      "loss": 0.0931,
      "step": 109460
    },
    {
      "epoch": 42.51,
      "learning_rate": 1.4331650485436895e-05,
      "loss": 0.1722,
      "step": 109470
    },
    {
      "epoch": 42.52,
      "learning_rate": 1.4331132686084144e-05,
      "loss": 0.1178,
      "step": 109480
    },
    {
      "epoch": 42.52,
      "learning_rate": 1.4330614886731394e-05,
      "loss": 0.1376,
      "step": 109490
    },
    {
      "epoch": 42.52,
      "learning_rate": 1.4330097087378642e-05,
      "loss": 0.0381,
      "step": 109500
    },
    {
      "epoch": 42.53,
      "learning_rate": 1.4329579288025891e-05,
      "loss": 0.0153,
      "step": 109510
    },
    {
      "epoch": 42.53,
      "learning_rate": 1.432906148867314e-05,
      "loss": 0.1925,
      "step": 109520
    },
    {
      "epoch": 42.54,
      "learning_rate": 1.4328543689320389e-05,
      "loss": 0.0316,
      "step": 109530
    },
    {
      "epoch": 42.54,
      "learning_rate": 1.4328025889967639e-05,
      "loss": 0.0603,
      "step": 109540
    },
    {
      "epoch": 42.54,
      "learning_rate": 1.4327508090614888e-05,
      "loss": 0.0873,
      "step": 109550
    },
    {
      "epoch": 42.55,
      "learning_rate": 1.4326990291262138e-05,
      "loss": 0.1117,
      "step": 109560
    },
    {
      "epoch": 42.55,
      "learning_rate": 1.4326472491909388e-05,
      "loss": 0.013,
      "step": 109570
    },
    {
      "epoch": 42.56,
      "learning_rate": 1.4325954692556635e-05,
      "loss": 0.099,
      "step": 109580
    },
    {
      "epoch": 42.56,
      "learning_rate": 1.4325436893203885e-05,
      "loss": 0.1597,
      "step": 109590
    },
    {
      "epoch": 42.56,
      "learning_rate": 1.4324919093851133e-05,
      "loss": 0.018,
      "step": 109600
    },
    {
      "epoch": 42.57,
      "learning_rate": 1.4324401294498383e-05,
      "loss": 0.0563,
      "step": 109610
    },
    {
      "epoch": 42.57,
      "learning_rate": 1.4323883495145632e-05,
      "loss": 0.189,
      "step": 109620
    },
    {
      "epoch": 42.57,
      "learning_rate": 1.4323365695792882e-05,
      "loss": 0.0139,
      "step": 109630
    },
    {
      "epoch": 42.58,
      "learning_rate": 1.4322847896440131e-05,
      "loss": 0.1256,
      "step": 109640
    },
    {
      "epoch": 42.58,
      "learning_rate": 1.432233009708738e-05,
      "loss": 0.0891,
      "step": 109650
    },
    {
      "epoch": 42.59,
      "learning_rate": 1.4321812297734629e-05,
      "loss": 0.0511,
      "step": 109660
    },
    {
      "epoch": 42.59,
      "learning_rate": 1.4321294498381879e-05,
      "loss": 0.0906,
      "step": 109670
    },
    {
      "epoch": 42.59,
      "learning_rate": 1.4320776699029127e-05,
      "loss": 0.1231,
      "step": 109680
    },
    {
      "epoch": 42.6,
      "learning_rate": 1.4320258899676376e-05,
      "loss": 0.0159,
      "step": 109690
    },
    {
      "epoch": 42.6,
      "learning_rate": 1.4319741100323626e-05,
      "loss": 0.0003,
      "step": 109700
    },
    {
      "epoch": 42.61,
      "learning_rate": 1.4319223300970875e-05,
      "loss": 0.269,
      "step": 109710
    },
    {
      "epoch": 42.61,
      "learning_rate": 1.4318705501618125e-05,
      "loss": 0.2327,
      "step": 109720
    },
    {
      "epoch": 42.61,
      "learning_rate": 1.4318187702265373e-05,
      "loss": 0.0519,
      "step": 109730
    },
    {
      "epoch": 42.62,
      "learning_rate": 1.4317669902912623e-05,
      "loss": 0.0884,
      "step": 109740
    },
    {
      "epoch": 42.62,
      "learning_rate": 1.4317152103559872e-05,
      "loss": 0.0247,
      "step": 109750
    },
    {
      "epoch": 42.63,
      "learning_rate": 1.431663430420712e-05,
      "loss": 0.0819,
      "step": 109760
    },
    {
      "epoch": 42.63,
      "learning_rate": 1.431611650485437e-05,
      "loss": 0.0719,
      "step": 109770
    },
    {
      "epoch": 42.63,
      "learning_rate": 1.431559870550162e-05,
      "loss": 0.0951,
      "step": 109780
    },
    {
      "epoch": 42.64,
      "learning_rate": 1.4315080906148869e-05,
      "loss": 0.0333,
      "step": 109790
    },
    {
      "epoch": 42.64,
      "learning_rate": 1.4314563106796117e-05,
      "loss": 0.1231,
      "step": 109800
    },
    {
      "epoch": 42.64,
      "learning_rate": 1.4314045307443366e-05,
      "loss": 0.0394,
      "step": 109810
    },
    {
      "epoch": 42.65,
      "learning_rate": 1.4313527508090616e-05,
      "loss": 0.1134,
      "step": 109820
    },
    {
      "epoch": 42.65,
      "learning_rate": 1.4313009708737866e-05,
      "loss": 0.0469,
      "step": 109830
    },
    {
      "epoch": 42.66,
      "learning_rate": 1.4312491909385114e-05,
      "loss": 0.1567,
      "step": 109840
    },
    {
      "epoch": 42.66,
      "learning_rate": 1.4311974110032363e-05,
      "loss": 0.1041,
      "step": 109850
    },
    {
      "epoch": 42.66,
      "learning_rate": 1.4311456310679613e-05,
      "loss": 0.0702,
      "step": 109860
    },
    {
      "epoch": 42.67,
      "learning_rate": 1.4310938511326863e-05,
      "loss": 0.0218,
      "step": 109870
    },
    {
      "epoch": 42.67,
      "learning_rate": 1.431042071197411e-05,
      "loss": 0.0908,
      "step": 109880
    },
    {
      "epoch": 42.68,
      "learning_rate": 1.430990291262136e-05,
      "loss": 0.006,
      "step": 109890
    },
    {
      "epoch": 42.68,
      "learning_rate": 1.430938511326861e-05,
      "loss": 0.0483,
      "step": 109900
    },
    {
      "epoch": 42.68,
      "learning_rate": 1.430886731391586e-05,
      "loss": 0.0312,
      "step": 109910
    },
    {
      "epoch": 42.69,
      "learning_rate": 1.4308349514563109e-05,
      "loss": 0.1271,
      "step": 109920
    },
    {
      "epoch": 42.69,
      "learning_rate": 1.4307831715210357e-05,
      "loss": 0.1347,
      "step": 109930
    },
    {
      "epoch": 42.7,
      "learning_rate": 1.4307313915857606e-05,
      "loss": 0.1291,
      "step": 109940
    },
    {
      "epoch": 42.7,
      "learning_rate": 1.4306796116504856e-05,
      "loss": 0.0849,
      "step": 109950
    },
    {
      "epoch": 42.7,
      "learning_rate": 1.4306278317152104e-05,
      "loss": 0.0307,
      "step": 109960
    },
    {
      "epoch": 42.71,
      "learning_rate": 1.4305760517799354e-05,
      "loss": 0.0835,
      "step": 109970
    },
    {
      "epoch": 42.71,
      "learning_rate": 1.4305242718446603e-05,
      "loss": 0.0686,
      "step": 109980
    },
    {
      "epoch": 42.71,
      "learning_rate": 1.4304724919093853e-05,
      "loss": 0.0107,
      "step": 109990
    },
    {
      "epoch": 42.72,
      "learning_rate": 1.4304207119741102e-05,
      "loss": 0.0375,
      "step": 110000
    },
    {
      "epoch": 42.72,
      "learning_rate": 1.430368932038835e-05,
      "loss": 0.0259,
      "step": 110010
    },
    {
      "epoch": 42.73,
      "learning_rate": 1.43031715210356e-05,
      "loss": 0.1343,
      "step": 110020
    },
    {
      "epoch": 42.73,
      "learning_rate": 1.4302653721682848e-05,
      "loss": 0.2079,
      "step": 110030
    },
    {
      "epoch": 42.73,
      "learning_rate": 1.4302135922330098e-05,
      "loss": 0.1439,
      "step": 110040
    },
    {
      "epoch": 42.74,
      "learning_rate": 1.4301618122977347e-05,
      "loss": 0.0051,
      "step": 110050
    },
    {
      "epoch": 42.74,
      "learning_rate": 1.4301100323624597e-05,
      "loss": 0.0445,
      "step": 110060
    },
    {
      "epoch": 42.75,
      "learning_rate": 1.4300582524271846e-05,
      "loss": 0.1062,
      "step": 110070
    },
    {
      "epoch": 42.75,
      "learning_rate": 1.4300064724919096e-05,
      "loss": 0.1363,
      "step": 110080
    },
    {
      "epoch": 42.75,
      "learning_rate": 1.4299546925566344e-05,
      "loss": 0.0486,
      "step": 110090
    },
    {
      "epoch": 42.76,
      "learning_rate": 1.4299029126213594e-05,
      "loss": 0.1843,
      "step": 110100
    },
    {
      "epoch": 42.76,
      "learning_rate": 1.4298511326860841e-05,
      "loss": 0.0623,
      "step": 110110
    },
    {
      "epoch": 42.77,
      "learning_rate": 1.4297993527508091e-05,
      "loss": 0.0773,
      "step": 110120
    },
    {
      "epoch": 42.77,
      "learning_rate": 1.429747572815534e-05,
      "loss": 0.2387,
      "step": 110130
    },
    {
      "epoch": 42.77,
      "learning_rate": 1.429695792880259e-05,
      "loss": 0.2098,
      "step": 110140
    },
    {
      "epoch": 42.78,
      "learning_rate": 1.429644012944984e-05,
      "loss": 0.1471,
      "step": 110150
    },
    {
      "epoch": 42.78,
      "learning_rate": 1.429592233009709e-05,
      "loss": 0.0314,
      "step": 110160
    },
    {
      "epoch": 42.78,
      "learning_rate": 1.4295404530744337e-05,
      "loss": 0.0697,
      "step": 110170
    },
    {
      "epoch": 42.79,
      "learning_rate": 1.4294886731391585e-05,
      "loss": 0.0047,
      "step": 110180
    },
    {
      "epoch": 42.79,
      "learning_rate": 1.4294368932038835e-05,
      "loss": 0.1226,
      "step": 110190
    },
    {
      "epoch": 42.8,
      "learning_rate": 1.4293851132686085e-05,
      "loss": 0.133,
      "step": 110200
    },
    {
      "epoch": 42.8,
      "learning_rate": 1.4293333333333334e-05,
      "loss": 0.2022,
      "step": 110210
    },
    {
      "epoch": 42.8,
      "learning_rate": 1.4292815533980584e-05,
      "loss": 0.1301,
      "step": 110220
    },
    {
      "epoch": 42.81,
      "learning_rate": 1.4292297734627834e-05,
      "loss": 0.1459,
      "step": 110230
    },
    {
      "epoch": 42.81,
      "learning_rate": 1.4291779935275083e-05,
      "loss": 0.0556,
      "step": 110240
    },
    {
      "epoch": 42.82,
      "learning_rate": 1.4291262135922331e-05,
      "loss": 0.0282,
      "step": 110250
    },
    {
      "epoch": 42.82,
      "learning_rate": 1.4290744336569579e-05,
      "loss": 0.2028,
      "step": 110260
    },
    {
      "epoch": 42.82,
      "learning_rate": 1.4290226537216829e-05,
      "loss": 0.0475,
      "step": 110270
    },
    {
      "epoch": 42.83,
      "learning_rate": 1.4289708737864078e-05,
      "loss": 0.0848,
      "step": 110280
    },
    {
      "epoch": 42.83,
      "learning_rate": 1.4289190938511328e-05,
      "loss": 0.0042,
      "step": 110290
    },
    {
      "epoch": 42.83,
      "learning_rate": 1.4288673139158577e-05,
      "loss": 0.0191,
      "step": 110300
    },
    {
      "epoch": 42.84,
      "learning_rate": 1.4288155339805827e-05,
      "loss": 0.142,
      "step": 110310
    },
    {
      "epoch": 42.84,
      "learning_rate": 1.4287637540453077e-05,
      "loss": 0.1659,
      "step": 110320
    },
    {
      "epoch": 42.85,
      "learning_rate": 1.4287119741100323e-05,
      "loss": 0.1027,
      "step": 110330
    },
    {
      "epoch": 42.85,
      "learning_rate": 1.4286601941747573e-05,
      "loss": 0.0458,
      "step": 110340
    },
    {
      "epoch": 42.85,
      "learning_rate": 1.4286084142394822e-05,
      "loss": 0.034,
      "step": 110350
    },
    {
      "epoch": 42.86,
      "learning_rate": 1.4285566343042072e-05,
      "loss": 0.2063,
      "step": 110360
    },
    {
      "epoch": 42.86,
      "learning_rate": 1.4285048543689321e-05,
      "loss": 0.0816,
      "step": 110370
    },
    {
      "epoch": 42.87,
      "learning_rate": 1.4284530744336571e-05,
      "loss": 0.0492,
      "step": 110380
    },
    {
      "epoch": 42.87,
      "learning_rate": 1.428401294498382e-05,
      "loss": 0.0252,
      "step": 110390
    },
    {
      "epoch": 42.87,
      "learning_rate": 1.428349514563107e-05,
      "loss": 0.026,
      "step": 110400
    },
    {
      "epoch": 42.88,
      "learning_rate": 1.4282977346278316e-05,
      "loss": 0.04,
      "step": 110410
    },
    {
      "epoch": 42.88,
      "learning_rate": 1.4282459546925566e-05,
      "loss": 0.1553,
      "step": 110420
    },
    {
      "epoch": 42.89,
      "learning_rate": 1.4281941747572816e-05,
      "loss": 0.128,
      "step": 110430
    },
    {
      "epoch": 42.89,
      "learning_rate": 1.4281423948220065e-05,
      "loss": 0.1954,
      "step": 110440
    },
    {
      "epoch": 42.89,
      "learning_rate": 1.4280906148867315e-05,
      "loss": 0.1066,
      "step": 110450
    },
    {
      "epoch": 42.9,
      "learning_rate": 1.4280388349514565e-05,
      "loss": 0.0824,
      "step": 110460
    },
    {
      "epoch": 42.9,
      "learning_rate": 1.4279870550161814e-05,
      "loss": 0.0309,
      "step": 110470
    },
    {
      "epoch": 42.9,
      "learning_rate": 1.4279352750809064e-05,
      "loss": 0.0743,
      "step": 110480
    },
    {
      "epoch": 42.91,
      "learning_rate": 1.4278834951456313e-05,
      "loss": 0.103,
      "step": 110490
    },
    {
      "epoch": 42.91,
      "learning_rate": 1.427831715210356e-05,
      "loss": 0.0618,
      "step": 110500
    },
    {
      "epoch": 42.92,
      "learning_rate": 1.427779935275081e-05,
      "loss": 0.0915,
      "step": 110510
    },
    {
      "epoch": 42.92,
      "learning_rate": 1.4277281553398059e-05,
      "loss": 0.0144,
      "step": 110520
    },
    {
      "epoch": 42.92,
      "learning_rate": 1.4276763754045308e-05,
      "loss": 0.0338,
      "step": 110530
    },
    {
      "epoch": 42.93,
      "learning_rate": 1.4276245954692558e-05,
      "loss": 0.084,
      "step": 110540
    },
    {
      "epoch": 42.93,
      "learning_rate": 1.4275728155339808e-05,
      "loss": 0.2384,
      "step": 110550
    },
    {
      "epoch": 42.94,
      "learning_rate": 1.4275210355987057e-05,
      "loss": 0.0006,
      "step": 110560
    },
    {
      "epoch": 42.94,
      "learning_rate": 1.4274692556634307e-05,
      "loss": 0.1204,
      "step": 110570
    },
    {
      "epoch": 42.94,
      "learning_rate": 1.4274174757281553e-05,
      "loss": 0.0492,
      "step": 110580
    },
    {
      "epoch": 42.95,
      "learning_rate": 1.4273656957928803e-05,
      "loss": 0.0316,
      "step": 110590
    },
    {
      "epoch": 42.95,
      "learning_rate": 1.4273139158576052e-05,
      "loss": 0.0651,
      "step": 110600
    },
    {
      "epoch": 42.96,
      "learning_rate": 1.4272621359223302e-05,
      "loss": 0.1892,
      "step": 110610
    },
    {
      "epoch": 42.96,
      "learning_rate": 1.4272103559870552e-05,
      "loss": 0.0738,
      "step": 110620
    },
    {
      "epoch": 42.96,
      "learning_rate": 1.4271585760517801e-05,
      "loss": 0.0482,
      "step": 110630
    },
    {
      "epoch": 42.97,
      "learning_rate": 1.4271067961165051e-05,
      "loss": 0.138,
      "step": 110640
    },
    {
      "epoch": 42.97,
      "learning_rate": 1.42705501618123e-05,
      "loss": 0.0794,
      "step": 110650
    },
    {
      "epoch": 42.97,
      "learning_rate": 1.4270032362459547e-05,
      "loss": 0.1457,
      "step": 110660
    },
    {
      "epoch": 42.98,
      "learning_rate": 1.4269514563106796e-05,
      "loss": 0.09,
      "step": 110670
    },
    {
      "epoch": 42.98,
      "learning_rate": 1.4268996763754046e-05,
      "loss": 0.0339,
      "step": 110680
    },
    {
      "epoch": 42.99,
      "learning_rate": 1.4268478964401296e-05,
      "loss": 0.1134,
      "step": 110690
    },
    {
      "epoch": 42.99,
      "learning_rate": 1.4267961165048545e-05,
      "loss": 0.0799,
      "step": 110700
    },
    {
      "epoch": 42.99,
      "learning_rate": 1.4267443365695795e-05,
      "loss": 0.2117,
      "step": 110710
    },
    {
      "epoch": 43.0,
      "learning_rate": 1.4266925566343044e-05,
      "loss": 0.0287,
      "step": 110720
    },
    {
      "epoch": 43.0,
      "eval_accuracy": 0.9469050894085282,
      "eval_loss": 0.3224157989025116,
      "eval_runtime": 8.2286,
      "eval_samples_per_second": 441.75,
      "eval_steps_per_second": 55.295,
      "step": 110725
    },
    {
      "epoch": 43.0,
      "learning_rate": 1.4266407766990294e-05,
      "loss": 0.1755,
      "step": 110730
    },
    {
      "epoch": 43.01,
      "learning_rate": 1.426588996763754e-05,
      "loss": 0.0774,
      "step": 110740
    },
    {
      "epoch": 43.01,
      "learning_rate": 1.426537216828479e-05,
      "loss": 0.0013,
      "step": 110750
    },
    {
      "epoch": 43.01,
      "learning_rate": 1.426485436893204e-05,
      "loss": 0.0362,
      "step": 110760
    },
    {
      "epoch": 43.02,
      "learning_rate": 1.426433656957929e-05,
      "loss": 0.1009,
      "step": 110770
    },
    {
      "epoch": 43.02,
      "learning_rate": 1.4263818770226539e-05,
      "loss": 0.0583,
      "step": 110780
    },
    {
      "epoch": 43.03,
      "learning_rate": 1.4263300970873788e-05,
      "loss": 0.0027,
      "step": 110790
    },
    {
      "epoch": 43.03,
      "learning_rate": 1.4262783171521038e-05,
      "loss": 0.1513,
      "step": 110800
    },
    {
      "epoch": 43.03,
      "learning_rate": 1.4262265372168288e-05,
      "loss": 0.1833,
      "step": 110810
    },
    {
      "epoch": 43.04,
      "learning_rate": 1.4261747572815534e-05,
      "loss": 0.1306,
      "step": 110820
    },
    {
      "epoch": 43.04,
      "learning_rate": 1.4261229773462783e-05,
      "loss": 0.1056,
      "step": 110830
    },
    {
      "epoch": 43.04,
      "learning_rate": 1.4260711974110033e-05,
      "loss": 0.0833,
      "step": 110840
    },
    {
      "epoch": 43.05,
      "learning_rate": 1.4260194174757283e-05,
      "loss": 0.0332,
      "step": 110850
    },
    {
      "epoch": 43.05,
      "learning_rate": 1.4259676375404532e-05,
      "loss": 0.0687,
      "step": 110860
    },
    {
      "epoch": 43.06,
      "learning_rate": 1.4259158576051782e-05,
      "loss": 0.0733,
      "step": 110870
    },
    {
      "epoch": 43.06,
      "learning_rate": 1.4258640776699032e-05,
      "loss": 0.0021,
      "step": 110880
    },
    {
      "epoch": 43.06,
      "learning_rate": 1.4258122977346281e-05,
      "loss": 0.0037,
      "step": 110890
    },
    {
      "epoch": 43.07,
      "learning_rate": 1.4257605177993527e-05,
      "loss": 0.0248,
      "step": 110900
    },
    {
      "epoch": 43.07,
      "learning_rate": 1.4257087378640777e-05,
      "loss": 0.0079,
      "step": 110910
    },
    {
      "epoch": 43.08,
      "learning_rate": 1.4256569579288027e-05,
      "loss": 0.037,
      "step": 110920
    },
    {
      "epoch": 43.08,
      "learning_rate": 1.4256051779935276e-05,
      "loss": 0.0703,
      "step": 110930
    },
    {
      "epoch": 43.08,
      "learning_rate": 1.4255533980582526e-05,
      "loss": 0.0452,
      "step": 110940
    },
    {
      "epoch": 43.09,
      "learning_rate": 1.4255016181229776e-05,
      "loss": 0.0452,
      "step": 110950
    },
    {
      "epoch": 43.09,
      "learning_rate": 1.4254498381877025e-05,
      "loss": 0.0781,
      "step": 110960
    },
    {
      "epoch": 43.1,
      "learning_rate": 1.4253980582524273e-05,
      "loss": 0.0027,
      "step": 110970
    },
    {
      "epoch": 43.1,
      "learning_rate": 1.4253462783171521e-05,
      "loss": 0.0635,
      "step": 110980
    },
    {
      "epoch": 43.1,
      "learning_rate": 1.425294498381877e-05,
      "loss": 0.0578,
      "step": 110990
    },
    {
      "epoch": 43.11,
      "learning_rate": 1.425242718446602e-05,
      "loss": 0.0858,
      "step": 111000
    },
    {
      "epoch": 43.11,
      "learning_rate": 1.425190938511327e-05,
      "loss": 0.0912,
      "step": 111010
    },
    {
      "epoch": 43.11,
      "learning_rate": 1.425139158576052e-05,
      "loss": 0.0547,
      "step": 111020
    },
    {
      "epoch": 43.12,
      "learning_rate": 1.4250873786407769e-05,
      "loss": 0.0836,
      "step": 111030
    },
    {
      "epoch": 43.12,
      "learning_rate": 1.4250355987055019e-05,
      "loss": 0.1152,
      "step": 111040
    },
    {
      "epoch": 43.13,
      "learning_rate": 1.4249838187702267e-05,
      "loss": 0.0006,
      "step": 111050
    },
    {
      "epoch": 43.13,
      "learning_rate": 1.4249320388349516e-05,
      "loss": 0.0921,
      "step": 111060
    },
    {
      "epoch": 43.13,
      "learning_rate": 1.4248802588996764e-05,
      "loss": 0.1454,
      "step": 111070
    },
    {
      "epoch": 43.14,
      "learning_rate": 1.4248284789644014e-05,
      "loss": 0.032,
      "step": 111080
    },
    {
      "epoch": 43.14,
      "learning_rate": 1.4247766990291263e-05,
      "loss": 0.1344,
      "step": 111090
    },
    {
      "epoch": 43.15,
      "learning_rate": 1.4247249190938513e-05,
      "loss": 0.1054,
      "step": 111100
    },
    {
      "epoch": 43.15,
      "learning_rate": 1.4246731391585763e-05,
      "loss": 0.0637,
      "step": 111110
    },
    {
      "epoch": 43.15,
      "learning_rate": 1.424621359223301e-05,
      "loss": 0.1001,
      "step": 111120
    },
    {
      "epoch": 43.16,
      "learning_rate": 1.424569579288026e-05,
      "loss": 0.1256,
      "step": 111130
    },
    {
      "epoch": 43.16,
      "learning_rate": 1.424517799352751e-05,
      "loss": 0.0185,
      "step": 111140
    },
    {
      "epoch": 43.17,
      "learning_rate": 1.4244660194174758e-05,
      "loss": 0.0715,
      "step": 111150
    },
    {
      "epoch": 43.17,
      "learning_rate": 1.4244142394822007e-05,
      "loss": 0.0727,
      "step": 111160
    },
    {
      "epoch": 43.17,
      "learning_rate": 1.4243624595469257e-05,
      "loss": 0.278,
      "step": 111170
    },
    {
      "epoch": 43.18,
      "learning_rate": 1.4243106796116507e-05,
      "loss": 0.1078,
      "step": 111180
    },
    {
      "epoch": 43.18,
      "learning_rate": 1.4242588996763756e-05,
      "loss": 0.0066,
      "step": 111190
    },
    {
      "epoch": 43.18,
      "learning_rate": 1.4242071197411004e-05,
      "loss": 0.0042,
      "step": 111200
    },
    {
      "epoch": 43.19,
      "learning_rate": 1.4241553398058254e-05,
      "loss": 0.1386,
      "step": 111210
    },
    {
      "epoch": 43.19,
      "learning_rate": 1.4241035598705503e-05,
      "loss": 0.0141,
      "step": 111220
    },
    {
      "epoch": 43.2,
      "learning_rate": 1.4240517799352751e-05,
      "loss": 0.0186,
      "step": 111230
    },
    {
      "epoch": 43.2,
      "learning_rate": 1.4240000000000001e-05,
      "loss": 0.0029,
      "step": 111240
    },
    {
      "epoch": 43.2,
      "learning_rate": 1.423948220064725e-05,
      "loss": 0.175,
      "step": 111250
    },
    {
      "epoch": 43.21,
      "learning_rate": 1.42389644012945e-05,
      "loss": 0.0408,
      "step": 111260
    },
    {
      "epoch": 43.21,
      "learning_rate": 1.4238446601941748e-05,
      "loss": 0.0476,
      "step": 111270
    },
    {
      "epoch": 43.22,
      "learning_rate": 1.4237928802588998e-05,
      "loss": 0.0341,
      "step": 111280
    },
    {
      "epoch": 43.22,
      "learning_rate": 1.4237411003236247e-05,
      "loss": 0.1651,
      "step": 111290
    },
    {
      "epoch": 43.22,
      "learning_rate": 1.4236893203883497e-05,
      "loss": 0.1432,
      "step": 111300
    },
    {
      "epoch": 43.23,
      "learning_rate": 1.4236375404530745e-05,
      "loss": 0.0465,
      "step": 111310
    },
    {
      "epoch": 43.23,
      "learning_rate": 1.4235857605177994e-05,
      "loss": 0.2105,
      "step": 111320
    },
    {
      "epoch": 43.23,
      "learning_rate": 1.4235339805825244e-05,
      "loss": 0.1131,
      "step": 111330
    },
    {
      "epoch": 43.24,
      "learning_rate": 1.4234822006472494e-05,
      "loss": 0.0727,
      "step": 111340
    },
    {
      "epoch": 43.24,
      "learning_rate": 1.4234304207119742e-05,
      "loss": 0.0228,
      "step": 111350
    },
    {
      "epoch": 43.25,
      "learning_rate": 1.4233786407766991e-05,
      "loss": 0.0015,
      "step": 111360
    },
    {
      "epoch": 43.25,
      "learning_rate": 1.423326860841424e-05,
      "loss": 0.1352,
      "step": 111370
    },
    {
      "epoch": 43.25,
      "learning_rate": 1.423275080906149e-05,
      "loss": 0.161,
      "step": 111380
    },
    {
      "epoch": 43.26,
      "learning_rate": 1.4232233009708738e-05,
      "loss": 0.2379,
      "step": 111390
    },
    {
      "epoch": 43.26,
      "learning_rate": 1.4231715210355988e-05,
      "loss": 0.0713,
      "step": 111400
    },
    {
      "epoch": 43.27,
      "learning_rate": 1.4231197411003238e-05,
      "loss": 0.021,
      "step": 111410
    },
    {
      "epoch": 43.27,
      "learning_rate": 1.4230679611650487e-05,
      "loss": 0.0351,
      "step": 111420
    },
    {
      "epoch": 43.27,
      "learning_rate": 1.4230161812297735e-05,
      "loss": 0.0637,
      "step": 111430
    },
    {
      "epoch": 43.28,
      "learning_rate": 1.4229644012944985e-05,
      "loss": 0.0437,
      "step": 111440
    },
    {
      "epoch": 43.28,
      "learning_rate": 1.4229126213592234e-05,
      "loss": 0.1011,
      "step": 111450
    },
    {
      "epoch": 43.29,
      "learning_rate": 1.4228608414239484e-05,
      "loss": 0.0055,
      "step": 111460
    },
    {
      "epoch": 43.29,
      "learning_rate": 1.4228090614886732e-05,
      "loss": 0.0802,
      "step": 111470
    },
    {
      "epoch": 43.29,
      "learning_rate": 1.4227572815533982e-05,
      "loss": 0.0858,
      "step": 111480
    },
    {
      "epoch": 43.3,
      "learning_rate": 1.4227055016181231e-05,
      "loss": 0.0177,
      "step": 111490
    },
    {
      "epoch": 43.3,
      "learning_rate": 1.4226537216828479e-05,
      "loss": 0.1429,
      "step": 111500
    },
    {
      "epoch": 43.3,
      "learning_rate": 1.4226019417475729e-05,
      "loss": 0.0273,
      "step": 111510
    },
    {
      "epoch": 43.31,
      "learning_rate": 1.4225501618122978e-05,
      "loss": 0.1542,
      "step": 111520
    },
    {
      "epoch": 43.31,
      "learning_rate": 1.4224983818770228e-05,
      "loss": 0.0346,
      "step": 111530
    },
    {
      "epoch": 43.32,
      "learning_rate": 1.4224466019417478e-05,
      "loss": 0.058,
      "step": 111540
    },
    {
      "epoch": 43.32,
      "learning_rate": 1.4223948220064725e-05,
      "loss": 0.2566,
      "step": 111550
    },
    {
      "epoch": 43.32,
      "learning_rate": 1.4223430420711975e-05,
      "loss": 0.0256,
      "step": 111560
    },
    {
      "epoch": 43.33,
      "learning_rate": 1.4222912621359225e-05,
      "loss": 0.0881,
      "step": 111570
    },
    {
      "epoch": 43.33,
      "learning_rate": 1.4222394822006473e-05,
      "loss": 0.0955,
      "step": 111580
    },
    {
      "epoch": 43.34,
      "learning_rate": 1.4221877022653722e-05,
      "loss": 0.0168,
      "step": 111590
    },
    {
      "epoch": 43.34,
      "learning_rate": 1.4221359223300972e-05,
      "loss": 0.1715,
      "step": 111600
    },
    {
      "epoch": 43.34,
      "learning_rate": 1.4220841423948222e-05,
      "loss": 0.0854,
      "step": 111610
    },
    {
      "epoch": 43.35,
      "learning_rate": 1.4220323624595471e-05,
      "loss": 0.2271,
      "step": 111620
    },
    {
      "epoch": 43.35,
      "learning_rate": 1.421980582524272e-05,
      "loss": 0.0229,
      "step": 111630
    },
    {
      "epoch": 43.36,
      "learning_rate": 1.4219288025889969e-05,
      "loss": 0.0886,
      "step": 111640
    },
    {
      "epoch": 43.36,
      "learning_rate": 1.4218770226537217e-05,
      "loss": 0.0807,
      "step": 111650
    },
    {
      "epoch": 43.36,
      "learning_rate": 1.4218252427184466e-05,
      "loss": 0.0946,
      "step": 111660
    },
    {
      "epoch": 43.37,
      "learning_rate": 1.4217734627831716e-05,
      "loss": 0.0908,
      "step": 111670
    },
    {
      "epoch": 43.37,
      "learning_rate": 1.4217216828478965e-05,
      "loss": 0.0714,
      "step": 111680
    },
    {
      "epoch": 43.37,
      "learning_rate": 1.4216699029126215e-05,
      "loss": 0.0825,
      "step": 111690
    },
    {
      "epoch": 43.38,
      "learning_rate": 1.4216181229773465e-05,
      "loss": 0.0664,
      "step": 111700
    },
    {
      "epoch": 43.38,
      "learning_rate": 1.4215663430420714e-05,
      "loss": 0.17,
      "step": 111710
    },
    {
      "epoch": 43.39,
      "learning_rate": 1.4215145631067962e-05,
      "loss": 0.0139,
      "step": 111720
    },
    {
      "epoch": 43.39,
      "learning_rate": 1.421462783171521e-05,
      "loss": 0.1287,
      "step": 111730
    },
    {
      "epoch": 43.39,
      "learning_rate": 1.421411003236246e-05,
      "loss": 0.126,
      "step": 111740
    },
    {
      "epoch": 43.4,
      "learning_rate": 1.421359223300971e-05,
      "loss": 0.107,
      "step": 111750
    },
    {
      "epoch": 43.4,
      "learning_rate": 1.4213074433656959e-05,
      "loss": 0.1543,
      "step": 111760
    },
    {
      "epoch": 43.41,
      "learning_rate": 1.4212556634304209e-05,
      "loss": 0.0516,
      "step": 111770
    },
    {
      "epoch": 43.41,
      "learning_rate": 1.4212038834951458e-05,
      "loss": 0.1143,
      "step": 111780
    },
    {
      "epoch": 43.41,
      "learning_rate": 1.4211521035598708e-05,
      "loss": 0.0764,
      "step": 111790
    },
    {
      "epoch": 43.42,
      "learning_rate": 1.4211003236245954e-05,
      "loss": 0.1029,
      "step": 111800
    },
    {
      "epoch": 43.42,
      "learning_rate": 1.4210485436893204e-05,
      "loss": 0.0721,
      "step": 111810
    },
    {
      "epoch": 43.43,
      "learning_rate": 1.4209967637540453e-05,
      "loss": 0.0454,
      "step": 111820
    },
    {
      "epoch": 43.43,
      "learning_rate": 1.4209449838187703e-05,
      "loss": 0.1338,
      "step": 111830
    },
    {
      "epoch": 43.43,
      "learning_rate": 1.4208932038834953e-05,
      "loss": 0.0141,
      "step": 111840
    },
    {
      "epoch": 43.44,
      "learning_rate": 1.4208414239482202e-05,
      "loss": 0.0211,
      "step": 111850
    },
    {
      "epoch": 43.44,
      "learning_rate": 1.4207896440129452e-05,
      "loss": 0.0348,
      "step": 111860
    },
    {
      "epoch": 43.44,
      "learning_rate": 1.4207378640776701e-05,
      "loss": 0.0446,
      "step": 111870
    },
    {
      "epoch": 43.45,
      "learning_rate": 1.4206860841423948e-05,
      "loss": 0.0291,
      "step": 111880
    },
    {
      "epoch": 43.45,
      "learning_rate": 1.4206343042071197e-05,
      "loss": 0.0198,
      "step": 111890
    },
    {
      "epoch": 43.46,
      "learning_rate": 1.4205825242718447e-05,
      "loss": 0.1119,
      "step": 111900
    },
    {
      "epoch": 43.46,
      "learning_rate": 1.4205307443365696e-05,
      "loss": 0.0273,
      "step": 111910
    },
    {
      "epoch": 43.46,
      "learning_rate": 1.4204789644012946e-05,
      "loss": 0.0047,
      "step": 111920
    },
    {
      "epoch": 43.47,
      "learning_rate": 1.4204271844660196e-05,
      "loss": 0.012,
      "step": 111930
    },
    {
      "epoch": 43.47,
      "learning_rate": 1.4203754045307445e-05,
      "loss": 0.0286,
      "step": 111940
    },
    {
      "epoch": 43.48,
      "learning_rate": 1.4203236245954695e-05,
      "loss": 0.1629,
      "step": 111950
    },
    {
      "epoch": 43.48,
      "learning_rate": 1.4202718446601941e-05,
      "loss": 0.0512,
      "step": 111960
    },
    {
      "epoch": 43.48,
      "learning_rate": 1.420220064724919e-05,
      "loss": 0.0764,
      "step": 111970
    },
    {
      "epoch": 43.49,
      "learning_rate": 1.420168284789644e-05,
      "loss": 0.0005,
      "step": 111980
    },
    {
      "epoch": 43.49,
      "learning_rate": 1.420116504854369e-05,
      "loss": 0.0116,
      "step": 111990
    },
    {
      "epoch": 43.5,
      "learning_rate": 1.420064724919094e-05,
      "loss": 0.1257,
      "step": 112000
    },
    {
      "epoch": 43.5,
      "learning_rate": 1.420012944983819e-05,
      "loss": 0.089,
      "step": 112010
    },
    {
      "epoch": 43.5,
      "learning_rate": 1.4199611650485439e-05,
      "loss": 0.0789,
      "step": 112020
    },
    {
      "epoch": 43.51,
      "learning_rate": 1.4199093851132689e-05,
      "loss": 0.1647,
      "step": 112030
    },
    {
      "epoch": 43.51,
      "learning_rate": 1.4198576051779935e-05,
      "loss": 0.0667,
      "step": 112040
    },
    {
      "epoch": 43.51,
      "learning_rate": 1.4198058252427184e-05,
      "loss": 0.1506,
      "step": 112050
    },
    {
      "epoch": 43.52,
      "learning_rate": 1.4197540453074434e-05,
      "loss": 0.0667,
      "step": 112060
    },
    {
      "epoch": 43.52,
      "learning_rate": 1.4197022653721684e-05,
      "loss": 0.0819,
      "step": 112070
    },
    {
      "epoch": 43.53,
      "learning_rate": 1.4196504854368933e-05,
      "loss": 0.2053,
      "step": 112080
    },
    {
      "epoch": 43.53,
      "learning_rate": 1.4195987055016183e-05,
      "loss": 0.1289,
      "step": 112090
    },
    {
      "epoch": 43.53,
      "learning_rate": 1.4195469255663432e-05,
      "loss": 0.1022,
      "step": 112100
    },
    {
      "epoch": 43.54,
      "learning_rate": 1.4194951456310682e-05,
      "loss": 0.1009,
      "step": 112110
    },
    {
      "epoch": 43.54,
      "learning_rate": 1.4194433656957932e-05,
      "loss": 0.1337,
      "step": 112120
    },
    {
      "epoch": 43.55,
      "learning_rate": 1.4193915857605178e-05,
      "loss": 0.0152,
      "step": 112130
    },
    {
      "epoch": 43.55,
      "learning_rate": 1.4193398058252428e-05,
      "loss": 0.1404,
      "step": 112140
    },
    {
      "epoch": 43.55,
      "learning_rate": 1.4192880258899677e-05,
      "loss": 0.1037,
      "step": 112150
    },
    {
      "epoch": 43.56,
      "learning_rate": 1.4192362459546927e-05,
      "loss": 0.0686,
      "step": 112160
    },
    {
      "epoch": 43.56,
      "learning_rate": 1.4191844660194176e-05,
      "loss": 0.0305,
      "step": 112170
    },
    {
      "epoch": 43.57,
      "learning_rate": 1.4191326860841426e-05,
      "loss": 0.0012,
      "step": 112180
    },
    {
      "epoch": 43.57,
      "learning_rate": 1.4190809061488676e-05,
      "loss": 0.075,
      "step": 112190
    },
    {
      "epoch": 43.57,
      "learning_rate": 1.4190291262135925e-05,
      "loss": 0.199,
      "step": 112200
    },
    {
      "epoch": 43.58,
      "learning_rate": 1.4189773462783171e-05,
      "loss": 0.1555,
      "step": 112210
    },
    {
      "epoch": 43.58,
      "learning_rate": 1.4189255663430421e-05,
      "loss": 0.0862,
      "step": 112220
    },
    {
      "epoch": 43.58,
      "learning_rate": 1.418873786407767e-05,
      "loss": 0.2112,
      "step": 112230
    },
    {
      "epoch": 43.59,
      "learning_rate": 1.418822006472492e-05,
      "loss": 0.0539,
      "step": 112240
    },
    {
      "epoch": 43.59,
      "learning_rate": 1.418770226537217e-05,
      "loss": 0.0137,
      "step": 112250
    },
    {
      "epoch": 43.6,
      "learning_rate": 1.418718446601942e-05,
      "loss": 0.0268,
      "step": 112260
    },
    {
      "epoch": 43.6,
      "learning_rate": 1.418666666666667e-05,
      "loss": 0.0303,
      "step": 112270
    },
    {
      "epoch": 43.6,
      "learning_rate": 1.4186148867313919e-05,
      "loss": 0.0217,
      "step": 112280
    },
    {
      "epoch": 43.61,
      "learning_rate": 1.4185631067961165e-05,
      "loss": 0.0929,
      "step": 112290
    },
    {
      "epoch": 43.61,
      "learning_rate": 1.4185113268608415e-05,
      "loss": 0.1031,
      "step": 112300
    },
    {
      "epoch": 43.62,
      "learning_rate": 1.4184595469255664e-05,
      "loss": 0.0526,
      "step": 112310
    },
    {
      "epoch": 43.62,
      "learning_rate": 1.4184077669902914e-05,
      "loss": 0.0237,
      "step": 112320
    },
    {
      "epoch": 43.62,
      "learning_rate": 1.4183559870550164e-05,
      "loss": 0.2315,
      "step": 112330
    },
    {
      "epoch": 43.63,
      "learning_rate": 1.4183042071197413e-05,
      "loss": 0.0712,
      "step": 112340
    },
    {
      "epoch": 43.63,
      "learning_rate": 1.4182524271844663e-05,
      "loss": 0.1062,
      "step": 112350
    },
    {
      "epoch": 43.63,
      "learning_rate": 1.4182006472491912e-05,
      "loss": 0.0526,
      "step": 112360
    },
    {
      "epoch": 43.64,
      "learning_rate": 1.4181488673139159e-05,
      "loss": 0.1261,
      "step": 112370
    },
    {
      "epoch": 43.64,
      "learning_rate": 1.4180970873786408e-05,
      "loss": 0.2217,
      "step": 112380
    },
    {
      "epoch": 43.65,
      "learning_rate": 1.4180453074433658e-05,
      "loss": 0.1111,
      "step": 112390
    },
    {
      "epoch": 43.65,
      "learning_rate": 1.4179935275080907e-05,
      "loss": 0.1611,
      "step": 112400
    },
    {
      "epoch": 43.65,
      "learning_rate": 1.4179417475728157e-05,
      "loss": 0.1256,
      "step": 112410
    },
    {
      "epoch": 43.66,
      "learning_rate": 1.4178899676375407e-05,
      "loss": 0.1157,
      "step": 112420
    },
    {
      "epoch": 43.66,
      "learning_rate": 1.4178381877022656e-05,
      "loss": 0.0919,
      "step": 112430
    },
    {
      "epoch": 43.67,
      "learning_rate": 1.4177864077669904e-05,
      "loss": 0.0885,
      "step": 112440
    },
    {
      "epoch": 43.67,
      "learning_rate": 1.4177346278317152e-05,
      "loss": 0.0106,
      "step": 112450
    },
    {
      "epoch": 43.67,
      "learning_rate": 1.4176828478964402e-05,
      "loss": 0.061,
      "step": 112460
    },
    {
      "epoch": 43.68,
      "learning_rate": 1.4176310679611651e-05,
      "loss": 0.0617,
      "step": 112470
    },
    {
      "epoch": 43.68,
      "learning_rate": 1.4175792880258901e-05,
      "loss": 0.0027,
      "step": 112480
    },
    {
      "epoch": 43.69,
      "learning_rate": 1.417527508090615e-05,
      "loss": 0.0721,
      "step": 112490
    },
    {
      "epoch": 43.69,
      "learning_rate": 1.41747572815534e-05,
      "loss": 0.0694,
      "step": 112500
    },
    {
      "epoch": 43.69,
      "learning_rate": 1.417423948220065e-05,
      "loss": 0.0511,
      "step": 112510
    },
    {
      "epoch": 43.7,
      "learning_rate": 1.4173721682847898e-05,
      "loss": 0.0768,
      "step": 112520
    },
    {
      "epoch": 43.7,
      "learning_rate": 1.4173203883495146e-05,
      "loss": 0.0235,
      "step": 112530
    },
    {
      "epoch": 43.7,
      "learning_rate": 1.4172686084142395e-05,
      "loss": 0.021,
      "step": 112540
    },
    {
      "epoch": 43.71,
      "learning_rate": 1.4172168284789645e-05,
      "loss": 0.1898,
      "step": 112550
    },
    {
      "epoch": 43.71,
      "learning_rate": 1.4171650485436895e-05,
      "loss": 0.1122,
      "step": 112560
    },
    {
      "epoch": 43.72,
      "learning_rate": 1.4171132686084144e-05,
      "loss": 0.1905,
      "step": 112570
    },
    {
      "epoch": 43.72,
      "learning_rate": 1.4170614886731394e-05,
      "loss": 0.3136,
      "step": 112580
    },
    {
      "epoch": 43.72,
      "learning_rate": 1.4170097087378642e-05,
      "loss": 0.0389,
      "step": 112590
    },
    {
      "epoch": 43.73,
      "learning_rate": 1.4169579288025891e-05,
      "loss": 0.0631,
      "step": 112600
    },
    {
      "epoch": 43.73,
      "learning_rate": 1.416906148867314e-05,
      "loss": 0.0663,
      "step": 112610
    },
    {
      "epoch": 43.74,
      "learning_rate": 1.4168543689320389e-05,
      "loss": 0.274,
      "step": 112620
    },
    {
      "epoch": 43.74,
      "learning_rate": 1.4168025889967638e-05,
      "loss": 0.1781,
      "step": 112630
    },
    {
      "epoch": 43.74,
      "learning_rate": 1.4167508090614888e-05,
      "loss": 0.0093,
      "step": 112640
    },
    {
      "epoch": 43.75,
      "learning_rate": 1.4166990291262138e-05,
      "loss": 0.0765,
      "step": 112650
    },
    {
      "epoch": 43.75,
      "learning_rate": 1.4166472491909387e-05,
      "loss": 0.0174,
      "step": 112660
    },
    {
      "epoch": 43.76,
      "learning_rate": 1.4165954692556635e-05,
      "loss": 0.0366,
      "step": 112670
    },
    {
      "epoch": 43.76,
      "learning_rate": 1.4165436893203885e-05,
      "loss": 0.0779,
      "step": 112680
    },
    {
      "epoch": 43.76,
      "learning_rate": 1.4164919093851135e-05,
      "loss": 0.1607,
      "step": 112690
    },
    {
      "epoch": 43.77,
      "learning_rate": 1.4164401294498382e-05,
      "loss": 0.1587,
      "step": 112700
    },
    {
      "epoch": 43.77,
      "learning_rate": 1.4163883495145632e-05,
      "loss": 0.0851,
      "step": 112710
    },
    {
      "epoch": 43.77,
      "learning_rate": 1.4163365695792882e-05,
      "loss": 0.0392,
      "step": 112720
    },
    {
      "epoch": 43.78,
      "learning_rate": 1.4162847896440131e-05,
      "loss": 0.021,
      "step": 112730
    },
    {
      "epoch": 43.78,
      "learning_rate": 1.416233009708738e-05,
      "loss": 0.0049,
      "step": 112740
    },
    {
      "epoch": 43.79,
      "learning_rate": 1.4161812297734629e-05,
      "loss": 0.1751,
      "step": 112750
    },
    {
      "epoch": 43.79,
      "learning_rate": 1.4161294498381878e-05,
      "loss": 0.0477,
      "step": 112760
    },
    {
      "epoch": 43.79,
      "learning_rate": 1.4160776699029128e-05,
      "loss": 0.0909,
      "step": 112770
    },
    {
      "epoch": 43.8,
      "learning_rate": 1.4160258899676376e-05,
      "loss": 0.0463,
      "step": 112780
    },
    {
      "epoch": 43.8,
      "learning_rate": 1.4159741100323626e-05,
      "loss": 0.1094,
      "step": 112790
    },
    {
      "epoch": 43.81,
      "learning_rate": 1.4159223300970875e-05,
      "loss": 0.0599,
      "step": 112800
    },
    {
      "epoch": 43.81,
      "learning_rate": 1.4158705501618125e-05,
      "loss": 0.0396,
      "step": 112810
    },
    {
      "epoch": 43.81,
      "learning_rate": 1.4158187702265373e-05,
      "loss": 0.1164,
      "step": 112820
    },
    {
      "epoch": 43.82,
      "learning_rate": 1.4157669902912622e-05,
      "loss": 0.0341,
      "step": 112830
    },
    {
      "epoch": 43.82,
      "learning_rate": 1.4157152103559872e-05,
      "loss": 0.0515,
      "step": 112840
    },
    {
      "epoch": 43.83,
      "learning_rate": 1.4156634304207122e-05,
      "loss": 0.187,
      "step": 112850
    },
    {
      "epoch": 43.83,
      "learning_rate": 1.415611650485437e-05,
      "loss": 0.0112,
      "step": 112860
    },
    {
      "epoch": 43.83,
      "learning_rate": 1.415559870550162e-05,
      "loss": 0.1974,
      "step": 112870
    },
    {
      "epoch": 43.84,
      "learning_rate": 1.4155080906148869e-05,
      "loss": 0.0152,
      "step": 112880
    },
    {
      "epoch": 43.84,
      "learning_rate": 1.4154563106796118e-05,
      "loss": 0.0226,
      "step": 112890
    },
    {
      "epoch": 43.84,
      "learning_rate": 1.4154045307443366e-05,
      "loss": 0.0748,
      "step": 112900
    },
    {
      "epoch": 43.85,
      "learning_rate": 1.4153527508090616e-05,
      "loss": 0.1564,
      "step": 112910
    },
    {
      "epoch": 43.85,
      "learning_rate": 1.4153009708737866e-05,
      "loss": 0.0366,
      "step": 112920
    },
    {
      "epoch": 43.86,
      "learning_rate": 1.4152491909385115e-05,
      "loss": 0.0575,
      "step": 112930
    },
    {
      "epoch": 43.86,
      "learning_rate": 1.4151974110032363e-05,
      "loss": 0.031,
      "step": 112940
    },
    {
      "epoch": 43.86,
      "learning_rate": 1.4151456310679613e-05,
      "loss": 0.197,
      "step": 112950
    },
    {
      "epoch": 43.87,
      "learning_rate": 1.4150938511326862e-05,
      "loss": 0.1475,
      "step": 112960
    },
    {
      "epoch": 43.87,
      "learning_rate": 1.415042071197411e-05,
      "loss": 0.0472,
      "step": 112970
    },
    {
      "epoch": 43.88,
      "learning_rate": 1.414990291262136e-05,
      "loss": 0.0223,
      "step": 112980
    },
    {
      "epoch": 43.88,
      "learning_rate": 1.414938511326861e-05,
      "loss": 0.0051,
      "step": 112990
    },
    {
      "epoch": 43.88,
      "learning_rate": 1.4148867313915859e-05,
      "loss": 0.12,
      "step": 113000
    },
    {
      "epoch": 43.89,
      "learning_rate": 1.4148349514563109e-05,
      "loss": 0.0016,
      "step": 113010
    },
    {
      "epoch": 43.89,
      "learning_rate": 1.4147831715210357e-05,
      "loss": 0.0949,
      "step": 113020
    },
    {
      "epoch": 43.9,
      "learning_rate": 1.4147313915857606e-05,
      "loss": 0.0623,
      "step": 113030
    },
    {
      "epoch": 43.9,
      "learning_rate": 1.4146796116504856e-05,
      "loss": 0.0139,
      "step": 113040
    },
    {
      "epoch": 43.9,
      "learning_rate": 1.4146278317152104e-05,
      "loss": 0.0452,
      "step": 113050
    },
    {
      "epoch": 43.91,
      "learning_rate": 1.4145760517799353e-05,
      "loss": 0.1457,
      "step": 113060
    },
    {
      "epoch": 43.91,
      "learning_rate": 1.4145242718446603e-05,
      "loss": 0.1514,
      "step": 113070
    },
    {
      "epoch": 43.91,
      "learning_rate": 1.4144724919093853e-05,
      "loss": 0.0865,
      "step": 113080
    },
    {
      "epoch": 43.92,
      "learning_rate": 1.4144207119741102e-05,
      "loss": 0.1589,
      "step": 113090
    },
    {
      "epoch": 43.92,
      "learning_rate": 1.414368932038835e-05,
      "loss": 0.2965,
      "step": 113100
    },
    {
      "epoch": 43.93,
      "learning_rate": 1.41431715210356e-05,
      "loss": 0.0699,
      "step": 113110
    },
    {
      "epoch": 43.93,
      "learning_rate": 1.4142653721682848e-05,
      "loss": 0.0106,
      "step": 113120
    },
    {
      "epoch": 43.93,
      "learning_rate": 1.4142135922330097e-05,
      "loss": 0.0676,
      "step": 113130
    },
    {
      "epoch": 43.94,
      "learning_rate": 1.4141618122977347e-05,
      "loss": 0.1224,
      "step": 113140
    },
    {
      "epoch": 43.94,
      "learning_rate": 1.4141100323624597e-05,
      "loss": 0.0325,
      "step": 113150
    },
    {
      "epoch": 43.95,
      "learning_rate": 1.4140582524271846e-05,
      "loss": 0.2108,
      "step": 113160
    },
    {
      "epoch": 43.95,
      "learning_rate": 1.4140064724919096e-05,
      "loss": 0.0361,
      "step": 113170
    },
    {
      "epoch": 43.95,
      "learning_rate": 1.4139546925566344e-05,
      "loss": 0.1428,
      "step": 113180
    },
    {
      "epoch": 43.96,
      "learning_rate": 1.4139029126213593e-05,
      "loss": 0.0689,
      "step": 113190
    },
    {
      "epoch": 43.96,
      "learning_rate": 1.4138511326860841e-05,
      "loss": 0.0957,
      "step": 113200
    },
    {
      "epoch": 43.97,
      "learning_rate": 1.4137993527508091e-05,
      "loss": 0.0075,
      "step": 113210
    },
    {
      "epoch": 43.97,
      "learning_rate": 1.413747572815534e-05,
      "loss": 0.2594,
      "step": 113220
    },
    {
      "epoch": 43.97,
      "learning_rate": 1.413695792880259e-05,
      "loss": 0.1103,
      "step": 113230
    },
    {
      "epoch": 43.98,
      "learning_rate": 1.413644012944984e-05,
      "loss": 0.0829,
      "step": 113240
    },
    {
      "epoch": 43.98,
      "learning_rate": 1.413592233009709e-05,
      "loss": 0.0528,
      "step": 113250
    },
    {
      "epoch": 43.98,
      "learning_rate": 1.4135404530744339e-05,
      "loss": 0.1021,
      "step": 113260
    },
    {
      "epoch": 43.99,
      "learning_rate": 1.4134886731391585e-05,
      "loss": 0.2524,
      "step": 113270
    },
    {
      "epoch": 43.99,
      "learning_rate": 1.4134368932038835e-05,
      "loss": 0.0952,
      "step": 113280
    },
    {
      "epoch": 44.0,
      "learning_rate": 1.4133851132686084e-05,
      "loss": 0.0713,
      "step": 113290
    },
    {
      "epoch": 44.0,
      "learning_rate": 1.4133333333333334e-05,
      "loss": 0.0384,
      "step": 113300
    },
    {
      "epoch": 44.0,
      "eval_accuracy": 0.9488308115543329,
      "eval_loss": 0.3085843324661255,
      "eval_runtime": 8.2092,
      "eval_samples_per_second": 442.797,
      "eval_steps_per_second": 55.426,
      "step": 113300
    },
    {
      "epoch": 44.0,
      "learning_rate": 1.4132815533980584e-05,
      "loss": 0.0434,
      "step": 113310
    },
    {
      "epoch": 44.01,
      "learning_rate": 1.4132297734627833e-05,
      "loss": 0.069,
      "step": 113320
    },
    {
      "epoch": 44.01,
      "learning_rate": 1.4131779935275083e-05,
      "loss": 0.1067,
      "step": 113330
    },
    {
      "epoch": 44.02,
      "learning_rate": 1.4131262135922333e-05,
      "loss": 0.0582,
      "step": 113340
    },
    {
      "epoch": 44.02,
      "learning_rate": 1.4130744336569579e-05,
      "loss": 0.0506,
      "step": 113350
    },
    {
      "epoch": 44.02,
      "learning_rate": 1.4130226537216828e-05,
      "loss": 0.0535,
      "step": 113360
    },
    {
      "epoch": 44.03,
      "learning_rate": 1.4129708737864078e-05,
      "loss": 0.1368,
      "step": 113370
    },
    {
      "epoch": 44.03,
      "learning_rate": 1.4129190938511328e-05,
      "loss": 0.2723,
      "step": 113380
    },
    {
      "epoch": 44.03,
      "learning_rate": 1.4128673139158577e-05,
      "loss": 0.0884,
      "step": 113390
    },
    {
      "epoch": 44.04,
      "learning_rate": 1.4128155339805827e-05,
      "loss": 0.0213,
      "step": 113400
    },
    {
      "epoch": 44.04,
      "learning_rate": 1.4127637540453077e-05,
      "loss": 0.038,
      "step": 113410
    },
    {
      "epoch": 44.05,
      "learning_rate": 1.4127119741100326e-05,
      "loss": 0.2221,
      "step": 113420
    },
    {
      "epoch": 44.05,
      "learning_rate": 1.4126601941747572e-05,
      "loss": 0.0761,
      "step": 113430
    },
    {
      "epoch": 44.05,
      "learning_rate": 1.4126084142394822e-05,
      "loss": 0.0666,
      "step": 113440
    },
    {
      "epoch": 44.06,
      "learning_rate": 1.4125566343042072e-05,
      "loss": 0.1397,
      "step": 113450
    },
    {
      "epoch": 44.06,
      "learning_rate": 1.4125048543689321e-05,
      "loss": 0.0349,
      "step": 113460
    },
    {
      "epoch": 44.07,
      "learning_rate": 1.412453074433657e-05,
      "loss": 0.0046,
      "step": 113470
    },
    {
      "epoch": 44.07,
      "learning_rate": 1.412401294498382e-05,
      "loss": 0.0532,
      "step": 113480
    },
    {
      "epoch": 44.07,
      "learning_rate": 1.412349514563107e-05,
      "loss": 0.1626,
      "step": 113490
    },
    {
      "epoch": 44.08,
      "learning_rate": 1.412297734627832e-05,
      "loss": 0.0062,
      "step": 113500
    },
    {
      "epoch": 44.08,
      "learning_rate": 1.4122459546925566e-05,
      "loss": 0.0983,
      "step": 113510
    },
    {
      "epoch": 44.09,
      "learning_rate": 1.4121941747572816e-05,
      "loss": 0.0535,
      "step": 113520
    },
    {
      "epoch": 44.09,
      "learning_rate": 1.4121423948220065e-05,
      "loss": 0.0605,
      "step": 113530
    },
    {
      "epoch": 44.09,
      "learning_rate": 1.4120906148867315e-05,
      "loss": 0.0757,
      "step": 113540
    },
    {
      "epoch": 44.1,
      "learning_rate": 1.4120388349514564e-05,
      "loss": 0.0291,
      "step": 113550
    },
    {
      "epoch": 44.1,
      "learning_rate": 1.4119870550161814e-05,
      "loss": 0.0417,
      "step": 113560
    },
    {
      "epoch": 44.1,
      "learning_rate": 1.4119352750809064e-05,
      "loss": 0.0725,
      "step": 113570
    },
    {
      "epoch": 44.11,
      "learning_rate": 1.4118834951456313e-05,
      "loss": 0.2589,
      "step": 113580
    },
    {
      "epoch": 44.11,
      "learning_rate": 1.411831715210356e-05,
      "loss": 0.1312,
      "step": 113590
    },
    {
      "epoch": 44.12,
      "learning_rate": 1.4117799352750809e-05,
      "loss": 0.0893,
      "step": 113600
    },
    {
      "epoch": 44.12,
      "learning_rate": 1.4117281553398059e-05,
      "loss": 0.1013,
      "step": 113610
    },
    {
      "epoch": 44.12,
      "learning_rate": 1.4116763754045308e-05,
      "loss": 0.0563,
      "step": 113620
    },
    {
      "epoch": 44.13,
      "learning_rate": 1.4116245954692558e-05,
      "loss": 0.0087,
      "step": 113630
    },
    {
      "epoch": 44.13,
      "learning_rate": 1.4115728155339808e-05,
      "loss": 0.1013,
      "step": 113640
    },
    {
      "epoch": 44.14,
      "learning_rate": 1.4115210355987057e-05,
      "loss": 0.0746,
      "step": 113650
    },
    {
      "epoch": 44.14,
      "learning_rate": 1.4114692556634307e-05,
      "loss": 0.073,
      "step": 113660
    },
    {
      "epoch": 44.14,
      "learning_rate": 1.4114174757281553e-05,
      "loss": 0.0642,
      "step": 113670
    },
    {
      "epoch": 44.15,
      "learning_rate": 1.4113656957928803e-05,
      "loss": 0.0349,
      "step": 113680
    },
    {
      "epoch": 44.15,
      "learning_rate": 1.4113139158576052e-05,
      "loss": 0.0355,
      "step": 113690
    },
    {
      "epoch": 44.16,
      "learning_rate": 1.4112621359223302e-05,
      "loss": 0.0387,
      "step": 113700
    },
    {
      "epoch": 44.16,
      "learning_rate": 1.4112103559870552e-05,
      "loss": 0.0082,
      "step": 113710
    },
    {
      "epoch": 44.16,
      "learning_rate": 1.4111585760517801e-05,
      "loss": 0.0051,
      "step": 113720
    },
    {
      "epoch": 44.17,
      "learning_rate": 1.411106796116505e-05,
      "loss": 0.0551,
      "step": 113730
    },
    {
      "epoch": 44.17,
      "learning_rate": 1.41105501618123e-05,
      "loss": 0.0827,
      "step": 113740
    },
    {
      "epoch": 44.17,
      "learning_rate": 1.4110032362459547e-05,
      "loss": 0.0745,
      "step": 113750
    },
    {
      "epoch": 44.18,
      "learning_rate": 1.4109514563106796e-05,
      "loss": 0.0107,
      "step": 113760
    },
    {
      "epoch": 44.18,
      "learning_rate": 1.4108996763754046e-05,
      "loss": 0.0711,
      "step": 113770
    },
    {
      "epoch": 44.19,
      "learning_rate": 1.4108478964401295e-05,
      "loss": 0.1428,
      "step": 113780
    },
    {
      "epoch": 44.19,
      "learning_rate": 1.4107961165048545e-05,
      "loss": 0.2399,
      "step": 113790
    },
    {
      "epoch": 44.19,
      "learning_rate": 1.4107443365695795e-05,
      "loss": 0.1052,
      "step": 113800
    },
    {
      "epoch": 44.2,
      "learning_rate": 1.4106925566343044e-05,
      "loss": 0.1278,
      "step": 113810
    },
    {
      "epoch": 44.2,
      "learning_rate": 1.4106407766990294e-05,
      "loss": 0.1221,
      "step": 113820
    },
    {
      "epoch": 44.21,
      "learning_rate": 1.4105889967637544e-05,
      "loss": 0.0763,
      "step": 113830
    },
    {
      "epoch": 44.21,
      "learning_rate": 1.410537216828479e-05,
      "loss": 0.2006,
      "step": 113840
    },
    {
      "epoch": 44.21,
      "learning_rate": 1.410485436893204e-05,
      "loss": 0.1274,
      "step": 113850
    },
    {
      "epoch": 44.22,
      "learning_rate": 1.4104336569579289e-05,
      "loss": 0.2367,
      "step": 113860
    },
    {
      "epoch": 44.22,
      "learning_rate": 1.4103818770226539e-05,
      "loss": 0.0236,
      "step": 113870
    },
    {
      "epoch": 44.23,
      "learning_rate": 1.4103300970873788e-05,
      "loss": 0.0081,
      "step": 113880
    },
    {
      "epoch": 44.23,
      "learning_rate": 1.4102783171521038e-05,
      "loss": 0.1016,
      "step": 113890
    },
    {
      "epoch": 44.23,
      "learning_rate": 1.4102265372168287e-05,
      "loss": 0.021,
      "step": 113900
    },
    {
      "epoch": 44.24,
      "learning_rate": 1.4101747572815535e-05,
      "loss": 0.0063,
      "step": 113910
    },
    {
      "epoch": 44.24,
      "learning_rate": 1.4101229773462783e-05,
      "loss": 0.0644,
      "step": 113920
    },
    {
      "epoch": 44.24,
      "learning_rate": 1.4100711974110033e-05,
      "loss": 0.1077,
      "step": 113930
    },
    {
      "epoch": 44.25,
      "learning_rate": 1.4100194174757283e-05,
      "loss": 0.0729,
      "step": 113940
    },
    {
      "epoch": 44.25,
      "learning_rate": 1.4099676375404532e-05,
      "loss": 0.0724,
      "step": 113950
    },
    {
      "epoch": 44.26,
      "learning_rate": 1.4099158576051782e-05,
      "loss": 0.0419,
      "step": 113960
    },
    {
      "epoch": 44.26,
      "learning_rate": 1.4098640776699031e-05,
      "loss": 0.0695,
      "step": 113970
    },
    {
      "epoch": 44.26,
      "learning_rate": 1.4098122977346281e-05,
      "loss": 0.1098,
      "step": 113980
    },
    {
      "epoch": 44.27,
      "learning_rate": 1.4097605177993529e-05,
      "loss": 0.0014,
      "step": 113990
    },
    {
      "epoch": 44.27,
      "learning_rate": 1.4097087378640777e-05,
      "loss": 0.1477,
      "step": 114000
    },
    {
      "epoch": 44.28,
      "learning_rate": 1.4096569579288026e-05,
      "loss": 0.0754,
      "step": 114010
    },
    {
      "epoch": 44.28,
      "learning_rate": 1.4096051779935276e-05,
      "loss": 0.167,
      "step": 114020
    },
    {
      "epoch": 44.28,
      "learning_rate": 1.4095533980582526e-05,
      "loss": 0.047,
      "step": 114030
    },
    {
      "epoch": 44.29,
      "learning_rate": 1.4095016181229775e-05,
      "loss": 0.0396,
      "step": 114040
    },
    {
      "epoch": 44.29,
      "learning_rate": 1.4094498381877025e-05,
      "loss": 0.0451,
      "step": 114050
    },
    {
      "epoch": 44.3,
      "learning_rate": 1.4093980582524273e-05,
      "loss": 0.0518,
      "step": 114060
    },
    {
      "epoch": 44.3,
      "learning_rate": 1.4093462783171523e-05,
      "loss": 0.0735,
      "step": 114070
    },
    {
      "epoch": 44.3,
      "learning_rate": 1.409294498381877e-05,
      "loss": 0.0418,
      "step": 114080
    },
    {
      "epoch": 44.31,
      "learning_rate": 1.409242718446602e-05,
      "loss": 0.0524,
      "step": 114090
    },
    {
      "epoch": 44.31,
      "learning_rate": 1.409190938511327e-05,
      "loss": 0.1514,
      "step": 114100
    },
    {
      "epoch": 44.31,
      "learning_rate": 1.409139158576052e-05,
      "loss": 0.1328,
      "step": 114110
    },
    {
      "epoch": 44.32,
      "learning_rate": 1.4090873786407769e-05,
      "loss": 0.0886,
      "step": 114120
    },
    {
      "epoch": 44.32,
      "learning_rate": 1.4090355987055019e-05,
      "loss": 0.1248,
      "step": 114130
    },
    {
      "epoch": 44.33,
      "learning_rate": 1.4089838187702266e-05,
      "loss": 0.0388,
      "step": 114140
    },
    {
      "epoch": 44.33,
      "learning_rate": 1.4089320388349516e-05,
      "loss": 0.0005,
      "step": 114150
    },
    {
      "epoch": 44.33,
      "learning_rate": 1.4088802588996764e-05,
      "loss": 0.1376,
      "step": 114160
    },
    {
      "epoch": 44.34,
      "learning_rate": 1.4088284789644014e-05,
      "loss": 0.1087,
      "step": 114170
    },
    {
      "epoch": 44.34,
      "learning_rate": 1.4087766990291263e-05,
      "loss": 0.0678,
      "step": 114180
    },
    {
      "epoch": 44.35,
      "learning_rate": 1.4087249190938513e-05,
      "loss": 0.1009,
      "step": 114190
    },
    {
      "epoch": 44.35,
      "learning_rate": 1.4086731391585762e-05,
      "loss": 0.0648,
      "step": 114200
    },
    {
      "epoch": 44.35,
      "learning_rate": 1.408621359223301e-05,
      "loss": 0.0849,
      "step": 114210
    },
    {
      "epoch": 44.36,
      "learning_rate": 1.408569579288026e-05,
      "loss": 0.0664,
      "step": 114220
    },
    {
      "epoch": 44.36,
      "learning_rate": 1.408517799352751e-05,
      "loss": 0.0206,
      "step": 114230
    },
    {
      "epoch": 44.37,
      "learning_rate": 1.4084660194174758e-05,
      "loss": 0.1469,
      "step": 114240
    },
    {
      "epoch": 44.37,
      "learning_rate": 1.4084142394822007e-05,
      "loss": 0.1996,
      "step": 114250
    },
    {
      "epoch": 44.37,
      "learning_rate": 1.4083624595469257e-05,
      "loss": 0.139,
      "step": 114260
    },
    {
      "epoch": 44.38,
      "learning_rate": 1.4083106796116506e-05,
      "loss": 0.0366,
      "step": 114270
    },
    {
      "epoch": 44.38,
      "learning_rate": 1.4082588996763756e-05,
      "loss": 0.0684,
      "step": 114280
    },
    {
      "epoch": 44.38,
      "learning_rate": 1.4082071197411004e-05,
      "loss": 0.0208,
      "step": 114290
    },
    {
      "epoch": 44.39,
      "learning_rate": 1.4081553398058254e-05,
      "loss": 0.1995,
      "step": 114300
    },
    {
      "epoch": 44.39,
      "learning_rate": 1.4081035598705503e-05,
      "loss": 0.068,
      "step": 114310
    },
    {
      "epoch": 44.4,
      "learning_rate": 1.4080517799352751e-05,
      "loss": 0.0285,
      "step": 114320
    },
    {
      "epoch": 44.4,
      "learning_rate": 1.408e-05,
      "loss": 0.0219,
      "step": 114330
    },
    {
      "epoch": 44.4,
      "learning_rate": 1.407948220064725e-05,
      "loss": 0.098,
      "step": 114340
    },
    {
      "epoch": 44.41,
      "learning_rate": 1.40789644012945e-05,
      "loss": 0.0554,
      "step": 114350
    },
    {
      "epoch": 44.41,
      "learning_rate": 1.4078446601941748e-05,
      "loss": 0.137,
      "step": 114360
    },
    {
      "epoch": 44.42,
      "learning_rate": 1.4077928802588997e-05,
      "loss": 0.0268,
      "step": 114370
    },
    {
      "epoch": 44.42,
      "learning_rate": 1.4077411003236247e-05,
      "loss": 0.0934,
      "step": 114380
    },
    {
      "epoch": 44.42,
      "learning_rate": 1.4076893203883497e-05,
      "loss": 0.079,
      "step": 114390
    },
    {
      "epoch": 44.43,
      "learning_rate": 1.4076375404530746e-05,
      "loss": 0.0471,
      "step": 114400
    },
    {
      "epoch": 44.43,
      "learning_rate": 1.4075857605177994e-05,
      "loss": 0.1365,
      "step": 114410
    },
    {
      "epoch": 44.43,
      "learning_rate": 1.4075339805825244e-05,
      "loss": 0.0496,
      "step": 114420
    },
    {
      "epoch": 44.44,
      "learning_rate": 1.4074822006472494e-05,
      "loss": 0.0592,
      "step": 114430
    },
    {
      "epoch": 44.44,
      "learning_rate": 1.4074304207119741e-05,
      "loss": 0.294,
      "step": 114440
    },
    {
      "epoch": 44.45,
      "learning_rate": 1.4073786407766991e-05,
      "loss": 0.0974,
      "step": 114450
    },
    {
      "epoch": 44.45,
      "learning_rate": 1.407326860841424e-05,
      "loss": 0.0113,
      "step": 114460
    },
    {
      "epoch": 44.45,
      "learning_rate": 1.407275080906149e-05,
      "loss": 0.0823,
      "step": 114470
    },
    {
      "epoch": 44.46,
      "learning_rate": 1.407223300970874e-05,
      "loss": 0.0593,
      "step": 114480
    },
    {
      "epoch": 44.46,
      "learning_rate": 1.4071715210355988e-05,
      "loss": 0.0337,
      "step": 114490
    },
    {
      "epoch": 44.47,
      "learning_rate": 1.4071197411003237e-05,
      "loss": 0.1532,
      "step": 114500
    },
    {
      "epoch": 44.47,
      "learning_rate": 1.4070679611650487e-05,
      "loss": 0.0775,
      "step": 114510
    },
    {
      "epoch": 44.47,
      "learning_rate": 1.4070161812297735e-05,
      "loss": 0.1341,
      "step": 114520
    },
    {
      "epoch": 44.48,
      "learning_rate": 1.4069644012944985e-05,
      "loss": 0.1074,
      "step": 114530
    },
    {
      "epoch": 44.48,
      "learning_rate": 1.4069126213592234e-05,
      "loss": 0.0266,
      "step": 114540
    },
    {
      "epoch": 44.49,
      "learning_rate": 1.4068608414239484e-05,
      "loss": 0.0532,
      "step": 114550
    },
    {
      "epoch": 44.49,
      "learning_rate": 1.4068090614886733e-05,
      "loss": 0.1694,
      "step": 114560
    },
    {
      "epoch": 44.49,
      "learning_rate": 1.4067572815533981e-05,
      "loss": 0.2494,
      "step": 114570
    },
    {
      "epoch": 44.5,
      "learning_rate": 1.4067055016181231e-05,
      "loss": 0.1343,
      "step": 114580
    },
    {
      "epoch": 44.5,
      "learning_rate": 1.4066537216828479e-05,
      "loss": 0.0549,
      "step": 114590
    },
    {
      "epoch": 44.5,
      "learning_rate": 1.4066019417475729e-05,
      "loss": 0.1193,
      "step": 114600
    },
    {
      "epoch": 44.51,
      "learning_rate": 1.4065501618122978e-05,
      "loss": 0.0503,
      "step": 114610
    },
    {
      "epoch": 44.51,
      "learning_rate": 1.4064983818770228e-05,
      "loss": 0.0159,
      "step": 114620
    },
    {
      "epoch": 44.52,
      "learning_rate": 1.4064466019417477e-05,
      "loss": 0.0855,
      "step": 114630
    },
    {
      "epoch": 44.52,
      "learning_rate": 1.4063948220064727e-05,
      "loss": 0.041,
      "step": 114640
    },
    {
      "epoch": 44.52,
      "learning_rate": 1.4063430420711975e-05,
      "loss": 0.1439,
      "step": 114650
    },
    {
      "epoch": 44.53,
      "learning_rate": 1.4062912621359225e-05,
      "loss": 0.0136,
      "step": 114660
    },
    {
      "epoch": 44.53,
      "learning_rate": 1.4062394822006472e-05,
      "loss": 0.0813,
      "step": 114670
    },
    {
      "epoch": 44.54,
      "learning_rate": 1.4061877022653722e-05,
      "loss": 0.0849,
      "step": 114680
    },
    {
      "epoch": 44.54,
      "learning_rate": 1.4061359223300972e-05,
      "loss": 0.0492,
      "step": 114690
    },
    {
      "epoch": 44.54,
      "learning_rate": 1.4060841423948221e-05,
      "loss": 0.0918,
      "step": 114700
    },
    {
      "epoch": 44.55,
      "learning_rate": 1.4060323624595471e-05,
      "loss": 0.1131,
      "step": 114710
    },
    {
      "epoch": 44.55,
      "learning_rate": 1.405980582524272e-05,
      "loss": 0.1241,
      "step": 114720
    },
    {
      "epoch": 44.56,
      "learning_rate": 1.4059288025889968e-05,
      "loss": 0.1851,
      "step": 114730
    },
    {
      "epoch": 44.56,
      "learning_rate": 1.4058770226537216e-05,
      "loss": 0.0527,
      "step": 114740
    },
    {
      "epoch": 44.56,
      "learning_rate": 1.4058252427184466e-05,
      "loss": 0.0111,
      "step": 114750
    },
    {
      "epoch": 44.57,
      "learning_rate": 1.4057734627831716e-05,
      "loss": 0.125,
      "step": 114760
    },
    {
      "epoch": 44.57,
      "learning_rate": 1.4057216828478965e-05,
      "loss": 0.0573,
      "step": 114770
    },
    {
      "epoch": 44.57,
      "learning_rate": 1.4056699029126215e-05,
      "loss": 0.1161,
      "step": 114780
    },
    {
      "epoch": 44.58,
      "learning_rate": 1.4056181229773465e-05,
      "loss": 0.1339,
      "step": 114790
    },
    {
      "epoch": 44.58,
      "learning_rate": 1.4055663430420714e-05,
      "loss": 0.1672,
      "step": 114800
    },
    {
      "epoch": 44.59,
      "learning_rate": 1.4055145631067962e-05,
      "loss": 0.0994,
      "step": 114810
    },
    {
      "epoch": 44.59,
      "learning_rate": 1.405462783171521e-05,
      "loss": 0.0948,
      "step": 114820
    },
    {
      "epoch": 44.59,
      "learning_rate": 1.405411003236246e-05,
      "loss": 0.0943,
      "step": 114830
    },
    {
      "epoch": 44.6,
      "learning_rate": 1.405359223300971e-05,
      "loss": 0.0546,
      "step": 114840
    },
    {
      "epoch": 44.6,
      "learning_rate": 1.4053074433656959e-05,
      "loss": 0.0712,
      "step": 114850
    },
    {
      "epoch": 44.61,
      "learning_rate": 1.4052556634304208e-05,
      "loss": 0.0551,
      "step": 114860
    },
    {
      "epoch": 44.61,
      "learning_rate": 1.4052038834951458e-05,
      "loss": 0.0426,
      "step": 114870
    },
    {
      "epoch": 44.61,
      "learning_rate": 1.4051521035598708e-05,
      "loss": 0.0764,
      "step": 114880
    },
    {
      "epoch": 44.62,
      "learning_rate": 1.4051003236245954e-05,
      "loss": 0.1159,
      "step": 114890
    },
    {
      "epoch": 44.62,
      "learning_rate": 1.4050485436893204e-05,
      "loss": 0.1709,
      "step": 114900
    },
    {
      "epoch": 44.63,
      "learning_rate": 1.4049967637540453e-05,
      "loss": 0.0483,
      "step": 114910
    },
    {
      "epoch": 44.63,
      "learning_rate": 1.4049449838187703e-05,
      "loss": 0.0699,
      "step": 114920
    },
    {
      "epoch": 44.63,
      "learning_rate": 1.4048932038834952e-05,
      "loss": 0.0514,
      "step": 114930
    },
    {
      "epoch": 44.64,
      "learning_rate": 1.4048414239482202e-05,
      "loss": 0.1164,
      "step": 114940
    },
    {
      "epoch": 44.64,
      "learning_rate": 1.4047896440129452e-05,
      "loss": 0.0034,
      "step": 114950
    },
    {
      "epoch": 44.64,
      "learning_rate": 1.4047378640776701e-05,
      "loss": 0.1484,
      "step": 114960
    },
    {
      "epoch": 44.65,
      "learning_rate": 1.4046860841423951e-05,
      "loss": 0.1233,
      "step": 114970
    },
    {
      "epoch": 44.65,
      "learning_rate": 1.4046343042071197e-05,
      "loss": 0.0892,
      "step": 114980
    },
    {
      "epoch": 44.66,
      "learning_rate": 1.4045825242718447e-05,
      "loss": 0.1505,
      "step": 114990
    },
    {
      "epoch": 44.66,
      "learning_rate": 1.4045307443365696e-05,
      "loss": 0.0822,
      "step": 115000
    },
    {
      "epoch": 44.66,
      "learning_rate": 1.4044789644012946e-05,
      "loss": 0.0719,
      "step": 115010
    },
    {
      "epoch": 44.67,
      "learning_rate": 1.4044271844660196e-05,
      "loss": 0.0897,
      "step": 115020
    },
    {
      "epoch": 44.67,
      "learning_rate": 1.4043754045307445e-05,
      "loss": 0.0033,
      "step": 115030
    },
    {
      "epoch": 44.68,
      "learning_rate": 1.4043236245954695e-05,
      "loss": 0.1836,
      "step": 115040
    },
    {
      "epoch": 44.68,
      "learning_rate": 1.4042718446601944e-05,
      "loss": 0.0395,
      "step": 115050
    },
    {
      "epoch": 44.68,
      "learning_rate": 1.404220064724919e-05,
      "loss": 0.174,
      "step": 115060
    },
    {
      "epoch": 44.69,
      "learning_rate": 1.404168284789644e-05,
      "loss": 0.0263,
      "step": 115070
    },
    {
      "epoch": 44.69,
      "learning_rate": 1.404116504854369e-05,
      "loss": 0.0434,
      "step": 115080
    },
    {
      "epoch": 44.7,
      "learning_rate": 1.404064724919094e-05,
      "loss": 0.0721,
      "step": 115090
    },
    {
      "epoch": 44.7,
      "learning_rate": 1.4040129449838189e-05,
      "loss": 0.0516,
      "step": 115100
    },
    {
      "epoch": 44.7,
      "learning_rate": 1.4039611650485439e-05,
      "loss": 0.0644,
      "step": 115110
    },
    {
      "epoch": 44.71,
      "learning_rate": 1.4039093851132688e-05,
      "loss": 0.1187,
      "step": 115120
    },
    {
      "epoch": 44.71,
      "learning_rate": 1.4038576051779938e-05,
      "loss": 0.2456,
      "step": 115130
    },
    {
      "epoch": 44.71,
      "learning_rate": 1.4038058252427184e-05,
      "loss": 0.2054,
      "step": 115140
    },
    {
      "epoch": 44.72,
      "learning_rate": 1.4037540453074434e-05,
      "loss": 0.1093,
      "step": 115150
    },
    {
      "epoch": 44.72,
      "learning_rate": 1.4037022653721683e-05,
      "loss": 0.014,
      "step": 115160
    },
    {
      "epoch": 44.73,
      "learning_rate": 1.4036504854368933e-05,
      "loss": 0.0013,
      "step": 115170
    },
    {
      "epoch": 44.73,
      "learning_rate": 1.4035987055016183e-05,
      "loss": 0.0542,
      "step": 115180
    },
    {
      "epoch": 44.73,
      "learning_rate": 1.4035469255663432e-05,
      "loss": 0.0375,
      "step": 115190
    },
    {
      "epoch": 44.74,
      "learning_rate": 1.4034951456310682e-05,
      "loss": 0.0019,
      "step": 115200
    },
    {
      "epoch": 44.74,
      "learning_rate": 1.4034433656957932e-05,
      "loss": 0.1169,
      "step": 115210
    },
    {
      "epoch": 44.75,
      "learning_rate": 1.4033915857605178e-05,
      "loss": 0.2151,
      "step": 115220
    },
    {
      "epoch": 44.75,
      "learning_rate": 1.4033398058252427e-05,
      "loss": 0.1614,
      "step": 115230
    },
    {
      "epoch": 44.75,
      "learning_rate": 1.4032880258899677e-05,
      "loss": 0.0432,
      "step": 115240
    },
    {
      "epoch": 44.76,
      "learning_rate": 1.4032362459546927e-05,
      "loss": 0.0175,
      "step": 115250
    },
    {
      "epoch": 44.76,
      "learning_rate": 1.4031844660194176e-05,
      "loss": 0.0672,
      "step": 115260
    },
    {
      "epoch": 44.77,
      "learning_rate": 1.4031326860841426e-05,
      "loss": 0.0391,
      "step": 115270
    },
    {
      "epoch": 44.77,
      "learning_rate": 1.4030809061488675e-05,
      "loss": 0.1352,
      "step": 115280
    },
    {
      "epoch": 44.77,
      "learning_rate": 1.4030291262135925e-05,
      "loss": 0.0438,
      "step": 115290
    },
    {
      "epoch": 44.78,
      "learning_rate": 1.4029773462783171e-05,
      "loss": 0.0168,
      "step": 115300
    },
    {
      "epoch": 44.78,
      "learning_rate": 1.4029255663430421e-05,
      "loss": 0.086,
      "step": 115310
    },
    {
      "epoch": 44.78,
      "learning_rate": 1.402873786407767e-05,
      "loss": 0.0712,
      "step": 115320
    },
    {
      "epoch": 44.79,
      "learning_rate": 1.402822006472492e-05,
      "loss": 0.0866,
      "step": 115330
    },
    {
      "epoch": 44.79,
      "learning_rate": 1.402770226537217e-05,
      "loss": 0.117,
      "step": 115340
    },
    {
      "epoch": 44.8,
      "learning_rate": 1.402718446601942e-05,
      "loss": 0.0535,
      "step": 115350
    },
    {
      "epoch": 44.8,
      "learning_rate": 1.4026666666666669e-05,
      "loss": 0.0582,
      "step": 115360
    },
    {
      "epoch": 44.8,
      "learning_rate": 1.4026148867313919e-05,
      "loss": 0.0776,
      "step": 115370
    },
    {
      "epoch": 44.81,
      "learning_rate": 1.4025631067961165e-05,
      "loss": 0.1571,
      "step": 115380
    },
    {
      "epoch": 44.81,
      "learning_rate": 1.4025113268608414e-05,
      "loss": 0.2169,
      "step": 115390
    },
    {
      "epoch": 44.82,
      "learning_rate": 1.4024595469255664e-05,
      "loss": 0.163,
      "step": 115400
    },
    {
      "epoch": 44.82,
      "learning_rate": 1.4024077669902914e-05,
      "loss": 0.1046,
      "step": 115410
    },
    {
      "epoch": 44.82,
      "learning_rate": 1.4023559870550163e-05,
      "loss": 0.0401,
      "step": 115420
    },
    {
      "epoch": 44.83,
      "learning_rate": 1.4023042071197413e-05,
      "loss": 0.0495,
      "step": 115430
    },
    {
      "epoch": 44.83,
      "learning_rate": 1.4022524271844663e-05,
      "loss": 0.0958,
      "step": 115440
    },
    {
      "epoch": 44.83,
      "learning_rate": 1.4022006472491912e-05,
      "loss": 0.0557,
      "step": 115450
    },
    {
      "epoch": 44.84,
      "learning_rate": 1.4021488673139158e-05,
      "loss": 0.1445,
      "step": 115460
    },
    {
      "epoch": 44.84,
      "learning_rate": 1.4020970873786408e-05,
      "loss": 0.0565,
      "step": 115470
    },
    {
      "epoch": 44.85,
      "learning_rate": 1.4020453074433658e-05,
      "loss": 0.1201,
      "step": 115480
    },
    {
      "epoch": 44.85,
      "learning_rate": 1.4019935275080907e-05,
      "loss": 0.0021,
      "step": 115490
    },
    {
      "epoch": 44.85,
      "learning_rate": 1.4019417475728157e-05,
      "loss": 0.2078,
      "step": 115500
    },
    {
      "epoch": 44.86,
      "learning_rate": 1.4018899676375407e-05,
      "loss": 0.1895,
      "step": 115510
    },
    {
      "epoch": 44.86,
      "learning_rate": 1.4018381877022656e-05,
      "loss": 0.123,
      "step": 115520
    },
    {
      "epoch": 44.87,
      "learning_rate": 1.4017864077669904e-05,
      "loss": 0.1512,
      "step": 115530
    },
    {
      "epoch": 44.87,
      "learning_rate": 1.4017346278317154e-05,
      "loss": 0.0336,
      "step": 115540
    },
    {
      "epoch": 44.87,
      "learning_rate": 1.4016828478964402e-05,
      "loss": 0.0686,
      "step": 115550
    },
    {
      "epoch": 44.88,
      "learning_rate": 1.4016310679611651e-05,
      "loss": 0.0505,
      "step": 115560
    },
    {
      "epoch": 44.88,
      "learning_rate": 1.40157928802589e-05,
      "loss": 0.1376,
      "step": 115570
    },
    {
      "epoch": 44.89,
      "learning_rate": 1.401527508090615e-05,
      "loss": 0.1276,
      "step": 115580
    },
    {
      "epoch": 44.89,
      "learning_rate": 1.40147572815534e-05,
      "loss": 0.099,
      "step": 115590
    },
    {
      "epoch": 44.89,
      "learning_rate": 1.401423948220065e-05,
      "loss": 0.0373,
      "step": 115600
    },
    {
      "epoch": 44.9,
      "learning_rate": 1.4013721682847898e-05,
      "loss": 0.0718,
      "step": 115610
    },
    {
      "epoch": 44.9,
      "learning_rate": 1.4013203883495147e-05,
      "loss": 0.1168,
      "step": 115620
    },
    {
      "epoch": 44.9,
      "learning_rate": 1.4012686084142395e-05,
      "loss": 0.1562,
      "step": 115630
    },
    {
      "epoch": 44.91,
      "learning_rate": 1.4012168284789645e-05,
      "loss": 0.0809,
      "step": 115640
    },
    {
      "epoch": 44.91,
      "learning_rate": 1.4011650485436894e-05,
      "loss": 0.1116,
      "step": 115650
    },
    {
      "epoch": 44.92,
      "learning_rate": 1.4011132686084144e-05,
      "loss": 0.0627,
      "step": 115660
    },
    {
      "epoch": 44.92,
      "learning_rate": 1.4010614886731394e-05,
      "loss": 0.0774,
      "step": 115670
    },
    {
      "epoch": 44.92,
      "learning_rate": 1.4010097087378642e-05,
      "loss": 0.112,
      "step": 115680
    },
    {
      "epoch": 44.93,
      "learning_rate": 1.4009579288025891e-05,
      "loss": 0.0716,
      "step": 115690
    },
    {
      "epoch": 44.93,
      "learning_rate": 1.400906148867314e-05,
      "loss": 0.1165,
      "step": 115700
    },
    {
      "epoch": 44.94,
      "learning_rate": 1.4008543689320389e-05,
      "loss": 0.1913,
      "step": 115710
    },
    {
      "epoch": 44.94,
      "learning_rate": 1.4008025889967638e-05,
      "loss": 0.0973,
      "step": 115720
    },
    {
      "epoch": 44.94,
      "learning_rate": 1.4007508090614888e-05,
      "loss": 0.0234,
      "step": 115730
    },
    {
      "epoch": 44.95,
      "learning_rate": 1.4006990291262138e-05,
      "loss": 0.1035,
      "step": 115740
    },
    {
      "epoch": 44.95,
      "learning_rate": 1.4006472491909387e-05,
      "loss": 0.0044,
      "step": 115750
    },
    {
      "epoch": 44.96,
      "learning_rate": 1.4005954692556635e-05,
      "loss": 0.0664,
      "step": 115760
    },
    {
      "epoch": 44.96,
      "learning_rate": 1.4005436893203885e-05,
      "loss": 0.0694,
      "step": 115770
    },
    {
      "epoch": 44.96,
      "learning_rate": 1.4004919093851134e-05,
      "loss": 0.0772,
      "step": 115780
    },
    {
      "epoch": 44.97,
      "learning_rate": 1.4004401294498382e-05,
      "loss": 0.1597,
      "step": 115790
    },
    {
      "epoch": 44.97,
      "learning_rate": 1.4003883495145632e-05,
      "loss": 0.0028,
      "step": 115800
    },
    {
      "epoch": 44.97,
      "learning_rate": 1.4003365695792882e-05,
      "loss": 0.0494,
      "step": 115810
    },
    {
      "epoch": 44.98,
      "learning_rate": 1.4002847896440131e-05,
      "loss": 0.0243,
      "step": 115820
    },
    {
      "epoch": 44.98,
      "learning_rate": 1.4002330097087379e-05,
      "loss": 0.012,
      "step": 115830
    },
    {
      "epoch": 44.99,
      "learning_rate": 1.4001812297734629e-05,
      "loss": 0.1421,
      "step": 115840
    },
    {
      "epoch": 44.99,
      "learning_rate": 1.4001294498381878e-05,
      "loss": 0.0327,
      "step": 115850
    },
    {
      "epoch": 44.99,
      "learning_rate": 1.4000776699029128e-05,
      "loss": 0.0467,
      "step": 115860
    },
    {
      "epoch": 45.0,
      "learning_rate": 1.4000258899676376e-05,
      "loss": 0.1214,
      "step": 115870
    },
    {
      "epoch": 45.0,
      "eval_accuracy": 0.9493810178817056,
      "eval_loss": 0.31446582078933716,
      "eval_runtime": 8.2323,
      "eval_samples_per_second": 441.553,
      "eval_steps_per_second": 55.27,
      "step": 115875
    },
    {
      "epoch": 45.0,
      "learning_rate": 1.3999741100323625e-05,
      "loss": 0.0036,
      "step": 115880
    },
    {
      "epoch": 45.01,
      "learning_rate": 1.3999223300970875e-05,
      "loss": 0.004,
      "step": 115890
    },
    {
      "epoch": 45.01,
      "learning_rate": 1.3998705501618125e-05,
      "loss": 0.0632,
      "step": 115900
    },
    {
      "epoch": 45.01,
      "learning_rate": 1.3998187702265373e-05,
      "loss": 0.0521,
      "step": 115910
    },
    {
      "epoch": 45.02,
      "learning_rate": 1.3997669902912622e-05,
      "loss": 0.1009,
      "step": 115920
    },
    {
      "epoch": 45.02,
      "learning_rate": 1.3997152103559872e-05,
      "loss": 0.1257,
      "step": 115930
    },
    {
      "epoch": 45.03,
      "learning_rate": 1.3996634304207121e-05,
      "loss": 0.0505,
      "step": 115940
    },
    {
      "epoch": 45.03,
      "learning_rate": 1.399611650485437e-05,
      "loss": 0.0768,
      "step": 115950
    },
    {
      "epoch": 45.03,
      "learning_rate": 1.3995598705501619e-05,
      "loss": 0.1131,
      "step": 115960
    },
    {
      "epoch": 45.04,
      "learning_rate": 1.3995080906148869e-05,
      "loss": 0.001,
      "step": 115970
    },
    {
      "epoch": 45.04,
      "learning_rate": 1.3994563106796118e-05,
      "loss": 0.0504,
      "step": 115980
    },
    {
      "epoch": 45.04,
      "learning_rate": 1.3994045307443366e-05,
      "loss": 0.1015,
      "step": 115990
    },
    {
      "epoch": 45.05,
      "learning_rate": 1.3993527508090616e-05,
      "loss": 0.1503,
      "step": 116000
    },
    {
      "epoch": 45.05,
      "learning_rate": 1.3993009708737865e-05,
      "loss": 0.005,
      "step": 116010
    },
    {
      "epoch": 45.06,
      "learning_rate": 1.3992491909385115e-05,
      "loss": 0.0009,
      "step": 116020
    },
    {
      "epoch": 45.06,
      "learning_rate": 1.3991974110032363e-05,
      "loss": 0.1056,
      "step": 116030
    },
    {
      "epoch": 45.06,
      "learning_rate": 1.3991456310679613e-05,
      "loss": 0.21,
      "step": 116040
    },
    {
      "epoch": 45.07,
      "learning_rate": 1.3990938511326862e-05,
      "loss": 0.0481,
      "step": 116050
    },
    {
      "epoch": 45.07,
      "learning_rate": 1.399042071197411e-05,
      "loss": 0.0307,
      "step": 116060
    },
    {
      "epoch": 45.08,
      "learning_rate": 1.398990291262136e-05,
      "loss": 0.0838,
      "step": 116070
    },
    {
      "epoch": 45.08,
      "learning_rate": 1.398938511326861e-05,
      "loss": 0.0826,
      "step": 116080
    },
    {
      "epoch": 45.08,
      "learning_rate": 1.3988867313915859e-05,
      "loss": 0.0999,
      "step": 116090
    },
    {
      "epoch": 45.09,
      "learning_rate": 1.3988349514563109e-05,
      "loss": 0.1512,
      "step": 116100
    },
    {
      "epoch": 45.09,
      "learning_rate": 1.3987831715210358e-05,
      "loss": 0.0328,
      "step": 116110
    },
    {
      "epoch": 45.1,
      "learning_rate": 1.3987313915857606e-05,
      "loss": 0.024,
      "step": 116120
    },
    {
      "epoch": 45.1,
      "learning_rate": 1.3986796116504856e-05,
      "loss": 0.063,
      "step": 116130
    },
    {
      "epoch": 45.1,
      "learning_rate": 1.3986278317152104e-05,
      "loss": 0.0032,
      "step": 116140
    },
    {
      "epoch": 45.11,
      "learning_rate": 1.3985760517799353e-05,
      "loss": 0.0223,
      "step": 116150
    },
    {
      "epoch": 45.11,
      "learning_rate": 1.3985242718446603e-05,
      "loss": 0.0875,
      "step": 116160
    },
    {
      "epoch": 45.11,
      "learning_rate": 1.3984724919093853e-05,
      "loss": 0.0738,
      "step": 116170
    },
    {
      "epoch": 45.12,
      "learning_rate": 1.3984207119741102e-05,
      "loss": 0.0656,
      "step": 116180
    },
    {
      "epoch": 45.12,
      "learning_rate": 1.3983689320388352e-05,
      "loss": 0.0983,
      "step": 116190
    },
    {
      "epoch": 45.13,
      "learning_rate": 1.39831715210356e-05,
      "loss": 0.0674,
      "step": 116200
    },
    {
      "epoch": 45.13,
      "learning_rate": 1.3982653721682848e-05,
      "loss": 0.0297,
      "step": 116210
    },
    {
      "epoch": 45.13,
      "learning_rate": 1.3982135922330097e-05,
      "loss": 0.106,
      "step": 116220
    },
    {
      "epoch": 45.14,
      "learning_rate": 1.3981618122977347e-05,
      "loss": 0.1697,
      "step": 116230
    },
    {
      "epoch": 45.14,
      "learning_rate": 1.3981100323624596e-05,
      "loss": 0.0053,
      "step": 116240
    },
    {
      "epoch": 45.15,
      "learning_rate": 1.3980582524271846e-05,
      "loss": 0.1963,
      "step": 116250
    },
    {
      "epoch": 45.15,
      "learning_rate": 1.3980064724919096e-05,
      "loss": 0.0881,
      "step": 116260
    },
    {
      "epoch": 45.15,
      "learning_rate": 1.3979546925566345e-05,
      "loss": 0.0164,
      "step": 116270
    },
    {
      "epoch": 45.16,
      "learning_rate": 1.3979029126213593e-05,
      "loss": 0.1287,
      "step": 116280
    },
    {
      "epoch": 45.16,
      "learning_rate": 1.3978511326860841e-05,
      "loss": 0.095,
      "step": 116290
    },
    {
      "epoch": 45.17,
      "learning_rate": 1.397799352750809e-05,
      "loss": 0.0097,
      "step": 116300
    },
    {
      "epoch": 45.17,
      "learning_rate": 1.397747572815534e-05,
      "loss": 0.0072,
      "step": 116310
    },
    {
      "epoch": 45.17,
      "learning_rate": 1.397695792880259e-05,
      "loss": 0.1621,
      "step": 116320
    },
    {
      "epoch": 45.18,
      "learning_rate": 1.397644012944984e-05,
      "loss": 0.1351,
      "step": 116330
    },
    {
      "epoch": 45.18,
      "learning_rate": 1.397592233009709e-05,
      "loss": 0.0203,
      "step": 116340
    },
    {
      "epoch": 45.18,
      "learning_rate": 1.3975404530744339e-05,
      "loss": 0.0521,
      "step": 116350
    },
    {
      "epoch": 45.19,
      "learning_rate": 1.3974886731391585e-05,
      "loss": 0.0707,
      "step": 116360
    },
    {
      "epoch": 45.19,
      "learning_rate": 1.3974368932038835e-05,
      "loss": 0.0619,
      "step": 116370
    },
    {
      "epoch": 45.2,
      "learning_rate": 1.3973851132686084e-05,
      "loss": 0.1363,
      "step": 116380
    },
    {
      "epoch": 45.2,
      "learning_rate": 1.3973333333333334e-05,
      "loss": 0.1135,
      "step": 116390
    },
    {
      "epoch": 45.2,
      "learning_rate": 1.3972815533980584e-05,
      "loss": 0.1156,
      "step": 116400
    },
    {
      "epoch": 45.21,
      "learning_rate": 1.3972297734627833e-05,
      "loss": 0.0857,
      "step": 116410
    },
    {
      "epoch": 45.21,
      "learning_rate": 1.3971779935275083e-05,
      "loss": 0.0119,
      "step": 116420
    },
    {
      "epoch": 45.22,
      "learning_rate": 1.3971262135922332e-05,
      "loss": 0.0203,
      "step": 116430
    },
    {
      "epoch": 45.22,
      "learning_rate": 1.3970744336569579e-05,
      "loss": 0.1167,
      "step": 116440
    },
    {
      "epoch": 45.22,
      "learning_rate": 1.3970226537216828e-05,
      "loss": 0.1133,
      "step": 116450
    },
    {
      "epoch": 45.23,
      "learning_rate": 1.3969708737864078e-05,
      "loss": 0.0045,
      "step": 116460
    },
    {
      "epoch": 45.23,
      "learning_rate": 1.3969190938511327e-05,
      "loss": 0.0469,
      "step": 116470
    },
    {
      "epoch": 45.23,
      "learning_rate": 1.3968673139158577e-05,
      "loss": 0.0801,
      "step": 116480
    },
    {
      "epoch": 45.24,
      "learning_rate": 1.3968155339805827e-05,
      "loss": 0.1006,
      "step": 116490
    },
    {
      "epoch": 45.24,
      "learning_rate": 1.3967637540453076e-05,
      "loss": 0.0369,
      "step": 116500
    },
    {
      "epoch": 45.25,
      "learning_rate": 1.3967119741100326e-05,
      "loss": 0.0233,
      "step": 116510
    },
    {
      "epoch": 45.25,
      "learning_rate": 1.3966601941747572e-05,
      "loss": 0.0523,
      "step": 116520
    },
    {
      "epoch": 45.25,
      "learning_rate": 1.3966084142394822e-05,
      "loss": 0.0075,
      "step": 116530
    },
    {
      "epoch": 45.26,
      "learning_rate": 1.3965566343042071e-05,
      "loss": 0.1466,
      "step": 116540
    },
    {
      "epoch": 45.26,
      "learning_rate": 1.3965048543689321e-05,
      "loss": 0.0494,
      "step": 116550
    },
    {
      "epoch": 45.27,
      "learning_rate": 1.396453074433657e-05,
      "loss": 0.0095,
      "step": 116560
    },
    {
      "epoch": 45.27,
      "learning_rate": 1.396401294498382e-05,
      "loss": 0.0797,
      "step": 116570
    },
    {
      "epoch": 45.27,
      "learning_rate": 1.396349514563107e-05,
      "loss": 0.0615,
      "step": 116580
    },
    {
      "epoch": 45.28,
      "learning_rate": 1.396297734627832e-05,
      "loss": 0.0351,
      "step": 116590
    },
    {
      "epoch": 45.28,
      "learning_rate": 1.3962459546925566e-05,
      "loss": 0.0857,
      "step": 116600
    },
    {
      "epoch": 45.29,
      "learning_rate": 1.3961941747572815e-05,
      "loss": 0.0252,
      "step": 116610
    },
    {
      "epoch": 45.29,
      "learning_rate": 1.3961423948220065e-05,
      "loss": 0.0938,
      "step": 116620
    },
    {
      "epoch": 45.29,
      "learning_rate": 1.3960906148867315e-05,
      "loss": 0.0451,
      "step": 116630
    },
    {
      "epoch": 45.3,
      "learning_rate": 1.3960388349514564e-05,
      "loss": 0.3091,
      "step": 116640
    },
    {
      "epoch": 45.3,
      "learning_rate": 1.3959870550161814e-05,
      "loss": 0.0788,
      "step": 116650
    },
    {
      "epoch": 45.3,
      "learning_rate": 1.3959352750809063e-05,
      "loss": 0.097,
      "step": 116660
    },
    {
      "epoch": 45.31,
      "learning_rate": 1.3958834951456313e-05,
      "loss": 0.0384,
      "step": 116670
    },
    {
      "epoch": 45.31,
      "learning_rate": 1.3958317152103563e-05,
      "loss": 0.0511,
      "step": 116680
    },
    {
      "epoch": 45.32,
      "learning_rate": 1.3957799352750809e-05,
      "loss": 0.0931,
      "step": 116690
    },
    {
      "epoch": 45.32,
      "learning_rate": 1.3957281553398059e-05,
      "loss": 0.1179,
      "step": 116700
    },
    {
      "epoch": 45.32,
      "learning_rate": 1.3956763754045308e-05,
      "loss": 0.0393,
      "step": 116710
    },
    {
      "epoch": 45.33,
      "learning_rate": 1.3956245954692558e-05,
      "loss": 0.0673,
      "step": 116720
    },
    {
      "epoch": 45.33,
      "learning_rate": 1.3955728155339807e-05,
      "loss": 0.0281,
      "step": 116730
    },
    {
      "epoch": 45.34,
      "learning_rate": 1.3955210355987057e-05,
      "loss": 0.0083,
      "step": 116740
    },
    {
      "epoch": 45.34,
      "learning_rate": 1.3954692556634307e-05,
      "loss": 0.0825,
      "step": 116750
    },
    {
      "epoch": 45.34,
      "learning_rate": 1.3954174757281556e-05,
      "loss": 0.0845,
      "step": 116760
    },
    {
      "epoch": 45.35,
      "learning_rate": 1.3953656957928802e-05,
      "loss": 0.03,
      "step": 116770
    },
    {
      "epoch": 45.35,
      "learning_rate": 1.3953139158576052e-05,
      "loss": 0.0291,
      "step": 116780
    },
    {
      "epoch": 45.36,
      "learning_rate": 1.3952621359223302e-05,
      "loss": 0.1124,
      "step": 116790
    },
    {
      "epoch": 45.36,
      "learning_rate": 1.3952103559870551e-05,
      "loss": 0.1395,
      "step": 116800
    },
    {
      "epoch": 45.36,
      "learning_rate": 1.3951585760517801e-05,
      "loss": 0.0475,
      "step": 116810
    },
    {
      "epoch": 45.37,
      "learning_rate": 1.395106796116505e-05,
      "loss": 0.0483,
      "step": 116820
    },
    {
      "epoch": 45.37,
      "learning_rate": 1.39505501618123e-05,
      "loss": 0.1144,
      "step": 116830
    },
    {
      "epoch": 45.37,
      "learning_rate": 1.395003236245955e-05,
      "loss": 0.0724,
      "step": 116840
    },
    {
      "epoch": 45.38,
      "learning_rate": 1.3949514563106796e-05,
      "loss": 0.0961,
      "step": 116850
    },
    {
      "epoch": 45.38,
      "learning_rate": 1.3948996763754046e-05,
      "loss": 0.0153,
      "step": 116860
    },
    {
      "epoch": 45.39,
      "learning_rate": 1.3948478964401295e-05,
      "loss": 0.1791,
      "step": 116870
    },
    {
      "epoch": 45.39,
      "learning_rate": 1.3947961165048545e-05,
      "loss": 0.0533,
      "step": 116880
    },
    {
      "epoch": 45.39,
      "learning_rate": 1.3947443365695795e-05,
      "loss": 0.1602,
      "step": 116890
    },
    {
      "epoch": 45.4,
      "learning_rate": 1.3946925566343044e-05,
      "loss": 0.0313,
      "step": 116900
    },
    {
      "epoch": 45.4,
      "learning_rate": 1.3946407766990294e-05,
      "loss": 0.0053,
      "step": 116910
    },
    {
      "epoch": 45.41,
      "learning_rate": 1.3945889967637543e-05,
      "loss": 0.0253,
      "step": 116920
    },
    {
      "epoch": 45.41,
      "learning_rate": 1.394537216828479e-05,
      "loss": 0.0062,
      "step": 116930
    },
    {
      "epoch": 45.41,
      "learning_rate": 1.394485436893204e-05,
      "loss": 0.0575,
      "step": 116940
    },
    {
      "epoch": 45.42,
      "learning_rate": 1.3944336569579289e-05,
      "loss": 0.0147,
      "step": 116950
    },
    {
      "epoch": 45.42,
      "learning_rate": 1.3943818770226538e-05,
      "loss": 0.0995,
      "step": 116960
    },
    {
      "epoch": 45.43,
      "learning_rate": 1.3943300970873788e-05,
      "loss": 0.0665,
      "step": 116970
    },
    {
      "epoch": 45.43,
      "learning_rate": 1.3942783171521038e-05,
      "loss": 0.0081,
      "step": 116980
    },
    {
      "epoch": 45.43,
      "learning_rate": 1.3942265372168287e-05,
      "loss": 0.0878,
      "step": 116990
    },
    {
      "epoch": 45.44,
      "learning_rate": 1.3941747572815535e-05,
      "loss": 0.1847,
      "step": 117000
    },
    {
      "epoch": 45.44,
      "learning_rate": 1.3941229773462783e-05,
      "loss": 0.3107,
      "step": 117010
    },
    {
      "epoch": 45.44,
      "learning_rate": 1.3940711974110033e-05,
      "loss": 0.0868,
      "step": 117020
    },
    {
      "epoch": 45.45,
      "learning_rate": 1.3940194174757282e-05,
      "loss": 0.0621,
      "step": 117030
    },
    {
      "epoch": 45.45,
      "learning_rate": 1.3939676375404532e-05,
      "loss": 0.0801,
      "step": 117040
    },
    {
      "epoch": 45.46,
      "learning_rate": 1.3939158576051782e-05,
      "loss": 0.0659,
      "step": 117050
    },
    {
      "epoch": 45.46,
      "learning_rate": 1.3938640776699031e-05,
      "loss": 0.0497,
      "step": 117060
    },
    {
      "epoch": 45.46,
      "learning_rate": 1.3938122977346281e-05,
      "loss": 0.2756,
      "step": 117070
    },
    {
      "epoch": 45.47,
      "learning_rate": 1.3937605177993529e-05,
      "loss": 0.1786,
      "step": 117080
    },
    {
      "epoch": 45.47,
      "learning_rate": 1.3937087378640777e-05,
      "loss": 0.0275,
      "step": 117090
    },
    {
      "epoch": 45.48,
      "learning_rate": 1.3936569579288026e-05,
      "loss": 0.1328,
      "step": 117100
    },
    {
      "epoch": 45.48,
      "learning_rate": 1.3936051779935276e-05,
      "loss": 0.0292,
      "step": 117110
    },
    {
      "epoch": 45.48,
      "learning_rate": 1.3935533980582526e-05,
      "loss": 0.0105,
      "step": 117120
    },
    {
      "epoch": 45.49,
      "learning_rate": 1.3935016181229775e-05,
      "loss": 0.1031,
      "step": 117130
    },
    {
      "epoch": 45.49,
      "learning_rate": 1.3934498381877025e-05,
      "loss": 0.0479,
      "step": 117140
    },
    {
      "epoch": 45.5,
      "learning_rate": 1.3933980582524273e-05,
      "loss": 0.054,
      "step": 117150
    },
    {
      "epoch": 45.5,
      "learning_rate": 1.3933462783171522e-05,
      "loss": 0.0063,
      "step": 117160
    },
    {
      "epoch": 45.5,
      "learning_rate": 1.3932944983818772e-05,
      "loss": 0.0017,
      "step": 117170
    },
    {
      "epoch": 45.51,
      "learning_rate": 1.393242718446602e-05,
      "loss": 0.0164,
      "step": 117180
    },
    {
      "epoch": 45.51,
      "learning_rate": 1.393190938511327e-05,
      "loss": 0.0997,
      "step": 117190
    },
    {
      "epoch": 45.51,
      "learning_rate": 1.3931391585760519e-05,
      "loss": 0.0876,
      "step": 117200
    },
    {
      "epoch": 45.52,
      "learning_rate": 1.3930873786407769e-05,
      "loss": 0.0883,
      "step": 117210
    },
    {
      "epoch": 45.52,
      "learning_rate": 1.3930355987055018e-05,
      "loss": 0.1892,
      "step": 117220
    },
    {
      "epoch": 45.53,
      "learning_rate": 1.3929838187702266e-05,
      "loss": 0.0922,
      "step": 117230
    },
    {
      "epoch": 45.53,
      "learning_rate": 1.3929320388349516e-05,
      "loss": 0.009,
      "step": 117240
    },
    {
      "epoch": 45.53,
      "learning_rate": 1.3928802588996766e-05,
      "loss": 0.0687,
      "step": 117250
    },
    {
      "epoch": 45.54,
      "learning_rate": 1.3928284789644013e-05,
      "loss": 0.1796,
      "step": 117260
    },
    {
      "epoch": 45.54,
      "learning_rate": 1.3927766990291263e-05,
      "loss": 0.0062,
      "step": 117270
    },
    {
      "epoch": 45.55,
      "learning_rate": 1.3927249190938513e-05,
      "loss": 0.1271,
      "step": 117280
    },
    {
      "epoch": 45.55,
      "learning_rate": 1.3926731391585762e-05,
      "loss": 0.013,
      "step": 117290
    },
    {
      "epoch": 45.55,
      "learning_rate": 1.392621359223301e-05,
      "loss": 0.1195,
      "step": 117300
    },
    {
      "epoch": 45.56,
      "learning_rate": 1.392569579288026e-05,
      "loss": 0.0542,
      "step": 117310
    },
    {
      "epoch": 45.56,
      "learning_rate": 1.392517799352751e-05,
      "loss": 0.1179,
      "step": 117320
    },
    {
      "epoch": 45.57,
      "learning_rate": 1.3924660194174759e-05,
      "loss": 0.1204,
      "step": 117330
    },
    {
      "epoch": 45.57,
      "learning_rate": 1.3924142394822007e-05,
      "loss": 0.1928,
      "step": 117340
    },
    {
      "epoch": 45.57,
      "learning_rate": 1.3923624595469257e-05,
      "loss": 0.0374,
      "step": 117350
    },
    {
      "epoch": 45.58,
      "learning_rate": 1.3923106796116506e-05,
      "loss": 0.0495,
      "step": 117360
    },
    {
      "epoch": 45.58,
      "learning_rate": 1.3922588996763756e-05,
      "loss": 0.1955,
      "step": 117370
    },
    {
      "epoch": 45.58,
      "learning_rate": 1.3922071197411004e-05,
      "loss": 0.2518,
      "step": 117380
    },
    {
      "epoch": 45.59,
      "learning_rate": 1.3921553398058253e-05,
      "loss": 0.1381,
      "step": 117390
    },
    {
      "epoch": 45.59,
      "learning_rate": 1.3921035598705503e-05,
      "loss": 0.0325,
      "step": 117400
    },
    {
      "epoch": 45.6,
      "learning_rate": 1.3920517799352753e-05,
      "loss": 0.1269,
      "step": 117410
    },
    {
      "epoch": 45.6,
      "learning_rate": 1.392e-05,
      "loss": 0.0307,
      "step": 117420
    },
    {
      "epoch": 45.6,
      "learning_rate": 1.391948220064725e-05,
      "loss": 0.047,
      "step": 117430
    },
    {
      "epoch": 45.61,
      "learning_rate": 1.39189644012945e-05,
      "loss": 0.0011,
      "step": 117440
    },
    {
      "epoch": 45.61,
      "learning_rate": 1.391844660194175e-05,
      "loss": 0.0656,
      "step": 117450
    },
    {
      "epoch": 45.62,
      "learning_rate": 1.3917928802588997e-05,
      "loss": 0.1009,
      "step": 117460
    },
    {
      "epoch": 45.62,
      "learning_rate": 1.3917411003236247e-05,
      "loss": 0.1032,
      "step": 117470
    },
    {
      "epoch": 45.62,
      "learning_rate": 1.3916893203883497e-05,
      "loss": 0.091,
      "step": 117480
    },
    {
      "epoch": 45.63,
      "learning_rate": 1.3916375404530746e-05,
      "loss": 0.0703,
      "step": 117490
    },
    {
      "epoch": 45.63,
      "learning_rate": 1.3915857605177994e-05,
      "loss": 0.0263,
      "step": 117500
    },
    {
      "epoch": 45.63,
      "learning_rate": 1.3915339805825244e-05,
      "loss": 0.0171,
      "step": 117510
    },
    {
      "epoch": 45.64,
      "learning_rate": 1.3914822006472493e-05,
      "loss": 0.0844,
      "step": 117520
    },
    {
      "epoch": 45.64,
      "learning_rate": 1.3914304207119741e-05,
      "loss": 0.1271,
      "step": 117530
    },
    {
      "epoch": 45.65,
      "learning_rate": 1.3913786407766991e-05,
      "loss": 0.1219,
      "step": 117540
    },
    {
      "epoch": 45.65,
      "learning_rate": 1.391326860841424e-05,
      "loss": 0.0296,
      "step": 117550
    },
    {
      "epoch": 45.65,
      "learning_rate": 1.391275080906149e-05,
      "loss": 0.1213,
      "step": 117560
    },
    {
      "epoch": 45.66,
      "learning_rate": 1.391223300970874e-05,
      "loss": 0.0666,
      "step": 117570
    },
    {
      "epoch": 45.66,
      "learning_rate": 1.3911715210355988e-05,
      "loss": 0.024,
      "step": 117580
    },
    {
      "epoch": 45.67,
      "learning_rate": 1.3911197411003237e-05,
      "loss": 0.0004,
      "step": 117590
    },
    {
      "epoch": 45.67,
      "learning_rate": 1.3910679611650487e-05,
      "loss": 0.0056,
      "step": 117600
    },
    {
      "epoch": 45.67,
      "learning_rate": 1.3910161812297735e-05,
      "loss": 0.1016,
      "step": 117610
    },
    {
      "epoch": 45.68,
      "learning_rate": 1.3909644012944984e-05,
      "loss": 0.1258,
      "step": 117620
    },
    {
      "epoch": 45.68,
      "learning_rate": 1.3909126213592234e-05,
      "loss": 0.1095,
      "step": 117630
    },
    {
      "epoch": 45.69,
      "learning_rate": 1.3908608414239484e-05,
      "loss": 0.1256,
      "step": 117640
    },
    {
      "epoch": 45.69,
      "learning_rate": 1.3908090614886733e-05,
      "loss": 0.0865,
      "step": 117650
    },
    {
      "epoch": 45.69,
      "learning_rate": 1.3907572815533981e-05,
      "loss": 0.1362,
      "step": 117660
    },
    {
      "epoch": 45.7,
      "learning_rate": 1.390705501618123e-05,
      "loss": 0.0035,
      "step": 117670
    },
    {
      "epoch": 45.7,
      "learning_rate": 1.3906537216828479e-05,
      "loss": 0.0056,
      "step": 117680
    },
    {
      "epoch": 45.7,
      "learning_rate": 1.3906019417475728e-05,
      "loss": 0.0847,
      "step": 117690
    },
    {
      "epoch": 45.71,
      "learning_rate": 1.3905501618122978e-05,
      "loss": 0.2662,
      "step": 117700
    },
    {
      "epoch": 45.71,
      "learning_rate": 1.3904983818770228e-05,
      "loss": 0.1463,
      "step": 117710
    },
    {
      "epoch": 45.72,
      "learning_rate": 1.3904466019417477e-05,
      "loss": 0.0732,
      "step": 117720
    },
    {
      "epoch": 45.72,
      "learning_rate": 1.3903948220064727e-05,
      "loss": 0.1418,
      "step": 117730
    },
    {
      "epoch": 45.72,
      "learning_rate": 1.3903430420711976e-05,
      "loss": 0.0011,
      "step": 117740
    },
    {
      "epoch": 45.73,
      "learning_rate": 1.3902912621359224e-05,
      "loss": 0.0794,
      "step": 117750
    },
    {
      "epoch": 45.73,
      "learning_rate": 1.3902394822006472e-05,
      "loss": 0.107,
      "step": 117760
    },
    {
      "epoch": 45.74,
      "learning_rate": 1.3901877022653722e-05,
      "loss": 0.2211,
      "step": 117770
    },
    {
      "epoch": 45.74,
      "learning_rate": 1.3901359223300972e-05,
      "loss": 0.0451,
      "step": 117780
    },
    {
      "epoch": 45.74,
      "learning_rate": 1.3900841423948221e-05,
      "loss": 0.0901,
      "step": 117790
    },
    {
      "epoch": 45.75,
      "learning_rate": 1.390032362459547e-05,
      "loss": 0.0855,
      "step": 117800
    },
    {
      "epoch": 45.75,
      "learning_rate": 1.389980582524272e-05,
      "loss": 0.011,
      "step": 117810
    },
    {
      "epoch": 45.76,
      "learning_rate": 1.389928802588997e-05,
      "loss": 0.0666,
      "step": 117820
    },
    {
      "epoch": 45.76,
      "learning_rate": 1.3898770226537216e-05,
      "loss": 0.0363,
      "step": 117830
    },
    {
      "epoch": 45.76,
      "learning_rate": 1.3898252427184466e-05,
      "loss": 0.111,
      "step": 117840
    },
    {
      "epoch": 45.77,
      "learning_rate": 1.3897734627831715e-05,
      "loss": 0.028,
      "step": 117850
    },
    {
      "epoch": 45.77,
      "learning_rate": 1.3897216828478965e-05,
      "loss": 0.1051,
      "step": 117860
    },
    {
      "epoch": 45.77,
      "learning_rate": 1.3896699029126215e-05,
      "loss": 0.1985,
      "step": 117870
    },
    {
      "epoch": 45.78,
      "learning_rate": 1.3896181229773464e-05,
      "loss": 0.1236,
      "step": 117880
    },
    {
      "epoch": 45.78,
      "learning_rate": 1.3895663430420714e-05,
      "loss": 0.2367,
      "step": 117890
    },
    {
      "epoch": 45.79,
      "learning_rate": 1.3895145631067964e-05,
      "loss": 0.0241,
      "step": 117900
    },
    {
      "epoch": 45.79,
      "learning_rate": 1.389462783171521e-05,
      "loss": 0.0558,
      "step": 117910
    },
    {
      "epoch": 45.79,
      "learning_rate": 1.389411003236246e-05,
      "loss": 0.0727,
      "step": 117920
    },
    {
      "epoch": 45.8,
      "learning_rate": 1.3893592233009709e-05,
      "loss": 0.1284,
      "step": 117930
    },
    {
      "epoch": 45.8,
      "learning_rate": 1.3893074433656959e-05,
      "loss": 0.0046,
      "step": 117940
    },
    {
      "epoch": 45.81,
      "learning_rate": 1.3892556634304208e-05,
      "loss": 0.0503,
      "step": 117950
    },
    {
      "epoch": 45.81,
      "learning_rate": 1.3892038834951458e-05,
      "loss": 0.0048,
      "step": 117960
    },
    {
      "epoch": 45.81,
      "learning_rate": 1.3891521035598708e-05,
      "loss": 0.1027,
      "step": 117970
    },
    {
      "epoch": 45.82,
      "learning_rate": 1.3891003236245957e-05,
      "loss": 0.1582,
      "step": 117980
    },
    {
      "epoch": 45.82,
      "learning_rate": 1.3890485436893203e-05,
      "loss": 0.0266,
      "step": 117990
    },
    {
      "epoch": 45.83,
      "learning_rate": 1.3889967637540453e-05,
      "loss": 0.0175,
      "step": 118000
    },
    {
      "epoch": 45.83,
      "learning_rate": 1.3889449838187703e-05,
      "loss": 0.0281,
      "step": 118010
    },
    {
      "epoch": 45.83,
      "learning_rate": 1.3888932038834952e-05,
      "loss": 0.0154,
      "step": 118020
    },
    {
      "epoch": 45.84,
      "learning_rate": 1.3888414239482202e-05,
      "loss": 0.0858,
      "step": 118030
    },
    {
      "epoch": 45.84,
      "learning_rate": 1.3887896440129451e-05,
      "loss": 0.0285,
      "step": 118040
    },
    {
      "epoch": 45.84,
      "learning_rate": 1.3887378640776701e-05,
      "loss": 0.2288,
      "step": 118050
    },
    {
      "epoch": 45.85,
      "learning_rate": 1.388686084142395e-05,
      "loss": 0.0499,
      "step": 118060
    },
    {
      "epoch": 45.85,
      "learning_rate": 1.3886343042071197e-05,
      "loss": 0.1386,
      "step": 118070
    },
    {
      "epoch": 45.86,
      "learning_rate": 1.3885825242718447e-05,
      "loss": 0.1937,
      "step": 118080
    },
    {
      "epoch": 45.86,
      "learning_rate": 1.3885307443365696e-05,
      "loss": 0.0323,
      "step": 118090
    },
    {
      "epoch": 45.86,
      "learning_rate": 1.3884789644012946e-05,
      "loss": 0.0052,
      "step": 118100
    },
    {
      "epoch": 45.87,
      "learning_rate": 1.3884271844660195e-05,
      "loss": 0.0721,
      "step": 118110
    },
    {
      "epoch": 45.87,
      "learning_rate": 1.3883754045307445e-05,
      "loss": 0.1239,
      "step": 118120
    },
    {
      "epoch": 45.88,
      "learning_rate": 1.3883236245954695e-05,
      "loss": 0.1845,
      "step": 118130
    },
    {
      "epoch": 45.88,
      "learning_rate": 1.3882718446601944e-05,
      "loss": 0.1274,
      "step": 118140
    },
    {
      "epoch": 45.88,
      "learning_rate": 1.388220064724919e-05,
      "loss": 0.0259,
      "step": 118150
    },
    {
      "epoch": 45.89,
      "learning_rate": 1.388168284789644e-05,
      "loss": 0.076,
      "step": 118160
    },
    {
      "epoch": 45.89,
      "learning_rate": 1.388116504854369e-05,
      "loss": 0.0454,
      "step": 118170
    },
    {
      "epoch": 45.9,
      "learning_rate": 1.388064724919094e-05,
      "loss": 0.1178,
      "step": 118180
    },
    {
      "epoch": 45.9,
      "learning_rate": 1.3880129449838189e-05,
      "loss": 0.1449,
      "step": 118190
    },
    {
      "epoch": 45.9,
      "learning_rate": 1.3879611650485439e-05,
      "loss": 0.0419,
      "step": 118200
    },
    {
      "epoch": 45.91,
      "learning_rate": 1.3879093851132688e-05,
      "loss": 0.036,
      "step": 118210
    },
    {
      "epoch": 45.91,
      "learning_rate": 1.3878576051779938e-05,
      "loss": 0.0529,
      "step": 118220
    },
    {
      "epoch": 45.91,
      "learning_rate": 1.3878058252427184e-05,
      "loss": 0.0808,
      "step": 118230
    },
    {
      "epoch": 45.92,
      "learning_rate": 1.3877540453074434e-05,
      "loss": 0.1587,
      "step": 118240
    },
    {
      "epoch": 45.92,
      "learning_rate": 1.3877022653721683e-05,
      "loss": 0.0013,
      "step": 118250
    },
    {
      "epoch": 45.93,
      "learning_rate": 1.3876504854368933e-05,
      "loss": 0.0618,
      "step": 118260
    },
    {
      "epoch": 45.93,
      "learning_rate": 1.3875987055016183e-05,
      "loss": 0.0178,
      "step": 118270
    },
    {
      "epoch": 45.93,
      "learning_rate": 1.3875469255663432e-05,
      "loss": 0.0355,
      "step": 118280
    },
    {
      "epoch": 45.94,
      "learning_rate": 1.3874951456310682e-05,
      "loss": 0.1488,
      "step": 118290
    },
    {
      "epoch": 45.94,
      "learning_rate": 1.3874433656957931e-05,
      "loss": 0.062,
      "step": 118300
    },
    {
      "epoch": 45.95,
      "learning_rate": 1.3873915857605181e-05,
      "loss": 0.0332,
      "step": 118310
    },
    {
      "epoch": 45.95,
      "learning_rate": 1.3873398058252427e-05,
      "loss": 0.16,
      "step": 118320
    },
    {
      "epoch": 45.95,
      "learning_rate": 1.3872880258899677e-05,
      "loss": 0.1126,
      "step": 118330
    },
    {
      "epoch": 45.96,
      "learning_rate": 1.3872362459546926e-05,
      "loss": 0.0179,
      "step": 118340
    },
    {
      "epoch": 45.96,
      "learning_rate": 1.3871844660194176e-05,
      "loss": 0.1076,
      "step": 118350
    },
    {
      "epoch": 45.97,
      "learning_rate": 1.3871326860841426e-05,
      "loss": 0.0104,
      "step": 118360
    },
    {
      "epoch": 45.97,
      "learning_rate": 1.3870809061488675e-05,
      "loss": 0.0687,
      "step": 118370
    },
    {
      "epoch": 45.97,
      "learning_rate": 1.3870291262135925e-05,
      "loss": 0.1351,
      "step": 118380
    },
    {
      "epoch": 45.98,
      "learning_rate": 1.3869773462783175e-05,
      "loss": 0.0399,
      "step": 118390
    },
    {
      "epoch": 45.98,
      "learning_rate": 1.386925566343042e-05,
      "loss": 0.1683,
      "step": 118400
    },
    {
      "epoch": 45.98,
      "learning_rate": 1.386873786407767e-05,
      "loss": 0.2905,
      "step": 118410
    },
    {
      "epoch": 45.99,
      "learning_rate": 1.386822006472492e-05,
      "loss": 0.0311,
      "step": 118420
    },
    {
      "epoch": 45.99,
      "learning_rate": 1.386770226537217e-05,
      "loss": 0.1109,
      "step": 118430
    },
    {
      "epoch": 46.0,
      "learning_rate": 1.386718446601942e-05,
      "loss": 0.0638,
      "step": 118440
    },
    {
      "epoch": 46.0,
      "learning_rate": 1.3866666666666669e-05,
      "loss": 0.1735,
      "step": 118450
    },
    {
      "epoch": 46.0,
      "eval_accuracy": 0.9493810178817056,
      "eval_loss": 0.31910738348960876,
      "eval_runtime": 8.1822,
      "eval_samples_per_second": 444.255,
      "eval_steps_per_second": 55.608,
      "step": 118450
    },
    {
      "epoch": 46.0,
      "learning_rate": 1.3866148867313918e-05,
      "loss": 0.0018,
      "step": 118460
    },
    {
      "epoch": 46.01,
      "learning_rate": 1.3865631067961166e-05,
      "loss": 0.0839,
      "step": 118470
    },
    {
      "epoch": 46.01,
      "learning_rate": 1.3865113268608414e-05,
      "loss": 0.147,
      "step": 118480
    },
    {
      "epoch": 46.02,
      "learning_rate": 1.3864595469255664e-05,
      "loss": 0.0284,
      "step": 118490
    },
    {
      "epoch": 46.02,
      "learning_rate": 1.3864077669902914e-05,
      "loss": 0.0714,
      "step": 118500
    },
    {
      "epoch": 46.02,
      "learning_rate": 1.3863559870550163e-05,
      "loss": 0.1089,
      "step": 118510
    },
    {
      "epoch": 46.03,
      "learning_rate": 1.3863042071197413e-05,
      "loss": 0.0324,
      "step": 118520
    },
    {
      "epoch": 46.03,
      "learning_rate": 1.3862524271844662e-05,
      "loss": 0.0931,
      "step": 118530
    },
    {
      "epoch": 46.03,
      "learning_rate": 1.3862006472491912e-05,
      "loss": 0.0398,
      "step": 118540
    },
    {
      "epoch": 46.04,
      "learning_rate": 1.386148867313916e-05,
      "loss": 0.1862,
      "step": 118550
    },
    {
      "epoch": 46.04,
      "learning_rate": 1.3860970873786408e-05,
      "loss": 0.0189,
      "step": 118560
    },
    {
      "epoch": 46.05,
      "learning_rate": 1.3860453074433658e-05,
      "loss": 0.0181,
      "step": 118570
    },
    {
      "epoch": 46.05,
      "learning_rate": 1.3859935275080907e-05,
      "loss": 0.0366,
      "step": 118580
    },
    {
      "epoch": 46.05,
      "learning_rate": 1.3859417475728157e-05,
      "loss": 0.1725,
      "step": 118590
    },
    {
      "epoch": 46.06,
      "learning_rate": 1.3858899676375406e-05,
      "loss": 0.0766,
      "step": 118600
    },
    {
      "epoch": 46.06,
      "learning_rate": 1.3858381877022656e-05,
      "loss": 0.0573,
      "step": 118610
    },
    {
      "epoch": 46.07,
      "learning_rate": 1.3857864077669904e-05,
      "loss": 0.0378,
      "step": 118620
    },
    {
      "epoch": 46.07,
      "learning_rate": 1.3857346278317154e-05,
      "loss": 0.0619,
      "step": 118630
    },
    {
      "epoch": 46.07,
      "learning_rate": 1.3856828478964401e-05,
      "loss": 0.1167,
      "step": 118640
    },
    {
      "epoch": 46.08,
      "learning_rate": 1.3856310679611651e-05,
      "loss": 0.0349,
      "step": 118650
    },
    {
      "epoch": 46.08,
      "learning_rate": 1.38557928802589e-05,
      "loss": 0.0919,
      "step": 118660
    },
    {
      "epoch": 46.09,
      "learning_rate": 1.385527508090615e-05,
      "loss": 0.016,
      "step": 118670
    },
    {
      "epoch": 46.09,
      "learning_rate": 1.38547572815534e-05,
      "loss": 0.0935,
      "step": 118680
    },
    {
      "epoch": 46.09,
      "learning_rate": 1.385423948220065e-05,
      "loss": 0.1195,
      "step": 118690
    },
    {
      "epoch": 46.1,
      "learning_rate": 1.3853721682847897e-05,
      "loss": 0.0789,
      "step": 118700
    },
    {
      "epoch": 46.1,
      "learning_rate": 1.3853203883495147e-05,
      "loss": 0.2546,
      "step": 118710
    },
    {
      "epoch": 46.1,
      "learning_rate": 1.3852686084142395e-05,
      "loss": 0.1461,
      "step": 118720
    },
    {
      "epoch": 46.11,
      "learning_rate": 1.3852168284789645e-05,
      "loss": 0.1101,
      "step": 118730
    },
    {
      "epoch": 46.11,
      "learning_rate": 1.3851650485436894e-05,
      "loss": 0.174,
      "step": 118740
    },
    {
      "epoch": 46.12,
      "learning_rate": 1.3851132686084144e-05,
      "loss": 0.002,
      "step": 118750
    },
    {
      "epoch": 46.12,
      "learning_rate": 1.3850614886731393e-05,
      "loss": 0.1688,
      "step": 118760
    },
    {
      "epoch": 46.12,
      "learning_rate": 1.3850097087378641e-05,
      "loss": 0.2182,
      "step": 118770
    },
    {
      "epoch": 46.13,
      "learning_rate": 1.3849579288025891e-05,
      "loss": 0.0789,
      "step": 118780
    },
    {
      "epoch": 46.13,
      "learning_rate": 1.384906148867314e-05,
      "loss": 0.0328,
      "step": 118790
    },
    {
      "epoch": 46.14,
      "learning_rate": 1.3848543689320389e-05,
      "loss": 0.0518,
      "step": 118800
    },
    {
      "epoch": 46.14,
      "learning_rate": 1.3848025889967638e-05,
      "loss": 0.1019,
      "step": 118810
    },
    {
      "epoch": 46.14,
      "learning_rate": 1.3847508090614888e-05,
      "loss": 0.1366,
      "step": 118820
    },
    {
      "epoch": 46.15,
      "learning_rate": 1.3846990291262137e-05,
      "loss": 0.0028,
      "step": 118830
    },
    {
      "epoch": 46.15,
      "learning_rate": 1.3846472491909387e-05,
      "loss": 0.1695,
      "step": 118840
    },
    {
      "epoch": 46.16,
      "learning_rate": 1.3845954692556635e-05,
      "loss": 0.1993,
      "step": 118850
    },
    {
      "epoch": 46.16,
      "learning_rate": 1.3845436893203885e-05,
      "loss": 0.1511,
      "step": 118860
    },
    {
      "epoch": 46.16,
      "learning_rate": 1.3844919093851134e-05,
      "loss": 0.0072,
      "step": 118870
    },
    {
      "epoch": 46.17,
      "learning_rate": 1.3844401294498384e-05,
      "loss": 0.0012,
      "step": 118880
    },
    {
      "epoch": 46.17,
      "learning_rate": 1.3843883495145632e-05,
      "loss": 0.0878,
      "step": 118890
    },
    {
      "epoch": 46.17,
      "learning_rate": 1.3843365695792881e-05,
      "loss": 0.0755,
      "step": 118900
    },
    {
      "epoch": 46.18,
      "learning_rate": 1.3842847896440131e-05,
      "loss": 0.0302,
      "step": 118910
    },
    {
      "epoch": 46.18,
      "learning_rate": 1.384233009708738e-05,
      "loss": 0.0009,
      "step": 118920
    },
    {
      "epoch": 46.19,
      "learning_rate": 1.3841812297734629e-05,
      "loss": 0.0222,
      "step": 118930
    },
    {
      "epoch": 46.19,
      "learning_rate": 1.3841294498381878e-05,
      "loss": 0.0112,
      "step": 118940
    },
    {
      "epoch": 46.19,
      "learning_rate": 1.3840776699029128e-05,
      "loss": 0.0209,
      "step": 118950
    },
    {
      "epoch": 46.2,
      "learning_rate": 1.3840258899676377e-05,
      "loss": 0.0561,
      "step": 118960
    },
    {
      "epoch": 46.2,
      "learning_rate": 1.3839741100323625e-05,
      "loss": 0.2384,
      "step": 118970
    },
    {
      "epoch": 46.21,
      "learning_rate": 1.3839223300970875e-05,
      "loss": 0.081,
      "step": 118980
    },
    {
      "epoch": 46.21,
      "learning_rate": 1.3838705501618125e-05,
      "loss": 0.0735,
      "step": 118990
    },
    {
      "epoch": 46.21,
      "learning_rate": 1.3838187702265372e-05,
      "loss": 0.001,
      "step": 119000
    },
    {
      "epoch": 46.22,
      "learning_rate": 1.3837669902912622e-05,
      "loss": 0.0398,
      "step": 119010
    },
    {
      "epoch": 46.22,
      "learning_rate": 1.3837152103559872e-05,
      "loss": 0.1107,
      "step": 119020
    },
    {
      "epoch": 46.23,
      "learning_rate": 1.3836634304207121e-05,
      "loss": 0.1387,
      "step": 119030
    },
    {
      "epoch": 46.23,
      "learning_rate": 1.3836116504854371e-05,
      "loss": 0.0449,
      "step": 119040
    },
    {
      "epoch": 46.23,
      "learning_rate": 1.3835598705501619e-05,
      "loss": 0.0339,
      "step": 119050
    },
    {
      "epoch": 46.24,
      "learning_rate": 1.3835080906148868e-05,
      "loss": 0.1461,
      "step": 119060
    },
    {
      "epoch": 46.24,
      "learning_rate": 1.3834563106796118e-05,
      "loss": 0.1617,
      "step": 119070
    },
    {
      "epoch": 46.24,
      "learning_rate": 1.3834045307443366e-05,
      "loss": 0.1343,
      "step": 119080
    },
    {
      "epoch": 46.25,
      "learning_rate": 1.3833527508090616e-05,
      "loss": 0.0019,
      "step": 119090
    },
    {
      "epoch": 46.25,
      "learning_rate": 1.3833009708737865e-05,
      "loss": 0.1805,
      "step": 119100
    },
    {
      "epoch": 46.26,
      "learning_rate": 1.3832491909385115e-05,
      "loss": 0.0314,
      "step": 119110
    },
    {
      "epoch": 46.26,
      "learning_rate": 1.3831974110032364e-05,
      "loss": 0.0784,
      "step": 119120
    },
    {
      "epoch": 46.26,
      "learning_rate": 1.3831456310679612e-05,
      "loss": 0.0379,
      "step": 119130
    },
    {
      "epoch": 46.27,
      "learning_rate": 1.3830938511326862e-05,
      "loss": 0.0424,
      "step": 119140
    },
    {
      "epoch": 46.27,
      "learning_rate": 1.383042071197411e-05,
      "loss": 0.0846,
      "step": 119150
    },
    {
      "epoch": 46.28,
      "learning_rate": 1.382990291262136e-05,
      "loss": 0.1702,
      "step": 119160
    },
    {
      "epoch": 46.28,
      "learning_rate": 1.382938511326861e-05,
      "loss": 0.0852,
      "step": 119170
    },
    {
      "epoch": 46.28,
      "learning_rate": 1.3828867313915859e-05,
      "loss": 0.0456,
      "step": 119180
    },
    {
      "epoch": 46.29,
      "learning_rate": 1.3828349514563108e-05,
      "loss": 0.034,
      "step": 119190
    },
    {
      "epoch": 46.29,
      "learning_rate": 1.3827831715210358e-05,
      "loss": 0.0389,
      "step": 119200
    },
    {
      "epoch": 46.3,
      "learning_rate": 1.3827313915857606e-05,
      "loss": 0.0139,
      "step": 119210
    },
    {
      "epoch": 46.3,
      "learning_rate": 1.3826796116504856e-05,
      "loss": 0.1255,
      "step": 119220
    },
    {
      "epoch": 46.3,
      "learning_rate": 1.3826278317152103e-05,
      "loss": 0.0251,
      "step": 119230
    },
    {
      "epoch": 46.31,
      "learning_rate": 1.3825760517799353e-05,
      "loss": 0.0606,
      "step": 119240
    },
    {
      "epoch": 46.31,
      "learning_rate": 1.3825242718446603e-05,
      "loss": 0.0554,
      "step": 119250
    },
    {
      "epoch": 46.31,
      "learning_rate": 1.3824724919093852e-05,
      "loss": 0.0347,
      "step": 119260
    },
    {
      "epoch": 46.32,
      "learning_rate": 1.3824207119741102e-05,
      "loss": 0.0175,
      "step": 119270
    },
    {
      "epoch": 46.32,
      "learning_rate": 1.3823689320388352e-05,
      "loss": 0.0472,
      "step": 119280
    },
    {
      "epoch": 46.33,
      "learning_rate": 1.38231715210356e-05,
      "loss": 0.0032,
      "step": 119290
    },
    {
      "epoch": 46.33,
      "learning_rate": 1.3822653721682847e-05,
      "loss": 0.0069,
      "step": 119300
    },
    {
      "epoch": 46.33,
      "learning_rate": 1.3822135922330097e-05,
      "loss": 0.0869,
      "step": 119310
    },
    {
      "epoch": 46.34,
      "learning_rate": 1.3821618122977347e-05,
      "loss": 0.1355,
      "step": 119320
    },
    {
      "epoch": 46.34,
      "learning_rate": 1.3821100323624596e-05,
      "loss": 0.1399,
      "step": 119330
    },
    {
      "epoch": 46.35,
      "learning_rate": 1.3820582524271846e-05,
      "loss": 0.0609,
      "step": 119340
    },
    {
      "epoch": 46.35,
      "learning_rate": 1.3820064724919096e-05,
      "loss": 0.0224,
      "step": 119350
    },
    {
      "epoch": 46.35,
      "learning_rate": 1.3819546925566345e-05,
      "loss": 0.0383,
      "step": 119360
    },
    {
      "epoch": 46.36,
      "learning_rate": 1.3819029126213593e-05,
      "loss": 0.2589,
      "step": 119370
    },
    {
      "epoch": 46.36,
      "learning_rate": 1.3818511326860841e-05,
      "loss": 0.1203,
      "step": 119380
    },
    {
      "epoch": 46.37,
      "learning_rate": 1.381799352750809e-05,
      "loss": 0.1043,
      "step": 119390
    },
    {
      "epoch": 46.37,
      "learning_rate": 1.381747572815534e-05,
      "loss": 0.0252,
      "step": 119400
    },
    {
      "epoch": 46.37,
      "learning_rate": 1.381695792880259e-05,
      "loss": 0.0961,
      "step": 119410
    },
    {
      "epoch": 46.38,
      "learning_rate": 1.381644012944984e-05,
      "loss": 0.0914,
      "step": 119420
    },
    {
      "epoch": 46.38,
      "learning_rate": 1.3815922330097089e-05,
      "loss": 0.1181,
      "step": 119430
    },
    {
      "epoch": 46.38,
      "learning_rate": 1.3815404530744339e-05,
      "loss": 0.0055,
      "step": 119440
    },
    {
      "epoch": 46.39,
      "learning_rate": 1.3814886731391588e-05,
      "loss": 0.0795,
      "step": 119450
    },
    {
      "epoch": 46.39,
      "learning_rate": 1.3814368932038835e-05,
      "loss": 0.0272,
      "step": 119460
    },
    {
      "epoch": 46.4,
      "learning_rate": 1.3813851132686084e-05,
      "loss": 0.0054,
      "step": 119470
    },
    {
      "epoch": 46.4,
      "learning_rate": 1.3813333333333334e-05,
      "loss": 0.0896,
      "step": 119480
    },
    {
      "epoch": 46.4,
      "learning_rate": 1.3812815533980583e-05,
      "loss": 0.0695,
      "step": 119490
    },
    {
      "epoch": 46.41,
      "learning_rate": 1.3812297734627833e-05,
      "loss": 0.0112,
      "step": 119500
    },
    {
      "epoch": 46.41,
      "learning_rate": 1.3811779935275083e-05,
      "loss": 0.0004,
      "step": 119510
    },
    {
      "epoch": 46.42,
      "learning_rate": 1.3811262135922332e-05,
      "loss": 0.1074,
      "step": 119520
    },
    {
      "epoch": 46.42,
      "learning_rate": 1.3810744336569582e-05,
      "loss": 0.036,
      "step": 119530
    },
    {
      "epoch": 46.42,
      "learning_rate": 1.3810226537216828e-05,
      "loss": 0.0579,
      "step": 119540
    },
    {
      "epoch": 46.43,
      "learning_rate": 1.3809708737864078e-05,
      "loss": 0.0402,
      "step": 119550
    },
    {
      "epoch": 46.43,
      "learning_rate": 1.3809190938511327e-05,
      "loss": 0.0845,
      "step": 119560
    },
    {
      "epoch": 46.43,
      "learning_rate": 1.3808673139158577e-05,
      "loss": 0.1101,
      "step": 119570
    },
    {
      "epoch": 46.44,
      "learning_rate": 1.3808155339805827e-05,
      "loss": 0.0447,
      "step": 119580
    },
    {
      "epoch": 46.44,
      "learning_rate": 1.3807637540453076e-05,
      "loss": 0.0387,
      "step": 119590
    },
    {
      "epoch": 46.45,
      "learning_rate": 1.3807119741100326e-05,
      "loss": 0.0685,
      "step": 119600
    },
    {
      "epoch": 46.45,
      "learning_rate": 1.3806601941747575e-05,
      "loss": 0.0284,
      "step": 119610
    },
    {
      "epoch": 46.45,
      "learning_rate": 1.3806084142394822e-05,
      "loss": 0.0319,
      "step": 119620
    },
    {
      "epoch": 46.46,
      "learning_rate": 1.3805566343042071e-05,
      "loss": 0.1088,
      "step": 119630
    },
    {
      "epoch": 46.46,
      "learning_rate": 1.3805048543689321e-05,
      "loss": 0.1034,
      "step": 119640
    },
    {
      "epoch": 46.47,
      "learning_rate": 1.380453074433657e-05,
      "loss": 0.069,
      "step": 119650
    },
    {
      "epoch": 46.47,
      "learning_rate": 1.380401294498382e-05,
      "loss": 0.1487,
      "step": 119660
    },
    {
      "epoch": 46.47,
      "learning_rate": 1.380349514563107e-05,
      "loss": 0.033,
      "step": 119670
    },
    {
      "epoch": 46.48,
      "learning_rate": 1.380297734627832e-05,
      "loss": 0.1628,
      "step": 119680
    },
    {
      "epoch": 46.48,
      "learning_rate": 1.3802459546925569e-05,
      "loss": 0.1152,
      "step": 119690
    },
    {
      "epoch": 46.49,
      "learning_rate": 1.3801941747572815e-05,
      "loss": 0.1103,
      "step": 119700
    },
    {
      "epoch": 46.49,
      "learning_rate": 1.3801423948220065e-05,
      "loss": 0.2213,
      "step": 119710
    },
    {
      "epoch": 46.49,
      "learning_rate": 1.3800906148867314e-05,
      "loss": 0.1662,
      "step": 119720
    },
    {
      "epoch": 46.5,
      "learning_rate": 1.3800388349514564e-05,
      "loss": 0.1643,
      "step": 119730
    },
    {
      "epoch": 46.5,
      "learning_rate": 1.3799870550161814e-05,
      "loss": 0.1132,
      "step": 119740
    },
    {
      "epoch": 46.5,
      "learning_rate": 1.3799352750809063e-05,
      "loss": 0.1096,
      "step": 119750
    },
    {
      "epoch": 46.51,
      "learning_rate": 1.3798834951456313e-05,
      "loss": 0.017,
      "step": 119760
    },
    {
      "epoch": 46.51,
      "learning_rate": 1.3798317152103563e-05,
      "loss": 0.0189,
      "step": 119770
    },
    {
      "epoch": 46.52,
      "learning_rate": 1.3797799352750809e-05,
      "loss": 0.0304,
      "step": 119780
    },
    {
      "epoch": 46.52,
      "learning_rate": 1.3797281553398058e-05,
      "loss": 0.1059,
      "step": 119790
    },
    {
      "epoch": 46.52,
      "learning_rate": 1.3796763754045308e-05,
      "loss": 0.1213,
      "step": 119800
    },
    {
      "epoch": 46.53,
      "learning_rate": 1.3796245954692558e-05,
      "loss": 0.0942,
      "step": 119810
    },
    {
      "epoch": 46.53,
      "learning_rate": 1.3795728155339807e-05,
      "loss": 0.162,
      "step": 119820
    },
    {
      "epoch": 46.54,
      "learning_rate": 1.3795210355987057e-05,
      "loss": 0.0228,
      "step": 119830
    },
    {
      "epoch": 46.54,
      "learning_rate": 1.3794692556634306e-05,
      "loss": 0.024,
      "step": 119840
    },
    {
      "epoch": 46.54,
      "learning_rate": 1.3794174757281556e-05,
      "loss": 0.1177,
      "step": 119850
    },
    {
      "epoch": 46.55,
      "learning_rate": 1.3793656957928802e-05,
      "loss": 0.0926,
      "step": 119860
    },
    {
      "epoch": 46.55,
      "learning_rate": 1.3793139158576052e-05,
      "loss": 0.0863,
      "step": 119870
    },
    {
      "epoch": 46.56,
      "learning_rate": 1.3792621359223302e-05,
      "loss": 0.249,
      "step": 119880
    },
    {
      "epoch": 46.56,
      "learning_rate": 1.3792103559870551e-05,
      "loss": 0.0375,
      "step": 119890
    },
    {
      "epoch": 46.56,
      "learning_rate": 1.37915857605178e-05,
      "loss": 0.0374,
      "step": 119900
    },
    {
      "epoch": 46.57,
      "learning_rate": 1.379106796116505e-05,
      "loss": 0.2026,
      "step": 119910
    },
    {
      "epoch": 46.57,
      "learning_rate": 1.37905501618123e-05,
      "loss": 0.0156,
      "step": 119920
    },
    {
      "epoch": 46.57,
      "learning_rate": 1.379003236245955e-05,
      "loss": 0.0186,
      "step": 119930
    },
    {
      "epoch": 46.58,
      "learning_rate": 1.3789514563106796e-05,
      "loss": 0.0192,
      "step": 119940
    },
    {
      "epoch": 46.58,
      "learning_rate": 1.3788996763754046e-05,
      "loss": 0.1303,
      "step": 119950
    },
    {
      "epoch": 46.59,
      "learning_rate": 1.3788478964401295e-05,
      "loss": 0.037,
      "step": 119960
    },
    {
      "epoch": 46.59,
      "learning_rate": 1.3787961165048545e-05,
      "loss": 0.0351,
      "step": 119970
    },
    {
      "epoch": 46.59,
      "learning_rate": 1.3787443365695794e-05,
      "loss": 0.1221,
      "step": 119980
    },
    {
      "epoch": 46.6,
      "learning_rate": 1.3786925566343044e-05,
      "loss": 0.0762,
      "step": 119990
    },
    {
      "epoch": 46.6,
      "learning_rate": 1.3786407766990294e-05,
      "loss": 0.3293,
      "step": 120000
    },
    {
      "epoch": 46.61,
      "learning_rate": 1.3785889967637543e-05,
      "loss": 0.0549,
      "step": 120010
    },
    {
      "epoch": 46.61,
      "learning_rate": 1.3785372168284791e-05,
      "loss": 0.0037,
      "step": 120020
    },
    {
      "epoch": 46.61,
      "learning_rate": 1.3784854368932039e-05,
      "loss": 0.0568,
      "step": 120030
    },
    {
      "epoch": 46.62,
      "learning_rate": 1.3784336569579289e-05,
      "loss": 0.1536,
      "step": 120040
    },
    {
      "epoch": 46.62,
      "learning_rate": 1.3783818770226538e-05,
      "loss": 0.0738,
      "step": 120050
    },
    {
      "epoch": 46.63,
      "learning_rate": 1.3783300970873788e-05,
      "loss": 0.1567,
      "step": 120060
    },
    {
      "epoch": 46.63,
      "learning_rate": 1.3782783171521038e-05,
      "loss": 0.0689,
      "step": 120070
    },
    {
      "epoch": 46.63,
      "learning_rate": 1.3782265372168287e-05,
      "loss": 0.0095,
      "step": 120080
    },
    {
      "epoch": 46.64,
      "learning_rate": 1.3781747572815535e-05,
      "loss": 0.1123,
      "step": 120090
    },
    {
      "epoch": 46.64,
      "learning_rate": 1.3781229773462785e-05,
      "loss": 0.1089,
      "step": 120100
    },
    {
      "epoch": 46.64,
      "learning_rate": 1.3780711974110033e-05,
      "loss": 0.1506,
      "step": 120110
    },
    {
      "epoch": 46.65,
      "learning_rate": 1.3780194174757282e-05,
      "loss": 0.0835,
      "step": 120120
    },
    {
      "epoch": 46.65,
      "learning_rate": 1.3779676375404532e-05,
      "loss": 0.1639,
      "step": 120130
    },
    {
      "epoch": 46.66,
      "learning_rate": 1.3779158576051781e-05,
      "loss": 0.1013,
      "step": 120140
    },
    {
      "epoch": 46.66,
      "learning_rate": 1.3778640776699031e-05,
      "loss": 0.0229,
      "step": 120150
    },
    {
      "epoch": 46.66,
      "learning_rate": 1.377812297734628e-05,
      "loss": 0.0813,
      "step": 120160
    },
    {
      "epoch": 46.67,
      "learning_rate": 1.3777605177993529e-05,
      "loss": 0.0726,
      "step": 120170
    },
    {
      "epoch": 46.67,
      "learning_rate": 1.3777087378640778e-05,
      "loss": 0.1203,
      "step": 120180
    },
    {
      "epoch": 46.68,
      "learning_rate": 1.3776569579288026e-05,
      "loss": 0.0983,
      "step": 120190
    },
    {
      "epoch": 46.68,
      "learning_rate": 1.3776051779935276e-05,
      "loss": 0.1232,
      "step": 120200
    },
    {
      "epoch": 46.68,
      "learning_rate": 1.3775533980582525e-05,
      "loss": 0.0544,
      "step": 120210
    },
    {
      "epoch": 46.69,
      "learning_rate": 1.3775016181229775e-05,
      "loss": 0.0855,
      "step": 120220
    },
    {
      "epoch": 46.69,
      "learning_rate": 1.3774498381877025e-05,
      "loss": 0.0232,
      "step": 120230
    },
    {
      "epoch": 46.7,
      "learning_rate": 1.3773980582524273e-05,
      "loss": 0.0419,
      "step": 120240
    },
    {
      "epoch": 46.7,
      "learning_rate": 1.3773462783171522e-05,
      "loss": 0.0223,
      "step": 120250
    },
    {
      "epoch": 46.7,
      "learning_rate": 1.3772944983818772e-05,
      "loss": 0.1979,
      "step": 120260
    },
    {
      "epoch": 46.71,
      "learning_rate": 1.377242718446602e-05,
      "loss": 0.0691,
      "step": 120270
    },
    {
      "epoch": 46.71,
      "learning_rate": 1.377190938511327e-05,
      "loss": 0.0016,
      "step": 120280
    },
    {
      "epoch": 46.71,
      "learning_rate": 1.3771391585760519e-05,
      "loss": 0.0748,
      "step": 120290
    },
    {
      "epoch": 46.72,
      "learning_rate": 1.3770873786407769e-05,
      "loss": 0.1024,
      "step": 120300
    },
    {
      "epoch": 46.72,
      "learning_rate": 1.3770355987055018e-05,
      "loss": 0.1291,
      "step": 120310
    },
    {
      "epoch": 46.73,
      "learning_rate": 1.3769838187702266e-05,
      "loss": 0.1112,
      "step": 120320
    },
    {
      "epoch": 46.73,
      "learning_rate": 1.3769320388349516e-05,
      "loss": 0.1231,
      "step": 120330
    },
    {
      "epoch": 46.73,
      "learning_rate": 1.3768802588996765e-05,
      "loss": 0.0339,
      "step": 120340
    },
    {
      "epoch": 46.74,
      "learning_rate": 1.3768284789644013e-05,
      "loss": 0.0308,
      "step": 120350
    },
    {
      "epoch": 46.74,
      "learning_rate": 1.3767766990291263e-05,
      "loss": 0.0417,
      "step": 120360
    },
    {
      "epoch": 46.75,
      "learning_rate": 1.3767249190938513e-05,
      "loss": 0.1599,
      "step": 120370
    },
    {
      "epoch": 46.75,
      "learning_rate": 1.3766731391585762e-05,
      "loss": 0.0031,
      "step": 120380
    },
    {
      "epoch": 46.75,
      "learning_rate": 1.376621359223301e-05,
      "loss": 0.1504,
      "step": 120390
    },
    {
      "epoch": 46.76,
      "learning_rate": 1.376569579288026e-05,
      "loss": 0.0284,
      "step": 120400
    },
    {
      "epoch": 46.76,
      "learning_rate": 1.376517799352751e-05,
      "loss": 0.0949,
      "step": 120410
    },
    {
      "epoch": 46.77,
      "learning_rate": 1.3764660194174759e-05,
      "loss": 0.0286,
      "step": 120420
    },
    {
      "epoch": 46.77,
      "learning_rate": 1.3764142394822007e-05,
      "loss": 0.1601,
      "step": 120430
    },
    {
      "epoch": 46.77,
      "learning_rate": 1.3763624595469256e-05,
      "loss": 0.0361,
      "step": 120440
    },
    {
      "epoch": 46.78,
      "learning_rate": 1.3763106796116506e-05,
      "loss": 0.0823,
      "step": 120450
    },
    {
      "epoch": 46.78,
      "learning_rate": 1.3762588996763756e-05,
      "loss": 0.0502,
      "step": 120460
    },
    {
      "epoch": 46.78,
      "learning_rate": 1.3762071197411004e-05,
      "loss": 0.0823,
      "step": 120470
    },
    {
      "epoch": 46.79,
      "learning_rate": 1.3761553398058253e-05,
      "loss": 0.0271,
      "step": 120480
    },
    {
      "epoch": 46.79,
      "learning_rate": 1.3761035598705503e-05,
      "loss": 0.0288,
      "step": 120490
    },
    {
      "epoch": 46.8,
      "learning_rate": 1.3760517799352752e-05,
      "loss": 0.0834,
      "step": 120500
    },
    {
      "epoch": 46.8,
      "learning_rate": 1.376e-05,
      "loss": 0.1004,
      "step": 120510
    },
    {
      "epoch": 46.8,
      "learning_rate": 1.375948220064725e-05,
      "loss": 0.0761,
      "step": 120520
    },
    {
      "epoch": 46.81,
      "learning_rate": 1.37589644012945e-05,
      "loss": 0.1262,
      "step": 120530
    },
    {
      "epoch": 46.81,
      "learning_rate": 1.375844660194175e-05,
      "loss": 0.03,
      "step": 120540
    },
    {
      "epoch": 46.82,
      "learning_rate": 1.3757928802588997e-05,
      "loss": 0.0024,
      "step": 120550
    },
    {
      "epoch": 46.82,
      "learning_rate": 1.3757411003236247e-05,
      "loss": 0.146,
      "step": 120560
    },
    {
      "epoch": 46.82,
      "learning_rate": 1.3756893203883496e-05,
      "loss": 0.0598,
      "step": 120570
    },
    {
      "epoch": 46.83,
      "learning_rate": 1.3756375404530746e-05,
      "loss": 0.0499,
      "step": 120580
    },
    {
      "epoch": 46.83,
      "learning_rate": 1.3755857605177996e-05,
      "loss": 0.1578,
      "step": 120590
    },
    {
      "epoch": 46.83,
      "learning_rate": 1.3755339805825244e-05,
      "loss": 0.0459,
      "step": 120600
    },
    {
      "epoch": 46.84,
      "learning_rate": 1.3754822006472493e-05,
      "loss": 0.0938,
      "step": 120610
    },
    {
      "epoch": 46.84,
      "learning_rate": 1.3754304207119741e-05,
      "loss": 0.0326,
      "step": 120620
    },
    {
      "epoch": 46.85,
      "learning_rate": 1.375378640776699e-05,
      "loss": 0.1172,
      "step": 120630
    },
    {
      "epoch": 46.85,
      "learning_rate": 1.375326860841424e-05,
      "loss": 0.075,
      "step": 120640
    },
    {
      "epoch": 46.85,
      "learning_rate": 1.375275080906149e-05,
      "loss": 0.0238,
      "step": 120650
    },
    {
      "epoch": 46.86,
      "learning_rate": 1.375223300970874e-05,
      "loss": 0.041,
      "step": 120660
    },
    {
      "epoch": 46.86,
      "learning_rate": 1.375171521035599e-05,
      "loss": 0.0608,
      "step": 120670
    },
    {
      "epoch": 46.87,
      "learning_rate": 1.3751197411003237e-05,
      "loss": 0.0704,
      "step": 120680
    },
    {
      "epoch": 46.87,
      "learning_rate": 1.3750679611650487e-05,
      "loss": 0.1618,
      "step": 120690
    },
    {
      "epoch": 46.87,
      "learning_rate": 1.3750161812297735e-05,
      "loss": 0.0539,
      "step": 120700
    },
    {
      "epoch": 46.88,
      "learning_rate": 1.3749644012944984e-05,
      "loss": 0.1621,
      "step": 120710
    },
    {
      "epoch": 46.88,
      "learning_rate": 1.3749126213592234e-05,
      "loss": 0.0229,
      "step": 120720
    },
    {
      "epoch": 46.89,
      "learning_rate": 1.3748608414239484e-05,
      "loss": 0.0604,
      "step": 120730
    },
    {
      "epoch": 46.89,
      "learning_rate": 1.3748090614886733e-05,
      "loss": 0.1737,
      "step": 120740
    },
    {
      "epoch": 46.89,
      "learning_rate": 1.3747572815533983e-05,
      "loss": 0.0282,
      "step": 120750
    },
    {
      "epoch": 46.9,
      "learning_rate": 1.374705501618123e-05,
      "loss": 0.0421,
      "step": 120760
    },
    {
      "epoch": 46.9,
      "learning_rate": 1.3746537216828479e-05,
      "loss": 0.0913,
      "step": 120770
    },
    {
      "epoch": 46.9,
      "learning_rate": 1.3746019417475728e-05,
      "loss": 0.0692,
      "step": 120780
    },
    {
      "epoch": 46.91,
      "learning_rate": 1.3745501618122978e-05,
      "loss": 0.0438,
      "step": 120790
    },
    {
      "epoch": 46.91,
      "learning_rate": 1.3744983818770227e-05,
      "loss": 0.0529,
      "step": 120800
    },
    {
      "epoch": 46.92,
      "learning_rate": 1.3744466019417477e-05,
      "loss": 0.0004,
      "step": 120810
    },
    {
      "epoch": 46.92,
      "learning_rate": 1.3743948220064727e-05,
      "loss": 0.0546,
      "step": 120820
    },
    {
      "epoch": 46.92,
      "learning_rate": 1.3743430420711976e-05,
      "loss": 0.02,
      "step": 120830
    },
    {
      "epoch": 46.93,
      "learning_rate": 1.3742912621359224e-05,
      "loss": 0.0799,
      "step": 120840
    },
    {
      "epoch": 46.93,
      "learning_rate": 1.3742394822006472e-05,
      "loss": 0.0964,
      "step": 120850
    },
    {
      "epoch": 46.94,
      "learning_rate": 1.3741877022653722e-05,
      "loss": 0.1591,
      "step": 120860
    },
    {
      "epoch": 46.94,
      "learning_rate": 1.3741359223300971e-05,
      "loss": 0.1087,
      "step": 120870
    },
    {
      "epoch": 46.94,
      "learning_rate": 1.3740841423948221e-05,
      "loss": 0.13,
      "step": 120880
    },
    {
      "epoch": 46.95,
      "learning_rate": 1.374032362459547e-05,
      "loss": 0.027,
      "step": 120890
    },
    {
      "epoch": 46.95,
      "learning_rate": 1.373980582524272e-05,
      "loss": 0.1496,
      "step": 120900
    },
    {
      "epoch": 46.96,
      "learning_rate": 1.373928802588997e-05,
      "loss": 0.0344,
      "step": 120910
    },
    {
      "epoch": 46.96,
      "learning_rate": 1.3738770226537216e-05,
      "loss": 0.0211,
      "step": 120920
    },
    {
      "epoch": 46.96,
      "learning_rate": 1.3738252427184466e-05,
      "loss": 0.0177,
      "step": 120930
    },
    {
      "epoch": 46.97,
      "learning_rate": 1.3737734627831715e-05,
      "loss": 0.0789,
      "step": 120940
    },
    {
      "epoch": 46.97,
      "learning_rate": 1.3737216828478965e-05,
      "loss": 0.0571,
      "step": 120950
    },
    {
      "epoch": 46.97,
      "learning_rate": 1.3736699029126215e-05,
      "loss": 0.0665,
      "step": 120960
    },
    {
      "epoch": 46.98,
      "learning_rate": 1.3736181229773464e-05,
      "loss": 0.0121,
      "step": 120970
    },
    {
      "epoch": 46.98,
      "learning_rate": 1.3735663430420714e-05,
      "loss": 0.0716,
      "step": 120980
    },
    {
      "epoch": 46.99,
      "learning_rate": 1.3735145631067963e-05,
      "loss": 0.0427,
      "step": 120990
    },
    {
      "epoch": 46.99,
      "learning_rate": 1.373462783171521e-05,
      "loss": 0.143,
      "step": 121000
    },
    {
      "epoch": 46.99,
      "learning_rate": 1.373411003236246e-05,
      "loss": 0.0167,
      "step": 121010
    },
    {
      "epoch": 47.0,
      "learning_rate": 1.3733592233009709e-05,
      "loss": 0.0477,
      "step": 121020
    },
    {
      "epoch": 47.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3004183769226074,
      "eval_runtime": 8.2571,
      "eval_samples_per_second": 440.227,
      "eval_steps_per_second": 55.104,
      "step": 121025
    },
    {
      "epoch": 47.0,
      "learning_rate": 1.3733074433656959e-05,
      "loss": 0.1371,
      "step": 121030
    },
    {
      "epoch": 47.01,
      "learning_rate": 1.3732556634304208e-05,
      "loss": 0.1228,
      "step": 121040
    },
    {
      "epoch": 47.01,
      "learning_rate": 1.3732038834951458e-05,
      "loss": 0.0104,
      "step": 121050
    },
    {
      "epoch": 47.01,
      "learning_rate": 1.3731521035598707e-05,
      "loss": 0.1207,
      "step": 121060
    },
    {
      "epoch": 47.02,
      "learning_rate": 1.3731003236245957e-05,
      "loss": 0.093,
      "step": 121070
    },
    {
      "epoch": 47.02,
      "learning_rate": 1.3730485436893203e-05,
      "loss": 0.0332,
      "step": 121080
    },
    {
      "epoch": 47.03,
      "learning_rate": 1.3729967637540453e-05,
      "loss": 0.1466,
      "step": 121090
    },
    {
      "epoch": 47.03,
      "learning_rate": 1.3729449838187702e-05,
      "loss": 0.0632,
      "step": 121100
    },
    {
      "epoch": 47.03,
      "learning_rate": 1.3728932038834952e-05,
      "loss": 0.041,
      "step": 121110
    },
    {
      "epoch": 47.04,
      "learning_rate": 1.3728414239482202e-05,
      "loss": 0.1267,
      "step": 121120
    },
    {
      "epoch": 47.04,
      "learning_rate": 1.3727896440129451e-05,
      "loss": 0.0043,
      "step": 121130
    },
    {
      "epoch": 47.04,
      "learning_rate": 1.3727378640776701e-05,
      "loss": 0.0776,
      "step": 121140
    },
    {
      "epoch": 47.05,
      "learning_rate": 1.372686084142395e-05,
      "loss": 0.2181,
      "step": 121150
    },
    {
      "epoch": 47.05,
      "learning_rate": 1.37263430420712e-05,
      "loss": 0.0791,
      "step": 121160
    },
    {
      "epoch": 47.06,
      "learning_rate": 1.3725825242718446e-05,
      "loss": 0.1001,
      "step": 121170
    },
    {
      "epoch": 47.06,
      "learning_rate": 1.3725307443365696e-05,
      "loss": 0.0686,
      "step": 121180
    },
    {
      "epoch": 47.06,
      "learning_rate": 1.3724789644012946e-05,
      "loss": 0.278,
      "step": 121190
    },
    {
      "epoch": 47.07,
      "learning_rate": 1.3724271844660195e-05,
      "loss": 0.0943,
      "step": 121200
    },
    {
      "epoch": 47.07,
      "learning_rate": 1.3723754045307445e-05,
      "loss": 0.0034,
      "step": 121210
    },
    {
      "epoch": 47.08,
      "learning_rate": 1.3723236245954694e-05,
      "loss": 0.0732,
      "step": 121220
    },
    {
      "epoch": 47.08,
      "learning_rate": 1.3722718446601944e-05,
      "loss": 0.1479,
      "step": 121230
    },
    {
      "epoch": 47.08,
      "learning_rate": 1.3722200647249194e-05,
      "loss": 0.0813,
      "step": 121240
    },
    {
      "epoch": 47.09,
      "learning_rate": 1.372168284789644e-05,
      "loss": 0.0418,
      "step": 121250
    },
    {
      "epoch": 47.09,
      "learning_rate": 1.372116504854369e-05,
      "loss": 0.1738,
      "step": 121260
    },
    {
      "epoch": 47.1,
      "learning_rate": 1.372064724919094e-05,
      "loss": 0.2138,
      "step": 121270
    },
    {
      "epoch": 47.1,
      "learning_rate": 1.3720129449838189e-05,
      "loss": 0.1742,
      "step": 121280
    },
    {
      "epoch": 47.1,
      "learning_rate": 1.3719611650485438e-05,
      "loss": 0.0257,
      "step": 121290
    },
    {
      "epoch": 47.11,
      "learning_rate": 1.3719093851132688e-05,
      "loss": 0.1741,
      "step": 121300
    },
    {
      "epoch": 47.11,
      "learning_rate": 1.3718576051779938e-05,
      "loss": 0.0047,
      "step": 121310
    },
    {
      "epoch": 47.11,
      "learning_rate": 1.3718058252427187e-05,
      "loss": 0.0185,
      "step": 121320
    },
    {
      "epoch": 47.12,
      "learning_rate": 1.3717540453074433e-05,
      "loss": 0.1538,
      "step": 121330
    },
    {
      "epoch": 47.12,
      "learning_rate": 1.3717022653721683e-05,
      "loss": 0.0455,
      "step": 121340
    },
    {
      "epoch": 47.13,
      "learning_rate": 1.3716504854368933e-05,
      "loss": 0.0921,
      "step": 121350
    },
    {
      "epoch": 47.13,
      "learning_rate": 1.3715987055016182e-05,
      "loss": 0.1305,
      "step": 121360
    },
    {
      "epoch": 47.13,
      "learning_rate": 1.3715469255663432e-05,
      "loss": 0.0091,
      "step": 121370
    },
    {
      "epoch": 47.14,
      "learning_rate": 1.3714951456310682e-05,
      "loss": 0.094,
      "step": 121380
    },
    {
      "epoch": 47.14,
      "learning_rate": 1.3714433656957931e-05,
      "loss": 0.0744,
      "step": 121390
    },
    {
      "epoch": 47.15,
      "learning_rate": 1.371391585760518e-05,
      "loss": 0.0416,
      "step": 121400
    },
    {
      "epoch": 47.15,
      "learning_rate": 1.3713398058252427e-05,
      "loss": 0.0254,
      "step": 121410
    },
    {
      "epoch": 47.15,
      "learning_rate": 1.3712880258899677e-05,
      "loss": 0.0472,
      "step": 121420
    },
    {
      "epoch": 47.16,
      "learning_rate": 1.3712362459546926e-05,
      "loss": 0.1175,
      "step": 121430
    },
    {
      "epoch": 47.16,
      "learning_rate": 1.3711844660194176e-05,
      "loss": 0.0746,
      "step": 121440
    },
    {
      "epoch": 47.17,
      "learning_rate": 1.3711326860841426e-05,
      "loss": 0.1241,
      "step": 121450
    },
    {
      "epoch": 47.17,
      "learning_rate": 1.3710809061488675e-05,
      "loss": 0.1389,
      "step": 121460
    },
    {
      "epoch": 47.17,
      "learning_rate": 1.3710291262135925e-05,
      "loss": 0.1224,
      "step": 121470
    },
    {
      "epoch": 47.18,
      "learning_rate": 1.3709773462783174e-05,
      "loss": 0.1481,
      "step": 121480
    },
    {
      "epoch": 47.18,
      "learning_rate": 1.370925566343042e-05,
      "loss": 0.034,
      "step": 121490
    },
    {
      "epoch": 47.18,
      "learning_rate": 1.370873786407767e-05,
      "loss": 0.0778,
      "step": 121500
    },
    {
      "epoch": 47.19,
      "learning_rate": 1.370822006472492e-05,
      "loss": 0.0572,
      "step": 121510
    },
    {
      "epoch": 47.19,
      "learning_rate": 1.370770226537217e-05,
      "loss": 0.0099,
      "step": 121520
    },
    {
      "epoch": 47.2,
      "learning_rate": 1.3707184466019419e-05,
      "loss": 0.031,
      "step": 121530
    },
    {
      "epoch": 47.2,
      "learning_rate": 1.3706666666666669e-05,
      "loss": 0.1592,
      "step": 121540
    },
    {
      "epoch": 47.2,
      "learning_rate": 1.3706148867313918e-05,
      "loss": 0.013,
      "step": 121550
    },
    {
      "epoch": 47.21,
      "learning_rate": 1.3705631067961166e-05,
      "loss": 0.0109,
      "step": 121560
    },
    {
      "epoch": 47.21,
      "learning_rate": 1.3705113268608414e-05,
      "loss": 0.071,
      "step": 121570
    },
    {
      "epoch": 47.22,
      "learning_rate": 1.3704595469255664e-05,
      "loss": 0.0339,
      "step": 121580
    },
    {
      "epoch": 47.22,
      "learning_rate": 1.3704077669902913e-05,
      "loss": 0.0363,
      "step": 121590
    },
    {
      "epoch": 47.22,
      "learning_rate": 1.3703559870550163e-05,
      "loss": 0.0596,
      "step": 121600
    },
    {
      "epoch": 47.23,
      "learning_rate": 1.3703042071197413e-05,
      "loss": 0.0996,
      "step": 121610
    },
    {
      "epoch": 47.23,
      "learning_rate": 1.3702524271844662e-05,
      "loss": 0.1272,
      "step": 121620
    },
    {
      "epoch": 47.23,
      "learning_rate": 1.3702006472491912e-05,
      "loss": 0.1305,
      "step": 121630
    },
    {
      "epoch": 47.24,
      "learning_rate": 1.370148867313916e-05,
      "loss": 0.1203,
      "step": 121640
    },
    {
      "epoch": 47.24,
      "learning_rate": 1.370097087378641e-05,
      "loss": 0.1075,
      "step": 121650
    },
    {
      "epoch": 47.25,
      "learning_rate": 1.3700453074433657e-05,
      "loss": 0.0821,
      "step": 121660
    },
    {
      "epoch": 47.25,
      "learning_rate": 1.3699935275080907e-05,
      "loss": 0.0819,
      "step": 121670
    },
    {
      "epoch": 47.25,
      "learning_rate": 1.3699417475728157e-05,
      "loss": 0.1786,
      "step": 121680
    },
    {
      "epoch": 47.26,
      "learning_rate": 1.3698899676375406e-05,
      "loss": 0.0382,
      "step": 121690
    },
    {
      "epoch": 47.26,
      "learning_rate": 1.3698381877022656e-05,
      "loss": 0.2385,
      "step": 121700
    },
    {
      "epoch": 47.27,
      "learning_rate": 1.3697864077669904e-05,
      "loss": 0.043,
      "step": 121710
    },
    {
      "epoch": 47.27,
      "learning_rate": 1.3697346278317153e-05,
      "loss": 0.0116,
      "step": 121720
    },
    {
      "epoch": 47.27,
      "learning_rate": 1.3696828478964403e-05,
      "loss": 0.0952,
      "step": 121730
    },
    {
      "epoch": 47.28,
      "learning_rate": 1.3696310679611651e-05,
      "loss": 0.0347,
      "step": 121740
    },
    {
      "epoch": 47.28,
      "learning_rate": 1.36957928802589e-05,
      "loss": 0.1355,
      "step": 121750
    },
    {
      "epoch": 47.29,
      "learning_rate": 1.369527508090615e-05,
      "loss": 0.0596,
      "step": 121760
    },
    {
      "epoch": 47.29,
      "learning_rate": 1.36947572815534e-05,
      "loss": 0.1499,
      "step": 121770
    },
    {
      "epoch": 47.29,
      "learning_rate": 1.369423948220065e-05,
      "loss": 0.0248,
      "step": 121780
    },
    {
      "epoch": 47.3,
      "learning_rate": 1.3693721682847897e-05,
      "loss": 0.0949,
      "step": 121790
    },
    {
      "epoch": 47.3,
      "learning_rate": 1.3693203883495147e-05,
      "loss": 0.075,
      "step": 121800
    },
    {
      "epoch": 47.3,
      "learning_rate": 1.3692686084142397e-05,
      "loss": 0.149,
      "step": 121810
    },
    {
      "epoch": 47.31,
      "learning_rate": 1.3692168284789644e-05,
      "loss": 0.2598,
      "step": 121820
    },
    {
      "epoch": 47.31,
      "learning_rate": 1.3691650485436894e-05,
      "loss": 0.1062,
      "step": 121830
    },
    {
      "epoch": 47.32,
      "learning_rate": 1.3691132686084144e-05,
      "loss": 0.0016,
      "step": 121840
    },
    {
      "epoch": 47.32,
      "learning_rate": 1.3690614886731393e-05,
      "loss": 0.1714,
      "step": 121850
    },
    {
      "epoch": 47.32,
      "learning_rate": 1.3690097087378641e-05,
      "loss": 0.0502,
      "step": 121860
    },
    {
      "epoch": 47.33,
      "learning_rate": 1.3689579288025891e-05,
      "loss": 0.0006,
      "step": 121870
    },
    {
      "epoch": 47.33,
      "learning_rate": 1.368906148867314e-05,
      "loss": 0.0922,
      "step": 121880
    },
    {
      "epoch": 47.34,
      "learning_rate": 1.368854368932039e-05,
      "loss": 0.0422,
      "step": 121890
    },
    {
      "epoch": 47.34,
      "learning_rate": 1.3688025889967638e-05,
      "loss": 0.1023,
      "step": 121900
    },
    {
      "epoch": 47.34,
      "learning_rate": 1.3687508090614888e-05,
      "loss": 0.1463,
      "step": 121910
    },
    {
      "epoch": 47.35,
      "learning_rate": 1.3686990291262137e-05,
      "loss": 0.1508,
      "step": 121920
    },
    {
      "epoch": 47.35,
      "learning_rate": 1.3686472491909387e-05,
      "loss": 0.0694,
      "step": 121930
    },
    {
      "epoch": 47.36,
      "learning_rate": 1.3685954692556635e-05,
      "loss": 0.0173,
      "step": 121940
    },
    {
      "epoch": 47.36,
      "learning_rate": 1.3685436893203884e-05,
      "loss": 0.0255,
      "step": 121950
    },
    {
      "epoch": 47.36,
      "learning_rate": 1.3684919093851134e-05,
      "loss": 0.0446,
      "step": 121960
    },
    {
      "epoch": 47.37,
      "learning_rate": 1.3684401294498384e-05,
      "loss": 0.1093,
      "step": 121970
    },
    {
      "epoch": 47.37,
      "learning_rate": 1.3683883495145632e-05,
      "loss": 0.2535,
      "step": 121980
    },
    {
      "epoch": 47.37,
      "learning_rate": 1.3683365695792881e-05,
      "loss": 0.0037,
      "step": 121990
    },
    {
      "epoch": 47.38,
      "learning_rate": 1.368284789644013e-05,
      "loss": 0.136,
      "step": 122000
    },
    {
      "epoch": 47.38,
      "learning_rate": 1.368233009708738e-05,
      "loss": 0.0708,
      "step": 122010
    },
    {
      "epoch": 47.39,
      "learning_rate": 1.3681812297734628e-05,
      "loss": 0.0052,
      "step": 122020
    },
    {
      "epoch": 47.39,
      "learning_rate": 1.3681294498381878e-05,
      "loss": 0.0027,
      "step": 122030
    },
    {
      "epoch": 47.39,
      "learning_rate": 1.3680776699029128e-05,
      "loss": 0.0638,
      "step": 122040
    },
    {
      "epoch": 47.4,
      "learning_rate": 1.3680258899676377e-05,
      "loss": 0.032,
      "step": 122050
    },
    {
      "epoch": 47.4,
      "learning_rate": 1.3679741100323625e-05,
      "loss": 0.0005,
      "step": 122060
    },
    {
      "epoch": 47.41,
      "learning_rate": 1.3679223300970875e-05,
      "loss": 0.0252,
      "step": 122070
    },
    {
      "epoch": 47.41,
      "learning_rate": 1.3678705501618124e-05,
      "loss": 0.0078,
      "step": 122080
    },
    {
      "epoch": 47.41,
      "learning_rate": 1.3678187702265372e-05,
      "loss": 0.0832,
      "step": 122090
    },
    {
      "epoch": 47.42,
      "learning_rate": 1.3677669902912622e-05,
      "loss": 0.1648,
      "step": 122100
    },
    {
      "epoch": 47.42,
      "learning_rate": 1.3677152103559872e-05,
      "loss": 0.0128,
      "step": 122110
    },
    {
      "epoch": 47.43,
      "learning_rate": 1.3676634304207121e-05,
      "loss": 0.0839,
      "step": 122120
    },
    {
      "epoch": 47.43,
      "learning_rate": 1.367611650485437e-05,
      "loss": 0.1285,
      "step": 122130
    },
    {
      "epoch": 47.43,
      "learning_rate": 1.3675598705501619e-05,
      "loss": 0.0202,
      "step": 122140
    },
    {
      "epoch": 47.44,
      "learning_rate": 1.3675080906148868e-05,
      "loss": 0.0689,
      "step": 122150
    },
    {
      "epoch": 47.44,
      "learning_rate": 1.3674563106796118e-05,
      "loss": 0.0645,
      "step": 122160
    },
    {
      "epoch": 47.44,
      "learning_rate": 1.3674045307443366e-05,
      "loss": 0.0404,
      "step": 122170
    },
    {
      "epoch": 47.45,
      "learning_rate": 1.3673527508090615e-05,
      "loss": 0.0626,
      "step": 122180
    },
    {
      "epoch": 47.45,
      "learning_rate": 1.3673009708737865e-05,
      "loss": 0.1205,
      "step": 122190
    },
    {
      "epoch": 47.46,
      "learning_rate": 1.3672491909385115e-05,
      "loss": 0.0424,
      "step": 122200
    },
    {
      "epoch": 47.46,
      "learning_rate": 1.3671974110032364e-05,
      "loss": 0.0212,
      "step": 122210
    },
    {
      "epoch": 47.46,
      "learning_rate": 1.3671456310679614e-05,
      "loss": 0.1278,
      "step": 122220
    },
    {
      "epoch": 47.47,
      "learning_rate": 1.3670938511326862e-05,
      "loss": 0.1588,
      "step": 122230
    },
    {
      "epoch": 47.47,
      "learning_rate": 1.367042071197411e-05,
      "loss": 0.1026,
      "step": 122240
    },
    {
      "epoch": 47.48,
      "learning_rate": 1.366990291262136e-05,
      "loss": 0.0783,
      "step": 122250
    },
    {
      "epoch": 47.48,
      "learning_rate": 1.3669385113268609e-05,
      "loss": 0.049,
      "step": 122260
    },
    {
      "epoch": 47.48,
      "learning_rate": 1.3668867313915859e-05,
      "loss": 0.0639,
      "step": 122270
    },
    {
      "epoch": 47.49,
      "learning_rate": 1.3668349514563108e-05,
      "loss": 0.0984,
      "step": 122280
    },
    {
      "epoch": 47.49,
      "learning_rate": 1.3667831715210358e-05,
      "loss": 0.1124,
      "step": 122290
    },
    {
      "epoch": 47.5,
      "learning_rate": 1.3667313915857607e-05,
      "loss": 0.0511,
      "step": 122300
    },
    {
      "epoch": 47.5,
      "learning_rate": 1.3666796116504855e-05,
      "loss": 0.057,
      "step": 122310
    },
    {
      "epoch": 47.5,
      "learning_rate": 1.3666278317152103e-05,
      "loss": 0.1541,
      "step": 122320
    },
    {
      "epoch": 47.51,
      "learning_rate": 1.3665760517799353e-05,
      "loss": 0.057,
      "step": 122330
    },
    {
      "epoch": 47.51,
      "learning_rate": 1.3665242718446603e-05,
      "loss": 0.2141,
      "step": 122340
    },
    {
      "epoch": 47.51,
      "learning_rate": 1.3664724919093852e-05,
      "loss": 0.2975,
      "step": 122350
    },
    {
      "epoch": 47.52,
      "learning_rate": 1.3664207119741102e-05,
      "loss": 0.1058,
      "step": 122360
    },
    {
      "epoch": 47.52,
      "learning_rate": 1.3663689320388351e-05,
      "loss": 0.0886,
      "step": 122370
    },
    {
      "epoch": 47.53,
      "learning_rate": 1.3663171521035601e-05,
      "loss": 0.0711,
      "step": 122380
    },
    {
      "epoch": 47.53,
      "learning_rate": 1.3662653721682847e-05,
      "loss": 0.0068,
      "step": 122390
    },
    {
      "epoch": 47.53,
      "learning_rate": 1.3662135922330097e-05,
      "loss": 0.0533,
      "step": 122400
    },
    {
      "epoch": 47.54,
      "learning_rate": 1.3661618122977347e-05,
      "loss": 0.0248,
      "step": 122410
    },
    {
      "epoch": 47.54,
      "learning_rate": 1.3661100323624596e-05,
      "loss": 0.1673,
      "step": 122420
    },
    {
      "epoch": 47.55,
      "learning_rate": 1.3660582524271846e-05,
      "loss": 0.0899,
      "step": 122430
    },
    {
      "epoch": 47.55,
      "learning_rate": 1.3660064724919095e-05,
      "loss": 0.1845,
      "step": 122440
    },
    {
      "epoch": 47.55,
      "learning_rate": 1.3659546925566345e-05,
      "loss": 0.0553,
      "step": 122450
    },
    {
      "epoch": 47.56,
      "learning_rate": 1.3659029126213595e-05,
      "loss": 0.126,
      "step": 122460
    },
    {
      "epoch": 47.56,
      "learning_rate": 1.365851132686084e-05,
      "loss": 0.0339,
      "step": 122470
    },
    {
      "epoch": 47.57,
      "learning_rate": 1.365799352750809e-05,
      "loss": 0.0196,
      "step": 122480
    },
    {
      "epoch": 47.57,
      "learning_rate": 1.365747572815534e-05,
      "loss": 0.0102,
      "step": 122490
    },
    {
      "epoch": 47.57,
      "learning_rate": 1.365695792880259e-05,
      "loss": 0.1043,
      "step": 122500
    },
    {
      "epoch": 47.58,
      "learning_rate": 1.365644012944984e-05,
      "loss": 0.0133,
      "step": 122510
    },
    {
      "epoch": 47.58,
      "learning_rate": 1.3655922330097089e-05,
      "loss": 0.0119,
      "step": 122520
    },
    {
      "epoch": 47.58,
      "learning_rate": 1.3655404530744339e-05,
      "loss": 0.0269,
      "step": 122530
    },
    {
      "epoch": 47.59,
      "learning_rate": 1.3654886731391588e-05,
      "loss": 0.0853,
      "step": 122540
    },
    {
      "epoch": 47.59,
      "learning_rate": 1.3654368932038834e-05,
      "loss": 0.0534,
      "step": 122550
    },
    {
      "epoch": 47.6,
      "learning_rate": 1.3653851132686084e-05,
      "loss": 0.0498,
      "step": 122560
    },
    {
      "epoch": 47.6,
      "learning_rate": 1.3653333333333334e-05,
      "loss": 0.0173,
      "step": 122570
    },
    {
      "epoch": 47.6,
      "learning_rate": 1.3652815533980583e-05,
      "loss": 0.0946,
      "step": 122580
    },
    {
      "epoch": 47.61,
      "learning_rate": 1.3652297734627833e-05,
      "loss": 0.1585,
      "step": 122590
    },
    {
      "epoch": 47.61,
      "learning_rate": 1.3651779935275082e-05,
      "loss": 0.0645,
      "step": 122600
    },
    {
      "epoch": 47.62,
      "learning_rate": 1.3651262135922332e-05,
      "loss": 0.1473,
      "step": 122610
    },
    {
      "epoch": 47.62,
      "learning_rate": 1.3650744336569582e-05,
      "loss": 0.1521,
      "step": 122620
    },
    {
      "epoch": 47.62,
      "learning_rate": 1.3650226537216828e-05,
      "loss": 0.0608,
      "step": 122630
    },
    {
      "epoch": 47.63,
      "learning_rate": 1.3649708737864078e-05,
      "loss": 0.1303,
      "step": 122640
    },
    {
      "epoch": 47.63,
      "learning_rate": 1.3649190938511327e-05,
      "loss": 0.0778,
      "step": 122650
    },
    {
      "epoch": 47.63,
      "learning_rate": 1.3648673139158577e-05,
      "loss": 0.087,
      "step": 122660
    },
    {
      "epoch": 47.64,
      "learning_rate": 1.3648155339805826e-05,
      "loss": 0.0299,
      "step": 122670
    },
    {
      "epoch": 47.64,
      "learning_rate": 1.3647637540453076e-05,
      "loss": 0.0314,
      "step": 122680
    },
    {
      "epoch": 47.65,
      "learning_rate": 1.3647119741100326e-05,
      "loss": 0.1367,
      "step": 122690
    },
    {
      "epoch": 47.65,
      "learning_rate": 1.3646601941747575e-05,
      "loss": 0.0128,
      "step": 122700
    },
    {
      "epoch": 47.65,
      "learning_rate": 1.3646084142394821e-05,
      "loss": 0.1053,
      "step": 122710
    },
    {
      "epoch": 47.66,
      "learning_rate": 1.3645566343042071e-05,
      "loss": 0.0099,
      "step": 122720
    },
    {
      "epoch": 47.66,
      "learning_rate": 1.364504854368932e-05,
      "loss": 0.1114,
      "step": 122730
    },
    {
      "epoch": 47.67,
      "learning_rate": 1.364453074433657e-05,
      "loss": 0.0359,
      "step": 122740
    },
    {
      "epoch": 47.67,
      "learning_rate": 1.364401294498382e-05,
      "loss": 0.0119,
      "step": 122750
    },
    {
      "epoch": 47.67,
      "learning_rate": 1.364349514563107e-05,
      "loss": 0.0754,
      "step": 122760
    },
    {
      "epoch": 47.68,
      "learning_rate": 1.364297734627832e-05,
      "loss": 0.0996,
      "step": 122770
    },
    {
      "epoch": 47.68,
      "learning_rate": 1.3642459546925569e-05,
      "loss": 0.0106,
      "step": 122780
    },
    {
      "epoch": 47.69,
      "learning_rate": 1.3641941747572818e-05,
      "loss": 0.1583,
      "step": 122790
    },
    {
      "epoch": 47.69,
      "learning_rate": 1.3641423948220065e-05,
      "loss": 0.1134,
      "step": 122800
    },
    {
      "epoch": 47.69,
      "learning_rate": 1.3640906148867314e-05,
      "loss": 0.1062,
      "step": 122810
    },
    {
      "epoch": 47.7,
      "learning_rate": 1.3640388349514564e-05,
      "loss": 0.0031,
      "step": 122820
    },
    {
      "epoch": 47.7,
      "learning_rate": 1.3639870550161814e-05,
      "loss": 0.0053,
      "step": 122830
    },
    {
      "epoch": 47.7,
      "learning_rate": 1.3639352750809063e-05,
      "loss": 0.1302,
      "step": 122840
    },
    {
      "epoch": 47.71,
      "learning_rate": 1.3638834951456313e-05,
      "loss": 0.0067,
      "step": 122850
    },
    {
      "epoch": 47.71,
      "learning_rate": 1.3638317152103562e-05,
      "loss": 0.0644,
      "step": 122860
    },
    {
      "epoch": 47.72,
      "learning_rate": 1.3637799352750812e-05,
      "loss": 0.0106,
      "step": 122870
    },
    {
      "epoch": 47.72,
      "learning_rate": 1.3637281553398058e-05,
      "loss": 0.2422,
      "step": 122880
    },
    {
      "epoch": 47.72,
      "learning_rate": 1.3636763754045308e-05,
      "loss": 0.0761,
      "step": 122890
    },
    {
      "epoch": 47.73,
      "learning_rate": 1.3636245954692557e-05,
      "loss": 0.0963,
      "step": 122900
    },
    {
      "epoch": 47.73,
      "learning_rate": 1.3635728155339807e-05,
      "loss": 0.1635,
      "step": 122910
    },
    {
      "epoch": 47.74,
      "learning_rate": 1.3635210355987057e-05,
      "loss": 0.175,
      "step": 122920
    },
    {
      "epoch": 47.74,
      "learning_rate": 1.3634692556634306e-05,
      "loss": 0.0721,
      "step": 122930
    },
    {
      "epoch": 47.74,
      "learning_rate": 1.3634174757281556e-05,
      "loss": 0.087,
      "step": 122940
    },
    {
      "epoch": 47.75,
      "learning_rate": 1.3633656957928806e-05,
      "loss": 0.0063,
      "step": 122950
    },
    {
      "epoch": 47.75,
      "learning_rate": 1.3633139158576052e-05,
      "loss": 0.0441,
      "step": 122960
    },
    {
      "epoch": 47.76,
      "learning_rate": 1.3632621359223301e-05,
      "loss": 0.0377,
      "step": 122970
    },
    {
      "epoch": 47.76,
      "learning_rate": 1.3632103559870551e-05,
      "loss": 0.1396,
      "step": 122980
    },
    {
      "epoch": 47.76,
      "learning_rate": 1.36315857605178e-05,
      "loss": 0.021,
      "step": 122990
    },
    {
      "epoch": 47.77,
      "learning_rate": 1.363106796116505e-05,
      "loss": 0.0529,
      "step": 123000
    },
    {
      "epoch": 47.77,
      "learning_rate": 1.36305501618123e-05,
      "loss": 0.0779,
      "step": 123010
    },
    {
      "epoch": 47.77,
      "learning_rate": 1.363003236245955e-05,
      "loss": 0.245,
      "step": 123020
    },
    {
      "epoch": 47.78,
      "learning_rate": 1.3629514563106797e-05,
      "loss": 0.0212,
      "step": 123030
    },
    {
      "epoch": 47.78,
      "learning_rate": 1.3628996763754045e-05,
      "loss": 0.0071,
      "step": 123040
    },
    {
      "epoch": 47.79,
      "learning_rate": 1.3628478964401295e-05,
      "loss": 0.1594,
      "step": 123050
    },
    {
      "epoch": 47.79,
      "learning_rate": 1.3627961165048545e-05,
      "loss": 0.0333,
      "step": 123060
    },
    {
      "epoch": 47.79,
      "learning_rate": 1.3627443365695794e-05,
      "loss": 0.0095,
      "step": 123070
    },
    {
      "epoch": 47.8,
      "learning_rate": 1.3626925566343044e-05,
      "loss": 0.0181,
      "step": 123080
    },
    {
      "epoch": 47.8,
      "learning_rate": 1.3626407766990293e-05,
      "loss": 0.0899,
      "step": 123090
    },
    {
      "epoch": 47.81,
      "learning_rate": 1.3625889967637543e-05,
      "loss": 0.1647,
      "step": 123100
    },
    {
      "epoch": 47.81,
      "learning_rate": 1.3625372168284791e-05,
      "loss": 0.1119,
      "step": 123110
    },
    {
      "epoch": 47.81,
      "learning_rate": 1.3624854368932039e-05,
      "loss": 0.1516,
      "step": 123120
    },
    {
      "epoch": 47.82,
      "learning_rate": 1.3624336569579289e-05,
      "loss": 0.1561,
      "step": 123130
    },
    {
      "epoch": 47.82,
      "learning_rate": 1.3623818770226538e-05,
      "loss": 0.0229,
      "step": 123140
    },
    {
      "epoch": 47.83,
      "learning_rate": 1.3623300970873788e-05,
      "loss": 0.1076,
      "step": 123150
    },
    {
      "epoch": 47.83,
      "learning_rate": 1.3622783171521037e-05,
      "loss": 0.1757,
      "step": 123160
    },
    {
      "epoch": 47.83,
      "learning_rate": 1.3622265372168287e-05,
      "loss": 0.061,
      "step": 123170
    },
    {
      "epoch": 47.84,
      "learning_rate": 1.3621747572815535e-05,
      "loss": 0.1153,
      "step": 123180
    },
    {
      "epoch": 47.84,
      "learning_rate": 1.3621229773462785e-05,
      "loss": 0.0342,
      "step": 123190
    },
    {
      "epoch": 47.84,
      "learning_rate": 1.3620711974110032e-05,
      "loss": 0.1088,
      "step": 123200
    },
    {
      "epoch": 47.85,
      "learning_rate": 1.3620194174757282e-05,
      "loss": 0.0344,
      "step": 123210
    },
    {
      "epoch": 47.85,
      "learning_rate": 1.3619676375404532e-05,
      "loss": 0.1193,
      "step": 123220
    },
    {
      "epoch": 47.86,
      "learning_rate": 1.3619158576051781e-05,
      "loss": 0.1406,
      "step": 123230
    },
    {
      "epoch": 47.86,
      "learning_rate": 1.3618640776699031e-05,
      "loss": 0.1434,
      "step": 123240
    },
    {
      "epoch": 47.86,
      "learning_rate": 1.361812297734628e-05,
      "loss": 0.052,
      "step": 123250
    },
    {
      "epoch": 47.87,
      "learning_rate": 1.3617605177993528e-05,
      "loss": 0.0928,
      "step": 123260
    },
    {
      "epoch": 47.87,
      "learning_rate": 1.3617087378640778e-05,
      "loss": 0.0081,
      "step": 123270
    },
    {
      "epoch": 47.88,
      "learning_rate": 1.3616569579288026e-05,
      "loss": 0.0496,
      "step": 123280
    },
    {
      "epoch": 47.88,
      "learning_rate": 1.3616051779935276e-05,
      "loss": 0.0538,
      "step": 123290
    },
    {
      "epoch": 47.88,
      "learning_rate": 1.3615533980582525e-05,
      "loss": 0.1032,
      "step": 123300
    },
    {
      "epoch": 47.89,
      "learning_rate": 1.3615016181229775e-05,
      "loss": 0.0959,
      "step": 123310
    },
    {
      "epoch": 47.89,
      "learning_rate": 1.3614498381877024e-05,
      "loss": 0.1043,
      "step": 123320
    },
    {
      "epoch": 47.9,
      "learning_rate": 1.3613980582524272e-05,
      "loss": 0.0884,
      "step": 123330
    },
    {
      "epoch": 47.9,
      "learning_rate": 1.3613462783171522e-05,
      "loss": 0.2509,
      "step": 123340
    },
    {
      "epoch": 47.9,
      "learning_rate": 1.3612944983818772e-05,
      "loss": 0.0407,
      "step": 123350
    },
    {
      "epoch": 47.91,
      "learning_rate": 1.3612427184466021e-05,
      "loss": 0.072,
      "step": 123360
    },
    {
      "epoch": 47.91,
      "learning_rate": 1.361190938511327e-05,
      "loss": 0.0425,
      "step": 123370
    },
    {
      "epoch": 47.91,
      "learning_rate": 1.3611391585760519e-05,
      "loss": 0.0003,
      "step": 123380
    },
    {
      "epoch": 47.92,
      "learning_rate": 1.3610873786407768e-05,
      "loss": 0.0014,
      "step": 123390
    },
    {
      "epoch": 47.92,
      "learning_rate": 1.3610355987055018e-05,
      "loss": 0.0919,
      "step": 123400
    },
    {
      "epoch": 47.93,
      "learning_rate": 1.3609838187702266e-05,
      "loss": 0.1265,
      "step": 123410
    },
    {
      "epoch": 47.93,
      "learning_rate": 1.3609320388349516e-05,
      "loss": 0.0009,
      "step": 123420
    },
    {
      "epoch": 47.93,
      "learning_rate": 1.3608802588996765e-05,
      "loss": 0.0782,
      "step": 123430
    },
    {
      "epoch": 47.94,
      "learning_rate": 1.3608284789644015e-05,
      "loss": 0.0963,
      "step": 123440
    },
    {
      "epoch": 47.94,
      "learning_rate": 1.3607766990291263e-05,
      "loss": 0.0444,
      "step": 123450
    },
    {
      "epoch": 47.95,
      "learning_rate": 1.3607249190938512e-05,
      "loss": 0.1148,
      "step": 123460
    },
    {
      "epoch": 47.95,
      "learning_rate": 1.3606731391585762e-05,
      "loss": 0.0076,
      "step": 123470
    },
    {
      "epoch": 47.95,
      "learning_rate": 1.3606213592233012e-05,
      "loss": 0.2529,
      "step": 123480
    },
    {
      "epoch": 47.96,
      "learning_rate": 1.360569579288026e-05,
      "loss": 0.0187,
      "step": 123490
    },
    {
      "epoch": 47.96,
      "learning_rate": 1.3605177993527509e-05,
      "loss": 0.1857,
      "step": 123500
    },
    {
      "epoch": 47.97,
      "learning_rate": 1.3604660194174759e-05,
      "loss": 0.0532,
      "step": 123510
    },
    {
      "epoch": 47.97,
      "learning_rate": 1.3604142394822008e-05,
      "loss": 0.01,
      "step": 123520
    },
    {
      "epoch": 47.97,
      "learning_rate": 1.3603624595469256e-05,
      "loss": 0.1938,
      "step": 123530
    },
    {
      "epoch": 47.98,
      "learning_rate": 1.3603106796116506e-05,
      "loss": 0.0742,
      "step": 123540
    },
    {
      "epoch": 47.98,
      "learning_rate": 1.3602588996763756e-05,
      "loss": 0.0604,
      "step": 123550
    },
    {
      "epoch": 47.98,
      "learning_rate": 1.3602071197411003e-05,
      "loss": 0.0784,
      "step": 123560
    },
    {
      "epoch": 47.99,
      "learning_rate": 1.3601553398058253e-05,
      "loss": 0.0182,
      "step": 123570
    },
    {
      "epoch": 47.99,
      "learning_rate": 1.3601035598705503e-05,
      "loss": 0.1701,
      "step": 123580
    },
    {
      "epoch": 48.0,
      "learning_rate": 1.3600517799352752e-05,
      "loss": 0.0932,
      "step": 123590
    },
    {
      "epoch": 48.0,
      "learning_rate": 1.3600000000000002e-05,
      "loss": 0.0221,
      "step": 123600
    },
    {
      "epoch": 48.0,
      "eval_accuracy": 0.9480055020632737,
      "eval_loss": 0.32050269842147827,
      "eval_runtime": 8.2485,
      "eval_samples_per_second": 440.685,
      "eval_steps_per_second": 55.161,
      "step": 123600
    },
    {
      "epoch": 48.0,
      "learning_rate": 1.359948220064725e-05,
      "loss": 0.0992,
      "step": 123610
    },
    {
      "epoch": 48.01,
      "learning_rate": 1.35989644012945e-05,
      "loss": 0.0346,
      "step": 123620
    },
    {
      "epoch": 48.01,
      "learning_rate": 1.3598446601941749e-05,
      "loss": 0.066,
      "step": 123630
    },
    {
      "epoch": 48.02,
      "learning_rate": 1.3597928802588997e-05,
      "loss": 0.1347,
      "step": 123640
    },
    {
      "epoch": 48.02,
      "learning_rate": 1.3597411003236247e-05,
      "loss": 0.029,
      "step": 123650
    },
    {
      "epoch": 48.02,
      "learning_rate": 1.3596893203883496e-05,
      "loss": 0.0241,
      "step": 123660
    },
    {
      "epoch": 48.03,
      "learning_rate": 1.3596375404530746e-05,
      "loss": 0.0115,
      "step": 123670
    },
    {
      "epoch": 48.03,
      "learning_rate": 1.3595857605177995e-05,
      "loss": 0.1223,
      "step": 123680
    },
    {
      "epoch": 48.03,
      "learning_rate": 1.3595339805825243e-05,
      "loss": 0.1311,
      "step": 123690
    },
    {
      "epoch": 48.04,
      "learning_rate": 1.3594822006472493e-05,
      "loss": 0.0632,
      "step": 123700
    },
    {
      "epoch": 48.04,
      "learning_rate": 1.3594304207119741e-05,
      "loss": 0.0287,
      "step": 123710
    },
    {
      "epoch": 48.05,
      "learning_rate": 1.359378640776699e-05,
      "loss": 0.0568,
      "step": 123720
    },
    {
      "epoch": 48.05,
      "learning_rate": 1.359326860841424e-05,
      "loss": 0.0805,
      "step": 123730
    },
    {
      "epoch": 48.05,
      "learning_rate": 1.359275080906149e-05,
      "loss": 0.0885,
      "step": 123740
    },
    {
      "epoch": 48.06,
      "learning_rate": 1.359223300970874e-05,
      "loss": 0.0385,
      "step": 123750
    },
    {
      "epoch": 48.06,
      "learning_rate": 1.3591715210355989e-05,
      "loss": 0.0407,
      "step": 123760
    },
    {
      "epoch": 48.07,
      "learning_rate": 1.3591197411003237e-05,
      "loss": 0.0514,
      "step": 123770
    },
    {
      "epoch": 48.07,
      "learning_rate": 1.3590679611650487e-05,
      "loss": 0.1457,
      "step": 123780
    },
    {
      "epoch": 48.07,
      "learning_rate": 1.3590161812297735e-05,
      "loss": 0.0113,
      "step": 123790
    },
    {
      "epoch": 48.08,
      "learning_rate": 1.3589644012944984e-05,
      "loss": 0.0014,
      "step": 123800
    },
    {
      "epoch": 48.08,
      "learning_rate": 1.3589126213592234e-05,
      "loss": 0.0857,
      "step": 123810
    },
    {
      "epoch": 48.09,
      "learning_rate": 1.3588608414239483e-05,
      "loss": 0.1111,
      "step": 123820
    },
    {
      "epoch": 48.09,
      "learning_rate": 1.3588090614886733e-05,
      "loss": 0.0866,
      "step": 123830
    },
    {
      "epoch": 48.09,
      "learning_rate": 1.3587572815533983e-05,
      "loss": 0.0227,
      "step": 123840
    },
    {
      "epoch": 48.1,
      "learning_rate": 1.358705501618123e-05,
      "loss": 0.1133,
      "step": 123850
    },
    {
      "epoch": 48.1,
      "learning_rate": 1.3586537216828478e-05,
      "loss": 0.0927,
      "step": 123860
    },
    {
      "epoch": 48.1,
      "learning_rate": 1.3586019417475728e-05,
      "loss": 0.0294,
      "step": 123870
    },
    {
      "epoch": 48.11,
      "learning_rate": 1.3585501618122978e-05,
      "loss": 0.141,
      "step": 123880
    },
    {
      "epoch": 48.11,
      "learning_rate": 1.3584983818770227e-05,
      "loss": 0.0935,
      "step": 123890
    },
    {
      "epoch": 48.12,
      "learning_rate": 1.3584466019417477e-05,
      "loss": 0.0002,
      "step": 123900
    },
    {
      "epoch": 48.12,
      "learning_rate": 1.3583948220064727e-05,
      "loss": 0.0005,
      "step": 123910
    },
    {
      "epoch": 48.12,
      "learning_rate": 1.3583430420711976e-05,
      "loss": 0.0589,
      "step": 123920
    },
    {
      "epoch": 48.13,
      "learning_rate": 1.3582912621359226e-05,
      "loss": 0.0015,
      "step": 123930
    },
    {
      "epoch": 48.13,
      "learning_rate": 1.3582394822006472e-05,
      "loss": 0.0272,
      "step": 123940
    },
    {
      "epoch": 48.14,
      "learning_rate": 1.3581877022653722e-05,
      "loss": 0.3285,
      "step": 123950
    },
    {
      "epoch": 48.14,
      "learning_rate": 1.3581359223300971e-05,
      "loss": 0.0486,
      "step": 123960
    },
    {
      "epoch": 48.14,
      "learning_rate": 1.3580841423948221e-05,
      "loss": 0.0523,
      "step": 123970
    },
    {
      "epoch": 48.15,
      "learning_rate": 1.358032362459547e-05,
      "loss": 0.0628,
      "step": 123980
    },
    {
      "epoch": 48.15,
      "learning_rate": 1.357980582524272e-05,
      "loss": 0.0848,
      "step": 123990
    },
    {
      "epoch": 48.16,
      "learning_rate": 1.357928802588997e-05,
      "loss": 0.0794,
      "step": 124000
    },
    {
      "epoch": 48.16,
      "learning_rate": 1.357877022653722e-05,
      "loss": 0.0214,
      "step": 124010
    },
    {
      "epoch": 48.16,
      "learning_rate": 1.3578252427184466e-05,
      "loss": 0.0272,
      "step": 124020
    },
    {
      "epoch": 48.17,
      "learning_rate": 1.3577734627831715e-05,
      "loss": 0.0188,
      "step": 124030
    },
    {
      "epoch": 48.17,
      "learning_rate": 1.3577216828478965e-05,
      "loss": 0.108,
      "step": 124040
    },
    {
      "epoch": 48.17,
      "learning_rate": 1.3576699029126214e-05,
      "loss": 0.1005,
      "step": 124050
    },
    {
      "epoch": 48.18,
      "learning_rate": 1.3576181229773464e-05,
      "loss": 0.0648,
      "step": 124060
    },
    {
      "epoch": 48.18,
      "learning_rate": 1.3575663430420714e-05,
      "loss": 0.0014,
      "step": 124070
    },
    {
      "epoch": 48.19,
      "learning_rate": 1.3575145631067963e-05,
      "loss": 0.0652,
      "step": 124080
    },
    {
      "epoch": 48.19,
      "learning_rate": 1.3574627831715213e-05,
      "loss": 0.0451,
      "step": 124090
    },
    {
      "epoch": 48.19,
      "learning_rate": 1.3574110032362459e-05,
      "loss": 0.0496,
      "step": 124100
    },
    {
      "epoch": 48.2,
      "learning_rate": 1.3573592233009709e-05,
      "loss": 0.2249,
      "step": 124110
    },
    {
      "epoch": 48.2,
      "learning_rate": 1.3573074433656958e-05,
      "loss": 0.0457,
      "step": 124120
    },
    {
      "epoch": 48.21,
      "learning_rate": 1.3572556634304208e-05,
      "loss": 0.0657,
      "step": 124130
    },
    {
      "epoch": 48.21,
      "learning_rate": 1.3572038834951458e-05,
      "loss": 0.0196,
      "step": 124140
    },
    {
      "epoch": 48.21,
      "learning_rate": 1.3571521035598707e-05,
      "loss": 0.0658,
      "step": 124150
    },
    {
      "epoch": 48.22,
      "learning_rate": 1.3571003236245957e-05,
      "loss": 0.0144,
      "step": 124160
    },
    {
      "epoch": 48.22,
      "learning_rate": 1.3570485436893206e-05,
      "loss": 0.0354,
      "step": 124170
    },
    {
      "epoch": 48.23,
      "learning_rate": 1.3569967637540453e-05,
      "loss": 0.0068,
      "step": 124180
    },
    {
      "epoch": 48.23,
      "learning_rate": 1.3569449838187702e-05,
      "loss": 0.0308,
      "step": 124190
    },
    {
      "epoch": 48.23,
      "learning_rate": 1.3568932038834952e-05,
      "loss": 0.1704,
      "step": 124200
    },
    {
      "epoch": 48.24,
      "learning_rate": 1.3568414239482202e-05,
      "loss": 0.2049,
      "step": 124210
    },
    {
      "epoch": 48.24,
      "learning_rate": 1.3567896440129451e-05,
      "loss": 0.2563,
      "step": 124220
    },
    {
      "epoch": 48.24,
      "learning_rate": 1.35673786407767e-05,
      "loss": 0.1956,
      "step": 124230
    },
    {
      "epoch": 48.25,
      "learning_rate": 1.356686084142395e-05,
      "loss": 0.1156,
      "step": 124240
    },
    {
      "epoch": 48.25,
      "learning_rate": 1.35663430420712e-05,
      "loss": 0.022,
      "step": 124250
    },
    {
      "epoch": 48.26,
      "learning_rate": 1.3565825242718446e-05,
      "loss": 0.1304,
      "step": 124260
    },
    {
      "epoch": 48.26,
      "learning_rate": 1.3565307443365696e-05,
      "loss": 0.1049,
      "step": 124270
    },
    {
      "epoch": 48.26,
      "learning_rate": 1.3564789644012945e-05,
      "loss": 0.0293,
      "step": 124280
    },
    {
      "epoch": 48.27,
      "learning_rate": 1.3564271844660195e-05,
      "loss": 0.0011,
      "step": 124290
    },
    {
      "epoch": 48.27,
      "learning_rate": 1.3563754045307445e-05,
      "loss": 0.0677,
      "step": 124300
    },
    {
      "epoch": 48.28,
      "learning_rate": 1.3563236245954694e-05,
      "loss": 0.0375,
      "step": 124310
    },
    {
      "epoch": 48.28,
      "learning_rate": 1.3562718446601944e-05,
      "loss": 0.0082,
      "step": 124320
    },
    {
      "epoch": 48.28,
      "learning_rate": 1.3562200647249194e-05,
      "loss": 0.1539,
      "step": 124330
    },
    {
      "epoch": 48.29,
      "learning_rate": 1.356168284789644e-05,
      "loss": 0.1383,
      "step": 124340
    },
    {
      "epoch": 48.29,
      "learning_rate": 1.356116504854369e-05,
      "loss": 0.0689,
      "step": 124350
    },
    {
      "epoch": 48.3,
      "learning_rate": 1.3560647249190939e-05,
      "loss": 0.0014,
      "step": 124360
    },
    {
      "epoch": 48.3,
      "learning_rate": 1.3560129449838189e-05,
      "loss": 0.1445,
      "step": 124370
    },
    {
      "epoch": 48.3,
      "learning_rate": 1.3559611650485438e-05,
      "loss": 0.0403,
      "step": 124380
    },
    {
      "epoch": 48.31,
      "learning_rate": 1.3559093851132688e-05,
      "loss": 0.164,
      "step": 124390
    },
    {
      "epoch": 48.31,
      "learning_rate": 1.3558576051779937e-05,
      "loss": 0.0682,
      "step": 124400
    },
    {
      "epoch": 48.31,
      "learning_rate": 1.3558058252427187e-05,
      "loss": 0.046,
      "step": 124410
    },
    {
      "epoch": 48.32,
      "learning_rate": 1.3557540453074433e-05,
      "loss": 0.0213,
      "step": 124420
    },
    {
      "epoch": 48.32,
      "learning_rate": 1.3557022653721683e-05,
      "loss": 0.1289,
      "step": 124430
    },
    {
      "epoch": 48.33,
      "learning_rate": 1.3556504854368933e-05,
      "loss": 0.0868,
      "step": 124440
    },
    {
      "epoch": 48.33,
      "learning_rate": 1.3555987055016182e-05,
      "loss": 0.1261,
      "step": 124450
    },
    {
      "epoch": 48.33,
      "learning_rate": 1.3555469255663432e-05,
      "loss": 0.0079,
      "step": 124460
    },
    {
      "epoch": 48.34,
      "learning_rate": 1.3554951456310681e-05,
      "loss": 0.0656,
      "step": 124470
    },
    {
      "epoch": 48.34,
      "learning_rate": 1.3554433656957931e-05,
      "loss": 0.0827,
      "step": 124480
    },
    {
      "epoch": 48.35,
      "learning_rate": 1.355391585760518e-05,
      "loss": 0.0681,
      "step": 124490
    },
    {
      "epoch": 48.35,
      "learning_rate": 1.3553398058252429e-05,
      "loss": 0.1584,
      "step": 124500
    },
    {
      "epoch": 48.35,
      "learning_rate": 1.3552880258899677e-05,
      "loss": 0.0024,
      "step": 124510
    },
    {
      "epoch": 48.36,
      "learning_rate": 1.3552362459546926e-05,
      "loss": 0.1401,
      "step": 124520
    },
    {
      "epoch": 48.36,
      "learning_rate": 1.3551844660194176e-05,
      "loss": 0.0968,
      "step": 124530
    },
    {
      "epoch": 48.37,
      "learning_rate": 1.3551326860841425e-05,
      "loss": 0.0003,
      "step": 124540
    },
    {
      "epoch": 48.37,
      "learning_rate": 1.3550809061488675e-05,
      "loss": 0.0298,
      "step": 124550
    },
    {
      "epoch": 48.37,
      "learning_rate": 1.3550291262135925e-05,
      "loss": 0.101,
      "step": 124560
    },
    {
      "epoch": 48.38,
      "learning_rate": 1.3549773462783174e-05,
      "loss": 0.1293,
      "step": 124570
    },
    {
      "epoch": 48.38,
      "learning_rate": 1.3549255663430422e-05,
      "loss": 0.0831,
      "step": 124580
    },
    {
      "epoch": 48.38,
      "learning_rate": 1.354873786407767e-05,
      "loss": 0.0306,
      "step": 124590
    },
    {
      "epoch": 48.39,
      "learning_rate": 1.354822006472492e-05,
      "loss": 0.0406,
      "step": 124600
    },
    {
      "epoch": 48.39,
      "learning_rate": 1.354770226537217e-05,
      "loss": 0.0456,
      "step": 124610
    },
    {
      "epoch": 48.4,
      "learning_rate": 1.3547184466019419e-05,
      "loss": 0.0659,
      "step": 124620
    },
    {
      "epoch": 48.4,
      "learning_rate": 1.3546666666666669e-05,
      "loss": 0.0404,
      "step": 124630
    },
    {
      "epoch": 48.4,
      "learning_rate": 1.3546148867313918e-05,
      "loss": 0.0725,
      "step": 124640
    },
    {
      "epoch": 48.41,
      "learning_rate": 1.3545631067961166e-05,
      "loss": 0.0416,
      "step": 124650
    },
    {
      "epoch": 48.41,
      "learning_rate": 1.3545113268608416e-05,
      "loss": 0.0704,
      "step": 124660
    },
    {
      "epoch": 48.42,
      "learning_rate": 1.3544595469255664e-05,
      "loss": 0.1281,
      "step": 124670
    },
    {
      "epoch": 48.42,
      "learning_rate": 1.3544077669902913e-05,
      "loss": 0.0504,
      "step": 124680
    },
    {
      "epoch": 48.42,
      "learning_rate": 1.3543559870550163e-05,
      "loss": 0.0833,
      "step": 124690
    },
    {
      "epoch": 48.43,
      "learning_rate": 1.3543042071197412e-05,
      "loss": 0.2,
      "step": 124700
    },
    {
      "epoch": 48.43,
      "learning_rate": 1.3542524271844662e-05,
      "loss": 0.1419,
      "step": 124710
    },
    {
      "epoch": 48.43,
      "learning_rate": 1.3542006472491912e-05,
      "loss": 0.0258,
      "step": 124720
    },
    {
      "epoch": 48.44,
      "learning_rate": 1.354148867313916e-05,
      "loss": 0.1503,
      "step": 124730
    },
    {
      "epoch": 48.44,
      "learning_rate": 1.354097087378641e-05,
      "loss": 0.0365,
      "step": 124740
    },
    {
      "epoch": 48.45,
      "learning_rate": 1.3540453074433657e-05,
      "loss": 0.0804,
      "step": 124750
    },
    {
      "epoch": 48.45,
      "learning_rate": 1.3539935275080907e-05,
      "loss": 0.0059,
      "step": 124760
    },
    {
      "epoch": 48.45,
      "learning_rate": 1.3539417475728156e-05,
      "loss": 0.0061,
      "step": 124770
    },
    {
      "epoch": 48.46,
      "learning_rate": 1.3538899676375406e-05,
      "loss": 0.0099,
      "step": 124780
    },
    {
      "epoch": 48.46,
      "learning_rate": 1.3538381877022656e-05,
      "loss": 0.0256,
      "step": 124790
    },
    {
      "epoch": 48.47,
      "learning_rate": 1.3537864077669904e-05,
      "loss": 0.1119,
      "step": 124800
    },
    {
      "epoch": 48.47,
      "learning_rate": 1.3537346278317153e-05,
      "loss": 0.0005,
      "step": 124810
    },
    {
      "epoch": 48.47,
      "learning_rate": 1.3536828478964403e-05,
      "loss": 0.0924,
      "step": 124820
    },
    {
      "epoch": 48.48,
      "learning_rate": 1.353631067961165e-05,
      "loss": 0.1597,
      "step": 124830
    },
    {
      "epoch": 48.48,
      "learning_rate": 1.35357928802589e-05,
      "loss": 0.1043,
      "step": 124840
    },
    {
      "epoch": 48.49,
      "learning_rate": 1.353527508090615e-05,
      "loss": 0.0333,
      "step": 124850
    },
    {
      "epoch": 48.49,
      "learning_rate": 1.35347572815534e-05,
      "loss": 0.2038,
      "step": 124860
    },
    {
      "epoch": 48.49,
      "learning_rate": 1.353423948220065e-05,
      "loss": 0.0507,
      "step": 124870
    },
    {
      "epoch": 48.5,
      "learning_rate": 1.3533721682847897e-05,
      "loss": 0.0928,
      "step": 124880
    },
    {
      "epoch": 48.5,
      "learning_rate": 1.3533203883495147e-05,
      "loss": 0.0206,
      "step": 124890
    },
    {
      "epoch": 48.5,
      "learning_rate": 1.3532686084142396e-05,
      "loss": 0.1349,
      "step": 124900
    },
    {
      "epoch": 48.51,
      "learning_rate": 1.3532168284789644e-05,
      "loss": 0.1557,
      "step": 124910
    },
    {
      "epoch": 48.51,
      "learning_rate": 1.3531650485436894e-05,
      "loss": 0.1238,
      "step": 124920
    },
    {
      "epoch": 48.52,
      "learning_rate": 1.3531132686084144e-05,
      "loss": 0.1107,
      "step": 124930
    },
    {
      "epoch": 48.52,
      "learning_rate": 1.3530614886731393e-05,
      "loss": 0.0173,
      "step": 124940
    },
    {
      "epoch": 48.52,
      "learning_rate": 1.3530097087378641e-05,
      "loss": 0.1518,
      "step": 124950
    },
    {
      "epoch": 48.53,
      "learning_rate": 1.352957928802589e-05,
      "loss": 0.0234,
      "step": 124960
    },
    {
      "epoch": 48.53,
      "learning_rate": 1.352906148867314e-05,
      "loss": 0.0899,
      "step": 124970
    },
    {
      "epoch": 48.54,
      "learning_rate": 1.352854368932039e-05,
      "loss": 0.0868,
      "step": 124980
    },
    {
      "epoch": 48.54,
      "learning_rate": 1.3528025889967638e-05,
      "loss": 0.111,
      "step": 124990
    },
    {
      "epoch": 48.54,
      "learning_rate": 1.3527508090614887e-05,
      "loss": 0.0305,
      "step": 125000
    },
    {
      "epoch": 48.55,
      "learning_rate": 1.3526990291262137e-05,
      "loss": 0.193,
      "step": 125010
    },
    {
      "epoch": 48.55,
      "learning_rate": 1.3526472491909387e-05,
      "loss": 0.0008,
      "step": 125020
    },
    {
      "epoch": 48.56,
      "learning_rate": 1.3525954692556635e-05,
      "loss": 0.0152,
      "step": 125030
    },
    {
      "epoch": 48.56,
      "learning_rate": 1.3525436893203884e-05,
      "loss": 0.0615,
      "step": 125040
    },
    {
      "epoch": 48.56,
      "learning_rate": 1.3524919093851134e-05,
      "loss": 0.2004,
      "step": 125050
    },
    {
      "epoch": 48.57,
      "learning_rate": 1.3524401294498383e-05,
      "loss": 0.1104,
      "step": 125060
    },
    {
      "epoch": 48.57,
      "learning_rate": 1.3523883495145633e-05,
      "loss": 0.1484,
      "step": 125070
    },
    {
      "epoch": 48.57,
      "learning_rate": 1.3523365695792881e-05,
      "loss": 0.0449,
      "step": 125080
    },
    {
      "epoch": 48.58,
      "learning_rate": 1.352284789644013e-05,
      "loss": 0.1092,
      "step": 125090
    },
    {
      "epoch": 48.58,
      "learning_rate": 1.352233009708738e-05,
      "loss": 0.0015,
      "step": 125100
    },
    {
      "epoch": 48.59,
      "learning_rate": 1.3521812297734628e-05,
      "loss": 0.0404,
      "step": 125110
    },
    {
      "epoch": 48.59,
      "learning_rate": 1.3521294498381878e-05,
      "loss": 0.0572,
      "step": 125120
    },
    {
      "epoch": 48.59,
      "learning_rate": 1.3520776699029127e-05,
      "loss": 0.1467,
      "step": 125130
    },
    {
      "epoch": 48.6,
      "learning_rate": 1.3520258899676377e-05,
      "loss": 0.1845,
      "step": 125140
    },
    {
      "epoch": 48.6,
      "learning_rate": 1.3519741100323627e-05,
      "loss": 0.0589,
      "step": 125150
    },
    {
      "epoch": 48.61,
      "learning_rate": 1.3519223300970875e-05,
      "loss": 0.0901,
      "step": 125160
    },
    {
      "epoch": 48.61,
      "learning_rate": 1.3518705501618124e-05,
      "loss": 0.0126,
      "step": 125170
    },
    {
      "epoch": 48.61,
      "learning_rate": 1.3518187702265372e-05,
      "loss": 0.1349,
      "step": 125180
    },
    {
      "epoch": 48.62,
      "learning_rate": 1.3517669902912622e-05,
      "loss": 0.0577,
      "step": 125190
    },
    {
      "epoch": 48.62,
      "learning_rate": 1.3517152103559871e-05,
      "loss": 0.0616,
      "step": 125200
    },
    {
      "epoch": 48.63,
      "learning_rate": 1.3516634304207121e-05,
      "loss": 0.018,
      "step": 125210
    },
    {
      "epoch": 48.63,
      "learning_rate": 1.351611650485437e-05,
      "loss": 0.0131,
      "step": 125220
    },
    {
      "epoch": 48.63,
      "learning_rate": 1.351559870550162e-05,
      "loss": 0.0069,
      "step": 125230
    },
    {
      "epoch": 48.64,
      "learning_rate": 1.3515080906148868e-05,
      "loss": 0.075,
      "step": 125240
    },
    {
      "epoch": 48.64,
      "learning_rate": 1.3514563106796118e-05,
      "loss": 0.1003,
      "step": 125250
    },
    {
      "epoch": 48.64,
      "learning_rate": 1.3514045307443366e-05,
      "loss": 0.0671,
      "step": 125260
    },
    {
      "epoch": 48.65,
      "learning_rate": 1.3513527508090615e-05,
      "loss": 0.0932,
      "step": 125270
    },
    {
      "epoch": 48.65,
      "learning_rate": 1.3513009708737865e-05,
      "loss": 0.0715,
      "step": 125280
    },
    {
      "epoch": 48.66,
      "learning_rate": 1.3512491909385115e-05,
      "loss": 0.1556,
      "step": 125290
    },
    {
      "epoch": 48.66,
      "learning_rate": 1.3511974110032364e-05,
      "loss": 0.056,
      "step": 125300
    },
    {
      "epoch": 48.66,
      "learning_rate": 1.3511456310679614e-05,
      "loss": 0.0124,
      "step": 125310
    },
    {
      "epoch": 48.67,
      "learning_rate": 1.3510938511326862e-05,
      "loss": 0.1419,
      "step": 125320
    },
    {
      "epoch": 48.67,
      "learning_rate": 1.351042071197411e-05,
      "loss": 0.014,
      "step": 125330
    },
    {
      "epoch": 48.68,
      "learning_rate": 1.350990291262136e-05,
      "loss": 0.129,
      "step": 125340
    },
    {
      "epoch": 48.68,
      "learning_rate": 1.3509385113268609e-05,
      "loss": 0.075,
      "step": 125350
    },
    {
      "epoch": 48.68,
      "learning_rate": 1.3508867313915858e-05,
      "loss": 0.04,
      "step": 125360
    },
    {
      "epoch": 48.69,
      "learning_rate": 1.3508349514563108e-05,
      "loss": 0.1133,
      "step": 125370
    },
    {
      "epoch": 48.69,
      "learning_rate": 1.3507831715210358e-05,
      "loss": 0.1857,
      "step": 125380
    },
    {
      "epoch": 48.7,
      "learning_rate": 1.3507313915857607e-05,
      "loss": 0.0863,
      "step": 125390
    },
    {
      "epoch": 48.7,
      "learning_rate": 1.3506796116504855e-05,
      "loss": 0.1328,
      "step": 125400
    },
    {
      "epoch": 48.7,
      "learning_rate": 1.3506278317152103e-05,
      "loss": 0.0005,
      "step": 125410
    },
    {
      "epoch": 48.71,
      "learning_rate": 1.3505760517799353e-05,
      "loss": 0.07,
      "step": 125420
    },
    {
      "epoch": 48.71,
      "learning_rate": 1.3505242718446602e-05,
      "loss": 0.1331,
      "step": 125430
    },
    {
      "epoch": 48.71,
      "learning_rate": 1.3504724919093852e-05,
      "loss": 0.0368,
      "step": 125440
    },
    {
      "epoch": 48.72,
      "learning_rate": 1.3504207119741102e-05,
      "loss": 0.0566,
      "step": 125450
    },
    {
      "epoch": 48.72,
      "learning_rate": 1.3503689320388351e-05,
      "loss": 0.0136,
      "step": 125460
    },
    {
      "epoch": 48.73,
      "learning_rate": 1.3503171521035601e-05,
      "loss": 0.0355,
      "step": 125470
    },
    {
      "epoch": 48.73,
      "learning_rate": 1.3502653721682847e-05,
      "loss": 0.0808,
      "step": 125480
    },
    {
      "epoch": 48.73,
      "learning_rate": 1.3502135922330097e-05,
      "loss": 0.0634,
      "step": 125490
    },
    {
      "epoch": 48.74,
      "learning_rate": 1.3501618122977346e-05,
      "loss": 0.1023,
      "step": 125500
    },
    {
      "epoch": 48.74,
      "learning_rate": 1.3501100323624596e-05,
      "loss": 0.0131,
      "step": 125510
    },
    {
      "epoch": 48.75,
      "learning_rate": 1.3500582524271846e-05,
      "loss": 0.0347,
      "step": 125520
    },
    {
      "epoch": 48.75,
      "learning_rate": 1.3500064724919095e-05,
      "loss": 0.0378,
      "step": 125530
    },
    {
      "epoch": 48.75,
      "learning_rate": 1.3499546925566345e-05,
      "loss": 0.1495,
      "step": 125540
    },
    {
      "epoch": 48.76,
      "learning_rate": 1.3499029126213594e-05,
      "loss": 0.0062,
      "step": 125550
    },
    {
      "epoch": 48.76,
      "learning_rate": 1.349851132686084e-05,
      "loss": 0.1176,
      "step": 125560
    },
    {
      "epoch": 48.77,
      "learning_rate": 1.349799352750809e-05,
      "loss": 0.082,
      "step": 125570
    },
    {
      "epoch": 48.77,
      "learning_rate": 1.349747572815534e-05,
      "loss": 0.1871,
      "step": 125580
    },
    {
      "epoch": 48.77,
      "learning_rate": 1.349695792880259e-05,
      "loss": 0.1503,
      "step": 125590
    },
    {
      "epoch": 48.78,
      "learning_rate": 1.3496440129449839e-05,
      "loss": 0.0503,
      "step": 125600
    },
    {
      "epoch": 48.78,
      "learning_rate": 1.3495922330097089e-05,
      "loss": 0.002,
      "step": 125610
    },
    {
      "epoch": 48.78,
      "learning_rate": 1.3495404530744338e-05,
      "loss": 0.1018,
      "step": 125620
    },
    {
      "epoch": 48.79,
      "learning_rate": 1.3494886731391588e-05,
      "loss": 0.0875,
      "step": 125630
    },
    {
      "epoch": 48.79,
      "learning_rate": 1.3494368932038838e-05,
      "loss": 0.1589,
      "step": 125640
    },
    {
      "epoch": 48.8,
      "learning_rate": 1.3493851132686084e-05,
      "loss": 0.0942,
      "step": 125650
    },
    {
      "epoch": 48.8,
      "learning_rate": 1.3493333333333333e-05,
      "loss": 0.2853,
      "step": 125660
    },
    {
      "epoch": 48.8,
      "learning_rate": 1.3492815533980583e-05,
      "loss": 0.0427,
      "step": 125670
    },
    {
      "epoch": 48.81,
      "learning_rate": 1.3492297734627833e-05,
      "loss": 0.1611,
      "step": 125680
    },
    {
      "epoch": 48.81,
      "learning_rate": 1.3491779935275082e-05,
      "loss": 0.064,
      "step": 125690
    },
    {
      "epoch": 48.82,
      "learning_rate": 1.3491262135922332e-05,
      "loss": 0.1089,
      "step": 125700
    },
    {
      "epoch": 48.82,
      "learning_rate": 1.3490744336569582e-05,
      "loss": 0.0263,
      "step": 125710
    },
    {
      "epoch": 48.82,
      "learning_rate": 1.3490226537216831e-05,
      "loss": 0.0322,
      "step": 125720
    },
    {
      "epoch": 48.83,
      "learning_rate": 1.3489708737864077e-05,
      "loss": 0.0624,
      "step": 125730
    },
    {
      "epoch": 48.83,
      "learning_rate": 1.3489190938511327e-05,
      "loss": 0.1727,
      "step": 125740
    },
    {
      "epoch": 48.83,
      "learning_rate": 1.3488673139158577e-05,
      "loss": 0.0412,
      "step": 125750
    },
    {
      "epoch": 48.84,
      "learning_rate": 1.3488155339805826e-05,
      "loss": 0.1111,
      "step": 125760
    },
    {
      "epoch": 48.84,
      "learning_rate": 1.3487637540453076e-05,
      "loss": 0.0448,
      "step": 125770
    },
    {
      "epoch": 48.85,
      "learning_rate": 1.3487119741100325e-05,
      "loss": 0.0952,
      "step": 125780
    },
    {
      "epoch": 48.85,
      "learning_rate": 1.3486601941747575e-05,
      "loss": 0.065,
      "step": 125790
    },
    {
      "epoch": 48.85,
      "learning_rate": 1.3486084142394825e-05,
      "loss": 0.0445,
      "step": 125800
    },
    {
      "epoch": 48.86,
      "learning_rate": 1.3485566343042071e-05,
      "loss": 0.0563,
      "step": 125810
    },
    {
      "epoch": 48.86,
      "learning_rate": 1.348504854368932e-05,
      "loss": 0.0329,
      "step": 125820
    },
    {
      "epoch": 48.87,
      "learning_rate": 1.348453074433657e-05,
      "loss": 0.0834,
      "step": 125830
    },
    {
      "epoch": 48.87,
      "learning_rate": 1.348401294498382e-05,
      "loss": 0.0074,
      "step": 125840
    },
    {
      "epoch": 48.87,
      "learning_rate": 1.348349514563107e-05,
      "loss": 0.0249,
      "step": 125850
    },
    {
      "epoch": 48.88,
      "learning_rate": 1.3482977346278319e-05,
      "loss": 0.1716,
      "step": 125860
    },
    {
      "epoch": 48.88,
      "learning_rate": 1.3482459546925569e-05,
      "loss": 0.0047,
      "step": 125870
    },
    {
      "epoch": 48.89,
      "learning_rate": 1.3481941747572818e-05,
      "loss": 0.1417,
      "step": 125880
    },
    {
      "epoch": 48.89,
      "learning_rate": 1.3481423948220065e-05,
      "loss": 0.0885,
      "step": 125890
    },
    {
      "epoch": 48.89,
      "learning_rate": 1.3480906148867314e-05,
      "loss": 0.0489,
      "step": 125900
    },
    {
      "epoch": 48.9,
      "learning_rate": 1.3480388349514564e-05,
      "loss": 0.0063,
      "step": 125910
    },
    {
      "epoch": 48.9,
      "learning_rate": 1.3479870550161813e-05,
      "loss": 0.023,
      "step": 125920
    },
    {
      "epoch": 48.9,
      "learning_rate": 1.3479352750809063e-05,
      "loss": 0.0416,
      "step": 125930
    },
    {
      "epoch": 48.91,
      "learning_rate": 1.3478834951456313e-05,
      "loss": 0.2327,
      "step": 125940
    },
    {
      "epoch": 48.91,
      "learning_rate": 1.3478317152103562e-05,
      "loss": 0.1167,
      "step": 125950
    },
    {
      "epoch": 48.92,
      "learning_rate": 1.3477799352750812e-05,
      "loss": 0.035,
      "step": 125960
    },
    {
      "epoch": 48.92,
      "learning_rate": 1.3477281553398058e-05,
      "loss": 0.0536,
      "step": 125970
    },
    {
      "epoch": 48.92,
      "learning_rate": 1.3476763754045308e-05,
      "loss": 0.1128,
      "step": 125980
    },
    {
      "epoch": 48.93,
      "learning_rate": 1.3476245954692557e-05,
      "loss": 0.2563,
      "step": 125990
    },
    {
      "epoch": 48.93,
      "learning_rate": 1.3475728155339807e-05,
      "loss": 0.1264,
      "step": 126000
    },
    {
      "epoch": 48.94,
      "learning_rate": 1.3475210355987057e-05,
      "loss": 0.1387,
      "step": 126010
    },
    {
      "epoch": 48.94,
      "learning_rate": 1.3474692556634306e-05,
      "loss": 0.1023,
      "step": 126020
    },
    {
      "epoch": 48.94,
      "learning_rate": 1.3474174757281556e-05,
      "loss": 0.0387,
      "step": 126030
    },
    {
      "epoch": 48.95,
      "learning_rate": 1.3473656957928805e-05,
      "loss": 0.0752,
      "step": 126040
    },
    {
      "epoch": 48.95,
      "learning_rate": 1.3473139158576052e-05,
      "loss": 0.1248,
      "step": 126050
    },
    {
      "epoch": 48.96,
      "learning_rate": 1.3472621359223301e-05,
      "loss": 0.0576,
      "step": 126060
    },
    {
      "epoch": 48.96,
      "learning_rate": 1.3472103559870551e-05,
      "loss": 0.0932,
      "step": 126070
    },
    {
      "epoch": 48.96,
      "learning_rate": 1.34715857605178e-05,
      "loss": 0.0213,
      "step": 126080
    },
    {
      "epoch": 48.97,
      "learning_rate": 1.347106796116505e-05,
      "loss": 0.068,
      "step": 126090
    },
    {
      "epoch": 48.97,
      "learning_rate": 1.34705501618123e-05,
      "loss": 0.082,
      "step": 126100
    },
    {
      "epoch": 48.97,
      "learning_rate": 1.347003236245955e-05,
      "loss": 0.2132,
      "step": 126110
    },
    {
      "epoch": 48.98,
      "learning_rate": 1.3469514563106797e-05,
      "loss": 0.0704,
      "step": 126120
    },
    {
      "epoch": 48.98,
      "learning_rate": 1.3468996763754045e-05,
      "loss": 0.0772,
      "step": 126130
    },
    {
      "epoch": 48.99,
      "learning_rate": 1.3468478964401295e-05,
      "loss": 0.0738,
      "step": 126140
    },
    {
      "epoch": 48.99,
      "learning_rate": 1.3467961165048544e-05,
      "loss": 0.1023,
      "step": 126150
    },
    {
      "epoch": 48.99,
      "learning_rate": 1.3467443365695794e-05,
      "loss": 0.123,
      "step": 126160
    },
    {
      "epoch": 49.0,
      "learning_rate": 1.3466925566343044e-05,
      "loss": 0.0939,
      "step": 126170
    },
    {
      "epoch": 49.0,
      "eval_accuracy": 0.9485557083906465,
      "eval_loss": 0.3431084454059601,
      "eval_runtime": 8.2153,
      "eval_samples_per_second": 442.467,
      "eval_steps_per_second": 55.384,
      "step": 126175
    },
    {
      "epoch": 49.0,
      "learning_rate": 1.3466407766990293e-05,
      "loss": 0.1448,
      "step": 126180
    },
    {
      "epoch": 49.01,
      "learning_rate": 1.3465889967637543e-05,
      "loss": 0.201,
      "step": 126190
    },
    {
      "epoch": 49.01,
      "learning_rate": 1.346537216828479e-05,
      "loss": 0.0787,
      "step": 126200
    },
    {
      "epoch": 49.01,
      "learning_rate": 1.346485436893204e-05,
      "loss": 0.0446,
      "step": 126210
    },
    {
      "epoch": 49.02,
      "learning_rate": 1.3464336569579288e-05,
      "loss": 0.0431,
      "step": 126220
    },
    {
      "epoch": 49.02,
      "learning_rate": 1.3463818770226538e-05,
      "loss": 0.108,
      "step": 126230
    },
    {
      "epoch": 49.03,
      "learning_rate": 1.3463300970873788e-05,
      "loss": 0.0901,
      "step": 126240
    },
    {
      "epoch": 49.03,
      "learning_rate": 1.3462783171521037e-05,
      "loss": 0.0603,
      "step": 126250
    },
    {
      "epoch": 49.03,
      "learning_rate": 1.3462265372168287e-05,
      "loss": 0.0156,
      "step": 126260
    },
    {
      "epoch": 49.04,
      "learning_rate": 1.3461747572815535e-05,
      "loss": 0.0329,
      "step": 126270
    },
    {
      "epoch": 49.04,
      "learning_rate": 1.3461229773462784e-05,
      "loss": 0.0831,
      "step": 126280
    },
    {
      "epoch": 49.04,
      "learning_rate": 1.3460711974110034e-05,
      "loss": 0.0132,
      "step": 126290
    },
    {
      "epoch": 49.05,
      "learning_rate": 1.3460194174757282e-05,
      "loss": 0.0908,
      "step": 126300
    },
    {
      "epoch": 49.05,
      "learning_rate": 1.3459676375404532e-05,
      "loss": 0.0813,
      "step": 126310
    },
    {
      "epoch": 49.06,
      "learning_rate": 1.3459158576051781e-05,
      "loss": 0.088,
      "step": 126320
    },
    {
      "epoch": 49.06,
      "learning_rate": 1.345864077669903e-05,
      "loss": 0.186,
      "step": 126330
    },
    {
      "epoch": 49.06,
      "learning_rate": 1.345812297734628e-05,
      "loss": 0.0888,
      "step": 126340
    },
    {
      "epoch": 49.07,
      "learning_rate": 1.3457605177993528e-05,
      "loss": 0.0658,
      "step": 126350
    },
    {
      "epoch": 49.07,
      "learning_rate": 1.3457087378640778e-05,
      "loss": 0.0547,
      "step": 126360
    },
    {
      "epoch": 49.08,
      "learning_rate": 1.3456569579288028e-05,
      "loss": 0.07,
      "step": 126370
    },
    {
      "epoch": 49.08,
      "learning_rate": 1.3456051779935275e-05,
      "loss": 0.1041,
      "step": 126380
    },
    {
      "epoch": 49.08,
      "learning_rate": 1.3455533980582525e-05,
      "loss": 0.1467,
      "step": 126390
    },
    {
      "epoch": 49.09,
      "learning_rate": 1.3455016181229775e-05,
      "loss": 0.0859,
      "step": 126400
    },
    {
      "epoch": 49.09,
      "learning_rate": 1.3454498381877024e-05,
      "loss": 0.1492,
      "step": 126410
    },
    {
      "epoch": 49.1,
      "learning_rate": 1.3453980582524272e-05,
      "loss": 0.0112,
      "step": 126420
    },
    {
      "epoch": 49.1,
      "learning_rate": 1.3453462783171522e-05,
      "loss": 0.009,
      "step": 126430
    },
    {
      "epoch": 49.1,
      "learning_rate": 1.3452944983818771e-05,
      "loss": 0.0247,
      "step": 126440
    },
    {
      "epoch": 49.11,
      "learning_rate": 1.3452427184466021e-05,
      "loss": 0.0012,
      "step": 126450
    },
    {
      "epoch": 49.11,
      "learning_rate": 1.3451909385113269e-05,
      "loss": 0.0196,
      "step": 126460
    },
    {
      "epoch": 49.11,
      "learning_rate": 1.3451391585760519e-05,
      "loss": 0.0093,
      "step": 126470
    },
    {
      "epoch": 49.12,
      "learning_rate": 1.3450873786407768e-05,
      "loss": 0.0665,
      "step": 126480
    },
    {
      "epoch": 49.12,
      "learning_rate": 1.3450355987055018e-05,
      "loss": 0.0313,
      "step": 126490
    },
    {
      "epoch": 49.13,
      "learning_rate": 1.3449838187702266e-05,
      "loss": 0.0215,
      "step": 126500
    },
    {
      "epoch": 49.13,
      "learning_rate": 1.3449320388349515e-05,
      "loss": 0.1794,
      "step": 126510
    },
    {
      "epoch": 49.13,
      "learning_rate": 1.3448802588996765e-05,
      "loss": 0.1555,
      "step": 126520
    },
    {
      "epoch": 49.14,
      "learning_rate": 1.3448284789644015e-05,
      "loss": 0.0261,
      "step": 126530
    },
    {
      "epoch": 49.14,
      "learning_rate": 1.3447766990291263e-05,
      "loss": 0.1589,
      "step": 126540
    },
    {
      "epoch": 49.15,
      "learning_rate": 1.3447249190938512e-05,
      "loss": 0.0074,
      "step": 126550
    },
    {
      "epoch": 49.15,
      "learning_rate": 1.3446731391585762e-05,
      "loss": 0.0512,
      "step": 126560
    },
    {
      "epoch": 49.15,
      "learning_rate": 1.3446213592233011e-05,
      "loss": 0.0801,
      "step": 126570
    },
    {
      "epoch": 49.16,
      "learning_rate": 1.344569579288026e-05,
      "loss": 0.0247,
      "step": 126580
    },
    {
      "epoch": 49.16,
      "learning_rate": 1.3445177993527509e-05,
      "loss": 0.1598,
      "step": 126590
    },
    {
      "epoch": 49.17,
      "learning_rate": 1.3444660194174759e-05,
      "loss": 0.0707,
      "step": 126600
    },
    {
      "epoch": 49.17,
      "learning_rate": 1.3444142394822008e-05,
      "loss": 0.1358,
      "step": 126610
    },
    {
      "epoch": 49.17,
      "learning_rate": 1.3443624595469256e-05,
      "loss": 0.1661,
      "step": 126620
    },
    {
      "epoch": 49.18,
      "learning_rate": 1.3443106796116506e-05,
      "loss": 0.0372,
      "step": 126630
    },
    {
      "epoch": 49.18,
      "learning_rate": 1.3442588996763755e-05,
      "loss": 0.1075,
      "step": 126640
    },
    {
      "epoch": 49.18,
      "learning_rate": 1.3442071197411003e-05,
      "loss": 0.0213,
      "step": 126650
    },
    {
      "epoch": 49.19,
      "learning_rate": 1.3441553398058253e-05,
      "loss": 0.0209,
      "step": 126660
    },
    {
      "epoch": 49.19,
      "learning_rate": 1.3441035598705503e-05,
      "loss": 0.0879,
      "step": 126670
    },
    {
      "epoch": 49.2,
      "learning_rate": 1.3440517799352752e-05,
      "loss": 0.0731,
      "step": 126680
    },
    {
      "epoch": 49.2,
      "learning_rate": 1.3440000000000002e-05,
      "loss": 0.0429,
      "step": 126690
    },
    {
      "epoch": 49.2,
      "learning_rate": 1.3439482200647251e-05,
      "loss": 0.2395,
      "step": 126700
    },
    {
      "epoch": 49.21,
      "learning_rate": 1.34389644012945e-05,
      "loss": 0.1148,
      "step": 126710
    },
    {
      "epoch": 49.21,
      "learning_rate": 1.3438446601941749e-05,
      "loss": 0.225,
      "step": 126720
    },
    {
      "epoch": 49.22,
      "learning_rate": 1.3437928802588997e-05,
      "loss": 0.1291,
      "step": 126730
    },
    {
      "epoch": 49.22,
      "learning_rate": 1.3437411003236246e-05,
      "loss": 0.0702,
      "step": 126740
    },
    {
      "epoch": 49.22,
      "learning_rate": 1.3436893203883496e-05,
      "loss": 0.102,
      "step": 126750
    },
    {
      "epoch": 49.23,
      "learning_rate": 1.3436375404530746e-05,
      "loss": 0.0703,
      "step": 126760
    },
    {
      "epoch": 49.23,
      "learning_rate": 1.3435857605177995e-05,
      "loss": 0.0655,
      "step": 126770
    },
    {
      "epoch": 49.23,
      "learning_rate": 1.3435339805825245e-05,
      "loss": 0.0044,
      "step": 126780
    },
    {
      "epoch": 49.24,
      "learning_rate": 1.3434822006472493e-05,
      "loss": 0.1531,
      "step": 126790
    },
    {
      "epoch": 49.24,
      "learning_rate": 1.343430420711974e-05,
      "loss": 0.0756,
      "step": 126800
    },
    {
      "epoch": 49.25,
      "learning_rate": 1.343378640776699e-05,
      "loss": 0.0577,
      "step": 126810
    },
    {
      "epoch": 49.25,
      "learning_rate": 1.343326860841424e-05,
      "loss": 0.0054,
      "step": 126820
    },
    {
      "epoch": 49.25,
      "learning_rate": 1.343275080906149e-05,
      "loss": 0.0418,
      "step": 126830
    },
    {
      "epoch": 49.26,
      "learning_rate": 1.343223300970874e-05,
      "loss": 0.0889,
      "step": 126840
    },
    {
      "epoch": 49.26,
      "learning_rate": 1.3431715210355989e-05,
      "loss": 0.1663,
      "step": 126850
    },
    {
      "epoch": 49.27,
      "learning_rate": 1.3431197411003238e-05,
      "loss": 0.0598,
      "step": 126860
    },
    {
      "epoch": 49.27,
      "learning_rate": 1.3430679611650486e-05,
      "loss": 0.0072,
      "step": 126870
    },
    {
      "epoch": 49.27,
      "learning_rate": 1.3430161812297734e-05,
      "loss": 0.0862,
      "step": 126880
    },
    {
      "epoch": 49.28,
      "learning_rate": 1.3429644012944984e-05,
      "loss": 0.0345,
      "step": 126890
    },
    {
      "epoch": 49.28,
      "learning_rate": 1.3429126213592234e-05,
      "loss": 0.0637,
      "step": 126900
    },
    {
      "epoch": 49.29,
      "learning_rate": 1.3428608414239483e-05,
      "loss": 0.071,
      "step": 126910
    },
    {
      "epoch": 49.29,
      "learning_rate": 1.3428090614886733e-05,
      "loss": 0.0163,
      "step": 126920
    },
    {
      "epoch": 49.29,
      "learning_rate": 1.3427572815533982e-05,
      "loss": 0.0114,
      "step": 126930
    },
    {
      "epoch": 49.3,
      "learning_rate": 1.3427055016181232e-05,
      "loss": 0.1619,
      "step": 126940
    },
    {
      "epoch": 49.3,
      "learning_rate": 1.3426537216828478e-05,
      "loss": 0.0723,
      "step": 126950
    },
    {
      "epoch": 49.3,
      "learning_rate": 1.3426019417475728e-05,
      "loss": 0.0248,
      "step": 126960
    },
    {
      "epoch": 49.31,
      "learning_rate": 1.3425501618122978e-05,
      "loss": 0.0522,
      "step": 126970
    },
    {
      "epoch": 49.31,
      "learning_rate": 1.3424983818770227e-05,
      "loss": 0.088,
      "step": 126980
    },
    {
      "epoch": 49.32,
      "learning_rate": 1.3424466019417477e-05,
      "loss": 0.1007,
      "step": 126990
    },
    {
      "epoch": 49.32,
      "learning_rate": 1.3423948220064726e-05,
      "loss": 0.1052,
      "step": 127000
    },
    {
      "epoch": 49.32,
      "learning_rate": 1.3423430420711976e-05,
      "loss": 0.0565,
      "step": 127010
    },
    {
      "epoch": 49.33,
      "learning_rate": 1.3422912621359226e-05,
      "loss": 0.1357,
      "step": 127020
    },
    {
      "epoch": 49.33,
      "learning_rate": 1.3422394822006472e-05,
      "loss": 0.02,
      "step": 127030
    },
    {
      "epoch": 49.34,
      "learning_rate": 1.3421877022653721e-05,
      "loss": 0.0195,
      "step": 127040
    },
    {
      "epoch": 49.34,
      "learning_rate": 1.3421359223300971e-05,
      "loss": 0.1843,
      "step": 127050
    },
    {
      "epoch": 49.34,
      "learning_rate": 1.342084142394822e-05,
      "loss": 0.0674,
      "step": 127060
    },
    {
      "epoch": 49.35,
      "learning_rate": 1.342032362459547e-05,
      "loss": 0.0353,
      "step": 127070
    },
    {
      "epoch": 49.35,
      "learning_rate": 1.341980582524272e-05,
      "loss": 0.0477,
      "step": 127080
    },
    {
      "epoch": 49.36,
      "learning_rate": 1.341928802588997e-05,
      "loss": 0.0152,
      "step": 127090
    },
    {
      "epoch": 49.36,
      "learning_rate": 1.341877022653722e-05,
      "loss": 0.038,
      "step": 127100
    },
    {
      "epoch": 49.36,
      "learning_rate": 1.3418252427184465e-05,
      "loss": 0.0757,
      "step": 127110
    },
    {
      "epoch": 49.37,
      "learning_rate": 1.3417734627831715e-05,
      "loss": 0.0316,
      "step": 127120
    },
    {
      "epoch": 49.37,
      "learning_rate": 1.3417216828478965e-05,
      "loss": 0.1288,
      "step": 127130
    },
    {
      "epoch": 49.37,
      "learning_rate": 1.3416699029126214e-05,
      "loss": 0.1889,
      "step": 127140
    },
    {
      "epoch": 49.38,
      "learning_rate": 1.3416181229773464e-05,
      "loss": 0.0597,
      "step": 127150
    },
    {
      "epoch": 49.38,
      "learning_rate": 1.3415663430420713e-05,
      "loss": 0.048,
      "step": 127160
    },
    {
      "epoch": 49.39,
      "learning_rate": 1.3415145631067963e-05,
      "loss": 0.1566,
      "step": 127170
    },
    {
      "epoch": 49.39,
      "learning_rate": 1.3414627831715213e-05,
      "loss": 0.0915,
      "step": 127180
    },
    {
      "epoch": 49.39,
      "learning_rate": 1.3414110032362459e-05,
      "loss": 0.0265,
      "step": 127190
    },
    {
      "epoch": 49.4,
      "learning_rate": 1.3413592233009709e-05,
      "loss": 0.0242,
      "step": 127200
    },
    {
      "epoch": 49.4,
      "learning_rate": 1.3413074433656958e-05,
      "loss": 0.0593,
      "step": 127210
    },
    {
      "epoch": 49.41,
      "learning_rate": 1.3412556634304208e-05,
      "loss": 0.0016,
      "step": 127220
    },
    {
      "epoch": 49.41,
      "learning_rate": 1.3412038834951457e-05,
      "loss": 0.0369,
      "step": 127230
    },
    {
      "epoch": 49.41,
      "learning_rate": 1.3411521035598707e-05,
      "loss": 0.0234,
      "step": 127240
    },
    {
      "epoch": 49.42,
      "learning_rate": 1.3411003236245957e-05,
      "loss": 0.0302,
      "step": 127250
    },
    {
      "epoch": 49.42,
      "learning_rate": 1.3410485436893206e-05,
      "loss": 0.0227,
      "step": 127260
    },
    {
      "epoch": 49.43,
      "learning_rate": 1.3409967637540456e-05,
      "loss": 0.0006,
      "step": 127270
    },
    {
      "epoch": 49.43,
      "learning_rate": 1.3409449838187702e-05,
      "loss": 0.021,
      "step": 127280
    },
    {
      "epoch": 49.43,
      "learning_rate": 1.3408932038834952e-05,
      "loss": 0.0284,
      "step": 127290
    },
    {
      "epoch": 49.44,
      "learning_rate": 1.3408414239482201e-05,
      "loss": 0.0008,
      "step": 127300
    },
    {
      "epoch": 49.44,
      "learning_rate": 1.3407896440129451e-05,
      "loss": 0.1253,
      "step": 127310
    },
    {
      "epoch": 49.44,
      "learning_rate": 1.34073786407767e-05,
      "loss": 0.0201,
      "step": 127320
    },
    {
      "epoch": 49.45,
      "learning_rate": 1.340686084142395e-05,
      "loss": 0.0737,
      "step": 127330
    },
    {
      "epoch": 49.45,
      "learning_rate": 1.34063430420712e-05,
      "loss": 0.1984,
      "step": 127340
    },
    {
      "epoch": 49.46,
      "learning_rate": 1.340582524271845e-05,
      "loss": 0.0782,
      "step": 127350
    },
    {
      "epoch": 49.46,
      "learning_rate": 1.3405307443365696e-05,
      "loss": 0.0442,
      "step": 127360
    },
    {
      "epoch": 49.46,
      "learning_rate": 1.3404789644012945e-05,
      "loss": 0.0182,
      "step": 127370
    },
    {
      "epoch": 49.47,
      "learning_rate": 1.3404271844660195e-05,
      "loss": 0.1399,
      "step": 127380
    },
    {
      "epoch": 49.47,
      "learning_rate": 1.3403754045307445e-05,
      "loss": 0.0723,
      "step": 127390
    },
    {
      "epoch": 49.48,
      "learning_rate": 1.3403236245954694e-05,
      "loss": 0.006,
      "step": 127400
    },
    {
      "epoch": 49.48,
      "learning_rate": 1.3402718446601944e-05,
      "loss": 0.054,
      "step": 127410
    },
    {
      "epoch": 49.48,
      "learning_rate": 1.3402200647249193e-05,
      "loss": 0.032,
      "step": 127420
    },
    {
      "epoch": 49.49,
      "learning_rate": 1.3401682847896443e-05,
      "loss": 0.0074,
      "step": 127430
    },
    {
      "epoch": 49.49,
      "learning_rate": 1.340116504854369e-05,
      "loss": 0.0062,
      "step": 127440
    },
    {
      "epoch": 49.5,
      "learning_rate": 1.3400647249190939e-05,
      "loss": 0.0556,
      "step": 127450
    },
    {
      "epoch": 49.5,
      "learning_rate": 1.3400129449838188e-05,
      "loss": 0.0005,
      "step": 127460
    },
    {
      "epoch": 49.5,
      "learning_rate": 1.3399611650485438e-05,
      "loss": 0.0628,
      "step": 127470
    },
    {
      "epoch": 49.51,
      "learning_rate": 1.3399093851132688e-05,
      "loss": 0.0017,
      "step": 127480
    },
    {
      "epoch": 49.51,
      "learning_rate": 1.3398576051779937e-05,
      "loss": 0.0237,
      "step": 127490
    },
    {
      "epoch": 49.51,
      "learning_rate": 1.3398058252427187e-05,
      "loss": 0.0159,
      "step": 127500
    },
    {
      "epoch": 49.52,
      "learning_rate": 1.3397540453074437e-05,
      "loss": 0.0963,
      "step": 127510
    },
    {
      "epoch": 49.52,
      "learning_rate": 1.3397022653721683e-05,
      "loss": 0.0155,
      "step": 127520
    },
    {
      "epoch": 49.53,
      "learning_rate": 1.3396504854368932e-05,
      "loss": 0.0259,
      "step": 127530
    },
    {
      "epoch": 49.53,
      "learning_rate": 1.3395987055016182e-05,
      "loss": 0.0669,
      "step": 127540
    },
    {
      "epoch": 49.53,
      "learning_rate": 1.3395469255663432e-05,
      "loss": 0.0182,
      "step": 127550
    },
    {
      "epoch": 49.54,
      "learning_rate": 1.3394951456310681e-05,
      "loss": 0.1163,
      "step": 127560
    },
    {
      "epoch": 49.54,
      "learning_rate": 1.3394433656957931e-05,
      "loss": 0.0983,
      "step": 127570
    },
    {
      "epoch": 49.55,
      "learning_rate": 1.339391585760518e-05,
      "loss": 0.0299,
      "step": 127580
    },
    {
      "epoch": 49.55,
      "learning_rate": 1.3393398058252428e-05,
      "loss": 0.0841,
      "step": 127590
    },
    {
      "epoch": 49.55,
      "learning_rate": 1.3392880258899676e-05,
      "loss": 0.0483,
      "step": 127600
    },
    {
      "epoch": 49.56,
      "learning_rate": 1.3392362459546926e-05,
      "loss": 0.0496,
      "step": 127610
    },
    {
      "epoch": 49.56,
      "learning_rate": 1.3391844660194176e-05,
      "loss": 0.3217,
      "step": 127620
    },
    {
      "epoch": 49.57,
      "learning_rate": 1.3391326860841425e-05,
      "loss": 0.0785,
      "step": 127630
    },
    {
      "epoch": 49.57,
      "learning_rate": 1.3390809061488675e-05,
      "loss": 0.0782,
      "step": 127640
    },
    {
      "epoch": 49.57,
      "learning_rate": 1.3390291262135924e-05,
      "loss": 0.007,
      "step": 127650
    },
    {
      "epoch": 49.58,
      "learning_rate": 1.3389773462783174e-05,
      "loss": 0.0479,
      "step": 127660
    },
    {
      "epoch": 49.58,
      "learning_rate": 1.3389255663430422e-05,
      "loss": 0.0379,
      "step": 127670
    },
    {
      "epoch": 49.58,
      "learning_rate": 1.338873786407767e-05,
      "loss": 0.0932,
      "step": 127680
    },
    {
      "epoch": 49.59,
      "learning_rate": 1.338822006472492e-05,
      "loss": 0.0505,
      "step": 127690
    },
    {
      "epoch": 49.59,
      "learning_rate": 1.3387702265372169e-05,
      "loss": 0.0869,
      "step": 127700
    },
    {
      "epoch": 49.6,
      "learning_rate": 1.3387184466019419e-05,
      "loss": 0.0273,
      "step": 127710
    },
    {
      "epoch": 49.6,
      "learning_rate": 1.3386666666666668e-05,
      "loss": 0.1132,
      "step": 127720
    },
    {
      "epoch": 49.6,
      "learning_rate": 1.3386148867313918e-05,
      "loss": 0.0221,
      "step": 127730
    },
    {
      "epoch": 49.61,
      "learning_rate": 1.3385631067961166e-05,
      "loss": 0.118,
      "step": 127740
    },
    {
      "epoch": 49.61,
      "learning_rate": 1.3385113268608416e-05,
      "loss": 0.3227,
      "step": 127750
    },
    {
      "epoch": 49.62,
      "learning_rate": 1.3384595469255663e-05,
      "loss": 0.0383,
      "step": 127760
    },
    {
      "epoch": 49.62,
      "learning_rate": 1.3384077669902913e-05,
      "loss": 0.178,
      "step": 127770
    },
    {
      "epoch": 49.62,
      "learning_rate": 1.3383559870550163e-05,
      "loss": 0.0758,
      "step": 127780
    },
    {
      "epoch": 49.63,
      "learning_rate": 1.3383042071197412e-05,
      "loss": 0.0314,
      "step": 127790
    },
    {
      "epoch": 49.63,
      "learning_rate": 1.3382524271844662e-05,
      "loss": 0.1465,
      "step": 127800
    },
    {
      "epoch": 49.63,
      "learning_rate": 1.3382006472491912e-05,
      "loss": 0.0182,
      "step": 127810
    },
    {
      "epoch": 49.64,
      "learning_rate": 1.338148867313916e-05,
      "loss": 0.0322,
      "step": 127820
    },
    {
      "epoch": 49.64,
      "learning_rate": 1.3380970873786409e-05,
      "loss": 0.3149,
      "step": 127830
    },
    {
      "epoch": 49.65,
      "learning_rate": 1.3380453074433659e-05,
      "loss": 0.0234,
      "step": 127840
    },
    {
      "epoch": 49.65,
      "learning_rate": 1.3379935275080907e-05,
      "loss": 0.2232,
      "step": 127850
    },
    {
      "epoch": 49.65,
      "learning_rate": 1.3379417475728156e-05,
      "loss": 0.0711,
      "step": 127860
    },
    {
      "epoch": 49.66,
      "learning_rate": 1.3378899676375406e-05,
      "loss": 0.1993,
      "step": 127870
    },
    {
      "epoch": 49.66,
      "learning_rate": 1.3378381877022655e-05,
      "loss": 0.0991,
      "step": 127880
    },
    {
      "epoch": 49.67,
      "learning_rate": 1.3377864077669903e-05,
      "loss": 0.029,
      "step": 127890
    },
    {
      "epoch": 49.67,
      "learning_rate": 1.3377346278317153e-05,
      "loss": 0.0872,
      "step": 127900
    },
    {
      "epoch": 49.67,
      "learning_rate": 1.3376828478964403e-05,
      "loss": 0.0229,
      "step": 127910
    },
    {
      "epoch": 49.68,
      "learning_rate": 1.3376310679611652e-05,
      "loss": 0.0385,
      "step": 127920
    },
    {
      "epoch": 49.68,
      "learning_rate": 1.33757928802589e-05,
      "loss": 0.1936,
      "step": 127930
    },
    {
      "epoch": 49.69,
      "learning_rate": 1.337527508090615e-05,
      "loss": 0.1104,
      "step": 127940
    },
    {
      "epoch": 49.69,
      "learning_rate": 1.33747572815534e-05,
      "loss": 0.0446,
      "step": 127950
    },
    {
      "epoch": 49.69,
      "learning_rate": 1.3374239482200649e-05,
      "loss": 0.1954,
      "step": 127960
    },
    {
      "epoch": 49.7,
      "learning_rate": 1.3373721682847897e-05,
      "loss": 0.1952,
      "step": 127970
    },
    {
      "epoch": 49.7,
      "learning_rate": 1.3373203883495147e-05,
      "loss": 0.1486,
      "step": 127980
    },
    {
      "epoch": 49.7,
      "learning_rate": 1.3372686084142396e-05,
      "loss": 0.0305,
      "step": 127990
    },
    {
      "epoch": 49.71,
      "learning_rate": 1.3372168284789646e-05,
      "loss": 0.1525,
      "step": 128000
    },
    {
      "epoch": 49.71,
      "learning_rate": 1.3371650485436894e-05,
      "loss": 0.1002,
      "step": 128010
    },
    {
      "epoch": 49.72,
      "learning_rate": 1.3371132686084143e-05,
      "loss": 0.0983,
      "step": 128020
    },
    {
      "epoch": 49.72,
      "learning_rate": 1.3370614886731393e-05,
      "loss": 0.1617,
      "step": 128030
    },
    {
      "epoch": 49.72,
      "learning_rate": 1.3370097087378643e-05,
      "loss": 0.2585,
      "step": 128040
    },
    {
      "epoch": 49.73,
      "learning_rate": 1.336957928802589e-05,
      "loss": 0.1167,
      "step": 128050
    },
    {
      "epoch": 49.73,
      "learning_rate": 1.336906148867314e-05,
      "loss": 0.0042,
      "step": 128060
    },
    {
      "epoch": 49.74,
      "learning_rate": 1.336854368932039e-05,
      "loss": 0.0106,
      "step": 128070
    },
    {
      "epoch": 49.74,
      "learning_rate": 1.336802588996764e-05,
      "loss": 0.0782,
      "step": 128080
    },
    {
      "epoch": 49.74,
      "learning_rate": 1.3367508090614887e-05,
      "loss": 0.0793,
      "step": 128090
    },
    {
      "epoch": 49.75,
      "learning_rate": 1.3366990291262137e-05,
      "loss": 0.1683,
      "step": 128100
    },
    {
      "epoch": 49.75,
      "learning_rate": 1.3366472491909387e-05,
      "loss": 0.1678,
      "step": 128110
    },
    {
      "epoch": 49.76,
      "learning_rate": 1.3365954692556634e-05,
      "loss": 0.0767,
      "step": 128120
    },
    {
      "epoch": 49.76,
      "learning_rate": 1.3365436893203884e-05,
      "loss": 0.0576,
      "step": 128130
    },
    {
      "epoch": 49.76,
      "learning_rate": 1.3364919093851134e-05,
      "loss": 0.0017,
      "step": 128140
    },
    {
      "epoch": 49.77,
      "learning_rate": 1.3364401294498383e-05,
      "loss": 0.177,
      "step": 128150
    },
    {
      "epoch": 49.77,
      "learning_rate": 1.3363883495145633e-05,
      "loss": 0.2572,
      "step": 128160
    },
    {
      "epoch": 49.77,
      "learning_rate": 1.3363365695792881e-05,
      "loss": 0.0596,
      "step": 128170
    },
    {
      "epoch": 49.78,
      "learning_rate": 1.336284789644013e-05,
      "loss": 0.0401,
      "step": 128180
    },
    {
      "epoch": 49.78,
      "learning_rate": 1.336233009708738e-05,
      "loss": 0.1306,
      "step": 128190
    },
    {
      "epoch": 49.79,
      "learning_rate": 1.3361812297734628e-05,
      "loss": 0.0642,
      "step": 128200
    },
    {
      "epoch": 49.79,
      "learning_rate": 1.3361294498381878e-05,
      "loss": 0.1035,
      "step": 128210
    },
    {
      "epoch": 49.79,
      "learning_rate": 1.3360776699029127e-05,
      "loss": 0.082,
      "step": 128220
    },
    {
      "epoch": 49.8,
      "learning_rate": 1.3360258899676377e-05,
      "loss": 0.0179,
      "step": 128230
    },
    {
      "epoch": 49.8,
      "learning_rate": 1.3359741100323626e-05,
      "loss": 0.004,
      "step": 128240
    },
    {
      "epoch": 49.81,
      "learning_rate": 1.3359223300970874e-05,
      "loss": 0.1409,
      "step": 128250
    },
    {
      "epoch": 49.81,
      "learning_rate": 1.3358705501618124e-05,
      "loss": 0.0705,
      "step": 128260
    },
    {
      "epoch": 49.81,
      "learning_rate": 1.3358187702265372e-05,
      "loss": 0.0459,
      "step": 128270
    },
    {
      "epoch": 49.82,
      "learning_rate": 1.3357669902912622e-05,
      "loss": 0.07,
      "step": 128280
    },
    {
      "epoch": 49.82,
      "learning_rate": 1.3357152103559871e-05,
      "loss": 0.1234,
      "step": 128290
    },
    {
      "epoch": 49.83,
      "learning_rate": 1.335663430420712e-05,
      "loss": 0.0956,
      "step": 128300
    },
    {
      "epoch": 49.83,
      "learning_rate": 1.335611650485437e-05,
      "loss": 0.0575,
      "step": 128310
    },
    {
      "epoch": 49.83,
      "learning_rate": 1.335559870550162e-05,
      "loss": 0.1417,
      "step": 128320
    },
    {
      "epoch": 49.84,
      "learning_rate": 1.3355080906148868e-05,
      "loss": 0.0942,
      "step": 128330
    },
    {
      "epoch": 49.84,
      "learning_rate": 1.3354563106796118e-05,
      "loss": 0.0604,
      "step": 128340
    },
    {
      "epoch": 49.84,
      "learning_rate": 1.3354045307443366e-05,
      "loss": 0.0588,
      "step": 128350
    },
    {
      "epoch": 49.85,
      "learning_rate": 1.3353527508090615e-05,
      "loss": 0.0025,
      "step": 128360
    },
    {
      "epoch": 49.85,
      "learning_rate": 1.3353009708737865e-05,
      "loss": 0.0556,
      "step": 128370
    },
    {
      "epoch": 49.86,
      "learning_rate": 1.3352491909385114e-05,
      "loss": 0.0023,
      "step": 128380
    },
    {
      "epoch": 49.86,
      "learning_rate": 1.3351974110032364e-05,
      "loss": 0.0805,
      "step": 128390
    },
    {
      "epoch": 49.86,
      "learning_rate": 1.3351456310679614e-05,
      "loss": 0.0012,
      "step": 128400
    },
    {
      "epoch": 49.87,
      "learning_rate": 1.3350938511326863e-05,
      "loss": 0.0022,
      "step": 128410
    },
    {
      "epoch": 49.87,
      "learning_rate": 1.335042071197411e-05,
      "loss": 0.0331,
      "step": 128420
    },
    {
      "epoch": 49.88,
      "learning_rate": 1.3349902912621359e-05,
      "loss": 0.1665,
      "step": 128430
    },
    {
      "epoch": 49.88,
      "learning_rate": 1.3349385113268609e-05,
      "loss": 0.0858,
      "step": 128440
    },
    {
      "epoch": 49.88,
      "learning_rate": 1.3348867313915858e-05,
      "loss": 0.1256,
      "step": 128450
    },
    {
      "epoch": 49.89,
      "learning_rate": 1.3348349514563108e-05,
      "loss": 0.2353,
      "step": 128460
    },
    {
      "epoch": 49.89,
      "learning_rate": 1.3347831715210358e-05,
      "loss": 0.0344,
      "step": 128470
    },
    {
      "epoch": 49.9,
      "learning_rate": 1.3347313915857607e-05,
      "loss": 0.0566,
      "step": 128480
    },
    {
      "epoch": 49.9,
      "learning_rate": 1.3346796116504857e-05,
      "loss": 0.1254,
      "step": 128490
    },
    {
      "epoch": 49.9,
      "learning_rate": 1.3346278317152103e-05,
      "loss": 0.0526,
      "step": 128500
    },
    {
      "epoch": 49.91,
      "learning_rate": 1.3345760517799353e-05,
      "loss": 0.1928,
      "step": 128510
    },
    {
      "epoch": 49.91,
      "learning_rate": 1.3345242718446602e-05,
      "loss": 0.0992,
      "step": 128520
    },
    {
      "epoch": 49.91,
      "learning_rate": 1.3344724919093852e-05,
      "loss": 0.1226,
      "step": 128530
    },
    {
      "epoch": 49.92,
      "learning_rate": 1.3344207119741101e-05,
      "loss": 0.1824,
      "step": 128540
    },
    {
      "epoch": 49.92,
      "learning_rate": 1.3343689320388351e-05,
      "loss": 0.078,
      "step": 128550
    },
    {
      "epoch": 49.93,
      "learning_rate": 1.33431715210356e-05,
      "loss": 0.305,
      "step": 128560
    },
    {
      "epoch": 49.93,
      "learning_rate": 1.334265372168285e-05,
      "loss": 0.0278,
      "step": 128570
    },
    {
      "epoch": 49.93,
      "learning_rate": 1.3342135922330097e-05,
      "loss": 0.1835,
      "step": 128580
    },
    {
      "epoch": 49.94,
      "learning_rate": 1.3341618122977346e-05,
      "loss": 0.0407,
      "step": 128590
    },
    {
      "epoch": 49.94,
      "learning_rate": 1.3341100323624596e-05,
      "loss": 0.0529,
      "step": 128600
    },
    {
      "epoch": 49.95,
      "learning_rate": 1.3340582524271845e-05,
      "loss": 0.0399,
      "step": 128610
    },
    {
      "epoch": 49.95,
      "learning_rate": 1.3340064724919095e-05,
      "loss": 0.1249,
      "step": 128620
    },
    {
      "epoch": 49.95,
      "learning_rate": 1.3339546925566345e-05,
      "loss": 0.0238,
      "step": 128630
    },
    {
      "epoch": 49.96,
      "learning_rate": 1.3339029126213594e-05,
      "loss": 0.168,
      "step": 128640
    },
    {
      "epoch": 49.96,
      "learning_rate": 1.3338511326860844e-05,
      "loss": 0.0496,
      "step": 128650
    },
    {
      "epoch": 49.97,
      "learning_rate": 1.333799352750809e-05,
      "loss": 0.0874,
      "step": 128660
    },
    {
      "epoch": 49.97,
      "learning_rate": 1.333747572815534e-05,
      "loss": 0.1442,
      "step": 128670
    },
    {
      "epoch": 49.97,
      "learning_rate": 1.333695792880259e-05,
      "loss": 0.0691,
      "step": 128680
    },
    {
      "epoch": 49.98,
      "learning_rate": 1.3336440129449839e-05,
      "loss": 0.081,
      "step": 128690
    },
    {
      "epoch": 49.98,
      "learning_rate": 1.3335922330097089e-05,
      "loss": 0.1192,
      "step": 128700
    },
    {
      "epoch": 49.98,
      "learning_rate": 1.3335404530744338e-05,
      "loss": 0.0461,
      "step": 128710
    },
    {
      "epoch": 49.99,
      "learning_rate": 1.3334886731391588e-05,
      "loss": 0.0862,
      "step": 128720
    },
    {
      "epoch": 49.99,
      "learning_rate": 1.3334368932038837e-05,
      "loss": 0.0132,
      "step": 128730
    },
    {
      "epoch": 50.0,
      "learning_rate": 1.3333851132686084e-05,
      "loss": 0.0503,
      "step": 128740
    },
    {
      "epoch": 50.0,
      "learning_rate": 1.3333333333333333e-05,
      "loss": 0.0599,
      "step": 128750
    },
    {
      "epoch": 50.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.3166683316230774,
      "eval_runtime": 8.2321,
      "eval_samples_per_second": 441.562,
      "eval_steps_per_second": 55.271,
      "step": 128750
    },
    {
      "epoch": 50.0,
      "learning_rate": 1.3332815533980583e-05,
      "loss": 0.0261,
      "step": 128760
    },
    {
      "epoch": 50.01,
      "learning_rate": 1.3332297734627833e-05,
      "loss": 0.0754,
      "step": 128770
    },
    {
      "epoch": 50.01,
      "learning_rate": 1.3331779935275082e-05,
      "loss": 0.084,
      "step": 128780
    },
    {
      "epoch": 50.02,
      "learning_rate": 1.3331262135922332e-05,
      "loss": 0.0728,
      "step": 128790
    },
    {
      "epoch": 50.02,
      "learning_rate": 1.3330744336569581e-05,
      "loss": 0.1192,
      "step": 128800
    },
    {
      "epoch": 50.02,
      "learning_rate": 1.3330226537216831e-05,
      "loss": 0.0007,
      "step": 128810
    },
    {
      "epoch": 50.03,
      "learning_rate": 1.3329708737864077e-05,
      "loss": 0.0995,
      "step": 128820
    },
    {
      "epoch": 50.03,
      "learning_rate": 1.3329190938511327e-05,
      "loss": 0.0798,
      "step": 128830
    },
    {
      "epoch": 50.03,
      "learning_rate": 1.3328673139158576e-05,
      "loss": 0.0798,
      "step": 128840
    },
    {
      "epoch": 50.04,
      "learning_rate": 1.3328155339805826e-05,
      "loss": 0.1158,
      "step": 128850
    },
    {
      "epoch": 50.04,
      "learning_rate": 1.3327637540453076e-05,
      "loss": 0.0274,
      "step": 128860
    },
    {
      "epoch": 50.05,
      "learning_rate": 1.3327119741100325e-05,
      "loss": 0.1302,
      "step": 128870
    },
    {
      "epoch": 50.05,
      "learning_rate": 1.3326601941747575e-05,
      "loss": 0.0411,
      "step": 128880
    },
    {
      "epoch": 50.05,
      "learning_rate": 1.3326084142394825e-05,
      "loss": 0.0733,
      "step": 128890
    },
    {
      "epoch": 50.06,
      "learning_rate": 1.332556634304207e-05,
      "loss": 0.1093,
      "step": 128900
    },
    {
      "epoch": 50.06,
      "learning_rate": 1.332504854368932e-05,
      "loss": 0.1115,
      "step": 128910
    },
    {
      "epoch": 50.07,
      "learning_rate": 1.332453074433657e-05,
      "loss": 0.0039,
      "step": 128920
    },
    {
      "epoch": 50.07,
      "learning_rate": 1.332401294498382e-05,
      "loss": 0.0897,
      "step": 128930
    },
    {
      "epoch": 50.07,
      "learning_rate": 1.332349514563107e-05,
      "loss": 0.1407,
      "step": 128940
    },
    {
      "epoch": 50.08,
      "learning_rate": 1.3322977346278319e-05,
      "loss": 0.1428,
      "step": 128950
    },
    {
      "epoch": 50.08,
      "learning_rate": 1.3322459546925568e-05,
      "loss": 0.1149,
      "step": 128960
    },
    {
      "epoch": 50.09,
      "learning_rate": 1.3321941747572818e-05,
      "loss": 0.0594,
      "step": 128970
    },
    {
      "epoch": 50.09,
      "learning_rate": 1.3321423948220068e-05,
      "loss": 0.1067,
      "step": 128980
    },
    {
      "epoch": 50.09,
      "learning_rate": 1.3320906148867314e-05,
      "loss": 0.0598,
      "step": 128990
    },
    {
      "epoch": 50.1,
      "learning_rate": 1.3320388349514564e-05,
      "loss": 0.292,
      "step": 129000
    },
    {
      "epoch": 50.1,
      "learning_rate": 1.3319870550161813e-05,
      "loss": 0.0886,
      "step": 129010
    },
    {
      "epoch": 50.1,
      "learning_rate": 1.3319352750809063e-05,
      "loss": 0.0814,
      "step": 129020
    },
    {
      "epoch": 50.11,
      "learning_rate": 1.3318834951456312e-05,
      "loss": 0.1103,
      "step": 129030
    },
    {
      "epoch": 50.11,
      "learning_rate": 1.3318317152103562e-05,
      "loss": 0.1528,
      "step": 129040
    },
    {
      "epoch": 50.12,
      "learning_rate": 1.3317799352750812e-05,
      "loss": 0.0282,
      "step": 129050
    },
    {
      "epoch": 50.12,
      "learning_rate": 1.331728155339806e-05,
      "loss": 0.0574,
      "step": 129060
    },
    {
      "epoch": 50.12,
      "learning_rate": 1.3316763754045308e-05,
      "loss": 0.0631,
      "step": 129070
    },
    {
      "epoch": 50.13,
      "learning_rate": 1.3316245954692557e-05,
      "loss": 0.1645,
      "step": 129080
    },
    {
      "epoch": 50.13,
      "learning_rate": 1.3315728155339807e-05,
      "loss": 0.1168,
      "step": 129090
    },
    {
      "epoch": 50.14,
      "learning_rate": 1.3315210355987056e-05,
      "loss": 0.0095,
      "step": 129100
    },
    {
      "epoch": 50.14,
      "learning_rate": 1.3314692556634306e-05,
      "loss": 0.0915,
      "step": 129110
    },
    {
      "epoch": 50.14,
      "learning_rate": 1.3314174757281556e-05,
      "loss": 0.0202,
      "step": 129120
    },
    {
      "epoch": 50.15,
      "learning_rate": 1.3313656957928805e-05,
      "loss": 0.0306,
      "step": 129130
    },
    {
      "epoch": 50.15,
      "learning_rate": 1.3313139158576053e-05,
      "loss": 0.0987,
      "step": 129140
    },
    {
      "epoch": 50.16,
      "learning_rate": 1.3312621359223301e-05,
      "loss": 0.1058,
      "step": 129150
    },
    {
      "epoch": 50.16,
      "learning_rate": 1.331210355987055e-05,
      "loss": 0.0585,
      "step": 129160
    },
    {
      "epoch": 50.16,
      "learning_rate": 1.33115857605178e-05,
      "loss": 0.0514,
      "step": 129170
    },
    {
      "epoch": 50.17,
      "learning_rate": 1.331106796116505e-05,
      "loss": 0.0555,
      "step": 129180
    },
    {
      "epoch": 50.17,
      "learning_rate": 1.33105501618123e-05,
      "loss": 0.1225,
      "step": 129190
    },
    {
      "epoch": 50.17,
      "learning_rate": 1.331003236245955e-05,
      "loss": 0.1178,
      "step": 129200
    },
    {
      "epoch": 50.18,
      "learning_rate": 1.3309514563106797e-05,
      "loss": 0.0803,
      "step": 129210
    },
    {
      "epoch": 50.18,
      "learning_rate": 1.3308996763754047e-05,
      "loss": 0.2603,
      "step": 129220
    },
    {
      "epoch": 50.19,
      "learning_rate": 1.3308478964401295e-05,
      "loss": 0.069,
      "step": 129230
    },
    {
      "epoch": 50.19,
      "learning_rate": 1.3307961165048544e-05,
      "loss": 0.1192,
      "step": 129240
    },
    {
      "epoch": 50.19,
      "learning_rate": 1.3307443365695794e-05,
      "loss": 0.0125,
      "step": 129250
    },
    {
      "epoch": 50.2,
      "learning_rate": 1.3306925566343043e-05,
      "loss": 0.0123,
      "step": 129260
    },
    {
      "epoch": 50.2,
      "learning_rate": 1.3306407766990293e-05,
      "loss": 0.0623,
      "step": 129270
    },
    {
      "epoch": 50.21,
      "learning_rate": 1.3305889967637543e-05,
      "loss": 0.0736,
      "step": 129280
    },
    {
      "epoch": 50.21,
      "learning_rate": 1.330537216828479e-05,
      "loss": 0.035,
      "step": 129290
    },
    {
      "epoch": 50.21,
      "learning_rate": 1.330485436893204e-05,
      "loss": 0.0956,
      "step": 129300
    },
    {
      "epoch": 50.22,
      "learning_rate": 1.3304336569579288e-05,
      "loss": 0.0582,
      "step": 129310
    },
    {
      "epoch": 50.22,
      "learning_rate": 1.3303818770226538e-05,
      "loss": 0.1461,
      "step": 129320
    },
    {
      "epoch": 50.23,
      "learning_rate": 1.3303300970873787e-05,
      "loss": 0.1733,
      "step": 129330
    },
    {
      "epoch": 50.23,
      "learning_rate": 1.3302783171521037e-05,
      "loss": 0.0071,
      "step": 129340
    },
    {
      "epoch": 50.23,
      "learning_rate": 1.3302265372168287e-05,
      "loss": 0.0469,
      "step": 129350
    },
    {
      "epoch": 50.24,
      "learning_rate": 1.3301747572815535e-05,
      "loss": 0.1987,
      "step": 129360
    },
    {
      "epoch": 50.24,
      "learning_rate": 1.3301229773462784e-05,
      "loss": 0.0406,
      "step": 129370
    },
    {
      "epoch": 50.24,
      "learning_rate": 1.3300711974110034e-05,
      "loss": 0.2309,
      "step": 129380
    },
    {
      "epoch": 50.25,
      "learning_rate": 1.3300194174757282e-05,
      "loss": 0.0156,
      "step": 129390
    },
    {
      "epoch": 50.25,
      "learning_rate": 1.3299676375404531e-05,
      "loss": 0.0949,
      "step": 129400
    },
    {
      "epoch": 50.26,
      "learning_rate": 1.3299158576051781e-05,
      "loss": 0.0887,
      "step": 129410
    },
    {
      "epoch": 50.26,
      "learning_rate": 1.329864077669903e-05,
      "loss": 0.1759,
      "step": 129420
    },
    {
      "epoch": 50.26,
      "learning_rate": 1.329812297734628e-05,
      "loss": 0.2203,
      "step": 129430
    },
    {
      "epoch": 50.27,
      "learning_rate": 1.3297605177993528e-05,
      "loss": 0.08,
      "step": 129440
    },
    {
      "epoch": 50.27,
      "learning_rate": 1.3297087378640778e-05,
      "loss": 0.0841,
      "step": 129450
    },
    {
      "epoch": 50.28,
      "learning_rate": 1.3296569579288027e-05,
      "loss": 0.0959,
      "step": 129460
    },
    {
      "epoch": 50.28,
      "learning_rate": 1.3296051779935275e-05,
      "loss": 0.0028,
      "step": 129470
    },
    {
      "epoch": 50.28,
      "learning_rate": 1.3295533980582525e-05,
      "loss": 0.1344,
      "step": 129480
    },
    {
      "epoch": 50.29,
      "learning_rate": 1.3295016181229775e-05,
      "loss": 0.0182,
      "step": 129490
    },
    {
      "epoch": 50.29,
      "learning_rate": 1.3294498381877024e-05,
      "loss": 0.2198,
      "step": 129500
    },
    {
      "epoch": 50.3,
      "learning_rate": 1.3293980582524274e-05,
      "loss": 0.2259,
      "step": 129510
    },
    {
      "epoch": 50.3,
      "learning_rate": 1.3293462783171522e-05,
      "loss": 0.033,
      "step": 129520
    },
    {
      "epoch": 50.3,
      "learning_rate": 1.3292944983818771e-05,
      "loss": 0.1704,
      "step": 129530
    },
    {
      "epoch": 50.31,
      "learning_rate": 1.3292427184466021e-05,
      "loss": 0.0936,
      "step": 129540
    },
    {
      "epoch": 50.31,
      "learning_rate": 1.329190938511327e-05,
      "loss": 0.0188,
      "step": 129550
    },
    {
      "epoch": 50.31,
      "learning_rate": 1.3291391585760518e-05,
      "loss": 0.0279,
      "step": 129560
    },
    {
      "epoch": 50.32,
      "learning_rate": 1.3290873786407768e-05,
      "loss": 0.0233,
      "step": 129570
    },
    {
      "epoch": 50.32,
      "learning_rate": 1.3290355987055018e-05,
      "loss": 0.192,
      "step": 129580
    },
    {
      "epoch": 50.33,
      "learning_rate": 1.3289838187702266e-05,
      "loss": 0.0575,
      "step": 129590
    },
    {
      "epoch": 50.33,
      "learning_rate": 1.3289320388349515e-05,
      "loss": 0.1414,
      "step": 129600
    },
    {
      "epoch": 50.33,
      "learning_rate": 1.3288802588996765e-05,
      "loss": 0.1369,
      "step": 129610
    },
    {
      "epoch": 50.34,
      "learning_rate": 1.3288284789644014e-05,
      "loss": 0.0468,
      "step": 129620
    },
    {
      "epoch": 50.34,
      "learning_rate": 1.3287766990291264e-05,
      "loss": 0.0872,
      "step": 129630
    },
    {
      "epoch": 50.35,
      "learning_rate": 1.3287249190938512e-05,
      "loss": 0.0378,
      "step": 129640
    },
    {
      "epoch": 50.35,
      "learning_rate": 1.3286731391585762e-05,
      "loss": 0.036,
      "step": 129650
    },
    {
      "epoch": 50.35,
      "learning_rate": 1.3286213592233011e-05,
      "loss": 0.0093,
      "step": 129660
    },
    {
      "epoch": 50.36,
      "learning_rate": 1.328569579288026e-05,
      "loss": 0.126,
      "step": 129670
    },
    {
      "epoch": 50.36,
      "learning_rate": 1.3285177993527509e-05,
      "loss": 0.0991,
      "step": 129680
    },
    {
      "epoch": 50.37,
      "learning_rate": 1.3284660194174758e-05,
      "loss": 0.0057,
      "step": 129690
    },
    {
      "epoch": 50.37,
      "learning_rate": 1.3284142394822008e-05,
      "loss": 0.1076,
      "step": 129700
    },
    {
      "epoch": 50.37,
      "learning_rate": 1.3283624595469258e-05,
      "loss": 0.0382,
      "step": 129710
    },
    {
      "epoch": 50.38,
      "learning_rate": 1.3283106796116506e-05,
      "loss": 0.0744,
      "step": 129720
    },
    {
      "epoch": 50.38,
      "learning_rate": 1.3282588996763755e-05,
      "loss": 0.0611,
      "step": 129730
    },
    {
      "epoch": 50.38,
      "learning_rate": 1.3282071197411003e-05,
      "loss": 0.0346,
      "step": 129740
    },
    {
      "epoch": 50.39,
      "learning_rate": 1.3281553398058253e-05,
      "loss": 0.1452,
      "step": 129750
    },
    {
      "epoch": 50.39,
      "learning_rate": 1.3281035598705502e-05,
      "loss": 0.0738,
      "step": 129760
    },
    {
      "epoch": 50.4,
      "learning_rate": 1.3280517799352752e-05,
      "loss": 0.1081,
      "step": 129770
    },
    {
      "epoch": 50.4,
      "learning_rate": 1.3280000000000002e-05,
      "loss": 0.0326,
      "step": 129780
    },
    {
      "epoch": 50.4,
      "learning_rate": 1.3279482200647251e-05,
      "loss": 0.0172,
      "step": 129790
    },
    {
      "epoch": 50.41,
      "learning_rate": 1.3278964401294499e-05,
      "loss": 0.063,
      "step": 129800
    },
    {
      "epoch": 50.41,
      "learning_rate": 1.3278446601941749e-05,
      "loss": 0.1515,
      "step": 129810
    },
    {
      "epoch": 50.42,
      "learning_rate": 1.3277928802588997e-05,
      "loss": 0.0551,
      "step": 129820
    },
    {
      "epoch": 50.42,
      "learning_rate": 1.3277411003236246e-05,
      "loss": 0.0867,
      "step": 129830
    },
    {
      "epoch": 50.42,
      "learning_rate": 1.3276893203883496e-05,
      "loss": 0.007,
      "step": 129840
    },
    {
      "epoch": 50.43,
      "learning_rate": 1.3276375404530746e-05,
      "loss": 0.0073,
      "step": 129850
    },
    {
      "epoch": 50.43,
      "learning_rate": 1.3275857605177995e-05,
      "loss": 0.0773,
      "step": 129860
    },
    {
      "epoch": 50.43,
      "learning_rate": 1.3275339805825245e-05,
      "loss": 0.0248,
      "step": 129870
    },
    {
      "epoch": 50.44,
      "learning_rate": 1.3274822006472493e-05,
      "loss": 0.0754,
      "step": 129880
    },
    {
      "epoch": 50.44,
      "learning_rate": 1.327430420711974e-05,
      "loss": 0.1557,
      "step": 129890
    },
    {
      "epoch": 50.45,
      "learning_rate": 1.327378640776699e-05,
      "loss": 0.0076,
      "step": 129900
    },
    {
      "epoch": 50.45,
      "learning_rate": 1.327326860841424e-05,
      "loss": 0.125,
      "step": 129910
    },
    {
      "epoch": 50.45,
      "learning_rate": 1.327275080906149e-05,
      "loss": 0.055,
      "step": 129920
    },
    {
      "epoch": 50.46,
      "learning_rate": 1.3272233009708739e-05,
      "loss": 0.0918,
      "step": 129930
    },
    {
      "epoch": 50.46,
      "learning_rate": 1.3271715210355989e-05,
      "loss": 0.1725,
      "step": 129940
    },
    {
      "epoch": 50.47,
      "learning_rate": 1.3271197411003238e-05,
      "loss": 0.1927,
      "step": 129950
    },
    {
      "epoch": 50.47,
      "learning_rate": 1.3270679611650486e-05,
      "loss": 0.2174,
      "step": 129960
    },
    {
      "epoch": 50.47,
      "learning_rate": 1.3270161812297734e-05,
      "loss": 0.0632,
      "step": 129970
    },
    {
      "epoch": 50.48,
      "learning_rate": 1.3269644012944984e-05,
      "loss": 0.0758,
      "step": 129980
    },
    {
      "epoch": 50.48,
      "learning_rate": 1.3269126213592233e-05,
      "loss": 0.012,
      "step": 129990
    },
    {
      "epoch": 50.49,
      "learning_rate": 1.3268608414239483e-05,
      "loss": 0.0932,
      "step": 130000
    },
    {
      "epoch": 50.49,
      "learning_rate": 1.3268090614886733e-05,
      "loss": 0.0605,
      "step": 130010
    },
    {
      "epoch": 50.49,
      "learning_rate": 1.3267572815533982e-05,
      "loss": 0.0221,
      "step": 130020
    },
    {
      "epoch": 50.5,
      "learning_rate": 1.3267055016181232e-05,
      "loss": 0.0311,
      "step": 130030
    },
    {
      "epoch": 50.5,
      "learning_rate": 1.3266537216828478e-05,
      "loss": 0.0721,
      "step": 130040
    },
    {
      "epoch": 50.5,
      "learning_rate": 1.3266019417475728e-05,
      "loss": 0.0138,
      "step": 130050
    },
    {
      "epoch": 50.51,
      "learning_rate": 1.3265501618122977e-05,
      "loss": 0.164,
      "step": 130060
    },
    {
      "epoch": 50.51,
      "learning_rate": 1.3264983818770227e-05,
      "loss": 0.1385,
      "step": 130070
    },
    {
      "epoch": 50.52,
      "learning_rate": 1.3264466019417477e-05,
      "loss": 0.0605,
      "step": 130080
    },
    {
      "epoch": 50.52,
      "learning_rate": 1.3263948220064726e-05,
      "loss": 0.0822,
      "step": 130090
    },
    {
      "epoch": 50.52,
      "learning_rate": 1.3263430420711976e-05,
      "loss": 0.0632,
      "step": 130100
    },
    {
      "epoch": 50.53,
      "learning_rate": 1.3262912621359225e-05,
      "loss": 0.0425,
      "step": 130110
    },
    {
      "epoch": 50.53,
      "learning_rate": 1.3262394822006475e-05,
      "loss": 0.0877,
      "step": 130120
    },
    {
      "epoch": 50.54,
      "learning_rate": 1.3261877022653721e-05,
      "loss": 0.1132,
      "step": 130130
    },
    {
      "epoch": 50.54,
      "learning_rate": 1.3261359223300971e-05,
      "loss": 0.0942,
      "step": 130140
    },
    {
      "epoch": 50.54,
      "learning_rate": 1.326084142394822e-05,
      "loss": 0.1912,
      "step": 130150
    },
    {
      "epoch": 50.55,
      "learning_rate": 1.326032362459547e-05,
      "loss": 0.0007,
      "step": 130160
    },
    {
      "epoch": 50.55,
      "learning_rate": 1.325980582524272e-05,
      "loss": 0.0679,
      "step": 130170
    },
    {
      "epoch": 50.56,
      "learning_rate": 1.325928802588997e-05,
      "loss": 0.1151,
      "step": 130180
    },
    {
      "epoch": 50.56,
      "learning_rate": 1.3258770226537219e-05,
      "loss": 0.0894,
      "step": 130190
    },
    {
      "epoch": 50.56,
      "learning_rate": 1.3258252427184469e-05,
      "loss": 0.1693,
      "step": 130200
    },
    {
      "epoch": 50.57,
      "learning_rate": 1.3257734627831715e-05,
      "loss": 0.0379,
      "step": 130210
    },
    {
      "epoch": 50.57,
      "learning_rate": 1.3257216828478964e-05,
      "loss": 0.0899,
      "step": 130220
    },
    {
      "epoch": 50.57,
      "learning_rate": 1.3256699029126214e-05,
      "loss": 0.0394,
      "step": 130230
    },
    {
      "epoch": 50.58,
      "learning_rate": 1.3256181229773464e-05,
      "loss": 0.1063,
      "step": 130240
    },
    {
      "epoch": 50.58,
      "learning_rate": 1.3255663430420713e-05,
      "loss": 0.1492,
      "step": 130250
    },
    {
      "epoch": 50.59,
      "learning_rate": 1.3255145631067963e-05,
      "loss": 0.1158,
      "step": 130260
    },
    {
      "epoch": 50.59,
      "learning_rate": 1.3254627831715213e-05,
      "loss": 0.0749,
      "step": 130270
    },
    {
      "epoch": 50.59,
      "learning_rate": 1.3254110032362462e-05,
      "loss": 0.2305,
      "step": 130280
    },
    {
      "epoch": 50.6,
      "learning_rate": 1.3253592233009708e-05,
      "loss": 0.0591,
      "step": 130290
    },
    {
      "epoch": 50.6,
      "learning_rate": 1.3253074433656958e-05,
      "loss": 0.0728,
      "step": 130300
    },
    {
      "epoch": 50.61,
      "learning_rate": 1.3252556634304208e-05,
      "loss": 0.0701,
      "step": 130310
    },
    {
      "epoch": 50.61,
      "learning_rate": 1.3252038834951457e-05,
      "loss": 0.0985,
      "step": 130320
    },
    {
      "epoch": 50.61,
      "learning_rate": 1.3251521035598707e-05,
      "loss": 0.0531,
      "step": 130330
    },
    {
      "epoch": 50.62,
      "learning_rate": 1.3251003236245956e-05,
      "loss": 0.1245,
      "step": 130340
    },
    {
      "epoch": 50.62,
      "learning_rate": 1.3250485436893206e-05,
      "loss": 0.2005,
      "step": 130350
    },
    {
      "epoch": 50.63,
      "learning_rate": 1.3249967637540456e-05,
      "loss": 0.1264,
      "step": 130360
    },
    {
      "epoch": 50.63,
      "learning_rate": 1.3249449838187702e-05,
      "loss": 0.0703,
      "step": 130370
    },
    {
      "epoch": 50.63,
      "learning_rate": 1.3248932038834952e-05,
      "loss": 0.1338,
      "step": 130380
    },
    {
      "epoch": 50.64,
      "learning_rate": 1.3248414239482201e-05,
      "loss": 0.0581,
      "step": 130390
    },
    {
      "epoch": 50.64,
      "learning_rate": 1.324789644012945e-05,
      "loss": 0.0534,
      "step": 130400
    },
    {
      "epoch": 50.64,
      "learning_rate": 1.32473786407767e-05,
      "loss": 0.1012,
      "step": 130410
    },
    {
      "epoch": 50.65,
      "learning_rate": 1.324686084142395e-05,
      "loss": 0.0748,
      "step": 130420
    },
    {
      "epoch": 50.65,
      "learning_rate": 1.32463430420712e-05,
      "loss": 0.0502,
      "step": 130430
    },
    {
      "epoch": 50.66,
      "learning_rate": 1.324582524271845e-05,
      "loss": 0.1573,
      "step": 130440
    },
    {
      "epoch": 50.66,
      "learning_rate": 1.3245307443365696e-05,
      "loss": 0.1206,
      "step": 130450
    },
    {
      "epoch": 50.66,
      "learning_rate": 1.3244789644012945e-05,
      "loss": 0.1632,
      "step": 130460
    },
    {
      "epoch": 50.67,
      "learning_rate": 1.3244271844660195e-05,
      "loss": 0.0636,
      "step": 130470
    },
    {
      "epoch": 50.67,
      "learning_rate": 1.3243754045307444e-05,
      "loss": 0.032,
      "step": 130480
    },
    {
      "epoch": 50.68,
      "learning_rate": 1.3243236245954694e-05,
      "loss": 0.1203,
      "step": 130490
    },
    {
      "epoch": 50.68,
      "learning_rate": 1.3242718446601944e-05,
      "loss": 0.049,
      "step": 130500
    },
    {
      "epoch": 50.68,
      "learning_rate": 1.3242200647249193e-05,
      "loss": 0.1083,
      "step": 130510
    },
    {
      "epoch": 50.69,
      "learning_rate": 1.3241682847896443e-05,
      "loss": 0.0462,
      "step": 130520
    },
    {
      "epoch": 50.69,
      "learning_rate": 1.3241165048543689e-05,
      "loss": 0.1291,
      "step": 130530
    },
    {
      "epoch": 50.7,
      "learning_rate": 1.3240647249190939e-05,
      "loss": 0.0675,
      "step": 130540
    },
    {
      "epoch": 50.7,
      "learning_rate": 1.3240129449838188e-05,
      "loss": 0.0426,
      "step": 130550
    },
    {
      "epoch": 50.7,
      "learning_rate": 1.3239611650485438e-05,
      "loss": 0.0631,
      "step": 130560
    },
    {
      "epoch": 50.71,
      "learning_rate": 1.3239093851132688e-05,
      "loss": 0.0097,
      "step": 130570
    },
    {
      "epoch": 50.71,
      "learning_rate": 1.3238576051779937e-05,
      "loss": 0.0568,
      "step": 130580
    },
    {
      "epoch": 50.71,
      "learning_rate": 1.3238058252427187e-05,
      "loss": 0.0212,
      "step": 130590
    },
    {
      "epoch": 50.72,
      "learning_rate": 1.3237540453074436e-05,
      "loss": 0.0537,
      "step": 130600
    },
    {
      "epoch": 50.72,
      "learning_rate": 1.3237022653721683e-05,
      "loss": 0.0198,
      "step": 130610
    },
    {
      "epoch": 50.73,
      "learning_rate": 1.3236504854368932e-05,
      "loss": 0.0491,
      "step": 130620
    },
    {
      "epoch": 50.73,
      "learning_rate": 1.3235987055016182e-05,
      "loss": 0.1215,
      "step": 130630
    },
    {
      "epoch": 50.73,
      "learning_rate": 1.3235469255663431e-05,
      "loss": 0.1103,
      "step": 130640
    },
    {
      "epoch": 50.74,
      "learning_rate": 1.3234951456310681e-05,
      "loss": 0.1061,
      "step": 130650
    },
    {
      "epoch": 50.74,
      "learning_rate": 1.323443365695793e-05,
      "loss": 0.0931,
      "step": 130660
    },
    {
      "epoch": 50.75,
      "learning_rate": 1.323391585760518e-05,
      "loss": 0.2891,
      "step": 130670
    },
    {
      "epoch": 50.75,
      "learning_rate": 1.3233398058252428e-05,
      "loss": 0.0821,
      "step": 130680
    },
    {
      "epoch": 50.75,
      "learning_rate": 1.3232880258899678e-05,
      "loss": 0.0026,
      "step": 130690
    },
    {
      "epoch": 50.76,
      "learning_rate": 1.3232362459546926e-05,
      "loss": 0.0261,
      "step": 130700
    },
    {
      "epoch": 50.76,
      "learning_rate": 1.3231844660194175e-05,
      "loss": 0.0551,
      "step": 130710
    },
    {
      "epoch": 50.77,
      "learning_rate": 1.3231326860841425e-05,
      "loss": 0.2067,
      "step": 130720
    },
    {
      "epoch": 50.77,
      "learning_rate": 1.3230809061488675e-05,
      "loss": 0.0722,
      "step": 130730
    },
    {
      "epoch": 50.77,
      "learning_rate": 1.3230291262135924e-05,
      "loss": 0.046,
      "step": 130740
    },
    {
      "epoch": 50.78,
      "learning_rate": 1.3229773462783174e-05,
      "loss": 0.0667,
      "step": 130750
    },
    {
      "epoch": 50.78,
      "learning_rate": 1.3229255663430422e-05,
      "loss": 0.1106,
      "step": 130760
    },
    {
      "epoch": 50.78,
      "learning_rate": 1.3228737864077671e-05,
      "loss": 0.056,
      "step": 130770
    },
    {
      "epoch": 50.79,
      "learning_rate": 1.322822006472492e-05,
      "loss": 0.0731,
      "step": 130780
    },
    {
      "epoch": 50.79,
      "learning_rate": 1.3227702265372169e-05,
      "loss": 0.129,
      "step": 130790
    },
    {
      "epoch": 50.8,
      "learning_rate": 1.3227184466019419e-05,
      "loss": 0.1332,
      "step": 130800
    },
    {
      "epoch": 50.8,
      "learning_rate": 1.3226666666666668e-05,
      "loss": 0.0781,
      "step": 130810
    },
    {
      "epoch": 50.8,
      "learning_rate": 1.3226148867313918e-05,
      "loss": 0.0099,
      "step": 130820
    },
    {
      "epoch": 50.81,
      "learning_rate": 1.3225631067961166e-05,
      "loss": 0.017,
      "step": 130830
    },
    {
      "epoch": 50.81,
      "learning_rate": 1.3225113268608415e-05,
      "loss": 0.0236,
      "step": 130840
    },
    {
      "epoch": 50.82,
      "learning_rate": 1.3224595469255665e-05,
      "loss": 0.1228,
      "step": 130850
    },
    {
      "epoch": 50.82,
      "learning_rate": 1.3224077669902913e-05,
      "loss": 0.1749,
      "step": 130860
    },
    {
      "epoch": 50.82,
      "learning_rate": 1.3223559870550163e-05,
      "loss": 0.0388,
      "step": 130870
    },
    {
      "epoch": 50.83,
      "learning_rate": 1.3223042071197412e-05,
      "loss": 0.0283,
      "step": 130880
    },
    {
      "epoch": 50.83,
      "learning_rate": 1.3222524271844662e-05,
      "loss": 0.2304,
      "step": 130890
    },
    {
      "epoch": 50.83,
      "learning_rate": 1.3222006472491911e-05,
      "loss": 0.1408,
      "step": 130900
    },
    {
      "epoch": 50.84,
      "learning_rate": 1.322148867313916e-05,
      "loss": 0.0254,
      "step": 130910
    },
    {
      "epoch": 50.84,
      "learning_rate": 1.3220970873786409e-05,
      "loss": 0.0363,
      "step": 130920
    },
    {
      "epoch": 50.85,
      "learning_rate": 1.3220453074433659e-05,
      "loss": 0.0514,
      "step": 130930
    },
    {
      "epoch": 50.85,
      "learning_rate": 1.3219935275080906e-05,
      "loss": 0.036,
      "step": 130940
    },
    {
      "epoch": 50.85,
      "learning_rate": 1.3219417475728156e-05,
      "loss": 0.0967,
      "step": 130950
    },
    {
      "epoch": 50.86,
      "learning_rate": 1.3218899676375406e-05,
      "loss": 0.155,
      "step": 130960
    },
    {
      "epoch": 50.86,
      "learning_rate": 1.3218381877022655e-05,
      "loss": 0.0285,
      "step": 130970
    },
    {
      "epoch": 50.87,
      "learning_rate": 1.3217864077669903e-05,
      "loss": 0.0763,
      "step": 130980
    },
    {
      "epoch": 50.87,
      "learning_rate": 1.3217346278317153e-05,
      "loss": 0.0046,
      "step": 130990
    },
    {
      "epoch": 50.87,
      "learning_rate": 1.3216828478964402e-05,
      "loss": 0.0159,
      "step": 131000
    },
    {
      "epoch": 50.88,
      "learning_rate": 1.3216310679611652e-05,
      "loss": 0.043,
      "step": 131010
    },
    {
      "epoch": 50.88,
      "learning_rate": 1.32157928802589e-05,
      "loss": 0.109,
      "step": 131020
    },
    {
      "epoch": 50.89,
      "learning_rate": 1.321527508090615e-05,
      "loss": 0.0951,
      "step": 131030
    },
    {
      "epoch": 50.89,
      "learning_rate": 1.32147572815534e-05,
      "loss": 0.0399,
      "step": 131040
    },
    {
      "epoch": 50.89,
      "learning_rate": 1.3214239482200649e-05,
      "loss": 0.0826,
      "step": 131050
    },
    {
      "epoch": 50.9,
      "learning_rate": 1.3213721682847897e-05,
      "loss": 0.0697,
      "step": 131060
    },
    {
      "epoch": 50.9,
      "learning_rate": 1.3213203883495146e-05,
      "loss": 0.0053,
      "step": 131070
    },
    {
      "epoch": 50.9,
      "learning_rate": 1.3212686084142396e-05,
      "loss": 0.1033,
      "step": 131080
    },
    {
      "epoch": 50.91,
      "learning_rate": 1.3212168284789646e-05,
      "loss": 0.0128,
      "step": 131090
    },
    {
      "epoch": 50.91,
      "learning_rate": 1.3211650485436894e-05,
      "loss": 0.001,
      "step": 131100
    },
    {
      "epoch": 50.92,
      "learning_rate": 1.3211132686084143e-05,
      "loss": 0.0116,
      "step": 131110
    },
    {
      "epoch": 50.92,
      "learning_rate": 1.3210614886731393e-05,
      "loss": 0.0821,
      "step": 131120
    },
    {
      "epoch": 50.92,
      "learning_rate": 1.3210097087378642e-05,
      "loss": 0.059,
      "step": 131130
    },
    {
      "epoch": 50.93,
      "learning_rate": 1.320957928802589e-05,
      "loss": 0.0591,
      "step": 131140
    },
    {
      "epoch": 50.93,
      "learning_rate": 1.320906148867314e-05,
      "loss": 0.0518,
      "step": 131150
    },
    {
      "epoch": 50.94,
      "learning_rate": 1.320854368932039e-05,
      "loss": 0.2752,
      "step": 131160
    },
    {
      "epoch": 50.94,
      "learning_rate": 1.320802588996764e-05,
      "loss": 0.0007,
      "step": 131170
    },
    {
      "epoch": 50.94,
      "learning_rate": 1.3207508090614889e-05,
      "loss": 0.0352,
      "step": 131180
    },
    {
      "epoch": 50.95,
      "learning_rate": 1.3206990291262137e-05,
      "loss": 0.0436,
      "step": 131190
    },
    {
      "epoch": 50.95,
      "learning_rate": 1.3206472491909386e-05,
      "loss": 0.1107,
      "step": 131200
    },
    {
      "epoch": 50.96,
      "learning_rate": 1.3205954692556634e-05,
      "loss": 0.1102,
      "step": 131210
    },
    {
      "epoch": 50.96,
      "learning_rate": 1.3205436893203884e-05,
      "loss": 0.0896,
      "step": 131220
    },
    {
      "epoch": 50.96,
      "learning_rate": 1.3204919093851134e-05,
      "loss": 0.122,
      "step": 131230
    },
    {
      "epoch": 50.97,
      "learning_rate": 1.3204401294498383e-05,
      "loss": 0.1999,
      "step": 131240
    },
    {
      "epoch": 50.97,
      "learning_rate": 1.3203883495145633e-05,
      "loss": 0.0604,
      "step": 131250
    },
    {
      "epoch": 50.97,
      "learning_rate": 1.3203365695792882e-05,
      "loss": 0.4277,
      "step": 131260
    },
    {
      "epoch": 50.98,
      "learning_rate": 1.320284789644013e-05,
      "loss": 0.0934,
      "step": 131270
    },
    {
      "epoch": 50.98,
      "learning_rate": 1.320233009708738e-05,
      "loss": 0.0234,
      "step": 131280
    },
    {
      "epoch": 50.99,
      "learning_rate": 1.3201812297734628e-05,
      "loss": 0.1346,
      "step": 131290
    },
    {
      "epoch": 50.99,
      "learning_rate": 1.3201294498381877e-05,
      "loss": 0.138,
      "step": 131300
    },
    {
      "epoch": 50.99,
      "learning_rate": 1.3200776699029127e-05,
      "loss": 0.0856,
      "step": 131310
    },
    {
      "epoch": 51.0,
      "learning_rate": 1.3200258899676377e-05,
      "loss": 0.1785,
      "step": 131320
    },
    {
      "epoch": 51.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.3273870646953583,
      "eval_runtime": 8.2475,
      "eval_samples_per_second": 440.741,
      "eval_steps_per_second": 55.168,
      "step": 131325
    },
    {
      "epoch": 51.0,
      "learning_rate": 1.3199741100323626e-05,
      "loss": 0.1529,
      "step": 131330
    },
    {
      "epoch": 51.01,
      "learning_rate": 1.3199223300970876e-05,
      "loss": 0.087,
      "step": 131340
    },
    {
      "epoch": 51.01,
      "learning_rate": 1.3198705501618124e-05,
      "loss": 0.0113,
      "step": 131350
    },
    {
      "epoch": 51.01,
      "learning_rate": 1.3198187702265372e-05,
      "loss": 0.0405,
      "step": 131360
    },
    {
      "epoch": 51.02,
      "learning_rate": 1.3197669902912621e-05,
      "loss": 0.0551,
      "step": 131370
    },
    {
      "epoch": 51.02,
      "learning_rate": 1.3197152103559871e-05,
      "loss": 0.0927,
      "step": 131380
    },
    {
      "epoch": 51.03,
      "learning_rate": 1.319663430420712e-05,
      "loss": 0.1212,
      "step": 131390
    },
    {
      "epoch": 51.03,
      "learning_rate": 1.319611650485437e-05,
      "loss": 0.0195,
      "step": 131400
    },
    {
      "epoch": 51.03,
      "learning_rate": 1.319559870550162e-05,
      "loss": 0.1405,
      "step": 131410
    },
    {
      "epoch": 51.04,
      "learning_rate": 1.319508090614887e-05,
      "loss": 0.0597,
      "step": 131420
    },
    {
      "epoch": 51.04,
      "learning_rate": 1.3194563106796117e-05,
      "loss": 0.2453,
      "step": 131430
    },
    {
      "epoch": 51.04,
      "learning_rate": 1.3194045307443365e-05,
      "loss": 0.1105,
      "step": 131440
    },
    {
      "epoch": 51.05,
      "learning_rate": 1.3193527508090615e-05,
      "loss": 0.1077,
      "step": 131450
    },
    {
      "epoch": 51.05,
      "learning_rate": 1.3193009708737865e-05,
      "loss": 0.0573,
      "step": 131460
    },
    {
      "epoch": 51.06,
      "learning_rate": 1.3192491909385114e-05,
      "loss": 0.0003,
      "step": 131470
    },
    {
      "epoch": 51.06,
      "learning_rate": 1.3191974110032364e-05,
      "loss": 0.0224,
      "step": 131480
    },
    {
      "epoch": 51.06,
      "learning_rate": 1.3191456310679613e-05,
      "loss": 0.0751,
      "step": 131490
    },
    {
      "epoch": 51.07,
      "learning_rate": 1.3190938511326863e-05,
      "loss": 0.0982,
      "step": 131500
    },
    {
      "epoch": 51.07,
      "learning_rate": 1.319042071197411e-05,
      "loss": 0.0719,
      "step": 131510
    },
    {
      "epoch": 51.08,
      "learning_rate": 1.3189902912621359e-05,
      "loss": 0.1023,
      "step": 131520
    },
    {
      "epoch": 51.08,
      "learning_rate": 1.3189385113268609e-05,
      "loss": 0.2425,
      "step": 131530
    },
    {
      "epoch": 51.08,
      "learning_rate": 1.3188867313915858e-05,
      "loss": 0.0245,
      "step": 131540
    },
    {
      "epoch": 51.09,
      "learning_rate": 1.3188349514563108e-05,
      "loss": 0.0976,
      "step": 131550
    },
    {
      "epoch": 51.09,
      "learning_rate": 1.3187831715210357e-05,
      "loss": 0.0864,
      "step": 131560
    },
    {
      "epoch": 51.1,
      "learning_rate": 1.3187313915857607e-05,
      "loss": 0.0928,
      "step": 131570
    },
    {
      "epoch": 51.1,
      "learning_rate": 1.3186796116504857e-05,
      "loss": 0.0568,
      "step": 131580
    },
    {
      "epoch": 51.1,
      "learning_rate": 1.3186278317152103e-05,
      "loss": 0.1058,
      "step": 131590
    },
    {
      "epoch": 51.11,
      "learning_rate": 1.3185760517799352e-05,
      "loss": 0.0882,
      "step": 131600
    },
    {
      "epoch": 51.11,
      "learning_rate": 1.3185242718446602e-05,
      "loss": 0.1466,
      "step": 131610
    },
    {
      "epoch": 51.11,
      "learning_rate": 1.3184724919093852e-05,
      "loss": 0.1514,
      "step": 131620
    },
    {
      "epoch": 51.12,
      "learning_rate": 1.3184207119741101e-05,
      "loss": 0.0326,
      "step": 131630
    },
    {
      "epoch": 51.12,
      "learning_rate": 1.3183689320388351e-05,
      "loss": 0.0873,
      "step": 131640
    },
    {
      "epoch": 51.13,
      "learning_rate": 1.31831715210356e-05,
      "loss": 0.0805,
      "step": 131650
    },
    {
      "epoch": 51.13,
      "learning_rate": 1.318265372168285e-05,
      "loss": 0.1236,
      "step": 131660
    },
    {
      "epoch": 51.13,
      "learning_rate": 1.3182135922330096e-05,
      "loss": 0.0734,
      "step": 131670
    },
    {
      "epoch": 51.14,
      "learning_rate": 1.3181618122977346e-05,
      "loss": 0.03,
      "step": 131680
    },
    {
      "epoch": 51.14,
      "learning_rate": 1.3181100323624596e-05,
      "loss": 0.0011,
      "step": 131690
    },
    {
      "epoch": 51.15,
      "learning_rate": 1.3180582524271845e-05,
      "loss": 0.1,
      "step": 131700
    },
    {
      "epoch": 51.15,
      "learning_rate": 1.3180064724919095e-05,
      "loss": 0.001,
      "step": 131710
    },
    {
      "epoch": 51.15,
      "learning_rate": 1.3179546925566344e-05,
      "loss": 0.0625,
      "step": 131720
    },
    {
      "epoch": 51.16,
      "learning_rate": 1.3179029126213594e-05,
      "loss": 0.0978,
      "step": 131730
    },
    {
      "epoch": 51.16,
      "learning_rate": 1.3178511326860844e-05,
      "loss": 0.0245,
      "step": 131740
    },
    {
      "epoch": 51.17,
      "learning_rate": 1.3177993527508093e-05,
      "loss": 0.1179,
      "step": 131750
    },
    {
      "epoch": 51.17,
      "learning_rate": 1.317747572815534e-05,
      "loss": 0.0261,
      "step": 131760
    },
    {
      "epoch": 51.17,
      "learning_rate": 1.317695792880259e-05,
      "loss": 0.0929,
      "step": 131770
    },
    {
      "epoch": 51.18,
      "learning_rate": 1.3176440129449839e-05,
      "loss": 0.0909,
      "step": 131780
    },
    {
      "epoch": 51.18,
      "learning_rate": 1.3175922330097088e-05,
      "loss": 0.098,
      "step": 131790
    },
    {
      "epoch": 51.18,
      "learning_rate": 1.3175404530744338e-05,
      "loss": 0.0432,
      "step": 131800
    },
    {
      "epoch": 51.19,
      "learning_rate": 1.3174886731391588e-05,
      "loss": 0.0686,
      "step": 131810
    },
    {
      "epoch": 51.19,
      "learning_rate": 1.3174368932038837e-05,
      "loss": 0.1541,
      "step": 131820
    },
    {
      "epoch": 51.2,
      "learning_rate": 1.3173851132686087e-05,
      "loss": 0.1656,
      "step": 131830
    },
    {
      "epoch": 51.2,
      "learning_rate": 1.3173333333333333e-05,
      "loss": 0.0859,
      "step": 131840
    },
    {
      "epoch": 51.2,
      "learning_rate": 1.3172815533980583e-05,
      "loss": 0.0727,
      "step": 131850
    },
    {
      "epoch": 51.21,
      "learning_rate": 1.3172297734627832e-05,
      "loss": 0.1848,
      "step": 131860
    },
    {
      "epoch": 51.21,
      "learning_rate": 1.3171779935275082e-05,
      "loss": 0.0209,
      "step": 131870
    },
    {
      "epoch": 51.22,
      "learning_rate": 1.3171262135922332e-05,
      "loss": 0.006,
      "step": 131880
    },
    {
      "epoch": 51.22,
      "learning_rate": 1.3170744336569581e-05,
      "loss": 0.0742,
      "step": 131890
    },
    {
      "epoch": 51.22,
      "learning_rate": 1.317022653721683e-05,
      "loss": 0.0711,
      "step": 131900
    },
    {
      "epoch": 51.23,
      "learning_rate": 1.316970873786408e-05,
      "loss": 0.0162,
      "step": 131910
    },
    {
      "epoch": 51.23,
      "learning_rate": 1.3169190938511327e-05,
      "loss": 0.1001,
      "step": 131920
    },
    {
      "epoch": 51.23,
      "learning_rate": 1.3168673139158576e-05,
      "loss": 0.0018,
      "step": 131930
    },
    {
      "epoch": 51.24,
      "learning_rate": 1.3168155339805826e-05,
      "loss": 0.1019,
      "step": 131940
    },
    {
      "epoch": 51.24,
      "learning_rate": 1.3167637540453076e-05,
      "loss": 0.0362,
      "step": 131950
    },
    {
      "epoch": 51.25,
      "learning_rate": 1.3167119741100325e-05,
      "loss": 0.0448,
      "step": 131960
    },
    {
      "epoch": 51.25,
      "learning_rate": 1.3166601941747575e-05,
      "loss": 0.0746,
      "step": 131970
    },
    {
      "epoch": 51.25,
      "learning_rate": 1.3166084142394824e-05,
      "loss": 0.0773,
      "step": 131980
    },
    {
      "epoch": 51.26,
      "learning_rate": 1.3165566343042074e-05,
      "loss": 0.1832,
      "step": 131990
    },
    {
      "epoch": 51.26,
      "learning_rate": 1.316504854368932e-05,
      "loss": 0.0048,
      "step": 132000
    },
    {
      "epoch": 51.27,
      "learning_rate": 1.316453074433657e-05,
      "loss": 0.0996,
      "step": 132010
    },
    {
      "epoch": 51.27,
      "learning_rate": 1.316401294498382e-05,
      "loss": 0.1454,
      "step": 132020
    },
    {
      "epoch": 51.27,
      "learning_rate": 1.3163495145631069e-05,
      "loss": 0.0055,
      "step": 132030
    },
    {
      "epoch": 51.28,
      "learning_rate": 1.3162977346278319e-05,
      "loss": 0.1092,
      "step": 132040
    },
    {
      "epoch": 51.28,
      "learning_rate": 1.3162459546925568e-05,
      "loss": 0.1275,
      "step": 132050
    },
    {
      "epoch": 51.29,
      "learning_rate": 1.3161941747572818e-05,
      "loss": 0.0388,
      "step": 132060
    },
    {
      "epoch": 51.29,
      "learning_rate": 1.3161423948220068e-05,
      "loss": 0.1498,
      "step": 132070
    },
    {
      "epoch": 51.29,
      "learning_rate": 1.3160906148867314e-05,
      "loss": 0.1005,
      "step": 132080
    },
    {
      "epoch": 51.3,
      "learning_rate": 1.3160388349514563e-05,
      "loss": 0.0317,
      "step": 132090
    },
    {
      "epoch": 51.3,
      "learning_rate": 1.3159870550161813e-05,
      "loss": 0.0373,
      "step": 132100
    },
    {
      "epoch": 51.3,
      "learning_rate": 1.3159352750809063e-05,
      "loss": 0.0155,
      "step": 132110
    },
    {
      "epoch": 51.31,
      "learning_rate": 1.3158834951456312e-05,
      "loss": 0.021,
      "step": 132120
    },
    {
      "epoch": 51.31,
      "learning_rate": 1.3158317152103562e-05,
      "loss": 0.0356,
      "step": 132130
    },
    {
      "epoch": 51.32,
      "learning_rate": 1.3157799352750812e-05,
      "loss": 0.0901,
      "step": 132140
    },
    {
      "epoch": 51.32,
      "learning_rate": 1.315728155339806e-05,
      "loss": 0.0418,
      "step": 132150
    },
    {
      "epoch": 51.32,
      "learning_rate": 1.3156763754045307e-05,
      "loss": 0.156,
      "step": 132160
    },
    {
      "epoch": 51.33,
      "learning_rate": 1.3156245954692557e-05,
      "loss": 0.0461,
      "step": 132170
    },
    {
      "epoch": 51.33,
      "learning_rate": 1.3155728155339807e-05,
      "loss": 0.068,
      "step": 132180
    },
    {
      "epoch": 51.34,
      "learning_rate": 1.3155210355987056e-05,
      "loss": 0.0952,
      "step": 132190
    },
    {
      "epoch": 51.34,
      "learning_rate": 1.3154692556634306e-05,
      "loss": 0.2315,
      "step": 132200
    },
    {
      "epoch": 51.34,
      "learning_rate": 1.3154174757281555e-05,
      "loss": 0.0729,
      "step": 132210
    },
    {
      "epoch": 51.35,
      "learning_rate": 1.3153656957928805e-05,
      "loss": 0.0691,
      "step": 132220
    },
    {
      "epoch": 51.35,
      "learning_rate": 1.3153139158576053e-05,
      "loss": 0.1652,
      "step": 132230
    },
    {
      "epoch": 51.36,
      "learning_rate": 1.3152621359223301e-05,
      "loss": 0.1815,
      "step": 132240
    },
    {
      "epoch": 51.36,
      "learning_rate": 1.315210355987055e-05,
      "loss": 0.004,
      "step": 132250
    },
    {
      "epoch": 51.36,
      "learning_rate": 1.31515857605178e-05,
      "loss": 0.1538,
      "step": 132260
    },
    {
      "epoch": 51.37,
      "learning_rate": 1.315106796116505e-05,
      "loss": 0.1328,
      "step": 132270
    },
    {
      "epoch": 51.37,
      "learning_rate": 1.31505501618123e-05,
      "loss": 0.0499,
      "step": 132280
    },
    {
      "epoch": 51.37,
      "learning_rate": 1.3150032362459549e-05,
      "loss": 0.1438,
      "step": 132290
    },
    {
      "epoch": 51.38,
      "learning_rate": 1.3149514563106797e-05,
      "loss": 0.1341,
      "step": 132300
    },
    {
      "epoch": 51.38,
      "learning_rate": 1.3148996763754047e-05,
      "loss": 0.0881,
      "step": 132310
    },
    {
      "epoch": 51.39,
      "learning_rate": 1.3148478964401296e-05,
      "loss": 0.0463,
      "step": 132320
    },
    {
      "epoch": 51.39,
      "learning_rate": 1.3147961165048544e-05,
      "loss": 0.0585,
      "step": 132330
    },
    {
      "epoch": 51.39,
      "learning_rate": 1.3147443365695794e-05,
      "loss": 0.0521,
      "step": 132340
    },
    {
      "epoch": 51.4,
      "learning_rate": 1.3146925566343043e-05,
      "loss": 0.0936,
      "step": 132350
    },
    {
      "epoch": 51.4,
      "learning_rate": 1.3146407766990293e-05,
      "loss": 0.0562,
      "step": 132360
    },
    {
      "epoch": 51.41,
      "learning_rate": 1.3145889967637543e-05,
      "loss": 0.0016,
      "step": 132370
    },
    {
      "epoch": 51.41,
      "learning_rate": 1.314537216828479e-05,
      "loss": 0.0572,
      "step": 132380
    },
    {
      "epoch": 51.41,
      "learning_rate": 1.314485436893204e-05,
      "loss": 0.1511,
      "step": 132390
    },
    {
      "epoch": 51.42,
      "learning_rate": 1.314433656957929e-05,
      "loss": 0.0648,
      "step": 132400
    },
    {
      "epoch": 51.42,
      "learning_rate": 1.3143818770226538e-05,
      "loss": 0.1046,
      "step": 132410
    },
    {
      "epoch": 51.43,
      "learning_rate": 1.3143300970873787e-05,
      "loss": 0.0633,
      "step": 132420
    },
    {
      "epoch": 51.43,
      "learning_rate": 1.3142783171521037e-05,
      "loss": 0.1827,
      "step": 132430
    },
    {
      "epoch": 51.43,
      "learning_rate": 1.3142265372168286e-05,
      "loss": 0.0316,
      "step": 132440
    },
    {
      "epoch": 51.44,
      "learning_rate": 1.3141747572815534e-05,
      "loss": 0.0719,
      "step": 132450
    },
    {
      "epoch": 51.44,
      "learning_rate": 1.3141229773462784e-05,
      "loss": 0.1907,
      "step": 132460
    },
    {
      "epoch": 51.44,
      "learning_rate": 1.3140711974110034e-05,
      "loss": 0.1051,
      "step": 132470
    },
    {
      "epoch": 51.45,
      "learning_rate": 1.3140194174757283e-05,
      "loss": 0.1064,
      "step": 132480
    },
    {
      "epoch": 51.45,
      "learning_rate": 1.3139676375404531e-05,
      "loss": 0.0908,
      "step": 132490
    },
    {
      "epoch": 51.46,
      "learning_rate": 1.313915857605178e-05,
      "loss": 0.1072,
      "step": 132500
    },
    {
      "epoch": 51.46,
      "learning_rate": 1.313864077669903e-05,
      "loss": 0.1335,
      "step": 132510
    },
    {
      "epoch": 51.46,
      "learning_rate": 1.313812297734628e-05,
      "loss": 0.0015,
      "step": 132520
    },
    {
      "epoch": 51.47,
      "learning_rate": 1.3137605177993528e-05,
      "loss": 0.0283,
      "step": 132530
    },
    {
      "epoch": 51.47,
      "learning_rate": 1.3137087378640778e-05,
      "loss": 0.0435,
      "step": 132540
    },
    {
      "epoch": 51.48,
      "learning_rate": 1.3136569579288027e-05,
      "loss": 0.0554,
      "step": 132550
    },
    {
      "epoch": 51.48,
      "learning_rate": 1.3136051779935277e-05,
      "loss": 0.0455,
      "step": 132560
    },
    {
      "epoch": 51.48,
      "learning_rate": 1.3135533980582525e-05,
      "loss": 0.0911,
      "step": 132570
    },
    {
      "epoch": 51.49,
      "learning_rate": 1.3135016181229774e-05,
      "loss": 0.0772,
      "step": 132580
    },
    {
      "epoch": 51.49,
      "learning_rate": 1.3134498381877024e-05,
      "loss": 0.1302,
      "step": 132590
    },
    {
      "epoch": 51.5,
      "learning_rate": 1.3133980582524274e-05,
      "loss": 0.0146,
      "step": 132600
    },
    {
      "epoch": 51.5,
      "learning_rate": 1.3133462783171522e-05,
      "loss": 0.0343,
      "step": 132610
    },
    {
      "epoch": 51.5,
      "learning_rate": 1.3132944983818771e-05,
      "loss": 0.1143,
      "step": 132620
    },
    {
      "epoch": 51.51,
      "learning_rate": 1.313242718446602e-05,
      "loss": 0.0393,
      "step": 132630
    },
    {
      "epoch": 51.51,
      "learning_rate": 1.313190938511327e-05,
      "loss": 0.1122,
      "step": 132640
    },
    {
      "epoch": 51.51,
      "learning_rate": 1.3131391585760518e-05,
      "loss": 0.0471,
      "step": 132650
    },
    {
      "epoch": 51.52,
      "learning_rate": 1.3130873786407768e-05,
      "loss": 0.0488,
      "step": 132660
    },
    {
      "epoch": 51.52,
      "learning_rate": 1.3130355987055018e-05,
      "loss": 0.12,
      "step": 132670
    },
    {
      "epoch": 51.53,
      "learning_rate": 1.3129838187702265e-05,
      "loss": 0.0108,
      "step": 132680
    },
    {
      "epoch": 51.53,
      "learning_rate": 1.3129320388349515e-05,
      "loss": 0.0216,
      "step": 132690
    },
    {
      "epoch": 51.53,
      "learning_rate": 1.3128802588996765e-05,
      "loss": 0.0028,
      "step": 132700
    },
    {
      "epoch": 51.54,
      "learning_rate": 1.3128284789644014e-05,
      "loss": 0.128,
      "step": 132710
    },
    {
      "epoch": 51.54,
      "learning_rate": 1.3127766990291264e-05,
      "loss": 0.0446,
      "step": 132720
    },
    {
      "epoch": 51.55,
      "learning_rate": 1.3127249190938512e-05,
      "loss": 0.0452,
      "step": 132730
    },
    {
      "epoch": 51.55,
      "learning_rate": 1.3126731391585761e-05,
      "loss": 0.185,
      "step": 132740
    },
    {
      "epoch": 51.55,
      "learning_rate": 1.3126213592233011e-05,
      "loss": 0.1568,
      "step": 132750
    },
    {
      "epoch": 51.56,
      "learning_rate": 1.3125695792880259e-05,
      "loss": 0.1241,
      "step": 132760
    },
    {
      "epoch": 51.56,
      "learning_rate": 1.3125177993527509e-05,
      "loss": 0.1309,
      "step": 132770
    },
    {
      "epoch": 51.57,
      "learning_rate": 1.3124660194174758e-05,
      "loss": 0.0132,
      "step": 132780
    },
    {
      "epoch": 51.57,
      "learning_rate": 1.3124142394822008e-05,
      "loss": 0.0448,
      "step": 132790
    },
    {
      "epoch": 51.57,
      "learning_rate": 1.3123624595469258e-05,
      "loss": 0.0428,
      "step": 132800
    },
    {
      "epoch": 51.58,
      "learning_rate": 1.3123106796116505e-05,
      "loss": 0.1283,
      "step": 132810
    },
    {
      "epoch": 51.58,
      "learning_rate": 1.3122588996763755e-05,
      "loss": 0.0358,
      "step": 132820
    },
    {
      "epoch": 51.58,
      "learning_rate": 1.3122071197411003e-05,
      "loss": 0.0291,
      "step": 132830
    },
    {
      "epoch": 51.59,
      "learning_rate": 1.3121553398058253e-05,
      "loss": 0.0604,
      "step": 132840
    },
    {
      "epoch": 51.59,
      "learning_rate": 1.3121035598705502e-05,
      "loss": 0.0336,
      "step": 132850
    },
    {
      "epoch": 51.6,
      "learning_rate": 1.3120517799352752e-05,
      "loss": 0.009,
      "step": 132860
    },
    {
      "epoch": 51.6,
      "learning_rate": 1.3120000000000001e-05,
      "loss": 0.1414,
      "step": 132870
    },
    {
      "epoch": 51.6,
      "learning_rate": 1.3119482200647251e-05,
      "loss": 0.0239,
      "step": 132880
    },
    {
      "epoch": 51.61,
      "learning_rate": 1.31189644012945e-05,
      "loss": 0.0507,
      "step": 132890
    },
    {
      "epoch": 51.61,
      "learning_rate": 1.3118446601941749e-05,
      "loss": 0.1254,
      "step": 132900
    },
    {
      "epoch": 51.62,
      "learning_rate": 1.3117928802588997e-05,
      "loss": 0.1163,
      "step": 132910
    },
    {
      "epoch": 51.62,
      "learning_rate": 1.3117411003236246e-05,
      "loss": 0.1097,
      "step": 132920
    },
    {
      "epoch": 51.62,
      "learning_rate": 1.3116893203883496e-05,
      "loss": 0.1226,
      "step": 132930
    },
    {
      "epoch": 51.63,
      "learning_rate": 1.3116375404530745e-05,
      "loss": 0.0275,
      "step": 132940
    },
    {
      "epoch": 51.63,
      "learning_rate": 1.3115857605177995e-05,
      "loss": 0.0182,
      "step": 132950
    },
    {
      "epoch": 51.63,
      "learning_rate": 1.3115339805825245e-05,
      "loss": 0.0746,
      "step": 132960
    },
    {
      "epoch": 51.64,
      "learning_rate": 1.3114822006472494e-05,
      "loss": 0.1684,
      "step": 132970
    },
    {
      "epoch": 51.64,
      "learning_rate": 1.311430420711974e-05,
      "loss": 0.0561,
      "step": 132980
    },
    {
      "epoch": 51.65,
      "learning_rate": 1.311378640776699e-05,
      "loss": 0.0927,
      "step": 132990
    },
    {
      "epoch": 51.65,
      "learning_rate": 1.311326860841424e-05,
      "loss": 0.0289,
      "step": 133000
    },
    {
      "epoch": 51.65,
      "learning_rate": 1.311275080906149e-05,
      "loss": 0.0015,
      "step": 133010
    },
    {
      "epoch": 51.66,
      "learning_rate": 1.3112233009708739e-05,
      "loss": 0.0375,
      "step": 133020
    },
    {
      "epoch": 51.66,
      "learning_rate": 1.3111715210355989e-05,
      "loss": 0.1232,
      "step": 133030
    },
    {
      "epoch": 51.67,
      "learning_rate": 1.3111197411003238e-05,
      "loss": 0.0712,
      "step": 133040
    },
    {
      "epoch": 51.67,
      "learning_rate": 1.3110679611650488e-05,
      "loss": 0.0764,
      "step": 133050
    },
    {
      "epoch": 51.67,
      "learning_rate": 1.3110161812297734e-05,
      "loss": 0.0538,
      "step": 133060
    },
    {
      "epoch": 51.68,
      "learning_rate": 1.3109644012944984e-05,
      "loss": 0.0732,
      "step": 133070
    },
    {
      "epoch": 51.68,
      "learning_rate": 1.3109126213592233e-05,
      "loss": 0.1258,
      "step": 133080
    },
    {
      "epoch": 51.69,
      "learning_rate": 1.3108608414239483e-05,
      "loss": 0.1584,
      "step": 133090
    },
    {
      "epoch": 51.69,
      "learning_rate": 1.3108090614886732e-05,
      "loss": 0.125,
      "step": 133100
    },
    {
      "epoch": 51.69,
      "learning_rate": 1.3107572815533982e-05,
      "loss": 0.1582,
      "step": 133110
    },
    {
      "epoch": 51.7,
      "learning_rate": 1.3107055016181232e-05,
      "loss": 0.1028,
      "step": 133120
    },
    {
      "epoch": 51.7,
      "learning_rate": 1.3106537216828481e-05,
      "loss": 0.0389,
      "step": 133130
    },
    {
      "epoch": 51.7,
      "learning_rate": 1.3106019417475728e-05,
      "loss": 0.2008,
      "step": 133140
    },
    {
      "epoch": 51.71,
      "learning_rate": 1.3105501618122977e-05,
      "loss": 0.0244,
      "step": 133150
    },
    {
      "epoch": 51.71,
      "learning_rate": 1.3104983818770227e-05,
      "loss": 0.0497,
      "step": 133160
    },
    {
      "epoch": 51.72,
      "learning_rate": 1.3104466019417476e-05,
      "loss": 0.0126,
      "step": 133170
    },
    {
      "epoch": 51.72,
      "learning_rate": 1.3103948220064726e-05,
      "loss": 0.0309,
      "step": 133180
    },
    {
      "epoch": 51.72,
      "learning_rate": 1.3103430420711976e-05,
      "loss": 0.0224,
      "step": 133190
    },
    {
      "epoch": 51.73,
      "learning_rate": 1.3102912621359225e-05,
      "loss": 0.0729,
      "step": 133200
    },
    {
      "epoch": 51.73,
      "learning_rate": 1.3102394822006475e-05,
      "loss": 0.075,
      "step": 133210
    },
    {
      "epoch": 51.74,
      "learning_rate": 1.3101877022653721e-05,
      "loss": 0.0716,
      "step": 133220
    },
    {
      "epoch": 51.74,
      "learning_rate": 1.310135922330097e-05,
      "loss": 0.2125,
      "step": 133230
    },
    {
      "epoch": 51.74,
      "learning_rate": 1.310084142394822e-05,
      "loss": 0.0269,
      "step": 133240
    },
    {
      "epoch": 51.75,
      "learning_rate": 1.310032362459547e-05,
      "loss": 0.0534,
      "step": 133250
    },
    {
      "epoch": 51.75,
      "learning_rate": 1.309980582524272e-05,
      "loss": 0.028,
      "step": 133260
    },
    {
      "epoch": 51.76,
      "learning_rate": 1.309928802588997e-05,
      "loss": 0.1072,
      "step": 133270
    },
    {
      "epoch": 51.76,
      "learning_rate": 1.3098770226537219e-05,
      "loss": 0.0014,
      "step": 133280
    },
    {
      "epoch": 51.76,
      "learning_rate": 1.3098252427184468e-05,
      "loss": 0.1145,
      "step": 133290
    },
    {
      "epoch": 51.77,
      "learning_rate": 1.3097734627831715e-05,
      "loss": 0.0625,
      "step": 133300
    },
    {
      "epoch": 51.77,
      "learning_rate": 1.3097216828478964e-05,
      "loss": 0.1866,
      "step": 133310
    },
    {
      "epoch": 51.77,
      "learning_rate": 1.3096699029126214e-05,
      "loss": 0.0403,
      "step": 133320
    },
    {
      "epoch": 51.78,
      "learning_rate": 1.3096181229773464e-05,
      "loss": 0.0692,
      "step": 133330
    },
    {
      "epoch": 51.78,
      "learning_rate": 1.3095663430420713e-05,
      "loss": 0.0954,
      "step": 133340
    },
    {
      "epoch": 51.79,
      "learning_rate": 1.3095145631067963e-05,
      "loss": 0.1215,
      "step": 133350
    },
    {
      "epoch": 51.79,
      "learning_rate": 1.3094627831715212e-05,
      "loss": 0.0034,
      "step": 133360
    },
    {
      "epoch": 51.79,
      "learning_rate": 1.3094110032362462e-05,
      "loss": 0.0281,
      "step": 133370
    },
    {
      "epoch": 51.8,
      "learning_rate": 1.3093592233009708e-05,
      "loss": 0.0149,
      "step": 133380
    },
    {
      "epoch": 51.8,
      "learning_rate": 1.3093074433656958e-05,
      "loss": 0.0438,
      "step": 133390
    },
    {
      "epoch": 51.81,
      "learning_rate": 1.3092556634304207e-05,
      "loss": 0.0329,
      "step": 133400
    },
    {
      "epoch": 51.81,
      "learning_rate": 1.3092038834951457e-05,
      "loss": 0.0221,
      "step": 133410
    },
    {
      "epoch": 51.81,
      "learning_rate": 1.3091521035598707e-05,
      "loss": 0.0403,
      "step": 133420
    },
    {
      "epoch": 51.82,
      "learning_rate": 1.3091003236245956e-05,
      "loss": 0.1138,
      "step": 133430
    },
    {
      "epoch": 51.82,
      "learning_rate": 1.3090485436893206e-05,
      "loss": 0.1323,
      "step": 133440
    },
    {
      "epoch": 51.83,
      "learning_rate": 1.3089967637540456e-05,
      "loss": 0.0286,
      "step": 133450
    },
    {
      "epoch": 51.83,
      "learning_rate": 1.3089449838187705e-05,
      "loss": 0.244,
      "step": 133460
    },
    {
      "epoch": 51.83,
      "learning_rate": 1.3088932038834951e-05,
      "loss": 0.0514,
      "step": 133470
    },
    {
      "epoch": 51.84,
      "learning_rate": 1.3088414239482201e-05,
      "loss": 0.081,
      "step": 133480
    },
    {
      "epoch": 51.84,
      "learning_rate": 1.308789644012945e-05,
      "loss": 0.0937,
      "step": 133490
    },
    {
      "epoch": 51.84,
      "learning_rate": 1.30873786407767e-05,
      "loss": 0.0447,
      "step": 133500
    },
    {
      "epoch": 51.85,
      "learning_rate": 1.308686084142395e-05,
      "loss": 0.0224,
      "step": 133510
    },
    {
      "epoch": 51.85,
      "learning_rate": 1.30863430420712e-05,
      "loss": 0.072,
      "step": 133520
    },
    {
      "epoch": 51.86,
      "learning_rate": 1.3085825242718449e-05,
      "loss": 0.0106,
      "step": 133530
    },
    {
      "epoch": 51.86,
      "learning_rate": 1.3085307443365699e-05,
      "loss": 0.1252,
      "step": 133540
    },
    {
      "epoch": 51.86,
      "learning_rate": 1.3084789644012945e-05,
      "loss": 0.0642,
      "step": 133550
    },
    {
      "epoch": 51.87,
      "learning_rate": 1.3084271844660195e-05,
      "loss": 0.0136,
      "step": 133560
    },
    {
      "epoch": 51.87,
      "learning_rate": 1.3083754045307444e-05,
      "loss": 0.0917,
      "step": 133570
    },
    {
      "epoch": 51.88,
      "learning_rate": 1.3083236245954694e-05,
      "loss": 0.2719,
      "step": 133580
    },
    {
      "epoch": 51.88,
      "learning_rate": 1.3082718446601943e-05,
      "loss": 0.1144,
      "step": 133590
    },
    {
      "epoch": 51.88,
      "learning_rate": 1.3082200647249193e-05,
      "loss": 0.2071,
      "step": 133600
    },
    {
      "epoch": 51.89,
      "learning_rate": 1.3081682847896443e-05,
      "loss": 0.0344,
      "step": 133610
    },
    {
      "epoch": 51.89,
      "learning_rate": 1.308116504854369e-05,
      "loss": 0.1053,
      "step": 133620
    },
    {
      "epoch": 51.9,
      "learning_rate": 1.3080647249190939e-05,
      "loss": 0.0284,
      "step": 133630
    },
    {
      "epoch": 51.9,
      "learning_rate": 1.3080129449838188e-05,
      "loss": 0.122,
      "step": 133640
    },
    {
      "epoch": 51.9,
      "learning_rate": 1.3079611650485438e-05,
      "loss": 0.0847,
      "step": 133650
    },
    {
      "epoch": 51.91,
      "learning_rate": 1.3079093851132687e-05,
      "loss": 0.1117,
      "step": 133660
    },
    {
      "epoch": 51.91,
      "learning_rate": 1.3078576051779937e-05,
      "loss": 0.0894,
      "step": 133670
    },
    {
      "epoch": 51.91,
      "learning_rate": 1.3078058252427187e-05,
      "loss": 0.0666,
      "step": 133680
    },
    {
      "epoch": 51.92,
      "learning_rate": 1.3077540453074436e-05,
      "loss": 0.0029,
      "step": 133690
    },
    {
      "epoch": 51.92,
      "learning_rate": 1.3077022653721684e-05,
      "loss": 0.048,
      "step": 133700
    },
    {
      "epoch": 51.93,
      "learning_rate": 1.3076504854368932e-05,
      "loss": 0.0152,
      "step": 133710
    },
    {
      "epoch": 51.93,
      "learning_rate": 1.3075987055016182e-05,
      "loss": 0.1696,
      "step": 133720
    },
    {
      "epoch": 51.93,
      "learning_rate": 1.3075469255663431e-05,
      "loss": 0.2976,
      "step": 133730
    },
    {
      "epoch": 51.94,
      "learning_rate": 1.3074951456310681e-05,
      "loss": 0.0914,
      "step": 133740
    },
    {
      "epoch": 51.94,
      "learning_rate": 1.307443365695793e-05,
      "loss": 0.1302,
      "step": 133750
    },
    {
      "epoch": 51.95,
      "learning_rate": 1.307391585760518e-05,
      "loss": 0.0947,
      "step": 133760
    },
    {
      "epoch": 51.95,
      "learning_rate": 1.3073398058252428e-05,
      "loss": 0.1116,
      "step": 133770
    },
    {
      "epoch": 51.95,
      "learning_rate": 1.3072880258899678e-05,
      "loss": 0.0864,
      "step": 133780
    },
    {
      "epoch": 51.96,
      "learning_rate": 1.3072362459546926e-05,
      "loss": 0.1238,
      "step": 133790
    },
    {
      "epoch": 51.96,
      "learning_rate": 1.3071844660194175e-05,
      "loss": 0.1392,
      "step": 133800
    },
    {
      "epoch": 51.97,
      "learning_rate": 1.3071326860841425e-05,
      "loss": 0.0167,
      "step": 133810
    },
    {
      "epoch": 51.97,
      "learning_rate": 1.3070809061488674e-05,
      "loss": 0.1485,
      "step": 133820
    },
    {
      "epoch": 51.97,
      "learning_rate": 1.3070291262135924e-05,
      "loss": 0.0613,
      "step": 133830
    },
    {
      "epoch": 51.98,
      "learning_rate": 1.3069773462783174e-05,
      "loss": 0.0806,
      "step": 133840
    },
    {
      "epoch": 51.98,
      "learning_rate": 1.3069255663430422e-05,
      "loss": 0.0922,
      "step": 133850
    },
    {
      "epoch": 51.98,
      "learning_rate": 1.3068737864077671e-05,
      "loss": 0.3644,
      "step": 133860
    },
    {
      "epoch": 51.99,
      "learning_rate": 1.306822006472492e-05,
      "loss": 0.1268,
      "step": 133870
    },
    {
      "epoch": 51.99,
      "learning_rate": 1.3067702265372169e-05,
      "loss": 0.1273,
      "step": 133880
    },
    {
      "epoch": 52.0,
      "learning_rate": 1.3067184466019418e-05,
      "loss": 0.0745,
      "step": 133890
    },
    {
      "epoch": 52.0,
      "learning_rate": 1.3066666666666668e-05,
      "loss": 0.1039,
      "step": 133900
    },
    {
      "epoch": 52.0,
      "eval_accuracy": 0.951856946354883,
      "eval_loss": 0.31140270829200745,
      "eval_runtime": 8.2636,
      "eval_samples_per_second": 439.878,
      "eval_steps_per_second": 55.06,
      "step": 133900
    },
    {
      "epoch": 52.0,
      "learning_rate": 1.3066148867313918e-05,
      "loss": 0.0709,
      "step": 133910
    },
    {
      "epoch": 52.01,
      "learning_rate": 1.3065631067961166e-05,
      "loss": 0.1453,
      "step": 133920
    },
    {
      "epoch": 52.01,
      "learning_rate": 1.3065113268608415e-05,
      "loss": 0.0447,
      "step": 133930
    },
    {
      "epoch": 52.02,
      "learning_rate": 1.3064595469255665e-05,
      "loss": 0.0609,
      "step": 133940
    },
    {
      "epoch": 52.02,
      "learning_rate": 1.3064077669902913e-05,
      "loss": 0.061,
      "step": 133950
    },
    {
      "epoch": 52.02,
      "learning_rate": 1.3063559870550162e-05,
      "loss": 0.1363,
      "step": 133960
    },
    {
      "epoch": 52.03,
      "learning_rate": 1.3063042071197412e-05,
      "loss": 0.1477,
      "step": 133970
    },
    {
      "epoch": 52.03,
      "learning_rate": 1.3062524271844662e-05,
      "loss": 0.0627,
      "step": 133980
    },
    {
      "epoch": 52.03,
      "learning_rate": 1.3062006472491911e-05,
      "loss": 0.1729,
      "step": 133990
    },
    {
      "epoch": 52.04,
      "learning_rate": 1.3061488673139159e-05,
      "loss": 0.042,
      "step": 134000
    },
    {
      "epoch": 52.04,
      "learning_rate": 1.3060970873786409e-05,
      "loss": 0.0873,
      "step": 134010
    },
    {
      "epoch": 52.05,
      "learning_rate": 1.3060453074433658e-05,
      "loss": 0.0053,
      "step": 134020
    },
    {
      "epoch": 52.05,
      "learning_rate": 1.3059935275080908e-05,
      "loss": 0.0174,
      "step": 134030
    },
    {
      "epoch": 52.05,
      "learning_rate": 1.3059417475728156e-05,
      "loss": 0.2504,
      "step": 134040
    },
    {
      "epoch": 52.06,
      "learning_rate": 1.3058899676375406e-05,
      "loss": 0.1361,
      "step": 134050
    },
    {
      "epoch": 52.06,
      "learning_rate": 1.3058381877022655e-05,
      "loss": 0.029,
      "step": 134060
    },
    {
      "epoch": 52.07,
      "learning_rate": 1.3057864077669905e-05,
      "loss": 0.0981,
      "step": 134070
    },
    {
      "epoch": 52.07,
      "learning_rate": 1.3057346278317153e-05,
      "loss": 0.0981,
      "step": 134080
    },
    {
      "epoch": 52.07,
      "learning_rate": 1.3056828478964402e-05,
      "loss": 0.0369,
      "step": 134090
    },
    {
      "epoch": 52.08,
      "learning_rate": 1.3056310679611652e-05,
      "loss": 0.0695,
      "step": 134100
    },
    {
      "epoch": 52.08,
      "learning_rate": 1.3055792880258902e-05,
      "loss": 0.1289,
      "step": 134110
    },
    {
      "epoch": 52.09,
      "learning_rate": 1.305527508090615e-05,
      "loss": 0.1293,
      "step": 134120
    },
    {
      "epoch": 52.09,
      "learning_rate": 1.3054757281553399e-05,
      "loss": 0.0436,
      "step": 134130
    },
    {
      "epoch": 52.09,
      "learning_rate": 1.3054239482200649e-05,
      "loss": 0.0321,
      "step": 134140
    },
    {
      "epoch": 52.1,
      "learning_rate": 1.3053721682847897e-05,
      "loss": 0.0723,
      "step": 134150
    },
    {
      "epoch": 52.1,
      "learning_rate": 1.3053203883495146e-05,
      "loss": 0.0005,
      "step": 134160
    },
    {
      "epoch": 52.1,
      "learning_rate": 1.3052686084142396e-05,
      "loss": 0.0055,
      "step": 134170
    },
    {
      "epoch": 52.11,
      "learning_rate": 1.3052168284789645e-05,
      "loss": 0.0259,
      "step": 134180
    },
    {
      "epoch": 52.11,
      "learning_rate": 1.3051650485436895e-05,
      "loss": 0.0629,
      "step": 134190
    },
    {
      "epoch": 52.12,
      "learning_rate": 1.3051132686084143e-05,
      "loss": 0.1509,
      "step": 134200
    },
    {
      "epoch": 52.12,
      "learning_rate": 1.3050614886731393e-05,
      "loss": 0.1206,
      "step": 134210
    },
    {
      "epoch": 52.12,
      "learning_rate": 1.3050097087378642e-05,
      "loss": 0.0348,
      "step": 134220
    },
    {
      "epoch": 52.13,
      "learning_rate": 1.304957928802589e-05,
      "loss": 0.0726,
      "step": 134230
    },
    {
      "epoch": 52.13,
      "learning_rate": 1.304906148867314e-05,
      "loss": 0.0638,
      "step": 134240
    },
    {
      "epoch": 52.14,
      "learning_rate": 1.304854368932039e-05,
      "loss": 0.0088,
      "step": 134250
    },
    {
      "epoch": 52.14,
      "learning_rate": 1.3048025889967639e-05,
      "loss": 0.1524,
      "step": 134260
    },
    {
      "epoch": 52.14,
      "learning_rate": 1.3047508090614889e-05,
      "loss": 0.0724,
      "step": 134270
    },
    {
      "epoch": 52.15,
      "learning_rate": 1.3046990291262137e-05,
      "loss": 0.0572,
      "step": 134280
    },
    {
      "epoch": 52.15,
      "learning_rate": 1.3046472491909386e-05,
      "loss": 0.0822,
      "step": 134290
    },
    {
      "epoch": 52.16,
      "learning_rate": 1.3045954692556634e-05,
      "loss": 0.1829,
      "step": 134300
    },
    {
      "epoch": 52.16,
      "learning_rate": 1.3045436893203884e-05,
      "loss": 0.1707,
      "step": 134310
    },
    {
      "epoch": 52.16,
      "learning_rate": 1.3044919093851133e-05,
      "loss": 0.0641,
      "step": 134320
    },
    {
      "epoch": 52.17,
      "learning_rate": 1.3044401294498383e-05,
      "loss": 0.0833,
      "step": 134330
    },
    {
      "epoch": 52.17,
      "learning_rate": 1.3043883495145633e-05,
      "loss": 0.1369,
      "step": 134340
    },
    {
      "epoch": 52.17,
      "learning_rate": 1.3043365695792882e-05,
      "loss": 0.0254,
      "step": 134350
    },
    {
      "epoch": 52.18,
      "learning_rate": 1.304284789644013e-05,
      "loss": 0.1059,
      "step": 134360
    },
    {
      "epoch": 52.18,
      "learning_rate": 1.304233009708738e-05,
      "loss": 0.1016,
      "step": 134370
    },
    {
      "epoch": 52.19,
      "learning_rate": 1.3041812297734628e-05,
      "loss": 0.0109,
      "step": 134380
    },
    {
      "epoch": 52.19,
      "learning_rate": 1.3041294498381877e-05,
      "loss": 0.0261,
      "step": 134390
    },
    {
      "epoch": 52.19,
      "learning_rate": 1.3040776699029127e-05,
      "loss": 0.0029,
      "step": 134400
    },
    {
      "epoch": 52.2,
      "learning_rate": 1.3040258899676377e-05,
      "loss": 0.1253,
      "step": 134410
    },
    {
      "epoch": 52.2,
      "learning_rate": 1.3039741100323626e-05,
      "loss": 0.0159,
      "step": 134420
    },
    {
      "epoch": 52.21,
      "learning_rate": 1.3039223300970876e-05,
      "loss": 0.0919,
      "step": 134430
    },
    {
      "epoch": 52.21,
      "learning_rate": 1.3038705501618124e-05,
      "loss": 0.0694,
      "step": 134440
    },
    {
      "epoch": 52.21,
      "learning_rate": 1.3038187702265372e-05,
      "loss": 0.0674,
      "step": 134450
    },
    {
      "epoch": 52.22,
      "learning_rate": 1.3037669902912621e-05,
      "loss": 0.0359,
      "step": 134460
    },
    {
      "epoch": 52.22,
      "learning_rate": 1.3037152103559871e-05,
      "loss": 0.109,
      "step": 134470
    },
    {
      "epoch": 52.23,
      "learning_rate": 1.303663430420712e-05,
      "loss": 0.0058,
      "step": 134480
    },
    {
      "epoch": 52.23,
      "learning_rate": 1.303611650485437e-05,
      "loss": 0.1067,
      "step": 134490
    },
    {
      "epoch": 52.23,
      "learning_rate": 1.303559870550162e-05,
      "loss": 0.0375,
      "step": 134500
    },
    {
      "epoch": 52.24,
      "learning_rate": 1.303508090614887e-05,
      "loss": 0.0849,
      "step": 134510
    },
    {
      "epoch": 52.24,
      "learning_rate": 1.3034563106796117e-05,
      "loss": 0.0449,
      "step": 134520
    },
    {
      "epoch": 52.24,
      "learning_rate": 1.3034045307443365e-05,
      "loss": 0.0362,
      "step": 134530
    },
    {
      "epoch": 52.25,
      "learning_rate": 1.3033527508090615e-05,
      "loss": 0.0151,
      "step": 134540
    },
    {
      "epoch": 52.25,
      "learning_rate": 1.3033009708737864e-05,
      "loss": 0.1025,
      "step": 134550
    },
    {
      "epoch": 52.26,
      "learning_rate": 1.3032491909385114e-05,
      "loss": 0.0572,
      "step": 134560
    },
    {
      "epoch": 52.26,
      "learning_rate": 1.3031974110032364e-05,
      "loss": 0.0565,
      "step": 134570
    },
    {
      "epoch": 52.26,
      "learning_rate": 1.3031456310679613e-05,
      "loss": 0.1472,
      "step": 134580
    },
    {
      "epoch": 52.27,
      "learning_rate": 1.3030938511326863e-05,
      "loss": 0.0674,
      "step": 134590
    },
    {
      "epoch": 52.27,
      "learning_rate": 1.3030420711974113e-05,
      "loss": 0.0268,
      "step": 134600
    },
    {
      "epoch": 52.28,
      "learning_rate": 1.3029902912621359e-05,
      "loss": 0.0342,
      "step": 134610
    },
    {
      "epoch": 52.28,
      "learning_rate": 1.3029385113268608e-05,
      "loss": 0.0858,
      "step": 134620
    },
    {
      "epoch": 52.28,
      "learning_rate": 1.3028867313915858e-05,
      "loss": 0.1257,
      "step": 134630
    },
    {
      "epoch": 52.29,
      "learning_rate": 1.3028349514563108e-05,
      "loss": 0.0227,
      "step": 134640
    },
    {
      "epoch": 52.29,
      "learning_rate": 1.3027831715210357e-05,
      "loss": 0.0486,
      "step": 134650
    },
    {
      "epoch": 52.3,
      "learning_rate": 1.3027313915857607e-05,
      "loss": 0.1537,
      "step": 134660
    },
    {
      "epoch": 52.3,
      "learning_rate": 1.3026796116504856e-05,
      "loss": 0.0272,
      "step": 134670
    },
    {
      "epoch": 52.3,
      "learning_rate": 1.3026278317152106e-05,
      "loss": 0.0683,
      "step": 134680
    },
    {
      "epoch": 52.31,
      "learning_rate": 1.3025760517799352e-05,
      "loss": 0.0204,
      "step": 134690
    },
    {
      "epoch": 52.31,
      "learning_rate": 1.3025242718446602e-05,
      "loss": 0.0561,
      "step": 134700
    },
    {
      "epoch": 52.31,
      "learning_rate": 1.3024724919093852e-05,
      "loss": 0.135,
      "step": 134710
    },
    {
      "epoch": 52.32,
      "learning_rate": 1.3024207119741101e-05,
      "loss": 0.045,
      "step": 134720
    },
    {
      "epoch": 52.32,
      "learning_rate": 1.302368932038835e-05,
      "loss": 0.037,
      "step": 134730
    },
    {
      "epoch": 52.33,
      "learning_rate": 1.30231715210356e-05,
      "loss": 0.183,
      "step": 134740
    },
    {
      "epoch": 52.33,
      "learning_rate": 1.302265372168285e-05,
      "loss": 0.048,
      "step": 134750
    },
    {
      "epoch": 52.33,
      "learning_rate": 1.30221359223301e-05,
      "loss": 0.1867,
      "step": 134760
    },
    {
      "epoch": 52.34,
      "learning_rate": 1.3021618122977346e-05,
      "loss": 0.0485,
      "step": 134770
    },
    {
      "epoch": 52.34,
      "learning_rate": 1.3021100323624595e-05,
      "loss": 0.0764,
      "step": 134780
    },
    {
      "epoch": 52.35,
      "learning_rate": 1.3020582524271845e-05,
      "loss": 0.0285,
      "step": 134790
    },
    {
      "epoch": 52.35,
      "learning_rate": 1.3020064724919095e-05,
      "loss": 0.0161,
      "step": 134800
    },
    {
      "epoch": 52.35,
      "learning_rate": 1.3019546925566344e-05,
      "loss": 0.0769,
      "step": 134810
    },
    {
      "epoch": 52.36,
      "learning_rate": 1.3019029126213594e-05,
      "loss": 0.0502,
      "step": 134820
    },
    {
      "epoch": 52.36,
      "learning_rate": 1.3018511326860844e-05,
      "loss": 0.0967,
      "step": 134830
    },
    {
      "epoch": 52.37,
      "learning_rate": 1.3017993527508093e-05,
      "loss": 0.085,
      "step": 134840
    },
    {
      "epoch": 52.37,
      "learning_rate": 1.301747572815534e-05,
      "loss": 0.0798,
      "step": 134850
    },
    {
      "epoch": 52.37,
      "learning_rate": 1.3016957928802589e-05,
      "loss": 0.0719,
      "step": 134860
    },
    {
      "epoch": 52.38,
      "learning_rate": 1.3016440129449839e-05,
      "loss": 0.1229,
      "step": 134870
    },
    {
      "epoch": 52.38,
      "learning_rate": 1.3015922330097088e-05,
      "loss": 0.0452,
      "step": 134880
    },
    {
      "epoch": 52.38,
      "learning_rate": 1.3015404530744338e-05,
      "loss": 0.0512,
      "step": 134890
    },
    {
      "epoch": 52.39,
      "learning_rate": 1.3014886731391588e-05,
      "loss": 0.0996,
      "step": 134900
    },
    {
      "epoch": 52.39,
      "learning_rate": 1.3014368932038837e-05,
      "loss": 0.0081,
      "step": 134910
    },
    {
      "epoch": 52.4,
      "learning_rate": 1.3013851132686087e-05,
      "loss": 0.1026,
      "step": 134920
    },
    {
      "epoch": 52.4,
      "learning_rate": 1.3013333333333333e-05,
      "loss": 0.0745,
      "step": 134930
    },
    {
      "epoch": 52.4,
      "learning_rate": 1.3012815533980583e-05,
      "loss": 0.112,
      "step": 134940
    },
    {
      "epoch": 52.41,
      "learning_rate": 1.3012297734627832e-05,
      "loss": 0.0348,
      "step": 134950
    },
    {
      "epoch": 52.41,
      "learning_rate": 1.3011779935275082e-05,
      "loss": 0.0235,
      "step": 134960
    },
    {
      "epoch": 52.42,
      "learning_rate": 1.3011262135922331e-05,
      "loss": 0.1054,
      "step": 134970
    },
    {
      "epoch": 52.42,
      "learning_rate": 1.3010744336569581e-05,
      "loss": 0.1154,
      "step": 134980
    },
    {
      "epoch": 52.42,
      "learning_rate": 1.301022653721683e-05,
      "loss": 0.0577,
      "step": 134990
    },
    {
      "epoch": 52.43,
      "learning_rate": 1.300970873786408e-05,
      "loss": 0.0688,
      "step": 135000
    },
    {
      "epoch": 52.43,
      "learning_rate": 1.3009190938511327e-05,
      "loss": 0.0457,
      "step": 135010
    },
    {
      "epoch": 52.43,
      "learning_rate": 1.3008673139158576e-05,
      "loss": 0.1648,
      "step": 135020
    },
    {
      "epoch": 52.44,
      "learning_rate": 1.3008155339805826e-05,
      "loss": 0.0941,
      "step": 135030
    },
    {
      "epoch": 52.44,
      "learning_rate": 1.3007637540453075e-05,
      "loss": 0.0871,
      "step": 135040
    },
    {
      "epoch": 52.45,
      "learning_rate": 1.3007119741100325e-05,
      "loss": 0.1435,
      "step": 135050
    },
    {
      "epoch": 52.45,
      "learning_rate": 1.3006601941747575e-05,
      "loss": 0.0563,
      "step": 135060
    },
    {
      "epoch": 52.45,
      "learning_rate": 1.3006084142394824e-05,
      "loss": 0.2036,
      "step": 135070
    },
    {
      "epoch": 52.46,
      "learning_rate": 1.3005566343042074e-05,
      "loss": 0.1116,
      "step": 135080
    },
    {
      "epoch": 52.46,
      "learning_rate": 1.300504854368932e-05,
      "loss": 0.0559,
      "step": 135090
    },
    {
      "epoch": 52.47,
      "learning_rate": 1.300453074433657e-05,
      "loss": 0.02,
      "step": 135100
    },
    {
      "epoch": 52.47,
      "learning_rate": 1.300401294498382e-05,
      "loss": 0.0059,
      "step": 135110
    },
    {
      "epoch": 52.47,
      "learning_rate": 1.3003495145631069e-05,
      "loss": 0.0094,
      "step": 135120
    },
    {
      "epoch": 52.48,
      "learning_rate": 1.3002977346278319e-05,
      "loss": 0.0581,
      "step": 135130
    },
    {
      "epoch": 52.48,
      "learning_rate": 1.3002459546925568e-05,
      "loss": 0.0288,
      "step": 135140
    },
    {
      "epoch": 52.49,
      "learning_rate": 1.3001941747572818e-05,
      "loss": 0.0712,
      "step": 135150
    },
    {
      "epoch": 52.49,
      "learning_rate": 1.3001423948220067e-05,
      "loss": 0.2684,
      "step": 135160
    },
    {
      "epoch": 52.49,
      "learning_rate": 1.3000906148867315e-05,
      "loss": 0.1627,
      "step": 135170
    },
    {
      "epoch": 52.5,
      "learning_rate": 1.3000388349514563e-05,
      "loss": 0.0744,
      "step": 135180
    },
    {
      "epoch": 52.5,
      "learning_rate": 1.2999870550161813e-05,
      "loss": 0.2232,
      "step": 135190
    },
    {
      "epoch": 52.5,
      "learning_rate": 1.2999352750809062e-05,
      "loss": 0.0296,
      "step": 135200
    },
    {
      "epoch": 52.51,
      "learning_rate": 1.2998834951456312e-05,
      "loss": 0.018,
      "step": 135210
    },
    {
      "epoch": 52.51,
      "learning_rate": 1.2998317152103562e-05,
      "loss": 0.1778,
      "step": 135220
    },
    {
      "epoch": 52.52,
      "learning_rate": 1.2997799352750811e-05,
      "loss": 0.0768,
      "step": 135230
    },
    {
      "epoch": 52.52,
      "learning_rate": 1.299728155339806e-05,
      "loss": 0.0269,
      "step": 135240
    },
    {
      "epoch": 52.52,
      "learning_rate": 1.2996763754045309e-05,
      "loss": 0.1571,
      "step": 135250
    },
    {
      "epoch": 52.53,
      "learning_rate": 1.2996245954692557e-05,
      "loss": 0.0436,
      "step": 135260
    },
    {
      "epoch": 52.53,
      "learning_rate": 1.2995728155339806e-05,
      "loss": 0.1329,
      "step": 135270
    },
    {
      "epoch": 52.54,
      "learning_rate": 1.2995210355987056e-05,
      "loss": 0.032,
      "step": 135280
    },
    {
      "epoch": 52.54,
      "learning_rate": 1.2994692556634306e-05,
      "loss": 0.0495,
      "step": 135290
    },
    {
      "epoch": 52.54,
      "learning_rate": 1.2994174757281555e-05,
      "loss": 0.0663,
      "step": 135300
    },
    {
      "epoch": 52.55,
      "learning_rate": 1.2993656957928805e-05,
      "loss": 0.0295,
      "step": 135310
    },
    {
      "epoch": 52.55,
      "learning_rate": 1.2993139158576053e-05,
      "loss": 0.0639,
      "step": 135320
    },
    {
      "epoch": 52.56,
      "learning_rate": 1.2992621359223302e-05,
      "loss": 0.0107,
      "step": 135330
    },
    {
      "epoch": 52.56,
      "learning_rate": 1.299210355987055e-05,
      "loss": 0.057,
      "step": 135340
    },
    {
      "epoch": 52.56,
      "learning_rate": 1.29915857605178e-05,
      "loss": 0.1533,
      "step": 135350
    },
    {
      "epoch": 52.57,
      "learning_rate": 1.299106796116505e-05,
      "loss": 0.0812,
      "step": 135360
    },
    {
      "epoch": 52.57,
      "learning_rate": 1.29905501618123e-05,
      "loss": 0.0565,
      "step": 135370
    },
    {
      "epoch": 52.57,
      "learning_rate": 1.2990032362459549e-05,
      "loss": 0.0389,
      "step": 135380
    },
    {
      "epoch": 52.58,
      "learning_rate": 1.2989514563106797e-05,
      "loss": 0.1392,
      "step": 135390
    },
    {
      "epoch": 52.58,
      "learning_rate": 1.2988996763754046e-05,
      "loss": 0.1665,
      "step": 135400
    },
    {
      "epoch": 52.59,
      "learning_rate": 1.2988478964401296e-05,
      "loss": 0.0269,
      "step": 135410
    },
    {
      "epoch": 52.59,
      "learning_rate": 1.2987961165048544e-05,
      "loss": 0.062,
      "step": 135420
    },
    {
      "epoch": 52.59,
      "learning_rate": 1.2987443365695794e-05,
      "loss": 0.1622,
      "step": 135430
    },
    {
      "epoch": 52.6,
      "learning_rate": 1.2986925566343043e-05,
      "loss": 0.071,
      "step": 135440
    },
    {
      "epoch": 52.6,
      "learning_rate": 1.2986407766990293e-05,
      "loss": 0.0471,
      "step": 135450
    },
    {
      "epoch": 52.61,
      "learning_rate": 1.2985889967637542e-05,
      "loss": 0.0835,
      "step": 135460
    },
    {
      "epoch": 52.61,
      "learning_rate": 1.298537216828479e-05,
      "loss": 0.0589,
      "step": 135470
    },
    {
      "epoch": 52.61,
      "learning_rate": 1.298485436893204e-05,
      "loss": 0.0195,
      "step": 135480
    },
    {
      "epoch": 52.62,
      "learning_rate": 1.298433656957929e-05,
      "loss": 0.0301,
      "step": 135490
    },
    {
      "epoch": 52.62,
      "learning_rate": 1.2983818770226537e-05,
      "loss": 0.1107,
      "step": 135500
    },
    {
      "epoch": 52.63,
      "learning_rate": 1.2983300970873787e-05,
      "loss": 0.1227,
      "step": 135510
    },
    {
      "epoch": 52.63,
      "learning_rate": 1.2982783171521037e-05,
      "loss": 0.0078,
      "step": 135520
    },
    {
      "epoch": 52.63,
      "learning_rate": 1.2982265372168286e-05,
      "loss": 0.1769,
      "step": 135530
    },
    {
      "epoch": 52.64,
      "learning_rate": 1.2981747572815534e-05,
      "loss": 0.0219,
      "step": 135540
    },
    {
      "epoch": 52.64,
      "learning_rate": 1.2981229773462784e-05,
      "loss": 0.1567,
      "step": 135550
    },
    {
      "epoch": 52.64,
      "learning_rate": 1.2980711974110033e-05,
      "loss": 0.0388,
      "step": 135560
    },
    {
      "epoch": 52.65,
      "learning_rate": 1.2980194174757283e-05,
      "loss": 0.1052,
      "step": 135570
    },
    {
      "epoch": 52.65,
      "learning_rate": 1.2979676375404531e-05,
      "loss": 0.0117,
      "step": 135580
    },
    {
      "epoch": 52.66,
      "learning_rate": 1.297915857605178e-05,
      "loss": 0.0276,
      "step": 135590
    },
    {
      "epoch": 52.66,
      "learning_rate": 1.297864077669903e-05,
      "loss": 0.0919,
      "step": 135600
    },
    {
      "epoch": 52.66,
      "learning_rate": 1.297812297734628e-05,
      "loss": 0.176,
      "step": 135610
    },
    {
      "epoch": 52.67,
      "learning_rate": 1.2977605177993528e-05,
      "loss": 0.0898,
      "step": 135620
    },
    {
      "epoch": 52.67,
      "learning_rate": 1.2977087378640777e-05,
      "loss": 0.0351,
      "step": 135630
    },
    {
      "epoch": 52.68,
      "learning_rate": 1.2976569579288027e-05,
      "loss": 0.0745,
      "step": 135640
    },
    {
      "epoch": 52.68,
      "learning_rate": 1.2976051779935277e-05,
      "loss": 0.0573,
      "step": 135650
    },
    {
      "epoch": 52.68,
      "learning_rate": 1.2975533980582525e-05,
      "loss": 0.1479,
      "step": 135660
    },
    {
      "epoch": 52.69,
      "learning_rate": 1.2975016181229774e-05,
      "loss": 0.0567,
      "step": 135670
    },
    {
      "epoch": 52.69,
      "learning_rate": 1.2974498381877024e-05,
      "loss": 0.0771,
      "step": 135680
    },
    {
      "epoch": 52.7,
      "learning_rate": 1.2973980582524273e-05,
      "loss": 0.0267,
      "step": 135690
    },
    {
      "epoch": 52.7,
      "learning_rate": 1.2973462783171521e-05,
      "loss": 0.0445,
      "step": 135700
    },
    {
      "epoch": 52.7,
      "learning_rate": 1.2972944983818771e-05,
      "loss": 0.1099,
      "step": 135710
    },
    {
      "epoch": 52.71,
      "learning_rate": 1.297242718446602e-05,
      "loss": 0.0609,
      "step": 135720
    },
    {
      "epoch": 52.71,
      "learning_rate": 1.297190938511327e-05,
      "loss": 0.1361,
      "step": 135730
    },
    {
      "epoch": 52.71,
      "learning_rate": 1.297139158576052e-05,
      "loss": 0.0444,
      "step": 135740
    },
    {
      "epoch": 52.72,
      "learning_rate": 1.2970873786407768e-05,
      "loss": 0.075,
      "step": 135750
    },
    {
      "epoch": 52.72,
      "learning_rate": 1.2970355987055017e-05,
      "loss": 0.0034,
      "step": 135760
    },
    {
      "epoch": 52.73,
      "learning_rate": 1.2969838187702265e-05,
      "loss": 0.0464,
      "step": 135770
    },
    {
      "epoch": 52.73,
      "learning_rate": 1.2969320388349515e-05,
      "loss": 0.0674,
      "step": 135780
    },
    {
      "epoch": 52.73,
      "learning_rate": 1.2968802588996765e-05,
      "loss": 0.0419,
      "step": 135790
    },
    {
      "epoch": 52.74,
      "learning_rate": 1.2968284789644014e-05,
      "loss": 0.0325,
      "step": 135800
    },
    {
      "epoch": 52.74,
      "learning_rate": 1.2967766990291264e-05,
      "loss": 0.0758,
      "step": 135810
    },
    {
      "epoch": 52.75,
      "learning_rate": 1.2967249190938513e-05,
      "loss": 0.198,
      "step": 135820
    },
    {
      "epoch": 52.75,
      "learning_rate": 1.2966731391585761e-05,
      "loss": 0.1668,
      "step": 135830
    },
    {
      "epoch": 52.75,
      "learning_rate": 1.2966213592233011e-05,
      "loss": 0.108,
      "step": 135840
    },
    {
      "epoch": 52.76,
      "learning_rate": 1.2965695792880259e-05,
      "loss": 0.058,
      "step": 135850
    },
    {
      "epoch": 52.76,
      "learning_rate": 1.2965177993527508e-05,
      "loss": 0.1207,
      "step": 135860
    },
    {
      "epoch": 52.77,
      "learning_rate": 1.2964660194174758e-05,
      "loss": 0.1166,
      "step": 135870
    },
    {
      "epoch": 52.77,
      "learning_rate": 1.2964142394822008e-05,
      "loss": 0.0348,
      "step": 135880
    },
    {
      "epoch": 52.77,
      "learning_rate": 1.2963624595469257e-05,
      "loss": 0.0258,
      "step": 135890
    },
    {
      "epoch": 52.78,
      "learning_rate": 1.2963106796116507e-05,
      "loss": 0.1234,
      "step": 135900
    },
    {
      "epoch": 52.78,
      "learning_rate": 1.2962588996763755e-05,
      "loss": 0.1146,
      "step": 135910
    },
    {
      "epoch": 52.78,
      "learning_rate": 1.2962071197411003e-05,
      "loss": 0.0231,
      "step": 135920
    },
    {
      "epoch": 52.79,
      "learning_rate": 1.2961553398058252e-05,
      "loss": 0.1982,
      "step": 135930
    },
    {
      "epoch": 52.79,
      "learning_rate": 1.2961035598705502e-05,
      "loss": 0.128,
      "step": 135940
    },
    {
      "epoch": 52.8,
      "learning_rate": 1.2960517799352752e-05,
      "loss": 0.0755,
      "step": 135950
    },
    {
      "epoch": 52.8,
      "learning_rate": 1.2960000000000001e-05,
      "loss": 0.0276,
      "step": 135960
    },
    {
      "epoch": 52.8,
      "learning_rate": 1.2959482200647251e-05,
      "loss": 0.1435,
      "step": 135970
    },
    {
      "epoch": 52.81,
      "learning_rate": 1.29589644012945e-05,
      "loss": 0.1442,
      "step": 135980
    },
    {
      "epoch": 52.81,
      "learning_rate": 1.2958446601941748e-05,
      "loss": 0.2039,
      "step": 135990
    },
    {
      "epoch": 52.82,
      "learning_rate": 1.2957928802588996e-05,
      "loss": 0.1069,
      "step": 136000
    },
    {
      "epoch": 52.82,
      "learning_rate": 1.2957411003236246e-05,
      "loss": 0.1514,
      "step": 136010
    },
    {
      "epoch": 52.82,
      "learning_rate": 1.2956893203883496e-05,
      "loss": 0.1401,
      "step": 136020
    },
    {
      "epoch": 52.83,
      "learning_rate": 1.2956375404530745e-05,
      "loss": 0.0493,
      "step": 136030
    },
    {
      "epoch": 52.83,
      "learning_rate": 1.2955857605177995e-05,
      "loss": 0.0407,
      "step": 136040
    },
    {
      "epoch": 52.83,
      "learning_rate": 1.2955339805825244e-05,
      "loss": 0.0254,
      "step": 136050
    },
    {
      "epoch": 52.84,
      "learning_rate": 1.2954822006472494e-05,
      "loss": 0.0193,
      "step": 136060
    },
    {
      "epoch": 52.84,
      "learning_rate": 1.295430420711974e-05,
      "loss": 0.0551,
      "step": 136070
    },
    {
      "epoch": 52.85,
      "learning_rate": 1.295378640776699e-05,
      "loss": 0.1208,
      "step": 136080
    },
    {
      "epoch": 52.85,
      "learning_rate": 1.295326860841424e-05,
      "loss": 0.0861,
      "step": 136090
    },
    {
      "epoch": 52.85,
      "learning_rate": 1.295275080906149e-05,
      "loss": 0.1217,
      "step": 136100
    },
    {
      "epoch": 52.86,
      "learning_rate": 1.2952233009708739e-05,
      "loss": 0.209,
      "step": 136110
    },
    {
      "epoch": 52.86,
      "learning_rate": 1.2951715210355988e-05,
      "loss": 0.1053,
      "step": 136120
    },
    {
      "epoch": 52.87,
      "learning_rate": 1.2951197411003238e-05,
      "loss": 0.0066,
      "step": 136130
    },
    {
      "epoch": 52.87,
      "learning_rate": 1.2950679611650488e-05,
      "loss": 0.0042,
      "step": 136140
    },
    {
      "epoch": 52.87,
      "learning_rate": 1.2950161812297734e-05,
      "loss": 0.0414,
      "step": 136150
    },
    {
      "epoch": 52.88,
      "learning_rate": 1.2949644012944983e-05,
      "loss": 0.0471,
      "step": 136160
    },
    {
      "epoch": 52.88,
      "learning_rate": 1.2949126213592233e-05,
      "loss": 0.0758,
      "step": 136170
    },
    {
      "epoch": 52.89,
      "learning_rate": 1.2948608414239483e-05,
      "loss": 0.0038,
      "step": 136180
    },
    {
      "epoch": 52.89,
      "learning_rate": 1.2948090614886732e-05,
      "loss": 0.0928,
      "step": 136190
    },
    {
      "epoch": 52.89,
      "learning_rate": 1.2947572815533982e-05,
      "loss": 0.1305,
      "step": 136200
    },
    {
      "epoch": 52.9,
      "learning_rate": 1.2947055016181232e-05,
      "loss": 0.115,
      "step": 136210
    },
    {
      "epoch": 52.9,
      "learning_rate": 1.2946537216828481e-05,
      "loss": 0.2102,
      "step": 136220
    },
    {
      "epoch": 52.9,
      "learning_rate": 1.294601941747573e-05,
      "loss": 0.0857,
      "step": 136230
    },
    {
      "epoch": 52.91,
      "learning_rate": 1.2945501618122977e-05,
      "loss": 0.0118,
      "step": 136240
    },
    {
      "epoch": 52.91,
      "learning_rate": 1.2944983818770227e-05,
      "loss": 0.0741,
      "step": 136250
    },
    {
      "epoch": 52.92,
      "learning_rate": 1.2944466019417476e-05,
      "loss": 0.0125,
      "step": 136260
    },
    {
      "epoch": 52.92,
      "learning_rate": 1.2943948220064726e-05,
      "loss": 0.0862,
      "step": 136270
    },
    {
      "epoch": 52.92,
      "learning_rate": 1.2943430420711976e-05,
      "loss": 0.0965,
      "step": 136280
    },
    {
      "epoch": 52.93,
      "learning_rate": 1.2942912621359225e-05,
      "loss": 0.0906,
      "step": 136290
    },
    {
      "epoch": 52.93,
      "learning_rate": 1.2942394822006475e-05,
      "loss": 0.0609,
      "step": 136300
    },
    {
      "epoch": 52.94,
      "learning_rate": 1.2941877022653724e-05,
      "loss": 0.1302,
      "step": 136310
    },
    {
      "epoch": 52.94,
      "learning_rate": 1.294135922330097e-05,
      "loss": 0.0822,
      "step": 136320
    },
    {
      "epoch": 52.94,
      "learning_rate": 1.294084142394822e-05,
      "loss": 0.0235,
      "step": 136330
    },
    {
      "epoch": 52.95,
      "learning_rate": 1.294032362459547e-05,
      "loss": 0.0097,
      "step": 136340
    },
    {
      "epoch": 52.95,
      "learning_rate": 1.293980582524272e-05,
      "loss": 0.1538,
      "step": 136350
    },
    {
      "epoch": 52.96,
      "learning_rate": 1.2939288025889969e-05,
      "loss": 0.1384,
      "step": 136360
    },
    {
      "epoch": 52.96,
      "learning_rate": 1.2938770226537219e-05,
      "loss": 0.1116,
      "step": 136370
    },
    {
      "epoch": 52.96,
      "learning_rate": 1.2938252427184468e-05,
      "loss": 0.0358,
      "step": 136380
    },
    {
      "epoch": 52.97,
      "learning_rate": 1.2937734627831718e-05,
      "loss": 0.211,
      "step": 136390
    },
    {
      "epoch": 52.97,
      "learning_rate": 1.2937216828478964e-05,
      "loss": 0.071,
      "step": 136400
    },
    {
      "epoch": 52.97,
      "learning_rate": 1.2936699029126214e-05,
      "loss": 0.048,
      "step": 136410
    },
    {
      "epoch": 52.98,
      "learning_rate": 1.2936181229773463e-05,
      "loss": 0.1028,
      "step": 136420
    },
    {
      "epoch": 52.98,
      "learning_rate": 1.2935663430420713e-05,
      "loss": 0.1078,
      "step": 136430
    },
    {
      "epoch": 52.99,
      "learning_rate": 1.2935145631067963e-05,
      "loss": 0.0413,
      "step": 136440
    },
    {
      "epoch": 52.99,
      "learning_rate": 1.2934627831715212e-05,
      "loss": 0.1997,
      "step": 136450
    },
    {
      "epoch": 52.99,
      "learning_rate": 1.2934110032362462e-05,
      "loss": 0.0489,
      "step": 136460
    },
    {
      "epoch": 53.0,
      "learning_rate": 1.2933592233009711e-05,
      "loss": 0.0527,
      "step": 136470
    },
    {
      "epoch": 53.0,
      "eval_accuracy": 0.9477303988995873,
      "eval_loss": 0.3252100348472595,
      "eval_runtime": 8.3172,
      "eval_samples_per_second": 437.044,
      "eval_steps_per_second": 54.706,
      "step": 136475
    },
    {
      "epoch": 53.0,
      "learning_rate": 1.2933074433656958e-05,
      "loss": 0.0636,
      "step": 136480
    },
    {
      "epoch": 53.01,
      "learning_rate": 1.2932556634304207e-05,
      "loss": 0.0243,
      "step": 136490
    },
    {
      "epoch": 53.01,
      "learning_rate": 1.2932038834951457e-05,
      "loss": 0.0908,
      "step": 136500
    },
    {
      "epoch": 53.01,
      "learning_rate": 1.2931521035598707e-05,
      "loss": 0.0429,
      "step": 136510
    },
    {
      "epoch": 53.02,
      "learning_rate": 1.2931003236245956e-05,
      "loss": 0.2102,
      "step": 136520
    },
    {
      "epoch": 53.02,
      "learning_rate": 1.2930485436893206e-05,
      "loss": 0.0568,
      "step": 136530
    },
    {
      "epoch": 53.03,
      "learning_rate": 1.2929967637540455e-05,
      "loss": 0.0271,
      "step": 136540
    },
    {
      "epoch": 53.03,
      "learning_rate": 1.2929449838187705e-05,
      "loss": 0.0194,
      "step": 136550
    },
    {
      "epoch": 53.03,
      "learning_rate": 1.2928932038834951e-05,
      "loss": 0.072,
      "step": 136560
    },
    {
      "epoch": 53.04,
      "learning_rate": 1.2928414239482201e-05,
      "loss": 0.155,
      "step": 136570
    },
    {
      "epoch": 53.04,
      "learning_rate": 1.292789644012945e-05,
      "loss": 0.0545,
      "step": 136580
    },
    {
      "epoch": 53.04,
      "learning_rate": 1.29273786407767e-05,
      "loss": 0.075,
      "step": 136590
    },
    {
      "epoch": 53.05,
      "learning_rate": 1.292686084142395e-05,
      "loss": 0.1236,
      "step": 136600
    },
    {
      "epoch": 53.05,
      "learning_rate": 1.29263430420712e-05,
      "loss": 0.066,
      "step": 136610
    },
    {
      "epoch": 53.06,
      "learning_rate": 1.2925825242718449e-05,
      "loss": 0.066,
      "step": 136620
    },
    {
      "epoch": 53.06,
      "learning_rate": 1.2925307443365699e-05,
      "loss": 0.0057,
      "step": 136630
    },
    {
      "epoch": 53.06,
      "learning_rate": 1.2924789644012945e-05,
      "loss": 0.0389,
      "step": 136640
    },
    {
      "epoch": 53.07,
      "learning_rate": 1.2924271844660194e-05,
      "loss": 0.0229,
      "step": 136650
    },
    {
      "epoch": 53.07,
      "learning_rate": 1.2923754045307444e-05,
      "loss": 0.0346,
      "step": 136660
    },
    {
      "epoch": 53.08,
      "learning_rate": 1.2923236245954694e-05,
      "loss": 0.112,
      "step": 136670
    },
    {
      "epoch": 53.08,
      "learning_rate": 1.2922718446601943e-05,
      "loss": 0.0854,
      "step": 136680
    },
    {
      "epoch": 53.08,
      "learning_rate": 1.2922200647249193e-05,
      "loss": 0.0109,
      "step": 136690
    },
    {
      "epoch": 53.09,
      "learning_rate": 1.2921682847896443e-05,
      "loss": 0.0381,
      "step": 136700
    },
    {
      "epoch": 53.09,
      "learning_rate": 1.292116504854369e-05,
      "loss": 0.2415,
      "step": 136710
    },
    {
      "epoch": 53.1,
      "learning_rate": 1.2920647249190938e-05,
      "loss": 0.0098,
      "step": 136720
    },
    {
      "epoch": 53.1,
      "learning_rate": 1.2920129449838188e-05,
      "loss": 0.0601,
      "step": 136730
    },
    {
      "epoch": 53.1,
      "learning_rate": 1.2919611650485438e-05,
      "loss": 0.0733,
      "step": 136740
    },
    {
      "epoch": 53.11,
      "learning_rate": 1.2919093851132687e-05,
      "loss": 0.0714,
      "step": 136750
    },
    {
      "epoch": 53.11,
      "learning_rate": 1.2918576051779937e-05,
      "loss": 0.0189,
      "step": 136760
    },
    {
      "epoch": 53.11,
      "learning_rate": 1.2918058252427186e-05,
      "loss": 0.1456,
      "step": 136770
    },
    {
      "epoch": 53.12,
      "learning_rate": 1.2917540453074436e-05,
      "loss": 0.058,
      "step": 136780
    },
    {
      "epoch": 53.12,
      "learning_rate": 1.2917022653721684e-05,
      "loss": 0.0703,
      "step": 136790
    },
    {
      "epoch": 53.13,
      "learning_rate": 1.2916504854368934e-05,
      "loss": 0.0353,
      "step": 136800
    },
    {
      "epoch": 53.13,
      "learning_rate": 1.2915987055016182e-05,
      "loss": 0.0254,
      "step": 136810
    },
    {
      "epoch": 53.13,
      "learning_rate": 1.2915469255663431e-05,
      "loss": 0.0158,
      "step": 136820
    },
    {
      "epoch": 53.14,
      "learning_rate": 1.291495145631068e-05,
      "loss": 0.0568,
      "step": 136830
    },
    {
      "epoch": 53.14,
      "learning_rate": 1.291443365695793e-05,
      "loss": 0.1673,
      "step": 136840
    },
    {
      "epoch": 53.15,
      "learning_rate": 1.291391585760518e-05,
      "loss": 0.0253,
      "step": 136850
    },
    {
      "epoch": 53.15,
      "learning_rate": 1.2913398058252428e-05,
      "loss": 0.1077,
      "step": 136860
    },
    {
      "epoch": 53.15,
      "learning_rate": 1.2912880258899678e-05,
      "loss": 0.0519,
      "step": 136870
    },
    {
      "epoch": 53.16,
      "learning_rate": 1.2912362459546927e-05,
      "loss": 0.1228,
      "step": 136880
    },
    {
      "epoch": 53.16,
      "learning_rate": 1.2911844660194175e-05,
      "loss": 0.0619,
      "step": 136890
    },
    {
      "epoch": 53.17,
      "learning_rate": 1.2911326860841425e-05,
      "loss": 0.0405,
      "step": 136900
    },
    {
      "epoch": 53.17,
      "learning_rate": 1.2910809061488674e-05,
      "loss": 0.0855,
      "step": 136910
    },
    {
      "epoch": 53.17,
      "learning_rate": 1.2910291262135924e-05,
      "loss": 0.1281,
      "step": 136920
    },
    {
      "epoch": 53.18,
      "learning_rate": 1.2909773462783174e-05,
      "loss": 0.0456,
      "step": 136930
    },
    {
      "epoch": 53.18,
      "learning_rate": 1.2909255663430421e-05,
      "loss": 0.0005,
      "step": 136940
    },
    {
      "epoch": 53.18,
      "learning_rate": 1.2908737864077671e-05,
      "loss": 0.0208,
      "step": 136950
    },
    {
      "epoch": 53.19,
      "learning_rate": 1.290822006472492e-05,
      "loss": 0.0622,
      "step": 136960
    },
    {
      "epoch": 53.19,
      "learning_rate": 1.2907702265372169e-05,
      "loss": 0.0312,
      "step": 136970
    },
    {
      "epoch": 53.2,
      "learning_rate": 1.2907184466019418e-05,
      "loss": 0.0224,
      "step": 136980
    },
    {
      "epoch": 53.2,
      "learning_rate": 1.2906666666666668e-05,
      "loss": 0.075,
      "step": 136990
    },
    {
      "epoch": 53.2,
      "learning_rate": 1.2906148867313918e-05,
      "loss": 0.0815,
      "step": 137000
    },
    {
      "epoch": 53.21,
      "learning_rate": 1.2905631067961165e-05,
      "loss": 0.1268,
      "step": 137010
    },
    {
      "epoch": 53.21,
      "learning_rate": 1.2905113268608415e-05,
      "loss": 0.0266,
      "step": 137020
    },
    {
      "epoch": 53.22,
      "learning_rate": 1.2904595469255665e-05,
      "loss": 0.066,
      "step": 137030
    },
    {
      "epoch": 53.22,
      "learning_rate": 1.2904077669902914e-05,
      "loss": 0.0169,
      "step": 137040
    },
    {
      "epoch": 53.22,
      "learning_rate": 1.2903559870550162e-05,
      "loss": 0.0717,
      "step": 137050
    },
    {
      "epoch": 53.23,
      "learning_rate": 1.2903042071197412e-05,
      "loss": 0.1607,
      "step": 137060
    },
    {
      "epoch": 53.23,
      "learning_rate": 1.2902524271844661e-05,
      "loss": 0.2586,
      "step": 137070
    },
    {
      "epoch": 53.23,
      "learning_rate": 1.2902006472491911e-05,
      "loss": 0.043,
      "step": 137080
    },
    {
      "epoch": 53.24,
      "learning_rate": 1.2901488673139159e-05,
      "loss": 0.1147,
      "step": 137090
    },
    {
      "epoch": 53.24,
      "learning_rate": 1.2900970873786409e-05,
      "loss": 0.0608,
      "step": 137100
    },
    {
      "epoch": 53.25,
      "learning_rate": 1.2900453074433658e-05,
      "loss": 0.2575,
      "step": 137110
    },
    {
      "epoch": 53.25,
      "learning_rate": 1.2899935275080908e-05,
      "loss": 0.0694,
      "step": 137120
    },
    {
      "epoch": 53.25,
      "learning_rate": 1.2899417475728156e-05,
      "loss": 0.0396,
      "step": 137130
    },
    {
      "epoch": 53.26,
      "learning_rate": 1.2898899676375405e-05,
      "loss": 0.1162,
      "step": 137140
    },
    {
      "epoch": 53.26,
      "learning_rate": 1.2898381877022655e-05,
      "loss": 0.1035,
      "step": 137150
    },
    {
      "epoch": 53.27,
      "learning_rate": 1.2897864077669905e-05,
      "loss": 0.2275,
      "step": 137160
    },
    {
      "epoch": 53.27,
      "learning_rate": 1.2897346278317153e-05,
      "loss": 0.0011,
      "step": 137170
    },
    {
      "epoch": 53.27,
      "learning_rate": 1.2896828478964402e-05,
      "loss": 0.0429,
      "step": 137180
    },
    {
      "epoch": 53.28,
      "learning_rate": 1.2896310679611652e-05,
      "loss": 0.0545,
      "step": 137190
    },
    {
      "epoch": 53.28,
      "learning_rate": 1.2895792880258901e-05,
      "loss": 0.1625,
      "step": 137200
    },
    {
      "epoch": 53.29,
      "learning_rate": 1.289527508090615e-05,
      "loss": 0.1022,
      "step": 137210
    },
    {
      "epoch": 53.29,
      "learning_rate": 1.2894757281553399e-05,
      "loss": 0.0015,
      "step": 137220
    },
    {
      "epoch": 53.29,
      "learning_rate": 1.2894239482200649e-05,
      "loss": 0.0211,
      "step": 137230
    },
    {
      "epoch": 53.3,
      "learning_rate": 1.2893721682847896e-05,
      "loss": 0.0294,
      "step": 137240
    },
    {
      "epoch": 53.3,
      "learning_rate": 1.2893203883495146e-05,
      "loss": 0.0945,
      "step": 137250
    },
    {
      "epoch": 53.3,
      "learning_rate": 1.2892686084142396e-05,
      "loss": 0.288,
      "step": 137260
    },
    {
      "epoch": 53.31,
      "learning_rate": 1.2892168284789645e-05,
      "loss": 0.1213,
      "step": 137270
    },
    {
      "epoch": 53.31,
      "learning_rate": 1.2891650485436895e-05,
      "loss": 0.0177,
      "step": 137280
    },
    {
      "epoch": 53.32,
      "learning_rate": 1.2891132686084143e-05,
      "loss": 0.0422,
      "step": 137290
    },
    {
      "epoch": 53.32,
      "learning_rate": 1.2890614886731392e-05,
      "loss": 0.1525,
      "step": 137300
    },
    {
      "epoch": 53.32,
      "learning_rate": 1.2890097087378642e-05,
      "loss": 0.1499,
      "step": 137310
    },
    {
      "epoch": 53.33,
      "learning_rate": 1.288957928802589e-05,
      "loss": 0.0256,
      "step": 137320
    },
    {
      "epoch": 53.33,
      "learning_rate": 1.288906148867314e-05,
      "loss": 0.1243,
      "step": 137330
    },
    {
      "epoch": 53.34,
      "learning_rate": 1.288854368932039e-05,
      "loss": 0.0513,
      "step": 137340
    },
    {
      "epoch": 53.34,
      "learning_rate": 1.2888025889967639e-05,
      "loss": 0.0813,
      "step": 137350
    },
    {
      "epoch": 53.34,
      "learning_rate": 1.2887508090614889e-05,
      "loss": 0.013,
      "step": 137360
    },
    {
      "epoch": 53.35,
      "learning_rate": 1.2886990291262138e-05,
      "loss": 0.1021,
      "step": 137370
    },
    {
      "epoch": 53.35,
      "learning_rate": 1.2886472491909386e-05,
      "loss": 0.1033,
      "step": 137380
    },
    {
      "epoch": 53.36,
      "learning_rate": 1.2885954692556634e-05,
      "loss": 0.0038,
      "step": 137390
    },
    {
      "epoch": 53.36,
      "learning_rate": 1.2885436893203884e-05,
      "loss": 0.0751,
      "step": 137400
    },
    {
      "epoch": 53.36,
      "learning_rate": 1.2884919093851133e-05,
      "loss": 0.0566,
      "step": 137410
    },
    {
      "epoch": 53.37,
      "learning_rate": 1.2884401294498383e-05,
      "loss": 0.0543,
      "step": 137420
    },
    {
      "epoch": 53.37,
      "learning_rate": 1.2883883495145632e-05,
      "loss": 0.0441,
      "step": 137430
    },
    {
      "epoch": 53.37,
      "learning_rate": 1.2883365695792882e-05,
      "loss": 0.1596,
      "step": 137440
    },
    {
      "epoch": 53.38,
      "learning_rate": 1.2882847896440132e-05,
      "loss": 0.0862,
      "step": 137450
    },
    {
      "epoch": 53.38,
      "learning_rate": 1.288233009708738e-05,
      "loss": 0.0169,
      "step": 137460
    },
    {
      "epoch": 53.39,
      "learning_rate": 1.2881812297734628e-05,
      "loss": 0.0776,
      "step": 137470
    },
    {
      "epoch": 53.39,
      "learning_rate": 1.2881294498381877e-05,
      "loss": 0.0004,
      "step": 137480
    },
    {
      "epoch": 53.39,
      "learning_rate": 1.2880776699029127e-05,
      "loss": 0.0188,
      "step": 137490
    },
    {
      "epoch": 53.4,
      "learning_rate": 1.2880258899676376e-05,
      "loss": 0.0332,
      "step": 137500
    },
    {
      "epoch": 53.4,
      "learning_rate": 1.2879741100323626e-05,
      "loss": 0.0672,
      "step": 137510
    },
    {
      "epoch": 53.41,
      "learning_rate": 1.2879223300970876e-05,
      "loss": 0.2091,
      "step": 137520
    },
    {
      "epoch": 53.41,
      "learning_rate": 1.2878705501618125e-05,
      "loss": 0.0898,
      "step": 137530
    },
    {
      "epoch": 53.41,
      "learning_rate": 1.2878187702265371e-05,
      "loss": 0.0942,
      "step": 137540
    },
    {
      "epoch": 53.42,
      "learning_rate": 1.2877669902912621e-05,
      "loss": 0.0846,
      "step": 137550
    },
    {
      "epoch": 53.42,
      "learning_rate": 1.287715210355987e-05,
      "loss": 0.0284,
      "step": 137560
    },
    {
      "epoch": 53.43,
      "learning_rate": 1.287663430420712e-05,
      "loss": 0.0101,
      "step": 137570
    },
    {
      "epoch": 53.43,
      "learning_rate": 1.287611650485437e-05,
      "loss": 0.1005,
      "step": 137580
    },
    {
      "epoch": 53.43,
      "learning_rate": 1.287559870550162e-05,
      "loss": 0.2141,
      "step": 137590
    },
    {
      "epoch": 53.44,
      "learning_rate": 1.287508090614887e-05,
      "loss": 0.0702,
      "step": 137600
    },
    {
      "epoch": 53.44,
      "learning_rate": 1.2874563106796119e-05,
      "loss": 0.0529,
      "step": 137610
    },
    {
      "epoch": 53.44,
      "learning_rate": 1.2874045307443365e-05,
      "loss": 0.0543,
      "step": 137620
    },
    {
      "epoch": 53.45,
      "learning_rate": 1.2873527508090615e-05,
      "loss": 0.014,
      "step": 137630
    },
    {
      "epoch": 53.45,
      "learning_rate": 1.2873009708737864e-05,
      "loss": 0.011,
      "step": 137640
    },
    {
      "epoch": 53.46,
      "learning_rate": 1.2872491909385114e-05,
      "loss": 0.0161,
      "step": 137650
    },
    {
      "epoch": 53.46,
      "learning_rate": 1.2871974110032363e-05,
      "loss": 0.0159,
      "step": 137660
    },
    {
      "epoch": 53.46,
      "learning_rate": 1.2871456310679613e-05,
      "loss": 0.2819,
      "step": 137670
    },
    {
      "epoch": 53.47,
      "learning_rate": 1.2870938511326863e-05,
      "loss": 0.1407,
      "step": 137680
    },
    {
      "epoch": 53.47,
      "learning_rate": 1.2870420711974112e-05,
      "loss": 0.207,
      "step": 137690
    },
    {
      "epoch": 53.48,
      "learning_rate": 1.2869902912621359e-05,
      "loss": 0.0357,
      "step": 137700
    },
    {
      "epoch": 53.48,
      "learning_rate": 1.2869385113268608e-05,
      "loss": 0.1306,
      "step": 137710
    },
    {
      "epoch": 53.48,
      "learning_rate": 1.2868867313915858e-05,
      "loss": 0.0915,
      "step": 137720
    },
    {
      "epoch": 53.49,
      "learning_rate": 1.2868349514563107e-05,
      "loss": 0.0052,
      "step": 137730
    },
    {
      "epoch": 53.49,
      "learning_rate": 1.2867831715210357e-05,
      "loss": 0.0564,
      "step": 137740
    },
    {
      "epoch": 53.5,
      "learning_rate": 1.2867313915857607e-05,
      "loss": 0.1943,
      "step": 137750
    },
    {
      "epoch": 53.5,
      "learning_rate": 1.2866796116504856e-05,
      "loss": 0.0716,
      "step": 137760
    },
    {
      "epoch": 53.5,
      "learning_rate": 1.2866278317152106e-05,
      "loss": 0.0589,
      "step": 137770
    },
    {
      "epoch": 53.51,
      "learning_rate": 1.2865760517799352e-05,
      "loss": 0.1383,
      "step": 137780
    },
    {
      "epoch": 53.51,
      "learning_rate": 1.2865242718446602e-05,
      "loss": 0.1479,
      "step": 137790
    },
    {
      "epoch": 53.51,
      "learning_rate": 1.2864724919093851e-05,
      "loss": 0.0332,
      "step": 137800
    },
    {
      "epoch": 53.52,
      "learning_rate": 1.2864207119741101e-05,
      "loss": 0.024,
      "step": 137810
    },
    {
      "epoch": 53.52,
      "learning_rate": 1.286368932038835e-05,
      "loss": 0.1534,
      "step": 137820
    },
    {
      "epoch": 53.53,
      "learning_rate": 1.28631715210356e-05,
      "loss": 0.1334,
      "step": 137830
    },
    {
      "epoch": 53.53,
      "learning_rate": 1.286265372168285e-05,
      "loss": 0.0825,
      "step": 137840
    },
    {
      "epoch": 53.53,
      "learning_rate": 1.28621359223301e-05,
      "loss": 0.0999,
      "step": 137850
    },
    {
      "epoch": 53.54,
      "learning_rate": 1.2861618122977346e-05,
      "loss": 0.1139,
      "step": 137860
    },
    {
      "epoch": 53.54,
      "learning_rate": 1.2861100323624595e-05,
      "loss": 0.3255,
      "step": 137870
    },
    {
      "epoch": 53.55,
      "learning_rate": 1.2860582524271845e-05,
      "loss": 0.1478,
      "step": 137880
    },
    {
      "epoch": 53.55,
      "learning_rate": 1.2860064724919095e-05,
      "loss": 0.0392,
      "step": 137890
    },
    {
      "epoch": 53.55,
      "learning_rate": 1.2859546925566344e-05,
      "loss": 0.0998,
      "step": 137900
    },
    {
      "epoch": 53.56,
      "learning_rate": 1.2859029126213594e-05,
      "loss": 0.0295,
      "step": 137910
    },
    {
      "epoch": 53.56,
      "learning_rate": 1.2858511326860843e-05,
      "loss": 0.1597,
      "step": 137920
    },
    {
      "epoch": 53.57,
      "learning_rate": 1.2857993527508093e-05,
      "loss": 0.0248,
      "step": 137930
    },
    {
      "epoch": 53.57,
      "learning_rate": 1.2857475728155343e-05,
      "loss": 0.0317,
      "step": 137940
    },
    {
      "epoch": 53.57,
      "learning_rate": 1.2856957928802589e-05,
      "loss": 0.0681,
      "step": 137950
    },
    {
      "epoch": 53.58,
      "learning_rate": 1.2856440129449838e-05,
      "loss": 0.0184,
      "step": 137960
    },
    {
      "epoch": 53.58,
      "learning_rate": 1.2855922330097088e-05,
      "loss": 0.0427,
      "step": 137970
    },
    {
      "epoch": 53.58,
      "learning_rate": 1.2855404530744338e-05,
      "loss": 0.1294,
      "step": 137980
    },
    {
      "epoch": 53.59,
      "learning_rate": 1.2854886731391587e-05,
      "loss": 0.1119,
      "step": 137990
    },
    {
      "epoch": 53.59,
      "learning_rate": 1.2854368932038837e-05,
      "loss": 0.0793,
      "step": 138000
    },
    {
      "epoch": 53.6,
      "learning_rate": 1.2853851132686087e-05,
      "loss": 0.0185,
      "step": 138010
    },
    {
      "epoch": 53.6,
      "learning_rate": 1.2853333333333336e-05,
      "loss": 0.037,
      "step": 138020
    },
    {
      "epoch": 53.6,
      "learning_rate": 1.2852815533980582e-05,
      "loss": 0.2302,
      "step": 138030
    },
    {
      "epoch": 53.61,
      "learning_rate": 1.2852297734627832e-05,
      "loss": 0.0784,
      "step": 138040
    },
    {
      "epoch": 53.61,
      "learning_rate": 1.2851779935275082e-05,
      "loss": 0.0632,
      "step": 138050
    },
    {
      "epoch": 53.62,
      "learning_rate": 1.2851262135922331e-05,
      "loss": 0.0689,
      "step": 138060
    },
    {
      "epoch": 53.62,
      "learning_rate": 1.2850744336569581e-05,
      "loss": 0.1287,
      "step": 138070
    },
    {
      "epoch": 53.62,
      "learning_rate": 1.285022653721683e-05,
      "loss": 0.0173,
      "step": 138080
    },
    {
      "epoch": 53.63,
      "learning_rate": 1.284970873786408e-05,
      "loss": 0.0499,
      "step": 138090
    },
    {
      "epoch": 53.63,
      "learning_rate": 1.284919093851133e-05,
      "loss": 0.1056,
      "step": 138100
    },
    {
      "epoch": 53.63,
      "learning_rate": 1.2848673139158576e-05,
      "loss": 0.0581,
      "step": 138110
    },
    {
      "epoch": 53.64,
      "learning_rate": 1.2848155339805826e-05,
      "loss": 0.0633,
      "step": 138120
    },
    {
      "epoch": 53.64,
      "learning_rate": 1.2847637540453075e-05,
      "loss": 0.0296,
      "step": 138130
    },
    {
      "epoch": 53.65,
      "learning_rate": 1.2847119741100325e-05,
      "loss": 0.2311,
      "step": 138140
    },
    {
      "epoch": 53.65,
      "learning_rate": 1.2846601941747574e-05,
      "loss": 0.0488,
      "step": 138150
    },
    {
      "epoch": 53.65,
      "learning_rate": 1.2846084142394824e-05,
      "loss": 0.0266,
      "step": 138160
    },
    {
      "epoch": 53.66,
      "learning_rate": 1.2845566343042074e-05,
      "loss": 0.0371,
      "step": 138170
    },
    {
      "epoch": 53.66,
      "learning_rate": 1.2845048543689322e-05,
      "loss": 0.1079,
      "step": 138180
    },
    {
      "epoch": 53.67,
      "learning_rate": 1.284453074433657e-05,
      "loss": 0.0381,
      "step": 138190
    },
    {
      "epoch": 53.67,
      "learning_rate": 1.284401294498382e-05,
      "loss": 0.0437,
      "step": 138200
    },
    {
      "epoch": 53.67,
      "learning_rate": 1.2843495145631069e-05,
      "loss": 0.1006,
      "step": 138210
    },
    {
      "epoch": 53.68,
      "learning_rate": 1.2842977346278318e-05,
      "loss": 0.0646,
      "step": 138220
    },
    {
      "epoch": 53.68,
      "learning_rate": 1.2842459546925568e-05,
      "loss": 0.1939,
      "step": 138230
    },
    {
      "epoch": 53.69,
      "learning_rate": 1.2841941747572818e-05,
      "loss": 0.2113,
      "step": 138240
    },
    {
      "epoch": 53.69,
      "learning_rate": 1.2841423948220067e-05,
      "loss": 0.0041,
      "step": 138250
    },
    {
      "epoch": 53.69,
      "learning_rate": 1.2840906148867315e-05,
      "loss": 0.0953,
      "step": 138260
    },
    {
      "epoch": 53.7,
      "learning_rate": 1.2840388349514563e-05,
      "loss": 0.085,
      "step": 138270
    },
    {
      "epoch": 53.7,
      "learning_rate": 1.2839870550161813e-05,
      "loss": 0.133,
      "step": 138280
    },
    {
      "epoch": 53.7,
      "learning_rate": 1.2839352750809062e-05,
      "loss": 0.1631,
      "step": 138290
    },
    {
      "epoch": 53.71,
      "learning_rate": 1.2838834951456312e-05,
      "loss": 0.1075,
      "step": 138300
    },
    {
      "epoch": 53.71,
      "learning_rate": 1.2838317152103562e-05,
      "loss": 0.0586,
      "step": 138310
    },
    {
      "epoch": 53.72,
      "learning_rate": 1.2837799352750811e-05,
      "loss": 0.1057,
      "step": 138320
    },
    {
      "epoch": 53.72,
      "learning_rate": 1.2837281553398059e-05,
      "loss": 0.0934,
      "step": 138330
    },
    {
      "epoch": 53.72,
      "learning_rate": 1.2836763754045309e-05,
      "loss": 0.1591,
      "step": 138340
    },
    {
      "epoch": 53.73,
      "learning_rate": 1.2836245954692557e-05,
      "loss": 0.0974,
      "step": 138350
    },
    {
      "epoch": 53.73,
      "learning_rate": 1.2835728155339806e-05,
      "loss": 0.0527,
      "step": 138360
    },
    {
      "epoch": 53.74,
      "learning_rate": 1.2835210355987056e-05,
      "loss": 0.0047,
      "step": 138370
    },
    {
      "epoch": 53.74,
      "learning_rate": 1.2834692556634306e-05,
      "loss": 0.1886,
      "step": 138380
    },
    {
      "epoch": 53.74,
      "learning_rate": 1.2834174757281555e-05,
      "loss": 0.0817,
      "step": 138390
    },
    {
      "epoch": 53.75,
      "learning_rate": 1.2833656957928805e-05,
      "loss": 0.0402,
      "step": 138400
    },
    {
      "epoch": 53.75,
      "learning_rate": 1.2833139158576053e-05,
      "loss": 0.0728,
      "step": 138410
    },
    {
      "epoch": 53.76,
      "learning_rate": 1.2832621359223302e-05,
      "loss": 0.0454,
      "step": 138420
    },
    {
      "epoch": 53.76,
      "learning_rate": 1.283210355987055e-05,
      "loss": 0.1984,
      "step": 138430
    },
    {
      "epoch": 53.76,
      "learning_rate": 1.28315857605178e-05,
      "loss": 0.112,
      "step": 138440
    },
    {
      "epoch": 53.77,
      "learning_rate": 1.283106796116505e-05,
      "loss": 0.0947,
      "step": 138450
    },
    {
      "epoch": 53.77,
      "learning_rate": 1.2830550161812299e-05,
      "loss": 0.2117,
      "step": 138460
    },
    {
      "epoch": 53.77,
      "learning_rate": 1.2830032362459549e-05,
      "loss": 0.0085,
      "step": 138470
    },
    {
      "epoch": 53.78,
      "learning_rate": 1.2829514563106797e-05,
      "loss": 0.1602,
      "step": 138480
    },
    {
      "epoch": 53.78,
      "learning_rate": 1.2828996763754046e-05,
      "loss": 0.0759,
      "step": 138490
    },
    {
      "epoch": 53.79,
      "learning_rate": 1.2828478964401296e-05,
      "loss": 0.0106,
      "step": 138500
    },
    {
      "epoch": 53.79,
      "learning_rate": 1.2827961165048545e-05,
      "loss": 0.0638,
      "step": 138510
    },
    {
      "epoch": 53.79,
      "learning_rate": 1.2827443365695793e-05,
      "loss": 0.0874,
      "step": 138520
    },
    {
      "epoch": 53.8,
      "learning_rate": 1.2826925566343043e-05,
      "loss": 0.1208,
      "step": 138530
    },
    {
      "epoch": 53.8,
      "learning_rate": 1.2826407766990293e-05,
      "loss": 0.038,
      "step": 138540
    },
    {
      "epoch": 53.81,
      "learning_rate": 1.2825889967637542e-05,
      "loss": 0.2173,
      "step": 138550
    },
    {
      "epoch": 53.81,
      "learning_rate": 1.282537216828479e-05,
      "loss": 0.1583,
      "step": 138560
    },
    {
      "epoch": 53.81,
      "learning_rate": 1.282485436893204e-05,
      "loss": 0.0147,
      "step": 138570
    },
    {
      "epoch": 53.82,
      "learning_rate": 1.282433656957929e-05,
      "loss": 0.0653,
      "step": 138580
    },
    {
      "epoch": 53.82,
      "learning_rate": 1.2823818770226539e-05,
      "loss": 0.0883,
      "step": 138590
    },
    {
      "epoch": 53.83,
      "learning_rate": 1.2823300970873787e-05,
      "loss": 0.0655,
      "step": 138600
    },
    {
      "epoch": 53.83,
      "learning_rate": 1.2822783171521037e-05,
      "loss": 0.0746,
      "step": 138610
    },
    {
      "epoch": 53.83,
      "learning_rate": 1.2822265372168286e-05,
      "loss": 0.0538,
      "step": 138620
    },
    {
      "epoch": 53.84,
      "learning_rate": 1.2821747572815536e-05,
      "loss": 0.0221,
      "step": 138630
    },
    {
      "epoch": 53.84,
      "learning_rate": 1.2821229773462784e-05,
      "loss": 0.0628,
      "step": 138640
    },
    {
      "epoch": 53.84,
      "learning_rate": 1.2820711974110033e-05,
      "loss": 0.1168,
      "step": 138650
    },
    {
      "epoch": 53.85,
      "learning_rate": 1.2820194174757283e-05,
      "loss": 0.0828,
      "step": 138660
    },
    {
      "epoch": 53.85,
      "learning_rate": 1.2819676375404533e-05,
      "loss": 0.2454,
      "step": 138670
    },
    {
      "epoch": 53.86,
      "learning_rate": 1.281915857605178e-05,
      "loss": 0.013,
      "step": 138680
    },
    {
      "epoch": 53.86,
      "learning_rate": 1.281864077669903e-05,
      "loss": 0.1341,
      "step": 138690
    },
    {
      "epoch": 53.86,
      "learning_rate": 1.281812297734628e-05,
      "loss": 0.1506,
      "step": 138700
    },
    {
      "epoch": 53.87,
      "learning_rate": 1.2817605177993528e-05,
      "loss": 0.0043,
      "step": 138710
    },
    {
      "epoch": 53.87,
      "learning_rate": 1.2817087378640777e-05,
      "loss": 0.0026,
      "step": 138720
    },
    {
      "epoch": 53.88,
      "learning_rate": 1.2816569579288027e-05,
      "loss": 0.1118,
      "step": 138730
    },
    {
      "epoch": 53.88,
      "learning_rate": 1.2816051779935277e-05,
      "loss": 0.1669,
      "step": 138740
    },
    {
      "epoch": 53.88,
      "learning_rate": 1.2815533980582526e-05,
      "loss": 0.0814,
      "step": 138750
    },
    {
      "epoch": 53.89,
      "learning_rate": 1.2815016181229774e-05,
      "loss": 0.1544,
      "step": 138760
    },
    {
      "epoch": 53.89,
      "learning_rate": 1.2814498381877024e-05,
      "loss": 0.0078,
      "step": 138770
    },
    {
      "epoch": 53.9,
      "learning_rate": 1.2813980582524273e-05,
      "loss": 0.0608,
      "step": 138780
    },
    {
      "epoch": 53.9,
      "learning_rate": 1.2813462783171521e-05,
      "loss": 0.1009,
      "step": 138790
    },
    {
      "epoch": 53.9,
      "learning_rate": 1.281294498381877e-05,
      "loss": 0.1198,
      "step": 138800
    },
    {
      "epoch": 53.91,
      "learning_rate": 1.281242718446602e-05,
      "loss": 0.058,
      "step": 138810
    },
    {
      "epoch": 53.91,
      "learning_rate": 1.281190938511327e-05,
      "loss": 0.0128,
      "step": 138820
    },
    {
      "epoch": 53.91,
      "learning_rate": 1.281139158576052e-05,
      "loss": 0.0508,
      "step": 138830
    },
    {
      "epoch": 53.92,
      "learning_rate": 1.2810873786407768e-05,
      "loss": 0.026,
      "step": 138840
    },
    {
      "epoch": 53.92,
      "learning_rate": 1.2810355987055017e-05,
      "loss": 0.1573,
      "step": 138850
    },
    {
      "epoch": 53.93,
      "learning_rate": 1.2809838187702265e-05,
      "loss": 0.0369,
      "step": 138860
    },
    {
      "epoch": 53.93,
      "learning_rate": 1.2809320388349515e-05,
      "loss": 0.0576,
      "step": 138870
    },
    {
      "epoch": 53.93,
      "learning_rate": 1.2808802588996764e-05,
      "loss": 0.0335,
      "step": 138880
    },
    {
      "epoch": 53.94,
      "learning_rate": 1.2808284789644014e-05,
      "loss": 0.101,
      "step": 138890
    },
    {
      "epoch": 53.94,
      "learning_rate": 1.2807766990291264e-05,
      "loss": 0.1526,
      "step": 138900
    },
    {
      "epoch": 53.95,
      "learning_rate": 1.2807249190938513e-05,
      "loss": 0.1037,
      "step": 138910
    },
    {
      "epoch": 53.95,
      "learning_rate": 1.2806731391585761e-05,
      "loss": 0.1148,
      "step": 138920
    },
    {
      "epoch": 53.95,
      "learning_rate": 1.280621359223301e-05,
      "loss": 0.1128,
      "step": 138930
    },
    {
      "epoch": 53.96,
      "learning_rate": 1.2805695792880259e-05,
      "loss": 0.0868,
      "step": 138940
    },
    {
      "epoch": 53.96,
      "learning_rate": 1.2805177993527508e-05,
      "loss": 0.0383,
      "step": 138950
    },
    {
      "epoch": 53.97,
      "learning_rate": 1.2804660194174758e-05,
      "loss": 0.0125,
      "step": 138960
    },
    {
      "epoch": 53.97,
      "learning_rate": 1.2804142394822008e-05,
      "loss": 0.1462,
      "step": 138970
    },
    {
      "epoch": 53.97,
      "learning_rate": 1.2803624595469257e-05,
      "loss": 0.1112,
      "step": 138980
    },
    {
      "epoch": 53.98,
      "learning_rate": 1.2803106796116507e-05,
      "loss": 0.0183,
      "step": 138990
    },
    {
      "epoch": 53.98,
      "learning_rate": 1.2802588996763755e-05,
      "loss": 0.0661,
      "step": 139000
    },
    {
      "epoch": 53.98,
      "learning_rate": 1.2802071197411003e-05,
      "loss": 0.0343,
      "step": 139010
    },
    {
      "epoch": 53.99,
      "learning_rate": 1.2801553398058252e-05,
      "loss": 0.1815,
      "step": 139020
    },
    {
      "epoch": 53.99,
      "learning_rate": 1.2801035598705502e-05,
      "loss": 0.0294,
      "step": 139030
    },
    {
      "epoch": 54.0,
      "learning_rate": 1.2800517799352751e-05,
      "loss": 0.0196,
      "step": 139040
    },
    {
      "epoch": 54.0,
      "learning_rate": 1.2800000000000001e-05,
      "loss": 0.0584,
      "step": 139050
    },
    {
      "epoch": 54.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.319997102022171,
      "eval_runtime": 8.2173,
      "eval_samples_per_second": 442.361,
      "eval_steps_per_second": 55.371,
      "step": 139050
    },
    {
      "epoch": 54.0,
      "learning_rate": 1.279948220064725e-05,
      "loss": 0.0565,
      "step": 139060
    },
    {
      "epoch": 54.01,
      "learning_rate": 1.27989644012945e-05,
      "loss": 0.0385,
      "step": 139070
    },
    {
      "epoch": 54.01,
      "learning_rate": 1.279844660194175e-05,
      "loss": 0.0364,
      "step": 139080
    },
    {
      "epoch": 54.02,
      "learning_rate": 1.2797928802588996e-05,
      "loss": 0.0796,
      "step": 139090
    },
    {
      "epoch": 54.02,
      "learning_rate": 1.2797411003236246e-05,
      "loss": 0.0429,
      "step": 139100
    },
    {
      "epoch": 54.02,
      "learning_rate": 1.2796893203883495e-05,
      "loss": 0.0148,
      "step": 139110
    },
    {
      "epoch": 54.03,
      "learning_rate": 1.2796375404530745e-05,
      "loss": 0.1017,
      "step": 139120
    },
    {
      "epoch": 54.03,
      "learning_rate": 1.2795857605177995e-05,
      "loss": 0.1885,
      "step": 139130
    },
    {
      "epoch": 54.03,
      "learning_rate": 1.2795339805825244e-05,
      "loss": 0.1088,
      "step": 139140
    },
    {
      "epoch": 54.04,
      "learning_rate": 1.2794822006472494e-05,
      "loss": 0.0837,
      "step": 139150
    },
    {
      "epoch": 54.04,
      "learning_rate": 1.2794304207119744e-05,
      "loss": 0.0096,
      "step": 139160
    },
    {
      "epoch": 54.05,
      "learning_rate": 1.279378640776699e-05,
      "loss": 0.0856,
      "step": 139170
    },
    {
      "epoch": 54.05,
      "learning_rate": 1.279326860841424e-05,
      "loss": 0.0407,
      "step": 139180
    },
    {
      "epoch": 54.05,
      "learning_rate": 1.2792750809061489e-05,
      "loss": 0.0196,
      "step": 139190
    },
    {
      "epoch": 54.06,
      "learning_rate": 1.2792233009708739e-05,
      "loss": 0.015,
      "step": 139200
    },
    {
      "epoch": 54.06,
      "learning_rate": 1.2791715210355988e-05,
      "loss": 0.0247,
      "step": 139210
    },
    {
      "epoch": 54.07,
      "learning_rate": 1.2791197411003238e-05,
      "loss": 0.0002,
      "step": 139220
    },
    {
      "epoch": 54.07,
      "learning_rate": 1.2790679611650487e-05,
      "loss": 0.0002,
      "step": 139230
    },
    {
      "epoch": 54.07,
      "learning_rate": 1.2790161812297737e-05,
      "loss": 0.1272,
      "step": 139240
    },
    {
      "epoch": 54.08,
      "learning_rate": 1.2789644012944983e-05,
      "loss": 0.2859,
      "step": 139250
    },
    {
      "epoch": 54.08,
      "learning_rate": 1.2789126213592233e-05,
      "loss": 0.0682,
      "step": 139260
    },
    {
      "epoch": 54.09,
      "learning_rate": 1.2788608414239483e-05,
      "loss": 0.1397,
      "step": 139270
    },
    {
      "epoch": 54.09,
      "learning_rate": 1.2788090614886732e-05,
      "loss": 0.031,
      "step": 139280
    },
    {
      "epoch": 54.09,
      "learning_rate": 1.2787572815533982e-05,
      "loss": 0.0286,
      "step": 139290
    },
    {
      "epoch": 54.1,
      "learning_rate": 1.2787055016181231e-05,
      "loss": 0.1421,
      "step": 139300
    },
    {
      "epoch": 54.1,
      "learning_rate": 1.2786537216828481e-05,
      "loss": 0.0031,
      "step": 139310
    },
    {
      "epoch": 54.1,
      "learning_rate": 1.278601941747573e-05,
      "loss": 0.0591,
      "step": 139320
    },
    {
      "epoch": 54.11,
      "learning_rate": 1.2785501618122977e-05,
      "loss": 0.034,
      "step": 139330
    },
    {
      "epoch": 54.11,
      "learning_rate": 1.2784983818770226e-05,
      "loss": 0.0696,
      "step": 139340
    },
    {
      "epoch": 54.12,
      "learning_rate": 1.2784466019417476e-05,
      "loss": 0.102,
      "step": 139350
    },
    {
      "epoch": 54.12,
      "learning_rate": 1.2783948220064726e-05,
      "loss": 0.0585,
      "step": 139360
    },
    {
      "epoch": 54.12,
      "learning_rate": 1.2783430420711975e-05,
      "loss": 0.1187,
      "step": 139370
    },
    {
      "epoch": 54.13,
      "learning_rate": 1.2782912621359225e-05,
      "loss": 0.0645,
      "step": 139380
    },
    {
      "epoch": 54.13,
      "learning_rate": 1.2782394822006475e-05,
      "loss": 0.1194,
      "step": 139390
    },
    {
      "epoch": 54.14,
      "learning_rate": 1.2781877022653724e-05,
      "loss": 0.0702,
      "step": 139400
    },
    {
      "epoch": 54.14,
      "learning_rate": 1.278135922330097e-05,
      "loss": 0.0202,
      "step": 139410
    },
    {
      "epoch": 54.14,
      "learning_rate": 1.278084142394822e-05,
      "loss": 0.0138,
      "step": 139420
    },
    {
      "epoch": 54.15,
      "learning_rate": 1.278032362459547e-05,
      "loss": 0.0431,
      "step": 139430
    },
    {
      "epoch": 54.15,
      "learning_rate": 1.277980582524272e-05,
      "loss": 0.0698,
      "step": 139440
    },
    {
      "epoch": 54.16,
      "learning_rate": 1.2779288025889969e-05,
      "loss": 0.12,
      "step": 139450
    },
    {
      "epoch": 54.16,
      "learning_rate": 1.2778770226537219e-05,
      "loss": 0.1108,
      "step": 139460
    },
    {
      "epoch": 54.16,
      "learning_rate": 1.2778252427184468e-05,
      "loss": 0.0521,
      "step": 139470
    },
    {
      "epoch": 54.17,
      "learning_rate": 1.2777734627831718e-05,
      "loss": 0.0555,
      "step": 139480
    },
    {
      "epoch": 54.17,
      "learning_rate": 1.2777216828478964e-05,
      "loss": 0.0876,
      "step": 139490
    },
    {
      "epoch": 54.17,
      "learning_rate": 1.2776699029126214e-05,
      "loss": 0.0496,
      "step": 139500
    },
    {
      "epoch": 54.18,
      "learning_rate": 1.2776181229773463e-05,
      "loss": 0.114,
      "step": 139510
    },
    {
      "epoch": 54.18,
      "learning_rate": 1.2775663430420713e-05,
      "loss": 0.0356,
      "step": 139520
    },
    {
      "epoch": 54.19,
      "learning_rate": 1.2775145631067962e-05,
      "loss": 0.1419,
      "step": 139530
    },
    {
      "epoch": 54.19,
      "learning_rate": 1.2774627831715212e-05,
      "loss": 0.009,
      "step": 139540
    },
    {
      "epoch": 54.19,
      "learning_rate": 1.2774110032362462e-05,
      "loss": 0.0053,
      "step": 139550
    },
    {
      "epoch": 54.2,
      "learning_rate": 1.2773592233009711e-05,
      "loss": 0.0997,
      "step": 139560
    },
    {
      "epoch": 54.2,
      "learning_rate": 1.2773074433656958e-05,
      "loss": 0.1274,
      "step": 139570
    },
    {
      "epoch": 54.21,
      "learning_rate": 1.2772556634304207e-05,
      "loss": 0.0722,
      "step": 139580
    },
    {
      "epoch": 54.21,
      "learning_rate": 1.2772038834951457e-05,
      "loss": 0.126,
      "step": 139590
    },
    {
      "epoch": 54.21,
      "learning_rate": 1.2771521035598706e-05,
      "loss": 0.0219,
      "step": 139600
    },
    {
      "epoch": 54.22,
      "learning_rate": 1.2771003236245956e-05,
      "loss": 0.1736,
      "step": 139610
    },
    {
      "epoch": 54.22,
      "learning_rate": 1.2770485436893206e-05,
      "loss": 0.3487,
      "step": 139620
    },
    {
      "epoch": 54.23,
      "learning_rate": 1.2769967637540455e-05,
      "loss": 0.1613,
      "step": 139630
    },
    {
      "epoch": 54.23,
      "learning_rate": 1.2769449838187705e-05,
      "loss": 0.0798,
      "step": 139640
    },
    {
      "epoch": 54.23,
      "learning_rate": 1.2768932038834953e-05,
      "loss": 0.1508,
      "step": 139650
    },
    {
      "epoch": 54.24,
      "learning_rate": 1.27684142394822e-05,
      "loss": 0.162,
      "step": 139660
    },
    {
      "epoch": 54.24,
      "learning_rate": 1.276789644012945e-05,
      "loss": 0.0639,
      "step": 139670
    },
    {
      "epoch": 54.24,
      "learning_rate": 1.27673786407767e-05,
      "loss": 0.0984,
      "step": 139680
    },
    {
      "epoch": 54.25,
      "learning_rate": 1.276686084142395e-05,
      "loss": 0.0422,
      "step": 139690
    },
    {
      "epoch": 54.25,
      "learning_rate": 1.27663430420712e-05,
      "loss": 0.1928,
      "step": 139700
    },
    {
      "epoch": 54.26,
      "learning_rate": 1.2765825242718449e-05,
      "loss": 0.1314,
      "step": 139710
    },
    {
      "epoch": 54.26,
      "learning_rate": 1.2765307443365698e-05,
      "loss": 0.1061,
      "step": 139720
    },
    {
      "epoch": 54.26,
      "learning_rate": 1.2764789644012946e-05,
      "loss": 0.0188,
      "step": 139730
    },
    {
      "epoch": 54.27,
      "learning_rate": 1.2764271844660194e-05,
      "loss": 0.1819,
      "step": 139740
    },
    {
      "epoch": 54.27,
      "learning_rate": 1.2763754045307444e-05,
      "loss": 0.0697,
      "step": 139750
    },
    {
      "epoch": 54.28,
      "learning_rate": 1.2763236245954694e-05,
      "loss": 0.0286,
      "step": 139760
    },
    {
      "epoch": 54.28,
      "learning_rate": 1.2762718446601943e-05,
      "loss": 0.0755,
      "step": 139770
    },
    {
      "epoch": 54.28,
      "learning_rate": 1.2762200647249193e-05,
      "loss": 0.1376,
      "step": 139780
    },
    {
      "epoch": 54.29,
      "learning_rate": 1.2761682847896442e-05,
      "loss": 0.0138,
      "step": 139790
    },
    {
      "epoch": 54.29,
      "learning_rate": 1.276116504854369e-05,
      "loss": 0.0357,
      "step": 139800
    },
    {
      "epoch": 54.3,
      "learning_rate": 1.276064724919094e-05,
      "loss": 0.0296,
      "step": 139810
    },
    {
      "epoch": 54.3,
      "learning_rate": 1.2760129449838188e-05,
      "loss": 0.0656,
      "step": 139820
    },
    {
      "epoch": 54.3,
      "learning_rate": 1.2759611650485437e-05,
      "loss": 0.0273,
      "step": 139830
    },
    {
      "epoch": 54.31,
      "learning_rate": 1.2759093851132687e-05,
      "loss": 0.0342,
      "step": 139840
    },
    {
      "epoch": 54.31,
      "learning_rate": 1.2758576051779937e-05,
      "loss": 0.1261,
      "step": 139850
    },
    {
      "epoch": 54.31,
      "learning_rate": 1.2758058252427186e-05,
      "loss": 0.0469,
      "step": 139860
    },
    {
      "epoch": 54.32,
      "learning_rate": 1.2757540453074436e-05,
      "loss": 0.0543,
      "step": 139870
    },
    {
      "epoch": 54.32,
      "learning_rate": 1.2757022653721684e-05,
      "loss": 0.0231,
      "step": 139880
    },
    {
      "epoch": 54.33,
      "learning_rate": 1.2756504854368933e-05,
      "loss": 0.0739,
      "step": 139890
    },
    {
      "epoch": 54.33,
      "learning_rate": 1.2755987055016181e-05,
      "loss": 0.1235,
      "step": 139900
    },
    {
      "epoch": 54.33,
      "learning_rate": 1.2755469255663431e-05,
      "loss": 0.1258,
      "step": 139910
    },
    {
      "epoch": 54.34,
      "learning_rate": 1.275495145631068e-05,
      "loss": 0.0774,
      "step": 139920
    },
    {
      "epoch": 54.34,
      "learning_rate": 1.275443365695793e-05,
      "loss": 0.0745,
      "step": 139930
    },
    {
      "epoch": 54.35,
      "learning_rate": 1.275391585760518e-05,
      "loss": 0.0837,
      "step": 139940
    },
    {
      "epoch": 54.35,
      "learning_rate": 1.2753398058252428e-05,
      "loss": 0.1456,
      "step": 139950
    },
    {
      "epoch": 54.35,
      "learning_rate": 1.2752880258899677e-05,
      "loss": 0.0531,
      "step": 139960
    },
    {
      "epoch": 54.36,
      "learning_rate": 1.2752362459546927e-05,
      "loss": 0.0865,
      "step": 139970
    },
    {
      "epoch": 54.36,
      "learning_rate": 1.2751844660194175e-05,
      "loss": 0.1078,
      "step": 139980
    },
    {
      "epoch": 54.37,
      "learning_rate": 1.2751326860841425e-05,
      "loss": 0.0327,
      "step": 139990
    },
    {
      "epoch": 54.37,
      "learning_rate": 1.2750809061488674e-05,
      "loss": 0.1181,
      "step": 140000
    },
    {
      "epoch": 54.37,
      "learning_rate": 1.2750291262135924e-05,
      "loss": 0.1126,
      "step": 140010
    },
    {
      "epoch": 54.38,
      "learning_rate": 1.2749773462783173e-05,
      "loss": 0.1063,
      "step": 140020
    },
    {
      "epoch": 54.38,
      "learning_rate": 1.2749255663430421e-05,
      "loss": 0.1153,
      "step": 140030
    },
    {
      "epoch": 54.38,
      "learning_rate": 1.2748737864077671e-05,
      "loss": 0.022,
      "step": 140040
    },
    {
      "epoch": 54.39,
      "learning_rate": 1.274822006472492e-05,
      "loss": 0.0945,
      "step": 140050
    },
    {
      "epoch": 54.39,
      "learning_rate": 1.2747702265372168e-05,
      "loss": 0.126,
      "step": 140060
    },
    {
      "epoch": 54.4,
      "learning_rate": 1.2747184466019418e-05,
      "loss": 0.0052,
      "step": 140070
    },
    {
      "epoch": 54.4,
      "learning_rate": 1.2746666666666668e-05,
      "loss": 0.1314,
      "step": 140080
    },
    {
      "epoch": 54.4,
      "learning_rate": 1.2746148867313917e-05,
      "loss": 0.0936,
      "step": 140090
    },
    {
      "epoch": 54.41,
      "learning_rate": 1.2745631067961165e-05,
      "loss": 0.1739,
      "step": 140100
    },
    {
      "epoch": 54.41,
      "learning_rate": 1.2745113268608415e-05,
      "loss": 0.0916,
      "step": 140110
    },
    {
      "epoch": 54.42,
      "learning_rate": 1.2744595469255665e-05,
      "loss": 0.0411,
      "step": 140120
    },
    {
      "epoch": 54.42,
      "learning_rate": 1.2744077669902914e-05,
      "loss": 0.0302,
      "step": 140130
    },
    {
      "epoch": 54.42,
      "learning_rate": 1.2743559870550162e-05,
      "loss": 0.0617,
      "step": 140140
    },
    {
      "epoch": 54.43,
      "learning_rate": 1.2743042071197412e-05,
      "loss": 0.098,
      "step": 140150
    },
    {
      "epoch": 54.43,
      "learning_rate": 1.2742524271844661e-05,
      "loss": 0.0351,
      "step": 140160
    },
    {
      "epoch": 54.43,
      "learning_rate": 1.2742006472491911e-05,
      "loss": 0.007,
      "step": 140170
    },
    {
      "epoch": 54.44,
      "learning_rate": 1.2741488673139159e-05,
      "loss": 0.0153,
      "step": 140180
    },
    {
      "epoch": 54.44,
      "learning_rate": 1.2740970873786408e-05,
      "loss": 0.0605,
      "step": 140190
    },
    {
      "epoch": 54.45,
      "learning_rate": 1.2740453074433658e-05,
      "loss": 0.2399,
      "step": 140200
    },
    {
      "epoch": 54.45,
      "learning_rate": 1.2739935275080908e-05,
      "loss": 0.0809,
      "step": 140210
    },
    {
      "epoch": 54.45,
      "learning_rate": 1.2739417475728157e-05,
      "loss": 0.0195,
      "step": 140220
    },
    {
      "epoch": 54.46,
      "learning_rate": 1.2738899676375405e-05,
      "loss": 0.0131,
      "step": 140230
    },
    {
      "epoch": 54.46,
      "learning_rate": 1.2738381877022655e-05,
      "loss": 0.0004,
      "step": 140240
    },
    {
      "epoch": 54.47,
      "learning_rate": 1.2737864077669904e-05,
      "loss": 0.0117,
      "step": 140250
    },
    {
      "epoch": 54.47,
      "learning_rate": 1.2737346278317152e-05,
      "loss": 0.137,
      "step": 140260
    },
    {
      "epoch": 54.47,
      "learning_rate": 1.2736828478964402e-05,
      "loss": 0.0349,
      "step": 140270
    },
    {
      "epoch": 54.48,
      "learning_rate": 1.2736310679611652e-05,
      "loss": 0.0738,
      "step": 140280
    },
    {
      "epoch": 54.48,
      "learning_rate": 1.2735792880258901e-05,
      "loss": 0.217,
      "step": 140290
    },
    {
      "epoch": 54.49,
      "learning_rate": 1.2735275080906151e-05,
      "loss": 0.0869,
      "step": 140300
    },
    {
      "epoch": 54.49,
      "learning_rate": 1.2734757281553399e-05,
      "loss": 0.0386,
      "step": 140310
    },
    {
      "epoch": 54.49,
      "learning_rate": 1.2734239482200648e-05,
      "loss": 0.0277,
      "step": 140320
    },
    {
      "epoch": 54.5,
      "learning_rate": 1.2733721682847896e-05,
      "loss": 0.0235,
      "step": 140330
    },
    {
      "epoch": 54.5,
      "learning_rate": 1.2733203883495146e-05,
      "loss": 0.0168,
      "step": 140340
    },
    {
      "epoch": 54.5,
      "learning_rate": 1.2732686084142396e-05,
      "loss": 0.1017,
      "step": 140350
    },
    {
      "epoch": 54.51,
      "learning_rate": 1.2732168284789645e-05,
      "loss": 0.03,
      "step": 140360
    },
    {
      "epoch": 54.51,
      "learning_rate": 1.2731650485436895e-05,
      "loss": 0.105,
      "step": 140370
    },
    {
      "epoch": 54.52,
      "learning_rate": 1.2731132686084144e-05,
      "loss": 0.0876,
      "step": 140380
    },
    {
      "epoch": 54.52,
      "learning_rate": 1.2730614886731392e-05,
      "loss": 0.1319,
      "step": 140390
    },
    {
      "epoch": 54.52,
      "learning_rate": 1.2730097087378642e-05,
      "loss": 0.1311,
      "step": 140400
    },
    {
      "epoch": 54.53,
      "learning_rate": 1.272957928802589e-05,
      "loss": 0.0358,
      "step": 140410
    },
    {
      "epoch": 54.53,
      "learning_rate": 1.272906148867314e-05,
      "loss": 0.0035,
      "step": 140420
    },
    {
      "epoch": 54.54,
      "learning_rate": 1.2728543689320389e-05,
      "loss": 0.0556,
      "step": 140430
    },
    {
      "epoch": 54.54,
      "learning_rate": 1.2728025889967639e-05,
      "loss": 0.0696,
      "step": 140440
    },
    {
      "epoch": 54.54,
      "learning_rate": 1.2727508090614888e-05,
      "loss": 0.0655,
      "step": 140450
    },
    {
      "epoch": 54.55,
      "learning_rate": 1.2726990291262138e-05,
      "loss": 0.1188,
      "step": 140460
    },
    {
      "epoch": 54.55,
      "learning_rate": 1.2726472491909386e-05,
      "loss": 0.024,
      "step": 140470
    },
    {
      "epoch": 54.56,
      "learning_rate": 1.2725954692556634e-05,
      "loss": 0.2087,
      "step": 140480
    },
    {
      "epoch": 54.56,
      "learning_rate": 1.2725436893203883e-05,
      "loss": 0.1394,
      "step": 140490
    },
    {
      "epoch": 54.56,
      "learning_rate": 1.2724919093851133e-05,
      "loss": 0.0673,
      "step": 140500
    },
    {
      "epoch": 54.57,
      "learning_rate": 1.2724401294498383e-05,
      "loss": 0.0985,
      "step": 140510
    },
    {
      "epoch": 54.57,
      "learning_rate": 1.2723883495145632e-05,
      "loss": 0.1547,
      "step": 140520
    },
    {
      "epoch": 54.57,
      "learning_rate": 1.2723365695792882e-05,
      "loss": 0.0498,
      "step": 140530
    },
    {
      "epoch": 54.58,
      "learning_rate": 1.2722847896440132e-05,
      "loss": 0.0723,
      "step": 140540
    },
    {
      "epoch": 54.58,
      "learning_rate": 1.272233009708738e-05,
      "loss": 0.076,
      "step": 140550
    },
    {
      "epoch": 54.59,
      "learning_rate": 1.2721812297734627e-05,
      "loss": 0.0792,
      "step": 140560
    },
    {
      "epoch": 54.59,
      "learning_rate": 1.2721294498381877e-05,
      "loss": 0.0379,
      "step": 140570
    },
    {
      "epoch": 54.59,
      "learning_rate": 1.2720776699029127e-05,
      "loss": 0.1347,
      "step": 140580
    },
    {
      "epoch": 54.6,
      "learning_rate": 1.2720258899676376e-05,
      "loss": 0.1407,
      "step": 140590
    },
    {
      "epoch": 54.6,
      "learning_rate": 1.2719741100323626e-05,
      "loss": 0.0301,
      "step": 140600
    },
    {
      "epoch": 54.61,
      "learning_rate": 1.2719223300970875e-05,
      "loss": 0.0676,
      "step": 140610
    },
    {
      "epoch": 54.61,
      "learning_rate": 1.2718705501618125e-05,
      "loss": 0.0422,
      "step": 140620
    },
    {
      "epoch": 54.61,
      "learning_rate": 1.2718187702265371e-05,
      "loss": 0.0146,
      "step": 140630
    },
    {
      "epoch": 54.62,
      "learning_rate": 1.2717669902912621e-05,
      "loss": 0.0747,
      "step": 140640
    },
    {
      "epoch": 54.62,
      "learning_rate": 1.271715210355987e-05,
      "loss": 0.1522,
      "step": 140650
    },
    {
      "epoch": 54.63,
      "learning_rate": 1.271663430420712e-05,
      "loss": 0.0264,
      "step": 140660
    },
    {
      "epoch": 54.63,
      "learning_rate": 1.271611650485437e-05,
      "loss": 0.0095,
      "step": 140670
    },
    {
      "epoch": 54.63,
      "learning_rate": 1.271559870550162e-05,
      "loss": 0.0634,
      "step": 140680
    },
    {
      "epoch": 54.64,
      "learning_rate": 1.2715080906148869e-05,
      "loss": 0.1337,
      "step": 140690
    },
    {
      "epoch": 54.64,
      "learning_rate": 1.2714563106796119e-05,
      "loss": 0.0376,
      "step": 140700
    },
    {
      "epoch": 54.64,
      "learning_rate": 1.2714045307443368e-05,
      "loss": 0.0419,
      "step": 140710
    },
    {
      "epoch": 54.65,
      "learning_rate": 1.2713527508090614e-05,
      "loss": 0.1088,
      "step": 140720
    },
    {
      "epoch": 54.65,
      "learning_rate": 1.2713009708737864e-05,
      "loss": 0.0156,
      "step": 140730
    },
    {
      "epoch": 54.66,
      "learning_rate": 1.2712491909385114e-05,
      "loss": 0.0229,
      "step": 140740
    },
    {
      "epoch": 54.66,
      "learning_rate": 1.2711974110032363e-05,
      "loss": 0.1092,
      "step": 140750
    },
    {
      "epoch": 54.66,
      "learning_rate": 1.2711456310679613e-05,
      "loss": 0.2339,
      "step": 140760
    },
    {
      "epoch": 54.67,
      "learning_rate": 1.2710938511326863e-05,
      "loss": 0.1252,
      "step": 140770
    },
    {
      "epoch": 54.67,
      "learning_rate": 1.2710420711974112e-05,
      "loss": 0.1215,
      "step": 140780
    },
    {
      "epoch": 54.68,
      "learning_rate": 1.2709902912621362e-05,
      "loss": 0.1302,
      "step": 140790
    },
    {
      "epoch": 54.68,
      "learning_rate": 1.2709385113268608e-05,
      "loss": 0.0498,
      "step": 140800
    },
    {
      "epoch": 54.68,
      "learning_rate": 1.2708867313915858e-05,
      "loss": 0.115,
      "step": 140810
    },
    {
      "epoch": 54.69,
      "learning_rate": 1.2708349514563107e-05,
      "loss": 0.1193,
      "step": 140820
    },
    {
      "epoch": 54.69,
      "learning_rate": 1.2707831715210357e-05,
      "loss": 0.1008,
      "step": 140830
    },
    {
      "epoch": 54.7,
      "learning_rate": 1.2707313915857607e-05,
      "loss": 0.1112,
      "step": 140840
    },
    {
      "epoch": 54.7,
      "learning_rate": 1.2706796116504856e-05,
      "loss": 0.0363,
      "step": 140850
    },
    {
      "epoch": 54.7,
      "learning_rate": 1.2706278317152106e-05,
      "loss": 0.0921,
      "step": 140860
    },
    {
      "epoch": 54.71,
      "learning_rate": 1.2705760517799355e-05,
      "loss": 0.058,
      "step": 140870
    },
    {
      "epoch": 54.71,
      "learning_rate": 1.2705242718446602e-05,
      "loss": 0.0453,
      "step": 140880
    },
    {
      "epoch": 54.71,
      "learning_rate": 1.2704724919093851e-05,
      "loss": 0.2195,
      "step": 140890
    },
    {
      "epoch": 54.72,
      "learning_rate": 1.27042071197411e-05,
      "loss": 0.0253,
      "step": 140900
    },
    {
      "epoch": 54.72,
      "learning_rate": 1.270368932038835e-05,
      "loss": 0.1713,
      "step": 140910
    },
    {
      "epoch": 54.73,
      "learning_rate": 1.27031715210356e-05,
      "loss": 0.1513,
      "step": 140920
    },
    {
      "epoch": 54.73,
      "learning_rate": 1.270265372168285e-05,
      "loss": 0.0296,
      "step": 140930
    },
    {
      "epoch": 54.73,
      "learning_rate": 1.27021359223301e-05,
      "loss": 0.0789,
      "step": 140940
    },
    {
      "epoch": 54.74,
      "learning_rate": 1.2701618122977349e-05,
      "loss": 0.0412,
      "step": 140950
    },
    {
      "epoch": 54.74,
      "learning_rate": 1.2701100323624595e-05,
      "loss": 0.0529,
      "step": 140960
    },
    {
      "epoch": 54.75,
      "learning_rate": 1.2700582524271845e-05,
      "loss": 0.1453,
      "step": 140970
    },
    {
      "epoch": 54.75,
      "learning_rate": 1.2700064724919094e-05,
      "loss": 0.0293,
      "step": 140980
    },
    {
      "epoch": 54.75,
      "learning_rate": 1.2699546925566344e-05,
      "loss": 0.101,
      "step": 140990
    },
    {
      "epoch": 54.76,
      "learning_rate": 1.2699029126213594e-05,
      "loss": 0.0993,
      "step": 141000
    },
    {
      "epoch": 54.76,
      "learning_rate": 1.2698511326860843e-05,
      "loss": 0.1448,
      "step": 141010
    },
    {
      "epoch": 54.77,
      "learning_rate": 1.2697993527508093e-05,
      "loss": 0.0479,
      "step": 141020
    },
    {
      "epoch": 54.77,
      "learning_rate": 1.2697475728155342e-05,
      "loss": 0.0767,
      "step": 141030
    },
    {
      "epoch": 54.77,
      "learning_rate": 1.2696957928802589e-05,
      "loss": 0.0499,
      "step": 141040
    },
    {
      "epoch": 54.78,
      "learning_rate": 1.2696440129449838e-05,
      "loss": 0.0792,
      "step": 141050
    },
    {
      "epoch": 54.78,
      "learning_rate": 1.2695922330097088e-05,
      "loss": 0.0571,
      "step": 141060
    },
    {
      "epoch": 54.78,
      "learning_rate": 1.2695404530744338e-05,
      "loss": 0.0493,
      "step": 141070
    },
    {
      "epoch": 54.79,
      "learning_rate": 1.2694886731391587e-05,
      "loss": 0.0181,
      "step": 141080
    },
    {
      "epoch": 54.79,
      "learning_rate": 1.2694368932038837e-05,
      "loss": 0.0123,
      "step": 141090
    },
    {
      "epoch": 54.8,
      "learning_rate": 1.2693851132686086e-05,
      "loss": 0.1277,
      "step": 141100
    },
    {
      "epoch": 54.8,
      "learning_rate": 1.2693333333333336e-05,
      "loss": 0.0715,
      "step": 141110
    },
    {
      "epoch": 54.8,
      "learning_rate": 1.2692815533980582e-05,
      "loss": 0.1867,
      "step": 141120
    },
    {
      "epoch": 54.81,
      "learning_rate": 1.2692297734627832e-05,
      "loss": 0.0112,
      "step": 141130
    },
    {
      "epoch": 54.81,
      "learning_rate": 1.2691779935275082e-05,
      "loss": 0.2275,
      "step": 141140
    },
    {
      "epoch": 54.82,
      "learning_rate": 1.2691262135922331e-05,
      "loss": 0.008,
      "step": 141150
    },
    {
      "epoch": 54.82,
      "learning_rate": 1.269074433656958e-05,
      "loss": 0.1353,
      "step": 141160
    },
    {
      "epoch": 54.82,
      "learning_rate": 1.269022653721683e-05,
      "loss": 0.1227,
      "step": 141170
    },
    {
      "epoch": 54.83,
      "learning_rate": 1.268970873786408e-05,
      "loss": 0.111,
      "step": 141180
    },
    {
      "epoch": 54.83,
      "learning_rate": 1.268919093851133e-05,
      "loss": 0.127,
      "step": 141190
    },
    {
      "epoch": 54.83,
      "learning_rate": 1.2688673139158576e-05,
      "loss": 0.0555,
      "step": 141200
    },
    {
      "epoch": 54.84,
      "learning_rate": 1.2688155339805825e-05,
      "loss": 0.1882,
      "step": 141210
    },
    {
      "epoch": 54.84,
      "learning_rate": 1.2687637540453075e-05,
      "loss": 0.1245,
      "step": 141220
    },
    {
      "epoch": 54.85,
      "learning_rate": 1.2687119741100325e-05,
      "loss": 0.0603,
      "step": 141230
    },
    {
      "epoch": 54.85,
      "learning_rate": 1.2686601941747574e-05,
      "loss": 0.1147,
      "step": 141240
    },
    {
      "epoch": 54.85,
      "learning_rate": 1.2686084142394824e-05,
      "loss": 0.0644,
      "step": 141250
    },
    {
      "epoch": 54.86,
      "learning_rate": 1.2685566343042074e-05,
      "loss": 0.0033,
      "step": 141260
    },
    {
      "epoch": 54.86,
      "learning_rate": 1.2685048543689321e-05,
      "loss": 0.1203,
      "step": 141270
    },
    {
      "epoch": 54.87,
      "learning_rate": 1.2684530744336571e-05,
      "loss": 0.1065,
      "step": 141280
    },
    {
      "epoch": 54.87,
      "learning_rate": 1.2684012944983819e-05,
      "loss": 0.1231,
      "step": 141290
    },
    {
      "epoch": 54.87,
      "learning_rate": 1.2683495145631069e-05,
      "loss": 0.0532,
      "step": 141300
    },
    {
      "epoch": 54.88,
      "learning_rate": 1.2682977346278318e-05,
      "loss": 0.0542,
      "step": 141310
    },
    {
      "epoch": 54.88,
      "learning_rate": 1.2682459546925568e-05,
      "loss": 0.0032,
      "step": 141320
    },
    {
      "epoch": 54.89,
      "learning_rate": 1.2681941747572817e-05,
      "loss": 0.0489,
      "step": 141330
    },
    {
      "epoch": 54.89,
      "learning_rate": 1.2681423948220067e-05,
      "loss": 0.0491,
      "step": 141340
    },
    {
      "epoch": 54.89,
      "learning_rate": 1.2680906148867315e-05,
      "loss": 0.0578,
      "step": 141350
    },
    {
      "epoch": 54.9,
      "learning_rate": 1.2680388349514565e-05,
      "loss": 0.0166,
      "step": 141360
    },
    {
      "epoch": 54.9,
      "learning_rate": 1.2679870550161813e-05,
      "loss": 0.0068,
      "step": 141370
    },
    {
      "epoch": 54.9,
      "learning_rate": 1.2679352750809062e-05,
      "loss": 0.1781,
      "step": 141380
    },
    {
      "epoch": 54.91,
      "learning_rate": 1.2678834951456312e-05,
      "loss": 0.0955,
      "step": 141390
    },
    {
      "epoch": 54.91,
      "learning_rate": 1.2678317152103561e-05,
      "loss": 0.0367,
      "step": 141400
    },
    {
      "epoch": 54.92,
      "learning_rate": 1.2677799352750811e-05,
      "loss": 0.0531,
      "step": 141410
    },
    {
      "epoch": 54.92,
      "learning_rate": 1.2677281553398059e-05,
      "loss": 0.0801,
      "step": 141420
    },
    {
      "epoch": 54.92,
      "learning_rate": 1.2676763754045309e-05,
      "loss": 0.0368,
      "step": 141430
    },
    {
      "epoch": 54.93,
      "learning_rate": 1.2676245954692558e-05,
      "loss": 0.035,
      "step": 141440
    },
    {
      "epoch": 54.93,
      "learning_rate": 1.2675728155339806e-05,
      "loss": 0.0912,
      "step": 141450
    },
    {
      "epoch": 54.94,
      "learning_rate": 1.2675210355987056e-05,
      "loss": 0.0645,
      "step": 141460
    },
    {
      "epoch": 54.94,
      "learning_rate": 1.2674692556634305e-05,
      "loss": 0.0715,
      "step": 141470
    },
    {
      "epoch": 54.94,
      "learning_rate": 1.2674174757281555e-05,
      "loss": 0.1044,
      "step": 141480
    },
    {
      "epoch": 54.95,
      "learning_rate": 1.2673656957928805e-05,
      "loss": 0.0528,
      "step": 141490
    },
    {
      "epoch": 54.95,
      "learning_rate": 1.2673139158576053e-05,
      "loss": 0.0399,
      "step": 141500
    },
    {
      "epoch": 54.96,
      "learning_rate": 1.2672621359223302e-05,
      "loss": 0.0863,
      "step": 141510
    },
    {
      "epoch": 54.96,
      "learning_rate": 1.2672103559870552e-05,
      "loss": 0.0386,
      "step": 141520
    },
    {
      "epoch": 54.96,
      "learning_rate": 1.26715857605178e-05,
      "loss": 0.1754,
      "step": 141530
    },
    {
      "epoch": 54.97,
      "learning_rate": 1.267106796116505e-05,
      "loss": 0.0556,
      "step": 141540
    },
    {
      "epoch": 54.97,
      "learning_rate": 1.2670550161812299e-05,
      "loss": 0.0203,
      "step": 141550
    },
    {
      "epoch": 54.97,
      "learning_rate": 1.2670032362459549e-05,
      "loss": 0.0545,
      "step": 141560
    },
    {
      "epoch": 54.98,
      "learning_rate": 1.2669514563106796e-05,
      "loss": 0.0999,
      "step": 141570
    },
    {
      "epoch": 54.98,
      "learning_rate": 1.2668996763754046e-05,
      "loss": 0.0427,
      "step": 141580
    },
    {
      "epoch": 54.99,
      "learning_rate": 1.2668478964401296e-05,
      "loss": 0.0003,
      "step": 141590
    },
    {
      "epoch": 54.99,
      "learning_rate": 1.2667961165048545e-05,
      "loss": 0.0296,
      "step": 141600
    },
    {
      "epoch": 54.99,
      "learning_rate": 1.2667443365695793e-05,
      "loss": 0.0801,
      "step": 141610
    },
    {
      "epoch": 55.0,
      "learning_rate": 1.2666925566343043e-05,
      "loss": 0.1022,
      "step": 141620
    },
    {
      "epoch": 55.0,
      "eval_accuracy": 0.9491059147180193,
      "eval_loss": 0.32835954427719116,
      "eval_runtime": 8.2469,
      "eval_samples_per_second": 440.774,
      "eval_steps_per_second": 55.173,
      "step": 141625
    },
    {
      "epoch": 55.0,
      "learning_rate": 1.2666407766990292e-05,
      "loss": 0.0212,
      "step": 141630
    },
    {
      "epoch": 55.01,
      "learning_rate": 1.2665889967637542e-05,
      "loss": 0.1452,
      "step": 141640
    },
    {
      "epoch": 55.01,
      "learning_rate": 1.266537216828479e-05,
      "loss": 0.059,
      "step": 141650
    },
    {
      "epoch": 55.01,
      "learning_rate": 1.266485436893204e-05,
      "loss": 0.0005,
      "step": 141660
    },
    {
      "epoch": 55.02,
      "learning_rate": 1.266433656957929e-05,
      "loss": 0.0512,
      "step": 141670
    },
    {
      "epoch": 55.02,
      "learning_rate": 1.2663818770226539e-05,
      "loss": 0.0275,
      "step": 141680
    },
    {
      "epoch": 55.03,
      "learning_rate": 1.2663300970873787e-05,
      "loss": 0.1011,
      "step": 141690
    },
    {
      "epoch": 55.03,
      "learning_rate": 1.2662783171521036e-05,
      "loss": 0.1663,
      "step": 141700
    },
    {
      "epoch": 55.03,
      "learning_rate": 1.2662265372168286e-05,
      "loss": 0.1308,
      "step": 141710
    },
    {
      "epoch": 55.04,
      "learning_rate": 1.2661747572815536e-05,
      "loss": 0.1457,
      "step": 141720
    },
    {
      "epoch": 55.04,
      "learning_rate": 1.2661229773462784e-05,
      "loss": 0.2276,
      "step": 141730
    },
    {
      "epoch": 55.04,
      "learning_rate": 1.2660711974110033e-05,
      "loss": 0.0246,
      "step": 141740
    },
    {
      "epoch": 55.05,
      "learning_rate": 1.2660194174757283e-05,
      "loss": 0.0748,
      "step": 141750
    },
    {
      "epoch": 55.05,
      "learning_rate": 1.2659676375404532e-05,
      "loss": 0.0374,
      "step": 141760
    },
    {
      "epoch": 55.06,
      "learning_rate": 1.265915857605178e-05,
      "loss": 0.0643,
      "step": 141770
    },
    {
      "epoch": 55.06,
      "learning_rate": 1.265864077669903e-05,
      "loss": 0.0519,
      "step": 141780
    },
    {
      "epoch": 55.06,
      "learning_rate": 1.265812297734628e-05,
      "loss": 0.0642,
      "step": 141790
    },
    {
      "epoch": 55.07,
      "learning_rate": 1.2657605177993527e-05,
      "loss": 0.0108,
      "step": 141800
    },
    {
      "epoch": 55.07,
      "learning_rate": 1.2657087378640777e-05,
      "loss": 0.0523,
      "step": 141810
    },
    {
      "epoch": 55.08,
      "learning_rate": 1.2656569579288027e-05,
      "loss": 0.0748,
      "step": 141820
    },
    {
      "epoch": 55.08,
      "learning_rate": 1.2656051779935276e-05,
      "loss": 0.1278,
      "step": 141830
    },
    {
      "epoch": 55.08,
      "learning_rate": 1.2655533980582526e-05,
      "loss": 0.0479,
      "step": 141840
    },
    {
      "epoch": 55.09,
      "learning_rate": 1.2655016181229776e-05,
      "loss": 0.0662,
      "step": 141850
    },
    {
      "epoch": 55.09,
      "learning_rate": 1.2654498381877024e-05,
      "loss": 0.0935,
      "step": 141860
    },
    {
      "epoch": 55.1,
      "learning_rate": 1.2653980582524273e-05,
      "loss": 0.0086,
      "step": 141870
    },
    {
      "epoch": 55.1,
      "learning_rate": 1.2653462783171521e-05,
      "loss": 0.0973,
      "step": 141880
    },
    {
      "epoch": 55.1,
      "learning_rate": 1.265294498381877e-05,
      "loss": 0.029,
      "step": 141890
    },
    {
      "epoch": 55.11,
      "learning_rate": 1.265242718446602e-05,
      "loss": 0.0711,
      "step": 141900
    },
    {
      "epoch": 55.11,
      "learning_rate": 1.265190938511327e-05,
      "loss": 0.0005,
      "step": 141910
    },
    {
      "epoch": 55.11,
      "learning_rate": 1.265139158576052e-05,
      "loss": 0.0653,
      "step": 141920
    },
    {
      "epoch": 55.12,
      "learning_rate": 1.2650873786407769e-05,
      "loss": 0.0394,
      "step": 141930
    },
    {
      "epoch": 55.12,
      "learning_rate": 1.2650355987055017e-05,
      "loss": 0.0981,
      "step": 141940
    },
    {
      "epoch": 55.13,
      "learning_rate": 1.2649838187702265e-05,
      "loss": 0.1034,
      "step": 141950
    },
    {
      "epoch": 55.13,
      "learning_rate": 1.2649320388349515e-05,
      "loss": 0.141,
      "step": 141960
    },
    {
      "epoch": 55.13,
      "learning_rate": 1.2648802588996764e-05,
      "loss": 0.0611,
      "step": 141970
    },
    {
      "epoch": 55.14,
      "learning_rate": 1.2648284789644014e-05,
      "loss": 0.1003,
      "step": 141980
    },
    {
      "epoch": 55.14,
      "learning_rate": 1.2647766990291263e-05,
      "loss": 0.0392,
      "step": 141990
    },
    {
      "epoch": 55.15,
      "learning_rate": 1.2647249190938513e-05,
      "loss": 0.2226,
      "step": 142000
    },
    {
      "epoch": 55.15,
      "learning_rate": 1.2646731391585763e-05,
      "loss": 0.1075,
      "step": 142010
    },
    {
      "epoch": 55.15,
      "learning_rate": 1.264621359223301e-05,
      "loss": 0.0325,
      "step": 142020
    },
    {
      "epoch": 55.16,
      "learning_rate": 1.2645695792880259e-05,
      "loss": 0.012,
      "step": 142030
    },
    {
      "epoch": 55.16,
      "learning_rate": 1.2645177993527508e-05,
      "loss": 0.0156,
      "step": 142040
    },
    {
      "epoch": 55.17,
      "learning_rate": 1.2644660194174758e-05,
      "loss": 0.0314,
      "step": 142050
    },
    {
      "epoch": 55.17,
      "learning_rate": 1.2644142394822007e-05,
      "loss": 0.0201,
      "step": 142060
    },
    {
      "epoch": 55.17,
      "learning_rate": 1.2643624595469257e-05,
      "loss": 0.0745,
      "step": 142070
    },
    {
      "epoch": 55.18,
      "learning_rate": 1.2643106796116507e-05,
      "loss": 0.0379,
      "step": 142080
    },
    {
      "epoch": 55.18,
      "learning_rate": 1.2642588996763756e-05,
      "loss": 0.09,
      "step": 142090
    },
    {
      "epoch": 55.18,
      "learning_rate": 1.2642071197411002e-05,
      "loss": 0.0408,
      "step": 142100
    },
    {
      "epoch": 55.19,
      "learning_rate": 1.2641553398058252e-05,
      "loss": 0.0881,
      "step": 142110
    },
    {
      "epoch": 55.19,
      "learning_rate": 1.2641035598705502e-05,
      "loss": 0.1241,
      "step": 142120
    },
    {
      "epoch": 55.2,
      "learning_rate": 1.2640517799352751e-05,
      "loss": 0.0149,
      "step": 142130
    },
    {
      "epoch": 55.2,
      "learning_rate": 1.2640000000000001e-05,
      "loss": 0.0759,
      "step": 142140
    },
    {
      "epoch": 55.2,
      "learning_rate": 1.263948220064725e-05,
      "loss": 0.0313,
      "step": 142150
    },
    {
      "epoch": 55.21,
      "learning_rate": 1.26389644012945e-05,
      "loss": 0.1009,
      "step": 142160
    },
    {
      "epoch": 55.21,
      "learning_rate": 1.263844660194175e-05,
      "loss": 0.0117,
      "step": 142170
    },
    {
      "epoch": 55.22,
      "learning_rate": 1.2637928802588996e-05,
      "loss": 0.0327,
      "step": 142180
    },
    {
      "epoch": 55.22,
      "learning_rate": 1.2637411003236246e-05,
      "loss": 0.1146,
      "step": 142190
    },
    {
      "epoch": 55.22,
      "learning_rate": 1.2636893203883495e-05,
      "loss": 0.0685,
      "step": 142200
    },
    {
      "epoch": 55.23,
      "learning_rate": 1.2636375404530745e-05,
      "loss": 0.149,
      "step": 142210
    },
    {
      "epoch": 55.23,
      "learning_rate": 1.2635857605177995e-05,
      "loss": 0.0395,
      "step": 142220
    },
    {
      "epoch": 55.23,
      "learning_rate": 1.2635339805825244e-05,
      "loss": 0.0004,
      "step": 142230
    },
    {
      "epoch": 55.24,
      "learning_rate": 1.2634822006472494e-05,
      "loss": 0.0472,
      "step": 142240
    },
    {
      "epoch": 55.24,
      "learning_rate": 1.2634304207119743e-05,
      "loss": 0.1354,
      "step": 142250
    },
    {
      "epoch": 55.25,
      "learning_rate": 1.263378640776699e-05,
      "loss": 0.0797,
      "step": 142260
    },
    {
      "epoch": 55.25,
      "learning_rate": 1.263326860841424e-05,
      "loss": 0.0769,
      "step": 142270
    },
    {
      "epoch": 55.25,
      "learning_rate": 1.2632750809061489e-05,
      "loss": 0.0836,
      "step": 142280
    },
    {
      "epoch": 55.26,
      "learning_rate": 1.2632233009708738e-05,
      "loss": 0.0599,
      "step": 142290
    },
    {
      "epoch": 55.26,
      "learning_rate": 1.2631715210355988e-05,
      "loss": 0.0379,
      "step": 142300
    },
    {
      "epoch": 55.27,
      "learning_rate": 1.2631197411003238e-05,
      "loss": 0.0316,
      "step": 142310
    },
    {
      "epoch": 55.27,
      "learning_rate": 1.2630679611650487e-05,
      "loss": 0.0511,
      "step": 142320
    },
    {
      "epoch": 55.27,
      "learning_rate": 1.2630161812297737e-05,
      "loss": 0.0691,
      "step": 142330
    },
    {
      "epoch": 55.28,
      "learning_rate": 1.2629644012944983e-05,
      "loss": 0.2658,
      "step": 142340
    },
    {
      "epoch": 55.28,
      "learning_rate": 1.2629126213592233e-05,
      "loss": 0.0397,
      "step": 142350
    },
    {
      "epoch": 55.29,
      "learning_rate": 1.2628608414239482e-05,
      "loss": 0.0961,
      "step": 142360
    },
    {
      "epoch": 55.29,
      "learning_rate": 1.2628090614886732e-05,
      "loss": 0.1,
      "step": 142370
    },
    {
      "epoch": 55.29,
      "learning_rate": 1.2627572815533982e-05,
      "loss": 0.0812,
      "step": 142380
    },
    {
      "epoch": 55.3,
      "learning_rate": 1.2627055016181231e-05,
      "loss": 0.172,
      "step": 142390
    },
    {
      "epoch": 55.3,
      "learning_rate": 1.2626537216828481e-05,
      "loss": 0.0806,
      "step": 142400
    },
    {
      "epoch": 55.3,
      "learning_rate": 1.262601941747573e-05,
      "loss": 0.0725,
      "step": 142410
    },
    {
      "epoch": 55.31,
      "learning_rate": 1.262550161812298e-05,
      "loss": 0.0079,
      "step": 142420
    },
    {
      "epoch": 55.31,
      "learning_rate": 1.2624983818770226e-05,
      "loss": 0.0249,
      "step": 142430
    },
    {
      "epoch": 55.32,
      "learning_rate": 1.2624466019417476e-05,
      "loss": 0.1251,
      "step": 142440
    },
    {
      "epoch": 55.32,
      "learning_rate": 1.2623948220064726e-05,
      "loss": 0.191,
      "step": 142450
    },
    {
      "epoch": 55.32,
      "learning_rate": 1.2623430420711975e-05,
      "loss": 0.0546,
      "step": 142460
    },
    {
      "epoch": 55.33,
      "learning_rate": 1.2622912621359225e-05,
      "loss": 0.0935,
      "step": 142470
    },
    {
      "epoch": 55.33,
      "learning_rate": 1.2622394822006474e-05,
      "loss": 0.1182,
      "step": 142480
    },
    {
      "epoch": 55.34,
      "learning_rate": 1.2621877022653724e-05,
      "loss": 0.1465,
      "step": 142490
    },
    {
      "epoch": 55.34,
      "learning_rate": 1.2621359223300974e-05,
      "loss": 0.0326,
      "step": 142500
    },
    {
      "epoch": 55.34,
      "learning_rate": 1.262084142394822e-05,
      "loss": 0.0882,
      "step": 142510
    },
    {
      "epoch": 55.35,
      "learning_rate": 1.262032362459547e-05,
      "loss": 0.0772,
      "step": 142520
    },
    {
      "epoch": 55.35,
      "learning_rate": 1.2619805825242719e-05,
      "loss": 0.0016,
      "step": 142530
    },
    {
      "epoch": 55.36,
      "learning_rate": 1.2619288025889969e-05,
      "loss": 0.0372,
      "step": 142540
    },
    {
      "epoch": 55.36,
      "learning_rate": 1.2618770226537218e-05,
      "loss": 0.2161,
      "step": 142550
    },
    {
      "epoch": 55.36,
      "learning_rate": 1.2618252427184468e-05,
      "loss": 0.0167,
      "step": 142560
    },
    {
      "epoch": 55.37,
      "learning_rate": 1.2617734627831718e-05,
      "loss": 0.0316,
      "step": 142570
    },
    {
      "epoch": 55.37,
      "learning_rate": 1.2617216828478967e-05,
      "loss": 0.1418,
      "step": 142580
    },
    {
      "epoch": 55.37,
      "learning_rate": 1.2616699029126213e-05,
      "loss": 0.1451,
      "step": 142590
    },
    {
      "epoch": 55.38,
      "learning_rate": 1.2616181229773463e-05,
      "loss": 0.2044,
      "step": 142600
    },
    {
      "epoch": 55.38,
      "learning_rate": 1.2615663430420713e-05,
      "loss": 0.022,
      "step": 142610
    },
    {
      "epoch": 55.39,
      "learning_rate": 1.2615145631067962e-05,
      "loss": 0.1969,
      "step": 142620
    },
    {
      "epoch": 55.39,
      "learning_rate": 1.2614627831715212e-05,
      "loss": 0.068,
      "step": 142630
    },
    {
      "epoch": 55.39,
      "learning_rate": 1.2614110032362462e-05,
      "loss": 0.076,
      "step": 142640
    },
    {
      "epoch": 55.4,
      "learning_rate": 1.2613592233009711e-05,
      "loss": 0.0839,
      "step": 142650
    },
    {
      "epoch": 55.4,
      "learning_rate": 1.261307443365696e-05,
      "loss": 0.1118,
      "step": 142660
    },
    {
      "epoch": 55.41,
      "learning_rate": 1.2612556634304207e-05,
      "loss": 0.0017,
      "step": 142670
    },
    {
      "epoch": 55.41,
      "learning_rate": 1.2612038834951457e-05,
      "loss": 0.0308,
      "step": 142680
    },
    {
      "epoch": 55.41,
      "learning_rate": 1.2611521035598706e-05,
      "loss": 0.0594,
      "step": 142690
    },
    {
      "epoch": 55.42,
      "learning_rate": 1.2611003236245956e-05,
      "loss": 0.0373,
      "step": 142700
    },
    {
      "epoch": 55.42,
      "learning_rate": 1.2610485436893205e-05,
      "loss": 0.0305,
      "step": 142710
    },
    {
      "epoch": 55.43,
      "learning_rate": 1.2609967637540455e-05,
      "loss": 0.0738,
      "step": 142720
    },
    {
      "epoch": 55.43,
      "learning_rate": 1.2609449838187705e-05,
      "loss": 0.0896,
      "step": 142730
    },
    {
      "epoch": 55.43,
      "learning_rate": 1.2608932038834953e-05,
      "loss": 0.0013,
      "step": 142740
    },
    {
      "epoch": 55.44,
      "learning_rate": 1.26084142394822e-05,
      "loss": 0.1317,
      "step": 142750
    },
    {
      "epoch": 55.44,
      "learning_rate": 1.260789644012945e-05,
      "loss": 0.0854,
      "step": 142760
    },
    {
      "epoch": 55.44,
      "learning_rate": 1.26073786407767e-05,
      "loss": 0.1681,
      "step": 142770
    },
    {
      "epoch": 55.45,
      "learning_rate": 1.260686084142395e-05,
      "loss": 0.1384,
      "step": 142780
    },
    {
      "epoch": 55.45,
      "learning_rate": 1.2606343042071199e-05,
      "loss": 0.0398,
      "step": 142790
    },
    {
      "epoch": 55.46,
      "learning_rate": 1.2605825242718449e-05,
      "loss": 0.0539,
      "step": 142800
    },
    {
      "epoch": 55.46,
      "learning_rate": 1.2605307443365698e-05,
      "loss": 0.0559,
      "step": 142810
    },
    {
      "epoch": 55.46,
      "learning_rate": 1.2604789644012946e-05,
      "loss": 0.098,
      "step": 142820
    },
    {
      "epoch": 55.47,
      "learning_rate": 1.2604271844660194e-05,
      "loss": 0.037,
      "step": 142830
    },
    {
      "epoch": 55.47,
      "learning_rate": 1.2603754045307444e-05,
      "loss": 0.3,
      "step": 142840
    },
    {
      "epoch": 55.48,
      "learning_rate": 1.2603236245954693e-05,
      "loss": 0.0654,
      "step": 142850
    },
    {
      "epoch": 55.48,
      "learning_rate": 1.2602718446601943e-05,
      "loss": 0.0864,
      "step": 142860
    },
    {
      "epoch": 55.48,
      "learning_rate": 1.2602200647249193e-05,
      "loss": 0.0465,
      "step": 142870
    },
    {
      "epoch": 55.49,
      "learning_rate": 1.2601682847896442e-05,
      "loss": 0.021,
      "step": 142880
    },
    {
      "epoch": 55.49,
      "learning_rate": 1.260116504854369e-05,
      "loss": 0.1219,
      "step": 142890
    },
    {
      "epoch": 55.5,
      "learning_rate": 1.260064724919094e-05,
      "loss": 0.0222,
      "step": 142900
    },
    {
      "epoch": 55.5,
      "learning_rate": 1.2600129449838188e-05,
      "loss": 0.2815,
      "step": 142910
    },
    {
      "epoch": 55.5,
      "learning_rate": 1.2599611650485437e-05,
      "loss": 0.0393,
      "step": 142920
    },
    {
      "epoch": 55.51,
      "learning_rate": 1.2599093851132687e-05,
      "loss": 0.134,
      "step": 142930
    },
    {
      "epoch": 55.51,
      "learning_rate": 1.2598576051779937e-05,
      "loss": 0.0007,
      "step": 142940
    },
    {
      "epoch": 55.51,
      "learning_rate": 1.2598058252427186e-05,
      "loss": 0.0808,
      "step": 142950
    },
    {
      "epoch": 55.52,
      "learning_rate": 1.2597540453074436e-05,
      "loss": 0.0899,
      "step": 142960
    },
    {
      "epoch": 55.52,
      "learning_rate": 1.2597022653721684e-05,
      "loss": 0.0457,
      "step": 142970
    },
    {
      "epoch": 55.53,
      "learning_rate": 1.2596504854368933e-05,
      "loss": 0.1032,
      "step": 142980
    },
    {
      "epoch": 55.53,
      "learning_rate": 1.2595987055016183e-05,
      "loss": 0.0363,
      "step": 142990
    },
    {
      "epoch": 55.53,
      "learning_rate": 1.259546925566343e-05,
      "loss": 0.0206,
      "step": 143000
    },
    {
      "epoch": 55.54,
      "learning_rate": 1.259495145631068e-05,
      "loss": 0.0031,
      "step": 143010
    },
    {
      "epoch": 55.54,
      "learning_rate": 1.259443365695793e-05,
      "loss": 0.0826,
      "step": 143020
    },
    {
      "epoch": 55.55,
      "learning_rate": 1.259391585760518e-05,
      "loss": 0.0495,
      "step": 143030
    },
    {
      "epoch": 55.55,
      "learning_rate": 1.2593398058252428e-05,
      "loss": 0.0423,
      "step": 143040
    },
    {
      "epoch": 55.55,
      "learning_rate": 1.2592880258899677e-05,
      "loss": 0.0232,
      "step": 143050
    },
    {
      "epoch": 55.56,
      "learning_rate": 1.2592362459546927e-05,
      "loss": 0.0047,
      "step": 143060
    },
    {
      "epoch": 55.56,
      "learning_rate": 1.2591844660194176e-05,
      "loss": 0.0665,
      "step": 143070
    },
    {
      "epoch": 55.57,
      "learning_rate": 1.2591326860841424e-05,
      "loss": 0.0061,
      "step": 143080
    },
    {
      "epoch": 55.57,
      "learning_rate": 1.2590809061488674e-05,
      "loss": 0.0218,
      "step": 143090
    },
    {
      "epoch": 55.57,
      "learning_rate": 1.2590291262135924e-05,
      "loss": 0.077,
      "step": 143100
    },
    {
      "epoch": 55.58,
      "learning_rate": 1.2589773462783173e-05,
      "loss": 0.0003,
      "step": 143110
    },
    {
      "epoch": 55.58,
      "learning_rate": 1.2589255663430421e-05,
      "loss": 0.2818,
      "step": 143120
    },
    {
      "epoch": 55.58,
      "learning_rate": 1.258873786407767e-05,
      "loss": 0.035,
      "step": 143130
    },
    {
      "epoch": 55.59,
      "learning_rate": 1.258822006472492e-05,
      "loss": 0.003,
      "step": 143140
    },
    {
      "epoch": 55.59,
      "learning_rate": 1.258770226537217e-05,
      "loss": 0.1882,
      "step": 143150
    },
    {
      "epoch": 55.6,
      "learning_rate": 1.2587184466019418e-05,
      "loss": 0.1245,
      "step": 143160
    },
    {
      "epoch": 55.6,
      "learning_rate": 1.2586666666666668e-05,
      "loss": 0.0603,
      "step": 143170
    },
    {
      "epoch": 55.6,
      "learning_rate": 1.2586148867313917e-05,
      "loss": 0.0704,
      "step": 143180
    },
    {
      "epoch": 55.61,
      "learning_rate": 1.2585631067961167e-05,
      "loss": 0.0881,
      "step": 143190
    },
    {
      "epoch": 55.61,
      "learning_rate": 1.2585113268608415e-05,
      "loss": 0.1678,
      "step": 143200
    },
    {
      "epoch": 55.62,
      "learning_rate": 1.2584595469255664e-05,
      "loss": 0.0404,
      "step": 143210
    },
    {
      "epoch": 55.62,
      "learning_rate": 1.2584077669902914e-05,
      "loss": 0.082,
      "step": 143220
    },
    {
      "epoch": 55.62,
      "learning_rate": 1.2583559870550164e-05,
      "loss": 0.0728,
      "step": 143230
    },
    {
      "epoch": 55.63,
      "learning_rate": 1.2583042071197412e-05,
      "loss": 0.1003,
      "step": 143240
    },
    {
      "epoch": 55.63,
      "learning_rate": 1.2582524271844661e-05,
      "loss": 0.028,
      "step": 143250
    },
    {
      "epoch": 55.63,
      "learning_rate": 1.258200647249191e-05,
      "loss": 0.0754,
      "step": 143260
    },
    {
      "epoch": 55.64,
      "learning_rate": 1.2581488673139159e-05,
      "loss": 0.1603,
      "step": 143270
    },
    {
      "epoch": 55.64,
      "learning_rate": 1.2580970873786408e-05,
      "loss": 0.1196,
      "step": 143280
    },
    {
      "epoch": 55.65,
      "learning_rate": 1.2580453074433658e-05,
      "loss": 0.0241,
      "step": 143290
    },
    {
      "epoch": 55.65,
      "learning_rate": 1.2579935275080908e-05,
      "loss": 0.0053,
      "step": 143300
    },
    {
      "epoch": 55.65,
      "learning_rate": 1.2579417475728157e-05,
      "loss": 0.0304,
      "step": 143310
    },
    {
      "epoch": 55.66,
      "learning_rate": 1.2578899676375405e-05,
      "loss": 0.2298,
      "step": 143320
    },
    {
      "epoch": 55.66,
      "learning_rate": 1.2578381877022655e-05,
      "loss": 0.1454,
      "step": 143330
    },
    {
      "epoch": 55.67,
      "learning_rate": 1.2577864077669904e-05,
      "loss": 0.1176,
      "step": 143340
    },
    {
      "epoch": 55.67,
      "learning_rate": 1.2577346278317152e-05,
      "loss": 0.0732,
      "step": 143350
    },
    {
      "epoch": 55.67,
      "learning_rate": 1.2576828478964402e-05,
      "loss": 0.1048,
      "step": 143360
    },
    {
      "epoch": 55.68,
      "learning_rate": 1.2576310679611651e-05,
      "loss": 0.0972,
      "step": 143370
    },
    {
      "epoch": 55.68,
      "learning_rate": 1.2575792880258901e-05,
      "loss": 0.0582,
      "step": 143380
    },
    {
      "epoch": 55.69,
      "learning_rate": 1.257527508090615e-05,
      "loss": 0.0973,
      "step": 143390
    },
    {
      "epoch": 55.69,
      "learning_rate": 1.2574757281553399e-05,
      "loss": 0.0533,
      "step": 143400
    },
    {
      "epoch": 55.69,
      "learning_rate": 1.2574239482200648e-05,
      "loss": 0.0311,
      "step": 143410
    },
    {
      "epoch": 55.7,
      "learning_rate": 1.2573721682847896e-05,
      "loss": 0.058,
      "step": 143420
    },
    {
      "epoch": 55.7,
      "learning_rate": 1.2573203883495146e-05,
      "loss": 0.0019,
      "step": 143430
    },
    {
      "epoch": 55.7,
      "learning_rate": 1.2572686084142395e-05,
      "loss": 0.1991,
      "step": 143440
    },
    {
      "epoch": 55.71,
      "learning_rate": 1.2572168284789645e-05,
      "loss": 0.0046,
      "step": 143450
    },
    {
      "epoch": 55.71,
      "learning_rate": 1.2571650485436895e-05,
      "loss": 0.0184,
      "step": 143460
    },
    {
      "epoch": 55.72,
      "learning_rate": 1.2571132686084144e-05,
      "loss": 0.0566,
      "step": 143470
    },
    {
      "epoch": 55.72,
      "learning_rate": 1.2570614886731392e-05,
      "loss": 0.0828,
      "step": 143480
    },
    {
      "epoch": 55.72,
      "learning_rate": 1.2570097087378642e-05,
      "loss": 0.0894,
      "step": 143490
    },
    {
      "epoch": 55.73,
      "learning_rate": 1.256957928802589e-05,
      "loss": 0.0206,
      "step": 143500
    },
    {
      "epoch": 55.73,
      "learning_rate": 1.256906148867314e-05,
      "loss": 0.018,
      "step": 143510
    },
    {
      "epoch": 55.74,
      "learning_rate": 1.2568543689320389e-05,
      "loss": 0.0895,
      "step": 143520
    },
    {
      "epoch": 55.74,
      "learning_rate": 1.2568025889967639e-05,
      "loss": 0.0143,
      "step": 143530
    },
    {
      "epoch": 55.74,
      "learning_rate": 1.2567508090614888e-05,
      "loss": 0.0649,
      "step": 143540
    },
    {
      "epoch": 55.75,
      "learning_rate": 1.2566990291262138e-05,
      "loss": 0.0847,
      "step": 143550
    },
    {
      "epoch": 55.75,
      "learning_rate": 1.2566472491909387e-05,
      "loss": 0.0232,
      "step": 143560
    },
    {
      "epoch": 55.76,
      "learning_rate": 1.2565954692556634e-05,
      "loss": 0.0274,
      "step": 143570
    },
    {
      "epoch": 55.76,
      "learning_rate": 1.2565436893203883e-05,
      "loss": 0.0137,
      "step": 143580
    },
    {
      "epoch": 55.76,
      "learning_rate": 1.2564919093851133e-05,
      "loss": 0.0215,
      "step": 143590
    },
    {
      "epoch": 55.77,
      "learning_rate": 1.2564401294498383e-05,
      "loss": 0.0162,
      "step": 143600
    },
    {
      "epoch": 55.77,
      "learning_rate": 1.2563883495145632e-05,
      "loss": 0.0606,
      "step": 143610
    },
    {
      "epoch": 55.77,
      "learning_rate": 1.2563365695792882e-05,
      "loss": 0.2037,
      "step": 143620
    },
    {
      "epoch": 55.78,
      "learning_rate": 1.2562847896440131e-05,
      "loss": 0.0794,
      "step": 143630
    },
    {
      "epoch": 55.78,
      "learning_rate": 1.2562330097087381e-05,
      "loss": 0.2215,
      "step": 143640
    },
    {
      "epoch": 55.79,
      "learning_rate": 1.2561812297734627e-05,
      "loss": 0.0536,
      "step": 143650
    },
    {
      "epoch": 55.79,
      "learning_rate": 1.2561294498381877e-05,
      "loss": 0.0517,
      "step": 143660
    },
    {
      "epoch": 55.79,
      "learning_rate": 1.2560776699029126e-05,
      "loss": 0.0543,
      "step": 143670
    },
    {
      "epoch": 55.8,
      "learning_rate": 1.2560258899676376e-05,
      "loss": 0.0237,
      "step": 143680
    },
    {
      "epoch": 55.8,
      "learning_rate": 1.2559741100323626e-05,
      "loss": 0.0298,
      "step": 143690
    },
    {
      "epoch": 55.81,
      "learning_rate": 1.2559223300970875e-05,
      "loss": 0.0414,
      "step": 143700
    },
    {
      "epoch": 55.81,
      "learning_rate": 1.2558705501618125e-05,
      "loss": 0.0625,
      "step": 143710
    },
    {
      "epoch": 55.81,
      "learning_rate": 1.2558187702265375e-05,
      "loss": 0.1638,
      "step": 143720
    },
    {
      "epoch": 55.82,
      "learning_rate": 1.255766990291262e-05,
      "loss": 0.0212,
      "step": 143730
    },
    {
      "epoch": 55.82,
      "learning_rate": 1.255715210355987e-05,
      "loss": 0.0752,
      "step": 143740
    },
    {
      "epoch": 55.83,
      "learning_rate": 1.255663430420712e-05,
      "loss": 0.1783,
      "step": 143750
    },
    {
      "epoch": 55.83,
      "learning_rate": 1.255611650485437e-05,
      "loss": 0.1245,
      "step": 143760
    },
    {
      "epoch": 55.83,
      "learning_rate": 1.255559870550162e-05,
      "loss": 0.2137,
      "step": 143770
    },
    {
      "epoch": 55.84,
      "learning_rate": 1.2555080906148869e-05,
      "loss": 0.0915,
      "step": 143780
    },
    {
      "epoch": 55.84,
      "learning_rate": 1.2554563106796118e-05,
      "loss": 0.0804,
      "step": 143790
    },
    {
      "epoch": 55.84,
      "learning_rate": 1.2554045307443368e-05,
      "loss": 0.0305,
      "step": 143800
    },
    {
      "epoch": 55.85,
      "learning_rate": 1.2553527508090614e-05,
      "loss": 0.0047,
      "step": 143810
    },
    {
      "epoch": 55.85,
      "learning_rate": 1.2553009708737864e-05,
      "loss": 0.1661,
      "step": 143820
    },
    {
      "epoch": 55.86,
      "learning_rate": 1.2552491909385114e-05,
      "loss": 0.0009,
      "step": 143830
    },
    {
      "epoch": 55.86,
      "learning_rate": 1.2551974110032363e-05,
      "loss": 0.0589,
      "step": 143840
    },
    {
      "epoch": 55.86,
      "learning_rate": 1.2551456310679613e-05,
      "loss": 0.0319,
      "step": 143850
    },
    {
      "epoch": 55.87,
      "learning_rate": 1.2550938511326862e-05,
      "loss": 0.0194,
      "step": 143860
    },
    {
      "epoch": 55.87,
      "learning_rate": 1.2550420711974112e-05,
      "loss": 0.0272,
      "step": 143870
    },
    {
      "epoch": 55.88,
      "learning_rate": 1.2549902912621362e-05,
      "loss": 0.0946,
      "step": 143880
    },
    {
      "epoch": 55.88,
      "learning_rate": 1.2549385113268608e-05,
      "loss": 0.1634,
      "step": 143890
    },
    {
      "epoch": 55.88,
      "learning_rate": 1.2548867313915857e-05,
      "loss": 0.007,
      "step": 143900
    },
    {
      "epoch": 55.89,
      "learning_rate": 1.2548349514563107e-05,
      "loss": 0.0028,
      "step": 143910
    },
    {
      "epoch": 55.89,
      "learning_rate": 1.2547831715210357e-05,
      "loss": 0.0465,
      "step": 143920
    },
    {
      "epoch": 55.9,
      "learning_rate": 1.2547313915857606e-05,
      "loss": 0.1184,
      "step": 143930
    },
    {
      "epoch": 55.9,
      "learning_rate": 1.2546796116504856e-05,
      "loss": 0.0128,
      "step": 143940
    },
    {
      "epoch": 55.9,
      "learning_rate": 1.2546278317152106e-05,
      "loss": 0.0457,
      "step": 143950
    },
    {
      "epoch": 55.91,
      "learning_rate": 1.2545760517799355e-05,
      "loss": 0.1039,
      "step": 143960
    },
    {
      "epoch": 55.91,
      "learning_rate": 1.2545242718446601e-05,
      "loss": 0.044,
      "step": 143970
    },
    {
      "epoch": 55.91,
      "learning_rate": 1.2544724919093851e-05,
      "loss": 0.0427,
      "step": 143980
    },
    {
      "epoch": 55.92,
      "learning_rate": 1.25442071197411e-05,
      "loss": 0.0011,
      "step": 143990
    },
    {
      "epoch": 55.92,
      "learning_rate": 1.254368932038835e-05,
      "loss": 0.114,
      "step": 144000
    },
    {
      "epoch": 55.93,
      "learning_rate": 1.25431715210356e-05,
      "loss": 0.1906,
      "step": 144010
    },
    {
      "epoch": 55.93,
      "learning_rate": 1.254265372168285e-05,
      "loss": 0.2447,
      "step": 144020
    },
    {
      "epoch": 55.93,
      "learning_rate": 1.2542135922330099e-05,
      "loss": 0.068,
      "step": 144030
    },
    {
      "epoch": 55.94,
      "learning_rate": 1.2541618122977349e-05,
      "loss": 0.0058,
      "step": 144040
    },
    {
      "epoch": 55.94,
      "learning_rate": 1.2541100323624595e-05,
      "loss": 0.0769,
      "step": 144050
    },
    {
      "epoch": 55.95,
      "learning_rate": 1.2540582524271845e-05,
      "loss": 0.0694,
      "step": 144060
    },
    {
      "epoch": 55.95,
      "learning_rate": 1.2540064724919094e-05,
      "loss": 0.0312,
      "step": 144070
    },
    {
      "epoch": 55.95,
      "learning_rate": 1.2539546925566344e-05,
      "loss": 0.0215,
      "step": 144080
    },
    {
      "epoch": 55.96,
      "learning_rate": 1.2539029126213593e-05,
      "loss": 0.0828,
      "step": 144090
    },
    {
      "epoch": 55.96,
      "learning_rate": 1.2538511326860843e-05,
      "loss": 0.0159,
      "step": 144100
    },
    {
      "epoch": 55.97,
      "learning_rate": 1.2537993527508093e-05,
      "loss": 0.0943,
      "step": 144110
    },
    {
      "epoch": 55.97,
      "learning_rate": 1.2537475728155342e-05,
      "loss": 0.0696,
      "step": 144120
    },
    {
      "epoch": 55.97,
      "learning_rate": 1.2536957928802592e-05,
      "loss": 0.0791,
      "step": 144130
    },
    {
      "epoch": 55.98,
      "learning_rate": 1.2536440129449838e-05,
      "loss": 0.0547,
      "step": 144140
    },
    {
      "epoch": 55.98,
      "learning_rate": 1.2535922330097088e-05,
      "loss": 0.025,
      "step": 144150
    },
    {
      "epoch": 55.98,
      "learning_rate": 1.2535404530744337e-05,
      "loss": 0.2083,
      "step": 144160
    },
    {
      "epoch": 55.99,
      "learning_rate": 1.2534886731391587e-05,
      "loss": 0.0352,
      "step": 144170
    },
    {
      "epoch": 55.99,
      "learning_rate": 1.2534368932038837e-05,
      "loss": 0.0056,
      "step": 144180
    },
    {
      "epoch": 56.0,
      "learning_rate": 1.2533851132686086e-05,
      "loss": 0.0707,
      "step": 144190
    },
    {
      "epoch": 56.0,
      "learning_rate": 1.2533333333333336e-05,
      "loss": 0.013,
      "step": 144200
    },
    {
      "epoch": 56.0,
      "eval_accuracy": 0.9474552957359009,
      "eval_loss": 0.33862391114234924,
      "eval_runtime": 8.2896,
      "eval_samples_per_second": 438.504,
      "eval_steps_per_second": 54.888,
      "step": 144200
    },
    {
      "epoch": 56.0,
      "learning_rate": 1.2532815533980584e-05,
      "loss": 0.1405,
      "step": 144210
    },
    {
      "epoch": 56.01,
      "learning_rate": 1.2532297734627832e-05,
      "loss": 0.1296,
      "step": 144220
    },
    {
      "epoch": 56.01,
      "learning_rate": 1.2531779935275081e-05,
      "loss": 0.0601,
      "step": 144230
    },
    {
      "epoch": 56.02,
      "learning_rate": 1.2531262135922331e-05,
      "loss": 0.0043,
      "step": 144240
    },
    {
      "epoch": 56.02,
      "learning_rate": 1.253074433656958e-05,
      "loss": 0.0137,
      "step": 144250
    },
    {
      "epoch": 56.02,
      "learning_rate": 1.253022653721683e-05,
      "loss": 0.0477,
      "step": 144260
    },
    {
      "epoch": 56.03,
      "learning_rate": 1.252970873786408e-05,
      "loss": 0.0662,
      "step": 144270
    },
    {
      "epoch": 56.03,
      "learning_rate": 1.252919093851133e-05,
      "loss": 0.063,
      "step": 144280
    },
    {
      "epoch": 56.03,
      "learning_rate": 1.2528673139158577e-05,
      "loss": 0.0343,
      "step": 144290
    },
    {
      "epoch": 56.04,
      "learning_rate": 1.2528155339805825e-05,
      "loss": 0.2374,
      "step": 144300
    },
    {
      "epoch": 56.04,
      "learning_rate": 1.2527637540453075e-05,
      "loss": 0.0435,
      "step": 144310
    },
    {
      "epoch": 56.05,
      "learning_rate": 1.2527119741100325e-05,
      "loss": 0.0995,
      "step": 144320
    },
    {
      "epoch": 56.05,
      "learning_rate": 1.2526601941747574e-05,
      "loss": 0.0177,
      "step": 144330
    },
    {
      "epoch": 56.05,
      "learning_rate": 1.2526084142394824e-05,
      "loss": 0.1354,
      "step": 144340
    },
    {
      "epoch": 56.06,
      "learning_rate": 1.2525566343042073e-05,
      "loss": 0.0589,
      "step": 144350
    },
    {
      "epoch": 56.06,
      "learning_rate": 1.2525048543689321e-05,
      "loss": 0.0906,
      "step": 144360
    },
    {
      "epoch": 56.07,
      "learning_rate": 1.2524530744336571e-05,
      "loss": 0.0372,
      "step": 144370
    },
    {
      "epoch": 56.07,
      "learning_rate": 1.2524012944983819e-05,
      "loss": 0.0242,
      "step": 144380
    },
    {
      "epoch": 56.07,
      "learning_rate": 1.2523495145631068e-05,
      "loss": 0.0014,
      "step": 144390
    },
    {
      "epoch": 56.08,
      "learning_rate": 1.2522977346278318e-05,
      "loss": 0.0441,
      "step": 144400
    },
    {
      "epoch": 56.08,
      "learning_rate": 1.2522459546925568e-05,
      "loss": 0.0789,
      "step": 144410
    },
    {
      "epoch": 56.09,
      "learning_rate": 1.2521941747572817e-05,
      "loss": 0.0829,
      "step": 144420
    },
    {
      "epoch": 56.09,
      "learning_rate": 1.2521423948220067e-05,
      "loss": 0.1228,
      "step": 144430
    },
    {
      "epoch": 56.09,
      "learning_rate": 1.2520906148867315e-05,
      "loss": 0.0412,
      "step": 144440
    },
    {
      "epoch": 56.1,
      "learning_rate": 1.2520388349514564e-05,
      "loss": 0.0311,
      "step": 144450
    },
    {
      "epoch": 56.1,
      "learning_rate": 1.2519870550161812e-05,
      "loss": 0.2098,
      "step": 144460
    },
    {
      "epoch": 56.1,
      "learning_rate": 1.2519352750809062e-05,
      "loss": 0.1536,
      "step": 144470
    },
    {
      "epoch": 56.11,
      "learning_rate": 1.2518834951456312e-05,
      "loss": 0.0453,
      "step": 144480
    },
    {
      "epoch": 56.11,
      "learning_rate": 1.2518317152103561e-05,
      "loss": 0.0191,
      "step": 144490
    },
    {
      "epoch": 56.12,
      "learning_rate": 1.2517799352750811e-05,
      "loss": 0.098,
      "step": 144500
    },
    {
      "epoch": 56.12,
      "learning_rate": 1.2517281553398059e-05,
      "loss": 0.1214,
      "step": 144510
    },
    {
      "epoch": 56.12,
      "learning_rate": 1.2516763754045308e-05,
      "loss": 0.0882,
      "step": 144520
    },
    {
      "epoch": 56.13,
      "learning_rate": 1.2516245954692558e-05,
      "loss": 0.0774,
      "step": 144530
    },
    {
      "epoch": 56.13,
      "learning_rate": 1.2515728155339806e-05,
      "loss": 0.1478,
      "step": 144540
    },
    {
      "epoch": 56.14,
      "learning_rate": 1.2515210355987056e-05,
      "loss": 0.0037,
      "step": 144550
    },
    {
      "epoch": 56.14,
      "learning_rate": 1.2514692556634305e-05,
      "loss": 0.0398,
      "step": 144560
    },
    {
      "epoch": 56.14,
      "learning_rate": 1.2514174757281555e-05,
      "loss": 0.0476,
      "step": 144570
    },
    {
      "epoch": 56.15,
      "learning_rate": 1.2513656957928804e-05,
      "loss": 0.0338,
      "step": 144580
    },
    {
      "epoch": 56.15,
      "learning_rate": 1.2513139158576052e-05,
      "loss": 0.0517,
      "step": 144590
    },
    {
      "epoch": 56.16,
      "learning_rate": 1.2512621359223302e-05,
      "loss": 0.0303,
      "step": 144600
    },
    {
      "epoch": 56.16,
      "learning_rate": 1.2512103559870552e-05,
      "loss": 0.0328,
      "step": 144610
    },
    {
      "epoch": 56.16,
      "learning_rate": 1.25115857605178e-05,
      "loss": 0.0506,
      "step": 144620
    },
    {
      "epoch": 56.17,
      "learning_rate": 1.2511067961165049e-05,
      "loss": 0.0563,
      "step": 144630
    },
    {
      "epoch": 56.17,
      "learning_rate": 1.2510550161812299e-05,
      "loss": 0.0519,
      "step": 144640
    },
    {
      "epoch": 56.17,
      "learning_rate": 1.2510032362459548e-05,
      "loss": 0.1511,
      "step": 144650
    },
    {
      "epoch": 56.18,
      "learning_rate": 1.2509514563106798e-05,
      "loss": 0.1289,
      "step": 144660
    },
    {
      "epoch": 56.18,
      "learning_rate": 1.2508996763754046e-05,
      "loss": 0.001,
      "step": 144670
    },
    {
      "epoch": 56.19,
      "learning_rate": 1.2508478964401296e-05,
      "loss": 0.0323,
      "step": 144680
    },
    {
      "epoch": 56.19,
      "learning_rate": 1.2507961165048545e-05,
      "loss": 0.042,
      "step": 144690
    },
    {
      "epoch": 56.19,
      "learning_rate": 1.2507443365695795e-05,
      "loss": 0.0226,
      "step": 144700
    },
    {
      "epoch": 56.2,
      "learning_rate": 1.2506925566343043e-05,
      "loss": 0.001,
      "step": 144710
    },
    {
      "epoch": 56.2,
      "learning_rate": 1.2506407766990292e-05,
      "loss": 0.0659,
      "step": 144720
    },
    {
      "epoch": 56.21,
      "learning_rate": 1.2505889967637542e-05,
      "loss": 0.1402,
      "step": 144730
    },
    {
      "epoch": 56.21,
      "learning_rate": 1.250537216828479e-05,
      "loss": 0.1722,
      "step": 144740
    },
    {
      "epoch": 56.21,
      "learning_rate": 1.250485436893204e-05,
      "loss": 0.1299,
      "step": 144750
    },
    {
      "epoch": 56.22,
      "learning_rate": 1.2504336569579289e-05,
      "loss": 0.1568,
      "step": 144760
    },
    {
      "epoch": 56.22,
      "learning_rate": 1.2503818770226539e-05,
      "loss": 0.0164,
      "step": 144770
    },
    {
      "epoch": 56.23,
      "learning_rate": 1.2503300970873788e-05,
      "loss": 0.0609,
      "step": 144780
    },
    {
      "epoch": 56.23,
      "learning_rate": 1.2502783171521036e-05,
      "loss": 0.0121,
      "step": 144790
    },
    {
      "epoch": 56.23,
      "learning_rate": 1.2502265372168286e-05,
      "loss": 0.1821,
      "step": 144800
    },
    {
      "epoch": 56.24,
      "learning_rate": 1.2501747572815535e-05,
      "loss": 0.0383,
      "step": 144810
    },
    {
      "epoch": 56.24,
      "learning_rate": 1.2501229773462783e-05,
      "loss": 0.0706,
      "step": 144820
    },
    {
      "epoch": 56.24,
      "learning_rate": 1.2500711974110033e-05,
      "loss": 0.0097,
      "step": 144830
    },
    {
      "epoch": 56.25,
      "learning_rate": 1.2500194174757283e-05,
      "loss": 0.1106,
      "step": 144840
    },
    {
      "epoch": 56.25,
      "learning_rate": 1.2499676375404532e-05,
      "loss": 0.0537,
      "step": 144850
    },
    {
      "epoch": 56.26,
      "learning_rate": 1.2499158576051782e-05,
      "loss": 0.0929,
      "step": 144860
    },
    {
      "epoch": 56.26,
      "learning_rate": 1.249864077669903e-05,
      "loss": 0.0985,
      "step": 144870
    },
    {
      "epoch": 56.26,
      "learning_rate": 1.249812297734628e-05,
      "loss": 0.0469,
      "step": 144880
    },
    {
      "epoch": 56.27,
      "learning_rate": 1.2497605177993527e-05,
      "loss": 0.0009,
      "step": 144890
    },
    {
      "epoch": 56.27,
      "learning_rate": 1.2497087378640777e-05,
      "loss": 0.1414,
      "step": 144900
    },
    {
      "epoch": 56.28,
      "learning_rate": 1.2496569579288027e-05,
      "loss": 0.1001,
      "step": 144910
    },
    {
      "epoch": 56.28,
      "learning_rate": 1.2496051779935276e-05,
      "loss": 0.0518,
      "step": 144920
    },
    {
      "epoch": 56.28,
      "learning_rate": 1.2495533980582526e-05,
      "loss": 0.0066,
      "step": 144930
    },
    {
      "epoch": 56.29,
      "learning_rate": 1.2495016181229775e-05,
      "loss": 0.0523,
      "step": 144940
    },
    {
      "epoch": 56.29,
      "learning_rate": 1.2494498381877023e-05,
      "loss": 0.1074,
      "step": 144950
    },
    {
      "epoch": 56.3,
      "learning_rate": 1.2493980582524273e-05,
      "loss": 0.1033,
      "step": 144960
    },
    {
      "epoch": 56.3,
      "learning_rate": 1.2493462783171521e-05,
      "loss": 0.0064,
      "step": 144970
    },
    {
      "epoch": 56.3,
      "learning_rate": 1.249294498381877e-05,
      "loss": 0.1705,
      "step": 144980
    },
    {
      "epoch": 56.31,
      "learning_rate": 1.249242718446602e-05,
      "loss": 0.1462,
      "step": 144990
    },
    {
      "epoch": 56.31,
      "learning_rate": 1.249190938511327e-05,
      "loss": 0.0427,
      "step": 145000
    },
    {
      "epoch": 56.31,
      "learning_rate": 1.249139158576052e-05,
      "loss": 0.0257,
      "step": 145010
    },
    {
      "epoch": 56.32,
      "learning_rate": 1.2490873786407769e-05,
      "loss": 0.0991,
      "step": 145020
    },
    {
      "epoch": 56.32,
      "learning_rate": 1.2490355987055017e-05,
      "loss": 0.0151,
      "step": 145030
    },
    {
      "epoch": 56.33,
      "learning_rate": 1.2489838187702265e-05,
      "loss": 0.0776,
      "step": 145040
    },
    {
      "epoch": 56.33,
      "learning_rate": 1.2489320388349514e-05,
      "loss": 0.1153,
      "step": 145050
    },
    {
      "epoch": 56.33,
      "learning_rate": 1.2488802588996764e-05,
      "loss": 0.0734,
      "step": 145060
    },
    {
      "epoch": 56.34,
      "learning_rate": 1.2488284789644014e-05,
      "loss": 0.0585,
      "step": 145070
    },
    {
      "epoch": 56.34,
      "learning_rate": 1.2487766990291263e-05,
      "loss": 0.0278,
      "step": 145080
    },
    {
      "epoch": 56.35,
      "learning_rate": 1.2487249190938513e-05,
      "loss": 0.0529,
      "step": 145090
    },
    {
      "epoch": 56.35,
      "learning_rate": 1.2486731391585763e-05,
      "loss": 0.1558,
      "step": 145100
    },
    {
      "epoch": 56.35,
      "learning_rate": 1.248621359223301e-05,
      "loss": 0.1343,
      "step": 145110
    },
    {
      "epoch": 56.36,
      "learning_rate": 1.2485695792880258e-05,
      "loss": 0.023,
      "step": 145120
    },
    {
      "epoch": 56.36,
      "learning_rate": 1.2485177993527508e-05,
      "loss": 0.0238,
      "step": 145130
    },
    {
      "epoch": 56.37,
      "learning_rate": 1.2484660194174758e-05,
      "loss": 0.1609,
      "step": 145140
    },
    {
      "epoch": 56.37,
      "learning_rate": 1.2484142394822007e-05,
      "loss": 0.1522,
      "step": 145150
    },
    {
      "epoch": 56.37,
      "learning_rate": 1.2483624595469257e-05,
      "loss": 0.0528,
      "step": 145160
    },
    {
      "epoch": 56.38,
      "learning_rate": 1.2483106796116506e-05,
      "loss": 0.0823,
      "step": 145170
    },
    {
      "epoch": 56.38,
      "learning_rate": 1.2482588996763756e-05,
      "loss": 0.1223,
      "step": 145180
    },
    {
      "epoch": 56.38,
      "learning_rate": 1.2482071197411002e-05,
      "loss": 0.2386,
      "step": 145190
    },
    {
      "epoch": 56.39,
      "learning_rate": 1.2481553398058252e-05,
      "loss": 0.1902,
      "step": 145200
    },
    {
      "epoch": 56.39,
      "learning_rate": 1.2481035598705502e-05,
      "loss": 0.0928,
      "step": 145210
    },
    {
      "epoch": 56.4,
      "learning_rate": 1.2480517799352751e-05,
      "loss": 0.0782,
      "step": 145220
    },
    {
      "epoch": 56.4,
      "learning_rate": 1.248e-05,
      "loss": 0.0974,
      "step": 145230
    },
    {
      "epoch": 56.4,
      "learning_rate": 1.247948220064725e-05,
      "loss": 0.1363,
      "step": 145240
    },
    {
      "epoch": 56.41,
      "learning_rate": 1.24789644012945e-05,
      "loss": 0.0192,
      "step": 145250
    },
    {
      "epoch": 56.41,
      "learning_rate": 1.247844660194175e-05,
      "loss": 0.05,
      "step": 145260
    },
    {
      "epoch": 56.42,
      "learning_rate": 1.2477928802589e-05,
      "loss": 0.0335,
      "step": 145270
    },
    {
      "epoch": 56.42,
      "learning_rate": 1.2477411003236245e-05,
      "loss": 0.077,
      "step": 145280
    },
    {
      "epoch": 56.42,
      "learning_rate": 1.2476893203883495e-05,
      "loss": 0.0285,
      "step": 145290
    },
    {
      "epoch": 56.43,
      "learning_rate": 1.2476375404530745e-05,
      "loss": 0.1448,
      "step": 145300
    },
    {
      "epoch": 56.43,
      "learning_rate": 1.2475857605177994e-05,
      "loss": 0.0026,
      "step": 145310
    },
    {
      "epoch": 56.43,
      "learning_rate": 1.2475339805825244e-05,
      "loss": 0.0894,
      "step": 145320
    },
    {
      "epoch": 56.44,
      "learning_rate": 1.2474822006472494e-05,
      "loss": 0.125,
      "step": 145330
    },
    {
      "epoch": 56.44,
      "learning_rate": 1.2474304207119743e-05,
      "loss": 0.065,
      "step": 145340
    },
    {
      "epoch": 56.45,
      "learning_rate": 1.2473786407766993e-05,
      "loss": 0.0414,
      "step": 145350
    },
    {
      "epoch": 56.45,
      "learning_rate": 1.2473268608414239e-05,
      "loss": 0.0133,
      "step": 145360
    },
    {
      "epoch": 56.45,
      "learning_rate": 1.2472750809061489e-05,
      "loss": 0.1055,
      "step": 145370
    },
    {
      "epoch": 56.46,
      "learning_rate": 1.2472233009708738e-05,
      "loss": 0.0154,
      "step": 145380
    },
    {
      "epoch": 56.46,
      "learning_rate": 1.2471715210355988e-05,
      "loss": 0.0679,
      "step": 145390
    },
    {
      "epoch": 56.47,
      "learning_rate": 1.2471197411003238e-05,
      "loss": 0.0034,
      "step": 145400
    },
    {
      "epoch": 56.47,
      "learning_rate": 1.2470679611650487e-05,
      "loss": 0.0218,
      "step": 145410
    },
    {
      "epoch": 56.47,
      "learning_rate": 1.2470161812297737e-05,
      "loss": 0.2282,
      "step": 145420
    },
    {
      "epoch": 56.48,
      "learning_rate": 1.2469644012944986e-05,
      "loss": 0.1077,
      "step": 145430
    },
    {
      "epoch": 56.48,
      "learning_rate": 1.2469126213592233e-05,
      "loss": 0.0683,
      "step": 145440
    },
    {
      "epoch": 56.49,
      "learning_rate": 1.2468608414239482e-05,
      "loss": 0.112,
      "step": 145450
    },
    {
      "epoch": 56.49,
      "learning_rate": 1.2468090614886732e-05,
      "loss": 0.0016,
      "step": 145460
    },
    {
      "epoch": 56.49,
      "learning_rate": 1.2467572815533981e-05,
      "loss": 0.0407,
      "step": 145470
    },
    {
      "epoch": 56.5,
      "learning_rate": 1.2467055016181231e-05,
      "loss": 0.0764,
      "step": 145480
    },
    {
      "epoch": 56.5,
      "learning_rate": 1.246653721682848e-05,
      "loss": 0.0518,
      "step": 145490
    },
    {
      "epoch": 56.5,
      "learning_rate": 1.246601941747573e-05,
      "loss": 0.0757,
      "step": 145500
    },
    {
      "epoch": 56.51,
      "learning_rate": 1.246550161812298e-05,
      "loss": 0.1609,
      "step": 145510
    },
    {
      "epoch": 56.51,
      "learning_rate": 1.2464983818770226e-05,
      "loss": 0.0666,
      "step": 145520
    },
    {
      "epoch": 56.52,
      "learning_rate": 1.2464466019417476e-05,
      "loss": 0.1055,
      "step": 145530
    },
    {
      "epoch": 56.52,
      "learning_rate": 1.2463948220064725e-05,
      "loss": 0.0554,
      "step": 145540
    },
    {
      "epoch": 56.52,
      "learning_rate": 1.2463430420711975e-05,
      "loss": 0.0167,
      "step": 145550
    },
    {
      "epoch": 56.53,
      "learning_rate": 1.2462912621359225e-05,
      "loss": 0.0073,
      "step": 145560
    },
    {
      "epoch": 56.53,
      "learning_rate": 1.2462394822006474e-05,
      "loss": 0.1386,
      "step": 145570
    },
    {
      "epoch": 56.54,
      "learning_rate": 1.2461877022653724e-05,
      "loss": 0.1172,
      "step": 145580
    },
    {
      "epoch": 56.54,
      "learning_rate": 1.2461359223300973e-05,
      "loss": 0.0151,
      "step": 145590
    },
    {
      "epoch": 56.54,
      "learning_rate": 1.246084142394822e-05,
      "loss": 0.1412,
      "step": 145600
    },
    {
      "epoch": 56.55,
      "learning_rate": 1.246032362459547e-05,
      "loss": 0.005,
      "step": 145610
    },
    {
      "epoch": 56.55,
      "learning_rate": 1.2459805825242719e-05,
      "loss": 0.0171,
      "step": 145620
    },
    {
      "epoch": 56.56,
      "learning_rate": 1.2459288025889969e-05,
      "loss": 0.0075,
      "step": 145630
    },
    {
      "epoch": 56.56,
      "learning_rate": 1.2458770226537218e-05,
      "loss": 0.1801,
      "step": 145640
    },
    {
      "epoch": 56.56,
      "learning_rate": 1.2458252427184468e-05,
      "loss": 0.097,
      "step": 145650
    },
    {
      "epoch": 56.57,
      "learning_rate": 1.2457734627831717e-05,
      "loss": 0.0083,
      "step": 145660
    },
    {
      "epoch": 56.57,
      "learning_rate": 1.2457216828478967e-05,
      "loss": 0.0776,
      "step": 145670
    },
    {
      "epoch": 56.57,
      "learning_rate": 1.2456699029126213e-05,
      "loss": 0.0644,
      "step": 145680
    },
    {
      "epoch": 56.58,
      "learning_rate": 1.2456181229773463e-05,
      "loss": 0.0409,
      "step": 145690
    },
    {
      "epoch": 56.58,
      "learning_rate": 1.2455663430420713e-05,
      "loss": 0.0985,
      "step": 145700
    },
    {
      "epoch": 56.59,
      "learning_rate": 1.2455145631067962e-05,
      "loss": 0.1275,
      "step": 145710
    },
    {
      "epoch": 56.59,
      "learning_rate": 1.2454627831715212e-05,
      "loss": 0.1726,
      "step": 145720
    },
    {
      "epoch": 56.59,
      "learning_rate": 1.2454110032362461e-05,
      "loss": 0.0899,
      "step": 145730
    },
    {
      "epoch": 56.6,
      "learning_rate": 1.2453592233009711e-05,
      "loss": 0.1321,
      "step": 145740
    },
    {
      "epoch": 56.6,
      "learning_rate": 1.245307443365696e-05,
      "loss": 0.0628,
      "step": 145750
    },
    {
      "epoch": 56.61,
      "learning_rate": 1.2452556634304209e-05,
      "loss": 0.0192,
      "step": 145760
    },
    {
      "epoch": 56.61,
      "learning_rate": 1.2452038834951456e-05,
      "loss": 0.0215,
      "step": 145770
    },
    {
      "epoch": 56.61,
      "learning_rate": 1.2451521035598706e-05,
      "loss": 0.0792,
      "step": 145780
    },
    {
      "epoch": 56.62,
      "learning_rate": 1.2451003236245956e-05,
      "loss": 0.0886,
      "step": 145790
    },
    {
      "epoch": 56.62,
      "learning_rate": 1.2450485436893205e-05,
      "loss": 0.0013,
      "step": 145800
    },
    {
      "epoch": 56.63,
      "learning_rate": 1.2449967637540455e-05,
      "loss": 0.146,
      "step": 145810
    },
    {
      "epoch": 56.63,
      "learning_rate": 1.2449449838187705e-05,
      "loss": 0.025,
      "step": 145820
    },
    {
      "epoch": 56.63,
      "learning_rate": 1.2448932038834952e-05,
      "loss": 0.0378,
      "step": 145830
    },
    {
      "epoch": 56.64,
      "learning_rate": 1.2448414239482202e-05,
      "loss": 0.0779,
      "step": 145840
    },
    {
      "epoch": 56.64,
      "learning_rate": 1.244789644012945e-05,
      "loss": 0.0331,
      "step": 145850
    },
    {
      "epoch": 56.64,
      "learning_rate": 1.24473786407767e-05,
      "loss": 0.0143,
      "step": 145860
    },
    {
      "epoch": 56.65,
      "learning_rate": 1.244686084142395e-05,
      "loss": 0.1706,
      "step": 145870
    },
    {
      "epoch": 56.65,
      "learning_rate": 1.2446343042071199e-05,
      "loss": 0.1078,
      "step": 145880
    },
    {
      "epoch": 56.66,
      "learning_rate": 1.2445825242718448e-05,
      "loss": 0.0256,
      "step": 145890
    },
    {
      "epoch": 56.66,
      "learning_rate": 1.2445307443365698e-05,
      "loss": 0.0561,
      "step": 145900
    },
    {
      "epoch": 56.66,
      "learning_rate": 1.2444789644012946e-05,
      "loss": 0.0901,
      "step": 145910
    },
    {
      "epoch": 56.67,
      "learning_rate": 1.2444271844660196e-05,
      "loss": 0.0444,
      "step": 145920
    },
    {
      "epoch": 56.67,
      "learning_rate": 1.2443754045307444e-05,
      "loss": 0.0166,
      "step": 145930
    },
    {
      "epoch": 56.68,
      "learning_rate": 1.2443236245954693e-05,
      "loss": 0.0555,
      "step": 145940
    },
    {
      "epoch": 56.68,
      "learning_rate": 1.2442718446601943e-05,
      "loss": 0.1718,
      "step": 145950
    },
    {
      "epoch": 56.68,
      "learning_rate": 1.2442200647249192e-05,
      "loss": 0.0022,
      "step": 145960
    },
    {
      "epoch": 56.69,
      "learning_rate": 1.2441682847896442e-05,
      "loss": 0.1049,
      "step": 145970
    },
    {
      "epoch": 56.69,
      "learning_rate": 1.244116504854369e-05,
      "loss": 0.074,
      "step": 145980
    },
    {
      "epoch": 56.7,
      "learning_rate": 1.244064724919094e-05,
      "loss": 0.1199,
      "step": 145990
    },
    {
      "epoch": 56.7,
      "learning_rate": 1.244012944983819e-05,
      "loss": 0.0177,
      "step": 146000
    },
    {
      "epoch": 56.7,
      "learning_rate": 1.2439611650485437e-05,
      "loss": 0.0922,
      "step": 146010
    },
    {
      "epoch": 56.71,
      "learning_rate": 1.2439093851132687e-05,
      "loss": 0.1021,
      "step": 146020
    },
    {
      "epoch": 56.71,
      "learning_rate": 1.2438576051779936e-05,
      "loss": 0.1059,
      "step": 146030
    },
    {
      "epoch": 56.71,
      "learning_rate": 1.2438058252427186e-05,
      "loss": 0.0174,
      "step": 146040
    },
    {
      "epoch": 56.72,
      "learning_rate": 1.2437540453074436e-05,
      "loss": 0.0574,
      "step": 146050
    },
    {
      "epoch": 56.72,
      "learning_rate": 1.2437022653721684e-05,
      "loss": 0.0244,
      "step": 146060
    },
    {
      "epoch": 56.73,
      "learning_rate": 1.2436504854368933e-05,
      "loss": 0.0312,
      "step": 146070
    },
    {
      "epoch": 56.73,
      "learning_rate": 1.2435987055016183e-05,
      "loss": 0.2167,
      "step": 146080
    },
    {
      "epoch": 56.73,
      "learning_rate": 1.243546925566343e-05,
      "loss": 0.0303,
      "step": 146090
    },
    {
      "epoch": 56.74,
      "learning_rate": 1.243495145631068e-05,
      "loss": 0.0338,
      "step": 146100
    },
    {
      "epoch": 56.74,
      "learning_rate": 1.243443365695793e-05,
      "loss": 0.1458,
      "step": 146110
    },
    {
      "epoch": 56.75,
      "learning_rate": 1.243391585760518e-05,
      "loss": 0.0193,
      "step": 146120
    },
    {
      "epoch": 56.75,
      "learning_rate": 1.2433398058252427e-05,
      "loss": 0.052,
      "step": 146130
    },
    {
      "epoch": 56.75,
      "learning_rate": 1.2432880258899677e-05,
      "loss": 0.2126,
      "step": 146140
    },
    {
      "epoch": 56.76,
      "learning_rate": 1.2432362459546927e-05,
      "loss": 0.1479,
      "step": 146150
    },
    {
      "epoch": 56.76,
      "learning_rate": 1.2431844660194176e-05,
      "loss": 0.067,
      "step": 146160
    },
    {
      "epoch": 56.77,
      "learning_rate": 1.2431326860841424e-05,
      "loss": 0.0446,
      "step": 146170
    },
    {
      "epoch": 56.77,
      "learning_rate": 1.2430809061488674e-05,
      "loss": 0.0296,
      "step": 146180
    },
    {
      "epoch": 56.77,
      "learning_rate": 1.2430291262135923e-05,
      "loss": 0.0359,
      "step": 146190
    },
    {
      "epoch": 56.78,
      "learning_rate": 1.2429773462783173e-05,
      "loss": 0.0748,
      "step": 146200
    },
    {
      "epoch": 56.78,
      "learning_rate": 1.2429255663430421e-05,
      "loss": 0.0648,
      "step": 146210
    },
    {
      "epoch": 56.78,
      "learning_rate": 1.242873786407767e-05,
      "loss": 0.0905,
      "step": 146220
    },
    {
      "epoch": 56.79,
      "learning_rate": 1.242822006472492e-05,
      "loss": 0.1142,
      "step": 146230
    },
    {
      "epoch": 56.79,
      "learning_rate": 1.242770226537217e-05,
      "loss": 0.1133,
      "step": 146240
    },
    {
      "epoch": 56.8,
      "learning_rate": 1.2427184466019418e-05,
      "loss": 0.1787,
      "step": 146250
    },
    {
      "epoch": 56.8,
      "learning_rate": 1.2426666666666667e-05,
      "loss": 0.0878,
      "step": 146260
    },
    {
      "epoch": 56.8,
      "learning_rate": 1.2426148867313917e-05,
      "loss": 0.0912,
      "step": 146270
    },
    {
      "epoch": 56.81,
      "learning_rate": 1.2425631067961167e-05,
      "loss": 0.0152,
      "step": 146280
    },
    {
      "epoch": 56.81,
      "learning_rate": 1.2425113268608415e-05,
      "loss": 0.0843,
      "step": 146290
    },
    {
      "epoch": 56.82,
      "learning_rate": 1.2424595469255664e-05,
      "loss": 0.1392,
      "step": 146300
    },
    {
      "epoch": 56.82,
      "learning_rate": 1.2424077669902914e-05,
      "loss": 0.0503,
      "step": 146310
    },
    {
      "epoch": 56.82,
      "learning_rate": 1.2423559870550163e-05,
      "loss": 0.0029,
      "step": 146320
    },
    {
      "epoch": 56.83,
      "learning_rate": 1.2423042071197413e-05,
      "loss": 0.1003,
      "step": 146330
    },
    {
      "epoch": 56.83,
      "learning_rate": 1.2422524271844661e-05,
      "loss": 0.0598,
      "step": 146340
    },
    {
      "epoch": 56.83,
      "learning_rate": 1.242200647249191e-05,
      "loss": 0.1111,
      "step": 146350
    },
    {
      "epoch": 56.84,
      "learning_rate": 1.2421488673139159e-05,
      "loss": 0.0496,
      "step": 146360
    },
    {
      "epoch": 56.84,
      "learning_rate": 1.2420970873786408e-05,
      "loss": 0.1087,
      "step": 146370
    },
    {
      "epoch": 56.85,
      "learning_rate": 1.2420453074433658e-05,
      "loss": 0.2092,
      "step": 146380
    },
    {
      "epoch": 56.85,
      "learning_rate": 1.2419935275080907e-05,
      "loss": 0.1021,
      "step": 146390
    },
    {
      "epoch": 56.85,
      "learning_rate": 1.2419417475728157e-05,
      "loss": 0.0052,
      "step": 146400
    },
    {
      "epoch": 56.86,
      "learning_rate": 1.2418899676375407e-05,
      "loss": 0.0113,
      "step": 146410
    },
    {
      "epoch": 56.86,
      "learning_rate": 1.2418381877022655e-05,
      "loss": 0.27,
      "step": 146420
    },
    {
      "epoch": 56.87,
      "learning_rate": 1.2417864077669904e-05,
      "loss": 0.0077,
      "step": 146430
    },
    {
      "epoch": 56.87,
      "learning_rate": 1.2417346278317152e-05,
      "loss": 0.0268,
      "step": 146440
    },
    {
      "epoch": 56.87,
      "learning_rate": 1.2416828478964402e-05,
      "loss": 0.1367,
      "step": 146450
    },
    {
      "epoch": 56.88,
      "learning_rate": 1.2416310679611651e-05,
      "loss": 0.1242,
      "step": 146460
    },
    {
      "epoch": 56.88,
      "learning_rate": 1.2415792880258901e-05,
      "loss": 0.1266,
      "step": 146470
    },
    {
      "epoch": 56.89,
      "learning_rate": 1.241527508090615e-05,
      "loss": 0.2708,
      "step": 146480
    },
    {
      "epoch": 56.89,
      "learning_rate": 1.24147572815534e-05,
      "loss": 0.2052,
      "step": 146490
    },
    {
      "epoch": 56.89,
      "learning_rate": 1.2414239482200648e-05,
      "loss": 0.0753,
      "step": 146500
    },
    {
      "epoch": 56.9,
      "learning_rate": 1.2413721682847896e-05,
      "loss": 0.0331,
      "step": 146510
    },
    {
      "epoch": 56.9,
      "learning_rate": 1.2413203883495146e-05,
      "loss": 0.0424,
      "step": 146520
    },
    {
      "epoch": 56.9,
      "learning_rate": 1.2412686084142395e-05,
      "loss": 0.1217,
      "step": 146530
    },
    {
      "epoch": 56.91,
      "learning_rate": 1.2412168284789645e-05,
      "loss": 0.021,
      "step": 146540
    },
    {
      "epoch": 56.91,
      "learning_rate": 1.2411650485436894e-05,
      "loss": 0.0602,
      "step": 146550
    },
    {
      "epoch": 56.92,
      "learning_rate": 1.2411132686084144e-05,
      "loss": 0.0227,
      "step": 146560
    },
    {
      "epoch": 56.92,
      "learning_rate": 1.2410614886731394e-05,
      "loss": 0.0426,
      "step": 146570
    },
    {
      "epoch": 56.92,
      "learning_rate": 1.2410097087378642e-05,
      "loss": 0.0595,
      "step": 146580
    },
    {
      "epoch": 56.93,
      "learning_rate": 1.240957928802589e-05,
      "loss": 0.1152,
      "step": 146590
    },
    {
      "epoch": 56.93,
      "learning_rate": 1.240906148867314e-05,
      "loss": 0.0169,
      "step": 146600
    },
    {
      "epoch": 56.94,
      "learning_rate": 1.2408543689320389e-05,
      "loss": 0.0962,
      "step": 146610
    },
    {
      "epoch": 56.94,
      "learning_rate": 1.2408025889967638e-05,
      "loss": 0.1029,
      "step": 146620
    },
    {
      "epoch": 56.94,
      "learning_rate": 1.2407508090614888e-05,
      "loss": 0.0006,
      "step": 146630
    },
    {
      "epoch": 56.95,
      "learning_rate": 1.2406990291262138e-05,
      "loss": 0.0467,
      "step": 146640
    },
    {
      "epoch": 56.95,
      "learning_rate": 1.2406472491909387e-05,
      "loss": 0.0608,
      "step": 146650
    },
    {
      "epoch": 56.96,
      "learning_rate": 1.2405954692556633e-05,
      "loss": 0.0212,
      "step": 146660
    },
    {
      "epoch": 56.96,
      "learning_rate": 1.2405436893203883e-05,
      "loss": 0.1184,
      "step": 146670
    },
    {
      "epoch": 56.96,
      "learning_rate": 1.2404919093851133e-05,
      "loss": 0.1716,
      "step": 146680
    },
    {
      "epoch": 56.97,
      "learning_rate": 1.2404401294498382e-05,
      "loss": 0.1621,
      "step": 146690
    },
    {
      "epoch": 56.97,
      "learning_rate": 1.2403883495145632e-05,
      "loss": 0.0233,
      "step": 146700
    },
    {
      "epoch": 56.97,
      "learning_rate": 1.2403365695792882e-05,
      "loss": 0.011,
      "step": 146710
    },
    {
      "epoch": 56.98,
      "learning_rate": 1.2402847896440131e-05,
      "loss": 0.0376,
      "step": 146720
    },
    {
      "epoch": 56.98,
      "learning_rate": 1.240233009708738e-05,
      "loss": 0.1659,
      "step": 146730
    },
    {
      "epoch": 56.99,
      "learning_rate": 1.2401812297734627e-05,
      "loss": 0.0811,
      "step": 146740
    },
    {
      "epoch": 56.99,
      "learning_rate": 1.2401294498381877e-05,
      "loss": 0.1475,
      "step": 146750
    },
    {
      "epoch": 56.99,
      "learning_rate": 1.2400776699029126e-05,
      "loss": 0.0554,
      "step": 146760
    },
    {
      "epoch": 57.0,
      "learning_rate": 1.2400258899676376e-05,
      "loss": 0.0488,
      "step": 146770
    },
    {
      "epoch": 57.0,
      "eval_accuracy": 0.9504814305364512,
      "eval_loss": 0.32903438806533813,
      "eval_runtime": 8.2078,
      "eval_samples_per_second": 442.874,
      "eval_steps_per_second": 55.435,
      "step": 146775
    },
    {
      "epoch": 57.0,
      "learning_rate": 1.2399741100323626e-05,
      "loss": 0.0551,
      "step": 146780
    },
    {
      "epoch": 57.01,
      "learning_rate": 1.2399223300970875e-05,
      "loss": 0.0517,
      "step": 146790
    },
    {
      "epoch": 57.01,
      "learning_rate": 1.2398705501618125e-05,
      "loss": 0.0426,
      "step": 146800
    },
    {
      "epoch": 57.01,
      "learning_rate": 1.2398187702265374e-05,
      "loss": 0.0879,
      "step": 146810
    },
    {
      "epoch": 57.02,
      "learning_rate": 1.239766990291262e-05,
      "loss": 0.0139,
      "step": 146820
    },
    {
      "epoch": 57.02,
      "learning_rate": 1.239715210355987e-05,
      "loss": 0.0321,
      "step": 146830
    },
    {
      "epoch": 57.03,
      "learning_rate": 1.239663430420712e-05,
      "loss": 0.1205,
      "step": 146840
    },
    {
      "epoch": 57.03,
      "learning_rate": 1.239611650485437e-05,
      "loss": 0.0827,
      "step": 146850
    },
    {
      "epoch": 57.03,
      "learning_rate": 1.2395598705501619e-05,
      "loss": 0.128,
      "step": 146860
    },
    {
      "epoch": 57.04,
      "learning_rate": 1.2395080906148869e-05,
      "loss": 0.1091,
      "step": 146870
    },
    {
      "epoch": 57.04,
      "learning_rate": 1.2394563106796118e-05,
      "loss": 0.0685,
      "step": 146880
    },
    {
      "epoch": 57.04,
      "learning_rate": 1.2394045307443368e-05,
      "loss": 0.0735,
      "step": 146890
    },
    {
      "epoch": 57.05,
      "learning_rate": 1.2393527508090618e-05,
      "loss": 0.0916,
      "step": 146900
    },
    {
      "epoch": 57.05,
      "learning_rate": 1.2393009708737864e-05,
      "loss": 0.0854,
      "step": 146910
    },
    {
      "epoch": 57.06,
      "learning_rate": 1.2392491909385113e-05,
      "loss": 0.0436,
      "step": 146920
    },
    {
      "epoch": 57.06,
      "learning_rate": 1.2391974110032363e-05,
      "loss": 0.0356,
      "step": 146930
    },
    {
      "epoch": 57.06,
      "learning_rate": 1.2391456310679613e-05,
      "loss": 0.0068,
      "step": 146940
    },
    {
      "epoch": 57.07,
      "learning_rate": 1.2390938511326862e-05,
      "loss": 0.0888,
      "step": 146950
    },
    {
      "epoch": 57.07,
      "learning_rate": 1.2390420711974112e-05,
      "loss": 0.0386,
      "step": 146960
    },
    {
      "epoch": 57.08,
      "learning_rate": 1.2389902912621361e-05,
      "loss": 0.0307,
      "step": 146970
    },
    {
      "epoch": 57.08,
      "learning_rate": 1.2389385113268611e-05,
      "loss": 0.0005,
      "step": 146980
    },
    {
      "epoch": 57.08,
      "learning_rate": 1.2388867313915857e-05,
      "loss": 0.2216,
      "step": 146990
    },
    {
      "epoch": 57.09,
      "learning_rate": 1.2388349514563107e-05,
      "loss": 0.0588,
      "step": 147000
    },
    {
      "epoch": 57.09,
      "learning_rate": 1.2387831715210357e-05,
      "loss": 0.1029,
      "step": 147010
    },
    {
      "epoch": 57.1,
      "learning_rate": 1.2387313915857606e-05,
      "loss": 0.0778,
      "step": 147020
    },
    {
      "epoch": 57.1,
      "learning_rate": 1.2386796116504856e-05,
      "loss": 0.1957,
      "step": 147030
    },
    {
      "epoch": 57.1,
      "learning_rate": 1.2386278317152105e-05,
      "loss": 0.0219,
      "step": 147040
    },
    {
      "epoch": 57.11,
      "learning_rate": 1.2385760517799355e-05,
      "loss": 0.0901,
      "step": 147050
    },
    {
      "epoch": 57.11,
      "learning_rate": 1.2385242718446605e-05,
      "loss": 0.0261,
      "step": 147060
    },
    {
      "epoch": 57.11,
      "learning_rate": 1.2384724919093851e-05,
      "loss": 0.1307,
      "step": 147070
    },
    {
      "epoch": 57.12,
      "learning_rate": 1.23842071197411e-05,
      "loss": 0.0759,
      "step": 147080
    },
    {
      "epoch": 57.12,
      "learning_rate": 1.238368932038835e-05,
      "loss": 0.0764,
      "step": 147090
    },
    {
      "epoch": 57.13,
      "learning_rate": 1.23831715210356e-05,
      "loss": 0.0933,
      "step": 147100
    },
    {
      "epoch": 57.13,
      "learning_rate": 1.238265372168285e-05,
      "loss": 0.0311,
      "step": 147110
    },
    {
      "epoch": 57.13,
      "learning_rate": 1.2382135922330099e-05,
      "loss": 0.0746,
      "step": 147120
    },
    {
      "epoch": 57.14,
      "learning_rate": 1.2381618122977349e-05,
      "loss": 0.0524,
      "step": 147130
    },
    {
      "epoch": 57.14,
      "learning_rate": 1.2381100323624598e-05,
      "loss": 0.047,
      "step": 147140
    },
    {
      "epoch": 57.15,
      "learning_rate": 1.2380582524271844e-05,
      "loss": 0.0625,
      "step": 147150
    },
    {
      "epoch": 57.15,
      "learning_rate": 1.2380064724919094e-05,
      "loss": 0.0667,
      "step": 147160
    },
    {
      "epoch": 57.15,
      "learning_rate": 1.2379546925566344e-05,
      "loss": 0.0895,
      "step": 147170
    },
    {
      "epoch": 57.16,
      "learning_rate": 1.2379029126213593e-05,
      "loss": 0.1018,
      "step": 147180
    },
    {
      "epoch": 57.16,
      "learning_rate": 1.2378511326860843e-05,
      "loss": 0.099,
      "step": 147190
    },
    {
      "epoch": 57.17,
      "learning_rate": 1.2377993527508093e-05,
      "loss": 0.1472,
      "step": 147200
    },
    {
      "epoch": 57.17,
      "learning_rate": 1.2377475728155342e-05,
      "loss": 0.0174,
      "step": 147210
    },
    {
      "epoch": 57.17,
      "learning_rate": 1.2376957928802592e-05,
      "loss": 0.0945,
      "step": 147220
    },
    {
      "epoch": 57.18,
      "learning_rate": 1.2376440129449838e-05,
      "loss": 0.116,
      "step": 147230
    },
    {
      "epoch": 57.18,
      "learning_rate": 1.2375922330097088e-05,
      "loss": 0.0737,
      "step": 147240
    },
    {
      "epoch": 57.18,
      "learning_rate": 1.2375404530744337e-05,
      "loss": 0.1798,
      "step": 147250
    },
    {
      "epoch": 57.19,
      "learning_rate": 1.2374886731391587e-05,
      "loss": 0.0071,
      "step": 147260
    },
    {
      "epoch": 57.19,
      "learning_rate": 1.2374368932038836e-05,
      "loss": 0.0407,
      "step": 147270
    },
    {
      "epoch": 57.2,
      "learning_rate": 1.2373851132686086e-05,
      "loss": 0.0915,
      "step": 147280
    },
    {
      "epoch": 57.2,
      "learning_rate": 1.2373333333333336e-05,
      "loss": 0.0086,
      "step": 147290
    },
    {
      "epoch": 57.2,
      "learning_rate": 1.2372815533980584e-05,
      "loss": 0.0395,
      "step": 147300
    },
    {
      "epoch": 57.21,
      "learning_rate": 1.2372297734627832e-05,
      "loss": 0.1053,
      "step": 147310
    },
    {
      "epoch": 57.21,
      "learning_rate": 1.2371779935275081e-05,
      "loss": 0.0202,
      "step": 147320
    },
    {
      "epoch": 57.22,
      "learning_rate": 1.237126213592233e-05,
      "loss": 0.1365,
      "step": 147330
    },
    {
      "epoch": 57.22,
      "learning_rate": 1.237074433656958e-05,
      "loss": 0.0012,
      "step": 147340
    },
    {
      "epoch": 57.22,
      "learning_rate": 1.237022653721683e-05,
      "loss": 0.0707,
      "step": 147350
    },
    {
      "epoch": 57.23,
      "learning_rate": 1.236970873786408e-05,
      "loss": 0.1063,
      "step": 147360
    },
    {
      "epoch": 57.23,
      "learning_rate": 1.236919093851133e-05,
      "loss": 0.1231,
      "step": 147370
    },
    {
      "epoch": 57.23,
      "learning_rate": 1.2368673139158577e-05,
      "loss": 0.0537,
      "step": 147380
    },
    {
      "epoch": 57.24,
      "learning_rate": 1.2368155339805825e-05,
      "loss": 0.0207,
      "step": 147390
    },
    {
      "epoch": 57.24,
      "learning_rate": 1.2367637540453075e-05,
      "loss": 0.1228,
      "step": 147400
    },
    {
      "epoch": 57.25,
      "learning_rate": 1.2367119741100324e-05,
      "loss": 0.0431,
      "step": 147410
    },
    {
      "epoch": 57.25,
      "learning_rate": 1.2366601941747574e-05,
      "loss": 0.1329,
      "step": 147420
    },
    {
      "epoch": 57.25,
      "learning_rate": 1.2366084142394824e-05,
      "loss": 0.0296,
      "step": 147430
    },
    {
      "epoch": 57.26,
      "learning_rate": 1.2365566343042073e-05,
      "loss": 0.0105,
      "step": 147440
    },
    {
      "epoch": 57.26,
      "learning_rate": 1.2365048543689321e-05,
      "loss": 0.1017,
      "step": 147450
    },
    {
      "epoch": 57.27,
      "learning_rate": 1.236453074433657e-05,
      "loss": 0.0009,
      "step": 147460
    },
    {
      "epoch": 57.27,
      "learning_rate": 1.236401294498382e-05,
      "loss": 0.0672,
      "step": 147470
    },
    {
      "epoch": 57.27,
      "learning_rate": 1.2363495145631068e-05,
      "loss": 0.1194,
      "step": 147480
    },
    {
      "epoch": 57.28,
      "learning_rate": 1.2362977346278318e-05,
      "loss": 0.0188,
      "step": 147490
    },
    {
      "epoch": 57.28,
      "learning_rate": 1.2362459546925568e-05,
      "loss": 0.0401,
      "step": 147500
    },
    {
      "epoch": 57.29,
      "learning_rate": 1.2361941747572817e-05,
      "loss": 0.1037,
      "step": 147510
    },
    {
      "epoch": 57.29,
      "learning_rate": 1.2361423948220067e-05,
      "loss": 0.037,
      "step": 147520
    },
    {
      "epoch": 57.29,
      "learning_rate": 1.2360906148867315e-05,
      "loss": 0.0926,
      "step": 147530
    },
    {
      "epoch": 57.3,
      "learning_rate": 1.2360388349514564e-05,
      "loss": 0.0182,
      "step": 147540
    },
    {
      "epoch": 57.3,
      "learning_rate": 1.2359870550161814e-05,
      "loss": 0.0561,
      "step": 147550
    },
    {
      "epoch": 57.3,
      "learning_rate": 1.2359352750809062e-05,
      "loss": 0.024,
      "step": 147560
    },
    {
      "epoch": 57.31,
      "learning_rate": 1.2358834951456311e-05,
      "loss": 0.0125,
      "step": 147570
    },
    {
      "epoch": 57.31,
      "learning_rate": 1.2358317152103561e-05,
      "loss": 0.1062,
      "step": 147580
    },
    {
      "epoch": 57.32,
      "learning_rate": 1.235779935275081e-05,
      "loss": 0.0931,
      "step": 147590
    },
    {
      "epoch": 57.32,
      "learning_rate": 1.2357281553398059e-05,
      "loss": 0.0292,
      "step": 147600
    },
    {
      "epoch": 57.32,
      "learning_rate": 1.2356763754045308e-05,
      "loss": 0.0671,
      "step": 147610
    },
    {
      "epoch": 57.33,
      "learning_rate": 1.2356245954692558e-05,
      "loss": 0.0012,
      "step": 147620
    },
    {
      "epoch": 57.33,
      "learning_rate": 1.2355728155339807e-05,
      "loss": 0.0211,
      "step": 147630
    },
    {
      "epoch": 57.34,
      "learning_rate": 1.2355210355987055e-05,
      "loss": 0.0436,
      "step": 147640
    },
    {
      "epoch": 57.34,
      "learning_rate": 1.2354692556634305e-05,
      "loss": 0.0404,
      "step": 147650
    },
    {
      "epoch": 57.34,
      "learning_rate": 1.2354174757281555e-05,
      "loss": 0.1809,
      "step": 147660
    },
    {
      "epoch": 57.35,
      "learning_rate": 1.2353656957928804e-05,
      "loss": 0.1913,
      "step": 147670
    },
    {
      "epoch": 57.35,
      "learning_rate": 1.2353139158576052e-05,
      "loss": 0.1422,
      "step": 147680
    },
    {
      "epoch": 57.36,
      "learning_rate": 1.2352621359223302e-05,
      "loss": 0.1429,
      "step": 147690
    },
    {
      "epoch": 57.36,
      "learning_rate": 1.2352103559870551e-05,
      "loss": 0.0068,
      "step": 147700
    },
    {
      "epoch": 57.36,
      "learning_rate": 1.2351585760517801e-05,
      "loss": 0.1333,
      "step": 147710
    },
    {
      "epoch": 57.37,
      "learning_rate": 1.2351067961165049e-05,
      "loss": 0.0336,
      "step": 147720
    },
    {
      "epoch": 57.37,
      "learning_rate": 1.2350550161812299e-05,
      "loss": 0.0035,
      "step": 147730
    },
    {
      "epoch": 57.37,
      "learning_rate": 1.2350032362459548e-05,
      "loss": 0.046,
      "step": 147740
    },
    {
      "epoch": 57.38,
      "learning_rate": 1.2349514563106798e-05,
      "loss": 0.0635,
      "step": 147750
    },
    {
      "epoch": 57.38,
      "learning_rate": 1.2348996763754046e-05,
      "loss": 0.0425,
      "step": 147760
    },
    {
      "epoch": 57.39,
      "learning_rate": 1.2348478964401295e-05,
      "loss": 0.1155,
      "step": 147770
    },
    {
      "epoch": 57.39,
      "learning_rate": 1.2347961165048545e-05,
      "loss": 0.1112,
      "step": 147780
    },
    {
      "epoch": 57.39,
      "learning_rate": 1.2347443365695795e-05,
      "loss": 0.1296,
      "step": 147790
    },
    {
      "epoch": 57.4,
      "learning_rate": 1.2346925566343043e-05,
      "loss": 0.167,
      "step": 147800
    },
    {
      "epoch": 57.4,
      "learning_rate": 1.2346407766990292e-05,
      "loss": 0.0332,
      "step": 147810
    },
    {
      "epoch": 57.41,
      "learning_rate": 1.2345889967637542e-05,
      "loss": 0.0608,
      "step": 147820
    },
    {
      "epoch": 57.41,
      "learning_rate": 1.234537216828479e-05,
      "loss": 0.1475,
      "step": 147830
    },
    {
      "epoch": 57.41,
      "learning_rate": 1.234485436893204e-05,
      "loss": 0.0088,
      "step": 147840
    },
    {
      "epoch": 57.42,
      "learning_rate": 1.2344336569579289e-05,
      "loss": 0.0875,
      "step": 147850
    },
    {
      "epoch": 57.42,
      "learning_rate": 1.2343818770226539e-05,
      "loss": 0.084,
      "step": 147860
    },
    {
      "epoch": 57.43,
      "learning_rate": 1.2343300970873788e-05,
      "loss": 0.02,
      "step": 147870
    },
    {
      "epoch": 57.43,
      "learning_rate": 1.2342783171521036e-05,
      "loss": 0.0574,
      "step": 147880
    },
    {
      "epoch": 57.43,
      "learning_rate": 1.2342265372168286e-05,
      "loss": 0.0008,
      "step": 147890
    },
    {
      "epoch": 57.44,
      "learning_rate": 1.2341747572815535e-05,
      "loss": 0.0327,
      "step": 147900
    },
    {
      "epoch": 57.44,
      "learning_rate": 1.2341229773462783e-05,
      "loss": 0.0745,
      "step": 147910
    },
    {
      "epoch": 57.44,
      "learning_rate": 1.2340711974110033e-05,
      "loss": 0.1114,
      "step": 147920
    },
    {
      "epoch": 57.45,
      "learning_rate": 1.2340194174757282e-05,
      "loss": 0.0494,
      "step": 147930
    },
    {
      "epoch": 57.45,
      "learning_rate": 1.2339676375404532e-05,
      "loss": 0.1438,
      "step": 147940
    },
    {
      "epoch": 57.46,
      "learning_rate": 1.2339158576051782e-05,
      "loss": 0.089,
      "step": 147950
    },
    {
      "epoch": 57.46,
      "learning_rate": 1.233864077669903e-05,
      "loss": 0.0908,
      "step": 147960
    },
    {
      "epoch": 57.46,
      "learning_rate": 1.233812297734628e-05,
      "loss": 0.245,
      "step": 147970
    },
    {
      "epoch": 57.47,
      "learning_rate": 1.2337605177993527e-05,
      "loss": 0.1421,
      "step": 147980
    },
    {
      "epoch": 57.47,
      "learning_rate": 1.2337087378640777e-05,
      "loss": 0.0779,
      "step": 147990
    },
    {
      "epoch": 57.48,
      "learning_rate": 1.2336569579288026e-05,
      "loss": 0.1637,
      "step": 148000
    },
    {
      "epoch": 57.48,
      "learning_rate": 1.2336051779935276e-05,
      "loss": 0.1809,
      "step": 148010
    },
    {
      "epoch": 57.48,
      "learning_rate": 1.2335533980582526e-05,
      "loss": 0.06,
      "step": 148020
    },
    {
      "epoch": 57.49,
      "learning_rate": 1.2335016181229775e-05,
      "loss": 0.0515,
      "step": 148030
    },
    {
      "epoch": 57.49,
      "learning_rate": 1.2334498381877025e-05,
      "loss": 0.0923,
      "step": 148040
    },
    {
      "epoch": 57.5,
      "learning_rate": 1.2333980582524273e-05,
      "loss": 0.0811,
      "step": 148050
    },
    {
      "epoch": 57.5,
      "learning_rate": 1.233346278317152e-05,
      "loss": 0.0508,
      "step": 148060
    },
    {
      "epoch": 57.5,
      "learning_rate": 1.233294498381877e-05,
      "loss": 0.1234,
      "step": 148070
    },
    {
      "epoch": 57.51,
      "learning_rate": 1.233242718446602e-05,
      "loss": 0.0969,
      "step": 148080
    },
    {
      "epoch": 57.51,
      "learning_rate": 1.233190938511327e-05,
      "loss": 0.0428,
      "step": 148090
    },
    {
      "epoch": 57.51,
      "learning_rate": 1.233139158576052e-05,
      "loss": 0.0189,
      "step": 148100
    },
    {
      "epoch": 57.52,
      "learning_rate": 1.2330873786407769e-05,
      "loss": 0.1193,
      "step": 148110
    },
    {
      "epoch": 57.52,
      "learning_rate": 1.2330355987055018e-05,
      "loss": 0.1869,
      "step": 148120
    },
    {
      "epoch": 57.53,
      "learning_rate": 1.2329838187702265e-05,
      "loss": 0.1211,
      "step": 148130
    },
    {
      "epoch": 57.53,
      "learning_rate": 1.2329320388349514e-05,
      "loss": 0.1472,
      "step": 148140
    },
    {
      "epoch": 57.53,
      "learning_rate": 1.2328802588996764e-05,
      "loss": 0.0072,
      "step": 148150
    },
    {
      "epoch": 57.54,
      "learning_rate": 1.2328284789644014e-05,
      "loss": 0.0161,
      "step": 148160
    },
    {
      "epoch": 57.54,
      "learning_rate": 1.2327766990291263e-05,
      "loss": 0.104,
      "step": 148170
    },
    {
      "epoch": 57.55,
      "learning_rate": 1.2327249190938513e-05,
      "loss": 0.2126,
      "step": 148180
    },
    {
      "epoch": 57.55,
      "learning_rate": 1.2326731391585762e-05,
      "loss": 0.0521,
      "step": 148190
    },
    {
      "epoch": 57.55,
      "learning_rate": 1.2326213592233012e-05,
      "loss": 0.1855,
      "step": 148200
    },
    {
      "epoch": 57.56,
      "learning_rate": 1.2325695792880258e-05,
      "loss": 0.0114,
      "step": 148210
    },
    {
      "epoch": 57.56,
      "learning_rate": 1.2325177993527508e-05,
      "loss": 0.1272,
      "step": 148220
    },
    {
      "epoch": 57.57,
      "learning_rate": 1.2324660194174757e-05,
      "loss": 0.148,
      "step": 148230
    },
    {
      "epoch": 57.57,
      "learning_rate": 1.2324142394822007e-05,
      "loss": 0.0155,
      "step": 148240
    },
    {
      "epoch": 57.57,
      "learning_rate": 1.2323624595469257e-05,
      "loss": 0.0602,
      "step": 148250
    },
    {
      "epoch": 57.58,
      "learning_rate": 1.2323106796116506e-05,
      "loss": 0.1456,
      "step": 148260
    },
    {
      "epoch": 57.58,
      "learning_rate": 1.2322588996763756e-05,
      "loss": 0.0952,
      "step": 148270
    },
    {
      "epoch": 57.58,
      "learning_rate": 1.2322071197411006e-05,
      "loss": 0.0098,
      "step": 148280
    },
    {
      "epoch": 57.59,
      "learning_rate": 1.2321553398058252e-05,
      "loss": 0.1108,
      "step": 148290
    },
    {
      "epoch": 57.59,
      "learning_rate": 1.2321035598705501e-05,
      "loss": 0.1067,
      "step": 148300
    },
    {
      "epoch": 57.6,
      "learning_rate": 1.2320517799352751e-05,
      "loss": 0.0641,
      "step": 148310
    },
    {
      "epoch": 57.6,
      "learning_rate": 1.232e-05,
      "loss": 0.202,
      "step": 148320
    },
    {
      "epoch": 57.6,
      "learning_rate": 1.231948220064725e-05,
      "loss": 0.1799,
      "step": 148330
    },
    {
      "epoch": 57.61,
      "learning_rate": 1.23189644012945e-05,
      "loss": 0.1125,
      "step": 148340
    },
    {
      "epoch": 57.61,
      "learning_rate": 1.231844660194175e-05,
      "loss": 0.0217,
      "step": 148350
    },
    {
      "epoch": 57.62,
      "learning_rate": 1.2317928802588999e-05,
      "loss": 0.1807,
      "step": 148360
    },
    {
      "epoch": 57.62,
      "learning_rate": 1.2317411003236245e-05,
      "loss": 0.0004,
      "step": 148370
    },
    {
      "epoch": 57.62,
      "learning_rate": 1.2316893203883495e-05,
      "loss": 0.0039,
      "step": 148380
    },
    {
      "epoch": 57.63,
      "learning_rate": 1.2316375404530745e-05,
      "loss": 0.0318,
      "step": 148390
    },
    {
      "epoch": 57.63,
      "learning_rate": 1.2315857605177994e-05,
      "loss": 0.0737,
      "step": 148400
    },
    {
      "epoch": 57.63,
      "learning_rate": 1.2315339805825244e-05,
      "loss": 0.0192,
      "step": 148410
    },
    {
      "epoch": 57.64,
      "learning_rate": 1.2314822006472493e-05,
      "loss": 0.0611,
      "step": 148420
    },
    {
      "epoch": 57.64,
      "learning_rate": 1.2314304207119743e-05,
      "loss": 0.1463,
      "step": 148430
    },
    {
      "epoch": 57.65,
      "learning_rate": 1.2313786407766993e-05,
      "loss": 0.0191,
      "step": 148440
    },
    {
      "epoch": 57.65,
      "learning_rate": 1.2313268608414239e-05,
      "loss": 0.0878,
      "step": 148450
    },
    {
      "epoch": 57.65,
      "learning_rate": 1.2312750809061489e-05,
      "loss": 0.0754,
      "step": 148460
    },
    {
      "epoch": 57.66,
      "learning_rate": 1.2312233009708738e-05,
      "loss": 0.1074,
      "step": 148470
    },
    {
      "epoch": 57.66,
      "learning_rate": 1.2311715210355988e-05,
      "loss": 0.06,
      "step": 148480
    },
    {
      "epoch": 57.67,
      "learning_rate": 1.2311197411003237e-05,
      "loss": 0.0649,
      "step": 148490
    },
    {
      "epoch": 57.67,
      "learning_rate": 1.2310679611650487e-05,
      "loss": 0.0267,
      "step": 148500
    },
    {
      "epoch": 57.67,
      "learning_rate": 1.2310161812297737e-05,
      "loss": 0.0227,
      "step": 148510
    },
    {
      "epoch": 57.68,
      "learning_rate": 1.2309644012944986e-05,
      "loss": 0.1277,
      "step": 148520
    },
    {
      "epoch": 57.68,
      "learning_rate": 1.2309126213592232e-05,
      "loss": 0.0187,
      "step": 148530
    },
    {
      "epoch": 57.69,
      "learning_rate": 1.2308608414239482e-05,
      "loss": 0.0385,
      "step": 148540
    },
    {
      "epoch": 57.69,
      "learning_rate": 1.2308090614886732e-05,
      "loss": 0.1231,
      "step": 148550
    },
    {
      "epoch": 57.69,
      "learning_rate": 1.2307572815533981e-05,
      "loss": 0.0289,
      "step": 148560
    },
    {
      "epoch": 57.7,
      "learning_rate": 1.2307055016181231e-05,
      "loss": 0.0701,
      "step": 148570
    },
    {
      "epoch": 57.7,
      "learning_rate": 1.230653721682848e-05,
      "loss": 0.062,
      "step": 148580
    },
    {
      "epoch": 57.7,
      "learning_rate": 1.230601941747573e-05,
      "loss": 0.0124,
      "step": 148590
    },
    {
      "epoch": 57.71,
      "learning_rate": 1.230550161812298e-05,
      "loss": 0.0185,
      "step": 148600
    },
    {
      "epoch": 57.71,
      "learning_rate": 1.230498381877023e-05,
      "loss": 0.1391,
      "step": 148610
    },
    {
      "epoch": 57.72,
      "learning_rate": 1.2304466019417476e-05,
      "loss": 0.1234,
      "step": 148620
    },
    {
      "epoch": 57.72,
      "learning_rate": 1.2303948220064725e-05,
      "loss": 0.0124,
      "step": 148630
    },
    {
      "epoch": 57.72,
      "learning_rate": 1.2303430420711975e-05,
      "loss": 0.1425,
      "step": 148640
    },
    {
      "epoch": 57.73,
      "learning_rate": 1.2302912621359224e-05,
      "loss": 0.0748,
      "step": 148650
    },
    {
      "epoch": 57.73,
      "learning_rate": 1.2302394822006474e-05,
      "loss": 0.06,
      "step": 148660
    },
    {
      "epoch": 57.74,
      "learning_rate": 1.2301877022653724e-05,
      "loss": 0.1168,
      "step": 148670
    },
    {
      "epoch": 57.74,
      "learning_rate": 1.2301359223300973e-05,
      "loss": 0.2276,
      "step": 148680
    },
    {
      "epoch": 57.74,
      "learning_rate": 1.2300841423948223e-05,
      "loss": 0.1543,
      "step": 148690
    },
    {
      "epoch": 57.75,
      "learning_rate": 1.230032362459547e-05,
      "loss": 0.0416,
      "step": 148700
    },
    {
      "epoch": 57.75,
      "learning_rate": 1.2299805825242719e-05,
      "loss": 0.0442,
      "step": 148710
    },
    {
      "epoch": 57.76,
      "learning_rate": 1.2299288025889968e-05,
      "loss": 0.1606,
      "step": 148720
    },
    {
      "epoch": 57.76,
      "learning_rate": 1.2298770226537218e-05,
      "loss": 0.0773,
      "step": 148730
    },
    {
      "epoch": 57.76,
      "learning_rate": 1.2298252427184468e-05,
      "loss": 0.0958,
      "step": 148740
    },
    {
      "epoch": 57.77,
      "learning_rate": 1.2297734627831717e-05,
      "loss": 0.0862,
      "step": 148750
    },
    {
      "epoch": 57.77,
      "learning_rate": 1.2297216828478967e-05,
      "loss": 0.0232,
      "step": 148760
    },
    {
      "epoch": 57.77,
      "learning_rate": 1.2296699029126215e-05,
      "loss": 0.0635,
      "step": 148770
    },
    {
      "epoch": 57.78,
      "learning_rate": 1.2296181229773463e-05,
      "loss": 0.1017,
      "step": 148780
    },
    {
      "epoch": 57.78,
      "learning_rate": 1.2295663430420712e-05,
      "loss": 0.0833,
      "step": 148790
    },
    {
      "epoch": 57.79,
      "learning_rate": 1.2295145631067962e-05,
      "loss": 0.1616,
      "step": 148800
    },
    {
      "epoch": 57.79,
      "learning_rate": 1.2294627831715212e-05,
      "loss": 0.0854,
      "step": 148810
    },
    {
      "epoch": 57.79,
      "learning_rate": 1.2294110032362461e-05,
      "loss": 0.083,
      "step": 148820
    },
    {
      "epoch": 57.8,
      "learning_rate": 1.229359223300971e-05,
      "loss": 0.2582,
      "step": 148830
    },
    {
      "epoch": 57.8,
      "learning_rate": 1.229307443365696e-05,
      "loss": 0.0247,
      "step": 148840
    },
    {
      "epoch": 57.81,
      "learning_rate": 1.2292556634304208e-05,
      "loss": 0.202,
      "step": 148850
    },
    {
      "epoch": 57.81,
      "learning_rate": 1.2292038834951456e-05,
      "loss": 0.1219,
      "step": 148860
    },
    {
      "epoch": 57.81,
      "learning_rate": 1.2291521035598706e-05,
      "loss": 0.0037,
      "step": 148870
    },
    {
      "epoch": 57.82,
      "learning_rate": 1.2291003236245956e-05,
      "loss": 0.0724,
      "step": 148880
    },
    {
      "epoch": 57.82,
      "learning_rate": 1.2290485436893205e-05,
      "loss": 0.0477,
      "step": 148890
    },
    {
      "epoch": 57.83,
      "learning_rate": 1.2289967637540455e-05,
      "loss": 0.0879,
      "step": 148900
    },
    {
      "epoch": 57.83,
      "learning_rate": 1.2289449838187704e-05,
      "loss": 0.0295,
      "step": 148910
    },
    {
      "epoch": 57.83,
      "learning_rate": 1.2288932038834952e-05,
      "loss": 0.0703,
      "step": 148920
    },
    {
      "epoch": 57.84,
      "learning_rate": 1.2288414239482202e-05,
      "loss": 0.0243,
      "step": 148930
    },
    {
      "epoch": 57.84,
      "learning_rate": 1.228789644012945e-05,
      "loss": 0.0455,
      "step": 148940
    },
    {
      "epoch": 57.84,
      "learning_rate": 1.22873786407767e-05,
      "loss": 0.0457,
      "step": 148950
    },
    {
      "epoch": 57.85,
      "learning_rate": 1.2286860841423949e-05,
      "loss": 0.0269,
      "step": 148960
    },
    {
      "epoch": 57.85,
      "learning_rate": 1.2286343042071199e-05,
      "loss": 0.1555,
      "step": 148970
    },
    {
      "epoch": 57.86,
      "learning_rate": 1.2285825242718448e-05,
      "loss": 0.1088,
      "step": 148980
    },
    {
      "epoch": 57.86,
      "learning_rate": 1.2285307443365698e-05,
      "loss": 0.0012,
      "step": 148990
    },
    {
      "epoch": 57.86,
      "learning_rate": 1.2284789644012946e-05,
      "loss": 0.1071,
      "step": 149000
    },
    {
      "epoch": 57.87,
      "learning_rate": 1.2284271844660195e-05,
      "loss": 0.0409,
      "step": 149010
    },
    {
      "epoch": 57.87,
      "learning_rate": 1.2283754045307443e-05,
      "loss": 0.0181,
      "step": 149020
    },
    {
      "epoch": 57.88,
      "learning_rate": 1.2283236245954693e-05,
      "loss": 0.0016,
      "step": 149030
    },
    {
      "epoch": 57.88,
      "learning_rate": 1.2282718446601943e-05,
      "loss": 0.0277,
      "step": 149040
    },
    {
      "epoch": 57.88,
      "learning_rate": 1.2282200647249192e-05,
      "loss": 0.1081,
      "step": 149050
    },
    {
      "epoch": 57.89,
      "learning_rate": 1.2281682847896442e-05,
      "loss": 0.0503,
      "step": 149060
    },
    {
      "epoch": 57.89,
      "learning_rate": 1.228116504854369e-05,
      "loss": 0.0805,
      "step": 149070
    },
    {
      "epoch": 57.9,
      "learning_rate": 1.228064724919094e-05,
      "loss": 0.1656,
      "step": 149080
    },
    {
      "epoch": 57.9,
      "learning_rate": 1.2280129449838189e-05,
      "loss": 0.1282,
      "step": 149090
    },
    {
      "epoch": 57.9,
      "learning_rate": 1.2279611650485437e-05,
      "loss": 0.0959,
      "step": 149100
    },
    {
      "epoch": 57.91,
      "learning_rate": 1.2279093851132687e-05,
      "loss": 0.0473,
      "step": 149110
    },
    {
      "epoch": 57.91,
      "learning_rate": 1.2278576051779936e-05,
      "loss": 0.0146,
      "step": 149120
    },
    {
      "epoch": 57.91,
      "learning_rate": 1.2278058252427186e-05,
      "loss": 0.0029,
      "step": 149130
    },
    {
      "epoch": 57.92,
      "learning_rate": 1.2277540453074435e-05,
      "loss": 0.1864,
      "step": 149140
    },
    {
      "epoch": 57.92,
      "learning_rate": 1.2277022653721683e-05,
      "loss": 0.0448,
      "step": 149150
    },
    {
      "epoch": 57.93,
      "learning_rate": 1.2276504854368933e-05,
      "loss": 0.243,
      "step": 149160
    },
    {
      "epoch": 57.93,
      "learning_rate": 1.2275987055016183e-05,
      "loss": 0.0515,
      "step": 149170
    },
    {
      "epoch": 57.93,
      "learning_rate": 1.2275469255663432e-05,
      "loss": 0.0092,
      "step": 149180
    },
    {
      "epoch": 57.94,
      "learning_rate": 1.227495145631068e-05,
      "loss": 0.0487,
      "step": 149190
    },
    {
      "epoch": 57.94,
      "learning_rate": 1.227443365695793e-05,
      "loss": 0.0298,
      "step": 149200
    },
    {
      "epoch": 57.95,
      "learning_rate": 1.227391585760518e-05,
      "loss": 0.0452,
      "step": 149210
    },
    {
      "epoch": 57.95,
      "learning_rate": 1.2273398058252429e-05,
      "loss": 0.097,
      "step": 149220
    },
    {
      "epoch": 57.95,
      "learning_rate": 1.2272880258899677e-05,
      "loss": 0.1084,
      "step": 149230
    },
    {
      "epoch": 57.96,
      "learning_rate": 1.2272362459546927e-05,
      "loss": 0.116,
      "step": 149240
    },
    {
      "epoch": 57.96,
      "learning_rate": 1.2271844660194176e-05,
      "loss": 0.0763,
      "step": 149250
    },
    {
      "epoch": 57.97,
      "learning_rate": 1.2271326860841426e-05,
      "loss": 0.2194,
      "step": 149260
    },
    {
      "epoch": 57.97,
      "learning_rate": 1.2270809061488674e-05,
      "loss": 0.0764,
      "step": 149270
    },
    {
      "epoch": 57.97,
      "learning_rate": 1.2270291262135923e-05,
      "loss": 0.0758,
      "step": 149280
    },
    {
      "epoch": 57.98,
      "learning_rate": 1.2269773462783173e-05,
      "loss": 0.0531,
      "step": 149290
    },
    {
      "epoch": 57.98,
      "learning_rate": 1.226925566343042e-05,
      "loss": 0.0161,
      "step": 149300
    },
    {
      "epoch": 57.98,
      "learning_rate": 1.226873786407767e-05,
      "loss": 0.1997,
      "step": 149310
    },
    {
      "epoch": 57.99,
      "learning_rate": 1.226822006472492e-05,
      "loss": 0.2084,
      "step": 149320
    },
    {
      "epoch": 57.99,
      "learning_rate": 1.226770226537217e-05,
      "loss": 0.0213,
      "step": 149330
    },
    {
      "epoch": 58.0,
      "learning_rate": 1.226718446601942e-05,
      "loss": 0.011,
      "step": 149340
    },
    {
      "epoch": 58.0,
      "learning_rate": 1.2266666666666667e-05,
      "loss": 0.0514,
      "step": 149350
    },
    {
      "epoch": 58.0,
      "eval_accuracy": 0.9535075653370014,
      "eval_loss": 0.3125966489315033,
      "eval_runtime": 8.2099,
      "eval_samples_per_second": 442.757,
      "eval_steps_per_second": 55.421,
      "step": 149350
    },
    {
      "epoch": 58.0,
      "learning_rate": 1.2266148867313917e-05,
      "loss": 0.0969,
      "step": 149360
    },
    {
      "epoch": 58.01,
      "learning_rate": 1.2265631067961166e-05,
      "loss": 0.0924,
      "step": 149370
    },
    {
      "epoch": 58.01,
      "learning_rate": 1.2265113268608414e-05,
      "loss": 0.0071,
      "step": 149380
    },
    {
      "epoch": 58.02,
      "learning_rate": 1.2264595469255664e-05,
      "loss": 0.1166,
      "step": 149390
    },
    {
      "epoch": 58.02,
      "learning_rate": 1.2264077669902914e-05,
      "loss": 0.0257,
      "step": 149400
    },
    {
      "epoch": 58.02,
      "learning_rate": 1.2263559870550163e-05,
      "loss": 0.1404,
      "step": 149410
    },
    {
      "epoch": 58.03,
      "learning_rate": 1.2263042071197413e-05,
      "loss": 0.0689,
      "step": 149420
    },
    {
      "epoch": 58.03,
      "learning_rate": 1.226252427184466e-05,
      "loss": 0.0527,
      "step": 149430
    },
    {
      "epoch": 58.03,
      "learning_rate": 1.226200647249191e-05,
      "loss": 0.0676,
      "step": 149440
    },
    {
      "epoch": 58.04,
      "learning_rate": 1.2261488673139158e-05,
      "loss": 0.1099,
      "step": 149450
    },
    {
      "epoch": 58.04,
      "learning_rate": 1.2260970873786408e-05,
      "loss": 0.0875,
      "step": 149460
    },
    {
      "epoch": 58.05,
      "learning_rate": 1.2260453074433658e-05,
      "loss": 0.0969,
      "step": 149470
    },
    {
      "epoch": 58.05,
      "learning_rate": 1.2259935275080907e-05,
      "loss": 0.0574,
      "step": 149480
    },
    {
      "epoch": 58.05,
      "learning_rate": 1.2259417475728157e-05,
      "loss": 0.062,
      "step": 149490
    },
    {
      "epoch": 58.06,
      "learning_rate": 1.2258899676375406e-05,
      "loss": 0.0681,
      "step": 149500
    },
    {
      "epoch": 58.06,
      "learning_rate": 1.2258381877022654e-05,
      "loss": 0.0496,
      "step": 149510
    },
    {
      "epoch": 58.07,
      "learning_rate": 1.2257864077669904e-05,
      "loss": 0.0508,
      "step": 149520
    },
    {
      "epoch": 58.07,
      "learning_rate": 1.2257346278317152e-05,
      "loss": 0.0015,
      "step": 149530
    },
    {
      "epoch": 58.07,
      "learning_rate": 1.2256828478964402e-05,
      "loss": 0.2114,
      "step": 149540
    },
    {
      "epoch": 58.08,
      "learning_rate": 1.2256310679611651e-05,
      "loss": 0.0917,
      "step": 149550
    },
    {
      "epoch": 58.08,
      "learning_rate": 1.22557928802589e-05,
      "loss": 0.1207,
      "step": 149560
    },
    {
      "epoch": 58.09,
      "learning_rate": 1.225527508090615e-05,
      "loss": 0.1091,
      "step": 149570
    },
    {
      "epoch": 58.09,
      "learning_rate": 1.22547572815534e-05,
      "loss": 0.0165,
      "step": 149580
    },
    {
      "epoch": 58.09,
      "learning_rate": 1.2254239482200648e-05,
      "loss": 0.001,
      "step": 149590
    },
    {
      "epoch": 58.1,
      "learning_rate": 1.2253721682847896e-05,
      "loss": 0.0298,
      "step": 149600
    },
    {
      "epoch": 58.1,
      "learning_rate": 1.2253203883495145e-05,
      "loss": 0.0437,
      "step": 149610
    },
    {
      "epoch": 58.1,
      "learning_rate": 1.2252686084142395e-05,
      "loss": 0.0076,
      "step": 149620
    },
    {
      "epoch": 58.11,
      "learning_rate": 1.2252168284789645e-05,
      "loss": 0.0249,
      "step": 149630
    },
    {
      "epoch": 58.11,
      "learning_rate": 1.2251650485436894e-05,
      "loss": 0.1105,
      "step": 149640
    },
    {
      "epoch": 58.12,
      "learning_rate": 1.2251132686084144e-05,
      "loss": 0.0889,
      "step": 149650
    },
    {
      "epoch": 58.12,
      "learning_rate": 1.2250614886731394e-05,
      "loss": 0.02,
      "step": 149660
    },
    {
      "epoch": 58.12,
      "learning_rate": 1.2250097087378641e-05,
      "loss": 0.0506,
      "step": 149670
    },
    {
      "epoch": 58.13,
      "learning_rate": 1.224957928802589e-05,
      "loss": 0.0707,
      "step": 149680
    },
    {
      "epoch": 58.13,
      "learning_rate": 1.2249061488673139e-05,
      "loss": 0.0242,
      "step": 149690
    },
    {
      "epoch": 58.14,
      "learning_rate": 1.2248543689320389e-05,
      "loss": 0.0552,
      "step": 149700
    },
    {
      "epoch": 58.14,
      "learning_rate": 1.2248025889967638e-05,
      "loss": 0.019,
      "step": 149710
    },
    {
      "epoch": 58.14,
      "learning_rate": 1.2247508090614888e-05,
      "loss": 0.1608,
      "step": 149720
    },
    {
      "epoch": 58.15,
      "learning_rate": 1.2246990291262137e-05,
      "loss": 0.015,
      "step": 149730
    },
    {
      "epoch": 58.15,
      "learning_rate": 1.2246472491909387e-05,
      "loss": 0.0862,
      "step": 149740
    },
    {
      "epoch": 58.16,
      "learning_rate": 1.2245954692556637e-05,
      "loss": 0.0512,
      "step": 149750
    },
    {
      "epoch": 58.16,
      "learning_rate": 1.2245436893203883e-05,
      "loss": 0.0802,
      "step": 149760
    },
    {
      "epoch": 58.16,
      "learning_rate": 1.2244919093851133e-05,
      "loss": 0.1479,
      "step": 149770
    },
    {
      "epoch": 58.17,
      "learning_rate": 1.2244401294498382e-05,
      "loss": 0.0624,
      "step": 149780
    },
    {
      "epoch": 58.17,
      "learning_rate": 1.2243883495145632e-05,
      "loss": 0.0204,
      "step": 149790
    },
    {
      "epoch": 58.17,
      "learning_rate": 1.2243365695792881e-05,
      "loss": 0.0149,
      "step": 149800
    },
    {
      "epoch": 58.18,
      "learning_rate": 1.2242847896440131e-05,
      "loss": 0.0259,
      "step": 149810
    },
    {
      "epoch": 58.18,
      "learning_rate": 1.224233009708738e-05,
      "loss": 0.009,
      "step": 149820
    },
    {
      "epoch": 58.19,
      "learning_rate": 1.224181229773463e-05,
      "loss": 0.094,
      "step": 149830
    },
    {
      "epoch": 58.19,
      "learning_rate": 1.2241294498381877e-05,
      "loss": 0.1391,
      "step": 149840
    },
    {
      "epoch": 58.19,
      "learning_rate": 1.2240776699029126e-05,
      "loss": 0.0831,
      "step": 149850
    },
    {
      "epoch": 58.2,
      "learning_rate": 1.2240258899676376e-05,
      "loss": 0.0509,
      "step": 149860
    },
    {
      "epoch": 58.2,
      "learning_rate": 1.2239741100323625e-05,
      "loss": 0.1053,
      "step": 149870
    },
    {
      "epoch": 58.21,
      "learning_rate": 1.2239223300970875e-05,
      "loss": 0.0775,
      "step": 149880
    },
    {
      "epoch": 58.21,
      "learning_rate": 1.2238705501618125e-05,
      "loss": 0.2851,
      "step": 149890
    },
    {
      "epoch": 58.21,
      "learning_rate": 1.2238187702265374e-05,
      "loss": 0.0003,
      "step": 149900
    },
    {
      "epoch": 58.22,
      "learning_rate": 1.2237669902912624e-05,
      "loss": 0.251,
      "step": 149910
    },
    {
      "epoch": 58.22,
      "learning_rate": 1.223715210355987e-05,
      "loss": 0.1444,
      "step": 149920
    },
    {
      "epoch": 58.23,
      "learning_rate": 1.223663430420712e-05,
      "loss": 0.0164,
      "step": 149930
    },
    {
      "epoch": 58.23,
      "learning_rate": 1.223611650485437e-05,
      "loss": 0.0451,
      "step": 149940
    },
    {
      "epoch": 58.23,
      "learning_rate": 1.2235598705501619e-05,
      "loss": 0.0836,
      "step": 149950
    },
    {
      "epoch": 58.24,
      "learning_rate": 1.2235080906148869e-05,
      "loss": 0.0407,
      "step": 149960
    },
    {
      "epoch": 58.24,
      "learning_rate": 1.2234563106796118e-05,
      "loss": 0.0953,
      "step": 149970
    },
    {
      "epoch": 58.24,
      "learning_rate": 1.2234045307443368e-05,
      "loss": 0.1638,
      "step": 149980
    },
    {
      "epoch": 58.25,
      "learning_rate": 1.2233527508090617e-05,
      "loss": 0.0996,
      "step": 149990
    },
    {
      "epoch": 58.25,
      "learning_rate": 1.2233009708737864e-05,
      "loss": 0.0044,
      "step": 150000
    },
    {
      "epoch": 58.26,
      "learning_rate": 1.2232491909385113e-05,
      "loss": 0.0921,
      "step": 150010
    },
    {
      "epoch": 58.26,
      "learning_rate": 1.2231974110032363e-05,
      "loss": 0.0377,
      "step": 150020
    },
    {
      "epoch": 58.26,
      "learning_rate": 1.2231456310679612e-05,
      "loss": 0.0647,
      "step": 150030
    },
    {
      "epoch": 58.27,
      "learning_rate": 1.2230938511326862e-05,
      "loss": 0.0779,
      "step": 150040
    },
    {
      "epoch": 58.27,
      "learning_rate": 1.2230420711974112e-05,
      "loss": 0.1152,
      "step": 150050
    },
    {
      "epoch": 58.28,
      "learning_rate": 1.2229902912621361e-05,
      "loss": 0.0137,
      "step": 150060
    },
    {
      "epoch": 58.28,
      "learning_rate": 1.2229385113268611e-05,
      "loss": 0.0831,
      "step": 150070
    },
    {
      "epoch": 58.28,
      "learning_rate": 1.2228867313915857e-05,
      "loss": 0.1087,
      "step": 150080
    },
    {
      "epoch": 58.29,
      "learning_rate": 1.2228349514563107e-05,
      "loss": 0.0191,
      "step": 150090
    },
    {
      "epoch": 58.29,
      "learning_rate": 1.2227831715210356e-05,
      "loss": 0.0375,
      "step": 150100
    },
    {
      "epoch": 58.3,
      "learning_rate": 1.2227313915857606e-05,
      "loss": 0.1038,
      "step": 150110
    },
    {
      "epoch": 58.3,
      "learning_rate": 1.2226796116504856e-05,
      "loss": 0.0185,
      "step": 150120
    },
    {
      "epoch": 58.3,
      "learning_rate": 1.2226278317152105e-05,
      "loss": 0.1499,
      "step": 150130
    },
    {
      "epoch": 58.31,
      "learning_rate": 1.2225760517799355e-05,
      "loss": 0.0625,
      "step": 150140
    },
    {
      "epoch": 58.31,
      "learning_rate": 1.2225242718446604e-05,
      "loss": 0.0387,
      "step": 150150
    },
    {
      "epoch": 58.31,
      "learning_rate": 1.222472491909385e-05,
      "loss": 0.117,
      "step": 150160
    },
    {
      "epoch": 58.32,
      "learning_rate": 1.22242071197411e-05,
      "loss": 0.2011,
      "step": 150170
    },
    {
      "epoch": 58.32,
      "learning_rate": 1.222368932038835e-05,
      "loss": 0.0756,
      "step": 150180
    },
    {
      "epoch": 58.33,
      "learning_rate": 1.22231715210356e-05,
      "loss": 0.0006,
      "step": 150190
    },
    {
      "epoch": 58.33,
      "learning_rate": 1.222265372168285e-05,
      "loss": 0.1178,
      "step": 150200
    },
    {
      "epoch": 58.33,
      "learning_rate": 1.2222135922330099e-05,
      "loss": 0.1524,
      "step": 150210
    },
    {
      "epoch": 58.34,
      "learning_rate": 1.2221618122977348e-05,
      "loss": 0.0763,
      "step": 150220
    },
    {
      "epoch": 58.34,
      "learning_rate": 1.2221100323624598e-05,
      "loss": 0.0164,
      "step": 150230
    },
    {
      "epoch": 58.35,
      "learning_rate": 1.2220582524271846e-05,
      "loss": 0.1098,
      "step": 150240
    },
    {
      "epoch": 58.35,
      "learning_rate": 1.2220064724919094e-05,
      "loss": 0.0699,
      "step": 150250
    },
    {
      "epoch": 58.35,
      "learning_rate": 1.2219546925566344e-05,
      "loss": 0.1229,
      "step": 150260
    },
    {
      "epoch": 58.36,
      "learning_rate": 1.2219029126213593e-05,
      "loss": 0.1069,
      "step": 150270
    },
    {
      "epoch": 58.36,
      "learning_rate": 1.2218511326860843e-05,
      "loss": 0.0643,
      "step": 150280
    },
    {
      "epoch": 58.37,
      "learning_rate": 1.2217993527508092e-05,
      "loss": 0.0125,
      "step": 150290
    },
    {
      "epoch": 58.37,
      "learning_rate": 1.2217475728155342e-05,
      "loss": 0.031,
      "step": 150300
    },
    {
      "epoch": 58.37,
      "learning_rate": 1.2216957928802592e-05,
      "loss": 0.1094,
      "step": 150310
    },
    {
      "epoch": 58.38,
      "learning_rate": 1.221644012944984e-05,
      "loss": 0.0484,
      "step": 150320
    },
    {
      "epoch": 58.38,
      "learning_rate": 1.2215922330097087e-05,
      "loss": 0.0433,
      "step": 150330
    },
    {
      "epoch": 58.38,
      "learning_rate": 1.2215404530744337e-05,
      "loss": 0.0893,
      "step": 150340
    },
    {
      "epoch": 58.39,
      "learning_rate": 1.2214886731391587e-05,
      "loss": 0.0823,
      "step": 150350
    },
    {
      "epoch": 58.39,
      "learning_rate": 1.2214368932038836e-05,
      "loss": 0.0833,
      "step": 150360
    },
    {
      "epoch": 58.4,
      "learning_rate": 1.2213851132686086e-05,
      "loss": 0.2186,
      "step": 150370
    },
    {
      "epoch": 58.4,
      "learning_rate": 1.2213333333333336e-05,
      "loss": 0.1108,
      "step": 150380
    },
    {
      "epoch": 58.4,
      "learning_rate": 1.2212815533980583e-05,
      "loss": 0.0427,
      "step": 150390
    },
    {
      "epoch": 58.41,
      "learning_rate": 1.2212297734627833e-05,
      "loss": 0.0596,
      "step": 150400
    },
    {
      "epoch": 58.41,
      "learning_rate": 1.2211779935275081e-05,
      "loss": 0.0315,
      "step": 150410
    },
    {
      "epoch": 58.42,
      "learning_rate": 1.221126213592233e-05,
      "loss": 0.1382,
      "step": 150420
    },
    {
      "epoch": 58.42,
      "learning_rate": 1.221074433656958e-05,
      "loss": 0.0009,
      "step": 150430
    },
    {
      "epoch": 58.42,
      "learning_rate": 1.221022653721683e-05,
      "loss": 0.0443,
      "step": 150440
    },
    {
      "epoch": 58.43,
      "learning_rate": 1.220970873786408e-05,
      "loss": 0.1066,
      "step": 150450
    },
    {
      "epoch": 58.43,
      "learning_rate": 1.2209190938511329e-05,
      "loss": 0.1195,
      "step": 150460
    },
    {
      "epoch": 58.43,
      "learning_rate": 1.2208673139158577e-05,
      "loss": 0.069,
      "step": 150470
    },
    {
      "epoch": 58.44,
      "learning_rate": 1.2208155339805827e-05,
      "loss": 0.0019,
      "step": 150480
    },
    {
      "epoch": 58.44,
      "learning_rate": 1.2207637540453075e-05,
      "loss": 0.0159,
      "step": 150490
    },
    {
      "epoch": 58.45,
      "learning_rate": 1.2207119741100324e-05,
      "loss": 0.0405,
      "step": 150500
    },
    {
      "epoch": 58.45,
      "learning_rate": 1.2206601941747574e-05,
      "loss": 0.0597,
      "step": 150510
    },
    {
      "epoch": 58.45,
      "learning_rate": 1.2206084142394823e-05,
      "loss": 0.2137,
      "step": 150520
    },
    {
      "epoch": 58.46,
      "learning_rate": 1.2205566343042073e-05,
      "loss": 0.0517,
      "step": 150530
    },
    {
      "epoch": 58.46,
      "learning_rate": 1.2205048543689321e-05,
      "loss": 0.0664,
      "step": 150540
    },
    {
      "epoch": 58.47,
      "learning_rate": 1.220453074433657e-05,
      "loss": 0.0418,
      "step": 150550
    },
    {
      "epoch": 58.47,
      "learning_rate": 1.220401294498382e-05,
      "loss": 0.1804,
      "step": 150560
    },
    {
      "epoch": 58.47,
      "learning_rate": 1.2203495145631068e-05,
      "loss": 0.0673,
      "step": 150570
    },
    {
      "epoch": 58.48,
      "learning_rate": 1.2202977346278318e-05,
      "loss": 0.0698,
      "step": 150580
    },
    {
      "epoch": 58.48,
      "learning_rate": 1.2202459546925567e-05,
      "loss": 0.0622,
      "step": 150590
    },
    {
      "epoch": 58.49,
      "learning_rate": 1.2201941747572817e-05,
      "loss": 0.0381,
      "step": 150600
    },
    {
      "epoch": 58.49,
      "learning_rate": 1.2201423948220067e-05,
      "loss": 0.0189,
      "step": 150610
    },
    {
      "epoch": 58.49,
      "learning_rate": 1.2200906148867315e-05,
      "loss": 0.0909,
      "step": 150620
    },
    {
      "epoch": 58.5,
      "learning_rate": 1.2200388349514564e-05,
      "loss": 0.0834,
      "step": 150630
    },
    {
      "epoch": 58.5,
      "learning_rate": 1.2199870550161814e-05,
      "loss": 0.0279,
      "step": 150640
    },
    {
      "epoch": 58.5,
      "learning_rate": 1.2199352750809062e-05,
      "loss": 0.0199,
      "step": 150650
    },
    {
      "epoch": 58.51,
      "learning_rate": 1.2198834951456311e-05,
      "loss": 0.0279,
      "step": 150660
    },
    {
      "epoch": 58.51,
      "learning_rate": 1.2198317152103561e-05,
      "loss": 0.289,
      "step": 150670
    },
    {
      "epoch": 58.52,
      "learning_rate": 1.219779935275081e-05,
      "loss": 0.0265,
      "step": 150680
    },
    {
      "epoch": 58.52,
      "learning_rate": 1.2197281553398058e-05,
      "loss": 0.0402,
      "step": 150690
    },
    {
      "epoch": 58.52,
      "learning_rate": 1.2196763754045308e-05,
      "loss": 0.1009,
      "step": 150700
    },
    {
      "epoch": 58.53,
      "learning_rate": 1.2196245954692558e-05,
      "loss": 0.0015,
      "step": 150710
    },
    {
      "epoch": 58.53,
      "learning_rate": 1.2195728155339807e-05,
      "loss": 0.1013,
      "step": 150720
    },
    {
      "epoch": 58.54,
      "learning_rate": 1.2195210355987055e-05,
      "loss": 0.0879,
      "step": 150730
    },
    {
      "epoch": 58.54,
      "learning_rate": 1.2194692556634305e-05,
      "loss": 0.0666,
      "step": 150740
    },
    {
      "epoch": 58.54,
      "learning_rate": 1.2194174757281554e-05,
      "loss": 0.0329,
      "step": 150750
    },
    {
      "epoch": 58.55,
      "learning_rate": 1.2193656957928804e-05,
      "loss": 0.097,
      "step": 150760
    },
    {
      "epoch": 58.55,
      "learning_rate": 1.2193139158576052e-05,
      "loss": 0.0225,
      "step": 150770
    },
    {
      "epoch": 58.56,
      "learning_rate": 1.2192621359223302e-05,
      "loss": 0.0553,
      "step": 150780
    },
    {
      "epoch": 58.56,
      "learning_rate": 1.2192103559870551e-05,
      "loss": 0.1021,
      "step": 150790
    },
    {
      "epoch": 58.56,
      "learning_rate": 1.2191585760517801e-05,
      "loss": 0.0382,
      "step": 150800
    },
    {
      "epoch": 58.57,
      "learning_rate": 1.219106796116505e-05,
      "loss": 0.1127,
      "step": 150810
    },
    {
      "epoch": 58.57,
      "learning_rate": 1.2190550161812298e-05,
      "loss": 0.0589,
      "step": 150820
    },
    {
      "epoch": 58.57,
      "learning_rate": 1.2190032362459548e-05,
      "loss": 0.0648,
      "step": 150830
    },
    {
      "epoch": 58.58,
      "learning_rate": 1.2189514563106798e-05,
      "loss": 0.0057,
      "step": 150840
    },
    {
      "epoch": 58.58,
      "learning_rate": 1.2188996763754046e-05,
      "loss": 0.0047,
      "step": 150850
    },
    {
      "epoch": 58.59,
      "learning_rate": 1.2188478964401295e-05,
      "loss": 0.1292,
      "step": 150860
    },
    {
      "epoch": 58.59,
      "learning_rate": 1.2187961165048545e-05,
      "loss": 0.0095,
      "step": 150870
    },
    {
      "epoch": 58.59,
      "learning_rate": 1.2187443365695794e-05,
      "loss": 0.2074,
      "step": 150880
    },
    {
      "epoch": 58.6,
      "learning_rate": 1.2186925566343044e-05,
      "loss": 0.0556,
      "step": 150890
    },
    {
      "epoch": 58.6,
      "learning_rate": 1.2186407766990292e-05,
      "loss": 0.0695,
      "step": 150900
    },
    {
      "epoch": 58.61,
      "learning_rate": 1.2185889967637542e-05,
      "loss": 0.2049,
      "step": 150910
    },
    {
      "epoch": 58.61,
      "learning_rate": 1.218537216828479e-05,
      "loss": 0.1577,
      "step": 150920
    },
    {
      "epoch": 58.61,
      "learning_rate": 1.2184854368932039e-05,
      "loss": 0.0471,
      "step": 150930
    },
    {
      "epoch": 58.62,
      "learning_rate": 1.2184336569579289e-05,
      "loss": 0.1259,
      "step": 150940
    },
    {
      "epoch": 58.62,
      "learning_rate": 1.2183818770226538e-05,
      "loss": 0.0359,
      "step": 150950
    },
    {
      "epoch": 58.63,
      "learning_rate": 1.2183300970873788e-05,
      "loss": 0.1161,
      "step": 150960
    },
    {
      "epoch": 58.63,
      "learning_rate": 1.2182783171521038e-05,
      "loss": 0.1067,
      "step": 150970
    },
    {
      "epoch": 58.63,
      "learning_rate": 1.2182265372168286e-05,
      "loss": 0.0598,
      "step": 150980
    },
    {
      "epoch": 58.64,
      "learning_rate": 1.2181747572815535e-05,
      "loss": 0.0471,
      "step": 150990
    },
    {
      "epoch": 58.64,
      "learning_rate": 1.2181229773462783e-05,
      "loss": 0.0836,
      "step": 151000
    },
    {
      "epoch": 58.64,
      "learning_rate": 1.2180711974110033e-05,
      "loss": 0.0427,
      "step": 151010
    },
    {
      "epoch": 58.65,
      "learning_rate": 1.2180194174757282e-05,
      "loss": 0.1646,
      "step": 151020
    },
    {
      "epoch": 58.65,
      "learning_rate": 1.2179676375404532e-05,
      "loss": 0.0452,
      "step": 151030
    },
    {
      "epoch": 58.66,
      "learning_rate": 1.2179158576051782e-05,
      "loss": 0.0509,
      "step": 151040
    },
    {
      "epoch": 58.66,
      "learning_rate": 1.2178640776699031e-05,
      "loss": 0.0709,
      "step": 151050
    },
    {
      "epoch": 58.66,
      "learning_rate": 1.2178122977346279e-05,
      "loss": 0.1116,
      "step": 151060
    },
    {
      "epoch": 58.67,
      "learning_rate": 1.2177605177993527e-05,
      "loss": 0.0455,
      "step": 151070
    },
    {
      "epoch": 58.67,
      "learning_rate": 1.2177087378640777e-05,
      "loss": 0.038,
      "step": 151080
    },
    {
      "epoch": 58.68,
      "learning_rate": 1.2176569579288026e-05,
      "loss": 0.0334,
      "step": 151090
    },
    {
      "epoch": 58.68,
      "learning_rate": 1.2176051779935276e-05,
      "loss": 0.1564,
      "step": 151100
    },
    {
      "epoch": 58.68,
      "learning_rate": 1.2175533980582525e-05,
      "loss": 0.0217,
      "step": 151110
    },
    {
      "epoch": 58.69,
      "learning_rate": 1.2175016181229775e-05,
      "loss": 0.1302,
      "step": 151120
    },
    {
      "epoch": 58.69,
      "learning_rate": 1.2174498381877025e-05,
      "loss": 0.0406,
      "step": 151130
    },
    {
      "epoch": 58.7,
      "learning_rate": 1.2173980582524273e-05,
      "loss": 0.0766,
      "step": 151140
    },
    {
      "epoch": 58.7,
      "learning_rate": 1.217346278317152e-05,
      "loss": 0.0921,
      "step": 151150
    },
    {
      "epoch": 58.7,
      "learning_rate": 1.217294498381877e-05,
      "loss": 0.0572,
      "step": 151160
    },
    {
      "epoch": 58.71,
      "learning_rate": 1.217242718446602e-05,
      "loss": 0.0255,
      "step": 151170
    },
    {
      "epoch": 58.71,
      "learning_rate": 1.217190938511327e-05,
      "loss": 0.1565,
      "step": 151180
    },
    {
      "epoch": 58.71,
      "learning_rate": 1.2171391585760519e-05,
      "loss": 0.0427,
      "step": 151190
    },
    {
      "epoch": 58.72,
      "learning_rate": 1.2170873786407769e-05,
      "loss": 0.0621,
      "step": 151200
    },
    {
      "epoch": 58.72,
      "learning_rate": 1.2170355987055018e-05,
      "loss": 0.1852,
      "step": 151210
    },
    {
      "epoch": 58.73,
      "learning_rate": 1.2169838187702264e-05,
      "loss": 0.0323,
      "step": 151220
    },
    {
      "epoch": 58.73,
      "learning_rate": 1.2169320388349514e-05,
      "loss": 0.1399,
      "step": 151230
    },
    {
      "epoch": 58.73,
      "learning_rate": 1.2168802588996764e-05,
      "loss": 0.1172,
      "step": 151240
    },
    {
      "epoch": 58.74,
      "learning_rate": 1.2168284789644013e-05,
      "loss": 0.0956,
      "step": 151250
    },
    {
      "epoch": 58.74,
      "learning_rate": 1.2167766990291263e-05,
      "loss": 0.0557,
      "step": 151260
    },
    {
      "epoch": 58.75,
      "learning_rate": 1.2167249190938513e-05,
      "loss": 0.0569,
      "step": 151270
    },
    {
      "epoch": 58.75,
      "learning_rate": 1.2166731391585762e-05,
      "loss": 0.0167,
      "step": 151280
    },
    {
      "epoch": 58.75,
      "learning_rate": 1.2166213592233012e-05,
      "loss": 0.0582,
      "step": 151290
    },
    {
      "epoch": 58.76,
      "learning_rate": 1.2165695792880258e-05,
      "loss": 0.0549,
      "step": 151300
    },
    {
      "epoch": 58.76,
      "learning_rate": 1.2165177993527508e-05,
      "loss": 0.0249,
      "step": 151310
    },
    {
      "epoch": 58.77,
      "learning_rate": 1.2164660194174757e-05,
      "loss": 0.0724,
      "step": 151320
    },
    {
      "epoch": 58.77,
      "learning_rate": 1.2164142394822007e-05,
      "loss": 0.1405,
      "step": 151330
    },
    {
      "epoch": 58.77,
      "learning_rate": 1.2163624595469257e-05,
      "loss": 0.0411,
      "step": 151340
    },
    {
      "epoch": 58.78,
      "learning_rate": 1.2163106796116506e-05,
      "loss": 0.0612,
      "step": 151350
    },
    {
      "epoch": 58.78,
      "learning_rate": 1.2162588996763756e-05,
      "loss": 0.005,
      "step": 151360
    },
    {
      "epoch": 58.78,
      "learning_rate": 1.2162071197411005e-05,
      "loss": 0.0424,
      "step": 151370
    },
    {
      "epoch": 58.79,
      "learning_rate": 1.2161553398058255e-05,
      "loss": 0.1032,
      "step": 151380
    },
    {
      "epoch": 58.79,
      "learning_rate": 1.2161035598705501e-05,
      "loss": 0.1223,
      "step": 151390
    },
    {
      "epoch": 58.8,
      "learning_rate": 1.2160517799352751e-05,
      "loss": 0.0095,
      "step": 151400
    },
    {
      "epoch": 58.8,
      "learning_rate": 1.216e-05,
      "loss": 0.0519,
      "step": 151410
    },
    {
      "epoch": 58.8,
      "learning_rate": 1.215948220064725e-05,
      "loss": 0.0576,
      "step": 151420
    },
    {
      "epoch": 58.81,
      "learning_rate": 1.21589644012945e-05,
      "loss": 0.0095,
      "step": 151430
    },
    {
      "epoch": 58.81,
      "learning_rate": 1.215844660194175e-05,
      "loss": 0.1768,
      "step": 151440
    },
    {
      "epoch": 58.82,
      "learning_rate": 1.2157928802588999e-05,
      "loss": 0.0222,
      "step": 151450
    },
    {
      "epoch": 58.82,
      "learning_rate": 1.2157411003236249e-05,
      "loss": 0.0005,
      "step": 151460
    },
    {
      "epoch": 58.82,
      "learning_rate": 1.2156893203883495e-05,
      "loss": 0.0159,
      "step": 151470
    },
    {
      "epoch": 58.83,
      "learning_rate": 1.2156375404530744e-05,
      "loss": 0.0107,
      "step": 151480
    },
    {
      "epoch": 58.83,
      "learning_rate": 1.2155857605177994e-05,
      "loss": 0.0186,
      "step": 151490
    },
    {
      "epoch": 58.83,
      "learning_rate": 1.2155339805825244e-05,
      "loss": 0.1654,
      "step": 151500
    },
    {
      "epoch": 58.84,
      "learning_rate": 1.2154822006472493e-05,
      "loss": 0.0557,
      "step": 151510
    },
    {
      "epoch": 58.84,
      "learning_rate": 1.2154304207119743e-05,
      "loss": 0.1145,
      "step": 151520
    },
    {
      "epoch": 58.85,
      "learning_rate": 1.2153786407766992e-05,
      "loss": 0.1287,
      "step": 151530
    },
    {
      "epoch": 58.85,
      "learning_rate": 1.2153268608414242e-05,
      "loss": 0.0684,
      "step": 151540
    },
    {
      "epoch": 58.85,
      "learning_rate": 1.2152750809061488e-05,
      "loss": 0.0016,
      "step": 151550
    },
    {
      "epoch": 58.86,
      "learning_rate": 1.2152233009708738e-05,
      "loss": 0.0752,
      "step": 151560
    },
    {
      "epoch": 58.86,
      "learning_rate": 1.2151715210355988e-05,
      "loss": 0.0323,
      "step": 151570
    },
    {
      "epoch": 58.87,
      "learning_rate": 1.2151197411003237e-05,
      "loss": 0.0292,
      "step": 151580
    },
    {
      "epoch": 58.87,
      "learning_rate": 1.2150679611650487e-05,
      "loss": 0.1222,
      "step": 151590
    },
    {
      "epoch": 58.87,
      "learning_rate": 1.2150161812297736e-05,
      "loss": 0.1622,
      "step": 151600
    },
    {
      "epoch": 58.88,
      "learning_rate": 1.2149644012944986e-05,
      "loss": 0.1783,
      "step": 151610
    },
    {
      "epoch": 58.88,
      "learning_rate": 1.2149126213592236e-05,
      "loss": 0.0229,
      "step": 151620
    },
    {
      "epoch": 58.89,
      "learning_rate": 1.2148608414239482e-05,
      "loss": 0.1725,
      "step": 151630
    },
    {
      "epoch": 58.89,
      "learning_rate": 1.2148090614886732e-05,
      "loss": 0.0427,
      "step": 151640
    },
    {
      "epoch": 58.89,
      "learning_rate": 1.2147572815533981e-05,
      "loss": 0.1389,
      "step": 151650
    },
    {
      "epoch": 58.9,
      "learning_rate": 1.214705501618123e-05,
      "loss": 0.2184,
      "step": 151660
    },
    {
      "epoch": 58.9,
      "learning_rate": 1.214653721682848e-05,
      "loss": 0.1255,
      "step": 151670
    },
    {
      "epoch": 58.9,
      "learning_rate": 1.214601941747573e-05,
      "loss": 0.0273,
      "step": 151680
    },
    {
      "epoch": 58.91,
      "learning_rate": 1.214550161812298e-05,
      "loss": 0.0318,
      "step": 151690
    },
    {
      "epoch": 58.91,
      "learning_rate": 1.214498381877023e-05,
      "loss": 0.0353,
      "step": 151700
    },
    {
      "epoch": 58.92,
      "learning_rate": 1.2144466019417475e-05,
      "loss": 0.2038,
      "step": 151710
    },
    {
      "epoch": 58.92,
      "learning_rate": 1.2143948220064725e-05,
      "loss": 0.1579,
      "step": 151720
    },
    {
      "epoch": 58.92,
      "learning_rate": 1.2143430420711975e-05,
      "loss": 0.1848,
      "step": 151730
    },
    {
      "epoch": 58.93,
      "learning_rate": 1.2142912621359224e-05,
      "loss": 0.0869,
      "step": 151740
    },
    {
      "epoch": 58.93,
      "learning_rate": 1.2142394822006474e-05,
      "loss": 0.0465,
      "step": 151750
    },
    {
      "epoch": 58.94,
      "learning_rate": 1.2141877022653724e-05,
      "loss": 0.0352,
      "step": 151760
    },
    {
      "epoch": 58.94,
      "learning_rate": 1.2141359223300973e-05,
      "loss": 0.0352,
      "step": 151770
    },
    {
      "epoch": 58.94,
      "learning_rate": 1.2140841423948223e-05,
      "loss": 0.0038,
      "step": 151780
    },
    {
      "epoch": 58.95,
      "learning_rate": 1.2140323624595469e-05,
      "loss": 0.1958,
      "step": 151790
    },
    {
      "epoch": 58.95,
      "learning_rate": 1.2139805825242719e-05,
      "loss": 0.0096,
      "step": 151800
    },
    {
      "epoch": 58.96,
      "learning_rate": 1.2139288025889968e-05,
      "loss": 0.0044,
      "step": 151810
    },
    {
      "epoch": 58.96,
      "learning_rate": 1.2138770226537218e-05,
      "loss": 0.1161,
      "step": 151820
    },
    {
      "epoch": 58.96,
      "learning_rate": 1.2138252427184467e-05,
      "loss": 0.0186,
      "step": 151830
    },
    {
      "epoch": 58.97,
      "learning_rate": 1.2137734627831717e-05,
      "loss": 0.0021,
      "step": 151840
    },
    {
      "epoch": 58.97,
      "learning_rate": 1.2137216828478967e-05,
      "loss": 0.0089,
      "step": 151850
    },
    {
      "epoch": 58.97,
      "learning_rate": 1.2136699029126215e-05,
      "loss": 0.1239,
      "step": 151860
    },
    {
      "epoch": 58.98,
      "learning_rate": 1.2136181229773463e-05,
      "loss": 0.0552,
      "step": 151870
    },
    {
      "epoch": 58.98,
      "learning_rate": 1.2135663430420712e-05,
      "loss": 0.0093,
      "step": 151880
    },
    {
      "epoch": 58.99,
      "learning_rate": 1.2135145631067962e-05,
      "loss": 0.0455,
      "step": 151890
    },
    {
      "epoch": 58.99,
      "learning_rate": 1.2134627831715211e-05,
      "loss": 0.0009,
      "step": 151900
    },
    {
      "epoch": 58.99,
      "learning_rate": 1.2134110032362461e-05,
      "loss": 0.0319,
      "step": 151910
    },
    {
      "epoch": 59.0,
      "learning_rate": 1.213359223300971e-05,
      "loss": 0.0184,
      "step": 151920
    },
    {
      "epoch": 59.0,
      "eval_accuracy": 0.953232462173315,
      "eval_loss": 0.31959620118141174,
      "eval_runtime": 8.2236,
      "eval_samples_per_second": 442.019,
      "eval_steps_per_second": 55.328,
      "step": 151925
    },
    {
      "epoch": 59.0,
      "learning_rate": 1.213307443365696e-05,
      "loss": 0.0603,
      "step": 151930
    },
    {
      "epoch": 59.01,
      "learning_rate": 1.2132556634304208e-05,
      "loss": 0.0134,
      "step": 151940
    },
    {
      "epoch": 59.01,
      "learning_rate": 1.2132038834951458e-05,
      "loss": 0.0138,
      "step": 151950
    },
    {
      "epoch": 59.01,
      "learning_rate": 1.2131521035598706e-05,
      "loss": 0.0068,
      "step": 151960
    },
    {
      "epoch": 59.02,
      "learning_rate": 1.2131003236245955e-05,
      "loss": 0.0337,
      "step": 151970
    },
    {
      "epoch": 59.02,
      "learning_rate": 1.2130485436893205e-05,
      "loss": 0.031,
      "step": 151980
    },
    {
      "epoch": 59.03,
      "learning_rate": 1.2129967637540455e-05,
      "loss": 0.2306,
      "step": 151990
    },
    {
      "epoch": 59.03,
      "learning_rate": 1.2129449838187704e-05,
      "loss": 0.078,
      "step": 152000
    },
    {
      "epoch": 59.03,
      "learning_rate": 1.2128932038834952e-05,
      "loss": 0.0008,
      "step": 152010
    },
    {
      "epoch": 59.04,
      "learning_rate": 1.2128414239482202e-05,
      "loss": 0.2069,
      "step": 152020
    },
    {
      "epoch": 59.04,
      "learning_rate": 1.2127896440129451e-05,
      "loss": 0.0187,
      "step": 152030
    },
    {
      "epoch": 59.04,
      "learning_rate": 1.21273786407767e-05,
      "loss": 0.0983,
      "step": 152040
    },
    {
      "epoch": 59.05,
      "learning_rate": 1.2126860841423949e-05,
      "loss": 0.0943,
      "step": 152050
    },
    {
      "epoch": 59.05,
      "learning_rate": 1.2126343042071199e-05,
      "loss": 0.1395,
      "step": 152060
    },
    {
      "epoch": 59.06,
      "learning_rate": 1.2125825242718448e-05,
      "loss": 0.1371,
      "step": 152070
    },
    {
      "epoch": 59.06,
      "learning_rate": 1.2125307443365698e-05,
      "loss": 0.0354,
      "step": 152080
    },
    {
      "epoch": 59.06,
      "learning_rate": 1.2124789644012946e-05,
      "loss": 0.0757,
      "step": 152090
    },
    {
      "epoch": 59.07,
      "learning_rate": 1.2124271844660195e-05,
      "loss": 0.18,
      "step": 152100
    },
    {
      "epoch": 59.07,
      "learning_rate": 1.2123754045307445e-05,
      "loss": 0.0141,
      "step": 152110
    },
    {
      "epoch": 59.08,
      "learning_rate": 1.2123236245954693e-05,
      "loss": 0.0078,
      "step": 152120
    },
    {
      "epoch": 59.08,
      "learning_rate": 1.2122718446601942e-05,
      "loss": 0.2501,
      "step": 152130
    },
    {
      "epoch": 59.08,
      "learning_rate": 1.2122200647249192e-05,
      "loss": 0.0062,
      "step": 152140
    },
    {
      "epoch": 59.09,
      "learning_rate": 1.2121682847896442e-05,
      "loss": 0.0942,
      "step": 152150
    },
    {
      "epoch": 59.09,
      "learning_rate": 1.212116504854369e-05,
      "loss": 0.2062,
      "step": 152160
    },
    {
      "epoch": 59.1,
      "learning_rate": 1.212064724919094e-05,
      "loss": 0.0428,
      "step": 152170
    },
    {
      "epoch": 59.1,
      "learning_rate": 1.2120129449838189e-05,
      "loss": 0.0624,
      "step": 152180
    },
    {
      "epoch": 59.1,
      "learning_rate": 1.2119611650485438e-05,
      "loss": 0.1801,
      "step": 152190
    },
    {
      "epoch": 59.11,
      "learning_rate": 1.2119093851132686e-05,
      "loss": 0.074,
      "step": 152200
    },
    {
      "epoch": 59.11,
      "learning_rate": 1.2118576051779936e-05,
      "loss": 0.0187,
      "step": 152210
    },
    {
      "epoch": 59.11,
      "learning_rate": 1.2118058252427186e-05,
      "loss": 0.1719,
      "step": 152220
    },
    {
      "epoch": 59.12,
      "learning_rate": 1.2117540453074435e-05,
      "loss": 0.0298,
      "step": 152230
    },
    {
      "epoch": 59.12,
      "learning_rate": 1.2117022653721683e-05,
      "loss": 0.0461,
      "step": 152240
    },
    {
      "epoch": 59.13,
      "learning_rate": 1.2116504854368933e-05,
      "loss": 0.0421,
      "step": 152250
    },
    {
      "epoch": 59.13,
      "learning_rate": 1.2115987055016182e-05,
      "loss": 0.1055,
      "step": 152260
    },
    {
      "epoch": 59.13,
      "learning_rate": 1.2115469255663432e-05,
      "loss": 0.0572,
      "step": 152270
    },
    {
      "epoch": 59.14,
      "learning_rate": 1.211495145631068e-05,
      "loss": 0.008,
      "step": 152280
    },
    {
      "epoch": 59.14,
      "learning_rate": 1.211443365695793e-05,
      "loss": 0.0329,
      "step": 152290
    },
    {
      "epoch": 59.15,
      "learning_rate": 1.211391585760518e-05,
      "loss": 0.0147,
      "step": 152300
    },
    {
      "epoch": 59.15,
      "learning_rate": 1.2113398058252429e-05,
      "loss": 0.1041,
      "step": 152310
    },
    {
      "epoch": 59.15,
      "learning_rate": 1.2112880258899677e-05,
      "loss": 0.0217,
      "step": 152320
    },
    {
      "epoch": 59.16,
      "learning_rate": 1.2112362459546926e-05,
      "loss": 0.1349,
      "step": 152330
    },
    {
      "epoch": 59.16,
      "learning_rate": 1.2111844660194176e-05,
      "loss": 0.0035,
      "step": 152340
    },
    {
      "epoch": 59.17,
      "learning_rate": 1.2111326860841426e-05,
      "loss": 0.0319,
      "step": 152350
    },
    {
      "epoch": 59.17,
      "learning_rate": 1.2110809061488674e-05,
      "loss": 0.0734,
      "step": 152360
    },
    {
      "epoch": 59.17,
      "learning_rate": 1.2110291262135923e-05,
      "loss": 0.0145,
      "step": 152370
    },
    {
      "epoch": 59.18,
      "learning_rate": 1.2109773462783173e-05,
      "loss": 0.0219,
      "step": 152380
    },
    {
      "epoch": 59.18,
      "learning_rate": 1.210925566343042e-05,
      "loss": 0.0104,
      "step": 152390
    },
    {
      "epoch": 59.18,
      "learning_rate": 1.210873786407767e-05,
      "loss": 0.023,
      "step": 152400
    },
    {
      "epoch": 59.19,
      "learning_rate": 1.210822006472492e-05,
      "loss": 0.0227,
      "step": 152410
    },
    {
      "epoch": 59.19,
      "learning_rate": 1.210770226537217e-05,
      "loss": 0.0475,
      "step": 152420
    },
    {
      "epoch": 59.2,
      "learning_rate": 1.210718446601942e-05,
      "loss": 0.0388,
      "step": 152430
    },
    {
      "epoch": 59.2,
      "learning_rate": 1.2106666666666667e-05,
      "loss": 0.1069,
      "step": 152440
    },
    {
      "epoch": 59.2,
      "learning_rate": 1.2106148867313917e-05,
      "loss": 0.1984,
      "step": 152450
    },
    {
      "epoch": 59.21,
      "learning_rate": 1.2105631067961166e-05,
      "loss": 0.0755,
      "step": 152460
    },
    {
      "epoch": 59.21,
      "learning_rate": 1.2105113268608414e-05,
      "loss": 0.1003,
      "step": 152470
    },
    {
      "epoch": 59.22,
      "learning_rate": 1.2104595469255664e-05,
      "loss": 0.0474,
      "step": 152480
    },
    {
      "epoch": 59.22,
      "learning_rate": 1.2104077669902913e-05,
      "loss": 0.1109,
      "step": 152490
    },
    {
      "epoch": 59.22,
      "learning_rate": 1.2103559870550163e-05,
      "loss": 0.1048,
      "step": 152500
    },
    {
      "epoch": 59.23,
      "learning_rate": 1.2103042071197413e-05,
      "loss": 0.0873,
      "step": 152510
    },
    {
      "epoch": 59.23,
      "learning_rate": 1.2102524271844662e-05,
      "loss": 0.1442,
      "step": 152520
    },
    {
      "epoch": 59.23,
      "learning_rate": 1.210200647249191e-05,
      "loss": 0.0271,
      "step": 152530
    },
    {
      "epoch": 59.24,
      "learning_rate": 1.2101488673139158e-05,
      "loss": 0.0509,
      "step": 152540
    },
    {
      "epoch": 59.24,
      "learning_rate": 1.2100970873786408e-05,
      "loss": 0.1766,
      "step": 152550
    },
    {
      "epoch": 59.25,
      "learning_rate": 1.2100453074433657e-05,
      "loss": 0.0083,
      "step": 152560
    },
    {
      "epoch": 59.25,
      "learning_rate": 1.2099935275080907e-05,
      "loss": 0.0193,
      "step": 152570
    },
    {
      "epoch": 59.25,
      "learning_rate": 1.2099417475728157e-05,
      "loss": 0.0652,
      "step": 152580
    },
    {
      "epoch": 59.26,
      "learning_rate": 1.2098899676375406e-05,
      "loss": 0.0277,
      "step": 152590
    },
    {
      "epoch": 59.26,
      "learning_rate": 1.2098381877022656e-05,
      "loss": 0.1498,
      "step": 152600
    },
    {
      "epoch": 59.27,
      "learning_rate": 1.2097864077669904e-05,
      "loss": 0.0566,
      "step": 152610
    },
    {
      "epoch": 59.27,
      "learning_rate": 1.2097346278317152e-05,
      "loss": 0.0985,
      "step": 152620
    },
    {
      "epoch": 59.27,
      "learning_rate": 1.2096828478964401e-05,
      "loss": 0.081,
      "step": 152630
    },
    {
      "epoch": 59.28,
      "learning_rate": 1.2096310679611651e-05,
      "loss": 0.0307,
      "step": 152640
    },
    {
      "epoch": 59.28,
      "learning_rate": 1.20957928802589e-05,
      "loss": 0.0067,
      "step": 152650
    },
    {
      "epoch": 59.29,
      "learning_rate": 1.209527508090615e-05,
      "loss": 0.3483,
      "step": 152660
    },
    {
      "epoch": 59.29,
      "learning_rate": 1.20947572815534e-05,
      "loss": 0.1129,
      "step": 152670
    },
    {
      "epoch": 59.29,
      "learning_rate": 1.209423948220065e-05,
      "loss": 0.1295,
      "step": 152680
    },
    {
      "epoch": 59.3,
      "learning_rate": 1.2093721682847896e-05,
      "loss": 0.0596,
      "step": 152690
    },
    {
      "epoch": 59.3,
      "learning_rate": 1.2093203883495145e-05,
      "loss": 0.0333,
      "step": 152700
    },
    {
      "epoch": 59.3,
      "learning_rate": 1.2092686084142395e-05,
      "loss": 0.0018,
      "step": 152710
    },
    {
      "epoch": 59.31,
      "learning_rate": 1.2092168284789645e-05,
      "loss": 0.1021,
      "step": 152720
    },
    {
      "epoch": 59.31,
      "learning_rate": 1.2091650485436894e-05,
      "loss": 0.0783,
      "step": 152730
    },
    {
      "epoch": 59.32,
      "learning_rate": 1.2091132686084144e-05,
      "loss": 0.005,
      "step": 152740
    },
    {
      "epoch": 59.32,
      "learning_rate": 1.2090614886731393e-05,
      "loss": 0.0311,
      "step": 152750
    },
    {
      "epoch": 59.32,
      "learning_rate": 1.2090097087378643e-05,
      "loss": 0.0906,
      "step": 152760
    },
    {
      "epoch": 59.33,
      "learning_rate": 1.208957928802589e-05,
      "loss": 0.0768,
      "step": 152770
    },
    {
      "epoch": 59.33,
      "learning_rate": 1.2089061488673139e-05,
      "loss": 0.0922,
      "step": 152780
    },
    {
      "epoch": 59.34,
      "learning_rate": 1.2088543689320388e-05,
      "loss": 0.0687,
      "step": 152790
    },
    {
      "epoch": 59.34,
      "learning_rate": 1.2088025889967638e-05,
      "loss": 0.0565,
      "step": 152800
    },
    {
      "epoch": 59.34,
      "learning_rate": 1.2087508090614888e-05,
      "loss": 0.0081,
      "step": 152810
    },
    {
      "epoch": 59.35,
      "learning_rate": 1.2086990291262137e-05,
      "loss": 0.1294,
      "step": 152820
    },
    {
      "epoch": 59.35,
      "learning_rate": 1.2086472491909387e-05,
      "loss": 0.0948,
      "step": 152830
    },
    {
      "epoch": 59.36,
      "learning_rate": 1.2085954692556637e-05,
      "loss": 0.0596,
      "step": 152840
    },
    {
      "epoch": 59.36,
      "learning_rate": 1.2085436893203883e-05,
      "loss": 0.02,
      "step": 152850
    },
    {
      "epoch": 59.36,
      "learning_rate": 1.2084919093851132e-05,
      "loss": 0.0218,
      "step": 152860
    },
    {
      "epoch": 59.37,
      "learning_rate": 1.2084401294498382e-05,
      "loss": 0.1019,
      "step": 152870
    },
    {
      "epoch": 59.37,
      "learning_rate": 1.2083883495145632e-05,
      "loss": 0.0145,
      "step": 152880
    },
    {
      "epoch": 59.37,
      "learning_rate": 1.2083365695792881e-05,
      "loss": 0.0338,
      "step": 152890
    },
    {
      "epoch": 59.38,
      "learning_rate": 1.2082847896440131e-05,
      "loss": 0.182,
      "step": 152900
    },
    {
      "epoch": 59.38,
      "learning_rate": 1.208233009708738e-05,
      "loss": 0.1098,
      "step": 152910
    },
    {
      "epoch": 59.39,
      "learning_rate": 1.208181229773463e-05,
      "loss": 0.0614,
      "step": 152920
    },
    {
      "epoch": 59.39,
      "learning_rate": 1.2081294498381876e-05,
      "loss": 0.095,
      "step": 152930
    },
    {
      "epoch": 59.39,
      "learning_rate": 1.2080776699029126e-05,
      "loss": 0.0516,
      "step": 152940
    },
    {
      "epoch": 59.4,
      "learning_rate": 1.2080258899676376e-05,
      "loss": 0.0418,
      "step": 152950
    },
    {
      "epoch": 59.4,
      "learning_rate": 1.2079741100323625e-05,
      "loss": 0.1119,
      "step": 152960
    },
    {
      "epoch": 59.41,
      "learning_rate": 1.2079223300970875e-05,
      "loss": 0.081,
      "step": 152970
    },
    {
      "epoch": 59.41,
      "learning_rate": 1.2078705501618124e-05,
      "loss": 0.0806,
      "step": 152980
    },
    {
      "epoch": 59.41,
      "learning_rate": 1.2078187702265374e-05,
      "loss": 0.0879,
      "step": 152990
    },
    {
      "epoch": 59.42,
      "learning_rate": 1.2077669902912624e-05,
      "loss": 0.1427,
      "step": 153000
    },
    {
      "epoch": 59.42,
      "learning_rate": 1.207715210355987e-05,
      "loss": 0.1138,
      "step": 153010
    },
    {
      "epoch": 59.43,
      "learning_rate": 1.207663430420712e-05,
      "loss": 0.0969,
      "step": 153020
    },
    {
      "epoch": 59.43,
      "learning_rate": 1.2076116504854369e-05,
      "loss": 0.1263,
      "step": 153030
    },
    {
      "epoch": 59.43,
      "learning_rate": 1.2075598705501619e-05,
      "loss": 0.0225,
      "step": 153040
    },
    {
      "epoch": 59.44,
      "learning_rate": 1.2075080906148868e-05,
      "loss": 0.0341,
      "step": 153050
    },
    {
      "epoch": 59.44,
      "learning_rate": 1.2074563106796118e-05,
      "loss": 0.1899,
      "step": 153060
    },
    {
      "epoch": 59.44,
      "learning_rate": 1.2074045307443368e-05,
      "loss": 0.0341,
      "step": 153070
    },
    {
      "epoch": 59.45,
      "learning_rate": 1.2073527508090617e-05,
      "loss": 0.1021,
      "step": 153080
    },
    {
      "epoch": 59.45,
      "learning_rate": 1.2073009708737867e-05,
      "loss": 0.1404,
      "step": 153090
    },
    {
      "epoch": 59.46,
      "learning_rate": 1.2072491909385113e-05,
      "loss": 0.0061,
      "step": 153100
    },
    {
      "epoch": 59.46,
      "learning_rate": 1.2071974110032363e-05,
      "loss": 0.0319,
      "step": 153110
    },
    {
      "epoch": 59.46,
      "learning_rate": 1.2071456310679612e-05,
      "loss": 0.0422,
      "step": 153120
    },
    {
      "epoch": 59.47,
      "learning_rate": 1.2070938511326862e-05,
      "loss": 0.0264,
      "step": 153130
    },
    {
      "epoch": 59.47,
      "learning_rate": 1.2070420711974112e-05,
      "loss": 0.0556,
      "step": 153140
    },
    {
      "epoch": 59.48,
      "learning_rate": 1.2069902912621361e-05,
      "loss": 0.1722,
      "step": 153150
    },
    {
      "epoch": 59.48,
      "learning_rate": 1.206938511326861e-05,
      "loss": 0.0184,
      "step": 153160
    },
    {
      "epoch": 59.48,
      "learning_rate": 1.206886731391586e-05,
      "loss": 0.0783,
      "step": 153170
    },
    {
      "epoch": 59.49,
      "learning_rate": 1.2068349514563107e-05,
      "loss": 0.1757,
      "step": 153180
    },
    {
      "epoch": 59.49,
      "learning_rate": 1.2067831715210356e-05,
      "loss": 0.2306,
      "step": 153190
    },
    {
      "epoch": 59.5,
      "learning_rate": 1.2067313915857606e-05,
      "loss": 0.0476,
      "step": 153200
    },
    {
      "epoch": 59.5,
      "learning_rate": 1.2066796116504855e-05,
      "loss": 0.0105,
      "step": 153210
    },
    {
      "epoch": 59.5,
      "learning_rate": 1.2066278317152105e-05,
      "loss": 0.1132,
      "step": 153220
    },
    {
      "epoch": 59.51,
      "learning_rate": 1.2065760517799355e-05,
      "loss": 0.0639,
      "step": 153230
    },
    {
      "epoch": 59.51,
      "learning_rate": 1.2065242718446604e-05,
      "loss": 0.1108,
      "step": 153240
    },
    {
      "epoch": 59.51,
      "learning_rate": 1.2064724919093854e-05,
      "loss": 0.0825,
      "step": 153250
    },
    {
      "epoch": 59.52,
      "learning_rate": 1.20642071197411e-05,
      "loss": 0.0598,
      "step": 153260
    },
    {
      "epoch": 59.52,
      "learning_rate": 1.206368932038835e-05,
      "loss": 0.042,
      "step": 153270
    },
    {
      "epoch": 59.53,
      "learning_rate": 1.20631715210356e-05,
      "loss": 0.1506,
      "step": 153280
    },
    {
      "epoch": 59.53,
      "learning_rate": 1.2062653721682849e-05,
      "loss": 0.1066,
      "step": 153290
    },
    {
      "epoch": 59.53,
      "learning_rate": 1.2062135922330099e-05,
      "loss": 0.087,
      "step": 153300
    },
    {
      "epoch": 59.54,
      "learning_rate": 1.2061618122977348e-05,
      "loss": 0.0815,
      "step": 153310
    },
    {
      "epoch": 59.54,
      "learning_rate": 1.2061100323624598e-05,
      "loss": 0.0448,
      "step": 153320
    },
    {
      "epoch": 59.55,
      "learning_rate": 1.2060582524271846e-05,
      "loss": 0.0053,
      "step": 153330
    },
    {
      "epoch": 59.55,
      "learning_rate": 1.2060064724919094e-05,
      "loss": 0.015,
      "step": 153340
    },
    {
      "epoch": 59.55,
      "learning_rate": 1.2059546925566343e-05,
      "loss": 0.0288,
      "step": 153350
    },
    {
      "epoch": 59.56,
      "learning_rate": 1.2059029126213593e-05,
      "loss": 0.2195,
      "step": 153360
    },
    {
      "epoch": 59.56,
      "learning_rate": 1.2058511326860843e-05,
      "loss": 0.0525,
      "step": 153370
    },
    {
      "epoch": 59.57,
      "learning_rate": 1.2057993527508092e-05,
      "loss": 0.0339,
      "step": 153380
    },
    {
      "epoch": 59.57,
      "learning_rate": 1.2057475728155342e-05,
      "loss": 0.1352,
      "step": 153390
    },
    {
      "epoch": 59.57,
      "learning_rate": 1.2056957928802591e-05,
      "loss": 0.1025,
      "step": 153400
    },
    {
      "epoch": 59.58,
      "learning_rate": 1.205644012944984e-05,
      "loss": 0.0748,
      "step": 153410
    },
    {
      "epoch": 59.58,
      "learning_rate": 1.2055922330097087e-05,
      "loss": 0.0241,
      "step": 153420
    },
    {
      "epoch": 59.58,
      "learning_rate": 1.2055404530744337e-05,
      "loss": 0.0476,
      "step": 153430
    },
    {
      "epoch": 59.59,
      "learning_rate": 1.2054886731391587e-05,
      "loss": 0.0217,
      "step": 153440
    },
    {
      "epoch": 59.59,
      "learning_rate": 1.2054368932038836e-05,
      "loss": 0.0954,
      "step": 153450
    },
    {
      "epoch": 59.6,
      "learning_rate": 1.2053851132686086e-05,
      "loss": 0.0556,
      "step": 153460
    },
    {
      "epoch": 59.6,
      "learning_rate": 1.2053333333333335e-05,
      "loss": 0.0162,
      "step": 153470
    },
    {
      "epoch": 59.6,
      "learning_rate": 1.2052815533980583e-05,
      "loss": 0.0091,
      "step": 153480
    },
    {
      "epoch": 59.61,
      "learning_rate": 1.2052297734627833e-05,
      "loss": 0.1533,
      "step": 153490
    },
    {
      "epoch": 59.61,
      "learning_rate": 1.2051779935275081e-05,
      "loss": 0.1139,
      "step": 153500
    },
    {
      "epoch": 59.62,
      "learning_rate": 1.205126213592233e-05,
      "loss": 0.1378,
      "step": 153510
    },
    {
      "epoch": 59.62,
      "learning_rate": 1.205074433656958e-05,
      "loss": 0.0895,
      "step": 153520
    },
    {
      "epoch": 59.62,
      "learning_rate": 1.205022653721683e-05,
      "loss": 0.095,
      "step": 153530
    },
    {
      "epoch": 59.63,
      "learning_rate": 1.204970873786408e-05,
      "loss": 0.0676,
      "step": 153540
    },
    {
      "epoch": 59.63,
      "learning_rate": 1.2049190938511329e-05,
      "loss": 0.125,
      "step": 153550
    },
    {
      "epoch": 59.63,
      "learning_rate": 1.2048673139158577e-05,
      "loss": 0.1613,
      "step": 153560
    },
    {
      "epoch": 59.64,
      "learning_rate": 1.2048155339805826e-05,
      "loss": 0.0601,
      "step": 153570
    },
    {
      "epoch": 59.64,
      "learning_rate": 1.2047637540453074e-05,
      "loss": 0.1631,
      "step": 153580
    },
    {
      "epoch": 59.65,
      "learning_rate": 1.2047119741100324e-05,
      "loss": 0.1871,
      "step": 153590
    },
    {
      "epoch": 59.65,
      "learning_rate": 1.2046601941747574e-05,
      "loss": 0.1252,
      "step": 153600
    },
    {
      "epoch": 59.65,
      "learning_rate": 1.2046084142394823e-05,
      "loss": 0.197,
      "step": 153610
    },
    {
      "epoch": 59.66,
      "learning_rate": 1.2045566343042073e-05,
      "loss": 0.016,
      "step": 153620
    },
    {
      "epoch": 59.66,
      "learning_rate": 1.204504854368932e-05,
      "loss": 0.0026,
      "step": 153630
    },
    {
      "epoch": 59.67,
      "learning_rate": 1.204453074433657e-05,
      "loss": 0.0565,
      "step": 153640
    },
    {
      "epoch": 59.67,
      "learning_rate": 1.204401294498382e-05,
      "loss": 0.0334,
      "step": 153650
    },
    {
      "epoch": 59.67,
      "learning_rate": 1.204349514563107e-05,
      "loss": 0.2449,
      "step": 153660
    },
    {
      "epoch": 59.68,
      "learning_rate": 1.2042977346278318e-05,
      "loss": 0.0606,
      "step": 153670
    },
    {
      "epoch": 59.68,
      "learning_rate": 1.2042459546925567e-05,
      "loss": 0.1644,
      "step": 153680
    },
    {
      "epoch": 59.69,
      "learning_rate": 1.2041941747572817e-05,
      "loss": 0.0005,
      "step": 153690
    },
    {
      "epoch": 59.69,
      "learning_rate": 1.2041423948220066e-05,
      "loss": 0.2303,
      "step": 153700
    },
    {
      "epoch": 59.69,
      "learning_rate": 1.2040906148867314e-05,
      "loss": 0.1502,
      "step": 153710
    },
    {
      "epoch": 59.7,
      "learning_rate": 1.2040388349514564e-05,
      "loss": 0.2184,
      "step": 153720
    },
    {
      "epoch": 59.7,
      "learning_rate": 1.2039870550161814e-05,
      "loss": 0.0693,
      "step": 153730
    },
    {
      "epoch": 59.7,
      "learning_rate": 1.2039352750809063e-05,
      "loss": 0.0086,
      "step": 153740
    },
    {
      "epoch": 59.71,
      "learning_rate": 1.2038834951456311e-05,
      "loss": 0.1077,
      "step": 153750
    },
    {
      "epoch": 59.71,
      "learning_rate": 1.203831715210356e-05,
      "loss": 0.017,
      "step": 153760
    },
    {
      "epoch": 59.72,
      "learning_rate": 1.203779935275081e-05,
      "loss": 0.0162,
      "step": 153770
    },
    {
      "epoch": 59.72,
      "learning_rate": 1.203728155339806e-05,
      "loss": 0.0541,
      "step": 153780
    },
    {
      "epoch": 59.72,
      "learning_rate": 1.2036763754045308e-05,
      "loss": 0.1001,
      "step": 153790
    },
    {
      "epoch": 59.73,
      "learning_rate": 1.2036245954692558e-05,
      "loss": 0.0735,
      "step": 153800
    },
    {
      "epoch": 59.73,
      "learning_rate": 1.2035728155339807e-05,
      "loss": 0.0468,
      "step": 153810
    },
    {
      "epoch": 59.74,
      "learning_rate": 1.2035210355987057e-05,
      "loss": 0.1216,
      "step": 153820
    },
    {
      "epoch": 59.74,
      "learning_rate": 1.2034692556634305e-05,
      "loss": 0.0959,
      "step": 153830
    },
    {
      "epoch": 59.74,
      "learning_rate": 1.2034174757281554e-05,
      "loss": 0.094,
      "step": 153840
    },
    {
      "epoch": 59.75,
      "learning_rate": 1.2033656957928804e-05,
      "loss": 0.144,
      "step": 153850
    },
    {
      "epoch": 59.75,
      "learning_rate": 1.2033139158576052e-05,
      "loss": 0.0402,
      "step": 153860
    },
    {
      "epoch": 59.76,
      "learning_rate": 1.2032621359223301e-05,
      "loss": 0.0998,
      "step": 153870
    },
    {
      "epoch": 59.76,
      "learning_rate": 1.2032103559870551e-05,
      "loss": 0.1523,
      "step": 153880
    },
    {
      "epoch": 59.76,
      "learning_rate": 1.20315857605178e-05,
      "loss": 0.1147,
      "step": 153890
    },
    {
      "epoch": 59.77,
      "learning_rate": 1.203106796116505e-05,
      "loss": 0.1817,
      "step": 153900
    },
    {
      "epoch": 59.77,
      "learning_rate": 1.2030550161812298e-05,
      "loss": 0.0591,
      "step": 153910
    },
    {
      "epoch": 59.77,
      "learning_rate": 1.2030032362459548e-05,
      "loss": 0.0687,
      "step": 153920
    },
    {
      "epoch": 59.78,
      "learning_rate": 1.2029514563106797e-05,
      "loss": 0.0806,
      "step": 153930
    },
    {
      "epoch": 59.78,
      "learning_rate": 1.2028996763754045e-05,
      "loss": 0.0372,
      "step": 153940
    },
    {
      "epoch": 59.79,
      "learning_rate": 1.2028478964401295e-05,
      "loss": 0.111,
      "step": 153950
    },
    {
      "epoch": 59.79,
      "learning_rate": 1.2027961165048545e-05,
      "loss": 0.0025,
      "step": 153960
    },
    {
      "epoch": 59.79,
      "learning_rate": 1.2027443365695794e-05,
      "loss": 0.1125,
      "step": 153970
    },
    {
      "epoch": 59.8,
      "learning_rate": 1.2026925566343044e-05,
      "loss": 0.0806,
      "step": 153980
    },
    {
      "epoch": 59.8,
      "learning_rate": 1.2026407766990292e-05,
      "loss": 0.0795,
      "step": 153990
    },
    {
      "epoch": 59.81,
      "learning_rate": 1.2025889967637541e-05,
      "loss": 0.0424,
      "step": 154000
    },
    {
      "epoch": 59.81,
      "learning_rate": 1.202537216828479e-05,
      "loss": 0.0071,
      "step": 154010
    },
    {
      "epoch": 59.81,
      "learning_rate": 1.2024854368932039e-05,
      "loss": 0.1018,
      "step": 154020
    },
    {
      "epoch": 59.82,
      "learning_rate": 1.2024336569579289e-05,
      "loss": 0.1193,
      "step": 154030
    },
    {
      "epoch": 59.82,
      "learning_rate": 1.2023818770226538e-05,
      "loss": 0.0379,
      "step": 154040
    },
    {
      "epoch": 59.83,
      "learning_rate": 1.2023300970873788e-05,
      "loss": 0.0061,
      "step": 154050
    },
    {
      "epoch": 59.83,
      "learning_rate": 1.2022783171521037e-05,
      "loss": 0.1379,
      "step": 154060
    },
    {
      "epoch": 59.83,
      "learning_rate": 1.2022265372168285e-05,
      "loss": 0.056,
      "step": 154070
    },
    {
      "epoch": 59.84,
      "learning_rate": 1.2021747572815535e-05,
      "loss": 0.0314,
      "step": 154080
    },
    {
      "epoch": 59.84,
      "learning_rate": 1.2021229773462783e-05,
      "loss": 0.0153,
      "step": 154090
    },
    {
      "epoch": 59.84,
      "learning_rate": 1.2020711974110033e-05,
      "loss": 0.0044,
      "step": 154100
    },
    {
      "epoch": 59.85,
      "learning_rate": 1.2020194174757282e-05,
      "loss": 0.0209,
      "step": 154110
    },
    {
      "epoch": 59.85,
      "learning_rate": 1.2019676375404532e-05,
      "loss": 0.0162,
      "step": 154120
    },
    {
      "epoch": 59.86,
      "learning_rate": 1.2019158576051781e-05,
      "loss": 0.229,
      "step": 154130
    },
    {
      "epoch": 59.86,
      "learning_rate": 1.2018640776699031e-05,
      "loss": 0.0471,
      "step": 154140
    },
    {
      "epoch": 59.86,
      "learning_rate": 1.2018122977346279e-05,
      "loss": 0.104,
      "step": 154150
    },
    {
      "epoch": 59.87,
      "learning_rate": 1.2017605177993527e-05,
      "loss": 0.1885,
      "step": 154160
    },
    {
      "epoch": 59.87,
      "learning_rate": 1.2017087378640776e-05,
      "loss": 0.053,
      "step": 154170
    },
    {
      "epoch": 59.88,
      "learning_rate": 1.2016569579288026e-05,
      "loss": 0.072,
      "step": 154180
    },
    {
      "epoch": 59.88,
      "learning_rate": 1.2016051779935276e-05,
      "loss": 0.0128,
      "step": 154190
    },
    {
      "epoch": 59.88,
      "learning_rate": 1.2015533980582525e-05,
      "loss": 0.0919,
      "step": 154200
    },
    {
      "epoch": 59.89,
      "learning_rate": 1.2015016181229775e-05,
      "loss": 0.0651,
      "step": 154210
    },
    {
      "epoch": 59.89,
      "learning_rate": 1.2014498381877025e-05,
      "loss": 0.0318,
      "step": 154220
    },
    {
      "epoch": 59.9,
      "learning_rate": 1.2013980582524274e-05,
      "loss": 0.0431,
      "step": 154230
    },
    {
      "epoch": 59.9,
      "learning_rate": 1.201346278317152e-05,
      "loss": 0.0049,
      "step": 154240
    },
    {
      "epoch": 59.9,
      "learning_rate": 1.201294498381877e-05,
      "loss": 0.1265,
      "step": 154250
    },
    {
      "epoch": 59.91,
      "learning_rate": 1.201242718446602e-05,
      "loss": 0.1034,
      "step": 154260
    },
    {
      "epoch": 59.91,
      "learning_rate": 1.201190938511327e-05,
      "loss": 0.0238,
      "step": 154270
    },
    {
      "epoch": 59.91,
      "learning_rate": 1.2011391585760519e-05,
      "loss": 0.1014,
      "step": 154280
    },
    {
      "epoch": 59.92,
      "learning_rate": 1.2010873786407768e-05,
      "loss": 0.1089,
      "step": 154290
    },
    {
      "epoch": 59.92,
      "learning_rate": 1.2010355987055018e-05,
      "loss": 0.2796,
      "step": 154300
    },
    {
      "epoch": 59.93,
      "learning_rate": 1.2009838187702268e-05,
      "loss": 0.0582,
      "step": 154310
    },
    {
      "epoch": 59.93,
      "learning_rate": 1.2009320388349514e-05,
      "loss": 0.0254,
      "step": 154320
    },
    {
      "epoch": 59.93,
      "learning_rate": 1.2008802588996764e-05,
      "loss": 0.1634,
      "step": 154330
    },
    {
      "epoch": 59.94,
      "learning_rate": 1.2008284789644013e-05,
      "loss": 0.0128,
      "step": 154340
    },
    {
      "epoch": 59.94,
      "learning_rate": 1.2007766990291263e-05,
      "loss": 0.0981,
      "step": 154350
    },
    {
      "epoch": 59.95,
      "learning_rate": 1.2007249190938512e-05,
      "loss": 0.0414,
      "step": 154360
    },
    {
      "epoch": 59.95,
      "learning_rate": 1.2006731391585762e-05,
      "loss": 0.1022,
      "step": 154370
    },
    {
      "epoch": 59.95,
      "learning_rate": 1.2006213592233012e-05,
      "loss": 0.0026,
      "step": 154380
    },
    {
      "epoch": 59.96,
      "learning_rate": 1.2005695792880261e-05,
      "loss": 0.0261,
      "step": 154390
    },
    {
      "epoch": 59.96,
      "learning_rate": 1.2005177993527508e-05,
      "loss": 0.1075,
      "step": 154400
    },
    {
      "epoch": 59.97,
      "learning_rate": 1.2004660194174757e-05,
      "loss": 0.0374,
      "step": 154410
    },
    {
      "epoch": 59.97,
      "learning_rate": 1.2004142394822007e-05,
      "loss": 0.0159,
      "step": 154420
    },
    {
      "epoch": 59.97,
      "learning_rate": 1.2003624595469256e-05,
      "loss": 0.1148,
      "step": 154430
    },
    {
      "epoch": 59.98,
      "learning_rate": 1.2003106796116506e-05,
      "loss": 0.1067,
      "step": 154440
    },
    {
      "epoch": 59.98,
      "learning_rate": 1.2002588996763756e-05,
      "loss": 0.0956,
      "step": 154450
    },
    {
      "epoch": 59.98,
      "learning_rate": 1.2002071197411005e-05,
      "loss": 0.0622,
      "step": 154460
    },
    {
      "epoch": 59.99,
      "learning_rate": 1.2001553398058255e-05,
      "loss": 0.0504,
      "step": 154470
    },
    {
      "epoch": 59.99,
      "learning_rate": 1.2001035598705501e-05,
      "loss": 0.0511,
      "step": 154480
    },
    {
      "epoch": 60.0,
      "learning_rate": 1.200051779935275e-05,
      "loss": 0.1265,
      "step": 154490
    },
    {
      "epoch": 60.0,
      "learning_rate": 1.2e-05,
      "loss": 0.1233,
      "step": 154500
    },
    {
      "epoch": 60.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.3270106315612793,
      "eval_runtime": 8.2702,
      "eval_samples_per_second": 439.527,
      "eval_steps_per_second": 55.016,
      "step": 154500
    },
    {
      "epoch": 60.0,
      "learning_rate": 1.199948220064725e-05,
      "loss": 0.0453,
      "step": 154510
    },
    {
      "epoch": 60.01,
      "learning_rate": 1.19989644012945e-05,
      "loss": 0.0006,
      "step": 154520
    },
    {
      "epoch": 60.01,
      "learning_rate": 1.199844660194175e-05,
      "loss": 0.0644,
      "step": 154530
    },
    {
      "epoch": 60.02,
      "learning_rate": 1.1997928802588999e-05,
      "loss": 0.1287,
      "step": 154540
    },
    {
      "epoch": 60.02,
      "learning_rate": 1.1997411003236248e-05,
      "loss": 0.043,
      "step": 154550
    },
    {
      "epoch": 60.02,
      "learning_rate": 1.1996893203883495e-05,
      "loss": 0.0771,
      "step": 154560
    },
    {
      "epoch": 60.03,
      "learning_rate": 1.1996375404530744e-05,
      "loss": 0.0051,
      "step": 154570
    },
    {
      "epoch": 60.03,
      "learning_rate": 1.1995857605177994e-05,
      "loss": 0.1302,
      "step": 154580
    },
    {
      "epoch": 60.03,
      "learning_rate": 1.1995339805825243e-05,
      "loss": 0.0181,
      "step": 154590
    },
    {
      "epoch": 60.04,
      "learning_rate": 1.1994822006472493e-05,
      "loss": 0.0152,
      "step": 154600
    },
    {
      "epoch": 60.04,
      "learning_rate": 1.1994304207119743e-05,
      "loss": 0.0528,
      "step": 154610
    },
    {
      "epoch": 60.05,
      "learning_rate": 1.1993786407766992e-05,
      "loss": 0.0407,
      "step": 154620
    },
    {
      "epoch": 60.05,
      "learning_rate": 1.1993268608414242e-05,
      "loss": 0.2022,
      "step": 154630
    },
    {
      "epoch": 60.05,
      "learning_rate": 1.1992750809061488e-05,
      "loss": 0.0917,
      "step": 154640
    },
    {
      "epoch": 60.06,
      "learning_rate": 1.1992233009708738e-05,
      "loss": 0.1391,
      "step": 154650
    },
    {
      "epoch": 60.06,
      "learning_rate": 1.1991715210355987e-05,
      "loss": 0.0237,
      "step": 154660
    },
    {
      "epoch": 60.07,
      "learning_rate": 1.1991197411003237e-05,
      "loss": 0.0566,
      "step": 154670
    },
    {
      "epoch": 60.07,
      "learning_rate": 1.1990679611650487e-05,
      "loss": 0.0641,
      "step": 154680
    },
    {
      "epoch": 60.07,
      "learning_rate": 1.1990161812297736e-05,
      "loss": 0.0027,
      "step": 154690
    },
    {
      "epoch": 60.08,
      "learning_rate": 1.1989644012944986e-05,
      "loss": 0.0512,
      "step": 154700
    },
    {
      "epoch": 60.08,
      "learning_rate": 1.1989126213592236e-05,
      "loss": 0.1369,
      "step": 154710
    },
    {
      "epoch": 60.09,
      "learning_rate": 1.1988608414239482e-05,
      "loss": 0.172,
      "step": 154720
    },
    {
      "epoch": 60.09,
      "learning_rate": 1.1988090614886731e-05,
      "loss": 0.0525,
      "step": 154730
    },
    {
      "epoch": 60.09,
      "learning_rate": 1.1987572815533981e-05,
      "loss": 0.104,
      "step": 154740
    },
    {
      "epoch": 60.1,
      "learning_rate": 1.198705501618123e-05,
      "loss": 0.0926,
      "step": 154750
    },
    {
      "epoch": 60.1,
      "learning_rate": 1.198653721682848e-05,
      "loss": 0.042,
      "step": 154760
    },
    {
      "epoch": 60.1,
      "learning_rate": 1.198601941747573e-05,
      "loss": 0.0003,
      "step": 154770
    },
    {
      "epoch": 60.11,
      "learning_rate": 1.198550161812298e-05,
      "loss": 0.0698,
      "step": 154780
    },
    {
      "epoch": 60.11,
      "learning_rate": 1.1984983818770229e-05,
      "loss": 0.1121,
      "step": 154790
    },
    {
      "epoch": 60.12,
      "learning_rate": 1.1984466019417477e-05,
      "loss": 0.0514,
      "step": 154800
    },
    {
      "epoch": 60.12,
      "learning_rate": 1.1983948220064725e-05,
      "loss": 0.0543,
      "step": 154810
    },
    {
      "epoch": 60.12,
      "learning_rate": 1.1983430420711975e-05,
      "loss": 0.1078,
      "step": 154820
    },
    {
      "epoch": 60.13,
      "learning_rate": 1.1982912621359224e-05,
      "loss": 0.0311,
      "step": 154830
    },
    {
      "epoch": 60.13,
      "learning_rate": 1.1982394822006474e-05,
      "loss": 0.0886,
      "step": 154840
    },
    {
      "epoch": 60.14,
      "learning_rate": 1.1981877022653723e-05,
      "loss": 0.2306,
      "step": 154850
    },
    {
      "epoch": 60.14,
      "learning_rate": 1.1981359223300973e-05,
      "loss": 0.0024,
      "step": 154860
    },
    {
      "epoch": 60.14,
      "learning_rate": 1.1980841423948223e-05,
      "loss": 0.0228,
      "step": 154870
    },
    {
      "epoch": 60.15,
      "learning_rate": 1.198032362459547e-05,
      "loss": 0.0768,
      "step": 154880
    },
    {
      "epoch": 60.15,
      "learning_rate": 1.1979805825242718e-05,
      "loss": 0.086,
      "step": 154890
    },
    {
      "epoch": 60.16,
      "learning_rate": 1.1979288025889968e-05,
      "loss": 0.019,
      "step": 154900
    },
    {
      "epoch": 60.16,
      "learning_rate": 1.1978770226537218e-05,
      "loss": 0.1197,
      "step": 154910
    },
    {
      "epoch": 60.16,
      "learning_rate": 1.1978252427184467e-05,
      "loss": 0.0945,
      "step": 154920
    },
    {
      "epoch": 60.17,
      "learning_rate": 1.1977734627831717e-05,
      "loss": 0.0324,
      "step": 154930
    },
    {
      "epoch": 60.17,
      "learning_rate": 1.1977216828478967e-05,
      "loss": 0.1092,
      "step": 154940
    },
    {
      "epoch": 60.17,
      "learning_rate": 1.1976699029126214e-05,
      "loss": 0.1088,
      "step": 154950
    },
    {
      "epoch": 60.18,
      "learning_rate": 1.1976181229773464e-05,
      "loss": 0.2394,
      "step": 154960
    },
    {
      "epoch": 60.18,
      "learning_rate": 1.1975663430420712e-05,
      "loss": 0.1018,
      "step": 154970
    },
    {
      "epoch": 60.19,
      "learning_rate": 1.1975145631067962e-05,
      "loss": 0.045,
      "step": 154980
    },
    {
      "epoch": 60.19,
      "learning_rate": 1.1974627831715211e-05,
      "loss": 0.1302,
      "step": 154990
    },
    {
      "epoch": 60.19,
      "learning_rate": 1.1974110032362461e-05,
      "loss": 0.0075,
      "step": 155000
    },
    {
      "epoch": 60.2,
      "learning_rate": 1.197359223300971e-05,
      "loss": 0.0149,
      "step": 155010
    },
    {
      "epoch": 60.2,
      "learning_rate": 1.197307443365696e-05,
      "loss": 0.0471,
      "step": 155020
    },
    {
      "epoch": 60.21,
      "learning_rate": 1.1972556634304208e-05,
      "loss": 0.0063,
      "step": 155030
    },
    {
      "epoch": 60.21,
      "learning_rate": 1.1972038834951458e-05,
      "loss": 0.0503,
      "step": 155040
    },
    {
      "epoch": 60.21,
      "learning_rate": 1.1971521035598706e-05,
      "loss": 0.0417,
      "step": 155050
    },
    {
      "epoch": 60.22,
      "learning_rate": 1.1971003236245955e-05,
      "loss": 0.0171,
      "step": 155060
    },
    {
      "epoch": 60.22,
      "learning_rate": 1.1970485436893205e-05,
      "loss": 0.0379,
      "step": 155070
    },
    {
      "epoch": 60.23,
      "learning_rate": 1.1969967637540454e-05,
      "loss": 0.0549,
      "step": 155080
    },
    {
      "epoch": 60.23,
      "learning_rate": 1.1969449838187704e-05,
      "loss": 0.0016,
      "step": 155090
    },
    {
      "epoch": 60.23,
      "learning_rate": 1.1968932038834952e-05,
      "loss": 0.1607,
      "step": 155100
    },
    {
      "epoch": 60.24,
      "learning_rate": 1.1968414239482202e-05,
      "loss": 0.0549,
      "step": 155110
    },
    {
      "epoch": 60.24,
      "learning_rate": 1.1967896440129451e-05,
      "loss": 0.0467,
      "step": 155120
    },
    {
      "epoch": 60.24,
      "learning_rate": 1.1967378640776699e-05,
      "loss": 0.0406,
      "step": 155130
    },
    {
      "epoch": 60.25,
      "learning_rate": 1.1966860841423949e-05,
      "loss": 0.1003,
      "step": 155140
    },
    {
      "epoch": 60.25,
      "learning_rate": 1.1966343042071198e-05,
      "loss": 0.0907,
      "step": 155150
    },
    {
      "epoch": 60.26,
      "learning_rate": 1.1965825242718448e-05,
      "loss": 0.0063,
      "step": 155160
    },
    {
      "epoch": 60.26,
      "learning_rate": 1.1965307443365698e-05,
      "loss": 0.0475,
      "step": 155170
    },
    {
      "epoch": 60.26,
      "learning_rate": 1.1964789644012946e-05,
      "loss": 0.2801,
      "step": 155180
    },
    {
      "epoch": 60.27,
      "learning_rate": 1.1964271844660195e-05,
      "loss": 0.0951,
      "step": 155190
    },
    {
      "epoch": 60.27,
      "learning_rate": 1.1963754045307445e-05,
      "loss": 0.0232,
      "step": 155200
    },
    {
      "epoch": 60.28,
      "learning_rate": 1.1963236245954693e-05,
      "loss": 0.05,
      "step": 155210
    },
    {
      "epoch": 60.28,
      "learning_rate": 1.1962718446601942e-05,
      "loss": 0.0019,
      "step": 155220
    },
    {
      "epoch": 60.28,
      "learning_rate": 1.1962200647249192e-05,
      "loss": 0.2717,
      "step": 155230
    },
    {
      "epoch": 60.29,
      "learning_rate": 1.1961682847896442e-05,
      "loss": 0.0993,
      "step": 155240
    },
    {
      "epoch": 60.29,
      "learning_rate": 1.196116504854369e-05,
      "loss": 0.0132,
      "step": 155250
    },
    {
      "epoch": 60.3,
      "learning_rate": 1.1960647249190939e-05,
      "loss": 0.1866,
      "step": 155260
    },
    {
      "epoch": 60.3,
      "learning_rate": 1.1960129449838189e-05,
      "loss": 0.0181,
      "step": 155270
    },
    {
      "epoch": 60.3,
      "learning_rate": 1.1959611650485438e-05,
      "loss": 0.1661,
      "step": 155280
    },
    {
      "epoch": 60.31,
      "learning_rate": 1.1959093851132688e-05,
      "loss": 0.2961,
      "step": 155290
    },
    {
      "epoch": 60.31,
      "learning_rate": 1.1958576051779936e-05,
      "loss": 0.0004,
      "step": 155300
    },
    {
      "epoch": 60.31,
      "learning_rate": 1.1958058252427185e-05,
      "loss": 0.0545,
      "step": 155310
    },
    {
      "epoch": 60.32,
      "learning_rate": 1.1957540453074435e-05,
      "loss": 0.0004,
      "step": 155320
    },
    {
      "epoch": 60.32,
      "learning_rate": 1.1957022653721683e-05,
      "loss": 0.1217,
      "step": 155330
    },
    {
      "epoch": 60.33,
      "learning_rate": 1.1956504854368933e-05,
      "loss": 0.0252,
      "step": 155340
    },
    {
      "epoch": 60.33,
      "learning_rate": 1.1955987055016182e-05,
      "loss": 0.0132,
      "step": 155350
    },
    {
      "epoch": 60.33,
      "learning_rate": 1.1955469255663432e-05,
      "loss": 0.0563,
      "step": 155360
    },
    {
      "epoch": 60.34,
      "learning_rate": 1.1954951456310681e-05,
      "loss": 0.0036,
      "step": 155370
    },
    {
      "epoch": 60.34,
      "learning_rate": 1.195443365695793e-05,
      "loss": 0.0909,
      "step": 155380
    },
    {
      "epoch": 60.35,
      "learning_rate": 1.1953915857605179e-05,
      "loss": 0.0221,
      "step": 155390
    },
    {
      "epoch": 60.35,
      "learning_rate": 1.1953398058252429e-05,
      "loss": 0.0862,
      "step": 155400
    },
    {
      "epoch": 60.35,
      "learning_rate": 1.1952880258899677e-05,
      "loss": 0.0288,
      "step": 155410
    },
    {
      "epoch": 60.36,
      "learning_rate": 1.1952362459546926e-05,
      "loss": 0.0198,
      "step": 155420
    },
    {
      "epoch": 60.36,
      "learning_rate": 1.1951844660194176e-05,
      "loss": 0.0052,
      "step": 155430
    },
    {
      "epoch": 60.37,
      "learning_rate": 1.1951326860841425e-05,
      "loss": 0.1604,
      "step": 155440
    },
    {
      "epoch": 60.37,
      "learning_rate": 1.1950809061488675e-05,
      "loss": 0.0219,
      "step": 155450
    },
    {
      "epoch": 60.37,
      "learning_rate": 1.1950291262135923e-05,
      "loss": 0.2094,
      "step": 155460
    },
    {
      "epoch": 60.38,
      "learning_rate": 1.1949773462783173e-05,
      "loss": 0.0231,
      "step": 155470
    },
    {
      "epoch": 60.38,
      "learning_rate": 1.194925566343042e-05,
      "loss": 0.0587,
      "step": 155480
    },
    {
      "epoch": 60.38,
      "learning_rate": 1.194873786407767e-05,
      "loss": 0.0771,
      "step": 155490
    },
    {
      "epoch": 60.39,
      "learning_rate": 1.194822006472492e-05,
      "loss": 0.0692,
      "step": 155500
    },
    {
      "epoch": 60.39,
      "learning_rate": 1.194770226537217e-05,
      "loss": 0.0096,
      "step": 155510
    },
    {
      "epoch": 60.4,
      "learning_rate": 1.1947184466019419e-05,
      "loss": 0.0981,
      "step": 155520
    },
    {
      "epoch": 60.4,
      "learning_rate": 1.1946666666666669e-05,
      "loss": 0.0483,
      "step": 155530
    },
    {
      "epoch": 60.4,
      "learning_rate": 1.1946148867313917e-05,
      "loss": 0.0623,
      "step": 155540
    },
    {
      "epoch": 60.41,
      "learning_rate": 1.1945631067961166e-05,
      "loss": 0.1517,
      "step": 155550
    },
    {
      "epoch": 60.41,
      "learning_rate": 1.1945113268608414e-05,
      "loss": 0.0877,
      "step": 155560
    },
    {
      "epoch": 60.42,
      "learning_rate": 1.1944595469255664e-05,
      "loss": 0.1436,
      "step": 155570
    },
    {
      "epoch": 60.42,
      "learning_rate": 1.1944077669902913e-05,
      "loss": 0.147,
      "step": 155580
    },
    {
      "epoch": 60.42,
      "learning_rate": 1.1943559870550163e-05,
      "loss": 0.0363,
      "step": 155590
    },
    {
      "epoch": 60.43,
      "learning_rate": 1.1943042071197413e-05,
      "loss": 0.0277,
      "step": 155600
    },
    {
      "epoch": 60.43,
      "learning_rate": 1.1942524271844662e-05,
      "loss": 0.0407,
      "step": 155610
    },
    {
      "epoch": 60.43,
      "learning_rate": 1.194200647249191e-05,
      "loss": 0.0183,
      "step": 155620
    },
    {
      "epoch": 60.44,
      "learning_rate": 1.1941488673139158e-05,
      "loss": 0.0699,
      "step": 155630
    },
    {
      "epoch": 60.44,
      "learning_rate": 1.1940970873786408e-05,
      "loss": 0.1696,
      "step": 155640
    },
    {
      "epoch": 60.45,
      "learning_rate": 1.1940453074433657e-05,
      "loss": 0.1493,
      "step": 155650
    },
    {
      "epoch": 60.45,
      "learning_rate": 1.1939935275080907e-05,
      "loss": 0.082,
      "step": 155660
    },
    {
      "epoch": 60.45,
      "learning_rate": 1.1939417475728156e-05,
      "loss": 0.0391,
      "step": 155670
    },
    {
      "epoch": 60.46,
      "learning_rate": 1.1938899676375406e-05,
      "loss": 0.0958,
      "step": 155680
    },
    {
      "epoch": 60.46,
      "learning_rate": 1.1938381877022656e-05,
      "loss": 0.0453,
      "step": 155690
    },
    {
      "epoch": 60.47,
      "learning_rate": 1.1937864077669904e-05,
      "loss": 0.0927,
      "step": 155700
    },
    {
      "epoch": 60.47,
      "learning_rate": 1.1937346278317152e-05,
      "loss": 0.0811,
      "step": 155710
    },
    {
      "epoch": 60.47,
      "learning_rate": 1.1936828478964401e-05,
      "loss": 0.0727,
      "step": 155720
    },
    {
      "epoch": 60.48,
      "learning_rate": 1.193631067961165e-05,
      "loss": 0.0226,
      "step": 155730
    },
    {
      "epoch": 60.48,
      "learning_rate": 1.19357928802589e-05,
      "loss": 0.0282,
      "step": 155740
    },
    {
      "epoch": 60.49,
      "learning_rate": 1.193527508090615e-05,
      "loss": 0.1121,
      "step": 155750
    },
    {
      "epoch": 60.49,
      "learning_rate": 1.19347572815534e-05,
      "loss": 0.0542,
      "step": 155760
    },
    {
      "epoch": 60.49,
      "learning_rate": 1.193423948220065e-05,
      "loss": 0.0732,
      "step": 155770
    },
    {
      "epoch": 60.5,
      "learning_rate": 1.1933721682847896e-05,
      "loss": 0.0282,
      "step": 155780
    },
    {
      "epoch": 60.5,
      "learning_rate": 1.1933203883495145e-05,
      "loss": 0.0186,
      "step": 155790
    },
    {
      "epoch": 60.5,
      "learning_rate": 1.1932686084142395e-05,
      "loss": 0.0301,
      "step": 155800
    },
    {
      "epoch": 60.51,
      "learning_rate": 1.1932168284789644e-05,
      "loss": 0.0706,
      "step": 155810
    },
    {
      "epoch": 60.51,
      "learning_rate": 1.1931650485436894e-05,
      "loss": 0.0203,
      "step": 155820
    },
    {
      "epoch": 60.52,
      "learning_rate": 1.1931132686084144e-05,
      "loss": 0.0767,
      "step": 155830
    },
    {
      "epoch": 60.52,
      "learning_rate": 1.1930614886731393e-05,
      "loss": 0.0311,
      "step": 155840
    },
    {
      "epoch": 60.52,
      "learning_rate": 1.1930097087378643e-05,
      "loss": 0.0315,
      "step": 155850
    },
    {
      "epoch": 60.53,
      "learning_rate": 1.1929579288025892e-05,
      "loss": 0.0253,
      "step": 155860
    },
    {
      "epoch": 60.53,
      "learning_rate": 1.1929061488673139e-05,
      "loss": 0.0602,
      "step": 155870
    },
    {
      "epoch": 60.54,
      "learning_rate": 1.1928543689320388e-05,
      "loss": 0.0861,
      "step": 155880
    },
    {
      "epoch": 60.54,
      "learning_rate": 1.1928025889967638e-05,
      "loss": 0.1734,
      "step": 155890
    },
    {
      "epoch": 60.54,
      "learning_rate": 1.1927508090614888e-05,
      "loss": 0.1657,
      "step": 155900
    },
    {
      "epoch": 60.55,
      "learning_rate": 1.1926990291262137e-05,
      "loss": 0.0209,
      "step": 155910
    },
    {
      "epoch": 60.55,
      "learning_rate": 1.1926472491909387e-05,
      "loss": 0.0264,
      "step": 155920
    },
    {
      "epoch": 60.56,
      "learning_rate": 1.1925954692556636e-05,
      "loss": 0.0187,
      "step": 155930
    },
    {
      "epoch": 60.56,
      "learning_rate": 1.1925436893203886e-05,
      "loss": 0.0072,
      "step": 155940
    },
    {
      "epoch": 60.56,
      "learning_rate": 1.1924919093851132e-05,
      "loss": 0.0009,
      "step": 155950
    },
    {
      "epoch": 60.57,
      "learning_rate": 1.1924401294498382e-05,
      "loss": 0.1241,
      "step": 155960
    },
    {
      "epoch": 60.57,
      "learning_rate": 1.1923883495145631e-05,
      "loss": 0.018,
      "step": 155970
    },
    {
      "epoch": 60.57,
      "learning_rate": 1.1923365695792881e-05,
      "loss": 0.0708,
      "step": 155980
    },
    {
      "epoch": 60.58,
      "learning_rate": 1.192284789644013e-05,
      "loss": 0.08,
      "step": 155990
    },
    {
      "epoch": 60.58,
      "learning_rate": 1.192233009708738e-05,
      "loss": 0.0178,
      "step": 156000
    },
    {
      "epoch": 60.59,
      "learning_rate": 1.192181229773463e-05,
      "loss": 0.1351,
      "step": 156010
    },
    {
      "epoch": 60.59,
      "learning_rate": 1.192129449838188e-05,
      "loss": 0.0056,
      "step": 156020
    },
    {
      "epoch": 60.59,
      "learning_rate": 1.1920776699029126e-05,
      "loss": 0.1967,
      "step": 156030
    },
    {
      "epoch": 60.6,
      "learning_rate": 1.1920258899676375e-05,
      "loss": 0.0755,
      "step": 156040
    },
    {
      "epoch": 60.6,
      "learning_rate": 1.1919741100323625e-05,
      "loss": 0.0042,
      "step": 156050
    },
    {
      "epoch": 60.61,
      "learning_rate": 1.1919223300970875e-05,
      "loss": 0.0967,
      "step": 156060
    },
    {
      "epoch": 60.61,
      "learning_rate": 1.1918705501618124e-05,
      "loss": 0.0829,
      "step": 156070
    },
    {
      "epoch": 60.61,
      "learning_rate": 1.1918187702265374e-05,
      "loss": 0.1516,
      "step": 156080
    },
    {
      "epoch": 60.62,
      "learning_rate": 1.1917669902912624e-05,
      "loss": 0.0673,
      "step": 156090
    },
    {
      "epoch": 60.62,
      "learning_rate": 1.1917152103559873e-05,
      "loss": 0.064,
      "step": 156100
    },
    {
      "epoch": 60.63,
      "learning_rate": 1.191663430420712e-05,
      "loss": 0.1235,
      "step": 156110
    },
    {
      "epoch": 60.63,
      "learning_rate": 1.1916116504854369e-05,
      "loss": 0.037,
      "step": 156120
    },
    {
      "epoch": 60.63,
      "learning_rate": 1.1915598705501619e-05,
      "loss": 0.2453,
      "step": 156130
    },
    {
      "epoch": 60.64,
      "learning_rate": 1.1915080906148868e-05,
      "loss": 0.2768,
      "step": 156140
    },
    {
      "epoch": 60.64,
      "learning_rate": 1.1914563106796118e-05,
      "loss": 0.1128,
      "step": 156150
    },
    {
      "epoch": 60.64,
      "learning_rate": 1.1914045307443367e-05,
      "loss": 0.0558,
      "step": 156160
    },
    {
      "epoch": 60.65,
      "learning_rate": 1.1913527508090617e-05,
      "loss": 0.0101,
      "step": 156170
    },
    {
      "epoch": 60.65,
      "learning_rate": 1.1913009708737867e-05,
      "loss": 0.0732,
      "step": 156180
    },
    {
      "epoch": 60.66,
      "learning_rate": 1.1912491909385113e-05,
      "loss": 0.0288,
      "step": 156190
    },
    {
      "epoch": 60.66,
      "learning_rate": 1.1911974110032363e-05,
      "loss": 0.0195,
      "step": 156200
    },
    {
      "epoch": 60.66,
      "learning_rate": 1.1911456310679612e-05,
      "loss": 0.0801,
      "step": 156210
    },
    {
      "epoch": 60.67,
      "learning_rate": 1.1910938511326862e-05,
      "loss": 0.0658,
      "step": 156220
    },
    {
      "epoch": 60.67,
      "learning_rate": 1.1910420711974111e-05,
      "loss": 0.1411,
      "step": 156230
    },
    {
      "epoch": 60.68,
      "learning_rate": 1.1909902912621361e-05,
      "loss": 0.0135,
      "step": 156240
    },
    {
      "epoch": 60.68,
      "learning_rate": 1.190938511326861e-05,
      "loss": 0.1086,
      "step": 156250
    },
    {
      "epoch": 60.68,
      "learning_rate": 1.190886731391586e-05,
      "loss": 0.1215,
      "step": 156260
    },
    {
      "epoch": 60.69,
      "learning_rate": 1.1908349514563106e-05,
      "loss": 0.1385,
      "step": 156270
    },
    {
      "epoch": 60.69,
      "learning_rate": 1.1907831715210356e-05,
      "loss": 0.0248,
      "step": 156280
    },
    {
      "epoch": 60.7,
      "learning_rate": 1.1907313915857606e-05,
      "loss": 0.0218,
      "step": 156290
    },
    {
      "epoch": 60.7,
      "learning_rate": 1.1906796116504855e-05,
      "loss": 0.0949,
      "step": 156300
    },
    {
      "epoch": 60.7,
      "learning_rate": 1.1906278317152105e-05,
      "loss": 0.1057,
      "step": 156310
    },
    {
      "epoch": 60.71,
      "learning_rate": 1.1905760517799355e-05,
      "loss": 0.0911,
      "step": 156320
    },
    {
      "epoch": 60.71,
      "learning_rate": 1.1905242718446604e-05,
      "loss": 0.0939,
      "step": 156330
    },
    {
      "epoch": 60.71,
      "learning_rate": 1.1904724919093854e-05,
      "loss": 0.0695,
      "step": 156340
    },
    {
      "epoch": 60.72,
      "learning_rate": 1.19042071197411e-05,
      "loss": 0.1471,
      "step": 156350
    },
    {
      "epoch": 60.72,
      "learning_rate": 1.190368932038835e-05,
      "loss": 0.005,
      "step": 156360
    },
    {
      "epoch": 60.73,
      "learning_rate": 1.19031715210356e-05,
      "loss": 0.1125,
      "step": 156370
    },
    {
      "epoch": 60.73,
      "learning_rate": 1.1902653721682849e-05,
      "loss": 0.0658,
      "step": 156380
    },
    {
      "epoch": 60.73,
      "learning_rate": 1.1902135922330098e-05,
      "loss": 0.1357,
      "step": 156390
    },
    {
      "epoch": 60.74,
      "learning_rate": 1.1901618122977348e-05,
      "loss": 0.0673,
      "step": 156400
    },
    {
      "epoch": 60.74,
      "learning_rate": 1.1901100323624598e-05,
      "loss": 0.0803,
      "step": 156410
    },
    {
      "epoch": 60.75,
      "learning_rate": 1.1900582524271846e-05,
      "loss": 0.049,
      "step": 156420
    },
    {
      "epoch": 60.75,
      "learning_rate": 1.1900064724919095e-05,
      "loss": 0.0329,
      "step": 156430
    },
    {
      "epoch": 60.75,
      "learning_rate": 1.1899546925566343e-05,
      "loss": 0.0598,
      "step": 156440
    },
    {
      "epoch": 60.76,
      "learning_rate": 1.1899029126213593e-05,
      "loss": 0.093,
      "step": 156450
    },
    {
      "epoch": 60.76,
      "learning_rate": 1.1898511326860842e-05,
      "loss": 0.1312,
      "step": 156460
    },
    {
      "epoch": 60.77,
      "learning_rate": 1.1897993527508092e-05,
      "loss": 0.0543,
      "step": 156470
    },
    {
      "epoch": 60.77,
      "learning_rate": 1.1897475728155342e-05,
      "loss": 0.0244,
      "step": 156480
    },
    {
      "epoch": 60.77,
      "learning_rate": 1.1896957928802591e-05,
      "loss": 0.0946,
      "step": 156490
    },
    {
      "epoch": 60.78,
      "learning_rate": 1.189644012944984e-05,
      "loss": 0.0711,
      "step": 156500
    },
    {
      "epoch": 60.78,
      "learning_rate": 1.1895922330097089e-05,
      "loss": 0.0575,
      "step": 156510
    },
    {
      "epoch": 60.78,
      "learning_rate": 1.1895404530744337e-05,
      "loss": 0.1781,
      "step": 156520
    },
    {
      "epoch": 60.79,
      "learning_rate": 1.1894886731391586e-05,
      "loss": 0.087,
      "step": 156530
    },
    {
      "epoch": 60.79,
      "learning_rate": 1.1894368932038836e-05,
      "loss": 0.0565,
      "step": 156540
    },
    {
      "epoch": 60.8,
      "learning_rate": 1.1893851132686086e-05,
      "loss": 0.1005,
      "step": 156550
    },
    {
      "epoch": 60.8,
      "learning_rate": 1.1893333333333335e-05,
      "loss": 0.0472,
      "step": 156560
    },
    {
      "epoch": 60.8,
      "learning_rate": 1.1892815533980583e-05,
      "loss": 0.0956,
      "step": 156570
    },
    {
      "epoch": 60.81,
      "learning_rate": 1.1892297734627833e-05,
      "loss": 0.0237,
      "step": 156580
    },
    {
      "epoch": 60.81,
      "learning_rate": 1.1891779935275082e-05,
      "loss": 0.2198,
      "step": 156590
    },
    {
      "epoch": 60.82,
      "learning_rate": 1.189126213592233e-05,
      "loss": 0.0244,
      "step": 156600
    },
    {
      "epoch": 60.82,
      "learning_rate": 1.189074433656958e-05,
      "loss": 0.0889,
      "step": 156610
    },
    {
      "epoch": 60.82,
      "learning_rate": 1.189022653721683e-05,
      "loss": 0.1039,
      "step": 156620
    },
    {
      "epoch": 60.83,
      "learning_rate": 1.188970873786408e-05,
      "loss": 0.0231,
      "step": 156630
    },
    {
      "epoch": 60.83,
      "learning_rate": 1.1889190938511329e-05,
      "loss": 0.2699,
      "step": 156640
    },
    {
      "epoch": 60.83,
      "learning_rate": 1.1888673139158577e-05,
      "loss": 0.0003,
      "step": 156650
    },
    {
      "epoch": 60.84,
      "learning_rate": 1.1888155339805826e-05,
      "loss": 0.0308,
      "step": 156660
    },
    {
      "epoch": 60.84,
      "learning_rate": 1.1887637540453076e-05,
      "loss": 0.0496,
      "step": 156670
    },
    {
      "epoch": 60.85,
      "learning_rate": 1.1887119741100324e-05,
      "loss": 0.0518,
      "step": 156680
    },
    {
      "epoch": 60.85,
      "learning_rate": 1.1886601941747573e-05,
      "loss": 0.0646,
      "step": 156690
    },
    {
      "epoch": 60.85,
      "learning_rate": 1.1886084142394823e-05,
      "loss": 0.0341,
      "step": 156700
    },
    {
      "epoch": 60.86,
      "learning_rate": 1.1885566343042073e-05,
      "loss": 0.1255,
      "step": 156710
    },
    {
      "epoch": 60.86,
      "learning_rate": 1.188504854368932e-05,
      "loss": 0.0772,
      "step": 156720
    },
    {
      "epoch": 60.87,
      "learning_rate": 1.188453074433657e-05,
      "loss": 0.2768,
      "step": 156730
    },
    {
      "epoch": 60.87,
      "learning_rate": 1.188401294498382e-05,
      "loss": 0.1339,
      "step": 156740
    },
    {
      "epoch": 60.87,
      "learning_rate": 1.188349514563107e-05,
      "loss": 0.0872,
      "step": 156750
    },
    {
      "epoch": 60.88,
      "learning_rate": 1.1882977346278317e-05,
      "loss": 0.1968,
      "step": 156760
    },
    {
      "epoch": 60.88,
      "learning_rate": 1.1882459546925567e-05,
      "loss": 0.0801,
      "step": 156770
    },
    {
      "epoch": 60.89,
      "learning_rate": 1.1881941747572817e-05,
      "loss": 0.0899,
      "step": 156780
    },
    {
      "epoch": 60.89,
      "learning_rate": 1.1881423948220066e-05,
      "loss": 0.2303,
      "step": 156790
    },
    {
      "epoch": 60.89,
      "learning_rate": 1.1880906148867314e-05,
      "loss": 0.1224,
      "step": 156800
    },
    {
      "epoch": 60.9,
      "learning_rate": 1.1880388349514564e-05,
      "loss": 0.0699,
      "step": 156810
    },
    {
      "epoch": 60.9,
      "learning_rate": 1.1879870550161813e-05,
      "loss": 0.0803,
      "step": 156820
    },
    {
      "epoch": 60.9,
      "learning_rate": 1.1879352750809063e-05,
      "loss": 0.0285,
      "step": 156830
    },
    {
      "epoch": 60.91,
      "learning_rate": 1.1878834951456311e-05,
      "loss": 0.0492,
      "step": 156840
    },
    {
      "epoch": 60.91,
      "learning_rate": 1.187831715210356e-05,
      "loss": 0.0337,
      "step": 156850
    },
    {
      "epoch": 60.92,
      "learning_rate": 1.187779935275081e-05,
      "loss": 0.162,
      "step": 156860
    },
    {
      "epoch": 60.92,
      "learning_rate": 1.187728155339806e-05,
      "loss": 0.055,
      "step": 156870
    },
    {
      "epoch": 60.92,
      "learning_rate": 1.1876763754045308e-05,
      "loss": 0.1634,
      "step": 156880
    },
    {
      "epoch": 60.93,
      "learning_rate": 1.1876245954692557e-05,
      "loss": 0.0333,
      "step": 156890
    },
    {
      "epoch": 60.93,
      "learning_rate": 1.1875728155339807e-05,
      "loss": 0.0619,
      "step": 156900
    },
    {
      "epoch": 60.94,
      "learning_rate": 1.1875210355987057e-05,
      "loss": 0.0288,
      "step": 156910
    },
    {
      "epoch": 60.94,
      "learning_rate": 1.1874692556634305e-05,
      "loss": 0.0368,
      "step": 156920
    },
    {
      "epoch": 60.94,
      "learning_rate": 1.1874174757281554e-05,
      "loss": 0.0874,
      "step": 156930
    },
    {
      "epoch": 60.95,
      "learning_rate": 1.1873656957928804e-05,
      "loss": 0.0169,
      "step": 156940
    },
    {
      "epoch": 60.95,
      "learning_rate": 1.1873139158576052e-05,
      "loss": 0.0823,
      "step": 156950
    },
    {
      "epoch": 60.96,
      "learning_rate": 1.1872621359223301e-05,
      "loss": 0.1826,
      "step": 156960
    },
    {
      "epoch": 60.96,
      "learning_rate": 1.1872103559870551e-05,
      "loss": 0.0922,
      "step": 156970
    },
    {
      "epoch": 60.96,
      "learning_rate": 1.18715857605178e-05,
      "loss": 0.1045,
      "step": 156980
    },
    {
      "epoch": 60.97,
      "learning_rate": 1.187106796116505e-05,
      "loss": 0.1131,
      "step": 156990
    },
    {
      "epoch": 60.97,
      "learning_rate": 1.18705501618123e-05,
      "loss": 0.2021,
      "step": 157000
    },
    {
      "epoch": 60.97,
      "learning_rate": 1.1870032362459548e-05,
      "loss": 0.0268,
      "step": 157010
    },
    {
      "epoch": 60.98,
      "learning_rate": 1.1869514563106797e-05,
      "loss": 0.168,
      "step": 157020
    },
    {
      "epoch": 60.98,
      "learning_rate": 1.1868996763754045e-05,
      "loss": 0.0073,
      "step": 157030
    },
    {
      "epoch": 60.99,
      "learning_rate": 1.1868478964401295e-05,
      "loss": 0.0758,
      "step": 157040
    },
    {
      "epoch": 60.99,
      "learning_rate": 1.1867961165048544e-05,
      "loss": 0.0572,
      "step": 157050
    },
    {
      "epoch": 60.99,
      "learning_rate": 1.1867443365695794e-05,
      "loss": 0.0617,
      "step": 157060
    },
    {
      "epoch": 61.0,
      "learning_rate": 1.1866925566343044e-05,
      "loss": 0.1667,
      "step": 157070
    },
    {
      "epoch": 61.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.3249916434288025,
      "eval_runtime": 8.2057,
      "eval_samples_per_second": 442.986,
      "eval_steps_per_second": 55.449,
      "step": 157075
    },
    {
      "epoch": 61.0,
      "learning_rate": 1.1866407766990293e-05,
      "loss": 0.0288,
      "step": 157080
    },
    {
      "epoch": 61.01,
      "learning_rate": 1.1865889967637541e-05,
      "loss": 0.0709,
      "step": 157090
    },
    {
      "epoch": 61.01,
      "learning_rate": 1.186537216828479e-05,
      "loss": 0.1363,
      "step": 157100
    },
    {
      "epoch": 61.01,
      "learning_rate": 1.1864854368932039e-05,
      "loss": 0.0433,
      "step": 157110
    },
    {
      "epoch": 61.02,
      "learning_rate": 1.1864336569579288e-05,
      "loss": 0.032,
      "step": 157120
    },
    {
      "epoch": 61.02,
      "learning_rate": 1.1863818770226538e-05,
      "loss": 0.0378,
      "step": 157130
    },
    {
      "epoch": 61.03,
      "learning_rate": 1.1863300970873788e-05,
      "loss": 0.0315,
      "step": 157140
    },
    {
      "epoch": 61.03,
      "learning_rate": 1.1862783171521037e-05,
      "loss": 0.213,
      "step": 157150
    },
    {
      "epoch": 61.03,
      "learning_rate": 1.1862265372168287e-05,
      "loss": 0.014,
      "step": 157160
    },
    {
      "epoch": 61.04,
      "learning_rate": 1.1861747572815535e-05,
      "loss": 0.0972,
      "step": 157170
    },
    {
      "epoch": 61.04,
      "learning_rate": 1.1861229773462783e-05,
      "loss": 0.005,
      "step": 157180
    },
    {
      "epoch": 61.04,
      "learning_rate": 1.1860711974110032e-05,
      "loss": 0.0274,
      "step": 157190
    },
    {
      "epoch": 61.05,
      "learning_rate": 1.1860194174757282e-05,
      "loss": 0.0462,
      "step": 157200
    },
    {
      "epoch": 61.05,
      "learning_rate": 1.1859676375404532e-05,
      "loss": 0.1357,
      "step": 157210
    },
    {
      "epoch": 61.06,
      "learning_rate": 1.1859158576051781e-05,
      "loss": 0.0478,
      "step": 157220
    },
    {
      "epoch": 61.06,
      "learning_rate": 1.185864077669903e-05,
      "loss": 0.1394,
      "step": 157230
    },
    {
      "epoch": 61.06,
      "learning_rate": 1.185812297734628e-05,
      "loss": 0.0446,
      "step": 157240
    },
    {
      "epoch": 61.07,
      "learning_rate": 1.1857605177993527e-05,
      "loss": 0.0333,
      "step": 157250
    },
    {
      "epoch": 61.07,
      "learning_rate": 1.1857087378640776e-05,
      "loss": 0.024,
      "step": 157260
    },
    {
      "epoch": 61.08,
      "learning_rate": 1.1856569579288026e-05,
      "loss": 0.0106,
      "step": 157270
    },
    {
      "epoch": 61.08,
      "learning_rate": 1.1856051779935276e-05,
      "loss": 0.0757,
      "step": 157280
    },
    {
      "epoch": 61.08,
      "learning_rate": 1.1855533980582525e-05,
      "loss": 0.2639,
      "step": 157290
    },
    {
      "epoch": 61.09,
      "learning_rate": 1.1855016181229775e-05,
      "loss": 0.0791,
      "step": 157300
    },
    {
      "epoch": 61.09,
      "learning_rate": 1.1854498381877024e-05,
      "loss": 0.0975,
      "step": 157310
    },
    {
      "epoch": 61.1,
      "learning_rate": 1.1853980582524274e-05,
      "loss": 0.0919,
      "step": 157320
    },
    {
      "epoch": 61.1,
      "learning_rate": 1.185346278317152e-05,
      "loss": 0.0501,
      "step": 157330
    },
    {
      "epoch": 61.1,
      "learning_rate": 1.185294498381877e-05,
      "loss": 0.1761,
      "step": 157340
    },
    {
      "epoch": 61.11,
      "learning_rate": 1.185242718446602e-05,
      "loss": 0.1919,
      "step": 157350
    },
    {
      "epoch": 61.11,
      "learning_rate": 1.1851909385113269e-05,
      "loss": 0.0587,
      "step": 157360
    },
    {
      "epoch": 61.11,
      "learning_rate": 1.1851391585760519e-05,
      "loss": 0.0197,
      "step": 157370
    },
    {
      "epoch": 61.12,
      "learning_rate": 1.1850873786407768e-05,
      "loss": 0.0271,
      "step": 157380
    },
    {
      "epoch": 61.12,
      "learning_rate": 1.1850355987055018e-05,
      "loss": 0.0718,
      "step": 157390
    },
    {
      "epoch": 61.13,
      "learning_rate": 1.1849838187702268e-05,
      "loss": 0.002,
      "step": 157400
    },
    {
      "epoch": 61.13,
      "learning_rate": 1.1849320388349514e-05,
      "loss": 0.0826,
      "step": 157410
    },
    {
      "epoch": 61.13,
      "learning_rate": 1.1848802588996763e-05,
      "loss": 0.087,
      "step": 157420
    },
    {
      "epoch": 61.14,
      "learning_rate": 1.1848284789644013e-05,
      "loss": 0.0038,
      "step": 157430
    },
    {
      "epoch": 61.14,
      "learning_rate": 1.1847766990291263e-05,
      "loss": 0.0663,
      "step": 157440
    },
    {
      "epoch": 61.15,
      "learning_rate": 1.1847249190938512e-05,
      "loss": 0.0216,
      "step": 157450
    },
    {
      "epoch": 61.15,
      "learning_rate": 1.1846731391585762e-05,
      "loss": 0.2073,
      "step": 157460
    },
    {
      "epoch": 61.15,
      "learning_rate": 1.1846213592233012e-05,
      "loss": 0.1304,
      "step": 157470
    },
    {
      "epoch": 61.16,
      "learning_rate": 1.1845695792880261e-05,
      "loss": 0.058,
      "step": 157480
    },
    {
      "epoch": 61.16,
      "learning_rate": 1.1845177993527507e-05,
      "loss": 0.0581,
      "step": 157490
    },
    {
      "epoch": 61.17,
      "learning_rate": 1.1844660194174757e-05,
      "loss": 0.0346,
      "step": 157500
    },
    {
      "epoch": 61.17,
      "learning_rate": 1.1844142394822007e-05,
      "loss": 0.1615,
      "step": 157510
    },
    {
      "epoch": 61.17,
      "learning_rate": 1.1843624595469256e-05,
      "loss": 0.1667,
      "step": 157520
    },
    {
      "epoch": 61.18,
      "learning_rate": 1.1843106796116506e-05,
      "loss": 0.1097,
      "step": 157530
    },
    {
      "epoch": 61.18,
      "learning_rate": 1.1842588996763755e-05,
      "loss": 0.1375,
      "step": 157540
    },
    {
      "epoch": 61.18,
      "learning_rate": 1.1842071197411005e-05,
      "loss": 0.0032,
      "step": 157550
    },
    {
      "epoch": 61.19,
      "learning_rate": 1.1841553398058255e-05,
      "loss": 0.0765,
      "step": 157560
    },
    {
      "epoch": 61.19,
      "learning_rate": 1.1841035598705504e-05,
      "loss": 0.0458,
      "step": 157570
    },
    {
      "epoch": 61.2,
      "learning_rate": 1.184051779935275e-05,
      "loss": 0.0089,
      "step": 157580
    },
    {
      "epoch": 61.2,
      "learning_rate": 1.184e-05,
      "loss": 0.0025,
      "step": 157590
    },
    {
      "epoch": 61.2,
      "learning_rate": 1.183948220064725e-05,
      "loss": 0.0066,
      "step": 157600
    },
    {
      "epoch": 61.21,
      "learning_rate": 1.18389644012945e-05,
      "loss": 0.08,
      "step": 157610
    },
    {
      "epoch": 61.21,
      "learning_rate": 1.1838446601941749e-05,
      "loss": 0.0757,
      "step": 157620
    },
    {
      "epoch": 61.22,
      "learning_rate": 1.1837928802588999e-05,
      "loss": 0.064,
      "step": 157630
    },
    {
      "epoch": 61.22,
      "learning_rate": 1.1837411003236248e-05,
      "loss": 0.0586,
      "step": 157640
    },
    {
      "epoch": 61.22,
      "learning_rate": 1.1836893203883498e-05,
      "loss": 0.0556,
      "step": 157650
    },
    {
      "epoch": 61.23,
      "learning_rate": 1.1836375404530744e-05,
      "loss": 0.0383,
      "step": 157660
    },
    {
      "epoch": 61.23,
      "learning_rate": 1.1835857605177994e-05,
      "loss": 0.0171,
      "step": 157670
    },
    {
      "epoch": 61.23,
      "learning_rate": 1.1835339805825243e-05,
      "loss": 0.0281,
      "step": 157680
    },
    {
      "epoch": 61.24,
      "learning_rate": 1.1834822006472493e-05,
      "loss": 0.057,
      "step": 157690
    },
    {
      "epoch": 61.24,
      "learning_rate": 1.1834304207119743e-05,
      "loss": 0.1338,
      "step": 157700
    },
    {
      "epoch": 61.25,
      "learning_rate": 1.1833786407766992e-05,
      "loss": 0.107,
      "step": 157710
    },
    {
      "epoch": 61.25,
      "learning_rate": 1.1833268608414242e-05,
      "loss": 0.0363,
      "step": 157720
    },
    {
      "epoch": 61.25,
      "learning_rate": 1.1832750809061491e-05,
      "loss": 0.1414,
      "step": 157730
    },
    {
      "epoch": 61.26,
      "learning_rate": 1.1832233009708738e-05,
      "loss": 0.0879,
      "step": 157740
    },
    {
      "epoch": 61.26,
      "learning_rate": 1.1831715210355987e-05,
      "loss": 0.0163,
      "step": 157750
    },
    {
      "epoch": 61.27,
      "learning_rate": 1.1831197411003237e-05,
      "loss": 0.0648,
      "step": 157760
    },
    {
      "epoch": 61.27,
      "learning_rate": 1.1830679611650486e-05,
      "loss": 0.0132,
      "step": 157770
    },
    {
      "epoch": 61.27,
      "learning_rate": 1.1830161812297736e-05,
      "loss": 0.0466,
      "step": 157780
    },
    {
      "epoch": 61.28,
      "learning_rate": 1.1829644012944986e-05,
      "loss": 0.1285,
      "step": 157790
    },
    {
      "epoch": 61.28,
      "learning_rate": 1.1829126213592235e-05,
      "loss": 0.0887,
      "step": 157800
    },
    {
      "epoch": 61.29,
      "learning_rate": 1.1828608414239485e-05,
      "loss": 0.0757,
      "step": 157810
    },
    {
      "epoch": 61.29,
      "learning_rate": 1.1828090614886731e-05,
      "loss": 0.0273,
      "step": 157820
    },
    {
      "epoch": 61.29,
      "learning_rate": 1.182757281553398e-05,
      "loss": 0.0724,
      "step": 157830
    },
    {
      "epoch": 61.3,
      "learning_rate": 1.182705501618123e-05,
      "loss": 0.0182,
      "step": 157840
    },
    {
      "epoch": 61.3,
      "learning_rate": 1.182653721682848e-05,
      "loss": 0.0378,
      "step": 157850
    },
    {
      "epoch": 61.3,
      "learning_rate": 1.182601941747573e-05,
      "loss": 0.093,
      "step": 157860
    },
    {
      "epoch": 61.31,
      "learning_rate": 1.182550161812298e-05,
      "loss": 0.0498,
      "step": 157870
    },
    {
      "epoch": 61.31,
      "learning_rate": 1.1824983818770229e-05,
      "loss": 0.0231,
      "step": 157880
    },
    {
      "epoch": 61.32,
      "learning_rate": 1.1824466019417477e-05,
      "loss": 0.1971,
      "step": 157890
    },
    {
      "epoch": 61.32,
      "learning_rate": 1.1823948220064725e-05,
      "loss": 0.0105,
      "step": 157900
    },
    {
      "epoch": 61.32,
      "learning_rate": 1.1823430420711974e-05,
      "loss": 0.0936,
      "step": 157910
    },
    {
      "epoch": 61.33,
      "learning_rate": 1.1822912621359224e-05,
      "loss": 0.0533,
      "step": 157920
    },
    {
      "epoch": 61.33,
      "learning_rate": 1.1822394822006474e-05,
      "loss": 0.0817,
      "step": 157930
    },
    {
      "epoch": 61.34,
      "learning_rate": 1.1821877022653723e-05,
      "loss": 0.0551,
      "step": 157940
    },
    {
      "epoch": 61.34,
      "learning_rate": 1.1821359223300973e-05,
      "loss": 0.0352,
      "step": 157950
    },
    {
      "epoch": 61.34,
      "learning_rate": 1.1820841423948222e-05,
      "loss": 0.0033,
      "step": 157960
    },
    {
      "epoch": 61.35,
      "learning_rate": 1.182032362459547e-05,
      "loss": 0.0906,
      "step": 157970
    },
    {
      "epoch": 61.35,
      "learning_rate": 1.1819805825242718e-05,
      "loss": 0.0855,
      "step": 157980
    },
    {
      "epoch": 61.36,
      "learning_rate": 1.1819288025889968e-05,
      "loss": 0.0081,
      "step": 157990
    },
    {
      "epoch": 61.36,
      "learning_rate": 1.1818770226537218e-05,
      "loss": 0.1398,
      "step": 158000
    },
    {
      "epoch": 61.36,
      "learning_rate": 1.1818252427184467e-05,
      "loss": 0.0598,
      "step": 158010
    },
    {
      "epoch": 61.37,
      "learning_rate": 1.1817734627831717e-05,
      "loss": 0.1733,
      "step": 158020
    },
    {
      "epoch": 61.37,
      "learning_rate": 1.1817216828478966e-05,
      "loss": 0.1032,
      "step": 158030
    },
    {
      "epoch": 61.37,
      "learning_rate": 1.1816699029126214e-05,
      "loss": 0.1334,
      "step": 158040
    },
    {
      "epoch": 61.38,
      "learning_rate": 1.1816181229773464e-05,
      "loss": 0.1433,
      "step": 158050
    },
    {
      "epoch": 61.38,
      "learning_rate": 1.1815663430420712e-05,
      "loss": 0.0595,
      "step": 158060
    },
    {
      "epoch": 61.39,
      "learning_rate": 1.1815145631067961e-05,
      "loss": 0.0122,
      "step": 158070
    },
    {
      "epoch": 61.39,
      "learning_rate": 1.1814627831715211e-05,
      "loss": 0.1087,
      "step": 158080
    },
    {
      "epoch": 61.39,
      "learning_rate": 1.181411003236246e-05,
      "loss": 0.0835,
      "step": 158090
    },
    {
      "epoch": 61.4,
      "learning_rate": 1.181359223300971e-05,
      "loss": 0.1839,
      "step": 158100
    },
    {
      "epoch": 61.4,
      "learning_rate": 1.181307443365696e-05,
      "loss": 0.1435,
      "step": 158110
    },
    {
      "epoch": 61.41,
      "learning_rate": 1.1812556634304208e-05,
      "loss": 0.0221,
      "step": 158120
    },
    {
      "epoch": 61.41,
      "learning_rate": 1.1812038834951457e-05,
      "loss": 0.0722,
      "step": 158130
    },
    {
      "epoch": 61.41,
      "learning_rate": 1.1811521035598707e-05,
      "loss": 0.1012,
      "step": 158140
    },
    {
      "epoch": 61.42,
      "learning_rate": 1.1811003236245955e-05,
      "loss": 0.1024,
      "step": 158150
    },
    {
      "epoch": 61.42,
      "learning_rate": 1.1810485436893205e-05,
      "loss": 0.111,
      "step": 158160
    },
    {
      "epoch": 61.43,
      "learning_rate": 1.1809967637540454e-05,
      "loss": 0.0436,
      "step": 158170
    },
    {
      "epoch": 61.43,
      "learning_rate": 1.1809449838187704e-05,
      "loss": 0.0482,
      "step": 158180
    },
    {
      "epoch": 61.43,
      "learning_rate": 1.1808932038834952e-05,
      "loss": 0.0499,
      "step": 158190
    },
    {
      "epoch": 61.44,
      "learning_rate": 1.1808414239482201e-05,
      "loss": 0.0687,
      "step": 158200
    },
    {
      "epoch": 61.44,
      "learning_rate": 1.1807896440129451e-05,
      "loss": 0.0585,
      "step": 158210
    },
    {
      "epoch": 61.44,
      "learning_rate": 1.18073786407767e-05,
      "loss": 0.059,
      "step": 158220
    },
    {
      "epoch": 61.45,
      "learning_rate": 1.1806860841423949e-05,
      "loss": 0.2574,
      "step": 158230
    },
    {
      "epoch": 61.45,
      "learning_rate": 1.1806343042071198e-05,
      "loss": 0.0391,
      "step": 158240
    },
    {
      "epoch": 61.46,
      "learning_rate": 1.1805825242718448e-05,
      "loss": 0.0202,
      "step": 158250
    },
    {
      "epoch": 61.46,
      "learning_rate": 1.1805307443365697e-05,
      "loss": 0.1099,
      "step": 158260
    },
    {
      "epoch": 61.46,
      "learning_rate": 1.1804789644012945e-05,
      "loss": 0.0453,
      "step": 158270
    },
    {
      "epoch": 61.47,
      "learning_rate": 1.1804271844660195e-05,
      "loss": 0.1198,
      "step": 158280
    },
    {
      "epoch": 61.47,
      "learning_rate": 1.1803754045307445e-05,
      "loss": 0.0562,
      "step": 158290
    },
    {
      "epoch": 61.48,
      "learning_rate": 1.1803236245954694e-05,
      "loss": 0.0322,
      "step": 158300
    },
    {
      "epoch": 61.48,
      "learning_rate": 1.1802718446601942e-05,
      "loss": 0.129,
      "step": 158310
    },
    {
      "epoch": 61.48,
      "learning_rate": 1.1802200647249192e-05,
      "loss": 0.0377,
      "step": 158320
    },
    {
      "epoch": 61.49,
      "learning_rate": 1.1801682847896441e-05,
      "loss": 0.0701,
      "step": 158330
    },
    {
      "epoch": 61.49,
      "learning_rate": 1.1801165048543691e-05,
      "loss": 0.1655,
      "step": 158340
    },
    {
      "epoch": 61.5,
      "learning_rate": 1.1800647249190939e-05,
      "loss": 0.1036,
      "step": 158350
    },
    {
      "epoch": 61.5,
      "learning_rate": 1.1800129449838189e-05,
      "loss": 0.0812,
      "step": 158360
    },
    {
      "epoch": 61.5,
      "learning_rate": 1.1799611650485438e-05,
      "loss": 0.1851,
      "step": 158370
    },
    {
      "epoch": 61.51,
      "learning_rate": 1.1799093851132688e-05,
      "loss": 0.0596,
      "step": 158380
    },
    {
      "epoch": 61.51,
      "learning_rate": 1.1798576051779936e-05,
      "loss": 0.0159,
      "step": 158390
    },
    {
      "epoch": 61.51,
      "learning_rate": 1.1798058252427185e-05,
      "loss": 0.0634,
      "step": 158400
    },
    {
      "epoch": 61.52,
      "learning_rate": 1.1797540453074435e-05,
      "loss": 0.0545,
      "step": 158410
    },
    {
      "epoch": 61.52,
      "learning_rate": 1.1797022653721683e-05,
      "loss": 0.0134,
      "step": 158420
    },
    {
      "epoch": 61.53,
      "learning_rate": 1.1796504854368932e-05,
      "loss": 0.049,
      "step": 158430
    },
    {
      "epoch": 61.53,
      "learning_rate": 1.1795987055016182e-05,
      "loss": 0.073,
      "step": 158440
    },
    {
      "epoch": 61.53,
      "learning_rate": 1.1795469255663432e-05,
      "loss": 0.0162,
      "step": 158450
    },
    {
      "epoch": 61.54,
      "learning_rate": 1.1794951456310681e-05,
      "loss": 0.0304,
      "step": 158460
    },
    {
      "epoch": 61.54,
      "learning_rate": 1.179443365695793e-05,
      "loss": 0.0602,
      "step": 158470
    },
    {
      "epoch": 61.55,
      "learning_rate": 1.1793915857605179e-05,
      "loss": 0.0357,
      "step": 158480
    },
    {
      "epoch": 61.55,
      "learning_rate": 1.1793398058252428e-05,
      "loss": 0.1271,
      "step": 158490
    },
    {
      "epoch": 61.55,
      "learning_rate": 1.1792880258899676e-05,
      "loss": 0.0545,
      "step": 158500
    },
    {
      "epoch": 61.56,
      "learning_rate": 1.1792362459546926e-05,
      "loss": 0.0633,
      "step": 158510
    },
    {
      "epoch": 61.56,
      "learning_rate": 1.1791844660194176e-05,
      "loss": 0.117,
      "step": 158520
    },
    {
      "epoch": 61.57,
      "learning_rate": 1.1791326860841425e-05,
      "loss": 0.0111,
      "step": 158530
    },
    {
      "epoch": 61.57,
      "learning_rate": 1.1790809061488675e-05,
      "loss": 0.0027,
      "step": 158540
    },
    {
      "epoch": 61.57,
      "learning_rate": 1.1790291262135923e-05,
      "loss": 0.0838,
      "step": 158550
    },
    {
      "epoch": 61.58,
      "learning_rate": 1.1789773462783172e-05,
      "loss": 0.0713,
      "step": 158560
    },
    {
      "epoch": 61.58,
      "learning_rate": 1.178925566343042e-05,
      "loss": 0.0617,
      "step": 158570
    },
    {
      "epoch": 61.58,
      "learning_rate": 1.178873786407767e-05,
      "loss": 0.0539,
      "step": 158580
    },
    {
      "epoch": 61.59,
      "learning_rate": 1.178822006472492e-05,
      "loss": 0.0006,
      "step": 158590
    },
    {
      "epoch": 61.59,
      "learning_rate": 1.178770226537217e-05,
      "loss": 0.3234,
      "step": 158600
    },
    {
      "epoch": 61.6,
      "learning_rate": 1.1787184466019419e-05,
      "loss": 0.0705,
      "step": 158610
    },
    {
      "epoch": 61.6,
      "learning_rate": 1.1786666666666668e-05,
      "loss": 0.1577,
      "step": 158620
    },
    {
      "epoch": 61.6,
      "learning_rate": 1.1786148867313916e-05,
      "loss": 0.2519,
      "step": 158630
    },
    {
      "epoch": 61.61,
      "learning_rate": 1.1785631067961166e-05,
      "loss": 0.1127,
      "step": 158640
    },
    {
      "epoch": 61.61,
      "learning_rate": 1.1785113268608414e-05,
      "loss": 0.0696,
      "step": 158650
    },
    {
      "epoch": 61.62,
      "learning_rate": 1.1784595469255664e-05,
      "loss": 0.1386,
      "step": 158660
    },
    {
      "epoch": 61.62,
      "learning_rate": 1.1784077669902913e-05,
      "loss": 0.2308,
      "step": 158670
    },
    {
      "epoch": 61.62,
      "learning_rate": 1.1783559870550163e-05,
      "loss": 0.0256,
      "step": 158680
    },
    {
      "epoch": 61.63,
      "learning_rate": 1.1783042071197412e-05,
      "loss": 0.0578,
      "step": 158690
    },
    {
      "epoch": 61.63,
      "learning_rate": 1.1782524271844662e-05,
      "loss": 0.0738,
      "step": 158700
    },
    {
      "epoch": 61.63,
      "learning_rate": 1.1782006472491912e-05,
      "loss": 0.0882,
      "step": 158710
    },
    {
      "epoch": 61.64,
      "learning_rate": 1.1781488673139158e-05,
      "loss": 0.0499,
      "step": 158720
    },
    {
      "epoch": 61.64,
      "learning_rate": 1.1780970873786407e-05,
      "loss": 0.1312,
      "step": 158730
    },
    {
      "epoch": 61.65,
      "learning_rate": 1.1780453074433657e-05,
      "loss": 0.0989,
      "step": 158740
    },
    {
      "epoch": 61.65,
      "learning_rate": 1.1779935275080907e-05,
      "loss": 0.0963,
      "step": 158750
    },
    {
      "epoch": 61.65,
      "learning_rate": 1.1779417475728156e-05,
      "loss": 0.0188,
      "step": 158760
    },
    {
      "epoch": 61.66,
      "learning_rate": 1.1778899676375406e-05,
      "loss": 0.0343,
      "step": 158770
    },
    {
      "epoch": 61.66,
      "learning_rate": 1.1778381877022656e-05,
      "loss": 0.0876,
      "step": 158780
    },
    {
      "epoch": 61.67,
      "learning_rate": 1.1777864077669905e-05,
      "loss": 0.0933,
      "step": 158790
    },
    {
      "epoch": 61.67,
      "learning_rate": 1.1777346278317151e-05,
      "loss": 0.1486,
      "step": 158800
    },
    {
      "epoch": 61.67,
      "learning_rate": 1.1776828478964401e-05,
      "loss": 0.0827,
      "step": 158810
    },
    {
      "epoch": 61.68,
      "learning_rate": 1.177631067961165e-05,
      "loss": 0.099,
      "step": 158820
    },
    {
      "epoch": 61.68,
      "learning_rate": 1.17757928802589e-05,
      "loss": 0.0519,
      "step": 158830
    },
    {
      "epoch": 61.69,
      "learning_rate": 1.177527508090615e-05,
      "loss": 0.0313,
      "step": 158840
    },
    {
      "epoch": 61.69,
      "learning_rate": 1.17747572815534e-05,
      "loss": 0.019,
      "step": 158850
    },
    {
      "epoch": 61.69,
      "learning_rate": 1.1774239482200649e-05,
      "loss": 0.1131,
      "step": 158860
    },
    {
      "epoch": 61.7,
      "learning_rate": 1.1773721682847899e-05,
      "loss": 0.1469,
      "step": 158870
    },
    {
      "epoch": 61.7,
      "learning_rate": 1.1773203883495145e-05,
      "loss": 0.1762,
      "step": 158880
    },
    {
      "epoch": 61.7,
      "learning_rate": 1.1772686084142395e-05,
      "loss": 0.0182,
      "step": 158890
    },
    {
      "epoch": 61.71,
      "learning_rate": 1.1772168284789644e-05,
      "loss": 0.0268,
      "step": 158900
    },
    {
      "epoch": 61.71,
      "learning_rate": 1.1771650485436894e-05,
      "loss": 0.0158,
      "step": 158910
    },
    {
      "epoch": 61.72,
      "learning_rate": 1.1771132686084143e-05,
      "loss": 0.0685,
      "step": 158920
    },
    {
      "epoch": 61.72,
      "learning_rate": 1.1770614886731393e-05,
      "loss": 0.0268,
      "step": 158930
    },
    {
      "epoch": 61.72,
      "learning_rate": 1.1770097087378643e-05,
      "loss": 0.0462,
      "step": 158940
    },
    {
      "epoch": 61.73,
      "learning_rate": 1.1769579288025892e-05,
      "loss": 0.0371,
      "step": 158950
    },
    {
      "epoch": 61.73,
      "learning_rate": 1.1769061488673139e-05,
      "loss": 0.0247,
      "step": 158960
    },
    {
      "epoch": 61.74,
      "learning_rate": 1.1768543689320388e-05,
      "loss": 0.0549,
      "step": 158970
    },
    {
      "epoch": 61.74,
      "learning_rate": 1.1768025889967638e-05,
      "loss": 0.0905,
      "step": 158980
    },
    {
      "epoch": 61.74,
      "learning_rate": 1.1767508090614887e-05,
      "loss": 0.0003,
      "step": 158990
    },
    {
      "epoch": 61.75,
      "learning_rate": 1.1766990291262137e-05,
      "loss": 0.0618,
      "step": 159000
    },
    {
      "epoch": 61.75,
      "learning_rate": 1.1766472491909387e-05,
      "loss": 0.0971,
      "step": 159010
    },
    {
      "epoch": 61.76,
      "learning_rate": 1.1765954692556636e-05,
      "loss": 0.043,
      "step": 159020
    },
    {
      "epoch": 61.76,
      "learning_rate": 1.1765436893203886e-05,
      "loss": 0.0469,
      "step": 159030
    },
    {
      "epoch": 61.76,
      "learning_rate": 1.1764919093851132e-05,
      "loss": 0.148,
      "step": 159040
    },
    {
      "epoch": 61.77,
      "learning_rate": 1.1764401294498382e-05,
      "loss": 0.0253,
      "step": 159050
    },
    {
      "epoch": 61.77,
      "learning_rate": 1.1763883495145631e-05,
      "loss": 0.047,
      "step": 159060
    },
    {
      "epoch": 61.77,
      "learning_rate": 1.1763365695792881e-05,
      "loss": 0.0921,
      "step": 159070
    },
    {
      "epoch": 61.78,
      "learning_rate": 1.176284789644013e-05,
      "loss": 0.0656,
      "step": 159080
    },
    {
      "epoch": 61.78,
      "learning_rate": 1.176233009708738e-05,
      "loss": 0.031,
      "step": 159090
    },
    {
      "epoch": 61.79,
      "learning_rate": 1.176181229773463e-05,
      "loss": 0.1548,
      "step": 159100
    },
    {
      "epoch": 61.79,
      "learning_rate": 1.176129449838188e-05,
      "loss": 0.1277,
      "step": 159110
    },
    {
      "epoch": 61.79,
      "learning_rate": 1.1760776699029126e-05,
      "loss": 0.0554,
      "step": 159120
    },
    {
      "epoch": 61.8,
      "learning_rate": 1.1760258899676375e-05,
      "loss": 0.0493,
      "step": 159130
    },
    {
      "epoch": 61.8,
      "learning_rate": 1.1759741100323625e-05,
      "loss": 0.0346,
      "step": 159140
    },
    {
      "epoch": 61.81,
      "learning_rate": 1.1759223300970874e-05,
      "loss": 0.0332,
      "step": 159150
    },
    {
      "epoch": 61.81,
      "learning_rate": 1.1758705501618124e-05,
      "loss": 0.1455,
      "step": 159160
    },
    {
      "epoch": 61.81,
      "learning_rate": 1.1758187702265374e-05,
      "loss": 0.003,
      "step": 159170
    },
    {
      "epoch": 61.82,
      "learning_rate": 1.1757669902912623e-05,
      "loss": 0.0016,
      "step": 159180
    },
    {
      "epoch": 61.82,
      "learning_rate": 1.1757152103559873e-05,
      "loss": 0.0646,
      "step": 159190
    },
    {
      "epoch": 61.83,
      "learning_rate": 1.175663430420712e-05,
      "loss": 0.0492,
      "step": 159200
    },
    {
      "epoch": 61.83,
      "learning_rate": 1.1756116504854369e-05,
      "loss": 0.1667,
      "step": 159210
    },
    {
      "epoch": 61.83,
      "learning_rate": 1.1755598705501618e-05,
      "loss": 0.0378,
      "step": 159220
    },
    {
      "epoch": 61.84,
      "learning_rate": 1.1755080906148868e-05,
      "loss": 0.1507,
      "step": 159230
    },
    {
      "epoch": 61.84,
      "learning_rate": 1.1754563106796118e-05,
      "loss": 0.0409,
      "step": 159240
    },
    {
      "epoch": 61.84,
      "learning_rate": 1.1754045307443367e-05,
      "loss": 0.0048,
      "step": 159250
    },
    {
      "epoch": 61.85,
      "learning_rate": 1.1753527508090617e-05,
      "loss": 0.0677,
      "step": 159260
    },
    {
      "epoch": 61.85,
      "learning_rate": 1.1753009708737867e-05,
      "loss": 0.0637,
      "step": 159270
    },
    {
      "epoch": 61.86,
      "learning_rate": 1.1752491909385116e-05,
      "loss": 0.1037,
      "step": 159280
    },
    {
      "epoch": 61.86,
      "learning_rate": 1.1751974110032362e-05,
      "loss": 0.0724,
      "step": 159290
    },
    {
      "epoch": 61.86,
      "learning_rate": 1.1751456310679612e-05,
      "loss": 0.087,
      "step": 159300
    },
    {
      "epoch": 61.87,
      "learning_rate": 1.1750938511326862e-05,
      "loss": 0.1013,
      "step": 159310
    },
    {
      "epoch": 61.87,
      "learning_rate": 1.1750420711974111e-05,
      "loss": 0.0971,
      "step": 159320
    },
    {
      "epoch": 61.88,
      "learning_rate": 1.174990291262136e-05,
      "loss": 0.044,
      "step": 159330
    },
    {
      "epoch": 61.88,
      "learning_rate": 1.174938511326861e-05,
      "loss": 0.0407,
      "step": 159340
    },
    {
      "epoch": 61.88,
      "learning_rate": 1.174886731391586e-05,
      "loss": 0.0866,
      "step": 159350
    },
    {
      "epoch": 61.89,
      "learning_rate": 1.1748349514563108e-05,
      "loss": 0.0331,
      "step": 159360
    },
    {
      "epoch": 61.89,
      "learning_rate": 1.1747831715210356e-05,
      "loss": 0.0297,
      "step": 159370
    },
    {
      "epoch": 61.9,
      "learning_rate": 1.1747313915857606e-05,
      "loss": 0.1338,
      "step": 159380
    },
    {
      "epoch": 61.9,
      "learning_rate": 1.1746796116504855e-05,
      "loss": 0.0109,
      "step": 159390
    },
    {
      "epoch": 61.9,
      "learning_rate": 1.1746278317152105e-05,
      "loss": 0.0343,
      "step": 159400
    },
    {
      "epoch": 61.91,
      "learning_rate": 1.1745760517799354e-05,
      "loss": 0.0107,
      "step": 159410
    },
    {
      "epoch": 61.91,
      "learning_rate": 1.1745242718446604e-05,
      "loss": 0.0006,
      "step": 159420
    },
    {
      "epoch": 61.91,
      "learning_rate": 1.1744724919093854e-05,
      "loss": 0.1806,
      "step": 159430
    },
    {
      "epoch": 61.92,
      "learning_rate": 1.1744207119741102e-05,
      "loss": 0.1237,
      "step": 159440
    },
    {
      "epoch": 61.92,
      "learning_rate": 1.174368932038835e-05,
      "loss": 0.0542,
      "step": 159450
    },
    {
      "epoch": 61.93,
      "learning_rate": 1.1743171521035599e-05,
      "loss": 0.0347,
      "step": 159460
    },
    {
      "epoch": 61.93,
      "learning_rate": 1.1742653721682849e-05,
      "loss": 0.0871,
      "step": 159470
    },
    {
      "epoch": 61.93,
      "learning_rate": 1.1742135922330098e-05,
      "loss": 0.07,
      "step": 159480
    },
    {
      "epoch": 61.94,
      "learning_rate": 1.1741618122977348e-05,
      "loss": 0.0381,
      "step": 159490
    },
    {
      "epoch": 61.94,
      "learning_rate": 1.1741100323624598e-05,
      "loss": 0.1865,
      "step": 159500
    },
    {
      "epoch": 61.95,
      "learning_rate": 1.1740582524271845e-05,
      "loss": 0.1099,
      "step": 159510
    },
    {
      "epoch": 61.95,
      "learning_rate": 1.1740064724919095e-05,
      "loss": 0.1264,
      "step": 159520
    },
    {
      "epoch": 61.95,
      "learning_rate": 1.1739546925566343e-05,
      "loss": 0.059,
      "step": 159530
    },
    {
      "epoch": 61.96,
      "learning_rate": 1.1739029126213593e-05,
      "loss": 0.0208,
      "step": 159540
    },
    {
      "epoch": 61.96,
      "learning_rate": 1.1738511326860842e-05,
      "loss": 0.048,
      "step": 159550
    },
    {
      "epoch": 61.97,
      "learning_rate": 1.1737993527508092e-05,
      "loss": 0.0655,
      "step": 159560
    },
    {
      "epoch": 61.97,
      "learning_rate": 1.1737475728155342e-05,
      "loss": 0.0386,
      "step": 159570
    },
    {
      "epoch": 61.97,
      "learning_rate": 1.1736957928802591e-05,
      "loss": 0.034,
      "step": 159580
    },
    {
      "epoch": 61.98,
      "learning_rate": 1.1736440129449839e-05,
      "loss": 0.0499,
      "step": 159590
    },
    {
      "epoch": 61.98,
      "learning_rate": 1.1735922330097089e-05,
      "loss": 0.0906,
      "step": 159600
    },
    {
      "epoch": 61.98,
      "learning_rate": 1.1735404530744337e-05,
      "loss": 0.1151,
      "step": 159610
    },
    {
      "epoch": 61.99,
      "learning_rate": 1.1734886731391586e-05,
      "loss": 0.3572,
      "step": 159620
    },
    {
      "epoch": 61.99,
      "learning_rate": 1.1734368932038836e-05,
      "loss": 0.0317,
      "step": 159630
    },
    {
      "epoch": 62.0,
      "learning_rate": 1.1733851132686085e-05,
      "loss": 0.1341,
      "step": 159640
    },
    {
      "epoch": 62.0,
      "learning_rate": 1.1733333333333335e-05,
      "loss": 0.0497,
      "step": 159650
    },
    {
      "epoch": 62.0,
      "eval_accuracy": 0.9466299862448418,
      "eval_loss": 0.33754032850265503,
      "eval_runtime": 8.2233,
      "eval_samples_per_second": 442.039,
      "eval_steps_per_second": 55.331,
      "step": 159650
    },
    {
      "epoch": 62.0,
      "learning_rate": 1.1732815533980583e-05,
      "loss": 0.1406,
      "step": 159660
    },
    {
      "epoch": 62.01,
      "learning_rate": 1.1732297734627833e-05,
      "loss": 0.0233,
      "step": 159670
    },
    {
      "epoch": 62.01,
      "learning_rate": 1.1731779935275082e-05,
      "loss": 0.0234,
      "step": 159680
    },
    {
      "epoch": 62.02,
      "learning_rate": 1.173126213592233e-05,
      "loss": 0.041,
      "step": 159690
    },
    {
      "epoch": 62.02,
      "learning_rate": 1.173074433656958e-05,
      "loss": 0.1109,
      "step": 159700
    },
    {
      "epoch": 62.02,
      "learning_rate": 1.173022653721683e-05,
      "loss": 0.0554,
      "step": 159710
    },
    {
      "epoch": 62.03,
      "learning_rate": 1.1729708737864079e-05,
      "loss": 0.0368,
      "step": 159720
    },
    {
      "epoch": 62.03,
      "learning_rate": 1.1729190938511329e-05,
      "loss": 0.168,
      "step": 159730
    },
    {
      "epoch": 62.03,
      "learning_rate": 1.1728673139158577e-05,
      "loss": 0.025,
      "step": 159740
    },
    {
      "epoch": 62.04,
      "learning_rate": 1.1728155339805826e-05,
      "loss": 0.0005,
      "step": 159750
    },
    {
      "epoch": 62.04,
      "learning_rate": 1.1727637540453076e-05,
      "loss": 0.0546,
      "step": 159760
    },
    {
      "epoch": 62.05,
      "learning_rate": 1.1727119741100325e-05,
      "loss": 0.0047,
      "step": 159770
    },
    {
      "epoch": 62.05,
      "learning_rate": 1.1726601941747573e-05,
      "loss": 0.106,
      "step": 159780
    },
    {
      "epoch": 62.05,
      "learning_rate": 1.1726084142394823e-05,
      "loss": 0.0217,
      "step": 159790
    },
    {
      "epoch": 62.06,
      "learning_rate": 1.1725566343042073e-05,
      "loss": 0.0271,
      "step": 159800
    },
    {
      "epoch": 62.06,
      "learning_rate": 1.1725048543689322e-05,
      "loss": 0.0968,
      "step": 159810
    },
    {
      "epoch": 62.07,
      "learning_rate": 1.172453074433657e-05,
      "loss": 0.0494,
      "step": 159820
    },
    {
      "epoch": 62.07,
      "learning_rate": 1.172401294498382e-05,
      "loss": 0.3305,
      "step": 159830
    },
    {
      "epoch": 62.07,
      "learning_rate": 1.172349514563107e-05,
      "loss": 0.1541,
      "step": 159840
    },
    {
      "epoch": 62.08,
      "learning_rate": 1.1722977346278319e-05,
      "loss": 0.1962,
      "step": 159850
    },
    {
      "epoch": 62.08,
      "learning_rate": 1.1722459546925567e-05,
      "loss": 0.0081,
      "step": 159860
    },
    {
      "epoch": 62.09,
      "learning_rate": 1.1721941747572816e-05,
      "loss": 0.1279,
      "step": 159870
    },
    {
      "epoch": 62.09,
      "learning_rate": 1.1721423948220066e-05,
      "loss": 0.1206,
      "step": 159880
    },
    {
      "epoch": 62.09,
      "learning_rate": 1.1720906148867314e-05,
      "loss": 0.002,
      "step": 159890
    },
    {
      "epoch": 62.1,
      "learning_rate": 1.1720388349514564e-05,
      "loss": 0.0952,
      "step": 159900
    },
    {
      "epoch": 62.1,
      "learning_rate": 1.1719870550161813e-05,
      "loss": 0.2279,
      "step": 159910
    },
    {
      "epoch": 62.1,
      "learning_rate": 1.1719352750809063e-05,
      "loss": 0.2049,
      "step": 159920
    },
    {
      "epoch": 62.11,
      "learning_rate": 1.1718834951456313e-05,
      "loss": 0.0516,
      "step": 159930
    },
    {
      "epoch": 62.11,
      "learning_rate": 1.171831715210356e-05,
      "loss": 0.0086,
      "step": 159940
    },
    {
      "epoch": 62.12,
      "learning_rate": 1.171779935275081e-05,
      "loss": 0.0817,
      "step": 159950
    },
    {
      "epoch": 62.12,
      "learning_rate": 1.171728155339806e-05,
      "loss": 0.0617,
      "step": 159960
    },
    {
      "epoch": 62.12,
      "learning_rate": 1.1716763754045308e-05,
      "loss": 0.0484,
      "step": 159970
    },
    {
      "epoch": 62.13,
      "learning_rate": 1.1716245954692557e-05,
      "loss": 0.0197,
      "step": 159980
    },
    {
      "epoch": 62.13,
      "learning_rate": 1.1715728155339807e-05,
      "loss": 0.1033,
      "step": 159990
    },
    {
      "epoch": 62.14,
      "learning_rate": 1.1715210355987056e-05,
      "loss": 0.0264,
      "step": 160000
    },
    {
      "epoch": 62.14,
      "learning_rate": 1.1714692556634306e-05,
      "loss": 0.0557,
      "step": 160010
    },
    {
      "epoch": 62.14,
      "learning_rate": 1.1714174757281554e-05,
      "loss": 0.0809,
      "step": 160020
    },
    {
      "epoch": 62.15,
      "learning_rate": 1.1713656957928804e-05,
      "loss": 0.1431,
      "step": 160030
    },
    {
      "epoch": 62.15,
      "learning_rate": 1.1713139158576052e-05,
      "loss": 0.0326,
      "step": 160040
    },
    {
      "epoch": 62.16,
      "learning_rate": 1.1712621359223301e-05,
      "loss": 0.0503,
      "step": 160050
    },
    {
      "epoch": 62.16,
      "learning_rate": 1.171210355987055e-05,
      "loss": 0.0215,
      "step": 160060
    },
    {
      "epoch": 62.16,
      "learning_rate": 1.17115857605178e-05,
      "loss": 0.1465,
      "step": 160070
    },
    {
      "epoch": 62.17,
      "learning_rate": 1.171106796116505e-05,
      "loss": 0.0962,
      "step": 160080
    },
    {
      "epoch": 62.17,
      "learning_rate": 1.17105501618123e-05,
      "loss": 0.0178,
      "step": 160090
    },
    {
      "epoch": 62.17,
      "learning_rate": 1.1710032362459548e-05,
      "loss": 0.0257,
      "step": 160100
    },
    {
      "epoch": 62.18,
      "learning_rate": 1.1709514563106797e-05,
      "loss": 0.0532,
      "step": 160110
    },
    {
      "epoch": 62.18,
      "learning_rate": 1.1708996763754045e-05,
      "loss": 0.1192,
      "step": 160120
    },
    {
      "epoch": 62.19,
      "learning_rate": 1.1708478964401295e-05,
      "loss": 0.0424,
      "step": 160130
    },
    {
      "epoch": 62.19,
      "learning_rate": 1.1707961165048544e-05,
      "loss": 0.0758,
      "step": 160140
    },
    {
      "epoch": 62.19,
      "learning_rate": 1.1707443365695794e-05,
      "loss": 0.1262,
      "step": 160150
    },
    {
      "epoch": 62.2,
      "learning_rate": 1.1706925566343044e-05,
      "loss": 0.0393,
      "step": 160160
    },
    {
      "epoch": 62.2,
      "learning_rate": 1.1706407766990293e-05,
      "loss": 0.1037,
      "step": 160170
    },
    {
      "epoch": 62.21,
      "learning_rate": 1.1705889967637541e-05,
      "loss": 0.0362,
      "step": 160180
    },
    {
      "epoch": 62.21,
      "learning_rate": 1.1705372168284789e-05,
      "loss": 0.0867,
      "step": 160190
    },
    {
      "epoch": 62.21,
      "learning_rate": 1.1704854368932039e-05,
      "loss": 0.1144,
      "step": 160200
    },
    {
      "epoch": 62.22,
      "learning_rate": 1.1704336569579288e-05,
      "loss": 0.087,
      "step": 160210
    },
    {
      "epoch": 62.22,
      "learning_rate": 1.1703818770226538e-05,
      "loss": 0.0356,
      "step": 160220
    },
    {
      "epoch": 62.23,
      "learning_rate": 1.1703300970873787e-05,
      "loss": 0.0624,
      "step": 160230
    },
    {
      "epoch": 62.23,
      "learning_rate": 1.1702783171521037e-05,
      "loss": 0.0434,
      "step": 160240
    },
    {
      "epoch": 62.23,
      "learning_rate": 1.1702265372168287e-05,
      "loss": 0.0009,
      "step": 160250
    },
    {
      "epoch": 62.24,
      "learning_rate": 1.1701747572815535e-05,
      "loss": 0.0181,
      "step": 160260
    },
    {
      "epoch": 62.24,
      "learning_rate": 1.1701229773462783e-05,
      "loss": 0.123,
      "step": 160270
    },
    {
      "epoch": 62.24,
      "learning_rate": 1.1700711974110032e-05,
      "loss": 0.0634,
      "step": 160280
    },
    {
      "epoch": 62.25,
      "learning_rate": 1.1700194174757282e-05,
      "loss": 0.0777,
      "step": 160290
    },
    {
      "epoch": 62.25,
      "learning_rate": 1.1699676375404531e-05,
      "loss": 0.0181,
      "step": 160300
    },
    {
      "epoch": 62.26,
      "learning_rate": 1.1699158576051781e-05,
      "loss": 0.0769,
      "step": 160310
    },
    {
      "epoch": 62.26,
      "learning_rate": 1.169864077669903e-05,
      "loss": 0.0425,
      "step": 160320
    },
    {
      "epoch": 62.26,
      "learning_rate": 1.169812297734628e-05,
      "loss": 0.1036,
      "step": 160330
    },
    {
      "epoch": 62.27,
      "learning_rate": 1.169760517799353e-05,
      "loss": 0.0013,
      "step": 160340
    },
    {
      "epoch": 62.27,
      "learning_rate": 1.1697087378640776e-05,
      "loss": 0.0923,
      "step": 160350
    },
    {
      "epoch": 62.28,
      "learning_rate": 1.1696569579288026e-05,
      "loss": 0.053,
      "step": 160360
    },
    {
      "epoch": 62.28,
      "learning_rate": 1.1696051779935275e-05,
      "loss": 0.0307,
      "step": 160370
    },
    {
      "epoch": 62.28,
      "learning_rate": 1.1695533980582525e-05,
      "loss": 0.0002,
      "step": 160380
    },
    {
      "epoch": 62.29,
      "learning_rate": 1.1695016181229775e-05,
      "loss": 0.1323,
      "step": 160390
    },
    {
      "epoch": 62.29,
      "learning_rate": 1.1694498381877024e-05,
      "loss": 0.0331,
      "step": 160400
    },
    {
      "epoch": 62.3,
      "learning_rate": 1.1693980582524274e-05,
      "loss": 0.1057,
      "step": 160410
    },
    {
      "epoch": 62.3,
      "learning_rate": 1.1693462783171523e-05,
      "loss": 0.0138,
      "step": 160420
    },
    {
      "epoch": 62.3,
      "learning_rate": 1.169294498381877e-05,
      "loss": 0.0436,
      "step": 160430
    },
    {
      "epoch": 62.31,
      "learning_rate": 1.169242718446602e-05,
      "loss": 0.0245,
      "step": 160440
    },
    {
      "epoch": 62.31,
      "learning_rate": 1.1691909385113269e-05,
      "loss": 0.1289,
      "step": 160450
    },
    {
      "epoch": 62.31,
      "learning_rate": 1.1691391585760519e-05,
      "loss": 0.0144,
      "step": 160460
    },
    {
      "epoch": 62.32,
      "learning_rate": 1.1690873786407768e-05,
      "loss": 0.0941,
      "step": 160470
    },
    {
      "epoch": 62.32,
      "learning_rate": 1.1690355987055018e-05,
      "loss": 0.0437,
      "step": 160480
    },
    {
      "epoch": 62.33,
      "learning_rate": 1.1689838187702267e-05,
      "loss": 0.3188,
      "step": 160490
    },
    {
      "epoch": 62.33,
      "learning_rate": 1.1689320388349517e-05,
      "loss": 0.0183,
      "step": 160500
    },
    {
      "epoch": 62.33,
      "learning_rate": 1.1688802588996763e-05,
      "loss": 0.1731,
      "step": 160510
    },
    {
      "epoch": 62.34,
      "learning_rate": 1.1688284789644013e-05,
      "loss": 0.0832,
      "step": 160520
    },
    {
      "epoch": 62.34,
      "learning_rate": 1.1687766990291262e-05,
      "loss": 0.1293,
      "step": 160530
    },
    {
      "epoch": 62.35,
      "learning_rate": 1.1687249190938512e-05,
      "loss": 0.1393,
      "step": 160540
    },
    {
      "epoch": 62.35,
      "learning_rate": 1.1686731391585762e-05,
      "loss": 0.0709,
      "step": 160550
    },
    {
      "epoch": 62.35,
      "learning_rate": 1.1686213592233011e-05,
      "loss": 0.0011,
      "step": 160560
    },
    {
      "epoch": 62.36,
      "learning_rate": 1.1685695792880261e-05,
      "loss": 0.0051,
      "step": 160570
    },
    {
      "epoch": 62.36,
      "learning_rate": 1.168517799352751e-05,
      "loss": 0.0704,
      "step": 160580
    },
    {
      "epoch": 62.37,
      "learning_rate": 1.1684660194174757e-05,
      "loss": 0.0175,
      "step": 160590
    },
    {
      "epoch": 62.37,
      "learning_rate": 1.1684142394822006e-05,
      "loss": 0.057,
      "step": 160600
    },
    {
      "epoch": 62.37,
      "learning_rate": 1.1683624595469256e-05,
      "loss": 0.0788,
      "step": 160610
    },
    {
      "epoch": 62.38,
      "learning_rate": 1.1683106796116506e-05,
      "loss": 0.1073,
      "step": 160620
    },
    {
      "epoch": 62.38,
      "learning_rate": 1.1682588996763755e-05,
      "loss": 0.065,
      "step": 160630
    },
    {
      "epoch": 62.38,
      "learning_rate": 1.1682071197411005e-05,
      "loss": 0.0454,
      "step": 160640
    },
    {
      "epoch": 62.39,
      "learning_rate": 1.1681553398058255e-05,
      "loss": 0.0352,
      "step": 160650
    },
    {
      "epoch": 62.39,
      "learning_rate": 1.1681035598705504e-05,
      "loss": 0.2588,
      "step": 160660
    },
    {
      "epoch": 62.4,
      "learning_rate": 1.168051779935275e-05,
      "loss": 0.1417,
      "step": 160670
    },
    {
      "epoch": 62.4,
      "learning_rate": 1.168e-05,
      "loss": 0.0188,
      "step": 160680
    },
    {
      "epoch": 62.4,
      "learning_rate": 1.167948220064725e-05,
      "loss": 0.0936,
      "step": 160690
    },
    {
      "epoch": 62.41,
      "learning_rate": 1.16789644012945e-05,
      "loss": 0.0251,
      "step": 160700
    },
    {
      "epoch": 62.41,
      "learning_rate": 1.1678446601941749e-05,
      "loss": 0.0663,
      "step": 160710
    },
    {
      "epoch": 62.42,
      "learning_rate": 1.1677928802588998e-05,
      "loss": 0.0834,
      "step": 160720
    },
    {
      "epoch": 62.42,
      "learning_rate": 1.1677411003236248e-05,
      "loss": 0.0544,
      "step": 160730
    },
    {
      "epoch": 62.42,
      "learning_rate": 1.1676893203883498e-05,
      "loss": 0.0421,
      "step": 160740
    },
    {
      "epoch": 62.43,
      "learning_rate": 1.1676375404530744e-05,
      "loss": 0.0105,
      "step": 160750
    },
    {
      "epoch": 62.43,
      "learning_rate": 1.1675857605177994e-05,
      "loss": 0.0942,
      "step": 160760
    },
    {
      "epoch": 62.43,
      "learning_rate": 1.1675339805825243e-05,
      "loss": 0.0843,
      "step": 160770
    },
    {
      "epoch": 62.44,
      "learning_rate": 1.1674822006472493e-05,
      "loss": 0.0362,
      "step": 160780
    },
    {
      "epoch": 62.44,
      "learning_rate": 1.1674304207119742e-05,
      "loss": 0.0593,
      "step": 160790
    },
    {
      "epoch": 62.45,
      "learning_rate": 1.1673786407766992e-05,
      "loss": 0.1021,
      "step": 160800
    },
    {
      "epoch": 62.45,
      "learning_rate": 1.1673268608414242e-05,
      "loss": 0.0887,
      "step": 160810
    },
    {
      "epoch": 62.45,
      "learning_rate": 1.1672750809061491e-05,
      "loss": 0.0649,
      "step": 160820
    },
    {
      "epoch": 62.46,
      "learning_rate": 1.1672233009708737e-05,
      "loss": 0.1001,
      "step": 160830
    },
    {
      "epoch": 62.46,
      "learning_rate": 1.1671715210355987e-05,
      "loss": 0.122,
      "step": 160840
    },
    {
      "epoch": 62.47,
      "learning_rate": 1.1671197411003237e-05,
      "loss": 0.0802,
      "step": 160850
    },
    {
      "epoch": 62.47,
      "learning_rate": 1.1670679611650486e-05,
      "loss": 0.0879,
      "step": 160860
    },
    {
      "epoch": 62.47,
      "learning_rate": 1.1670161812297736e-05,
      "loss": 0.0866,
      "step": 160870
    },
    {
      "epoch": 62.48,
      "learning_rate": 1.1669644012944986e-05,
      "loss": 0.0098,
      "step": 160880
    },
    {
      "epoch": 62.48,
      "learning_rate": 1.1669126213592235e-05,
      "loss": 0.1141,
      "step": 160890
    },
    {
      "epoch": 62.49,
      "learning_rate": 1.1668608414239485e-05,
      "loss": 0.0715,
      "step": 160900
    },
    {
      "epoch": 62.49,
      "learning_rate": 1.1668090614886733e-05,
      "loss": 0.0002,
      "step": 160910
    },
    {
      "epoch": 62.49,
      "learning_rate": 1.166757281553398e-05,
      "loss": 0.1001,
      "step": 160920
    },
    {
      "epoch": 62.5,
      "learning_rate": 1.166705501618123e-05,
      "loss": 0.0102,
      "step": 160930
    },
    {
      "epoch": 62.5,
      "learning_rate": 1.166653721682848e-05,
      "loss": 0.3731,
      "step": 160940
    },
    {
      "epoch": 62.5,
      "learning_rate": 1.166601941747573e-05,
      "loss": 0.136,
      "step": 160950
    },
    {
      "epoch": 62.51,
      "learning_rate": 1.1665501618122979e-05,
      "loss": 0.0472,
      "step": 160960
    },
    {
      "epoch": 62.51,
      "learning_rate": 1.1664983818770229e-05,
      "loss": 0.0711,
      "step": 160970
    },
    {
      "epoch": 62.52,
      "learning_rate": 1.1664466019417477e-05,
      "loss": 0.067,
      "step": 160980
    },
    {
      "epoch": 62.52,
      "learning_rate": 1.1663948220064726e-05,
      "loss": 0.0764,
      "step": 160990
    },
    {
      "epoch": 62.52,
      "learning_rate": 1.1663430420711974e-05,
      "loss": 0.0366,
      "step": 161000
    },
    {
      "epoch": 62.53,
      "learning_rate": 1.1662912621359224e-05,
      "loss": 0.0207,
      "step": 161010
    },
    {
      "epoch": 62.53,
      "learning_rate": 1.1662394822006473e-05,
      "loss": 0.1033,
      "step": 161020
    },
    {
      "epoch": 62.54,
      "learning_rate": 1.1661877022653723e-05,
      "loss": 0.0594,
      "step": 161030
    },
    {
      "epoch": 62.54,
      "learning_rate": 1.1661359223300973e-05,
      "loss": 0.0033,
      "step": 161040
    },
    {
      "epoch": 62.54,
      "learning_rate": 1.1660841423948222e-05,
      "loss": 0.1152,
      "step": 161050
    },
    {
      "epoch": 62.55,
      "learning_rate": 1.166032362459547e-05,
      "loss": 0.1302,
      "step": 161060
    },
    {
      "epoch": 62.55,
      "learning_rate": 1.165980582524272e-05,
      "loss": 0.0229,
      "step": 161070
    },
    {
      "epoch": 62.56,
      "learning_rate": 1.1659288025889968e-05,
      "loss": 0.0514,
      "step": 161080
    },
    {
      "epoch": 62.56,
      "learning_rate": 1.1658770226537217e-05,
      "loss": 0.0692,
      "step": 161090
    },
    {
      "epoch": 62.56,
      "learning_rate": 1.1658252427184467e-05,
      "loss": 0.0272,
      "step": 161100
    },
    {
      "epoch": 62.57,
      "learning_rate": 1.1657734627831717e-05,
      "loss": 0.0605,
      "step": 161110
    },
    {
      "epoch": 62.57,
      "learning_rate": 1.1657216828478966e-05,
      "loss": 0.0227,
      "step": 161120
    },
    {
      "epoch": 62.57,
      "learning_rate": 1.1656699029126214e-05,
      "loss": 0.0598,
      "step": 161130
    },
    {
      "epoch": 62.58,
      "learning_rate": 1.1656181229773464e-05,
      "loss": 0.059,
      "step": 161140
    },
    {
      "epoch": 62.58,
      "learning_rate": 1.1655663430420713e-05,
      "loss": 0.1037,
      "step": 161150
    },
    {
      "epoch": 62.59,
      "learning_rate": 1.1655145631067961e-05,
      "loss": 0.1128,
      "step": 161160
    },
    {
      "epoch": 62.59,
      "learning_rate": 1.1654627831715211e-05,
      "loss": 0.1446,
      "step": 161170
    },
    {
      "epoch": 62.59,
      "learning_rate": 1.165411003236246e-05,
      "loss": 0.0477,
      "step": 161180
    },
    {
      "epoch": 62.6,
      "learning_rate": 1.165359223300971e-05,
      "loss": 0.2412,
      "step": 161190
    },
    {
      "epoch": 62.6,
      "learning_rate": 1.165307443365696e-05,
      "loss": 0.0159,
      "step": 161200
    },
    {
      "epoch": 62.61,
      "learning_rate": 1.1652556634304208e-05,
      "loss": 0.0736,
      "step": 161210
    },
    {
      "epoch": 62.61,
      "learning_rate": 1.1652038834951457e-05,
      "loss": 0.0741,
      "step": 161220
    },
    {
      "epoch": 62.61,
      "learning_rate": 1.1651521035598707e-05,
      "loss": 0.0131,
      "step": 161230
    },
    {
      "epoch": 62.62,
      "learning_rate": 1.1651003236245955e-05,
      "loss": 0.0037,
      "step": 161240
    },
    {
      "epoch": 62.62,
      "learning_rate": 1.1650485436893204e-05,
      "loss": 0.102,
      "step": 161250
    },
    {
      "epoch": 62.63,
      "learning_rate": 1.1649967637540454e-05,
      "loss": 0.1175,
      "step": 161260
    },
    {
      "epoch": 62.63,
      "learning_rate": 1.1649449838187704e-05,
      "loss": 0.0133,
      "step": 161270
    },
    {
      "epoch": 62.63,
      "learning_rate": 1.1648932038834952e-05,
      "loss": 0.0872,
      "step": 161280
    },
    {
      "epoch": 62.64,
      "learning_rate": 1.1648414239482201e-05,
      "loss": 0.0774,
      "step": 161290
    },
    {
      "epoch": 62.64,
      "learning_rate": 1.1647896440129451e-05,
      "loss": 0.1211,
      "step": 161300
    },
    {
      "epoch": 62.64,
      "learning_rate": 1.16473786407767e-05,
      "loss": 0.0042,
      "step": 161310
    },
    {
      "epoch": 62.65,
      "learning_rate": 1.1646860841423948e-05,
      "loss": 0.001,
      "step": 161320
    },
    {
      "epoch": 62.65,
      "learning_rate": 1.1646343042071198e-05,
      "loss": 0.1241,
      "step": 161330
    },
    {
      "epoch": 62.66,
      "learning_rate": 1.1645825242718448e-05,
      "loss": 0.0166,
      "step": 161340
    },
    {
      "epoch": 62.66,
      "learning_rate": 1.1645307443365697e-05,
      "loss": 0.1305,
      "step": 161350
    },
    {
      "epoch": 62.66,
      "learning_rate": 1.1644789644012945e-05,
      "loss": 0.0179,
      "step": 161360
    },
    {
      "epoch": 62.67,
      "learning_rate": 1.1644271844660195e-05,
      "loss": 0.0035,
      "step": 161370
    },
    {
      "epoch": 62.67,
      "learning_rate": 1.1643754045307444e-05,
      "loss": 0.0282,
      "step": 161380
    },
    {
      "epoch": 62.68,
      "learning_rate": 1.1643236245954694e-05,
      "loss": 0.1037,
      "step": 161390
    },
    {
      "epoch": 62.68,
      "learning_rate": 1.1642718446601942e-05,
      "loss": 0.1124,
      "step": 161400
    },
    {
      "epoch": 62.68,
      "learning_rate": 1.1642200647249192e-05,
      "loss": 0.0868,
      "step": 161410
    },
    {
      "epoch": 62.69,
      "learning_rate": 1.1641682847896441e-05,
      "loss": 0.1238,
      "step": 161420
    },
    {
      "epoch": 62.69,
      "learning_rate": 1.164116504854369e-05,
      "loss": 0.0396,
      "step": 161430
    },
    {
      "epoch": 62.7,
      "learning_rate": 1.1640647249190939e-05,
      "loss": 0.0723,
      "step": 161440
    },
    {
      "epoch": 62.7,
      "learning_rate": 1.1640129449838188e-05,
      "loss": 0.0699,
      "step": 161450
    },
    {
      "epoch": 62.7,
      "learning_rate": 1.1639611650485438e-05,
      "loss": 0.0002,
      "step": 161460
    },
    {
      "epoch": 62.71,
      "learning_rate": 1.1639093851132688e-05,
      "loss": 0.018,
      "step": 161470
    },
    {
      "epoch": 62.71,
      "learning_rate": 1.1638576051779937e-05,
      "loss": 0.0615,
      "step": 161480
    },
    {
      "epoch": 62.71,
      "learning_rate": 1.1638058252427185e-05,
      "loss": 0.0293,
      "step": 161490
    },
    {
      "epoch": 62.72,
      "learning_rate": 1.1637540453074435e-05,
      "loss": 0.1037,
      "step": 161500
    },
    {
      "epoch": 62.72,
      "learning_rate": 1.1637022653721683e-05,
      "loss": 0.0902,
      "step": 161510
    },
    {
      "epoch": 62.73,
      "learning_rate": 1.1636504854368932e-05,
      "loss": 0.0508,
      "step": 161520
    },
    {
      "epoch": 62.73,
      "learning_rate": 1.1635987055016182e-05,
      "loss": 0.1576,
      "step": 161530
    },
    {
      "epoch": 62.73,
      "learning_rate": 1.1635469255663432e-05,
      "loss": 0.1609,
      "step": 161540
    },
    {
      "epoch": 62.74,
      "learning_rate": 1.1634951456310681e-05,
      "loss": 0.0626,
      "step": 161550
    },
    {
      "epoch": 62.74,
      "learning_rate": 1.163443365695793e-05,
      "loss": 0.1047,
      "step": 161560
    },
    {
      "epoch": 62.75,
      "learning_rate": 1.1633915857605179e-05,
      "loss": 0.0038,
      "step": 161570
    },
    {
      "epoch": 62.75,
      "learning_rate": 1.1633398058252428e-05,
      "loss": 0.1391,
      "step": 161580
    },
    {
      "epoch": 62.75,
      "learning_rate": 1.1632880258899676e-05,
      "loss": 0.047,
      "step": 161590
    },
    {
      "epoch": 62.76,
      "learning_rate": 1.1632362459546926e-05,
      "loss": 0.1438,
      "step": 161600
    },
    {
      "epoch": 62.76,
      "learning_rate": 1.1631844660194175e-05,
      "loss": 0.1288,
      "step": 161610
    },
    {
      "epoch": 62.77,
      "learning_rate": 1.1631326860841425e-05,
      "loss": 0.0121,
      "step": 161620
    },
    {
      "epoch": 62.77,
      "learning_rate": 1.1630809061488675e-05,
      "loss": 0.1414,
      "step": 161630
    },
    {
      "epoch": 62.77,
      "learning_rate": 1.1630291262135924e-05,
      "loss": 0.1377,
      "step": 161640
    },
    {
      "epoch": 62.78,
      "learning_rate": 1.1629773462783172e-05,
      "loss": 0.0324,
      "step": 161650
    },
    {
      "epoch": 62.78,
      "learning_rate": 1.162925566343042e-05,
      "loss": 0.0624,
      "step": 161660
    },
    {
      "epoch": 62.78,
      "learning_rate": 1.162873786407767e-05,
      "loss": 0.0004,
      "step": 161670
    },
    {
      "epoch": 62.79,
      "learning_rate": 1.162822006472492e-05,
      "loss": 0.1327,
      "step": 161680
    },
    {
      "epoch": 62.79,
      "learning_rate": 1.1627702265372169e-05,
      "loss": 0.0141,
      "step": 161690
    },
    {
      "epoch": 62.8,
      "learning_rate": 1.1627184466019419e-05,
      "loss": 0.1224,
      "step": 161700
    },
    {
      "epoch": 62.8,
      "learning_rate": 1.1626666666666668e-05,
      "loss": 0.0706,
      "step": 161710
    },
    {
      "epoch": 62.8,
      "learning_rate": 1.1626148867313918e-05,
      "loss": 0.0458,
      "step": 161720
    },
    {
      "epoch": 62.81,
      "learning_rate": 1.1625631067961166e-05,
      "loss": 0.0183,
      "step": 161730
    },
    {
      "epoch": 62.81,
      "learning_rate": 1.1625113268608414e-05,
      "loss": 0.05,
      "step": 161740
    },
    {
      "epoch": 62.82,
      "learning_rate": 1.1624595469255663e-05,
      "loss": 0.0702,
      "step": 161750
    },
    {
      "epoch": 62.82,
      "learning_rate": 1.1624077669902913e-05,
      "loss": 0.101,
      "step": 161760
    },
    {
      "epoch": 62.82,
      "learning_rate": 1.1623559870550163e-05,
      "loss": 0.0455,
      "step": 161770
    },
    {
      "epoch": 62.83,
      "learning_rate": 1.1623042071197412e-05,
      "loss": 0.0074,
      "step": 161780
    },
    {
      "epoch": 62.83,
      "learning_rate": 1.1622524271844662e-05,
      "loss": 0.0237,
      "step": 161790
    },
    {
      "epoch": 62.83,
      "learning_rate": 1.1622006472491911e-05,
      "loss": 0.141,
      "step": 161800
    },
    {
      "epoch": 62.84,
      "learning_rate": 1.1621488673139158e-05,
      "loss": 0.0845,
      "step": 161810
    },
    {
      "epoch": 62.84,
      "learning_rate": 1.1620970873786407e-05,
      "loss": 0.0577,
      "step": 161820
    },
    {
      "epoch": 62.85,
      "learning_rate": 1.1620453074433657e-05,
      "loss": 0.0075,
      "step": 161830
    },
    {
      "epoch": 62.85,
      "learning_rate": 1.1619935275080907e-05,
      "loss": 0.0389,
      "step": 161840
    },
    {
      "epoch": 62.85,
      "learning_rate": 1.1619417475728156e-05,
      "loss": 0.0006,
      "step": 161850
    },
    {
      "epoch": 62.86,
      "learning_rate": 1.1618899676375406e-05,
      "loss": 0.0037,
      "step": 161860
    },
    {
      "epoch": 62.86,
      "learning_rate": 1.1618381877022655e-05,
      "loss": 0.0061,
      "step": 161870
    },
    {
      "epoch": 62.87,
      "learning_rate": 1.1617864077669905e-05,
      "loss": 0.215,
      "step": 161880
    },
    {
      "epoch": 62.87,
      "learning_rate": 1.1617346278317151e-05,
      "loss": 0.1077,
      "step": 161890
    },
    {
      "epoch": 62.87,
      "learning_rate": 1.1616828478964401e-05,
      "loss": 0.0727,
      "step": 161900
    },
    {
      "epoch": 62.88,
      "learning_rate": 1.161631067961165e-05,
      "loss": 0.1936,
      "step": 161910
    },
    {
      "epoch": 62.88,
      "learning_rate": 1.16157928802589e-05,
      "loss": 0.0056,
      "step": 161920
    },
    {
      "epoch": 62.89,
      "learning_rate": 1.161527508090615e-05,
      "loss": 0.069,
      "step": 161930
    },
    {
      "epoch": 62.89,
      "learning_rate": 1.16147572815534e-05,
      "loss": 0.1229,
      "step": 161940
    },
    {
      "epoch": 62.89,
      "learning_rate": 1.1614239482200649e-05,
      "loss": 0.16,
      "step": 161950
    },
    {
      "epoch": 62.9,
      "learning_rate": 1.1613721682847899e-05,
      "loss": 0.0868,
      "step": 161960
    },
    {
      "epoch": 62.9,
      "learning_rate": 1.1613203883495145e-05,
      "loss": 0.0934,
      "step": 161970
    },
    {
      "epoch": 62.9,
      "learning_rate": 1.1612686084142394e-05,
      "loss": 0.1242,
      "step": 161980
    },
    {
      "epoch": 62.91,
      "learning_rate": 1.1612168284789644e-05,
      "loss": 0.0063,
      "step": 161990
    },
    {
      "epoch": 62.91,
      "learning_rate": 1.1611650485436894e-05,
      "loss": 0.0335,
      "step": 162000
    },
    {
      "epoch": 62.92,
      "learning_rate": 1.1611132686084143e-05,
      "loss": 0.1096,
      "step": 162010
    },
    {
      "epoch": 62.92,
      "learning_rate": 1.1610614886731393e-05,
      "loss": 0.1381,
      "step": 162020
    },
    {
      "epoch": 62.92,
      "learning_rate": 1.1610097087378643e-05,
      "loss": 0.0962,
      "step": 162030
    },
    {
      "epoch": 62.93,
      "learning_rate": 1.1609579288025892e-05,
      "loss": 0.015,
      "step": 162040
    },
    {
      "epoch": 62.93,
      "learning_rate": 1.1609061488673142e-05,
      "loss": 0.0563,
      "step": 162050
    },
    {
      "epoch": 62.94,
      "learning_rate": 1.1608543689320388e-05,
      "loss": 0.0204,
      "step": 162060
    },
    {
      "epoch": 62.94,
      "learning_rate": 1.1608025889967638e-05,
      "loss": 0.1194,
      "step": 162070
    },
    {
      "epoch": 62.94,
      "learning_rate": 1.1607508090614887e-05,
      "loss": 0.0202,
      "step": 162080
    },
    {
      "epoch": 62.95,
      "learning_rate": 1.1606990291262137e-05,
      "loss": 0.0841,
      "step": 162090
    },
    {
      "epoch": 62.95,
      "learning_rate": 1.1606472491909386e-05,
      "loss": 0.0231,
      "step": 162100
    },
    {
      "epoch": 62.96,
      "learning_rate": 1.1605954692556636e-05,
      "loss": 0.0652,
      "step": 162110
    },
    {
      "epoch": 62.96,
      "learning_rate": 1.1605436893203886e-05,
      "loss": 0.0267,
      "step": 162120
    },
    {
      "epoch": 62.96,
      "learning_rate": 1.1604919093851135e-05,
      "loss": 0.0009,
      "step": 162130
    },
    {
      "epoch": 62.97,
      "learning_rate": 1.1604401294498382e-05,
      "loss": 0.0943,
      "step": 162140
    },
    {
      "epoch": 62.97,
      "learning_rate": 1.1603883495145631e-05,
      "loss": 0.0009,
      "step": 162150
    },
    {
      "epoch": 62.97,
      "learning_rate": 1.160336569579288e-05,
      "loss": 0.0972,
      "step": 162160
    },
    {
      "epoch": 62.98,
      "learning_rate": 1.160284789644013e-05,
      "loss": 0.0077,
      "step": 162170
    },
    {
      "epoch": 62.98,
      "learning_rate": 1.160233009708738e-05,
      "loss": 0.0875,
      "step": 162180
    },
    {
      "epoch": 62.99,
      "learning_rate": 1.160181229773463e-05,
      "loss": 0.0965,
      "step": 162190
    },
    {
      "epoch": 62.99,
      "learning_rate": 1.160129449838188e-05,
      "loss": 0.0094,
      "step": 162200
    },
    {
      "epoch": 62.99,
      "learning_rate": 1.1600776699029129e-05,
      "loss": 0.0729,
      "step": 162210
    },
    {
      "epoch": 63.0,
      "learning_rate": 1.1600258899676375e-05,
      "loss": 0.0445,
      "step": 162220
    },
    {
      "epoch": 63.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.34933972358703613,
      "eval_runtime": 8.1937,
      "eval_samples_per_second": 443.635,
      "eval_steps_per_second": 55.531,
      "step": 162225
    },
    {
      "epoch": 63.0,
      "learning_rate": 1.1599741100323625e-05,
      "loss": 0.1234,
      "step": 162230
    },
    {
      "epoch": 63.01,
      "learning_rate": 1.1599223300970874e-05,
      "loss": 0.1836,
      "step": 162240
    },
    {
      "epoch": 63.01,
      "learning_rate": 1.1598705501618124e-05,
      "loss": 0.1387,
      "step": 162250
    },
    {
      "epoch": 63.01,
      "learning_rate": 1.1598187702265374e-05,
      "loss": 0.0033,
      "step": 162260
    },
    {
      "epoch": 63.02,
      "learning_rate": 1.1597669902912623e-05,
      "loss": 0.01,
      "step": 162270
    },
    {
      "epoch": 63.02,
      "learning_rate": 1.1597152103559873e-05,
      "loss": 0.0965,
      "step": 162280
    },
    {
      "epoch": 63.03,
      "learning_rate": 1.1596634304207122e-05,
      "loss": 0.0452,
      "step": 162290
    },
    {
      "epoch": 63.03,
      "learning_rate": 1.1596116504854369e-05,
      "loss": 0.0174,
      "step": 162300
    },
    {
      "epoch": 63.03,
      "learning_rate": 1.1595598705501618e-05,
      "loss": 0.0437,
      "step": 162310
    },
    {
      "epoch": 63.04,
      "learning_rate": 1.1595080906148868e-05,
      "loss": 0.0892,
      "step": 162320
    },
    {
      "epoch": 63.04,
      "learning_rate": 1.1594563106796117e-05,
      "loss": 0.0024,
      "step": 162330
    },
    {
      "epoch": 63.04,
      "learning_rate": 1.1594045307443367e-05,
      "loss": 0.3035,
      "step": 162340
    },
    {
      "epoch": 63.05,
      "learning_rate": 1.1593527508090617e-05,
      "loss": 0.007,
      "step": 162350
    },
    {
      "epoch": 63.05,
      "learning_rate": 1.1593009708737866e-05,
      "loss": 0.0165,
      "step": 162360
    },
    {
      "epoch": 63.06,
      "learning_rate": 1.1592491909385116e-05,
      "loss": 0.0145,
      "step": 162370
    },
    {
      "epoch": 63.06,
      "learning_rate": 1.1591974110032362e-05,
      "loss": 0.123,
      "step": 162380
    },
    {
      "epoch": 63.06,
      "learning_rate": 1.1591456310679612e-05,
      "loss": 0.013,
      "step": 162390
    },
    {
      "epoch": 63.07,
      "learning_rate": 1.1590938511326861e-05,
      "loss": 0.0853,
      "step": 162400
    },
    {
      "epoch": 63.07,
      "learning_rate": 1.1590420711974111e-05,
      "loss": 0.1146,
      "step": 162410
    },
    {
      "epoch": 63.08,
      "learning_rate": 1.158990291262136e-05,
      "loss": 0.0007,
      "step": 162420
    },
    {
      "epoch": 63.08,
      "learning_rate": 1.158938511326861e-05,
      "loss": 0.0025,
      "step": 162430
    },
    {
      "epoch": 63.08,
      "learning_rate": 1.158886731391586e-05,
      "loss": 0.0161,
      "step": 162440
    },
    {
      "epoch": 63.09,
      "learning_rate": 1.1588349514563108e-05,
      "loss": 0.0796,
      "step": 162450
    },
    {
      "epoch": 63.09,
      "learning_rate": 1.1587831715210356e-05,
      "loss": 0.0487,
      "step": 162460
    },
    {
      "epoch": 63.1,
      "learning_rate": 1.1587313915857605e-05,
      "loss": 0.0814,
      "step": 162470
    },
    {
      "epoch": 63.1,
      "learning_rate": 1.1586796116504855e-05,
      "loss": 0.1035,
      "step": 162480
    },
    {
      "epoch": 63.1,
      "learning_rate": 1.1586278317152105e-05,
      "loss": 0.1078,
      "step": 162490
    },
    {
      "epoch": 63.11,
      "learning_rate": 1.1585760517799354e-05,
      "loss": 0.0898,
      "step": 162500
    },
    {
      "epoch": 63.11,
      "learning_rate": 1.1585242718446604e-05,
      "loss": 0.0471,
      "step": 162510
    },
    {
      "epoch": 63.11,
      "learning_rate": 1.1584724919093853e-05,
      "loss": 0.0917,
      "step": 162520
    },
    {
      "epoch": 63.12,
      "learning_rate": 1.1584207119741101e-05,
      "loss": 0.0382,
      "step": 162530
    },
    {
      "epoch": 63.12,
      "learning_rate": 1.158368932038835e-05,
      "loss": 0.0087,
      "step": 162540
    },
    {
      "epoch": 63.13,
      "learning_rate": 1.1583171521035599e-05,
      "loss": 0.0337,
      "step": 162550
    },
    {
      "epoch": 63.13,
      "learning_rate": 1.1582653721682849e-05,
      "loss": 0.0576,
      "step": 162560
    },
    {
      "epoch": 63.13,
      "learning_rate": 1.1582135922330098e-05,
      "loss": 0.0011,
      "step": 162570
    },
    {
      "epoch": 63.14,
      "learning_rate": 1.1581618122977348e-05,
      "loss": 0.0369,
      "step": 162580
    },
    {
      "epoch": 63.14,
      "learning_rate": 1.1581100323624597e-05,
      "loss": 0.0984,
      "step": 162590
    },
    {
      "epoch": 63.15,
      "learning_rate": 1.1580582524271845e-05,
      "loss": 0.062,
      "step": 162600
    },
    {
      "epoch": 63.15,
      "learning_rate": 1.1580064724919095e-05,
      "loss": 0.0089,
      "step": 162610
    },
    {
      "epoch": 63.15,
      "learning_rate": 1.1579546925566345e-05,
      "loss": 0.0878,
      "step": 162620
    },
    {
      "epoch": 63.16,
      "learning_rate": 1.1579029126213592e-05,
      "loss": 0.0446,
      "step": 162630
    },
    {
      "epoch": 63.16,
      "learning_rate": 1.1578511326860842e-05,
      "loss": 0.0064,
      "step": 162640
    },
    {
      "epoch": 63.17,
      "learning_rate": 1.1577993527508092e-05,
      "loss": 0.1143,
      "step": 162650
    },
    {
      "epoch": 63.17,
      "learning_rate": 1.1577475728155341e-05,
      "loss": 0.1178,
      "step": 162660
    },
    {
      "epoch": 63.17,
      "learning_rate": 1.1576957928802591e-05,
      "loss": 0.0475,
      "step": 162670
    },
    {
      "epoch": 63.18,
      "learning_rate": 1.1576440129449839e-05,
      "loss": 0.156,
      "step": 162680
    },
    {
      "epoch": 63.18,
      "learning_rate": 1.1575922330097089e-05,
      "loss": 0.0269,
      "step": 162690
    },
    {
      "epoch": 63.18,
      "learning_rate": 1.1575404530744338e-05,
      "loss": 0.1417,
      "step": 162700
    },
    {
      "epoch": 63.19,
      "learning_rate": 1.1574886731391586e-05,
      "loss": 0.0492,
      "step": 162710
    },
    {
      "epoch": 63.19,
      "learning_rate": 1.1574368932038836e-05,
      "loss": 0.0785,
      "step": 162720
    },
    {
      "epoch": 63.2,
      "learning_rate": 1.1573851132686085e-05,
      "loss": 0.0663,
      "step": 162730
    },
    {
      "epoch": 63.2,
      "learning_rate": 1.1573333333333335e-05,
      "loss": 0.0844,
      "step": 162740
    },
    {
      "epoch": 63.2,
      "learning_rate": 1.1572815533980583e-05,
      "loss": 0.1656,
      "step": 162750
    },
    {
      "epoch": 63.21,
      "learning_rate": 1.1572297734627832e-05,
      "loss": 0.0459,
      "step": 162760
    },
    {
      "epoch": 63.21,
      "learning_rate": 1.1571779935275082e-05,
      "loss": 0.096,
      "step": 162770
    },
    {
      "epoch": 63.22,
      "learning_rate": 1.1571262135922332e-05,
      "loss": 0.1477,
      "step": 162780
    },
    {
      "epoch": 63.22,
      "learning_rate": 1.157074433656958e-05,
      "loss": 0.0711,
      "step": 162790
    },
    {
      "epoch": 63.22,
      "learning_rate": 1.157022653721683e-05,
      "loss": 0.0119,
      "step": 162800
    },
    {
      "epoch": 63.23,
      "learning_rate": 1.1569708737864079e-05,
      "loss": 0.116,
      "step": 162810
    },
    {
      "epoch": 63.23,
      "learning_rate": 1.1569190938511328e-05,
      "loss": 0.1007,
      "step": 162820
    },
    {
      "epoch": 63.23,
      "learning_rate": 1.1568673139158576e-05,
      "loss": 0.0507,
      "step": 162830
    },
    {
      "epoch": 63.24,
      "learning_rate": 1.1568155339805826e-05,
      "loss": 0.2073,
      "step": 162840
    },
    {
      "epoch": 63.24,
      "learning_rate": 1.1567637540453076e-05,
      "loss": 0.0518,
      "step": 162850
    },
    {
      "epoch": 63.25,
      "learning_rate": 1.1567119741100325e-05,
      "loss": 0.1393,
      "step": 162860
    },
    {
      "epoch": 63.25,
      "learning_rate": 1.1566601941747573e-05,
      "loss": 0.0166,
      "step": 162870
    },
    {
      "epoch": 63.25,
      "learning_rate": 1.1566084142394823e-05,
      "loss": 0.1809,
      "step": 162880
    },
    {
      "epoch": 63.26,
      "learning_rate": 1.1565566343042072e-05,
      "loss": 0.3335,
      "step": 162890
    },
    {
      "epoch": 63.26,
      "learning_rate": 1.1565048543689322e-05,
      "loss": 0.0297,
      "step": 162900
    },
    {
      "epoch": 63.27,
      "learning_rate": 1.156453074433657e-05,
      "loss": 0.0537,
      "step": 162910
    },
    {
      "epoch": 63.27,
      "learning_rate": 1.156401294498382e-05,
      "loss": 0.0898,
      "step": 162920
    },
    {
      "epoch": 63.27,
      "learning_rate": 1.156349514563107e-05,
      "loss": 0.0523,
      "step": 162930
    },
    {
      "epoch": 63.28,
      "learning_rate": 1.1562977346278319e-05,
      "loss": 0.1789,
      "step": 162940
    },
    {
      "epoch": 63.28,
      "learning_rate": 1.1562459546925567e-05,
      "loss": 0.158,
      "step": 162950
    },
    {
      "epoch": 63.29,
      "learning_rate": 1.1561941747572816e-05,
      "loss": 0.2038,
      "step": 162960
    },
    {
      "epoch": 63.29,
      "learning_rate": 1.1561423948220066e-05,
      "loss": 0.1665,
      "step": 162970
    },
    {
      "epoch": 63.29,
      "learning_rate": 1.1560906148867314e-05,
      "loss": 0.2001,
      "step": 162980
    },
    {
      "epoch": 63.3,
      "learning_rate": 1.1560388349514563e-05,
      "loss": 0.034,
      "step": 162990
    },
    {
      "epoch": 63.3,
      "learning_rate": 1.1559870550161813e-05,
      "loss": 0.09,
      "step": 163000
    },
    {
      "epoch": 63.3,
      "learning_rate": 1.1559352750809063e-05,
      "loss": 0.0228,
      "step": 163010
    },
    {
      "epoch": 63.31,
      "learning_rate": 1.1558834951456312e-05,
      "loss": 0.0875,
      "step": 163020
    },
    {
      "epoch": 63.31,
      "learning_rate": 1.155831715210356e-05,
      "loss": 0.0498,
      "step": 163030
    },
    {
      "epoch": 63.32,
      "learning_rate": 1.155779935275081e-05,
      "loss": 0.1129,
      "step": 163040
    },
    {
      "epoch": 63.32,
      "learning_rate": 1.155728155339806e-05,
      "loss": 0.1949,
      "step": 163050
    },
    {
      "epoch": 63.32,
      "learning_rate": 1.1556763754045307e-05,
      "loss": 0.1952,
      "step": 163060
    },
    {
      "epoch": 63.33,
      "learning_rate": 1.1556245954692557e-05,
      "loss": 0.0191,
      "step": 163070
    },
    {
      "epoch": 63.33,
      "learning_rate": 1.1555728155339807e-05,
      "loss": 0.0619,
      "step": 163080
    },
    {
      "epoch": 63.34,
      "learning_rate": 1.1555210355987056e-05,
      "loss": 0.0632,
      "step": 163090
    },
    {
      "epoch": 63.34,
      "learning_rate": 1.1554692556634306e-05,
      "loss": 0.1438,
      "step": 163100
    },
    {
      "epoch": 63.34,
      "learning_rate": 1.1554174757281554e-05,
      "loss": 0.0732,
      "step": 163110
    },
    {
      "epoch": 63.35,
      "learning_rate": 1.1553656957928803e-05,
      "loss": 0.1305,
      "step": 163120
    },
    {
      "epoch": 63.35,
      "learning_rate": 1.1553139158576051e-05,
      "loss": 0.0217,
      "step": 163130
    },
    {
      "epoch": 63.36,
      "learning_rate": 1.1552621359223301e-05,
      "loss": 0.0279,
      "step": 163140
    },
    {
      "epoch": 63.36,
      "learning_rate": 1.155210355987055e-05,
      "loss": 0.0151,
      "step": 163150
    },
    {
      "epoch": 63.36,
      "learning_rate": 1.15515857605178e-05,
      "loss": 0.0268,
      "step": 163160
    },
    {
      "epoch": 63.37,
      "learning_rate": 1.155106796116505e-05,
      "loss": 0.0455,
      "step": 163170
    },
    {
      "epoch": 63.37,
      "learning_rate": 1.15505501618123e-05,
      "loss": 0.0212,
      "step": 163180
    },
    {
      "epoch": 63.37,
      "learning_rate": 1.1550032362459549e-05,
      "loss": 0.0888,
      "step": 163190
    },
    {
      "epoch": 63.38,
      "learning_rate": 1.1549514563106797e-05,
      "loss": 0.0219,
      "step": 163200
    },
    {
      "epoch": 63.38,
      "learning_rate": 1.1548996763754045e-05,
      "loss": 0.0525,
      "step": 163210
    },
    {
      "epoch": 63.39,
      "learning_rate": 1.1548478964401295e-05,
      "loss": 0.029,
      "step": 163220
    },
    {
      "epoch": 63.39,
      "learning_rate": 1.1547961165048544e-05,
      "loss": 0.0825,
      "step": 163230
    },
    {
      "epoch": 63.39,
      "learning_rate": 1.1547443365695794e-05,
      "loss": 0.0557,
      "step": 163240
    },
    {
      "epoch": 63.4,
      "learning_rate": 1.1546925566343043e-05,
      "loss": 0.0108,
      "step": 163250
    },
    {
      "epoch": 63.4,
      "learning_rate": 1.1546407766990293e-05,
      "loss": 0.0684,
      "step": 163260
    },
    {
      "epoch": 63.41,
      "learning_rate": 1.1545889967637543e-05,
      "loss": 0.1003,
      "step": 163270
    },
    {
      "epoch": 63.41,
      "learning_rate": 1.1545372168284789e-05,
      "loss": 0.0638,
      "step": 163280
    },
    {
      "epoch": 63.41,
      "learning_rate": 1.1544854368932038e-05,
      "loss": 0.1211,
      "step": 163290
    },
    {
      "epoch": 63.42,
      "learning_rate": 1.1544336569579288e-05,
      "loss": 0.096,
      "step": 163300
    },
    {
      "epoch": 63.42,
      "learning_rate": 1.1543818770226538e-05,
      "loss": 0.0424,
      "step": 163310
    },
    {
      "epoch": 63.43,
      "learning_rate": 1.1543300970873787e-05,
      "loss": 0.1519,
      "step": 163320
    },
    {
      "epoch": 63.43,
      "learning_rate": 1.1542783171521037e-05,
      "loss": 0.0011,
      "step": 163330
    },
    {
      "epoch": 63.43,
      "learning_rate": 1.1542265372168287e-05,
      "loss": 0.1671,
      "step": 163340
    },
    {
      "epoch": 63.44,
      "learning_rate": 1.1541747572815536e-05,
      "loss": 0.1156,
      "step": 163350
    },
    {
      "epoch": 63.44,
      "learning_rate": 1.1541229773462782e-05,
      "loss": 0.0263,
      "step": 163360
    },
    {
      "epoch": 63.44,
      "learning_rate": 1.1540711974110032e-05,
      "loss": 0.0694,
      "step": 163370
    },
    {
      "epoch": 63.45,
      "learning_rate": 1.1540194174757282e-05,
      "loss": 0.1245,
      "step": 163380
    },
    {
      "epoch": 63.45,
      "learning_rate": 1.1539676375404531e-05,
      "loss": 0.0511,
      "step": 163390
    },
    {
      "epoch": 63.46,
      "learning_rate": 1.1539158576051781e-05,
      "loss": 0.1245,
      "step": 163400
    },
    {
      "epoch": 63.46,
      "learning_rate": 1.153864077669903e-05,
      "loss": 0.0038,
      "step": 163410
    },
    {
      "epoch": 63.46,
      "learning_rate": 1.153812297734628e-05,
      "loss": 0.1716,
      "step": 163420
    },
    {
      "epoch": 63.47,
      "learning_rate": 1.153760517799353e-05,
      "loss": 0.0747,
      "step": 163430
    },
    {
      "epoch": 63.47,
      "learning_rate": 1.1537087378640776e-05,
      "loss": 0.3062,
      "step": 163440
    },
    {
      "epoch": 63.48,
      "learning_rate": 1.1536569579288026e-05,
      "loss": 0.1054,
      "step": 163450
    },
    {
      "epoch": 63.48,
      "learning_rate": 1.1536051779935275e-05,
      "loss": 0.054,
      "step": 163460
    },
    {
      "epoch": 63.48,
      "learning_rate": 1.1535533980582525e-05,
      "loss": 0.0402,
      "step": 163470
    },
    {
      "epoch": 63.49,
      "learning_rate": 1.1535016181229774e-05,
      "loss": 0.1091,
      "step": 163480
    },
    {
      "epoch": 63.49,
      "learning_rate": 1.1534498381877024e-05,
      "loss": 0.0744,
      "step": 163490
    },
    {
      "epoch": 63.5,
      "learning_rate": 1.1533980582524274e-05,
      "loss": 0.0292,
      "step": 163500
    },
    {
      "epoch": 63.5,
      "learning_rate": 1.1533462783171523e-05,
      "loss": 0.0344,
      "step": 163510
    },
    {
      "epoch": 63.5,
      "learning_rate": 1.153294498381877e-05,
      "loss": 0.0831,
      "step": 163520
    },
    {
      "epoch": 63.51,
      "learning_rate": 1.1532427184466019e-05,
      "loss": 0.0082,
      "step": 163530
    },
    {
      "epoch": 63.51,
      "learning_rate": 1.1531909385113269e-05,
      "loss": 0.0046,
      "step": 163540
    },
    {
      "epoch": 63.51,
      "learning_rate": 1.1531391585760518e-05,
      "loss": 0.0526,
      "step": 163550
    },
    {
      "epoch": 63.52,
      "learning_rate": 1.1530873786407768e-05,
      "loss": 0.1092,
      "step": 163560
    },
    {
      "epoch": 63.52,
      "learning_rate": 1.1530355987055018e-05,
      "loss": 0.0216,
      "step": 163570
    },
    {
      "epoch": 63.53,
      "learning_rate": 1.1529838187702267e-05,
      "loss": 0.0196,
      "step": 163580
    },
    {
      "epoch": 63.53,
      "learning_rate": 1.1529320388349517e-05,
      "loss": 0.0276,
      "step": 163590
    },
    {
      "epoch": 63.53,
      "learning_rate": 1.1528802588996763e-05,
      "loss": 0.0088,
      "step": 163600
    },
    {
      "epoch": 63.54,
      "learning_rate": 1.1528284789644013e-05,
      "loss": 0.038,
      "step": 163610
    },
    {
      "epoch": 63.54,
      "learning_rate": 1.1527766990291262e-05,
      "loss": 0.0187,
      "step": 163620
    },
    {
      "epoch": 63.55,
      "learning_rate": 1.1527249190938512e-05,
      "loss": 0.0719,
      "step": 163630
    },
    {
      "epoch": 63.55,
      "learning_rate": 1.1526731391585762e-05,
      "loss": 0.0495,
      "step": 163640
    },
    {
      "epoch": 63.55,
      "learning_rate": 1.1526213592233011e-05,
      "loss": 0.0252,
      "step": 163650
    },
    {
      "epoch": 63.56,
      "learning_rate": 1.152569579288026e-05,
      "loss": 0.037,
      "step": 163660
    },
    {
      "epoch": 63.56,
      "learning_rate": 1.152517799352751e-05,
      "loss": 0.0388,
      "step": 163670
    },
    {
      "epoch": 63.57,
      "learning_rate": 1.1524660194174757e-05,
      "loss": 0.0324,
      "step": 163680
    },
    {
      "epoch": 63.57,
      "learning_rate": 1.1524142394822006e-05,
      "loss": 0.0243,
      "step": 163690
    },
    {
      "epoch": 63.57,
      "learning_rate": 1.1523624595469256e-05,
      "loss": 0.0017,
      "step": 163700
    },
    {
      "epoch": 63.58,
      "learning_rate": 1.1523106796116505e-05,
      "loss": 0.1349,
      "step": 163710
    },
    {
      "epoch": 63.58,
      "learning_rate": 1.1522588996763755e-05,
      "loss": 0.0003,
      "step": 163720
    },
    {
      "epoch": 63.58,
      "learning_rate": 1.1522071197411005e-05,
      "loss": 0.0208,
      "step": 163730
    },
    {
      "epoch": 63.59,
      "learning_rate": 1.1521553398058254e-05,
      "loss": 0.0356,
      "step": 163740
    },
    {
      "epoch": 63.59,
      "learning_rate": 1.1521035598705504e-05,
      "loss": 0.0214,
      "step": 163750
    },
    {
      "epoch": 63.6,
      "learning_rate": 1.1520517799352754e-05,
      "loss": 0.0548,
      "step": 163760
    },
    {
      "epoch": 63.6,
      "learning_rate": 1.152e-05,
      "loss": 0.1039,
      "step": 163770
    },
    {
      "epoch": 63.6,
      "learning_rate": 1.151948220064725e-05,
      "loss": 0.0434,
      "step": 163780
    },
    {
      "epoch": 63.61,
      "learning_rate": 1.1518964401294499e-05,
      "loss": 0.1026,
      "step": 163790
    },
    {
      "epoch": 63.61,
      "learning_rate": 1.1518446601941749e-05,
      "loss": 0.129,
      "step": 163800
    },
    {
      "epoch": 63.62,
      "learning_rate": 1.1517928802588998e-05,
      "loss": 0.0426,
      "step": 163810
    },
    {
      "epoch": 63.62,
      "learning_rate": 1.1517411003236248e-05,
      "loss": 0.0095,
      "step": 163820
    },
    {
      "epoch": 63.62,
      "learning_rate": 1.1516893203883498e-05,
      "loss": 0.0461,
      "step": 163830
    },
    {
      "epoch": 63.63,
      "learning_rate": 1.1516375404530747e-05,
      "loss": 0.0248,
      "step": 163840
    },
    {
      "epoch": 63.63,
      "learning_rate": 1.1515857605177993e-05,
      "loss": 0.1225,
      "step": 163850
    },
    {
      "epoch": 63.63,
      "learning_rate": 1.1515339805825243e-05,
      "loss": 0.0301,
      "step": 163860
    },
    {
      "epoch": 63.64,
      "learning_rate": 1.1514822006472493e-05,
      "loss": 0.0736,
      "step": 163870
    },
    {
      "epoch": 63.64,
      "learning_rate": 1.1514304207119742e-05,
      "loss": 0.1236,
      "step": 163880
    },
    {
      "epoch": 63.65,
      "learning_rate": 1.1513786407766992e-05,
      "loss": 0.0182,
      "step": 163890
    },
    {
      "epoch": 63.65,
      "learning_rate": 1.1513268608414241e-05,
      "loss": 0.0398,
      "step": 163900
    },
    {
      "epoch": 63.65,
      "learning_rate": 1.1512750809061491e-05,
      "loss": 0.1578,
      "step": 163910
    },
    {
      "epoch": 63.66,
      "learning_rate": 1.1512233009708739e-05,
      "loss": 0.0342,
      "step": 163920
    },
    {
      "epoch": 63.66,
      "learning_rate": 1.1511715210355987e-05,
      "loss": 0.0011,
      "step": 163930
    },
    {
      "epoch": 63.67,
      "learning_rate": 1.1511197411003237e-05,
      "loss": 0.1475,
      "step": 163940
    },
    {
      "epoch": 63.67,
      "learning_rate": 1.1510679611650486e-05,
      "loss": 0.1054,
      "step": 163950
    },
    {
      "epoch": 63.67,
      "learning_rate": 1.1510161812297736e-05,
      "loss": 0.0015,
      "step": 163960
    },
    {
      "epoch": 63.68,
      "learning_rate": 1.1509644012944985e-05,
      "loss": 0.1701,
      "step": 163970
    },
    {
      "epoch": 63.68,
      "learning_rate": 1.1509126213592235e-05,
      "loss": 0.1073,
      "step": 163980
    },
    {
      "epoch": 63.69,
      "learning_rate": 1.1508608414239485e-05,
      "loss": 0.0335,
      "step": 163990
    },
    {
      "epoch": 63.69,
      "learning_rate": 1.1508090614886733e-05,
      "loss": 0.0332,
      "step": 164000
    },
    {
      "epoch": 63.69,
      "learning_rate": 1.150757281553398e-05,
      "loss": 0.1068,
      "step": 164010
    },
    {
      "epoch": 63.7,
      "learning_rate": 1.150705501618123e-05,
      "loss": 0.0242,
      "step": 164020
    },
    {
      "epoch": 63.7,
      "learning_rate": 1.150653721682848e-05,
      "loss": 0.0551,
      "step": 164030
    },
    {
      "epoch": 63.7,
      "learning_rate": 1.150601941747573e-05,
      "loss": 0.3086,
      "step": 164040
    },
    {
      "epoch": 63.71,
      "learning_rate": 1.1505501618122979e-05,
      "loss": 0.0616,
      "step": 164050
    },
    {
      "epoch": 63.71,
      "learning_rate": 1.1504983818770229e-05,
      "loss": 0.0706,
      "step": 164060
    },
    {
      "epoch": 63.72,
      "learning_rate": 1.1504466019417477e-05,
      "loss": 0.1485,
      "step": 164070
    },
    {
      "epoch": 63.72,
      "learning_rate": 1.1503948220064726e-05,
      "loss": 0.0486,
      "step": 164080
    },
    {
      "epoch": 63.72,
      "learning_rate": 1.1503430420711974e-05,
      "loss": 0.0479,
      "step": 164090
    },
    {
      "epoch": 63.73,
      "learning_rate": 1.1502912621359224e-05,
      "loss": 0.0578,
      "step": 164100
    },
    {
      "epoch": 63.73,
      "learning_rate": 1.1502394822006473e-05,
      "loss": 0.0027,
      "step": 164110
    },
    {
      "epoch": 63.74,
      "learning_rate": 1.1501877022653723e-05,
      "loss": 0.2392,
      "step": 164120
    },
    {
      "epoch": 63.74,
      "learning_rate": 1.1501359223300973e-05,
      "loss": 0.0635,
      "step": 164130
    },
    {
      "epoch": 63.74,
      "learning_rate": 1.1500841423948222e-05,
      "loss": 0.0821,
      "step": 164140
    },
    {
      "epoch": 63.75,
      "learning_rate": 1.150032362459547e-05,
      "loss": 0.1873,
      "step": 164150
    },
    {
      "epoch": 63.75,
      "learning_rate": 1.149980582524272e-05,
      "loss": 0.0426,
      "step": 164160
    },
    {
      "epoch": 63.76,
      "learning_rate": 1.1499288025889968e-05,
      "loss": 0.1033,
      "step": 164170
    },
    {
      "epoch": 63.76,
      "learning_rate": 1.1498770226537217e-05,
      "loss": 0.0594,
      "step": 164180
    },
    {
      "epoch": 63.76,
      "learning_rate": 1.1498252427184467e-05,
      "loss": 0.0253,
      "step": 164190
    },
    {
      "epoch": 63.77,
      "learning_rate": 1.1497734627831716e-05,
      "loss": 0.0751,
      "step": 164200
    },
    {
      "epoch": 63.77,
      "learning_rate": 1.1497216828478966e-05,
      "loss": 0.0528,
      "step": 164210
    },
    {
      "epoch": 63.77,
      "learning_rate": 1.1496699029126214e-05,
      "loss": 0.1324,
      "step": 164220
    },
    {
      "epoch": 63.78,
      "learning_rate": 1.1496181229773464e-05,
      "loss": 0.0314,
      "step": 164230
    },
    {
      "epoch": 63.78,
      "learning_rate": 1.1495663430420713e-05,
      "loss": 0.0185,
      "step": 164240
    },
    {
      "epoch": 63.79,
      "learning_rate": 1.1495145631067961e-05,
      "loss": 0.0002,
      "step": 164250
    },
    {
      "epoch": 63.79,
      "learning_rate": 1.149462783171521e-05,
      "loss": 0.086,
      "step": 164260
    },
    {
      "epoch": 63.79,
      "learning_rate": 1.149411003236246e-05,
      "loss": 0.0896,
      "step": 164270
    },
    {
      "epoch": 63.8,
      "learning_rate": 1.149359223300971e-05,
      "loss": 0.0094,
      "step": 164280
    },
    {
      "epoch": 63.8,
      "learning_rate": 1.149307443365696e-05,
      "loss": 0.0852,
      "step": 164290
    },
    {
      "epoch": 63.81,
      "learning_rate": 1.1492556634304208e-05,
      "loss": 0.0319,
      "step": 164300
    },
    {
      "epoch": 63.81,
      "learning_rate": 1.1492038834951457e-05,
      "loss": 0.0714,
      "step": 164310
    },
    {
      "epoch": 63.81,
      "learning_rate": 1.1491521035598707e-05,
      "loss": 0.0075,
      "step": 164320
    },
    {
      "epoch": 63.82,
      "learning_rate": 1.1491003236245956e-05,
      "loss": 0.0401,
      "step": 164330
    },
    {
      "epoch": 63.82,
      "learning_rate": 1.1490485436893204e-05,
      "loss": 0.2353,
      "step": 164340
    },
    {
      "epoch": 63.83,
      "learning_rate": 1.1489967637540454e-05,
      "loss": 0.0856,
      "step": 164350
    },
    {
      "epoch": 63.83,
      "learning_rate": 1.1489449838187704e-05,
      "loss": 0.0727,
      "step": 164360
    },
    {
      "epoch": 63.83,
      "learning_rate": 1.1488932038834953e-05,
      "loss": 0.0368,
      "step": 164370
    },
    {
      "epoch": 63.84,
      "learning_rate": 1.1488414239482201e-05,
      "loss": 0.0117,
      "step": 164380
    },
    {
      "epoch": 63.84,
      "learning_rate": 1.148789644012945e-05,
      "loss": 0.0058,
      "step": 164390
    },
    {
      "epoch": 63.84,
      "learning_rate": 1.14873786407767e-05,
      "loss": 0.0671,
      "step": 164400
    },
    {
      "epoch": 63.85,
      "learning_rate": 1.148686084142395e-05,
      "loss": 0.0198,
      "step": 164410
    },
    {
      "epoch": 63.85,
      "learning_rate": 1.1486343042071198e-05,
      "loss": 0.1179,
      "step": 164420
    },
    {
      "epoch": 63.86,
      "learning_rate": 1.1485825242718448e-05,
      "loss": 0.1665,
      "step": 164430
    },
    {
      "epoch": 63.86,
      "learning_rate": 1.1485307443365697e-05,
      "loss": 0.1671,
      "step": 164440
    },
    {
      "epoch": 63.86,
      "learning_rate": 1.1484789644012945e-05,
      "loss": 0.0015,
      "step": 164450
    },
    {
      "epoch": 63.87,
      "learning_rate": 1.1484271844660195e-05,
      "loss": 0.0706,
      "step": 164460
    },
    {
      "epoch": 63.87,
      "learning_rate": 1.1483754045307444e-05,
      "loss": 0.0075,
      "step": 164470
    },
    {
      "epoch": 63.88,
      "learning_rate": 1.1483236245954694e-05,
      "loss": 0.0392,
      "step": 164480
    },
    {
      "epoch": 63.88,
      "learning_rate": 1.1482718446601944e-05,
      "loss": 0.0317,
      "step": 164490
    },
    {
      "epoch": 63.88,
      "learning_rate": 1.1482200647249191e-05,
      "loss": 0.1332,
      "step": 164500
    },
    {
      "epoch": 63.89,
      "learning_rate": 1.1481682847896441e-05,
      "loss": 0.0305,
      "step": 164510
    },
    {
      "epoch": 63.89,
      "learning_rate": 1.148116504854369e-05,
      "loss": 0.0688,
      "step": 164520
    },
    {
      "epoch": 63.9,
      "learning_rate": 1.1480647249190939e-05,
      "loss": 0.0588,
      "step": 164530
    },
    {
      "epoch": 63.9,
      "learning_rate": 1.1480129449838188e-05,
      "loss": 0.0558,
      "step": 164540
    },
    {
      "epoch": 63.9,
      "learning_rate": 1.1479611650485438e-05,
      "loss": 0.1488,
      "step": 164550
    },
    {
      "epoch": 63.91,
      "learning_rate": 1.1479093851132687e-05,
      "loss": 0.1101,
      "step": 164560
    },
    {
      "epoch": 63.91,
      "learning_rate": 1.1478576051779937e-05,
      "loss": 0.0942,
      "step": 164570
    },
    {
      "epoch": 63.91,
      "learning_rate": 1.1478058252427185e-05,
      "loss": 0.0881,
      "step": 164580
    },
    {
      "epoch": 63.92,
      "learning_rate": 1.1477540453074435e-05,
      "loss": 0.0563,
      "step": 164590
    },
    {
      "epoch": 63.92,
      "learning_rate": 1.1477022653721683e-05,
      "loss": 0.1307,
      "step": 164600
    },
    {
      "epoch": 63.93,
      "learning_rate": 1.1476504854368932e-05,
      "loss": 0.05,
      "step": 164610
    },
    {
      "epoch": 63.93,
      "learning_rate": 1.1475987055016182e-05,
      "loss": 0.1104,
      "step": 164620
    },
    {
      "epoch": 63.93,
      "learning_rate": 1.1475469255663431e-05,
      "loss": 0.0195,
      "step": 164630
    },
    {
      "epoch": 63.94,
      "learning_rate": 1.1474951456310681e-05,
      "loss": 0.1457,
      "step": 164640
    },
    {
      "epoch": 63.94,
      "learning_rate": 1.147443365695793e-05,
      "loss": 0.1289,
      "step": 164650
    },
    {
      "epoch": 63.95,
      "learning_rate": 1.1473915857605179e-05,
      "loss": 0.056,
      "step": 164660
    },
    {
      "epoch": 63.95,
      "learning_rate": 1.1473398058252428e-05,
      "loss": 0.0269,
      "step": 164670
    },
    {
      "epoch": 63.95,
      "learning_rate": 1.1472880258899676e-05,
      "loss": 0.246,
      "step": 164680
    },
    {
      "epoch": 63.96,
      "learning_rate": 1.1472362459546926e-05,
      "loss": 0.0964,
      "step": 164690
    },
    {
      "epoch": 63.96,
      "learning_rate": 1.1471844660194175e-05,
      "loss": 0.1809,
      "step": 164700
    },
    {
      "epoch": 63.97,
      "learning_rate": 1.1471326860841425e-05,
      "loss": 0.0892,
      "step": 164710
    },
    {
      "epoch": 63.97,
      "learning_rate": 1.1470809061488675e-05,
      "loss": 0.1482,
      "step": 164720
    },
    {
      "epoch": 63.97,
      "learning_rate": 1.1470291262135924e-05,
      "loss": 0.0147,
      "step": 164730
    },
    {
      "epoch": 63.98,
      "learning_rate": 1.1469773462783172e-05,
      "loss": 0.1311,
      "step": 164740
    },
    {
      "epoch": 63.98,
      "learning_rate": 1.146925566343042e-05,
      "loss": 0.1914,
      "step": 164750
    },
    {
      "epoch": 63.98,
      "learning_rate": 1.146873786407767e-05,
      "loss": 0.0791,
      "step": 164760
    },
    {
      "epoch": 63.99,
      "learning_rate": 1.146822006472492e-05,
      "loss": 0.1097,
      "step": 164770
    },
    {
      "epoch": 63.99,
      "learning_rate": 1.1467702265372169e-05,
      "loss": 0.0787,
      "step": 164780
    },
    {
      "epoch": 64.0,
      "learning_rate": 1.1467184466019419e-05,
      "loss": 0.0004,
      "step": 164790
    },
    {
      "epoch": 64.0,
      "learning_rate": 1.1466666666666668e-05,
      "loss": 0.114,
      "step": 164800
    },
    {
      "epoch": 64.0,
      "eval_accuracy": 0.9488308115543329,
      "eval_loss": 0.3368075489997864,
      "eval_runtime": 8.2161,
      "eval_samples_per_second": 442.424,
      "eval_steps_per_second": 55.379,
      "step": 164800
    },
    {
      "epoch": 64.0,
      "learning_rate": 1.1466148867313918e-05,
      "loss": 0.0828,
      "step": 164810
    },
    {
      "epoch": 64.01,
      "learning_rate": 1.1465631067961167e-05,
      "loss": 0.0741,
      "step": 164820
    },
    {
      "epoch": 64.01,
      "learning_rate": 1.1465113268608414e-05,
      "loss": 0.012,
      "step": 164830
    },
    {
      "epoch": 64.02,
      "learning_rate": 1.1464595469255663e-05,
      "loss": 0.0484,
      "step": 164840
    },
    {
      "epoch": 64.02,
      "learning_rate": 1.1464077669902913e-05,
      "loss": 0.0615,
      "step": 164850
    },
    {
      "epoch": 64.02,
      "learning_rate": 1.1463559870550162e-05,
      "loss": 0.0826,
      "step": 164860
    },
    {
      "epoch": 64.03,
      "learning_rate": 1.1463042071197412e-05,
      "loss": 0.0693,
      "step": 164870
    },
    {
      "epoch": 64.03,
      "learning_rate": 1.1462524271844662e-05,
      "loss": 0.054,
      "step": 164880
    },
    {
      "epoch": 64.03,
      "learning_rate": 1.1462006472491911e-05,
      "loss": 0.0352,
      "step": 164890
    },
    {
      "epoch": 64.04,
      "learning_rate": 1.1461488673139161e-05,
      "loss": 0.115,
      "step": 164900
    },
    {
      "epoch": 64.04,
      "learning_rate": 1.1460970873786407e-05,
      "loss": 0.0208,
      "step": 164910
    },
    {
      "epoch": 64.05,
      "learning_rate": 1.1460453074433657e-05,
      "loss": 0.1814,
      "step": 164920
    },
    {
      "epoch": 64.05,
      "learning_rate": 1.1459935275080906e-05,
      "loss": 0.2075,
      "step": 164930
    },
    {
      "epoch": 64.05,
      "learning_rate": 1.1459417475728156e-05,
      "loss": 0.0606,
      "step": 164940
    },
    {
      "epoch": 64.06,
      "learning_rate": 1.1458899676375406e-05,
      "loss": 0.0448,
      "step": 164950
    },
    {
      "epoch": 64.06,
      "learning_rate": 1.1458381877022655e-05,
      "loss": 0.0169,
      "step": 164960
    },
    {
      "epoch": 64.07,
      "learning_rate": 1.1457864077669905e-05,
      "loss": 0.0686,
      "step": 164970
    },
    {
      "epoch": 64.07,
      "learning_rate": 1.1457346278317154e-05,
      "loss": 0.0356,
      "step": 164980
    },
    {
      "epoch": 64.07,
      "learning_rate": 1.14568284789644e-05,
      "loss": 0.0005,
      "step": 164990
    },
    {
      "epoch": 64.08,
      "learning_rate": 1.145631067961165e-05,
      "loss": 0.0287,
      "step": 165000
    },
    {
      "epoch": 64.08,
      "learning_rate": 1.14557928802589e-05,
      "loss": 0.0384,
      "step": 165010
    },
    {
      "epoch": 64.09,
      "learning_rate": 1.145527508090615e-05,
      "loss": 0.087,
      "step": 165020
    },
    {
      "epoch": 64.09,
      "learning_rate": 1.14547572815534e-05,
      "loss": 0.0227,
      "step": 165030
    },
    {
      "epoch": 64.09,
      "learning_rate": 1.1454239482200649e-05,
      "loss": 0.1596,
      "step": 165040
    },
    {
      "epoch": 64.1,
      "learning_rate": 1.1453721682847898e-05,
      "loss": 0.0483,
      "step": 165050
    },
    {
      "epoch": 64.1,
      "learning_rate": 1.1453203883495148e-05,
      "loss": 0.0002,
      "step": 165060
    },
    {
      "epoch": 64.1,
      "learning_rate": 1.1452686084142394e-05,
      "loss": 0.0002,
      "step": 165070
    },
    {
      "epoch": 64.11,
      "learning_rate": 1.1452168284789644e-05,
      "loss": 0.0439,
      "step": 165080
    },
    {
      "epoch": 64.11,
      "learning_rate": 1.1451650485436893e-05,
      "loss": 0.0172,
      "step": 165090
    },
    {
      "epoch": 64.12,
      "learning_rate": 1.1451132686084143e-05,
      "loss": 0.0175,
      "step": 165100
    },
    {
      "epoch": 64.12,
      "learning_rate": 1.1450614886731393e-05,
      "loss": 0.1803,
      "step": 165110
    },
    {
      "epoch": 64.12,
      "learning_rate": 1.1450097087378642e-05,
      "loss": 0.0463,
      "step": 165120
    },
    {
      "epoch": 64.13,
      "learning_rate": 1.1449579288025892e-05,
      "loss": 0.1294,
      "step": 165130
    },
    {
      "epoch": 64.13,
      "learning_rate": 1.1449061488673142e-05,
      "loss": 0.1067,
      "step": 165140
    },
    {
      "epoch": 64.14,
      "learning_rate": 1.1448543689320388e-05,
      "loss": 0.0637,
      "step": 165150
    },
    {
      "epoch": 64.14,
      "learning_rate": 1.1448025889967637e-05,
      "loss": 0.1315,
      "step": 165160
    },
    {
      "epoch": 64.14,
      "learning_rate": 1.1447508090614887e-05,
      "loss": 0.0193,
      "step": 165170
    },
    {
      "epoch": 64.15,
      "learning_rate": 1.1446990291262137e-05,
      "loss": 0.0972,
      "step": 165180
    },
    {
      "epoch": 64.15,
      "learning_rate": 1.1446472491909386e-05,
      "loss": 0.0452,
      "step": 165190
    },
    {
      "epoch": 64.16,
      "learning_rate": 1.1445954692556636e-05,
      "loss": 0.055,
      "step": 165200
    },
    {
      "epoch": 64.16,
      "learning_rate": 1.1445436893203886e-05,
      "loss": 0.1493,
      "step": 165210
    },
    {
      "epoch": 64.16,
      "learning_rate": 1.1444919093851135e-05,
      "loss": 0.0295,
      "step": 165220
    },
    {
      "epoch": 64.17,
      "learning_rate": 1.1444401294498381e-05,
      "loss": 0.0511,
      "step": 165230
    },
    {
      "epoch": 64.17,
      "learning_rate": 1.1443883495145631e-05,
      "loss": 0.2345,
      "step": 165240
    },
    {
      "epoch": 64.17,
      "learning_rate": 1.144336569579288e-05,
      "loss": 0.079,
      "step": 165250
    },
    {
      "epoch": 64.18,
      "learning_rate": 1.144284789644013e-05,
      "loss": 0.0353,
      "step": 165260
    },
    {
      "epoch": 64.18,
      "learning_rate": 1.144233009708738e-05,
      "loss": 0.0608,
      "step": 165270
    },
    {
      "epoch": 64.19,
      "learning_rate": 1.144181229773463e-05,
      "loss": 0.0979,
      "step": 165280
    },
    {
      "epoch": 64.19,
      "learning_rate": 1.1441294498381879e-05,
      "loss": 0.0617,
      "step": 165290
    },
    {
      "epoch": 64.19,
      "learning_rate": 1.1440776699029129e-05,
      "loss": 0.1291,
      "step": 165300
    },
    {
      "epoch": 64.2,
      "learning_rate": 1.1440258899676375e-05,
      "loss": 0.0107,
      "step": 165310
    },
    {
      "epoch": 64.2,
      "learning_rate": 1.1439741100323625e-05,
      "loss": 0.0563,
      "step": 165320
    },
    {
      "epoch": 64.21,
      "learning_rate": 1.1439223300970874e-05,
      "loss": 0.04,
      "step": 165330
    },
    {
      "epoch": 64.21,
      "learning_rate": 1.1438705501618124e-05,
      "loss": 0.0829,
      "step": 165340
    },
    {
      "epoch": 64.21,
      "learning_rate": 1.1438187702265373e-05,
      "loss": 0.01,
      "step": 165350
    },
    {
      "epoch": 64.22,
      "learning_rate": 1.1437669902912623e-05,
      "loss": 0.0673,
      "step": 165360
    },
    {
      "epoch": 64.22,
      "learning_rate": 1.1437152103559873e-05,
      "loss": 0.0507,
      "step": 165370
    },
    {
      "epoch": 64.23,
      "learning_rate": 1.1436634304207122e-05,
      "loss": 0.0076,
      "step": 165380
    },
    {
      "epoch": 64.23,
      "learning_rate": 1.143611650485437e-05,
      "loss": 0.1252,
      "step": 165390
    },
    {
      "epoch": 64.23,
      "learning_rate": 1.1435598705501618e-05,
      "loss": 0.1142,
      "step": 165400
    },
    {
      "epoch": 64.24,
      "learning_rate": 1.1435080906148868e-05,
      "loss": 0.0003,
      "step": 165410
    },
    {
      "epoch": 64.24,
      "learning_rate": 1.1434563106796117e-05,
      "loss": 0.1224,
      "step": 165420
    },
    {
      "epoch": 64.24,
      "learning_rate": 1.1434045307443367e-05,
      "loss": 0.0138,
      "step": 165430
    },
    {
      "epoch": 64.25,
      "learning_rate": 1.1433527508090617e-05,
      "loss": 0.0143,
      "step": 165440
    },
    {
      "epoch": 64.25,
      "learning_rate": 1.1433009708737866e-05,
      "loss": 0.0583,
      "step": 165450
    },
    {
      "epoch": 64.26,
      "learning_rate": 1.1432491909385116e-05,
      "loss": 0.0723,
      "step": 165460
    },
    {
      "epoch": 64.26,
      "learning_rate": 1.1431974110032364e-05,
      "loss": 0.0148,
      "step": 165470
    },
    {
      "epoch": 64.26,
      "learning_rate": 1.1431456310679612e-05,
      "loss": 0.0609,
      "step": 165480
    },
    {
      "epoch": 64.27,
      "learning_rate": 1.1430938511326861e-05,
      "loss": 0.0139,
      "step": 165490
    },
    {
      "epoch": 64.27,
      "learning_rate": 1.1430420711974111e-05,
      "loss": 0.1054,
      "step": 165500
    },
    {
      "epoch": 64.28,
      "learning_rate": 1.142990291262136e-05,
      "loss": 0.0378,
      "step": 165510
    },
    {
      "epoch": 64.28,
      "learning_rate": 1.142938511326861e-05,
      "loss": 0.0131,
      "step": 165520
    },
    {
      "epoch": 64.28,
      "learning_rate": 1.142886731391586e-05,
      "loss": 0.0002,
      "step": 165530
    },
    {
      "epoch": 64.29,
      "learning_rate": 1.1428349514563108e-05,
      "loss": 0.1214,
      "step": 165540
    },
    {
      "epoch": 64.29,
      "learning_rate": 1.1427831715210357e-05,
      "loss": 0.1354,
      "step": 165550
    },
    {
      "epoch": 64.3,
      "learning_rate": 1.1427313915857605e-05,
      "loss": 0.1821,
      "step": 165560
    },
    {
      "epoch": 64.3,
      "learning_rate": 1.1426796116504855e-05,
      "loss": 0.038,
      "step": 165570
    },
    {
      "epoch": 64.3,
      "learning_rate": 1.1426278317152104e-05,
      "loss": 0.0587,
      "step": 165580
    },
    {
      "epoch": 64.31,
      "learning_rate": 1.1425760517799354e-05,
      "loss": 0.0083,
      "step": 165590
    },
    {
      "epoch": 64.31,
      "learning_rate": 1.1425242718446604e-05,
      "loss": 0.0288,
      "step": 165600
    },
    {
      "epoch": 64.31,
      "learning_rate": 1.1424724919093853e-05,
      "loss": 0.0261,
      "step": 165610
    },
    {
      "epoch": 64.32,
      "learning_rate": 1.1424207119741101e-05,
      "loss": 0.0451,
      "step": 165620
    },
    {
      "epoch": 64.32,
      "learning_rate": 1.142368932038835e-05,
      "loss": 0.0038,
      "step": 165630
    },
    {
      "epoch": 64.33,
      "learning_rate": 1.1423171521035599e-05,
      "loss": 0.1513,
      "step": 165640
    },
    {
      "epoch": 64.33,
      "learning_rate": 1.1422653721682848e-05,
      "loss": 0.1076,
      "step": 165650
    },
    {
      "epoch": 64.33,
      "learning_rate": 1.1422135922330098e-05,
      "loss": 0.0071,
      "step": 165660
    },
    {
      "epoch": 64.34,
      "learning_rate": 1.1421618122977348e-05,
      "loss": 0.0175,
      "step": 165670
    },
    {
      "epoch": 64.34,
      "learning_rate": 1.1421100323624597e-05,
      "loss": 0.0004,
      "step": 165680
    },
    {
      "epoch": 64.35,
      "learning_rate": 1.1420582524271845e-05,
      "loss": 0.0067,
      "step": 165690
    },
    {
      "epoch": 64.35,
      "learning_rate": 1.1420064724919095e-05,
      "loss": 0.0915,
      "step": 165700
    },
    {
      "epoch": 64.35,
      "learning_rate": 1.1419546925566344e-05,
      "loss": 0.0018,
      "step": 165710
    },
    {
      "epoch": 64.36,
      "learning_rate": 1.1419029126213592e-05,
      "loss": 0.1316,
      "step": 165720
    },
    {
      "epoch": 64.36,
      "learning_rate": 1.1418511326860842e-05,
      "loss": 0.1427,
      "step": 165730
    },
    {
      "epoch": 64.37,
      "learning_rate": 1.1417993527508092e-05,
      "loss": 0.0443,
      "step": 165740
    },
    {
      "epoch": 64.37,
      "learning_rate": 1.1417475728155341e-05,
      "loss": 0.0944,
      "step": 165750
    },
    {
      "epoch": 64.37,
      "learning_rate": 1.141695792880259e-05,
      "loss": 0.0527,
      "step": 165760
    },
    {
      "epoch": 64.38,
      "learning_rate": 1.1416440129449839e-05,
      "loss": 0.0845,
      "step": 165770
    },
    {
      "epoch": 64.38,
      "learning_rate": 1.1415922330097088e-05,
      "loss": 0.1426,
      "step": 165780
    },
    {
      "epoch": 64.38,
      "learning_rate": 1.1415404530744338e-05,
      "loss": 0.0723,
      "step": 165790
    },
    {
      "epoch": 64.39,
      "learning_rate": 1.1414886731391586e-05,
      "loss": 0.0658,
      "step": 165800
    },
    {
      "epoch": 64.39,
      "learning_rate": 1.1414368932038836e-05,
      "loss": 0.0018,
      "step": 165810
    },
    {
      "epoch": 64.4,
      "learning_rate": 1.1413851132686085e-05,
      "loss": 0.0397,
      "step": 165820
    },
    {
      "epoch": 64.4,
      "learning_rate": 1.1413333333333335e-05,
      "loss": 0.0643,
      "step": 165830
    },
    {
      "epoch": 64.4,
      "learning_rate": 1.1412815533980583e-05,
      "loss": 0.1348,
      "step": 165840
    },
    {
      "epoch": 64.41,
      "learning_rate": 1.1412297734627832e-05,
      "loss": 0.1665,
      "step": 165850
    },
    {
      "epoch": 64.41,
      "learning_rate": 1.1411779935275082e-05,
      "loss": 0.1429,
      "step": 165860
    },
    {
      "epoch": 64.42,
      "learning_rate": 1.1411262135922332e-05,
      "loss": 0.0631,
      "step": 165870
    },
    {
      "epoch": 64.42,
      "learning_rate": 1.141074433656958e-05,
      "loss": 0.0701,
      "step": 165880
    },
    {
      "epoch": 64.42,
      "learning_rate": 1.1410226537216829e-05,
      "loss": 0.1045,
      "step": 165890
    },
    {
      "epoch": 64.43,
      "learning_rate": 1.1409708737864079e-05,
      "loss": 0.0402,
      "step": 165900
    },
    {
      "epoch": 64.43,
      "learning_rate": 1.1409190938511328e-05,
      "loss": 0.0795,
      "step": 165910
    },
    {
      "epoch": 64.43,
      "learning_rate": 1.1408673139158576e-05,
      "loss": 0.2038,
      "step": 165920
    },
    {
      "epoch": 64.44,
      "learning_rate": 1.1408155339805826e-05,
      "loss": 0.0314,
      "step": 165930
    },
    {
      "epoch": 64.44,
      "learning_rate": 1.1407637540453075e-05,
      "loss": 0.0439,
      "step": 165940
    },
    {
      "epoch": 64.45,
      "learning_rate": 1.1407119741100325e-05,
      "loss": 0.001,
      "step": 165950
    },
    {
      "epoch": 64.45,
      "learning_rate": 1.1406601941747575e-05,
      "loss": 0.0451,
      "step": 165960
    },
    {
      "epoch": 64.45,
      "learning_rate": 1.1406084142394823e-05,
      "loss": 0.0467,
      "step": 165970
    },
    {
      "epoch": 64.46,
      "learning_rate": 1.1405566343042072e-05,
      "loss": 0.1373,
      "step": 165980
    },
    {
      "epoch": 64.46,
      "learning_rate": 1.1405048543689322e-05,
      "loss": 0.0186,
      "step": 165990
    },
    {
      "epoch": 64.47,
      "learning_rate": 1.140453074433657e-05,
      "loss": 0.1583,
      "step": 166000
    },
    {
      "epoch": 64.47,
      "learning_rate": 1.140401294498382e-05,
      "loss": 0.1107,
      "step": 166010
    },
    {
      "epoch": 64.47,
      "learning_rate": 1.1403495145631069e-05,
      "loss": 0.0367,
      "step": 166020
    },
    {
      "epoch": 64.48,
      "learning_rate": 1.1402977346278319e-05,
      "loss": 0.0342,
      "step": 166030
    },
    {
      "epoch": 64.48,
      "learning_rate": 1.1402459546925568e-05,
      "loss": 0.0225,
      "step": 166040
    },
    {
      "epoch": 64.49,
      "learning_rate": 1.1401941747572816e-05,
      "loss": 0.0561,
      "step": 166050
    },
    {
      "epoch": 64.49,
      "learning_rate": 1.1401423948220066e-05,
      "loss": 0.0461,
      "step": 166060
    },
    {
      "epoch": 64.49,
      "learning_rate": 1.1400906148867314e-05,
      "loss": 0.0572,
      "step": 166070
    },
    {
      "epoch": 64.5,
      "learning_rate": 1.1400388349514563e-05,
      "loss": 0.0205,
      "step": 166080
    },
    {
      "epoch": 64.5,
      "learning_rate": 1.1399870550161813e-05,
      "loss": 0.0258,
      "step": 166090
    },
    {
      "epoch": 64.5,
      "learning_rate": 1.1399352750809063e-05,
      "loss": 0.0257,
      "step": 166100
    },
    {
      "epoch": 64.51,
      "learning_rate": 1.1398834951456312e-05,
      "loss": 0.0105,
      "step": 166110
    },
    {
      "epoch": 64.51,
      "learning_rate": 1.1398317152103562e-05,
      "loss": 0.0554,
      "step": 166120
    },
    {
      "epoch": 64.52,
      "learning_rate": 1.139779935275081e-05,
      "loss": 0.0413,
      "step": 166130
    },
    {
      "epoch": 64.52,
      "learning_rate": 1.139728155339806e-05,
      "loss": 0.0516,
      "step": 166140
    },
    {
      "epoch": 64.52,
      "learning_rate": 1.1396763754045307e-05,
      "loss": 0.1315,
      "step": 166150
    },
    {
      "epoch": 64.53,
      "learning_rate": 1.1396245954692557e-05,
      "loss": 0.0008,
      "step": 166160
    },
    {
      "epoch": 64.53,
      "learning_rate": 1.1395728155339807e-05,
      "loss": 0.0623,
      "step": 166170
    },
    {
      "epoch": 64.54,
      "learning_rate": 1.1395210355987056e-05,
      "loss": 0.1403,
      "step": 166180
    },
    {
      "epoch": 64.54,
      "learning_rate": 1.1394692556634306e-05,
      "loss": 0.1247,
      "step": 166190
    },
    {
      "epoch": 64.54,
      "learning_rate": 1.1394174757281555e-05,
      "loss": 0.1378,
      "step": 166200
    },
    {
      "epoch": 64.55,
      "learning_rate": 1.1393656957928803e-05,
      "loss": 0.048,
      "step": 166210
    },
    {
      "epoch": 64.55,
      "learning_rate": 1.1393139158576051e-05,
      "loss": 0.1123,
      "step": 166220
    },
    {
      "epoch": 64.56,
      "learning_rate": 1.13926213592233e-05,
      "loss": 0.0461,
      "step": 166230
    },
    {
      "epoch": 64.56,
      "learning_rate": 1.139210355987055e-05,
      "loss": 0.072,
      "step": 166240
    },
    {
      "epoch": 64.56,
      "learning_rate": 1.13915857605178e-05,
      "loss": 0.1131,
      "step": 166250
    },
    {
      "epoch": 64.57,
      "learning_rate": 1.139106796116505e-05,
      "loss": 0.0385,
      "step": 166260
    },
    {
      "epoch": 64.57,
      "learning_rate": 1.13905501618123e-05,
      "loss": 0.0569,
      "step": 166270
    },
    {
      "epoch": 64.57,
      "learning_rate": 1.1390032362459549e-05,
      "loss": 0.0725,
      "step": 166280
    },
    {
      "epoch": 64.58,
      "learning_rate": 1.1389514563106797e-05,
      "loss": 0.0872,
      "step": 166290
    },
    {
      "epoch": 64.58,
      "learning_rate": 1.1388996763754045e-05,
      "loss": 0.0989,
      "step": 166300
    },
    {
      "epoch": 64.59,
      "learning_rate": 1.1388478964401294e-05,
      "loss": 0.0565,
      "step": 166310
    },
    {
      "epoch": 64.59,
      "learning_rate": 1.1387961165048544e-05,
      "loss": 0.0733,
      "step": 166320
    },
    {
      "epoch": 64.59,
      "learning_rate": 1.1387443365695794e-05,
      "loss": 0.1651,
      "step": 166330
    },
    {
      "epoch": 64.6,
      "learning_rate": 1.1386925566343043e-05,
      "loss": 0.1265,
      "step": 166340
    },
    {
      "epoch": 64.6,
      "learning_rate": 1.1386407766990293e-05,
      "loss": 0.058,
      "step": 166350
    },
    {
      "epoch": 64.61,
      "learning_rate": 1.1385889967637542e-05,
      "loss": 0.0014,
      "step": 166360
    },
    {
      "epoch": 64.61,
      "learning_rate": 1.1385372168284789e-05,
      "loss": 0.0742,
      "step": 166370
    },
    {
      "epoch": 64.61,
      "learning_rate": 1.1384854368932038e-05,
      "loss": 0.0413,
      "step": 166380
    },
    {
      "epoch": 64.62,
      "learning_rate": 1.1384336569579288e-05,
      "loss": 0.0889,
      "step": 166390
    },
    {
      "epoch": 64.62,
      "learning_rate": 1.1383818770226538e-05,
      "loss": 0.0754,
      "step": 166400
    },
    {
      "epoch": 64.63,
      "learning_rate": 1.1383300970873787e-05,
      "loss": 0.0931,
      "step": 166410
    },
    {
      "epoch": 64.63,
      "learning_rate": 1.1382783171521037e-05,
      "loss": 0.0624,
      "step": 166420
    },
    {
      "epoch": 64.63,
      "learning_rate": 1.1382265372168286e-05,
      "loss": 0.0697,
      "step": 166430
    },
    {
      "epoch": 64.64,
      "learning_rate": 1.1381747572815536e-05,
      "loss": 0.1209,
      "step": 166440
    },
    {
      "epoch": 64.64,
      "learning_rate": 1.1381229773462782e-05,
      "loss": 0.0231,
      "step": 166450
    },
    {
      "epoch": 64.64,
      "learning_rate": 1.1380711974110032e-05,
      "loss": 0.0667,
      "step": 166460
    },
    {
      "epoch": 64.65,
      "learning_rate": 1.1380194174757281e-05,
      "loss": 0.1146,
      "step": 166470
    },
    {
      "epoch": 64.65,
      "learning_rate": 1.1379676375404531e-05,
      "loss": 0.1356,
      "step": 166480
    },
    {
      "epoch": 64.66,
      "learning_rate": 1.137915857605178e-05,
      "loss": 0.1053,
      "step": 166490
    },
    {
      "epoch": 64.66,
      "learning_rate": 1.137864077669903e-05,
      "loss": 0.0141,
      "step": 166500
    },
    {
      "epoch": 64.66,
      "learning_rate": 1.137812297734628e-05,
      "loss": 0.1087,
      "step": 166510
    },
    {
      "epoch": 64.67,
      "learning_rate": 1.137760517799353e-05,
      "loss": 0.0185,
      "step": 166520
    },
    {
      "epoch": 64.67,
      "learning_rate": 1.137708737864078e-05,
      "loss": 0.0103,
      "step": 166530
    },
    {
      "epoch": 64.68,
      "learning_rate": 1.1376569579288025e-05,
      "loss": 0.0034,
      "step": 166540
    },
    {
      "epoch": 64.68,
      "learning_rate": 1.1376051779935275e-05,
      "loss": 0.0585,
      "step": 166550
    },
    {
      "epoch": 64.68,
      "learning_rate": 1.1375533980582525e-05,
      "loss": 0.1668,
      "step": 166560
    },
    {
      "epoch": 64.69,
      "learning_rate": 1.1375016181229774e-05,
      "loss": 0.0128,
      "step": 166570
    },
    {
      "epoch": 64.69,
      "learning_rate": 1.1374498381877024e-05,
      "loss": 0.1597,
      "step": 166580
    },
    {
      "epoch": 64.7,
      "learning_rate": 1.1373980582524274e-05,
      "loss": 0.072,
      "step": 166590
    },
    {
      "epoch": 64.7,
      "learning_rate": 1.1373462783171523e-05,
      "loss": 0.0415,
      "step": 166600
    },
    {
      "epoch": 64.7,
      "learning_rate": 1.1372944983818773e-05,
      "loss": 0.052,
      "step": 166610
    },
    {
      "epoch": 64.71,
      "learning_rate": 1.1372427184466019e-05,
      "loss": 0.1109,
      "step": 166620
    },
    {
      "epoch": 64.71,
      "learning_rate": 1.1371909385113269e-05,
      "loss": 0.1262,
      "step": 166630
    },
    {
      "epoch": 64.71,
      "learning_rate": 1.1371391585760518e-05,
      "loss": 0.1002,
      "step": 166640
    },
    {
      "epoch": 64.72,
      "learning_rate": 1.1370873786407768e-05,
      "loss": 0.1112,
      "step": 166650
    },
    {
      "epoch": 64.72,
      "learning_rate": 1.1370355987055017e-05,
      "loss": 0.0005,
      "step": 166660
    },
    {
      "epoch": 64.73,
      "learning_rate": 1.1369838187702267e-05,
      "loss": 0.1112,
      "step": 166670
    },
    {
      "epoch": 64.73,
      "learning_rate": 1.1369320388349517e-05,
      "loss": 0.1059,
      "step": 166680
    },
    {
      "epoch": 64.73,
      "learning_rate": 1.1368802588996766e-05,
      "loss": 0.0232,
      "step": 166690
    },
    {
      "epoch": 64.74,
      "learning_rate": 1.1368284789644013e-05,
      "loss": 0.0167,
      "step": 166700
    },
    {
      "epoch": 64.74,
      "learning_rate": 1.1367766990291262e-05,
      "loss": 0.0604,
      "step": 166710
    },
    {
      "epoch": 64.75,
      "learning_rate": 1.1367249190938512e-05,
      "loss": 0.0724,
      "step": 166720
    },
    {
      "epoch": 64.75,
      "learning_rate": 1.1366731391585761e-05,
      "loss": 0.0007,
      "step": 166730
    },
    {
      "epoch": 64.75,
      "learning_rate": 1.1366213592233011e-05,
      "loss": 0.038,
      "step": 166740
    },
    {
      "epoch": 64.76,
      "learning_rate": 1.136569579288026e-05,
      "loss": 0.007,
      "step": 166750
    },
    {
      "epoch": 64.76,
      "learning_rate": 1.136517799352751e-05,
      "loss": 0.0366,
      "step": 166760
    },
    {
      "epoch": 64.77,
      "learning_rate": 1.136466019417476e-05,
      "loss": 0.0085,
      "step": 166770
    },
    {
      "epoch": 64.77,
      "learning_rate": 1.1364142394822006e-05,
      "loss": 0.0601,
      "step": 166780
    },
    {
      "epoch": 64.77,
      "learning_rate": 1.1363624595469256e-05,
      "loss": 0.1358,
      "step": 166790
    },
    {
      "epoch": 64.78,
      "learning_rate": 1.1363106796116505e-05,
      "loss": 0.1318,
      "step": 166800
    },
    {
      "epoch": 64.78,
      "learning_rate": 1.1362588996763755e-05,
      "loss": 0.0495,
      "step": 166810
    },
    {
      "epoch": 64.78,
      "learning_rate": 1.1362071197411005e-05,
      "loss": 0.0338,
      "step": 166820
    },
    {
      "epoch": 64.79,
      "learning_rate": 1.1361553398058254e-05,
      "loss": 0.065,
      "step": 166830
    },
    {
      "epoch": 64.79,
      "learning_rate": 1.1361035598705504e-05,
      "loss": 0.1135,
      "step": 166840
    },
    {
      "epoch": 64.8,
      "learning_rate": 1.1360517799352753e-05,
      "loss": 0.0287,
      "step": 166850
    },
    {
      "epoch": 64.8,
      "learning_rate": 1.136e-05,
      "loss": 0.0541,
      "step": 166860
    },
    {
      "epoch": 64.8,
      "learning_rate": 1.135948220064725e-05,
      "loss": 0.0745,
      "step": 166870
    },
    {
      "epoch": 64.81,
      "learning_rate": 1.1358964401294499e-05,
      "loss": 0.032,
      "step": 166880
    },
    {
      "epoch": 64.81,
      "learning_rate": 1.1358446601941749e-05,
      "loss": 0.0895,
      "step": 166890
    },
    {
      "epoch": 64.82,
      "learning_rate": 1.1357928802588998e-05,
      "loss": 0.0477,
      "step": 166900
    },
    {
      "epoch": 64.82,
      "learning_rate": 1.1357411003236248e-05,
      "loss": 0.0468,
      "step": 166910
    },
    {
      "epoch": 64.82,
      "learning_rate": 1.1356893203883497e-05,
      "loss": 0.0288,
      "step": 166920
    },
    {
      "epoch": 64.83,
      "learning_rate": 1.1356375404530747e-05,
      "loss": 0.0034,
      "step": 166930
    },
    {
      "epoch": 64.83,
      "learning_rate": 1.1355857605177993e-05,
      "loss": 0.1734,
      "step": 166940
    },
    {
      "epoch": 64.83,
      "learning_rate": 1.1355339805825243e-05,
      "loss": 0.1708,
      "step": 166950
    },
    {
      "epoch": 64.84,
      "learning_rate": 1.1354822006472492e-05,
      "loss": 0.0479,
      "step": 166960
    },
    {
      "epoch": 64.84,
      "learning_rate": 1.1354304207119742e-05,
      "loss": 0.0997,
      "step": 166970
    },
    {
      "epoch": 64.85,
      "learning_rate": 1.1353786407766992e-05,
      "loss": 0.1767,
      "step": 166980
    },
    {
      "epoch": 64.85,
      "learning_rate": 1.1353268608414241e-05,
      "loss": 0.0436,
      "step": 166990
    },
    {
      "epoch": 64.85,
      "learning_rate": 1.1352750809061491e-05,
      "loss": 0.104,
      "step": 167000
    },
    {
      "epoch": 64.86,
      "learning_rate": 1.1352233009708739e-05,
      "loss": 0.0016,
      "step": 167010
    },
    {
      "epoch": 64.86,
      "learning_rate": 1.1351715210355987e-05,
      "loss": 0.0211,
      "step": 167020
    },
    {
      "epoch": 64.87,
      "learning_rate": 1.1351197411003236e-05,
      "loss": 0.1665,
      "step": 167030
    },
    {
      "epoch": 64.87,
      "learning_rate": 1.1350679611650486e-05,
      "loss": 0.1075,
      "step": 167040
    },
    {
      "epoch": 64.87,
      "learning_rate": 1.1350161812297736e-05,
      "loss": 0.067,
      "step": 167050
    },
    {
      "epoch": 64.88,
      "learning_rate": 1.1349644012944985e-05,
      "loss": 0.0363,
      "step": 167060
    },
    {
      "epoch": 64.88,
      "learning_rate": 1.1349126213592235e-05,
      "loss": 0.006,
      "step": 167070
    },
    {
      "epoch": 64.89,
      "learning_rate": 1.1348608414239484e-05,
      "loss": 0.0737,
      "step": 167080
    },
    {
      "epoch": 64.89,
      "learning_rate": 1.1348090614886732e-05,
      "loss": 0.2195,
      "step": 167090
    },
    {
      "epoch": 64.89,
      "learning_rate": 1.1347572815533982e-05,
      "loss": 0.0128,
      "step": 167100
    },
    {
      "epoch": 64.9,
      "learning_rate": 1.134705501618123e-05,
      "loss": 0.1553,
      "step": 167110
    },
    {
      "epoch": 64.9,
      "learning_rate": 1.134653721682848e-05,
      "loss": 0.03,
      "step": 167120
    },
    {
      "epoch": 64.9,
      "learning_rate": 1.134601941747573e-05,
      "loss": 0.1351,
      "step": 167130
    },
    {
      "epoch": 64.91,
      "learning_rate": 1.1345501618122979e-05,
      "loss": 0.0308,
      "step": 167140
    },
    {
      "epoch": 64.91,
      "learning_rate": 1.1344983818770228e-05,
      "loss": 0.1303,
      "step": 167150
    },
    {
      "epoch": 64.92,
      "learning_rate": 1.1344466019417476e-05,
      "loss": 0.1481,
      "step": 167160
    },
    {
      "epoch": 64.92,
      "learning_rate": 1.1343948220064726e-05,
      "loss": 0.0676,
      "step": 167170
    },
    {
      "epoch": 64.92,
      "learning_rate": 1.1343430420711976e-05,
      "loss": 0.1199,
      "step": 167180
    },
    {
      "epoch": 64.93,
      "learning_rate": 1.1342912621359223e-05,
      "loss": 0.1081,
      "step": 167190
    },
    {
      "epoch": 64.93,
      "learning_rate": 1.1342394822006473e-05,
      "loss": 0.0298,
      "step": 167200
    },
    {
      "epoch": 64.94,
      "learning_rate": 1.1341877022653723e-05,
      "loss": 0.0627,
      "step": 167210
    },
    {
      "epoch": 64.94,
      "learning_rate": 1.1341359223300972e-05,
      "loss": 0.0153,
      "step": 167220
    },
    {
      "epoch": 64.94,
      "learning_rate": 1.1340841423948222e-05,
      "loss": 0.1145,
      "step": 167230
    },
    {
      "epoch": 64.95,
      "learning_rate": 1.134032362459547e-05,
      "loss": 0.0292,
      "step": 167240
    },
    {
      "epoch": 64.95,
      "learning_rate": 1.133980582524272e-05,
      "loss": 0.0319,
      "step": 167250
    },
    {
      "epoch": 64.96,
      "learning_rate": 1.1339288025889969e-05,
      "loss": 0.0475,
      "step": 167260
    },
    {
      "epoch": 64.96,
      "learning_rate": 1.1338770226537217e-05,
      "loss": 0.0255,
      "step": 167270
    },
    {
      "epoch": 64.96,
      "learning_rate": 1.1338252427184467e-05,
      "loss": 0.0354,
      "step": 167280
    },
    {
      "epoch": 64.97,
      "learning_rate": 1.1337734627831716e-05,
      "loss": 0.0615,
      "step": 167290
    },
    {
      "epoch": 64.97,
      "learning_rate": 1.1337216828478966e-05,
      "loss": 0.0663,
      "step": 167300
    },
    {
      "epoch": 64.97,
      "learning_rate": 1.1336699029126214e-05,
      "loss": 0.0246,
      "step": 167310
    },
    {
      "epoch": 64.98,
      "learning_rate": 1.1336181229773463e-05,
      "loss": 0.008,
      "step": 167320
    },
    {
      "epoch": 64.98,
      "learning_rate": 1.1335663430420713e-05,
      "loss": 0.1021,
      "step": 167330
    },
    {
      "epoch": 64.99,
      "learning_rate": 1.1335145631067963e-05,
      "loss": 0.0746,
      "step": 167340
    },
    {
      "epoch": 64.99,
      "learning_rate": 1.133462783171521e-05,
      "loss": 0.0203,
      "step": 167350
    },
    {
      "epoch": 64.99,
      "learning_rate": 1.133411003236246e-05,
      "loss": 0.0834,
      "step": 167360
    },
    {
      "epoch": 65.0,
      "learning_rate": 1.133359223300971e-05,
      "loss": 0.048,
      "step": 167370
    },
    {
      "epoch": 65.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.335758239030838,
      "eval_runtime": 8.2116,
      "eval_samples_per_second": 442.667,
      "eval_steps_per_second": 55.41,
      "step": 167375
    },
    {
      "epoch": 65.0,
      "learning_rate": 1.133307443365696e-05,
      "loss": 0.0573,
      "step": 167380
    },
    {
      "epoch": 65.01,
      "learning_rate": 1.1332556634304207e-05,
      "loss": 0.0145,
      "step": 167390
    },
    {
      "epoch": 65.01,
      "learning_rate": 1.1332038834951457e-05,
      "loss": 0.0201,
      "step": 167400
    },
    {
      "epoch": 65.01,
      "learning_rate": 1.1331521035598707e-05,
      "loss": 0.0482,
      "step": 167410
    },
    {
      "epoch": 65.02,
      "learning_rate": 1.1331003236245956e-05,
      "loss": 0.1077,
      "step": 167420
    },
    {
      "epoch": 65.02,
      "learning_rate": 1.1330485436893204e-05,
      "loss": 0.0676,
      "step": 167430
    },
    {
      "epoch": 65.03,
      "learning_rate": 1.1329967637540454e-05,
      "loss": 0.1153,
      "step": 167440
    },
    {
      "epoch": 65.03,
      "learning_rate": 1.1329449838187703e-05,
      "loss": 0.2046,
      "step": 167450
    },
    {
      "epoch": 65.03,
      "learning_rate": 1.1328932038834953e-05,
      "loss": 0.052,
      "step": 167460
    },
    {
      "epoch": 65.04,
      "learning_rate": 1.1328414239482201e-05,
      "loss": 0.0173,
      "step": 167470
    },
    {
      "epoch": 65.04,
      "learning_rate": 1.132789644012945e-05,
      "loss": 0.1732,
      "step": 167480
    },
    {
      "epoch": 65.04,
      "learning_rate": 1.13273786407767e-05,
      "loss": 0.1,
      "step": 167490
    },
    {
      "epoch": 65.05,
      "learning_rate": 1.132686084142395e-05,
      "loss": 0.0024,
      "step": 167500
    },
    {
      "epoch": 65.05,
      "learning_rate": 1.1326343042071198e-05,
      "loss": 0.0454,
      "step": 167510
    },
    {
      "epoch": 65.06,
      "learning_rate": 1.1325825242718447e-05,
      "loss": 0.0184,
      "step": 167520
    },
    {
      "epoch": 65.06,
      "learning_rate": 1.1325307443365697e-05,
      "loss": 0.0399,
      "step": 167530
    },
    {
      "epoch": 65.06,
      "learning_rate": 1.1324789644012945e-05,
      "loss": 0.1152,
      "step": 167540
    },
    {
      "epoch": 65.07,
      "learning_rate": 1.1324271844660195e-05,
      "loss": 0.0153,
      "step": 167550
    },
    {
      "epoch": 65.07,
      "learning_rate": 1.1323754045307444e-05,
      "loss": 0.0308,
      "step": 167560
    },
    {
      "epoch": 65.08,
      "learning_rate": 1.1323236245954694e-05,
      "loss": 0.0821,
      "step": 167570
    },
    {
      "epoch": 65.08,
      "learning_rate": 1.1322718446601943e-05,
      "loss": 0.1258,
      "step": 167580
    },
    {
      "epoch": 65.08,
      "learning_rate": 1.1322200647249191e-05,
      "loss": 0.0155,
      "step": 167590
    },
    {
      "epoch": 65.09,
      "learning_rate": 1.1321682847896441e-05,
      "loss": 0.033,
      "step": 167600
    },
    {
      "epoch": 65.09,
      "learning_rate": 1.132116504854369e-05,
      "loss": 0.0125,
      "step": 167610
    },
    {
      "epoch": 65.1,
      "learning_rate": 1.1320647249190938e-05,
      "loss": 0.0879,
      "step": 167620
    },
    {
      "epoch": 65.1,
      "learning_rate": 1.1320129449838188e-05,
      "loss": 0.056,
      "step": 167630
    },
    {
      "epoch": 65.1,
      "learning_rate": 1.1319611650485438e-05,
      "loss": 0.0985,
      "step": 167640
    },
    {
      "epoch": 65.11,
      "learning_rate": 1.1319093851132687e-05,
      "loss": 0.0935,
      "step": 167650
    },
    {
      "epoch": 65.11,
      "learning_rate": 1.1318576051779937e-05,
      "loss": 0.0002,
      "step": 167660
    },
    {
      "epoch": 65.11,
      "learning_rate": 1.1318058252427187e-05,
      "loss": 0.1038,
      "step": 167670
    },
    {
      "epoch": 65.12,
      "learning_rate": 1.1317540453074434e-05,
      "loss": 0.0836,
      "step": 167680
    },
    {
      "epoch": 65.12,
      "learning_rate": 1.1317022653721682e-05,
      "loss": 0.0438,
      "step": 167690
    },
    {
      "epoch": 65.13,
      "learning_rate": 1.1316504854368932e-05,
      "loss": 0.045,
      "step": 167700
    },
    {
      "epoch": 65.13,
      "learning_rate": 1.1315987055016182e-05,
      "loss": 0.1049,
      "step": 167710
    },
    {
      "epoch": 65.13,
      "learning_rate": 1.1315469255663431e-05,
      "loss": 0.0325,
      "step": 167720
    },
    {
      "epoch": 65.14,
      "learning_rate": 1.1314951456310681e-05,
      "loss": 0.0329,
      "step": 167730
    },
    {
      "epoch": 65.14,
      "learning_rate": 1.131443365695793e-05,
      "loss": 0.0501,
      "step": 167740
    },
    {
      "epoch": 65.15,
      "learning_rate": 1.131391585760518e-05,
      "loss": 0.1619,
      "step": 167750
    },
    {
      "epoch": 65.15,
      "learning_rate": 1.1313398058252428e-05,
      "loss": 0.1382,
      "step": 167760
    },
    {
      "epoch": 65.15,
      "learning_rate": 1.1312880258899676e-05,
      "loss": 0.0668,
      "step": 167770
    },
    {
      "epoch": 65.16,
      "learning_rate": 1.1312362459546926e-05,
      "loss": 0.0045,
      "step": 167780
    },
    {
      "epoch": 65.16,
      "learning_rate": 1.1311844660194175e-05,
      "loss": 0.017,
      "step": 167790
    },
    {
      "epoch": 65.17,
      "learning_rate": 1.1311326860841425e-05,
      "loss": 0.0117,
      "step": 167800
    },
    {
      "epoch": 65.17,
      "learning_rate": 1.1310809061488674e-05,
      "loss": 0.0605,
      "step": 167810
    },
    {
      "epoch": 65.17,
      "learning_rate": 1.1310291262135924e-05,
      "loss": 0.0933,
      "step": 167820
    },
    {
      "epoch": 65.18,
      "learning_rate": 1.1309773462783174e-05,
      "loss": 0.2166,
      "step": 167830
    },
    {
      "epoch": 65.18,
      "learning_rate": 1.130925566343042e-05,
      "loss": 0.0865,
      "step": 167840
    },
    {
      "epoch": 65.18,
      "learning_rate": 1.130873786407767e-05,
      "loss": 0.024,
      "step": 167850
    },
    {
      "epoch": 65.19,
      "learning_rate": 1.1308220064724919e-05,
      "loss": 0.0007,
      "step": 167860
    },
    {
      "epoch": 65.19,
      "learning_rate": 1.1307702265372169e-05,
      "loss": 0.0484,
      "step": 167870
    },
    {
      "epoch": 65.2,
      "learning_rate": 1.1307184466019418e-05,
      "loss": 0.0372,
      "step": 167880
    },
    {
      "epoch": 65.2,
      "learning_rate": 1.1306666666666668e-05,
      "loss": 0.0549,
      "step": 167890
    },
    {
      "epoch": 65.2,
      "learning_rate": 1.1306148867313918e-05,
      "loss": 0.0207,
      "step": 167900
    },
    {
      "epoch": 65.21,
      "learning_rate": 1.1305631067961167e-05,
      "loss": 0.1246,
      "step": 167910
    },
    {
      "epoch": 65.21,
      "learning_rate": 1.1305113268608413e-05,
      "loss": 0.112,
      "step": 167920
    },
    {
      "epoch": 65.22,
      "learning_rate": 1.1304595469255663e-05,
      "loss": 0.068,
      "step": 167930
    },
    {
      "epoch": 65.22,
      "learning_rate": 1.1304077669902913e-05,
      "loss": 0.0906,
      "step": 167940
    },
    {
      "epoch": 65.22,
      "learning_rate": 1.1303559870550162e-05,
      "loss": 0.1214,
      "step": 167950
    },
    {
      "epoch": 65.23,
      "learning_rate": 1.1303042071197412e-05,
      "loss": 0.0113,
      "step": 167960
    },
    {
      "epoch": 65.23,
      "learning_rate": 1.1302524271844662e-05,
      "loss": 0.0307,
      "step": 167970
    },
    {
      "epoch": 65.23,
      "learning_rate": 1.1302006472491911e-05,
      "loss": 0.0507,
      "step": 167980
    },
    {
      "epoch": 65.24,
      "learning_rate": 1.130148867313916e-05,
      "loss": 0.0555,
      "step": 167990
    },
    {
      "epoch": 65.24,
      "learning_rate": 1.1300970873786407e-05,
      "loss": 0.0003,
      "step": 168000
    },
    {
      "epoch": 65.25,
      "learning_rate": 1.1300453074433657e-05,
      "loss": 0.0746,
      "step": 168010
    },
    {
      "epoch": 65.25,
      "learning_rate": 1.1299935275080906e-05,
      "loss": 0.0357,
      "step": 168020
    },
    {
      "epoch": 65.25,
      "learning_rate": 1.1299417475728156e-05,
      "loss": 0.0657,
      "step": 168030
    },
    {
      "epoch": 65.26,
      "learning_rate": 1.1298899676375405e-05,
      "loss": 0.0024,
      "step": 168040
    },
    {
      "epoch": 65.26,
      "learning_rate": 1.1298381877022655e-05,
      "loss": 0.0037,
      "step": 168050
    },
    {
      "epoch": 65.27,
      "learning_rate": 1.1297864077669905e-05,
      "loss": 0.0722,
      "step": 168060
    },
    {
      "epoch": 65.27,
      "learning_rate": 1.1297346278317154e-05,
      "loss": 0.0629,
      "step": 168070
    },
    {
      "epoch": 65.27,
      "learning_rate": 1.12968284789644e-05,
      "loss": 0.1243,
      "step": 168080
    },
    {
      "epoch": 65.28,
      "learning_rate": 1.129631067961165e-05,
      "loss": 0.1355,
      "step": 168090
    },
    {
      "epoch": 65.28,
      "learning_rate": 1.12957928802589e-05,
      "loss": 0.0599,
      "step": 168100
    },
    {
      "epoch": 65.29,
      "learning_rate": 1.129527508090615e-05,
      "loss": 0.044,
      "step": 168110
    },
    {
      "epoch": 65.29,
      "learning_rate": 1.1294757281553399e-05,
      "loss": 0.0071,
      "step": 168120
    },
    {
      "epoch": 65.29,
      "learning_rate": 1.1294239482200649e-05,
      "loss": 0.0547,
      "step": 168130
    },
    {
      "epoch": 65.3,
      "learning_rate": 1.1293721682847898e-05,
      "loss": 0.0762,
      "step": 168140
    },
    {
      "epoch": 65.3,
      "learning_rate": 1.1293203883495148e-05,
      "loss": 0.138,
      "step": 168150
    },
    {
      "epoch": 65.3,
      "learning_rate": 1.1292686084142394e-05,
      "loss": 0.0519,
      "step": 168160
    },
    {
      "epoch": 65.31,
      "learning_rate": 1.1292168284789644e-05,
      "loss": 0.111,
      "step": 168170
    },
    {
      "epoch": 65.31,
      "learning_rate": 1.1291650485436893e-05,
      "loss": 0.0014,
      "step": 168180
    },
    {
      "epoch": 65.32,
      "learning_rate": 1.1291132686084143e-05,
      "loss": 0.0133,
      "step": 168190
    },
    {
      "epoch": 65.32,
      "learning_rate": 1.1290614886731393e-05,
      "loss": 0.0237,
      "step": 168200
    },
    {
      "epoch": 65.32,
      "learning_rate": 1.1290097087378642e-05,
      "loss": 0.0907,
      "step": 168210
    },
    {
      "epoch": 65.33,
      "learning_rate": 1.1289579288025892e-05,
      "loss": 0.0739,
      "step": 168220
    },
    {
      "epoch": 65.33,
      "learning_rate": 1.1289061488673141e-05,
      "loss": 0.0824,
      "step": 168230
    },
    {
      "epoch": 65.34,
      "learning_rate": 1.1288543689320391e-05,
      "loss": 0.0269,
      "step": 168240
    },
    {
      "epoch": 65.34,
      "learning_rate": 1.1288025889967637e-05,
      "loss": 0.1133,
      "step": 168250
    },
    {
      "epoch": 65.34,
      "learning_rate": 1.1287508090614887e-05,
      "loss": 0.0316,
      "step": 168260
    },
    {
      "epoch": 65.35,
      "learning_rate": 1.1286990291262137e-05,
      "loss": 0.1447,
      "step": 168270
    },
    {
      "epoch": 65.35,
      "learning_rate": 1.1286472491909386e-05,
      "loss": 0.0954,
      "step": 168280
    },
    {
      "epoch": 65.36,
      "learning_rate": 1.1285954692556636e-05,
      "loss": 0.2526,
      "step": 168290
    },
    {
      "epoch": 65.36,
      "learning_rate": 1.1285436893203885e-05,
      "loss": 0.0695,
      "step": 168300
    },
    {
      "epoch": 65.36,
      "learning_rate": 1.1284919093851135e-05,
      "loss": 0.0094,
      "step": 168310
    },
    {
      "epoch": 65.37,
      "learning_rate": 1.1284401294498385e-05,
      "loss": 0.1178,
      "step": 168320
    },
    {
      "epoch": 65.37,
      "learning_rate": 1.128388349514563e-05,
      "loss": 0.0469,
      "step": 168330
    },
    {
      "epoch": 65.37,
      "learning_rate": 1.128336569579288e-05,
      "loss": 0.062,
      "step": 168340
    },
    {
      "epoch": 65.38,
      "learning_rate": 1.128284789644013e-05,
      "loss": 0.009,
      "step": 168350
    },
    {
      "epoch": 65.38,
      "learning_rate": 1.128233009708738e-05,
      "loss": 0.0542,
      "step": 168360
    },
    {
      "epoch": 65.39,
      "learning_rate": 1.128181229773463e-05,
      "loss": 0.018,
      "step": 168370
    },
    {
      "epoch": 65.39,
      "learning_rate": 1.1281294498381879e-05,
      "loss": 0.1299,
      "step": 168380
    },
    {
      "epoch": 65.39,
      "learning_rate": 1.1280776699029129e-05,
      "loss": 0.0741,
      "step": 168390
    },
    {
      "epoch": 65.4,
      "learning_rate": 1.1280258899676378e-05,
      "loss": 0.0339,
      "step": 168400
    },
    {
      "epoch": 65.4,
      "learning_rate": 1.1279741100323624e-05,
      "loss": 0.0088,
      "step": 168410
    },
    {
      "epoch": 65.41,
      "learning_rate": 1.1279223300970874e-05,
      "loss": 0.1032,
      "step": 168420
    },
    {
      "epoch": 65.41,
      "learning_rate": 1.1278705501618124e-05,
      "loss": 0.0219,
      "step": 168430
    },
    {
      "epoch": 65.41,
      "learning_rate": 1.1278187702265373e-05,
      "loss": 0.1528,
      "step": 168440
    },
    {
      "epoch": 65.42,
      "learning_rate": 1.1277669902912623e-05,
      "loss": 0.0576,
      "step": 168450
    },
    {
      "epoch": 65.42,
      "learning_rate": 1.1277152103559872e-05,
      "loss": 0.1169,
      "step": 168460
    },
    {
      "epoch": 65.43,
      "learning_rate": 1.1276634304207122e-05,
      "loss": 0.1258,
      "step": 168470
    },
    {
      "epoch": 65.43,
      "learning_rate": 1.127611650485437e-05,
      "loss": 0.0647,
      "step": 168480
    },
    {
      "epoch": 65.43,
      "learning_rate": 1.1275598705501618e-05,
      "loss": 0.1146,
      "step": 168490
    },
    {
      "epoch": 65.44,
      "learning_rate": 1.1275080906148868e-05,
      "loss": 0.0372,
      "step": 168500
    },
    {
      "epoch": 65.44,
      "learning_rate": 1.1274563106796117e-05,
      "loss": 0.0205,
      "step": 168510
    },
    {
      "epoch": 65.44,
      "learning_rate": 1.1274045307443367e-05,
      "loss": 0.0475,
      "step": 168520
    },
    {
      "epoch": 65.45,
      "learning_rate": 1.1273527508090616e-05,
      "loss": 0.1392,
      "step": 168530
    },
    {
      "epoch": 65.45,
      "learning_rate": 1.1273009708737866e-05,
      "loss": 0.114,
      "step": 168540
    },
    {
      "epoch": 65.46,
      "learning_rate": 1.1272491909385116e-05,
      "loss": 0.0761,
      "step": 168550
    },
    {
      "epoch": 65.46,
      "learning_rate": 1.1271974110032364e-05,
      "loss": 0.0006,
      "step": 168560
    },
    {
      "epoch": 65.46,
      "learning_rate": 1.1271456310679611e-05,
      "loss": 0.1811,
      "step": 168570
    },
    {
      "epoch": 65.47,
      "learning_rate": 1.1270938511326861e-05,
      "loss": 0.0123,
      "step": 168580
    },
    {
      "epoch": 65.47,
      "learning_rate": 1.127042071197411e-05,
      "loss": 0.148,
      "step": 168590
    },
    {
      "epoch": 65.48,
      "learning_rate": 1.126990291262136e-05,
      "loss": 0.1811,
      "step": 168600
    },
    {
      "epoch": 65.48,
      "learning_rate": 1.126938511326861e-05,
      "loss": 0.1067,
      "step": 168610
    },
    {
      "epoch": 65.48,
      "learning_rate": 1.126886731391586e-05,
      "loss": 0.0556,
      "step": 168620
    },
    {
      "epoch": 65.49,
      "learning_rate": 1.1268349514563108e-05,
      "loss": 0.1919,
      "step": 168630
    },
    {
      "epoch": 65.49,
      "learning_rate": 1.1267831715210357e-05,
      "loss": 0.0088,
      "step": 168640
    },
    {
      "epoch": 65.5,
      "learning_rate": 1.1267313915857605e-05,
      "loss": 0.0519,
      "step": 168650
    },
    {
      "epoch": 65.5,
      "learning_rate": 1.1266796116504855e-05,
      "loss": 0.0923,
      "step": 168660
    },
    {
      "epoch": 65.5,
      "learning_rate": 1.1266278317152104e-05,
      "loss": 0.0012,
      "step": 168670
    },
    {
      "epoch": 65.51,
      "learning_rate": 1.1265760517799354e-05,
      "loss": 0.0154,
      "step": 168680
    },
    {
      "epoch": 65.51,
      "learning_rate": 1.1265242718446604e-05,
      "loss": 0.0317,
      "step": 168690
    },
    {
      "epoch": 65.51,
      "learning_rate": 1.1264724919093853e-05,
      "loss": 0.0025,
      "step": 168700
    },
    {
      "epoch": 65.52,
      "learning_rate": 1.1264207119741101e-05,
      "loss": 0.0201,
      "step": 168710
    },
    {
      "epoch": 65.52,
      "learning_rate": 1.126368932038835e-05,
      "loss": 0.1575,
      "step": 168720
    },
    {
      "epoch": 65.53,
      "learning_rate": 1.1263171521035599e-05,
      "loss": 0.0148,
      "step": 168730
    },
    {
      "epoch": 65.53,
      "learning_rate": 1.1262653721682848e-05,
      "loss": 0.1472,
      "step": 168740
    },
    {
      "epoch": 65.53,
      "learning_rate": 1.1262135922330098e-05,
      "loss": 0.1445,
      "step": 168750
    },
    {
      "epoch": 65.54,
      "learning_rate": 1.1261618122977347e-05,
      "loss": 0.1112,
      "step": 168760
    },
    {
      "epoch": 65.54,
      "learning_rate": 1.1261100323624597e-05,
      "loss": 0.0839,
      "step": 168770
    },
    {
      "epoch": 65.55,
      "learning_rate": 1.1260582524271845e-05,
      "loss": 0.0653,
      "step": 168780
    },
    {
      "epoch": 65.55,
      "learning_rate": 1.1260064724919095e-05,
      "loss": 0.0879,
      "step": 168790
    },
    {
      "epoch": 65.55,
      "learning_rate": 1.1259546925566344e-05,
      "loss": 0.0429,
      "step": 168800
    },
    {
      "epoch": 65.56,
      "learning_rate": 1.1259029126213594e-05,
      "loss": 0.042,
      "step": 168810
    },
    {
      "epoch": 65.56,
      "learning_rate": 1.1258511326860842e-05,
      "loss": 0.0183,
      "step": 168820
    },
    {
      "epoch": 65.57,
      "learning_rate": 1.1257993527508091e-05,
      "loss": 0.0681,
      "step": 168830
    },
    {
      "epoch": 65.57,
      "learning_rate": 1.1257475728155341e-05,
      "loss": 0.2782,
      "step": 168840
    },
    {
      "epoch": 65.57,
      "learning_rate": 1.125695792880259e-05,
      "loss": 0.0346,
      "step": 168850
    },
    {
      "epoch": 65.58,
      "learning_rate": 1.1256440129449839e-05,
      "loss": 0.198,
      "step": 168860
    },
    {
      "epoch": 65.58,
      "learning_rate": 1.1255922330097088e-05,
      "loss": 0.1432,
      "step": 168870
    },
    {
      "epoch": 65.58,
      "learning_rate": 1.1255404530744338e-05,
      "loss": 0.0696,
      "step": 168880
    },
    {
      "epoch": 65.59,
      "learning_rate": 1.1254886731391587e-05,
      "loss": 0.0462,
      "step": 168890
    },
    {
      "epoch": 65.59,
      "learning_rate": 1.1254368932038835e-05,
      "loss": 0.0881,
      "step": 168900
    },
    {
      "epoch": 65.6,
      "learning_rate": 1.1253851132686085e-05,
      "loss": 0.182,
      "step": 168910
    },
    {
      "epoch": 65.6,
      "learning_rate": 1.1253333333333335e-05,
      "loss": 0.0385,
      "step": 168920
    },
    {
      "epoch": 65.6,
      "learning_rate": 1.1252815533980584e-05,
      "loss": 0.0313,
      "step": 168930
    },
    {
      "epoch": 65.61,
      "learning_rate": 1.1252297734627832e-05,
      "loss": 0.0632,
      "step": 168940
    },
    {
      "epoch": 65.61,
      "learning_rate": 1.1251779935275082e-05,
      "loss": 0.0622,
      "step": 168950
    },
    {
      "epoch": 65.62,
      "learning_rate": 1.1251262135922331e-05,
      "loss": 0.0925,
      "step": 168960
    },
    {
      "epoch": 65.62,
      "learning_rate": 1.1250744336569581e-05,
      "loss": 0.1,
      "step": 168970
    },
    {
      "epoch": 65.62,
      "learning_rate": 1.1250226537216829e-05,
      "loss": 0.1298,
      "step": 168980
    },
    {
      "epoch": 65.63,
      "learning_rate": 1.1249708737864079e-05,
      "loss": 0.0903,
      "step": 168990
    },
    {
      "epoch": 65.63,
      "learning_rate": 1.1249190938511328e-05,
      "loss": 0.1195,
      "step": 169000
    },
    {
      "epoch": 65.63,
      "learning_rate": 1.1248673139158576e-05,
      "loss": 0.0175,
      "step": 169010
    },
    {
      "epoch": 65.64,
      "learning_rate": 1.1248155339805826e-05,
      "loss": 0.0088,
      "step": 169020
    },
    {
      "epoch": 65.64,
      "learning_rate": 1.1247637540453075e-05,
      "loss": 0.0502,
      "step": 169030
    },
    {
      "epoch": 65.65,
      "learning_rate": 1.1247119741100325e-05,
      "loss": 0.0193,
      "step": 169040
    },
    {
      "epoch": 65.65,
      "learning_rate": 1.1246601941747575e-05,
      "loss": 0.0613,
      "step": 169050
    },
    {
      "epoch": 65.65,
      "learning_rate": 1.1246084142394822e-05,
      "loss": 0.0065,
      "step": 169060
    },
    {
      "epoch": 65.66,
      "learning_rate": 1.1245566343042072e-05,
      "loss": 0.0012,
      "step": 169070
    },
    {
      "epoch": 65.66,
      "learning_rate": 1.1245048543689322e-05,
      "loss": 0.0226,
      "step": 169080
    },
    {
      "epoch": 65.67,
      "learning_rate": 1.124453074433657e-05,
      "loss": 0.0369,
      "step": 169090
    },
    {
      "epoch": 65.67,
      "learning_rate": 1.124401294498382e-05,
      "loss": 0.0997,
      "step": 169100
    },
    {
      "epoch": 65.67,
      "learning_rate": 1.1243495145631069e-05,
      "loss": 0.121,
      "step": 169110
    },
    {
      "epoch": 65.68,
      "learning_rate": 1.1242977346278318e-05,
      "loss": 0.0911,
      "step": 169120
    },
    {
      "epoch": 65.68,
      "learning_rate": 1.1242459546925568e-05,
      "loss": 0.11,
      "step": 169130
    },
    {
      "epoch": 65.69,
      "learning_rate": 1.1241941747572816e-05,
      "loss": 0.0579,
      "step": 169140
    },
    {
      "epoch": 65.69,
      "learning_rate": 1.1241423948220066e-05,
      "loss": 0.0054,
      "step": 169150
    },
    {
      "epoch": 65.69,
      "learning_rate": 1.1240906148867314e-05,
      "loss": 0.1151,
      "step": 169160
    },
    {
      "epoch": 65.7,
      "learning_rate": 1.1240388349514563e-05,
      "loss": 0.0751,
      "step": 169170
    },
    {
      "epoch": 65.7,
      "learning_rate": 1.1239870550161813e-05,
      "loss": 0.0127,
      "step": 169180
    },
    {
      "epoch": 65.7,
      "learning_rate": 1.1239352750809062e-05,
      "loss": 0.0365,
      "step": 169190
    },
    {
      "epoch": 65.71,
      "learning_rate": 1.1238834951456312e-05,
      "loss": 0.1651,
      "step": 169200
    },
    {
      "epoch": 65.71,
      "learning_rate": 1.1238317152103562e-05,
      "loss": 0.0014,
      "step": 169210
    },
    {
      "epoch": 65.72,
      "learning_rate": 1.123779935275081e-05,
      "loss": 0.0186,
      "step": 169220
    },
    {
      "epoch": 65.72,
      "learning_rate": 1.123728155339806e-05,
      "loss": 0.0912,
      "step": 169230
    },
    {
      "epoch": 65.72,
      "learning_rate": 1.1236763754045307e-05,
      "loss": 0.1584,
      "step": 169240
    },
    {
      "epoch": 65.73,
      "learning_rate": 1.1236245954692557e-05,
      "loss": 0.2207,
      "step": 169250
    },
    {
      "epoch": 65.73,
      "learning_rate": 1.1235728155339806e-05,
      "loss": 0.0913,
      "step": 169260
    },
    {
      "epoch": 65.74,
      "learning_rate": 1.1235210355987056e-05,
      "loss": 0.0852,
      "step": 169270
    },
    {
      "epoch": 65.74,
      "learning_rate": 1.1234692556634306e-05,
      "loss": 0.0498,
      "step": 169280
    },
    {
      "epoch": 65.74,
      "learning_rate": 1.1234174757281555e-05,
      "loss": 0.0784,
      "step": 169290
    },
    {
      "epoch": 65.75,
      "learning_rate": 1.1233656957928805e-05,
      "loss": 0.0513,
      "step": 169300
    },
    {
      "epoch": 65.75,
      "learning_rate": 1.1233139158576051e-05,
      "loss": 0.0756,
      "step": 169310
    },
    {
      "epoch": 65.76,
      "learning_rate": 1.12326213592233e-05,
      "loss": 0.1284,
      "step": 169320
    },
    {
      "epoch": 65.76,
      "learning_rate": 1.123210355987055e-05,
      "loss": 0.0289,
      "step": 169330
    },
    {
      "epoch": 65.76,
      "learning_rate": 1.12315857605178e-05,
      "loss": 0.0618,
      "step": 169340
    },
    {
      "epoch": 65.77,
      "learning_rate": 1.123106796116505e-05,
      "loss": 0.0217,
      "step": 169350
    },
    {
      "epoch": 65.77,
      "learning_rate": 1.1230550161812299e-05,
      "loss": 0.0328,
      "step": 169360
    },
    {
      "epoch": 65.77,
      "learning_rate": 1.1230032362459549e-05,
      "loss": 0.0212,
      "step": 169370
    },
    {
      "epoch": 65.78,
      "learning_rate": 1.1229514563106798e-05,
      "loss": 0.0109,
      "step": 169380
    },
    {
      "epoch": 65.78,
      "learning_rate": 1.1228996763754045e-05,
      "loss": 0.0803,
      "step": 169390
    },
    {
      "epoch": 65.79,
      "learning_rate": 1.1228478964401294e-05,
      "loss": 0.0755,
      "step": 169400
    },
    {
      "epoch": 65.79,
      "learning_rate": 1.1227961165048544e-05,
      "loss": 0.2043,
      "step": 169410
    },
    {
      "epoch": 65.79,
      "learning_rate": 1.1227443365695793e-05,
      "loss": 0.0455,
      "step": 169420
    },
    {
      "epoch": 65.8,
      "learning_rate": 1.1226925566343043e-05,
      "loss": 0.0441,
      "step": 169430
    },
    {
      "epoch": 65.8,
      "learning_rate": 1.1226407766990293e-05,
      "loss": 0.0118,
      "step": 169440
    },
    {
      "epoch": 65.81,
      "learning_rate": 1.1225889967637542e-05,
      "loss": 0.0548,
      "step": 169450
    },
    {
      "epoch": 65.81,
      "learning_rate": 1.1225372168284792e-05,
      "loss": 0.0754,
      "step": 169460
    },
    {
      "epoch": 65.81,
      "learning_rate": 1.1224854368932038e-05,
      "loss": 0.0244,
      "step": 169470
    },
    {
      "epoch": 65.82,
      "learning_rate": 1.1224336569579288e-05,
      "loss": 0.0006,
      "step": 169480
    },
    {
      "epoch": 65.82,
      "learning_rate": 1.1223818770226537e-05,
      "loss": 0.0816,
      "step": 169490
    },
    {
      "epoch": 65.83,
      "learning_rate": 1.1223300970873787e-05,
      "loss": 0.0633,
      "step": 169500
    },
    {
      "epoch": 65.83,
      "learning_rate": 1.1222783171521037e-05,
      "loss": 0.1546,
      "step": 169510
    },
    {
      "epoch": 65.83,
      "learning_rate": 1.1222265372168286e-05,
      "loss": 0.0533,
      "step": 169520
    },
    {
      "epoch": 65.84,
      "learning_rate": 1.1221747572815536e-05,
      "loss": 0.0747,
      "step": 169530
    },
    {
      "epoch": 65.84,
      "learning_rate": 1.1221229773462785e-05,
      "loss": 0.1262,
      "step": 169540
    },
    {
      "epoch": 65.84,
      "learning_rate": 1.1220711974110032e-05,
      "loss": 0.2004,
      "step": 169550
    },
    {
      "epoch": 65.85,
      "learning_rate": 1.1220194174757281e-05,
      "loss": 0.2035,
      "step": 169560
    },
    {
      "epoch": 65.85,
      "learning_rate": 1.1219676375404531e-05,
      "loss": 0.0413,
      "step": 169570
    },
    {
      "epoch": 65.86,
      "learning_rate": 1.121915857605178e-05,
      "loss": 0.0593,
      "step": 169580
    },
    {
      "epoch": 65.86,
      "learning_rate": 1.121864077669903e-05,
      "loss": 0.0388,
      "step": 169590
    },
    {
      "epoch": 65.86,
      "learning_rate": 1.121812297734628e-05,
      "loss": 0.066,
      "step": 169600
    },
    {
      "epoch": 65.87,
      "learning_rate": 1.121760517799353e-05,
      "loss": 0.0605,
      "step": 169610
    },
    {
      "epoch": 65.87,
      "learning_rate": 1.1217087378640779e-05,
      "loss": 0.0115,
      "step": 169620
    },
    {
      "epoch": 65.88,
      "learning_rate": 1.1216569579288025e-05,
      "loss": 0.0758,
      "step": 169630
    },
    {
      "epoch": 65.88,
      "learning_rate": 1.1216051779935275e-05,
      "loss": 0.0616,
      "step": 169640
    },
    {
      "epoch": 65.88,
      "learning_rate": 1.1215533980582525e-05,
      "loss": 0.0375,
      "step": 169650
    },
    {
      "epoch": 65.89,
      "learning_rate": 1.1215016181229774e-05,
      "loss": 0.0406,
      "step": 169660
    },
    {
      "epoch": 65.89,
      "learning_rate": 1.1214498381877024e-05,
      "loss": 0.0152,
      "step": 169670
    },
    {
      "epoch": 65.9,
      "learning_rate": 1.1213980582524273e-05,
      "loss": 0.0099,
      "step": 169680
    },
    {
      "epoch": 65.9,
      "learning_rate": 1.1213462783171523e-05,
      "loss": 0.0812,
      "step": 169690
    },
    {
      "epoch": 65.9,
      "learning_rate": 1.1212944983818773e-05,
      "loss": 0.0689,
      "step": 169700
    },
    {
      "epoch": 65.91,
      "learning_rate": 1.1212427184466019e-05,
      "loss": 0.0514,
      "step": 169710
    },
    {
      "epoch": 65.91,
      "learning_rate": 1.1211909385113268e-05,
      "loss": 0.0185,
      "step": 169720
    },
    {
      "epoch": 65.91,
      "learning_rate": 1.1211391585760518e-05,
      "loss": 0.0466,
      "step": 169730
    },
    {
      "epoch": 65.92,
      "learning_rate": 1.1210873786407768e-05,
      "loss": 0.0966,
      "step": 169740
    },
    {
      "epoch": 65.92,
      "learning_rate": 1.1210355987055017e-05,
      "loss": 0.1537,
      "step": 169750
    },
    {
      "epoch": 65.93,
      "learning_rate": 1.1209838187702267e-05,
      "loss": 0.1197,
      "step": 169760
    },
    {
      "epoch": 65.93,
      "learning_rate": 1.1209320388349517e-05,
      "loss": 0.1824,
      "step": 169770
    },
    {
      "epoch": 65.93,
      "learning_rate": 1.1208802588996766e-05,
      "loss": 0.0434,
      "step": 169780
    },
    {
      "epoch": 65.94,
      "learning_rate": 1.1208284789644012e-05,
      "loss": 0.0166,
      "step": 169790
    },
    {
      "epoch": 65.94,
      "learning_rate": 1.1207766990291262e-05,
      "loss": 0.0263,
      "step": 169800
    },
    {
      "epoch": 65.95,
      "learning_rate": 1.1207249190938512e-05,
      "loss": 0.0941,
      "step": 169810
    },
    {
      "epoch": 65.95,
      "learning_rate": 1.1206731391585761e-05,
      "loss": 0.1069,
      "step": 169820
    },
    {
      "epoch": 65.95,
      "learning_rate": 1.1206213592233011e-05,
      "loss": 0.0086,
      "step": 169830
    },
    {
      "epoch": 65.96,
      "learning_rate": 1.120569579288026e-05,
      "loss": 0.0293,
      "step": 169840
    },
    {
      "epoch": 65.96,
      "learning_rate": 1.120517799352751e-05,
      "loss": 0.0317,
      "step": 169850
    },
    {
      "epoch": 65.97,
      "learning_rate": 1.120466019417476e-05,
      "loss": 0.0217,
      "step": 169860
    },
    {
      "epoch": 65.97,
      "learning_rate": 1.120414239482201e-05,
      "loss": 0.0499,
      "step": 169870
    },
    {
      "epoch": 65.97,
      "learning_rate": 1.1203624595469256e-05,
      "loss": 0.2496,
      "step": 169880
    },
    {
      "epoch": 65.98,
      "learning_rate": 1.1203106796116505e-05,
      "loss": 0.1562,
      "step": 169890
    },
    {
      "epoch": 65.98,
      "learning_rate": 1.1202588996763755e-05,
      "loss": 0.0027,
      "step": 169900
    },
    {
      "epoch": 65.98,
      "learning_rate": 1.1202071197411004e-05,
      "loss": 0.0026,
      "step": 169910
    },
    {
      "epoch": 65.99,
      "learning_rate": 1.1201553398058254e-05,
      "loss": 0.0693,
      "step": 169920
    },
    {
      "epoch": 65.99,
      "learning_rate": 1.1201035598705504e-05,
      "loss": 0.0842,
      "step": 169930
    },
    {
      "epoch": 66.0,
      "learning_rate": 1.1200517799352753e-05,
      "loss": 0.1062,
      "step": 169940
    },
    {
      "epoch": 66.0,
      "learning_rate": 1.1200000000000001e-05,
      "loss": 0.2337,
      "step": 169950
    },
    {
      "epoch": 66.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.3330360949039459,
      "eval_runtime": 8.1861,
      "eval_samples_per_second": 444.046,
      "eval_steps_per_second": 55.582,
      "step": 169950
    },
    {
      "epoch": 66.0,
      "learning_rate": 1.1199482200647249e-05,
      "loss": 0.0164,
      "step": 169960
    },
    {
      "epoch": 66.01,
      "learning_rate": 1.1198964401294499e-05,
      "loss": 0.0503,
      "step": 169970
    },
    {
      "epoch": 66.01,
      "learning_rate": 1.1198446601941748e-05,
      "loss": 0.0116,
      "step": 169980
    },
    {
      "epoch": 66.02,
      "learning_rate": 1.1197928802588998e-05,
      "loss": 0.0566,
      "step": 169990
    },
    {
      "epoch": 66.02,
      "learning_rate": 1.1197411003236248e-05,
      "loss": 0.0158,
      "step": 170000
    },
    {
      "epoch": 66.02,
      "learning_rate": 1.1196893203883497e-05,
      "loss": 0.0013,
      "step": 170010
    },
    {
      "epoch": 66.03,
      "learning_rate": 1.1196375404530747e-05,
      "loss": 0.0211,
      "step": 170020
    },
    {
      "epoch": 66.03,
      "learning_rate": 1.1195857605177995e-05,
      "loss": 0.0365,
      "step": 170030
    },
    {
      "epoch": 66.03,
      "learning_rate": 1.1195339805825243e-05,
      "loss": 0.063,
      "step": 170040
    },
    {
      "epoch": 66.04,
      "learning_rate": 1.1194822006472492e-05,
      "loss": 0.0729,
      "step": 170050
    },
    {
      "epoch": 66.04,
      "learning_rate": 1.1194304207119742e-05,
      "loss": 0.1178,
      "step": 170060
    },
    {
      "epoch": 66.05,
      "learning_rate": 1.1193786407766992e-05,
      "loss": 0.0285,
      "step": 170070
    },
    {
      "epoch": 66.05,
      "learning_rate": 1.1193268608414241e-05,
      "loss": 0.0782,
      "step": 170080
    },
    {
      "epoch": 66.05,
      "learning_rate": 1.119275080906149e-05,
      "loss": 0.0034,
      "step": 170090
    },
    {
      "epoch": 66.06,
      "learning_rate": 1.1192233009708739e-05,
      "loss": 0.143,
      "step": 170100
    },
    {
      "epoch": 66.06,
      "learning_rate": 1.1191715210355988e-05,
      "loss": 0.0576,
      "step": 170110
    },
    {
      "epoch": 66.07,
      "learning_rate": 1.1191197411003236e-05,
      "loss": 0.0518,
      "step": 170120
    },
    {
      "epoch": 66.07,
      "learning_rate": 1.1190679611650486e-05,
      "loss": 0.2071,
      "step": 170130
    },
    {
      "epoch": 66.07,
      "learning_rate": 1.1190161812297735e-05,
      "loss": 0.0009,
      "step": 170140
    },
    {
      "epoch": 66.08,
      "learning_rate": 1.1189644012944985e-05,
      "loss": 0.1267,
      "step": 170150
    },
    {
      "epoch": 66.08,
      "learning_rate": 1.1189126213592235e-05,
      "loss": 0.0587,
      "step": 170160
    },
    {
      "epoch": 66.09,
      "learning_rate": 1.1188608414239484e-05,
      "loss": 0.0269,
      "step": 170170
    },
    {
      "epoch": 66.09,
      "learning_rate": 1.1188090614886732e-05,
      "loss": 0.066,
      "step": 170180
    },
    {
      "epoch": 66.09,
      "learning_rate": 1.1187572815533982e-05,
      "loss": 0.0908,
      "step": 170190
    },
    {
      "epoch": 66.1,
      "learning_rate": 1.118705501618123e-05,
      "loss": 0.0442,
      "step": 170200
    },
    {
      "epoch": 66.1,
      "learning_rate": 1.118653721682848e-05,
      "loss": 0.0482,
      "step": 170210
    },
    {
      "epoch": 66.1,
      "learning_rate": 1.1186019417475729e-05,
      "loss": 0.0749,
      "step": 170220
    },
    {
      "epoch": 66.11,
      "learning_rate": 1.1185501618122979e-05,
      "loss": 0.0209,
      "step": 170230
    },
    {
      "epoch": 66.11,
      "learning_rate": 1.1184983818770228e-05,
      "loss": 0.0889,
      "step": 170240
    },
    {
      "epoch": 66.12,
      "learning_rate": 1.1184466019417476e-05,
      "loss": 0.0182,
      "step": 170250
    },
    {
      "epoch": 66.12,
      "learning_rate": 1.1183948220064726e-05,
      "loss": 0.002,
      "step": 170260
    },
    {
      "epoch": 66.12,
      "learning_rate": 1.1183430420711975e-05,
      "loss": 0.0438,
      "step": 170270
    },
    {
      "epoch": 66.13,
      "learning_rate": 1.1182912621359223e-05,
      "loss": 0.1083,
      "step": 170280
    },
    {
      "epoch": 66.13,
      "learning_rate": 1.1182394822006473e-05,
      "loss": 0.1071,
      "step": 170290
    },
    {
      "epoch": 66.14,
      "learning_rate": 1.1181877022653723e-05,
      "loss": 0.0628,
      "step": 170300
    },
    {
      "epoch": 66.14,
      "learning_rate": 1.1181359223300972e-05,
      "loss": 0.0773,
      "step": 170310
    },
    {
      "epoch": 66.14,
      "learning_rate": 1.1180841423948222e-05,
      "loss": 0.1106,
      "step": 170320
    },
    {
      "epoch": 66.15,
      "learning_rate": 1.118032362459547e-05,
      "loss": 0.0828,
      "step": 170330
    },
    {
      "epoch": 66.15,
      "learning_rate": 1.117980582524272e-05,
      "loss": 0.0522,
      "step": 170340
    },
    {
      "epoch": 66.16,
      "learning_rate": 1.1179288025889969e-05,
      "loss": 0.1227,
      "step": 170350
    },
    {
      "epoch": 66.16,
      "learning_rate": 1.1178770226537217e-05,
      "loss": 0.0621,
      "step": 170360
    },
    {
      "epoch": 66.16,
      "learning_rate": 1.1178252427184467e-05,
      "loss": 0.0111,
      "step": 170370
    },
    {
      "epoch": 66.17,
      "learning_rate": 1.1177734627831716e-05,
      "loss": 0.0042,
      "step": 170380
    },
    {
      "epoch": 66.17,
      "learning_rate": 1.1177216828478966e-05,
      "loss": 0.0007,
      "step": 170390
    },
    {
      "epoch": 66.17,
      "learning_rate": 1.1176699029126214e-05,
      "loss": 0.0772,
      "step": 170400
    },
    {
      "epoch": 66.18,
      "learning_rate": 1.1176181229773463e-05,
      "loss": 0.0739,
      "step": 170410
    },
    {
      "epoch": 66.18,
      "learning_rate": 1.1175663430420713e-05,
      "loss": 0.0437,
      "step": 170420
    },
    {
      "epoch": 66.19,
      "learning_rate": 1.1175145631067963e-05,
      "loss": 0.1578,
      "step": 170430
    },
    {
      "epoch": 66.19,
      "learning_rate": 1.1174627831715212e-05,
      "loss": 0.0933,
      "step": 170440
    },
    {
      "epoch": 66.19,
      "learning_rate": 1.117411003236246e-05,
      "loss": 0.077,
      "step": 170450
    },
    {
      "epoch": 66.2,
      "learning_rate": 1.117359223300971e-05,
      "loss": 0.0463,
      "step": 170460
    },
    {
      "epoch": 66.2,
      "learning_rate": 1.117307443365696e-05,
      "loss": 0.0957,
      "step": 170470
    },
    {
      "epoch": 66.21,
      "learning_rate": 1.1172556634304207e-05,
      "loss": 0.0933,
      "step": 170480
    },
    {
      "epoch": 66.21,
      "learning_rate": 1.1172038834951457e-05,
      "loss": 0.0436,
      "step": 170490
    },
    {
      "epoch": 66.21,
      "learning_rate": 1.1171521035598706e-05,
      "loss": 0.0162,
      "step": 170500
    },
    {
      "epoch": 66.22,
      "learning_rate": 1.1171003236245956e-05,
      "loss": 0.0407,
      "step": 170510
    },
    {
      "epoch": 66.22,
      "learning_rate": 1.1170485436893206e-05,
      "loss": 0.0451,
      "step": 170520
    },
    {
      "epoch": 66.23,
      "learning_rate": 1.1169967637540454e-05,
      "loss": 0.0996,
      "step": 170530
    },
    {
      "epoch": 66.23,
      "learning_rate": 1.1169449838187703e-05,
      "loss": 0.2134,
      "step": 170540
    },
    {
      "epoch": 66.23,
      "learning_rate": 1.1168932038834953e-05,
      "loss": 0.0433,
      "step": 170550
    },
    {
      "epoch": 66.24,
      "learning_rate": 1.11684142394822e-05,
      "loss": 0.0977,
      "step": 170560
    },
    {
      "epoch": 66.24,
      "learning_rate": 1.116789644012945e-05,
      "loss": 0.1031,
      "step": 170570
    },
    {
      "epoch": 66.24,
      "learning_rate": 1.11673786407767e-05,
      "loss": 0.0335,
      "step": 170580
    },
    {
      "epoch": 66.25,
      "learning_rate": 1.116686084142395e-05,
      "loss": 0.0237,
      "step": 170590
    },
    {
      "epoch": 66.25,
      "learning_rate": 1.11663430420712e-05,
      "loss": 0.0059,
      "step": 170600
    },
    {
      "epoch": 66.26,
      "learning_rate": 1.1165825242718447e-05,
      "loss": 0.0984,
      "step": 170610
    },
    {
      "epoch": 66.26,
      "learning_rate": 1.1165307443365697e-05,
      "loss": 0.0079,
      "step": 170620
    },
    {
      "epoch": 66.26,
      "learning_rate": 1.1164789644012945e-05,
      "loss": 0.0672,
      "step": 170630
    },
    {
      "epoch": 66.27,
      "learning_rate": 1.1164271844660194e-05,
      "loss": 0.0109,
      "step": 170640
    },
    {
      "epoch": 66.27,
      "learning_rate": 1.1163754045307444e-05,
      "loss": 0.0446,
      "step": 170650
    },
    {
      "epoch": 66.28,
      "learning_rate": 1.1163236245954694e-05,
      "loss": 0.0308,
      "step": 170660
    },
    {
      "epoch": 66.28,
      "learning_rate": 1.1162718446601943e-05,
      "loss": 0.048,
      "step": 170670
    },
    {
      "epoch": 66.28,
      "learning_rate": 1.1162200647249193e-05,
      "loss": 0.1441,
      "step": 170680
    },
    {
      "epoch": 66.29,
      "learning_rate": 1.116168284789644e-05,
      "loss": 0.0582,
      "step": 170690
    },
    {
      "epoch": 66.29,
      "learning_rate": 1.116116504854369e-05,
      "loss": 0.1504,
      "step": 170700
    },
    {
      "epoch": 66.3,
      "learning_rate": 1.1160647249190938e-05,
      "loss": 0.1974,
      "step": 170710
    },
    {
      "epoch": 66.3,
      "learning_rate": 1.1160129449838188e-05,
      "loss": 0.0586,
      "step": 170720
    },
    {
      "epoch": 66.3,
      "learning_rate": 1.1159611650485438e-05,
      "loss": 0.083,
      "step": 170730
    },
    {
      "epoch": 66.31,
      "learning_rate": 1.1159093851132687e-05,
      "loss": 0.1323,
      "step": 170740
    },
    {
      "epoch": 66.31,
      "learning_rate": 1.1158576051779937e-05,
      "loss": 0.0673,
      "step": 170750
    },
    {
      "epoch": 66.31,
      "learning_rate": 1.1158058252427186e-05,
      "loss": 0.0625,
      "step": 170760
    },
    {
      "epoch": 66.32,
      "learning_rate": 1.1157540453074434e-05,
      "loss": 0.0664,
      "step": 170770
    },
    {
      "epoch": 66.32,
      "learning_rate": 1.1157022653721682e-05,
      "loss": 0.0022,
      "step": 170780
    },
    {
      "epoch": 66.33,
      "learning_rate": 1.1156504854368932e-05,
      "loss": 0.1038,
      "step": 170790
    },
    {
      "epoch": 66.33,
      "learning_rate": 1.1155987055016181e-05,
      "loss": 0.1543,
      "step": 170800
    },
    {
      "epoch": 66.33,
      "learning_rate": 1.1155469255663431e-05,
      "loss": 0.0405,
      "step": 170810
    },
    {
      "epoch": 66.34,
      "learning_rate": 1.115495145631068e-05,
      "loss": 0.1025,
      "step": 170820
    },
    {
      "epoch": 66.34,
      "learning_rate": 1.115443365695793e-05,
      "loss": 0.0546,
      "step": 170830
    },
    {
      "epoch": 66.35,
      "learning_rate": 1.115391585760518e-05,
      "loss": 0.2046,
      "step": 170840
    },
    {
      "epoch": 66.35,
      "learning_rate": 1.1153398058252428e-05,
      "loss": 0.1409,
      "step": 170850
    },
    {
      "epoch": 66.35,
      "learning_rate": 1.1152880258899676e-05,
      "loss": 0.0124,
      "step": 170860
    },
    {
      "epoch": 66.36,
      "learning_rate": 1.1152362459546925e-05,
      "loss": 0.0861,
      "step": 170870
    },
    {
      "epoch": 66.36,
      "learning_rate": 1.1151844660194175e-05,
      "loss": 0.037,
      "step": 170880
    },
    {
      "epoch": 66.37,
      "learning_rate": 1.1151326860841425e-05,
      "loss": 0.0237,
      "step": 170890
    },
    {
      "epoch": 66.37,
      "learning_rate": 1.1150809061488674e-05,
      "loss": 0.07,
      "step": 170900
    },
    {
      "epoch": 66.37,
      "learning_rate": 1.1150291262135924e-05,
      "loss": 0.0118,
      "step": 170910
    },
    {
      "epoch": 66.38,
      "learning_rate": 1.1149773462783173e-05,
      "loss": 0.0501,
      "step": 170920
    },
    {
      "epoch": 66.38,
      "learning_rate": 1.114925566343042e-05,
      "loss": 0.0415,
      "step": 170930
    },
    {
      "epoch": 66.38,
      "learning_rate": 1.114873786407767e-05,
      "loss": 0.0208,
      "step": 170940
    },
    {
      "epoch": 66.39,
      "learning_rate": 1.1148220064724919e-05,
      "loss": 0.0374,
      "step": 170950
    },
    {
      "epoch": 66.39,
      "learning_rate": 1.1147702265372169e-05,
      "loss": 0.068,
      "step": 170960
    },
    {
      "epoch": 66.4,
      "learning_rate": 1.1147184466019418e-05,
      "loss": 0.0405,
      "step": 170970
    },
    {
      "epoch": 66.4,
      "learning_rate": 1.1146666666666668e-05,
      "loss": 0.0754,
      "step": 170980
    },
    {
      "epoch": 66.4,
      "learning_rate": 1.1146148867313917e-05,
      "loss": 0.1058,
      "step": 170990
    },
    {
      "epoch": 66.41,
      "learning_rate": 1.1145631067961167e-05,
      "loss": 0.0191,
      "step": 171000
    },
    {
      "epoch": 66.41,
      "learning_rate": 1.1145113268608417e-05,
      "loss": 0.1645,
      "step": 171010
    },
    {
      "epoch": 66.42,
      "learning_rate": 1.1144595469255663e-05,
      "loss": 0.0795,
      "step": 171020
    },
    {
      "epoch": 66.42,
      "learning_rate": 1.1144077669902913e-05,
      "loss": 0.1664,
      "step": 171030
    },
    {
      "epoch": 66.42,
      "learning_rate": 1.1143559870550162e-05,
      "loss": 0.0609,
      "step": 171040
    },
    {
      "epoch": 66.43,
      "learning_rate": 1.1143042071197412e-05,
      "loss": 0.1361,
      "step": 171050
    },
    {
      "epoch": 66.43,
      "learning_rate": 1.1142524271844661e-05,
      "loss": 0.061,
      "step": 171060
    },
    {
      "epoch": 66.43,
      "learning_rate": 1.1142006472491911e-05,
      "loss": 0.0413,
      "step": 171070
    },
    {
      "epoch": 66.44,
      "learning_rate": 1.114148867313916e-05,
      "loss": 0.0726,
      "step": 171080
    },
    {
      "epoch": 66.44,
      "learning_rate": 1.114097087378641e-05,
      "loss": 0.1758,
      "step": 171090
    },
    {
      "epoch": 66.45,
      "learning_rate": 1.1140453074433656e-05,
      "loss": 0.1055,
      "step": 171100
    },
    {
      "epoch": 66.45,
      "learning_rate": 1.1139935275080906e-05,
      "loss": 0.0907,
      "step": 171110
    },
    {
      "epoch": 66.45,
      "learning_rate": 1.1139417475728156e-05,
      "loss": 0.0158,
      "step": 171120
    },
    {
      "epoch": 66.46,
      "learning_rate": 1.1138899676375405e-05,
      "loss": 0.0445,
      "step": 171130
    },
    {
      "epoch": 66.46,
      "learning_rate": 1.1138381877022655e-05,
      "loss": 0.0049,
      "step": 171140
    },
    {
      "epoch": 66.47,
      "learning_rate": 1.1137864077669905e-05,
      "loss": 0.0002,
      "step": 171150
    },
    {
      "epoch": 66.47,
      "learning_rate": 1.1137346278317154e-05,
      "loss": 0.0015,
      "step": 171160
    },
    {
      "epoch": 66.47,
      "learning_rate": 1.1136828478964404e-05,
      "loss": 0.1283,
      "step": 171170
    },
    {
      "epoch": 66.48,
      "learning_rate": 1.113631067961165e-05,
      "loss": 0.2223,
      "step": 171180
    },
    {
      "epoch": 66.48,
      "learning_rate": 1.11357928802589e-05,
      "loss": 0.0186,
      "step": 171190
    },
    {
      "epoch": 66.49,
      "learning_rate": 1.113527508090615e-05,
      "loss": 0.0635,
      "step": 171200
    },
    {
      "epoch": 66.49,
      "learning_rate": 1.1134757281553399e-05,
      "loss": 0.0117,
      "step": 171210
    },
    {
      "epoch": 66.49,
      "learning_rate": 1.1134239482200648e-05,
      "loss": 0.0341,
      "step": 171220
    },
    {
      "epoch": 66.5,
      "learning_rate": 1.1133721682847898e-05,
      "loss": 0.1455,
      "step": 171230
    },
    {
      "epoch": 66.5,
      "learning_rate": 1.1133203883495148e-05,
      "loss": 0.2285,
      "step": 171240
    },
    {
      "epoch": 66.5,
      "learning_rate": 1.1132686084142397e-05,
      "loss": 0.2105,
      "step": 171250
    },
    {
      "epoch": 66.51,
      "learning_rate": 1.1132168284789644e-05,
      "loss": 0.0279,
      "step": 171260
    },
    {
      "epoch": 66.51,
      "learning_rate": 1.1131650485436893e-05,
      "loss": 0.1267,
      "step": 171270
    },
    {
      "epoch": 66.52,
      "learning_rate": 1.1131132686084143e-05,
      "loss": 0.0851,
      "step": 171280
    },
    {
      "epoch": 66.52,
      "learning_rate": 1.1130614886731392e-05,
      "loss": 0.2046,
      "step": 171290
    },
    {
      "epoch": 66.52,
      "learning_rate": 1.1130097087378642e-05,
      "loss": 0.0394,
      "step": 171300
    },
    {
      "epoch": 66.53,
      "learning_rate": 1.1129579288025892e-05,
      "loss": 0.1169,
      "step": 171310
    },
    {
      "epoch": 66.53,
      "learning_rate": 1.1129061488673141e-05,
      "loss": 0.0013,
      "step": 171320
    },
    {
      "epoch": 66.54,
      "learning_rate": 1.1128543689320391e-05,
      "loss": 0.0015,
      "step": 171330
    },
    {
      "epoch": 66.54,
      "learning_rate": 1.1128025889967637e-05,
      "loss": 0.1745,
      "step": 171340
    },
    {
      "epoch": 66.54,
      "learning_rate": 1.1127508090614887e-05,
      "loss": 0.1446,
      "step": 171350
    },
    {
      "epoch": 66.55,
      "learning_rate": 1.1126990291262136e-05,
      "loss": 0.1358,
      "step": 171360
    },
    {
      "epoch": 66.55,
      "learning_rate": 1.1126472491909386e-05,
      "loss": 0.0247,
      "step": 171370
    },
    {
      "epoch": 66.56,
      "learning_rate": 1.1125954692556636e-05,
      "loss": 0.0125,
      "step": 171380
    },
    {
      "epoch": 66.56,
      "learning_rate": 1.1125436893203885e-05,
      "loss": 0.0887,
      "step": 171390
    },
    {
      "epoch": 66.56,
      "learning_rate": 1.1124919093851135e-05,
      "loss": 0.0498,
      "step": 171400
    },
    {
      "epoch": 66.57,
      "learning_rate": 1.1124401294498384e-05,
      "loss": 0.0127,
      "step": 171410
    },
    {
      "epoch": 66.57,
      "learning_rate": 1.112388349514563e-05,
      "loss": 0.0069,
      "step": 171420
    },
    {
      "epoch": 66.57,
      "learning_rate": 1.112336569579288e-05,
      "loss": 0.0324,
      "step": 171430
    },
    {
      "epoch": 66.58,
      "learning_rate": 1.112284789644013e-05,
      "loss": 0.0845,
      "step": 171440
    },
    {
      "epoch": 66.58,
      "learning_rate": 1.112233009708738e-05,
      "loss": 0.1043,
      "step": 171450
    },
    {
      "epoch": 66.59,
      "learning_rate": 1.1121812297734629e-05,
      "loss": 0.0753,
      "step": 171460
    },
    {
      "epoch": 66.59,
      "learning_rate": 1.1121294498381879e-05,
      "loss": 0.0674,
      "step": 171470
    },
    {
      "epoch": 66.59,
      "learning_rate": 1.1120776699029128e-05,
      "loss": 0.025,
      "step": 171480
    },
    {
      "epoch": 66.6,
      "learning_rate": 1.1120258899676378e-05,
      "loss": 0.1231,
      "step": 171490
    },
    {
      "epoch": 66.6,
      "learning_rate": 1.1119741100323624e-05,
      "loss": 0.1318,
      "step": 171500
    },
    {
      "epoch": 66.61,
      "learning_rate": 1.1119223300970874e-05,
      "loss": 0.101,
      "step": 171510
    },
    {
      "epoch": 66.61,
      "learning_rate": 1.1118705501618123e-05,
      "loss": 0.2335,
      "step": 171520
    },
    {
      "epoch": 66.61,
      "learning_rate": 1.1118187702265373e-05,
      "loss": 0.0701,
      "step": 171530
    },
    {
      "epoch": 66.62,
      "learning_rate": 1.1117669902912623e-05,
      "loss": 0.0102,
      "step": 171540
    },
    {
      "epoch": 66.62,
      "learning_rate": 1.1117152103559872e-05,
      "loss": 0.1691,
      "step": 171550
    },
    {
      "epoch": 66.63,
      "learning_rate": 1.1116634304207122e-05,
      "loss": 0.1114,
      "step": 171560
    },
    {
      "epoch": 66.63,
      "learning_rate": 1.111611650485437e-05,
      "loss": 0.026,
      "step": 171570
    },
    {
      "epoch": 66.63,
      "learning_rate": 1.111559870550162e-05,
      "loss": 0.1072,
      "step": 171580
    },
    {
      "epoch": 66.64,
      "learning_rate": 1.1115080906148867e-05,
      "loss": 0.1495,
      "step": 171590
    },
    {
      "epoch": 66.64,
      "learning_rate": 1.1114563106796117e-05,
      "loss": 0.0911,
      "step": 171600
    },
    {
      "epoch": 66.64,
      "learning_rate": 1.1114045307443367e-05,
      "loss": 0.0192,
      "step": 171610
    },
    {
      "epoch": 66.65,
      "learning_rate": 1.1113527508090616e-05,
      "loss": 0.0081,
      "step": 171620
    },
    {
      "epoch": 66.65,
      "learning_rate": 1.1113009708737866e-05,
      "loss": 0.0451,
      "step": 171630
    },
    {
      "epoch": 66.66,
      "learning_rate": 1.1112491909385115e-05,
      "loss": 0.0809,
      "step": 171640
    },
    {
      "epoch": 66.66,
      "learning_rate": 1.1111974110032363e-05,
      "loss": 0.0974,
      "step": 171650
    },
    {
      "epoch": 66.66,
      "learning_rate": 1.1111456310679613e-05,
      "loss": 0.017,
      "step": 171660
    },
    {
      "epoch": 66.67,
      "learning_rate": 1.1110938511326861e-05,
      "loss": 0.0308,
      "step": 171670
    },
    {
      "epoch": 66.67,
      "learning_rate": 1.111042071197411e-05,
      "loss": 0.0975,
      "step": 171680
    },
    {
      "epoch": 66.68,
      "learning_rate": 1.110990291262136e-05,
      "loss": 0.0897,
      "step": 171690
    },
    {
      "epoch": 66.68,
      "learning_rate": 1.110938511326861e-05,
      "loss": 0.0152,
      "step": 171700
    },
    {
      "epoch": 66.68,
      "learning_rate": 1.110886731391586e-05,
      "loss": 0.0015,
      "step": 171710
    },
    {
      "epoch": 66.69,
      "learning_rate": 1.1108349514563107e-05,
      "loss": 0.0261,
      "step": 171720
    },
    {
      "epoch": 66.69,
      "learning_rate": 1.1107831715210357e-05,
      "loss": 0.0923,
      "step": 171730
    },
    {
      "epoch": 66.7,
      "learning_rate": 1.1107313915857607e-05,
      "loss": 0.0017,
      "step": 171740
    },
    {
      "epoch": 66.7,
      "learning_rate": 1.1106796116504855e-05,
      "loss": 0.122,
      "step": 171750
    },
    {
      "epoch": 66.7,
      "learning_rate": 1.1106278317152104e-05,
      "loss": 0.0738,
      "step": 171760
    },
    {
      "epoch": 66.71,
      "learning_rate": 1.1105760517799354e-05,
      "loss": 0.0724,
      "step": 171770
    },
    {
      "epoch": 66.71,
      "learning_rate": 1.1105242718446603e-05,
      "loss": 0.2447,
      "step": 171780
    },
    {
      "epoch": 66.71,
      "learning_rate": 1.1104724919093853e-05,
      "loss": 0.0292,
      "step": 171790
    },
    {
      "epoch": 66.72,
      "learning_rate": 1.1104207119741101e-05,
      "loss": 0.1115,
      "step": 171800
    },
    {
      "epoch": 66.72,
      "learning_rate": 1.110368932038835e-05,
      "loss": 0.2066,
      "step": 171810
    },
    {
      "epoch": 66.73,
      "learning_rate": 1.11031715210356e-05,
      "loss": 0.0004,
      "step": 171820
    },
    {
      "epoch": 66.73,
      "learning_rate": 1.1102653721682848e-05,
      "loss": 0.0207,
      "step": 171830
    },
    {
      "epoch": 66.73,
      "learning_rate": 1.1102135922330098e-05,
      "loss": 0.0908,
      "step": 171840
    },
    {
      "epoch": 66.74,
      "learning_rate": 1.1101618122977347e-05,
      "loss": 0.1119,
      "step": 171850
    },
    {
      "epoch": 66.74,
      "learning_rate": 1.1101100323624597e-05,
      "loss": 0.0711,
      "step": 171860
    },
    {
      "epoch": 66.75,
      "learning_rate": 1.1100582524271845e-05,
      "loss": 0.0339,
      "step": 171870
    },
    {
      "epoch": 66.75,
      "learning_rate": 1.1100064724919094e-05,
      "loss": 0.0448,
      "step": 171880
    },
    {
      "epoch": 66.75,
      "learning_rate": 1.1099546925566344e-05,
      "loss": 0.0539,
      "step": 171890
    },
    {
      "epoch": 66.76,
      "learning_rate": 1.1099029126213594e-05,
      "loss": 0.0219,
      "step": 171900
    },
    {
      "epoch": 66.76,
      "learning_rate": 1.1098511326860842e-05,
      "loss": 0.0474,
      "step": 171910
    },
    {
      "epoch": 66.77,
      "learning_rate": 1.1097993527508091e-05,
      "loss": 0.0536,
      "step": 171920
    },
    {
      "epoch": 66.77,
      "learning_rate": 1.1097475728155341e-05,
      "loss": 0.0206,
      "step": 171930
    },
    {
      "epoch": 66.77,
      "learning_rate": 1.109695792880259e-05,
      "loss": 0.0395,
      "step": 171940
    },
    {
      "epoch": 66.78,
      "learning_rate": 1.1096440129449838e-05,
      "loss": 0.0552,
      "step": 171950
    },
    {
      "epoch": 66.78,
      "learning_rate": 1.1095922330097088e-05,
      "loss": 0.0133,
      "step": 171960
    },
    {
      "epoch": 66.78,
      "learning_rate": 1.1095404530744338e-05,
      "loss": 0.1071,
      "step": 171970
    },
    {
      "epoch": 66.79,
      "learning_rate": 1.1094886731391587e-05,
      "loss": 0.1399,
      "step": 171980
    },
    {
      "epoch": 66.79,
      "learning_rate": 1.1094368932038835e-05,
      "loss": 0.0279,
      "step": 171990
    },
    {
      "epoch": 66.8,
      "learning_rate": 1.1093851132686085e-05,
      "loss": 0.0141,
      "step": 172000
    },
    {
      "epoch": 66.8,
      "learning_rate": 1.1093333333333334e-05,
      "loss": 0.0029,
      "step": 172010
    },
    {
      "epoch": 66.8,
      "learning_rate": 1.1092815533980584e-05,
      "loss": 0.1476,
      "step": 172020
    },
    {
      "epoch": 66.81,
      "learning_rate": 1.1092297734627832e-05,
      "loss": 0.098,
      "step": 172030
    },
    {
      "epoch": 66.81,
      "learning_rate": 1.1091779935275082e-05,
      "loss": 0.0175,
      "step": 172040
    },
    {
      "epoch": 66.82,
      "learning_rate": 1.1091262135922331e-05,
      "loss": 0.0053,
      "step": 172050
    },
    {
      "epoch": 66.82,
      "learning_rate": 1.109074433656958e-05,
      "loss": 0.0947,
      "step": 172060
    },
    {
      "epoch": 66.82,
      "learning_rate": 1.1090226537216829e-05,
      "loss": 0.015,
      "step": 172070
    },
    {
      "epoch": 66.83,
      "learning_rate": 1.1089708737864078e-05,
      "loss": 0.0501,
      "step": 172080
    },
    {
      "epoch": 66.83,
      "learning_rate": 1.1089190938511328e-05,
      "loss": 0.186,
      "step": 172090
    },
    {
      "epoch": 66.83,
      "learning_rate": 1.1088673139158576e-05,
      "loss": 0.1847,
      "step": 172100
    },
    {
      "epoch": 66.84,
      "learning_rate": 1.1088155339805826e-05,
      "loss": 0.002,
      "step": 172110
    },
    {
      "epoch": 66.84,
      "learning_rate": 1.1087637540453075e-05,
      "loss": 0.0267,
      "step": 172120
    },
    {
      "epoch": 66.85,
      "learning_rate": 1.1087119741100325e-05,
      "loss": 0.0082,
      "step": 172130
    },
    {
      "epoch": 66.85,
      "learning_rate": 1.1086601941747574e-05,
      "loss": 0.0771,
      "step": 172140
    },
    {
      "epoch": 66.85,
      "learning_rate": 1.1086084142394824e-05,
      "loss": 0.0003,
      "step": 172150
    },
    {
      "epoch": 66.86,
      "learning_rate": 1.1085566343042072e-05,
      "loss": 0.0011,
      "step": 172160
    },
    {
      "epoch": 66.86,
      "learning_rate": 1.1085048543689322e-05,
      "loss": 0.0365,
      "step": 172170
    },
    {
      "epoch": 66.87,
      "learning_rate": 1.108453074433657e-05,
      "loss": 0.0116,
      "step": 172180
    },
    {
      "epoch": 66.87,
      "learning_rate": 1.1084012944983819e-05,
      "loss": 0.0195,
      "step": 172190
    },
    {
      "epoch": 66.87,
      "learning_rate": 1.1083495145631069e-05,
      "loss": 0.1264,
      "step": 172200
    },
    {
      "epoch": 66.88,
      "learning_rate": 1.1082977346278318e-05,
      "loss": 0.1395,
      "step": 172210
    },
    {
      "epoch": 66.88,
      "learning_rate": 1.1082459546925568e-05,
      "loss": 0.0242,
      "step": 172220
    },
    {
      "epoch": 66.89,
      "learning_rate": 1.1081941747572818e-05,
      "loss": 0.1665,
      "step": 172230
    },
    {
      "epoch": 66.89,
      "learning_rate": 1.1081423948220065e-05,
      "loss": 0.076,
      "step": 172240
    },
    {
      "epoch": 66.89,
      "learning_rate": 1.1080906148867313e-05,
      "loss": 0.1804,
      "step": 172250
    },
    {
      "epoch": 66.9,
      "learning_rate": 1.1080388349514563e-05,
      "loss": 0.095,
      "step": 172260
    },
    {
      "epoch": 66.9,
      "learning_rate": 1.1079870550161813e-05,
      "loss": 0.2112,
      "step": 172270
    },
    {
      "epoch": 66.9,
      "learning_rate": 1.1079352750809062e-05,
      "loss": 0.0839,
      "step": 172280
    },
    {
      "epoch": 66.91,
      "learning_rate": 1.1078834951456312e-05,
      "loss": 0.2411,
      "step": 172290
    },
    {
      "epoch": 66.91,
      "learning_rate": 1.1078317152103561e-05,
      "loss": 0.1247,
      "step": 172300
    },
    {
      "epoch": 66.92,
      "learning_rate": 1.1077799352750811e-05,
      "loss": 0.0033,
      "step": 172310
    },
    {
      "epoch": 66.92,
      "learning_rate": 1.1077281553398059e-05,
      "loss": 0.067,
      "step": 172320
    },
    {
      "epoch": 66.92,
      "learning_rate": 1.1076763754045307e-05,
      "loss": 0.1136,
      "step": 172330
    },
    {
      "epoch": 66.93,
      "learning_rate": 1.1076245954692557e-05,
      "loss": 0.1222,
      "step": 172340
    },
    {
      "epoch": 66.93,
      "learning_rate": 1.1075728155339806e-05,
      "loss": 0.0321,
      "step": 172350
    },
    {
      "epoch": 66.94,
      "learning_rate": 1.1075210355987056e-05,
      "loss": 0.1564,
      "step": 172360
    },
    {
      "epoch": 66.94,
      "learning_rate": 1.1074692556634305e-05,
      "loss": 0.0377,
      "step": 172370
    },
    {
      "epoch": 66.94,
      "learning_rate": 1.1074174757281555e-05,
      "loss": 0.0049,
      "step": 172380
    },
    {
      "epoch": 66.95,
      "learning_rate": 1.1073656957928805e-05,
      "loss": 0.1062,
      "step": 172390
    },
    {
      "epoch": 66.95,
      "learning_rate": 1.1073139158576051e-05,
      "loss": 0.0274,
      "step": 172400
    },
    {
      "epoch": 66.96,
      "learning_rate": 1.10726213592233e-05,
      "loss": 0.0274,
      "step": 172410
    },
    {
      "epoch": 66.96,
      "learning_rate": 1.107210355987055e-05,
      "loss": 0.0852,
      "step": 172420
    },
    {
      "epoch": 66.96,
      "learning_rate": 1.10715857605178e-05,
      "loss": 0.0612,
      "step": 172430
    },
    {
      "epoch": 66.97,
      "learning_rate": 1.107106796116505e-05,
      "loss": 0.0247,
      "step": 172440
    },
    {
      "epoch": 66.97,
      "learning_rate": 1.1070550161812299e-05,
      "loss": 0.1255,
      "step": 172450
    },
    {
      "epoch": 66.97,
      "learning_rate": 1.1070032362459549e-05,
      "loss": 0.1527,
      "step": 172460
    },
    {
      "epoch": 66.98,
      "learning_rate": 1.1069514563106798e-05,
      "loss": 0.0109,
      "step": 172470
    },
    {
      "epoch": 66.98,
      "learning_rate": 1.1068996763754044e-05,
      "loss": 0.0006,
      "step": 172480
    },
    {
      "epoch": 66.99,
      "learning_rate": 1.1068478964401294e-05,
      "loss": 0.0131,
      "step": 172490
    },
    {
      "epoch": 66.99,
      "learning_rate": 1.1067961165048544e-05,
      "loss": 0.1492,
      "step": 172500
    },
    {
      "epoch": 66.99,
      "learning_rate": 1.1067443365695793e-05,
      "loss": 0.0331,
      "step": 172510
    },
    {
      "epoch": 67.0,
      "learning_rate": 1.1066925566343043e-05,
      "loss": 0.0705,
      "step": 172520
    },
    {
      "epoch": 67.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.3480420410633087,
      "eval_runtime": 8.2142,
      "eval_samples_per_second": 442.526,
      "eval_steps_per_second": 55.392,
      "step": 172525
    },
    {
      "epoch": 67.0,
      "learning_rate": 1.1066407766990293e-05,
      "loss": 0.0972,
      "step": 172530
    },
    {
      "epoch": 67.01,
      "learning_rate": 1.1065889967637542e-05,
      "loss": 0.0404,
      "step": 172540
    },
    {
      "epoch": 67.01,
      "learning_rate": 1.1065372168284792e-05,
      "loss": 0.0024,
      "step": 172550
    },
    {
      "epoch": 67.01,
      "learning_rate": 1.1064854368932038e-05,
      "loss": 0.0102,
      "step": 172560
    },
    {
      "epoch": 67.02,
      "learning_rate": 1.1064336569579288e-05,
      "loss": 0.0512,
      "step": 172570
    },
    {
      "epoch": 67.02,
      "learning_rate": 1.1063818770226537e-05,
      "loss": 0.0289,
      "step": 172580
    },
    {
      "epoch": 67.03,
      "learning_rate": 1.1063300970873787e-05,
      "loss": 0.0619,
      "step": 172590
    },
    {
      "epoch": 67.03,
      "learning_rate": 1.1062783171521036e-05,
      "loss": 0.0193,
      "step": 172600
    },
    {
      "epoch": 67.03,
      "learning_rate": 1.1062265372168286e-05,
      "loss": 0.0691,
      "step": 172610
    },
    {
      "epoch": 67.04,
      "learning_rate": 1.1061747572815536e-05,
      "loss": 0.0774,
      "step": 172620
    },
    {
      "epoch": 67.04,
      "learning_rate": 1.1061229773462785e-05,
      "loss": 0.2546,
      "step": 172630
    },
    {
      "epoch": 67.04,
      "learning_rate": 1.1060711974110032e-05,
      "loss": 0.1219,
      "step": 172640
    },
    {
      "epoch": 67.05,
      "learning_rate": 1.1060194174757281e-05,
      "loss": 0.0488,
      "step": 172650
    },
    {
      "epoch": 67.05,
      "learning_rate": 1.105967637540453e-05,
      "loss": 0.1328,
      "step": 172660
    },
    {
      "epoch": 67.06,
      "learning_rate": 1.105915857605178e-05,
      "loss": 0.0984,
      "step": 172670
    },
    {
      "epoch": 67.06,
      "learning_rate": 1.105864077669903e-05,
      "loss": 0.0712,
      "step": 172680
    },
    {
      "epoch": 67.06,
      "learning_rate": 1.105812297734628e-05,
      "loss": 0.0644,
      "step": 172690
    },
    {
      "epoch": 67.07,
      "learning_rate": 1.105760517799353e-05,
      "loss": 0.0639,
      "step": 172700
    },
    {
      "epoch": 67.07,
      "learning_rate": 1.1057087378640779e-05,
      "loss": 0.0782,
      "step": 172710
    },
    {
      "epoch": 67.08,
      "learning_rate": 1.1056569579288028e-05,
      "loss": 0.0248,
      "step": 172720
    },
    {
      "epoch": 67.08,
      "learning_rate": 1.1056051779935275e-05,
      "loss": 0.0133,
      "step": 172730
    },
    {
      "epoch": 67.08,
      "learning_rate": 1.1055533980582524e-05,
      "loss": 0.0332,
      "step": 172740
    },
    {
      "epoch": 67.09,
      "learning_rate": 1.1055016181229774e-05,
      "loss": 0.0132,
      "step": 172750
    },
    {
      "epoch": 67.09,
      "learning_rate": 1.1054498381877024e-05,
      "loss": 0.0821,
      "step": 172760
    },
    {
      "epoch": 67.1,
      "learning_rate": 1.1053980582524273e-05,
      "loss": 0.1764,
      "step": 172770
    },
    {
      "epoch": 67.1,
      "learning_rate": 1.1053462783171523e-05,
      "loss": 0.1969,
      "step": 172780
    },
    {
      "epoch": 67.1,
      "learning_rate": 1.1052944983818772e-05,
      "loss": 0.0243,
      "step": 172790
    },
    {
      "epoch": 67.11,
      "learning_rate": 1.1052427184466022e-05,
      "loss": 0.0519,
      "step": 172800
    },
    {
      "epoch": 67.11,
      "learning_rate": 1.1051909385113268e-05,
      "loss": 0.0094,
      "step": 172810
    },
    {
      "epoch": 67.11,
      "learning_rate": 1.1051391585760518e-05,
      "loss": 0.0488,
      "step": 172820
    },
    {
      "epoch": 67.12,
      "learning_rate": 1.1050873786407768e-05,
      "loss": 0.0763,
      "step": 172830
    },
    {
      "epoch": 67.12,
      "learning_rate": 1.1050355987055017e-05,
      "loss": 0.1604,
      "step": 172840
    },
    {
      "epoch": 67.13,
      "learning_rate": 1.1049838187702267e-05,
      "loss": 0.092,
      "step": 172850
    },
    {
      "epoch": 67.13,
      "learning_rate": 1.1049320388349516e-05,
      "loss": 0.001,
      "step": 172860
    },
    {
      "epoch": 67.13,
      "learning_rate": 1.1048802588996766e-05,
      "loss": 0.0004,
      "step": 172870
    },
    {
      "epoch": 67.14,
      "learning_rate": 1.1048284789644016e-05,
      "loss": 0.0261,
      "step": 172880
    },
    {
      "epoch": 67.14,
      "learning_rate": 1.1047766990291262e-05,
      "loss": 0.0539,
      "step": 172890
    },
    {
      "epoch": 67.15,
      "learning_rate": 1.1047249190938511e-05,
      "loss": 0.0009,
      "step": 172900
    },
    {
      "epoch": 67.15,
      "learning_rate": 1.1046731391585761e-05,
      "loss": 0.1068,
      "step": 172910
    },
    {
      "epoch": 67.15,
      "learning_rate": 1.104621359223301e-05,
      "loss": 0.012,
      "step": 172920
    },
    {
      "epoch": 67.16,
      "learning_rate": 1.104569579288026e-05,
      "loss": 0.0639,
      "step": 172930
    },
    {
      "epoch": 67.16,
      "learning_rate": 1.104517799352751e-05,
      "loss": 0.051,
      "step": 172940
    },
    {
      "epoch": 67.17,
      "learning_rate": 1.104466019417476e-05,
      "loss": 0.1596,
      "step": 172950
    },
    {
      "epoch": 67.17,
      "learning_rate": 1.104414239482201e-05,
      "loss": 0.1444,
      "step": 172960
    },
    {
      "epoch": 67.17,
      "learning_rate": 1.1043624595469255e-05,
      "loss": 0.0764,
      "step": 172970
    },
    {
      "epoch": 67.18,
      "learning_rate": 1.1043106796116505e-05,
      "loss": 0.0726,
      "step": 172980
    },
    {
      "epoch": 67.18,
      "learning_rate": 1.1042588996763755e-05,
      "loss": 0.0977,
      "step": 172990
    },
    {
      "epoch": 67.18,
      "learning_rate": 1.1042071197411004e-05,
      "loss": 0.1265,
      "step": 173000
    },
    {
      "epoch": 67.19,
      "learning_rate": 1.1041553398058254e-05,
      "loss": 0.0793,
      "step": 173010
    },
    {
      "epoch": 67.19,
      "learning_rate": 1.1041035598705503e-05,
      "loss": 0.1177,
      "step": 173020
    },
    {
      "epoch": 67.2,
      "learning_rate": 1.1040517799352753e-05,
      "loss": 0.0953,
      "step": 173030
    },
    {
      "epoch": 67.2,
      "learning_rate": 1.1040000000000001e-05,
      "loss": 0.0671,
      "step": 173040
    },
    {
      "epoch": 67.2,
      "learning_rate": 1.1039482200647249e-05,
      "loss": 0.1165,
      "step": 173050
    },
    {
      "epoch": 67.21,
      "learning_rate": 1.1038964401294499e-05,
      "loss": 0.0599,
      "step": 173060
    },
    {
      "epoch": 67.21,
      "learning_rate": 1.1038446601941748e-05,
      "loss": 0.0238,
      "step": 173070
    },
    {
      "epoch": 67.22,
      "learning_rate": 1.1037928802588998e-05,
      "loss": 0.1281,
      "step": 173080
    },
    {
      "epoch": 67.22,
      "learning_rate": 1.1037411003236247e-05,
      "loss": 0.1539,
      "step": 173090
    },
    {
      "epoch": 67.22,
      "learning_rate": 1.1036893203883497e-05,
      "loss": 0.1116,
      "step": 173100
    },
    {
      "epoch": 67.23,
      "learning_rate": 1.1036375404530747e-05,
      "loss": 0.0604,
      "step": 173110
    },
    {
      "epoch": 67.23,
      "learning_rate": 1.1035857605177995e-05,
      "loss": 0.0868,
      "step": 173120
    },
    {
      "epoch": 67.23,
      "learning_rate": 1.1035339805825243e-05,
      "loss": 0.0786,
      "step": 173130
    },
    {
      "epoch": 67.24,
      "learning_rate": 1.1034822006472492e-05,
      "loss": 0.0601,
      "step": 173140
    },
    {
      "epoch": 67.24,
      "learning_rate": 1.1034304207119742e-05,
      "loss": 0.0702,
      "step": 173150
    },
    {
      "epoch": 67.25,
      "learning_rate": 1.1033786407766991e-05,
      "loss": 0.0052,
      "step": 173160
    },
    {
      "epoch": 67.25,
      "learning_rate": 1.1033268608414241e-05,
      "loss": 0.1556,
      "step": 173170
    },
    {
      "epoch": 67.25,
      "learning_rate": 1.103275080906149e-05,
      "loss": 0.1069,
      "step": 173180
    },
    {
      "epoch": 67.26,
      "learning_rate": 1.1032233009708739e-05,
      "loss": 0.055,
      "step": 173190
    },
    {
      "epoch": 67.26,
      "learning_rate": 1.1031715210355988e-05,
      "loss": 0.0427,
      "step": 173200
    },
    {
      "epoch": 67.27,
      "learning_rate": 1.1031197411003236e-05,
      "loss": 0.0285,
      "step": 173210
    },
    {
      "epoch": 67.27,
      "learning_rate": 1.1030679611650486e-05,
      "loss": 0.1133,
      "step": 173220
    },
    {
      "epoch": 67.27,
      "learning_rate": 1.1030161812297735e-05,
      "loss": 0.0929,
      "step": 173230
    },
    {
      "epoch": 67.28,
      "learning_rate": 1.1029644012944985e-05,
      "loss": 0.1059,
      "step": 173240
    },
    {
      "epoch": 67.28,
      "learning_rate": 1.1029126213592235e-05,
      "loss": 0.0419,
      "step": 173250
    },
    {
      "epoch": 67.29,
      "learning_rate": 1.1028608414239484e-05,
      "loss": 0.0102,
      "step": 173260
    },
    {
      "epoch": 67.29,
      "learning_rate": 1.1028090614886732e-05,
      "loss": 0.0928,
      "step": 173270
    },
    {
      "epoch": 67.29,
      "learning_rate": 1.1027572815533982e-05,
      "loss": 0.0689,
      "step": 173280
    },
    {
      "epoch": 67.3,
      "learning_rate": 1.1027055016181231e-05,
      "loss": 0.0609,
      "step": 173290
    },
    {
      "epoch": 67.3,
      "learning_rate": 1.102653721682848e-05,
      "loss": 0.1797,
      "step": 173300
    },
    {
      "epoch": 67.3,
      "learning_rate": 1.1026019417475729e-05,
      "loss": 0.1429,
      "step": 173310
    },
    {
      "epoch": 67.31,
      "learning_rate": 1.1025501618122978e-05,
      "loss": 0.0004,
      "step": 173320
    },
    {
      "epoch": 67.31,
      "learning_rate": 1.1024983818770228e-05,
      "loss": 0.0121,
      "step": 173330
    },
    {
      "epoch": 67.32,
      "learning_rate": 1.1024466019417476e-05,
      "loss": 0.0663,
      "step": 173340
    },
    {
      "epoch": 67.32,
      "learning_rate": 1.1023948220064726e-05,
      "loss": 0.0226,
      "step": 173350
    },
    {
      "epoch": 67.32,
      "learning_rate": 1.1023430420711975e-05,
      "loss": 0.1278,
      "step": 173360
    },
    {
      "epoch": 67.33,
      "learning_rate": 1.1022912621359225e-05,
      "loss": 0.0542,
      "step": 173370
    },
    {
      "epoch": 67.33,
      "learning_rate": 1.1022394822006473e-05,
      "loss": 0.0536,
      "step": 173380
    },
    {
      "epoch": 67.34,
      "learning_rate": 1.1021877022653722e-05,
      "loss": 0.0638,
      "step": 173390
    },
    {
      "epoch": 67.34,
      "learning_rate": 1.1021359223300972e-05,
      "loss": 0.0365,
      "step": 173400
    },
    {
      "epoch": 67.34,
      "learning_rate": 1.1020841423948222e-05,
      "loss": 0.0777,
      "step": 173410
    },
    {
      "epoch": 67.35,
      "learning_rate": 1.102032362459547e-05,
      "loss": 0.082,
      "step": 173420
    },
    {
      "epoch": 67.35,
      "learning_rate": 1.101980582524272e-05,
      "loss": 0.0333,
      "step": 173430
    },
    {
      "epoch": 67.36,
      "learning_rate": 1.1019288025889969e-05,
      "loss": 0.0322,
      "step": 173440
    },
    {
      "epoch": 67.36,
      "learning_rate": 1.1018770226537218e-05,
      "loss": 0.0205,
      "step": 173450
    },
    {
      "epoch": 67.36,
      "learning_rate": 1.1018252427184466e-05,
      "loss": 0.0049,
      "step": 173460
    },
    {
      "epoch": 67.37,
      "learning_rate": 1.1017734627831716e-05,
      "loss": 0.0334,
      "step": 173470
    },
    {
      "epoch": 67.37,
      "learning_rate": 1.1017216828478966e-05,
      "loss": 0.0287,
      "step": 173480
    },
    {
      "epoch": 67.37,
      "learning_rate": 1.1016699029126215e-05,
      "loss": 0.1505,
      "step": 173490
    },
    {
      "epoch": 67.38,
      "learning_rate": 1.1016181229773463e-05,
      "loss": 0.1216,
      "step": 173500
    },
    {
      "epoch": 67.38,
      "learning_rate": 1.1015663430420713e-05,
      "loss": 0.0119,
      "step": 173510
    },
    {
      "epoch": 67.39,
      "learning_rate": 1.1015145631067962e-05,
      "loss": 0.0738,
      "step": 173520
    },
    {
      "epoch": 67.39,
      "learning_rate": 1.1014627831715212e-05,
      "loss": 0.001,
      "step": 173530
    },
    {
      "epoch": 67.39,
      "learning_rate": 1.101411003236246e-05,
      "loss": 0.0871,
      "step": 173540
    },
    {
      "epoch": 67.4,
      "learning_rate": 1.101359223300971e-05,
      "loss": 0.1601,
      "step": 173550
    },
    {
      "epoch": 67.4,
      "learning_rate": 1.1013074433656959e-05,
      "loss": 0.0548,
      "step": 173560
    },
    {
      "epoch": 67.41,
      "learning_rate": 1.1012556634304207e-05,
      "loss": 0.2329,
      "step": 173570
    },
    {
      "epoch": 67.41,
      "learning_rate": 1.1012038834951457e-05,
      "loss": 0.0693,
      "step": 173580
    },
    {
      "epoch": 67.41,
      "learning_rate": 1.1011521035598706e-05,
      "loss": 0.0354,
      "step": 173590
    },
    {
      "epoch": 67.42,
      "learning_rate": 1.1011003236245956e-05,
      "loss": 0.0194,
      "step": 173600
    },
    {
      "epoch": 67.42,
      "learning_rate": 1.1010485436893206e-05,
      "loss": 0.0887,
      "step": 173610
    },
    {
      "epoch": 67.43,
      "learning_rate": 1.1009967637540453e-05,
      "loss": 0.01,
      "step": 173620
    },
    {
      "epoch": 67.43,
      "learning_rate": 1.1009449838187703e-05,
      "loss": 0.0429,
      "step": 173630
    },
    {
      "epoch": 67.43,
      "learning_rate": 1.1008932038834953e-05,
      "loss": 0.0305,
      "step": 173640
    },
    {
      "epoch": 67.44,
      "learning_rate": 1.10084142394822e-05,
      "loss": 0.039,
      "step": 173650
    },
    {
      "epoch": 67.44,
      "learning_rate": 1.100789644012945e-05,
      "loss": 0.1512,
      "step": 173660
    },
    {
      "epoch": 67.44,
      "learning_rate": 1.10073786407767e-05,
      "loss": 0.0337,
      "step": 173670
    },
    {
      "epoch": 67.45,
      "learning_rate": 1.100686084142395e-05,
      "loss": 0.0353,
      "step": 173680
    },
    {
      "epoch": 67.45,
      "learning_rate": 1.1006343042071199e-05,
      "loss": 0.1227,
      "step": 173690
    },
    {
      "epoch": 67.46,
      "learning_rate": 1.1005825242718447e-05,
      "loss": 0.002,
      "step": 173700
    },
    {
      "epoch": 67.46,
      "learning_rate": 1.1005307443365697e-05,
      "loss": 0.0141,
      "step": 173710
    },
    {
      "epoch": 67.46,
      "learning_rate": 1.1004789644012945e-05,
      "loss": 0.1152,
      "step": 173720
    },
    {
      "epoch": 67.47,
      "learning_rate": 1.1004271844660194e-05,
      "loss": 0.0058,
      "step": 173730
    },
    {
      "epoch": 67.47,
      "learning_rate": 1.1003754045307444e-05,
      "loss": 0.0916,
      "step": 173740
    },
    {
      "epoch": 67.48,
      "learning_rate": 1.1003236245954693e-05,
      "loss": 0.0836,
      "step": 173750
    },
    {
      "epoch": 67.48,
      "learning_rate": 1.1002718446601943e-05,
      "loss": 0.0419,
      "step": 173760
    },
    {
      "epoch": 67.48,
      "learning_rate": 1.1002200647249193e-05,
      "loss": 0.0541,
      "step": 173770
    },
    {
      "epoch": 67.49,
      "learning_rate": 1.100168284789644e-05,
      "loss": 0.0057,
      "step": 173780
    },
    {
      "epoch": 67.49,
      "learning_rate": 1.100116504854369e-05,
      "loss": 0.2058,
      "step": 173790
    },
    {
      "epoch": 67.5,
      "learning_rate": 1.1000647249190938e-05,
      "loss": 0.0403,
      "step": 173800
    },
    {
      "epoch": 67.5,
      "learning_rate": 1.1000129449838188e-05,
      "loss": 0.0302,
      "step": 173810
    },
    {
      "epoch": 67.5,
      "learning_rate": 1.0999611650485437e-05,
      "loss": 0.0827,
      "step": 173820
    },
    {
      "epoch": 67.51,
      "learning_rate": 1.0999093851132687e-05,
      "loss": 0.1238,
      "step": 173830
    },
    {
      "epoch": 67.51,
      "learning_rate": 1.0998576051779937e-05,
      "loss": 0.0191,
      "step": 173840
    },
    {
      "epoch": 67.51,
      "learning_rate": 1.0998058252427186e-05,
      "loss": 0.0286,
      "step": 173850
    },
    {
      "epoch": 67.52,
      "learning_rate": 1.0997540453074436e-05,
      "loss": 0.0597,
      "step": 173860
    },
    {
      "epoch": 67.52,
      "learning_rate": 1.0997022653721682e-05,
      "loss": 0.225,
      "step": 173870
    },
    {
      "epoch": 67.53,
      "learning_rate": 1.0996504854368932e-05,
      "loss": 0.0507,
      "step": 173880
    },
    {
      "epoch": 67.53,
      "learning_rate": 1.0995987055016181e-05,
      "loss": 0.122,
      "step": 173890
    },
    {
      "epoch": 67.53,
      "learning_rate": 1.0995469255663431e-05,
      "loss": 0.0625,
      "step": 173900
    },
    {
      "epoch": 67.54,
      "learning_rate": 1.099495145631068e-05,
      "loss": 0.0768,
      "step": 173910
    },
    {
      "epoch": 67.54,
      "learning_rate": 1.099443365695793e-05,
      "loss": 0.1315,
      "step": 173920
    },
    {
      "epoch": 67.55,
      "learning_rate": 1.099391585760518e-05,
      "loss": 0.0007,
      "step": 173930
    },
    {
      "epoch": 67.55,
      "learning_rate": 1.099339805825243e-05,
      "loss": 0.105,
      "step": 173940
    },
    {
      "epoch": 67.55,
      "learning_rate": 1.0992880258899676e-05,
      "loss": 0.1466,
      "step": 173950
    },
    {
      "epoch": 67.56,
      "learning_rate": 1.0992362459546925e-05,
      "loss": 0.0134,
      "step": 173960
    },
    {
      "epoch": 67.56,
      "learning_rate": 1.0991844660194175e-05,
      "loss": 0.2258,
      "step": 173970
    },
    {
      "epoch": 67.57,
      "learning_rate": 1.0991326860841424e-05,
      "loss": 0.0361,
      "step": 173980
    },
    {
      "epoch": 67.57,
      "learning_rate": 1.0990809061488674e-05,
      "loss": 0.0927,
      "step": 173990
    },
    {
      "epoch": 67.57,
      "learning_rate": 1.0990291262135924e-05,
      "loss": 0.0536,
      "step": 174000
    },
    {
      "epoch": 67.58,
      "learning_rate": 1.0989773462783173e-05,
      "loss": 0.1109,
      "step": 174010
    },
    {
      "epoch": 67.58,
      "learning_rate": 1.0989255663430423e-05,
      "loss": 0.1175,
      "step": 174020
    },
    {
      "epoch": 67.58,
      "learning_rate": 1.098873786407767e-05,
      "loss": 0.0653,
      "step": 174030
    },
    {
      "epoch": 67.59,
      "learning_rate": 1.0988220064724919e-05,
      "loss": 0.0895,
      "step": 174040
    },
    {
      "epoch": 67.59,
      "learning_rate": 1.0987702265372168e-05,
      "loss": 0.0161,
      "step": 174050
    },
    {
      "epoch": 67.6,
      "learning_rate": 1.0987184466019418e-05,
      "loss": 0.0266,
      "step": 174060
    },
    {
      "epoch": 67.6,
      "learning_rate": 1.0986666666666668e-05,
      "loss": 0.0368,
      "step": 174070
    },
    {
      "epoch": 67.6,
      "learning_rate": 1.0986148867313917e-05,
      "loss": 0.041,
      "step": 174080
    },
    {
      "epoch": 67.61,
      "learning_rate": 1.0985631067961167e-05,
      "loss": 0.0173,
      "step": 174090
    },
    {
      "epoch": 67.61,
      "learning_rate": 1.0985113268608416e-05,
      "loss": 0.028,
      "step": 174100
    },
    {
      "epoch": 67.62,
      "learning_rate": 1.0984595469255663e-05,
      "loss": 0.105,
      "step": 174110
    },
    {
      "epoch": 67.62,
      "learning_rate": 1.0984077669902912e-05,
      "loss": 0.022,
      "step": 174120
    },
    {
      "epoch": 67.62,
      "learning_rate": 1.0983559870550162e-05,
      "loss": 0.0826,
      "step": 174130
    },
    {
      "epoch": 67.63,
      "learning_rate": 1.0983042071197412e-05,
      "loss": 0.0467,
      "step": 174140
    },
    {
      "epoch": 67.63,
      "learning_rate": 1.0982524271844661e-05,
      "loss": 0.0681,
      "step": 174150
    },
    {
      "epoch": 67.63,
      "learning_rate": 1.098200647249191e-05,
      "loss": 0.0338,
      "step": 174160
    },
    {
      "epoch": 67.64,
      "learning_rate": 1.098148867313916e-05,
      "loss": 0.068,
      "step": 174170
    },
    {
      "epoch": 67.64,
      "learning_rate": 1.098097087378641e-05,
      "loss": 0.0117,
      "step": 174180
    },
    {
      "epoch": 67.65,
      "learning_rate": 1.0980453074433656e-05,
      "loss": 0.0271,
      "step": 174190
    },
    {
      "epoch": 67.65,
      "learning_rate": 1.0979935275080906e-05,
      "loss": 0.0645,
      "step": 174200
    },
    {
      "epoch": 67.65,
      "learning_rate": 1.0979417475728156e-05,
      "loss": 0.0871,
      "step": 174210
    },
    {
      "epoch": 67.66,
      "learning_rate": 1.0978899676375405e-05,
      "loss": 0.0719,
      "step": 174220
    },
    {
      "epoch": 67.66,
      "learning_rate": 1.0978381877022655e-05,
      "loss": 0.0986,
      "step": 174230
    },
    {
      "epoch": 67.67,
      "learning_rate": 1.0977864077669904e-05,
      "loss": 0.0489,
      "step": 174240
    },
    {
      "epoch": 67.67,
      "learning_rate": 1.0977346278317154e-05,
      "loss": 0.0224,
      "step": 174250
    },
    {
      "epoch": 67.67,
      "learning_rate": 1.0976828478964404e-05,
      "loss": 0.0329,
      "step": 174260
    },
    {
      "epoch": 67.68,
      "learning_rate": 1.097631067961165e-05,
      "loss": 0.0007,
      "step": 174270
    },
    {
      "epoch": 67.68,
      "learning_rate": 1.09757928802589e-05,
      "loss": 0.0655,
      "step": 174280
    },
    {
      "epoch": 67.69,
      "learning_rate": 1.0975275080906149e-05,
      "loss": 0.1875,
      "step": 174290
    },
    {
      "epoch": 67.69,
      "learning_rate": 1.0974757281553399e-05,
      "loss": 0.0012,
      "step": 174300
    },
    {
      "epoch": 67.69,
      "learning_rate": 1.0974239482200648e-05,
      "loss": 0.0186,
      "step": 174310
    },
    {
      "epoch": 67.7,
      "learning_rate": 1.0973721682847898e-05,
      "loss": 0.0825,
      "step": 174320
    },
    {
      "epoch": 67.7,
      "learning_rate": 1.0973203883495148e-05,
      "loss": 0.0991,
      "step": 174330
    },
    {
      "epoch": 67.7,
      "learning_rate": 1.0972686084142397e-05,
      "loss": 0.1659,
      "step": 174340
    },
    {
      "epoch": 67.71,
      "learning_rate": 1.0972168284789647e-05,
      "loss": 0.0492,
      "step": 174350
    },
    {
      "epoch": 67.71,
      "learning_rate": 1.0971650485436893e-05,
      "loss": 0.1596,
      "step": 174360
    },
    {
      "epoch": 67.72,
      "learning_rate": 1.0971132686084143e-05,
      "loss": 0.0243,
      "step": 174370
    },
    {
      "epoch": 67.72,
      "learning_rate": 1.0970614886731392e-05,
      "loss": 0.0065,
      "step": 174380
    },
    {
      "epoch": 67.72,
      "learning_rate": 1.0970097087378642e-05,
      "loss": 0.1524,
      "step": 174390
    },
    {
      "epoch": 67.73,
      "learning_rate": 1.0969579288025891e-05,
      "loss": 0.1073,
      "step": 174400
    },
    {
      "epoch": 67.73,
      "learning_rate": 1.0969061488673141e-05,
      "loss": 0.1084,
      "step": 174410
    },
    {
      "epoch": 67.74,
      "learning_rate": 1.096854368932039e-05,
      "loss": 0.0105,
      "step": 174420
    },
    {
      "epoch": 67.74,
      "learning_rate": 1.096802588996764e-05,
      "loss": 0.0454,
      "step": 174430
    },
    {
      "epoch": 67.74,
      "learning_rate": 1.0967508090614887e-05,
      "loss": 0.0047,
      "step": 174440
    },
    {
      "epoch": 67.75,
      "learning_rate": 1.0966990291262136e-05,
      "loss": 0.0076,
      "step": 174450
    },
    {
      "epoch": 67.75,
      "learning_rate": 1.0966472491909386e-05,
      "loss": 0.0769,
      "step": 174460
    },
    {
      "epoch": 67.76,
      "learning_rate": 1.0965954692556635e-05,
      "loss": 0.0381,
      "step": 174470
    },
    {
      "epoch": 67.76,
      "learning_rate": 1.0965436893203885e-05,
      "loss": 0.0514,
      "step": 174480
    },
    {
      "epoch": 67.76,
      "learning_rate": 1.0964919093851135e-05,
      "loss": 0.0331,
      "step": 174490
    },
    {
      "epoch": 67.77,
      "learning_rate": 1.0964401294498384e-05,
      "loss": 0.0526,
      "step": 174500
    },
    {
      "epoch": 67.77,
      "learning_rate": 1.0963883495145632e-05,
      "loss": 0.0555,
      "step": 174510
    },
    {
      "epoch": 67.77,
      "learning_rate": 1.096336569579288e-05,
      "loss": 0.0015,
      "step": 174520
    },
    {
      "epoch": 67.78,
      "learning_rate": 1.096284789644013e-05,
      "loss": 0.1176,
      "step": 174530
    },
    {
      "epoch": 67.78,
      "learning_rate": 1.096233009708738e-05,
      "loss": 0.0623,
      "step": 174540
    },
    {
      "epoch": 67.79,
      "learning_rate": 1.0961812297734629e-05,
      "loss": 0.0521,
      "step": 174550
    },
    {
      "epoch": 67.79,
      "learning_rate": 1.0961294498381879e-05,
      "loss": 0.0197,
      "step": 174560
    },
    {
      "epoch": 67.79,
      "learning_rate": 1.0960776699029128e-05,
      "loss": 0.1512,
      "step": 174570
    },
    {
      "epoch": 67.8,
      "learning_rate": 1.0960258899676378e-05,
      "loss": 0.1197,
      "step": 174580
    },
    {
      "epoch": 67.8,
      "learning_rate": 1.0959741100323626e-05,
      "loss": 0.0872,
      "step": 174590
    },
    {
      "epoch": 67.81,
      "learning_rate": 1.0959223300970874e-05,
      "loss": 0.0945,
      "step": 174600
    },
    {
      "epoch": 67.81,
      "learning_rate": 1.0958705501618123e-05,
      "loss": 0.2249,
      "step": 174610
    },
    {
      "epoch": 67.81,
      "learning_rate": 1.0958187702265373e-05,
      "loss": 0.0177,
      "step": 174620
    },
    {
      "epoch": 67.82,
      "learning_rate": 1.0957669902912623e-05,
      "loss": 0.0366,
      "step": 174630
    },
    {
      "epoch": 67.82,
      "learning_rate": 1.0957152103559872e-05,
      "loss": 0.0075,
      "step": 174640
    },
    {
      "epoch": 67.83,
      "learning_rate": 1.0956634304207122e-05,
      "loss": 0.0324,
      "step": 174650
    },
    {
      "epoch": 67.83,
      "learning_rate": 1.095611650485437e-05,
      "loss": 0.0184,
      "step": 174660
    },
    {
      "epoch": 67.83,
      "learning_rate": 1.095559870550162e-05,
      "loss": 0.12,
      "step": 174670
    },
    {
      "epoch": 67.84,
      "learning_rate": 1.0955080906148867e-05,
      "loss": 0.1216,
      "step": 174680
    },
    {
      "epoch": 67.84,
      "learning_rate": 1.0954563106796117e-05,
      "loss": 0.1556,
      "step": 174690
    },
    {
      "epoch": 67.84,
      "learning_rate": 1.0954045307443366e-05,
      "loss": 0.1248,
      "step": 174700
    },
    {
      "epoch": 67.85,
      "learning_rate": 1.0953527508090616e-05,
      "loss": 0.1028,
      "step": 174710
    },
    {
      "epoch": 67.85,
      "learning_rate": 1.0953009708737866e-05,
      "loss": 0.1275,
      "step": 174720
    },
    {
      "epoch": 67.86,
      "learning_rate": 1.0952491909385115e-05,
      "loss": 0.0037,
      "step": 174730
    },
    {
      "epoch": 67.86,
      "learning_rate": 1.0951974110032363e-05,
      "loss": 0.1086,
      "step": 174740
    },
    {
      "epoch": 67.86,
      "learning_rate": 1.0951456310679613e-05,
      "loss": 0.0905,
      "step": 174750
    },
    {
      "epoch": 67.87,
      "learning_rate": 1.095093851132686e-05,
      "loss": 0.0035,
      "step": 174760
    },
    {
      "epoch": 67.87,
      "learning_rate": 1.095042071197411e-05,
      "loss": 0.1989,
      "step": 174770
    },
    {
      "epoch": 67.88,
      "learning_rate": 1.094990291262136e-05,
      "loss": 0.172,
      "step": 174780
    },
    {
      "epoch": 67.88,
      "learning_rate": 1.094938511326861e-05,
      "loss": 0.0627,
      "step": 174790
    },
    {
      "epoch": 67.88,
      "learning_rate": 1.094886731391586e-05,
      "loss": 0.0894,
      "step": 174800
    },
    {
      "epoch": 67.89,
      "learning_rate": 1.0948349514563107e-05,
      "loss": 0.0966,
      "step": 174810
    },
    {
      "epoch": 67.89,
      "learning_rate": 1.0947831715210357e-05,
      "loss": 0.0185,
      "step": 174820
    },
    {
      "epoch": 67.9,
      "learning_rate": 1.0947313915857606e-05,
      "loss": 0.0361,
      "step": 174830
    },
    {
      "epoch": 67.9,
      "learning_rate": 1.0946796116504854e-05,
      "loss": 0.0787,
      "step": 174840
    },
    {
      "epoch": 67.9,
      "learning_rate": 1.0946278317152104e-05,
      "loss": 0.0844,
      "step": 174850
    },
    {
      "epoch": 67.91,
      "learning_rate": 1.0945760517799354e-05,
      "loss": 0.0022,
      "step": 174860
    },
    {
      "epoch": 67.91,
      "learning_rate": 1.0945242718446603e-05,
      "loss": 0.0022,
      "step": 174870
    },
    {
      "epoch": 67.91,
      "learning_rate": 1.0944724919093853e-05,
      "loss": 0.0693,
      "step": 174880
    },
    {
      "epoch": 67.92,
      "learning_rate": 1.09442071197411e-05,
      "loss": 0.0325,
      "step": 174890
    },
    {
      "epoch": 67.92,
      "learning_rate": 1.094368932038835e-05,
      "loss": 0.0149,
      "step": 174900
    },
    {
      "epoch": 67.93,
      "learning_rate": 1.09431715210356e-05,
      "loss": 0.0519,
      "step": 174910
    },
    {
      "epoch": 67.93,
      "learning_rate": 1.094265372168285e-05,
      "loss": 0.0177,
      "step": 174920
    },
    {
      "epoch": 67.93,
      "learning_rate": 1.0942135922330098e-05,
      "loss": 0.1525,
      "step": 174930
    },
    {
      "epoch": 67.94,
      "learning_rate": 1.0941618122977347e-05,
      "loss": 0.0111,
      "step": 174940
    },
    {
      "epoch": 67.94,
      "learning_rate": 1.0941100323624597e-05,
      "loss": 0.0507,
      "step": 174950
    },
    {
      "epoch": 67.95,
      "learning_rate": 1.0940582524271846e-05,
      "loss": 0.0704,
      "step": 174960
    },
    {
      "epoch": 67.95,
      "learning_rate": 1.0940064724919094e-05,
      "loss": 0.0076,
      "step": 174970
    },
    {
      "epoch": 67.95,
      "learning_rate": 1.0939546925566344e-05,
      "loss": 0.043,
      "step": 174980
    },
    {
      "epoch": 67.96,
      "learning_rate": 1.0939029126213594e-05,
      "loss": 0.0614,
      "step": 174990
    },
    {
      "epoch": 67.96,
      "learning_rate": 1.0938511326860843e-05,
      "loss": 0.0258,
      "step": 175000
    },
    {
      "epoch": 67.97,
      "learning_rate": 1.0937993527508091e-05,
      "loss": 0.0131,
      "step": 175010
    },
    {
      "epoch": 67.97,
      "learning_rate": 1.093747572815534e-05,
      "loss": 0.1103,
      "step": 175020
    },
    {
      "epoch": 67.97,
      "learning_rate": 1.093695792880259e-05,
      "loss": 0.0285,
      "step": 175030
    },
    {
      "epoch": 67.98,
      "learning_rate": 1.0936440129449838e-05,
      "loss": 0.1335,
      "step": 175040
    },
    {
      "epoch": 67.98,
      "learning_rate": 1.0935922330097088e-05,
      "loss": 0.217,
      "step": 175050
    },
    {
      "epoch": 67.98,
      "learning_rate": 1.0935404530744337e-05,
      "loss": 0.0311,
      "step": 175060
    },
    {
      "epoch": 67.99,
      "learning_rate": 1.0934886731391587e-05,
      "loss": 0.0021,
      "step": 175070
    },
    {
      "epoch": 67.99,
      "learning_rate": 1.0934368932038837e-05,
      "loss": 0.0594,
      "step": 175080
    },
    {
      "epoch": 68.0,
      "learning_rate": 1.0933851132686085e-05,
      "loss": 0.1347,
      "step": 175090
    },
    {
      "epoch": 68.0,
      "learning_rate": 1.0933333333333334e-05,
      "loss": 0.094,
      "step": 175100
    },
    {
      "epoch": 68.0,
      "eval_accuracy": 0.949656121045392,
      "eval_loss": 0.3507804572582245,
      "eval_runtime": 8.2022,
      "eval_samples_per_second": 443.174,
      "eval_steps_per_second": 55.473,
      "step": 175100
    },
    {
      "epoch": 68.0,
      "learning_rate": 1.0932815533980584e-05,
      "loss": 0.1268,
      "step": 175110
    },
    {
      "epoch": 68.01,
      "learning_rate": 1.0932297734627832e-05,
      "loss": 0.023,
      "step": 175120
    },
    {
      "epoch": 68.01,
      "learning_rate": 1.0931779935275081e-05,
      "loss": 0.0272,
      "step": 175130
    },
    {
      "epoch": 68.02,
      "learning_rate": 1.0931262135922331e-05,
      "loss": 0.0036,
      "step": 175140
    },
    {
      "epoch": 68.02,
      "learning_rate": 1.093074433656958e-05,
      "loss": 0.0312,
      "step": 175150
    },
    {
      "epoch": 68.02,
      "learning_rate": 1.093022653721683e-05,
      "loss": 0.0873,
      "step": 175160
    },
    {
      "epoch": 68.03,
      "learning_rate": 1.0929708737864078e-05,
      "loss": 0.0882,
      "step": 175170
    },
    {
      "epoch": 68.03,
      "learning_rate": 1.0929190938511328e-05,
      "loss": 0.0077,
      "step": 175180
    },
    {
      "epoch": 68.03,
      "learning_rate": 1.0928673139158576e-05,
      "loss": 0.2152,
      "step": 175190
    },
    {
      "epoch": 68.04,
      "learning_rate": 1.0928155339805825e-05,
      "loss": 0.0742,
      "step": 175200
    },
    {
      "epoch": 68.04,
      "learning_rate": 1.0927637540453075e-05,
      "loss": 0.0224,
      "step": 175210
    },
    {
      "epoch": 68.05,
      "learning_rate": 1.0927119741100325e-05,
      "loss": 0.0432,
      "step": 175220
    },
    {
      "epoch": 68.05,
      "learning_rate": 1.0926601941747574e-05,
      "loss": 0.1055,
      "step": 175230
    },
    {
      "epoch": 68.05,
      "learning_rate": 1.0926084142394824e-05,
      "loss": 0.2127,
      "step": 175240
    },
    {
      "epoch": 68.06,
      "learning_rate": 1.0925566343042072e-05,
      "loss": 0.0594,
      "step": 175250
    },
    {
      "epoch": 68.06,
      "learning_rate": 1.0925048543689321e-05,
      "loss": 0.0412,
      "step": 175260
    },
    {
      "epoch": 68.07,
      "learning_rate": 1.092453074433657e-05,
      "loss": 0.1809,
      "step": 175270
    },
    {
      "epoch": 68.07,
      "learning_rate": 1.0924012944983819e-05,
      "loss": 0.0103,
      "step": 175280
    },
    {
      "epoch": 68.07,
      "learning_rate": 1.0923495145631069e-05,
      "loss": 0.0009,
      "step": 175290
    },
    {
      "epoch": 68.08,
      "learning_rate": 1.0922977346278318e-05,
      "loss": 0.0182,
      "step": 175300
    },
    {
      "epoch": 68.08,
      "learning_rate": 1.0922459546925568e-05,
      "loss": 0.0814,
      "step": 175310
    },
    {
      "epoch": 68.09,
      "learning_rate": 1.0921941747572817e-05,
      "loss": 0.0982,
      "step": 175320
    },
    {
      "epoch": 68.09,
      "learning_rate": 1.0921423948220065e-05,
      "loss": 0.1108,
      "step": 175330
    },
    {
      "epoch": 68.09,
      "learning_rate": 1.0920906148867313e-05,
      "loss": 0.0066,
      "step": 175340
    },
    {
      "epoch": 68.1,
      "learning_rate": 1.0920388349514563e-05,
      "loss": 0.1458,
      "step": 175350
    },
    {
      "epoch": 68.1,
      "learning_rate": 1.0919870550161812e-05,
      "loss": 0.0225,
      "step": 175360
    },
    {
      "epoch": 68.1,
      "learning_rate": 1.0919352750809062e-05,
      "loss": 0.1302,
      "step": 175370
    },
    {
      "epoch": 68.11,
      "learning_rate": 1.0918834951456312e-05,
      "loss": 0.0604,
      "step": 175380
    },
    {
      "epoch": 68.11,
      "learning_rate": 1.0918317152103561e-05,
      "loss": 0.0152,
      "step": 175390
    },
    {
      "epoch": 68.12,
      "learning_rate": 1.0917799352750811e-05,
      "loss": 0.0382,
      "step": 175400
    },
    {
      "epoch": 68.12,
      "learning_rate": 1.0917281553398059e-05,
      "loss": 0.0351,
      "step": 175410
    },
    {
      "epoch": 68.12,
      "learning_rate": 1.0916763754045307e-05,
      "loss": 0.0026,
      "step": 175420
    },
    {
      "epoch": 68.13,
      "learning_rate": 1.0916245954692556e-05,
      "loss": 0.0184,
      "step": 175430
    },
    {
      "epoch": 68.13,
      "learning_rate": 1.0915728155339806e-05,
      "loss": 0.0446,
      "step": 175440
    },
    {
      "epoch": 68.14,
      "learning_rate": 1.0915210355987056e-05,
      "loss": 0.0398,
      "step": 175450
    },
    {
      "epoch": 68.14,
      "learning_rate": 1.0914692556634305e-05,
      "loss": 0.1502,
      "step": 175460
    },
    {
      "epoch": 68.14,
      "learning_rate": 1.0914174757281555e-05,
      "loss": 0.0603,
      "step": 175470
    },
    {
      "epoch": 68.15,
      "learning_rate": 1.0913656957928804e-05,
      "loss": 0.0744,
      "step": 175480
    },
    {
      "epoch": 68.15,
      "learning_rate": 1.0913139158576054e-05,
      "loss": 0.0476,
      "step": 175490
    },
    {
      "epoch": 68.16,
      "learning_rate": 1.09126213592233e-05,
      "loss": 0.1489,
      "step": 175500
    },
    {
      "epoch": 68.16,
      "learning_rate": 1.091210355987055e-05,
      "loss": 0.0398,
      "step": 175510
    },
    {
      "epoch": 68.16,
      "learning_rate": 1.09115857605178e-05,
      "loss": 0.0674,
      "step": 175520
    },
    {
      "epoch": 68.17,
      "learning_rate": 1.091106796116505e-05,
      "loss": 0.0382,
      "step": 175530
    },
    {
      "epoch": 68.17,
      "learning_rate": 1.0910550161812299e-05,
      "loss": 0.0106,
      "step": 175540
    },
    {
      "epoch": 68.17,
      "learning_rate": 1.0910032362459548e-05,
      "loss": 0.1437,
      "step": 175550
    },
    {
      "epoch": 68.18,
      "learning_rate": 1.0909514563106798e-05,
      "loss": 0.0534,
      "step": 175560
    },
    {
      "epoch": 68.18,
      "learning_rate": 1.0908996763754048e-05,
      "loss": 0.1641,
      "step": 175570
    },
    {
      "epoch": 68.19,
      "learning_rate": 1.0908478964401294e-05,
      "loss": 0.0682,
      "step": 175580
    },
    {
      "epoch": 68.19,
      "learning_rate": 1.0907961165048544e-05,
      "loss": 0.0419,
      "step": 175590
    },
    {
      "epoch": 68.19,
      "learning_rate": 1.0907443365695793e-05,
      "loss": 0.0015,
      "step": 175600
    },
    {
      "epoch": 68.2,
      "learning_rate": 1.0906925566343043e-05,
      "loss": 0.0253,
      "step": 175610
    },
    {
      "epoch": 68.2,
      "learning_rate": 1.0906407766990292e-05,
      "loss": 0.0112,
      "step": 175620
    },
    {
      "epoch": 68.21,
      "learning_rate": 1.0905889967637542e-05,
      "loss": 0.0184,
      "step": 175630
    },
    {
      "epoch": 68.21,
      "learning_rate": 1.0905372168284792e-05,
      "loss": 0.1615,
      "step": 175640
    },
    {
      "epoch": 68.21,
      "learning_rate": 1.0904854368932041e-05,
      "loss": 0.0835,
      "step": 175650
    },
    {
      "epoch": 68.22,
      "learning_rate": 1.0904336569579287e-05,
      "loss": 0.0785,
      "step": 175660
    },
    {
      "epoch": 68.22,
      "learning_rate": 1.0903818770226537e-05,
      "loss": 0.1597,
      "step": 175670
    },
    {
      "epoch": 68.23,
      "learning_rate": 1.0903300970873787e-05,
      "loss": 0.0104,
      "step": 175680
    },
    {
      "epoch": 68.23,
      "learning_rate": 1.0902783171521036e-05,
      "loss": 0.0179,
      "step": 175690
    },
    {
      "epoch": 68.23,
      "learning_rate": 1.0902265372168286e-05,
      "loss": 0.0391,
      "step": 175700
    },
    {
      "epoch": 68.24,
      "learning_rate": 1.0901747572815536e-05,
      "loss": 0.1177,
      "step": 175710
    },
    {
      "epoch": 68.24,
      "learning_rate": 1.0901229773462785e-05,
      "loss": 0.0068,
      "step": 175720
    },
    {
      "epoch": 68.24,
      "learning_rate": 1.0900711974110035e-05,
      "loss": 0.0249,
      "step": 175730
    },
    {
      "epoch": 68.25,
      "learning_rate": 1.0900194174757281e-05,
      "loss": 0.0003,
      "step": 175740
    },
    {
      "epoch": 68.25,
      "learning_rate": 1.089967637540453e-05,
      "loss": 0.0835,
      "step": 175750
    },
    {
      "epoch": 68.26,
      "learning_rate": 1.089915857605178e-05,
      "loss": 0.2499,
      "step": 175760
    },
    {
      "epoch": 68.26,
      "learning_rate": 1.089864077669903e-05,
      "loss": 0.0268,
      "step": 175770
    },
    {
      "epoch": 68.26,
      "learning_rate": 1.089812297734628e-05,
      "loss": 0.04,
      "step": 175780
    },
    {
      "epoch": 68.27,
      "learning_rate": 1.0897605177993529e-05,
      "loss": 0.0974,
      "step": 175790
    },
    {
      "epoch": 68.27,
      "learning_rate": 1.0897087378640779e-05,
      "loss": 0.0668,
      "step": 175800
    },
    {
      "epoch": 68.28,
      "learning_rate": 1.0896569579288028e-05,
      "loss": 0.014,
      "step": 175810
    },
    {
      "epoch": 68.28,
      "learning_rate": 1.0896051779935275e-05,
      "loss": 0.1303,
      "step": 175820
    },
    {
      "epoch": 68.28,
      "learning_rate": 1.0895533980582524e-05,
      "loss": 0.0618,
      "step": 175830
    },
    {
      "epoch": 68.29,
      "learning_rate": 1.0895016181229774e-05,
      "loss": 0.1218,
      "step": 175840
    },
    {
      "epoch": 68.29,
      "learning_rate": 1.0894498381877023e-05,
      "loss": 0.0725,
      "step": 175850
    },
    {
      "epoch": 68.3,
      "learning_rate": 1.0893980582524273e-05,
      "loss": 0.0596,
      "step": 175860
    },
    {
      "epoch": 68.3,
      "learning_rate": 1.0893462783171523e-05,
      "loss": 0.0105,
      "step": 175870
    },
    {
      "epoch": 68.3,
      "learning_rate": 1.0892944983818772e-05,
      "loss": 0.1382,
      "step": 175880
    },
    {
      "epoch": 68.31,
      "learning_rate": 1.0892427184466022e-05,
      "loss": 0.0886,
      "step": 175890
    },
    {
      "epoch": 68.31,
      "learning_rate": 1.0891909385113268e-05,
      "loss": 0.0085,
      "step": 175900
    },
    {
      "epoch": 68.31,
      "learning_rate": 1.0891391585760518e-05,
      "loss": 0.1796,
      "step": 175910
    },
    {
      "epoch": 68.32,
      "learning_rate": 1.0890873786407767e-05,
      "loss": 0.0245,
      "step": 175920
    },
    {
      "epoch": 68.32,
      "learning_rate": 1.0890355987055017e-05,
      "loss": 0.0148,
      "step": 175930
    },
    {
      "epoch": 68.33,
      "learning_rate": 1.0889838187702267e-05,
      "loss": 0.1071,
      "step": 175940
    },
    {
      "epoch": 68.33,
      "learning_rate": 1.0889320388349516e-05,
      "loss": 0.0517,
      "step": 175950
    },
    {
      "epoch": 68.33,
      "learning_rate": 1.0888802588996766e-05,
      "loss": 0.0274,
      "step": 175960
    },
    {
      "epoch": 68.34,
      "learning_rate": 1.0888284789644015e-05,
      "loss": 0.0223,
      "step": 175970
    },
    {
      "epoch": 68.34,
      "learning_rate": 1.0887766990291262e-05,
      "loss": 0.0003,
      "step": 175980
    },
    {
      "epoch": 68.35,
      "learning_rate": 1.0887249190938511e-05,
      "loss": 0.0831,
      "step": 175990
    },
    {
      "epoch": 68.35,
      "learning_rate": 1.0886731391585761e-05,
      "loss": 0.0306,
      "step": 176000
    },
    {
      "epoch": 68.35,
      "learning_rate": 1.088621359223301e-05,
      "loss": 0.1181,
      "step": 176010
    },
    {
      "epoch": 68.36,
      "learning_rate": 1.088569579288026e-05,
      "loss": 0.1923,
      "step": 176020
    },
    {
      "epoch": 68.36,
      "learning_rate": 1.088517799352751e-05,
      "loss": 0.0445,
      "step": 176030
    },
    {
      "epoch": 68.37,
      "learning_rate": 1.088466019417476e-05,
      "loss": 0.0224,
      "step": 176040
    },
    {
      "epoch": 68.37,
      "learning_rate": 1.0884142394822009e-05,
      "loss": 0.2267,
      "step": 176050
    },
    {
      "epoch": 68.37,
      "learning_rate": 1.0883624595469257e-05,
      "loss": 0.12,
      "step": 176060
    },
    {
      "epoch": 68.38,
      "learning_rate": 1.0883106796116505e-05,
      "loss": 0.0119,
      "step": 176070
    },
    {
      "epoch": 68.38,
      "learning_rate": 1.0882588996763754e-05,
      "loss": 0.1754,
      "step": 176080
    },
    {
      "epoch": 68.38,
      "learning_rate": 1.0882071197411004e-05,
      "loss": 0.0499,
      "step": 176090
    },
    {
      "epoch": 68.39,
      "learning_rate": 1.0881553398058254e-05,
      "loss": 0.0342,
      "step": 176100
    },
    {
      "epoch": 68.39,
      "learning_rate": 1.0881035598705503e-05,
      "loss": 0.16,
      "step": 176110
    },
    {
      "epoch": 68.4,
      "learning_rate": 1.0880517799352753e-05,
      "loss": 0.1582,
      "step": 176120
    },
    {
      "epoch": 68.4,
      "learning_rate": 1.0880000000000001e-05,
      "loss": 0.1044,
      "step": 176130
    },
    {
      "epoch": 68.4,
      "learning_rate": 1.087948220064725e-05,
      "loss": 0.1002,
      "step": 176140
    },
    {
      "epoch": 68.41,
      "learning_rate": 1.0878964401294498e-05,
      "loss": 0.113,
      "step": 176150
    },
    {
      "epoch": 68.41,
      "learning_rate": 1.0878446601941748e-05,
      "loss": 0.1205,
      "step": 176160
    },
    {
      "epoch": 68.42,
      "learning_rate": 1.0877928802588998e-05,
      "loss": 0.0524,
      "step": 176170
    },
    {
      "epoch": 68.42,
      "learning_rate": 1.0877411003236247e-05,
      "loss": 0.0024,
      "step": 176180
    },
    {
      "epoch": 68.42,
      "learning_rate": 1.0876893203883497e-05,
      "loss": 0.0171,
      "step": 176190
    },
    {
      "epoch": 68.43,
      "learning_rate": 1.0876375404530746e-05,
      "loss": 0.1116,
      "step": 176200
    },
    {
      "epoch": 68.43,
      "learning_rate": 1.0875857605177994e-05,
      "loss": 0.1023,
      "step": 176210
    },
    {
      "epoch": 68.43,
      "learning_rate": 1.0875339805825244e-05,
      "loss": 0.006,
      "step": 176220
    },
    {
      "epoch": 68.44,
      "learning_rate": 1.0874822006472492e-05,
      "loss": 0.0794,
      "step": 176230
    },
    {
      "epoch": 68.44,
      "learning_rate": 1.0874304207119742e-05,
      "loss": 0.0415,
      "step": 176240
    },
    {
      "epoch": 68.45,
      "learning_rate": 1.0873786407766991e-05,
      "loss": 0.0509,
      "step": 176250
    },
    {
      "epoch": 68.45,
      "learning_rate": 1.087326860841424e-05,
      "loss": 0.0555,
      "step": 176260
    },
    {
      "epoch": 68.45,
      "learning_rate": 1.087275080906149e-05,
      "loss": 0.0165,
      "step": 176270
    },
    {
      "epoch": 68.46,
      "learning_rate": 1.0872233009708738e-05,
      "loss": 0.1446,
      "step": 176280
    },
    {
      "epoch": 68.46,
      "learning_rate": 1.0871715210355988e-05,
      "loss": 0.0943,
      "step": 176290
    },
    {
      "epoch": 68.47,
      "learning_rate": 1.0871197411003238e-05,
      "loss": 0.1089,
      "step": 176300
    },
    {
      "epoch": 68.47,
      "learning_rate": 1.0870679611650486e-05,
      "loss": 0.0963,
      "step": 176310
    },
    {
      "epoch": 68.47,
      "learning_rate": 1.0870161812297735e-05,
      "loss": 0.1373,
      "step": 176320
    },
    {
      "epoch": 68.48,
      "learning_rate": 1.0869644012944985e-05,
      "loss": 0.0358,
      "step": 176330
    },
    {
      "epoch": 68.48,
      "learning_rate": 1.0869126213592234e-05,
      "loss": 0.0195,
      "step": 176340
    },
    {
      "epoch": 68.49,
      "learning_rate": 1.0868608414239484e-05,
      "loss": 0.0173,
      "step": 176350
    },
    {
      "epoch": 68.49,
      "learning_rate": 1.0868090614886732e-05,
      "loss": 0.097,
      "step": 176360
    },
    {
      "epoch": 68.49,
      "learning_rate": 1.0867572815533982e-05,
      "loss": 0.037,
      "step": 176370
    },
    {
      "epoch": 68.5,
      "learning_rate": 1.0867055016181231e-05,
      "loss": 0.1286,
      "step": 176380
    },
    {
      "epoch": 68.5,
      "learning_rate": 1.0866537216828479e-05,
      "loss": 0.0545,
      "step": 176390
    },
    {
      "epoch": 68.5,
      "learning_rate": 1.0866019417475729e-05,
      "loss": 0.0145,
      "step": 176400
    },
    {
      "epoch": 68.51,
      "learning_rate": 1.0865501618122978e-05,
      "loss": 0.0301,
      "step": 176410
    },
    {
      "epoch": 68.51,
      "learning_rate": 1.0864983818770228e-05,
      "loss": 0.0827,
      "step": 176420
    },
    {
      "epoch": 68.52,
      "learning_rate": 1.0864466019417476e-05,
      "loss": 0.0323,
      "step": 176430
    },
    {
      "epoch": 68.52,
      "learning_rate": 1.0863948220064725e-05,
      "loss": 0.0335,
      "step": 176440
    },
    {
      "epoch": 68.52,
      "learning_rate": 1.0863430420711975e-05,
      "loss": 0.0515,
      "step": 176450
    },
    {
      "epoch": 68.53,
      "learning_rate": 1.0862912621359225e-05,
      "loss": 0.0087,
      "step": 176460
    },
    {
      "epoch": 68.53,
      "learning_rate": 1.0862394822006473e-05,
      "loss": 0.0513,
      "step": 176470
    },
    {
      "epoch": 68.54,
      "learning_rate": 1.0861877022653722e-05,
      "loss": 0.0603,
      "step": 176480
    },
    {
      "epoch": 68.54,
      "learning_rate": 1.0861359223300972e-05,
      "loss": 0.0283,
      "step": 176490
    },
    {
      "epoch": 68.54,
      "learning_rate": 1.0860841423948221e-05,
      "loss": 0.0754,
      "step": 176500
    },
    {
      "epoch": 68.55,
      "learning_rate": 1.086032362459547e-05,
      "loss": 0.1206,
      "step": 176510
    },
    {
      "epoch": 68.55,
      "learning_rate": 1.0859805825242719e-05,
      "loss": 0.0577,
      "step": 176520
    },
    {
      "epoch": 68.56,
      "learning_rate": 1.0859288025889969e-05,
      "loss": 0.1396,
      "step": 176530
    },
    {
      "epoch": 68.56,
      "learning_rate": 1.0858770226537218e-05,
      "loss": 0.0952,
      "step": 176540
    },
    {
      "epoch": 68.56,
      "learning_rate": 1.0858252427184466e-05,
      "loss": 0.0532,
      "step": 176550
    },
    {
      "epoch": 68.57,
      "learning_rate": 1.0857734627831716e-05,
      "loss": 0.2163,
      "step": 176560
    },
    {
      "epoch": 68.57,
      "learning_rate": 1.0857216828478965e-05,
      "loss": 0.0643,
      "step": 176570
    },
    {
      "epoch": 68.57,
      "learning_rate": 1.0856699029126215e-05,
      "loss": 0.127,
      "step": 176580
    },
    {
      "epoch": 68.58,
      "learning_rate": 1.0856181229773463e-05,
      "loss": 0.0464,
      "step": 176590
    },
    {
      "epoch": 68.58,
      "learning_rate": 1.0855663430420713e-05,
      "loss": 0.0584,
      "step": 176600
    },
    {
      "epoch": 68.59,
      "learning_rate": 1.0855145631067962e-05,
      "loss": 0.0207,
      "step": 176610
    },
    {
      "epoch": 68.59,
      "learning_rate": 1.0854627831715212e-05,
      "loss": 0.0003,
      "step": 176620
    },
    {
      "epoch": 68.59,
      "learning_rate": 1.0854110032362461e-05,
      "loss": 0.027,
      "step": 176630
    },
    {
      "epoch": 68.6,
      "learning_rate": 1.085359223300971e-05,
      "loss": 0.0681,
      "step": 176640
    },
    {
      "epoch": 68.6,
      "learning_rate": 1.0853074433656959e-05,
      "loss": 0.1082,
      "step": 176650
    },
    {
      "epoch": 68.61,
      "learning_rate": 1.0852556634304207e-05,
      "loss": 0.0022,
      "step": 176660
    },
    {
      "epoch": 68.61,
      "learning_rate": 1.0852038834951457e-05,
      "loss": 0.1848,
      "step": 176670
    },
    {
      "epoch": 68.61,
      "learning_rate": 1.0851521035598706e-05,
      "loss": 0.1273,
      "step": 176680
    },
    {
      "epoch": 68.62,
      "learning_rate": 1.0851003236245956e-05,
      "loss": 0.0643,
      "step": 176690
    },
    {
      "epoch": 68.62,
      "learning_rate": 1.0850485436893205e-05,
      "loss": 0.0011,
      "step": 176700
    },
    {
      "epoch": 68.63,
      "learning_rate": 1.0849967637540455e-05,
      "loss": 0.0879,
      "step": 176710
    },
    {
      "epoch": 68.63,
      "learning_rate": 1.0849449838187703e-05,
      "loss": 0.0869,
      "step": 176720
    },
    {
      "epoch": 68.63,
      "learning_rate": 1.0848932038834953e-05,
      "loss": 0.0364,
      "step": 176730
    },
    {
      "epoch": 68.64,
      "learning_rate": 1.08484142394822e-05,
      "loss": 0.1076,
      "step": 176740
    },
    {
      "epoch": 68.64,
      "learning_rate": 1.084789644012945e-05,
      "loss": 0.0308,
      "step": 176750
    },
    {
      "epoch": 68.64,
      "learning_rate": 1.08473786407767e-05,
      "loss": 0.0212,
      "step": 176760
    },
    {
      "epoch": 68.65,
      "learning_rate": 1.084686084142395e-05,
      "loss": 0.0497,
      "step": 176770
    },
    {
      "epoch": 68.65,
      "learning_rate": 1.0846343042071199e-05,
      "loss": 0.0671,
      "step": 176780
    },
    {
      "epoch": 68.66,
      "learning_rate": 1.0845825242718449e-05,
      "loss": 0.0386,
      "step": 176790
    },
    {
      "epoch": 68.66,
      "learning_rate": 1.0845307443365696e-05,
      "loss": 0.0953,
      "step": 176800
    },
    {
      "epoch": 68.66,
      "learning_rate": 1.0844789644012944e-05,
      "loss": 0.1142,
      "step": 176810
    },
    {
      "epoch": 68.67,
      "learning_rate": 1.0844271844660194e-05,
      "loss": 0.0146,
      "step": 176820
    },
    {
      "epoch": 68.67,
      "learning_rate": 1.0843754045307444e-05,
      "loss": 0.0471,
      "step": 176830
    },
    {
      "epoch": 68.68,
      "learning_rate": 1.0843236245954693e-05,
      "loss": 0.0383,
      "step": 176840
    },
    {
      "epoch": 68.68,
      "learning_rate": 1.0842718446601943e-05,
      "loss": 0.2526,
      "step": 176850
    },
    {
      "epoch": 68.68,
      "learning_rate": 1.0842200647249192e-05,
      "loss": 0.0973,
      "step": 176860
    },
    {
      "epoch": 68.69,
      "learning_rate": 1.0841682847896442e-05,
      "loss": 0.0801,
      "step": 176870
    },
    {
      "epoch": 68.69,
      "learning_rate": 1.084116504854369e-05,
      "loss": 0.0289,
      "step": 176880
    },
    {
      "epoch": 68.7,
      "learning_rate": 1.0840647249190938e-05,
      "loss": 0.0101,
      "step": 176890
    },
    {
      "epoch": 68.7,
      "learning_rate": 1.0840129449838188e-05,
      "loss": 0.029,
      "step": 176900
    },
    {
      "epoch": 68.7,
      "learning_rate": 1.0839611650485437e-05,
      "loss": 0.1888,
      "step": 176910
    },
    {
      "epoch": 68.71,
      "learning_rate": 1.0839093851132687e-05,
      "loss": 0.1239,
      "step": 176920
    },
    {
      "epoch": 68.71,
      "learning_rate": 1.0838576051779936e-05,
      "loss": 0.1638,
      "step": 176930
    },
    {
      "epoch": 68.71,
      "learning_rate": 1.0838058252427186e-05,
      "loss": 0.1342,
      "step": 176940
    },
    {
      "epoch": 68.72,
      "learning_rate": 1.0837540453074436e-05,
      "loss": 0.0663,
      "step": 176950
    },
    {
      "epoch": 68.72,
      "learning_rate": 1.0837022653721682e-05,
      "loss": 0.1194,
      "step": 176960
    },
    {
      "epoch": 68.73,
      "learning_rate": 1.0836504854368932e-05,
      "loss": 0.107,
      "step": 176970
    },
    {
      "epoch": 68.73,
      "learning_rate": 1.0835987055016181e-05,
      "loss": 0.1353,
      "step": 176980
    },
    {
      "epoch": 68.73,
      "learning_rate": 1.083546925566343e-05,
      "loss": 0.0566,
      "step": 176990
    },
    {
      "epoch": 68.74,
      "learning_rate": 1.083495145631068e-05,
      "loss": 0.0216,
      "step": 177000
    },
    {
      "epoch": 68.74,
      "learning_rate": 1.083443365695793e-05,
      "loss": 0.0767,
      "step": 177010
    },
    {
      "epoch": 68.75,
      "learning_rate": 1.083391585760518e-05,
      "loss": 0.0287,
      "step": 177020
    },
    {
      "epoch": 68.75,
      "learning_rate": 1.083339805825243e-05,
      "loss": 0.0925,
      "step": 177030
    },
    {
      "epoch": 68.75,
      "learning_rate": 1.0832880258899675e-05,
      "loss": 0.0998,
      "step": 177040
    },
    {
      "epoch": 68.76,
      "learning_rate": 1.0832362459546925e-05,
      "loss": 0.1122,
      "step": 177050
    },
    {
      "epoch": 68.76,
      "learning_rate": 1.0831844660194175e-05,
      "loss": 0.1447,
      "step": 177060
    },
    {
      "epoch": 68.77,
      "learning_rate": 1.0831326860841424e-05,
      "loss": 0.1263,
      "step": 177070
    },
    {
      "epoch": 68.77,
      "learning_rate": 1.0830809061488674e-05,
      "loss": 0.1273,
      "step": 177080
    },
    {
      "epoch": 68.77,
      "learning_rate": 1.0830291262135924e-05,
      "loss": 0.0712,
      "step": 177090
    },
    {
      "epoch": 68.78,
      "learning_rate": 1.0829773462783173e-05,
      "loss": 0.1277,
      "step": 177100
    },
    {
      "epoch": 68.78,
      "learning_rate": 1.0829255663430423e-05,
      "loss": 0.0065,
      "step": 177110
    },
    {
      "epoch": 68.78,
      "learning_rate": 1.0828737864077669e-05,
      "loss": 0.0494,
      "step": 177120
    },
    {
      "epoch": 68.79,
      "learning_rate": 1.0828220064724919e-05,
      "loss": 0.0305,
      "step": 177130
    },
    {
      "epoch": 68.79,
      "learning_rate": 1.0827702265372168e-05,
      "loss": 0.0432,
      "step": 177140
    },
    {
      "epoch": 68.8,
      "learning_rate": 1.0827184466019418e-05,
      "loss": 0.0539,
      "step": 177150
    },
    {
      "epoch": 68.8,
      "learning_rate": 1.0826666666666667e-05,
      "loss": 0.0641,
      "step": 177160
    },
    {
      "epoch": 68.8,
      "learning_rate": 1.0826148867313917e-05,
      "loss": 0.0672,
      "step": 177170
    },
    {
      "epoch": 68.81,
      "learning_rate": 1.0825631067961167e-05,
      "loss": 0.0319,
      "step": 177180
    },
    {
      "epoch": 68.81,
      "learning_rate": 1.0825113268608416e-05,
      "loss": 0.002,
      "step": 177190
    },
    {
      "epoch": 68.82,
      "learning_rate": 1.0824595469255666e-05,
      "loss": 0.1161,
      "step": 177200
    },
    {
      "epoch": 68.82,
      "learning_rate": 1.0824077669902912e-05,
      "loss": 0.0952,
      "step": 177210
    },
    {
      "epoch": 68.82,
      "learning_rate": 1.0823559870550162e-05,
      "loss": 0.1003,
      "step": 177220
    },
    {
      "epoch": 68.83,
      "learning_rate": 1.0823042071197411e-05,
      "loss": 0.0415,
      "step": 177230
    },
    {
      "epoch": 68.83,
      "learning_rate": 1.0822524271844661e-05,
      "loss": 0.025,
      "step": 177240
    },
    {
      "epoch": 68.83,
      "learning_rate": 1.082200647249191e-05,
      "loss": 0.0895,
      "step": 177250
    },
    {
      "epoch": 68.84,
      "learning_rate": 1.082148867313916e-05,
      "loss": 0.091,
      "step": 177260
    },
    {
      "epoch": 68.84,
      "learning_rate": 1.082097087378641e-05,
      "loss": 0.0838,
      "step": 177270
    },
    {
      "epoch": 68.85,
      "learning_rate": 1.082045307443366e-05,
      "loss": 0.019,
      "step": 177280
    },
    {
      "epoch": 68.85,
      "learning_rate": 1.0819935275080906e-05,
      "loss": 0.0676,
      "step": 177290
    },
    {
      "epoch": 68.85,
      "learning_rate": 1.0819417475728155e-05,
      "loss": 0.07,
      "step": 177300
    },
    {
      "epoch": 68.86,
      "learning_rate": 1.0818899676375405e-05,
      "loss": 0.117,
      "step": 177310
    },
    {
      "epoch": 68.86,
      "learning_rate": 1.0818381877022655e-05,
      "loss": 0.1152,
      "step": 177320
    },
    {
      "epoch": 68.87,
      "learning_rate": 1.0817864077669904e-05,
      "loss": 0.1645,
      "step": 177330
    },
    {
      "epoch": 68.87,
      "learning_rate": 1.0817346278317154e-05,
      "loss": 0.0091,
      "step": 177340
    },
    {
      "epoch": 68.87,
      "learning_rate": 1.0816828478964403e-05,
      "loss": 0.0685,
      "step": 177350
    },
    {
      "epoch": 68.88,
      "learning_rate": 1.0816310679611653e-05,
      "loss": 0.0752,
      "step": 177360
    },
    {
      "epoch": 68.88,
      "learning_rate": 1.08157928802589e-05,
      "loss": 0.245,
      "step": 177370
    },
    {
      "epoch": 68.89,
      "learning_rate": 1.0815275080906149e-05,
      "loss": 0.0709,
      "step": 177380
    },
    {
      "epoch": 68.89,
      "learning_rate": 1.0814757281553399e-05,
      "loss": 0.0201,
      "step": 177390
    },
    {
      "epoch": 68.89,
      "learning_rate": 1.0814239482200648e-05,
      "loss": 0.0722,
      "step": 177400
    },
    {
      "epoch": 68.9,
      "learning_rate": 1.0813721682847898e-05,
      "loss": 0.0664,
      "step": 177410
    },
    {
      "epoch": 68.9,
      "learning_rate": 1.0813203883495147e-05,
      "loss": 0.1465,
      "step": 177420
    },
    {
      "epoch": 68.9,
      "learning_rate": 1.0812686084142397e-05,
      "loss": 0.0293,
      "step": 177430
    },
    {
      "epoch": 68.91,
      "learning_rate": 1.0812168284789647e-05,
      "loss": 0.001,
      "step": 177440
    },
    {
      "epoch": 68.91,
      "learning_rate": 1.0811650485436893e-05,
      "loss": 0.1311,
      "step": 177450
    },
    {
      "epoch": 68.92,
      "learning_rate": 1.0811132686084142e-05,
      "loss": 0.0317,
      "step": 177460
    },
    {
      "epoch": 68.92,
      "learning_rate": 1.0810614886731392e-05,
      "loss": 0.0285,
      "step": 177470
    },
    {
      "epoch": 68.92,
      "learning_rate": 1.0810097087378642e-05,
      "loss": 0.0709,
      "step": 177480
    },
    {
      "epoch": 68.93,
      "learning_rate": 1.0809579288025891e-05,
      "loss": 0.0025,
      "step": 177490
    },
    {
      "epoch": 68.93,
      "learning_rate": 1.0809061488673141e-05,
      "loss": 0.0072,
      "step": 177500
    },
    {
      "epoch": 68.94,
      "learning_rate": 1.080854368932039e-05,
      "loss": 0.0954,
      "step": 177510
    },
    {
      "epoch": 68.94,
      "learning_rate": 1.080802588996764e-05,
      "loss": 0.1739,
      "step": 177520
    },
    {
      "epoch": 68.94,
      "learning_rate": 1.0807508090614886e-05,
      "loss": 0.028,
      "step": 177530
    },
    {
      "epoch": 68.95,
      "learning_rate": 1.0806990291262136e-05,
      "loss": 0.0358,
      "step": 177540
    },
    {
      "epoch": 68.95,
      "learning_rate": 1.0806472491909386e-05,
      "loss": 0.0596,
      "step": 177550
    },
    {
      "epoch": 68.96,
      "learning_rate": 1.0805954692556635e-05,
      "loss": 0.0302,
      "step": 177560
    },
    {
      "epoch": 68.96,
      "learning_rate": 1.0805436893203885e-05,
      "loss": 0.049,
      "step": 177570
    },
    {
      "epoch": 68.96,
      "learning_rate": 1.0804919093851134e-05,
      "loss": 0.0378,
      "step": 177580
    },
    {
      "epoch": 68.97,
      "learning_rate": 1.0804401294498384e-05,
      "loss": 0.1051,
      "step": 177590
    },
    {
      "epoch": 68.97,
      "learning_rate": 1.0803883495145632e-05,
      "loss": 0.0987,
      "step": 177600
    },
    {
      "epoch": 68.97,
      "learning_rate": 1.080336569579288e-05,
      "loss": 0.1052,
      "step": 177610
    },
    {
      "epoch": 68.98,
      "learning_rate": 1.080284789644013e-05,
      "loss": 0.0704,
      "step": 177620
    },
    {
      "epoch": 68.98,
      "learning_rate": 1.080233009708738e-05,
      "loss": 0.0278,
      "step": 177630
    },
    {
      "epoch": 68.99,
      "learning_rate": 1.0801812297734629e-05,
      "loss": 0.0106,
      "step": 177640
    },
    {
      "epoch": 68.99,
      "learning_rate": 1.0801294498381878e-05,
      "loss": 0.1588,
      "step": 177650
    },
    {
      "epoch": 68.99,
      "learning_rate": 1.0800776699029128e-05,
      "loss": 0.0632,
      "step": 177660
    },
    {
      "epoch": 69.0,
      "learning_rate": 1.0800258899676378e-05,
      "loss": 0.0498,
      "step": 177670
    },
    {
      "epoch": 69.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.33282846212387085,
      "eval_runtime": 8.256,
      "eval_samples_per_second": 440.285,
      "eval_steps_per_second": 55.111,
      "step": 177675
    },
    {
      "epoch": 69.0,
      "learning_rate": 1.0799741100323626e-05,
      "loss": 0.0004,
      "step": 177680
    },
    {
      "epoch": 69.01,
      "learning_rate": 1.0799223300970874e-05,
      "loss": 0.1271,
      "step": 177690
    },
    {
      "epoch": 69.01,
      "learning_rate": 1.0798705501618123e-05,
      "loss": 0.0668,
      "step": 177700
    },
    {
      "epoch": 69.01,
      "learning_rate": 1.0798187702265373e-05,
      "loss": 0.0604,
      "step": 177710
    },
    {
      "epoch": 69.02,
      "learning_rate": 1.0797669902912622e-05,
      "loss": 0.1426,
      "step": 177720
    },
    {
      "epoch": 69.02,
      "learning_rate": 1.0797152103559872e-05,
      "loss": 0.1536,
      "step": 177730
    },
    {
      "epoch": 69.03,
      "learning_rate": 1.0796634304207122e-05,
      "loss": 0.1229,
      "step": 177740
    },
    {
      "epoch": 69.03,
      "learning_rate": 1.079611650485437e-05,
      "loss": 0.0915,
      "step": 177750
    },
    {
      "epoch": 69.03,
      "learning_rate": 1.0795598705501619e-05,
      "loss": 0.045,
      "step": 177760
    },
    {
      "epoch": 69.04,
      "learning_rate": 1.0795080906148869e-05,
      "loss": 0.1046,
      "step": 177770
    },
    {
      "epoch": 69.04,
      "learning_rate": 1.0794563106796117e-05,
      "loss": 0.1974,
      "step": 177780
    },
    {
      "epoch": 69.04,
      "learning_rate": 1.0794045307443366e-05,
      "loss": 0.0502,
      "step": 177790
    },
    {
      "epoch": 69.05,
      "learning_rate": 1.0793527508090616e-05,
      "loss": 0.0079,
      "step": 177800
    },
    {
      "epoch": 69.05,
      "learning_rate": 1.0793009708737866e-05,
      "loss": 0.1,
      "step": 177810
    },
    {
      "epoch": 69.06,
      "learning_rate": 1.0792491909385115e-05,
      "loss": 0.1038,
      "step": 177820
    },
    {
      "epoch": 69.06,
      "learning_rate": 1.0791974110032363e-05,
      "loss": 0.1187,
      "step": 177830
    },
    {
      "epoch": 69.06,
      "learning_rate": 1.0791456310679613e-05,
      "loss": 0.0792,
      "step": 177840
    },
    {
      "epoch": 69.07,
      "learning_rate": 1.0790938511326862e-05,
      "loss": 0.0016,
      "step": 177850
    },
    {
      "epoch": 69.07,
      "learning_rate": 1.079042071197411e-05,
      "loss": 0.0075,
      "step": 177860
    },
    {
      "epoch": 69.08,
      "learning_rate": 1.078990291262136e-05,
      "loss": 0.0954,
      "step": 177870
    },
    {
      "epoch": 69.08,
      "learning_rate": 1.078938511326861e-05,
      "loss": 0.0456,
      "step": 177880
    },
    {
      "epoch": 69.08,
      "learning_rate": 1.0788867313915859e-05,
      "loss": 0.153,
      "step": 177890
    },
    {
      "epoch": 69.09,
      "learning_rate": 1.0788349514563107e-05,
      "loss": 0.0574,
      "step": 177900
    },
    {
      "epoch": 69.09,
      "learning_rate": 1.0787831715210357e-05,
      "loss": 0.0272,
      "step": 177910
    },
    {
      "epoch": 69.1,
      "learning_rate": 1.0787313915857606e-05,
      "loss": 0.1048,
      "step": 177920
    },
    {
      "epoch": 69.1,
      "learning_rate": 1.0786796116504856e-05,
      "loss": 0.1543,
      "step": 177930
    },
    {
      "epoch": 69.1,
      "learning_rate": 1.0786278317152104e-05,
      "loss": 0.2584,
      "step": 177940
    },
    {
      "epoch": 69.11,
      "learning_rate": 1.0785760517799353e-05,
      "loss": 0.103,
      "step": 177950
    },
    {
      "epoch": 69.11,
      "learning_rate": 1.0785242718446603e-05,
      "loss": 0.0818,
      "step": 177960
    },
    {
      "epoch": 69.11,
      "learning_rate": 1.0784724919093853e-05,
      "loss": 0.1342,
      "step": 177970
    },
    {
      "epoch": 69.12,
      "learning_rate": 1.07842071197411e-05,
      "loss": 0.063,
      "step": 177980
    },
    {
      "epoch": 69.12,
      "learning_rate": 1.078368932038835e-05,
      "loss": 0.0202,
      "step": 177990
    },
    {
      "epoch": 69.13,
      "learning_rate": 1.07831715210356e-05,
      "loss": 0.0502,
      "step": 178000
    },
    {
      "epoch": 69.13,
      "learning_rate": 1.078265372168285e-05,
      "loss": 0.041,
      "step": 178010
    },
    {
      "epoch": 69.13,
      "learning_rate": 1.0782135922330097e-05,
      "loss": 0.0574,
      "step": 178020
    },
    {
      "epoch": 69.14,
      "learning_rate": 1.0781618122977347e-05,
      "loss": 0.1061,
      "step": 178030
    },
    {
      "epoch": 69.14,
      "learning_rate": 1.0781100323624597e-05,
      "loss": 0.0853,
      "step": 178040
    },
    {
      "epoch": 69.15,
      "learning_rate": 1.0780582524271846e-05,
      "loss": 0.0372,
      "step": 178050
    },
    {
      "epoch": 69.15,
      "learning_rate": 1.0780064724919094e-05,
      "loss": 0.0567,
      "step": 178060
    },
    {
      "epoch": 69.15,
      "learning_rate": 1.0779546925566344e-05,
      "loss": 0.032,
      "step": 178070
    },
    {
      "epoch": 69.16,
      "learning_rate": 1.0779029126213593e-05,
      "loss": 0.0958,
      "step": 178080
    },
    {
      "epoch": 69.16,
      "learning_rate": 1.0778511326860843e-05,
      "loss": 0.0527,
      "step": 178090
    },
    {
      "epoch": 69.17,
      "learning_rate": 1.0777993527508091e-05,
      "loss": 0.0767,
      "step": 178100
    },
    {
      "epoch": 69.17,
      "learning_rate": 1.077747572815534e-05,
      "loss": 0.0693,
      "step": 178110
    },
    {
      "epoch": 69.17,
      "learning_rate": 1.077695792880259e-05,
      "loss": 0.0463,
      "step": 178120
    },
    {
      "epoch": 69.18,
      "learning_rate": 1.0776440129449838e-05,
      "loss": 0.0519,
      "step": 178130
    },
    {
      "epoch": 69.18,
      "learning_rate": 1.0775922330097088e-05,
      "loss": 0.0942,
      "step": 178140
    },
    {
      "epoch": 69.18,
      "learning_rate": 1.0775404530744337e-05,
      "loss": 0.0674,
      "step": 178150
    },
    {
      "epoch": 69.19,
      "learning_rate": 1.0774886731391587e-05,
      "loss": 0.0598,
      "step": 178160
    },
    {
      "epoch": 69.19,
      "learning_rate": 1.0774368932038837e-05,
      "loss": 0.028,
      "step": 178170
    },
    {
      "epoch": 69.2,
      "learning_rate": 1.0773851132686084e-05,
      "loss": 0.0215,
      "step": 178180
    },
    {
      "epoch": 69.2,
      "learning_rate": 1.0773333333333334e-05,
      "loss": 0.0839,
      "step": 178190
    },
    {
      "epoch": 69.2,
      "learning_rate": 1.0772815533980584e-05,
      "loss": 0.3014,
      "step": 178200
    },
    {
      "epoch": 69.21,
      "learning_rate": 1.0772297734627832e-05,
      "loss": 0.236,
      "step": 178210
    },
    {
      "epoch": 69.21,
      "learning_rate": 1.0771779935275081e-05,
      "loss": 0.0307,
      "step": 178220
    },
    {
      "epoch": 69.22,
      "learning_rate": 1.0771262135922331e-05,
      "loss": 0.0519,
      "step": 178230
    },
    {
      "epoch": 69.22,
      "learning_rate": 1.077074433656958e-05,
      "loss": 0.0687,
      "step": 178240
    },
    {
      "epoch": 69.22,
      "learning_rate": 1.077022653721683e-05,
      "loss": 0.1202,
      "step": 178250
    },
    {
      "epoch": 69.23,
      "learning_rate": 1.0769708737864078e-05,
      "loss": 0.0461,
      "step": 178260
    },
    {
      "epoch": 69.23,
      "learning_rate": 1.0769190938511328e-05,
      "loss": 0.1181,
      "step": 178270
    },
    {
      "epoch": 69.23,
      "learning_rate": 1.0768673139158576e-05,
      "loss": 0.0549,
      "step": 178280
    },
    {
      "epoch": 69.24,
      "learning_rate": 1.0768155339805825e-05,
      "loss": 0.0559,
      "step": 178290
    },
    {
      "epoch": 69.24,
      "learning_rate": 1.0767637540453075e-05,
      "loss": 0.0522,
      "step": 178300
    },
    {
      "epoch": 69.25,
      "learning_rate": 1.0767119741100324e-05,
      "loss": 0.0243,
      "step": 178310
    },
    {
      "epoch": 69.25,
      "learning_rate": 1.0766601941747574e-05,
      "loss": 0.0987,
      "step": 178320
    },
    {
      "epoch": 69.25,
      "learning_rate": 1.0766084142394824e-05,
      "loss": 0.0872,
      "step": 178330
    },
    {
      "epoch": 69.26,
      "learning_rate": 1.0765566343042073e-05,
      "loss": 0.0602,
      "step": 178340
    },
    {
      "epoch": 69.26,
      "learning_rate": 1.0765048543689321e-05,
      "loss": 0.2384,
      "step": 178350
    },
    {
      "epoch": 69.27,
      "learning_rate": 1.0764530744336569e-05,
      "loss": 0.0138,
      "step": 178360
    },
    {
      "epoch": 69.27,
      "learning_rate": 1.0764012944983819e-05,
      "loss": 0.0438,
      "step": 178370
    },
    {
      "epoch": 69.27,
      "learning_rate": 1.0763495145631068e-05,
      "loss": 0.0398,
      "step": 178380
    },
    {
      "epoch": 69.28,
      "learning_rate": 1.0762977346278318e-05,
      "loss": 0.0626,
      "step": 178390
    },
    {
      "epoch": 69.28,
      "learning_rate": 1.0762459546925568e-05,
      "loss": 0.0913,
      "step": 178400
    },
    {
      "epoch": 69.29,
      "learning_rate": 1.0761941747572817e-05,
      "loss": 0.1047,
      "step": 178410
    },
    {
      "epoch": 69.29,
      "learning_rate": 1.0761423948220067e-05,
      "loss": 0.1467,
      "step": 178420
    },
    {
      "epoch": 69.29,
      "learning_rate": 1.0760906148867313e-05,
      "loss": 0.0673,
      "step": 178430
    },
    {
      "epoch": 69.3,
      "learning_rate": 1.0760388349514563e-05,
      "loss": 0.0432,
      "step": 178440
    },
    {
      "epoch": 69.3,
      "learning_rate": 1.0759870550161812e-05,
      "loss": 0.0514,
      "step": 178450
    },
    {
      "epoch": 69.3,
      "learning_rate": 1.0759352750809062e-05,
      "loss": 0.0801,
      "step": 178460
    },
    {
      "epoch": 69.31,
      "learning_rate": 1.0758834951456312e-05,
      "loss": 0.0207,
      "step": 178470
    },
    {
      "epoch": 69.31,
      "learning_rate": 1.0758317152103561e-05,
      "loss": 0.0197,
      "step": 178480
    },
    {
      "epoch": 69.32,
      "learning_rate": 1.075779935275081e-05,
      "loss": 0.0015,
      "step": 178490
    },
    {
      "epoch": 69.32,
      "learning_rate": 1.075728155339806e-05,
      "loss": 0.0506,
      "step": 178500
    },
    {
      "epoch": 69.32,
      "learning_rate": 1.0756763754045307e-05,
      "loss": 0.03,
      "step": 178510
    },
    {
      "epoch": 69.33,
      "learning_rate": 1.0756245954692556e-05,
      "loss": 0.0066,
      "step": 178520
    },
    {
      "epoch": 69.33,
      "learning_rate": 1.0755728155339806e-05,
      "loss": 0.1083,
      "step": 178530
    },
    {
      "epoch": 69.34,
      "learning_rate": 1.0755210355987055e-05,
      "loss": 0.0321,
      "step": 178540
    },
    {
      "epoch": 69.34,
      "learning_rate": 1.0754692556634305e-05,
      "loss": 0.0527,
      "step": 178550
    },
    {
      "epoch": 69.34,
      "learning_rate": 1.0754174757281555e-05,
      "loss": 0.1802,
      "step": 178560
    },
    {
      "epoch": 69.35,
      "learning_rate": 1.0753656957928804e-05,
      "loss": 0.0904,
      "step": 178570
    },
    {
      "epoch": 69.35,
      "learning_rate": 1.0753139158576054e-05,
      "loss": 0.0056,
      "step": 178580
    },
    {
      "epoch": 69.36,
      "learning_rate": 1.07526213592233e-05,
      "loss": 0.0635,
      "step": 178590
    },
    {
      "epoch": 69.36,
      "learning_rate": 1.075210355987055e-05,
      "loss": 0.0423,
      "step": 178600
    },
    {
      "epoch": 69.36,
      "learning_rate": 1.07515857605178e-05,
      "loss": 0.0275,
      "step": 178610
    },
    {
      "epoch": 69.37,
      "learning_rate": 1.0751067961165049e-05,
      "loss": 0.0482,
      "step": 178620
    },
    {
      "epoch": 69.37,
      "learning_rate": 1.0750550161812299e-05,
      "loss": 0.0796,
      "step": 178630
    },
    {
      "epoch": 69.37,
      "learning_rate": 1.0750032362459548e-05,
      "loss": 0.0749,
      "step": 178640
    },
    {
      "epoch": 69.38,
      "learning_rate": 1.0749514563106798e-05,
      "loss": 0.0885,
      "step": 178650
    },
    {
      "epoch": 69.38,
      "learning_rate": 1.0748996763754048e-05,
      "loss": 0.1116,
      "step": 178660
    },
    {
      "epoch": 69.39,
      "learning_rate": 1.0748478964401294e-05,
      "loss": 0.0176,
      "step": 178670
    },
    {
      "epoch": 69.39,
      "learning_rate": 1.0747961165048543e-05,
      "loss": 0.0166,
      "step": 178680
    },
    {
      "epoch": 69.39,
      "learning_rate": 1.0747443365695793e-05,
      "loss": 0.0036,
      "step": 178690
    },
    {
      "epoch": 69.4,
      "learning_rate": 1.0746925566343043e-05,
      "loss": 0.0031,
      "step": 178700
    },
    {
      "epoch": 69.4,
      "learning_rate": 1.0746407766990292e-05,
      "loss": 0.0665,
      "step": 178710
    },
    {
      "epoch": 69.41,
      "learning_rate": 1.0745889967637542e-05,
      "loss": 0.05,
      "step": 178720
    },
    {
      "epoch": 69.41,
      "learning_rate": 1.0745372168284791e-05,
      "loss": 0.0988,
      "step": 178730
    },
    {
      "epoch": 69.41,
      "learning_rate": 1.0744854368932041e-05,
      "loss": 0.1728,
      "step": 178740
    },
    {
      "epoch": 69.42,
      "learning_rate": 1.0744336569579287e-05,
      "loss": 0.0007,
      "step": 178750
    },
    {
      "epoch": 69.42,
      "learning_rate": 1.0743818770226537e-05,
      "loss": 0.0871,
      "step": 178760
    },
    {
      "epoch": 69.43,
      "learning_rate": 1.0743300970873787e-05,
      "loss": 0.041,
      "step": 178770
    },
    {
      "epoch": 69.43,
      "learning_rate": 1.0742783171521036e-05,
      "loss": 0.1185,
      "step": 178780
    },
    {
      "epoch": 69.43,
      "learning_rate": 1.0742265372168286e-05,
      "loss": 0.0338,
      "step": 178790
    },
    {
      "epoch": 69.44,
      "learning_rate": 1.0741747572815535e-05,
      "loss": 0.0459,
      "step": 178800
    },
    {
      "epoch": 69.44,
      "learning_rate": 1.0741229773462785e-05,
      "loss": 0.0442,
      "step": 178810
    },
    {
      "epoch": 69.44,
      "learning_rate": 1.0740711974110035e-05,
      "loss": 0.0116,
      "step": 178820
    },
    {
      "epoch": 69.45,
      "learning_rate": 1.0740194174757284e-05,
      "loss": 0.0857,
      "step": 178830
    },
    {
      "epoch": 69.45,
      "learning_rate": 1.073967637540453e-05,
      "loss": 0.0857,
      "step": 178840
    },
    {
      "epoch": 69.46,
      "learning_rate": 1.073915857605178e-05,
      "loss": 0.0068,
      "step": 178850
    },
    {
      "epoch": 69.46,
      "learning_rate": 1.073864077669903e-05,
      "loss": 0.0911,
      "step": 178860
    },
    {
      "epoch": 69.46,
      "learning_rate": 1.073812297734628e-05,
      "loss": 0.0927,
      "step": 178870
    },
    {
      "epoch": 69.47,
      "learning_rate": 1.0737605177993529e-05,
      "loss": 0.0317,
      "step": 178880
    },
    {
      "epoch": 69.47,
      "learning_rate": 1.0737087378640779e-05,
      "loss": 0.0077,
      "step": 178890
    },
    {
      "epoch": 69.48,
      "learning_rate": 1.0736569579288028e-05,
      "loss": 0.0222,
      "step": 178900
    },
    {
      "epoch": 69.48,
      "learning_rate": 1.0736051779935278e-05,
      "loss": 0.0424,
      "step": 178910
    },
    {
      "epoch": 69.48,
      "learning_rate": 1.0735533980582524e-05,
      "loss": 0.0542,
      "step": 178920
    },
    {
      "epoch": 69.49,
      "learning_rate": 1.0735016181229774e-05,
      "loss": 0.0371,
      "step": 178930
    },
    {
      "epoch": 69.49,
      "learning_rate": 1.0734498381877023e-05,
      "loss": 0.0099,
      "step": 178940
    },
    {
      "epoch": 69.5,
      "learning_rate": 1.0733980582524273e-05,
      "loss": 0.0036,
      "step": 178950
    },
    {
      "epoch": 69.5,
      "learning_rate": 1.0733462783171522e-05,
      "loss": 0.0405,
      "step": 178960
    },
    {
      "epoch": 69.5,
      "learning_rate": 1.0732944983818772e-05,
      "loss": 0.0015,
      "step": 178970
    },
    {
      "epoch": 69.51,
      "learning_rate": 1.0732427184466022e-05,
      "loss": 0.0234,
      "step": 178980
    },
    {
      "epoch": 69.51,
      "learning_rate": 1.0731909385113271e-05,
      "loss": 0.1003,
      "step": 178990
    },
    {
      "epoch": 69.51,
      "learning_rate": 1.0731391585760518e-05,
      "loss": 0.0623,
      "step": 179000
    },
    {
      "epoch": 69.52,
      "learning_rate": 1.0730873786407767e-05,
      "loss": 0.0625,
      "step": 179010
    },
    {
      "epoch": 69.52,
      "learning_rate": 1.0730355987055017e-05,
      "loss": 0.023,
      "step": 179020
    },
    {
      "epoch": 69.53,
      "learning_rate": 1.0729838187702266e-05,
      "loss": 0.0514,
      "step": 179030
    },
    {
      "epoch": 69.53,
      "learning_rate": 1.0729320388349516e-05,
      "loss": 0.0327,
      "step": 179040
    },
    {
      "epoch": 69.53,
      "learning_rate": 1.0728802588996766e-05,
      "loss": 0.0703,
      "step": 179050
    },
    {
      "epoch": 69.54,
      "learning_rate": 1.0728284789644015e-05,
      "loss": 0.1096,
      "step": 179060
    },
    {
      "epoch": 69.54,
      "learning_rate": 1.0727766990291263e-05,
      "loss": 0.0177,
      "step": 179070
    },
    {
      "epoch": 69.55,
      "learning_rate": 1.0727249190938511e-05,
      "loss": 0.0192,
      "step": 179080
    },
    {
      "epoch": 69.55,
      "learning_rate": 1.072673139158576e-05,
      "loss": 0.0768,
      "step": 179090
    },
    {
      "epoch": 69.55,
      "learning_rate": 1.072621359223301e-05,
      "loss": 0.0889,
      "step": 179100
    },
    {
      "epoch": 69.56,
      "learning_rate": 1.072569579288026e-05,
      "loss": 0.1635,
      "step": 179110
    },
    {
      "epoch": 69.56,
      "learning_rate": 1.072517799352751e-05,
      "loss": 0.0406,
      "step": 179120
    },
    {
      "epoch": 69.57,
      "learning_rate": 1.072466019417476e-05,
      "loss": 0.015,
      "step": 179130
    },
    {
      "epoch": 69.57,
      "learning_rate": 1.0724142394822009e-05,
      "loss": 0.0315,
      "step": 179140
    },
    {
      "epoch": 69.57,
      "learning_rate": 1.0723624595469257e-05,
      "loss": 0.0311,
      "step": 179150
    },
    {
      "epoch": 69.58,
      "learning_rate": 1.0723106796116505e-05,
      "loss": 0.1396,
      "step": 179160
    },
    {
      "epoch": 69.58,
      "learning_rate": 1.0722588996763754e-05,
      "loss": 0.0037,
      "step": 179170
    },
    {
      "epoch": 69.58,
      "learning_rate": 1.0722071197411004e-05,
      "loss": 0.0996,
      "step": 179180
    },
    {
      "epoch": 69.59,
      "learning_rate": 1.0721553398058254e-05,
      "loss": 0.1059,
      "step": 179190
    },
    {
      "epoch": 69.59,
      "learning_rate": 1.0721035598705503e-05,
      "loss": 0.0001,
      "step": 179200
    },
    {
      "epoch": 69.6,
      "learning_rate": 1.0720517799352753e-05,
      "loss": 0.0086,
      "step": 179210
    },
    {
      "epoch": 69.6,
      "learning_rate": 1.072e-05,
      "loss": 0.0029,
      "step": 179220
    },
    {
      "epoch": 69.6,
      "learning_rate": 1.071948220064725e-05,
      "loss": 0.0936,
      "step": 179230
    },
    {
      "epoch": 69.61,
      "learning_rate": 1.0718964401294498e-05,
      "loss": 0.0296,
      "step": 179240
    },
    {
      "epoch": 69.61,
      "learning_rate": 1.0718446601941748e-05,
      "loss": 0.0968,
      "step": 179250
    },
    {
      "epoch": 69.62,
      "learning_rate": 1.0717928802588997e-05,
      "loss": 0.0793,
      "step": 179260
    },
    {
      "epoch": 69.62,
      "learning_rate": 1.0717411003236247e-05,
      "loss": 0.2012,
      "step": 179270
    },
    {
      "epoch": 69.62,
      "learning_rate": 1.0716893203883497e-05,
      "loss": 0.0627,
      "step": 179280
    },
    {
      "epoch": 69.63,
      "learning_rate": 1.0716375404530746e-05,
      "loss": 0.0525,
      "step": 179290
    },
    {
      "epoch": 69.63,
      "learning_rate": 1.0715857605177994e-05,
      "loss": 0.0552,
      "step": 179300
    },
    {
      "epoch": 69.63,
      "learning_rate": 1.0715339805825244e-05,
      "loss": 0.1928,
      "step": 179310
    },
    {
      "epoch": 69.64,
      "learning_rate": 1.0714822006472492e-05,
      "loss": 0.0413,
      "step": 179320
    },
    {
      "epoch": 69.64,
      "learning_rate": 1.0714304207119741e-05,
      "loss": 0.0633,
      "step": 179330
    },
    {
      "epoch": 69.65,
      "learning_rate": 1.0713786407766991e-05,
      "loss": 0.0962,
      "step": 179340
    },
    {
      "epoch": 69.65,
      "learning_rate": 1.071326860841424e-05,
      "loss": 0.1175,
      "step": 179350
    },
    {
      "epoch": 69.65,
      "learning_rate": 1.071275080906149e-05,
      "loss": 0.1767,
      "step": 179360
    },
    {
      "epoch": 69.66,
      "learning_rate": 1.0712233009708738e-05,
      "loss": 0.0007,
      "step": 179370
    },
    {
      "epoch": 69.66,
      "learning_rate": 1.0711715210355988e-05,
      "loss": 0.0738,
      "step": 179380
    },
    {
      "epoch": 69.67,
      "learning_rate": 1.0711197411003237e-05,
      "loss": 0.128,
      "step": 179390
    },
    {
      "epoch": 69.67,
      "learning_rate": 1.0710679611650487e-05,
      "loss": 0.0399,
      "step": 179400
    },
    {
      "epoch": 69.67,
      "learning_rate": 1.0710161812297735e-05,
      "loss": 0.0863,
      "step": 179410
    },
    {
      "epoch": 69.68,
      "learning_rate": 1.0709644012944985e-05,
      "loss": 0.0779,
      "step": 179420
    },
    {
      "epoch": 69.68,
      "learning_rate": 1.0709126213592234e-05,
      "loss": 0.1155,
      "step": 179430
    },
    {
      "epoch": 69.69,
      "learning_rate": 1.0708608414239484e-05,
      "loss": 0.1061,
      "step": 179440
    },
    {
      "epoch": 69.69,
      "learning_rate": 1.0708090614886732e-05,
      "loss": 0.0009,
      "step": 179450
    },
    {
      "epoch": 69.69,
      "learning_rate": 1.0707572815533981e-05,
      "loss": 0.0429,
      "step": 179460
    },
    {
      "epoch": 69.7,
      "learning_rate": 1.0707055016181231e-05,
      "loss": 0.0769,
      "step": 179470
    },
    {
      "epoch": 69.7,
      "learning_rate": 1.070653721682848e-05,
      "loss": 0.0032,
      "step": 179480
    },
    {
      "epoch": 69.7,
      "learning_rate": 1.0706019417475729e-05,
      "loss": 0.1538,
      "step": 179490
    },
    {
      "epoch": 69.71,
      "learning_rate": 1.0705501618122978e-05,
      "loss": 0.2242,
      "step": 179500
    },
    {
      "epoch": 69.71,
      "learning_rate": 1.0704983818770228e-05,
      "loss": 0.0618,
      "step": 179510
    },
    {
      "epoch": 69.72,
      "learning_rate": 1.0704466019417477e-05,
      "loss": 0.0218,
      "step": 179520
    },
    {
      "epoch": 69.72,
      "learning_rate": 1.0703948220064725e-05,
      "loss": 0.0212,
      "step": 179530
    },
    {
      "epoch": 69.72,
      "learning_rate": 1.0703430420711975e-05,
      "loss": 0.049,
      "step": 179540
    },
    {
      "epoch": 69.73,
      "learning_rate": 1.0702912621359225e-05,
      "loss": 0.0338,
      "step": 179550
    },
    {
      "epoch": 69.73,
      "learning_rate": 1.0702394822006474e-05,
      "loss": 0.1114,
      "step": 179560
    },
    {
      "epoch": 69.74,
      "learning_rate": 1.0701877022653722e-05,
      "loss": 0.1169,
      "step": 179570
    },
    {
      "epoch": 69.74,
      "learning_rate": 1.0701359223300972e-05,
      "loss": 0.1006,
      "step": 179580
    },
    {
      "epoch": 69.74,
      "learning_rate": 1.0700841423948221e-05,
      "loss": 0.1587,
      "step": 179590
    },
    {
      "epoch": 69.75,
      "learning_rate": 1.070032362459547e-05,
      "loss": 0.169,
      "step": 179600
    },
    {
      "epoch": 69.75,
      "learning_rate": 1.0699805825242719e-05,
      "loss": 0.0215,
      "step": 179610
    },
    {
      "epoch": 69.76,
      "learning_rate": 1.0699288025889968e-05,
      "loss": 0.0094,
      "step": 179620
    },
    {
      "epoch": 69.76,
      "learning_rate": 1.0698770226537218e-05,
      "loss": 0.1559,
      "step": 179630
    },
    {
      "epoch": 69.76,
      "learning_rate": 1.0698252427184468e-05,
      "loss": 0.0109,
      "step": 179640
    },
    {
      "epoch": 69.77,
      "learning_rate": 1.0697734627831716e-05,
      "loss": 0.1409,
      "step": 179650
    },
    {
      "epoch": 69.77,
      "learning_rate": 1.0697216828478965e-05,
      "loss": 0.1355,
      "step": 179660
    },
    {
      "epoch": 69.77,
      "learning_rate": 1.0696699029126215e-05,
      "loss": 0.0298,
      "step": 179670
    },
    {
      "epoch": 69.78,
      "learning_rate": 1.0696181229773463e-05,
      "loss": 0.1295,
      "step": 179680
    },
    {
      "epoch": 69.78,
      "learning_rate": 1.0695663430420712e-05,
      "loss": 0.0692,
      "step": 179690
    },
    {
      "epoch": 69.79,
      "learning_rate": 1.0695145631067962e-05,
      "loss": 0.1102,
      "step": 179700
    },
    {
      "epoch": 69.79,
      "learning_rate": 1.0694627831715212e-05,
      "loss": 0.0398,
      "step": 179710
    },
    {
      "epoch": 69.79,
      "learning_rate": 1.0694110032362461e-05,
      "loss": 0.0611,
      "step": 179720
    },
    {
      "epoch": 69.8,
      "learning_rate": 1.069359223300971e-05,
      "loss": 0.1066,
      "step": 179730
    },
    {
      "epoch": 69.8,
      "learning_rate": 1.0693074433656959e-05,
      "loss": 0.0552,
      "step": 179740
    },
    {
      "epoch": 69.81,
      "learning_rate": 1.0692556634304207e-05,
      "loss": 0.0577,
      "step": 179750
    },
    {
      "epoch": 69.81,
      "learning_rate": 1.0692038834951456e-05,
      "loss": 0.0602,
      "step": 179760
    },
    {
      "epoch": 69.81,
      "learning_rate": 1.0691521035598706e-05,
      "loss": 0.0356,
      "step": 179770
    },
    {
      "epoch": 69.82,
      "learning_rate": 1.0691003236245956e-05,
      "loss": 0.1042,
      "step": 179780
    },
    {
      "epoch": 69.82,
      "learning_rate": 1.0690485436893205e-05,
      "loss": 0.0514,
      "step": 179790
    },
    {
      "epoch": 69.83,
      "learning_rate": 1.0689967637540455e-05,
      "loss": 0.057,
      "step": 179800
    },
    {
      "epoch": 69.83,
      "learning_rate": 1.0689449838187703e-05,
      "loss": 0.0294,
      "step": 179810
    },
    {
      "epoch": 69.83,
      "learning_rate": 1.0688932038834952e-05,
      "loss": 0.0002,
      "step": 179820
    },
    {
      "epoch": 69.84,
      "learning_rate": 1.06884142394822e-05,
      "loss": 0.0745,
      "step": 179830
    },
    {
      "epoch": 69.84,
      "learning_rate": 1.068789644012945e-05,
      "loss": 0.0042,
      "step": 179840
    },
    {
      "epoch": 69.84,
      "learning_rate": 1.06873786407767e-05,
      "loss": 0.167,
      "step": 179850
    },
    {
      "epoch": 69.85,
      "learning_rate": 1.0686860841423949e-05,
      "loss": 0.009,
      "step": 179860
    },
    {
      "epoch": 69.85,
      "learning_rate": 1.0686343042071199e-05,
      "loss": 0.193,
      "step": 179870
    },
    {
      "epoch": 69.86,
      "learning_rate": 1.0685825242718448e-05,
      "loss": 0.0712,
      "step": 179880
    },
    {
      "epoch": 69.86,
      "learning_rate": 1.0685307443365696e-05,
      "loss": 0.0619,
      "step": 179890
    },
    {
      "epoch": 69.86,
      "learning_rate": 1.0684789644012944e-05,
      "loss": 0.0177,
      "step": 179900
    },
    {
      "epoch": 69.87,
      "learning_rate": 1.0684271844660194e-05,
      "loss": 0.0033,
      "step": 179910
    },
    {
      "epoch": 69.87,
      "learning_rate": 1.0683754045307443e-05,
      "loss": 0.0528,
      "step": 179920
    },
    {
      "epoch": 69.88,
      "learning_rate": 1.0683236245954693e-05,
      "loss": 0.0754,
      "step": 179930
    },
    {
      "epoch": 69.88,
      "learning_rate": 1.0682718446601943e-05,
      "loss": 0.1956,
      "step": 179940
    },
    {
      "epoch": 69.88,
      "learning_rate": 1.0682200647249192e-05,
      "loss": 0.0623,
      "step": 179950
    },
    {
      "epoch": 69.89,
      "learning_rate": 1.0681682847896442e-05,
      "loss": 0.0167,
      "step": 179960
    },
    {
      "epoch": 69.89,
      "learning_rate": 1.0681165048543692e-05,
      "loss": 0.1324,
      "step": 179970
    },
    {
      "epoch": 69.9,
      "learning_rate": 1.0680647249190938e-05,
      "loss": 0.0345,
      "step": 179980
    },
    {
      "epoch": 69.9,
      "learning_rate": 1.0680129449838187e-05,
      "loss": 0.0711,
      "step": 179990
    },
    {
      "epoch": 69.9,
      "learning_rate": 1.0679611650485437e-05,
      "loss": 0.0772,
      "step": 180000
    },
    {
      "epoch": 69.91,
      "learning_rate": 1.0679093851132687e-05,
      "loss": 0.0465,
      "step": 180010
    },
    {
      "epoch": 69.91,
      "learning_rate": 1.0678576051779936e-05,
      "loss": 0.0376,
      "step": 180020
    },
    {
      "epoch": 69.91,
      "learning_rate": 1.0678058252427186e-05,
      "loss": 0.0316,
      "step": 180030
    },
    {
      "epoch": 69.92,
      "learning_rate": 1.0677540453074435e-05,
      "loss": 0.0507,
      "step": 180040
    },
    {
      "epoch": 69.92,
      "learning_rate": 1.0677022653721685e-05,
      "loss": 0.1756,
      "step": 180050
    },
    {
      "epoch": 69.93,
      "learning_rate": 1.0676504854368931e-05,
      "loss": 0.1412,
      "step": 180060
    },
    {
      "epoch": 69.93,
      "learning_rate": 1.0675987055016181e-05,
      "loss": 0.0138,
      "step": 180070
    },
    {
      "epoch": 69.93,
      "learning_rate": 1.067546925566343e-05,
      "loss": 0.0967,
      "step": 180080
    },
    {
      "epoch": 69.94,
      "learning_rate": 1.067495145631068e-05,
      "loss": 0.0026,
      "step": 180090
    },
    {
      "epoch": 69.94,
      "learning_rate": 1.067443365695793e-05,
      "loss": 0.0295,
      "step": 180100
    },
    {
      "epoch": 69.95,
      "learning_rate": 1.067391585760518e-05,
      "loss": 0.0279,
      "step": 180110
    },
    {
      "epoch": 69.95,
      "learning_rate": 1.0673398058252429e-05,
      "loss": 0.0883,
      "step": 180120
    },
    {
      "epoch": 69.95,
      "learning_rate": 1.0672880258899679e-05,
      "loss": 0.0196,
      "step": 180130
    },
    {
      "epoch": 69.96,
      "learning_rate": 1.0672362459546925e-05,
      "loss": 0.0159,
      "step": 180140
    },
    {
      "epoch": 69.96,
      "learning_rate": 1.0671844660194175e-05,
      "loss": 0.0602,
      "step": 180150
    },
    {
      "epoch": 69.97,
      "learning_rate": 1.0671326860841424e-05,
      "loss": 0.0699,
      "step": 180160
    },
    {
      "epoch": 69.97,
      "learning_rate": 1.0670809061488674e-05,
      "loss": 0.1291,
      "step": 180170
    },
    {
      "epoch": 69.97,
      "learning_rate": 1.0670291262135923e-05,
      "loss": 0.0144,
      "step": 180180
    },
    {
      "epoch": 69.98,
      "learning_rate": 1.0669773462783173e-05,
      "loss": 0.0077,
      "step": 180190
    },
    {
      "epoch": 69.98,
      "learning_rate": 1.0669255663430423e-05,
      "loss": 0.1981,
      "step": 180200
    },
    {
      "epoch": 69.98,
      "learning_rate": 1.0668737864077672e-05,
      "loss": 0.0833,
      "step": 180210
    },
    {
      "epoch": 69.99,
      "learning_rate": 1.0668220064724918e-05,
      "loss": 0.1865,
      "step": 180220
    },
    {
      "epoch": 69.99,
      "learning_rate": 1.0667702265372168e-05,
      "loss": 0.005,
      "step": 180230
    },
    {
      "epoch": 70.0,
      "learning_rate": 1.0667184466019418e-05,
      "loss": 0.2284,
      "step": 180240
    },
    {
      "epoch": 70.0,
      "learning_rate": 1.0666666666666667e-05,
      "loss": 0.0535,
      "step": 180250
    },
    {
      "epoch": 70.0,
      "eval_accuracy": 0.9499312242090784,
      "eval_loss": 0.3557586073875427,
      "eval_runtime": 8.2114,
      "eval_samples_per_second": 442.677,
      "eval_steps_per_second": 55.411,
      "step": 180250
    },
    {
      "epoch": 70.0,
      "learning_rate": 1.0666148867313917e-05,
      "loss": 0.0776,
      "step": 180260
    },
    {
      "epoch": 70.01,
      "learning_rate": 1.0665631067961167e-05,
      "loss": 0.0792,
      "step": 180270
    },
    {
      "epoch": 70.01,
      "learning_rate": 1.0665113268608416e-05,
      "loss": 0.0006,
      "step": 180280
    },
    {
      "epoch": 70.02,
      "learning_rate": 1.0664595469255666e-05,
      "loss": 0.0909,
      "step": 180290
    },
    {
      "epoch": 70.02,
      "learning_rate": 1.0664077669902912e-05,
      "loss": 0.044,
      "step": 180300
    },
    {
      "epoch": 70.02,
      "learning_rate": 1.0663559870550162e-05,
      "loss": 0.0509,
      "step": 180310
    },
    {
      "epoch": 70.03,
      "learning_rate": 1.0663042071197411e-05,
      "loss": 0.0522,
      "step": 180320
    },
    {
      "epoch": 70.03,
      "learning_rate": 1.0662524271844661e-05,
      "loss": 0.0662,
      "step": 180330
    },
    {
      "epoch": 70.03,
      "learning_rate": 1.066200647249191e-05,
      "loss": 0.1108,
      "step": 180340
    },
    {
      "epoch": 70.04,
      "learning_rate": 1.066148867313916e-05,
      "loss": 0.0607,
      "step": 180350
    },
    {
      "epoch": 70.04,
      "learning_rate": 1.066097087378641e-05,
      "loss": 0.0527,
      "step": 180360
    },
    {
      "epoch": 70.05,
      "learning_rate": 1.066045307443366e-05,
      "loss": 0.0929,
      "step": 180370
    },
    {
      "epoch": 70.05,
      "learning_rate": 1.0659935275080906e-05,
      "loss": 0.0645,
      "step": 180380
    },
    {
      "epoch": 70.05,
      "learning_rate": 1.0659417475728155e-05,
      "loss": 0.0156,
      "step": 180390
    },
    {
      "epoch": 70.06,
      "learning_rate": 1.0658899676375405e-05,
      "loss": 0.1432,
      "step": 180400
    },
    {
      "epoch": 70.06,
      "learning_rate": 1.0658381877022654e-05,
      "loss": 0.0607,
      "step": 180410
    },
    {
      "epoch": 70.07,
      "learning_rate": 1.0657864077669904e-05,
      "loss": 0.1211,
      "step": 180420
    },
    {
      "epoch": 70.07,
      "learning_rate": 1.0657346278317154e-05,
      "loss": 0.0047,
      "step": 180430
    },
    {
      "epoch": 70.07,
      "learning_rate": 1.0656828478964403e-05,
      "loss": 0.0951,
      "step": 180440
    },
    {
      "epoch": 70.08,
      "learning_rate": 1.0656310679611653e-05,
      "loss": 0.1895,
      "step": 180450
    },
    {
      "epoch": 70.08,
      "learning_rate": 1.0655792880258899e-05,
      "loss": 0.0253,
      "step": 180460
    },
    {
      "epoch": 70.09,
      "learning_rate": 1.0655275080906149e-05,
      "loss": 0.125,
      "step": 180470
    },
    {
      "epoch": 70.09,
      "learning_rate": 1.0654757281553398e-05,
      "loss": 0.0908,
      "step": 180480
    },
    {
      "epoch": 70.09,
      "learning_rate": 1.0654239482200648e-05,
      "loss": 0.0676,
      "step": 180490
    },
    {
      "epoch": 70.1,
      "learning_rate": 1.0653721682847898e-05,
      "loss": 0.1533,
      "step": 180500
    },
    {
      "epoch": 70.1,
      "learning_rate": 1.0653203883495147e-05,
      "loss": 0.0626,
      "step": 180510
    },
    {
      "epoch": 70.1,
      "learning_rate": 1.0652686084142397e-05,
      "loss": 0.0676,
      "step": 180520
    },
    {
      "epoch": 70.11,
      "learning_rate": 1.0652168284789646e-05,
      "loss": 0.0458,
      "step": 180530
    },
    {
      "epoch": 70.11,
      "learning_rate": 1.0651650485436894e-05,
      "loss": 0.0576,
      "step": 180540
    },
    {
      "epoch": 70.12,
      "learning_rate": 1.0651132686084142e-05,
      "loss": 0.0674,
      "step": 180550
    },
    {
      "epoch": 70.12,
      "learning_rate": 1.0650614886731392e-05,
      "loss": 0.0064,
      "step": 180560
    },
    {
      "epoch": 70.12,
      "learning_rate": 1.0650097087378642e-05,
      "loss": 0.0322,
      "step": 180570
    },
    {
      "epoch": 70.13,
      "learning_rate": 1.0649579288025891e-05,
      "loss": 0.1584,
      "step": 180580
    },
    {
      "epoch": 70.13,
      "learning_rate": 1.064906148867314e-05,
      "loss": 0.018,
      "step": 180590
    },
    {
      "epoch": 70.14,
      "learning_rate": 1.064854368932039e-05,
      "loss": 0.0612,
      "step": 180600
    },
    {
      "epoch": 70.14,
      "learning_rate": 1.064802588996764e-05,
      "loss": 0.0768,
      "step": 180610
    },
    {
      "epoch": 70.14,
      "learning_rate": 1.0647508090614888e-05,
      "loss": 0.0011,
      "step": 180620
    },
    {
      "epoch": 70.15,
      "learning_rate": 1.0646990291262136e-05,
      "loss": 0.0725,
      "step": 180630
    },
    {
      "epoch": 70.15,
      "learning_rate": 1.0646472491909385e-05,
      "loss": 0.033,
      "step": 180640
    },
    {
      "epoch": 70.16,
      "learning_rate": 1.0645954692556635e-05,
      "loss": 0.0915,
      "step": 180650
    },
    {
      "epoch": 70.16,
      "learning_rate": 1.0645436893203885e-05,
      "loss": 0.0632,
      "step": 180660
    },
    {
      "epoch": 70.16,
      "learning_rate": 1.0644919093851134e-05,
      "loss": 0.0752,
      "step": 180670
    },
    {
      "epoch": 70.17,
      "learning_rate": 1.0644401294498384e-05,
      "loss": 0.0688,
      "step": 180680
    },
    {
      "epoch": 70.17,
      "learning_rate": 1.0643883495145632e-05,
      "loss": 0.0155,
      "step": 180690
    },
    {
      "epoch": 70.17,
      "learning_rate": 1.0643365695792881e-05,
      "loss": 0.0709,
      "step": 180700
    },
    {
      "epoch": 70.18,
      "learning_rate": 1.064284789644013e-05,
      "loss": 0.1874,
      "step": 180710
    },
    {
      "epoch": 70.18,
      "learning_rate": 1.0642330097087379e-05,
      "loss": 0.0868,
      "step": 180720
    },
    {
      "epoch": 70.19,
      "learning_rate": 1.0641812297734629e-05,
      "loss": 0.1348,
      "step": 180730
    },
    {
      "epoch": 70.19,
      "learning_rate": 1.0641294498381878e-05,
      "loss": 0.039,
      "step": 180740
    },
    {
      "epoch": 70.19,
      "learning_rate": 1.0640776699029128e-05,
      "loss": 0.0343,
      "step": 180750
    },
    {
      "epoch": 70.2,
      "learning_rate": 1.0640258899676378e-05,
      "loss": 0.0025,
      "step": 180760
    },
    {
      "epoch": 70.2,
      "learning_rate": 1.0639741100323625e-05,
      "loss": 0.0929,
      "step": 180770
    },
    {
      "epoch": 70.21,
      "learning_rate": 1.0639223300970875e-05,
      "loss": 0.0931,
      "step": 180780
    },
    {
      "epoch": 70.21,
      "learning_rate": 1.0638705501618123e-05,
      "loss": 0.0076,
      "step": 180790
    },
    {
      "epoch": 70.21,
      "learning_rate": 1.0638187702265373e-05,
      "loss": 0.0563,
      "step": 180800
    },
    {
      "epoch": 70.22,
      "learning_rate": 1.0637669902912622e-05,
      "loss": 0.1776,
      "step": 180810
    },
    {
      "epoch": 70.22,
      "learning_rate": 1.0637152103559872e-05,
      "loss": 0.1674,
      "step": 180820
    },
    {
      "epoch": 70.23,
      "learning_rate": 1.0636634304207121e-05,
      "loss": 0.1463,
      "step": 180830
    },
    {
      "epoch": 70.23,
      "learning_rate": 1.063611650485437e-05,
      "loss": 0.0486,
      "step": 180840
    },
    {
      "epoch": 70.23,
      "learning_rate": 1.0635598705501619e-05,
      "loss": 0.0281,
      "step": 180850
    },
    {
      "epoch": 70.24,
      "learning_rate": 1.0635080906148869e-05,
      "loss": 0.0035,
      "step": 180860
    },
    {
      "epoch": 70.24,
      "learning_rate": 1.0634563106796117e-05,
      "loss": 0.0187,
      "step": 180870
    },
    {
      "epoch": 70.24,
      "learning_rate": 1.0634045307443366e-05,
      "loss": 0.091,
      "step": 180880
    },
    {
      "epoch": 70.25,
      "learning_rate": 1.0633527508090616e-05,
      "loss": 0.0641,
      "step": 180890
    },
    {
      "epoch": 70.25,
      "learning_rate": 1.0633009708737865e-05,
      "loss": 0.1093,
      "step": 180900
    },
    {
      "epoch": 70.26,
      "learning_rate": 1.0632491909385115e-05,
      "loss": 0.1469,
      "step": 180910
    },
    {
      "epoch": 70.26,
      "learning_rate": 1.0631974110032363e-05,
      "loss": 0.0112,
      "step": 180920
    },
    {
      "epoch": 70.26,
      "learning_rate": 1.0631456310679613e-05,
      "loss": 0.0728,
      "step": 180930
    },
    {
      "epoch": 70.27,
      "learning_rate": 1.0630938511326862e-05,
      "loss": 0.015,
      "step": 180940
    },
    {
      "epoch": 70.27,
      "learning_rate": 1.063042071197411e-05,
      "loss": 0.0371,
      "step": 180950
    },
    {
      "epoch": 70.28,
      "learning_rate": 1.062990291262136e-05,
      "loss": 0.0141,
      "step": 180960
    },
    {
      "epoch": 70.28,
      "learning_rate": 1.062938511326861e-05,
      "loss": 0.0523,
      "step": 180970
    },
    {
      "epoch": 70.28,
      "learning_rate": 1.0628867313915859e-05,
      "loss": 0.185,
      "step": 180980
    },
    {
      "epoch": 70.29,
      "learning_rate": 1.0628349514563107e-05,
      "loss": 0.0559,
      "step": 180990
    },
    {
      "epoch": 70.29,
      "learning_rate": 1.0627831715210356e-05,
      "loss": 0.1456,
      "step": 181000
    },
    {
      "epoch": 70.3,
      "learning_rate": 1.0627313915857606e-05,
      "loss": 0.1343,
      "step": 181010
    },
    {
      "epoch": 70.3,
      "learning_rate": 1.0626796116504856e-05,
      "loss": 0.0243,
      "step": 181020
    },
    {
      "epoch": 70.3,
      "learning_rate": 1.0626278317152104e-05,
      "loss": 0.0198,
      "step": 181030
    },
    {
      "epoch": 70.31,
      "learning_rate": 1.0625760517799353e-05,
      "loss": 0.0022,
      "step": 181040
    },
    {
      "epoch": 70.31,
      "learning_rate": 1.0625242718446603e-05,
      "loss": 0.0457,
      "step": 181050
    },
    {
      "epoch": 70.31,
      "learning_rate": 1.0624724919093852e-05,
      "loss": 0.1107,
      "step": 181060
    },
    {
      "epoch": 70.32,
      "learning_rate": 1.06242071197411e-05,
      "loss": 0.1717,
      "step": 181070
    },
    {
      "epoch": 70.32,
      "learning_rate": 1.062368932038835e-05,
      "loss": 0.0457,
      "step": 181080
    },
    {
      "epoch": 70.33,
      "learning_rate": 1.06231715210356e-05,
      "loss": 0.092,
      "step": 181090
    },
    {
      "epoch": 70.33,
      "learning_rate": 1.062265372168285e-05,
      "loss": 0.0513,
      "step": 181100
    },
    {
      "epoch": 70.33,
      "learning_rate": 1.0622135922330099e-05,
      "loss": 0.0384,
      "step": 181110
    },
    {
      "epoch": 70.34,
      "learning_rate": 1.0621618122977347e-05,
      "loss": 0.0658,
      "step": 181120
    },
    {
      "epoch": 70.34,
      "learning_rate": 1.0621100323624596e-05,
      "loss": 0.0604,
      "step": 181130
    },
    {
      "epoch": 70.35,
      "learning_rate": 1.0620582524271846e-05,
      "loss": 0.084,
      "step": 181140
    },
    {
      "epoch": 70.35,
      "learning_rate": 1.0620064724919094e-05,
      "loss": 0.0428,
      "step": 181150
    },
    {
      "epoch": 70.35,
      "learning_rate": 1.0619546925566344e-05,
      "loss": 0.0741,
      "step": 181160
    },
    {
      "epoch": 70.36,
      "learning_rate": 1.0619029126213593e-05,
      "loss": 0.0411,
      "step": 181170
    },
    {
      "epoch": 70.36,
      "learning_rate": 1.0618511326860843e-05,
      "loss": 0.0358,
      "step": 181180
    },
    {
      "epoch": 70.37,
      "learning_rate": 1.0617993527508092e-05,
      "loss": 0.1684,
      "step": 181190
    },
    {
      "epoch": 70.37,
      "learning_rate": 1.061747572815534e-05,
      "loss": 0.0372,
      "step": 181200
    },
    {
      "epoch": 70.37,
      "learning_rate": 1.061695792880259e-05,
      "loss": 0.0321,
      "step": 181210
    },
    {
      "epoch": 70.38,
      "learning_rate": 1.0616440129449838e-05,
      "loss": 0.1459,
      "step": 181220
    },
    {
      "epoch": 70.38,
      "learning_rate": 1.0615922330097088e-05,
      "loss": 0.0325,
      "step": 181230
    },
    {
      "epoch": 70.38,
      "learning_rate": 1.0615404530744337e-05,
      "loss": 0.023,
      "step": 181240
    },
    {
      "epoch": 70.39,
      "learning_rate": 1.0614886731391587e-05,
      "loss": 0.1089,
      "step": 181250
    },
    {
      "epoch": 70.39,
      "learning_rate": 1.0614368932038836e-05,
      "loss": 0.102,
      "step": 181260
    },
    {
      "epoch": 70.4,
      "learning_rate": 1.0613851132686086e-05,
      "loss": 0.064,
      "step": 181270
    },
    {
      "epoch": 70.4,
      "learning_rate": 1.0613333333333334e-05,
      "loss": 0.2323,
      "step": 181280
    },
    {
      "epoch": 70.4,
      "learning_rate": 1.0612815533980584e-05,
      "loss": 0.0232,
      "step": 181290
    },
    {
      "epoch": 70.41,
      "learning_rate": 1.0612297734627831e-05,
      "loss": 0.1379,
      "step": 181300
    },
    {
      "epoch": 70.41,
      "learning_rate": 1.0611779935275081e-05,
      "loss": 0.0997,
      "step": 181310
    },
    {
      "epoch": 70.42,
      "learning_rate": 1.061126213592233e-05,
      "loss": 0.0448,
      "step": 181320
    },
    {
      "epoch": 70.42,
      "learning_rate": 1.061074433656958e-05,
      "loss": 0.1354,
      "step": 181330
    },
    {
      "epoch": 70.42,
      "learning_rate": 1.061022653721683e-05,
      "loss": 0.0324,
      "step": 181340
    },
    {
      "epoch": 70.43,
      "learning_rate": 1.060970873786408e-05,
      "loss": 0.0649,
      "step": 181350
    },
    {
      "epoch": 70.43,
      "learning_rate": 1.0609190938511327e-05,
      "loss": 0.0132,
      "step": 181360
    },
    {
      "epoch": 70.43,
      "learning_rate": 1.0608673139158575e-05,
      "loss": 0.005,
      "step": 181370
    },
    {
      "epoch": 70.44,
      "learning_rate": 1.0608155339805825e-05,
      "loss": 0.0269,
      "step": 181380
    },
    {
      "epoch": 70.44,
      "learning_rate": 1.0607637540453075e-05,
      "loss": 0.0618,
      "step": 181390
    },
    {
      "epoch": 70.45,
      "learning_rate": 1.0607119741100324e-05,
      "loss": 0.0234,
      "step": 181400
    },
    {
      "epoch": 70.45,
      "learning_rate": 1.0606601941747574e-05,
      "loss": 0.0548,
      "step": 181410
    },
    {
      "epoch": 70.45,
      "learning_rate": 1.0606084142394823e-05,
      "loss": 0.0654,
      "step": 181420
    },
    {
      "epoch": 70.46,
      "learning_rate": 1.0605566343042073e-05,
      "loss": 0.0294,
      "step": 181430
    },
    {
      "epoch": 70.46,
      "learning_rate": 1.0605048543689321e-05,
      "loss": 0.0181,
      "step": 181440
    },
    {
      "epoch": 70.47,
      "learning_rate": 1.0604530744336569e-05,
      "loss": 0.0012,
      "step": 181450
    },
    {
      "epoch": 70.47,
      "learning_rate": 1.0604012944983819e-05,
      "loss": 0.1554,
      "step": 181460
    },
    {
      "epoch": 70.47,
      "learning_rate": 1.0603495145631068e-05,
      "loss": 0.0886,
      "step": 181470
    },
    {
      "epoch": 70.48,
      "learning_rate": 1.0602977346278318e-05,
      "loss": 0.0501,
      "step": 181480
    },
    {
      "epoch": 70.48,
      "learning_rate": 1.0602459546925567e-05,
      "loss": 0.0204,
      "step": 181490
    },
    {
      "epoch": 70.49,
      "learning_rate": 1.0601941747572817e-05,
      "loss": 0.0053,
      "step": 181500
    },
    {
      "epoch": 70.49,
      "learning_rate": 1.0601423948220067e-05,
      "loss": 0.0004,
      "step": 181510
    },
    {
      "epoch": 70.49,
      "learning_rate": 1.0600906148867313e-05,
      "loss": 0.0781,
      "step": 181520
    },
    {
      "epoch": 70.5,
      "learning_rate": 1.0600388349514563e-05,
      "loss": 0.0771,
      "step": 181530
    },
    {
      "epoch": 70.5,
      "learning_rate": 1.0599870550161812e-05,
      "loss": 0.0046,
      "step": 181540
    },
    {
      "epoch": 70.5,
      "learning_rate": 1.0599352750809062e-05,
      "loss": 0.1046,
      "step": 181550
    },
    {
      "epoch": 70.51,
      "learning_rate": 1.0598834951456311e-05,
      "loss": 0.21,
      "step": 181560
    },
    {
      "epoch": 70.51,
      "learning_rate": 1.0598317152103561e-05,
      "loss": 0.0298,
      "step": 181570
    },
    {
      "epoch": 70.52,
      "learning_rate": 1.059779935275081e-05,
      "loss": 0.0729,
      "step": 181580
    },
    {
      "epoch": 70.52,
      "learning_rate": 1.059728155339806e-05,
      "loss": 0.0074,
      "step": 181590
    },
    {
      "epoch": 70.52,
      "learning_rate": 1.0596763754045306e-05,
      "loss": 0.0262,
      "step": 181600
    },
    {
      "epoch": 70.53,
      "learning_rate": 1.0596245954692556e-05,
      "loss": 0.0331,
      "step": 181610
    },
    {
      "epoch": 70.53,
      "learning_rate": 1.0595728155339806e-05,
      "loss": 0.0796,
      "step": 181620
    },
    {
      "epoch": 70.54,
      "learning_rate": 1.0595210355987055e-05,
      "loss": 0.0408,
      "step": 181630
    },
    {
      "epoch": 70.54,
      "learning_rate": 1.0594692556634305e-05,
      "loss": 0.0032,
      "step": 181640
    },
    {
      "epoch": 70.54,
      "learning_rate": 1.0594174757281555e-05,
      "loss": 0.0888,
      "step": 181650
    },
    {
      "epoch": 70.55,
      "learning_rate": 1.0593656957928804e-05,
      "loss": 0.1528,
      "step": 181660
    },
    {
      "epoch": 70.55,
      "learning_rate": 1.0593139158576054e-05,
      "loss": 0.2742,
      "step": 181670
    },
    {
      "epoch": 70.56,
      "learning_rate": 1.0592621359223303e-05,
      "loss": 0.0995,
      "step": 181680
    },
    {
      "epoch": 70.56,
      "learning_rate": 1.059210355987055e-05,
      "loss": 0.0751,
      "step": 181690
    },
    {
      "epoch": 70.56,
      "learning_rate": 1.05915857605178e-05,
      "loss": 0.0818,
      "step": 181700
    },
    {
      "epoch": 70.57,
      "learning_rate": 1.0591067961165049e-05,
      "loss": 0.0878,
      "step": 181710
    },
    {
      "epoch": 70.57,
      "learning_rate": 1.0590550161812298e-05,
      "loss": 0.0602,
      "step": 181720
    },
    {
      "epoch": 70.57,
      "learning_rate": 1.0590032362459548e-05,
      "loss": 0.0711,
      "step": 181730
    },
    {
      "epoch": 70.58,
      "learning_rate": 1.0589514563106798e-05,
      "loss": 0.0392,
      "step": 181740
    },
    {
      "epoch": 70.58,
      "learning_rate": 1.0588996763754047e-05,
      "loss": 0.0599,
      "step": 181750
    },
    {
      "epoch": 70.59,
      "learning_rate": 1.0588478964401297e-05,
      "loss": 0.2649,
      "step": 181760
    },
    {
      "epoch": 70.59,
      "learning_rate": 1.0587961165048543e-05,
      "loss": 0.0328,
      "step": 181770
    },
    {
      "epoch": 70.59,
      "learning_rate": 1.0587443365695793e-05,
      "loss": 0.0075,
      "step": 181780
    },
    {
      "epoch": 70.6,
      "learning_rate": 1.0586925566343042e-05,
      "loss": 0.1296,
      "step": 181790
    },
    {
      "epoch": 70.6,
      "learning_rate": 1.0586407766990292e-05,
      "loss": 0.0169,
      "step": 181800
    },
    {
      "epoch": 70.61,
      "learning_rate": 1.0585889967637542e-05,
      "loss": 0.0536,
      "step": 181810
    },
    {
      "epoch": 70.61,
      "learning_rate": 1.0585372168284791e-05,
      "loss": 0.1202,
      "step": 181820
    },
    {
      "epoch": 70.61,
      "learning_rate": 1.0584854368932041e-05,
      "loss": 0.0007,
      "step": 181830
    },
    {
      "epoch": 70.62,
      "learning_rate": 1.058433656957929e-05,
      "loss": 0.1145,
      "step": 181840
    },
    {
      "epoch": 70.62,
      "learning_rate": 1.0583818770226537e-05,
      "loss": 0.0289,
      "step": 181850
    },
    {
      "epoch": 70.63,
      "learning_rate": 1.0583300970873786e-05,
      "loss": 0.0397,
      "step": 181860
    },
    {
      "epoch": 70.63,
      "learning_rate": 1.0582783171521036e-05,
      "loss": 0.202,
      "step": 181870
    },
    {
      "epoch": 70.63,
      "learning_rate": 1.0582265372168286e-05,
      "loss": 0.0678,
      "step": 181880
    },
    {
      "epoch": 70.64,
      "learning_rate": 1.0581747572815535e-05,
      "loss": 0.0312,
      "step": 181890
    },
    {
      "epoch": 70.64,
      "learning_rate": 1.0581229773462785e-05,
      "loss": 0.0941,
      "step": 181900
    },
    {
      "epoch": 70.64,
      "learning_rate": 1.0580711974110034e-05,
      "loss": 0.0384,
      "step": 181910
    },
    {
      "epoch": 70.65,
      "learning_rate": 1.0580194174757284e-05,
      "loss": 0.1796,
      "step": 181920
    },
    {
      "epoch": 70.65,
      "learning_rate": 1.057967637540453e-05,
      "loss": 0.0642,
      "step": 181930
    },
    {
      "epoch": 70.66,
      "learning_rate": 1.057915857605178e-05,
      "loss": 0.0766,
      "step": 181940
    },
    {
      "epoch": 70.66,
      "learning_rate": 1.057864077669903e-05,
      "loss": 0.0739,
      "step": 181950
    },
    {
      "epoch": 70.66,
      "learning_rate": 1.057812297734628e-05,
      "loss": 0.1495,
      "step": 181960
    },
    {
      "epoch": 70.67,
      "learning_rate": 1.0577605177993529e-05,
      "loss": 0.0088,
      "step": 181970
    },
    {
      "epoch": 70.67,
      "learning_rate": 1.0577087378640778e-05,
      "loss": 0.1097,
      "step": 181980
    },
    {
      "epoch": 70.68,
      "learning_rate": 1.0576569579288028e-05,
      "loss": 0.1385,
      "step": 181990
    },
    {
      "epoch": 70.68,
      "learning_rate": 1.0576051779935278e-05,
      "loss": 0.0446,
      "step": 182000
    },
    {
      "epoch": 70.68,
      "learning_rate": 1.0575533980582524e-05,
      "loss": 0.0658,
      "step": 182010
    },
    {
      "epoch": 70.69,
      "learning_rate": 1.0575016181229773e-05,
      "loss": 0.0334,
      "step": 182020
    },
    {
      "epoch": 70.69,
      "learning_rate": 1.0574498381877023e-05,
      "loss": 0.0145,
      "step": 182030
    },
    {
      "epoch": 70.7,
      "learning_rate": 1.0573980582524273e-05,
      "loss": 0.0007,
      "step": 182040
    },
    {
      "epoch": 70.7,
      "learning_rate": 1.0573462783171522e-05,
      "loss": 0.0015,
      "step": 182050
    },
    {
      "epoch": 70.7,
      "learning_rate": 1.0572944983818772e-05,
      "loss": 0.1117,
      "step": 182060
    },
    {
      "epoch": 70.71,
      "learning_rate": 1.0572427184466022e-05,
      "loss": 0.0466,
      "step": 182070
    },
    {
      "epoch": 70.71,
      "learning_rate": 1.0571909385113271e-05,
      "loss": 0.0556,
      "step": 182080
    },
    {
      "epoch": 70.71,
      "learning_rate": 1.0571391585760517e-05,
      "loss": 0.032,
      "step": 182090
    },
    {
      "epoch": 70.72,
      "learning_rate": 1.0570873786407767e-05,
      "loss": 0.1218,
      "step": 182100
    },
    {
      "epoch": 70.72,
      "learning_rate": 1.0570355987055017e-05,
      "loss": 0.1239,
      "step": 182110
    },
    {
      "epoch": 70.73,
      "learning_rate": 1.0569838187702266e-05,
      "loss": 0.1025,
      "step": 182120
    },
    {
      "epoch": 70.73,
      "learning_rate": 1.0569320388349516e-05,
      "loss": 0.0091,
      "step": 182130
    },
    {
      "epoch": 70.73,
      "learning_rate": 1.0568802588996766e-05,
      "loss": 0.0691,
      "step": 182140
    },
    {
      "epoch": 70.74,
      "learning_rate": 1.0568284789644015e-05,
      "loss": 0.0963,
      "step": 182150
    },
    {
      "epoch": 70.74,
      "learning_rate": 1.0567766990291263e-05,
      "loss": 0.1049,
      "step": 182160
    },
    {
      "epoch": 70.75,
      "learning_rate": 1.0567249190938511e-05,
      "loss": 0.1234,
      "step": 182170
    },
    {
      "epoch": 70.75,
      "learning_rate": 1.056673139158576e-05,
      "loss": 0.0564,
      "step": 182180
    },
    {
      "epoch": 70.75,
      "learning_rate": 1.056621359223301e-05,
      "loss": 0.1231,
      "step": 182190
    },
    {
      "epoch": 70.76,
      "learning_rate": 1.056569579288026e-05,
      "loss": 0.058,
      "step": 182200
    },
    {
      "epoch": 70.76,
      "learning_rate": 1.056517799352751e-05,
      "loss": 0.0496,
      "step": 182210
    },
    {
      "epoch": 70.77,
      "learning_rate": 1.0564660194174759e-05,
      "loss": 0.0152,
      "step": 182220
    },
    {
      "epoch": 70.77,
      "learning_rate": 1.0564142394822009e-05,
      "loss": 0.0679,
      "step": 182230
    },
    {
      "epoch": 70.77,
      "learning_rate": 1.0563624595469257e-05,
      "loss": 0.1545,
      "step": 182240
    },
    {
      "epoch": 70.78,
      "learning_rate": 1.0563106796116506e-05,
      "loss": 0.0072,
      "step": 182250
    },
    {
      "epoch": 70.78,
      "learning_rate": 1.0562588996763754e-05,
      "loss": 0.1236,
      "step": 182260
    },
    {
      "epoch": 70.78,
      "learning_rate": 1.0562071197411004e-05,
      "loss": 0.2174,
      "step": 182270
    },
    {
      "epoch": 70.79,
      "learning_rate": 1.0561553398058253e-05,
      "loss": 0.0418,
      "step": 182280
    },
    {
      "epoch": 70.79,
      "learning_rate": 1.0561035598705503e-05,
      "loss": 0.0607,
      "step": 182290
    },
    {
      "epoch": 70.8,
      "learning_rate": 1.0560517799352753e-05,
      "loss": 0.0709,
      "step": 182300
    },
    {
      "epoch": 70.8,
      "learning_rate": 1.056e-05,
      "loss": 0.1377,
      "step": 182310
    },
    {
      "epoch": 70.8,
      "learning_rate": 1.055948220064725e-05,
      "loss": 0.0465,
      "step": 182320
    },
    {
      "epoch": 70.81,
      "learning_rate": 1.05589644012945e-05,
      "loss": 0.0325,
      "step": 182330
    },
    {
      "epoch": 70.81,
      "learning_rate": 1.0558446601941748e-05,
      "loss": 0.065,
      "step": 182340
    },
    {
      "epoch": 70.82,
      "learning_rate": 1.0557928802588997e-05,
      "loss": 0.0127,
      "step": 182350
    },
    {
      "epoch": 70.82,
      "learning_rate": 1.0557411003236247e-05,
      "loss": 0.0454,
      "step": 182360
    },
    {
      "epoch": 70.82,
      "learning_rate": 1.0556893203883497e-05,
      "loss": 0.03,
      "step": 182370
    },
    {
      "epoch": 70.83,
      "learning_rate": 1.0556375404530746e-05,
      "loss": 0.0113,
      "step": 182380
    },
    {
      "epoch": 70.83,
      "learning_rate": 1.0555857605177994e-05,
      "loss": 0.0504,
      "step": 182390
    },
    {
      "epoch": 70.83,
      "learning_rate": 1.0555339805825244e-05,
      "loss": 0.0952,
      "step": 182400
    },
    {
      "epoch": 70.84,
      "learning_rate": 1.0554822006472493e-05,
      "loss": 0.0294,
      "step": 182410
    },
    {
      "epoch": 70.84,
      "learning_rate": 1.0554304207119741e-05,
      "loss": 0.0558,
      "step": 182420
    },
    {
      "epoch": 70.85,
      "learning_rate": 1.0553786407766991e-05,
      "loss": 0.1178,
      "step": 182430
    },
    {
      "epoch": 70.85,
      "learning_rate": 1.055326860841424e-05,
      "loss": 0.1503,
      "step": 182440
    },
    {
      "epoch": 70.85,
      "learning_rate": 1.055275080906149e-05,
      "loss": 0.1709,
      "step": 182450
    },
    {
      "epoch": 70.86,
      "learning_rate": 1.0552233009708738e-05,
      "loss": 0.0143,
      "step": 182460
    },
    {
      "epoch": 70.86,
      "learning_rate": 1.0551715210355988e-05,
      "loss": 0.0299,
      "step": 182470
    },
    {
      "epoch": 70.87,
      "learning_rate": 1.0551197411003237e-05,
      "loss": 0.0592,
      "step": 182480
    },
    {
      "epoch": 70.87,
      "learning_rate": 1.0550679611650487e-05,
      "loss": 0.0814,
      "step": 182490
    },
    {
      "epoch": 70.87,
      "learning_rate": 1.0550161812297735e-05,
      "loss": 0.0793,
      "step": 182500
    },
    {
      "epoch": 70.88,
      "learning_rate": 1.0549644012944984e-05,
      "loss": 0.0793,
      "step": 182510
    },
    {
      "epoch": 70.88,
      "learning_rate": 1.0549126213592234e-05,
      "loss": 0.0534,
      "step": 182520
    },
    {
      "epoch": 70.89,
      "learning_rate": 1.0548608414239484e-05,
      "loss": 0.0718,
      "step": 182530
    },
    {
      "epoch": 70.89,
      "learning_rate": 1.0548090614886732e-05,
      "loss": 0.0234,
      "step": 182540
    },
    {
      "epoch": 70.89,
      "learning_rate": 1.0547572815533981e-05,
      "loss": 0.1224,
      "step": 182550
    },
    {
      "epoch": 70.9,
      "learning_rate": 1.054705501618123e-05,
      "loss": 0.1524,
      "step": 182560
    },
    {
      "epoch": 70.9,
      "learning_rate": 1.054653721682848e-05,
      "loss": 0.0506,
      "step": 182570
    },
    {
      "epoch": 70.9,
      "learning_rate": 1.0546019417475728e-05,
      "loss": 0.0143,
      "step": 182580
    },
    {
      "epoch": 70.91,
      "learning_rate": 1.0545501618122978e-05,
      "loss": 0.1028,
      "step": 182590
    },
    {
      "epoch": 70.91,
      "learning_rate": 1.0544983818770228e-05,
      "loss": 0.0118,
      "step": 182600
    },
    {
      "epoch": 70.92,
      "learning_rate": 1.0544466019417477e-05,
      "loss": 0.0175,
      "step": 182610
    },
    {
      "epoch": 70.92,
      "learning_rate": 1.0543948220064725e-05,
      "loss": 0.0527,
      "step": 182620
    },
    {
      "epoch": 70.92,
      "learning_rate": 1.0543430420711975e-05,
      "loss": 0.0273,
      "step": 182630
    },
    {
      "epoch": 70.93,
      "learning_rate": 1.0542912621359224e-05,
      "loss": 0.0185,
      "step": 182640
    },
    {
      "epoch": 70.93,
      "learning_rate": 1.0542394822006474e-05,
      "loss": 0.1144,
      "step": 182650
    },
    {
      "epoch": 70.94,
      "learning_rate": 1.0541877022653722e-05,
      "loss": 0.1295,
      "step": 182660
    },
    {
      "epoch": 70.94,
      "learning_rate": 1.0541359223300972e-05,
      "loss": 0.0135,
      "step": 182670
    },
    {
      "epoch": 70.94,
      "learning_rate": 1.0540841423948221e-05,
      "loss": 0.0197,
      "step": 182680
    },
    {
      "epoch": 70.95,
      "learning_rate": 1.0540323624595469e-05,
      "loss": 0.2627,
      "step": 182690
    },
    {
      "epoch": 70.95,
      "learning_rate": 1.0539805825242719e-05,
      "loss": 0.1789,
      "step": 182700
    },
    {
      "epoch": 70.96,
      "learning_rate": 1.0539288025889968e-05,
      "loss": 0.1568,
      "step": 182710
    },
    {
      "epoch": 70.96,
      "learning_rate": 1.0538770226537218e-05,
      "loss": 0.0134,
      "step": 182720
    },
    {
      "epoch": 70.96,
      "learning_rate": 1.0538252427184468e-05,
      "loss": 0.0443,
      "step": 182730
    },
    {
      "epoch": 70.97,
      "learning_rate": 1.0537734627831715e-05,
      "loss": 0.0037,
      "step": 182740
    },
    {
      "epoch": 70.97,
      "learning_rate": 1.0537216828478965e-05,
      "loss": 0.1315,
      "step": 182750
    },
    {
      "epoch": 70.97,
      "learning_rate": 1.0536699029126215e-05,
      "loss": 0.0414,
      "step": 182760
    },
    {
      "epoch": 70.98,
      "learning_rate": 1.0536181229773463e-05,
      "loss": 0.0014,
      "step": 182770
    },
    {
      "epoch": 70.98,
      "learning_rate": 1.0535663430420712e-05,
      "loss": 0.2767,
      "step": 182780
    },
    {
      "epoch": 70.99,
      "learning_rate": 1.0535145631067962e-05,
      "loss": 0.131,
      "step": 182790
    },
    {
      "epoch": 70.99,
      "learning_rate": 1.0534627831715211e-05,
      "loss": 0.182,
      "step": 182800
    },
    {
      "epoch": 70.99,
      "learning_rate": 1.0534110032362461e-05,
      "loss": 0.0684,
      "step": 182810
    },
    {
      "epoch": 71.0,
      "learning_rate": 1.053359223300971e-05,
      "loss": 0.0217,
      "step": 182820
    },
    {
      "epoch": 71.0,
      "eval_accuracy": 0.9488308115543329,
      "eval_loss": 0.35828545689582825,
      "eval_runtime": 8.1754,
      "eval_samples_per_second": 444.628,
      "eval_steps_per_second": 55.655,
      "step": 182825
    },
    {
      "epoch": 71.0,
      "learning_rate": 1.0533074433656959e-05,
      "loss": 0.0338,
      "step": 182830
    },
    {
      "epoch": 71.01,
      "learning_rate": 1.0532556634304207e-05,
      "loss": 0.0241,
      "step": 182840
    },
    {
      "epoch": 71.01,
      "learning_rate": 1.0532038834951456e-05,
      "loss": 0.1219,
      "step": 182850
    },
    {
      "epoch": 71.01,
      "learning_rate": 1.0531521035598706e-05,
      "loss": 0.0288,
      "step": 182860
    },
    {
      "epoch": 71.02,
      "learning_rate": 1.0531003236245955e-05,
      "loss": 0.0197,
      "step": 182870
    },
    {
      "epoch": 71.02,
      "learning_rate": 1.0530485436893205e-05,
      "loss": 0.0403,
      "step": 182880
    },
    {
      "epoch": 71.03,
      "learning_rate": 1.0529967637540455e-05,
      "loss": 0.0246,
      "step": 182890
    },
    {
      "epoch": 71.03,
      "learning_rate": 1.0529449838187704e-05,
      "loss": 0.1104,
      "step": 182900
    },
    {
      "epoch": 71.03,
      "learning_rate": 1.0528932038834952e-05,
      "loss": 0.1113,
      "step": 182910
    },
    {
      "epoch": 71.04,
      "learning_rate": 1.05284142394822e-05,
      "loss": 0.0149,
      "step": 182920
    },
    {
      "epoch": 71.04,
      "learning_rate": 1.052789644012945e-05,
      "loss": 0.0733,
      "step": 182930
    },
    {
      "epoch": 71.04,
      "learning_rate": 1.05273786407767e-05,
      "loss": 0.0449,
      "step": 182940
    },
    {
      "epoch": 71.05,
      "learning_rate": 1.0526860841423949e-05,
      "loss": 0.0306,
      "step": 182950
    },
    {
      "epoch": 71.05,
      "learning_rate": 1.0526343042071199e-05,
      "loss": 0.0157,
      "step": 182960
    },
    {
      "epoch": 71.06,
      "learning_rate": 1.0525825242718448e-05,
      "loss": 0.0017,
      "step": 182970
    },
    {
      "epoch": 71.06,
      "learning_rate": 1.0525307443365698e-05,
      "loss": 0.009,
      "step": 182980
    },
    {
      "epoch": 71.06,
      "learning_rate": 1.0524789644012944e-05,
      "loss": 0.1288,
      "step": 182990
    },
    {
      "epoch": 71.07,
      "learning_rate": 1.0524271844660194e-05,
      "loss": 0.0353,
      "step": 183000
    },
    {
      "epoch": 71.07,
      "learning_rate": 1.0523754045307443e-05,
      "loss": 0.0372,
      "step": 183010
    },
    {
      "epoch": 71.08,
      "learning_rate": 1.0523236245954693e-05,
      "loss": 0.0913,
      "step": 183020
    },
    {
      "epoch": 71.08,
      "learning_rate": 1.0522718446601943e-05,
      "loss": 0.0389,
      "step": 183030
    },
    {
      "epoch": 71.08,
      "learning_rate": 1.0522200647249192e-05,
      "loss": 0.0641,
      "step": 183040
    },
    {
      "epoch": 71.09,
      "learning_rate": 1.0521682847896442e-05,
      "loss": 0.0197,
      "step": 183050
    },
    {
      "epoch": 71.09,
      "learning_rate": 1.0521165048543691e-05,
      "loss": 0.103,
      "step": 183060
    },
    {
      "epoch": 71.1,
      "learning_rate": 1.0520647249190938e-05,
      "loss": 0.1023,
      "step": 183070
    },
    {
      "epoch": 71.1,
      "learning_rate": 1.0520129449838187e-05,
      "loss": 0.0596,
      "step": 183080
    },
    {
      "epoch": 71.1,
      "learning_rate": 1.0519611650485437e-05,
      "loss": 0.0348,
      "step": 183090
    },
    {
      "epoch": 71.11,
      "learning_rate": 1.0519093851132686e-05,
      "loss": 0.0003,
      "step": 183100
    },
    {
      "epoch": 71.11,
      "learning_rate": 1.0518576051779936e-05,
      "loss": 0.0225,
      "step": 183110
    },
    {
      "epoch": 71.11,
      "learning_rate": 1.0518058252427186e-05,
      "loss": 0.0869,
      "step": 183120
    },
    {
      "epoch": 71.12,
      "learning_rate": 1.0517540453074435e-05,
      "loss": 0.1911,
      "step": 183130
    },
    {
      "epoch": 71.12,
      "learning_rate": 1.0517022653721685e-05,
      "loss": 0.0004,
      "step": 183140
    },
    {
      "epoch": 71.13,
      "learning_rate": 1.0516504854368931e-05,
      "loss": 0.0246,
      "step": 183150
    },
    {
      "epoch": 71.13,
      "learning_rate": 1.051598705501618e-05,
      "loss": 0.0969,
      "step": 183160
    },
    {
      "epoch": 71.13,
      "learning_rate": 1.051546925566343e-05,
      "loss": 0.0827,
      "step": 183170
    },
    {
      "epoch": 71.14,
      "learning_rate": 1.051495145631068e-05,
      "loss": 0.0343,
      "step": 183180
    },
    {
      "epoch": 71.14,
      "learning_rate": 1.051443365695793e-05,
      "loss": 0.0929,
      "step": 183190
    },
    {
      "epoch": 71.15,
      "learning_rate": 1.051391585760518e-05,
      "loss": 0.0964,
      "step": 183200
    },
    {
      "epoch": 71.15,
      "learning_rate": 1.0513398058252429e-05,
      "loss": 0.0972,
      "step": 183210
    },
    {
      "epoch": 71.15,
      "learning_rate": 1.0512880258899679e-05,
      "loss": 0.0383,
      "step": 183220
    },
    {
      "epoch": 71.16,
      "learning_rate": 1.0512362459546925e-05,
      "loss": 0.209,
      "step": 183230
    },
    {
      "epoch": 71.16,
      "learning_rate": 1.0511844660194174e-05,
      "loss": 0.0594,
      "step": 183240
    },
    {
      "epoch": 71.17,
      "learning_rate": 1.0511326860841424e-05,
      "loss": 0.0746,
      "step": 183250
    },
    {
      "epoch": 71.17,
      "learning_rate": 1.0510809061488674e-05,
      "loss": 0.0955,
      "step": 183260
    },
    {
      "epoch": 71.17,
      "learning_rate": 1.0510291262135923e-05,
      "loss": 0.0318,
      "step": 183270
    },
    {
      "epoch": 71.18,
      "learning_rate": 1.0509773462783173e-05,
      "loss": 0.1788,
      "step": 183280
    },
    {
      "epoch": 71.18,
      "learning_rate": 1.0509255663430422e-05,
      "loss": 0.103,
      "step": 183290
    },
    {
      "epoch": 71.18,
      "learning_rate": 1.0508737864077672e-05,
      "loss": 0.0325,
      "step": 183300
    },
    {
      "epoch": 71.19,
      "learning_rate": 1.0508220064724918e-05,
      "loss": 0.0011,
      "step": 183310
    },
    {
      "epoch": 71.19,
      "learning_rate": 1.0507702265372168e-05,
      "loss": 0.0914,
      "step": 183320
    },
    {
      "epoch": 71.2,
      "learning_rate": 1.0507184466019418e-05,
      "loss": 0.0532,
      "step": 183330
    },
    {
      "epoch": 71.2,
      "learning_rate": 1.0506666666666667e-05,
      "loss": 0.0183,
      "step": 183340
    },
    {
      "epoch": 71.2,
      "learning_rate": 1.0506148867313917e-05,
      "loss": 0.0173,
      "step": 183350
    },
    {
      "epoch": 71.21,
      "learning_rate": 1.0505631067961166e-05,
      "loss": 0.1288,
      "step": 183360
    },
    {
      "epoch": 71.21,
      "learning_rate": 1.0505113268608416e-05,
      "loss": 0.028,
      "step": 183370
    },
    {
      "epoch": 71.22,
      "learning_rate": 1.0504595469255666e-05,
      "loss": 0.0798,
      "step": 183380
    },
    {
      "epoch": 71.22,
      "learning_rate": 1.0504077669902915e-05,
      "loss": 0.0026,
      "step": 183390
    },
    {
      "epoch": 71.22,
      "learning_rate": 1.0503559870550161e-05,
      "loss": 0.1988,
      "step": 183400
    },
    {
      "epoch": 71.23,
      "learning_rate": 1.0503042071197411e-05,
      "loss": 0.0904,
      "step": 183410
    },
    {
      "epoch": 71.23,
      "learning_rate": 1.050252427184466e-05,
      "loss": 0.0959,
      "step": 183420
    },
    {
      "epoch": 71.23,
      "learning_rate": 1.050200647249191e-05,
      "loss": 0.0828,
      "step": 183430
    },
    {
      "epoch": 71.24,
      "learning_rate": 1.050148867313916e-05,
      "loss": 0.1064,
      "step": 183440
    },
    {
      "epoch": 71.24,
      "learning_rate": 1.050097087378641e-05,
      "loss": 0.1591,
      "step": 183450
    },
    {
      "epoch": 71.25,
      "learning_rate": 1.050045307443366e-05,
      "loss": 0.0582,
      "step": 183460
    },
    {
      "epoch": 71.25,
      "learning_rate": 1.0499935275080909e-05,
      "loss": 0.0207,
      "step": 183470
    },
    {
      "epoch": 71.25,
      "learning_rate": 1.0499417475728155e-05,
      "loss": 0.0148,
      "step": 183480
    },
    {
      "epoch": 71.26,
      "learning_rate": 1.0498899676375405e-05,
      "loss": 0.1408,
      "step": 183490
    },
    {
      "epoch": 71.26,
      "learning_rate": 1.0498381877022654e-05,
      "loss": 0.0765,
      "step": 183500
    },
    {
      "epoch": 71.27,
      "learning_rate": 1.0497864077669904e-05,
      "loss": 0.0649,
      "step": 183510
    },
    {
      "epoch": 71.27,
      "learning_rate": 1.0497346278317153e-05,
      "loss": 0.1207,
      "step": 183520
    },
    {
      "epoch": 71.27,
      "learning_rate": 1.0496828478964403e-05,
      "loss": 0.1496,
      "step": 183530
    },
    {
      "epoch": 71.28,
      "learning_rate": 1.0496310679611653e-05,
      "loss": 0.0654,
      "step": 183540
    },
    {
      "epoch": 71.28,
      "learning_rate": 1.0495792880258902e-05,
      "loss": 0.0734,
      "step": 183550
    },
    {
      "epoch": 71.29,
      "learning_rate": 1.0495275080906149e-05,
      "loss": 0.1172,
      "step": 183560
    },
    {
      "epoch": 71.29,
      "learning_rate": 1.0494757281553398e-05,
      "loss": 0.021,
      "step": 183570
    },
    {
      "epoch": 71.29,
      "learning_rate": 1.0494239482200648e-05,
      "loss": 0.1245,
      "step": 183580
    },
    {
      "epoch": 71.3,
      "learning_rate": 1.0493721682847897e-05,
      "loss": 0.0475,
      "step": 183590
    },
    {
      "epoch": 71.3,
      "learning_rate": 1.0493203883495147e-05,
      "loss": 0.0441,
      "step": 183600
    },
    {
      "epoch": 71.3,
      "learning_rate": 1.0492686084142397e-05,
      "loss": 0.0645,
      "step": 183610
    },
    {
      "epoch": 71.31,
      "learning_rate": 1.0492168284789646e-05,
      "loss": 0.0123,
      "step": 183620
    },
    {
      "epoch": 71.31,
      "learning_rate": 1.0491650485436894e-05,
      "loss": 0.1279,
      "step": 183630
    },
    {
      "epoch": 71.32,
      "learning_rate": 1.0491132686084142e-05,
      "loss": 0.0004,
      "step": 183640
    },
    {
      "epoch": 71.32,
      "learning_rate": 1.0490614886731392e-05,
      "loss": 0.0672,
      "step": 183650
    },
    {
      "epoch": 71.32,
      "learning_rate": 1.0490097087378641e-05,
      "loss": 0.0752,
      "step": 183660
    },
    {
      "epoch": 71.33,
      "learning_rate": 1.0489579288025891e-05,
      "loss": 0.0622,
      "step": 183670
    },
    {
      "epoch": 71.33,
      "learning_rate": 1.048906148867314e-05,
      "loss": 0.0165,
      "step": 183680
    },
    {
      "epoch": 71.34,
      "learning_rate": 1.048854368932039e-05,
      "loss": 0.0099,
      "step": 183690
    },
    {
      "epoch": 71.34,
      "learning_rate": 1.048802588996764e-05,
      "loss": 0.1289,
      "step": 183700
    },
    {
      "epoch": 71.34,
      "learning_rate": 1.0487508090614888e-05,
      "loss": 0.1947,
      "step": 183710
    },
    {
      "epoch": 71.35,
      "learning_rate": 1.0486990291262136e-05,
      "loss": 0.0011,
      "step": 183720
    },
    {
      "epoch": 71.35,
      "learning_rate": 1.0486472491909385e-05,
      "loss": 0.0617,
      "step": 183730
    },
    {
      "epoch": 71.36,
      "learning_rate": 1.0485954692556635e-05,
      "loss": 0.1271,
      "step": 183740
    },
    {
      "epoch": 71.36,
      "learning_rate": 1.0485436893203885e-05,
      "loss": 0.0866,
      "step": 183750
    },
    {
      "epoch": 71.36,
      "learning_rate": 1.0484919093851134e-05,
      "loss": 0.0946,
      "step": 183760
    },
    {
      "epoch": 71.37,
      "learning_rate": 1.0484401294498384e-05,
      "loss": 0.0961,
      "step": 183770
    },
    {
      "epoch": 71.37,
      "learning_rate": 1.0483883495145632e-05,
      "loss": 0.1469,
      "step": 183780
    },
    {
      "epoch": 71.37,
      "learning_rate": 1.0483365695792881e-05,
      "loss": 0.05,
      "step": 183790
    },
    {
      "epoch": 71.38,
      "learning_rate": 1.048284789644013e-05,
      "loss": 0.0723,
      "step": 183800
    },
    {
      "epoch": 71.38,
      "learning_rate": 1.0482330097087379e-05,
      "loss": 0.0046,
      "step": 183810
    },
    {
      "epoch": 71.39,
      "learning_rate": 1.0481812297734628e-05,
      "loss": 0.1058,
      "step": 183820
    },
    {
      "epoch": 71.39,
      "learning_rate": 1.0481294498381878e-05,
      "loss": 0.069,
      "step": 183830
    },
    {
      "epoch": 71.39,
      "learning_rate": 1.0480776699029128e-05,
      "loss": 0.0112,
      "step": 183840
    },
    {
      "epoch": 71.4,
      "learning_rate": 1.0480258899676377e-05,
      "loss": 0.0226,
      "step": 183850
    },
    {
      "epoch": 71.4,
      "learning_rate": 1.0479741100323625e-05,
      "loss": 0.0015,
      "step": 183860
    },
    {
      "epoch": 71.41,
      "learning_rate": 1.0479223300970875e-05,
      "loss": 0.0029,
      "step": 183870
    },
    {
      "epoch": 71.41,
      "learning_rate": 1.0478705501618125e-05,
      "loss": 0.0968,
      "step": 183880
    },
    {
      "epoch": 71.41,
      "learning_rate": 1.0478187702265372e-05,
      "loss": 0.0174,
      "step": 183890
    },
    {
      "epoch": 71.42,
      "learning_rate": 1.0477669902912622e-05,
      "loss": 0.0864,
      "step": 183900
    },
    {
      "epoch": 71.42,
      "learning_rate": 1.0477152103559872e-05,
      "loss": 0.0618,
      "step": 183910
    },
    {
      "epoch": 71.43,
      "learning_rate": 1.0476634304207121e-05,
      "loss": 0.0897,
      "step": 183920
    },
    {
      "epoch": 71.43,
      "learning_rate": 1.047611650485437e-05,
      "loss": 0.0644,
      "step": 183930
    },
    {
      "epoch": 71.43,
      "learning_rate": 1.0475598705501619e-05,
      "loss": 0.0337,
      "step": 183940
    },
    {
      "epoch": 71.44,
      "learning_rate": 1.0475080906148868e-05,
      "loss": 0.028,
      "step": 183950
    },
    {
      "epoch": 71.44,
      "learning_rate": 1.0474563106796118e-05,
      "loss": 0.0003,
      "step": 183960
    },
    {
      "epoch": 71.44,
      "learning_rate": 1.0474045307443366e-05,
      "loss": 0.0232,
      "step": 183970
    },
    {
      "epoch": 71.45,
      "learning_rate": 1.0473527508090616e-05,
      "loss": 0.0549,
      "step": 183980
    },
    {
      "epoch": 71.45,
      "learning_rate": 1.0473009708737865e-05,
      "loss": 0.0692,
      "step": 183990
    },
    {
      "epoch": 71.46,
      "learning_rate": 1.0472491909385115e-05,
      "loss": 0.0019,
      "step": 184000
    },
    {
      "epoch": 71.46,
      "learning_rate": 1.0471974110032363e-05,
      "loss": 0.0645,
      "step": 184010
    },
    {
      "epoch": 71.46,
      "learning_rate": 1.0471456310679612e-05,
      "loss": 0.095,
      "step": 184020
    },
    {
      "epoch": 71.47,
      "learning_rate": 1.0470938511326862e-05,
      "loss": 0.082,
      "step": 184030
    },
    {
      "epoch": 71.47,
      "learning_rate": 1.0470420711974112e-05,
      "loss": 0.0492,
      "step": 184040
    },
    {
      "epoch": 71.48,
      "learning_rate": 1.046990291262136e-05,
      "loss": 0.0232,
      "step": 184050
    },
    {
      "epoch": 71.48,
      "learning_rate": 1.046938511326861e-05,
      "loss": 0.0615,
      "step": 184060
    },
    {
      "epoch": 71.48,
      "learning_rate": 1.0468867313915859e-05,
      "loss": 0.0003,
      "step": 184070
    },
    {
      "epoch": 71.49,
      "learning_rate": 1.0468349514563108e-05,
      "loss": 0.0754,
      "step": 184080
    },
    {
      "epoch": 71.49,
      "learning_rate": 1.0467831715210356e-05,
      "loss": 0.1469,
      "step": 184090
    },
    {
      "epoch": 71.5,
      "learning_rate": 1.0467313915857606e-05,
      "loss": 0.0004,
      "step": 184100
    },
    {
      "epoch": 71.5,
      "learning_rate": 1.0466796116504856e-05,
      "loss": 0.0313,
      "step": 184110
    },
    {
      "epoch": 71.5,
      "learning_rate": 1.0466278317152105e-05,
      "loss": 0.0152,
      "step": 184120
    },
    {
      "epoch": 71.51,
      "learning_rate": 1.0465760517799353e-05,
      "loss": 0.0714,
      "step": 184130
    },
    {
      "epoch": 71.51,
      "learning_rate": 1.0465242718446603e-05,
      "loss": 0.0439,
      "step": 184140
    },
    {
      "epoch": 71.51,
      "learning_rate": 1.0464724919093852e-05,
      "loss": 0.0658,
      "step": 184150
    },
    {
      "epoch": 71.52,
      "learning_rate": 1.04642071197411e-05,
      "loss": 0.0524,
      "step": 184160
    },
    {
      "epoch": 71.52,
      "learning_rate": 1.046368932038835e-05,
      "loss": 0.046,
      "step": 184170
    },
    {
      "epoch": 71.53,
      "learning_rate": 1.04631715210356e-05,
      "loss": 0.0274,
      "step": 184180
    },
    {
      "epoch": 71.53,
      "learning_rate": 1.0462653721682849e-05,
      "loss": 0.021,
      "step": 184190
    },
    {
      "epoch": 71.53,
      "learning_rate": 1.0462135922330099e-05,
      "loss": 0.1134,
      "step": 184200
    },
    {
      "epoch": 71.54,
      "learning_rate": 1.0461618122977347e-05,
      "loss": 0.0833,
      "step": 184210
    },
    {
      "epoch": 71.54,
      "learning_rate": 1.0461100323624596e-05,
      "loss": 0.0137,
      "step": 184220
    },
    {
      "epoch": 71.55,
      "learning_rate": 1.0460582524271846e-05,
      "loss": 0.1086,
      "step": 184230
    },
    {
      "epoch": 71.55,
      "learning_rate": 1.0460064724919094e-05,
      "loss": 0.0032,
      "step": 184240
    },
    {
      "epoch": 71.55,
      "learning_rate": 1.0459546925566343e-05,
      "loss": 0.0111,
      "step": 184250
    },
    {
      "epoch": 71.56,
      "learning_rate": 1.0459029126213593e-05,
      "loss": 0.0029,
      "step": 184260
    },
    {
      "epoch": 71.56,
      "learning_rate": 1.0458511326860843e-05,
      "loss": 0.0754,
      "step": 184270
    },
    {
      "epoch": 71.57,
      "learning_rate": 1.0457993527508092e-05,
      "loss": 0.0852,
      "step": 184280
    },
    {
      "epoch": 71.57,
      "learning_rate": 1.045747572815534e-05,
      "loss": 0.1493,
      "step": 184290
    },
    {
      "epoch": 71.57,
      "learning_rate": 1.045695792880259e-05,
      "loss": 0.1759,
      "step": 184300
    },
    {
      "epoch": 71.58,
      "learning_rate": 1.0456440129449838e-05,
      "loss": 0.0454,
      "step": 184310
    },
    {
      "epoch": 71.58,
      "learning_rate": 1.0455922330097087e-05,
      "loss": 0.0738,
      "step": 184320
    },
    {
      "epoch": 71.58,
      "learning_rate": 1.0455404530744337e-05,
      "loss": 0.1595,
      "step": 184330
    },
    {
      "epoch": 71.59,
      "learning_rate": 1.0454886731391587e-05,
      "loss": 0.018,
      "step": 184340
    },
    {
      "epoch": 71.59,
      "learning_rate": 1.0454368932038836e-05,
      "loss": 0.1258,
      "step": 184350
    },
    {
      "epoch": 71.6,
      "learning_rate": 1.0453851132686086e-05,
      "loss": 0.0036,
      "step": 184360
    },
    {
      "epoch": 71.6,
      "learning_rate": 1.0453333333333334e-05,
      "loss": 0.1143,
      "step": 184370
    },
    {
      "epoch": 71.6,
      "learning_rate": 1.0452815533980583e-05,
      "loss": 0.0166,
      "step": 184380
    },
    {
      "epoch": 71.61,
      "learning_rate": 1.0452297734627831e-05,
      "loss": 0.0703,
      "step": 184390
    },
    {
      "epoch": 71.61,
      "learning_rate": 1.0451779935275081e-05,
      "loss": 0.0428,
      "step": 184400
    },
    {
      "epoch": 71.62,
      "learning_rate": 1.045126213592233e-05,
      "loss": 0.0933,
      "step": 184410
    },
    {
      "epoch": 71.62,
      "learning_rate": 1.045074433656958e-05,
      "loss": 0.0172,
      "step": 184420
    },
    {
      "epoch": 71.62,
      "learning_rate": 1.045022653721683e-05,
      "loss": 0.0409,
      "step": 184430
    },
    {
      "epoch": 71.63,
      "learning_rate": 1.044970873786408e-05,
      "loss": 0.081,
      "step": 184440
    },
    {
      "epoch": 71.63,
      "learning_rate": 1.0449190938511329e-05,
      "loss": 0.1183,
      "step": 184450
    },
    {
      "epoch": 71.63,
      "learning_rate": 1.0448673139158575e-05,
      "loss": 0.0407,
      "step": 184460
    },
    {
      "epoch": 71.64,
      "learning_rate": 1.0448155339805825e-05,
      "loss": 0.0443,
      "step": 184470
    },
    {
      "epoch": 71.64,
      "learning_rate": 1.0447637540453074e-05,
      "loss": 0.1215,
      "step": 184480
    },
    {
      "epoch": 71.65,
      "learning_rate": 1.0447119741100324e-05,
      "loss": 0.0415,
      "step": 184490
    },
    {
      "epoch": 71.65,
      "learning_rate": 1.0446601941747574e-05,
      "loss": 0.0436,
      "step": 184500
    },
    {
      "epoch": 71.65,
      "learning_rate": 1.0446084142394823e-05,
      "loss": 0.0307,
      "step": 184510
    },
    {
      "epoch": 71.66,
      "learning_rate": 1.0445566343042073e-05,
      "loss": 0.1242,
      "step": 184520
    },
    {
      "epoch": 71.66,
      "learning_rate": 1.0445048543689323e-05,
      "loss": 0.0444,
      "step": 184530
    },
    {
      "epoch": 71.67,
      "learning_rate": 1.0444530744336569e-05,
      "loss": 0.004,
      "step": 184540
    },
    {
      "epoch": 71.67,
      "learning_rate": 1.0444012944983818e-05,
      "loss": 0.1965,
      "step": 184550
    },
    {
      "epoch": 71.67,
      "learning_rate": 1.0443495145631068e-05,
      "loss": 0.0458,
      "step": 184560
    },
    {
      "epoch": 71.68,
      "learning_rate": 1.0442977346278318e-05,
      "loss": 0.0444,
      "step": 184570
    },
    {
      "epoch": 71.68,
      "learning_rate": 1.0442459546925567e-05,
      "loss": 0.0422,
      "step": 184580
    },
    {
      "epoch": 71.69,
      "learning_rate": 1.0441941747572817e-05,
      "loss": 0.1339,
      "step": 184590
    },
    {
      "epoch": 71.69,
      "learning_rate": 1.0441423948220067e-05,
      "loss": 0.0275,
      "step": 184600
    },
    {
      "epoch": 71.69,
      "learning_rate": 1.0440906148867316e-05,
      "loss": 0.0221,
      "step": 184610
    },
    {
      "epoch": 71.7,
      "learning_rate": 1.0440388349514562e-05,
      "loss": 0.006,
      "step": 184620
    },
    {
      "epoch": 71.7,
      "learning_rate": 1.0439870550161812e-05,
      "loss": 0.0135,
      "step": 184630
    },
    {
      "epoch": 71.7,
      "learning_rate": 1.0439352750809062e-05,
      "loss": 0.0111,
      "step": 184640
    },
    {
      "epoch": 71.71,
      "learning_rate": 1.0438834951456311e-05,
      "loss": 0.0343,
      "step": 184650
    },
    {
      "epoch": 71.71,
      "learning_rate": 1.043831715210356e-05,
      "loss": 0.0217,
      "step": 184660
    },
    {
      "epoch": 71.72,
      "learning_rate": 1.043779935275081e-05,
      "loss": 0.0053,
      "step": 184670
    },
    {
      "epoch": 71.72,
      "learning_rate": 1.043728155339806e-05,
      "loss": 0.0433,
      "step": 184680
    },
    {
      "epoch": 71.72,
      "learning_rate": 1.043676375404531e-05,
      "loss": 0.0014,
      "step": 184690
    },
    {
      "epoch": 71.73,
      "learning_rate": 1.0436245954692556e-05,
      "loss": 0.0664,
      "step": 184700
    },
    {
      "epoch": 71.73,
      "learning_rate": 1.0435728155339806e-05,
      "loss": 0.0137,
      "step": 184710
    },
    {
      "epoch": 71.74,
      "learning_rate": 1.0435210355987055e-05,
      "loss": 0.0012,
      "step": 184720
    },
    {
      "epoch": 71.74,
      "learning_rate": 1.0434692556634305e-05,
      "loss": 0.1119,
      "step": 184730
    },
    {
      "epoch": 71.74,
      "learning_rate": 1.0434174757281554e-05,
      "loss": 0.0428,
      "step": 184740
    },
    {
      "epoch": 71.75,
      "learning_rate": 1.0433656957928804e-05,
      "loss": 0.0241,
      "step": 184750
    },
    {
      "epoch": 71.75,
      "learning_rate": 1.0433139158576054e-05,
      "loss": 0.067,
      "step": 184760
    },
    {
      "epoch": 71.76,
      "learning_rate": 1.0432621359223303e-05,
      "loss": 0.1071,
      "step": 184770
    },
    {
      "epoch": 71.76,
      "learning_rate": 1.043210355987055e-05,
      "loss": 0.0458,
      "step": 184780
    },
    {
      "epoch": 71.76,
      "learning_rate": 1.0431585760517799e-05,
      "loss": 0.0056,
      "step": 184790
    },
    {
      "epoch": 71.77,
      "learning_rate": 1.0431067961165049e-05,
      "loss": 0.0578,
      "step": 184800
    },
    {
      "epoch": 71.77,
      "learning_rate": 1.0430550161812298e-05,
      "loss": 0.0326,
      "step": 184810
    },
    {
      "epoch": 71.77,
      "learning_rate": 1.0430032362459548e-05,
      "loss": 0.0734,
      "step": 184820
    },
    {
      "epoch": 71.78,
      "learning_rate": 1.0429514563106798e-05,
      "loss": 0.0234,
      "step": 184830
    },
    {
      "epoch": 71.78,
      "learning_rate": 1.0428996763754047e-05,
      "loss": 0.0108,
      "step": 184840
    },
    {
      "epoch": 71.79,
      "learning_rate": 1.0428478964401297e-05,
      "loss": 0.0396,
      "step": 184850
    },
    {
      "epoch": 71.79,
      "learning_rate": 1.0427961165048543e-05,
      "loss": 0.0657,
      "step": 184860
    },
    {
      "epoch": 71.79,
      "learning_rate": 1.0427443365695793e-05,
      "loss": 0.0724,
      "step": 184870
    },
    {
      "epoch": 71.8,
      "learning_rate": 1.0426925566343042e-05,
      "loss": 0.1535,
      "step": 184880
    },
    {
      "epoch": 71.8,
      "learning_rate": 1.0426407766990292e-05,
      "loss": 0.1264,
      "step": 184890
    },
    {
      "epoch": 71.81,
      "learning_rate": 1.0425889967637541e-05,
      "loss": 0.0574,
      "step": 184900
    },
    {
      "epoch": 71.81,
      "learning_rate": 1.0425372168284791e-05,
      "loss": 0.049,
      "step": 184910
    },
    {
      "epoch": 71.81,
      "learning_rate": 1.042485436893204e-05,
      "loss": 0.0585,
      "step": 184920
    },
    {
      "epoch": 71.82,
      "learning_rate": 1.042433656957929e-05,
      "loss": 0.1314,
      "step": 184930
    },
    {
      "epoch": 71.82,
      "learning_rate": 1.0423818770226537e-05,
      "loss": 0.2103,
      "step": 184940
    },
    {
      "epoch": 71.83,
      "learning_rate": 1.0423300970873786e-05,
      "loss": 0.1295,
      "step": 184950
    },
    {
      "epoch": 71.83,
      "learning_rate": 1.0422783171521036e-05,
      "loss": 0.0577,
      "step": 184960
    },
    {
      "epoch": 71.83,
      "learning_rate": 1.0422265372168285e-05,
      "loss": 0.0942,
      "step": 184970
    },
    {
      "epoch": 71.84,
      "learning_rate": 1.0421747572815535e-05,
      "loss": 0.1349,
      "step": 184980
    },
    {
      "epoch": 71.84,
      "learning_rate": 1.0421229773462785e-05,
      "loss": 0.2063,
      "step": 184990
    },
    {
      "epoch": 71.84,
      "learning_rate": 1.0420711974110034e-05,
      "loss": 0.0368,
      "step": 185000
    },
    {
      "epoch": 71.85,
      "learning_rate": 1.0420194174757284e-05,
      "loss": 0.0146,
      "step": 185010
    },
    {
      "epoch": 71.85,
      "learning_rate": 1.0419676375404534e-05,
      "loss": 0.0134,
      "step": 185020
    },
    {
      "epoch": 71.86,
      "learning_rate": 1.041915857605178e-05,
      "loss": 0.046,
      "step": 185030
    },
    {
      "epoch": 71.86,
      "learning_rate": 1.041864077669903e-05,
      "loss": 0.0403,
      "step": 185040
    },
    {
      "epoch": 71.86,
      "learning_rate": 1.0418122977346279e-05,
      "loss": 0.1825,
      "step": 185050
    },
    {
      "epoch": 71.87,
      "learning_rate": 1.0417605177993529e-05,
      "loss": 0.1602,
      "step": 185060
    },
    {
      "epoch": 71.87,
      "learning_rate": 1.0417087378640778e-05,
      "loss": 0.107,
      "step": 185070
    },
    {
      "epoch": 71.88,
      "learning_rate": 1.0416569579288028e-05,
      "loss": 0.0002,
      "step": 185080
    },
    {
      "epoch": 71.88,
      "learning_rate": 1.0416051779935277e-05,
      "loss": 0.3094,
      "step": 185090
    },
    {
      "epoch": 71.88,
      "learning_rate": 1.0415533980582525e-05,
      "loss": 0.0132,
      "step": 185100
    },
    {
      "epoch": 71.89,
      "learning_rate": 1.0415016181229773e-05,
      "loss": 0.0222,
      "step": 185110
    },
    {
      "epoch": 71.89,
      "learning_rate": 1.0414498381877023e-05,
      "loss": 0.0764,
      "step": 185120
    },
    {
      "epoch": 71.9,
      "learning_rate": 1.0413980582524273e-05,
      "loss": 0.0111,
      "step": 185130
    },
    {
      "epoch": 71.9,
      "learning_rate": 1.0413462783171522e-05,
      "loss": 0.0795,
      "step": 185140
    },
    {
      "epoch": 71.9,
      "learning_rate": 1.0412944983818772e-05,
      "loss": 0.011,
      "step": 185150
    },
    {
      "epoch": 71.91,
      "learning_rate": 1.0412427184466021e-05,
      "loss": 0.0429,
      "step": 185160
    },
    {
      "epoch": 71.91,
      "learning_rate": 1.0411909385113271e-05,
      "loss": 0.0396,
      "step": 185170
    },
    {
      "epoch": 71.91,
      "learning_rate": 1.0411391585760519e-05,
      "loss": 0.1042,
      "step": 185180
    },
    {
      "epoch": 71.92,
      "learning_rate": 1.0410873786407767e-05,
      "loss": 0.1204,
      "step": 185190
    },
    {
      "epoch": 71.92,
      "learning_rate": 1.0410355987055016e-05,
      "loss": 0.0679,
      "step": 185200
    },
    {
      "epoch": 71.93,
      "learning_rate": 1.0409838187702266e-05,
      "loss": 0.0514,
      "step": 185210
    },
    {
      "epoch": 71.93,
      "learning_rate": 1.0409320388349516e-05,
      "loss": 0.0622,
      "step": 185220
    },
    {
      "epoch": 71.93,
      "learning_rate": 1.0408802588996765e-05,
      "loss": 0.047,
      "step": 185230
    },
    {
      "epoch": 71.94,
      "learning_rate": 1.0408284789644015e-05,
      "loss": 0.0006,
      "step": 185240
    },
    {
      "epoch": 71.94,
      "learning_rate": 1.0407766990291263e-05,
      "loss": 0.004,
      "step": 185250
    },
    {
      "epoch": 71.95,
      "learning_rate": 1.0407249190938513e-05,
      "loss": 0.3037,
      "step": 185260
    },
    {
      "epoch": 71.95,
      "learning_rate": 1.040673139158576e-05,
      "loss": 0.1149,
      "step": 185270
    },
    {
      "epoch": 71.95,
      "learning_rate": 1.040621359223301e-05,
      "loss": 0.064,
      "step": 185280
    },
    {
      "epoch": 71.96,
      "learning_rate": 1.040569579288026e-05,
      "loss": 0.0004,
      "step": 185290
    },
    {
      "epoch": 71.96,
      "learning_rate": 1.040517799352751e-05,
      "loss": 0.0348,
      "step": 185300
    },
    {
      "epoch": 71.97,
      "learning_rate": 1.0404660194174759e-05,
      "loss": 0.063,
      "step": 185310
    },
    {
      "epoch": 71.97,
      "learning_rate": 1.0404142394822009e-05,
      "loss": 0.0506,
      "step": 185320
    },
    {
      "epoch": 71.97,
      "learning_rate": 1.0403624595469256e-05,
      "loss": 0.0897,
      "step": 185330
    },
    {
      "epoch": 71.98,
      "learning_rate": 1.0403106796116506e-05,
      "loss": 0.1156,
      "step": 185340
    },
    {
      "epoch": 71.98,
      "learning_rate": 1.0402588996763754e-05,
      "loss": 0.0018,
      "step": 185350
    },
    {
      "epoch": 71.98,
      "learning_rate": 1.0402071197411004e-05,
      "loss": 0.1471,
      "step": 185360
    },
    {
      "epoch": 71.99,
      "learning_rate": 1.0401553398058253e-05,
      "loss": 0.0025,
      "step": 185370
    },
    {
      "epoch": 71.99,
      "learning_rate": 1.0401035598705503e-05,
      "loss": 0.1567,
      "step": 185380
    },
    {
      "epoch": 72.0,
      "learning_rate": 1.0400517799352752e-05,
      "loss": 0.0831,
      "step": 185390
    },
    {
      "epoch": 72.0,
      "learning_rate": 1.04e-05,
      "loss": 0.0264,
      "step": 185400
    },
    {
      "epoch": 72.0,
      "eval_accuracy": 0.9477303988995873,
      "eval_loss": 0.35998329520225525,
      "eval_runtime": 8.2027,
      "eval_samples_per_second": 443.148,
      "eval_steps_per_second": 55.47,
      "step": 185400
    },
    {
      "epoch": 72.0,
      "learning_rate": 1.039948220064725e-05,
      "loss": 0.012,
      "step": 185410
    },
    {
      "epoch": 72.01,
      "learning_rate": 1.03989644012945e-05,
      "loss": 0.0154,
      "step": 185420
    },
    {
      "epoch": 72.01,
      "learning_rate": 1.0398446601941748e-05,
      "loss": 0.0256,
      "step": 185430
    },
    {
      "epoch": 72.02,
      "learning_rate": 1.0397928802588997e-05,
      "loss": 0.0205,
      "step": 185440
    },
    {
      "epoch": 72.02,
      "learning_rate": 1.0397411003236247e-05,
      "loss": 0.0003,
      "step": 185450
    },
    {
      "epoch": 72.02,
      "learning_rate": 1.0396893203883496e-05,
      "loss": 0.059,
      "step": 185460
    },
    {
      "epoch": 72.03,
      "learning_rate": 1.0396375404530746e-05,
      "loss": 0.2431,
      "step": 185470
    },
    {
      "epoch": 72.03,
      "learning_rate": 1.0395857605177994e-05,
      "loss": 0.1409,
      "step": 185480
    },
    {
      "epoch": 72.03,
      "learning_rate": 1.0395339805825244e-05,
      "loss": 0.053,
      "step": 185490
    },
    {
      "epoch": 72.04,
      "learning_rate": 1.0394822006472493e-05,
      "loss": 0.0015,
      "step": 185500
    },
    {
      "epoch": 72.04,
      "learning_rate": 1.0394304207119741e-05,
      "loss": 0.1503,
      "step": 185510
    },
    {
      "epoch": 72.05,
      "learning_rate": 1.039378640776699e-05,
      "loss": 0.1318,
      "step": 185520
    },
    {
      "epoch": 72.05,
      "learning_rate": 1.039326860841424e-05,
      "loss": 0.0409,
      "step": 185530
    },
    {
      "epoch": 72.05,
      "learning_rate": 1.039275080906149e-05,
      "loss": 0.1381,
      "step": 185540
    },
    {
      "epoch": 72.06,
      "learning_rate": 1.0392233009708738e-05,
      "loss": 0.0198,
      "step": 185550
    },
    {
      "epoch": 72.06,
      "learning_rate": 1.0391715210355987e-05,
      "loss": 0.1336,
      "step": 185560
    },
    {
      "epoch": 72.07,
      "learning_rate": 1.0391197411003237e-05,
      "loss": 0.0395,
      "step": 185570
    },
    {
      "epoch": 72.07,
      "learning_rate": 1.0390679611650487e-05,
      "loss": 0.0347,
      "step": 185580
    },
    {
      "epoch": 72.07,
      "learning_rate": 1.0390161812297736e-05,
      "loss": 0.1868,
      "step": 185590
    },
    {
      "epoch": 72.08,
      "learning_rate": 1.0389644012944984e-05,
      "loss": 0.1334,
      "step": 185600
    },
    {
      "epoch": 72.08,
      "learning_rate": 1.0389126213592234e-05,
      "loss": 0.0394,
      "step": 185610
    },
    {
      "epoch": 72.09,
      "learning_rate": 1.0388608414239484e-05,
      "loss": 0.001,
      "step": 185620
    },
    {
      "epoch": 72.09,
      "learning_rate": 1.0388090614886731e-05,
      "loss": 0.0937,
      "step": 185630
    },
    {
      "epoch": 72.09,
      "learning_rate": 1.0387572815533981e-05,
      "loss": 0.0803,
      "step": 185640
    },
    {
      "epoch": 72.1,
      "learning_rate": 1.038705501618123e-05,
      "loss": 0.0173,
      "step": 185650
    },
    {
      "epoch": 72.1,
      "learning_rate": 1.038653721682848e-05,
      "loss": 0.0562,
      "step": 185660
    },
    {
      "epoch": 72.1,
      "learning_rate": 1.038601941747573e-05,
      "loss": 0.0802,
      "step": 185670
    },
    {
      "epoch": 72.11,
      "learning_rate": 1.0385501618122978e-05,
      "loss": 0.0299,
      "step": 185680
    },
    {
      "epoch": 72.11,
      "learning_rate": 1.0384983818770227e-05,
      "loss": 0.0973,
      "step": 185690
    },
    {
      "epoch": 72.12,
      "learning_rate": 1.0384466019417477e-05,
      "loss": 0.0617,
      "step": 185700
    },
    {
      "epoch": 72.12,
      "learning_rate": 1.0383948220064725e-05,
      "loss": 0.0018,
      "step": 185710
    },
    {
      "epoch": 72.12,
      "learning_rate": 1.0383430420711975e-05,
      "loss": 0.0608,
      "step": 185720
    },
    {
      "epoch": 72.13,
      "learning_rate": 1.0382912621359224e-05,
      "loss": 0.1126,
      "step": 185730
    },
    {
      "epoch": 72.13,
      "learning_rate": 1.0382394822006474e-05,
      "loss": 0.0452,
      "step": 185740
    },
    {
      "epoch": 72.14,
      "learning_rate": 1.0381877022653723e-05,
      "loss": 0.0945,
      "step": 185750
    },
    {
      "epoch": 72.14,
      "learning_rate": 1.0381359223300971e-05,
      "loss": 0.0438,
      "step": 185760
    },
    {
      "epoch": 72.14,
      "learning_rate": 1.0380841423948221e-05,
      "loss": 0.0061,
      "step": 185770
    },
    {
      "epoch": 72.15,
      "learning_rate": 1.0380323624595469e-05,
      "loss": 0.0968,
      "step": 185780
    },
    {
      "epoch": 72.15,
      "learning_rate": 1.0379805825242719e-05,
      "loss": 0.0004,
      "step": 185790
    },
    {
      "epoch": 72.16,
      "learning_rate": 1.0379288025889968e-05,
      "loss": 0.0219,
      "step": 185800
    },
    {
      "epoch": 72.16,
      "learning_rate": 1.0378770226537218e-05,
      "loss": 0.0759,
      "step": 185810
    },
    {
      "epoch": 72.16,
      "learning_rate": 1.0378252427184467e-05,
      "loss": 0.062,
      "step": 185820
    },
    {
      "epoch": 72.17,
      "learning_rate": 1.0377734627831717e-05,
      "loss": 0.0478,
      "step": 185830
    },
    {
      "epoch": 72.17,
      "learning_rate": 1.0377216828478965e-05,
      "loss": 0.0992,
      "step": 185840
    },
    {
      "epoch": 72.17,
      "learning_rate": 1.0376699029126215e-05,
      "loss": 0.0819,
      "step": 185850
    },
    {
      "epoch": 72.18,
      "learning_rate": 1.0376181229773462e-05,
      "loss": 0.0184,
      "step": 185860
    },
    {
      "epoch": 72.18,
      "learning_rate": 1.0375663430420712e-05,
      "loss": 0.0745,
      "step": 185870
    },
    {
      "epoch": 72.19,
      "learning_rate": 1.0375145631067962e-05,
      "loss": 0.0498,
      "step": 185880
    },
    {
      "epoch": 72.19,
      "learning_rate": 1.0374627831715211e-05,
      "loss": 0.0868,
      "step": 185890
    },
    {
      "epoch": 72.19,
      "learning_rate": 1.0374110032362461e-05,
      "loss": 0.0203,
      "step": 185900
    },
    {
      "epoch": 72.2,
      "learning_rate": 1.037359223300971e-05,
      "loss": 0.0886,
      "step": 185910
    },
    {
      "epoch": 72.2,
      "learning_rate": 1.0373074433656958e-05,
      "loss": 0.019,
      "step": 185920
    },
    {
      "epoch": 72.21,
      "learning_rate": 1.0372556634304206e-05,
      "loss": 0.0766,
      "step": 185930
    },
    {
      "epoch": 72.21,
      "learning_rate": 1.0372038834951456e-05,
      "loss": 0.0652,
      "step": 185940
    },
    {
      "epoch": 72.21,
      "learning_rate": 1.0371521035598706e-05,
      "loss": 0.0824,
      "step": 185950
    },
    {
      "epoch": 72.22,
      "learning_rate": 1.0371003236245955e-05,
      "loss": 0.002,
      "step": 185960
    },
    {
      "epoch": 72.22,
      "learning_rate": 1.0370485436893205e-05,
      "loss": 0.0251,
      "step": 185970
    },
    {
      "epoch": 72.23,
      "learning_rate": 1.0369967637540455e-05,
      "loss": 0.0765,
      "step": 185980
    },
    {
      "epoch": 72.23,
      "learning_rate": 1.0369449838187704e-05,
      "loss": 0.021,
      "step": 185990
    },
    {
      "epoch": 72.23,
      "learning_rate": 1.0368932038834952e-05,
      "loss": 0.0004,
      "step": 186000
    },
    {
      "epoch": 72.24,
      "learning_rate": 1.03684142394822e-05,
      "loss": 0.0152,
      "step": 186010
    },
    {
      "epoch": 72.24,
      "learning_rate": 1.036789644012945e-05,
      "loss": 0.0196,
      "step": 186020
    },
    {
      "epoch": 72.24,
      "learning_rate": 1.03673786407767e-05,
      "loss": 0.1437,
      "step": 186030
    },
    {
      "epoch": 72.25,
      "learning_rate": 1.0366860841423949e-05,
      "loss": 0.1571,
      "step": 186040
    },
    {
      "epoch": 72.25,
      "learning_rate": 1.0366343042071198e-05,
      "loss": 0.0752,
      "step": 186050
    },
    {
      "epoch": 72.26,
      "learning_rate": 1.0365825242718448e-05,
      "loss": 0.174,
      "step": 186060
    },
    {
      "epoch": 72.26,
      "learning_rate": 1.0365307443365698e-05,
      "loss": 0.1298,
      "step": 186070
    },
    {
      "epoch": 72.26,
      "learning_rate": 1.0364789644012944e-05,
      "loss": 0.0466,
      "step": 186080
    },
    {
      "epoch": 72.27,
      "learning_rate": 1.0364271844660194e-05,
      "loss": 0.0999,
      "step": 186090
    },
    {
      "epoch": 72.27,
      "learning_rate": 1.0363754045307443e-05,
      "loss": 0.081,
      "step": 186100
    },
    {
      "epoch": 72.28,
      "learning_rate": 1.0363236245954693e-05,
      "loss": 0.0827,
      "step": 186110
    },
    {
      "epoch": 72.28,
      "learning_rate": 1.0362718446601942e-05,
      "loss": 0.0459,
      "step": 186120
    },
    {
      "epoch": 72.28,
      "learning_rate": 1.0362200647249192e-05,
      "loss": 0.1077,
      "step": 186130
    },
    {
      "epoch": 72.29,
      "learning_rate": 1.0361682847896442e-05,
      "loss": 0.1913,
      "step": 186140
    },
    {
      "epoch": 72.29,
      "learning_rate": 1.0361165048543691e-05,
      "loss": 0.1541,
      "step": 186150
    },
    {
      "epoch": 72.3,
      "learning_rate": 1.0360647249190941e-05,
      "loss": 0.0006,
      "step": 186160
    },
    {
      "epoch": 72.3,
      "learning_rate": 1.0360129449838187e-05,
      "loss": 0.1568,
      "step": 186170
    },
    {
      "epoch": 72.3,
      "learning_rate": 1.0359611650485437e-05,
      "loss": 0.0118,
      "step": 186180
    },
    {
      "epoch": 72.31,
      "learning_rate": 1.0359093851132686e-05,
      "loss": 0.0012,
      "step": 186190
    },
    {
      "epoch": 72.31,
      "learning_rate": 1.0358576051779936e-05,
      "loss": 0.1331,
      "step": 186200
    },
    {
      "epoch": 72.31,
      "learning_rate": 1.0358058252427186e-05,
      "loss": 0.0121,
      "step": 186210
    },
    {
      "epoch": 72.32,
      "learning_rate": 1.0357540453074435e-05,
      "loss": 0.0435,
      "step": 186220
    },
    {
      "epoch": 72.32,
      "learning_rate": 1.0357022653721685e-05,
      "loss": 0.0252,
      "step": 186230
    },
    {
      "epoch": 72.33,
      "learning_rate": 1.0356504854368934e-05,
      "loss": 0.025,
      "step": 186240
    },
    {
      "epoch": 72.33,
      "learning_rate": 1.035598705501618e-05,
      "loss": 0.0953,
      "step": 186250
    },
    {
      "epoch": 72.33,
      "learning_rate": 1.035546925566343e-05,
      "loss": 0.0324,
      "step": 186260
    },
    {
      "epoch": 72.34,
      "learning_rate": 1.035495145631068e-05,
      "loss": 0.0173,
      "step": 186270
    },
    {
      "epoch": 72.34,
      "learning_rate": 1.035443365695793e-05,
      "loss": 0.0352,
      "step": 186280
    },
    {
      "epoch": 72.35,
      "learning_rate": 1.0353915857605179e-05,
      "loss": 0.0012,
      "step": 186290
    },
    {
      "epoch": 72.35,
      "learning_rate": 1.0353398058252429e-05,
      "loss": 0.1352,
      "step": 186300
    },
    {
      "epoch": 72.35,
      "learning_rate": 1.0352880258899678e-05,
      "loss": 0.0297,
      "step": 186310
    },
    {
      "epoch": 72.36,
      "learning_rate": 1.0352362459546928e-05,
      "loss": 0.0697,
      "step": 186320
    },
    {
      "epoch": 72.36,
      "learning_rate": 1.0351844660194174e-05,
      "loss": 0.0503,
      "step": 186330
    },
    {
      "epoch": 72.37,
      "learning_rate": 1.0351326860841424e-05,
      "loss": 0.0168,
      "step": 186340
    },
    {
      "epoch": 72.37,
      "learning_rate": 1.0350809061488673e-05,
      "loss": 0.0059,
      "step": 186350
    },
    {
      "epoch": 72.37,
      "learning_rate": 1.0350291262135923e-05,
      "loss": 0.0308,
      "step": 186360
    },
    {
      "epoch": 72.38,
      "learning_rate": 1.0349773462783173e-05,
      "loss": 0.0633,
      "step": 186370
    },
    {
      "epoch": 72.38,
      "learning_rate": 1.0349255663430422e-05,
      "loss": 0.1008,
      "step": 186380
    },
    {
      "epoch": 72.38,
      "learning_rate": 1.0348737864077672e-05,
      "loss": 0.143,
      "step": 186390
    },
    {
      "epoch": 72.39,
      "learning_rate": 1.0348220064724922e-05,
      "loss": 0.0224,
      "step": 186400
    },
    {
      "epoch": 72.39,
      "learning_rate": 1.0347702265372168e-05,
      "loss": 0.0025,
      "step": 186410
    },
    {
      "epoch": 72.4,
      "learning_rate": 1.0347184466019417e-05,
      "loss": 0.0117,
      "step": 186420
    },
    {
      "epoch": 72.4,
      "learning_rate": 1.0346666666666667e-05,
      "loss": 0.0591,
      "step": 186430
    },
    {
      "epoch": 72.4,
      "learning_rate": 1.0346148867313917e-05,
      "loss": 0.0856,
      "step": 186440
    },
    {
      "epoch": 72.41,
      "learning_rate": 1.0345631067961166e-05,
      "loss": 0.0878,
      "step": 186450
    },
    {
      "epoch": 72.41,
      "learning_rate": 1.0345113268608416e-05,
      "loss": 0.0257,
      "step": 186460
    },
    {
      "epoch": 72.42,
      "learning_rate": 1.0344595469255665e-05,
      "loss": 0.0579,
      "step": 186470
    },
    {
      "epoch": 72.42,
      "learning_rate": 1.0344077669902915e-05,
      "loss": 0.0475,
      "step": 186480
    },
    {
      "epoch": 72.42,
      "learning_rate": 1.0343559870550161e-05,
      "loss": 0.0135,
      "step": 186490
    },
    {
      "epoch": 72.43,
      "learning_rate": 1.0343042071197411e-05,
      "loss": 0.1235,
      "step": 186500
    },
    {
      "epoch": 72.43,
      "learning_rate": 1.034252427184466e-05,
      "loss": 0.0621,
      "step": 186510
    },
    {
      "epoch": 72.43,
      "learning_rate": 1.034200647249191e-05,
      "loss": 0.0661,
      "step": 186520
    },
    {
      "epoch": 72.44,
      "learning_rate": 1.034148867313916e-05,
      "loss": 0.0105,
      "step": 186530
    },
    {
      "epoch": 72.44,
      "learning_rate": 1.034097087378641e-05,
      "loss": 0.0774,
      "step": 186540
    },
    {
      "epoch": 72.45,
      "learning_rate": 1.0340453074433659e-05,
      "loss": 0.1572,
      "step": 186550
    },
    {
      "epoch": 72.45,
      "learning_rate": 1.0339935275080909e-05,
      "loss": 0.0076,
      "step": 186560
    },
    {
      "epoch": 72.45,
      "learning_rate": 1.0339417475728155e-05,
      "loss": 0.0705,
      "step": 186570
    },
    {
      "epoch": 72.46,
      "learning_rate": 1.0338899676375404e-05,
      "loss": 0.1513,
      "step": 186580
    },
    {
      "epoch": 72.46,
      "learning_rate": 1.0338381877022654e-05,
      "loss": 0.0628,
      "step": 186590
    },
    {
      "epoch": 72.47,
      "learning_rate": 1.0337864077669904e-05,
      "loss": 0.0977,
      "step": 186600
    },
    {
      "epoch": 72.47,
      "learning_rate": 1.0337346278317153e-05,
      "loss": 0.0234,
      "step": 186610
    },
    {
      "epoch": 72.47,
      "learning_rate": 1.0336828478964403e-05,
      "loss": 0.1736,
      "step": 186620
    },
    {
      "epoch": 72.48,
      "learning_rate": 1.0336310679611653e-05,
      "loss": 0.0636,
      "step": 186630
    },
    {
      "epoch": 72.48,
      "learning_rate": 1.0335792880258902e-05,
      "loss": 0.09,
      "step": 186640
    },
    {
      "epoch": 72.49,
      "learning_rate": 1.0335275080906148e-05,
      "loss": 0.0268,
      "step": 186650
    },
    {
      "epoch": 72.49,
      "learning_rate": 1.0334757281553398e-05,
      "loss": 0.1216,
      "step": 186660
    },
    {
      "epoch": 72.49,
      "learning_rate": 1.0334239482200648e-05,
      "loss": 0.1387,
      "step": 186670
    },
    {
      "epoch": 72.5,
      "learning_rate": 1.0333721682847897e-05,
      "loss": 0.0194,
      "step": 186680
    },
    {
      "epoch": 72.5,
      "learning_rate": 1.0333203883495147e-05,
      "loss": 0.0673,
      "step": 186690
    },
    {
      "epoch": 72.5,
      "learning_rate": 1.0332686084142397e-05,
      "loss": 0.0588,
      "step": 186700
    },
    {
      "epoch": 72.51,
      "learning_rate": 1.0332168284789646e-05,
      "loss": 0.0056,
      "step": 186710
    },
    {
      "epoch": 72.51,
      "learning_rate": 1.0331650485436894e-05,
      "loss": 0.0968,
      "step": 186720
    },
    {
      "epoch": 72.52,
      "learning_rate": 1.0331132686084144e-05,
      "loss": 0.2472,
      "step": 186730
    },
    {
      "epoch": 72.52,
      "learning_rate": 1.0330614886731392e-05,
      "loss": 0.0999,
      "step": 186740
    },
    {
      "epoch": 72.52,
      "learning_rate": 1.0330097087378641e-05,
      "loss": 0.063,
      "step": 186750
    },
    {
      "epoch": 72.53,
      "learning_rate": 1.032957928802589e-05,
      "loss": 0.1234,
      "step": 186760
    },
    {
      "epoch": 72.53,
      "learning_rate": 1.032906148867314e-05,
      "loss": 0.0214,
      "step": 186770
    },
    {
      "epoch": 72.54,
      "learning_rate": 1.032854368932039e-05,
      "loss": 0.0634,
      "step": 186780
    },
    {
      "epoch": 72.54,
      "learning_rate": 1.032802588996764e-05,
      "loss": 0.0329,
      "step": 186790
    },
    {
      "epoch": 72.54,
      "learning_rate": 1.0327508090614888e-05,
      "loss": 0.0115,
      "step": 186800
    },
    {
      "epoch": 72.55,
      "learning_rate": 1.0326990291262137e-05,
      "loss": 0.0087,
      "step": 186810
    },
    {
      "epoch": 72.55,
      "learning_rate": 1.0326472491909385e-05,
      "loss": 0.0176,
      "step": 186820
    },
    {
      "epoch": 72.56,
      "learning_rate": 1.0325954692556635e-05,
      "loss": 0.0006,
      "step": 186830
    },
    {
      "epoch": 72.56,
      "learning_rate": 1.0325436893203884e-05,
      "loss": 0.0286,
      "step": 186840
    },
    {
      "epoch": 72.56,
      "learning_rate": 1.0324919093851134e-05,
      "loss": 0.028,
      "step": 186850
    },
    {
      "epoch": 72.57,
      "learning_rate": 1.0324401294498384e-05,
      "loss": 0.0523,
      "step": 186860
    },
    {
      "epoch": 72.57,
      "learning_rate": 1.0323883495145632e-05,
      "loss": 0.0101,
      "step": 186870
    },
    {
      "epoch": 72.57,
      "learning_rate": 1.0323365695792881e-05,
      "loss": 0.0007,
      "step": 186880
    },
    {
      "epoch": 72.58,
      "learning_rate": 1.032284789644013e-05,
      "loss": 0.0651,
      "step": 186890
    },
    {
      "epoch": 72.58,
      "learning_rate": 1.0322330097087379e-05,
      "loss": 0.1259,
      "step": 186900
    },
    {
      "epoch": 72.59,
      "learning_rate": 1.0321812297734628e-05,
      "loss": 0.1662,
      "step": 186910
    },
    {
      "epoch": 72.59,
      "learning_rate": 1.0321294498381878e-05,
      "loss": 0.1143,
      "step": 186920
    },
    {
      "epoch": 72.59,
      "learning_rate": 1.0320776699029128e-05,
      "loss": 0.1189,
      "step": 186930
    },
    {
      "epoch": 72.6,
      "learning_rate": 1.0320258899676377e-05,
      "loss": 0.1191,
      "step": 186940
    },
    {
      "epoch": 72.6,
      "learning_rate": 1.0319741100323625e-05,
      "loss": 0.0184,
      "step": 186950
    },
    {
      "epoch": 72.61,
      "learning_rate": 1.0319223300970875e-05,
      "loss": 0.0741,
      "step": 186960
    },
    {
      "epoch": 72.61,
      "learning_rate": 1.0318705501618124e-05,
      "loss": 0.0183,
      "step": 186970
    },
    {
      "epoch": 72.61,
      "learning_rate": 1.0318187702265372e-05,
      "loss": 0.0281,
      "step": 186980
    },
    {
      "epoch": 72.62,
      "learning_rate": 1.0317669902912622e-05,
      "loss": 0.1602,
      "step": 186990
    },
    {
      "epoch": 72.62,
      "learning_rate": 1.0317152103559872e-05,
      "loss": 0.0272,
      "step": 187000
    },
    {
      "epoch": 72.63,
      "learning_rate": 1.0316634304207121e-05,
      "loss": 0.0108,
      "step": 187010
    },
    {
      "epoch": 72.63,
      "learning_rate": 1.0316116504854369e-05,
      "loss": 0.0691,
      "step": 187020
    },
    {
      "epoch": 72.63,
      "learning_rate": 1.0315598705501619e-05,
      "loss": 0.0024,
      "step": 187030
    },
    {
      "epoch": 72.64,
      "learning_rate": 1.0315080906148868e-05,
      "loss": 0.06,
      "step": 187040
    },
    {
      "epoch": 72.64,
      "learning_rate": 1.0314563106796118e-05,
      "loss": 0.319,
      "step": 187050
    },
    {
      "epoch": 72.64,
      "learning_rate": 1.0314045307443366e-05,
      "loss": 0.1279,
      "step": 187060
    },
    {
      "epoch": 72.65,
      "learning_rate": 1.0313527508090615e-05,
      "loss": 0.0005,
      "step": 187070
    },
    {
      "epoch": 72.65,
      "learning_rate": 1.0313009708737865e-05,
      "loss": 0.1704,
      "step": 187080
    },
    {
      "epoch": 72.66,
      "learning_rate": 1.0312491909385115e-05,
      "loss": 0.0003,
      "step": 187090
    },
    {
      "epoch": 72.66,
      "learning_rate": 1.0311974110032363e-05,
      "loss": 0.0313,
      "step": 187100
    },
    {
      "epoch": 72.66,
      "learning_rate": 1.0311456310679612e-05,
      "loss": 0.0893,
      "step": 187110
    },
    {
      "epoch": 72.67,
      "learning_rate": 1.0310938511326862e-05,
      "loss": 0.0642,
      "step": 187120
    },
    {
      "epoch": 72.67,
      "learning_rate": 1.0310420711974111e-05,
      "loss": 0.0561,
      "step": 187130
    },
    {
      "epoch": 72.68,
      "learning_rate": 1.030990291262136e-05,
      "loss": 0.1443,
      "step": 187140
    },
    {
      "epoch": 72.68,
      "learning_rate": 1.0309385113268609e-05,
      "loss": 0.0937,
      "step": 187150
    },
    {
      "epoch": 72.68,
      "learning_rate": 1.0308867313915859e-05,
      "loss": 0.0704,
      "step": 187160
    },
    {
      "epoch": 72.69,
      "learning_rate": 1.0308349514563108e-05,
      "loss": 0.035,
      "step": 187170
    },
    {
      "epoch": 72.69,
      "learning_rate": 1.0307831715210356e-05,
      "loss": 0.002,
      "step": 187180
    },
    {
      "epoch": 72.7,
      "learning_rate": 1.0307313915857606e-05,
      "loss": 0.0466,
      "step": 187190
    },
    {
      "epoch": 72.7,
      "learning_rate": 1.0306796116504855e-05,
      "loss": 0.0887,
      "step": 187200
    },
    {
      "epoch": 72.7,
      "learning_rate": 1.0306278317152105e-05,
      "loss": 0.0885,
      "step": 187210
    },
    {
      "epoch": 72.71,
      "learning_rate": 1.0305760517799353e-05,
      "loss": 0.0317,
      "step": 187220
    },
    {
      "epoch": 72.71,
      "learning_rate": 1.0305242718446603e-05,
      "loss": 0.0175,
      "step": 187230
    },
    {
      "epoch": 72.71,
      "learning_rate": 1.0304724919093852e-05,
      "loss": 0.0332,
      "step": 187240
    },
    {
      "epoch": 72.72,
      "learning_rate": 1.03042071197411e-05,
      "loss": 0.0724,
      "step": 187250
    },
    {
      "epoch": 72.72,
      "learning_rate": 1.030368932038835e-05,
      "loss": 0.0011,
      "step": 187260
    },
    {
      "epoch": 72.73,
      "learning_rate": 1.03031715210356e-05,
      "loss": 0.0195,
      "step": 187270
    },
    {
      "epoch": 72.73,
      "learning_rate": 1.0302653721682849e-05,
      "loss": 0.0287,
      "step": 187280
    },
    {
      "epoch": 72.73,
      "learning_rate": 1.0302135922330099e-05,
      "loss": 0.0093,
      "step": 187290
    },
    {
      "epoch": 72.74,
      "learning_rate": 1.0301618122977348e-05,
      "loss": 0.0002,
      "step": 187300
    },
    {
      "epoch": 72.74,
      "learning_rate": 1.0301100323624596e-05,
      "loss": 0.0002,
      "step": 187310
    },
    {
      "epoch": 72.75,
      "learning_rate": 1.0300582524271846e-05,
      "loss": 0.017,
      "step": 187320
    },
    {
      "epoch": 72.75,
      "learning_rate": 1.0300064724919094e-05,
      "loss": 0.0954,
      "step": 187330
    },
    {
      "epoch": 72.75,
      "learning_rate": 1.0299546925566343e-05,
      "loss": 0.0691,
      "step": 187340
    },
    {
      "epoch": 72.76,
      "learning_rate": 1.0299029126213593e-05,
      "loss": 0.1751,
      "step": 187350
    },
    {
      "epoch": 72.76,
      "learning_rate": 1.0298511326860843e-05,
      "loss": 0.0569,
      "step": 187360
    },
    {
      "epoch": 72.77,
      "learning_rate": 1.0297993527508092e-05,
      "loss": 0.1162,
      "step": 187370
    },
    {
      "epoch": 72.77,
      "learning_rate": 1.0297475728155342e-05,
      "loss": 0.037,
      "step": 187380
    },
    {
      "epoch": 72.77,
      "learning_rate": 1.029695792880259e-05,
      "loss": 0.0659,
      "step": 187390
    },
    {
      "epoch": 72.78,
      "learning_rate": 1.0296440129449838e-05,
      "loss": 0.0379,
      "step": 187400
    },
    {
      "epoch": 72.78,
      "learning_rate": 1.0295922330097087e-05,
      "loss": 0.1346,
      "step": 187410
    },
    {
      "epoch": 72.78,
      "learning_rate": 1.0295404530744337e-05,
      "loss": 0.1368,
      "step": 187420
    },
    {
      "epoch": 72.79,
      "learning_rate": 1.0294886731391586e-05,
      "loss": 0.0507,
      "step": 187430
    },
    {
      "epoch": 72.79,
      "learning_rate": 1.0294368932038836e-05,
      "loss": 0.0462,
      "step": 187440
    },
    {
      "epoch": 72.8,
      "learning_rate": 1.0293851132686086e-05,
      "loss": 0.0005,
      "step": 187450
    },
    {
      "epoch": 72.8,
      "learning_rate": 1.0293333333333335e-05,
      "loss": 0.1229,
      "step": 187460
    },
    {
      "epoch": 72.8,
      "learning_rate": 1.0292815533980583e-05,
      "loss": 0.0613,
      "step": 187470
    },
    {
      "epoch": 72.81,
      "learning_rate": 1.0292297734627831e-05,
      "loss": 0.0976,
      "step": 187480
    },
    {
      "epoch": 72.81,
      "learning_rate": 1.029177993527508e-05,
      "loss": 0.0051,
      "step": 187490
    },
    {
      "epoch": 72.82,
      "learning_rate": 1.029126213592233e-05,
      "loss": 0.1282,
      "step": 187500
    },
    {
      "epoch": 72.82,
      "learning_rate": 1.029074433656958e-05,
      "loss": 0.0027,
      "step": 187510
    },
    {
      "epoch": 72.82,
      "learning_rate": 1.029022653721683e-05,
      "loss": 0.0386,
      "step": 187520
    },
    {
      "epoch": 72.83,
      "learning_rate": 1.028970873786408e-05,
      "loss": 0.0001,
      "step": 187530
    },
    {
      "epoch": 72.83,
      "learning_rate": 1.0289190938511329e-05,
      "loss": 0.021,
      "step": 187540
    },
    {
      "epoch": 72.83,
      "learning_rate": 1.0288673139158575e-05,
      "loss": 0.0176,
      "step": 187550
    },
    {
      "epoch": 72.84,
      "learning_rate": 1.0288155339805825e-05,
      "loss": 0.0138,
      "step": 187560
    },
    {
      "epoch": 72.84,
      "learning_rate": 1.0287637540453074e-05,
      "loss": 0.0177,
      "step": 187570
    },
    {
      "epoch": 72.85,
      "learning_rate": 1.0287119741100324e-05,
      "loss": 0.0177,
      "step": 187580
    },
    {
      "epoch": 72.85,
      "learning_rate": 1.0286601941747574e-05,
      "loss": 0.0856,
      "step": 187590
    },
    {
      "epoch": 72.85,
      "learning_rate": 1.0286084142394823e-05,
      "loss": 0.0533,
      "step": 187600
    },
    {
      "epoch": 72.86,
      "learning_rate": 1.0285566343042073e-05,
      "loss": 0.0249,
      "step": 187610
    },
    {
      "epoch": 72.86,
      "learning_rate": 1.0285048543689322e-05,
      "loss": 0.0519,
      "step": 187620
    },
    {
      "epoch": 72.87,
      "learning_rate": 1.0284530744336569e-05,
      "loss": 0.0756,
      "step": 187630
    },
    {
      "epoch": 72.87,
      "learning_rate": 1.0284012944983818e-05,
      "loss": 0.0193,
      "step": 187640
    },
    {
      "epoch": 72.87,
      "learning_rate": 1.0283495145631068e-05,
      "loss": 0.1042,
      "step": 187650
    },
    {
      "epoch": 72.88,
      "learning_rate": 1.0282977346278317e-05,
      "loss": 0.0434,
      "step": 187660
    },
    {
      "epoch": 72.88,
      "learning_rate": 1.0282459546925567e-05,
      "loss": 0.0923,
      "step": 187670
    },
    {
      "epoch": 72.89,
      "learning_rate": 1.0281941747572817e-05,
      "loss": 0.1582,
      "step": 187680
    },
    {
      "epoch": 72.89,
      "learning_rate": 1.0281423948220066e-05,
      "loss": 0.0108,
      "step": 187690
    },
    {
      "epoch": 72.89,
      "learning_rate": 1.0280906148867316e-05,
      "loss": 0.0255,
      "step": 187700
    },
    {
      "epoch": 72.9,
      "learning_rate": 1.0280388349514562e-05,
      "loss": 0.0204,
      "step": 187710
    },
    {
      "epoch": 72.9,
      "learning_rate": 1.0279870550161812e-05,
      "loss": 0.1345,
      "step": 187720
    },
    {
      "epoch": 72.9,
      "learning_rate": 1.0279352750809061e-05,
      "loss": 0.0376,
      "step": 187730
    },
    {
      "epoch": 72.91,
      "learning_rate": 1.0278834951456311e-05,
      "loss": 0.0522,
      "step": 187740
    },
    {
      "epoch": 72.91,
      "learning_rate": 1.027831715210356e-05,
      "loss": 0.0324,
      "step": 187750
    },
    {
      "epoch": 72.92,
      "learning_rate": 1.027779935275081e-05,
      "loss": 0.0936,
      "step": 187760
    },
    {
      "epoch": 72.92,
      "learning_rate": 1.027728155339806e-05,
      "loss": 0.2088,
      "step": 187770
    },
    {
      "epoch": 72.92,
      "learning_rate": 1.027676375404531e-05,
      "loss": 0.0494,
      "step": 187780
    },
    {
      "epoch": 72.93,
      "learning_rate": 1.0276245954692556e-05,
      "loss": 0.047,
      "step": 187790
    },
    {
      "epoch": 72.93,
      "learning_rate": 1.0275728155339805e-05,
      "loss": 0.0374,
      "step": 187800
    },
    {
      "epoch": 72.94,
      "learning_rate": 1.0275210355987055e-05,
      "loss": 0.0365,
      "step": 187810
    },
    {
      "epoch": 72.94,
      "learning_rate": 1.0274692556634305e-05,
      "loss": 0.0856,
      "step": 187820
    },
    {
      "epoch": 72.94,
      "learning_rate": 1.0274174757281554e-05,
      "loss": 0.0446,
      "step": 187830
    },
    {
      "epoch": 72.95,
      "learning_rate": 1.0273656957928804e-05,
      "loss": 0.048,
      "step": 187840
    },
    {
      "epoch": 72.95,
      "learning_rate": 1.0273139158576053e-05,
      "loss": 0.0062,
      "step": 187850
    },
    {
      "epoch": 72.96,
      "learning_rate": 1.0272621359223303e-05,
      "loss": 0.0405,
      "step": 187860
    },
    {
      "epoch": 72.96,
      "learning_rate": 1.0272103559870553e-05,
      "loss": 0.1084,
      "step": 187870
    },
    {
      "epoch": 72.96,
      "learning_rate": 1.0271585760517799e-05,
      "loss": 0.0416,
      "step": 187880
    },
    {
      "epoch": 72.97,
      "learning_rate": 1.0271067961165049e-05,
      "loss": 0.2291,
      "step": 187890
    },
    {
      "epoch": 72.97,
      "learning_rate": 1.0270550161812298e-05,
      "loss": 0.1103,
      "step": 187900
    },
    {
      "epoch": 72.97,
      "learning_rate": 1.0270032362459548e-05,
      "loss": 0.0067,
      "step": 187910
    },
    {
      "epoch": 72.98,
      "learning_rate": 1.0269514563106797e-05,
      "loss": 0.0453,
      "step": 187920
    },
    {
      "epoch": 72.98,
      "learning_rate": 1.0268996763754047e-05,
      "loss": 0.1366,
      "step": 187930
    },
    {
      "epoch": 72.99,
      "learning_rate": 1.0268478964401297e-05,
      "loss": 0.0481,
      "step": 187940
    },
    {
      "epoch": 72.99,
      "learning_rate": 1.0267961165048546e-05,
      "loss": 0.0473,
      "step": 187950
    },
    {
      "epoch": 72.99,
      "learning_rate": 1.0267443365695792e-05,
      "loss": 0.0945,
      "step": 187960
    },
    {
      "epoch": 73.0,
      "learning_rate": 1.0266925566343042e-05,
      "loss": 0.0108,
      "step": 187970
    },
    {
      "epoch": 73.0,
      "eval_accuracy": 0.9491059147180193,
      "eval_loss": 0.36288994550704956,
      "eval_runtime": 8.2044,
      "eval_samples_per_second": 443.055,
      "eval_steps_per_second": 55.458,
      "step": 187975
    },
    {
      "epoch": 73.0,
      "learning_rate": 1.0266407766990292e-05,
      "loss": 0.0536,
      "step": 187980
    },
    {
      "epoch": 73.01,
      "learning_rate": 1.0265889967637541e-05,
      "loss": 0.1152,
      "step": 187990
    },
    {
      "epoch": 73.01,
      "learning_rate": 1.0265372168284791e-05,
      "loss": 0.1347,
      "step": 188000
    },
    {
      "epoch": 73.01,
      "learning_rate": 1.026485436893204e-05,
      "loss": 0.1505,
      "step": 188010
    },
    {
      "epoch": 73.02,
      "learning_rate": 1.026433656957929e-05,
      "loss": 0.078,
      "step": 188020
    },
    {
      "epoch": 73.02,
      "learning_rate": 1.026381877022654e-05,
      "loss": 0.0423,
      "step": 188030
    },
    {
      "epoch": 73.03,
      "learning_rate": 1.0263300970873786e-05,
      "loss": 0.0243,
      "step": 188040
    },
    {
      "epoch": 73.03,
      "learning_rate": 1.0262783171521036e-05,
      "loss": 0.1491,
      "step": 188050
    },
    {
      "epoch": 73.03,
      "learning_rate": 1.0262265372168285e-05,
      "loss": 0.1088,
      "step": 188060
    },
    {
      "epoch": 73.04,
      "learning_rate": 1.0261747572815535e-05,
      "loss": 0.1759,
      "step": 188070
    },
    {
      "epoch": 73.04,
      "learning_rate": 1.0261229773462785e-05,
      "loss": 0.0036,
      "step": 188080
    },
    {
      "epoch": 73.04,
      "learning_rate": 1.0260711974110034e-05,
      "loss": 0.0294,
      "step": 188090
    },
    {
      "epoch": 73.05,
      "learning_rate": 1.0260194174757284e-05,
      "loss": 0.0936,
      "step": 188100
    },
    {
      "epoch": 73.05,
      "learning_rate": 1.0259676375404533e-05,
      "loss": 0.009,
      "step": 188110
    },
    {
      "epoch": 73.06,
      "learning_rate": 1.025915857605178e-05,
      "loss": 0.0592,
      "step": 188120
    },
    {
      "epoch": 73.06,
      "learning_rate": 1.025864077669903e-05,
      "loss": 0.066,
      "step": 188130
    },
    {
      "epoch": 73.06,
      "learning_rate": 1.0258122977346279e-05,
      "loss": 0.0091,
      "step": 188140
    },
    {
      "epoch": 73.07,
      "learning_rate": 1.0257605177993528e-05,
      "loss": 0.0588,
      "step": 188150
    },
    {
      "epoch": 73.07,
      "learning_rate": 1.0257087378640778e-05,
      "loss": 0.142,
      "step": 188160
    },
    {
      "epoch": 73.08,
      "learning_rate": 1.0256569579288028e-05,
      "loss": 0.0153,
      "step": 188170
    },
    {
      "epoch": 73.08,
      "learning_rate": 1.0256051779935277e-05,
      "loss": 0.0628,
      "step": 188180
    },
    {
      "epoch": 73.08,
      "learning_rate": 1.0255533980582525e-05,
      "loss": 0.096,
      "step": 188190
    },
    {
      "epoch": 73.09,
      "learning_rate": 1.0255016181229773e-05,
      "loss": 0.2315,
      "step": 188200
    },
    {
      "epoch": 73.09,
      "learning_rate": 1.0254498381877023e-05,
      "loss": 0.0139,
      "step": 188210
    },
    {
      "epoch": 73.1,
      "learning_rate": 1.0253980582524272e-05,
      "loss": 0.0328,
      "step": 188220
    },
    {
      "epoch": 73.1,
      "learning_rate": 1.0253462783171522e-05,
      "loss": 0.1674,
      "step": 188230
    },
    {
      "epoch": 73.1,
      "learning_rate": 1.0252944983818772e-05,
      "loss": 0.1621,
      "step": 188240
    },
    {
      "epoch": 73.11,
      "learning_rate": 1.0252427184466021e-05,
      "loss": 0.1289,
      "step": 188250
    },
    {
      "epoch": 73.11,
      "learning_rate": 1.0251909385113271e-05,
      "loss": 0.0508,
      "step": 188260
    },
    {
      "epoch": 73.11,
      "learning_rate": 1.0251391585760519e-05,
      "loss": 0.0602,
      "step": 188270
    },
    {
      "epoch": 73.12,
      "learning_rate": 1.0250873786407767e-05,
      "loss": 0.0443,
      "step": 188280
    },
    {
      "epoch": 73.12,
      "learning_rate": 1.0250355987055016e-05,
      "loss": 0.0119,
      "step": 188290
    },
    {
      "epoch": 73.13,
      "learning_rate": 1.0249838187702266e-05,
      "loss": 0.0991,
      "step": 188300
    },
    {
      "epoch": 73.13,
      "learning_rate": 1.0249320388349516e-05,
      "loss": 0.0339,
      "step": 188310
    },
    {
      "epoch": 73.13,
      "learning_rate": 1.0248802588996765e-05,
      "loss": 0.0285,
      "step": 188320
    },
    {
      "epoch": 73.14,
      "learning_rate": 1.0248284789644015e-05,
      "loss": 0.033,
      "step": 188330
    },
    {
      "epoch": 73.14,
      "learning_rate": 1.0247766990291263e-05,
      "loss": 0.1006,
      "step": 188340
    },
    {
      "epoch": 73.15,
      "learning_rate": 1.0247249190938512e-05,
      "loss": 0.0496,
      "step": 188350
    },
    {
      "epoch": 73.15,
      "learning_rate": 1.0246731391585762e-05,
      "loss": 0.0657,
      "step": 188360
    },
    {
      "epoch": 73.15,
      "learning_rate": 1.024621359223301e-05,
      "loss": 0.1309,
      "step": 188370
    },
    {
      "epoch": 73.16,
      "learning_rate": 1.024569579288026e-05,
      "loss": 0.2257,
      "step": 188380
    },
    {
      "epoch": 73.16,
      "learning_rate": 1.0245177993527509e-05,
      "loss": 0.0405,
      "step": 188390
    },
    {
      "epoch": 73.17,
      "learning_rate": 1.0244660194174759e-05,
      "loss": 0.0012,
      "step": 188400
    },
    {
      "epoch": 73.17,
      "learning_rate": 1.0244142394822008e-05,
      "loss": 0.1353,
      "step": 188410
    },
    {
      "epoch": 73.17,
      "learning_rate": 1.0243624595469256e-05,
      "loss": 0.0254,
      "step": 188420
    },
    {
      "epoch": 73.18,
      "learning_rate": 1.0243106796116506e-05,
      "loss": 0.0668,
      "step": 188430
    },
    {
      "epoch": 73.18,
      "learning_rate": 1.0242588996763756e-05,
      "loss": 0.0131,
      "step": 188440
    },
    {
      "epoch": 73.18,
      "learning_rate": 1.0242071197411003e-05,
      "loss": 0.1939,
      "step": 188450
    },
    {
      "epoch": 73.19,
      "learning_rate": 1.0241553398058253e-05,
      "loss": 0.1155,
      "step": 188460
    },
    {
      "epoch": 73.19,
      "learning_rate": 1.0241035598705503e-05,
      "loss": 0.0192,
      "step": 188470
    },
    {
      "epoch": 73.2,
      "learning_rate": 1.0240517799352752e-05,
      "loss": 0.0315,
      "step": 188480
    },
    {
      "epoch": 73.2,
      "learning_rate": 1.024e-05,
      "loss": 0.1691,
      "step": 188490
    },
    {
      "epoch": 73.2,
      "learning_rate": 1.023948220064725e-05,
      "loss": 0.0169,
      "step": 188500
    },
    {
      "epoch": 73.21,
      "learning_rate": 1.02389644012945e-05,
      "loss": 0.1071,
      "step": 188510
    },
    {
      "epoch": 73.21,
      "learning_rate": 1.0238446601941749e-05,
      "loss": 0.0521,
      "step": 188520
    },
    {
      "epoch": 73.22,
      "learning_rate": 1.0237928802588997e-05,
      "loss": 0.1492,
      "step": 188530
    },
    {
      "epoch": 73.22,
      "learning_rate": 1.0237411003236247e-05,
      "loss": 0.0101,
      "step": 188540
    },
    {
      "epoch": 73.22,
      "learning_rate": 1.0236893203883496e-05,
      "loss": 0.0385,
      "step": 188550
    },
    {
      "epoch": 73.23,
      "learning_rate": 1.0236375404530746e-05,
      "loss": 0.0073,
      "step": 188560
    },
    {
      "epoch": 73.23,
      "learning_rate": 1.0235857605177994e-05,
      "loss": 0.0256,
      "step": 188570
    },
    {
      "epoch": 73.23,
      "learning_rate": 1.0235339805825243e-05,
      "loss": 0.0137,
      "step": 188580
    },
    {
      "epoch": 73.24,
      "learning_rate": 1.0234822006472493e-05,
      "loss": 0.0545,
      "step": 188590
    },
    {
      "epoch": 73.24,
      "learning_rate": 1.0234304207119743e-05,
      "loss": 0.11,
      "step": 188600
    },
    {
      "epoch": 73.25,
      "learning_rate": 1.023378640776699e-05,
      "loss": 0.0748,
      "step": 188610
    },
    {
      "epoch": 73.25,
      "learning_rate": 1.023326860841424e-05,
      "loss": 0.0369,
      "step": 188620
    },
    {
      "epoch": 73.25,
      "learning_rate": 1.023275080906149e-05,
      "loss": 0.1298,
      "step": 188630
    },
    {
      "epoch": 73.26,
      "learning_rate": 1.023223300970874e-05,
      "loss": 0.0671,
      "step": 188640
    },
    {
      "epoch": 73.26,
      "learning_rate": 1.0231715210355987e-05,
      "loss": 0.1757,
      "step": 188650
    },
    {
      "epoch": 73.27,
      "learning_rate": 1.0231197411003237e-05,
      "loss": 0.0816,
      "step": 188660
    },
    {
      "epoch": 73.27,
      "learning_rate": 1.0230679611650487e-05,
      "loss": 0.1489,
      "step": 188670
    },
    {
      "epoch": 73.27,
      "learning_rate": 1.0230161812297736e-05,
      "loss": 0.1148,
      "step": 188680
    },
    {
      "epoch": 73.28,
      "learning_rate": 1.0229644012944984e-05,
      "loss": 0.1654,
      "step": 188690
    },
    {
      "epoch": 73.28,
      "learning_rate": 1.0229126213592234e-05,
      "loss": 0.0283,
      "step": 188700
    },
    {
      "epoch": 73.29,
      "learning_rate": 1.0228608414239483e-05,
      "loss": 0.0399,
      "step": 188710
    },
    {
      "epoch": 73.29,
      "learning_rate": 1.0228090614886731e-05,
      "loss": 0.001,
      "step": 188720
    },
    {
      "epoch": 73.29,
      "learning_rate": 1.0227572815533981e-05,
      "loss": 0.038,
      "step": 188730
    },
    {
      "epoch": 73.3,
      "learning_rate": 1.022705501618123e-05,
      "loss": 0.0202,
      "step": 188740
    },
    {
      "epoch": 73.3,
      "learning_rate": 1.022653721682848e-05,
      "loss": 0.0488,
      "step": 188750
    },
    {
      "epoch": 73.3,
      "learning_rate": 1.022601941747573e-05,
      "loss": 0.0873,
      "step": 188760
    },
    {
      "epoch": 73.31,
      "learning_rate": 1.0225501618122978e-05,
      "loss": 0.0607,
      "step": 188770
    },
    {
      "epoch": 73.31,
      "learning_rate": 1.0224983818770227e-05,
      "loss": 0.0098,
      "step": 188780
    },
    {
      "epoch": 73.32,
      "learning_rate": 1.0224466019417477e-05,
      "loss": 0.0308,
      "step": 188790
    },
    {
      "epoch": 73.32,
      "learning_rate": 1.0223948220064725e-05,
      "loss": 0.0015,
      "step": 188800
    },
    {
      "epoch": 73.32,
      "learning_rate": 1.0223430420711974e-05,
      "loss": 0.1135,
      "step": 188810
    },
    {
      "epoch": 73.33,
      "learning_rate": 1.0222912621359224e-05,
      "loss": 0.0787,
      "step": 188820
    },
    {
      "epoch": 73.33,
      "learning_rate": 1.0222394822006474e-05,
      "loss": 0.0562,
      "step": 188830
    },
    {
      "epoch": 73.34,
      "learning_rate": 1.0221877022653723e-05,
      "loss": 0.0948,
      "step": 188840
    },
    {
      "epoch": 73.34,
      "learning_rate": 1.0221359223300971e-05,
      "loss": 0.0903,
      "step": 188850
    },
    {
      "epoch": 73.34,
      "learning_rate": 1.022084142394822e-05,
      "loss": 0.0433,
      "step": 188860
    },
    {
      "epoch": 73.35,
      "learning_rate": 1.0220323624595469e-05,
      "loss": 0.0489,
      "step": 188870
    },
    {
      "epoch": 73.35,
      "learning_rate": 1.0219805825242718e-05,
      "loss": 0.0269,
      "step": 188880
    },
    {
      "epoch": 73.36,
      "learning_rate": 1.0219288025889968e-05,
      "loss": 0.0989,
      "step": 188890
    },
    {
      "epoch": 73.36,
      "learning_rate": 1.0218770226537218e-05,
      "loss": 0.0139,
      "step": 188900
    },
    {
      "epoch": 73.36,
      "learning_rate": 1.0218252427184467e-05,
      "loss": 0.0354,
      "step": 188910
    },
    {
      "epoch": 73.37,
      "learning_rate": 1.0217734627831717e-05,
      "loss": 0.0184,
      "step": 188920
    },
    {
      "epoch": 73.37,
      "learning_rate": 1.0217216828478966e-05,
      "loss": 0.0061,
      "step": 188930
    },
    {
      "epoch": 73.37,
      "learning_rate": 1.0216699029126214e-05,
      "loss": 0.0899,
      "step": 188940
    },
    {
      "epoch": 73.38,
      "learning_rate": 1.0216181229773462e-05,
      "loss": 0.0587,
      "step": 188950
    },
    {
      "epoch": 73.38,
      "learning_rate": 1.0215663430420712e-05,
      "loss": 0.0314,
      "step": 188960
    },
    {
      "epoch": 73.39,
      "learning_rate": 1.0215145631067962e-05,
      "loss": 0.0899,
      "step": 188970
    },
    {
      "epoch": 73.39,
      "learning_rate": 1.0214627831715211e-05,
      "loss": 0.2649,
      "step": 188980
    },
    {
      "epoch": 73.39,
      "learning_rate": 1.021411003236246e-05,
      "loss": 0.0978,
      "step": 188990
    },
    {
      "epoch": 73.4,
      "learning_rate": 1.021359223300971e-05,
      "loss": 0.0847,
      "step": 189000
    },
    {
      "epoch": 73.4,
      "learning_rate": 1.021307443365696e-05,
      "loss": 0.0598,
      "step": 189010
    },
    {
      "epoch": 73.41,
      "learning_rate": 1.0212556634304206e-05,
      "loss": 0.1057,
      "step": 189020
    },
    {
      "epoch": 73.41,
      "learning_rate": 1.0212038834951456e-05,
      "loss": 0.0484,
      "step": 189030
    },
    {
      "epoch": 73.41,
      "learning_rate": 1.0211521035598705e-05,
      "loss": 0.2021,
      "step": 189040
    },
    {
      "epoch": 73.42,
      "learning_rate": 1.0211003236245955e-05,
      "loss": 0.0065,
      "step": 189050
    },
    {
      "epoch": 73.42,
      "learning_rate": 1.0210485436893205e-05,
      "loss": 0.0083,
      "step": 189060
    },
    {
      "epoch": 73.43,
      "learning_rate": 1.0209967637540454e-05,
      "loss": 0.1749,
      "step": 189070
    },
    {
      "epoch": 73.43,
      "learning_rate": 1.0209449838187704e-05,
      "loss": 0.0435,
      "step": 189080
    },
    {
      "epoch": 73.43,
      "learning_rate": 1.0208932038834954e-05,
      "loss": 0.1296,
      "step": 189090
    },
    {
      "epoch": 73.44,
      "learning_rate": 1.02084142394822e-05,
      "loss": 0.0235,
      "step": 189100
    },
    {
      "epoch": 73.44,
      "learning_rate": 1.020789644012945e-05,
      "loss": 0.0297,
      "step": 189110
    },
    {
      "epoch": 73.44,
      "learning_rate": 1.0207378640776699e-05,
      "loss": 0.0133,
      "step": 189120
    },
    {
      "epoch": 73.45,
      "learning_rate": 1.0206860841423949e-05,
      "loss": 0.0768,
      "step": 189130
    },
    {
      "epoch": 73.45,
      "learning_rate": 1.0206343042071198e-05,
      "loss": 0.0839,
      "step": 189140
    },
    {
      "epoch": 73.46,
      "learning_rate": 1.0205825242718448e-05,
      "loss": 0.053,
      "step": 189150
    },
    {
      "epoch": 73.46,
      "learning_rate": 1.0205307443365698e-05,
      "loss": 0.0297,
      "step": 189160
    },
    {
      "epoch": 73.46,
      "learning_rate": 1.0204789644012947e-05,
      "loss": 0.0222,
      "step": 189170
    },
    {
      "epoch": 73.47,
      "learning_rate": 1.0204271844660193e-05,
      "loss": 0.0625,
      "step": 189180
    },
    {
      "epoch": 73.47,
      "learning_rate": 1.0203754045307443e-05,
      "loss": 0.0252,
      "step": 189190
    },
    {
      "epoch": 73.48,
      "learning_rate": 1.0203236245954693e-05,
      "loss": 0.0584,
      "step": 189200
    },
    {
      "epoch": 73.48,
      "learning_rate": 1.0202718446601942e-05,
      "loss": 0.2124,
      "step": 189210
    },
    {
      "epoch": 73.48,
      "learning_rate": 1.0202200647249192e-05,
      "loss": 0.0691,
      "step": 189220
    },
    {
      "epoch": 73.49,
      "learning_rate": 1.0201682847896441e-05,
      "loss": 0.0988,
      "step": 189230
    },
    {
      "epoch": 73.49,
      "learning_rate": 1.0201165048543691e-05,
      "loss": 0.0849,
      "step": 189240
    },
    {
      "epoch": 73.5,
      "learning_rate": 1.020064724919094e-05,
      "loss": 0.0811,
      "step": 189250
    },
    {
      "epoch": 73.5,
      "learning_rate": 1.0200129449838187e-05,
      "loss": 0.0638,
      "step": 189260
    },
    {
      "epoch": 73.5,
      "learning_rate": 1.0199611650485437e-05,
      "loss": 0.1045,
      "step": 189270
    },
    {
      "epoch": 73.51,
      "learning_rate": 1.0199093851132686e-05,
      "loss": 0.1435,
      "step": 189280
    },
    {
      "epoch": 73.51,
      "learning_rate": 1.0198576051779936e-05,
      "loss": 0.1562,
      "step": 189290
    },
    {
      "epoch": 73.51,
      "learning_rate": 1.0198058252427185e-05,
      "loss": 0.0218,
      "step": 189300
    },
    {
      "epoch": 73.52,
      "learning_rate": 1.0197540453074435e-05,
      "loss": 0.0666,
      "step": 189310
    },
    {
      "epoch": 73.52,
      "learning_rate": 1.0197022653721685e-05,
      "loss": 0.1,
      "step": 189320
    },
    {
      "epoch": 73.53,
      "learning_rate": 1.0196504854368934e-05,
      "loss": 0.1259,
      "step": 189330
    },
    {
      "epoch": 73.53,
      "learning_rate": 1.019598705501618e-05,
      "loss": 0.0013,
      "step": 189340
    },
    {
      "epoch": 73.53,
      "learning_rate": 1.019546925566343e-05,
      "loss": 0.0151,
      "step": 189350
    },
    {
      "epoch": 73.54,
      "learning_rate": 1.019495145631068e-05,
      "loss": 0.0712,
      "step": 189360
    },
    {
      "epoch": 73.54,
      "learning_rate": 1.019443365695793e-05,
      "loss": 0.0476,
      "step": 189370
    },
    {
      "epoch": 73.55,
      "learning_rate": 1.0193915857605179e-05,
      "loss": 0.1017,
      "step": 189380
    },
    {
      "epoch": 73.55,
      "learning_rate": 1.0193398058252429e-05,
      "loss": 0.045,
      "step": 189390
    },
    {
      "epoch": 73.55,
      "learning_rate": 1.0192880258899678e-05,
      "loss": 0.0204,
      "step": 189400
    },
    {
      "epoch": 73.56,
      "learning_rate": 1.0192362459546928e-05,
      "loss": 0.0201,
      "step": 189410
    },
    {
      "epoch": 73.56,
      "learning_rate": 1.0191844660194174e-05,
      "loss": 0.0592,
      "step": 189420
    },
    {
      "epoch": 73.57,
      "learning_rate": 1.0191326860841424e-05,
      "loss": 0.0077,
      "step": 189430
    },
    {
      "epoch": 73.57,
      "learning_rate": 1.0190809061488673e-05,
      "loss": 0.0233,
      "step": 189440
    },
    {
      "epoch": 73.57,
      "learning_rate": 1.0190291262135923e-05,
      "loss": 0.1624,
      "step": 189450
    },
    {
      "epoch": 73.58,
      "learning_rate": 1.0189773462783173e-05,
      "loss": 0.0003,
      "step": 189460
    },
    {
      "epoch": 73.58,
      "learning_rate": 1.0189255663430422e-05,
      "loss": 0.0088,
      "step": 189470
    },
    {
      "epoch": 73.58,
      "learning_rate": 1.0188737864077672e-05,
      "loss": 0.0235,
      "step": 189480
    },
    {
      "epoch": 73.59,
      "learning_rate": 1.0188220064724921e-05,
      "loss": 0.1561,
      "step": 189490
    },
    {
      "epoch": 73.59,
      "learning_rate": 1.0187702265372171e-05,
      "loss": 0.0148,
      "step": 189500
    },
    {
      "epoch": 73.6,
      "learning_rate": 1.0187184466019417e-05,
      "loss": 0.0328,
      "step": 189510
    },
    {
      "epoch": 73.6,
      "learning_rate": 1.0186666666666667e-05,
      "loss": 0.0043,
      "step": 189520
    },
    {
      "epoch": 73.6,
      "learning_rate": 1.0186148867313916e-05,
      "loss": 0.0403,
      "step": 189530
    },
    {
      "epoch": 73.61,
      "learning_rate": 1.0185631067961166e-05,
      "loss": 0.1375,
      "step": 189540
    },
    {
      "epoch": 73.61,
      "learning_rate": 1.0185113268608416e-05,
      "loss": 0.0293,
      "step": 189550
    },
    {
      "epoch": 73.62,
      "learning_rate": 1.0184595469255665e-05,
      "loss": 0.038,
      "step": 189560
    },
    {
      "epoch": 73.62,
      "learning_rate": 1.0184077669902915e-05,
      "loss": 0.0833,
      "step": 189570
    },
    {
      "epoch": 73.62,
      "learning_rate": 1.0183559870550165e-05,
      "loss": 0.0018,
      "step": 189580
    },
    {
      "epoch": 73.63,
      "learning_rate": 1.018304207119741e-05,
      "loss": 0.1619,
      "step": 189590
    },
    {
      "epoch": 73.63,
      "learning_rate": 1.018252427184466e-05,
      "loss": 0.0054,
      "step": 189600
    },
    {
      "epoch": 73.63,
      "learning_rate": 1.018200647249191e-05,
      "loss": 0.1262,
      "step": 189610
    },
    {
      "epoch": 73.64,
      "learning_rate": 1.018148867313916e-05,
      "loss": 0.2621,
      "step": 189620
    },
    {
      "epoch": 73.64,
      "learning_rate": 1.018097087378641e-05,
      "loss": 0.0981,
      "step": 189630
    },
    {
      "epoch": 73.65,
      "learning_rate": 1.0180453074433659e-05,
      "loss": 0.0227,
      "step": 189640
    },
    {
      "epoch": 73.65,
      "learning_rate": 1.0179935275080908e-05,
      "loss": 0.0764,
      "step": 189650
    },
    {
      "epoch": 73.65,
      "learning_rate": 1.0179417475728156e-05,
      "loss": 0.0143,
      "step": 189660
    },
    {
      "epoch": 73.66,
      "learning_rate": 1.0178899676375404e-05,
      "loss": 0.2687,
      "step": 189670
    },
    {
      "epoch": 73.66,
      "learning_rate": 1.0178381877022654e-05,
      "loss": 0.0533,
      "step": 189680
    },
    {
      "epoch": 73.67,
      "learning_rate": 1.0177864077669904e-05,
      "loss": 0.1215,
      "step": 189690
    },
    {
      "epoch": 73.67,
      "learning_rate": 1.0177346278317153e-05,
      "loss": 0.0849,
      "step": 189700
    },
    {
      "epoch": 73.67,
      "learning_rate": 1.0176828478964403e-05,
      "loss": 0.0544,
      "step": 189710
    },
    {
      "epoch": 73.68,
      "learning_rate": 1.0176310679611652e-05,
      "loss": 0.0253,
      "step": 189720
    },
    {
      "epoch": 73.68,
      "learning_rate": 1.0175792880258902e-05,
      "loss": 0.0282,
      "step": 189730
    },
    {
      "epoch": 73.69,
      "learning_rate": 1.017527508090615e-05,
      "loss": 0.0308,
      "step": 189740
    },
    {
      "epoch": 73.69,
      "learning_rate": 1.0174757281553398e-05,
      "loss": 0.1439,
      "step": 189750
    },
    {
      "epoch": 73.69,
      "learning_rate": 1.0174239482200647e-05,
      "loss": 0.1147,
      "step": 189760
    },
    {
      "epoch": 73.7,
      "learning_rate": 1.0173721682847897e-05,
      "loss": 0.0121,
      "step": 189770
    },
    {
      "epoch": 73.7,
      "learning_rate": 1.0173203883495147e-05,
      "loss": 0.0388,
      "step": 189780
    },
    {
      "epoch": 73.7,
      "learning_rate": 1.0172686084142396e-05,
      "loss": 0.1018,
      "step": 189790
    },
    {
      "epoch": 73.71,
      "learning_rate": 1.0172168284789646e-05,
      "loss": 0.0011,
      "step": 189800
    },
    {
      "epoch": 73.71,
      "learning_rate": 1.0171650485436894e-05,
      "loss": 0.0573,
      "step": 189810
    },
    {
      "epoch": 73.72,
      "learning_rate": 1.0171132686084144e-05,
      "loss": 0.0317,
      "step": 189820
    },
    {
      "epoch": 73.72,
      "learning_rate": 1.0170614886731391e-05,
      "loss": 0.184,
      "step": 189830
    },
    {
      "epoch": 73.72,
      "learning_rate": 1.0170097087378641e-05,
      "loss": 0.0597,
      "step": 189840
    },
    {
      "epoch": 73.73,
      "learning_rate": 1.016957928802589e-05,
      "loss": 0.0122,
      "step": 189850
    },
    {
      "epoch": 73.73,
      "learning_rate": 1.016906148867314e-05,
      "loss": 0.0445,
      "step": 189860
    },
    {
      "epoch": 73.74,
      "learning_rate": 1.016854368932039e-05,
      "loss": 0.0008,
      "step": 189870
    },
    {
      "epoch": 73.74,
      "learning_rate": 1.016802588996764e-05,
      "loss": 0.0436,
      "step": 189880
    },
    {
      "epoch": 73.74,
      "learning_rate": 1.0167508090614887e-05,
      "loss": 0.089,
      "step": 189890
    },
    {
      "epoch": 73.75,
      "learning_rate": 1.0166990291262137e-05,
      "loss": 0.0002,
      "step": 189900
    },
    {
      "epoch": 73.75,
      "learning_rate": 1.0166472491909385e-05,
      "loss": 0.1984,
      "step": 189910
    },
    {
      "epoch": 73.76,
      "learning_rate": 1.0165954692556635e-05,
      "loss": 0.1084,
      "step": 189920
    },
    {
      "epoch": 73.76,
      "learning_rate": 1.0165436893203884e-05,
      "loss": 0.0902,
      "step": 189930
    },
    {
      "epoch": 73.76,
      "learning_rate": 1.0164919093851134e-05,
      "loss": 0.0847,
      "step": 189940
    },
    {
      "epoch": 73.77,
      "learning_rate": 1.0164401294498383e-05,
      "loss": 0.0269,
      "step": 189950
    },
    {
      "epoch": 73.77,
      "learning_rate": 1.0163883495145631e-05,
      "loss": 0.0448,
      "step": 189960
    },
    {
      "epoch": 73.77,
      "learning_rate": 1.0163365695792881e-05,
      "loss": 0.0358,
      "step": 189970
    },
    {
      "epoch": 73.78,
      "learning_rate": 1.016284789644013e-05,
      "loss": 0.0623,
      "step": 189980
    },
    {
      "epoch": 73.78,
      "learning_rate": 1.0162330097087379e-05,
      "loss": 0.1839,
      "step": 189990
    },
    {
      "epoch": 73.79,
      "learning_rate": 1.0161812297734628e-05,
      "loss": 0.0624,
      "step": 190000
    },
    {
      "epoch": 73.79,
      "learning_rate": 1.0161294498381878e-05,
      "loss": 0.0843,
      "step": 190010
    },
    {
      "epoch": 73.79,
      "learning_rate": 1.0160776699029127e-05,
      "loss": 0.0168,
      "step": 190020
    },
    {
      "epoch": 73.8,
      "learning_rate": 1.0160258899676377e-05,
      "loss": 0.0706,
      "step": 190030
    },
    {
      "epoch": 73.8,
      "learning_rate": 1.0159741100323625e-05,
      "loss": 0.0418,
      "step": 190040
    },
    {
      "epoch": 73.81,
      "learning_rate": 1.0159223300970875e-05,
      "loss": 0.0647,
      "step": 190050
    },
    {
      "epoch": 73.81,
      "learning_rate": 1.0158705501618124e-05,
      "loss": 0.0327,
      "step": 190060
    },
    {
      "epoch": 73.81,
      "learning_rate": 1.0158187702265374e-05,
      "loss": 0.1731,
      "step": 190070
    },
    {
      "epoch": 73.82,
      "learning_rate": 1.0157669902912622e-05,
      "loss": 0.011,
      "step": 190080
    },
    {
      "epoch": 73.82,
      "learning_rate": 1.0157152103559871e-05,
      "loss": 0.1228,
      "step": 190090
    },
    {
      "epoch": 73.83,
      "learning_rate": 1.0156634304207121e-05,
      "loss": 0.0369,
      "step": 190100
    },
    {
      "epoch": 73.83,
      "learning_rate": 1.015611650485437e-05,
      "loss": 0.0699,
      "step": 190110
    },
    {
      "epoch": 73.83,
      "learning_rate": 1.0155598705501618e-05,
      "loss": 0.0437,
      "step": 190120
    },
    {
      "epoch": 73.84,
      "learning_rate": 1.0155080906148868e-05,
      "loss": 0.0428,
      "step": 190130
    },
    {
      "epoch": 73.84,
      "learning_rate": 1.0154563106796118e-05,
      "loss": 0.0012,
      "step": 190140
    },
    {
      "epoch": 73.84,
      "learning_rate": 1.0154045307443367e-05,
      "loss": 0.1259,
      "step": 190150
    },
    {
      "epoch": 73.85,
      "learning_rate": 1.0153527508090615e-05,
      "loss": 0.1472,
      "step": 190160
    },
    {
      "epoch": 73.85,
      "learning_rate": 1.0153009708737865e-05,
      "loss": 0.0323,
      "step": 190170
    },
    {
      "epoch": 73.86,
      "learning_rate": 1.0152491909385115e-05,
      "loss": 0.0016,
      "step": 190180
    },
    {
      "epoch": 73.86,
      "learning_rate": 1.0151974110032362e-05,
      "loss": 0.0308,
      "step": 190190
    },
    {
      "epoch": 73.86,
      "learning_rate": 1.0151456310679612e-05,
      "loss": 0.1118,
      "step": 190200
    },
    {
      "epoch": 73.87,
      "learning_rate": 1.0150938511326862e-05,
      "loss": 0.1181,
      "step": 190210
    },
    {
      "epoch": 73.87,
      "learning_rate": 1.0150420711974111e-05,
      "loss": 0.1537,
      "step": 190220
    },
    {
      "epoch": 73.88,
      "learning_rate": 1.0149902912621361e-05,
      "loss": 0.0007,
      "step": 190230
    },
    {
      "epoch": 73.88,
      "learning_rate": 1.0149385113268609e-05,
      "loss": 0.1252,
      "step": 190240
    },
    {
      "epoch": 73.88,
      "learning_rate": 1.0148867313915858e-05,
      "loss": 0.0044,
      "step": 190250
    },
    {
      "epoch": 73.89,
      "learning_rate": 1.0148349514563108e-05,
      "loss": 0.084,
      "step": 190260
    },
    {
      "epoch": 73.89,
      "learning_rate": 1.0147831715210356e-05,
      "loss": 0.0036,
      "step": 190270
    },
    {
      "epoch": 73.9,
      "learning_rate": 1.0147313915857606e-05,
      "loss": 0.0516,
      "step": 190280
    },
    {
      "epoch": 73.9,
      "learning_rate": 1.0146796116504855e-05,
      "loss": 0.0052,
      "step": 190290
    },
    {
      "epoch": 73.9,
      "learning_rate": 1.0146278317152105e-05,
      "loss": 0.137,
      "step": 190300
    },
    {
      "epoch": 73.91,
      "learning_rate": 1.0145760517799354e-05,
      "loss": 0.0509,
      "step": 190310
    },
    {
      "epoch": 73.91,
      "learning_rate": 1.0145242718446602e-05,
      "loss": 0.0288,
      "step": 190320
    },
    {
      "epoch": 73.91,
      "learning_rate": 1.0144724919093852e-05,
      "loss": 0.0786,
      "step": 190330
    },
    {
      "epoch": 73.92,
      "learning_rate": 1.01442071197411e-05,
      "loss": 0.035,
      "step": 190340
    },
    {
      "epoch": 73.92,
      "learning_rate": 1.014368932038835e-05,
      "loss": 0.0098,
      "step": 190350
    },
    {
      "epoch": 73.93,
      "learning_rate": 1.01431715210356e-05,
      "loss": 0.0757,
      "step": 190360
    },
    {
      "epoch": 73.93,
      "learning_rate": 1.0142653721682849e-05,
      "loss": 0.0334,
      "step": 190370
    },
    {
      "epoch": 73.93,
      "learning_rate": 1.0142135922330098e-05,
      "loss": 0.1023,
      "step": 190380
    },
    {
      "epoch": 73.94,
      "learning_rate": 1.0141618122977348e-05,
      "loss": 0.0236,
      "step": 190390
    },
    {
      "epoch": 73.94,
      "learning_rate": 1.0141100323624596e-05,
      "loss": 0.0585,
      "step": 190400
    },
    {
      "epoch": 73.95,
      "learning_rate": 1.0140582524271846e-05,
      "loss": 0.0615,
      "step": 190410
    },
    {
      "epoch": 73.95,
      "learning_rate": 1.0140064724919093e-05,
      "loss": 0.0525,
      "step": 190420
    },
    {
      "epoch": 73.95,
      "learning_rate": 1.0139546925566343e-05,
      "loss": 0.0601,
      "step": 190430
    },
    {
      "epoch": 73.96,
      "learning_rate": 1.0139029126213593e-05,
      "loss": 0.1023,
      "step": 190440
    },
    {
      "epoch": 73.96,
      "learning_rate": 1.0138511326860842e-05,
      "loss": 0.1699,
      "step": 190450
    },
    {
      "epoch": 73.97,
      "learning_rate": 1.0137993527508092e-05,
      "loss": 0.1158,
      "step": 190460
    },
    {
      "epoch": 73.97,
      "learning_rate": 1.0137475728155342e-05,
      "loss": 0.055,
      "step": 190470
    },
    {
      "epoch": 73.97,
      "learning_rate": 1.013695792880259e-05,
      "loss": 0.0509,
      "step": 190480
    },
    {
      "epoch": 73.98,
      "learning_rate": 1.0136440129449837e-05,
      "loss": 0.07,
      "step": 190490
    },
    {
      "epoch": 73.98,
      "learning_rate": 1.0135922330097087e-05,
      "loss": 0.0216,
      "step": 190500
    },
    {
      "epoch": 73.98,
      "learning_rate": 1.0135404530744337e-05,
      "loss": 0.0681,
      "step": 190510
    },
    {
      "epoch": 73.99,
      "learning_rate": 1.0134886731391586e-05,
      "loss": 0.0175,
      "step": 190520
    },
    {
      "epoch": 73.99,
      "learning_rate": 1.0134368932038836e-05,
      "loss": 0.0829,
      "step": 190530
    },
    {
      "epoch": 74.0,
      "learning_rate": 1.0133851132686086e-05,
      "loss": 0.1317,
      "step": 190540
    },
    {
      "epoch": 74.0,
      "learning_rate": 1.0133333333333335e-05,
      "loss": 0.0446,
      "step": 190550
    },
    {
      "epoch": 74.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.35697141289711,
      "eval_runtime": 8.2637,
      "eval_samples_per_second": 439.875,
      "eval_steps_per_second": 55.06,
      "step": 190550
    },
    {
      "epoch": 74.0,
      "learning_rate": 1.0132815533980583e-05,
      "loss": 0.0938,
      "step": 190560
    },
    {
      "epoch": 74.01,
      "learning_rate": 1.0132297734627831e-05,
      "loss": 0.084,
      "step": 190570
    },
    {
      "epoch": 74.01,
      "learning_rate": 1.013177993527508e-05,
      "loss": 0.0003,
      "step": 190580
    },
    {
      "epoch": 74.02,
      "learning_rate": 1.013126213592233e-05,
      "loss": 0.0807,
      "step": 190590
    },
    {
      "epoch": 74.02,
      "learning_rate": 1.013074433656958e-05,
      "loss": 0.066,
      "step": 190600
    },
    {
      "epoch": 74.02,
      "learning_rate": 1.013022653721683e-05,
      "loss": 0.0293,
      "step": 190610
    },
    {
      "epoch": 74.03,
      "learning_rate": 1.0129708737864079e-05,
      "loss": 0.0082,
      "step": 190620
    },
    {
      "epoch": 74.03,
      "learning_rate": 1.0129190938511329e-05,
      "loss": 0.0405,
      "step": 190630
    },
    {
      "epoch": 74.03,
      "learning_rate": 1.0128673139158578e-05,
      "loss": 0.0277,
      "step": 190640
    },
    {
      "epoch": 74.04,
      "learning_rate": 1.0128155339805825e-05,
      "loss": 0.0017,
      "step": 190650
    },
    {
      "epoch": 74.04,
      "learning_rate": 1.0127637540453074e-05,
      "loss": 0.072,
      "step": 190660
    },
    {
      "epoch": 74.05,
      "learning_rate": 1.0127119741100324e-05,
      "loss": 0.1266,
      "step": 190670
    },
    {
      "epoch": 74.05,
      "learning_rate": 1.0126601941747573e-05,
      "loss": 0.1322,
      "step": 190680
    },
    {
      "epoch": 74.05,
      "learning_rate": 1.0126084142394823e-05,
      "loss": 0.1008,
      "step": 190690
    },
    {
      "epoch": 74.06,
      "learning_rate": 1.0125566343042073e-05,
      "loss": 0.0665,
      "step": 190700
    },
    {
      "epoch": 74.06,
      "learning_rate": 1.0125048543689322e-05,
      "loss": 0.0088,
      "step": 190710
    },
    {
      "epoch": 74.07,
      "learning_rate": 1.0124530744336572e-05,
      "loss": 0.0016,
      "step": 190720
    },
    {
      "epoch": 74.07,
      "learning_rate": 1.0124012944983818e-05,
      "loss": 0.3053,
      "step": 190730
    },
    {
      "epoch": 74.07,
      "learning_rate": 1.0123495145631068e-05,
      "loss": 0.0174,
      "step": 190740
    },
    {
      "epoch": 74.08,
      "learning_rate": 1.0122977346278317e-05,
      "loss": 0.0005,
      "step": 190750
    },
    {
      "epoch": 74.08,
      "learning_rate": 1.0122459546925567e-05,
      "loss": 0.0028,
      "step": 190760
    },
    {
      "epoch": 74.09,
      "learning_rate": 1.0121941747572817e-05,
      "loss": 0.0902,
      "step": 190770
    },
    {
      "epoch": 74.09,
      "learning_rate": 1.0121423948220066e-05,
      "loss": 0.1061,
      "step": 190780
    },
    {
      "epoch": 74.09,
      "learning_rate": 1.0120906148867316e-05,
      "loss": 0.0038,
      "step": 190790
    },
    {
      "epoch": 74.1,
      "learning_rate": 1.0120388349514565e-05,
      "loss": 0.0173,
      "step": 190800
    },
    {
      "epoch": 74.1,
      "learning_rate": 1.0119870550161812e-05,
      "loss": 0.0362,
      "step": 190810
    },
    {
      "epoch": 74.1,
      "learning_rate": 1.0119352750809061e-05,
      "loss": 0.0003,
      "step": 190820
    },
    {
      "epoch": 74.11,
      "learning_rate": 1.0118834951456311e-05,
      "loss": 0.197,
      "step": 190830
    },
    {
      "epoch": 74.11,
      "learning_rate": 1.011831715210356e-05,
      "loss": 0.0197,
      "step": 190840
    },
    {
      "epoch": 74.12,
      "learning_rate": 1.011779935275081e-05,
      "loss": 0.084,
      "step": 190850
    },
    {
      "epoch": 74.12,
      "learning_rate": 1.011728155339806e-05,
      "loss": 0.0616,
      "step": 190860
    },
    {
      "epoch": 74.12,
      "learning_rate": 1.011676375404531e-05,
      "loss": 0.0763,
      "step": 190870
    },
    {
      "epoch": 74.13,
      "learning_rate": 1.0116245954692559e-05,
      "loss": 0.1039,
      "step": 190880
    },
    {
      "epoch": 74.13,
      "learning_rate": 1.0115728155339805e-05,
      "loss": 0.1235,
      "step": 190890
    },
    {
      "epoch": 74.14,
      "learning_rate": 1.0115210355987055e-05,
      "loss": 0.0714,
      "step": 190900
    },
    {
      "epoch": 74.14,
      "learning_rate": 1.0114692556634304e-05,
      "loss": 0.0981,
      "step": 190910
    },
    {
      "epoch": 74.14,
      "learning_rate": 1.0114174757281554e-05,
      "loss": 0.1792,
      "step": 190920
    },
    {
      "epoch": 74.15,
      "learning_rate": 1.0113656957928804e-05,
      "loss": 0.1034,
      "step": 190930
    },
    {
      "epoch": 74.15,
      "learning_rate": 1.0113139158576053e-05,
      "loss": 0.0027,
      "step": 190940
    },
    {
      "epoch": 74.16,
      "learning_rate": 1.0112621359223303e-05,
      "loss": 0.0288,
      "step": 190950
    },
    {
      "epoch": 74.16,
      "learning_rate": 1.0112103559870553e-05,
      "loss": 0.0835,
      "step": 190960
    },
    {
      "epoch": 74.16,
      "learning_rate": 1.0111585760517799e-05,
      "loss": 0.0434,
      "step": 190970
    },
    {
      "epoch": 74.17,
      "learning_rate": 1.0111067961165048e-05,
      "loss": 0.1468,
      "step": 190980
    },
    {
      "epoch": 74.17,
      "learning_rate": 1.0110550161812298e-05,
      "loss": 0.0148,
      "step": 190990
    },
    {
      "epoch": 74.17,
      "learning_rate": 1.0110032362459548e-05,
      "loss": 0.084,
      "step": 191000
    },
    {
      "epoch": 74.18,
      "learning_rate": 1.0109514563106797e-05,
      "loss": 0.0941,
      "step": 191010
    },
    {
      "epoch": 74.18,
      "learning_rate": 1.0108996763754047e-05,
      "loss": 0.1014,
      "step": 191020
    },
    {
      "epoch": 74.19,
      "learning_rate": 1.0108478964401296e-05,
      "loss": 0.1619,
      "step": 191030
    },
    {
      "epoch": 74.19,
      "learning_rate": 1.0107961165048546e-05,
      "loss": 0.1233,
      "step": 191040
    },
    {
      "epoch": 74.19,
      "learning_rate": 1.0107443365695792e-05,
      "loss": 0.2314,
      "step": 191050
    },
    {
      "epoch": 74.2,
      "learning_rate": 1.0106925566343042e-05,
      "loss": 0.1445,
      "step": 191060
    },
    {
      "epoch": 74.2,
      "learning_rate": 1.0106407766990292e-05,
      "loss": 0.0533,
      "step": 191070
    },
    {
      "epoch": 74.21,
      "learning_rate": 1.0105889967637541e-05,
      "loss": 0.0044,
      "step": 191080
    },
    {
      "epoch": 74.21,
      "learning_rate": 1.010537216828479e-05,
      "loss": 0.0574,
      "step": 191090
    },
    {
      "epoch": 74.21,
      "learning_rate": 1.010485436893204e-05,
      "loss": 0.0498,
      "step": 191100
    },
    {
      "epoch": 74.22,
      "learning_rate": 1.010433656957929e-05,
      "loss": 0.0583,
      "step": 191110
    },
    {
      "epoch": 74.22,
      "learning_rate": 1.010381877022654e-05,
      "loss": 0.0413,
      "step": 191120
    },
    {
      "epoch": 74.23,
      "learning_rate": 1.0103300970873786e-05,
      "loss": 0.0446,
      "step": 191130
    },
    {
      "epoch": 74.23,
      "learning_rate": 1.0102783171521035e-05,
      "loss": 0.0003,
      "step": 191140
    },
    {
      "epoch": 74.23,
      "learning_rate": 1.0102265372168285e-05,
      "loss": 0.125,
      "step": 191150
    },
    {
      "epoch": 74.24,
      "learning_rate": 1.0101747572815535e-05,
      "loss": 0.1559,
      "step": 191160
    },
    {
      "epoch": 74.24,
      "learning_rate": 1.0101229773462784e-05,
      "loss": 0.1756,
      "step": 191170
    },
    {
      "epoch": 74.24,
      "learning_rate": 1.0100711974110034e-05,
      "loss": 0.0673,
      "step": 191180
    },
    {
      "epoch": 74.25,
      "learning_rate": 1.0100194174757284e-05,
      "loss": 0.0476,
      "step": 191190
    },
    {
      "epoch": 74.25,
      "learning_rate": 1.0099676375404533e-05,
      "loss": 0.1097,
      "step": 191200
    },
    {
      "epoch": 74.26,
      "learning_rate": 1.0099158576051781e-05,
      "loss": 0.0571,
      "step": 191210
    },
    {
      "epoch": 74.26,
      "learning_rate": 1.0098640776699029e-05,
      "loss": 0.0956,
      "step": 191220
    },
    {
      "epoch": 74.26,
      "learning_rate": 1.0098122977346279e-05,
      "loss": 0.0206,
      "step": 191230
    },
    {
      "epoch": 74.27,
      "learning_rate": 1.0097605177993528e-05,
      "loss": 0.0459,
      "step": 191240
    },
    {
      "epoch": 74.27,
      "learning_rate": 1.0097087378640778e-05,
      "loss": 0.0396,
      "step": 191250
    },
    {
      "epoch": 74.28,
      "learning_rate": 1.0096569579288028e-05,
      "loss": 0.0096,
      "step": 191260
    },
    {
      "epoch": 74.28,
      "learning_rate": 1.0096051779935277e-05,
      "loss": 0.0151,
      "step": 191270
    },
    {
      "epoch": 74.28,
      "learning_rate": 1.0095533980582525e-05,
      "loss": 0.135,
      "step": 191280
    },
    {
      "epoch": 74.29,
      "learning_rate": 1.0095016181229775e-05,
      "loss": 0.0193,
      "step": 191290
    },
    {
      "epoch": 74.29,
      "learning_rate": 1.0094498381877023e-05,
      "loss": 0.0202,
      "step": 191300
    },
    {
      "epoch": 74.3,
      "learning_rate": 1.0093980582524272e-05,
      "loss": 0.0003,
      "step": 191310
    },
    {
      "epoch": 74.3,
      "learning_rate": 1.0093462783171522e-05,
      "loss": 0.0669,
      "step": 191320
    },
    {
      "epoch": 74.3,
      "learning_rate": 1.0092944983818771e-05,
      "loss": 0.2528,
      "step": 191330
    },
    {
      "epoch": 74.31,
      "learning_rate": 1.0092427184466021e-05,
      "loss": 0.0043,
      "step": 191340
    },
    {
      "epoch": 74.31,
      "learning_rate": 1.009190938511327e-05,
      "loss": 0.0394,
      "step": 191350
    },
    {
      "epoch": 74.31,
      "learning_rate": 1.0091391585760519e-05,
      "loss": 0.0137,
      "step": 191360
    },
    {
      "epoch": 74.32,
      "learning_rate": 1.0090873786407768e-05,
      "loss": 0.2087,
      "step": 191370
    },
    {
      "epoch": 74.32,
      "learning_rate": 1.0090355987055016e-05,
      "loss": 0.0488,
      "step": 191380
    },
    {
      "epoch": 74.33,
      "learning_rate": 1.0089838187702266e-05,
      "loss": 0.0243,
      "step": 191390
    },
    {
      "epoch": 74.33,
      "learning_rate": 1.0089320388349515e-05,
      "loss": 0.0184,
      "step": 191400
    },
    {
      "epoch": 74.33,
      "learning_rate": 1.0088802588996765e-05,
      "loss": 0.0591,
      "step": 191410
    },
    {
      "epoch": 74.34,
      "learning_rate": 1.0088284789644015e-05,
      "loss": 0.0932,
      "step": 191420
    },
    {
      "epoch": 74.34,
      "learning_rate": 1.0087766990291263e-05,
      "loss": 0.004,
      "step": 191430
    },
    {
      "epoch": 74.35,
      "learning_rate": 1.0087249190938512e-05,
      "loss": 0.0245,
      "step": 191440
    },
    {
      "epoch": 74.35,
      "learning_rate": 1.0086731391585762e-05,
      "loss": 0.089,
      "step": 191450
    },
    {
      "epoch": 74.35,
      "learning_rate": 1.008621359223301e-05,
      "loss": 0.0212,
      "step": 191460
    },
    {
      "epoch": 74.36,
      "learning_rate": 1.008569579288026e-05,
      "loss": 0.0775,
      "step": 191470
    },
    {
      "epoch": 74.36,
      "learning_rate": 1.0085177993527509e-05,
      "loss": 0.0036,
      "step": 191480
    },
    {
      "epoch": 74.37,
      "learning_rate": 1.0084660194174759e-05,
      "loss": 0.1239,
      "step": 191490
    },
    {
      "epoch": 74.37,
      "learning_rate": 1.0084142394822008e-05,
      "loss": 0.0206,
      "step": 191500
    },
    {
      "epoch": 74.37,
      "learning_rate": 1.0083624595469256e-05,
      "loss": 0.1132,
      "step": 191510
    },
    {
      "epoch": 74.38,
      "learning_rate": 1.0083106796116506e-05,
      "loss": 0.1228,
      "step": 191520
    },
    {
      "epoch": 74.38,
      "learning_rate": 1.0082588996763755e-05,
      "loss": 0.038,
      "step": 191530
    },
    {
      "epoch": 74.38,
      "learning_rate": 1.0082071197411003e-05,
      "loss": 0.1063,
      "step": 191540
    },
    {
      "epoch": 74.39,
      "learning_rate": 1.0081553398058253e-05,
      "loss": 0.0733,
      "step": 191550
    },
    {
      "epoch": 74.39,
      "learning_rate": 1.0081035598705503e-05,
      "loss": 0.0902,
      "step": 191560
    },
    {
      "epoch": 74.4,
      "learning_rate": 1.0080517799352752e-05,
      "loss": 0.1961,
      "step": 191570
    },
    {
      "epoch": 74.4,
      "learning_rate": 1.008e-05,
      "loss": 0.0568,
      "step": 191580
    },
    {
      "epoch": 74.4,
      "learning_rate": 1.007948220064725e-05,
      "loss": 0.0267,
      "step": 191590
    },
    {
      "epoch": 74.41,
      "learning_rate": 1.00789644012945e-05,
      "loss": 0.072,
      "step": 191600
    },
    {
      "epoch": 74.41,
      "learning_rate": 1.0078446601941749e-05,
      "loss": 0.0076,
      "step": 191610
    },
    {
      "epoch": 74.42,
      "learning_rate": 1.0077928802588997e-05,
      "loss": 0.0024,
      "step": 191620
    },
    {
      "epoch": 74.42,
      "learning_rate": 1.0077411003236246e-05,
      "loss": 0.0711,
      "step": 191630
    },
    {
      "epoch": 74.42,
      "learning_rate": 1.0076893203883496e-05,
      "loss": 0.0035,
      "step": 191640
    },
    {
      "epoch": 74.43,
      "learning_rate": 1.0076375404530746e-05,
      "loss": 0.0646,
      "step": 191650
    },
    {
      "epoch": 74.43,
      "learning_rate": 1.0075857605177994e-05,
      "loss": 0.0236,
      "step": 191660
    },
    {
      "epoch": 74.43,
      "learning_rate": 1.0075339805825243e-05,
      "loss": 0.0776,
      "step": 191670
    },
    {
      "epoch": 74.44,
      "learning_rate": 1.0074822006472493e-05,
      "loss": 0.0407,
      "step": 191680
    },
    {
      "epoch": 74.44,
      "learning_rate": 1.0074304207119742e-05,
      "loss": 0.0167,
      "step": 191690
    },
    {
      "epoch": 74.45,
      "learning_rate": 1.007378640776699e-05,
      "loss": 0.1497,
      "step": 191700
    },
    {
      "epoch": 74.45,
      "learning_rate": 1.007326860841424e-05,
      "loss": 0.1452,
      "step": 191710
    },
    {
      "epoch": 74.45,
      "learning_rate": 1.007275080906149e-05,
      "loss": 0.1229,
      "step": 191720
    },
    {
      "epoch": 74.46,
      "learning_rate": 1.007223300970874e-05,
      "loss": 0.0764,
      "step": 191730
    },
    {
      "epoch": 74.46,
      "learning_rate": 1.0071715210355987e-05,
      "loss": 0.0957,
      "step": 191740
    },
    {
      "epoch": 74.47,
      "learning_rate": 1.0071197411003237e-05,
      "loss": 0.0017,
      "step": 191750
    },
    {
      "epoch": 74.47,
      "learning_rate": 1.0070679611650486e-05,
      "loss": 0.0541,
      "step": 191760
    },
    {
      "epoch": 74.47,
      "learning_rate": 1.0070161812297736e-05,
      "loss": 0.1046,
      "step": 191770
    },
    {
      "epoch": 74.48,
      "learning_rate": 1.0069644012944986e-05,
      "loss": 0.0064,
      "step": 191780
    },
    {
      "epoch": 74.48,
      "learning_rate": 1.0069126213592234e-05,
      "loss": 0.0713,
      "step": 191790
    },
    {
      "epoch": 74.49,
      "learning_rate": 1.0068608414239483e-05,
      "loss": 0.0763,
      "step": 191800
    },
    {
      "epoch": 74.49,
      "learning_rate": 1.0068090614886731e-05,
      "loss": 0.0825,
      "step": 191810
    },
    {
      "epoch": 74.49,
      "learning_rate": 1.006757281553398e-05,
      "loss": 0.0324,
      "step": 191820
    },
    {
      "epoch": 74.5,
      "learning_rate": 1.006705501618123e-05,
      "loss": 0.0007,
      "step": 191830
    },
    {
      "epoch": 74.5,
      "learning_rate": 1.006653721682848e-05,
      "loss": 0.0372,
      "step": 191840
    },
    {
      "epoch": 74.5,
      "learning_rate": 1.006601941747573e-05,
      "loss": 0.0357,
      "step": 191850
    },
    {
      "epoch": 74.51,
      "learning_rate": 1.006550161812298e-05,
      "loss": 0.0138,
      "step": 191860
    },
    {
      "epoch": 74.51,
      "learning_rate": 1.0064983818770227e-05,
      "loss": 0.0432,
      "step": 191870
    },
    {
      "epoch": 74.52,
      "learning_rate": 1.0064466019417477e-05,
      "loss": 0.0337,
      "step": 191880
    },
    {
      "epoch": 74.52,
      "learning_rate": 1.0063948220064725e-05,
      "loss": 0.0033,
      "step": 191890
    },
    {
      "epoch": 74.52,
      "learning_rate": 1.0063430420711974e-05,
      "loss": 0.1304,
      "step": 191900
    },
    {
      "epoch": 74.53,
      "learning_rate": 1.0062912621359224e-05,
      "loss": 0.0143,
      "step": 191910
    },
    {
      "epoch": 74.53,
      "learning_rate": 1.0062394822006474e-05,
      "loss": 0.0128,
      "step": 191920
    },
    {
      "epoch": 74.54,
      "learning_rate": 1.0061877022653723e-05,
      "loss": 0.0519,
      "step": 191930
    },
    {
      "epoch": 74.54,
      "learning_rate": 1.0061359223300973e-05,
      "loss": 0.0658,
      "step": 191940
    },
    {
      "epoch": 74.54,
      "learning_rate": 1.006084142394822e-05,
      "loss": 0.0775,
      "step": 191950
    },
    {
      "epoch": 74.55,
      "learning_rate": 1.0060323624595469e-05,
      "loss": 0.0057,
      "step": 191960
    },
    {
      "epoch": 74.55,
      "learning_rate": 1.0059805825242718e-05,
      "loss": 0.0482,
      "step": 191970
    },
    {
      "epoch": 74.56,
      "learning_rate": 1.0059288025889968e-05,
      "loss": 0.0225,
      "step": 191980
    },
    {
      "epoch": 74.56,
      "learning_rate": 1.0058770226537217e-05,
      "loss": 0.1523,
      "step": 191990
    },
    {
      "epoch": 74.56,
      "learning_rate": 1.0058252427184467e-05,
      "loss": 0.0159,
      "step": 192000
    },
    {
      "epoch": 74.57,
      "learning_rate": 1.0057734627831717e-05,
      "loss": 0.1058,
      "step": 192010
    },
    {
      "epoch": 74.57,
      "learning_rate": 1.0057216828478966e-05,
      "loss": 0.0588,
      "step": 192020
    },
    {
      "epoch": 74.57,
      "learning_rate": 1.0056699029126214e-05,
      "loss": 0.0363,
      "step": 192030
    },
    {
      "epoch": 74.58,
      "learning_rate": 1.0056181229773462e-05,
      "loss": 0.0124,
      "step": 192040
    },
    {
      "epoch": 74.58,
      "learning_rate": 1.0055663430420712e-05,
      "loss": 0.0079,
      "step": 192050
    },
    {
      "epoch": 74.59,
      "learning_rate": 1.0055145631067961e-05,
      "loss": 0.0935,
      "step": 192060
    },
    {
      "epoch": 74.59,
      "learning_rate": 1.0054627831715211e-05,
      "loss": 0.0016,
      "step": 192070
    },
    {
      "epoch": 74.59,
      "learning_rate": 1.005411003236246e-05,
      "loss": 0.0579,
      "step": 192080
    },
    {
      "epoch": 74.6,
      "learning_rate": 1.005359223300971e-05,
      "loss": 0.1021,
      "step": 192090
    },
    {
      "epoch": 74.6,
      "learning_rate": 1.005307443365696e-05,
      "loss": 0.0584,
      "step": 192100
    },
    {
      "epoch": 74.61,
      "learning_rate": 1.0052556634304206e-05,
      "loss": 0.1834,
      "step": 192110
    },
    {
      "epoch": 74.61,
      "learning_rate": 1.0052038834951456e-05,
      "loss": 0.1713,
      "step": 192120
    },
    {
      "epoch": 74.61,
      "learning_rate": 1.0051521035598705e-05,
      "loss": 0.0489,
      "step": 192130
    },
    {
      "epoch": 74.62,
      "learning_rate": 1.0051003236245955e-05,
      "loss": 0.0599,
      "step": 192140
    },
    {
      "epoch": 74.62,
      "learning_rate": 1.0050485436893205e-05,
      "loss": 0.0927,
      "step": 192150
    },
    {
      "epoch": 74.63,
      "learning_rate": 1.0049967637540454e-05,
      "loss": 0.0033,
      "step": 192160
    },
    {
      "epoch": 74.63,
      "learning_rate": 1.0049449838187704e-05,
      "loss": 0.1239,
      "step": 192170
    },
    {
      "epoch": 74.63,
      "learning_rate": 1.0048932038834953e-05,
      "loss": 0.0225,
      "step": 192180
    },
    {
      "epoch": 74.64,
      "learning_rate": 1.00484142394822e-05,
      "loss": 0.0005,
      "step": 192190
    },
    {
      "epoch": 74.64,
      "learning_rate": 1.004789644012945e-05,
      "loss": 0.0901,
      "step": 192200
    },
    {
      "epoch": 74.64,
      "learning_rate": 1.0047378640776699e-05,
      "loss": 0.1268,
      "step": 192210
    },
    {
      "epoch": 74.65,
      "learning_rate": 1.0046860841423949e-05,
      "loss": 0.1025,
      "step": 192220
    },
    {
      "epoch": 74.65,
      "learning_rate": 1.0046343042071198e-05,
      "loss": 0.0168,
      "step": 192230
    },
    {
      "epoch": 74.66,
      "learning_rate": 1.0045825242718448e-05,
      "loss": 0.1188,
      "step": 192240
    },
    {
      "epoch": 74.66,
      "learning_rate": 1.0045307443365697e-05,
      "loss": 0.0263,
      "step": 192250
    },
    {
      "epoch": 74.66,
      "learning_rate": 1.0044789644012947e-05,
      "loss": 0.0077,
      "step": 192260
    },
    {
      "epoch": 74.67,
      "learning_rate": 1.0044271844660193e-05,
      "loss": 0.0921,
      "step": 192270
    },
    {
      "epoch": 74.67,
      "learning_rate": 1.0043754045307443e-05,
      "loss": 0.009,
      "step": 192280
    },
    {
      "epoch": 74.68,
      "learning_rate": 1.0043236245954692e-05,
      "loss": 0.0143,
      "step": 192290
    },
    {
      "epoch": 74.68,
      "learning_rate": 1.0042718446601942e-05,
      "loss": 0.0253,
      "step": 192300
    },
    {
      "epoch": 74.68,
      "learning_rate": 1.0042200647249192e-05,
      "loss": 0.058,
      "step": 192310
    },
    {
      "epoch": 74.69,
      "learning_rate": 1.0041682847896441e-05,
      "loss": 0.05,
      "step": 192320
    },
    {
      "epoch": 74.69,
      "learning_rate": 1.0041165048543691e-05,
      "loss": 0.0147,
      "step": 192330
    },
    {
      "epoch": 74.7,
      "learning_rate": 1.004064724919094e-05,
      "loss": 0.0075,
      "step": 192340
    },
    {
      "epoch": 74.7,
      "learning_rate": 1.004012944983819e-05,
      "loss": 0.0398,
      "step": 192350
    },
    {
      "epoch": 74.7,
      "learning_rate": 1.0039611650485436e-05,
      "loss": 0.0471,
      "step": 192360
    },
    {
      "epoch": 74.71,
      "learning_rate": 1.0039093851132686e-05,
      "loss": 0.082,
      "step": 192370
    },
    {
      "epoch": 74.71,
      "learning_rate": 1.0038576051779936e-05,
      "loss": 0.0403,
      "step": 192380
    },
    {
      "epoch": 74.71,
      "learning_rate": 1.0038058252427185e-05,
      "loss": 0.0817,
      "step": 192390
    },
    {
      "epoch": 74.72,
      "learning_rate": 1.0037540453074435e-05,
      "loss": 0.136,
      "step": 192400
    },
    {
      "epoch": 74.72,
      "learning_rate": 1.0037022653721684e-05,
      "loss": 0.138,
      "step": 192410
    },
    {
      "epoch": 74.73,
      "learning_rate": 1.0036504854368934e-05,
      "loss": 0.1607,
      "step": 192420
    },
    {
      "epoch": 74.73,
      "learning_rate": 1.0035987055016184e-05,
      "loss": 0.1907,
      "step": 192430
    },
    {
      "epoch": 74.73,
      "learning_rate": 1.003546925566343e-05,
      "loss": 0.0185,
      "step": 192440
    },
    {
      "epoch": 74.74,
      "learning_rate": 1.003495145631068e-05,
      "loss": 0.0715,
      "step": 192450
    },
    {
      "epoch": 74.74,
      "learning_rate": 1.003443365695793e-05,
      "loss": 0.0025,
      "step": 192460
    },
    {
      "epoch": 74.75,
      "learning_rate": 1.0033915857605179e-05,
      "loss": 0.0311,
      "step": 192470
    },
    {
      "epoch": 74.75,
      "learning_rate": 1.0033398058252428e-05,
      "loss": 0.0642,
      "step": 192480
    },
    {
      "epoch": 74.75,
      "learning_rate": 1.0032880258899678e-05,
      "loss": 0.0819,
      "step": 192490
    },
    {
      "epoch": 74.76,
      "learning_rate": 1.0032362459546928e-05,
      "loss": 0.0041,
      "step": 192500
    },
    {
      "epoch": 74.76,
      "learning_rate": 1.0031844660194177e-05,
      "loss": 0.0309,
      "step": 192510
    },
    {
      "epoch": 74.77,
      "learning_rate": 1.0031326860841423e-05,
      "loss": 0.0757,
      "step": 192520
    },
    {
      "epoch": 74.77,
      "learning_rate": 1.0030809061488673e-05,
      "loss": 0.1517,
      "step": 192530
    },
    {
      "epoch": 74.77,
      "learning_rate": 1.0030291262135923e-05,
      "loss": 0.0057,
      "step": 192540
    },
    {
      "epoch": 74.78,
      "learning_rate": 1.0029773462783172e-05,
      "loss": 0.0709,
      "step": 192550
    },
    {
      "epoch": 74.78,
      "learning_rate": 1.0029255663430422e-05,
      "loss": 0.1319,
      "step": 192560
    },
    {
      "epoch": 74.78,
      "learning_rate": 1.0028737864077672e-05,
      "loss": 0.1234,
      "step": 192570
    },
    {
      "epoch": 74.79,
      "learning_rate": 1.0028220064724921e-05,
      "loss": 0.1415,
      "step": 192580
    },
    {
      "epoch": 74.79,
      "learning_rate": 1.002770226537217e-05,
      "loss": 0.0542,
      "step": 192590
    },
    {
      "epoch": 74.8,
      "learning_rate": 1.0027184466019417e-05,
      "loss": 0.055,
      "step": 192600
    },
    {
      "epoch": 74.8,
      "learning_rate": 1.0026666666666667e-05,
      "loss": 0.0143,
      "step": 192610
    },
    {
      "epoch": 74.8,
      "learning_rate": 1.0026148867313916e-05,
      "loss": 0.0928,
      "step": 192620
    },
    {
      "epoch": 74.81,
      "learning_rate": 1.0025631067961166e-05,
      "loss": 0.0971,
      "step": 192630
    },
    {
      "epoch": 74.81,
      "learning_rate": 1.0025113268608416e-05,
      "loss": 0.0688,
      "step": 192640
    },
    {
      "epoch": 74.82,
      "learning_rate": 1.0024595469255665e-05,
      "loss": 0.1474,
      "step": 192650
    },
    {
      "epoch": 74.82,
      "learning_rate": 1.0024077669902915e-05,
      "loss": 0.1121,
      "step": 192660
    },
    {
      "epoch": 74.82,
      "learning_rate": 1.0023559870550164e-05,
      "loss": 0.0431,
      "step": 192670
    },
    {
      "epoch": 74.83,
      "learning_rate": 1.002304207119741e-05,
      "loss": 0.0157,
      "step": 192680
    },
    {
      "epoch": 74.83,
      "learning_rate": 1.002252427184466e-05,
      "loss": 0.0427,
      "step": 192690
    },
    {
      "epoch": 74.83,
      "learning_rate": 1.002200647249191e-05,
      "loss": 0.0264,
      "step": 192700
    },
    {
      "epoch": 74.84,
      "learning_rate": 1.002148867313916e-05,
      "loss": 0.0338,
      "step": 192710
    },
    {
      "epoch": 74.84,
      "learning_rate": 1.0020970873786409e-05,
      "loss": 0.1123,
      "step": 192720
    },
    {
      "epoch": 74.85,
      "learning_rate": 1.0020453074433659e-05,
      "loss": 0.0161,
      "step": 192730
    },
    {
      "epoch": 74.85,
      "learning_rate": 1.0019935275080908e-05,
      "loss": 0.0025,
      "step": 192740
    },
    {
      "epoch": 74.85,
      "learning_rate": 1.0019417475728156e-05,
      "loss": 0.0187,
      "step": 192750
    },
    {
      "epoch": 74.86,
      "learning_rate": 1.0018899676375404e-05,
      "loss": 0.0003,
      "step": 192760
    },
    {
      "epoch": 74.86,
      "learning_rate": 1.0018381877022654e-05,
      "loss": 0.0763,
      "step": 192770
    },
    {
      "epoch": 74.87,
      "learning_rate": 1.0017864077669903e-05,
      "loss": 0.0877,
      "step": 192780
    },
    {
      "epoch": 74.87,
      "learning_rate": 1.0017346278317153e-05,
      "loss": 0.0623,
      "step": 192790
    },
    {
      "epoch": 74.87,
      "learning_rate": 1.0016828478964403e-05,
      "loss": 0.1048,
      "step": 192800
    },
    {
      "epoch": 74.88,
      "learning_rate": 1.0016310679611652e-05,
      "loss": 0.046,
      "step": 192810
    },
    {
      "epoch": 74.88,
      "learning_rate": 1.0015792880258902e-05,
      "loss": 0.1502,
      "step": 192820
    },
    {
      "epoch": 74.89,
      "learning_rate": 1.001527508090615e-05,
      "loss": 0.0348,
      "step": 192830
    },
    {
      "epoch": 74.89,
      "learning_rate": 1.0014757281553398e-05,
      "loss": 0.0539,
      "step": 192840
    },
    {
      "epoch": 74.89,
      "learning_rate": 1.0014239482200647e-05,
      "loss": 0.1101,
      "step": 192850
    },
    {
      "epoch": 74.9,
      "learning_rate": 1.0013721682847897e-05,
      "loss": 0.0726,
      "step": 192860
    },
    {
      "epoch": 74.9,
      "learning_rate": 1.0013203883495147e-05,
      "loss": 0.0696,
      "step": 192870
    },
    {
      "epoch": 74.9,
      "learning_rate": 1.0012686084142396e-05,
      "loss": 0.0402,
      "step": 192880
    },
    {
      "epoch": 74.91,
      "learning_rate": 1.0012168284789646e-05,
      "loss": 0.0519,
      "step": 192890
    },
    {
      "epoch": 74.91,
      "learning_rate": 1.0011650485436894e-05,
      "loss": 0.0362,
      "step": 192900
    },
    {
      "epoch": 74.92,
      "learning_rate": 1.0011132686084143e-05,
      "loss": 0.1967,
      "step": 192910
    },
    {
      "epoch": 74.92,
      "learning_rate": 1.0010614886731393e-05,
      "loss": 0.1408,
      "step": 192920
    },
    {
      "epoch": 74.92,
      "learning_rate": 1.0010097087378641e-05,
      "loss": 0.1049,
      "step": 192930
    },
    {
      "epoch": 74.93,
      "learning_rate": 1.000957928802589e-05,
      "loss": 0.0526,
      "step": 192940
    },
    {
      "epoch": 74.93,
      "learning_rate": 1.000906148867314e-05,
      "loss": 0.0089,
      "step": 192950
    },
    {
      "epoch": 74.94,
      "learning_rate": 1.000854368932039e-05,
      "loss": 0.0074,
      "step": 192960
    },
    {
      "epoch": 74.94,
      "learning_rate": 1.000802588996764e-05,
      "loss": 0.0969,
      "step": 192970
    },
    {
      "epoch": 74.94,
      "learning_rate": 1.0007508090614887e-05,
      "loss": 0.0211,
      "step": 192980
    },
    {
      "epoch": 74.95,
      "learning_rate": 1.0006990291262137e-05,
      "loss": 0.0663,
      "step": 192990
    },
    {
      "epoch": 74.95,
      "learning_rate": 1.0006472491909387e-05,
      "loss": 0.0741,
      "step": 193000
    },
    {
      "epoch": 74.96,
      "learning_rate": 1.0005954692556634e-05,
      "loss": 0.0683,
      "step": 193010
    },
    {
      "epoch": 74.96,
      "learning_rate": 1.0005436893203884e-05,
      "loss": 0.1097,
      "step": 193020
    },
    {
      "epoch": 74.96,
      "learning_rate": 1.0004919093851134e-05,
      "loss": 0.0691,
      "step": 193030
    },
    {
      "epoch": 74.97,
      "learning_rate": 1.0004401294498383e-05,
      "loss": 0.1019,
      "step": 193040
    },
    {
      "epoch": 74.97,
      "learning_rate": 1.0003883495145631e-05,
      "loss": 0.0875,
      "step": 193050
    },
    {
      "epoch": 74.97,
      "learning_rate": 1.000336569579288e-05,
      "loss": 0.0345,
      "step": 193060
    },
    {
      "epoch": 74.98,
      "learning_rate": 1.000284789644013e-05,
      "loss": 0.0373,
      "step": 193070
    },
    {
      "epoch": 74.98,
      "learning_rate": 1.000233009708738e-05,
      "loss": 0.1415,
      "step": 193080
    },
    {
      "epoch": 74.99,
      "learning_rate": 1.0001812297734628e-05,
      "loss": 0.0051,
      "step": 193090
    },
    {
      "epoch": 74.99,
      "learning_rate": 1.0001294498381878e-05,
      "loss": 0.028,
      "step": 193100
    },
    {
      "epoch": 74.99,
      "learning_rate": 1.0000776699029127e-05,
      "loss": 0.0484,
      "step": 193110
    },
    {
      "epoch": 75.0,
      "learning_rate": 1.0000258899676377e-05,
      "loss": 0.0702,
      "step": 193120
    },
    {
      "epoch": 75.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.3600085973739624,
      "eval_runtime": 8.1812,
      "eval_samples_per_second": 444.312,
      "eval_steps_per_second": 55.615,
      "step": 193125
    },
    {
      "epoch": 75.0,
      "learning_rate": 9.999741100323625e-06,
      "loss": 0.0287,
      "step": 193130
    },
    {
      "epoch": 75.01,
      "learning_rate": 9.999223300970874e-06,
      "loss": 0.1373,
      "step": 193140
    },
    {
      "epoch": 75.01,
      "learning_rate": 9.998705501618124e-06,
      "loss": 0.0446,
      "step": 193150
    },
    {
      "epoch": 75.01,
      "learning_rate": 9.998187702265374e-06,
      "loss": 0.1161,
      "step": 193160
    },
    {
      "epoch": 75.02,
      "learning_rate": 9.997669902912622e-06,
      "loss": 0.1237,
      "step": 193170
    },
    {
      "epoch": 75.02,
      "learning_rate": 9.997152103559871e-06,
      "loss": 0.0003,
      "step": 193180
    },
    {
      "epoch": 75.03,
      "learning_rate": 9.99663430420712e-06,
      "loss": 0.0163,
      "step": 193190
    },
    {
      "epoch": 75.03,
      "learning_rate": 9.99611650485437e-06,
      "loss": 0.0835,
      "step": 193200
    },
    {
      "epoch": 75.03,
      "learning_rate": 9.995598705501618e-06,
      "loss": 0.0394,
      "step": 193210
    },
    {
      "epoch": 75.04,
      "learning_rate": 9.995080906148868e-06,
      "loss": 0.0698,
      "step": 193220
    },
    {
      "epoch": 75.04,
      "learning_rate": 9.994563106796118e-06,
      "loss": 0.052,
      "step": 193230
    },
    {
      "epoch": 75.04,
      "learning_rate": 9.994045307443365e-06,
      "loss": 0.055,
      "step": 193240
    },
    {
      "epoch": 75.05,
      "learning_rate": 9.993527508090615e-06,
      "loss": 0.177,
      "step": 193250
    },
    {
      "epoch": 75.05,
      "learning_rate": 9.993009708737865e-06,
      "loss": 0.0341,
      "step": 193260
    },
    {
      "epoch": 75.06,
      "learning_rate": 9.992491909385114e-06,
      "loss": 0.0431,
      "step": 193270
    },
    {
      "epoch": 75.06,
      "learning_rate": 9.991974110032362e-06,
      "loss": 0.0077,
      "step": 193280
    },
    {
      "epoch": 75.06,
      "learning_rate": 9.991456310679612e-06,
      "loss": 0.1538,
      "step": 193290
    },
    {
      "epoch": 75.07,
      "learning_rate": 9.990938511326862e-06,
      "loss": 0.0356,
      "step": 193300
    },
    {
      "epoch": 75.07,
      "learning_rate": 9.990420711974111e-06,
      "loss": 0.0008,
      "step": 193310
    },
    {
      "epoch": 75.08,
      "learning_rate": 9.989902912621359e-06,
      "loss": 0.067,
      "step": 193320
    },
    {
      "epoch": 75.08,
      "learning_rate": 9.989385113268609e-06,
      "loss": 0.0005,
      "step": 193330
    },
    {
      "epoch": 75.08,
      "learning_rate": 9.988867313915858e-06,
      "loss": 0.1441,
      "step": 193340
    },
    {
      "epoch": 75.09,
      "learning_rate": 9.988349514563108e-06,
      "loss": 0.0011,
      "step": 193350
    },
    {
      "epoch": 75.09,
      "learning_rate": 9.987831715210356e-06,
      "loss": 0.1236,
      "step": 193360
    },
    {
      "epoch": 75.1,
      "learning_rate": 9.987313915857605e-06,
      "loss": 0.1442,
      "step": 193370
    },
    {
      "epoch": 75.1,
      "learning_rate": 9.986796116504855e-06,
      "loss": 0.1062,
      "step": 193380
    },
    {
      "epoch": 75.1,
      "learning_rate": 9.986278317152103e-06,
      "loss": 0.1093,
      "step": 193390
    },
    {
      "epoch": 75.11,
      "learning_rate": 9.985760517799353e-06,
      "loss": 0.1529,
      "step": 193400
    },
    {
      "epoch": 75.11,
      "learning_rate": 9.985242718446602e-06,
      "loss": 0.0003,
      "step": 193410
    },
    {
      "epoch": 75.11,
      "learning_rate": 9.984724919093852e-06,
      "loss": 0.0171,
      "step": 193420
    },
    {
      "epoch": 75.12,
      "learning_rate": 9.984207119741101e-06,
      "loss": 0.0532,
      "step": 193430
    },
    {
      "epoch": 75.12,
      "learning_rate": 9.98368932038835e-06,
      "loss": 0.291,
      "step": 193440
    },
    {
      "epoch": 75.13,
      "learning_rate": 9.983171521035599e-06,
      "loss": 0.0603,
      "step": 193450
    },
    {
      "epoch": 75.13,
      "learning_rate": 9.982653721682849e-06,
      "loss": 0.062,
      "step": 193460
    },
    {
      "epoch": 75.13,
      "learning_rate": 9.982135922330098e-06,
      "loss": 0.0421,
      "step": 193470
    },
    {
      "epoch": 75.14,
      "learning_rate": 9.981618122977346e-06,
      "loss": 0.0061,
      "step": 193480
    },
    {
      "epoch": 75.14,
      "learning_rate": 9.981100323624596e-06,
      "loss": 0.1542,
      "step": 193490
    },
    {
      "epoch": 75.15,
      "learning_rate": 9.980582524271845e-06,
      "loss": 0.0034,
      "step": 193500
    },
    {
      "epoch": 75.15,
      "learning_rate": 9.980064724919095e-06,
      "loss": 0.0158,
      "step": 193510
    },
    {
      "epoch": 75.15,
      "learning_rate": 9.979546925566343e-06,
      "loss": 0.0565,
      "step": 193520
    },
    {
      "epoch": 75.16,
      "learning_rate": 9.979029126213593e-06,
      "loss": 0.1973,
      "step": 193530
    },
    {
      "epoch": 75.16,
      "learning_rate": 9.978511326860842e-06,
      "loss": 0.0283,
      "step": 193540
    },
    {
      "epoch": 75.17,
      "learning_rate": 9.977993527508092e-06,
      "loss": 0.2026,
      "step": 193550
    },
    {
      "epoch": 75.17,
      "learning_rate": 9.97747572815534e-06,
      "loss": 0.0131,
      "step": 193560
    },
    {
      "epoch": 75.17,
      "learning_rate": 9.97695792880259e-06,
      "loss": 0.0897,
      "step": 193570
    },
    {
      "epoch": 75.18,
      "learning_rate": 9.976440129449839e-06,
      "loss": 0.015,
      "step": 193580
    },
    {
      "epoch": 75.18,
      "learning_rate": 9.975922330097089e-06,
      "loss": 0.0649,
      "step": 193590
    },
    {
      "epoch": 75.18,
      "learning_rate": 9.975404530744336e-06,
      "loss": 0.085,
      "step": 193600
    },
    {
      "epoch": 75.19,
      "learning_rate": 9.974886731391586e-06,
      "loss": 0.1915,
      "step": 193610
    },
    {
      "epoch": 75.19,
      "learning_rate": 9.974368932038836e-06,
      "loss": 0.0079,
      "step": 193620
    },
    {
      "epoch": 75.2,
      "learning_rate": 9.973851132686085e-06,
      "loss": 0.0003,
      "step": 193630
    },
    {
      "epoch": 75.2,
      "learning_rate": 9.973333333333333e-06,
      "loss": 0.1474,
      "step": 193640
    },
    {
      "epoch": 75.2,
      "learning_rate": 9.972815533980583e-06,
      "loss": 0.1088,
      "step": 193650
    },
    {
      "epoch": 75.21,
      "learning_rate": 9.972297734627833e-06,
      "loss": 0.1359,
      "step": 193660
    },
    {
      "epoch": 75.21,
      "learning_rate": 9.971779935275082e-06,
      "loss": 0.0008,
      "step": 193670
    },
    {
      "epoch": 75.22,
      "learning_rate": 9.97126213592233e-06,
      "loss": 0.0642,
      "step": 193680
    },
    {
      "epoch": 75.22,
      "learning_rate": 9.97074433656958e-06,
      "loss": 0.0206,
      "step": 193690
    },
    {
      "epoch": 75.22,
      "learning_rate": 9.97022653721683e-06,
      "loss": 0.0676,
      "step": 193700
    },
    {
      "epoch": 75.23,
      "learning_rate": 9.969708737864079e-06,
      "loss": 0.0265,
      "step": 193710
    },
    {
      "epoch": 75.23,
      "learning_rate": 9.969190938511327e-06,
      "loss": 0.1958,
      "step": 193720
    },
    {
      "epoch": 75.23,
      "learning_rate": 9.968673139158576e-06,
      "loss": 0.0846,
      "step": 193730
    },
    {
      "epoch": 75.24,
      "learning_rate": 9.968155339805826e-06,
      "loss": 0.0315,
      "step": 193740
    },
    {
      "epoch": 75.24,
      "learning_rate": 9.967637540453076e-06,
      "loss": 0.1283,
      "step": 193750
    },
    {
      "epoch": 75.25,
      "learning_rate": 9.967119741100324e-06,
      "loss": 0.0015,
      "step": 193760
    },
    {
      "epoch": 75.25,
      "learning_rate": 9.966601941747573e-06,
      "loss": 0.0018,
      "step": 193770
    },
    {
      "epoch": 75.25,
      "learning_rate": 9.966084142394823e-06,
      "loss": 0.0222,
      "step": 193780
    },
    {
      "epoch": 75.26,
      "learning_rate": 9.965566343042072e-06,
      "loss": 0.0498,
      "step": 193790
    },
    {
      "epoch": 75.26,
      "learning_rate": 9.96504854368932e-06,
      "loss": 0.0078,
      "step": 193800
    },
    {
      "epoch": 75.27,
      "learning_rate": 9.96453074433657e-06,
      "loss": 0.1294,
      "step": 193810
    },
    {
      "epoch": 75.27,
      "learning_rate": 9.96401294498382e-06,
      "loss": 0.0547,
      "step": 193820
    },
    {
      "epoch": 75.27,
      "learning_rate": 9.96349514563107e-06,
      "loss": 0.0382,
      "step": 193830
    },
    {
      "epoch": 75.28,
      "learning_rate": 9.962977346278317e-06,
      "loss": 0.005,
      "step": 193840
    },
    {
      "epoch": 75.28,
      "learning_rate": 9.962459546925567e-06,
      "loss": 0.1343,
      "step": 193850
    },
    {
      "epoch": 75.29,
      "learning_rate": 9.961941747572816e-06,
      "loss": 0.0177,
      "step": 193860
    },
    {
      "epoch": 75.29,
      "learning_rate": 9.961423948220066e-06,
      "loss": 0.0295,
      "step": 193870
    },
    {
      "epoch": 75.29,
      "learning_rate": 9.960906148867314e-06,
      "loss": 0.0668,
      "step": 193880
    },
    {
      "epoch": 75.3,
      "learning_rate": 9.960388349514564e-06,
      "loss": 0.1177,
      "step": 193890
    },
    {
      "epoch": 75.3,
      "learning_rate": 9.959870550161813e-06,
      "loss": 0.035,
      "step": 193900
    },
    {
      "epoch": 75.3,
      "learning_rate": 9.959352750809063e-06,
      "loss": 0.1088,
      "step": 193910
    },
    {
      "epoch": 75.31,
      "learning_rate": 9.95883495145631e-06,
      "loss": 0.0635,
      "step": 193920
    },
    {
      "epoch": 75.31,
      "learning_rate": 9.95831715210356e-06,
      "loss": 0.0007,
      "step": 193930
    },
    {
      "epoch": 75.32,
      "learning_rate": 9.95779935275081e-06,
      "loss": 0.0242,
      "step": 193940
    },
    {
      "epoch": 75.32,
      "learning_rate": 9.95728155339806e-06,
      "loss": 0.0917,
      "step": 193950
    },
    {
      "epoch": 75.32,
      "learning_rate": 9.956763754045308e-06,
      "loss": 0.2047,
      "step": 193960
    },
    {
      "epoch": 75.33,
      "learning_rate": 9.956245954692557e-06,
      "loss": 0.1515,
      "step": 193970
    },
    {
      "epoch": 75.33,
      "learning_rate": 9.955728155339807e-06,
      "loss": 0.0192,
      "step": 193980
    },
    {
      "epoch": 75.34,
      "learning_rate": 9.955210355987056e-06,
      "loss": 0.0158,
      "step": 193990
    },
    {
      "epoch": 75.34,
      "learning_rate": 9.954692556634306e-06,
      "loss": 0.0196,
      "step": 194000
    },
    {
      "epoch": 75.34,
      "learning_rate": 9.954174757281554e-06,
      "loss": 0.0687,
      "step": 194010
    },
    {
      "epoch": 75.35,
      "learning_rate": 9.953656957928804e-06,
      "loss": 0.1105,
      "step": 194020
    },
    {
      "epoch": 75.35,
      "learning_rate": 9.953139158576053e-06,
      "loss": 0.1153,
      "step": 194030
    },
    {
      "epoch": 75.36,
      "learning_rate": 9.952621359223303e-06,
      "loss": 0.1458,
      "step": 194040
    },
    {
      "epoch": 75.36,
      "learning_rate": 9.95210355987055e-06,
      "loss": 0.2594,
      "step": 194050
    },
    {
      "epoch": 75.36,
      "learning_rate": 9.9515857605178e-06,
      "loss": 0.1193,
      "step": 194060
    },
    {
      "epoch": 75.37,
      "learning_rate": 9.95106796116505e-06,
      "loss": 0.0782,
      "step": 194070
    },
    {
      "epoch": 75.37,
      "learning_rate": 9.9505501618123e-06,
      "loss": 0.1256,
      "step": 194080
    },
    {
      "epoch": 75.37,
      "learning_rate": 9.950032362459547e-06,
      "loss": 0.0702,
      "step": 194090
    },
    {
      "epoch": 75.38,
      "learning_rate": 9.949514563106797e-06,
      "loss": 0.0441,
      "step": 194100
    },
    {
      "epoch": 75.38,
      "learning_rate": 9.948996763754047e-06,
      "loss": 0.055,
      "step": 194110
    },
    {
      "epoch": 75.39,
      "learning_rate": 9.948478964401296e-06,
      "loss": 0.0179,
      "step": 194120
    },
    {
      "epoch": 75.39,
      "learning_rate": 9.947961165048544e-06,
      "loss": 0.1141,
      "step": 194130
    },
    {
      "epoch": 75.39,
      "learning_rate": 9.947443365695794e-06,
      "loss": 0.0002,
      "step": 194140
    },
    {
      "epoch": 75.4,
      "learning_rate": 9.946925566343043e-06,
      "loss": 0.069,
      "step": 194150
    },
    {
      "epoch": 75.4,
      "learning_rate": 9.946407766990293e-06,
      "loss": 0.0577,
      "step": 194160
    },
    {
      "epoch": 75.41,
      "learning_rate": 9.945889967637541e-06,
      "loss": 0.0279,
      "step": 194170
    },
    {
      "epoch": 75.41,
      "learning_rate": 9.94537216828479e-06,
      "loss": 0.0652,
      "step": 194180
    },
    {
      "epoch": 75.41,
      "learning_rate": 9.94485436893204e-06,
      "loss": 0.0375,
      "step": 194190
    },
    {
      "epoch": 75.42,
      "learning_rate": 9.94433656957929e-06,
      "loss": 0.1226,
      "step": 194200
    },
    {
      "epoch": 75.42,
      "learning_rate": 9.943818770226538e-06,
      "loss": 0.1914,
      "step": 194210
    },
    {
      "epoch": 75.43,
      "learning_rate": 9.943300970873787e-06,
      "loss": 0.0542,
      "step": 194220
    },
    {
      "epoch": 75.43,
      "learning_rate": 9.942783171521037e-06,
      "loss": 0.1068,
      "step": 194230
    },
    {
      "epoch": 75.43,
      "learning_rate": 9.942265372168287e-06,
      "loss": 0.0315,
      "step": 194240
    },
    {
      "epoch": 75.44,
      "learning_rate": 9.941747572815535e-06,
      "loss": 0.1178,
      "step": 194250
    },
    {
      "epoch": 75.44,
      "learning_rate": 9.941229773462784e-06,
      "loss": 0.0582,
      "step": 194260
    },
    {
      "epoch": 75.44,
      "learning_rate": 9.940711974110034e-06,
      "loss": 0.0324,
      "step": 194270
    },
    {
      "epoch": 75.45,
      "learning_rate": 9.940194174757283e-06,
      "loss": 0.0895,
      "step": 194280
    },
    {
      "epoch": 75.45,
      "learning_rate": 9.939676375404531e-06,
      "loss": 0.0734,
      "step": 194290
    },
    {
      "epoch": 75.46,
      "learning_rate": 9.939158576051781e-06,
      "loss": 0.0325,
      "step": 194300
    },
    {
      "epoch": 75.46,
      "learning_rate": 9.93864077669903e-06,
      "loss": 0.0867,
      "step": 194310
    },
    {
      "epoch": 75.46,
      "learning_rate": 9.93812297734628e-06,
      "loss": 0.0884,
      "step": 194320
    },
    {
      "epoch": 75.47,
      "learning_rate": 9.937605177993528e-06,
      "loss": 0.088,
      "step": 194330
    },
    {
      "epoch": 75.47,
      "learning_rate": 9.937087378640778e-06,
      "loss": 0.0252,
      "step": 194340
    },
    {
      "epoch": 75.48,
      "learning_rate": 9.936569579288027e-06,
      "loss": 0.2104,
      "step": 194350
    },
    {
      "epoch": 75.48,
      "learning_rate": 9.936051779935277e-06,
      "loss": 0.0523,
      "step": 194360
    },
    {
      "epoch": 75.48,
      "learning_rate": 9.935533980582525e-06,
      "loss": 0.0376,
      "step": 194370
    },
    {
      "epoch": 75.49,
      "learning_rate": 9.935016181229775e-06,
      "loss": 0.1502,
      "step": 194380
    },
    {
      "epoch": 75.49,
      "learning_rate": 9.934498381877024e-06,
      "loss": 0.0684,
      "step": 194390
    },
    {
      "epoch": 75.5,
      "learning_rate": 9.933980582524274e-06,
      "loss": 0.0782,
      "step": 194400
    },
    {
      "epoch": 75.5,
      "learning_rate": 9.933462783171522e-06,
      "loss": 0.0573,
      "step": 194410
    },
    {
      "epoch": 75.5,
      "learning_rate": 9.932944983818771e-06,
      "loss": 0.0327,
      "step": 194420
    },
    {
      "epoch": 75.51,
      "learning_rate": 9.932427184466021e-06,
      "loss": 0.1462,
      "step": 194430
    },
    {
      "epoch": 75.51,
      "learning_rate": 9.93190938511327e-06,
      "loss": 0.0642,
      "step": 194440
    },
    {
      "epoch": 75.51,
      "learning_rate": 9.931391585760518e-06,
      "loss": 0.0702,
      "step": 194450
    },
    {
      "epoch": 75.52,
      "learning_rate": 9.930873786407768e-06,
      "loss": 0.0219,
      "step": 194460
    },
    {
      "epoch": 75.52,
      "learning_rate": 9.930355987055018e-06,
      "loss": 0.0025,
      "step": 194470
    },
    {
      "epoch": 75.53,
      "learning_rate": 9.929838187702267e-06,
      "loss": 0.0191,
      "step": 194480
    },
    {
      "epoch": 75.53,
      "learning_rate": 9.929320388349515e-06,
      "loss": 0.0806,
      "step": 194490
    },
    {
      "epoch": 75.53,
      "learning_rate": 9.928802588996765e-06,
      "loss": 0.0015,
      "step": 194500
    },
    {
      "epoch": 75.54,
      "learning_rate": 9.928284789644014e-06,
      "loss": 0.1148,
      "step": 194510
    },
    {
      "epoch": 75.54,
      "learning_rate": 9.927766990291262e-06,
      "loss": 0.0029,
      "step": 194520
    },
    {
      "epoch": 75.55,
      "learning_rate": 9.927249190938512e-06,
      "loss": 0.1723,
      "step": 194530
    },
    {
      "epoch": 75.55,
      "learning_rate": 9.926731391585762e-06,
      "loss": 0.0011,
      "step": 194540
    },
    {
      "epoch": 75.55,
      "learning_rate": 9.926213592233011e-06,
      "loss": 0.0182,
      "step": 194550
    },
    {
      "epoch": 75.56,
      "learning_rate": 9.92569579288026e-06,
      "loss": 0.0585,
      "step": 194560
    },
    {
      "epoch": 75.56,
      "learning_rate": 9.925177993527509e-06,
      "loss": 0.1335,
      "step": 194570
    },
    {
      "epoch": 75.57,
      "learning_rate": 9.924660194174758e-06,
      "loss": 0.0854,
      "step": 194580
    },
    {
      "epoch": 75.57,
      "learning_rate": 9.924142394822008e-06,
      "loss": 0.0595,
      "step": 194590
    },
    {
      "epoch": 75.57,
      "learning_rate": 9.923624595469256e-06,
      "loss": 0.1089,
      "step": 194600
    },
    {
      "epoch": 75.58,
      "learning_rate": 9.923106796116506e-06,
      "loss": 0.0672,
      "step": 194610
    },
    {
      "epoch": 75.58,
      "learning_rate": 9.922588996763755e-06,
      "loss": 0.0096,
      "step": 194620
    },
    {
      "epoch": 75.58,
      "learning_rate": 9.922071197411005e-06,
      "loss": 0.0307,
      "step": 194630
    },
    {
      "epoch": 75.59,
      "learning_rate": 9.921553398058253e-06,
      "loss": 0.0015,
      "step": 194640
    },
    {
      "epoch": 75.59,
      "learning_rate": 9.921035598705502e-06,
      "loss": 0.0319,
      "step": 194650
    },
    {
      "epoch": 75.6,
      "learning_rate": 9.920517799352752e-06,
      "loss": 0.081,
      "step": 194660
    },
    {
      "epoch": 75.6,
      "learning_rate": 9.920000000000002e-06,
      "loss": 0.0465,
      "step": 194670
    },
    {
      "epoch": 75.6,
      "learning_rate": 9.91948220064725e-06,
      "loss": 0.1223,
      "step": 194680
    },
    {
      "epoch": 75.61,
      "learning_rate": 9.918964401294499e-06,
      "loss": 0.04,
      "step": 194690
    },
    {
      "epoch": 75.61,
      "learning_rate": 9.918446601941749e-06,
      "loss": 0.0843,
      "step": 194700
    },
    {
      "epoch": 75.62,
      "learning_rate": 9.917928802588997e-06,
      "loss": 0.1177,
      "step": 194710
    },
    {
      "epoch": 75.62,
      "learning_rate": 9.917411003236246e-06,
      "loss": 0.0748,
      "step": 194720
    },
    {
      "epoch": 75.62,
      "learning_rate": 9.916893203883496e-06,
      "loss": 0.1224,
      "step": 194730
    },
    {
      "epoch": 75.63,
      "learning_rate": 9.916375404530746e-06,
      "loss": 0.0761,
      "step": 194740
    },
    {
      "epoch": 75.63,
      "learning_rate": 9.915857605177993e-06,
      "loss": 0.0806,
      "step": 194750
    },
    {
      "epoch": 75.63,
      "learning_rate": 9.915339805825243e-06,
      "loss": 0.1322,
      "step": 194760
    },
    {
      "epoch": 75.64,
      "learning_rate": 9.914822006472493e-06,
      "loss": 0.046,
      "step": 194770
    },
    {
      "epoch": 75.64,
      "learning_rate": 9.914304207119742e-06,
      "loss": 0.0004,
      "step": 194780
    },
    {
      "epoch": 75.65,
      "learning_rate": 9.91378640776699e-06,
      "loss": 0.0144,
      "step": 194790
    },
    {
      "epoch": 75.65,
      "learning_rate": 9.91326860841424e-06,
      "loss": 0.0777,
      "step": 194800
    },
    {
      "epoch": 75.65,
      "learning_rate": 9.91275080906149e-06,
      "loss": 0.0797,
      "step": 194810
    },
    {
      "epoch": 75.66,
      "learning_rate": 9.912233009708739e-06,
      "loss": 0.0244,
      "step": 194820
    },
    {
      "epoch": 75.66,
      "learning_rate": 9.911715210355987e-06,
      "loss": 0.0156,
      "step": 194830
    },
    {
      "epoch": 75.67,
      "learning_rate": 9.911197411003237e-06,
      "loss": 0.0033,
      "step": 194840
    },
    {
      "epoch": 75.67,
      "learning_rate": 9.910679611650486e-06,
      "loss": 0.0326,
      "step": 194850
    },
    {
      "epoch": 75.67,
      "learning_rate": 9.910161812297734e-06,
      "loss": 0.0025,
      "step": 194860
    },
    {
      "epoch": 75.68,
      "learning_rate": 9.909644012944984e-06,
      "loss": 0.0718,
      "step": 194870
    },
    {
      "epoch": 75.68,
      "learning_rate": 9.909126213592233e-06,
      "loss": 0.0814,
      "step": 194880
    },
    {
      "epoch": 75.69,
      "learning_rate": 9.908608414239483e-06,
      "loss": 0.0133,
      "step": 194890
    },
    {
      "epoch": 75.69,
      "learning_rate": 9.908090614886731e-06,
      "loss": 0.0566,
      "step": 194900
    },
    {
      "epoch": 75.69,
      "learning_rate": 9.90757281553398e-06,
      "loss": 0.0081,
      "step": 194910
    },
    {
      "epoch": 75.7,
      "learning_rate": 9.90705501618123e-06,
      "loss": 0.053,
      "step": 194920
    },
    {
      "epoch": 75.7,
      "learning_rate": 9.90653721682848e-06,
      "loss": 0.0095,
      "step": 194930
    },
    {
      "epoch": 75.7,
      "learning_rate": 9.906019417475728e-06,
      "loss": 0.0507,
      "step": 194940
    },
    {
      "epoch": 75.71,
      "learning_rate": 9.905501618122977e-06,
      "loss": 0.0016,
      "step": 194950
    },
    {
      "epoch": 75.71,
      "learning_rate": 9.904983818770227e-06,
      "loss": 0.1255,
      "step": 194960
    },
    {
      "epoch": 75.72,
      "learning_rate": 9.904466019417477e-06,
      "loss": 0.1314,
      "step": 194970
    },
    {
      "epoch": 75.72,
      "learning_rate": 9.903948220064724e-06,
      "loss": 0.0243,
      "step": 194980
    },
    {
      "epoch": 75.72,
      "learning_rate": 9.903430420711974e-06,
      "loss": 0.0262,
      "step": 194990
    },
    {
      "epoch": 75.73,
      "learning_rate": 9.902912621359224e-06,
      "loss": 0.0179,
      "step": 195000
    },
    {
      "epoch": 75.73,
      "learning_rate": 9.902394822006473e-06,
      "loss": 0.0155,
      "step": 195010
    },
    {
      "epoch": 75.74,
      "learning_rate": 9.901877022653721e-06,
      "loss": 0.0016,
      "step": 195020
    },
    {
      "epoch": 75.74,
      "learning_rate": 9.901359223300971e-06,
      "loss": 0.0128,
      "step": 195030
    },
    {
      "epoch": 75.74,
      "learning_rate": 9.90084142394822e-06,
      "loss": 0.1132,
      "step": 195040
    },
    {
      "epoch": 75.75,
      "learning_rate": 9.90032362459547e-06,
      "loss": 0.0045,
      "step": 195050
    },
    {
      "epoch": 75.75,
      "learning_rate": 9.899805825242718e-06,
      "loss": 0.0462,
      "step": 195060
    },
    {
      "epoch": 75.76,
      "learning_rate": 9.899288025889968e-06,
      "loss": 0.1335,
      "step": 195070
    },
    {
      "epoch": 75.76,
      "learning_rate": 9.898770226537217e-06,
      "loss": 0.1145,
      "step": 195080
    },
    {
      "epoch": 75.76,
      "learning_rate": 9.898252427184467e-06,
      "loss": 0.0346,
      "step": 195090
    },
    {
      "epoch": 75.77,
      "learning_rate": 9.897734627831717e-06,
      "loss": 0.1411,
      "step": 195100
    },
    {
      "epoch": 75.77,
      "learning_rate": 9.897216828478964e-06,
      "loss": 0.0594,
      "step": 195110
    },
    {
      "epoch": 75.77,
      "learning_rate": 9.896699029126214e-06,
      "loss": 0.0424,
      "step": 195120
    },
    {
      "epoch": 75.78,
      "learning_rate": 9.896181229773464e-06,
      "loss": 0.2705,
      "step": 195130
    },
    {
      "epoch": 75.78,
      "learning_rate": 9.895663430420713e-06,
      "loss": 0.0372,
      "step": 195140
    },
    {
      "epoch": 75.79,
      "learning_rate": 9.895145631067961e-06,
      "loss": 0.001,
      "step": 195150
    },
    {
      "epoch": 75.79,
      "learning_rate": 9.89462783171521e-06,
      "loss": 0.0495,
      "step": 195160
    },
    {
      "epoch": 75.79,
      "learning_rate": 9.89411003236246e-06,
      "loss": 0.1372,
      "step": 195170
    },
    {
      "epoch": 75.8,
      "learning_rate": 9.89359223300971e-06,
      "loss": 0.1401,
      "step": 195180
    },
    {
      "epoch": 75.8,
      "learning_rate": 9.893074433656958e-06,
      "loss": 0.1624,
      "step": 195190
    },
    {
      "epoch": 75.81,
      "learning_rate": 9.892556634304208e-06,
      "loss": 0.0783,
      "step": 195200
    },
    {
      "epoch": 75.81,
      "learning_rate": 9.892038834951457e-06,
      "loss": 0.0258,
      "step": 195210
    },
    {
      "epoch": 75.81,
      "learning_rate": 9.891521035598707e-06,
      "loss": 0.0629,
      "step": 195220
    },
    {
      "epoch": 75.82,
      "learning_rate": 9.891003236245955e-06,
      "loss": 0.0476,
      "step": 195230
    },
    {
      "epoch": 75.82,
      "learning_rate": 9.890485436893204e-06,
      "loss": 0.036,
      "step": 195240
    },
    {
      "epoch": 75.83,
      "learning_rate": 9.889967637540454e-06,
      "loss": 0.0972,
      "step": 195250
    },
    {
      "epoch": 75.83,
      "learning_rate": 9.889449838187704e-06,
      "loss": 0.0015,
      "step": 195260
    },
    {
      "epoch": 75.83,
      "learning_rate": 9.888932038834952e-06,
      "loss": 0.0171,
      "step": 195270
    },
    {
      "epoch": 75.84,
      "learning_rate": 9.888414239482201e-06,
      "loss": 0.025,
      "step": 195280
    },
    {
      "epoch": 75.84,
      "learning_rate": 9.88789644012945e-06,
      "loss": 0.0042,
      "step": 195290
    },
    {
      "epoch": 75.84,
      "learning_rate": 9.8873786407767e-06,
      "loss": 0.1762,
      "step": 195300
    },
    {
      "epoch": 75.85,
      "learning_rate": 9.886860841423948e-06,
      "loss": 0.0315,
      "step": 195310
    },
    {
      "epoch": 75.85,
      "learning_rate": 9.886343042071198e-06,
      "loss": 0.0551,
      "step": 195320
    },
    {
      "epoch": 75.86,
      "learning_rate": 9.885825242718448e-06,
      "loss": 0.0929,
      "step": 195330
    },
    {
      "epoch": 75.86,
      "learning_rate": 9.885307443365697e-06,
      "loss": 0.0393,
      "step": 195340
    },
    {
      "epoch": 75.86,
      "learning_rate": 9.884789644012945e-06,
      "loss": 0.0161,
      "step": 195350
    },
    {
      "epoch": 75.87,
      "learning_rate": 9.884271844660195e-06,
      "loss": 0.0289,
      "step": 195360
    },
    {
      "epoch": 75.87,
      "learning_rate": 9.883754045307444e-06,
      "loss": 0.1172,
      "step": 195370
    },
    {
      "epoch": 75.88,
      "learning_rate": 9.883236245954694e-06,
      "loss": 0.021,
      "step": 195380
    },
    {
      "epoch": 75.88,
      "learning_rate": 9.882718446601942e-06,
      "loss": 0.0103,
      "step": 195390
    },
    {
      "epoch": 75.88,
      "learning_rate": 9.882200647249192e-06,
      "loss": 0.0002,
      "step": 195400
    },
    {
      "epoch": 75.89,
      "learning_rate": 9.881682847896441e-06,
      "loss": 0.2606,
      "step": 195410
    },
    {
      "epoch": 75.89,
      "learning_rate": 9.88116504854369e-06,
      "loss": 0.0002,
      "step": 195420
    },
    {
      "epoch": 75.9,
      "learning_rate": 9.880647249190939e-06,
      "loss": 0.1573,
      "step": 195430
    },
    {
      "epoch": 75.9,
      "learning_rate": 9.880129449838188e-06,
      "loss": 0.1398,
      "step": 195440
    },
    {
      "epoch": 75.9,
      "learning_rate": 9.879611650485438e-06,
      "loss": 0.0479,
      "step": 195450
    },
    {
      "epoch": 75.91,
      "learning_rate": 9.879093851132688e-06,
      "loss": 0.0061,
      "step": 195460
    },
    {
      "epoch": 75.91,
      "learning_rate": 9.878576051779935e-06,
      "loss": 0.0036,
      "step": 195470
    },
    {
      "epoch": 75.91,
      "learning_rate": 9.878058252427185e-06,
      "loss": 0.0913,
      "step": 195480
    },
    {
      "epoch": 75.92,
      "learning_rate": 9.877540453074435e-06,
      "loss": 0.0651,
      "step": 195490
    },
    {
      "epoch": 75.92,
      "learning_rate": 9.877022653721684e-06,
      "loss": 0.2052,
      "step": 195500
    },
    {
      "epoch": 75.93,
      "learning_rate": 9.876504854368932e-06,
      "loss": 0.0023,
      "step": 195510
    },
    {
      "epoch": 75.93,
      "learning_rate": 9.875987055016182e-06,
      "loss": 0.0613,
      "step": 195520
    },
    {
      "epoch": 75.93,
      "learning_rate": 9.875469255663431e-06,
      "loss": 0.1094,
      "step": 195530
    },
    {
      "epoch": 75.94,
      "learning_rate": 9.874951456310681e-06,
      "loss": 0.1584,
      "step": 195540
    },
    {
      "epoch": 75.94,
      "learning_rate": 9.874433656957929e-06,
      "loss": 0.0037,
      "step": 195550
    },
    {
      "epoch": 75.95,
      "learning_rate": 9.873915857605179e-06,
      "loss": 0.0179,
      "step": 195560
    },
    {
      "epoch": 75.95,
      "learning_rate": 9.873398058252428e-06,
      "loss": 0.0281,
      "step": 195570
    },
    {
      "epoch": 75.95,
      "learning_rate": 9.872880258899678e-06,
      "loss": 0.1004,
      "step": 195580
    },
    {
      "epoch": 75.96,
      "learning_rate": 9.872362459546926e-06,
      "loss": 0.08,
      "step": 195590
    },
    {
      "epoch": 75.96,
      "learning_rate": 9.871844660194175e-06,
      "loss": 0.0575,
      "step": 195600
    },
    {
      "epoch": 75.97,
      "learning_rate": 9.871326860841425e-06,
      "loss": 0.0345,
      "step": 195610
    },
    {
      "epoch": 75.97,
      "learning_rate": 9.870809061488675e-06,
      "loss": 0.1059,
      "step": 195620
    },
    {
      "epoch": 75.97,
      "learning_rate": 9.870291262135923e-06,
      "loss": 0.0832,
      "step": 195630
    },
    {
      "epoch": 75.98,
      "learning_rate": 9.869773462783172e-06,
      "loss": 0.0513,
      "step": 195640
    },
    {
      "epoch": 75.98,
      "learning_rate": 9.869255663430422e-06,
      "loss": 0.1186,
      "step": 195650
    },
    {
      "epoch": 75.98,
      "learning_rate": 9.868737864077671e-06,
      "loss": 0.2331,
      "step": 195660
    },
    {
      "epoch": 75.99,
      "learning_rate": 9.868220064724921e-06,
      "loss": 0.0023,
      "step": 195670
    },
    {
      "epoch": 75.99,
      "learning_rate": 9.867702265372169e-06,
      "loss": 0.0786,
      "step": 195680
    },
    {
      "epoch": 76.0,
      "learning_rate": 9.867184466019419e-06,
      "loss": 0.0378,
      "step": 195690
    },
    {
      "epoch": 76.0,
      "learning_rate": 9.866666666666668e-06,
      "loss": 0.141,
      "step": 195700
    },
    {
      "epoch": 76.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.34276077151298523,
      "eval_runtime": 8.1771,
      "eval_samples_per_second": 444.532,
      "eval_steps_per_second": 55.643,
      "step": 195700
    },
    {
      "epoch": 76.0,
      "learning_rate": 9.866148867313918e-06,
      "loss": 0.1148,
      "step": 195710
    },
    {
      "epoch": 76.01,
      "learning_rate": 9.865631067961166e-06,
      "loss": 0.0486,
      "step": 195720
    },
    {
      "epoch": 76.01,
      "learning_rate": 9.865113268608415e-06,
      "loss": 0.0333,
      "step": 195730
    },
    {
      "epoch": 76.02,
      "learning_rate": 9.864595469255665e-06,
      "loss": 0.1143,
      "step": 195740
    },
    {
      "epoch": 76.02,
      "learning_rate": 9.864077669902915e-06,
      "loss": 0.2066,
      "step": 195750
    },
    {
      "epoch": 76.02,
      "learning_rate": 9.863559870550163e-06,
      "loss": 0.1544,
      "step": 195760
    },
    {
      "epoch": 76.03,
      "learning_rate": 9.863042071197412e-06,
      "loss": 0.1269,
      "step": 195770
    },
    {
      "epoch": 76.03,
      "learning_rate": 9.862524271844662e-06,
      "loss": 0.0393,
      "step": 195780
    },
    {
      "epoch": 76.03,
      "learning_rate": 9.862006472491911e-06,
      "loss": 0.042,
      "step": 195790
    },
    {
      "epoch": 76.04,
      "learning_rate": 9.86148867313916e-06,
      "loss": 0.0617,
      "step": 195800
    },
    {
      "epoch": 76.04,
      "learning_rate": 9.860970873786409e-06,
      "loss": 0.0391,
      "step": 195810
    },
    {
      "epoch": 76.05,
      "learning_rate": 9.860453074433659e-06,
      "loss": 0.0324,
      "step": 195820
    },
    {
      "epoch": 76.05,
      "learning_rate": 9.859935275080908e-06,
      "loss": 0.0486,
      "step": 195830
    },
    {
      "epoch": 76.05,
      "learning_rate": 9.859417475728156e-06,
      "loss": 0.2899,
      "step": 195840
    },
    {
      "epoch": 76.06,
      "learning_rate": 9.858899676375406e-06,
      "loss": 0.0588,
      "step": 195850
    },
    {
      "epoch": 76.06,
      "learning_rate": 9.858381877022655e-06,
      "loss": 0.0443,
      "step": 195860
    },
    {
      "epoch": 76.07,
      "learning_rate": 9.857864077669905e-06,
      "loss": 0.0441,
      "step": 195870
    },
    {
      "epoch": 76.07,
      "learning_rate": 9.857346278317153e-06,
      "loss": 0.0779,
      "step": 195880
    },
    {
      "epoch": 76.07,
      "learning_rate": 9.856828478964402e-06,
      "loss": 0.0575,
      "step": 195890
    },
    {
      "epoch": 76.08,
      "learning_rate": 9.856310679611652e-06,
      "loss": 0.0153,
      "step": 195900
    },
    {
      "epoch": 76.08,
      "learning_rate": 9.855792880258902e-06,
      "loss": 0.0002,
      "step": 195910
    },
    {
      "epoch": 76.09,
      "learning_rate": 9.85527508090615e-06,
      "loss": 0.0722,
      "step": 195920
    },
    {
      "epoch": 76.09,
      "learning_rate": 9.8547572815534e-06,
      "loss": 0.0874,
      "step": 195930
    },
    {
      "epoch": 76.09,
      "learning_rate": 9.854239482200649e-06,
      "loss": 0.0715,
      "step": 195940
    },
    {
      "epoch": 76.1,
      "learning_rate": 9.853721682847898e-06,
      "loss": 0.0063,
      "step": 195950
    },
    {
      "epoch": 76.1,
      "learning_rate": 9.853203883495146e-06,
      "loss": 0.0321,
      "step": 195960
    },
    {
      "epoch": 76.1,
      "learning_rate": 9.852686084142396e-06,
      "loss": 0.1362,
      "step": 195970
    },
    {
      "epoch": 76.11,
      "learning_rate": 9.852168284789646e-06,
      "loss": 0.0636,
      "step": 195980
    },
    {
      "epoch": 76.11,
      "learning_rate": 9.851650485436894e-06,
      "loss": 0.0161,
      "step": 195990
    },
    {
      "epoch": 76.12,
      "learning_rate": 9.851132686084143e-06,
      "loss": 0.0111,
      "step": 196000
    },
    {
      "epoch": 76.12,
      "learning_rate": 9.850614886731393e-06,
      "loss": 0.0551,
      "step": 196010
    },
    {
      "epoch": 76.12,
      "learning_rate": 9.850097087378642e-06,
      "loss": 0.1109,
      "step": 196020
    },
    {
      "epoch": 76.13,
      "learning_rate": 9.84957928802589e-06,
      "loss": 0.142,
      "step": 196030
    },
    {
      "epoch": 76.13,
      "learning_rate": 9.84906148867314e-06,
      "loss": 0.0684,
      "step": 196040
    },
    {
      "epoch": 76.14,
      "learning_rate": 9.84854368932039e-06,
      "loss": 0.059,
      "step": 196050
    },
    {
      "epoch": 76.14,
      "learning_rate": 9.84802588996764e-06,
      "loss": 0.0179,
      "step": 196060
    },
    {
      "epoch": 76.14,
      "learning_rate": 9.847508090614887e-06,
      "loss": 0.0562,
      "step": 196070
    },
    {
      "epoch": 76.15,
      "learning_rate": 9.846990291262137e-06,
      "loss": 0.0885,
      "step": 196080
    },
    {
      "epoch": 76.15,
      "learning_rate": 9.846472491909386e-06,
      "loss": 0.051,
      "step": 196090
    },
    {
      "epoch": 76.16,
      "learning_rate": 9.845954692556636e-06,
      "loss": 0.0215,
      "step": 196100
    },
    {
      "epoch": 76.16,
      "learning_rate": 9.845436893203884e-06,
      "loss": 0.0078,
      "step": 196110
    },
    {
      "epoch": 76.16,
      "learning_rate": 9.844919093851134e-06,
      "loss": 0.0525,
      "step": 196120
    },
    {
      "epoch": 76.17,
      "learning_rate": 9.844401294498383e-06,
      "loss": 0.022,
      "step": 196130
    },
    {
      "epoch": 76.17,
      "learning_rate": 9.843883495145631e-06,
      "loss": 0.0417,
      "step": 196140
    },
    {
      "epoch": 76.17,
      "learning_rate": 9.84336569579288e-06,
      "loss": 0.0425,
      "step": 196150
    },
    {
      "epoch": 76.18,
      "learning_rate": 9.84284789644013e-06,
      "loss": 0.1461,
      "step": 196160
    },
    {
      "epoch": 76.18,
      "learning_rate": 9.84233009708738e-06,
      "loss": 0.0154,
      "step": 196170
    },
    {
      "epoch": 76.19,
      "learning_rate": 9.841812297734628e-06,
      "loss": 0.1002,
      "step": 196180
    },
    {
      "epoch": 76.19,
      "learning_rate": 9.841294498381877e-06,
      "loss": 0.0003,
      "step": 196190
    },
    {
      "epoch": 76.19,
      "learning_rate": 9.840776699029127e-06,
      "loss": 0.1564,
      "step": 196200
    },
    {
      "epoch": 76.2,
      "learning_rate": 9.840258899676377e-06,
      "loss": 0.0471,
      "step": 196210
    },
    {
      "epoch": 76.2,
      "learning_rate": 9.839741100323625e-06,
      "loss": 0.001,
      "step": 196220
    },
    {
      "epoch": 76.21,
      "learning_rate": 9.839223300970874e-06,
      "loss": 0.0002,
      "step": 196230
    },
    {
      "epoch": 76.21,
      "learning_rate": 9.838705501618124e-06,
      "loss": 0.0304,
      "step": 196240
    },
    {
      "epoch": 76.21,
      "learning_rate": 9.838187702265373e-06,
      "loss": 0.2077,
      "step": 196250
    },
    {
      "epoch": 76.22,
      "learning_rate": 9.837669902912621e-06,
      "loss": 0.0408,
      "step": 196260
    },
    {
      "epoch": 76.22,
      "learning_rate": 9.837152103559871e-06,
      "loss": 0.0533,
      "step": 196270
    },
    {
      "epoch": 76.23,
      "learning_rate": 9.83663430420712e-06,
      "loss": 0.1076,
      "step": 196280
    },
    {
      "epoch": 76.23,
      "learning_rate": 9.83611650485437e-06,
      "loss": 0.1653,
      "step": 196290
    },
    {
      "epoch": 76.23,
      "learning_rate": 9.835598705501618e-06,
      "loss": 0.1596,
      "step": 196300
    },
    {
      "epoch": 76.24,
      "learning_rate": 9.835080906148868e-06,
      "loss": 0.0463,
      "step": 196310
    },
    {
      "epoch": 76.24,
      "learning_rate": 9.834563106796117e-06,
      "loss": 0.0719,
      "step": 196320
    },
    {
      "epoch": 76.24,
      "learning_rate": 9.834045307443365e-06,
      "loss": 0.0864,
      "step": 196330
    },
    {
      "epoch": 76.25,
      "learning_rate": 9.833527508090615e-06,
      "loss": 0.0579,
      "step": 196340
    },
    {
      "epoch": 76.25,
      "learning_rate": 9.833009708737865e-06,
      "loss": 0.0018,
      "step": 196350
    },
    {
      "epoch": 76.26,
      "learning_rate": 9.832491909385114e-06,
      "loss": 0.0054,
      "step": 196360
    },
    {
      "epoch": 76.26,
      "learning_rate": 9.831974110032362e-06,
      "loss": 0.0877,
      "step": 196370
    },
    {
      "epoch": 76.26,
      "learning_rate": 9.831456310679612e-06,
      "loss": 0.0626,
      "step": 196380
    },
    {
      "epoch": 76.27,
      "learning_rate": 9.830938511326861e-06,
      "loss": 0.0508,
      "step": 196390
    },
    {
      "epoch": 76.27,
      "learning_rate": 9.830420711974111e-06,
      "loss": 0.032,
      "step": 196400
    },
    {
      "epoch": 76.28,
      "learning_rate": 9.829902912621359e-06,
      "loss": 0.1934,
      "step": 196410
    },
    {
      "epoch": 76.28,
      "learning_rate": 9.829385113268609e-06,
      "loss": 0.0775,
      "step": 196420
    },
    {
      "epoch": 76.28,
      "learning_rate": 9.828867313915858e-06,
      "loss": 0.1724,
      "step": 196430
    },
    {
      "epoch": 76.29,
      "learning_rate": 9.828349514563108e-06,
      "loss": 0.0489,
      "step": 196440
    },
    {
      "epoch": 76.29,
      "learning_rate": 9.827831715210356e-06,
      "loss": 0.0401,
      "step": 196450
    },
    {
      "epoch": 76.3,
      "learning_rate": 9.827313915857605e-06,
      "loss": 0.004,
      "step": 196460
    },
    {
      "epoch": 76.3,
      "learning_rate": 9.826796116504855e-06,
      "loss": 0.1798,
      "step": 196470
    },
    {
      "epoch": 76.3,
      "learning_rate": 9.826278317152105e-06,
      "loss": 0.0612,
      "step": 196480
    },
    {
      "epoch": 76.31,
      "learning_rate": 9.825760517799352e-06,
      "loss": 0.0044,
      "step": 196490
    },
    {
      "epoch": 76.31,
      "learning_rate": 9.825242718446602e-06,
      "loss": 0.142,
      "step": 196500
    },
    {
      "epoch": 76.31,
      "learning_rate": 9.824724919093852e-06,
      "loss": 0.0878,
      "step": 196510
    },
    {
      "epoch": 76.32,
      "learning_rate": 9.824207119741101e-06,
      "loss": 0.0078,
      "step": 196520
    },
    {
      "epoch": 76.32,
      "learning_rate": 9.82368932038835e-06,
      "loss": 0.0849,
      "step": 196530
    },
    {
      "epoch": 76.33,
      "learning_rate": 9.823171521035599e-06,
      "loss": 0.0321,
      "step": 196540
    },
    {
      "epoch": 76.33,
      "learning_rate": 9.822653721682848e-06,
      "loss": 0.0326,
      "step": 196550
    },
    {
      "epoch": 76.33,
      "learning_rate": 9.822135922330098e-06,
      "loss": 0.1192,
      "step": 196560
    },
    {
      "epoch": 76.34,
      "learning_rate": 9.821618122977346e-06,
      "loss": 0.1195,
      "step": 196570
    },
    {
      "epoch": 76.34,
      "learning_rate": 9.821100323624596e-06,
      "loss": 0.0583,
      "step": 196580
    },
    {
      "epoch": 76.35,
      "learning_rate": 9.820582524271845e-06,
      "loss": 0.132,
      "step": 196590
    },
    {
      "epoch": 76.35,
      "learning_rate": 9.820064724919095e-06,
      "loss": 0.053,
      "step": 196600
    },
    {
      "epoch": 76.35,
      "learning_rate": 9.819546925566343e-06,
      "loss": 0.011,
      "step": 196610
    },
    {
      "epoch": 76.36,
      "learning_rate": 9.819029126213592e-06,
      "loss": 0.0239,
      "step": 196620
    },
    {
      "epoch": 76.36,
      "learning_rate": 9.818511326860842e-06,
      "loss": 0.0176,
      "step": 196630
    },
    {
      "epoch": 76.37,
      "learning_rate": 9.817993527508092e-06,
      "loss": 0.1626,
      "step": 196640
    },
    {
      "epoch": 76.37,
      "learning_rate": 9.81747572815534e-06,
      "loss": 0.0172,
      "step": 196650
    },
    {
      "epoch": 76.37,
      "learning_rate": 9.81695792880259e-06,
      "loss": 0.0629,
      "step": 196660
    },
    {
      "epoch": 76.38,
      "learning_rate": 9.816440129449839e-06,
      "loss": 0.1979,
      "step": 196670
    },
    {
      "epoch": 76.38,
      "learning_rate": 9.815922330097088e-06,
      "loss": 0.0274,
      "step": 196680
    },
    {
      "epoch": 76.38,
      "learning_rate": 9.815404530744336e-06,
      "loss": 0.0239,
      "step": 196690
    },
    {
      "epoch": 76.39,
      "learning_rate": 9.814886731391586e-06,
      "loss": 0.0273,
      "step": 196700
    },
    {
      "epoch": 76.39,
      "learning_rate": 9.814368932038836e-06,
      "loss": 0.0758,
      "step": 196710
    },
    {
      "epoch": 76.4,
      "learning_rate": 9.813851132686085e-06,
      "loss": 0.0033,
      "step": 196720
    },
    {
      "epoch": 76.4,
      "learning_rate": 9.813333333333333e-06,
      "loss": 0.0177,
      "step": 196730
    },
    {
      "epoch": 76.4,
      "learning_rate": 9.812815533980583e-06,
      "loss": 0.0915,
      "step": 196740
    },
    {
      "epoch": 76.41,
      "learning_rate": 9.812297734627832e-06,
      "loss": 0.0634,
      "step": 196750
    },
    {
      "epoch": 76.41,
      "learning_rate": 9.811779935275082e-06,
      "loss": 0.0346,
      "step": 196760
    },
    {
      "epoch": 76.42,
      "learning_rate": 9.811262135922332e-06,
      "loss": 0.0476,
      "step": 196770
    },
    {
      "epoch": 76.42,
      "learning_rate": 9.81074433656958e-06,
      "loss": 0.0019,
      "step": 196780
    },
    {
      "epoch": 76.42,
      "learning_rate": 9.810226537216829e-06,
      "loss": 0.0657,
      "step": 196790
    },
    {
      "epoch": 76.43,
      "learning_rate": 9.809708737864079e-06,
      "loss": 0.2223,
      "step": 196800
    },
    {
      "epoch": 76.43,
      "learning_rate": 9.809190938511328e-06,
      "loss": 0.0219,
      "step": 196810
    },
    {
      "epoch": 76.43,
      "learning_rate": 9.808673139158576e-06,
      "loss": 0.0527,
      "step": 196820
    },
    {
      "epoch": 76.44,
      "learning_rate": 9.808155339805826e-06,
      "loss": 0.0032,
      "step": 196830
    },
    {
      "epoch": 76.44,
      "learning_rate": 9.807637540453076e-06,
      "loss": 0.0976,
      "step": 196840
    },
    {
      "epoch": 76.45,
      "learning_rate": 9.807119741100325e-06,
      "loss": 0.0436,
      "step": 196850
    },
    {
      "epoch": 76.45,
      "learning_rate": 9.806601941747573e-06,
      "loss": 0.014,
      "step": 196860
    },
    {
      "epoch": 76.45,
      "learning_rate": 9.806084142394823e-06,
      "loss": 0.0096,
      "step": 196870
    },
    {
      "epoch": 76.46,
      "learning_rate": 9.805566343042072e-06,
      "loss": 0.0143,
      "step": 196880
    },
    {
      "epoch": 76.46,
      "learning_rate": 9.805048543689322e-06,
      "loss": 0.0268,
      "step": 196890
    },
    {
      "epoch": 76.47,
      "learning_rate": 9.80453074433657e-06,
      "loss": 0.0147,
      "step": 196900
    },
    {
      "epoch": 76.47,
      "learning_rate": 9.80401294498382e-06,
      "loss": 0.1641,
      "step": 196910
    },
    {
      "epoch": 76.47,
      "learning_rate": 9.803495145631069e-06,
      "loss": 0.028,
      "step": 196920
    },
    {
      "epoch": 76.48,
      "learning_rate": 9.802977346278319e-06,
      "loss": 0.0855,
      "step": 196930
    },
    {
      "epoch": 76.48,
      "learning_rate": 9.802459546925567e-06,
      "loss": 0.1244,
      "step": 196940
    },
    {
      "epoch": 76.49,
      "learning_rate": 9.801941747572816e-06,
      "loss": 0.1262,
      "step": 196950
    },
    {
      "epoch": 76.49,
      "learning_rate": 9.801423948220066e-06,
      "loss": 0.0512,
      "step": 196960
    },
    {
      "epoch": 76.49,
      "learning_rate": 9.800906148867315e-06,
      "loss": 0.0005,
      "step": 196970
    },
    {
      "epoch": 76.5,
      "learning_rate": 9.800388349514563e-06,
      "loss": 0.0638,
      "step": 196980
    },
    {
      "epoch": 76.5,
      "learning_rate": 9.799870550161813e-06,
      "loss": 0.1059,
      "step": 196990
    },
    {
      "epoch": 76.5,
      "learning_rate": 9.799352750809063e-06,
      "loss": 0.136,
      "step": 197000
    },
    {
      "epoch": 76.51,
      "learning_rate": 9.798834951456312e-06,
      "loss": 0.0762,
      "step": 197010
    },
    {
      "epoch": 76.51,
      "learning_rate": 9.79831715210356e-06,
      "loss": 0.1048,
      "step": 197020
    },
    {
      "epoch": 76.52,
      "learning_rate": 9.79779935275081e-06,
      "loss": 0.0175,
      "step": 197030
    },
    {
      "epoch": 76.52,
      "learning_rate": 9.79728155339806e-06,
      "loss": 0.3074,
      "step": 197040
    },
    {
      "epoch": 76.52,
      "learning_rate": 9.796763754045309e-06,
      "loss": 0.0898,
      "step": 197050
    },
    {
      "epoch": 76.53,
      "learning_rate": 9.796245954692557e-06,
      "loss": 0.0318,
      "step": 197060
    },
    {
      "epoch": 76.53,
      "learning_rate": 9.795728155339807e-06,
      "loss": 0.0425,
      "step": 197070
    },
    {
      "epoch": 76.54,
      "learning_rate": 9.795210355987056e-06,
      "loss": 0.0209,
      "step": 197080
    },
    {
      "epoch": 76.54,
      "learning_rate": 9.794692556634306e-06,
      "loss": 0.0093,
      "step": 197090
    },
    {
      "epoch": 76.54,
      "learning_rate": 9.794174757281554e-06,
      "loss": 0.1393,
      "step": 197100
    },
    {
      "epoch": 76.55,
      "learning_rate": 9.793656957928803e-06,
      "loss": 0.0416,
      "step": 197110
    },
    {
      "epoch": 76.55,
      "learning_rate": 9.793139158576053e-06,
      "loss": 0.0225,
      "step": 197120
    },
    {
      "epoch": 76.56,
      "learning_rate": 9.792621359223303e-06,
      "loss": 0.0459,
      "step": 197130
    },
    {
      "epoch": 76.56,
      "learning_rate": 9.79210355987055e-06,
      "loss": 0.0555,
      "step": 197140
    },
    {
      "epoch": 76.56,
      "learning_rate": 9.7915857605178e-06,
      "loss": 0.136,
      "step": 197150
    },
    {
      "epoch": 76.57,
      "learning_rate": 9.79106796116505e-06,
      "loss": 0.1385,
      "step": 197160
    },
    {
      "epoch": 76.57,
      "learning_rate": 9.7905501618123e-06,
      "loss": 0.0381,
      "step": 197170
    },
    {
      "epoch": 76.57,
      "learning_rate": 9.790032362459547e-06,
      "loss": 0.1206,
      "step": 197180
    },
    {
      "epoch": 76.58,
      "learning_rate": 9.789514563106797e-06,
      "loss": 0.1444,
      "step": 197190
    },
    {
      "epoch": 76.58,
      "learning_rate": 9.788996763754047e-06,
      "loss": 0.0153,
      "step": 197200
    },
    {
      "epoch": 76.59,
      "learning_rate": 9.788478964401296e-06,
      "loss": 0.1742,
      "step": 197210
    },
    {
      "epoch": 76.59,
      "learning_rate": 9.787961165048544e-06,
      "loss": 0.2147,
      "step": 197220
    },
    {
      "epoch": 76.59,
      "learning_rate": 9.787443365695794e-06,
      "loss": 0.1913,
      "step": 197230
    },
    {
      "epoch": 76.6,
      "learning_rate": 9.786925566343043e-06,
      "loss": 0.0598,
      "step": 197240
    },
    {
      "epoch": 76.6,
      "learning_rate": 9.786407766990293e-06,
      "loss": 0.0689,
      "step": 197250
    },
    {
      "epoch": 76.61,
      "learning_rate": 9.785889967637541e-06,
      "loss": 0.065,
      "step": 197260
    },
    {
      "epoch": 76.61,
      "learning_rate": 9.78537216828479e-06,
      "loss": 0.0171,
      "step": 197270
    },
    {
      "epoch": 76.61,
      "learning_rate": 9.78485436893204e-06,
      "loss": 0.0893,
      "step": 197280
    },
    {
      "epoch": 76.62,
      "learning_rate": 9.78433656957929e-06,
      "loss": 0.032,
      "step": 197290
    },
    {
      "epoch": 76.62,
      "learning_rate": 9.783818770226538e-06,
      "loss": 0.0953,
      "step": 197300
    },
    {
      "epoch": 76.63,
      "learning_rate": 9.783300970873787e-06,
      "loss": 0.0576,
      "step": 197310
    },
    {
      "epoch": 76.63,
      "learning_rate": 9.782783171521037e-06,
      "loss": 0.0251,
      "step": 197320
    },
    {
      "epoch": 76.63,
      "learning_rate": 9.782265372168286e-06,
      "loss": 0.07,
      "step": 197330
    },
    {
      "epoch": 76.64,
      "learning_rate": 9.781747572815536e-06,
      "loss": 0.0647,
      "step": 197340
    },
    {
      "epoch": 76.64,
      "learning_rate": 9.781229773462784e-06,
      "loss": 0.0099,
      "step": 197350
    },
    {
      "epoch": 76.64,
      "learning_rate": 9.780711974110034e-06,
      "loss": 0.0395,
      "step": 197360
    },
    {
      "epoch": 76.65,
      "learning_rate": 9.780194174757283e-06,
      "loss": 0.085,
      "step": 197370
    },
    {
      "epoch": 76.65,
      "learning_rate": 9.779676375404533e-06,
      "loss": 0.0512,
      "step": 197380
    },
    {
      "epoch": 76.66,
      "learning_rate": 9.77915857605178e-06,
      "loss": 0.0021,
      "step": 197390
    },
    {
      "epoch": 76.66,
      "learning_rate": 9.77864077669903e-06,
      "loss": 0.0927,
      "step": 197400
    },
    {
      "epoch": 76.66,
      "learning_rate": 9.77812297734628e-06,
      "loss": 0.1181,
      "step": 197410
    },
    {
      "epoch": 76.67,
      "learning_rate": 9.77760517799353e-06,
      "loss": 0.0163,
      "step": 197420
    },
    {
      "epoch": 76.67,
      "learning_rate": 9.777087378640778e-06,
      "loss": 0.0871,
      "step": 197430
    },
    {
      "epoch": 76.68,
      "learning_rate": 9.776569579288027e-06,
      "loss": 0.1003,
      "step": 197440
    },
    {
      "epoch": 76.68,
      "learning_rate": 9.776051779935277e-06,
      "loss": 0.0062,
      "step": 197450
    },
    {
      "epoch": 76.68,
      "learning_rate": 9.775533980582525e-06,
      "loss": 0.0825,
      "step": 197460
    },
    {
      "epoch": 76.69,
      "learning_rate": 9.775016181229774e-06,
      "loss": 0.1042,
      "step": 197470
    },
    {
      "epoch": 76.69,
      "learning_rate": 9.774498381877024e-06,
      "loss": 0.0282,
      "step": 197480
    },
    {
      "epoch": 76.7,
      "learning_rate": 9.773980582524274e-06,
      "loss": 0.0675,
      "step": 197490
    },
    {
      "epoch": 76.7,
      "learning_rate": 9.773462783171522e-06,
      "loss": 0.0721,
      "step": 197500
    },
    {
      "epoch": 76.7,
      "learning_rate": 9.772944983818771e-06,
      "loss": 0.0066,
      "step": 197510
    },
    {
      "epoch": 76.71,
      "learning_rate": 9.77242718446602e-06,
      "loss": 0.026,
      "step": 197520
    },
    {
      "epoch": 76.71,
      "learning_rate": 9.77190938511327e-06,
      "loss": 0.0315,
      "step": 197530
    },
    {
      "epoch": 76.71,
      "learning_rate": 9.771391585760518e-06,
      "loss": 0.0723,
      "step": 197540
    },
    {
      "epoch": 76.72,
      "learning_rate": 9.770873786407768e-06,
      "loss": 0.0289,
      "step": 197550
    },
    {
      "epoch": 76.72,
      "learning_rate": 9.770355987055018e-06,
      "loss": 0.0149,
      "step": 197560
    },
    {
      "epoch": 76.73,
      "learning_rate": 9.769838187702267e-06,
      "loss": 0.0996,
      "step": 197570
    },
    {
      "epoch": 76.73,
      "learning_rate": 9.769320388349515e-06,
      "loss": 0.0892,
      "step": 197580
    },
    {
      "epoch": 76.73,
      "learning_rate": 9.768802588996765e-06,
      "loss": 0.1379,
      "step": 197590
    },
    {
      "epoch": 76.74,
      "learning_rate": 9.768284789644014e-06,
      "loss": 0.1011,
      "step": 197600
    },
    {
      "epoch": 76.74,
      "learning_rate": 9.767766990291262e-06,
      "loss": 0.0853,
      "step": 197610
    },
    {
      "epoch": 76.75,
      "learning_rate": 9.767249190938512e-06,
      "loss": 0.0152,
      "step": 197620
    },
    {
      "epoch": 76.75,
      "learning_rate": 9.766731391585761e-06,
      "loss": 0.1629,
      "step": 197630
    },
    {
      "epoch": 76.75,
      "learning_rate": 9.766213592233011e-06,
      "loss": 0.2364,
      "step": 197640
    },
    {
      "epoch": 76.76,
      "learning_rate": 9.765695792880259e-06,
      "loss": 0.2121,
      "step": 197650
    },
    {
      "epoch": 76.76,
      "learning_rate": 9.765177993527509e-06,
      "loss": 0.0667,
      "step": 197660
    },
    {
      "epoch": 76.77,
      "learning_rate": 9.764660194174758e-06,
      "loss": 0.0715,
      "step": 197670
    },
    {
      "epoch": 76.77,
      "learning_rate": 9.764142394822008e-06,
      "loss": 0.0689,
      "step": 197680
    },
    {
      "epoch": 76.77,
      "learning_rate": 9.763624595469256e-06,
      "loss": 0.0123,
      "step": 197690
    },
    {
      "epoch": 76.78,
      "learning_rate": 9.763106796116505e-06,
      "loss": 0.1291,
      "step": 197700
    },
    {
      "epoch": 76.78,
      "learning_rate": 9.762588996763755e-06,
      "loss": 0.014,
      "step": 197710
    },
    {
      "epoch": 76.78,
      "learning_rate": 9.762071197411005e-06,
      "loss": 0.0616,
      "step": 197720
    },
    {
      "epoch": 76.79,
      "learning_rate": 9.761553398058253e-06,
      "loss": 0.0455,
      "step": 197730
    },
    {
      "epoch": 76.79,
      "learning_rate": 9.761035598705502e-06,
      "loss": 0.0376,
      "step": 197740
    },
    {
      "epoch": 76.8,
      "learning_rate": 9.760517799352752e-06,
      "loss": 0.0637,
      "step": 197750
    },
    {
      "epoch": 76.8,
      "learning_rate": 9.760000000000001e-06,
      "loss": 0.0248,
      "step": 197760
    },
    {
      "epoch": 76.8,
      "learning_rate": 9.75948220064725e-06,
      "loss": 0.0662,
      "step": 197770
    },
    {
      "epoch": 76.81,
      "learning_rate": 9.758964401294499e-06,
      "loss": 0.0676,
      "step": 197780
    },
    {
      "epoch": 76.81,
      "learning_rate": 9.758446601941749e-06,
      "loss": 0.0091,
      "step": 197790
    },
    {
      "epoch": 76.82,
      "learning_rate": 9.757928802588997e-06,
      "loss": 0.0867,
      "step": 197800
    },
    {
      "epoch": 76.82,
      "learning_rate": 9.757411003236246e-06,
      "loss": 0.087,
      "step": 197810
    },
    {
      "epoch": 76.82,
      "learning_rate": 9.756893203883496e-06,
      "loss": 0.0652,
      "step": 197820
    },
    {
      "epoch": 76.83,
      "learning_rate": 9.756375404530745e-06,
      "loss": 0.0667,
      "step": 197830
    },
    {
      "epoch": 76.83,
      "learning_rate": 9.755857605177993e-06,
      "loss": 0.1265,
      "step": 197840
    },
    {
      "epoch": 76.83,
      "learning_rate": 9.755339805825243e-06,
      "loss": 0.0323,
      "step": 197850
    },
    {
      "epoch": 76.84,
      "learning_rate": 9.754822006472493e-06,
      "loss": 0.0003,
      "step": 197860
    },
    {
      "epoch": 76.84,
      "learning_rate": 9.754304207119742e-06,
      "loss": 0.0114,
      "step": 197870
    },
    {
      "epoch": 76.85,
      "learning_rate": 9.75378640776699e-06,
      "loss": 0.093,
      "step": 197880
    },
    {
      "epoch": 76.85,
      "learning_rate": 9.75326860841424e-06,
      "loss": 0.0008,
      "step": 197890
    },
    {
      "epoch": 76.85,
      "learning_rate": 9.75275080906149e-06,
      "loss": 0.0367,
      "step": 197900
    },
    {
      "epoch": 76.86,
      "learning_rate": 9.752233009708739e-06,
      "loss": 0.0519,
      "step": 197910
    },
    {
      "epoch": 76.86,
      "learning_rate": 9.751715210355987e-06,
      "loss": 0.0791,
      "step": 197920
    },
    {
      "epoch": 76.87,
      "learning_rate": 9.751197411003236e-06,
      "loss": 0.0088,
      "step": 197930
    },
    {
      "epoch": 76.87,
      "learning_rate": 9.750679611650486e-06,
      "loss": 0.072,
      "step": 197940
    },
    {
      "epoch": 76.87,
      "learning_rate": 9.750161812297736e-06,
      "loss": 0.0008,
      "step": 197950
    },
    {
      "epoch": 76.88,
      "learning_rate": 9.749644012944984e-06,
      "loss": 0.0484,
      "step": 197960
    },
    {
      "epoch": 76.88,
      "learning_rate": 9.749126213592233e-06,
      "loss": 0.0351,
      "step": 197970
    },
    {
      "epoch": 76.89,
      "learning_rate": 9.748608414239483e-06,
      "loss": 0.0134,
      "step": 197980
    },
    {
      "epoch": 76.89,
      "learning_rate": 9.748090614886732e-06,
      "loss": 0.0455,
      "step": 197990
    },
    {
      "epoch": 76.89,
      "learning_rate": 9.74757281553398e-06,
      "loss": 0.0054,
      "step": 198000
    },
    {
      "epoch": 76.9,
      "learning_rate": 9.74705501618123e-06,
      "loss": 0.0483,
      "step": 198010
    },
    {
      "epoch": 76.9,
      "learning_rate": 9.74653721682848e-06,
      "loss": 0.0007,
      "step": 198020
    },
    {
      "epoch": 76.9,
      "learning_rate": 9.74601941747573e-06,
      "loss": 0.0483,
      "step": 198030
    },
    {
      "epoch": 76.91,
      "learning_rate": 9.745501618122977e-06,
      "loss": 0.0378,
      "step": 198040
    },
    {
      "epoch": 76.91,
      "learning_rate": 9.744983818770227e-06,
      "loss": 0.0772,
      "step": 198050
    },
    {
      "epoch": 76.92,
      "learning_rate": 9.744466019417476e-06,
      "loss": 0.0543,
      "step": 198060
    },
    {
      "epoch": 76.92,
      "learning_rate": 9.743948220064726e-06,
      "loss": 0.0244,
      "step": 198070
    },
    {
      "epoch": 76.92,
      "learning_rate": 9.743430420711974e-06,
      "loss": 0.0822,
      "step": 198080
    },
    {
      "epoch": 76.93,
      "learning_rate": 9.742912621359224e-06,
      "loss": 0.0003,
      "step": 198090
    },
    {
      "epoch": 76.93,
      "learning_rate": 9.742394822006473e-06,
      "loss": 0.0641,
      "step": 198100
    },
    {
      "epoch": 76.94,
      "learning_rate": 9.741877022653723e-06,
      "loss": 0.103,
      "step": 198110
    },
    {
      "epoch": 76.94,
      "learning_rate": 9.74135922330097e-06,
      "loss": 0.0601,
      "step": 198120
    },
    {
      "epoch": 76.94,
      "learning_rate": 9.74084142394822e-06,
      "loss": 0.1037,
      "step": 198130
    },
    {
      "epoch": 76.95,
      "learning_rate": 9.74032362459547e-06,
      "loss": 0.0691,
      "step": 198140
    },
    {
      "epoch": 76.95,
      "learning_rate": 9.73980582524272e-06,
      "loss": 0.0171,
      "step": 198150
    },
    {
      "epoch": 76.96,
      "learning_rate": 9.739288025889968e-06,
      "loss": 0.0052,
      "step": 198160
    },
    {
      "epoch": 76.96,
      "learning_rate": 9.738770226537217e-06,
      "loss": 0.0667,
      "step": 198170
    },
    {
      "epoch": 76.96,
      "learning_rate": 9.738252427184467e-06,
      "loss": 0.0743,
      "step": 198180
    },
    {
      "epoch": 76.97,
      "learning_rate": 9.737734627831716e-06,
      "loss": 0.0731,
      "step": 198190
    },
    {
      "epoch": 76.97,
      "learning_rate": 9.737216828478964e-06,
      "loss": 0.0141,
      "step": 198200
    },
    {
      "epoch": 76.97,
      "learning_rate": 9.736699029126214e-06,
      "loss": 0.0012,
      "step": 198210
    },
    {
      "epoch": 76.98,
      "learning_rate": 9.736181229773464e-06,
      "loss": 0.0026,
      "step": 198220
    },
    {
      "epoch": 76.98,
      "learning_rate": 9.735663430420713e-06,
      "loss": 0.0692,
      "step": 198230
    },
    {
      "epoch": 76.99,
      "learning_rate": 9.735145631067961e-06,
      "loss": 0.1095,
      "step": 198240
    },
    {
      "epoch": 76.99,
      "learning_rate": 9.73462783171521e-06,
      "loss": 0.1161,
      "step": 198250
    },
    {
      "epoch": 76.99,
      "learning_rate": 9.73411003236246e-06,
      "loss": 0.1324,
      "step": 198260
    },
    {
      "epoch": 77.0,
      "learning_rate": 9.73359223300971e-06,
      "loss": 0.0226,
      "step": 198270
    },
    {
      "epoch": 77.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.3594486117362976,
      "eval_runtime": 8.1373,
      "eval_samples_per_second": 446.707,
      "eval_steps_per_second": 55.915,
      "step": 198275
    },
    {
      "epoch": 77.0,
      "learning_rate": 9.733074433656958e-06,
      "loss": 0.0019,
      "step": 198280
    },
    {
      "epoch": 77.01,
      "learning_rate": 9.732556634304207e-06,
      "loss": 0.0467,
      "step": 198290
    },
    {
      "epoch": 77.01,
      "learning_rate": 9.732038834951457e-06,
      "loss": 0.0712,
      "step": 198300
    },
    {
      "epoch": 77.01,
      "learning_rate": 9.731521035598707e-06,
      "loss": 0.0859,
      "step": 198310
    },
    {
      "epoch": 77.02,
      "learning_rate": 9.731003236245955e-06,
      "loss": 0.2678,
      "step": 198320
    },
    {
      "epoch": 77.02,
      "learning_rate": 9.730485436893204e-06,
      "loss": 0.1927,
      "step": 198330
    },
    {
      "epoch": 77.03,
      "learning_rate": 9.729967637540454e-06,
      "loss": 0.0548,
      "step": 198340
    },
    {
      "epoch": 77.03,
      "learning_rate": 9.729449838187703e-06,
      "loss": 0.0046,
      "step": 198350
    },
    {
      "epoch": 77.03,
      "learning_rate": 9.728932038834951e-06,
      "loss": 0.1073,
      "step": 198360
    },
    {
      "epoch": 77.04,
      "learning_rate": 9.728414239482201e-06,
      "loss": 0.2912,
      "step": 198370
    },
    {
      "epoch": 77.04,
      "learning_rate": 9.72789644012945e-06,
      "loss": 0.0927,
      "step": 198380
    },
    {
      "epoch": 77.04,
      "learning_rate": 9.7273786407767e-06,
      "loss": 0.044,
      "step": 198390
    },
    {
      "epoch": 77.05,
      "learning_rate": 9.726860841423948e-06,
      "loss": 0.0503,
      "step": 198400
    },
    {
      "epoch": 77.05,
      "learning_rate": 9.726343042071198e-06,
      "loss": 0.1347,
      "step": 198410
    },
    {
      "epoch": 77.06,
      "learning_rate": 9.725825242718447e-06,
      "loss": 0.0156,
      "step": 198420
    },
    {
      "epoch": 77.06,
      "learning_rate": 9.725307443365697e-06,
      "loss": 0.0223,
      "step": 198430
    },
    {
      "epoch": 77.06,
      "learning_rate": 9.724789644012945e-06,
      "loss": 0.039,
      "step": 198440
    },
    {
      "epoch": 77.07,
      "learning_rate": 9.724271844660195e-06,
      "loss": 0.0259,
      "step": 198450
    },
    {
      "epoch": 77.07,
      "learning_rate": 9.723754045307444e-06,
      "loss": 0.0607,
      "step": 198460
    },
    {
      "epoch": 77.08,
      "learning_rate": 9.723236245954694e-06,
      "loss": 0.0177,
      "step": 198470
    },
    {
      "epoch": 77.08,
      "learning_rate": 9.722718446601943e-06,
      "loss": 0.0103,
      "step": 198480
    },
    {
      "epoch": 77.08,
      "learning_rate": 9.722200647249191e-06,
      "loss": 0.0412,
      "step": 198490
    },
    {
      "epoch": 77.09,
      "learning_rate": 9.721682847896441e-06,
      "loss": 0.0528,
      "step": 198500
    },
    {
      "epoch": 77.09,
      "learning_rate": 9.72116504854369e-06,
      "loss": 0.0233,
      "step": 198510
    },
    {
      "epoch": 77.1,
      "learning_rate": 9.72064724919094e-06,
      "loss": 0.0563,
      "step": 198520
    },
    {
      "epoch": 77.1,
      "learning_rate": 9.720129449838188e-06,
      "loss": 0.01,
      "step": 198530
    },
    {
      "epoch": 77.1,
      "learning_rate": 9.719611650485438e-06,
      "loss": 0.0697,
      "step": 198540
    },
    {
      "epoch": 77.11,
      "learning_rate": 9.719093851132687e-06,
      "loss": 0.0157,
      "step": 198550
    },
    {
      "epoch": 77.11,
      "learning_rate": 9.718576051779937e-06,
      "loss": 0.1156,
      "step": 198560
    },
    {
      "epoch": 77.11,
      "learning_rate": 9.718058252427185e-06,
      "loss": 0.0004,
      "step": 198570
    },
    {
      "epoch": 77.12,
      "learning_rate": 9.717540453074435e-06,
      "loss": 0.0143,
      "step": 198580
    },
    {
      "epoch": 77.12,
      "learning_rate": 9.717022653721684e-06,
      "loss": 0.2524,
      "step": 198590
    },
    {
      "epoch": 77.13,
      "learning_rate": 9.716504854368934e-06,
      "loss": 0.1034,
      "step": 198600
    },
    {
      "epoch": 77.13,
      "learning_rate": 9.715987055016182e-06,
      "loss": 0.1952,
      "step": 198610
    },
    {
      "epoch": 77.13,
      "learning_rate": 9.715469255663431e-06,
      "loss": 0.1269,
      "step": 198620
    },
    {
      "epoch": 77.14,
      "learning_rate": 9.714951456310681e-06,
      "loss": 0.001,
      "step": 198630
    },
    {
      "epoch": 77.14,
      "learning_rate": 9.71443365695793e-06,
      "loss": 0.0385,
      "step": 198640
    },
    {
      "epoch": 77.15,
      "learning_rate": 9.713915857605178e-06,
      "loss": 0.1623,
      "step": 198650
    },
    {
      "epoch": 77.15,
      "learning_rate": 9.713398058252428e-06,
      "loss": 0.0917,
      "step": 198660
    },
    {
      "epoch": 77.15,
      "learning_rate": 9.712880258899678e-06,
      "loss": 0.0223,
      "step": 198670
    },
    {
      "epoch": 77.16,
      "learning_rate": 9.712362459546927e-06,
      "loss": 0.0167,
      "step": 198680
    },
    {
      "epoch": 77.16,
      "learning_rate": 9.711844660194175e-06,
      "loss": 0.0674,
      "step": 198690
    },
    {
      "epoch": 77.17,
      "learning_rate": 9.711326860841425e-06,
      "loss": 0.1744,
      "step": 198700
    },
    {
      "epoch": 77.17,
      "learning_rate": 9.710809061488674e-06,
      "loss": 0.0611,
      "step": 198710
    },
    {
      "epoch": 77.17,
      "learning_rate": 9.710291262135924e-06,
      "loss": 0.0786,
      "step": 198720
    },
    {
      "epoch": 77.18,
      "learning_rate": 9.709773462783172e-06,
      "loss": 0.1056,
      "step": 198730
    },
    {
      "epoch": 77.18,
      "learning_rate": 9.709255663430422e-06,
      "loss": 0.1144,
      "step": 198740
    },
    {
      "epoch": 77.18,
      "learning_rate": 9.708737864077671e-06,
      "loss": 0.0326,
      "step": 198750
    },
    {
      "epoch": 77.19,
      "learning_rate": 9.708220064724921e-06,
      "loss": 0.0413,
      "step": 198760
    },
    {
      "epoch": 77.19,
      "learning_rate": 9.707702265372169e-06,
      "loss": 0.0236,
      "step": 198770
    },
    {
      "epoch": 77.2,
      "learning_rate": 9.707184466019418e-06,
      "loss": 0.0473,
      "step": 198780
    },
    {
      "epoch": 77.2,
      "learning_rate": 9.706666666666668e-06,
      "loss": 0.0739,
      "step": 198790
    },
    {
      "epoch": 77.2,
      "learning_rate": 9.706148867313918e-06,
      "loss": 0.1098,
      "step": 198800
    },
    {
      "epoch": 77.21,
      "learning_rate": 9.705631067961166e-06,
      "loss": 0.0559,
      "step": 198810
    },
    {
      "epoch": 77.21,
      "learning_rate": 9.705113268608415e-06,
      "loss": 0.071,
      "step": 198820
    },
    {
      "epoch": 77.22,
      "learning_rate": 9.704595469255665e-06,
      "loss": 0.0946,
      "step": 198830
    },
    {
      "epoch": 77.22,
      "learning_rate": 9.704077669902914e-06,
      "loss": 0.0001,
      "step": 198840
    },
    {
      "epoch": 77.22,
      "learning_rate": 9.703559870550162e-06,
      "loss": 0.0307,
      "step": 198850
    },
    {
      "epoch": 77.23,
      "learning_rate": 9.703042071197412e-06,
      "loss": 0.0551,
      "step": 198860
    },
    {
      "epoch": 77.23,
      "learning_rate": 9.702524271844662e-06,
      "loss": 0.1406,
      "step": 198870
    },
    {
      "epoch": 77.23,
      "learning_rate": 9.702006472491911e-06,
      "loss": 0.0295,
      "step": 198880
    },
    {
      "epoch": 77.24,
      "learning_rate": 9.701488673139159e-06,
      "loss": 0.1565,
      "step": 198890
    },
    {
      "epoch": 77.24,
      "learning_rate": 9.700970873786409e-06,
      "loss": 0.0718,
      "step": 198900
    },
    {
      "epoch": 77.25,
      "learning_rate": 9.700453074433658e-06,
      "loss": 0.1579,
      "step": 198910
    },
    {
      "epoch": 77.25,
      "learning_rate": 9.699935275080908e-06,
      "loss": 0.0779,
      "step": 198920
    },
    {
      "epoch": 77.25,
      "learning_rate": 9.699417475728156e-06,
      "loss": 0.0082,
      "step": 198930
    },
    {
      "epoch": 77.26,
      "learning_rate": 9.698899676375406e-06,
      "loss": 0.0255,
      "step": 198940
    },
    {
      "epoch": 77.26,
      "learning_rate": 9.698381877022655e-06,
      "loss": 0.0658,
      "step": 198950
    },
    {
      "epoch": 77.27,
      "learning_rate": 9.697864077669905e-06,
      "loss": 0.0192,
      "step": 198960
    },
    {
      "epoch": 77.27,
      "learning_rate": 9.697346278317153e-06,
      "loss": 0.0505,
      "step": 198970
    },
    {
      "epoch": 77.27,
      "learning_rate": 9.696828478964402e-06,
      "loss": 0.0876,
      "step": 198980
    },
    {
      "epoch": 77.28,
      "learning_rate": 9.696310679611652e-06,
      "loss": 0.066,
      "step": 198990
    },
    {
      "epoch": 77.28,
      "learning_rate": 9.695792880258902e-06,
      "loss": 0.0565,
      "step": 199000
    },
    {
      "epoch": 77.29,
      "learning_rate": 9.69527508090615e-06,
      "loss": 0.0728,
      "step": 199010
    },
    {
      "epoch": 77.29,
      "learning_rate": 9.694757281553399e-06,
      "loss": 0.0571,
      "step": 199020
    },
    {
      "epoch": 77.29,
      "learning_rate": 9.694239482200649e-06,
      "loss": 0.0979,
      "step": 199030
    },
    {
      "epoch": 77.3,
      "learning_rate": 9.693721682847898e-06,
      "loss": 0.0189,
      "step": 199040
    },
    {
      "epoch": 77.3,
      "learning_rate": 9.693203883495146e-06,
      "loss": 0.1149,
      "step": 199050
    },
    {
      "epoch": 77.3,
      "learning_rate": 9.692686084142396e-06,
      "loss": 0.0487,
      "step": 199060
    },
    {
      "epoch": 77.31,
      "learning_rate": 9.692168284789645e-06,
      "loss": 0.0173,
      "step": 199070
    },
    {
      "epoch": 77.31,
      "learning_rate": 9.691650485436893e-06,
      "loss": 0.0246,
      "step": 199080
    },
    {
      "epoch": 77.32,
      "learning_rate": 9.691132686084143e-06,
      "loss": 0.0014,
      "step": 199090
    },
    {
      "epoch": 77.32,
      "learning_rate": 9.690614886731393e-06,
      "loss": 0.0741,
      "step": 199100
    },
    {
      "epoch": 77.32,
      "learning_rate": 9.690097087378642e-06,
      "loss": 0.0707,
      "step": 199110
    },
    {
      "epoch": 77.33,
      "learning_rate": 9.68957928802589e-06,
      "loss": 0.1339,
      "step": 199120
    },
    {
      "epoch": 77.33,
      "learning_rate": 9.68906148867314e-06,
      "loss": 0.14,
      "step": 199130
    },
    {
      "epoch": 77.34,
      "learning_rate": 9.68854368932039e-06,
      "loss": 0.0592,
      "step": 199140
    },
    {
      "epoch": 77.34,
      "learning_rate": 9.688025889967639e-06,
      "loss": 0.0619,
      "step": 199150
    },
    {
      "epoch": 77.34,
      "learning_rate": 9.687508090614887e-06,
      "loss": 0.0477,
      "step": 199160
    },
    {
      "epoch": 77.35,
      "learning_rate": 9.686990291262137e-06,
      "loss": 0.0106,
      "step": 199170
    },
    {
      "epoch": 77.35,
      "learning_rate": 9.686472491909386e-06,
      "loss": 0.0233,
      "step": 199180
    },
    {
      "epoch": 77.36,
      "learning_rate": 9.685954692556636e-06,
      "loss": 0.0443,
      "step": 199190
    },
    {
      "epoch": 77.36,
      "learning_rate": 9.685436893203884e-06,
      "loss": 0.001,
      "step": 199200
    },
    {
      "epoch": 77.36,
      "learning_rate": 9.684919093851133e-06,
      "loss": 0.1172,
      "step": 199210
    },
    {
      "epoch": 77.37,
      "learning_rate": 9.684401294498383e-06,
      "loss": 0.0075,
      "step": 199220
    },
    {
      "epoch": 77.37,
      "learning_rate": 9.683883495145633e-06,
      "loss": 0.0785,
      "step": 199230
    },
    {
      "epoch": 77.37,
      "learning_rate": 9.68336569579288e-06,
      "loss": 0.0919,
      "step": 199240
    },
    {
      "epoch": 77.38,
      "learning_rate": 9.68284789644013e-06,
      "loss": 0.0428,
      "step": 199250
    },
    {
      "epoch": 77.38,
      "learning_rate": 9.68233009708738e-06,
      "loss": 0.1062,
      "step": 199260
    },
    {
      "epoch": 77.39,
      "learning_rate": 9.681812297734628e-06,
      "loss": 0.1699,
      "step": 199270
    },
    {
      "epoch": 77.39,
      "learning_rate": 9.681294498381877e-06,
      "loss": 0.06,
      "step": 199280
    },
    {
      "epoch": 77.39,
      "learning_rate": 9.680776699029127e-06,
      "loss": 0.1057,
      "step": 199290
    },
    {
      "epoch": 77.4,
      "learning_rate": 9.680258899676377e-06,
      "loss": 0.117,
      "step": 199300
    },
    {
      "epoch": 77.4,
      "learning_rate": 9.679741100323624e-06,
      "loss": 0.0008,
      "step": 199310
    },
    {
      "epoch": 77.41,
      "learning_rate": 9.679223300970874e-06,
      "loss": 0.0617,
      "step": 199320
    },
    {
      "epoch": 77.41,
      "learning_rate": 9.678705501618124e-06,
      "loss": 0.125,
      "step": 199330
    },
    {
      "epoch": 77.41,
      "learning_rate": 9.678187702265373e-06,
      "loss": 0.0043,
      "step": 199340
    },
    {
      "epoch": 77.42,
      "learning_rate": 9.677669902912621e-06,
      "loss": 0.0152,
      "step": 199350
    },
    {
      "epoch": 77.42,
      "learning_rate": 9.677152103559871e-06,
      "loss": 0.013,
      "step": 199360
    },
    {
      "epoch": 77.43,
      "learning_rate": 9.67663430420712e-06,
      "loss": 0.0389,
      "step": 199370
    },
    {
      "epoch": 77.43,
      "learning_rate": 9.67611650485437e-06,
      "loss": 0.0626,
      "step": 199380
    },
    {
      "epoch": 77.43,
      "learning_rate": 9.675598705501618e-06,
      "loss": 0.048,
      "step": 199390
    },
    {
      "epoch": 77.44,
      "learning_rate": 9.675080906148868e-06,
      "loss": 0.0734,
      "step": 199400
    },
    {
      "epoch": 77.44,
      "learning_rate": 9.674563106796117e-06,
      "loss": 0.0831,
      "step": 199410
    },
    {
      "epoch": 77.44,
      "learning_rate": 9.674045307443365e-06,
      "loss": 0.027,
      "step": 199420
    },
    {
      "epoch": 77.45,
      "learning_rate": 9.673527508090615e-06,
      "loss": 0.0654,
      "step": 199430
    },
    {
      "epoch": 77.45,
      "learning_rate": 9.673009708737864e-06,
      "loss": 0.0079,
      "step": 199440
    },
    {
      "epoch": 77.46,
      "learning_rate": 9.672491909385114e-06,
      "loss": 0.0019,
      "step": 199450
    },
    {
      "epoch": 77.46,
      "learning_rate": 9.671974110032362e-06,
      "loss": 0.0299,
      "step": 199460
    },
    {
      "epoch": 77.46,
      "learning_rate": 9.671456310679612e-06,
      "loss": 0.0796,
      "step": 199470
    },
    {
      "epoch": 77.47,
      "learning_rate": 9.670938511326861e-06,
      "loss": 0.0794,
      "step": 199480
    },
    {
      "epoch": 77.47,
      "learning_rate": 9.67042071197411e-06,
      "loss": 0.0915,
      "step": 199490
    },
    {
      "epoch": 77.48,
      "learning_rate": 9.669902912621359e-06,
      "loss": 0.1295,
      "step": 199500
    },
    {
      "epoch": 77.48,
      "learning_rate": 9.669385113268608e-06,
      "loss": 0.094,
      "step": 199510
    },
    {
      "epoch": 77.48,
      "learning_rate": 9.668867313915858e-06,
      "loss": 0.0131,
      "step": 199520
    },
    {
      "epoch": 77.49,
      "learning_rate": 9.668349514563108e-06,
      "loss": 0.1824,
      "step": 199530
    },
    {
      "epoch": 77.49,
      "learning_rate": 9.667831715210356e-06,
      "loss": 0.1163,
      "step": 199540
    },
    {
      "epoch": 77.5,
      "learning_rate": 9.667313915857605e-06,
      "loss": 0.0915,
      "step": 199550
    },
    {
      "epoch": 77.5,
      "learning_rate": 9.666796116504855e-06,
      "loss": 0.0969,
      "step": 199560
    },
    {
      "epoch": 77.5,
      "learning_rate": 9.666278317152104e-06,
      "loss": 0.0121,
      "step": 199570
    },
    {
      "epoch": 77.51,
      "learning_rate": 9.665760517799354e-06,
      "loss": 0.0083,
      "step": 199580
    },
    {
      "epoch": 77.51,
      "learning_rate": 9.665242718446602e-06,
      "loss": 0.0937,
      "step": 199590
    },
    {
      "epoch": 77.51,
      "learning_rate": 9.664724919093852e-06,
      "loss": 0.1614,
      "step": 199600
    },
    {
      "epoch": 77.52,
      "learning_rate": 9.664207119741101e-06,
      "loss": 0.0956,
      "step": 199610
    },
    {
      "epoch": 77.52,
      "learning_rate": 9.66368932038835e-06,
      "loss": 0.0034,
      "step": 199620
    },
    {
      "epoch": 77.53,
      "learning_rate": 9.663171521035599e-06,
      "loss": 0.0405,
      "step": 199630
    },
    {
      "epoch": 77.53,
      "learning_rate": 9.662653721682848e-06,
      "loss": 0.0703,
      "step": 199640
    },
    {
      "epoch": 77.53,
      "learning_rate": 9.662135922330098e-06,
      "loss": 0.029,
      "step": 199650
    },
    {
      "epoch": 77.54,
      "learning_rate": 9.661618122977348e-06,
      "loss": 0.0034,
      "step": 199660
    },
    {
      "epoch": 77.54,
      "learning_rate": 9.661100323624595e-06,
      "loss": 0.0246,
      "step": 199670
    },
    {
      "epoch": 77.55,
      "learning_rate": 9.660582524271845e-06,
      "loss": 0.0008,
      "step": 199680
    },
    {
      "epoch": 77.55,
      "learning_rate": 9.660064724919095e-06,
      "loss": 0.1044,
      "step": 199690
    },
    {
      "epoch": 77.55,
      "learning_rate": 9.659546925566344e-06,
      "loss": 0.0463,
      "step": 199700
    },
    {
      "epoch": 77.56,
      "learning_rate": 9.659029126213592e-06,
      "loss": 0.0096,
      "step": 199710
    },
    {
      "epoch": 77.56,
      "learning_rate": 9.658511326860842e-06,
      "loss": 0.045,
      "step": 199720
    },
    {
      "epoch": 77.57,
      "learning_rate": 9.657993527508091e-06,
      "loss": 0.0274,
      "step": 199730
    },
    {
      "epoch": 77.57,
      "learning_rate": 9.657475728155341e-06,
      "loss": 0.0695,
      "step": 199740
    },
    {
      "epoch": 77.57,
      "learning_rate": 9.656957928802589e-06,
      "loss": 0.1581,
      "step": 199750
    },
    {
      "epoch": 77.58,
      "learning_rate": 9.656440129449839e-06,
      "loss": 0.1469,
      "step": 199760
    },
    {
      "epoch": 77.58,
      "learning_rate": 9.655922330097088e-06,
      "loss": 0.0542,
      "step": 199770
    },
    {
      "epoch": 77.58,
      "learning_rate": 9.655404530744338e-06,
      "loss": 0.0658,
      "step": 199780
    },
    {
      "epoch": 77.59,
      "learning_rate": 9.654886731391586e-06,
      "loss": 0.0874,
      "step": 199790
    },
    {
      "epoch": 77.59,
      "learning_rate": 9.654368932038835e-06,
      "loss": 0.0307,
      "step": 199800
    },
    {
      "epoch": 77.6,
      "learning_rate": 9.653851132686085e-06,
      "loss": 0.0722,
      "step": 199810
    },
    {
      "epoch": 77.6,
      "learning_rate": 9.653333333333335e-06,
      "loss": 0.0349,
      "step": 199820
    },
    {
      "epoch": 77.6,
      "learning_rate": 9.652815533980583e-06,
      "loss": 0.0819,
      "step": 199830
    },
    {
      "epoch": 77.61,
      "learning_rate": 9.652297734627832e-06,
      "loss": 0.0938,
      "step": 199840
    },
    {
      "epoch": 77.61,
      "learning_rate": 9.651779935275082e-06,
      "loss": 0.0473,
      "step": 199850
    },
    {
      "epoch": 77.62,
      "learning_rate": 9.651262135922331e-06,
      "loss": 0.0405,
      "step": 199860
    },
    {
      "epoch": 77.62,
      "learning_rate": 9.65074433656958e-06,
      "loss": 0.0243,
      "step": 199870
    },
    {
      "epoch": 77.62,
      "learning_rate": 9.650226537216829e-06,
      "loss": 0.1544,
      "step": 199880
    },
    {
      "epoch": 77.63,
      "learning_rate": 9.649708737864079e-06,
      "loss": 0.0863,
      "step": 199890
    },
    {
      "epoch": 77.63,
      "learning_rate": 9.649190938511328e-06,
      "loss": 0.027,
      "step": 199900
    },
    {
      "epoch": 77.63,
      "learning_rate": 9.648673139158576e-06,
      "loss": 0.0396,
      "step": 199910
    },
    {
      "epoch": 77.64,
      "learning_rate": 9.648155339805826e-06,
      "loss": 0.0014,
      "step": 199920
    },
    {
      "epoch": 77.64,
      "learning_rate": 9.647637540453075e-06,
      "loss": 0.0248,
      "step": 199930
    },
    {
      "epoch": 77.65,
      "learning_rate": 9.647119741100325e-06,
      "loss": 0.1174,
      "step": 199940
    },
    {
      "epoch": 77.65,
      "learning_rate": 9.646601941747573e-06,
      "loss": 0.0102,
      "step": 199950
    },
    {
      "epoch": 77.65,
      "learning_rate": 9.646084142394823e-06,
      "loss": 0.1626,
      "step": 199960
    },
    {
      "epoch": 77.66,
      "learning_rate": 9.645566343042072e-06,
      "loss": 0.0371,
      "step": 199970
    },
    {
      "epoch": 77.66,
      "learning_rate": 9.645048543689322e-06,
      "loss": 0.1089,
      "step": 199980
    },
    {
      "epoch": 77.67,
      "learning_rate": 9.64453074433657e-06,
      "loss": 0.01,
      "step": 199990
    },
    {
      "epoch": 77.67,
      "learning_rate": 9.64401294498382e-06,
      "loss": 0.0995,
      "step": 200000
    },
    {
      "epoch": 77.67,
      "learning_rate": 9.643495145631069e-06,
      "loss": 0.0488,
      "step": 200010
    },
    {
      "epoch": 77.68,
      "learning_rate": 9.642977346278319e-06,
      "loss": 0.0027,
      "step": 200020
    },
    {
      "epoch": 77.68,
      "learning_rate": 9.642459546925566e-06,
      "loss": 0.0096,
      "step": 200030
    },
    {
      "epoch": 77.69,
      "learning_rate": 9.641941747572816e-06,
      "loss": 0.0822,
      "step": 200040
    },
    {
      "epoch": 77.69,
      "learning_rate": 9.641423948220066e-06,
      "loss": 0.0003,
      "step": 200050
    },
    {
      "epoch": 77.69,
      "learning_rate": 9.640906148867315e-06,
      "loss": 0.2158,
      "step": 200060
    },
    {
      "epoch": 77.7,
      "learning_rate": 9.640388349514563e-06,
      "loss": 0.0653,
      "step": 200070
    },
    {
      "epoch": 77.7,
      "learning_rate": 9.639870550161813e-06,
      "loss": 0.0304,
      "step": 200080
    },
    {
      "epoch": 77.7,
      "learning_rate": 9.639352750809062e-06,
      "loss": 0.0468,
      "step": 200090
    },
    {
      "epoch": 77.71,
      "learning_rate": 9.638834951456312e-06,
      "loss": 0.2144,
      "step": 200100
    },
    {
      "epoch": 77.71,
      "learning_rate": 9.63831715210356e-06,
      "loss": 0.0135,
      "step": 200110
    },
    {
      "epoch": 77.72,
      "learning_rate": 9.63779935275081e-06,
      "loss": 0.0188,
      "step": 200120
    },
    {
      "epoch": 77.72,
      "learning_rate": 9.63728155339806e-06,
      "loss": 0.0096,
      "step": 200130
    },
    {
      "epoch": 77.72,
      "learning_rate": 9.636763754045309e-06,
      "loss": 0.0403,
      "step": 200140
    },
    {
      "epoch": 77.73,
      "learning_rate": 9.636245954692558e-06,
      "loss": 0.1316,
      "step": 200150
    },
    {
      "epoch": 77.73,
      "learning_rate": 9.635728155339806e-06,
      "loss": 0.0192,
      "step": 200160
    },
    {
      "epoch": 77.74,
      "learning_rate": 9.635210355987056e-06,
      "loss": 0.0463,
      "step": 200170
    },
    {
      "epoch": 77.74,
      "learning_rate": 9.634692556634306e-06,
      "loss": 0.068,
      "step": 200180
    },
    {
      "epoch": 77.74,
      "learning_rate": 9.634174757281555e-06,
      "loss": 0.0271,
      "step": 200190
    },
    {
      "epoch": 77.75,
      "learning_rate": 9.633656957928803e-06,
      "loss": 0.0721,
      "step": 200200
    },
    {
      "epoch": 77.75,
      "learning_rate": 9.633139158576053e-06,
      "loss": 0.0834,
      "step": 200210
    },
    {
      "epoch": 77.76,
      "learning_rate": 9.632621359223302e-06,
      "loss": 0.0998,
      "step": 200220
    },
    {
      "epoch": 77.76,
      "learning_rate": 9.632103559870552e-06,
      "loss": 0.0534,
      "step": 200230
    },
    {
      "epoch": 77.76,
      "learning_rate": 9.6315857605178e-06,
      "loss": 0.0552,
      "step": 200240
    },
    {
      "epoch": 77.77,
      "learning_rate": 9.63106796116505e-06,
      "loss": 0.0207,
      "step": 200250
    },
    {
      "epoch": 77.77,
      "learning_rate": 9.6305501618123e-06,
      "loss": 0.0062,
      "step": 200260
    },
    {
      "epoch": 77.77,
      "learning_rate": 9.630032362459549e-06,
      "loss": 0.131,
      "step": 200270
    },
    {
      "epoch": 77.78,
      "learning_rate": 9.629514563106797e-06,
      "loss": 0.1376,
      "step": 200280
    },
    {
      "epoch": 77.78,
      "learning_rate": 9.628996763754046e-06,
      "loss": 0.0591,
      "step": 200290
    },
    {
      "epoch": 77.79,
      "learning_rate": 9.628478964401296e-06,
      "loss": 0.0789,
      "step": 200300
    },
    {
      "epoch": 77.79,
      "learning_rate": 9.627961165048546e-06,
      "loss": 0.0456,
      "step": 200310
    },
    {
      "epoch": 77.79,
      "learning_rate": 9.627443365695794e-06,
      "loss": 0.0428,
      "step": 200320
    },
    {
      "epoch": 77.8,
      "learning_rate": 9.626925566343043e-06,
      "loss": 0.0025,
      "step": 200330
    },
    {
      "epoch": 77.8,
      "learning_rate": 9.626407766990293e-06,
      "loss": 0.0831,
      "step": 200340
    },
    {
      "epoch": 77.81,
      "learning_rate": 9.625889967637542e-06,
      "loss": 0.0165,
      "step": 200350
    },
    {
      "epoch": 77.81,
      "learning_rate": 9.62537216828479e-06,
      "loss": 0.0713,
      "step": 200360
    },
    {
      "epoch": 77.81,
      "learning_rate": 9.62485436893204e-06,
      "loss": 0.1156,
      "step": 200370
    },
    {
      "epoch": 77.82,
      "learning_rate": 9.62433656957929e-06,
      "loss": 0.0204,
      "step": 200380
    },
    {
      "epoch": 77.82,
      "learning_rate": 9.62381877022654e-06,
      "loss": 0.0338,
      "step": 200390
    },
    {
      "epoch": 77.83,
      "learning_rate": 9.623300970873787e-06,
      "loss": 0.09,
      "step": 200400
    },
    {
      "epoch": 77.83,
      "learning_rate": 9.622783171521037e-06,
      "loss": 0.059,
      "step": 200410
    },
    {
      "epoch": 77.83,
      "learning_rate": 9.622265372168286e-06,
      "loss": 0.0646,
      "step": 200420
    },
    {
      "epoch": 77.84,
      "learning_rate": 9.621747572815536e-06,
      "loss": 0.0042,
      "step": 200430
    },
    {
      "epoch": 77.84,
      "learning_rate": 9.621229773462784e-06,
      "loss": 0.0767,
      "step": 200440
    },
    {
      "epoch": 77.84,
      "learning_rate": 9.620711974110033e-06,
      "loss": 0.1534,
      "step": 200450
    },
    {
      "epoch": 77.85,
      "learning_rate": 9.620194174757283e-06,
      "loss": 0.0959,
      "step": 200460
    },
    {
      "epoch": 77.85,
      "learning_rate": 9.619676375404533e-06,
      "loss": 0.1026,
      "step": 200470
    },
    {
      "epoch": 77.86,
      "learning_rate": 9.61915857605178e-06,
      "loss": 0.1404,
      "step": 200480
    },
    {
      "epoch": 77.86,
      "learning_rate": 9.61864077669903e-06,
      "loss": 0.1637,
      "step": 200490
    },
    {
      "epoch": 77.86,
      "learning_rate": 9.61812297734628e-06,
      "loss": 0.0012,
      "step": 200500
    },
    {
      "epoch": 77.87,
      "learning_rate": 9.61760517799353e-06,
      "loss": 0.0088,
      "step": 200510
    },
    {
      "epoch": 77.87,
      "learning_rate": 9.617087378640777e-06,
      "loss": 0.1094,
      "step": 200520
    },
    {
      "epoch": 77.88,
      "learning_rate": 9.616569579288027e-06,
      "loss": 0.0181,
      "step": 200530
    },
    {
      "epoch": 77.88,
      "learning_rate": 9.616051779935277e-06,
      "loss": 0.0517,
      "step": 200540
    },
    {
      "epoch": 77.88,
      "learning_rate": 9.615533980582525e-06,
      "loss": 0.0264,
      "step": 200550
    },
    {
      "epoch": 77.89,
      "learning_rate": 9.615016181229774e-06,
      "loss": 0.0147,
      "step": 200560
    },
    {
      "epoch": 77.89,
      "learning_rate": 9.614498381877024e-06,
      "loss": 0.0444,
      "step": 200570
    },
    {
      "epoch": 77.9,
      "learning_rate": 9.613980582524273e-06,
      "loss": 0.0784,
      "step": 200580
    },
    {
      "epoch": 77.9,
      "learning_rate": 9.613462783171521e-06,
      "loss": 0.02,
      "step": 200590
    },
    {
      "epoch": 77.9,
      "learning_rate": 9.612944983818771e-06,
      "loss": 0.0883,
      "step": 200600
    },
    {
      "epoch": 77.91,
      "learning_rate": 9.61242718446602e-06,
      "loss": 0.0587,
      "step": 200610
    },
    {
      "epoch": 77.91,
      "learning_rate": 9.61190938511327e-06,
      "loss": 0.0342,
      "step": 200620
    },
    {
      "epoch": 77.91,
      "learning_rate": 9.611391585760518e-06,
      "loss": 0.1098,
      "step": 200630
    },
    {
      "epoch": 77.92,
      "learning_rate": 9.610873786407768e-06,
      "loss": 0.0721,
      "step": 200640
    },
    {
      "epoch": 77.92,
      "learning_rate": 9.610355987055017e-06,
      "loss": 0.1282,
      "step": 200650
    },
    {
      "epoch": 77.93,
      "learning_rate": 9.609838187702267e-06,
      "loss": 0.0471,
      "step": 200660
    },
    {
      "epoch": 77.93,
      "learning_rate": 9.609320388349515e-06,
      "loss": 0.0335,
      "step": 200670
    },
    {
      "epoch": 77.93,
      "learning_rate": 9.608802588996765e-06,
      "loss": 0.0836,
      "step": 200680
    },
    {
      "epoch": 77.94,
      "learning_rate": 9.608284789644014e-06,
      "loss": 0.0967,
      "step": 200690
    },
    {
      "epoch": 77.94,
      "learning_rate": 9.607766990291264e-06,
      "loss": 0.0292,
      "step": 200700
    },
    {
      "epoch": 77.95,
      "learning_rate": 9.607249190938512e-06,
      "loss": 0.0132,
      "step": 200710
    },
    {
      "epoch": 77.95,
      "learning_rate": 9.606731391585761e-06,
      "loss": 0.021,
      "step": 200720
    },
    {
      "epoch": 77.95,
      "learning_rate": 9.606213592233011e-06,
      "loss": 0.1618,
      "step": 200730
    },
    {
      "epoch": 77.96,
      "learning_rate": 9.605695792880259e-06,
      "loss": 0.0706,
      "step": 200740
    },
    {
      "epoch": 77.96,
      "learning_rate": 9.605177993527508e-06,
      "loss": 0.1346,
      "step": 200750
    },
    {
      "epoch": 77.97,
      "learning_rate": 9.604660194174758e-06,
      "loss": 0.0009,
      "step": 200760
    },
    {
      "epoch": 77.97,
      "learning_rate": 9.604142394822008e-06,
      "loss": 0.0367,
      "step": 200770
    },
    {
      "epoch": 77.97,
      "learning_rate": 9.603624595469256e-06,
      "loss": 0.1164,
      "step": 200780
    },
    {
      "epoch": 77.98,
      "learning_rate": 9.603106796116505e-06,
      "loss": 0.012,
      "step": 200790
    },
    {
      "epoch": 77.98,
      "learning_rate": 9.602588996763755e-06,
      "loss": 0.0222,
      "step": 200800
    },
    {
      "epoch": 77.98,
      "learning_rate": 9.602071197411004e-06,
      "loss": 0.0746,
      "step": 200810
    },
    {
      "epoch": 77.99,
      "learning_rate": 9.601553398058252e-06,
      "loss": 0.0677,
      "step": 200820
    },
    {
      "epoch": 77.99,
      "learning_rate": 9.601035598705502e-06,
      "loss": 0.04,
      "step": 200830
    },
    {
      "epoch": 78.0,
      "learning_rate": 9.600517799352752e-06,
      "loss": 0.0946,
      "step": 200840
    },
    {
      "epoch": 78.0,
      "learning_rate": 9.600000000000001e-06,
      "loss": 0.0055,
      "step": 200850
    },
    {
      "epoch": 78.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.36534062027931213,
      "eval_runtime": 8.2121,
      "eval_samples_per_second": 442.637,
      "eval_steps_per_second": 55.406,
      "step": 200850
    },
    {
      "epoch": 78.0,
      "learning_rate": 9.59948220064725e-06,
      "loss": 0.0779,
      "step": 200860
    },
    {
      "epoch": 78.01,
      "learning_rate": 9.598964401294499e-06,
      "loss": 0.095,
      "step": 200870
    },
    {
      "epoch": 78.01,
      "learning_rate": 9.598446601941748e-06,
      "loss": 0.107,
      "step": 200880
    },
    {
      "epoch": 78.02,
      "learning_rate": 9.597928802588996e-06,
      "loss": 0.1115,
      "step": 200890
    },
    {
      "epoch": 78.02,
      "learning_rate": 9.597411003236246e-06,
      "loss": 0.153,
      "step": 200900
    },
    {
      "epoch": 78.02,
      "learning_rate": 9.596893203883496e-06,
      "loss": 0.0514,
      "step": 200910
    },
    {
      "epoch": 78.03,
      "learning_rate": 9.596375404530745e-06,
      "loss": 0.0118,
      "step": 200920
    },
    {
      "epoch": 78.03,
      "learning_rate": 9.595857605177993e-06,
      "loss": 0.056,
      "step": 200930
    },
    {
      "epoch": 78.03,
      "learning_rate": 9.595339805825243e-06,
      "loss": 0.0336,
      "step": 200940
    },
    {
      "epoch": 78.04,
      "learning_rate": 9.594822006472492e-06,
      "loss": 0.0412,
      "step": 200950
    },
    {
      "epoch": 78.04,
      "learning_rate": 9.594304207119742e-06,
      "loss": 0.0638,
      "step": 200960
    },
    {
      "epoch": 78.05,
      "learning_rate": 9.59378640776699e-06,
      "loss": 0.0609,
      "step": 200970
    },
    {
      "epoch": 78.05,
      "learning_rate": 9.59326860841424e-06,
      "loss": 0.0672,
      "step": 200980
    },
    {
      "epoch": 78.05,
      "learning_rate": 9.592750809061489e-06,
      "loss": 0.1036,
      "step": 200990
    },
    {
      "epoch": 78.06,
      "learning_rate": 9.592233009708739e-06,
      "loss": 0.0163,
      "step": 201000
    },
    {
      "epoch": 78.06,
      "learning_rate": 9.591715210355987e-06,
      "loss": 0.0105,
      "step": 201010
    },
    {
      "epoch": 78.07,
      "learning_rate": 9.591197411003236e-06,
      "loss": 0.0137,
      "step": 201020
    },
    {
      "epoch": 78.07,
      "learning_rate": 9.590679611650486e-06,
      "loss": 0.0015,
      "step": 201030
    },
    {
      "epoch": 78.07,
      "learning_rate": 9.590161812297736e-06,
      "loss": 0.0531,
      "step": 201040
    },
    {
      "epoch": 78.08,
      "learning_rate": 9.589644012944983e-06,
      "loss": 0.1231,
      "step": 201050
    },
    {
      "epoch": 78.08,
      "learning_rate": 9.589126213592233e-06,
      "loss": 0.0237,
      "step": 201060
    },
    {
      "epoch": 78.09,
      "learning_rate": 9.588608414239483e-06,
      "loss": 0.0499,
      "step": 201070
    },
    {
      "epoch": 78.09,
      "learning_rate": 9.588090614886732e-06,
      "loss": 0.1048,
      "step": 201080
    },
    {
      "epoch": 78.09,
      "learning_rate": 9.58757281553398e-06,
      "loss": 0.1215,
      "step": 201090
    },
    {
      "epoch": 78.1,
      "learning_rate": 9.58705501618123e-06,
      "loss": 0.1036,
      "step": 201100
    },
    {
      "epoch": 78.1,
      "learning_rate": 9.58653721682848e-06,
      "loss": 0.0824,
      "step": 201110
    },
    {
      "epoch": 78.1,
      "learning_rate": 9.586019417475729e-06,
      "loss": 0.0212,
      "step": 201120
    },
    {
      "epoch": 78.11,
      "learning_rate": 9.585501618122977e-06,
      "loss": 0.0559,
      "step": 201130
    },
    {
      "epoch": 78.11,
      "learning_rate": 9.584983818770227e-06,
      "loss": 0.2121,
      "step": 201140
    },
    {
      "epoch": 78.12,
      "learning_rate": 9.584466019417476e-06,
      "loss": 0.0083,
      "step": 201150
    },
    {
      "epoch": 78.12,
      "learning_rate": 9.583948220064726e-06,
      "loss": 0.0503,
      "step": 201160
    },
    {
      "epoch": 78.12,
      "learning_rate": 9.583430420711974e-06,
      "loss": 0.051,
      "step": 201170
    },
    {
      "epoch": 78.13,
      "learning_rate": 9.582912621359223e-06,
      "loss": 0.1026,
      "step": 201180
    },
    {
      "epoch": 78.13,
      "learning_rate": 9.582394822006473e-06,
      "loss": 0.0143,
      "step": 201190
    },
    {
      "epoch": 78.14,
      "learning_rate": 9.581877022653723e-06,
      "loss": 0.0014,
      "step": 201200
    },
    {
      "epoch": 78.14,
      "learning_rate": 9.58135922330097e-06,
      "loss": 0.0295,
      "step": 201210
    },
    {
      "epoch": 78.14,
      "learning_rate": 9.58084142394822e-06,
      "loss": 0.0004,
      "step": 201220
    },
    {
      "epoch": 78.15,
      "learning_rate": 9.58032362459547e-06,
      "loss": 0.0008,
      "step": 201230
    },
    {
      "epoch": 78.15,
      "learning_rate": 9.57980582524272e-06,
      "loss": 0.0735,
      "step": 201240
    },
    {
      "epoch": 78.16,
      "learning_rate": 9.579288025889967e-06,
      "loss": 0.0074,
      "step": 201250
    },
    {
      "epoch": 78.16,
      "learning_rate": 9.578770226537217e-06,
      "loss": 0.0009,
      "step": 201260
    },
    {
      "epoch": 78.16,
      "learning_rate": 9.578252427184467e-06,
      "loss": 0.1236,
      "step": 201270
    },
    {
      "epoch": 78.17,
      "learning_rate": 9.577734627831716e-06,
      "loss": 0.0111,
      "step": 201280
    },
    {
      "epoch": 78.17,
      "learning_rate": 9.577216828478966e-06,
      "loss": 0.0358,
      "step": 201290
    },
    {
      "epoch": 78.17,
      "learning_rate": 9.576699029126214e-06,
      "loss": 0.0171,
      "step": 201300
    },
    {
      "epoch": 78.18,
      "learning_rate": 9.576181229773463e-06,
      "loss": 0.002,
      "step": 201310
    },
    {
      "epoch": 78.18,
      "learning_rate": 9.575663430420713e-06,
      "loss": 0.0341,
      "step": 201320
    },
    {
      "epoch": 78.19,
      "learning_rate": 9.575145631067963e-06,
      "loss": 0.1217,
      "step": 201330
    },
    {
      "epoch": 78.19,
      "learning_rate": 9.57462783171521e-06,
      "loss": 0.0121,
      "step": 201340
    },
    {
      "epoch": 78.19,
      "learning_rate": 9.57411003236246e-06,
      "loss": 0.0964,
      "step": 201350
    },
    {
      "epoch": 78.2,
      "learning_rate": 9.57359223300971e-06,
      "loss": 0.0011,
      "step": 201360
    },
    {
      "epoch": 78.2,
      "learning_rate": 9.57307443365696e-06,
      "loss": 0.0413,
      "step": 201370
    },
    {
      "epoch": 78.21,
      "learning_rate": 9.572556634304207e-06,
      "loss": 0.0321,
      "step": 201380
    },
    {
      "epoch": 78.21,
      "learning_rate": 9.572038834951457e-06,
      "loss": 0.0676,
      "step": 201390
    },
    {
      "epoch": 78.21,
      "learning_rate": 9.571521035598707e-06,
      "loss": 0.0256,
      "step": 201400
    },
    {
      "epoch": 78.22,
      "learning_rate": 9.571003236245956e-06,
      "loss": 0.0048,
      "step": 201410
    },
    {
      "epoch": 78.22,
      "learning_rate": 9.570485436893204e-06,
      "loss": 0.1065,
      "step": 201420
    },
    {
      "epoch": 78.23,
      "learning_rate": 9.569967637540454e-06,
      "loss": 0.0668,
      "step": 201430
    },
    {
      "epoch": 78.23,
      "learning_rate": 9.569449838187703e-06,
      "loss": 0.0308,
      "step": 201440
    },
    {
      "epoch": 78.23,
      "learning_rate": 9.568932038834953e-06,
      "loss": 0.0195,
      "step": 201450
    },
    {
      "epoch": 78.24,
      "learning_rate": 9.568414239482201e-06,
      "loss": 0.1277,
      "step": 201460
    },
    {
      "epoch": 78.24,
      "learning_rate": 9.56789644012945e-06,
      "loss": 0.0897,
      "step": 201470
    },
    {
      "epoch": 78.24,
      "learning_rate": 9.5673786407767e-06,
      "loss": 0.0873,
      "step": 201480
    },
    {
      "epoch": 78.25,
      "learning_rate": 9.56686084142395e-06,
      "loss": 0.0118,
      "step": 201490
    },
    {
      "epoch": 78.25,
      "learning_rate": 9.566343042071198e-06,
      "loss": 0.1278,
      "step": 201500
    },
    {
      "epoch": 78.26,
      "learning_rate": 9.565825242718447e-06,
      "loss": 0.1043,
      "step": 201510
    },
    {
      "epoch": 78.26,
      "learning_rate": 9.565307443365697e-06,
      "loss": 0.0449,
      "step": 201520
    },
    {
      "epoch": 78.26,
      "learning_rate": 9.564789644012946e-06,
      "loss": 0.0062,
      "step": 201530
    },
    {
      "epoch": 78.27,
      "learning_rate": 9.564271844660194e-06,
      "loss": 0.0303,
      "step": 201540
    },
    {
      "epoch": 78.27,
      "learning_rate": 9.563754045307444e-06,
      "loss": 0.0546,
      "step": 201550
    },
    {
      "epoch": 78.28,
      "learning_rate": 9.563236245954694e-06,
      "loss": 0.0175,
      "step": 201560
    },
    {
      "epoch": 78.28,
      "learning_rate": 9.562718446601943e-06,
      "loss": 0.0059,
      "step": 201570
    },
    {
      "epoch": 78.28,
      "learning_rate": 9.562200647249191e-06,
      "loss": 0.1638,
      "step": 201580
    },
    {
      "epoch": 78.29,
      "learning_rate": 9.56168284789644e-06,
      "loss": 0.0789,
      "step": 201590
    },
    {
      "epoch": 78.29,
      "learning_rate": 9.56116504854369e-06,
      "loss": 0.1419,
      "step": 201600
    },
    {
      "epoch": 78.3,
      "learning_rate": 9.56064724919094e-06,
      "loss": 0.056,
      "step": 201610
    },
    {
      "epoch": 78.3,
      "learning_rate": 9.560129449838188e-06,
      "loss": 0.0284,
      "step": 201620
    },
    {
      "epoch": 78.3,
      "learning_rate": 9.559611650485438e-06,
      "loss": 0.0095,
      "step": 201630
    },
    {
      "epoch": 78.31,
      "learning_rate": 9.559093851132687e-06,
      "loss": 0.067,
      "step": 201640
    },
    {
      "epoch": 78.31,
      "learning_rate": 9.558576051779937e-06,
      "loss": 0.1047,
      "step": 201650
    },
    {
      "epoch": 78.31,
      "learning_rate": 9.558058252427185e-06,
      "loss": 0.0016,
      "step": 201660
    },
    {
      "epoch": 78.32,
      "learning_rate": 9.557540453074434e-06,
      "loss": 0.0404,
      "step": 201670
    },
    {
      "epoch": 78.32,
      "learning_rate": 9.557022653721684e-06,
      "loss": 0.0386,
      "step": 201680
    },
    {
      "epoch": 78.33,
      "learning_rate": 9.556504854368934e-06,
      "loss": 0.0266,
      "step": 201690
    },
    {
      "epoch": 78.33,
      "learning_rate": 9.555987055016182e-06,
      "loss": 0.0586,
      "step": 201700
    },
    {
      "epoch": 78.33,
      "learning_rate": 9.555469255663431e-06,
      "loss": 0.0505,
      "step": 201710
    },
    {
      "epoch": 78.34,
      "learning_rate": 9.55495145631068e-06,
      "loss": 0.2207,
      "step": 201720
    },
    {
      "epoch": 78.34,
      "learning_rate": 9.55443365695793e-06,
      "loss": 0.021,
      "step": 201730
    },
    {
      "epoch": 78.35,
      "learning_rate": 9.553915857605178e-06,
      "loss": 0.0674,
      "step": 201740
    },
    {
      "epoch": 78.35,
      "learning_rate": 9.553398058252428e-06,
      "loss": 0.0953,
      "step": 201750
    },
    {
      "epoch": 78.35,
      "learning_rate": 9.552880258899678e-06,
      "loss": 0.0167,
      "step": 201760
    },
    {
      "epoch": 78.36,
      "learning_rate": 9.552362459546927e-06,
      "loss": 0.068,
      "step": 201770
    },
    {
      "epoch": 78.36,
      "learning_rate": 9.551844660194175e-06,
      "loss": 0.0209,
      "step": 201780
    },
    {
      "epoch": 78.37,
      "learning_rate": 9.551326860841425e-06,
      "loss": 0.0109,
      "step": 201790
    },
    {
      "epoch": 78.37,
      "learning_rate": 9.550809061488674e-06,
      "loss": 0.0478,
      "step": 201800
    },
    {
      "epoch": 78.37,
      "learning_rate": 9.550291262135924e-06,
      "loss": 0.0004,
      "step": 201810
    },
    {
      "epoch": 78.38,
      "learning_rate": 9.549773462783174e-06,
      "loss": 0.0771,
      "step": 201820
    },
    {
      "epoch": 78.38,
      "learning_rate": 9.549255663430421e-06,
      "loss": 0.1698,
      "step": 201830
    },
    {
      "epoch": 78.38,
      "learning_rate": 9.548737864077671e-06,
      "loss": 0.0782,
      "step": 201840
    },
    {
      "epoch": 78.39,
      "learning_rate": 9.54822006472492e-06,
      "loss": 0.0365,
      "step": 201850
    },
    {
      "epoch": 78.39,
      "learning_rate": 9.54770226537217e-06,
      "loss": 0.1406,
      "step": 201860
    },
    {
      "epoch": 78.4,
      "learning_rate": 9.547184466019418e-06,
      "loss": 0.1312,
      "step": 201870
    },
    {
      "epoch": 78.4,
      "learning_rate": 9.546666666666668e-06,
      "loss": 0.0169,
      "step": 201880
    },
    {
      "epoch": 78.4,
      "learning_rate": 9.546148867313917e-06,
      "loss": 0.0269,
      "step": 201890
    },
    {
      "epoch": 78.41,
      "learning_rate": 9.545631067961167e-06,
      "loss": 0.0433,
      "step": 201900
    },
    {
      "epoch": 78.41,
      "learning_rate": 9.545113268608415e-06,
      "loss": 0.3131,
      "step": 201910
    },
    {
      "epoch": 78.42,
      "learning_rate": 9.544595469255665e-06,
      "loss": 0.007,
      "step": 201920
    },
    {
      "epoch": 78.42,
      "learning_rate": 9.544077669902914e-06,
      "loss": 0.1338,
      "step": 201930
    },
    {
      "epoch": 78.42,
      "learning_rate": 9.543559870550164e-06,
      "loss": 0.0357,
      "step": 201940
    },
    {
      "epoch": 78.43,
      "learning_rate": 9.543042071197412e-06,
      "loss": 0.0385,
      "step": 201950
    },
    {
      "epoch": 78.43,
      "learning_rate": 9.542524271844661e-06,
      "loss": 0.0562,
      "step": 201960
    },
    {
      "epoch": 78.43,
      "learning_rate": 9.542006472491911e-06,
      "loss": 0.0997,
      "step": 201970
    },
    {
      "epoch": 78.44,
      "learning_rate": 9.54148867313916e-06,
      "loss": 0.088,
      "step": 201980
    },
    {
      "epoch": 78.44,
      "learning_rate": 9.540970873786409e-06,
      "loss": 0.0048,
      "step": 201990
    },
    {
      "epoch": 78.45,
      "learning_rate": 9.540453074433658e-06,
      "loss": 0.1222,
      "step": 202000
    },
    {
      "epoch": 78.45,
      "learning_rate": 9.539935275080908e-06,
      "loss": 0.0402,
      "step": 202010
    },
    {
      "epoch": 78.45,
      "learning_rate": 9.539417475728156e-06,
      "loss": 0.0481,
      "step": 202020
    },
    {
      "epoch": 78.46,
      "learning_rate": 9.538899676375405e-06,
      "loss": 0.026,
      "step": 202030
    },
    {
      "epoch": 78.46,
      "learning_rate": 9.538381877022655e-06,
      "loss": 0.02,
      "step": 202040
    },
    {
      "epoch": 78.47,
      "learning_rate": 9.537864077669905e-06,
      "loss": 0.03,
      "step": 202050
    },
    {
      "epoch": 78.47,
      "learning_rate": 9.537346278317153e-06,
      "loss": 0.1672,
      "step": 202060
    },
    {
      "epoch": 78.47,
      "learning_rate": 9.536828478964402e-06,
      "loss": 0.015,
      "step": 202070
    },
    {
      "epoch": 78.48,
      "learning_rate": 9.536310679611652e-06,
      "loss": 0.141,
      "step": 202080
    },
    {
      "epoch": 78.48,
      "learning_rate": 9.535792880258901e-06,
      "loss": 0.054,
      "step": 202090
    },
    {
      "epoch": 78.49,
      "learning_rate": 9.53527508090615e-06,
      "loss": 0.0587,
      "step": 202100
    },
    {
      "epoch": 78.49,
      "learning_rate": 9.534757281553399e-06,
      "loss": 0.1602,
      "step": 202110
    },
    {
      "epoch": 78.49,
      "learning_rate": 9.534239482200649e-06,
      "loss": 0.0943,
      "step": 202120
    },
    {
      "epoch": 78.5,
      "learning_rate": 9.533721682847898e-06,
      "loss": 0.0348,
      "step": 202130
    },
    {
      "epoch": 78.5,
      "learning_rate": 9.533203883495146e-06,
      "loss": 0.1064,
      "step": 202140
    },
    {
      "epoch": 78.5,
      "learning_rate": 9.532686084142396e-06,
      "loss": 0.0455,
      "step": 202150
    },
    {
      "epoch": 78.51,
      "learning_rate": 9.532168284789645e-06,
      "loss": 0.1528,
      "step": 202160
    },
    {
      "epoch": 78.51,
      "learning_rate": 9.531650485436893e-06,
      "loss": 0.0234,
      "step": 202170
    },
    {
      "epoch": 78.52,
      "learning_rate": 9.531132686084143e-06,
      "loss": 0.085,
      "step": 202180
    },
    {
      "epoch": 78.52,
      "learning_rate": 9.530614886731392e-06,
      "loss": 0.0751,
      "step": 202190
    },
    {
      "epoch": 78.52,
      "learning_rate": 9.530097087378642e-06,
      "loss": 0.0284,
      "step": 202200
    },
    {
      "epoch": 78.53,
      "learning_rate": 9.52957928802589e-06,
      "loss": 0.1681,
      "step": 202210
    },
    {
      "epoch": 78.53,
      "learning_rate": 9.52906148867314e-06,
      "loss": 0.0205,
      "step": 202220
    },
    {
      "epoch": 78.54,
      "learning_rate": 9.52854368932039e-06,
      "loss": 0.1132,
      "step": 202230
    },
    {
      "epoch": 78.54,
      "learning_rate": 9.528025889967639e-06,
      "loss": 0.0539,
      "step": 202240
    },
    {
      "epoch": 78.54,
      "learning_rate": 9.527508090614887e-06,
      "loss": 0.1909,
      "step": 202250
    },
    {
      "epoch": 78.55,
      "learning_rate": 9.526990291262136e-06,
      "loss": 0.0579,
      "step": 202260
    },
    {
      "epoch": 78.55,
      "learning_rate": 9.526472491909386e-06,
      "loss": 0.0678,
      "step": 202270
    },
    {
      "epoch": 78.56,
      "learning_rate": 9.525954692556636e-06,
      "loss": 0.0003,
      "step": 202280
    },
    {
      "epoch": 78.56,
      "learning_rate": 9.525436893203884e-06,
      "loss": 0.0228,
      "step": 202290
    },
    {
      "epoch": 78.56,
      "learning_rate": 9.524919093851133e-06,
      "loss": 0.0261,
      "step": 202300
    },
    {
      "epoch": 78.57,
      "learning_rate": 9.524401294498383e-06,
      "loss": 0.0643,
      "step": 202310
    },
    {
      "epoch": 78.57,
      "learning_rate": 9.523883495145632e-06,
      "loss": 0.1013,
      "step": 202320
    },
    {
      "epoch": 78.57,
      "learning_rate": 9.52336569579288e-06,
      "loss": 0.025,
      "step": 202330
    },
    {
      "epoch": 78.58,
      "learning_rate": 9.52284789644013e-06,
      "loss": 0.0003,
      "step": 202340
    },
    {
      "epoch": 78.58,
      "learning_rate": 9.52233009708738e-06,
      "loss": 0.075,
      "step": 202350
    },
    {
      "epoch": 78.59,
      "learning_rate": 9.521812297734628e-06,
      "loss": 0.0434,
      "step": 202360
    },
    {
      "epoch": 78.59,
      "learning_rate": 9.521294498381877e-06,
      "loss": 0.0556,
      "step": 202370
    },
    {
      "epoch": 78.59,
      "learning_rate": 9.520776699029127e-06,
      "loss": 0.0928,
      "step": 202380
    },
    {
      "epoch": 78.6,
      "learning_rate": 9.520258899676376e-06,
      "loss": 0.126,
      "step": 202390
    },
    {
      "epoch": 78.6,
      "learning_rate": 9.519741100323624e-06,
      "loss": 0.0267,
      "step": 202400
    },
    {
      "epoch": 78.61,
      "learning_rate": 9.519223300970874e-06,
      "loss": 0.0683,
      "step": 202410
    },
    {
      "epoch": 78.61,
      "learning_rate": 9.518705501618124e-06,
      "loss": 0.0825,
      "step": 202420
    },
    {
      "epoch": 78.61,
      "learning_rate": 9.518187702265373e-06,
      "loss": 0.0492,
      "step": 202430
    },
    {
      "epoch": 78.62,
      "learning_rate": 9.517669902912621e-06,
      "loss": 0.057,
      "step": 202440
    },
    {
      "epoch": 78.62,
      "learning_rate": 9.51715210355987e-06,
      "loss": 0.0022,
      "step": 202450
    },
    {
      "epoch": 78.63,
      "learning_rate": 9.51663430420712e-06,
      "loss": 0.0499,
      "step": 202460
    },
    {
      "epoch": 78.63,
      "learning_rate": 9.51611650485437e-06,
      "loss": 0.0014,
      "step": 202470
    },
    {
      "epoch": 78.63,
      "learning_rate": 9.515598705501618e-06,
      "loss": 0.0425,
      "step": 202480
    },
    {
      "epoch": 78.64,
      "learning_rate": 9.515080906148867e-06,
      "loss": 0.1058,
      "step": 202490
    },
    {
      "epoch": 78.64,
      "learning_rate": 9.514563106796117e-06,
      "loss": 0.0913,
      "step": 202500
    },
    {
      "epoch": 78.64,
      "learning_rate": 9.514045307443367e-06,
      "loss": 0.027,
      "step": 202510
    },
    {
      "epoch": 78.65,
      "learning_rate": 9.513527508090615e-06,
      "loss": 0.0787,
      "step": 202520
    },
    {
      "epoch": 78.65,
      "learning_rate": 9.513009708737864e-06,
      "loss": 0.1203,
      "step": 202530
    },
    {
      "epoch": 78.66,
      "learning_rate": 9.512491909385114e-06,
      "loss": 0.047,
      "step": 202540
    },
    {
      "epoch": 78.66,
      "learning_rate": 9.511974110032363e-06,
      "loss": 0.0011,
      "step": 202550
    },
    {
      "epoch": 78.66,
      "learning_rate": 9.511456310679611e-06,
      "loss": 0.0698,
      "step": 202560
    },
    {
      "epoch": 78.67,
      "learning_rate": 9.510938511326861e-06,
      "loss": 0.0651,
      "step": 202570
    },
    {
      "epoch": 78.67,
      "learning_rate": 9.51042071197411e-06,
      "loss": 0.0631,
      "step": 202580
    },
    {
      "epoch": 78.68,
      "learning_rate": 9.50990291262136e-06,
      "loss": 0.0101,
      "step": 202590
    },
    {
      "epoch": 78.68,
      "learning_rate": 9.509385113268608e-06,
      "loss": 0.0445,
      "step": 202600
    },
    {
      "epoch": 78.68,
      "learning_rate": 9.508867313915858e-06,
      "loss": 0.0536,
      "step": 202610
    },
    {
      "epoch": 78.69,
      "learning_rate": 9.508349514563107e-06,
      "loss": 0.035,
      "step": 202620
    },
    {
      "epoch": 78.69,
      "learning_rate": 9.507831715210357e-06,
      "loss": 0.0666,
      "step": 202630
    },
    {
      "epoch": 78.7,
      "learning_rate": 9.507313915857605e-06,
      "loss": 0.1132,
      "step": 202640
    },
    {
      "epoch": 78.7,
      "learning_rate": 9.506796116504855e-06,
      "loss": 0.0834,
      "step": 202650
    },
    {
      "epoch": 78.7,
      "learning_rate": 9.506278317152104e-06,
      "loss": 0.1332,
      "step": 202660
    },
    {
      "epoch": 78.71,
      "learning_rate": 9.505760517799354e-06,
      "loss": 0.0664,
      "step": 202670
    },
    {
      "epoch": 78.71,
      "learning_rate": 9.505242718446602e-06,
      "loss": 0.0582,
      "step": 202680
    },
    {
      "epoch": 78.71,
      "learning_rate": 9.504724919093851e-06,
      "loss": 0.1476,
      "step": 202690
    },
    {
      "epoch": 78.72,
      "learning_rate": 9.504207119741101e-06,
      "loss": 0.0935,
      "step": 202700
    },
    {
      "epoch": 78.72,
      "learning_rate": 9.50368932038835e-06,
      "loss": 0.1115,
      "step": 202710
    },
    {
      "epoch": 78.73,
      "learning_rate": 9.503171521035599e-06,
      "loss": 0.047,
      "step": 202720
    },
    {
      "epoch": 78.73,
      "learning_rate": 9.502653721682848e-06,
      "loss": 0.0031,
      "step": 202730
    },
    {
      "epoch": 78.73,
      "learning_rate": 9.502135922330098e-06,
      "loss": 0.0584,
      "step": 202740
    },
    {
      "epoch": 78.74,
      "learning_rate": 9.501618122977347e-06,
      "loss": 0.0966,
      "step": 202750
    },
    {
      "epoch": 78.74,
      "learning_rate": 9.501100323624595e-06,
      "loss": 0.1354,
      "step": 202760
    },
    {
      "epoch": 78.75,
      "learning_rate": 9.500582524271845e-06,
      "loss": 0.0788,
      "step": 202770
    },
    {
      "epoch": 78.75,
      "learning_rate": 9.500064724919095e-06,
      "loss": 0.0725,
      "step": 202780
    },
    {
      "epoch": 78.75,
      "learning_rate": 9.499546925566344e-06,
      "loss": 0.1238,
      "step": 202790
    },
    {
      "epoch": 78.76,
      "learning_rate": 9.499029126213592e-06,
      "loss": 0.0355,
      "step": 202800
    },
    {
      "epoch": 78.76,
      "learning_rate": 9.498511326860842e-06,
      "loss": 0.0259,
      "step": 202810
    },
    {
      "epoch": 78.77,
      "learning_rate": 9.497993527508091e-06,
      "loss": 0.0919,
      "step": 202820
    },
    {
      "epoch": 78.77,
      "learning_rate": 9.497475728155341e-06,
      "loss": 0.0402,
      "step": 202830
    },
    {
      "epoch": 78.77,
      "learning_rate": 9.496957928802589e-06,
      "loss": 0.1655,
      "step": 202840
    },
    {
      "epoch": 78.78,
      "learning_rate": 9.496440129449838e-06,
      "loss": 0.0887,
      "step": 202850
    },
    {
      "epoch": 78.78,
      "learning_rate": 9.495922330097088e-06,
      "loss": 0.1879,
      "step": 202860
    },
    {
      "epoch": 78.78,
      "learning_rate": 9.495404530744338e-06,
      "loss": 0.0686,
      "step": 202870
    },
    {
      "epoch": 78.79,
      "learning_rate": 9.494886731391586e-06,
      "loss": 0.0364,
      "step": 202880
    },
    {
      "epoch": 78.79,
      "learning_rate": 9.494368932038835e-06,
      "loss": 0.0234,
      "step": 202890
    },
    {
      "epoch": 78.8,
      "learning_rate": 9.493851132686085e-06,
      "loss": 0.1594,
      "step": 202900
    },
    {
      "epoch": 78.8,
      "learning_rate": 9.493333333333334e-06,
      "loss": 0.0551,
      "step": 202910
    },
    {
      "epoch": 78.8,
      "learning_rate": 9.492815533980582e-06,
      "loss": 0.0562,
      "step": 202920
    },
    {
      "epoch": 78.81,
      "learning_rate": 9.492297734627832e-06,
      "loss": 0.4693,
      "step": 202930
    },
    {
      "epoch": 78.81,
      "learning_rate": 9.491779935275082e-06,
      "loss": 0.0929,
      "step": 202940
    },
    {
      "epoch": 78.82,
      "learning_rate": 9.491262135922331e-06,
      "loss": 0.1009,
      "step": 202950
    },
    {
      "epoch": 78.82,
      "learning_rate": 9.490744336569581e-06,
      "loss": 0.1005,
      "step": 202960
    },
    {
      "epoch": 78.82,
      "learning_rate": 9.490226537216829e-06,
      "loss": 0.0393,
      "step": 202970
    },
    {
      "epoch": 78.83,
      "learning_rate": 9.489708737864078e-06,
      "loss": 0.0623,
      "step": 202980
    },
    {
      "epoch": 78.83,
      "learning_rate": 9.489190938511328e-06,
      "loss": 0.0271,
      "step": 202990
    },
    {
      "epoch": 78.83,
      "learning_rate": 9.488673139158578e-06,
      "loss": 0.0198,
      "step": 203000
    },
    {
      "epoch": 78.84,
      "learning_rate": 9.488155339805826e-06,
      "loss": 0.0465,
      "step": 203010
    },
    {
      "epoch": 78.84,
      "learning_rate": 9.487637540453075e-06,
      "loss": 0.0123,
      "step": 203020
    },
    {
      "epoch": 78.85,
      "learning_rate": 9.487119741100325e-06,
      "loss": 0.0394,
      "step": 203030
    },
    {
      "epoch": 78.85,
      "learning_rate": 9.486601941747574e-06,
      "loss": 0.1156,
      "step": 203040
    },
    {
      "epoch": 78.85,
      "learning_rate": 9.486084142394822e-06,
      "loss": 0.0494,
      "step": 203050
    },
    {
      "epoch": 78.86,
      "learning_rate": 9.485566343042072e-06,
      "loss": 0.2171,
      "step": 203060
    },
    {
      "epoch": 78.86,
      "learning_rate": 9.485048543689322e-06,
      "loss": 0.0489,
      "step": 203070
    },
    {
      "epoch": 78.87,
      "learning_rate": 9.484530744336571e-06,
      "loss": 0.0827,
      "step": 203080
    },
    {
      "epoch": 78.87,
      "learning_rate": 9.484012944983819e-06,
      "loss": 0.0045,
      "step": 203090
    },
    {
      "epoch": 78.87,
      "learning_rate": 9.483495145631069e-06,
      "loss": 0.0045,
      "step": 203100
    },
    {
      "epoch": 78.88,
      "learning_rate": 9.482977346278318e-06,
      "loss": 0.0674,
      "step": 203110
    },
    {
      "epoch": 78.88,
      "learning_rate": 9.482459546925568e-06,
      "loss": 0.0246,
      "step": 203120
    },
    {
      "epoch": 78.89,
      "learning_rate": 9.481941747572816e-06,
      "loss": 0.0093,
      "step": 203130
    },
    {
      "epoch": 78.89,
      "learning_rate": 9.481423948220066e-06,
      "loss": 0.09,
      "step": 203140
    },
    {
      "epoch": 78.89,
      "learning_rate": 9.480906148867315e-06,
      "loss": 0.0052,
      "step": 203150
    },
    {
      "epoch": 78.9,
      "learning_rate": 9.480388349514565e-06,
      "loss": 0.1577,
      "step": 203160
    },
    {
      "epoch": 78.9,
      "learning_rate": 9.479870550161813e-06,
      "loss": 0.1233,
      "step": 203170
    },
    {
      "epoch": 78.9,
      "learning_rate": 9.479352750809062e-06,
      "loss": 0.1578,
      "step": 203180
    },
    {
      "epoch": 78.91,
      "learning_rate": 9.478834951456312e-06,
      "loss": 0.064,
      "step": 203190
    },
    {
      "epoch": 78.91,
      "learning_rate": 9.478317152103562e-06,
      "loss": 0.031,
      "step": 203200
    },
    {
      "epoch": 78.92,
      "learning_rate": 9.47779935275081e-06,
      "loss": 0.0732,
      "step": 203210
    },
    {
      "epoch": 78.92,
      "learning_rate": 9.477281553398059e-06,
      "loss": 0.0991,
      "step": 203220
    },
    {
      "epoch": 78.92,
      "learning_rate": 9.476763754045309e-06,
      "loss": 0.0566,
      "step": 203230
    },
    {
      "epoch": 78.93,
      "learning_rate": 9.476245954692558e-06,
      "loss": 0.0271,
      "step": 203240
    },
    {
      "epoch": 78.93,
      "learning_rate": 9.475728155339806e-06,
      "loss": 0.0617,
      "step": 203250
    },
    {
      "epoch": 78.94,
      "learning_rate": 9.475210355987056e-06,
      "loss": 0.0155,
      "step": 203260
    },
    {
      "epoch": 78.94,
      "learning_rate": 9.474692556634305e-06,
      "loss": 0.0194,
      "step": 203270
    },
    {
      "epoch": 78.94,
      "learning_rate": 9.474174757281555e-06,
      "loss": 0.0711,
      "step": 203280
    },
    {
      "epoch": 78.95,
      "learning_rate": 9.473656957928803e-06,
      "loss": 0.0493,
      "step": 203290
    },
    {
      "epoch": 78.95,
      "learning_rate": 9.473139158576053e-06,
      "loss": 0.0831,
      "step": 203300
    },
    {
      "epoch": 78.96,
      "learning_rate": 9.472621359223302e-06,
      "loss": 0.002,
      "step": 203310
    },
    {
      "epoch": 78.96,
      "learning_rate": 9.472103559870552e-06,
      "loss": 0.0491,
      "step": 203320
    },
    {
      "epoch": 78.96,
      "learning_rate": 9.4715857605178e-06,
      "loss": 0.0478,
      "step": 203330
    },
    {
      "epoch": 78.97,
      "learning_rate": 9.47106796116505e-06,
      "loss": 0.0745,
      "step": 203340
    },
    {
      "epoch": 78.97,
      "learning_rate": 9.470550161812299e-06,
      "loss": 0.079,
      "step": 203350
    },
    {
      "epoch": 78.97,
      "learning_rate": 9.470032362459549e-06,
      "loss": 0.2838,
      "step": 203360
    },
    {
      "epoch": 78.98,
      "learning_rate": 9.469514563106797e-06,
      "loss": 0.2464,
      "step": 203370
    },
    {
      "epoch": 78.98,
      "learning_rate": 9.468996763754046e-06,
      "loss": 0.0393,
      "step": 203380
    },
    {
      "epoch": 78.99,
      "learning_rate": 9.468478964401296e-06,
      "loss": 0.0001,
      "step": 203390
    },
    {
      "epoch": 78.99,
      "learning_rate": 9.467961165048545e-06,
      "loss": 0.0093,
      "step": 203400
    },
    {
      "epoch": 78.99,
      "learning_rate": 9.467443365695793e-06,
      "loss": 0.0123,
      "step": 203410
    },
    {
      "epoch": 79.0,
      "learning_rate": 9.466925566343043e-06,
      "loss": 0.1442,
      "step": 203420
    },
    {
      "epoch": 79.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.34370800852775574,
      "eval_runtime": 8.2133,
      "eval_samples_per_second": 442.572,
      "eval_steps_per_second": 55.398,
      "step": 203425
    },
    {
      "epoch": 79.0,
      "learning_rate": 9.466407766990293e-06,
      "loss": 0.0215,
      "step": 203430
    },
    {
      "epoch": 79.01,
      "learning_rate": 9.465889967637542e-06,
      "loss": 0.0243,
      "step": 203440
    },
    {
      "epoch": 79.01,
      "learning_rate": 9.46537216828479e-06,
      "loss": 0.0405,
      "step": 203450
    },
    {
      "epoch": 79.01,
      "learning_rate": 9.46485436893204e-06,
      "loss": 0.0826,
      "step": 203460
    },
    {
      "epoch": 79.02,
      "learning_rate": 9.46433656957929e-06,
      "loss": 0.0668,
      "step": 203470
    },
    {
      "epoch": 79.02,
      "learning_rate": 9.463818770226539e-06,
      "loss": 0.0086,
      "step": 203480
    },
    {
      "epoch": 79.03,
      "learning_rate": 9.463300970873787e-06,
      "loss": 0.0773,
      "step": 203490
    },
    {
      "epoch": 79.03,
      "learning_rate": 9.462783171521037e-06,
      "loss": 0.091,
      "step": 203500
    },
    {
      "epoch": 79.03,
      "learning_rate": 9.462265372168286e-06,
      "loss": 0.0131,
      "step": 203510
    },
    {
      "epoch": 79.04,
      "learning_rate": 9.461747572815536e-06,
      "loss": 0.0552,
      "step": 203520
    },
    {
      "epoch": 79.04,
      "learning_rate": 9.461229773462784e-06,
      "loss": 0.1489,
      "step": 203530
    },
    {
      "epoch": 79.04,
      "learning_rate": 9.460711974110033e-06,
      "loss": 0.2338,
      "step": 203540
    },
    {
      "epoch": 79.05,
      "learning_rate": 9.460194174757283e-06,
      "loss": 0.0773,
      "step": 203550
    },
    {
      "epoch": 79.05,
      "learning_rate": 9.459676375404533e-06,
      "loss": 0.0177,
      "step": 203560
    },
    {
      "epoch": 79.06,
      "learning_rate": 9.45915857605178e-06,
      "loss": 0.0946,
      "step": 203570
    },
    {
      "epoch": 79.06,
      "learning_rate": 9.45864077669903e-06,
      "loss": 0.0895,
      "step": 203580
    },
    {
      "epoch": 79.06,
      "learning_rate": 9.45812297734628e-06,
      "loss": 0.0428,
      "step": 203590
    },
    {
      "epoch": 79.07,
      "learning_rate": 9.45760517799353e-06,
      "loss": 0.012,
      "step": 203600
    },
    {
      "epoch": 79.07,
      "learning_rate": 9.457087378640777e-06,
      "loss": 0.0153,
      "step": 203610
    },
    {
      "epoch": 79.08,
      "learning_rate": 9.456569579288027e-06,
      "loss": 0.0332,
      "step": 203620
    },
    {
      "epoch": 79.08,
      "learning_rate": 9.456051779935276e-06,
      "loss": 0.0001,
      "step": 203630
    },
    {
      "epoch": 79.08,
      "learning_rate": 9.455533980582524e-06,
      "loss": 0.0975,
      "step": 203640
    },
    {
      "epoch": 79.09,
      "learning_rate": 9.455016181229774e-06,
      "loss": 0.0544,
      "step": 203650
    },
    {
      "epoch": 79.09,
      "learning_rate": 9.454498381877024e-06,
      "loss": 0.1182,
      "step": 203660
    },
    {
      "epoch": 79.1,
      "learning_rate": 9.453980582524273e-06,
      "loss": 0.0481,
      "step": 203670
    },
    {
      "epoch": 79.1,
      "learning_rate": 9.453462783171521e-06,
      "loss": 0.0123,
      "step": 203680
    },
    {
      "epoch": 79.1,
      "learning_rate": 9.45294498381877e-06,
      "loss": 0.1767,
      "step": 203690
    },
    {
      "epoch": 79.11,
      "learning_rate": 9.45242718446602e-06,
      "loss": 0.1113,
      "step": 203700
    },
    {
      "epoch": 79.11,
      "learning_rate": 9.45190938511327e-06,
      "loss": 0.0023,
      "step": 203710
    },
    {
      "epoch": 79.11,
      "learning_rate": 9.451391585760518e-06,
      "loss": 0.0117,
      "step": 203720
    },
    {
      "epoch": 79.12,
      "learning_rate": 9.450873786407768e-06,
      "loss": 0.0651,
      "step": 203730
    },
    {
      "epoch": 79.12,
      "learning_rate": 9.450355987055017e-06,
      "loss": 0.0793,
      "step": 203740
    },
    {
      "epoch": 79.13,
      "learning_rate": 9.449838187702267e-06,
      "loss": 0.0913,
      "step": 203750
    },
    {
      "epoch": 79.13,
      "learning_rate": 9.449320388349515e-06,
      "loss": 0.008,
      "step": 203760
    },
    {
      "epoch": 79.13,
      "learning_rate": 9.448802588996764e-06,
      "loss": 0.0823,
      "step": 203770
    },
    {
      "epoch": 79.14,
      "learning_rate": 9.448284789644014e-06,
      "loss": 0.1207,
      "step": 203780
    },
    {
      "epoch": 79.14,
      "learning_rate": 9.447766990291264e-06,
      "loss": 0.0821,
      "step": 203790
    },
    {
      "epoch": 79.15,
      "learning_rate": 9.447249190938512e-06,
      "loss": 0.0655,
      "step": 203800
    },
    {
      "epoch": 79.15,
      "learning_rate": 9.446731391585761e-06,
      "loss": 0.1306,
      "step": 203810
    },
    {
      "epoch": 79.15,
      "learning_rate": 9.44621359223301e-06,
      "loss": 0.0578,
      "step": 203820
    },
    {
      "epoch": 79.16,
      "learning_rate": 9.445695792880259e-06,
      "loss": 0.0612,
      "step": 203830
    },
    {
      "epoch": 79.16,
      "learning_rate": 9.445177993527508e-06,
      "loss": 0.0676,
      "step": 203840
    },
    {
      "epoch": 79.17,
      "learning_rate": 9.444660194174758e-06,
      "loss": 0.0574,
      "step": 203850
    },
    {
      "epoch": 79.17,
      "learning_rate": 9.444142394822008e-06,
      "loss": 0.0633,
      "step": 203860
    },
    {
      "epoch": 79.17,
      "learning_rate": 9.443624595469255e-06,
      "loss": 0.1024,
      "step": 203870
    },
    {
      "epoch": 79.18,
      "learning_rate": 9.443106796116505e-06,
      "loss": 0.0316,
      "step": 203880
    },
    {
      "epoch": 79.18,
      "learning_rate": 9.442588996763755e-06,
      "loss": 0.0098,
      "step": 203890
    },
    {
      "epoch": 79.18,
      "learning_rate": 9.442071197411004e-06,
      "loss": 0.0003,
      "step": 203900
    },
    {
      "epoch": 79.19,
      "learning_rate": 9.441553398058252e-06,
      "loss": 0.0588,
      "step": 203910
    },
    {
      "epoch": 79.19,
      "learning_rate": 9.441035598705502e-06,
      "loss": 0.0797,
      "step": 203920
    },
    {
      "epoch": 79.2,
      "learning_rate": 9.440517799352751e-06,
      "loss": 0.0092,
      "step": 203930
    },
    {
      "epoch": 79.2,
      "learning_rate": 9.440000000000001e-06,
      "loss": 0.0717,
      "step": 203940
    },
    {
      "epoch": 79.2,
      "learning_rate": 9.439482200647249e-06,
      "loss": 0.0446,
      "step": 203950
    },
    {
      "epoch": 79.21,
      "learning_rate": 9.438964401294499e-06,
      "loss": 0.0004,
      "step": 203960
    },
    {
      "epoch": 79.21,
      "learning_rate": 9.438446601941748e-06,
      "loss": 0.0452,
      "step": 203970
    },
    {
      "epoch": 79.22,
      "learning_rate": 9.437928802588996e-06,
      "loss": 0.0564,
      "step": 203980
    },
    {
      "epoch": 79.22,
      "learning_rate": 9.437411003236246e-06,
      "loss": 0.0034,
      "step": 203990
    },
    {
      "epoch": 79.22,
      "learning_rate": 9.436893203883495e-06,
      "loss": 0.092,
      "step": 204000
    },
    {
      "epoch": 79.23,
      "learning_rate": 9.436375404530745e-06,
      "loss": 0.1499,
      "step": 204010
    },
    {
      "epoch": 79.23,
      "learning_rate": 9.435857605177993e-06,
      "loss": 0.0012,
      "step": 204020
    },
    {
      "epoch": 79.23,
      "learning_rate": 9.435339805825243e-06,
      "loss": 0.0254,
      "step": 204030
    },
    {
      "epoch": 79.24,
      "learning_rate": 9.434822006472492e-06,
      "loss": 0.1441,
      "step": 204040
    },
    {
      "epoch": 79.24,
      "learning_rate": 9.434304207119742e-06,
      "loss": 0.0348,
      "step": 204050
    },
    {
      "epoch": 79.25,
      "learning_rate": 9.433786407766991e-06,
      "loss": 0.0188,
      "step": 204060
    },
    {
      "epoch": 79.25,
      "learning_rate": 9.43326860841424e-06,
      "loss": 0.0662,
      "step": 204070
    },
    {
      "epoch": 79.25,
      "learning_rate": 9.432750809061489e-06,
      "loss": 0.1723,
      "step": 204080
    },
    {
      "epoch": 79.26,
      "learning_rate": 9.432233009708739e-06,
      "loss": 0.0952,
      "step": 204090
    },
    {
      "epoch": 79.26,
      "learning_rate": 9.431715210355988e-06,
      "loss": 0.0452,
      "step": 204100
    },
    {
      "epoch": 79.27,
      "learning_rate": 9.431197411003236e-06,
      "loss": 0.0775,
      "step": 204110
    },
    {
      "epoch": 79.27,
      "learning_rate": 9.430679611650486e-06,
      "loss": 0.0424,
      "step": 204120
    },
    {
      "epoch": 79.27,
      "learning_rate": 9.430161812297735e-06,
      "loss": 0.1063,
      "step": 204130
    },
    {
      "epoch": 79.28,
      "learning_rate": 9.429644012944985e-06,
      "loss": 0.0951,
      "step": 204140
    },
    {
      "epoch": 79.28,
      "learning_rate": 9.429126213592233e-06,
      "loss": 0.0609,
      "step": 204150
    },
    {
      "epoch": 79.29,
      "learning_rate": 9.428608414239483e-06,
      "loss": 0.0816,
      "step": 204160
    },
    {
      "epoch": 79.29,
      "learning_rate": 9.428090614886732e-06,
      "loss": 0.2035,
      "step": 204170
    },
    {
      "epoch": 79.29,
      "learning_rate": 9.427572815533982e-06,
      "loss": 0.0802,
      "step": 204180
    },
    {
      "epoch": 79.3,
      "learning_rate": 9.42705501618123e-06,
      "loss": 0.0143,
      "step": 204190
    },
    {
      "epoch": 79.3,
      "learning_rate": 9.42653721682848e-06,
      "loss": 0.0109,
      "step": 204200
    },
    {
      "epoch": 79.3,
      "learning_rate": 9.426019417475729e-06,
      "loss": 0.0668,
      "step": 204210
    },
    {
      "epoch": 79.31,
      "learning_rate": 9.425501618122979e-06,
      "loss": 0.0482,
      "step": 204220
    },
    {
      "epoch": 79.31,
      "learning_rate": 9.424983818770226e-06,
      "loss": 0.1096,
      "step": 204230
    },
    {
      "epoch": 79.32,
      "learning_rate": 9.424466019417476e-06,
      "loss": 0.002,
      "step": 204240
    },
    {
      "epoch": 79.32,
      "learning_rate": 9.423948220064726e-06,
      "loss": 0.0594,
      "step": 204250
    },
    {
      "epoch": 79.32,
      "learning_rate": 9.423430420711975e-06,
      "loss": 0.0215,
      "step": 204260
    },
    {
      "epoch": 79.33,
      "learning_rate": 9.422912621359223e-06,
      "loss": 0.1454,
      "step": 204270
    },
    {
      "epoch": 79.33,
      "learning_rate": 9.422394822006473e-06,
      "loss": 0.0388,
      "step": 204280
    },
    {
      "epoch": 79.34,
      "learning_rate": 9.421877022653722e-06,
      "loss": 0.1316,
      "step": 204290
    },
    {
      "epoch": 79.34,
      "learning_rate": 9.421359223300972e-06,
      "loss": 0.0149,
      "step": 204300
    },
    {
      "epoch": 79.34,
      "learning_rate": 9.42084142394822e-06,
      "loss": 0.0736,
      "step": 204310
    },
    {
      "epoch": 79.35,
      "learning_rate": 9.42032362459547e-06,
      "loss": 0.0941,
      "step": 204320
    },
    {
      "epoch": 79.35,
      "learning_rate": 9.41980582524272e-06,
      "loss": 0.0467,
      "step": 204330
    },
    {
      "epoch": 79.36,
      "learning_rate": 9.419288025889969e-06,
      "loss": 0.0352,
      "step": 204340
    },
    {
      "epoch": 79.36,
      "learning_rate": 9.418770226537217e-06,
      "loss": 0.0983,
      "step": 204350
    },
    {
      "epoch": 79.36,
      "learning_rate": 9.418252427184466e-06,
      "loss": 0.0121,
      "step": 204360
    },
    {
      "epoch": 79.37,
      "learning_rate": 9.417734627831716e-06,
      "loss": 0.0323,
      "step": 204370
    },
    {
      "epoch": 79.37,
      "learning_rate": 9.417216828478966e-06,
      "loss": 0.026,
      "step": 204380
    },
    {
      "epoch": 79.37,
      "learning_rate": 9.416699029126214e-06,
      "loss": 0.0925,
      "step": 204390
    },
    {
      "epoch": 79.38,
      "learning_rate": 9.416181229773463e-06,
      "loss": 0.0121,
      "step": 204400
    },
    {
      "epoch": 79.38,
      "learning_rate": 9.415663430420713e-06,
      "loss": 0.1486,
      "step": 204410
    },
    {
      "epoch": 79.39,
      "learning_rate": 9.415145631067962e-06,
      "loss": 0.0494,
      "step": 204420
    },
    {
      "epoch": 79.39,
      "learning_rate": 9.41462783171521e-06,
      "loss": 0.1412,
      "step": 204430
    },
    {
      "epoch": 79.39,
      "learning_rate": 9.41411003236246e-06,
      "loss": 0.1564,
      "step": 204440
    },
    {
      "epoch": 79.4,
      "learning_rate": 9.41359223300971e-06,
      "loss": 0.0455,
      "step": 204450
    },
    {
      "epoch": 79.4,
      "learning_rate": 9.41307443365696e-06,
      "loss": 0.1792,
      "step": 204460
    },
    {
      "epoch": 79.41,
      "learning_rate": 9.412556634304207e-06,
      "loss": 0.119,
      "step": 204470
    },
    {
      "epoch": 79.41,
      "learning_rate": 9.412038834951457e-06,
      "loss": 0.0333,
      "step": 204480
    },
    {
      "epoch": 79.41,
      "learning_rate": 9.411521035598706e-06,
      "loss": 0.0799,
      "step": 204490
    },
    {
      "epoch": 79.42,
      "learning_rate": 9.411003236245956e-06,
      "loss": 0.1046,
      "step": 204500
    },
    {
      "epoch": 79.42,
      "learning_rate": 9.410485436893204e-06,
      "loss": 0.0444,
      "step": 204510
    },
    {
      "epoch": 79.43,
      "learning_rate": 9.409967637540454e-06,
      "loss": 0.0618,
      "step": 204520
    },
    {
      "epoch": 79.43,
      "learning_rate": 9.409449838187703e-06,
      "loss": 0.092,
      "step": 204530
    },
    {
      "epoch": 79.43,
      "learning_rate": 9.408932038834953e-06,
      "loss": 0.0721,
      "step": 204540
    },
    {
      "epoch": 79.44,
      "learning_rate": 9.4084142394822e-06,
      "loss": 0.0582,
      "step": 204550
    },
    {
      "epoch": 79.44,
      "learning_rate": 9.40789644012945e-06,
      "loss": 0.0133,
      "step": 204560
    },
    {
      "epoch": 79.44,
      "learning_rate": 9.4073786407767e-06,
      "loss": 0.2454,
      "step": 204570
    },
    {
      "epoch": 79.45,
      "learning_rate": 9.40686084142395e-06,
      "loss": 0.1449,
      "step": 204580
    },
    {
      "epoch": 79.45,
      "learning_rate": 9.406343042071197e-06,
      "loss": 0.0703,
      "step": 204590
    },
    {
      "epoch": 79.46,
      "learning_rate": 9.405825242718447e-06,
      "loss": 0.0587,
      "step": 204600
    },
    {
      "epoch": 79.46,
      "learning_rate": 9.405307443365697e-06,
      "loss": 0.0351,
      "step": 204610
    },
    {
      "epoch": 79.46,
      "learning_rate": 9.404789644012946e-06,
      "loss": 0.0488,
      "step": 204620
    },
    {
      "epoch": 79.47,
      "learning_rate": 9.404271844660196e-06,
      "loss": 0.0469,
      "step": 204630
    },
    {
      "epoch": 79.47,
      "learning_rate": 9.403754045307444e-06,
      "loss": 0.0325,
      "step": 204640
    },
    {
      "epoch": 79.48,
      "learning_rate": 9.403236245954693e-06,
      "loss": 0.1569,
      "step": 204650
    },
    {
      "epoch": 79.48,
      "learning_rate": 9.402718446601943e-06,
      "loss": 0.1724,
      "step": 204660
    },
    {
      "epoch": 79.48,
      "learning_rate": 9.402200647249193e-06,
      "loss": 0.0611,
      "step": 204670
    },
    {
      "epoch": 79.49,
      "learning_rate": 9.40168284789644e-06,
      "loss": 0.0665,
      "step": 204680
    },
    {
      "epoch": 79.49,
      "learning_rate": 9.40116504854369e-06,
      "loss": 0.0588,
      "step": 204690
    },
    {
      "epoch": 79.5,
      "learning_rate": 9.40064724919094e-06,
      "loss": 0.0473,
      "step": 204700
    },
    {
      "epoch": 79.5,
      "learning_rate": 9.40012944983819e-06,
      "loss": 0.1229,
      "step": 204710
    },
    {
      "epoch": 79.5,
      "learning_rate": 9.399611650485437e-06,
      "loss": 0.0701,
      "step": 204720
    },
    {
      "epoch": 79.51,
      "learning_rate": 9.399093851132687e-06,
      "loss": 0.0505,
      "step": 204730
    },
    {
      "epoch": 79.51,
      "learning_rate": 9.398576051779937e-06,
      "loss": 0.0267,
      "step": 204740
    },
    {
      "epoch": 79.51,
      "learning_rate": 9.398058252427186e-06,
      "loss": 0.059,
      "step": 204750
    },
    {
      "epoch": 79.52,
      "learning_rate": 9.397540453074434e-06,
      "loss": 0.0815,
      "step": 204760
    },
    {
      "epoch": 79.52,
      "learning_rate": 9.397022653721684e-06,
      "loss": 0.0207,
      "step": 204770
    },
    {
      "epoch": 79.53,
      "learning_rate": 9.396504854368933e-06,
      "loss": 0.0948,
      "step": 204780
    },
    {
      "epoch": 79.53,
      "learning_rate": 9.395987055016183e-06,
      "loss": 0.0263,
      "step": 204790
    },
    {
      "epoch": 79.53,
      "learning_rate": 9.395469255663431e-06,
      "loss": 0.0982,
      "step": 204800
    },
    {
      "epoch": 79.54,
      "learning_rate": 9.39495145631068e-06,
      "loss": 0.0094,
      "step": 204810
    },
    {
      "epoch": 79.54,
      "learning_rate": 9.39443365695793e-06,
      "loss": 0.0636,
      "step": 204820
    },
    {
      "epoch": 79.55,
      "learning_rate": 9.39391585760518e-06,
      "loss": 0.0495,
      "step": 204830
    },
    {
      "epoch": 79.55,
      "learning_rate": 9.393398058252428e-06,
      "loss": 0.0671,
      "step": 204840
    },
    {
      "epoch": 79.55,
      "learning_rate": 9.392880258899677e-06,
      "loss": 0.1453,
      "step": 204850
    },
    {
      "epoch": 79.56,
      "learning_rate": 9.392362459546927e-06,
      "loss": 0.0352,
      "step": 204860
    },
    {
      "epoch": 79.56,
      "learning_rate": 9.391844660194177e-06,
      "loss": 0.099,
      "step": 204870
    },
    {
      "epoch": 79.57,
      "learning_rate": 9.391326860841425e-06,
      "loss": 0.0908,
      "step": 204880
    },
    {
      "epoch": 79.57,
      "learning_rate": 9.390809061488674e-06,
      "loss": 0.017,
      "step": 204890
    },
    {
      "epoch": 79.57,
      "learning_rate": 9.390291262135924e-06,
      "loss": 0.05,
      "step": 204900
    },
    {
      "epoch": 79.58,
      "learning_rate": 9.389773462783173e-06,
      "loss": 0.0569,
      "step": 204910
    },
    {
      "epoch": 79.58,
      "learning_rate": 9.389255663430421e-06,
      "loss": 0.0199,
      "step": 204920
    },
    {
      "epoch": 79.58,
      "learning_rate": 9.388737864077671e-06,
      "loss": 0.1493,
      "step": 204930
    },
    {
      "epoch": 79.59,
      "learning_rate": 9.38822006472492e-06,
      "loss": 0.1191,
      "step": 204940
    },
    {
      "epoch": 79.59,
      "learning_rate": 9.38770226537217e-06,
      "loss": 0.0598,
      "step": 204950
    },
    {
      "epoch": 79.6,
      "learning_rate": 9.387184466019418e-06,
      "loss": 0.1,
      "step": 204960
    },
    {
      "epoch": 79.6,
      "learning_rate": 9.386666666666668e-06,
      "loss": 0.0244,
      "step": 204970
    },
    {
      "epoch": 79.6,
      "learning_rate": 9.386148867313917e-06,
      "loss": 0.0878,
      "step": 204980
    },
    {
      "epoch": 79.61,
      "learning_rate": 9.385631067961167e-06,
      "loss": 0.04,
      "step": 204990
    },
    {
      "epoch": 79.61,
      "learning_rate": 9.385113268608415e-06,
      "loss": 0.0689,
      "step": 205000
    },
    {
      "epoch": 79.62,
      "learning_rate": 9.384595469255664e-06,
      "loss": 0.0059,
      "step": 205010
    },
    {
      "epoch": 79.62,
      "learning_rate": 9.384077669902914e-06,
      "loss": 0.1311,
      "step": 205020
    },
    {
      "epoch": 79.62,
      "learning_rate": 9.383559870550164e-06,
      "loss": 0.0787,
      "step": 205030
    },
    {
      "epoch": 79.63,
      "learning_rate": 9.383042071197412e-06,
      "loss": 0.1991,
      "step": 205040
    },
    {
      "epoch": 79.63,
      "learning_rate": 9.382524271844661e-06,
      "loss": 0.0569,
      "step": 205050
    },
    {
      "epoch": 79.63,
      "learning_rate": 9.382006472491911e-06,
      "loss": 0.0109,
      "step": 205060
    },
    {
      "epoch": 79.64,
      "learning_rate": 9.38148867313916e-06,
      "loss": 0.0835,
      "step": 205070
    },
    {
      "epoch": 79.64,
      "learning_rate": 9.380970873786408e-06,
      "loss": 0.013,
      "step": 205080
    },
    {
      "epoch": 79.65,
      "learning_rate": 9.380453074433658e-06,
      "loss": 0.0883,
      "step": 205090
    },
    {
      "epoch": 79.65,
      "learning_rate": 9.379935275080908e-06,
      "loss": 0.0013,
      "step": 205100
    },
    {
      "epoch": 79.65,
      "learning_rate": 9.379417475728156e-06,
      "loss": 0.1321,
      "step": 205110
    },
    {
      "epoch": 79.66,
      "learning_rate": 9.378899676375405e-06,
      "loss": 0.0129,
      "step": 205120
    },
    {
      "epoch": 79.66,
      "learning_rate": 9.378381877022655e-06,
      "loss": 0.0003,
      "step": 205130
    },
    {
      "epoch": 79.67,
      "learning_rate": 9.377864077669904e-06,
      "loss": 0.033,
      "step": 205140
    },
    {
      "epoch": 79.67,
      "learning_rate": 9.377346278317152e-06,
      "loss": 0.084,
      "step": 205150
    },
    {
      "epoch": 79.67,
      "learning_rate": 9.376828478964402e-06,
      "loss": 0.0898,
      "step": 205160
    },
    {
      "epoch": 79.68,
      "learning_rate": 9.376310679611652e-06,
      "loss": 0.0446,
      "step": 205170
    },
    {
      "epoch": 79.68,
      "learning_rate": 9.375792880258901e-06,
      "loss": 0.0671,
      "step": 205180
    },
    {
      "epoch": 79.69,
      "learning_rate": 9.375275080906149e-06,
      "loss": 0.0332,
      "step": 205190
    },
    {
      "epoch": 79.69,
      "learning_rate": 9.374757281553399e-06,
      "loss": 0.074,
      "step": 205200
    },
    {
      "epoch": 79.69,
      "learning_rate": 9.374239482200648e-06,
      "loss": 0.0534,
      "step": 205210
    },
    {
      "epoch": 79.7,
      "learning_rate": 9.373721682847898e-06,
      "loss": 0.1099,
      "step": 205220
    },
    {
      "epoch": 79.7,
      "learning_rate": 9.373203883495146e-06,
      "loss": 0.1084,
      "step": 205230
    },
    {
      "epoch": 79.7,
      "learning_rate": 9.372686084142396e-06,
      "loss": 0.0561,
      "step": 205240
    },
    {
      "epoch": 79.71,
      "learning_rate": 9.372168284789645e-06,
      "loss": 0.0249,
      "step": 205250
    },
    {
      "epoch": 79.71,
      "learning_rate": 9.371650485436895e-06,
      "loss": 0.0369,
      "step": 205260
    },
    {
      "epoch": 79.72,
      "learning_rate": 9.371132686084143e-06,
      "loss": 0.0109,
      "step": 205270
    },
    {
      "epoch": 79.72,
      "learning_rate": 9.370614886731392e-06,
      "loss": 0.0589,
      "step": 205280
    },
    {
      "epoch": 79.72,
      "learning_rate": 9.370097087378642e-06,
      "loss": 0.1434,
      "step": 205290
    },
    {
      "epoch": 79.73,
      "learning_rate": 9.36957928802589e-06,
      "loss": 0.0061,
      "step": 205300
    },
    {
      "epoch": 79.73,
      "learning_rate": 9.36906148867314e-06,
      "loss": 0.0846,
      "step": 205310
    },
    {
      "epoch": 79.74,
      "learning_rate": 9.368543689320389e-06,
      "loss": 0.0393,
      "step": 205320
    },
    {
      "epoch": 79.74,
      "learning_rate": 9.368025889967639e-06,
      "loss": 0.0396,
      "step": 205330
    },
    {
      "epoch": 79.74,
      "learning_rate": 9.367508090614887e-06,
      "loss": 0.0806,
      "step": 205340
    },
    {
      "epoch": 79.75,
      "learning_rate": 9.366990291262136e-06,
      "loss": 0.0018,
      "step": 205350
    },
    {
      "epoch": 79.75,
      "learning_rate": 9.366472491909386e-06,
      "loss": 0.0677,
      "step": 205360
    },
    {
      "epoch": 79.76,
      "learning_rate": 9.365954692556635e-06,
      "loss": 0.0392,
      "step": 205370
    },
    {
      "epoch": 79.76,
      "learning_rate": 9.365436893203883e-06,
      "loss": 0.0529,
      "step": 205380
    },
    {
      "epoch": 79.76,
      "learning_rate": 9.364919093851133e-06,
      "loss": 0.0232,
      "step": 205390
    },
    {
      "epoch": 79.77,
      "learning_rate": 9.364401294498383e-06,
      "loss": 0.0162,
      "step": 205400
    },
    {
      "epoch": 79.77,
      "learning_rate": 9.363883495145632e-06,
      "loss": 0.0366,
      "step": 205410
    },
    {
      "epoch": 79.77,
      "learning_rate": 9.36336569579288e-06,
      "loss": 0.044,
      "step": 205420
    },
    {
      "epoch": 79.78,
      "learning_rate": 9.36284789644013e-06,
      "loss": 0.0997,
      "step": 205430
    },
    {
      "epoch": 79.78,
      "learning_rate": 9.36233009708738e-06,
      "loss": 0.0627,
      "step": 205440
    },
    {
      "epoch": 79.79,
      "learning_rate": 9.361812297734627e-06,
      "loss": 0.0558,
      "step": 205450
    },
    {
      "epoch": 79.79,
      "learning_rate": 9.361294498381877e-06,
      "loss": 0.0453,
      "step": 205460
    },
    {
      "epoch": 79.79,
      "learning_rate": 9.360776699029127e-06,
      "loss": 0.0619,
      "step": 205470
    },
    {
      "epoch": 79.8,
      "learning_rate": 9.360258899676376e-06,
      "loss": 0.0439,
      "step": 205480
    },
    {
      "epoch": 79.8,
      "learning_rate": 9.359741100323624e-06,
      "loss": 0.133,
      "step": 205490
    },
    {
      "epoch": 79.81,
      "learning_rate": 9.359223300970874e-06,
      "loss": 0.0712,
      "step": 205500
    },
    {
      "epoch": 79.81,
      "learning_rate": 9.358705501618123e-06,
      "loss": 0.1207,
      "step": 205510
    },
    {
      "epoch": 79.81,
      "learning_rate": 9.358187702265373e-06,
      "loss": 0.0508,
      "step": 205520
    },
    {
      "epoch": 79.82,
      "learning_rate": 9.357669902912621e-06,
      "loss": 0.1441,
      "step": 205530
    },
    {
      "epoch": 79.82,
      "learning_rate": 9.35715210355987e-06,
      "loss": 0.0574,
      "step": 205540
    },
    {
      "epoch": 79.83,
      "learning_rate": 9.35663430420712e-06,
      "loss": 0.0427,
      "step": 205550
    },
    {
      "epoch": 79.83,
      "learning_rate": 9.35611650485437e-06,
      "loss": 0.0061,
      "step": 205560
    },
    {
      "epoch": 79.83,
      "learning_rate": 9.355598705501618e-06,
      "loss": 0.0251,
      "step": 205570
    },
    {
      "epoch": 79.84,
      "learning_rate": 9.355080906148867e-06,
      "loss": 0.1727,
      "step": 205580
    },
    {
      "epoch": 79.84,
      "learning_rate": 9.354563106796117e-06,
      "loss": 0.0073,
      "step": 205590
    },
    {
      "epoch": 79.84,
      "learning_rate": 9.354045307443367e-06,
      "loss": 0.0162,
      "step": 205600
    },
    {
      "epoch": 79.85,
      "learning_rate": 9.353527508090614e-06,
      "loss": 0.0669,
      "step": 205610
    },
    {
      "epoch": 79.85,
      "learning_rate": 9.353009708737864e-06,
      "loss": 0.0325,
      "step": 205620
    },
    {
      "epoch": 79.86,
      "learning_rate": 9.352491909385114e-06,
      "loss": 0.0622,
      "step": 205630
    },
    {
      "epoch": 79.86,
      "learning_rate": 9.351974110032363e-06,
      "loss": 0.0537,
      "step": 205640
    },
    {
      "epoch": 79.86,
      "learning_rate": 9.351456310679611e-06,
      "loss": 0.021,
      "step": 205650
    },
    {
      "epoch": 79.87,
      "learning_rate": 9.350938511326861e-06,
      "loss": 0.0978,
      "step": 205660
    },
    {
      "epoch": 79.87,
      "learning_rate": 9.35042071197411e-06,
      "loss": 0.0339,
      "step": 205670
    },
    {
      "epoch": 79.88,
      "learning_rate": 9.34990291262136e-06,
      "loss": 0.0694,
      "step": 205680
    },
    {
      "epoch": 79.88,
      "learning_rate": 9.349385113268608e-06,
      "loss": 0.4598,
      "step": 205690
    },
    {
      "epoch": 79.88,
      "learning_rate": 9.348867313915858e-06,
      "loss": 0.0146,
      "step": 205700
    },
    {
      "epoch": 79.89,
      "learning_rate": 9.348349514563107e-06,
      "loss": 0.0211,
      "step": 205710
    },
    {
      "epoch": 79.89,
      "learning_rate": 9.347831715210357e-06,
      "loss": 0.0019,
      "step": 205720
    },
    {
      "epoch": 79.9,
      "learning_rate": 9.347313915857605e-06,
      "loss": 0.0669,
      "step": 205730
    },
    {
      "epoch": 79.9,
      "learning_rate": 9.346796116504854e-06,
      "loss": 0.0844,
      "step": 205740
    },
    {
      "epoch": 79.9,
      "learning_rate": 9.346278317152104e-06,
      "loss": 0.0593,
      "step": 205750
    },
    {
      "epoch": 79.91,
      "learning_rate": 9.345760517799354e-06,
      "loss": 0.0837,
      "step": 205760
    },
    {
      "epoch": 79.91,
      "learning_rate": 9.345242718446603e-06,
      "loss": 0.028,
      "step": 205770
    },
    {
      "epoch": 79.91,
      "learning_rate": 9.344724919093851e-06,
      "loss": 0.0122,
      "step": 205780
    },
    {
      "epoch": 79.92,
      "learning_rate": 9.3442071197411e-06,
      "loss": 0.0547,
      "step": 205790
    },
    {
      "epoch": 79.92,
      "learning_rate": 9.34368932038835e-06,
      "loss": 0.2912,
      "step": 205800
    },
    {
      "epoch": 79.93,
      "learning_rate": 9.3431715210356e-06,
      "loss": 0.0863,
      "step": 205810
    },
    {
      "epoch": 79.93,
      "learning_rate": 9.342653721682848e-06,
      "loss": 0.0759,
      "step": 205820
    },
    {
      "epoch": 79.93,
      "learning_rate": 9.342135922330098e-06,
      "loss": 0.1011,
      "step": 205830
    },
    {
      "epoch": 79.94,
      "learning_rate": 9.341618122977347e-06,
      "loss": 0.0452,
      "step": 205840
    },
    {
      "epoch": 79.94,
      "learning_rate": 9.341100323624597e-06,
      "loss": 0.034,
      "step": 205850
    },
    {
      "epoch": 79.95,
      "learning_rate": 9.340582524271845e-06,
      "loss": 0.1132,
      "step": 205860
    },
    {
      "epoch": 79.95,
      "learning_rate": 9.340064724919094e-06,
      "loss": 0.0465,
      "step": 205870
    },
    {
      "epoch": 79.95,
      "learning_rate": 9.339546925566344e-06,
      "loss": 0.003,
      "step": 205880
    },
    {
      "epoch": 79.96,
      "learning_rate": 9.339029126213594e-06,
      "loss": 0.1065,
      "step": 205890
    },
    {
      "epoch": 79.96,
      "learning_rate": 9.338511326860842e-06,
      "loss": 0.1086,
      "step": 205900
    },
    {
      "epoch": 79.97,
      "learning_rate": 9.337993527508091e-06,
      "loss": 0.1116,
      "step": 205910
    },
    {
      "epoch": 79.97,
      "learning_rate": 9.33747572815534e-06,
      "loss": 0.0332,
      "step": 205920
    },
    {
      "epoch": 79.97,
      "learning_rate": 9.33695792880259e-06,
      "loss": 0.0698,
      "step": 205930
    },
    {
      "epoch": 79.98,
      "learning_rate": 9.336440129449838e-06,
      "loss": 0.0369,
      "step": 205940
    },
    {
      "epoch": 79.98,
      "learning_rate": 9.335922330097088e-06,
      "loss": 0.0163,
      "step": 205950
    },
    {
      "epoch": 79.98,
      "learning_rate": 9.335404530744338e-06,
      "loss": 0.0516,
      "step": 205960
    },
    {
      "epoch": 79.99,
      "learning_rate": 9.334886731391587e-06,
      "loss": 0.0105,
      "step": 205970
    },
    {
      "epoch": 79.99,
      "learning_rate": 9.334368932038835e-06,
      "loss": 0.0852,
      "step": 205980
    },
    {
      "epoch": 80.0,
      "learning_rate": 9.333851132686085e-06,
      "loss": 0.1047,
      "step": 205990
    },
    {
      "epoch": 80.0,
      "learning_rate": 9.333333333333334e-06,
      "loss": 0.0834,
      "step": 206000
    },
    {
      "epoch": 80.0,
      "eval_accuracy": 0.9524071526822558,
      "eval_loss": 0.3430519104003906,
      "eval_runtime": 8.1895,
      "eval_samples_per_second": 443.859,
      "eval_steps_per_second": 55.559,
      "step": 206000
    },
    {
      "epoch": 80.0,
      "learning_rate": 9.332815533980584e-06,
      "loss": 0.0701,
      "step": 206010
    },
    {
      "epoch": 80.01,
      "learning_rate": 9.332297734627832e-06,
      "loss": 0.0704,
      "step": 206020
    },
    {
      "epoch": 80.01,
      "learning_rate": 9.331779935275081e-06,
      "loss": 0.0757,
      "step": 206030
    },
    {
      "epoch": 80.02,
      "learning_rate": 9.331262135922331e-06,
      "loss": 0.034,
      "step": 206040
    },
    {
      "epoch": 80.02,
      "learning_rate": 9.33074433656958e-06,
      "loss": 0.2888,
      "step": 206050
    },
    {
      "epoch": 80.02,
      "learning_rate": 9.330226537216829e-06,
      "loss": 0.0143,
      "step": 206060
    },
    {
      "epoch": 80.03,
      "learning_rate": 9.329708737864078e-06,
      "loss": 0.0434,
      "step": 206070
    },
    {
      "epoch": 80.03,
      "learning_rate": 9.329190938511328e-06,
      "loss": 0.0008,
      "step": 206080
    },
    {
      "epoch": 80.03,
      "learning_rate": 9.328673139158577e-06,
      "loss": 0.0093,
      "step": 206090
    },
    {
      "epoch": 80.04,
      "learning_rate": 9.328155339805825e-06,
      "loss": 0.0194,
      "step": 206100
    },
    {
      "epoch": 80.04,
      "learning_rate": 9.327637540453075e-06,
      "loss": 0.0629,
      "step": 206110
    },
    {
      "epoch": 80.05,
      "learning_rate": 9.327119741100325e-06,
      "loss": 0.007,
      "step": 206120
    },
    {
      "epoch": 80.05,
      "learning_rate": 9.326601941747574e-06,
      "loss": 0.0005,
      "step": 206130
    },
    {
      "epoch": 80.05,
      "learning_rate": 9.326084142394822e-06,
      "loss": 0.0686,
      "step": 206140
    },
    {
      "epoch": 80.06,
      "learning_rate": 9.325566343042072e-06,
      "loss": 0.0946,
      "step": 206150
    },
    {
      "epoch": 80.06,
      "learning_rate": 9.325048543689321e-06,
      "loss": 0.0264,
      "step": 206160
    },
    {
      "epoch": 80.07,
      "learning_rate": 9.324530744336571e-06,
      "loss": 0.1151,
      "step": 206170
    },
    {
      "epoch": 80.07,
      "learning_rate": 9.324012944983819e-06,
      "loss": 0.0529,
      "step": 206180
    },
    {
      "epoch": 80.07,
      "learning_rate": 9.323495145631069e-06,
      "loss": 0.0466,
      "step": 206190
    },
    {
      "epoch": 80.08,
      "learning_rate": 9.322977346278318e-06,
      "loss": 0.0387,
      "step": 206200
    },
    {
      "epoch": 80.08,
      "learning_rate": 9.322459546925568e-06,
      "loss": 0.0217,
      "step": 206210
    },
    {
      "epoch": 80.09,
      "learning_rate": 9.321941747572816e-06,
      "loss": 0.0031,
      "step": 206220
    },
    {
      "epoch": 80.09,
      "learning_rate": 9.321423948220065e-06,
      "loss": 0.092,
      "step": 206230
    },
    {
      "epoch": 80.09,
      "learning_rate": 9.320906148867315e-06,
      "loss": 0.0472,
      "step": 206240
    },
    {
      "epoch": 80.1,
      "learning_rate": 9.320388349514565e-06,
      "loss": 0.0616,
      "step": 206250
    },
    {
      "epoch": 80.1,
      "learning_rate": 9.319870550161813e-06,
      "loss": 0.0198,
      "step": 206260
    },
    {
      "epoch": 80.1,
      "learning_rate": 9.319352750809062e-06,
      "loss": 0.0416,
      "step": 206270
    },
    {
      "epoch": 80.11,
      "learning_rate": 9.318834951456312e-06,
      "loss": 0.0336,
      "step": 206280
    },
    {
      "epoch": 80.11,
      "learning_rate": 9.318317152103561e-06,
      "loss": 0.0587,
      "step": 206290
    },
    {
      "epoch": 80.12,
      "learning_rate": 9.317799352750811e-06,
      "loss": 0.0071,
      "step": 206300
    },
    {
      "epoch": 80.12,
      "learning_rate": 9.317281553398059e-06,
      "loss": 0.0108,
      "step": 206310
    },
    {
      "epoch": 80.12,
      "learning_rate": 9.316763754045309e-06,
      "loss": 0.0472,
      "step": 206320
    },
    {
      "epoch": 80.13,
      "learning_rate": 9.316245954692558e-06,
      "loss": 0.1397,
      "step": 206330
    },
    {
      "epoch": 80.13,
      "learning_rate": 9.315728155339808e-06,
      "loss": 0.036,
      "step": 206340
    },
    {
      "epoch": 80.14,
      "learning_rate": 9.315210355987056e-06,
      "loss": 0.0259,
      "step": 206350
    },
    {
      "epoch": 80.14,
      "learning_rate": 9.314692556634305e-06,
      "loss": 0.0474,
      "step": 206360
    },
    {
      "epoch": 80.14,
      "learning_rate": 9.314174757281555e-06,
      "loss": 0.0389,
      "step": 206370
    },
    {
      "epoch": 80.15,
      "learning_rate": 9.313656957928805e-06,
      "loss": 0.0001,
      "step": 206380
    },
    {
      "epoch": 80.15,
      "learning_rate": 9.313139158576052e-06,
      "loss": 0.0247,
      "step": 206390
    },
    {
      "epoch": 80.16,
      "learning_rate": 9.312621359223302e-06,
      "loss": 0.0372,
      "step": 206400
    },
    {
      "epoch": 80.16,
      "learning_rate": 9.312103559870552e-06,
      "loss": 0.0001,
      "step": 206410
    },
    {
      "epoch": 80.16,
      "learning_rate": 9.311585760517801e-06,
      "loss": 0.0214,
      "step": 206420
    },
    {
      "epoch": 80.17,
      "learning_rate": 9.31106796116505e-06,
      "loss": 0.0004,
      "step": 206430
    },
    {
      "epoch": 80.17,
      "learning_rate": 9.310550161812299e-06,
      "loss": 0.0835,
      "step": 206440
    },
    {
      "epoch": 80.17,
      "learning_rate": 9.310032362459548e-06,
      "loss": 0.0956,
      "step": 206450
    },
    {
      "epoch": 80.18,
      "learning_rate": 9.309514563106798e-06,
      "loss": 0.0154,
      "step": 206460
    },
    {
      "epoch": 80.18,
      "learning_rate": 9.308996763754046e-06,
      "loss": 0.0262,
      "step": 206470
    },
    {
      "epoch": 80.19,
      "learning_rate": 9.308478964401296e-06,
      "loss": 0.0639,
      "step": 206480
    },
    {
      "epoch": 80.19,
      "learning_rate": 9.307961165048545e-06,
      "loss": 0.0122,
      "step": 206490
    },
    {
      "epoch": 80.19,
      "learning_rate": 9.307443365695795e-06,
      "loss": 0.0215,
      "step": 206500
    },
    {
      "epoch": 80.2,
      "learning_rate": 9.306925566343043e-06,
      "loss": 0.0119,
      "step": 206510
    },
    {
      "epoch": 80.2,
      "learning_rate": 9.306407766990292e-06,
      "loss": 0.0823,
      "step": 206520
    },
    {
      "epoch": 80.21,
      "learning_rate": 9.305889967637542e-06,
      "loss": 0.0192,
      "step": 206530
    },
    {
      "epoch": 80.21,
      "learning_rate": 9.305372168284792e-06,
      "loss": 0.0344,
      "step": 206540
    },
    {
      "epoch": 80.21,
      "learning_rate": 9.30485436893204e-06,
      "loss": 0.0009,
      "step": 206550
    },
    {
      "epoch": 80.22,
      "learning_rate": 9.30433656957929e-06,
      "loss": 0.0768,
      "step": 206560
    },
    {
      "epoch": 80.22,
      "learning_rate": 9.303818770226539e-06,
      "loss": 0.0163,
      "step": 206570
    },
    {
      "epoch": 80.23,
      "learning_rate": 9.303300970873787e-06,
      "loss": 0.0747,
      "step": 206580
    },
    {
      "epoch": 80.23,
      "learning_rate": 9.302783171521036e-06,
      "loss": 0.0438,
      "step": 206590
    },
    {
      "epoch": 80.23,
      "learning_rate": 9.302265372168286e-06,
      "loss": 0.0003,
      "step": 206600
    },
    {
      "epoch": 80.24,
      "learning_rate": 9.301747572815536e-06,
      "loss": 0.0741,
      "step": 206610
    },
    {
      "epoch": 80.24,
      "learning_rate": 9.301229773462784e-06,
      "loss": 0.0271,
      "step": 206620
    },
    {
      "epoch": 80.24,
      "learning_rate": 9.300711974110033e-06,
      "loss": 0.0811,
      "step": 206630
    },
    {
      "epoch": 80.25,
      "learning_rate": 9.300194174757283e-06,
      "loss": 0.0252,
      "step": 206640
    },
    {
      "epoch": 80.25,
      "learning_rate": 9.299676375404532e-06,
      "loss": 0.0125,
      "step": 206650
    },
    {
      "epoch": 80.26,
      "learning_rate": 9.29915857605178e-06,
      "loss": 0.067,
      "step": 206660
    },
    {
      "epoch": 80.26,
      "learning_rate": 9.29864077669903e-06,
      "loss": 0.0631,
      "step": 206670
    },
    {
      "epoch": 80.26,
      "learning_rate": 9.29812297734628e-06,
      "loss": 0.0195,
      "step": 206680
    },
    {
      "epoch": 80.27,
      "learning_rate": 9.29760517799353e-06,
      "loss": 0.0368,
      "step": 206690
    },
    {
      "epoch": 80.27,
      "learning_rate": 9.297087378640777e-06,
      "loss": 0.0005,
      "step": 206700
    },
    {
      "epoch": 80.28,
      "learning_rate": 9.296569579288027e-06,
      "loss": 0.071,
      "step": 206710
    },
    {
      "epoch": 80.28,
      "learning_rate": 9.296051779935276e-06,
      "loss": 0.1663,
      "step": 206720
    },
    {
      "epoch": 80.28,
      "learning_rate": 9.295533980582524e-06,
      "loss": 0.105,
      "step": 206730
    },
    {
      "epoch": 80.29,
      "learning_rate": 9.295016181229774e-06,
      "loss": 0.0188,
      "step": 206740
    },
    {
      "epoch": 80.29,
      "learning_rate": 9.294498381877023e-06,
      "loss": 0.0592,
      "step": 206750
    },
    {
      "epoch": 80.3,
      "learning_rate": 9.293980582524273e-06,
      "loss": 0.0124,
      "step": 206760
    },
    {
      "epoch": 80.3,
      "learning_rate": 9.293462783171521e-06,
      "loss": 0.0478,
      "step": 206770
    },
    {
      "epoch": 80.3,
      "learning_rate": 9.29294498381877e-06,
      "loss": 0.133,
      "step": 206780
    },
    {
      "epoch": 80.31,
      "learning_rate": 9.29242718446602e-06,
      "loss": 0.0165,
      "step": 206790
    },
    {
      "epoch": 80.31,
      "learning_rate": 9.29190938511327e-06,
      "loss": 0.0259,
      "step": 206800
    },
    {
      "epoch": 80.31,
      "learning_rate": 9.291391585760518e-06,
      "loss": 0.0191,
      "step": 206810
    },
    {
      "epoch": 80.32,
      "learning_rate": 9.290873786407767e-06,
      "loss": 0.0486,
      "step": 206820
    },
    {
      "epoch": 80.32,
      "learning_rate": 9.290355987055017e-06,
      "loss": 0.0168,
      "step": 206830
    },
    {
      "epoch": 80.33,
      "learning_rate": 9.289838187702267e-06,
      "loss": 0.0132,
      "step": 206840
    },
    {
      "epoch": 80.33,
      "learning_rate": 9.289320388349515e-06,
      "loss": 0.0143,
      "step": 206850
    },
    {
      "epoch": 80.33,
      "learning_rate": 9.288802588996764e-06,
      "loss": 0.0639,
      "step": 206860
    },
    {
      "epoch": 80.34,
      "learning_rate": 9.288284789644014e-06,
      "loss": 0.1073,
      "step": 206870
    },
    {
      "epoch": 80.34,
      "learning_rate": 9.287766990291263e-06,
      "loss": 0.0424,
      "step": 206880
    },
    {
      "epoch": 80.35,
      "learning_rate": 9.287249190938511e-06,
      "loss": 0.0181,
      "step": 206890
    },
    {
      "epoch": 80.35,
      "learning_rate": 9.286731391585761e-06,
      "loss": 0.1596,
      "step": 206900
    },
    {
      "epoch": 80.35,
      "learning_rate": 9.28621359223301e-06,
      "loss": 0.0136,
      "step": 206910
    },
    {
      "epoch": 80.36,
      "learning_rate": 9.285695792880259e-06,
      "loss": 0.1907,
      "step": 206920
    },
    {
      "epoch": 80.36,
      "learning_rate": 9.285177993527508e-06,
      "loss": 0.1622,
      "step": 206930
    },
    {
      "epoch": 80.37,
      "learning_rate": 9.284660194174758e-06,
      "loss": 0.0721,
      "step": 206940
    },
    {
      "epoch": 80.37,
      "learning_rate": 9.284142394822007e-06,
      "loss": 0.04,
      "step": 206950
    },
    {
      "epoch": 80.37,
      "learning_rate": 9.283624595469255e-06,
      "loss": 0.0215,
      "step": 206960
    },
    {
      "epoch": 80.38,
      "learning_rate": 9.283106796116505e-06,
      "loss": 0.0804,
      "step": 206970
    },
    {
      "epoch": 80.38,
      "learning_rate": 9.282588996763755e-06,
      "loss": 0.149,
      "step": 206980
    },
    {
      "epoch": 80.38,
      "learning_rate": 9.282071197411004e-06,
      "loss": 0.2217,
      "step": 206990
    },
    {
      "epoch": 80.39,
      "learning_rate": 9.281553398058252e-06,
      "loss": 0.03,
      "step": 207000
    },
    {
      "epoch": 80.39,
      "learning_rate": 9.281035598705502e-06,
      "loss": 0.1564,
      "step": 207010
    },
    {
      "epoch": 80.4,
      "learning_rate": 9.280517799352751e-06,
      "loss": 0.0085,
      "step": 207020
    },
    {
      "epoch": 80.4,
      "learning_rate": 9.280000000000001e-06,
      "loss": 0.137,
      "step": 207030
    },
    {
      "epoch": 80.4,
      "learning_rate": 9.279482200647249e-06,
      "loss": 0.0545,
      "step": 207040
    },
    {
      "epoch": 80.41,
      "learning_rate": 9.278964401294498e-06,
      "loss": 0.0845,
      "step": 207050
    },
    {
      "epoch": 80.41,
      "learning_rate": 9.278446601941748e-06,
      "loss": 0.124,
      "step": 207060
    },
    {
      "epoch": 80.42,
      "learning_rate": 9.277928802588998e-06,
      "loss": 0.0863,
      "step": 207070
    },
    {
      "epoch": 80.42,
      "learning_rate": 9.277411003236246e-06,
      "loss": 0.0183,
      "step": 207080
    },
    {
      "epoch": 80.42,
      "learning_rate": 9.276893203883495e-06,
      "loss": 0.1179,
      "step": 207090
    },
    {
      "epoch": 80.43,
      "learning_rate": 9.276375404530745e-06,
      "loss": 0.0426,
      "step": 207100
    },
    {
      "epoch": 80.43,
      "learning_rate": 9.275857605177994e-06,
      "loss": 0.0661,
      "step": 207110
    },
    {
      "epoch": 80.43,
      "learning_rate": 9.275339805825242e-06,
      "loss": 0.0107,
      "step": 207120
    },
    {
      "epoch": 80.44,
      "learning_rate": 9.274822006472492e-06,
      "loss": 0.1471,
      "step": 207130
    },
    {
      "epoch": 80.44,
      "learning_rate": 9.274304207119742e-06,
      "loss": 0.0474,
      "step": 207140
    },
    {
      "epoch": 80.45,
      "learning_rate": 9.273786407766991e-06,
      "loss": 0.0199,
      "step": 207150
    },
    {
      "epoch": 80.45,
      "learning_rate": 9.27326860841424e-06,
      "loss": 0.0492,
      "step": 207160
    },
    {
      "epoch": 80.45,
      "learning_rate": 9.272750809061489e-06,
      "loss": 0.017,
      "step": 207170
    },
    {
      "epoch": 80.46,
      "learning_rate": 9.272233009708738e-06,
      "loss": 0.0448,
      "step": 207180
    },
    {
      "epoch": 80.46,
      "learning_rate": 9.271715210355988e-06,
      "loss": 0.0566,
      "step": 207190
    },
    {
      "epoch": 80.47,
      "learning_rate": 9.271197411003236e-06,
      "loss": 0.0049,
      "step": 207200
    },
    {
      "epoch": 80.47,
      "learning_rate": 9.270679611650486e-06,
      "loss": 0.0205,
      "step": 207210
    },
    {
      "epoch": 80.47,
      "learning_rate": 9.270161812297735e-06,
      "loss": 0.0346,
      "step": 207220
    },
    {
      "epoch": 80.48,
      "learning_rate": 9.269644012944985e-06,
      "loss": 0.0728,
      "step": 207230
    },
    {
      "epoch": 80.48,
      "learning_rate": 9.269126213592233e-06,
      "loss": 0.1606,
      "step": 207240
    },
    {
      "epoch": 80.49,
      "learning_rate": 9.268608414239482e-06,
      "loss": 0.0498,
      "step": 207250
    },
    {
      "epoch": 80.49,
      "learning_rate": 9.268090614886732e-06,
      "loss": 0.2175,
      "step": 207260
    },
    {
      "epoch": 80.49,
      "learning_rate": 9.267572815533982e-06,
      "loss": 0.0157,
      "step": 207270
    },
    {
      "epoch": 80.5,
      "learning_rate": 9.26705501618123e-06,
      "loss": 0.0827,
      "step": 207280
    },
    {
      "epoch": 80.5,
      "learning_rate": 9.266537216828479e-06,
      "loss": 0.0209,
      "step": 207290
    },
    {
      "epoch": 80.5,
      "learning_rate": 9.266019417475729e-06,
      "loss": 0.0094,
      "step": 207300
    },
    {
      "epoch": 80.51,
      "learning_rate": 9.265501618122978e-06,
      "loss": 0.0663,
      "step": 207310
    },
    {
      "epoch": 80.51,
      "learning_rate": 9.264983818770226e-06,
      "loss": 0.0755,
      "step": 207320
    },
    {
      "epoch": 80.52,
      "learning_rate": 9.264466019417476e-06,
      "loss": 0.0646,
      "step": 207330
    },
    {
      "epoch": 80.52,
      "learning_rate": 9.263948220064726e-06,
      "loss": 0.0281,
      "step": 207340
    },
    {
      "epoch": 80.52,
      "learning_rate": 9.263430420711975e-06,
      "loss": 0.0113,
      "step": 207350
    },
    {
      "epoch": 80.53,
      "learning_rate": 9.262912621359223e-06,
      "loss": 0.0719,
      "step": 207360
    },
    {
      "epoch": 80.53,
      "learning_rate": 9.262394822006473e-06,
      "loss": 0.0665,
      "step": 207370
    },
    {
      "epoch": 80.54,
      "learning_rate": 9.261877022653722e-06,
      "loss": 0.06,
      "step": 207380
    },
    {
      "epoch": 80.54,
      "learning_rate": 9.261359223300972e-06,
      "loss": 0.0889,
      "step": 207390
    },
    {
      "epoch": 80.54,
      "learning_rate": 9.26084142394822e-06,
      "loss": 0.0839,
      "step": 207400
    },
    {
      "epoch": 80.55,
      "learning_rate": 9.26032362459547e-06,
      "loss": 0.1781,
      "step": 207410
    },
    {
      "epoch": 80.55,
      "learning_rate": 9.259805825242719e-06,
      "loss": 0.1102,
      "step": 207420
    },
    {
      "epoch": 80.56,
      "learning_rate": 9.259288025889969e-06,
      "loss": 0.0007,
      "step": 207430
    },
    {
      "epoch": 80.56,
      "learning_rate": 9.258770226537218e-06,
      "loss": 0.0094,
      "step": 207440
    },
    {
      "epoch": 80.56,
      "learning_rate": 9.258252427184466e-06,
      "loss": 0.0667,
      "step": 207450
    },
    {
      "epoch": 80.57,
      "learning_rate": 9.257734627831716e-06,
      "loss": 0.0471,
      "step": 207460
    },
    {
      "epoch": 80.57,
      "learning_rate": 9.257216828478965e-06,
      "loss": 0.0168,
      "step": 207470
    },
    {
      "epoch": 80.57,
      "learning_rate": 9.256699029126215e-06,
      "loss": 0.0364,
      "step": 207480
    },
    {
      "epoch": 80.58,
      "learning_rate": 9.256181229773463e-06,
      "loss": 0.0962,
      "step": 207490
    },
    {
      "epoch": 80.58,
      "learning_rate": 9.255663430420713e-06,
      "loss": 0.0302,
      "step": 207500
    },
    {
      "epoch": 80.59,
      "learning_rate": 9.255145631067962e-06,
      "loss": 0.096,
      "step": 207510
    },
    {
      "epoch": 80.59,
      "learning_rate": 9.254627831715212e-06,
      "loss": 0.153,
      "step": 207520
    },
    {
      "epoch": 80.59,
      "learning_rate": 9.25411003236246e-06,
      "loss": 0.0833,
      "step": 207530
    },
    {
      "epoch": 80.6,
      "learning_rate": 9.25359223300971e-06,
      "loss": 0.0342,
      "step": 207540
    },
    {
      "epoch": 80.6,
      "learning_rate": 9.253074433656959e-06,
      "loss": 0.0393,
      "step": 207550
    },
    {
      "epoch": 80.61,
      "learning_rate": 9.252556634304209e-06,
      "loss": 0.0266,
      "step": 207560
    },
    {
      "epoch": 80.61,
      "learning_rate": 9.252038834951457e-06,
      "loss": 0.0884,
      "step": 207570
    },
    {
      "epoch": 80.61,
      "learning_rate": 9.251521035598706e-06,
      "loss": 0.1696,
      "step": 207580
    },
    {
      "epoch": 80.62,
      "learning_rate": 9.251003236245956e-06,
      "loss": 0.0737,
      "step": 207590
    },
    {
      "epoch": 80.62,
      "learning_rate": 9.250485436893205e-06,
      "loss": 0.0293,
      "step": 207600
    },
    {
      "epoch": 80.63,
      "learning_rate": 9.249967637540453e-06,
      "loss": 0.0542,
      "step": 207610
    },
    {
      "epoch": 80.63,
      "learning_rate": 9.249449838187703e-06,
      "loss": 0.1616,
      "step": 207620
    },
    {
      "epoch": 80.63,
      "learning_rate": 9.248932038834953e-06,
      "loss": 0.1097,
      "step": 207630
    },
    {
      "epoch": 80.64,
      "learning_rate": 9.248414239482202e-06,
      "loss": 0.1761,
      "step": 207640
    },
    {
      "epoch": 80.64,
      "learning_rate": 9.24789644012945e-06,
      "loss": 0.0247,
      "step": 207650
    },
    {
      "epoch": 80.64,
      "learning_rate": 9.2473786407767e-06,
      "loss": 0.0522,
      "step": 207660
    },
    {
      "epoch": 80.65,
      "learning_rate": 9.24686084142395e-06,
      "loss": 0.0081,
      "step": 207670
    },
    {
      "epoch": 80.65,
      "learning_rate": 9.246343042071199e-06,
      "loss": 0.082,
      "step": 207680
    },
    {
      "epoch": 80.66,
      "learning_rate": 9.245825242718447e-06,
      "loss": 0.1285,
      "step": 207690
    },
    {
      "epoch": 80.66,
      "learning_rate": 9.245307443365697e-06,
      "loss": 0.0021,
      "step": 207700
    },
    {
      "epoch": 80.66,
      "learning_rate": 9.244789644012946e-06,
      "loss": 0.0772,
      "step": 207710
    },
    {
      "epoch": 80.67,
      "learning_rate": 9.244271844660196e-06,
      "loss": 0.0161,
      "step": 207720
    },
    {
      "epoch": 80.67,
      "learning_rate": 9.243754045307444e-06,
      "loss": 0.0839,
      "step": 207730
    },
    {
      "epoch": 80.68,
      "learning_rate": 9.243236245954693e-06,
      "loss": 0.0367,
      "step": 207740
    },
    {
      "epoch": 80.68,
      "learning_rate": 9.242718446601943e-06,
      "loss": 0.0133,
      "step": 207750
    },
    {
      "epoch": 80.68,
      "learning_rate": 9.242200647249193e-06,
      "loss": 0.073,
      "step": 207760
    },
    {
      "epoch": 80.69,
      "learning_rate": 9.24168284789644e-06,
      "loss": 0.1004,
      "step": 207770
    },
    {
      "epoch": 80.69,
      "learning_rate": 9.24116504854369e-06,
      "loss": 0.02,
      "step": 207780
    },
    {
      "epoch": 80.7,
      "learning_rate": 9.24064724919094e-06,
      "loss": 0.0353,
      "step": 207790
    },
    {
      "epoch": 80.7,
      "learning_rate": 9.24012944983819e-06,
      "loss": 0.0485,
      "step": 207800
    },
    {
      "epoch": 80.7,
      "learning_rate": 9.239611650485437e-06,
      "loss": 0.1412,
      "step": 207810
    },
    {
      "epoch": 80.71,
      "learning_rate": 9.239093851132687e-06,
      "loss": 0.1189,
      "step": 207820
    },
    {
      "epoch": 80.71,
      "learning_rate": 9.238576051779936e-06,
      "loss": 0.1168,
      "step": 207830
    },
    {
      "epoch": 80.71,
      "learning_rate": 9.238058252427186e-06,
      "loss": 0.0429,
      "step": 207840
    },
    {
      "epoch": 80.72,
      "learning_rate": 9.237540453074434e-06,
      "loss": 0.0115,
      "step": 207850
    },
    {
      "epoch": 80.72,
      "learning_rate": 9.237022653721684e-06,
      "loss": 0.015,
      "step": 207860
    },
    {
      "epoch": 80.73,
      "learning_rate": 9.236504854368933e-06,
      "loss": 0.0372,
      "step": 207870
    },
    {
      "epoch": 80.73,
      "learning_rate": 9.235987055016183e-06,
      "loss": 0.0676,
      "step": 207880
    },
    {
      "epoch": 80.73,
      "learning_rate": 9.23546925566343e-06,
      "loss": 0.0002,
      "step": 207890
    },
    {
      "epoch": 80.74,
      "learning_rate": 9.23495145631068e-06,
      "loss": 0.0874,
      "step": 207900
    },
    {
      "epoch": 80.74,
      "learning_rate": 9.23443365695793e-06,
      "loss": 0.1168,
      "step": 207910
    },
    {
      "epoch": 80.75,
      "learning_rate": 9.23391585760518e-06,
      "loss": 0.0601,
      "step": 207920
    },
    {
      "epoch": 80.75,
      "learning_rate": 9.233398058252428e-06,
      "loss": 0.021,
      "step": 207930
    },
    {
      "epoch": 80.75,
      "learning_rate": 9.232880258899677e-06,
      "loss": 0.0731,
      "step": 207940
    },
    {
      "epoch": 80.76,
      "learning_rate": 9.232362459546927e-06,
      "loss": 0.0869,
      "step": 207950
    },
    {
      "epoch": 80.76,
      "learning_rate": 9.231844660194176e-06,
      "loss": 0.0092,
      "step": 207960
    },
    {
      "epoch": 80.77,
      "learning_rate": 9.231326860841424e-06,
      "loss": 0.0391,
      "step": 207970
    },
    {
      "epoch": 80.77,
      "learning_rate": 9.230809061488674e-06,
      "loss": 0.0067,
      "step": 207980
    },
    {
      "epoch": 80.77,
      "learning_rate": 9.230291262135924e-06,
      "loss": 0.0004,
      "step": 207990
    },
    {
      "epoch": 80.78,
      "learning_rate": 9.229773462783173e-06,
      "loss": 0.0276,
      "step": 208000
    },
    {
      "epoch": 80.78,
      "learning_rate": 9.229255663430423e-06,
      "loss": 0.0314,
      "step": 208010
    },
    {
      "epoch": 80.78,
      "learning_rate": 9.22873786407767e-06,
      "loss": 0.0753,
      "step": 208020
    },
    {
      "epoch": 80.79,
      "learning_rate": 9.22822006472492e-06,
      "loss": 0.1325,
      "step": 208030
    },
    {
      "epoch": 80.79,
      "learning_rate": 9.22770226537217e-06,
      "loss": 0.1429,
      "step": 208040
    },
    {
      "epoch": 80.8,
      "learning_rate": 9.227184466019418e-06,
      "loss": 0.0194,
      "step": 208050
    },
    {
      "epoch": 80.8,
      "learning_rate": 9.226666666666668e-06,
      "loss": 0.1394,
      "step": 208060
    },
    {
      "epoch": 80.8,
      "learning_rate": 9.226148867313917e-06,
      "loss": 0.044,
      "step": 208070
    },
    {
      "epoch": 80.81,
      "learning_rate": 9.225631067961167e-06,
      "loss": 0.0592,
      "step": 208080
    },
    {
      "epoch": 80.81,
      "learning_rate": 9.225113268608415e-06,
      "loss": 0.1318,
      "step": 208090
    },
    {
      "epoch": 80.82,
      "learning_rate": 9.224595469255664e-06,
      "loss": 0.0746,
      "step": 208100
    },
    {
      "epoch": 80.82,
      "learning_rate": 9.224077669902914e-06,
      "loss": 0.012,
      "step": 208110
    },
    {
      "epoch": 80.82,
      "learning_rate": 9.223559870550164e-06,
      "loss": 0.0265,
      "step": 208120
    },
    {
      "epoch": 80.83,
      "learning_rate": 9.223042071197411e-06,
      "loss": 0.0002,
      "step": 208130
    },
    {
      "epoch": 80.83,
      "learning_rate": 9.222524271844661e-06,
      "loss": 0.0733,
      "step": 208140
    },
    {
      "epoch": 80.83,
      "learning_rate": 9.22200647249191e-06,
      "loss": 0.0761,
      "step": 208150
    },
    {
      "epoch": 80.84,
      "learning_rate": 9.22148867313916e-06,
      "loss": 0.0024,
      "step": 208160
    },
    {
      "epoch": 80.84,
      "learning_rate": 9.220970873786408e-06,
      "loss": 0.0169,
      "step": 208170
    },
    {
      "epoch": 80.85,
      "learning_rate": 9.220453074433658e-06,
      "loss": 0.0029,
      "step": 208180
    },
    {
      "epoch": 80.85,
      "learning_rate": 9.219935275080908e-06,
      "loss": 0.1233,
      "step": 208190
    },
    {
      "epoch": 80.85,
      "learning_rate": 9.219417475728155e-06,
      "loss": 0.0875,
      "step": 208200
    },
    {
      "epoch": 80.86,
      "learning_rate": 9.218899676375405e-06,
      "loss": 0.0253,
      "step": 208210
    },
    {
      "epoch": 80.86,
      "learning_rate": 9.218381877022655e-06,
      "loss": 0.0405,
      "step": 208220
    },
    {
      "epoch": 80.87,
      "learning_rate": 9.217864077669904e-06,
      "loss": 0.0819,
      "step": 208230
    },
    {
      "epoch": 80.87,
      "learning_rate": 9.217346278317152e-06,
      "loss": 0.0119,
      "step": 208240
    },
    {
      "epoch": 80.87,
      "learning_rate": 9.216828478964402e-06,
      "loss": 0.0319,
      "step": 208250
    },
    {
      "epoch": 80.88,
      "learning_rate": 9.216310679611651e-06,
      "loss": 0.1023,
      "step": 208260
    },
    {
      "epoch": 80.88,
      "learning_rate": 9.215792880258901e-06,
      "loss": 0.0343,
      "step": 208270
    },
    {
      "epoch": 80.89,
      "learning_rate": 9.215275080906149e-06,
      "loss": 0.0745,
      "step": 208280
    },
    {
      "epoch": 80.89,
      "learning_rate": 9.214757281553399e-06,
      "loss": 0.0082,
      "step": 208290
    },
    {
      "epoch": 80.89,
      "learning_rate": 9.214239482200648e-06,
      "loss": 0.092,
      "step": 208300
    },
    {
      "epoch": 80.9,
      "learning_rate": 9.213721682847898e-06,
      "loss": 0.0317,
      "step": 208310
    },
    {
      "epoch": 80.9,
      "learning_rate": 9.213203883495146e-06,
      "loss": 0.0504,
      "step": 208320
    },
    {
      "epoch": 80.9,
      "learning_rate": 9.212686084142395e-06,
      "loss": 0.0709,
      "step": 208330
    },
    {
      "epoch": 80.91,
      "learning_rate": 9.212168284789645e-06,
      "loss": 0.0015,
      "step": 208340
    },
    {
      "epoch": 80.91,
      "learning_rate": 9.211650485436895e-06,
      "loss": 0.048,
      "step": 208350
    },
    {
      "epoch": 80.92,
      "learning_rate": 9.211132686084143e-06,
      "loss": 0.0651,
      "step": 208360
    },
    {
      "epoch": 80.92,
      "learning_rate": 9.210614886731392e-06,
      "loss": 0.0243,
      "step": 208370
    },
    {
      "epoch": 80.92,
      "learning_rate": 9.210097087378642e-06,
      "loss": 0.0051,
      "step": 208380
    },
    {
      "epoch": 80.93,
      "learning_rate": 9.20957928802589e-06,
      "loss": 0.0253,
      "step": 208390
    },
    {
      "epoch": 80.93,
      "learning_rate": 9.20906148867314e-06,
      "loss": 0.0173,
      "step": 208400
    },
    {
      "epoch": 80.94,
      "learning_rate": 9.208543689320389e-06,
      "loss": 0.1477,
      "step": 208410
    },
    {
      "epoch": 80.94,
      "learning_rate": 9.208025889967639e-06,
      "loss": 0.0066,
      "step": 208420
    },
    {
      "epoch": 80.94,
      "learning_rate": 9.207508090614886e-06,
      "loss": 0.1031,
      "step": 208430
    },
    {
      "epoch": 80.95,
      "learning_rate": 9.206990291262136e-06,
      "loss": 0.1412,
      "step": 208440
    },
    {
      "epoch": 80.95,
      "learning_rate": 9.206472491909386e-06,
      "loss": 0.042,
      "step": 208450
    },
    {
      "epoch": 80.96,
      "learning_rate": 9.205954692556635e-06,
      "loss": 0.078,
      "step": 208460
    },
    {
      "epoch": 80.96,
      "learning_rate": 9.205436893203883e-06,
      "loss": 0.177,
      "step": 208470
    },
    {
      "epoch": 80.96,
      "learning_rate": 9.204919093851133e-06,
      "loss": 0.0672,
      "step": 208480
    },
    {
      "epoch": 80.97,
      "learning_rate": 9.204401294498382e-06,
      "loss": 0.0305,
      "step": 208490
    },
    {
      "epoch": 80.97,
      "learning_rate": 9.203883495145632e-06,
      "loss": 0.0087,
      "step": 208500
    },
    {
      "epoch": 80.97,
      "learning_rate": 9.20336569579288e-06,
      "loss": 0.0501,
      "step": 208510
    },
    {
      "epoch": 80.98,
      "learning_rate": 9.20284789644013e-06,
      "loss": 0.036,
      "step": 208520
    },
    {
      "epoch": 80.98,
      "learning_rate": 9.20233009708738e-06,
      "loss": 0.0439,
      "step": 208530
    },
    {
      "epoch": 80.99,
      "learning_rate": 9.201812297734627e-06,
      "loss": 0.0631,
      "step": 208540
    },
    {
      "epoch": 80.99,
      "learning_rate": 9.201294498381877e-06,
      "loss": 0.0006,
      "step": 208550
    },
    {
      "epoch": 80.99,
      "learning_rate": 9.200776699029126e-06,
      "loss": 0.2947,
      "step": 208560
    },
    {
      "epoch": 81.0,
      "learning_rate": 9.200258899676376e-06,
      "loss": 0.0388,
      "step": 208570
    },
    {
      "epoch": 81.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.34262311458587646,
      "eval_runtime": 8.2561,
      "eval_samples_per_second": 440.282,
      "eval_steps_per_second": 55.111,
      "step": 208575
    },
    {
      "epoch": 81.0,
      "learning_rate": 9.199741100323626e-06,
      "loss": 0.0242,
      "step": 208580
    },
    {
      "epoch": 81.01,
      "learning_rate": 9.199223300970874e-06,
      "loss": 0.0981,
      "step": 208590
    },
    {
      "epoch": 81.01,
      "learning_rate": 9.198705501618123e-06,
      "loss": 0.0065,
      "step": 208600
    },
    {
      "epoch": 81.01,
      "learning_rate": 9.198187702265373e-06,
      "loss": 0.0776,
      "step": 208610
    },
    {
      "epoch": 81.02,
      "learning_rate": 9.197669902912622e-06,
      "loss": 0.0646,
      "step": 208620
    },
    {
      "epoch": 81.02,
      "learning_rate": 9.19715210355987e-06,
      "loss": 0.0574,
      "step": 208630
    },
    {
      "epoch": 81.03,
      "learning_rate": 9.19663430420712e-06,
      "loss": 0.1101,
      "step": 208640
    },
    {
      "epoch": 81.03,
      "learning_rate": 9.19611650485437e-06,
      "loss": 0.0355,
      "step": 208650
    },
    {
      "epoch": 81.03,
      "learning_rate": 9.19559870550162e-06,
      "loss": 0.0132,
      "step": 208660
    },
    {
      "epoch": 81.04,
      "learning_rate": 9.195080906148867e-06,
      "loss": 0.02,
      "step": 208670
    },
    {
      "epoch": 81.04,
      "learning_rate": 9.194563106796117e-06,
      "loss": 0.0384,
      "step": 208680
    },
    {
      "epoch": 81.04,
      "learning_rate": 9.194045307443366e-06,
      "loss": 0.0798,
      "step": 208690
    },
    {
      "epoch": 81.05,
      "learning_rate": 9.193527508090616e-06,
      "loss": 0.0205,
      "step": 208700
    },
    {
      "epoch": 81.05,
      "learning_rate": 9.193009708737864e-06,
      "loss": 0.0153,
      "step": 208710
    },
    {
      "epoch": 81.06,
      "learning_rate": 9.192491909385114e-06,
      "loss": 0.0209,
      "step": 208720
    },
    {
      "epoch": 81.06,
      "learning_rate": 9.191974110032363e-06,
      "loss": 0.1864,
      "step": 208730
    },
    {
      "epoch": 81.06,
      "learning_rate": 9.191456310679613e-06,
      "loss": 0.1199,
      "step": 208740
    },
    {
      "epoch": 81.07,
      "learning_rate": 9.19093851132686e-06,
      "loss": 0.0086,
      "step": 208750
    },
    {
      "epoch": 81.07,
      "learning_rate": 9.19042071197411e-06,
      "loss": 0.0301,
      "step": 208760
    },
    {
      "epoch": 81.08,
      "learning_rate": 9.18990291262136e-06,
      "loss": 0.0916,
      "step": 208770
    },
    {
      "epoch": 81.08,
      "learning_rate": 9.18938511326861e-06,
      "loss": 0.0906,
      "step": 208780
    },
    {
      "epoch": 81.08,
      "learning_rate": 9.188867313915857e-06,
      "loss": 0.1145,
      "step": 208790
    },
    {
      "epoch": 81.09,
      "learning_rate": 9.188349514563107e-06,
      "loss": 0.0149,
      "step": 208800
    },
    {
      "epoch": 81.09,
      "learning_rate": 9.187831715210357e-06,
      "loss": 0.0699,
      "step": 208810
    },
    {
      "epoch": 81.1,
      "learning_rate": 9.187313915857606e-06,
      "loss": 0.0238,
      "step": 208820
    },
    {
      "epoch": 81.1,
      "learning_rate": 9.186796116504854e-06,
      "loss": 0.0318,
      "step": 208830
    },
    {
      "epoch": 81.1,
      "learning_rate": 9.186278317152104e-06,
      "loss": 0.0179,
      "step": 208840
    },
    {
      "epoch": 81.11,
      "learning_rate": 9.185760517799353e-06,
      "loss": 0.1295,
      "step": 208850
    },
    {
      "epoch": 81.11,
      "learning_rate": 9.185242718446603e-06,
      "loss": 0.0097,
      "step": 208860
    },
    {
      "epoch": 81.11,
      "learning_rate": 9.184724919093851e-06,
      "loss": 0.0127,
      "step": 208870
    },
    {
      "epoch": 81.12,
      "learning_rate": 9.1842071197411e-06,
      "loss": 0.0056,
      "step": 208880
    },
    {
      "epoch": 81.12,
      "learning_rate": 9.18368932038835e-06,
      "loss": 0.1129,
      "step": 208890
    },
    {
      "epoch": 81.13,
      "learning_rate": 9.1831715210356e-06,
      "loss": 0.0577,
      "step": 208900
    },
    {
      "epoch": 81.13,
      "learning_rate": 9.182653721682848e-06,
      "loss": 0.1776,
      "step": 208910
    },
    {
      "epoch": 81.13,
      "learning_rate": 9.182135922330097e-06,
      "loss": 0.0275,
      "step": 208920
    },
    {
      "epoch": 81.14,
      "learning_rate": 9.181618122977347e-06,
      "loss": 0.0813,
      "step": 208930
    },
    {
      "epoch": 81.14,
      "learning_rate": 9.181100323624597e-06,
      "loss": 0.0009,
      "step": 208940
    },
    {
      "epoch": 81.15,
      "learning_rate": 9.180582524271845e-06,
      "loss": 0.0358,
      "step": 208950
    },
    {
      "epoch": 81.15,
      "learning_rate": 9.180064724919094e-06,
      "loss": 0.0269,
      "step": 208960
    },
    {
      "epoch": 81.15,
      "learning_rate": 9.179546925566344e-06,
      "loss": 0.1834,
      "step": 208970
    },
    {
      "epoch": 81.16,
      "learning_rate": 9.179029126213593e-06,
      "loss": 0.0799,
      "step": 208980
    },
    {
      "epoch": 81.16,
      "learning_rate": 9.178511326860841e-06,
      "loss": 0.0396,
      "step": 208990
    },
    {
      "epoch": 81.17,
      "learning_rate": 9.177993527508091e-06,
      "loss": 0.0719,
      "step": 209000
    },
    {
      "epoch": 81.17,
      "learning_rate": 9.17747572815534e-06,
      "loss": 0.0841,
      "step": 209010
    },
    {
      "epoch": 81.17,
      "learning_rate": 9.17695792880259e-06,
      "loss": 0.1213,
      "step": 209020
    },
    {
      "epoch": 81.18,
      "learning_rate": 9.176440129449838e-06,
      "loss": 0.1153,
      "step": 209030
    },
    {
      "epoch": 81.18,
      "learning_rate": 9.175922330097088e-06,
      "loss": 0.1099,
      "step": 209040
    },
    {
      "epoch": 81.18,
      "learning_rate": 9.175404530744337e-06,
      "loss": 0.0229,
      "step": 209050
    },
    {
      "epoch": 81.19,
      "learning_rate": 9.174886731391587e-06,
      "loss": 0.0355,
      "step": 209060
    },
    {
      "epoch": 81.19,
      "learning_rate": 9.174368932038835e-06,
      "loss": 0.1444,
      "step": 209070
    },
    {
      "epoch": 81.2,
      "learning_rate": 9.173851132686085e-06,
      "loss": 0.0092,
      "step": 209080
    },
    {
      "epoch": 81.2,
      "learning_rate": 9.173333333333334e-06,
      "loss": 0.1225,
      "step": 209090
    },
    {
      "epoch": 81.2,
      "learning_rate": 9.172815533980584e-06,
      "loss": 0.0139,
      "step": 209100
    },
    {
      "epoch": 81.21,
      "learning_rate": 9.172297734627833e-06,
      "loss": 0.0443,
      "step": 209110
    },
    {
      "epoch": 81.21,
      "learning_rate": 9.171779935275081e-06,
      "loss": 0.0218,
      "step": 209120
    },
    {
      "epoch": 81.22,
      "learning_rate": 9.171262135922331e-06,
      "loss": 0.0003,
      "step": 209130
    },
    {
      "epoch": 81.22,
      "learning_rate": 9.17074433656958e-06,
      "loss": 0.0958,
      "step": 209140
    },
    {
      "epoch": 81.22,
      "learning_rate": 9.17022653721683e-06,
      "loss": 0.0502,
      "step": 209150
    },
    {
      "epoch": 81.23,
      "learning_rate": 9.169708737864078e-06,
      "loss": 0.0535,
      "step": 209160
    },
    {
      "epoch": 81.23,
      "learning_rate": 9.169190938511328e-06,
      "loss": 0.0429,
      "step": 209170
    },
    {
      "epoch": 81.23,
      "learning_rate": 9.168673139158577e-06,
      "loss": 0.0937,
      "step": 209180
    },
    {
      "epoch": 81.24,
      "learning_rate": 9.168155339805827e-06,
      "loss": 0.1716,
      "step": 209190
    },
    {
      "epoch": 81.24,
      "learning_rate": 9.167637540453075e-06,
      "loss": 0.0284,
      "step": 209200
    },
    {
      "epoch": 81.25,
      "learning_rate": 9.167119741100324e-06,
      "loss": 0.1401,
      "step": 209210
    },
    {
      "epoch": 81.25,
      "learning_rate": 9.166601941747574e-06,
      "loss": 0.0167,
      "step": 209220
    },
    {
      "epoch": 81.25,
      "learning_rate": 9.166084142394824e-06,
      "loss": 0.0184,
      "step": 209230
    },
    {
      "epoch": 81.26,
      "learning_rate": 9.165566343042072e-06,
      "loss": 0.0867,
      "step": 209240
    },
    {
      "epoch": 81.26,
      "learning_rate": 9.165048543689321e-06,
      "loss": 0.1053,
      "step": 209250
    },
    {
      "epoch": 81.27,
      "learning_rate": 9.164530744336571e-06,
      "loss": 0.0308,
      "step": 209260
    },
    {
      "epoch": 81.27,
      "learning_rate": 9.16401294498382e-06,
      "loss": 0.0209,
      "step": 209270
    },
    {
      "epoch": 81.27,
      "learning_rate": 9.163495145631068e-06,
      "loss": 0.1603,
      "step": 209280
    },
    {
      "epoch": 81.28,
      "learning_rate": 9.162977346278318e-06,
      "loss": 0.0241,
      "step": 209290
    },
    {
      "epoch": 81.28,
      "learning_rate": 9.162459546925568e-06,
      "loss": 0.0839,
      "step": 209300
    },
    {
      "epoch": 81.29,
      "learning_rate": 9.161941747572817e-06,
      "loss": 0.04,
      "step": 209310
    },
    {
      "epoch": 81.29,
      "learning_rate": 9.161423948220065e-06,
      "loss": 0.0767,
      "step": 209320
    },
    {
      "epoch": 81.29,
      "learning_rate": 9.160906148867315e-06,
      "loss": 0.0016,
      "step": 209330
    },
    {
      "epoch": 81.3,
      "learning_rate": 9.160388349514564e-06,
      "loss": 0.0458,
      "step": 209340
    },
    {
      "epoch": 81.3,
      "learning_rate": 9.159870550161814e-06,
      "loss": 0.0231,
      "step": 209350
    },
    {
      "epoch": 81.3,
      "learning_rate": 9.159352750809062e-06,
      "loss": 0.1401,
      "step": 209360
    },
    {
      "epoch": 81.31,
      "learning_rate": 9.158834951456312e-06,
      "loss": 0.047,
      "step": 209370
    },
    {
      "epoch": 81.31,
      "learning_rate": 9.158317152103561e-06,
      "loss": 0.0217,
      "step": 209380
    },
    {
      "epoch": 81.32,
      "learning_rate": 9.15779935275081e-06,
      "loss": 0.024,
      "step": 209390
    },
    {
      "epoch": 81.32,
      "learning_rate": 9.157281553398059e-06,
      "loss": 0.1227,
      "step": 209400
    },
    {
      "epoch": 81.32,
      "learning_rate": 9.156763754045308e-06,
      "loss": 0.0896,
      "step": 209410
    },
    {
      "epoch": 81.33,
      "learning_rate": 9.156245954692558e-06,
      "loss": 0.0342,
      "step": 209420
    },
    {
      "epoch": 81.33,
      "learning_rate": 9.155728155339808e-06,
      "loss": 0.0993,
      "step": 209430
    },
    {
      "epoch": 81.34,
      "learning_rate": 9.155210355987056e-06,
      "loss": 0.0824,
      "step": 209440
    },
    {
      "epoch": 81.34,
      "learning_rate": 9.154692556634305e-06,
      "loss": 0.0675,
      "step": 209450
    },
    {
      "epoch": 81.34,
      "learning_rate": 9.154174757281555e-06,
      "loss": 0.0412,
      "step": 209460
    },
    {
      "epoch": 81.35,
      "learning_rate": 9.153656957928804e-06,
      "loss": 0.0137,
      "step": 209470
    },
    {
      "epoch": 81.35,
      "learning_rate": 9.153139158576052e-06,
      "loss": 0.089,
      "step": 209480
    },
    {
      "epoch": 81.36,
      "learning_rate": 9.152621359223302e-06,
      "loss": 0.0012,
      "step": 209490
    },
    {
      "epoch": 81.36,
      "learning_rate": 9.152103559870552e-06,
      "loss": 0.0309,
      "step": 209500
    },
    {
      "epoch": 81.36,
      "learning_rate": 9.151585760517801e-06,
      "loss": 0.0457,
      "step": 209510
    },
    {
      "epoch": 81.37,
      "learning_rate": 9.151067961165049e-06,
      "loss": 0.047,
      "step": 209520
    },
    {
      "epoch": 81.37,
      "learning_rate": 9.150550161812299e-06,
      "loss": 0.1341,
      "step": 209530
    },
    {
      "epoch": 81.37,
      "learning_rate": 9.150032362459548e-06,
      "loss": 0.0799,
      "step": 209540
    },
    {
      "epoch": 81.38,
      "learning_rate": 9.149514563106798e-06,
      "loss": 0.0378,
      "step": 209550
    },
    {
      "epoch": 81.38,
      "learning_rate": 9.148996763754046e-06,
      "loss": 0.138,
      "step": 209560
    },
    {
      "epoch": 81.39,
      "learning_rate": 9.148478964401295e-06,
      "loss": 0.0597,
      "step": 209570
    },
    {
      "epoch": 81.39,
      "learning_rate": 9.147961165048545e-06,
      "loss": 0.0434,
      "step": 209580
    },
    {
      "epoch": 81.39,
      "learning_rate": 9.147443365695795e-06,
      "loss": 0.0991,
      "step": 209590
    },
    {
      "epoch": 81.4,
      "learning_rate": 9.146925566343043e-06,
      "loss": 0.0164,
      "step": 209600
    },
    {
      "epoch": 81.4,
      "learning_rate": 9.146407766990292e-06,
      "loss": 0.0872,
      "step": 209610
    },
    {
      "epoch": 81.41,
      "learning_rate": 9.145889967637542e-06,
      "loss": 0.0789,
      "step": 209620
    },
    {
      "epoch": 81.41,
      "learning_rate": 9.145372168284792e-06,
      "loss": 0.1037,
      "step": 209630
    },
    {
      "epoch": 81.41,
      "learning_rate": 9.14485436893204e-06,
      "loss": 0.0906,
      "step": 209640
    },
    {
      "epoch": 81.42,
      "learning_rate": 9.144336569579289e-06,
      "loss": 0.03,
      "step": 209650
    },
    {
      "epoch": 81.42,
      "learning_rate": 9.143818770226539e-06,
      "loss": 0.0242,
      "step": 209660
    },
    {
      "epoch": 81.43,
      "learning_rate": 9.143300970873787e-06,
      "loss": 0.0235,
      "step": 209670
    },
    {
      "epoch": 81.43,
      "learning_rate": 9.142783171521036e-06,
      "loss": 0.047,
      "step": 209680
    },
    {
      "epoch": 81.43,
      "learning_rate": 9.142265372168286e-06,
      "loss": 0.215,
      "step": 209690
    },
    {
      "epoch": 81.44,
      "learning_rate": 9.141747572815535e-06,
      "loss": 0.0194,
      "step": 209700
    },
    {
      "epoch": 81.44,
      "learning_rate": 9.141229773462783e-06,
      "loss": 0.0436,
      "step": 209710
    },
    {
      "epoch": 81.44,
      "learning_rate": 9.140711974110033e-06,
      "loss": 0.1308,
      "step": 209720
    },
    {
      "epoch": 81.45,
      "learning_rate": 9.140194174757283e-06,
      "loss": 0.0424,
      "step": 209730
    },
    {
      "epoch": 81.45,
      "learning_rate": 9.139676375404532e-06,
      "loss": 0.0585,
      "step": 209740
    },
    {
      "epoch": 81.46,
      "learning_rate": 9.13915857605178e-06,
      "loss": 0.028,
      "step": 209750
    },
    {
      "epoch": 81.46,
      "learning_rate": 9.13864077669903e-06,
      "loss": 0.0008,
      "step": 209760
    },
    {
      "epoch": 81.46,
      "learning_rate": 9.13812297734628e-06,
      "loss": 0.1022,
      "step": 209770
    },
    {
      "epoch": 81.47,
      "learning_rate": 9.137605177993529e-06,
      "loss": 0.0421,
      "step": 209780
    },
    {
      "epoch": 81.47,
      "learning_rate": 9.137087378640777e-06,
      "loss": 0.165,
      "step": 209790
    },
    {
      "epoch": 81.48,
      "learning_rate": 9.136569579288027e-06,
      "loss": 0.1437,
      "step": 209800
    },
    {
      "epoch": 81.48,
      "learning_rate": 9.136051779935276e-06,
      "loss": 0.0365,
      "step": 209810
    },
    {
      "epoch": 81.48,
      "learning_rate": 9.135533980582526e-06,
      "loss": 0.0082,
      "step": 209820
    },
    {
      "epoch": 81.49,
      "learning_rate": 9.135016181229774e-06,
      "loss": 0.0216,
      "step": 209830
    },
    {
      "epoch": 81.49,
      "learning_rate": 9.134498381877023e-06,
      "loss": 0.0339,
      "step": 209840
    },
    {
      "epoch": 81.5,
      "learning_rate": 9.133980582524273e-06,
      "loss": 0.0881,
      "step": 209850
    },
    {
      "epoch": 81.5,
      "learning_rate": 9.133462783171521e-06,
      "loss": 0.012,
      "step": 209860
    },
    {
      "epoch": 81.5,
      "learning_rate": 9.13294498381877e-06,
      "loss": 0.0115,
      "step": 209870
    },
    {
      "epoch": 81.51,
      "learning_rate": 9.13242718446602e-06,
      "loss": 0.046,
      "step": 209880
    },
    {
      "epoch": 81.51,
      "learning_rate": 9.13190938511327e-06,
      "loss": 0.1129,
      "step": 209890
    },
    {
      "epoch": 81.51,
      "learning_rate": 9.131391585760518e-06,
      "loss": 0.1034,
      "step": 209900
    },
    {
      "epoch": 81.52,
      "learning_rate": 9.130873786407767e-06,
      "loss": 0.0152,
      "step": 209910
    },
    {
      "epoch": 81.52,
      "learning_rate": 9.130355987055017e-06,
      "loss": 0.0189,
      "step": 209920
    },
    {
      "epoch": 81.53,
      "learning_rate": 9.129838187702267e-06,
      "loss": 0.0115,
      "step": 209930
    },
    {
      "epoch": 81.53,
      "learning_rate": 9.129320388349514e-06,
      "loss": 0.0454,
      "step": 209940
    },
    {
      "epoch": 81.53,
      "learning_rate": 9.128802588996764e-06,
      "loss": 0.1421,
      "step": 209950
    },
    {
      "epoch": 81.54,
      "learning_rate": 9.128284789644014e-06,
      "loss": 0.0506,
      "step": 209960
    },
    {
      "epoch": 81.54,
      "learning_rate": 9.127766990291263e-06,
      "loss": 0.1097,
      "step": 209970
    },
    {
      "epoch": 81.55,
      "learning_rate": 9.127249190938511e-06,
      "loss": 0.0647,
      "step": 209980
    },
    {
      "epoch": 81.55,
      "learning_rate": 9.12673139158576e-06,
      "loss": 0.0255,
      "step": 209990
    },
    {
      "epoch": 81.55,
      "learning_rate": 9.12621359223301e-06,
      "loss": 0.0072,
      "step": 210000
    },
    {
      "epoch": 81.56,
      "learning_rate": 9.125695792880258e-06,
      "loss": 0.0136,
      "step": 210010
    },
    {
      "epoch": 81.56,
      "learning_rate": 9.125177993527508e-06,
      "loss": 0.024,
      "step": 210020
    },
    {
      "epoch": 81.57,
      "learning_rate": 9.124660194174758e-06,
      "loss": 0.108,
      "step": 210030
    },
    {
      "epoch": 81.57,
      "learning_rate": 9.124142394822007e-06,
      "loss": 0.0754,
      "step": 210040
    },
    {
      "epoch": 81.57,
      "learning_rate": 9.123624595469255e-06,
      "loss": 0.0357,
      "step": 210050
    },
    {
      "epoch": 81.58,
      "learning_rate": 9.123106796116505e-06,
      "loss": 0.0921,
      "step": 210060
    },
    {
      "epoch": 81.58,
      "learning_rate": 9.122588996763754e-06,
      "loss": 0.0465,
      "step": 210070
    },
    {
      "epoch": 81.58,
      "learning_rate": 9.122071197411004e-06,
      "loss": 0.0185,
      "step": 210080
    },
    {
      "epoch": 81.59,
      "learning_rate": 9.121553398058252e-06,
      "loss": 0.0019,
      "step": 210090
    },
    {
      "epoch": 81.59,
      "learning_rate": 9.121035598705502e-06,
      "loss": 0.0135,
      "step": 210100
    },
    {
      "epoch": 81.6,
      "learning_rate": 9.120517799352751e-06,
      "loss": 0.1091,
      "step": 210110
    },
    {
      "epoch": 81.6,
      "learning_rate": 9.12e-06,
      "loss": 0.0403,
      "step": 210120
    },
    {
      "epoch": 81.6,
      "learning_rate": 9.119482200647249e-06,
      "loss": 0.005,
      "step": 210130
    },
    {
      "epoch": 81.61,
      "learning_rate": 9.118964401294498e-06,
      "loss": 0.0959,
      "step": 210140
    },
    {
      "epoch": 81.61,
      "learning_rate": 9.118446601941748e-06,
      "loss": 0.0873,
      "step": 210150
    },
    {
      "epoch": 81.62,
      "learning_rate": 9.117928802588998e-06,
      "loss": 0.087,
      "step": 210160
    },
    {
      "epoch": 81.62,
      "learning_rate": 9.117411003236245e-06,
      "loss": 0.1089,
      "step": 210170
    },
    {
      "epoch": 81.62,
      "learning_rate": 9.116893203883495e-06,
      "loss": 0.0338,
      "step": 210180
    },
    {
      "epoch": 81.63,
      "learning_rate": 9.116375404530745e-06,
      "loss": 0.0353,
      "step": 210190
    },
    {
      "epoch": 81.63,
      "learning_rate": 9.115857605177994e-06,
      "loss": 0.0346,
      "step": 210200
    },
    {
      "epoch": 81.63,
      "learning_rate": 9.115339805825242e-06,
      "loss": 0.0011,
      "step": 210210
    },
    {
      "epoch": 81.64,
      "learning_rate": 9.114822006472492e-06,
      "loss": 0.0371,
      "step": 210220
    },
    {
      "epoch": 81.64,
      "learning_rate": 9.114304207119741e-06,
      "loss": 0.0602,
      "step": 210230
    },
    {
      "epoch": 81.65,
      "learning_rate": 9.113786407766991e-06,
      "loss": 0.1011,
      "step": 210240
    },
    {
      "epoch": 81.65,
      "learning_rate": 9.11326860841424e-06,
      "loss": 0.0124,
      "step": 210250
    },
    {
      "epoch": 81.65,
      "learning_rate": 9.112750809061489e-06,
      "loss": 0.0372,
      "step": 210260
    },
    {
      "epoch": 81.66,
      "learning_rate": 9.112233009708738e-06,
      "loss": 0.01,
      "step": 210270
    },
    {
      "epoch": 81.66,
      "learning_rate": 9.111715210355988e-06,
      "loss": 0.0043,
      "step": 210280
    },
    {
      "epoch": 81.67,
      "learning_rate": 9.111197411003238e-06,
      "loss": 0.0772,
      "step": 210290
    },
    {
      "epoch": 81.67,
      "learning_rate": 9.110679611650485e-06,
      "loss": 0.1096,
      "step": 210300
    },
    {
      "epoch": 81.67,
      "learning_rate": 9.110161812297735e-06,
      "loss": 0.079,
      "step": 210310
    },
    {
      "epoch": 81.68,
      "learning_rate": 9.109644012944985e-06,
      "loss": 0.0135,
      "step": 210320
    },
    {
      "epoch": 81.68,
      "learning_rate": 9.109126213592234e-06,
      "loss": 0.0474,
      "step": 210330
    },
    {
      "epoch": 81.69,
      "learning_rate": 9.108608414239482e-06,
      "loss": 0.0581,
      "step": 210340
    },
    {
      "epoch": 81.69,
      "learning_rate": 9.108090614886732e-06,
      "loss": 0.0645,
      "step": 210350
    },
    {
      "epoch": 81.69,
      "learning_rate": 9.107572815533981e-06,
      "loss": 0.0177,
      "step": 210360
    },
    {
      "epoch": 81.7,
      "learning_rate": 9.107055016181231e-06,
      "loss": 0.1155,
      "step": 210370
    },
    {
      "epoch": 81.7,
      "learning_rate": 9.106537216828479e-06,
      "loss": 0.0374,
      "step": 210380
    },
    {
      "epoch": 81.7,
      "learning_rate": 9.106019417475729e-06,
      "loss": 0.0263,
      "step": 210390
    },
    {
      "epoch": 81.71,
      "learning_rate": 9.105501618122978e-06,
      "loss": 0.0069,
      "step": 210400
    },
    {
      "epoch": 81.71,
      "learning_rate": 9.104983818770228e-06,
      "loss": 0.0522,
      "step": 210410
    },
    {
      "epoch": 81.72,
      "learning_rate": 9.104466019417476e-06,
      "loss": 0.0276,
      "step": 210420
    },
    {
      "epoch": 81.72,
      "learning_rate": 9.103948220064725e-06,
      "loss": 0.0218,
      "step": 210430
    },
    {
      "epoch": 81.72,
      "learning_rate": 9.103430420711975e-06,
      "loss": 0.0411,
      "step": 210440
    },
    {
      "epoch": 81.73,
      "learning_rate": 9.102912621359225e-06,
      "loss": 0.002,
      "step": 210450
    },
    {
      "epoch": 81.73,
      "learning_rate": 9.102394822006473e-06,
      "loss": 0.1061,
      "step": 210460
    },
    {
      "epoch": 81.74,
      "learning_rate": 9.101877022653722e-06,
      "loss": 0.146,
      "step": 210470
    },
    {
      "epoch": 81.74,
      "learning_rate": 9.101359223300972e-06,
      "loss": 0.1651,
      "step": 210480
    },
    {
      "epoch": 81.74,
      "learning_rate": 9.100841423948221e-06,
      "loss": 0.1358,
      "step": 210490
    },
    {
      "epoch": 81.75,
      "learning_rate": 9.10032362459547e-06,
      "loss": 0.1671,
      "step": 210500
    },
    {
      "epoch": 81.75,
      "learning_rate": 9.099805825242719e-06,
      "loss": 0.1055,
      "step": 210510
    },
    {
      "epoch": 81.76,
      "learning_rate": 9.099288025889969e-06,
      "loss": 0.0268,
      "step": 210520
    },
    {
      "epoch": 81.76,
      "learning_rate": 9.098770226537218e-06,
      "loss": 0.0252,
      "step": 210530
    },
    {
      "epoch": 81.76,
      "learning_rate": 9.098252427184466e-06,
      "loss": 0.0029,
      "step": 210540
    },
    {
      "epoch": 81.77,
      "learning_rate": 9.097734627831716e-06,
      "loss": 0.0056,
      "step": 210550
    },
    {
      "epoch": 81.77,
      "learning_rate": 9.097216828478965e-06,
      "loss": 0.1349,
      "step": 210560
    },
    {
      "epoch": 81.77,
      "learning_rate": 9.096699029126215e-06,
      "loss": 0.1305,
      "step": 210570
    },
    {
      "epoch": 81.78,
      "learning_rate": 9.096181229773463e-06,
      "loss": 0.1104,
      "step": 210580
    },
    {
      "epoch": 81.78,
      "learning_rate": 9.095663430420712e-06,
      "loss": 0.1126,
      "step": 210590
    },
    {
      "epoch": 81.79,
      "learning_rate": 9.095145631067962e-06,
      "loss": 0.0817,
      "step": 210600
    },
    {
      "epoch": 81.79,
      "learning_rate": 9.094627831715212e-06,
      "loss": 0.0354,
      "step": 210610
    },
    {
      "epoch": 81.79,
      "learning_rate": 9.09411003236246e-06,
      "loss": 0.0133,
      "step": 210620
    },
    {
      "epoch": 81.8,
      "learning_rate": 9.09359223300971e-06,
      "loss": 0.0544,
      "step": 210630
    },
    {
      "epoch": 81.8,
      "learning_rate": 9.093074433656959e-06,
      "loss": 0.0407,
      "step": 210640
    },
    {
      "epoch": 81.81,
      "learning_rate": 9.092556634304209e-06,
      "loss": 0.0703,
      "step": 210650
    },
    {
      "epoch": 81.81,
      "learning_rate": 9.092038834951456e-06,
      "loss": 0.0258,
      "step": 210660
    },
    {
      "epoch": 81.81,
      "learning_rate": 9.091521035598706e-06,
      "loss": 0.0516,
      "step": 210670
    },
    {
      "epoch": 81.82,
      "learning_rate": 9.091003236245956e-06,
      "loss": 0.0009,
      "step": 210680
    },
    {
      "epoch": 81.82,
      "learning_rate": 9.090485436893205e-06,
      "loss": 0.0928,
      "step": 210690
    },
    {
      "epoch": 81.83,
      "learning_rate": 9.089967637540453e-06,
      "loss": 0.3581,
      "step": 210700
    },
    {
      "epoch": 81.83,
      "learning_rate": 9.089449838187703e-06,
      "loss": 0.0727,
      "step": 210710
    },
    {
      "epoch": 81.83,
      "learning_rate": 9.088932038834952e-06,
      "loss": 0.0618,
      "step": 210720
    },
    {
      "epoch": 81.84,
      "learning_rate": 9.088414239482202e-06,
      "loss": 0.1099,
      "step": 210730
    },
    {
      "epoch": 81.84,
      "learning_rate": 9.08789644012945e-06,
      "loss": 0.1867,
      "step": 210740
    },
    {
      "epoch": 81.84,
      "learning_rate": 9.0873786407767e-06,
      "loss": 0.1348,
      "step": 210750
    },
    {
      "epoch": 81.85,
      "learning_rate": 9.08686084142395e-06,
      "loss": 0.0294,
      "step": 210760
    },
    {
      "epoch": 81.85,
      "learning_rate": 9.086343042071199e-06,
      "loss": 0.0264,
      "step": 210770
    },
    {
      "epoch": 81.86,
      "learning_rate": 9.085825242718447e-06,
      "loss": 0.1793,
      "step": 210780
    },
    {
      "epoch": 81.86,
      "learning_rate": 9.085307443365696e-06,
      "loss": 0.1212,
      "step": 210790
    },
    {
      "epoch": 81.86,
      "learning_rate": 9.084789644012946e-06,
      "loss": 0.005,
      "step": 210800
    },
    {
      "epoch": 81.87,
      "learning_rate": 9.084271844660196e-06,
      "loss": 0.0122,
      "step": 210810
    },
    {
      "epoch": 81.87,
      "learning_rate": 9.083754045307445e-06,
      "loss": 0.0244,
      "step": 210820
    },
    {
      "epoch": 81.88,
      "learning_rate": 9.083236245954693e-06,
      "loss": 0.1112,
      "step": 210830
    },
    {
      "epoch": 81.88,
      "learning_rate": 9.082718446601943e-06,
      "loss": 0.1243,
      "step": 210840
    },
    {
      "epoch": 81.88,
      "learning_rate": 9.082200647249192e-06,
      "loss": 0.0241,
      "step": 210850
    },
    {
      "epoch": 81.89,
      "learning_rate": 9.081682847896442e-06,
      "loss": 0.027,
      "step": 210860
    },
    {
      "epoch": 81.89,
      "learning_rate": 9.08116504854369e-06,
      "loss": 0.0661,
      "step": 210870
    },
    {
      "epoch": 81.9,
      "learning_rate": 9.08064724919094e-06,
      "loss": 0.1182,
      "step": 210880
    },
    {
      "epoch": 81.9,
      "learning_rate": 9.08012944983819e-06,
      "loss": 0.0814,
      "step": 210890
    },
    {
      "epoch": 81.9,
      "learning_rate": 9.079611650485439e-06,
      "loss": 0.0671,
      "step": 210900
    },
    {
      "epoch": 81.91,
      "learning_rate": 9.079093851132687e-06,
      "loss": 0.0003,
      "step": 210910
    },
    {
      "epoch": 81.91,
      "learning_rate": 9.078576051779936e-06,
      "loss": 0.2047,
      "step": 210920
    },
    {
      "epoch": 81.91,
      "learning_rate": 9.078058252427186e-06,
      "loss": 0.0569,
      "step": 210930
    },
    {
      "epoch": 81.92,
      "learning_rate": 9.077540453074436e-06,
      "loss": 0.0635,
      "step": 210940
    },
    {
      "epoch": 81.92,
      "learning_rate": 9.077022653721683e-06,
      "loss": 0.0896,
      "step": 210950
    },
    {
      "epoch": 81.93,
      "learning_rate": 9.076504854368933e-06,
      "loss": 0.2268,
      "step": 210960
    },
    {
      "epoch": 81.93,
      "learning_rate": 9.075987055016183e-06,
      "loss": 0.0624,
      "step": 210970
    },
    {
      "epoch": 81.93,
      "learning_rate": 9.075469255663432e-06,
      "loss": 0.0749,
      "step": 210980
    },
    {
      "epoch": 81.94,
      "learning_rate": 9.07495145631068e-06,
      "loss": 0.072,
      "step": 210990
    },
    {
      "epoch": 81.94,
      "learning_rate": 9.07443365695793e-06,
      "loss": 0.0435,
      "step": 211000
    },
    {
      "epoch": 81.95,
      "learning_rate": 9.07391585760518e-06,
      "loss": 0.032,
      "step": 211010
    },
    {
      "epoch": 81.95,
      "learning_rate": 9.073398058252429e-06,
      "loss": 0.0183,
      "step": 211020
    },
    {
      "epoch": 81.95,
      "learning_rate": 9.072880258899677e-06,
      "loss": 0.0316,
      "step": 211030
    },
    {
      "epoch": 81.96,
      "learning_rate": 9.072362459546927e-06,
      "loss": 0.0575,
      "step": 211040
    },
    {
      "epoch": 81.96,
      "learning_rate": 9.071844660194176e-06,
      "loss": 0.013,
      "step": 211050
    },
    {
      "epoch": 81.97,
      "learning_rate": 9.071326860841426e-06,
      "loss": 0.0757,
      "step": 211060
    },
    {
      "epoch": 81.97,
      "learning_rate": 9.070809061488674e-06,
      "loss": 0.0686,
      "step": 211070
    },
    {
      "epoch": 81.97,
      "learning_rate": 9.070291262135923e-06,
      "loss": 0.017,
      "step": 211080
    },
    {
      "epoch": 81.98,
      "learning_rate": 9.069773462783173e-06,
      "loss": 0.0461,
      "step": 211090
    },
    {
      "epoch": 81.98,
      "learning_rate": 9.069255663430423e-06,
      "loss": 0.0391,
      "step": 211100
    },
    {
      "epoch": 81.98,
      "learning_rate": 9.06873786407767e-06,
      "loss": 0.024,
      "step": 211110
    },
    {
      "epoch": 81.99,
      "learning_rate": 9.06822006472492e-06,
      "loss": 0.0004,
      "step": 211120
    },
    {
      "epoch": 81.99,
      "learning_rate": 9.06770226537217e-06,
      "loss": 0.0019,
      "step": 211130
    },
    {
      "epoch": 82.0,
      "learning_rate": 9.067184466019418e-06,
      "loss": 0.0783,
      "step": 211140
    },
    {
      "epoch": 82.0,
      "learning_rate": 9.066666666666667e-06,
      "loss": 0.0321,
      "step": 211150
    },
    {
      "epoch": 82.0,
      "eval_accuracy": 0.949656121045392,
      "eval_loss": 0.35550665855407715,
      "eval_runtime": 8.1719,
      "eval_samples_per_second": 444.817,
      "eval_steps_per_second": 55.679,
      "step": 211150
    },
    {
      "epoch": 82.0,
      "learning_rate": 9.066148867313917e-06,
      "loss": 0.0241,
      "step": 211160
    },
    {
      "epoch": 82.01,
      "learning_rate": 9.065631067961167e-06,
      "loss": 0.1045,
      "step": 211170
    },
    {
      "epoch": 82.01,
      "learning_rate": 9.065113268608415e-06,
      "loss": 0.0207,
      "step": 211180
    },
    {
      "epoch": 82.02,
      "learning_rate": 9.064595469255664e-06,
      "loss": 0.0713,
      "step": 211190
    },
    {
      "epoch": 82.02,
      "learning_rate": 9.064077669902914e-06,
      "loss": 0.0396,
      "step": 211200
    },
    {
      "epoch": 82.02,
      "learning_rate": 9.063559870550163e-06,
      "loss": 0.0562,
      "step": 211210
    },
    {
      "epoch": 82.03,
      "learning_rate": 9.063042071197411e-06,
      "loss": 0.0004,
      "step": 211220
    },
    {
      "epoch": 82.03,
      "learning_rate": 9.062524271844661e-06,
      "loss": 0.0576,
      "step": 211230
    },
    {
      "epoch": 82.03,
      "learning_rate": 9.06200647249191e-06,
      "loss": 0.0888,
      "step": 211240
    },
    {
      "epoch": 82.04,
      "learning_rate": 9.06148867313916e-06,
      "loss": 0.0095,
      "step": 211250
    },
    {
      "epoch": 82.04,
      "learning_rate": 9.060970873786408e-06,
      "loss": 0.0886,
      "step": 211260
    },
    {
      "epoch": 82.05,
      "learning_rate": 9.060453074433658e-06,
      "loss": 0.0436,
      "step": 211270
    },
    {
      "epoch": 82.05,
      "learning_rate": 9.059935275080907e-06,
      "loss": 0.0707,
      "step": 211280
    },
    {
      "epoch": 82.05,
      "learning_rate": 9.059417475728155e-06,
      "loss": 0.0105,
      "step": 211290
    },
    {
      "epoch": 82.06,
      "learning_rate": 9.058899676375405e-06,
      "loss": 0.0677,
      "step": 211300
    },
    {
      "epoch": 82.06,
      "learning_rate": 9.058381877022654e-06,
      "loss": 0.0201,
      "step": 211310
    },
    {
      "epoch": 82.07,
      "learning_rate": 9.057864077669904e-06,
      "loss": 0.0574,
      "step": 211320
    },
    {
      "epoch": 82.07,
      "learning_rate": 9.057346278317152e-06,
      "loss": 0.035,
      "step": 211330
    },
    {
      "epoch": 82.07,
      "learning_rate": 9.056828478964402e-06,
      "loss": 0.0047,
      "step": 211340
    },
    {
      "epoch": 82.08,
      "learning_rate": 9.056310679611651e-06,
      "loss": 0.0218,
      "step": 211350
    },
    {
      "epoch": 82.08,
      "learning_rate": 9.055792880258901e-06,
      "loss": 0.097,
      "step": 211360
    },
    {
      "epoch": 82.09,
      "learning_rate": 9.055275080906149e-06,
      "loss": 0.0028,
      "step": 211370
    },
    {
      "epoch": 82.09,
      "learning_rate": 9.054757281553398e-06,
      "loss": 0.0154,
      "step": 211380
    },
    {
      "epoch": 82.09,
      "learning_rate": 9.054239482200648e-06,
      "loss": 0.0207,
      "step": 211390
    },
    {
      "epoch": 82.1,
      "learning_rate": 9.053721682847898e-06,
      "loss": 0.0379,
      "step": 211400
    },
    {
      "epoch": 82.1,
      "learning_rate": 9.053203883495146e-06,
      "loss": 0.1271,
      "step": 211410
    },
    {
      "epoch": 82.1,
      "learning_rate": 9.052686084142395e-06,
      "loss": 0.0581,
      "step": 211420
    },
    {
      "epoch": 82.11,
      "learning_rate": 9.052168284789645e-06,
      "loss": 0.0021,
      "step": 211430
    },
    {
      "epoch": 82.11,
      "learning_rate": 9.051650485436894e-06,
      "loss": 0.053,
      "step": 211440
    },
    {
      "epoch": 82.12,
      "learning_rate": 9.051132686084142e-06,
      "loss": 0.1199,
      "step": 211450
    },
    {
      "epoch": 82.12,
      "learning_rate": 9.050614886731392e-06,
      "loss": 0.0897,
      "step": 211460
    },
    {
      "epoch": 82.12,
      "learning_rate": 9.050097087378642e-06,
      "loss": 0.0248,
      "step": 211470
    },
    {
      "epoch": 82.13,
      "learning_rate": 9.04957928802589e-06,
      "loss": 0.0002,
      "step": 211480
    },
    {
      "epoch": 82.13,
      "learning_rate": 9.049061488673139e-06,
      "loss": 0.011,
      "step": 211490
    },
    {
      "epoch": 82.14,
      "learning_rate": 9.048543689320389e-06,
      "loss": 0.0079,
      "step": 211500
    },
    {
      "epoch": 82.14,
      "learning_rate": 9.048025889967638e-06,
      "loss": 0.0261,
      "step": 211510
    },
    {
      "epoch": 82.14,
      "learning_rate": 9.047508090614886e-06,
      "loss": 0.0934,
      "step": 211520
    },
    {
      "epoch": 82.15,
      "learning_rate": 9.046990291262136e-06,
      "loss": 0.0397,
      "step": 211530
    },
    {
      "epoch": 82.15,
      "learning_rate": 9.046472491909386e-06,
      "loss": 0.1331,
      "step": 211540
    },
    {
      "epoch": 82.16,
      "learning_rate": 9.045954692556635e-06,
      "loss": 0.1355,
      "step": 211550
    },
    {
      "epoch": 82.16,
      "learning_rate": 9.045436893203883e-06,
      "loss": 0.0405,
      "step": 211560
    },
    {
      "epoch": 82.16,
      "learning_rate": 9.044919093851133e-06,
      "loss": 0.0278,
      "step": 211570
    },
    {
      "epoch": 82.17,
      "learning_rate": 9.044401294498382e-06,
      "loss": 0.0489,
      "step": 211580
    },
    {
      "epoch": 82.17,
      "learning_rate": 9.043883495145632e-06,
      "loss": 0.0309,
      "step": 211590
    },
    {
      "epoch": 82.17,
      "learning_rate": 9.04336569579288e-06,
      "loss": 0.0204,
      "step": 211600
    },
    {
      "epoch": 82.18,
      "learning_rate": 9.04284789644013e-06,
      "loss": 0.0465,
      "step": 211610
    },
    {
      "epoch": 82.18,
      "learning_rate": 9.042330097087379e-06,
      "loss": 0.0363,
      "step": 211620
    },
    {
      "epoch": 82.19,
      "learning_rate": 9.041812297734629e-06,
      "loss": 0.0163,
      "step": 211630
    },
    {
      "epoch": 82.19,
      "learning_rate": 9.041294498381877e-06,
      "loss": 0.0312,
      "step": 211640
    },
    {
      "epoch": 82.19,
      "learning_rate": 9.040776699029126e-06,
      "loss": 0.0157,
      "step": 211650
    },
    {
      "epoch": 82.2,
      "learning_rate": 9.040258899676376e-06,
      "loss": 0.1325,
      "step": 211660
    },
    {
      "epoch": 82.2,
      "learning_rate": 9.039741100323626e-06,
      "loss": 0.0054,
      "step": 211670
    },
    {
      "epoch": 82.21,
      "learning_rate": 9.039223300970873e-06,
      "loss": 0.0016,
      "step": 211680
    },
    {
      "epoch": 82.21,
      "learning_rate": 9.038705501618123e-06,
      "loss": 0.1406,
      "step": 211690
    },
    {
      "epoch": 82.21,
      "learning_rate": 9.038187702265373e-06,
      "loss": 0.0277,
      "step": 211700
    },
    {
      "epoch": 82.22,
      "learning_rate": 9.037669902912622e-06,
      "loss": 0.0084,
      "step": 211710
    },
    {
      "epoch": 82.22,
      "learning_rate": 9.03715210355987e-06,
      "loss": 0.2167,
      "step": 211720
    },
    {
      "epoch": 82.23,
      "learning_rate": 9.03663430420712e-06,
      "loss": 0.0309,
      "step": 211730
    },
    {
      "epoch": 82.23,
      "learning_rate": 9.03611650485437e-06,
      "loss": 0.0438,
      "step": 211740
    },
    {
      "epoch": 82.23,
      "learning_rate": 9.035598705501619e-06,
      "loss": 0.0599,
      "step": 211750
    },
    {
      "epoch": 82.24,
      "learning_rate": 9.035080906148867e-06,
      "loss": 0.0358,
      "step": 211760
    },
    {
      "epoch": 82.24,
      "learning_rate": 9.034563106796117e-06,
      "loss": 0.113,
      "step": 211770
    },
    {
      "epoch": 82.24,
      "learning_rate": 9.034045307443366e-06,
      "loss": 0.1344,
      "step": 211780
    },
    {
      "epoch": 82.25,
      "learning_rate": 9.033527508090616e-06,
      "loss": 0.0623,
      "step": 211790
    },
    {
      "epoch": 82.25,
      "learning_rate": 9.033009708737864e-06,
      "loss": 0.0137,
      "step": 211800
    },
    {
      "epoch": 82.26,
      "learning_rate": 9.032491909385113e-06,
      "loss": 0.1124,
      "step": 211810
    },
    {
      "epoch": 82.26,
      "learning_rate": 9.031974110032363e-06,
      "loss": 0.0806,
      "step": 211820
    },
    {
      "epoch": 82.26,
      "learning_rate": 9.031456310679613e-06,
      "loss": 0.1408,
      "step": 211830
    },
    {
      "epoch": 82.27,
      "learning_rate": 9.03093851132686e-06,
      "loss": 0.0017,
      "step": 211840
    },
    {
      "epoch": 82.27,
      "learning_rate": 9.03042071197411e-06,
      "loss": 0.0675,
      "step": 211850
    },
    {
      "epoch": 82.28,
      "learning_rate": 9.02990291262136e-06,
      "loss": 0.1094,
      "step": 211860
    },
    {
      "epoch": 82.28,
      "learning_rate": 9.02938511326861e-06,
      "loss": 0.0353,
      "step": 211870
    },
    {
      "epoch": 82.28,
      "learning_rate": 9.028867313915857e-06,
      "loss": 0.0652,
      "step": 211880
    },
    {
      "epoch": 82.29,
      "learning_rate": 9.028349514563107e-06,
      "loss": 0.028,
      "step": 211890
    },
    {
      "epoch": 82.29,
      "learning_rate": 9.027831715210357e-06,
      "loss": 0.0708,
      "step": 211900
    },
    {
      "epoch": 82.3,
      "learning_rate": 9.027313915857606e-06,
      "loss": 0.0203,
      "step": 211910
    },
    {
      "epoch": 82.3,
      "learning_rate": 9.026796116504856e-06,
      "loss": 0.0826,
      "step": 211920
    },
    {
      "epoch": 82.3,
      "learning_rate": 9.026278317152104e-06,
      "loss": 0.0273,
      "step": 211930
    },
    {
      "epoch": 82.31,
      "learning_rate": 9.025760517799353e-06,
      "loss": 0.1086,
      "step": 211940
    },
    {
      "epoch": 82.31,
      "learning_rate": 9.025242718446603e-06,
      "loss": 0.0659,
      "step": 211950
    },
    {
      "epoch": 82.31,
      "learning_rate": 9.024724919093853e-06,
      "loss": 0.178,
      "step": 211960
    },
    {
      "epoch": 82.32,
      "learning_rate": 9.0242071197411e-06,
      "loss": 0.1838,
      "step": 211970
    },
    {
      "epoch": 82.32,
      "learning_rate": 9.02368932038835e-06,
      "loss": 0.0267,
      "step": 211980
    },
    {
      "epoch": 82.33,
      "learning_rate": 9.0231715210356e-06,
      "loss": 0.1512,
      "step": 211990
    },
    {
      "epoch": 82.33,
      "learning_rate": 9.02265372168285e-06,
      "loss": 0.0264,
      "step": 212000
    },
    {
      "epoch": 82.33,
      "learning_rate": 9.022135922330097e-06,
      "loss": 0.0568,
      "step": 212010
    },
    {
      "epoch": 82.34,
      "learning_rate": 9.021618122977347e-06,
      "loss": 0.0186,
      "step": 212020
    },
    {
      "epoch": 82.34,
      "learning_rate": 9.021100323624597e-06,
      "loss": 0.117,
      "step": 212030
    },
    {
      "epoch": 82.35,
      "learning_rate": 9.020582524271846e-06,
      "loss": 0.0709,
      "step": 212040
    },
    {
      "epoch": 82.35,
      "learning_rate": 9.020064724919094e-06,
      "loss": 0.0904,
      "step": 212050
    },
    {
      "epoch": 82.35,
      "learning_rate": 9.019546925566344e-06,
      "loss": 0.0468,
      "step": 212060
    },
    {
      "epoch": 82.36,
      "learning_rate": 9.019029126213593e-06,
      "loss": 0.0135,
      "step": 212070
    },
    {
      "epoch": 82.36,
      "learning_rate": 9.018511326860843e-06,
      "loss": 0.0223,
      "step": 212080
    },
    {
      "epoch": 82.37,
      "learning_rate": 9.01799352750809e-06,
      "loss": 0.0432,
      "step": 212090
    },
    {
      "epoch": 82.37,
      "learning_rate": 9.01747572815534e-06,
      "loss": 0.0593,
      "step": 212100
    },
    {
      "epoch": 82.37,
      "learning_rate": 9.01695792880259e-06,
      "loss": 0.1741,
      "step": 212110
    },
    {
      "epoch": 82.38,
      "learning_rate": 9.01644012944984e-06,
      "loss": 0.1354,
      "step": 212120
    },
    {
      "epoch": 82.38,
      "learning_rate": 9.015922330097088e-06,
      "loss": 0.0121,
      "step": 212130
    },
    {
      "epoch": 82.38,
      "learning_rate": 9.015404530744337e-06,
      "loss": 0.0837,
      "step": 212140
    },
    {
      "epoch": 82.39,
      "learning_rate": 9.014886731391587e-06,
      "loss": 0.0132,
      "step": 212150
    },
    {
      "epoch": 82.39,
      "learning_rate": 9.014368932038836e-06,
      "loss": 0.0204,
      "step": 212160
    },
    {
      "epoch": 82.4,
      "learning_rate": 9.013851132686084e-06,
      "loss": 0.0167,
      "step": 212170
    },
    {
      "epoch": 82.4,
      "learning_rate": 9.013333333333334e-06,
      "loss": 0.0189,
      "step": 212180
    },
    {
      "epoch": 82.4,
      "learning_rate": 9.012815533980584e-06,
      "loss": 0.1586,
      "step": 212190
    },
    {
      "epoch": 82.41,
      "learning_rate": 9.012297734627833e-06,
      "loss": 0.04,
      "step": 212200
    },
    {
      "epoch": 82.41,
      "learning_rate": 9.011779935275081e-06,
      "loss": 0.1638,
      "step": 212210
    },
    {
      "epoch": 82.42,
      "learning_rate": 9.01126213592233e-06,
      "loss": 0.0396,
      "step": 212220
    },
    {
      "epoch": 82.42,
      "learning_rate": 9.01074433656958e-06,
      "loss": 0.1615,
      "step": 212230
    },
    {
      "epoch": 82.42,
      "learning_rate": 9.01022653721683e-06,
      "loss": 0.0937,
      "step": 212240
    },
    {
      "epoch": 82.43,
      "learning_rate": 9.009708737864078e-06,
      "loss": 0.0645,
      "step": 212250
    },
    {
      "epoch": 82.43,
      "learning_rate": 9.009190938511328e-06,
      "loss": 0.001,
      "step": 212260
    },
    {
      "epoch": 82.43,
      "learning_rate": 9.008673139158577e-06,
      "loss": 0.0688,
      "step": 212270
    },
    {
      "epoch": 82.44,
      "learning_rate": 9.008155339805827e-06,
      "loss": 0.0239,
      "step": 212280
    },
    {
      "epoch": 82.44,
      "learning_rate": 9.007637540453075e-06,
      "loss": 0.0328,
      "step": 212290
    },
    {
      "epoch": 82.45,
      "learning_rate": 9.007119741100324e-06,
      "loss": 0.1198,
      "step": 212300
    },
    {
      "epoch": 82.45,
      "learning_rate": 9.006601941747574e-06,
      "loss": 0.0408,
      "step": 212310
    },
    {
      "epoch": 82.45,
      "learning_rate": 9.006084142394824e-06,
      "loss": 0.0184,
      "step": 212320
    },
    {
      "epoch": 82.46,
      "learning_rate": 9.005566343042071e-06,
      "loss": 0.0565,
      "step": 212330
    },
    {
      "epoch": 82.46,
      "learning_rate": 9.005048543689321e-06,
      "loss": 0.1639,
      "step": 212340
    },
    {
      "epoch": 82.47,
      "learning_rate": 9.00453074433657e-06,
      "loss": 0.0607,
      "step": 212350
    },
    {
      "epoch": 82.47,
      "learning_rate": 9.00401294498382e-06,
      "loss": 0.0212,
      "step": 212360
    },
    {
      "epoch": 82.47,
      "learning_rate": 9.003495145631068e-06,
      "loss": 0.0018,
      "step": 212370
    },
    {
      "epoch": 82.48,
      "learning_rate": 9.002977346278318e-06,
      "loss": 0.0323,
      "step": 212380
    },
    {
      "epoch": 82.48,
      "learning_rate": 9.002459546925568e-06,
      "loss": 0.0869,
      "step": 212390
    },
    {
      "epoch": 82.49,
      "learning_rate": 9.001941747572817e-06,
      "loss": 0.3047,
      "step": 212400
    },
    {
      "epoch": 82.49,
      "learning_rate": 9.001423948220065e-06,
      "loss": 0.0465,
      "step": 212410
    },
    {
      "epoch": 82.49,
      "learning_rate": 9.000906148867315e-06,
      "loss": 0.146,
      "step": 212420
    },
    {
      "epoch": 82.5,
      "learning_rate": 9.000388349514564e-06,
      "loss": 0.1513,
      "step": 212430
    },
    {
      "epoch": 82.5,
      "learning_rate": 8.999870550161814e-06,
      "loss": 0.0419,
      "step": 212440
    },
    {
      "epoch": 82.5,
      "learning_rate": 8.999352750809062e-06,
      "loss": 0.0806,
      "step": 212450
    },
    {
      "epoch": 82.51,
      "learning_rate": 8.998834951456311e-06,
      "loss": 0.0665,
      "step": 212460
    },
    {
      "epoch": 82.51,
      "learning_rate": 8.998317152103561e-06,
      "loss": 0.08,
      "step": 212470
    },
    {
      "epoch": 82.52,
      "learning_rate": 8.99779935275081e-06,
      "loss": 0.0198,
      "step": 212480
    },
    {
      "epoch": 82.52,
      "learning_rate": 8.99728155339806e-06,
      "loss": 0.0628,
      "step": 212490
    },
    {
      "epoch": 82.52,
      "learning_rate": 8.996763754045308e-06,
      "loss": 0.0426,
      "step": 212500
    },
    {
      "epoch": 82.53,
      "learning_rate": 8.996245954692558e-06,
      "loss": 0.0481,
      "step": 212510
    },
    {
      "epoch": 82.53,
      "learning_rate": 8.995728155339807e-06,
      "loss": 0.1077,
      "step": 212520
    },
    {
      "epoch": 82.54,
      "learning_rate": 8.995210355987057e-06,
      "loss": 0.0173,
      "step": 212530
    },
    {
      "epoch": 82.54,
      "learning_rate": 8.994692556634305e-06,
      "loss": 0.1043,
      "step": 212540
    },
    {
      "epoch": 82.54,
      "learning_rate": 8.994174757281555e-06,
      "loss": 0.0956,
      "step": 212550
    },
    {
      "epoch": 82.55,
      "learning_rate": 8.993656957928804e-06,
      "loss": 0.018,
      "step": 212560
    },
    {
      "epoch": 82.55,
      "learning_rate": 8.993139158576054e-06,
      "loss": 0.1036,
      "step": 212570
    },
    {
      "epoch": 82.56,
      "learning_rate": 8.992621359223302e-06,
      "loss": 0.0644,
      "step": 212580
    },
    {
      "epoch": 82.56,
      "learning_rate": 8.992103559870551e-06,
      "loss": 0.1074,
      "step": 212590
    },
    {
      "epoch": 82.56,
      "learning_rate": 8.991585760517801e-06,
      "loss": 0.0843,
      "step": 212600
    },
    {
      "epoch": 82.57,
      "learning_rate": 8.991067961165049e-06,
      "loss": 0.1225,
      "step": 212610
    },
    {
      "epoch": 82.57,
      "learning_rate": 8.990550161812299e-06,
      "loss": 0.1554,
      "step": 212620
    },
    {
      "epoch": 82.57,
      "learning_rate": 8.990032362459548e-06,
      "loss": 0.0372,
      "step": 212630
    },
    {
      "epoch": 82.58,
      "learning_rate": 8.989514563106798e-06,
      "loss": 0.078,
      "step": 212640
    },
    {
      "epoch": 82.58,
      "learning_rate": 8.988996763754046e-06,
      "loss": 0.0154,
      "step": 212650
    },
    {
      "epoch": 82.59,
      "learning_rate": 8.988478964401295e-06,
      "loss": 0.0019,
      "step": 212660
    },
    {
      "epoch": 82.59,
      "learning_rate": 8.987961165048545e-06,
      "loss": 0.157,
      "step": 212670
    },
    {
      "epoch": 82.59,
      "learning_rate": 8.987443365695795e-06,
      "loss": 0.0182,
      "step": 212680
    },
    {
      "epoch": 82.6,
      "learning_rate": 8.986925566343042e-06,
      "loss": 0.0191,
      "step": 212690
    },
    {
      "epoch": 82.6,
      "learning_rate": 8.986407766990292e-06,
      "loss": 0.103,
      "step": 212700
    },
    {
      "epoch": 82.61,
      "learning_rate": 8.985889967637542e-06,
      "loss": 0.0267,
      "step": 212710
    },
    {
      "epoch": 82.61,
      "learning_rate": 8.985372168284791e-06,
      "loss": 0.1026,
      "step": 212720
    },
    {
      "epoch": 82.61,
      "learning_rate": 8.98485436893204e-06,
      "loss": 0.0027,
      "step": 212730
    },
    {
      "epoch": 82.62,
      "learning_rate": 8.984336569579289e-06,
      "loss": 0.0478,
      "step": 212740
    },
    {
      "epoch": 82.62,
      "learning_rate": 8.983818770226539e-06,
      "loss": 0.0285,
      "step": 212750
    },
    {
      "epoch": 82.63,
      "learning_rate": 8.983300970873786e-06,
      "loss": 0.0001,
      "step": 212760
    },
    {
      "epoch": 82.63,
      "learning_rate": 8.982783171521036e-06,
      "loss": 0.0358,
      "step": 212770
    },
    {
      "epoch": 82.63,
      "learning_rate": 8.982265372168286e-06,
      "loss": 0.0189,
      "step": 212780
    },
    {
      "epoch": 82.64,
      "learning_rate": 8.981747572815535e-06,
      "loss": 0.2348,
      "step": 212790
    },
    {
      "epoch": 82.64,
      "learning_rate": 8.981229773462783e-06,
      "loss": 0.0319,
      "step": 212800
    },
    {
      "epoch": 82.64,
      "learning_rate": 8.980711974110033e-06,
      "loss": 0.2198,
      "step": 212810
    },
    {
      "epoch": 82.65,
      "learning_rate": 8.980194174757282e-06,
      "loss": 0.0057,
      "step": 212820
    },
    {
      "epoch": 82.65,
      "learning_rate": 8.979676375404532e-06,
      "loss": 0.0712,
      "step": 212830
    },
    {
      "epoch": 82.66,
      "learning_rate": 8.97915857605178e-06,
      "loss": 0.0627,
      "step": 212840
    },
    {
      "epoch": 82.66,
      "learning_rate": 8.97864077669903e-06,
      "loss": 0.1685,
      "step": 212850
    },
    {
      "epoch": 82.66,
      "learning_rate": 8.97812297734628e-06,
      "loss": 0.0003,
      "step": 212860
    },
    {
      "epoch": 82.67,
      "learning_rate": 8.977605177993529e-06,
      "loss": 0.0006,
      "step": 212870
    },
    {
      "epoch": 82.67,
      "learning_rate": 8.977087378640777e-06,
      "loss": 0.009,
      "step": 212880
    },
    {
      "epoch": 82.68,
      "learning_rate": 8.976569579288026e-06,
      "loss": 0.0667,
      "step": 212890
    },
    {
      "epoch": 82.68,
      "learning_rate": 8.976051779935276e-06,
      "loss": 0.1025,
      "step": 212900
    },
    {
      "epoch": 82.68,
      "learning_rate": 8.975533980582526e-06,
      "loss": 0.0291,
      "step": 212910
    },
    {
      "epoch": 82.69,
      "learning_rate": 8.975016181229774e-06,
      "loss": 0.0093,
      "step": 212920
    },
    {
      "epoch": 82.69,
      "learning_rate": 8.974498381877023e-06,
      "loss": 0.0136,
      "step": 212930
    },
    {
      "epoch": 82.7,
      "learning_rate": 8.973980582524273e-06,
      "loss": 0.0856,
      "step": 212940
    },
    {
      "epoch": 82.7,
      "learning_rate": 8.97346278317152e-06,
      "loss": 0.0352,
      "step": 212950
    },
    {
      "epoch": 82.7,
      "learning_rate": 8.97294498381877e-06,
      "loss": 0.1215,
      "step": 212960
    },
    {
      "epoch": 82.71,
      "learning_rate": 8.97242718446602e-06,
      "loss": 0.0463,
      "step": 212970
    },
    {
      "epoch": 82.71,
      "learning_rate": 8.97190938511327e-06,
      "loss": 0.135,
      "step": 212980
    },
    {
      "epoch": 82.71,
      "learning_rate": 8.971391585760517e-06,
      "loss": 0.0673,
      "step": 212990
    },
    {
      "epoch": 82.72,
      "learning_rate": 8.970873786407767e-06,
      "loss": 0.0139,
      "step": 213000
    },
    {
      "epoch": 82.72,
      "learning_rate": 8.970355987055017e-06,
      "loss": 0.0096,
      "step": 213010
    },
    {
      "epoch": 82.73,
      "learning_rate": 8.969838187702266e-06,
      "loss": 0.0751,
      "step": 213020
    },
    {
      "epoch": 82.73,
      "learning_rate": 8.969320388349514e-06,
      "loss": 0.0044,
      "step": 213030
    },
    {
      "epoch": 82.73,
      "learning_rate": 8.968802588996764e-06,
      "loss": 0.0649,
      "step": 213040
    },
    {
      "epoch": 82.74,
      "learning_rate": 8.968284789644013e-06,
      "loss": 0.044,
      "step": 213050
    },
    {
      "epoch": 82.74,
      "learning_rate": 8.967766990291263e-06,
      "loss": 0.0372,
      "step": 213060
    },
    {
      "epoch": 82.75,
      "learning_rate": 8.967249190938511e-06,
      "loss": 0.0583,
      "step": 213070
    },
    {
      "epoch": 82.75,
      "learning_rate": 8.96673139158576e-06,
      "loss": 0.03,
      "step": 213080
    },
    {
      "epoch": 82.75,
      "learning_rate": 8.96621359223301e-06,
      "loss": 0.0181,
      "step": 213090
    },
    {
      "epoch": 82.76,
      "learning_rate": 8.96569579288026e-06,
      "loss": 0.0073,
      "step": 213100
    },
    {
      "epoch": 82.76,
      "learning_rate": 8.965177993527508e-06,
      "loss": 0.0627,
      "step": 213110
    },
    {
      "epoch": 82.77,
      "learning_rate": 8.964660194174757e-06,
      "loss": 0.0673,
      "step": 213120
    },
    {
      "epoch": 82.77,
      "learning_rate": 8.964142394822007e-06,
      "loss": 0.0934,
      "step": 213130
    },
    {
      "epoch": 82.77,
      "learning_rate": 8.963624595469257e-06,
      "loss": 0.1125,
      "step": 213140
    },
    {
      "epoch": 82.78,
      "learning_rate": 8.963106796116505e-06,
      "loss": 0.1144,
      "step": 213150
    },
    {
      "epoch": 82.78,
      "learning_rate": 8.962588996763754e-06,
      "loss": 0.0726,
      "step": 213160
    },
    {
      "epoch": 82.78,
      "learning_rate": 8.962071197411004e-06,
      "loss": 0.0017,
      "step": 213170
    },
    {
      "epoch": 82.79,
      "learning_rate": 8.961553398058253e-06,
      "loss": 0.0803,
      "step": 213180
    },
    {
      "epoch": 82.79,
      "learning_rate": 8.961035598705501e-06,
      "loss": 0.0041,
      "step": 213190
    },
    {
      "epoch": 82.8,
      "learning_rate": 8.960517799352751e-06,
      "loss": 0.0138,
      "step": 213200
    },
    {
      "epoch": 82.8,
      "learning_rate": 8.96e-06,
      "loss": 0.0677,
      "step": 213210
    },
    {
      "epoch": 82.8,
      "learning_rate": 8.95948220064725e-06,
      "loss": 0.0248,
      "step": 213220
    },
    {
      "epoch": 82.81,
      "learning_rate": 8.958964401294498e-06,
      "loss": 0.0842,
      "step": 213230
    },
    {
      "epoch": 82.81,
      "learning_rate": 8.958446601941748e-06,
      "loss": 0.1126,
      "step": 213240
    },
    {
      "epoch": 82.82,
      "learning_rate": 8.957928802588997e-06,
      "loss": 0.0559,
      "step": 213250
    },
    {
      "epoch": 82.82,
      "learning_rate": 8.957411003236247e-06,
      "loss": 0.0081,
      "step": 213260
    },
    {
      "epoch": 82.82,
      "learning_rate": 8.956893203883495e-06,
      "loss": 0.0808,
      "step": 213270
    },
    {
      "epoch": 82.83,
      "learning_rate": 8.956375404530745e-06,
      "loss": 0.0314,
      "step": 213280
    },
    {
      "epoch": 82.83,
      "learning_rate": 8.955857605177994e-06,
      "loss": 0.1403,
      "step": 213290
    },
    {
      "epoch": 82.83,
      "learning_rate": 8.955339805825244e-06,
      "loss": 0.0942,
      "step": 213300
    },
    {
      "epoch": 82.84,
      "learning_rate": 8.954822006472492e-06,
      "loss": 0.0063,
      "step": 213310
    },
    {
      "epoch": 82.84,
      "learning_rate": 8.954304207119741e-06,
      "loss": 0.0713,
      "step": 213320
    },
    {
      "epoch": 82.85,
      "learning_rate": 8.953786407766991e-06,
      "loss": 0.0682,
      "step": 213330
    },
    {
      "epoch": 82.85,
      "learning_rate": 8.95326860841424e-06,
      "loss": 0.0189,
      "step": 213340
    },
    {
      "epoch": 82.85,
      "learning_rate": 8.952750809061488e-06,
      "loss": 0.1195,
      "step": 213350
    },
    {
      "epoch": 82.86,
      "learning_rate": 8.952233009708738e-06,
      "loss": 0.0127,
      "step": 213360
    },
    {
      "epoch": 82.86,
      "learning_rate": 8.951715210355988e-06,
      "loss": 0.0124,
      "step": 213370
    },
    {
      "epoch": 82.87,
      "learning_rate": 8.951197411003237e-06,
      "loss": 0.0196,
      "step": 213380
    },
    {
      "epoch": 82.87,
      "learning_rate": 8.950679611650485e-06,
      "loss": 0.1042,
      "step": 213390
    },
    {
      "epoch": 82.87,
      "learning_rate": 8.950161812297735e-06,
      "loss": 0.0088,
      "step": 213400
    },
    {
      "epoch": 82.88,
      "learning_rate": 8.949644012944985e-06,
      "loss": 0.1113,
      "step": 213410
    },
    {
      "epoch": 82.88,
      "learning_rate": 8.949126213592234e-06,
      "loss": 0.1065,
      "step": 213420
    },
    {
      "epoch": 82.89,
      "learning_rate": 8.948608414239482e-06,
      "loss": 0.1004,
      "step": 213430
    },
    {
      "epoch": 82.89,
      "learning_rate": 8.948090614886732e-06,
      "loss": 0.0067,
      "step": 213440
    },
    {
      "epoch": 82.89,
      "learning_rate": 8.947572815533981e-06,
      "loss": 0.0155,
      "step": 213450
    },
    {
      "epoch": 82.9,
      "learning_rate": 8.947055016181231e-06,
      "loss": 0.0392,
      "step": 213460
    },
    {
      "epoch": 82.9,
      "learning_rate": 8.946537216828479e-06,
      "loss": 0.0231,
      "step": 213470
    },
    {
      "epoch": 82.9,
      "learning_rate": 8.946019417475728e-06,
      "loss": 0.0122,
      "step": 213480
    },
    {
      "epoch": 82.91,
      "learning_rate": 8.945501618122978e-06,
      "loss": 0.0395,
      "step": 213490
    },
    {
      "epoch": 82.91,
      "learning_rate": 8.944983818770228e-06,
      "loss": 0.1326,
      "step": 213500
    },
    {
      "epoch": 82.92,
      "learning_rate": 8.944466019417476e-06,
      "loss": 0.0062,
      "step": 213510
    },
    {
      "epoch": 82.92,
      "learning_rate": 8.943948220064725e-06,
      "loss": 0.0466,
      "step": 213520
    },
    {
      "epoch": 82.92,
      "learning_rate": 8.943430420711975e-06,
      "loss": 0.0094,
      "step": 213530
    },
    {
      "epoch": 82.93,
      "learning_rate": 8.942912621359224e-06,
      "loss": 0.0109,
      "step": 213540
    },
    {
      "epoch": 82.93,
      "learning_rate": 8.942394822006472e-06,
      "loss": 0.0495,
      "step": 213550
    },
    {
      "epoch": 82.94,
      "learning_rate": 8.941877022653722e-06,
      "loss": 0.0615,
      "step": 213560
    },
    {
      "epoch": 82.94,
      "learning_rate": 8.941359223300972e-06,
      "loss": 0.0085,
      "step": 213570
    },
    {
      "epoch": 82.94,
      "learning_rate": 8.940841423948221e-06,
      "loss": 0.1322,
      "step": 213580
    },
    {
      "epoch": 82.95,
      "learning_rate": 8.940323624595471e-06,
      "loss": 0.0167,
      "step": 213590
    },
    {
      "epoch": 82.95,
      "learning_rate": 8.939805825242719e-06,
      "loss": 0.022,
      "step": 213600
    },
    {
      "epoch": 82.96,
      "learning_rate": 8.939288025889968e-06,
      "loss": 0.0616,
      "step": 213610
    },
    {
      "epoch": 82.96,
      "learning_rate": 8.938770226537218e-06,
      "loss": 0.1776,
      "step": 213620
    },
    {
      "epoch": 82.96,
      "learning_rate": 8.938252427184468e-06,
      "loss": 0.0204,
      "step": 213630
    },
    {
      "epoch": 82.97,
      "learning_rate": 8.937734627831716e-06,
      "loss": 0.0776,
      "step": 213640
    },
    {
      "epoch": 82.97,
      "learning_rate": 8.937216828478965e-06,
      "loss": 0.101,
      "step": 213650
    },
    {
      "epoch": 82.97,
      "learning_rate": 8.936699029126215e-06,
      "loss": 0.0555,
      "step": 213660
    },
    {
      "epoch": 82.98,
      "learning_rate": 8.936181229773464e-06,
      "loss": 0.0359,
      "step": 213670
    },
    {
      "epoch": 82.98,
      "learning_rate": 8.935663430420712e-06,
      "loss": 0.0002,
      "step": 213680
    },
    {
      "epoch": 82.99,
      "learning_rate": 8.935145631067962e-06,
      "loss": 0.1289,
      "step": 213690
    },
    {
      "epoch": 82.99,
      "learning_rate": 8.934627831715212e-06,
      "loss": 0.0442,
      "step": 213700
    },
    {
      "epoch": 82.99,
      "learning_rate": 8.934110032362461e-06,
      "loss": 0.014,
      "step": 213710
    },
    {
      "epoch": 83.0,
      "learning_rate": 8.933592233009709e-06,
      "loss": 0.051,
      "step": 213720
    },
    {
      "epoch": 83.0,
      "eval_accuracy": 0.9504814305364512,
      "eval_loss": 0.3729594647884369,
      "eval_runtime": 8.1912,
      "eval_samples_per_second": 443.767,
      "eval_steps_per_second": 55.547,
      "step": 213725
    },
    {
      "epoch": 83.0,
      "learning_rate": 8.933074433656959e-06,
      "loss": 0.1098,
      "step": 213730
    },
    {
      "epoch": 83.01,
      "learning_rate": 8.932556634304208e-06,
      "loss": 0.0499,
      "step": 213740
    },
    {
      "epoch": 83.01,
      "learning_rate": 8.932038834951458e-06,
      "loss": 0.0066,
      "step": 213750
    },
    {
      "epoch": 83.01,
      "learning_rate": 8.931521035598706e-06,
      "loss": 0.0163,
      "step": 213760
    },
    {
      "epoch": 83.02,
      "learning_rate": 8.931003236245956e-06,
      "loss": 0.0374,
      "step": 213770
    },
    {
      "epoch": 83.02,
      "learning_rate": 8.930485436893205e-06,
      "loss": 0.0599,
      "step": 213780
    },
    {
      "epoch": 83.03,
      "learning_rate": 8.929967637540455e-06,
      "loss": 0.0071,
      "step": 213790
    },
    {
      "epoch": 83.03,
      "learning_rate": 8.929449838187703e-06,
      "loss": 0.0014,
      "step": 213800
    },
    {
      "epoch": 83.03,
      "learning_rate": 8.928932038834952e-06,
      "loss": 0.0833,
      "step": 213810
    },
    {
      "epoch": 83.04,
      "learning_rate": 8.928414239482202e-06,
      "loss": 0.0328,
      "step": 213820
    },
    {
      "epoch": 83.04,
      "learning_rate": 8.927896440129452e-06,
      "loss": 0.1229,
      "step": 213830
    },
    {
      "epoch": 83.04,
      "learning_rate": 8.9273786407767e-06,
      "loss": 0.0003,
      "step": 213840
    },
    {
      "epoch": 83.05,
      "learning_rate": 8.926860841423949e-06,
      "loss": 0.0003,
      "step": 213850
    },
    {
      "epoch": 83.05,
      "learning_rate": 8.926343042071199e-06,
      "loss": 0.0988,
      "step": 213860
    },
    {
      "epoch": 83.06,
      "learning_rate": 8.925825242718448e-06,
      "loss": 0.0204,
      "step": 213870
    },
    {
      "epoch": 83.06,
      "learning_rate": 8.925307443365696e-06,
      "loss": 0.009,
      "step": 213880
    },
    {
      "epoch": 83.06,
      "learning_rate": 8.924789644012946e-06,
      "loss": 0.022,
      "step": 213890
    },
    {
      "epoch": 83.07,
      "learning_rate": 8.924271844660195e-06,
      "loss": 0.0489,
      "step": 213900
    },
    {
      "epoch": 83.07,
      "learning_rate": 8.923754045307445e-06,
      "loss": 0.0012,
      "step": 213910
    },
    {
      "epoch": 83.08,
      "learning_rate": 8.923236245954693e-06,
      "loss": 0.0163,
      "step": 213920
    },
    {
      "epoch": 83.08,
      "learning_rate": 8.922718446601943e-06,
      "loss": 0.134,
      "step": 213930
    },
    {
      "epoch": 83.08,
      "learning_rate": 8.922200647249192e-06,
      "loss": 0.0222,
      "step": 213940
    },
    {
      "epoch": 83.09,
      "learning_rate": 8.921682847896442e-06,
      "loss": 0.0367,
      "step": 213950
    },
    {
      "epoch": 83.09,
      "learning_rate": 8.92116504854369e-06,
      "loss": 0.0072,
      "step": 213960
    },
    {
      "epoch": 83.1,
      "learning_rate": 8.92064724919094e-06,
      "loss": 0.0987,
      "step": 213970
    },
    {
      "epoch": 83.1,
      "learning_rate": 8.920129449838189e-06,
      "loss": 0.045,
      "step": 213980
    },
    {
      "epoch": 83.1,
      "learning_rate": 8.919611650485439e-06,
      "loss": 0.0079,
      "step": 213990
    },
    {
      "epoch": 83.11,
      "learning_rate": 8.919093851132687e-06,
      "loss": 0.0264,
      "step": 214000
    },
    {
      "epoch": 83.11,
      "learning_rate": 8.918576051779936e-06,
      "loss": 0.0023,
      "step": 214010
    },
    {
      "epoch": 83.11,
      "learning_rate": 8.918058252427186e-06,
      "loss": 0.0706,
      "step": 214020
    },
    {
      "epoch": 83.12,
      "learning_rate": 8.917540453074435e-06,
      "loss": 0.1012,
      "step": 214030
    },
    {
      "epoch": 83.12,
      "learning_rate": 8.917022653721683e-06,
      "loss": 0.4475,
      "step": 214040
    },
    {
      "epoch": 83.13,
      "learning_rate": 8.916504854368933e-06,
      "loss": 0.0534,
      "step": 214050
    },
    {
      "epoch": 83.13,
      "learning_rate": 8.915987055016183e-06,
      "loss": 0.0013,
      "step": 214060
    },
    {
      "epoch": 83.13,
      "learning_rate": 8.915469255663432e-06,
      "loss": 0.0358,
      "step": 214070
    },
    {
      "epoch": 83.14,
      "learning_rate": 8.91495145631068e-06,
      "loss": 0.1854,
      "step": 214080
    },
    {
      "epoch": 83.14,
      "learning_rate": 8.91443365695793e-06,
      "loss": 0.1059,
      "step": 214090
    },
    {
      "epoch": 83.15,
      "learning_rate": 8.91391585760518e-06,
      "loss": 0.0113,
      "step": 214100
    },
    {
      "epoch": 83.15,
      "learning_rate": 8.913398058252429e-06,
      "loss": 0.0003,
      "step": 214110
    },
    {
      "epoch": 83.15,
      "learning_rate": 8.912880258899677e-06,
      "loss": 0.0219,
      "step": 214120
    },
    {
      "epoch": 83.16,
      "learning_rate": 8.912362459546927e-06,
      "loss": 0.0044,
      "step": 214130
    },
    {
      "epoch": 83.16,
      "learning_rate": 8.911844660194176e-06,
      "loss": 0.0986,
      "step": 214140
    },
    {
      "epoch": 83.17,
      "learning_rate": 8.911326860841426e-06,
      "loss": 0.0152,
      "step": 214150
    },
    {
      "epoch": 83.17,
      "learning_rate": 8.910809061488674e-06,
      "loss": 0.156,
      "step": 214160
    },
    {
      "epoch": 83.17,
      "learning_rate": 8.910291262135923e-06,
      "loss": 0.0292,
      "step": 214170
    },
    {
      "epoch": 83.18,
      "learning_rate": 8.909773462783173e-06,
      "loss": 0.0572,
      "step": 214180
    },
    {
      "epoch": 83.18,
      "learning_rate": 8.909255663430423e-06,
      "loss": 0.1479,
      "step": 214190
    },
    {
      "epoch": 83.18,
      "learning_rate": 8.90873786407767e-06,
      "loss": 0.1086,
      "step": 214200
    },
    {
      "epoch": 83.19,
      "learning_rate": 8.90822006472492e-06,
      "loss": 0.1411,
      "step": 214210
    },
    {
      "epoch": 83.19,
      "learning_rate": 8.90770226537217e-06,
      "loss": 0.0868,
      "step": 214220
    },
    {
      "epoch": 83.2,
      "learning_rate": 8.907184466019418e-06,
      "loss": 0.0156,
      "step": 214230
    },
    {
      "epoch": 83.2,
      "learning_rate": 8.906666666666667e-06,
      "loss": 0.0269,
      "step": 214240
    },
    {
      "epoch": 83.2,
      "learning_rate": 8.906148867313917e-06,
      "loss": 0.0622,
      "step": 214250
    },
    {
      "epoch": 83.21,
      "learning_rate": 8.905631067961166e-06,
      "loss": 0.0629,
      "step": 214260
    },
    {
      "epoch": 83.21,
      "learning_rate": 8.905113268608414e-06,
      "loss": 0.0429,
      "step": 214270
    },
    {
      "epoch": 83.22,
      "learning_rate": 8.904595469255664e-06,
      "loss": 0.0393,
      "step": 214280
    },
    {
      "epoch": 83.22,
      "learning_rate": 8.904077669902914e-06,
      "loss": 0.0764,
      "step": 214290
    },
    {
      "epoch": 83.22,
      "learning_rate": 8.903559870550163e-06,
      "loss": 0.0189,
      "step": 214300
    },
    {
      "epoch": 83.23,
      "learning_rate": 8.903042071197411e-06,
      "loss": 0.0022,
      "step": 214310
    },
    {
      "epoch": 83.23,
      "learning_rate": 8.90252427184466e-06,
      "loss": 0.1649,
      "step": 214320
    },
    {
      "epoch": 83.23,
      "learning_rate": 8.90200647249191e-06,
      "loss": 0.0724,
      "step": 214330
    },
    {
      "epoch": 83.24,
      "learning_rate": 8.90148867313916e-06,
      "loss": 0.054,
      "step": 214340
    },
    {
      "epoch": 83.24,
      "learning_rate": 8.900970873786408e-06,
      "loss": 0.0529,
      "step": 214350
    },
    {
      "epoch": 83.25,
      "learning_rate": 8.900453074433658e-06,
      "loss": 0.0237,
      "step": 214360
    },
    {
      "epoch": 83.25,
      "learning_rate": 8.899935275080907e-06,
      "loss": 0.0179,
      "step": 214370
    },
    {
      "epoch": 83.25,
      "learning_rate": 8.899417475728157e-06,
      "loss": 0.0168,
      "step": 214380
    },
    {
      "epoch": 83.26,
      "learning_rate": 8.898899676375405e-06,
      "loss": 0.0198,
      "step": 214390
    },
    {
      "epoch": 83.26,
      "learning_rate": 8.898381877022654e-06,
      "loss": 0.0487,
      "step": 214400
    },
    {
      "epoch": 83.27,
      "learning_rate": 8.897864077669904e-06,
      "loss": 0.0765,
      "step": 214410
    },
    {
      "epoch": 83.27,
      "learning_rate": 8.897346278317152e-06,
      "loss": 0.0168,
      "step": 214420
    },
    {
      "epoch": 83.27,
      "learning_rate": 8.896828478964401e-06,
      "loss": 0.0656,
      "step": 214430
    },
    {
      "epoch": 83.28,
      "learning_rate": 8.896310679611651e-06,
      "loss": 0.0444,
      "step": 214440
    },
    {
      "epoch": 83.28,
      "learning_rate": 8.8957928802589e-06,
      "loss": 0.053,
      "step": 214450
    },
    {
      "epoch": 83.29,
      "learning_rate": 8.895275080906149e-06,
      "loss": 0.0092,
      "step": 214460
    },
    {
      "epoch": 83.29,
      "learning_rate": 8.894757281553398e-06,
      "loss": 0.1552,
      "step": 214470
    },
    {
      "epoch": 83.29,
      "learning_rate": 8.894239482200648e-06,
      "loss": 0.0823,
      "step": 214480
    },
    {
      "epoch": 83.3,
      "learning_rate": 8.893721682847898e-06,
      "loss": 0.0751,
      "step": 214490
    },
    {
      "epoch": 83.3,
      "learning_rate": 8.893203883495145e-06,
      "loss": 0.0404,
      "step": 214500
    },
    {
      "epoch": 83.3,
      "learning_rate": 8.892686084142395e-06,
      "loss": 0.0222,
      "step": 214510
    },
    {
      "epoch": 83.31,
      "learning_rate": 8.892168284789645e-06,
      "loss": 0.0298,
      "step": 214520
    },
    {
      "epoch": 83.31,
      "learning_rate": 8.891650485436894e-06,
      "loss": 0.0313,
      "step": 214530
    },
    {
      "epoch": 83.32,
      "learning_rate": 8.891132686084142e-06,
      "loss": 0.1718,
      "step": 214540
    },
    {
      "epoch": 83.32,
      "learning_rate": 8.890614886731392e-06,
      "loss": 0.1025,
      "step": 214550
    },
    {
      "epoch": 83.32,
      "learning_rate": 8.890097087378641e-06,
      "loss": 0.0083,
      "step": 214560
    },
    {
      "epoch": 83.33,
      "learning_rate": 8.88957928802589e-06,
      "loss": 0.0212,
      "step": 214570
    },
    {
      "epoch": 83.33,
      "learning_rate": 8.889061488673139e-06,
      "loss": 0.0219,
      "step": 214580
    },
    {
      "epoch": 83.34,
      "learning_rate": 8.888543689320389e-06,
      "loss": 0.0744,
      "step": 214590
    },
    {
      "epoch": 83.34,
      "learning_rate": 8.888025889967638e-06,
      "loss": 0.1091,
      "step": 214600
    },
    {
      "epoch": 83.34,
      "learning_rate": 8.887508090614886e-06,
      "loss": 0.0032,
      "step": 214610
    },
    {
      "epoch": 83.35,
      "learning_rate": 8.886990291262136e-06,
      "loss": 0.0749,
      "step": 214620
    },
    {
      "epoch": 83.35,
      "learning_rate": 8.886472491909385e-06,
      "loss": 0.0822,
      "step": 214630
    },
    {
      "epoch": 83.36,
      "learning_rate": 8.885954692556635e-06,
      "loss": 0.0545,
      "step": 214640
    },
    {
      "epoch": 83.36,
      "learning_rate": 8.885436893203883e-06,
      "loss": 0.0108,
      "step": 214650
    },
    {
      "epoch": 83.36,
      "learning_rate": 8.884919093851133e-06,
      "loss": 0.0044,
      "step": 214660
    },
    {
      "epoch": 83.37,
      "learning_rate": 8.884401294498382e-06,
      "loss": 0.0693,
      "step": 214670
    },
    {
      "epoch": 83.37,
      "learning_rate": 8.883883495145632e-06,
      "loss": 0.0358,
      "step": 214680
    },
    {
      "epoch": 83.37,
      "learning_rate": 8.88336569579288e-06,
      "loss": 0.0135,
      "step": 214690
    },
    {
      "epoch": 83.38,
      "learning_rate": 8.88284789644013e-06,
      "loss": 0.0189,
      "step": 214700
    },
    {
      "epoch": 83.38,
      "learning_rate": 8.882330097087379e-06,
      "loss": 0.0468,
      "step": 214710
    },
    {
      "epoch": 83.39,
      "learning_rate": 8.881812297734629e-06,
      "loss": 0.0198,
      "step": 214720
    },
    {
      "epoch": 83.39,
      "learning_rate": 8.881294498381878e-06,
      "loss": 0.0001,
      "step": 214730
    },
    {
      "epoch": 83.39,
      "learning_rate": 8.880776699029126e-06,
      "loss": 0.1825,
      "step": 214740
    },
    {
      "epoch": 83.4,
      "learning_rate": 8.880258899676376e-06,
      "loss": 0.2035,
      "step": 214750
    },
    {
      "epoch": 83.4,
      "learning_rate": 8.879741100323625e-06,
      "loss": 0.066,
      "step": 214760
    },
    {
      "epoch": 83.41,
      "learning_rate": 8.879223300970875e-06,
      "loss": 0.1557,
      "step": 214770
    },
    {
      "epoch": 83.41,
      "learning_rate": 8.878705501618123e-06,
      "loss": 0.0627,
      "step": 214780
    },
    {
      "epoch": 83.41,
      "learning_rate": 8.878187702265372e-06,
      "loss": 0.0228,
      "step": 214790
    },
    {
      "epoch": 83.42,
      "learning_rate": 8.877669902912622e-06,
      "loss": 0.0078,
      "step": 214800
    },
    {
      "epoch": 83.42,
      "learning_rate": 8.877152103559872e-06,
      "loss": 0.0004,
      "step": 214810
    },
    {
      "epoch": 83.43,
      "learning_rate": 8.87663430420712e-06,
      "loss": 0.1188,
      "step": 214820
    },
    {
      "epoch": 83.43,
      "learning_rate": 8.87611650485437e-06,
      "loss": 0.0374,
      "step": 214830
    },
    {
      "epoch": 83.43,
      "learning_rate": 8.875598705501619e-06,
      "loss": 0.0693,
      "step": 214840
    },
    {
      "epoch": 83.44,
      "learning_rate": 8.875080906148869e-06,
      "loss": 0.0133,
      "step": 214850
    },
    {
      "epoch": 83.44,
      "learning_rate": 8.874563106796116e-06,
      "loss": 0.0125,
      "step": 214860
    },
    {
      "epoch": 83.44,
      "learning_rate": 8.874045307443366e-06,
      "loss": 0.0244,
      "step": 214870
    },
    {
      "epoch": 83.45,
      "learning_rate": 8.873527508090616e-06,
      "loss": 0.1499,
      "step": 214880
    },
    {
      "epoch": 83.45,
      "learning_rate": 8.873009708737865e-06,
      "loss": 0.0141,
      "step": 214890
    },
    {
      "epoch": 83.46,
      "learning_rate": 8.872491909385113e-06,
      "loss": 0.0682,
      "step": 214900
    },
    {
      "epoch": 83.46,
      "learning_rate": 8.871974110032363e-06,
      "loss": 0.0065,
      "step": 214910
    },
    {
      "epoch": 83.46,
      "learning_rate": 8.871456310679612e-06,
      "loss": 0.0311,
      "step": 214920
    },
    {
      "epoch": 83.47,
      "learning_rate": 8.870938511326862e-06,
      "loss": 0.0119,
      "step": 214930
    },
    {
      "epoch": 83.47,
      "learning_rate": 8.87042071197411e-06,
      "loss": 0.0419,
      "step": 214940
    },
    {
      "epoch": 83.48,
      "learning_rate": 8.86990291262136e-06,
      "loss": 0.2019,
      "step": 214950
    },
    {
      "epoch": 83.48,
      "learning_rate": 8.86938511326861e-06,
      "loss": 0.1012,
      "step": 214960
    },
    {
      "epoch": 83.48,
      "learning_rate": 8.868867313915859e-06,
      "loss": 0.0468,
      "step": 214970
    },
    {
      "epoch": 83.49,
      "learning_rate": 8.868349514563107e-06,
      "loss": 0.0715,
      "step": 214980
    },
    {
      "epoch": 83.49,
      "learning_rate": 8.867831715210356e-06,
      "loss": 0.0452,
      "step": 214990
    },
    {
      "epoch": 83.5,
      "learning_rate": 8.867313915857606e-06,
      "loss": 0.1162,
      "step": 215000
    },
    {
      "epoch": 83.5,
      "learning_rate": 8.866796116504856e-06,
      "loss": 0.063,
      "step": 215010
    },
    {
      "epoch": 83.5,
      "learning_rate": 8.866278317152104e-06,
      "loss": 0.2283,
      "step": 215020
    },
    {
      "epoch": 83.51,
      "learning_rate": 8.865760517799353e-06,
      "loss": 0.0829,
      "step": 215030
    },
    {
      "epoch": 83.51,
      "learning_rate": 8.865242718446603e-06,
      "loss": 0.0742,
      "step": 215040
    },
    {
      "epoch": 83.51,
      "learning_rate": 8.864724919093852e-06,
      "loss": 0.0088,
      "step": 215050
    },
    {
      "epoch": 83.52,
      "learning_rate": 8.8642071197411e-06,
      "loss": 0.0195,
      "step": 215060
    },
    {
      "epoch": 83.52,
      "learning_rate": 8.86368932038835e-06,
      "loss": 0.03,
      "step": 215070
    },
    {
      "epoch": 83.53,
      "learning_rate": 8.8631715210356e-06,
      "loss": 0.0994,
      "step": 215080
    },
    {
      "epoch": 83.53,
      "learning_rate": 8.86265372168285e-06,
      "loss": 0.1239,
      "step": 215090
    },
    {
      "epoch": 83.53,
      "learning_rate": 8.862135922330097e-06,
      "loss": 0.0786,
      "step": 215100
    },
    {
      "epoch": 83.54,
      "learning_rate": 8.861618122977347e-06,
      "loss": 0.1017,
      "step": 215110
    },
    {
      "epoch": 83.54,
      "learning_rate": 8.861100323624596e-06,
      "loss": 0.0781,
      "step": 215120
    },
    {
      "epoch": 83.55,
      "learning_rate": 8.860582524271846e-06,
      "loss": 0.0346,
      "step": 215130
    },
    {
      "epoch": 83.55,
      "learning_rate": 8.860064724919094e-06,
      "loss": 0.122,
      "step": 215140
    },
    {
      "epoch": 83.55,
      "learning_rate": 8.859546925566344e-06,
      "loss": 0.0039,
      "step": 215150
    },
    {
      "epoch": 83.56,
      "learning_rate": 8.859029126213593e-06,
      "loss": 0.1319,
      "step": 215160
    },
    {
      "epoch": 83.56,
      "learning_rate": 8.858511326860843e-06,
      "loss": 0.016,
      "step": 215170
    },
    {
      "epoch": 83.57,
      "learning_rate": 8.85799352750809e-06,
      "loss": 0.0245,
      "step": 215180
    },
    {
      "epoch": 83.57,
      "learning_rate": 8.85747572815534e-06,
      "loss": 0.047,
      "step": 215190
    },
    {
      "epoch": 83.57,
      "learning_rate": 8.85695792880259e-06,
      "loss": 0.0337,
      "step": 215200
    },
    {
      "epoch": 83.58,
      "learning_rate": 8.85644012944984e-06,
      "loss": 0.1295,
      "step": 215210
    },
    {
      "epoch": 83.58,
      "learning_rate": 8.855922330097087e-06,
      "loss": 0.0565,
      "step": 215220
    },
    {
      "epoch": 83.58,
      "learning_rate": 8.855404530744337e-06,
      "loss": 0.0267,
      "step": 215230
    },
    {
      "epoch": 83.59,
      "learning_rate": 8.854886731391587e-06,
      "loss": 0.016,
      "step": 215240
    },
    {
      "epoch": 83.59,
      "learning_rate": 8.854368932038836e-06,
      "loss": 0.0772,
      "step": 215250
    },
    {
      "epoch": 83.6,
      "learning_rate": 8.853851132686084e-06,
      "loss": 0.0591,
      "step": 215260
    },
    {
      "epoch": 83.6,
      "learning_rate": 8.853333333333334e-06,
      "loss": 0.1231,
      "step": 215270
    },
    {
      "epoch": 83.6,
      "learning_rate": 8.852815533980583e-06,
      "loss": 0.0927,
      "step": 215280
    },
    {
      "epoch": 83.61,
      "learning_rate": 8.852297734627833e-06,
      "loss": 0.0002,
      "step": 215290
    },
    {
      "epoch": 83.61,
      "learning_rate": 8.851779935275083e-06,
      "loss": 0.1272,
      "step": 215300
    },
    {
      "epoch": 83.62,
      "learning_rate": 8.85126213592233e-06,
      "loss": 0.0886,
      "step": 215310
    },
    {
      "epoch": 83.62,
      "learning_rate": 8.85074433656958e-06,
      "loss": 0.0854,
      "step": 215320
    },
    {
      "epoch": 83.62,
      "learning_rate": 8.85022653721683e-06,
      "loss": 0.0002,
      "step": 215330
    },
    {
      "epoch": 83.63,
      "learning_rate": 8.84970873786408e-06,
      "loss": 0.0048,
      "step": 215340
    },
    {
      "epoch": 83.63,
      "learning_rate": 8.849190938511327e-06,
      "loss": 0.0356,
      "step": 215350
    },
    {
      "epoch": 83.63,
      "learning_rate": 8.848673139158577e-06,
      "loss": 0.0718,
      "step": 215360
    },
    {
      "epoch": 83.64,
      "learning_rate": 8.848155339805827e-06,
      "loss": 0.0477,
      "step": 215370
    },
    {
      "epoch": 83.64,
      "learning_rate": 8.847637540453076e-06,
      "loss": 0.0518,
      "step": 215380
    },
    {
      "epoch": 83.65,
      "learning_rate": 8.847119741100324e-06,
      "loss": 0.0438,
      "step": 215390
    },
    {
      "epoch": 83.65,
      "learning_rate": 8.846601941747574e-06,
      "loss": 0.0002,
      "step": 215400
    },
    {
      "epoch": 83.65,
      "learning_rate": 8.846084142394823e-06,
      "loss": 0.0879,
      "step": 215410
    },
    {
      "epoch": 83.66,
      "learning_rate": 8.845566343042073e-06,
      "loss": 0.0007,
      "step": 215420
    },
    {
      "epoch": 83.66,
      "learning_rate": 8.845048543689321e-06,
      "loss": 0.1372,
      "step": 215430
    },
    {
      "epoch": 83.67,
      "learning_rate": 8.84453074433657e-06,
      "loss": 0.119,
      "step": 215440
    },
    {
      "epoch": 83.67,
      "learning_rate": 8.84401294498382e-06,
      "loss": 0.0847,
      "step": 215450
    },
    {
      "epoch": 83.67,
      "learning_rate": 8.84349514563107e-06,
      "loss": 0.0223,
      "step": 215460
    },
    {
      "epoch": 83.68,
      "learning_rate": 8.842977346278318e-06,
      "loss": 0.0241,
      "step": 215470
    },
    {
      "epoch": 83.68,
      "learning_rate": 8.842459546925567e-06,
      "loss": 0.0065,
      "step": 215480
    },
    {
      "epoch": 83.69,
      "learning_rate": 8.841941747572817e-06,
      "loss": 0.0808,
      "step": 215490
    },
    {
      "epoch": 83.69,
      "learning_rate": 8.841423948220067e-06,
      "loss": 0.0267,
      "step": 215500
    },
    {
      "epoch": 83.69,
      "learning_rate": 8.840906148867315e-06,
      "loss": 0.011,
      "step": 215510
    },
    {
      "epoch": 83.7,
      "learning_rate": 8.840388349514564e-06,
      "loss": 0.0317,
      "step": 215520
    },
    {
      "epoch": 83.7,
      "learning_rate": 8.839870550161814e-06,
      "loss": 0.1552,
      "step": 215530
    },
    {
      "epoch": 83.7,
      "learning_rate": 8.839352750809063e-06,
      "loss": 0.0489,
      "step": 215540
    },
    {
      "epoch": 83.71,
      "learning_rate": 8.838834951456311e-06,
      "loss": 0.0951,
      "step": 215550
    },
    {
      "epoch": 83.71,
      "learning_rate": 8.838317152103561e-06,
      "loss": 0.114,
      "step": 215560
    },
    {
      "epoch": 83.72,
      "learning_rate": 8.83779935275081e-06,
      "loss": 0.0991,
      "step": 215570
    },
    {
      "epoch": 83.72,
      "learning_rate": 8.83728155339806e-06,
      "loss": 0.0182,
      "step": 215580
    },
    {
      "epoch": 83.72,
      "learning_rate": 8.836763754045308e-06,
      "loss": 0.0762,
      "step": 215590
    },
    {
      "epoch": 83.73,
      "learning_rate": 8.836245954692558e-06,
      "loss": 0.0306,
      "step": 215600
    },
    {
      "epoch": 83.73,
      "learning_rate": 8.835728155339807e-06,
      "loss": 0.095,
      "step": 215610
    },
    {
      "epoch": 83.74,
      "learning_rate": 8.835210355987057e-06,
      "loss": 0.0215,
      "step": 215620
    },
    {
      "epoch": 83.74,
      "learning_rate": 8.834692556634305e-06,
      "loss": 0.1118,
      "step": 215630
    },
    {
      "epoch": 83.74,
      "learning_rate": 8.834174757281554e-06,
      "loss": 0.0015,
      "step": 215640
    },
    {
      "epoch": 83.75,
      "learning_rate": 8.833656957928804e-06,
      "loss": 0.0333,
      "step": 215650
    },
    {
      "epoch": 83.75,
      "learning_rate": 8.833139158576054e-06,
      "loss": 0.1615,
      "step": 215660
    },
    {
      "epoch": 83.76,
      "learning_rate": 8.832621359223302e-06,
      "loss": 0.0735,
      "step": 215670
    },
    {
      "epoch": 83.76,
      "learning_rate": 8.832103559870551e-06,
      "loss": 0.0251,
      "step": 215680
    },
    {
      "epoch": 83.76,
      "learning_rate": 8.831585760517801e-06,
      "loss": 0.1706,
      "step": 215690
    },
    {
      "epoch": 83.77,
      "learning_rate": 8.831067961165049e-06,
      "loss": 0.0211,
      "step": 215700
    },
    {
      "epoch": 83.77,
      "learning_rate": 8.830550161812298e-06,
      "loss": 0.0278,
      "step": 215710
    },
    {
      "epoch": 83.77,
      "learning_rate": 8.830032362459548e-06,
      "loss": 0.0661,
      "step": 215720
    },
    {
      "epoch": 83.78,
      "learning_rate": 8.829514563106798e-06,
      "loss": 0.1933,
      "step": 215730
    },
    {
      "epoch": 83.78,
      "learning_rate": 8.828996763754046e-06,
      "loss": 0.0851,
      "step": 215740
    },
    {
      "epoch": 83.79,
      "learning_rate": 8.828478964401295e-06,
      "loss": 0.0726,
      "step": 215750
    },
    {
      "epoch": 83.79,
      "learning_rate": 8.827961165048545e-06,
      "loss": 0.0646,
      "step": 215760
    },
    {
      "epoch": 83.79,
      "learning_rate": 8.827443365695794e-06,
      "loss": 0.0011,
      "step": 215770
    },
    {
      "epoch": 83.8,
      "learning_rate": 8.826925566343042e-06,
      "loss": 0.0132,
      "step": 215780
    },
    {
      "epoch": 83.8,
      "learning_rate": 8.826407766990292e-06,
      "loss": 0.0184,
      "step": 215790
    },
    {
      "epoch": 83.81,
      "learning_rate": 8.825889967637542e-06,
      "loss": 0.0629,
      "step": 215800
    },
    {
      "epoch": 83.81,
      "learning_rate": 8.825372168284791e-06,
      "loss": 0.1478,
      "step": 215810
    },
    {
      "epoch": 83.81,
      "learning_rate": 8.824854368932039e-06,
      "loss": 0.0122,
      "step": 215820
    },
    {
      "epoch": 83.82,
      "learning_rate": 8.824336569579289e-06,
      "loss": 0.0446,
      "step": 215830
    },
    {
      "epoch": 83.82,
      "learning_rate": 8.823818770226538e-06,
      "loss": 0.0767,
      "step": 215840
    },
    {
      "epoch": 83.83,
      "learning_rate": 8.823300970873788e-06,
      "loss": 0.0982,
      "step": 215850
    },
    {
      "epoch": 83.83,
      "learning_rate": 8.822783171521036e-06,
      "loss": 0.0168,
      "step": 215860
    },
    {
      "epoch": 83.83,
      "learning_rate": 8.822265372168286e-06,
      "loss": 0.0118,
      "step": 215870
    },
    {
      "epoch": 83.84,
      "learning_rate": 8.821747572815535e-06,
      "loss": 0.0273,
      "step": 215880
    },
    {
      "epoch": 83.84,
      "learning_rate": 8.821229773462783e-06,
      "loss": 0.0015,
      "step": 215890
    },
    {
      "epoch": 83.84,
      "learning_rate": 8.820711974110033e-06,
      "loss": 0.0967,
      "step": 215900
    },
    {
      "epoch": 83.85,
      "learning_rate": 8.820194174757282e-06,
      "loss": 0.013,
      "step": 215910
    },
    {
      "epoch": 83.85,
      "learning_rate": 8.819676375404532e-06,
      "loss": 0.0319,
      "step": 215920
    },
    {
      "epoch": 83.86,
      "learning_rate": 8.81915857605178e-06,
      "loss": 0.0492,
      "step": 215930
    },
    {
      "epoch": 83.86,
      "learning_rate": 8.81864077669903e-06,
      "loss": 0.0377,
      "step": 215940
    },
    {
      "epoch": 83.86,
      "learning_rate": 8.818122977346279e-06,
      "loss": 0.0511,
      "step": 215950
    },
    {
      "epoch": 83.87,
      "learning_rate": 8.817605177993529e-06,
      "loss": 0.0575,
      "step": 215960
    },
    {
      "epoch": 83.87,
      "learning_rate": 8.817087378640777e-06,
      "loss": 0.084,
      "step": 215970
    },
    {
      "epoch": 83.88,
      "learning_rate": 8.816569579288026e-06,
      "loss": 0.1589,
      "step": 215980
    },
    {
      "epoch": 83.88,
      "learning_rate": 8.816051779935276e-06,
      "loss": 0.0389,
      "step": 215990
    },
    {
      "epoch": 83.88,
      "learning_rate": 8.815533980582525e-06,
      "loss": 0.002,
      "step": 216000
    },
    {
      "epoch": 83.89,
      "learning_rate": 8.815016181229773e-06,
      "loss": 0.0727,
      "step": 216010
    },
    {
      "epoch": 83.89,
      "learning_rate": 8.814498381877023e-06,
      "loss": 0.0472,
      "step": 216020
    },
    {
      "epoch": 83.9,
      "learning_rate": 8.813980582524273e-06,
      "loss": 0.1136,
      "step": 216030
    },
    {
      "epoch": 83.9,
      "learning_rate": 8.81346278317152e-06,
      "loss": 0.0428,
      "step": 216040
    },
    {
      "epoch": 83.9,
      "learning_rate": 8.81294498381877e-06,
      "loss": 0.05,
      "step": 216050
    },
    {
      "epoch": 83.91,
      "learning_rate": 8.81242718446602e-06,
      "loss": 0.0922,
      "step": 216060
    },
    {
      "epoch": 83.91,
      "learning_rate": 8.81190938511327e-06,
      "loss": 0.2485,
      "step": 216070
    },
    {
      "epoch": 83.91,
      "learning_rate": 8.811391585760517e-06,
      "loss": 0.0204,
      "step": 216080
    },
    {
      "epoch": 83.92,
      "learning_rate": 8.810873786407767e-06,
      "loss": 0.1234,
      "step": 216090
    },
    {
      "epoch": 83.92,
      "learning_rate": 8.810355987055017e-06,
      "loss": 0.0004,
      "step": 216100
    },
    {
      "epoch": 83.93,
      "learning_rate": 8.809838187702266e-06,
      "loss": 0.0003,
      "step": 216110
    },
    {
      "epoch": 83.93,
      "learning_rate": 8.809320388349514e-06,
      "loss": 0.0002,
      "step": 216120
    },
    {
      "epoch": 83.93,
      "learning_rate": 8.808802588996764e-06,
      "loss": 0.018,
      "step": 216130
    },
    {
      "epoch": 83.94,
      "learning_rate": 8.808284789644013e-06,
      "loss": 0.0285,
      "step": 216140
    },
    {
      "epoch": 83.94,
      "learning_rate": 8.807766990291263e-06,
      "loss": 0.0996,
      "step": 216150
    },
    {
      "epoch": 83.95,
      "learning_rate": 8.807249190938511e-06,
      "loss": 0.0603,
      "step": 216160
    },
    {
      "epoch": 83.95,
      "learning_rate": 8.80673139158576e-06,
      "loss": 0.0576,
      "step": 216170
    },
    {
      "epoch": 83.95,
      "learning_rate": 8.80621359223301e-06,
      "loss": 0.0375,
      "step": 216180
    },
    {
      "epoch": 83.96,
      "learning_rate": 8.80569579288026e-06,
      "loss": 0.0823,
      "step": 216190
    },
    {
      "epoch": 83.96,
      "learning_rate": 8.805177993527508e-06,
      "loss": 0.0122,
      "step": 216200
    },
    {
      "epoch": 83.97,
      "learning_rate": 8.804660194174757e-06,
      "loss": 0.1507,
      "step": 216210
    },
    {
      "epoch": 83.97,
      "learning_rate": 8.804142394822007e-06,
      "loss": 0.2317,
      "step": 216220
    },
    {
      "epoch": 83.97,
      "learning_rate": 8.803624595469257e-06,
      "loss": 0.1536,
      "step": 216230
    },
    {
      "epoch": 83.98,
      "learning_rate": 8.803106796116504e-06,
      "loss": 0.0143,
      "step": 216240
    },
    {
      "epoch": 83.98,
      "learning_rate": 8.802588996763754e-06,
      "loss": 0.0279,
      "step": 216250
    },
    {
      "epoch": 83.98,
      "learning_rate": 8.802071197411004e-06,
      "loss": 0.1296,
      "step": 216260
    },
    {
      "epoch": 83.99,
      "learning_rate": 8.801553398058253e-06,
      "loss": 0.0577,
      "step": 216270
    },
    {
      "epoch": 83.99,
      "learning_rate": 8.801035598705501e-06,
      "loss": 0.0102,
      "step": 216280
    },
    {
      "epoch": 84.0,
      "learning_rate": 8.80051779935275e-06,
      "loss": 0.0372,
      "step": 216290
    },
    {
      "epoch": 84.0,
      "learning_rate": 8.8e-06,
      "loss": 0.0049,
      "step": 216300
    },
    {
      "epoch": 84.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.35490280389785767,
      "eval_runtime": 8.159,
      "eval_samples_per_second": 445.52,
      "eval_steps_per_second": 55.767,
      "step": 216300
    },
    {
      "epoch": 84.0,
      "learning_rate": 8.79948220064725e-06,
      "loss": 0.1428,
      "step": 216310
    },
    {
      "epoch": 84.01,
      "learning_rate": 8.798964401294498e-06,
      "loss": 0.02,
      "step": 216320
    },
    {
      "epoch": 84.01,
      "learning_rate": 8.798446601941748e-06,
      "loss": 0.0963,
      "step": 216330
    },
    {
      "epoch": 84.02,
      "learning_rate": 8.797928802588997e-06,
      "loss": 0.1017,
      "step": 216340
    },
    {
      "epoch": 84.02,
      "learning_rate": 8.797411003236247e-06,
      "loss": 0.0261,
      "step": 216350
    },
    {
      "epoch": 84.02,
      "learning_rate": 8.796893203883495e-06,
      "loss": 0.0274,
      "step": 216360
    },
    {
      "epoch": 84.03,
      "learning_rate": 8.796375404530744e-06,
      "loss": 0.1114,
      "step": 216370
    },
    {
      "epoch": 84.03,
      "learning_rate": 8.795857605177994e-06,
      "loss": 0.0248,
      "step": 216380
    },
    {
      "epoch": 84.03,
      "learning_rate": 8.795339805825244e-06,
      "loss": 0.1006,
      "step": 216390
    },
    {
      "epoch": 84.04,
      "learning_rate": 8.794822006472493e-06,
      "loss": 0.0223,
      "step": 216400
    },
    {
      "epoch": 84.04,
      "learning_rate": 8.794304207119741e-06,
      "loss": 0.195,
      "step": 216410
    },
    {
      "epoch": 84.05,
      "learning_rate": 8.79378640776699e-06,
      "loss": 0.116,
      "step": 216420
    },
    {
      "epoch": 84.05,
      "learning_rate": 8.79326860841424e-06,
      "loss": 0.0439,
      "step": 216430
    },
    {
      "epoch": 84.05,
      "learning_rate": 8.79275080906149e-06,
      "loss": 0.038,
      "step": 216440
    },
    {
      "epoch": 84.06,
      "learning_rate": 8.792233009708738e-06,
      "loss": 0.1371,
      "step": 216450
    },
    {
      "epoch": 84.06,
      "learning_rate": 8.791715210355988e-06,
      "loss": 0.0164,
      "step": 216460
    },
    {
      "epoch": 84.07,
      "learning_rate": 8.791197411003237e-06,
      "loss": 0.0555,
      "step": 216470
    },
    {
      "epoch": 84.07,
      "learning_rate": 8.790679611650487e-06,
      "loss": 0.0181,
      "step": 216480
    },
    {
      "epoch": 84.07,
      "learning_rate": 8.790161812297735e-06,
      "loss": 0.1066,
      "step": 216490
    },
    {
      "epoch": 84.08,
      "learning_rate": 8.789644012944984e-06,
      "loss": 0.0909,
      "step": 216500
    },
    {
      "epoch": 84.08,
      "learning_rate": 8.789126213592234e-06,
      "loss": 0.1629,
      "step": 216510
    },
    {
      "epoch": 84.09,
      "learning_rate": 8.788608414239484e-06,
      "loss": 0.1455,
      "step": 216520
    },
    {
      "epoch": 84.09,
      "learning_rate": 8.788090614886731e-06,
      "loss": 0.0657,
      "step": 216530
    },
    {
      "epoch": 84.09,
      "learning_rate": 8.787572815533981e-06,
      "loss": 0.0439,
      "step": 216540
    },
    {
      "epoch": 84.1,
      "learning_rate": 8.78705501618123e-06,
      "loss": 0.1125,
      "step": 216550
    },
    {
      "epoch": 84.1,
      "learning_rate": 8.78653721682848e-06,
      "loss": 0.0748,
      "step": 216560
    },
    {
      "epoch": 84.1,
      "learning_rate": 8.786019417475728e-06,
      "loss": 0.045,
      "step": 216570
    },
    {
      "epoch": 84.11,
      "learning_rate": 8.785501618122978e-06,
      "loss": 0.1503,
      "step": 216580
    },
    {
      "epoch": 84.11,
      "learning_rate": 8.784983818770228e-06,
      "loss": 0.0876,
      "step": 216590
    },
    {
      "epoch": 84.12,
      "learning_rate": 8.784466019417477e-06,
      "loss": 0.0132,
      "step": 216600
    },
    {
      "epoch": 84.12,
      "learning_rate": 8.783948220064725e-06,
      "loss": 0.1259,
      "step": 216610
    },
    {
      "epoch": 84.12,
      "learning_rate": 8.783430420711975e-06,
      "loss": 0.0339,
      "step": 216620
    },
    {
      "epoch": 84.13,
      "learning_rate": 8.782912621359224e-06,
      "loss": 0.107,
      "step": 216630
    },
    {
      "epoch": 84.13,
      "learning_rate": 8.782394822006474e-06,
      "loss": 0.016,
      "step": 216640
    },
    {
      "epoch": 84.14,
      "learning_rate": 8.781877022653722e-06,
      "loss": 0.0445,
      "step": 216650
    },
    {
      "epoch": 84.14,
      "learning_rate": 8.781359223300971e-06,
      "loss": 0.0286,
      "step": 216660
    },
    {
      "epoch": 84.14,
      "learning_rate": 8.780841423948221e-06,
      "loss": 0.0399,
      "step": 216670
    },
    {
      "epoch": 84.15,
      "learning_rate": 8.78032362459547e-06,
      "loss": 0.031,
      "step": 216680
    },
    {
      "epoch": 84.15,
      "learning_rate": 8.779805825242719e-06,
      "loss": 0.0032,
      "step": 216690
    },
    {
      "epoch": 84.16,
      "learning_rate": 8.779288025889968e-06,
      "loss": 0.0718,
      "step": 216700
    },
    {
      "epoch": 84.16,
      "learning_rate": 8.778770226537218e-06,
      "loss": 0.081,
      "step": 216710
    },
    {
      "epoch": 84.16,
      "learning_rate": 8.778252427184467e-06,
      "loss": 0.0455,
      "step": 216720
    },
    {
      "epoch": 84.17,
      "learning_rate": 8.777734627831715e-06,
      "loss": 0.0001,
      "step": 216730
    },
    {
      "epoch": 84.17,
      "learning_rate": 8.777216828478965e-06,
      "loss": 0.0315,
      "step": 216740
    },
    {
      "epoch": 84.17,
      "learning_rate": 8.776699029126215e-06,
      "loss": 0.017,
      "step": 216750
    },
    {
      "epoch": 84.18,
      "learning_rate": 8.776181229773464e-06,
      "loss": 0.1011,
      "step": 216760
    },
    {
      "epoch": 84.18,
      "learning_rate": 8.775663430420712e-06,
      "loss": 0.0818,
      "step": 216770
    },
    {
      "epoch": 84.19,
      "learning_rate": 8.775145631067962e-06,
      "loss": 0.0232,
      "step": 216780
    },
    {
      "epoch": 84.19,
      "learning_rate": 8.774627831715211e-06,
      "loss": 0.0331,
      "step": 216790
    },
    {
      "epoch": 84.19,
      "learning_rate": 8.774110032362461e-06,
      "loss": 0.0793,
      "step": 216800
    },
    {
      "epoch": 84.2,
      "learning_rate": 8.773592233009709e-06,
      "loss": 0.0746,
      "step": 216810
    },
    {
      "epoch": 84.2,
      "learning_rate": 8.773074433656959e-06,
      "loss": 0.0976,
      "step": 216820
    },
    {
      "epoch": 84.21,
      "learning_rate": 8.772556634304208e-06,
      "loss": 0.067,
      "step": 216830
    },
    {
      "epoch": 84.21,
      "learning_rate": 8.772038834951458e-06,
      "loss": 0.0884,
      "step": 216840
    },
    {
      "epoch": 84.21,
      "learning_rate": 8.771521035598706e-06,
      "loss": 0.0868,
      "step": 216850
    },
    {
      "epoch": 84.22,
      "learning_rate": 8.771003236245955e-06,
      "loss": 0.001,
      "step": 216860
    },
    {
      "epoch": 84.22,
      "learning_rate": 8.770485436893205e-06,
      "loss": 0.002,
      "step": 216870
    },
    {
      "epoch": 84.23,
      "learning_rate": 8.769967637540455e-06,
      "loss": 0.0474,
      "step": 216880
    },
    {
      "epoch": 84.23,
      "learning_rate": 8.769449838187703e-06,
      "loss": 0.1686,
      "step": 216890
    },
    {
      "epoch": 84.23,
      "learning_rate": 8.768932038834952e-06,
      "loss": 0.0803,
      "step": 216900
    },
    {
      "epoch": 84.24,
      "learning_rate": 8.768414239482202e-06,
      "loss": 0.0061,
      "step": 216910
    },
    {
      "epoch": 84.24,
      "learning_rate": 8.767896440129451e-06,
      "loss": 0.0091,
      "step": 216920
    },
    {
      "epoch": 84.24,
      "learning_rate": 8.7673786407767e-06,
      "loss": 0.0074,
      "step": 216930
    },
    {
      "epoch": 84.25,
      "learning_rate": 8.766860841423949e-06,
      "loss": 0.0012,
      "step": 216940
    },
    {
      "epoch": 84.25,
      "learning_rate": 8.766343042071199e-06,
      "loss": 0.0125,
      "step": 216950
    },
    {
      "epoch": 84.26,
      "learning_rate": 8.765825242718448e-06,
      "loss": 0.0371,
      "step": 216960
    },
    {
      "epoch": 84.26,
      "learning_rate": 8.765307443365698e-06,
      "loss": 0.0866,
      "step": 216970
    },
    {
      "epoch": 84.26,
      "learning_rate": 8.764789644012946e-06,
      "loss": 0.0515,
      "step": 216980
    },
    {
      "epoch": 84.27,
      "learning_rate": 8.764271844660195e-06,
      "loss": 0.0427,
      "step": 216990
    },
    {
      "epoch": 84.27,
      "learning_rate": 8.763754045307445e-06,
      "loss": 0.0635,
      "step": 217000
    },
    {
      "epoch": 84.28,
      "learning_rate": 8.763236245954695e-06,
      "loss": 0.0636,
      "step": 217010
    },
    {
      "epoch": 84.28,
      "learning_rate": 8.762718446601942e-06,
      "loss": 0.0156,
      "step": 217020
    },
    {
      "epoch": 84.28,
      "learning_rate": 8.762200647249192e-06,
      "loss": 0.038,
      "step": 217030
    },
    {
      "epoch": 84.29,
      "learning_rate": 8.761682847896442e-06,
      "loss": 0.0449,
      "step": 217040
    },
    {
      "epoch": 84.29,
      "learning_rate": 8.761165048543691e-06,
      "loss": 0.0042,
      "step": 217050
    },
    {
      "epoch": 84.3,
      "learning_rate": 8.76064724919094e-06,
      "loss": 0.0721,
      "step": 217060
    },
    {
      "epoch": 84.3,
      "learning_rate": 8.760129449838189e-06,
      "loss": 0.0233,
      "step": 217070
    },
    {
      "epoch": 84.3,
      "learning_rate": 8.759611650485438e-06,
      "loss": 0.1353,
      "step": 217080
    },
    {
      "epoch": 84.31,
      "learning_rate": 8.759093851132688e-06,
      "loss": 0.0975,
      "step": 217090
    },
    {
      "epoch": 84.31,
      "learning_rate": 8.758576051779936e-06,
      "loss": 0.0681,
      "step": 217100
    },
    {
      "epoch": 84.31,
      "learning_rate": 8.758058252427186e-06,
      "loss": 0.0273,
      "step": 217110
    },
    {
      "epoch": 84.32,
      "learning_rate": 8.757540453074435e-06,
      "loss": 0.1793,
      "step": 217120
    },
    {
      "epoch": 84.32,
      "learning_rate": 8.757022653721685e-06,
      "loss": 0.1479,
      "step": 217130
    },
    {
      "epoch": 84.33,
      "learning_rate": 8.756504854368933e-06,
      "loss": 0.1144,
      "step": 217140
    },
    {
      "epoch": 84.33,
      "learning_rate": 8.755987055016182e-06,
      "loss": 0.0196,
      "step": 217150
    },
    {
      "epoch": 84.33,
      "learning_rate": 8.755469255663432e-06,
      "loss": 0.15,
      "step": 217160
    },
    {
      "epoch": 84.34,
      "learning_rate": 8.75495145631068e-06,
      "loss": 0.0434,
      "step": 217170
    },
    {
      "epoch": 84.34,
      "learning_rate": 8.75443365695793e-06,
      "loss": 0.1618,
      "step": 217180
    },
    {
      "epoch": 84.35,
      "learning_rate": 8.75391585760518e-06,
      "loss": 0.0708,
      "step": 217190
    },
    {
      "epoch": 84.35,
      "learning_rate": 8.753398058252429e-06,
      "loss": 0.0901,
      "step": 217200
    },
    {
      "epoch": 84.35,
      "learning_rate": 8.752880258899677e-06,
      "loss": 0.0628,
      "step": 217210
    },
    {
      "epoch": 84.36,
      "learning_rate": 8.752362459546926e-06,
      "loss": 0.1114,
      "step": 217220
    },
    {
      "epoch": 84.36,
      "learning_rate": 8.751844660194176e-06,
      "loss": 0.0763,
      "step": 217230
    },
    {
      "epoch": 84.37,
      "learning_rate": 8.751326860841426e-06,
      "loss": 0.0352,
      "step": 217240
    },
    {
      "epoch": 84.37,
      "learning_rate": 8.750809061488674e-06,
      "loss": 0.0816,
      "step": 217250
    },
    {
      "epoch": 84.37,
      "learning_rate": 8.750291262135923e-06,
      "loss": 0.0877,
      "step": 217260
    },
    {
      "epoch": 84.38,
      "learning_rate": 8.749773462783173e-06,
      "loss": 0.0328,
      "step": 217270
    },
    {
      "epoch": 84.38,
      "learning_rate": 8.749255663430422e-06,
      "loss": 0.1195,
      "step": 217280
    },
    {
      "epoch": 84.38,
      "learning_rate": 8.74873786407767e-06,
      "loss": 0.0867,
      "step": 217290
    },
    {
      "epoch": 84.39,
      "learning_rate": 8.74822006472492e-06,
      "loss": 0.0692,
      "step": 217300
    },
    {
      "epoch": 84.39,
      "learning_rate": 8.74770226537217e-06,
      "loss": 0.0863,
      "step": 217310
    },
    {
      "epoch": 84.4,
      "learning_rate": 8.747184466019417e-06,
      "loss": 0.1513,
      "step": 217320
    },
    {
      "epoch": 84.4,
      "learning_rate": 8.746666666666667e-06,
      "loss": 0.0147,
      "step": 217330
    },
    {
      "epoch": 84.4,
      "learning_rate": 8.746148867313917e-06,
      "loss": 0.0003,
      "step": 217340
    },
    {
      "epoch": 84.41,
      "learning_rate": 8.745631067961166e-06,
      "loss": 0.0639,
      "step": 217350
    },
    {
      "epoch": 84.41,
      "learning_rate": 8.745113268608414e-06,
      "loss": 0.0519,
      "step": 217360
    },
    {
      "epoch": 84.42,
      "learning_rate": 8.744595469255664e-06,
      "loss": 0.0527,
      "step": 217370
    },
    {
      "epoch": 84.42,
      "learning_rate": 8.744077669902913e-06,
      "loss": 0.0447,
      "step": 217380
    },
    {
      "epoch": 84.42,
      "learning_rate": 8.743559870550163e-06,
      "loss": 0.0295,
      "step": 217390
    },
    {
      "epoch": 84.43,
      "learning_rate": 8.743042071197411e-06,
      "loss": 0.1065,
      "step": 217400
    },
    {
      "epoch": 84.43,
      "learning_rate": 8.74252427184466e-06,
      "loss": 0.0124,
      "step": 217410
    },
    {
      "epoch": 84.43,
      "learning_rate": 8.74200647249191e-06,
      "loss": 0.0042,
      "step": 217420
    },
    {
      "epoch": 84.44,
      "learning_rate": 8.74148867313916e-06,
      "loss": 0.0754,
      "step": 217430
    },
    {
      "epoch": 84.44,
      "learning_rate": 8.740970873786408e-06,
      "loss": 0.0052,
      "step": 217440
    },
    {
      "epoch": 84.45,
      "learning_rate": 8.740453074433657e-06,
      "loss": 0.0923,
      "step": 217450
    },
    {
      "epoch": 84.45,
      "learning_rate": 8.739935275080907e-06,
      "loss": 0.0206,
      "step": 217460
    },
    {
      "epoch": 84.45,
      "learning_rate": 8.739417475728157e-06,
      "loss": 0.0666,
      "step": 217470
    },
    {
      "epoch": 84.46,
      "learning_rate": 8.738899676375405e-06,
      "loss": 0.0713,
      "step": 217480
    },
    {
      "epoch": 84.46,
      "learning_rate": 8.738381877022654e-06,
      "loss": 0.0126,
      "step": 217490
    },
    {
      "epoch": 84.47,
      "learning_rate": 8.737864077669904e-06,
      "loss": 0.0005,
      "step": 217500
    },
    {
      "epoch": 84.47,
      "learning_rate": 8.737346278317152e-06,
      "loss": 0.1545,
      "step": 217510
    },
    {
      "epoch": 84.47,
      "learning_rate": 8.736828478964401e-06,
      "loss": 0.0166,
      "step": 217520
    },
    {
      "epoch": 84.48,
      "learning_rate": 8.736310679611651e-06,
      "loss": 0.0358,
      "step": 217530
    },
    {
      "epoch": 84.48,
      "learning_rate": 8.7357928802589e-06,
      "loss": 0.0331,
      "step": 217540
    },
    {
      "epoch": 84.49,
      "learning_rate": 8.735275080906148e-06,
      "loss": 0.0097,
      "step": 217550
    },
    {
      "epoch": 84.49,
      "learning_rate": 8.734757281553398e-06,
      "loss": 0.1009,
      "step": 217560
    },
    {
      "epoch": 84.49,
      "learning_rate": 8.734239482200648e-06,
      "loss": 0.0648,
      "step": 217570
    },
    {
      "epoch": 84.5,
      "learning_rate": 8.733721682847897e-06,
      "loss": 0.0189,
      "step": 217580
    },
    {
      "epoch": 84.5,
      "learning_rate": 8.733203883495145e-06,
      "loss": 0.0449,
      "step": 217590
    },
    {
      "epoch": 84.5,
      "learning_rate": 8.732686084142395e-06,
      "loss": 0.056,
      "step": 217600
    },
    {
      "epoch": 84.51,
      "learning_rate": 8.732168284789645e-06,
      "loss": 0.0946,
      "step": 217610
    },
    {
      "epoch": 84.51,
      "learning_rate": 8.731650485436894e-06,
      "loss": 0.0687,
      "step": 217620
    },
    {
      "epoch": 84.52,
      "learning_rate": 8.731132686084142e-06,
      "loss": 0.0817,
      "step": 217630
    },
    {
      "epoch": 84.52,
      "learning_rate": 8.730614886731392e-06,
      "loss": 0.1189,
      "step": 217640
    },
    {
      "epoch": 84.52,
      "learning_rate": 8.730097087378641e-06,
      "loss": 0.0893,
      "step": 217650
    },
    {
      "epoch": 84.53,
      "learning_rate": 8.729579288025891e-06,
      "loss": 0.0001,
      "step": 217660
    },
    {
      "epoch": 84.53,
      "learning_rate": 8.729061488673139e-06,
      "loss": 0.0159,
      "step": 217670
    },
    {
      "epoch": 84.54,
      "learning_rate": 8.728543689320388e-06,
      "loss": 0.0485,
      "step": 217680
    },
    {
      "epoch": 84.54,
      "learning_rate": 8.728025889967638e-06,
      "loss": 0.0937,
      "step": 217690
    },
    {
      "epoch": 84.54,
      "learning_rate": 8.727508090614888e-06,
      "loss": 0.0894,
      "step": 217700
    },
    {
      "epoch": 84.55,
      "learning_rate": 8.726990291262136e-06,
      "loss": 0.1008,
      "step": 217710
    },
    {
      "epoch": 84.55,
      "learning_rate": 8.726472491909385e-06,
      "loss": 0.0005,
      "step": 217720
    },
    {
      "epoch": 84.56,
      "learning_rate": 8.725954692556635e-06,
      "loss": 0.0004,
      "step": 217730
    },
    {
      "epoch": 84.56,
      "learning_rate": 8.725436893203884e-06,
      "loss": 0.0662,
      "step": 217740
    },
    {
      "epoch": 84.56,
      "learning_rate": 8.724919093851132e-06,
      "loss": 0.1215,
      "step": 217750
    },
    {
      "epoch": 84.57,
      "learning_rate": 8.724401294498382e-06,
      "loss": 0.0011,
      "step": 217760
    },
    {
      "epoch": 84.57,
      "learning_rate": 8.723883495145632e-06,
      "loss": 0.0452,
      "step": 217770
    },
    {
      "epoch": 84.57,
      "learning_rate": 8.723365695792881e-06,
      "loss": 0.0338,
      "step": 217780
    },
    {
      "epoch": 84.58,
      "learning_rate": 8.72284789644013e-06,
      "loss": 0.0005,
      "step": 217790
    },
    {
      "epoch": 84.58,
      "learning_rate": 8.722330097087379e-06,
      "loss": 0.0741,
      "step": 217800
    },
    {
      "epoch": 84.59,
      "learning_rate": 8.721812297734628e-06,
      "loss": 0.0395,
      "step": 217810
    },
    {
      "epoch": 84.59,
      "learning_rate": 8.721294498381878e-06,
      "loss": 0.2081,
      "step": 217820
    },
    {
      "epoch": 84.59,
      "learning_rate": 8.720776699029126e-06,
      "loss": 0.0311,
      "step": 217830
    },
    {
      "epoch": 84.6,
      "learning_rate": 8.720258899676376e-06,
      "loss": 0.0512,
      "step": 217840
    },
    {
      "epoch": 84.6,
      "learning_rate": 8.719741100323625e-06,
      "loss": 0.1498,
      "step": 217850
    },
    {
      "epoch": 84.61,
      "learning_rate": 8.719223300970875e-06,
      "loss": 0.1302,
      "step": 217860
    },
    {
      "epoch": 84.61,
      "learning_rate": 8.718705501618123e-06,
      "loss": 0.0002,
      "step": 217870
    },
    {
      "epoch": 84.61,
      "learning_rate": 8.718187702265372e-06,
      "loss": 0.1226,
      "step": 217880
    },
    {
      "epoch": 84.62,
      "learning_rate": 8.717669902912622e-06,
      "loss": 0.0303,
      "step": 217890
    },
    {
      "epoch": 84.62,
      "learning_rate": 8.717152103559872e-06,
      "loss": 0.0673,
      "step": 217900
    },
    {
      "epoch": 84.63,
      "learning_rate": 8.71663430420712e-06,
      "loss": 0.1485,
      "step": 217910
    },
    {
      "epoch": 84.63,
      "learning_rate": 8.716116504854369e-06,
      "loss": 0.0125,
      "step": 217920
    },
    {
      "epoch": 84.63,
      "learning_rate": 8.715598705501619e-06,
      "loss": 0.1405,
      "step": 217930
    },
    {
      "epoch": 84.64,
      "learning_rate": 8.715080906148868e-06,
      "loss": 0.041,
      "step": 217940
    },
    {
      "epoch": 84.64,
      "learning_rate": 8.714563106796116e-06,
      "loss": 0.0159,
      "step": 217950
    },
    {
      "epoch": 84.64,
      "learning_rate": 8.714045307443366e-06,
      "loss": 0.1471,
      "step": 217960
    },
    {
      "epoch": 84.65,
      "learning_rate": 8.713527508090616e-06,
      "loss": 0.0051,
      "step": 217970
    },
    {
      "epoch": 84.65,
      "learning_rate": 8.713009708737865e-06,
      "loss": 0.0796,
      "step": 217980
    },
    {
      "epoch": 84.66,
      "learning_rate": 8.712491909385113e-06,
      "loss": 0.0603,
      "step": 217990
    },
    {
      "epoch": 84.66,
      "learning_rate": 8.711974110032363e-06,
      "loss": 0.0078,
      "step": 218000
    },
    {
      "epoch": 84.66,
      "learning_rate": 8.711456310679612e-06,
      "loss": 0.0374,
      "step": 218010
    },
    {
      "epoch": 84.67,
      "learning_rate": 8.710938511326862e-06,
      "loss": 0.0116,
      "step": 218020
    },
    {
      "epoch": 84.67,
      "learning_rate": 8.71042071197411e-06,
      "loss": 0.1142,
      "step": 218030
    },
    {
      "epoch": 84.68,
      "learning_rate": 8.70990291262136e-06,
      "loss": 0.1628,
      "step": 218040
    },
    {
      "epoch": 84.68,
      "learning_rate": 8.709385113268609e-06,
      "loss": 0.0564,
      "step": 218050
    },
    {
      "epoch": 84.68,
      "learning_rate": 8.708867313915859e-06,
      "loss": 0.0161,
      "step": 218060
    },
    {
      "epoch": 84.69,
      "learning_rate": 8.708349514563107e-06,
      "loss": 0.0096,
      "step": 218070
    },
    {
      "epoch": 84.69,
      "learning_rate": 8.707831715210356e-06,
      "loss": 0.0164,
      "step": 218080
    },
    {
      "epoch": 84.7,
      "learning_rate": 8.707313915857606e-06,
      "loss": 0.0659,
      "step": 218090
    },
    {
      "epoch": 84.7,
      "learning_rate": 8.706796116504855e-06,
      "loss": 0.0141,
      "step": 218100
    },
    {
      "epoch": 84.7,
      "learning_rate": 8.706278317152105e-06,
      "loss": 0.0036,
      "step": 218110
    },
    {
      "epoch": 84.71,
      "learning_rate": 8.705760517799353e-06,
      "loss": 0.0607,
      "step": 218120
    },
    {
      "epoch": 84.71,
      "learning_rate": 8.705242718446603e-06,
      "loss": 0.0737,
      "step": 218130
    },
    {
      "epoch": 84.71,
      "learning_rate": 8.704724919093852e-06,
      "loss": 0.0557,
      "step": 218140
    },
    {
      "epoch": 84.72,
      "learning_rate": 8.704207119741102e-06,
      "loss": 0.1873,
      "step": 218150
    },
    {
      "epoch": 84.72,
      "learning_rate": 8.70368932038835e-06,
      "loss": 0.0148,
      "step": 218160
    },
    {
      "epoch": 84.73,
      "learning_rate": 8.7031715210356e-06,
      "loss": 0.0011,
      "step": 218170
    },
    {
      "epoch": 84.73,
      "learning_rate": 8.702653721682849e-06,
      "loss": 0.1181,
      "step": 218180
    },
    {
      "epoch": 84.73,
      "learning_rate": 8.702135922330099e-06,
      "loss": 0.0616,
      "step": 218190
    },
    {
      "epoch": 84.74,
      "learning_rate": 8.701618122977347e-06,
      "loss": 0.0215,
      "step": 218200
    },
    {
      "epoch": 84.74,
      "learning_rate": 8.701100323624596e-06,
      "loss": 0.1044,
      "step": 218210
    },
    {
      "epoch": 84.75,
      "learning_rate": 8.700582524271846e-06,
      "loss": 0.0301,
      "step": 218220
    },
    {
      "epoch": 84.75,
      "learning_rate": 8.700064724919095e-06,
      "loss": 0.0488,
      "step": 218230
    },
    {
      "epoch": 84.75,
      "learning_rate": 8.699546925566343e-06,
      "loss": 0.0014,
      "step": 218240
    },
    {
      "epoch": 84.76,
      "learning_rate": 8.699029126213593e-06,
      "loss": 0.0617,
      "step": 218250
    },
    {
      "epoch": 84.76,
      "learning_rate": 8.698511326860843e-06,
      "loss": 0.0535,
      "step": 218260
    },
    {
      "epoch": 84.77,
      "learning_rate": 8.697993527508092e-06,
      "loss": 0.0036,
      "step": 218270
    },
    {
      "epoch": 84.77,
      "learning_rate": 8.69747572815534e-06,
      "loss": 0.0569,
      "step": 218280
    },
    {
      "epoch": 84.77,
      "learning_rate": 8.69695792880259e-06,
      "loss": 0.0016,
      "step": 218290
    },
    {
      "epoch": 84.78,
      "learning_rate": 8.69644012944984e-06,
      "loss": 0.101,
      "step": 218300
    },
    {
      "epoch": 84.78,
      "learning_rate": 8.695922330097089e-06,
      "loss": 0.0077,
      "step": 218310
    },
    {
      "epoch": 84.78,
      "learning_rate": 8.695404530744337e-06,
      "loss": 0.0393,
      "step": 218320
    },
    {
      "epoch": 84.79,
      "learning_rate": 8.694886731391587e-06,
      "loss": 0.0282,
      "step": 218330
    },
    {
      "epoch": 84.79,
      "learning_rate": 8.694368932038836e-06,
      "loss": 0.0006,
      "step": 218340
    },
    {
      "epoch": 84.8,
      "learning_rate": 8.693851132686086e-06,
      "loss": 0.0091,
      "step": 218350
    },
    {
      "epoch": 84.8,
      "learning_rate": 8.693333333333334e-06,
      "loss": 0.0381,
      "step": 218360
    },
    {
      "epoch": 84.8,
      "learning_rate": 8.692815533980583e-06,
      "loss": 0.0081,
      "step": 218370
    },
    {
      "epoch": 84.81,
      "learning_rate": 8.692297734627833e-06,
      "loss": 0.0486,
      "step": 218380
    },
    {
      "epoch": 84.81,
      "learning_rate": 8.691779935275083e-06,
      "loss": 0.2375,
      "step": 218390
    },
    {
      "epoch": 84.82,
      "learning_rate": 8.69126213592233e-06,
      "loss": 0.0454,
      "step": 218400
    },
    {
      "epoch": 84.82,
      "learning_rate": 8.69074433656958e-06,
      "loss": 0.0624,
      "step": 218410
    },
    {
      "epoch": 84.82,
      "learning_rate": 8.69022653721683e-06,
      "loss": 0.0011,
      "step": 218420
    },
    {
      "epoch": 84.83,
      "learning_rate": 8.68970873786408e-06,
      "loss": 0.0106,
      "step": 218430
    },
    {
      "epoch": 84.83,
      "learning_rate": 8.689190938511327e-06,
      "loss": 0.1086,
      "step": 218440
    },
    {
      "epoch": 84.83,
      "learning_rate": 8.688673139158577e-06,
      "loss": 0.0467,
      "step": 218450
    },
    {
      "epoch": 84.84,
      "learning_rate": 8.688155339805826e-06,
      "loss": 0.0002,
      "step": 218460
    },
    {
      "epoch": 84.84,
      "learning_rate": 8.687637540453076e-06,
      "loss": 0.0427,
      "step": 218470
    },
    {
      "epoch": 84.85,
      "learning_rate": 8.687119741100324e-06,
      "loss": 0.1414,
      "step": 218480
    },
    {
      "epoch": 84.85,
      "learning_rate": 8.686601941747574e-06,
      "loss": 0.1014,
      "step": 218490
    },
    {
      "epoch": 84.85,
      "learning_rate": 8.686084142394823e-06,
      "loss": 0.0795,
      "step": 218500
    },
    {
      "epoch": 84.86,
      "learning_rate": 8.685566343042073e-06,
      "loss": 0.1764,
      "step": 218510
    },
    {
      "epoch": 84.86,
      "learning_rate": 8.68504854368932e-06,
      "loss": 0.0545,
      "step": 218520
    },
    {
      "epoch": 84.87,
      "learning_rate": 8.68453074433657e-06,
      "loss": 0.0603,
      "step": 218530
    },
    {
      "epoch": 84.87,
      "learning_rate": 8.68401294498382e-06,
      "loss": 0.0208,
      "step": 218540
    },
    {
      "epoch": 84.87,
      "learning_rate": 8.68349514563107e-06,
      "loss": 0.0018,
      "step": 218550
    },
    {
      "epoch": 84.88,
      "learning_rate": 8.682977346278318e-06,
      "loss": 0.0552,
      "step": 218560
    },
    {
      "epoch": 84.88,
      "learning_rate": 8.682459546925567e-06,
      "loss": 0.1088,
      "step": 218570
    },
    {
      "epoch": 84.89,
      "learning_rate": 8.681941747572817e-06,
      "loss": 0.0162,
      "step": 218580
    },
    {
      "epoch": 84.89,
      "learning_rate": 8.681423948220066e-06,
      "loss": 0.1549,
      "step": 218590
    },
    {
      "epoch": 84.89,
      "learning_rate": 8.680906148867314e-06,
      "loss": 0.0232,
      "step": 218600
    },
    {
      "epoch": 84.9,
      "learning_rate": 8.680388349514564e-06,
      "loss": 0.031,
      "step": 218610
    },
    {
      "epoch": 84.9,
      "learning_rate": 8.679870550161814e-06,
      "loss": 0.0956,
      "step": 218620
    },
    {
      "epoch": 84.9,
      "learning_rate": 8.679352750809063e-06,
      "loss": 0.1014,
      "step": 218630
    },
    {
      "epoch": 84.91,
      "learning_rate": 8.678834951456311e-06,
      "loss": 0.1449,
      "step": 218640
    },
    {
      "epoch": 84.91,
      "learning_rate": 8.67831715210356e-06,
      "loss": 0.015,
      "step": 218650
    },
    {
      "epoch": 84.92,
      "learning_rate": 8.67779935275081e-06,
      "loss": 0.0103,
      "step": 218660
    },
    {
      "epoch": 84.92,
      "learning_rate": 8.67728155339806e-06,
      "loss": 0.0809,
      "step": 218670
    },
    {
      "epoch": 84.92,
      "learning_rate": 8.676763754045308e-06,
      "loss": 0.0363,
      "step": 218680
    },
    {
      "epoch": 84.93,
      "learning_rate": 8.676245954692558e-06,
      "loss": 0.0817,
      "step": 218690
    },
    {
      "epoch": 84.93,
      "learning_rate": 8.675728155339807e-06,
      "loss": 0.0588,
      "step": 218700
    },
    {
      "epoch": 84.94,
      "learning_rate": 8.675210355987057e-06,
      "loss": 0.1293,
      "step": 218710
    },
    {
      "epoch": 84.94,
      "learning_rate": 8.674692556634305e-06,
      "loss": 0.1119,
      "step": 218720
    },
    {
      "epoch": 84.94,
      "learning_rate": 8.674174757281554e-06,
      "loss": 0.0087,
      "step": 218730
    },
    {
      "epoch": 84.95,
      "learning_rate": 8.673656957928804e-06,
      "loss": 0.0328,
      "step": 218740
    },
    {
      "epoch": 84.95,
      "learning_rate": 8.673139158576054e-06,
      "loss": 0.0284,
      "step": 218750
    },
    {
      "epoch": 84.96,
      "learning_rate": 8.672621359223301e-06,
      "loss": 0.0258,
      "step": 218760
    },
    {
      "epoch": 84.96,
      "learning_rate": 8.672103559870551e-06,
      "loss": 0.0349,
      "step": 218770
    },
    {
      "epoch": 84.96,
      "learning_rate": 8.6715857605178e-06,
      "loss": 0.0191,
      "step": 218780
    },
    {
      "epoch": 84.97,
      "learning_rate": 8.671067961165049e-06,
      "loss": 0.0619,
      "step": 218790
    },
    {
      "epoch": 84.97,
      "learning_rate": 8.670550161812298e-06,
      "loss": 0.0293,
      "step": 218800
    },
    {
      "epoch": 84.97,
      "learning_rate": 8.670032362459548e-06,
      "loss": 0.0529,
      "step": 218810
    },
    {
      "epoch": 84.98,
      "learning_rate": 8.669514563106797e-06,
      "loss": 0.0748,
      "step": 218820
    },
    {
      "epoch": 84.98,
      "learning_rate": 8.668996763754045e-06,
      "loss": 0.0754,
      "step": 218830
    },
    {
      "epoch": 84.99,
      "learning_rate": 8.668478964401295e-06,
      "loss": 0.0671,
      "step": 218840
    },
    {
      "epoch": 84.99,
      "learning_rate": 8.667961165048545e-06,
      "loss": 0.0402,
      "step": 218850
    },
    {
      "epoch": 84.99,
      "learning_rate": 8.667443365695794e-06,
      "loss": 0.0009,
      "step": 218860
    },
    {
      "epoch": 85.0,
      "learning_rate": 8.666925566343042e-06,
      "loss": 0.043,
      "step": 218870
    },
    {
      "epoch": 85.0,
      "eval_accuracy": 0.9524071526822558,
      "eval_loss": 0.35918599367141724,
      "eval_runtime": 8.1935,
      "eval_samples_per_second": 443.646,
      "eval_steps_per_second": 55.532,
      "step": 218875
    },
    {
      "epoch": 85.0,
      "learning_rate": 8.666407766990292e-06,
      "loss": 0.001,
      "step": 218880
    },
    {
      "epoch": 85.01,
      "learning_rate": 8.665889967637541e-06,
      "loss": 0.1453,
      "step": 218890
    },
    {
      "epoch": 85.01,
      "learning_rate": 8.665372168284791e-06,
      "loss": 0.018,
      "step": 218900
    },
    {
      "epoch": 85.01,
      "learning_rate": 8.664854368932039e-06,
      "loss": 0.125,
      "step": 218910
    },
    {
      "epoch": 85.02,
      "learning_rate": 8.664336569579289e-06,
      "loss": 0.0006,
      "step": 218920
    },
    {
      "epoch": 85.02,
      "learning_rate": 8.663818770226538e-06,
      "loss": 0.1196,
      "step": 218930
    },
    {
      "epoch": 85.03,
      "learning_rate": 8.663300970873788e-06,
      "loss": 0.0656,
      "step": 218940
    },
    {
      "epoch": 85.03,
      "learning_rate": 8.662783171521036e-06,
      "loss": 0.014,
      "step": 218950
    },
    {
      "epoch": 85.03,
      "learning_rate": 8.662265372168285e-06,
      "loss": 0.0271,
      "step": 218960
    },
    {
      "epoch": 85.04,
      "learning_rate": 8.661747572815535e-06,
      "loss": 0.1233,
      "step": 218970
    },
    {
      "epoch": 85.04,
      "learning_rate": 8.661229773462783e-06,
      "loss": 0.0518,
      "step": 218980
    },
    {
      "epoch": 85.04,
      "learning_rate": 8.660711974110033e-06,
      "loss": 0.0131,
      "step": 218990
    },
    {
      "epoch": 85.05,
      "learning_rate": 8.660194174757282e-06,
      "loss": 0.1272,
      "step": 219000
    },
    {
      "epoch": 85.05,
      "learning_rate": 8.659676375404532e-06,
      "loss": 0.1033,
      "step": 219010
    },
    {
      "epoch": 85.06,
      "learning_rate": 8.65915857605178e-06,
      "loss": 0.0531,
      "step": 219020
    },
    {
      "epoch": 85.06,
      "learning_rate": 8.65864077669903e-06,
      "loss": 0.1653,
      "step": 219030
    },
    {
      "epoch": 85.06,
      "learning_rate": 8.658122977346279e-06,
      "loss": 0.0346,
      "step": 219040
    },
    {
      "epoch": 85.07,
      "learning_rate": 8.657605177993529e-06,
      "loss": 0.0261,
      "step": 219050
    },
    {
      "epoch": 85.07,
      "learning_rate": 8.657087378640776e-06,
      "loss": 0.1185,
      "step": 219060
    },
    {
      "epoch": 85.08,
      "learning_rate": 8.656569579288026e-06,
      "loss": 0.1434,
      "step": 219070
    },
    {
      "epoch": 85.08,
      "learning_rate": 8.656051779935276e-06,
      "loss": 0.0559,
      "step": 219080
    },
    {
      "epoch": 85.08,
      "learning_rate": 8.655533980582525e-06,
      "loss": 0.0001,
      "step": 219090
    },
    {
      "epoch": 85.09,
      "learning_rate": 8.655016181229773e-06,
      "loss": 0.0129,
      "step": 219100
    },
    {
      "epoch": 85.09,
      "learning_rate": 8.654498381877023e-06,
      "loss": 0.1186,
      "step": 219110
    },
    {
      "epoch": 85.1,
      "learning_rate": 8.653980582524272e-06,
      "loss": 0.0148,
      "step": 219120
    },
    {
      "epoch": 85.1,
      "learning_rate": 8.65346278317152e-06,
      "loss": 0.048,
      "step": 219130
    },
    {
      "epoch": 85.1,
      "learning_rate": 8.65294498381877e-06,
      "loss": 0.0062,
      "step": 219140
    },
    {
      "epoch": 85.11,
      "learning_rate": 8.65242718446602e-06,
      "loss": 0.0658,
      "step": 219150
    },
    {
      "epoch": 85.11,
      "learning_rate": 8.65190938511327e-06,
      "loss": 0.0089,
      "step": 219160
    },
    {
      "epoch": 85.11,
      "learning_rate": 8.651391585760517e-06,
      "loss": 0.0469,
      "step": 219170
    },
    {
      "epoch": 85.12,
      "learning_rate": 8.650873786407767e-06,
      "loss": 0.0822,
      "step": 219180
    },
    {
      "epoch": 85.12,
      "learning_rate": 8.650355987055016e-06,
      "loss": 0.0388,
      "step": 219190
    },
    {
      "epoch": 85.13,
      "learning_rate": 8.649838187702266e-06,
      "loss": 0.0624,
      "step": 219200
    },
    {
      "epoch": 85.13,
      "learning_rate": 8.649320388349516e-06,
      "loss": 0.0133,
      "step": 219210
    },
    {
      "epoch": 85.13,
      "learning_rate": 8.648802588996764e-06,
      "loss": 0.0049,
      "step": 219220
    },
    {
      "epoch": 85.14,
      "learning_rate": 8.648284789644013e-06,
      "loss": 0.0051,
      "step": 219230
    },
    {
      "epoch": 85.14,
      "learning_rate": 8.647766990291263e-06,
      "loss": 0.0692,
      "step": 219240
    },
    {
      "epoch": 85.15,
      "learning_rate": 8.647249190938512e-06,
      "loss": 0.0421,
      "step": 219250
    },
    {
      "epoch": 85.15,
      "learning_rate": 8.64673139158576e-06,
      "loss": 0.0335,
      "step": 219260
    },
    {
      "epoch": 85.15,
      "learning_rate": 8.64621359223301e-06,
      "loss": 0.0186,
      "step": 219270
    },
    {
      "epoch": 85.16,
      "learning_rate": 8.64569579288026e-06,
      "loss": 0.0055,
      "step": 219280
    },
    {
      "epoch": 85.16,
      "learning_rate": 8.64517799352751e-06,
      "loss": 0.0274,
      "step": 219290
    },
    {
      "epoch": 85.17,
      "learning_rate": 8.644660194174757e-06,
      "loss": 0.0169,
      "step": 219300
    },
    {
      "epoch": 85.17,
      "learning_rate": 8.644142394822007e-06,
      "loss": 0.0429,
      "step": 219310
    },
    {
      "epoch": 85.17,
      "learning_rate": 8.643624595469256e-06,
      "loss": 0.0432,
      "step": 219320
    },
    {
      "epoch": 85.18,
      "learning_rate": 8.643106796116506e-06,
      "loss": 0.0447,
      "step": 219330
    },
    {
      "epoch": 85.18,
      "learning_rate": 8.642588996763754e-06,
      "loss": 0.0067,
      "step": 219340
    },
    {
      "epoch": 85.18,
      "learning_rate": 8.642071197411004e-06,
      "loss": 0.04,
      "step": 219350
    },
    {
      "epoch": 85.19,
      "learning_rate": 8.641553398058253e-06,
      "loss": 0.0018,
      "step": 219360
    },
    {
      "epoch": 85.19,
      "learning_rate": 8.641035598705503e-06,
      "loss": 0.0073,
      "step": 219370
    },
    {
      "epoch": 85.2,
      "learning_rate": 8.64051779935275e-06,
      "loss": 0.1129,
      "step": 219380
    },
    {
      "epoch": 85.2,
      "learning_rate": 8.64e-06,
      "loss": 0.0565,
      "step": 219390
    },
    {
      "epoch": 85.2,
      "learning_rate": 8.63948220064725e-06,
      "loss": 0.0376,
      "step": 219400
    },
    {
      "epoch": 85.21,
      "learning_rate": 8.6389644012945e-06,
      "loss": 0.0082,
      "step": 219410
    },
    {
      "epoch": 85.21,
      "learning_rate": 8.638446601941747e-06,
      "loss": 0.0347,
      "step": 219420
    },
    {
      "epoch": 85.22,
      "learning_rate": 8.637928802588997e-06,
      "loss": 0.1035,
      "step": 219430
    },
    {
      "epoch": 85.22,
      "learning_rate": 8.637411003236247e-06,
      "loss": 0.1807,
      "step": 219440
    },
    {
      "epoch": 85.22,
      "learning_rate": 8.636893203883496e-06,
      "loss": 0.0125,
      "step": 219450
    },
    {
      "epoch": 85.23,
      "learning_rate": 8.636375404530744e-06,
      "loss": 0.0553,
      "step": 219460
    },
    {
      "epoch": 85.23,
      "learning_rate": 8.635857605177994e-06,
      "loss": 0.0963,
      "step": 219470
    },
    {
      "epoch": 85.23,
      "learning_rate": 8.635339805825243e-06,
      "loss": 0.0538,
      "step": 219480
    },
    {
      "epoch": 85.24,
      "learning_rate": 8.634822006472493e-06,
      "loss": 0.1437,
      "step": 219490
    },
    {
      "epoch": 85.24,
      "learning_rate": 8.634304207119741e-06,
      "loss": 0.0294,
      "step": 219500
    },
    {
      "epoch": 85.25,
      "learning_rate": 8.63378640776699e-06,
      "loss": 0.1208,
      "step": 219510
    },
    {
      "epoch": 85.25,
      "learning_rate": 8.63326860841424e-06,
      "loss": 0.0196,
      "step": 219520
    },
    {
      "epoch": 85.25,
      "learning_rate": 8.63275080906149e-06,
      "loss": 0.0951,
      "step": 219530
    },
    {
      "epoch": 85.26,
      "learning_rate": 8.632233009708738e-06,
      "loss": 0.1345,
      "step": 219540
    },
    {
      "epoch": 85.26,
      "learning_rate": 8.631715210355987e-06,
      "loss": 0.0938,
      "step": 219550
    },
    {
      "epoch": 85.27,
      "learning_rate": 8.631197411003237e-06,
      "loss": 0.0295,
      "step": 219560
    },
    {
      "epoch": 85.27,
      "learning_rate": 8.630679611650487e-06,
      "loss": 0.0015,
      "step": 219570
    },
    {
      "epoch": 85.27,
      "learning_rate": 8.630161812297735e-06,
      "loss": 0.0454,
      "step": 219580
    },
    {
      "epoch": 85.28,
      "learning_rate": 8.629644012944984e-06,
      "loss": 0.1038,
      "step": 219590
    },
    {
      "epoch": 85.28,
      "learning_rate": 8.629126213592234e-06,
      "loss": 0.0561,
      "step": 219600
    },
    {
      "epoch": 85.29,
      "learning_rate": 8.628608414239483e-06,
      "loss": 0.0002,
      "step": 219610
    },
    {
      "epoch": 85.29,
      "learning_rate": 8.628090614886731e-06,
      "loss": 0.1394,
      "step": 219620
    },
    {
      "epoch": 85.29,
      "learning_rate": 8.627572815533981e-06,
      "loss": 0.1044,
      "step": 219630
    },
    {
      "epoch": 85.3,
      "learning_rate": 8.62705501618123e-06,
      "loss": 0.0281,
      "step": 219640
    },
    {
      "epoch": 85.3,
      "learning_rate": 8.62653721682848e-06,
      "loss": 0.002,
      "step": 219650
    },
    {
      "epoch": 85.3,
      "learning_rate": 8.626019417475728e-06,
      "loss": 0.014,
      "step": 219660
    },
    {
      "epoch": 85.31,
      "learning_rate": 8.625501618122978e-06,
      "loss": 0.1707,
      "step": 219670
    },
    {
      "epoch": 85.31,
      "learning_rate": 8.624983818770227e-06,
      "loss": 0.0916,
      "step": 219680
    },
    {
      "epoch": 85.32,
      "learning_rate": 8.624466019417477e-06,
      "loss": 0.0069,
      "step": 219690
    },
    {
      "epoch": 85.32,
      "learning_rate": 8.623948220064725e-06,
      "loss": 0.1947,
      "step": 219700
    },
    {
      "epoch": 85.32,
      "learning_rate": 8.623430420711975e-06,
      "loss": 0.0025,
      "step": 219710
    },
    {
      "epoch": 85.33,
      "learning_rate": 8.622912621359224e-06,
      "loss": 0.073,
      "step": 219720
    },
    {
      "epoch": 85.33,
      "learning_rate": 8.622394822006474e-06,
      "loss": 0.1819,
      "step": 219730
    },
    {
      "epoch": 85.34,
      "learning_rate": 8.621877022653722e-06,
      "loss": 0.0635,
      "step": 219740
    },
    {
      "epoch": 85.34,
      "learning_rate": 8.621359223300971e-06,
      "loss": 0.0528,
      "step": 219750
    },
    {
      "epoch": 85.34,
      "learning_rate": 8.620841423948221e-06,
      "loss": 0.0619,
      "step": 219760
    },
    {
      "epoch": 85.35,
      "learning_rate": 8.62032362459547e-06,
      "loss": 0.0126,
      "step": 219770
    },
    {
      "epoch": 85.35,
      "learning_rate": 8.61980582524272e-06,
      "loss": 0.0033,
      "step": 219780
    },
    {
      "epoch": 85.36,
      "learning_rate": 8.619288025889968e-06,
      "loss": 0.1256,
      "step": 219790
    },
    {
      "epoch": 85.36,
      "learning_rate": 8.618770226537218e-06,
      "loss": 0.0075,
      "step": 219800
    },
    {
      "epoch": 85.36,
      "learning_rate": 8.618252427184467e-06,
      "loss": 0.0821,
      "step": 219810
    },
    {
      "epoch": 85.37,
      "learning_rate": 8.617734627831717e-06,
      "loss": 0.0203,
      "step": 219820
    },
    {
      "epoch": 85.37,
      "learning_rate": 8.617216828478965e-06,
      "loss": 0.0095,
      "step": 219830
    },
    {
      "epoch": 85.37,
      "learning_rate": 8.616699029126214e-06,
      "loss": 0.0187,
      "step": 219840
    },
    {
      "epoch": 85.38,
      "learning_rate": 8.616181229773464e-06,
      "loss": 0.122,
      "step": 219850
    },
    {
      "epoch": 85.38,
      "learning_rate": 8.615663430420714e-06,
      "loss": 0.021,
      "step": 219860
    },
    {
      "epoch": 85.39,
      "learning_rate": 8.615145631067962e-06,
      "loss": 0.03,
      "step": 219870
    },
    {
      "epoch": 85.39,
      "learning_rate": 8.614627831715211e-06,
      "loss": 0.0689,
      "step": 219880
    },
    {
      "epoch": 85.39,
      "learning_rate": 8.614110032362461e-06,
      "loss": 0.0569,
      "step": 219890
    },
    {
      "epoch": 85.4,
      "learning_rate": 8.61359223300971e-06,
      "loss": 0.0159,
      "step": 219900
    },
    {
      "epoch": 85.4,
      "learning_rate": 8.613074433656958e-06,
      "loss": 0.0094,
      "step": 219910
    },
    {
      "epoch": 85.41,
      "learning_rate": 8.612556634304208e-06,
      "loss": 0.1133,
      "step": 219920
    },
    {
      "epoch": 85.41,
      "learning_rate": 8.612038834951458e-06,
      "loss": 0.0361,
      "step": 219930
    },
    {
      "epoch": 85.41,
      "learning_rate": 8.611521035598707e-06,
      "loss": 0.0168,
      "step": 219940
    },
    {
      "epoch": 85.42,
      "learning_rate": 8.611003236245955e-06,
      "loss": 0.1061,
      "step": 219950
    },
    {
      "epoch": 85.42,
      "learning_rate": 8.610485436893205e-06,
      "loss": 0.0717,
      "step": 219960
    },
    {
      "epoch": 85.43,
      "learning_rate": 8.609967637540454e-06,
      "loss": 0.0782,
      "step": 219970
    },
    {
      "epoch": 85.43,
      "learning_rate": 8.609449838187704e-06,
      "loss": 0.0376,
      "step": 219980
    },
    {
      "epoch": 85.43,
      "learning_rate": 8.608932038834952e-06,
      "loss": 0.0114,
      "step": 219990
    },
    {
      "epoch": 85.44,
      "learning_rate": 8.608414239482202e-06,
      "loss": 0.0675,
      "step": 220000
    },
    {
      "epoch": 85.44,
      "learning_rate": 8.607896440129451e-06,
      "loss": 0.1431,
      "step": 220010
    },
    {
      "epoch": 85.44,
      "learning_rate": 8.6073786407767e-06,
      "loss": 0.1466,
      "step": 220020
    },
    {
      "epoch": 85.45,
      "learning_rate": 8.606860841423949e-06,
      "loss": 0.0331,
      "step": 220030
    },
    {
      "epoch": 85.45,
      "learning_rate": 8.606343042071198e-06,
      "loss": 0.0648,
      "step": 220040
    },
    {
      "epoch": 85.46,
      "learning_rate": 8.605825242718448e-06,
      "loss": 0.0406,
      "step": 220050
    },
    {
      "epoch": 85.46,
      "learning_rate": 8.605307443365698e-06,
      "loss": 0.0333,
      "step": 220060
    },
    {
      "epoch": 85.46,
      "learning_rate": 8.604789644012946e-06,
      "loss": 0.0449,
      "step": 220070
    },
    {
      "epoch": 85.47,
      "learning_rate": 8.604271844660195e-06,
      "loss": 0.0047,
      "step": 220080
    },
    {
      "epoch": 85.47,
      "learning_rate": 8.603754045307445e-06,
      "loss": 0.042,
      "step": 220090
    },
    {
      "epoch": 85.48,
      "learning_rate": 8.603236245954694e-06,
      "loss": 0.0001,
      "step": 220100
    },
    {
      "epoch": 85.48,
      "learning_rate": 8.602718446601942e-06,
      "loss": 0.0744,
      "step": 220110
    },
    {
      "epoch": 85.48,
      "learning_rate": 8.602200647249192e-06,
      "loss": 0.0337,
      "step": 220120
    },
    {
      "epoch": 85.49,
      "learning_rate": 8.601682847896442e-06,
      "loss": 0.0499,
      "step": 220130
    },
    {
      "epoch": 85.49,
      "learning_rate": 8.601165048543691e-06,
      "loss": 0.0336,
      "step": 220140
    },
    {
      "epoch": 85.5,
      "learning_rate": 8.600647249190939e-06,
      "loss": 0.0094,
      "step": 220150
    },
    {
      "epoch": 85.5,
      "learning_rate": 8.600129449838189e-06,
      "loss": 0.0003,
      "step": 220160
    },
    {
      "epoch": 85.5,
      "learning_rate": 8.599611650485438e-06,
      "loss": 0.1371,
      "step": 220170
    },
    {
      "epoch": 85.51,
      "learning_rate": 8.599093851132688e-06,
      "loss": 0.0814,
      "step": 220180
    },
    {
      "epoch": 85.51,
      "learning_rate": 8.598576051779936e-06,
      "loss": 0.0267,
      "step": 220190
    },
    {
      "epoch": 85.51,
      "learning_rate": 8.598058252427185e-06,
      "loss": 0.026,
      "step": 220200
    },
    {
      "epoch": 85.52,
      "learning_rate": 8.597540453074435e-06,
      "loss": 0.1749,
      "step": 220210
    },
    {
      "epoch": 85.52,
      "learning_rate": 8.597022653721685e-06,
      "loss": 0.1153,
      "step": 220220
    },
    {
      "epoch": 85.53,
      "learning_rate": 8.596504854368933e-06,
      "loss": 0.0095,
      "step": 220230
    },
    {
      "epoch": 85.53,
      "learning_rate": 8.595987055016182e-06,
      "loss": 0.0319,
      "step": 220240
    },
    {
      "epoch": 85.53,
      "learning_rate": 8.595469255663432e-06,
      "loss": 0.1418,
      "step": 220250
    },
    {
      "epoch": 85.54,
      "learning_rate": 8.59495145631068e-06,
      "loss": 0.26,
      "step": 220260
    },
    {
      "epoch": 85.54,
      "learning_rate": 8.59443365695793e-06,
      "loss": 0.0455,
      "step": 220270
    },
    {
      "epoch": 85.55,
      "learning_rate": 8.593915857605179e-06,
      "loss": 0.0354,
      "step": 220280
    },
    {
      "epoch": 85.55,
      "learning_rate": 8.593398058252429e-06,
      "loss": 0.1276,
      "step": 220290
    },
    {
      "epoch": 85.55,
      "learning_rate": 8.592880258899677e-06,
      "loss": 0.1517,
      "step": 220300
    },
    {
      "epoch": 85.56,
      "learning_rate": 8.592362459546926e-06,
      "loss": 0.0834,
      "step": 220310
    },
    {
      "epoch": 85.56,
      "learning_rate": 8.591844660194176e-06,
      "loss": 0.11,
      "step": 220320
    },
    {
      "epoch": 85.57,
      "learning_rate": 8.591326860841425e-06,
      "loss": 0.0065,
      "step": 220330
    },
    {
      "epoch": 85.57,
      "learning_rate": 8.590809061488673e-06,
      "loss": 0.013,
      "step": 220340
    },
    {
      "epoch": 85.57,
      "learning_rate": 8.590291262135923e-06,
      "loss": 0.0362,
      "step": 220350
    },
    {
      "epoch": 85.58,
      "learning_rate": 8.589773462783173e-06,
      "loss": 0.0298,
      "step": 220360
    },
    {
      "epoch": 85.58,
      "learning_rate": 8.589255663430422e-06,
      "loss": 0.0526,
      "step": 220370
    },
    {
      "epoch": 85.58,
      "learning_rate": 8.58873786407767e-06,
      "loss": 0.0418,
      "step": 220380
    },
    {
      "epoch": 85.59,
      "learning_rate": 8.58822006472492e-06,
      "loss": 0.0617,
      "step": 220390
    },
    {
      "epoch": 85.59,
      "learning_rate": 8.58770226537217e-06,
      "loss": 0.0404,
      "step": 220400
    },
    {
      "epoch": 85.6,
      "learning_rate": 8.587184466019419e-06,
      "loss": 0.0572,
      "step": 220410
    },
    {
      "epoch": 85.6,
      "learning_rate": 8.586666666666667e-06,
      "loss": 0.0831,
      "step": 220420
    },
    {
      "epoch": 85.6,
      "learning_rate": 8.586148867313917e-06,
      "loss": 0.0293,
      "step": 220430
    },
    {
      "epoch": 85.61,
      "learning_rate": 8.585631067961166e-06,
      "loss": 0.0021,
      "step": 220440
    },
    {
      "epoch": 85.61,
      "learning_rate": 8.585113268608414e-06,
      "loss": 0.1545,
      "step": 220450
    },
    {
      "epoch": 85.62,
      "learning_rate": 8.584595469255664e-06,
      "loss": 0.1351,
      "step": 220460
    },
    {
      "epoch": 85.62,
      "learning_rate": 8.584077669902913e-06,
      "loss": 0.0169,
      "step": 220470
    },
    {
      "epoch": 85.62,
      "learning_rate": 8.583559870550163e-06,
      "loss": 0.0927,
      "step": 220480
    },
    {
      "epoch": 85.63,
      "learning_rate": 8.58304207119741e-06,
      "loss": 0.1434,
      "step": 220490
    },
    {
      "epoch": 85.63,
      "learning_rate": 8.58252427184466e-06,
      "loss": 0.1431,
      "step": 220500
    },
    {
      "epoch": 85.63,
      "learning_rate": 8.58200647249191e-06,
      "loss": 0.0542,
      "step": 220510
    },
    {
      "epoch": 85.64,
      "learning_rate": 8.58148867313916e-06,
      "loss": 0.0451,
      "step": 220520
    },
    {
      "epoch": 85.64,
      "learning_rate": 8.580970873786408e-06,
      "loss": 0.2238,
      "step": 220530
    },
    {
      "epoch": 85.65,
      "learning_rate": 8.580453074433657e-06,
      "loss": 0.0109,
      "step": 220540
    },
    {
      "epoch": 85.65,
      "learning_rate": 8.579935275080907e-06,
      "loss": 0.0476,
      "step": 220550
    },
    {
      "epoch": 85.65,
      "learning_rate": 8.579417475728156e-06,
      "loss": 0.0437,
      "step": 220560
    },
    {
      "epoch": 85.66,
      "learning_rate": 8.578899676375404e-06,
      "loss": 0.1187,
      "step": 220570
    },
    {
      "epoch": 85.66,
      "learning_rate": 8.578381877022654e-06,
      "loss": 0.0348,
      "step": 220580
    },
    {
      "epoch": 85.67,
      "learning_rate": 8.577864077669904e-06,
      "loss": 0.013,
      "step": 220590
    },
    {
      "epoch": 85.67,
      "learning_rate": 8.577346278317152e-06,
      "loss": 0.0261,
      "step": 220600
    },
    {
      "epoch": 85.67,
      "learning_rate": 8.576828478964401e-06,
      "loss": 0.0108,
      "step": 220610
    },
    {
      "epoch": 85.68,
      "learning_rate": 8.57631067961165e-06,
      "loss": 0.1254,
      "step": 220620
    },
    {
      "epoch": 85.68,
      "learning_rate": 8.5757928802589e-06,
      "loss": 0.079,
      "step": 220630
    },
    {
      "epoch": 85.69,
      "learning_rate": 8.575275080906148e-06,
      "loss": 0.1107,
      "step": 220640
    },
    {
      "epoch": 85.69,
      "learning_rate": 8.574757281553398e-06,
      "loss": 0.017,
      "step": 220650
    },
    {
      "epoch": 85.69,
      "learning_rate": 8.574239482200648e-06,
      "loss": 0.0248,
      "step": 220660
    },
    {
      "epoch": 85.7,
      "learning_rate": 8.573721682847897e-06,
      "loss": 0.1514,
      "step": 220670
    },
    {
      "epoch": 85.7,
      "learning_rate": 8.573203883495145e-06,
      "loss": 0.2548,
      "step": 220680
    },
    {
      "epoch": 85.7,
      "learning_rate": 8.572686084142395e-06,
      "loss": 0.0486,
      "step": 220690
    },
    {
      "epoch": 85.71,
      "learning_rate": 8.572168284789644e-06,
      "loss": 0.0838,
      "step": 220700
    },
    {
      "epoch": 85.71,
      "learning_rate": 8.571650485436894e-06,
      "loss": 0.0164,
      "step": 220710
    },
    {
      "epoch": 85.72,
      "learning_rate": 8.571132686084142e-06,
      "loss": 0.0788,
      "step": 220720
    },
    {
      "epoch": 85.72,
      "learning_rate": 8.570614886731392e-06,
      "loss": 0.0829,
      "step": 220730
    },
    {
      "epoch": 85.72,
      "learning_rate": 8.570097087378641e-06,
      "loss": 0.0527,
      "step": 220740
    },
    {
      "epoch": 85.73,
      "learning_rate": 8.56957928802589e-06,
      "loss": 0.0417,
      "step": 220750
    },
    {
      "epoch": 85.73,
      "learning_rate": 8.569061488673139e-06,
      "loss": 0.0719,
      "step": 220760
    },
    {
      "epoch": 85.74,
      "learning_rate": 8.568543689320388e-06,
      "loss": 0.0735,
      "step": 220770
    },
    {
      "epoch": 85.74,
      "learning_rate": 8.568025889967638e-06,
      "loss": 0.1058,
      "step": 220780
    },
    {
      "epoch": 85.74,
      "learning_rate": 8.567508090614888e-06,
      "loss": 0.0548,
      "step": 220790
    },
    {
      "epoch": 85.75,
      "learning_rate": 8.566990291262135e-06,
      "loss": 0.0672,
      "step": 220800
    },
    {
      "epoch": 85.75,
      "learning_rate": 8.566472491909385e-06,
      "loss": 0.0378,
      "step": 220810
    },
    {
      "epoch": 85.76,
      "learning_rate": 8.565954692556635e-06,
      "loss": 0.0833,
      "step": 220820
    },
    {
      "epoch": 85.76,
      "learning_rate": 8.565436893203884e-06,
      "loss": 0.0024,
      "step": 220830
    },
    {
      "epoch": 85.76,
      "learning_rate": 8.564919093851132e-06,
      "loss": 0.1122,
      "step": 220840
    },
    {
      "epoch": 85.77,
      "learning_rate": 8.564401294498382e-06,
      "loss": 0.0341,
      "step": 220850
    },
    {
      "epoch": 85.77,
      "learning_rate": 8.563883495145631e-06,
      "loss": 0.0369,
      "step": 220860
    },
    {
      "epoch": 85.77,
      "learning_rate": 8.563365695792881e-06,
      "loss": 0.2044,
      "step": 220870
    },
    {
      "epoch": 85.78,
      "learning_rate": 8.56284789644013e-06,
      "loss": 0.0217,
      "step": 220880
    },
    {
      "epoch": 85.78,
      "learning_rate": 8.562330097087379e-06,
      "loss": 0.0396,
      "step": 220890
    },
    {
      "epoch": 85.79,
      "learning_rate": 8.561812297734628e-06,
      "loss": 0.0442,
      "step": 220900
    },
    {
      "epoch": 85.79,
      "learning_rate": 8.561294498381878e-06,
      "loss": 0.0037,
      "step": 220910
    },
    {
      "epoch": 85.79,
      "learning_rate": 8.560776699029127e-06,
      "loss": 0.1072,
      "step": 220920
    },
    {
      "epoch": 85.8,
      "learning_rate": 8.560258899676375e-06,
      "loss": 0.0335,
      "step": 220930
    },
    {
      "epoch": 85.8,
      "learning_rate": 8.559741100323625e-06,
      "loss": 0.035,
      "step": 220940
    },
    {
      "epoch": 85.81,
      "learning_rate": 8.559223300970875e-06,
      "loss": 0.0672,
      "step": 220950
    },
    {
      "epoch": 85.81,
      "learning_rate": 8.558705501618124e-06,
      "loss": 0.147,
      "step": 220960
    },
    {
      "epoch": 85.81,
      "learning_rate": 8.558187702265372e-06,
      "loss": 0.0317,
      "step": 220970
    },
    {
      "epoch": 85.82,
      "learning_rate": 8.557669902912622e-06,
      "loss": 0.0714,
      "step": 220980
    },
    {
      "epoch": 85.82,
      "learning_rate": 8.557152103559871e-06,
      "loss": 0.0846,
      "step": 220990
    },
    {
      "epoch": 85.83,
      "learning_rate": 8.556634304207121e-06,
      "loss": 0.0021,
      "step": 221000
    },
    {
      "epoch": 85.83,
      "learning_rate": 8.556116504854369e-06,
      "loss": 0.0935,
      "step": 221010
    },
    {
      "epoch": 85.83,
      "learning_rate": 8.555598705501619e-06,
      "loss": 0.0004,
      "step": 221020
    },
    {
      "epoch": 85.84,
      "learning_rate": 8.555080906148868e-06,
      "loss": 0.0868,
      "step": 221030
    },
    {
      "epoch": 85.84,
      "learning_rate": 8.554563106796118e-06,
      "loss": 0.0655,
      "step": 221040
    },
    {
      "epoch": 85.84,
      "learning_rate": 8.554045307443366e-06,
      "loss": 0.093,
      "step": 221050
    },
    {
      "epoch": 85.85,
      "learning_rate": 8.553527508090615e-06,
      "loss": 0.1758,
      "step": 221060
    },
    {
      "epoch": 85.85,
      "learning_rate": 8.553009708737865e-06,
      "loss": 0.0098,
      "step": 221070
    },
    {
      "epoch": 85.86,
      "learning_rate": 8.552491909385115e-06,
      "loss": 0.1004,
      "step": 221080
    },
    {
      "epoch": 85.86,
      "learning_rate": 8.551974110032363e-06,
      "loss": 0.0669,
      "step": 221090
    },
    {
      "epoch": 85.86,
      "learning_rate": 8.551456310679612e-06,
      "loss": 0.0028,
      "step": 221100
    },
    {
      "epoch": 85.87,
      "learning_rate": 8.550938511326862e-06,
      "loss": 0.0441,
      "step": 221110
    },
    {
      "epoch": 85.87,
      "learning_rate": 8.550420711974111e-06,
      "loss": 0.0128,
      "step": 221120
    },
    {
      "epoch": 85.88,
      "learning_rate": 8.54990291262136e-06,
      "loss": 0.1313,
      "step": 221130
    },
    {
      "epoch": 85.88,
      "learning_rate": 8.549385113268609e-06,
      "loss": 0.1093,
      "step": 221140
    },
    {
      "epoch": 85.88,
      "learning_rate": 8.548867313915859e-06,
      "loss": 0.0202,
      "step": 221150
    },
    {
      "epoch": 85.89,
      "learning_rate": 8.548349514563108e-06,
      "loss": 0.057,
      "step": 221160
    },
    {
      "epoch": 85.89,
      "learning_rate": 8.547831715210356e-06,
      "loss": 0.058,
      "step": 221170
    },
    {
      "epoch": 85.9,
      "learning_rate": 8.547313915857606e-06,
      "loss": 0.015,
      "step": 221180
    },
    {
      "epoch": 85.9,
      "learning_rate": 8.546796116504855e-06,
      "loss": 0.0333,
      "step": 221190
    },
    {
      "epoch": 85.9,
      "learning_rate": 8.546278317152105e-06,
      "loss": 0.0504,
      "step": 221200
    },
    {
      "epoch": 85.91,
      "learning_rate": 8.545760517799353e-06,
      "loss": 0.068,
      "step": 221210
    },
    {
      "epoch": 85.91,
      "learning_rate": 8.545242718446602e-06,
      "loss": 0.1287,
      "step": 221220
    },
    {
      "epoch": 85.91,
      "learning_rate": 8.544724919093852e-06,
      "loss": 0.0399,
      "step": 221230
    },
    {
      "epoch": 85.92,
      "learning_rate": 8.544207119741102e-06,
      "loss": 0.0175,
      "step": 221240
    },
    {
      "epoch": 85.92,
      "learning_rate": 8.54368932038835e-06,
      "loss": 0.009,
      "step": 221250
    },
    {
      "epoch": 85.93,
      "learning_rate": 8.5431715210356e-06,
      "loss": 0.0787,
      "step": 221260
    },
    {
      "epoch": 85.93,
      "learning_rate": 8.542653721682849e-06,
      "loss": 0.1314,
      "step": 221270
    },
    {
      "epoch": 85.93,
      "learning_rate": 8.542135922330098e-06,
      "loss": 0.0589,
      "step": 221280
    },
    {
      "epoch": 85.94,
      "learning_rate": 8.541618122977346e-06,
      "loss": 0.0115,
      "step": 221290
    },
    {
      "epoch": 85.94,
      "learning_rate": 8.541100323624596e-06,
      "loss": 0.0043,
      "step": 221300
    },
    {
      "epoch": 85.95,
      "learning_rate": 8.540582524271846e-06,
      "loss": 0.0636,
      "step": 221310
    },
    {
      "epoch": 85.95,
      "learning_rate": 8.540064724919095e-06,
      "loss": 0.0693,
      "step": 221320
    },
    {
      "epoch": 85.95,
      "learning_rate": 8.539546925566343e-06,
      "loss": 0.0021,
      "step": 221330
    },
    {
      "epoch": 85.96,
      "learning_rate": 8.539029126213593e-06,
      "loss": 0.0276,
      "step": 221340
    },
    {
      "epoch": 85.96,
      "learning_rate": 8.538511326860842e-06,
      "loss": 0.0411,
      "step": 221350
    },
    {
      "epoch": 85.97,
      "learning_rate": 8.537993527508092e-06,
      "loss": 0.0163,
      "step": 221360
    },
    {
      "epoch": 85.97,
      "learning_rate": 8.53747572815534e-06,
      "loss": 0.1259,
      "step": 221370
    },
    {
      "epoch": 85.97,
      "learning_rate": 8.53695792880259e-06,
      "loss": 0.014,
      "step": 221380
    },
    {
      "epoch": 85.98,
      "learning_rate": 8.53644012944984e-06,
      "loss": 0.1157,
      "step": 221390
    },
    {
      "epoch": 85.98,
      "learning_rate": 8.535922330097089e-06,
      "loss": 0.0116,
      "step": 221400
    },
    {
      "epoch": 85.98,
      "learning_rate": 8.535404530744337e-06,
      "loss": 0.165,
      "step": 221410
    },
    {
      "epoch": 85.99,
      "learning_rate": 8.534886731391586e-06,
      "loss": 0.0999,
      "step": 221420
    },
    {
      "epoch": 85.99,
      "learning_rate": 8.534368932038836e-06,
      "loss": 0.094,
      "step": 221430
    },
    {
      "epoch": 86.0,
      "learning_rate": 8.533851132686086e-06,
      "loss": 0.1481,
      "step": 221440
    },
    {
      "epoch": 86.0,
      "learning_rate": 8.533333333333335e-06,
      "loss": 0.0284,
      "step": 221450
    },
    {
      "epoch": 86.0,
      "eval_accuracy": 0.9499312242090784,
      "eval_loss": 0.3749178946018219,
      "eval_runtime": 8.1935,
      "eval_samples_per_second": 443.645,
      "eval_steps_per_second": 55.532,
      "step": 221450
    },
    {
      "epoch": 86.0,
      "learning_rate": 8.532815533980583e-06,
      "loss": 0.0116,
      "step": 221460
    },
    {
      "epoch": 86.01,
      "learning_rate": 8.532297734627833e-06,
      "loss": 0.1044,
      "step": 221470
    },
    {
      "epoch": 86.01,
      "learning_rate": 8.531779935275082e-06,
      "loss": 0.0464,
      "step": 221480
    },
    {
      "epoch": 86.02,
      "learning_rate": 8.531262135922332e-06,
      "loss": 0.011,
      "step": 221490
    },
    {
      "epoch": 86.02,
      "learning_rate": 8.53074433656958e-06,
      "loss": 0.0044,
      "step": 221500
    },
    {
      "epoch": 86.02,
      "learning_rate": 8.53022653721683e-06,
      "loss": 0.2006,
      "step": 221510
    },
    {
      "epoch": 86.03,
      "learning_rate": 8.529708737864079e-06,
      "loss": 0.053,
      "step": 221520
    },
    {
      "epoch": 86.03,
      "learning_rate": 8.529190938511329e-06,
      "loss": 0.0808,
      "step": 221530
    },
    {
      "epoch": 86.03,
      "learning_rate": 8.528673139158577e-06,
      "loss": 0.1504,
      "step": 221540
    },
    {
      "epoch": 86.04,
      "learning_rate": 8.528155339805826e-06,
      "loss": 0.1345,
      "step": 221550
    },
    {
      "epoch": 86.04,
      "learning_rate": 8.527637540453076e-06,
      "loss": 0.0276,
      "step": 221560
    },
    {
      "epoch": 86.05,
      "learning_rate": 8.527119741100326e-06,
      "loss": 0.0668,
      "step": 221570
    },
    {
      "epoch": 86.05,
      "learning_rate": 8.526601941747573e-06,
      "loss": 0.0452,
      "step": 221580
    },
    {
      "epoch": 86.05,
      "learning_rate": 8.526084142394823e-06,
      "loss": 0.0217,
      "step": 221590
    },
    {
      "epoch": 86.06,
      "learning_rate": 8.525566343042073e-06,
      "loss": 0.01,
      "step": 221600
    },
    {
      "epoch": 86.06,
      "learning_rate": 8.525048543689322e-06,
      "loss": 0.0358,
      "step": 221610
    },
    {
      "epoch": 86.07,
      "learning_rate": 8.52453074433657e-06,
      "loss": 0.0481,
      "step": 221620
    },
    {
      "epoch": 86.07,
      "learning_rate": 8.52401294498382e-06,
      "loss": 0.098,
      "step": 221630
    },
    {
      "epoch": 86.07,
      "learning_rate": 8.52349514563107e-06,
      "loss": 0.068,
      "step": 221640
    },
    {
      "epoch": 86.08,
      "learning_rate": 8.522977346278319e-06,
      "loss": 0.0619,
      "step": 221650
    },
    {
      "epoch": 86.08,
      "learning_rate": 8.522459546925567e-06,
      "loss": 0.0599,
      "step": 221660
    },
    {
      "epoch": 86.09,
      "learning_rate": 8.521941747572817e-06,
      "loss": 0.0086,
      "step": 221670
    },
    {
      "epoch": 86.09,
      "learning_rate": 8.521423948220066e-06,
      "loss": 0.0331,
      "step": 221680
    },
    {
      "epoch": 86.09,
      "learning_rate": 8.520906148867316e-06,
      "loss": 0.0763,
      "step": 221690
    },
    {
      "epoch": 86.1,
      "learning_rate": 8.520388349514564e-06,
      "loss": 0.1874,
      "step": 221700
    },
    {
      "epoch": 86.1,
      "learning_rate": 8.519870550161813e-06,
      "loss": 0.0007,
      "step": 221710
    },
    {
      "epoch": 86.1,
      "learning_rate": 8.519352750809063e-06,
      "loss": 0.013,
      "step": 221720
    },
    {
      "epoch": 86.11,
      "learning_rate": 8.518834951456311e-06,
      "loss": 0.0152,
      "step": 221730
    },
    {
      "epoch": 86.11,
      "learning_rate": 8.51831715210356e-06,
      "loss": 0.0543,
      "step": 221740
    },
    {
      "epoch": 86.12,
      "learning_rate": 8.51779935275081e-06,
      "loss": 0.023,
      "step": 221750
    },
    {
      "epoch": 86.12,
      "learning_rate": 8.51728155339806e-06,
      "loss": 0.1119,
      "step": 221760
    },
    {
      "epoch": 86.12,
      "learning_rate": 8.516763754045308e-06,
      "loss": 0.0116,
      "step": 221770
    },
    {
      "epoch": 86.13,
      "learning_rate": 8.516245954692557e-06,
      "loss": 0.0264,
      "step": 221780
    },
    {
      "epoch": 86.13,
      "learning_rate": 8.515728155339807e-06,
      "loss": 0.0003,
      "step": 221790
    },
    {
      "epoch": 86.14,
      "learning_rate": 8.515210355987057e-06,
      "loss": 0.0723,
      "step": 221800
    },
    {
      "epoch": 86.14,
      "learning_rate": 8.514692556634305e-06,
      "loss": 0.0229,
      "step": 221810
    },
    {
      "epoch": 86.14,
      "learning_rate": 8.514174757281554e-06,
      "loss": 0.0715,
      "step": 221820
    },
    {
      "epoch": 86.15,
      "learning_rate": 8.513656957928804e-06,
      "loss": 0.0554,
      "step": 221830
    },
    {
      "epoch": 86.15,
      "learning_rate": 8.513139158576053e-06,
      "loss": 0.0339,
      "step": 221840
    },
    {
      "epoch": 86.16,
      "learning_rate": 8.512621359223301e-06,
      "loss": 0.0086,
      "step": 221850
    },
    {
      "epoch": 86.16,
      "learning_rate": 8.512103559870551e-06,
      "loss": 0.0538,
      "step": 221860
    },
    {
      "epoch": 86.16,
      "learning_rate": 8.5115857605178e-06,
      "loss": 0.076,
      "step": 221870
    },
    {
      "epoch": 86.17,
      "learning_rate": 8.511067961165048e-06,
      "loss": 0.0174,
      "step": 221880
    },
    {
      "epoch": 86.17,
      "learning_rate": 8.510550161812298e-06,
      "loss": 0.0708,
      "step": 221890
    },
    {
      "epoch": 86.17,
      "learning_rate": 8.510032362459548e-06,
      "loss": 0.0091,
      "step": 221900
    },
    {
      "epoch": 86.18,
      "learning_rate": 8.509514563106797e-06,
      "loss": 0.0157,
      "step": 221910
    },
    {
      "epoch": 86.18,
      "learning_rate": 8.508996763754045e-06,
      "loss": 0.0733,
      "step": 221920
    },
    {
      "epoch": 86.19,
      "learning_rate": 8.508478964401295e-06,
      "loss": 0.1329,
      "step": 221930
    },
    {
      "epoch": 86.19,
      "learning_rate": 8.507961165048544e-06,
      "loss": 0.0348,
      "step": 221940
    },
    {
      "epoch": 86.19,
      "learning_rate": 8.507443365695794e-06,
      "loss": 0.0628,
      "step": 221950
    },
    {
      "epoch": 86.2,
      "learning_rate": 8.506925566343042e-06,
      "loss": 0.1106,
      "step": 221960
    },
    {
      "epoch": 86.2,
      "learning_rate": 8.506407766990292e-06,
      "loss": 0.0638,
      "step": 221970
    },
    {
      "epoch": 86.21,
      "learning_rate": 8.505889967637541e-06,
      "loss": 0.0795,
      "step": 221980
    },
    {
      "epoch": 86.21,
      "learning_rate": 8.505372168284791e-06,
      "loss": 0.0317,
      "step": 221990
    },
    {
      "epoch": 86.21,
      "learning_rate": 8.504854368932039e-06,
      "loss": 0.0663,
      "step": 222000
    },
    {
      "epoch": 86.22,
      "learning_rate": 8.504336569579288e-06,
      "loss": 0.0414,
      "step": 222010
    },
    {
      "epoch": 86.22,
      "learning_rate": 8.503818770226538e-06,
      "loss": 0.0403,
      "step": 222020
    },
    {
      "epoch": 86.23,
      "learning_rate": 8.503300970873788e-06,
      "loss": 0.0325,
      "step": 222030
    },
    {
      "epoch": 86.23,
      "learning_rate": 8.502783171521036e-06,
      "loss": 0.0086,
      "step": 222040
    },
    {
      "epoch": 86.23,
      "learning_rate": 8.502265372168285e-06,
      "loss": 0.01,
      "step": 222050
    },
    {
      "epoch": 86.24,
      "learning_rate": 8.501747572815535e-06,
      "loss": 0.0634,
      "step": 222060
    },
    {
      "epoch": 86.24,
      "learning_rate": 8.501229773462783e-06,
      "loss": 0.034,
      "step": 222070
    },
    {
      "epoch": 86.24,
      "learning_rate": 8.500711974110032e-06,
      "loss": 0.1227,
      "step": 222080
    },
    {
      "epoch": 86.25,
      "learning_rate": 8.500194174757282e-06,
      "loss": 0.0145,
      "step": 222090
    },
    {
      "epoch": 86.25,
      "learning_rate": 8.499676375404532e-06,
      "loss": 0.019,
      "step": 222100
    },
    {
      "epoch": 86.26,
      "learning_rate": 8.49915857605178e-06,
      "loss": 0.0522,
      "step": 222110
    },
    {
      "epoch": 86.26,
      "learning_rate": 8.498640776699029e-06,
      "loss": 0.048,
      "step": 222120
    },
    {
      "epoch": 86.26,
      "learning_rate": 8.498122977346279e-06,
      "loss": 0.01,
      "step": 222130
    },
    {
      "epoch": 86.27,
      "learning_rate": 8.497605177993528e-06,
      "loss": 0.0908,
      "step": 222140
    },
    {
      "epoch": 86.27,
      "learning_rate": 8.497087378640776e-06,
      "loss": 0.083,
      "step": 222150
    },
    {
      "epoch": 86.28,
      "learning_rate": 8.496569579288026e-06,
      "loss": 0.0309,
      "step": 222160
    },
    {
      "epoch": 86.28,
      "learning_rate": 8.496051779935276e-06,
      "loss": 0.0478,
      "step": 222170
    },
    {
      "epoch": 86.28,
      "learning_rate": 8.495533980582525e-06,
      "loss": 0.0396,
      "step": 222180
    },
    {
      "epoch": 86.29,
      "learning_rate": 8.495016181229773e-06,
      "loss": 0.0735,
      "step": 222190
    },
    {
      "epoch": 86.29,
      "learning_rate": 8.494498381877023e-06,
      "loss": 0.0202,
      "step": 222200
    },
    {
      "epoch": 86.3,
      "learning_rate": 8.493980582524272e-06,
      "loss": 0.0003,
      "step": 222210
    },
    {
      "epoch": 86.3,
      "learning_rate": 8.493462783171522e-06,
      "loss": 0.0214,
      "step": 222220
    },
    {
      "epoch": 86.3,
      "learning_rate": 8.49294498381877e-06,
      "loss": 0.091,
      "step": 222230
    },
    {
      "epoch": 86.31,
      "learning_rate": 8.49242718446602e-06,
      "loss": 0.0344,
      "step": 222240
    },
    {
      "epoch": 86.31,
      "learning_rate": 8.491909385113269e-06,
      "loss": 0.0398,
      "step": 222250
    },
    {
      "epoch": 86.31,
      "learning_rate": 8.491391585760519e-06,
      "loss": 0.0004,
      "step": 222260
    },
    {
      "epoch": 86.32,
      "learning_rate": 8.490873786407767e-06,
      "loss": 0.1373,
      "step": 222270
    },
    {
      "epoch": 86.32,
      "learning_rate": 8.490355987055016e-06,
      "loss": 0.014,
      "step": 222280
    },
    {
      "epoch": 86.33,
      "learning_rate": 8.489838187702266e-06,
      "loss": 0.2258,
      "step": 222290
    },
    {
      "epoch": 86.33,
      "learning_rate": 8.489320388349515e-06,
      "loss": 0.0074,
      "step": 222300
    },
    {
      "epoch": 86.33,
      "learning_rate": 8.488802588996763e-06,
      "loss": 0.0382,
      "step": 222310
    },
    {
      "epoch": 86.34,
      "learning_rate": 8.488284789644013e-06,
      "loss": 0.0773,
      "step": 222320
    },
    {
      "epoch": 86.34,
      "learning_rate": 8.487766990291263e-06,
      "loss": 0.0011,
      "step": 222330
    },
    {
      "epoch": 86.35,
      "learning_rate": 8.487249190938512e-06,
      "loss": 0.0087,
      "step": 222340
    },
    {
      "epoch": 86.35,
      "learning_rate": 8.48673139158576e-06,
      "loss": 0.0064,
      "step": 222350
    },
    {
      "epoch": 86.35,
      "learning_rate": 8.48621359223301e-06,
      "loss": 0.0292,
      "step": 222360
    },
    {
      "epoch": 86.36,
      "learning_rate": 8.48569579288026e-06,
      "loss": 0.0007,
      "step": 222370
    },
    {
      "epoch": 86.36,
      "learning_rate": 8.485177993527509e-06,
      "loss": 0.0105,
      "step": 222380
    },
    {
      "epoch": 86.37,
      "learning_rate": 8.484660194174757e-06,
      "loss": 0.1652,
      "step": 222390
    },
    {
      "epoch": 86.37,
      "learning_rate": 8.484142394822007e-06,
      "loss": 0.135,
      "step": 222400
    },
    {
      "epoch": 86.37,
      "learning_rate": 8.483624595469256e-06,
      "loss": 0.2559,
      "step": 222410
    },
    {
      "epoch": 86.38,
      "learning_rate": 8.483106796116506e-06,
      "loss": 0.2507,
      "step": 222420
    },
    {
      "epoch": 86.38,
      "learning_rate": 8.482588996763754e-06,
      "loss": 0.0002,
      "step": 222430
    },
    {
      "epoch": 86.38,
      "learning_rate": 8.482071197411003e-06,
      "loss": 0.0647,
      "step": 222440
    },
    {
      "epoch": 86.39,
      "learning_rate": 8.481553398058253e-06,
      "loss": 0.0118,
      "step": 222450
    },
    {
      "epoch": 86.39,
      "learning_rate": 8.481035598705503e-06,
      "loss": 0.0118,
      "step": 222460
    },
    {
      "epoch": 86.4,
      "learning_rate": 8.48051779935275e-06,
      "loss": 0.0465,
      "step": 222470
    },
    {
      "epoch": 86.4,
      "learning_rate": 8.48e-06,
      "loss": 0.0053,
      "step": 222480
    },
    {
      "epoch": 86.4,
      "learning_rate": 8.47948220064725e-06,
      "loss": 0.0015,
      "step": 222490
    },
    {
      "epoch": 86.41,
      "learning_rate": 8.4789644012945e-06,
      "loss": 0.0072,
      "step": 222500
    },
    {
      "epoch": 86.41,
      "learning_rate": 8.478446601941747e-06,
      "loss": 0.0215,
      "step": 222510
    },
    {
      "epoch": 86.42,
      "learning_rate": 8.477928802588997e-06,
      "loss": 0.0061,
      "step": 222520
    },
    {
      "epoch": 86.42,
      "learning_rate": 8.477411003236247e-06,
      "loss": 0.0518,
      "step": 222530
    },
    {
      "epoch": 86.42,
      "learning_rate": 8.476893203883496e-06,
      "loss": 0.0121,
      "step": 222540
    },
    {
      "epoch": 86.43,
      "learning_rate": 8.476375404530744e-06,
      "loss": 0.0146,
      "step": 222550
    },
    {
      "epoch": 86.43,
      "learning_rate": 8.475857605177994e-06,
      "loss": 0.1413,
      "step": 222560
    },
    {
      "epoch": 86.43,
      "learning_rate": 8.475339805825243e-06,
      "loss": 0.0014,
      "step": 222570
    },
    {
      "epoch": 86.44,
      "learning_rate": 8.474822006472493e-06,
      "loss": 0.0357,
      "step": 222580
    },
    {
      "epoch": 86.44,
      "learning_rate": 8.474304207119743e-06,
      "loss": 0.0593,
      "step": 222590
    },
    {
      "epoch": 86.45,
      "learning_rate": 8.47378640776699e-06,
      "loss": 0.0486,
      "step": 222600
    },
    {
      "epoch": 86.45,
      "learning_rate": 8.47326860841424e-06,
      "loss": 0.0722,
      "step": 222610
    },
    {
      "epoch": 86.45,
      "learning_rate": 8.47275080906149e-06,
      "loss": 0.0628,
      "step": 222620
    },
    {
      "epoch": 86.46,
      "learning_rate": 8.47223300970874e-06,
      "loss": 0.0345,
      "step": 222630
    },
    {
      "epoch": 86.46,
      "learning_rate": 8.471715210355987e-06,
      "loss": 0.0439,
      "step": 222640
    },
    {
      "epoch": 86.47,
      "learning_rate": 8.471197411003237e-06,
      "loss": 0.0288,
      "step": 222650
    },
    {
      "epoch": 86.47,
      "learning_rate": 8.470679611650486e-06,
      "loss": 0.0083,
      "step": 222660
    },
    {
      "epoch": 86.47,
      "learning_rate": 8.470161812297736e-06,
      "loss": 0.0409,
      "step": 222670
    },
    {
      "epoch": 86.48,
      "learning_rate": 8.469644012944984e-06,
      "loss": 0.0149,
      "step": 222680
    },
    {
      "epoch": 86.48,
      "learning_rate": 8.469126213592234e-06,
      "loss": 0.0847,
      "step": 222690
    },
    {
      "epoch": 86.49,
      "learning_rate": 8.468608414239483e-06,
      "loss": 0.1681,
      "step": 222700
    },
    {
      "epoch": 86.49,
      "learning_rate": 8.468090614886733e-06,
      "loss": 0.0229,
      "step": 222710
    },
    {
      "epoch": 86.49,
      "learning_rate": 8.46757281553398e-06,
      "loss": 0.0841,
      "step": 222720
    },
    {
      "epoch": 86.5,
      "learning_rate": 8.46705501618123e-06,
      "loss": 0.0253,
      "step": 222730
    },
    {
      "epoch": 86.5,
      "learning_rate": 8.46653721682848e-06,
      "loss": 0.0577,
      "step": 222740
    },
    {
      "epoch": 86.5,
      "learning_rate": 8.46601941747573e-06,
      "loss": 0.0145,
      "step": 222750
    },
    {
      "epoch": 86.51,
      "learning_rate": 8.465501618122978e-06,
      "loss": 0.0352,
      "step": 222760
    },
    {
      "epoch": 86.51,
      "learning_rate": 8.464983818770227e-06,
      "loss": 0.0545,
      "step": 222770
    },
    {
      "epoch": 86.52,
      "learning_rate": 8.464466019417477e-06,
      "loss": 0.032,
      "step": 222780
    },
    {
      "epoch": 86.52,
      "learning_rate": 8.463948220064726e-06,
      "loss": 0.0355,
      "step": 222790
    },
    {
      "epoch": 86.52,
      "learning_rate": 8.463430420711974e-06,
      "loss": 0.0819,
      "step": 222800
    },
    {
      "epoch": 86.53,
      "learning_rate": 8.462912621359224e-06,
      "loss": 0.0079,
      "step": 222810
    },
    {
      "epoch": 86.53,
      "learning_rate": 8.462394822006474e-06,
      "loss": 0.1389,
      "step": 222820
    },
    {
      "epoch": 86.54,
      "learning_rate": 8.461877022653723e-06,
      "loss": 0.0111,
      "step": 222830
    },
    {
      "epoch": 86.54,
      "learning_rate": 8.461359223300971e-06,
      "loss": 0.0041,
      "step": 222840
    },
    {
      "epoch": 86.54,
      "learning_rate": 8.46084142394822e-06,
      "loss": 0.1657,
      "step": 222850
    },
    {
      "epoch": 86.55,
      "learning_rate": 8.46032362459547e-06,
      "loss": 0.0967,
      "step": 222860
    },
    {
      "epoch": 86.55,
      "learning_rate": 8.45980582524272e-06,
      "loss": 0.0016,
      "step": 222870
    },
    {
      "epoch": 86.56,
      "learning_rate": 8.459288025889968e-06,
      "loss": 0.0011,
      "step": 222880
    },
    {
      "epoch": 86.56,
      "learning_rate": 8.458770226537218e-06,
      "loss": 0.0028,
      "step": 222890
    },
    {
      "epoch": 86.56,
      "learning_rate": 8.458252427184467e-06,
      "loss": 0.065,
      "step": 222900
    },
    {
      "epoch": 86.57,
      "learning_rate": 8.457734627831717e-06,
      "loss": 0.0802,
      "step": 222910
    },
    {
      "epoch": 86.57,
      "learning_rate": 8.457216828478965e-06,
      "loss": 0.0115,
      "step": 222920
    },
    {
      "epoch": 86.57,
      "learning_rate": 8.456699029126214e-06,
      "loss": 0.1321,
      "step": 222930
    },
    {
      "epoch": 86.58,
      "learning_rate": 8.456181229773464e-06,
      "loss": 0.105,
      "step": 222940
    },
    {
      "epoch": 86.58,
      "learning_rate": 8.455663430420714e-06,
      "loss": 0.0445,
      "step": 222950
    },
    {
      "epoch": 86.59,
      "learning_rate": 8.455145631067961e-06,
      "loss": 0.0478,
      "step": 222960
    },
    {
      "epoch": 86.59,
      "learning_rate": 8.454627831715211e-06,
      "loss": 0.2093,
      "step": 222970
    },
    {
      "epoch": 86.59,
      "learning_rate": 8.45411003236246e-06,
      "loss": 0.0468,
      "step": 222980
    },
    {
      "epoch": 86.6,
      "learning_rate": 8.45359223300971e-06,
      "loss": 0.0963,
      "step": 222990
    },
    {
      "epoch": 86.6,
      "learning_rate": 8.453074433656958e-06,
      "loss": 0.0207,
      "step": 223000
    },
    {
      "epoch": 86.61,
      "learning_rate": 8.452556634304208e-06,
      "loss": 0.0615,
      "step": 223010
    },
    {
      "epoch": 86.61,
      "learning_rate": 8.452038834951457e-06,
      "loss": 0.0899,
      "step": 223020
    },
    {
      "epoch": 86.61,
      "learning_rate": 8.451521035598707e-06,
      "loss": 0.0269,
      "step": 223030
    },
    {
      "epoch": 86.62,
      "learning_rate": 8.451003236245955e-06,
      "loss": 0.0119,
      "step": 223040
    },
    {
      "epoch": 86.62,
      "learning_rate": 8.450485436893205e-06,
      "loss": 0.0856,
      "step": 223050
    },
    {
      "epoch": 86.63,
      "learning_rate": 8.449967637540454e-06,
      "loss": 0.0132,
      "step": 223060
    },
    {
      "epoch": 86.63,
      "learning_rate": 8.449449838187704e-06,
      "loss": 0.0475,
      "step": 223070
    },
    {
      "epoch": 86.63,
      "learning_rate": 8.448932038834952e-06,
      "loss": 0.0422,
      "step": 223080
    },
    {
      "epoch": 86.64,
      "learning_rate": 8.448414239482201e-06,
      "loss": 0.0844,
      "step": 223090
    },
    {
      "epoch": 86.64,
      "learning_rate": 8.447896440129451e-06,
      "loss": 0.0971,
      "step": 223100
    },
    {
      "epoch": 86.64,
      "learning_rate": 8.4473786407767e-06,
      "loss": 0.0508,
      "step": 223110
    },
    {
      "epoch": 86.65,
      "learning_rate": 8.44686084142395e-06,
      "loss": 0.0832,
      "step": 223120
    },
    {
      "epoch": 86.65,
      "learning_rate": 8.446343042071198e-06,
      "loss": 0.0911,
      "step": 223130
    },
    {
      "epoch": 86.66,
      "learning_rate": 8.445825242718448e-06,
      "loss": 0.0476,
      "step": 223140
    },
    {
      "epoch": 86.66,
      "learning_rate": 8.445307443365697e-06,
      "loss": 0.0892,
      "step": 223150
    },
    {
      "epoch": 86.66,
      "learning_rate": 8.444789644012947e-06,
      "loss": 0.0012,
      "step": 223160
    },
    {
      "epoch": 86.67,
      "learning_rate": 8.444271844660195e-06,
      "loss": 0.1167,
      "step": 223170
    },
    {
      "epoch": 86.67,
      "learning_rate": 8.443754045307445e-06,
      "loss": 0.002,
      "step": 223180
    },
    {
      "epoch": 86.68,
      "learning_rate": 8.443236245954694e-06,
      "loss": 0.0595,
      "step": 223190
    },
    {
      "epoch": 86.68,
      "learning_rate": 8.442718446601942e-06,
      "loss": 0.002,
      "step": 223200
    },
    {
      "epoch": 86.68,
      "learning_rate": 8.442200647249192e-06,
      "loss": 0.1019,
      "step": 223210
    },
    {
      "epoch": 86.69,
      "learning_rate": 8.441682847896441e-06,
      "loss": 0.1163,
      "step": 223220
    },
    {
      "epoch": 86.69,
      "learning_rate": 8.441165048543691e-06,
      "loss": 0.0472,
      "step": 223230
    },
    {
      "epoch": 86.7,
      "learning_rate": 8.440647249190939e-06,
      "loss": 0.1311,
      "step": 223240
    },
    {
      "epoch": 86.7,
      "learning_rate": 8.440129449838189e-06,
      "loss": 0.2311,
      "step": 223250
    },
    {
      "epoch": 86.7,
      "learning_rate": 8.439611650485438e-06,
      "loss": 0.1632,
      "step": 223260
    },
    {
      "epoch": 86.71,
      "learning_rate": 8.439093851132688e-06,
      "loss": 0.0245,
      "step": 223270
    },
    {
      "epoch": 86.71,
      "learning_rate": 8.438576051779936e-06,
      "loss": 0.0516,
      "step": 223280
    },
    {
      "epoch": 86.71,
      "learning_rate": 8.438058252427185e-06,
      "loss": 0.0398,
      "step": 223290
    },
    {
      "epoch": 86.72,
      "learning_rate": 8.437540453074435e-06,
      "loss": 0.0357,
      "step": 223300
    },
    {
      "epoch": 86.72,
      "learning_rate": 8.437022653721685e-06,
      "loss": 0.1833,
      "step": 223310
    },
    {
      "epoch": 86.73,
      "learning_rate": 8.436504854368932e-06,
      "loss": 0.0093,
      "step": 223320
    },
    {
      "epoch": 86.73,
      "learning_rate": 8.435987055016182e-06,
      "loss": 0.0111,
      "step": 223330
    },
    {
      "epoch": 86.73,
      "learning_rate": 8.435469255663432e-06,
      "loss": 0.0628,
      "step": 223340
    },
    {
      "epoch": 86.74,
      "learning_rate": 8.43495145631068e-06,
      "loss": 0.0401,
      "step": 223350
    },
    {
      "epoch": 86.74,
      "learning_rate": 8.43443365695793e-06,
      "loss": 0.0446,
      "step": 223360
    },
    {
      "epoch": 86.75,
      "learning_rate": 8.433915857605179e-06,
      "loss": 0.0202,
      "step": 223370
    },
    {
      "epoch": 86.75,
      "learning_rate": 8.433398058252428e-06,
      "loss": 0.0106,
      "step": 223380
    },
    {
      "epoch": 86.75,
      "learning_rate": 8.432880258899676e-06,
      "loss": 0.0027,
      "step": 223390
    },
    {
      "epoch": 86.76,
      "learning_rate": 8.432362459546926e-06,
      "loss": 0.1113,
      "step": 223400
    },
    {
      "epoch": 86.76,
      "learning_rate": 8.431844660194176e-06,
      "loss": 0.0806,
      "step": 223410
    },
    {
      "epoch": 86.77,
      "learning_rate": 8.431326860841425e-06,
      "loss": 0.083,
      "step": 223420
    },
    {
      "epoch": 86.77,
      "learning_rate": 8.430809061488673e-06,
      "loss": 0.0755,
      "step": 223430
    },
    {
      "epoch": 86.77,
      "learning_rate": 8.430291262135923e-06,
      "loss": 0.0189,
      "step": 223440
    },
    {
      "epoch": 86.78,
      "learning_rate": 8.429773462783172e-06,
      "loss": 0.0725,
      "step": 223450
    },
    {
      "epoch": 86.78,
      "learning_rate": 8.429255663430422e-06,
      "loss": 0.024,
      "step": 223460
    },
    {
      "epoch": 86.78,
      "learning_rate": 8.42873786407767e-06,
      "loss": 0.0088,
      "step": 223470
    },
    {
      "epoch": 86.79,
      "learning_rate": 8.42822006472492e-06,
      "loss": 0.0364,
      "step": 223480
    },
    {
      "epoch": 86.79,
      "learning_rate": 8.42770226537217e-06,
      "loss": 0.0204,
      "step": 223490
    },
    {
      "epoch": 86.8,
      "learning_rate": 8.427184466019419e-06,
      "loss": 0.1066,
      "step": 223500
    },
    {
      "epoch": 86.8,
      "learning_rate": 8.426666666666667e-06,
      "loss": 0.0957,
      "step": 223510
    },
    {
      "epoch": 86.8,
      "learning_rate": 8.426148867313916e-06,
      "loss": 0.0838,
      "step": 223520
    },
    {
      "epoch": 86.81,
      "learning_rate": 8.425631067961166e-06,
      "loss": 0.1694,
      "step": 223530
    },
    {
      "epoch": 86.81,
      "learning_rate": 8.425113268608414e-06,
      "loss": 0.0624,
      "step": 223540
    },
    {
      "epoch": 86.82,
      "learning_rate": 8.424595469255664e-06,
      "loss": 0.0248,
      "step": 223550
    },
    {
      "epoch": 86.82,
      "learning_rate": 8.424077669902913e-06,
      "loss": 0.1287,
      "step": 223560
    },
    {
      "epoch": 86.82,
      "learning_rate": 8.423559870550163e-06,
      "loss": 0.0522,
      "step": 223570
    },
    {
      "epoch": 86.83,
      "learning_rate": 8.42304207119741e-06,
      "loss": 0.0191,
      "step": 223580
    },
    {
      "epoch": 86.83,
      "learning_rate": 8.42252427184466e-06,
      "loss": 0.0344,
      "step": 223590
    },
    {
      "epoch": 86.83,
      "learning_rate": 8.42200647249191e-06,
      "loss": 0.0106,
      "step": 223600
    },
    {
      "epoch": 86.84,
      "learning_rate": 8.42148867313916e-06,
      "loss": 0.0673,
      "step": 223610
    },
    {
      "epoch": 86.84,
      "learning_rate": 8.420970873786407e-06,
      "loss": 0.0428,
      "step": 223620
    },
    {
      "epoch": 86.85,
      "learning_rate": 8.420453074433657e-06,
      "loss": 0.1552,
      "step": 223630
    },
    {
      "epoch": 86.85,
      "learning_rate": 8.419935275080907e-06,
      "loss": 0.0835,
      "step": 223640
    },
    {
      "epoch": 86.85,
      "learning_rate": 8.419417475728156e-06,
      "loss": 0.0665,
      "step": 223650
    },
    {
      "epoch": 86.86,
      "learning_rate": 8.418899676375404e-06,
      "loss": 0.0489,
      "step": 223660
    },
    {
      "epoch": 86.86,
      "learning_rate": 8.418381877022654e-06,
      "loss": 0.0006,
      "step": 223670
    },
    {
      "epoch": 86.87,
      "learning_rate": 8.417864077669903e-06,
      "loss": 0.0769,
      "step": 223680
    },
    {
      "epoch": 86.87,
      "learning_rate": 8.417346278317153e-06,
      "loss": 0.1649,
      "step": 223690
    },
    {
      "epoch": 86.87,
      "learning_rate": 8.416828478964401e-06,
      "loss": 0.0712,
      "step": 223700
    },
    {
      "epoch": 86.88,
      "learning_rate": 8.41631067961165e-06,
      "loss": 0.0379,
      "step": 223710
    },
    {
      "epoch": 86.88,
      "learning_rate": 8.4157928802589e-06,
      "loss": 0.1122,
      "step": 223720
    },
    {
      "epoch": 86.89,
      "learning_rate": 8.41527508090615e-06,
      "loss": 0.0957,
      "step": 223730
    },
    {
      "epoch": 86.89,
      "learning_rate": 8.414757281553398e-06,
      "loss": 0.0081,
      "step": 223740
    },
    {
      "epoch": 86.89,
      "learning_rate": 8.414239482200647e-06,
      "loss": 0.0154,
      "step": 223750
    },
    {
      "epoch": 86.9,
      "learning_rate": 8.413721682847897e-06,
      "loss": 0.0979,
      "step": 223760
    },
    {
      "epoch": 86.9,
      "learning_rate": 8.413203883495147e-06,
      "loss": 0.0171,
      "step": 223770
    },
    {
      "epoch": 86.9,
      "learning_rate": 8.412686084142395e-06,
      "loss": 0.1473,
      "step": 223780
    },
    {
      "epoch": 86.91,
      "learning_rate": 8.412168284789644e-06,
      "loss": 0.1239,
      "step": 223790
    },
    {
      "epoch": 86.91,
      "learning_rate": 8.411650485436894e-06,
      "loss": 0.026,
      "step": 223800
    },
    {
      "epoch": 86.92,
      "learning_rate": 8.411132686084143e-06,
      "loss": 0.0664,
      "step": 223810
    },
    {
      "epoch": 86.92,
      "learning_rate": 8.410614886731391e-06,
      "loss": 0.0413,
      "step": 223820
    },
    {
      "epoch": 86.92,
      "learning_rate": 8.410097087378641e-06,
      "loss": 0.0655,
      "step": 223830
    },
    {
      "epoch": 86.93,
      "learning_rate": 8.40957928802589e-06,
      "loss": 0.0653,
      "step": 223840
    },
    {
      "epoch": 86.93,
      "learning_rate": 8.40906148867314e-06,
      "loss": 0.0003,
      "step": 223850
    },
    {
      "epoch": 86.94,
      "learning_rate": 8.408543689320388e-06,
      "loss": 0.008,
      "step": 223860
    },
    {
      "epoch": 86.94,
      "learning_rate": 8.408025889967638e-06,
      "loss": 0.0388,
      "step": 223870
    },
    {
      "epoch": 86.94,
      "learning_rate": 8.407508090614887e-06,
      "loss": 0.0762,
      "step": 223880
    },
    {
      "epoch": 86.95,
      "learning_rate": 8.406990291262137e-06,
      "loss": 0.0265,
      "step": 223890
    },
    {
      "epoch": 86.95,
      "learning_rate": 8.406472491909385e-06,
      "loss": 0.1949,
      "step": 223900
    },
    {
      "epoch": 86.96,
      "learning_rate": 8.405954692556635e-06,
      "loss": 0.0312,
      "step": 223910
    },
    {
      "epoch": 86.96,
      "learning_rate": 8.405436893203884e-06,
      "loss": 0.0711,
      "step": 223920
    },
    {
      "epoch": 86.96,
      "learning_rate": 8.404919093851134e-06,
      "loss": 0.0967,
      "step": 223930
    },
    {
      "epoch": 86.97,
      "learning_rate": 8.404401294498382e-06,
      "loss": 0.0721,
      "step": 223940
    },
    {
      "epoch": 86.97,
      "learning_rate": 8.403883495145631e-06,
      "loss": 0.03,
      "step": 223950
    },
    {
      "epoch": 86.97,
      "learning_rate": 8.403365695792881e-06,
      "loss": 0.022,
      "step": 223960
    },
    {
      "epoch": 86.98,
      "learning_rate": 8.40284789644013e-06,
      "loss": 0.002,
      "step": 223970
    },
    {
      "epoch": 86.98,
      "learning_rate": 8.402330097087378e-06,
      "loss": 0.0506,
      "step": 223980
    },
    {
      "epoch": 86.99,
      "learning_rate": 8.401812297734628e-06,
      "loss": 0.197,
      "step": 223990
    },
    {
      "epoch": 86.99,
      "learning_rate": 8.401294498381878e-06,
      "loss": 0.1186,
      "step": 224000
    },
    {
      "epoch": 86.99,
      "learning_rate": 8.400776699029127e-06,
      "loss": 0.03,
      "step": 224010
    },
    {
      "epoch": 87.0,
      "learning_rate": 8.400258899676375e-06,
      "loss": 0.0923,
      "step": 224020
    },
    {
      "epoch": 87.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.35266244411468506,
      "eval_runtime": 8.1691,
      "eval_samples_per_second": 444.971,
      "eval_steps_per_second": 55.698,
      "step": 224025
    },
    {
      "epoch": 87.0,
      "learning_rate": 8.399741100323625e-06,
      "loss": 0.1982,
      "step": 224030
    },
    {
      "epoch": 87.01,
      "learning_rate": 8.399223300970874e-06,
      "loss": 0.0222,
      "step": 224040
    },
    {
      "epoch": 87.01,
      "learning_rate": 8.398705501618124e-06,
      "loss": 0.0012,
      "step": 224050
    },
    {
      "epoch": 87.01,
      "learning_rate": 8.398187702265372e-06,
      "loss": 0.1271,
      "step": 224060
    },
    {
      "epoch": 87.02,
      "learning_rate": 8.397669902912622e-06,
      "loss": 0.0492,
      "step": 224070
    },
    {
      "epoch": 87.02,
      "learning_rate": 8.397152103559871e-06,
      "loss": 0.0158,
      "step": 224080
    },
    {
      "epoch": 87.03,
      "learning_rate": 8.396634304207121e-06,
      "loss": 0.0628,
      "step": 224090
    },
    {
      "epoch": 87.03,
      "learning_rate": 8.396116504854369e-06,
      "loss": 0.0428,
      "step": 224100
    },
    {
      "epoch": 87.03,
      "learning_rate": 8.395598705501618e-06,
      "loss": 0.0486,
      "step": 224110
    },
    {
      "epoch": 87.04,
      "learning_rate": 8.395080906148868e-06,
      "loss": 0.0104,
      "step": 224120
    },
    {
      "epoch": 87.04,
      "learning_rate": 8.394563106796118e-06,
      "loss": 0.0591,
      "step": 224130
    },
    {
      "epoch": 87.04,
      "learning_rate": 8.394045307443366e-06,
      "loss": 0.0153,
      "step": 224140
    },
    {
      "epoch": 87.05,
      "learning_rate": 8.393527508090615e-06,
      "loss": 0.0597,
      "step": 224150
    },
    {
      "epoch": 87.05,
      "learning_rate": 8.393009708737865e-06,
      "loss": 0.0198,
      "step": 224160
    },
    {
      "epoch": 87.06,
      "learning_rate": 8.392491909385114e-06,
      "loss": 0.0007,
      "step": 224170
    },
    {
      "epoch": 87.06,
      "learning_rate": 8.391974110032362e-06,
      "loss": 0.049,
      "step": 224180
    },
    {
      "epoch": 87.06,
      "learning_rate": 8.391456310679612e-06,
      "loss": 0.068,
      "step": 224190
    },
    {
      "epoch": 87.07,
      "learning_rate": 8.390938511326862e-06,
      "loss": 0.0818,
      "step": 224200
    },
    {
      "epoch": 87.07,
      "learning_rate": 8.390420711974111e-06,
      "loss": 0.0041,
      "step": 224210
    },
    {
      "epoch": 87.08,
      "learning_rate": 8.389902912621359e-06,
      "loss": 0.0219,
      "step": 224220
    },
    {
      "epoch": 87.08,
      "learning_rate": 8.389385113268609e-06,
      "loss": 0.0581,
      "step": 224230
    },
    {
      "epoch": 87.08,
      "learning_rate": 8.388867313915858e-06,
      "loss": 0.0361,
      "step": 224240
    },
    {
      "epoch": 87.09,
      "learning_rate": 8.388349514563108e-06,
      "loss": 0.019,
      "step": 224250
    },
    {
      "epoch": 87.09,
      "learning_rate": 8.387831715210358e-06,
      "loss": 0.0736,
      "step": 224260
    },
    {
      "epoch": 87.1,
      "learning_rate": 8.387313915857606e-06,
      "loss": 0.0264,
      "step": 224270
    },
    {
      "epoch": 87.1,
      "learning_rate": 8.386796116504855e-06,
      "loss": 0.1067,
      "step": 224280
    },
    {
      "epoch": 87.1,
      "learning_rate": 8.386278317152105e-06,
      "loss": 0.1028,
      "step": 224290
    },
    {
      "epoch": 87.11,
      "learning_rate": 8.385760517799354e-06,
      "loss": 0.0497,
      "step": 224300
    },
    {
      "epoch": 87.11,
      "learning_rate": 8.385242718446602e-06,
      "loss": 0.0485,
      "step": 224310
    },
    {
      "epoch": 87.11,
      "learning_rate": 8.384724919093852e-06,
      "loss": 0.1011,
      "step": 224320
    },
    {
      "epoch": 87.12,
      "learning_rate": 8.384207119741102e-06,
      "loss": 0.0709,
      "step": 224330
    },
    {
      "epoch": 87.12,
      "learning_rate": 8.383689320388351e-06,
      "loss": 0.0383,
      "step": 224340
    },
    {
      "epoch": 87.13,
      "learning_rate": 8.383171521035599e-06,
      "loss": 0.0121,
      "step": 224350
    },
    {
      "epoch": 87.13,
      "learning_rate": 8.382653721682849e-06,
      "loss": 0.0617,
      "step": 224360
    },
    {
      "epoch": 87.13,
      "learning_rate": 8.382135922330098e-06,
      "loss": 0.158,
      "step": 224370
    },
    {
      "epoch": 87.14,
      "learning_rate": 8.381618122977348e-06,
      "loss": 0.0436,
      "step": 224380
    },
    {
      "epoch": 87.14,
      "learning_rate": 8.381100323624596e-06,
      "loss": 0.0524,
      "step": 224390
    },
    {
      "epoch": 87.15,
      "learning_rate": 8.380582524271845e-06,
      "loss": 0.0466,
      "step": 224400
    },
    {
      "epoch": 87.15,
      "learning_rate": 8.380064724919095e-06,
      "loss": 0.0789,
      "step": 224410
    },
    {
      "epoch": 87.15,
      "learning_rate": 8.379546925566345e-06,
      "loss": 0.1797,
      "step": 224420
    },
    {
      "epoch": 87.16,
      "learning_rate": 8.379029126213593e-06,
      "loss": 0.0976,
      "step": 224430
    },
    {
      "epoch": 87.16,
      "learning_rate": 8.378511326860842e-06,
      "loss": 0.0574,
      "step": 224440
    },
    {
      "epoch": 87.17,
      "learning_rate": 8.377993527508092e-06,
      "loss": 0.0044,
      "step": 224450
    },
    {
      "epoch": 87.17,
      "learning_rate": 8.377475728155341e-06,
      "loss": 0.1577,
      "step": 224460
    },
    {
      "epoch": 87.17,
      "learning_rate": 8.37695792880259e-06,
      "loss": 0.0241,
      "step": 224470
    },
    {
      "epoch": 87.18,
      "learning_rate": 8.376440129449839e-06,
      "loss": 0.024,
      "step": 224480
    },
    {
      "epoch": 87.18,
      "learning_rate": 8.375922330097089e-06,
      "loss": 0.1509,
      "step": 224490
    },
    {
      "epoch": 87.18,
      "learning_rate": 8.375404530744338e-06,
      "loss": 0.0006,
      "step": 224500
    },
    {
      "epoch": 87.19,
      "learning_rate": 8.374886731391586e-06,
      "loss": 0.0007,
      "step": 224510
    },
    {
      "epoch": 87.19,
      "learning_rate": 8.374368932038836e-06,
      "loss": 0.1691,
      "step": 224520
    },
    {
      "epoch": 87.2,
      "learning_rate": 8.373851132686085e-06,
      "loss": 0.0416,
      "step": 224530
    },
    {
      "epoch": 87.2,
      "learning_rate": 8.373333333333335e-06,
      "loss": 0.013,
      "step": 224540
    },
    {
      "epoch": 87.2,
      "learning_rate": 8.372815533980583e-06,
      "loss": 0.0501,
      "step": 224550
    },
    {
      "epoch": 87.21,
      "learning_rate": 8.372297734627833e-06,
      "loss": 0.0623,
      "step": 224560
    },
    {
      "epoch": 87.21,
      "learning_rate": 8.371779935275082e-06,
      "loss": 0.0011,
      "step": 224570
    },
    {
      "epoch": 87.22,
      "learning_rate": 8.371262135922332e-06,
      "loss": 0.1161,
      "step": 224580
    },
    {
      "epoch": 87.22,
      "learning_rate": 8.37074433656958e-06,
      "loss": 0.0697,
      "step": 224590
    },
    {
      "epoch": 87.22,
      "learning_rate": 8.37022653721683e-06,
      "loss": 0.0754,
      "step": 224600
    },
    {
      "epoch": 87.23,
      "learning_rate": 8.369708737864079e-06,
      "loss": 0.1084,
      "step": 224610
    },
    {
      "epoch": 87.23,
      "learning_rate": 8.369190938511329e-06,
      "loss": 0.159,
      "step": 224620
    },
    {
      "epoch": 87.23,
      "learning_rate": 8.368673139158577e-06,
      "loss": 0.0477,
      "step": 224630
    },
    {
      "epoch": 87.24,
      "learning_rate": 8.368155339805826e-06,
      "loss": 0.0673,
      "step": 224640
    },
    {
      "epoch": 87.24,
      "learning_rate": 8.367637540453076e-06,
      "loss": 0.0483,
      "step": 224650
    },
    {
      "epoch": 87.25,
      "learning_rate": 8.367119741100325e-06,
      "loss": 0.0701,
      "step": 224660
    },
    {
      "epoch": 87.25,
      "learning_rate": 8.366601941747573e-06,
      "loss": 0.0728,
      "step": 224670
    },
    {
      "epoch": 87.25,
      "learning_rate": 8.366084142394823e-06,
      "loss": 0.0538,
      "step": 224680
    },
    {
      "epoch": 87.26,
      "learning_rate": 8.365566343042073e-06,
      "loss": 0.0027,
      "step": 224690
    },
    {
      "epoch": 87.26,
      "learning_rate": 8.365048543689322e-06,
      "loss": 0.0203,
      "step": 224700
    },
    {
      "epoch": 87.27,
      "learning_rate": 8.36453074433657e-06,
      "loss": 0.0184,
      "step": 224710
    },
    {
      "epoch": 87.27,
      "learning_rate": 8.36401294498382e-06,
      "loss": 0.0022,
      "step": 224720
    },
    {
      "epoch": 87.27,
      "learning_rate": 8.36349514563107e-06,
      "loss": 0.0762,
      "step": 224730
    },
    {
      "epoch": 87.28,
      "learning_rate": 8.362977346278319e-06,
      "loss": 0.0505,
      "step": 224740
    },
    {
      "epoch": 87.28,
      "learning_rate": 8.362459546925567e-06,
      "loss": 0.0099,
      "step": 224750
    },
    {
      "epoch": 87.29,
      "learning_rate": 8.361941747572816e-06,
      "loss": 0.1424,
      "step": 224760
    },
    {
      "epoch": 87.29,
      "learning_rate": 8.361423948220066e-06,
      "loss": 0.0004,
      "step": 224770
    },
    {
      "epoch": 87.29,
      "learning_rate": 8.360906148867316e-06,
      "loss": 0.0136,
      "step": 224780
    },
    {
      "epoch": 87.3,
      "learning_rate": 8.360388349514564e-06,
      "loss": 0.0749,
      "step": 224790
    },
    {
      "epoch": 87.3,
      "learning_rate": 8.359870550161813e-06,
      "loss": 0.1446,
      "step": 224800
    },
    {
      "epoch": 87.3,
      "learning_rate": 8.359352750809063e-06,
      "loss": 0.0642,
      "step": 224810
    },
    {
      "epoch": 87.31,
      "learning_rate": 8.35883495145631e-06,
      "loss": 0.0085,
      "step": 224820
    },
    {
      "epoch": 87.31,
      "learning_rate": 8.35831715210356e-06,
      "loss": 0.0539,
      "step": 224830
    },
    {
      "epoch": 87.32,
      "learning_rate": 8.35779935275081e-06,
      "loss": 0.0871,
      "step": 224840
    },
    {
      "epoch": 87.32,
      "learning_rate": 8.35728155339806e-06,
      "loss": 0.0675,
      "step": 224850
    },
    {
      "epoch": 87.32,
      "learning_rate": 8.356763754045308e-06,
      "loss": 0.0614,
      "step": 224860
    },
    {
      "epoch": 87.33,
      "learning_rate": 8.356245954692557e-06,
      "loss": 0.0091,
      "step": 224870
    },
    {
      "epoch": 87.33,
      "learning_rate": 8.355728155339807e-06,
      "loss": 0.1359,
      "step": 224880
    },
    {
      "epoch": 87.34,
      "learning_rate": 8.355210355987056e-06,
      "loss": 0.0748,
      "step": 224890
    },
    {
      "epoch": 87.34,
      "learning_rate": 8.354692556634304e-06,
      "loss": 0.1367,
      "step": 224900
    },
    {
      "epoch": 87.34,
      "learning_rate": 8.354174757281554e-06,
      "loss": 0.0049,
      "step": 224910
    },
    {
      "epoch": 87.35,
      "learning_rate": 8.353656957928804e-06,
      "loss": 0.1359,
      "step": 224920
    },
    {
      "epoch": 87.35,
      "learning_rate": 8.353139158576053e-06,
      "loss": 0.2223,
      "step": 224930
    },
    {
      "epoch": 87.36,
      "learning_rate": 8.352621359223301e-06,
      "loss": 0.0086,
      "step": 224940
    },
    {
      "epoch": 87.36,
      "learning_rate": 8.35210355987055e-06,
      "loss": 0.188,
      "step": 224950
    },
    {
      "epoch": 87.36,
      "learning_rate": 8.3515857605178e-06,
      "loss": 0.2138,
      "step": 224960
    },
    {
      "epoch": 87.37,
      "learning_rate": 8.35106796116505e-06,
      "loss": 0.143,
      "step": 224970
    },
    {
      "epoch": 87.37,
      "learning_rate": 8.350550161812298e-06,
      "loss": 0.0142,
      "step": 224980
    },
    {
      "epoch": 87.37,
      "learning_rate": 8.350032362459548e-06,
      "loss": 0.0815,
      "step": 224990
    },
    {
      "epoch": 87.38,
      "learning_rate": 8.349514563106797e-06,
      "loss": 0.0159,
      "step": 225000
    },
    {
      "epoch": 87.38,
      "learning_rate": 8.348996763754045e-06,
      "loss": 0.0003,
      "step": 225010
    },
    {
      "epoch": 87.39,
      "learning_rate": 8.348478964401295e-06,
      "loss": 0.1051,
      "step": 225020
    },
    {
      "epoch": 87.39,
      "learning_rate": 8.347961165048544e-06,
      "loss": 0.0337,
      "step": 225030
    },
    {
      "epoch": 87.39,
      "learning_rate": 8.347443365695794e-06,
      "loss": 0.0109,
      "step": 225040
    },
    {
      "epoch": 87.4,
      "learning_rate": 8.346925566343042e-06,
      "loss": 0.0062,
      "step": 225050
    },
    {
      "epoch": 87.4,
      "learning_rate": 8.346407766990291e-06,
      "loss": 0.0165,
      "step": 225060
    },
    {
      "epoch": 87.41,
      "learning_rate": 8.345889967637541e-06,
      "loss": 0.0642,
      "step": 225070
    },
    {
      "epoch": 87.41,
      "learning_rate": 8.34537216828479e-06,
      "loss": 0.0335,
      "step": 225080
    },
    {
      "epoch": 87.41,
      "learning_rate": 8.344854368932039e-06,
      "loss": 0.0906,
      "step": 225090
    },
    {
      "epoch": 87.42,
      "learning_rate": 8.344336569579288e-06,
      "loss": 0.058,
      "step": 225100
    },
    {
      "epoch": 87.42,
      "learning_rate": 8.343818770226538e-06,
      "loss": 0.0254,
      "step": 225110
    },
    {
      "epoch": 87.43,
      "learning_rate": 8.343300970873787e-06,
      "loss": 0.1002,
      "step": 225120
    },
    {
      "epoch": 87.43,
      "learning_rate": 8.342783171521035e-06,
      "loss": 0.012,
      "step": 225130
    },
    {
      "epoch": 87.43,
      "learning_rate": 8.342265372168285e-06,
      "loss": 0.0162,
      "step": 225140
    },
    {
      "epoch": 87.44,
      "learning_rate": 8.341747572815535e-06,
      "loss": 0.0329,
      "step": 225150
    },
    {
      "epoch": 87.44,
      "learning_rate": 8.341229773462783e-06,
      "loss": 0.0448,
      "step": 225160
    },
    {
      "epoch": 87.44,
      "learning_rate": 8.340711974110032e-06,
      "loss": 0.0515,
      "step": 225170
    },
    {
      "epoch": 87.45,
      "learning_rate": 8.340194174757282e-06,
      "loss": 0.1003,
      "step": 225180
    },
    {
      "epoch": 87.45,
      "learning_rate": 8.339676375404531e-06,
      "loss": 0.0001,
      "step": 225190
    },
    {
      "epoch": 87.46,
      "learning_rate": 8.33915857605178e-06,
      "loss": 0.0031,
      "step": 225200
    },
    {
      "epoch": 87.46,
      "learning_rate": 8.338640776699029e-06,
      "loss": 0.0786,
      "step": 225210
    },
    {
      "epoch": 87.46,
      "learning_rate": 8.338122977346279e-06,
      "loss": 0.1756,
      "step": 225220
    },
    {
      "epoch": 87.47,
      "learning_rate": 8.337605177993528e-06,
      "loss": 0.0029,
      "step": 225230
    },
    {
      "epoch": 87.47,
      "learning_rate": 8.337087378640776e-06,
      "loss": 0.118,
      "step": 225240
    },
    {
      "epoch": 87.48,
      "learning_rate": 8.336569579288026e-06,
      "loss": 0.0257,
      "step": 225250
    },
    {
      "epoch": 87.48,
      "learning_rate": 8.336051779935275e-06,
      "loss": 0.0178,
      "step": 225260
    },
    {
      "epoch": 87.48,
      "learning_rate": 8.335533980582525e-06,
      "loss": 0.1261,
      "step": 225270
    },
    {
      "epoch": 87.49,
      "learning_rate": 8.335016181229773e-06,
      "loss": 0.0708,
      "step": 225280
    },
    {
      "epoch": 87.49,
      "learning_rate": 8.334498381877023e-06,
      "loss": 0.0226,
      "step": 225290
    },
    {
      "epoch": 87.5,
      "learning_rate": 8.333980582524272e-06,
      "loss": 0.021,
      "step": 225300
    },
    {
      "epoch": 87.5,
      "learning_rate": 8.333462783171522e-06,
      "loss": 0.1167,
      "step": 225310
    },
    {
      "epoch": 87.5,
      "learning_rate": 8.33294498381877e-06,
      "loss": 0.042,
      "step": 225320
    },
    {
      "epoch": 87.51,
      "learning_rate": 8.33242718446602e-06,
      "loss": 0.059,
      "step": 225330
    },
    {
      "epoch": 87.51,
      "learning_rate": 8.331909385113269e-06,
      "loss": 0.1111,
      "step": 225340
    },
    {
      "epoch": 87.51,
      "learning_rate": 8.331391585760519e-06,
      "loss": 0.1122,
      "step": 225350
    },
    {
      "epoch": 87.52,
      "learning_rate": 8.330873786407768e-06,
      "loss": 0.1031,
      "step": 225360
    },
    {
      "epoch": 87.52,
      "learning_rate": 8.330355987055016e-06,
      "loss": 0.0152,
      "step": 225370
    },
    {
      "epoch": 87.53,
      "learning_rate": 8.329838187702266e-06,
      "loss": 0.1022,
      "step": 225380
    },
    {
      "epoch": 87.53,
      "learning_rate": 8.329320388349515e-06,
      "loss": 0.084,
      "step": 225390
    },
    {
      "epoch": 87.53,
      "learning_rate": 8.328802588996765e-06,
      "loss": 0.1629,
      "step": 225400
    },
    {
      "epoch": 87.54,
      "learning_rate": 8.328284789644013e-06,
      "loss": 0.0147,
      "step": 225410
    },
    {
      "epoch": 87.54,
      "learning_rate": 8.327766990291262e-06,
      "loss": 0.0897,
      "step": 225420
    },
    {
      "epoch": 87.55,
      "learning_rate": 8.327249190938512e-06,
      "loss": 0.0568,
      "step": 225430
    },
    {
      "epoch": 87.55,
      "learning_rate": 8.326731391585762e-06,
      "loss": 0.0327,
      "step": 225440
    },
    {
      "epoch": 87.55,
      "learning_rate": 8.32621359223301e-06,
      "loss": 0.034,
      "step": 225450
    },
    {
      "epoch": 87.56,
      "learning_rate": 8.32569579288026e-06,
      "loss": 0.0574,
      "step": 225460
    },
    {
      "epoch": 87.56,
      "learning_rate": 8.325177993527509e-06,
      "loss": 0.0511,
      "step": 225470
    },
    {
      "epoch": 87.57,
      "learning_rate": 8.324660194174758e-06,
      "loss": 0.1085,
      "step": 225480
    },
    {
      "epoch": 87.57,
      "learning_rate": 8.324142394822006e-06,
      "loss": 0.0612,
      "step": 225490
    },
    {
      "epoch": 87.57,
      "learning_rate": 8.323624595469256e-06,
      "loss": 0.0952,
      "step": 225500
    },
    {
      "epoch": 87.58,
      "learning_rate": 8.323106796116506e-06,
      "loss": 0.0757,
      "step": 225510
    },
    {
      "epoch": 87.58,
      "learning_rate": 8.322588996763755e-06,
      "loss": 0.0769,
      "step": 225520
    },
    {
      "epoch": 87.58,
      "learning_rate": 8.322071197411003e-06,
      "loss": 0.0817,
      "step": 225530
    },
    {
      "epoch": 87.59,
      "learning_rate": 8.321553398058253e-06,
      "loss": 0.0426,
      "step": 225540
    },
    {
      "epoch": 87.59,
      "learning_rate": 8.321035598705502e-06,
      "loss": 0.0504,
      "step": 225550
    },
    {
      "epoch": 87.6,
      "learning_rate": 8.320517799352752e-06,
      "loss": 0.144,
      "step": 225560
    },
    {
      "epoch": 87.6,
      "learning_rate": 8.32e-06,
      "loss": 0.0893,
      "step": 225570
    },
    {
      "epoch": 87.6,
      "learning_rate": 8.31948220064725e-06,
      "loss": 0.0962,
      "step": 225580
    },
    {
      "epoch": 87.61,
      "learning_rate": 8.3189644012945e-06,
      "loss": 0.0004,
      "step": 225590
    },
    {
      "epoch": 87.61,
      "learning_rate": 8.318446601941749e-06,
      "loss": 0.017,
      "step": 225600
    },
    {
      "epoch": 87.62,
      "learning_rate": 8.317928802588997e-06,
      "loss": 0.0785,
      "step": 225610
    },
    {
      "epoch": 87.62,
      "learning_rate": 8.317411003236246e-06,
      "loss": 0.0021,
      "step": 225620
    },
    {
      "epoch": 87.62,
      "learning_rate": 8.316893203883496e-06,
      "loss": 0.0678,
      "step": 225630
    },
    {
      "epoch": 87.63,
      "learning_rate": 8.316375404530746e-06,
      "loss": 0.0246,
      "step": 225640
    },
    {
      "epoch": 87.63,
      "learning_rate": 8.315857605177994e-06,
      "loss": 0.0184,
      "step": 225650
    },
    {
      "epoch": 87.63,
      "learning_rate": 8.315339805825243e-06,
      "loss": 0.0039,
      "step": 225660
    },
    {
      "epoch": 87.64,
      "learning_rate": 8.314822006472493e-06,
      "loss": 0.0635,
      "step": 225670
    },
    {
      "epoch": 87.64,
      "learning_rate": 8.314304207119742e-06,
      "loss": 0.0339,
      "step": 225680
    },
    {
      "epoch": 87.65,
      "learning_rate": 8.31378640776699e-06,
      "loss": 0.0128,
      "step": 225690
    },
    {
      "epoch": 87.65,
      "learning_rate": 8.31326860841424e-06,
      "loss": 0.1059,
      "step": 225700
    },
    {
      "epoch": 87.65,
      "learning_rate": 8.31275080906149e-06,
      "loss": 0.0993,
      "step": 225710
    },
    {
      "epoch": 87.66,
      "learning_rate": 8.312233009708739e-06,
      "loss": 0.0884,
      "step": 225720
    },
    {
      "epoch": 87.66,
      "learning_rate": 8.311715210355987e-06,
      "loss": 0.0188,
      "step": 225730
    },
    {
      "epoch": 87.67,
      "learning_rate": 8.311197411003237e-06,
      "loss": 0.0034,
      "step": 225740
    },
    {
      "epoch": 87.67,
      "learning_rate": 8.310679611650486e-06,
      "loss": 0.1298,
      "step": 225750
    },
    {
      "epoch": 87.67,
      "learning_rate": 8.310161812297736e-06,
      "loss": 0.0149,
      "step": 225760
    },
    {
      "epoch": 87.68,
      "learning_rate": 8.309644012944984e-06,
      "loss": 0.0407,
      "step": 225770
    },
    {
      "epoch": 87.68,
      "learning_rate": 8.309126213592233e-06,
      "loss": 0.034,
      "step": 225780
    },
    {
      "epoch": 87.69,
      "learning_rate": 8.308608414239483e-06,
      "loss": 0.0374,
      "step": 225790
    },
    {
      "epoch": 87.69,
      "learning_rate": 8.308090614886733e-06,
      "loss": 0.0997,
      "step": 225800
    },
    {
      "epoch": 87.69,
      "learning_rate": 8.30757281553398e-06,
      "loss": 0.0132,
      "step": 225810
    },
    {
      "epoch": 87.7,
      "learning_rate": 8.30705501618123e-06,
      "loss": 0.0844,
      "step": 225820
    },
    {
      "epoch": 87.7,
      "learning_rate": 8.30653721682848e-06,
      "loss": 0.0579,
      "step": 225830
    },
    {
      "epoch": 87.7,
      "learning_rate": 8.30601941747573e-06,
      "loss": 0.0779,
      "step": 225840
    },
    {
      "epoch": 87.71,
      "learning_rate": 8.305501618122977e-06,
      "loss": 0.0445,
      "step": 225850
    },
    {
      "epoch": 87.71,
      "learning_rate": 8.304983818770227e-06,
      "loss": 0.2079,
      "step": 225860
    },
    {
      "epoch": 87.72,
      "learning_rate": 8.304466019417477e-06,
      "loss": 0.0104,
      "step": 225870
    },
    {
      "epoch": 87.72,
      "learning_rate": 8.303948220064726e-06,
      "loss": 0.0471,
      "step": 225880
    },
    {
      "epoch": 87.72,
      "learning_rate": 8.303430420711974e-06,
      "loss": 0.0148,
      "step": 225890
    },
    {
      "epoch": 87.73,
      "learning_rate": 8.302912621359224e-06,
      "loss": 0.0006,
      "step": 225900
    },
    {
      "epoch": 87.73,
      "learning_rate": 8.302394822006473e-06,
      "loss": 0.0014,
      "step": 225910
    },
    {
      "epoch": 87.74,
      "learning_rate": 8.301877022653723e-06,
      "loss": 0.0181,
      "step": 225920
    },
    {
      "epoch": 87.74,
      "learning_rate": 8.301359223300973e-06,
      "loss": 0.0363,
      "step": 225930
    },
    {
      "epoch": 87.74,
      "learning_rate": 8.30084142394822e-06,
      "loss": 0.0397,
      "step": 225940
    },
    {
      "epoch": 87.75,
      "learning_rate": 8.30032362459547e-06,
      "loss": 0.0868,
      "step": 225950
    },
    {
      "epoch": 87.75,
      "learning_rate": 8.29980582524272e-06,
      "loss": 0.0093,
      "step": 225960
    },
    {
      "epoch": 87.76,
      "learning_rate": 8.29928802588997e-06,
      "loss": 0.0196,
      "step": 225970
    },
    {
      "epoch": 87.76,
      "learning_rate": 8.298770226537217e-06,
      "loss": 0.0492,
      "step": 225980
    },
    {
      "epoch": 87.76,
      "learning_rate": 8.298252427184467e-06,
      "loss": 0.015,
      "step": 225990
    },
    {
      "epoch": 87.77,
      "learning_rate": 8.297734627831717e-06,
      "loss": 0.0349,
      "step": 226000
    },
    {
      "epoch": 87.77,
      "learning_rate": 8.297216828478966e-06,
      "loss": 0.0269,
      "step": 226010
    },
    {
      "epoch": 87.77,
      "learning_rate": 8.296699029126214e-06,
      "loss": 0.0849,
      "step": 226020
    },
    {
      "epoch": 87.78,
      "learning_rate": 8.296181229773464e-06,
      "loss": 0.044,
      "step": 226030
    },
    {
      "epoch": 87.78,
      "learning_rate": 8.295663430420713e-06,
      "loss": 0.0869,
      "step": 226040
    },
    {
      "epoch": 87.79,
      "learning_rate": 8.295145631067963e-06,
      "loss": 0.0504,
      "step": 226050
    },
    {
      "epoch": 87.79,
      "learning_rate": 8.294627831715211e-06,
      "loss": 0.1417,
      "step": 226060
    },
    {
      "epoch": 87.79,
      "learning_rate": 8.29411003236246e-06,
      "loss": 0.0126,
      "step": 226070
    },
    {
      "epoch": 87.8,
      "learning_rate": 8.29359223300971e-06,
      "loss": 0.0033,
      "step": 226080
    },
    {
      "epoch": 87.8,
      "learning_rate": 8.29307443365696e-06,
      "loss": 0.1094,
      "step": 226090
    },
    {
      "epoch": 87.81,
      "learning_rate": 8.292556634304208e-06,
      "loss": 0.0509,
      "step": 226100
    },
    {
      "epoch": 87.81,
      "learning_rate": 8.292038834951457e-06,
      "loss": 0.1086,
      "step": 226110
    },
    {
      "epoch": 87.81,
      "learning_rate": 8.291521035598707e-06,
      "loss": 0.0032,
      "step": 226120
    },
    {
      "epoch": 87.82,
      "learning_rate": 8.291003236245957e-06,
      "loss": 0.0006,
      "step": 226130
    },
    {
      "epoch": 87.82,
      "learning_rate": 8.290485436893204e-06,
      "loss": 0.0532,
      "step": 226140
    },
    {
      "epoch": 87.83,
      "learning_rate": 8.289967637540454e-06,
      "loss": 0.058,
      "step": 226150
    },
    {
      "epoch": 87.83,
      "learning_rate": 8.289449838187704e-06,
      "loss": 0.06,
      "step": 226160
    },
    {
      "epoch": 87.83,
      "learning_rate": 8.288932038834953e-06,
      "loss": 0.087,
      "step": 226170
    },
    {
      "epoch": 87.84,
      "learning_rate": 8.288414239482201e-06,
      "loss": 0.09,
      "step": 226180
    },
    {
      "epoch": 87.84,
      "learning_rate": 8.287896440129451e-06,
      "loss": 0.0272,
      "step": 226190
    },
    {
      "epoch": 87.84,
      "learning_rate": 8.2873786407767e-06,
      "loss": 0.0455,
      "step": 226200
    },
    {
      "epoch": 87.85,
      "learning_rate": 8.28686084142395e-06,
      "loss": 0.0259,
      "step": 226210
    },
    {
      "epoch": 87.85,
      "learning_rate": 8.286343042071198e-06,
      "loss": 0.0214,
      "step": 226220
    },
    {
      "epoch": 87.86,
      "learning_rate": 8.285825242718448e-06,
      "loss": 0.0333,
      "step": 226230
    },
    {
      "epoch": 87.86,
      "learning_rate": 8.285307443365697e-06,
      "loss": 0.0189,
      "step": 226240
    },
    {
      "epoch": 87.86,
      "learning_rate": 8.284789644012947e-06,
      "loss": 0.1101,
      "step": 226250
    },
    {
      "epoch": 87.87,
      "learning_rate": 8.284271844660195e-06,
      "loss": 0.0059,
      "step": 226260
    },
    {
      "epoch": 87.87,
      "learning_rate": 8.283754045307444e-06,
      "loss": 0.0162,
      "step": 226270
    },
    {
      "epoch": 87.88,
      "learning_rate": 8.283236245954694e-06,
      "loss": 0.2011,
      "step": 226280
    },
    {
      "epoch": 87.88,
      "learning_rate": 8.282718446601942e-06,
      "loss": 0.0198,
      "step": 226290
    },
    {
      "epoch": 87.88,
      "learning_rate": 8.282200647249192e-06,
      "loss": 0.0154,
      "step": 226300
    },
    {
      "epoch": 87.89,
      "learning_rate": 8.281682847896441e-06,
      "loss": 0.0074,
      "step": 226310
    },
    {
      "epoch": 87.89,
      "learning_rate": 8.28116504854369e-06,
      "loss": 0.3087,
      "step": 226320
    },
    {
      "epoch": 87.9,
      "learning_rate": 8.280647249190939e-06,
      "loss": 0.0541,
      "step": 226330
    },
    {
      "epoch": 87.9,
      "learning_rate": 8.280129449838188e-06,
      "loss": 0.0392,
      "step": 226340
    },
    {
      "epoch": 87.9,
      "learning_rate": 8.279611650485438e-06,
      "loss": 0.0144,
      "step": 226350
    },
    {
      "epoch": 87.91,
      "learning_rate": 8.279093851132688e-06,
      "loss": 0.0531,
      "step": 226360
    },
    {
      "epoch": 87.91,
      "learning_rate": 8.278576051779936e-06,
      "loss": 0.0218,
      "step": 226370
    },
    {
      "epoch": 87.91,
      "learning_rate": 8.278058252427185e-06,
      "loss": 0.0174,
      "step": 226380
    },
    {
      "epoch": 87.92,
      "learning_rate": 8.277540453074435e-06,
      "loss": 0.1345,
      "step": 226390
    },
    {
      "epoch": 87.92,
      "learning_rate": 8.277022653721684e-06,
      "loss": 0.0504,
      "step": 226400
    },
    {
      "epoch": 87.93,
      "learning_rate": 8.276504854368932e-06,
      "loss": 0.093,
      "step": 226410
    },
    {
      "epoch": 87.93,
      "learning_rate": 8.275987055016182e-06,
      "loss": 0.0386,
      "step": 226420
    },
    {
      "epoch": 87.93,
      "learning_rate": 8.275469255663432e-06,
      "loss": 0.0815,
      "step": 226430
    },
    {
      "epoch": 87.94,
      "learning_rate": 8.27495145631068e-06,
      "loss": 0.1022,
      "step": 226440
    },
    {
      "epoch": 87.94,
      "learning_rate": 8.274433656957929e-06,
      "loss": 0.1055,
      "step": 226450
    },
    {
      "epoch": 87.95,
      "learning_rate": 8.273915857605179e-06,
      "loss": 0.0004,
      "step": 226460
    },
    {
      "epoch": 87.95,
      "learning_rate": 8.273398058252428e-06,
      "loss": 0.1114,
      "step": 226470
    },
    {
      "epoch": 87.95,
      "learning_rate": 8.272880258899676e-06,
      "loss": 0.0086,
      "step": 226480
    },
    {
      "epoch": 87.96,
      "learning_rate": 8.272362459546926e-06,
      "loss": 0.0493,
      "step": 226490
    },
    {
      "epoch": 87.96,
      "learning_rate": 8.271844660194175e-06,
      "loss": 0.0995,
      "step": 226500
    },
    {
      "epoch": 87.97,
      "learning_rate": 8.271326860841425e-06,
      "loss": 0.0006,
      "step": 226510
    },
    {
      "epoch": 87.97,
      "learning_rate": 8.270809061488673e-06,
      "loss": 0.0074,
      "step": 226520
    },
    {
      "epoch": 87.97,
      "learning_rate": 8.270291262135923e-06,
      "loss": 0.0104,
      "step": 226530
    },
    {
      "epoch": 87.98,
      "learning_rate": 8.269773462783172e-06,
      "loss": 0.1458,
      "step": 226540
    },
    {
      "epoch": 87.98,
      "learning_rate": 8.269255663430422e-06,
      "loss": 0.0094,
      "step": 226550
    },
    {
      "epoch": 87.98,
      "learning_rate": 8.26873786407767e-06,
      "loss": 0.035,
      "step": 226560
    },
    {
      "epoch": 87.99,
      "learning_rate": 8.26822006472492e-06,
      "loss": 0.0944,
      "step": 226570
    },
    {
      "epoch": 87.99,
      "learning_rate": 8.267702265372169e-06,
      "loss": 0.0648,
      "step": 226580
    },
    {
      "epoch": 88.0,
      "learning_rate": 8.267184466019419e-06,
      "loss": 0.1342,
      "step": 226590
    },
    {
      "epoch": 88.0,
      "learning_rate": 8.266666666666667e-06,
      "loss": 0.1188,
      "step": 226600
    },
    {
      "epoch": 88.0,
      "eval_accuracy": 0.9485557083906465,
      "eval_loss": 0.37250131368637085,
      "eval_runtime": 8.1938,
      "eval_samples_per_second": 443.626,
      "eval_steps_per_second": 55.529,
      "step": 226600
    },
    {
      "epoch": 88.0,
      "learning_rate": 8.266148867313916e-06,
      "loss": 0.0775,
      "step": 226610
    },
    {
      "epoch": 88.01,
      "learning_rate": 8.265631067961166e-06,
      "loss": 0.024,
      "step": 226620
    },
    {
      "epoch": 88.01,
      "learning_rate": 8.265113268608414e-06,
      "loss": 0.0093,
      "step": 226630
    },
    {
      "epoch": 88.02,
      "learning_rate": 8.264595469255663e-06,
      "loss": 0.0931,
      "step": 226640
    },
    {
      "epoch": 88.02,
      "learning_rate": 8.264077669902913e-06,
      "loss": 0.1535,
      "step": 226650
    },
    {
      "epoch": 88.02,
      "learning_rate": 8.263559870550163e-06,
      "loss": 0.0054,
      "step": 226660
    },
    {
      "epoch": 88.03,
      "learning_rate": 8.26304207119741e-06,
      "loss": 0.1043,
      "step": 226670
    },
    {
      "epoch": 88.03,
      "learning_rate": 8.26252427184466e-06,
      "loss": 0.0002,
      "step": 226680
    },
    {
      "epoch": 88.03,
      "learning_rate": 8.26200647249191e-06,
      "loss": 0.0286,
      "step": 226690
    },
    {
      "epoch": 88.04,
      "learning_rate": 8.26148867313916e-06,
      "loss": 0.0088,
      "step": 226700
    },
    {
      "epoch": 88.04,
      "learning_rate": 8.260970873786407e-06,
      "loss": 0.05,
      "step": 226710
    },
    {
      "epoch": 88.05,
      "learning_rate": 8.260453074433657e-06,
      "loss": 0.057,
      "step": 226720
    },
    {
      "epoch": 88.05,
      "learning_rate": 8.259935275080907e-06,
      "loss": 0.045,
      "step": 226730
    },
    {
      "epoch": 88.05,
      "learning_rate": 8.259417475728156e-06,
      "loss": 0.001,
      "step": 226740
    },
    {
      "epoch": 88.06,
      "learning_rate": 8.258899676375404e-06,
      "loss": 0.0076,
      "step": 226750
    },
    {
      "epoch": 88.06,
      "learning_rate": 8.258381877022654e-06,
      "loss": 0.0288,
      "step": 226760
    },
    {
      "epoch": 88.07,
      "learning_rate": 8.257864077669903e-06,
      "loss": 0.0431,
      "step": 226770
    },
    {
      "epoch": 88.07,
      "learning_rate": 8.257346278317153e-06,
      "loss": 0.0211,
      "step": 226780
    },
    {
      "epoch": 88.07,
      "learning_rate": 8.256828478964401e-06,
      "loss": 0.0313,
      "step": 226790
    },
    {
      "epoch": 88.08,
      "learning_rate": 8.25631067961165e-06,
      "loss": 0.0515,
      "step": 226800
    },
    {
      "epoch": 88.08,
      "learning_rate": 8.2557928802589e-06,
      "loss": 0.0916,
      "step": 226810
    },
    {
      "epoch": 88.09,
      "learning_rate": 8.25527508090615e-06,
      "loss": 0.1283,
      "step": 226820
    },
    {
      "epoch": 88.09,
      "learning_rate": 8.254757281553398e-06,
      "loss": 0.1278,
      "step": 226830
    },
    {
      "epoch": 88.09,
      "learning_rate": 8.254239482200647e-06,
      "loss": 0.0133,
      "step": 226840
    },
    {
      "epoch": 88.1,
      "learning_rate": 8.253721682847897e-06,
      "loss": 0.1005,
      "step": 226850
    },
    {
      "epoch": 88.1,
      "learning_rate": 8.253203883495146e-06,
      "loss": 0.0509,
      "step": 226860
    },
    {
      "epoch": 88.1,
      "learning_rate": 8.252686084142394e-06,
      "loss": 0.1255,
      "step": 226870
    },
    {
      "epoch": 88.11,
      "learning_rate": 8.252168284789644e-06,
      "loss": 0.001,
      "step": 226880
    },
    {
      "epoch": 88.11,
      "learning_rate": 8.251650485436894e-06,
      "loss": 0.1471,
      "step": 226890
    },
    {
      "epoch": 88.12,
      "learning_rate": 8.251132686084143e-06,
      "loss": 0.002,
      "step": 226900
    },
    {
      "epoch": 88.12,
      "learning_rate": 8.250614886731391e-06,
      "loss": 0.1261,
      "step": 226910
    },
    {
      "epoch": 88.12,
      "learning_rate": 8.25009708737864e-06,
      "loss": 0.1305,
      "step": 226920
    },
    {
      "epoch": 88.13,
      "learning_rate": 8.24957928802589e-06,
      "loss": 0.1085,
      "step": 226930
    },
    {
      "epoch": 88.13,
      "learning_rate": 8.24906148867314e-06,
      "loss": 0.0133,
      "step": 226940
    },
    {
      "epoch": 88.14,
      "learning_rate": 8.248543689320388e-06,
      "loss": 0.0481,
      "step": 226950
    },
    {
      "epoch": 88.14,
      "learning_rate": 8.248025889967638e-06,
      "loss": 0.0506,
      "step": 226960
    },
    {
      "epoch": 88.14,
      "learning_rate": 8.247508090614887e-06,
      "loss": 0.0914,
      "step": 226970
    },
    {
      "epoch": 88.15,
      "learning_rate": 8.246990291262137e-06,
      "loss": 0.0583,
      "step": 226980
    },
    {
      "epoch": 88.15,
      "learning_rate": 8.246472491909385e-06,
      "loss": 0.0495,
      "step": 226990
    },
    {
      "epoch": 88.16,
      "learning_rate": 8.245954692556634e-06,
      "loss": 0.1721,
      "step": 227000
    },
    {
      "epoch": 88.16,
      "learning_rate": 8.245436893203884e-06,
      "loss": 0.1916,
      "step": 227010
    },
    {
      "epoch": 88.16,
      "learning_rate": 8.244919093851134e-06,
      "loss": 0.0168,
      "step": 227020
    },
    {
      "epoch": 88.17,
      "learning_rate": 8.244401294498382e-06,
      "loss": 0.0234,
      "step": 227030
    },
    {
      "epoch": 88.17,
      "learning_rate": 8.243883495145631e-06,
      "loss": 0.0539,
      "step": 227040
    },
    {
      "epoch": 88.17,
      "learning_rate": 8.24336569579288e-06,
      "loss": 0.0056,
      "step": 227050
    },
    {
      "epoch": 88.18,
      "learning_rate": 8.24284789644013e-06,
      "loss": 0.0752,
      "step": 227060
    },
    {
      "epoch": 88.18,
      "learning_rate": 8.24233009708738e-06,
      "loss": 0.0163,
      "step": 227070
    },
    {
      "epoch": 88.19,
      "learning_rate": 8.241812297734628e-06,
      "loss": 0.0029,
      "step": 227080
    },
    {
      "epoch": 88.19,
      "learning_rate": 8.241294498381878e-06,
      "loss": 0.1332,
      "step": 227090
    },
    {
      "epoch": 88.19,
      "learning_rate": 8.240776699029127e-06,
      "loss": 0.0012,
      "step": 227100
    },
    {
      "epoch": 88.2,
      "learning_rate": 8.240258899676377e-06,
      "loss": 0.0388,
      "step": 227110
    },
    {
      "epoch": 88.2,
      "learning_rate": 8.239741100323625e-06,
      "loss": 0.109,
      "step": 227120
    },
    {
      "epoch": 88.21,
      "learning_rate": 8.239223300970874e-06,
      "loss": 0.0824,
      "step": 227130
    },
    {
      "epoch": 88.21,
      "learning_rate": 8.238705501618124e-06,
      "loss": 0.0497,
      "step": 227140
    },
    {
      "epoch": 88.21,
      "learning_rate": 8.238187702265374e-06,
      "loss": 0.0262,
      "step": 227150
    },
    {
      "epoch": 88.22,
      "learning_rate": 8.237669902912621e-06,
      "loss": 0.052,
      "step": 227160
    },
    {
      "epoch": 88.22,
      "learning_rate": 8.237152103559871e-06,
      "loss": 0.0172,
      "step": 227170
    },
    {
      "epoch": 88.23,
      "learning_rate": 8.23663430420712e-06,
      "loss": 0.0013,
      "step": 227180
    },
    {
      "epoch": 88.23,
      "learning_rate": 8.23611650485437e-06,
      "loss": 0.0703,
      "step": 227190
    },
    {
      "epoch": 88.23,
      "learning_rate": 8.235598705501618e-06,
      "loss": 0.0171,
      "step": 227200
    },
    {
      "epoch": 88.24,
      "learning_rate": 8.235080906148868e-06,
      "loss": 0.0173,
      "step": 227210
    },
    {
      "epoch": 88.24,
      "learning_rate": 8.234563106796117e-06,
      "loss": 0.238,
      "step": 227220
    },
    {
      "epoch": 88.24,
      "learning_rate": 8.234045307443367e-06,
      "loss": 0.0215,
      "step": 227230
    },
    {
      "epoch": 88.25,
      "learning_rate": 8.233527508090615e-06,
      "loss": 0.0236,
      "step": 227240
    },
    {
      "epoch": 88.25,
      "learning_rate": 8.233009708737865e-06,
      "loss": 0.0332,
      "step": 227250
    },
    {
      "epoch": 88.26,
      "learning_rate": 8.232491909385114e-06,
      "loss": 0.1584,
      "step": 227260
    },
    {
      "epoch": 88.26,
      "learning_rate": 8.231974110032364e-06,
      "loss": 0.0264,
      "step": 227270
    },
    {
      "epoch": 88.26,
      "learning_rate": 8.231456310679612e-06,
      "loss": 0.04,
      "step": 227280
    },
    {
      "epoch": 88.27,
      "learning_rate": 8.230938511326861e-06,
      "loss": 0.1187,
      "step": 227290
    },
    {
      "epoch": 88.27,
      "learning_rate": 8.230420711974111e-06,
      "loss": 0.0107,
      "step": 227300
    },
    {
      "epoch": 88.28,
      "learning_rate": 8.22990291262136e-06,
      "loss": 0.061,
      "step": 227310
    },
    {
      "epoch": 88.28,
      "learning_rate": 8.229385113268609e-06,
      "loss": 0.0618,
      "step": 227320
    },
    {
      "epoch": 88.28,
      "learning_rate": 8.228867313915858e-06,
      "loss": 0.0749,
      "step": 227330
    },
    {
      "epoch": 88.29,
      "learning_rate": 8.228349514563108e-06,
      "loss": 0.0315,
      "step": 227340
    },
    {
      "epoch": 88.29,
      "learning_rate": 8.227831715210357e-06,
      "loss": 0.0405,
      "step": 227350
    },
    {
      "epoch": 88.3,
      "learning_rate": 8.227313915857605e-06,
      "loss": 0.0334,
      "step": 227360
    },
    {
      "epoch": 88.3,
      "learning_rate": 8.226796116504855e-06,
      "loss": 0.0011,
      "step": 227370
    },
    {
      "epoch": 88.3,
      "learning_rate": 8.226278317152105e-06,
      "loss": 0.0256,
      "step": 227380
    },
    {
      "epoch": 88.31,
      "learning_rate": 8.225760517799354e-06,
      "loss": 0.0459,
      "step": 227390
    },
    {
      "epoch": 88.31,
      "learning_rate": 8.225242718446602e-06,
      "loss": 0.067,
      "step": 227400
    },
    {
      "epoch": 88.31,
      "learning_rate": 8.224724919093852e-06,
      "loss": 0.0345,
      "step": 227410
    },
    {
      "epoch": 88.32,
      "learning_rate": 8.224207119741101e-06,
      "loss": 0.0107,
      "step": 227420
    },
    {
      "epoch": 88.32,
      "learning_rate": 8.223689320388351e-06,
      "loss": 0.1011,
      "step": 227430
    },
    {
      "epoch": 88.33,
      "learning_rate": 8.223171521035599e-06,
      "loss": 0.027,
      "step": 227440
    },
    {
      "epoch": 88.33,
      "learning_rate": 8.222653721682849e-06,
      "loss": 0.0809,
      "step": 227450
    },
    {
      "epoch": 88.33,
      "learning_rate": 8.222135922330098e-06,
      "loss": 0.0047,
      "step": 227460
    },
    {
      "epoch": 88.34,
      "learning_rate": 8.221618122977348e-06,
      "loss": 0.0671,
      "step": 227470
    },
    {
      "epoch": 88.34,
      "learning_rate": 8.221100323624596e-06,
      "loss": 0.0743,
      "step": 227480
    },
    {
      "epoch": 88.35,
      "learning_rate": 8.220582524271845e-06,
      "loss": 0.0168,
      "step": 227490
    },
    {
      "epoch": 88.35,
      "learning_rate": 8.220064724919095e-06,
      "loss": 0.0421,
      "step": 227500
    },
    {
      "epoch": 88.35,
      "learning_rate": 8.219546925566345e-06,
      "loss": 0.0094,
      "step": 227510
    },
    {
      "epoch": 88.36,
      "learning_rate": 8.219029126213592e-06,
      "loss": 0.0114,
      "step": 227520
    },
    {
      "epoch": 88.36,
      "learning_rate": 8.218511326860842e-06,
      "loss": 0.0833,
      "step": 227530
    },
    {
      "epoch": 88.37,
      "learning_rate": 8.217993527508092e-06,
      "loss": 0.0115,
      "step": 227540
    },
    {
      "epoch": 88.37,
      "learning_rate": 8.217475728155341e-06,
      "loss": 0.0239,
      "step": 227550
    },
    {
      "epoch": 88.37,
      "learning_rate": 8.21695792880259e-06,
      "loss": 0.0644,
      "step": 227560
    },
    {
      "epoch": 88.38,
      "learning_rate": 8.216440129449839e-06,
      "loss": 0.0289,
      "step": 227570
    },
    {
      "epoch": 88.38,
      "learning_rate": 8.215922330097088e-06,
      "loss": 0.0001,
      "step": 227580
    },
    {
      "epoch": 88.38,
      "learning_rate": 8.215404530744338e-06,
      "loss": 0.088,
      "step": 227590
    },
    {
      "epoch": 88.39,
      "learning_rate": 8.214886731391586e-06,
      "loss": 0.0602,
      "step": 227600
    },
    {
      "epoch": 88.39,
      "learning_rate": 8.214368932038836e-06,
      "loss": 0.0155,
      "step": 227610
    },
    {
      "epoch": 88.4,
      "learning_rate": 8.213851132686085e-06,
      "loss": 0.1322,
      "step": 227620
    },
    {
      "epoch": 88.4,
      "learning_rate": 8.213333333333335e-06,
      "loss": 0.0656,
      "step": 227630
    },
    {
      "epoch": 88.4,
      "learning_rate": 8.212815533980584e-06,
      "loss": 0.0035,
      "step": 227640
    },
    {
      "epoch": 88.41,
      "learning_rate": 8.212297734627832e-06,
      "loss": 0.0682,
      "step": 227650
    },
    {
      "epoch": 88.41,
      "learning_rate": 8.211779935275082e-06,
      "loss": 0.0331,
      "step": 227660
    },
    {
      "epoch": 88.42,
      "learning_rate": 8.211262135922332e-06,
      "loss": 0.0009,
      "step": 227670
    },
    {
      "epoch": 88.42,
      "learning_rate": 8.210744336569581e-06,
      "loss": 0.0396,
      "step": 227680
    },
    {
      "epoch": 88.42,
      "learning_rate": 8.21022653721683e-06,
      "loss": 0.0729,
      "step": 227690
    },
    {
      "epoch": 88.43,
      "learning_rate": 8.209708737864079e-06,
      "loss": 0.2013,
      "step": 227700
    },
    {
      "epoch": 88.43,
      "learning_rate": 8.209190938511328e-06,
      "loss": 0.0376,
      "step": 227710
    },
    {
      "epoch": 88.43,
      "learning_rate": 8.208673139158578e-06,
      "loss": 0.1394,
      "step": 227720
    },
    {
      "epoch": 88.44,
      "learning_rate": 8.208155339805826e-06,
      "loss": 0.1549,
      "step": 227730
    },
    {
      "epoch": 88.44,
      "learning_rate": 8.207637540453076e-06,
      "loss": 0.0201,
      "step": 227740
    },
    {
      "epoch": 88.45,
      "learning_rate": 8.207119741100325e-06,
      "loss": 0.1594,
      "step": 227750
    },
    {
      "epoch": 88.45,
      "learning_rate": 8.206601941747573e-06,
      "loss": 0.0761,
      "step": 227760
    },
    {
      "epoch": 88.45,
      "learning_rate": 8.206084142394823e-06,
      "loss": 0.0014,
      "step": 227770
    },
    {
      "epoch": 88.46,
      "learning_rate": 8.205566343042072e-06,
      "loss": 0.0688,
      "step": 227780
    },
    {
      "epoch": 88.46,
      "learning_rate": 8.205048543689322e-06,
      "loss": 0.0007,
      "step": 227790
    },
    {
      "epoch": 88.47,
      "learning_rate": 8.20453074433657e-06,
      "loss": 0.0562,
      "step": 227800
    },
    {
      "epoch": 88.47,
      "learning_rate": 8.20401294498382e-06,
      "loss": 0.0522,
      "step": 227810
    },
    {
      "epoch": 88.47,
      "learning_rate": 8.20349514563107e-06,
      "loss": 0.0765,
      "step": 227820
    },
    {
      "epoch": 88.48,
      "learning_rate": 8.202977346278319e-06,
      "loss": 0.0378,
      "step": 227830
    },
    {
      "epoch": 88.48,
      "learning_rate": 8.202459546925567e-06,
      "loss": 0.0191,
      "step": 227840
    },
    {
      "epoch": 88.49,
      "learning_rate": 8.201941747572816e-06,
      "loss": 0.0022,
      "step": 227850
    },
    {
      "epoch": 88.49,
      "learning_rate": 8.201423948220066e-06,
      "loss": 0.0521,
      "step": 227860
    },
    {
      "epoch": 88.49,
      "learning_rate": 8.200906148867316e-06,
      "loss": 0.0921,
      "step": 227870
    },
    {
      "epoch": 88.5,
      "learning_rate": 8.200388349514563e-06,
      "loss": 0.0787,
      "step": 227880
    },
    {
      "epoch": 88.5,
      "learning_rate": 8.199870550161813e-06,
      "loss": 0.0442,
      "step": 227890
    },
    {
      "epoch": 88.5,
      "learning_rate": 8.199352750809063e-06,
      "loss": 0.0375,
      "step": 227900
    },
    {
      "epoch": 88.51,
      "learning_rate": 8.19883495145631e-06,
      "loss": 0.0529,
      "step": 227910
    },
    {
      "epoch": 88.51,
      "learning_rate": 8.19831715210356e-06,
      "loss": 0.0384,
      "step": 227920
    },
    {
      "epoch": 88.52,
      "learning_rate": 8.19779935275081e-06,
      "loss": 0.0466,
      "step": 227930
    },
    {
      "epoch": 88.52,
      "learning_rate": 8.19728155339806e-06,
      "loss": 0.0002,
      "step": 227940
    },
    {
      "epoch": 88.52,
      "learning_rate": 8.196763754045307e-06,
      "loss": 0.0391,
      "step": 227950
    },
    {
      "epoch": 88.53,
      "learning_rate": 8.196245954692557e-06,
      "loss": 0.048,
      "step": 227960
    },
    {
      "epoch": 88.53,
      "learning_rate": 8.195728155339807e-06,
      "loss": 0.2114,
      "step": 227970
    },
    {
      "epoch": 88.54,
      "learning_rate": 8.195210355987056e-06,
      "loss": 0.0548,
      "step": 227980
    },
    {
      "epoch": 88.54,
      "learning_rate": 8.194692556634304e-06,
      "loss": 0.2075,
      "step": 227990
    },
    {
      "epoch": 88.54,
      "learning_rate": 8.194174757281554e-06,
      "loss": 0.0908,
      "step": 228000
    },
    {
      "epoch": 88.55,
      "learning_rate": 8.193656957928803e-06,
      "loss": 0.0008,
      "step": 228010
    },
    {
      "epoch": 88.55,
      "learning_rate": 8.193139158576053e-06,
      "loss": 0.0365,
      "step": 228020
    },
    {
      "epoch": 88.56,
      "learning_rate": 8.192621359223301e-06,
      "loss": 0.0332,
      "step": 228030
    },
    {
      "epoch": 88.56,
      "learning_rate": 8.19210355987055e-06,
      "loss": 0.0261,
      "step": 228040
    },
    {
      "epoch": 88.56,
      "learning_rate": 8.1915857605178e-06,
      "loss": 0.0102,
      "step": 228050
    },
    {
      "epoch": 88.57,
      "learning_rate": 8.19106796116505e-06,
      "loss": 0.0125,
      "step": 228060
    },
    {
      "epoch": 88.57,
      "learning_rate": 8.190550161812298e-06,
      "loss": 0.1339,
      "step": 228070
    },
    {
      "epoch": 88.57,
      "learning_rate": 8.190032362459547e-06,
      "loss": 0.0934,
      "step": 228080
    },
    {
      "epoch": 88.58,
      "learning_rate": 8.189514563106797e-06,
      "loss": 0.1305,
      "step": 228090
    },
    {
      "epoch": 88.58,
      "learning_rate": 8.188996763754045e-06,
      "loss": 0.0626,
      "step": 228100
    },
    {
      "epoch": 88.59,
      "learning_rate": 8.188478964401295e-06,
      "loss": 0.0111,
      "step": 228110
    },
    {
      "epoch": 88.59,
      "learning_rate": 8.187961165048544e-06,
      "loss": 0.0005,
      "step": 228120
    },
    {
      "epoch": 88.59,
      "learning_rate": 8.187443365695794e-06,
      "loss": 0.0001,
      "step": 228130
    },
    {
      "epoch": 88.6,
      "learning_rate": 8.186925566343042e-06,
      "loss": 0.0028,
      "step": 228140
    },
    {
      "epoch": 88.6,
      "learning_rate": 8.186407766990291e-06,
      "loss": 0.0215,
      "step": 228150
    },
    {
      "epoch": 88.61,
      "learning_rate": 8.185889967637541e-06,
      "loss": 0.0381,
      "step": 228160
    },
    {
      "epoch": 88.61,
      "learning_rate": 8.18537216828479e-06,
      "loss": 0.0151,
      "step": 228170
    },
    {
      "epoch": 88.61,
      "learning_rate": 8.184854368932038e-06,
      "loss": 0.1006,
      "step": 228180
    },
    {
      "epoch": 88.62,
      "learning_rate": 8.184336569579288e-06,
      "loss": 0.0393,
      "step": 228190
    },
    {
      "epoch": 88.62,
      "learning_rate": 8.183818770226538e-06,
      "loss": 0.0786,
      "step": 228200
    },
    {
      "epoch": 88.63,
      "learning_rate": 8.183300970873787e-06,
      "loss": 0.0358,
      "step": 228210
    },
    {
      "epoch": 88.63,
      "learning_rate": 8.182783171521035e-06,
      "loss": 0.0474,
      "step": 228220
    },
    {
      "epoch": 88.63,
      "learning_rate": 8.182265372168285e-06,
      "loss": 0.1346,
      "step": 228230
    },
    {
      "epoch": 88.64,
      "learning_rate": 8.181747572815534e-06,
      "loss": 0.0075,
      "step": 228240
    },
    {
      "epoch": 88.64,
      "learning_rate": 8.181229773462784e-06,
      "loss": 0.237,
      "step": 228250
    },
    {
      "epoch": 88.64,
      "learning_rate": 8.180711974110032e-06,
      "loss": 0.1597,
      "step": 228260
    },
    {
      "epoch": 88.65,
      "learning_rate": 8.180194174757282e-06,
      "loss": 0.0276,
      "step": 228270
    },
    {
      "epoch": 88.65,
      "learning_rate": 8.179676375404531e-06,
      "loss": 0.1664,
      "step": 228280
    },
    {
      "epoch": 88.66,
      "learning_rate": 8.179158576051781e-06,
      "loss": 0.0483,
      "step": 228290
    },
    {
      "epoch": 88.66,
      "learning_rate": 8.178640776699029e-06,
      "loss": 0.0003,
      "step": 228300
    },
    {
      "epoch": 88.66,
      "learning_rate": 8.178122977346278e-06,
      "loss": 0.0716,
      "step": 228310
    },
    {
      "epoch": 88.67,
      "learning_rate": 8.177605177993528e-06,
      "loss": 0.0861,
      "step": 228320
    },
    {
      "epoch": 88.67,
      "learning_rate": 8.177087378640778e-06,
      "loss": 0.0034,
      "step": 228330
    },
    {
      "epoch": 88.68,
      "learning_rate": 8.176569579288026e-06,
      "loss": 0.0412,
      "step": 228340
    },
    {
      "epoch": 88.68,
      "learning_rate": 8.176051779935275e-06,
      "loss": 0.0092,
      "step": 228350
    },
    {
      "epoch": 88.68,
      "learning_rate": 8.175533980582525e-06,
      "loss": 0.0397,
      "step": 228360
    },
    {
      "epoch": 88.69,
      "learning_rate": 8.175016181229774e-06,
      "loss": 0.0817,
      "step": 228370
    },
    {
      "epoch": 88.69,
      "learning_rate": 8.174498381877022e-06,
      "loss": 0.0264,
      "step": 228380
    },
    {
      "epoch": 88.7,
      "learning_rate": 8.173980582524272e-06,
      "loss": 0.0001,
      "step": 228390
    },
    {
      "epoch": 88.7,
      "learning_rate": 8.173462783171522e-06,
      "loss": 0.0535,
      "step": 228400
    },
    {
      "epoch": 88.7,
      "learning_rate": 8.172944983818771e-06,
      "loss": 0.0345,
      "step": 228410
    },
    {
      "epoch": 88.71,
      "learning_rate": 8.172427184466019e-06,
      "loss": 0.0075,
      "step": 228420
    },
    {
      "epoch": 88.71,
      "learning_rate": 8.171909385113269e-06,
      "loss": 0.0015,
      "step": 228430
    },
    {
      "epoch": 88.71,
      "learning_rate": 8.171391585760518e-06,
      "loss": 0.0196,
      "step": 228440
    },
    {
      "epoch": 88.72,
      "learning_rate": 8.170873786407768e-06,
      "loss": 0.0039,
      "step": 228450
    },
    {
      "epoch": 88.72,
      "learning_rate": 8.170355987055016e-06,
      "loss": 0.0111,
      "step": 228460
    },
    {
      "epoch": 88.73,
      "learning_rate": 8.169838187702266e-06,
      "loss": 0.0016,
      "step": 228470
    },
    {
      "epoch": 88.73,
      "learning_rate": 8.169320388349515e-06,
      "loss": 0.0073,
      "step": 228480
    },
    {
      "epoch": 88.73,
      "learning_rate": 8.168802588996765e-06,
      "loss": 0.0006,
      "step": 228490
    },
    {
      "epoch": 88.74,
      "learning_rate": 8.168284789644013e-06,
      "loss": 0.0633,
      "step": 228500
    },
    {
      "epoch": 88.74,
      "learning_rate": 8.167766990291262e-06,
      "loss": 0.1453,
      "step": 228510
    },
    {
      "epoch": 88.75,
      "learning_rate": 8.167249190938512e-06,
      "loss": 0.0979,
      "step": 228520
    },
    {
      "epoch": 88.75,
      "learning_rate": 8.166731391585762e-06,
      "loss": 0.1108,
      "step": 228530
    },
    {
      "epoch": 88.75,
      "learning_rate": 8.16621359223301e-06,
      "loss": 0.0001,
      "step": 228540
    },
    {
      "epoch": 88.76,
      "learning_rate": 8.165695792880259e-06,
      "loss": 0.1057,
      "step": 228550
    },
    {
      "epoch": 88.76,
      "learning_rate": 8.165177993527509e-06,
      "loss": 0.1027,
      "step": 228560
    },
    {
      "epoch": 88.77,
      "learning_rate": 8.164660194174758e-06,
      "loss": 0.0493,
      "step": 228570
    },
    {
      "epoch": 88.77,
      "learning_rate": 8.164142394822006e-06,
      "loss": 0.0286,
      "step": 228580
    },
    {
      "epoch": 88.77,
      "learning_rate": 8.163624595469256e-06,
      "loss": 0.0203,
      "step": 228590
    },
    {
      "epoch": 88.78,
      "learning_rate": 8.163106796116505e-06,
      "loss": 0.0174,
      "step": 228600
    },
    {
      "epoch": 88.78,
      "learning_rate": 8.162588996763755e-06,
      "loss": 0.0773,
      "step": 228610
    },
    {
      "epoch": 88.78,
      "learning_rate": 8.162071197411003e-06,
      "loss": 0.0336,
      "step": 228620
    },
    {
      "epoch": 88.79,
      "learning_rate": 8.161553398058253e-06,
      "loss": 0.0458,
      "step": 228630
    },
    {
      "epoch": 88.79,
      "learning_rate": 8.161035598705502e-06,
      "loss": 0.0001,
      "step": 228640
    },
    {
      "epoch": 88.8,
      "learning_rate": 8.160517799352752e-06,
      "loss": 0.0387,
      "step": 228650
    },
    {
      "epoch": 88.8,
      "learning_rate": 8.16e-06,
      "loss": 0.0024,
      "step": 228660
    },
    {
      "epoch": 88.8,
      "learning_rate": 8.15948220064725e-06,
      "loss": 0.0002,
      "step": 228670
    },
    {
      "epoch": 88.81,
      "learning_rate": 8.158964401294499e-06,
      "loss": 0.0098,
      "step": 228680
    },
    {
      "epoch": 88.81,
      "learning_rate": 8.158446601941749e-06,
      "loss": 0.0088,
      "step": 228690
    },
    {
      "epoch": 88.82,
      "learning_rate": 8.157928802588997e-06,
      "loss": 0.061,
      "step": 228700
    },
    {
      "epoch": 88.82,
      "learning_rate": 8.157411003236246e-06,
      "loss": 0.0195,
      "step": 228710
    },
    {
      "epoch": 88.82,
      "learning_rate": 8.156893203883496e-06,
      "loss": 0.0336,
      "step": 228720
    },
    {
      "epoch": 88.83,
      "learning_rate": 8.156375404530745e-06,
      "loss": 0.0084,
      "step": 228730
    },
    {
      "epoch": 88.83,
      "learning_rate": 8.155857605177995e-06,
      "loss": 0.0708,
      "step": 228740
    },
    {
      "epoch": 88.83,
      "learning_rate": 8.155339805825243e-06,
      "loss": 0.0942,
      "step": 228750
    },
    {
      "epoch": 88.84,
      "learning_rate": 8.154822006472493e-06,
      "loss": 0.0375,
      "step": 228760
    },
    {
      "epoch": 88.84,
      "learning_rate": 8.154304207119742e-06,
      "loss": 0.0162,
      "step": 228770
    },
    {
      "epoch": 88.85,
      "learning_rate": 8.153786407766992e-06,
      "loss": 0.0287,
      "step": 228780
    },
    {
      "epoch": 88.85,
      "learning_rate": 8.15326860841424e-06,
      "loss": 0.0004,
      "step": 228790
    },
    {
      "epoch": 88.85,
      "learning_rate": 8.15275080906149e-06,
      "loss": 0.0266,
      "step": 228800
    },
    {
      "epoch": 88.86,
      "learning_rate": 8.152233009708739e-06,
      "loss": 0.0165,
      "step": 228810
    },
    {
      "epoch": 88.86,
      "learning_rate": 8.151715210355989e-06,
      "loss": 0.0549,
      "step": 228820
    },
    {
      "epoch": 88.87,
      "learning_rate": 8.151197411003237e-06,
      "loss": 0.0189,
      "step": 228830
    },
    {
      "epoch": 88.87,
      "learning_rate": 8.150679611650486e-06,
      "loss": 0.0002,
      "step": 228840
    },
    {
      "epoch": 88.87,
      "learning_rate": 8.150161812297736e-06,
      "loss": 0.0346,
      "step": 228850
    },
    {
      "epoch": 88.88,
      "learning_rate": 8.149644012944985e-06,
      "loss": 0.0075,
      "step": 228860
    },
    {
      "epoch": 88.88,
      "learning_rate": 8.149126213592233e-06,
      "loss": 0.0613,
      "step": 228870
    },
    {
      "epoch": 88.89,
      "learning_rate": 8.148608414239483e-06,
      "loss": 0.0158,
      "step": 228880
    },
    {
      "epoch": 88.89,
      "learning_rate": 8.148090614886733e-06,
      "loss": 0.0141,
      "step": 228890
    },
    {
      "epoch": 88.89,
      "learning_rate": 8.147572815533982e-06,
      "loss": 0.0405,
      "step": 228900
    },
    {
      "epoch": 88.9,
      "learning_rate": 8.14705501618123e-06,
      "loss": 0.0288,
      "step": 228910
    },
    {
      "epoch": 88.9,
      "learning_rate": 8.14653721682848e-06,
      "loss": 0.0731,
      "step": 228920
    },
    {
      "epoch": 88.9,
      "learning_rate": 8.14601941747573e-06,
      "loss": 0.0477,
      "step": 228930
    },
    {
      "epoch": 88.91,
      "learning_rate": 8.145501618122979e-06,
      "loss": 0.0499,
      "step": 228940
    },
    {
      "epoch": 88.91,
      "learning_rate": 8.144983818770227e-06,
      "loss": 0.0216,
      "step": 228950
    },
    {
      "epoch": 88.92,
      "learning_rate": 8.144466019417476e-06,
      "loss": 0.0215,
      "step": 228960
    },
    {
      "epoch": 88.92,
      "learning_rate": 8.143948220064726e-06,
      "loss": 0.0891,
      "step": 228970
    },
    {
      "epoch": 88.92,
      "learning_rate": 8.143430420711976e-06,
      "loss": 0.038,
      "step": 228980
    },
    {
      "epoch": 88.93,
      "learning_rate": 8.142912621359224e-06,
      "loss": 0.0089,
      "step": 228990
    },
    {
      "epoch": 88.93,
      "learning_rate": 8.142394822006473e-06,
      "loss": 0.0777,
      "step": 229000
    },
    {
      "epoch": 88.94,
      "learning_rate": 8.141877022653723e-06,
      "loss": 0.1021,
      "step": 229010
    },
    {
      "epoch": 88.94,
      "learning_rate": 8.141359223300972e-06,
      "loss": 0.0446,
      "step": 229020
    },
    {
      "epoch": 88.94,
      "learning_rate": 8.14084142394822e-06,
      "loss": 0.0077,
      "step": 229030
    },
    {
      "epoch": 88.95,
      "learning_rate": 8.14032362459547e-06,
      "loss": 0.0108,
      "step": 229040
    },
    {
      "epoch": 88.95,
      "learning_rate": 8.13980582524272e-06,
      "loss": 0.0446,
      "step": 229050
    },
    {
      "epoch": 88.96,
      "learning_rate": 8.13928802588997e-06,
      "loss": 0.2494,
      "step": 229060
    },
    {
      "epoch": 88.96,
      "learning_rate": 8.138770226537217e-06,
      "loss": 0.1199,
      "step": 229070
    },
    {
      "epoch": 88.96,
      "learning_rate": 8.138252427184467e-06,
      "loss": 0.0352,
      "step": 229080
    },
    {
      "epoch": 88.97,
      "learning_rate": 8.137734627831716e-06,
      "loss": 0.0506,
      "step": 229090
    },
    {
      "epoch": 88.97,
      "learning_rate": 8.137216828478966e-06,
      "loss": 0.1272,
      "step": 229100
    },
    {
      "epoch": 88.97,
      "learning_rate": 8.136699029126214e-06,
      "loss": 0.0888,
      "step": 229110
    },
    {
      "epoch": 88.98,
      "learning_rate": 8.136181229773464e-06,
      "loss": 0.1335,
      "step": 229120
    },
    {
      "epoch": 88.98,
      "learning_rate": 8.135663430420713e-06,
      "loss": 0.0383,
      "step": 229130
    },
    {
      "epoch": 88.99,
      "learning_rate": 8.135145631067963e-06,
      "loss": 0.0239,
      "step": 229140
    },
    {
      "epoch": 88.99,
      "learning_rate": 8.13462783171521e-06,
      "loss": 0.0002,
      "step": 229150
    },
    {
      "epoch": 88.99,
      "learning_rate": 8.13411003236246e-06,
      "loss": 0.1077,
      "step": 229160
    },
    {
      "epoch": 89.0,
      "learning_rate": 8.13359223300971e-06,
      "loss": 0.1493,
      "step": 229170
    },
    {
      "epoch": 89.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3559674322605133,
      "eval_runtime": 8.2123,
      "eval_samples_per_second": 442.628,
      "eval_steps_per_second": 55.405,
      "step": 229175
    },
    {
      "epoch": 89.0,
      "learning_rate": 8.13307443365696e-06,
      "loss": 0.0005,
      "step": 229180
    },
    {
      "epoch": 89.01,
      "learning_rate": 8.132556634304208e-06,
      "loss": 0.0232,
      "step": 229190
    },
    {
      "epoch": 89.01,
      "learning_rate": 8.132038834951457e-06,
      "loss": 0.0862,
      "step": 229200
    },
    {
      "epoch": 89.01,
      "learning_rate": 8.131521035598707e-06,
      "loss": 0.0457,
      "step": 229210
    },
    {
      "epoch": 89.02,
      "learning_rate": 8.131003236245956e-06,
      "loss": 0.0373,
      "step": 229220
    },
    {
      "epoch": 89.02,
      "learning_rate": 8.130485436893204e-06,
      "loss": 0.1482,
      "step": 229230
    },
    {
      "epoch": 89.03,
      "learning_rate": 8.129967637540454e-06,
      "loss": 0.014,
      "step": 229240
    },
    {
      "epoch": 89.03,
      "learning_rate": 8.129449838187704e-06,
      "loss": 0.0172,
      "step": 229250
    },
    {
      "epoch": 89.03,
      "learning_rate": 8.128932038834953e-06,
      "loss": 0.0105,
      "step": 229260
    },
    {
      "epoch": 89.04,
      "learning_rate": 8.128414239482201e-06,
      "loss": 0.0966,
      "step": 229270
    },
    {
      "epoch": 89.04,
      "learning_rate": 8.12789644012945e-06,
      "loss": 0.0411,
      "step": 229280
    },
    {
      "epoch": 89.04,
      "learning_rate": 8.1273786407767e-06,
      "loss": 0.1419,
      "step": 229290
    },
    {
      "epoch": 89.05,
      "learning_rate": 8.12686084142395e-06,
      "loss": 0.1712,
      "step": 229300
    },
    {
      "epoch": 89.05,
      "learning_rate": 8.126343042071198e-06,
      "loss": 0.0085,
      "step": 229310
    },
    {
      "epoch": 89.06,
      "learning_rate": 8.125825242718447e-06,
      "loss": 0.0041,
      "step": 229320
    },
    {
      "epoch": 89.06,
      "learning_rate": 8.125307443365697e-06,
      "loss": 0.0279,
      "step": 229330
    },
    {
      "epoch": 89.06,
      "learning_rate": 8.124789644012947e-06,
      "loss": 0.1184,
      "step": 229340
    },
    {
      "epoch": 89.07,
      "learning_rate": 8.124271844660195e-06,
      "loss": 0.0003,
      "step": 229350
    },
    {
      "epoch": 89.07,
      "learning_rate": 8.123754045307444e-06,
      "loss": 0.13,
      "step": 229360
    },
    {
      "epoch": 89.08,
      "learning_rate": 8.123236245954694e-06,
      "loss": 0.1883,
      "step": 229370
    },
    {
      "epoch": 89.08,
      "learning_rate": 8.122718446601942e-06,
      "loss": 0.1589,
      "step": 229380
    },
    {
      "epoch": 89.08,
      "learning_rate": 8.122200647249191e-06,
      "loss": 0.0355,
      "step": 229390
    },
    {
      "epoch": 89.09,
      "learning_rate": 8.121682847896441e-06,
      "loss": 0.0522,
      "step": 229400
    },
    {
      "epoch": 89.09,
      "learning_rate": 8.12116504854369e-06,
      "loss": 0.1907,
      "step": 229410
    },
    {
      "epoch": 89.1,
      "learning_rate": 8.120647249190939e-06,
      "loss": 0.002,
      "step": 229420
    },
    {
      "epoch": 89.1,
      "learning_rate": 8.120129449838188e-06,
      "loss": 0.0042,
      "step": 229430
    },
    {
      "epoch": 89.1,
      "learning_rate": 8.119611650485438e-06,
      "loss": 0.0855,
      "step": 229440
    },
    {
      "epoch": 89.11,
      "learning_rate": 8.119093851132687e-06,
      "loss": 0.1521,
      "step": 229450
    },
    {
      "epoch": 89.11,
      "learning_rate": 8.118576051779935e-06,
      "loss": 0.0688,
      "step": 229460
    },
    {
      "epoch": 89.11,
      "learning_rate": 8.118058252427185e-06,
      "loss": 0.0799,
      "step": 229470
    },
    {
      "epoch": 89.12,
      "learning_rate": 8.117540453074435e-06,
      "loss": 0.1158,
      "step": 229480
    },
    {
      "epoch": 89.12,
      "learning_rate": 8.117022653721684e-06,
      "loss": 0.0566,
      "step": 229490
    },
    {
      "epoch": 89.13,
      "learning_rate": 8.116504854368932e-06,
      "loss": 0.0009,
      "step": 229500
    },
    {
      "epoch": 89.13,
      "learning_rate": 8.115987055016182e-06,
      "loss": 0.0012,
      "step": 229510
    },
    {
      "epoch": 89.13,
      "learning_rate": 8.115469255663431e-06,
      "loss": 0.0514,
      "step": 229520
    },
    {
      "epoch": 89.14,
      "learning_rate": 8.114951456310681e-06,
      "loss": 0.0026,
      "step": 229530
    },
    {
      "epoch": 89.14,
      "learning_rate": 8.114433656957929e-06,
      "loss": 0.0282,
      "step": 229540
    },
    {
      "epoch": 89.15,
      "learning_rate": 8.113915857605179e-06,
      "loss": 0.0271,
      "step": 229550
    },
    {
      "epoch": 89.15,
      "learning_rate": 8.113398058252428e-06,
      "loss": 0.0982,
      "step": 229560
    },
    {
      "epoch": 89.15,
      "learning_rate": 8.112880258899676e-06,
      "loss": 0.0202,
      "step": 229570
    },
    {
      "epoch": 89.16,
      "learning_rate": 8.112362459546926e-06,
      "loss": 0.0705,
      "step": 229580
    },
    {
      "epoch": 89.16,
      "learning_rate": 8.111844660194175e-06,
      "loss": 0.015,
      "step": 229590
    },
    {
      "epoch": 89.17,
      "learning_rate": 8.111326860841425e-06,
      "loss": 0.0684,
      "step": 229600
    },
    {
      "epoch": 89.17,
      "learning_rate": 8.110809061488673e-06,
      "loss": 0.1066,
      "step": 229610
    },
    {
      "epoch": 89.17,
      "learning_rate": 8.110291262135922e-06,
      "loss": 0.0008,
      "step": 229620
    },
    {
      "epoch": 89.18,
      "learning_rate": 8.109773462783172e-06,
      "loss": 0.001,
      "step": 229630
    },
    {
      "epoch": 89.18,
      "learning_rate": 8.109255663430422e-06,
      "loss": 0.0544,
      "step": 229640
    },
    {
      "epoch": 89.18,
      "learning_rate": 8.10873786407767e-06,
      "loss": 0.1639,
      "step": 229650
    },
    {
      "epoch": 89.19,
      "learning_rate": 8.10822006472492e-06,
      "loss": 0.0535,
      "step": 229660
    },
    {
      "epoch": 89.19,
      "learning_rate": 8.107702265372169e-06,
      "loss": 0.0448,
      "step": 229670
    },
    {
      "epoch": 89.2,
      "learning_rate": 8.107184466019418e-06,
      "loss": 0.0885,
      "step": 229680
    },
    {
      "epoch": 89.2,
      "learning_rate": 8.106666666666666e-06,
      "loss": 0.0829,
      "step": 229690
    },
    {
      "epoch": 89.2,
      "learning_rate": 8.106148867313916e-06,
      "loss": 0.0538,
      "step": 229700
    },
    {
      "epoch": 89.21,
      "learning_rate": 8.105631067961166e-06,
      "loss": 0.0467,
      "step": 229710
    },
    {
      "epoch": 89.21,
      "learning_rate": 8.105113268608414e-06,
      "loss": 0.19,
      "step": 229720
    },
    {
      "epoch": 89.22,
      "learning_rate": 8.104595469255663e-06,
      "loss": 0.0549,
      "step": 229730
    },
    {
      "epoch": 89.22,
      "learning_rate": 8.104077669902913e-06,
      "loss": 0.0375,
      "step": 229740
    },
    {
      "epoch": 89.22,
      "learning_rate": 8.103559870550162e-06,
      "loss": 0.0332,
      "step": 229750
    },
    {
      "epoch": 89.23,
      "learning_rate": 8.10304207119741e-06,
      "loss": 0.0358,
      "step": 229760
    },
    {
      "epoch": 89.23,
      "learning_rate": 8.10252427184466e-06,
      "loss": 0.0389,
      "step": 229770
    },
    {
      "epoch": 89.23,
      "learning_rate": 8.10200647249191e-06,
      "loss": 0.1136,
      "step": 229780
    },
    {
      "epoch": 89.24,
      "learning_rate": 8.10148867313916e-06,
      "loss": 0.0036,
      "step": 229790
    },
    {
      "epoch": 89.24,
      "learning_rate": 8.100970873786407e-06,
      "loss": 0.1126,
      "step": 229800
    },
    {
      "epoch": 89.25,
      "learning_rate": 8.100453074433657e-06,
      "loss": 0.0225,
      "step": 229810
    },
    {
      "epoch": 89.25,
      "learning_rate": 8.099935275080906e-06,
      "loss": 0.2919,
      "step": 229820
    },
    {
      "epoch": 89.25,
      "learning_rate": 8.099417475728156e-06,
      "loss": 0.0138,
      "step": 229830
    },
    {
      "epoch": 89.26,
      "learning_rate": 8.098899676375404e-06,
      "loss": 0.0413,
      "step": 229840
    },
    {
      "epoch": 89.26,
      "learning_rate": 8.098381877022654e-06,
      "loss": 0.0202,
      "step": 229850
    },
    {
      "epoch": 89.27,
      "learning_rate": 8.097864077669903e-06,
      "loss": 0.0173,
      "step": 229860
    },
    {
      "epoch": 89.27,
      "learning_rate": 8.097346278317153e-06,
      "loss": 0.055,
      "step": 229870
    },
    {
      "epoch": 89.27,
      "learning_rate": 8.096828478964402e-06,
      "loss": 0.1036,
      "step": 229880
    },
    {
      "epoch": 89.28,
      "learning_rate": 8.09631067961165e-06,
      "loss": 0.2198,
      "step": 229890
    },
    {
      "epoch": 89.28,
      "learning_rate": 8.0957928802589e-06,
      "loss": 0.036,
      "step": 229900
    },
    {
      "epoch": 89.29,
      "learning_rate": 8.09527508090615e-06,
      "loss": 0.1316,
      "step": 229910
    },
    {
      "epoch": 89.29,
      "learning_rate": 8.0947572815534e-06,
      "loss": 0.0079,
      "step": 229920
    },
    {
      "epoch": 89.29,
      "learning_rate": 8.094239482200647e-06,
      "loss": 0.109,
      "step": 229930
    },
    {
      "epoch": 89.3,
      "learning_rate": 8.093721682847897e-06,
      "loss": 0.0643,
      "step": 229940
    },
    {
      "epoch": 89.3,
      "learning_rate": 8.093203883495146e-06,
      "loss": 0.029,
      "step": 229950
    },
    {
      "epoch": 89.3,
      "learning_rate": 8.092686084142396e-06,
      "loss": 0.1209,
      "step": 229960
    },
    {
      "epoch": 89.31,
      "learning_rate": 8.092168284789644e-06,
      "loss": 0.0383,
      "step": 229970
    },
    {
      "epoch": 89.31,
      "learning_rate": 8.091650485436893e-06,
      "loss": 0.0404,
      "step": 229980
    },
    {
      "epoch": 89.32,
      "learning_rate": 8.091132686084143e-06,
      "loss": 0.1191,
      "step": 229990
    },
    {
      "epoch": 89.32,
      "learning_rate": 8.090614886731393e-06,
      "loss": 0.0105,
      "step": 230000
    },
    {
      "epoch": 89.32,
      "learning_rate": 8.09009708737864e-06,
      "loss": 0.016,
      "step": 230010
    },
    {
      "epoch": 89.33,
      "learning_rate": 8.08957928802589e-06,
      "loss": 0.0116,
      "step": 230020
    },
    {
      "epoch": 89.33,
      "learning_rate": 8.08906148867314e-06,
      "loss": 0.0169,
      "step": 230030
    },
    {
      "epoch": 89.34,
      "learning_rate": 8.08854368932039e-06,
      "loss": 0.029,
      "step": 230040
    },
    {
      "epoch": 89.34,
      "learning_rate": 8.088025889967637e-06,
      "loss": 0.0105,
      "step": 230050
    },
    {
      "epoch": 89.34,
      "learning_rate": 8.087508090614887e-06,
      "loss": 0.1755,
      "step": 230060
    },
    {
      "epoch": 89.35,
      "learning_rate": 8.086990291262137e-06,
      "loss": 0.0381,
      "step": 230070
    },
    {
      "epoch": 89.35,
      "learning_rate": 8.086472491909386e-06,
      "loss": 0.0255,
      "step": 230080
    },
    {
      "epoch": 89.36,
      "learning_rate": 8.085954692556634e-06,
      "loss": 0.0643,
      "step": 230090
    },
    {
      "epoch": 89.36,
      "learning_rate": 8.085436893203884e-06,
      "loss": 0.0535,
      "step": 230100
    },
    {
      "epoch": 89.36,
      "learning_rate": 8.084919093851133e-06,
      "loss": 0.0432,
      "step": 230110
    },
    {
      "epoch": 89.37,
      "learning_rate": 8.084401294498383e-06,
      "loss": 0.069,
      "step": 230120
    },
    {
      "epoch": 89.37,
      "learning_rate": 8.083883495145631e-06,
      "loss": 0.0134,
      "step": 230130
    },
    {
      "epoch": 89.37,
      "learning_rate": 8.08336569579288e-06,
      "loss": 0.0146,
      "step": 230140
    },
    {
      "epoch": 89.38,
      "learning_rate": 8.08284789644013e-06,
      "loss": 0.02,
      "step": 230150
    },
    {
      "epoch": 89.38,
      "learning_rate": 8.08233009708738e-06,
      "loss": 0.0707,
      "step": 230160
    },
    {
      "epoch": 89.39,
      "learning_rate": 8.081812297734628e-06,
      "loss": 0.0582,
      "step": 230170
    },
    {
      "epoch": 89.39,
      "learning_rate": 8.081294498381877e-06,
      "loss": 0.035,
      "step": 230180
    },
    {
      "epoch": 89.39,
      "learning_rate": 8.080776699029127e-06,
      "loss": 0.0697,
      "step": 230190
    },
    {
      "epoch": 89.4,
      "learning_rate": 8.080258899676377e-06,
      "loss": 0.0004,
      "step": 230200
    },
    {
      "epoch": 89.4,
      "learning_rate": 8.079741100323625e-06,
      "loss": 0.0178,
      "step": 230210
    },
    {
      "epoch": 89.41,
      "learning_rate": 8.079223300970874e-06,
      "loss": 0.0795,
      "step": 230220
    },
    {
      "epoch": 89.41,
      "learning_rate": 8.078705501618124e-06,
      "loss": 0.0311,
      "step": 230230
    },
    {
      "epoch": 89.41,
      "learning_rate": 8.078187702265373e-06,
      "loss": 0.0228,
      "step": 230240
    },
    {
      "epoch": 89.42,
      "learning_rate": 8.077669902912621e-06,
      "loss": 0.001,
      "step": 230250
    },
    {
      "epoch": 89.42,
      "learning_rate": 8.077152103559871e-06,
      "loss": 0.0514,
      "step": 230260
    },
    {
      "epoch": 89.43,
      "learning_rate": 8.07663430420712e-06,
      "loss": 0.0646,
      "step": 230270
    },
    {
      "epoch": 89.43,
      "learning_rate": 8.07611650485437e-06,
      "loss": 0.0129,
      "step": 230280
    },
    {
      "epoch": 89.43,
      "learning_rate": 8.075598705501618e-06,
      "loss": 0.0294,
      "step": 230290
    },
    {
      "epoch": 89.44,
      "learning_rate": 8.075080906148868e-06,
      "loss": 0.1105,
      "step": 230300
    },
    {
      "epoch": 89.44,
      "learning_rate": 8.074563106796117e-06,
      "loss": 0.1377,
      "step": 230310
    },
    {
      "epoch": 89.44,
      "learning_rate": 8.074045307443367e-06,
      "loss": 0.023,
      "step": 230320
    },
    {
      "epoch": 89.45,
      "learning_rate": 8.073527508090615e-06,
      "loss": 0.1767,
      "step": 230330
    },
    {
      "epoch": 89.45,
      "learning_rate": 8.073009708737864e-06,
      "loss": 0.0378,
      "step": 230340
    },
    {
      "epoch": 89.46,
      "learning_rate": 8.072491909385114e-06,
      "loss": 0.0732,
      "step": 230350
    },
    {
      "epoch": 89.46,
      "learning_rate": 8.071974110032364e-06,
      "loss": 0.0488,
      "step": 230360
    },
    {
      "epoch": 89.46,
      "learning_rate": 8.071456310679612e-06,
      "loss": 0.0525,
      "step": 230370
    },
    {
      "epoch": 89.47,
      "learning_rate": 8.070938511326861e-06,
      "loss": 0.0422,
      "step": 230380
    },
    {
      "epoch": 89.47,
      "learning_rate": 8.070420711974111e-06,
      "loss": 0.0311,
      "step": 230390
    },
    {
      "epoch": 89.48,
      "learning_rate": 8.06990291262136e-06,
      "loss": 0.0118,
      "step": 230400
    },
    {
      "epoch": 89.48,
      "learning_rate": 8.06938511326861e-06,
      "loss": 0.1679,
      "step": 230410
    },
    {
      "epoch": 89.48,
      "learning_rate": 8.068867313915858e-06,
      "loss": 0.0081,
      "step": 230420
    },
    {
      "epoch": 89.49,
      "learning_rate": 8.068349514563108e-06,
      "loss": 0.0216,
      "step": 230430
    },
    {
      "epoch": 89.49,
      "learning_rate": 8.067831715210357e-06,
      "loss": 0.0191,
      "step": 230440
    },
    {
      "epoch": 89.5,
      "learning_rate": 8.067313915857607e-06,
      "loss": 0.0337,
      "step": 230450
    },
    {
      "epoch": 89.5,
      "learning_rate": 8.066796116504855e-06,
      "loss": 0.029,
      "step": 230460
    },
    {
      "epoch": 89.5,
      "learning_rate": 8.066278317152104e-06,
      "loss": 0.1171,
      "step": 230470
    },
    {
      "epoch": 89.51,
      "learning_rate": 8.065760517799354e-06,
      "loss": 0.2098,
      "step": 230480
    },
    {
      "epoch": 89.51,
      "learning_rate": 8.065242718446604e-06,
      "loss": 0.0441,
      "step": 230490
    },
    {
      "epoch": 89.51,
      "learning_rate": 8.064724919093852e-06,
      "loss": 0.1055,
      "step": 230500
    },
    {
      "epoch": 89.52,
      "learning_rate": 8.064207119741101e-06,
      "loss": 0.0045,
      "step": 230510
    },
    {
      "epoch": 89.52,
      "learning_rate": 8.06368932038835e-06,
      "loss": 0.07,
      "step": 230520
    },
    {
      "epoch": 89.53,
      "learning_rate": 8.0631715210356e-06,
      "loss": 0.0558,
      "step": 230530
    },
    {
      "epoch": 89.53,
      "learning_rate": 8.062653721682848e-06,
      "loss": 0.1042,
      "step": 230540
    },
    {
      "epoch": 89.53,
      "learning_rate": 8.062135922330098e-06,
      "loss": 0.0074,
      "step": 230550
    },
    {
      "epoch": 89.54,
      "learning_rate": 8.061618122977348e-06,
      "loss": 0.1678,
      "step": 230560
    },
    {
      "epoch": 89.54,
      "learning_rate": 8.061100323624597e-06,
      "loss": 0.0022,
      "step": 230570
    },
    {
      "epoch": 89.55,
      "learning_rate": 8.060582524271845e-06,
      "loss": 0.0186,
      "step": 230580
    },
    {
      "epoch": 89.55,
      "learning_rate": 8.060064724919095e-06,
      "loss": 0.0299,
      "step": 230590
    },
    {
      "epoch": 89.55,
      "learning_rate": 8.059546925566344e-06,
      "loss": 0.0469,
      "step": 230600
    },
    {
      "epoch": 89.56,
      "learning_rate": 8.059029126213594e-06,
      "loss": 0.0068,
      "step": 230610
    },
    {
      "epoch": 89.56,
      "learning_rate": 8.058511326860842e-06,
      "loss": 0.0099,
      "step": 230620
    },
    {
      "epoch": 89.57,
      "learning_rate": 8.057993527508092e-06,
      "loss": 0.0715,
      "step": 230630
    },
    {
      "epoch": 89.57,
      "learning_rate": 8.057475728155341e-06,
      "loss": 0.0332,
      "step": 230640
    },
    {
      "epoch": 89.57,
      "learning_rate": 8.05695792880259e-06,
      "loss": 0.1022,
      "step": 230650
    },
    {
      "epoch": 89.58,
      "learning_rate": 8.056440129449839e-06,
      "loss": 0.1461,
      "step": 230660
    },
    {
      "epoch": 89.58,
      "learning_rate": 8.055922330097088e-06,
      "loss": 0.0069,
      "step": 230670
    },
    {
      "epoch": 89.58,
      "learning_rate": 8.055404530744338e-06,
      "loss": 0.0484,
      "step": 230680
    },
    {
      "epoch": 89.59,
      "learning_rate": 8.054886731391588e-06,
      "loss": 0.0896,
      "step": 230690
    },
    {
      "epoch": 89.59,
      "learning_rate": 8.054368932038835e-06,
      "loss": 0.002,
      "step": 230700
    },
    {
      "epoch": 89.6,
      "learning_rate": 8.053851132686085e-06,
      "loss": 0.0217,
      "step": 230710
    },
    {
      "epoch": 89.6,
      "learning_rate": 8.053333333333335e-06,
      "loss": 0.0505,
      "step": 230720
    },
    {
      "epoch": 89.6,
      "learning_rate": 8.052815533980584e-06,
      "loss": 0.2077,
      "step": 230730
    },
    {
      "epoch": 89.61,
      "learning_rate": 8.052297734627832e-06,
      "loss": 0.0226,
      "step": 230740
    },
    {
      "epoch": 89.61,
      "learning_rate": 8.051779935275082e-06,
      "loss": 0.1332,
      "step": 230750
    },
    {
      "epoch": 89.62,
      "learning_rate": 8.051262135922331e-06,
      "loss": 0.0896,
      "step": 230760
    },
    {
      "epoch": 89.62,
      "learning_rate": 8.050744336569581e-06,
      "loss": 0.0055,
      "step": 230770
    },
    {
      "epoch": 89.62,
      "learning_rate": 8.050226537216829e-06,
      "loss": 0.0425,
      "step": 230780
    },
    {
      "epoch": 89.63,
      "learning_rate": 8.049708737864079e-06,
      "loss": 0.0267,
      "step": 230790
    },
    {
      "epoch": 89.63,
      "learning_rate": 8.049190938511328e-06,
      "loss": 0.0618,
      "step": 230800
    },
    {
      "epoch": 89.63,
      "learning_rate": 8.048673139158578e-06,
      "loss": 0.0886,
      "step": 230810
    },
    {
      "epoch": 89.64,
      "learning_rate": 8.048155339805826e-06,
      "loss": 0.0014,
      "step": 230820
    },
    {
      "epoch": 89.64,
      "learning_rate": 8.047637540453075e-06,
      "loss": 0.0331,
      "step": 230830
    },
    {
      "epoch": 89.65,
      "learning_rate": 8.047119741100325e-06,
      "loss": 0.0288,
      "step": 230840
    },
    {
      "epoch": 89.65,
      "learning_rate": 8.046601941747573e-06,
      "loss": 0.0189,
      "step": 230850
    },
    {
      "epoch": 89.65,
      "learning_rate": 8.046084142394823e-06,
      "loss": 0.1266,
      "step": 230860
    },
    {
      "epoch": 89.66,
      "learning_rate": 8.045566343042072e-06,
      "loss": 0.0278,
      "step": 230870
    },
    {
      "epoch": 89.66,
      "learning_rate": 8.045048543689322e-06,
      "loss": 0.009,
      "step": 230880
    },
    {
      "epoch": 89.67,
      "learning_rate": 8.04453074433657e-06,
      "loss": 0.0035,
      "step": 230890
    },
    {
      "epoch": 89.67,
      "learning_rate": 8.04401294498382e-06,
      "loss": 0.1982,
      "step": 230900
    },
    {
      "epoch": 89.67,
      "learning_rate": 8.043495145631069e-06,
      "loss": 0.1187,
      "step": 230910
    },
    {
      "epoch": 89.68,
      "learning_rate": 8.042977346278319e-06,
      "loss": 0.0007,
      "step": 230920
    },
    {
      "epoch": 89.68,
      "learning_rate": 8.042459546925567e-06,
      "loss": 0.0466,
      "step": 230930
    },
    {
      "epoch": 89.69,
      "learning_rate": 8.041941747572816e-06,
      "loss": 0.0397,
      "step": 230940
    },
    {
      "epoch": 89.69,
      "learning_rate": 8.041423948220066e-06,
      "loss": 0.0801,
      "step": 230950
    },
    {
      "epoch": 89.69,
      "learning_rate": 8.040906148867315e-06,
      "loss": 0.0876,
      "step": 230960
    },
    {
      "epoch": 89.7,
      "learning_rate": 8.040388349514563e-06,
      "loss": 0.0012,
      "step": 230970
    },
    {
      "epoch": 89.7,
      "learning_rate": 8.039870550161813e-06,
      "loss": 0.0019,
      "step": 230980
    },
    {
      "epoch": 89.7,
      "learning_rate": 8.039352750809063e-06,
      "loss": 0.0329,
      "step": 230990
    },
    {
      "epoch": 89.71,
      "learning_rate": 8.038834951456312e-06,
      "loss": 0.0184,
      "step": 231000
    },
    {
      "epoch": 89.71,
      "learning_rate": 8.03831715210356e-06,
      "loss": 0.0722,
      "step": 231010
    },
    {
      "epoch": 89.72,
      "learning_rate": 8.03779935275081e-06,
      "loss": 0.2023,
      "step": 231020
    },
    {
      "epoch": 89.72,
      "learning_rate": 8.03728155339806e-06,
      "loss": 0.0179,
      "step": 231030
    },
    {
      "epoch": 89.72,
      "learning_rate": 8.036763754045307e-06,
      "loss": 0.0735,
      "step": 231040
    },
    {
      "epoch": 89.73,
      "learning_rate": 8.036245954692557e-06,
      "loss": 0.1068,
      "step": 231050
    },
    {
      "epoch": 89.73,
      "learning_rate": 8.035728155339806e-06,
      "loss": 0.0208,
      "step": 231060
    },
    {
      "epoch": 89.74,
      "learning_rate": 8.035210355987056e-06,
      "loss": 0.0097,
      "step": 231070
    },
    {
      "epoch": 89.74,
      "learning_rate": 8.034692556634304e-06,
      "loss": 0.0548,
      "step": 231080
    },
    {
      "epoch": 89.74,
      "learning_rate": 8.034174757281554e-06,
      "loss": 0.0436,
      "step": 231090
    },
    {
      "epoch": 89.75,
      "learning_rate": 8.033656957928803e-06,
      "loss": 0.022,
      "step": 231100
    },
    {
      "epoch": 89.75,
      "learning_rate": 8.033139158576053e-06,
      "loss": 0.135,
      "step": 231110
    },
    {
      "epoch": 89.76,
      "learning_rate": 8.0326213592233e-06,
      "loss": 0.0277,
      "step": 231120
    },
    {
      "epoch": 89.76,
      "learning_rate": 8.03210355987055e-06,
      "loss": 0.004,
      "step": 231130
    },
    {
      "epoch": 89.76,
      "learning_rate": 8.0315857605178e-06,
      "loss": 0.1233,
      "step": 231140
    },
    {
      "epoch": 89.77,
      "learning_rate": 8.03106796116505e-06,
      "loss": 0.129,
      "step": 231150
    },
    {
      "epoch": 89.77,
      "learning_rate": 8.030550161812298e-06,
      "loss": 0.0628,
      "step": 231160
    },
    {
      "epoch": 89.77,
      "learning_rate": 8.030032362459547e-06,
      "loss": 0.0831,
      "step": 231170
    },
    {
      "epoch": 89.78,
      "learning_rate": 8.029514563106797e-06,
      "loss": 0.0896,
      "step": 231180
    },
    {
      "epoch": 89.78,
      "learning_rate": 8.028996763754045e-06,
      "loss": 0.158,
      "step": 231190
    },
    {
      "epoch": 89.79,
      "learning_rate": 8.028478964401294e-06,
      "loss": 0.0874,
      "step": 231200
    },
    {
      "epoch": 89.79,
      "learning_rate": 8.027961165048544e-06,
      "loss": 0.0976,
      "step": 231210
    },
    {
      "epoch": 89.79,
      "learning_rate": 8.027443365695794e-06,
      "loss": 0.0494,
      "step": 231220
    },
    {
      "epoch": 89.8,
      "learning_rate": 8.026925566343042e-06,
      "loss": 0.0351,
      "step": 231230
    },
    {
      "epoch": 89.8,
      "learning_rate": 8.026407766990291e-06,
      "loss": 0.009,
      "step": 231240
    },
    {
      "epoch": 89.81,
      "learning_rate": 8.02588996763754e-06,
      "loss": 0.1917,
      "step": 231250
    },
    {
      "epoch": 89.81,
      "learning_rate": 8.02537216828479e-06,
      "loss": 0.0227,
      "step": 231260
    },
    {
      "epoch": 89.81,
      "learning_rate": 8.024854368932038e-06,
      "loss": 0.05,
      "step": 231270
    },
    {
      "epoch": 89.82,
      "learning_rate": 8.024336569579288e-06,
      "loss": 0.0341,
      "step": 231280
    },
    {
      "epoch": 89.82,
      "learning_rate": 8.023818770226538e-06,
      "loss": 0.1045,
      "step": 231290
    },
    {
      "epoch": 89.83,
      "learning_rate": 8.023300970873787e-06,
      "loss": 0.099,
      "step": 231300
    },
    {
      "epoch": 89.83,
      "learning_rate": 8.022783171521035e-06,
      "loss": 0.0581,
      "step": 231310
    },
    {
      "epoch": 89.83,
      "learning_rate": 8.022265372168285e-06,
      "loss": 0.0525,
      "step": 231320
    },
    {
      "epoch": 89.84,
      "learning_rate": 8.021747572815534e-06,
      "loss": 0.1286,
      "step": 231330
    },
    {
      "epoch": 89.84,
      "learning_rate": 8.021229773462784e-06,
      "loss": 0.0689,
      "step": 231340
    },
    {
      "epoch": 89.84,
      "learning_rate": 8.020711974110032e-06,
      "loss": 0.1125,
      "step": 231350
    },
    {
      "epoch": 89.85,
      "learning_rate": 8.020194174757281e-06,
      "loss": 0.0768,
      "step": 231360
    },
    {
      "epoch": 89.85,
      "learning_rate": 8.019676375404531e-06,
      "loss": 0.0663,
      "step": 231370
    },
    {
      "epoch": 89.86,
      "learning_rate": 8.01915857605178e-06,
      "loss": 0.0033,
      "step": 231380
    },
    {
      "epoch": 89.86,
      "learning_rate": 8.018640776699029e-06,
      "loss": 0.1355,
      "step": 231390
    },
    {
      "epoch": 89.86,
      "learning_rate": 8.018122977346278e-06,
      "loss": 0.158,
      "step": 231400
    },
    {
      "epoch": 89.87,
      "learning_rate": 8.017605177993528e-06,
      "loss": 0.0008,
      "step": 231410
    },
    {
      "epoch": 89.87,
      "learning_rate": 8.017087378640777e-06,
      "loss": 0.169,
      "step": 231420
    },
    {
      "epoch": 89.88,
      "learning_rate": 8.016569579288025e-06,
      "loss": 0.0593,
      "step": 231430
    },
    {
      "epoch": 89.88,
      "learning_rate": 8.016051779935275e-06,
      "loss": 0.0584,
      "step": 231440
    },
    {
      "epoch": 89.88,
      "learning_rate": 8.015533980582525e-06,
      "loss": 0.1459,
      "step": 231450
    },
    {
      "epoch": 89.89,
      "learning_rate": 8.015016181229774e-06,
      "loss": 0.1572,
      "step": 231460
    },
    {
      "epoch": 89.89,
      "learning_rate": 8.014498381877022e-06,
      "loss": 0.0076,
      "step": 231470
    },
    {
      "epoch": 89.9,
      "learning_rate": 8.013980582524272e-06,
      "loss": 0.0516,
      "step": 231480
    },
    {
      "epoch": 89.9,
      "learning_rate": 8.013462783171521e-06,
      "loss": 0.0315,
      "step": 231490
    },
    {
      "epoch": 89.9,
      "learning_rate": 8.012944983818771e-06,
      "loss": 0.1207,
      "step": 231500
    },
    {
      "epoch": 89.91,
      "learning_rate": 8.012427184466019e-06,
      "loss": 0.2,
      "step": 231510
    },
    {
      "epoch": 89.91,
      "learning_rate": 8.011909385113269e-06,
      "loss": 0.0008,
      "step": 231520
    },
    {
      "epoch": 89.91,
      "learning_rate": 8.011391585760518e-06,
      "loss": 0.0633,
      "step": 231530
    },
    {
      "epoch": 89.92,
      "learning_rate": 8.010873786407768e-06,
      "loss": 0.0126,
      "step": 231540
    },
    {
      "epoch": 89.92,
      "learning_rate": 8.010355987055017e-06,
      "loss": 0.0062,
      "step": 231550
    },
    {
      "epoch": 89.93,
      "learning_rate": 8.009838187702265e-06,
      "loss": 0.0946,
      "step": 231560
    },
    {
      "epoch": 89.93,
      "learning_rate": 8.009320388349515e-06,
      "loss": 0.2503,
      "step": 231570
    },
    {
      "epoch": 89.93,
      "learning_rate": 8.008802588996765e-06,
      "loss": 0.0033,
      "step": 231580
    },
    {
      "epoch": 89.94,
      "learning_rate": 8.008284789644014e-06,
      "loss": 0.0722,
      "step": 231590
    },
    {
      "epoch": 89.94,
      "learning_rate": 8.007766990291262e-06,
      "loss": 0.1168,
      "step": 231600
    },
    {
      "epoch": 89.95,
      "learning_rate": 8.007249190938512e-06,
      "loss": 0.0325,
      "step": 231610
    },
    {
      "epoch": 89.95,
      "learning_rate": 8.006731391585761e-06,
      "loss": 0.1396,
      "step": 231620
    },
    {
      "epoch": 89.95,
      "learning_rate": 8.006213592233011e-06,
      "loss": 0.1214,
      "step": 231630
    },
    {
      "epoch": 89.96,
      "learning_rate": 8.005695792880259e-06,
      "loss": 0.1364,
      "step": 231640
    },
    {
      "epoch": 89.96,
      "learning_rate": 8.005177993527509e-06,
      "loss": 0.1105,
      "step": 231650
    },
    {
      "epoch": 89.97,
      "learning_rate": 8.004660194174758e-06,
      "loss": 0.0365,
      "step": 231660
    },
    {
      "epoch": 89.97,
      "learning_rate": 8.004142394822008e-06,
      "loss": 0.1072,
      "step": 231670
    },
    {
      "epoch": 89.97,
      "learning_rate": 8.003624595469256e-06,
      "loss": 0.1143,
      "step": 231680
    },
    {
      "epoch": 89.98,
      "learning_rate": 8.003106796116505e-06,
      "loss": 0.0029,
      "step": 231690
    },
    {
      "epoch": 89.98,
      "learning_rate": 8.002588996763755e-06,
      "loss": 0.0909,
      "step": 231700
    },
    {
      "epoch": 89.98,
      "learning_rate": 8.002071197411005e-06,
      "loss": 0.0898,
      "step": 231710
    },
    {
      "epoch": 89.99,
      "learning_rate": 8.001553398058252e-06,
      "loss": 0.0676,
      "step": 231720
    },
    {
      "epoch": 89.99,
      "learning_rate": 8.001035598705502e-06,
      "loss": 0.0568,
      "step": 231730
    },
    {
      "epoch": 90.0,
      "learning_rate": 8.000517799352752e-06,
      "loss": 0.0094,
      "step": 231740
    },
    {
      "epoch": 90.0,
      "learning_rate": 8.000000000000001e-06,
      "loss": 0.0164,
      "step": 231750
    },
    {
      "epoch": 90.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.3572826385498047,
      "eval_runtime": 8.1761,
      "eval_samples_per_second": 444.589,
      "eval_steps_per_second": 55.65,
      "step": 231750
    },
    {
      "epoch": 90.0,
      "learning_rate": 7.99948220064725e-06,
      "loss": 0.1009,
      "step": 231760
    },
    {
      "epoch": 90.01,
      "learning_rate": 7.998964401294499e-06,
      "loss": 0.114,
      "step": 231770
    },
    {
      "epoch": 90.01,
      "learning_rate": 7.998446601941748e-06,
      "loss": 0.0455,
      "step": 231780
    },
    {
      "epoch": 90.02,
      "learning_rate": 7.997928802588998e-06,
      "loss": 0.0555,
      "step": 231790
    },
    {
      "epoch": 90.02,
      "learning_rate": 7.997411003236246e-06,
      "loss": 0.0476,
      "step": 231800
    },
    {
      "epoch": 90.02,
      "learning_rate": 7.996893203883496e-06,
      "loss": 0.0476,
      "step": 231810
    },
    {
      "epoch": 90.03,
      "learning_rate": 7.996375404530745e-06,
      "loss": 0.1748,
      "step": 231820
    },
    {
      "epoch": 90.03,
      "learning_rate": 7.995857605177995e-06,
      "loss": 0.1309,
      "step": 231830
    },
    {
      "epoch": 90.03,
      "learning_rate": 7.995339805825243e-06,
      "loss": 0.0164,
      "step": 231840
    },
    {
      "epoch": 90.04,
      "learning_rate": 7.994822006472492e-06,
      "loss": 0.1648,
      "step": 231850
    },
    {
      "epoch": 90.04,
      "learning_rate": 7.994304207119742e-06,
      "loss": 0.0611,
      "step": 231860
    },
    {
      "epoch": 90.05,
      "learning_rate": 7.993786407766992e-06,
      "loss": 0.0111,
      "step": 231870
    },
    {
      "epoch": 90.05,
      "learning_rate": 7.99326860841424e-06,
      "loss": 0.0334,
      "step": 231880
    },
    {
      "epoch": 90.05,
      "learning_rate": 7.99275080906149e-06,
      "loss": 0.1384,
      "step": 231890
    },
    {
      "epoch": 90.06,
      "learning_rate": 7.992233009708739e-06,
      "loss": 0.0222,
      "step": 231900
    },
    {
      "epoch": 90.06,
      "learning_rate": 7.991715210355988e-06,
      "loss": 0.0058,
      "step": 231910
    },
    {
      "epoch": 90.07,
      "learning_rate": 7.991197411003236e-06,
      "loss": 0.2447,
      "step": 231920
    },
    {
      "epoch": 90.07,
      "learning_rate": 7.990679611650486e-06,
      "loss": 0.1321,
      "step": 231930
    },
    {
      "epoch": 90.07,
      "learning_rate": 7.990161812297736e-06,
      "loss": 0.1063,
      "step": 231940
    },
    {
      "epoch": 90.08,
      "learning_rate": 7.989644012944985e-06,
      "loss": 0.0582,
      "step": 231950
    },
    {
      "epoch": 90.08,
      "learning_rate": 7.989126213592233e-06,
      "loss": 0.0146,
      "step": 231960
    },
    {
      "epoch": 90.09,
      "learning_rate": 7.988608414239483e-06,
      "loss": 0.0015,
      "step": 231970
    },
    {
      "epoch": 90.09,
      "learning_rate": 7.988090614886732e-06,
      "loss": 0.1494,
      "step": 231980
    },
    {
      "epoch": 90.09,
      "learning_rate": 7.987572815533982e-06,
      "loss": 0.0556,
      "step": 231990
    },
    {
      "epoch": 90.1,
      "learning_rate": 7.98705501618123e-06,
      "loss": 0.1115,
      "step": 232000
    },
    {
      "epoch": 90.1,
      "learning_rate": 7.98653721682848e-06,
      "loss": 0.0179,
      "step": 232010
    },
    {
      "epoch": 90.1,
      "learning_rate": 7.98601941747573e-06,
      "loss": 0.046,
      "step": 232020
    },
    {
      "epoch": 90.11,
      "learning_rate": 7.985501618122979e-06,
      "loss": 0.0019,
      "step": 232030
    },
    {
      "epoch": 90.11,
      "learning_rate": 7.984983818770227e-06,
      "loss": 0.0489,
      "step": 232040
    },
    {
      "epoch": 90.12,
      "learning_rate": 7.984466019417476e-06,
      "loss": 0.051,
      "step": 232050
    },
    {
      "epoch": 90.12,
      "learning_rate": 7.983948220064726e-06,
      "loss": 0.0661,
      "step": 232060
    },
    {
      "epoch": 90.12,
      "learning_rate": 7.983430420711976e-06,
      "loss": 0.0083,
      "step": 232070
    },
    {
      "epoch": 90.13,
      "learning_rate": 7.982912621359223e-06,
      "loss": 0.002,
      "step": 232080
    },
    {
      "epoch": 90.13,
      "learning_rate": 7.982394822006473e-06,
      "loss": 0.0574,
      "step": 232090
    },
    {
      "epoch": 90.14,
      "learning_rate": 7.981877022653723e-06,
      "loss": 0.0386,
      "step": 232100
    },
    {
      "epoch": 90.14,
      "learning_rate": 7.981359223300972e-06,
      "loss": 0.0434,
      "step": 232110
    },
    {
      "epoch": 90.14,
      "learning_rate": 7.980841423948222e-06,
      "loss": 0.053,
      "step": 232120
    },
    {
      "epoch": 90.15,
      "learning_rate": 7.98032362459547e-06,
      "loss": 0.034,
      "step": 232130
    },
    {
      "epoch": 90.15,
      "learning_rate": 7.97980582524272e-06,
      "loss": 0.0287,
      "step": 232140
    },
    {
      "epoch": 90.16,
      "learning_rate": 7.979288025889969e-06,
      "loss": 0.0547,
      "step": 232150
    },
    {
      "epoch": 90.16,
      "learning_rate": 7.978770226537219e-06,
      "loss": 0.0163,
      "step": 232160
    },
    {
      "epoch": 90.16,
      "learning_rate": 7.978252427184467e-06,
      "loss": 0.0907,
      "step": 232170
    },
    {
      "epoch": 90.17,
      "learning_rate": 7.977734627831716e-06,
      "loss": 0.206,
      "step": 232180
    },
    {
      "epoch": 90.17,
      "learning_rate": 7.977216828478966e-06,
      "loss": 0.0978,
      "step": 232190
    },
    {
      "epoch": 90.17,
      "learning_rate": 7.976699029126216e-06,
      "loss": 0.0795,
      "step": 232200
    },
    {
      "epoch": 90.18,
      "learning_rate": 7.976181229773463e-06,
      "loss": 0.0347,
      "step": 232210
    },
    {
      "epoch": 90.18,
      "learning_rate": 7.975663430420713e-06,
      "loss": 0.0516,
      "step": 232220
    },
    {
      "epoch": 90.19,
      "learning_rate": 7.975145631067963e-06,
      "loss": 0.0487,
      "step": 232230
    },
    {
      "epoch": 90.19,
      "learning_rate": 7.974627831715212e-06,
      "loss": 0.0098,
      "step": 232240
    },
    {
      "epoch": 90.19,
      "learning_rate": 7.97411003236246e-06,
      "loss": 0.1611,
      "step": 232250
    },
    {
      "epoch": 90.2,
      "learning_rate": 7.97359223300971e-06,
      "loss": 0.0523,
      "step": 232260
    },
    {
      "epoch": 90.2,
      "learning_rate": 7.97307443365696e-06,
      "loss": 0.1814,
      "step": 232270
    },
    {
      "epoch": 90.21,
      "learning_rate": 7.972556634304209e-06,
      "loss": 0.1037,
      "step": 232280
    },
    {
      "epoch": 90.21,
      "learning_rate": 7.972038834951457e-06,
      "loss": 0.0167,
      "step": 232290
    },
    {
      "epoch": 90.21,
      "learning_rate": 7.971521035598707e-06,
      "loss": 0.0369,
      "step": 232300
    },
    {
      "epoch": 90.22,
      "learning_rate": 7.971003236245956e-06,
      "loss": 0.0041,
      "step": 232310
    },
    {
      "epoch": 90.22,
      "learning_rate": 7.970485436893204e-06,
      "loss": 0.0922,
      "step": 232320
    },
    {
      "epoch": 90.23,
      "learning_rate": 7.969967637540454e-06,
      "loss": 0.0347,
      "step": 232330
    },
    {
      "epoch": 90.23,
      "learning_rate": 7.969449838187703e-06,
      "loss": 0.0984,
      "step": 232340
    },
    {
      "epoch": 90.23,
      "learning_rate": 7.968932038834953e-06,
      "loss": 0.0091,
      "step": 232350
    },
    {
      "epoch": 90.24,
      "learning_rate": 7.968414239482201e-06,
      "loss": 0.0831,
      "step": 232360
    },
    {
      "epoch": 90.24,
      "learning_rate": 7.96789644012945e-06,
      "loss": 0.0019,
      "step": 232370
    },
    {
      "epoch": 90.24,
      "learning_rate": 7.9673786407767e-06,
      "loss": 0.0018,
      "step": 232380
    },
    {
      "epoch": 90.25,
      "learning_rate": 7.96686084142395e-06,
      "loss": 0.0483,
      "step": 232390
    },
    {
      "epoch": 90.25,
      "learning_rate": 7.966343042071198e-06,
      "loss": 0.0976,
      "step": 232400
    },
    {
      "epoch": 90.26,
      "learning_rate": 7.965825242718447e-06,
      "loss": 0.1454,
      "step": 232410
    },
    {
      "epoch": 90.26,
      "learning_rate": 7.965307443365697e-06,
      "loss": 0.0024,
      "step": 232420
    },
    {
      "epoch": 90.26,
      "learning_rate": 7.964789644012947e-06,
      "loss": 0.035,
      "step": 232430
    },
    {
      "epoch": 90.27,
      "learning_rate": 7.964271844660194e-06,
      "loss": 0.0155,
      "step": 232440
    },
    {
      "epoch": 90.27,
      "learning_rate": 7.963754045307444e-06,
      "loss": 0.0828,
      "step": 232450
    },
    {
      "epoch": 90.28,
      "learning_rate": 7.963236245954694e-06,
      "loss": 0.0098,
      "step": 232460
    },
    {
      "epoch": 90.28,
      "learning_rate": 7.962718446601942e-06,
      "loss": 0.0362,
      "step": 232470
    },
    {
      "epoch": 90.28,
      "learning_rate": 7.962200647249191e-06,
      "loss": 0.0009,
      "step": 232480
    },
    {
      "epoch": 90.29,
      "learning_rate": 7.961682847896441e-06,
      "loss": 0.0541,
      "step": 232490
    },
    {
      "epoch": 90.29,
      "learning_rate": 7.96116504854369e-06,
      "loss": 0.1354,
      "step": 232500
    },
    {
      "epoch": 90.3,
      "learning_rate": 7.960647249190938e-06,
      "loss": 0.1718,
      "step": 232510
    },
    {
      "epoch": 90.3,
      "learning_rate": 7.960129449838188e-06,
      "loss": 0.0984,
      "step": 232520
    },
    {
      "epoch": 90.3,
      "learning_rate": 7.959611650485438e-06,
      "loss": 0.0637,
      "step": 232530
    },
    {
      "epoch": 90.31,
      "learning_rate": 7.959093851132687e-06,
      "loss": 0.0285,
      "step": 232540
    },
    {
      "epoch": 90.31,
      "learning_rate": 7.958576051779935e-06,
      "loss": 0.056,
      "step": 232550
    },
    {
      "epoch": 90.31,
      "learning_rate": 7.958058252427185e-06,
      "loss": 0.0142,
      "step": 232560
    },
    {
      "epoch": 90.32,
      "learning_rate": 7.957540453074434e-06,
      "loss": 0.1653,
      "step": 232570
    },
    {
      "epoch": 90.32,
      "learning_rate": 7.957022653721684e-06,
      "loss": 0.0126,
      "step": 232580
    },
    {
      "epoch": 90.33,
      "learning_rate": 7.956504854368932e-06,
      "loss": 0.0867,
      "step": 232590
    },
    {
      "epoch": 90.33,
      "learning_rate": 7.955987055016182e-06,
      "loss": 0.0593,
      "step": 232600
    },
    {
      "epoch": 90.33,
      "learning_rate": 7.955469255663431e-06,
      "loss": 0.0466,
      "step": 232610
    },
    {
      "epoch": 90.34,
      "learning_rate": 7.95495145631068e-06,
      "loss": 0.0513,
      "step": 232620
    },
    {
      "epoch": 90.34,
      "learning_rate": 7.954433656957929e-06,
      "loss": 0.1138,
      "step": 232630
    },
    {
      "epoch": 90.35,
      "learning_rate": 7.953915857605178e-06,
      "loss": 0.1154,
      "step": 232640
    },
    {
      "epoch": 90.35,
      "learning_rate": 7.953398058252428e-06,
      "loss": 0.055,
      "step": 232650
    },
    {
      "epoch": 90.35,
      "learning_rate": 7.952880258899676e-06,
      "loss": 0.0941,
      "step": 232660
    },
    {
      "epoch": 90.36,
      "learning_rate": 7.952362459546926e-06,
      "loss": 0.0832,
      "step": 232670
    },
    {
      "epoch": 90.36,
      "learning_rate": 7.951844660194175e-06,
      "loss": 0.0283,
      "step": 232680
    },
    {
      "epoch": 90.37,
      "learning_rate": 7.951326860841425e-06,
      "loss": 0.0097,
      "step": 232690
    },
    {
      "epoch": 90.37,
      "learning_rate": 7.950809061488673e-06,
      "loss": 0.1393,
      "step": 232700
    },
    {
      "epoch": 90.37,
      "learning_rate": 7.950291262135922e-06,
      "loss": 0.0852,
      "step": 232710
    },
    {
      "epoch": 90.38,
      "learning_rate": 7.949773462783172e-06,
      "loss": 0.1222,
      "step": 232720
    },
    {
      "epoch": 90.38,
      "learning_rate": 7.949255663430422e-06,
      "loss": 0.0431,
      "step": 232730
    },
    {
      "epoch": 90.38,
      "learning_rate": 7.94873786407767e-06,
      "loss": 0.0396,
      "step": 232740
    },
    {
      "epoch": 90.39,
      "learning_rate": 7.948220064724919e-06,
      "loss": 0.0193,
      "step": 232750
    },
    {
      "epoch": 90.39,
      "learning_rate": 7.947702265372169e-06,
      "loss": 0.1236,
      "step": 232760
    },
    {
      "epoch": 90.4,
      "learning_rate": 7.947184466019418e-06,
      "loss": 0.0508,
      "step": 232770
    },
    {
      "epoch": 90.4,
      "learning_rate": 7.946666666666666e-06,
      "loss": 0.0868,
      "step": 232780
    },
    {
      "epoch": 90.4,
      "learning_rate": 7.946148867313916e-06,
      "loss": 0.0045,
      "step": 232790
    },
    {
      "epoch": 90.41,
      "learning_rate": 7.945631067961165e-06,
      "loss": 0.0512,
      "step": 232800
    },
    {
      "epoch": 90.41,
      "learning_rate": 7.945113268608415e-06,
      "loss": 0.092,
      "step": 232810
    },
    {
      "epoch": 90.42,
      "learning_rate": 7.944595469255663e-06,
      "loss": 0.0948,
      "step": 232820
    },
    {
      "epoch": 90.42,
      "learning_rate": 7.944077669902913e-06,
      "loss": 0.0693,
      "step": 232830
    },
    {
      "epoch": 90.42,
      "learning_rate": 7.943559870550162e-06,
      "loss": 0.1007,
      "step": 232840
    },
    {
      "epoch": 90.43,
      "learning_rate": 7.943042071197412e-06,
      "loss": 0.0309,
      "step": 232850
    },
    {
      "epoch": 90.43,
      "learning_rate": 7.94252427184466e-06,
      "loss": 0.0102,
      "step": 232860
    },
    {
      "epoch": 90.43,
      "learning_rate": 7.94200647249191e-06,
      "loss": 0.2096,
      "step": 232870
    },
    {
      "epoch": 90.44,
      "learning_rate": 7.941488673139159e-06,
      "loss": 0.0824,
      "step": 232880
    },
    {
      "epoch": 90.44,
      "learning_rate": 7.940970873786409e-06,
      "loss": 0.0004,
      "step": 232890
    },
    {
      "epoch": 90.45,
      "learning_rate": 7.940453074433657e-06,
      "loss": 0.0632,
      "step": 232900
    },
    {
      "epoch": 90.45,
      "learning_rate": 7.939935275080906e-06,
      "loss": 0.1391,
      "step": 232910
    },
    {
      "epoch": 90.45,
      "learning_rate": 7.939417475728156e-06,
      "loss": 0.1039,
      "step": 232920
    },
    {
      "epoch": 90.46,
      "learning_rate": 7.938899676375405e-06,
      "loss": 0.0213,
      "step": 232930
    },
    {
      "epoch": 90.46,
      "learning_rate": 7.938381877022653e-06,
      "loss": 0.0137,
      "step": 232940
    },
    {
      "epoch": 90.47,
      "learning_rate": 7.937864077669903e-06,
      "loss": 0.0243,
      "step": 232950
    },
    {
      "epoch": 90.47,
      "learning_rate": 7.937346278317153e-06,
      "loss": 0.0339,
      "step": 232960
    },
    {
      "epoch": 90.47,
      "learning_rate": 7.936828478964402e-06,
      "loss": 0.0369,
      "step": 232970
    },
    {
      "epoch": 90.48,
      "learning_rate": 7.93631067961165e-06,
      "loss": 0.065,
      "step": 232980
    },
    {
      "epoch": 90.48,
      "learning_rate": 7.9357928802589e-06,
      "loss": 0.1356,
      "step": 232990
    },
    {
      "epoch": 90.49,
      "learning_rate": 7.93527508090615e-06,
      "loss": 0.0599,
      "step": 233000
    },
    {
      "epoch": 90.49,
      "learning_rate": 7.934757281553399e-06,
      "loss": 0.0183,
      "step": 233010
    },
    {
      "epoch": 90.49,
      "learning_rate": 7.934239482200647e-06,
      "loss": 0.0926,
      "step": 233020
    },
    {
      "epoch": 90.5,
      "learning_rate": 7.933721682847897e-06,
      "loss": 0.024,
      "step": 233030
    },
    {
      "epoch": 90.5,
      "learning_rate": 7.933203883495146e-06,
      "loss": 0.0777,
      "step": 233040
    },
    {
      "epoch": 90.5,
      "learning_rate": 7.932686084142396e-06,
      "loss": 0.1844,
      "step": 233050
    },
    {
      "epoch": 90.51,
      "learning_rate": 7.932168284789644e-06,
      "loss": 0.0158,
      "step": 233060
    },
    {
      "epoch": 90.51,
      "learning_rate": 7.931650485436893e-06,
      "loss": 0.1348,
      "step": 233070
    },
    {
      "epoch": 90.52,
      "learning_rate": 7.931132686084143e-06,
      "loss": 0.0998,
      "step": 233080
    },
    {
      "epoch": 90.52,
      "learning_rate": 7.930614886731393e-06,
      "loss": 0.174,
      "step": 233090
    },
    {
      "epoch": 90.52,
      "learning_rate": 7.93009708737864e-06,
      "loss": 0.1779,
      "step": 233100
    },
    {
      "epoch": 90.53,
      "learning_rate": 7.92957928802589e-06,
      "loss": 0.065,
      "step": 233110
    },
    {
      "epoch": 90.53,
      "learning_rate": 7.92906148867314e-06,
      "loss": 0.0804,
      "step": 233120
    },
    {
      "epoch": 90.54,
      "learning_rate": 7.92854368932039e-06,
      "loss": 0.0002,
      "step": 233130
    },
    {
      "epoch": 90.54,
      "learning_rate": 7.928025889967637e-06,
      "loss": 0.0404,
      "step": 233140
    },
    {
      "epoch": 90.54,
      "learning_rate": 7.927508090614887e-06,
      "loss": 0.0141,
      "step": 233150
    },
    {
      "epoch": 90.55,
      "learning_rate": 7.926990291262136e-06,
      "loss": 0.0448,
      "step": 233160
    },
    {
      "epoch": 90.55,
      "learning_rate": 7.926472491909386e-06,
      "loss": 0.0111,
      "step": 233170
    },
    {
      "epoch": 90.56,
      "learning_rate": 7.925954692556634e-06,
      "loss": 0.0385,
      "step": 233180
    },
    {
      "epoch": 90.56,
      "learning_rate": 7.925436893203884e-06,
      "loss": 0.1138,
      "step": 233190
    },
    {
      "epoch": 90.56,
      "learning_rate": 7.924919093851133e-06,
      "loss": 0.0517,
      "step": 233200
    },
    {
      "epoch": 90.57,
      "learning_rate": 7.924401294498383e-06,
      "loss": 0.0461,
      "step": 233210
    },
    {
      "epoch": 90.57,
      "learning_rate": 7.923883495145633e-06,
      "loss": 0.0427,
      "step": 233220
    },
    {
      "epoch": 90.57,
      "learning_rate": 7.92336569579288e-06,
      "loss": 0.0125,
      "step": 233230
    },
    {
      "epoch": 90.58,
      "learning_rate": 7.92284789644013e-06,
      "loss": 0.011,
      "step": 233240
    },
    {
      "epoch": 90.58,
      "learning_rate": 7.92233009708738e-06,
      "loss": 0.0179,
      "step": 233250
    },
    {
      "epoch": 90.59,
      "learning_rate": 7.92181229773463e-06,
      "loss": 0.0629,
      "step": 233260
    },
    {
      "epoch": 90.59,
      "learning_rate": 7.921294498381877e-06,
      "loss": 0.093,
      "step": 233270
    },
    {
      "epoch": 90.59,
      "learning_rate": 7.920776699029127e-06,
      "loss": 0.0011,
      "step": 233280
    },
    {
      "epoch": 90.6,
      "learning_rate": 7.920258899676376e-06,
      "loss": 0.0062,
      "step": 233290
    },
    {
      "epoch": 90.6,
      "learning_rate": 7.919741100323626e-06,
      "loss": 0.1268,
      "step": 233300
    },
    {
      "epoch": 90.61,
      "learning_rate": 7.919223300970874e-06,
      "loss": 0.1453,
      "step": 233310
    },
    {
      "epoch": 90.61,
      "learning_rate": 7.918705501618124e-06,
      "loss": 0.0524,
      "step": 233320
    },
    {
      "epoch": 90.61,
      "learning_rate": 7.918187702265373e-06,
      "loss": 0.1196,
      "step": 233330
    },
    {
      "epoch": 90.62,
      "learning_rate": 7.917669902912623e-06,
      "loss": 0.1316,
      "step": 233340
    },
    {
      "epoch": 90.62,
      "learning_rate": 7.91715210355987e-06,
      "loss": 0.0514,
      "step": 233350
    },
    {
      "epoch": 90.63,
      "learning_rate": 7.91663430420712e-06,
      "loss": 0.0199,
      "step": 233360
    },
    {
      "epoch": 90.63,
      "learning_rate": 7.91611650485437e-06,
      "loss": 0.0511,
      "step": 233370
    },
    {
      "epoch": 90.63,
      "learning_rate": 7.91559870550162e-06,
      "loss": 0.0747,
      "step": 233380
    },
    {
      "epoch": 90.64,
      "learning_rate": 7.915080906148868e-06,
      "loss": 0.0072,
      "step": 233390
    },
    {
      "epoch": 90.64,
      "learning_rate": 7.914563106796117e-06,
      "loss": 0.0853,
      "step": 233400
    },
    {
      "epoch": 90.64,
      "learning_rate": 7.914045307443367e-06,
      "loss": 0.0454,
      "step": 233410
    },
    {
      "epoch": 90.65,
      "learning_rate": 7.913527508090616e-06,
      "loss": 0.0026,
      "step": 233420
    },
    {
      "epoch": 90.65,
      "learning_rate": 7.913009708737864e-06,
      "loss": 0.0604,
      "step": 233430
    },
    {
      "epoch": 90.66,
      "learning_rate": 7.912491909385114e-06,
      "loss": 0.0195,
      "step": 233440
    },
    {
      "epoch": 90.66,
      "learning_rate": 7.911974110032364e-06,
      "loss": 0.0524,
      "step": 233450
    },
    {
      "epoch": 90.66,
      "learning_rate": 7.911456310679613e-06,
      "loss": 0.0788,
      "step": 233460
    },
    {
      "epoch": 90.67,
      "learning_rate": 7.910938511326861e-06,
      "loss": 0.0387,
      "step": 233470
    },
    {
      "epoch": 90.67,
      "learning_rate": 7.91042071197411e-06,
      "loss": 0.0015,
      "step": 233480
    },
    {
      "epoch": 90.68,
      "learning_rate": 7.90990291262136e-06,
      "loss": 0.1358,
      "step": 233490
    },
    {
      "epoch": 90.68,
      "learning_rate": 7.90938511326861e-06,
      "loss": 0.0027,
      "step": 233500
    },
    {
      "epoch": 90.68,
      "learning_rate": 7.908867313915858e-06,
      "loss": 0.0066,
      "step": 233510
    },
    {
      "epoch": 90.69,
      "learning_rate": 7.908349514563107e-06,
      "loss": 0.0179,
      "step": 233520
    },
    {
      "epoch": 90.69,
      "learning_rate": 7.907831715210357e-06,
      "loss": 0.1197,
      "step": 233530
    },
    {
      "epoch": 90.7,
      "learning_rate": 7.907313915857607e-06,
      "loss": 0.0627,
      "step": 233540
    },
    {
      "epoch": 90.7,
      "learning_rate": 7.906796116504855e-06,
      "loss": 0.0825,
      "step": 233550
    },
    {
      "epoch": 90.7,
      "learning_rate": 7.906278317152104e-06,
      "loss": 0.0233,
      "step": 233560
    },
    {
      "epoch": 90.71,
      "learning_rate": 7.905760517799354e-06,
      "loss": 0.0152,
      "step": 233570
    },
    {
      "epoch": 90.71,
      "learning_rate": 7.905242718446604e-06,
      "loss": 0.0442,
      "step": 233580
    },
    {
      "epoch": 90.71,
      "learning_rate": 7.904724919093851e-06,
      "loss": 0.0142,
      "step": 233590
    },
    {
      "epoch": 90.72,
      "learning_rate": 7.904207119741101e-06,
      "loss": 0.0066,
      "step": 233600
    },
    {
      "epoch": 90.72,
      "learning_rate": 7.90368932038835e-06,
      "loss": 0.0029,
      "step": 233610
    },
    {
      "epoch": 90.73,
      "learning_rate": 7.9031715210356e-06,
      "loss": 0.0238,
      "step": 233620
    },
    {
      "epoch": 90.73,
      "learning_rate": 7.902653721682848e-06,
      "loss": 0.01,
      "step": 233630
    },
    {
      "epoch": 90.73,
      "learning_rate": 7.902135922330098e-06,
      "loss": 0.079,
      "step": 233640
    },
    {
      "epoch": 90.74,
      "learning_rate": 7.901618122977347e-06,
      "loss": 0.0138,
      "step": 233650
    },
    {
      "epoch": 90.74,
      "learning_rate": 7.901100323624597e-06,
      "loss": 0.0098,
      "step": 233660
    },
    {
      "epoch": 90.75,
      "learning_rate": 7.900582524271845e-06,
      "loss": 0.0187,
      "step": 233670
    },
    {
      "epoch": 90.75,
      "learning_rate": 7.900064724919095e-06,
      "loss": 0.0643,
      "step": 233680
    },
    {
      "epoch": 90.75,
      "learning_rate": 7.899546925566344e-06,
      "loss": 0.0178,
      "step": 233690
    },
    {
      "epoch": 90.76,
      "learning_rate": 7.899029126213594e-06,
      "loss": 0.0315,
      "step": 233700
    },
    {
      "epoch": 90.76,
      "learning_rate": 7.898511326860842e-06,
      "loss": 0.0776,
      "step": 233710
    },
    {
      "epoch": 90.77,
      "learning_rate": 7.897993527508091e-06,
      "loss": 0.0192,
      "step": 233720
    },
    {
      "epoch": 90.77,
      "learning_rate": 7.897475728155341e-06,
      "loss": 0.1483,
      "step": 233730
    },
    {
      "epoch": 90.77,
      "learning_rate": 7.89695792880259e-06,
      "loss": 0.0697,
      "step": 233740
    },
    {
      "epoch": 90.78,
      "learning_rate": 7.896440129449839e-06,
      "loss": 0.0947,
      "step": 233750
    },
    {
      "epoch": 90.78,
      "learning_rate": 7.895922330097088e-06,
      "loss": 0.0592,
      "step": 233760
    },
    {
      "epoch": 90.78,
      "learning_rate": 7.895404530744338e-06,
      "loss": 0.0155,
      "step": 233770
    },
    {
      "epoch": 90.79,
      "learning_rate": 7.894886731391587e-06,
      "loss": 0.0361,
      "step": 233780
    },
    {
      "epoch": 90.79,
      "learning_rate": 7.894368932038835e-06,
      "loss": 0.0222,
      "step": 233790
    },
    {
      "epoch": 90.8,
      "learning_rate": 7.893851132686085e-06,
      "loss": 0.0937,
      "step": 233800
    },
    {
      "epoch": 90.8,
      "learning_rate": 7.893333333333335e-06,
      "loss": 0.0578,
      "step": 233810
    },
    {
      "epoch": 90.8,
      "learning_rate": 7.892815533980584e-06,
      "loss": 0.1153,
      "step": 233820
    },
    {
      "epoch": 90.81,
      "learning_rate": 7.892297734627832e-06,
      "loss": 0.0719,
      "step": 233830
    },
    {
      "epoch": 90.81,
      "learning_rate": 7.891779935275082e-06,
      "loss": 0.0211,
      "step": 233840
    },
    {
      "epoch": 90.82,
      "learning_rate": 7.891262135922331e-06,
      "loss": 0.0097,
      "step": 233850
    },
    {
      "epoch": 90.82,
      "learning_rate": 7.890744336569581e-06,
      "loss": 0.0765,
      "step": 233860
    },
    {
      "epoch": 90.82,
      "learning_rate": 7.890226537216829e-06,
      "loss": 0.0325,
      "step": 233870
    },
    {
      "epoch": 90.83,
      "learning_rate": 7.889708737864078e-06,
      "loss": 0.0588,
      "step": 233880
    },
    {
      "epoch": 90.83,
      "learning_rate": 7.889190938511328e-06,
      "loss": 0.0109,
      "step": 233890
    },
    {
      "epoch": 90.83,
      "learning_rate": 7.888673139158578e-06,
      "loss": 0.0088,
      "step": 233900
    },
    {
      "epoch": 90.84,
      "learning_rate": 7.888155339805826e-06,
      "loss": 0.0055,
      "step": 233910
    },
    {
      "epoch": 90.84,
      "learning_rate": 7.887637540453075e-06,
      "loss": 0.1153,
      "step": 233920
    },
    {
      "epoch": 90.85,
      "learning_rate": 7.887119741100325e-06,
      "loss": 0.0407,
      "step": 233930
    },
    {
      "epoch": 90.85,
      "learning_rate": 7.886601941747573e-06,
      "loss": 0.0249,
      "step": 233940
    },
    {
      "epoch": 90.85,
      "learning_rate": 7.886084142394822e-06,
      "loss": 0.0638,
      "step": 233950
    },
    {
      "epoch": 90.86,
      "learning_rate": 7.885566343042072e-06,
      "loss": 0.1586,
      "step": 233960
    },
    {
      "epoch": 90.86,
      "learning_rate": 7.885048543689322e-06,
      "loss": 0.0523,
      "step": 233970
    },
    {
      "epoch": 90.87,
      "learning_rate": 7.88453074433657e-06,
      "loss": 0.0003,
      "step": 233980
    },
    {
      "epoch": 90.87,
      "learning_rate": 7.88401294498382e-06,
      "loss": 0.0433,
      "step": 233990
    },
    {
      "epoch": 90.87,
      "learning_rate": 7.883495145631069e-06,
      "loss": 0.009,
      "step": 234000
    },
    {
      "epoch": 90.88,
      "learning_rate": 7.882977346278318e-06,
      "loss": 0.0981,
      "step": 234010
    },
    {
      "epoch": 90.88,
      "learning_rate": 7.882459546925566e-06,
      "loss": 0.0218,
      "step": 234020
    },
    {
      "epoch": 90.89,
      "learning_rate": 7.881941747572816e-06,
      "loss": 0.0033,
      "step": 234030
    },
    {
      "epoch": 90.89,
      "learning_rate": 7.881423948220066e-06,
      "loss": 0.0044,
      "step": 234040
    },
    {
      "epoch": 90.89,
      "learning_rate": 7.880906148867315e-06,
      "loss": 0.093,
      "step": 234050
    },
    {
      "epoch": 90.9,
      "learning_rate": 7.880388349514563e-06,
      "loss": 0.0289,
      "step": 234060
    },
    {
      "epoch": 90.9,
      "learning_rate": 7.879870550161813e-06,
      "loss": 0.0832,
      "step": 234070
    },
    {
      "epoch": 90.9,
      "learning_rate": 7.879352750809062e-06,
      "loss": 0.0789,
      "step": 234080
    },
    {
      "epoch": 90.91,
      "learning_rate": 7.878834951456312e-06,
      "loss": 0.0197,
      "step": 234090
    },
    {
      "epoch": 90.91,
      "learning_rate": 7.87831715210356e-06,
      "loss": 0.0791,
      "step": 234100
    },
    {
      "epoch": 90.92,
      "learning_rate": 7.87779935275081e-06,
      "loss": 0.097,
      "step": 234110
    },
    {
      "epoch": 90.92,
      "learning_rate": 7.87728155339806e-06,
      "loss": 0.0159,
      "step": 234120
    },
    {
      "epoch": 90.92,
      "learning_rate": 7.876763754045307e-06,
      "loss": 0.0304,
      "step": 234130
    },
    {
      "epoch": 90.93,
      "learning_rate": 7.876245954692557e-06,
      "loss": 0.1623,
      "step": 234140
    },
    {
      "epoch": 90.93,
      "learning_rate": 7.875728155339806e-06,
      "loss": 0.0429,
      "step": 234150
    },
    {
      "epoch": 90.94,
      "learning_rate": 7.875210355987056e-06,
      "loss": 0.0159,
      "step": 234160
    },
    {
      "epoch": 90.94,
      "learning_rate": 7.874692556634304e-06,
      "loss": 0.0242,
      "step": 234170
    },
    {
      "epoch": 90.94,
      "learning_rate": 7.874174757281553e-06,
      "loss": 0.0353,
      "step": 234180
    },
    {
      "epoch": 90.95,
      "learning_rate": 7.873656957928803e-06,
      "loss": 0.1684,
      "step": 234190
    },
    {
      "epoch": 90.95,
      "learning_rate": 7.873139158576053e-06,
      "loss": 0.0683,
      "step": 234200
    },
    {
      "epoch": 90.96,
      "learning_rate": 7.8726213592233e-06,
      "loss": 0.0759,
      "step": 234210
    },
    {
      "epoch": 90.96,
      "learning_rate": 7.87210355987055e-06,
      "loss": 0.0514,
      "step": 234220
    },
    {
      "epoch": 90.96,
      "learning_rate": 7.8715857605178e-06,
      "loss": 0.1981,
      "step": 234230
    },
    {
      "epoch": 90.97,
      "learning_rate": 7.87106796116505e-06,
      "loss": 0.0504,
      "step": 234240
    },
    {
      "epoch": 90.97,
      "learning_rate": 7.870550161812297e-06,
      "loss": 0.1595,
      "step": 234250
    },
    {
      "epoch": 90.97,
      "learning_rate": 7.870032362459547e-06,
      "loss": 0.0851,
      "step": 234260
    },
    {
      "epoch": 90.98,
      "learning_rate": 7.869514563106797e-06,
      "loss": 0.002,
      "step": 234270
    },
    {
      "epoch": 90.98,
      "learning_rate": 7.868996763754045e-06,
      "loss": 0.0181,
      "step": 234280
    },
    {
      "epoch": 90.99,
      "learning_rate": 7.868478964401294e-06,
      "loss": 0.0005,
      "step": 234290
    },
    {
      "epoch": 90.99,
      "learning_rate": 7.867961165048544e-06,
      "loss": 0.0215,
      "step": 234300
    },
    {
      "epoch": 90.99,
      "learning_rate": 7.867443365695793e-06,
      "loss": 0.0274,
      "step": 234310
    },
    {
      "epoch": 91.0,
      "learning_rate": 7.866925566343041e-06,
      "loss": 0.0477,
      "step": 234320
    },
    {
      "epoch": 91.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.3678571879863739,
      "eval_runtime": 8.2007,
      "eval_samples_per_second": 443.255,
      "eval_steps_per_second": 55.483,
      "step": 234325
    },
    {
      "epoch": 91.0,
      "learning_rate": 7.866407766990291e-06,
      "loss": 0.1098,
      "step": 234330
    },
    {
      "epoch": 91.01,
      "learning_rate": 7.86588996763754e-06,
      "loss": 0.0449,
      "step": 234340
    },
    {
      "epoch": 91.01,
      "learning_rate": 7.86537216828479e-06,
      "loss": 0.0774,
      "step": 234350
    },
    {
      "epoch": 91.01,
      "learning_rate": 7.86485436893204e-06,
      "loss": 0.0132,
      "step": 234360
    },
    {
      "epoch": 91.02,
      "learning_rate": 7.864336569579288e-06,
      "loss": 0.0546,
      "step": 234370
    },
    {
      "epoch": 91.02,
      "learning_rate": 7.863818770226537e-06,
      "loss": 0.0267,
      "step": 234380
    },
    {
      "epoch": 91.03,
      "learning_rate": 7.863300970873787e-06,
      "loss": 0.0015,
      "step": 234390
    },
    {
      "epoch": 91.03,
      "learning_rate": 7.862783171521037e-06,
      "loss": 0.0046,
      "step": 234400
    },
    {
      "epoch": 91.03,
      "learning_rate": 7.862265372168285e-06,
      "loss": 0.0043,
      "step": 234410
    },
    {
      "epoch": 91.04,
      "learning_rate": 7.861747572815534e-06,
      "loss": 0.041,
      "step": 234420
    },
    {
      "epoch": 91.04,
      "learning_rate": 7.861229773462784e-06,
      "loss": 0.058,
      "step": 234430
    },
    {
      "epoch": 91.04,
      "learning_rate": 7.860711974110033e-06,
      "loss": 0.0292,
      "step": 234440
    },
    {
      "epoch": 91.05,
      "learning_rate": 7.860194174757281e-06,
      "loss": 0.0813,
      "step": 234450
    },
    {
      "epoch": 91.05,
      "learning_rate": 7.859676375404531e-06,
      "loss": 0.0777,
      "step": 234460
    },
    {
      "epoch": 91.06,
      "learning_rate": 7.85915857605178e-06,
      "loss": 0.0607,
      "step": 234470
    },
    {
      "epoch": 91.06,
      "learning_rate": 7.85864077669903e-06,
      "loss": 0.0013,
      "step": 234480
    },
    {
      "epoch": 91.06,
      "learning_rate": 7.858122977346278e-06,
      "loss": 0.1373,
      "step": 234490
    },
    {
      "epoch": 91.07,
      "learning_rate": 7.857605177993528e-06,
      "loss": 0.0005,
      "step": 234500
    },
    {
      "epoch": 91.07,
      "learning_rate": 7.857087378640777e-06,
      "loss": 0.0388,
      "step": 234510
    },
    {
      "epoch": 91.08,
      "learning_rate": 7.856569579288027e-06,
      "loss": 0.0981,
      "step": 234520
    },
    {
      "epoch": 91.08,
      "learning_rate": 7.856051779935275e-06,
      "loss": 0.0509,
      "step": 234530
    },
    {
      "epoch": 91.08,
      "learning_rate": 7.855533980582524e-06,
      "loss": 0.1173,
      "step": 234540
    },
    {
      "epoch": 91.09,
      "learning_rate": 7.855016181229774e-06,
      "loss": 0.0609,
      "step": 234550
    },
    {
      "epoch": 91.09,
      "learning_rate": 7.854498381877024e-06,
      "loss": 0.0013,
      "step": 234560
    },
    {
      "epoch": 91.1,
      "learning_rate": 7.853980582524272e-06,
      "loss": 0.0639,
      "step": 234570
    },
    {
      "epoch": 91.1,
      "learning_rate": 7.853462783171521e-06,
      "loss": 0.0321,
      "step": 234580
    },
    {
      "epoch": 91.1,
      "learning_rate": 7.852944983818771e-06,
      "loss": 0.026,
      "step": 234590
    },
    {
      "epoch": 91.11,
      "learning_rate": 7.85242718446602e-06,
      "loss": 0.1283,
      "step": 234600
    },
    {
      "epoch": 91.11,
      "learning_rate": 7.851909385113268e-06,
      "loss": 0.0572,
      "step": 234610
    },
    {
      "epoch": 91.11,
      "learning_rate": 7.851391585760518e-06,
      "loss": 0.1822,
      "step": 234620
    },
    {
      "epoch": 91.12,
      "learning_rate": 7.850873786407768e-06,
      "loss": 0.065,
      "step": 234630
    },
    {
      "epoch": 91.12,
      "learning_rate": 7.850355987055017e-06,
      "loss": 0.0147,
      "step": 234640
    },
    {
      "epoch": 91.13,
      "learning_rate": 7.849838187702265e-06,
      "loss": 0.0398,
      "step": 234650
    },
    {
      "epoch": 91.13,
      "learning_rate": 7.849320388349515e-06,
      "loss": 0.0007,
      "step": 234660
    },
    {
      "epoch": 91.13,
      "learning_rate": 7.848802588996764e-06,
      "loss": 0.0755,
      "step": 234670
    },
    {
      "epoch": 91.14,
      "learning_rate": 7.848284789644014e-06,
      "loss": 0.0019,
      "step": 234680
    },
    {
      "epoch": 91.14,
      "learning_rate": 7.847766990291262e-06,
      "loss": 0.0407,
      "step": 234690
    },
    {
      "epoch": 91.15,
      "learning_rate": 7.847249190938512e-06,
      "loss": 0.1187,
      "step": 234700
    },
    {
      "epoch": 91.15,
      "learning_rate": 7.846731391585761e-06,
      "loss": 0.0034,
      "step": 234710
    },
    {
      "epoch": 91.15,
      "learning_rate": 7.84621359223301e-06,
      "loss": 0.0488,
      "step": 234720
    },
    {
      "epoch": 91.16,
      "learning_rate": 7.845695792880259e-06,
      "loss": 0.0922,
      "step": 234730
    },
    {
      "epoch": 91.16,
      "learning_rate": 7.845177993527508e-06,
      "loss": 0.0194,
      "step": 234740
    },
    {
      "epoch": 91.17,
      "learning_rate": 7.844660194174758e-06,
      "loss": 0.0493,
      "step": 234750
    },
    {
      "epoch": 91.17,
      "learning_rate": 7.844142394822008e-06,
      "loss": 0.1157,
      "step": 234760
    },
    {
      "epoch": 91.17,
      "learning_rate": 7.843624595469256e-06,
      "loss": 0.0457,
      "step": 234770
    },
    {
      "epoch": 91.18,
      "learning_rate": 7.843106796116505e-06,
      "loss": 0.0093,
      "step": 234780
    },
    {
      "epoch": 91.18,
      "learning_rate": 7.842588996763755e-06,
      "loss": 0.0649,
      "step": 234790
    },
    {
      "epoch": 91.18,
      "learning_rate": 7.842071197411004e-06,
      "loss": 0.0534,
      "step": 234800
    },
    {
      "epoch": 91.19,
      "learning_rate": 7.841553398058252e-06,
      "loss": 0.0053,
      "step": 234810
    },
    {
      "epoch": 91.19,
      "learning_rate": 7.841035598705502e-06,
      "loss": 0.0557,
      "step": 234820
    },
    {
      "epoch": 91.2,
      "learning_rate": 7.840517799352752e-06,
      "loss": 0.0488,
      "step": 234830
    },
    {
      "epoch": 91.2,
      "learning_rate": 7.840000000000001e-06,
      "loss": 0.0292,
      "step": 234840
    },
    {
      "epoch": 91.2,
      "learning_rate": 7.839482200647249e-06,
      "loss": 0.0364,
      "step": 234850
    },
    {
      "epoch": 91.21,
      "learning_rate": 7.838964401294499e-06,
      "loss": 0.008,
      "step": 234860
    },
    {
      "epoch": 91.21,
      "learning_rate": 7.838446601941748e-06,
      "loss": 0.0735,
      "step": 234870
    },
    {
      "epoch": 91.22,
      "learning_rate": 7.837928802588998e-06,
      "loss": 0.0082,
      "step": 234880
    },
    {
      "epoch": 91.22,
      "learning_rate": 7.837411003236248e-06,
      "loss": 0.0728,
      "step": 234890
    },
    {
      "epoch": 91.22,
      "learning_rate": 7.836893203883495e-06,
      "loss": 0.0533,
      "step": 234900
    },
    {
      "epoch": 91.23,
      "learning_rate": 7.836375404530745e-06,
      "loss": 0.0134,
      "step": 234910
    },
    {
      "epoch": 91.23,
      "learning_rate": 7.835857605177995e-06,
      "loss": 0.0008,
      "step": 234920
    },
    {
      "epoch": 91.23,
      "learning_rate": 7.835339805825244e-06,
      "loss": 0.1104,
      "step": 234930
    },
    {
      "epoch": 91.24,
      "learning_rate": 7.834822006472492e-06,
      "loss": 0.0227,
      "step": 234940
    },
    {
      "epoch": 91.24,
      "learning_rate": 7.834304207119742e-06,
      "loss": 0.1052,
      "step": 234950
    },
    {
      "epoch": 91.25,
      "learning_rate": 7.833786407766992e-06,
      "loss": 0.0317,
      "step": 234960
    },
    {
      "epoch": 91.25,
      "learning_rate": 7.833268608414241e-06,
      "loss": 0.0681,
      "step": 234970
    },
    {
      "epoch": 91.25,
      "learning_rate": 7.832750809061489e-06,
      "loss": 0.0239,
      "step": 234980
    },
    {
      "epoch": 91.26,
      "learning_rate": 7.832233009708739e-06,
      "loss": 0.0966,
      "step": 234990
    },
    {
      "epoch": 91.26,
      "learning_rate": 7.831715210355988e-06,
      "loss": 0.0844,
      "step": 235000
    },
    {
      "epoch": 91.27,
      "learning_rate": 7.831197411003238e-06,
      "loss": 0.0048,
      "step": 235010
    },
    {
      "epoch": 91.27,
      "learning_rate": 7.830679611650486e-06,
      "loss": 0.1476,
      "step": 235020
    },
    {
      "epoch": 91.27,
      "learning_rate": 7.830161812297735e-06,
      "loss": 0.0084,
      "step": 235030
    },
    {
      "epoch": 91.28,
      "learning_rate": 7.829644012944985e-06,
      "loss": 0.0074,
      "step": 235040
    },
    {
      "epoch": 91.28,
      "learning_rate": 7.829126213592235e-06,
      "loss": 0.111,
      "step": 235050
    },
    {
      "epoch": 91.29,
      "learning_rate": 7.828608414239483e-06,
      "loss": 0.0579,
      "step": 235060
    },
    {
      "epoch": 91.29,
      "learning_rate": 7.828090614886732e-06,
      "loss": 0.0514,
      "step": 235070
    },
    {
      "epoch": 91.29,
      "learning_rate": 7.827572815533982e-06,
      "loss": 0.0268,
      "step": 235080
    },
    {
      "epoch": 91.3,
      "learning_rate": 7.827055016181231e-06,
      "loss": 0.1087,
      "step": 235090
    },
    {
      "epoch": 91.3,
      "learning_rate": 7.82653721682848e-06,
      "loss": 0.0099,
      "step": 235100
    },
    {
      "epoch": 91.3,
      "learning_rate": 7.826019417475729e-06,
      "loss": 0.0011,
      "step": 235110
    },
    {
      "epoch": 91.31,
      "learning_rate": 7.825501618122979e-06,
      "loss": 0.0427,
      "step": 235120
    },
    {
      "epoch": 91.31,
      "learning_rate": 7.824983818770228e-06,
      "loss": 0.0061,
      "step": 235130
    },
    {
      "epoch": 91.32,
      "learning_rate": 7.824466019417476e-06,
      "loss": 0.0614,
      "step": 235140
    },
    {
      "epoch": 91.32,
      "learning_rate": 7.823948220064726e-06,
      "loss": 0.1145,
      "step": 235150
    },
    {
      "epoch": 91.32,
      "learning_rate": 7.823430420711975e-06,
      "loss": 0.1097,
      "step": 235160
    },
    {
      "epoch": 91.33,
      "learning_rate": 7.822912621359225e-06,
      "loss": 0.0351,
      "step": 235170
    },
    {
      "epoch": 91.33,
      "learning_rate": 7.822394822006473e-06,
      "loss": 0.0164,
      "step": 235180
    },
    {
      "epoch": 91.34,
      "learning_rate": 7.821877022653723e-06,
      "loss": 0.0591,
      "step": 235190
    },
    {
      "epoch": 91.34,
      "learning_rate": 7.821359223300972e-06,
      "loss": 0.088,
      "step": 235200
    },
    {
      "epoch": 91.34,
      "learning_rate": 7.820841423948222e-06,
      "loss": 0.0225,
      "step": 235210
    },
    {
      "epoch": 91.35,
      "learning_rate": 7.82032362459547e-06,
      "loss": 0.0832,
      "step": 235220
    },
    {
      "epoch": 91.35,
      "learning_rate": 7.81980582524272e-06,
      "loss": 0.0174,
      "step": 235230
    },
    {
      "epoch": 91.36,
      "learning_rate": 7.819288025889969e-06,
      "loss": 0.0251,
      "step": 235240
    },
    {
      "epoch": 91.36,
      "learning_rate": 7.818770226537219e-06,
      "loss": 0.1409,
      "step": 235250
    },
    {
      "epoch": 91.36,
      "learning_rate": 7.818252427184466e-06,
      "loss": 0.0008,
      "step": 235260
    },
    {
      "epoch": 91.37,
      "learning_rate": 7.817734627831716e-06,
      "loss": 0.0578,
      "step": 235270
    },
    {
      "epoch": 91.37,
      "learning_rate": 7.817216828478966e-06,
      "loss": 0.0366,
      "step": 235280
    },
    {
      "epoch": 91.37,
      "learning_rate": 7.816699029126215e-06,
      "loss": 0.0253,
      "step": 235290
    },
    {
      "epoch": 91.38,
      "learning_rate": 7.816181229773463e-06,
      "loss": 0.0719,
      "step": 235300
    },
    {
      "epoch": 91.38,
      "learning_rate": 7.815663430420713e-06,
      "loss": 0.0862,
      "step": 235310
    },
    {
      "epoch": 91.39,
      "learning_rate": 7.815145631067963e-06,
      "loss": 0.0604,
      "step": 235320
    },
    {
      "epoch": 91.39,
      "learning_rate": 7.814627831715212e-06,
      "loss": 0.0177,
      "step": 235330
    },
    {
      "epoch": 91.39,
      "learning_rate": 7.81411003236246e-06,
      "loss": 0.0961,
      "step": 235340
    },
    {
      "epoch": 91.4,
      "learning_rate": 7.81359223300971e-06,
      "loss": 0.0002,
      "step": 235350
    },
    {
      "epoch": 91.4,
      "learning_rate": 7.81307443365696e-06,
      "loss": 0.0099,
      "step": 235360
    },
    {
      "epoch": 91.41,
      "learning_rate": 7.812556634304209e-06,
      "loss": 0.2257,
      "step": 235370
    },
    {
      "epoch": 91.41,
      "learning_rate": 7.812038834951457e-06,
      "loss": 0.09,
      "step": 235380
    },
    {
      "epoch": 91.41,
      "learning_rate": 7.811521035598706e-06,
      "loss": 0.0096,
      "step": 235390
    },
    {
      "epoch": 91.42,
      "learning_rate": 7.811003236245956e-06,
      "loss": 0.0094,
      "step": 235400
    },
    {
      "epoch": 91.42,
      "learning_rate": 7.810485436893204e-06,
      "loss": 0.0909,
      "step": 235410
    },
    {
      "epoch": 91.43,
      "learning_rate": 7.809967637540454e-06,
      "loss": 0.1653,
      "step": 235420
    },
    {
      "epoch": 91.43,
      "learning_rate": 7.809449838187703e-06,
      "loss": 0.0482,
      "step": 235430
    },
    {
      "epoch": 91.43,
      "learning_rate": 7.808932038834953e-06,
      "loss": 0.1097,
      "step": 235440
    },
    {
      "epoch": 91.44,
      "learning_rate": 7.8084142394822e-06,
      "loss": 0.095,
      "step": 235450
    },
    {
      "epoch": 91.44,
      "learning_rate": 7.80789644012945e-06,
      "loss": 0.0581,
      "step": 235460
    },
    {
      "epoch": 91.44,
      "learning_rate": 7.8073786407767e-06,
      "loss": 0.0003,
      "step": 235470
    },
    {
      "epoch": 91.45,
      "learning_rate": 7.80686084142395e-06,
      "loss": 0.0029,
      "step": 235480
    },
    {
      "epoch": 91.45,
      "learning_rate": 7.806343042071198e-06,
      "loss": 0.0614,
      "step": 235490
    },
    {
      "epoch": 91.46,
      "learning_rate": 7.805825242718447e-06,
      "loss": 0.02,
      "step": 235500
    },
    {
      "epoch": 91.46,
      "learning_rate": 7.805307443365697e-06,
      "loss": 0.1414,
      "step": 235510
    },
    {
      "epoch": 91.46,
      "learning_rate": 7.804789644012946e-06,
      "loss": 0.1167,
      "step": 235520
    },
    {
      "epoch": 91.47,
      "learning_rate": 7.804271844660194e-06,
      "loss": 0.0198,
      "step": 235530
    },
    {
      "epoch": 91.47,
      "learning_rate": 7.803754045307444e-06,
      "loss": 0.052,
      "step": 235540
    },
    {
      "epoch": 91.48,
      "learning_rate": 7.803236245954694e-06,
      "loss": 0.0034,
      "step": 235550
    },
    {
      "epoch": 91.48,
      "learning_rate": 7.802718446601943e-06,
      "loss": 0.0084,
      "step": 235560
    },
    {
      "epoch": 91.48,
      "learning_rate": 7.802200647249191e-06,
      "loss": 0.0866,
      "step": 235570
    },
    {
      "epoch": 91.49,
      "learning_rate": 7.80168284789644e-06,
      "loss": 0.0012,
      "step": 235580
    },
    {
      "epoch": 91.49,
      "learning_rate": 7.80116504854369e-06,
      "loss": 0.2861,
      "step": 235590
    },
    {
      "epoch": 91.5,
      "learning_rate": 7.800647249190938e-06,
      "loss": 0.0184,
      "step": 235600
    },
    {
      "epoch": 91.5,
      "learning_rate": 7.800129449838188e-06,
      "loss": 0.037,
      "step": 235610
    },
    {
      "epoch": 91.5,
      "learning_rate": 7.799611650485437e-06,
      "loss": 0.0848,
      "step": 235620
    },
    {
      "epoch": 91.51,
      "learning_rate": 7.799093851132687e-06,
      "loss": 0.0222,
      "step": 235630
    },
    {
      "epoch": 91.51,
      "learning_rate": 7.798576051779935e-06,
      "loss": 0.0802,
      "step": 235640
    },
    {
      "epoch": 91.51,
      "learning_rate": 7.798058252427185e-06,
      "loss": 0.0039,
      "step": 235650
    },
    {
      "epoch": 91.52,
      "learning_rate": 7.797540453074434e-06,
      "loss": 0.069,
      "step": 235660
    },
    {
      "epoch": 91.52,
      "learning_rate": 7.797022653721684e-06,
      "loss": 0.0747,
      "step": 235670
    },
    {
      "epoch": 91.53,
      "learning_rate": 7.796504854368932e-06,
      "loss": 0.011,
      "step": 235680
    },
    {
      "epoch": 91.53,
      "learning_rate": 7.795987055016181e-06,
      "loss": 0.0031,
      "step": 235690
    },
    {
      "epoch": 91.53,
      "learning_rate": 7.795469255663431e-06,
      "loss": 0.0034,
      "step": 235700
    },
    {
      "epoch": 91.54,
      "learning_rate": 7.79495145631068e-06,
      "loss": 0.0794,
      "step": 235710
    },
    {
      "epoch": 91.54,
      "learning_rate": 7.794433656957929e-06,
      "loss": 0.0607,
      "step": 235720
    },
    {
      "epoch": 91.55,
      "learning_rate": 7.793915857605178e-06,
      "loss": 0.0003,
      "step": 235730
    },
    {
      "epoch": 91.55,
      "learning_rate": 7.793398058252428e-06,
      "loss": 0.0953,
      "step": 235740
    },
    {
      "epoch": 91.55,
      "learning_rate": 7.792880258899676e-06,
      "loss": 0.0462,
      "step": 235750
    },
    {
      "epoch": 91.56,
      "learning_rate": 7.792362459546925e-06,
      "loss": 0.0929,
      "step": 235760
    },
    {
      "epoch": 91.56,
      "learning_rate": 7.791844660194175e-06,
      "loss": 0.1277,
      "step": 235770
    },
    {
      "epoch": 91.57,
      "learning_rate": 7.791326860841425e-06,
      "loss": 0.1522,
      "step": 235780
    },
    {
      "epoch": 91.57,
      "learning_rate": 7.790809061488673e-06,
      "loss": 0.0239,
      "step": 235790
    },
    {
      "epoch": 91.57,
      "learning_rate": 7.790291262135922e-06,
      "loss": 0.0129,
      "step": 235800
    },
    {
      "epoch": 91.58,
      "learning_rate": 7.789773462783172e-06,
      "loss": 0.0741,
      "step": 235810
    },
    {
      "epoch": 91.58,
      "learning_rate": 7.789255663430421e-06,
      "loss": 0.0841,
      "step": 235820
    },
    {
      "epoch": 91.58,
      "learning_rate": 7.78873786407767e-06,
      "loss": 0.1555,
      "step": 235830
    },
    {
      "epoch": 91.59,
      "learning_rate": 7.788220064724919e-06,
      "loss": 0.0144,
      "step": 235840
    },
    {
      "epoch": 91.59,
      "learning_rate": 7.787702265372169e-06,
      "loss": 0.0493,
      "step": 235850
    },
    {
      "epoch": 91.6,
      "learning_rate": 7.787184466019418e-06,
      "loss": 0.1034,
      "step": 235860
    },
    {
      "epoch": 91.6,
      "learning_rate": 7.786666666666666e-06,
      "loss": 0.0092,
      "step": 235870
    },
    {
      "epoch": 91.6,
      "learning_rate": 7.786148867313916e-06,
      "loss": 0.0513,
      "step": 235880
    },
    {
      "epoch": 91.61,
      "learning_rate": 7.785631067961165e-06,
      "loss": 0.0242,
      "step": 235890
    },
    {
      "epoch": 91.61,
      "learning_rate": 7.785113268608415e-06,
      "loss": 0.0154,
      "step": 235900
    },
    {
      "epoch": 91.62,
      "learning_rate": 7.784595469255663e-06,
      "loss": 0.0001,
      "step": 235910
    },
    {
      "epoch": 91.62,
      "learning_rate": 7.784077669902912e-06,
      "loss": 0.0011,
      "step": 235920
    },
    {
      "epoch": 91.62,
      "learning_rate": 7.783559870550162e-06,
      "loss": 0.138,
      "step": 235930
    },
    {
      "epoch": 91.63,
      "learning_rate": 7.783042071197412e-06,
      "loss": 0.0243,
      "step": 235940
    },
    {
      "epoch": 91.63,
      "learning_rate": 7.78252427184466e-06,
      "loss": 0.1508,
      "step": 235950
    },
    {
      "epoch": 91.63,
      "learning_rate": 7.78200647249191e-06,
      "loss": 0.0194,
      "step": 235960
    },
    {
      "epoch": 91.64,
      "learning_rate": 7.781488673139159e-06,
      "loss": 0.0141,
      "step": 235970
    },
    {
      "epoch": 91.64,
      "learning_rate": 7.780970873786408e-06,
      "loss": 0.0293,
      "step": 235980
    },
    {
      "epoch": 91.65,
      "learning_rate": 7.780453074433656e-06,
      "loss": 0.0333,
      "step": 235990
    },
    {
      "epoch": 91.65,
      "learning_rate": 7.779935275080906e-06,
      "loss": 0.0841,
      "step": 236000
    },
    {
      "epoch": 91.65,
      "learning_rate": 7.779417475728156e-06,
      "loss": 0.0676,
      "step": 236010
    },
    {
      "epoch": 91.66,
      "learning_rate": 7.778899676375405e-06,
      "loss": 0.01,
      "step": 236020
    },
    {
      "epoch": 91.66,
      "learning_rate": 7.778381877022655e-06,
      "loss": 0.063,
      "step": 236030
    },
    {
      "epoch": 91.67,
      "learning_rate": 7.777864077669903e-06,
      "loss": 0.0484,
      "step": 236040
    },
    {
      "epoch": 91.67,
      "learning_rate": 7.777346278317152e-06,
      "loss": 0.0003,
      "step": 236050
    },
    {
      "epoch": 91.67,
      "learning_rate": 7.776828478964402e-06,
      "loss": 0.0007,
      "step": 236060
    },
    {
      "epoch": 91.68,
      "learning_rate": 7.776310679611652e-06,
      "loss": 0.0875,
      "step": 236070
    },
    {
      "epoch": 91.68,
      "learning_rate": 7.7757928802589e-06,
      "loss": 0.0304,
      "step": 236080
    },
    {
      "epoch": 91.69,
      "learning_rate": 7.77527508090615e-06,
      "loss": 0.0834,
      "step": 236090
    },
    {
      "epoch": 91.69,
      "learning_rate": 7.774757281553399e-06,
      "loss": 0.0927,
      "step": 236100
    },
    {
      "epoch": 91.69,
      "learning_rate": 7.774239482200648e-06,
      "loss": 0.0375,
      "step": 236110
    },
    {
      "epoch": 91.7,
      "learning_rate": 7.773721682847896e-06,
      "loss": 0.0039,
      "step": 236120
    },
    {
      "epoch": 91.7,
      "learning_rate": 7.773203883495146e-06,
      "loss": 0.0931,
      "step": 236130
    },
    {
      "epoch": 91.7,
      "learning_rate": 7.772686084142396e-06,
      "loss": 0.1058,
      "step": 236140
    },
    {
      "epoch": 91.71,
      "learning_rate": 7.772168284789645e-06,
      "loss": 0.0066,
      "step": 236150
    },
    {
      "epoch": 91.71,
      "learning_rate": 7.771650485436893e-06,
      "loss": 0.0017,
      "step": 236160
    },
    {
      "epoch": 91.72,
      "learning_rate": 7.771132686084143e-06,
      "loss": 0.0951,
      "step": 236170
    },
    {
      "epoch": 91.72,
      "learning_rate": 7.770614886731392e-06,
      "loss": 0.053,
      "step": 236180
    },
    {
      "epoch": 91.72,
      "learning_rate": 7.770097087378642e-06,
      "loss": 0.0583,
      "step": 236190
    },
    {
      "epoch": 91.73,
      "learning_rate": 7.76957928802589e-06,
      "loss": 0.0002,
      "step": 236200
    },
    {
      "epoch": 91.73,
      "learning_rate": 7.76906148867314e-06,
      "loss": 0.0504,
      "step": 236210
    },
    {
      "epoch": 91.74,
      "learning_rate": 7.76854368932039e-06,
      "loss": 0.0593,
      "step": 236220
    },
    {
      "epoch": 91.74,
      "learning_rate": 7.768025889967639e-06,
      "loss": 0.0285,
      "step": 236230
    },
    {
      "epoch": 91.74,
      "learning_rate": 7.767508090614887e-06,
      "loss": 0.0793,
      "step": 236240
    },
    {
      "epoch": 91.75,
      "learning_rate": 7.766990291262136e-06,
      "loss": 0.1508,
      "step": 236250
    },
    {
      "epoch": 91.75,
      "learning_rate": 7.766472491909386e-06,
      "loss": 0.0901,
      "step": 236260
    },
    {
      "epoch": 91.76,
      "learning_rate": 7.765954692556636e-06,
      "loss": 0.0599,
      "step": 236270
    },
    {
      "epoch": 91.76,
      "learning_rate": 7.765436893203883e-06,
      "loss": 0.0283,
      "step": 236280
    },
    {
      "epoch": 91.76,
      "learning_rate": 7.764919093851133e-06,
      "loss": 0.0022,
      "step": 236290
    },
    {
      "epoch": 91.77,
      "learning_rate": 7.764401294498383e-06,
      "loss": 0.0254,
      "step": 236300
    },
    {
      "epoch": 91.77,
      "learning_rate": 7.763883495145632e-06,
      "loss": 0.1088,
      "step": 236310
    },
    {
      "epoch": 91.77,
      "learning_rate": 7.76336569579288e-06,
      "loss": 0.1425,
      "step": 236320
    },
    {
      "epoch": 91.78,
      "learning_rate": 7.76284789644013e-06,
      "loss": 0.0146,
      "step": 236330
    },
    {
      "epoch": 91.78,
      "learning_rate": 7.76233009708738e-06,
      "loss": 0.0732,
      "step": 236340
    },
    {
      "epoch": 91.79,
      "learning_rate": 7.761812297734629e-06,
      "loss": 0.0636,
      "step": 236350
    },
    {
      "epoch": 91.79,
      "learning_rate": 7.761294498381877e-06,
      "loss": 0.0991,
      "step": 236360
    },
    {
      "epoch": 91.79,
      "learning_rate": 7.760776699029127e-06,
      "loss": 0.0554,
      "step": 236370
    },
    {
      "epoch": 91.8,
      "learning_rate": 7.760258899676376e-06,
      "loss": 0.0206,
      "step": 236380
    },
    {
      "epoch": 91.8,
      "learning_rate": 7.759741100323626e-06,
      "loss": 0.1408,
      "step": 236390
    },
    {
      "epoch": 91.81,
      "learning_rate": 7.759223300970874e-06,
      "loss": 0.0529,
      "step": 236400
    },
    {
      "epoch": 91.81,
      "learning_rate": 7.758705501618123e-06,
      "loss": 0.1708,
      "step": 236410
    },
    {
      "epoch": 91.81,
      "learning_rate": 7.758187702265373e-06,
      "loss": 0.0312,
      "step": 236420
    },
    {
      "epoch": 91.82,
      "learning_rate": 7.757669902912623e-06,
      "loss": 0.0168,
      "step": 236430
    },
    {
      "epoch": 91.82,
      "learning_rate": 7.75715210355987e-06,
      "loss": 0.084,
      "step": 236440
    },
    {
      "epoch": 91.83,
      "learning_rate": 7.75663430420712e-06,
      "loss": 0.1027,
      "step": 236450
    },
    {
      "epoch": 91.83,
      "learning_rate": 7.75611650485437e-06,
      "loss": 0.0483,
      "step": 236460
    },
    {
      "epoch": 91.83,
      "learning_rate": 7.75559870550162e-06,
      "loss": 0.1206,
      "step": 236470
    },
    {
      "epoch": 91.84,
      "learning_rate": 7.755080906148867e-06,
      "loss": 0.2143,
      "step": 236480
    },
    {
      "epoch": 91.84,
      "learning_rate": 7.754563106796117e-06,
      "loss": 0.0002,
      "step": 236490
    },
    {
      "epoch": 91.84,
      "learning_rate": 7.754045307443367e-06,
      "loss": 0.172,
      "step": 236500
    },
    {
      "epoch": 91.85,
      "learning_rate": 7.753527508090616e-06,
      "loss": 0.0004,
      "step": 236510
    },
    {
      "epoch": 91.85,
      "learning_rate": 7.753009708737864e-06,
      "loss": 0.0206,
      "step": 236520
    },
    {
      "epoch": 91.86,
      "learning_rate": 7.752491909385114e-06,
      "loss": 0.0508,
      "step": 236530
    },
    {
      "epoch": 91.86,
      "learning_rate": 7.751974110032363e-06,
      "loss": 0.0851,
      "step": 236540
    },
    {
      "epoch": 91.86,
      "learning_rate": 7.751456310679613e-06,
      "loss": 0.0373,
      "step": 236550
    },
    {
      "epoch": 91.87,
      "learning_rate": 7.750938511326861e-06,
      "loss": 0.0074,
      "step": 236560
    },
    {
      "epoch": 91.87,
      "learning_rate": 7.75042071197411e-06,
      "loss": 0.0874,
      "step": 236570
    },
    {
      "epoch": 91.88,
      "learning_rate": 7.74990291262136e-06,
      "loss": 0.0879,
      "step": 236580
    },
    {
      "epoch": 91.88,
      "learning_rate": 7.74938511326861e-06,
      "loss": 0.0507,
      "step": 236590
    },
    {
      "epoch": 91.88,
      "learning_rate": 7.74886731391586e-06,
      "loss": 0.1319,
      "step": 236600
    },
    {
      "epoch": 91.89,
      "learning_rate": 7.748349514563107e-06,
      "loss": 0.019,
      "step": 236610
    },
    {
      "epoch": 91.89,
      "learning_rate": 7.747831715210357e-06,
      "loss": 0.0448,
      "step": 236620
    },
    {
      "epoch": 91.9,
      "learning_rate": 7.747313915857607e-06,
      "loss": 0.0497,
      "step": 236630
    },
    {
      "epoch": 91.9,
      "learning_rate": 7.746796116504856e-06,
      "loss": 0.0381,
      "step": 236640
    },
    {
      "epoch": 91.9,
      "learning_rate": 7.746278317152104e-06,
      "loss": 0.058,
      "step": 236650
    },
    {
      "epoch": 91.91,
      "learning_rate": 7.745760517799354e-06,
      "loss": 0.0583,
      "step": 236660
    },
    {
      "epoch": 91.91,
      "learning_rate": 7.745242718446603e-06,
      "loss": 0.0566,
      "step": 236670
    },
    {
      "epoch": 91.91,
      "learning_rate": 7.744724919093853e-06,
      "loss": 0.1002,
      "step": 236680
    },
    {
      "epoch": 91.92,
      "learning_rate": 7.744207119741101e-06,
      "loss": 0.0182,
      "step": 236690
    },
    {
      "epoch": 91.92,
      "learning_rate": 7.74368932038835e-06,
      "loss": 0.0375,
      "step": 236700
    },
    {
      "epoch": 91.93,
      "learning_rate": 7.7431715210356e-06,
      "loss": 0.1384,
      "step": 236710
    },
    {
      "epoch": 91.93,
      "learning_rate": 7.74265372168285e-06,
      "loss": 0.2688,
      "step": 236720
    },
    {
      "epoch": 91.93,
      "learning_rate": 7.742135922330098e-06,
      "loss": 0.0798,
      "step": 236730
    },
    {
      "epoch": 91.94,
      "learning_rate": 7.741618122977347e-06,
      "loss": 0.0603,
      "step": 236740
    },
    {
      "epoch": 91.94,
      "learning_rate": 7.741100323624597e-06,
      "loss": 0.0607,
      "step": 236750
    },
    {
      "epoch": 91.95,
      "learning_rate": 7.740582524271847e-06,
      "loss": 0.0338,
      "step": 236760
    },
    {
      "epoch": 91.95,
      "learning_rate": 7.740064724919094e-06,
      "loss": 0.1021,
      "step": 236770
    },
    {
      "epoch": 91.95,
      "learning_rate": 7.739546925566344e-06,
      "loss": 0.0432,
      "step": 236780
    },
    {
      "epoch": 91.96,
      "learning_rate": 7.739029126213594e-06,
      "loss": 0.0145,
      "step": 236790
    },
    {
      "epoch": 91.96,
      "learning_rate": 7.738511326860843e-06,
      "loss": 0.0347,
      "step": 236800
    },
    {
      "epoch": 91.97,
      "learning_rate": 7.737993527508091e-06,
      "loss": 0.0034,
      "step": 236810
    },
    {
      "epoch": 91.97,
      "learning_rate": 7.73747572815534e-06,
      "loss": 0.0005,
      "step": 236820
    },
    {
      "epoch": 91.97,
      "learning_rate": 7.73695792880259e-06,
      "loss": 0.0709,
      "step": 236830
    },
    {
      "epoch": 91.98,
      "learning_rate": 7.73644012944984e-06,
      "loss": 0.0033,
      "step": 236840
    },
    {
      "epoch": 91.98,
      "learning_rate": 7.735922330097088e-06,
      "loss": 0.0981,
      "step": 236850
    },
    {
      "epoch": 91.98,
      "learning_rate": 7.735404530744338e-06,
      "loss": 0.1462,
      "step": 236860
    },
    {
      "epoch": 91.99,
      "learning_rate": 7.734886731391587e-06,
      "loss": 0.1284,
      "step": 236870
    },
    {
      "epoch": 91.99,
      "learning_rate": 7.734368932038835e-06,
      "loss": 0.0125,
      "step": 236880
    },
    {
      "epoch": 92.0,
      "learning_rate": 7.733851132686085e-06,
      "loss": 0.0002,
      "step": 236890
    },
    {
      "epoch": 92.0,
      "learning_rate": 7.733333333333334e-06,
      "loss": 0.0827,
      "step": 236900
    },
    {
      "epoch": 92.0,
      "eval_accuracy": 0.9485557083906465,
      "eval_loss": 0.36826103925704956,
      "eval_runtime": 8.2034,
      "eval_samples_per_second": 443.11,
      "eval_steps_per_second": 55.465,
      "step": 236900
    },
    {
      "epoch": 92.0,
      "learning_rate": 7.732815533980584e-06,
      "loss": 0.0974,
      "step": 236910
    },
    {
      "epoch": 92.01,
      "learning_rate": 7.732297734627832e-06,
      "loss": 0.0106,
      "step": 236920
    },
    {
      "epoch": 92.01,
      "learning_rate": 7.731779935275082e-06,
      "loss": 0.0321,
      "step": 236930
    },
    {
      "epoch": 92.02,
      "learning_rate": 7.731262135922331e-06,
      "loss": 0.0937,
      "step": 236940
    },
    {
      "epoch": 92.02,
      "learning_rate": 7.73074433656958e-06,
      "loss": 0.1883,
      "step": 236950
    },
    {
      "epoch": 92.02,
      "learning_rate": 7.730226537216829e-06,
      "loss": 0.1077,
      "step": 236960
    },
    {
      "epoch": 92.03,
      "learning_rate": 7.729708737864078e-06,
      "loss": 0.0939,
      "step": 236970
    },
    {
      "epoch": 92.03,
      "learning_rate": 7.729190938511328e-06,
      "loss": 0.0282,
      "step": 236980
    },
    {
      "epoch": 92.03,
      "learning_rate": 7.728673139158578e-06,
      "loss": 0.0423,
      "step": 236990
    },
    {
      "epoch": 92.04,
      "learning_rate": 7.728155339805825e-06,
      "loss": 0.0149,
      "step": 237000
    },
    {
      "epoch": 92.04,
      "learning_rate": 7.727637540453075e-06,
      "loss": 0.0824,
      "step": 237010
    },
    {
      "epoch": 92.05,
      "learning_rate": 7.727119741100325e-06,
      "loss": 0.2211,
      "step": 237020
    },
    {
      "epoch": 92.05,
      "learning_rate": 7.726601941747573e-06,
      "loss": 0.0225,
      "step": 237030
    },
    {
      "epoch": 92.05,
      "learning_rate": 7.726084142394822e-06,
      "loss": 0.0088,
      "step": 237040
    },
    {
      "epoch": 92.06,
      "learning_rate": 7.725566343042072e-06,
      "loss": 0.0897,
      "step": 237050
    },
    {
      "epoch": 92.06,
      "learning_rate": 7.725048543689322e-06,
      "loss": 0.001,
      "step": 237060
    },
    {
      "epoch": 92.07,
      "learning_rate": 7.72453074433657e-06,
      "loss": 0.0825,
      "step": 237070
    },
    {
      "epoch": 92.07,
      "learning_rate": 7.724012944983819e-06,
      "loss": 0.0379,
      "step": 237080
    },
    {
      "epoch": 92.07,
      "learning_rate": 7.723495145631069e-06,
      "loss": 0.1259,
      "step": 237090
    },
    {
      "epoch": 92.08,
      "learning_rate": 7.722977346278318e-06,
      "loss": 0.0326,
      "step": 237100
    },
    {
      "epoch": 92.08,
      "learning_rate": 7.722459546925566e-06,
      "loss": 0.0181,
      "step": 237110
    },
    {
      "epoch": 92.09,
      "learning_rate": 7.721941747572816e-06,
      "loss": 0.0864,
      "step": 237120
    },
    {
      "epoch": 92.09,
      "learning_rate": 7.721423948220065e-06,
      "loss": 0.0024,
      "step": 237130
    },
    {
      "epoch": 92.09,
      "learning_rate": 7.720906148867315e-06,
      "loss": 0.0174,
      "step": 237140
    },
    {
      "epoch": 92.1,
      "learning_rate": 7.720388349514563e-06,
      "loss": 0.0197,
      "step": 237150
    },
    {
      "epoch": 92.1,
      "learning_rate": 7.719870550161813e-06,
      "loss": 0.0565,
      "step": 237160
    },
    {
      "epoch": 92.1,
      "learning_rate": 7.719352750809062e-06,
      "loss": 0.0458,
      "step": 237170
    },
    {
      "epoch": 92.11,
      "learning_rate": 7.718834951456312e-06,
      "loss": 0.0213,
      "step": 237180
    },
    {
      "epoch": 92.11,
      "learning_rate": 7.71831715210356e-06,
      "loss": 0.1253,
      "step": 237190
    },
    {
      "epoch": 92.12,
      "learning_rate": 7.71779935275081e-06,
      "loss": 0.1019,
      "step": 237200
    },
    {
      "epoch": 92.12,
      "learning_rate": 7.717281553398059e-06,
      "loss": 0.0098,
      "step": 237210
    },
    {
      "epoch": 92.12,
      "learning_rate": 7.716763754045307e-06,
      "loss": 0.0894,
      "step": 237220
    },
    {
      "epoch": 92.13,
      "learning_rate": 7.716245954692557e-06,
      "loss": 0.0264,
      "step": 237230
    },
    {
      "epoch": 92.13,
      "learning_rate": 7.715728155339806e-06,
      "loss": 0.0449,
      "step": 237240
    },
    {
      "epoch": 92.14,
      "learning_rate": 7.715210355987056e-06,
      "loss": 0.0207,
      "step": 237250
    },
    {
      "epoch": 92.14,
      "learning_rate": 7.714692556634304e-06,
      "loss": 0.0313,
      "step": 237260
    },
    {
      "epoch": 92.14,
      "learning_rate": 7.714174757281553e-06,
      "loss": 0.0043,
      "step": 237270
    },
    {
      "epoch": 92.15,
      "learning_rate": 7.713656957928803e-06,
      "loss": 0.0114,
      "step": 237280
    },
    {
      "epoch": 92.15,
      "learning_rate": 7.713139158576053e-06,
      "loss": 0.0036,
      "step": 237290
    },
    {
      "epoch": 92.16,
      "learning_rate": 7.7126213592233e-06,
      "loss": 0.1209,
      "step": 237300
    },
    {
      "epoch": 92.16,
      "learning_rate": 7.71210355987055e-06,
      "loss": 0.0834,
      "step": 237310
    },
    {
      "epoch": 92.16,
      "learning_rate": 7.7115857605178e-06,
      "loss": 0.0004,
      "step": 237320
    },
    {
      "epoch": 92.17,
      "learning_rate": 7.71106796116505e-06,
      "loss": 0.0792,
      "step": 237330
    },
    {
      "epoch": 92.17,
      "learning_rate": 7.710550161812297e-06,
      "loss": 0.034,
      "step": 237340
    },
    {
      "epoch": 92.17,
      "learning_rate": 7.710032362459547e-06,
      "loss": 0.0698,
      "step": 237350
    },
    {
      "epoch": 92.18,
      "learning_rate": 7.709514563106796e-06,
      "loss": 0.001,
      "step": 237360
    },
    {
      "epoch": 92.18,
      "learning_rate": 7.708996763754046e-06,
      "loss": 0.0154,
      "step": 237370
    },
    {
      "epoch": 92.19,
      "learning_rate": 7.708478964401294e-06,
      "loss": 0.1349,
      "step": 237380
    },
    {
      "epoch": 92.19,
      "learning_rate": 7.707961165048544e-06,
      "loss": 0.0537,
      "step": 237390
    },
    {
      "epoch": 92.19,
      "learning_rate": 7.707443365695793e-06,
      "loss": 0.1664,
      "step": 237400
    },
    {
      "epoch": 92.2,
      "learning_rate": 7.706925566343043e-06,
      "loss": 0.0622,
      "step": 237410
    },
    {
      "epoch": 92.2,
      "learning_rate": 7.70640776699029e-06,
      "loss": 0.0015,
      "step": 237420
    },
    {
      "epoch": 92.21,
      "learning_rate": 7.70588996763754e-06,
      "loss": 0.0927,
      "step": 237430
    },
    {
      "epoch": 92.21,
      "learning_rate": 7.70537216828479e-06,
      "loss": 0.0009,
      "step": 237440
    },
    {
      "epoch": 92.21,
      "learning_rate": 7.70485436893204e-06,
      "loss": 0.092,
      "step": 237450
    },
    {
      "epoch": 92.22,
      "learning_rate": 7.704336569579288e-06,
      "loss": 0.1238,
      "step": 237460
    },
    {
      "epoch": 92.22,
      "learning_rate": 7.703818770226537e-06,
      "loss": 0.0139,
      "step": 237470
    },
    {
      "epoch": 92.23,
      "learning_rate": 7.703300970873787e-06,
      "loss": 0.0205,
      "step": 237480
    },
    {
      "epoch": 92.23,
      "learning_rate": 7.702783171521036e-06,
      "loss": 0.0041,
      "step": 237490
    },
    {
      "epoch": 92.23,
      "learning_rate": 7.702265372168284e-06,
      "loss": 0.0833,
      "step": 237500
    },
    {
      "epoch": 92.24,
      "learning_rate": 7.701747572815534e-06,
      "loss": 0.0208,
      "step": 237510
    },
    {
      "epoch": 92.24,
      "learning_rate": 7.701229773462784e-06,
      "loss": 0.0273,
      "step": 237520
    },
    {
      "epoch": 92.24,
      "learning_rate": 7.700711974110033e-06,
      "loss": 0.0794,
      "step": 237530
    },
    {
      "epoch": 92.25,
      "learning_rate": 7.700194174757281e-06,
      "loss": 0.0688,
      "step": 237540
    },
    {
      "epoch": 92.25,
      "learning_rate": 7.69967637540453e-06,
      "loss": 0.0875,
      "step": 237550
    },
    {
      "epoch": 92.26,
      "learning_rate": 7.69915857605178e-06,
      "loss": 0.0017,
      "step": 237560
    },
    {
      "epoch": 92.26,
      "learning_rate": 7.69864077669903e-06,
      "loss": 0.0368,
      "step": 237570
    },
    {
      "epoch": 92.26,
      "learning_rate": 7.698122977346278e-06,
      "loss": 0.0415,
      "step": 237580
    },
    {
      "epoch": 92.27,
      "learning_rate": 7.697605177993528e-06,
      "loss": 0.1454,
      "step": 237590
    },
    {
      "epoch": 92.27,
      "learning_rate": 7.697087378640777e-06,
      "loss": 0.0029,
      "step": 237600
    },
    {
      "epoch": 92.28,
      "learning_rate": 7.696569579288027e-06,
      "loss": 0.0267,
      "step": 237610
    },
    {
      "epoch": 92.28,
      "learning_rate": 7.696051779935275e-06,
      "loss": 0.0362,
      "step": 237620
    },
    {
      "epoch": 92.28,
      "learning_rate": 7.695533980582524e-06,
      "loss": 0.0457,
      "step": 237630
    },
    {
      "epoch": 92.29,
      "learning_rate": 7.695016181229774e-06,
      "loss": 0.0128,
      "step": 237640
    },
    {
      "epoch": 92.29,
      "learning_rate": 7.694498381877024e-06,
      "loss": 0.047,
      "step": 237650
    },
    {
      "epoch": 92.3,
      "learning_rate": 7.693980582524271e-06,
      "loss": 0.0306,
      "step": 237660
    },
    {
      "epoch": 92.3,
      "learning_rate": 7.693462783171521e-06,
      "loss": 0.2175,
      "step": 237670
    },
    {
      "epoch": 92.3,
      "learning_rate": 7.69294498381877e-06,
      "loss": 0.0173,
      "step": 237680
    },
    {
      "epoch": 92.31,
      "learning_rate": 7.69242718446602e-06,
      "loss": 0.0219,
      "step": 237690
    },
    {
      "epoch": 92.31,
      "learning_rate": 7.69190938511327e-06,
      "loss": 0.0889,
      "step": 237700
    },
    {
      "epoch": 92.31,
      "learning_rate": 7.691391585760518e-06,
      "loss": 0.0869,
      "step": 237710
    },
    {
      "epoch": 92.32,
      "learning_rate": 7.690873786407767e-06,
      "loss": 0.0708,
      "step": 237720
    },
    {
      "epoch": 92.32,
      "learning_rate": 7.690355987055017e-06,
      "loss": 0.1184,
      "step": 237730
    },
    {
      "epoch": 92.33,
      "learning_rate": 7.689838187702267e-06,
      "loss": 0.0523,
      "step": 237740
    },
    {
      "epoch": 92.33,
      "learning_rate": 7.689320388349515e-06,
      "loss": 0.1246,
      "step": 237750
    },
    {
      "epoch": 92.33,
      "learning_rate": 7.688802588996764e-06,
      "loss": 0.0025,
      "step": 237760
    },
    {
      "epoch": 92.34,
      "learning_rate": 7.688284789644014e-06,
      "loss": 0.0921,
      "step": 237770
    },
    {
      "epoch": 92.34,
      "learning_rate": 7.687766990291264e-06,
      "loss": 0.0325,
      "step": 237780
    },
    {
      "epoch": 92.35,
      "learning_rate": 7.687249190938511e-06,
      "loss": 0.0609,
      "step": 237790
    },
    {
      "epoch": 92.35,
      "learning_rate": 7.686731391585761e-06,
      "loss": 0.0473,
      "step": 237800
    },
    {
      "epoch": 92.35,
      "learning_rate": 7.68621359223301e-06,
      "loss": 0.0608,
      "step": 237810
    },
    {
      "epoch": 92.36,
      "learning_rate": 7.68569579288026e-06,
      "loss": 0.0321,
      "step": 237820
    },
    {
      "epoch": 92.36,
      "learning_rate": 7.685177993527508e-06,
      "loss": 0.0532,
      "step": 237830
    },
    {
      "epoch": 92.37,
      "learning_rate": 7.684660194174758e-06,
      "loss": 0.1558,
      "step": 237840
    },
    {
      "epoch": 92.37,
      "learning_rate": 7.684142394822007e-06,
      "loss": 0.0567,
      "step": 237850
    },
    {
      "epoch": 92.37,
      "learning_rate": 7.683624595469257e-06,
      "loss": 0.1126,
      "step": 237860
    },
    {
      "epoch": 92.38,
      "learning_rate": 7.683106796116505e-06,
      "loss": 0.011,
      "step": 237870
    },
    {
      "epoch": 92.38,
      "learning_rate": 7.682588996763755e-06,
      "loss": 0.005,
      "step": 237880
    },
    {
      "epoch": 92.38,
      "learning_rate": 7.682071197411004e-06,
      "loss": 0.0134,
      "step": 237890
    },
    {
      "epoch": 92.39,
      "learning_rate": 7.681553398058254e-06,
      "loss": 0.0311,
      "step": 237900
    },
    {
      "epoch": 92.39,
      "learning_rate": 7.681035598705502e-06,
      "loss": 0.1547,
      "step": 237910
    },
    {
      "epoch": 92.4,
      "learning_rate": 7.680517799352751e-06,
      "loss": 0.046,
      "step": 237920
    },
    {
      "epoch": 92.4,
      "learning_rate": 7.680000000000001e-06,
      "loss": 0.1096,
      "step": 237930
    },
    {
      "epoch": 92.4,
      "learning_rate": 7.67948220064725e-06,
      "loss": 0.0911,
      "step": 237940
    },
    {
      "epoch": 92.41,
      "learning_rate": 7.678964401294499e-06,
      "loss": 0.0146,
      "step": 237950
    },
    {
      "epoch": 92.41,
      "learning_rate": 7.678446601941748e-06,
      "loss": 0.0572,
      "step": 237960
    },
    {
      "epoch": 92.42,
      "learning_rate": 7.677928802588998e-06,
      "loss": 0.0707,
      "step": 237970
    },
    {
      "epoch": 92.42,
      "learning_rate": 7.677411003236247e-06,
      "loss": 0.0534,
      "step": 237980
    },
    {
      "epoch": 92.42,
      "learning_rate": 7.676893203883495e-06,
      "loss": 0.096,
      "step": 237990
    },
    {
      "epoch": 92.43,
      "learning_rate": 7.676375404530745e-06,
      "loss": 0.0756,
      "step": 238000
    },
    {
      "epoch": 92.43,
      "learning_rate": 7.675857605177995e-06,
      "loss": 0.1056,
      "step": 238010
    },
    {
      "epoch": 92.43,
      "learning_rate": 7.675339805825244e-06,
      "loss": 0.0636,
      "step": 238020
    },
    {
      "epoch": 92.44,
      "learning_rate": 7.674822006472492e-06,
      "loss": 0.1009,
      "step": 238030
    },
    {
      "epoch": 92.44,
      "learning_rate": 7.674304207119742e-06,
      "loss": 0.0944,
      "step": 238040
    },
    {
      "epoch": 92.45,
      "learning_rate": 7.673786407766991e-06,
      "loss": 0.1125,
      "step": 238050
    },
    {
      "epoch": 92.45,
      "learning_rate": 7.673268608414241e-06,
      "loss": 0.0968,
      "step": 238060
    },
    {
      "epoch": 92.45,
      "learning_rate": 7.672750809061489e-06,
      "loss": 0.039,
      "step": 238070
    },
    {
      "epoch": 92.46,
      "learning_rate": 7.672233009708739e-06,
      "loss": 0.0399,
      "step": 238080
    },
    {
      "epoch": 92.46,
      "learning_rate": 7.671715210355988e-06,
      "loss": 0.0108,
      "step": 238090
    },
    {
      "epoch": 92.47,
      "learning_rate": 7.671197411003238e-06,
      "loss": 0.1491,
      "step": 238100
    },
    {
      "epoch": 92.47,
      "learning_rate": 7.670679611650486e-06,
      "loss": 0.0402,
      "step": 238110
    },
    {
      "epoch": 92.47,
      "learning_rate": 7.670161812297735e-06,
      "loss": 0.0495,
      "step": 238120
    },
    {
      "epoch": 92.48,
      "learning_rate": 7.669644012944985e-06,
      "loss": 0.0005,
      "step": 238130
    },
    {
      "epoch": 92.48,
      "learning_rate": 7.669126213592235e-06,
      "loss": 0.0387,
      "step": 238140
    },
    {
      "epoch": 92.49,
      "learning_rate": 7.668608414239482e-06,
      "loss": 0.0668,
      "step": 238150
    },
    {
      "epoch": 92.49,
      "learning_rate": 7.668090614886732e-06,
      "loss": 0.0071,
      "step": 238160
    },
    {
      "epoch": 92.49,
      "learning_rate": 7.667572815533982e-06,
      "loss": 0.0795,
      "step": 238170
    },
    {
      "epoch": 92.5,
      "learning_rate": 7.667055016181231e-06,
      "loss": 0.0007,
      "step": 238180
    },
    {
      "epoch": 92.5,
      "learning_rate": 7.66653721682848e-06,
      "loss": 0.0419,
      "step": 238190
    },
    {
      "epoch": 92.5,
      "learning_rate": 7.666019417475729e-06,
      "loss": 0.0694,
      "step": 238200
    },
    {
      "epoch": 92.51,
      "learning_rate": 7.665501618122978e-06,
      "loss": 0.0203,
      "step": 238210
    },
    {
      "epoch": 92.51,
      "learning_rate": 7.664983818770228e-06,
      "loss": 0.0906,
      "step": 238220
    },
    {
      "epoch": 92.52,
      "learning_rate": 7.664466019417476e-06,
      "loss": 0.0057,
      "step": 238230
    },
    {
      "epoch": 92.52,
      "learning_rate": 7.663948220064726e-06,
      "loss": 0.0059,
      "step": 238240
    },
    {
      "epoch": 92.52,
      "learning_rate": 7.663430420711975e-06,
      "loss": 0.0015,
      "step": 238250
    },
    {
      "epoch": 92.53,
      "learning_rate": 7.662912621359225e-06,
      "loss": 0.1447,
      "step": 238260
    },
    {
      "epoch": 92.53,
      "learning_rate": 7.662394822006474e-06,
      "loss": 0.1366,
      "step": 238270
    },
    {
      "epoch": 92.54,
      "learning_rate": 7.661877022653722e-06,
      "loss": 0.061,
      "step": 238280
    },
    {
      "epoch": 92.54,
      "learning_rate": 7.661359223300972e-06,
      "loss": 0.0374,
      "step": 238290
    },
    {
      "epoch": 92.54,
      "learning_rate": 7.660841423948222e-06,
      "loss": 0.1163,
      "step": 238300
    },
    {
      "epoch": 92.55,
      "learning_rate": 7.660323624595471e-06,
      "loss": 0.1245,
      "step": 238310
    },
    {
      "epoch": 92.55,
      "learning_rate": 7.65980582524272e-06,
      "loss": 0.1311,
      "step": 238320
    },
    {
      "epoch": 92.56,
      "learning_rate": 7.659288025889969e-06,
      "loss": 0.1952,
      "step": 238330
    },
    {
      "epoch": 92.56,
      "learning_rate": 7.658770226537218e-06,
      "loss": 0.0376,
      "step": 238340
    },
    {
      "epoch": 92.56,
      "learning_rate": 7.658252427184466e-06,
      "loss": 0.0172,
      "step": 238350
    },
    {
      "epoch": 92.57,
      "learning_rate": 7.657734627831716e-06,
      "loss": 0.0108,
      "step": 238360
    },
    {
      "epoch": 92.57,
      "learning_rate": 7.657216828478966e-06,
      "loss": 0.0301,
      "step": 238370
    },
    {
      "epoch": 92.57,
      "learning_rate": 7.656699029126215e-06,
      "loss": 0.0289,
      "step": 238380
    },
    {
      "epoch": 92.58,
      "learning_rate": 7.656181229773463e-06,
      "loss": 0.0492,
      "step": 238390
    },
    {
      "epoch": 92.58,
      "learning_rate": 7.655663430420713e-06,
      "loss": 0.0005,
      "step": 238400
    },
    {
      "epoch": 92.59,
      "learning_rate": 7.655145631067962e-06,
      "loss": 0.0277,
      "step": 238410
    },
    {
      "epoch": 92.59,
      "learning_rate": 7.654627831715212e-06,
      "loss": 0.0508,
      "step": 238420
    },
    {
      "epoch": 92.59,
      "learning_rate": 7.65411003236246e-06,
      "loss": 0.106,
      "step": 238430
    },
    {
      "epoch": 92.6,
      "learning_rate": 7.65359223300971e-06,
      "loss": 0.0622,
      "step": 238440
    },
    {
      "epoch": 92.6,
      "learning_rate": 7.653074433656959e-06,
      "loss": 0.0002,
      "step": 238450
    },
    {
      "epoch": 92.61,
      "learning_rate": 7.652556634304209e-06,
      "loss": 0.0129,
      "step": 238460
    },
    {
      "epoch": 92.61,
      "learning_rate": 7.652038834951457e-06,
      "loss": 0.0739,
      "step": 238470
    },
    {
      "epoch": 92.61,
      "learning_rate": 7.651521035598706e-06,
      "loss": 0.0244,
      "step": 238480
    },
    {
      "epoch": 92.62,
      "learning_rate": 7.651003236245956e-06,
      "loss": 0.0608,
      "step": 238490
    },
    {
      "epoch": 92.62,
      "learning_rate": 7.650485436893204e-06,
      "loss": 0.0017,
      "step": 238500
    },
    {
      "epoch": 92.63,
      "learning_rate": 7.649967637540453e-06,
      "loss": 0.0221,
      "step": 238510
    },
    {
      "epoch": 92.63,
      "learning_rate": 7.649449838187703e-06,
      "loss": 0.0659,
      "step": 238520
    },
    {
      "epoch": 92.63,
      "learning_rate": 7.648932038834953e-06,
      "loss": 0.0827,
      "step": 238530
    },
    {
      "epoch": 92.64,
      "learning_rate": 7.6484142394822e-06,
      "loss": 0.0669,
      "step": 238540
    },
    {
      "epoch": 92.64,
      "learning_rate": 7.64789644012945e-06,
      "loss": 0.0288,
      "step": 238550
    },
    {
      "epoch": 92.64,
      "learning_rate": 7.6473786407767e-06,
      "loss": 0.0187,
      "step": 238560
    },
    {
      "epoch": 92.65,
      "learning_rate": 7.64686084142395e-06,
      "loss": 0.0985,
      "step": 238570
    },
    {
      "epoch": 92.65,
      "learning_rate": 7.646343042071197e-06,
      "loss": 0.0005,
      "step": 238580
    },
    {
      "epoch": 92.66,
      "learning_rate": 7.645825242718447e-06,
      "loss": 0.0323,
      "step": 238590
    },
    {
      "epoch": 92.66,
      "learning_rate": 7.645307443365697e-06,
      "loss": 0.0205,
      "step": 238600
    },
    {
      "epoch": 92.66,
      "learning_rate": 7.644789644012946e-06,
      "loss": 0.0677,
      "step": 238610
    },
    {
      "epoch": 92.67,
      "learning_rate": 7.644271844660194e-06,
      "loss": 0.0007,
      "step": 238620
    },
    {
      "epoch": 92.67,
      "learning_rate": 7.643754045307444e-06,
      "loss": 0.0554,
      "step": 238630
    },
    {
      "epoch": 92.68,
      "learning_rate": 7.643236245954693e-06,
      "loss": 0.0004,
      "step": 238640
    },
    {
      "epoch": 92.68,
      "learning_rate": 7.642718446601943e-06,
      "loss": 0.0195,
      "step": 238650
    },
    {
      "epoch": 92.68,
      "learning_rate": 7.642200647249191e-06,
      "loss": 0.1397,
      "step": 238660
    },
    {
      "epoch": 92.69,
      "learning_rate": 7.64168284789644e-06,
      "loss": 0.0809,
      "step": 238670
    },
    {
      "epoch": 92.69,
      "learning_rate": 7.64116504854369e-06,
      "loss": 0.0303,
      "step": 238680
    },
    {
      "epoch": 92.7,
      "learning_rate": 7.640647249190938e-06,
      "loss": 0.0091,
      "step": 238690
    },
    {
      "epoch": 92.7,
      "learning_rate": 7.640129449838188e-06,
      "loss": 0.0639,
      "step": 238700
    },
    {
      "epoch": 92.7,
      "learning_rate": 7.639611650485437e-06,
      "loss": 0.0002,
      "step": 238710
    },
    {
      "epoch": 92.71,
      "learning_rate": 7.639093851132687e-06,
      "loss": 0.0669,
      "step": 238720
    },
    {
      "epoch": 92.71,
      "learning_rate": 7.638576051779935e-06,
      "loss": 0.1333,
      "step": 238730
    },
    {
      "epoch": 92.71,
      "learning_rate": 7.638058252427184e-06,
      "loss": 0.003,
      "step": 238740
    },
    {
      "epoch": 92.72,
      "learning_rate": 7.637540453074434e-06,
      "loss": 0.0119,
      "step": 238750
    },
    {
      "epoch": 92.72,
      "learning_rate": 7.637022653721684e-06,
      "loss": 0.1363,
      "step": 238760
    },
    {
      "epoch": 92.73,
      "learning_rate": 7.636504854368932e-06,
      "loss": 0.1757,
      "step": 238770
    },
    {
      "epoch": 92.73,
      "learning_rate": 7.635987055016181e-06,
      "loss": 0.1835,
      "step": 238780
    },
    {
      "epoch": 92.73,
      "learning_rate": 7.635469255663431e-06,
      "loss": 0.1767,
      "step": 238790
    },
    {
      "epoch": 92.74,
      "learning_rate": 7.63495145631068e-06,
      "loss": 0.0853,
      "step": 238800
    },
    {
      "epoch": 92.74,
      "learning_rate": 7.634433656957928e-06,
      "loss": 0.0384,
      "step": 238810
    },
    {
      "epoch": 92.75,
      "learning_rate": 7.633915857605178e-06,
      "loss": 0.0187,
      "step": 238820
    },
    {
      "epoch": 92.75,
      "learning_rate": 7.633398058252428e-06,
      "loss": 0.0313,
      "step": 238830
    },
    {
      "epoch": 92.75,
      "learning_rate": 7.632880258899677e-06,
      "loss": 0.1071,
      "step": 238840
    },
    {
      "epoch": 92.76,
      "learning_rate": 7.632362459546925e-06,
      "loss": 0.0627,
      "step": 238850
    },
    {
      "epoch": 92.76,
      "learning_rate": 7.631844660194175e-06,
      "loss": 0.0168,
      "step": 238860
    },
    {
      "epoch": 92.77,
      "learning_rate": 7.631326860841424e-06,
      "loss": 0.0971,
      "step": 238870
    },
    {
      "epoch": 92.77,
      "learning_rate": 7.630809061488674e-06,
      "loss": 0.0023,
      "step": 238880
    },
    {
      "epoch": 92.77,
      "learning_rate": 7.630291262135922e-06,
      "loss": 0.106,
      "step": 238890
    },
    {
      "epoch": 92.78,
      "learning_rate": 7.629773462783172e-06,
      "loss": 0.0364,
      "step": 238900
    },
    {
      "epoch": 92.78,
      "learning_rate": 7.629255663430421e-06,
      "loss": 0.0457,
      "step": 238910
    },
    {
      "epoch": 92.78,
      "learning_rate": 7.628737864077671e-06,
      "loss": 0.1308,
      "step": 238920
    },
    {
      "epoch": 92.79,
      "learning_rate": 7.62822006472492e-06,
      "loss": 0.0657,
      "step": 238930
    },
    {
      "epoch": 92.79,
      "learning_rate": 7.627702265372169e-06,
      "loss": 0.0876,
      "step": 238940
    },
    {
      "epoch": 92.8,
      "learning_rate": 7.627184466019418e-06,
      "loss": 0.0063,
      "step": 238950
    },
    {
      "epoch": 92.8,
      "learning_rate": 7.626666666666668e-06,
      "loss": 0.1407,
      "step": 238960
    },
    {
      "epoch": 92.8,
      "learning_rate": 7.626148867313916e-06,
      "loss": 0.1009,
      "step": 238970
    },
    {
      "epoch": 92.81,
      "learning_rate": 7.625631067961166e-06,
      "loss": 0.0758,
      "step": 238980
    },
    {
      "epoch": 92.81,
      "learning_rate": 7.625113268608415e-06,
      "loss": 0.0229,
      "step": 238990
    },
    {
      "epoch": 92.82,
      "learning_rate": 7.624595469255664e-06,
      "loss": 0.0026,
      "step": 239000
    },
    {
      "epoch": 92.82,
      "learning_rate": 7.624077669902913e-06,
      "loss": 0.0155,
      "step": 239010
    },
    {
      "epoch": 92.82,
      "learning_rate": 7.623559870550162e-06,
      "loss": 0.0131,
      "step": 239020
    },
    {
      "epoch": 92.83,
      "learning_rate": 7.6230420711974116e-06,
      "loss": 0.0801,
      "step": 239030
    },
    {
      "epoch": 92.83,
      "learning_rate": 7.622524271844661e-06,
      "loss": 0.1329,
      "step": 239040
    },
    {
      "epoch": 92.83,
      "learning_rate": 7.62200647249191e-06,
      "loss": 0.1669,
      "step": 239050
    },
    {
      "epoch": 92.84,
      "learning_rate": 7.621488673139159e-06,
      "loss": 0.0315,
      "step": 239060
    },
    {
      "epoch": 92.84,
      "learning_rate": 7.620970873786408e-06,
      "loss": 0.0436,
      "step": 239070
    },
    {
      "epoch": 92.85,
      "learning_rate": 7.620453074433658e-06,
      "loss": 0.0561,
      "step": 239080
    },
    {
      "epoch": 92.85,
      "learning_rate": 7.619935275080907e-06,
      "loss": 0.0783,
      "step": 239090
    },
    {
      "epoch": 92.85,
      "learning_rate": 7.6194174757281555e-06,
      "loss": 0.0005,
      "step": 239100
    },
    {
      "epoch": 92.86,
      "learning_rate": 7.618899676375405e-06,
      "loss": 0.0796,
      "step": 239110
    },
    {
      "epoch": 92.86,
      "learning_rate": 7.618381877022655e-06,
      "loss": 0.0023,
      "step": 239120
    },
    {
      "epoch": 92.87,
      "learning_rate": 7.6178640776699035e-06,
      "loss": 0.0733,
      "step": 239130
    },
    {
      "epoch": 92.87,
      "learning_rate": 7.617346278317152e-06,
      "loss": 0.0119,
      "step": 239140
    },
    {
      "epoch": 92.87,
      "learning_rate": 7.616828478964402e-06,
      "loss": 0.0239,
      "step": 239150
    },
    {
      "epoch": 92.88,
      "learning_rate": 7.6163106796116515e-06,
      "loss": 0.0402,
      "step": 239160
    },
    {
      "epoch": 92.88,
      "learning_rate": 7.6157928802588994e-06,
      "loss": 0.0839,
      "step": 239170
    },
    {
      "epoch": 92.89,
      "learning_rate": 7.615275080906149e-06,
      "loss": 0.0443,
      "step": 239180
    },
    {
      "epoch": 92.89,
      "learning_rate": 7.614757281553399e-06,
      "loss": 0.1275,
      "step": 239190
    },
    {
      "epoch": 92.89,
      "learning_rate": 7.614239482200648e-06,
      "loss": 0.042,
      "step": 239200
    },
    {
      "epoch": 92.9,
      "learning_rate": 7.613721682847896e-06,
      "loss": 0.0322,
      "step": 239210
    },
    {
      "epoch": 92.9,
      "learning_rate": 7.613203883495146e-06,
      "loss": 0.0338,
      "step": 239220
    },
    {
      "epoch": 92.9,
      "learning_rate": 7.6126860841423954e-06,
      "loss": 0.0064,
      "step": 239230
    },
    {
      "epoch": 92.91,
      "learning_rate": 7.612168284789645e-06,
      "loss": 0.0434,
      "step": 239240
    },
    {
      "epoch": 92.91,
      "learning_rate": 7.611650485436893e-06,
      "loss": 0.0641,
      "step": 239250
    },
    {
      "epoch": 92.92,
      "learning_rate": 7.611132686084143e-06,
      "loss": 0.1536,
      "step": 239260
    },
    {
      "epoch": 92.92,
      "learning_rate": 7.610614886731392e-06,
      "loss": 0.0675,
      "step": 239270
    },
    {
      "epoch": 92.92,
      "learning_rate": 7.610097087378642e-06,
      "loss": 0.1626,
      "step": 239280
    },
    {
      "epoch": 92.93,
      "learning_rate": 7.60957928802589e-06,
      "loss": 0.017,
      "step": 239290
    },
    {
      "epoch": 92.93,
      "learning_rate": 7.609061488673139e-06,
      "loss": 0.0004,
      "step": 239300
    },
    {
      "epoch": 92.94,
      "learning_rate": 7.608543689320389e-06,
      "loss": 0.1239,
      "step": 239310
    },
    {
      "epoch": 92.94,
      "learning_rate": 7.608025889967639e-06,
      "loss": 0.1049,
      "step": 239320
    },
    {
      "epoch": 92.94,
      "learning_rate": 7.6075080906148865e-06,
      "loss": 0.0223,
      "step": 239330
    },
    {
      "epoch": 92.95,
      "learning_rate": 7.606990291262136e-06,
      "loss": 0.1044,
      "step": 239340
    },
    {
      "epoch": 92.95,
      "learning_rate": 7.606472491909386e-06,
      "loss": 0.0002,
      "step": 239350
    },
    {
      "epoch": 92.96,
      "learning_rate": 7.605954692556635e-06,
      "loss": 0.0946,
      "step": 239360
    },
    {
      "epoch": 92.96,
      "learning_rate": 7.605436893203883e-06,
      "loss": 0.0798,
      "step": 239370
    },
    {
      "epoch": 92.96,
      "learning_rate": 7.604919093851133e-06,
      "loss": 0.0269,
      "step": 239380
    },
    {
      "epoch": 92.97,
      "learning_rate": 7.6044012944983826e-06,
      "loss": 0.0575,
      "step": 239390
    },
    {
      "epoch": 92.97,
      "learning_rate": 7.603883495145632e-06,
      "loss": 0.1086,
      "step": 239400
    },
    {
      "epoch": 92.97,
      "learning_rate": 7.603365695792882e-06,
      "loss": 0.0727,
      "step": 239410
    },
    {
      "epoch": 92.98,
      "learning_rate": 7.60284789644013e-06,
      "loss": 0.0003,
      "step": 239420
    },
    {
      "epoch": 92.98,
      "learning_rate": 7.602330097087379e-06,
      "loss": 0.2123,
      "step": 239430
    },
    {
      "epoch": 92.99,
      "learning_rate": 7.601812297734629e-06,
      "loss": 0.0421,
      "step": 239440
    },
    {
      "epoch": 92.99,
      "learning_rate": 7.6012944983818786e-06,
      "loss": 0.0249,
      "step": 239450
    },
    {
      "epoch": 92.99,
      "learning_rate": 7.6007766990291265e-06,
      "loss": 0.0528,
      "step": 239460
    },
    {
      "epoch": 93.0,
      "learning_rate": 7.600258899676376e-06,
      "loss": 0.0799,
      "step": 239470
    },
    {
      "epoch": 93.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.3666875958442688,
      "eval_runtime": 8.1844,
      "eval_samples_per_second": 444.135,
      "eval_steps_per_second": 55.593,
      "step": 239475
    },
    {
      "epoch": 93.0,
      "learning_rate": 7.599741100323626e-06,
      "loss": 0.0847,
      "step": 239480
    },
    {
      "epoch": 93.01,
      "learning_rate": 7.5992233009708745e-06,
      "loss": 0.0177,
      "step": 239490
    },
    {
      "epoch": 93.01,
      "learning_rate": 7.598705501618123e-06,
      "loss": 0.0653,
      "step": 239500
    },
    {
      "epoch": 93.01,
      "learning_rate": 7.598187702265373e-06,
      "loss": 0.0104,
      "step": 239510
    },
    {
      "epoch": 93.02,
      "learning_rate": 7.5976699029126225e-06,
      "loss": 0.155,
      "step": 239520
    },
    {
      "epoch": 93.02,
      "learning_rate": 7.597152103559871e-06,
      "loss": 0.0309,
      "step": 239530
    },
    {
      "epoch": 93.03,
      "learning_rate": 7.59663430420712e-06,
      "loss": 0.022,
      "step": 239540
    },
    {
      "epoch": 93.03,
      "learning_rate": 7.59611650485437e-06,
      "loss": 0.0003,
      "step": 239550
    },
    {
      "epoch": 93.03,
      "learning_rate": 7.595598705501619e-06,
      "loss": 0.091,
      "step": 239560
    },
    {
      "epoch": 93.04,
      "learning_rate": 7.595080906148868e-06,
      "loss": 0.0241,
      "step": 239570
    },
    {
      "epoch": 93.04,
      "learning_rate": 7.594563106796117e-06,
      "loss": 0.0645,
      "step": 239580
    },
    {
      "epoch": 93.04,
      "learning_rate": 7.5940453074433665e-06,
      "loss": 0.0503,
      "step": 239590
    },
    {
      "epoch": 93.05,
      "learning_rate": 7.593527508090616e-06,
      "loss": 0.0121,
      "step": 239600
    },
    {
      "epoch": 93.05,
      "learning_rate": 7.593009708737865e-06,
      "loss": 0.0461,
      "step": 239610
    },
    {
      "epoch": 93.06,
      "learning_rate": 7.592491909385114e-06,
      "loss": 0.009,
      "step": 239620
    },
    {
      "epoch": 93.06,
      "learning_rate": 7.591974110032363e-06,
      "loss": 0.0636,
      "step": 239630
    },
    {
      "epoch": 93.06,
      "learning_rate": 7.591456310679612e-06,
      "loss": 0.0899,
      "step": 239640
    },
    {
      "epoch": 93.07,
      "learning_rate": 7.590938511326862e-06,
      "loss": 0.0886,
      "step": 239650
    },
    {
      "epoch": 93.07,
      "learning_rate": 7.59042071197411e-06,
      "loss": 0.0136,
      "step": 239660
    },
    {
      "epoch": 93.08,
      "learning_rate": 7.58990291262136e-06,
      "loss": 0.1004,
      "step": 239670
    },
    {
      "epoch": 93.08,
      "learning_rate": 7.589385113268609e-06,
      "loss": 0.0558,
      "step": 239680
    },
    {
      "epoch": 93.08,
      "learning_rate": 7.588867313915858e-06,
      "loss": 0.0874,
      "step": 239690
    },
    {
      "epoch": 93.09,
      "learning_rate": 7.588349514563107e-06,
      "loss": 0.0249,
      "step": 239700
    },
    {
      "epoch": 93.09,
      "learning_rate": 7.587831715210357e-06,
      "loss": 0.0273,
      "step": 239710
    },
    {
      "epoch": 93.1,
      "learning_rate": 7.5873139158576056e-06,
      "loss": 0.0765,
      "step": 239720
    },
    {
      "epoch": 93.1,
      "learning_rate": 7.586796116504855e-06,
      "loss": 0.0842,
      "step": 239730
    },
    {
      "epoch": 93.1,
      "learning_rate": 7.586278317152104e-06,
      "loss": 0.019,
      "step": 239740
    },
    {
      "epoch": 93.11,
      "learning_rate": 7.5857605177993536e-06,
      "loss": 0.0001,
      "step": 239750
    },
    {
      "epoch": 93.11,
      "learning_rate": 7.585242718446602e-06,
      "loss": 0.1205,
      "step": 239760
    },
    {
      "epoch": 93.11,
      "learning_rate": 7.584724919093852e-06,
      "loss": 0.0064,
      "step": 239770
    },
    {
      "epoch": 93.12,
      "learning_rate": 7.584207119741101e-06,
      "loss": 0.0874,
      "step": 239780
    },
    {
      "epoch": 93.12,
      "learning_rate": 7.58368932038835e-06,
      "loss": 0.0045,
      "step": 239790
    },
    {
      "epoch": 93.13,
      "learning_rate": 7.583171521035599e-06,
      "loss": 0.088,
      "step": 239800
    },
    {
      "epoch": 93.13,
      "learning_rate": 7.582653721682849e-06,
      "loss": 0.2661,
      "step": 239810
    },
    {
      "epoch": 93.13,
      "learning_rate": 7.5821359223300975e-06,
      "loss": 0.0294,
      "step": 239820
    },
    {
      "epoch": 93.14,
      "learning_rate": 7.581618122977346e-06,
      "loss": 0.0255,
      "step": 239830
    },
    {
      "epoch": 93.14,
      "learning_rate": 7.581100323624596e-06,
      "loss": 0.0162,
      "step": 239840
    },
    {
      "epoch": 93.15,
      "learning_rate": 7.5805825242718455e-06,
      "loss": 0.0377,
      "step": 239850
    },
    {
      "epoch": 93.15,
      "learning_rate": 7.580064724919094e-06,
      "loss": 0.0382,
      "step": 239860
    },
    {
      "epoch": 93.15,
      "learning_rate": 7.579546925566343e-06,
      "loss": 0.0008,
      "step": 239870
    },
    {
      "epoch": 93.16,
      "learning_rate": 7.579029126213593e-06,
      "loss": 0.0974,
      "step": 239880
    },
    {
      "epoch": 93.16,
      "learning_rate": 7.578511326860842e-06,
      "loss": 0.0823,
      "step": 239890
    },
    {
      "epoch": 93.17,
      "learning_rate": 7.577993527508091e-06,
      "loss": 0.1039,
      "step": 239900
    },
    {
      "epoch": 93.17,
      "learning_rate": 7.57747572815534e-06,
      "loss": 0.1412,
      "step": 239910
    },
    {
      "epoch": 93.17,
      "learning_rate": 7.5769579288025894e-06,
      "loss": 0.0023,
      "step": 239920
    },
    {
      "epoch": 93.18,
      "learning_rate": 7.576440129449839e-06,
      "loss": 0.1495,
      "step": 239930
    },
    {
      "epoch": 93.18,
      "learning_rate": 7.575922330097089e-06,
      "loss": 0.3411,
      "step": 239940
    },
    {
      "epoch": 93.18,
      "learning_rate": 7.575404530744337e-06,
      "loss": 0.0223,
      "step": 239950
    },
    {
      "epoch": 93.19,
      "learning_rate": 7.574886731391586e-06,
      "loss": 0.0142,
      "step": 239960
    },
    {
      "epoch": 93.19,
      "learning_rate": 7.574368932038836e-06,
      "loss": 0.0438,
      "step": 239970
    },
    {
      "epoch": 93.2,
      "learning_rate": 7.5738511326860855e-06,
      "loss": 0.1137,
      "step": 239980
    },
    {
      "epoch": 93.2,
      "learning_rate": 7.573333333333333e-06,
      "loss": 0.0114,
      "step": 239990
    },
    {
      "epoch": 93.2,
      "learning_rate": 7.572815533980583e-06,
      "loss": 0.1191,
      "step": 240000
    },
    {
      "epoch": 93.21,
      "learning_rate": 7.572297734627833e-06,
      "loss": 0.1699,
      "step": 240010
    },
    {
      "epoch": 93.21,
      "learning_rate": 7.571779935275082e-06,
      "loss": 0.0154,
      "step": 240020
    },
    {
      "epoch": 93.22,
      "learning_rate": 7.57126213592233e-06,
      "loss": 0.0317,
      "step": 240030
    },
    {
      "epoch": 93.22,
      "learning_rate": 7.57074433656958e-06,
      "loss": 0.1658,
      "step": 240040
    },
    {
      "epoch": 93.22,
      "learning_rate": 7.570226537216829e-06,
      "loss": 0.1383,
      "step": 240050
    },
    {
      "epoch": 93.23,
      "learning_rate": 7.569708737864079e-06,
      "loss": 0.0081,
      "step": 240060
    },
    {
      "epoch": 93.23,
      "learning_rate": 7.569190938511327e-06,
      "loss": 0.0856,
      "step": 240070
    },
    {
      "epoch": 93.23,
      "learning_rate": 7.5686731391585766e-06,
      "loss": 0.0685,
      "step": 240080
    },
    {
      "epoch": 93.24,
      "learning_rate": 7.568155339805826e-06,
      "loss": 0.098,
      "step": 240090
    },
    {
      "epoch": 93.24,
      "learning_rate": 7.567637540453076e-06,
      "loss": 0.0179,
      "step": 240100
    },
    {
      "epoch": 93.25,
      "learning_rate": 7.567119741100324e-06,
      "loss": 0.0124,
      "step": 240110
    },
    {
      "epoch": 93.25,
      "learning_rate": 7.566601941747573e-06,
      "loss": 0.0172,
      "step": 240120
    },
    {
      "epoch": 93.25,
      "learning_rate": 7.566084142394823e-06,
      "loss": 0.0002,
      "step": 240130
    },
    {
      "epoch": 93.26,
      "learning_rate": 7.5655663430420726e-06,
      "loss": 0.012,
      "step": 240140
    },
    {
      "epoch": 93.26,
      "learning_rate": 7.5650485436893205e-06,
      "loss": 0.0619,
      "step": 240150
    },
    {
      "epoch": 93.27,
      "learning_rate": 7.56453074433657e-06,
      "loss": 0.1086,
      "step": 240160
    },
    {
      "epoch": 93.27,
      "learning_rate": 7.56401294498382e-06,
      "loss": 0.0752,
      "step": 240170
    },
    {
      "epoch": 93.27,
      "learning_rate": 7.563495145631069e-06,
      "loss": 0.0002,
      "step": 240180
    },
    {
      "epoch": 93.28,
      "learning_rate": 7.562977346278317e-06,
      "loss": 0.1772,
      "step": 240190
    },
    {
      "epoch": 93.28,
      "learning_rate": 7.562459546925567e-06,
      "loss": 0.0747,
      "step": 240200
    },
    {
      "epoch": 93.29,
      "learning_rate": 7.5619417475728165e-06,
      "loss": 0.014,
      "step": 240210
    },
    {
      "epoch": 93.29,
      "learning_rate": 7.561423948220066e-06,
      "loss": 0.0308,
      "step": 240220
    },
    {
      "epoch": 93.29,
      "learning_rate": 7.560906148867314e-06,
      "loss": 0.0016,
      "step": 240230
    },
    {
      "epoch": 93.3,
      "learning_rate": 7.560388349514564e-06,
      "loss": 0.0352,
      "step": 240240
    },
    {
      "epoch": 93.3,
      "learning_rate": 7.559870550161813e-06,
      "loss": 0.1753,
      "step": 240250
    },
    {
      "epoch": 93.3,
      "learning_rate": 7.559352750809063e-06,
      "loss": 0.1696,
      "step": 240260
    },
    {
      "epoch": 93.31,
      "learning_rate": 7.558834951456311e-06,
      "loss": 0.015,
      "step": 240270
    },
    {
      "epoch": 93.31,
      "learning_rate": 7.5583171521035605e-06,
      "loss": 0.0003,
      "step": 240280
    },
    {
      "epoch": 93.32,
      "learning_rate": 7.55779935275081e-06,
      "loss": 0.0192,
      "step": 240290
    },
    {
      "epoch": 93.32,
      "learning_rate": 7.557281553398059e-06,
      "loss": 0.0031,
      "step": 240300
    },
    {
      "epoch": 93.32,
      "learning_rate": 7.556763754045308e-06,
      "loss": 0.0145,
      "step": 240310
    },
    {
      "epoch": 93.33,
      "learning_rate": 7.556245954692557e-06,
      "loss": 0.0267,
      "step": 240320
    },
    {
      "epoch": 93.33,
      "learning_rate": 7.555728155339807e-06,
      "loss": 0.1786,
      "step": 240330
    },
    {
      "epoch": 93.34,
      "learning_rate": 7.555210355987056e-06,
      "loss": 0.0021,
      "step": 240340
    },
    {
      "epoch": 93.34,
      "learning_rate": 7.554692556634304e-06,
      "loss": 0.0002,
      "step": 240350
    },
    {
      "epoch": 93.34,
      "learning_rate": 7.554174757281554e-06,
      "loss": 0.042,
      "step": 240360
    },
    {
      "epoch": 93.35,
      "learning_rate": 7.553656957928804e-06,
      "loss": 0.0099,
      "step": 240370
    },
    {
      "epoch": 93.35,
      "learning_rate": 7.553139158576052e-06,
      "loss": 0.0591,
      "step": 240380
    },
    {
      "epoch": 93.36,
      "learning_rate": 7.552621359223301e-06,
      "loss": 0.1944,
      "step": 240390
    },
    {
      "epoch": 93.36,
      "learning_rate": 7.552103559870551e-06,
      "loss": 0.0894,
      "step": 240400
    },
    {
      "epoch": 93.36,
      "learning_rate": 7.5515857605178e-06,
      "loss": 0.0043,
      "step": 240410
    },
    {
      "epoch": 93.37,
      "learning_rate": 7.551067961165049e-06,
      "loss": 0.0029,
      "step": 240420
    },
    {
      "epoch": 93.37,
      "learning_rate": 7.550550161812298e-06,
      "loss": 0.0022,
      "step": 240430
    },
    {
      "epoch": 93.37,
      "learning_rate": 7.5500323624595476e-06,
      "loss": 0.0932,
      "step": 240440
    },
    {
      "epoch": 93.38,
      "learning_rate": 7.549514563106796e-06,
      "loss": 0.0232,
      "step": 240450
    },
    {
      "epoch": 93.38,
      "learning_rate": 7.548996763754046e-06,
      "loss": 0.0162,
      "step": 240460
    },
    {
      "epoch": 93.39,
      "learning_rate": 7.548478964401295e-06,
      "loss": 0.1003,
      "step": 240470
    },
    {
      "epoch": 93.39,
      "learning_rate": 7.547961165048544e-06,
      "loss": 0.0134,
      "step": 240480
    },
    {
      "epoch": 93.39,
      "learning_rate": 7.547443365695793e-06,
      "loss": 0.0194,
      "step": 240490
    },
    {
      "epoch": 93.4,
      "learning_rate": 7.546925566343043e-06,
      "loss": 0.077,
      "step": 240500
    },
    {
      "epoch": 93.4,
      "learning_rate": 7.546407766990292e-06,
      "loss": 0.0186,
      "step": 240510
    },
    {
      "epoch": 93.41,
      "learning_rate": 7.545889967637541e-06,
      "loss": 0.0992,
      "step": 240520
    },
    {
      "epoch": 93.41,
      "learning_rate": 7.54537216828479e-06,
      "loss": 0.0529,
      "step": 240530
    },
    {
      "epoch": 93.41,
      "learning_rate": 7.5448543689320395e-06,
      "loss": 0.0242,
      "step": 240540
    },
    {
      "epoch": 93.42,
      "learning_rate": 7.544336569579289e-06,
      "loss": 0.0992,
      "step": 240550
    },
    {
      "epoch": 93.42,
      "learning_rate": 7.543818770226538e-06,
      "loss": 0.01,
      "step": 240560
    },
    {
      "epoch": 93.43,
      "learning_rate": 7.543300970873787e-06,
      "loss": 0.0235,
      "step": 240570
    },
    {
      "epoch": 93.43,
      "learning_rate": 7.542783171521036e-06,
      "loss": 0.0125,
      "step": 240580
    },
    {
      "epoch": 93.43,
      "learning_rate": 7.542265372168286e-06,
      "loss": 0.01,
      "step": 240590
    },
    {
      "epoch": 93.44,
      "learning_rate": 7.541747572815535e-06,
      "loss": 0.084,
      "step": 240600
    },
    {
      "epoch": 93.44,
      "learning_rate": 7.5412297734627834e-06,
      "loss": 0.2215,
      "step": 240610
    },
    {
      "epoch": 93.44,
      "learning_rate": 7.540711974110033e-06,
      "loss": 0.0557,
      "step": 240620
    },
    {
      "epoch": 93.45,
      "learning_rate": 7.540194174757283e-06,
      "loss": 0.0161,
      "step": 240630
    },
    {
      "epoch": 93.45,
      "learning_rate": 7.539676375404531e-06,
      "loss": 0.0151,
      "step": 240640
    },
    {
      "epoch": 93.46,
      "learning_rate": 7.53915857605178e-06,
      "loss": 0.0087,
      "step": 240650
    },
    {
      "epoch": 93.46,
      "learning_rate": 7.53864077669903e-06,
      "loss": 0.0725,
      "step": 240660
    },
    {
      "epoch": 93.46,
      "learning_rate": 7.5381229773462795e-06,
      "loss": 0.1285,
      "step": 240670
    },
    {
      "epoch": 93.47,
      "learning_rate": 7.537605177993527e-06,
      "loss": 0.0007,
      "step": 240680
    },
    {
      "epoch": 93.47,
      "learning_rate": 7.537087378640777e-06,
      "loss": 0.0937,
      "step": 240690
    },
    {
      "epoch": 93.48,
      "learning_rate": 7.536569579288027e-06,
      "loss": 0.0471,
      "step": 240700
    },
    {
      "epoch": 93.48,
      "learning_rate": 7.536051779935276e-06,
      "loss": 0.0323,
      "step": 240710
    },
    {
      "epoch": 93.48,
      "learning_rate": 7.535533980582524e-06,
      "loss": 0.0824,
      "step": 240720
    },
    {
      "epoch": 93.49,
      "learning_rate": 7.535016181229774e-06,
      "loss": 0.0074,
      "step": 240730
    },
    {
      "epoch": 93.49,
      "learning_rate": 7.534498381877023e-06,
      "loss": 0.0254,
      "step": 240740
    },
    {
      "epoch": 93.5,
      "learning_rate": 7.533980582524273e-06,
      "loss": 0.0317,
      "step": 240750
    },
    {
      "epoch": 93.5,
      "learning_rate": 7.533462783171521e-06,
      "loss": 0.1572,
      "step": 240760
    },
    {
      "epoch": 93.5,
      "learning_rate": 7.5329449838187706e-06,
      "loss": 0.0002,
      "step": 240770
    },
    {
      "epoch": 93.51,
      "learning_rate": 7.53242718446602e-06,
      "loss": 0.0099,
      "step": 240780
    },
    {
      "epoch": 93.51,
      "learning_rate": 7.53190938511327e-06,
      "loss": 0.0886,
      "step": 240790
    },
    {
      "epoch": 93.51,
      "learning_rate": 7.531391585760518e-06,
      "loss": 0.0727,
      "step": 240800
    },
    {
      "epoch": 93.52,
      "learning_rate": 7.530873786407767e-06,
      "loss": 0.0702,
      "step": 240810
    },
    {
      "epoch": 93.52,
      "learning_rate": 7.530355987055017e-06,
      "loss": 0.0111,
      "step": 240820
    },
    {
      "epoch": 93.53,
      "learning_rate": 7.5298381877022666e-06,
      "loss": 0.0014,
      "step": 240830
    },
    {
      "epoch": 93.53,
      "learning_rate": 7.5293203883495145e-06,
      "loss": 0.0437,
      "step": 240840
    },
    {
      "epoch": 93.53,
      "learning_rate": 7.528802588996764e-06,
      "loss": 0.035,
      "step": 240850
    },
    {
      "epoch": 93.54,
      "learning_rate": 7.528284789644014e-06,
      "loss": 0.0566,
      "step": 240860
    },
    {
      "epoch": 93.54,
      "learning_rate": 7.527766990291263e-06,
      "loss": 0.0729,
      "step": 240870
    },
    {
      "epoch": 93.55,
      "learning_rate": 7.527249190938511e-06,
      "loss": 0.091,
      "step": 240880
    },
    {
      "epoch": 93.55,
      "learning_rate": 7.526731391585761e-06,
      "loss": 0.045,
      "step": 240890
    },
    {
      "epoch": 93.55,
      "learning_rate": 7.5262135922330105e-06,
      "loss": 0.0162,
      "step": 240900
    },
    {
      "epoch": 93.56,
      "learning_rate": 7.52569579288026e-06,
      "loss": 0.1122,
      "step": 240910
    },
    {
      "epoch": 93.56,
      "learning_rate": 7.525177993527508e-06,
      "loss": 0.1686,
      "step": 240920
    },
    {
      "epoch": 93.57,
      "learning_rate": 7.524660194174758e-06,
      "loss": 0.1209,
      "step": 240930
    },
    {
      "epoch": 93.57,
      "learning_rate": 7.524142394822007e-06,
      "loss": 0.0456,
      "step": 240940
    },
    {
      "epoch": 93.57,
      "learning_rate": 7.523624595469257e-06,
      "loss": 0.0682,
      "step": 240950
    },
    {
      "epoch": 93.58,
      "learning_rate": 7.523106796116505e-06,
      "loss": 0.0096,
      "step": 240960
    },
    {
      "epoch": 93.58,
      "learning_rate": 7.5225889967637544e-06,
      "loss": 0.0973,
      "step": 240970
    },
    {
      "epoch": 93.58,
      "learning_rate": 7.522071197411004e-06,
      "loss": 0.0833,
      "step": 240980
    },
    {
      "epoch": 93.59,
      "learning_rate": 7.521553398058254e-06,
      "loss": 0.1033,
      "step": 240990
    },
    {
      "epoch": 93.59,
      "learning_rate": 7.521035598705502e-06,
      "loss": 0.0123,
      "step": 241000
    },
    {
      "epoch": 93.6,
      "learning_rate": 7.520517799352751e-06,
      "loss": 0.0023,
      "step": 241010
    },
    {
      "epoch": 93.6,
      "learning_rate": 7.520000000000001e-06,
      "loss": 0.0271,
      "step": 241020
    },
    {
      "epoch": 93.6,
      "learning_rate": 7.5194822006472505e-06,
      "loss": 0.0617,
      "step": 241030
    },
    {
      "epoch": 93.61,
      "learning_rate": 7.518964401294498e-06,
      "loss": 0.1628,
      "step": 241040
    },
    {
      "epoch": 93.61,
      "learning_rate": 7.518446601941748e-06,
      "loss": 0.1281,
      "step": 241050
    },
    {
      "epoch": 93.62,
      "learning_rate": 7.517928802588998e-06,
      "loss": 0.1009,
      "step": 241060
    },
    {
      "epoch": 93.62,
      "learning_rate": 7.517411003236247e-06,
      "loss": 0.1651,
      "step": 241070
    },
    {
      "epoch": 93.62,
      "learning_rate": 7.516893203883496e-06,
      "loss": 0.0052,
      "step": 241080
    },
    {
      "epoch": 93.63,
      "learning_rate": 7.516375404530745e-06,
      "loss": 0.1503,
      "step": 241090
    },
    {
      "epoch": 93.63,
      "learning_rate": 7.515857605177994e-06,
      "loss": 0.0387,
      "step": 241100
    },
    {
      "epoch": 93.63,
      "learning_rate": 7.515339805825243e-06,
      "loss": 0.0873,
      "step": 241110
    },
    {
      "epoch": 93.64,
      "learning_rate": 7.514822006472493e-06,
      "loss": 0.0743,
      "step": 241120
    },
    {
      "epoch": 93.64,
      "learning_rate": 7.5143042071197416e-06,
      "loss": 0.041,
      "step": 241130
    },
    {
      "epoch": 93.65,
      "learning_rate": 7.513786407766991e-06,
      "loss": 0.0272,
      "step": 241140
    },
    {
      "epoch": 93.65,
      "learning_rate": 7.51326860841424e-06,
      "loss": 0.0215,
      "step": 241150
    },
    {
      "epoch": 93.65,
      "learning_rate": 7.5127508090614896e-06,
      "loss": 0.0326,
      "step": 241160
    },
    {
      "epoch": 93.66,
      "learning_rate": 7.512233009708738e-06,
      "loss": 0.0009,
      "step": 241170
    },
    {
      "epoch": 93.66,
      "learning_rate": 7.511715210355988e-06,
      "loss": 0.0856,
      "step": 241180
    },
    {
      "epoch": 93.67,
      "learning_rate": 7.511197411003237e-06,
      "loss": 0.086,
      "step": 241190
    },
    {
      "epoch": 93.67,
      "learning_rate": 7.510679611650486e-06,
      "loss": 0.0112,
      "step": 241200
    },
    {
      "epoch": 93.67,
      "learning_rate": 7.510161812297735e-06,
      "loss": 0.0083,
      "step": 241210
    },
    {
      "epoch": 93.68,
      "learning_rate": 7.509644012944985e-06,
      "loss": 0.0375,
      "step": 241220
    },
    {
      "epoch": 93.68,
      "learning_rate": 7.5091262135922335e-06,
      "loss": 0.0962,
      "step": 241230
    },
    {
      "epoch": 93.69,
      "learning_rate": 7.508608414239483e-06,
      "loss": 0.0248,
      "step": 241240
    },
    {
      "epoch": 93.69,
      "learning_rate": 7.508090614886732e-06,
      "loss": 0.0005,
      "step": 241250
    },
    {
      "epoch": 93.69,
      "learning_rate": 7.5075728155339815e-06,
      "loss": 0.034,
      "step": 241260
    },
    {
      "epoch": 93.7,
      "learning_rate": 7.50705501618123e-06,
      "loss": 0.0133,
      "step": 241270
    },
    {
      "epoch": 93.7,
      "learning_rate": 7.50653721682848e-06,
      "loss": 0.0233,
      "step": 241280
    },
    {
      "epoch": 93.7,
      "learning_rate": 7.506019417475729e-06,
      "loss": 0.0687,
      "step": 241290
    },
    {
      "epoch": 93.71,
      "learning_rate": 7.5055016181229774e-06,
      "loss": 0.033,
      "step": 241300
    },
    {
      "epoch": 93.71,
      "learning_rate": 7.504983818770227e-06,
      "loss": 0.0888,
      "step": 241310
    },
    {
      "epoch": 93.72,
      "learning_rate": 7.504466019417477e-06,
      "loss": 0.1394,
      "step": 241320
    },
    {
      "epoch": 93.72,
      "learning_rate": 7.5039482200647255e-06,
      "loss": 0.105,
      "step": 241330
    },
    {
      "epoch": 93.72,
      "learning_rate": 7.503430420711974e-06,
      "loss": 0.0807,
      "step": 241340
    },
    {
      "epoch": 93.73,
      "learning_rate": 7.502912621359224e-06,
      "loss": 0.0185,
      "step": 241350
    },
    {
      "epoch": 93.73,
      "learning_rate": 7.5023948220064735e-06,
      "loss": 0.1281,
      "step": 241360
    },
    {
      "epoch": 93.74,
      "learning_rate": 7.501877022653722e-06,
      "loss": 0.1563,
      "step": 241370
    },
    {
      "epoch": 93.74,
      "learning_rate": 7.501359223300971e-06,
      "loss": 0.0643,
      "step": 241380
    },
    {
      "epoch": 93.74,
      "learning_rate": 7.500841423948221e-06,
      "loss": 0.0445,
      "step": 241390
    },
    {
      "epoch": 93.75,
      "learning_rate": 7.50032362459547e-06,
      "loss": 0.0522,
      "step": 241400
    },
    {
      "epoch": 93.75,
      "learning_rate": 7.499805825242719e-06,
      "loss": 0.1032,
      "step": 241410
    },
    {
      "epoch": 93.76,
      "learning_rate": 7.499288025889968e-06,
      "loss": 0.065,
      "step": 241420
    },
    {
      "epoch": 93.76,
      "learning_rate": 7.498770226537217e-06,
      "loss": 0.0091,
      "step": 241430
    },
    {
      "epoch": 93.76,
      "learning_rate": 7.498252427184467e-06,
      "loss": 0.0704,
      "step": 241440
    },
    {
      "epoch": 93.77,
      "learning_rate": 7.497734627831715e-06,
      "loss": 0.0997,
      "step": 241450
    },
    {
      "epoch": 93.77,
      "learning_rate": 7.4972168284789646e-06,
      "loss": 0.0282,
      "step": 241460
    },
    {
      "epoch": 93.77,
      "learning_rate": 7.496699029126214e-06,
      "loss": 0.0573,
      "step": 241470
    },
    {
      "epoch": 93.78,
      "learning_rate": 7.496181229773464e-06,
      "loss": 0.0088,
      "step": 241480
    },
    {
      "epoch": 93.78,
      "learning_rate": 7.495663430420712e-06,
      "loss": 0.0809,
      "step": 241490
    },
    {
      "epoch": 93.79,
      "learning_rate": 7.495145631067961e-06,
      "loss": 0.1551,
      "step": 241500
    },
    {
      "epoch": 93.79,
      "learning_rate": 7.494627831715211e-06,
      "loss": 0.0007,
      "step": 241510
    },
    {
      "epoch": 93.79,
      "learning_rate": 7.4941100323624606e-06,
      "loss": 0.2059,
      "step": 241520
    },
    {
      "epoch": 93.8,
      "learning_rate": 7.4935922330097085e-06,
      "loss": 0.0006,
      "step": 241530
    },
    {
      "epoch": 93.8,
      "learning_rate": 7.493074433656958e-06,
      "loss": 0.0767,
      "step": 241540
    },
    {
      "epoch": 93.81,
      "learning_rate": 7.492556634304208e-06,
      "loss": 0.0301,
      "step": 241550
    },
    {
      "epoch": 93.81,
      "learning_rate": 7.492038834951457e-06,
      "loss": 0.0008,
      "step": 241560
    },
    {
      "epoch": 93.81,
      "learning_rate": 7.491521035598705e-06,
      "loss": 0.027,
      "step": 241570
    },
    {
      "epoch": 93.82,
      "learning_rate": 7.491003236245955e-06,
      "loss": 0.0486,
      "step": 241580
    },
    {
      "epoch": 93.82,
      "learning_rate": 7.4904854368932045e-06,
      "loss": 0.0921,
      "step": 241590
    },
    {
      "epoch": 93.83,
      "learning_rate": 7.489967637540454e-06,
      "loss": 0.0381,
      "step": 241600
    },
    {
      "epoch": 93.83,
      "learning_rate": 7.489449838187702e-06,
      "loss": 0.1182,
      "step": 241610
    },
    {
      "epoch": 93.83,
      "learning_rate": 7.488932038834952e-06,
      "loss": 0.0598,
      "step": 241620
    },
    {
      "epoch": 93.84,
      "learning_rate": 7.488414239482201e-06,
      "loss": 0.0314,
      "step": 241630
    },
    {
      "epoch": 93.84,
      "learning_rate": 7.487896440129451e-06,
      "loss": 0.073,
      "step": 241640
    },
    {
      "epoch": 93.84,
      "learning_rate": 7.4873786407767005e-06,
      "loss": 0.2092,
      "step": 241650
    },
    {
      "epoch": 93.85,
      "learning_rate": 7.4868608414239484e-06,
      "loss": 0.0529,
      "step": 241660
    },
    {
      "epoch": 93.85,
      "learning_rate": 7.486343042071198e-06,
      "loss": 0.0385,
      "step": 241670
    },
    {
      "epoch": 93.86,
      "learning_rate": 7.485825242718448e-06,
      "loss": 0.024,
      "step": 241680
    },
    {
      "epoch": 93.86,
      "learning_rate": 7.485307443365697e-06,
      "loss": 0.0363,
      "step": 241690
    },
    {
      "epoch": 93.86,
      "learning_rate": 7.484789644012945e-06,
      "loss": 0.0069,
      "step": 241700
    },
    {
      "epoch": 93.87,
      "learning_rate": 7.484271844660195e-06,
      "loss": 0.115,
      "step": 241710
    },
    {
      "epoch": 93.87,
      "learning_rate": 7.4837540453074445e-06,
      "loss": 0.1405,
      "step": 241720
    },
    {
      "epoch": 93.88,
      "learning_rate": 7.483236245954694e-06,
      "loss": 0.1307,
      "step": 241730
    },
    {
      "epoch": 93.88,
      "learning_rate": 7.482718446601942e-06,
      "loss": 0.0536,
      "step": 241740
    },
    {
      "epoch": 93.88,
      "learning_rate": 7.482200647249192e-06,
      "loss": 0.0132,
      "step": 241750
    },
    {
      "epoch": 93.89,
      "learning_rate": 7.481682847896441e-06,
      "loss": 0.025,
      "step": 241760
    },
    {
      "epoch": 93.89,
      "learning_rate": 7.48116504854369e-06,
      "loss": 0.0211,
      "step": 241770
    },
    {
      "epoch": 93.9,
      "learning_rate": 7.480647249190939e-06,
      "loss": 0.1992,
      "step": 241780
    },
    {
      "epoch": 93.9,
      "learning_rate": 7.480129449838188e-06,
      "loss": 0.0371,
      "step": 241790
    },
    {
      "epoch": 93.9,
      "learning_rate": 7.479611650485438e-06,
      "loss": 0.0007,
      "step": 241800
    },
    {
      "epoch": 93.91,
      "learning_rate": 7.479093851132687e-06,
      "loss": 0.0184,
      "step": 241810
    },
    {
      "epoch": 93.91,
      "learning_rate": 7.4785760517799356e-06,
      "loss": 0.011,
      "step": 241820
    },
    {
      "epoch": 93.91,
      "learning_rate": 7.478058252427185e-06,
      "loss": 0.102,
      "step": 241830
    },
    {
      "epoch": 93.92,
      "learning_rate": 7.477540453074435e-06,
      "loss": 0.0038,
      "step": 241840
    },
    {
      "epoch": 93.92,
      "learning_rate": 7.4770226537216836e-06,
      "loss": 0.0261,
      "step": 241850
    },
    {
      "epoch": 93.93,
      "learning_rate": 7.476504854368932e-06,
      "loss": 0.0107,
      "step": 241860
    },
    {
      "epoch": 93.93,
      "learning_rate": 7.475987055016182e-06,
      "loss": 0.0162,
      "step": 241870
    },
    {
      "epoch": 93.93,
      "learning_rate": 7.4754692556634316e-06,
      "loss": 0.0137,
      "step": 241880
    },
    {
      "epoch": 93.94,
      "learning_rate": 7.47495145631068e-06,
      "loss": 0.0035,
      "step": 241890
    },
    {
      "epoch": 93.94,
      "learning_rate": 7.474433656957929e-06,
      "loss": 0.0058,
      "step": 241900
    },
    {
      "epoch": 93.95,
      "learning_rate": 7.473915857605179e-06,
      "loss": 0.0493,
      "step": 241910
    },
    {
      "epoch": 93.95,
      "learning_rate": 7.4733980582524275e-06,
      "loss": 0.1736,
      "step": 241920
    },
    {
      "epoch": 93.95,
      "learning_rate": 7.472880258899677e-06,
      "loss": 0.0522,
      "step": 241930
    },
    {
      "epoch": 93.96,
      "learning_rate": 7.472362459546926e-06,
      "loss": 0.1151,
      "step": 241940
    },
    {
      "epoch": 93.96,
      "learning_rate": 7.4718446601941755e-06,
      "loss": 0.0801,
      "step": 241950
    },
    {
      "epoch": 93.97,
      "learning_rate": 7.471326860841424e-06,
      "loss": 0.0097,
      "step": 241960
    },
    {
      "epoch": 93.97,
      "learning_rate": 7.470809061488674e-06,
      "loss": 0.0285,
      "step": 241970
    },
    {
      "epoch": 93.97,
      "learning_rate": 7.470291262135923e-06,
      "loss": 0.0894,
      "step": 241980
    },
    {
      "epoch": 93.98,
      "learning_rate": 7.469773462783172e-06,
      "loss": 0.092,
      "step": 241990
    },
    {
      "epoch": 93.98,
      "learning_rate": 7.469255663430421e-06,
      "loss": 0.0461,
      "step": 242000
    },
    {
      "epoch": 93.98,
      "learning_rate": 7.468737864077671e-06,
      "loss": 0.0457,
      "step": 242010
    },
    {
      "epoch": 93.99,
      "learning_rate": 7.4682200647249195e-06,
      "loss": 0.0706,
      "step": 242020
    },
    {
      "epoch": 93.99,
      "learning_rate": 7.467702265372169e-06,
      "loss": 0.218,
      "step": 242030
    },
    {
      "epoch": 94.0,
      "learning_rate": 7.467184466019418e-06,
      "loss": 0.0713,
      "step": 242040
    },
    {
      "epoch": 94.0,
      "learning_rate": 7.4666666666666675e-06,
      "loss": 0.0413,
      "step": 242050
    },
    {
      "epoch": 94.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.3603576719760895,
      "eval_runtime": 8.2048,
      "eval_samples_per_second": 443.032,
      "eval_steps_per_second": 55.455,
      "step": 242050
    },
    {
      "epoch": 94.0,
      "learning_rate": 7.466148867313916e-06,
      "loss": 0.0198,
      "step": 242060
    },
    {
      "epoch": 94.01,
      "learning_rate": 7.465631067961166e-06,
      "loss": 0.0501,
      "step": 242070
    },
    {
      "epoch": 94.01,
      "learning_rate": 7.465113268608415e-06,
      "loss": 0.0263,
      "step": 242080
    },
    {
      "epoch": 94.02,
      "learning_rate": 7.464595469255664e-06,
      "loss": 0.0108,
      "step": 242090
    },
    {
      "epoch": 94.02,
      "learning_rate": 7.464077669902913e-06,
      "loss": 0.0484,
      "step": 242100
    },
    {
      "epoch": 94.02,
      "learning_rate": 7.463559870550162e-06,
      "loss": 0.0825,
      "step": 242110
    },
    {
      "epoch": 94.03,
      "learning_rate": 7.463042071197411e-06,
      "loss": 0.1299,
      "step": 242120
    },
    {
      "epoch": 94.03,
      "learning_rate": 7.462524271844661e-06,
      "loss": 0.0002,
      "step": 242130
    },
    {
      "epoch": 94.03,
      "learning_rate": 7.46200647249191e-06,
      "loss": 0.0347,
      "step": 242140
    },
    {
      "epoch": 94.04,
      "learning_rate": 7.4614886731391586e-06,
      "loss": 0.0961,
      "step": 242150
    },
    {
      "epoch": 94.04,
      "learning_rate": 7.460970873786408e-06,
      "loss": 0.0229,
      "step": 242160
    },
    {
      "epoch": 94.05,
      "learning_rate": 7.460453074433658e-06,
      "loss": 0.0746,
      "step": 242170
    },
    {
      "epoch": 94.05,
      "learning_rate": 7.459935275080907e-06,
      "loss": 0.0715,
      "step": 242180
    },
    {
      "epoch": 94.05,
      "learning_rate": 7.459417475728155e-06,
      "loss": 0.0558,
      "step": 242190
    },
    {
      "epoch": 94.06,
      "learning_rate": 7.458899676375405e-06,
      "loss": 0.0072,
      "step": 242200
    },
    {
      "epoch": 94.06,
      "learning_rate": 7.4583818770226546e-06,
      "loss": 0.0921,
      "step": 242210
    },
    {
      "epoch": 94.07,
      "learning_rate": 7.457864077669904e-06,
      "loss": 0.1365,
      "step": 242220
    },
    {
      "epoch": 94.07,
      "learning_rate": 7.457346278317152e-06,
      "loss": 0.0721,
      "step": 242230
    },
    {
      "epoch": 94.07,
      "learning_rate": 7.456828478964402e-06,
      "loss": 0.0445,
      "step": 242240
    },
    {
      "epoch": 94.08,
      "learning_rate": 7.456310679611651e-06,
      "loss": 0.0086,
      "step": 242250
    },
    {
      "epoch": 94.08,
      "learning_rate": 7.455792880258901e-06,
      "loss": 0.0115,
      "step": 242260
    },
    {
      "epoch": 94.09,
      "learning_rate": 7.455275080906149e-06,
      "loss": 0.2317,
      "step": 242270
    },
    {
      "epoch": 94.09,
      "learning_rate": 7.4547572815533985e-06,
      "loss": 0.1381,
      "step": 242280
    },
    {
      "epoch": 94.09,
      "learning_rate": 7.454239482200648e-06,
      "loss": 0.0487,
      "step": 242290
    },
    {
      "epoch": 94.1,
      "learning_rate": 7.453721682847898e-06,
      "loss": 0.0002,
      "step": 242300
    },
    {
      "epoch": 94.1,
      "learning_rate": 7.453203883495146e-06,
      "loss": 0.002,
      "step": 242310
    },
    {
      "epoch": 94.1,
      "learning_rate": 7.452686084142395e-06,
      "loss": 0.0118,
      "step": 242320
    },
    {
      "epoch": 94.11,
      "learning_rate": 7.452168284789645e-06,
      "loss": 0.0993,
      "step": 242330
    },
    {
      "epoch": 94.11,
      "learning_rate": 7.4516504854368945e-06,
      "loss": 0.0226,
      "step": 242340
    },
    {
      "epoch": 94.12,
      "learning_rate": 7.4511326860841424e-06,
      "loss": 0.0863,
      "step": 242350
    },
    {
      "epoch": 94.12,
      "learning_rate": 7.450614886731392e-06,
      "loss": 0.0113,
      "step": 242360
    },
    {
      "epoch": 94.12,
      "learning_rate": 7.450097087378642e-06,
      "loss": 0.0726,
      "step": 242370
    },
    {
      "epoch": 94.13,
      "learning_rate": 7.449579288025891e-06,
      "loss": 0.0101,
      "step": 242380
    },
    {
      "epoch": 94.13,
      "learning_rate": 7.449061488673139e-06,
      "loss": 0.0206,
      "step": 242390
    },
    {
      "epoch": 94.14,
      "learning_rate": 7.448543689320389e-06,
      "loss": 0.0318,
      "step": 242400
    },
    {
      "epoch": 94.14,
      "learning_rate": 7.4480258899676385e-06,
      "loss": 0.0396,
      "step": 242410
    },
    {
      "epoch": 94.14,
      "learning_rate": 7.447508090614888e-06,
      "loss": 0.0006,
      "step": 242420
    },
    {
      "epoch": 94.15,
      "learning_rate": 7.446990291262136e-06,
      "loss": 0.1187,
      "step": 242430
    },
    {
      "epoch": 94.15,
      "learning_rate": 7.446472491909386e-06,
      "loss": 0.0118,
      "step": 242440
    },
    {
      "epoch": 94.16,
      "learning_rate": 7.445954692556635e-06,
      "loss": 0.0432,
      "step": 242450
    },
    {
      "epoch": 94.16,
      "learning_rate": 7.445436893203885e-06,
      "loss": 0.0161,
      "step": 242460
    },
    {
      "epoch": 94.16,
      "learning_rate": 7.444919093851133e-06,
      "loss": 0.0688,
      "step": 242470
    },
    {
      "epoch": 94.17,
      "learning_rate": 7.444401294498382e-06,
      "loss": 0.0901,
      "step": 242480
    },
    {
      "epoch": 94.17,
      "learning_rate": 7.443883495145632e-06,
      "loss": 0.1486,
      "step": 242490
    },
    {
      "epoch": 94.17,
      "learning_rate": 7.443365695792882e-06,
      "loss": 0.0165,
      "step": 242500
    },
    {
      "epoch": 94.18,
      "learning_rate": 7.4428478964401296e-06,
      "loss": 0.0394,
      "step": 242510
    },
    {
      "epoch": 94.18,
      "learning_rate": 7.442330097087379e-06,
      "loss": 0.0128,
      "step": 242520
    },
    {
      "epoch": 94.19,
      "learning_rate": 7.441812297734629e-06,
      "loss": 0.0329,
      "step": 242530
    },
    {
      "epoch": 94.19,
      "learning_rate": 7.441294498381878e-06,
      "loss": 0.1185,
      "step": 242540
    },
    {
      "epoch": 94.19,
      "learning_rate": 7.440776699029126e-06,
      "loss": 0.0101,
      "step": 242550
    },
    {
      "epoch": 94.2,
      "learning_rate": 7.440258899676376e-06,
      "loss": 0.0752,
      "step": 242560
    },
    {
      "epoch": 94.2,
      "learning_rate": 7.4397411003236256e-06,
      "loss": 0.0668,
      "step": 242570
    },
    {
      "epoch": 94.21,
      "learning_rate": 7.439223300970874e-06,
      "loss": 0.068,
      "step": 242580
    },
    {
      "epoch": 94.21,
      "learning_rate": 7.438705501618123e-06,
      "loss": 0.0001,
      "step": 242590
    },
    {
      "epoch": 94.21,
      "learning_rate": 7.438187702265373e-06,
      "loss": 0.0212,
      "step": 242600
    },
    {
      "epoch": 94.22,
      "learning_rate": 7.437669902912622e-06,
      "loss": 0.0235,
      "step": 242610
    },
    {
      "epoch": 94.22,
      "learning_rate": 7.437152103559871e-06,
      "loss": 0.1304,
      "step": 242620
    },
    {
      "epoch": 94.23,
      "learning_rate": 7.43663430420712e-06,
      "loss": 0.115,
      "step": 242630
    },
    {
      "epoch": 94.23,
      "learning_rate": 7.4361165048543695e-06,
      "loss": 0.0859,
      "step": 242640
    },
    {
      "epoch": 94.23,
      "learning_rate": 7.435598705501619e-06,
      "loss": 0.0499,
      "step": 242650
    },
    {
      "epoch": 94.24,
      "learning_rate": 7.435080906148868e-06,
      "loss": 0.0459,
      "step": 242660
    },
    {
      "epoch": 94.24,
      "learning_rate": 7.434563106796117e-06,
      "loss": 0.0347,
      "step": 242670
    },
    {
      "epoch": 94.24,
      "learning_rate": 7.434045307443366e-06,
      "loss": 0.0438,
      "step": 242680
    },
    {
      "epoch": 94.25,
      "learning_rate": 7.433527508090616e-06,
      "loss": 0.0493,
      "step": 242690
    },
    {
      "epoch": 94.25,
      "learning_rate": 7.433009708737865e-06,
      "loss": 0.1844,
      "step": 242700
    },
    {
      "epoch": 94.26,
      "learning_rate": 7.4324919093851134e-06,
      "loss": 0.0648,
      "step": 242710
    },
    {
      "epoch": 94.26,
      "learning_rate": 7.431974110032363e-06,
      "loss": 0.0302,
      "step": 242720
    },
    {
      "epoch": 94.26,
      "learning_rate": 7.431456310679612e-06,
      "loss": 0.1572,
      "step": 242730
    },
    {
      "epoch": 94.27,
      "learning_rate": 7.4309385113268615e-06,
      "loss": 0.0766,
      "step": 242740
    },
    {
      "epoch": 94.27,
      "learning_rate": 7.430420711974111e-06,
      "loss": 0.1026,
      "step": 242750
    },
    {
      "epoch": 94.28,
      "learning_rate": 7.42990291262136e-06,
      "loss": 0.0133,
      "step": 242760
    },
    {
      "epoch": 94.28,
      "learning_rate": 7.429385113268609e-06,
      "loss": 0.0824,
      "step": 242770
    },
    {
      "epoch": 94.28,
      "learning_rate": 7.428867313915858e-06,
      "loss": 0.0358,
      "step": 242780
    },
    {
      "epoch": 94.29,
      "learning_rate": 7.428349514563108e-06,
      "loss": 0.0023,
      "step": 242790
    },
    {
      "epoch": 94.29,
      "learning_rate": 7.427831715210357e-06,
      "loss": 0.0188,
      "step": 242800
    },
    {
      "epoch": 94.3,
      "learning_rate": 7.427313915857605e-06,
      "loss": 0.0494,
      "step": 242810
    },
    {
      "epoch": 94.3,
      "learning_rate": 7.426796116504855e-06,
      "loss": 0.0173,
      "step": 242820
    },
    {
      "epoch": 94.3,
      "learning_rate": 7.426278317152105e-06,
      "loss": 0.0806,
      "step": 242830
    },
    {
      "epoch": 94.31,
      "learning_rate": 7.425760517799353e-06,
      "loss": 0.0699,
      "step": 242840
    },
    {
      "epoch": 94.31,
      "learning_rate": 7.425242718446602e-06,
      "loss": 0.1913,
      "step": 242850
    },
    {
      "epoch": 94.31,
      "learning_rate": 7.424724919093852e-06,
      "loss": 0.0362,
      "step": 242860
    },
    {
      "epoch": 94.32,
      "learning_rate": 7.424207119741101e-06,
      "loss": 0.1441,
      "step": 242870
    },
    {
      "epoch": 94.32,
      "learning_rate": 7.42368932038835e-06,
      "loss": 0.0524,
      "step": 242880
    },
    {
      "epoch": 94.33,
      "learning_rate": 7.423171521035599e-06,
      "loss": 0.2376,
      "step": 242890
    },
    {
      "epoch": 94.33,
      "learning_rate": 7.4226537216828486e-06,
      "loss": 0.1154,
      "step": 242900
    },
    {
      "epoch": 94.33,
      "learning_rate": 7.422135922330098e-06,
      "loss": 0.1111,
      "step": 242910
    },
    {
      "epoch": 94.34,
      "learning_rate": 7.421618122977346e-06,
      "loss": 0.0556,
      "step": 242920
    },
    {
      "epoch": 94.34,
      "learning_rate": 7.421100323624596e-06,
      "loss": 0.0924,
      "step": 242930
    },
    {
      "epoch": 94.35,
      "learning_rate": 7.420582524271845e-06,
      "loss": 0.033,
      "step": 242940
    },
    {
      "epoch": 94.35,
      "learning_rate": 7.420064724919095e-06,
      "loss": 0.0792,
      "step": 242950
    },
    {
      "epoch": 94.35,
      "learning_rate": 7.419546925566343e-06,
      "loss": 0.0021,
      "step": 242960
    },
    {
      "epoch": 94.36,
      "learning_rate": 7.4190291262135925e-06,
      "loss": 0.0366,
      "step": 242970
    },
    {
      "epoch": 94.36,
      "learning_rate": 7.418511326860842e-06,
      "loss": 0.0206,
      "step": 242980
    },
    {
      "epoch": 94.37,
      "learning_rate": 7.417993527508092e-06,
      "loss": 0.0444,
      "step": 242990
    },
    {
      "epoch": 94.37,
      "learning_rate": 7.41747572815534e-06,
      "loss": 0.0006,
      "step": 243000
    },
    {
      "epoch": 94.37,
      "learning_rate": 7.416957928802589e-06,
      "loss": 0.0547,
      "step": 243010
    },
    {
      "epoch": 94.38,
      "learning_rate": 7.416440129449839e-06,
      "loss": 0.0004,
      "step": 243020
    },
    {
      "epoch": 94.38,
      "learning_rate": 7.4159223300970885e-06,
      "loss": 0.0784,
      "step": 243030
    },
    {
      "epoch": 94.38,
      "learning_rate": 7.4154045307443364e-06,
      "loss": 0.1938,
      "step": 243040
    },
    {
      "epoch": 94.39,
      "learning_rate": 7.414886731391586e-06,
      "loss": 0.0177,
      "step": 243050
    },
    {
      "epoch": 94.39,
      "learning_rate": 7.414368932038836e-06,
      "loss": 0.0962,
      "step": 243060
    },
    {
      "epoch": 94.4,
      "learning_rate": 7.413851132686085e-06,
      "loss": 0.0455,
      "step": 243070
    },
    {
      "epoch": 94.4,
      "learning_rate": 7.413333333333333e-06,
      "loss": 0.0561,
      "step": 243080
    },
    {
      "epoch": 94.4,
      "learning_rate": 7.412815533980583e-06,
      "loss": 0.1252,
      "step": 243090
    },
    {
      "epoch": 94.41,
      "learning_rate": 7.4122977346278325e-06,
      "loss": 0.0772,
      "step": 243100
    },
    {
      "epoch": 94.41,
      "learning_rate": 7.411779935275082e-06,
      "loss": 0.0996,
      "step": 243110
    },
    {
      "epoch": 94.42,
      "learning_rate": 7.41126213592233e-06,
      "loss": 0.0301,
      "step": 243120
    },
    {
      "epoch": 94.42,
      "learning_rate": 7.41074433656958e-06,
      "loss": 0.0204,
      "step": 243130
    },
    {
      "epoch": 94.42,
      "learning_rate": 7.410226537216829e-06,
      "loss": 0.0887,
      "step": 243140
    },
    {
      "epoch": 94.43,
      "learning_rate": 7.409708737864079e-06,
      "loss": 0.0015,
      "step": 243150
    },
    {
      "epoch": 94.43,
      "learning_rate": 7.409190938511327e-06,
      "loss": 0.0666,
      "step": 243160
    },
    {
      "epoch": 94.43,
      "learning_rate": 7.408673139158576e-06,
      "loss": 0.047,
      "step": 243170
    },
    {
      "epoch": 94.44,
      "learning_rate": 7.408155339805826e-06,
      "loss": 0.0275,
      "step": 243180
    },
    {
      "epoch": 94.44,
      "learning_rate": 7.407637540453076e-06,
      "loss": 0.1671,
      "step": 243190
    },
    {
      "epoch": 94.45,
      "learning_rate": 7.4071197411003236e-06,
      "loss": 0.0111,
      "step": 243200
    },
    {
      "epoch": 94.45,
      "learning_rate": 7.406601941747573e-06,
      "loss": 0.0607,
      "step": 243210
    },
    {
      "epoch": 94.45,
      "learning_rate": 7.406084142394823e-06,
      "loss": 0.0274,
      "step": 243220
    },
    {
      "epoch": 94.46,
      "learning_rate": 7.405566343042072e-06,
      "loss": 0.0507,
      "step": 243230
    },
    {
      "epoch": 94.46,
      "learning_rate": 7.40504854368932e-06,
      "loss": 0.0096,
      "step": 243240
    },
    {
      "epoch": 94.47,
      "learning_rate": 7.40453074433657e-06,
      "loss": 0.0671,
      "step": 243250
    },
    {
      "epoch": 94.47,
      "learning_rate": 7.4040129449838196e-06,
      "loss": 0.0248,
      "step": 243260
    },
    {
      "epoch": 94.47,
      "learning_rate": 7.403495145631069e-06,
      "loss": 0.1207,
      "step": 243270
    },
    {
      "epoch": 94.48,
      "learning_rate": 7.402977346278317e-06,
      "loss": 0.1018,
      "step": 243280
    },
    {
      "epoch": 94.48,
      "learning_rate": 7.402459546925567e-06,
      "loss": 0.0452,
      "step": 243290
    },
    {
      "epoch": 94.49,
      "learning_rate": 7.401941747572816e-06,
      "loss": 0.0102,
      "step": 243300
    },
    {
      "epoch": 94.49,
      "learning_rate": 7.401423948220066e-06,
      "loss": 0.0067,
      "step": 243310
    },
    {
      "epoch": 94.49,
      "learning_rate": 7.400906148867315e-06,
      "loss": 0.0334,
      "step": 243320
    },
    {
      "epoch": 94.5,
      "learning_rate": 7.4003883495145635e-06,
      "loss": 0.0589,
      "step": 243330
    },
    {
      "epoch": 94.5,
      "learning_rate": 7.399870550161813e-06,
      "loss": 0.0195,
      "step": 243340
    },
    {
      "epoch": 94.5,
      "learning_rate": 7.399352750809063e-06,
      "loss": 0.0184,
      "step": 243350
    },
    {
      "epoch": 94.51,
      "learning_rate": 7.3988349514563115e-06,
      "loss": 0.045,
      "step": 243360
    },
    {
      "epoch": 94.51,
      "learning_rate": 7.39831715210356e-06,
      "loss": 0.0934,
      "step": 243370
    },
    {
      "epoch": 94.52,
      "learning_rate": 7.39779935275081e-06,
      "loss": 0.0564,
      "step": 243380
    },
    {
      "epoch": 94.52,
      "learning_rate": 7.397281553398059e-06,
      "loss": 0.037,
      "step": 243390
    },
    {
      "epoch": 94.52,
      "learning_rate": 7.396763754045308e-06,
      "loss": 0.0275,
      "step": 243400
    },
    {
      "epoch": 94.53,
      "learning_rate": 7.396245954692557e-06,
      "loss": 0.0267,
      "step": 243410
    },
    {
      "epoch": 94.53,
      "learning_rate": 7.395728155339807e-06,
      "loss": 0.0023,
      "step": 243420
    },
    {
      "epoch": 94.54,
      "learning_rate": 7.3952103559870555e-06,
      "loss": 0.0005,
      "step": 243430
    },
    {
      "epoch": 94.54,
      "learning_rate": 7.394692556634305e-06,
      "loss": 0.1122,
      "step": 243440
    },
    {
      "epoch": 94.54,
      "learning_rate": 7.394174757281554e-06,
      "loss": 0.1022,
      "step": 243450
    },
    {
      "epoch": 94.55,
      "learning_rate": 7.3936569579288035e-06,
      "loss": 0.0326,
      "step": 243460
    },
    {
      "epoch": 94.55,
      "learning_rate": 7.393139158576052e-06,
      "loss": 0.126,
      "step": 243470
    },
    {
      "epoch": 94.56,
      "learning_rate": 7.392621359223302e-06,
      "loss": 0.0749,
      "step": 243480
    },
    {
      "epoch": 94.56,
      "learning_rate": 7.392103559870551e-06,
      "loss": 0.0295,
      "step": 243490
    },
    {
      "epoch": 94.56,
      "learning_rate": 7.3915857605178e-06,
      "loss": 0.0006,
      "step": 243500
    },
    {
      "epoch": 94.57,
      "learning_rate": 7.391067961165049e-06,
      "loss": 0.0452,
      "step": 243510
    },
    {
      "epoch": 94.57,
      "learning_rate": 7.390550161812299e-06,
      "loss": 0.1028,
      "step": 243520
    },
    {
      "epoch": 94.57,
      "learning_rate": 7.390032362459547e-06,
      "loss": 0.0435,
      "step": 243530
    },
    {
      "epoch": 94.58,
      "learning_rate": 7.389514563106797e-06,
      "loss": 0.0138,
      "step": 243540
    },
    {
      "epoch": 94.58,
      "learning_rate": 7.388996763754046e-06,
      "loss": 0.0365,
      "step": 243550
    },
    {
      "epoch": 94.59,
      "learning_rate": 7.388478964401295e-06,
      "loss": 0.1402,
      "step": 243560
    },
    {
      "epoch": 94.59,
      "learning_rate": 7.387961165048544e-06,
      "loss": 0.0579,
      "step": 243570
    },
    {
      "epoch": 94.59,
      "learning_rate": 7.387443365695793e-06,
      "loss": 0.0315,
      "step": 243580
    },
    {
      "epoch": 94.6,
      "learning_rate": 7.3869255663430426e-06,
      "loss": 0.1554,
      "step": 243590
    },
    {
      "epoch": 94.6,
      "learning_rate": 7.386407766990292e-06,
      "loss": 0.0127,
      "step": 243600
    },
    {
      "epoch": 94.61,
      "learning_rate": 7.385889967637541e-06,
      "loss": 0.0429,
      "step": 243610
    },
    {
      "epoch": 94.61,
      "learning_rate": 7.38537216828479e-06,
      "loss": 0.1281,
      "step": 243620
    },
    {
      "epoch": 94.61,
      "learning_rate": 7.384854368932039e-06,
      "loss": 0.0954,
      "step": 243630
    },
    {
      "epoch": 94.62,
      "learning_rate": 7.384336569579289e-06,
      "loss": 0.0832,
      "step": 243640
    },
    {
      "epoch": 94.62,
      "learning_rate": 7.383818770226538e-06,
      "loss": 0.0323,
      "step": 243650
    },
    {
      "epoch": 94.63,
      "learning_rate": 7.3833009708737865e-06,
      "loss": 0.1274,
      "step": 243660
    },
    {
      "epoch": 94.63,
      "learning_rate": 7.382783171521036e-06,
      "loss": 0.0286,
      "step": 243670
    },
    {
      "epoch": 94.63,
      "learning_rate": 7.382265372168286e-06,
      "loss": 0.0102,
      "step": 243680
    },
    {
      "epoch": 94.64,
      "learning_rate": 7.3817475728155345e-06,
      "loss": 0.0469,
      "step": 243690
    },
    {
      "epoch": 94.64,
      "learning_rate": 7.381229773462783e-06,
      "loss": 0.0142,
      "step": 243700
    },
    {
      "epoch": 94.64,
      "learning_rate": 7.380711974110033e-06,
      "loss": 0.0975,
      "step": 243710
    },
    {
      "epoch": 94.65,
      "learning_rate": 7.3801941747572825e-06,
      "loss": 0.0021,
      "step": 243720
    },
    {
      "epoch": 94.65,
      "learning_rate": 7.3796763754045304e-06,
      "loss": 0.0935,
      "step": 243730
    },
    {
      "epoch": 94.66,
      "learning_rate": 7.37915857605178e-06,
      "loss": 0.0919,
      "step": 243740
    },
    {
      "epoch": 94.66,
      "learning_rate": 7.37864077669903e-06,
      "loss": 0.0516,
      "step": 243750
    },
    {
      "epoch": 94.66,
      "learning_rate": 7.378122977346279e-06,
      "loss": 0.0002,
      "step": 243760
    },
    {
      "epoch": 94.67,
      "learning_rate": 7.377605177993527e-06,
      "loss": 0.021,
      "step": 243770
    },
    {
      "epoch": 94.67,
      "learning_rate": 7.377087378640777e-06,
      "loss": 0.1205,
      "step": 243780
    },
    {
      "epoch": 94.68,
      "learning_rate": 7.3765695792880265e-06,
      "loss": 0.0206,
      "step": 243790
    },
    {
      "epoch": 94.68,
      "learning_rate": 7.376051779935276e-06,
      "loss": 0.0074,
      "step": 243800
    },
    {
      "epoch": 94.68,
      "learning_rate": 7.375533980582524e-06,
      "loss": 0.0939,
      "step": 243810
    },
    {
      "epoch": 94.69,
      "learning_rate": 7.375016181229774e-06,
      "loss": 0.0458,
      "step": 243820
    },
    {
      "epoch": 94.69,
      "learning_rate": 7.374498381877023e-06,
      "loss": 0.2001,
      "step": 243830
    },
    {
      "epoch": 94.7,
      "learning_rate": 7.373980582524273e-06,
      "loss": 0.0313,
      "step": 243840
    },
    {
      "epoch": 94.7,
      "learning_rate": 7.373462783171521e-06,
      "loss": 0.0236,
      "step": 243850
    },
    {
      "epoch": 94.7,
      "learning_rate": 7.37294498381877e-06,
      "loss": 0.0662,
      "step": 243860
    },
    {
      "epoch": 94.71,
      "learning_rate": 7.37242718446602e-06,
      "loss": 0.0725,
      "step": 243870
    },
    {
      "epoch": 94.71,
      "learning_rate": 7.37190938511327e-06,
      "loss": 0.0803,
      "step": 243880
    },
    {
      "epoch": 94.71,
      "learning_rate": 7.371391585760519e-06,
      "loss": 0.0198,
      "step": 243890
    },
    {
      "epoch": 94.72,
      "learning_rate": 7.370873786407767e-06,
      "loss": 0.036,
      "step": 243900
    },
    {
      "epoch": 94.72,
      "learning_rate": 7.370355987055017e-06,
      "loss": 0.0223,
      "step": 243910
    },
    {
      "epoch": 94.73,
      "learning_rate": 7.369838187702266e-06,
      "loss": 0.1279,
      "step": 243920
    },
    {
      "epoch": 94.73,
      "learning_rate": 7.369320388349516e-06,
      "loss": 0.1346,
      "step": 243930
    },
    {
      "epoch": 94.73,
      "learning_rate": 7.368802588996764e-06,
      "loss": 0.06,
      "step": 243940
    },
    {
      "epoch": 94.74,
      "learning_rate": 7.3682847896440136e-06,
      "loss": 0.0458,
      "step": 243950
    },
    {
      "epoch": 94.74,
      "learning_rate": 7.367766990291263e-06,
      "loss": 0.0377,
      "step": 243960
    },
    {
      "epoch": 94.75,
      "learning_rate": 7.367249190938513e-06,
      "loss": 0.0005,
      "step": 243970
    },
    {
      "epoch": 94.75,
      "learning_rate": 7.366731391585761e-06,
      "loss": 0.0806,
      "step": 243980
    },
    {
      "epoch": 94.75,
      "learning_rate": 7.36621359223301e-06,
      "loss": 0.0291,
      "step": 243990
    },
    {
      "epoch": 94.76,
      "learning_rate": 7.36569579288026e-06,
      "loss": 0.0002,
      "step": 244000
    },
    {
      "epoch": 94.76,
      "learning_rate": 7.36517799352751e-06,
      "loss": 0.0655,
      "step": 244010
    },
    {
      "epoch": 94.77,
      "learning_rate": 7.3646601941747575e-06,
      "loss": 0.0405,
      "step": 244020
    },
    {
      "epoch": 94.77,
      "learning_rate": 7.364142394822007e-06,
      "loss": 0.0513,
      "step": 244030
    },
    {
      "epoch": 94.77,
      "learning_rate": 7.363624595469257e-06,
      "loss": 0.1117,
      "step": 244040
    },
    {
      "epoch": 94.78,
      "learning_rate": 7.3631067961165055e-06,
      "loss": 0.0486,
      "step": 244050
    },
    {
      "epoch": 94.78,
      "learning_rate": 7.362588996763754e-06,
      "loss": 0.0505,
      "step": 244060
    },
    {
      "epoch": 94.78,
      "learning_rate": 7.362071197411004e-06,
      "loss": 0.0003,
      "step": 244070
    },
    {
      "epoch": 94.79,
      "learning_rate": 7.3615533980582535e-06,
      "loss": 0.0228,
      "step": 244080
    },
    {
      "epoch": 94.79,
      "learning_rate": 7.361035598705502e-06,
      "loss": 0.081,
      "step": 244090
    },
    {
      "epoch": 94.8,
      "learning_rate": 7.360517799352751e-06,
      "loss": 0.1203,
      "step": 244100
    },
    {
      "epoch": 94.8,
      "learning_rate": 7.360000000000001e-06,
      "loss": 0.0015,
      "step": 244110
    },
    {
      "epoch": 94.8,
      "learning_rate": 7.35948220064725e-06,
      "loss": 0.2574,
      "step": 244120
    },
    {
      "epoch": 94.81,
      "learning_rate": 7.358964401294499e-06,
      "loss": 0.0946,
      "step": 244130
    },
    {
      "epoch": 94.81,
      "learning_rate": 7.358446601941748e-06,
      "loss": 0.1121,
      "step": 244140
    },
    {
      "epoch": 94.82,
      "learning_rate": 7.3579288025889975e-06,
      "loss": 0.0139,
      "step": 244150
    },
    {
      "epoch": 94.82,
      "learning_rate": 7.357411003236247e-06,
      "loss": 0.0333,
      "step": 244160
    },
    {
      "epoch": 94.82,
      "learning_rate": 7.356893203883496e-06,
      "loss": 0.061,
      "step": 244170
    },
    {
      "epoch": 94.83,
      "learning_rate": 7.356375404530745e-06,
      "loss": 0.0609,
      "step": 244180
    },
    {
      "epoch": 94.83,
      "learning_rate": 7.355857605177994e-06,
      "loss": 0.1075,
      "step": 244190
    },
    {
      "epoch": 94.83,
      "learning_rate": 7.355339805825243e-06,
      "loss": 0.0027,
      "step": 244200
    },
    {
      "epoch": 94.84,
      "learning_rate": 7.354822006472493e-06,
      "loss": 0.0942,
      "step": 244210
    },
    {
      "epoch": 94.84,
      "learning_rate": 7.354304207119741e-06,
      "loss": 0.0492,
      "step": 244220
    },
    {
      "epoch": 94.85,
      "learning_rate": 7.353786407766991e-06,
      "loss": 0.1011,
      "step": 244230
    },
    {
      "epoch": 94.85,
      "learning_rate": 7.35326860841424e-06,
      "loss": 0.1131,
      "step": 244240
    },
    {
      "epoch": 94.85,
      "learning_rate": 7.352750809061489e-06,
      "loss": 0.0527,
      "step": 244250
    },
    {
      "epoch": 94.86,
      "learning_rate": 7.352233009708738e-06,
      "loss": 0.0057,
      "step": 244260
    },
    {
      "epoch": 94.86,
      "learning_rate": 7.351715210355988e-06,
      "loss": 0.0873,
      "step": 244270
    },
    {
      "epoch": 94.87,
      "learning_rate": 7.3511974110032366e-06,
      "loss": 0.0618,
      "step": 244280
    },
    {
      "epoch": 94.87,
      "learning_rate": 7.350679611650486e-06,
      "loss": 0.0265,
      "step": 244290
    },
    {
      "epoch": 94.87,
      "learning_rate": 7.350161812297735e-06,
      "loss": 0.0204,
      "step": 244300
    },
    {
      "epoch": 94.88,
      "learning_rate": 7.3496440129449846e-06,
      "loss": 0.0133,
      "step": 244310
    },
    {
      "epoch": 94.88,
      "learning_rate": 7.349126213592233e-06,
      "loss": 0.0059,
      "step": 244320
    },
    {
      "epoch": 94.89,
      "learning_rate": 7.348608414239483e-06,
      "loss": 0.1071,
      "step": 244330
    },
    {
      "epoch": 94.89,
      "learning_rate": 7.348090614886732e-06,
      "loss": 0.0071,
      "step": 244340
    },
    {
      "epoch": 94.89,
      "learning_rate": 7.347572815533981e-06,
      "loss": 0.0827,
      "step": 244350
    },
    {
      "epoch": 94.9,
      "learning_rate": 7.34705501618123e-06,
      "loss": 0.0307,
      "step": 244360
    },
    {
      "epoch": 94.9,
      "learning_rate": 7.34653721682848e-06,
      "loss": 0.1318,
      "step": 244370
    },
    {
      "epoch": 94.9,
      "learning_rate": 7.3460194174757285e-06,
      "loss": 0.0132,
      "step": 244380
    },
    {
      "epoch": 94.91,
      "learning_rate": 7.345501618122977e-06,
      "loss": 0.0182,
      "step": 244390
    },
    {
      "epoch": 94.91,
      "learning_rate": 7.344983818770227e-06,
      "loss": 0.0235,
      "step": 244400
    },
    {
      "epoch": 94.92,
      "learning_rate": 7.3444660194174765e-06,
      "loss": 0.0696,
      "step": 244410
    },
    {
      "epoch": 94.92,
      "learning_rate": 7.343948220064726e-06,
      "loss": 0.0419,
      "step": 244420
    },
    {
      "epoch": 94.92,
      "learning_rate": 7.343430420711974e-06,
      "loss": 0.0008,
      "step": 244430
    },
    {
      "epoch": 94.93,
      "learning_rate": 7.342912621359224e-06,
      "loss": 0.159,
      "step": 244440
    },
    {
      "epoch": 94.93,
      "learning_rate": 7.342394822006473e-06,
      "loss": 0.1616,
      "step": 244450
    },
    {
      "epoch": 94.94,
      "learning_rate": 7.341877022653723e-06,
      "loss": 0.0915,
      "step": 244460
    },
    {
      "epoch": 94.94,
      "learning_rate": 7.341359223300971e-06,
      "loss": 0.0336,
      "step": 244470
    },
    {
      "epoch": 94.94,
      "learning_rate": 7.3408414239482205e-06,
      "loss": 0.0266,
      "step": 244480
    },
    {
      "epoch": 94.95,
      "learning_rate": 7.34032362459547e-06,
      "loss": 0.1053,
      "step": 244490
    },
    {
      "epoch": 94.95,
      "learning_rate": 7.33980582524272e-06,
      "loss": 0.069,
      "step": 244500
    },
    {
      "epoch": 94.96,
      "learning_rate": 7.339288025889968e-06,
      "loss": 0.0234,
      "step": 244510
    },
    {
      "epoch": 94.96,
      "learning_rate": 7.338770226537217e-06,
      "loss": 0.0412,
      "step": 244520
    },
    {
      "epoch": 94.96,
      "learning_rate": 7.338252427184467e-06,
      "loss": 0.0917,
      "step": 244530
    },
    {
      "epoch": 94.97,
      "learning_rate": 7.3377346278317165e-06,
      "loss": 0.2571,
      "step": 244540
    },
    {
      "epoch": 94.97,
      "learning_rate": 7.337216828478964e-06,
      "loss": 0.1234,
      "step": 244550
    },
    {
      "epoch": 94.97,
      "learning_rate": 7.336699029126214e-06,
      "loss": 0.0302,
      "step": 244560
    },
    {
      "epoch": 94.98,
      "learning_rate": 7.336181229773464e-06,
      "loss": 0.0668,
      "step": 244570
    },
    {
      "epoch": 94.98,
      "learning_rate": 7.335663430420713e-06,
      "loss": 0.0854,
      "step": 244580
    },
    {
      "epoch": 94.99,
      "learning_rate": 7.335145631067961e-06,
      "loss": 0.0138,
      "step": 244590
    },
    {
      "epoch": 94.99,
      "learning_rate": 7.334627831715211e-06,
      "loss": 0.0103,
      "step": 244600
    },
    {
      "epoch": 94.99,
      "learning_rate": 7.33411003236246e-06,
      "loss": 0.0001,
      "step": 244610
    },
    {
      "epoch": 95.0,
      "learning_rate": 7.33359223300971e-06,
      "loss": 0.071,
      "step": 244620
    },
    {
      "epoch": 95.0,
      "eval_accuracy": 0.9482806052269601,
      "eval_loss": 0.3724736273288727,
      "eval_runtime": 8.291,
      "eval_samples_per_second": 438.427,
      "eval_steps_per_second": 54.879,
      "step": 244625
    },
    {
      "epoch": 95.0,
      "learning_rate": 7.333074433656958e-06,
      "loss": 0.0049,
      "step": 244630
    },
    {
      "epoch": 95.01,
      "learning_rate": 7.3325566343042076e-06,
      "loss": 0.0385,
      "step": 244640
    },
    {
      "epoch": 95.01,
      "learning_rate": 7.332038834951457e-06,
      "loss": 0.0183,
      "step": 244650
    },
    {
      "epoch": 95.01,
      "learning_rate": 7.331521035598707e-06,
      "loss": 0.0379,
      "step": 244660
    },
    {
      "epoch": 95.02,
      "learning_rate": 7.331003236245955e-06,
      "loss": 0.1306,
      "step": 244670
    },
    {
      "epoch": 95.02,
      "learning_rate": 7.330485436893204e-06,
      "loss": 0.0612,
      "step": 244680
    },
    {
      "epoch": 95.03,
      "learning_rate": 7.329967637540454e-06,
      "loss": 0.003,
      "step": 244690
    },
    {
      "epoch": 95.03,
      "learning_rate": 7.329449838187704e-06,
      "loss": 0.041,
      "step": 244700
    },
    {
      "epoch": 95.03,
      "learning_rate": 7.3289320388349515e-06,
      "loss": 0.1253,
      "step": 244710
    },
    {
      "epoch": 95.04,
      "learning_rate": 7.328414239482201e-06,
      "loss": 0.0237,
      "step": 244720
    },
    {
      "epoch": 95.04,
      "learning_rate": 7.327896440129451e-06,
      "loss": 0.0886,
      "step": 244730
    },
    {
      "epoch": 95.04,
      "learning_rate": 7.3273786407767e-06,
      "loss": 0.0273,
      "step": 244740
    },
    {
      "epoch": 95.05,
      "learning_rate": 7.326860841423948e-06,
      "loss": 0.0085,
      "step": 244750
    },
    {
      "epoch": 95.05,
      "learning_rate": 7.326343042071198e-06,
      "loss": 0.0156,
      "step": 244760
    },
    {
      "epoch": 95.06,
      "learning_rate": 7.3258252427184475e-06,
      "loss": 0.0104,
      "step": 244770
    },
    {
      "epoch": 95.06,
      "learning_rate": 7.325307443365697e-06,
      "loss": 0.0363,
      "step": 244780
    },
    {
      "epoch": 95.06,
      "learning_rate": 7.324789644012945e-06,
      "loss": 0.0601,
      "step": 244790
    },
    {
      "epoch": 95.07,
      "learning_rate": 7.324271844660195e-06,
      "loss": 0.0623,
      "step": 244800
    },
    {
      "epoch": 95.07,
      "learning_rate": 7.323754045307444e-06,
      "loss": 0.0008,
      "step": 244810
    },
    {
      "epoch": 95.08,
      "learning_rate": 7.323236245954694e-06,
      "loss": 0.1423,
      "step": 244820
    },
    {
      "epoch": 95.08,
      "learning_rate": 7.322718446601942e-06,
      "loss": 0.0314,
      "step": 244830
    },
    {
      "epoch": 95.08,
      "learning_rate": 7.3222006472491915e-06,
      "loss": 0.0365,
      "step": 244840
    },
    {
      "epoch": 95.09,
      "learning_rate": 7.321682847896441e-06,
      "loss": 0.0727,
      "step": 244850
    },
    {
      "epoch": 95.09,
      "learning_rate": 7.32116504854369e-06,
      "loss": 0.0521,
      "step": 244860
    },
    {
      "epoch": 95.1,
      "learning_rate": 7.320647249190939e-06,
      "loss": 0.0211,
      "step": 244870
    },
    {
      "epoch": 95.1,
      "learning_rate": 7.320129449838188e-06,
      "loss": 0.0272,
      "step": 244880
    },
    {
      "epoch": 95.1,
      "learning_rate": 7.319611650485438e-06,
      "loss": 0.0519,
      "step": 244890
    },
    {
      "epoch": 95.11,
      "learning_rate": 7.319093851132687e-06,
      "loss": 0.1658,
      "step": 244900
    },
    {
      "epoch": 95.11,
      "learning_rate": 7.318576051779935e-06,
      "loss": 0.2328,
      "step": 244910
    },
    {
      "epoch": 95.11,
      "learning_rate": 7.318058252427185e-06,
      "loss": 0.1275,
      "step": 244920
    },
    {
      "epoch": 95.12,
      "learning_rate": 7.317540453074435e-06,
      "loss": 0.1211,
      "step": 244930
    },
    {
      "epoch": 95.12,
      "learning_rate": 7.317022653721683e-06,
      "loss": 0.0023,
      "step": 244940
    },
    {
      "epoch": 95.13,
      "learning_rate": 7.316504854368932e-06,
      "loss": 0.042,
      "step": 244950
    },
    {
      "epoch": 95.13,
      "learning_rate": 7.315987055016182e-06,
      "loss": 0.0146,
      "step": 244960
    },
    {
      "epoch": 95.13,
      "learning_rate": 7.315469255663431e-06,
      "loss": 0.0147,
      "step": 244970
    },
    {
      "epoch": 95.14,
      "learning_rate": 7.31495145631068e-06,
      "loss": 0.1993,
      "step": 244980
    },
    {
      "epoch": 95.14,
      "learning_rate": 7.31443365695793e-06,
      "loss": 0.0432,
      "step": 244990
    },
    {
      "epoch": 95.15,
      "learning_rate": 7.3139158576051786e-06,
      "loss": 0.1412,
      "step": 245000
    },
    {
      "epoch": 95.15,
      "learning_rate": 7.313398058252428e-06,
      "loss": 0.0385,
      "step": 245010
    },
    {
      "epoch": 95.15,
      "learning_rate": 7.312880258899677e-06,
      "loss": 0.0292,
      "step": 245020
    },
    {
      "epoch": 95.16,
      "learning_rate": 7.312362459546927e-06,
      "loss": 0.0912,
      "step": 245030
    },
    {
      "epoch": 95.16,
      "learning_rate": 7.311844660194175e-06,
      "loss": 0.0002,
      "step": 245040
    },
    {
      "epoch": 95.17,
      "learning_rate": 7.311326860841424e-06,
      "loss": 0.0233,
      "step": 245050
    },
    {
      "epoch": 95.17,
      "learning_rate": 7.310809061488674e-06,
      "loss": 0.1104,
      "step": 245060
    },
    {
      "epoch": 95.17,
      "learning_rate": 7.310291262135923e-06,
      "loss": 0.0611,
      "step": 245070
    },
    {
      "epoch": 95.18,
      "learning_rate": 7.309773462783172e-06,
      "loss": 0.118,
      "step": 245080
    },
    {
      "epoch": 95.18,
      "learning_rate": 7.309255663430421e-06,
      "loss": 0.1542,
      "step": 245090
    },
    {
      "epoch": 95.18,
      "learning_rate": 7.3087378640776705e-06,
      "loss": 0.0146,
      "step": 245100
    },
    {
      "epoch": 95.19,
      "learning_rate": 7.30822006472492e-06,
      "loss": 0.0145,
      "step": 245110
    },
    {
      "epoch": 95.19,
      "learning_rate": 7.307702265372169e-06,
      "loss": 0.0113,
      "step": 245120
    },
    {
      "epoch": 95.2,
      "learning_rate": 7.307184466019418e-06,
      "loss": 0.0812,
      "step": 245130
    },
    {
      "epoch": 95.2,
      "learning_rate": 7.306666666666667e-06,
      "loss": 0.0333,
      "step": 245140
    },
    {
      "epoch": 95.2,
      "learning_rate": 7.306148867313917e-06,
      "loss": 0.0034,
      "step": 245150
    },
    {
      "epoch": 95.21,
      "learning_rate": 7.305631067961166e-06,
      "loss": 0.0264,
      "step": 245160
    },
    {
      "epoch": 95.21,
      "learning_rate": 7.3051132686084145e-06,
      "loss": 0.0714,
      "step": 245170
    },
    {
      "epoch": 95.22,
      "learning_rate": 7.304595469255664e-06,
      "loss": 0.0442,
      "step": 245180
    },
    {
      "epoch": 95.22,
      "learning_rate": 7.304077669902914e-06,
      "loss": 0.0357,
      "step": 245190
    },
    {
      "epoch": 95.22,
      "learning_rate": 7.303559870550162e-06,
      "loss": 0.0224,
      "step": 245200
    },
    {
      "epoch": 95.23,
      "learning_rate": 7.303042071197411e-06,
      "loss": 0.093,
      "step": 245210
    },
    {
      "epoch": 95.23,
      "learning_rate": 7.302524271844661e-06,
      "loss": 0.0163,
      "step": 245220
    },
    {
      "epoch": 95.23,
      "learning_rate": 7.3020064724919105e-06,
      "loss": 0.0381,
      "step": 245230
    },
    {
      "epoch": 95.24,
      "learning_rate": 7.301488673139158e-06,
      "loss": 0.0388,
      "step": 245240
    },
    {
      "epoch": 95.24,
      "learning_rate": 7.300970873786408e-06,
      "loss": 0.0002,
      "step": 245250
    },
    {
      "epoch": 95.25,
      "learning_rate": 7.300453074433658e-06,
      "loss": 0.0911,
      "step": 245260
    },
    {
      "epoch": 95.25,
      "learning_rate": 7.299935275080907e-06,
      "loss": 0.0089,
      "step": 245270
    },
    {
      "epoch": 95.25,
      "learning_rate": 7.299417475728155e-06,
      "loss": 0.023,
      "step": 245280
    },
    {
      "epoch": 95.26,
      "learning_rate": 7.298899676375405e-06,
      "loss": 0.1346,
      "step": 245290
    },
    {
      "epoch": 95.26,
      "learning_rate": 7.298381877022654e-06,
      "loss": 0.1675,
      "step": 245300
    },
    {
      "epoch": 95.27,
      "learning_rate": 7.297864077669904e-06,
      "loss": 0.0554,
      "step": 245310
    },
    {
      "epoch": 95.27,
      "learning_rate": 7.297346278317152e-06,
      "loss": 0.0379,
      "step": 245320
    },
    {
      "epoch": 95.27,
      "learning_rate": 7.2968284789644016e-06,
      "loss": 0.0337,
      "step": 245330
    },
    {
      "epoch": 95.28,
      "learning_rate": 7.296310679611651e-06,
      "loss": 0.1455,
      "step": 245340
    },
    {
      "epoch": 95.28,
      "learning_rate": 7.295792880258901e-06,
      "loss": 0.0911,
      "step": 245350
    },
    {
      "epoch": 95.29,
      "learning_rate": 7.295275080906149e-06,
      "loss": 0.0008,
      "step": 245360
    },
    {
      "epoch": 95.29,
      "learning_rate": 7.294757281553398e-06,
      "loss": 0.108,
      "step": 245370
    },
    {
      "epoch": 95.29,
      "learning_rate": 7.294239482200648e-06,
      "loss": 0.0589,
      "step": 245380
    },
    {
      "epoch": 95.3,
      "learning_rate": 7.293721682847898e-06,
      "loss": 0.0667,
      "step": 245390
    },
    {
      "epoch": 95.3,
      "learning_rate": 7.2932038834951455e-06,
      "loss": 0.2048,
      "step": 245400
    },
    {
      "epoch": 95.3,
      "learning_rate": 7.292686084142395e-06,
      "loss": 0.073,
      "step": 245410
    },
    {
      "epoch": 95.31,
      "learning_rate": 7.292168284789645e-06,
      "loss": 0.0331,
      "step": 245420
    },
    {
      "epoch": 95.31,
      "learning_rate": 7.291650485436894e-06,
      "loss": 0.0531,
      "step": 245430
    },
    {
      "epoch": 95.32,
      "learning_rate": 7.291132686084142e-06,
      "loss": 0.0003,
      "step": 245440
    },
    {
      "epoch": 95.32,
      "learning_rate": 7.290614886731392e-06,
      "loss": 0.0279,
      "step": 245450
    },
    {
      "epoch": 95.32,
      "learning_rate": 7.2900970873786415e-06,
      "loss": 0.0509,
      "step": 245460
    },
    {
      "epoch": 95.33,
      "learning_rate": 7.289579288025891e-06,
      "loss": 0.0225,
      "step": 245470
    },
    {
      "epoch": 95.33,
      "learning_rate": 7.289061488673139e-06,
      "loss": 0.0047,
      "step": 245480
    },
    {
      "epoch": 95.34,
      "learning_rate": 7.288543689320389e-06,
      "loss": 0.0204,
      "step": 245490
    },
    {
      "epoch": 95.34,
      "learning_rate": 7.288025889967638e-06,
      "loss": 0.0928,
      "step": 245500
    },
    {
      "epoch": 95.34,
      "learning_rate": 7.287508090614888e-06,
      "loss": 0.0571,
      "step": 245510
    },
    {
      "epoch": 95.35,
      "learning_rate": 7.286990291262136e-06,
      "loss": 0.0144,
      "step": 245520
    },
    {
      "epoch": 95.35,
      "learning_rate": 7.2864724919093855e-06,
      "loss": 0.1381,
      "step": 245530
    },
    {
      "epoch": 95.36,
      "learning_rate": 7.285954692556635e-06,
      "loss": 0.026,
      "step": 245540
    },
    {
      "epoch": 95.36,
      "learning_rate": 7.285436893203885e-06,
      "loss": 0.1009,
      "step": 245550
    },
    {
      "epoch": 95.36,
      "learning_rate": 7.2849190938511335e-06,
      "loss": 0.0314,
      "step": 245560
    },
    {
      "epoch": 95.37,
      "learning_rate": 7.284401294498382e-06,
      "loss": 0.1501,
      "step": 245570
    },
    {
      "epoch": 95.37,
      "learning_rate": 7.283883495145632e-06,
      "loss": 0.0393,
      "step": 245580
    },
    {
      "epoch": 95.37,
      "learning_rate": 7.2833656957928815e-06,
      "loss": 0.0725,
      "step": 245590
    },
    {
      "epoch": 95.38,
      "learning_rate": 7.28284789644013e-06,
      "loss": 0.172,
      "step": 245600
    },
    {
      "epoch": 95.38,
      "learning_rate": 7.282330097087379e-06,
      "loss": 0.0422,
      "step": 245610
    },
    {
      "epoch": 95.39,
      "learning_rate": 7.281812297734629e-06,
      "loss": 0.024,
      "step": 245620
    },
    {
      "epoch": 95.39,
      "learning_rate": 7.281294498381878e-06,
      "loss": 0.112,
      "step": 245630
    },
    {
      "epoch": 95.39,
      "learning_rate": 7.280776699029127e-06,
      "loss": 0.149,
      "step": 245640
    },
    {
      "epoch": 95.4,
      "learning_rate": 7.280258899676376e-06,
      "loss": 0.0339,
      "step": 245650
    },
    {
      "epoch": 95.4,
      "learning_rate": 7.279741100323625e-06,
      "loss": 0.0955,
      "step": 245660
    },
    {
      "epoch": 95.41,
      "learning_rate": 7.279223300970874e-06,
      "loss": 0.0897,
      "step": 245670
    },
    {
      "epoch": 95.41,
      "learning_rate": 7.278705501618124e-06,
      "loss": 0.043,
      "step": 245680
    },
    {
      "epoch": 95.41,
      "learning_rate": 7.2781877022653726e-06,
      "loss": 0.0914,
      "step": 245690
    },
    {
      "epoch": 95.42,
      "learning_rate": 7.277669902912622e-06,
      "loss": 0.1133,
      "step": 245700
    },
    {
      "epoch": 95.42,
      "learning_rate": 7.277152103559871e-06,
      "loss": 0.0971,
      "step": 245710
    },
    {
      "epoch": 95.43,
      "learning_rate": 7.276634304207121e-06,
      "loss": 0.033,
      "step": 245720
    },
    {
      "epoch": 95.43,
      "learning_rate": 7.276116504854369e-06,
      "loss": 0.047,
      "step": 245730
    },
    {
      "epoch": 95.43,
      "learning_rate": 7.275598705501619e-06,
      "loss": 0.0484,
      "step": 245740
    },
    {
      "epoch": 95.44,
      "learning_rate": 7.275080906148868e-06,
      "loss": 0.0011,
      "step": 245750
    },
    {
      "epoch": 95.44,
      "learning_rate": 7.274563106796117e-06,
      "loss": 0.1063,
      "step": 245760
    },
    {
      "epoch": 95.44,
      "learning_rate": 7.274045307443366e-06,
      "loss": 0.1546,
      "step": 245770
    },
    {
      "epoch": 95.45,
      "learning_rate": 7.273527508090616e-06,
      "loss": 0.0508,
      "step": 245780
    },
    {
      "epoch": 95.45,
      "learning_rate": 7.2730097087378645e-06,
      "loss": 0.1168,
      "step": 245790
    },
    {
      "epoch": 95.46,
      "learning_rate": 7.272491909385114e-06,
      "loss": 0.0242,
      "step": 245800
    },
    {
      "epoch": 95.46,
      "learning_rate": 7.271974110032363e-06,
      "loss": 0.0135,
      "step": 245810
    },
    {
      "epoch": 95.46,
      "learning_rate": 7.2714563106796125e-06,
      "loss": 0.0446,
      "step": 245820
    },
    {
      "epoch": 95.47,
      "learning_rate": 7.270938511326861e-06,
      "loss": 0.1222,
      "step": 245830
    },
    {
      "epoch": 95.47,
      "learning_rate": 7.270420711974111e-06,
      "loss": 0.0665,
      "step": 245840
    },
    {
      "epoch": 95.48,
      "learning_rate": 7.26990291262136e-06,
      "loss": 0.1829,
      "step": 245850
    },
    {
      "epoch": 95.48,
      "learning_rate": 7.2693851132686085e-06,
      "loss": 0.0298,
      "step": 245860
    },
    {
      "epoch": 95.48,
      "learning_rate": 7.268867313915858e-06,
      "loss": 0.1288,
      "step": 245870
    },
    {
      "epoch": 95.49,
      "learning_rate": 7.268349514563108e-06,
      "loss": 0.1544,
      "step": 245880
    },
    {
      "epoch": 95.49,
      "learning_rate": 7.2678317152103565e-06,
      "loss": 0.0296,
      "step": 245890
    },
    {
      "epoch": 95.5,
      "learning_rate": 7.267313915857605e-06,
      "loss": 0.0698,
      "step": 245900
    },
    {
      "epoch": 95.5,
      "learning_rate": 7.266796116504855e-06,
      "loss": 0.113,
      "step": 245910
    },
    {
      "epoch": 95.5,
      "learning_rate": 7.2662783171521045e-06,
      "loss": 0.0402,
      "step": 245920
    },
    {
      "epoch": 95.51,
      "learning_rate": 7.265760517799353e-06,
      "loss": 0.0471,
      "step": 245930
    },
    {
      "epoch": 95.51,
      "learning_rate": 7.265242718446602e-06,
      "loss": 0.0138,
      "step": 245940
    },
    {
      "epoch": 95.51,
      "learning_rate": 7.264724919093852e-06,
      "loss": 0.0555,
      "step": 245950
    },
    {
      "epoch": 95.52,
      "learning_rate": 7.264207119741101e-06,
      "loss": 0.0625,
      "step": 245960
    },
    {
      "epoch": 95.52,
      "learning_rate": 7.26368932038835e-06,
      "loss": 0.0003,
      "step": 245970
    },
    {
      "epoch": 95.53,
      "learning_rate": 7.263171521035599e-06,
      "loss": 0.1187,
      "step": 245980
    },
    {
      "epoch": 95.53,
      "learning_rate": 7.262653721682848e-06,
      "loss": 0.0399,
      "step": 245990
    },
    {
      "epoch": 95.53,
      "learning_rate": 7.262135922330098e-06,
      "loss": 0.0194,
      "step": 246000
    },
    {
      "epoch": 95.54,
      "learning_rate": 7.261618122977346e-06,
      "loss": 0.0762,
      "step": 246010
    },
    {
      "epoch": 95.54,
      "learning_rate": 7.2611003236245956e-06,
      "loss": 0.1149,
      "step": 246020
    },
    {
      "epoch": 95.55,
      "learning_rate": 7.260582524271845e-06,
      "loss": 0.0111,
      "step": 246030
    },
    {
      "epoch": 95.55,
      "learning_rate": 7.260064724919095e-06,
      "loss": 0.0216,
      "step": 246040
    },
    {
      "epoch": 95.55,
      "learning_rate": 7.259546925566343e-06,
      "loss": 0.1582,
      "step": 246050
    },
    {
      "epoch": 95.56,
      "learning_rate": 7.259029126213592e-06,
      "loss": 0.113,
      "step": 246060
    },
    {
      "epoch": 95.56,
      "learning_rate": 7.258511326860842e-06,
      "loss": 0.0207,
      "step": 246070
    },
    {
      "epoch": 95.57,
      "learning_rate": 7.257993527508092e-06,
      "loss": 0.0522,
      "step": 246080
    },
    {
      "epoch": 95.57,
      "learning_rate": 7.2574757281553395e-06,
      "loss": 0.1453,
      "step": 246090
    },
    {
      "epoch": 95.57,
      "learning_rate": 7.256957928802589e-06,
      "loss": 0.0619,
      "step": 246100
    },
    {
      "epoch": 95.58,
      "learning_rate": 7.256440129449839e-06,
      "loss": 0.0292,
      "step": 246110
    },
    {
      "epoch": 95.58,
      "learning_rate": 7.255922330097088e-06,
      "loss": 0.0181,
      "step": 246120
    },
    {
      "epoch": 95.58,
      "learning_rate": 7.255404530744338e-06,
      "loss": 0.0787,
      "step": 246130
    },
    {
      "epoch": 95.59,
      "learning_rate": 7.254886731391586e-06,
      "loss": 0.1276,
      "step": 246140
    },
    {
      "epoch": 95.59,
      "learning_rate": 7.2543689320388355e-06,
      "loss": 0.0256,
      "step": 246150
    },
    {
      "epoch": 95.6,
      "learning_rate": 7.253851132686085e-06,
      "loss": 0.0245,
      "step": 246160
    },
    {
      "epoch": 95.6,
      "learning_rate": 7.253333333333335e-06,
      "loss": 0.0542,
      "step": 246170
    },
    {
      "epoch": 95.6,
      "learning_rate": 7.252815533980583e-06,
      "loss": 0.1703,
      "step": 246180
    },
    {
      "epoch": 95.61,
      "learning_rate": 7.252297734627832e-06,
      "loss": 0.0797,
      "step": 246190
    },
    {
      "epoch": 95.61,
      "learning_rate": 7.251779935275082e-06,
      "loss": 0.0938,
      "step": 246200
    },
    {
      "epoch": 95.62,
      "learning_rate": 7.2512621359223315e-06,
      "loss": 0.0736,
      "step": 246210
    },
    {
      "epoch": 95.62,
      "learning_rate": 7.2507443365695795e-06,
      "loss": 0.0429,
      "step": 246220
    },
    {
      "epoch": 95.62,
      "learning_rate": 7.250226537216829e-06,
      "loss": 0.0833,
      "step": 246230
    },
    {
      "epoch": 95.63,
      "learning_rate": 7.249708737864079e-06,
      "loss": 0.174,
      "step": 246240
    },
    {
      "epoch": 95.63,
      "learning_rate": 7.249190938511328e-06,
      "loss": 0.0499,
      "step": 246250
    },
    {
      "epoch": 95.63,
      "learning_rate": 7.248673139158576e-06,
      "loss": 0.0458,
      "step": 246260
    },
    {
      "epoch": 95.64,
      "learning_rate": 7.248155339805826e-06,
      "loss": 0.061,
      "step": 246270
    },
    {
      "epoch": 95.64,
      "learning_rate": 7.2476375404530755e-06,
      "loss": 0.0004,
      "step": 246280
    },
    {
      "epoch": 95.65,
      "learning_rate": 7.247119741100325e-06,
      "loss": 0.1346,
      "step": 246290
    },
    {
      "epoch": 95.65,
      "learning_rate": 7.246601941747573e-06,
      "loss": 0.0189,
      "step": 246300
    },
    {
      "epoch": 95.65,
      "learning_rate": 7.246084142394823e-06,
      "loss": 0.0581,
      "step": 246310
    },
    {
      "epoch": 95.66,
      "learning_rate": 7.245566343042072e-06,
      "loss": 0.0969,
      "step": 246320
    },
    {
      "epoch": 95.66,
      "learning_rate": 7.245048543689321e-06,
      "loss": 0.0275,
      "step": 246330
    },
    {
      "epoch": 95.67,
      "learning_rate": 7.24453074433657e-06,
      "loss": 0.0323,
      "step": 246340
    },
    {
      "epoch": 95.67,
      "learning_rate": 7.244012944983819e-06,
      "loss": 0.0595,
      "step": 246350
    },
    {
      "epoch": 95.67,
      "learning_rate": 7.243495145631069e-06,
      "loss": 0.0963,
      "step": 246360
    },
    {
      "epoch": 95.68,
      "learning_rate": 7.242977346278318e-06,
      "loss": 0.0536,
      "step": 246370
    },
    {
      "epoch": 95.68,
      "learning_rate": 7.2424595469255666e-06,
      "loss": 0.0002,
      "step": 246380
    },
    {
      "epoch": 95.69,
      "learning_rate": 7.241941747572816e-06,
      "loss": 0.0885,
      "step": 246390
    },
    {
      "epoch": 95.69,
      "learning_rate": 7.241423948220066e-06,
      "loss": 0.0626,
      "step": 246400
    },
    {
      "epoch": 95.69,
      "learning_rate": 7.240906148867315e-06,
      "loss": 0.1123,
      "step": 246410
    },
    {
      "epoch": 95.7,
      "learning_rate": 7.240388349514563e-06,
      "loss": 0.1517,
      "step": 246420
    },
    {
      "epoch": 95.7,
      "learning_rate": 7.239870550161813e-06,
      "loss": 0.0767,
      "step": 246430
    },
    {
      "epoch": 95.7,
      "learning_rate": 7.239352750809063e-06,
      "loss": 0.0441,
      "step": 246440
    },
    {
      "epoch": 95.71,
      "learning_rate": 7.238834951456311e-06,
      "loss": 0.0603,
      "step": 246450
    },
    {
      "epoch": 95.71,
      "learning_rate": 7.23831715210356e-06,
      "loss": 0.131,
      "step": 246460
    },
    {
      "epoch": 95.72,
      "learning_rate": 7.23779935275081e-06,
      "loss": 0.0352,
      "step": 246470
    },
    {
      "epoch": 95.72,
      "learning_rate": 7.2372815533980585e-06,
      "loss": 0.0181,
      "step": 246480
    },
    {
      "epoch": 95.72,
      "learning_rate": 7.236763754045308e-06,
      "loss": 0.0187,
      "step": 246490
    },
    {
      "epoch": 95.73,
      "learning_rate": 7.236245954692557e-06,
      "loss": 0.0732,
      "step": 246500
    },
    {
      "epoch": 95.73,
      "learning_rate": 7.2357281553398065e-06,
      "loss": 0.0788,
      "step": 246510
    },
    {
      "epoch": 95.74,
      "learning_rate": 7.235210355987055e-06,
      "loss": 0.0341,
      "step": 246520
    },
    {
      "epoch": 95.74,
      "learning_rate": 7.234692556634305e-06,
      "loss": 0.1159,
      "step": 246530
    },
    {
      "epoch": 95.74,
      "learning_rate": 7.234174757281554e-06,
      "loss": 0.2356,
      "step": 246540
    },
    {
      "epoch": 95.75,
      "learning_rate": 7.233656957928803e-06,
      "loss": 0.0041,
      "step": 246550
    },
    {
      "epoch": 95.75,
      "learning_rate": 7.233139158576052e-06,
      "loss": 0.0111,
      "step": 246560
    },
    {
      "epoch": 95.76,
      "learning_rate": 7.232621359223302e-06,
      "loss": 0.0641,
      "step": 246570
    },
    {
      "epoch": 95.76,
      "learning_rate": 7.2321035598705505e-06,
      "loss": 0.1044,
      "step": 246580
    },
    {
      "epoch": 95.76,
      "learning_rate": 7.2315857605178e-06,
      "loss": 0.0005,
      "step": 246590
    },
    {
      "epoch": 95.77,
      "learning_rate": 7.231067961165049e-06,
      "loss": 0.0523,
      "step": 246600
    },
    {
      "epoch": 95.77,
      "learning_rate": 7.2305501618122985e-06,
      "loss": 0.0525,
      "step": 246610
    },
    {
      "epoch": 95.77,
      "learning_rate": 7.230032362459547e-06,
      "loss": 0.0911,
      "step": 246620
    },
    {
      "epoch": 95.78,
      "learning_rate": 7.229514563106797e-06,
      "loss": 0.1091,
      "step": 246630
    },
    {
      "epoch": 95.78,
      "learning_rate": 7.228996763754046e-06,
      "loss": 0.0457,
      "step": 246640
    },
    {
      "epoch": 95.79,
      "learning_rate": 7.228478964401295e-06,
      "loss": 0.0365,
      "step": 246650
    },
    {
      "epoch": 95.79,
      "learning_rate": 7.227961165048544e-06,
      "loss": 0.0369,
      "step": 246660
    },
    {
      "epoch": 95.79,
      "learning_rate": 7.227443365695793e-06,
      "loss": 0.0456,
      "step": 246670
    },
    {
      "epoch": 95.8,
      "learning_rate": 7.226925566343042e-06,
      "loss": 0.091,
      "step": 246680
    },
    {
      "epoch": 95.8,
      "learning_rate": 7.226407766990292e-06,
      "loss": 0.037,
      "step": 246690
    },
    {
      "epoch": 95.81,
      "learning_rate": 7.225889967637542e-06,
      "loss": 0.0381,
      "step": 246700
    },
    {
      "epoch": 95.81,
      "learning_rate": 7.2253721682847896e-06,
      "loss": 0.0078,
      "step": 246710
    },
    {
      "epoch": 95.81,
      "learning_rate": 7.224854368932039e-06,
      "loss": 0.0085,
      "step": 246720
    },
    {
      "epoch": 95.82,
      "learning_rate": 7.224336569579289e-06,
      "loss": 0.0524,
      "step": 246730
    },
    {
      "epoch": 95.82,
      "learning_rate": 7.223818770226538e-06,
      "loss": 0.0325,
      "step": 246740
    },
    {
      "epoch": 95.83,
      "learning_rate": 7.223300970873786e-06,
      "loss": 0.0107,
      "step": 246750
    },
    {
      "epoch": 95.83,
      "learning_rate": 7.222783171521036e-06,
      "loss": 0.0523,
      "step": 246760
    },
    {
      "epoch": 95.83,
      "learning_rate": 7.222265372168286e-06,
      "loss": 0.0529,
      "step": 246770
    },
    {
      "epoch": 95.84,
      "learning_rate": 7.221747572815535e-06,
      "loss": 0.0606,
      "step": 246780
    },
    {
      "epoch": 95.84,
      "learning_rate": 7.221229773462783e-06,
      "loss": 0.016,
      "step": 246790
    },
    {
      "epoch": 95.84,
      "learning_rate": 7.220711974110033e-06,
      "loss": 0.0627,
      "step": 246800
    },
    {
      "epoch": 95.85,
      "learning_rate": 7.220194174757282e-06,
      "loss": 0.0002,
      "step": 246810
    },
    {
      "epoch": 95.85,
      "learning_rate": 7.219676375404532e-06,
      "loss": 0.062,
      "step": 246820
    },
    {
      "epoch": 95.86,
      "learning_rate": 7.21915857605178e-06,
      "loss": 0.1087,
      "step": 246830
    },
    {
      "epoch": 95.86,
      "learning_rate": 7.2186407766990295e-06,
      "loss": 0.0852,
      "step": 246840
    },
    {
      "epoch": 95.86,
      "learning_rate": 7.218122977346279e-06,
      "loss": 0.0002,
      "step": 246850
    },
    {
      "epoch": 95.87,
      "learning_rate": 7.217605177993529e-06,
      "loss": 0.1184,
      "step": 246860
    },
    {
      "epoch": 95.87,
      "learning_rate": 7.217087378640777e-06,
      "loss": 0.0368,
      "step": 246870
    },
    {
      "epoch": 95.88,
      "learning_rate": 7.216569579288026e-06,
      "loss": 0.1051,
      "step": 246880
    },
    {
      "epoch": 95.88,
      "learning_rate": 7.216051779935276e-06,
      "loss": 0.1661,
      "step": 246890
    },
    {
      "epoch": 95.88,
      "learning_rate": 7.2155339805825255e-06,
      "loss": 0.0513,
      "step": 246900
    },
    {
      "epoch": 95.89,
      "learning_rate": 7.2150161812297735e-06,
      "loss": 0.0354,
      "step": 246910
    },
    {
      "epoch": 95.89,
      "learning_rate": 7.214498381877023e-06,
      "loss": 0.0887,
      "step": 246920
    },
    {
      "epoch": 95.9,
      "learning_rate": 7.213980582524273e-06,
      "loss": 0.0697,
      "step": 246930
    },
    {
      "epoch": 95.9,
      "learning_rate": 7.213462783171522e-06,
      "loss": 0.0464,
      "step": 246940
    },
    {
      "epoch": 95.9,
      "learning_rate": 7.21294498381877e-06,
      "loss": 0.0716,
      "step": 246950
    },
    {
      "epoch": 95.91,
      "learning_rate": 7.21242718446602e-06,
      "loss": 0.065,
      "step": 246960
    },
    {
      "epoch": 95.91,
      "learning_rate": 7.2119093851132695e-06,
      "loss": 0.0833,
      "step": 246970
    },
    {
      "epoch": 95.91,
      "learning_rate": 7.211391585760519e-06,
      "loss": 0.0228,
      "step": 246980
    },
    {
      "epoch": 95.92,
      "learning_rate": 7.210873786407767e-06,
      "loss": 0.0016,
      "step": 246990
    },
    {
      "epoch": 95.92,
      "learning_rate": 7.210355987055017e-06,
      "loss": 0.0003,
      "step": 247000
    },
    {
      "epoch": 95.93,
      "learning_rate": 7.209838187702266e-06,
      "loss": 0.0665,
      "step": 247010
    },
    {
      "epoch": 95.93,
      "learning_rate": 7.209320388349516e-06,
      "loss": 0.006,
      "step": 247020
    },
    {
      "epoch": 95.93,
      "learning_rate": 7.208802588996764e-06,
      "loss": 0.0606,
      "step": 247030
    },
    {
      "epoch": 95.94,
      "learning_rate": 7.208284789644013e-06,
      "loss": 0.0434,
      "step": 247040
    },
    {
      "epoch": 95.94,
      "learning_rate": 7.207766990291263e-06,
      "loss": 0.0156,
      "step": 247050
    },
    {
      "epoch": 95.95,
      "learning_rate": 7.207249190938513e-06,
      "loss": 0.0765,
      "step": 247060
    },
    {
      "epoch": 95.95,
      "learning_rate": 7.2067313915857606e-06,
      "loss": 0.0931,
      "step": 247070
    },
    {
      "epoch": 95.95,
      "learning_rate": 7.20621359223301e-06,
      "loss": 0.1202,
      "step": 247080
    },
    {
      "epoch": 95.96,
      "learning_rate": 7.20569579288026e-06,
      "loss": 0.0226,
      "step": 247090
    },
    {
      "epoch": 95.96,
      "learning_rate": 7.205177993527509e-06,
      "loss": 0.0421,
      "step": 247100
    },
    {
      "epoch": 95.97,
      "learning_rate": 7.204660194174757e-06,
      "loss": 0.0209,
      "step": 247110
    },
    {
      "epoch": 95.97,
      "learning_rate": 7.204142394822007e-06,
      "loss": 0.0313,
      "step": 247120
    },
    {
      "epoch": 95.97,
      "learning_rate": 7.203624595469257e-06,
      "loss": 0.0268,
      "step": 247130
    },
    {
      "epoch": 95.98,
      "learning_rate": 7.203106796116505e-06,
      "loss": 0.0032,
      "step": 247140
    },
    {
      "epoch": 95.98,
      "learning_rate": 7.202588996763754e-06,
      "loss": 0.0272,
      "step": 247150
    },
    {
      "epoch": 95.98,
      "learning_rate": 7.202071197411004e-06,
      "loss": 0.0433,
      "step": 247160
    },
    {
      "epoch": 95.99,
      "learning_rate": 7.201553398058253e-06,
      "loss": 0.0104,
      "step": 247170
    },
    {
      "epoch": 95.99,
      "learning_rate": 7.201035598705502e-06,
      "loss": 0.0582,
      "step": 247180
    },
    {
      "epoch": 96.0,
      "learning_rate": 7.200517799352751e-06,
      "loss": 0.0336,
      "step": 247190
    },
    {
      "epoch": 96.0,
      "learning_rate": 7.2000000000000005e-06,
      "loss": 0.2079,
      "step": 247200
    },
    {
      "epoch": 96.0,
      "eval_accuracy": 0.9482806052269601,
      "eval_loss": 0.36884090304374695,
      "eval_runtime": 8.2807,
      "eval_samples_per_second": 438.975,
      "eval_steps_per_second": 54.947,
      "step": 247200
    },
    {
      "epoch": 96.0,
      "learning_rate": 7.19948220064725e-06,
      "loss": 0.0115,
      "step": 247210
    },
    {
      "epoch": 96.01,
      "learning_rate": 7.198964401294499e-06,
      "loss": 0.0944,
      "step": 247220
    },
    {
      "epoch": 96.01,
      "learning_rate": 7.1984466019417485e-06,
      "loss": 0.0194,
      "step": 247230
    },
    {
      "epoch": 96.02,
      "learning_rate": 7.197928802588997e-06,
      "loss": 0.0206,
      "step": 247240
    },
    {
      "epoch": 96.02,
      "learning_rate": 7.197411003236247e-06,
      "loss": 0.0509,
      "step": 247250
    },
    {
      "epoch": 96.02,
      "learning_rate": 7.196893203883496e-06,
      "loss": 0.0372,
      "step": 247260
    },
    {
      "epoch": 96.03,
      "learning_rate": 7.196375404530745e-06,
      "loss": 0.0815,
      "step": 247270
    },
    {
      "epoch": 96.03,
      "learning_rate": 7.195857605177994e-06,
      "loss": 0.0019,
      "step": 247280
    },
    {
      "epoch": 96.03,
      "learning_rate": 7.195339805825244e-06,
      "loss": 0.095,
      "step": 247290
    },
    {
      "epoch": 96.04,
      "learning_rate": 7.1948220064724925e-06,
      "loss": 0.0135,
      "step": 247300
    },
    {
      "epoch": 96.04,
      "learning_rate": 7.194304207119742e-06,
      "loss": 0.01,
      "step": 247310
    },
    {
      "epoch": 96.05,
      "learning_rate": 7.193786407766991e-06,
      "loss": 0.0174,
      "step": 247320
    },
    {
      "epoch": 96.05,
      "learning_rate": 7.19326860841424e-06,
      "loss": 0.0552,
      "step": 247330
    },
    {
      "epoch": 96.05,
      "learning_rate": 7.192750809061489e-06,
      "loss": 0.1177,
      "step": 247340
    },
    {
      "epoch": 96.06,
      "learning_rate": 7.192233009708739e-06,
      "loss": 0.0865,
      "step": 247350
    },
    {
      "epoch": 96.06,
      "learning_rate": 7.191715210355988e-06,
      "loss": 0.0885,
      "step": 247360
    },
    {
      "epoch": 96.07,
      "learning_rate": 7.191197411003236e-06,
      "loss": 0.0114,
      "step": 247370
    },
    {
      "epoch": 96.07,
      "learning_rate": 7.190679611650486e-06,
      "loss": 0.0001,
      "step": 247380
    },
    {
      "epoch": 96.07,
      "learning_rate": 7.190161812297736e-06,
      "loss": 0.1163,
      "step": 247390
    },
    {
      "epoch": 96.08,
      "learning_rate": 7.189644012944984e-06,
      "loss": 0.047,
      "step": 247400
    },
    {
      "epoch": 96.08,
      "learning_rate": 7.189126213592233e-06,
      "loss": 0.1093,
      "step": 247410
    },
    {
      "epoch": 96.09,
      "learning_rate": 7.188608414239483e-06,
      "loss": 0.0845,
      "step": 247420
    },
    {
      "epoch": 96.09,
      "learning_rate": 7.188090614886732e-06,
      "loss": 0.0521,
      "step": 247430
    },
    {
      "epoch": 96.09,
      "learning_rate": 7.187572815533981e-06,
      "loss": 0.0514,
      "step": 247440
    },
    {
      "epoch": 96.1,
      "learning_rate": 7.18705501618123e-06,
      "loss": 0.0385,
      "step": 247450
    },
    {
      "epoch": 96.1,
      "learning_rate": 7.18653721682848e-06,
      "loss": 0.004,
      "step": 247460
    },
    {
      "epoch": 96.1,
      "learning_rate": 7.186019417475729e-06,
      "loss": 0.1586,
      "step": 247470
    },
    {
      "epoch": 96.11,
      "learning_rate": 7.185501618122977e-06,
      "loss": 0.0027,
      "step": 247480
    },
    {
      "epoch": 96.11,
      "learning_rate": 7.184983818770227e-06,
      "loss": 0.0364,
      "step": 247490
    },
    {
      "epoch": 96.12,
      "learning_rate": 7.184466019417476e-06,
      "loss": 0.0447,
      "step": 247500
    },
    {
      "epoch": 96.12,
      "learning_rate": 7.183948220064726e-06,
      "loss": 0.0228,
      "step": 247510
    },
    {
      "epoch": 96.12,
      "learning_rate": 7.183430420711974e-06,
      "loss": 0.0812,
      "step": 247520
    },
    {
      "epoch": 96.13,
      "learning_rate": 7.1829126213592235e-06,
      "loss": 0.054,
      "step": 247530
    },
    {
      "epoch": 96.13,
      "learning_rate": 7.182394822006473e-06,
      "loss": 0.0872,
      "step": 247540
    },
    {
      "epoch": 96.14,
      "learning_rate": 7.181877022653723e-06,
      "loss": 0.0234,
      "step": 247550
    },
    {
      "epoch": 96.14,
      "learning_rate": 7.181359223300971e-06,
      "loss": 0.0471,
      "step": 247560
    },
    {
      "epoch": 96.14,
      "learning_rate": 7.18084142394822e-06,
      "loss": 0.0147,
      "step": 247570
    },
    {
      "epoch": 96.15,
      "learning_rate": 7.18032362459547e-06,
      "loss": 0.082,
      "step": 247580
    },
    {
      "epoch": 96.15,
      "learning_rate": 7.1798058252427195e-06,
      "loss": 0.0611,
      "step": 247590
    },
    {
      "epoch": 96.16,
      "learning_rate": 7.1792880258899675e-06,
      "loss": 0.0309,
      "step": 247600
    },
    {
      "epoch": 96.16,
      "learning_rate": 7.178770226537217e-06,
      "loss": 0.0054,
      "step": 247610
    },
    {
      "epoch": 96.16,
      "learning_rate": 7.178252427184467e-06,
      "loss": 0.014,
      "step": 247620
    },
    {
      "epoch": 96.17,
      "learning_rate": 7.177734627831716e-06,
      "loss": 0.0001,
      "step": 247630
    },
    {
      "epoch": 96.17,
      "learning_rate": 7.177216828478964e-06,
      "loss": 0.0272,
      "step": 247640
    },
    {
      "epoch": 96.17,
      "learning_rate": 7.176699029126214e-06,
      "loss": 0.0316,
      "step": 247650
    },
    {
      "epoch": 96.18,
      "learning_rate": 7.1761812297734635e-06,
      "loss": 0.0974,
      "step": 247660
    },
    {
      "epoch": 96.18,
      "learning_rate": 7.175663430420713e-06,
      "loss": 0.0412,
      "step": 247670
    },
    {
      "epoch": 96.19,
      "learning_rate": 7.175145631067961e-06,
      "loss": 0.1019,
      "step": 247680
    },
    {
      "epoch": 96.19,
      "learning_rate": 7.174627831715211e-06,
      "loss": 0.0394,
      "step": 247690
    },
    {
      "epoch": 96.19,
      "learning_rate": 7.17411003236246e-06,
      "loss": 0.0074,
      "step": 247700
    },
    {
      "epoch": 96.2,
      "learning_rate": 7.17359223300971e-06,
      "loss": 0.1972,
      "step": 247710
    },
    {
      "epoch": 96.2,
      "learning_rate": 7.173074433656958e-06,
      "loss": 0.0999,
      "step": 247720
    },
    {
      "epoch": 96.21,
      "learning_rate": 7.172556634304207e-06,
      "loss": 0.0584,
      "step": 247730
    },
    {
      "epoch": 96.21,
      "learning_rate": 7.172038834951457e-06,
      "loss": 0.0002,
      "step": 247740
    },
    {
      "epoch": 96.21,
      "learning_rate": 7.171521035598707e-06,
      "loss": 0.0374,
      "step": 247750
    },
    {
      "epoch": 96.22,
      "learning_rate": 7.1710032362459546e-06,
      "loss": 0.1763,
      "step": 247760
    },
    {
      "epoch": 96.22,
      "learning_rate": 7.170485436893204e-06,
      "loss": 0.0472,
      "step": 247770
    },
    {
      "epoch": 96.23,
      "learning_rate": 7.169967637540454e-06,
      "loss": 0.0876,
      "step": 247780
    },
    {
      "epoch": 96.23,
      "learning_rate": 7.169449838187703e-06,
      "loss": 0.0608,
      "step": 247790
    },
    {
      "epoch": 96.23,
      "learning_rate": 7.168932038834952e-06,
      "loss": 0.0241,
      "step": 247800
    },
    {
      "epoch": 96.24,
      "learning_rate": 7.168414239482201e-06,
      "loss": 0.027,
      "step": 247810
    },
    {
      "epoch": 96.24,
      "learning_rate": 7.167896440129451e-06,
      "loss": 0.0006,
      "step": 247820
    },
    {
      "epoch": 96.24,
      "learning_rate": 7.1673786407767e-06,
      "loss": 0.0821,
      "step": 247830
    },
    {
      "epoch": 96.25,
      "learning_rate": 7.166860841423949e-06,
      "loss": 0.0501,
      "step": 247840
    },
    {
      "epoch": 96.25,
      "learning_rate": 7.166343042071198e-06,
      "loss": 0.0482,
      "step": 247850
    },
    {
      "epoch": 96.26,
      "learning_rate": 7.165825242718447e-06,
      "loss": 0.0002,
      "step": 247860
    },
    {
      "epoch": 96.26,
      "learning_rate": 7.165307443365697e-06,
      "loss": 0.0433,
      "step": 247870
    },
    {
      "epoch": 96.26,
      "learning_rate": 7.164789644012946e-06,
      "loss": 0.0376,
      "step": 247880
    },
    {
      "epoch": 96.27,
      "learning_rate": 7.1642718446601945e-06,
      "loss": 0.1897,
      "step": 247890
    },
    {
      "epoch": 96.27,
      "learning_rate": 7.163754045307444e-06,
      "loss": 0.045,
      "step": 247900
    },
    {
      "epoch": 96.28,
      "learning_rate": 7.163236245954694e-06,
      "loss": 0.0159,
      "step": 247910
    },
    {
      "epoch": 96.28,
      "learning_rate": 7.1627184466019425e-06,
      "loss": 0.1948,
      "step": 247920
    },
    {
      "epoch": 96.28,
      "learning_rate": 7.162200647249191e-06,
      "loss": 0.0099,
      "step": 247930
    },
    {
      "epoch": 96.29,
      "learning_rate": 7.161682847896441e-06,
      "loss": 0.0708,
      "step": 247940
    },
    {
      "epoch": 96.29,
      "learning_rate": 7.16116504854369e-06,
      "loss": 0.1057,
      "step": 247950
    },
    {
      "epoch": 96.3,
      "learning_rate": 7.160647249190939e-06,
      "loss": 0.0038,
      "step": 247960
    },
    {
      "epoch": 96.3,
      "learning_rate": 7.160129449838188e-06,
      "loss": 0.0236,
      "step": 247970
    },
    {
      "epoch": 96.3,
      "learning_rate": 7.159611650485438e-06,
      "loss": 0.0692,
      "step": 247980
    },
    {
      "epoch": 96.31,
      "learning_rate": 7.1590938511326865e-06,
      "loss": 0.1074,
      "step": 247990
    },
    {
      "epoch": 96.31,
      "learning_rate": 7.158576051779936e-06,
      "loss": 0.0571,
      "step": 248000
    },
    {
      "epoch": 96.31,
      "learning_rate": 7.158058252427185e-06,
      "loss": 0.0557,
      "step": 248010
    },
    {
      "epoch": 96.32,
      "learning_rate": 7.1575404530744345e-06,
      "loss": 0.1576,
      "step": 248020
    },
    {
      "epoch": 96.32,
      "learning_rate": 7.157022653721683e-06,
      "loss": 0.0176,
      "step": 248030
    },
    {
      "epoch": 96.33,
      "learning_rate": 7.156504854368933e-06,
      "loss": 0.0459,
      "step": 248040
    },
    {
      "epoch": 96.33,
      "learning_rate": 7.155987055016182e-06,
      "loss": 0.1444,
      "step": 248050
    },
    {
      "epoch": 96.33,
      "learning_rate": 7.155469255663431e-06,
      "loss": 0.0125,
      "step": 248060
    },
    {
      "epoch": 96.34,
      "learning_rate": 7.15495145631068e-06,
      "loss": 0.0667,
      "step": 248070
    },
    {
      "epoch": 96.34,
      "learning_rate": 7.15443365695793e-06,
      "loss": 0.0868,
      "step": 248080
    },
    {
      "epoch": 96.35,
      "learning_rate": 7.153915857605178e-06,
      "loss": 0.1282,
      "step": 248090
    },
    {
      "epoch": 96.35,
      "learning_rate": 7.153398058252428e-06,
      "loss": 0.0337,
      "step": 248100
    },
    {
      "epoch": 96.35,
      "learning_rate": 7.152880258899677e-06,
      "loss": 0.0128,
      "step": 248110
    },
    {
      "epoch": 96.36,
      "learning_rate": 7.152362459546926e-06,
      "loss": 0.0661,
      "step": 248120
    },
    {
      "epoch": 96.36,
      "learning_rate": 7.151844660194175e-06,
      "loss": 0.0992,
      "step": 248130
    },
    {
      "epoch": 96.37,
      "learning_rate": 7.151326860841424e-06,
      "loss": 0.1058,
      "step": 248140
    },
    {
      "epoch": 96.37,
      "learning_rate": 7.150809061488674e-06,
      "loss": 0.0173,
      "step": 248150
    },
    {
      "epoch": 96.37,
      "learning_rate": 7.150291262135923e-06,
      "loss": 0.0943,
      "step": 248160
    },
    {
      "epoch": 96.38,
      "learning_rate": 7.149773462783172e-06,
      "loss": 0.0996,
      "step": 248170
    },
    {
      "epoch": 96.38,
      "learning_rate": 7.149255663430421e-06,
      "loss": 0.0376,
      "step": 248180
    },
    {
      "epoch": 96.38,
      "learning_rate": 7.14873786407767e-06,
      "loss": 0.0226,
      "step": 248190
    },
    {
      "epoch": 96.39,
      "learning_rate": 7.14822006472492e-06,
      "loss": 0.0542,
      "step": 248200
    },
    {
      "epoch": 96.39,
      "learning_rate": 7.147702265372169e-06,
      "loss": 0.0413,
      "step": 248210
    },
    {
      "epoch": 96.4,
      "learning_rate": 7.1471844660194175e-06,
      "loss": 0.1502,
      "step": 248220
    },
    {
      "epoch": 96.4,
      "learning_rate": 7.146666666666667e-06,
      "loss": 0.0124,
      "step": 248230
    },
    {
      "epoch": 96.4,
      "learning_rate": 7.146148867313917e-06,
      "loss": 0.0535,
      "step": 248240
    },
    {
      "epoch": 96.41,
      "learning_rate": 7.1456310679611655e-06,
      "loss": 0.0032,
      "step": 248250
    },
    {
      "epoch": 96.41,
      "learning_rate": 7.145113268608414e-06,
      "loss": 0.022,
      "step": 248260
    },
    {
      "epoch": 96.42,
      "learning_rate": 7.144595469255664e-06,
      "loss": 0.0848,
      "step": 248270
    },
    {
      "epoch": 96.42,
      "learning_rate": 7.1440776699029135e-06,
      "loss": 0.0009,
      "step": 248280
    },
    {
      "epoch": 96.42,
      "learning_rate": 7.1435598705501615e-06,
      "loss": 0.0429,
      "step": 248290
    },
    {
      "epoch": 96.43,
      "learning_rate": 7.143042071197411e-06,
      "loss": 0.0245,
      "step": 248300
    },
    {
      "epoch": 96.43,
      "learning_rate": 7.142524271844661e-06,
      "loss": 0.0604,
      "step": 248310
    },
    {
      "epoch": 96.43,
      "learning_rate": 7.14200647249191e-06,
      "loss": 0.0618,
      "step": 248320
    },
    {
      "epoch": 96.44,
      "learning_rate": 7.141488673139158e-06,
      "loss": 0.0712,
      "step": 248330
    },
    {
      "epoch": 96.44,
      "learning_rate": 7.140970873786408e-06,
      "loss": 0.0482,
      "step": 248340
    },
    {
      "epoch": 96.45,
      "learning_rate": 7.1404530744336575e-06,
      "loss": 0.0775,
      "step": 248350
    },
    {
      "epoch": 96.45,
      "learning_rate": 7.139935275080907e-06,
      "loss": 0.0865,
      "step": 248360
    },
    {
      "epoch": 96.45,
      "learning_rate": 7.139417475728157e-06,
      "loss": 0.1499,
      "step": 248370
    },
    {
      "epoch": 96.46,
      "learning_rate": 7.138899676375405e-06,
      "loss": 0.0912,
      "step": 248380
    },
    {
      "epoch": 96.46,
      "learning_rate": 7.138381877022654e-06,
      "loss": 0.0621,
      "step": 248390
    },
    {
      "epoch": 96.47,
      "learning_rate": 7.137864077669904e-06,
      "loss": 0.1173,
      "step": 248400
    },
    {
      "epoch": 96.47,
      "learning_rate": 7.1373462783171535e-06,
      "loss": 0.0212,
      "step": 248410
    },
    {
      "epoch": 96.47,
      "learning_rate": 7.136828478964401e-06,
      "loss": 0.0707,
      "step": 248420
    },
    {
      "epoch": 96.48,
      "learning_rate": 7.136310679611651e-06,
      "loss": 0.0034,
      "step": 248430
    },
    {
      "epoch": 96.48,
      "learning_rate": 7.135792880258901e-06,
      "loss": 0.0112,
      "step": 248440
    },
    {
      "epoch": 96.49,
      "learning_rate": 7.13527508090615e-06,
      "loss": 0.0347,
      "step": 248450
    },
    {
      "epoch": 96.49,
      "learning_rate": 7.134757281553398e-06,
      "loss": 0.0198,
      "step": 248460
    },
    {
      "epoch": 96.49,
      "learning_rate": 7.134239482200648e-06,
      "loss": 0.0573,
      "step": 248470
    },
    {
      "epoch": 96.5,
      "learning_rate": 7.133721682847897e-06,
      "loss": 0.0111,
      "step": 248480
    },
    {
      "epoch": 96.5,
      "learning_rate": 7.133203883495147e-06,
      "loss": 0.0065,
      "step": 248490
    },
    {
      "epoch": 96.5,
      "learning_rate": 7.132686084142395e-06,
      "loss": 0.0243,
      "step": 248500
    },
    {
      "epoch": 96.51,
      "learning_rate": 7.132168284789645e-06,
      "loss": 0.0511,
      "step": 248510
    },
    {
      "epoch": 96.51,
      "learning_rate": 7.131650485436894e-06,
      "loss": 0.0274,
      "step": 248520
    },
    {
      "epoch": 96.52,
      "learning_rate": 7.131132686084144e-06,
      "loss": 0.0796,
      "step": 248530
    },
    {
      "epoch": 96.52,
      "learning_rate": 7.130614886731392e-06,
      "loss": 0.0351,
      "step": 248540
    },
    {
      "epoch": 96.52,
      "learning_rate": 7.130097087378641e-06,
      "loss": 0.0693,
      "step": 248550
    },
    {
      "epoch": 96.53,
      "learning_rate": 7.129579288025891e-06,
      "loss": 0.1469,
      "step": 248560
    },
    {
      "epoch": 96.53,
      "learning_rate": 7.129061488673141e-06,
      "loss": 0.0034,
      "step": 248570
    },
    {
      "epoch": 96.54,
      "learning_rate": 7.1285436893203885e-06,
      "loss": 0.0005,
      "step": 248580
    },
    {
      "epoch": 96.54,
      "learning_rate": 7.128025889967638e-06,
      "loss": 0.0283,
      "step": 248590
    },
    {
      "epoch": 96.54,
      "learning_rate": 7.127508090614888e-06,
      "loss": 0.0145,
      "step": 248600
    },
    {
      "epoch": 96.55,
      "learning_rate": 7.1269902912621365e-06,
      "loss": 0.0506,
      "step": 248610
    },
    {
      "epoch": 96.55,
      "learning_rate": 7.126472491909385e-06,
      "loss": 0.0526,
      "step": 248620
    },
    {
      "epoch": 96.56,
      "learning_rate": 7.125954692556635e-06,
      "loss": 0.0332,
      "step": 248630
    },
    {
      "epoch": 96.56,
      "learning_rate": 7.1254368932038845e-06,
      "loss": 0.0002,
      "step": 248640
    },
    {
      "epoch": 96.56,
      "learning_rate": 7.124919093851133e-06,
      "loss": 0.0231,
      "step": 248650
    },
    {
      "epoch": 96.57,
      "learning_rate": 7.124401294498382e-06,
      "loss": 0.0241,
      "step": 248660
    },
    {
      "epoch": 96.57,
      "learning_rate": 7.123883495145632e-06,
      "loss": 0.0655,
      "step": 248670
    },
    {
      "epoch": 96.57,
      "learning_rate": 7.123365695792881e-06,
      "loss": 0.0712,
      "step": 248680
    },
    {
      "epoch": 96.58,
      "learning_rate": 7.12284789644013e-06,
      "loss": 0.1203,
      "step": 248690
    },
    {
      "epoch": 96.58,
      "learning_rate": 7.122330097087379e-06,
      "loss": 0.0005,
      "step": 248700
    },
    {
      "epoch": 96.59,
      "learning_rate": 7.1218122977346285e-06,
      "loss": 0.0002,
      "step": 248710
    },
    {
      "epoch": 96.59,
      "learning_rate": 7.121294498381878e-06,
      "loss": 0.0439,
      "step": 248720
    },
    {
      "epoch": 96.59,
      "learning_rate": 7.120776699029127e-06,
      "loss": 0.0776,
      "step": 248730
    },
    {
      "epoch": 96.6,
      "learning_rate": 7.120258899676376e-06,
      "loss": 0.1602,
      "step": 248740
    },
    {
      "epoch": 96.6,
      "learning_rate": 7.119741100323625e-06,
      "loss": 0.0134,
      "step": 248750
    },
    {
      "epoch": 96.61,
      "learning_rate": 7.119223300970874e-06,
      "loss": 0.11,
      "step": 248760
    },
    {
      "epoch": 96.61,
      "learning_rate": 7.118705501618124e-06,
      "loss": 0.0118,
      "step": 248770
    },
    {
      "epoch": 96.61,
      "learning_rate": 7.118187702265372e-06,
      "loss": 0.0038,
      "step": 248780
    },
    {
      "epoch": 96.62,
      "learning_rate": 7.117669902912622e-06,
      "loss": 0.0061,
      "step": 248790
    },
    {
      "epoch": 96.62,
      "learning_rate": 7.117152103559871e-06,
      "loss": 0.0283,
      "step": 248800
    },
    {
      "epoch": 96.63,
      "learning_rate": 7.11663430420712e-06,
      "loss": 0.0005,
      "step": 248810
    },
    {
      "epoch": 96.63,
      "learning_rate": 7.116116504854369e-06,
      "loss": 0.1237,
      "step": 248820
    },
    {
      "epoch": 96.63,
      "learning_rate": 7.115598705501619e-06,
      "loss": 0.1111,
      "step": 248830
    },
    {
      "epoch": 96.64,
      "learning_rate": 7.1150809061488676e-06,
      "loss": 0.039,
      "step": 248840
    },
    {
      "epoch": 96.64,
      "learning_rate": 7.114563106796117e-06,
      "loss": 0.0541,
      "step": 248850
    },
    {
      "epoch": 96.64,
      "learning_rate": 7.114045307443366e-06,
      "loss": 0.083,
      "step": 248860
    },
    {
      "epoch": 96.65,
      "learning_rate": 7.113527508090616e-06,
      "loss": 0.0186,
      "step": 248870
    },
    {
      "epoch": 96.65,
      "learning_rate": 7.113009708737864e-06,
      "loss": 0.0012,
      "step": 248880
    },
    {
      "epoch": 96.66,
      "learning_rate": 7.112491909385114e-06,
      "loss": 0.0392,
      "step": 248890
    },
    {
      "epoch": 96.66,
      "learning_rate": 7.111974110032363e-06,
      "loss": 0.0093,
      "step": 248900
    },
    {
      "epoch": 96.66,
      "learning_rate": 7.111456310679612e-06,
      "loss": 0.1545,
      "step": 248910
    },
    {
      "epoch": 96.67,
      "learning_rate": 7.110938511326861e-06,
      "loss": 0.1075,
      "step": 248920
    },
    {
      "epoch": 96.67,
      "learning_rate": 7.110420711974111e-06,
      "loss": 0.0119,
      "step": 248930
    },
    {
      "epoch": 96.68,
      "learning_rate": 7.10990291262136e-06,
      "loss": 0.0307,
      "step": 248940
    },
    {
      "epoch": 96.68,
      "learning_rate": 7.109385113268608e-06,
      "loss": 0.0497,
      "step": 248950
    },
    {
      "epoch": 96.68,
      "learning_rate": 7.108867313915858e-06,
      "loss": 0.0252,
      "step": 248960
    },
    {
      "epoch": 96.69,
      "learning_rate": 7.1083495145631075e-06,
      "loss": 0.089,
      "step": 248970
    },
    {
      "epoch": 96.69,
      "learning_rate": 7.107831715210357e-06,
      "loss": 0.0501,
      "step": 248980
    },
    {
      "epoch": 96.7,
      "learning_rate": 7.107313915857605e-06,
      "loss": 0.0745,
      "step": 248990
    },
    {
      "epoch": 96.7,
      "learning_rate": 7.106796116504855e-06,
      "loss": 0.0375,
      "step": 249000
    },
    {
      "epoch": 96.7,
      "learning_rate": 7.106278317152104e-06,
      "loss": 0.1,
      "step": 249010
    },
    {
      "epoch": 96.71,
      "learning_rate": 7.105760517799354e-06,
      "loss": 0.043,
      "step": 249020
    },
    {
      "epoch": 96.71,
      "learning_rate": 7.105242718446602e-06,
      "loss": 0.0309,
      "step": 249030
    },
    {
      "epoch": 96.71,
      "learning_rate": 7.1047249190938515e-06,
      "loss": 0.0462,
      "step": 249040
    },
    {
      "epoch": 96.72,
      "learning_rate": 7.104207119741101e-06,
      "loss": 0.0824,
      "step": 249050
    },
    {
      "epoch": 96.72,
      "learning_rate": 7.103689320388351e-06,
      "loss": 0.0669,
      "step": 249060
    },
    {
      "epoch": 96.73,
      "learning_rate": 7.103171521035599e-06,
      "loss": 0.0759,
      "step": 249070
    },
    {
      "epoch": 96.73,
      "learning_rate": 7.102653721682848e-06,
      "loss": 0.0592,
      "step": 249080
    },
    {
      "epoch": 96.73,
      "learning_rate": 7.102135922330098e-06,
      "loss": 0.0317,
      "step": 249090
    },
    {
      "epoch": 96.74,
      "learning_rate": 7.1016181229773475e-06,
      "loss": 0.0005,
      "step": 249100
    },
    {
      "epoch": 96.74,
      "learning_rate": 7.101100323624595e-06,
      "loss": 0.0163,
      "step": 249110
    },
    {
      "epoch": 96.75,
      "learning_rate": 7.100582524271845e-06,
      "loss": 0.0443,
      "step": 249120
    },
    {
      "epoch": 96.75,
      "learning_rate": 7.100064724919095e-06,
      "loss": 0.1801,
      "step": 249130
    },
    {
      "epoch": 96.75,
      "learning_rate": 7.099546925566344e-06,
      "loss": 0.0523,
      "step": 249140
    },
    {
      "epoch": 96.76,
      "learning_rate": 7.099029126213592e-06,
      "loss": 0.0226,
      "step": 249150
    },
    {
      "epoch": 96.76,
      "learning_rate": 7.098511326860842e-06,
      "loss": 0.0277,
      "step": 249160
    },
    {
      "epoch": 96.77,
      "learning_rate": 7.097993527508091e-06,
      "loss": 0.0012,
      "step": 249170
    },
    {
      "epoch": 96.77,
      "learning_rate": 7.097475728155341e-06,
      "loss": 0.0077,
      "step": 249180
    },
    {
      "epoch": 96.77,
      "learning_rate": 7.096957928802589e-06,
      "loss": 0.0731,
      "step": 249190
    },
    {
      "epoch": 96.78,
      "learning_rate": 7.096440129449839e-06,
      "loss": 0.114,
      "step": 249200
    },
    {
      "epoch": 96.78,
      "learning_rate": 7.095922330097088e-06,
      "loss": 0.0102,
      "step": 249210
    },
    {
      "epoch": 96.78,
      "learning_rate": 7.095404530744338e-06,
      "loss": 0.0664,
      "step": 249220
    },
    {
      "epoch": 96.79,
      "learning_rate": 7.094886731391586e-06,
      "loss": 0.0995,
      "step": 249230
    },
    {
      "epoch": 96.79,
      "learning_rate": 7.094368932038835e-06,
      "loss": 0.0846,
      "step": 249240
    },
    {
      "epoch": 96.8,
      "learning_rate": 7.093851132686085e-06,
      "loss": 0.0125,
      "step": 249250
    },
    {
      "epoch": 96.8,
      "learning_rate": 7.093333333333335e-06,
      "loss": 0.0392,
      "step": 249260
    },
    {
      "epoch": 96.8,
      "learning_rate": 7.0928155339805825e-06,
      "loss": 0.0491,
      "step": 249270
    },
    {
      "epoch": 96.81,
      "learning_rate": 7.092297734627832e-06,
      "loss": 0.1044,
      "step": 249280
    },
    {
      "epoch": 96.81,
      "learning_rate": 7.091779935275082e-06,
      "loss": 0.0864,
      "step": 249290
    },
    {
      "epoch": 96.82,
      "learning_rate": 7.091262135922331e-06,
      "loss": 0.0788,
      "step": 249300
    },
    {
      "epoch": 96.82,
      "learning_rate": 7.090744336569579e-06,
      "loss": 0.0428,
      "step": 249310
    },
    {
      "epoch": 96.82,
      "learning_rate": 7.090226537216829e-06,
      "loss": 0.0228,
      "step": 249320
    },
    {
      "epoch": 96.83,
      "learning_rate": 7.0897087378640785e-06,
      "loss": 0.0301,
      "step": 249330
    },
    {
      "epoch": 96.83,
      "learning_rate": 7.089190938511328e-06,
      "loss": 0.0316,
      "step": 249340
    },
    {
      "epoch": 96.83,
      "learning_rate": 7.088673139158576e-06,
      "loss": 0.0596,
      "step": 249350
    },
    {
      "epoch": 96.84,
      "learning_rate": 7.088155339805826e-06,
      "loss": 0.0914,
      "step": 249360
    },
    {
      "epoch": 96.84,
      "learning_rate": 7.087637540453075e-06,
      "loss": 0.0754,
      "step": 249370
    },
    {
      "epoch": 96.85,
      "learning_rate": 7.087119741100325e-06,
      "loss": 0.0159,
      "step": 249380
    },
    {
      "epoch": 96.85,
      "learning_rate": 7.086601941747573e-06,
      "loss": 0.0839,
      "step": 249390
    },
    {
      "epoch": 96.85,
      "learning_rate": 7.0860841423948225e-06,
      "loss": 0.1048,
      "step": 249400
    },
    {
      "epoch": 96.86,
      "learning_rate": 7.085566343042072e-06,
      "loss": 0.0733,
      "step": 249410
    },
    {
      "epoch": 96.86,
      "learning_rate": 7.085048543689321e-06,
      "loss": 0.0001,
      "step": 249420
    },
    {
      "epoch": 96.87,
      "learning_rate": 7.08453074433657e-06,
      "loss": 0.0075,
      "step": 249430
    },
    {
      "epoch": 96.87,
      "learning_rate": 7.084012944983819e-06,
      "loss": 0.0707,
      "step": 249440
    },
    {
      "epoch": 96.87,
      "learning_rate": 7.083495145631069e-06,
      "loss": 0.0854,
      "step": 249450
    },
    {
      "epoch": 96.88,
      "learning_rate": 7.082977346278318e-06,
      "loss": 0.0242,
      "step": 249460
    },
    {
      "epoch": 96.88,
      "learning_rate": 7.082459546925567e-06,
      "loss": 0.0013,
      "step": 249470
    },
    {
      "epoch": 96.89,
      "learning_rate": 7.081941747572816e-06,
      "loss": 0.0659,
      "step": 249480
    },
    {
      "epoch": 96.89,
      "learning_rate": 7.081423948220066e-06,
      "loss": 0.1134,
      "step": 249490
    },
    {
      "epoch": 96.89,
      "learning_rate": 7.080906148867314e-06,
      "loss": 0.0451,
      "step": 249500
    },
    {
      "epoch": 96.9,
      "learning_rate": 7.080388349514564e-06,
      "loss": 0.0397,
      "step": 249510
    },
    {
      "epoch": 96.9,
      "learning_rate": 7.079870550161813e-06,
      "loss": 0.0898,
      "step": 249520
    },
    {
      "epoch": 96.9,
      "learning_rate": 7.079352750809062e-06,
      "loss": 0.1404,
      "step": 249530
    },
    {
      "epoch": 96.91,
      "learning_rate": 7.078834951456311e-06,
      "loss": 0.0727,
      "step": 249540
    },
    {
      "epoch": 96.91,
      "learning_rate": 7.078317152103561e-06,
      "loss": 0.03,
      "step": 249550
    },
    {
      "epoch": 96.92,
      "learning_rate": 7.07779935275081e-06,
      "loss": 0.1876,
      "step": 249560
    },
    {
      "epoch": 96.92,
      "learning_rate": 7.077281553398059e-06,
      "loss": 0.0037,
      "step": 249570
    },
    {
      "epoch": 96.92,
      "learning_rate": 7.076763754045308e-06,
      "loss": 0.0544,
      "step": 249580
    },
    {
      "epoch": 96.93,
      "learning_rate": 7.076245954692558e-06,
      "loss": 0.0016,
      "step": 249590
    },
    {
      "epoch": 96.93,
      "learning_rate": 7.075728155339806e-06,
      "loss": 0.0035,
      "step": 249600
    },
    {
      "epoch": 96.94,
      "learning_rate": 7.075210355987055e-06,
      "loss": 0.0667,
      "step": 249610
    },
    {
      "epoch": 96.94,
      "learning_rate": 7.074692556634305e-06,
      "loss": 0.0899,
      "step": 249620
    },
    {
      "epoch": 96.94,
      "learning_rate": 7.074174757281554e-06,
      "loss": 0.0006,
      "step": 249630
    },
    {
      "epoch": 96.95,
      "learning_rate": 7.073656957928803e-06,
      "loss": 0.0018,
      "step": 249640
    },
    {
      "epoch": 96.95,
      "learning_rate": 7.073139158576052e-06,
      "loss": 0.0411,
      "step": 249650
    },
    {
      "epoch": 96.96,
      "learning_rate": 7.0726213592233015e-06,
      "loss": 0.0527,
      "step": 249660
    },
    {
      "epoch": 96.96,
      "learning_rate": 7.072103559870551e-06,
      "loss": 0.0551,
      "step": 249670
    },
    {
      "epoch": 96.96,
      "learning_rate": 7.0715857605178e-06,
      "loss": 0.0623,
      "step": 249680
    },
    {
      "epoch": 96.97,
      "learning_rate": 7.071067961165049e-06,
      "loss": 0.1161,
      "step": 249690
    },
    {
      "epoch": 96.97,
      "learning_rate": 7.070550161812298e-06,
      "loss": 0.0724,
      "step": 249700
    },
    {
      "epoch": 96.97,
      "learning_rate": 7.070032362459548e-06,
      "loss": 0.0005,
      "step": 249710
    },
    {
      "epoch": 96.98,
      "learning_rate": 7.069514563106797e-06,
      "loss": 0.0003,
      "step": 249720
    },
    {
      "epoch": 96.98,
      "learning_rate": 7.0689967637540455e-06,
      "loss": 0.0628,
      "step": 249730
    },
    {
      "epoch": 96.99,
      "learning_rate": 7.068478964401295e-06,
      "loss": 0.0925,
      "step": 249740
    },
    {
      "epoch": 96.99,
      "learning_rate": 7.067961165048545e-06,
      "loss": 0.0772,
      "step": 249750
    },
    {
      "epoch": 96.99,
      "learning_rate": 7.067443365695793e-06,
      "loss": 0.0257,
      "step": 249760
    },
    {
      "epoch": 97.0,
      "learning_rate": 7.066925566343042e-06,
      "loss": 0.0665,
      "step": 249770
    },
    {
      "epoch": 97.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3576328158378601,
      "eval_runtime": 8.1723,
      "eval_samples_per_second": 444.796,
      "eval_steps_per_second": 55.676,
      "step": 249775
    },
    {
      "epoch": 97.0,
      "learning_rate": 7.066407766990292e-06,
      "loss": 0.0156,
      "step": 249780
    },
    {
      "epoch": 97.01,
      "learning_rate": 7.0658899676375415e-06,
      "loss": 0.015,
      "step": 249790
    },
    {
      "epoch": 97.01,
      "learning_rate": 7.065372168284789e-06,
      "loss": 0.2007,
      "step": 249800
    },
    {
      "epoch": 97.01,
      "learning_rate": 7.064854368932039e-06,
      "loss": 0.0234,
      "step": 249810
    },
    {
      "epoch": 97.02,
      "learning_rate": 7.064336569579289e-06,
      "loss": 0.0006,
      "step": 249820
    },
    {
      "epoch": 97.02,
      "learning_rate": 7.063818770226538e-06,
      "loss": 0.0446,
      "step": 249830
    },
    {
      "epoch": 97.03,
      "learning_rate": 7.063300970873786e-06,
      "loss": 0.0098,
      "step": 249840
    },
    {
      "epoch": 97.03,
      "learning_rate": 7.062783171521036e-06,
      "loss": 0.0435,
      "step": 249850
    },
    {
      "epoch": 97.03,
      "learning_rate": 7.062265372168285e-06,
      "loss": 0.0108,
      "step": 249860
    },
    {
      "epoch": 97.04,
      "learning_rate": 7.061747572815535e-06,
      "loss": 0.0628,
      "step": 249870
    },
    {
      "epoch": 97.04,
      "learning_rate": 7.061229773462783e-06,
      "loss": 0.0472,
      "step": 249880
    },
    {
      "epoch": 97.04,
      "learning_rate": 7.060711974110033e-06,
      "loss": 0.0185,
      "step": 249890
    },
    {
      "epoch": 97.05,
      "learning_rate": 7.060194174757282e-06,
      "loss": 0.1263,
      "step": 249900
    },
    {
      "epoch": 97.05,
      "learning_rate": 7.059676375404532e-06,
      "loss": 0.1439,
      "step": 249910
    },
    {
      "epoch": 97.06,
      "learning_rate": 7.05915857605178e-06,
      "loss": 0.0012,
      "step": 249920
    },
    {
      "epoch": 97.06,
      "learning_rate": 7.058640776699029e-06,
      "loss": 0.1363,
      "step": 249930
    },
    {
      "epoch": 97.06,
      "learning_rate": 7.058122977346279e-06,
      "loss": 0.1063,
      "step": 249940
    },
    {
      "epoch": 97.07,
      "learning_rate": 7.057605177993529e-06,
      "loss": 0.0229,
      "step": 249950
    },
    {
      "epoch": 97.07,
      "learning_rate": 7.0570873786407765e-06,
      "loss": 0.082,
      "step": 249960
    },
    {
      "epoch": 97.08,
      "learning_rate": 7.056569579288026e-06,
      "loss": 0.066,
      "step": 249970
    },
    {
      "epoch": 97.08,
      "learning_rate": 7.056051779935276e-06,
      "loss": 0.0678,
      "step": 249980
    },
    {
      "epoch": 97.08,
      "learning_rate": 7.055533980582525e-06,
      "loss": 0.0016,
      "step": 249990
    },
    {
      "epoch": 97.09,
      "learning_rate": 7.055016181229773e-06,
      "loss": 0.0535,
      "step": 250000
    },
    {
      "epoch": 97.09,
      "learning_rate": 7.054498381877023e-06,
      "loss": 0.1558,
      "step": 250010
    },
    {
      "epoch": 97.1,
      "learning_rate": 7.0539805825242725e-06,
      "loss": 0.0039,
      "step": 250020
    },
    {
      "epoch": 97.1,
      "learning_rate": 7.053462783171522e-06,
      "loss": 0.0698,
      "step": 250030
    },
    {
      "epoch": 97.1,
      "learning_rate": 7.052944983818772e-06,
      "loss": 0.0165,
      "step": 250040
    },
    {
      "epoch": 97.11,
      "learning_rate": 7.05242718446602e-06,
      "loss": 0.0476,
      "step": 250050
    },
    {
      "epoch": 97.11,
      "learning_rate": 7.051909385113269e-06,
      "loss": 0.0131,
      "step": 250060
    },
    {
      "epoch": 97.11,
      "learning_rate": 7.051391585760519e-06,
      "loss": 0.0001,
      "step": 250070
    },
    {
      "epoch": 97.12,
      "learning_rate": 7.050873786407768e-06,
      "loss": 0.0118,
      "step": 250080
    },
    {
      "epoch": 97.12,
      "learning_rate": 7.0503559870550165e-06,
      "loss": 0.0299,
      "step": 250090
    },
    {
      "epoch": 97.13,
      "learning_rate": 7.049838187702266e-06,
      "loss": 0.0807,
      "step": 250100
    },
    {
      "epoch": 97.13,
      "learning_rate": 7.049320388349516e-06,
      "loss": 0.0568,
      "step": 250110
    },
    {
      "epoch": 97.13,
      "learning_rate": 7.0488025889967645e-06,
      "loss": 0.0091,
      "step": 250120
    },
    {
      "epoch": 97.14,
      "learning_rate": 7.048284789644013e-06,
      "loss": 0.0004,
      "step": 250130
    },
    {
      "epoch": 97.14,
      "learning_rate": 7.047766990291263e-06,
      "loss": 0.0008,
      "step": 250140
    },
    {
      "epoch": 97.15,
      "learning_rate": 7.0472491909385125e-06,
      "loss": 0.033,
      "step": 250150
    },
    {
      "epoch": 97.15,
      "learning_rate": 7.046731391585761e-06,
      "loss": 0.0106,
      "step": 250160
    },
    {
      "epoch": 97.15,
      "learning_rate": 7.04621359223301e-06,
      "loss": 0.0193,
      "step": 250170
    },
    {
      "epoch": 97.16,
      "learning_rate": 7.04569579288026e-06,
      "loss": 0.0122,
      "step": 250180
    },
    {
      "epoch": 97.16,
      "learning_rate": 7.045177993527509e-06,
      "loss": 0.1215,
      "step": 250190
    },
    {
      "epoch": 97.17,
      "learning_rate": 7.044660194174758e-06,
      "loss": 0.0606,
      "step": 250200
    },
    {
      "epoch": 97.17,
      "learning_rate": 7.044142394822007e-06,
      "loss": 0.0117,
      "step": 250210
    },
    {
      "epoch": 97.17,
      "learning_rate": 7.043624595469256e-06,
      "loss": 0.1112,
      "step": 250220
    },
    {
      "epoch": 97.18,
      "learning_rate": 7.043106796116505e-06,
      "loss": 0.0239,
      "step": 250230
    },
    {
      "epoch": 97.18,
      "learning_rate": 7.042588996763755e-06,
      "loss": 0.065,
      "step": 250240
    },
    {
      "epoch": 97.18,
      "learning_rate": 7.042071197411004e-06,
      "loss": 0.06,
      "step": 250250
    },
    {
      "epoch": 97.19,
      "learning_rate": 7.041553398058253e-06,
      "loss": 0.1445,
      "step": 250260
    },
    {
      "epoch": 97.19,
      "learning_rate": 7.041035598705502e-06,
      "loss": 0.0003,
      "step": 250270
    },
    {
      "epoch": 97.2,
      "learning_rate": 7.040517799352752e-06,
      "loss": 0.0003,
      "step": 250280
    },
    {
      "epoch": 97.2,
      "learning_rate": 7.04e-06,
      "loss": 0.0394,
      "step": 250290
    },
    {
      "epoch": 97.2,
      "learning_rate": 7.03948220064725e-06,
      "loss": 0.0004,
      "step": 250300
    },
    {
      "epoch": 97.21,
      "learning_rate": 7.038964401294499e-06,
      "loss": 0.0149,
      "step": 250310
    },
    {
      "epoch": 97.21,
      "learning_rate": 7.038446601941748e-06,
      "loss": 0.0311,
      "step": 250320
    },
    {
      "epoch": 97.22,
      "learning_rate": 7.037928802588997e-06,
      "loss": 0.0439,
      "step": 250330
    },
    {
      "epoch": 97.22,
      "learning_rate": 7.037411003236247e-06,
      "loss": 0.0191,
      "step": 250340
    },
    {
      "epoch": 97.22,
      "learning_rate": 7.0368932038834955e-06,
      "loss": 0.0537,
      "step": 250350
    },
    {
      "epoch": 97.23,
      "learning_rate": 7.036375404530745e-06,
      "loss": 0.048,
      "step": 250360
    },
    {
      "epoch": 97.23,
      "learning_rate": 7.035857605177994e-06,
      "loss": 0.0298,
      "step": 250370
    },
    {
      "epoch": 97.23,
      "learning_rate": 7.0353398058252435e-06,
      "loss": 0.0105,
      "step": 250380
    },
    {
      "epoch": 97.24,
      "learning_rate": 7.034822006472492e-06,
      "loss": 0.0232,
      "step": 250390
    },
    {
      "epoch": 97.24,
      "learning_rate": 7.034304207119742e-06,
      "loss": 0.0721,
      "step": 250400
    },
    {
      "epoch": 97.25,
      "learning_rate": 7.033786407766991e-06,
      "loss": 0.0775,
      "step": 250410
    },
    {
      "epoch": 97.25,
      "learning_rate": 7.0332686084142395e-06,
      "loss": 0.0017,
      "step": 250420
    },
    {
      "epoch": 97.25,
      "learning_rate": 7.032750809061489e-06,
      "loss": 0.0354,
      "step": 250430
    },
    {
      "epoch": 97.26,
      "learning_rate": 7.032233009708739e-06,
      "loss": 0.003,
      "step": 250440
    },
    {
      "epoch": 97.26,
      "learning_rate": 7.0317152103559875e-06,
      "loss": 0.0004,
      "step": 250450
    },
    {
      "epoch": 97.27,
      "learning_rate": 7.031197411003236e-06,
      "loss": 0.0249,
      "step": 250460
    },
    {
      "epoch": 97.27,
      "learning_rate": 7.030679611650486e-06,
      "loss": 0.1807,
      "step": 250470
    },
    {
      "epoch": 97.27,
      "learning_rate": 7.0301618122977355e-06,
      "loss": 0.0047,
      "step": 250480
    },
    {
      "epoch": 97.28,
      "learning_rate": 7.029644012944984e-06,
      "loss": 0.1428,
      "step": 250490
    },
    {
      "epoch": 97.28,
      "learning_rate": 7.029126213592233e-06,
      "loss": 0.0692,
      "step": 250500
    },
    {
      "epoch": 97.29,
      "learning_rate": 7.028608414239483e-06,
      "loss": 0.1244,
      "step": 250510
    },
    {
      "epoch": 97.29,
      "learning_rate": 7.028090614886732e-06,
      "loss": 0.0133,
      "step": 250520
    },
    {
      "epoch": 97.29,
      "learning_rate": 7.027572815533981e-06,
      "loss": 0.0806,
      "step": 250530
    },
    {
      "epoch": 97.3,
      "learning_rate": 7.02705501618123e-06,
      "loss": 0.001,
      "step": 250540
    },
    {
      "epoch": 97.3,
      "learning_rate": 7.026537216828479e-06,
      "loss": 0.0094,
      "step": 250550
    },
    {
      "epoch": 97.3,
      "learning_rate": 7.026019417475729e-06,
      "loss": 0.0061,
      "step": 250560
    },
    {
      "epoch": 97.31,
      "learning_rate": 7.025501618122977e-06,
      "loss": 0.0187,
      "step": 250570
    },
    {
      "epoch": 97.31,
      "learning_rate": 7.024983818770227e-06,
      "loss": 0.1196,
      "step": 250580
    },
    {
      "epoch": 97.32,
      "learning_rate": 7.024466019417476e-06,
      "loss": 0.0028,
      "step": 250590
    },
    {
      "epoch": 97.32,
      "learning_rate": 7.023948220064726e-06,
      "loss": 0.0001,
      "step": 250600
    },
    {
      "epoch": 97.32,
      "learning_rate": 7.0234304207119754e-06,
      "loss": 0.0761,
      "step": 250610
    },
    {
      "epoch": 97.33,
      "learning_rate": 7.022912621359223e-06,
      "loss": 0.049,
      "step": 250620
    },
    {
      "epoch": 97.33,
      "learning_rate": 7.022394822006473e-06,
      "loss": 0.0236,
      "step": 250630
    },
    {
      "epoch": 97.34,
      "learning_rate": 7.021877022653723e-06,
      "loss": 0.0902,
      "step": 250640
    },
    {
      "epoch": 97.34,
      "learning_rate": 7.021359223300972e-06,
      "loss": 0.0924,
      "step": 250650
    },
    {
      "epoch": 97.34,
      "learning_rate": 7.02084142394822e-06,
      "loss": 0.0686,
      "step": 250660
    },
    {
      "epoch": 97.35,
      "learning_rate": 7.02032362459547e-06,
      "loss": 0.0575,
      "step": 250670
    },
    {
      "epoch": 97.35,
      "learning_rate": 7.019805825242719e-06,
      "loss": 0.0481,
      "step": 250680
    },
    {
      "epoch": 97.36,
      "learning_rate": 7.019288025889969e-06,
      "loss": 0.0108,
      "step": 250690
    },
    {
      "epoch": 97.36,
      "learning_rate": 7.018770226537217e-06,
      "loss": 0.0203,
      "step": 250700
    },
    {
      "epoch": 97.36,
      "learning_rate": 7.0182524271844665e-06,
      "loss": 0.1255,
      "step": 250710
    },
    {
      "epoch": 97.37,
      "learning_rate": 7.017734627831716e-06,
      "loss": 0.0781,
      "step": 250720
    },
    {
      "epoch": 97.37,
      "learning_rate": 7.017216828478966e-06,
      "loss": 0.016,
      "step": 250730
    },
    {
      "epoch": 97.37,
      "learning_rate": 7.016699029126214e-06,
      "loss": 0.0504,
      "step": 250740
    },
    {
      "epoch": 97.38,
      "learning_rate": 7.016181229773463e-06,
      "loss": 0.03,
      "step": 250750
    },
    {
      "epoch": 97.38,
      "learning_rate": 7.015663430420713e-06,
      "loss": 0.038,
      "step": 250760
    },
    {
      "epoch": 97.39,
      "learning_rate": 7.0151456310679625e-06,
      "loss": 0.1098,
      "step": 250770
    },
    {
      "epoch": 97.39,
      "learning_rate": 7.0146278317152105e-06,
      "loss": 0.1365,
      "step": 250780
    },
    {
      "epoch": 97.39,
      "learning_rate": 7.01411003236246e-06,
      "loss": 0.0931,
      "step": 250790
    },
    {
      "epoch": 97.4,
      "learning_rate": 7.01359223300971e-06,
      "loss": 0.0074,
      "step": 250800
    },
    {
      "epoch": 97.4,
      "learning_rate": 7.013074433656959e-06,
      "loss": 0.0635,
      "step": 250810
    },
    {
      "epoch": 97.41,
      "learning_rate": 7.012556634304207e-06,
      "loss": 0.0081,
      "step": 250820
    },
    {
      "epoch": 97.41,
      "learning_rate": 7.012038834951457e-06,
      "loss": 0.008,
      "step": 250830
    },
    {
      "epoch": 97.41,
      "learning_rate": 7.0115210355987065e-06,
      "loss": 0.0246,
      "step": 250840
    },
    {
      "epoch": 97.42,
      "learning_rate": 7.011003236245956e-06,
      "loss": 0.0563,
      "step": 250850
    },
    {
      "epoch": 97.42,
      "learning_rate": 7.010485436893204e-06,
      "loss": 0.0365,
      "step": 250860
    },
    {
      "epoch": 97.43,
      "learning_rate": 7.009967637540454e-06,
      "loss": 0.0092,
      "step": 250870
    },
    {
      "epoch": 97.43,
      "learning_rate": 7.009449838187703e-06,
      "loss": 0.007,
      "step": 250880
    },
    {
      "epoch": 97.43,
      "learning_rate": 7.008932038834952e-06,
      "loss": 0.0298,
      "step": 250890
    },
    {
      "epoch": 97.44,
      "learning_rate": 7.008414239482201e-06,
      "loss": 0.033,
      "step": 250900
    },
    {
      "epoch": 97.44,
      "learning_rate": 7.00789644012945e-06,
      "loss": 0.001,
      "step": 250910
    },
    {
      "epoch": 97.44,
      "learning_rate": 7.0073786407767e-06,
      "loss": 0.1066,
      "step": 250920
    },
    {
      "epoch": 97.45,
      "learning_rate": 7.006860841423949e-06,
      "loss": 0.0002,
      "step": 250930
    },
    {
      "epoch": 97.45,
      "learning_rate": 7.006343042071198e-06,
      "loss": 0.0495,
      "step": 250940
    },
    {
      "epoch": 97.46,
      "learning_rate": 7.005825242718447e-06,
      "loss": 0.0534,
      "step": 250950
    },
    {
      "epoch": 97.46,
      "learning_rate": 7.005307443365697e-06,
      "loss": 0.0172,
      "step": 250960
    },
    {
      "epoch": 97.46,
      "learning_rate": 7.004789644012946e-06,
      "loss": 0.0648,
      "step": 250970
    },
    {
      "epoch": 97.47,
      "learning_rate": 7.004271844660194e-06,
      "loss": 0.0451,
      "step": 250980
    },
    {
      "epoch": 97.47,
      "learning_rate": 7.003754045307444e-06,
      "loss": 0.0003,
      "step": 250990
    },
    {
      "epoch": 97.48,
      "learning_rate": 7.003236245954694e-06,
      "loss": 0.0095,
      "step": 251000
    },
    {
      "epoch": 97.48,
      "learning_rate": 7.002718446601942e-06,
      "loss": 0.1241,
      "step": 251010
    },
    {
      "epoch": 97.48,
      "learning_rate": 7.002200647249191e-06,
      "loss": 0.077,
      "step": 251020
    },
    {
      "epoch": 97.49,
      "learning_rate": 7.001682847896441e-06,
      "loss": 0.0632,
      "step": 251030
    },
    {
      "epoch": 97.49,
      "learning_rate": 7.0011650485436895e-06,
      "loss": 0.0038,
      "step": 251040
    },
    {
      "epoch": 97.5,
      "learning_rate": 7.000647249190939e-06,
      "loss": 0.0595,
      "step": 251050
    },
    {
      "epoch": 97.5,
      "learning_rate": 7.000129449838188e-06,
      "loss": 0.017,
      "step": 251060
    },
    {
      "epoch": 97.5,
      "learning_rate": 6.9996116504854375e-06,
      "loss": 0.0654,
      "step": 251070
    },
    {
      "epoch": 97.51,
      "learning_rate": 6.999093851132686e-06,
      "loss": 0.2574,
      "step": 251080
    },
    {
      "epoch": 97.51,
      "learning_rate": 6.998576051779936e-06,
      "loss": 0.0002,
      "step": 251090
    },
    {
      "epoch": 97.51,
      "learning_rate": 6.998058252427185e-06,
      "loss": 0.0753,
      "step": 251100
    },
    {
      "epoch": 97.52,
      "learning_rate": 6.997540453074434e-06,
      "loss": 0.0671,
      "step": 251110
    },
    {
      "epoch": 97.52,
      "learning_rate": 6.997022653721683e-06,
      "loss": 0.0595,
      "step": 251120
    },
    {
      "epoch": 97.53,
      "learning_rate": 6.996504854368933e-06,
      "loss": 0.0621,
      "step": 251130
    },
    {
      "epoch": 97.53,
      "learning_rate": 6.9959870550161815e-06,
      "loss": 0.049,
      "step": 251140
    },
    {
      "epoch": 97.53,
      "learning_rate": 6.995469255663431e-06,
      "loss": 0.0285,
      "step": 251150
    },
    {
      "epoch": 97.54,
      "learning_rate": 6.99495145631068e-06,
      "loss": 0.0642,
      "step": 251160
    },
    {
      "epoch": 97.54,
      "learning_rate": 6.9944336569579295e-06,
      "loss": 0.0081,
      "step": 251170
    },
    {
      "epoch": 97.55,
      "learning_rate": 6.993915857605179e-06,
      "loss": 0.0002,
      "step": 251180
    },
    {
      "epoch": 97.55,
      "learning_rate": 6.993398058252428e-06,
      "loss": 0.0044,
      "step": 251190
    },
    {
      "epoch": 97.55,
      "learning_rate": 6.992880258899677e-06,
      "loss": 0.037,
      "step": 251200
    },
    {
      "epoch": 97.56,
      "learning_rate": 6.992362459546926e-06,
      "loss": 0.041,
      "step": 251210
    },
    {
      "epoch": 97.56,
      "learning_rate": 6.991844660194176e-06,
      "loss": 0.0217,
      "step": 251220
    },
    {
      "epoch": 97.57,
      "learning_rate": 6.991326860841424e-06,
      "loss": 0.0007,
      "step": 251230
    },
    {
      "epoch": 97.57,
      "learning_rate": 6.990809061488673e-06,
      "loss": 0.0387,
      "step": 251240
    },
    {
      "epoch": 97.57,
      "learning_rate": 6.990291262135923e-06,
      "loss": 0.059,
      "step": 251250
    },
    {
      "epoch": 97.58,
      "learning_rate": 6.989773462783173e-06,
      "loss": 0.0001,
      "step": 251260
    },
    {
      "epoch": 97.58,
      "learning_rate": 6.9892556634304206e-06,
      "loss": 0.033,
      "step": 251270
    },
    {
      "epoch": 97.58,
      "learning_rate": 6.98873786407767e-06,
      "loss": 0.0031,
      "step": 251280
    },
    {
      "epoch": 97.59,
      "learning_rate": 6.98822006472492e-06,
      "loss": 0.0757,
      "step": 251290
    },
    {
      "epoch": 97.59,
      "learning_rate": 6.9877022653721694e-06,
      "loss": 0.0112,
      "step": 251300
    },
    {
      "epoch": 97.6,
      "learning_rate": 6.987184466019417e-06,
      "loss": 0.0805,
      "step": 251310
    },
    {
      "epoch": 97.6,
      "learning_rate": 6.986666666666667e-06,
      "loss": 0.0536,
      "step": 251320
    },
    {
      "epoch": 97.6,
      "learning_rate": 6.986148867313917e-06,
      "loss": 0.0482,
      "step": 251330
    },
    {
      "epoch": 97.61,
      "learning_rate": 6.985631067961166e-06,
      "loss": 0.1667,
      "step": 251340
    },
    {
      "epoch": 97.61,
      "learning_rate": 6.985113268608414e-06,
      "loss": 0.0453,
      "step": 251350
    },
    {
      "epoch": 97.62,
      "learning_rate": 6.984595469255664e-06,
      "loss": 0.0177,
      "step": 251360
    },
    {
      "epoch": 97.62,
      "learning_rate": 6.984077669902913e-06,
      "loss": 0.0851,
      "step": 251370
    },
    {
      "epoch": 97.62,
      "learning_rate": 6.983559870550163e-06,
      "loss": 0.0152,
      "step": 251380
    },
    {
      "epoch": 97.63,
      "learning_rate": 6.983042071197411e-06,
      "loss": 0.1217,
      "step": 251390
    },
    {
      "epoch": 97.63,
      "learning_rate": 6.9825242718446605e-06,
      "loss": 0.0635,
      "step": 251400
    },
    {
      "epoch": 97.63,
      "learning_rate": 6.98200647249191e-06,
      "loss": 0.1394,
      "step": 251410
    },
    {
      "epoch": 97.64,
      "learning_rate": 6.98148867313916e-06,
      "loss": 0.1144,
      "step": 251420
    },
    {
      "epoch": 97.64,
      "learning_rate": 6.980970873786408e-06,
      "loss": 0.0493,
      "step": 251430
    },
    {
      "epoch": 97.65,
      "learning_rate": 6.980453074433657e-06,
      "loss": 0.1289,
      "step": 251440
    },
    {
      "epoch": 97.65,
      "learning_rate": 6.979935275080907e-06,
      "loss": 0.042,
      "step": 251450
    },
    {
      "epoch": 97.65,
      "learning_rate": 6.9794174757281565e-06,
      "loss": 0.1034,
      "step": 251460
    },
    {
      "epoch": 97.66,
      "learning_rate": 6.9788996763754045e-06,
      "loss": 0.0789,
      "step": 251470
    },
    {
      "epoch": 97.66,
      "learning_rate": 6.978381877022654e-06,
      "loss": 0.0414,
      "step": 251480
    },
    {
      "epoch": 97.67,
      "learning_rate": 6.977864077669904e-06,
      "loss": 0.0195,
      "step": 251490
    },
    {
      "epoch": 97.67,
      "learning_rate": 6.977346278317153e-06,
      "loss": 0.0599,
      "step": 251500
    },
    {
      "epoch": 97.67,
      "learning_rate": 6.976828478964401e-06,
      "loss": 0.0519,
      "step": 251510
    },
    {
      "epoch": 97.68,
      "learning_rate": 6.976310679611651e-06,
      "loss": 0.1099,
      "step": 251520
    },
    {
      "epoch": 97.68,
      "learning_rate": 6.9757928802589005e-06,
      "loss": 0.069,
      "step": 251530
    },
    {
      "epoch": 97.69,
      "learning_rate": 6.97527508090615e-06,
      "loss": 0.0048,
      "step": 251540
    },
    {
      "epoch": 97.69,
      "learning_rate": 6.974757281553398e-06,
      "loss": 0.0852,
      "step": 251550
    },
    {
      "epoch": 97.69,
      "learning_rate": 6.974239482200648e-06,
      "loss": 0.1081,
      "step": 251560
    },
    {
      "epoch": 97.7,
      "learning_rate": 6.973721682847897e-06,
      "loss": 0.0165,
      "step": 251570
    },
    {
      "epoch": 97.7,
      "learning_rate": 6.973203883495147e-06,
      "loss": 0.0706,
      "step": 251580
    },
    {
      "epoch": 97.7,
      "learning_rate": 6.972686084142395e-06,
      "loss": 0.0773,
      "step": 251590
    },
    {
      "epoch": 97.71,
      "learning_rate": 6.972168284789644e-06,
      "loss": 0.1595,
      "step": 251600
    },
    {
      "epoch": 97.71,
      "learning_rate": 6.971650485436894e-06,
      "loss": 0.0441,
      "step": 251610
    },
    {
      "epoch": 97.72,
      "learning_rate": 6.971132686084144e-06,
      "loss": 0.1397,
      "step": 251620
    },
    {
      "epoch": 97.72,
      "learning_rate": 6.970614886731392e-06,
      "loss": 0.004,
      "step": 251630
    },
    {
      "epoch": 97.72,
      "learning_rate": 6.970097087378641e-06,
      "loss": 0.0118,
      "step": 251640
    },
    {
      "epoch": 97.73,
      "learning_rate": 6.969579288025891e-06,
      "loss": 0.1189,
      "step": 251650
    },
    {
      "epoch": 97.73,
      "learning_rate": 6.9690614886731404e-06,
      "loss": 0.0162,
      "step": 251660
    },
    {
      "epoch": 97.74,
      "learning_rate": 6.968543689320388e-06,
      "loss": 0.0036,
      "step": 251670
    },
    {
      "epoch": 97.74,
      "learning_rate": 6.968025889967638e-06,
      "loss": 0.0944,
      "step": 251680
    },
    {
      "epoch": 97.74,
      "learning_rate": 6.967508090614888e-06,
      "loss": 0.0285,
      "step": 251690
    },
    {
      "epoch": 97.75,
      "learning_rate": 6.966990291262136e-06,
      "loss": 0.1372,
      "step": 251700
    },
    {
      "epoch": 97.75,
      "learning_rate": 6.966472491909386e-06,
      "loss": 0.0716,
      "step": 251710
    },
    {
      "epoch": 97.76,
      "learning_rate": 6.965954692556635e-06,
      "loss": 0.0615,
      "step": 251720
    },
    {
      "epoch": 97.76,
      "learning_rate": 6.965436893203884e-06,
      "loss": 0.0783,
      "step": 251730
    },
    {
      "epoch": 97.76,
      "learning_rate": 6.964919093851133e-06,
      "loss": 0.0369,
      "step": 251740
    },
    {
      "epoch": 97.77,
      "learning_rate": 6.964401294498383e-06,
      "loss": 0.1647,
      "step": 251750
    },
    {
      "epoch": 97.77,
      "learning_rate": 6.9638834951456315e-06,
      "loss": 0.0095,
      "step": 251760
    },
    {
      "epoch": 97.77,
      "learning_rate": 6.963365695792881e-06,
      "loss": 0.1303,
      "step": 251770
    },
    {
      "epoch": 97.78,
      "learning_rate": 6.96284789644013e-06,
      "loss": 0.1212,
      "step": 251780
    },
    {
      "epoch": 97.78,
      "learning_rate": 6.9623300970873795e-06,
      "loss": 0.0557,
      "step": 251790
    },
    {
      "epoch": 97.79,
      "learning_rate": 6.961812297734628e-06,
      "loss": 0.0854,
      "step": 251800
    },
    {
      "epoch": 97.79,
      "learning_rate": 6.961294498381878e-06,
      "loss": 0.0233,
      "step": 251810
    },
    {
      "epoch": 97.79,
      "learning_rate": 6.960776699029127e-06,
      "loss": 0.0046,
      "step": 251820
    },
    {
      "epoch": 97.8,
      "learning_rate": 6.960258899676376e-06,
      "loss": 0.0312,
      "step": 251830
    },
    {
      "epoch": 97.8,
      "learning_rate": 6.959741100323625e-06,
      "loss": 0.1149,
      "step": 251840
    },
    {
      "epoch": 97.81,
      "learning_rate": 6.959223300970875e-06,
      "loss": 0.1044,
      "step": 251850
    },
    {
      "epoch": 97.81,
      "learning_rate": 6.9587055016181235e-06,
      "loss": 0.1299,
      "step": 251860
    },
    {
      "epoch": 97.81,
      "learning_rate": 6.958187702265373e-06,
      "loss": 0.032,
      "step": 251870
    },
    {
      "epoch": 97.82,
      "learning_rate": 6.957669902912622e-06,
      "loss": 0.0169,
      "step": 251880
    },
    {
      "epoch": 97.82,
      "learning_rate": 6.957152103559871e-06,
      "loss": 0.0344,
      "step": 251890
    },
    {
      "epoch": 97.83,
      "learning_rate": 6.95663430420712e-06,
      "loss": 0.0566,
      "step": 251900
    },
    {
      "epoch": 97.83,
      "learning_rate": 6.95611650485437e-06,
      "loss": 0.0093,
      "step": 251910
    },
    {
      "epoch": 97.83,
      "learning_rate": 6.955598705501619e-06,
      "loss": 0.0619,
      "step": 251920
    },
    {
      "epoch": 97.84,
      "learning_rate": 6.955080906148867e-06,
      "loss": 0.0225,
      "step": 251930
    },
    {
      "epoch": 97.84,
      "learning_rate": 6.954563106796117e-06,
      "loss": 0.0171,
      "step": 251940
    },
    {
      "epoch": 97.84,
      "learning_rate": 6.954045307443367e-06,
      "loss": 0.0098,
      "step": 251950
    },
    {
      "epoch": 97.85,
      "learning_rate": 6.953527508090615e-06,
      "loss": 0.0507,
      "step": 251960
    },
    {
      "epoch": 97.85,
      "learning_rate": 6.953009708737864e-06,
      "loss": 0.0036,
      "step": 251970
    },
    {
      "epoch": 97.86,
      "learning_rate": 6.952491909385114e-06,
      "loss": 0.0854,
      "step": 251980
    },
    {
      "epoch": 97.86,
      "learning_rate": 6.9519741100323634e-06,
      "loss": 0.0401,
      "step": 251990
    },
    {
      "epoch": 97.86,
      "learning_rate": 6.951456310679612e-06,
      "loss": 0.0539,
      "step": 252000
    },
    {
      "epoch": 97.87,
      "learning_rate": 6.950938511326861e-06,
      "loss": 0.069,
      "step": 252010
    },
    {
      "epoch": 97.87,
      "learning_rate": 6.950420711974111e-06,
      "loss": 0.0916,
      "step": 252020
    },
    {
      "epoch": 97.88,
      "learning_rate": 6.94990291262136e-06,
      "loss": 0.007,
      "step": 252030
    },
    {
      "epoch": 97.88,
      "learning_rate": 6.949385113268608e-06,
      "loss": 0.0043,
      "step": 252040
    },
    {
      "epoch": 97.88,
      "learning_rate": 6.948867313915858e-06,
      "loss": 0.0681,
      "step": 252050
    },
    {
      "epoch": 97.89,
      "learning_rate": 6.948349514563107e-06,
      "loss": 0.0323,
      "step": 252060
    },
    {
      "epoch": 97.89,
      "learning_rate": 6.947831715210357e-06,
      "loss": 0.0259,
      "step": 252070
    },
    {
      "epoch": 97.9,
      "learning_rate": 6.947313915857605e-06,
      "loss": 0.0898,
      "step": 252080
    },
    {
      "epoch": 97.9,
      "learning_rate": 6.9467961165048545e-06,
      "loss": 0.0221,
      "step": 252090
    },
    {
      "epoch": 97.9,
      "learning_rate": 6.946278317152104e-06,
      "loss": 0.0263,
      "step": 252100
    },
    {
      "epoch": 97.91,
      "learning_rate": 6.945760517799354e-06,
      "loss": 0.0403,
      "step": 252110
    },
    {
      "epoch": 97.91,
      "learning_rate": 6.945242718446602e-06,
      "loss": 0.0382,
      "step": 252120
    },
    {
      "epoch": 97.91,
      "learning_rate": 6.944724919093851e-06,
      "loss": 0.1629,
      "step": 252130
    },
    {
      "epoch": 97.92,
      "learning_rate": 6.944207119741101e-06,
      "loss": 0.0226,
      "step": 252140
    },
    {
      "epoch": 97.92,
      "learning_rate": 6.9436893203883505e-06,
      "loss": 0.0038,
      "step": 252150
    },
    {
      "epoch": 97.93,
      "learning_rate": 6.9431715210355985e-06,
      "loss": 0.0407,
      "step": 252160
    },
    {
      "epoch": 97.93,
      "learning_rate": 6.942653721682848e-06,
      "loss": 0.0835,
      "step": 252170
    },
    {
      "epoch": 97.93,
      "learning_rate": 6.942135922330098e-06,
      "loss": 0.0485,
      "step": 252180
    },
    {
      "epoch": 97.94,
      "learning_rate": 6.941618122977347e-06,
      "loss": 0.0392,
      "step": 252190
    },
    {
      "epoch": 97.94,
      "learning_rate": 6.941100323624595e-06,
      "loss": 0.0211,
      "step": 252200
    },
    {
      "epoch": 97.95,
      "learning_rate": 6.940582524271845e-06,
      "loss": 0.0569,
      "step": 252210
    },
    {
      "epoch": 97.95,
      "learning_rate": 6.9400647249190945e-06,
      "loss": 0.0287,
      "step": 252220
    },
    {
      "epoch": 97.95,
      "learning_rate": 6.939546925566344e-06,
      "loss": 0.0504,
      "step": 252230
    },
    {
      "epoch": 97.96,
      "learning_rate": 6.939029126213592e-06,
      "loss": 0.0388,
      "step": 252240
    },
    {
      "epoch": 97.96,
      "learning_rate": 6.938511326860842e-06,
      "loss": 0.0003,
      "step": 252250
    },
    {
      "epoch": 97.97,
      "learning_rate": 6.937993527508091e-06,
      "loss": 0.0617,
      "step": 252260
    },
    {
      "epoch": 97.97,
      "learning_rate": 6.937475728155341e-06,
      "loss": 0.0791,
      "step": 252270
    },
    {
      "epoch": 97.97,
      "learning_rate": 6.9369579288025905e-06,
      "loss": 0.1061,
      "step": 252280
    },
    {
      "epoch": 97.98,
      "learning_rate": 6.936440129449838e-06,
      "loss": 0.1122,
      "step": 252290
    },
    {
      "epoch": 97.98,
      "learning_rate": 6.935922330097088e-06,
      "loss": 0.0449,
      "step": 252300
    },
    {
      "epoch": 97.98,
      "learning_rate": 6.935404530744338e-06,
      "loss": 0.1041,
      "step": 252310
    },
    {
      "epoch": 97.99,
      "learning_rate": 6.934886731391587e-06,
      "loss": 0.054,
      "step": 252320
    },
    {
      "epoch": 97.99,
      "learning_rate": 6.934368932038835e-06,
      "loss": 0.0873,
      "step": 252330
    },
    {
      "epoch": 98.0,
      "learning_rate": 6.933851132686085e-06,
      "loss": 0.1139,
      "step": 252340
    },
    {
      "epoch": 98.0,
      "learning_rate": 6.9333333333333344e-06,
      "loss": 0.0673,
      "step": 252350
    },
    {
      "epoch": 98.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.3636351227760315,
      "eval_runtime": 8.1971,
      "eval_samples_per_second": 443.452,
      "eval_steps_per_second": 55.508,
      "step": 252350
    },
    {
      "epoch": 98.0,
      "learning_rate": 6.932815533980583e-06,
      "loss": 0.0216,
      "step": 252360
    },
    {
      "epoch": 98.01,
      "learning_rate": 6.932297734627832e-06,
      "loss": 0.0353,
      "step": 252370
    },
    {
      "epoch": 98.01,
      "learning_rate": 6.931779935275082e-06,
      "loss": 0.2583,
      "step": 252380
    },
    {
      "epoch": 98.02,
      "learning_rate": 6.931262135922331e-06,
      "loss": 0.0064,
      "step": 252390
    },
    {
      "epoch": 98.02,
      "learning_rate": 6.93074433656958e-06,
      "loss": 0.0661,
      "step": 252400
    },
    {
      "epoch": 98.02,
      "learning_rate": 6.930226537216829e-06,
      "loss": 0.0232,
      "step": 252410
    },
    {
      "epoch": 98.03,
      "learning_rate": 6.929708737864078e-06,
      "loss": 0.092,
      "step": 252420
    },
    {
      "epoch": 98.03,
      "learning_rate": 6.929190938511328e-06,
      "loss": 0.0324,
      "step": 252430
    },
    {
      "epoch": 98.03,
      "learning_rate": 6.928673139158577e-06,
      "loss": 0.042,
      "step": 252440
    },
    {
      "epoch": 98.04,
      "learning_rate": 6.9281553398058255e-06,
      "loss": 0.1064,
      "step": 252450
    },
    {
      "epoch": 98.04,
      "learning_rate": 6.927637540453075e-06,
      "loss": 0.0626,
      "step": 252460
    },
    {
      "epoch": 98.05,
      "learning_rate": 6.927119741100325e-06,
      "loss": 0.0527,
      "step": 252470
    },
    {
      "epoch": 98.05,
      "learning_rate": 6.9266019417475735e-06,
      "loss": 0.1293,
      "step": 252480
    },
    {
      "epoch": 98.05,
      "learning_rate": 6.926084142394822e-06,
      "loss": 0.0301,
      "step": 252490
    },
    {
      "epoch": 98.06,
      "learning_rate": 6.925566343042072e-06,
      "loss": 0.0187,
      "step": 252500
    },
    {
      "epoch": 98.06,
      "learning_rate": 6.925048543689321e-06,
      "loss": 0.0149,
      "step": 252510
    },
    {
      "epoch": 98.07,
      "learning_rate": 6.92453074433657e-06,
      "loss": 0.0337,
      "step": 252520
    },
    {
      "epoch": 98.07,
      "learning_rate": 6.924012944983819e-06,
      "loss": 0.0069,
      "step": 252530
    },
    {
      "epoch": 98.07,
      "learning_rate": 6.923495145631069e-06,
      "loss": 0.0031,
      "step": 252540
    },
    {
      "epoch": 98.08,
      "learning_rate": 6.9229773462783175e-06,
      "loss": 0.1159,
      "step": 252550
    },
    {
      "epoch": 98.08,
      "learning_rate": 6.922459546925567e-06,
      "loss": 0.0222,
      "step": 252560
    },
    {
      "epoch": 98.09,
      "learning_rate": 6.921941747572816e-06,
      "loss": 0.0342,
      "step": 252570
    },
    {
      "epoch": 98.09,
      "learning_rate": 6.9214239482200655e-06,
      "loss": 0.016,
      "step": 252580
    },
    {
      "epoch": 98.09,
      "learning_rate": 6.920906148867314e-06,
      "loss": 0.1221,
      "step": 252590
    },
    {
      "epoch": 98.1,
      "learning_rate": 6.920388349514564e-06,
      "loss": 0.0344,
      "step": 252600
    },
    {
      "epoch": 98.1,
      "learning_rate": 6.919870550161813e-06,
      "loss": 0.0167,
      "step": 252610
    },
    {
      "epoch": 98.1,
      "learning_rate": 6.919352750809062e-06,
      "loss": 0.0573,
      "step": 252620
    },
    {
      "epoch": 98.11,
      "learning_rate": 6.918834951456311e-06,
      "loss": 0.0975,
      "step": 252630
    },
    {
      "epoch": 98.11,
      "learning_rate": 6.918317152103561e-06,
      "loss": 0.0049,
      "step": 252640
    },
    {
      "epoch": 98.12,
      "learning_rate": 6.917799352750809e-06,
      "loss": 0.059,
      "step": 252650
    },
    {
      "epoch": 98.12,
      "learning_rate": 6.917281553398059e-06,
      "loss": 0.0681,
      "step": 252660
    },
    {
      "epoch": 98.12,
      "learning_rate": 6.916763754045308e-06,
      "loss": 0.1316,
      "step": 252670
    },
    {
      "epoch": 98.13,
      "learning_rate": 6.9162459546925574e-06,
      "loss": 0.1364,
      "step": 252680
    },
    {
      "epoch": 98.13,
      "learning_rate": 6.915728155339806e-06,
      "loss": 0.0026,
      "step": 252690
    },
    {
      "epoch": 98.14,
      "learning_rate": 6.915210355987055e-06,
      "loss": 0.0425,
      "step": 252700
    },
    {
      "epoch": 98.14,
      "learning_rate": 6.914692556634305e-06,
      "loss": 0.1669,
      "step": 252710
    },
    {
      "epoch": 98.14,
      "learning_rate": 6.914174757281554e-06,
      "loss": 0.104,
      "step": 252720
    },
    {
      "epoch": 98.15,
      "learning_rate": 6.913656957928803e-06,
      "loss": 0.0103,
      "step": 252730
    },
    {
      "epoch": 98.15,
      "learning_rate": 6.913139158576052e-06,
      "loss": 0.0185,
      "step": 252740
    },
    {
      "epoch": 98.16,
      "learning_rate": 6.912621359223301e-06,
      "loss": 0.1722,
      "step": 252750
    },
    {
      "epoch": 98.16,
      "learning_rate": 6.912103559870551e-06,
      "loss": 0.0395,
      "step": 252760
    },
    {
      "epoch": 98.16,
      "learning_rate": 6.9115857605178e-06,
      "loss": 0.0053,
      "step": 252770
    },
    {
      "epoch": 98.17,
      "learning_rate": 6.9110679611650485e-06,
      "loss": 0.0776,
      "step": 252780
    },
    {
      "epoch": 98.17,
      "learning_rate": 6.910550161812298e-06,
      "loss": 0.0593,
      "step": 252790
    },
    {
      "epoch": 98.17,
      "learning_rate": 6.910032362459548e-06,
      "loss": 0.1146,
      "step": 252800
    },
    {
      "epoch": 98.18,
      "learning_rate": 6.9095145631067965e-06,
      "loss": 0.1055,
      "step": 252810
    },
    {
      "epoch": 98.18,
      "learning_rate": 6.908996763754045e-06,
      "loss": 0.0226,
      "step": 252820
    },
    {
      "epoch": 98.19,
      "learning_rate": 6.908478964401295e-06,
      "loss": 0.1461,
      "step": 252830
    },
    {
      "epoch": 98.19,
      "learning_rate": 6.9079611650485445e-06,
      "loss": 0.0618,
      "step": 252840
    },
    {
      "epoch": 98.19,
      "learning_rate": 6.907443365695794e-06,
      "loss": 0.0069,
      "step": 252850
    },
    {
      "epoch": 98.2,
      "learning_rate": 6.906925566343042e-06,
      "loss": 0.0136,
      "step": 252860
    },
    {
      "epoch": 98.2,
      "learning_rate": 6.906407766990292e-06,
      "loss": 0.0007,
      "step": 252870
    },
    {
      "epoch": 98.21,
      "learning_rate": 6.905889967637541e-06,
      "loss": 0.0054,
      "step": 252880
    },
    {
      "epoch": 98.21,
      "learning_rate": 6.905372168284791e-06,
      "loss": 0.0395,
      "step": 252890
    },
    {
      "epoch": 98.21,
      "learning_rate": 6.904854368932039e-06,
      "loss": 0.0943,
      "step": 252900
    },
    {
      "epoch": 98.22,
      "learning_rate": 6.9043365695792885e-06,
      "loss": 0.1115,
      "step": 252910
    },
    {
      "epoch": 98.22,
      "learning_rate": 6.903818770226538e-06,
      "loss": 0.0067,
      "step": 252920
    },
    {
      "epoch": 98.23,
      "learning_rate": 6.903300970873788e-06,
      "loss": 0.032,
      "step": 252930
    },
    {
      "epoch": 98.23,
      "learning_rate": 6.902783171521036e-06,
      "loss": 0.0187,
      "step": 252940
    },
    {
      "epoch": 98.23,
      "learning_rate": 6.902265372168285e-06,
      "loss": 0.1301,
      "step": 252950
    },
    {
      "epoch": 98.24,
      "learning_rate": 6.901747572815535e-06,
      "loss": 0.0596,
      "step": 252960
    },
    {
      "epoch": 98.24,
      "learning_rate": 6.9012297734627845e-06,
      "loss": 0.1034,
      "step": 252970
    },
    {
      "epoch": 98.24,
      "learning_rate": 6.900711974110032e-06,
      "loss": 0.0198,
      "step": 252980
    },
    {
      "epoch": 98.25,
      "learning_rate": 6.900194174757282e-06,
      "loss": 0.0238,
      "step": 252990
    },
    {
      "epoch": 98.25,
      "learning_rate": 6.899676375404532e-06,
      "loss": 0.0093,
      "step": 253000
    },
    {
      "epoch": 98.26,
      "learning_rate": 6.899158576051781e-06,
      "loss": 0.0553,
      "step": 253010
    },
    {
      "epoch": 98.26,
      "learning_rate": 6.898640776699029e-06,
      "loss": 0.0421,
      "step": 253020
    },
    {
      "epoch": 98.26,
      "learning_rate": 6.898122977346279e-06,
      "loss": 0.0963,
      "step": 253030
    },
    {
      "epoch": 98.27,
      "learning_rate": 6.8976051779935284e-06,
      "loss": 0.0003,
      "step": 253040
    },
    {
      "epoch": 98.27,
      "learning_rate": 6.897087378640778e-06,
      "loss": 0.001,
      "step": 253050
    },
    {
      "epoch": 98.28,
      "learning_rate": 6.896569579288026e-06,
      "loss": 0.0764,
      "step": 253060
    },
    {
      "epoch": 98.28,
      "learning_rate": 6.896051779935276e-06,
      "loss": 0.2061,
      "step": 253070
    },
    {
      "epoch": 98.28,
      "learning_rate": 6.895533980582525e-06,
      "loss": 0.079,
      "step": 253080
    },
    {
      "epoch": 98.29,
      "learning_rate": 6.895016181229775e-06,
      "loss": 0.0591,
      "step": 253090
    },
    {
      "epoch": 98.29,
      "learning_rate": 6.894498381877023e-06,
      "loss": 0.035,
      "step": 253100
    },
    {
      "epoch": 98.3,
      "learning_rate": 6.893980582524272e-06,
      "loss": 0.0336,
      "step": 253110
    },
    {
      "epoch": 98.3,
      "learning_rate": 6.893462783171522e-06,
      "loss": 0.072,
      "step": 253120
    },
    {
      "epoch": 98.3,
      "learning_rate": 6.892944983818772e-06,
      "loss": 0.1616,
      "step": 253130
    },
    {
      "epoch": 98.31,
      "learning_rate": 6.8924271844660195e-06,
      "loss": 0.0587,
      "step": 253140
    },
    {
      "epoch": 98.31,
      "learning_rate": 6.891909385113269e-06,
      "loss": 0.0543,
      "step": 253150
    },
    {
      "epoch": 98.31,
      "learning_rate": 6.891391585760519e-06,
      "loss": 0.0264,
      "step": 253160
    },
    {
      "epoch": 98.32,
      "learning_rate": 6.8908737864077675e-06,
      "loss": 0.015,
      "step": 253170
    },
    {
      "epoch": 98.32,
      "learning_rate": 6.890355987055016e-06,
      "loss": 0.1798,
      "step": 253180
    },
    {
      "epoch": 98.33,
      "learning_rate": 6.889838187702266e-06,
      "loss": 0.018,
      "step": 253190
    },
    {
      "epoch": 98.33,
      "learning_rate": 6.8893203883495155e-06,
      "loss": 0.0507,
      "step": 253200
    },
    {
      "epoch": 98.33,
      "learning_rate": 6.888802588996764e-06,
      "loss": 0.0054,
      "step": 253210
    },
    {
      "epoch": 98.34,
      "learning_rate": 6.888284789644013e-06,
      "loss": 0.0509,
      "step": 253220
    },
    {
      "epoch": 98.34,
      "learning_rate": 6.887766990291263e-06,
      "loss": 0.1517,
      "step": 253230
    },
    {
      "epoch": 98.35,
      "learning_rate": 6.887249190938512e-06,
      "loss": 0.2042,
      "step": 253240
    },
    {
      "epoch": 98.35,
      "learning_rate": 6.886731391585761e-06,
      "loss": 0.0294,
      "step": 253250
    },
    {
      "epoch": 98.35,
      "learning_rate": 6.88621359223301e-06,
      "loss": 0.2049,
      "step": 253260
    },
    {
      "epoch": 98.36,
      "learning_rate": 6.8856957928802595e-06,
      "loss": 0.0392,
      "step": 253270
    },
    {
      "epoch": 98.36,
      "learning_rate": 6.885177993527509e-06,
      "loss": 0.0026,
      "step": 253280
    },
    {
      "epoch": 98.37,
      "learning_rate": 6.884660194174758e-06,
      "loss": 0.0681,
      "step": 253290
    },
    {
      "epoch": 98.37,
      "learning_rate": 6.884142394822007e-06,
      "loss": 0.1597,
      "step": 253300
    },
    {
      "epoch": 98.37,
      "learning_rate": 6.883624595469256e-06,
      "loss": 0.0728,
      "step": 253310
    },
    {
      "epoch": 98.38,
      "learning_rate": 6.883106796116505e-06,
      "loss": 0.0452,
      "step": 253320
    },
    {
      "epoch": 98.38,
      "learning_rate": 6.882588996763755e-06,
      "loss": 0.0025,
      "step": 253330
    },
    {
      "epoch": 98.38,
      "learning_rate": 6.882071197411003e-06,
      "loss": 0.0163,
      "step": 253340
    },
    {
      "epoch": 98.39,
      "learning_rate": 6.881553398058253e-06,
      "loss": 0.0012,
      "step": 253350
    },
    {
      "epoch": 98.39,
      "learning_rate": 6.881035598705502e-06,
      "loss": 0.0263,
      "step": 253360
    },
    {
      "epoch": 98.4,
      "learning_rate": 6.8805177993527514e-06,
      "loss": 0.0202,
      "step": 253370
    },
    {
      "epoch": 98.4,
      "learning_rate": 6.88e-06,
      "loss": 0.0666,
      "step": 253380
    },
    {
      "epoch": 98.4,
      "learning_rate": 6.87948220064725e-06,
      "loss": 0.061,
      "step": 253390
    },
    {
      "epoch": 98.41,
      "learning_rate": 6.878964401294499e-06,
      "loss": 0.0227,
      "step": 253400
    },
    {
      "epoch": 98.41,
      "learning_rate": 6.878446601941748e-06,
      "loss": 0.0004,
      "step": 253410
    },
    {
      "epoch": 98.42,
      "learning_rate": 6.877928802588998e-06,
      "loss": 0.0354,
      "step": 253420
    },
    {
      "epoch": 98.42,
      "learning_rate": 6.877411003236247e-06,
      "loss": 0.0602,
      "step": 253430
    },
    {
      "epoch": 98.42,
      "learning_rate": 6.876893203883495e-06,
      "loss": 0.0545,
      "step": 253440
    },
    {
      "epoch": 98.43,
      "learning_rate": 6.876375404530745e-06,
      "loss": 0.0505,
      "step": 253450
    },
    {
      "epoch": 98.43,
      "learning_rate": 6.875857605177995e-06,
      "loss": 0.0194,
      "step": 253460
    },
    {
      "epoch": 98.43,
      "learning_rate": 6.875339805825243e-06,
      "loss": 0.149,
      "step": 253470
    },
    {
      "epoch": 98.44,
      "learning_rate": 6.874822006472492e-06,
      "loss": 0.0376,
      "step": 253480
    },
    {
      "epoch": 98.44,
      "learning_rate": 6.874304207119742e-06,
      "loss": 0.0354,
      "step": 253490
    },
    {
      "epoch": 98.45,
      "learning_rate": 6.873786407766991e-06,
      "loss": 0.0677,
      "step": 253500
    },
    {
      "epoch": 98.45,
      "learning_rate": 6.873268608414239e-06,
      "loss": 0.0812,
      "step": 253510
    },
    {
      "epoch": 98.45,
      "learning_rate": 6.872750809061489e-06,
      "loss": 0.0219,
      "step": 253520
    },
    {
      "epoch": 98.46,
      "learning_rate": 6.8722330097087385e-06,
      "loss": 0.0309,
      "step": 253530
    },
    {
      "epoch": 98.46,
      "learning_rate": 6.871715210355988e-06,
      "loss": 0.0037,
      "step": 253540
    },
    {
      "epoch": 98.47,
      "learning_rate": 6.871197411003236e-06,
      "loss": 0.0381,
      "step": 253550
    },
    {
      "epoch": 98.47,
      "learning_rate": 6.870679611650486e-06,
      "loss": 0.1144,
      "step": 253560
    },
    {
      "epoch": 98.47,
      "learning_rate": 6.870161812297735e-06,
      "loss": 0.0352,
      "step": 253570
    },
    {
      "epoch": 98.48,
      "learning_rate": 6.869644012944985e-06,
      "loss": 0.0003,
      "step": 253580
    },
    {
      "epoch": 98.48,
      "learning_rate": 6.869126213592233e-06,
      "loss": 0.0419,
      "step": 253590
    },
    {
      "epoch": 98.49,
      "learning_rate": 6.8686084142394825e-06,
      "loss": 0.0672,
      "step": 253600
    },
    {
      "epoch": 98.49,
      "learning_rate": 6.868090614886732e-06,
      "loss": 0.0335,
      "step": 253610
    },
    {
      "epoch": 98.49,
      "learning_rate": 6.867572815533982e-06,
      "loss": 0.0562,
      "step": 253620
    },
    {
      "epoch": 98.5,
      "learning_rate": 6.86705501618123e-06,
      "loss": 0.1066,
      "step": 253630
    },
    {
      "epoch": 98.5,
      "learning_rate": 6.866537216828479e-06,
      "loss": 0.124,
      "step": 253640
    },
    {
      "epoch": 98.5,
      "learning_rate": 6.866019417475729e-06,
      "loss": 0.0123,
      "step": 253650
    },
    {
      "epoch": 98.51,
      "learning_rate": 6.8655016181229785e-06,
      "loss": 0.0017,
      "step": 253660
    },
    {
      "epoch": 98.51,
      "learning_rate": 6.864983818770226e-06,
      "loss": 0.0341,
      "step": 253670
    },
    {
      "epoch": 98.52,
      "learning_rate": 6.864466019417476e-06,
      "loss": 0.0001,
      "step": 253680
    },
    {
      "epoch": 98.52,
      "learning_rate": 6.863948220064726e-06,
      "loss": 0.1197,
      "step": 253690
    },
    {
      "epoch": 98.52,
      "learning_rate": 6.863430420711975e-06,
      "loss": 0.0942,
      "step": 253700
    },
    {
      "epoch": 98.53,
      "learning_rate": 6.862912621359223e-06,
      "loss": 0.0001,
      "step": 253710
    },
    {
      "epoch": 98.53,
      "learning_rate": 6.862394822006473e-06,
      "loss": 0.0921,
      "step": 253720
    },
    {
      "epoch": 98.54,
      "learning_rate": 6.8618770226537224e-06,
      "loss": 0.0053,
      "step": 253730
    },
    {
      "epoch": 98.54,
      "learning_rate": 6.861359223300972e-06,
      "loss": 0.0088,
      "step": 253740
    },
    {
      "epoch": 98.54,
      "learning_rate": 6.86084142394822e-06,
      "loss": 0.0412,
      "step": 253750
    },
    {
      "epoch": 98.55,
      "learning_rate": 6.86032362459547e-06,
      "loss": 0.006,
      "step": 253760
    },
    {
      "epoch": 98.55,
      "learning_rate": 6.859805825242719e-06,
      "loss": 0.0816,
      "step": 253770
    },
    {
      "epoch": 98.56,
      "learning_rate": 6.859288025889969e-06,
      "loss": 0.0119,
      "step": 253780
    },
    {
      "epoch": 98.56,
      "learning_rate": 6.858770226537217e-06,
      "loss": 0.0229,
      "step": 253790
    },
    {
      "epoch": 98.56,
      "learning_rate": 6.858252427184466e-06,
      "loss": 0.0131,
      "step": 253800
    },
    {
      "epoch": 98.57,
      "learning_rate": 6.857734627831716e-06,
      "loss": 0.0907,
      "step": 253810
    },
    {
      "epoch": 98.57,
      "learning_rate": 6.857216828478966e-06,
      "loss": 0.0487,
      "step": 253820
    },
    {
      "epoch": 98.57,
      "learning_rate": 6.8566990291262135e-06,
      "loss": 0.0722,
      "step": 253830
    },
    {
      "epoch": 98.58,
      "learning_rate": 6.856181229773463e-06,
      "loss": 0.0148,
      "step": 253840
    },
    {
      "epoch": 98.58,
      "learning_rate": 6.855663430420713e-06,
      "loss": 0.0211,
      "step": 253850
    },
    {
      "epoch": 98.59,
      "learning_rate": 6.855145631067962e-06,
      "loss": 0.0144,
      "step": 253860
    },
    {
      "epoch": 98.59,
      "learning_rate": 6.85462783171521e-06,
      "loss": 0.0537,
      "step": 253870
    },
    {
      "epoch": 98.59,
      "learning_rate": 6.85411003236246e-06,
      "loss": 0.0018,
      "step": 253880
    },
    {
      "epoch": 98.6,
      "learning_rate": 6.8535922330097095e-06,
      "loss": 0.0536,
      "step": 253890
    },
    {
      "epoch": 98.6,
      "learning_rate": 6.853074433656959e-06,
      "loss": 0.0591,
      "step": 253900
    },
    {
      "epoch": 98.61,
      "learning_rate": 6.852556634304207e-06,
      "loss": 0.0305,
      "step": 253910
    },
    {
      "epoch": 98.61,
      "learning_rate": 6.852038834951457e-06,
      "loss": 0.0174,
      "step": 253920
    },
    {
      "epoch": 98.61,
      "learning_rate": 6.851521035598706e-06,
      "loss": 0.0554,
      "step": 253930
    },
    {
      "epoch": 98.62,
      "learning_rate": 6.851003236245956e-06,
      "loss": 0.044,
      "step": 253940
    },
    {
      "epoch": 98.62,
      "learning_rate": 6.850485436893205e-06,
      "loss": 0.0807,
      "step": 253950
    },
    {
      "epoch": 98.63,
      "learning_rate": 6.8499676375404535e-06,
      "loss": 0.0206,
      "step": 253960
    },
    {
      "epoch": 98.63,
      "learning_rate": 6.849449838187703e-06,
      "loss": 0.0345,
      "step": 253970
    },
    {
      "epoch": 98.63,
      "learning_rate": 6.848932038834952e-06,
      "loss": 0.0019,
      "step": 253980
    },
    {
      "epoch": 98.64,
      "learning_rate": 6.8484142394822015e-06,
      "loss": 0.0032,
      "step": 253990
    },
    {
      "epoch": 98.64,
      "learning_rate": 6.84789644012945e-06,
      "loss": 0.0617,
      "step": 254000
    },
    {
      "epoch": 98.64,
      "learning_rate": 6.8473786407767e-06,
      "loss": 0.0695,
      "step": 254010
    },
    {
      "epoch": 98.65,
      "learning_rate": 6.846860841423949e-06,
      "loss": 0.0784,
      "step": 254020
    },
    {
      "epoch": 98.65,
      "learning_rate": 6.846343042071198e-06,
      "loss": 0.072,
      "step": 254030
    },
    {
      "epoch": 98.66,
      "learning_rate": 6.845825242718447e-06,
      "loss": 0.023,
      "step": 254040
    },
    {
      "epoch": 98.66,
      "learning_rate": 6.845307443365697e-06,
      "loss": 0.087,
      "step": 254050
    },
    {
      "epoch": 98.66,
      "learning_rate": 6.8447896440129454e-06,
      "loss": 0.0322,
      "step": 254060
    },
    {
      "epoch": 98.67,
      "learning_rate": 6.844271844660195e-06,
      "loss": 0.0345,
      "step": 254070
    },
    {
      "epoch": 98.67,
      "learning_rate": 6.843754045307444e-06,
      "loss": 0.0844,
      "step": 254080
    },
    {
      "epoch": 98.68,
      "learning_rate": 6.8432362459546934e-06,
      "loss": 0.2505,
      "step": 254090
    },
    {
      "epoch": 98.68,
      "learning_rate": 6.842718446601942e-06,
      "loss": 0.0206,
      "step": 254100
    },
    {
      "epoch": 98.68,
      "learning_rate": 6.842200647249192e-06,
      "loss": 0.1007,
      "step": 254110
    },
    {
      "epoch": 98.69,
      "learning_rate": 6.841682847896441e-06,
      "loss": 0.0018,
      "step": 254120
    },
    {
      "epoch": 98.69,
      "learning_rate": 6.84116504854369e-06,
      "loss": 0.0575,
      "step": 254130
    },
    {
      "epoch": 98.7,
      "learning_rate": 6.840647249190939e-06,
      "loss": 0.0236,
      "step": 254140
    },
    {
      "epoch": 98.7,
      "learning_rate": 6.840129449838189e-06,
      "loss": 0.1194,
      "step": 254150
    },
    {
      "epoch": 98.7,
      "learning_rate": 6.839611650485437e-06,
      "loss": 0.1679,
      "step": 254160
    },
    {
      "epoch": 98.71,
      "learning_rate": 6.839093851132686e-06,
      "loss": 0.0507,
      "step": 254170
    },
    {
      "epoch": 98.71,
      "learning_rate": 6.838576051779936e-06,
      "loss": 0.0176,
      "step": 254180
    },
    {
      "epoch": 98.71,
      "learning_rate": 6.838058252427185e-06,
      "loss": 0.0009,
      "step": 254190
    },
    {
      "epoch": 98.72,
      "learning_rate": 6.837540453074434e-06,
      "loss": 0.0306,
      "step": 254200
    },
    {
      "epoch": 98.72,
      "learning_rate": 6.837022653721683e-06,
      "loss": 0.0346,
      "step": 254210
    },
    {
      "epoch": 98.73,
      "learning_rate": 6.8365048543689325e-06,
      "loss": 0.0002,
      "step": 254220
    },
    {
      "epoch": 98.73,
      "learning_rate": 6.835987055016182e-06,
      "loss": 0.1106,
      "step": 254230
    },
    {
      "epoch": 98.73,
      "learning_rate": 6.835469255663431e-06,
      "loss": 0.0108,
      "step": 254240
    },
    {
      "epoch": 98.74,
      "learning_rate": 6.83495145631068e-06,
      "loss": 0.0897,
      "step": 254250
    },
    {
      "epoch": 98.74,
      "learning_rate": 6.834433656957929e-06,
      "loss": 0.0035,
      "step": 254260
    },
    {
      "epoch": 98.75,
      "learning_rate": 6.833915857605179e-06,
      "loss": 0.0456,
      "step": 254270
    },
    {
      "epoch": 98.75,
      "learning_rate": 6.833398058252428e-06,
      "loss": 0.0176,
      "step": 254280
    },
    {
      "epoch": 98.75,
      "learning_rate": 6.8328802588996765e-06,
      "loss": 0.0188,
      "step": 254290
    },
    {
      "epoch": 98.76,
      "learning_rate": 6.832362459546926e-06,
      "loss": 0.0073,
      "step": 254300
    },
    {
      "epoch": 98.76,
      "learning_rate": 6.831844660194176e-06,
      "loss": 0.0375,
      "step": 254310
    },
    {
      "epoch": 98.77,
      "learning_rate": 6.831326860841424e-06,
      "loss": 0.0691,
      "step": 254320
    },
    {
      "epoch": 98.77,
      "learning_rate": 6.830809061488673e-06,
      "loss": 0.0881,
      "step": 254330
    },
    {
      "epoch": 98.77,
      "learning_rate": 6.830291262135923e-06,
      "loss": 0.0114,
      "step": 254340
    },
    {
      "epoch": 98.78,
      "learning_rate": 6.8297734627831725e-06,
      "loss": 0.0555,
      "step": 254350
    },
    {
      "epoch": 98.78,
      "learning_rate": 6.82925566343042e-06,
      "loss": 0.1366,
      "step": 254360
    },
    {
      "epoch": 98.78,
      "learning_rate": 6.82873786407767e-06,
      "loss": 0.0593,
      "step": 254370
    },
    {
      "epoch": 98.79,
      "learning_rate": 6.82822006472492e-06,
      "loss": 0.0777,
      "step": 254380
    },
    {
      "epoch": 98.79,
      "learning_rate": 6.827702265372169e-06,
      "loss": 0.0007,
      "step": 254390
    },
    {
      "epoch": 98.8,
      "learning_rate": 6.827184466019417e-06,
      "loss": 0.1249,
      "step": 254400
    },
    {
      "epoch": 98.8,
      "learning_rate": 6.826666666666667e-06,
      "loss": 0.0708,
      "step": 254410
    },
    {
      "epoch": 98.8,
      "learning_rate": 6.8261488673139164e-06,
      "loss": 0.0477,
      "step": 254420
    },
    {
      "epoch": 98.81,
      "learning_rate": 6.825631067961166e-06,
      "loss": 0.0974,
      "step": 254430
    },
    {
      "epoch": 98.81,
      "learning_rate": 6.825113268608414e-06,
      "loss": 0.0316,
      "step": 254440
    },
    {
      "epoch": 98.82,
      "learning_rate": 6.824595469255664e-06,
      "loss": 0.1103,
      "step": 254450
    },
    {
      "epoch": 98.82,
      "learning_rate": 6.824077669902913e-06,
      "loss": 0.0439,
      "step": 254460
    },
    {
      "epoch": 98.82,
      "learning_rate": 6.823559870550163e-06,
      "loss": 0.1616,
      "step": 254470
    },
    {
      "epoch": 98.83,
      "learning_rate": 6.823042071197411e-06,
      "loss": 0.0203,
      "step": 254480
    },
    {
      "epoch": 98.83,
      "learning_rate": 6.82252427184466e-06,
      "loss": 0.0641,
      "step": 254490
    },
    {
      "epoch": 98.83,
      "learning_rate": 6.82200647249191e-06,
      "loss": 0.0978,
      "step": 254500
    },
    {
      "epoch": 98.84,
      "learning_rate": 6.82148867313916e-06,
      "loss": 0.0452,
      "step": 254510
    },
    {
      "epoch": 98.84,
      "learning_rate": 6.820970873786409e-06,
      "loss": 0.0573,
      "step": 254520
    },
    {
      "epoch": 98.85,
      "learning_rate": 6.820453074433657e-06,
      "loss": 0.0465,
      "step": 254530
    },
    {
      "epoch": 98.85,
      "learning_rate": 6.819935275080907e-06,
      "loss": 0.1753,
      "step": 254540
    },
    {
      "epoch": 98.85,
      "learning_rate": 6.819417475728156e-06,
      "loss": 0.0354,
      "step": 254550
    },
    {
      "epoch": 98.86,
      "learning_rate": 6.818899676375406e-06,
      "loss": 0.0399,
      "step": 254560
    },
    {
      "epoch": 98.86,
      "learning_rate": 6.818381877022654e-06,
      "loss": 0.1956,
      "step": 254570
    },
    {
      "epoch": 98.87,
      "learning_rate": 6.8178640776699035e-06,
      "loss": 0.0479,
      "step": 254580
    },
    {
      "epoch": 98.87,
      "learning_rate": 6.817346278317153e-06,
      "loss": 0.1183,
      "step": 254590
    },
    {
      "epoch": 98.87,
      "learning_rate": 6.816828478964403e-06,
      "loss": 0.0677,
      "step": 254600
    },
    {
      "epoch": 98.88,
      "learning_rate": 6.816310679611651e-06,
      "loss": 0.0062,
      "step": 254610
    },
    {
      "epoch": 98.88,
      "learning_rate": 6.8157928802589e-06,
      "loss": 0.0783,
      "step": 254620
    },
    {
      "epoch": 98.89,
      "learning_rate": 6.81527508090615e-06,
      "loss": 0.1224,
      "step": 254630
    },
    {
      "epoch": 98.89,
      "learning_rate": 6.814757281553399e-06,
      "loss": 0.0381,
      "step": 254640
    },
    {
      "epoch": 98.89,
      "learning_rate": 6.8142394822006475e-06,
      "loss": 0.102,
      "step": 254650
    },
    {
      "epoch": 98.9,
      "learning_rate": 6.813721682847897e-06,
      "loss": 0.0047,
      "step": 254660
    },
    {
      "epoch": 98.9,
      "learning_rate": 6.813203883495147e-06,
      "loss": 0.0997,
      "step": 254670
    },
    {
      "epoch": 98.9,
      "learning_rate": 6.8126860841423955e-06,
      "loss": 0.0382,
      "step": 254680
    },
    {
      "epoch": 98.91,
      "learning_rate": 6.812168284789644e-06,
      "loss": 0.0012,
      "step": 254690
    },
    {
      "epoch": 98.91,
      "learning_rate": 6.811650485436894e-06,
      "loss": 0.0901,
      "step": 254700
    },
    {
      "epoch": 98.92,
      "learning_rate": 6.8111326860841435e-06,
      "loss": 0.0698,
      "step": 254710
    },
    {
      "epoch": 98.92,
      "learning_rate": 6.810614886731392e-06,
      "loss": 0.0578,
      "step": 254720
    },
    {
      "epoch": 98.92,
      "learning_rate": 6.810097087378641e-06,
      "loss": 0.0622,
      "step": 254730
    },
    {
      "epoch": 98.93,
      "learning_rate": 6.809579288025891e-06,
      "loss": 0.1823,
      "step": 254740
    },
    {
      "epoch": 98.93,
      "learning_rate": 6.80906148867314e-06,
      "loss": 0.064,
      "step": 254750
    },
    {
      "epoch": 98.94,
      "learning_rate": 6.808543689320389e-06,
      "loss": 0.0647,
      "step": 254760
    },
    {
      "epoch": 98.94,
      "learning_rate": 6.808025889967638e-06,
      "loss": 0.0456,
      "step": 254770
    },
    {
      "epoch": 98.94,
      "learning_rate": 6.8075080906148874e-06,
      "loss": 0.0335,
      "step": 254780
    },
    {
      "epoch": 98.95,
      "learning_rate": 6.806990291262136e-06,
      "loss": 0.1062,
      "step": 254790
    },
    {
      "epoch": 98.95,
      "learning_rate": 6.806472491909386e-06,
      "loss": 0.0281,
      "step": 254800
    },
    {
      "epoch": 98.96,
      "learning_rate": 6.805954692556635e-06,
      "loss": 0.0194,
      "step": 254810
    },
    {
      "epoch": 98.96,
      "learning_rate": 6.805436893203884e-06,
      "loss": 0.0797,
      "step": 254820
    },
    {
      "epoch": 98.96,
      "learning_rate": 6.804919093851133e-06,
      "loss": 0.0433,
      "step": 254830
    },
    {
      "epoch": 98.97,
      "learning_rate": 6.804401294498383e-06,
      "loss": 0.0509,
      "step": 254840
    },
    {
      "epoch": 98.97,
      "learning_rate": 6.803883495145631e-06,
      "loss": 0.0578,
      "step": 254850
    },
    {
      "epoch": 98.97,
      "learning_rate": 6.803365695792881e-06,
      "loss": 0.0162,
      "step": 254860
    },
    {
      "epoch": 98.98,
      "learning_rate": 6.80284789644013e-06,
      "loss": 0.0013,
      "step": 254870
    },
    {
      "epoch": 98.98,
      "learning_rate": 6.802330097087379e-06,
      "loss": 0.0004,
      "step": 254880
    },
    {
      "epoch": 98.99,
      "learning_rate": 6.801812297734628e-06,
      "loss": 0.017,
      "step": 254890
    },
    {
      "epoch": 98.99,
      "learning_rate": 6.801294498381878e-06,
      "loss": 0.0297,
      "step": 254900
    },
    {
      "epoch": 98.99,
      "learning_rate": 6.8007766990291265e-06,
      "loss": 0.1409,
      "step": 254910
    },
    {
      "epoch": 99.0,
      "learning_rate": 6.800258899676376e-06,
      "loss": 0.062,
      "step": 254920
    },
    {
      "epoch": 99.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.3688491880893707,
      "eval_runtime": 8.2271,
      "eval_samples_per_second": 441.835,
      "eval_steps_per_second": 55.305,
      "step": 254925
    },
    {
      "epoch": 99.0,
      "learning_rate": 6.799741100323625e-06,
      "loss": 0.0874,
      "step": 254930
    },
    {
      "epoch": 99.01,
      "learning_rate": 6.7992233009708745e-06,
      "loss": 0.1264,
      "step": 254940
    },
    {
      "epoch": 99.01,
      "learning_rate": 6.798705501618123e-06,
      "loss": 0.0868,
      "step": 254950
    },
    {
      "epoch": 99.01,
      "learning_rate": 6.798187702265373e-06,
      "loss": 0.0552,
      "step": 254960
    },
    {
      "epoch": 99.02,
      "learning_rate": 6.797669902912622e-06,
      "loss": 0.0306,
      "step": 254970
    },
    {
      "epoch": 99.02,
      "learning_rate": 6.7971521035598705e-06,
      "loss": 0.0732,
      "step": 254980
    },
    {
      "epoch": 99.03,
      "learning_rate": 6.79663430420712e-06,
      "loss": 0.0319,
      "step": 254990
    },
    {
      "epoch": 99.03,
      "learning_rate": 6.79611650485437e-06,
      "loss": 0.0391,
      "step": 255000
    },
    {
      "epoch": 99.03,
      "learning_rate": 6.7955987055016185e-06,
      "loss": 0.0139,
      "step": 255010
    },
    {
      "epoch": 99.04,
      "learning_rate": 6.795080906148867e-06,
      "loss": 0.0267,
      "step": 255020
    },
    {
      "epoch": 99.04,
      "learning_rate": 6.794563106796117e-06,
      "loss": 0.0088,
      "step": 255030
    },
    {
      "epoch": 99.04,
      "learning_rate": 6.7940453074433665e-06,
      "loss": 0.0388,
      "step": 255040
    },
    {
      "epoch": 99.05,
      "learning_rate": 6.793527508090615e-06,
      "loss": 0.0369,
      "step": 255050
    },
    {
      "epoch": 99.05,
      "learning_rate": 6.793009708737864e-06,
      "loss": 0.1204,
      "step": 255060
    },
    {
      "epoch": 99.06,
      "learning_rate": 6.792491909385114e-06,
      "loss": 0.1541,
      "step": 255070
    },
    {
      "epoch": 99.06,
      "learning_rate": 6.791974110032363e-06,
      "loss": 0.1017,
      "step": 255080
    },
    {
      "epoch": 99.06,
      "learning_rate": 6.791456310679613e-06,
      "loss": 0.0132,
      "step": 255090
    },
    {
      "epoch": 99.07,
      "learning_rate": 6.790938511326861e-06,
      "loss": 0.0648,
      "step": 255100
    },
    {
      "epoch": 99.07,
      "learning_rate": 6.7904207119741104e-06,
      "loss": 0.0967,
      "step": 255110
    },
    {
      "epoch": 99.08,
      "learning_rate": 6.78990291262136e-06,
      "loss": 0.0359,
      "step": 255120
    },
    {
      "epoch": 99.08,
      "learning_rate": 6.78938511326861e-06,
      "loss": 0.0748,
      "step": 255130
    },
    {
      "epoch": 99.08,
      "learning_rate": 6.788867313915858e-06,
      "loss": 0.113,
      "step": 255140
    },
    {
      "epoch": 99.09,
      "learning_rate": 6.788349514563107e-06,
      "loss": 0.0062,
      "step": 255150
    },
    {
      "epoch": 99.09,
      "learning_rate": 6.787831715210357e-06,
      "loss": 0.0904,
      "step": 255160
    },
    {
      "epoch": 99.1,
      "learning_rate": 6.7873139158576064e-06,
      "loss": 0.1337,
      "step": 255170
    },
    {
      "epoch": 99.1,
      "learning_rate": 6.786796116504854e-06,
      "loss": 0.0975,
      "step": 255180
    },
    {
      "epoch": 99.1,
      "learning_rate": 6.786278317152104e-06,
      "loss": 0.0411,
      "step": 255190
    },
    {
      "epoch": 99.11,
      "learning_rate": 6.785760517799354e-06,
      "loss": 0.0049,
      "step": 255200
    },
    {
      "epoch": 99.11,
      "learning_rate": 6.785242718446603e-06,
      "loss": 0.0501,
      "step": 255210
    },
    {
      "epoch": 99.11,
      "learning_rate": 6.784724919093851e-06,
      "loss": 0.0684,
      "step": 255220
    },
    {
      "epoch": 99.12,
      "learning_rate": 6.784207119741101e-06,
      "loss": 0.0091,
      "step": 255230
    },
    {
      "epoch": 99.12,
      "learning_rate": 6.78368932038835e-06,
      "loss": 0.0039,
      "step": 255240
    },
    {
      "epoch": 99.13,
      "learning_rate": 6.7831715210356e-06,
      "loss": 0.1021,
      "step": 255250
    },
    {
      "epoch": 99.13,
      "learning_rate": 6.782653721682848e-06,
      "loss": 0.0027,
      "step": 255260
    },
    {
      "epoch": 99.13,
      "learning_rate": 6.7821359223300975e-06,
      "loss": 0.0012,
      "step": 255270
    },
    {
      "epoch": 99.14,
      "learning_rate": 6.781618122977347e-06,
      "loss": 0.0244,
      "step": 255280
    },
    {
      "epoch": 99.14,
      "learning_rate": 6.781100323624597e-06,
      "loss": 0.0103,
      "step": 255290
    },
    {
      "epoch": 99.15,
      "learning_rate": 6.780582524271845e-06,
      "loss": 0.0179,
      "step": 255300
    },
    {
      "epoch": 99.15,
      "learning_rate": 6.780064724919094e-06,
      "loss": 0.0008,
      "step": 255310
    },
    {
      "epoch": 99.15,
      "learning_rate": 6.779546925566344e-06,
      "loss": 0.1987,
      "step": 255320
    },
    {
      "epoch": 99.16,
      "learning_rate": 6.7790291262135936e-06,
      "loss": 0.005,
      "step": 255330
    },
    {
      "epoch": 99.16,
      "learning_rate": 6.7785113268608415e-06,
      "loss": 0.0417,
      "step": 255340
    },
    {
      "epoch": 99.17,
      "learning_rate": 6.777993527508091e-06,
      "loss": 0.0625,
      "step": 255350
    },
    {
      "epoch": 99.17,
      "learning_rate": 6.777475728155341e-06,
      "loss": 0.0335,
      "step": 255360
    },
    {
      "epoch": 99.17,
      "learning_rate": 6.77695792880259e-06,
      "loss": 0.1089,
      "step": 255370
    },
    {
      "epoch": 99.18,
      "learning_rate": 6.776440129449838e-06,
      "loss": 0.0567,
      "step": 255380
    },
    {
      "epoch": 99.18,
      "learning_rate": 6.775922330097088e-06,
      "loss": 0.0084,
      "step": 255390
    },
    {
      "epoch": 99.18,
      "learning_rate": 6.7754045307443375e-06,
      "loss": 0.1237,
      "step": 255400
    },
    {
      "epoch": 99.19,
      "learning_rate": 6.774886731391587e-06,
      "loss": 0.0139,
      "step": 255410
    },
    {
      "epoch": 99.19,
      "learning_rate": 6.774368932038835e-06,
      "loss": 0.0351,
      "step": 255420
    },
    {
      "epoch": 99.2,
      "learning_rate": 6.773851132686085e-06,
      "loss": 0.0104,
      "step": 255430
    },
    {
      "epoch": 99.2,
      "learning_rate": 6.773333333333334e-06,
      "loss": 0.0405,
      "step": 255440
    },
    {
      "epoch": 99.2,
      "learning_rate": 6.772815533980583e-06,
      "loss": 0.0534,
      "step": 255450
    },
    {
      "epoch": 99.21,
      "learning_rate": 6.772297734627832e-06,
      "loss": 0.0338,
      "step": 255460
    },
    {
      "epoch": 99.21,
      "learning_rate": 6.7717799352750814e-06,
      "loss": 0.0011,
      "step": 255470
    },
    {
      "epoch": 99.22,
      "learning_rate": 6.771262135922331e-06,
      "loss": 0.0167,
      "step": 255480
    },
    {
      "epoch": 99.22,
      "learning_rate": 6.77074433656958e-06,
      "loss": 0.1182,
      "step": 255490
    },
    {
      "epoch": 99.22,
      "learning_rate": 6.770226537216829e-06,
      "loss": 0.2223,
      "step": 255500
    },
    {
      "epoch": 99.23,
      "learning_rate": 6.769708737864078e-06,
      "loss": 0.0764,
      "step": 255510
    },
    {
      "epoch": 99.23,
      "learning_rate": 6.769190938511328e-06,
      "loss": 0.1797,
      "step": 255520
    },
    {
      "epoch": 99.23,
      "learning_rate": 6.768673139158577e-06,
      "loss": 0.0548,
      "step": 255530
    },
    {
      "epoch": 99.24,
      "learning_rate": 6.768155339805825e-06,
      "loss": 0.0853,
      "step": 255540
    },
    {
      "epoch": 99.24,
      "learning_rate": 6.767637540453075e-06,
      "loss": 0.0092,
      "step": 255550
    },
    {
      "epoch": 99.25,
      "learning_rate": 6.767119741100325e-06,
      "loss": 0.0895,
      "step": 255560
    },
    {
      "epoch": 99.25,
      "learning_rate": 6.766601941747573e-06,
      "loss": 0.0482,
      "step": 255570
    },
    {
      "epoch": 99.25,
      "learning_rate": 6.766084142394822e-06,
      "loss": 0.017,
      "step": 255580
    },
    {
      "epoch": 99.26,
      "learning_rate": 6.765566343042072e-06,
      "loss": 0.0415,
      "step": 255590
    },
    {
      "epoch": 99.26,
      "learning_rate": 6.7650485436893205e-06,
      "loss": 0.0215,
      "step": 255600
    },
    {
      "epoch": 99.27,
      "learning_rate": 6.76453074433657e-06,
      "loss": 0.083,
      "step": 255610
    },
    {
      "epoch": 99.27,
      "learning_rate": 6.764012944983819e-06,
      "loss": 0.0656,
      "step": 255620
    },
    {
      "epoch": 99.27,
      "learning_rate": 6.7634951456310685e-06,
      "loss": 0.0535,
      "step": 255630
    },
    {
      "epoch": 99.28,
      "learning_rate": 6.762977346278317e-06,
      "loss": 0.0928,
      "step": 255640
    },
    {
      "epoch": 99.28,
      "learning_rate": 6.762459546925567e-06,
      "loss": 0.0009,
      "step": 255650
    },
    {
      "epoch": 99.29,
      "learning_rate": 6.7619417475728165e-06,
      "loss": 0.1442,
      "step": 255660
    },
    {
      "epoch": 99.29,
      "learning_rate": 6.761423948220065e-06,
      "loss": 0.0214,
      "step": 255670
    },
    {
      "epoch": 99.29,
      "learning_rate": 6.760906148867314e-06,
      "loss": 0.0126,
      "step": 255680
    },
    {
      "epoch": 99.3,
      "learning_rate": 6.760388349514564e-06,
      "loss": 0.0817,
      "step": 255690
    },
    {
      "epoch": 99.3,
      "learning_rate": 6.759870550161813e-06,
      "loss": 0.1712,
      "step": 255700
    },
    {
      "epoch": 99.3,
      "learning_rate": 6.759352750809062e-06,
      "loss": 0.0095,
      "step": 255710
    },
    {
      "epoch": 99.31,
      "learning_rate": 6.758834951456311e-06,
      "loss": 0.0833,
      "step": 255720
    },
    {
      "epoch": 99.31,
      "learning_rate": 6.7583171521035605e-06,
      "loss": 0.0041,
      "step": 255730
    },
    {
      "epoch": 99.32,
      "learning_rate": 6.75779935275081e-06,
      "loss": 0.0093,
      "step": 255740
    },
    {
      "epoch": 99.32,
      "learning_rate": 6.757281553398059e-06,
      "loss": 0.1266,
      "step": 255750
    },
    {
      "epoch": 99.32,
      "learning_rate": 6.756763754045308e-06,
      "loss": 0.0112,
      "step": 255760
    },
    {
      "epoch": 99.33,
      "learning_rate": 6.756245954692557e-06,
      "loss": 0.0136,
      "step": 255770
    },
    {
      "epoch": 99.33,
      "learning_rate": 6.755728155339807e-06,
      "loss": 0.0805,
      "step": 255780
    },
    {
      "epoch": 99.34,
      "learning_rate": 6.755210355987055e-06,
      "loss": 0.0736,
      "step": 255790
    },
    {
      "epoch": 99.34,
      "learning_rate": 6.7546925566343044e-06,
      "loss": 0.1187,
      "step": 255800
    },
    {
      "epoch": 99.34,
      "learning_rate": 6.754174757281554e-06,
      "loss": 0.0606,
      "step": 255810
    },
    {
      "epoch": 99.35,
      "learning_rate": 6.753656957928804e-06,
      "loss": 0.0941,
      "step": 255820
    },
    {
      "epoch": 99.35,
      "learning_rate": 6.753139158576052e-06,
      "loss": 0.0115,
      "step": 255830
    },
    {
      "epoch": 99.36,
      "learning_rate": 6.752621359223301e-06,
      "loss": 0.0078,
      "step": 255840
    },
    {
      "epoch": 99.36,
      "learning_rate": 6.752103559870551e-06,
      "loss": 0.0927,
      "step": 255850
    },
    {
      "epoch": 99.36,
      "learning_rate": 6.7515857605178004e-06,
      "loss": 0.061,
      "step": 255860
    },
    {
      "epoch": 99.37,
      "learning_rate": 6.751067961165048e-06,
      "loss": 0.0228,
      "step": 255870
    },
    {
      "epoch": 99.37,
      "learning_rate": 6.750550161812298e-06,
      "loss": 0.1399,
      "step": 255880
    },
    {
      "epoch": 99.37,
      "learning_rate": 6.750032362459548e-06,
      "loss": 0.0118,
      "step": 255890
    },
    {
      "epoch": 99.38,
      "learning_rate": 6.749514563106797e-06,
      "loss": 0.0315,
      "step": 255900
    },
    {
      "epoch": 99.38,
      "learning_rate": 6.748996763754045e-06,
      "loss": 0.0168,
      "step": 255910
    },
    {
      "epoch": 99.39,
      "learning_rate": 6.748478964401295e-06,
      "loss": 0.0926,
      "step": 255920
    },
    {
      "epoch": 99.39,
      "learning_rate": 6.747961165048544e-06,
      "loss": 0.0147,
      "step": 255930
    },
    {
      "epoch": 99.39,
      "learning_rate": 6.747443365695794e-06,
      "loss": 0.0009,
      "step": 255940
    },
    {
      "epoch": 99.4,
      "learning_rate": 6.746925566343042e-06,
      "loss": 0.0346,
      "step": 255950
    },
    {
      "epoch": 99.4,
      "learning_rate": 6.7464077669902915e-06,
      "loss": 0.1183,
      "step": 255960
    },
    {
      "epoch": 99.41,
      "learning_rate": 6.745889967637541e-06,
      "loss": 0.0365,
      "step": 255970
    },
    {
      "epoch": 99.41,
      "learning_rate": 6.745372168284791e-06,
      "loss": 0.0001,
      "step": 255980
    },
    {
      "epoch": 99.41,
      "learning_rate": 6.744854368932039e-06,
      "loss": 0.0609,
      "step": 255990
    },
    {
      "epoch": 99.42,
      "learning_rate": 6.744336569579288e-06,
      "loss": 0.0173,
      "step": 256000
    },
    {
      "epoch": 99.42,
      "learning_rate": 6.743818770226538e-06,
      "loss": 0.066,
      "step": 256010
    },
    {
      "epoch": 99.43,
      "learning_rate": 6.7433009708737876e-06,
      "loss": 0.0849,
      "step": 256020
    },
    {
      "epoch": 99.43,
      "learning_rate": 6.7427831715210355e-06,
      "loss": 0.0719,
      "step": 256030
    },
    {
      "epoch": 99.43,
      "learning_rate": 6.742265372168285e-06,
      "loss": 0.0004,
      "step": 256040
    },
    {
      "epoch": 99.44,
      "learning_rate": 6.741747572815535e-06,
      "loss": 0.0261,
      "step": 256050
    },
    {
      "epoch": 99.44,
      "learning_rate": 6.741229773462784e-06,
      "loss": 0.0102,
      "step": 256060
    },
    {
      "epoch": 99.44,
      "learning_rate": 6.740711974110032e-06,
      "loss": 0.0363,
      "step": 256070
    },
    {
      "epoch": 99.45,
      "learning_rate": 6.740194174757282e-06,
      "loss": 0.1431,
      "step": 256080
    },
    {
      "epoch": 99.45,
      "learning_rate": 6.7396763754045315e-06,
      "loss": 0.0625,
      "step": 256090
    },
    {
      "epoch": 99.46,
      "learning_rate": 6.739158576051781e-06,
      "loss": 0.0199,
      "step": 256100
    },
    {
      "epoch": 99.46,
      "learning_rate": 6.738640776699029e-06,
      "loss": 0.0234,
      "step": 256110
    },
    {
      "epoch": 99.46,
      "learning_rate": 6.738122977346279e-06,
      "loss": 0.1127,
      "step": 256120
    },
    {
      "epoch": 99.47,
      "learning_rate": 6.737605177993528e-06,
      "loss": 0.0528,
      "step": 256130
    },
    {
      "epoch": 99.47,
      "learning_rate": 6.737087378640778e-06,
      "loss": 0.0465,
      "step": 256140
    },
    {
      "epoch": 99.48,
      "learning_rate": 6.736569579288026e-06,
      "loss": 0.0456,
      "step": 256150
    },
    {
      "epoch": 99.48,
      "learning_rate": 6.7360517799352754e-06,
      "loss": 0.0634,
      "step": 256160
    },
    {
      "epoch": 99.48,
      "learning_rate": 6.735533980582525e-06,
      "loss": 0.0434,
      "step": 256170
    },
    {
      "epoch": 99.49,
      "learning_rate": 6.735016181229775e-06,
      "loss": 0.0183,
      "step": 256180
    },
    {
      "epoch": 99.49,
      "learning_rate": 6.734498381877023e-06,
      "loss": 0.2522,
      "step": 256190
    },
    {
      "epoch": 99.5,
      "learning_rate": 6.733980582524272e-06,
      "loss": 0.0095,
      "step": 256200
    },
    {
      "epoch": 99.5,
      "learning_rate": 6.733462783171522e-06,
      "loss": 0.0443,
      "step": 256210
    },
    {
      "epoch": 99.5,
      "learning_rate": 6.7329449838187714e-06,
      "loss": 0.0476,
      "step": 256220
    },
    {
      "epoch": 99.51,
      "learning_rate": 6.73242718446602e-06,
      "loss": 0.0195,
      "step": 256230
    },
    {
      "epoch": 99.51,
      "learning_rate": 6.731909385113269e-06,
      "loss": 0.0305,
      "step": 256240
    },
    {
      "epoch": 99.51,
      "learning_rate": 6.731391585760519e-06,
      "loss": 0.079,
      "step": 256250
    },
    {
      "epoch": 99.52,
      "learning_rate": 6.730873786407767e-06,
      "loss": 0.0273,
      "step": 256260
    },
    {
      "epoch": 99.52,
      "learning_rate": 6.730355987055017e-06,
      "loss": 0.1401,
      "step": 256270
    },
    {
      "epoch": 99.53,
      "learning_rate": 6.729838187702266e-06,
      "loss": 0.0119,
      "step": 256280
    },
    {
      "epoch": 99.53,
      "learning_rate": 6.729320388349515e-06,
      "loss": 0.0367,
      "step": 256290
    },
    {
      "epoch": 99.53,
      "learning_rate": 6.728802588996764e-06,
      "loss": 0.0172,
      "step": 256300
    },
    {
      "epoch": 99.54,
      "learning_rate": 6.728284789644014e-06,
      "loss": 0.0008,
      "step": 256310
    },
    {
      "epoch": 99.54,
      "learning_rate": 6.7277669902912625e-06,
      "loss": 0.1096,
      "step": 256320
    },
    {
      "epoch": 99.55,
      "learning_rate": 6.727249190938512e-06,
      "loss": 0.017,
      "step": 256330
    },
    {
      "epoch": 99.55,
      "learning_rate": 6.726731391585761e-06,
      "loss": 0.0038,
      "step": 256340
    },
    {
      "epoch": 99.55,
      "learning_rate": 6.7262135922330105e-06,
      "loss": 0.0467,
      "step": 256350
    },
    {
      "epoch": 99.56,
      "learning_rate": 6.725695792880259e-06,
      "loss": 0.024,
      "step": 256360
    },
    {
      "epoch": 99.56,
      "learning_rate": 6.725177993527509e-06,
      "loss": 0.0238,
      "step": 256370
    },
    {
      "epoch": 99.57,
      "learning_rate": 6.724660194174758e-06,
      "loss": 0.0402,
      "step": 256380
    },
    {
      "epoch": 99.57,
      "learning_rate": 6.724142394822007e-06,
      "loss": 0.0012,
      "step": 256390
    },
    {
      "epoch": 99.57,
      "learning_rate": 6.723624595469256e-06,
      "loss": 0.0912,
      "step": 256400
    },
    {
      "epoch": 99.58,
      "learning_rate": 6.723106796116506e-06,
      "loss": 0.0849,
      "step": 256410
    },
    {
      "epoch": 99.58,
      "learning_rate": 6.7225889967637545e-06,
      "loss": 0.0165,
      "step": 256420
    },
    {
      "epoch": 99.58,
      "learning_rate": 6.722071197411004e-06,
      "loss": 0.0996,
      "step": 256430
    },
    {
      "epoch": 99.59,
      "learning_rate": 6.721553398058253e-06,
      "loss": 0.0385,
      "step": 256440
    },
    {
      "epoch": 99.59,
      "learning_rate": 6.721035598705502e-06,
      "loss": 0.0439,
      "step": 256450
    },
    {
      "epoch": 99.6,
      "learning_rate": 6.720517799352751e-06,
      "loss": 0.0555,
      "step": 256460
    },
    {
      "epoch": 99.6,
      "learning_rate": 6.720000000000001e-06,
      "loss": 0.0202,
      "step": 256470
    },
    {
      "epoch": 99.6,
      "learning_rate": 6.71948220064725e-06,
      "loss": 0.0203,
      "step": 256480
    },
    {
      "epoch": 99.61,
      "learning_rate": 6.7189644012944984e-06,
      "loss": 0.059,
      "step": 256490
    },
    {
      "epoch": 99.61,
      "learning_rate": 6.718446601941748e-06,
      "loss": 0.0323,
      "step": 256500
    },
    {
      "epoch": 99.62,
      "learning_rate": 6.717928802588998e-06,
      "loss": 0.012,
      "step": 256510
    },
    {
      "epoch": 99.62,
      "learning_rate": 6.7174110032362464e-06,
      "loss": 0.0259,
      "step": 256520
    },
    {
      "epoch": 99.62,
      "learning_rate": 6.716893203883495e-06,
      "loss": 0.087,
      "step": 256530
    },
    {
      "epoch": 99.63,
      "learning_rate": 6.716375404530745e-06,
      "loss": 0.0435,
      "step": 256540
    },
    {
      "epoch": 99.63,
      "learning_rate": 6.7158576051779944e-06,
      "loss": 0.0802,
      "step": 256550
    },
    {
      "epoch": 99.63,
      "learning_rate": 6.715339805825243e-06,
      "loss": 0.0464,
      "step": 256560
    },
    {
      "epoch": 99.64,
      "learning_rate": 6.714822006472492e-06,
      "loss": 0.0958,
      "step": 256570
    },
    {
      "epoch": 99.64,
      "learning_rate": 6.714304207119742e-06,
      "loss": 0.0571,
      "step": 256580
    },
    {
      "epoch": 99.65,
      "learning_rate": 6.713786407766991e-06,
      "loss": 0.005,
      "step": 256590
    },
    {
      "epoch": 99.65,
      "learning_rate": 6.713268608414239e-06,
      "loss": 0.044,
      "step": 256600
    },
    {
      "epoch": 99.65,
      "learning_rate": 6.712750809061489e-06,
      "loss": 0.0069,
      "step": 256610
    },
    {
      "epoch": 99.66,
      "learning_rate": 6.712233009708738e-06,
      "loss": 0.0103,
      "step": 256620
    },
    {
      "epoch": 99.66,
      "learning_rate": 6.711715210355988e-06,
      "loss": 0.0832,
      "step": 256630
    },
    {
      "epoch": 99.67,
      "learning_rate": 6.711197411003236e-06,
      "loss": 0.0996,
      "step": 256640
    },
    {
      "epoch": 99.67,
      "learning_rate": 6.7106796116504855e-06,
      "loss": 0.0666,
      "step": 256650
    },
    {
      "epoch": 99.67,
      "learning_rate": 6.710161812297735e-06,
      "loss": 0.0128,
      "step": 256660
    },
    {
      "epoch": 99.68,
      "learning_rate": 6.709644012944985e-06,
      "loss": 0.0069,
      "step": 256670
    },
    {
      "epoch": 99.68,
      "learning_rate": 6.709126213592233e-06,
      "loss": 0.0141,
      "step": 256680
    },
    {
      "epoch": 99.69,
      "learning_rate": 6.708608414239482e-06,
      "loss": 0.0277,
      "step": 256690
    },
    {
      "epoch": 99.69,
      "learning_rate": 6.708090614886732e-06,
      "loss": 0.1315,
      "step": 256700
    },
    {
      "epoch": 99.69,
      "learning_rate": 6.7075728155339816e-06,
      "loss": 0.0206,
      "step": 256710
    },
    {
      "epoch": 99.7,
      "learning_rate": 6.7070550161812295e-06,
      "loss": 0.0273,
      "step": 256720
    },
    {
      "epoch": 99.7,
      "learning_rate": 6.706537216828479e-06,
      "loss": 0.0003,
      "step": 256730
    },
    {
      "epoch": 99.7,
      "learning_rate": 6.706019417475729e-06,
      "loss": 0.0278,
      "step": 256740
    },
    {
      "epoch": 99.71,
      "learning_rate": 6.705501618122978e-06,
      "loss": 0.1183,
      "step": 256750
    },
    {
      "epoch": 99.71,
      "learning_rate": 6.704983818770228e-06,
      "loss": 0.0165,
      "step": 256760
    },
    {
      "epoch": 99.72,
      "learning_rate": 6.704466019417476e-06,
      "loss": 0.095,
      "step": 256770
    },
    {
      "epoch": 99.72,
      "learning_rate": 6.7039482200647255e-06,
      "loss": 0.0739,
      "step": 256780
    },
    {
      "epoch": 99.72,
      "learning_rate": 6.703430420711975e-06,
      "loss": 0.0528,
      "step": 256790
    },
    {
      "epoch": 99.73,
      "learning_rate": 6.702912621359225e-06,
      "loss": 0.0629,
      "step": 256800
    },
    {
      "epoch": 99.73,
      "learning_rate": 6.702394822006473e-06,
      "loss": 0.1046,
      "step": 256810
    },
    {
      "epoch": 99.74,
      "learning_rate": 6.701877022653722e-06,
      "loss": 0.0089,
      "step": 256820
    },
    {
      "epoch": 99.74,
      "learning_rate": 6.701359223300972e-06,
      "loss": 0.1678,
      "step": 256830
    },
    {
      "epoch": 99.74,
      "learning_rate": 6.7008414239482215e-06,
      "loss": 0.0192,
      "step": 256840
    },
    {
      "epoch": 99.75,
      "learning_rate": 6.7003236245954694e-06,
      "loss": 0.0008,
      "step": 256850
    },
    {
      "epoch": 99.75,
      "learning_rate": 6.699805825242719e-06,
      "loss": 0.0149,
      "step": 256860
    },
    {
      "epoch": 99.76,
      "learning_rate": 6.699288025889969e-06,
      "loss": 0.0098,
      "step": 256870
    },
    {
      "epoch": 99.76,
      "learning_rate": 6.698770226537218e-06,
      "loss": 0.0102,
      "step": 256880
    },
    {
      "epoch": 99.76,
      "learning_rate": 6.698252427184466e-06,
      "loss": 0.0503,
      "step": 256890
    },
    {
      "epoch": 99.77,
      "learning_rate": 6.697734627831716e-06,
      "loss": 0.0418,
      "step": 256900
    },
    {
      "epoch": 99.77,
      "learning_rate": 6.6972168284789654e-06,
      "loss": 0.0775,
      "step": 256910
    },
    {
      "epoch": 99.77,
      "learning_rate": 6.696699029126214e-06,
      "loss": 0.0031,
      "step": 256920
    },
    {
      "epoch": 99.78,
      "learning_rate": 6.696181229773463e-06,
      "loss": 0.0746,
      "step": 256930
    },
    {
      "epoch": 99.78,
      "learning_rate": 6.695663430420713e-06,
      "loss": 0.1522,
      "step": 256940
    },
    {
      "epoch": 99.79,
      "learning_rate": 6.695145631067962e-06,
      "loss": 0.0564,
      "step": 256950
    },
    {
      "epoch": 99.79,
      "learning_rate": 6.694627831715211e-06,
      "loss": 0.0284,
      "step": 256960
    },
    {
      "epoch": 99.79,
      "learning_rate": 6.69411003236246e-06,
      "loss": 0.059,
      "step": 256970
    },
    {
      "epoch": 99.8,
      "learning_rate": 6.693592233009709e-06,
      "loss": 0.0467,
      "step": 256980
    },
    {
      "epoch": 99.8,
      "learning_rate": 6.693074433656959e-06,
      "loss": 0.0558,
      "step": 256990
    },
    {
      "epoch": 99.81,
      "learning_rate": 6.692556634304208e-06,
      "loss": 0.0473,
      "step": 257000
    },
    {
      "epoch": 99.81,
      "learning_rate": 6.6920388349514565e-06,
      "loss": 0.215,
      "step": 257010
    },
    {
      "epoch": 99.81,
      "learning_rate": 6.691521035598706e-06,
      "loss": 0.0739,
      "step": 257020
    },
    {
      "epoch": 99.82,
      "learning_rate": 6.691003236245956e-06,
      "loss": 0.0324,
      "step": 257030
    },
    {
      "epoch": 99.82,
      "learning_rate": 6.6904854368932045e-06,
      "loss": 0.0297,
      "step": 257040
    },
    {
      "epoch": 99.83,
      "learning_rate": 6.689967637540453e-06,
      "loss": 0.0845,
      "step": 257050
    },
    {
      "epoch": 99.83,
      "learning_rate": 6.689449838187703e-06,
      "loss": 0.0119,
      "step": 257060
    },
    {
      "epoch": 99.83,
      "learning_rate": 6.688932038834952e-06,
      "loss": 0.0921,
      "step": 257070
    },
    {
      "epoch": 99.84,
      "learning_rate": 6.688414239482201e-06,
      "loss": 0.0497,
      "step": 257080
    },
    {
      "epoch": 99.84,
      "learning_rate": 6.68789644012945e-06,
      "loss": 0.0501,
      "step": 257090
    },
    {
      "epoch": 99.84,
      "learning_rate": 6.6873786407767e-06,
      "loss": 0.0595,
      "step": 257100
    },
    {
      "epoch": 99.85,
      "learning_rate": 6.6868608414239485e-06,
      "loss": 0.0022,
      "step": 257110
    },
    {
      "epoch": 99.85,
      "learning_rate": 6.686343042071198e-06,
      "loss": 0.0135,
      "step": 257120
    },
    {
      "epoch": 99.86,
      "learning_rate": 6.685825242718447e-06,
      "loss": 0.1788,
      "step": 257130
    },
    {
      "epoch": 99.86,
      "learning_rate": 6.6853074433656965e-06,
      "loss": 0.0391,
      "step": 257140
    },
    {
      "epoch": 99.86,
      "learning_rate": 6.684789644012945e-06,
      "loss": 0.0009,
      "step": 257150
    },
    {
      "epoch": 99.87,
      "learning_rate": 6.684271844660195e-06,
      "loss": 0.0333,
      "step": 257160
    },
    {
      "epoch": 99.87,
      "learning_rate": 6.683754045307444e-06,
      "loss": 0.0221,
      "step": 257170
    },
    {
      "epoch": 99.88,
      "learning_rate": 6.683236245954693e-06,
      "loss": 0.0489,
      "step": 257180
    },
    {
      "epoch": 99.88,
      "learning_rate": 6.682718446601942e-06,
      "loss": 0.0596,
      "step": 257190
    },
    {
      "epoch": 99.88,
      "learning_rate": 6.682200647249192e-06,
      "loss": 0.0201,
      "step": 257200
    },
    {
      "epoch": 99.89,
      "learning_rate": 6.6816828478964404e-06,
      "loss": 0.1118,
      "step": 257210
    },
    {
      "epoch": 99.89,
      "learning_rate": 6.68116504854369e-06,
      "loss": 0.0611,
      "step": 257220
    },
    {
      "epoch": 99.9,
      "learning_rate": 6.680647249190939e-06,
      "loss": 0.0209,
      "step": 257230
    },
    {
      "epoch": 99.9,
      "learning_rate": 6.6801294498381884e-06,
      "loss": 0.0784,
      "step": 257240
    },
    {
      "epoch": 99.9,
      "learning_rate": 6.679611650485437e-06,
      "loss": 0.0227,
      "step": 257250
    },
    {
      "epoch": 99.91,
      "learning_rate": 6.679093851132686e-06,
      "loss": 0.0104,
      "step": 257260
    },
    {
      "epoch": 99.91,
      "learning_rate": 6.678576051779936e-06,
      "loss": 0.045,
      "step": 257270
    },
    {
      "epoch": 99.91,
      "learning_rate": 6.678058252427185e-06,
      "loss": 0.0877,
      "step": 257280
    },
    {
      "epoch": 99.92,
      "learning_rate": 6.677540453074434e-06,
      "loss": 0.0299,
      "step": 257290
    },
    {
      "epoch": 99.92,
      "learning_rate": 6.677022653721683e-06,
      "loss": 0.0692,
      "step": 257300
    },
    {
      "epoch": 99.93,
      "learning_rate": 6.676504854368932e-06,
      "loss": 0.0588,
      "step": 257310
    },
    {
      "epoch": 99.93,
      "learning_rate": 6.675987055016182e-06,
      "loss": 0.0033,
      "step": 257320
    },
    {
      "epoch": 99.93,
      "learning_rate": 6.675469255663432e-06,
      "loss": 0.0009,
      "step": 257330
    },
    {
      "epoch": 99.94,
      "learning_rate": 6.6749514563106795e-06,
      "loss": 0.0257,
      "step": 257340
    },
    {
      "epoch": 99.94,
      "learning_rate": 6.674433656957929e-06,
      "loss": 0.0338,
      "step": 257350
    },
    {
      "epoch": 99.95,
      "learning_rate": 6.673915857605179e-06,
      "loss": 0.0657,
      "step": 257360
    },
    {
      "epoch": 99.95,
      "learning_rate": 6.673398058252428e-06,
      "loss": 0.046,
      "step": 257370
    },
    {
      "epoch": 99.95,
      "learning_rate": 6.672880258899676e-06,
      "loss": 0.0422,
      "step": 257380
    },
    {
      "epoch": 99.96,
      "learning_rate": 6.672362459546926e-06,
      "loss": 0.0733,
      "step": 257390
    },
    {
      "epoch": 99.96,
      "learning_rate": 6.6718446601941755e-06,
      "loss": 0.0702,
      "step": 257400
    },
    {
      "epoch": 99.97,
      "learning_rate": 6.671326860841425e-06,
      "loss": 0.0109,
      "step": 257410
    },
    {
      "epoch": 99.97,
      "learning_rate": 6.670809061488673e-06,
      "loss": 0.0084,
      "step": 257420
    },
    {
      "epoch": 99.97,
      "learning_rate": 6.670291262135923e-06,
      "loss": 0.0247,
      "step": 257430
    },
    {
      "epoch": 99.98,
      "learning_rate": 6.669773462783172e-06,
      "loss": 0.0815,
      "step": 257440
    },
    {
      "epoch": 99.98,
      "learning_rate": 6.669255663430422e-06,
      "loss": 0.0207,
      "step": 257450
    },
    {
      "epoch": 99.98,
      "learning_rate": 6.66873786407767e-06,
      "loss": 0.1644,
      "step": 257460
    },
    {
      "epoch": 99.99,
      "learning_rate": 6.6682200647249195e-06,
      "loss": 0.0304,
      "step": 257470
    },
    {
      "epoch": 99.99,
      "learning_rate": 6.667702265372169e-06,
      "loss": 0.054,
      "step": 257480
    },
    {
      "epoch": 100.0,
      "learning_rate": 6.667184466019419e-06,
      "loss": 0.0016,
      "step": 257490
    },
    {
      "epoch": 100.0,
      "learning_rate": 6.666666666666667e-06,
      "loss": 0.1217,
      "step": 257500
    },
    {
      "epoch": 100.0,
      "eval_accuracy": 0.9507565337001376,
      "eval_loss": 0.3741964101791382,
      "eval_runtime": 8.2071,
      "eval_samples_per_second": 442.91,
      "eval_steps_per_second": 55.44,
      "step": 257500
    },
    {
      "epoch": 100.0,
      "learning_rate": 6.666148867313916e-06,
      "loss": 0.0838,
      "step": 257510
    },
    {
      "epoch": 100.01,
      "learning_rate": 6.665631067961166e-06,
      "loss": 0.0515,
      "step": 257520
    },
    {
      "epoch": 100.01,
      "learning_rate": 6.6651132686084155e-06,
      "loss": 0.0096,
      "step": 257530
    },
    {
      "epoch": 100.02,
      "learning_rate": 6.6645954692556634e-06,
      "loss": 0.0482,
      "step": 257540
    },
    {
      "epoch": 100.02,
      "learning_rate": 6.664077669902913e-06,
      "loss": 0.0186,
      "step": 257550
    },
    {
      "epoch": 100.02,
      "learning_rate": 6.663559870550163e-06,
      "loss": 0.0087,
      "step": 257560
    },
    {
      "epoch": 100.03,
      "learning_rate": 6.663042071197412e-06,
      "loss": 0.0068,
      "step": 257570
    },
    {
      "epoch": 100.03,
      "learning_rate": 6.66252427184466e-06,
      "loss": 0.0458,
      "step": 257580
    },
    {
      "epoch": 100.03,
      "learning_rate": 6.66200647249191e-06,
      "loss": 0.0442,
      "step": 257590
    },
    {
      "epoch": 100.04,
      "learning_rate": 6.6614886731391594e-06,
      "loss": 0.0011,
      "step": 257600
    },
    {
      "epoch": 100.04,
      "learning_rate": 6.660970873786409e-06,
      "loss": 0.0644,
      "step": 257610
    },
    {
      "epoch": 100.05,
      "learning_rate": 6.660453074433657e-06,
      "loss": 0.1092,
      "step": 257620
    },
    {
      "epoch": 100.05,
      "learning_rate": 6.659935275080907e-06,
      "loss": 0.0176,
      "step": 257630
    },
    {
      "epoch": 100.05,
      "learning_rate": 6.659417475728156e-06,
      "loss": 0.1484,
      "step": 257640
    },
    {
      "epoch": 100.06,
      "learning_rate": 6.658899676375406e-06,
      "loss": 0.0423,
      "step": 257650
    },
    {
      "epoch": 100.06,
      "learning_rate": 6.658381877022654e-06,
      "loss": 0.032,
      "step": 257660
    },
    {
      "epoch": 100.07,
      "learning_rate": 6.657864077669903e-06,
      "loss": 0.0023,
      "step": 257670
    },
    {
      "epoch": 100.07,
      "learning_rate": 6.657346278317153e-06,
      "loss": 0.0026,
      "step": 257680
    },
    {
      "epoch": 100.07,
      "learning_rate": 6.656828478964403e-06,
      "loss": 0.0556,
      "step": 257690
    },
    {
      "epoch": 100.08,
      "learning_rate": 6.6563106796116505e-06,
      "loss": 0.0002,
      "step": 257700
    },
    {
      "epoch": 100.08,
      "learning_rate": 6.6557928802589e-06,
      "loss": 0.008,
      "step": 257710
    },
    {
      "epoch": 100.09,
      "learning_rate": 6.65527508090615e-06,
      "loss": 0.0809,
      "step": 257720
    },
    {
      "epoch": 100.09,
      "learning_rate": 6.6547572815533985e-06,
      "loss": 0.0174,
      "step": 257730
    },
    {
      "epoch": 100.09,
      "learning_rate": 6.654239482200647e-06,
      "loss": 0.0591,
      "step": 257740
    },
    {
      "epoch": 100.1,
      "learning_rate": 6.653721682847897e-06,
      "loss": 0.0094,
      "step": 257750
    },
    {
      "epoch": 100.1,
      "learning_rate": 6.6532038834951466e-06,
      "loss": 0.0723,
      "step": 257760
    },
    {
      "epoch": 100.1,
      "learning_rate": 6.652686084142395e-06,
      "loss": 0.0305,
      "step": 257770
    },
    {
      "epoch": 100.11,
      "learning_rate": 6.652168284789644e-06,
      "loss": 0.0409,
      "step": 257780
    },
    {
      "epoch": 100.11,
      "learning_rate": 6.651650485436894e-06,
      "loss": 0.1209,
      "step": 257790
    },
    {
      "epoch": 100.12,
      "learning_rate": 6.651132686084143e-06,
      "loss": 0.0308,
      "step": 257800
    },
    {
      "epoch": 100.12,
      "learning_rate": 6.650614886731392e-06,
      "loss": 0.1677,
      "step": 257810
    },
    {
      "epoch": 100.12,
      "learning_rate": 6.650097087378641e-06,
      "loss": 0.0111,
      "step": 257820
    },
    {
      "epoch": 100.13,
      "learning_rate": 6.6495792880258905e-06,
      "loss": 0.006,
      "step": 257830
    },
    {
      "epoch": 100.13,
      "learning_rate": 6.64906148867314e-06,
      "loss": 0.0495,
      "step": 257840
    },
    {
      "epoch": 100.14,
      "learning_rate": 6.648543689320389e-06,
      "loss": 0.0951,
      "step": 257850
    },
    {
      "epoch": 100.14,
      "learning_rate": 6.648025889967638e-06,
      "loss": 0.0869,
      "step": 257860
    },
    {
      "epoch": 100.14,
      "learning_rate": 6.647508090614887e-06,
      "loss": 0.0631,
      "step": 257870
    },
    {
      "epoch": 100.15,
      "learning_rate": 6.646990291262137e-06,
      "loss": 0.0146,
      "step": 257880
    },
    {
      "epoch": 100.15,
      "learning_rate": 6.646472491909386e-06,
      "loss": 0.1394,
      "step": 257890
    },
    {
      "epoch": 100.16,
      "learning_rate": 6.645954692556635e-06,
      "loss": 0.0651,
      "step": 257900
    },
    {
      "epoch": 100.16,
      "learning_rate": 6.645436893203884e-06,
      "loss": 0.0429,
      "step": 257910
    },
    {
      "epoch": 100.16,
      "learning_rate": 6.644919093851133e-06,
      "loss": 0.0071,
      "step": 257920
    },
    {
      "epoch": 100.17,
      "learning_rate": 6.6444012944983824e-06,
      "loss": 0.034,
      "step": 257930
    },
    {
      "epoch": 100.17,
      "learning_rate": 6.643883495145632e-06,
      "loss": 0.0249,
      "step": 257940
    },
    {
      "epoch": 100.17,
      "learning_rate": 6.643365695792881e-06,
      "loss": 0.0549,
      "step": 257950
    },
    {
      "epoch": 100.18,
      "learning_rate": 6.64284789644013e-06,
      "loss": 0.0049,
      "step": 257960
    },
    {
      "epoch": 100.18,
      "learning_rate": 6.642330097087379e-06,
      "loss": 0.0825,
      "step": 257970
    },
    {
      "epoch": 100.19,
      "learning_rate": 6.641812297734629e-06,
      "loss": 0.0631,
      "step": 257980
    },
    {
      "epoch": 100.19,
      "learning_rate": 6.641294498381878e-06,
      "loss": 0.0599,
      "step": 257990
    },
    {
      "epoch": 100.19,
      "learning_rate": 6.640776699029126e-06,
      "loss": 0.0703,
      "step": 258000
    },
    {
      "epoch": 100.2,
      "learning_rate": 6.640258899676376e-06,
      "loss": 0.0203,
      "step": 258010
    },
    {
      "epoch": 100.2,
      "learning_rate": 6.639741100323626e-06,
      "loss": 0.0309,
      "step": 258020
    },
    {
      "epoch": 100.21,
      "learning_rate": 6.639223300970874e-06,
      "loss": 0.1547,
      "step": 258030
    },
    {
      "epoch": 100.21,
      "learning_rate": 6.638705501618123e-06,
      "loss": 0.0331,
      "step": 258040
    },
    {
      "epoch": 100.21,
      "learning_rate": 6.638187702265373e-06,
      "loss": 0.1308,
      "step": 258050
    },
    {
      "epoch": 100.22,
      "learning_rate": 6.637669902912622e-06,
      "loss": 0.0087,
      "step": 258060
    },
    {
      "epoch": 100.22,
      "learning_rate": 6.63715210355987e-06,
      "loss": 0.0462,
      "step": 258070
    },
    {
      "epoch": 100.23,
      "learning_rate": 6.63663430420712e-06,
      "loss": 0.0108,
      "step": 258080
    },
    {
      "epoch": 100.23,
      "learning_rate": 6.6361165048543695e-06,
      "loss": 0.0013,
      "step": 258090
    },
    {
      "epoch": 100.23,
      "learning_rate": 6.635598705501619e-06,
      "loss": 0.0728,
      "step": 258100
    },
    {
      "epoch": 100.24,
      "learning_rate": 6.635080906148867e-06,
      "loss": 0.0741,
      "step": 258110
    },
    {
      "epoch": 100.24,
      "learning_rate": 6.634563106796117e-06,
      "loss": 0.072,
      "step": 258120
    },
    {
      "epoch": 100.24,
      "learning_rate": 6.634045307443366e-06,
      "loss": 0.0549,
      "step": 258130
    },
    {
      "epoch": 100.25,
      "learning_rate": 6.633527508090616e-06,
      "loss": 0.0001,
      "step": 258140
    },
    {
      "epoch": 100.25,
      "learning_rate": 6.633009708737864e-06,
      "loss": 0.0572,
      "step": 258150
    },
    {
      "epoch": 100.26,
      "learning_rate": 6.6324919093851135e-06,
      "loss": 0.0986,
      "step": 258160
    },
    {
      "epoch": 100.26,
      "learning_rate": 6.631974110032363e-06,
      "loss": 0.0581,
      "step": 258170
    },
    {
      "epoch": 100.26,
      "learning_rate": 6.631456310679613e-06,
      "loss": 0.1109,
      "step": 258180
    },
    {
      "epoch": 100.27,
      "learning_rate": 6.630938511326861e-06,
      "loss": 0.001,
      "step": 258190
    },
    {
      "epoch": 100.27,
      "learning_rate": 6.63042071197411e-06,
      "loss": 0.0699,
      "step": 258200
    },
    {
      "epoch": 100.28,
      "learning_rate": 6.62990291262136e-06,
      "loss": 0.0379,
      "step": 258210
    },
    {
      "epoch": 100.28,
      "learning_rate": 6.6293851132686095e-06,
      "loss": 0.0303,
      "step": 258220
    },
    {
      "epoch": 100.28,
      "learning_rate": 6.6288673139158574e-06,
      "loss": 0.0425,
      "step": 258230
    },
    {
      "epoch": 100.29,
      "learning_rate": 6.628349514563107e-06,
      "loss": 0.0006,
      "step": 258240
    },
    {
      "epoch": 100.29,
      "learning_rate": 6.627831715210357e-06,
      "loss": 0.0973,
      "step": 258250
    },
    {
      "epoch": 100.3,
      "learning_rate": 6.627313915857606e-06,
      "loss": 0.0348,
      "step": 258260
    },
    {
      "epoch": 100.3,
      "learning_rate": 6.626796116504854e-06,
      "loss": 0.0489,
      "step": 258270
    },
    {
      "epoch": 100.3,
      "learning_rate": 6.626278317152104e-06,
      "loss": 0.0321,
      "step": 258280
    },
    {
      "epoch": 100.31,
      "learning_rate": 6.6257605177993534e-06,
      "loss": 0.1461,
      "step": 258290
    },
    {
      "epoch": 100.31,
      "learning_rate": 6.625242718446603e-06,
      "loss": 0.074,
      "step": 258300
    },
    {
      "epoch": 100.31,
      "learning_rate": 6.624724919093851e-06,
      "loss": 0.0729,
      "step": 258310
    },
    {
      "epoch": 100.32,
      "learning_rate": 6.624207119741101e-06,
      "loss": 0.0245,
      "step": 258320
    },
    {
      "epoch": 100.32,
      "learning_rate": 6.62368932038835e-06,
      "loss": 0.0669,
      "step": 258330
    },
    {
      "epoch": 100.33,
      "learning_rate": 6.6231715210356e-06,
      "loss": 0.0654,
      "step": 258340
    },
    {
      "epoch": 100.33,
      "learning_rate": 6.622653721682848e-06,
      "loss": 0.0088,
      "step": 258350
    },
    {
      "epoch": 100.33,
      "learning_rate": 6.622135922330097e-06,
      "loss": 0.0343,
      "step": 258360
    },
    {
      "epoch": 100.34,
      "learning_rate": 6.621618122977347e-06,
      "loss": 0.0964,
      "step": 258370
    },
    {
      "epoch": 100.34,
      "learning_rate": 6.621100323624597e-06,
      "loss": 0.0271,
      "step": 258380
    },
    {
      "epoch": 100.35,
      "learning_rate": 6.6205825242718445e-06,
      "loss": 0.0734,
      "step": 258390
    },
    {
      "epoch": 100.35,
      "learning_rate": 6.620064724919094e-06,
      "loss": 0.125,
      "step": 258400
    },
    {
      "epoch": 100.35,
      "learning_rate": 6.619546925566344e-06,
      "loss": 0.0087,
      "step": 258410
    },
    {
      "epoch": 100.36,
      "learning_rate": 6.619029126213593e-06,
      "loss": 0.1169,
      "step": 258420
    },
    {
      "epoch": 100.36,
      "learning_rate": 6.618511326860841e-06,
      "loss": 0.0255,
      "step": 258430
    },
    {
      "epoch": 100.37,
      "learning_rate": 6.617993527508091e-06,
      "loss": 0.0347,
      "step": 258440
    },
    {
      "epoch": 100.37,
      "learning_rate": 6.6174757281553406e-06,
      "loss": 0.0844,
      "step": 258450
    },
    {
      "epoch": 100.37,
      "learning_rate": 6.61695792880259e-06,
      "loss": 0.014,
      "step": 258460
    },
    {
      "epoch": 100.38,
      "learning_rate": 6.616440129449839e-06,
      "loss": 0.0358,
      "step": 258470
    },
    {
      "epoch": 100.38,
      "learning_rate": 6.615922330097088e-06,
      "loss": 0.0538,
      "step": 258480
    },
    {
      "epoch": 100.38,
      "learning_rate": 6.615404530744337e-06,
      "loss": 0.0293,
      "step": 258490
    },
    {
      "epoch": 100.39,
      "learning_rate": 6.614886731391587e-06,
      "loss": 0.0487,
      "step": 258500
    },
    {
      "epoch": 100.39,
      "learning_rate": 6.614368932038836e-06,
      "loss": 0.05,
      "step": 258510
    },
    {
      "epoch": 100.4,
      "learning_rate": 6.6138511326860845e-06,
      "loss": 0.0174,
      "step": 258520
    },
    {
      "epoch": 100.4,
      "learning_rate": 6.613333333333334e-06,
      "loss": 0.0465,
      "step": 258530
    },
    {
      "epoch": 100.4,
      "learning_rate": 6.612815533980583e-06,
      "loss": 0.052,
      "step": 258540
    },
    {
      "epoch": 100.41,
      "learning_rate": 6.6122977346278325e-06,
      "loss": 0.0247,
      "step": 258550
    },
    {
      "epoch": 100.41,
      "learning_rate": 6.611779935275081e-06,
      "loss": 0.0581,
      "step": 258560
    },
    {
      "epoch": 100.42,
      "learning_rate": 6.611262135922331e-06,
      "loss": 0.0283,
      "step": 258570
    },
    {
      "epoch": 100.42,
      "learning_rate": 6.61074433656958e-06,
      "loss": 0.0264,
      "step": 258580
    },
    {
      "epoch": 100.42,
      "learning_rate": 6.610226537216829e-06,
      "loss": 0.0016,
      "step": 258590
    },
    {
      "epoch": 100.43,
      "learning_rate": 6.609708737864078e-06,
      "loss": 0.1041,
      "step": 258600
    },
    {
      "epoch": 100.43,
      "learning_rate": 6.609190938511328e-06,
      "loss": 0.1571,
      "step": 258610
    },
    {
      "epoch": 100.43,
      "learning_rate": 6.6086731391585764e-06,
      "loss": 0.0002,
      "step": 258620
    },
    {
      "epoch": 100.44,
      "learning_rate": 6.608155339805826e-06,
      "loss": 0.0557,
      "step": 258630
    },
    {
      "epoch": 100.44,
      "learning_rate": 6.607637540453075e-06,
      "loss": 0.1335,
      "step": 258640
    },
    {
      "epoch": 100.45,
      "learning_rate": 6.6071197411003244e-06,
      "loss": 0.0007,
      "step": 258650
    },
    {
      "epoch": 100.45,
      "learning_rate": 6.606601941747573e-06,
      "loss": 0.0814,
      "step": 258660
    },
    {
      "epoch": 100.45,
      "learning_rate": 6.606084142394823e-06,
      "loss": 0.0422,
      "step": 258670
    },
    {
      "epoch": 100.46,
      "learning_rate": 6.605566343042072e-06,
      "loss": 0.1398,
      "step": 258680
    },
    {
      "epoch": 100.46,
      "learning_rate": 6.605048543689321e-06,
      "loss": 0.0104,
      "step": 258690
    },
    {
      "epoch": 100.47,
      "learning_rate": 6.60453074433657e-06,
      "loss": 0.0812,
      "step": 258700
    },
    {
      "epoch": 100.47,
      "learning_rate": 6.60401294498382e-06,
      "loss": 0.0538,
      "step": 258710
    },
    {
      "epoch": 100.47,
      "learning_rate": 6.603495145631068e-06,
      "loss": 0.0881,
      "step": 258720
    },
    {
      "epoch": 100.48,
      "learning_rate": 6.602977346278317e-06,
      "loss": 0.1046,
      "step": 258730
    },
    {
      "epoch": 100.48,
      "learning_rate": 6.602459546925567e-06,
      "loss": 0.1103,
      "step": 258740
    },
    {
      "epoch": 100.49,
      "learning_rate": 6.601941747572816e-06,
      "loss": 0.0117,
      "step": 258750
    },
    {
      "epoch": 100.49,
      "learning_rate": 6.601423948220065e-06,
      "loss": 0.0934,
      "step": 258760
    },
    {
      "epoch": 100.49,
      "learning_rate": 6.600906148867314e-06,
      "loss": 0.0576,
      "step": 258770
    },
    {
      "epoch": 100.5,
      "learning_rate": 6.6003883495145635e-06,
      "loss": 0.0319,
      "step": 258780
    },
    {
      "epoch": 100.5,
      "learning_rate": 6.599870550161813e-06,
      "loss": 0.0234,
      "step": 258790
    },
    {
      "epoch": 100.5,
      "learning_rate": 6.599352750809062e-06,
      "loss": 0.0541,
      "step": 258800
    },
    {
      "epoch": 100.51,
      "learning_rate": 6.598834951456311e-06,
      "loss": 0.0048,
      "step": 258810
    },
    {
      "epoch": 100.51,
      "learning_rate": 6.59831715210356e-06,
      "loss": 0.0596,
      "step": 258820
    },
    {
      "epoch": 100.52,
      "learning_rate": 6.59779935275081e-06,
      "loss": 0.0295,
      "step": 258830
    },
    {
      "epoch": 100.52,
      "learning_rate": 6.597281553398059e-06,
      "loss": 0.0008,
      "step": 258840
    },
    {
      "epoch": 100.52,
      "learning_rate": 6.5967637540453075e-06,
      "loss": 0.0562,
      "step": 258850
    },
    {
      "epoch": 100.53,
      "learning_rate": 6.596245954692557e-06,
      "loss": 0.0185,
      "step": 258860
    },
    {
      "epoch": 100.53,
      "learning_rate": 6.595728155339807e-06,
      "loss": 0.0597,
      "step": 258870
    },
    {
      "epoch": 100.54,
      "learning_rate": 6.595210355987055e-06,
      "loss": 0.0002,
      "step": 258880
    },
    {
      "epoch": 100.54,
      "learning_rate": 6.594692556634304e-06,
      "loss": 0.0612,
      "step": 258890
    },
    {
      "epoch": 100.54,
      "learning_rate": 6.594174757281554e-06,
      "loss": 0.0883,
      "step": 258900
    },
    {
      "epoch": 100.55,
      "learning_rate": 6.5936569579288035e-06,
      "loss": 0.0046,
      "step": 258910
    },
    {
      "epoch": 100.55,
      "learning_rate": 6.593139158576051e-06,
      "loss": 0.0002,
      "step": 258920
    },
    {
      "epoch": 100.56,
      "learning_rate": 6.592621359223301e-06,
      "loss": 0.0116,
      "step": 258930
    },
    {
      "epoch": 100.56,
      "learning_rate": 6.592103559870551e-06,
      "loss": 0.1956,
      "step": 258940
    },
    {
      "epoch": 100.56,
      "learning_rate": 6.5915857605178e-06,
      "loss": 0.0337,
      "step": 258950
    },
    {
      "epoch": 100.57,
      "learning_rate": 6.591067961165048e-06,
      "loss": 0.111,
      "step": 258960
    },
    {
      "epoch": 100.57,
      "learning_rate": 6.590550161812298e-06,
      "loss": 0.003,
      "step": 258970
    },
    {
      "epoch": 100.57,
      "learning_rate": 6.5900323624595474e-06,
      "loss": 0.1574,
      "step": 258980
    },
    {
      "epoch": 100.58,
      "learning_rate": 6.589514563106797e-06,
      "loss": 0.0117,
      "step": 258990
    },
    {
      "epoch": 100.58,
      "learning_rate": 6.588996763754047e-06,
      "loss": 0.0059,
      "step": 259000
    },
    {
      "epoch": 100.59,
      "learning_rate": 6.588478964401295e-06,
      "loss": 0.0266,
      "step": 259010
    },
    {
      "epoch": 100.59,
      "learning_rate": 6.587961165048544e-06,
      "loss": 0.0228,
      "step": 259020
    },
    {
      "epoch": 100.59,
      "learning_rate": 6.587443365695794e-06,
      "loss": 0.1159,
      "step": 259030
    },
    {
      "epoch": 100.6,
      "learning_rate": 6.5869255663430435e-06,
      "loss": 0.1334,
      "step": 259040
    },
    {
      "epoch": 100.6,
      "learning_rate": 6.586407766990291e-06,
      "loss": 0.0112,
      "step": 259050
    },
    {
      "epoch": 100.61,
      "learning_rate": 6.585889967637541e-06,
      "loss": 0.0811,
      "step": 259060
    },
    {
      "epoch": 100.61,
      "learning_rate": 6.585372168284791e-06,
      "loss": 0.014,
      "step": 259070
    },
    {
      "epoch": 100.61,
      "learning_rate": 6.58485436893204e-06,
      "loss": 0.1534,
      "step": 259080
    },
    {
      "epoch": 100.62,
      "learning_rate": 6.584336569579288e-06,
      "loss": 0.0251,
      "step": 259090
    },
    {
      "epoch": 100.62,
      "learning_rate": 6.583818770226538e-06,
      "loss": 0.0001,
      "step": 259100
    },
    {
      "epoch": 100.63,
      "learning_rate": 6.583300970873787e-06,
      "loss": 0.068,
      "step": 259110
    },
    {
      "epoch": 100.63,
      "learning_rate": 6.582783171521037e-06,
      "loss": 0.0035,
      "step": 259120
    },
    {
      "epoch": 100.63,
      "learning_rate": 6.582265372168285e-06,
      "loss": 0.076,
      "step": 259130
    },
    {
      "epoch": 100.64,
      "learning_rate": 6.5817475728155345e-06,
      "loss": 0.0814,
      "step": 259140
    },
    {
      "epoch": 100.64,
      "learning_rate": 6.581229773462784e-06,
      "loss": 0.0164,
      "step": 259150
    },
    {
      "epoch": 100.64,
      "learning_rate": 6.580711974110034e-06,
      "loss": 0.0636,
      "step": 259160
    },
    {
      "epoch": 100.65,
      "learning_rate": 6.580194174757282e-06,
      "loss": 0.0896,
      "step": 259170
    },
    {
      "epoch": 100.65,
      "learning_rate": 6.579676375404531e-06,
      "loss": 0.1134,
      "step": 259180
    },
    {
      "epoch": 100.66,
      "learning_rate": 6.579158576051781e-06,
      "loss": 0.0006,
      "step": 259190
    },
    {
      "epoch": 100.66,
      "learning_rate": 6.57864077669903e-06,
      "loss": 0.0581,
      "step": 259200
    },
    {
      "epoch": 100.66,
      "learning_rate": 6.5781229773462785e-06,
      "loss": 0.0419,
      "step": 259210
    },
    {
      "epoch": 100.67,
      "learning_rate": 6.577605177993528e-06,
      "loss": 0.0681,
      "step": 259220
    },
    {
      "epoch": 100.67,
      "learning_rate": 6.577087378640778e-06,
      "loss": 0.1661,
      "step": 259230
    },
    {
      "epoch": 100.68,
      "learning_rate": 6.5765695792880265e-06,
      "loss": 0.0506,
      "step": 259240
    },
    {
      "epoch": 100.68,
      "learning_rate": 6.576051779935275e-06,
      "loss": 0.0099,
      "step": 259250
    },
    {
      "epoch": 100.68,
      "learning_rate": 6.575533980582525e-06,
      "loss": 0.0156,
      "step": 259260
    },
    {
      "epoch": 100.69,
      "learning_rate": 6.5750161812297745e-06,
      "loss": 0.0246,
      "step": 259270
    },
    {
      "epoch": 100.69,
      "learning_rate": 6.574498381877023e-06,
      "loss": 0.0429,
      "step": 259280
    },
    {
      "epoch": 100.7,
      "learning_rate": 6.573980582524272e-06,
      "loss": 0.1408,
      "step": 259290
    },
    {
      "epoch": 100.7,
      "learning_rate": 6.573462783171522e-06,
      "loss": 0.0536,
      "step": 259300
    },
    {
      "epoch": 100.7,
      "learning_rate": 6.572944983818771e-06,
      "loss": 0.1049,
      "step": 259310
    },
    {
      "epoch": 100.71,
      "learning_rate": 6.57242718446602e-06,
      "loss": 0.0055,
      "step": 259320
    },
    {
      "epoch": 100.71,
      "learning_rate": 6.571909385113269e-06,
      "loss": 0.0001,
      "step": 259330
    },
    {
      "epoch": 100.71,
      "learning_rate": 6.5713915857605184e-06,
      "loss": 0.0001,
      "step": 259340
    },
    {
      "epoch": 100.72,
      "learning_rate": 6.570873786407767e-06,
      "loss": 0.0684,
      "step": 259350
    },
    {
      "epoch": 100.72,
      "learning_rate": 6.570355987055017e-06,
      "loss": 0.0013,
      "step": 259360
    },
    {
      "epoch": 100.73,
      "learning_rate": 6.569838187702266e-06,
      "loss": 0.0286,
      "step": 259370
    },
    {
      "epoch": 100.73,
      "learning_rate": 6.569320388349515e-06,
      "loss": 0.0262,
      "step": 259380
    },
    {
      "epoch": 100.73,
      "learning_rate": 6.568802588996764e-06,
      "loss": 0.1304,
      "step": 259390
    },
    {
      "epoch": 100.74,
      "learning_rate": 6.568284789644014e-06,
      "loss": 0.0857,
      "step": 259400
    },
    {
      "epoch": 100.74,
      "learning_rate": 6.567766990291262e-06,
      "loss": 0.003,
      "step": 259410
    },
    {
      "epoch": 100.75,
      "learning_rate": 6.567249190938512e-06,
      "loss": 0.0319,
      "step": 259420
    },
    {
      "epoch": 100.75,
      "learning_rate": 6.566731391585761e-06,
      "loss": 0.1103,
      "step": 259430
    },
    {
      "epoch": 100.75,
      "learning_rate": 6.56621359223301e-06,
      "loss": 0.0027,
      "step": 259440
    },
    {
      "epoch": 100.76,
      "learning_rate": 6.565695792880259e-06,
      "loss": 0.0349,
      "step": 259450
    },
    {
      "epoch": 100.76,
      "learning_rate": 6.565177993527509e-06,
      "loss": 0.0458,
      "step": 259460
    },
    {
      "epoch": 100.77,
      "learning_rate": 6.5646601941747575e-06,
      "loss": 0.0272,
      "step": 259470
    },
    {
      "epoch": 100.77,
      "learning_rate": 6.564142394822007e-06,
      "loss": 0.0027,
      "step": 259480
    },
    {
      "epoch": 100.77,
      "learning_rate": 6.563624595469256e-06,
      "loss": 0.0578,
      "step": 259490
    },
    {
      "epoch": 100.78,
      "learning_rate": 6.5631067961165056e-06,
      "loss": 0.0938,
      "step": 259500
    },
    {
      "epoch": 100.78,
      "learning_rate": 6.562588996763754e-06,
      "loss": 0.0548,
      "step": 259510
    },
    {
      "epoch": 100.78,
      "learning_rate": 6.562071197411004e-06,
      "loss": 0.107,
      "step": 259520
    },
    {
      "epoch": 100.79,
      "learning_rate": 6.561553398058253e-06,
      "loss": 0.0128,
      "step": 259530
    },
    {
      "epoch": 100.79,
      "learning_rate": 6.5610355987055015e-06,
      "loss": 0.0516,
      "step": 259540
    },
    {
      "epoch": 100.8,
      "learning_rate": 6.560517799352751e-06,
      "loss": 0.1079,
      "step": 259550
    },
    {
      "epoch": 100.8,
      "learning_rate": 6.560000000000001e-06,
      "loss": 0.0758,
      "step": 259560
    },
    {
      "epoch": 100.8,
      "learning_rate": 6.55948220064725e-06,
      "loss": 0.0125,
      "step": 259570
    },
    {
      "epoch": 100.81,
      "learning_rate": 6.558964401294498e-06,
      "loss": 0.0321,
      "step": 259580
    },
    {
      "epoch": 100.81,
      "learning_rate": 6.558446601941748e-06,
      "loss": 0.0373,
      "step": 259590
    },
    {
      "epoch": 100.82,
      "learning_rate": 6.5579288025889975e-06,
      "loss": 0.0557,
      "step": 259600
    },
    {
      "epoch": 100.82,
      "learning_rate": 6.557411003236247e-06,
      "loss": 0.0451,
      "step": 259610
    },
    {
      "epoch": 100.82,
      "learning_rate": 6.556893203883495e-06,
      "loss": 0.0065,
      "step": 259620
    },
    {
      "epoch": 100.83,
      "learning_rate": 6.556375404530745e-06,
      "loss": 0.0052,
      "step": 259630
    },
    {
      "epoch": 100.83,
      "learning_rate": 6.555857605177994e-06,
      "loss": 0.0098,
      "step": 259640
    },
    {
      "epoch": 100.83,
      "learning_rate": 6.555339805825244e-06,
      "loss": 0.0198,
      "step": 259650
    },
    {
      "epoch": 100.84,
      "learning_rate": 6.554822006472492e-06,
      "loss": 0.0756,
      "step": 259660
    },
    {
      "epoch": 100.84,
      "learning_rate": 6.5543042071197414e-06,
      "loss": 0.0005,
      "step": 259670
    },
    {
      "epoch": 100.85,
      "learning_rate": 6.553786407766991e-06,
      "loss": 0.0655,
      "step": 259680
    },
    {
      "epoch": 100.85,
      "learning_rate": 6.553268608414241e-06,
      "loss": 0.0127,
      "step": 259690
    },
    {
      "epoch": 100.85,
      "learning_rate": 6.552750809061489e-06,
      "loss": 0.0925,
      "step": 259700
    },
    {
      "epoch": 100.86,
      "learning_rate": 6.552233009708738e-06,
      "loss": 0.0102,
      "step": 259710
    },
    {
      "epoch": 100.86,
      "learning_rate": 6.551715210355988e-06,
      "loss": 0.0392,
      "step": 259720
    },
    {
      "epoch": 100.87,
      "learning_rate": 6.5511974110032375e-06,
      "loss": 0.0585,
      "step": 259730
    },
    {
      "epoch": 100.87,
      "learning_rate": 6.550679611650485e-06,
      "loss": 0.219,
      "step": 259740
    },
    {
      "epoch": 100.87,
      "learning_rate": 6.550161812297735e-06,
      "loss": 0.0001,
      "step": 259750
    },
    {
      "epoch": 100.88,
      "learning_rate": 6.549644012944985e-06,
      "loss": 0.0187,
      "step": 259760
    },
    {
      "epoch": 100.88,
      "learning_rate": 6.549126213592234e-06,
      "loss": 0.0712,
      "step": 259770
    },
    {
      "epoch": 100.89,
      "learning_rate": 6.548608414239482e-06,
      "loss": 0.0862,
      "step": 259780
    },
    {
      "epoch": 100.89,
      "learning_rate": 6.548090614886732e-06,
      "loss": 0.0591,
      "step": 259790
    },
    {
      "epoch": 100.89,
      "learning_rate": 6.547572815533981e-06,
      "loss": 0.0105,
      "step": 259800
    },
    {
      "epoch": 100.9,
      "learning_rate": 6.547055016181231e-06,
      "loss": 0.0107,
      "step": 259810
    },
    {
      "epoch": 100.9,
      "learning_rate": 6.546537216828479e-06,
      "loss": 0.0348,
      "step": 259820
    },
    {
      "epoch": 100.9,
      "learning_rate": 6.5460194174757285e-06,
      "loss": 0.0984,
      "step": 259830
    },
    {
      "epoch": 100.91,
      "learning_rate": 6.545501618122978e-06,
      "loss": 0.0028,
      "step": 259840
    },
    {
      "epoch": 100.91,
      "learning_rate": 6.544983818770228e-06,
      "loss": 0.0656,
      "step": 259850
    },
    {
      "epoch": 100.92,
      "learning_rate": 6.544466019417476e-06,
      "loss": 0.1759,
      "step": 259860
    },
    {
      "epoch": 100.92,
      "learning_rate": 6.543948220064725e-06,
      "loss": 0.1096,
      "step": 259870
    },
    {
      "epoch": 100.92,
      "learning_rate": 6.543430420711975e-06,
      "loss": 0.0824,
      "step": 259880
    },
    {
      "epoch": 100.93,
      "learning_rate": 6.5429126213592246e-06,
      "loss": 0.0714,
      "step": 259890
    },
    {
      "epoch": 100.93,
      "learning_rate": 6.5423948220064725e-06,
      "loss": 0.0373,
      "step": 259900
    },
    {
      "epoch": 100.94,
      "learning_rate": 6.541877022653722e-06,
      "loss": 0.0671,
      "step": 259910
    },
    {
      "epoch": 100.94,
      "learning_rate": 6.541359223300972e-06,
      "loss": 0.0243,
      "step": 259920
    },
    {
      "epoch": 100.94,
      "learning_rate": 6.540841423948221e-06,
      "loss": 0.001,
      "step": 259930
    },
    {
      "epoch": 100.95,
      "learning_rate": 6.540323624595469e-06,
      "loss": 0.0847,
      "step": 259940
    },
    {
      "epoch": 100.95,
      "learning_rate": 6.539805825242719e-06,
      "loss": 0.0469,
      "step": 259950
    },
    {
      "epoch": 100.96,
      "learning_rate": 6.5392880258899685e-06,
      "loss": 0.039,
      "step": 259960
    },
    {
      "epoch": 100.96,
      "learning_rate": 6.538770226537218e-06,
      "loss": 0.0041,
      "step": 259970
    },
    {
      "epoch": 100.96,
      "learning_rate": 6.538252427184466e-06,
      "loss": 0.0717,
      "step": 259980
    },
    {
      "epoch": 100.97,
      "learning_rate": 6.537734627831716e-06,
      "loss": 0.0019,
      "step": 259990
    },
    {
      "epoch": 100.97,
      "learning_rate": 6.537216828478965e-06,
      "loss": 0.0039,
      "step": 260000
    },
    {
      "epoch": 100.97,
      "learning_rate": 6.536699029126214e-06,
      "loss": 0.0344,
      "step": 260010
    },
    {
      "epoch": 100.98,
      "learning_rate": 6.536181229773463e-06,
      "loss": 0.0943,
      "step": 260020
    },
    {
      "epoch": 100.98,
      "learning_rate": 6.5356634304207124e-06,
      "loss": 0.0077,
      "step": 260030
    },
    {
      "epoch": 100.99,
      "learning_rate": 6.535145631067962e-06,
      "loss": 0.1502,
      "step": 260040
    },
    {
      "epoch": 100.99,
      "learning_rate": 6.534627831715211e-06,
      "loss": 0.1233,
      "step": 260050
    },
    {
      "epoch": 100.99,
      "learning_rate": 6.53411003236246e-06,
      "loss": 0.0772,
      "step": 260060
    },
    {
      "epoch": 101.0,
      "learning_rate": 6.533592233009709e-06,
      "loss": 0.0951,
      "step": 260070
    },
    {
      "epoch": 101.0,
      "eval_accuracy": 0.9491059147180193,
      "eval_loss": 0.3718360364437103,
      "eval_runtime": 8.2256,
      "eval_samples_per_second": 441.912,
      "eval_steps_per_second": 55.315,
      "step": 260075
    },
    {
      "epoch": 101.0,
      "learning_rate": 6.533074433656959e-06,
      "loss": 0.0625,
      "step": 260080
    },
    {
      "epoch": 101.01,
      "learning_rate": 6.532556634304208e-06,
      "loss": 0.2079,
      "step": 260090
    },
    {
      "epoch": 101.01,
      "learning_rate": 6.532038834951456e-06,
      "loss": 0.0924,
      "step": 260100
    },
    {
      "epoch": 101.01,
      "learning_rate": 6.531521035598706e-06,
      "loss": 0.1342,
      "step": 260110
    },
    {
      "epoch": 101.02,
      "learning_rate": 6.531003236245956e-06,
      "loss": 0.0001,
      "step": 260120
    },
    {
      "epoch": 101.02,
      "learning_rate": 6.530485436893204e-06,
      "loss": 0.0028,
      "step": 260130
    },
    {
      "epoch": 101.03,
      "learning_rate": 6.529967637540454e-06,
      "loss": 0.0665,
      "step": 260140
    },
    {
      "epoch": 101.03,
      "learning_rate": 6.529449838187703e-06,
      "loss": 0.0387,
      "step": 260150
    },
    {
      "epoch": 101.03,
      "learning_rate": 6.528932038834952e-06,
      "loss": 0.0073,
      "step": 260160
    },
    {
      "epoch": 101.04,
      "learning_rate": 6.528414239482201e-06,
      "loss": 0.0611,
      "step": 260170
    },
    {
      "epoch": 101.04,
      "learning_rate": 6.527896440129451e-06,
      "loss": 0.0693,
      "step": 260180
    },
    {
      "epoch": 101.04,
      "learning_rate": 6.5273786407766996e-06,
      "loss": 0.0398,
      "step": 260190
    },
    {
      "epoch": 101.05,
      "learning_rate": 6.526860841423948e-06,
      "loss": 0.1098,
      "step": 260200
    },
    {
      "epoch": 101.05,
      "learning_rate": 6.526343042071198e-06,
      "loss": 0.0138,
      "step": 260210
    },
    {
      "epoch": 101.06,
      "learning_rate": 6.5258252427184476e-06,
      "loss": 0.0628,
      "step": 260220
    },
    {
      "epoch": 101.06,
      "learning_rate": 6.525307443365696e-06,
      "loss": 0.1302,
      "step": 260230
    },
    {
      "epoch": 101.06,
      "learning_rate": 6.524789644012945e-06,
      "loss": 0.0366,
      "step": 260240
    },
    {
      "epoch": 101.07,
      "learning_rate": 6.524271844660195e-06,
      "loss": 0.0537,
      "step": 260250
    },
    {
      "epoch": 101.07,
      "learning_rate": 6.523754045307444e-06,
      "loss": 0.01,
      "step": 260260
    },
    {
      "epoch": 101.08,
      "learning_rate": 6.523236245954693e-06,
      "loss": 0.082,
      "step": 260270
    },
    {
      "epoch": 101.08,
      "learning_rate": 6.522718446601942e-06,
      "loss": 0.0397,
      "step": 260280
    },
    {
      "epoch": 101.08,
      "learning_rate": 6.5222006472491915e-06,
      "loss": 0.011,
      "step": 260290
    },
    {
      "epoch": 101.09,
      "learning_rate": 6.521682847896441e-06,
      "loss": 0.0161,
      "step": 260300
    },
    {
      "epoch": 101.09,
      "learning_rate": 6.52116504854369e-06,
      "loss": 0.0405,
      "step": 260310
    },
    {
      "epoch": 101.1,
      "learning_rate": 6.520647249190939e-06,
      "loss": 0.0653,
      "step": 260320
    },
    {
      "epoch": 101.1,
      "learning_rate": 6.520129449838188e-06,
      "loss": 0.0371,
      "step": 260330
    },
    {
      "epoch": 101.1,
      "learning_rate": 6.519611650485438e-06,
      "loss": 0.1711,
      "step": 260340
    },
    {
      "epoch": 101.11,
      "learning_rate": 6.519093851132686e-06,
      "loss": 0.0118,
      "step": 260350
    },
    {
      "epoch": 101.11,
      "learning_rate": 6.5185760517799354e-06,
      "loss": 0.0007,
      "step": 260360
    },
    {
      "epoch": 101.11,
      "learning_rate": 6.518058252427185e-06,
      "loss": 0.06,
      "step": 260370
    },
    {
      "epoch": 101.12,
      "learning_rate": 6.517540453074435e-06,
      "loss": 0.0146,
      "step": 260380
    },
    {
      "epoch": 101.12,
      "learning_rate": 6.517022653721683e-06,
      "loss": 0.0032,
      "step": 260390
    },
    {
      "epoch": 101.13,
      "learning_rate": 6.516504854368932e-06,
      "loss": 0.0894,
      "step": 260400
    },
    {
      "epoch": 101.13,
      "learning_rate": 6.515987055016182e-06,
      "loss": 0.137,
      "step": 260410
    },
    {
      "epoch": 101.13,
      "learning_rate": 6.5154692556634314e-06,
      "loss": 0.1694,
      "step": 260420
    },
    {
      "epoch": 101.14,
      "learning_rate": 6.514951456310679e-06,
      "loss": 0.1914,
      "step": 260430
    },
    {
      "epoch": 101.14,
      "learning_rate": 6.514433656957929e-06,
      "loss": 0.0391,
      "step": 260440
    },
    {
      "epoch": 101.15,
      "learning_rate": 6.513915857605179e-06,
      "loss": 0.0759,
      "step": 260450
    },
    {
      "epoch": 101.15,
      "learning_rate": 6.513398058252428e-06,
      "loss": 0.0417,
      "step": 260460
    },
    {
      "epoch": 101.15,
      "learning_rate": 6.512880258899676e-06,
      "loss": 0.0577,
      "step": 260470
    },
    {
      "epoch": 101.16,
      "learning_rate": 6.512362459546926e-06,
      "loss": 0.0253,
      "step": 260480
    },
    {
      "epoch": 101.16,
      "learning_rate": 6.511844660194175e-06,
      "loss": 0.0267,
      "step": 260490
    },
    {
      "epoch": 101.17,
      "learning_rate": 6.511326860841425e-06,
      "loss": 0.0193,
      "step": 260500
    },
    {
      "epoch": 101.17,
      "learning_rate": 6.510809061488673e-06,
      "loss": 0.0241,
      "step": 260510
    },
    {
      "epoch": 101.17,
      "learning_rate": 6.5102912621359225e-06,
      "loss": 0.0926,
      "step": 260520
    },
    {
      "epoch": 101.18,
      "learning_rate": 6.509773462783172e-06,
      "loss": 0.2238,
      "step": 260530
    },
    {
      "epoch": 101.18,
      "learning_rate": 6.509255663430422e-06,
      "loss": 0.0012,
      "step": 260540
    },
    {
      "epoch": 101.18,
      "learning_rate": 6.50873786407767e-06,
      "loss": 0.0644,
      "step": 260550
    },
    {
      "epoch": 101.19,
      "learning_rate": 6.508220064724919e-06,
      "loss": 0.0732,
      "step": 260560
    },
    {
      "epoch": 101.19,
      "learning_rate": 6.507702265372169e-06,
      "loss": 0.0089,
      "step": 260570
    },
    {
      "epoch": 101.2,
      "learning_rate": 6.5071844660194186e-06,
      "loss": 0.0124,
      "step": 260580
    },
    {
      "epoch": 101.2,
      "learning_rate": 6.5066666666666665e-06,
      "loss": 0.0913,
      "step": 260590
    },
    {
      "epoch": 101.2,
      "learning_rate": 6.506148867313916e-06,
      "loss": 0.027,
      "step": 260600
    },
    {
      "epoch": 101.21,
      "learning_rate": 6.505631067961166e-06,
      "loss": 0.1625,
      "step": 260610
    },
    {
      "epoch": 101.21,
      "learning_rate": 6.505113268608415e-06,
      "loss": 0.0025,
      "step": 260620
    },
    {
      "epoch": 101.22,
      "learning_rate": 6.504595469255663e-06,
      "loss": 0.0857,
      "step": 260630
    },
    {
      "epoch": 101.22,
      "learning_rate": 6.504077669902913e-06,
      "loss": 0.0254,
      "step": 260640
    },
    {
      "epoch": 101.22,
      "learning_rate": 6.5035598705501625e-06,
      "loss": 0.002,
      "step": 260650
    },
    {
      "epoch": 101.23,
      "learning_rate": 6.503042071197412e-06,
      "loss": 0.0567,
      "step": 260660
    },
    {
      "epoch": 101.23,
      "learning_rate": 6.50252427184466e-06,
      "loss": 0.0062,
      "step": 260670
    },
    {
      "epoch": 101.23,
      "learning_rate": 6.50200647249191e-06,
      "loss": 0.0776,
      "step": 260680
    },
    {
      "epoch": 101.24,
      "learning_rate": 6.501488673139159e-06,
      "loss": 0.0815,
      "step": 260690
    },
    {
      "epoch": 101.24,
      "learning_rate": 6.500970873786409e-06,
      "loss": 0.069,
      "step": 260700
    },
    {
      "epoch": 101.25,
      "learning_rate": 6.500453074433658e-06,
      "loss": 0.2334,
      "step": 260710
    },
    {
      "epoch": 101.25,
      "learning_rate": 6.4999352750809064e-06,
      "loss": 0.0279,
      "step": 260720
    },
    {
      "epoch": 101.25,
      "learning_rate": 6.499417475728156e-06,
      "loss": 0.0022,
      "step": 260730
    },
    {
      "epoch": 101.26,
      "learning_rate": 6.498899676375406e-06,
      "loss": 0.013,
      "step": 260740
    },
    {
      "epoch": 101.26,
      "learning_rate": 6.4983818770226544e-06,
      "loss": 0.0742,
      "step": 260750
    },
    {
      "epoch": 101.27,
      "learning_rate": 6.497864077669903e-06,
      "loss": 0.075,
      "step": 260760
    },
    {
      "epoch": 101.27,
      "learning_rate": 6.497346278317153e-06,
      "loss": 0.139,
      "step": 260770
    },
    {
      "epoch": 101.27,
      "learning_rate": 6.4968284789644025e-06,
      "loss": 0.0309,
      "step": 260780
    },
    {
      "epoch": 101.28,
      "learning_rate": 6.496310679611651e-06,
      "loss": 0.0562,
      "step": 260790
    },
    {
      "epoch": 101.28,
      "learning_rate": 6.4957928802589e-06,
      "loss": 0.0088,
      "step": 260800
    },
    {
      "epoch": 101.29,
      "learning_rate": 6.49527508090615e-06,
      "loss": 0.0233,
      "step": 260810
    },
    {
      "epoch": 101.29,
      "learning_rate": 6.494757281553398e-06,
      "loss": 0.0003,
      "step": 260820
    },
    {
      "epoch": 101.29,
      "learning_rate": 6.494239482200648e-06,
      "loss": 0.0686,
      "step": 260830
    },
    {
      "epoch": 101.3,
      "learning_rate": 6.493721682847897e-06,
      "loss": 0.001,
      "step": 260840
    },
    {
      "epoch": 101.3,
      "learning_rate": 6.493203883495146e-06,
      "loss": 0.0002,
      "step": 260850
    },
    {
      "epoch": 101.3,
      "learning_rate": 6.492686084142395e-06,
      "loss": 0.0517,
      "step": 260860
    },
    {
      "epoch": 101.31,
      "learning_rate": 6.492168284789645e-06,
      "loss": 0.0898,
      "step": 260870
    },
    {
      "epoch": 101.31,
      "learning_rate": 6.4916504854368935e-06,
      "loss": 0.1048,
      "step": 260880
    },
    {
      "epoch": 101.32,
      "learning_rate": 6.491132686084143e-06,
      "loss": 0.0551,
      "step": 260890
    },
    {
      "epoch": 101.32,
      "learning_rate": 6.490614886731392e-06,
      "loss": 0.0155,
      "step": 260900
    },
    {
      "epoch": 101.32,
      "learning_rate": 6.4900970873786416e-06,
      "loss": 0.0742,
      "step": 260910
    },
    {
      "epoch": 101.33,
      "learning_rate": 6.48957928802589e-06,
      "loss": 0.0148,
      "step": 260920
    },
    {
      "epoch": 101.33,
      "learning_rate": 6.48906148867314e-06,
      "loss": 0.0128,
      "step": 260930
    },
    {
      "epoch": 101.34,
      "learning_rate": 6.488543689320389e-06,
      "loss": 0.064,
      "step": 260940
    },
    {
      "epoch": 101.34,
      "learning_rate": 6.488025889967638e-06,
      "loss": 0.1186,
      "step": 260950
    },
    {
      "epoch": 101.34,
      "learning_rate": 6.487508090614887e-06,
      "loss": 0.0344,
      "step": 260960
    },
    {
      "epoch": 101.35,
      "learning_rate": 6.486990291262137e-06,
      "loss": 0.0011,
      "step": 260970
    },
    {
      "epoch": 101.35,
      "learning_rate": 6.4864724919093855e-06,
      "loss": 0.1031,
      "step": 260980
    },
    {
      "epoch": 101.36,
      "learning_rate": 6.485954692556635e-06,
      "loss": 0.0837,
      "step": 260990
    },
    {
      "epoch": 101.36,
      "learning_rate": 6.485436893203884e-06,
      "loss": 0.031,
      "step": 261000
    },
    {
      "epoch": 101.36,
      "learning_rate": 6.484919093851133e-06,
      "loss": 0.0005,
      "step": 261010
    },
    {
      "epoch": 101.37,
      "learning_rate": 6.484401294498382e-06,
      "loss": 0.0142,
      "step": 261020
    },
    {
      "epoch": 101.37,
      "learning_rate": 6.483883495145632e-06,
      "loss": 0.0277,
      "step": 261030
    },
    {
      "epoch": 101.37,
      "learning_rate": 6.483365695792881e-06,
      "loss": 0.0152,
      "step": 261040
    },
    {
      "epoch": 101.38,
      "learning_rate": 6.4828478964401294e-06,
      "loss": 0.0053,
      "step": 261050
    },
    {
      "epoch": 101.38,
      "learning_rate": 6.482330097087379e-06,
      "loss": 0.0352,
      "step": 261060
    },
    {
      "epoch": 101.39,
      "learning_rate": 6.481812297734629e-06,
      "loss": 0.0566,
      "step": 261070
    },
    {
      "epoch": 101.39,
      "learning_rate": 6.4812944983818774e-06,
      "loss": 0.1071,
      "step": 261080
    },
    {
      "epoch": 101.39,
      "learning_rate": 6.480776699029126e-06,
      "loss": 0.0467,
      "step": 261090
    },
    {
      "epoch": 101.4,
      "learning_rate": 6.480258899676376e-06,
      "loss": 0.0385,
      "step": 261100
    },
    {
      "epoch": 101.4,
      "learning_rate": 6.4797411003236254e-06,
      "loss": 0.0001,
      "step": 261110
    },
    {
      "epoch": 101.41,
      "learning_rate": 6.479223300970874e-06,
      "loss": 0.0005,
      "step": 261120
    },
    {
      "epoch": 101.41,
      "learning_rate": 6.478705501618123e-06,
      "loss": 0.0871,
      "step": 261130
    },
    {
      "epoch": 101.41,
      "learning_rate": 6.478187702265373e-06,
      "loss": 0.0675,
      "step": 261140
    },
    {
      "epoch": 101.42,
      "learning_rate": 6.477669902912622e-06,
      "loss": 0.0334,
      "step": 261150
    },
    {
      "epoch": 101.42,
      "learning_rate": 6.47715210355987e-06,
      "loss": 0.1113,
      "step": 261160
    },
    {
      "epoch": 101.43,
      "learning_rate": 6.47663430420712e-06,
      "loss": 0.0889,
      "step": 261170
    },
    {
      "epoch": 101.43,
      "learning_rate": 6.476116504854369e-06,
      "loss": 0.0134,
      "step": 261180
    },
    {
      "epoch": 101.43,
      "learning_rate": 6.475598705501619e-06,
      "loss": 0.0553,
      "step": 261190
    },
    {
      "epoch": 101.44,
      "learning_rate": 6.475080906148867e-06,
      "loss": 0.0338,
      "step": 261200
    },
    {
      "epoch": 101.44,
      "learning_rate": 6.4745631067961165e-06,
      "loss": 0.0485,
      "step": 261210
    },
    {
      "epoch": 101.44,
      "learning_rate": 6.474045307443366e-06,
      "loss": 0.0078,
      "step": 261220
    },
    {
      "epoch": 101.45,
      "learning_rate": 6.473527508090616e-06,
      "loss": 0.0388,
      "step": 261230
    },
    {
      "epoch": 101.45,
      "learning_rate": 6.473009708737865e-06,
      "loss": 0.0339,
      "step": 261240
    },
    {
      "epoch": 101.46,
      "learning_rate": 6.472491909385113e-06,
      "loss": 0.0001,
      "step": 261250
    },
    {
      "epoch": 101.46,
      "learning_rate": 6.471974110032363e-06,
      "loss": 0.0613,
      "step": 261260
    },
    {
      "epoch": 101.46,
      "learning_rate": 6.4714563106796126e-06,
      "loss": 0.052,
      "step": 261270
    },
    {
      "epoch": 101.47,
      "learning_rate": 6.470938511326862e-06,
      "loss": 0.0125,
      "step": 261280
    },
    {
      "epoch": 101.47,
      "learning_rate": 6.47042071197411e-06,
      "loss": 0.0283,
      "step": 261290
    },
    {
      "epoch": 101.48,
      "learning_rate": 6.46990291262136e-06,
      "loss": 0.0268,
      "step": 261300
    },
    {
      "epoch": 101.48,
      "learning_rate": 6.469385113268609e-06,
      "loss": 0.0744,
      "step": 261310
    },
    {
      "epoch": 101.48,
      "learning_rate": 6.468867313915859e-06,
      "loss": 0.0133,
      "step": 261320
    },
    {
      "epoch": 101.49,
      "learning_rate": 6.468349514563107e-06,
      "loss": 0.0157,
      "step": 261330
    },
    {
      "epoch": 101.49,
      "learning_rate": 6.4678317152103565e-06,
      "loss": 0.0004,
      "step": 261340
    },
    {
      "epoch": 101.5,
      "learning_rate": 6.467313915857606e-06,
      "loss": 0.0636,
      "step": 261350
    },
    {
      "epoch": 101.5,
      "learning_rate": 6.466796116504856e-06,
      "loss": 0.0012,
      "step": 261360
    },
    {
      "epoch": 101.5,
      "learning_rate": 6.466278317152104e-06,
      "loss": 0.0099,
      "step": 261370
    },
    {
      "epoch": 101.51,
      "learning_rate": 6.465760517799353e-06,
      "loss": 0.0672,
      "step": 261380
    },
    {
      "epoch": 101.51,
      "learning_rate": 6.465242718446603e-06,
      "loss": 0.034,
      "step": 261390
    },
    {
      "epoch": 101.51,
      "learning_rate": 6.4647249190938525e-06,
      "loss": 0.0334,
      "step": 261400
    },
    {
      "epoch": 101.52,
      "learning_rate": 6.4642071197411004e-06,
      "loss": 0.0437,
      "step": 261410
    },
    {
      "epoch": 101.52,
      "learning_rate": 6.46368932038835e-06,
      "loss": 0.0195,
      "step": 261420
    },
    {
      "epoch": 101.53,
      "learning_rate": 6.4631715210356e-06,
      "loss": 0.1053,
      "step": 261430
    },
    {
      "epoch": 101.53,
      "learning_rate": 6.462653721682849e-06,
      "loss": 0.1218,
      "step": 261440
    },
    {
      "epoch": 101.53,
      "learning_rate": 6.462135922330097e-06,
      "loss": 0.1588,
      "step": 261450
    },
    {
      "epoch": 101.54,
      "learning_rate": 6.461618122977347e-06,
      "loss": 0.0456,
      "step": 261460
    },
    {
      "epoch": 101.54,
      "learning_rate": 6.4611003236245965e-06,
      "loss": 0.0834,
      "step": 261470
    },
    {
      "epoch": 101.55,
      "learning_rate": 6.460582524271845e-06,
      "loss": 0.0891,
      "step": 261480
    },
    {
      "epoch": 101.55,
      "learning_rate": 6.460064724919094e-06,
      "loss": 0.019,
      "step": 261490
    },
    {
      "epoch": 101.55,
      "learning_rate": 6.459546925566344e-06,
      "loss": 0.0843,
      "step": 261500
    },
    {
      "epoch": 101.56,
      "learning_rate": 6.459029126213593e-06,
      "loss": 0.0429,
      "step": 261510
    },
    {
      "epoch": 101.56,
      "learning_rate": 6.458511326860842e-06,
      "loss": 0.0432,
      "step": 261520
    },
    {
      "epoch": 101.57,
      "learning_rate": 6.457993527508091e-06,
      "loss": 0.0114,
      "step": 261530
    },
    {
      "epoch": 101.57,
      "learning_rate": 6.45747572815534e-06,
      "loss": 0.0212,
      "step": 261540
    },
    {
      "epoch": 101.57,
      "learning_rate": 6.45695792880259e-06,
      "loss": 0.0672,
      "step": 261550
    },
    {
      "epoch": 101.58,
      "learning_rate": 6.456440129449839e-06,
      "loss": 0.0461,
      "step": 261560
    },
    {
      "epoch": 101.58,
      "learning_rate": 6.4559223300970875e-06,
      "loss": 0.0417,
      "step": 261570
    },
    {
      "epoch": 101.58,
      "learning_rate": 6.455404530744337e-06,
      "loss": 0.0351,
      "step": 261580
    },
    {
      "epoch": 101.59,
      "learning_rate": 6.454886731391587e-06,
      "loss": 0.0339,
      "step": 261590
    },
    {
      "epoch": 101.59,
      "learning_rate": 6.4543689320388356e-06,
      "loss": 0.0456,
      "step": 261600
    },
    {
      "epoch": 101.6,
      "learning_rate": 6.453851132686084e-06,
      "loss": 0.0818,
      "step": 261610
    },
    {
      "epoch": 101.6,
      "learning_rate": 6.453333333333334e-06,
      "loss": 0.1812,
      "step": 261620
    },
    {
      "epoch": 101.6,
      "learning_rate": 6.452815533980583e-06,
      "loss": 0.0013,
      "step": 261630
    },
    {
      "epoch": 101.61,
      "learning_rate": 6.452297734627832e-06,
      "loss": 0.0102,
      "step": 261640
    },
    {
      "epoch": 101.61,
      "learning_rate": 6.451779935275081e-06,
      "loss": 0.054,
      "step": 261650
    },
    {
      "epoch": 101.62,
      "learning_rate": 6.451262135922331e-06,
      "loss": 0.0018,
      "step": 261660
    },
    {
      "epoch": 101.62,
      "learning_rate": 6.4507443365695795e-06,
      "loss": 0.0116,
      "step": 261670
    },
    {
      "epoch": 101.62,
      "learning_rate": 6.450226537216829e-06,
      "loss": 0.0269,
      "step": 261680
    },
    {
      "epoch": 101.63,
      "learning_rate": 6.449708737864078e-06,
      "loss": 0.0778,
      "step": 261690
    },
    {
      "epoch": 101.63,
      "learning_rate": 6.4491909385113275e-06,
      "loss": 0.1439,
      "step": 261700
    },
    {
      "epoch": 101.63,
      "learning_rate": 6.448673139158576e-06,
      "loss": 0.0009,
      "step": 261710
    },
    {
      "epoch": 101.64,
      "learning_rate": 6.448155339805826e-06,
      "loss": 0.0054,
      "step": 261720
    },
    {
      "epoch": 101.64,
      "learning_rate": 6.447637540453075e-06,
      "loss": 0.0003,
      "step": 261730
    },
    {
      "epoch": 101.65,
      "learning_rate": 6.447119741100324e-06,
      "loss": 0.0005,
      "step": 261740
    },
    {
      "epoch": 101.65,
      "learning_rate": 6.446601941747573e-06,
      "loss": 0.1327,
      "step": 261750
    },
    {
      "epoch": 101.65,
      "learning_rate": 6.446084142394823e-06,
      "loss": 0.1778,
      "step": 261760
    },
    {
      "epoch": 101.66,
      "learning_rate": 6.4455663430420714e-06,
      "loss": 0.0427,
      "step": 261770
    },
    {
      "epoch": 101.66,
      "learning_rate": 6.445048543689321e-06,
      "loss": 0.0199,
      "step": 261780
    },
    {
      "epoch": 101.67,
      "learning_rate": 6.44453074433657e-06,
      "loss": 0.0323,
      "step": 261790
    },
    {
      "epoch": 101.67,
      "learning_rate": 6.4440129449838194e-06,
      "loss": 0.0571,
      "step": 261800
    },
    {
      "epoch": 101.67,
      "learning_rate": 6.443495145631069e-06,
      "loss": 0.0635,
      "step": 261810
    },
    {
      "epoch": 101.68,
      "learning_rate": 6.442977346278317e-06,
      "loss": 0.0114,
      "step": 261820
    },
    {
      "epoch": 101.68,
      "learning_rate": 6.442459546925567e-06,
      "loss": 0.0149,
      "step": 261830
    },
    {
      "epoch": 101.69,
      "learning_rate": 6.441941747572816e-06,
      "loss": 0.0228,
      "step": 261840
    },
    {
      "epoch": 101.69,
      "learning_rate": 6.441423948220066e-06,
      "loss": 0.0149,
      "step": 261850
    },
    {
      "epoch": 101.69,
      "learning_rate": 6.440906148867314e-06,
      "loss": 0.0776,
      "step": 261860
    },
    {
      "epoch": 101.7,
      "learning_rate": 6.440388349514563e-06,
      "loss": 0.0004,
      "step": 261870
    },
    {
      "epoch": 101.7,
      "learning_rate": 6.439870550161813e-06,
      "loss": 0.0991,
      "step": 261880
    },
    {
      "epoch": 101.7,
      "learning_rate": 6.439352750809063e-06,
      "loss": 0.002,
      "step": 261890
    },
    {
      "epoch": 101.71,
      "learning_rate": 6.4388349514563105e-06,
      "loss": 0.0122,
      "step": 261900
    },
    {
      "epoch": 101.71,
      "learning_rate": 6.43831715210356e-06,
      "loss": 0.1291,
      "step": 261910
    },
    {
      "epoch": 101.72,
      "learning_rate": 6.43779935275081e-06,
      "loss": 0.0283,
      "step": 261920
    },
    {
      "epoch": 101.72,
      "learning_rate": 6.437281553398059e-06,
      "loss": 0.0799,
      "step": 261930
    },
    {
      "epoch": 101.72,
      "learning_rate": 6.436763754045307e-06,
      "loss": 0.0695,
      "step": 261940
    },
    {
      "epoch": 101.73,
      "learning_rate": 6.436245954692557e-06,
      "loss": 0.0316,
      "step": 261950
    },
    {
      "epoch": 101.73,
      "learning_rate": 6.4357281553398066e-06,
      "loss": 0.0452,
      "step": 261960
    },
    {
      "epoch": 101.74,
      "learning_rate": 6.435210355987056e-06,
      "loss": 0.0005,
      "step": 261970
    },
    {
      "epoch": 101.74,
      "learning_rate": 6.434692556634304e-06,
      "loss": 0.1686,
      "step": 261980
    },
    {
      "epoch": 101.74,
      "learning_rate": 6.434174757281554e-06,
      "loss": 0.2042,
      "step": 261990
    },
    {
      "epoch": 101.75,
      "learning_rate": 6.433656957928803e-06,
      "loss": 0.0458,
      "step": 262000
    },
    {
      "epoch": 101.75,
      "learning_rate": 6.433139158576053e-06,
      "loss": 0.0243,
      "step": 262010
    },
    {
      "epoch": 101.76,
      "learning_rate": 6.432621359223301e-06,
      "loss": 0.0002,
      "step": 262020
    },
    {
      "epoch": 101.76,
      "learning_rate": 6.4321035598705505e-06,
      "loss": 0.0117,
      "step": 262030
    },
    {
      "epoch": 101.76,
      "learning_rate": 6.4315857605178e-06,
      "loss": 0.1202,
      "step": 262040
    },
    {
      "epoch": 101.77,
      "learning_rate": 6.43106796116505e-06,
      "loss": 0.1342,
      "step": 262050
    },
    {
      "epoch": 101.77,
      "learning_rate": 6.430550161812298e-06,
      "loss": 0.0317,
      "step": 262060
    },
    {
      "epoch": 101.77,
      "learning_rate": 6.430032362459547e-06,
      "loss": 0.0523,
      "step": 262070
    },
    {
      "epoch": 101.78,
      "learning_rate": 6.429514563106797e-06,
      "loss": 0.0986,
      "step": 262080
    },
    {
      "epoch": 101.78,
      "learning_rate": 6.4289967637540465e-06,
      "loss": 0.0367,
      "step": 262090
    },
    {
      "epoch": 101.79,
      "learning_rate": 6.4284789644012944e-06,
      "loss": 0.0391,
      "step": 262100
    },
    {
      "epoch": 101.79,
      "learning_rate": 6.427961165048544e-06,
      "loss": 0.0331,
      "step": 262110
    },
    {
      "epoch": 101.79,
      "learning_rate": 6.427443365695794e-06,
      "loss": 0.0256,
      "step": 262120
    },
    {
      "epoch": 101.8,
      "learning_rate": 6.426925566343043e-06,
      "loss": 0.0633,
      "step": 262130
    },
    {
      "epoch": 101.8,
      "learning_rate": 6.426407766990291e-06,
      "loss": 0.0745,
      "step": 262140
    },
    {
      "epoch": 101.81,
      "learning_rate": 6.425889967637541e-06,
      "loss": 0.0193,
      "step": 262150
    },
    {
      "epoch": 101.81,
      "learning_rate": 6.4253721682847904e-06,
      "loss": 0.0756,
      "step": 262160
    },
    {
      "epoch": 101.81,
      "learning_rate": 6.42485436893204e-06,
      "loss": 0.0923,
      "step": 262170
    },
    {
      "epoch": 101.82,
      "learning_rate": 6.424336569579288e-06,
      "loss": 0.0213,
      "step": 262180
    },
    {
      "epoch": 101.82,
      "learning_rate": 6.423818770226538e-06,
      "loss": 0.0083,
      "step": 262190
    },
    {
      "epoch": 101.83,
      "learning_rate": 6.423300970873787e-06,
      "loss": 0.0084,
      "step": 262200
    },
    {
      "epoch": 101.83,
      "learning_rate": 6.422783171521037e-06,
      "loss": 0.0106,
      "step": 262210
    },
    {
      "epoch": 101.83,
      "learning_rate": 6.422265372168285e-06,
      "loss": 0.0002,
      "step": 262220
    },
    {
      "epoch": 101.84,
      "learning_rate": 6.421747572815534e-06,
      "loss": 0.0475,
      "step": 262230
    },
    {
      "epoch": 101.84,
      "learning_rate": 6.421229773462784e-06,
      "loss": 0.016,
      "step": 262240
    },
    {
      "epoch": 101.84,
      "learning_rate": 6.420711974110034e-06,
      "loss": 0.0883,
      "step": 262250
    },
    {
      "epoch": 101.85,
      "learning_rate": 6.4201941747572815e-06,
      "loss": 0.078,
      "step": 262260
    },
    {
      "epoch": 101.85,
      "learning_rate": 6.419676375404531e-06,
      "loss": 0.0736,
      "step": 262270
    },
    {
      "epoch": 101.86,
      "learning_rate": 6.419158576051781e-06,
      "loss": 0.015,
      "step": 262280
    },
    {
      "epoch": 101.86,
      "learning_rate": 6.4186407766990296e-06,
      "loss": 0.0547,
      "step": 262290
    },
    {
      "epoch": 101.86,
      "learning_rate": 6.418122977346278e-06,
      "loss": 0.1039,
      "step": 262300
    },
    {
      "epoch": 101.87,
      "learning_rate": 6.417605177993528e-06,
      "loss": 0.0781,
      "step": 262310
    },
    {
      "epoch": 101.87,
      "learning_rate": 6.4170873786407776e-06,
      "loss": 0.062,
      "step": 262320
    },
    {
      "epoch": 101.88,
      "learning_rate": 6.416569579288026e-06,
      "loss": 0.0351,
      "step": 262330
    },
    {
      "epoch": 101.88,
      "learning_rate": 6.416051779935275e-06,
      "loss": 0.0693,
      "step": 262340
    },
    {
      "epoch": 101.88,
      "learning_rate": 6.415533980582525e-06,
      "loss": 0.0009,
      "step": 262350
    },
    {
      "epoch": 101.89,
      "learning_rate": 6.415016181229774e-06,
      "loss": 0.0597,
      "step": 262360
    },
    {
      "epoch": 101.89,
      "learning_rate": 6.414498381877023e-06,
      "loss": 0.098,
      "step": 262370
    },
    {
      "epoch": 101.9,
      "learning_rate": 6.413980582524273e-06,
      "loss": 0.0441,
      "step": 262380
    },
    {
      "epoch": 101.9,
      "learning_rate": 6.4134627831715215e-06,
      "loss": 0.0322,
      "step": 262390
    },
    {
      "epoch": 101.9,
      "learning_rate": 6.412944983818771e-06,
      "loss": 0.0169,
      "step": 262400
    },
    {
      "epoch": 101.91,
      "learning_rate": 6.41242718446602e-06,
      "loss": 0.0944,
      "step": 262410
    },
    {
      "epoch": 101.91,
      "learning_rate": 6.4119093851132695e-06,
      "loss": 0.0981,
      "step": 262420
    },
    {
      "epoch": 101.91,
      "learning_rate": 6.411391585760518e-06,
      "loss": 0.0014,
      "step": 262430
    },
    {
      "epoch": 101.92,
      "learning_rate": 6.410873786407768e-06,
      "loss": 0.0002,
      "step": 262440
    },
    {
      "epoch": 101.92,
      "learning_rate": 6.410355987055017e-06,
      "loss": 0.2191,
      "step": 262450
    },
    {
      "epoch": 101.93,
      "learning_rate": 6.409838187702266e-06,
      "loss": 0.0951,
      "step": 262460
    },
    {
      "epoch": 101.93,
      "learning_rate": 6.409320388349515e-06,
      "loss": 0.0133,
      "step": 262470
    },
    {
      "epoch": 101.93,
      "learning_rate": 6.408802588996764e-06,
      "loss": 0.0292,
      "step": 262480
    },
    {
      "epoch": 101.94,
      "learning_rate": 6.4082847896440134e-06,
      "loss": 0.012,
      "step": 262490
    },
    {
      "epoch": 101.94,
      "learning_rate": 6.407766990291263e-06,
      "loss": 0.0276,
      "step": 262500
    },
    {
      "epoch": 101.95,
      "learning_rate": 6.407249190938512e-06,
      "loss": 0.012,
      "step": 262510
    },
    {
      "epoch": 101.95,
      "learning_rate": 6.406731391585761e-06,
      "loss": 0.0014,
      "step": 262520
    },
    {
      "epoch": 101.95,
      "learning_rate": 6.40621359223301e-06,
      "loss": 0.0783,
      "step": 262530
    },
    {
      "epoch": 101.96,
      "learning_rate": 6.40569579288026e-06,
      "loss": 0.0174,
      "step": 262540
    },
    {
      "epoch": 101.96,
      "learning_rate": 6.405177993527509e-06,
      "loss": 0.0165,
      "step": 262550
    },
    {
      "epoch": 101.97,
      "learning_rate": 6.404660194174757e-06,
      "loss": 0.0024,
      "step": 262560
    },
    {
      "epoch": 101.97,
      "learning_rate": 6.404142394822007e-06,
      "loss": 0.0139,
      "step": 262570
    },
    {
      "epoch": 101.97,
      "learning_rate": 6.403624595469257e-06,
      "loss": 0.0319,
      "step": 262580
    },
    {
      "epoch": 101.98,
      "learning_rate": 6.403106796116505e-06,
      "loss": 0.004,
      "step": 262590
    },
    {
      "epoch": 101.98,
      "learning_rate": 6.402588996763754e-06,
      "loss": 0.1765,
      "step": 262600
    },
    {
      "epoch": 101.98,
      "learning_rate": 6.402071197411004e-06,
      "loss": 0.0189,
      "step": 262610
    },
    {
      "epoch": 101.99,
      "learning_rate": 6.401553398058253e-06,
      "loss": 0.0384,
      "step": 262620
    },
    {
      "epoch": 101.99,
      "learning_rate": 6.401035598705501e-06,
      "loss": 0.0366,
      "step": 262630
    },
    {
      "epoch": 102.0,
      "learning_rate": 6.400517799352751e-06,
      "loss": 0.0207,
      "step": 262640
    },
    {
      "epoch": 102.0,
      "learning_rate": 6.4000000000000006e-06,
      "loss": 0.0118,
      "step": 262650
    },
    {
      "epoch": 102.0,
      "eval_accuracy": 0.9491059147180193,
      "eval_loss": 0.38485851883888245,
      "eval_runtime": 8.2149,
      "eval_samples_per_second": 442.49,
      "eval_steps_per_second": 55.387,
      "step": 262650
    },
    {
      "epoch": 102.0,
      "learning_rate": 6.39948220064725e-06,
      "loss": 0.0217,
      "step": 262660
    },
    {
      "epoch": 102.01,
      "learning_rate": 6.398964401294498e-06,
      "loss": 0.0995,
      "step": 262670
    },
    {
      "epoch": 102.01,
      "learning_rate": 6.398446601941748e-06,
      "loss": 0.0006,
      "step": 262680
    },
    {
      "epoch": 102.02,
      "learning_rate": 6.397928802588997e-06,
      "loss": 0.0928,
      "step": 262690
    },
    {
      "epoch": 102.02,
      "learning_rate": 6.397411003236247e-06,
      "loss": 0.042,
      "step": 262700
    },
    {
      "epoch": 102.02,
      "learning_rate": 6.396893203883495e-06,
      "loss": 0.0485,
      "step": 262710
    },
    {
      "epoch": 102.03,
      "learning_rate": 6.3963754045307445e-06,
      "loss": 0.0903,
      "step": 262720
    },
    {
      "epoch": 102.03,
      "learning_rate": 6.395857605177994e-06,
      "loss": 0.0601,
      "step": 262730
    },
    {
      "epoch": 102.03,
      "learning_rate": 6.395339805825244e-06,
      "loss": 0.0897,
      "step": 262740
    },
    {
      "epoch": 102.04,
      "learning_rate": 6.394822006472492e-06,
      "loss": 0.0518,
      "step": 262750
    },
    {
      "epoch": 102.04,
      "learning_rate": 6.394304207119741e-06,
      "loss": 0.0204,
      "step": 262760
    },
    {
      "epoch": 102.05,
      "learning_rate": 6.393786407766991e-06,
      "loss": 0.0749,
      "step": 262770
    },
    {
      "epoch": 102.05,
      "learning_rate": 6.3932686084142405e-06,
      "loss": 0.0199,
      "step": 262780
    },
    {
      "epoch": 102.05,
      "learning_rate": 6.3927508090614884e-06,
      "loss": 0.0088,
      "step": 262790
    },
    {
      "epoch": 102.06,
      "learning_rate": 6.392233009708738e-06,
      "loss": 0.0347,
      "step": 262800
    },
    {
      "epoch": 102.06,
      "learning_rate": 6.391715210355988e-06,
      "loss": 0.0467,
      "step": 262810
    },
    {
      "epoch": 102.07,
      "learning_rate": 6.391197411003237e-06,
      "loss": 0.1711,
      "step": 262820
    },
    {
      "epoch": 102.07,
      "learning_rate": 6.390679611650485e-06,
      "loss": 0.0022,
      "step": 262830
    },
    {
      "epoch": 102.07,
      "learning_rate": 6.390161812297735e-06,
      "loss": 0.049,
      "step": 262840
    },
    {
      "epoch": 102.08,
      "learning_rate": 6.3896440129449844e-06,
      "loss": 0.0572,
      "step": 262850
    },
    {
      "epoch": 102.08,
      "learning_rate": 6.389126213592234e-06,
      "loss": 0.0925,
      "step": 262860
    },
    {
      "epoch": 102.09,
      "learning_rate": 6.388608414239482e-06,
      "loss": 0.0001,
      "step": 262870
    },
    {
      "epoch": 102.09,
      "learning_rate": 6.388090614886732e-06,
      "loss": 0.085,
      "step": 262880
    },
    {
      "epoch": 102.09,
      "learning_rate": 6.387572815533981e-06,
      "loss": 0.0365,
      "step": 262890
    },
    {
      "epoch": 102.1,
      "learning_rate": 6.387055016181231e-06,
      "loss": 0.0596,
      "step": 262900
    },
    {
      "epoch": 102.1,
      "learning_rate": 6.386537216828479e-06,
      "loss": 0.1671,
      "step": 262910
    },
    {
      "epoch": 102.1,
      "learning_rate": 6.386019417475728e-06,
      "loss": 0.1103,
      "step": 262920
    },
    {
      "epoch": 102.11,
      "learning_rate": 6.385501618122978e-06,
      "loss": 0.0497,
      "step": 262930
    },
    {
      "epoch": 102.11,
      "learning_rate": 6.384983818770228e-06,
      "loss": 0.0673,
      "step": 262940
    },
    {
      "epoch": 102.12,
      "learning_rate": 6.384466019417476e-06,
      "loss": 0.0004,
      "step": 262950
    },
    {
      "epoch": 102.12,
      "learning_rate": 6.383948220064725e-06,
      "loss": 0.0525,
      "step": 262960
    },
    {
      "epoch": 102.12,
      "learning_rate": 6.383430420711975e-06,
      "loss": 0.0098,
      "step": 262970
    },
    {
      "epoch": 102.13,
      "learning_rate": 6.382912621359224e-06,
      "loss": 0.0015,
      "step": 262980
    },
    {
      "epoch": 102.13,
      "learning_rate": 6.382394822006473e-06,
      "loss": 0.0335,
      "step": 262990
    },
    {
      "epoch": 102.14,
      "learning_rate": 6.381877022653722e-06,
      "loss": 0.0611,
      "step": 263000
    },
    {
      "epoch": 102.14,
      "learning_rate": 6.3813592233009716e-06,
      "loss": 0.0027,
      "step": 263010
    },
    {
      "epoch": 102.14,
      "learning_rate": 6.380841423948221e-06,
      "loss": 0.0135,
      "step": 263020
    },
    {
      "epoch": 102.15,
      "learning_rate": 6.38032362459547e-06,
      "loss": 0.0563,
      "step": 263030
    },
    {
      "epoch": 102.15,
      "learning_rate": 6.379805825242719e-06,
      "loss": 0.1217,
      "step": 263040
    },
    {
      "epoch": 102.16,
      "learning_rate": 6.379288025889968e-06,
      "loss": 0.0859,
      "step": 263050
    },
    {
      "epoch": 102.16,
      "learning_rate": 6.378770226537218e-06,
      "loss": 0.0127,
      "step": 263060
    },
    {
      "epoch": 102.16,
      "learning_rate": 6.378252427184467e-06,
      "loss": 0.0133,
      "step": 263070
    },
    {
      "epoch": 102.17,
      "learning_rate": 6.3777346278317155e-06,
      "loss": 0.0639,
      "step": 263080
    },
    {
      "epoch": 102.17,
      "learning_rate": 6.377216828478965e-06,
      "loss": 0.0526,
      "step": 263090
    },
    {
      "epoch": 102.17,
      "learning_rate": 6.376699029126214e-06,
      "loss": 0.0923,
      "step": 263100
    },
    {
      "epoch": 102.18,
      "learning_rate": 6.3761812297734635e-06,
      "loss": 0.0905,
      "step": 263110
    },
    {
      "epoch": 102.18,
      "learning_rate": 6.375663430420712e-06,
      "loss": 0.0534,
      "step": 263120
    },
    {
      "epoch": 102.19,
      "learning_rate": 6.375145631067962e-06,
      "loss": 0.0043,
      "step": 263130
    },
    {
      "epoch": 102.19,
      "learning_rate": 6.374627831715211e-06,
      "loss": 0.0862,
      "step": 263140
    },
    {
      "epoch": 102.19,
      "learning_rate": 6.37411003236246e-06,
      "loss": 0.0605,
      "step": 263150
    },
    {
      "epoch": 102.2,
      "learning_rate": 6.373592233009709e-06,
      "loss": 0.0412,
      "step": 263160
    },
    {
      "epoch": 102.2,
      "learning_rate": 6.373074433656959e-06,
      "loss": 0.0682,
      "step": 263170
    },
    {
      "epoch": 102.21,
      "learning_rate": 6.3725566343042074e-06,
      "loss": 0.1837,
      "step": 263180
    },
    {
      "epoch": 102.21,
      "learning_rate": 6.372038834951457e-06,
      "loss": 0.0568,
      "step": 263190
    },
    {
      "epoch": 102.21,
      "learning_rate": 6.371521035598706e-06,
      "loss": 0.0499,
      "step": 263200
    },
    {
      "epoch": 102.22,
      "learning_rate": 6.3710032362459555e-06,
      "loss": 0.0094,
      "step": 263210
    },
    {
      "epoch": 102.22,
      "learning_rate": 6.370485436893204e-06,
      "loss": 0.0015,
      "step": 263220
    },
    {
      "epoch": 102.23,
      "learning_rate": 6.369967637540454e-06,
      "loss": 0.0256,
      "step": 263230
    },
    {
      "epoch": 102.23,
      "learning_rate": 6.369449838187703e-06,
      "loss": 0.0008,
      "step": 263240
    },
    {
      "epoch": 102.23,
      "learning_rate": 6.368932038834952e-06,
      "loss": 0.0418,
      "step": 263250
    },
    {
      "epoch": 102.24,
      "learning_rate": 6.368414239482201e-06,
      "loss": 0.026,
      "step": 263260
    },
    {
      "epoch": 102.24,
      "learning_rate": 6.367896440129451e-06,
      "loss": 0.0123,
      "step": 263270
    },
    {
      "epoch": 102.24,
      "learning_rate": 6.367378640776699e-06,
      "loss": 0.0264,
      "step": 263280
    },
    {
      "epoch": 102.25,
      "learning_rate": 6.366860841423948e-06,
      "loss": 0.0022,
      "step": 263290
    },
    {
      "epoch": 102.25,
      "learning_rate": 6.366343042071198e-06,
      "loss": 0.0811,
      "step": 263300
    },
    {
      "epoch": 102.26,
      "learning_rate": 6.365825242718447e-06,
      "loss": 0.088,
      "step": 263310
    },
    {
      "epoch": 102.26,
      "learning_rate": 6.365307443365696e-06,
      "loss": 0.0077,
      "step": 263320
    },
    {
      "epoch": 102.26,
      "learning_rate": 6.364789644012945e-06,
      "loss": 0.0309,
      "step": 263330
    },
    {
      "epoch": 102.27,
      "learning_rate": 6.3642718446601946e-06,
      "loss": 0.0011,
      "step": 263340
    },
    {
      "epoch": 102.27,
      "learning_rate": 6.363754045307444e-06,
      "loss": 0.0004,
      "step": 263350
    },
    {
      "epoch": 102.28,
      "learning_rate": 6.363236245954693e-06,
      "loss": 0.0259,
      "step": 263360
    },
    {
      "epoch": 102.28,
      "learning_rate": 6.362718446601942e-06,
      "loss": 0.0849,
      "step": 263370
    },
    {
      "epoch": 102.28,
      "learning_rate": 6.362200647249191e-06,
      "loss": 0.0216,
      "step": 263380
    },
    {
      "epoch": 102.29,
      "learning_rate": 6.361682847896441e-06,
      "loss": 0.0299,
      "step": 263390
    },
    {
      "epoch": 102.29,
      "learning_rate": 6.36116504854369e-06,
      "loss": 0.0305,
      "step": 263400
    },
    {
      "epoch": 102.3,
      "learning_rate": 6.3606472491909385e-06,
      "loss": 0.003,
      "step": 263410
    },
    {
      "epoch": 102.3,
      "learning_rate": 6.360129449838188e-06,
      "loss": 0.0785,
      "step": 263420
    },
    {
      "epoch": 102.3,
      "learning_rate": 6.359611650485438e-06,
      "loss": 0.0424,
      "step": 263430
    },
    {
      "epoch": 102.31,
      "learning_rate": 6.359093851132686e-06,
      "loss": 0.0231,
      "step": 263440
    },
    {
      "epoch": 102.31,
      "learning_rate": 6.358576051779935e-06,
      "loss": 0.0546,
      "step": 263450
    },
    {
      "epoch": 102.31,
      "learning_rate": 6.358058252427185e-06,
      "loss": 0.0648,
      "step": 263460
    },
    {
      "epoch": 102.32,
      "learning_rate": 6.3575404530744345e-06,
      "loss": 0.0195,
      "step": 263470
    },
    {
      "epoch": 102.32,
      "learning_rate": 6.357022653721684e-06,
      "loss": 0.0234,
      "step": 263480
    },
    {
      "epoch": 102.33,
      "learning_rate": 6.356504854368932e-06,
      "loss": 0.0219,
      "step": 263490
    },
    {
      "epoch": 102.33,
      "learning_rate": 6.355987055016182e-06,
      "loss": 0.0315,
      "step": 263500
    },
    {
      "epoch": 102.33,
      "learning_rate": 6.355469255663431e-06,
      "loss": 0.0113,
      "step": 263510
    },
    {
      "epoch": 102.34,
      "learning_rate": 6.354951456310681e-06,
      "loss": 0.029,
      "step": 263520
    },
    {
      "epoch": 102.34,
      "learning_rate": 6.354433656957929e-06,
      "loss": 0.0161,
      "step": 263530
    },
    {
      "epoch": 102.35,
      "learning_rate": 6.3539158576051784e-06,
      "loss": 0.0374,
      "step": 263540
    },
    {
      "epoch": 102.35,
      "learning_rate": 6.353398058252428e-06,
      "loss": 0.0032,
      "step": 263550
    },
    {
      "epoch": 102.35,
      "learning_rate": 6.352880258899678e-06,
      "loss": 0.0229,
      "step": 263560
    },
    {
      "epoch": 102.36,
      "learning_rate": 6.352362459546926e-06,
      "loss": 0.1049,
      "step": 263570
    },
    {
      "epoch": 102.36,
      "learning_rate": 6.351844660194175e-06,
      "loss": 0.0469,
      "step": 263580
    },
    {
      "epoch": 102.37,
      "learning_rate": 6.351326860841425e-06,
      "loss": 0.0148,
      "step": 263590
    },
    {
      "epoch": 102.37,
      "learning_rate": 6.3508090614886745e-06,
      "loss": 0.0719,
      "step": 263600
    },
    {
      "epoch": 102.37,
      "learning_rate": 6.350291262135922e-06,
      "loss": 0.1212,
      "step": 263610
    },
    {
      "epoch": 102.38,
      "learning_rate": 6.349773462783172e-06,
      "loss": 0.0213,
      "step": 263620
    },
    {
      "epoch": 102.38,
      "learning_rate": 6.349255663430422e-06,
      "loss": 0.1691,
      "step": 263630
    },
    {
      "epoch": 102.38,
      "learning_rate": 6.348737864077671e-06,
      "loss": 0.0019,
      "step": 263640
    },
    {
      "epoch": 102.39,
      "learning_rate": 6.348220064724919e-06,
      "loss": 0.1392,
      "step": 263650
    },
    {
      "epoch": 102.39,
      "learning_rate": 6.347702265372169e-06,
      "loss": 0.065,
      "step": 263660
    },
    {
      "epoch": 102.4,
      "learning_rate": 6.347184466019418e-06,
      "loss": 0.0668,
      "step": 263670
    },
    {
      "epoch": 102.4,
      "learning_rate": 6.346666666666668e-06,
      "loss": 0.0752,
      "step": 263680
    },
    {
      "epoch": 102.4,
      "learning_rate": 6.346148867313916e-06,
      "loss": 0.0275,
      "step": 263690
    },
    {
      "epoch": 102.41,
      "learning_rate": 6.3456310679611656e-06,
      "loss": 0.0333,
      "step": 263700
    },
    {
      "epoch": 102.41,
      "learning_rate": 6.345113268608415e-06,
      "loss": 0.0327,
      "step": 263710
    },
    {
      "epoch": 102.42,
      "learning_rate": 6.344595469255665e-06,
      "loss": 0.002,
      "step": 263720
    },
    {
      "epoch": 102.42,
      "learning_rate": 6.344077669902913e-06,
      "loss": 0.0768,
      "step": 263730
    },
    {
      "epoch": 102.42,
      "learning_rate": 6.343559870550162e-06,
      "loss": 0.0087,
      "step": 263740
    },
    {
      "epoch": 102.43,
      "learning_rate": 6.343042071197412e-06,
      "loss": 0.0497,
      "step": 263750
    },
    {
      "epoch": 102.43,
      "learning_rate": 6.342524271844661e-06,
      "loss": 0.0903,
      "step": 263760
    },
    {
      "epoch": 102.43,
      "learning_rate": 6.3420064724919095e-06,
      "loss": 0.0307,
      "step": 263770
    },
    {
      "epoch": 102.44,
      "learning_rate": 6.341488673139159e-06,
      "loss": 0.0004,
      "step": 263780
    },
    {
      "epoch": 102.44,
      "learning_rate": 6.340970873786409e-06,
      "loss": 0.0402,
      "step": 263790
    },
    {
      "epoch": 102.45,
      "learning_rate": 6.3404530744336575e-06,
      "loss": 0.0998,
      "step": 263800
    },
    {
      "epoch": 102.45,
      "learning_rate": 6.339935275080906e-06,
      "loss": 0.043,
      "step": 263810
    },
    {
      "epoch": 102.45,
      "learning_rate": 6.339417475728156e-06,
      "loss": 0.0657,
      "step": 263820
    },
    {
      "epoch": 102.46,
      "learning_rate": 6.3388996763754055e-06,
      "loss": 0.1726,
      "step": 263830
    },
    {
      "epoch": 102.46,
      "learning_rate": 6.338381877022654e-06,
      "loss": 0.0342,
      "step": 263840
    },
    {
      "epoch": 102.47,
      "learning_rate": 6.337864077669903e-06,
      "loss": 0.016,
      "step": 263850
    },
    {
      "epoch": 102.47,
      "learning_rate": 6.337346278317153e-06,
      "loss": 0.1193,
      "step": 263860
    },
    {
      "epoch": 102.47,
      "learning_rate": 6.336828478964402e-06,
      "loss": 0.0705,
      "step": 263870
    },
    {
      "epoch": 102.48,
      "learning_rate": 6.336310679611651e-06,
      "loss": 0.0197,
      "step": 263880
    },
    {
      "epoch": 102.48,
      "learning_rate": 6.3357928802589e-06,
      "loss": 0.104,
      "step": 263890
    },
    {
      "epoch": 102.49,
      "learning_rate": 6.3352750809061494e-06,
      "loss": 0.1033,
      "step": 263900
    },
    {
      "epoch": 102.49,
      "learning_rate": 6.334757281553398e-06,
      "loss": 0.0083,
      "step": 263910
    },
    {
      "epoch": 102.49,
      "learning_rate": 6.334239482200648e-06,
      "loss": 0.0005,
      "step": 263920
    },
    {
      "epoch": 102.5,
      "learning_rate": 6.333721682847897e-06,
      "loss": 0.0438,
      "step": 263930
    },
    {
      "epoch": 102.5,
      "learning_rate": 6.333203883495146e-06,
      "loss": 0.0121,
      "step": 263940
    },
    {
      "epoch": 102.5,
      "learning_rate": 6.332686084142395e-06,
      "loss": 0.0017,
      "step": 263950
    },
    {
      "epoch": 102.51,
      "learning_rate": 6.332168284789645e-06,
      "loss": 0.0487,
      "step": 263960
    },
    {
      "epoch": 102.51,
      "learning_rate": 6.331650485436893e-06,
      "loss": 0.0974,
      "step": 263970
    },
    {
      "epoch": 102.52,
      "learning_rate": 6.331132686084143e-06,
      "loss": 0.0623,
      "step": 263980
    },
    {
      "epoch": 102.52,
      "learning_rate": 6.330614886731392e-06,
      "loss": 0.0155,
      "step": 263990
    },
    {
      "epoch": 102.52,
      "learning_rate": 6.330097087378641e-06,
      "loss": 0.0133,
      "step": 264000
    },
    {
      "epoch": 102.53,
      "learning_rate": 6.32957928802589e-06,
      "loss": 0.0722,
      "step": 264010
    },
    {
      "epoch": 102.53,
      "learning_rate": 6.32906148867314e-06,
      "loss": 0.0654,
      "step": 264020
    },
    {
      "epoch": 102.54,
      "learning_rate": 6.3285436893203886e-06,
      "loss": 0.0009,
      "step": 264030
    },
    {
      "epoch": 102.54,
      "learning_rate": 6.328025889967638e-06,
      "loss": 0.062,
      "step": 264040
    },
    {
      "epoch": 102.54,
      "learning_rate": 6.327508090614888e-06,
      "loss": 0.0852,
      "step": 264050
    },
    {
      "epoch": 102.55,
      "learning_rate": 6.3269902912621366e-06,
      "loss": 0.0283,
      "step": 264060
    },
    {
      "epoch": 102.55,
      "learning_rate": 6.326472491909385e-06,
      "loss": 0.0563,
      "step": 264070
    },
    {
      "epoch": 102.56,
      "learning_rate": 6.325954692556635e-06,
      "loss": 0.051,
      "step": 264080
    },
    {
      "epoch": 102.56,
      "learning_rate": 6.3254368932038846e-06,
      "loss": 0.029,
      "step": 264090
    },
    {
      "epoch": 102.56,
      "learning_rate": 6.3249190938511325e-06,
      "loss": 0.0226,
      "step": 264100
    },
    {
      "epoch": 102.57,
      "learning_rate": 6.324401294498382e-06,
      "loss": 0.0522,
      "step": 264110
    },
    {
      "epoch": 102.57,
      "learning_rate": 6.323883495145632e-06,
      "loss": 0.0225,
      "step": 264120
    },
    {
      "epoch": 102.57,
      "learning_rate": 6.323365695792881e-06,
      "loss": 0.0049,
      "step": 264130
    },
    {
      "epoch": 102.58,
      "learning_rate": 6.322847896440129e-06,
      "loss": 0.1292,
      "step": 264140
    },
    {
      "epoch": 102.58,
      "learning_rate": 6.322330097087379e-06,
      "loss": 0.0728,
      "step": 264150
    },
    {
      "epoch": 102.59,
      "learning_rate": 6.3218122977346285e-06,
      "loss": 0.1119,
      "step": 264160
    },
    {
      "epoch": 102.59,
      "learning_rate": 6.321294498381878e-06,
      "loss": 0.081,
      "step": 264170
    },
    {
      "epoch": 102.59,
      "learning_rate": 6.320776699029126e-06,
      "loss": 0.0864,
      "step": 264180
    },
    {
      "epoch": 102.6,
      "learning_rate": 6.320258899676376e-06,
      "loss": 0.0084,
      "step": 264190
    },
    {
      "epoch": 102.6,
      "learning_rate": 6.319741100323625e-06,
      "loss": 0.0125,
      "step": 264200
    },
    {
      "epoch": 102.61,
      "learning_rate": 6.319223300970875e-06,
      "loss": 0.1179,
      "step": 264210
    },
    {
      "epoch": 102.61,
      "learning_rate": 6.318705501618123e-06,
      "loss": 0.0192,
      "step": 264220
    },
    {
      "epoch": 102.61,
      "learning_rate": 6.3181877022653724e-06,
      "loss": 0.0002,
      "step": 264230
    },
    {
      "epoch": 102.62,
      "learning_rate": 6.317669902912622e-06,
      "loss": 0.0091,
      "step": 264240
    },
    {
      "epoch": 102.62,
      "learning_rate": 6.317152103559872e-06,
      "loss": 0.0213,
      "step": 264250
    },
    {
      "epoch": 102.63,
      "learning_rate": 6.31663430420712e-06,
      "loss": 0.0125,
      "step": 264260
    },
    {
      "epoch": 102.63,
      "learning_rate": 6.316116504854369e-06,
      "loss": 0.0026,
      "step": 264270
    },
    {
      "epoch": 102.63,
      "learning_rate": 6.315598705501619e-06,
      "loss": 0.1897,
      "step": 264280
    },
    {
      "epoch": 102.64,
      "learning_rate": 6.3150809061488685e-06,
      "loss": 0.0181,
      "step": 264290
    },
    {
      "epoch": 102.64,
      "learning_rate": 6.314563106796116e-06,
      "loss": 0.0832,
      "step": 264300
    },
    {
      "epoch": 102.64,
      "learning_rate": 6.314045307443366e-06,
      "loss": 0.0177,
      "step": 264310
    },
    {
      "epoch": 102.65,
      "learning_rate": 6.313527508090616e-06,
      "loss": 0.0941,
      "step": 264320
    },
    {
      "epoch": 102.65,
      "learning_rate": 6.313009708737865e-06,
      "loss": 0.1718,
      "step": 264330
    },
    {
      "epoch": 102.66,
      "learning_rate": 6.312491909385113e-06,
      "loss": 0.0382,
      "step": 264340
    },
    {
      "epoch": 102.66,
      "learning_rate": 6.311974110032363e-06,
      "loss": 0.0025,
      "step": 264350
    },
    {
      "epoch": 102.66,
      "learning_rate": 6.311456310679612e-06,
      "loss": 0.0891,
      "step": 264360
    },
    {
      "epoch": 102.67,
      "learning_rate": 6.310938511326862e-06,
      "loss": 0.0984,
      "step": 264370
    },
    {
      "epoch": 102.67,
      "learning_rate": 6.31042071197411e-06,
      "loss": 0.1473,
      "step": 264380
    },
    {
      "epoch": 102.68,
      "learning_rate": 6.3099029126213596e-06,
      "loss": 0.0437,
      "step": 264390
    },
    {
      "epoch": 102.68,
      "learning_rate": 6.309385113268609e-06,
      "loss": 0.001,
      "step": 264400
    },
    {
      "epoch": 102.68,
      "learning_rate": 6.308867313915859e-06,
      "loss": 0.1302,
      "step": 264410
    },
    {
      "epoch": 102.69,
      "learning_rate": 6.308349514563107e-06,
      "loss": 0.0248,
      "step": 264420
    },
    {
      "epoch": 102.69,
      "learning_rate": 6.307831715210356e-06,
      "loss": 0.0131,
      "step": 264430
    },
    {
      "epoch": 102.7,
      "learning_rate": 6.307313915857606e-06,
      "loss": 0.0139,
      "step": 264440
    },
    {
      "epoch": 102.7,
      "learning_rate": 6.3067961165048556e-06,
      "loss": 0.0142,
      "step": 264450
    },
    {
      "epoch": 102.7,
      "learning_rate": 6.3062783171521035e-06,
      "loss": 0.0071,
      "step": 264460
    },
    {
      "epoch": 102.71,
      "learning_rate": 6.305760517799353e-06,
      "loss": 0.1486,
      "step": 264470
    },
    {
      "epoch": 102.71,
      "learning_rate": 6.305242718446603e-06,
      "loss": 0.1743,
      "step": 264480
    },
    {
      "epoch": 102.71,
      "learning_rate": 6.304724919093852e-06,
      "loss": 0.0938,
      "step": 264490
    },
    {
      "epoch": 102.72,
      "learning_rate": 6.3042071197411e-06,
      "loss": 0.0717,
      "step": 264500
    },
    {
      "epoch": 102.72,
      "learning_rate": 6.30368932038835e-06,
      "loss": 0.0386,
      "step": 264510
    },
    {
      "epoch": 102.73,
      "learning_rate": 6.3031715210355995e-06,
      "loss": 0.075,
      "step": 264520
    },
    {
      "epoch": 102.73,
      "learning_rate": 6.302653721682849e-06,
      "loss": 0.1006,
      "step": 264530
    },
    {
      "epoch": 102.73,
      "learning_rate": 6.302135922330097e-06,
      "loss": 0.1097,
      "step": 264540
    },
    {
      "epoch": 102.74,
      "learning_rate": 6.301618122977347e-06,
      "loss": 0.1012,
      "step": 264550
    },
    {
      "epoch": 102.74,
      "learning_rate": 6.301100323624596e-06,
      "loss": 0.0013,
      "step": 264560
    },
    {
      "epoch": 102.75,
      "learning_rate": 6.300582524271845e-06,
      "loss": 0.1417,
      "step": 264570
    },
    {
      "epoch": 102.75,
      "learning_rate": 6.300064724919094e-06,
      "loss": 0.0099,
      "step": 264580
    },
    {
      "epoch": 102.75,
      "learning_rate": 6.2995469255663434e-06,
      "loss": 0.0176,
      "step": 264590
    },
    {
      "epoch": 102.76,
      "learning_rate": 6.299029126213593e-06,
      "loss": 0.0889,
      "step": 264600
    },
    {
      "epoch": 102.76,
      "learning_rate": 6.298511326860842e-06,
      "loss": 0.0016,
      "step": 264610
    },
    {
      "epoch": 102.77,
      "learning_rate": 6.2979935275080915e-06,
      "loss": 0.0011,
      "step": 264620
    },
    {
      "epoch": 102.77,
      "learning_rate": 6.29747572815534e-06,
      "loss": 0.0178,
      "step": 264630
    },
    {
      "epoch": 102.77,
      "learning_rate": 6.29695792880259e-06,
      "loss": 0.0596,
      "step": 264640
    },
    {
      "epoch": 102.78,
      "learning_rate": 6.296440129449839e-06,
      "loss": 0.0566,
      "step": 264650
    },
    {
      "epoch": 102.78,
      "learning_rate": 6.295922330097088e-06,
      "loss": 0.1539,
      "step": 264660
    },
    {
      "epoch": 102.78,
      "learning_rate": 6.295404530744337e-06,
      "loss": 0.116,
      "step": 264670
    },
    {
      "epoch": 102.79,
      "learning_rate": 6.294886731391587e-06,
      "loss": 0.0462,
      "step": 264680
    },
    {
      "epoch": 102.79,
      "learning_rate": 6.294368932038835e-06,
      "loss": 0.0303,
      "step": 264690
    },
    {
      "epoch": 102.8,
      "learning_rate": 6.293851132686085e-06,
      "loss": 0.0365,
      "step": 264700
    },
    {
      "epoch": 102.8,
      "learning_rate": 6.293333333333334e-06,
      "loss": 0.0002,
      "step": 264710
    },
    {
      "epoch": 102.8,
      "learning_rate": 6.292815533980583e-06,
      "loss": 0.0269,
      "step": 264720
    },
    {
      "epoch": 102.81,
      "learning_rate": 6.292297734627832e-06,
      "loss": 0.0259,
      "step": 264730
    },
    {
      "epoch": 102.81,
      "learning_rate": 6.291779935275082e-06,
      "loss": 0.114,
      "step": 264740
    },
    {
      "epoch": 102.82,
      "learning_rate": 6.2912621359223306e-06,
      "loss": 0.0373,
      "step": 264750
    },
    {
      "epoch": 102.82,
      "learning_rate": 6.290744336569579e-06,
      "loss": 0.1162,
      "step": 264760
    },
    {
      "epoch": 102.82,
      "learning_rate": 6.290226537216829e-06,
      "loss": 0.0913,
      "step": 264770
    },
    {
      "epoch": 102.83,
      "learning_rate": 6.2897087378640786e-06,
      "loss": 0.1233,
      "step": 264780
    },
    {
      "epoch": 102.83,
      "learning_rate": 6.289190938511327e-06,
      "loss": 0.0677,
      "step": 264790
    },
    {
      "epoch": 102.83,
      "learning_rate": 6.288673139158576e-06,
      "loss": 0.1268,
      "step": 264800
    },
    {
      "epoch": 102.84,
      "learning_rate": 6.288155339805826e-06,
      "loss": 0.1093,
      "step": 264810
    },
    {
      "epoch": 102.84,
      "learning_rate": 6.287637540453075e-06,
      "loss": 0.0512,
      "step": 264820
    },
    {
      "epoch": 102.85,
      "learning_rate": 6.287119741100324e-06,
      "loss": 0.0002,
      "step": 264830
    },
    {
      "epoch": 102.85,
      "learning_rate": 6.286601941747573e-06,
      "loss": 0.0303,
      "step": 264840
    },
    {
      "epoch": 102.85,
      "learning_rate": 6.2860841423948225e-06,
      "loss": 0.0471,
      "step": 264850
    },
    {
      "epoch": 102.86,
      "learning_rate": 6.285566343042072e-06,
      "loss": 0.0197,
      "step": 264860
    },
    {
      "epoch": 102.86,
      "learning_rate": 6.285048543689321e-06,
      "loss": 0.0644,
      "step": 264870
    },
    {
      "epoch": 102.87,
      "learning_rate": 6.28453074433657e-06,
      "loss": 0.0294,
      "step": 264880
    },
    {
      "epoch": 102.87,
      "learning_rate": 6.284012944983819e-06,
      "loss": 0.0942,
      "step": 264890
    },
    {
      "epoch": 102.87,
      "learning_rate": 6.283495145631069e-06,
      "loss": 0.0267,
      "step": 264900
    },
    {
      "epoch": 102.88,
      "learning_rate": 6.282977346278317e-06,
      "loss": 0.0182,
      "step": 264910
    },
    {
      "epoch": 102.88,
      "learning_rate": 6.2824595469255664e-06,
      "loss": 0.0462,
      "step": 264920
    },
    {
      "epoch": 102.89,
      "learning_rate": 6.281941747572816e-06,
      "loss": 0.0004,
      "step": 264930
    },
    {
      "epoch": 102.89,
      "learning_rate": 6.281423948220066e-06,
      "loss": 0.0944,
      "step": 264940
    },
    {
      "epoch": 102.89,
      "learning_rate": 6.280906148867314e-06,
      "loss": 0.0089,
      "step": 264950
    },
    {
      "epoch": 102.9,
      "learning_rate": 6.280388349514563e-06,
      "loss": 0.1006,
      "step": 264960
    },
    {
      "epoch": 102.9,
      "learning_rate": 6.279870550161813e-06,
      "loss": 0.0004,
      "step": 264970
    },
    {
      "epoch": 102.9,
      "learning_rate": 6.2793527508090625e-06,
      "loss": 0.0814,
      "step": 264980
    },
    {
      "epoch": 102.91,
      "learning_rate": 6.27883495145631e-06,
      "loss": 0.0191,
      "step": 264990
    },
    {
      "epoch": 102.91,
      "learning_rate": 6.27831715210356e-06,
      "loss": 0.0127,
      "step": 265000
    },
    {
      "epoch": 102.92,
      "learning_rate": 6.27779935275081e-06,
      "loss": 0.0811,
      "step": 265010
    },
    {
      "epoch": 102.92,
      "learning_rate": 6.277281553398059e-06,
      "loss": 0.2498,
      "step": 265020
    },
    {
      "epoch": 102.92,
      "learning_rate": 6.276763754045307e-06,
      "loss": 0.0915,
      "step": 265030
    },
    {
      "epoch": 102.93,
      "learning_rate": 6.276245954692557e-06,
      "loss": 0.0143,
      "step": 265040
    },
    {
      "epoch": 102.93,
      "learning_rate": 6.275728155339806e-06,
      "loss": 0.1296,
      "step": 265050
    },
    {
      "epoch": 102.94,
      "learning_rate": 6.275210355987056e-06,
      "loss": 0.1452,
      "step": 265060
    },
    {
      "epoch": 102.94,
      "learning_rate": 6.274692556634304e-06,
      "loss": 0.0481,
      "step": 265070
    },
    {
      "epoch": 102.94,
      "learning_rate": 6.2741747572815536e-06,
      "loss": 0.0763,
      "step": 265080
    },
    {
      "epoch": 102.95,
      "learning_rate": 6.273656957928803e-06,
      "loss": 0.0006,
      "step": 265090
    },
    {
      "epoch": 102.95,
      "learning_rate": 6.273139158576053e-06,
      "loss": 0.0547,
      "step": 265100
    },
    {
      "epoch": 102.96,
      "learning_rate": 6.272621359223301e-06,
      "loss": 0.2334,
      "step": 265110
    },
    {
      "epoch": 102.96,
      "learning_rate": 6.27210355987055e-06,
      "loss": 0.0418,
      "step": 265120
    },
    {
      "epoch": 102.96,
      "learning_rate": 6.2715857605178e-06,
      "loss": 0.0317,
      "step": 265130
    },
    {
      "epoch": 102.97,
      "learning_rate": 6.2710679611650496e-06,
      "loss": 0.0739,
      "step": 265140
    },
    {
      "epoch": 102.97,
      "learning_rate": 6.2705501618122975e-06,
      "loss": 0.091,
      "step": 265150
    },
    {
      "epoch": 102.97,
      "learning_rate": 6.270032362459547e-06,
      "loss": 0.0142,
      "step": 265160
    },
    {
      "epoch": 102.98,
      "learning_rate": 6.269514563106797e-06,
      "loss": 0.101,
      "step": 265170
    },
    {
      "epoch": 102.98,
      "learning_rate": 6.268996763754046e-06,
      "loss": 0.1589,
      "step": 265180
    },
    {
      "epoch": 102.99,
      "learning_rate": 6.268478964401296e-06,
      "loss": 0.0397,
      "step": 265190
    },
    {
      "epoch": 102.99,
      "learning_rate": 6.267961165048544e-06,
      "loss": 0.0148,
      "step": 265200
    },
    {
      "epoch": 102.99,
      "learning_rate": 6.2674433656957935e-06,
      "loss": 0.0528,
      "step": 265210
    },
    {
      "epoch": 103.0,
      "learning_rate": 6.266925566343043e-06,
      "loss": 0.0307,
      "step": 265220
    },
    {
      "epoch": 103.0,
      "eval_accuracy": 0.9535075653370014,
      "eval_loss": 0.36436453461647034,
      "eval_runtime": 8.1887,
      "eval_samples_per_second": 443.903,
      "eval_steps_per_second": 55.564,
      "step": 265225
    },
    {
      "epoch": 103.0,
      "learning_rate": 6.266407766990292e-06,
      "loss": 0.0408,
      "step": 265230
    },
    {
      "epoch": 103.01,
      "learning_rate": 6.265889967637541e-06,
      "loss": 0.0114,
      "step": 265240
    },
    {
      "epoch": 103.01,
      "learning_rate": 6.26537216828479e-06,
      "loss": 0.0002,
      "step": 265250
    },
    {
      "epoch": 103.01,
      "learning_rate": 6.26485436893204e-06,
      "loss": 0.0371,
      "step": 265260
    },
    {
      "epoch": 103.02,
      "learning_rate": 6.264336569579289e-06,
      "loss": 0.1479,
      "step": 265270
    },
    {
      "epoch": 103.02,
      "learning_rate": 6.2638187702265374e-06,
      "loss": 0.0824,
      "step": 265280
    },
    {
      "epoch": 103.03,
      "learning_rate": 6.263300970873787e-06,
      "loss": 0.1319,
      "step": 265290
    },
    {
      "epoch": 103.03,
      "learning_rate": 6.262783171521037e-06,
      "loss": 0.1564,
      "step": 265300
    },
    {
      "epoch": 103.03,
      "learning_rate": 6.2622653721682855e-06,
      "loss": 0.0194,
      "step": 265310
    },
    {
      "epoch": 103.04,
      "learning_rate": 6.261747572815534e-06,
      "loss": 0.015,
      "step": 265320
    },
    {
      "epoch": 103.04,
      "learning_rate": 6.261229773462784e-06,
      "loss": 0.0448,
      "step": 265330
    },
    {
      "epoch": 103.04,
      "learning_rate": 6.2607119741100335e-06,
      "loss": 0.0535,
      "step": 265340
    },
    {
      "epoch": 103.05,
      "learning_rate": 6.260194174757282e-06,
      "loss": 0.0865,
      "step": 265350
    },
    {
      "epoch": 103.05,
      "learning_rate": 6.259676375404531e-06,
      "loss": 0.0313,
      "step": 265360
    },
    {
      "epoch": 103.06,
      "learning_rate": 6.259158576051781e-06,
      "loss": 0.2875,
      "step": 265370
    },
    {
      "epoch": 103.06,
      "learning_rate": 6.258640776699029e-06,
      "loss": 0.0088,
      "step": 265380
    },
    {
      "epoch": 103.06,
      "learning_rate": 6.258122977346279e-06,
      "loss": 0.0233,
      "step": 265390
    },
    {
      "epoch": 103.07,
      "learning_rate": 6.257605177993528e-06,
      "loss": 0.0228,
      "step": 265400
    },
    {
      "epoch": 103.07,
      "learning_rate": 6.257087378640777e-06,
      "loss": 0.028,
      "step": 265410
    },
    {
      "epoch": 103.08,
      "learning_rate": 6.256569579288026e-06,
      "loss": 0.0996,
      "step": 265420
    },
    {
      "epoch": 103.08,
      "learning_rate": 6.256051779935276e-06,
      "loss": 0.076,
      "step": 265430
    },
    {
      "epoch": 103.08,
      "learning_rate": 6.2555339805825246e-06,
      "loss": 0.1366,
      "step": 265440
    },
    {
      "epoch": 103.09,
      "learning_rate": 6.255016181229774e-06,
      "loss": 0.116,
      "step": 265450
    },
    {
      "epoch": 103.09,
      "learning_rate": 6.254498381877023e-06,
      "loss": 0.037,
      "step": 265460
    },
    {
      "epoch": 103.1,
      "learning_rate": 6.2539805825242726e-06,
      "loss": 0.0094,
      "step": 265470
    },
    {
      "epoch": 103.1,
      "learning_rate": 6.253462783171521e-06,
      "loss": 0.0154,
      "step": 265480
    },
    {
      "epoch": 103.1,
      "learning_rate": 6.252944983818771e-06,
      "loss": 0.0088,
      "step": 265490
    },
    {
      "epoch": 103.11,
      "learning_rate": 6.25242718446602e-06,
      "loss": 0.0597,
      "step": 265500
    },
    {
      "epoch": 103.11,
      "learning_rate": 6.251909385113269e-06,
      "loss": 0.2524,
      "step": 265510
    },
    {
      "epoch": 103.11,
      "learning_rate": 6.251391585760518e-06,
      "loss": 0.0892,
      "step": 265520
    },
    {
      "epoch": 103.12,
      "learning_rate": 6.250873786407768e-06,
      "loss": 0.013,
      "step": 265530
    },
    {
      "epoch": 103.12,
      "learning_rate": 6.2503559870550165e-06,
      "loss": 0.0528,
      "step": 265540
    },
    {
      "epoch": 103.13,
      "learning_rate": 6.249838187702266e-06,
      "loss": 0.0254,
      "step": 265550
    },
    {
      "epoch": 103.13,
      "learning_rate": 6.249320388349515e-06,
      "loss": 0.0505,
      "step": 265560
    },
    {
      "epoch": 103.13,
      "learning_rate": 6.248802588996764e-06,
      "loss": 0.0703,
      "step": 265570
    },
    {
      "epoch": 103.14,
      "learning_rate": 6.248284789644013e-06,
      "loss": 0.0516,
      "step": 265580
    },
    {
      "epoch": 103.14,
      "learning_rate": 6.247766990291263e-06,
      "loss": 0.0575,
      "step": 265590
    },
    {
      "epoch": 103.15,
      "learning_rate": 6.247249190938512e-06,
      "loss": 0.0234,
      "step": 265600
    },
    {
      "epoch": 103.15,
      "learning_rate": 6.2467313915857604e-06,
      "loss": 0.0026,
      "step": 265610
    },
    {
      "epoch": 103.15,
      "learning_rate": 6.24621359223301e-06,
      "loss": 0.0399,
      "step": 265620
    },
    {
      "epoch": 103.16,
      "learning_rate": 6.24569579288026e-06,
      "loss": 0.0188,
      "step": 265630
    },
    {
      "epoch": 103.16,
      "learning_rate": 6.2451779935275084e-06,
      "loss": 0.1378,
      "step": 265640
    },
    {
      "epoch": 103.17,
      "learning_rate": 6.244660194174757e-06,
      "loss": 0.1106,
      "step": 265650
    },
    {
      "epoch": 103.17,
      "learning_rate": 6.244142394822007e-06,
      "loss": 0.0564,
      "step": 265660
    },
    {
      "epoch": 103.17,
      "learning_rate": 6.2436245954692565e-06,
      "loss": 0.1288,
      "step": 265670
    },
    {
      "epoch": 103.18,
      "learning_rate": 6.243106796116505e-06,
      "loss": 0.0121,
      "step": 265680
    },
    {
      "epoch": 103.18,
      "learning_rate": 6.242588996763754e-06,
      "loss": 0.0427,
      "step": 265690
    },
    {
      "epoch": 103.18,
      "learning_rate": 6.242071197411004e-06,
      "loss": 0.034,
      "step": 265700
    },
    {
      "epoch": 103.19,
      "learning_rate": 6.241553398058253e-06,
      "loss": 0.0097,
      "step": 265710
    },
    {
      "epoch": 103.19,
      "learning_rate": 6.241035598705501e-06,
      "loss": 0.0836,
      "step": 265720
    },
    {
      "epoch": 103.2,
      "learning_rate": 6.240517799352751e-06,
      "loss": 0.0313,
      "step": 265730
    },
    {
      "epoch": 103.2,
      "learning_rate": 6.24e-06,
      "loss": 0.0329,
      "step": 265740
    },
    {
      "epoch": 103.2,
      "learning_rate": 6.23948220064725e-06,
      "loss": 0.0446,
      "step": 265750
    },
    {
      "epoch": 103.21,
      "learning_rate": 6.2389644012945e-06,
      "loss": 0.0269,
      "step": 265760
    },
    {
      "epoch": 103.21,
      "learning_rate": 6.2384466019417476e-06,
      "loss": 0.0457,
      "step": 265770
    },
    {
      "epoch": 103.22,
      "learning_rate": 6.237928802588997e-06,
      "loss": 0.0993,
      "step": 265780
    },
    {
      "epoch": 103.22,
      "learning_rate": 6.237411003236247e-06,
      "loss": 0.0063,
      "step": 265790
    },
    {
      "epoch": 103.22,
      "learning_rate": 6.236893203883496e-06,
      "loss": 0.0411,
      "step": 265800
    },
    {
      "epoch": 103.23,
      "learning_rate": 6.236375404530744e-06,
      "loss": 0.0334,
      "step": 265810
    },
    {
      "epoch": 103.23,
      "learning_rate": 6.235857605177994e-06,
      "loss": 0.0476,
      "step": 265820
    },
    {
      "epoch": 103.23,
      "learning_rate": 6.2353398058252436e-06,
      "loss": 0.09,
      "step": 265830
    },
    {
      "epoch": 103.24,
      "learning_rate": 6.234822006472493e-06,
      "loss": 0.0147,
      "step": 265840
    },
    {
      "epoch": 103.24,
      "learning_rate": 6.234304207119741e-06,
      "loss": 0.026,
      "step": 265850
    },
    {
      "epoch": 103.25,
      "learning_rate": 6.233786407766991e-06,
      "loss": 0.1581,
      "step": 265860
    },
    {
      "epoch": 103.25,
      "learning_rate": 6.23326860841424e-06,
      "loss": 0.0542,
      "step": 265870
    },
    {
      "epoch": 103.25,
      "learning_rate": 6.23275080906149e-06,
      "loss": 0.0244,
      "step": 265880
    },
    {
      "epoch": 103.26,
      "learning_rate": 6.232233009708738e-06,
      "loss": 0.1223,
      "step": 265890
    },
    {
      "epoch": 103.26,
      "learning_rate": 6.2317152103559875e-06,
      "loss": 0.0121,
      "step": 265900
    },
    {
      "epoch": 103.27,
      "learning_rate": 6.231197411003237e-06,
      "loss": 0.063,
      "step": 265910
    },
    {
      "epoch": 103.27,
      "learning_rate": 6.230679611650487e-06,
      "loss": 0.1375,
      "step": 265920
    },
    {
      "epoch": 103.27,
      "learning_rate": 6.230161812297735e-06,
      "loss": 0.0872,
      "step": 265930
    },
    {
      "epoch": 103.28,
      "learning_rate": 6.229644012944984e-06,
      "loss": 0.114,
      "step": 265940
    },
    {
      "epoch": 103.28,
      "learning_rate": 6.229126213592234e-06,
      "loss": 0.0104,
      "step": 265950
    },
    {
      "epoch": 103.29,
      "learning_rate": 6.2286084142394835e-06,
      "loss": 0.2218,
      "step": 265960
    },
    {
      "epoch": 103.29,
      "learning_rate": 6.2280906148867314e-06,
      "loss": 0.0227,
      "step": 265970
    },
    {
      "epoch": 103.29,
      "learning_rate": 6.227572815533981e-06,
      "loss": 0.0817,
      "step": 265980
    },
    {
      "epoch": 103.3,
      "learning_rate": 6.227055016181231e-06,
      "loss": 0.0381,
      "step": 265990
    },
    {
      "epoch": 103.3,
      "learning_rate": 6.22653721682848e-06,
      "loss": 0.0368,
      "step": 266000
    },
    {
      "epoch": 103.3,
      "learning_rate": 6.226019417475728e-06,
      "loss": 0.0333,
      "step": 266010
    },
    {
      "epoch": 103.31,
      "learning_rate": 6.225501618122978e-06,
      "loss": 0.009,
      "step": 266020
    },
    {
      "epoch": 103.31,
      "learning_rate": 6.2249838187702275e-06,
      "loss": 0.0549,
      "step": 266030
    },
    {
      "epoch": 103.32,
      "learning_rate": 6.224466019417476e-06,
      "loss": 0.0035,
      "step": 266040
    },
    {
      "epoch": 103.32,
      "learning_rate": 6.223948220064725e-06,
      "loss": 0.0497,
      "step": 266050
    },
    {
      "epoch": 103.32,
      "learning_rate": 6.223430420711975e-06,
      "loss": 0.0723,
      "step": 266060
    },
    {
      "epoch": 103.33,
      "learning_rate": 6.222912621359224e-06,
      "loss": 0.0272,
      "step": 266070
    },
    {
      "epoch": 103.33,
      "learning_rate": 6.222394822006473e-06,
      "loss": 0.0177,
      "step": 266080
    },
    {
      "epoch": 103.34,
      "learning_rate": 6.221877022653722e-06,
      "loss": 0.0184,
      "step": 266090
    },
    {
      "epoch": 103.34,
      "learning_rate": 6.221359223300971e-06,
      "loss": 0.0742,
      "step": 266100
    },
    {
      "epoch": 103.34,
      "learning_rate": 6.220841423948221e-06,
      "loss": 0.0107,
      "step": 266110
    },
    {
      "epoch": 103.35,
      "learning_rate": 6.22032362459547e-06,
      "loss": 0.1515,
      "step": 266120
    },
    {
      "epoch": 103.35,
      "learning_rate": 6.2198058252427186e-06,
      "loss": 0.0539,
      "step": 266130
    },
    {
      "epoch": 103.36,
      "learning_rate": 6.219288025889968e-06,
      "loss": 0.0698,
      "step": 266140
    },
    {
      "epoch": 103.36,
      "learning_rate": 6.218770226537218e-06,
      "loss": 0.1244,
      "step": 266150
    },
    {
      "epoch": 103.36,
      "learning_rate": 6.2182524271844666e-06,
      "loss": 0.0699,
      "step": 266160
    },
    {
      "epoch": 103.37,
      "learning_rate": 6.217734627831715e-06,
      "loss": 0.0369,
      "step": 266170
    },
    {
      "epoch": 103.37,
      "learning_rate": 6.217216828478965e-06,
      "loss": 0.1518,
      "step": 266180
    },
    {
      "epoch": 103.37,
      "learning_rate": 6.216699029126214e-06,
      "loss": 0.0382,
      "step": 266190
    },
    {
      "epoch": 103.38,
      "learning_rate": 6.216181229773463e-06,
      "loss": 0.0644,
      "step": 266200
    },
    {
      "epoch": 103.38,
      "learning_rate": 6.215663430420712e-06,
      "loss": 0.1128,
      "step": 266210
    },
    {
      "epoch": 103.39,
      "learning_rate": 6.215145631067962e-06,
      "loss": 0.0266,
      "step": 266220
    },
    {
      "epoch": 103.39,
      "learning_rate": 6.2146278317152105e-06,
      "loss": 0.0316,
      "step": 266230
    },
    {
      "epoch": 103.39,
      "learning_rate": 6.21411003236246e-06,
      "loss": 0.0277,
      "step": 266240
    },
    {
      "epoch": 103.4,
      "learning_rate": 6.213592233009709e-06,
      "loss": 0.0149,
      "step": 266250
    },
    {
      "epoch": 103.4,
      "learning_rate": 6.2130744336569585e-06,
      "loss": 0.0075,
      "step": 266260
    },
    {
      "epoch": 103.41,
      "learning_rate": 6.212556634304207e-06,
      "loss": 0.0409,
      "step": 266270
    },
    {
      "epoch": 103.41,
      "learning_rate": 6.212038834951457e-06,
      "loss": 0.0875,
      "step": 266280
    },
    {
      "epoch": 103.41,
      "learning_rate": 6.2115210355987065e-06,
      "loss": 0.1334,
      "step": 266290
    },
    {
      "epoch": 103.42,
      "learning_rate": 6.211003236245955e-06,
      "loss": 0.1269,
      "step": 266300
    },
    {
      "epoch": 103.42,
      "learning_rate": 6.210485436893204e-06,
      "loss": 0.016,
      "step": 266310
    },
    {
      "epoch": 103.43,
      "learning_rate": 6.209967637540454e-06,
      "loss": 0.0126,
      "step": 266320
    },
    {
      "epoch": 103.43,
      "learning_rate": 6.209449838187703e-06,
      "loss": 0.0497,
      "step": 266330
    },
    {
      "epoch": 103.43,
      "learning_rate": 6.208932038834952e-06,
      "loss": 0.1382,
      "step": 266340
    },
    {
      "epoch": 103.44,
      "learning_rate": 6.208414239482201e-06,
      "loss": 0.08,
      "step": 266350
    },
    {
      "epoch": 103.44,
      "learning_rate": 6.2078964401294505e-06,
      "loss": 0.0654,
      "step": 266360
    },
    {
      "epoch": 103.44,
      "learning_rate": 6.2073786407767e-06,
      "loss": 0.0025,
      "step": 266370
    },
    {
      "epoch": 103.45,
      "learning_rate": 6.206860841423948e-06,
      "loss": 0.0797,
      "step": 266380
    },
    {
      "epoch": 103.45,
      "learning_rate": 6.206343042071198e-06,
      "loss": 0.0645,
      "step": 266390
    },
    {
      "epoch": 103.46,
      "learning_rate": 6.205825242718447e-06,
      "loss": 0.0025,
      "step": 266400
    },
    {
      "epoch": 103.46,
      "learning_rate": 6.205307443365697e-06,
      "loss": 0.0769,
      "step": 266410
    },
    {
      "epoch": 103.46,
      "learning_rate": 6.204789644012945e-06,
      "loss": 0.0184,
      "step": 266420
    },
    {
      "epoch": 103.47,
      "learning_rate": 6.204271844660194e-06,
      "loss": 0.0203,
      "step": 266430
    },
    {
      "epoch": 103.47,
      "learning_rate": 6.203754045307444e-06,
      "loss": 0.1077,
      "step": 266440
    },
    {
      "epoch": 103.48,
      "learning_rate": 6.203236245954694e-06,
      "loss": 0.0496,
      "step": 266450
    },
    {
      "epoch": 103.48,
      "learning_rate": 6.2027184466019416e-06,
      "loss": 0.0251,
      "step": 266460
    },
    {
      "epoch": 103.48,
      "learning_rate": 6.202200647249191e-06,
      "loss": 0.0242,
      "step": 266470
    },
    {
      "epoch": 103.49,
      "learning_rate": 6.201682847896441e-06,
      "loss": 0.0443,
      "step": 266480
    },
    {
      "epoch": 103.49,
      "learning_rate": 6.20116504854369e-06,
      "loss": 0.0007,
      "step": 266490
    },
    {
      "epoch": 103.5,
      "learning_rate": 6.200647249190938e-06,
      "loss": 0.1417,
      "step": 266500
    },
    {
      "epoch": 103.5,
      "learning_rate": 6.200129449838188e-06,
      "loss": 0.1165,
      "step": 266510
    },
    {
      "epoch": 103.5,
      "learning_rate": 6.1996116504854376e-06,
      "loss": 0.0156,
      "step": 266520
    },
    {
      "epoch": 103.51,
      "learning_rate": 6.199093851132687e-06,
      "loss": 0.0253,
      "step": 266530
    },
    {
      "epoch": 103.51,
      "learning_rate": 6.198576051779935e-06,
      "loss": 0.0356,
      "step": 266540
    },
    {
      "epoch": 103.51,
      "learning_rate": 6.198058252427185e-06,
      "loss": 0.0093,
      "step": 266550
    },
    {
      "epoch": 103.52,
      "learning_rate": 6.197540453074434e-06,
      "loss": 0.0103,
      "step": 266560
    },
    {
      "epoch": 103.52,
      "learning_rate": 6.197022653721684e-06,
      "loss": 0.0811,
      "step": 266570
    },
    {
      "epoch": 103.53,
      "learning_rate": 6.196504854368932e-06,
      "loss": 0.0386,
      "step": 266580
    },
    {
      "epoch": 103.53,
      "learning_rate": 6.1959870550161815e-06,
      "loss": 0.0001,
      "step": 266590
    },
    {
      "epoch": 103.53,
      "learning_rate": 6.195469255663431e-06,
      "loss": 0.0825,
      "step": 266600
    },
    {
      "epoch": 103.54,
      "learning_rate": 6.194951456310681e-06,
      "loss": 0.0545,
      "step": 266610
    },
    {
      "epoch": 103.54,
      "learning_rate": 6.194433656957929e-06,
      "loss": 0.1119,
      "step": 266620
    },
    {
      "epoch": 103.55,
      "learning_rate": 6.193915857605178e-06,
      "loss": 0.041,
      "step": 266630
    },
    {
      "epoch": 103.55,
      "learning_rate": 6.193398058252428e-06,
      "loss": 0.1184,
      "step": 266640
    },
    {
      "epoch": 103.55,
      "learning_rate": 6.1928802588996775e-06,
      "loss": 0.0234,
      "step": 266650
    },
    {
      "epoch": 103.56,
      "learning_rate": 6.1923624595469254e-06,
      "loss": 0.0704,
      "step": 266660
    },
    {
      "epoch": 103.56,
      "learning_rate": 6.191844660194175e-06,
      "loss": 0.0045,
      "step": 266670
    },
    {
      "epoch": 103.57,
      "learning_rate": 6.191326860841425e-06,
      "loss": 0.005,
      "step": 266680
    },
    {
      "epoch": 103.57,
      "learning_rate": 6.190809061488674e-06,
      "loss": 0.0266,
      "step": 266690
    },
    {
      "epoch": 103.57,
      "learning_rate": 6.190291262135922e-06,
      "loss": 0.1506,
      "step": 266700
    },
    {
      "epoch": 103.58,
      "learning_rate": 6.189773462783172e-06,
      "loss": 0.0286,
      "step": 266710
    },
    {
      "epoch": 103.58,
      "learning_rate": 6.1892556634304215e-06,
      "loss": 0.0058,
      "step": 266720
    },
    {
      "epoch": 103.58,
      "learning_rate": 6.188737864077671e-06,
      "loss": 0.0373,
      "step": 266730
    },
    {
      "epoch": 103.59,
      "learning_rate": 6.188220064724919e-06,
      "loss": 0.0345,
      "step": 266740
    },
    {
      "epoch": 103.59,
      "learning_rate": 6.187702265372169e-06,
      "loss": 0.0022,
      "step": 266750
    },
    {
      "epoch": 103.6,
      "learning_rate": 6.187184466019418e-06,
      "loss": 0.0576,
      "step": 266760
    },
    {
      "epoch": 103.6,
      "learning_rate": 6.186666666666668e-06,
      "loss": 0.0247,
      "step": 266770
    },
    {
      "epoch": 103.6,
      "learning_rate": 6.186148867313916e-06,
      "loss": 0.1273,
      "step": 266780
    },
    {
      "epoch": 103.61,
      "learning_rate": 6.185631067961165e-06,
      "loss": 0.0133,
      "step": 266790
    },
    {
      "epoch": 103.61,
      "learning_rate": 6.185113268608415e-06,
      "loss": 0.2013,
      "step": 266800
    },
    {
      "epoch": 103.62,
      "learning_rate": 6.184595469255665e-06,
      "loss": 0.072,
      "step": 266810
    },
    {
      "epoch": 103.62,
      "learning_rate": 6.1840776699029126e-06,
      "loss": 0.0655,
      "step": 266820
    },
    {
      "epoch": 103.62,
      "learning_rate": 6.183559870550162e-06,
      "loss": 0.0227,
      "step": 266830
    },
    {
      "epoch": 103.63,
      "learning_rate": 6.183042071197412e-06,
      "loss": 0.0013,
      "step": 266840
    },
    {
      "epoch": 103.63,
      "learning_rate": 6.1825242718446606e-06,
      "loss": 0.0012,
      "step": 266850
    },
    {
      "epoch": 103.63,
      "learning_rate": 6.18200647249191e-06,
      "loss": 0.0584,
      "step": 266860
    },
    {
      "epoch": 103.64,
      "learning_rate": 6.181488673139159e-06,
      "loss": 0.0224,
      "step": 266870
    },
    {
      "epoch": 103.64,
      "learning_rate": 6.1809708737864086e-06,
      "loss": 0.1171,
      "step": 266880
    },
    {
      "epoch": 103.65,
      "learning_rate": 6.180453074433657e-06,
      "loss": 0.0066,
      "step": 266890
    },
    {
      "epoch": 103.65,
      "learning_rate": 6.179935275080907e-06,
      "loss": 0.1067,
      "step": 266900
    },
    {
      "epoch": 103.65,
      "learning_rate": 6.179417475728156e-06,
      "loss": 0.0375,
      "step": 266910
    },
    {
      "epoch": 103.66,
      "learning_rate": 6.178899676375405e-06,
      "loss": 0.0578,
      "step": 266920
    },
    {
      "epoch": 103.66,
      "learning_rate": 6.178381877022654e-06,
      "loss": 0.0424,
      "step": 266930
    },
    {
      "epoch": 103.67,
      "learning_rate": 6.177864077669904e-06,
      "loss": 0.114,
      "step": 266940
    },
    {
      "epoch": 103.67,
      "learning_rate": 6.1773462783171525e-06,
      "loss": 0.0249,
      "step": 266950
    },
    {
      "epoch": 103.67,
      "learning_rate": 6.176828478964402e-06,
      "loss": 0.0676,
      "step": 266960
    },
    {
      "epoch": 103.68,
      "learning_rate": 6.176310679611651e-06,
      "loss": 0.0262,
      "step": 266970
    },
    {
      "epoch": 103.68,
      "learning_rate": 6.1757928802589005e-06,
      "loss": 0.0196,
      "step": 266980
    },
    {
      "epoch": 103.69,
      "learning_rate": 6.175275080906149e-06,
      "loss": 0.0009,
      "step": 266990
    },
    {
      "epoch": 103.69,
      "learning_rate": 6.174757281553399e-06,
      "loss": 0.0012,
      "step": 267000
    },
    {
      "epoch": 103.69,
      "learning_rate": 6.174239482200648e-06,
      "loss": 0.0873,
      "step": 267010
    },
    {
      "epoch": 103.7,
      "learning_rate": 6.173721682847897e-06,
      "loss": 0.0407,
      "step": 267020
    },
    {
      "epoch": 103.7,
      "learning_rate": 6.173203883495146e-06,
      "loss": 0.0846,
      "step": 267030
    },
    {
      "epoch": 103.7,
      "learning_rate": 6.172686084142395e-06,
      "loss": 0.0321,
      "step": 267040
    },
    {
      "epoch": 103.71,
      "learning_rate": 6.1721682847896445e-06,
      "loss": 0.0137,
      "step": 267050
    },
    {
      "epoch": 103.71,
      "learning_rate": 6.171650485436894e-06,
      "loss": 0.0329,
      "step": 267060
    },
    {
      "epoch": 103.72,
      "learning_rate": 6.171132686084143e-06,
      "loss": 0.0003,
      "step": 267070
    },
    {
      "epoch": 103.72,
      "learning_rate": 6.170614886731392e-06,
      "loss": 0.0466,
      "step": 267080
    },
    {
      "epoch": 103.72,
      "learning_rate": 6.170097087378641e-06,
      "loss": 0.0132,
      "step": 267090
    },
    {
      "epoch": 103.73,
      "learning_rate": 6.169579288025891e-06,
      "loss": 0.0003,
      "step": 267100
    },
    {
      "epoch": 103.73,
      "learning_rate": 6.16906148867314e-06,
      "loss": 0.039,
      "step": 267110
    },
    {
      "epoch": 103.74,
      "learning_rate": 6.168543689320388e-06,
      "loss": 0.076,
      "step": 267120
    },
    {
      "epoch": 103.74,
      "learning_rate": 6.168025889967638e-06,
      "loss": 0.0854,
      "step": 267130
    },
    {
      "epoch": 103.74,
      "learning_rate": 6.167508090614888e-06,
      "loss": 0.0052,
      "step": 267140
    },
    {
      "epoch": 103.75,
      "learning_rate": 6.166990291262136e-06,
      "loss": 0.0942,
      "step": 267150
    },
    {
      "epoch": 103.75,
      "learning_rate": 6.166472491909385e-06,
      "loss": 0.0603,
      "step": 267160
    },
    {
      "epoch": 103.76,
      "learning_rate": 6.165954692556635e-06,
      "loss": 0.0295,
      "step": 267170
    },
    {
      "epoch": 103.76,
      "learning_rate": 6.165436893203884e-06,
      "loss": 0.1152,
      "step": 267180
    },
    {
      "epoch": 103.76,
      "learning_rate": 6.164919093851132e-06,
      "loss": 0.0175,
      "step": 267190
    },
    {
      "epoch": 103.77,
      "learning_rate": 6.164401294498382e-06,
      "loss": 0.0078,
      "step": 267200
    },
    {
      "epoch": 103.77,
      "learning_rate": 6.1638834951456316e-06,
      "loss": 0.001,
      "step": 267210
    },
    {
      "epoch": 103.77,
      "learning_rate": 6.163365695792881e-06,
      "loss": 0.1355,
      "step": 267220
    },
    {
      "epoch": 103.78,
      "learning_rate": 6.162847896440129e-06,
      "loss": 0.1037,
      "step": 267230
    },
    {
      "epoch": 103.78,
      "learning_rate": 6.162330097087379e-06,
      "loss": 0.034,
      "step": 267240
    },
    {
      "epoch": 103.79,
      "learning_rate": 6.161812297734628e-06,
      "loss": 0.031,
      "step": 267250
    },
    {
      "epoch": 103.79,
      "learning_rate": 6.161294498381878e-06,
      "loss": 0.0018,
      "step": 267260
    },
    {
      "epoch": 103.79,
      "learning_rate": 6.160776699029126e-06,
      "loss": 0.0008,
      "step": 267270
    },
    {
      "epoch": 103.8,
      "learning_rate": 6.1602588996763755e-06,
      "loss": 0.0283,
      "step": 267280
    },
    {
      "epoch": 103.8,
      "learning_rate": 6.159741100323625e-06,
      "loss": 0.0239,
      "step": 267290
    },
    {
      "epoch": 103.81,
      "learning_rate": 6.159223300970875e-06,
      "loss": 0.0541,
      "step": 267300
    },
    {
      "epoch": 103.81,
      "learning_rate": 6.158705501618123e-06,
      "loss": 0.103,
      "step": 267310
    },
    {
      "epoch": 103.81,
      "learning_rate": 6.158187702265372e-06,
      "loss": 0.0132,
      "step": 267320
    },
    {
      "epoch": 103.82,
      "learning_rate": 6.157669902912622e-06,
      "loss": 0.0013,
      "step": 267330
    },
    {
      "epoch": 103.82,
      "learning_rate": 6.1571521035598715e-06,
      "loss": 0.0215,
      "step": 267340
    },
    {
      "epoch": 103.83,
      "learning_rate": 6.1566343042071194e-06,
      "loss": 0.0156,
      "step": 267350
    },
    {
      "epoch": 103.83,
      "learning_rate": 6.156116504854369e-06,
      "loss": 0.0717,
      "step": 267360
    },
    {
      "epoch": 103.83,
      "learning_rate": 6.155598705501619e-06,
      "loss": 0.0678,
      "step": 267370
    },
    {
      "epoch": 103.84,
      "learning_rate": 6.155080906148868e-06,
      "loss": 0.0017,
      "step": 267380
    },
    {
      "epoch": 103.84,
      "learning_rate": 6.154563106796116e-06,
      "loss": 0.0133,
      "step": 267390
    },
    {
      "epoch": 103.84,
      "learning_rate": 6.154045307443366e-06,
      "loss": 0.008,
      "step": 267400
    },
    {
      "epoch": 103.85,
      "learning_rate": 6.1535275080906155e-06,
      "loss": 0.0408,
      "step": 267410
    },
    {
      "epoch": 103.85,
      "learning_rate": 6.153009708737865e-06,
      "loss": 0.003,
      "step": 267420
    },
    {
      "epoch": 103.86,
      "learning_rate": 6.152491909385115e-06,
      "loss": 0.0003,
      "step": 267430
    },
    {
      "epoch": 103.86,
      "learning_rate": 6.151974110032363e-06,
      "loss": 0.0006,
      "step": 267440
    },
    {
      "epoch": 103.86,
      "learning_rate": 6.151456310679612e-06,
      "loss": 0.0302,
      "step": 267450
    },
    {
      "epoch": 103.87,
      "learning_rate": 6.150938511326862e-06,
      "loss": 0.0019,
      "step": 267460
    },
    {
      "epoch": 103.87,
      "learning_rate": 6.1504207119741115e-06,
      "loss": 0.1092,
      "step": 267470
    },
    {
      "epoch": 103.88,
      "learning_rate": 6.149902912621359e-06,
      "loss": 0.0109,
      "step": 267480
    },
    {
      "epoch": 103.88,
      "learning_rate": 6.149385113268609e-06,
      "loss": 0.0042,
      "step": 267490
    },
    {
      "epoch": 103.88,
      "learning_rate": 6.148867313915859e-06,
      "loss": 0.0153,
      "step": 267500
    },
    {
      "epoch": 103.89,
      "learning_rate": 6.148349514563107e-06,
      "loss": 0.0308,
      "step": 267510
    },
    {
      "epoch": 103.89,
      "learning_rate": 6.147831715210356e-06,
      "loss": 0.078,
      "step": 267520
    },
    {
      "epoch": 103.9,
      "learning_rate": 6.147313915857606e-06,
      "loss": 0.0088,
      "step": 267530
    },
    {
      "epoch": 103.9,
      "learning_rate": 6.146796116504855e-06,
      "loss": 0.0702,
      "step": 267540
    },
    {
      "epoch": 103.9,
      "learning_rate": 6.146278317152104e-06,
      "loss": 0.02,
      "step": 267550
    },
    {
      "epoch": 103.91,
      "learning_rate": 6.145760517799353e-06,
      "loss": 0.0227,
      "step": 267560
    },
    {
      "epoch": 103.91,
      "learning_rate": 6.1452427184466026e-06,
      "loss": 0.1399,
      "step": 267570
    },
    {
      "epoch": 103.91,
      "learning_rate": 6.144724919093852e-06,
      "loss": 0.0578,
      "step": 267580
    },
    {
      "epoch": 103.92,
      "learning_rate": 6.144207119741101e-06,
      "loss": 0.1169,
      "step": 267590
    },
    {
      "epoch": 103.92,
      "learning_rate": 6.14368932038835e-06,
      "loss": 0.052,
      "step": 267600
    },
    {
      "epoch": 103.93,
      "learning_rate": 6.143171521035599e-06,
      "loss": 0.0868,
      "step": 267610
    },
    {
      "epoch": 103.93,
      "learning_rate": 6.142653721682849e-06,
      "loss": 0.1188,
      "step": 267620
    },
    {
      "epoch": 103.93,
      "learning_rate": 6.142135922330098e-06,
      "loss": 0.0574,
      "step": 267630
    },
    {
      "epoch": 103.94,
      "learning_rate": 6.1416181229773465e-06,
      "loss": 0.0194,
      "step": 267640
    },
    {
      "epoch": 103.94,
      "learning_rate": 6.141100323624596e-06,
      "loss": 0.0115,
      "step": 267650
    },
    {
      "epoch": 103.95,
      "learning_rate": 6.140582524271845e-06,
      "loss": 0.122,
      "step": 267660
    },
    {
      "epoch": 103.95,
      "learning_rate": 6.1400647249190945e-06,
      "loss": 0.0216,
      "step": 267670
    },
    {
      "epoch": 103.95,
      "learning_rate": 6.139546925566343e-06,
      "loss": 0.0182,
      "step": 267680
    },
    {
      "epoch": 103.96,
      "learning_rate": 6.139029126213593e-06,
      "loss": 0.0001,
      "step": 267690
    },
    {
      "epoch": 103.96,
      "learning_rate": 6.138511326860842e-06,
      "loss": 0.0764,
      "step": 267700
    },
    {
      "epoch": 103.97,
      "learning_rate": 6.137993527508091e-06,
      "loss": 0.1348,
      "step": 267710
    },
    {
      "epoch": 103.97,
      "learning_rate": 6.13747572815534e-06,
      "loss": 0.1209,
      "step": 267720
    },
    {
      "epoch": 103.97,
      "learning_rate": 6.13695792880259e-06,
      "loss": 0.0626,
      "step": 267730
    },
    {
      "epoch": 103.98,
      "learning_rate": 6.1364401294498385e-06,
      "loss": 0.0399,
      "step": 267740
    },
    {
      "epoch": 103.98,
      "learning_rate": 6.135922330097088e-06,
      "loss": 0.007,
      "step": 267750
    },
    {
      "epoch": 103.98,
      "learning_rate": 6.135404530744337e-06,
      "loss": 0.0003,
      "step": 267760
    },
    {
      "epoch": 103.99,
      "learning_rate": 6.1348867313915865e-06,
      "loss": 0.0613,
      "step": 267770
    },
    {
      "epoch": 103.99,
      "learning_rate": 6.134368932038835e-06,
      "loss": 0.0017,
      "step": 267780
    },
    {
      "epoch": 104.0,
      "learning_rate": 6.133851132686085e-06,
      "loss": 0.0982,
      "step": 267790
    },
    {
      "epoch": 104.0,
      "learning_rate": 6.133333333333334e-06,
      "loss": 0.0157,
      "step": 267800
    },
    {
      "epoch": 104.0,
      "eval_accuracy": 0.9524071526822558,
      "eval_loss": 0.3646901249885559,
      "eval_runtime": 8.2362,
      "eval_samples_per_second": 441.346,
      "eval_steps_per_second": 55.244,
      "step": 267800
    },
    {
      "epoch": 104.0,
      "learning_rate": 6.132815533980583e-06,
      "loss": 0.0003,
      "step": 267810
    },
    {
      "epoch": 104.01,
      "learning_rate": 6.132297734627832e-06,
      "loss": 0.0245,
      "step": 267820
    },
    {
      "epoch": 104.01,
      "learning_rate": 6.131779935275082e-06,
      "loss": 0.0006,
      "step": 267830
    },
    {
      "epoch": 104.02,
      "learning_rate": 6.13126213592233e-06,
      "loss": 0.0102,
      "step": 267840
    },
    {
      "epoch": 104.02,
      "learning_rate": 6.130744336569579e-06,
      "loss": 0.0646,
      "step": 267850
    },
    {
      "epoch": 104.02,
      "learning_rate": 6.130226537216829e-06,
      "loss": 0.0549,
      "step": 267860
    },
    {
      "epoch": 104.03,
      "learning_rate": 6.129708737864078e-06,
      "loss": 0.0832,
      "step": 267870
    },
    {
      "epoch": 104.03,
      "learning_rate": 6.129190938511327e-06,
      "loss": 0.0409,
      "step": 267880
    },
    {
      "epoch": 104.03,
      "learning_rate": 6.128673139158576e-06,
      "loss": 0.0947,
      "step": 267890
    },
    {
      "epoch": 104.04,
      "learning_rate": 6.1281553398058256e-06,
      "loss": 0.0167,
      "step": 267900
    },
    {
      "epoch": 104.04,
      "learning_rate": 6.127637540453075e-06,
      "loss": 0.0718,
      "step": 267910
    },
    {
      "epoch": 104.05,
      "learning_rate": 6.127119741100324e-06,
      "loss": 0.074,
      "step": 267920
    },
    {
      "epoch": 104.05,
      "learning_rate": 6.126601941747573e-06,
      "loss": 0.0537,
      "step": 267930
    },
    {
      "epoch": 104.05,
      "learning_rate": 6.126084142394822e-06,
      "loss": 0.0127,
      "step": 267940
    },
    {
      "epoch": 104.06,
      "learning_rate": 6.125566343042072e-06,
      "loss": 0.0994,
      "step": 267950
    },
    {
      "epoch": 104.06,
      "learning_rate": 6.125048543689321e-06,
      "loss": 0.0019,
      "step": 267960
    },
    {
      "epoch": 104.07,
      "learning_rate": 6.1245307443365695e-06,
      "loss": 0.1137,
      "step": 267970
    },
    {
      "epoch": 104.07,
      "learning_rate": 6.124012944983819e-06,
      "loss": 0.0628,
      "step": 267980
    },
    {
      "epoch": 104.07,
      "learning_rate": 6.123495145631069e-06,
      "loss": 0.0756,
      "step": 267990
    },
    {
      "epoch": 104.08,
      "learning_rate": 6.122977346278318e-06,
      "loss": 0.0524,
      "step": 268000
    },
    {
      "epoch": 104.08,
      "learning_rate": 6.122459546925566e-06,
      "loss": 0.0585,
      "step": 268010
    },
    {
      "epoch": 104.09,
      "learning_rate": 6.121941747572816e-06,
      "loss": 0.0701,
      "step": 268020
    },
    {
      "epoch": 104.09,
      "learning_rate": 6.1214239482200655e-06,
      "loss": 0.0101,
      "step": 268030
    },
    {
      "epoch": 104.09,
      "learning_rate": 6.120906148867315e-06,
      "loss": 0.0013,
      "step": 268040
    },
    {
      "epoch": 104.1,
      "learning_rate": 6.120388349514563e-06,
      "loss": 0.0127,
      "step": 268050
    },
    {
      "epoch": 104.1,
      "learning_rate": 6.119870550161813e-06,
      "loss": 0.1034,
      "step": 268060
    },
    {
      "epoch": 104.1,
      "learning_rate": 6.119352750809062e-06,
      "loss": 0.0002,
      "step": 268070
    },
    {
      "epoch": 104.11,
      "learning_rate": 6.118834951456312e-06,
      "loss": 0.0188,
      "step": 268080
    },
    {
      "epoch": 104.11,
      "learning_rate": 6.11831715210356e-06,
      "loss": 0.0002,
      "step": 268090
    },
    {
      "epoch": 104.12,
      "learning_rate": 6.1177993527508095e-06,
      "loss": 0.0032,
      "step": 268100
    },
    {
      "epoch": 104.12,
      "learning_rate": 6.117281553398059e-06,
      "loss": 0.1262,
      "step": 268110
    },
    {
      "epoch": 104.12,
      "learning_rate": 6.116763754045309e-06,
      "loss": 0.0109,
      "step": 268120
    },
    {
      "epoch": 104.13,
      "learning_rate": 6.116245954692557e-06,
      "loss": 0.0069,
      "step": 268130
    },
    {
      "epoch": 104.13,
      "learning_rate": 6.115728155339806e-06,
      "loss": 0.084,
      "step": 268140
    },
    {
      "epoch": 104.14,
      "learning_rate": 6.115210355987056e-06,
      "loss": 0.0152,
      "step": 268150
    },
    {
      "epoch": 104.14,
      "learning_rate": 6.1146925566343055e-06,
      "loss": 0.0561,
      "step": 268160
    },
    {
      "epoch": 104.14,
      "learning_rate": 6.114174757281553e-06,
      "loss": 0.0036,
      "step": 268170
    },
    {
      "epoch": 104.15,
      "learning_rate": 6.113656957928803e-06,
      "loss": 0.072,
      "step": 268180
    },
    {
      "epoch": 104.15,
      "learning_rate": 6.113139158576053e-06,
      "loss": 0.0097,
      "step": 268190
    },
    {
      "epoch": 104.16,
      "learning_rate": 6.112621359223302e-06,
      "loss": 0.1761,
      "step": 268200
    },
    {
      "epoch": 104.16,
      "learning_rate": 6.11210355987055e-06,
      "loss": 0.0627,
      "step": 268210
    },
    {
      "epoch": 104.16,
      "learning_rate": 6.1115857605178e-06,
      "loss": 0.092,
      "step": 268220
    },
    {
      "epoch": 104.17,
      "learning_rate": 6.111067961165049e-06,
      "loss": 0.031,
      "step": 268230
    },
    {
      "epoch": 104.17,
      "learning_rate": 6.110550161812299e-06,
      "loss": 0.0014,
      "step": 268240
    },
    {
      "epoch": 104.17,
      "learning_rate": 6.110032362459547e-06,
      "loss": 0.0924,
      "step": 268250
    },
    {
      "epoch": 104.18,
      "learning_rate": 6.1095145631067966e-06,
      "loss": 0.024,
      "step": 268260
    },
    {
      "epoch": 104.18,
      "learning_rate": 6.108996763754046e-06,
      "loss": 0.0005,
      "step": 268270
    },
    {
      "epoch": 104.19,
      "learning_rate": 6.108478964401296e-06,
      "loss": 0.0006,
      "step": 268280
    },
    {
      "epoch": 104.19,
      "learning_rate": 6.107961165048544e-06,
      "loss": 0.0813,
      "step": 268290
    },
    {
      "epoch": 104.19,
      "learning_rate": 6.107443365695793e-06,
      "loss": 0.0687,
      "step": 268300
    },
    {
      "epoch": 104.2,
      "learning_rate": 6.106925566343043e-06,
      "loss": 0.0471,
      "step": 268310
    },
    {
      "epoch": 104.2,
      "learning_rate": 6.106407766990292e-06,
      "loss": 0.0014,
      "step": 268320
    },
    {
      "epoch": 104.21,
      "learning_rate": 6.1058899676375405e-06,
      "loss": 0.0628,
      "step": 268330
    },
    {
      "epoch": 104.21,
      "learning_rate": 6.10537216828479e-06,
      "loss": 0.0967,
      "step": 268340
    },
    {
      "epoch": 104.21,
      "learning_rate": 6.10485436893204e-06,
      "loss": 0.0377,
      "step": 268350
    },
    {
      "epoch": 104.22,
      "learning_rate": 6.1043365695792885e-06,
      "loss": 0.1229,
      "step": 268360
    },
    {
      "epoch": 104.22,
      "learning_rate": 6.103818770226537e-06,
      "loss": 0.0575,
      "step": 268370
    },
    {
      "epoch": 104.23,
      "learning_rate": 6.103300970873787e-06,
      "loss": 0.0198,
      "step": 268380
    },
    {
      "epoch": 104.23,
      "learning_rate": 6.1027831715210365e-06,
      "loss": 0.1508,
      "step": 268390
    },
    {
      "epoch": 104.23,
      "learning_rate": 6.102265372168285e-06,
      "loss": 0.1169,
      "step": 268400
    },
    {
      "epoch": 104.24,
      "learning_rate": 6.101747572815534e-06,
      "loss": 0.1612,
      "step": 268410
    },
    {
      "epoch": 104.24,
      "learning_rate": 6.101229773462784e-06,
      "loss": 0.0838,
      "step": 268420
    },
    {
      "epoch": 104.24,
      "learning_rate": 6.100711974110033e-06,
      "loss": 0.0473,
      "step": 268430
    },
    {
      "epoch": 104.25,
      "learning_rate": 6.100194174757282e-06,
      "loss": 0.031,
      "step": 268440
    },
    {
      "epoch": 104.25,
      "learning_rate": 6.099676375404531e-06,
      "loss": 0.0432,
      "step": 268450
    },
    {
      "epoch": 104.26,
      "learning_rate": 6.0991585760517805e-06,
      "loss": 0.0467,
      "step": 268460
    },
    {
      "epoch": 104.26,
      "learning_rate": 6.098640776699029e-06,
      "loss": 0.0351,
      "step": 268470
    },
    {
      "epoch": 104.26,
      "learning_rate": 6.098122977346279e-06,
      "loss": 0.0004,
      "step": 268480
    },
    {
      "epoch": 104.27,
      "learning_rate": 6.097605177993528e-06,
      "loss": 0.0189,
      "step": 268490
    },
    {
      "epoch": 104.27,
      "learning_rate": 6.097087378640777e-06,
      "loss": 0.0009,
      "step": 268500
    },
    {
      "epoch": 104.28,
      "learning_rate": 6.096569579288026e-06,
      "loss": 0.0242,
      "step": 268510
    },
    {
      "epoch": 104.28,
      "learning_rate": 6.096051779935276e-06,
      "loss": 0.0018,
      "step": 268520
    },
    {
      "epoch": 104.28,
      "learning_rate": 6.095533980582525e-06,
      "loss": 0.0342,
      "step": 268530
    },
    {
      "epoch": 104.29,
      "learning_rate": 6.095016181229774e-06,
      "loss": 0.0869,
      "step": 268540
    },
    {
      "epoch": 104.29,
      "learning_rate": 6.094498381877023e-06,
      "loss": 0.0049,
      "step": 268550
    },
    {
      "epoch": 104.3,
      "learning_rate": 6.093980582524272e-06,
      "loss": 0.1104,
      "step": 268560
    },
    {
      "epoch": 104.3,
      "learning_rate": 6.093462783171522e-06,
      "loss": 0.0338,
      "step": 268570
    },
    {
      "epoch": 104.3,
      "learning_rate": 6.092944983818771e-06,
      "loss": 0.1001,
      "step": 268580
    },
    {
      "epoch": 104.31,
      "learning_rate": 6.0924271844660196e-06,
      "loss": 0.0393,
      "step": 268590
    },
    {
      "epoch": 104.31,
      "learning_rate": 6.091909385113269e-06,
      "loss": 0.0612,
      "step": 268600
    },
    {
      "epoch": 104.31,
      "learning_rate": 6.091391585760519e-06,
      "loss": 0.0667,
      "step": 268610
    },
    {
      "epoch": 104.32,
      "learning_rate": 6.0908737864077676e-06,
      "loss": 0.0224,
      "step": 268620
    },
    {
      "epoch": 104.32,
      "learning_rate": 6.090355987055016e-06,
      "loss": 0.1132,
      "step": 268630
    },
    {
      "epoch": 104.33,
      "learning_rate": 6.089838187702266e-06,
      "loss": 0.0145,
      "step": 268640
    },
    {
      "epoch": 104.33,
      "learning_rate": 6.089320388349516e-06,
      "loss": 0.0001,
      "step": 268650
    },
    {
      "epoch": 104.33,
      "learning_rate": 6.0888025889967635e-06,
      "loss": 0.0631,
      "step": 268660
    },
    {
      "epoch": 104.34,
      "learning_rate": 6.088284789644013e-06,
      "loss": 0.0137,
      "step": 268670
    },
    {
      "epoch": 104.34,
      "learning_rate": 6.087766990291263e-06,
      "loss": 0.0291,
      "step": 268680
    },
    {
      "epoch": 104.35,
      "learning_rate": 6.087249190938512e-06,
      "loss": 0.0918,
      "step": 268690
    },
    {
      "epoch": 104.35,
      "learning_rate": 6.08673139158576e-06,
      "loss": 0.0774,
      "step": 268700
    },
    {
      "epoch": 104.35,
      "learning_rate": 6.08621359223301e-06,
      "loss": 0.0225,
      "step": 268710
    },
    {
      "epoch": 104.36,
      "learning_rate": 6.0856957928802595e-06,
      "loss": 0.0912,
      "step": 268720
    },
    {
      "epoch": 104.36,
      "learning_rate": 6.085177993527509e-06,
      "loss": 0.0006,
      "step": 268730
    },
    {
      "epoch": 104.37,
      "learning_rate": 6.084660194174757e-06,
      "loss": 0.1603,
      "step": 268740
    },
    {
      "epoch": 104.37,
      "learning_rate": 6.084142394822007e-06,
      "loss": 0.029,
      "step": 268750
    },
    {
      "epoch": 104.37,
      "learning_rate": 6.083624595469256e-06,
      "loss": 0.0139,
      "step": 268760
    },
    {
      "epoch": 104.38,
      "learning_rate": 6.083106796116506e-06,
      "loss": 0.0131,
      "step": 268770
    },
    {
      "epoch": 104.38,
      "learning_rate": 6.082588996763754e-06,
      "loss": 0.0278,
      "step": 268780
    },
    {
      "epoch": 104.38,
      "learning_rate": 6.0820711974110035e-06,
      "loss": 0.0676,
      "step": 268790
    },
    {
      "epoch": 104.39,
      "learning_rate": 6.081553398058253e-06,
      "loss": 0.0346,
      "step": 268800
    },
    {
      "epoch": 104.39,
      "learning_rate": 6.081035598705503e-06,
      "loss": 0.1286,
      "step": 268810
    },
    {
      "epoch": 104.4,
      "learning_rate": 6.080517799352751e-06,
      "loss": 0.0223,
      "step": 268820
    },
    {
      "epoch": 104.4,
      "learning_rate": 6.08e-06,
      "loss": 0.0001,
      "step": 268830
    },
    {
      "epoch": 104.4,
      "learning_rate": 6.07948220064725e-06,
      "loss": 0.026,
      "step": 268840
    },
    {
      "epoch": 104.41,
      "learning_rate": 6.0789644012944995e-06,
      "loss": 0.0003,
      "step": 268850
    },
    {
      "epoch": 104.41,
      "learning_rate": 6.078446601941747e-06,
      "loss": 0.1449,
      "step": 268860
    },
    {
      "epoch": 104.42,
      "learning_rate": 6.077928802588997e-06,
      "loss": 0.1069,
      "step": 268870
    },
    {
      "epoch": 104.42,
      "learning_rate": 6.077411003236247e-06,
      "loss": 0.0004,
      "step": 268880
    },
    {
      "epoch": 104.42,
      "learning_rate": 6.076893203883496e-06,
      "loss": 0.0576,
      "step": 268890
    },
    {
      "epoch": 104.43,
      "learning_rate": 6.076375404530744e-06,
      "loss": 0.0002,
      "step": 268900
    },
    {
      "epoch": 104.43,
      "learning_rate": 6.075857605177994e-06,
      "loss": 0.0134,
      "step": 268910
    },
    {
      "epoch": 104.43,
      "learning_rate": 6.075339805825243e-06,
      "loss": 0.0346,
      "step": 268920
    },
    {
      "epoch": 104.44,
      "learning_rate": 6.074822006472493e-06,
      "loss": 0.0705,
      "step": 268930
    },
    {
      "epoch": 104.44,
      "learning_rate": 6.074304207119741e-06,
      "loss": 0.0755,
      "step": 268940
    },
    {
      "epoch": 104.45,
      "learning_rate": 6.0737864077669906e-06,
      "loss": 0.0001,
      "step": 268950
    },
    {
      "epoch": 104.45,
      "learning_rate": 6.07326860841424e-06,
      "loss": 0.0545,
      "step": 268960
    },
    {
      "epoch": 104.45,
      "learning_rate": 6.07275080906149e-06,
      "loss": 0.0069,
      "step": 268970
    },
    {
      "epoch": 104.46,
      "learning_rate": 6.072233009708738e-06,
      "loss": 0.0157,
      "step": 268980
    },
    {
      "epoch": 104.46,
      "learning_rate": 6.071715210355987e-06,
      "loss": 0.031,
      "step": 268990
    },
    {
      "epoch": 104.47,
      "learning_rate": 6.071197411003237e-06,
      "loss": 0.119,
      "step": 269000
    },
    {
      "epoch": 104.47,
      "learning_rate": 6.070679611650487e-06,
      "loss": 0.1231,
      "step": 269010
    },
    {
      "epoch": 104.47,
      "learning_rate": 6.0701618122977345e-06,
      "loss": 0.0721,
      "step": 269020
    },
    {
      "epoch": 104.48,
      "learning_rate": 6.069644012944984e-06,
      "loss": 0.0351,
      "step": 269030
    },
    {
      "epoch": 104.48,
      "learning_rate": 6.069126213592234e-06,
      "loss": 0.0268,
      "step": 269040
    },
    {
      "epoch": 104.49,
      "learning_rate": 6.068608414239483e-06,
      "loss": 0.0126,
      "step": 269050
    },
    {
      "epoch": 104.49,
      "learning_rate": 6.068090614886731e-06,
      "loss": 0.0281,
      "step": 269060
    },
    {
      "epoch": 104.49,
      "learning_rate": 6.067572815533981e-06,
      "loss": 0.0105,
      "step": 269070
    },
    {
      "epoch": 104.5,
      "learning_rate": 6.0670550161812305e-06,
      "loss": 0.1915,
      "step": 269080
    },
    {
      "epoch": 104.5,
      "learning_rate": 6.06653721682848e-06,
      "loss": 0.0002,
      "step": 269090
    },
    {
      "epoch": 104.5,
      "learning_rate": 6.066019417475729e-06,
      "loss": 0.0522,
      "step": 269100
    },
    {
      "epoch": 104.51,
      "learning_rate": 6.065501618122978e-06,
      "loss": 0.0284,
      "step": 269110
    },
    {
      "epoch": 104.51,
      "learning_rate": 6.064983818770227e-06,
      "loss": 0.0051,
      "step": 269120
    },
    {
      "epoch": 104.52,
      "learning_rate": 6.064466019417476e-06,
      "loss": 0.003,
      "step": 269130
    },
    {
      "epoch": 104.52,
      "learning_rate": 6.063948220064726e-06,
      "loss": 0.0448,
      "step": 269140
    },
    {
      "epoch": 104.52,
      "learning_rate": 6.0634304207119745e-06,
      "loss": 0.21,
      "step": 269150
    },
    {
      "epoch": 104.53,
      "learning_rate": 6.062912621359224e-06,
      "loss": 0.0845,
      "step": 269160
    },
    {
      "epoch": 104.53,
      "learning_rate": 6.062394822006473e-06,
      "loss": 0.0148,
      "step": 269170
    },
    {
      "epoch": 104.54,
      "learning_rate": 6.0618770226537225e-06,
      "loss": 0.0479,
      "step": 269180
    },
    {
      "epoch": 104.54,
      "learning_rate": 6.061359223300971e-06,
      "loss": 0.0297,
      "step": 269190
    },
    {
      "epoch": 104.54,
      "learning_rate": 6.060841423948221e-06,
      "loss": 0.0187,
      "step": 269200
    },
    {
      "epoch": 104.55,
      "learning_rate": 6.06032362459547e-06,
      "loss": 0.1585,
      "step": 269210
    },
    {
      "epoch": 104.55,
      "learning_rate": 6.059805825242719e-06,
      "loss": 0.0206,
      "step": 269220
    },
    {
      "epoch": 104.56,
      "learning_rate": 6.059288025889968e-06,
      "loss": 0.1669,
      "step": 269230
    },
    {
      "epoch": 104.56,
      "learning_rate": 6.058770226537218e-06,
      "loss": 0.0258,
      "step": 269240
    },
    {
      "epoch": 104.56,
      "learning_rate": 6.058252427184466e-06,
      "loss": 0.0225,
      "step": 269250
    },
    {
      "epoch": 104.57,
      "learning_rate": 6.057734627831716e-06,
      "loss": 0.1364,
      "step": 269260
    },
    {
      "epoch": 104.57,
      "learning_rate": 6.057216828478965e-06,
      "loss": 0.0076,
      "step": 269270
    },
    {
      "epoch": 104.57,
      "learning_rate": 6.056699029126214e-06,
      "loss": 0.0295,
      "step": 269280
    },
    {
      "epoch": 104.58,
      "learning_rate": 6.056181229773463e-06,
      "loss": 0.1768,
      "step": 269290
    },
    {
      "epoch": 104.58,
      "learning_rate": 6.055663430420713e-06,
      "loss": 0.0659,
      "step": 269300
    },
    {
      "epoch": 104.59,
      "learning_rate": 6.0551456310679616e-06,
      "loss": 0.0286,
      "step": 269310
    },
    {
      "epoch": 104.59,
      "learning_rate": 6.05462783171521e-06,
      "loss": 0.0096,
      "step": 269320
    },
    {
      "epoch": 104.59,
      "learning_rate": 6.05411003236246e-06,
      "loss": 0.1171,
      "step": 269330
    },
    {
      "epoch": 104.6,
      "learning_rate": 6.05359223300971e-06,
      "loss": 0.058,
      "step": 269340
    },
    {
      "epoch": 104.6,
      "learning_rate": 6.053074433656958e-06,
      "loss": 0.0981,
      "step": 269350
    },
    {
      "epoch": 104.61,
      "learning_rate": 6.052556634304207e-06,
      "loss": 0.0468,
      "step": 269360
    },
    {
      "epoch": 104.61,
      "learning_rate": 6.052038834951457e-06,
      "loss": 0.1009,
      "step": 269370
    },
    {
      "epoch": 104.61,
      "learning_rate": 6.051521035598706e-06,
      "loss": 0.0693,
      "step": 269380
    },
    {
      "epoch": 104.62,
      "learning_rate": 6.051003236245955e-06,
      "loss": 0.0743,
      "step": 269390
    },
    {
      "epoch": 104.62,
      "learning_rate": 6.050485436893204e-06,
      "loss": 0.0001,
      "step": 269400
    },
    {
      "epoch": 104.63,
      "learning_rate": 6.0499676375404535e-06,
      "loss": 0.0304,
      "step": 269410
    },
    {
      "epoch": 104.63,
      "learning_rate": 6.049449838187703e-06,
      "loss": 0.0011,
      "step": 269420
    },
    {
      "epoch": 104.63,
      "learning_rate": 6.048932038834952e-06,
      "loss": 0.0164,
      "step": 269430
    },
    {
      "epoch": 104.64,
      "learning_rate": 6.048414239482201e-06,
      "loss": 0.0417,
      "step": 269440
    },
    {
      "epoch": 104.64,
      "learning_rate": 6.04789644012945e-06,
      "loss": 0.0065,
      "step": 269450
    },
    {
      "epoch": 104.64,
      "learning_rate": 6.0473786407767e-06,
      "loss": 0.0256,
      "step": 269460
    },
    {
      "epoch": 104.65,
      "learning_rate": 6.046860841423948e-06,
      "loss": 0.0324,
      "step": 269470
    },
    {
      "epoch": 104.65,
      "learning_rate": 6.0463430420711975e-06,
      "loss": 0.0325,
      "step": 269480
    },
    {
      "epoch": 104.66,
      "learning_rate": 6.045825242718447e-06,
      "loss": 0.1281,
      "step": 269490
    },
    {
      "epoch": 104.66,
      "learning_rate": 6.045307443365697e-06,
      "loss": 0.1012,
      "step": 269500
    },
    {
      "epoch": 104.66,
      "learning_rate": 6.044789644012945e-06,
      "loss": 0.0005,
      "step": 269510
    },
    {
      "epoch": 104.67,
      "learning_rate": 6.044271844660194e-06,
      "loss": 0.0269,
      "step": 269520
    },
    {
      "epoch": 104.67,
      "learning_rate": 6.043754045307444e-06,
      "loss": 0.0152,
      "step": 269530
    },
    {
      "epoch": 104.68,
      "learning_rate": 6.0432362459546935e-06,
      "loss": 0.0217,
      "step": 269540
    },
    {
      "epoch": 104.68,
      "learning_rate": 6.042718446601941e-06,
      "loss": 0.0669,
      "step": 269550
    },
    {
      "epoch": 104.68,
      "learning_rate": 6.042200647249191e-06,
      "loss": 0.0047,
      "step": 269560
    },
    {
      "epoch": 104.69,
      "learning_rate": 6.041682847896441e-06,
      "loss": 0.0012,
      "step": 269570
    },
    {
      "epoch": 104.69,
      "learning_rate": 6.04116504854369e-06,
      "loss": 0.0502,
      "step": 269580
    },
    {
      "epoch": 104.7,
      "learning_rate": 6.040647249190938e-06,
      "loss": 0.0895,
      "step": 269590
    },
    {
      "epoch": 104.7,
      "learning_rate": 6.040129449838188e-06,
      "loss": 0.0814,
      "step": 269600
    },
    {
      "epoch": 104.7,
      "learning_rate": 6.039611650485437e-06,
      "loss": 0.0201,
      "step": 269610
    },
    {
      "epoch": 104.71,
      "learning_rate": 6.039093851132687e-06,
      "loss": 0.0133,
      "step": 269620
    },
    {
      "epoch": 104.71,
      "learning_rate": 6.038576051779935e-06,
      "loss": 0.0716,
      "step": 269630
    },
    {
      "epoch": 104.71,
      "learning_rate": 6.0380582524271846e-06,
      "loss": 0.0835,
      "step": 269640
    },
    {
      "epoch": 104.72,
      "learning_rate": 6.037540453074434e-06,
      "loss": 0.0299,
      "step": 269650
    },
    {
      "epoch": 104.72,
      "learning_rate": 6.037022653721684e-06,
      "loss": 0.0575,
      "step": 269660
    },
    {
      "epoch": 104.73,
      "learning_rate": 6.036504854368933e-06,
      "loss": 0.1117,
      "step": 269670
    },
    {
      "epoch": 104.73,
      "learning_rate": 6.035987055016181e-06,
      "loss": 0.0413,
      "step": 269680
    },
    {
      "epoch": 104.73,
      "learning_rate": 6.035469255663431e-06,
      "loss": 0.0157,
      "step": 269690
    },
    {
      "epoch": 104.74,
      "learning_rate": 6.034951456310681e-06,
      "loss": 0.0193,
      "step": 269700
    },
    {
      "epoch": 104.74,
      "learning_rate": 6.03443365695793e-06,
      "loss": 0.0114,
      "step": 269710
    },
    {
      "epoch": 104.75,
      "learning_rate": 6.033915857605178e-06,
      "loss": 0.0409,
      "step": 269720
    },
    {
      "epoch": 104.75,
      "learning_rate": 6.033398058252428e-06,
      "loss": 0.0337,
      "step": 269730
    },
    {
      "epoch": 104.75,
      "learning_rate": 6.032880258899677e-06,
      "loss": 0.0526,
      "step": 269740
    },
    {
      "epoch": 104.76,
      "learning_rate": 6.032362459546927e-06,
      "loss": 0.0171,
      "step": 269750
    },
    {
      "epoch": 104.76,
      "learning_rate": 6.031844660194175e-06,
      "loss": 0.0075,
      "step": 269760
    },
    {
      "epoch": 104.77,
      "learning_rate": 6.0313268608414245e-06,
      "loss": 0.0951,
      "step": 269770
    },
    {
      "epoch": 104.77,
      "learning_rate": 6.030809061488674e-06,
      "loss": 0.0903,
      "step": 269780
    },
    {
      "epoch": 104.77,
      "learning_rate": 6.030291262135923e-06,
      "loss": 0.0346,
      "step": 269790
    },
    {
      "epoch": 104.78,
      "learning_rate": 6.029773462783172e-06,
      "loss": 0.009,
      "step": 269800
    },
    {
      "epoch": 104.78,
      "learning_rate": 6.029255663430421e-06,
      "loss": 0.0765,
      "step": 269810
    },
    {
      "epoch": 104.78,
      "learning_rate": 6.028737864077671e-06,
      "loss": 0.0485,
      "step": 269820
    },
    {
      "epoch": 104.79,
      "learning_rate": 6.02822006472492e-06,
      "loss": 0.1918,
      "step": 269830
    },
    {
      "epoch": 104.79,
      "learning_rate": 6.0277022653721685e-06,
      "loss": 0.0007,
      "step": 269840
    },
    {
      "epoch": 104.8,
      "learning_rate": 6.027184466019418e-06,
      "loss": 0.0103,
      "step": 269850
    },
    {
      "epoch": 104.8,
      "learning_rate": 6.026666666666668e-06,
      "loss": 0.0211,
      "step": 269860
    },
    {
      "epoch": 104.8,
      "learning_rate": 6.0261488673139165e-06,
      "loss": 0.1003,
      "step": 269870
    },
    {
      "epoch": 104.81,
      "learning_rate": 6.025631067961165e-06,
      "loss": 0.0441,
      "step": 269880
    },
    {
      "epoch": 104.81,
      "learning_rate": 6.025113268608415e-06,
      "loss": 0.0006,
      "step": 269890
    },
    {
      "epoch": 104.82,
      "learning_rate": 6.0245954692556645e-06,
      "loss": 0.0205,
      "step": 269900
    },
    {
      "epoch": 104.82,
      "learning_rate": 6.024077669902913e-06,
      "loss": 0.0206,
      "step": 269910
    },
    {
      "epoch": 104.82,
      "learning_rate": 6.023559870550162e-06,
      "loss": 0.0288,
      "step": 269920
    },
    {
      "epoch": 104.83,
      "learning_rate": 6.023042071197412e-06,
      "loss": 0.0003,
      "step": 269930
    },
    {
      "epoch": 104.83,
      "learning_rate": 6.02252427184466e-06,
      "loss": 0.0109,
      "step": 269940
    },
    {
      "epoch": 104.83,
      "learning_rate": 6.02200647249191e-06,
      "loss": 0.0863,
      "step": 269950
    },
    {
      "epoch": 104.84,
      "learning_rate": 6.021488673139159e-06,
      "loss": 0.0008,
      "step": 269960
    },
    {
      "epoch": 104.84,
      "learning_rate": 6.020970873786408e-06,
      "loss": 0.0978,
      "step": 269970
    },
    {
      "epoch": 104.85,
      "learning_rate": 6.020453074433657e-06,
      "loss": 0.0231,
      "step": 269980
    },
    {
      "epoch": 104.85,
      "learning_rate": 6.019935275080907e-06,
      "loss": 0.0428,
      "step": 269990
    },
    {
      "epoch": 104.85,
      "learning_rate": 6.0194174757281556e-06,
      "loss": 0.039,
      "step": 270000
    },
    {
      "epoch": 104.86,
      "learning_rate": 6.018899676375405e-06,
      "loss": 0.0337,
      "step": 270010
    },
    {
      "epoch": 104.86,
      "learning_rate": 6.018381877022654e-06,
      "loss": 0.0373,
      "step": 270020
    },
    {
      "epoch": 104.87,
      "learning_rate": 6.0178640776699036e-06,
      "loss": 0.1072,
      "step": 270030
    },
    {
      "epoch": 104.87,
      "learning_rate": 6.017346278317152e-06,
      "loss": 0.0472,
      "step": 270040
    },
    {
      "epoch": 104.87,
      "learning_rate": 6.016828478964402e-06,
      "loss": 0.039,
      "step": 270050
    },
    {
      "epoch": 104.88,
      "learning_rate": 6.016310679611651e-06,
      "loss": 0.0234,
      "step": 270060
    },
    {
      "epoch": 104.88,
      "learning_rate": 6.0157928802589e-06,
      "loss": 0.133,
      "step": 270070
    },
    {
      "epoch": 104.89,
      "learning_rate": 6.015275080906149e-06,
      "loss": 0.0156,
      "step": 270080
    },
    {
      "epoch": 104.89,
      "learning_rate": 6.014757281553399e-06,
      "loss": 0.1553,
      "step": 270090
    },
    {
      "epoch": 104.89,
      "learning_rate": 6.0142394822006475e-06,
      "loss": 0.0503,
      "step": 270100
    },
    {
      "epoch": 104.9,
      "learning_rate": 6.013721682847897e-06,
      "loss": 0.0001,
      "step": 270110
    },
    {
      "epoch": 104.9,
      "learning_rate": 6.013203883495146e-06,
      "loss": 0.0001,
      "step": 270120
    },
    {
      "epoch": 104.9,
      "learning_rate": 6.012686084142395e-06,
      "loss": 0.0051,
      "step": 270130
    },
    {
      "epoch": 104.91,
      "learning_rate": 6.012168284789644e-06,
      "loss": 0.0116,
      "step": 270140
    },
    {
      "epoch": 104.91,
      "learning_rate": 6.011650485436894e-06,
      "loss": 0.0536,
      "step": 270150
    },
    {
      "epoch": 104.92,
      "learning_rate": 6.011132686084143e-06,
      "loss": 0.0404,
      "step": 270160
    },
    {
      "epoch": 104.92,
      "learning_rate": 6.0106148867313915e-06,
      "loss": 0.0428,
      "step": 270170
    },
    {
      "epoch": 104.92,
      "learning_rate": 6.010097087378641e-06,
      "loss": 0.0201,
      "step": 270180
    },
    {
      "epoch": 104.93,
      "learning_rate": 6.009579288025891e-06,
      "loss": 0.0663,
      "step": 270190
    },
    {
      "epoch": 104.93,
      "learning_rate": 6.0090614886731395e-06,
      "loss": 0.0001,
      "step": 270200
    },
    {
      "epoch": 104.94,
      "learning_rate": 6.008543689320388e-06,
      "loss": 0.0015,
      "step": 270210
    },
    {
      "epoch": 104.94,
      "learning_rate": 6.008025889967638e-06,
      "loss": 0.1012,
      "step": 270220
    },
    {
      "epoch": 104.94,
      "learning_rate": 6.0075080906148875e-06,
      "loss": 0.1549,
      "step": 270230
    },
    {
      "epoch": 104.95,
      "learning_rate": 6.006990291262137e-06,
      "loss": 0.0453,
      "step": 270240
    },
    {
      "epoch": 104.95,
      "learning_rate": 6.006472491909385e-06,
      "loss": 0.0104,
      "step": 270250
    },
    {
      "epoch": 104.96,
      "learning_rate": 6.005954692556635e-06,
      "loss": 0.0886,
      "step": 270260
    },
    {
      "epoch": 104.96,
      "learning_rate": 6.005436893203884e-06,
      "loss": 0.0421,
      "step": 270270
    },
    {
      "epoch": 104.96,
      "learning_rate": 6.004919093851134e-06,
      "loss": 0.0075,
      "step": 270280
    },
    {
      "epoch": 104.97,
      "learning_rate": 6.004401294498382e-06,
      "loss": 0.1413,
      "step": 270290
    },
    {
      "epoch": 104.97,
      "learning_rate": 6.003883495145631e-06,
      "loss": 0.0768,
      "step": 270300
    },
    {
      "epoch": 104.97,
      "learning_rate": 6.003365695792881e-06,
      "loss": 0.0026,
      "step": 270310
    },
    {
      "epoch": 104.98,
      "learning_rate": 6.002847896440131e-06,
      "loss": 0.0008,
      "step": 270320
    },
    {
      "epoch": 104.98,
      "learning_rate": 6.0023300970873786e-06,
      "loss": 0.0548,
      "step": 270330
    },
    {
      "epoch": 104.99,
      "learning_rate": 6.001812297734628e-06,
      "loss": 0.1933,
      "step": 270340
    },
    {
      "epoch": 104.99,
      "learning_rate": 6.001294498381878e-06,
      "loss": 0.0104,
      "step": 270350
    },
    {
      "epoch": 104.99,
      "learning_rate": 6.000776699029127e-06,
      "loss": 0.0119,
      "step": 270360
    },
    {
      "epoch": 105.0,
      "learning_rate": 6.000258899676375e-06,
      "loss": 0.0125,
      "step": 270370
    },
    {
      "epoch": 105.0,
      "eval_accuracy": 0.9485557083906465,
      "eval_loss": 0.3993532061576843,
      "eval_runtime": 8.1978,
      "eval_samples_per_second": 443.411,
      "eval_steps_per_second": 55.503,
      "step": 270375
    },
    {
      "epoch": 105.0,
      "learning_rate": 5.999741100323625e-06,
      "loss": 0.0015,
      "step": 270380
    },
    {
      "epoch": 105.01,
      "learning_rate": 5.999223300970875e-06,
      "loss": 0.0724,
      "step": 270390
    },
    {
      "epoch": 105.01,
      "learning_rate": 5.998705501618124e-06,
      "loss": 0.0099,
      "step": 270400
    },
    {
      "epoch": 105.01,
      "learning_rate": 5.998187702265372e-06,
      "loss": 0.0693,
      "step": 270410
    },
    {
      "epoch": 105.02,
      "learning_rate": 5.997669902912622e-06,
      "loss": 0.1983,
      "step": 270420
    },
    {
      "epoch": 105.02,
      "learning_rate": 5.997152103559871e-06,
      "loss": 0.0094,
      "step": 270430
    },
    {
      "epoch": 105.03,
      "learning_rate": 5.996634304207121e-06,
      "loss": 0.0005,
      "step": 270440
    },
    {
      "epoch": 105.03,
      "learning_rate": 5.996116504854369e-06,
      "loss": 0.0333,
      "step": 270450
    },
    {
      "epoch": 105.03,
      "learning_rate": 5.9955987055016185e-06,
      "loss": 0.0833,
      "step": 270460
    },
    {
      "epoch": 105.04,
      "learning_rate": 5.995080906148868e-06,
      "loss": 0.049,
      "step": 270470
    },
    {
      "epoch": 105.04,
      "learning_rate": 5.994563106796118e-06,
      "loss": 0.1508,
      "step": 270480
    },
    {
      "epoch": 105.04,
      "learning_rate": 5.994045307443366e-06,
      "loss": 0.0012,
      "step": 270490
    },
    {
      "epoch": 105.05,
      "learning_rate": 5.993527508090615e-06,
      "loss": 0.0438,
      "step": 270500
    },
    {
      "epoch": 105.05,
      "learning_rate": 5.993009708737865e-06,
      "loss": 0.0173,
      "step": 270510
    },
    {
      "epoch": 105.06,
      "learning_rate": 5.9924919093851145e-06,
      "loss": 0.0662,
      "step": 270520
    },
    {
      "epoch": 105.06,
      "learning_rate": 5.9919741100323625e-06,
      "loss": 0.0251,
      "step": 270530
    },
    {
      "epoch": 105.06,
      "learning_rate": 5.991456310679612e-06,
      "loss": 0.0108,
      "step": 270540
    },
    {
      "epoch": 105.07,
      "learning_rate": 5.990938511326862e-06,
      "loss": 0.2184,
      "step": 270550
    },
    {
      "epoch": 105.07,
      "learning_rate": 5.990420711974111e-06,
      "loss": 0.0107,
      "step": 270560
    },
    {
      "epoch": 105.08,
      "learning_rate": 5.989902912621359e-06,
      "loss": 0.0153,
      "step": 270570
    },
    {
      "epoch": 105.08,
      "learning_rate": 5.989385113268609e-06,
      "loss": 0.0742,
      "step": 270580
    },
    {
      "epoch": 105.08,
      "learning_rate": 5.9888673139158585e-06,
      "loss": 0.0103,
      "step": 270590
    },
    {
      "epoch": 105.09,
      "learning_rate": 5.988349514563107e-06,
      "loss": 0.0079,
      "step": 270600
    },
    {
      "epoch": 105.09,
      "learning_rate": 5.987831715210356e-06,
      "loss": 0.148,
      "step": 270610
    },
    {
      "epoch": 105.1,
      "learning_rate": 5.987313915857606e-06,
      "loss": 0.0589,
      "step": 270620
    },
    {
      "epoch": 105.1,
      "learning_rate": 5.986796116504855e-06,
      "loss": 0.0163,
      "step": 270630
    },
    {
      "epoch": 105.1,
      "learning_rate": 5.986278317152104e-06,
      "loss": 0.0448,
      "step": 270640
    },
    {
      "epoch": 105.11,
      "learning_rate": 5.985760517799353e-06,
      "loss": 0.1316,
      "step": 270650
    },
    {
      "epoch": 105.11,
      "learning_rate": 5.985242718446602e-06,
      "loss": 0.1383,
      "step": 270660
    },
    {
      "epoch": 105.11,
      "learning_rate": 5.984724919093852e-06,
      "loss": 0.0251,
      "step": 270670
    },
    {
      "epoch": 105.12,
      "learning_rate": 5.984207119741101e-06,
      "loss": 0.0178,
      "step": 270680
    },
    {
      "epoch": 105.12,
      "learning_rate": 5.9836893203883496e-06,
      "loss": 0.0083,
      "step": 270690
    },
    {
      "epoch": 105.13,
      "learning_rate": 5.983171521035599e-06,
      "loss": 0.0196,
      "step": 270700
    },
    {
      "epoch": 105.13,
      "learning_rate": 5.982653721682849e-06,
      "loss": 0.0414,
      "step": 270710
    },
    {
      "epoch": 105.13,
      "learning_rate": 5.9821359223300976e-06,
      "loss": 0.1205,
      "step": 270720
    },
    {
      "epoch": 105.14,
      "learning_rate": 5.981618122977346e-06,
      "loss": 0.0864,
      "step": 270730
    },
    {
      "epoch": 105.14,
      "learning_rate": 5.981100323624596e-06,
      "loss": 0.0264,
      "step": 270740
    },
    {
      "epoch": 105.15,
      "learning_rate": 5.980582524271845e-06,
      "loss": 0.0285,
      "step": 270750
    },
    {
      "epoch": 105.15,
      "learning_rate": 5.980064724919094e-06,
      "loss": 0.0002,
      "step": 270760
    },
    {
      "epoch": 105.15,
      "learning_rate": 5.979546925566344e-06,
      "loss": 0.0539,
      "step": 270770
    },
    {
      "epoch": 105.16,
      "learning_rate": 5.979029126213593e-06,
      "loss": 0.0475,
      "step": 270780
    },
    {
      "epoch": 105.16,
      "learning_rate": 5.9785113268608415e-06,
      "loss": 0.0084,
      "step": 270790
    },
    {
      "epoch": 105.17,
      "learning_rate": 5.977993527508091e-06,
      "loss": 0.0002,
      "step": 270800
    },
    {
      "epoch": 105.17,
      "learning_rate": 5.977475728155341e-06,
      "loss": 0.0116,
      "step": 270810
    },
    {
      "epoch": 105.17,
      "learning_rate": 5.9769579288025895e-06,
      "loss": 0.0298,
      "step": 270820
    },
    {
      "epoch": 105.18,
      "learning_rate": 5.976440129449838e-06,
      "loss": 0.0722,
      "step": 270830
    },
    {
      "epoch": 105.18,
      "learning_rate": 5.975922330097088e-06,
      "loss": 0.0825,
      "step": 270840
    },
    {
      "epoch": 105.18,
      "learning_rate": 5.9754045307443375e-06,
      "loss": 0.0294,
      "step": 270850
    },
    {
      "epoch": 105.19,
      "learning_rate": 5.974886731391586e-06,
      "loss": 0.1402,
      "step": 270860
    },
    {
      "epoch": 105.19,
      "learning_rate": 5.974368932038835e-06,
      "loss": 0.0156,
      "step": 270870
    },
    {
      "epoch": 105.2,
      "learning_rate": 5.973851132686085e-06,
      "loss": 0.0822,
      "step": 270880
    },
    {
      "epoch": 105.2,
      "learning_rate": 5.973333333333334e-06,
      "loss": 0.0227,
      "step": 270890
    },
    {
      "epoch": 105.2,
      "learning_rate": 5.972815533980583e-06,
      "loss": 0.0029,
      "step": 270900
    },
    {
      "epoch": 105.21,
      "learning_rate": 5.972297734627832e-06,
      "loss": 0.1005,
      "step": 270910
    },
    {
      "epoch": 105.21,
      "learning_rate": 5.9717799352750815e-06,
      "loss": 0.0476,
      "step": 270920
    },
    {
      "epoch": 105.22,
      "learning_rate": 5.971262135922331e-06,
      "loss": 0.0648,
      "step": 270930
    },
    {
      "epoch": 105.22,
      "learning_rate": 5.970744336569579e-06,
      "loss": 0.048,
      "step": 270940
    },
    {
      "epoch": 105.22,
      "learning_rate": 5.970226537216829e-06,
      "loss": 0.0043,
      "step": 270950
    },
    {
      "epoch": 105.23,
      "learning_rate": 5.969708737864078e-06,
      "loss": 0.0399,
      "step": 270960
    },
    {
      "epoch": 105.23,
      "learning_rate": 5.969190938511328e-06,
      "loss": 0.0639,
      "step": 270970
    },
    {
      "epoch": 105.23,
      "learning_rate": 5.968673139158576e-06,
      "loss": 0.1068,
      "step": 270980
    },
    {
      "epoch": 105.24,
      "learning_rate": 5.968155339805825e-06,
      "loss": 0.0916,
      "step": 270990
    },
    {
      "epoch": 105.24,
      "learning_rate": 5.967637540453075e-06,
      "loss": 0.0439,
      "step": 271000
    },
    {
      "epoch": 105.25,
      "learning_rate": 5.967119741100325e-06,
      "loss": 0.1391,
      "step": 271010
    },
    {
      "epoch": 105.25,
      "learning_rate": 5.9666019417475726e-06,
      "loss": 0.0145,
      "step": 271020
    },
    {
      "epoch": 105.25,
      "learning_rate": 5.966084142394822e-06,
      "loss": 0.0004,
      "step": 271030
    },
    {
      "epoch": 105.26,
      "learning_rate": 5.965566343042072e-06,
      "loss": 0.0007,
      "step": 271040
    },
    {
      "epoch": 105.26,
      "learning_rate": 5.965048543689321e-06,
      "loss": 0.0113,
      "step": 271050
    },
    {
      "epoch": 105.27,
      "learning_rate": 5.964530744336569e-06,
      "loss": 0.0044,
      "step": 271060
    },
    {
      "epoch": 105.27,
      "learning_rate": 5.964012944983819e-06,
      "loss": 0.0829,
      "step": 271070
    },
    {
      "epoch": 105.27,
      "learning_rate": 5.963495145631069e-06,
      "loss": 0.0572,
      "step": 271080
    },
    {
      "epoch": 105.28,
      "learning_rate": 5.962977346278318e-06,
      "loss": 0.1516,
      "step": 271090
    },
    {
      "epoch": 105.28,
      "learning_rate": 5.962459546925566e-06,
      "loss": 0.0617,
      "step": 271100
    },
    {
      "epoch": 105.29,
      "learning_rate": 5.961941747572816e-06,
      "loss": 0.0235,
      "step": 271110
    },
    {
      "epoch": 105.29,
      "learning_rate": 5.961423948220065e-06,
      "loss": 0.0089,
      "step": 271120
    },
    {
      "epoch": 105.29,
      "learning_rate": 5.960906148867315e-06,
      "loss": 0.0599,
      "step": 271130
    },
    {
      "epoch": 105.3,
      "learning_rate": 5.960388349514563e-06,
      "loss": 0.0769,
      "step": 271140
    },
    {
      "epoch": 105.3,
      "learning_rate": 5.9598705501618125e-06,
      "loss": 0.0143,
      "step": 271150
    },
    {
      "epoch": 105.3,
      "learning_rate": 5.959352750809062e-06,
      "loss": 0.0179,
      "step": 271160
    },
    {
      "epoch": 105.31,
      "learning_rate": 5.958834951456312e-06,
      "loss": 0.0179,
      "step": 271170
    },
    {
      "epoch": 105.31,
      "learning_rate": 5.95831715210356e-06,
      "loss": 0.099,
      "step": 271180
    },
    {
      "epoch": 105.32,
      "learning_rate": 5.957799352750809e-06,
      "loss": 0.0748,
      "step": 271190
    },
    {
      "epoch": 105.32,
      "learning_rate": 5.957281553398059e-06,
      "loss": 0.022,
      "step": 271200
    },
    {
      "epoch": 105.32,
      "learning_rate": 5.9567637540453085e-06,
      "loss": 0.0004,
      "step": 271210
    },
    {
      "epoch": 105.33,
      "learning_rate": 5.9562459546925565e-06,
      "loss": 0.0283,
      "step": 271220
    },
    {
      "epoch": 105.33,
      "learning_rate": 5.955728155339806e-06,
      "loss": 0.0993,
      "step": 271230
    },
    {
      "epoch": 105.34,
      "learning_rate": 5.955210355987056e-06,
      "loss": 0.0539,
      "step": 271240
    },
    {
      "epoch": 105.34,
      "learning_rate": 5.954692556634305e-06,
      "loss": 0.0199,
      "step": 271250
    },
    {
      "epoch": 105.34,
      "learning_rate": 5.954174757281553e-06,
      "loss": 0.126,
      "step": 271260
    },
    {
      "epoch": 105.35,
      "learning_rate": 5.953656957928803e-06,
      "loss": 0.0762,
      "step": 271270
    },
    {
      "epoch": 105.35,
      "learning_rate": 5.9531391585760525e-06,
      "loss": 0.0281,
      "step": 271280
    },
    {
      "epoch": 105.36,
      "learning_rate": 5.952621359223302e-06,
      "loss": 0.0268,
      "step": 271290
    },
    {
      "epoch": 105.36,
      "learning_rate": 5.95210355987055e-06,
      "loss": 0.0462,
      "step": 271300
    },
    {
      "epoch": 105.36,
      "learning_rate": 5.9515857605178e-06,
      "loss": 0.0201,
      "step": 271310
    },
    {
      "epoch": 105.37,
      "learning_rate": 5.951067961165049e-06,
      "loss": 0.0478,
      "step": 271320
    },
    {
      "epoch": 105.37,
      "learning_rate": 5.950550161812299e-06,
      "loss": 0.0017,
      "step": 271330
    },
    {
      "epoch": 105.37,
      "learning_rate": 5.950032362459548e-06,
      "loss": 0.0491,
      "step": 271340
    },
    {
      "epoch": 105.38,
      "learning_rate": 5.949514563106796e-06,
      "loss": 0.0007,
      "step": 271350
    },
    {
      "epoch": 105.38,
      "learning_rate": 5.948996763754046e-06,
      "loss": 0.0203,
      "step": 271360
    },
    {
      "epoch": 105.39,
      "learning_rate": 5.948478964401296e-06,
      "loss": 0.0087,
      "step": 271370
    },
    {
      "epoch": 105.39,
      "learning_rate": 5.947961165048544e-06,
      "loss": 0.0005,
      "step": 271380
    },
    {
      "epoch": 105.39,
      "learning_rate": 5.947443365695793e-06,
      "loss": 0.0149,
      "step": 271390
    },
    {
      "epoch": 105.4,
      "learning_rate": 5.946925566343043e-06,
      "loss": 0.004,
      "step": 271400
    },
    {
      "epoch": 105.4,
      "learning_rate": 5.9464077669902916e-06,
      "loss": 0.0001,
      "step": 271410
    },
    {
      "epoch": 105.41,
      "learning_rate": 5.945889967637541e-06,
      "loss": 0.0533,
      "step": 271420
    },
    {
      "epoch": 105.41,
      "learning_rate": 5.94537216828479e-06,
      "loss": 0.0327,
      "step": 271430
    },
    {
      "epoch": 105.41,
      "learning_rate": 5.94485436893204e-06,
      "loss": 0.0036,
      "step": 271440
    },
    {
      "epoch": 105.42,
      "learning_rate": 5.944336569579288e-06,
      "loss": 0.144,
      "step": 271450
    },
    {
      "epoch": 105.42,
      "learning_rate": 5.943818770226538e-06,
      "loss": 0.0671,
      "step": 271460
    },
    {
      "epoch": 105.43,
      "learning_rate": 5.943300970873787e-06,
      "loss": 0.1067,
      "step": 271470
    },
    {
      "epoch": 105.43,
      "learning_rate": 5.942783171521036e-06,
      "loss": 0.1026,
      "step": 271480
    },
    {
      "epoch": 105.43,
      "learning_rate": 5.942265372168285e-06,
      "loss": 0.0557,
      "step": 271490
    },
    {
      "epoch": 105.44,
      "learning_rate": 5.941747572815535e-06,
      "loss": 0.3014,
      "step": 271500
    },
    {
      "epoch": 105.44,
      "learning_rate": 5.9412297734627835e-06,
      "loss": 0.0746,
      "step": 271510
    },
    {
      "epoch": 105.44,
      "learning_rate": 5.940711974110033e-06,
      "loss": 0.0381,
      "step": 271520
    },
    {
      "epoch": 105.45,
      "learning_rate": 5.940194174757282e-06,
      "loss": 0.1234,
      "step": 271530
    },
    {
      "epoch": 105.45,
      "learning_rate": 5.9396763754045315e-06,
      "loss": 0.0197,
      "step": 271540
    },
    {
      "epoch": 105.46,
      "learning_rate": 5.93915857605178e-06,
      "loss": 0.0143,
      "step": 271550
    },
    {
      "epoch": 105.46,
      "learning_rate": 5.93864077669903e-06,
      "loss": 0.0858,
      "step": 271560
    },
    {
      "epoch": 105.46,
      "learning_rate": 5.938122977346279e-06,
      "loss": 0.0287,
      "step": 271570
    },
    {
      "epoch": 105.47,
      "learning_rate": 5.937605177993528e-06,
      "loss": 0.0866,
      "step": 271580
    },
    {
      "epoch": 105.47,
      "learning_rate": 5.937087378640777e-06,
      "loss": 0.0006,
      "step": 271590
    },
    {
      "epoch": 105.48,
      "learning_rate": 5.936569579288026e-06,
      "loss": 0.0727,
      "step": 271600
    },
    {
      "epoch": 105.48,
      "learning_rate": 5.9360517799352755e-06,
      "loss": 0.0374,
      "step": 271610
    },
    {
      "epoch": 105.48,
      "learning_rate": 5.935533980582525e-06,
      "loss": 0.0131,
      "step": 271620
    },
    {
      "epoch": 105.49,
      "learning_rate": 5.935016181229774e-06,
      "loss": 0.0307,
      "step": 271630
    },
    {
      "epoch": 105.49,
      "learning_rate": 5.934498381877023e-06,
      "loss": 0.0199,
      "step": 271640
    },
    {
      "epoch": 105.5,
      "learning_rate": 5.933980582524272e-06,
      "loss": 0.1011,
      "step": 271650
    },
    {
      "epoch": 105.5,
      "learning_rate": 5.933462783171522e-06,
      "loss": 0.0407,
      "step": 271660
    },
    {
      "epoch": 105.5,
      "learning_rate": 5.932944983818771e-06,
      "loss": 0.0006,
      "step": 271670
    },
    {
      "epoch": 105.51,
      "learning_rate": 5.932427184466019e-06,
      "loss": 0.0346,
      "step": 271680
    },
    {
      "epoch": 105.51,
      "learning_rate": 5.931909385113269e-06,
      "loss": 0.0684,
      "step": 271690
    },
    {
      "epoch": 105.51,
      "learning_rate": 5.931391585760519e-06,
      "loss": 0.0706,
      "step": 271700
    },
    {
      "epoch": 105.52,
      "learning_rate": 5.930873786407767e-06,
      "loss": 0.0576,
      "step": 271710
    },
    {
      "epoch": 105.52,
      "learning_rate": 5.930355987055016e-06,
      "loss": 0.1102,
      "step": 271720
    },
    {
      "epoch": 105.53,
      "learning_rate": 5.929838187702266e-06,
      "loss": 0.0047,
      "step": 271730
    },
    {
      "epoch": 105.53,
      "learning_rate": 5.929320388349515e-06,
      "loss": 0.0989,
      "step": 271740
    },
    {
      "epoch": 105.53,
      "learning_rate": 5.928802588996763e-06,
      "loss": 0.0658,
      "step": 271750
    },
    {
      "epoch": 105.54,
      "learning_rate": 5.928284789644013e-06,
      "loss": 0.0109,
      "step": 271760
    },
    {
      "epoch": 105.54,
      "learning_rate": 5.927766990291263e-06,
      "loss": 0.0346,
      "step": 271770
    },
    {
      "epoch": 105.55,
      "learning_rate": 5.927249190938512e-06,
      "loss": 0.0393,
      "step": 271780
    },
    {
      "epoch": 105.55,
      "learning_rate": 5.92673139158576e-06,
      "loss": 0.0323,
      "step": 271790
    },
    {
      "epoch": 105.55,
      "learning_rate": 5.92621359223301e-06,
      "loss": 0.0106,
      "step": 271800
    },
    {
      "epoch": 105.56,
      "learning_rate": 5.925695792880259e-06,
      "loss": 0.0472,
      "step": 271810
    },
    {
      "epoch": 105.56,
      "learning_rate": 5.925177993527509e-06,
      "loss": 0.092,
      "step": 271820
    },
    {
      "epoch": 105.57,
      "learning_rate": 5.924660194174757e-06,
      "loss": 0.0769,
      "step": 271830
    },
    {
      "epoch": 105.57,
      "learning_rate": 5.9241423948220065e-06,
      "loss": 0.1222,
      "step": 271840
    },
    {
      "epoch": 105.57,
      "learning_rate": 5.923624595469256e-06,
      "loss": 0.0804,
      "step": 271850
    },
    {
      "epoch": 105.58,
      "learning_rate": 5.923106796116506e-06,
      "loss": 0.0038,
      "step": 271860
    },
    {
      "epoch": 105.58,
      "learning_rate": 5.922588996763754e-06,
      "loss": 0.0244,
      "step": 271870
    },
    {
      "epoch": 105.58,
      "learning_rate": 5.922071197411003e-06,
      "loss": 0.0389,
      "step": 271880
    },
    {
      "epoch": 105.59,
      "learning_rate": 5.921553398058253e-06,
      "loss": 0.1394,
      "step": 271890
    },
    {
      "epoch": 105.59,
      "learning_rate": 5.9210355987055025e-06,
      "loss": 0.1384,
      "step": 271900
    },
    {
      "epoch": 105.6,
      "learning_rate": 5.920517799352752e-06,
      "loss": 0.0353,
      "step": 271910
    },
    {
      "epoch": 105.6,
      "learning_rate": 5.92e-06,
      "loss": 0.0038,
      "step": 271920
    },
    {
      "epoch": 105.6,
      "learning_rate": 5.91948220064725e-06,
      "loss": 0.2257,
      "step": 271930
    },
    {
      "epoch": 105.61,
      "learning_rate": 5.918964401294499e-06,
      "loss": 0.1107,
      "step": 271940
    },
    {
      "epoch": 105.61,
      "learning_rate": 5.918446601941749e-06,
      "loss": 0.1484,
      "step": 271950
    },
    {
      "epoch": 105.62,
      "learning_rate": 5.917928802588997e-06,
      "loss": 0.0121,
      "step": 271960
    },
    {
      "epoch": 105.62,
      "learning_rate": 5.9174110032362465e-06,
      "loss": 0.1342,
      "step": 271970
    },
    {
      "epoch": 105.62,
      "learning_rate": 5.916893203883496e-06,
      "loss": 0.0429,
      "step": 271980
    },
    {
      "epoch": 105.63,
      "learning_rate": 5.916375404530746e-06,
      "loss": 0.1303,
      "step": 271990
    },
    {
      "epoch": 105.63,
      "learning_rate": 5.915857605177994e-06,
      "loss": 0.2571,
      "step": 272000
    },
    {
      "epoch": 105.63,
      "learning_rate": 5.915339805825243e-06,
      "loss": 0.0565,
      "step": 272010
    },
    {
      "epoch": 105.64,
      "learning_rate": 5.914822006472493e-06,
      "loss": 0.0594,
      "step": 272020
    },
    {
      "epoch": 105.64,
      "learning_rate": 5.9143042071197425e-06,
      "loss": 0.2857,
      "step": 272030
    },
    {
      "epoch": 105.65,
      "learning_rate": 5.91378640776699e-06,
      "loss": 0.0024,
      "step": 272040
    },
    {
      "epoch": 105.65,
      "learning_rate": 5.91326860841424e-06,
      "loss": 0.1181,
      "step": 272050
    },
    {
      "epoch": 105.65,
      "learning_rate": 5.91275080906149e-06,
      "loss": 0.0624,
      "step": 272060
    },
    {
      "epoch": 105.66,
      "learning_rate": 5.912233009708738e-06,
      "loss": 0.0336,
      "step": 272070
    },
    {
      "epoch": 105.66,
      "learning_rate": 5.911715210355987e-06,
      "loss": 0.045,
      "step": 272080
    },
    {
      "epoch": 105.67,
      "learning_rate": 5.911197411003237e-06,
      "loss": 0.164,
      "step": 272090
    },
    {
      "epoch": 105.67,
      "learning_rate": 5.910679611650486e-06,
      "loss": 0.0536,
      "step": 272100
    },
    {
      "epoch": 105.67,
      "learning_rate": 5.910161812297735e-06,
      "loss": 0.0335,
      "step": 272110
    },
    {
      "epoch": 105.68,
      "learning_rate": 5.909644012944984e-06,
      "loss": 0.0377,
      "step": 272120
    },
    {
      "epoch": 105.68,
      "learning_rate": 5.909126213592234e-06,
      "loss": 0.1199,
      "step": 272130
    },
    {
      "epoch": 105.69,
      "learning_rate": 5.908608414239483e-06,
      "loss": 0.1187,
      "step": 272140
    },
    {
      "epoch": 105.69,
      "learning_rate": 5.908090614886732e-06,
      "loss": 0.0164,
      "step": 272150
    },
    {
      "epoch": 105.69,
      "learning_rate": 5.907572815533981e-06,
      "loss": 0.1135,
      "step": 272160
    },
    {
      "epoch": 105.7,
      "learning_rate": 5.90705501618123e-06,
      "loss": 0.0682,
      "step": 272170
    },
    {
      "epoch": 105.7,
      "learning_rate": 5.90653721682848e-06,
      "loss": 0.0919,
      "step": 272180
    },
    {
      "epoch": 105.7,
      "learning_rate": 5.906019417475729e-06,
      "loss": 0.0081,
      "step": 272190
    },
    {
      "epoch": 105.71,
      "learning_rate": 5.9055016181229775e-06,
      "loss": 0.0223,
      "step": 272200
    },
    {
      "epoch": 105.71,
      "learning_rate": 5.904983818770227e-06,
      "loss": 0.0456,
      "step": 272210
    },
    {
      "epoch": 105.72,
      "learning_rate": 5.904466019417476e-06,
      "loss": 0.1161,
      "step": 272220
    },
    {
      "epoch": 105.72,
      "learning_rate": 5.9039482200647255e-06,
      "loss": 0.0116,
      "step": 272230
    },
    {
      "epoch": 105.72,
      "learning_rate": 5.903430420711974e-06,
      "loss": 0.1161,
      "step": 272240
    },
    {
      "epoch": 105.73,
      "learning_rate": 5.902912621359224e-06,
      "loss": 0.0991,
      "step": 272250
    },
    {
      "epoch": 105.73,
      "learning_rate": 5.902394822006473e-06,
      "loss": 0.0209,
      "step": 272260
    },
    {
      "epoch": 105.74,
      "learning_rate": 5.901877022653722e-06,
      "loss": 0.0588,
      "step": 272270
    },
    {
      "epoch": 105.74,
      "learning_rate": 5.901359223300971e-06,
      "loss": 0.0203,
      "step": 272280
    },
    {
      "epoch": 105.74,
      "learning_rate": 5.900841423948221e-06,
      "loss": 0.009,
      "step": 272290
    },
    {
      "epoch": 105.75,
      "learning_rate": 5.9003236245954695e-06,
      "loss": 0.0295,
      "step": 272300
    },
    {
      "epoch": 105.75,
      "learning_rate": 5.899805825242719e-06,
      "loss": 0.1001,
      "step": 272310
    },
    {
      "epoch": 105.76,
      "learning_rate": 5.899288025889968e-06,
      "loss": 0.0644,
      "step": 272320
    },
    {
      "epoch": 105.76,
      "learning_rate": 5.8987702265372175e-06,
      "loss": 0.0004,
      "step": 272330
    },
    {
      "epoch": 105.76,
      "learning_rate": 5.898252427184466e-06,
      "loss": 0.0788,
      "step": 272340
    },
    {
      "epoch": 105.77,
      "learning_rate": 5.897734627831716e-06,
      "loss": 0.0407,
      "step": 272350
    },
    {
      "epoch": 105.77,
      "learning_rate": 5.897216828478965e-06,
      "loss": 0.0285,
      "step": 272360
    },
    {
      "epoch": 105.77,
      "learning_rate": 5.896699029126214e-06,
      "loss": 0.0291,
      "step": 272370
    },
    {
      "epoch": 105.78,
      "learning_rate": 5.896181229773463e-06,
      "loss": 0.0021,
      "step": 272380
    },
    {
      "epoch": 105.78,
      "learning_rate": 5.895663430420713e-06,
      "loss": 0.043,
      "step": 272390
    },
    {
      "epoch": 105.79,
      "learning_rate": 5.895145631067961e-06,
      "loss": 0.0227,
      "step": 272400
    },
    {
      "epoch": 105.79,
      "learning_rate": 5.89462783171521e-06,
      "loss": 0.0217,
      "step": 272410
    },
    {
      "epoch": 105.79,
      "learning_rate": 5.89411003236246e-06,
      "loss": 0.0138,
      "step": 272420
    },
    {
      "epoch": 105.8,
      "learning_rate": 5.893592233009709e-06,
      "loss": 0.0316,
      "step": 272430
    },
    {
      "epoch": 105.8,
      "learning_rate": 5.893074433656958e-06,
      "loss": 0.07,
      "step": 272440
    },
    {
      "epoch": 105.81,
      "learning_rate": 5.892556634304207e-06,
      "loss": 0.0355,
      "step": 272450
    },
    {
      "epoch": 105.81,
      "learning_rate": 5.8920388349514566e-06,
      "loss": 0.1566,
      "step": 272460
    },
    {
      "epoch": 105.81,
      "learning_rate": 5.891521035598706e-06,
      "loss": 0.0012,
      "step": 272470
    },
    {
      "epoch": 105.82,
      "learning_rate": 5.891003236245956e-06,
      "loss": 0.1116,
      "step": 272480
    },
    {
      "epoch": 105.82,
      "learning_rate": 5.890485436893204e-06,
      "loss": 0.049,
      "step": 272490
    },
    {
      "epoch": 105.83,
      "learning_rate": 5.889967637540453e-06,
      "loss": 0.1103,
      "step": 272500
    },
    {
      "epoch": 105.83,
      "learning_rate": 5.889449838187703e-06,
      "loss": 0.0399,
      "step": 272510
    },
    {
      "epoch": 105.83,
      "learning_rate": 5.888932038834953e-06,
      "loss": 0.1547,
      "step": 272520
    },
    {
      "epoch": 105.84,
      "learning_rate": 5.8884142394822005e-06,
      "loss": 0.0697,
      "step": 272530
    },
    {
      "epoch": 105.84,
      "learning_rate": 5.88789644012945e-06,
      "loss": 0.0122,
      "step": 272540
    },
    {
      "epoch": 105.84,
      "learning_rate": 5.8873786407767e-06,
      "loss": 0.0129,
      "step": 272550
    },
    {
      "epoch": 105.85,
      "learning_rate": 5.886860841423949e-06,
      "loss": 0.0442,
      "step": 272560
    },
    {
      "epoch": 105.85,
      "learning_rate": 5.886343042071197e-06,
      "loss": 0.0351,
      "step": 272570
    },
    {
      "epoch": 105.86,
      "learning_rate": 5.885825242718447e-06,
      "loss": 0.0015,
      "step": 272580
    },
    {
      "epoch": 105.86,
      "learning_rate": 5.8853074433656965e-06,
      "loss": 0.0755,
      "step": 272590
    },
    {
      "epoch": 105.86,
      "learning_rate": 5.884789644012946e-06,
      "loss": 0.0077,
      "step": 272600
    },
    {
      "epoch": 105.87,
      "learning_rate": 5.884271844660194e-06,
      "loss": 0.1655,
      "step": 272610
    },
    {
      "epoch": 105.87,
      "learning_rate": 5.883754045307444e-06,
      "loss": 0.0547,
      "step": 272620
    },
    {
      "epoch": 105.88,
      "learning_rate": 5.883236245954693e-06,
      "loss": 0.0238,
      "step": 272630
    },
    {
      "epoch": 105.88,
      "learning_rate": 5.882718446601943e-06,
      "loss": 0.0851,
      "step": 272640
    },
    {
      "epoch": 105.88,
      "learning_rate": 5.882200647249191e-06,
      "loss": 0.0032,
      "step": 272650
    },
    {
      "epoch": 105.89,
      "learning_rate": 5.8816828478964405e-06,
      "loss": 0.0742,
      "step": 272660
    },
    {
      "epoch": 105.89,
      "learning_rate": 5.88116504854369e-06,
      "loss": 0.0168,
      "step": 272670
    },
    {
      "epoch": 105.9,
      "learning_rate": 5.88064724919094e-06,
      "loss": 0.0475,
      "step": 272680
    },
    {
      "epoch": 105.9,
      "learning_rate": 5.880129449838188e-06,
      "loss": 0.0419,
      "step": 272690
    },
    {
      "epoch": 105.9,
      "learning_rate": 5.879611650485437e-06,
      "loss": 0.056,
      "step": 272700
    },
    {
      "epoch": 105.91,
      "learning_rate": 5.879093851132687e-06,
      "loss": 0.1533,
      "step": 272710
    },
    {
      "epoch": 105.91,
      "learning_rate": 5.8785760517799365e-06,
      "loss": 0.0983,
      "step": 272720
    },
    {
      "epoch": 105.91,
      "learning_rate": 5.878058252427184e-06,
      "loss": 0.0488,
      "step": 272730
    },
    {
      "epoch": 105.92,
      "learning_rate": 5.877540453074434e-06,
      "loss": 0.0621,
      "step": 272740
    },
    {
      "epoch": 105.92,
      "learning_rate": 5.877022653721684e-06,
      "loss": 0.0167,
      "step": 272750
    },
    {
      "epoch": 105.93,
      "learning_rate": 5.876504854368933e-06,
      "loss": 0.0842,
      "step": 272760
    },
    {
      "epoch": 105.93,
      "learning_rate": 5.875987055016181e-06,
      "loss": 0.0105,
      "step": 272770
    },
    {
      "epoch": 105.93,
      "learning_rate": 5.875469255663431e-06,
      "loss": 0.1693,
      "step": 272780
    },
    {
      "epoch": 105.94,
      "learning_rate": 5.87495145631068e-06,
      "loss": 0.0673,
      "step": 272790
    },
    {
      "epoch": 105.94,
      "learning_rate": 5.87443365695793e-06,
      "loss": 0.0284,
      "step": 272800
    },
    {
      "epoch": 105.95,
      "learning_rate": 5.873915857605178e-06,
      "loss": 0.0178,
      "step": 272810
    },
    {
      "epoch": 105.95,
      "learning_rate": 5.873398058252428e-06,
      "loss": 0.0757,
      "step": 272820
    },
    {
      "epoch": 105.95,
      "learning_rate": 5.872880258899677e-06,
      "loss": 0.0082,
      "step": 272830
    },
    {
      "epoch": 105.96,
      "learning_rate": 5.872362459546927e-06,
      "loss": 0.0099,
      "step": 272840
    },
    {
      "epoch": 105.96,
      "learning_rate": 5.871844660194175e-06,
      "loss": 0.071,
      "step": 272850
    },
    {
      "epoch": 105.97,
      "learning_rate": 5.871326860841424e-06,
      "loss": 0.0396,
      "step": 272860
    },
    {
      "epoch": 105.97,
      "learning_rate": 5.870809061488674e-06,
      "loss": 0.017,
      "step": 272870
    },
    {
      "epoch": 105.97,
      "learning_rate": 5.870291262135923e-06,
      "loss": 0.003,
      "step": 272880
    },
    {
      "epoch": 105.98,
      "learning_rate": 5.8697734627831715e-06,
      "loss": 0.0024,
      "step": 272890
    },
    {
      "epoch": 105.98,
      "learning_rate": 5.869255663430421e-06,
      "loss": 0.091,
      "step": 272900
    },
    {
      "epoch": 105.98,
      "learning_rate": 5.868737864077671e-06,
      "loss": 0.0246,
      "step": 272910
    },
    {
      "epoch": 105.99,
      "learning_rate": 5.8682200647249195e-06,
      "loss": 0.0993,
      "step": 272920
    },
    {
      "epoch": 105.99,
      "learning_rate": 5.867702265372168e-06,
      "loss": 0.0108,
      "step": 272930
    },
    {
      "epoch": 106.0,
      "learning_rate": 5.867184466019418e-06,
      "loss": 0.0105,
      "step": 272940
    },
    {
      "epoch": 106.0,
      "learning_rate": 5.8666666666666675e-06,
      "loss": 0.0213,
      "step": 272950
    },
    {
      "epoch": 106.0,
      "eval_accuracy": 0.9499312242090784,
      "eval_loss": 0.37747612595558167,
      "eval_runtime": 8.1704,
      "eval_samples_per_second": 444.896,
      "eval_steps_per_second": 55.689,
      "step": 272950
    },
    {
      "epoch": 106.0,
      "learning_rate": 5.866148867313916e-06,
      "loss": 0.0153,
      "step": 272960
    },
    {
      "epoch": 106.01,
      "learning_rate": 5.865631067961165e-06,
      "loss": 0.0341,
      "step": 272970
    },
    {
      "epoch": 106.01,
      "learning_rate": 5.865113268608415e-06,
      "loss": 0.1719,
      "step": 272980
    },
    {
      "epoch": 106.02,
      "learning_rate": 5.864595469255664e-06,
      "loss": 0.0515,
      "step": 272990
    },
    {
      "epoch": 106.02,
      "learning_rate": 5.864077669902913e-06,
      "loss": 0.0739,
      "step": 273000
    },
    {
      "epoch": 106.02,
      "learning_rate": 5.863559870550163e-06,
      "loss": 0.0181,
      "step": 273010
    },
    {
      "epoch": 106.03,
      "learning_rate": 5.8630420711974115e-06,
      "loss": 0.0836,
      "step": 273020
    },
    {
      "epoch": 106.03,
      "learning_rate": 5.862524271844661e-06,
      "loss": 0.1646,
      "step": 273030
    },
    {
      "epoch": 106.03,
      "learning_rate": 5.86200647249191e-06,
      "loss": 0.0002,
      "step": 273040
    },
    {
      "epoch": 106.04,
      "learning_rate": 5.8614886731391595e-06,
      "loss": 0.0922,
      "step": 273050
    },
    {
      "epoch": 106.04,
      "learning_rate": 5.860970873786408e-06,
      "loss": 0.001,
      "step": 273060
    },
    {
      "epoch": 106.05,
      "learning_rate": 5.860453074433657e-06,
      "loss": 0.0011,
      "step": 273070
    },
    {
      "epoch": 106.05,
      "learning_rate": 5.859935275080907e-06,
      "loss": 0.0795,
      "step": 273080
    },
    {
      "epoch": 106.05,
      "learning_rate": 5.859417475728156e-06,
      "loss": 0.0527,
      "step": 273090
    },
    {
      "epoch": 106.06,
      "learning_rate": 5.858899676375405e-06,
      "loss": 0.0473,
      "step": 273100
    },
    {
      "epoch": 106.06,
      "learning_rate": 5.858381877022654e-06,
      "loss": 0.0001,
      "step": 273110
    },
    {
      "epoch": 106.07,
      "learning_rate": 5.857864077669903e-06,
      "loss": 0.0065,
      "step": 273120
    },
    {
      "epoch": 106.07,
      "learning_rate": 5.857346278317153e-06,
      "loss": 0.0288,
      "step": 273130
    },
    {
      "epoch": 106.07,
      "learning_rate": 5.856828478964402e-06,
      "loss": 0.0127,
      "step": 273140
    },
    {
      "epoch": 106.08,
      "learning_rate": 5.8563106796116506e-06,
      "loss": 0.0336,
      "step": 273150
    },
    {
      "epoch": 106.08,
      "learning_rate": 5.8557928802589e-06,
      "loss": 0.0458,
      "step": 273160
    },
    {
      "epoch": 106.09,
      "learning_rate": 5.85527508090615e-06,
      "loss": 0.0151,
      "step": 273170
    },
    {
      "epoch": 106.09,
      "learning_rate": 5.854757281553399e-06,
      "loss": 0.0245,
      "step": 273180
    },
    {
      "epoch": 106.09,
      "learning_rate": 5.854239482200647e-06,
      "loss": 0.1152,
      "step": 273190
    },
    {
      "epoch": 106.1,
      "learning_rate": 5.853721682847897e-06,
      "loss": 0.1609,
      "step": 273200
    },
    {
      "epoch": 106.1,
      "learning_rate": 5.853203883495147e-06,
      "loss": 0.0096,
      "step": 273210
    },
    {
      "epoch": 106.1,
      "learning_rate": 5.8526860841423945e-06,
      "loss": 0.0283,
      "step": 273220
    },
    {
      "epoch": 106.11,
      "learning_rate": 5.852168284789644e-06,
      "loss": 0.0514,
      "step": 273230
    },
    {
      "epoch": 106.11,
      "learning_rate": 5.851650485436894e-06,
      "loss": 0.0126,
      "step": 273240
    },
    {
      "epoch": 106.12,
      "learning_rate": 5.851132686084143e-06,
      "loss": 0.0158,
      "step": 273250
    },
    {
      "epoch": 106.12,
      "learning_rate": 5.850614886731391e-06,
      "loss": 0.0014,
      "step": 273260
    },
    {
      "epoch": 106.12,
      "learning_rate": 5.850097087378641e-06,
      "loss": 0.1306,
      "step": 273270
    },
    {
      "epoch": 106.13,
      "learning_rate": 5.8495792880258905e-06,
      "loss": 0.0177,
      "step": 273280
    },
    {
      "epoch": 106.13,
      "learning_rate": 5.84906148867314e-06,
      "loss": 0.0005,
      "step": 273290
    },
    {
      "epoch": 106.14,
      "learning_rate": 5.848543689320388e-06,
      "loss": 0.0022,
      "step": 273300
    },
    {
      "epoch": 106.14,
      "learning_rate": 5.848025889967638e-06,
      "loss": 0.0113,
      "step": 273310
    },
    {
      "epoch": 106.14,
      "learning_rate": 5.847508090614887e-06,
      "loss": 0.0981,
      "step": 273320
    },
    {
      "epoch": 106.15,
      "learning_rate": 5.846990291262137e-06,
      "loss": 0.0215,
      "step": 273330
    },
    {
      "epoch": 106.15,
      "learning_rate": 5.846472491909385e-06,
      "loss": 0.0134,
      "step": 273340
    },
    {
      "epoch": 106.16,
      "learning_rate": 5.8459546925566345e-06,
      "loss": 0.0603,
      "step": 273350
    },
    {
      "epoch": 106.16,
      "learning_rate": 5.845436893203884e-06,
      "loss": 0.0726,
      "step": 273360
    },
    {
      "epoch": 106.16,
      "learning_rate": 5.844919093851134e-06,
      "loss": 0.0705,
      "step": 273370
    },
    {
      "epoch": 106.17,
      "learning_rate": 5.844401294498382e-06,
      "loss": 0.0002,
      "step": 273380
    },
    {
      "epoch": 106.17,
      "learning_rate": 5.843883495145631e-06,
      "loss": 0.0162,
      "step": 273390
    },
    {
      "epoch": 106.17,
      "learning_rate": 5.843365695792881e-06,
      "loss": 0.0177,
      "step": 273400
    },
    {
      "epoch": 106.18,
      "learning_rate": 5.8428478964401305e-06,
      "loss": 0.1124,
      "step": 273410
    },
    {
      "epoch": 106.18,
      "learning_rate": 5.842330097087378e-06,
      "loss": 0.1101,
      "step": 273420
    },
    {
      "epoch": 106.19,
      "learning_rate": 5.841812297734628e-06,
      "loss": 0.0001,
      "step": 273430
    },
    {
      "epoch": 106.19,
      "learning_rate": 5.841294498381878e-06,
      "loss": 0.0291,
      "step": 273440
    },
    {
      "epoch": 106.19,
      "learning_rate": 5.840776699029127e-06,
      "loss": 0.0722,
      "step": 273450
    },
    {
      "epoch": 106.2,
      "learning_rate": 5.840258899676375e-06,
      "loss": 0.0599,
      "step": 273460
    },
    {
      "epoch": 106.2,
      "learning_rate": 5.839741100323625e-06,
      "loss": 0.1002,
      "step": 273470
    },
    {
      "epoch": 106.21,
      "learning_rate": 5.839223300970874e-06,
      "loss": 0.0199,
      "step": 273480
    },
    {
      "epoch": 106.21,
      "learning_rate": 5.838705501618124e-06,
      "loss": 0.0055,
      "step": 273490
    },
    {
      "epoch": 106.21,
      "learning_rate": 5.838187702265372e-06,
      "loss": 0.004,
      "step": 273500
    },
    {
      "epoch": 106.22,
      "learning_rate": 5.837669902912622e-06,
      "loss": 0.0037,
      "step": 273510
    },
    {
      "epoch": 106.22,
      "learning_rate": 5.837152103559871e-06,
      "loss": 0.0128,
      "step": 273520
    },
    {
      "epoch": 106.23,
      "learning_rate": 5.836634304207121e-06,
      "loss": 0.01,
      "step": 273530
    },
    {
      "epoch": 106.23,
      "learning_rate": 5.836116504854369e-06,
      "loss": 0.0578,
      "step": 273540
    },
    {
      "epoch": 106.23,
      "learning_rate": 5.835598705501618e-06,
      "loss": 0.0425,
      "step": 273550
    },
    {
      "epoch": 106.24,
      "learning_rate": 5.835080906148868e-06,
      "loss": 0.0427,
      "step": 273560
    },
    {
      "epoch": 106.24,
      "learning_rate": 5.834563106796118e-06,
      "loss": 0.0098,
      "step": 273570
    },
    {
      "epoch": 106.24,
      "learning_rate": 5.834045307443366e-06,
      "loss": 0.049,
      "step": 273580
    },
    {
      "epoch": 106.25,
      "learning_rate": 5.833527508090615e-06,
      "loss": 0.0127,
      "step": 273590
    },
    {
      "epoch": 106.25,
      "learning_rate": 5.833009708737865e-06,
      "loss": 0.0182,
      "step": 273600
    },
    {
      "epoch": 106.26,
      "learning_rate": 5.832491909385114e-06,
      "loss": 0.0918,
      "step": 273610
    },
    {
      "epoch": 106.26,
      "learning_rate": 5.831974110032363e-06,
      "loss": 0.0012,
      "step": 273620
    },
    {
      "epoch": 106.26,
      "learning_rate": 5.831456310679612e-06,
      "loss": 0.175,
      "step": 273630
    },
    {
      "epoch": 106.27,
      "learning_rate": 5.8309385113268615e-06,
      "loss": 0.018,
      "step": 273640
    },
    {
      "epoch": 106.27,
      "learning_rate": 5.830420711974111e-06,
      "loss": 0.0216,
      "step": 273650
    },
    {
      "epoch": 106.28,
      "learning_rate": 5.82990291262136e-06,
      "loss": 0.0386,
      "step": 273660
    },
    {
      "epoch": 106.28,
      "learning_rate": 5.829385113268609e-06,
      "loss": 0.0437,
      "step": 273670
    },
    {
      "epoch": 106.28,
      "learning_rate": 5.828867313915858e-06,
      "loss": 0.0634,
      "step": 273680
    },
    {
      "epoch": 106.29,
      "learning_rate": 5.828349514563107e-06,
      "loss": 0.0257,
      "step": 273690
    },
    {
      "epoch": 106.29,
      "learning_rate": 5.827831715210357e-06,
      "loss": 0.0752,
      "step": 273700
    },
    {
      "epoch": 106.3,
      "learning_rate": 5.8273139158576055e-06,
      "loss": 0.0298,
      "step": 273710
    },
    {
      "epoch": 106.3,
      "learning_rate": 5.826796116504855e-06,
      "loss": 0.1016,
      "step": 273720
    },
    {
      "epoch": 106.3,
      "learning_rate": 5.826278317152104e-06,
      "loss": 0.0717,
      "step": 273730
    },
    {
      "epoch": 106.31,
      "learning_rate": 5.8257605177993535e-06,
      "loss": 0.0172,
      "step": 273740
    },
    {
      "epoch": 106.31,
      "learning_rate": 5.825242718446602e-06,
      "loss": 0.0165,
      "step": 273750
    },
    {
      "epoch": 106.31,
      "learning_rate": 5.824724919093852e-06,
      "loss": 0.1044,
      "step": 273760
    },
    {
      "epoch": 106.32,
      "learning_rate": 5.824207119741101e-06,
      "loss": 0.0013,
      "step": 273770
    },
    {
      "epoch": 106.32,
      "learning_rate": 5.82368932038835e-06,
      "loss": 0.0213,
      "step": 273780
    },
    {
      "epoch": 106.33,
      "learning_rate": 5.823171521035599e-06,
      "loss": 0.2549,
      "step": 273790
    },
    {
      "epoch": 106.33,
      "learning_rate": 5.822653721682849e-06,
      "loss": 0.0915,
      "step": 273800
    },
    {
      "epoch": 106.33,
      "learning_rate": 5.822135922330097e-06,
      "loss": 0.1249,
      "step": 273810
    },
    {
      "epoch": 106.34,
      "learning_rate": 5.821618122977347e-06,
      "loss": 0.0666,
      "step": 273820
    },
    {
      "epoch": 106.34,
      "learning_rate": 5.821100323624596e-06,
      "loss": 0.1315,
      "step": 273830
    },
    {
      "epoch": 106.35,
      "learning_rate": 5.820582524271845e-06,
      "loss": 0.0265,
      "step": 273840
    },
    {
      "epoch": 106.35,
      "learning_rate": 5.820064724919094e-06,
      "loss": 0.0526,
      "step": 273850
    },
    {
      "epoch": 106.35,
      "learning_rate": 5.819546925566344e-06,
      "loss": 0.089,
      "step": 273860
    },
    {
      "epoch": 106.36,
      "learning_rate": 5.819029126213593e-06,
      "loss": 0.0126,
      "step": 273870
    },
    {
      "epoch": 106.36,
      "learning_rate": 5.818511326860841e-06,
      "loss": 0.0098,
      "step": 273880
    },
    {
      "epoch": 106.37,
      "learning_rate": 5.817993527508091e-06,
      "loss": 0.0152,
      "step": 273890
    },
    {
      "epoch": 106.37,
      "learning_rate": 5.817475728155341e-06,
      "loss": 0.0042,
      "step": 273900
    },
    {
      "epoch": 106.37,
      "learning_rate": 5.816957928802589e-06,
      "loss": 0.035,
      "step": 273910
    },
    {
      "epoch": 106.38,
      "learning_rate": 5.816440129449838e-06,
      "loss": 0.0126,
      "step": 273920
    },
    {
      "epoch": 106.38,
      "learning_rate": 5.815922330097088e-06,
      "loss": 0.0232,
      "step": 273930
    },
    {
      "epoch": 106.38,
      "learning_rate": 5.815404530744337e-06,
      "loss": 0.0433,
      "step": 273940
    },
    {
      "epoch": 106.39,
      "learning_rate": 5.814886731391586e-06,
      "loss": 0.0616,
      "step": 273950
    },
    {
      "epoch": 106.39,
      "learning_rate": 5.814368932038835e-06,
      "loss": 0.0114,
      "step": 273960
    },
    {
      "epoch": 106.4,
      "learning_rate": 5.8138511326860845e-06,
      "loss": 0.0235,
      "step": 273970
    },
    {
      "epoch": 106.4,
      "learning_rate": 5.813333333333334e-06,
      "loss": 0.0257,
      "step": 273980
    },
    {
      "epoch": 106.4,
      "learning_rate": 5.812815533980583e-06,
      "loss": 0.1049,
      "step": 273990
    },
    {
      "epoch": 106.41,
      "learning_rate": 5.812297734627832e-06,
      "loss": 0.1609,
      "step": 274000
    },
    {
      "epoch": 106.41,
      "learning_rate": 5.811779935275081e-06,
      "loss": 0.0842,
      "step": 274010
    },
    {
      "epoch": 106.42,
      "learning_rate": 5.811262135922331e-06,
      "loss": 0.0116,
      "step": 274020
    },
    {
      "epoch": 106.42,
      "learning_rate": 5.810744336569579e-06,
      "loss": 0.0153,
      "step": 274030
    },
    {
      "epoch": 106.42,
      "learning_rate": 5.8102265372168285e-06,
      "loss": 0.0009,
      "step": 274040
    },
    {
      "epoch": 106.43,
      "learning_rate": 5.809708737864078e-06,
      "loss": 0.1575,
      "step": 274050
    },
    {
      "epoch": 106.43,
      "learning_rate": 5.809190938511328e-06,
      "loss": 0.0002,
      "step": 274060
    },
    {
      "epoch": 106.43,
      "learning_rate": 5.808673139158576e-06,
      "loss": 0.0104,
      "step": 274070
    },
    {
      "epoch": 106.44,
      "learning_rate": 5.808155339805825e-06,
      "loss": 0.0043,
      "step": 274080
    },
    {
      "epoch": 106.44,
      "learning_rate": 5.807637540453075e-06,
      "loss": 0.0311,
      "step": 274090
    },
    {
      "epoch": 106.45,
      "learning_rate": 5.8071197411003245e-06,
      "loss": 0.0811,
      "step": 274100
    },
    {
      "epoch": 106.45,
      "learning_rate": 5.806601941747572e-06,
      "loss": 0.0299,
      "step": 274110
    },
    {
      "epoch": 106.45,
      "learning_rate": 5.806084142394822e-06,
      "loss": 0.0002,
      "step": 274120
    },
    {
      "epoch": 106.46,
      "learning_rate": 5.805566343042072e-06,
      "loss": 0.0931,
      "step": 274130
    },
    {
      "epoch": 106.46,
      "learning_rate": 5.805048543689321e-06,
      "loss": 0.1359,
      "step": 274140
    },
    {
      "epoch": 106.47,
      "learning_rate": 5.804530744336571e-06,
      "loss": 0.1135,
      "step": 274150
    },
    {
      "epoch": 106.47,
      "learning_rate": 5.804012944983819e-06,
      "loss": 0.0621,
      "step": 274160
    },
    {
      "epoch": 106.47,
      "learning_rate": 5.803495145631068e-06,
      "loss": 0.1754,
      "step": 274170
    },
    {
      "epoch": 106.48,
      "learning_rate": 5.802977346278318e-06,
      "loss": 0.0443,
      "step": 274180
    },
    {
      "epoch": 106.48,
      "learning_rate": 5.802459546925568e-06,
      "loss": 0.0501,
      "step": 274190
    },
    {
      "epoch": 106.49,
      "learning_rate": 5.8019417475728156e-06,
      "loss": 0.0476,
      "step": 274200
    },
    {
      "epoch": 106.49,
      "learning_rate": 5.801423948220065e-06,
      "loss": 0.08,
      "step": 274210
    },
    {
      "epoch": 106.49,
      "learning_rate": 5.800906148867315e-06,
      "loss": 0.0191,
      "step": 274220
    },
    {
      "epoch": 106.5,
      "learning_rate": 5.8003883495145644e-06,
      "loss": 0.0776,
      "step": 274230
    },
    {
      "epoch": 106.5,
      "learning_rate": 5.799870550161812e-06,
      "loss": 0.1239,
      "step": 274240
    },
    {
      "epoch": 106.5,
      "learning_rate": 5.799352750809062e-06,
      "loss": 0.0601,
      "step": 274250
    },
    {
      "epoch": 106.51,
      "learning_rate": 5.798834951456312e-06,
      "loss": 0.0453,
      "step": 274260
    },
    {
      "epoch": 106.51,
      "learning_rate": 5.798317152103561e-06,
      "loss": 0.0714,
      "step": 274270
    },
    {
      "epoch": 106.52,
      "learning_rate": 5.797799352750809e-06,
      "loss": 0.0304,
      "step": 274280
    },
    {
      "epoch": 106.52,
      "learning_rate": 5.797281553398059e-06,
      "loss": 0.0192,
      "step": 274290
    },
    {
      "epoch": 106.52,
      "learning_rate": 5.796763754045308e-06,
      "loss": 0.0346,
      "step": 274300
    },
    {
      "epoch": 106.53,
      "learning_rate": 5.796245954692558e-06,
      "loss": 0.0143,
      "step": 274310
    },
    {
      "epoch": 106.53,
      "learning_rate": 5.795728155339806e-06,
      "loss": 0.1402,
      "step": 274320
    },
    {
      "epoch": 106.54,
      "learning_rate": 5.7952103559870555e-06,
      "loss": 0.0663,
      "step": 274330
    },
    {
      "epoch": 106.54,
      "learning_rate": 5.794692556634305e-06,
      "loss": 0.002,
      "step": 274340
    },
    {
      "epoch": 106.54,
      "learning_rate": 5.794174757281554e-06,
      "loss": 0.0006,
      "step": 274350
    },
    {
      "epoch": 106.55,
      "learning_rate": 5.793656957928803e-06,
      "loss": 0.0276,
      "step": 274360
    },
    {
      "epoch": 106.55,
      "learning_rate": 5.793139158576052e-06,
      "loss": 0.2616,
      "step": 274370
    },
    {
      "epoch": 106.56,
      "learning_rate": 5.792621359223302e-06,
      "loss": 0.0003,
      "step": 274380
    },
    {
      "epoch": 106.56,
      "learning_rate": 5.792103559870551e-06,
      "loss": 0.037,
      "step": 274390
    },
    {
      "epoch": 106.56,
      "learning_rate": 5.7915857605177995e-06,
      "loss": 0.0577,
      "step": 274400
    },
    {
      "epoch": 106.57,
      "learning_rate": 5.791067961165049e-06,
      "loss": 0.0614,
      "step": 274410
    },
    {
      "epoch": 106.57,
      "learning_rate": 5.790550161812299e-06,
      "loss": 0.0901,
      "step": 274420
    },
    {
      "epoch": 106.57,
      "learning_rate": 5.7900323624595475e-06,
      "loss": 0.0958,
      "step": 274430
    },
    {
      "epoch": 106.58,
      "learning_rate": 5.789514563106796e-06,
      "loss": 0.0501,
      "step": 274440
    },
    {
      "epoch": 106.58,
      "learning_rate": 5.788996763754046e-06,
      "loss": 0.0245,
      "step": 274450
    },
    {
      "epoch": 106.59,
      "learning_rate": 5.7884789644012955e-06,
      "loss": 0.0554,
      "step": 274460
    },
    {
      "epoch": 106.59,
      "learning_rate": 5.787961165048544e-06,
      "loss": 0.0921,
      "step": 274470
    },
    {
      "epoch": 106.59,
      "learning_rate": 5.787443365695793e-06,
      "loss": 0.0093,
      "step": 274480
    },
    {
      "epoch": 106.6,
      "learning_rate": 5.786925566343043e-06,
      "loss": 0.0196,
      "step": 274490
    },
    {
      "epoch": 106.6,
      "learning_rate": 5.786407766990291e-06,
      "loss": 0.0176,
      "step": 274500
    },
    {
      "epoch": 106.61,
      "learning_rate": 5.785889967637541e-06,
      "loss": 0.0619,
      "step": 274510
    },
    {
      "epoch": 106.61,
      "learning_rate": 5.78537216828479e-06,
      "loss": 0.0185,
      "step": 274520
    },
    {
      "epoch": 106.61,
      "learning_rate": 5.784854368932039e-06,
      "loss": 0.1206,
      "step": 274530
    },
    {
      "epoch": 106.62,
      "learning_rate": 5.784336569579288e-06,
      "loss": 0.1073,
      "step": 274540
    },
    {
      "epoch": 106.62,
      "learning_rate": 5.783818770226538e-06,
      "loss": 0.0949,
      "step": 274550
    },
    {
      "epoch": 106.63,
      "learning_rate": 5.783300970873787e-06,
      "loss": 0.0346,
      "step": 274560
    },
    {
      "epoch": 106.63,
      "learning_rate": 5.782783171521036e-06,
      "loss": 0.0045,
      "step": 274570
    },
    {
      "epoch": 106.63,
      "learning_rate": 5.782265372168285e-06,
      "loss": 0.0834,
      "step": 274580
    },
    {
      "epoch": 106.64,
      "learning_rate": 5.781747572815535e-06,
      "loss": 0.0589,
      "step": 274590
    },
    {
      "epoch": 106.64,
      "learning_rate": 5.781229773462783e-06,
      "loss": 0.0461,
      "step": 274600
    },
    {
      "epoch": 106.64,
      "learning_rate": 5.780711974110033e-06,
      "loss": 0.1046,
      "step": 274610
    },
    {
      "epoch": 106.65,
      "learning_rate": 5.780194174757282e-06,
      "loss": 0.1322,
      "step": 274620
    },
    {
      "epoch": 106.65,
      "learning_rate": 5.779676375404531e-06,
      "loss": 0.0296,
      "step": 274630
    },
    {
      "epoch": 106.66,
      "learning_rate": 5.77915857605178e-06,
      "loss": 0.0808,
      "step": 274640
    },
    {
      "epoch": 106.66,
      "learning_rate": 5.77864077669903e-06,
      "loss": 0.0054,
      "step": 274650
    },
    {
      "epoch": 106.66,
      "learning_rate": 5.7781229773462785e-06,
      "loss": 0.0472,
      "step": 274660
    },
    {
      "epoch": 106.67,
      "learning_rate": 5.777605177993528e-06,
      "loss": 0.169,
      "step": 274670
    },
    {
      "epoch": 106.67,
      "learning_rate": 5.777087378640777e-06,
      "loss": 0.1849,
      "step": 274680
    },
    {
      "epoch": 106.68,
      "learning_rate": 5.776569579288026e-06,
      "loss": 0.0914,
      "step": 274690
    },
    {
      "epoch": 106.68,
      "learning_rate": 5.776051779935275e-06,
      "loss": 0.0005,
      "step": 274700
    },
    {
      "epoch": 106.68,
      "learning_rate": 5.775533980582525e-06,
      "loss": 0.0453,
      "step": 274710
    },
    {
      "epoch": 106.69,
      "learning_rate": 5.7750161812297745e-06,
      "loss": 0.0968,
      "step": 274720
    },
    {
      "epoch": 106.69,
      "learning_rate": 5.7744983818770225e-06,
      "loss": 0.093,
      "step": 274730
    },
    {
      "epoch": 106.7,
      "learning_rate": 5.773980582524272e-06,
      "loss": 0.0758,
      "step": 274740
    },
    {
      "epoch": 106.7,
      "learning_rate": 5.773462783171522e-06,
      "loss": 0.0574,
      "step": 274750
    },
    {
      "epoch": 106.7,
      "learning_rate": 5.772944983818771e-06,
      "loss": 0.012,
      "step": 274760
    },
    {
      "epoch": 106.71,
      "learning_rate": 5.772427184466019e-06,
      "loss": 0.1484,
      "step": 274770
    },
    {
      "epoch": 106.71,
      "learning_rate": 5.771909385113269e-06,
      "loss": 0.0156,
      "step": 274780
    },
    {
      "epoch": 106.71,
      "learning_rate": 5.7713915857605185e-06,
      "loss": 0.0517,
      "step": 274790
    },
    {
      "epoch": 106.72,
      "learning_rate": 5.770873786407768e-06,
      "loss": 0.1327,
      "step": 274800
    },
    {
      "epoch": 106.72,
      "learning_rate": 5.770355987055016e-06,
      "loss": 0.1137,
      "step": 274810
    },
    {
      "epoch": 106.73,
      "learning_rate": 5.769838187702266e-06,
      "loss": 0.0518,
      "step": 274820
    },
    {
      "epoch": 106.73,
      "learning_rate": 5.769320388349515e-06,
      "loss": 0.0005,
      "step": 274830
    },
    {
      "epoch": 106.73,
      "learning_rate": 5.768802588996765e-06,
      "loss": 0.0745,
      "step": 274840
    },
    {
      "epoch": 106.74,
      "learning_rate": 5.768284789644013e-06,
      "loss": 0.0551,
      "step": 274850
    },
    {
      "epoch": 106.74,
      "learning_rate": 5.767766990291262e-06,
      "loss": 0.1052,
      "step": 274860
    },
    {
      "epoch": 106.75,
      "learning_rate": 5.767249190938512e-06,
      "loss": 0.0002,
      "step": 274870
    },
    {
      "epoch": 106.75,
      "learning_rate": 5.766731391585762e-06,
      "loss": 0.0146,
      "step": 274880
    },
    {
      "epoch": 106.75,
      "learning_rate": 5.7662135922330096e-06,
      "loss": 0.0156,
      "step": 274890
    },
    {
      "epoch": 106.76,
      "learning_rate": 5.765695792880259e-06,
      "loss": 0.0584,
      "step": 274900
    },
    {
      "epoch": 106.76,
      "learning_rate": 5.765177993527509e-06,
      "loss": 0.0962,
      "step": 274910
    },
    {
      "epoch": 106.77,
      "learning_rate": 5.7646601941747584e-06,
      "loss": 0.0647,
      "step": 274920
    },
    {
      "epoch": 106.77,
      "learning_rate": 5.764142394822006e-06,
      "loss": 0.0009,
      "step": 274930
    },
    {
      "epoch": 106.77,
      "learning_rate": 5.763624595469256e-06,
      "loss": 0.0763,
      "step": 274940
    },
    {
      "epoch": 106.78,
      "learning_rate": 5.763106796116506e-06,
      "loss": 0.0846,
      "step": 274950
    },
    {
      "epoch": 106.78,
      "learning_rate": 5.762588996763755e-06,
      "loss": 0.0258,
      "step": 274960
    },
    {
      "epoch": 106.78,
      "learning_rate": 5.762071197411003e-06,
      "loss": 0.0368,
      "step": 274970
    },
    {
      "epoch": 106.79,
      "learning_rate": 5.761553398058253e-06,
      "loss": 0.0175,
      "step": 274980
    },
    {
      "epoch": 106.79,
      "learning_rate": 5.761035598705502e-06,
      "loss": 0.0575,
      "step": 274990
    },
    {
      "epoch": 106.8,
      "learning_rate": 5.760517799352752e-06,
      "loss": 0.0379,
      "step": 275000
    },
    {
      "epoch": 106.8,
      "learning_rate": 5.76e-06,
      "loss": 0.0676,
      "step": 275010
    },
    {
      "epoch": 106.8,
      "learning_rate": 5.7594822006472495e-06,
      "loss": 0.0122,
      "step": 275020
    },
    {
      "epoch": 106.81,
      "learning_rate": 5.758964401294499e-06,
      "loss": 0.2,
      "step": 275030
    },
    {
      "epoch": 106.81,
      "learning_rate": 5.758446601941749e-06,
      "loss": 0.0607,
      "step": 275040
    },
    {
      "epoch": 106.82,
      "learning_rate": 5.757928802588997e-06,
      "loss": 0.0292,
      "step": 275050
    },
    {
      "epoch": 106.82,
      "learning_rate": 5.757411003236246e-06,
      "loss": 0.0334,
      "step": 275060
    },
    {
      "epoch": 106.82,
      "learning_rate": 5.756893203883496e-06,
      "loss": 0.0587,
      "step": 275070
    },
    {
      "epoch": 106.83,
      "learning_rate": 5.7563754045307455e-06,
      "loss": 0.0284,
      "step": 275080
    },
    {
      "epoch": 106.83,
      "learning_rate": 5.7558576051779935e-06,
      "loss": 0.0683,
      "step": 275090
    },
    {
      "epoch": 106.83,
      "learning_rate": 5.755339805825243e-06,
      "loss": 0.0305,
      "step": 275100
    },
    {
      "epoch": 106.84,
      "learning_rate": 5.754822006472493e-06,
      "loss": 0.0124,
      "step": 275110
    },
    {
      "epoch": 106.84,
      "learning_rate": 5.754304207119742e-06,
      "loss": 0.0339,
      "step": 275120
    },
    {
      "epoch": 106.85,
      "learning_rate": 5.75378640776699e-06,
      "loss": 0.0621,
      "step": 275130
    },
    {
      "epoch": 106.85,
      "learning_rate": 5.75326860841424e-06,
      "loss": 0.0059,
      "step": 275140
    },
    {
      "epoch": 106.85,
      "learning_rate": 5.7527508090614895e-06,
      "loss": 0.1002,
      "step": 275150
    },
    {
      "epoch": 106.86,
      "learning_rate": 5.752233009708738e-06,
      "loss": 0.0953,
      "step": 275160
    },
    {
      "epoch": 106.86,
      "learning_rate": 5.751715210355987e-06,
      "loss": 0.0945,
      "step": 275170
    },
    {
      "epoch": 106.87,
      "learning_rate": 5.751197411003237e-06,
      "loss": 0.0131,
      "step": 275180
    },
    {
      "epoch": 106.87,
      "learning_rate": 5.750679611650486e-06,
      "loss": 0.0578,
      "step": 275190
    },
    {
      "epoch": 106.87,
      "learning_rate": 5.750161812297735e-06,
      "loss": 0.0697,
      "step": 275200
    },
    {
      "epoch": 106.88,
      "learning_rate": 5.749644012944984e-06,
      "loss": 0.0245,
      "step": 275210
    },
    {
      "epoch": 106.88,
      "learning_rate": 5.749126213592233e-06,
      "loss": 0.0428,
      "step": 275220
    },
    {
      "epoch": 106.89,
      "learning_rate": 5.748608414239483e-06,
      "loss": 0.0644,
      "step": 275230
    },
    {
      "epoch": 106.89,
      "learning_rate": 5.748090614886732e-06,
      "loss": 0.0291,
      "step": 275240
    },
    {
      "epoch": 106.89,
      "learning_rate": 5.747572815533981e-06,
      "loss": 0.0001,
      "step": 275250
    },
    {
      "epoch": 106.9,
      "learning_rate": 5.74705501618123e-06,
      "loss": 0.0196,
      "step": 275260
    },
    {
      "epoch": 106.9,
      "learning_rate": 5.74653721682848e-06,
      "loss": 0.0559,
      "step": 275270
    },
    {
      "epoch": 106.9,
      "learning_rate": 5.746019417475729e-06,
      "loss": 0.0273,
      "step": 275280
    },
    {
      "epoch": 106.91,
      "learning_rate": 5.745501618122978e-06,
      "loss": 0.039,
      "step": 275290
    },
    {
      "epoch": 106.91,
      "learning_rate": 5.744983818770227e-06,
      "loss": 0.0355,
      "step": 275300
    },
    {
      "epoch": 106.92,
      "learning_rate": 5.744466019417477e-06,
      "loss": 0.074,
      "step": 275310
    },
    {
      "epoch": 106.92,
      "learning_rate": 5.743948220064725e-06,
      "loss": 0.1195,
      "step": 275320
    },
    {
      "epoch": 106.92,
      "learning_rate": 5.743430420711975e-06,
      "loss": 0.1055,
      "step": 275330
    },
    {
      "epoch": 106.93,
      "learning_rate": 5.742912621359224e-06,
      "loss": 0.0517,
      "step": 275340
    },
    {
      "epoch": 106.93,
      "learning_rate": 5.7423948220064725e-06,
      "loss": 0.0093,
      "step": 275350
    },
    {
      "epoch": 106.94,
      "learning_rate": 5.741877022653722e-06,
      "loss": 0.0061,
      "step": 275360
    },
    {
      "epoch": 106.94,
      "learning_rate": 5.741359223300972e-06,
      "loss": 0.067,
      "step": 275370
    },
    {
      "epoch": 106.94,
      "learning_rate": 5.7408414239482205e-06,
      "loss": 0.0308,
      "step": 275380
    },
    {
      "epoch": 106.95,
      "learning_rate": 5.740323624595469e-06,
      "loss": 0.0009,
      "step": 275390
    },
    {
      "epoch": 106.95,
      "learning_rate": 5.739805825242719e-06,
      "loss": 0.0651,
      "step": 275400
    },
    {
      "epoch": 106.96,
      "learning_rate": 5.7392880258899685e-06,
      "loss": 0.0015,
      "step": 275410
    },
    {
      "epoch": 106.96,
      "learning_rate": 5.738770226537217e-06,
      "loss": 0.0014,
      "step": 275420
    },
    {
      "epoch": 106.96,
      "learning_rate": 5.738252427184466e-06,
      "loss": 0.0114,
      "step": 275430
    },
    {
      "epoch": 106.97,
      "learning_rate": 5.737734627831716e-06,
      "loss": 0.0235,
      "step": 275440
    },
    {
      "epoch": 106.97,
      "learning_rate": 5.737216828478965e-06,
      "loss": 0.0888,
      "step": 275450
    },
    {
      "epoch": 106.97,
      "learning_rate": 5.736699029126214e-06,
      "loss": 0.1199,
      "step": 275460
    },
    {
      "epoch": 106.98,
      "learning_rate": 5.736181229773463e-06,
      "loss": 0.0066,
      "step": 275470
    },
    {
      "epoch": 106.98,
      "learning_rate": 5.7356634304207125e-06,
      "loss": 0.0316,
      "step": 275480
    },
    {
      "epoch": 106.99,
      "learning_rate": 5.735145631067962e-06,
      "loss": 0.0001,
      "step": 275490
    },
    {
      "epoch": 106.99,
      "learning_rate": 5.73462783171521e-06,
      "loss": 0.0423,
      "step": 275500
    },
    {
      "epoch": 106.99,
      "learning_rate": 5.73411003236246e-06,
      "loss": 0.0198,
      "step": 275510
    },
    {
      "epoch": 107.0,
      "learning_rate": 5.733592233009709e-06,
      "loss": 0.1249,
      "step": 275520
    },
    {
      "epoch": 107.0,
      "eval_accuracy": 0.9491059147180193,
      "eval_loss": 0.3901929259300232,
      "eval_runtime": 8.2173,
      "eval_samples_per_second": 442.361,
      "eval_steps_per_second": 55.371,
      "step": 275525
    },
    {
      "epoch": 107.0,
      "learning_rate": 5.733074433656959e-06,
      "loss": 0.1031,
      "step": 275530
    },
    {
      "epoch": 107.01,
      "learning_rate": 5.732556634304207e-06,
      "loss": 0.0002,
      "step": 275540
    },
    {
      "epoch": 107.01,
      "learning_rate": 5.732038834951456e-06,
      "loss": 0.1034,
      "step": 275550
    },
    {
      "epoch": 107.01,
      "learning_rate": 5.731521035598706e-06,
      "loss": 0.0122,
      "step": 275560
    },
    {
      "epoch": 107.02,
      "learning_rate": 5.731003236245956e-06,
      "loss": 0.0983,
      "step": 275570
    },
    {
      "epoch": 107.02,
      "learning_rate": 5.7304854368932036e-06,
      "loss": 0.0972,
      "step": 275580
    },
    {
      "epoch": 107.03,
      "learning_rate": 5.729967637540453e-06,
      "loss": 0.06,
      "step": 275590
    },
    {
      "epoch": 107.03,
      "learning_rate": 5.729449838187703e-06,
      "loss": 0.0017,
      "step": 275600
    },
    {
      "epoch": 107.03,
      "learning_rate": 5.7289320388349524e-06,
      "loss": 0.0626,
      "step": 275610
    },
    {
      "epoch": 107.04,
      "learning_rate": 5.7284142394822e-06,
      "loss": 0.0025,
      "step": 275620
    },
    {
      "epoch": 107.04,
      "learning_rate": 5.72789644012945e-06,
      "loss": 0.0115,
      "step": 275630
    },
    {
      "epoch": 107.04,
      "learning_rate": 5.7273786407767e-06,
      "loss": 0.0127,
      "step": 275640
    },
    {
      "epoch": 107.05,
      "learning_rate": 5.726860841423949e-06,
      "loss": 0.0001,
      "step": 275650
    },
    {
      "epoch": 107.05,
      "learning_rate": 5.726343042071197e-06,
      "loss": 0.0775,
      "step": 275660
    },
    {
      "epoch": 107.06,
      "learning_rate": 5.725825242718447e-06,
      "loss": 0.0543,
      "step": 275670
    },
    {
      "epoch": 107.06,
      "learning_rate": 5.725307443365696e-06,
      "loss": 0.0988,
      "step": 275680
    },
    {
      "epoch": 107.06,
      "learning_rate": 5.724789644012946e-06,
      "loss": 0.0936,
      "step": 275690
    },
    {
      "epoch": 107.07,
      "learning_rate": 5.724271844660194e-06,
      "loss": 0.0099,
      "step": 275700
    },
    {
      "epoch": 107.07,
      "learning_rate": 5.7237540453074435e-06,
      "loss": 0.0219,
      "step": 275710
    },
    {
      "epoch": 107.08,
      "learning_rate": 5.723236245954693e-06,
      "loss": 0.0326,
      "step": 275720
    },
    {
      "epoch": 107.08,
      "learning_rate": 5.722718446601943e-06,
      "loss": 0.0602,
      "step": 275730
    },
    {
      "epoch": 107.08,
      "learning_rate": 5.722200647249191e-06,
      "loss": 0.0398,
      "step": 275740
    },
    {
      "epoch": 107.09,
      "learning_rate": 5.72168284789644e-06,
      "loss": 0.0368,
      "step": 275750
    },
    {
      "epoch": 107.09,
      "learning_rate": 5.72116504854369e-06,
      "loss": 0.1156,
      "step": 275760
    },
    {
      "epoch": 107.1,
      "learning_rate": 5.7206472491909395e-06,
      "loss": 0.044,
      "step": 275770
    },
    {
      "epoch": 107.1,
      "learning_rate": 5.7201294498381875e-06,
      "loss": 0.0361,
      "step": 275780
    },
    {
      "epoch": 107.1,
      "learning_rate": 5.719611650485437e-06,
      "loss": 0.0524,
      "step": 275790
    },
    {
      "epoch": 107.11,
      "learning_rate": 5.719093851132687e-06,
      "loss": 0.0783,
      "step": 275800
    },
    {
      "epoch": 107.11,
      "learning_rate": 5.718576051779936e-06,
      "loss": 0.0582,
      "step": 275810
    },
    {
      "epoch": 107.11,
      "learning_rate": 5.718058252427185e-06,
      "loss": 0.0853,
      "step": 275820
    },
    {
      "epoch": 107.12,
      "learning_rate": 5.717540453074434e-06,
      "loss": 0.0739,
      "step": 275830
    },
    {
      "epoch": 107.12,
      "learning_rate": 5.7170226537216835e-06,
      "loss": 0.1153,
      "step": 275840
    },
    {
      "epoch": 107.13,
      "learning_rate": 5.716504854368933e-06,
      "loss": 0.0516,
      "step": 275850
    },
    {
      "epoch": 107.13,
      "learning_rate": 5.715987055016182e-06,
      "loss": 0.0087,
      "step": 275860
    },
    {
      "epoch": 107.13,
      "learning_rate": 5.715469255663431e-06,
      "loss": 0.0213,
      "step": 275870
    },
    {
      "epoch": 107.14,
      "learning_rate": 5.71495145631068e-06,
      "loss": 0.0374,
      "step": 275880
    },
    {
      "epoch": 107.14,
      "learning_rate": 5.71443365695793e-06,
      "loss": 0.0565,
      "step": 275890
    },
    {
      "epoch": 107.15,
      "learning_rate": 5.713915857605179e-06,
      "loss": 0.0995,
      "step": 275900
    },
    {
      "epoch": 107.15,
      "learning_rate": 5.713398058252427e-06,
      "loss": 0.0172,
      "step": 275910
    },
    {
      "epoch": 107.15,
      "learning_rate": 5.712880258899677e-06,
      "loss": 0.0748,
      "step": 275920
    },
    {
      "epoch": 107.16,
      "learning_rate": 5.712362459546927e-06,
      "loss": 0.0593,
      "step": 275930
    },
    {
      "epoch": 107.16,
      "learning_rate": 5.711844660194175e-06,
      "loss": 0.1275,
      "step": 275940
    },
    {
      "epoch": 107.17,
      "learning_rate": 5.711326860841424e-06,
      "loss": 0.077,
      "step": 275950
    },
    {
      "epoch": 107.17,
      "learning_rate": 5.710809061488674e-06,
      "loss": 0.0153,
      "step": 275960
    },
    {
      "epoch": 107.17,
      "learning_rate": 5.710291262135923e-06,
      "loss": 0.0007,
      "step": 275970
    },
    {
      "epoch": 107.18,
      "learning_rate": 5.709773462783172e-06,
      "loss": 0.047,
      "step": 275980
    },
    {
      "epoch": 107.18,
      "learning_rate": 5.709255663430421e-06,
      "loss": 0.1218,
      "step": 275990
    },
    {
      "epoch": 107.18,
      "learning_rate": 5.708737864077671e-06,
      "loss": 0.0706,
      "step": 276000
    },
    {
      "epoch": 107.19,
      "learning_rate": 5.708220064724919e-06,
      "loss": 0.0164,
      "step": 276010
    },
    {
      "epoch": 107.19,
      "learning_rate": 5.707702265372169e-06,
      "loss": 0.1985,
      "step": 276020
    },
    {
      "epoch": 107.2,
      "learning_rate": 5.707184466019418e-06,
      "loss": 0.0759,
      "step": 276030
    },
    {
      "epoch": 107.2,
      "learning_rate": 5.706666666666667e-06,
      "loss": 0.0004,
      "step": 276040
    },
    {
      "epoch": 107.2,
      "learning_rate": 5.706148867313916e-06,
      "loss": 0.0625,
      "step": 276050
    },
    {
      "epoch": 107.21,
      "learning_rate": 5.705631067961166e-06,
      "loss": 0.0443,
      "step": 276060
    },
    {
      "epoch": 107.21,
      "learning_rate": 5.7051132686084145e-06,
      "loss": 0.0085,
      "step": 276070
    },
    {
      "epoch": 107.22,
      "learning_rate": 5.704595469255664e-06,
      "loss": 0.0103,
      "step": 276080
    },
    {
      "epoch": 107.22,
      "learning_rate": 5.704077669902913e-06,
      "loss": 0.0721,
      "step": 276090
    },
    {
      "epoch": 107.22,
      "learning_rate": 5.7035598705501625e-06,
      "loss": 0.0015,
      "step": 276100
    },
    {
      "epoch": 107.23,
      "learning_rate": 5.703042071197411e-06,
      "loss": 0.2234,
      "step": 276110
    },
    {
      "epoch": 107.23,
      "learning_rate": 5.702524271844661e-06,
      "loss": 0.0001,
      "step": 276120
    },
    {
      "epoch": 107.23,
      "learning_rate": 5.70200647249191e-06,
      "loss": 0.0608,
      "step": 276130
    },
    {
      "epoch": 107.24,
      "learning_rate": 5.701488673139159e-06,
      "loss": 0.0651,
      "step": 276140
    },
    {
      "epoch": 107.24,
      "learning_rate": 5.700970873786408e-06,
      "loss": 0.1088,
      "step": 276150
    },
    {
      "epoch": 107.25,
      "learning_rate": 5.700453074433657e-06,
      "loss": 0.219,
      "step": 276160
    },
    {
      "epoch": 107.25,
      "learning_rate": 5.6999352750809065e-06,
      "loss": 0.0067,
      "step": 276170
    },
    {
      "epoch": 107.25,
      "learning_rate": 5.699417475728156e-06,
      "loss": 0.029,
      "step": 276180
    },
    {
      "epoch": 107.26,
      "learning_rate": 5.698899676375405e-06,
      "loss": 0.1035,
      "step": 276190
    },
    {
      "epoch": 107.26,
      "learning_rate": 5.698381877022654e-06,
      "loss": 0.0877,
      "step": 276200
    },
    {
      "epoch": 107.27,
      "learning_rate": 5.697864077669903e-06,
      "loss": 0.0328,
      "step": 276210
    },
    {
      "epoch": 107.27,
      "learning_rate": 5.697346278317153e-06,
      "loss": 0.0964,
      "step": 276220
    },
    {
      "epoch": 107.27,
      "learning_rate": 5.696828478964402e-06,
      "loss": 0.0526,
      "step": 276230
    },
    {
      "epoch": 107.28,
      "learning_rate": 5.69631067961165e-06,
      "loss": 0.0621,
      "step": 276240
    },
    {
      "epoch": 107.28,
      "learning_rate": 5.6957928802589e-06,
      "loss": 0.1284,
      "step": 276250
    },
    {
      "epoch": 107.29,
      "learning_rate": 5.69527508090615e-06,
      "loss": 0.0765,
      "step": 276260
    },
    {
      "epoch": 107.29,
      "learning_rate": 5.694757281553398e-06,
      "loss": 0.0132,
      "step": 276270
    },
    {
      "epoch": 107.29,
      "learning_rate": 5.694239482200647e-06,
      "loss": 0.0729,
      "step": 276280
    },
    {
      "epoch": 107.3,
      "learning_rate": 5.693721682847897e-06,
      "loss": 0.0313,
      "step": 276290
    },
    {
      "epoch": 107.3,
      "learning_rate": 5.6932038834951464e-06,
      "loss": 0.0072,
      "step": 276300
    },
    {
      "epoch": 107.3,
      "learning_rate": 5.692686084142394e-06,
      "loss": 0.0125,
      "step": 276310
    },
    {
      "epoch": 107.31,
      "learning_rate": 5.692168284789644e-06,
      "loss": 0.1654,
      "step": 276320
    },
    {
      "epoch": 107.31,
      "learning_rate": 5.691650485436894e-06,
      "loss": 0.1318,
      "step": 276330
    },
    {
      "epoch": 107.32,
      "learning_rate": 5.691132686084143e-06,
      "loss": 0.1228,
      "step": 276340
    },
    {
      "epoch": 107.32,
      "learning_rate": 5.690614886731391e-06,
      "loss": 0.0951,
      "step": 276350
    },
    {
      "epoch": 107.32,
      "learning_rate": 5.690097087378641e-06,
      "loss": 0.0165,
      "step": 276360
    },
    {
      "epoch": 107.33,
      "learning_rate": 5.68957928802589e-06,
      "loss": 0.0059,
      "step": 276370
    },
    {
      "epoch": 107.33,
      "learning_rate": 5.68906148867314e-06,
      "loss": 0.0078,
      "step": 276380
    },
    {
      "epoch": 107.34,
      "learning_rate": 5.68854368932039e-06,
      "loss": 0.0306,
      "step": 276390
    },
    {
      "epoch": 107.34,
      "learning_rate": 5.6880258899676375e-06,
      "loss": 0.0032,
      "step": 276400
    },
    {
      "epoch": 107.34,
      "learning_rate": 5.687508090614887e-06,
      "loss": 0.0805,
      "step": 276410
    },
    {
      "epoch": 107.35,
      "learning_rate": 5.686990291262137e-06,
      "loss": 0.0727,
      "step": 276420
    },
    {
      "epoch": 107.35,
      "learning_rate": 5.686472491909386e-06,
      "loss": 0.0218,
      "step": 276430
    },
    {
      "epoch": 107.36,
      "learning_rate": 5.685954692556634e-06,
      "loss": 0.0795,
      "step": 276440
    },
    {
      "epoch": 107.36,
      "learning_rate": 5.685436893203884e-06,
      "loss": 0.0482,
      "step": 276450
    },
    {
      "epoch": 107.36,
      "learning_rate": 5.6849190938511335e-06,
      "loss": 0.0965,
      "step": 276460
    },
    {
      "epoch": 107.37,
      "learning_rate": 5.684401294498383e-06,
      "loss": 0.0416,
      "step": 276470
    },
    {
      "epoch": 107.37,
      "learning_rate": 5.683883495145631e-06,
      "loss": 0.1095,
      "step": 276480
    },
    {
      "epoch": 107.37,
      "learning_rate": 5.683365695792881e-06,
      "loss": 0.0424,
      "step": 276490
    },
    {
      "epoch": 107.38,
      "learning_rate": 5.68284789644013e-06,
      "loss": 0.0561,
      "step": 276500
    },
    {
      "epoch": 107.38,
      "learning_rate": 5.68233009708738e-06,
      "loss": 0.0197,
      "step": 276510
    },
    {
      "epoch": 107.39,
      "learning_rate": 5.681812297734628e-06,
      "loss": 0.0268,
      "step": 276520
    },
    {
      "epoch": 107.39,
      "learning_rate": 5.6812944983818775e-06,
      "loss": 0.0159,
      "step": 276530
    },
    {
      "epoch": 107.39,
      "learning_rate": 5.680776699029127e-06,
      "loss": 0.0104,
      "step": 276540
    },
    {
      "epoch": 107.4,
      "learning_rate": 5.680258899676377e-06,
      "loss": 0.086,
      "step": 276550
    },
    {
      "epoch": 107.4,
      "learning_rate": 5.679741100323625e-06,
      "loss": 0.0143,
      "step": 276560
    },
    {
      "epoch": 107.41,
      "learning_rate": 5.679223300970874e-06,
      "loss": 0.0318,
      "step": 276570
    },
    {
      "epoch": 107.41,
      "learning_rate": 5.678705501618124e-06,
      "loss": 0.0745,
      "step": 276580
    },
    {
      "epoch": 107.41,
      "learning_rate": 5.6781877022653735e-06,
      "loss": 0.1146,
      "step": 276590
    },
    {
      "epoch": 107.42,
      "learning_rate": 5.677669902912621e-06,
      "loss": 0.0099,
      "step": 276600
    },
    {
      "epoch": 107.42,
      "learning_rate": 5.677152103559871e-06,
      "loss": 0.0993,
      "step": 276610
    },
    {
      "epoch": 107.43,
      "learning_rate": 5.676634304207121e-06,
      "loss": 0.1012,
      "step": 276620
    },
    {
      "epoch": 107.43,
      "learning_rate": 5.676116504854369e-06,
      "loss": 0.1315,
      "step": 276630
    },
    {
      "epoch": 107.43,
      "learning_rate": 5.675598705501618e-06,
      "loss": 0.0008,
      "step": 276640
    },
    {
      "epoch": 107.44,
      "learning_rate": 5.675080906148868e-06,
      "loss": 0.1136,
      "step": 276650
    },
    {
      "epoch": 107.44,
      "learning_rate": 5.6745631067961174e-06,
      "loss": 0.114,
      "step": 276660
    },
    {
      "epoch": 107.44,
      "learning_rate": 5.674045307443366e-06,
      "loss": 0.041,
      "step": 276670
    },
    {
      "epoch": 107.45,
      "learning_rate": 5.673527508090615e-06,
      "loss": 0.1168,
      "step": 276680
    },
    {
      "epoch": 107.45,
      "learning_rate": 5.673009708737865e-06,
      "loss": 0.098,
      "step": 276690
    },
    {
      "epoch": 107.46,
      "learning_rate": 5.672491909385114e-06,
      "loss": 0.1129,
      "step": 276700
    },
    {
      "epoch": 107.46,
      "learning_rate": 5.671974110032363e-06,
      "loss": 0.0003,
      "step": 276710
    },
    {
      "epoch": 107.46,
      "learning_rate": 5.671456310679612e-06,
      "loss": 0.0316,
      "step": 276720
    },
    {
      "epoch": 107.47,
      "learning_rate": 5.670938511326861e-06,
      "loss": 0.0149,
      "step": 276730
    },
    {
      "epoch": 107.47,
      "learning_rate": 5.670420711974111e-06,
      "loss": 0.0139,
      "step": 276740
    },
    {
      "epoch": 107.48,
      "learning_rate": 5.66990291262136e-06,
      "loss": 0.0054,
      "step": 276750
    },
    {
      "epoch": 107.48,
      "learning_rate": 5.6693851132686085e-06,
      "loss": 0.0363,
      "step": 276760
    },
    {
      "epoch": 107.48,
      "learning_rate": 5.668867313915858e-06,
      "loss": 0.0847,
      "step": 276770
    },
    {
      "epoch": 107.49,
      "learning_rate": 5.668349514563107e-06,
      "loss": 0.059,
      "step": 276780
    },
    {
      "epoch": 107.49,
      "learning_rate": 5.6678317152103565e-06,
      "loss": 0.0449,
      "step": 276790
    },
    {
      "epoch": 107.5,
      "learning_rate": 5.667313915857605e-06,
      "loss": 0.0992,
      "step": 276800
    },
    {
      "epoch": 107.5,
      "learning_rate": 5.666796116504855e-06,
      "loss": 0.1517,
      "step": 276810
    },
    {
      "epoch": 107.5,
      "learning_rate": 5.666278317152104e-06,
      "loss": 0.0948,
      "step": 276820
    },
    {
      "epoch": 107.51,
      "learning_rate": 5.665760517799353e-06,
      "loss": 0.0986,
      "step": 276830
    },
    {
      "epoch": 107.51,
      "learning_rate": 5.665242718446602e-06,
      "loss": 0.0343,
      "step": 276840
    },
    {
      "epoch": 107.51,
      "learning_rate": 5.664724919093852e-06,
      "loss": 0.0274,
      "step": 276850
    },
    {
      "epoch": 107.52,
      "learning_rate": 5.6642071197411005e-06,
      "loss": 0.0304,
      "step": 276860
    },
    {
      "epoch": 107.52,
      "learning_rate": 5.66368932038835e-06,
      "loss": 0.0955,
      "step": 276870
    },
    {
      "epoch": 107.53,
      "learning_rate": 5.663171521035599e-06,
      "loss": 0.0263,
      "step": 276880
    },
    {
      "epoch": 107.53,
      "learning_rate": 5.6626537216828485e-06,
      "loss": 0.0271,
      "step": 276890
    },
    {
      "epoch": 107.53,
      "learning_rate": 5.662135922330097e-06,
      "loss": 0.048,
      "step": 276900
    },
    {
      "epoch": 107.54,
      "learning_rate": 5.661618122977347e-06,
      "loss": 0.0344,
      "step": 276910
    },
    {
      "epoch": 107.54,
      "learning_rate": 5.661100323624596e-06,
      "loss": 0.0157,
      "step": 276920
    },
    {
      "epoch": 107.55,
      "learning_rate": 5.660582524271845e-06,
      "loss": 0.0006,
      "step": 276930
    },
    {
      "epoch": 107.55,
      "learning_rate": 5.660064724919094e-06,
      "loss": 0.1053,
      "step": 276940
    },
    {
      "epoch": 107.55,
      "learning_rate": 5.659546925566344e-06,
      "loss": 0.071,
      "step": 276950
    },
    {
      "epoch": 107.56,
      "learning_rate": 5.659029126213593e-06,
      "loss": 0.0201,
      "step": 276960
    },
    {
      "epoch": 107.56,
      "learning_rate": 5.658511326860841e-06,
      "loss": 0.0412,
      "step": 276970
    },
    {
      "epoch": 107.57,
      "learning_rate": 5.657993527508091e-06,
      "loss": 0.0136,
      "step": 276980
    },
    {
      "epoch": 107.57,
      "learning_rate": 5.6574757281553404e-06,
      "loss": 0.0251,
      "step": 276990
    },
    {
      "epoch": 107.57,
      "learning_rate": 5.65695792880259e-06,
      "loss": 0.0829,
      "step": 277000
    },
    {
      "epoch": 107.58,
      "learning_rate": 5.656440129449838e-06,
      "loss": 0.0422,
      "step": 277010
    },
    {
      "epoch": 107.58,
      "learning_rate": 5.655922330097088e-06,
      "loss": 0.0376,
      "step": 277020
    },
    {
      "epoch": 107.58,
      "learning_rate": 5.655404530744337e-06,
      "loss": 0.0142,
      "step": 277030
    },
    {
      "epoch": 107.59,
      "learning_rate": 5.654886731391587e-06,
      "loss": 0.0481,
      "step": 277040
    },
    {
      "epoch": 107.59,
      "learning_rate": 5.654368932038835e-06,
      "loss": 0.0032,
      "step": 277050
    },
    {
      "epoch": 107.6,
      "learning_rate": 5.653851132686084e-06,
      "loss": 0.06,
      "step": 277060
    },
    {
      "epoch": 107.6,
      "learning_rate": 5.653333333333334e-06,
      "loss": 0.0754,
      "step": 277070
    },
    {
      "epoch": 107.6,
      "learning_rate": 5.652815533980584e-06,
      "loss": 0.0003,
      "step": 277080
    },
    {
      "epoch": 107.61,
      "learning_rate": 5.6522977346278315e-06,
      "loss": 0.0099,
      "step": 277090
    },
    {
      "epoch": 107.61,
      "learning_rate": 5.651779935275081e-06,
      "loss": 0.0528,
      "step": 277100
    },
    {
      "epoch": 107.62,
      "learning_rate": 5.651262135922331e-06,
      "loss": 0.0074,
      "step": 277110
    },
    {
      "epoch": 107.62,
      "learning_rate": 5.65074433656958e-06,
      "loss": 0.0393,
      "step": 277120
    },
    {
      "epoch": 107.62,
      "learning_rate": 5.650226537216828e-06,
      "loss": 0.1259,
      "step": 277130
    },
    {
      "epoch": 107.63,
      "learning_rate": 5.649708737864078e-06,
      "loss": 0.0811,
      "step": 277140
    },
    {
      "epoch": 107.63,
      "learning_rate": 5.6491909385113275e-06,
      "loss": 0.201,
      "step": 277150
    },
    {
      "epoch": 107.63,
      "learning_rate": 5.648673139158577e-06,
      "loss": 0.1425,
      "step": 277160
    },
    {
      "epoch": 107.64,
      "learning_rate": 5.648155339805825e-06,
      "loss": 0.0097,
      "step": 277170
    },
    {
      "epoch": 107.64,
      "learning_rate": 5.647637540453075e-06,
      "loss": 0.0155,
      "step": 277180
    },
    {
      "epoch": 107.65,
      "learning_rate": 5.647119741100324e-06,
      "loss": 0.0714,
      "step": 277190
    },
    {
      "epoch": 107.65,
      "learning_rate": 5.646601941747574e-06,
      "loss": 0.0341,
      "step": 277200
    },
    {
      "epoch": 107.65,
      "learning_rate": 5.646084142394822e-06,
      "loss": 0.0086,
      "step": 277210
    },
    {
      "epoch": 107.66,
      "learning_rate": 5.6455663430420715e-06,
      "loss": 0.0325,
      "step": 277220
    },
    {
      "epoch": 107.66,
      "learning_rate": 5.645048543689321e-06,
      "loss": 0.1432,
      "step": 277230
    },
    {
      "epoch": 107.67,
      "learning_rate": 5.644530744336571e-06,
      "loss": 0.041,
      "step": 277240
    },
    {
      "epoch": 107.67,
      "learning_rate": 5.644012944983819e-06,
      "loss": 0.1665,
      "step": 277250
    },
    {
      "epoch": 107.67,
      "learning_rate": 5.643495145631068e-06,
      "loss": 0.0439,
      "step": 277260
    },
    {
      "epoch": 107.68,
      "learning_rate": 5.642977346278318e-06,
      "loss": 0.0683,
      "step": 277270
    },
    {
      "epoch": 107.68,
      "learning_rate": 5.6424595469255675e-06,
      "loss": 0.0277,
      "step": 277280
    },
    {
      "epoch": 107.69,
      "learning_rate": 5.641941747572815e-06,
      "loss": 0.0292,
      "step": 277290
    },
    {
      "epoch": 107.69,
      "learning_rate": 5.641423948220065e-06,
      "loss": 0.0226,
      "step": 277300
    },
    {
      "epoch": 107.69,
      "learning_rate": 5.640906148867315e-06,
      "loss": 0.0232,
      "step": 277310
    },
    {
      "epoch": 107.7,
      "learning_rate": 5.640388349514564e-06,
      "loss": 0.0602,
      "step": 277320
    },
    {
      "epoch": 107.7,
      "learning_rate": 5.639870550161812e-06,
      "loss": 0.0462,
      "step": 277330
    },
    {
      "epoch": 107.7,
      "learning_rate": 5.639352750809062e-06,
      "loss": 0.0719,
      "step": 277340
    },
    {
      "epoch": 107.71,
      "learning_rate": 5.6388349514563114e-06,
      "loss": 0.1397,
      "step": 277350
    },
    {
      "epoch": 107.71,
      "learning_rate": 5.638317152103561e-06,
      "loss": 0.0833,
      "step": 277360
    },
    {
      "epoch": 107.72,
      "learning_rate": 5.637799352750809e-06,
      "loss": 0.0243,
      "step": 277370
    },
    {
      "epoch": 107.72,
      "learning_rate": 5.637281553398059e-06,
      "loss": 0.1004,
      "step": 277380
    },
    {
      "epoch": 107.72,
      "learning_rate": 5.636763754045308e-06,
      "loss": 0.0097,
      "step": 277390
    },
    {
      "epoch": 107.73,
      "learning_rate": 5.636245954692558e-06,
      "loss": 0.033,
      "step": 277400
    },
    {
      "epoch": 107.73,
      "learning_rate": 5.635728155339806e-06,
      "loss": 0.0384,
      "step": 277410
    },
    {
      "epoch": 107.74,
      "learning_rate": 5.635210355987055e-06,
      "loss": 0.0001,
      "step": 277420
    },
    {
      "epoch": 107.74,
      "learning_rate": 5.634692556634305e-06,
      "loss": 0.0797,
      "step": 277430
    },
    {
      "epoch": 107.74,
      "learning_rate": 5.634174757281554e-06,
      "loss": 0.0223,
      "step": 277440
    },
    {
      "epoch": 107.75,
      "learning_rate": 5.6336569579288025e-06,
      "loss": 0.0571,
      "step": 277450
    },
    {
      "epoch": 107.75,
      "learning_rate": 5.633139158576052e-06,
      "loss": 0.0223,
      "step": 277460
    },
    {
      "epoch": 107.76,
      "learning_rate": 5.632621359223302e-06,
      "loss": 0.0219,
      "step": 277470
    },
    {
      "epoch": 107.76,
      "learning_rate": 5.6321035598705505e-06,
      "loss": 0.0005,
      "step": 277480
    },
    {
      "epoch": 107.76,
      "learning_rate": 5.631585760517799e-06,
      "loss": 0.0884,
      "step": 277490
    },
    {
      "epoch": 107.77,
      "learning_rate": 5.631067961165049e-06,
      "loss": 0.0001,
      "step": 277500
    },
    {
      "epoch": 107.77,
      "learning_rate": 5.6305501618122985e-06,
      "loss": 0.0372,
      "step": 277510
    },
    {
      "epoch": 107.77,
      "learning_rate": 5.630032362459547e-06,
      "loss": 0.0667,
      "step": 277520
    },
    {
      "epoch": 107.78,
      "learning_rate": 5.629514563106797e-06,
      "loss": 0.0417,
      "step": 277530
    },
    {
      "epoch": 107.78,
      "learning_rate": 5.628996763754046e-06,
      "loss": 0.0141,
      "step": 277540
    },
    {
      "epoch": 107.79,
      "learning_rate": 5.628478964401295e-06,
      "loss": 0.1067,
      "step": 277550
    },
    {
      "epoch": 107.79,
      "learning_rate": 5.627961165048544e-06,
      "loss": 0.0138,
      "step": 277560
    },
    {
      "epoch": 107.79,
      "learning_rate": 5.627443365695794e-06,
      "loss": 0.0059,
      "step": 277570
    },
    {
      "epoch": 107.8,
      "learning_rate": 5.6269255663430425e-06,
      "loss": 0.0781,
      "step": 277580
    },
    {
      "epoch": 107.8,
      "learning_rate": 5.626407766990292e-06,
      "loss": 0.0417,
      "step": 277590
    },
    {
      "epoch": 107.81,
      "learning_rate": 5.625889967637541e-06,
      "loss": 0.0474,
      "step": 277600
    },
    {
      "epoch": 107.81,
      "learning_rate": 5.6253721682847905e-06,
      "loss": 0.1032,
      "step": 277610
    },
    {
      "epoch": 107.81,
      "learning_rate": 5.624854368932039e-06,
      "loss": 0.0088,
      "step": 277620
    },
    {
      "epoch": 107.82,
      "learning_rate": 5.624336569579288e-06,
      "loss": 0.0123,
      "step": 277630
    },
    {
      "epoch": 107.82,
      "learning_rate": 5.623818770226538e-06,
      "loss": 0.034,
      "step": 277640
    },
    {
      "epoch": 107.83,
      "learning_rate": 5.623300970873787e-06,
      "loss": 0.0032,
      "step": 277650
    },
    {
      "epoch": 107.83,
      "learning_rate": 5.622783171521036e-06,
      "loss": 0.0338,
      "step": 277660
    },
    {
      "epoch": 107.83,
      "learning_rate": 5.622265372168285e-06,
      "loss": 0.1775,
      "step": 277670
    },
    {
      "epoch": 107.84,
      "learning_rate": 5.6217475728155344e-06,
      "loss": 0.0509,
      "step": 277680
    },
    {
      "epoch": 107.84,
      "learning_rate": 5.621229773462784e-06,
      "loss": 0.0002,
      "step": 277690
    },
    {
      "epoch": 107.84,
      "learning_rate": 5.620711974110033e-06,
      "loss": 0.1802,
      "step": 277700
    },
    {
      "epoch": 107.85,
      "learning_rate": 5.620194174757282e-06,
      "loss": 0.0453,
      "step": 277710
    },
    {
      "epoch": 107.85,
      "learning_rate": 5.619676375404531e-06,
      "loss": 0.0685,
      "step": 277720
    },
    {
      "epoch": 107.86,
      "learning_rate": 5.619158576051781e-06,
      "loss": 0.0058,
      "step": 277730
    },
    {
      "epoch": 107.86,
      "learning_rate": 5.61864077669903e-06,
      "loss": 0.068,
      "step": 277740
    },
    {
      "epoch": 107.86,
      "learning_rate": 5.618122977346278e-06,
      "loss": 0.0527,
      "step": 277750
    },
    {
      "epoch": 107.87,
      "learning_rate": 5.617605177993528e-06,
      "loss": 0.001,
      "step": 277760
    },
    {
      "epoch": 107.87,
      "learning_rate": 5.617087378640778e-06,
      "loss": 0.0245,
      "step": 277770
    },
    {
      "epoch": 107.88,
      "learning_rate": 5.6165695792880255e-06,
      "loss": 0.0836,
      "step": 277780
    },
    {
      "epoch": 107.88,
      "learning_rate": 5.616051779935275e-06,
      "loss": 0.047,
      "step": 277790
    },
    {
      "epoch": 107.88,
      "learning_rate": 5.615533980582525e-06,
      "loss": 0.0708,
      "step": 277800
    },
    {
      "epoch": 107.89,
      "learning_rate": 5.615016181229774e-06,
      "loss": 0.1361,
      "step": 277810
    },
    {
      "epoch": 107.89,
      "learning_rate": 5.614498381877022e-06,
      "loss": 0.0756,
      "step": 277820
    },
    {
      "epoch": 107.9,
      "learning_rate": 5.613980582524272e-06,
      "loss": 0.0389,
      "step": 277830
    },
    {
      "epoch": 107.9,
      "learning_rate": 5.6134627831715215e-06,
      "loss": 0.0359,
      "step": 277840
    },
    {
      "epoch": 107.9,
      "learning_rate": 5.612944983818771e-06,
      "loss": 0.0006,
      "step": 277850
    },
    {
      "epoch": 107.91,
      "learning_rate": 5.612427184466019e-06,
      "loss": 0.0663,
      "step": 277860
    },
    {
      "epoch": 107.91,
      "learning_rate": 5.611909385113269e-06,
      "loss": 0.0005,
      "step": 277870
    },
    {
      "epoch": 107.91,
      "learning_rate": 5.611391585760518e-06,
      "loss": 0.0024,
      "step": 277880
    },
    {
      "epoch": 107.92,
      "learning_rate": 5.610873786407768e-06,
      "loss": 0.0147,
      "step": 277890
    },
    {
      "epoch": 107.92,
      "learning_rate": 5.610355987055016e-06,
      "loss": 0.0155,
      "step": 277900
    },
    {
      "epoch": 107.93,
      "learning_rate": 5.6098381877022655e-06,
      "loss": 0.0794,
      "step": 277910
    },
    {
      "epoch": 107.93,
      "learning_rate": 5.609320388349515e-06,
      "loss": 0.1,
      "step": 277920
    },
    {
      "epoch": 107.93,
      "learning_rate": 5.608802588996765e-06,
      "loss": 0.0197,
      "step": 277930
    },
    {
      "epoch": 107.94,
      "learning_rate": 5.608284789644013e-06,
      "loss": 0.0745,
      "step": 277940
    },
    {
      "epoch": 107.94,
      "learning_rate": 5.607766990291262e-06,
      "loss": 0.0975,
      "step": 277950
    },
    {
      "epoch": 107.95,
      "learning_rate": 5.607249190938512e-06,
      "loss": 0.1889,
      "step": 277960
    },
    {
      "epoch": 107.95,
      "learning_rate": 5.6067313915857615e-06,
      "loss": 0.0482,
      "step": 277970
    },
    {
      "epoch": 107.95,
      "learning_rate": 5.606213592233009e-06,
      "loss": 0.0367,
      "step": 277980
    },
    {
      "epoch": 107.96,
      "learning_rate": 5.605695792880259e-06,
      "loss": 0.0176,
      "step": 277990
    },
    {
      "epoch": 107.96,
      "learning_rate": 5.605177993527509e-06,
      "loss": 0.0406,
      "step": 278000
    },
    {
      "epoch": 107.97,
      "learning_rate": 5.604660194174758e-06,
      "loss": 0.1117,
      "step": 278010
    },
    {
      "epoch": 107.97,
      "learning_rate": 5.604142394822006e-06,
      "loss": 0.014,
      "step": 278020
    },
    {
      "epoch": 107.97,
      "learning_rate": 5.603624595469256e-06,
      "loss": 0.2085,
      "step": 278030
    },
    {
      "epoch": 107.98,
      "learning_rate": 5.6031067961165054e-06,
      "loss": 0.0758,
      "step": 278040
    },
    {
      "epoch": 107.98,
      "learning_rate": 5.602588996763755e-06,
      "loss": 0.085,
      "step": 278050
    },
    {
      "epoch": 107.98,
      "learning_rate": 5.602071197411005e-06,
      "loss": 0.0017,
      "step": 278060
    },
    {
      "epoch": 107.99,
      "learning_rate": 5.601553398058253e-06,
      "loss": 0.1027,
      "step": 278070
    },
    {
      "epoch": 107.99,
      "learning_rate": 5.601035598705502e-06,
      "loss": 0.0336,
      "step": 278080
    },
    {
      "epoch": 108.0,
      "learning_rate": 5.600517799352752e-06,
      "loss": 0.0552,
      "step": 278090
    },
    {
      "epoch": 108.0,
      "learning_rate": 5.600000000000001e-06,
      "loss": 0.0333,
      "step": 278100
    },
    {
      "epoch": 108.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.3637031316757202,
      "eval_runtime": 8.2062,
      "eval_samples_per_second": 442.96,
      "eval_steps_per_second": 55.446,
      "step": 278100
    },
    {
      "epoch": 108.0,
      "learning_rate": 5.599482200647249e-06,
      "loss": 0.0526,
      "step": 278110
    },
    {
      "epoch": 108.01,
      "learning_rate": 5.598964401294499e-06,
      "loss": 0.0057,
      "step": 278120
    },
    {
      "epoch": 108.01,
      "learning_rate": 5.598446601941749e-06,
      "loss": 0.0653,
      "step": 278130
    },
    {
      "epoch": 108.02,
      "learning_rate": 5.597928802588997e-06,
      "loss": 0.1102,
      "step": 278140
    },
    {
      "epoch": 108.02,
      "learning_rate": 5.597411003236246e-06,
      "loss": 0.1764,
      "step": 278150
    },
    {
      "epoch": 108.02,
      "learning_rate": 5.596893203883496e-06,
      "loss": 0.0548,
      "step": 278160
    },
    {
      "epoch": 108.03,
      "learning_rate": 5.596375404530745e-06,
      "loss": 0.0283,
      "step": 278170
    },
    {
      "epoch": 108.03,
      "learning_rate": 5.595857605177994e-06,
      "loss": 0.0037,
      "step": 278180
    },
    {
      "epoch": 108.03,
      "learning_rate": 5.595339805825243e-06,
      "loss": 0.0899,
      "step": 278190
    },
    {
      "epoch": 108.04,
      "learning_rate": 5.5948220064724925e-06,
      "loss": 0.0559,
      "step": 278200
    },
    {
      "epoch": 108.04,
      "learning_rate": 5.594304207119742e-06,
      "loss": 0.0007,
      "step": 278210
    },
    {
      "epoch": 108.05,
      "learning_rate": 5.593786407766991e-06,
      "loss": 0.0999,
      "step": 278220
    },
    {
      "epoch": 108.05,
      "learning_rate": 5.59326860841424e-06,
      "loss": 0.0477,
      "step": 278230
    },
    {
      "epoch": 108.05,
      "learning_rate": 5.592750809061489e-06,
      "loss": 0.0012,
      "step": 278240
    },
    {
      "epoch": 108.06,
      "learning_rate": 5.592233009708738e-06,
      "loss": 0.0673,
      "step": 278250
    },
    {
      "epoch": 108.06,
      "learning_rate": 5.591715210355988e-06,
      "loss": 0.001,
      "step": 278260
    },
    {
      "epoch": 108.07,
      "learning_rate": 5.5911974110032365e-06,
      "loss": 0.0303,
      "step": 278270
    },
    {
      "epoch": 108.07,
      "learning_rate": 5.590679611650486e-06,
      "loss": 0.0592,
      "step": 278280
    },
    {
      "epoch": 108.07,
      "learning_rate": 5.590161812297735e-06,
      "loss": 0.0001,
      "step": 278290
    },
    {
      "epoch": 108.08,
      "learning_rate": 5.5896440129449845e-06,
      "loss": 0.0358,
      "step": 278300
    },
    {
      "epoch": 108.08,
      "learning_rate": 5.589126213592233e-06,
      "loss": 0.0675,
      "step": 278310
    },
    {
      "epoch": 108.09,
      "learning_rate": 5.588608414239483e-06,
      "loss": 0.022,
      "step": 278320
    },
    {
      "epoch": 108.09,
      "learning_rate": 5.588090614886732e-06,
      "loss": 0.0866,
      "step": 278330
    },
    {
      "epoch": 108.09,
      "learning_rate": 5.587572815533981e-06,
      "loss": 0.0641,
      "step": 278340
    },
    {
      "epoch": 108.1,
      "learning_rate": 5.58705501618123e-06,
      "loss": 0.0385,
      "step": 278350
    },
    {
      "epoch": 108.1,
      "learning_rate": 5.58653721682848e-06,
      "loss": 0.042,
      "step": 278360
    },
    {
      "epoch": 108.1,
      "learning_rate": 5.586019417475728e-06,
      "loss": 0.0284,
      "step": 278370
    },
    {
      "epoch": 108.11,
      "learning_rate": 5.585501618122978e-06,
      "loss": 0.0183,
      "step": 278380
    },
    {
      "epoch": 108.11,
      "learning_rate": 5.584983818770227e-06,
      "loss": 0.0627,
      "step": 278390
    },
    {
      "epoch": 108.12,
      "learning_rate": 5.5844660194174764e-06,
      "loss": 0.0629,
      "step": 278400
    },
    {
      "epoch": 108.12,
      "learning_rate": 5.583948220064725e-06,
      "loss": 0.1208,
      "step": 278410
    },
    {
      "epoch": 108.12,
      "learning_rate": 5.583430420711975e-06,
      "loss": 0.0871,
      "step": 278420
    },
    {
      "epoch": 108.13,
      "learning_rate": 5.582912621359224e-06,
      "loss": 0.0002,
      "step": 278430
    },
    {
      "epoch": 108.13,
      "learning_rate": 5.582394822006472e-06,
      "loss": 0.0272,
      "step": 278440
    },
    {
      "epoch": 108.14,
      "learning_rate": 5.581877022653722e-06,
      "loss": 0.0598,
      "step": 278450
    },
    {
      "epoch": 108.14,
      "learning_rate": 5.581359223300972e-06,
      "loss": 0.0915,
      "step": 278460
    },
    {
      "epoch": 108.14,
      "learning_rate": 5.58084142394822e-06,
      "loss": 0.1158,
      "step": 278470
    },
    {
      "epoch": 108.15,
      "learning_rate": 5.580323624595469e-06,
      "loss": 0.037,
      "step": 278480
    },
    {
      "epoch": 108.15,
      "learning_rate": 5.579805825242719e-06,
      "loss": 0.0733,
      "step": 278490
    },
    {
      "epoch": 108.16,
      "learning_rate": 5.579288025889968e-06,
      "loss": 0.0008,
      "step": 278500
    },
    {
      "epoch": 108.16,
      "learning_rate": 5.578770226537217e-06,
      "loss": 0.0126,
      "step": 278510
    },
    {
      "epoch": 108.16,
      "learning_rate": 5.578252427184466e-06,
      "loss": 0.0304,
      "step": 278520
    },
    {
      "epoch": 108.17,
      "learning_rate": 5.5777346278317155e-06,
      "loss": 0.0022,
      "step": 278530
    },
    {
      "epoch": 108.17,
      "learning_rate": 5.577216828478965e-06,
      "loss": 0.0906,
      "step": 278540
    },
    {
      "epoch": 108.17,
      "learning_rate": 5.576699029126214e-06,
      "loss": 0.0607,
      "step": 278550
    },
    {
      "epoch": 108.18,
      "learning_rate": 5.576181229773463e-06,
      "loss": 0.101,
      "step": 278560
    },
    {
      "epoch": 108.18,
      "learning_rate": 5.575663430420712e-06,
      "loss": 0.0507,
      "step": 278570
    },
    {
      "epoch": 108.19,
      "learning_rate": 5.575145631067962e-06,
      "loss": 0.0013,
      "step": 278580
    },
    {
      "epoch": 108.19,
      "learning_rate": 5.57462783171521e-06,
      "loss": 0.14,
      "step": 278590
    },
    {
      "epoch": 108.19,
      "learning_rate": 5.5741100323624595e-06,
      "loss": 0.0035,
      "step": 278600
    },
    {
      "epoch": 108.2,
      "learning_rate": 5.573592233009709e-06,
      "loss": 0.0579,
      "step": 278610
    },
    {
      "epoch": 108.2,
      "learning_rate": 5.573074433656959e-06,
      "loss": 0.1343,
      "step": 278620
    },
    {
      "epoch": 108.21,
      "learning_rate": 5.572556634304208e-06,
      "loss": 0.0951,
      "step": 278630
    },
    {
      "epoch": 108.21,
      "learning_rate": 5.572038834951456e-06,
      "loss": 0.0058,
      "step": 278640
    },
    {
      "epoch": 108.21,
      "learning_rate": 5.571521035598706e-06,
      "loss": 0.0699,
      "step": 278650
    },
    {
      "epoch": 108.22,
      "learning_rate": 5.5710032362459555e-06,
      "loss": 0.009,
      "step": 278660
    },
    {
      "epoch": 108.22,
      "learning_rate": 5.570485436893205e-06,
      "loss": 0.0445,
      "step": 278670
    },
    {
      "epoch": 108.23,
      "learning_rate": 5.569967637540453e-06,
      "loss": 0.1145,
      "step": 278680
    },
    {
      "epoch": 108.23,
      "learning_rate": 5.569449838187703e-06,
      "loss": 0.0609,
      "step": 278690
    },
    {
      "epoch": 108.23,
      "learning_rate": 5.568932038834952e-06,
      "loss": 0.0561,
      "step": 278700
    },
    {
      "epoch": 108.24,
      "learning_rate": 5.568414239482202e-06,
      "loss": 0.1192,
      "step": 278710
    },
    {
      "epoch": 108.24,
      "learning_rate": 5.56789644012945e-06,
      "loss": 0.0007,
      "step": 278720
    },
    {
      "epoch": 108.24,
      "learning_rate": 5.5673786407766994e-06,
      "loss": 0.0233,
      "step": 278730
    },
    {
      "epoch": 108.25,
      "learning_rate": 5.566860841423949e-06,
      "loss": 0.0729,
      "step": 278740
    },
    {
      "epoch": 108.25,
      "learning_rate": 5.566343042071199e-06,
      "loss": 0.003,
      "step": 278750
    },
    {
      "epoch": 108.26,
      "learning_rate": 5.565825242718447e-06,
      "loss": 0.0328,
      "step": 278760
    },
    {
      "epoch": 108.26,
      "learning_rate": 5.565307443365696e-06,
      "loss": 0.0302,
      "step": 278770
    },
    {
      "epoch": 108.26,
      "learning_rate": 5.564789644012946e-06,
      "loss": 0.0388,
      "step": 278780
    },
    {
      "epoch": 108.27,
      "learning_rate": 5.5642718446601954e-06,
      "loss": 0.0003,
      "step": 278790
    },
    {
      "epoch": 108.27,
      "learning_rate": 5.563754045307443e-06,
      "loss": 0.1385,
      "step": 278800
    },
    {
      "epoch": 108.28,
      "learning_rate": 5.563236245954693e-06,
      "loss": 0.0084,
      "step": 278810
    },
    {
      "epoch": 108.28,
      "learning_rate": 5.562718446601943e-06,
      "loss": 0.0888,
      "step": 278820
    },
    {
      "epoch": 108.28,
      "learning_rate": 5.562200647249192e-06,
      "loss": 0.0464,
      "step": 278830
    },
    {
      "epoch": 108.29,
      "learning_rate": 5.56168284789644e-06,
      "loss": 0.0477,
      "step": 278840
    },
    {
      "epoch": 108.29,
      "learning_rate": 5.56116504854369e-06,
      "loss": 0.0017,
      "step": 278850
    },
    {
      "epoch": 108.3,
      "learning_rate": 5.560647249190939e-06,
      "loss": 0.1876,
      "step": 278860
    },
    {
      "epoch": 108.3,
      "learning_rate": 5.560129449838189e-06,
      "loss": 0.0828,
      "step": 278870
    },
    {
      "epoch": 108.3,
      "learning_rate": 5.559611650485437e-06,
      "loss": 0.0214,
      "step": 278880
    },
    {
      "epoch": 108.31,
      "learning_rate": 5.5590938511326865e-06,
      "loss": 0.0035,
      "step": 278890
    },
    {
      "epoch": 108.31,
      "learning_rate": 5.558576051779936e-06,
      "loss": 0.0157,
      "step": 278900
    },
    {
      "epoch": 108.31,
      "learning_rate": 5.558058252427185e-06,
      "loss": 0.0581,
      "step": 278910
    },
    {
      "epoch": 108.32,
      "learning_rate": 5.557540453074434e-06,
      "loss": 0.0659,
      "step": 278920
    },
    {
      "epoch": 108.32,
      "learning_rate": 5.557022653721683e-06,
      "loss": 0.0085,
      "step": 278930
    },
    {
      "epoch": 108.33,
      "learning_rate": 5.556504854368933e-06,
      "loss": 0.072,
      "step": 278940
    },
    {
      "epoch": 108.33,
      "learning_rate": 5.555987055016182e-06,
      "loss": 0.0018,
      "step": 278950
    },
    {
      "epoch": 108.33,
      "learning_rate": 5.5554692556634305e-06,
      "loss": 0.0369,
      "step": 278960
    },
    {
      "epoch": 108.34,
      "learning_rate": 5.55495145631068e-06,
      "loss": 0.0108,
      "step": 278970
    },
    {
      "epoch": 108.34,
      "learning_rate": 5.55443365695793e-06,
      "loss": 0.035,
      "step": 278980
    },
    {
      "epoch": 108.35,
      "learning_rate": 5.5539158576051785e-06,
      "loss": 0.0323,
      "step": 278990
    },
    {
      "epoch": 108.35,
      "learning_rate": 5.553398058252427e-06,
      "loss": 0.113,
      "step": 279000
    },
    {
      "epoch": 108.35,
      "learning_rate": 5.552880258899677e-06,
      "loss": 0.0266,
      "step": 279010
    },
    {
      "epoch": 108.36,
      "learning_rate": 5.5523624595469265e-06,
      "loss": 0.1914,
      "step": 279020
    },
    {
      "epoch": 108.36,
      "learning_rate": 5.551844660194175e-06,
      "loss": 0.044,
      "step": 279030
    },
    {
      "epoch": 108.37,
      "learning_rate": 5.551326860841424e-06,
      "loss": 0.0655,
      "step": 279040
    },
    {
      "epoch": 108.37,
      "learning_rate": 5.550809061488674e-06,
      "loss": 0.0674,
      "step": 279050
    },
    {
      "epoch": 108.37,
      "learning_rate": 5.550291262135922e-06,
      "loss": 0.0369,
      "step": 279060
    },
    {
      "epoch": 108.38,
      "learning_rate": 5.549773462783172e-06,
      "loss": 0.0595,
      "step": 279070
    },
    {
      "epoch": 108.38,
      "learning_rate": 5.549255663430421e-06,
      "loss": 0.0087,
      "step": 279080
    },
    {
      "epoch": 108.38,
      "learning_rate": 5.5487378640776704e-06,
      "loss": 0.249,
      "step": 279090
    },
    {
      "epoch": 108.39,
      "learning_rate": 5.548220064724919e-06,
      "loss": 0.0308,
      "step": 279100
    },
    {
      "epoch": 108.39,
      "learning_rate": 5.547702265372169e-06,
      "loss": 0.0633,
      "step": 279110
    },
    {
      "epoch": 108.4,
      "learning_rate": 5.547184466019418e-06,
      "loss": 0.0747,
      "step": 279120
    },
    {
      "epoch": 108.4,
      "learning_rate": 5.546666666666667e-06,
      "loss": 0.0003,
      "step": 279130
    },
    {
      "epoch": 108.4,
      "learning_rate": 5.546148867313916e-06,
      "loss": 0.0101,
      "step": 279140
    },
    {
      "epoch": 108.41,
      "learning_rate": 5.545631067961166e-06,
      "loss": 0.0192,
      "step": 279150
    },
    {
      "epoch": 108.41,
      "learning_rate": 5.545113268608414e-06,
      "loss": 0.0651,
      "step": 279160
    },
    {
      "epoch": 108.42,
      "learning_rate": 5.544595469255664e-06,
      "loss": 0.0374,
      "step": 279170
    },
    {
      "epoch": 108.42,
      "learning_rate": 5.544077669902913e-06,
      "loss": 0.2018,
      "step": 279180
    },
    {
      "epoch": 108.42,
      "learning_rate": 5.543559870550162e-06,
      "loss": 0.0151,
      "step": 279190
    },
    {
      "epoch": 108.43,
      "learning_rate": 5.543042071197412e-06,
      "loss": 0.0246,
      "step": 279200
    },
    {
      "epoch": 108.43,
      "learning_rate": 5.542524271844661e-06,
      "loss": 0.0001,
      "step": 279210
    },
    {
      "epoch": 108.43,
      "learning_rate": 5.5420064724919095e-06,
      "loss": 0.0153,
      "step": 279220
    },
    {
      "epoch": 108.44,
      "learning_rate": 5.541488673139159e-06,
      "loss": 0.0098,
      "step": 279230
    },
    {
      "epoch": 108.44,
      "learning_rate": 5.540970873786409e-06,
      "loss": 0.1474,
      "step": 279240
    },
    {
      "epoch": 108.45,
      "learning_rate": 5.540453074433657e-06,
      "loss": 0.0603,
      "step": 279250
    },
    {
      "epoch": 108.45,
      "learning_rate": 5.539935275080906e-06,
      "loss": 0.0165,
      "step": 279260
    },
    {
      "epoch": 108.45,
      "learning_rate": 5.539417475728156e-06,
      "loss": 0.0678,
      "step": 279270
    },
    {
      "epoch": 108.46,
      "learning_rate": 5.5388996763754055e-06,
      "loss": 0.0072,
      "step": 279280
    },
    {
      "epoch": 108.46,
      "learning_rate": 5.5383818770226535e-06,
      "loss": 0.0347,
      "step": 279290
    },
    {
      "epoch": 108.47,
      "learning_rate": 5.537864077669903e-06,
      "loss": 0.0388,
      "step": 279300
    },
    {
      "epoch": 108.47,
      "learning_rate": 5.537346278317153e-06,
      "loss": 0.014,
      "step": 279310
    },
    {
      "epoch": 108.47,
      "learning_rate": 5.536828478964402e-06,
      "loss": 0.0011,
      "step": 279320
    },
    {
      "epoch": 108.48,
      "learning_rate": 5.53631067961165e-06,
      "loss": 0.0663,
      "step": 279330
    },
    {
      "epoch": 108.48,
      "learning_rate": 5.5357928802589e-06,
      "loss": 0.1285,
      "step": 279340
    },
    {
      "epoch": 108.49,
      "learning_rate": 5.5352750809061495e-06,
      "loss": 0.0715,
      "step": 279350
    },
    {
      "epoch": 108.49,
      "learning_rate": 5.534757281553399e-06,
      "loss": 0.0329,
      "step": 279360
    },
    {
      "epoch": 108.49,
      "learning_rate": 5.534239482200647e-06,
      "loss": 0.0489,
      "step": 279370
    },
    {
      "epoch": 108.5,
      "learning_rate": 5.533721682847897e-06,
      "loss": 0.0233,
      "step": 279380
    },
    {
      "epoch": 108.5,
      "learning_rate": 5.533203883495146e-06,
      "loss": 0.0486,
      "step": 279390
    },
    {
      "epoch": 108.5,
      "learning_rate": 5.532686084142396e-06,
      "loss": 0.0014,
      "step": 279400
    },
    {
      "epoch": 108.51,
      "learning_rate": 5.532168284789644e-06,
      "loss": 0.0048,
      "step": 279410
    },
    {
      "epoch": 108.51,
      "learning_rate": 5.5316504854368934e-06,
      "loss": 0.1202,
      "step": 279420
    },
    {
      "epoch": 108.52,
      "learning_rate": 5.531132686084143e-06,
      "loss": 0.1714,
      "step": 279430
    },
    {
      "epoch": 108.52,
      "learning_rate": 5.530614886731393e-06,
      "loss": 0.0995,
      "step": 279440
    },
    {
      "epoch": 108.52,
      "learning_rate": 5.530097087378641e-06,
      "loss": 0.05,
      "step": 279450
    },
    {
      "epoch": 108.53,
      "learning_rate": 5.52957928802589e-06,
      "loss": 0.1,
      "step": 279460
    },
    {
      "epoch": 108.53,
      "learning_rate": 5.52906148867314e-06,
      "loss": 0.0763,
      "step": 279470
    },
    {
      "epoch": 108.54,
      "learning_rate": 5.5285436893203894e-06,
      "loss": 0.1219,
      "step": 279480
    },
    {
      "epoch": 108.54,
      "learning_rate": 5.528025889967637e-06,
      "loss": 0.102,
      "step": 279490
    },
    {
      "epoch": 108.54,
      "learning_rate": 5.527508090614887e-06,
      "loss": 0.0533,
      "step": 279500
    },
    {
      "epoch": 108.55,
      "learning_rate": 5.526990291262137e-06,
      "loss": 0.1294,
      "step": 279510
    },
    {
      "epoch": 108.55,
      "learning_rate": 5.526472491909386e-06,
      "loss": 0.0305,
      "step": 279520
    },
    {
      "epoch": 108.56,
      "learning_rate": 5.525954692556634e-06,
      "loss": 0.1597,
      "step": 279530
    },
    {
      "epoch": 108.56,
      "learning_rate": 5.525436893203884e-06,
      "loss": 0.0019,
      "step": 279540
    },
    {
      "epoch": 108.56,
      "learning_rate": 5.524919093851133e-06,
      "loss": 0.0848,
      "step": 279550
    },
    {
      "epoch": 108.57,
      "learning_rate": 5.524401294498383e-06,
      "loss": 0.0865,
      "step": 279560
    },
    {
      "epoch": 108.57,
      "learning_rate": 5.523883495145631e-06,
      "loss": 0.0173,
      "step": 279570
    },
    {
      "epoch": 108.57,
      "learning_rate": 5.5233656957928805e-06,
      "loss": 0.0056,
      "step": 279580
    },
    {
      "epoch": 108.58,
      "learning_rate": 5.52284789644013e-06,
      "loss": 0.0162,
      "step": 279590
    },
    {
      "epoch": 108.58,
      "learning_rate": 5.52233009708738e-06,
      "loss": 0.0524,
      "step": 279600
    },
    {
      "epoch": 108.59,
      "learning_rate": 5.521812297734628e-06,
      "loss": 0.1795,
      "step": 279610
    },
    {
      "epoch": 108.59,
      "learning_rate": 5.521294498381877e-06,
      "loss": 0.1228,
      "step": 279620
    },
    {
      "epoch": 108.59,
      "learning_rate": 5.520776699029127e-06,
      "loss": 0.0943,
      "step": 279630
    },
    {
      "epoch": 108.6,
      "learning_rate": 5.5202588996763766e-06,
      "loss": 0.0134,
      "step": 279640
    },
    {
      "epoch": 108.6,
      "learning_rate": 5.5197411003236245e-06,
      "loss": 0.0042,
      "step": 279650
    },
    {
      "epoch": 108.61,
      "learning_rate": 5.519223300970874e-06,
      "loss": 0.1291,
      "step": 279660
    },
    {
      "epoch": 108.61,
      "learning_rate": 5.518705501618124e-06,
      "loss": 0.0358,
      "step": 279670
    },
    {
      "epoch": 108.61,
      "learning_rate": 5.518187702265373e-06,
      "loss": 0.0547,
      "step": 279680
    },
    {
      "epoch": 108.62,
      "learning_rate": 5.517669902912621e-06,
      "loss": 0.0184,
      "step": 279690
    },
    {
      "epoch": 108.62,
      "learning_rate": 5.517152103559871e-06,
      "loss": 0.0075,
      "step": 279700
    },
    {
      "epoch": 108.63,
      "learning_rate": 5.5166343042071205e-06,
      "loss": 0.0571,
      "step": 279710
    },
    {
      "epoch": 108.63,
      "learning_rate": 5.516116504854369e-06,
      "loss": 0.0548,
      "step": 279720
    },
    {
      "epoch": 108.63,
      "learning_rate": 5.515598705501618e-06,
      "loss": 0.0965,
      "step": 279730
    },
    {
      "epoch": 108.64,
      "learning_rate": 5.515080906148868e-06,
      "loss": 0.0053,
      "step": 279740
    },
    {
      "epoch": 108.64,
      "learning_rate": 5.514563106796117e-06,
      "loss": 0.074,
      "step": 279750
    },
    {
      "epoch": 108.64,
      "learning_rate": 5.514045307443366e-06,
      "loss": 0.0044,
      "step": 279760
    },
    {
      "epoch": 108.65,
      "learning_rate": 5.513527508090616e-06,
      "loss": 0.0126,
      "step": 279770
    },
    {
      "epoch": 108.65,
      "learning_rate": 5.5130097087378644e-06,
      "loss": 0.0004,
      "step": 279780
    },
    {
      "epoch": 108.66,
      "learning_rate": 5.512491909385114e-06,
      "loss": 0.0084,
      "step": 279790
    },
    {
      "epoch": 108.66,
      "learning_rate": 5.511974110032363e-06,
      "loss": 0.1216,
      "step": 279800
    },
    {
      "epoch": 108.66,
      "learning_rate": 5.5114563106796124e-06,
      "loss": 0.0167,
      "step": 279810
    },
    {
      "epoch": 108.67,
      "learning_rate": 5.510938511326861e-06,
      "loss": 0.0007,
      "step": 279820
    },
    {
      "epoch": 108.67,
      "learning_rate": 5.510420711974111e-06,
      "loss": 0.059,
      "step": 279830
    },
    {
      "epoch": 108.68,
      "learning_rate": 5.50990291262136e-06,
      "loss": 0.0362,
      "step": 279840
    },
    {
      "epoch": 108.68,
      "learning_rate": 5.509385113268609e-06,
      "loss": 0.0004,
      "step": 279850
    },
    {
      "epoch": 108.68,
      "learning_rate": 5.508867313915858e-06,
      "loss": 0.0017,
      "step": 279860
    },
    {
      "epoch": 108.69,
      "learning_rate": 5.508349514563108e-06,
      "loss": 0.1115,
      "step": 279870
    },
    {
      "epoch": 108.69,
      "learning_rate": 5.507831715210356e-06,
      "loss": 0.0996,
      "step": 279880
    },
    {
      "epoch": 108.7,
      "learning_rate": 5.507313915857606e-06,
      "loss": 0.1305,
      "step": 279890
    },
    {
      "epoch": 108.7,
      "learning_rate": 5.506796116504855e-06,
      "loss": 0.1093,
      "step": 279900
    },
    {
      "epoch": 108.7,
      "learning_rate": 5.5062783171521035e-06,
      "loss": 0.0529,
      "step": 279910
    },
    {
      "epoch": 108.71,
      "learning_rate": 5.505760517799353e-06,
      "loss": 0.1208,
      "step": 279920
    },
    {
      "epoch": 108.71,
      "learning_rate": 5.505242718446603e-06,
      "loss": 0.0724,
      "step": 279930
    },
    {
      "epoch": 108.71,
      "learning_rate": 5.5047249190938515e-06,
      "loss": 0.0132,
      "step": 279940
    },
    {
      "epoch": 108.72,
      "learning_rate": 5.5042071197411e-06,
      "loss": 0.0109,
      "step": 279950
    },
    {
      "epoch": 108.72,
      "learning_rate": 5.50368932038835e-06,
      "loss": 0.0237,
      "step": 279960
    },
    {
      "epoch": 108.73,
      "learning_rate": 5.5031715210355995e-06,
      "loss": 0.0167,
      "step": 279970
    },
    {
      "epoch": 108.73,
      "learning_rate": 5.502653721682848e-06,
      "loss": 0.0003,
      "step": 279980
    },
    {
      "epoch": 108.73,
      "learning_rate": 5.502135922330097e-06,
      "loss": 0.1302,
      "step": 279990
    },
    {
      "epoch": 108.74,
      "learning_rate": 5.501618122977347e-06,
      "loss": 0.0131,
      "step": 280000
    },
    {
      "epoch": 108.74,
      "learning_rate": 5.501100323624596e-06,
      "loss": 0.0807,
      "step": 280010
    },
    {
      "epoch": 108.75,
      "learning_rate": 5.500582524271845e-06,
      "loss": 0.091,
      "step": 280020
    },
    {
      "epoch": 108.75,
      "learning_rate": 5.500064724919094e-06,
      "loss": 0.088,
      "step": 280030
    },
    {
      "epoch": 108.75,
      "learning_rate": 5.4995469255663435e-06,
      "loss": 0.0087,
      "step": 280040
    },
    {
      "epoch": 108.76,
      "learning_rate": 5.499029126213593e-06,
      "loss": 0.0014,
      "step": 280050
    },
    {
      "epoch": 108.76,
      "learning_rate": 5.498511326860841e-06,
      "loss": 0.0033,
      "step": 280060
    },
    {
      "epoch": 108.77,
      "learning_rate": 5.497993527508091e-06,
      "loss": 0.1208,
      "step": 280070
    },
    {
      "epoch": 108.77,
      "learning_rate": 5.49747572815534e-06,
      "loss": 0.1336,
      "step": 280080
    },
    {
      "epoch": 108.77,
      "learning_rate": 5.49695792880259e-06,
      "loss": 0.063,
      "step": 280090
    },
    {
      "epoch": 108.78,
      "learning_rate": 5.496440129449838e-06,
      "loss": 0.0146,
      "step": 280100
    },
    {
      "epoch": 108.78,
      "learning_rate": 5.495922330097087e-06,
      "loss": 0.0246,
      "step": 280110
    },
    {
      "epoch": 108.78,
      "learning_rate": 5.495404530744337e-06,
      "loss": 0.0262,
      "step": 280120
    },
    {
      "epoch": 108.79,
      "learning_rate": 5.494886731391587e-06,
      "loss": 0.0585,
      "step": 280130
    },
    {
      "epoch": 108.79,
      "learning_rate": 5.494368932038835e-06,
      "loss": 0.0002,
      "step": 280140
    },
    {
      "epoch": 108.8,
      "learning_rate": 5.493851132686084e-06,
      "loss": 0.0275,
      "step": 280150
    },
    {
      "epoch": 108.8,
      "learning_rate": 5.493333333333334e-06,
      "loss": 0.0731,
      "step": 280160
    },
    {
      "epoch": 108.8,
      "learning_rate": 5.4928155339805834e-06,
      "loss": 0.0963,
      "step": 280170
    },
    {
      "epoch": 108.81,
      "learning_rate": 5.492297734627831e-06,
      "loss": 0.1446,
      "step": 280180
    },
    {
      "epoch": 108.81,
      "learning_rate": 5.491779935275081e-06,
      "loss": 0.0248,
      "step": 280190
    },
    {
      "epoch": 108.82,
      "learning_rate": 5.491262135922331e-06,
      "loss": 0.1315,
      "step": 280200
    },
    {
      "epoch": 108.82,
      "learning_rate": 5.49074433656958e-06,
      "loss": 0.1841,
      "step": 280210
    },
    {
      "epoch": 108.82,
      "learning_rate": 5.490226537216828e-06,
      "loss": 0.0406,
      "step": 280220
    },
    {
      "epoch": 108.83,
      "learning_rate": 5.489708737864078e-06,
      "loss": 0.0111,
      "step": 280230
    },
    {
      "epoch": 108.83,
      "learning_rate": 5.489190938511327e-06,
      "loss": 0.0936,
      "step": 280240
    },
    {
      "epoch": 108.83,
      "learning_rate": 5.488673139158577e-06,
      "loss": 0.1842,
      "step": 280250
    },
    {
      "epoch": 108.84,
      "learning_rate": 5.488155339805825e-06,
      "loss": 0.0385,
      "step": 280260
    },
    {
      "epoch": 108.84,
      "learning_rate": 5.4876375404530745e-06,
      "loss": 0.0177,
      "step": 280270
    },
    {
      "epoch": 108.85,
      "learning_rate": 5.487119741100324e-06,
      "loss": 0.0424,
      "step": 280280
    },
    {
      "epoch": 108.85,
      "learning_rate": 5.486601941747574e-06,
      "loss": 0.0249,
      "step": 280290
    },
    {
      "epoch": 108.85,
      "learning_rate": 5.486084142394823e-06,
      "loss": 0.0393,
      "step": 280300
    },
    {
      "epoch": 108.86,
      "learning_rate": 5.485566343042071e-06,
      "loss": 0.0395,
      "step": 280310
    },
    {
      "epoch": 108.86,
      "learning_rate": 5.485048543689321e-06,
      "loss": 0.0101,
      "step": 280320
    },
    {
      "epoch": 108.87,
      "learning_rate": 5.4845307443365705e-06,
      "loss": 0.0013,
      "step": 280330
    },
    {
      "epoch": 108.87,
      "learning_rate": 5.48401294498382e-06,
      "loss": 0.0005,
      "step": 280340
    },
    {
      "epoch": 108.87,
      "learning_rate": 5.483495145631068e-06,
      "loss": 0.0238,
      "step": 280350
    },
    {
      "epoch": 108.88,
      "learning_rate": 5.482977346278318e-06,
      "loss": 0.0093,
      "step": 280360
    },
    {
      "epoch": 108.88,
      "learning_rate": 5.482459546925567e-06,
      "loss": 0.0556,
      "step": 280370
    },
    {
      "epoch": 108.89,
      "learning_rate": 5.481941747572816e-06,
      "loss": 0.1537,
      "step": 280380
    },
    {
      "epoch": 108.89,
      "learning_rate": 5.481423948220065e-06,
      "loss": 0.0973,
      "step": 280390
    },
    {
      "epoch": 108.89,
      "learning_rate": 5.4809061488673145e-06,
      "loss": 0.1341,
      "step": 280400
    },
    {
      "epoch": 108.9,
      "learning_rate": 5.480388349514564e-06,
      "loss": 0.0087,
      "step": 280410
    },
    {
      "epoch": 108.9,
      "learning_rate": 5.479870550161813e-06,
      "loss": 0.0449,
      "step": 280420
    },
    {
      "epoch": 108.9,
      "learning_rate": 5.479352750809062e-06,
      "loss": 0.0431,
      "step": 280430
    },
    {
      "epoch": 108.91,
      "learning_rate": 5.478834951456311e-06,
      "loss": 0.0204,
      "step": 280440
    },
    {
      "epoch": 108.91,
      "learning_rate": 5.478317152103561e-06,
      "loss": 0.0149,
      "step": 280450
    },
    {
      "epoch": 108.92,
      "learning_rate": 5.47779935275081e-06,
      "loss": 0.0068,
      "step": 280460
    },
    {
      "epoch": 108.92,
      "learning_rate": 5.4772815533980584e-06,
      "loss": 0.0661,
      "step": 280470
    },
    {
      "epoch": 108.92,
      "learning_rate": 5.476763754045308e-06,
      "loss": 0.0724,
      "step": 280480
    },
    {
      "epoch": 108.93,
      "learning_rate": 5.476245954692558e-06,
      "loss": 0.1044,
      "step": 280490
    },
    {
      "epoch": 108.93,
      "learning_rate": 5.4757281553398064e-06,
      "loss": 0.014,
      "step": 280500
    },
    {
      "epoch": 108.94,
      "learning_rate": 5.475210355987055e-06,
      "loss": 0.0778,
      "step": 280510
    },
    {
      "epoch": 108.94,
      "learning_rate": 5.474692556634305e-06,
      "loss": 0.0131,
      "step": 280520
    },
    {
      "epoch": 108.94,
      "learning_rate": 5.474174757281554e-06,
      "loss": 0.001,
      "step": 280530
    },
    {
      "epoch": 108.95,
      "learning_rate": 5.473656957928803e-06,
      "loss": 0.008,
      "step": 280540
    },
    {
      "epoch": 108.95,
      "learning_rate": 5.473139158576052e-06,
      "loss": 0.2071,
      "step": 280550
    },
    {
      "epoch": 108.96,
      "learning_rate": 5.472621359223302e-06,
      "loss": 0.0014,
      "step": 280560
    },
    {
      "epoch": 108.96,
      "learning_rate": 5.47210355987055e-06,
      "loss": 0.0398,
      "step": 280570
    },
    {
      "epoch": 108.96,
      "learning_rate": 5.4715857605178e-06,
      "loss": 0.0233,
      "step": 280580
    },
    {
      "epoch": 108.97,
      "learning_rate": 5.471067961165049e-06,
      "loss": 0.1256,
      "step": 280590
    },
    {
      "epoch": 108.97,
      "learning_rate": 5.470550161812298e-06,
      "loss": 0.1392,
      "step": 280600
    },
    {
      "epoch": 108.97,
      "learning_rate": 5.470032362459547e-06,
      "loss": 0.1028,
      "step": 280610
    },
    {
      "epoch": 108.98,
      "learning_rate": 5.469514563106797e-06,
      "loss": 0.0235,
      "step": 280620
    },
    {
      "epoch": 108.98,
      "learning_rate": 5.4689967637540455e-06,
      "loss": 0.141,
      "step": 280630
    },
    {
      "epoch": 108.99,
      "learning_rate": 5.468478964401295e-06,
      "loss": 0.0313,
      "step": 280640
    },
    {
      "epoch": 108.99,
      "learning_rate": 5.467961165048544e-06,
      "loss": 0.0323,
      "step": 280650
    },
    {
      "epoch": 108.99,
      "learning_rate": 5.4674433656957935e-06,
      "loss": 0.1213,
      "step": 280660
    },
    {
      "epoch": 109.0,
      "learning_rate": 5.466925566343042e-06,
      "loss": 0.0545,
      "step": 280670
    },
    {
      "epoch": 109.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3662911355495453,
      "eval_runtime": 8.2186,
      "eval_samples_per_second": 442.287,
      "eval_steps_per_second": 55.362,
      "step": 280675
    },
    {
      "epoch": 109.0,
      "learning_rate": 5.466407766990292e-06,
      "loss": 0.1973,
      "step": 280680
    },
    {
      "epoch": 109.01,
      "learning_rate": 5.465889967637541e-06,
      "loss": 0.0873,
      "step": 280690
    },
    {
      "epoch": 109.01,
      "learning_rate": 5.46537216828479e-06,
      "loss": 0.0596,
      "step": 280700
    },
    {
      "epoch": 109.01,
      "learning_rate": 5.464854368932039e-06,
      "loss": 0.0102,
      "step": 280710
    },
    {
      "epoch": 109.02,
      "learning_rate": 5.464336569579288e-06,
      "loss": 0.0003,
      "step": 280720
    },
    {
      "epoch": 109.02,
      "learning_rate": 5.4638187702265375e-06,
      "loss": 0.1827,
      "step": 280730
    },
    {
      "epoch": 109.03,
      "learning_rate": 5.463300970873787e-06,
      "loss": 0.0718,
      "step": 280740
    },
    {
      "epoch": 109.03,
      "learning_rate": 5.462783171521036e-06,
      "loss": 0.1124,
      "step": 280750
    },
    {
      "epoch": 109.03,
      "learning_rate": 5.462265372168285e-06,
      "loss": 0.0361,
      "step": 280760
    },
    {
      "epoch": 109.04,
      "learning_rate": 5.461747572815534e-06,
      "loss": 0.0814,
      "step": 280770
    },
    {
      "epoch": 109.04,
      "learning_rate": 5.461229773462784e-06,
      "loss": 0.035,
      "step": 280780
    },
    {
      "epoch": 109.04,
      "learning_rate": 5.460711974110033e-06,
      "loss": 0.0153,
      "step": 280790
    },
    {
      "epoch": 109.05,
      "learning_rate": 5.460194174757281e-06,
      "loss": 0.1077,
      "step": 280800
    },
    {
      "epoch": 109.05,
      "learning_rate": 5.459676375404531e-06,
      "loss": 0.0113,
      "step": 280810
    },
    {
      "epoch": 109.06,
      "learning_rate": 5.459158576051781e-06,
      "loss": 0.0018,
      "step": 280820
    },
    {
      "epoch": 109.06,
      "learning_rate": 5.4586407766990294e-06,
      "loss": 0.026,
      "step": 280830
    },
    {
      "epoch": 109.06,
      "learning_rate": 5.458122977346278e-06,
      "loss": 0.1534,
      "step": 280840
    },
    {
      "epoch": 109.07,
      "learning_rate": 5.457605177993528e-06,
      "loss": 0.0605,
      "step": 280850
    },
    {
      "epoch": 109.07,
      "learning_rate": 5.4570873786407774e-06,
      "loss": 0.0366,
      "step": 280860
    },
    {
      "epoch": 109.08,
      "learning_rate": 5.456569579288027e-06,
      "loss": 0.0001,
      "step": 280870
    },
    {
      "epoch": 109.08,
      "learning_rate": 5.456051779935275e-06,
      "loss": 0.0457,
      "step": 280880
    },
    {
      "epoch": 109.08,
      "learning_rate": 5.455533980582525e-06,
      "loss": 0.0748,
      "step": 280890
    },
    {
      "epoch": 109.09,
      "learning_rate": 5.455016181229774e-06,
      "loss": 0.0609,
      "step": 280900
    },
    {
      "epoch": 109.09,
      "learning_rate": 5.454498381877024e-06,
      "loss": 0.1361,
      "step": 280910
    },
    {
      "epoch": 109.1,
      "learning_rate": 5.453980582524272e-06,
      "loss": 0.0793,
      "step": 280920
    },
    {
      "epoch": 109.1,
      "learning_rate": 5.453462783171521e-06,
      "loss": 0.0328,
      "step": 280930
    },
    {
      "epoch": 109.1,
      "learning_rate": 5.452944983818771e-06,
      "loss": 0.0031,
      "step": 280940
    },
    {
      "epoch": 109.11,
      "learning_rate": 5.452427184466021e-06,
      "loss": 0.0442,
      "step": 280950
    },
    {
      "epoch": 109.11,
      "learning_rate": 5.4519093851132685e-06,
      "loss": 0.0848,
      "step": 280960
    },
    {
      "epoch": 109.11,
      "learning_rate": 5.451391585760518e-06,
      "loss": 0.0021,
      "step": 280970
    },
    {
      "epoch": 109.12,
      "learning_rate": 5.450873786407768e-06,
      "loss": 0.0457,
      "step": 280980
    },
    {
      "epoch": 109.12,
      "learning_rate": 5.450355987055017e-06,
      "loss": 0.0528,
      "step": 280990
    },
    {
      "epoch": 109.13,
      "learning_rate": 5.449838187702265e-06,
      "loss": 0.0869,
      "step": 281000
    },
    {
      "epoch": 109.13,
      "learning_rate": 5.449320388349515e-06,
      "loss": 0.0684,
      "step": 281010
    },
    {
      "epoch": 109.13,
      "learning_rate": 5.4488025889967645e-06,
      "loss": 0.0013,
      "step": 281020
    },
    {
      "epoch": 109.14,
      "learning_rate": 5.448284789644014e-06,
      "loss": 0.0818,
      "step": 281030
    },
    {
      "epoch": 109.14,
      "learning_rate": 5.447766990291262e-06,
      "loss": 0.0647,
      "step": 281040
    },
    {
      "epoch": 109.15,
      "learning_rate": 5.447249190938512e-06,
      "loss": 0.0285,
      "step": 281050
    },
    {
      "epoch": 109.15,
      "learning_rate": 5.446731391585761e-06,
      "loss": 0.0687,
      "step": 281060
    },
    {
      "epoch": 109.15,
      "learning_rate": 5.446213592233011e-06,
      "loss": 0.0105,
      "step": 281070
    },
    {
      "epoch": 109.16,
      "learning_rate": 5.445695792880259e-06,
      "loss": 0.1491,
      "step": 281080
    },
    {
      "epoch": 109.16,
      "learning_rate": 5.4451779935275085e-06,
      "loss": 0.0306,
      "step": 281090
    },
    {
      "epoch": 109.17,
      "learning_rate": 5.444660194174758e-06,
      "loss": 0.0425,
      "step": 281100
    },
    {
      "epoch": 109.17,
      "learning_rate": 5.444142394822008e-06,
      "loss": 0.1132,
      "step": 281110
    },
    {
      "epoch": 109.17,
      "learning_rate": 5.443624595469256e-06,
      "loss": 0.0001,
      "step": 281120
    },
    {
      "epoch": 109.18,
      "learning_rate": 5.443106796116505e-06,
      "loss": 0.0861,
      "step": 281130
    },
    {
      "epoch": 109.18,
      "learning_rate": 5.442588996763755e-06,
      "loss": 0.0379,
      "step": 281140
    },
    {
      "epoch": 109.18,
      "learning_rate": 5.4420711974110045e-06,
      "loss": 0.0359,
      "step": 281150
    },
    {
      "epoch": 109.19,
      "learning_rate": 5.4415533980582524e-06,
      "loss": 0.0862,
      "step": 281160
    },
    {
      "epoch": 109.19,
      "learning_rate": 5.441035598705502e-06,
      "loss": 0.085,
      "step": 281170
    },
    {
      "epoch": 109.2,
      "learning_rate": 5.440517799352752e-06,
      "loss": 0.0623,
      "step": 281180
    },
    {
      "epoch": 109.2,
      "learning_rate": 5.4400000000000004e-06,
      "loss": 0.0917,
      "step": 281190
    },
    {
      "epoch": 109.2,
      "learning_rate": 5.439482200647249e-06,
      "loss": 0.0242,
      "step": 281200
    },
    {
      "epoch": 109.21,
      "learning_rate": 5.438964401294499e-06,
      "loss": 0.0444,
      "step": 281210
    },
    {
      "epoch": 109.21,
      "learning_rate": 5.4384466019417484e-06,
      "loss": 0.1922,
      "step": 281220
    },
    {
      "epoch": 109.22,
      "learning_rate": 5.437928802588997e-06,
      "loss": 0.0431,
      "step": 281230
    },
    {
      "epoch": 109.22,
      "learning_rate": 5.437411003236246e-06,
      "loss": 0.0444,
      "step": 281240
    },
    {
      "epoch": 109.22,
      "learning_rate": 5.436893203883496e-06,
      "loss": 0.1075,
      "step": 281250
    },
    {
      "epoch": 109.23,
      "learning_rate": 5.436375404530745e-06,
      "loss": 0.0225,
      "step": 281260
    },
    {
      "epoch": 109.23,
      "learning_rate": 5.435857605177994e-06,
      "loss": 0.0703,
      "step": 281270
    },
    {
      "epoch": 109.23,
      "learning_rate": 5.435339805825243e-06,
      "loss": 0.0315,
      "step": 281280
    },
    {
      "epoch": 109.24,
      "learning_rate": 5.434822006472492e-06,
      "loss": 0.0429,
      "step": 281290
    },
    {
      "epoch": 109.24,
      "learning_rate": 5.434304207119742e-06,
      "loss": 0.0182,
      "step": 281300
    },
    {
      "epoch": 109.25,
      "learning_rate": 5.433786407766991e-06,
      "loss": 0.0963,
      "step": 281310
    },
    {
      "epoch": 109.25,
      "learning_rate": 5.4332686084142395e-06,
      "loss": 0.1197,
      "step": 281320
    },
    {
      "epoch": 109.25,
      "learning_rate": 5.432750809061489e-06,
      "loss": 0.0082,
      "step": 281330
    },
    {
      "epoch": 109.26,
      "learning_rate": 5.432233009708738e-06,
      "loss": 0.0096,
      "step": 281340
    },
    {
      "epoch": 109.26,
      "learning_rate": 5.4317152103559875e-06,
      "loss": 0.0141,
      "step": 281350
    },
    {
      "epoch": 109.27,
      "learning_rate": 5.431197411003236e-06,
      "loss": 0.1285,
      "step": 281360
    },
    {
      "epoch": 109.27,
      "learning_rate": 5.430679611650486e-06,
      "loss": 0.0442,
      "step": 281370
    },
    {
      "epoch": 109.27,
      "learning_rate": 5.430161812297735e-06,
      "loss": 0.0464,
      "step": 281380
    },
    {
      "epoch": 109.28,
      "learning_rate": 5.429644012944984e-06,
      "loss": 0.0131,
      "step": 281390
    },
    {
      "epoch": 109.28,
      "learning_rate": 5.429126213592233e-06,
      "loss": 0.029,
      "step": 281400
    },
    {
      "epoch": 109.29,
      "learning_rate": 5.428608414239483e-06,
      "loss": 0.0109,
      "step": 281410
    },
    {
      "epoch": 109.29,
      "learning_rate": 5.4280906148867315e-06,
      "loss": 0.0678,
      "step": 281420
    },
    {
      "epoch": 109.29,
      "learning_rate": 5.427572815533981e-06,
      "loss": 0.0208,
      "step": 281430
    },
    {
      "epoch": 109.3,
      "learning_rate": 5.427055016181231e-06,
      "loss": 0.093,
      "step": 281440
    },
    {
      "epoch": 109.3,
      "learning_rate": 5.4265372168284795e-06,
      "loss": 0.2169,
      "step": 281450
    },
    {
      "epoch": 109.3,
      "learning_rate": 5.426019417475728e-06,
      "loss": 0.098,
      "step": 281460
    },
    {
      "epoch": 109.31,
      "learning_rate": 5.425501618122978e-06,
      "loss": 0.0101,
      "step": 281470
    },
    {
      "epoch": 109.31,
      "learning_rate": 5.4249838187702275e-06,
      "loss": 0.0156,
      "step": 281480
    },
    {
      "epoch": 109.32,
      "learning_rate": 5.424466019417476e-06,
      "loss": 0.0062,
      "step": 281490
    },
    {
      "epoch": 109.32,
      "learning_rate": 5.423948220064725e-06,
      "loss": 0.0032,
      "step": 281500
    },
    {
      "epoch": 109.32,
      "learning_rate": 5.423430420711975e-06,
      "loss": 0.0173,
      "step": 281510
    },
    {
      "epoch": 109.33,
      "learning_rate": 5.422912621359224e-06,
      "loss": 0.0981,
      "step": 281520
    },
    {
      "epoch": 109.33,
      "learning_rate": 5.422394822006472e-06,
      "loss": 0.0546,
      "step": 281530
    },
    {
      "epoch": 109.34,
      "learning_rate": 5.421877022653722e-06,
      "loss": 0.019,
      "step": 281540
    },
    {
      "epoch": 109.34,
      "learning_rate": 5.4213592233009714e-06,
      "loss": 0.027,
      "step": 281550
    },
    {
      "epoch": 109.34,
      "learning_rate": 5.420841423948221e-06,
      "loss": 0.0787,
      "step": 281560
    },
    {
      "epoch": 109.35,
      "learning_rate": 5.420323624595469e-06,
      "loss": 0.0142,
      "step": 281570
    },
    {
      "epoch": 109.35,
      "learning_rate": 5.419805825242719e-06,
      "loss": 0.0794,
      "step": 281580
    },
    {
      "epoch": 109.36,
      "learning_rate": 5.419288025889968e-06,
      "loss": 0.0122,
      "step": 281590
    },
    {
      "epoch": 109.36,
      "learning_rate": 5.418770226537218e-06,
      "loss": 0.0297,
      "step": 281600
    },
    {
      "epoch": 109.36,
      "learning_rate": 5.418252427184466e-06,
      "loss": 0.2012,
      "step": 281610
    },
    {
      "epoch": 109.37,
      "learning_rate": 5.417734627831715e-06,
      "loss": 0.0507,
      "step": 281620
    },
    {
      "epoch": 109.37,
      "learning_rate": 5.417216828478965e-06,
      "loss": 0.1149,
      "step": 281630
    },
    {
      "epoch": 109.37,
      "learning_rate": 5.416699029126215e-06,
      "loss": 0.043,
      "step": 281640
    },
    {
      "epoch": 109.38,
      "learning_rate": 5.4161812297734625e-06,
      "loss": 0.0256,
      "step": 281650
    },
    {
      "epoch": 109.38,
      "learning_rate": 5.415663430420712e-06,
      "loss": 0.035,
      "step": 281660
    },
    {
      "epoch": 109.39,
      "learning_rate": 5.415145631067962e-06,
      "loss": 0.0607,
      "step": 281670
    },
    {
      "epoch": 109.39,
      "learning_rate": 5.414627831715211e-06,
      "loss": 0.0253,
      "step": 281680
    },
    {
      "epoch": 109.39,
      "learning_rate": 5.414110032362459e-06,
      "loss": 0.0421,
      "step": 281690
    },
    {
      "epoch": 109.4,
      "learning_rate": 5.413592233009709e-06,
      "loss": 0.0307,
      "step": 281700
    },
    {
      "epoch": 109.4,
      "learning_rate": 5.4130744336569585e-06,
      "loss": 0.0333,
      "step": 281710
    },
    {
      "epoch": 109.41,
      "learning_rate": 5.412556634304208e-06,
      "loss": 0.0232,
      "step": 281720
    },
    {
      "epoch": 109.41,
      "learning_rate": 5.412038834951456e-06,
      "loss": 0.0788,
      "step": 281730
    },
    {
      "epoch": 109.41,
      "learning_rate": 5.411521035598706e-06,
      "loss": 0.0786,
      "step": 281740
    },
    {
      "epoch": 109.42,
      "learning_rate": 5.411003236245955e-06,
      "loss": 0.011,
      "step": 281750
    },
    {
      "epoch": 109.42,
      "learning_rate": 5.410485436893205e-06,
      "loss": 0.0001,
      "step": 281760
    },
    {
      "epoch": 109.43,
      "learning_rate": 5.409967637540453e-06,
      "loss": 0.0026,
      "step": 281770
    },
    {
      "epoch": 109.43,
      "learning_rate": 5.4094498381877025e-06,
      "loss": 0.0948,
      "step": 281780
    },
    {
      "epoch": 109.43,
      "learning_rate": 5.408932038834952e-06,
      "loss": 0.0512,
      "step": 281790
    },
    {
      "epoch": 109.44,
      "learning_rate": 5.408414239482202e-06,
      "loss": 0.0117,
      "step": 281800
    },
    {
      "epoch": 109.44,
      "learning_rate": 5.40789644012945e-06,
      "loss": 0.0025,
      "step": 281810
    },
    {
      "epoch": 109.44,
      "learning_rate": 5.407378640776699e-06,
      "loss": 0.0991,
      "step": 281820
    },
    {
      "epoch": 109.45,
      "learning_rate": 5.406860841423949e-06,
      "loss": 0.1095,
      "step": 281830
    },
    {
      "epoch": 109.45,
      "learning_rate": 5.4063430420711985e-06,
      "loss": 0.1055,
      "step": 281840
    },
    {
      "epoch": 109.46,
      "learning_rate": 5.405825242718446e-06,
      "loss": 0.0288,
      "step": 281850
    },
    {
      "epoch": 109.46,
      "learning_rate": 5.405307443365696e-06,
      "loss": 0.0729,
      "step": 281860
    },
    {
      "epoch": 109.46,
      "learning_rate": 5.404789644012946e-06,
      "loss": 0.0002,
      "step": 281870
    },
    {
      "epoch": 109.47,
      "learning_rate": 5.404271844660195e-06,
      "loss": 0.0028,
      "step": 281880
    },
    {
      "epoch": 109.47,
      "learning_rate": 5.403754045307443e-06,
      "loss": 0.0689,
      "step": 281890
    },
    {
      "epoch": 109.48,
      "learning_rate": 5.403236245954693e-06,
      "loss": 0.0596,
      "step": 281900
    },
    {
      "epoch": 109.48,
      "learning_rate": 5.4027184466019424e-06,
      "loss": 0.0177,
      "step": 281910
    },
    {
      "epoch": 109.48,
      "learning_rate": 5.402200647249192e-06,
      "loss": 0.0201,
      "step": 281920
    },
    {
      "epoch": 109.49,
      "learning_rate": 5.40168284789644e-06,
      "loss": 0.0253,
      "step": 281930
    },
    {
      "epoch": 109.49,
      "learning_rate": 5.40116504854369e-06,
      "loss": 0.013,
      "step": 281940
    },
    {
      "epoch": 109.5,
      "learning_rate": 5.400647249190939e-06,
      "loss": 0.0042,
      "step": 281950
    },
    {
      "epoch": 109.5,
      "learning_rate": 5.400129449838189e-06,
      "loss": 0.0386,
      "step": 281960
    },
    {
      "epoch": 109.5,
      "learning_rate": 5.399611650485437e-06,
      "loss": 0.0324,
      "step": 281970
    },
    {
      "epoch": 109.51,
      "learning_rate": 5.399093851132686e-06,
      "loss": 0.0108,
      "step": 281980
    },
    {
      "epoch": 109.51,
      "learning_rate": 5.398576051779936e-06,
      "loss": 0.0928,
      "step": 281990
    },
    {
      "epoch": 109.51,
      "learning_rate": 5.398058252427185e-06,
      "loss": 0.0032,
      "step": 282000
    },
    {
      "epoch": 109.52,
      "learning_rate": 5.397540453074434e-06,
      "loss": 0.046,
      "step": 282010
    },
    {
      "epoch": 109.52,
      "learning_rate": 5.397022653721683e-06,
      "loss": 0.1248,
      "step": 282020
    },
    {
      "epoch": 109.53,
      "learning_rate": 5.396504854368933e-06,
      "loss": 0.0009,
      "step": 282030
    },
    {
      "epoch": 109.53,
      "learning_rate": 5.3959870550161815e-06,
      "loss": 0.0724,
      "step": 282040
    },
    {
      "epoch": 109.53,
      "learning_rate": 5.395469255663431e-06,
      "loss": 0.0554,
      "step": 282050
    },
    {
      "epoch": 109.54,
      "learning_rate": 5.39495145631068e-06,
      "loss": 0.0312,
      "step": 282060
    },
    {
      "epoch": 109.54,
      "learning_rate": 5.3944336569579295e-06,
      "loss": 0.1031,
      "step": 282070
    },
    {
      "epoch": 109.55,
      "learning_rate": 5.393915857605178e-06,
      "loss": 0.0612,
      "step": 282080
    },
    {
      "epoch": 109.55,
      "learning_rate": 5.393398058252428e-06,
      "loss": 0.0093,
      "step": 282090
    },
    {
      "epoch": 109.55,
      "learning_rate": 5.392880258899677e-06,
      "loss": 0.0296,
      "step": 282100
    },
    {
      "epoch": 109.56,
      "learning_rate": 5.392362459546926e-06,
      "loss": 0.0315,
      "step": 282110
    },
    {
      "epoch": 109.56,
      "learning_rate": 5.391844660194175e-06,
      "loss": 0.0466,
      "step": 282120
    },
    {
      "epoch": 109.57,
      "learning_rate": 5.391326860841425e-06,
      "loss": 0.1441,
      "step": 282130
    },
    {
      "epoch": 109.57,
      "learning_rate": 5.3908090614886735e-06,
      "loss": 0.1307,
      "step": 282140
    },
    {
      "epoch": 109.57,
      "learning_rate": 5.390291262135923e-06,
      "loss": 0.0151,
      "step": 282150
    },
    {
      "epoch": 109.58,
      "learning_rate": 5.389773462783172e-06,
      "loss": 0.0668,
      "step": 282160
    },
    {
      "epoch": 109.58,
      "learning_rate": 5.3892556634304215e-06,
      "loss": 0.0623,
      "step": 282170
    },
    {
      "epoch": 109.58,
      "learning_rate": 5.38873786407767e-06,
      "loss": 0.0718,
      "step": 282180
    },
    {
      "epoch": 109.59,
      "learning_rate": 5.388220064724919e-06,
      "loss": 0.071,
      "step": 282190
    },
    {
      "epoch": 109.59,
      "learning_rate": 5.387702265372169e-06,
      "loss": 0.0479,
      "step": 282200
    },
    {
      "epoch": 109.6,
      "learning_rate": 5.387184466019418e-06,
      "loss": 0.0448,
      "step": 282210
    },
    {
      "epoch": 109.6,
      "learning_rate": 5.386666666666667e-06,
      "loss": 0.0203,
      "step": 282220
    },
    {
      "epoch": 109.6,
      "learning_rate": 5.386148867313916e-06,
      "loss": 0.0577,
      "step": 282230
    },
    {
      "epoch": 109.61,
      "learning_rate": 5.3856310679611654e-06,
      "loss": 0.1391,
      "step": 282240
    },
    {
      "epoch": 109.61,
      "learning_rate": 5.385113268608415e-06,
      "loss": 0.0011,
      "step": 282250
    },
    {
      "epoch": 109.62,
      "learning_rate": 5.384595469255664e-06,
      "loss": 0.0898,
      "step": 282260
    },
    {
      "epoch": 109.62,
      "learning_rate": 5.384077669902913e-06,
      "loss": 0.0067,
      "step": 282270
    },
    {
      "epoch": 109.62,
      "learning_rate": 5.383559870550162e-06,
      "loss": 0.1435,
      "step": 282280
    },
    {
      "epoch": 109.63,
      "learning_rate": 5.383042071197412e-06,
      "loss": 0.0744,
      "step": 282290
    },
    {
      "epoch": 109.63,
      "learning_rate": 5.382524271844661e-06,
      "loss": 0.0009,
      "step": 282300
    },
    {
      "epoch": 109.63,
      "learning_rate": 5.382006472491909e-06,
      "loss": 0.0856,
      "step": 282310
    },
    {
      "epoch": 109.64,
      "learning_rate": 5.381488673139159e-06,
      "loss": 0.0538,
      "step": 282320
    },
    {
      "epoch": 109.64,
      "learning_rate": 5.380970873786409e-06,
      "loss": 0.0496,
      "step": 282330
    },
    {
      "epoch": 109.65,
      "learning_rate": 5.3804530744336565e-06,
      "loss": 0.0874,
      "step": 282340
    },
    {
      "epoch": 109.65,
      "learning_rate": 5.379935275080906e-06,
      "loss": 0.1401,
      "step": 282350
    },
    {
      "epoch": 109.65,
      "learning_rate": 5.379417475728156e-06,
      "loss": 0.0808,
      "step": 282360
    },
    {
      "epoch": 109.66,
      "learning_rate": 5.378899676375405e-06,
      "loss": 0.0252,
      "step": 282370
    },
    {
      "epoch": 109.66,
      "learning_rate": 5.378381877022653e-06,
      "loss": 0.04,
      "step": 282380
    },
    {
      "epoch": 109.67,
      "learning_rate": 5.377864077669903e-06,
      "loss": 0.0702,
      "step": 282390
    },
    {
      "epoch": 109.67,
      "learning_rate": 5.3773462783171525e-06,
      "loss": 0.0973,
      "step": 282400
    },
    {
      "epoch": 109.67,
      "learning_rate": 5.376828478964402e-06,
      "loss": 0.0259,
      "step": 282410
    },
    {
      "epoch": 109.68,
      "learning_rate": 5.37631067961165e-06,
      "loss": 0.0443,
      "step": 282420
    },
    {
      "epoch": 109.68,
      "learning_rate": 5.3757928802589e-06,
      "loss": 0.0402,
      "step": 282430
    },
    {
      "epoch": 109.69,
      "learning_rate": 5.375275080906149e-06,
      "loss": 0.0266,
      "step": 282440
    },
    {
      "epoch": 109.69,
      "learning_rate": 5.374757281553399e-06,
      "loss": 0.0082,
      "step": 282450
    },
    {
      "epoch": 109.69,
      "learning_rate": 5.374239482200647e-06,
      "loss": 0.0118,
      "step": 282460
    },
    {
      "epoch": 109.7,
      "learning_rate": 5.3737216828478965e-06,
      "loss": 0.1169,
      "step": 282470
    },
    {
      "epoch": 109.7,
      "learning_rate": 5.373203883495146e-06,
      "loss": 0.0197,
      "step": 282480
    },
    {
      "epoch": 109.7,
      "learning_rate": 5.372686084142396e-06,
      "loss": 0.0589,
      "step": 282490
    },
    {
      "epoch": 109.71,
      "learning_rate": 5.372168284789644e-06,
      "loss": 0.0075,
      "step": 282500
    },
    {
      "epoch": 109.71,
      "learning_rate": 5.371650485436893e-06,
      "loss": 0.1597,
      "step": 282510
    },
    {
      "epoch": 109.72,
      "learning_rate": 5.371132686084143e-06,
      "loss": 0.0642,
      "step": 282520
    },
    {
      "epoch": 109.72,
      "learning_rate": 5.3706148867313925e-06,
      "loss": 0.0602,
      "step": 282530
    },
    {
      "epoch": 109.72,
      "learning_rate": 5.370097087378642e-06,
      "loss": 0.0635,
      "step": 282540
    },
    {
      "epoch": 109.73,
      "learning_rate": 5.36957928802589e-06,
      "loss": 0.002,
      "step": 282550
    },
    {
      "epoch": 109.73,
      "learning_rate": 5.36906148867314e-06,
      "loss": 0.0715,
      "step": 282560
    },
    {
      "epoch": 109.74,
      "learning_rate": 5.368543689320389e-06,
      "loss": 0.1458,
      "step": 282570
    },
    {
      "epoch": 109.74,
      "learning_rate": 5.368025889967639e-06,
      "loss": 0.1386,
      "step": 282580
    },
    {
      "epoch": 109.74,
      "learning_rate": 5.367508090614887e-06,
      "loss": 0.0021,
      "step": 282590
    },
    {
      "epoch": 109.75,
      "learning_rate": 5.3669902912621364e-06,
      "loss": 0.0162,
      "step": 282600
    },
    {
      "epoch": 109.75,
      "learning_rate": 5.366472491909386e-06,
      "loss": 0.1106,
      "step": 282610
    },
    {
      "epoch": 109.76,
      "learning_rate": 5.365954692556636e-06,
      "loss": 0.0659,
      "step": 282620
    },
    {
      "epoch": 109.76,
      "learning_rate": 5.365436893203884e-06,
      "loss": 0.0108,
      "step": 282630
    },
    {
      "epoch": 109.76,
      "learning_rate": 5.364919093851133e-06,
      "loss": 0.0407,
      "step": 282640
    },
    {
      "epoch": 109.77,
      "learning_rate": 5.364401294498383e-06,
      "loss": 0.0944,
      "step": 282650
    },
    {
      "epoch": 109.77,
      "learning_rate": 5.363883495145632e-06,
      "loss": 0.1025,
      "step": 282660
    },
    {
      "epoch": 109.77,
      "learning_rate": 5.36336569579288e-06,
      "loss": 0.0065,
      "step": 282670
    },
    {
      "epoch": 109.78,
      "learning_rate": 5.36284789644013e-06,
      "loss": 0.0002,
      "step": 282680
    },
    {
      "epoch": 109.78,
      "learning_rate": 5.36233009708738e-06,
      "loss": 0.0485,
      "step": 282690
    },
    {
      "epoch": 109.79,
      "learning_rate": 5.361812297734628e-06,
      "loss": 0.0112,
      "step": 282700
    },
    {
      "epoch": 109.79,
      "learning_rate": 5.361294498381877e-06,
      "loss": 0.0133,
      "step": 282710
    },
    {
      "epoch": 109.79,
      "learning_rate": 5.360776699029127e-06,
      "loss": 0.0215,
      "step": 282720
    },
    {
      "epoch": 109.8,
      "learning_rate": 5.360258899676376e-06,
      "loss": 0.014,
      "step": 282730
    },
    {
      "epoch": 109.8,
      "learning_rate": 5.359741100323625e-06,
      "loss": 0.0325,
      "step": 282740
    },
    {
      "epoch": 109.81,
      "learning_rate": 5.359223300970874e-06,
      "loss": 0.0535,
      "step": 282750
    },
    {
      "epoch": 109.81,
      "learning_rate": 5.3587055016181235e-06,
      "loss": 0.0724,
      "step": 282760
    },
    {
      "epoch": 109.81,
      "learning_rate": 5.358187702265373e-06,
      "loss": 0.0195,
      "step": 282770
    },
    {
      "epoch": 109.82,
      "learning_rate": 5.357669902912622e-06,
      "loss": 0.0957,
      "step": 282780
    },
    {
      "epoch": 109.82,
      "learning_rate": 5.357152103559871e-06,
      "loss": 0.004,
      "step": 282790
    },
    {
      "epoch": 109.83,
      "learning_rate": 5.35663430420712e-06,
      "loss": 0.0461,
      "step": 282800
    },
    {
      "epoch": 109.83,
      "learning_rate": 5.356116504854369e-06,
      "loss": 0.0126,
      "step": 282810
    },
    {
      "epoch": 109.83,
      "learning_rate": 5.355598705501619e-06,
      "loss": 0.1068,
      "step": 282820
    },
    {
      "epoch": 109.84,
      "learning_rate": 5.3550809061488675e-06,
      "loss": 0.0786,
      "step": 282830
    },
    {
      "epoch": 109.84,
      "learning_rate": 5.354563106796117e-06,
      "loss": 0.0135,
      "step": 282840
    },
    {
      "epoch": 109.84,
      "learning_rate": 5.354045307443366e-06,
      "loss": 0.0485,
      "step": 282850
    },
    {
      "epoch": 109.85,
      "learning_rate": 5.3535275080906155e-06,
      "loss": 0.1129,
      "step": 282860
    },
    {
      "epoch": 109.85,
      "learning_rate": 5.353009708737864e-06,
      "loss": 0.04,
      "step": 282870
    },
    {
      "epoch": 109.86,
      "learning_rate": 5.352491909385114e-06,
      "loss": 0.0002,
      "step": 282880
    },
    {
      "epoch": 109.86,
      "learning_rate": 5.351974110032363e-06,
      "loss": 0.0904,
      "step": 282890
    },
    {
      "epoch": 109.86,
      "learning_rate": 5.351456310679612e-06,
      "loss": 0.0292,
      "step": 282900
    },
    {
      "epoch": 109.87,
      "learning_rate": 5.350938511326861e-06,
      "loss": 0.0643,
      "step": 282910
    },
    {
      "epoch": 109.87,
      "learning_rate": 5.350420711974111e-06,
      "loss": 0.098,
      "step": 282920
    },
    {
      "epoch": 109.88,
      "learning_rate": 5.3499029126213594e-06,
      "loss": 0.0082,
      "step": 282930
    },
    {
      "epoch": 109.88,
      "learning_rate": 5.349385113268609e-06,
      "loss": 0.0099,
      "step": 282940
    },
    {
      "epoch": 109.88,
      "learning_rate": 5.348867313915858e-06,
      "loss": 0.1181,
      "step": 282950
    },
    {
      "epoch": 109.89,
      "learning_rate": 5.3483495145631074e-06,
      "loss": 0.0973,
      "step": 282960
    },
    {
      "epoch": 109.89,
      "learning_rate": 5.347831715210356e-06,
      "loss": 0.1662,
      "step": 282970
    },
    {
      "epoch": 109.9,
      "learning_rate": 5.347313915857606e-06,
      "loss": 0.0387,
      "step": 282980
    },
    {
      "epoch": 109.9,
      "learning_rate": 5.346796116504855e-06,
      "loss": 0.0009,
      "step": 282990
    },
    {
      "epoch": 109.9,
      "learning_rate": 5.346278317152103e-06,
      "loss": 0.0601,
      "step": 283000
    },
    {
      "epoch": 109.91,
      "learning_rate": 5.345760517799353e-06,
      "loss": 0.0613,
      "step": 283010
    },
    {
      "epoch": 109.91,
      "learning_rate": 5.345242718446603e-06,
      "loss": 0.01,
      "step": 283020
    },
    {
      "epoch": 109.91,
      "learning_rate": 5.344724919093851e-06,
      "loss": 0.0155,
      "step": 283030
    },
    {
      "epoch": 109.92,
      "learning_rate": 5.3442071197411e-06,
      "loss": 0.1427,
      "step": 283040
    },
    {
      "epoch": 109.92,
      "learning_rate": 5.34368932038835e-06,
      "loss": 0.1307,
      "step": 283050
    },
    {
      "epoch": 109.93,
      "learning_rate": 5.343171521035599e-06,
      "loss": 0.0202,
      "step": 283060
    },
    {
      "epoch": 109.93,
      "learning_rate": 5.342653721682848e-06,
      "loss": 0.0657,
      "step": 283070
    },
    {
      "epoch": 109.93,
      "learning_rate": 5.342135922330097e-06,
      "loss": 0.1823,
      "step": 283080
    },
    {
      "epoch": 109.94,
      "learning_rate": 5.3416181229773465e-06,
      "loss": 0.0781,
      "step": 283090
    },
    {
      "epoch": 109.94,
      "learning_rate": 5.341100323624596e-06,
      "loss": 0.0411,
      "step": 283100
    },
    {
      "epoch": 109.95,
      "learning_rate": 5.340582524271846e-06,
      "loss": 0.0816,
      "step": 283110
    },
    {
      "epoch": 109.95,
      "learning_rate": 5.340064724919094e-06,
      "loss": 0.0325,
      "step": 283120
    },
    {
      "epoch": 109.95,
      "learning_rate": 5.339546925566343e-06,
      "loss": 0.0648,
      "step": 283130
    },
    {
      "epoch": 109.96,
      "learning_rate": 5.339029126213593e-06,
      "loss": 0.0428,
      "step": 283140
    },
    {
      "epoch": 109.96,
      "learning_rate": 5.3385113268608426e-06,
      "loss": 0.0683,
      "step": 283150
    },
    {
      "epoch": 109.97,
      "learning_rate": 5.3379935275080905e-06,
      "loss": 0.0418,
      "step": 283160
    },
    {
      "epoch": 109.97,
      "learning_rate": 5.33747572815534e-06,
      "loss": 0.1464,
      "step": 283170
    },
    {
      "epoch": 109.97,
      "learning_rate": 5.33695792880259e-06,
      "loss": 0.0247,
      "step": 283180
    },
    {
      "epoch": 109.98,
      "learning_rate": 5.336440129449839e-06,
      "loss": 0.0344,
      "step": 283190
    },
    {
      "epoch": 109.98,
      "learning_rate": 5.335922330097087e-06,
      "loss": 0.0831,
      "step": 283200
    },
    {
      "epoch": 109.98,
      "learning_rate": 5.335404530744337e-06,
      "loss": 0.0752,
      "step": 283210
    },
    {
      "epoch": 109.99,
      "learning_rate": 5.3348867313915865e-06,
      "loss": 0.0429,
      "step": 283220
    },
    {
      "epoch": 109.99,
      "learning_rate": 5.334368932038836e-06,
      "loss": 0.0018,
      "step": 283230
    },
    {
      "epoch": 110.0,
      "learning_rate": 5.333851132686084e-06,
      "loss": 0.0084,
      "step": 283240
    },
    {
      "epoch": 110.0,
      "learning_rate": 5.333333333333334e-06,
      "loss": 0.1136,
      "step": 283250
    },
    {
      "epoch": 110.0,
      "eval_accuracy": 0.9502063273727648,
      "eval_loss": 0.3846931755542755,
      "eval_runtime": 8.1817,
      "eval_samples_per_second": 444.286,
      "eval_steps_per_second": 55.612,
      "step": 283250
    },
    {
      "epoch": 110.0,
      "learning_rate": 5.332815533980583e-06,
      "loss": 0.0531,
      "step": 283260
    },
    {
      "epoch": 110.01,
      "learning_rate": 5.332297734627833e-06,
      "loss": 0.1532,
      "step": 283270
    },
    {
      "epoch": 110.01,
      "learning_rate": 5.331779935275081e-06,
      "loss": 0.0001,
      "step": 283280
    },
    {
      "epoch": 110.02,
      "learning_rate": 5.3312621359223304e-06,
      "loss": 0.1421,
      "step": 283290
    },
    {
      "epoch": 110.02,
      "learning_rate": 5.33074433656958e-06,
      "loss": 0.0315,
      "step": 283300
    },
    {
      "epoch": 110.02,
      "learning_rate": 5.33022653721683e-06,
      "loss": 0.0232,
      "step": 283310
    },
    {
      "epoch": 110.03,
      "learning_rate": 5.329708737864078e-06,
      "loss": 0.0896,
      "step": 283320
    },
    {
      "epoch": 110.03,
      "learning_rate": 5.329190938511327e-06,
      "loss": 0.0093,
      "step": 283330
    },
    {
      "epoch": 110.03,
      "learning_rate": 5.328673139158577e-06,
      "loss": 0.0633,
      "step": 283340
    },
    {
      "epoch": 110.04,
      "learning_rate": 5.3281553398058264e-06,
      "loss": 0.0128,
      "step": 283350
    },
    {
      "epoch": 110.04,
      "learning_rate": 5.327637540453074e-06,
      "loss": 0.161,
      "step": 283360
    },
    {
      "epoch": 110.05,
      "learning_rate": 5.327119741100324e-06,
      "loss": 0.0283,
      "step": 283370
    },
    {
      "epoch": 110.05,
      "learning_rate": 5.326601941747574e-06,
      "loss": 0.0351,
      "step": 283380
    },
    {
      "epoch": 110.05,
      "learning_rate": 5.326084142394823e-06,
      "loss": 0.0736,
      "step": 283390
    },
    {
      "epoch": 110.06,
      "learning_rate": 5.325566343042071e-06,
      "loss": 0.0346,
      "step": 283400
    },
    {
      "epoch": 110.06,
      "learning_rate": 5.325048543689321e-06,
      "loss": 0.1902,
      "step": 283410
    },
    {
      "epoch": 110.07,
      "learning_rate": 5.32453074433657e-06,
      "loss": 0.0757,
      "step": 283420
    },
    {
      "epoch": 110.07,
      "learning_rate": 5.32401294498382e-06,
      "loss": 0.0017,
      "step": 283430
    },
    {
      "epoch": 110.07,
      "learning_rate": 5.323495145631068e-06,
      "loss": 0.009,
      "step": 283440
    },
    {
      "epoch": 110.08,
      "learning_rate": 5.3229773462783175e-06,
      "loss": 0.0251,
      "step": 283450
    },
    {
      "epoch": 110.08,
      "learning_rate": 5.322459546925567e-06,
      "loss": 0.0045,
      "step": 283460
    },
    {
      "epoch": 110.09,
      "learning_rate": 5.321941747572816e-06,
      "loss": 0.0918,
      "step": 283470
    },
    {
      "epoch": 110.09,
      "learning_rate": 5.321423948220065e-06,
      "loss": 0.0636,
      "step": 283480
    },
    {
      "epoch": 110.09,
      "learning_rate": 5.320906148867314e-06,
      "loss": 0.1418,
      "step": 283490
    },
    {
      "epoch": 110.1,
      "learning_rate": 5.320388349514564e-06,
      "loss": 0.0135,
      "step": 283500
    },
    {
      "epoch": 110.1,
      "learning_rate": 5.319870550161813e-06,
      "loss": 0.1267,
      "step": 283510
    },
    {
      "epoch": 110.1,
      "learning_rate": 5.3193527508090615e-06,
      "loss": 0.0379,
      "step": 283520
    },
    {
      "epoch": 110.11,
      "learning_rate": 5.318834951456311e-06,
      "loss": 0.1027,
      "step": 283530
    },
    {
      "epoch": 110.11,
      "learning_rate": 5.318317152103561e-06,
      "loss": 0.014,
      "step": 283540
    },
    {
      "epoch": 110.12,
      "learning_rate": 5.3177993527508095e-06,
      "loss": 0.1096,
      "step": 283550
    },
    {
      "epoch": 110.12,
      "learning_rate": 5.317281553398058e-06,
      "loss": 0.0306,
      "step": 283560
    },
    {
      "epoch": 110.12,
      "learning_rate": 5.316763754045308e-06,
      "loss": 0.0003,
      "step": 283570
    },
    {
      "epoch": 110.13,
      "learning_rate": 5.3162459546925575e-06,
      "loss": 0.0464,
      "step": 283580
    },
    {
      "epoch": 110.13,
      "learning_rate": 5.315728155339806e-06,
      "loss": 0.0098,
      "step": 283590
    },
    {
      "epoch": 110.14,
      "learning_rate": 5.315210355987055e-06,
      "loss": 0.1423,
      "step": 283600
    },
    {
      "epoch": 110.14,
      "learning_rate": 5.314692556634305e-06,
      "loss": 0.0131,
      "step": 283610
    },
    {
      "epoch": 110.14,
      "learning_rate": 5.3141747572815534e-06,
      "loss": 0.0393,
      "step": 283620
    },
    {
      "epoch": 110.15,
      "learning_rate": 5.313656957928803e-06,
      "loss": 0.0003,
      "step": 283630
    },
    {
      "epoch": 110.15,
      "learning_rate": 5.313139158576052e-06,
      "loss": 0.0926,
      "step": 283640
    },
    {
      "epoch": 110.16,
      "learning_rate": 5.3126213592233014e-06,
      "loss": 0.0331,
      "step": 283650
    },
    {
      "epoch": 110.16,
      "learning_rate": 5.31210355987055e-06,
      "loss": 0.0344,
      "step": 283660
    },
    {
      "epoch": 110.16,
      "learning_rate": 5.3115857605178e-06,
      "loss": 0.0005,
      "step": 283670
    },
    {
      "epoch": 110.17,
      "learning_rate": 5.3110679611650494e-06,
      "loss": 0.0947,
      "step": 283680
    },
    {
      "epoch": 110.17,
      "learning_rate": 5.310550161812298e-06,
      "loss": 0.0344,
      "step": 283690
    },
    {
      "epoch": 110.17,
      "learning_rate": 5.310032362459547e-06,
      "loss": 0.0915,
      "step": 283700
    },
    {
      "epoch": 110.18,
      "learning_rate": 5.309514563106797e-06,
      "loss": 0.03,
      "step": 283710
    },
    {
      "epoch": 110.18,
      "learning_rate": 5.308996763754046e-06,
      "loss": 0.0163,
      "step": 283720
    },
    {
      "epoch": 110.19,
      "learning_rate": 5.308478964401295e-06,
      "loss": 0.0562,
      "step": 283730
    },
    {
      "epoch": 110.19,
      "learning_rate": 5.307961165048544e-06,
      "loss": 0.0087,
      "step": 283740
    },
    {
      "epoch": 110.19,
      "learning_rate": 5.307443365695793e-06,
      "loss": 0.1461,
      "step": 283750
    },
    {
      "epoch": 110.2,
      "learning_rate": 5.306925566343043e-06,
      "loss": 0.0196,
      "step": 283760
    },
    {
      "epoch": 110.2,
      "learning_rate": 5.306407766990292e-06,
      "loss": 0.0954,
      "step": 283770
    },
    {
      "epoch": 110.21,
      "learning_rate": 5.3058899676375405e-06,
      "loss": 0.129,
      "step": 283780
    },
    {
      "epoch": 110.21,
      "learning_rate": 5.30537216828479e-06,
      "loss": 0.2165,
      "step": 283790
    },
    {
      "epoch": 110.21,
      "learning_rate": 5.30485436893204e-06,
      "loss": 0.0196,
      "step": 283800
    },
    {
      "epoch": 110.22,
      "learning_rate": 5.304336569579288e-06,
      "loss": 0.0069,
      "step": 283810
    },
    {
      "epoch": 110.22,
      "learning_rate": 5.303818770226537e-06,
      "loss": 0.0255,
      "step": 283820
    },
    {
      "epoch": 110.23,
      "learning_rate": 5.303300970873787e-06,
      "loss": 0.0164,
      "step": 283830
    },
    {
      "epoch": 110.23,
      "learning_rate": 5.3027831715210366e-06,
      "loss": 0.0595,
      "step": 283840
    },
    {
      "epoch": 110.23,
      "learning_rate": 5.3022653721682845e-06,
      "loss": 0.018,
      "step": 283850
    },
    {
      "epoch": 110.24,
      "learning_rate": 5.301747572815534e-06,
      "loss": 0.0403,
      "step": 283860
    },
    {
      "epoch": 110.24,
      "learning_rate": 5.301229773462784e-06,
      "loss": 0.018,
      "step": 283870
    },
    {
      "epoch": 110.24,
      "learning_rate": 5.300711974110033e-06,
      "loss": 0.0187,
      "step": 283880
    },
    {
      "epoch": 110.25,
      "learning_rate": 5.300194174757281e-06,
      "loss": 0.0565,
      "step": 283890
    },
    {
      "epoch": 110.25,
      "learning_rate": 5.299676375404531e-06,
      "loss": 0.092,
      "step": 283900
    },
    {
      "epoch": 110.26,
      "learning_rate": 5.2991585760517805e-06,
      "loss": 0.0025,
      "step": 283910
    },
    {
      "epoch": 110.26,
      "learning_rate": 5.29864077669903e-06,
      "loss": 0.0131,
      "step": 283920
    },
    {
      "epoch": 110.26,
      "learning_rate": 5.298122977346278e-06,
      "loss": 0.0405,
      "step": 283930
    },
    {
      "epoch": 110.27,
      "learning_rate": 5.297605177993528e-06,
      "loss": 0.0371,
      "step": 283940
    },
    {
      "epoch": 110.27,
      "learning_rate": 5.297087378640777e-06,
      "loss": 0.0301,
      "step": 283950
    },
    {
      "epoch": 110.28,
      "learning_rate": 5.296569579288027e-06,
      "loss": 0.0178,
      "step": 283960
    },
    {
      "epoch": 110.28,
      "learning_rate": 5.296051779935275e-06,
      "loss": 0.0001,
      "step": 283970
    },
    {
      "epoch": 110.28,
      "learning_rate": 5.2955339805825244e-06,
      "loss": 0.074,
      "step": 283980
    },
    {
      "epoch": 110.29,
      "learning_rate": 5.295016181229774e-06,
      "loss": 0.0035,
      "step": 283990
    },
    {
      "epoch": 110.29,
      "learning_rate": 5.294498381877024e-06,
      "loss": 0.0173,
      "step": 284000
    },
    {
      "epoch": 110.3,
      "learning_rate": 5.293980582524272e-06,
      "loss": 0.0005,
      "step": 284010
    },
    {
      "epoch": 110.3,
      "learning_rate": 5.293462783171521e-06,
      "loss": 0.0749,
      "step": 284020
    },
    {
      "epoch": 110.3,
      "learning_rate": 5.292944983818771e-06,
      "loss": 0.0429,
      "step": 284030
    },
    {
      "epoch": 110.31,
      "learning_rate": 5.2924271844660204e-06,
      "loss": 0.1101,
      "step": 284040
    },
    {
      "epoch": 110.31,
      "learning_rate": 5.291909385113268e-06,
      "loss": 0.0165,
      "step": 284050
    },
    {
      "epoch": 110.31,
      "learning_rate": 5.291391585760518e-06,
      "loss": 0.0642,
      "step": 284060
    },
    {
      "epoch": 110.32,
      "learning_rate": 5.290873786407768e-06,
      "loss": 0.013,
      "step": 284070
    },
    {
      "epoch": 110.32,
      "learning_rate": 5.290355987055017e-06,
      "loss": 0.001,
      "step": 284080
    },
    {
      "epoch": 110.33,
      "learning_rate": 5.289838187702265e-06,
      "loss": 0.0712,
      "step": 284090
    },
    {
      "epoch": 110.33,
      "learning_rate": 5.289320388349515e-06,
      "loss": 0.0017,
      "step": 284100
    },
    {
      "epoch": 110.33,
      "learning_rate": 5.288802588996764e-06,
      "loss": 0.002,
      "step": 284110
    },
    {
      "epoch": 110.34,
      "learning_rate": 5.288284789644014e-06,
      "loss": 0.0944,
      "step": 284120
    },
    {
      "epoch": 110.34,
      "learning_rate": 5.287766990291262e-06,
      "loss": 0.0017,
      "step": 284130
    },
    {
      "epoch": 110.35,
      "learning_rate": 5.2872491909385115e-06,
      "loss": 0.0061,
      "step": 284140
    },
    {
      "epoch": 110.35,
      "learning_rate": 5.286731391585761e-06,
      "loss": 0.0008,
      "step": 284150
    },
    {
      "epoch": 110.35,
      "learning_rate": 5.286213592233011e-06,
      "loss": 0.0094,
      "step": 284160
    },
    {
      "epoch": 110.36,
      "learning_rate": 5.285695792880259e-06,
      "loss": 0.0458,
      "step": 284170
    },
    {
      "epoch": 110.36,
      "learning_rate": 5.285177993527508e-06,
      "loss": 0.0783,
      "step": 284180
    },
    {
      "epoch": 110.37,
      "learning_rate": 5.284660194174758e-06,
      "loss": 0.0065,
      "step": 284190
    },
    {
      "epoch": 110.37,
      "learning_rate": 5.2841423948220076e-06,
      "loss": 0.0143,
      "step": 284200
    },
    {
      "epoch": 110.37,
      "learning_rate": 5.2836245954692555e-06,
      "loss": 0.0281,
      "step": 284210
    },
    {
      "epoch": 110.38,
      "learning_rate": 5.283106796116505e-06,
      "loss": 0.0083,
      "step": 284220
    },
    {
      "epoch": 110.38,
      "learning_rate": 5.282588996763755e-06,
      "loss": 0.0987,
      "step": 284230
    },
    {
      "epoch": 110.38,
      "learning_rate": 5.282071197411004e-06,
      "loss": 0.0004,
      "step": 284240
    },
    {
      "epoch": 110.39,
      "learning_rate": 5.281553398058253e-06,
      "loss": 0.0622,
      "step": 284250
    },
    {
      "epoch": 110.39,
      "learning_rate": 5.281035598705502e-06,
      "loss": 0.0449,
      "step": 284260
    },
    {
      "epoch": 110.4,
      "learning_rate": 5.2805177993527515e-06,
      "loss": 0.0527,
      "step": 284270
    },
    {
      "epoch": 110.4,
      "learning_rate": 5.28e-06,
      "loss": 0.0001,
      "step": 284280
    },
    {
      "epoch": 110.4,
      "learning_rate": 5.27948220064725e-06,
      "loss": 0.0008,
      "step": 284290
    },
    {
      "epoch": 110.41,
      "learning_rate": 5.278964401294499e-06,
      "loss": 0.0385,
      "step": 284300
    },
    {
      "epoch": 110.41,
      "learning_rate": 5.278446601941748e-06,
      "loss": 0.09,
      "step": 284310
    },
    {
      "epoch": 110.42,
      "learning_rate": 5.277928802588997e-06,
      "loss": 0.027,
      "step": 284320
    },
    {
      "epoch": 110.42,
      "learning_rate": 5.277411003236247e-06,
      "loss": 0.0053,
      "step": 284330
    },
    {
      "epoch": 110.42,
      "learning_rate": 5.2768932038834954e-06,
      "loss": 0.0469,
      "step": 284340
    },
    {
      "epoch": 110.43,
      "learning_rate": 5.276375404530745e-06,
      "loss": 0.0259,
      "step": 284350
    },
    {
      "epoch": 110.43,
      "learning_rate": 5.275857605177994e-06,
      "loss": 0.1013,
      "step": 284360
    },
    {
      "epoch": 110.43,
      "learning_rate": 5.2753398058252434e-06,
      "loss": 0.1427,
      "step": 284370
    },
    {
      "epoch": 110.44,
      "learning_rate": 5.274822006472492e-06,
      "loss": 0.042,
      "step": 284380
    },
    {
      "epoch": 110.44,
      "learning_rate": 5.274304207119742e-06,
      "loss": 0.034,
      "step": 284390
    },
    {
      "epoch": 110.45,
      "learning_rate": 5.273786407766991e-06,
      "loss": 0.0459,
      "step": 284400
    },
    {
      "epoch": 110.45,
      "learning_rate": 5.27326860841424e-06,
      "loss": 0.0769,
      "step": 284410
    },
    {
      "epoch": 110.45,
      "learning_rate": 5.272750809061489e-06,
      "loss": 0.0006,
      "step": 284420
    },
    {
      "epoch": 110.46,
      "learning_rate": 5.272233009708739e-06,
      "loss": 0.0975,
      "step": 284430
    },
    {
      "epoch": 110.46,
      "learning_rate": 5.271715210355987e-06,
      "loss": 0.1041,
      "step": 284440
    },
    {
      "epoch": 110.47,
      "learning_rate": 5.271197411003237e-06,
      "loss": 0.0036,
      "step": 284450
    },
    {
      "epoch": 110.47,
      "learning_rate": 5.270679611650486e-06,
      "loss": 0.0015,
      "step": 284460
    },
    {
      "epoch": 110.47,
      "learning_rate": 5.2701618122977345e-06,
      "loss": 0.081,
      "step": 284470
    },
    {
      "epoch": 110.48,
      "learning_rate": 5.269644012944984e-06,
      "loss": 0.0456,
      "step": 284480
    },
    {
      "epoch": 110.48,
      "learning_rate": 5.269126213592234e-06,
      "loss": 0.0397,
      "step": 284490
    },
    {
      "epoch": 110.49,
      "learning_rate": 5.2686084142394825e-06,
      "loss": 0.0723,
      "step": 284500
    },
    {
      "epoch": 110.49,
      "learning_rate": 5.268090614886731e-06,
      "loss": 0.1289,
      "step": 284510
    },
    {
      "epoch": 110.49,
      "learning_rate": 5.267572815533981e-06,
      "loss": 0.0001,
      "step": 284520
    },
    {
      "epoch": 110.5,
      "learning_rate": 5.2670550161812306e-06,
      "loss": 0.0003,
      "step": 284530
    },
    {
      "epoch": 110.5,
      "learning_rate": 5.266537216828479e-06,
      "loss": 0.0713,
      "step": 284540
    },
    {
      "epoch": 110.5,
      "learning_rate": 5.266019417475728e-06,
      "loss": 0.001,
      "step": 284550
    },
    {
      "epoch": 110.51,
      "learning_rate": 5.265501618122978e-06,
      "loss": 0.0078,
      "step": 284560
    },
    {
      "epoch": 110.51,
      "learning_rate": 5.264983818770227e-06,
      "loss": 0.037,
      "step": 284570
    },
    {
      "epoch": 110.52,
      "learning_rate": 5.264466019417476e-06,
      "loss": 0.0599,
      "step": 284580
    },
    {
      "epoch": 110.52,
      "learning_rate": 5.263948220064725e-06,
      "loss": 0.1222,
      "step": 284590
    },
    {
      "epoch": 110.52,
      "learning_rate": 5.2634304207119745e-06,
      "loss": 0.0187,
      "step": 284600
    },
    {
      "epoch": 110.53,
      "learning_rate": 5.262912621359224e-06,
      "loss": 0.0423,
      "step": 284610
    },
    {
      "epoch": 110.53,
      "learning_rate": 5.262394822006472e-06,
      "loss": 0.0607,
      "step": 284620
    },
    {
      "epoch": 110.54,
      "learning_rate": 5.261877022653722e-06,
      "loss": 0.0028,
      "step": 284630
    },
    {
      "epoch": 110.54,
      "learning_rate": 5.261359223300971e-06,
      "loss": 0.059,
      "step": 284640
    },
    {
      "epoch": 110.54,
      "learning_rate": 5.260841423948221e-06,
      "loss": 0.0454,
      "step": 284650
    },
    {
      "epoch": 110.55,
      "learning_rate": 5.260323624595469e-06,
      "loss": 0.1094,
      "step": 284660
    },
    {
      "epoch": 110.55,
      "learning_rate": 5.2598058252427184e-06,
      "loss": 0.021,
      "step": 284670
    },
    {
      "epoch": 110.56,
      "learning_rate": 5.259288025889968e-06,
      "loss": 0.0366,
      "step": 284680
    },
    {
      "epoch": 110.56,
      "learning_rate": 5.258770226537218e-06,
      "loss": 0.1077,
      "step": 284690
    },
    {
      "epoch": 110.56,
      "learning_rate": 5.258252427184466e-06,
      "loss": 0.0973,
      "step": 284700
    },
    {
      "epoch": 110.57,
      "learning_rate": 5.257734627831715e-06,
      "loss": 0.0003,
      "step": 284710
    },
    {
      "epoch": 110.57,
      "learning_rate": 5.257216828478965e-06,
      "loss": 0.0431,
      "step": 284720
    },
    {
      "epoch": 110.57,
      "learning_rate": 5.2566990291262144e-06,
      "loss": 0.0975,
      "step": 284730
    },
    {
      "epoch": 110.58,
      "learning_rate": 5.256181229773462e-06,
      "loss": 0.0856,
      "step": 284740
    },
    {
      "epoch": 110.58,
      "learning_rate": 5.255663430420712e-06,
      "loss": 0.0964,
      "step": 284750
    },
    {
      "epoch": 110.59,
      "learning_rate": 5.255145631067962e-06,
      "loss": 0.0029,
      "step": 284760
    },
    {
      "epoch": 110.59,
      "learning_rate": 5.254627831715211e-06,
      "loss": 0.0445,
      "step": 284770
    },
    {
      "epoch": 110.59,
      "learning_rate": 5.254110032362459e-06,
      "loss": 0.0279,
      "step": 284780
    },
    {
      "epoch": 110.6,
      "learning_rate": 5.253592233009709e-06,
      "loss": 0.0669,
      "step": 284790
    },
    {
      "epoch": 110.6,
      "learning_rate": 5.253074433656958e-06,
      "loss": 0.0193,
      "step": 284800
    },
    {
      "epoch": 110.61,
      "learning_rate": 5.252556634304208e-06,
      "loss": 0.0876,
      "step": 284810
    },
    {
      "epoch": 110.61,
      "learning_rate": 5.252038834951458e-06,
      "loss": 0.0832,
      "step": 284820
    },
    {
      "epoch": 110.61,
      "learning_rate": 5.2515210355987055e-06,
      "loss": 0.1484,
      "step": 284830
    },
    {
      "epoch": 110.62,
      "learning_rate": 5.251003236245955e-06,
      "loss": 0.0008,
      "step": 284840
    },
    {
      "epoch": 110.62,
      "learning_rate": 5.250485436893205e-06,
      "loss": 0.0623,
      "step": 284850
    },
    {
      "epoch": 110.63,
      "learning_rate": 5.249967637540454e-06,
      "loss": 0.0441,
      "step": 284860
    },
    {
      "epoch": 110.63,
      "learning_rate": 5.249449838187702e-06,
      "loss": 0.0361,
      "step": 284870
    },
    {
      "epoch": 110.63,
      "learning_rate": 5.248932038834952e-06,
      "loss": 0.0148,
      "step": 284880
    },
    {
      "epoch": 110.64,
      "learning_rate": 5.2484142394822016e-06,
      "loss": 0.0557,
      "step": 284890
    },
    {
      "epoch": 110.64,
      "learning_rate": 5.247896440129451e-06,
      "loss": 0.1154,
      "step": 284900
    },
    {
      "epoch": 110.64,
      "learning_rate": 5.247378640776699e-06,
      "loss": 0.067,
      "step": 284910
    },
    {
      "epoch": 110.65,
      "learning_rate": 5.246860841423949e-06,
      "loss": 0.0976,
      "step": 284920
    },
    {
      "epoch": 110.65,
      "learning_rate": 5.246343042071198e-06,
      "loss": 0.1133,
      "step": 284930
    },
    {
      "epoch": 110.66,
      "learning_rate": 5.245825242718447e-06,
      "loss": 0.1607,
      "step": 284940
    },
    {
      "epoch": 110.66,
      "learning_rate": 5.245307443365696e-06,
      "loss": 0.1014,
      "step": 284950
    },
    {
      "epoch": 110.66,
      "learning_rate": 5.2447896440129455e-06,
      "loss": 0.0259,
      "step": 284960
    },
    {
      "epoch": 110.67,
      "learning_rate": 5.244271844660195e-06,
      "loss": 0.0323,
      "step": 284970
    },
    {
      "epoch": 110.67,
      "learning_rate": 5.243754045307444e-06,
      "loss": 0.0344,
      "step": 284980
    },
    {
      "epoch": 110.68,
      "learning_rate": 5.243236245954693e-06,
      "loss": 0.0017,
      "step": 284990
    },
    {
      "epoch": 110.68,
      "learning_rate": 5.242718446601942e-06,
      "loss": 0.0002,
      "step": 285000
    },
    {
      "epoch": 110.68,
      "learning_rate": 5.242200647249192e-06,
      "loss": 0.0704,
      "step": 285010
    },
    {
      "epoch": 110.69,
      "learning_rate": 5.241682847896441e-06,
      "loss": 0.0757,
      "step": 285020
    },
    {
      "epoch": 110.69,
      "learning_rate": 5.2411650485436894e-06,
      "loss": 0.0769,
      "step": 285030
    },
    {
      "epoch": 110.7,
      "learning_rate": 5.240647249190939e-06,
      "loss": 0.0098,
      "step": 285040
    },
    {
      "epoch": 110.7,
      "learning_rate": 5.240129449838189e-06,
      "loss": 0.0786,
      "step": 285050
    },
    {
      "epoch": 110.7,
      "learning_rate": 5.2396116504854374e-06,
      "loss": 0.3239,
      "step": 285060
    },
    {
      "epoch": 110.71,
      "learning_rate": 5.239093851132686e-06,
      "loss": 0.1818,
      "step": 285070
    },
    {
      "epoch": 110.71,
      "learning_rate": 5.238576051779936e-06,
      "loss": 0.0144,
      "step": 285080
    },
    {
      "epoch": 110.71,
      "learning_rate": 5.238058252427185e-06,
      "loss": 0.039,
      "step": 285090
    },
    {
      "epoch": 110.72,
      "learning_rate": 5.237540453074434e-06,
      "loss": 0.0244,
      "step": 285100
    },
    {
      "epoch": 110.72,
      "learning_rate": 5.237022653721683e-06,
      "loss": 0.0797,
      "step": 285110
    },
    {
      "epoch": 110.73,
      "learning_rate": 5.236504854368933e-06,
      "loss": 0.0837,
      "step": 285120
    },
    {
      "epoch": 110.73,
      "learning_rate": 5.235987055016181e-06,
      "loss": 0.0571,
      "step": 285130
    },
    {
      "epoch": 110.73,
      "learning_rate": 5.235469255663431e-06,
      "loss": 0.0517,
      "step": 285140
    },
    {
      "epoch": 110.74,
      "learning_rate": 5.23495145631068e-06,
      "loss": 0.0431,
      "step": 285150
    },
    {
      "epoch": 110.74,
      "learning_rate": 5.234433656957929e-06,
      "loss": 0.0501,
      "step": 285160
    },
    {
      "epoch": 110.75,
      "learning_rate": 5.233915857605178e-06,
      "loss": 0.0115,
      "step": 285170
    },
    {
      "epoch": 110.75,
      "learning_rate": 5.233398058252428e-06,
      "loss": 0.0012,
      "step": 285180
    },
    {
      "epoch": 110.75,
      "learning_rate": 5.2328802588996765e-06,
      "loss": 0.1219,
      "step": 285190
    },
    {
      "epoch": 110.76,
      "learning_rate": 5.232362459546926e-06,
      "loss": 0.0886,
      "step": 285200
    },
    {
      "epoch": 110.76,
      "learning_rate": 5.231844660194175e-06,
      "loss": 0.0206,
      "step": 285210
    },
    {
      "epoch": 110.77,
      "learning_rate": 5.2313268608414246e-06,
      "loss": 0.0433,
      "step": 285220
    },
    {
      "epoch": 110.77,
      "learning_rate": 5.230809061488673e-06,
      "loss": 0.0577,
      "step": 285230
    },
    {
      "epoch": 110.77,
      "learning_rate": 5.230291262135923e-06,
      "loss": 0.0042,
      "step": 285240
    },
    {
      "epoch": 110.78,
      "learning_rate": 5.229773462783172e-06,
      "loss": 0.2058,
      "step": 285250
    },
    {
      "epoch": 110.78,
      "learning_rate": 5.229255663430421e-06,
      "loss": 0.0187,
      "step": 285260
    },
    {
      "epoch": 110.78,
      "learning_rate": 5.22873786407767e-06,
      "loss": 0.066,
      "step": 285270
    },
    {
      "epoch": 110.79,
      "learning_rate": 5.228220064724919e-06,
      "loss": 0.0576,
      "step": 285280
    },
    {
      "epoch": 110.79,
      "learning_rate": 5.2277022653721685e-06,
      "loss": 0.1293,
      "step": 285290
    },
    {
      "epoch": 110.8,
      "learning_rate": 5.227184466019418e-06,
      "loss": 0.0888,
      "step": 285300
    },
    {
      "epoch": 110.8,
      "learning_rate": 5.226666666666667e-06,
      "loss": 0.0129,
      "step": 285310
    },
    {
      "epoch": 110.8,
      "learning_rate": 5.226148867313916e-06,
      "loss": 0.0002,
      "step": 285320
    },
    {
      "epoch": 110.81,
      "learning_rate": 5.225631067961165e-06,
      "loss": 0.007,
      "step": 285330
    },
    {
      "epoch": 110.81,
      "learning_rate": 5.225113268608415e-06,
      "loss": 0.0176,
      "step": 285340
    },
    {
      "epoch": 110.82,
      "learning_rate": 5.2245954692556645e-06,
      "loss": 0.0692,
      "step": 285350
    },
    {
      "epoch": 110.82,
      "learning_rate": 5.2240776699029124e-06,
      "loss": 0.0456,
      "step": 285360
    },
    {
      "epoch": 110.82,
      "learning_rate": 5.223559870550162e-06,
      "loss": 0.0436,
      "step": 285370
    },
    {
      "epoch": 110.83,
      "learning_rate": 5.223042071197412e-06,
      "loss": 0.1016,
      "step": 285380
    },
    {
      "epoch": 110.83,
      "learning_rate": 5.222524271844661e-06,
      "loss": 0.1118,
      "step": 285390
    },
    {
      "epoch": 110.83,
      "learning_rate": 5.222006472491909e-06,
      "loss": 0.0857,
      "step": 285400
    },
    {
      "epoch": 110.84,
      "learning_rate": 5.221488673139159e-06,
      "loss": 0.066,
      "step": 285410
    },
    {
      "epoch": 110.84,
      "learning_rate": 5.2209708737864084e-06,
      "loss": 0.0934,
      "step": 285420
    },
    {
      "epoch": 110.85,
      "learning_rate": 5.220453074433658e-06,
      "loss": 0.0498,
      "step": 285430
    },
    {
      "epoch": 110.85,
      "learning_rate": 5.219935275080906e-06,
      "loss": 0.0283,
      "step": 285440
    },
    {
      "epoch": 110.85,
      "learning_rate": 5.219417475728156e-06,
      "loss": 0.0634,
      "step": 285450
    },
    {
      "epoch": 110.86,
      "learning_rate": 5.218899676375405e-06,
      "loss": 0.0293,
      "step": 285460
    },
    {
      "epoch": 110.86,
      "learning_rate": 5.218381877022655e-06,
      "loss": 0.1879,
      "step": 285470
    },
    {
      "epoch": 110.87,
      "learning_rate": 5.217864077669903e-06,
      "loss": 0.0196,
      "step": 285480
    },
    {
      "epoch": 110.87,
      "learning_rate": 5.217346278317152e-06,
      "loss": 0.0089,
      "step": 285490
    },
    {
      "epoch": 110.87,
      "learning_rate": 5.216828478964402e-06,
      "loss": 0.0175,
      "step": 285500
    },
    {
      "epoch": 110.88,
      "learning_rate": 5.216310679611652e-06,
      "loss": 0.0282,
      "step": 285510
    },
    {
      "epoch": 110.88,
      "learning_rate": 5.2157928802588995e-06,
      "loss": 0.0507,
      "step": 285520
    },
    {
      "epoch": 110.89,
      "learning_rate": 5.215275080906149e-06,
      "loss": 0.0017,
      "step": 285530
    },
    {
      "epoch": 110.89,
      "learning_rate": 5.214757281553399e-06,
      "loss": 0.0379,
      "step": 285540
    },
    {
      "epoch": 110.89,
      "learning_rate": 5.214239482200648e-06,
      "loss": 0.0471,
      "step": 285550
    },
    {
      "epoch": 110.9,
      "learning_rate": 5.213721682847896e-06,
      "loss": 0.07,
      "step": 285560
    },
    {
      "epoch": 110.9,
      "learning_rate": 5.213203883495146e-06,
      "loss": 0.0257,
      "step": 285570
    },
    {
      "epoch": 110.9,
      "learning_rate": 5.2126860841423956e-06,
      "loss": 0.0388,
      "step": 285580
    },
    {
      "epoch": 110.91,
      "learning_rate": 5.212168284789645e-06,
      "loss": 0.0478,
      "step": 285590
    },
    {
      "epoch": 110.91,
      "learning_rate": 5.211650485436893e-06,
      "loss": 0.0438,
      "step": 285600
    },
    {
      "epoch": 110.92,
      "learning_rate": 5.211132686084143e-06,
      "loss": 0.004,
      "step": 285610
    },
    {
      "epoch": 110.92,
      "learning_rate": 5.210614886731392e-06,
      "loss": 0.0035,
      "step": 285620
    },
    {
      "epoch": 110.92,
      "learning_rate": 5.210097087378642e-06,
      "loss": 0.1101,
      "step": 285630
    },
    {
      "epoch": 110.93,
      "learning_rate": 5.20957928802589e-06,
      "loss": 0.0938,
      "step": 285640
    },
    {
      "epoch": 110.93,
      "learning_rate": 5.2090614886731395e-06,
      "loss": 0.0484,
      "step": 285650
    },
    {
      "epoch": 110.94,
      "learning_rate": 5.208543689320389e-06,
      "loss": 0.0389,
      "step": 285660
    },
    {
      "epoch": 110.94,
      "learning_rate": 5.208025889967639e-06,
      "loss": 0.0224,
      "step": 285670
    },
    {
      "epoch": 110.94,
      "learning_rate": 5.207508090614887e-06,
      "loss": 0.0525,
      "step": 285680
    },
    {
      "epoch": 110.95,
      "learning_rate": 5.206990291262136e-06,
      "loss": 0.0159,
      "step": 285690
    },
    {
      "epoch": 110.95,
      "learning_rate": 5.206472491909386e-06,
      "loss": 0.1414,
      "step": 285700
    },
    {
      "epoch": 110.96,
      "learning_rate": 5.2059546925566355e-06,
      "loss": 0.0184,
      "step": 285710
    },
    {
      "epoch": 110.96,
      "learning_rate": 5.2054368932038834e-06,
      "loss": 0.0124,
      "step": 285720
    },
    {
      "epoch": 110.96,
      "learning_rate": 5.204919093851133e-06,
      "loss": 0.1025,
      "step": 285730
    },
    {
      "epoch": 110.97,
      "learning_rate": 5.204401294498383e-06,
      "loss": 0.0951,
      "step": 285740
    },
    {
      "epoch": 110.97,
      "learning_rate": 5.2038834951456314e-06,
      "loss": 0.0387,
      "step": 285750
    },
    {
      "epoch": 110.97,
      "learning_rate": 5.20336569579288e-06,
      "loss": 0.0006,
      "step": 285760
    },
    {
      "epoch": 110.98,
      "learning_rate": 5.20284789644013e-06,
      "loss": 0.0101,
      "step": 285770
    },
    {
      "epoch": 110.98,
      "learning_rate": 5.2023300970873794e-06,
      "loss": 0.0035,
      "step": 285780
    },
    {
      "epoch": 110.99,
      "learning_rate": 5.201812297734628e-06,
      "loss": 0.1124,
      "step": 285790
    },
    {
      "epoch": 110.99,
      "learning_rate": 5.201294498381877e-06,
      "loss": 0.0001,
      "step": 285800
    },
    {
      "epoch": 110.99,
      "learning_rate": 5.200776699029127e-06,
      "loss": 0.01,
      "step": 285810
    },
    {
      "epoch": 111.0,
      "learning_rate": 5.200258899676376e-06,
      "loss": 0.0751,
      "step": 285820
    },
    {
      "epoch": 111.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.38175803422927856,
      "eval_runtime": 8.1715,
      "eval_samples_per_second": 444.84,
      "eval_steps_per_second": 55.682,
      "step": 285825
    },
    {
      "epoch": 111.0,
      "learning_rate": 5.199741100323625e-06,
      "loss": 0.0085,
      "step": 285830
    },
    {
      "epoch": 111.01,
      "learning_rate": 5.199223300970874e-06,
      "loss": 0.1059,
      "step": 285840
    },
    {
      "epoch": 111.01,
      "learning_rate": 5.198705501618123e-06,
      "loss": 0.0087,
      "step": 285850
    },
    {
      "epoch": 111.01,
      "learning_rate": 5.198187702265373e-06,
      "loss": 0.0117,
      "step": 285860
    },
    {
      "epoch": 111.02,
      "learning_rate": 5.197669902912622e-06,
      "loss": 0.0482,
      "step": 285870
    },
    {
      "epoch": 111.02,
      "learning_rate": 5.1971521035598705e-06,
      "loss": 0.017,
      "step": 285880
    },
    {
      "epoch": 111.03,
      "learning_rate": 5.19663430420712e-06,
      "loss": 0.0399,
      "step": 285890
    },
    {
      "epoch": 111.03,
      "learning_rate": 5.196116504854369e-06,
      "loss": 0.0227,
      "step": 285900
    },
    {
      "epoch": 111.03,
      "learning_rate": 5.1955987055016186e-06,
      "loss": 0.0179,
      "step": 285910
    },
    {
      "epoch": 111.04,
      "learning_rate": 5.195080906148868e-06,
      "loss": 0.0356,
      "step": 285920
    },
    {
      "epoch": 111.04,
      "learning_rate": 5.194563106796117e-06,
      "loss": 0.0113,
      "step": 285930
    },
    {
      "epoch": 111.04,
      "learning_rate": 5.194045307443366e-06,
      "loss": 0.0112,
      "step": 285940
    },
    {
      "epoch": 111.05,
      "learning_rate": 5.193527508090615e-06,
      "loss": 0.1683,
      "step": 285950
    },
    {
      "epoch": 111.05,
      "learning_rate": 5.193009708737865e-06,
      "loss": 0.0212,
      "step": 285960
    },
    {
      "epoch": 111.06,
      "learning_rate": 5.192491909385114e-06,
      "loss": 0.0321,
      "step": 285970
    },
    {
      "epoch": 111.06,
      "learning_rate": 5.1919741100323625e-06,
      "loss": 0.0042,
      "step": 285980
    },
    {
      "epoch": 111.06,
      "learning_rate": 5.191456310679612e-06,
      "loss": 0.0076,
      "step": 285990
    },
    {
      "epoch": 111.07,
      "learning_rate": 5.190938511326862e-06,
      "loss": 0.0819,
      "step": 286000
    },
    {
      "epoch": 111.07,
      "learning_rate": 5.1904207119741105e-06,
      "loss": 0.0277,
      "step": 286010
    },
    {
      "epoch": 111.08,
      "learning_rate": 5.189902912621359e-06,
      "loss": 0.0044,
      "step": 286020
    },
    {
      "epoch": 111.08,
      "learning_rate": 5.189385113268609e-06,
      "loss": 0.0483,
      "step": 286030
    },
    {
      "epoch": 111.08,
      "learning_rate": 5.1888673139158585e-06,
      "loss": 0.0143,
      "step": 286040
    },
    {
      "epoch": 111.09,
      "learning_rate": 5.188349514563107e-06,
      "loss": 0.0428,
      "step": 286050
    },
    {
      "epoch": 111.09,
      "learning_rate": 5.187831715210356e-06,
      "loss": 0.0276,
      "step": 286060
    },
    {
      "epoch": 111.1,
      "learning_rate": 5.187313915857606e-06,
      "loss": 0.1004,
      "step": 286070
    },
    {
      "epoch": 111.1,
      "learning_rate": 5.186796116504855e-06,
      "loss": 0.0435,
      "step": 286080
    },
    {
      "epoch": 111.1,
      "learning_rate": 5.186278317152103e-06,
      "loss": 0.132,
      "step": 286090
    },
    {
      "epoch": 111.11,
      "learning_rate": 5.185760517799353e-06,
      "loss": 0.058,
      "step": 286100
    },
    {
      "epoch": 111.11,
      "learning_rate": 5.1852427184466024e-06,
      "loss": 0.0674,
      "step": 286110
    },
    {
      "epoch": 111.11,
      "learning_rate": 5.184724919093852e-06,
      "loss": 0.0171,
      "step": 286120
    },
    {
      "epoch": 111.12,
      "learning_rate": 5.1842071197411e-06,
      "loss": 0.0113,
      "step": 286130
    },
    {
      "epoch": 111.12,
      "learning_rate": 5.18368932038835e-06,
      "loss": 0.061,
      "step": 286140
    },
    {
      "epoch": 111.13,
      "learning_rate": 5.183171521035599e-06,
      "loss": 0.0561,
      "step": 286150
    },
    {
      "epoch": 111.13,
      "learning_rate": 5.182653721682849e-06,
      "loss": 0.0093,
      "step": 286160
    },
    {
      "epoch": 111.13,
      "learning_rate": 5.182135922330097e-06,
      "loss": 0.0827,
      "step": 286170
    },
    {
      "epoch": 111.14,
      "learning_rate": 5.181618122977346e-06,
      "loss": 0.0454,
      "step": 286180
    },
    {
      "epoch": 111.14,
      "learning_rate": 5.181100323624596e-06,
      "loss": 0.0156,
      "step": 286190
    },
    {
      "epoch": 111.15,
      "learning_rate": 5.180582524271846e-06,
      "loss": 0.0001,
      "step": 286200
    },
    {
      "epoch": 111.15,
      "learning_rate": 5.1800647249190935e-06,
      "loss": 0.1181,
      "step": 286210
    },
    {
      "epoch": 111.15,
      "learning_rate": 5.179546925566343e-06,
      "loss": 0.0779,
      "step": 286220
    },
    {
      "epoch": 111.16,
      "learning_rate": 5.179029126213593e-06,
      "loss": 0.1052,
      "step": 286230
    },
    {
      "epoch": 111.16,
      "learning_rate": 5.178511326860842e-06,
      "loss": 0.0289,
      "step": 286240
    },
    {
      "epoch": 111.17,
      "learning_rate": 5.17799352750809e-06,
      "loss": 0.0261,
      "step": 286250
    },
    {
      "epoch": 111.17,
      "learning_rate": 5.17747572815534e-06,
      "loss": 0.0188,
      "step": 286260
    },
    {
      "epoch": 111.17,
      "learning_rate": 5.1769579288025896e-06,
      "loss": 0.1851,
      "step": 286270
    },
    {
      "epoch": 111.18,
      "learning_rate": 5.176440129449839e-06,
      "loss": 0.004,
      "step": 286280
    },
    {
      "epoch": 111.18,
      "learning_rate": 5.175922330097087e-06,
      "loss": 0.0017,
      "step": 286290
    },
    {
      "epoch": 111.18,
      "learning_rate": 5.175404530744337e-06,
      "loss": 0.0029,
      "step": 286300
    },
    {
      "epoch": 111.19,
      "learning_rate": 5.174886731391586e-06,
      "loss": 0.0946,
      "step": 286310
    },
    {
      "epoch": 111.19,
      "learning_rate": 5.174368932038836e-06,
      "loss": 0.0679,
      "step": 286320
    },
    {
      "epoch": 111.2,
      "learning_rate": 5.173851132686084e-06,
      "loss": 0.0494,
      "step": 286330
    },
    {
      "epoch": 111.2,
      "learning_rate": 5.1733333333333335e-06,
      "loss": 0.0644,
      "step": 286340
    },
    {
      "epoch": 111.2,
      "learning_rate": 5.172815533980583e-06,
      "loss": 0.0266,
      "step": 286350
    },
    {
      "epoch": 111.21,
      "learning_rate": 5.172297734627833e-06,
      "loss": 0.069,
      "step": 286360
    },
    {
      "epoch": 111.21,
      "learning_rate": 5.171779935275081e-06,
      "loss": 0.1173,
      "step": 286370
    },
    {
      "epoch": 111.22,
      "learning_rate": 5.17126213592233e-06,
      "loss": 0.0001,
      "step": 286380
    },
    {
      "epoch": 111.22,
      "learning_rate": 5.17074433656958e-06,
      "loss": 0.0015,
      "step": 286390
    },
    {
      "epoch": 111.22,
      "learning_rate": 5.1702265372168295e-06,
      "loss": 0.0235,
      "step": 286400
    },
    {
      "epoch": 111.23,
      "learning_rate": 5.1697087378640774e-06,
      "loss": 0.0007,
      "step": 286410
    },
    {
      "epoch": 111.23,
      "learning_rate": 5.169190938511327e-06,
      "loss": 0.0682,
      "step": 286420
    },
    {
      "epoch": 111.23,
      "learning_rate": 5.168673139158577e-06,
      "loss": 0.0396,
      "step": 286430
    },
    {
      "epoch": 111.24,
      "learning_rate": 5.168155339805826e-06,
      "loss": 0.0187,
      "step": 286440
    },
    {
      "epoch": 111.24,
      "learning_rate": 5.167637540453074e-06,
      "loss": 0.0651,
      "step": 286450
    },
    {
      "epoch": 111.25,
      "learning_rate": 5.167119741100324e-06,
      "loss": 0.0253,
      "step": 286460
    },
    {
      "epoch": 111.25,
      "learning_rate": 5.1666019417475734e-06,
      "loss": 0.0149,
      "step": 286470
    },
    {
      "epoch": 111.25,
      "learning_rate": 5.166084142394823e-06,
      "loss": 0.0314,
      "step": 286480
    },
    {
      "epoch": 111.26,
      "learning_rate": 5.165566343042072e-06,
      "loss": 0.0591,
      "step": 286490
    },
    {
      "epoch": 111.26,
      "learning_rate": 5.165048543689321e-06,
      "loss": 0.025,
      "step": 286500
    },
    {
      "epoch": 111.27,
      "learning_rate": 5.16453074433657e-06,
      "loss": 0.0961,
      "step": 286510
    },
    {
      "epoch": 111.27,
      "learning_rate": 5.16401294498382e-06,
      "loss": 0.0323,
      "step": 286520
    },
    {
      "epoch": 111.27,
      "learning_rate": 5.163495145631069e-06,
      "loss": 0.2161,
      "step": 286530
    },
    {
      "epoch": 111.28,
      "learning_rate": 5.162977346278317e-06,
      "loss": 0.0004,
      "step": 286540
    },
    {
      "epoch": 111.28,
      "learning_rate": 5.162459546925567e-06,
      "loss": 0.0766,
      "step": 286550
    },
    {
      "epoch": 111.29,
      "learning_rate": 5.161941747572816e-06,
      "loss": 0.0387,
      "step": 286560
    },
    {
      "epoch": 111.29,
      "learning_rate": 5.161423948220065e-06,
      "loss": 0.0718,
      "step": 286570
    },
    {
      "epoch": 111.29,
      "learning_rate": 5.160906148867314e-06,
      "loss": 0.0019,
      "step": 286580
    },
    {
      "epoch": 111.3,
      "learning_rate": 5.160388349514564e-06,
      "loss": 0.0156,
      "step": 286590
    },
    {
      "epoch": 111.3,
      "learning_rate": 5.1598705501618126e-06,
      "loss": 0.0086,
      "step": 286600
    },
    {
      "epoch": 111.3,
      "learning_rate": 5.159352750809062e-06,
      "loss": 0.0412,
      "step": 286610
    },
    {
      "epoch": 111.31,
      "learning_rate": 5.158834951456311e-06,
      "loss": 0.0952,
      "step": 286620
    },
    {
      "epoch": 111.31,
      "learning_rate": 5.1583171521035606e-06,
      "loss": 0.0724,
      "step": 286630
    },
    {
      "epoch": 111.32,
      "learning_rate": 5.157799352750809e-06,
      "loss": 0.0547,
      "step": 286640
    },
    {
      "epoch": 111.32,
      "learning_rate": 5.157281553398059e-06,
      "loss": 0.0089,
      "step": 286650
    },
    {
      "epoch": 111.32,
      "learning_rate": 5.156763754045308e-06,
      "loss": 0.0953,
      "step": 286660
    },
    {
      "epoch": 111.33,
      "learning_rate": 5.156245954692557e-06,
      "loss": 0.0714,
      "step": 286670
    },
    {
      "epoch": 111.33,
      "learning_rate": 5.155728155339806e-06,
      "loss": 0.0487,
      "step": 286680
    },
    {
      "epoch": 111.34,
      "learning_rate": 5.155210355987056e-06,
      "loss": 0.0109,
      "step": 286690
    },
    {
      "epoch": 111.34,
      "learning_rate": 5.1546925566343045e-06,
      "loss": 0.0281,
      "step": 286700
    },
    {
      "epoch": 111.34,
      "learning_rate": 5.154174757281554e-06,
      "loss": 0.0397,
      "step": 286710
    },
    {
      "epoch": 111.35,
      "learning_rate": 5.153656957928803e-06,
      "loss": 0.0131,
      "step": 286720
    },
    {
      "epoch": 111.35,
      "learning_rate": 5.1531391585760525e-06,
      "loss": 0.0162,
      "step": 286730
    },
    {
      "epoch": 111.36,
      "learning_rate": 5.152621359223301e-06,
      "loss": 0.1855,
      "step": 286740
    },
    {
      "epoch": 111.36,
      "learning_rate": 5.15210355987055e-06,
      "loss": 0.0002,
      "step": 286750
    },
    {
      "epoch": 111.36,
      "learning_rate": 5.1515857605178e-06,
      "loss": 0.0288,
      "step": 286760
    },
    {
      "epoch": 111.37,
      "learning_rate": 5.151067961165049e-06,
      "loss": 0.0024,
      "step": 286770
    },
    {
      "epoch": 111.37,
      "learning_rate": 5.150550161812298e-06,
      "loss": 0.0003,
      "step": 286780
    },
    {
      "epoch": 111.37,
      "learning_rate": 5.150032362459547e-06,
      "loss": 0.0814,
      "step": 286790
    },
    {
      "epoch": 111.38,
      "learning_rate": 5.1495145631067964e-06,
      "loss": 0.0099,
      "step": 286800
    },
    {
      "epoch": 111.38,
      "learning_rate": 5.148996763754046e-06,
      "loss": 0.0005,
      "step": 286810
    },
    {
      "epoch": 111.39,
      "learning_rate": 5.148478964401295e-06,
      "loss": 0.0701,
      "step": 286820
    },
    {
      "epoch": 111.39,
      "learning_rate": 5.147961165048544e-06,
      "loss": 0.0477,
      "step": 286830
    },
    {
      "epoch": 111.39,
      "learning_rate": 5.147443365695793e-06,
      "loss": 0.0091,
      "step": 286840
    },
    {
      "epoch": 111.4,
      "learning_rate": 5.146925566343043e-06,
      "loss": 0.0204,
      "step": 286850
    },
    {
      "epoch": 111.4,
      "learning_rate": 5.146407766990292e-06,
      "loss": 0.0412,
      "step": 286860
    },
    {
      "epoch": 111.41,
      "learning_rate": 5.14588996763754e-06,
      "loss": 0.167,
      "step": 286870
    },
    {
      "epoch": 111.41,
      "learning_rate": 5.14537216828479e-06,
      "loss": 0.0688,
      "step": 286880
    },
    {
      "epoch": 111.41,
      "learning_rate": 5.14485436893204e-06,
      "loss": 0.0668,
      "step": 286890
    },
    {
      "epoch": 111.42,
      "learning_rate": 5.1443365695792875e-06,
      "loss": 0.1327,
      "step": 286900
    },
    {
      "epoch": 111.42,
      "learning_rate": 5.143818770226537e-06,
      "loss": 0.0214,
      "step": 286910
    },
    {
      "epoch": 111.43,
      "learning_rate": 5.143300970873787e-06,
      "loss": 0.0365,
      "step": 286920
    },
    {
      "epoch": 111.43,
      "learning_rate": 5.142783171521036e-06,
      "loss": 0.0838,
      "step": 286930
    },
    {
      "epoch": 111.43,
      "learning_rate": 5.142265372168284e-06,
      "loss": 0.1303,
      "step": 286940
    },
    {
      "epoch": 111.44,
      "learning_rate": 5.141747572815534e-06,
      "loss": 0.0333,
      "step": 286950
    },
    {
      "epoch": 111.44,
      "learning_rate": 5.1412297734627836e-06,
      "loss": 0.0152,
      "step": 286960
    },
    {
      "epoch": 111.44,
      "learning_rate": 5.140711974110033e-06,
      "loss": 0.01,
      "step": 286970
    },
    {
      "epoch": 111.45,
      "learning_rate": 5.140194174757281e-06,
      "loss": 0.0368,
      "step": 286980
    },
    {
      "epoch": 111.45,
      "learning_rate": 5.139676375404531e-06,
      "loss": 0.0748,
      "step": 286990
    },
    {
      "epoch": 111.46,
      "learning_rate": 5.13915857605178e-06,
      "loss": 0.0571,
      "step": 287000
    },
    {
      "epoch": 111.46,
      "learning_rate": 5.13864077669903e-06,
      "loss": 0.0008,
      "step": 287010
    },
    {
      "epoch": 111.46,
      "learning_rate": 5.138122977346278e-06,
      "loss": 0.0979,
      "step": 287020
    },
    {
      "epoch": 111.47,
      "learning_rate": 5.1376051779935275e-06,
      "loss": 0.0304,
      "step": 287030
    },
    {
      "epoch": 111.47,
      "learning_rate": 5.137087378640777e-06,
      "loss": 0.0792,
      "step": 287040
    },
    {
      "epoch": 111.48,
      "learning_rate": 5.136569579288027e-06,
      "loss": 0.0324,
      "step": 287050
    },
    {
      "epoch": 111.48,
      "learning_rate": 5.136051779935276e-06,
      "loss": 0.1361,
      "step": 287060
    },
    {
      "epoch": 111.48,
      "learning_rate": 5.135533980582524e-06,
      "loss": 0.0975,
      "step": 287070
    },
    {
      "epoch": 111.49,
      "learning_rate": 5.135016181229774e-06,
      "loss": 0.0126,
      "step": 287080
    },
    {
      "epoch": 111.49,
      "learning_rate": 5.1344983818770235e-06,
      "loss": 0.1261,
      "step": 287090
    },
    {
      "epoch": 111.5,
      "learning_rate": 5.133980582524273e-06,
      "loss": 0.0225,
      "step": 287100
    },
    {
      "epoch": 111.5,
      "learning_rate": 5.133462783171521e-06,
      "loss": 0.0047,
      "step": 287110
    },
    {
      "epoch": 111.5,
      "learning_rate": 5.132944983818771e-06,
      "loss": 0.0636,
      "step": 287120
    },
    {
      "epoch": 111.51,
      "learning_rate": 5.13242718446602e-06,
      "loss": 0.0115,
      "step": 287130
    },
    {
      "epoch": 111.51,
      "learning_rate": 5.13190938511327e-06,
      "loss": 0.0151,
      "step": 287140
    },
    {
      "epoch": 111.51,
      "learning_rate": 5.131391585760518e-06,
      "loss": 0.1257,
      "step": 287150
    },
    {
      "epoch": 111.52,
      "learning_rate": 5.1308737864077674e-06,
      "loss": 0.0146,
      "step": 287160
    },
    {
      "epoch": 111.52,
      "learning_rate": 5.130355987055017e-06,
      "loss": 0.0371,
      "step": 287170
    },
    {
      "epoch": 111.53,
      "learning_rate": 5.129838187702267e-06,
      "loss": 0.0603,
      "step": 287180
    },
    {
      "epoch": 111.53,
      "learning_rate": 5.129320388349515e-06,
      "loss": 0.0362,
      "step": 287190
    },
    {
      "epoch": 111.53,
      "learning_rate": 5.128802588996764e-06,
      "loss": 0.016,
      "step": 287200
    },
    {
      "epoch": 111.54,
      "learning_rate": 5.128284789644014e-06,
      "loss": 0.0008,
      "step": 287210
    },
    {
      "epoch": 111.54,
      "learning_rate": 5.127766990291263e-06,
      "loss": 0.0725,
      "step": 287220
    },
    {
      "epoch": 111.55,
      "learning_rate": 5.127249190938511e-06,
      "loss": 0.0352,
      "step": 287230
    },
    {
      "epoch": 111.55,
      "learning_rate": 5.126731391585761e-06,
      "loss": 0.0221,
      "step": 287240
    },
    {
      "epoch": 111.55,
      "learning_rate": 5.126213592233011e-06,
      "loss": 0.0594,
      "step": 287250
    },
    {
      "epoch": 111.56,
      "learning_rate": 5.125695792880259e-06,
      "loss": 0.0781,
      "step": 287260
    },
    {
      "epoch": 111.56,
      "learning_rate": 5.125177993527508e-06,
      "loss": 0.0001,
      "step": 287270
    },
    {
      "epoch": 111.57,
      "learning_rate": 5.124660194174758e-06,
      "loss": 0.0946,
      "step": 287280
    },
    {
      "epoch": 111.57,
      "learning_rate": 5.124142394822007e-06,
      "loss": 0.0085,
      "step": 287290
    },
    {
      "epoch": 111.57,
      "learning_rate": 5.123624595469256e-06,
      "loss": 0.0107,
      "step": 287300
    },
    {
      "epoch": 111.58,
      "learning_rate": 5.123106796116505e-06,
      "loss": 0.0859,
      "step": 287310
    },
    {
      "epoch": 111.58,
      "learning_rate": 5.1225889967637546e-06,
      "loss": 0.0331,
      "step": 287320
    },
    {
      "epoch": 111.58,
      "learning_rate": 5.122071197411004e-06,
      "loss": 0.0244,
      "step": 287330
    },
    {
      "epoch": 111.59,
      "learning_rate": 5.121553398058253e-06,
      "loss": 0.0319,
      "step": 287340
    },
    {
      "epoch": 111.59,
      "learning_rate": 5.121035598705502e-06,
      "loss": 0.0674,
      "step": 287350
    },
    {
      "epoch": 111.6,
      "learning_rate": 5.120517799352751e-06,
      "loss": 0.0857,
      "step": 287360
    },
    {
      "epoch": 111.6,
      "learning_rate": 5.12e-06,
      "loss": 0.0024,
      "step": 287370
    },
    {
      "epoch": 111.6,
      "learning_rate": 5.11948220064725e-06,
      "loss": 0.0001,
      "step": 287380
    },
    {
      "epoch": 111.61,
      "learning_rate": 5.1189644012944985e-06,
      "loss": 0.0297,
      "step": 287390
    },
    {
      "epoch": 111.61,
      "learning_rate": 5.118446601941748e-06,
      "loss": 0.0386,
      "step": 287400
    },
    {
      "epoch": 111.62,
      "learning_rate": 5.117928802588997e-06,
      "loss": 0.0091,
      "step": 287410
    },
    {
      "epoch": 111.62,
      "learning_rate": 5.1174110032362465e-06,
      "loss": 0.033,
      "step": 287420
    },
    {
      "epoch": 111.62,
      "learning_rate": 5.116893203883495e-06,
      "loss": 0.0396,
      "step": 287430
    },
    {
      "epoch": 111.63,
      "learning_rate": 5.116375404530745e-06,
      "loss": 0.0557,
      "step": 287440
    },
    {
      "epoch": 111.63,
      "learning_rate": 5.115857605177994e-06,
      "loss": 0.0808,
      "step": 287450
    },
    {
      "epoch": 111.63,
      "learning_rate": 5.115339805825243e-06,
      "loss": 0.0511,
      "step": 287460
    },
    {
      "epoch": 111.64,
      "learning_rate": 5.114822006472492e-06,
      "loss": 0.0001,
      "step": 287470
    },
    {
      "epoch": 111.64,
      "learning_rate": 5.114304207119742e-06,
      "loss": 0.031,
      "step": 287480
    },
    {
      "epoch": 111.65,
      "learning_rate": 5.1137864077669904e-06,
      "loss": 0.0023,
      "step": 287490
    },
    {
      "epoch": 111.65,
      "learning_rate": 5.11326860841424e-06,
      "loss": 0.046,
      "step": 287500
    },
    {
      "epoch": 111.65,
      "learning_rate": 5.112750809061489e-06,
      "loss": 0.0003,
      "step": 287510
    },
    {
      "epoch": 111.66,
      "learning_rate": 5.1122330097087384e-06,
      "loss": 0.0287,
      "step": 287520
    },
    {
      "epoch": 111.66,
      "learning_rate": 5.111715210355987e-06,
      "loss": 0.0443,
      "step": 287530
    },
    {
      "epoch": 111.67,
      "learning_rate": 5.111197411003237e-06,
      "loss": 0.0087,
      "step": 287540
    },
    {
      "epoch": 111.67,
      "learning_rate": 5.110679611650486e-06,
      "loss": 0.0764,
      "step": 287550
    },
    {
      "epoch": 111.67,
      "learning_rate": 5.110161812297734e-06,
      "loss": 0.0028,
      "step": 287560
    },
    {
      "epoch": 111.68,
      "learning_rate": 5.109644012944984e-06,
      "loss": 0.0001,
      "step": 287570
    },
    {
      "epoch": 111.68,
      "learning_rate": 5.109126213592234e-06,
      "loss": 0.011,
      "step": 287580
    },
    {
      "epoch": 111.69,
      "learning_rate": 5.108608414239483e-06,
      "loss": 0.0821,
      "step": 287590
    },
    {
      "epoch": 111.69,
      "learning_rate": 5.108090614886731e-06,
      "loss": 0.0745,
      "step": 287600
    },
    {
      "epoch": 111.69,
      "learning_rate": 5.107572815533981e-06,
      "loss": 0.1662,
      "step": 287610
    },
    {
      "epoch": 111.7,
      "learning_rate": 5.10705501618123e-06,
      "loss": 0.0483,
      "step": 287620
    },
    {
      "epoch": 111.7,
      "learning_rate": 5.10653721682848e-06,
      "loss": 0.0236,
      "step": 287630
    },
    {
      "epoch": 111.7,
      "learning_rate": 5.106019417475728e-06,
      "loss": 0.0129,
      "step": 287640
    },
    {
      "epoch": 111.71,
      "learning_rate": 5.1055016181229776e-06,
      "loss": 0.0002,
      "step": 287650
    },
    {
      "epoch": 111.71,
      "learning_rate": 5.104983818770227e-06,
      "loss": 0.0739,
      "step": 287660
    },
    {
      "epoch": 111.72,
      "learning_rate": 5.104466019417477e-06,
      "loss": 0.0644,
      "step": 287670
    },
    {
      "epoch": 111.72,
      "learning_rate": 5.103948220064725e-06,
      "loss": 0.0735,
      "step": 287680
    },
    {
      "epoch": 111.72,
      "learning_rate": 5.103430420711974e-06,
      "loss": 0.0028,
      "step": 287690
    },
    {
      "epoch": 111.73,
      "learning_rate": 5.102912621359224e-06,
      "loss": 0.0106,
      "step": 287700
    },
    {
      "epoch": 111.73,
      "learning_rate": 5.1023948220064736e-06,
      "loss": 0.0066,
      "step": 287710
    },
    {
      "epoch": 111.74,
      "learning_rate": 5.1018770226537215e-06,
      "loss": 0.022,
      "step": 287720
    },
    {
      "epoch": 111.74,
      "learning_rate": 5.101359223300971e-06,
      "loss": 0.0264,
      "step": 287730
    },
    {
      "epoch": 111.74,
      "learning_rate": 5.100841423948221e-06,
      "loss": 0.0464,
      "step": 287740
    },
    {
      "epoch": 111.75,
      "learning_rate": 5.10032362459547e-06,
      "loss": 0.0726,
      "step": 287750
    },
    {
      "epoch": 111.75,
      "learning_rate": 5.099805825242718e-06,
      "loss": 0.1451,
      "step": 287760
    },
    {
      "epoch": 111.76,
      "learning_rate": 5.099288025889968e-06,
      "loss": 0.034,
      "step": 287770
    },
    {
      "epoch": 111.76,
      "learning_rate": 5.0987702265372175e-06,
      "loss": 0.1463,
      "step": 287780
    },
    {
      "epoch": 111.76,
      "learning_rate": 5.098252427184467e-06,
      "loss": 0.089,
      "step": 287790
    },
    {
      "epoch": 111.77,
      "learning_rate": 5.097734627831715e-06,
      "loss": 0.0004,
      "step": 287800
    },
    {
      "epoch": 111.77,
      "learning_rate": 5.097216828478965e-06,
      "loss": 0.1094,
      "step": 287810
    },
    {
      "epoch": 111.77,
      "learning_rate": 5.096699029126214e-06,
      "loss": 0.0216,
      "step": 287820
    },
    {
      "epoch": 111.78,
      "learning_rate": 5.096181229773464e-06,
      "loss": 0.014,
      "step": 287830
    },
    {
      "epoch": 111.78,
      "learning_rate": 5.095663430420712e-06,
      "loss": 0.0018,
      "step": 287840
    },
    {
      "epoch": 111.79,
      "learning_rate": 5.0951456310679614e-06,
      "loss": 0.0271,
      "step": 287850
    },
    {
      "epoch": 111.79,
      "learning_rate": 5.094627831715211e-06,
      "loss": 0.0298,
      "step": 287860
    },
    {
      "epoch": 111.79,
      "learning_rate": 5.094110032362461e-06,
      "loss": 0.033,
      "step": 287870
    },
    {
      "epoch": 111.8,
      "learning_rate": 5.093592233009709e-06,
      "loss": 0.1338,
      "step": 287880
    },
    {
      "epoch": 111.8,
      "learning_rate": 5.093074433656958e-06,
      "loss": 0.0556,
      "step": 287890
    },
    {
      "epoch": 111.81,
      "learning_rate": 5.092556634304208e-06,
      "loss": 0.0237,
      "step": 287900
    },
    {
      "epoch": 111.81,
      "learning_rate": 5.0920388349514575e-06,
      "loss": 0.0402,
      "step": 287910
    },
    {
      "epoch": 111.81,
      "learning_rate": 5.091521035598705e-06,
      "loss": 0.0109,
      "step": 287920
    },
    {
      "epoch": 111.82,
      "learning_rate": 5.091003236245955e-06,
      "loss": 0.0312,
      "step": 287930
    },
    {
      "epoch": 111.82,
      "learning_rate": 5.090485436893205e-06,
      "loss": 0.0001,
      "step": 287940
    },
    {
      "epoch": 111.83,
      "learning_rate": 5.089967637540454e-06,
      "loss": 0.002,
      "step": 287950
    },
    {
      "epoch": 111.83,
      "learning_rate": 5.089449838187702e-06,
      "loss": 0.0896,
      "step": 287960
    },
    {
      "epoch": 111.83,
      "learning_rate": 5.088932038834952e-06,
      "loss": 0.116,
      "step": 287970
    },
    {
      "epoch": 111.84,
      "learning_rate": 5.088414239482201e-06,
      "loss": 0.0003,
      "step": 287980
    },
    {
      "epoch": 111.84,
      "learning_rate": 5.087896440129451e-06,
      "loss": 0.1049,
      "step": 287990
    },
    {
      "epoch": 111.84,
      "learning_rate": 5.087378640776699e-06,
      "loss": 0.1248,
      "step": 288000
    },
    {
      "epoch": 111.85,
      "learning_rate": 5.0868608414239486e-06,
      "loss": 0.0004,
      "step": 288010
    },
    {
      "epoch": 111.85,
      "learning_rate": 5.086343042071198e-06,
      "loss": 0.0917,
      "step": 288020
    },
    {
      "epoch": 111.86,
      "learning_rate": 5.085825242718447e-06,
      "loss": 0.0008,
      "step": 288030
    },
    {
      "epoch": 111.86,
      "learning_rate": 5.085307443365696e-06,
      "loss": 0.0945,
      "step": 288040
    },
    {
      "epoch": 111.86,
      "learning_rate": 5.084789644012945e-06,
      "loss": 0.0064,
      "step": 288050
    },
    {
      "epoch": 111.87,
      "learning_rate": 5.084271844660195e-06,
      "loss": 0.046,
      "step": 288060
    },
    {
      "epoch": 111.87,
      "learning_rate": 5.083754045307444e-06,
      "loss": 0.057,
      "step": 288070
    },
    {
      "epoch": 111.88,
      "learning_rate": 5.0832362459546925e-06,
      "loss": 0.04,
      "step": 288080
    },
    {
      "epoch": 111.88,
      "learning_rate": 5.082718446601942e-06,
      "loss": 0.0108,
      "step": 288090
    },
    {
      "epoch": 111.88,
      "learning_rate": 5.082200647249192e-06,
      "loss": 0.0409,
      "step": 288100
    },
    {
      "epoch": 111.89,
      "learning_rate": 5.0816828478964405e-06,
      "loss": 0.058,
      "step": 288110
    },
    {
      "epoch": 111.89,
      "learning_rate": 5.081165048543689e-06,
      "loss": 0.0098,
      "step": 288120
    },
    {
      "epoch": 111.9,
      "learning_rate": 5.080647249190939e-06,
      "loss": 0.1203,
      "step": 288130
    },
    {
      "epoch": 111.9,
      "learning_rate": 5.0801294498381885e-06,
      "loss": 0.0348,
      "step": 288140
    },
    {
      "epoch": 111.9,
      "learning_rate": 5.079611650485437e-06,
      "loss": 0.009,
      "step": 288150
    },
    {
      "epoch": 111.91,
      "learning_rate": 5.079093851132687e-06,
      "loss": 0.0128,
      "step": 288160
    },
    {
      "epoch": 111.91,
      "learning_rate": 5.078576051779936e-06,
      "loss": 0.0684,
      "step": 288170
    },
    {
      "epoch": 111.91,
      "learning_rate": 5.078058252427185e-06,
      "loss": 0.0464,
      "step": 288180
    },
    {
      "epoch": 111.92,
      "learning_rate": 5.077540453074434e-06,
      "loss": 0.0218,
      "step": 288190
    },
    {
      "epoch": 111.92,
      "learning_rate": 5.077022653721684e-06,
      "loss": 0.1688,
      "step": 288200
    },
    {
      "epoch": 111.93,
      "learning_rate": 5.0765048543689324e-06,
      "loss": 0.1379,
      "step": 288210
    },
    {
      "epoch": 111.93,
      "learning_rate": 5.075987055016181e-06,
      "loss": 0.0335,
      "step": 288220
    },
    {
      "epoch": 111.93,
      "learning_rate": 5.075469255663431e-06,
      "loss": 0.0135,
      "step": 288230
    },
    {
      "epoch": 111.94,
      "learning_rate": 5.0749514563106805e-06,
      "loss": 0.0147,
      "step": 288240
    },
    {
      "epoch": 111.94,
      "learning_rate": 5.074433656957929e-06,
      "loss": 0.1257,
      "step": 288250
    },
    {
      "epoch": 111.95,
      "learning_rate": 5.073915857605178e-06,
      "loss": 0.0036,
      "step": 288260
    },
    {
      "epoch": 111.95,
      "learning_rate": 5.073398058252428e-06,
      "loss": 0.0188,
      "step": 288270
    },
    {
      "epoch": 111.95,
      "learning_rate": 5.072880258899677e-06,
      "loss": 0.0911,
      "step": 288280
    },
    {
      "epoch": 111.96,
      "learning_rate": 5.072362459546926e-06,
      "loss": 0.0074,
      "step": 288290
    },
    {
      "epoch": 111.96,
      "learning_rate": 5.071844660194175e-06,
      "loss": 0.0001,
      "step": 288300
    },
    {
      "epoch": 111.97,
      "learning_rate": 5.071326860841424e-06,
      "loss": 0.0101,
      "step": 288310
    },
    {
      "epoch": 111.97,
      "learning_rate": 5.070809061488674e-06,
      "loss": 0.0009,
      "step": 288320
    },
    {
      "epoch": 111.97,
      "learning_rate": 5.070291262135923e-06,
      "loss": 0.1474,
      "step": 288330
    },
    {
      "epoch": 111.98,
      "learning_rate": 5.0697734627831716e-06,
      "loss": 0.161,
      "step": 288340
    },
    {
      "epoch": 111.98,
      "learning_rate": 5.069255663430421e-06,
      "loss": 0.088,
      "step": 288350
    },
    {
      "epoch": 111.98,
      "learning_rate": 5.068737864077671e-06,
      "loss": 0.0103,
      "step": 288360
    },
    {
      "epoch": 111.99,
      "learning_rate": 5.068220064724919e-06,
      "loss": 0.1057,
      "step": 288370
    },
    {
      "epoch": 111.99,
      "learning_rate": 5.067702265372168e-06,
      "loss": 0.0452,
      "step": 288380
    },
    {
      "epoch": 112.0,
      "learning_rate": 5.067184466019418e-06,
      "loss": 0.1511,
      "step": 288390
    },
    {
      "epoch": 112.0,
      "learning_rate": 5.0666666666666676e-06,
      "loss": 0.001,
      "step": 288400
    },
    {
      "epoch": 112.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3811458349227905,
      "eval_runtime": 8.2582,
      "eval_samples_per_second": 440.168,
      "eval_steps_per_second": 55.097,
      "step": 288400
    },
    {
      "epoch": 112.0,
      "learning_rate": 5.0661488673139155e-06,
      "loss": 0.0133,
      "step": 288410
    },
    {
      "epoch": 112.01,
      "learning_rate": 5.065631067961165e-06,
      "loss": 0.0474,
      "step": 288420
    },
    {
      "epoch": 112.01,
      "learning_rate": 5.065113268608415e-06,
      "loss": 0.0331,
      "step": 288430
    },
    {
      "epoch": 112.02,
      "learning_rate": 5.064595469255664e-06,
      "loss": 0.0031,
      "step": 288440
    },
    {
      "epoch": 112.02,
      "learning_rate": 5.064077669902912e-06,
      "loss": 0.0658,
      "step": 288450
    },
    {
      "epoch": 112.02,
      "learning_rate": 5.063559870550162e-06,
      "loss": 0.05,
      "step": 288460
    },
    {
      "epoch": 112.03,
      "learning_rate": 5.0630420711974115e-06,
      "loss": 0.0369,
      "step": 288470
    },
    {
      "epoch": 112.03,
      "learning_rate": 5.062524271844661e-06,
      "loss": 0.0539,
      "step": 288480
    },
    {
      "epoch": 112.03,
      "learning_rate": 5.062006472491909e-06,
      "loss": 0.0221,
      "step": 288490
    },
    {
      "epoch": 112.04,
      "learning_rate": 5.061488673139159e-06,
      "loss": 0.0006,
      "step": 288500
    },
    {
      "epoch": 112.04,
      "learning_rate": 5.060970873786408e-06,
      "loss": 0.0273,
      "step": 288510
    },
    {
      "epoch": 112.05,
      "learning_rate": 5.060453074433658e-06,
      "loss": 0.0001,
      "step": 288520
    },
    {
      "epoch": 112.05,
      "learning_rate": 5.059935275080906e-06,
      "loss": 0.0098,
      "step": 288530
    },
    {
      "epoch": 112.05,
      "learning_rate": 5.0594174757281554e-06,
      "loss": 0.0208,
      "step": 288540
    },
    {
      "epoch": 112.06,
      "learning_rate": 5.058899676375405e-06,
      "loss": 0.0106,
      "step": 288550
    },
    {
      "epoch": 112.06,
      "learning_rate": 5.058381877022655e-06,
      "loss": 0.0851,
      "step": 288560
    },
    {
      "epoch": 112.07,
      "learning_rate": 5.057864077669903e-06,
      "loss": 0.0597,
      "step": 288570
    },
    {
      "epoch": 112.07,
      "learning_rate": 5.057346278317152e-06,
      "loss": 0.0583,
      "step": 288580
    },
    {
      "epoch": 112.07,
      "learning_rate": 5.056828478964402e-06,
      "loss": 0.0226,
      "step": 288590
    },
    {
      "epoch": 112.08,
      "learning_rate": 5.0563106796116515e-06,
      "loss": 0.0003,
      "step": 288600
    },
    {
      "epoch": 112.08,
      "learning_rate": 5.055792880258899e-06,
      "loss": 0.1237,
      "step": 288610
    },
    {
      "epoch": 112.09,
      "learning_rate": 5.055275080906149e-06,
      "loss": 0.025,
      "step": 288620
    },
    {
      "epoch": 112.09,
      "learning_rate": 5.054757281553399e-06,
      "loss": 0.0011,
      "step": 288630
    },
    {
      "epoch": 112.09,
      "learning_rate": 5.054239482200648e-06,
      "loss": 0.0162,
      "step": 288640
    },
    {
      "epoch": 112.1,
      "learning_rate": 5.053721682847896e-06,
      "loss": 0.0002,
      "step": 288650
    },
    {
      "epoch": 112.1,
      "learning_rate": 5.053203883495146e-06,
      "loss": 0.0625,
      "step": 288660
    },
    {
      "epoch": 112.1,
      "learning_rate": 5.052686084142395e-06,
      "loss": 0.0004,
      "step": 288670
    },
    {
      "epoch": 112.11,
      "learning_rate": 5.052168284789645e-06,
      "loss": 0.1508,
      "step": 288680
    },
    {
      "epoch": 112.11,
      "learning_rate": 5.051650485436893e-06,
      "loss": 0.0418,
      "step": 288690
    },
    {
      "epoch": 112.12,
      "learning_rate": 5.0511326860841426e-06,
      "loss": 0.0716,
      "step": 288700
    },
    {
      "epoch": 112.12,
      "learning_rate": 5.050614886731392e-06,
      "loss": 0.0956,
      "step": 288710
    },
    {
      "epoch": 112.12,
      "learning_rate": 5.050097087378642e-06,
      "loss": 0.1314,
      "step": 288720
    },
    {
      "epoch": 112.13,
      "learning_rate": 5.0495792880258906e-06,
      "loss": 0.0369,
      "step": 288730
    },
    {
      "epoch": 112.13,
      "learning_rate": 5.049061488673139e-06,
      "loss": 0.0462,
      "step": 288740
    },
    {
      "epoch": 112.14,
      "learning_rate": 5.048543689320389e-06,
      "loss": 0.1149,
      "step": 288750
    },
    {
      "epoch": 112.14,
      "learning_rate": 5.0480258899676386e-06,
      "loss": 0.0179,
      "step": 288760
    },
    {
      "epoch": 112.14,
      "learning_rate": 5.047508090614887e-06,
      "loss": 0.0203,
      "step": 288770
    },
    {
      "epoch": 112.15,
      "learning_rate": 5.046990291262136e-06,
      "loss": 0.0496,
      "step": 288780
    },
    {
      "epoch": 112.15,
      "learning_rate": 5.046472491909386e-06,
      "loss": 0.1111,
      "step": 288790
    },
    {
      "epoch": 112.16,
      "learning_rate": 5.045954692556635e-06,
      "loss": 0.0163,
      "step": 288800
    },
    {
      "epoch": 112.16,
      "learning_rate": 5.045436893203884e-06,
      "loss": 0.1297,
      "step": 288810
    },
    {
      "epoch": 112.16,
      "learning_rate": 5.044919093851133e-06,
      "loss": 0.0171,
      "step": 288820
    },
    {
      "epoch": 112.17,
      "learning_rate": 5.0444012944983825e-06,
      "loss": 0.0001,
      "step": 288830
    },
    {
      "epoch": 112.17,
      "learning_rate": 5.043883495145631e-06,
      "loss": 0.0234,
      "step": 288840
    },
    {
      "epoch": 112.17,
      "learning_rate": 5.043365695792881e-06,
      "loss": 0.0283,
      "step": 288850
    },
    {
      "epoch": 112.18,
      "learning_rate": 5.04284789644013e-06,
      "loss": 0.0002,
      "step": 288860
    },
    {
      "epoch": 112.18,
      "learning_rate": 5.042330097087379e-06,
      "loss": 0.0565,
      "step": 288870
    },
    {
      "epoch": 112.19,
      "learning_rate": 5.041812297734628e-06,
      "loss": 0.0042,
      "step": 288880
    },
    {
      "epoch": 112.19,
      "learning_rate": 5.041294498381878e-06,
      "loss": 0.0011,
      "step": 288890
    },
    {
      "epoch": 112.19,
      "learning_rate": 5.0407766990291264e-06,
      "loss": 0.0486,
      "step": 288900
    },
    {
      "epoch": 112.2,
      "learning_rate": 5.040258899676376e-06,
      "loss": 0.0274,
      "step": 288910
    },
    {
      "epoch": 112.2,
      "learning_rate": 5.039741100323625e-06,
      "loss": 0.0574,
      "step": 288920
    },
    {
      "epoch": 112.21,
      "learning_rate": 5.0392233009708745e-06,
      "loss": 0.0129,
      "step": 288930
    },
    {
      "epoch": 112.21,
      "learning_rate": 5.038705501618123e-06,
      "loss": 0.0167,
      "step": 288940
    },
    {
      "epoch": 112.21,
      "learning_rate": 5.038187702265373e-06,
      "loss": 0.177,
      "step": 288950
    },
    {
      "epoch": 112.22,
      "learning_rate": 5.037669902912622e-06,
      "loss": 0.0167,
      "step": 288960
    },
    {
      "epoch": 112.22,
      "learning_rate": 5.037152103559871e-06,
      "loss": 0.1575,
      "step": 288970
    },
    {
      "epoch": 112.23,
      "learning_rate": 5.03663430420712e-06,
      "loss": 0.1229,
      "step": 288980
    },
    {
      "epoch": 112.23,
      "learning_rate": 5.03611650485437e-06,
      "loss": 0.0435,
      "step": 288990
    },
    {
      "epoch": 112.23,
      "learning_rate": 5.035598705501618e-06,
      "loss": 0.0266,
      "step": 289000
    },
    {
      "epoch": 112.24,
      "learning_rate": 5.035080906148868e-06,
      "loss": 0.0106,
      "step": 289010
    },
    {
      "epoch": 112.24,
      "learning_rate": 5.034563106796117e-06,
      "loss": 0.1688,
      "step": 289020
    },
    {
      "epoch": 112.24,
      "learning_rate": 5.0340453074433656e-06,
      "loss": 0.047,
      "step": 289030
    },
    {
      "epoch": 112.25,
      "learning_rate": 5.033527508090615e-06,
      "loss": 0.0088,
      "step": 289040
    },
    {
      "epoch": 112.25,
      "learning_rate": 5.033009708737865e-06,
      "loss": 0.0492,
      "step": 289050
    },
    {
      "epoch": 112.26,
      "learning_rate": 5.0324919093851136e-06,
      "loss": 0.0422,
      "step": 289060
    },
    {
      "epoch": 112.26,
      "learning_rate": 5.031974110032362e-06,
      "loss": 0.0142,
      "step": 289070
    },
    {
      "epoch": 112.26,
      "learning_rate": 5.031456310679612e-06,
      "loss": 0.0991,
      "step": 289080
    },
    {
      "epoch": 112.27,
      "learning_rate": 5.0309385113268616e-06,
      "loss": 0.0016,
      "step": 289090
    },
    {
      "epoch": 112.27,
      "learning_rate": 5.03042071197411e-06,
      "loss": 0.0006,
      "step": 289100
    },
    {
      "epoch": 112.28,
      "learning_rate": 5.029902912621359e-06,
      "loss": 0.043,
      "step": 289110
    },
    {
      "epoch": 112.28,
      "learning_rate": 5.029385113268609e-06,
      "loss": 0.0085,
      "step": 289120
    },
    {
      "epoch": 112.28,
      "learning_rate": 5.028867313915858e-06,
      "loss": 0.0834,
      "step": 289130
    },
    {
      "epoch": 112.29,
      "learning_rate": 5.028349514563107e-06,
      "loss": 0.0003,
      "step": 289140
    },
    {
      "epoch": 112.29,
      "learning_rate": 5.027831715210356e-06,
      "loss": 0.0511,
      "step": 289150
    },
    {
      "epoch": 112.3,
      "learning_rate": 5.0273139158576055e-06,
      "loss": 0.0735,
      "step": 289160
    },
    {
      "epoch": 112.3,
      "learning_rate": 5.026796116504855e-06,
      "loss": 0.0235,
      "step": 289170
    },
    {
      "epoch": 112.3,
      "learning_rate": 5.026278317152103e-06,
      "loss": 0.0138,
      "step": 289180
    },
    {
      "epoch": 112.31,
      "learning_rate": 5.025760517799353e-06,
      "loss": 0.0788,
      "step": 289190
    },
    {
      "epoch": 112.31,
      "learning_rate": 5.025242718446602e-06,
      "loss": 0.1815,
      "step": 289200
    },
    {
      "epoch": 112.31,
      "learning_rate": 5.024724919093852e-06,
      "loss": 0.0007,
      "step": 289210
    },
    {
      "epoch": 112.32,
      "learning_rate": 5.0242071197411e-06,
      "loss": 0.0832,
      "step": 289220
    },
    {
      "epoch": 112.32,
      "learning_rate": 5.0236893203883494e-06,
      "loss": 0.0569,
      "step": 289230
    },
    {
      "epoch": 112.33,
      "learning_rate": 5.023171521035599e-06,
      "loss": 0.0447,
      "step": 289240
    },
    {
      "epoch": 112.33,
      "learning_rate": 5.022653721682849e-06,
      "loss": 0.0797,
      "step": 289250
    },
    {
      "epoch": 112.33,
      "learning_rate": 5.022135922330097e-06,
      "loss": 0.0095,
      "step": 289260
    },
    {
      "epoch": 112.34,
      "learning_rate": 5.021618122977346e-06,
      "loss": 0.0486,
      "step": 289270
    },
    {
      "epoch": 112.34,
      "learning_rate": 5.021100323624596e-06,
      "loss": 0.0581,
      "step": 289280
    },
    {
      "epoch": 112.35,
      "learning_rate": 5.0205825242718455e-06,
      "loss": 0.1283,
      "step": 289290
    },
    {
      "epoch": 112.35,
      "learning_rate": 5.020064724919095e-06,
      "loss": 0.1163,
      "step": 289300
    },
    {
      "epoch": 112.35,
      "learning_rate": 5.019546925566343e-06,
      "loss": 0.0301,
      "step": 289310
    },
    {
      "epoch": 112.36,
      "learning_rate": 5.019029126213593e-06,
      "loss": 0.027,
      "step": 289320
    },
    {
      "epoch": 112.36,
      "learning_rate": 5.018511326860842e-06,
      "loss": 0.0116,
      "step": 289330
    },
    {
      "epoch": 112.37,
      "learning_rate": 5.017993527508092e-06,
      "loss": 0.0491,
      "step": 289340
    },
    {
      "epoch": 112.37,
      "learning_rate": 5.01747572815534e-06,
      "loss": 0.0254,
      "step": 289350
    },
    {
      "epoch": 112.37,
      "learning_rate": 5.016957928802589e-06,
      "loss": 0.0761,
      "step": 289360
    },
    {
      "epoch": 112.38,
      "learning_rate": 5.016440129449839e-06,
      "loss": 0.0255,
      "step": 289370
    },
    {
      "epoch": 112.38,
      "learning_rate": 5.015922330097089e-06,
      "loss": 0.143,
      "step": 289380
    },
    {
      "epoch": 112.38,
      "learning_rate": 5.0154045307443366e-06,
      "loss": 0.0269,
      "step": 289390
    },
    {
      "epoch": 112.39,
      "learning_rate": 5.014886731391586e-06,
      "loss": 0.0312,
      "step": 289400
    },
    {
      "epoch": 112.39,
      "learning_rate": 5.014368932038836e-06,
      "loss": 0.0473,
      "step": 289410
    },
    {
      "epoch": 112.4,
      "learning_rate": 5.013851132686085e-06,
      "loss": 0.0605,
      "step": 289420
    },
    {
      "epoch": 112.4,
      "learning_rate": 5.013333333333333e-06,
      "loss": 0.0208,
      "step": 289430
    },
    {
      "epoch": 112.4,
      "learning_rate": 5.012815533980583e-06,
      "loss": 0.0552,
      "step": 289440
    },
    {
      "epoch": 112.41,
      "learning_rate": 5.0122977346278326e-06,
      "loss": 0.0078,
      "step": 289450
    },
    {
      "epoch": 112.41,
      "learning_rate": 5.011779935275082e-06,
      "loss": 0.1286,
      "step": 289460
    },
    {
      "epoch": 112.42,
      "learning_rate": 5.01126213592233e-06,
      "loss": 0.0431,
      "step": 289470
    },
    {
      "epoch": 112.42,
      "learning_rate": 5.01074433656958e-06,
      "loss": 0.1587,
      "step": 289480
    },
    {
      "epoch": 112.42,
      "learning_rate": 5.010226537216829e-06,
      "loss": 0.0003,
      "step": 289490
    },
    {
      "epoch": 112.43,
      "learning_rate": 5.009708737864078e-06,
      "loss": 0.0243,
      "step": 289500
    },
    {
      "epoch": 112.43,
      "learning_rate": 5.009190938511327e-06,
      "loss": 0.0016,
      "step": 289510
    },
    {
      "epoch": 112.43,
      "learning_rate": 5.0086731391585765e-06,
      "loss": 0.0003,
      "step": 289520
    },
    {
      "epoch": 112.44,
      "learning_rate": 5.008155339805826e-06,
      "loss": 0.0001,
      "step": 289530
    },
    {
      "epoch": 112.44,
      "learning_rate": 5.007637540453075e-06,
      "loss": 0.0003,
      "step": 289540
    },
    {
      "epoch": 112.45,
      "learning_rate": 5.007119741100324e-06,
      "loss": 0.0108,
      "step": 289550
    },
    {
      "epoch": 112.45,
      "learning_rate": 5.006601941747573e-06,
      "loss": 0.0445,
      "step": 289560
    },
    {
      "epoch": 112.45,
      "learning_rate": 5.006084142394823e-06,
      "loss": 0.0437,
      "step": 289570
    },
    {
      "epoch": 112.46,
      "learning_rate": 5.005566343042072e-06,
      "loss": 0.0261,
      "step": 289580
    },
    {
      "epoch": 112.46,
      "learning_rate": 5.0050485436893204e-06,
      "loss": 0.014,
      "step": 289590
    },
    {
      "epoch": 112.47,
      "learning_rate": 5.00453074433657e-06,
      "loss": 0.0008,
      "step": 289600
    },
    {
      "epoch": 112.47,
      "learning_rate": 5.00401294498382e-06,
      "loss": 0.0669,
      "step": 289610
    },
    {
      "epoch": 112.47,
      "learning_rate": 5.0034951456310685e-06,
      "loss": 0.0015,
      "step": 289620
    },
    {
      "epoch": 112.48,
      "learning_rate": 5.002977346278317e-06,
      "loss": 0.0097,
      "step": 289630
    },
    {
      "epoch": 112.48,
      "learning_rate": 5.002459546925567e-06,
      "loss": 0.1313,
      "step": 289640
    },
    {
      "epoch": 112.49,
      "learning_rate": 5.001941747572816e-06,
      "loss": 0.134,
      "step": 289650
    },
    {
      "epoch": 112.49,
      "learning_rate": 5.001423948220065e-06,
      "loss": 0.0003,
      "step": 289660
    },
    {
      "epoch": 112.49,
      "learning_rate": 5.000906148867314e-06,
      "loss": 0.0664,
      "step": 289670
    },
    {
      "epoch": 112.5,
      "learning_rate": 5.000388349514564e-06,
      "loss": 0.0808,
      "step": 289680
    },
    {
      "epoch": 112.5,
      "learning_rate": 4.999870550161812e-06,
      "loss": 0.0217,
      "step": 289690
    },
    {
      "epoch": 112.5,
      "learning_rate": 4.999352750809062e-06,
      "loss": 0.0001,
      "step": 289700
    },
    {
      "epoch": 112.51,
      "learning_rate": 4.998834951456311e-06,
      "loss": 0.0254,
      "step": 289710
    },
    {
      "epoch": 112.51,
      "learning_rate": 4.99831715210356e-06,
      "loss": 0.1837,
      "step": 289720
    },
    {
      "epoch": 112.52,
      "learning_rate": 4.997799352750809e-06,
      "loss": 0.0852,
      "step": 289730
    },
    {
      "epoch": 112.52,
      "learning_rate": 4.997281553398059e-06,
      "loss": 0.0987,
      "step": 289740
    },
    {
      "epoch": 112.52,
      "learning_rate": 4.9967637540453076e-06,
      "loss": 0.0481,
      "step": 289750
    },
    {
      "epoch": 112.53,
      "learning_rate": 4.996245954692557e-06,
      "loss": 0.1156,
      "step": 289760
    },
    {
      "epoch": 112.53,
      "learning_rate": 4.995728155339806e-06,
      "loss": 0.0036,
      "step": 289770
    },
    {
      "epoch": 112.54,
      "learning_rate": 4.9952103559870556e-06,
      "loss": 0.1142,
      "step": 289780
    },
    {
      "epoch": 112.54,
      "learning_rate": 4.994692556634304e-06,
      "loss": 0.0588,
      "step": 289790
    },
    {
      "epoch": 112.54,
      "learning_rate": 4.994174757281554e-06,
      "loss": 0.0034,
      "step": 289800
    },
    {
      "epoch": 112.55,
      "learning_rate": 4.993656957928803e-06,
      "loss": 0.0021,
      "step": 289810
    },
    {
      "epoch": 112.55,
      "learning_rate": 4.9931391585760515e-06,
      "loss": 0.029,
      "step": 289820
    },
    {
      "epoch": 112.56,
      "learning_rate": 4.992621359223301e-06,
      "loss": 0.0025,
      "step": 289830
    },
    {
      "epoch": 112.56,
      "learning_rate": 4.992103559870551e-06,
      "loss": 0.0407,
      "step": 289840
    },
    {
      "epoch": 112.56,
      "learning_rate": 4.9915857605177995e-06,
      "loss": 0.0873,
      "step": 289850
    },
    {
      "epoch": 112.57,
      "learning_rate": 4.991067961165049e-06,
      "loss": 0.0117,
      "step": 289860
    },
    {
      "epoch": 112.57,
      "learning_rate": 4.990550161812298e-06,
      "loss": 0.0817,
      "step": 289870
    },
    {
      "epoch": 112.57,
      "learning_rate": 4.9900323624595475e-06,
      "loss": 0.0181,
      "step": 289880
    },
    {
      "epoch": 112.58,
      "learning_rate": 4.989514563106796e-06,
      "loss": 0.0002,
      "step": 289890
    },
    {
      "epoch": 112.58,
      "learning_rate": 4.988996763754046e-06,
      "loss": 0.0002,
      "step": 289900
    },
    {
      "epoch": 112.59,
      "learning_rate": 4.988478964401295e-06,
      "loss": 0.0601,
      "step": 289910
    },
    {
      "epoch": 112.59,
      "learning_rate": 4.987961165048544e-06,
      "loss": 0.1102,
      "step": 289920
    },
    {
      "epoch": 112.59,
      "learning_rate": 4.987443365695793e-06,
      "loss": 0.0599,
      "step": 289930
    },
    {
      "epoch": 112.6,
      "learning_rate": 4.986925566343043e-06,
      "loss": 0.1024,
      "step": 289940
    },
    {
      "epoch": 112.6,
      "learning_rate": 4.9864077669902914e-06,
      "loss": 0.0087,
      "step": 289950
    },
    {
      "epoch": 112.61,
      "learning_rate": 4.985889967637541e-06,
      "loss": 0.1288,
      "step": 289960
    },
    {
      "epoch": 112.61,
      "learning_rate": 4.98537216828479e-06,
      "loss": 0.0879,
      "step": 289970
    },
    {
      "epoch": 112.61,
      "learning_rate": 4.9848543689320395e-06,
      "loss": 0.0006,
      "step": 289980
    },
    {
      "epoch": 112.62,
      "learning_rate": 4.984336569579288e-06,
      "loss": 0.0578,
      "step": 289990
    },
    {
      "epoch": 112.62,
      "learning_rate": 4.983818770226538e-06,
      "loss": 0.1399,
      "step": 290000
    },
    {
      "epoch": 112.63,
      "learning_rate": 4.983300970873787e-06,
      "loss": 0.0993,
      "step": 290010
    },
    {
      "epoch": 112.63,
      "learning_rate": 4.982783171521036e-06,
      "loss": 0.0517,
      "step": 290020
    },
    {
      "epoch": 112.63,
      "learning_rate": 4.982265372168285e-06,
      "loss": 0.0197,
      "step": 290030
    },
    {
      "epoch": 112.64,
      "learning_rate": 4.981747572815535e-06,
      "loss": 0.0741,
      "step": 290040
    },
    {
      "epoch": 112.64,
      "learning_rate": 4.981229773462783e-06,
      "loss": 0.0003,
      "step": 290050
    },
    {
      "epoch": 112.64,
      "learning_rate": 4.980711974110033e-06,
      "loss": 0.0042,
      "step": 290060
    },
    {
      "epoch": 112.65,
      "learning_rate": 4.980194174757282e-06,
      "loss": 0.0589,
      "step": 290070
    },
    {
      "epoch": 112.65,
      "learning_rate": 4.979676375404531e-06,
      "loss": 0.0051,
      "step": 290080
    },
    {
      "epoch": 112.66,
      "learning_rate": 4.97915857605178e-06,
      "loss": 0.0536,
      "step": 290090
    },
    {
      "epoch": 112.66,
      "learning_rate": 4.97864077669903e-06,
      "loss": 0.0116,
      "step": 290100
    },
    {
      "epoch": 112.66,
      "learning_rate": 4.9781229773462786e-06,
      "loss": 0.0012,
      "step": 290110
    },
    {
      "epoch": 112.67,
      "learning_rate": 4.977605177993528e-06,
      "loss": 0.0978,
      "step": 290120
    },
    {
      "epoch": 112.67,
      "learning_rate": 4.977087378640777e-06,
      "loss": 0.0147,
      "step": 290130
    },
    {
      "epoch": 112.68,
      "learning_rate": 4.9765695792880266e-06,
      "loss": 0.1607,
      "step": 290140
    },
    {
      "epoch": 112.68,
      "learning_rate": 4.976051779935275e-06,
      "loss": 0.0756,
      "step": 290150
    },
    {
      "epoch": 112.68,
      "learning_rate": 4.975533980582525e-06,
      "loss": 0.0322,
      "step": 290160
    },
    {
      "epoch": 112.69,
      "learning_rate": 4.975016181229774e-06,
      "loss": 0.1384,
      "step": 290170
    },
    {
      "epoch": 112.69,
      "learning_rate": 4.974498381877023e-06,
      "loss": 0.087,
      "step": 290180
    },
    {
      "epoch": 112.7,
      "learning_rate": 4.973980582524272e-06,
      "loss": 0.0851,
      "step": 290190
    },
    {
      "epoch": 112.7,
      "learning_rate": 4.973462783171522e-06,
      "loss": 0.0231,
      "step": 290200
    },
    {
      "epoch": 112.7,
      "learning_rate": 4.9729449838187705e-06,
      "loss": 0.0818,
      "step": 290210
    },
    {
      "epoch": 112.71,
      "learning_rate": 4.97242718446602e-06,
      "loss": 0.0062,
      "step": 290220
    },
    {
      "epoch": 112.71,
      "learning_rate": 4.971909385113269e-06,
      "loss": 0.0266,
      "step": 290230
    },
    {
      "epoch": 112.71,
      "learning_rate": 4.9713915857605185e-06,
      "loss": 0.0014,
      "step": 290240
    },
    {
      "epoch": 112.72,
      "learning_rate": 4.970873786407767e-06,
      "loss": 0.0347,
      "step": 290250
    },
    {
      "epoch": 112.72,
      "learning_rate": 4.970355987055017e-06,
      "loss": 0.1288,
      "step": 290260
    },
    {
      "epoch": 112.73,
      "learning_rate": 4.969838187702266e-06,
      "loss": 0.0304,
      "step": 290270
    },
    {
      "epoch": 112.73,
      "learning_rate": 4.969320388349515e-06,
      "loss": 0.0325,
      "step": 290280
    },
    {
      "epoch": 112.73,
      "learning_rate": 4.968802588996764e-06,
      "loss": 0.1261,
      "step": 290290
    },
    {
      "epoch": 112.74,
      "learning_rate": 4.968284789644014e-06,
      "loss": 0.0967,
      "step": 290300
    },
    {
      "epoch": 112.74,
      "learning_rate": 4.9677669902912625e-06,
      "loss": 0.1161,
      "step": 290310
    },
    {
      "epoch": 112.75,
      "learning_rate": 4.967249190938512e-06,
      "loss": 0.0003,
      "step": 290320
    },
    {
      "epoch": 112.75,
      "learning_rate": 4.966731391585761e-06,
      "loss": 0.0001,
      "step": 290330
    },
    {
      "epoch": 112.75,
      "learning_rate": 4.9662135922330105e-06,
      "loss": 0.0284,
      "step": 290340
    },
    {
      "epoch": 112.76,
      "learning_rate": 4.965695792880259e-06,
      "loss": 0.1949,
      "step": 290350
    },
    {
      "epoch": 112.76,
      "learning_rate": 4.965177993527509e-06,
      "loss": 0.0066,
      "step": 290360
    },
    {
      "epoch": 112.77,
      "learning_rate": 4.964660194174758e-06,
      "loss": 0.0008,
      "step": 290370
    },
    {
      "epoch": 112.77,
      "learning_rate": 4.964142394822007e-06,
      "loss": 0.0578,
      "step": 290380
    },
    {
      "epoch": 112.77,
      "learning_rate": 4.963624595469256e-06,
      "loss": 0.1878,
      "step": 290390
    },
    {
      "epoch": 112.78,
      "learning_rate": 4.963106796116506e-06,
      "loss": 0.0329,
      "step": 290400
    },
    {
      "epoch": 112.78,
      "learning_rate": 4.962588996763754e-06,
      "loss": 0.0417,
      "step": 290410
    },
    {
      "epoch": 112.78,
      "learning_rate": 4.962071197411004e-06,
      "loss": 0.1084,
      "step": 290420
    },
    {
      "epoch": 112.79,
      "learning_rate": 4.961553398058253e-06,
      "loss": 0.0568,
      "step": 290430
    },
    {
      "epoch": 112.79,
      "learning_rate": 4.961035598705502e-06,
      "loss": 0.0232,
      "step": 290440
    },
    {
      "epoch": 112.8,
      "learning_rate": 4.960517799352751e-06,
      "loss": 0.0067,
      "step": 290450
    },
    {
      "epoch": 112.8,
      "learning_rate": 4.960000000000001e-06,
      "loss": 0.012,
      "step": 290460
    },
    {
      "epoch": 112.8,
      "learning_rate": 4.9594822006472496e-06,
      "loss": 0.0934,
      "step": 290470
    },
    {
      "epoch": 112.81,
      "learning_rate": 4.958964401294498e-06,
      "loss": 0.0194,
      "step": 290480
    },
    {
      "epoch": 112.81,
      "learning_rate": 4.958446601941748e-06,
      "loss": 0.0323,
      "step": 290490
    },
    {
      "epoch": 112.82,
      "learning_rate": 4.957928802588997e-06,
      "loss": 0.0228,
      "step": 290500
    },
    {
      "epoch": 112.82,
      "learning_rate": 4.957411003236246e-06,
      "loss": 0.0021,
      "step": 290510
    },
    {
      "epoch": 112.82,
      "learning_rate": 4.956893203883495e-06,
      "loss": 0.0508,
      "step": 290520
    },
    {
      "epoch": 112.83,
      "learning_rate": 4.956375404530745e-06,
      "loss": 0.05,
      "step": 290530
    },
    {
      "epoch": 112.83,
      "learning_rate": 4.9558576051779935e-06,
      "loss": 0.0445,
      "step": 290540
    },
    {
      "epoch": 112.83,
      "learning_rate": 4.955339805825243e-06,
      "loss": 0.1436,
      "step": 290550
    },
    {
      "epoch": 112.84,
      "learning_rate": 4.954822006472492e-06,
      "loss": 0.018,
      "step": 290560
    },
    {
      "epoch": 112.84,
      "learning_rate": 4.9543042071197415e-06,
      "loss": 0.0531,
      "step": 290570
    },
    {
      "epoch": 112.85,
      "learning_rate": 4.95378640776699e-06,
      "loss": 0.07,
      "step": 290580
    },
    {
      "epoch": 112.85,
      "learning_rate": 4.95326860841424e-06,
      "loss": 0.1176,
      "step": 290590
    },
    {
      "epoch": 112.85,
      "learning_rate": 4.952750809061489e-06,
      "loss": 0.0215,
      "step": 290600
    },
    {
      "epoch": 112.86,
      "learning_rate": 4.952233009708738e-06,
      "loss": 0.0001,
      "step": 290610
    },
    {
      "epoch": 112.86,
      "learning_rate": 4.951715210355987e-06,
      "loss": 0.0799,
      "step": 290620
    },
    {
      "epoch": 112.87,
      "learning_rate": 4.951197411003237e-06,
      "loss": 0.1248,
      "step": 290630
    },
    {
      "epoch": 112.87,
      "learning_rate": 4.9506796116504854e-06,
      "loss": 0.0028,
      "step": 290640
    },
    {
      "epoch": 112.87,
      "learning_rate": 4.950161812297735e-06,
      "loss": 0.0815,
      "step": 290650
    },
    {
      "epoch": 112.88,
      "learning_rate": 4.949644012944984e-06,
      "loss": 0.0527,
      "step": 290660
    },
    {
      "epoch": 112.88,
      "learning_rate": 4.9491262135922335e-06,
      "loss": 0.0317,
      "step": 290670
    },
    {
      "epoch": 112.89,
      "learning_rate": 4.948608414239482e-06,
      "loss": 0.0274,
      "step": 290680
    },
    {
      "epoch": 112.89,
      "learning_rate": 4.948090614886732e-06,
      "loss": 0.039,
      "step": 290690
    },
    {
      "epoch": 112.89,
      "learning_rate": 4.947572815533981e-06,
      "loss": 0.0859,
      "step": 290700
    },
    {
      "epoch": 112.9,
      "learning_rate": 4.94705501618123e-06,
      "loss": 0.0363,
      "step": 290710
    },
    {
      "epoch": 112.9,
      "learning_rate": 4.946537216828479e-06,
      "loss": 0.0733,
      "step": 290720
    },
    {
      "epoch": 112.9,
      "learning_rate": 4.946019417475729e-06,
      "loss": 0.0081,
      "step": 290730
    },
    {
      "epoch": 112.91,
      "learning_rate": 4.945501618122977e-06,
      "loss": 0.0224,
      "step": 290740
    },
    {
      "epoch": 112.91,
      "learning_rate": 4.944983818770227e-06,
      "loss": 0.0639,
      "step": 290750
    },
    {
      "epoch": 112.92,
      "learning_rate": 4.944466019417476e-06,
      "loss": 0.0366,
      "step": 290760
    },
    {
      "epoch": 112.92,
      "learning_rate": 4.943948220064725e-06,
      "loss": 0.0395,
      "step": 290770
    },
    {
      "epoch": 112.92,
      "learning_rate": 4.943430420711974e-06,
      "loss": 0.0396,
      "step": 290780
    },
    {
      "epoch": 112.93,
      "learning_rate": 4.942912621359224e-06,
      "loss": 0.0074,
      "step": 290790
    },
    {
      "epoch": 112.93,
      "learning_rate": 4.9423948220064726e-06,
      "loss": 0.0118,
      "step": 290800
    },
    {
      "epoch": 112.94,
      "learning_rate": 4.941877022653722e-06,
      "loss": 0.1131,
      "step": 290810
    },
    {
      "epoch": 112.94,
      "learning_rate": 4.941359223300971e-06,
      "loss": 0.0755,
      "step": 290820
    },
    {
      "epoch": 112.94,
      "learning_rate": 4.9408414239482206e-06,
      "loss": 0.0124,
      "step": 290830
    },
    {
      "epoch": 112.95,
      "learning_rate": 4.940323624595469e-06,
      "loss": 0.0125,
      "step": 290840
    },
    {
      "epoch": 112.95,
      "learning_rate": 4.939805825242719e-06,
      "loss": 0.014,
      "step": 290850
    },
    {
      "epoch": 112.96,
      "learning_rate": 4.939288025889968e-06,
      "loss": 0.0922,
      "step": 290860
    },
    {
      "epoch": 112.96,
      "learning_rate": 4.938770226537217e-06,
      "loss": 0.0247,
      "step": 290870
    },
    {
      "epoch": 112.96,
      "learning_rate": 4.938252427184466e-06,
      "loss": 0.0792,
      "step": 290880
    },
    {
      "epoch": 112.97,
      "learning_rate": 4.937734627831716e-06,
      "loss": 0.0608,
      "step": 290890
    },
    {
      "epoch": 112.97,
      "learning_rate": 4.9372168284789645e-06,
      "loss": 0.0791,
      "step": 290900
    },
    {
      "epoch": 112.97,
      "learning_rate": 4.936699029126214e-06,
      "loss": 0.0161,
      "step": 290910
    },
    {
      "epoch": 112.98,
      "learning_rate": 4.936181229773463e-06,
      "loss": 0.0962,
      "step": 290920
    },
    {
      "epoch": 112.98,
      "learning_rate": 4.9356634304207125e-06,
      "loss": 0.0445,
      "step": 290930
    },
    {
      "epoch": 112.99,
      "learning_rate": 4.935145631067961e-06,
      "loss": 0.0005,
      "step": 290940
    },
    {
      "epoch": 112.99,
      "learning_rate": 4.934627831715211e-06,
      "loss": 0.0416,
      "step": 290950
    },
    {
      "epoch": 112.99,
      "learning_rate": 4.9341100323624605e-06,
      "loss": 0.0347,
      "step": 290960
    },
    {
      "epoch": 113.0,
      "learning_rate": 4.933592233009709e-06,
      "loss": 0.0282,
      "step": 290970
    },
    {
      "epoch": 113.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.38432401418685913,
      "eval_runtime": 8.2159,
      "eval_samples_per_second": 442.437,
      "eval_steps_per_second": 55.381,
      "step": 290975
    },
    {
      "epoch": 113.0,
      "learning_rate": 4.933074433656959e-06,
      "loss": 0.0301,
      "step": 290980
    },
    {
      "epoch": 113.01,
      "learning_rate": 4.932556634304208e-06,
      "loss": 0.1055,
      "step": 290990
    },
    {
      "epoch": 113.01,
      "learning_rate": 4.932038834951457e-06,
      "loss": 0.0006,
      "step": 291000
    },
    {
      "epoch": 113.01,
      "learning_rate": 4.931521035598706e-06,
      "loss": 0.0918,
      "step": 291010
    },
    {
      "epoch": 113.02,
      "learning_rate": 4.931003236245956e-06,
      "loss": 0.1132,
      "step": 291020
    },
    {
      "epoch": 113.02,
      "learning_rate": 4.9304854368932045e-06,
      "loss": 0.0033,
      "step": 291030
    },
    {
      "epoch": 113.03,
      "learning_rate": 4.929967637540454e-06,
      "loss": 0.0304,
      "step": 291040
    },
    {
      "epoch": 113.03,
      "learning_rate": 4.929449838187703e-06,
      "loss": 0.081,
      "step": 291050
    },
    {
      "epoch": 113.03,
      "learning_rate": 4.9289320388349525e-06,
      "loss": 0.2264,
      "step": 291060
    },
    {
      "epoch": 113.04,
      "learning_rate": 4.928414239482201e-06,
      "loss": 0.0445,
      "step": 291070
    },
    {
      "epoch": 113.04,
      "learning_rate": 4.927896440129451e-06,
      "loss": 0.0697,
      "step": 291080
    },
    {
      "epoch": 113.04,
      "learning_rate": 4.9273786407767e-06,
      "loss": 0.0417,
      "step": 291090
    },
    {
      "epoch": 113.05,
      "learning_rate": 4.926860841423949e-06,
      "loss": 0.0043,
      "step": 291100
    },
    {
      "epoch": 113.05,
      "learning_rate": 4.926343042071198e-06,
      "loss": 0.0163,
      "step": 291110
    },
    {
      "epoch": 113.06,
      "learning_rate": 4.925825242718447e-06,
      "loss": 0.0691,
      "step": 291120
    },
    {
      "epoch": 113.06,
      "learning_rate": 4.925307443365696e-06,
      "loss": 0.0034,
      "step": 291130
    },
    {
      "epoch": 113.06,
      "learning_rate": 4.924789644012945e-06,
      "loss": 0.1025,
      "step": 291140
    },
    {
      "epoch": 113.07,
      "learning_rate": 4.924271844660195e-06,
      "loss": 0.0243,
      "step": 291150
    },
    {
      "epoch": 113.07,
      "learning_rate": 4.9237540453074436e-06,
      "loss": 0.1543,
      "step": 291160
    },
    {
      "epoch": 113.08,
      "learning_rate": 4.923236245954693e-06,
      "loss": 0.0021,
      "step": 291170
    },
    {
      "epoch": 113.08,
      "learning_rate": 4.922718446601942e-06,
      "loss": 0.0066,
      "step": 291180
    },
    {
      "epoch": 113.08,
      "learning_rate": 4.9222006472491916e-06,
      "loss": 0.0645,
      "step": 291190
    },
    {
      "epoch": 113.09,
      "learning_rate": 4.92168284789644e-06,
      "loss": 0.011,
      "step": 291200
    },
    {
      "epoch": 113.09,
      "learning_rate": 4.92116504854369e-06,
      "loss": 0.0248,
      "step": 291210
    },
    {
      "epoch": 113.1,
      "learning_rate": 4.920647249190939e-06,
      "loss": 0.1875,
      "step": 291220
    },
    {
      "epoch": 113.1,
      "learning_rate": 4.920129449838188e-06,
      "loss": 0.0269,
      "step": 291230
    },
    {
      "epoch": 113.1,
      "learning_rate": 4.919611650485437e-06,
      "loss": 0.0808,
      "step": 291240
    },
    {
      "epoch": 113.11,
      "learning_rate": 4.919093851132687e-06,
      "loss": 0.0259,
      "step": 291250
    },
    {
      "epoch": 113.11,
      "learning_rate": 4.9185760517799355e-06,
      "loss": 0.0738,
      "step": 291260
    },
    {
      "epoch": 113.11,
      "learning_rate": 4.918058252427185e-06,
      "loss": 0.0614,
      "step": 291270
    },
    {
      "epoch": 113.12,
      "learning_rate": 4.917540453074434e-06,
      "loss": 0.0081,
      "step": 291280
    },
    {
      "epoch": 113.12,
      "learning_rate": 4.917022653721683e-06,
      "loss": 0.0004,
      "step": 291290
    },
    {
      "epoch": 113.13,
      "learning_rate": 4.916504854368932e-06,
      "loss": 0.05,
      "step": 291300
    },
    {
      "epoch": 113.13,
      "learning_rate": 4.915987055016181e-06,
      "loss": 0.0805,
      "step": 291310
    },
    {
      "epoch": 113.13,
      "learning_rate": 4.915469255663431e-06,
      "loss": 0.0421,
      "step": 291320
    },
    {
      "epoch": 113.14,
      "learning_rate": 4.9149514563106794e-06,
      "loss": 0.0459,
      "step": 291330
    },
    {
      "epoch": 113.14,
      "learning_rate": 4.914433656957929e-06,
      "loss": 0.1232,
      "step": 291340
    },
    {
      "epoch": 113.15,
      "learning_rate": 4.913915857605178e-06,
      "loss": 0.0513,
      "step": 291350
    },
    {
      "epoch": 113.15,
      "learning_rate": 4.9133980582524275e-06,
      "loss": 0.0379,
      "step": 291360
    },
    {
      "epoch": 113.15,
      "learning_rate": 4.912880258899676e-06,
      "loss": 0.0222,
      "step": 291370
    },
    {
      "epoch": 113.16,
      "learning_rate": 4.912362459546926e-06,
      "loss": 0.0359,
      "step": 291380
    },
    {
      "epoch": 113.16,
      "learning_rate": 4.911844660194175e-06,
      "loss": 0.0156,
      "step": 291390
    },
    {
      "epoch": 113.17,
      "learning_rate": 4.911326860841424e-06,
      "loss": 0.0007,
      "step": 291400
    },
    {
      "epoch": 113.17,
      "learning_rate": 4.910809061488673e-06,
      "loss": 0.0118,
      "step": 291410
    },
    {
      "epoch": 113.17,
      "learning_rate": 4.910291262135923e-06,
      "loss": 0.001,
      "step": 291420
    },
    {
      "epoch": 113.18,
      "learning_rate": 4.909773462783171e-06,
      "loss": 0.0352,
      "step": 291430
    },
    {
      "epoch": 113.18,
      "learning_rate": 4.909255663430421e-06,
      "loss": 0.0461,
      "step": 291440
    },
    {
      "epoch": 113.18,
      "learning_rate": 4.90873786407767e-06,
      "loss": 0.0259,
      "step": 291450
    },
    {
      "epoch": 113.19,
      "learning_rate": 4.908220064724919e-06,
      "loss": 0.0162,
      "step": 291460
    },
    {
      "epoch": 113.19,
      "learning_rate": 4.907702265372168e-06,
      "loss": 0.0976,
      "step": 291470
    },
    {
      "epoch": 113.2,
      "learning_rate": 4.907184466019418e-06,
      "loss": 0.0192,
      "step": 291480
    },
    {
      "epoch": 113.2,
      "learning_rate": 4.9066666666666666e-06,
      "loss": 0.0177,
      "step": 291490
    },
    {
      "epoch": 113.2,
      "learning_rate": 4.906148867313916e-06,
      "loss": 0.114,
      "step": 291500
    },
    {
      "epoch": 113.21,
      "learning_rate": 4.905631067961166e-06,
      "loss": 0.0179,
      "step": 291510
    },
    {
      "epoch": 113.21,
      "learning_rate": 4.9051132686084146e-06,
      "loss": 0.0141,
      "step": 291520
    },
    {
      "epoch": 113.22,
      "learning_rate": 4.904595469255664e-06,
      "loss": 0.0362,
      "step": 291530
    },
    {
      "epoch": 113.22,
      "learning_rate": 4.904077669902913e-06,
      "loss": 0.1245,
      "step": 291540
    },
    {
      "epoch": 113.22,
      "learning_rate": 4.9035598705501626e-06,
      "loss": 0.0002,
      "step": 291550
    },
    {
      "epoch": 113.23,
      "learning_rate": 4.903042071197411e-06,
      "loss": 0.0173,
      "step": 291560
    },
    {
      "epoch": 113.23,
      "learning_rate": 4.902524271844661e-06,
      "loss": 0.015,
      "step": 291570
    },
    {
      "epoch": 113.23,
      "learning_rate": 4.90200647249191e-06,
      "loss": 0.0114,
      "step": 291580
    },
    {
      "epoch": 113.24,
      "learning_rate": 4.901488673139159e-06,
      "loss": 0.0218,
      "step": 291590
    },
    {
      "epoch": 113.24,
      "learning_rate": 4.900970873786408e-06,
      "loss": 0.0612,
      "step": 291600
    },
    {
      "epoch": 113.25,
      "learning_rate": 4.900453074433658e-06,
      "loss": 0.0798,
      "step": 291610
    },
    {
      "epoch": 113.25,
      "learning_rate": 4.8999352750809065e-06,
      "loss": 0.031,
      "step": 291620
    },
    {
      "epoch": 113.25,
      "learning_rate": 4.899417475728156e-06,
      "loss": 0.0233,
      "step": 291630
    },
    {
      "epoch": 113.26,
      "learning_rate": 4.898899676375405e-06,
      "loss": 0.0527,
      "step": 291640
    },
    {
      "epoch": 113.26,
      "learning_rate": 4.8983818770226545e-06,
      "loss": 0.0002,
      "step": 291650
    },
    {
      "epoch": 113.27,
      "learning_rate": 4.897864077669903e-06,
      "loss": 0.0926,
      "step": 291660
    },
    {
      "epoch": 113.27,
      "learning_rate": 4.897346278317153e-06,
      "loss": 0.0375,
      "step": 291670
    },
    {
      "epoch": 113.27,
      "learning_rate": 4.896828478964402e-06,
      "loss": 0.0468,
      "step": 291680
    },
    {
      "epoch": 113.28,
      "learning_rate": 4.896310679611651e-06,
      "loss": 0.0174,
      "step": 291690
    },
    {
      "epoch": 113.28,
      "learning_rate": 4.8957928802589e-06,
      "loss": 0.0627,
      "step": 291700
    },
    {
      "epoch": 113.29,
      "learning_rate": 4.89527508090615e-06,
      "loss": 0.1028,
      "step": 291710
    },
    {
      "epoch": 113.29,
      "learning_rate": 4.8947572815533985e-06,
      "loss": 0.0792,
      "step": 291720
    },
    {
      "epoch": 113.29,
      "learning_rate": 4.894239482200648e-06,
      "loss": 0.0078,
      "step": 291730
    },
    {
      "epoch": 113.3,
      "learning_rate": 4.893721682847897e-06,
      "loss": 0.0782,
      "step": 291740
    },
    {
      "epoch": 113.3,
      "learning_rate": 4.8932038834951465e-06,
      "loss": 0.0984,
      "step": 291750
    },
    {
      "epoch": 113.3,
      "learning_rate": 4.892686084142395e-06,
      "loss": 0.0111,
      "step": 291760
    },
    {
      "epoch": 113.31,
      "learning_rate": 4.892168284789645e-06,
      "loss": 0.0317,
      "step": 291770
    },
    {
      "epoch": 113.31,
      "learning_rate": 4.891650485436894e-06,
      "loss": 0.0317,
      "step": 291780
    },
    {
      "epoch": 113.32,
      "learning_rate": 4.891132686084143e-06,
      "loss": 0.0744,
      "step": 291790
    },
    {
      "epoch": 113.32,
      "learning_rate": 4.890614886731392e-06,
      "loss": 0.1261,
      "step": 291800
    },
    {
      "epoch": 113.32,
      "learning_rate": 4.890097087378642e-06,
      "loss": 0.0529,
      "step": 291810
    },
    {
      "epoch": 113.33,
      "learning_rate": 4.88957928802589e-06,
      "loss": 0.0414,
      "step": 291820
    },
    {
      "epoch": 113.33,
      "learning_rate": 4.88906148867314e-06,
      "loss": 0.0337,
      "step": 291830
    },
    {
      "epoch": 113.34,
      "learning_rate": 4.888543689320389e-06,
      "loss": 0.026,
      "step": 291840
    },
    {
      "epoch": 113.34,
      "learning_rate": 4.888025889967638e-06,
      "loss": 0.061,
      "step": 291850
    },
    {
      "epoch": 113.34,
      "learning_rate": 4.887508090614887e-06,
      "loss": 0.0325,
      "step": 291860
    },
    {
      "epoch": 113.35,
      "learning_rate": 4.886990291262137e-06,
      "loss": 0.0115,
      "step": 291870
    },
    {
      "epoch": 113.35,
      "learning_rate": 4.8864724919093856e-06,
      "loss": 0.1056,
      "step": 291880
    },
    {
      "epoch": 113.36,
      "learning_rate": 4.885954692556635e-06,
      "loss": 0.1041,
      "step": 291890
    },
    {
      "epoch": 113.36,
      "learning_rate": 4.885436893203884e-06,
      "loss": 0.0007,
      "step": 291900
    },
    {
      "epoch": 113.36,
      "learning_rate": 4.8849190938511336e-06,
      "loss": 0.0079,
      "step": 291910
    },
    {
      "epoch": 113.37,
      "learning_rate": 4.884401294498382e-06,
      "loss": 0.2211,
      "step": 291920
    },
    {
      "epoch": 113.37,
      "learning_rate": 4.883883495145631e-06,
      "loss": 0.0007,
      "step": 291930
    },
    {
      "epoch": 113.37,
      "learning_rate": 4.883365695792881e-06,
      "loss": 0.0087,
      "step": 291940
    },
    {
      "epoch": 113.38,
      "learning_rate": 4.8828478964401295e-06,
      "loss": 0.029,
      "step": 291950
    },
    {
      "epoch": 113.38,
      "learning_rate": 4.882330097087379e-06,
      "loss": 0.0925,
      "step": 291960
    },
    {
      "epoch": 113.39,
      "learning_rate": 4.881812297734628e-06,
      "loss": 0.0002,
      "step": 291970
    },
    {
      "epoch": 113.39,
      "learning_rate": 4.8812944983818775e-06,
      "loss": 0.0784,
      "step": 291980
    },
    {
      "epoch": 113.39,
      "learning_rate": 4.880776699029126e-06,
      "loss": 0.0496,
      "step": 291990
    },
    {
      "epoch": 113.4,
      "learning_rate": 4.880258899676376e-06,
      "loss": 0.0505,
      "step": 292000
    },
    {
      "epoch": 113.4,
      "learning_rate": 4.879741100323625e-06,
      "loss": 0.051,
      "step": 292010
    },
    {
      "epoch": 113.41,
      "learning_rate": 4.879223300970874e-06,
      "loss": 0.0275,
      "step": 292020
    },
    {
      "epoch": 113.41,
      "learning_rate": 4.878705501618123e-06,
      "loss": 0.0779,
      "step": 292030
    },
    {
      "epoch": 113.41,
      "learning_rate": 4.878187702265373e-06,
      "loss": 0.0763,
      "step": 292040
    },
    {
      "epoch": 113.42,
      "learning_rate": 4.8776699029126215e-06,
      "loss": 0.0238,
      "step": 292050
    },
    {
      "epoch": 113.42,
      "learning_rate": 4.877152103559871e-06,
      "loss": 0.0159,
      "step": 292060
    },
    {
      "epoch": 113.43,
      "learning_rate": 4.87663430420712e-06,
      "loss": 0.087,
      "step": 292070
    },
    {
      "epoch": 113.43,
      "learning_rate": 4.8761165048543695e-06,
      "loss": 0.0351,
      "step": 292080
    },
    {
      "epoch": 113.43,
      "learning_rate": 4.875598705501618e-06,
      "loss": 0.0097,
      "step": 292090
    },
    {
      "epoch": 113.44,
      "learning_rate": 4.875080906148868e-06,
      "loss": 0.0544,
      "step": 292100
    },
    {
      "epoch": 113.44,
      "learning_rate": 4.874563106796117e-06,
      "loss": 0.1541,
      "step": 292110
    },
    {
      "epoch": 113.44,
      "learning_rate": 4.874045307443366e-06,
      "loss": 0.0871,
      "step": 292120
    },
    {
      "epoch": 113.45,
      "learning_rate": 4.873527508090615e-06,
      "loss": 0.1023,
      "step": 292130
    },
    {
      "epoch": 113.45,
      "learning_rate": 4.873009708737865e-06,
      "loss": 0.1073,
      "step": 292140
    },
    {
      "epoch": 113.46,
      "learning_rate": 4.872491909385113e-06,
      "loss": 0.0174,
      "step": 292150
    },
    {
      "epoch": 113.46,
      "learning_rate": 4.871974110032363e-06,
      "loss": 0.0507,
      "step": 292160
    },
    {
      "epoch": 113.46,
      "learning_rate": 4.871456310679612e-06,
      "loss": 0.0056,
      "step": 292170
    },
    {
      "epoch": 113.47,
      "learning_rate": 4.870938511326861e-06,
      "loss": 0.0645,
      "step": 292180
    },
    {
      "epoch": 113.47,
      "learning_rate": 4.87042071197411e-06,
      "loss": 0.0494,
      "step": 292190
    },
    {
      "epoch": 113.48,
      "learning_rate": 4.86990291262136e-06,
      "loss": 0.023,
      "step": 292200
    },
    {
      "epoch": 113.48,
      "learning_rate": 4.8693851132686086e-06,
      "loss": 0.056,
      "step": 292210
    },
    {
      "epoch": 113.48,
      "learning_rate": 4.868867313915858e-06,
      "loss": 0.0168,
      "step": 292220
    },
    {
      "epoch": 113.49,
      "learning_rate": 4.868349514563107e-06,
      "loss": 0.0833,
      "step": 292230
    },
    {
      "epoch": 113.49,
      "learning_rate": 4.8678317152103566e-06,
      "loss": 0.0872,
      "step": 292240
    },
    {
      "epoch": 113.5,
      "learning_rate": 4.867313915857605e-06,
      "loss": 0.0181,
      "step": 292250
    },
    {
      "epoch": 113.5,
      "learning_rate": 4.866796116504855e-06,
      "loss": 0.0787,
      "step": 292260
    },
    {
      "epoch": 113.5,
      "learning_rate": 4.866278317152104e-06,
      "loss": 0.0094,
      "step": 292270
    },
    {
      "epoch": 113.51,
      "learning_rate": 4.865760517799353e-06,
      "loss": 0.0276,
      "step": 292280
    },
    {
      "epoch": 113.51,
      "learning_rate": 4.865242718446602e-06,
      "loss": 0.0616,
      "step": 292290
    },
    {
      "epoch": 113.51,
      "learning_rate": 4.864724919093852e-06,
      "loss": 0.0225,
      "step": 292300
    },
    {
      "epoch": 113.52,
      "learning_rate": 4.8642071197411005e-06,
      "loss": 0.1236,
      "step": 292310
    },
    {
      "epoch": 113.52,
      "learning_rate": 4.86368932038835e-06,
      "loss": 0.0044,
      "step": 292320
    },
    {
      "epoch": 113.53,
      "learning_rate": 4.863171521035599e-06,
      "loss": 0.0136,
      "step": 292330
    },
    {
      "epoch": 113.53,
      "learning_rate": 4.8626537216828485e-06,
      "loss": 0.059,
      "step": 292340
    },
    {
      "epoch": 113.53,
      "learning_rate": 4.862135922330097e-06,
      "loss": 0.0137,
      "step": 292350
    },
    {
      "epoch": 113.54,
      "learning_rate": 4.861618122977347e-06,
      "loss": 0.0286,
      "step": 292360
    },
    {
      "epoch": 113.54,
      "learning_rate": 4.861100323624596e-06,
      "loss": 0.1999,
      "step": 292370
    },
    {
      "epoch": 113.55,
      "learning_rate": 4.860582524271845e-06,
      "loss": 0.1766,
      "step": 292380
    },
    {
      "epoch": 113.55,
      "learning_rate": 4.860064724919094e-06,
      "loss": 0.0995,
      "step": 292390
    },
    {
      "epoch": 113.55,
      "learning_rate": 4.859546925566344e-06,
      "loss": 0.0136,
      "step": 292400
    },
    {
      "epoch": 113.56,
      "learning_rate": 4.8590291262135925e-06,
      "loss": 0.0839,
      "step": 292410
    },
    {
      "epoch": 113.56,
      "learning_rate": 4.858511326860842e-06,
      "loss": 0.0009,
      "step": 292420
    },
    {
      "epoch": 113.57,
      "learning_rate": 4.857993527508091e-06,
      "loss": 0.1132,
      "step": 292430
    },
    {
      "epoch": 113.57,
      "learning_rate": 4.8574757281553405e-06,
      "loss": 0.0565,
      "step": 292440
    },
    {
      "epoch": 113.57,
      "learning_rate": 4.856957928802589e-06,
      "loss": 0.0077,
      "step": 292450
    },
    {
      "epoch": 113.58,
      "learning_rate": 4.856440129449839e-06,
      "loss": 0.0229,
      "step": 292460
    },
    {
      "epoch": 113.58,
      "learning_rate": 4.855922330097088e-06,
      "loss": 0.0278,
      "step": 292470
    },
    {
      "epoch": 113.58,
      "learning_rate": 4.855404530744337e-06,
      "loss": 0.0001,
      "step": 292480
    },
    {
      "epoch": 113.59,
      "learning_rate": 4.854886731391586e-06,
      "loss": 0.0266,
      "step": 292490
    },
    {
      "epoch": 113.59,
      "learning_rate": 4.854368932038836e-06,
      "loss": 0.0178,
      "step": 292500
    },
    {
      "epoch": 113.6,
      "learning_rate": 4.853851132686084e-06,
      "loss": 0.1188,
      "step": 292510
    },
    {
      "epoch": 113.6,
      "learning_rate": 4.853333333333334e-06,
      "loss": 0.0654,
      "step": 292520
    },
    {
      "epoch": 113.6,
      "learning_rate": 4.852815533980583e-06,
      "loss": 0.1424,
      "step": 292530
    },
    {
      "epoch": 113.61,
      "learning_rate": 4.852297734627832e-06,
      "loss": 0.0099,
      "step": 292540
    },
    {
      "epoch": 113.61,
      "learning_rate": 4.851779935275081e-06,
      "loss": 0.0482,
      "step": 292550
    },
    {
      "epoch": 113.62,
      "learning_rate": 4.851262135922331e-06,
      "loss": 0.0162,
      "step": 292560
    },
    {
      "epoch": 113.62,
      "learning_rate": 4.8507443365695796e-06,
      "loss": 0.0725,
      "step": 292570
    },
    {
      "epoch": 113.62,
      "learning_rate": 4.850226537216829e-06,
      "loss": 0.0894,
      "step": 292580
    },
    {
      "epoch": 113.63,
      "learning_rate": 4.849708737864078e-06,
      "loss": 0.0584,
      "step": 292590
    },
    {
      "epoch": 113.63,
      "learning_rate": 4.8491909385113276e-06,
      "loss": 0.0537,
      "step": 292600
    },
    {
      "epoch": 113.63,
      "learning_rate": 4.848673139158576e-06,
      "loss": 0.0017,
      "step": 292610
    },
    {
      "epoch": 113.64,
      "learning_rate": 4.848155339805826e-06,
      "loss": 0.035,
      "step": 292620
    },
    {
      "epoch": 113.64,
      "learning_rate": 4.847637540453075e-06,
      "loss": 0.004,
      "step": 292630
    },
    {
      "epoch": 113.65,
      "learning_rate": 4.847119741100324e-06,
      "loss": 0.1286,
      "step": 292640
    },
    {
      "epoch": 113.65,
      "learning_rate": 4.846601941747573e-06,
      "loss": 0.1134,
      "step": 292650
    },
    {
      "epoch": 113.65,
      "learning_rate": 4.846084142394823e-06,
      "loss": 0.0107,
      "step": 292660
    },
    {
      "epoch": 113.66,
      "learning_rate": 4.8455663430420715e-06,
      "loss": 0.0265,
      "step": 292670
    },
    {
      "epoch": 113.66,
      "learning_rate": 4.845048543689321e-06,
      "loss": 0.0517,
      "step": 292680
    },
    {
      "epoch": 113.67,
      "learning_rate": 4.84453074433657e-06,
      "loss": 0.0934,
      "step": 292690
    },
    {
      "epoch": 113.67,
      "learning_rate": 4.8440129449838195e-06,
      "loss": 0.2096,
      "step": 292700
    },
    {
      "epoch": 113.67,
      "learning_rate": 4.843495145631068e-06,
      "loss": 0.1428,
      "step": 292710
    },
    {
      "epoch": 113.68,
      "learning_rate": 4.842977346278318e-06,
      "loss": 0.0763,
      "step": 292720
    },
    {
      "epoch": 113.68,
      "learning_rate": 4.842459546925567e-06,
      "loss": 0.0886,
      "step": 292730
    },
    {
      "epoch": 113.69,
      "learning_rate": 4.841941747572816e-06,
      "loss": 0.0816,
      "step": 292740
    },
    {
      "epoch": 113.69,
      "learning_rate": 4.841423948220065e-06,
      "loss": 0.0016,
      "step": 292750
    },
    {
      "epoch": 113.69,
      "learning_rate": 4.840906148867314e-06,
      "loss": 0.0567,
      "step": 292760
    },
    {
      "epoch": 113.7,
      "learning_rate": 4.8403883495145635e-06,
      "loss": 0.0002,
      "step": 292770
    },
    {
      "epoch": 113.7,
      "learning_rate": 4.839870550161812e-06,
      "loss": 0.0617,
      "step": 292780
    },
    {
      "epoch": 113.7,
      "learning_rate": 4.839352750809062e-06,
      "loss": 0.0778,
      "step": 292790
    },
    {
      "epoch": 113.71,
      "learning_rate": 4.838834951456311e-06,
      "loss": 0.0559,
      "step": 292800
    },
    {
      "epoch": 113.71,
      "learning_rate": 4.83831715210356e-06,
      "loss": 0.003,
      "step": 292810
    },
    {
      "epoch": 113.72,
      "learning_rate": 4.837799352750809e-06,
      "loss": 0.0805,
      "step": 292820
    },
    {
      "epoch": 113.72,
      "learning_rate": 4.837281553398059e-06,
      "loss": 0.1329,
      "step": 292830
    },
    {
      "epoch": 113.72,
      "learning_rate": 4.836763754045307e-06,
      "loss": 0.0389,
      "step": 292840
    },
    {
      "epoch": 113.73,
      "learning_rate": 4.836245954692557e-06,
      "loss": 0.0955,
      "step": 292850
    },
    {
      "epoch": 113.73,
      "learning_rate": 4.835728155339806e-06,
      "loss": 0.0201,
      "step": 292860
    },
    {
      "epoch": 113.74,
      "learning_rate": 4.835210355987055e-06,
      "loss": 0.0908,
      "step": 292870
    },
    {
      "epoch": 113.74,
      "learning_rate": 4.834692556634304e-06,
      "loss": 0.0007,
      "step": 292880
    },
    {
      "epoch": 113.74,
      "learning_rate": 4.834174757281554e-06,
      "loss": 0.0011,
      "step": 292890
    },
    {
      "epoch": 113.75,
      "learning_rate": 4.8336569579288026e-06,
      "loss": 0.0899,
      "step": 292900
    },
    {
      "epoch": 113.75,
      "learning_rate": 4.833139158576052e-06,
      "loss": 0.007,
      "step": 292910
    },
    {
      "epoch": 113.76,
      "learning_rate": 4.832621359223301e-06,
      "loss": 0.0214,
      "step": 292920
    },
    {
      "epoch": 113.76,
      "learning_rate": 4.8321035598705506e-06,
      "loss": 0.187,
      "step": 292930
    },
    {
      "epoch": 113.76,
      "learning_rate": 4.831585760517799e-06,
      "loss": 0.01,
      "step": 292940
    },
    {
      "epoch": 113.77,
      "learning_rate": 4.831067961165049e-06,
      "loss": 0.0255,
      "step": 292950
    },
    {
      "epoch": 113.77,
      "learning_rate": 4.830550161812298e-06,
      "loss": 0.0273,
      "step": 292960
    },
    {
      "epoch": 113.77,
      "learning_rate": 4.830032362459547e-06,
      "loss": 0.0419,
      "step": 292970
    },
    {
      "epoch": 113.78,
      "learning_rate": 4.829514563106796e-06,
      "loss": 0.0008,
      "step": 292980
    },
    {
      "epoch": 113.78,
      "learning_rate": 4.828996763754046e-06,
      "loss": 0.0256,
      "step": 292990
    },
    {
      "epoch": 113.79,
      "learning_rate": 4.8284789644012945e-06,
      "loss": 0.0813,
      "step": 293000
    },
    {
      "epoch": 113.79,
      "learning_rate": 4.827961165048544e-06,
      "loss": 0.0012,
      "step": 293010
    },
    {
      "epoch": 113.79,
      "learning_rate": 4.827443365695793e-06,
      "loss": 0.0149,
      "step": 293020
    },
    {
      "epoch": 113.8,
      "learning_rate": 4.8269255663430425e-06,
      "loss": 0.0186,
      "step": 293030
    },
    {
      "epoch": 113.8,
      "learning_rate": 4.826407766990291e-06,
      "loss": 0.0002,
      "step": 293040
    },
    {
      "epoch": 113.81,
      "learning_rate": 4.825889967637541e-06,
      "loss": 0.0242,
      "step": 293050
    },
    {
      "epoch": 113.81,
      "learning_rate": 4.82537216828479e-06,
      "loss": 0.0497,
      "step": 293060
    },
    {
      "epoch": 113.81,
      "learning_rate": 4.824854368932039e-06,
      "loss": 0.0607,
      "step": 293070
    },
    {
      "epoch": 113.82,
      "learning_rate": 4.824336569579288e-06,
      "loss": 0.0655,
      "step": 293080
    },
    {
      "epoch": 113.82,
      "learning_rate": 4.823818770226538e-06,
      "loss": 0.0004,
      "step": 293090
    },
    {
      "epoch": 113.83,
      "learning_rate": 4.8233009708737865e-06,
      "loss": 0.0599,
      "step": 293100
    },
    {
      "epoch": 113.83,
      "learning_rate": 4.822783171521036e-06,
      "loss": 0.026,
      "step": 293110
    },
    {
      "epoch": 113.83,
      "learning_rate": 4.822265372168285e-06,
      "loss": 0.0093,
      "step": 293120
    },
    {
      "epoch": 113.84,
      "learning_rate": 4.8217475728155345e-06,
      "loss": 0.0299,
      "step": 293130
    },
    {
      "epoch": 113.84,
      "learning_rate": 4.821229773462783e-06,
      "loss": 0.0252,
      "step": 293140
    },
    {
      "epoch": 113.84,
      "learning_rate": 4.820711974110033e-06,
      "loss": 0.0337,
      "step": 293150
    },
    {
      "epoch": 113.85,
      "learning_rate": 4.820194174757282e-06,
      "loss": 0.0171,
      "step": 293160
    },
    {
      "epoch": 113.85,
      "learning_rate": 4.819676375404531e-06,
      "loss": 0.0545,
      "step": 293170
    },
    {
      "epoch": 113.86,
      "learning_rate": 4.81915857605178e-06,
      "loss": 0.0236,
      "step": 293180
    },
    {
      "epoch": 113.86,
      "learning_rate": 4.81864077669903e-06,
      "loss": 0.2279,
      "step": 293190
    },
    {
      "epoch": 113.86,
      "learning_rate": 4.818122977346279e-06,
      "loss": 0.103,
      "step": 293200
    },
    {
      "epoch": 113.87,
      "learning_rate": 4.817605177993528e-06,
      "loss": 0.0013,
      "step": 293210
    },
    {
      "epoch": 113.87,
      "learning_rate": 4.817087378640778e-06,
      "loss": 0.0855,
      "step": 293220
    },
    {
      "epoch": 113.88,
      "learning_rate": 4.816569579288026e-06,
      "loss": 0.0008,
      "step": 293230
    },
    {
      "epoch": 113.88,
      "learning_rate": 4.816051779935276e-06,
      "loss": 0.0001,
      "step": 293240
    },
    {
      "epoch": 113.88,
      "learning_rate": 4.815533980582525e-06,
      "loss": 0.1217,
      "step": 293250
    },
    {
      "epoch": 113.89,
      "learning_rate": 4.815016181229774e-06,
      "loss": 0.0383,
      "step": 293260
    },
    {
      "epoch": 113.89,
      "learning_rate": 4.814498381877023e-06,
      "loss": 0.0222,
      "step": 293270
    },
    {
      "epoch": 113.9,
      "learning_rate": 4.813980582524273e-06,
      "loss": 0.0817,
      "step": 293280
    },
    {
      "epoch": 113.9,
      "learning_rate": 4.8134627831715216e-06,
      "loss": 0.0021,
      "step": 293290
    },
    {
      "epoch": 113.9,
      "learning_rate": 4.812944983818771e-06,
      "loss": 0.1562,
      "step": 293300
    },
    {
      "epoch": 113.91,
      "learning_rate": 4.81242718446602e-06,
      "loss": 0.068,
      "step": 293310
    },
    {
      "epoch": 113.91,
      "learning_rate": 4.81190938511327e-06,
      "loss": 0.0605,
      "step": 293320
    },
    {
      "epoch": 113.91,
      "learning_rate": 4.811391585760518e-06,
      "loss": 0.055,
      "step": 293330
    },
    {
      "epoch": 113.92,
      "learning_rate": 4.810873786407768e-06,
      "loss": 0.0085,
      "step": 293340
    },
    {
      "epoch": 113.92,
      "learning_rate": 4.810355987055017e-06,
      "loss": 0.0003,
      "step": 293350
    },
    {
      "epoch": 113.93,
      "learning_rate": 4.809838187702266e-06,
      "loss": 0.1281,
      "step": 293360
    },
    {
      "epoch": 113.93,
      "learning_rate": 4.809320388349515e-06,
      "loss": 0.252,
      "step": 293370
    },
    {
      "epoch": 113.93,
      "learning_rate": 4.808802588996765e-06,
      "loss": 0.0896,
      "step": 293380
    },
    {
      "epoch": 113.94,
      "learning_rate": 4.8082847896440135e-06,
      "loss": 0.1008,
      "step": 293390
    },
    {
      "epoch": 113.94,
      "learning_rate": 4.807766990291262e-06,
      "loss": 0.0029,
      "step": 293400
    },
    {
      "epoch": 113.95,
      "learning_rate": 4.807249190938512e-06,
      "loss": 0.01,
      "step": 293410
    },
    {
      "epoch": 113.95,
      "learning_rate": 4.806731391585761e-06,
      "loss": 0.001,
      "step": 293420
    },
    {
      "epoch": 113.95,
      "learning_rate": 4.80621359223301e-06,
      "loss": 0.0002,
      "step": 293430
    },
    {
      "epoch": 113.96,
      "learning_rate": 4.805695792880259e-06,
      "loss": 0.0353,
      "step": 293440
    },
    {
      "epoch": 113.96,
      "learning_rate": 4.805177993527509e-06,
      "loss": 0.0329,
      "step": 293450
    },
    {
      "epoch": 113.97,
      "learning_rate": 4.8046601941747575e-06,
      "loss": 0.0432,
      "step": 293460
    },
    {
      "epoch": 113.97,
      "learning_rate": 4.804142394822007e-06,
      "loss": 0.0234,
      "step": 293470
    },
    {
      "epoch": 113.97,
      "learning_rate": 4.803624595469256e-06,
      "loss": 0.0789,
      "step": 293480
    },
    {
      "epoch": 113.98,
      "learning_rate": 4.8031067961165055e-06,
      "loss": 0.0384,
      "step": 293490
    },
    {
      "epoch": 113.98,
      "learning_rate": 4.802588996763754e-06,
      "loss": 0.0238,
      "step": 293500
    },
    {
      "epoch": 113.98,
      "learning_rate": 4.802071197411004e-06,
      "loss": 0.0455,
      "step": 293510
    },
    {
      "epoch": 113.99,
      "learning_rate": 4.801553398058253e-06,
      "loss": 0.0003,
      "step": 293520
    },
    {
      "epoch": 113.99,
      "learning_rate": 4.801035598705502e-06,
      "loss": 0.0133,
      "step": 293530
    },
    {
      "epoch": 114.0,
      "learning_rate": 4.800517799352751e-06,
      "loss": 0.1137,
      "step": 293540
    },
    {
      "epoch": 114.0,
      "learning_rate": 4.800000000000001e-06,
      "loss": 0.1117,
      "step": 293550
    },
    {
      "epoch": 114.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3789585828781128,
      "eval_runtime": 8.1818,
      "eval_samples_per_second": 444.276,
      "eval_steps_per_second": 55.611,
      "step": 293550
    },
    {
      "epoch": 114.0,
      "learning_rate": 4.799482200647249e-06,
      "loss": 0.023,
      "step": 293560
    },
    {
      "epoch": 114.01,
      "learning_rate": 4.798964401294498e-06,
      "loss": 0.0116,
      "step": 293570
    },
    {
      "epoch": 114.01,
      "learning_rate": 4.798446601941748e-06,
      "loss": 0.0012,
      "step": 293580
    },
    {
      "epoch": 114.02,
      "learning_rate": 4.7979288025889966e-06,
      "loss": 0.0399,
      "step": 293590
    },
    {
      "epoch": 114.02,
      "learning_rate": 4.797411003236246e-06,
      "loss": 0.1014,
      "step": 293600
    },
    {
      "epoch": 114.02,
      "learning_rate": 4.796893203883495e-06,
      "loss": 0.0733,
      "step": 293610
    },
    {
      "epoch": 114.03,
      "learning_rate": 4.7963754045307446e-06,
      "loss": 0.0312,
      "step": 293620
    },
    {
      "epoch": 114.03,
      "learning_rate": 4.795857605177993e-06,
      "loss": 0.008,
      "step": 293630
    },
    {
      "epoch": 114.03,
      "learning_rate": 4.795339805825243e-06,
      "loss": 0.0444,
      "step": 293640
    },
    {
      "epoch": 114.04,
      "learning_rate": 4.794822006472492e-06,
      "loss": 0.0082,
      "step": 293650
    },
    {
      "epoch": 114.04,
      "learning_rate": 4.794304207119741e-06,
      "loss": 0.0784,
      "step": 293660
    },
    {
      "epoch": 114.05,
      "learning_rate": 4.79378640776699e-06,
      "loss": 0.0107,
      "step": 293670
    },
    {
      "epoch": 114.05,
      "learning_rate": 4.79326860841424e-06,
      "loss": 0.1249,
      "step": 293680
    },
    {
      "epoch": 114.05,
      "learning_rate": 4.7927508090614885e-06,
      "loss": 0.0275,
      "step": 293690
    },
    {
      "epoch": 114.06,
      "learning_rate": 4.792233009708738e-06,
      "loss": 0.0318,
      "step": 293700
    },
    {
      "epoch": 114.06,
      "learning_rate": 4.791715210355987e-06,
      "loss": 0.0181,
      "step": 293710
    },
    {
      "epoch": 114.07,
      "learning_rate": 4.7911974110032365e-06,
      "loss": 0.046,
      "step": 293720
    },
    {
      "epoch": 114.07,
      "learning_rate": 4.790679611650485e-06,
      "loss": 0.1164,
      "step": 293730
    },
    {
      "epoch": 114.07,
      "learning_rate": 4.790161812297735e-06,
      "loss": 0.0163,
      "step": 293740
    },
    {
      "epoch": 114.08,
      "learning_rate": 4.789644012944984e-06,
      "loss": 0.0534,
      "step": 293750
    },
    {
      "epoch": 114.08,
      "learning_rate": 4.789126213592233e-06,
      "loss": 0.0404,
      "step": 293760
    },
    {
      "epoch": 114.09,
      "learning_rate": 4.788608414239483e-06,
      "loss": 0.1018,
      "step": 293770
    },
    {
      "epoch": 114.09,
      "learning_rate": 4.788090614886732e-06,
      "loss": 0.0004,
      "step": 293780
    },
    {
      "epoch": 114.09,
      "learning_rate": 4.787572815533981e-06,
      "loss": 0.0583,
      "step": 293790
    },
    {
      "epoch": 114.1,
      "learning_rate": 4.78705501618123e-06,
      "loss": 0.0004,
      "step": 293800
    },
    {
      "epoch": 114.1,
      "learning_rate": 4.78653721682848e-06,
      "loss": 0.0088,
      "step": 293810
    },
    {
      "epoch": 114.1,
      "learning_rate": 4.7860194174757285e-06,
      "loss": 0.1356,
      "step": 293820
    },
    {
      "epoch": 114.11,
      "learning_rate": 4.785501618122978e-06,
      "loss": 0.0348,
      "step": 293830
    },
    {
      "epoch": 114.11,
      "learning_rate": 4.784983818770227e-06,
      "loss": 0.0173,
      "step": 293840
    },
    {
      "epoch": 114.12,
      "learning_rate": 4.7844660194174765e-06,
      "loss": 0.0703,
      "step": 293850
    },
    {
      "epoch": 114.12,
      "learning_rate": 4.783948220064725e-06,
      "loss": 0.0386,
      "step": 293860
    },
    {
      "epoch": 114.12,
      "learning_rate": 4.783430420711975e-06,
      "loss": 0.0969,
      "step": 293870
    },
    {
      "epoch": 114.13,
      "learning_rate": 4.782912621359224e-06,
      "loss": 0.1254,
      "step": 293880
    },
    {
      "epoch": 114.13,
      "learning_rate": 4.782394822006473e-06,
      "loss": 0.0827,
      "step": 293890
    },
    {
      "epoch": 114.14,
      "learning_rate": 4.781877022653722e-06,
      "loss": 0.1176,
      "step": 293900
    },
    {
      "epoch": 114.14,
      "learning_rate": 4.781359223300972e-06,
      "loss": 0.0529,
      "step": 293910
    },
    {
      "epoch": 114.14,
      "learning_rate": 4.78084142394822e-06,
      "loss": 0.0126,
      "step": 293920
    },
    {
      "epoch": 114.15,
      "learning_rate": 4.78032362459547e-06,
      "loss": 0.0452,
      "step": 293930
    },
    {
      "epoch": 114.15,
      "learning_rate": 4.779805825242719e-06,
      "loss": 0.0488,
      "step": 293940
    },
    {
      "epoch": 114.16,
      "learning_rate": 4.779288025889968e-06,
      "loss": 0.0831,
      "step": 293950
    },
    {
      "epoch": 114.16,
      "learning_rate": 4.778770226537217e-06,
      "loss": 0.0011,
      "step": 293960
    },
    {
      "epoch": 114.16,
      "learning_rate": 4.778252427184467e-06,
      "loss": 0.0692,
      "step": 293970
    },
    {
      "epoch": 114.17,
      "learning_rate": 4.7777346278317156e-06,
      "loss": 0.0318,
      "step": 293980
    },
    {
      "epoch": 114.17,
      "learning_rate": 4.777216828478965e-06,
      "loss": 0.0317,
      "step": 293990
    },
    {
      "epoch": 114.17,
      "learning_rate": 4.776699029126214e-06,
      "loss": 0.0665,
      "step": 294000
    },
    {
      "epoch": 114.18,
      "learning_rate": 4.776181229773464e-06,
      "loss": 0.0038,
      "step": 294010
    },
    {
      "epoch": 114.18,
      "learning_rate": 4.775663430420712e-06,
      "loss": 0.021,
      "step": 294020
    },
    {
      "epoch": 114.19,
      "learning_rate": 4.775145631067962e-06,
      "loss": 0.0312,
      "step": 294030
    },
    {
      "epoch": 114.19,
      "learning_rate": 4.774627831715211e-06,
      "loss": 0.07,
      "step": 294040
    },
    {
      "epoch": 114.19,
      "learning_rate": 4.77411003236246e-06,
      "loss": 0.0163,
      "step": 294050
    },
    {
      "epoch": 114.2,
      "learning_rate": 4.773592233009709e-06,
      "loss": 0.0175,
      "step": 294060
    },
    {
      "epoch": 114.2,
      "learning_rate": 4.773074433656959e-06,
      "loss": 0.0013,
      "step": 294070
    },
    {
      "epoch": 114.21,
      "learning_rate": 4.7725566343042075e-06,
      "loss": 0.2251,
      "step": 294080
    },
    {
      "epoch": 114.21,
      "learning_rate": 4.772038834951457e-06,
      "loss": 0.0122,
      "step": 294090
    },
    {
      "epoch": 114.21,
      "learning_rate": 4.771521035598706e-06,
      "loss": 0.0312,
      "step": 294100
    },
    {
      "epoch": 114.22,
      "learning_rate": 4.7710032362459555e-06,
      "loss": 0.0591,
      "step": 294110
    },
    {
      "epoch": 114.22,
      "learning_rate": 4.770485436893204e-06,
      "loss": 0.031,
      "step": 294120
    },
    {
      "epoch": 114.23,
      "learning_rate": 4.769967637540454e-06,
      "loss": 0.0259,
      "step": 294130
    },
    {
      "epoch": 114.23,
      "learning_rate": 4.769449838187703e-06,
      "loss": 0.0007,
      "step": 294140
    },
    {
      "epoch": 114.23,
      "learning_rate": 4.768932038834952e-06,
      "loss": 0.0915,
      "step": 294150
    },
    {
      "epoch": 114.24,
      "learning_rate": 4.768414239482201e-06,
      "loss": 0.0418,
      "step": 294160
    },
    {
      "epoch": 114.24,
      "learning_rate": 4.767896440129451e-06,
      "loss": 0.0825,
      "step": 294170
    },
    {
      "epoch": 114.24,
      "learning_rate": 4.7673786407766995e-06,
      "loss": 0.2062,
      "step": 294180
    },
    {
      "epoch": 114.25,
      "learning_rate": 4.766860841423949e-06,
      "loss": 0.1862,
      "step": 294190
    },
    {
      "epoch": 114.25,
      "learning_rate": 4.766343042071198e-06,
      "loss": 0.0013,
      "step": 294200
    },
    {
      "epoch": 114.26,
      "learning_rate": 4.765825242718447e-06,
      "loss": 0.0966,
      "step": 294210
    },
    {
      "epoch": 114.26,
      "learning_rate": 4.765307443365696e-06,
      "loss": 0.0148,
      "step": 294220
    },
    {
      "epoch": 114.26,
      "learning_rate": 4.764789644012945e-06,
      "loss": 0.0474,
      "step": 294230
    },
    {
      "epoch": 114.27,
      "learning_rate": 4.764271844660195e-06,
      "loss": 0.0479,
      "step": 294240
    },
    {
      "epoch": 114.27,
      "learning_rate": 4.763754045307443e-06,
      "loss": 0.0219,
      "step": 294250
    },
    {
      "epoch": 114.28,
      "learning_rate": 4.763236245954693e-06,
      "loss": 0.0003,
      "step": 294260
    },
    {
      "epoch": 114.28,
      "learning_rate": 4.762718446601942e-06,
      "loss": 0.0795,
      "step": 294270
    },
    {
      "epoch": 114.28,
      "learning_rate": 4.762200647249191e-06,
      "loss": 0.1411,
      "step": 294280
    },
    {
      "epoch": 114.29,
      "learning_rate": 4.76168284789644e-06,
      "loss": 0.0592,
      "step": 294290
    },
    {
      "epoch": 114.29,
      "learning_rate": 4.76116504854369e-06,
      "loss": 0.1206,
      "step": 294300
    },
    {
      "epoch": 114.3,
      "learning_rate": 4.7606472491909386e-06,
      "loss": 0.0001,
      "step": 294310
    },
    {
      "epoch": 114.3,
      "learning_rate": 4.760129449838188e-06,
      "loss": 0.0152,
      "step": 294320
    },
    {
      "epoch": 114.3,
      "learning_rate": 4.759611650485437e-06,
      "loss": 0.0589,
      "step": 294330
    },
    {
      "epoch": 114.31,
      "learning_rate": 4.7590938511326866e-06,
      "loss": 0.0092,
      "step": 294340
    },
    {
      "epoch": 114.31,
      "learning_rate": 4.758576051779935e-06,
      "loss": 0.0325,
      "step": 294350
    },
    {
      "epoch": 114.31,
      "learning_rate": 4.758058252427185e-06,
      "loss": 0.0856,
      "step": 294360
    },
    {
      "epoch": 114.32,
      "learning_rate": 4.757540453074434e-06,
      "loss": 0.0416,
      "step": 294370
    },
    {
      "epoch": 114.32,
      "learning_rate": 4.757022653721683e-06,
      "loss": 0.0443,
      "step": 294380
    },
    {
      "epoch": 114.33,
      "learning_rate": 4.756504854368932e-06,
      "loss": 0.0169,
      "step": 294390
    },
    {
      "epoch": 114.33,
      "learning_rate": 4.755987055016182e-06,
      "loss": 0.0004,
      "step": 294400
    },
    {
      "epoch": 114.33,
      "learning_rate": 4.7554692556634305e-06,
      "loss": 0.0243,
      "step": 294410
    },
    {
      "epoch": 114.34,
      "learning_rate": 4.75495145631068e-06,
      "loss": 0.1376,
      "step": 294420
    },
    {
      "epoch": 114.34,
      "learning_rate": 4.754433656957929e-06,
      "loss": 0.0178,
      "step": 294430
    },
    {
      "epoch": 114.35,
      "learning_rate": 4.7539158576051785e-06,
      "loss": 0.0895,
      "step": 294440
    },
    {
      "epoch": 114.35,
      "learning_rate": 4.753398058252427e-06,
      "loss": 0.0183,
      "step": 294450
    },
    {
      "epoch": 114.35,
      "learning_rate": 4.752880258899677e-06,
      "loss": 0.0148,
      "step": 294460
    },
    {
      "epoch": 114.36,
      "learning_rate": 4.752362459546926e-06,
      "loss": 0.0022,
      "step": 294470
    },
    {
      "epoch": 114.36,
      "learning_rate": 4.751844660194175e-06,
      "loss": 0.0096,
      "step": 294480
    },
    {
      "epoch": 114.37,
      "learning_rate": 4.751326860841424e-06,
      "loss": 0.0291,
      "step": 294490
    },
    {
      "epoch": 114.37,
      "learning_rate": 4.750809061488674e-06,
      "loss": 0.068,
      "step": 294500
    },
    {
      "epoch": 114.37,
      "learning_rate": 4.7502912621359225e-06,
      "loss": 0.0358,
      "step": 294510
    },
    {
      "epoch": 114.38,
      "learning_rate": 4.749773462783172e-06,
      "loss": 0.0273,
      "step": 294520
    },
    {
      "epoch": 114.38,
      "learning_rate": 4.749255663430421e-06,
      "loss": 0.0149,
      "step": 294530
    },
    {
      "epoch": 114.38,
      "learning_rate": 4.7487378640776705e-06,
      "loss": 0.0963,
      "step": 294540
    },
    {
      "epoch": 114.39,
      "learning_rate": 4.748220064724919e-06,
      "loss": 0.0681,
      "step": 294550
    },
    {
      "epoch": 114.39,
      "learning_rate": 4.747702265372169e-06,
      "loss": 0.0584,
      "step": 294560
    },
    {
      "epoch": 114.4,
      "learning_rate": 4.747184466019418e-06,
      "loss": 0.0326,
      "step": 294570
    },
    {
      "epoch": 114.4,
      "learning_rate": 4.746666666666667e-06,
      "loss": 0.0225,
      "step": 294580
    },
    {
      "epoch": 114.4,
      "learning_rate": 4.746148867313916e-06,
      "loss": 0.0147,
      "step": 294590
    },
    {
      "epoch": 114.41,
      "learning_rate": 4.745631067961166e-06,
      "loss": 0.0454,
      "step": 294600
    },
    {
      "epoch": 114.41,
      "learning_rate": 4.745113268608414e-06,
      "loss": 0.0019,
      "step": 294610
    },
    {
      "epoch": 114.42,
      "learning_rate": 4.744595469255664e-06,
      "loss": 0.0391,
      "step": 294620
    },
    {
      "epoch": 114.42,
      "learning_rate": 4.744077669902913e-06,
      "loss": 0.0505,
      "step": 294630
    },
    {
      "epoch": 114.42,
      "learning_rate": 4.743559870550162e-06,
      "loss": 0.0461,
      "step": 294640
    },
    {
      "epoch": 114.43,
      "learning_rate": 4.743042071197411e-06,
      "loss": 0.0099,
      "step": 294650
    },
    {
      "epoch": 114.43,
      "learning_rate": 4.742524271844661e-06,
      "loss": 0.1036,
      "step": 294660
    },
    {
      "epoch": 114.43,
      "learning_rate": 4.7420064724919096e-06,
      "loss": 0.0805,
      "step": 294670
    },
    {
      "epoch": 114.44,
      "learning_rate": 4.741488673139159e-06,
      "loss": 0.0001,
      "step": 294680
    },
    {
      "epoch": 114.44,
      "learning_rate": 4.740970873786408e-06,
      "loss": 0.0468,
      "step": 294690
    },
    {
      "epoch": 114.45,
      "learning_rate": 4.740453074433658e-06,
      "loss": 0.0229,
      "step": 294700
    },
    {
      "epoch": 114.45,
      "learning_rate": 4.739935275080906e-06,
      "loss": 0.1052,
      "step": 294710
    },
    {
      "epoch": 114.45,
      "learning_rate": 4.739417475728156e-06,
      "loss": 0.0077,
      "step": 294720
    },
    {
      "epoch": 114.46,
      "learning_rate": 4.738899676375405e-06,
      "loss": 0.0001,
      "step": 294730
    },
    {
      "epoch": 114.46,
      "learning_rate": 4.738381877022654e-06,
      "loss": 0.0865,
      "step": 294740
    },
    {
      "epoch": 114.47,
      "learning_rate": 4.737864077669903e-06,
      "loss": 0.1698,
      "step": 294750
    },
    {
      "epoch": 114.47,
      "learning_rate": 4.737346278317153e-06,
      "loss": 0.0091,
      "step": 294760
    },
    {
      "epoch": 114.47,
      "learning_rate": 4.7368284789644015e-06,
      "loss": 0.0912,
      "step": 294770
    },
    {
      "epoch": 114.48,
      "learning_rate": 4.736310679611651e-06,
      "loss": 0.0194,
      "step": 294780
    },
    {
      "epoch": 114.48,
      "learning_rate": 4.7357928802589e-06,
      "loss": 0.0841,
      "step": 294790
    },
    {
      "epoch": 114.49,
      "learning_rate": 4.7352750809061495e-06,
      "loss": 0.0001,
      "step": 294800
    },
    {
      "epoch": 114.49,
      "learning_rate": 4.734757281553398e-06,
      "loss": 0.0103,
      "step": 294810
    },
    {
      "epoch": 114.49,
      "learning_rate": 4.734239482200648e-06,
      "loss": 0.0666,
      "step": 294820
    },
    {
      "epoch": 114.5,
      "learning_rate": 4.733721682847897e-06,
      "loss": 0.0246,
      "step": 294830
    },
    {
      "epoch": 114.5,
      "learning_rate": 4.733203883495146e-06,
      "loss": 0.0638,
      "step": 294840
    },
    {
      "epoch": 114.5,
      "learning_rate": 4.732686084142395e-06,
      "loss": 0.1281,
      "step": 294850
    },
    {
      "epoch": 114.51,
      "learning_rate": 4.732168284789645e-06,
      "loss": 0.0123,
      "step": 294860
    },
    {
      "epoch": 114.51,
      "learning_rate": 4.7316504854368935e-06,
      "loss": 0.0667,
      "step": 294870
    },
    {
      "epoch": 114.52,
      "learning_rate": 4.731132686084143e-06,
      "loss": 0.0494,
      "step": 294880
    },
    {
      "epoch": 114.52,
      "learning_rate": 4.730614886731392e-06,
      "loss": 0.0007,
      "step": 294890
    },
    {
      "epoch": 114.52,
      "learning_rate": 4.7300970873786415e-06,
      "loss": 0.0156,
      "step": 294900
    },
    {
      "epoch": 114.53,
      "learning_rate": 4.72957928802589e-06,
      "loss": 0.0529,
      "step": 294910
    },
    {
      "epoch": 114.53,
      "learning_rate": 4.72906148867314e-06,
      "loss": 0.0186,
      "step": 294920
    },
    {
      "epoch": 114.54,
      "learning_rate": 4.728543689320389e-06,
      "loss": 0.0535,
      "step": 294930
    },
    {
      "epoch": 114.54,
      "learning_rate": 4.728025889967638e-06,
      "loss": 0.0084,
      "step": 294940
    },
    {
      "epoch": 114.54,
      "learning_rate": 4.727508090614887e-06,
      "loss": 0.0088,
      "step": 294950
    },
    {
      "epoch": 114.55,
      "learning_rate": 4.726990291262137e-06,
      "loss": 0.0172,
      "step": 294960
    },
    {
      "epoch": 114.55,
      "learning_rate": 4.726472491909385e-06,
      "loss": 0.0108,
      "step": 294970
    },
    {
      "epoch": 114.56,
      "learning_rate": 4.725954692556635e-06,
      "loss": 0.028,
      "step": 294980
    },
    {
      "epoch": 114.56,
      "learning_rate": 4.725436893203884e-06,
      "loss": 0.0621,
      "step": 294990
    },
    {
      "epoch": 114.56,
      "learning_rate": 4.724919093851133e-06,
      "loss": 0.2592,
      "step": 295000
    },
    {
      "epoch": 114.57,
      "learning_rate": 4.724401294498382e-06,
      "loss": 0.0002,
      "step": 295010
    },
    {
      "epoch": 114.57,
      "learning_rate": 4.723883495145632e-06,
      "loss": 0.0361,
      "step": 295020
    },
    {
      "epoch": 114.57,
      "learning_rate": 4.7233656957928806e-06,
      "loss": 0.0037,
      "step": 295030
    },
    {
      "epoch": 114.58,
      "learning_rate": 4.722847896440129e-06,
      "loss": 0.0174,
      "step": 295040
    },
    {
      "epoch": 114.58,
      "learning_rate": 4.722330097087379e-06,
      "loss": 0.0004,
      "step": 295050
    },
    {
      "epoch": 114.59,
      "learning_rate": 4.721812297734628e-06,
      "loss": 0.0193,
      "step": 295060
    },
    {
      "epoch": 114.59,
      "learning_rate": 4.721294498381877e-06,
      "loss": 0.1299,
      "step": 295070
    },
    {
      "epoch": 114.59,
      "learning_rate": 4.720776699029126e-06,
      "loss": 0.1156,
      "step": 295080
    },
    {
      "epoch": 114.6,
      "learning_rate": 4.720258899676376e-06,
      "loss": 0.0622,
      "step": 295090
    },
    {
      "epoch": 114.6,
      "learning_rate": 4.7197411003236245e-06,
      "loss": 0.001,
      "step": 295100
    },
    {
      "epoch": 114.61,
      "learning_rate": 4.719223300970874e-06,
      "loss": 0.027,
      "step": 295110
    },
    {
      "epoch": 114.61,
      "learning_rate": 4.718705501618123e-06,
      "loss": 0.0077,
      "step": 295120
    },
    {
      "epoch": 114.61,
      "learning_rate": 4.7181877022653725e-06,
      "loss": 0.0018,
      "step": 295130
    },
    {
      "epoch": 114.62,
      "learning_rate": 4.717669902912621e-06,
      "loss": 0.0012,
      "step": 295140
    },
    {
      "epoch": 114.62,
      "learning_rate": 4.717152103559871e-06,
      "loss": 0.0268,
      "step": 295150
    },
    {
      "epoch": 114.63,
      "learning_rate": 4.71663430420712e-06,
      "loss": 0.0088,
      "step": 295160
    },
    {
      "epoch": 114.63,
      "learning_rate": 4.716116504854369e-06,
      "loss": 0.0157,
      "step": 295170
    },
    {
      "epoch": 114.63,
      "learning_rate": 4.715598705501618e-06,
      "loss": 0.0164,
      "step": 295180
    },
    {
      "epoch": 114.64,
      "learning_rate": 4.715080906148868e-06,
      "loss": 0.0293,
      "step": 295190
    },
    {
      "epoch": 114.64,
      "learning_rate": 4.7145631067961165e-06,
      "loss": 0.0015,
      "step": 295200
    },
    {
      "epoch": 114.64,
      "learning_rate": 4.714045307443366e-06,
      "loss": 0.1605,
      "step": 295210
    },
    {
      "epoch": 114.65,
      "learning_rate": 4.713527508090615e-06,
      "loss": 0.1028,
      "step": 295220
    },
    {
      "epoch": 114.65,
      "learning_rate": 4.7130097087378645e-06,
      "loss": 0.0289,
      "step": 295230
    },
    {
      "epoch": 114.66,
      "learning_rate": 4.712491909385113e-06,
      "loss": 0.0378,
      "step": 295240
    },
    {
      "epoch": 114.66,
      "learning_rate": 4.711974110032363e-06,
      "loss": 0.0104,
      "step": 295250
    },
    {
      "epoch": 114.66,
      "learning_rate": 4.711456310679612e-06,
      "loss": 0.0066,
      "step": 295260
    },
    {
      "epoch": 114.67,
      "learning_rate": 4.710938511326861e-06,
      "loss": 0.1991,
      "step": 295270
    },
    {
      "epoch": 114.67,
      "learning_rate": 4.71042071197411e-06,
      "loss": 0.0999,
      "step": 295280
    },
    {
      "epoch": 114.68,
      "learning_rate": 4.70990291262136e-06,
      "loss": 0.0011,
      "step": 295290
    },
    {
      "epoch": 114.68,
      "learning_rate": 4.709385113268608e-06,
      "loss": 0.0629,
      "step": 295300
    },
    {
      "epoch": 114.68,
      "learning_rate": 4.708867313915858e-06,
      "loss": 0.0069,
      "step": 295310
    },
    {
      "epoch": 114.69,
      "learning_rate": 4.708349514563107e-06,
      "loss": 0.0023,
      "step": 295320
    },
    {
      "epoch": 114.69,
      "learning_rate": 4.707831715210356e-06,
      "loss": 0.0254,
      "step": 295330
    },
    {
      "epoch": 114.7,
      "learning_rate": 4.707313915857605e-06,
      "loss": 0.0228,
      "step": 295340
    },
    {
      "epoch": 114.7,
      "learning_rate": 4.706796116504855e-06,
      "loss": 0.0008,
      "step": 295350
    },
    {
      "epoch": 114.7,
      "learning_rate": 4.7062783171521036e-06,
      "loss": 0.0429,
      "step": 295360
    },
    {
      "epoch": 114.71,
      "learning_rate": 4.705760517799353e-06,
      "loss": 0.1593,
      "step": 295370
    },
    {
      "epoch": 114.71,
      "learning_rate": 4.705242718446602e-06,
      "loss": 0.1689,
      "step": 295380
    },
    {
      "epoch": 114.71,
      "learning_rate": 4.7047249190938516e-06,
      "loss": 0.0976,
      "step": 295390
    },
    {
      "epoch": 114.72,
      "learning_rate": 4.7042071197411e-06,
      "loss": 0.0002,
      "step": 295400
    },
    {
      "epoch": 114.72,
      "learning_rate": 4.70368932038835e-06,
      "loss": 0.1808,
      "step": 295410
    },
    {
      "epoch": 114.73,
      "learning_rate": 4.703171521035599e-06,
      "loss": 0.1095,
      "step": 295420
    },
    {
      "epoch": 114.73,
      "learning_rate": 4.702653721682848e-06,
      "loss": 0.0246,
      "step": 295430
    },
    {
      "epoch": 114.73,
      "learning_rate": 4.702135922330098e-06,
      "loss": 0.0002,
      "step": 295440
    },
    {
      "epoch": 114.74,
      "learning_rate": 4.701618122977347e-06,
      "loss": 0.0828,
      "step": 295450
    },
    {
      "epoch": 114.74,
      "learning_rate": 4.701100323624596e-06,
      "loss": 0.0964,
      "step": 295460
    },
    {
      "epoch": 114.75,
      "learning_rate": 4.700582524271845e-06,
      "loss": 0.0735,
      "step": 295470
    },
    {
      "epoch": 114.75,
      "learning_rate": 4.700064724919095e-06,
      "loss": 0.0354,
      "step": 295480
    },
    {
      "epoch": 114.75,
      "learning_rate": 4.6995469255663435e-06,
      "loss": 0.071,
      "step": 295490
    },
    {
      "epoch": 114.76,
      "learning_rate": 4.699029126213593e-06,
      "loss": 0.0064,
      "step": 295500
    },
    {
      "epoch": 114.76,
      "learning_rate": 4.698511326860842e-06,
      "loss": 0.011,
      "step": 295510
    },
    {
      "epoch": 114.77,
      "learning_rate": 4.6979935275080915e-06,
      "loss": 0.0458,
      "step": 295520
    },
    {
      "epoch": 114.77,
      "learning_rate": 4.69747572815534e-06,
      "loss": 0.0984,
      "step": 295530
    },
    {
      "epoch": 114.77,
      "learning_rate": 4.69695792880259e-06,
      "loss": 0.0137,
      "step": 295540
    },
    {
      "epoch": 114.78,
      "learning_rate": 4.696440129449839e-06,
      "loss": 0.0211,
      "step": 295550
    },
    {
      "epoch": 114.78,
      "learning_rate": 4.695922330097088e-06,
      "loss": 0.1298,
      "step": 295560
    },
    {
      "epoch": 114.78,
      "learning_rate": 4.695404530744337e-06,
      "loss": 0.1281,
      "step": 295570
    },
    {
      "epoch": 114.79,
      "learning_rate": 4.694886731391587e-06,
      "loss": 0.0364,
      "step": 295580
    },
    {
      "epoch": 114.79,
      "learning_rate": 4.6943689320388355e-06,
      "loss": 0.0004,
      "step": 295590
    },
    {
      "epoch": 114.8,
      "learning_rate": 4.693851132686085e-06,
      "loss": 0.0875,
      "step": 295600
    },
    {
      "epoch": 114.8,
      "learning_rate": 4.693333333333334e-06,
      "loss": 0.0214,
      "step": 295610
    },
    {
      "epoch": 114.8,
      "learning_rate": 4.6928155339805835e-06,
      "loss": 0.0206,
      "step": 295620
    },
    {
      "epoch": 114.81,
      "learning_rate": 4.692297734627832e-06,
      "loss": 0.0774,
      "step": 295630
    },
    {
      "epoch": 114.81,
      "learning_rate": 4.691779935275082e-06,
      "loss": 0.0472,
      "step": 295640
    },
    {
      "epoch": 114.82,
      "learning_rate": 4.691262135922331e-06,
      "loss": 0.0426,
      "step": 295650
    },
    {
      "epoch": 114.82,
      "learning_rate": 4.69074433656958e-06,
      "loss": 0.0703,
      "step": 295660
    },
    {
      "epoch": 114.82,
      "learning_rate": 4.690226537216829e-06,
      "loss": 0.0924,
      "step": 295670
    },
    {
      "epoch": 114.83,
      "learning_rate": 4.689708737864078e-06,
      "loss": 0.0677,
      "step": 295680
    },
    {
      "epoch": 114.83,
      "learning_rate": 4.689190938511327e-06,
      "loss": 0.0157,
      "step": 295690
    },
    {
      "epoch": 114.83,
      "learning_rate": 4.688673139158576e-06,
      "loss": 0.0123,
      "step": 295700
    },
    {
      "epoch": 114.84,
      "learning_rate": 4.688155339805826e-06,
      "loss": 0.0416,
      "step": 295710
    },
    {
      "epoch": 114.84,
      "learning_rate": 4.6876375404530746e-06,
      "loss": 0.1002,
      "step": 295720
    },
    {
      "epoch": 114.85,
      "learning_rate": 4.687119741100324e-06,
      "loss": 0.0422,
      "step": 295730
    },
    {
      "epoch": 114.85,
      "learning_rate": 4.686601941747573e-06,
      "loss": 0.0496,
      "step": 295740
    },
    {
      "epoch": 114.85,
      "learning_rate": 4.686084142394823e-06,
      "loss": 0.0495,
      "step": 295750
    },
    {
      "epoch": 114.86,
      "learning_rate": 4.685566343042071e-06,
      "loss": 0.018,
      "step": 295760
    },
    {
      "epoch": 114.86,
      "learning_rate": 4.685048543689321e-06,
      "loss": 0.0573,
      "step": 295770
    },
    {
      "epoch": 114.87,
      "learning_rate": 4.68453074433657e-06,
      "loss": 0.0236,
      "step": 295780
    },
    {
      "epoch": 114.87,
      "learning_rate": 4.684012944983819e-06,
      "loss": 0.0205,
      "step": 295790
    },
    {
      "epoch": 114.87,
      "learning_rate": 4.683495145631068e-06,
      "loss": 0.0643,
      "step": 295800
    },
    {
      "epoch": 114.88,
      "learning_rate": 4.682977346278318e-06,
      "loss": 0.0132,
      "step": 295810
    },
    {
      "epoch": 114.88,
      "learning_rate": 4.6824595469255665e-06,
      "loss": 0.0372,
      "step": 295820
    },
    {
      "epoch": 114.89,
      "learning_rate": 4.681941747572816e-06,
      "loss": 0.0016,
      "step": 295830
    },
    {
      "epoch": 114.89,
      "learning_rate": 4.681423948220065e-06,
      "loss": 0.1709,
      "step": 295840
    },
    {
      "epoch": 114.89,
      "learning_rate": 4.680906148867314e-06,
      "loss": 0.0696,
      "step": 295850
    },
    {
      "epoch": 114.9,
      "learning_rate": 4.680388349514563e-06,
      "loss": 0.0191,
      "step": 295860
    },
    {
      "epoch": 114.9,
      "learning_rate": 4.679870550161812e-06,
      "loss": 0.0197,
      "step": 295870
    },
    {
      "epoch": 114.9,
      "learning_rate": 4.679352750809062e-06,
      "loss": 0.0191,
      "step": 295880
    },
    {
      "epoch": 114.91,
      "learning_rate": 4.6788349514563105e-06,
      "loss": 0.079,
      "step": 295890
    },
    {
      "epoch": 114.91,
      "learning_rate": 4.67831715210356e-06,
      "loss": 0.1457,
      "step": 295900
    },
    {
      "epoch": 114.92,
      "learning_rate": 4.677799352750809e-06,
      "loss": 0.117,
      "step": 295910
    },
    {
      "epoch": 114.92,
      "learning_rate": 4.6772815533980585e-06,
      "loss": 0.0716,
      "step": 295920
    },
    {
      "epoch": 114.92,
      "learning_rate": 4.676763754045307e-06,
      "loss": 0.0001,
      "step": 295930
    },
    {
      "epoch": 114.93,
      "learning_rate": 4.676245954692557e-06,
      "loss": 0.0954,
      "step": 295940
    },
    {
      "epoch": 114.93,
      "learning_rate": 4.675728155339806e-06,
      "loss": 0.0225,
      "step": 295950
    },
    {
      "epoch": 114.94,
      "learning_rate": 4.675210355987055e-06,
      "loss": 0.0005,
      "step": 295960
    },
    {
      "epoch": 114.94,
      "learning_rate": 4.674692556634304e-06,
      "loss": 0.0025,
      "step": 295970
    },
    {
      "epoch": 114.94,
      "learning_rate": 4.674174757281554e-06,
      "loss": 0.0345,
      "step": 295980
    },
    {
      "epoch": 114.95,
      "learning_rate": 4.673656957928802e-06,
      "loss": 0.0486,
      "step": 295990
    },
    {
      "epoch": 114.95,
      "learning_rate": 4.673139158576052e-06,
      "loss": 0.0015,
      "step": 296000
    },
    {
      "epoch": 114.96,
      "learning_rate": 4.672621359223302e-06,
      "loss": 0.1916,
      "step": 296010
    },
    {
      "epoch": 114.96,
      "learning_rate": 4.67210355987055e-06,
      "loss": 0.022,
      "step": 296020
    },
    {
      "epoch": 114.96,
      "learning_rate": 4.6715857605178e-06,
      "loss": 0.0021,
      "step": 296030
    },
    {
      "epoch": 114.97,
      "learning_rate": 4.671067961165049e-06,
      "loss": 0.0294,
      "step": 296040
    },
    {
      "epoch": 114.97,
      "learning_rate": 4.670550161812298e-06,
      "loss": 0.0645,
      "step": 296050
    },
    {
      "epoch": 114.97,
      "learning_rate": 4.670032362459547e-06,
      "loss": 0.0255,
      "step": 296060
    },
    {
      "epoch": 114.98,
      "learning_rate": 4.669514563106797e-06,
      "loss": 0.0579,
      "step": 296070
    },
    {
      "epoch": 114.98,
      "learning_rate": 4.6689967637540456e-06,
      "loss": 0.0296,
      "step": 296080
    },
    {
      "epoch": 114.99,
      "learning_rate": 4.668478964401295e-06,
      "loss": 0.0821,
      "step": 296090
    },
    {
      "epoch": 114.99,
      "learning_rate": 4.667961165048544e-06,
      "loss": 0.0307,
      "step": 296100
    },
    {
      "epoch": 114.99,
      "learning_rate": 4.667443365695794e-06,
      "loss": 0.0338,
      "step": 296110
    },
    {
      "epoch": 115.0,
      "learning_rate": 4.666925566343042e-06,
      "loss": 0.0022,
      "step": 296120
    },
    {
      "epoch": 115.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.3717191815376282,
      "eval_runtime": 8.1946,
      "eval_samples_per_second": 443.583,
      "eval_steps_per_second": 55.524,
      "step": 296125
    },
    {
      "epoch": 115.0,
      "learning_rate": 4.666407766990292e-06,
      "loss": 0.0316,
      "step": 296130
    },
    {
      "epoch": 115.01,
      "learning_rate": 4.665889967637541e-06,
      "loss": 0.0021,
      "step": 296140
    },
    {
      "epoch": 115.01,
      "learning_rate": 4.66537216828479e-06,
      "loss": 0.0237,
      "step": 296150
    },
    {
      "epoch": 115.01,
      "learning_rate": 4.664854368932039e-06,
      "loss": 0.0174,
      "step": 296160
    },
    {
      "epoch": 115.02,
      "learning_rate": 4.664336569579289e-06,
      "loss": 0.0662,
      "step": 296170
    },
    {
      "epoch": 115.02,
      "learning_rate": 4.6638187702265375e-06,
      "loss": 0.0727,
      "step": 296180
    },
    {
      "epoch": 115.03,
      "learning_rate": 4.663300970873787e-06,
      "loss": 0.21,
      "step": 296190
    },
    {
      "epoch": 115.03,
      "learning_rate": 4.662783171521036e-06,
      "loss": 0.0009,
      "step": 296200
    },
    {
      "epoch": 115.03,
      "learning_rate": 4.6622653721682855e-06,
      "loss": 0.0403,
      "step": 296210
    },
    {
      "epoch": 115.04,
      "learning_rate": 4.661747572815534e-06,
      "loss": 0.0164,
      "step": 296220
    },
    {
      "epoch": 115.04,
      "learning_rate": 4.661229773462784e-06,
      "loss": 0.0567,
      "step": 296230
    },
    {
      "epoch": 115.04,
      "learning_rate": 4.660711974110033e-06,
      "loss": 0.0011,
      "step": 296240
    },
    {
      "epoch": 115.05,
      "learning_rate": 4.660194174757282e-06,
      "loss": 0.0119,
      "step": 296250
    },
    {
      "epoch": 115.05,
      "learning_rate": 4.659676375404531e-06,
      "loss": 0.1424,
      "step": 296260
    },
    {
      "epoch": 115.06,
      "learning_rate": 4.659158576051781e-06,
      "loss": 0.0615,
      "step": 296270
    },
    {
      "epoch": 115.06,
      "learning_rate": 4.6586407766990295e-06,
      "loss": 0.0728,
      "step": 296280
    },
    {
      "epoch": 115.06,
      "learning_rate": 4.658122977346279e-06,
      "loss": 0.0114,
      "step": 296290
    },
    {
      "epoch": 115.07,
      "learning_rate": 4.657605177993528e-06,
      "loss": 0.144,
      "step": 296300
    },
    {
      "epoch": 115.07,
      "learning_rate": 4.6570873786407775e-06,
      "loss": 0.0411,
      "step": 296310
    },
    {
      "epoch": 115.08,
      "learning_rate": 4.656569579288026e-06,
      "loss": 0.0271,
      "step": 296320
    },
    {
      "epoch": 115.08,
      "learning_rate": 4.656051779935276e-06,
      "loss": 0.0009,
      "step": 296330
    },
    {
      "epoch": 115.08,
      "learning_rate": 4.655533980582525e-06,
      "loss": 0.089,
      "step": 296340
    },
    {
      "epoch": 115.09,
      "learning_rate": 4.655016181229774e-06,
      "loss": 0.0259,
      "step": 296350
    },
    {
      "epoch": 115.09,
      "learning_rate": 4.654498381877023e-06,
      "loss": 0.0016,
      "step": 296360
    },
    {
      "epoch": 115.1,
      "learning_rate": 4.653980582524273e-06,
      "loss": 0.0405,
      "step": 296370
    },
    {
      "epoch": 115.1,
      "learning_rate": 4.653462783171521e-06,
      "loss": 0.2202,
      "step": 296380
    },
    {
      "epoch": 115.1,
      "learning_rate": 4.652944983818771e-06,
      "loss": 0.0696,
      "step": 296390
    },
    {
      "epoch": 115.11,
      "learning_rate": 4.65242718446602e-06,
      "loss": 0.1528,
      "step": 296400
    },
    {
      "epoch": 115.11,
      "learning_rate": 4.651909385113269e-06,
      "loss": 0.0358,
      "step": 296410
    },
    {
      "epoch": 115.11,
      "learning_rate": 4.651391585760518e-06,
      "loss": 0.0894,
      "step": 296420
    },
    {
      "epoch": 115.12,
      "learning_rate": 4.650873786407768e-06,
      "loss": 0.0375,
      "step": 296430
    },
    {
      "epoch": 115.12,
      "learning_rate": 4.650355987055017e-06,
      "loss": 0.046,
      "step": 296440
    },
    {
      "epoch": 115.13,
      "learning_rate": 4.649838187702266e-06,
      "loss": 0.0101,
      "step": 296450
    },
    {
      "epoch": 115.13,
      "learning_rate": 4.649320388349515e-06,
      "loss": 0.0514,
      "step": 296460
    },
    {
      "epoch": 115.13,
      "learning_rate": 4.648802588996765e-06,
      "loss": 0.0014,
      "step": 296470
    },
    {
      "epoch": 115.14,
      "learning_rate": 4.648284789644013e-06,
      "loss": 0.0769,
      "step": 296480
    },
    {
      "epoch": 115.14,
      "learning_rate": 4.647766990291262e-06,
      "loss": 0.122,
      "step": 296490
    },
    {
      "epoch": 115.15,
      "learning_rate": 4.647249190938512e-06,
      "loss": 0.0601,
      "step": 296500
    },
    {
      "epoch": 115.15,
      "learning_rate": 4.6467313915857605e-06,
      "loss": 0.0459,
      "step": 296510
    },
    {
      "epoch": 115.15,
      "learning_rate": 4.64621359223301e-06,
      "loss": 0.0678,
      "step": 296520
    },
    {
      "epoch": 115.16,
      "learning_rate": 4.645695792880259e-06,
      "loss": 0.1038,
      "step": 296530
    },
    {
      "epoch": 115.16,
      "learning_rate": 4.6451779935275085e-06,
      "loss": 0.0489,
      "step": 296540
    },
    {
      "epoch": 115.17,
      "learning_rate": 4.644660194174757e-06,
      "loss": 0.0072,
      "step": 296550
    },
    {
      "epoch": 115.17,
      "learning_rate": 4.644142394822007e-06,
      "loss": 0.0119,
      "step": 296560
    },
    {
      "epoch": 115.17,
      "learning_rate": 4.643624595469256e-06,
      "loss": 0.0249,
      "step": 296570
    },
    {
      "epoch": 115.18,
      "learning_rate": 4.643106796116505e-06,
      "loss": 0.0001,
      "step": 296580
    },
    {
      "epoch": 115.18,
      "learning_rate": 4.642588996763754e-06,
      "loss": 0.1036,
      "step": 296590
    },
    {
      "epoch": 115.18,
      "learning_rate": 4.642071197411004e-06,
      "loss": 0.0181,
      "step": 296600
    },
    {
      "epoch": 115.19,
      "learning_rate": 4.6415533980582525e-06,
      "loss": 0.0103,
      "step": 296610
    },
    {
      "epoch": 115.19,
      "learning_rate": 4.641035598705502e-06,
      "loss": 0.0165,
      "step": 296620
    },
    {
      "epoch": 115.2,
      "learning_rate": 4.640517799352751e-06,
      "loss": 0.0567,
      "step": 296630
    },
    {
      "epoch": 115.2,
      "learning_rate": 4.6400000000000005e-06,
      "loss": 0.0092,
      "step": 296640
    },
    {
      "epoch": 115.2,
      "learning_rate": 4.639482200647249e-06,
      "loss": 0.0061,
      "step": 296650
    },
    {
      "epoch": 115.21,
      "learning_rate": 4.638964401294499e-06,
      "loss": 0.0005,
      "step": 296660
    },
    {
      "epoch": 115.21,
      "learning_rate": 4.638446601941748e-06,
      "loss": 0.0944,
      "step": 296670
    },
    {
      "epoch": 115.22,
      "learning_rate": 4.637928802588997e-06,
      "loss": 0.0444,
      "step": 296680
    },
    {
      "epoch": 115.22,
      "learning_rate": 4.637411003236246e-06,
      "loss": 0.0524,
      "step": 296690
    },
    {
      "epoch": 115.22,
      "learning_rate": 4.636893203883496e-06,
      "loss": 0.0115,
      "step": 296700
    },
    {
      "epoch": 115.23,
      "learning_rate": 4.636375404530744e-06,
      "loss": 0.0867,
      "step": 296710
    },
    {
      "epoch": 115.23,
      "learning_rate": 4.635857605177994e-06,
      "loss": 0.0007,
      "step": 296720
    },
    {
      "epoch": 115.23,
      "learning_rate": 4.635339805825243e-06,
      "loss": 0.0361,
      "step": 296730
    },
    {
      "epoch": 115.24,
      "learning_rate": 4.634822006472492e-06,
      "loss": 0.0815,
      "step": 296740
    },
    {
      "epoch": 115.24,
      "learning_rate": 4.634304207119741e-06,
      "loss": 0.0666,
      "step": 296750
    },
    {
      "epoch": 115.25,
      "learning_rate": 4.633786407766991e-06,
      "loss": 0.0026,
      "step": 296760
    },
    {
      "epoch": 115.25,
      "learning_rate": 4.6332686084142396e-06,
      "loss": 0.0697,
      "step": 296770
    },
    {
      "epoch": 115.25,
      "learning_rate": 4.632750809061489e-06,
      "loss": 0.0514,
      "step": 296780
    },
    {
      "epoch": 115.26,
      "learning_rate": 4.632233009708738e-06,
      "loss": 0.0808,
      "step": 296790
    },
    {
      "epoch": 115.26,
      "learning_rate": 4.631715210355988e-06,
      "loss": 0.1025,
      "step": 296800
    },
    {
      "epoch": 115.27,
      "learning_rate": 4.631197411003236e-06,
      "loss": 0.0193,
      "step": 296810
    },
    {
      "epoch": 115.27,
      "learning_rate": 4.630679611650486e-06,
      "loss": 0.0193,
      "step": 296820
    },
    {
      "epoch": 115.27,
      "learning_rate": 4.630161812297735e-06,
      "loss": 0.0478,
      "step": 296830
    },
    {
      "epoch": 115.28,
      "learning_rate": 4.629644012944984e-06,
      "loss": 0.0226,
      "step": 296840
    },
    {
      "epoch": 115.28,
      "learning_rate": 4.629126213592233e-06,
      "loss": 0.0626,
      "step": 296850
    },
    {
      "epoch": 115.29,
      "learning_rate": 4.628608414239483e-06,
      "loss": 0.0651,
      "step": 296860
    },
    {
      "epoch": 115.29,
      "learning_rate": 4.6280906148867315e-06,
      "loss": 0.0006,
      "step": 296870
    },
    {
      "epoch": 115.29,
      "learning_rate": 4.627572815533981e-06,
      "loss": 0.086,
      "step": 296880
    },
    {
      "epoch": 115.3,
      "learning_rate": 4.62705501618123e-06,
      "loss": 0.0006,
      "step": 296890
    },
    {
      "epoch": 115.3,
      "learning_rate": 4.6265372168284795e-06,
      "loss": 0.1143,
      "step": 296900
    },
    {
      "epoch": 115.3,
      "learning_rate": 4.626019417475728e-06,
      "loss": 0.0931,
      "step": 296910
    },
    {
      "epoch": 115.31,
      "learning_rate": 4.625501618122978e-06,
      "loss": 0.0201,
      "step": 296920
    },
    {
      "epoch": 115.31,
      "learning_rate": 4.624983818770227e-06,
      "loss": 0.0009,
      "step": 296930
    },
    {
      "epoch": 115.32,
      "learning_rate": 4.624466019417476e-06,
      "loss": 0.0024,
      "step": 296940
    },
    {
      "epoch": 115.32,
      "learning_rate": 4.623948220064725e-06,
      "loss": 0.0005,
      "step": 296950
    },
    {
      "epoch": 115.32,
      "learning_rate": 4.623430420711975e-06,
      "loss": 0.0454,
      "step": 296960
    },
    {
      "epoch": 115.33,
      "learning_rate": 4.6229126213592235e-06,
      "loss": 0.0231,
      "step": 296970
    },
    {
      "epoch": 115.33,
      "learning_rate": 4.622394822006473e-06,
      "loss": 0.0744,
      "step": 296980
    },
    {
      "epoch": 115.34,
      "learning_rate": 4.621877022653722e-06,
      "loss": 0.0008,
      "step": 296990
    },
    {
      "epoch": 115.34,
      "learning_rate": 4.6213592233009715e-06,
      "loss": 0.0001,
      "step": 297000
    },
    {
      "epoch": 115.34,
      "learning_rate": 4.62084142394822e-06,
      "loss": 0.0084,
      "step": 297010
    },
    {
      "epoch": 115.35,
      "learning_rate": 4.62032362459547e-06,
      "loss": 0.0711,
      "step": 297020
    },
    {
      "epoch": 115.35,
      "learning_rate": 4.619805825242719e-06,
      "loss": 0.1702,
      "step": 297030
    },
    {
      "epoch": 115.36,
      "learning_rate": 4.619288025889968e-06,
      "loss": 0.0207,
      "step": 297040
    },
    {
      "epoch": 115.36,
      "learning_rate": 4.618770226537217e-06,
      "loss": 0.0441,
      "step": 297050
    },
    {
      "epoch": 115.36,
      "learning_rate": 4.618252427184467e-06,
      "loss": 0.0308,
      "step": 297060
    },
    {
      "epoch": 115.37,
      "learning_rate": 4.617734627831715e-06,
      "loss": 0.0195,
      "step": 297070
    },
    {
      "epoch": 115.37,
      "learning_rate": 4.617216828478965e-06,
      "loss": 0.0003,
      "step": 297080
    },
    {
      "epoch": 115.37,
      "learning_rate": 4.616699029126214e-06,
      "loss": 0.001,
      "step": 297090
    },
    {
      "epoch": 115.38,
      "learning_rate": 4.616181229773463e-06,
      "loss": 0.0422,
      "step": 297100
    },
    {
      "epoch": 115.38,
      "learning_rate": 4.615663430420712e-06,
      "loss": 0.0113,
      "step": 297110
    },
    {
      "epoch": 115.39,
      "learning_rate": 4.615145631067962e-06,
      "loss": 0.0712,
      "step": 297120
    },
    {
      "epoch": 115.39,
      "learning_rate": 4.614627831715211e-06,
      "loss": 0.0136,
      "step": 297130
    },
    {
      "epoch": 115.39,
      "learning_rate": 4.61411003236246e-06,
      "loss": 0.03,
      "step": 297140
    },
    {
      "epoch": 115.4,
      "learning_rate": 4.613592233009709e-06,
      "loss": 0.0028,
      "step": 297150
    },
    {
      "epoch": 115.4,
      "learning_rate": 4.613074433656959e-06,
      "loss": 0.1817,
      "step": 297160
    },
    {
      "epoch": 115.41,
      "learning_rate": 4.612556634304207e-06,
      "loss": 0.0554,
      "step": 297170
    },
    {
      "epoch": 115.41,
      "learning_rate": 4.612038834951457e-06,
      "loss": 0.028,
      "step": 297180
    },
    {
      "epoch": 115.41,
      "learning_rate": 4.611521035598706e-06,
      "loss": 0.0084,
      "step": 297190
    },
    {
      "epoch": 115.42,
      "learning_rate": 4.611003236245955e-06,
      "loss": 0.0017,
      "step": 297200
    },
    {
      "epoch": 115.42,
      "learning_rate": 4.610485436893204e-06,
      "loss": 0.0435,
      "step": 297210
    },
    {
      "epoch": 115.43,
      "learning_rate": 4.609967637540454e-06,
      "loss": 0.0979,
      "step": 297220
    },
    {
      "epoch": 115.43,
      "learning_rate": 4.6094498381877025e-06,
      "loss": 0.1138,
      "step": 297230
    },
    {
      "epoch": 115.43,
      "learning_rate": 4.608932038834952e-06,
      "loss": 0.0479,
      "step": 297240
    },
    {
      "epoch": 115.44,
      "learning_rate": 4.608414239482201e-06,
      "loss": 0.112,
      "step": 297250
    },
    {
      "epoch": 115.44,
      "learning_rate": 4.6078964401294505e-06,
      "loss": 0.0191,
      "step": 297260
    },
    {
      "epoch": 115.44,
      "learning_rate": 4.607378640776699e-06,
      "loss": 0.0285,
      "step": 297270
    },
    {
      "epoch": 115.45,
      "learning_rate": 4.606860841423949e-06,
      "loss": 0.0153,
      "step": 297280
    },
    {
      "epoch": 115.45,
      "learning_rate": 4.606343042071198e-06,
      "loss": 0.2271,
      "step": 297290
    },
    {
      "epoch": 115.46,
      "learning_rate": 4.605825242718447e-06,
      "loss": 0.0792,
      "step": 297300
    },
    {
      "epoch": 115.46,
      "learning_rate": 4.605307443365696e-06,
      "loss": 0.0825,
      "step": 297310
    },
    {
      "epoch": 115.46,
      "learning_rate": 4.604789644012945e-06,
      "loss": 0.0061,
      "step": 297320
    },
    {
      "epoch": 115.47,
      "learning_rate": 4.6042718446601945e-06,
      "loss": 0.0669,
      "step": 297330
    },
    {
      "epoch": 115.47,
      "learning_rate": 4.603754045307443e-06,
      "loss": 0.0302,
      "step": 297340
    },
    {
      "epoch": 115.48,
      "learning_rate": 4.603236245954693e-06,
      "loss": 0.0008,
      "step": 297350
    },
    {
      "epoch": 115.48,
      "learning_rate": 4.602718446601942e-06,
      "loss": 0.0411,
      "step": 297360
    },
    {
      "epoch": 115.48,
      "learning_rate": 4.602200647249191e-06,
      "loss": 0.0275,
      "step": 297370
    },
    {
      "epoch": 115.49,
      "learning_rate": 4.60168284789644e-06,
      "loss": 0.0733,
      "step": 297380
    },
    {
      "epoch": 115.49,
      "learning_rate": 4.60116504854369e-06,
      "loss": 0.0249,
      "step": 297390
    },
    {
      "epoch": 115.5,
      "learning_rate": 4.600647249190938e-06,
      "loss": 0.0622,
      "step": 297400
    },
    {
      "epoch": 115.5,
      "learning_rate": 4.600129449838188e-06,
      "loss": 0.0164,
      "step": 297410
    },
    {
      "epoch": 115.5,
      "learning_rate": 4.599611650485437e-06,
      "loss": 0.051,
      "step": 297420
    },
    {
      "epoch": 115.51,
      "learning_rate": 4.599093851132686e-06,
      "loss": 0.1509,
      "step": 297430
    },
    {
      "epoch": 115.51,
      "learning_rate": 4.598576051779935e-06,
      "loss": 0.0602,
      "step": 297440
    },
    {
      "epoch": 115.51,
      "learning_rate": 4.598058252427185e-06,
      "loss": 0.3053,
      "step": 297450
    },
    {
      "epoch": 115.52,
      "learning_rate": 4.5975404530744336e-06,
      "loss": 0.0214,
      "step": 297460
    },
    {
      "epoch": 115.52,
      "learning_rate": 4.597022653721683e-06,
      "loss": 0.0753,
      "step": 297470
    },
    {
      "epoch": 115.53,
      "learning_rate": 4.596504854368932e-06,
      "loss": 0.0473,
      "step": 297480
    },
    {
      "epoch": 115.53,
      "learning_rate": 4.595987055016182e-06,
      "loss": 0.0245,
      "step": 297490
    },
    {
      "epoch": 115.53,
      "learning_rate": 4.59546925566343e-06,
      "loss": 0.037,
      "step": 297500
    },
    {
      "epoch": 115.54,
      "learning_rate": 4.59495145631068e-06,
      "loss": 0.0454,
      "step": 297510
    },
    {
      "epoch": 115.54,
      "learning_rate": 4.594433656957929e-06,
      "loss": 0.1222,
      "step": 297520
    },
    {
      "epoch": 115.55,
      "learning_rate": 4.593915857605178e-06,
      "loss": 0.0762,
      "step": 297530
    },
    {
      "epoch": 115.55,
      "learning_rate": 4.593398058252427e-06,
      "loss": 0.0007,
      "step": 297540
    },
    {
      "epoch": 115.55,
      "learning_rate": 4.592880258899677e-06,
      "loss": 0.1968,
      "step": 297550
    },
    {
      "epoch": 115.56,
      "learning_rate": 4.5923624595469255e-06,
      "loss": 0.0475,
      "step": 297560
    },
    {
      "epoch": 115.56,
      "learning_rate": 4.591844660194175e-06,
      "loss": 0.1319,
      "step": 297570
    },
    {
      "epoch": 115.57,
      "learning_rate": 4.591326860841424e-06,
      "loss": 0.0353,
      "step": 297580
    },
    {
      "epoch": 115.57,
      "learning_rate": 4.5908090614886735e-06,
      "loss": 0.0353,
      "step": 297590
    },
    {
      "epoch": 115.57,
      "learning_rate": 4.590291262135922e-06,
      "loss": 0.0479,
      "step": 297600
    },
    {
      "epoch": 115.58,
      "learning_rate": 4.589773462783172e-06,
      "loss": 0.0265,
      "step": 297610
    },
    {
      "epoch": 115.58,
      "learning_rate": 4.589255663430421e-06,
      "loss": 0.0007,
      "step": 297620
    },
    {
      "epoch": 115.58,
      "learning_rate": 4.58873786407767e-06,
      "loss": 0.0528,
      "step": 297630
    },
    {
      "epoch": 115.59,
      "learning_rate": 4.588220064724919e-06,
      "loss": 0.0183,
      "step": 297640
    },
    {
      "epoch": 115.59,
      "learning_rate": 4.587702265372169e-06,
      "loss": 0.0191,
      "step": 297650
    },
    {
      "epoch": 115.6,
      "learning_rate": 4.5871844660194175e-06,
      "loss": 0.1009,
      "step": 297660
    },
    {
      "epoch": 115.6,
      "learning_rate": 4.586666666666667e-06,
      "loss": 0.1474,
      "step": 297670
    },
    {
      "epoch": 115.6,
      "learning_rate": 4.586148867313917e-06,
      "loss": 0.0027,
      "step": 297680
    },
    {
      "epoch": 115.61,
      "learning_rate": 4.5856310679611655e-06,
      "loss": 0.0256,
      "step": 297690
    },
    {
      "epoch": 115.61,
      "learning_rate": 4.585113268608415e-06,
      "loss": 0.0004,
      "step": 297700
    },
    {
      "epoch": 115.62,
      "learning_rate": 4.584595469255664e-06,
      "loss": 0.0757,
      "step": 297710
    },
    {
      "epoch": 115.62,
      "learning_rate": 4.5840776699029135e-06,
      "loss": 0.0632,
      "step": 297720
    },
    {
      "epoch": 115.62,
      "learning_rate": 4.583559870550162e-06,
      "loss": 0.022,
      "step": 297730
    },
    {
      "epoch": 115.63,
      "learning_rate": 4.583042071197412e-06,
      "loss": 0.157,
      "step": 297740
    },
    {
      "epoch": 115.63,
      "learning_rate": 4.582524271844661e-06,
      "loss": 0.0114,
      "step": 297750
    },
    {
      "epoch": 115.63,
      "learning_rate": 4.58200647249191e-06,
      "loss": 0.0002,
      "step": 297760
    },
    {
      "epoch": 115.64,
      "learning_rate": 4.581488673139159e-06,
      "loss": 0.1108,
      "step": 297770
    },
    {
      "epoch": 115.64,
      "learning_rate": 4.580970873786409e-06,
      "loss": 0.001,
      "step": 297780
    },
    {
      "epoch": 115.65,
      "learning_rate": 4.580453074433657e-06,
      "loss": 0.0119,
      "step": 297790
    },
    {
      "epoch": 115.65,
      "learning_rate": 4.579935275080907e-06,
      "loss": 0.0092,
      "step": 297800
    },
    {
      "epoch": 115.65,
      "learning_rate": 4.579417475728156e-06,
      "loss": 0.0454,
      "step": 297810
    },
    {
      "epoch": 115.66,
      "learning_rate": 4.578899676375405e-06,
      "loss": 0.0329,
      "step": 297820
    },
    {
      "epoch": 115.66,
      "learning_rate": 4.578381877022654e-06,
      "loss": 0.1952,
      "step": 297830
    },
    {
      "epoch": 115.67,
      "learning_rate": 4.577864077669904e-06,
      "loss": 0.0328,
      "step": 297840
    },
    {
      "epoch": 115.67,
      "learning_rate": 4.577346278317153e-06,
      "loss": 0.1247,
      "step": 297850
    },
    {
      "epoch": 115.67,
      "learning_rate": 4.576828478964402e-06,
      "loss": 0.0004,
      "step": 297860
    },
    {
      "epoch": 115.68,
      "learning_rate": 4.576310679611651e-06,
      "loss": 0.0253,
      "step": 297870
    },
    {
      "epoch": 115.68,
      "learning_rate": 4.575792880258901e-06,
      "loss": 0.0011,
      "step": 297880
    },
    {
      "epoch": 115.69,
      "learning_rate": 4.575275080906149e-06,
      "loss": 0.1075,
      "step": 297890
    },
    {
      "epoch": 115.69,
      "learning_rate": 4.574757281553399e-06,
      "loss": 0.0002,
      "step": 297900
    },
    {
      "epoch": 115.69,
      "learning_rate": 4.574239482200648e-06,
      "loss": 0.0022,
      "step": 297910
    },
    {
      "epoch": 115.7,
      "learning_rate": 4.573721682847897e-06,
      "loss": 0.2034,
      "step": 297920
    },
    {
      "epoch": 115.7,
      "learning_rate": 4.573203883495146e-06,
      "loss": 0.1292,
      "step": 297930
    },
    {
      "epoch": 115.7,
      "learning_rate": 4.572686084142396e-06,
      "loss": 0.0522,
      "step": 297940
    },
    {
      "epoch": 115.71,
      "learning_rate": 4.5721682847896445e-06,
      "loss": 0.0063,
      "step": 297950
    },
    {
      "epoch": 115.71,
      "learning_rate": 4.571650485436893e-06,
      "loss": 0.0381,
      "step": 297960
    },
    {
      "epoch": 115.72,
      "learning_rate": 4.571132686084143e-06,
      "loss": 0.0001,
      "step": 297970
    },
    {
      "epoch": 115.72,
      "learning_rate": 4.570614886731392e-06,
      "loss": 0.0399,
      "step": 297980
    },
    {
      "epoch": 115.72,
      "learning_rate": 4.570097087378641e-06,
      "loss": 0.0206,
      "step": 297990
    },
    {
      "epoch": 115.73,
      "learning_rate": 4.56957928802589e-06,
      "loss": 0.0299,
      "step": 298000
    },
    {
      "epoch": 115.73,
      "learning_rate": 4.56906148867314e-06,
      "loss": 0.0667,
      "step": 298010
    },
    {
      "epoch": 115.74,
      "learning_rate": 4.5685436893203885e-06,
      "loss": 0.1092,
      "step": 298020
    },
    {
      "epoch": 115.74,
      "learning_rate": 4.568025889967638e-06,
      "loss": 0.0077,
      "step": 298030
    },
    {
      "epoch": 115.74,
      "learning_rate": 4.567508090614887e-06,
      "loss": 0.0606,
      "step": 298040
    },
    {
      "epoch": 115.75,
      "learning_rate": 4.5669902912621365e-06,
      "loss": 0.0344,
      "step": 298050
    },
    {
      "epoch": 115.75,
      "learning_rate": 4.566472491909385e-06,
      "loss": 0.0794,
      "step": 298060
    },
    {
      "epoch": 115.76,
      "learning_rate": 4.565954692556635e-06,
      "loss": 0.0781,
      "step": 298070
    },
    {
      "epoch": 115.76,
      "learning_rate": 4.565436893203884e-06,
      "loss": 0.0802,
      "step": 298080
    },
    {
      "epoch": 115.76,
      "learning_rate": 4.564919093851133e-06,
      "loss": 0.0362,
      "step": 298090
    },
    {
      "epoch": 115.77,
      "learning_rate": 4.564401294498382e-06,
      "loss": 0.0203,
      "step": 298100
    },
    {
      "epoch": 115.77,
      "learning_rate": 4.563883495145632e-06,
      "loss": 0.0002,
      "step": 298110
    },
    {
      "epoch": 115.77,
      "learning_rate": 4.56336569579288e-06,
      "loss": 0.073,
      "step": 298120
    },
    {
      "epoch": 115.78,
      "learning_rate": 4.562847896440129e-06,
      "loss": 0.1433,
      "step": 298130
    },
    {
      "epoch": 115.78,
      "learning_rate": 4.562330097087379e-06,
      "loss": 0.0003,
      "step": 298140
    },
    {
      "epoch": 115.79,
      "learning_rate": 4.5618122977346276e-06,
      "loss": 0.0232,
      "step": 298150
    },
    {
      "epoch": 115.79,
      "learning_rate": 4.561294498381877e-06,
      "loss": 0.0291,
      "step": 298160
    },
    {
      "epoch": 115.79,
      "learning_rate": 4.560776699029126e-06,
      "loss": 0.0171,
      "step": 298170
    },
    {
      "epoch": 115.8,
      "learning_rate": 4.560258899676376e-06,
      "loss": 0.0114,
      "step": 298180
    },
    {
      "epoch": 115.8,
      "learning_rate": 4.559741100323624e-06,
      "loss": 0.0002,
      "step": 298190
    },
    {
      "epoch": 115.81,
      "learning_rate": 4.559223300970874e-06,
      "loss": 0.0374,
      "step": 298200
    },
    {
      "epoch": 115.81,
      "learning_rate": 4.558705501618123e-06,
      "loss": 0.0869,
      "step": 298210
    },
    {
      "epoch": 115.81,
      "learning_rate": 4.558187702265372e-06,
      "loss": 0.0261,
      "step": 298220
    },
    {
      "epoch": 115.82,
      "learning_rate": 4.557669902912621e-06,
      "loss": 0.0007,
      "step": 298230
    },
    {
      "epoch": 115.82,
      "learning_rate": 4.557152103559871e-06,
      "loss": 0.0012,
      "step": 298240
    },
    {
      "epoch": 115.83,
      "learning_rate": 4.55663430420712e-06,
      "loss": 0.0001,
      "step": 298250
    },
    {
      "epoch": 115.83,
      "learning_rate": 4.556116504854369e-06,
      "loss": 0.0752,
      "step": 298260
    },
    {
      "epoch": 115.83,
      "learning_rate": 4.555598705501619e-06,
      "loss": 0.0882,
      "step": 298270
    },
    {
      "epoch": 115.84,
      "learning_rate": 4.5550809061488675e-06,
      "loss": 0.0173,
      "step": 298280
    },
    {
      "epoch": 115.84,
      "learning_rate": 4.554563106796117e-06,
      "loss": 0.1275,
      "step": 298290
    },
    {
      "epoch": 115.84,
      "learning_rate": 4.554045307443366e-06,
      "loss": 0.1937,
      "step": 298300
    },
    {
      "epoch": 115.85,
      "learning_rate": 4.5535275080906155e-06,
      "loss": 0.0159,
      "step": 298310
    },
    {
      "epoch": 115.85,
      "learning_rate": 4.553009708737864e-06,
      "loss": 0.1134,
      "step": 298320
    },
    {
      "epoch": 115.86,
      "learning_rate": 4.552491909385114e-06,
      "loss": 0.0878,
      "step": 298330
    },
    {
      "epoch": 115.86,
      "learning_rate": 4.551974110032363e-06,
      "loss": 0.0133,
      "step": 298340
    },
    {
      "epoch": 115.86,
      "learning_rate": 4.551456310679612e-06,
      "loss": 0.0228,
      "step": 298350
    },
    {
      "epoch": 115.87,
      "learning_rate": 4.550938511326861e-06,
      "loss": 0.1829,
      "step": 298360
    },
    {
      "epoch": 115.87,
      "learning_rate": 4.550420711974111e-06,
      "loss": 0.0829,
      "step": 298370
    },
    {
      "epoch": 115.88,
      "learning_rate": 4.5499029126213595e-06,
      "loss": 0.0125,
      "step": 298380
    },
    {
      "epoch": 115.88,
      "learning_rate": 4.549385113268609e-06,
      "loss": 0.0459,
      "step": 298390
    },
    {
      "epoch": 115.88,
      "learning_rate": 4.548867313915858e-06,
      "loss": 0.0296,
      "step": 298400
    },
    {
      "epoch": 115.89,
      "learning_rate": 4.5483495145631075e-06,
      "loss": 0.0515,
      "step": 298410
    },
    {
      "epoch": 115.89,
      "learning_rate": 4.547831715210356e-06,
      "loss": 0.0344,
      "step": 298420
    },
    {
      "epoch": 115.9,
      "learning_rate": 4.547313915857606e-06,
      "loss": 0.0566,
      "step": 298430
    },
    {
      "epoch": 115.9,
      "learning_rate": 4.546796116504855e-06,
      "loss": 0.0276,
      "step": 298440
    },
    {
      "epoch": 115.9,
      "learning_rate": 4.546278317152104e-06,
      "loss": 0.0206,
      "step": 298450
    },
    {
      "epoch": 115.91,
      "learning_rate": 4.545760517799353e-06,
      "loss": 0.0355,
      "step": 298460
    },
    {
      "epoch": 115.91,
      "learning_rate": 4.545242718446603e-06,
      "loss": 0.0126,
      "step": 298470
    },
    {
      "epoch": 115.91,
      "learning_rate": 4.544724919093851e-06,
      "loss": 0.0023,
      "step": 298480
    },
    {
      "epoch": 115.92,
      "learning_rate": 4.544207119741101e-06,
      "loss": 0.0583,
      "step": 298490
    },
    {
      "epoch": 115.92,
      "learning_rate": 4.54368932038835e-06,
      "loss": 0.0097,
      "step": 298500
    },
    {
      "epoch": 115.93,
      "learning_rate": 4.543171521035599e-06,
      "loss": 0.0279,
      "step": 298510
    },
    {
      "epoch": 115.93,
      "learning_rate": 4.542653721682848e-06,
      "loss": 0.021,
      "step": 298520
    },
    {
      "epoch": 115.93,
      "learning_rate": 4.542135922330098e-06,
      "loss": 0.1048,
      "step": 298530
    },
    {
      "epoch": 115.94,
      "learning_rate": 4.541618122977347e-06,
      "loss": 0.0803,
      "step": 298540
    },
    {
      "epoch": 115.94,
      "learning_rate": 4.541100323624596e-06,
      "loss": 0.0958,
      "step": 298550
    },
    {
      "epoch": 115.95,
      "learning_rate": 4.540582524271845e-06,
      "loss": 0.0017,
      "step": 298560
    },
    {
      "epoch": 115.95,
      "learning_rate": 4.540064724919095e-06,
      "loss": 0.0295,
      "step": 298570
    },
    {
      "epoch": 115.95,
      "learning_rate": 4.539546925566343e-06,
      "loss": 0.1096,
      "step": 298580
    },
    {
      "epoch": 115.96,
      "learning_rate": 4.539029126213593e-06,
      "loss": 0.0395,
      "step": 298590
    },
    {
      "epoch": 115.96,
      "learning_rate": 4.538511326860842e-06,
      "loss": 0.1657,
      "step": 298600
    },
    {
      "epoch": 115.97,
      "learning_rate": 4.537993527508091e-06,
      "loss": 0.0762,
      "step": 298610
    },
    {
      "epoch": 115.97,
      "learning_rate": 4.53747572815534e-06,
      "loss": 0.0083,
      "step": 298620
    },
    {
      "epoch": 115.97,
      "learning_rate": 4.53695792880259e-06,
      "loss": 0.1394,
      "step": 298630
    },
    {
      "epoch": 115.98,
      "learning_rate": 4.5364401294498385e-06,
      "loss": 0.0176,
      "step": 298640
    },
    {
      "epoch": 115.98,
      "learning_rate": 4.535922330097088e-06,
      "loss": 0.1013,
      "step": 298650
    },
    {
      "epoch": 115.98,
      "learning_rate": 4.535404530744337e-06,
      "loss": 0.0002,
      "step": 298660
    },
    {
      "epoch": 115.99,
      "learning_rate": 4.5348867313915865e-06,
      "loss": 0.0107,
      "step": 298670
    },
    {
      "epoch": 115.99,
      "learning_rate": 4.534368932038835e-06,
      "loss": 0.0096,
      "step": 298680
    },
    {
      "epoch": 116.0,
      "learning_rate": 4.533851132686085e-06,
      "loss": 0.0742,
      "step": 298690
    },
    {
      "epoch": 116.0,
      "learning_rate": 4.533333333333334e-06,
      "loss": 0.0203,
      "step": 298700
    },
    {
      "epoch": 116.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.37935107946395874,
      "eval_runtime": 8.2562,
      "eval_samples_per_second": 440.273,
      "eval_steps_per_second": 55.11,
      "step": 298700
    },
    {
      "epoch": 116.0,
      "learning_rate": 4.532815533980583e-06,
      "loss": 0.0373,
      "step": 298710
    },
    {
      "epoch": 116.01,
      "learning_rate": 4.532297734627832e-06,
      "loss": 0.0795,
      "step": 298720
    },
    {
      "epoch": 116.01,
      "learning_rate": 4.531779935275082e-06,
      "loss": 0.0023,
      "step": 298730
    },
    {
      "epoch": 116.02,
      "learning_rate": 4.5312621359223305e-06,
      "loss": 0.0417,
      "step": 298740
    },
    {
      "epoch": 116.02,
      "learning_rate": 4.53074433656958e-06,
      "loss": 0.0005,
      "step": 298750
    },
    {
      "epoch": 116.02,
      "learning_rate": 4.530226537216829e-06,
      "loss": 0.019,
      "step": 298760
    },
    {
      "epoch": 116.03,
      "learning_rate": 4.529708737864078e-06,
      "loss": 0.017,
      "step": 298770
    },
    {
      "epoch": 116.03,
      "learning_rate": 4.529190938511327e-06,
      "loss": 0.0003,
      "step": 298780
    },
    {
      "epoch": 116.03,
      "learning_rate": 4.528673139158576e-06,
      "loss": 0.0016,
      "step": 298790
    },
    {
      "epoch": 116.04,
      "learning_rate": 4.528155339805826e-06,
      "loss": 0.0219,
      "step": 298800
    },
    {
      "epoch": 116.04,
      "learning_rate": 4.527637540453074e-06,
      "loss": 0.0092,
      "step": 298810
    },
    {
      "epoch": 116.05,
      "learning_rate": 4.527119741100324e-06,
      "loss": 0.0015,
      "step": 298820
    },
    {
      "epoch": 116.05,
      "learning_rate": 4.526601941747573e-06,
      "loss": 0.0542,
      "step": 298830
    },
    {
      "epoch": 116.05,
      "learning_rate": 4.526084142394822e-06,
      "loss": 0.0408,
      "step": 298840
    },
    {
      "epoch": 116.06,
      "learning_rate": 4.525566343042071e-06,
      "loss": 0.0858,
      "step": 298850
    },
    {
      "epoch": 116.06,
      "learning_rate": 4.525048543689321e-06,
      "loss": 0.0039,
      "step": 298860
    },
    {
      "epoch": 116.07,
      "learning_rate": 4.5245307443365696e-06,
      "loss": 0.07,
      "step": 298870
    },
    {
      "epoch": 116.07,
      "learning_rate": 4.524012944983819e-06,
      "loss": 0.0049,
      "step": 298880
    },
    {
      "epoch": 116.07,
      "learning_rate": 4.523495145631068e-06,
      "loss": 0.054,
      "step": 298890
    },
    {
      "epoch": 116.08,
      "learning_rate": 4.522977346278318e-06,
      "loss": 0.0335,
      "step": 298900
    },
    {
      "epoch": 116.08,
      "learning_rate": 4.522459546925566e-06,
      "loss": 0.0118,
      "step": 298910
    },
    {
      "epoch": 116.09,
      "learning_rate": 4.521941747572816e-06,
      "loss": 0.0534,
      "step": 298920
    },
    {
      "epoch": 116.09,
      "learning_rate": 4.521423948220065e-06,
      "loss": 0.008,
      "step": 298930
    },
    {
      "epoch": 116.09,
      "learning_rate": 4.520906148867314e-06,
      "loss": 0.0876,
      "step": 298940
    },
    {
      "epoch": 116.1,
      "learning_rate": 4.520388349514563e-06,
      "loss": 0.0794,
      "step": 298950
    },
    {
      "epoch": 116.1,
      "learning_rate": 4.519870550161813e-06,
      "loss": 0.1365,
      "step": 298960
    },
    {
      "epoch": 116.1,
      "learning_rate": 4.5193527508090615e-06,
      "loss": 0.0957,
      "step": 298970
    },
    {
      "epoch": 116.11,
      "learning_rate": 4.518834951456311e-06,
      "loss": 0.0628,
      "step": 298980
    },
    {
      "epoch": 116.11,
      "learning_rate": 4.51831715210356e-06,
      "loss": 0.0608,
      "step": 298990
    },
    {
      "epoch": 116.12,
      "learning_rate": 4.5177993527508095e-06,
      "loss": 0.0007,
      "step": 299000
    },
    {
      "epoch": 116.12,
      "learning_rate": 4.517281553398058e-06,
      "loss": 0.025,
      "step": 299010
    },
    {
      "epoch": 116.12,
      "learning_rate": 4.516763754045308e-06,
      "loss": 0.0491,
      "step": 299020
    },
    {
      "epoch": 116.13,
      "learning_rate": 4.516245954692557e-06,
      "loss": 0.0249,
      "step": 299030
    },
    {
      "epoch": 116.13,
      "learning_rate": 4.515728155339806e-06,
      "loss": 0.0203,
      "step": 299040
    },
    {
      "epoch": 116.14,
      "learning_rate": 4.515210355987055e-06,
      "loss": 0.0002,
      "step": 299050
    },
    {
      "epoch": 116.14,
      "learning_rate": 4.514692556634305e-06,
      "loss": 0.007,
      "step": 299060
    },
    {
      "epoch": 116.14,
      "learning_rate": 4.5141747572815535e-06,
      "loss": 0.0279,
      "step": 299070
    },
    {
      "epoch": 116.15,
      "learning_rate": 4.513656957928803e-06,
      "loss": 0.0917,
      "step": 299080
    },
    {
      "epoch": 116.15,
      "learning_rate": 4.513139158576052e-06,
      "loss": 0.0259,
      "step": 299090
    },
    {
      "epoch": 116.16,
      "learning_rate": 4.5126213592233015e-06,
      "loss": 0.025,
      "step": 299100
    },
    {
      "epoch": 116.16,
      "learning_rate": 4.51210355987055e-06,
      "loss": 0.0735,
      "step": 299110
    },
    {
      "epoch": 116.16,
      "learning_rate": 4.5115857605178e-06,
      "loss": 0.1106,
      "step": 299120
    },
    {
      "epoch": 116.17,
      "learning_rate": 4.511067961165049e-06,
      "loss": 0.0145,
      "step": 299130
    },
    {
      "epoch": 116.17,
      "learning_rate": 4.510550161812298e-06,
      "loss": 0.0374,
      "step": 299140
    },
    {
      "epoch": 116.17,
      "learning_rate": 4.510032362459547e-06,
      "loss": 0.1241,
      "step": 299150
    },
    {
      "epoch": 116.18,
      "learning_rate": 4.509514563106797e-06,
      "loss": 0.0017,
      "step": 299160
    },
    {
      "epoch": 116.18,
      "learning_rate": 4.508996763754045e-06,
      "loss": 0.0948,
      "step": 299170
    },
    {
      "epoch": 116.19,
      "learning_rate": 4.508478964401295e-06,
      "loss": 0.0465,
      "step": 299180
    },
    {
      "epoch": 116.19,
      "learning_rate": 4.507961165048544e-06,
      "loss": 0.2227,
      "step": 299190
    },
    {
      "epoch": 116.19,
      "learning_rate": 4.507443365695793e-06,
      "loss": 0.1159,
      "step": 299200
    },
    {
      "epoch": 116.2,
      "learning_rate": 4.506925566343042e-06,
      "loss": 0.0024,
      "step": 299210
    },
    {
      "epoch": 116.2,
      "learning_rate": 4.506407766990292e-06,
      "loss": 0.0709,
      "step": 299220
    },
    {
      "epoch": 116.21,
      "learning_rate": 4.505889967637541e-06,
      "loss": 0.0003,
      "step": 299230
    },
    {
      "epoch": 116.21,
      "learning_rate": 4.50537216828479e-06,
      "loss": 0.0331,
      "step": 299240
    },
    {
      "epoch": 116.21,
      "learning_rate": 4.504854368932039e-06,
      "loss": 0.0022,
      "step": 299250
    },
    {
      "epoch": 116.22,
      "learning_rate": 4.504336569579289e-06,
      "loss": 0.0866,
      "step": 299260
    },
    {
      "epoch": 116.22,
      "learning_rate": 4.503818770226537e-06,
      "loss": 0.1539,
      "step": 299270
    },
    {
      "epoch": 116.23,
      "learning_rate": 4.503300970873787e-06,
      "loss": 0.012,
      "step": 299280
    },
    {
      "epoch": 116.23,
      "learning_rate": 4.502783171521036e-06,
      "loss": 0.0919,
      "step": 299290
    },
    {
      "epoch": 116.23,
      "learning_rate": 4.502265372168285e-06,
      "loss": 0.0494,
      "step": 299300
    },
    {
      "epoch": 116.24,
      "learning_rate": 4.501747572815534e-06,
      "loss": 0.0771,
      "step": 299310
    },
    {
      "epoch": 116.24,
      "learning_rate": 4.501229773462784e-06,
      "loss": 0.0545,
      "step": 299320
    },
    {
      "epoch": 116.24,
      "learning_rate": 4.5007119741100325e-06,
      "loss": 0.0165,
      "step": 299330
    },
    {
      "epoch": 116.25,
      "learning_rate": 4.500194174757282e-06,
      "loss": 0.0463,
      "step": 299340
    },
    {
      "epoch": 116.25,
      "learning_rate": 4.499676375404531e-06,
      "loss": 0.0261,
      "step": 299350
    },
    {
      "epoch": 116.26,
      "learning_rate": 4.4991585760517805e-06,
      "loss": 0.1425,
      "step": 299360
    },
    {
      "epoch": 116.26,
      "learning_rate": 4.49864077669903e-06,
      "loss": 0.018,
      "step": 299370
    },
    {
      "epoch": 116.26,
      "learning_rate": 4.498122977346279e-06,
      "loss": 0.0809,
      "step": 299380
    },
    {
      "epoch": 116.27,
      "learning_rate": 4.4976051779935285e-06,
      "loss": 0.0059,
      "step": 299390
    },
    {
      "epoch": 116.27,
      "learning_rate": 4.497087378640777e-06,
      "loss": 0.0193,
      "step": 299400
    },
    {
      "epoch": 116.28,
      "learning_rate": 4.496569579288027e-06,
      "loss": 0.0055,
      "step": 299410
    },
    {
      "epoch": 116.28,
      "learning_rate": 4.496051779935276e-06,
      "loss": 0.0257,
      "step": 299420
    },
    {
      "epoch": 116.28,
      "learning_rate": 4.4955339805825245e-06,
      "loss": 0.0249,
      "step": 299430
    },
    {
      "epoch": 116.29,
      "learning_rate": 4.495016181229774e-06,
      "loss": 0.0552,
      "step": 299440
    },
    {
      "epoch": 116.29,
      "learning_rate": 4.494498381877023e-06,
      "loss": 0.015,
      "step": 299450
    },
    {
      "epoch": 116.3,
      "learning_rate": 4.4939805825242725e-06,
      "loss": 0.0667,
      "step": 299460
    },
    {
      "epoch": 116.3,
      "learning_rate": 4.493462783171521e-06,
      "loss": 0.0134,
      "step": 299470
    },
    {
      "epoch": 116.3,
      "learning_rate": 4.492944983818771e-06,
      "loss": 0.0671,
      "step": 299480
    },
    {
      "epoch": 116.31,
      "learning_rate": 4.49242718446602e-06,
      "loss": 0.0319,
      "step": 299490
    },
    {
      "epoch": 116.31,
      "learning_rate": 4.491909385113269e-06,
      "loss": 0.0468,
      "step": 299500
    },
    {
      "epoch": 116.31,
      "learning_rate": 4.491391585760518e-06,
      "loss": 0.0952,
      "step": 299510
    },
    {
      "epoch": 116.32,
      "learning_rate": 4.490873786407768e-06,
      "loss": 0.0319,
      "step": 299520
    },
    {
      "epoch": 116.32,
      "learning_rate": 4.490355987055016e-06,
      "loss": 0.0938,
      "step": 299530
    },
    {
      "epoch": 116.33,
      "learning_rate": 4.489838187702266e-06,
      "loss": 0.057,
      "step": 299540
    },
    {
      "epoch": 116.33,
      "learning_rate": 4.489320388349515e-06,
      "loss": 0.0547,
      "step": 299550
    },
    {
      "epoch": 116.33,
      "learning_rate": 4.488802588996764e-06,
      "loss": 0.0011,
      "step": 299560
    },
    {
      "epoch": 116.34,
      "learning_rate": 4.488284789644013e-06,
      "loss": 0.0766,
      "step": 299570
    },
    {
      "epoch": 116.34,
      "learning_rate": 4.487766990291263e-06,
      "loss": 0.0716,
      "step": 299580
    },
    {
      "epoch": 116.35,
      "learning_rate": 4.487249190938512e-06,
      "loss": 0.0984,
      "step": 299590
    },
    {
      "epoch": 116.35,
      "learning_rate": 4.48673139158576e-06,
      "loss": 0.1083,
      "step": 299600
    },
    {
      "epoch": 116.35,
      "learning_rate": 4.48621359223301e-06,
      "loss": 0.0217,
      "step": 299610
    },
    {
      "epoch": 116.36,
      "learning_rate": 4.485695792880259e-06,
      "loss": 0.0437,
      "step": 299620
    },
    {
      "epoch": 116.36,
      "learning_rate": 4.485177993527508e-06,
      "loss": 0.024,
      "step": 299630
    },
    {
      "epoch": 116.37,
      "learning_rate": 4.484660194174757e-06,
      "loss": 0.0812,
      "step": 299640
    },
    {
      "epoch": 116.37,
      "learning_rate": 4.484142394822007e-06,
      "loss": 0.0641,
      "step": 299650
    },
    {
      "epoch": 116.37,
      "learning_rate": 4.4836245954692555e-06,
      "loss": 0.035,
      "step": 299660
    },
    {
      "epoch": 116.38,
      "learning_rate": 4.483106796116505e-06,
      "loss": 0.1,
      "step": 299670
    },
    {
      "epoch": 116.38,
      "learning_rate": 4.482588996763754e-06,
      "loss": 0.0094,
      "step": 299680
    },
    {
      "epoch": 116.38,
      "learning_rate": 4.4820711974110035e-06,
      "loss": 0.0172,
      "step": 299690
    },
    {
      "epoch": 116.39,
      "learning_rate": 4.481553398058252e-06,
      "loss": 0.0897,
      "step": 299700
    },
    {
      "epoch": 116.39,
      "learning_rate": 4.481035598705502e-06,
      "loss": 0.0003,
      "step": 299710
    },
    {
      "epoch": 116.4,
      "learning_rate": 4.480517799352751e-06,
      "loss": 0.085,
      "step": 299720
    },
    {
      "epoch": 116.4,
      "learning_rate": 4.48e-06,
      "loss": 0.0052,
      "step": 299730
    },
    {
      "epoch": 116.4,
      "learning_rate": 4.479482200647249e-06,
      "loss": 0.0171,
      "step": 299740
    },
    {
      "epoch": 116.41,
      "learning_rate": 4.478964401294499e-06,
      "loss": 0.0936,
      "step": 299750
    },
    {
      "epoch": 116.41,
      "learning_rate": 4.4784466019417475e-06,
      "loss": 0.0084,
      "step": 299760
    },
    {
      "epoch": 116.42,
      "learning_rate": 4.477928802588997e-06,
      "loss": 0.0417,
      "step": 299770
    },
    {
      "epoch": 116.42,
      "learning_rate": 4.477411003236246e-06,
      "loss": 0.1116,
      "step": 299780
    },
    {
      "epoch": 116.42,
      "learning_rate": 4.4768932038834955e-06,
      "loss": 0.0171,
      "step": 299790
    },
    {
      "epoch": 116.43,
      "learning_rate": 4.476375404530744e-06,
      "loss": 0.0118,
      "step": 299800
    },
    {
      "epoch": 116.43,
      "learning_rate": 4.475857605177994e-06,
      "loss": 0.1502,
      "step": 299810
    },
    {
      "epoch": 116.43,
      "learning_rate": 4.475339805825243e-06,
      "loss": 0.0095,
      "step": 299820
    },
    {
      "epoch": 116.44,
      "learning_rate": 4.474822006472492e-06,
      "loss": 0.0266,
      "step": 299830
    },
    {
      "epoch": 116.44,
      "learning_rate": 4.474304207119741e-06,
      "loss": 0.1747,
      "step": 299840
    },
    {
      "epoch": 116.45,
      "learning_rate": 4.473786407766991e-06,
      "loss": 0.0322,
      "step": 299850
    },
    {
      "epoch": 116.45,
      "learning_rate": 4.473268608414239e-06,
      "loss": 0.0068,
      "step": 299860
    },
    {
      "epoch": 116.45,
      "learning_rate": 4.472750809061489e-06,
      "loss": 0.0007,
      "step": 299870
    },
    {
      "epoch": 116.46,
      "learning_rate": 4.472233009708738e-06,
      "loss": 0.2051,
      "step": 299880
    },
    {
      "epoch": 116.46,
      "learning_rate": 4.471715210355987e-06,
      "loss": 0.0329,
      "step": 299890
    },
    {
      "epoch": 116.47,
      "learning_rate": 4.471197411003236e-06,
      "loss": 0.0328,
      "step": 299900
    },
    {
      "epoch": 116.47,
      "learning_rate": 4.470679611650486e-06,
      "loss": 0.0478,
      "step": 299910
    },
    {
      "epoch": 116.47,
      "learning_rate": 4.4701618122977354e-06,
      "loss": 0.0165,
      "step": 299920
    },
    {
      "epoch": 116.48,
      "learning_rate": 4.469644012944984e-06,
      "loss": 0.0009,
      "step": 299930
    },
    {
      "epoch": 116.48,
      "learning_rate": 4.469126213592234e-06,
      "loss": 0.0435,
      "step": 299940
    },
    {
      "epoch": 116.49,
      "learning_rate": 4.468608414239483e-06,
      "loss": 0.1248,
      "step": 299950
    },
    {
      "epoch": 116.49,
      "learning_rate": 4.468090614886732e-06,
      "loss": 0.0867,
      "step": 299960
    },
    {
      "epoch": 116.49,
      "learning_rate": 4.467572815533981e-06,
      "loss": 0.0337,
      "step": 299970
    },
    {
      "epoch": 116.5,
      "learning_rate": 4.467055016181231e-06,
      "loss": 0.0196,
      "step": 299980
    },
    {
      "epoch": 116.5,
      "learning_rate": 4.466537216828479e-06,
      "loss": 0.0809,
      "step": 299990
    },
    {
      "epoch": 116.5,
      "learning_rate": 4.466019417475729e-06,
      "loss": 0.0393,
      "step": 300000
    },
    {
      "epoch": 116.51,
      "learning_rate": 4.465501618122978e-06,
      "loss": 0.0333,
      "step": 300010
    },
    {
      "epoch": 116.51,
      "learning_rate": 4.464983818770227e-06,
      "loss": 0.1105,
      "step": 300020
    },
    {
      "epoch": 116.52,
      "learning_rate": 4.464466019417476e-06,
      "loss": 0.0455,
      "step": 300030
    },
    {
      "epoch": 116.52,
      "learning_rate": 4.463948220064726e-06,
      "loss": 0.0004,
      "step": 300040
    },
    {
      "epoch": 116.52,
      "learning_rate": 4.4634304207119745e-06,
      "loss": 0.081,
      "step": 300050
    },
    {
      "epoch": 116.53,
      "learning_rate": 4.462912621359224e-06,
      "loss": 0.0003,
      "step": 300060
    },
    {
      "epoch": 116.53,
      "learning_rate": 4.462394822006473e-06,
      "loss": 0.0643,
      "step": 300070
    },
    {
      "epoch": 116.54,
      "learning_rate": 4.4618770226537225e-06,
      "loss": 0.0873,
      "step": 300080
    },
    {
      "epoch": 116.54,
      "learning_rate": 4.461359223300971e-06,
      "loss": 0.1652,
      "step": 300090
    },
    {
      "epoch": 116.54,
      "learning_rate": 4.460841423948221e-06,
      "loss": 0.0641,
      "step": 300100
    },
    {
      "epoch": 116.55,
      "learning_rate": 4.46032362459547e-06,
      "loss": 0.0024,
      "step": 300110
    },
    {
      "epoch": 116.55,
      "learning_rate": 4.459805825242719e-06,
      "loss": 0.1002,
      "step": 300120
    },
    {
      "epoch": 116.56,
      "learning_rate": 4.459288025889968e-06,
      "loss": 0.0002,
      "step": 300130
    },
    {
      "epoch": 116.56,
      "learning_rate": 4.458770226537218e-06,
      "loss": 0.0063,
      "step": 300140
    },
    {
      "epoch": 116.56,
      "learning_rate": 4.4582524271844665e-06,
      "loss": 0.0929,
      "step": 300150
    },
    {
      "epoch": 116.57,
      "learning_rate": 4.457734627831716e-06,
      "loss": 0.0004,
      "step": 300160
    },
    {
      "epoch": 116.57,
      "learning_rate": 4.457216828478965e-06,
      "loss": 0.0119,
      "step": 300170
    },
    {
      "epoch": 116.57,
      "learning_rate": 4.4566990291262145e-06,
      "loss": 0.0088,
      "step": 300180
    },
    {
      "epoch": 116.58,
      "learning_rate": 4.456181229773463e-06,
      "loss": 0.0131,
      "step": 300190
    },
    {
      "epoch": 116.58,
      "learning_rate": 4.455663430420713e-06,
      "loss": 0.0214,
      "step": 300200
    },
    {
      "epoch": 116.59,
      "learning_rate": 4.455145631067962e-06,
      "loss": 0.1053,
      "step": 300210
    },
    {
      "epoch": 116.59,
      "learning_rate": 4.454627831715211e-06,
      "loss": 0.1004,
      "step": 300220
    },
    {
      "epoch": 116.59,
      "learning_rate": 4.45411003236246e-06,
      "loss": 0.0315,
      "step": 300230
    },
    {
      "epoch": 116.6,
      "learning_rate": 4.453592233009709e-06,
      "loss": 0.0001,
      "step": 300240
    },
    {
      "epoch": 116.6,
      "learning_rate": 4.453074433656958e-06,
      "loss": 0.0111,
      "step": 300250
    },
    {
      "epoch": 116.61,
      "learning_rate": 4.452556634304207e-06,
      "loss": 0.0044,
      "step": 300260
    },
    {
      "epoch": 116.61,
      "learning_rate": 4.452038834951457e-06,
      "loss": 0.0078,
      "step": 300270
    },
    {
      "epoch": 116.61,
      "learning_rate": 4.451521035598706e-06,
      "loss": 0.0466,
      "step": 300280
    },
    {
      "epoch": 116.62,
      "learning_rate": 4.451003236245955e-06,
      "loss": 0.0928,
      "step": 300290
    },
    {
      "epoch": 116.62,
      "learning_rate": 4.450485436893204e-06,
      "loss": 0.102,
      "step": 300300
    },
    {
      "epoch": 116.63,
      "learning_rate": 4.449967637540454e-06,
      "loss": 0.0977,
      "step": 300310
    },
    {
      "epoch": 116.63,
      "learning_rate": 4.449449838187702e-06,
      "loss": 0.0213,
      "step": 300320
    },
    {
      "epoch": 116.63,
      "learning_rate": 4.448932038834952e-06,
      "loss": 0.0611,
      "step": 300330
    },
    {
      "epoch": 116.64,
      "learning_rate": 4.448414239482201e-06,
      "loss": 0.0184,
      "step": 300340
    },
    {
      "epoch": 116.64,
      "learning_rate": 4.44789644012945e-06,
      "loss": 0.039,
      "step": 300350
    },
    {
      "epoch": 116.64,
      "learning_rate": 4.447378640776699e-06,
      "loss": 0.0931,
      "step": 300360
    },
    {
      "epoch": 116.65,
      "learning_rate": 4.446860841423949e-06,
      "loss": 0.048,
      "step": 300370
    },
    {
      "epoch": 116.65,
      "learning_rate": 4.4463430420711975e-06,
      "loss": 0.0508,
      "step": 300380
    },
    {
      "epoch": 116.66,
      "learning_rate": 4.445825242718447e-06,
      "loss": 0.0772,
      "step": 300390
    },
    {
      "epoch": 116.66,
      "learning_rate": 4.445307443365696e-06,
      "loss": 0.0469,
      "step": 300400
    },
    {
      "epoch": 116.66,
      "learning_rate": 4.444789644012945e-06,
      "loss": 0.0045,
      "step": 300410
    },
    {
      "epoch": 116.67,
      "learning_rate": 4.444271844660194e-06,
      "loss": 0.101,
      "step": 300420
    },
    {
      "epoch": 116.67,
      "learning_rate": 4.443754045307443e-06,
      "loss": 0.0151,
      "step": 300430
    },
    {
      "epoch": 116.68,
      "learning_rate": 4.443236245954693e-06,
      "loss": 0.0022,
      "step": 300440
    },
    {
      "epoch": 116.68,
      "learning_rate": 4.4427184466019415e-06,
      "loss": 0.0115,
      "step": 300450
    },
    {
      "epoch": 116.68,
      "learning_rate": 4.442200647249191e-06,
      "loss": 0.0213,
      "step": 300460
    },
    {
      "epoch": 116.69,
      "learning_rate": 4.44168284789644e-06,
      "loss": 0.0008,
      "step": 300470
    },
    {
      "epoch": 116.69,
      "learning_rate": 4.4411650485436895e-06,
      "loss": 0.0124,
      "step": 300480
    },
    {
      "epoch": 116.7,
      "learning_rate": 4.440647249190939e-06,
      "loss": 0.0265,
      "step": 300490
    },
    {
      "epoch": 116.7,
      "learning_rate": 4.440129449838188e-06,
      "loss": 0.1345,
      "step": 300500
    },
    {
      "epoch": 116.7,
      "learning_rate": 4.4396116504854375e-06,
      "loss": 0.066,
      "step": 300510
    },
    {
      "epoch": 116.71,
      "learning_rate": 4.439093851132686e-06,
      "loss": 0.0521,
      "step": 300520
    },
    {
      "epoch": 116.71,
      "learning_rate": 4.438576051779936e-06,
      "loss": 0.0991,
      "step": 300530
    },
    {
      "epoch": 116.71,
      "learning_rate": 4.438058252427185e-06,
      "loss": 0.0079,
      "step": 300540
    },
    {
      "epoch": 116.72,
      "learning_rate": 4.437540453074434e-06,
      "loss": 0.042,
      "step": 300550
    },
    {
      "epoch": 116.72,
      "learning_rate": 4.437022653721683e-06,
      "loss": 0.0221,
      "step": 300560
    },
    {
      "epoch": 116.73,
      "learning_rate": 4.436504854368933e-06,
      "loss": 0.0079,
      "step": 300570
    },
    {
      "epoch": 116.73,
      "learning_rate": 4.435987055016181e-06,
      "loss": 0.071,
      "step": 300580
    },
    {
      "epoch": 116.73,
      "learning_rate": 4.435469255663431e-06,
      "loss": 0.0355,
      "step": 300590
    },
    {
      "epoch": 116.74,
      "learning_rate": 4.43495145631068e-06,
      "loss": 0.006,
      "step": 300600
    },
    {
      "epoch": 116.74,
      "learning_rate": 4.4344336569579294e-06,
      "loss": 0.1479,
      "step": 300610
    },
    {
      "epoch": 116.75,
      "learning_rate": 4.433915857605178e-06,
      "loss": 0.0105,
      "step": 300620
    },
    {
      "epoch": 116.75,
      "learning_rate": 4.433398058252428e-06,
      "loss": 0.001,
      "step": 300630
    },
    {
      "epoch": 116.75,
      "learning_rate": 4.432880258899677e-06,
      "loss": 0.0436,
      "step": 300640
    },
    {
      "epoch": 116.76,
      "learning_rate": 4.432362459546926e-06,
      "loss": 0.0315,
      "step": 300650
    },
    {
      "epoch": 116.76,
      "learning_rate": 4.431844660194175e-06,
      "loss": 0.002,
      "step": 300660
    },
    {
      "epoch": 116.77,
      "learning_rate": 4.431326860841425e-06,
      "loss": 0.0035,
      "step": 300670
    },
    {
      "epoch": 116.77,
      "learning_rate": 4.430809061488673e-06,
      "loss": 0.0511,
      "step": 300680
    },
    {
      "epoch": 116.77,
      "learning_rate": 4.430291262135923e-06,
      "loss": 0.2045,
      "step": 300690
    },
    {
      "epoch": 116.78,
      "learning_rate": 4.429773462783172e-06,
      "loss": 0.0799,
      "step": 300700
    },
    {
      "epoch": 116.78,
      "learning_rate": 4.429255663430421e-06,
      "loss": 0.0447,
      "step": 300710
    },
    {
      "epoch": 116.78,
      "learning_rate": 4.42873786407767e-06,
      "loss": 0.0187,
      "step": 300720
    },
    {
      "epoch": 116.79,
      "learning_rate": 4.42822006472492e-06,
      "loss": 0.0703,
      "step": 300730
    },
    {
      "epoch": 116.79,
      "learning_rate": 4.4277022653721685e-06,
      "loss": 0.0298,
      "step": 300740
    },
    {
      "epoch": 116.8,
      "learning_rate": 4.427184466019418e-06,
      "loss": 0.0878,
      "step": 300750
    },
    {
      "epoch": 116.8,
      "learning_rate": 4.426666666666667e-06,
      "loss": 0.0034,
      "step": 300760
    },
    {
      "epoch": 116.8,
      "learning_rate": 4.4261488673139165e-06,
      "loss": 0.009,
      "step": 300770
    },
    {
      "epoch": 116.81,
      "learning_rate": 4.425631067961165e-06,
      "loss": 0.1492,
      "step": 300780
    },
    {
      "epoch": 116.81,
      "learning_rate": 4.425113268608415e-06,
      "loss": 0.0201,
      "step": 300790
    },
    {
      "epoch": 116.82,
      "learning_rate": 4.424595469255664e-06,
      "loss": 0.0175,
      "step": 300800
    },
    {
      "epoch": 116.82,
      "learning_rate": 4.424077669902913e-06,
      "loss": 0.0004,
      "step": 300810
    },
    {
      "epoch": 116.82,
      "learning_rate": 4.423559870550162e-06,
      "loss": 0.0101,
      "step": 300820
    },
    {
      "epoch": 116.83,
      "learning_rate": 4.423042071197412e-06,
      "loss": 0.0948,
      "step": 300830
    },
    {
      "epoch": 116.83,
      "learning_rate": 4.4225242718446605e-06,
      "loss": 0.0405,
      "step": 300840
    },
    {
      "epoch": 116.83,
      "learning_rate": 4.42200647249191e-06,
      "loss": 0.1598,
      "step": 300850
    },
    {
      "epoch": 116.84,
      "learning_rate": 4.421488673139159e-06,
      "loss": 0.0317,
      "step": 300860
    },
    {
      "epoch": 116.84,
      "learning_rate": 4.4209708737864085e-06,
      "loss": 0.0252,
      "step": 300870
    },
    {
      "epoch": 116.85,
      "learning_rate": 4.420453074433657e-06,
      "loss": 0.0225,
      "step": 300880
    },
    {
      "epoch": 116.85,
      "learning_rate": 4.419935275080907e-06,
      "loss": 0.0956,
      "step": 300890
    },
    {
      "epoch": 116.85,
      "learning_rate": 4.419417475728156e-06,
      "loss": 0.0278,
      "step": 300900
    },
    {
      "epoch": 116.86,
      "learning_rate": 4.418899676375405e-06,
      "loss": 0.0208,
      "step": 300910
    },
    {
      "epoch": 116.86,
      "learning_rate": 4.418381877022654e-06,
      "loss": 0.021,
      "step": 300920
    },
    {
      "epoch": 116.87,
      "learning_rate": 4.417864077669904e-06,
      "loss": 0.121,
      "step": 300930
    },
    {
      "epoch": 116.87,
      "learning_rate": 4.417346278317152e-06,
      "loss": 0.0678,
      "step": 300940
    },
    {
      "epoch": 116.87,
      "learning_rate": 4.416828478964402e-06,
      "loss": 0.0988,
      "step": 300950
    },
    {
      "epoch": 116.88,
      "learning_rate": 4.416310679611651e-06,
      "loss": 0.0021,
      "step": 300960
    },
    {
      "epoch": 116.88,
      "learning_rate": 4.4157928802589004e-06,
      "loss": 0.094,
      "step": 300970
    },
    {
      "epoch": 116.89,
      "learning_rate": 4.415275080906149e-06,
      "loss": 0.0926,
      "step": 300980
    },
    {
      "epoch": 116.89,
      "learning_rate": 4.414757281553399e-06,
      "loss": 0.1016,
      "step": 300990
    },
    {
      "epoch": 116.89,
      "learning_rate": 4.414239482200648e-06,
      "loss": 0.0228,
      "step": 301000
    },
    {
      "epoch": 116.9,
      "learning_rate": 4.413721682847897e-06,
      "loss": 0.0078,
      "step": 301010
    },
    {
      "epoch": 116.9,
      "learning_rate": 4.413203883495146e-06,
      "loss": 0.0001,
      "step": 301020
    },
    {
      "epoch": 116.9,
      "learning_rate": 4.412686084142396e-06,
      "loss": 0.0257,
      "step": 301030
    },
    {
      "epoch": 116.91,
      "learning_rate": 4.412168284789644e-06,
      "loss": 0.0599,
      "step": 301040
    },
    {
      "epoch": 116.91,
      "learning_rate": 4.411650485436894e-06,
      "loss": 0.0008,
      "step": 301050
    },
    {
      "epoch": 116.92,
      "learning_rate": 4.411132686084143e-06,
      "loss": 0.0001,
      "step": 301060
    },
    {
      "epoch": 116.92,
      "learning_rate": 4.4106148867313915e-06,
      "loss": 0.039,
      "step": 301070
    },
    {
      "epoch": 116.92,
      "learning_rate": 4.410097087378641e-06,
      "loss": 0.0697,
      "step": 301080
    },
    {
      "epoch": 116.93,
      "learning_rate": 4.40957928802589e-06,
      "loss": 0.0142,
      "step": 301090
    },
    {
      "epoch": 116.93,
      "learning_rate": 4.4090614886731395e-06,
      "loss": 0.0325,
      "step": 301100
    },
    {
      "epoch": 116.94,
      "learning_rate": 4.408543689320388e-06,
      "loss": 0.0279,
      "step": 301110
    },
    {
      "epoch": 116.94,
      "learning_rate": 4.408025889967638e-06,
      "loss": 0.2032,
      "step": 301120
    },
    {
      "epoch": 116.94,
      "learning_rate": 4.407508090614887e-06,
      "loss": 0.0297,
      "step": 301130
    },
    {
      "epoch": 116.95,
      "learning_rate": 4.406990291262136e-06,
      "loss": 0.0002,
      "step": 301140
    },
    {
      "epoch": 116.95,
      "learning_rate": 4.406472491909385e-06,
      "loss": 0.1513,
      "step": 301150
    },
    {
      "epoch": 116.96,
      "learning_rate": 4.405954692556635e-06,
      "loss": 0.0395,
      "step": 301160
    },
    {
      "epoch": 116.96,
      "learning_rate": 4.4054368932038835e-06,
      "loss": 0.1121,
      "step": 301170
    },
    {
      "epoch": 116.96,
      "learning_rate": 4.404919093851133e-06,
      "loss": 0.0158,
      "step": 301180
    },
    {
      "epoch": 116.97,
      "learning_rate": 4.404401294498382e-06,
      "loss": 0.037,
      "step": 301190
    },
    {
      "epoch": 116.97,
      "learning_rate": 4.4038834951456315e-06,
      "loss": 0.0156,
      "step": 301200
    },
    {
      "epoch": 116.97,
      "learning_rate": 4.40336569579288e-06,
      "loss": 0.056,
      "step": 301210
    },
    {
      "epoch": 116.98,
      "learning_rate": 4.40284789644013e-06,
      "loss": 0.009,
      "step": 301220
    },
    {
      "epoch": 116.98,
      "learning_rate": 4.402330097087379e-06,
      "loss": 0.0031,
      "step": 301230
    },
    {
      "epoch": 116.99,
      "learning_rate": 4.401812297734628e-06,
      "loss": 0.0228,
      "step": 301240
    },
    {
      "epoch": 116.99,
      "learning_rate": 4.401294498381877e-06,
      "loss": 0.0005,
      "step": 301250
    },
    {
      "epoch": 116.99,
      "learning_rate": 4.400776699029127e-06,
      "loss": 0.1003,
      "step": 301260
    },
    {
      "epoch": 117.0,
      "learning_rate": 4.400258899676375e-06,
      "loss": 0.0437,
      "step": 301270
    },
    {
      "epoch": 117.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.3807455897331238,
      "eval_runtime": 8.1742,
      "eval_samples_per_second": 444.693,
      "eval_steps_per_second": 55.663,
      "step": 301275
    },
    {
      "epoch": 117.0,
      "learning_rate": 4.399741100323625e-06,
      "loss": 0.0054,
      "step": 301280
    },
    {
      "epoch": 117.01,
      "learning_rate": 4.399223300970874e-06,
      "loss": 0.1195,
      "step": 301290
    },
    {
      "epoch": 117.01,
      "learning_rate": 4.398705501618123e-06,
      "loss": 0.0154,
      "step": 301300
    },
    {
      "epoch": 117.01,
      "learning_rate": 4.398187702265372e-06,
      "loss": 0.0301,
      "step": 301310
    },
    {
      "epoch": 117.02,
      "learning_rate": 4.397669902912622e-06,
      "loss": 0.0003,
      "step": 301320
    },
    {
      "epoch": 117.02,
      "learning_rate": 4.397152103559871e-06,
      "loss": 0.0659,
      "step": 301330
    },
    {
      "epoch": 117.03,
      "learning_rate": 4.39663430420712e-06,
      "loss": 0.0549,
      "step": 301340
    },
    {
      "epoch": 117.03,
      "learning_rate": 4.396116504854369e-06,
      "loss": 0.0354,
      "step": 301350
    },
    {
      "epoch": 117.03,
      "learning_rate": 4.395598705501619e-06,
      "loss": 0.0231,
      "step": 301360
    },
    {
      "epoch": 117.04,
      "learning_rate": 4.395080906148867e-06,
      "loss": 0.0231,
      "step": 301370
    },
    {
      "epoch": 117.04,
      "learning_rate": 4.394563106796117e-06,
      "loss": 0.0005,
      "step": 301380
    },
    {
      "epoch": 117.04,
      "learning_rate": 4.394045307443366e-06,
      "loss": 0.0208,
      "step": 301390
    },
    {
      "epoch": 117.05,
      "learning_rate": 4.393527508090615e-06,
      "loss": 0.0397,
      "step": 301400
    },
    {
      "epoch": 117.05,
      "learning_rate": 4.393009708737864e-06,
      "loss": 0.0398,
      "step": 301410
    },
    {
      "epoch": 117.06,
      "learning_rate": 4.392491909385114e-06,
      "loss": 0.1056,
      "step": 301420
    },
    {
      "epoch": 117.06,
      "learning_rate": 4.3919741100323625e-06,
      "loss": 0.0367,
      "step": 301430
    },
    {
      "epoch": 117.06,
      "learning_rate": 4.391456310679612e-06,
      "loss": 0.0201,
      "step": 301440
    },
    {
      "epoch": 117.07,
      "learning_rate": 4.390938511326861e-06,
      "loss": 0.0099,
      "step": 301450
    },
    {
      "epoch": 117.07,
      "learning_rate": 4.3904207119741105e-06,
      "loss": 0.021,
      "step": 301460
    },
    {
      "epoch": 117.08,
      "learning_rate": 4.389902912621359e-06,
      "loss": 0.0411,
      "step": 301470
    },
    {
      "epoch": 117.08,
      "learning_rate": 4.389385113268609e-06,
      "loss": 0.1234,
      "step": 301480
    },
    {
      "epoch": 117.08,
      "learning_rate": 4.388867313915858e-06,
      "loss": 0.0517,
      "step": 301490
    },
    {
      "epoch": 117.09,
      "learning_rate": 4.388349514563107e-06,
      "loss": 0.0117,
      "step": 301500
    },
    {
      "epoch": 117.09,
      "learning_rate": 4.387831715210356e-06,
      "loss": 0.0967,
      "step": 301510
    },
    {
      "epoch": 117.1,
      "learning_rate": 4.387313915857606e-06,
      "loss": 0.0081,
      "step": 301520
    },
    {
      "epoch": 117.1,
      "learning_rate": 4.3867961165048545e-06,
      "loss": 0.0413,
      "step": 301530
    },
    {
      "epoch": 117.1,
      "learning_rate": 4.386278317152104e-06,
      "loss": 0.0034,
      "step": 301540
    },
    {
      "epoch": 117.11,
      "learning_rate": 4.385760517799353e-06,
      "loss": 0.0025,
      "step": 301550
    },
    {
      "epoch": 117.11,
      "learning_rate": 4.3852427184466025e-06,
      "loss": 0.0951,
      "step": 301560
    },
    {
      "epoch": 117.11,
      "learning_rate": 4.384724919093851e-06,
      "loss": 0.0074,
      "step": 301570
    },
    {
      "epoch": 117.12,
      "learning_rate": 4.384207119741101e-06,
      "loss": 0.0869,
      "step": 301580
    },
    {
      "epoch": 117.12,
      "learning_rate": 4.38368932038835e-06,
      "loss": 0.0115,
      "step": 301590
    },
    {
      "epoch": 117.13,
      "learning_rate": 4.383171521035599e-06,
      "loss": 0.1069,
      "step": 301600
    },
    {
      "epoch": 117.13,
      "learning_rate": 4.382653721682849e-06,
      "loss": 0.0251,
      "step": 301610
    },
    {
      "epoch": 117.13,
      "learning_rate": 4.382135922330098e-06,
      "loss": 0.1173,
      "step": 301620
    },
    {
      "epoch": 117.14,
      "learning_rate": 4.381618122977347e-06,
      "loss": 0.0443,
      "step": 301630
    },
    {
      "epoch": 117.14,
      "learning_rate": 4.381100323624596e-06,
      "loss": 0.0547,
      "step": 301640
    },
    {
      "epoch": 117.15,
      "learning_rate": 4.380582524271846e-06,
      "loss": 0.0771,
      "step": 301650
    },
    {
      "epoch": 117.15,
      "learning_rate": 4.3800647249190944e-06,
      "loss": 0.0402,
      "step": 301660
    },
    {
      "epoch": 117.15,
      "learning_rate": 4.379546925566344e-06,
      "loss": 0.0627,
      "step": 301670
    },
    {
      "epoch": 117.16,
      "learning_rate": 4.379029126213593e-06,
      "loss": 0.0697,
      "step": 301680
    },
    {
      "epoch": 117.16,
      "learning_rate": 4.3785113268608424e-06,
      "loss": 0.0063,
      "step": 301690
    },
    {
      "epoch": 117.17,
      "learning_rate": 4.377993527508091e-06,
      "loss": 0.0157,
      "step": 301700
    },
    {
      "epoch": 117.17,
      "learning_rate": 4.37747572815534e-06,
      "loss": 0.0037,
      "step": 301710
    },
    {
      "epoch": 117.17,
      "learning_rate": 4.37695792880259e-06,
      "loss": 0.0842,
      "step": 301720
    },
    {
      "epoch": 117.18,
      "learning_rate": 4.376440129449838e-06,
      "loss": 0.0966,
      "step": 301730
    },
    {
      "epoch": 117.18,
      "learning_rate": 4.375922330097088e-06,
      "loss": 0.0132,
      "step": 301740
    },
    {
      "epoch": 117.18,
      "learning_rate": 4.375404530744337e-06,
      "loss": 0.0015,
      "step": 301750
    },
    {
      "epoch": 117.19,
      "learning_rate": 4.374886731391586e-06,
      "loss": 0.1428,
      "step": 301760
    },
    {
      "epoch": 117.19,
      "learning_rate": 4.374368932038835e-06,
      "loss": 0.0841,
      "step": 301770
    },
    {
      "epoch": 117.2,
      "learning_rate": 4.373851132686085e-06,
      "loss": 0.025,
      "step": 301780
    },
    {
      "epoch": 117.2,
      "learning_rate": 4.3733333333333335e-06,
      "loss": 0.0475,
      "step": 301790
    },
    {
      "epoch": 117.2,
      "learning_rate": 4.372815533980583e-06,
      "loss": 0.0235,
      "step": 301800
    },
    {
      "epoch": 117.21,
      "learning_rate": 4.372297734627832e-06,
      "loss": 0.0279,
      "step": 301810
    },
    {
      "epoch": 117.21,
      "learning_rate": 4.3717799352750815e-06,
      "loss": 0.0283,
      "step": 301820
    },
    {
      "epoch": 117.22,
      "learning_rate": 4.37126213592233e-06,
      "loss": 0.0282,
      "step": 301830
    },
    {
      "epoch": 117.22,
      "learning_rate": 4.37074433656958e-06,
      "loss": 0.248,
      "step": 301840
    },
    {
      "epoch": 117.22,
      "learning_rate": 4.370226537216829e-06,
      "loss": 0.0486,
      "step": 301850
    },
    {
      "epoch": 117.23,
      "learning_rate": 4.369708737864078e-06,
      "loss": 0.067,
      "step": 301860
    },
    {
      "epoch": 117.23,
      "learning_rate": 4.369190938511327e-06,
      "loss": 0.0785,
      "step": 301870
    },
    {
      "epoch": 117.23,
      "learning_rate": 4.368673139158576e-06,
      "loss": 0.0272,
      "step": 301880
    },
    {
      "epoch": 117.24,
      "learning_rate": 4.3681553398058255e-06,
      "loss": 0.0376,
      "step": 301890
    },
    {
      "epoch": 117.24,
      "learning_rate": 4.367637540453074e-06,
      "loss": 0.0713,
      "step": 301900
    },
    {
      "epoch": 117.25,
      "learning_rate": 4.367119741100324e-06,
      "loss": 0.0409,
      "step": 301910
    },
    {
      "epoch": 117.25,
      "learning_rate": 4.366601941747573e-06,
      "loss": 0.1008,
      "step": 301920
    },
    {
      "epoch": 117.25,
      "learning_rate": 4.366084142394822e-06,
      "loss": 0.0479,
      "step": 301930
    },
    {
      "epoch": 117.26,
      "learning_rate": 4.365566343042071e-06,
      "loss": 0.032,
      "step": 301940
    },
    {
      "epoch": 117.26,
      "learning_rate": 4.365048543689321e-06,
      "loss": 0.0823,
      "step": 301950
    },
    {
      "epoch": 117.27,
      "learning_rate": 4.364530744336569e-06,
      "loss": 0.0011,
      "step": 301960
    },
    {
      "epoch": 117.27,
      "learning_rate": 4.364012944983819e-06,
      "loss": 0.0039,
      "step": 301970
    },
    {
      "epoch": 117.27,
      "learning_rate": 4.363495145631068e-06,
      "loss": 0.0517,
      "step": 301980
    },
    {
      "epoch": 117.28,
      "learning_rate": 4.362977346278317e-06,
      "loss": 0.049,
      "step": 301990
    },
    {
      "epoch": 117.28,
      "learning_rate": 4.362459546925566e-06,
      "loss": 0.0004,
      "step": 302000
    },
    {
      "epoch": 117.29,
      "learning_rate": 4.361941747572816e-06,
      "loss": 0.0924,
      "step": 302010
    },
    {
      "epoch": 117.29,
      "learning_rate": 4.361423948220065e-06,
      "loss": 0.0334,
      "step": 302020
    },
    {
      "epoch": 117.29,
      "learning_rate": 4.360906148867314e-06,
      "loss": 0.1643,
      "step": 302030
    },
    {
      "epoch": 117.3,
      "learning_rate": 4.360388349514563e-06,
      "loss": 0.0246,
      "step": 302040
    },
    {
      "epoch": 117.3,
      "learning_rate": 4.359870550161813e-06,
      "loss": 0.0856,
      "step": 302050
    },
    {
      "epoch": 117.3,
      "learning_rate": 4.359352750809061e-06,
      "loss": 0.0771,
      "step": 302060
    },
    {
      "epoch": 117.31,
      "learning_rate": 4.358834951456311e-06,
      "loss": 0.0388,
      "step": 302070
    },
    {
      "epoch": 117.31,
      "learning_rate": 4.35831715210356e-06,
      "loss": 0.0006,
      "step": 302080
    },
    {
      "epoch": 117.32,
      "learning_rate": 4.357799352750809e-06,
      "loss": 0.0356,
      "step": 302090
    },
    {
      "epoch": 117.32,
      "learning_rate": 4.357281553398058e-06,
      "loss": 0.0816,
      "step": 302100
    },
    {
      "epoch": 117.32,
      "learning_rate": 4.356763754045308e-06,
      "loss": 0.0846,
      "step": 302110
    },
    {
      "epoch": 117.33,
      "learning_rate": 4.3562459546925565e-06,
      "loss": 0.0895,
      "step": 302120
    },
    {
      "epoch": 117.33,
      "learning_rate": 4.355728155339806e-06,
      "loss": 0.0012,
      "step": 302130
    },
    {
      "epoch": 117.34,
      "learning_rate": 4.355210355987055e-06,
      "loss": 0.02,
      "step": 302140
    },
    {
      "epoch": 117.34,
      "learning_rate": 4.3546925566343045e-06,
      "loss": 0.0529,
      "step": 302150
    },
    {
      "epoch": 117.34,
      "learning_rate": 4.354174757281553e-06,
      "loss": 0.0109,
      "step": 302160
    },
    {
      "epoch": 117.35,
      "learning_rate": 4.353656957928803e-06,
      "loss": 0.0008,
      "step": 302170
    },
    {
      "epoch": 117.35,
      "learning_rate": 4.3531391585760525e-06,
      "loss": 0.0337,
      "step": 302180
    },
    {
      "epoch": 117.36,
      "learning_rate": 4.352621359223301e-06,
      "loss": 0.0631,
      "step": 302190
    },
    {
      "epoch": 117.36,
      "learning_rate": 4.352103559870551e-06,
      "loss": 0.0298,
      "step": 302200
    },
    {
      "epoch": 117.36,
      "learning_rate": 4.3515857605178e-06,
      "loss": 0.0162,
      "step": 302210
    },
    {
      "epoch": 117.37,
      "learning_rate": 4.351067961165049e-06,
      "loss": 0.0149,
      "step": 302220
    },
    {
      "epoch": 117.37,
      "learning_rate": 4.350550161812298e-06,
      "loss": 0.2614,
      "step": 302230
    },
    {
      "epoch": 117.37,
      "learning_rate": 4.350032362459548e-06,
      "loss": 0.0691,
      "step": 302240
    },
    {
      "epoch": 117.38,
      "learning_rate": 4.3495145631067965e-06,
      "loss": 0.0073,
      "step": 302250
    },
    {
      "epoch": 117.38,
      "learning_rate": 4.348996763754046e-06,
      "loss": 0.0321,
      "step": 302260
    },
    {
      "epoch": 117.39,
      "learning_rate": 4.348478964401295e-06,
      "loss": 0.0465,
      "step": 302270
    },
    {
      "epoch": 117.39,
      "learning_rate": 4.3479611650485445e-06,
      "loss": 0.0006,
      "step": 302280
    },
    {
      "epoch": 117.39,
      "learning_rate": 4.347443365695793e-06,
      "loss": 0.0081,
      "step": 302290
    },
    {
      "epoch": 117.4,
      "learning_rate": 4.346925566343043e-06,
      "loss": 0.063,
      "step": 302300
    },
    {
      "epoch": 117.4,
      "learning_rate": 4.346407766990292e-06,
      "loss": 0.1007,
      "step": 302310
    },
    {
      "epoch": 117.41,
      "learning_rate": 4.345889967637541e-06,
      "loss": 0.001,
      "step": 302320
    },
    {
      "epoch": 117.41,
      "learning_rate": 4.34537216828479e-06,
      "loss": 0.1071,
      "step": 302330
    },
    {
      "epoch": 117.41,
      "learning_rate": 4.34485436893204e-06,
      "loss": 0.0412,
      "step": 302340
    },
    {
      "epoch": 117.42,
      "learning_rate": 4.3443365695792884e-06,
      "loss": 0.0971,
      "step": 302350
    },
    {
      "epoch": 117.42,
      "learning_rate": 4.343818770226538e-06,
      "loss": 0.1249,
      "step": 302360
    },
    {
      "epoch": 117.43,
      "learning_rate": 4.343300970873787e-06,
      "loss": 0.0729,
      "step": 302370
    },
    {
      "epoch": 117.43,
      "learning_rate": 4.3427831715210364e-06,
      "loss": 0.0291,
      "step": 302380
    },
    {
      "epoch": 117.43,
      "learning_rate": 4.342265372168285e-06,
      "loss": 0.0332,
      "step": 302390
    },
    {
      "epoch": 117.44,
      "learning_rate": 4.341747572815535e-06,
      "loss": 0.1386,
      "step": 302400
    },
    {
      "epoch": 117.44,
      "learning_rate": 4.341229773462784e-06,
      "loss": 0.0701,
      "step": 302410
    },
    {
      "epoch": 117.44,
      "learning_rate": 4.340711974110033e-06,
      "loss": 0.0271,
      "step": 302420
    },
    {
      "epoch": 117.45,
      "learning_rate": 4.340194174757282e-06,
      "loss": 0.0456,
      "step": 302430
    },
    {
      "epoch": 117.45,
      "learning_rate": 4.339676375404532e-06,
      "loss": 0.0307,
      "step": 302440
    },
    {
      "epoch": 117.46,
      "learning_rate": 4.33915857605178e-06,
      "loss": 0.0377,
      "step": 302450
    },
    {
      "epoch": 117.46,
      "learning_rate": 4.33864077669903e-06,
      "loss": 0.0522,
      "step": 302460
    },
    {
      "epoch": 117.46,
      "learning_rate": 4.338122977346279e-06,
      "loss": 0.065,
      "step": 302470
    },
    {
      "epoch": 117.47,
      "learning_rate": 4.337605177993528e-06,
      "loss": 0.0014,
      "step": 302480
    },
    {
      "epoch": 117.47,
      "learning_rate": 4.337087378640777e-06,
      "loss": 0.071,
      "step": 302490
    },
    {
      "epoch": 117.48,
      "learning_rate": 4.336569579288027e-06,
      "loss": 0.0925,
      "step": 302500
    },
    {
      "epoch": 117.48,
      "learning_rate": 4.3360517799352755e-06,
      "loss": 0.054,
      "step": 302510
    },
    {
      "epoch": 117.48,
      "learning_rate": 4.335533980582524e-06,
      "loss": 0.0582,
      "step": 302520
    },
    {
      "epoch": 117.49,
      "learning_rate": 4.335016181229774e-06,
      "loss": 0.0114,
      "step": 302530
    },
    {
      "epoch": 117.49,
      "learning_rate": 4.334498381877023e-06,
      "loss": 0.0934,
      "step": 302540
    },
    {
      "epoch": 117.5,
      "learning_rate": 4.333980582524272e-06,
      "loss": 0.0483,
      "step": 302550
    },
    {
      "epoch": 117.5,
      "learning_rate": 4.333462783171521e-06,
      "loss": 0.0022,
      "step": 302560
    },
    {
      "epoch": 117.5,
      "learning_rate": 4.332944983818771e-06,
      "loss": 0.0728,
      "step": 302570
    },
    {
      "epoch": 117.51,
      "learning_rate": 4.3324271844660195e-06,
      "loss": 0.0302,
      "step": 302580
    },
    {
      "epoch": 117.51,
      "learning_rate": 4.331909385113269e-06,
      "loss": 0.0155,
      "step": 302590
    },
    {
      "epoch": 117.51,
      "learning_rate": 4.331391585760518e-06,
      "loss": 0.0429,
      "step": 302600
    },
    {
      "epoch": 117.52,
      "learning_rate": 4.3308737864077675e-06,
      "loss": 0.0001,
      "step": 302610
    },
    {
      "epoch": 117.52,
      "learning_rate": 4.330355987055016e-06,
      "loss": 0.0121,
      "step": 302620
    },
    {
      "epoch": 117.53,
      "learning_rate": 4.329838187702266e-06,
      "loss": 0.0107,
      "step": 302630
    },
    {
      "epoch": 117.53,
      "learning_rate": 4.329320388349515e-06,
      "loss": 0.2105,
      "step": 302640
    },
    {
      "epoch": 117.53,
      "learning_rate": 4.328802588996764e-06,
      "loss": 0.0306,
      "step": 302650
    },
    {
      "epoch": 117.54,
      "learning_rate": 4.328284789644013e-06,
      "loss": 0.0732,
      "step": 302660
    },
    {
      "epoch": 117.54,
      "learning_rate": 4.327766990291263e-06,
      "loss": 0.0966,
      "step": 302670
    },
    {
      "epoch": 117.55,
      "learning_rate": 4.327249190938511e-06,
      "loss": 0.0426,
      "step": 302680
    },
    {
      "epoch": 117.55,
      "learning_rate": 4.32673139158576e-06,
      "loss": 0.0398,
      "step": 302690
    },
    {
      "epoch": 117.55,
      "learning_rate": 4.32621359223301e-06,
      "loss": 0.03,
      "step": 302700
    },
    {
      "epoch": 117.56,
      "learning_rate": 4.325695792880259e-06,
      "loss": 0.0699,
      "step": 302710
    },
    {
      "epoch": 117.56,
      "learning_rate": 4.325177993527508e-06,
      "loss": 0.0007,
      "step": 302720
    },
    {
      "epoch": 117.57,
      "learning_rate": 4.324660194174758e-06,
      "loss": 0.0135,
      "step": 302730
    },
    {
      "epoch": 117.57,
      "learning_rate": 4.324142394822007e-06,
      "loss": 0.0017,
      "step": 302740
    },
    {
      "epoch": 117.57,
      "learning_rate": 4.323624595469256e-06,
      "loss": 0.0003,
      "step": 302750
    },
    {
      "epoch": 117.58,
      "learning_rate": 4.323106796116505e-06,
      "loss": 0.0001,
      "step": 302760
    },
    {
      "epoch": 117.58,
      "learning_rate": 4.322588996763755e-06,
      "loss": 0.0085,
      "step": 302770
    },
    {
      "epoch": 117.58,
      "learning_rate": 4.322071197411003e-06,
      "loss": 0.0204,
      "step": 302780
    },
    {
      "epoch": 117.59,
      "learning_rate": 4.321553398058253e-06,
      "loss": 0.0956,
      "step": 302790
    },
    {
      "epoch": 117.59,
      "learning_rate": 4.321035598705502e-06,
      "loss": 0.0897,
      "step": 302800
    },
    {
      "epoch": 117.6,
      "learning_rate": 4.320517799352751e-06,
      "loss": 0.0286,
      "step": 302810
    },
    {
      "epoch": 117.6,
      "learning_rate": 4.32e-06,
      "loss": 0.0962,
      "step": 302820
    },
    {
      "epoch": 117.6,
      "learning_rate": 4.31948220064725e-06,
      "loss": 0.0867,
      "step": 302830
    },
    {
      "epoch": 117.61,
      "learning_rate": 4.3189644012944985e-06,
      "loss": 0.0093,
      "step": 302840
    },
    {
      "epoch": 117.61,
      "learning_rate": 4.318446601941748e-06,
      "loss": 0.028,
      "step": 302850
    },
    {
      "epoch": 117.62,
      "learning_rate": 4.317928802588997e-06,
      "loss": 0.037,
      "step": 302860
    },
    {
      "epoch": 117.62,
      "learning_rate": 4.3174110032362465e-06,
      "loss": 0.0125,
      "step": 302870
    },
    {
      "epoch": 117.62,
      "learning_rate": 4.316893203883495e-06,
      "loss": 0.01,
      "step": 302880
    },
    {
      "epoch": 117.63,
      "learning_rate": 4.316375404530745e-06,
      "loss": 0.0155,
      "step": 302890
    },
    {
      "epoch": 117.63,
      "learning_rate": 4.315857605177994e-06,
      "loss": 0.0109,
      "step": 302900
    },
    {
      "epoch": 117.63,
      "learning_rate": 4.315339805825243e-06,
      "loss": 0.0011,
      "step": 302910
    },
    {
      "epoch": 117.64,
      "learning_rate": 4.314822006472492e-06,
      "loss": 0.0204,
      "step": 302920
    },
    {
      "epoch": 117.64,
      "learning_rate": 4.314304207119742e-06,
      "loss": 0.0911,
      "step": 302930
    },
    {
      "epoch": 117.65,
      "learning_rate": 4.3137864077669905e-06,
      "loss": 0.0657,
      "step": 302940
    },
    {
      "epoch": 117.65,
      "learning_rate": 4.31326860841424e-06,
      "loss": 0.0379,
      "step": 302950
    },
    {
      "epoch": 117.65,
      "learning_rate": 4.312750809061489e-06,
      "loss": 0.0004,
      "step": 302960
    },
    {
      "epoch": 117.66,
      "learning_rate": 4.3122330097087385e-06,
      "loss": 0.0248,
      "step": 302970
    },
    {
      "epoch": 117.66,
      "learning_rate": 4.311715210355987e-06,
      "loss": 0.0341,
      "step": 302980
    },
    {
      "epoch": 117.67,
      "learning_rate": 4.311197411003237e-06,
      "loss": 0.1404,
      "step": 302990
    },
    {
      "epoch": 117.67,
      "learning_rate": 4.310679611650486e-06,
      "loss": 0.0374,
      "step": 303000
    },
    {
      "epoch": 117.67,
      "learning_rate": 4.310161812297735e-06,
      "loss": 0.0081,
      "step": 303010
    },
    {
      "epoch": 117.68,
      "learning_rate": 4.309644012944984e-06,
      "loss": 0.0086,
      "step": 303020
    },
    {
      "epoch": 117.68,
      "learning_rate": 4.309126213592234e-06,
      "loss": 0.0978,
      "step": 303030
    },
    {
      "epoch": 117.69,
      "learning_rate": 4.308608414239482e-06,
      "loss": 0.0009,
      "step": 303040
    },
    {
      "epoch": 117.69,
      "learning_rate": 4.308090614886732e-06,
      "loss": 0.0737,
      "step": 303050
    },
    {
      "epoch": 117.69,
      "learning_rate": 4.307572815533981e-06,
      "loss": 0.0632,
      "step": 303060
    },
    {
      "epoch": 117.7,
      "learning_rate": 4.3070550161812304e-06,
      "loss": 0.1006,
      "step": 303070
    },
    {
      "epoch": 117.7,
      "learning_rate": 4.306537216828479e-06,
      "loss": 0.0001,
      "step": 303080
    },
    {
      "epoch": 117.7,
      "learning_rate": 4.306019417475729e-06,
      "loss": 0.0162,
      "step": 303090
    },
    {
      "epoch": 117.71,
      "learning_rate": 4.305501618122978e-06,
      "loss": 0.0756,
      "step": 303100
    },
    {
      "epoch": 117.71,
      "learning_rate": 4.304983818770227e-06,
      "loss": 0.0004,
      "step": 303110
    },
    {
      "epoch": 117.72,
      "learning_rate": 4.304466019417476e-06,
      "loss": 0.0002,
      "step": 303120
    },
    {
      "epoch": 117.72,
      "learning_rate": 4.303948220064726e-06,
      "loss": 0.1409,
      "step": 303130
    },
    {
      "epoch": 117.72,
      "learning_rate": 4.303430420711974e-06,
      "loss": 0.0015,
      "step": 303140
    },
    {
      "epoch": 117.73,
      "learning_rate": 4.302912621359224e-06,
      "loss": 0.0766,
      "step": 303150
    },
    {
      "epoch": 117.73,
      "learning_rate": 4.302394822006473e-06,
      "loss": 0.0085,
      "step": 303160
    },
    {
      "epoch": 117.74,
      "learning_rate": 4.301877022653722e-06,
      "loss": 0.051,
      "step": 303170
    },
    {
      "epoch": 117.74,
      "learning_rate": 4.301359223300971e-06,
      "loss": 0.0129,
      "step": 303180
    },
    {
      "epoch": 117.74,
      "learning_rate": 4.300841423948221e-06,
      "loss": 0.0554,
      "step": 303190
    },
    {
      "epoch": 117.75,
      "learning_rate": 4.3003236245954695e-06,
      "loss": 0.0662,
      "step": 303200
    },
    {
      "epoch": 117.75,
      "learning_rate": 4.299805825242719e-06,
      "loss": 0.0523,
      "step": 303210
    },
    {
      "epoch": 117.76,
      "learning_rate": 4.299288025889968e-06,
      "loss": 0.0213,
      "step": 303220
    },
    {
      "epoch": 117.76,
      "learning_rate": 4.2987702265372175e-06,
      "loss": 0.0591,
      "step": 303230
    },
    {
      "epoch": 117.76,
      "learning_rate": 4.298252427184466e-06,
      "loss": 0.0151,
      "step": 303240
    },
    {
      "epoch": 117.77,
      "learning_rate": 4.297734627831716e-06,
      "loss": 0.0605,
      "step": 303250
    },
    {
      "epoch": 117.77,
      "learning_rate": 4.297216828478965e-06,
      "loss": 0.0586,
      "step": 303260
    },
    {
      "epoch": 117.77,
      "learning_rate": 4.296699029126214e-06,
      "loss": 0.1048,
      "step": 303270
    },
    {
      "epoch": 117.78,
      "learning_rate": 4.296181229773463e-06,
      "loss": 0.0563,
      "step": 303280
    },
    {
      "epoch": 117.78,
      "learning_rate": 4.295663430420713e-06,
      "loss": 0.0564,
      "step": 303290
    },
    {
      "epoch": 117.79,
      "learning_rate": 4.2951456310679615e-06,
      "loss": 0.0357,
      "step": 303300
    },
    {
      "epoch": 117.79,
      "learning_rate": 4.294627831715211e-06,
      "loss": 0.0383,
      "step": 303310
    },
    {
      "epoch": 117.79,
      "learning_rate": 4.29411003236246e-06,
      "loss": 0.0126,
      "step": 303320
    },
    {
      "epoch": 117.8,
      "learning_rate": 4.2935922330097095e-06,
      "loss": 0.0793,
      "step": 303330
    },
    {
      "epoch": 117.8,
      "learning_rate": 4.293074433656958e-06,
      "loss": 0.0254,
      "step": 303340
    },
    {
      "epoch": 117.81,
      "learning_rate": 4.292556634304207e-06,
      "loss": 0.0342,
      "step": 303350
    },
    {
      "epoch": 117.81,
      "learning_rate": 4.292038834951457e-06,
      "loss": 0.0966,
      "step": 303360
    },
    {
      "epoch": 117.81,
      "learning_rate": 4.291521035598705e-06,
      "loss": 0.0195,
      "step": 303370
    },
    {
      "epoch": 117.82,
      "learning_rate": 4.291003236245955e-06,
      "loss": 0.0106,
      "step": 303380
    },
    {
      "epoch": 117.82,
      "learning_rate": 4.290485436893204e-06,
      "loss": 0.044,
      "step": 303390
    },
    {
      "epoch": 117.83,
      "learning_rate": 4.2899676375404534e-06,
      "loss": 0.0039,
      "step": 303400
    },
    {
      "epoch": 117.83,
      "learning_rate": 4.289449838187702e-06,
      "loss": 0.0178,
      "step": 303410
    },
    {
      "epoch": 117.83,
      "learning_rate": 4.288932038834952e-06,
      "loss": 0.0066,
      "step": 303420
    },
    {
      "epoch": 117.84,
      "learning_rate": 4.288414239482201e-06,
      "loss": 0.0001,
      "step": 303430
    },
    {
      "epoch": 117.84,
      "learning_rate": 4.28789644012945e-06,
      "loss": 0.0142,
      "step": 303440
    },
    {
      "epoch": 117.84,
      "learning_rate": 4.287378640776699e-06,
      "loss": 0.0006,
      "step": 303450
    },
    {
      "epoch": 117.85,
      "learning_rate": 4.286860841423949e-06,
      "loss": 0.0003,
      "step": 303460
    },
    {
      "epoch": 117.85,
      "learning_rate": 4.286343042071197e-06,
      "loss": 0.0577,
      "step": 303470
    },
    {
      "epoch": 117.86,
      "learning_rate": 4.285825242718447e-06,
      "loss": 0.0182,
      "step": 303480
    },
    {
      "epoch": 117.86,
      "learning_rate": 4.285307443365696e-06,
      "loss": 0.0944,
      "step": 303490
    },
    {
      "epoch": 117.86,
      "learning_rate": 4.284789644012945e-06,
      "loss": 0.0002,
      "step": 303500
    },
    {
      "epoch": 117.87,
      "learning_rate": 4.284271844660194e-06,
      "loss": 0.0004,
      "step": 303510
    },
    {
      "epoch": 117.87,
      "learning_rate": 4.283754045307444e-06,
      "loss": 0.0996,
      "step": 303520
    },
    {
      "epoch": 117.88,
      "learning_rate": 4.2832362459546925e-06,
      "loss": 0.0091,
      "step": 303530
    },
    {
      "epoch": 117.88,
      "learning_rate": 4.282718446601942e-06,
      "loss": 0.0112,
      "step": 303540
    },
    {
      "epoch": 117.88,
      "learning_rate": 4.282200647249191e-06,
      "loss": 0.0318,
      "step": 303550
    },
    {
      "epoch": 117.89,
      "learning_rate": 4.2816828478964405e-06,
      "loss": 0.0203,
      "step": 303560
    },
    {
      "epoch": 117.89,
      "learning_rate": 4.281165048543689e-06,
      "loss": 0.032,
      "step": 303570
    },
    {
      "epoch": 117.9,
      "learning_rate": 4.280647249190939e-06,
      "loss": 0.0645,
      "step": 303580
    },
    {
      "epoch": 117.9,
      "learning_rate": 4.280129449838188e-06,
      "loss": 0.0208,
      "step": 303590
    },
    {
      "epoch": 117.9,
      "learning_rate": 4.279611650485437e-06,
      "loss": 0.0009,
      "step": 303600
    },
    {
      "epoch": 117.91,
      "learning_rate": 4.279093851132686e-06,
      "loss": 0.0002,
      "step": 303610
    },
    {
      "epoch": 117.91,
      "learning_rate": 4.278576051779936e-06,
      "loss": 0.1258,
      "step": 303620
    },
    {
      "epoch": 117.91,
      "learning_rate": 4.2780582524271845e-06,
      "loss": 0.0001,
      "step": 303630
    },
    {
      "epoch": 117.92,
      "learning_rate": 4.277540453074434e-06,
      "loss": 0.0218,
      "step": 303640
    },
    {
      "epoch": 117.92,
      "learning_rate": 4.277022653721683e-06,
      "loss": 0.1255,
      "step": 303650
    },
    {
      "epoch": 117.93,
      "learning_rate": 4.2765048543689325e-06,
      "loss": 0.0002,
      "step": 303660
    },
    {
      "epoch": 117.93,
      "learning_rate": 4.275987055016181e-06,
      "loss": 0.0048,
      "step": 303670
    },
    {
      "epoch": 117.93,
      "learning_rate": 4.275469255663431e-06,
      "loss": 0.0098,
      "step": 303680
    },
    {
      "epoch": 117.94,
      "learning_rate": 4.27495145631068e-06,
      "loss": 0.043,
      "step": 303690
    },
    {
      "epoch": 117.94,
      "learning_rate": 4.274433656957929e-06,
      "loss": 0.117,
      "step": 303700
    },
    {
      "epoch": 117.95,
      "learning_rate": 4.273915857605178e-06,
      "loss": 0.1114,
      "step": 303710
    },
    {
      "epoch": 117.95,
      "learning_rate": 4.273398058252428e-06,
      "loss": 0.0208,
      "step": 303720
    },
    {
      "epoch": 117.95,
      "learning_rate": 4.272880258899676e-06,
      "loss": 0.0094,
      "step": 303730
    },
    {
      "epoch": 117.96,
      "learning_rate": 4.272362459546926e-06,
      "loss": 0.0897,
      "step": 303740
    },
    {
      "epoch": 117.96,
      "learning_rate": 4.271844660194175e-06,
      "loss": 0.0774,
      "step": 303750
    },
    {
      "epoch": 117.97,
      "learning_rate": 4.2713268608414244e-06,
      "loss": 0.0479,
      "step": 303760
    },
    {
      "epoch": 117.97,
      "learning_rate": 4.270809061488673e-06,
      "loss": 0.0201,
      "step": 303770
    },
    {
      "epoch": 117.97,
      "learning_rate": 4.270291262135923e-06,
      "loss": 0.0931,
      "step": 303780
    },
    {
      "epoch": 117.98,
      "learning_rate": 4.269773462783172e-06,
      "loss": 0.0348,
      "step": 303790
    },
    {
      "epoch": 117.98,
      "learning_rate": 4.269255663430421e-06,
      "loss": 0.059,
      "step": 303800
    },
    {
      "epoch": 117.98,
      "learning_rate": 4.26873786407767e-06,
      "loss": 0.0446,
      "step": 303810
    },
    {
      "epoch": 117.99,
      "learning_rate": 4.26822006472492e-06,
      "loss": 0.0946,
      "step": 303820
    },
    {
      "epoch": 117.99,
      "learning_rate": 4.267702265372168e-06,
      "loss": 0.0402,
      "step": 303830
    },
    {
      "epoch": 118.0,
      "learning_rate": 4.267184466019418e-06,
      "loss": 0.0587,
      "step": 303840
    },
    {
      "epoch": 118.0,
      "learning_rate": 4.266666666666668e-06,
      "loss": 0.0045,
      "step": 303850
    },
    {
      "epoch": 118.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.3821144700050354,
      "eval_runtime": 8.1852,
      "eval_samples_per_second": 444.092,
      "eval_steps_per_second": 55.588,
      "step": 303850
    },
    {
      "epoch": 118.0,
      "learning_rate": 4.266148867313916e-06,
      "loss": 0.0002,
      "step": 303860
    },
    {
      "epoch": 118.01,
      "learning_rate": 4.265631067961166e-06,
      "loss": 0.1503,
      "step": 303870
    },
    {
      "epoch": 118.01,
      "learning_rate": 4.265113268608415e-06,
      "loss": 0.0005,
      "step": 303880
    },
    {
      "epoch": 118.02,
      "learning_rate": 4.264595469255664e-06,
      "loss": 0.0208,
      "step": 303890
    },
    {
      "epoch": 118.02,
      "learning_rate": 4.264077669902913e-06,
      "loss": 0.1188,
      "step": 303900
    },
    {
      "epoch": 118.02,
      "learning_rate": 4.263559870550163e-06,
      "loss": 0.0119,
      "step": 303910
    },
    {
      "epoch": 118.03,
      "learning_rate": 4.2630420711974115e-06,
      "loss": 0.0114,
      "step": 303920
    },
    {
      "epoch": 118.03,
      "learning_rate": 4.262524271844661e-06,
      "loss": 0.0096,
      "step": 303930
    },
    {
      "epoch": 118.03,
      "learning_rate": 4.26200647249191e-06,
      "loss": 0.0455,
      "step": 303940
    },
    {
      "epoch": 118.04,
      "learning_rate": 4.2614886731391595e-06,
      "loss": 0.0003,
      "step": 303950
    },
    {
      "epoch": 118.04,
      "learning_rate": 4.260970873786408e-06,
      "loss": 0.0546,
      "step": 303960
    },
    {
      "epoch": 118.05,
      "learning_rate": 4.260453074433658e-06,
      "loss": 0.1516,
      "step": 303970
    },
    {
      "epoch": 118.05,
      "learning_rate": 4.259935275080907e-06,
      "loss": 0.1272,
      "step": 303980
    },
    {
      "epoch": 118.05,
      "learning_rate": 4.2594174757281555e-06,
      "loss": 0.0329,
      "step": 303990
    },
    {
      "epoch": 118.06,
      "learning_rate": 4.258899676375405e-06,
      "loss": 0.042,
      "step": 304000
    },
    {
      "epoch": 118.06,
      "learning_rate": 4.258381877022654e-06,
      "loss": 0.0538,
      "step": 304010
    },
    {
      "epoch": 118.07,
      "learning_rate": 4.2578640776699035e-06,
      "loss": 0.0007,
      "step": 304020
    },
    {
      "epoch": 118.07,
      "learning_rate": 4.257346278317152e-06,
      "loss": 0.0212,
      "step": 304030
    },
    {
      "epoch": 118.07,
      "learning_rate": 4.256828478964402e-06,
      "loss": 0.0011,
      "step": 304040
    },
    {
      "epoch": 118.08,
      "learning_rate": 4.256310679611651e-06,
      "loss": 0.0517,
      "step": 304050
    },
    {
      "epoch": 118.08,
      "learning_rate": 4.2557928802589e-06,
      "loss": 0.0417,
      "step": 304060
    },
    {
      "epoch": 118.09,
      "learning_rate": 4.255275080906149e-06,
      "loss": 0.0395,
      "step": 304070
    },
    {
      "epoch": 118.09,
      "learning_rate": 4.254757281553399e-06,
      "loss": 0.004,
      "step": 304080
    },
    {
      "epoch": 118.09,
      "learning_rate": 4.2542394822006474e-06,
      "loss": 0.0103,
      "step": 304090
    },
    {
      "epoch": 118.1,
      "learning_rate": 4.253721682847897e-06,
      "loss": 0.0593,
      "step": 304100
    },
    {
      "epoch": 118.1,
      "learning_rate": 4.253203883495146e-06,
      "loss": 0.0007,
      "step": 304110
    },
    {
      "epoch": 118.1,
      "learning_rate": 4.2526860841423954e-06,
      "loss": 0.0007,
      "step": 304120
    },
    {
      "epoch": 118.11,
      "learning_rate": 4.252168284789644e-06,
      "loss": 0.0326,
      "step": 304130
    },
    {
      "epoch": 118.11,
      "learning_rate": 4.251650485436894e-06,
      "loss": 0.0007,
      "step": 304140
    },
    {
      "epoch": 118.12,
      "learning_rate": 4.251132686084143e-06,
      "loss": 0.0466,
      "step": 304150
    },
    {
      "epoch": 118.12,
      "learning_rate": 4.250614886731391e-06,
      "loss": 0.0431,
      "step": 304160
    },
    {
      "epoch": 118.12,
      "learning_rate": 4.250097087378641e-06,
      "loss": 0.0285,
      "step": 304170
    },
    {
      "epoch": 118.13,
      "learning_rate": 4.24957928802589e-06,
      "loss": 0.0241,
      "step": 304180
    },
    {
      "epoch": 118.13,
      "learning_rate": 4.249061488673139e-06,
      "loss": 0.0015,
      "step": 304190
    },
    {
      "epoch": 118.14,
      "learning_rate": 4.248543689320388e-06,
      "loss": 0.023,
      "step": 304200
    },
    {
      "epoch": 118.14,
      "learning_rate": 4.248025889967638e-06,
      "loss": 0.0692,
      "step": 304210
    },
    {
      "epoch": 118.14,
      "learning_rate": 4.2475080906148865e-06,
      "loss": 0.0015,
      "step": 304220
    },
    {
      "epoch": 118.15,
      "learning_rate": 4.246990291262136e-06,
      "loss": 0.0541,
      "step": 304230
    },
    {
      "epoch": 118.15,
      "learning_rate": 4.246472491909385e-06,
      "loss": 0.0779,
      "step": 304240
    },
    {
      "epoch": 118.16,
      "learning_rate": 4.2459546925566345e-06,
      "loss": 0.0012,
      "step": 304250
    },
    {
      "epoch": 118.16,
      "learning_rate": 4.245436893203883e-06,
      "loss": 0.0217,
      "step": 304260
    },
    {
      "epoch": 118.16,
      "learning_rate": 4.244919093851133e-06,
      "loss": 0.0784,
      "step": 304270
    },
    {
      "epoch": 118.17,
      "learning_rate": 4.244401294498382e-06,
      "loss": 0.1067,
      "step": 304280
    },
    {
      "epoch": 118.17,
      "learning_rate": 4.243883495145631e-06,
      "loss": 0.0602,
      "step": 304290
    },
    {
      "epoch": 118.17,
      "learning_rate": 4.24336569579288e-06,
      "loss": 0.0045,
      "step": 304300
    },
    {
      "epoch": 118.18,
      "learning_rate": 4.24284789644013e-06,
      "loss": 0.0201,
      "step": 304310
    },
    {
      "epoch": 118.18,
      "learning_rate": 4.2423300970873785e-06,
      "loss": 0.0451,
      "step": 304320
    },
    {
      "epoch": 118.19,
      "learning_rate": 4.241812297734628e-06,
      "loss": 0.0361,
      "step": 304330
    },
    {
      "epoch": 118.19,
      "learning_rate": 4.241294498381877e-06,
      "loss": 0.0171,
      "step": 304340
    },
    {
      "epoch": 118.19,
      "learning_rate": 4.2407766990291265e-06,
      "loss": 0.1083,
      "step": 304350
    },
    {
      "epoch": 118.2,
      "learning_rate": 4.240258899676375e-06,
      "loss": 0.0357,
      "step": 304360
    },
    {
      "epoch": 118.2,
      "learning_rate": 4.239741100323625e-06,
      "loss": 0.065,
      "step": 304370
    },
    {
      "epoch": 118.21,
      "learning_rate": 4.239223300970874e-06,
      "loss": 0.0057,
      "step": 304380
    },
    {
      "epoch": 118.21,
      "learning_rate": 4.238705501618123e-06,
      "loss": 0.0747,
      "step": 304390
    },
    {
      "epoch": 118.21,
      "learning_rate": 4.238187702265372e-06,
      "loss": 0.0098,
      "step": 304400
    },
    {
      "epoch": 118.22,
      "learning_rate": 4.237669902912622e-06,
      "loss": 0.0465,
      "step": 304410
    },
    {
      "epoch": 118.22,
      "learning_rate": 4.237152103559871e-06,
      "loss": 0.0995,
      "step": 304420
    },
    {
      "epoch": 118.23,
      "learning_rate": 4.23663430420712e-06,
      "loss": 0.0428,
      "step": 304430
    },
    {
      "epoch": 118.23,
      "learning_rate": 4.23611650485437e-06,
      "loss": 0.0807,
      "step": 304440
    },
    {
      "epoch": 118.23,
      "learning_rate": 4.2355987055016184e-06,
      "loss": 0.0828,
      "step": 304450
    },
    {
      "epoch": 118.24,
      "learning_rate": 4.235080906148868e-06,
      "loss": 0.0266,
      "step": 304460
    },
    {
      "epoch": 118.24,
      "learning_rate": 4.234563106796117e-06,
      "loss": 0.0473,
      "step": 304470
    },
    {
      "epoch": 118.24,
      "learning_rate": 4.2340453074433664e-06,
      "loss": 0.0352,
      "step": 304480
    },
    {
      "epoch": 118.25,
      "learning_rate": 4.233527508090615e-06,
      "loss": 0.0162,
      "step": 304490
    },
    {
      "epoch": 118.25,
      "learning_rate": 4.233009708737865e-06,
      "loss": 0.0332,
      "step": 304500
    },
    {
      "epoch": 118.26,
      "learning_rate": 4.232491909385114e-06,
      "loss": 0.0091,
      "step": 304510
    },
    {
      "epoch": 118.26,
      "learning_rate": 4.231974110032363e-06,
      "loss": 0.0275,
      "step": 304520
    },
    {
      "epoch": 118.26,
      "learning_rate": 4.231456310679612e-06,
      "loss": 0.0921,
      "step": 304530
    },
    {
      "epoch": 118.27,
      "learning_rate": 4.230938511326862e-06,
      "loss": 0.0402,
      "step": 304540
    },
    {
      "epoch": 118.27,
      "learning_rate": 4.23042071197411e-06,
      "loss": 0.1986,
      "step": 304550
    },
    {
      "epoch": 118.28,
      "learning_rate": 4.22990291262136e-06,
      "loss": 0.0733,
      "step": 304560
    },
    {
      "epoch": 118.28,
      "learning_rate": 4.229385113268609e-06,
      "loss": 0.0523,
      "step": 304570
    },
    {
      "epoch": 118.28,
      "learning_rate": 4.228867313915858e-06,
      "loss": 0.1387,
      "step": 304580
    },
    {
      "epoch": 118.29,
      "learning_rate": 4.228349514563107e-06,
      "loss": 0.009,
      "step": 304590
    },
    {
      "epoch": 118.29,
      "learning_rate": 4.227831715210357e-06,
      "loss": 0.0562,
      "step": 304600
    },
    {
      "epoch": 118.3,
      "learning_rate": 4.2273139158576055e-06,
      "loss": 0.1713,
      "step": 304610
    },
    {
      "epoch": 118.3,
      "learning_rate": 4.226796116504855e-06,
      "loss": 0.1227,
      "step": 304620
    },
    {
      "epoch": 118.3,
      "learning_rate": 4.226278317152104e-06,
      "loss": 0.1221,
      "step": 304630
    },
    {
      "epoch": 118.31,
      "learning_rate": 4.2257605177993535e-06,
      "loss": 0.0108,
      "step": 304640
    },
    {
      "epoch": 118.31,
      "learning_rate": 4.225242718446602e-06,
      "loss": 0.1494,
      "step": 304650
    },
    {
      "epoch": 118.31,
      "learning_rate": 4.224724919093852e-06,
      "loss": 0.1221,
      "step": 304660
    },
    {
      "epoch": 118.32,
      "learning_rate": 4.224207119741101e-06,
      "loss": 0.0123,
      "step": 304670
    },
    {
      "epoch": 118.32,
      "learning_rate": 4.22368932038835e-06,
      "loss": 0.0788,
      "step": 304680
    },
    {
      "epoch": 118.33,
      "learning_rate": 4.223171521035599e-06,
      "loss": 0.0095,
      "step": 304690
    },
    {
      "epoch": 118.33,
      "learning_rate": 4.222653721682849e-06,
      "loss": 0.0616,
      "step": 304700
    },
    {
      "epoch": 118.33,
      "learning_rate": 4.2221359223300975e-06,
      "loss": 0.133,
      "step": 304710
    },
    {
      "epoch": 118.34,
      "learning_rate": 4.221618122977347e-06,
      "loss": 0.0174,
      "step": 304720
    },
    {
      "epoch": 118.34,
      "learning_rate": 4.221100323624596e-06,
      "loss": 0.0079,
      "step": 304730
    },
    {
      "epoch": 118.35,
      "learning_rate": 4.2205825242718455e-06,
      "loss": 0.0004,
      "step": 304740
    },
    {
      "epoch": 118.35,
      "learning_rate": 4.220064724919094e-06,
      "loss": 0.1355,
      "step": 304750
    },
    {
      "epoch": 118.35,
      "learning_rate": 4.219546925566344e-06,
      "loss": 0.0346,
      "step": 304760
    },
    {
      "epoch": 118.36,
      "learning_rate": 4.219029126213593e-06,
      "loss": 0.0341,
      "step": 304770
    },
    {
      "epoch": 118.36,
      "learning_rate": 4.218511326860842e-06,
      "loss": 0.0807,
      "step": 304780
    },
    {
      "epoch": 118.37,
      "learning_rate": 4.217993527508091e-06,
      "loss": 0.0427,
      "step": 304790
    },
    {
      "epoch": 118.37,
      "learning_rate": 4.21747572815534e-06,
      "loss": 0.0801,
      "step": 304800
    },
    {
      "epoch": 118.37,
      "learning_rate": 4.2169579288025894e-06,
      "loss": 0.0468,
      "step": 304810
    },
    {
      "epoch": 118.38,
      "learning_rate": 4.216440129449838e-06,
      "loss": 0.144,
      "step": 304820
    },
    {
      "epoch": 118.38,
      "learning_rate": 4.215922330097088e-06,
      "loss": 0.0357,
      "step": 304830
    },
    {
      "epoch": 118.38,
      "learning_rate": 4.215404530744337e-06,
      "loss": 0.0089,
      "step": 304840
    },
    {
      "epoch": 118.39,
      "learning_rate": 4.214886731391586e-06,
      "loss": 0.0089,
      "step": 304850
    },
    {
      "epoch": 118.39,
      "learning_rate": 4.214368932038835e-06,
      "loss": 0.1129,
      "step": 304860
    },
    {
      "epoch": 118.4,
      "learning_rate": 4.213851132686085e-06,
      "loss": 0.1442,
      "step": 304870
    },
    {
      "epoch": 118.4,
      "learning_rate": 4.213333333333333e-06,
      "loss": 0.0954,
      "step": 304880
    },
    {
      "epoch": 118.4,
      "learning_rate": 4.212815533980583e-06,
      "loss": 0.0475,
      "step": 304890
    },
    {
      "epoch": 118.41,
      "learning_rate": 4.212297734627832e-06,
      "loss": 0.0468,
      "step": 304900
    },
    {
      "epoch": 118.41,
      "learning_rate": 4.211779935275081e-06,
      "loss": 0.0051,
      "step": 304910
    },
    {
      "epoch": 118.42,
      "learning_rate": 4.21126213592233e-06,
      "loss": 0.0085,
      "step": 304920
    },
    {
      "epoch": 118.42,
      "learning_rate": 4.21074433656958e-06,
      "loss": 0.0608,
      "step": 304930
    },
    {
      "epoch": 118.42,
      "learning_rate": 4.2102265372168285e-06,
      "loss": 0.0959,
      "step": 304940
    },
    {
      "epoch": 118.43,
      "learning_rate": 4.209708737864078e-06,
      "loss": 0.0552,
      "step": 304950
    },
    {
      "epoch": 118.43,
      "learning_rate": 4.209190938511327e-06,
      "loss": 0.0216,
      "step": 304960
    },
    {
      "epoch": 118.43,
      "learning_rate": 4.2086731391585765e-06,
      "loss": 0.0514,
      "step": 304970
    },
    {
      "epoch": 118.44,
      "learning_rate": 4.208155339805825e-06,
      "loss": 0.0717,
      "step": 304980
    },
    {
      "epoch": 118.44,
      "learning_rate": 4.207637540453075e-06,
      "loss": 0.0279,
      "step": 304990
    },
    {
      "epoch": 118.45,
      "learning_rate": 4.207119741100324e-06,
      "loss": 0.0001,
      "step": 305000
    },
    {
      "epoch": 118.45,
      "learning_rate": 4.206601941747573e-06,
      "loss": 0.0001,
      "step": 305010
    },
    {
      "epoch": 118.45,
      "learning_rate": 4.206084142394822e-06,
      "loss": 0.0642,
      "step": 305020
    },
    {
      "epoch": 118.46,
      "learning_rate": 4.205566343042072e-06,
      "loss": 0.0986,
      "step": 305030
    },
    {
      "epoch": 118.46,
      "learning_rate": 4.2050485436893205e-06,
      "loss": 0.1188,
      "step": 305040
    },
    {
      "epoch": 118.47,
      "learning_rate": 4.20453074433657e-06,
      "loss": 0.0729,
      "step": 305050
    },
    {
      "epoch": 118.47,
      "learning_rate": 4.204012944983819e-06,
      "loss": 0.0003,
      "step": 305060
    },
    {
      "epoch": 118.47,
      "learning_rate": 4.2034951456310685e-06,
      "loss": 0.1166,
      "step": 305070
    },
    {
      "epoch": 118.48,
      "learning_rate": 4.202977346278317e-06,
      "loss": 0.0235,
      "step": 305080
    },
    {
      "epoch": 118.48,
      "learning_rate": 4.202459546925567e-06,
      "loss": 0.0001,
      "step": 305090
    },
    {
      "epoch": 118.49,
      "learning_rate": 4.201941747572816e-06,
      "loss": 0.1036,
      "step": 305100
    },
    {
      "epoch": 118.49,
      "learning_rate": 4.201423948220065e-06,
      "loss": 0.0062,
      "step": 305110
    },
    {
      "epoch": 118.49,
      "learning_rate": 4.200906148867314e-06,
      "loss": 0.0414,
      "step": 305120
    },
    {
      "epoch": 118.5,
      "learning_rate": 4.200388349514564e-06,
      "loss": 0.0174,
      "step": 305130
    },
    {
      "epoch": 118.5,
      "learning_rate": 4.1998705501618124e-06,
      "loss": 0.0274,
      "step": 305140
    },
    {
      "epoch": 118.5,
      "learning_rate": 4.199352750809062e-06,
      "loss": 0.0925,
      "step": 305150
    },
    {
      "epoch": 118.51,
      "learning_rate": 4.198834951456311e-06,
      "loss": 0.0006,
      "step": 305160
    },
    {
      "epoch": 118.51,
      "learning_rate": 4.1983171521035604e-06,
      "loss": 0.0119,
      "step": 305170
    },
    {
      "epoch": 118.52,
      "learning_rate": 4.197799352750809e-06,
      "loss": 0.0769,
      "step": 305180
    },
    {
      "epoch": 118.52,
      "learning_rate": 4.197281553398059e-06,
      "loss": 0.0295,
      "step": 305190
    },
    {
      "epoch": 118.52,
      "learning_rate": 4.196763754045308e-06,
      "loss": 0.0265,
      "step": 305200
    },
    {
      "epoch": 118.53,
      "learning_rate": 4.196245954692557e-06,
      "loss": 0.0002,
      "step": 305210
    },
    {
      "epoch": 118.53,
      "learning_rate": 4.195728155339806e-06,
      "loss": 0.0607,
      "step": 305220
    },
    {
      "epoch": 118.54,
      "learning_rate": 4.195210355987056e-06,
      "loss": 0.0715,
      "step": 305230
    },
    {
      "epoch": 118.54,
      "learning_rate": 4.194692556634304e-06,
      "loss": 0.0241,
      "step": 305240
    },
    {
      "epoch": 118.54,
      "learning_rate": 4.194174757281554e-06,
      "loss": 0.027,
      "step": 305250
    },
    {
      "epoch": 118.55,
      "learning_rate": 4.193656957928803e-06,
      "loss": 0.0843,
      "step": 305260
    },
    {
      "epoch": 118.55,
      "learning_rate": 4.193139158576052e-06,
      "loss": 0.0001,
      "step": 305270
    },
    {
      "epoch": 118.56,
      "learning_rate": 4.192621359223301e-06,
      "loss": 0.0227,
      "step": 305280
    },
    {
      "epoch": 118.56,
      "learning_rate": 4.192103559870551e-06,
      "loss": 0.0885,
      "step": 305290
    },
    {
      "epoch": 118.56,
      "learning_rate": 4.1915857605177995e-06,
      "loss": 0.003,
      "step": 305300
    },
    {
      "epoch": 118.57,
      "learning_rate": 4.191067961165049e-06,
      "loss": 0.0637,
      "step": 305310
    },
    {
      "epoch": 118.57,
      "learning_rate": 4.190550161812298e-06,
      "loss": 0.049,
      "step": 305320
    },
    {
      "epoch": 118.57,
      "learning_rate": 4.1900323624595475e-06,
      "loss": 0.0568,
      "step": 305330
    },
    {
      "epoch": 118.58,
      "learning_rate": 4.189514563106796e-06,
      "loss": 0.1109,
      "step": 305340
    },
    {
      "epoch": 118.58,
      "learning_rate": 4.188996763754046e-06,
      "loss": 0.0183,
      "step": 305350
    },
    {
      "epoch": 118.59,
      "learning_rate": 4.188478964401295e-06,
      "loss": 0.0189,
      "step": 305360
    },
    {
      "epoch": 118.59,
      "learning_rate": 4.187961165048544e-06,
      "loss": 0.0005,
      "step": 305370
    },
    {
      "epoch": 118.59,
      "learning_rate": 4.187443365695793e-06,
      "loss": 0.0545,
      "step": 305380
    },
    {
      "epoch": 118.6,
      "learning_rate": 4.186925566343043e-06,
      "loss": 0.0882,
      "step": 305390
    },
    {
      "epoch": 118.6,
      "learning_rate": 4.1864077669902915e-06,
      "loss": 0.0024,
      "step": 305400
    },
    {
      "epoch": 118.61,
      "learning_rate": 4.185889967637541e-06,
      "loss": 0.1319,
      "step": 305410
    },
    {
      "epoch": 118.61,
      "learning_rate": 4.18537216828479e-06,
      "loss": 0.0582,
      "step": 305420
    },
    {
      "epoch": 118.61,
      "learning_rate": 4.1848543689320395e-06,
      "loss": 0.0006,
      "step": 305430
    },
    {
      "epoch": 118.62,
      "learning_rate": 4.184336569579288e-06,
      "loss": 0.0554,
      "step": 305440
    },
    {
      "epoch": 118.62,
      "learning_rate": 4.183818770226538e-06,
      "loss": 0.0792,
      "step": 305450
    },
    {
      "epoch": 118.63,
      "learning_rate": 4.183300970873787e-06,
      "loss": 0.0244,
      "step": 305460
    },
    {
      "epoch": 118.63,
      "learning_rate": 4.182783171521036e-06,
      "loss": 0.1235,
      "step": 305470
    },
    {
      "epoch": 118.63,
      "learning_rate": 4.182265372168285e-06,
      "loss": 0.006,
      "step": 305480
    },
    {
      "epoch": 118.64,
      "learning_rate": 4.181747572815535e-06,
      "loss": 0.0041,
      "step": 305490
    },
    {
      "epoch": 118.64,
      "learning_rate": 4.1812297734627834e-06,
      "loss": 0.0005,
      "step": 305500
    },
    {
      "epoch": 118.64,
      "learning_rate": 4.180711974110033e-06,
      "loss": 0.0979,
      "step": 305510
    },
    {
      "epoch": 118.65,
      "learning_rate": 4.180194174757282e-06,
      "loss": 0.0397,
      "step": 305520
    },
    {
      "epoch": 118.65,
      "learning_rate": 4.1796763754045314e-06,
      "loss": 0.1189,
      "step": 305530
    },
    {
      "epoch": 118.66,
      "learning_rate": 4.17915857605178e-06,
      "loss": 0.0145,
      "step": 305540
    },
    {
      "epoch": 118.66,
      "learning_rate": 4.17864077669903e-06,
      "loss": 0.0725,
      "step": 305550
    },
    {
      "epoch": 118.66,
      "learning_rate": 4.178122977346279e-06,
      "loss": 0.0131,
      "step": 305560
    },
    {
      "epoch": 118.67,
      "learning_rate": 4.177605177993528e-06,
      "loss": 0.1086,
      "step": 305570
    },
    {
      "epoch": 118.67,
      "learning_rate": 4.177087378640777e-06,
      "loss": 0.1194,
      "step": 305580
    },
    {
      "epoch": 118.68,
      "learning_rate": 4.176569579288027e-06,
      "loss": 0.0221,
      "step": 305590
    },
    {
      "epoch": 118.68,
      "learning_rate": 4.176051779935275e-06,
      "loss": 0.0527,
      "step": 305600
    },
    {
      "epoch": 118.68,
      "learning_rate": 4.175533980582525e-06,
      "loss": 0.1023,
      "step": 305610
    },
    {
      "epoch": 118.69,
      "learning_rate": 4.175016181229774e-06,
      "loss": 0.1207,
      "step": 305620
    },
    {
      "epoch": 118.69,
      "learning_rate": 4.1744983818770225e-06,
      "loss": 0.0185,
      "step": 305630
    },
    {
      "epoch": 118.7,
      "learning_rate": 4.173980582524272e-06,
      "loss": 0.0212,
      "step": 305640
    },
    {
      "epoch": 118.7,
      "learning_rate": 4.173462783171521e-06,
      "loss": 0.0644,
      "step": 305650
    },
    {
      "epoch": 118.7,
      "learning_rate": 4.1729449838187705e-06,
      "loss": 0.083,
      "step": 305660
    },
    {
      "epoch": 118.71,
      "learning_rate": 4.172427184466019e-06,
      "loss": 0.0897,
      "step": 305670
    },
    {
      "epoch": 118.71,
      "learning_rate": 4.171909385113269e-06,
      "loss": 0.0191,
      "step": 305680
    },
    {
      "epoch": 118.71,
      "learning_rate": 4.171391585760518e-06,
      "loss": 0.157,
      "step": 305690
    },
    {
      "epoch": 118.72,
      "learning_rate": 4.170873786407767e-06,
      "loss": 0.1191,
      "step": 305700
    },
    {
      "epoch": 118.72,
      "learning_rate": 4.170355987055016e-06,
      "loss": 0.0805,
      "step": 305710
    },
    {
      "epoch": 118.73,
      "learning_rate": 4.169838187702266e-06,
      "loss": 0.0028,
      "step": 305720
    },
    {
      "epoch": 118.73,
      "learning_rate": 4.1693203883495145e-06,
      "loss": 0.0411,
      "step": 305730
    },
    {
      "epoch": 118.73,
      "learning_rate": 4.168802588996764e-06,
      "loss": 0.0039,
      "step": 305740
    },
    {
      "epoch": 118.74,
      "learning_rate": 4.168284789644013e-06,
      "loss": 0.0145,
      "step": 305750
    },
    {
      "epoch": 118.74,
      "learning_rate": 4.1677669902912625e-06,
      "loss": 0.0601,
      "step": 305760
    },
    {
      "epoch": 118.75,
      "learning_rate": 4.167249190938511e-06,
      "loss": 0.0033,
      "step": 305770
    },
    {
      "epoch": 118.75,
      "learning_rate": 4.166731391585761e-06,
      "loss": 0.0185,
      "step": 305780
    },
    {
      "epoch": 118.75,
      "learning_rate": 4.16621359223301e-06,
      "loss": 0.1451,
      "step": 305790
    },
    {
      "epoch": 118.76,
      "learning_rate": 4.165695792880259e-06,
      "loss": 0.0002,
      "step": 305800
    },
    {
      "epoch": 118.76,
      "learning_rate": 4.165177993527508e-06,
      "loss": 0.0005,
      "step": 305810
    },
    {
      "epoch": 118.77,
      "learning_rate": 4.164660194174758e-06,
      "loss": 0.0253,
      "step": 305820
    },
    {
      "epoch": 118.77,
      "learning_rate": 4.1641423948220064e-06,
      "loss": 0.0351,
      "step": 305830
    },
    {
      "epoch": 118.77,
      "learning_rate": 4.163624595469256e-06,
      "loss": 0.0942,
      "step": 305840
    },
    {
      "epoch": 118.78,
      "learning_rate": 4.163106796116505e-06,
      "loss": 0.1116,
      "step": 305850
    },
    {
      "epoch": 118.78,
      "learning_rate": 4.1625889967637544e-06,
      "loss": 0.0565,
      "step": 305860
    },
    {
      "epoch": 118.78,
      "learning_rate": 4.162071197411003e-06,
      "loss": 0.0381,
      "step": 305870
    },
    {
      "epoch": 118.79,
      "learning_rate": 4.161553398058253e-06,
      "loss": 0.0009,
      "step": 305880
    },
    {
      "epoch": 118.79,
      "learning_rate": 4.161035598705502e-06,
      "loss": 0.0158,
      "step": 305890
    },
    {
      "epoch": 118.8,
      "learning_rate": 4.160517799352751e-06,
      "loss": 0.1461,
      "step": 305900
    },
    {
      "epoch": 118.8,
      "learning_rate": 4.16e-06,
      "loss": 0.0218,
      "step": 305910
    },
    {
      "epoch": 118.8,
      "learning_rate": 4.15948220064725e-06,
      "loss": 0.0665,
      "step": 305920
    },
    {
      "epoch": 118.81,
      "learning_rate": 4.158964401294498e-06,
      "loss": 0.0784,
      "step": 305930
    },
    {
      "epoch": 118.81,
      "learning_rate": 4.158446601941748e-06,
      "loss": 0.0355,
      "step": 305940
    },
    {
      "epoch": 118.82,
      "learning_rate": 4.157928802588997e-06,
      "loss": 0.0138,
      "step": 305950
    },
    {
      "epoch": 118.82,
      "learning_rate": 4.157411003236246e-06,
      "loss": 0.0007,
      "step": 305960
    },
    {
      "epoch": 118.82,
      "learning_rate": 4.156893203883495e-06,
      "loss": 0.0271,
      "step": 305970
    },
    {
      "epoch": 118.83,
      "learning_rate": 4.156375404530745e-06,
      "loss": 0.0218,
      "step": 305980
    },
    {
      "epoch": 118.83,
      "learning_rate": 4.1558576051779935e-06,
      "loss": 0.1237,
      "step": 305990
    },
    {
      "epoch": 118.83,
      "learning_rate": 4.155339805825243e-06,
      "loss": 0.1401,
      "step": 306000
    },
    {
      "epoch": 118.84,
      "learning_rate": 4.154822006472492e-06,
      "loss": 0.0014,
      "step": 306010
    },
    {
      "epoch": 118.84,
      "learning_rate": 4.1543042071197415e-06,
      "loss": 0.0373,
      "step": 306020
    },
    {
      "epoch": 118.85,
      "learning_rate": 4.15378640776699e-06,
      "loss": 0.0186,
      "step": 306030
    },
    {
      "epoch": 118.85,
      "learning_rate": 4.15326860841424e-06,
      "loss": 0.1324,
      "step": 306040
    },
    {
      "epoch": 118.85,
      "learning_rate": 4.152750809061489e-06,
      "loss": 0.0242,
      "step": 306050
    },
    {
      "epoch": 118.86,
      "learning_rate": 4.152233009708738e-06,
      "loss": 0.0107,
      "step": 306060
    },
    {
      "epoch": 118.86,
      "learning_rate": 4.151715210355987e-06,
      "loss": 0.0234,
      "step": 306070
    },
    {
      "epoch": 118.87,
      "learning_rate": 4.151197411003237e-06,
      "loss": 0.0605,
      "step": 306080
    },
    {
      "epoch": 118.87,
      "learning_rate": 4.150679611650486e-06,
      "loss": 0.073,
      "step": 306090
    },
    {
      "epoch": 118.87,
      "learning_rate": 4.150161812297735e-06,
      "loss": 0.0009,
      "step": 306100
    },
    {
      "epoch": 118.88,
      "learning_rate": 4.149644012944985e-06,
      "loss": 0.0087,
      "step": 306110
    },
    {
      "epoch": 118.88,
      "learning_rate": 4.1491262135922335e-06,
      "loss": 0.1229,
      "step": 306120
    },
    {
      "epoch": 118.89,
      "learning_rate": 4.148608414239483e-06,
      "loss": 0.035,
      "step": 306130
    },
    {
      "epoch": 118.89,
      "learning_rate": 4.148090614886732e-06,
      "loss": 0.0116,
      "step": 306140
    },
    {
      "epoch": 118.89,
      "learning_rate": 4.1475728155339815e-06,
      "loss": 0.0346,
      "step": 306150
    },
    {
      "epoch": 118.9,
      "learning_rate": 4.14705501618123e-06,
      "loss": 0.0109,
      "step": 306160
    },
    {
      "epoch": 118.9,
      "learning_rate": 4.14653721682848e-06,
      "loss": 0.0556,
      "step": 306170
    },
    {
      "epoch": 118.9,
      "learning_rate": 4.146019417475729e-06,
      "loss": 0.0864,
      "step": 306180
    },
    {
      "epoch": 118.91,
      "learning_rate": 4.145501618122978e-06,
      "loss": 0.0704,
      "step": 306190
    },
    {
      "epoch": 118.91,
      "learning_rate": 4.144983818770227e-06,
      "loss": 0.0025,
      "step": 306200
    },
    {
      "epoch": 118.92,
      "learning_rate": 4.144466019417477e-06,
      "loss": 0.0759,
      "step": 306210
    },
    {
      "epoch": 118.92,
      "learning_rate": 4.1439482200647254e-06,
      "loss": 0.0472,
      "step": 306220
    },
    {
      "epoch": 118.92,
      "learning_rate": 4.143430420711975e-06,
      "loss": 0.0013,
      "step": 306230
    },
    {
      "epoch": 118.93,
      "learning_rate": 4.142912621359224e-06,
      "loss": 0.0202,
      "step": 306240
    },
    {
      "epoch": 118.93,
      "learning_rate": 4.1423948220064734e-06,
      "loss": 0.0166,
      "step": 306250
    },
    {
      "epoch": 118.94,
      "learning_rate": 4.141877022653722e-06,
      "loss": 0.0499,
      "step": 306260
    },
    {
      "epoch": 118.94,
      "learning_rate": 4.141359223300971e-06,
      "loss": 0.0176,
      "step": 306270
    },
    {
      "epoch": 118.94,
      "learning_rate": 4.140841423948221e-06,
      "loss": 0.0106,
      "step": 306280
    },
    {
      "epoch": 118.95,
      "learning_rate": 4.140323624595469e-06,
      "loss": 0.0687,
      "step": 306290
    },
    {
      "epoch": 118.95,
      "learning_rate": 4.139805825242719e-06,
      "loss": 0.0861,
      "step": 306300
    },
    {
      "epoch": 118.96,
      "learning_rate": 4.139288025889968e-06,
      "loss": 0.0203,
      "step": 306310
    },
    {
      "epoch": 118.96,
      "learning_rate": 4.138770226537217e-06,
      "loss": 0.0128,
      "step": 306320
    },
    {
      "epoch": 118.96,
      "learning_rate": 4.138252427184466e-06,
      "loss": 0.0001,
      "step": 306330
    },
    {
      "epoch": 118.97,
      "learning_rate": 4.137734627831716e-06,
      "loss": 0.0974,
      "step": 306340
    },
    {
      "epoch": 118.97,
      "learning_rate": 4.1372168284789645e-06,
      "loss": 0.0945,
      "step": 306350
    },
    {
      "epoch": 118.97,
      "learning_rate": 4.136699029126214e-06,
      "loss": 0.0162,
      "step": 306360
    },
    {
      "epoch": 118.98,
      "learning_rate": 4.136181229773463e-06,
      "loss": 0.0396,
      "step": 306370
    },
    {
      "epoch": 118.98,
      "learning_rate": 4.1356634304207125e-06,
      "loss": 0.028,
      "step": 306380
    },
    {
      "epoch": 118.99,
      "learning_rate": 4.135145631067961e-06,
      "loss": 0.0649,
      "step": 306390
    },
    {
      "epoch": 118.99,
      "learning_rate": 4.134627831715211e-06,
      "loss": 0.0676,
      "step": 306400
    },
    {
      "epoch": 118.99,
      "learning_rate": 4.13411003236246e-06,
      "loss": 0.0206,
      "step": 306410
    },
    {
      "epoch": 119.0,
      "learning_rate": 4.133592233009709e-06,
      "loss": 0.0015,
      "step": 306420
    },
    {
      "epoch": 119.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.38671842217445374,
      "eval_runtime": 8.2613,
      "eval_samples_per_second": 440.001,
      "eval_steps_per_second": 55.076,
      "step": 306425
    },
    {
      "epoch": 119.0,
      "learning_rate": 4.133074433656958e-06,
      "loss": 0.1241,
      "step": 306430
    },
    {
      "epoch": 119.01,
      "learning_rate": 4.132556634304207e-06,
      "loss": 0.0918,
      "step": 306440
    },
    {
      "epoch": 119.01,
      "learning_rate": 4.1320388349514565e-06,
      "loss": 0.0206,
      "step": 306450
    },
    {
      "epoch": 119.01,
      "learning_rate": 4.131521035598705e-06,
      "loss": 0.017,
      "step": 306460
    },
    {
      "epoch": 119.02,
      "learning_rate": 4.131003236245955e-06,
      "loss": 0.104,
      "step": 306470
    },
    {
      "epoch": 119.02,
      "learning_rate": 4.130485436893204e-06,
      "loss": 0.0909,
      "step": 306480
    },
    {
      "epoch": 119.03,
      "learning_rate": 4.129967637540453e-06,
      "loss": 0.0111,
      "step": 306490
    },
    {
      "epoch": 119.03,
      "learning_rate": 4.129449838187702e-06,
      "loss": 0.0571,
      "step": 306500
    },
    {
      "epoch": 119.03,
      "learning_rate": 4.128932038834952e-06,
      "loss": 0.0859,
      "step": 306510
    },
    {
      "epoch": 119.04,
      "learning_rate": 4.1284142394822004e-06,
      "loss": 0.0287,
      "step": 306520
    },
    {
      "epoch": 119.04,
      "learning_rate": 4.12789644012945e-06,
      "loss": 0.0436,
      "step": 306530
    },
    {
      "epoch": 119.04,
      "learning_rate": 4.127378640776699e-06,
      "loss": 0.0402,
      "step": 306540
    },
    {
      "epoch": 119.05,
      "learning_rate": 4.1268608414239484e-06,
      "loss": 0.0194,
      "step": 306550
    },
    {
      "epoch": 119.05,
      "learning_rate": 4.126343042071197e-06,
      "loss": 0.0212,
      "step": 306560
    },
    {
      "epoch": 119.06,
      "learning_rate": 4.125825242718447e-06,
      "loss": 0.1084,
      "step": 306570
    },
    {
      "epoch": 119.06,
      "learning_rate": 4.125307443365696e-06,
      "loss": 0.0069,
      "step": 306580
    },
    {
      "epoch": 119.06,
      "learning_rate": 4.124789644012945e-06,
      "loss": 0.0594,
      "step": 306590
    },
    {
      "epoch": 119.07,
      "learning_rate": 4.124271844660194e-06,
      "loss": 0.0232,
      "step": 306600
    },
    {
      "epoch": 119.07,
      "learning_rate": 4.123754045307444e-06,
      "loss": 0.1606,
      "step": 306610
    },
    {
      "epoch": 119.08,
      "learning_rate": 4.123236245954692e-06,
      "loss": 0.0362,
      "step": 306620
    },
    {
      "epoch": 119.08,
      "learning_rate": 4.122718446601942e-06,
      "loss": 0.053,
      "step": 306630
    },
    {
      "epoch": 119.08,
      "learning_rate": 4.122200647249191e-06,
      "loss": 0.0511,
      "step": 306640
    },
    {
      "epoch": 119.09,
      "learning_rate": 4.12168284789644e-06,
      "loss": 0.0165,
      "step": 306650
    },
    {
      "epoch": 119.09,
      "learning_rate": 4.12116504854369e-06,
      "loss": 0.0001,
      "step": 306660
    },
    {
      "epoch": 119.1,
      "learning_rate": 4.120647249190939e-06,
      "loss": 0.0766,
      "step": 306670
    },
    {
      "epoch": 119.1,
      "learning_rate": 4.120129449838188e-06,
      "loss": 0.014,
      "step": 306680
    },
    {
      "epoch": 119.1,
      "learning_rate": 4.119611650485437e-06,
      "loss": 0.0073,
      "step": 306690
    },
    {
      "epoch": 119.11,
      "learning_rate": 4.119093851132687e-06,
      "loss": 0.0608,
      "step": 306700
    },
    {
      "epoch": 119.11,
      "learning_rate": 4.1185760517799355e-06,
      "loss": 0.0647,
      "step": 306710
    },
    {
      "epoch": 119.11,
      "learning_rate": 4.118058252427185e-06,
      "loss": 0.0724,
      "step": 306720
    },
    {
      "epoch": 119.12,
      "learning_rate": 4.117540453074434e-06,
      "loss": 0.0392,
      "step": 306730
    },
    {
      "epoch": 119.12,
      "learning_rate": 4.1170226537216836e-06,
      "loss": 0.0351,
      "step": 306740
    },
    {
      "epoch": 119.13,
      "learning_rate": 4.116504854368932e-06,
      "loss": 0.0262,
      "step": 306750
    },
    {
      "epoch": 119.13,
      "learning_rate": 4.115987055016182e-06,
      "loss": 0.069,
      "step": 306760
    },
    {
      "epoch": 119.13,
      "learning_rate": 4.115469255663431e-06,
      "loss": 0.0189,
      "step": 306770
    },
    {
      "epoch": 119.14,
      "learning_rate": 4.11495145631068e-06,
      "loss": 0.019,
      "step": 306780
    },
    {
      "epoch": 119.14,
      "learning_rate": 4.114433656957929e-06,
      "loss": 0.0055,
      "step": 306790
    },
    {
      "epoch": 119.15,
      "learning_rate": 4.113915857605179e-06,
      "loss": 0.0639,
      "step": 306800
    },
    {
      "epoch": 119.15,
      "learning_rate": 4.1133980582524275e-06,
      "loss": 0.0474,
      "step": 306810
    },
    {
      "epoch": 119.15,
      "learning_rate": 4.112880258899677e-06,
      "loss": 0.0548,
      "step": 306820
    },
    {
      "epoch": 119.16,
      "learning_rate": 4.112362459546926e-06,
      "loss": 0.0141,
      "step": 306830
    },
    {
      "epoch": 119.16,
      "learning_rate": 4.1118446601941755e-06,
      "loss": 0.0004,
      "step": 306840
    },
    {
      "epoch": 119.17,
      "learning_rate": 4.111326860841424e-06,
      "loss": 0.0454,
      "step": 306850
    },
    {
      "epoch": 119.17,
      "learning_rate": 4.110809061488674e-06,
      "loss": 0.0711,
      "step": 306860
    },
    {
      "epoch": 119.17,
      "learning_rate": 4.110291262135923e-06,
      "loss": 0.1555,
      "step": 306870
    },
    {
      "epoch": 119.18,
      "learning_rate": 4.109773462783172e-06,
      "loss": 0.0215,
      "step": 306880
    },
    {
      "epoch": 119.18,
      "learning_rate": 4.109255663430421e-06,
      "loss": 0.0926,
      "step": 306890
    },
    {
      "epoch": 119.18,
      "learning_rate": 4.108737864077671e-06,
      "loss": 0.0001,
      "step": 306900
    },
    {
      "epoch": 119.19,
      "learning_rate": 4.1082200647249194e-06,
      "loss": 0.0552,
      "step": 306910
    },
    {
      "epoch": 119.19,
      "learning_rate": 4.107702265372169e-06,
      "loss": 0.0724,
      "step": 306920
    },
    {
      "epoch": 119.2,
      "learning_rate": 4.107184466019418e-06,
      "loss": 0.0123,
      "step": 306930
    },
    {
      "epoch": 119.2,
      "learning_rate": 4.1066666666666674e-06,
      "loss": 0.0136,
      "step": 306940
    },
    {
      "epoch": 119.2,
      "learning_rate": 4.106148867313916e-06,
      "loss": 0.0829,
      "step": 306950
    },
    {
      "epoch": 119.21,
      "learning_rate": 4.105631067961166e-06,
      "loss": 0.0025,
      "step": 306960
    },
    {
      "epoch": 119.21,
      "learning_rate": 4.105113268608415e-06,
      "loss": 0.0334,
      "step": 306970
    },
    {
      "epoch": 119.22,
      "learning_rate": 4.104595469255664e-06,
      "loss": 0.0019,
      "step": 306980
    },
    {
      "epoch": 119.22,
      "learning_rate": 4.104077669902913e-06,
      "loss": 0.0009,
      "step": 306990
    },
    {
      "epoch": 119.22,
      "learning_rate": 4.103559870550163e-06,
      "loss": 0.0706,
      "step": 307000
    },
    {
      "epoch": 119.23,
      "learning_rate": 4.103042071197411e-06,
      "loss": 0.0726,
      "step": 307010
    },
    {
      "epoch": 119.23,
      "learning_rate": 4.102524271844661e-06,
      "loss": 0.0475,
      "step": 307020
    },
    {
      "epoch": 119.23,
      "learning_rate": 4.10200647249191e-06,
      "loss": 0.0171,
      "step": 307030
    },
    {
      "epoch": 119.24,
      "learning_rate": 4.101488673139159e-06,
      "loss": 0.0207,
      "step": 307040
    },
    {
      "epoch": 119.24,
      "learning_rate": 4.100970873786408e-06,
      "loss": 0.034,
      "step": 307050
    },
    {
      "epoch": 119.25,
      "learning_rate": 4.100453074433658e-06,
      "loss": 0.0997,
      "step": 307060
    },
    {
      "epoch": 119.25,
      "learning_rate": 4.0999352750809065e-06,
      "loss": 0.0283,
      "step": 307070
    },
    {
      "epoch": 119.25,
      "learning_rate": 4.099417475728155e-06,
      "loss": 0.0649,
      "step": 307080
    },
    {
      "epoch": 119.26,
      "learning_rate": 4.098899676375405e-06,
      "loss": 0.0914,
      "step": 307090
    },
    {
      "epoch": 119.26,
      "learning_rate": 4.098381877022654e-06,
      "loss": 0.0317,
      "step": 307100
    },
    {
      "epoch": 119.27,
      "learning_rate": 4.097864077669903e-06,
      "loss": 0.0287,
      "step": 307110
    },
    {
      "epoch": 119.27,
      "learning_rate": 4.097346278317152e-06,
      "loss": 0.0108,
      "step": 307120
    },
    {
      "epoch": 119.27,
      "learning_rate": 4.096828478964402e-06,
      "loss": 0.0003,
      "step": 307130
    },
    {
      "epoch": 119.28,
      "learning_rate": 4.0963106796116505e-06,
      "loss": 0.0789,
      "step": 307140
    },
    {
      "epoch": 119.28,
      "learning_rate": 4.0957928802589e-06,
      "loss": 0.0668,
      "step": 307150
    },
    {
      "epoch": 119.29,
      "learning_rate": 4.095275080906149e-06,
      "loss": 0.0091,
      "step": 307160
    },
    {
      "epoch": 119.29,
      "learning_rate": 4.0947572815533985e-06,
      "loss": 0.0027,
      "step": 307170
    },
    {
      "epoch": 119.29,
      "learning_rate": 4.094239482200647e-06,
      "loss": 0.0008,
      "step": 307180
    },
    {
      "epoch": 119.3,
      "learning_rate": 4.093721682847897e-06,
      "loss": 0.0078,
      "step": 307190
    },
    {
      "epoch": 119.3,
      "learning_rate": 4.093203883495146e-06,
      "loss": 0.0468,
      "step": 307200
    },
    {
      "epoch": 119.3,
      "learning_rate": 4.092686084142395e-06,
      "loss": 0.045,
      "step": 307210
    },
    {
      "epoch": 119.31,
      "learning_rate": 4.092168284789644e-06,
      "loss": 0.0067,
      "step": 307220
    },
    {
      "epoch": 119.31,
      "learning_rate": 4.091650485436894e-06,
      "loss": 0.0946,
      "step": 307230
    },
    {
      "epoch": 119.32,
      "learning_rate": 4.0911326860841424e-06,
      "loss": 0.1151,
      "step": 307240
    },
    {
      "epoch": 119.32,
      "learning_rate": 4.090614886731392e-06,
      "loss": 0.0557,
      "step": 307250
    },
    {
      "epoch": 119.32,
      "learning_rate": 4.090097087378641e-06,
      "loss": 0.0006,
      "step": 307260
    },
    {
      "epoch": 119.33,
      "learning_rate": 4.0895792880258904e-06,
      "loss": 0.0039,
      "step": 307270
    },
    {
      "epoch": 119.33,
      "learning_rate": 4.089061488673139e-06,
      "loss": 0.0289,
      "step": 307280
    },
    {
      "epoch": 119.34,
      "learning_rate": 4.088543689320389e-06,
      "loss": 0.0002,
      "step": 307290
    },
    {
      "epoch": 119.34,
      "learning_rate": 4.088025889967638e-06,
      "loss": 0.098,
      "step": 307300
    },
    {
      "epoch": 119.34,
      "learning_rate": 4.087508090614887e-06,
      "loss": 0.0009,
      "step": 307310
    },
    {
      "epoch": 119.35,
      "learning_rate": 4.086990291262136e-06,
      "loss": 0.0397,
      "step": 307320
    },
    {
      "epoch": 119.35,
      "learning_rate": 4.086472491909386e-06,
      "loss": 0.0319,
      "step": 307330
    },
    {
      "epoch": 119.36,
      "learning_rate": 4.085954692556634e-06,
      "loss": 0.0126,
      "step": 307340
    },
    {
      "epoch": 119.36,
      "learning_rate": 4.085436893203884e-06,
      "loss": 0.0163,
      "step": 307350
    },
    {
      "epoch": 119.36,
      "learning_rate": 4.084919093851133e-06,
      "loss": 0.1154,
      "step": 307360
    },
    {
      "epoch": 119.37,
      "learning_rate": 4.084401294498382e-06,
      "loss": 0.0076,
      "step": 307370
    },
    {
      "epoch": 119.37,
      "learning_rate": 4.083883495145631e-06,
      "loss": 0.0089,
      "step": 307380
    },
    {
      "epoch": 119.37,
      "learning_rate": 4.083365695792881e-06,
      "loss": 0.0822,
      "step": 307390
    },
    {
      "epoch": 119.38,
      "learning_rate": 4.0828478964401295e-06,
      "loss": 0.141,
      "step": 307400
    },
    {
      "epoch": 119.38,
      "learning_rate": 4.082330097087379e-06,
      "loss": 0.0561,
      "step": 307410
    },
    {
      "epoch": 119.39,
      "learning_rate": 4.081812297734628e-06,
      "loss": 0.0112,
      "step": 307420
    },
    {
      "epoch": 119.39,
      "learning_rate": 4.0812944983818775e-06,
      "loss": 0.0022,
      "step": 307430
    },
    {
      "epoch": 119.39,
      "learning_rate": 4.080776699029126e-06,
      "loss": 0.0573,
      "step": 307440
    },
    {
      "epoch": 119.4,
      "learning_rate": 4.080258899676376e-06,
      "loss": 0.0002,
      "step": 307450
    },
    {
      "epoch": 119.4,
      "learning_rate": 4.079741100323625e-06,
      "loss": 0.0968,
      "step": 307460
    },
    {
      "epoch": 119.41,
      "learning_rate": 4.079223300970874e-06,
      "loss": 0.0001,
      "step": 307470
    },
    {
      "epoch": 119.41,
      "learning_rate": 4.078705501618123e-06,
      "loss": 0.0195,
      "step": 307480
    },
    {
      "epoch": 119.41,
      "learning_rate": 4.078187702265373e-06,
      "loss": 0.0265,
      "step": 307490
    },
    {
      "epoch": 119.42,
      "learning_rate": 4.0776699029126215e-06,
      "loss": 0.0759,
      "step": 307500
    },
    {
      "epoch": 119.42,
      "learning_rate": 4.077152103559871e-06,
      "loss": 0.1338,
      "step": 307510
    },
    {
      "epoch": 119.43,
      "learning_rate": 4.07663430420712e-06,
      "loss": 0.0728,
      "step": 307520
    },
    {
      "epoch": 119.43,
      "learning_rate": 4.0761165048543695e-06,
      "loss": 0.0817,
      "step": 307530
    },
    {
      "epoch": 119.43,
      "learning_rate": 4.075598705501618e-06,
      "loss": 0.0916,
      "step": 307540
    },
    {
      "epoch": 119.44,
      "learning_rate": 4.075080906148868e-06,
      "loss": 0.0787,
      "step": 307550
    },
    {
      "epoch": 119.44,
      "learning_rate": 4.074563106796117e-06,
      "loss": 0.0087,
      "step": 307560
    },
    {
      "epoch": 119.44,
      "learning_rate": 4.074045307443366e-06,
      "loss": 0.0006,
      "step": 307570
    },
    {
      "epoch": 119.45,
      "learning_rate": 4.073527508090615e-06,
      "loss": 0.0456,
      "step": 307580
    },
    {
      "epoch": 119.45,
      "learning_rate": 4.073009708737865e-06,
      "loss": 0.01,
      "step": 307590
    },
    {
      "epoch": 119.46,
      "learning_rate": 4.0724919093851134e-06,
      "loss": 0.0018,
      "step": 307600
    },
    {
      "epoch": 119.46,
      "learning_rate": 4.071974110032363e-06,
      "loss": 0.0171,
      "step": 307610
    },
    {
      "epoch": 119.46,
      "learning_rate": 4.071456310679612e-06,
      "loss": 0.0175,
      "step": 307620
    },
    {
      "epoch": 119.47,
      "learning_rate": 4.0709385113268614e-06,
      "loss": 0.0936,
      "step": 307630
    },
    {
      "epoch": 119.47,
      "learning_rate": 4.07042071197411e-06,
      "loss": 0.0504,
      "step": 307640
    },
    {
      "epoch": 119.48,
      "learning_rate": 4.06990291262136e-06,
      "loss": 0.0891,
      "step": 307650
    },
    {
      "epoch": 119.48,
      "learning_rate": 4.069385113268609e-06,
      "loss": 0.0934,
      "step": 307660
    },
    {
      "epoch": 119.48,
      "learning_rate": 4.068867313915858e-06,
      "loss": 0.0326,
      "step": 307670
    },
    {
      "epoch": 119.49,
      "learning_rate": 4.068349514563107e-06,
      "loss": 0.0499,
      "step": 307680
    },
    {
      "epoch": 119.49,
      "learning_rate": 4.067831715210357e-06,
      "loss": 0.1323,
      "step": 307690
    },
    {
      "epoch": 119.5,
      "learning_rate": 4.067313915857605e-06,
      "loss": 0.1009,
      "step": 307700
    },
    {
      "epoch": 119.5,
      "learning_rate": 4.066796116504855e-06,
      "loss": 0.0562,
      "step": 307710
    },
    {
      "epoch": 119.5,
      "learning_rate": 4.066278317152104e-06,
      "loss": 0.0001,
      "step": 307720
    },
    {
      "epoch": 119.51,
      "learning_rate": 4.065760517799353e-06,
      "loss": 0.0347,
      "step": 307730
    },
    {
      "epoch": 119.51,
      "learning_rate": 4.065242718446602e-06,
      "loss": 0.0327,
      "step": 307740
    },
    {
      "epoch": 119.51,
      "learning_rate": 4.064724919093852e-06,
      "loss": 0.0072,
      "step": 307750
    },
    {
      "epoch": 119.52,
      "learning_rate": 4.0642071197411005e-06,
      "loss": 0.0722,
      "step": 307760
    },
    {
      "epoch": 119.52,
      "learning_rate": 4.06368932038835e-06,
      "loss": 0.0277,
      "step": 307770
    },
    {
      "epoch": 119.53,
      "learning_rate": 4.063171521035599e-06,
      "loss": 0.06,
      "step": 307780
    },
    {
      "epoch": 119.53,
      "learning_rate": 4.0626537216828486e-06,
      "loss": 0.1064,
      "step": 307790
    },
    {
      "epoch": 119.53,
      "learning_rate": 4.062135922330097e-06,
      "loss": 0.0053,
      "step": 307800
    },
    {
      "epoch": 119.54,
      "learning_rate": 4.061618122977347e-06,
      "loss": 0.0537,
      "step": 307810
    },
    {
      "epoch": 119.54,
      "learning_rate": 4.061100323624596e-06,
      "loss": 0.0127,
      "step": 307820
    },
    {
      "epoch": 119.55,
      "learning_rate": 4.060582524271845e-06,
      "loss": 0.0523,
      "step": 307830
    },
    {
      "epoch": 119.55,
      "learning_rate": 4.060064724919094e-06,
      "loss": 0.0243,
      "step": 307840
    },
    {
      "epoch": 119.55,
      "learning_rate": 4.059546925566344e-06,
      "loss": 0.0804,
      "step": 307850
    },
    {
      "epoch": 119.56,
      "learning_rate": 4.0590291262135925e-06,
      "loss": 0.0274,
      "step": 307860
    },
    {
      "epoch": 119.56,
      "learning_rate": 4.058511326860842e-06,
      "loss": 0.0016,
      "step": 307870
    },
    {
      "epoch": 119.57,
      "learning_rate": 4.057993527508091e-06,
      "loss": 0.0209,
      "step": 307880
    },
    {
      "epoch": 119.57,
      "learning_rate": 4.0574757281553405e-06,
      "loss": 0.0006,
      "step": 307890
    },
    {
      "epoch": 119.57,
      "learning_rate": 4.056957928802589e-06,
      "loss": 0.0877,
      "step": 307900
    },
    {
      "epoch": 119.58,
      "learning_rate": 4.056440129449838e-06,
      "loss": 0.0012,
      "step": 307910
    },
    {
      "epoch": 119.58,
      "learning_rate": 4.055922330097088e-06,
      "loss": 0.0294,
      "step": 307920
    },
    {
      "epoch": 119.58,
      "learning_rate": 4.0554045307443364e-06,
      "loss": 0.0107,
      "step": 307930
    },
    {
      "epoch": 119.59,
      "learning_rate": 4.054886731391586e-06,
      "loss": 0.0022,
      "step": 307940
    },
    {
      "epoch": 119.59,
      "learning_rate": 4.054368932038835e-06,
      "loss": 0.0536,
      "step": 307950
    },
    {
      "epoch": 119.6,
      "learning_rate": 4.0538511326860844e-06,
      "loss": 0.0256,
      "step": 307960
    },
    {
      "epoch": 119.6,
      "learning_rate": 4.053333333333333e-06,
      "loss": 0.0139,
      "step": 307970
    },
    {
      "epoch": 119.6,
      "learning_rate": 4.052815533980583e-06,
      "loss": 0.0283,
      "step": 307980
    },
    {
      "epoch": 119.61,
      "learning_rate": 4.052297734627832e-06,
      "loss": 0.0828,
      "step": 307990
    },
    {
      "epoch": 119.61,
      "learning_rate": 4.051779935275081e-06,
      "loss": 0.1688,
      "step": 308000
    },
    {
      "epoch": 119.62,
      "learning_rate": 4.05126213592233e-06,
      "loss": 0.0328,
      "step": 308010
    },
    {
      "epoch": 119.62,
      "learning_rate": 4.05074433656958e-06,
      "loss": 0.017,
      "step": 308020
    },
    {
      "epoch": 119.62,
      "learning_rate": 4.050226537216828e-06,
      "loss": 0.0001,
      "step": 308030
    },
    {
      "epoch": 119.63,
      "learning_rate": 4.049708737864078e-06,
      "loss": 0.0092,
      "step": 308040
    },
    {
      "epoch": 119.63,
      "learning_rate": 4.049190938511327e-06,
      "loss": 0.0212,
      "step": 308050
    },
    {
      "epoch": 119.63,
      "learning_rate": 4.048673139158576e-06,
      "loss": 0.0356,
      "step": 308060
    },
    {
      "epoch": 119.64,
      "learning_rate": 4.048155339805825e-06,
      "loss": 0.0104,
      "step": 308070
    },
    {
      "epoch": 119.64,
      "learning_rate": 4.047637540453075e-06,
      "loss": 0.0558,
      "step": 308080
    },
    {
      "epoch": 119.65,
      "learning_rate": 4.0471197411003235e-06,
      "loss": 0.0286,
      "step": 308090
    },
    {
      "epoch": 119.65,
      "learning_rate": 4.046601941747573e-06,
      "loss": 0.0076,
      "step": 308100
    },
    {
      "epoch": 119.65,
      "learning_rate": 4.046084142394822e-06,
      "loss": 0.1497,
      "step": 308110
    },
    {
      "epoch": 119.66,
      "learning_rate": 4.0455663430420715e-06,
      "loss": 0.0001,
      "step": 308120
    },
    {
      "epoch": 119.66,
      "learning_rate": 4.04504854368932e-06,
      "loss": 0.0376,
      "step": 308130
    },
    {
      "epoch": 119.67,
      "learning_rate": 4.04453074433657e-06,
      "loss": 0.0714,
      "step": 308140
    },
    {
      "epoch": 119.67,
      "learning_rate": 4.044012944983819e-06,
      "loss": 0.0292,
      "step": 308150
    },
    {
      "epoch": 119.67,
      "learning_rate": 4.043495145631068e-06,
      "loss": 0.001,
      "step": 308160
    },
    {
      "epoch": 119.68,
      "learning_rate": 4.042977346278317e-06,
      "loss": 0.001,
      "step": 308170
    },
    {
      "epoch": 119.68,
      "learning_rate": 4.042459546925567e-06,
      "loss": 0.0402,
      "step": 308180
    },
    {
      "epoch": 119.69,
      "learning_rate": 4.0419417475728155e-06,
      "loss": 0.002,
      "step": 308190
    },
    {
      "epoch": 119.69,
      "learning_rate": 4.041423948220065e-06,
      "loss": 0.0007,
      "step": 308200
    },
    {
      "epoch": 119.69,
      "learning_rate": 4.040906148867314e-06,
      "loss": 0.0417,
      "step": 308210
    },
    {
      "epoch": 119.7,
      "learning_rate": 4.0403883495145635e-06,
      "loss": 0.0249,
      "step": 308220
    },
    {
      "epoch": 119.7,
      "learning_rate": 4.039870550161812e-06,
      "loss": 0.022,
      "step": 308230
    },
    {
      "epoch": 119.7,
      "learning_rate": 4.039352750809062e-06,
      "loss": 0.0258,
      "step": 308240
    },
    {
      "epoch": 119.71,
      "learning_rate": 4.038834951456311e-06,
      "loss": 0.1113,
      "step": 308250
    },
    {
      "epoch": 119.71,
      "learning_rate": 4.03831715210356e-06,
      "loss": 0.1169,
      "step": 308260
    },
    {
      "epoch": 119.72,
      "learning_rate": 4.037799352750809e-06,
      "loss": 0.0002,
      "step": 308270
    },
    {
      "epoch": 119.72,
      "learning_rate": 4.037281553398059e-06,
      "loss": 0.0005,
      "step": 308280
    },
    {
      "epoch": 119.72,
      "learning_rate": 4.0367637540453074e-06,
      "loss": 0.0002,
      "step": 308290
    },
    {
      "epoch": 119.73,
      "learning_rate": 4.036245954692557e-06,
      "loss": 0.1028,
      "step": 308300
    },
    {
      "epoch": 119.73,
      "learning_rate": 4.035728155339806e-06,
      "loss": 0.0025,
      "step": 308310
    },
    {
      "epoch": 119.74,
      "learning_rate": 4.0352103559870554e-06,
      "loss": 0.0912,
      "step": 308320
    },
    {
      "epoch": 119.74,
      "learning_rate": 4.034692556634305e-06,
      "loss": 0.0242,
      "step": 308330
    },
    {
      "epoch": 119.74,
      "learning_rate": 4.034174757281554e-06,
      "loss": 0.0373,
      "step": 308340
    },
    {
      "epoch": 119.75,
      "learning_rate": 4.0336569579288034e-06,
      "loss": 0.0166,
      "step": 308350
    },
    {
      "epoch": 119.75,
      "learning_rate": 4.033139158576052e-06,
      "loss": 0.0517,
      "step": 308360
    },
    {
      "epoch": 119.76,
      "learning_rate": 4.032621359223302e-06,
      "loss": 0.0258,
      "step": 308370
    },
    {
      "epoch": 119.76,
      "learning_rate": 4.032103559870551e-06,
      "loss": 0.002,
      "step": 308380
    },
    {
      "epoch": 119.76,
      "learning_rate": 4.0315857605178e-06,
      "loss": 0.0318,
      "step": 308390
    },
    {
      "epoch": 119.77,
      "learning_rate": 4.031067961165049e-06,
      "loss": 0.0502,
      "step": 308400
    },
    {
      "epoch": 119.77,
      "learning_rate": 4.030550161812299e-06,
      "loss": 0.0826,
      "step": 308410
    },
    {
      "epoch": 119.77,
      "learning_rate": 4.030032362459547e-06,
      "loss": 0.0584,
      "step": 308420
    },
    {
      "epoch": 119.78,
      "learning_rate": 4.029514563106797e-06,
      "loss": 0.0171,
      "step": 308430
    },
    {
      "epoch": 119.78,
      "learning_rate": 4.028996763754046e-06,
      "loss": 0.1051,
      "step": 308440
    },
    {
      "epoch": 119.79,
      "learning_rate": 4.028478964401295e-06,
      "loss": 0.0499,
      "step": 308450
    },
    {
      "epoch": 119.79,
      "learning_rate": 4.027961165048544e-06,
      "loss": 0.0305,
      "step": 308460
    },
    {
      "epoch": 119.79,
      "learning_rate": 4.027443365695794e-06,
      "loss": 0.0124,
      "step": 308470
    },
    {
      "epoch": 119.8,
      "learning_rate": 4.0269255663430426e-06,
      "loss": 0.0804,
      "step": 308480
    },
    {
      "epoch": 119.8,
      "learning_rate": 4.026407766990292e-06,
      "loss": 0.0622,
      "step": 308490
    },
    {
      "epoch": 119.81,
      "learning_rate": 4.025889967637541e-06,
      "loss": 0.0432,
      "step": 308500
    },
    {
      "epoch": 119.81,
      "learning_rate": 4.0253721682847906e-06,
      "loss": 0.0081,
      "step": 308510
    },
    {
      "epoch": 119.81,
      "learning_rate": 4.024854368932039e-06,
      "loss": 0.0002,
      "step": 308520
    },
    {
      "epoch": 119.82,
      "learning_rate": 4.024336569579289e-06,
      "loss": 0.0312,
      "step": 308530
    },
    {
      "epoch": 119.82,
      "learning_rate": 4.023818770226538e-06,
      "loss": 0.1119,
      "step": 308540
    },
    {
      "epoch": 119.83,
      "learning_rate": 4.0233009708737865e-06,
      "loss": 0.0014,
      "step": 308550
    },
    {
      "epoch": 119.83,
      "learning_rate": 4.022783171521036e-06,
      "loss": 0.0232,
      "step": 308560
    },
    {
      "epoch": 119.83,
      "learning_rate": 4.022265372168285e-06,
      "loss": 0.0008,
      "step": 308570
    },
    {
      "epoch": 119.84,
      "learning_rate": 4.0217475728155345e-06,
      "loss": 0.0098,
      "step": 308580
    },
    {
      "epoch": 119.84,
      "learning_rate": 4.021229773462783e-06,
      "loss": 0.0531,
      "step": 308590
    },
    {
      "epoch": 119.84,
      "learning_rate": 4.020711974110033e-06,
      "loss": 0.0743,
      "step": 308600
    },
    {
      "epoch": 119.85,
      "learning_rate": 4.020194174757282e-06,
      "loss": 0.0759,
      "step": 308610
    },
    {
      "epoch": 119.85,
      "learning_rate": 4.019676375404531e-06,
      "loss": 0.0199,
      "step": 308620
    },
    {
      "epoch": 119.86,
      "learning_rate": 4.01915857605178e-06,
      "loss": 0.0176,
      "step": 308630
    },
    {
      "epoch": 119.86,
      "learning_rate": 4.01864077669903e-06,
      "loss": 0.0557,
      "step": 308640
    },
    {
      "epoch": 119.86,
      "learning_rate": 4.0181229773462784e-06,
      "loss": 0.001,
      "step": 308650
    },
    {
      "epoch": 119.87,
      "learning_rate": 4.017605177993528e-06,
      "loss": 0.0003,
      "step": 308660
    },
    {
      "epoch": 119.87,
      "learning_rate": 4.017087378640777e-06,
      "loss": 0.009,
      "step": 308670
    },
    {
      "epoch": 119.88,
      "learning_rate": 4.0165695792880264e-06,
      "loss": 0.0473,
      "step": 308680
    },
    {
      "epoch": 119.88,
      "learning_rate": 4.016051779935275e-06,
      "loss": 0.1472,
      "step": 308690
    },
    {
      "epoch": 119.88,
      "learning_rate": 4.015533980582525e-06,
      "loss": 0.0503,
      "step": 308700
    },
    {
      "epoch": 119.89,
      "learning_rate": 4.015016181229774e-06,
      "loss": 0.0244,
      "step": 308710
    },
    {
      "epoch": 119.89,
      "learning_rate": 4.014498381877022e-06,
      "loss": 0.0692,
      "step": 308720
    },
    {
      "epoch": 119.9,
      "learning_rate": 4.013980582524272e-06,
      "loss": 0.0827,
      "step": 308730
    },
    {
      "epoch": 119.9,
      "learning_rate": 4.013462783171521e-06,
      "loss": 0.0254,
      "step": 308740
    },
    {
      "epoch": 119.9,
      "learning_rate": 4.01294498381877e-06,
      "loss": 0.1048,
      "step": 308750
    },
    {
      "epoch": 119.91,
      "learning_rate": 4.012427184466019e-06,
      "loss": 0.0287,
      "step": 308760
    },
    {
      "epoch": 119.91,
      "learning_rate": 4.011909385113269e-06,
      "loss": 0.0825,
      "step": 308770
    },
    {
      "epoch": 119.91,
      "learning_rate": 4.0113915857605175e-06,
      "loss": 0.0207,
      "step": 308780
    },
    {
      "epoch": 119.92,
      "learning_rate": 4.010873786407767e-06,
      "loss": 0.0272,
      "step": 308790
    },
    {
      "epoch": 119.92,
      "learning_rate": 4.010355987055016e-06,
      "loss": 0.2165,
      "step": 308800
    },
    {
      "epoch": 119.93,
      "learning_rate": 4.0098381877022655e-06,
      "loss": 0.0034,
      "step": 308810
    },
    {
      "epoch": 119.93,
      "learning_rate": 4.009320388349514e-06,
      "loss": 0.0831,
      "step": 308820
    },
    {
      "epoch": 119.93,
      "learning_rate": 4.008802588996764e-06,
      "loss": 0.0905,
      "step": 308830
    },
    {
      "epoch": 119.94,
      "learning_rate": 4.008284789644013e-06,
      "loss": 0.0001,
      "step": 308840
    },
    {
      "epoch": 119.94,
      "learning_rate": 4.007766990291262e-06,
      "loss": 0.0549,
      "step": 308850
    },
    {
      "epoch": 119.95,
      "learning_rate": 4.007249190938511e-06,
      "loss": 0.0026,
      "step": 308860
    },
    {
      "epoch": 119.95,
      "learning_rate": 4.006731391585761e-06,
      "loss": 0.1377,
      "step": 308870
    },
    {
      "epoch": 119.95,
      "learning_rate": 4.0062135922330095e-06,
      "loss": 0.0348,
      "step": 308880
    },
    {
      "epoch": 119.96,
      "learning_rate": 4.005695792880259e-06,
      "loss": 0.0661,
      "step": 308890
    },
    {
      "epoch": 119.96,
      "learning_rate": 4.005177993527509e-06,
      "loss": 0.1015,
      "step": 308900
    },
    {
      "epoch": 119.97,
      "learning_rate": 4.0046601941747575e-06,
      "loss": 0.1233,
      "step": 308910
    },
    {
      "epoch": 119.97,
      "learning_rate": 4.004142394822007e-06,
      "loss": 0.044,
      "step": 308920
    },
    {
      "epoch": 119.97,
      "learning_rate": 4.003624595469256e-06,
      "loss": 0.0094,
      "step": 308930
    },
    {
      "epoch": 119.98,
      "learning_rate": 4.0031067961165055e-06,
      "loss": 0.0125,
      "step": 308940
    },
    {
      "epoch": 119.98,
      "learning_rate": 4.002588996763754e-06,
      "loss": 0.0628,
      "step": 308950
    },
    {
      "epoch": 119.98,
      "learning_rate": 4.002071197411004e-06,
      "loss": 0.0008,
      "step": 308960
    },
    {
      "epoch": 119.99,
      "learning_rate": 4.001553398058253e-06,
      "loss": 0.0512,
      "step": 308970
    },
    {
      "epoch": 119.99,
      "learning_rate": 4.001035598705502e-06,
      "loss": 0.0168,
      "step": 308980
    },
    {
      "epoch": 120.0,
      "learning_rate": 4.000517799352751e-06,
      "loss": 0.0195,
      "step": 308990
    },
    {
      "epoch": 120.0,
      "learning_rate": 4.000000000000001e-06,
      "loss": 0.1152,
      "step": 309000
    },
    {
      "epoch": 120.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.38419002294540405,
      "eval_runtime": 8.3091,
      "eval_samples_per_second": 437.472,
      "eval_steps_per_second": 54.759,
      "step": 309000
    },
    {
      "epoch": 120.0,
      "learning_rate": 3.9994822006472494e-06,
      "loss": 0.0991,
      "step": 309010
    },
    {
      "epoch": 120.01,
      "learning_rate": 3.998964401294499e-06,
      "loss": 0.1227,
      "step": 309020
    },
    {
      "epoch": 120.01,
      "learning_rate": 3.998446601941748e-06,
      "loss": 0.0653,
      "step": 309030
    },
    {
      "epoch": 120.02,
      "learning_rate": 3.9979288025889974e-06,
      "loss": 0.139,
      "step": 309040
    },
    {
      "epoch": 120.02,
      "learning_rate": 3.997411003236246e-06,
      "loss": 0.1291,
      "step": 309050
    },
    {
      "epoch": 120.02,
      "learning_rate": 3.996893203883496e-06,
      "loss": 0.0001,
      "step": 309060
    },
    {
      "epoch": 120.03,
      "learning_rate": 3.996375404530745e-06,
      "loss": 0.0379,
      "step": 309070
    },
    {
      "epoch": 120.03,
      "learning_rate": 3.995857605177994e-06,
      "loss": 0.0552,
      "step": 309080
    },
    {
      "epoch": 120.03,
      "learning_rate": 3.995339805825243e-06,
      "loss": 0.0618,
      "step": 309090
    },
    {
      "epoch": 120.04,
      "learning_rate": 3.994822006472493e-06,
      "loss": 0.1623,
      "step": 309100
    },
    {
      "epoch": 120.04,
      "learning_rate": 3.994304207119741e-06,
      "loss": 0.0001,
      "step": 309110
    },
    {
      "epoch": 120.05,
      "learning_rate": 3.993786407766991e-06,
      "loss": 0.033,
      "step": 309120
    },
    {
      "epoch": 120.05,
      "learning_rate": 3.99326860841424e-06,
      "loss": 0.0223,
      "step": 309130
    },
    {
      "epoch": 120.05,
      "learning_rate": 3.992750809061489e-06,
      "loss": 0.0015,
      "step": 309140
    },
    {
      "epoch": 120.06,
      "learning_rate": 3.992233009708738e-06,
      "loss": 0.0372,
      "step": 309150
    },
    {
      "epoch": 120.06,
      "learning_rate": 3.991715210355988e-06,
      "loss": 0.1175,
      "step": 309160
    },
    {
      "epoch": 120.07,
      "learning_rate": 3.9911974110032365e-06,
      "loss": 0.1884,
      "step": 309170
    },
    {
      "epoch": 120.07,
      "learning_rate": 3.990679611650486e-06,
      "loss": 0.0226,
      "step": 309180
    },
    {
      "epoch": 120.07,
      "learning_rate": 3.990161812297735e-06,
      "loss": 0.0549,
      "step": 309190
    },
    {
      "epoch": 120.08,
      "learning_rate": 3.9896440129449846e-06,
      "loss": 0.0871,
      "step": 309200
    },
    {
      "epoch": 120.08,
      "learning_rate": 3.989126213592233e-06,
      "loss": 0.1964,
      "step": 309210
    },
    {
      "epoch": 120.09,
      "learning_rate": 3.988608414239483e-06,
      "loss": 0.0671,
      "step": 309220
    },
    {
      "epoch": 120.09,
      "learning_rate": 3.988090614886732e-06,
      "loss": 0.0345,
      "step": 309230
    },
    {
      "epoch": 120.09,
      "learning_rate": 3.987572815533981e-06,
      "loss": 0.0273,
      "step": 309240
    },
    {
      "epoch": 120.1,
      "learning_rate": 3.98705501618123e-06,
      "loss": 0.0198,
      "step": 309250
    },
    {
      "epoch": 120.1,
      "learning_rate": 3.98653721682848e-06,
      "loss": 0.0088,
      "step": 309260
    },
    {
      "epoch": 120.1,
      "learning_rate": 3.9860194174757285e-06,
      "loss": 0.063,
      "step": 309270
    },
    {
      "epoch": 120.11,
      "learning_rate": 3.985501618122978e-06,
      "loss": 0.047,
      "step": 309280
    },
    {
      "epoch": 120.11,
      "learning_rate": 3.984983818770227e-06,
      "loss": 0.0092,
      "step": 309290
    },
    {
      "epoch": 120.12,
      "learning_rate": 3.9844660194174765e-06,
      "loss": 0.0571,
      "step": 309300
    },
    {
      "epoch": 120.12,
      "learning_rate": 3.983948220064725e-06,
      "loss": 0.0746,
      "step": 309310
    },
    {
      "epoch": 120.12,
      "learning_rate": 3.983430420711975e-06,
      "loss": 0.025,
      "step": 309320
    },
    {
      "epoch": 120.13,
      "learning_rate": 3.982912621359224e-06,
      "loss": 0.0226,
      "step": 309330
    },
    {
      "epoch": 120.13,
      "learning_rate": 3.982394822006473e-06,
      "loss": 0.1043,
      "step": 309340
    },
    {
      "epoch": 120.14,
      "learning_rate": 3.981877022653722e-06,
      "loss": 0.0517,
      "step": 309350
    },
    {
      "epoch": 120.14,
      "learning_rate": 3.981359223300971e-06,
      "loss": 0.0497,
      "step": 309360
    },
    {
      "epoch": 120.14,
      "learning_rate": 3.9808414239482204e-06,
      "loss": 0.0075,
      "step": 309370
    },
    {
      "epoch": 120.15,
      "learning_rate": 3.980323624595469e-06,
      "loss": 0.0088,
      "step": 309380
    },
    {
      "epoch": 120.15,
      "learning_rate": 3.979805825242719e-06,
      "loss": 0.1232,
      "step": 309390
    },
    {
      "epoch": 120.16,
      "learning_rate": 3.979288025889968e-06,
      "loss": 0.0004,
      "step": 309400
    },
    {
      "epoch": 120.16,
      "learning_rate": 3.978770226537217e-06,
      "loss": 0.1045,
      "step": 309410
    },
    {
      "epoch": 120.16,
      "learning_rate": 3.978252427184466e-06,
      "loss": 0.0988,
      "step": 309420
    },
    {
      "epoch": 120.17,
      "learning_rate": 3.977734627831716e-06,
      "loss": 0.0009,
      "step": 309430
    },
    {
      "epoch": 120.17,
      "learning_rate": 3.977216828478964e-06,
      "loss": 0.0608,
      "step": 309440
    },
    {
      "epoch": 120.17,
      "learning_rate": 3.976699029126214e-06,
      "loss": 0.0651,
      "step": 309450
    },
    {
      "epoch": 120.18,
      "learning_rate": 3.976181229773463e-06,
      "loss": 0.004,
      "step": 309460
    },
    {
      "epoch": 120.18,
      "learning_rate": 3.975663430420712e-06,
      "loss": 0.0423,
      "step": 309470
    },
    {
      "epoch": 120.19,
      "learning_rate": 3.975145631067961e-06,
      "loss": 0.0643,
      "step": 309480
    },
    {
      "epoch": 120.19,
      "learning_rate": 3.974627831715211e-06,
      "loss": 0.042,
      "step": 309490
    },
    {
      "epoch": 120.19,
      "learning_rate": 3.9741100323624595e-06,
      "loss": 0.0447,
      "step": 309500
    },
    {
      "epoch": 120.2,
      "learning_rate": 3.973592233009709e-06,
      "loss": 0.0612,
      "step": 309510
    },
    {
      "epoch": 120.2,
      "learning_rate": 3.973074433656958e-06,
      "loss": 0.0008,
      "step": 309520
    },
    {
      "epoch": 120.21,
      "learning_rate": 3.9725566343042076e-06,
      "loss": 0.109,
      "step": 309530
    },
    {
      "epoch": 120.21,
      "learning_rate": 3.972038834951456e-06,
      "loss": 0.0013,
      "step": 309540
    },
    {
      "epoch": 120.21,
      "learning_rate": 3.971521035598706e-06,
      "loss": 0.0799,
      "step": 309550
    },
    {
      "epoch": 120.22,
      "learning_rate": 3.971003236245955e-06,
      "loss": 0.0152,
      "step": 309560
    },
    {
      "epoch": 120.22,
      "learning_rate": 3.970485436893204e-06,
      "loss": 0.0951,
      "step": 309570
    },
    {
      "epoch": 120.23,
      "learning_rate": 3.969967637540453e-06,
      "loss": 0.0104,
      "step": 309580
    },
    {
      "epoch": 120.23,
      "learning_rate": 3.969449838187703e-06,
      "loss": 0.0356,
      "step": 309590
    },
    {
      "epoch": 120.23,
      "learning_rate": 3.9689320388349515e-06,
      "loss": 0.0837,
      "step": 309600
    },
    {
      "epoch": 120.24,
      "learning_rate": 3.968414239482201e-06,
      "loss": 0.0182,
      "step": 309610
    },
    {
      "epoch": 120.24,
      "learning_rate": 3.96789644012945e-06,
      "loss": 0.1028,
      "step": 309620
    },
    {
      "epoch": 120.24,
      "learning_rate": 3.9673786407766995e-06,
      "loss": 0.1984,
      "step": 309630
    },
    {
      "epoch": 120.25,
      "learning_rate": 3.966860841423948e-06,
      "loss": 0.0561,
      "step": 309640
    },
    {
      "epoch": 120.25,
      "learning_rate": 3.966343042071198e-06,
      "loss": 0.019,
      "step": 309650
    },
    {
      "epoch": 120.26,
      "learning_rate": 3.965825242718447e-06,
      "loss": 0.115,
      "step": 309660
    },
    {
      "epoch": 120.26,
      "learning_rate": 3.965307443365696e-06,
      "loss": 0.0393,
      "step": 309670
    },
    {
      "epoch": 120.26,
      "learning_rate": 3.964789644012945e-06,
      "loss": 0.0822,
      "step": 309680
    },
    {
      "epoch": 120.27,
      "learning_rate": 3.964271844660195e-06,
      "loss": 0.0858,
      "step": 309690
    },
    {
      "epoch": 120.27,
      "learning_rate": 3.9637540453074434e-06,
      "loss": 0.0085,
      "step": 309700
    },
    {
      "epoch": 120.28,
      "learning_rate": 3.963236245954693e-06,
      "loss": 0.0644,
      "step": 309710
    },
    {
      "epoch": 120.28,
      "learning_rate": 3.962718446601942e-06,
      "loss": 0.1916,
      "step": 309720
    },
    {
      "epoch": 120.28,
      "learning_rate": 3.9622006472491914e-06,
      "loss": 0.1138,
      "step": 309730
    },
    {
      "epoch": 120.29,
      "learning_rate": 3.96168284789644e-06,
      "loss": 0.0336,
      "step": 309740
    },
    {
      "epoch": 120.29,
      "learning_rate": 3.96116504854369e-06,
      "loss": 0.0164,
      "step": 309750
    },
    {
      "epoch": 120.3,
      "learning_rate": 3.960647249190939e-06,
      "loss": 0.0296,
      "step": 309760
    },
    {
      "epoch": 120.3,
      "learning_rate": 3.960129449838188e-06,
      "loss": 0.1213,
      "step": 309770
    },
    {
      "epoch": 120.3,
      "learning_rate": 3.959611650485437e-06,
      "loss": 0.0393,
      "step": 309780
    },
    {
      "epoch": 120.31,
      "learning_rate": 3.959093851132687e-06,
      "loss": 0.0003,
      "step": 309790
    },
    {
      "epoch": 120.31,
      "learning_rate": 3.958576051779935e-06,
      "loss": 0.1484,
      "step": 309800
    },
    {
      "epoch": 120.31,
      "learning_rate": 3.958058252427185e-06,
      "loss": 0.0008,
      "step": 309810
    },
    {
      "epoch": 120.32,
      "learning_rate": 3.957540453074434e-06,
      "loss": 0.012,
      "step": 309820
    },
    {
      "epoch": 120.32,
      "learning_rate": 3.957022653721683e-06,
      "loss": 0.0438,
      "step": 309830
    },
    {
      "epoch": 120.33,
      "learning_rate": 3.956504854368932e-06,
      "loss": 0.0162,
      "step": 309840
    },
    {
      "epoch": 120.33,
      "learning_rate": 3.955987055016182e-06,
      "loss": 0.0646,
      "step": 309850
    },
    {
      "epoch": 120.33,
      "learning_rate": 3.9554692556634305e-06,
      "loss": 0.0738,
      "step": 309860
    },
    {
      "epoch": 120.34,
      "learning_rate": 3.95495145631068e-06,
      "loss": 0.0769,
      "step": 309870
    },
    {
      "epoch": 120.34,
      "learning_rate": 3.954433656957929e-06,
      "loss": 0.0524,
      "step": 309880
    },
    {
      "epoch": 120.35,
      "learning_rate": 3.9539158576051786e-06,
      "loss": 0.0555,
      "step": 309890
    },
    {
      "epoch": 120.35,
      "learning_rate": 3.953398058252427e-06,
      "loss": 0.0893,
      "step": 309900
    },
    {
      "epoch": 120.35,
      "learning_rate": 3.952880258899677e-06,
      "loss": 0.1085,
      "step": 309910
    },
    {
      "epoch": 120.36,
      "learning_rate": 3.952362459546926e-06,
      "loss": 0.1031,
      "step": 309920
    },
    {
      "epoch": 120.36,
      "learning_rate": 3.951844660194175e-06,
      "loss": 0.0242,
      "step": 309930
    },
    {
      "epoch": 120.37,
      "learning_rate": 3.951326860841424e-06,
      "loss": 0.0641,
      "step": 309940
    },
    {
      "epoch": 120.37,
      "learning_rate": 3.950809061488674e-06,
      "loss": 0.012,
      "step": 309950
    },
    {
      "epoch": 120.37,
      "learning_rate": 3.9502912621359225e-06,
      "loss": 0.009,
      "step": 309960
    },
    {
      "epoch": 120.38,
      "learning_rate": 3.949773462783172e-06,
      "loss": 0.0599,
      "step": 309970
    },
    {
      "epoch": 120.38,
      "learning_rate": 3.949255663430421e-06,
      "loss": 0.0095,
      "step": 309980
    },
    {
      "epoch": 120.38,
      "learning_rate": 3.9487378640776705e-06,
      "loss": 0.0274,
      "step": 309990
    },
    {
      "epoch": 120.39,
      "learning_rate": 3.948220064724919e-06,
      "loss": 0.0268,
      "step": 310000
    },
    {
      "epoch": 120.39,
      "learning_rate": 3.947702265372169e-06,
      "loss": 0.0111,
      "step": 310010
    },
    {
      "epoch": 120.4,
      "learning_rate": 3.947184466019418e-06,
      "loss": 0.0018,
      "step": 310020
    },
    {
      "epoch": 120.4,
      "learning_rate": 3.946666666666667e-06,
      "loss": 0.0581,
      "step": 310030
    },
    {
      "epoch": 120.4,
      "learning_rate": 3.946148867313916e-06,
      "loss": 0.0004,
      "step": 310040
    },
    {
      "epoch": 120.41,
      "learning_rate": 3.945631067961166e-06,
      "loss": 0.0353,
      "step": 310050
    },
    {
      "epoch": 120.41,
      "learning_rate": 3.9451132686084144e-06,
      "loss": 0.0691,
      "step": 310060
    },
    {
      "epoch": 120.42,
      "learning_rate": 3.944595469255664e-06,
      "loss": 0.0812,
      "step": 310070
    },
    {
      "epoch": 120.42,
      "learning_rate": 3.944077669902913e-06,
      "loss": 0.0009,
      "step": 310080
    },
    {
      "epoch": 120.42,
      "learning_rate": 3.9435598705501624e-06,
      "loss": 0.029,
      "step": 310090
    },
    {
      "epoch": 120.43,
      "learning_rate": 3.943042071197411e-06,
      "loss": 0.0476,
      "step": 310100
    },
    {
      "epoch": 120.43,
      "learning_rate": 3.942524271844661e-06,
      "loss": 0.1185,
      "step": 310110
    },
    {
      "epoch": 120.43,
      "learning_rate": 3.94200647249191e-06,
      "loss": 0.039,
      "step": 310120
    },
    {
      "epoch": 120.44,
      "learning_rate": 3.941488673139159e-06,
      "loss": 0.0474,
      "step": 310130
    },
    {
      "epoch": 120.44,
      "learning_rate": 3.940970873786408e-06,
      "loss": 0.0021,
      "step": 310140
    },
    {
      "epoch": 120.45,
      "learning_rate": 3.940453074433658e-06,
      "loss": 0.0267,
      "step": 310150
    },
    {
      "epoch": 120.45,
      "learning_rate": 3.939935275080906e-06,
      "loss": 0.0298,
      "step": 310160
    },
    {
      "epoch": 120.45,
      "learning_rate": 3.939417475728156e-06,
      "loss": 0.0253,
      "step": 310170
    },
    {
      "epoch": 120.46,
      "learning_rate": 3.938899676375405e-06,
      "loss": 0.0215,
      "step": 310180
    },
    {
      "epoch": 120.46,
      "learning_rate": 3.9383818770226535e-06,
      "loss": 0.1199,
      "step": 310190
    },
    {
      "epoch": 120.47,
      "learning_rate": 3.937864077669903e-06,
      "loss": 0.01,
      "step": 310200
    },
    {
      "epoch": 120.47,
      "learning_rate": 3.937346278317152e-06,
      "loss": 0.0228,
      "step": 310210
    },
    {
      "epoch": 120.47,
      "learning_rate": 3.9368284789644016e-06,
      "loss": 0.0005,
      "step": 310220
    },
    {
      "epoch": 120.48,
      "learning_rate": 3.93631067961165e-06,
      "loss": 0.0408,
      "step": 310230
    },
    {
      "epoch": 120.48,
      "learning_rate": 3.9357928802589e-06,
      "loss": 0.0137,
      "step": 310240
    },
    {
      "epoch": 120.49,
      "learning_rate": 3.935275080906149e-06,
      "loss": 0.072,
      "step": 310250
    },
    {
      "epoch": 120.49,
      "learning_rate": 3.934757281553398e-06,
      "loss": 0.0013,
      "step": 310260
    },
    {
      "epoch": 120.49,
      "learning_rate": 3.934239482200647e-06,
      "loss": 0.0001,
      "step": 310270
    },
    {
      "epoch": 120.5,
      "learning_rate": 3.933721682847897e-06,
      "loss": 0.0776,
      "step": 310280
    },
    {
      "epoch": 120.5,
      "learning_rate": 3.9332038834951455e-06,
      "loss": 0.0891,
      "step": 310290
    },
    {
      "epoch": 120.5,
      "learning_rate": 3.932686084142395e-06,
      "loss": 0.1022,
      "step": 310300
    },
    {
      "epoch": 120.51,
      "learning_rate": 3.932168284789644e-06,
      "loss": 0.0229,
      "step": 310310
    },
    {
      "epoch": 120.51,
      "learning_rate": 3.9316504854368935e-06,
      "loss": 0.0968,
      "step": 310320
    },
    {
      "epoch": 120.52,
      "learning_rate": 3.931132686084142e-06,
      "loss": 0.0639,
      "step": 310330
    },
    {
      "epoch": 120.52,
      "learning_rate": 3.930614886731392e-06,
      "loss": 0.0247,
      "step": 310340
    },
    {
      "epoch": 120.52,
      "learning_rate": 3.930097087378641e-06,
      "loss": 0.0444,
      "step": 310350
    },
    {
      "epoch": 120.53,
      "learning_rate": 3.92957928802589e-06,
      "loss": 0.0191,
      "step": 310360
    },
    {
      "epoch": 120.53,
      "learning_rate": 3.929061488673139e-06,
      "loss": 0.0067,
      "step": 310370
    },
    {
      "epoch": 120.54,
      "learning_rate": 3.928543689320389e-06,
      "loss": 0.028,
      "step": 310380
    },
    {
      "epoch": 120.54,
      "learning_rate": 3.9280258899676374e-06,
      "loss": 0.0686,
      "step": 310390
    },
    {
      "epoch": 120.54,
      "learning_rate": 3.927508090614887e-06,
      "loss": 0.0686,
      "step": 310400
    },
    {
      "epoch": 120.55,
      "learning_rate": 3.926990291262136e-06,
      "loss": 0.0642,
      "step": 310410
    },
    {
      "epoch": 120.55,
      "learning_rate": 3.9264724919093854e-06,
      "loss": 0.0103,
      "step": 310420
    },
    {
      "epoch": 120.56,
      "learning_rate": 3.925954692556634e-06,
      "loss": 0.0868,
      "step": 310430
    },
    {
      "epoch": 120.56,
      "learning_rate": 3.925436893203884e-06,
      "loss": 0.013,
      "step": 310440
    },
    {
      "epoch": 120.56,
      "learning_rate": 3.924919093851133e-06,
      "loss": 0.001,
      "step": 310450
    },
    {
      "epoch": 120.57,
      "learning_rate": 3.924401294498382e-06,
      "loss": 0.0443,
      "step": 310460
    },
    {
      "epoch": 120.57,
      "learning_rate": 3.923883495145631e-06,
      "loss": 0.0304,
      "step": 310470
    },
    {
      "epoch": 120.57,
      "learning_rate": 3.923365695792881e-06,
      "loss": 0.1446,
      "step": 310480
    },
    {
      "epoch": 120.58,
      "learning_rate": 3.922847896440129e-06,
      "loss": 0.0019,
      "step": 310490
    },
    {
      "epoch": 120.58,
      "learning_rate": 3.922330097087379e-06,
      "loss": 0.0017,
      "step": 310500
    },
    {
      "epoch": 120.59,
      "learning_rate": 3.921812297734628e-06,
      "loss": 0.0251,
      "step": 310510
    },
    {
      "epoch": 120.59,
      "learning_rate": 3.921294498381877e-06,
      "loss": 0.0001,
      "step": 310520
    },
    {
      "epoch": 120.59,
      "learning_rate": 3.920776699029126e-06,
      "loss": 0.0092,
      "step": 310530
    },
    {
      "epoch": 120.6,
      "learning_rate": 3.920258899676376e-06,
      "loss": 0.0511,
      "step": 310540
    },
    {
      "epoch": 120.6,
      "learning_rate": 3.9197411003236245e-06,
      "loss": 0.0777,
      "step": 310550
    },
    {
      "epoch": 120.61,
      "learning_rate": 3.919223300970874e-06,
      "loss": 0.002,
      "step": 310560
    },
    {
      "epoch": 120.61,
      "learning_rate": 3.918705501618124e-06,
      "loss": 0.074,
      "step": 310570
    },
    {
      "epoch": 120.61,
      "learning_rate": 3.9181877022653726e-06,
      "loss": 0.0773,
      "step": 310580
    },
    {
      "epoch": 120.62,
      "learning_rate": 3.917669902912622e-06,
      "loss": 0.0945,
      "step": 310590
    },
    {
      "epoch": 120.62,
      "learning_rate": 3.917152103559871e-06,
      "loss": 0.0564,
      "step": 310600
    },
    {
      "epoch": 120.63,
      "learning_rate": 3.9166343042071206e-06,
      "loss": 0.1149,
      "step": 310610
    },
    {
      "epoch": 120.63,
      "learning_rate": 3.916116504854369e-06,
      "loss": 0.0801,
      "step": 310620
    },
    {
      "epoch": 120.63,
      "learning_rate": 3.915598705501619e-06,
      "loss": 0.0393,
      "step": 310630
    },
    {
      "epoch": 120.64,
      "learning_rate": 3.915080906148868e-06,
      "loss": 0.0008,
      "step": 310640
    },
    {
      "epoch": 120.64,
      "learning_rate": 3.914563106796117e-06,
      "loss": 0.0004,
      "step": 310650
    },
    {
      "epoch": 120.64,
      "learning_rate": 3.914045307443366e-06,
      "loss": 0.0655,
      "step": 310660
    },
    {
      "epoch": 120.65,
      "learning_rate": 3.913527508090616e-06,
      "loss": 0.1668,
      "step": 310670
    },
    {
      "epoch": 120.65,
      "learning_rate": 3.9130097087378645e-06,
      "loss": 0.071,
      "step": 310680
    },
    {
      "epoch": 120.66,
      "learning_rate": 3.912491909385114e-06,
      "loss": 0.0205,
      "step": 310690
    },
    {
      "epoch": 120.66,
      "learning_rate": 3.911974110032363e-06,
      "loss": 0.0004,
      "step": 310700
    },
    {
      "epoch": 120.66,
      "learning_rate": 3.9114563106796125e-06,
      "loss": 0.0244,
      "step": 310710
    },
    {
      "epoch": 120.67,
      "learning_rate": 3.910938511326861e-06,
      "loss": 0.082,
      "step": 310720
    },
    {
      "epoch": 120.67,
      "learning_rate": 3.910420711974111e-06,
      "loss": 0.0556,
      "step": 310730
    },
    {
      "epoch": 120.68,
      "learning_rate": 3.90990291262136e-06,
      "loss": 0.0794,
      "step": 310740
    },
    {
      "epoch": 120.68,
      "learning_rate": 3.909385113268609e-06,
      "loss": 0.1021,
      "step": 310750
    },
    {
      "epoch": 120.68,
      "learning_rate": 3.908867313915858e-06,
      "loss": 0.0287,
      "step": 310760
    },
    {
      "epoch": 120.69,
      "learning_rate": 3.908349514563108e-06,
      "loss": 0.0104,
      "step": 310770
    },
    {
      "epoch": 120.69,
      "learning_rate": 3.9078317152103564e-06,
      "loss": 0.0119,
      "step": 310780
    },
    {
      "epoch": 120.7,
      "learning_rate": 3.907313915857606e-06,
      "loss": 0.0432,
      "step": 310790
    },
    {
      "epoch": 120.7,
      "learning_rate": 3.906796116504855e-06,
      "loss": 0.0329,
      "step": 310800
    },
    {
      "epoch": 120.7,
      "learning_rate": 3.9062783171521045e-06,
      "loss": 0.0483,
      "step": 310810
    },
    {
      "epoch": 120.71,
      "learning_rate": 3.905760517799353e-06,
      "loss": 0.0799,
      "step": 310820
    },
    {
      "epoch": 120.71,
      "learning_rate": 3.905242718446602e-06,
      "loss": 0.0445,
      "step": 310830
    },
    {
      "epoch": 120.71,
      "learning_rate": 3.904724919093852e-06,
      "loss": 0.0173,
      "step": 310840
    },
    {
      "epoch": 120.72,
      "learning_rate": 3.9042071197411e-06,
      "loss": 0.0164,
      "step": 310850
    },
    {
      "epoch": 120.72,
      "learning_rate": 3.90368932038835e-06,
      "loss": 0.0321,
      "step": 310860
    },
    {
      "epoch": 120.73,
      "learning_rate": 3.903171521035599e-06,
      "loss": 0.1408,
      "step": 310870
    },
    {
      "epoch": 120.73,
      "learning_rate": 3.902653721682848e-06,
      "loss": 0.0638,
      "step": 310880
    },
    {
      "epoch": 120.73,
      "learning_rate": 3.902135922330097e-06,
      "loss": 0.0544,
      "step": 310890
    },
    {
      "epoch": 120.74,
      "learning_rate": 3.901618122977347e-06,
      "loss": 0.0547,
      "step": 310900
    },
    {
      "epoch": 120.74,
      "learning_rate": 3.9011003236245955e-06,
      "loss": 0.0487,
      "step": 310910
    },
    {
      "epoch": 120.75,
      "learning_rate": 3.900582524271845e-06,
      "loss": 0.0001,
      "step": 310920
    },
    {
      "epoch": 120.75,
      "learning_rate": 3.900064724919094e-06,
      "loss": 0.0019,
      "step": 310930
    },
    {
      "epoch": 120.75,
      "learning_rate": 3.8995469255663436e-06,
      "loss": 0.0482,
      "step": 310940
    },
    {
      "epoch": 120.76,
      "learning_rate": 3.899029126213592e-06,
      "loss": 0.0643,
      "step": 310950
    },
    {
      "epoch": 120.76,
      "learning_rate": 3.898511326860842e-06,
      "loss": 0.0272,
      "step": 310960
    },
    {
      "epoch": 120.77,
      "learning_rate": 3.897993527508091e-06,
      "loss": 0.0316,
      "step": 310970
    },
    {
      "epoch": 120.77,
      "learning_rate": 3.89747572815534e-06,
      "loss": 0.0001,
      "step": 310980
    },
    {
      "epoch": 120.77,
      "learning_rate": 3.896957928802589e-06,
      "loss": 0.0393,
      "step": 310990
    },
    {
      "epoch": 120.78,
      "learning_rate": 3.896440129449838e-06,
      "loss": 0.0419,
      "step": 311000
    },
    {
      "epoch": 120.78,
      "learning_rate": 3.8959223300970875e-06,
      "loss": 0.0058,
      "step": 311010
    },
    {
      "epoch": 120.78,
      "learning_rate": 3.895404530744336e-06,
      "loss": 0.0799,
      "step": 311020
    },
    {
      "epoch": 120.79,
      "learning_rate": 3.894886731391586e-06,
      "loss": 0.0186,
      "step": 311030
    },
    {
      "epoch": 120.79,
      "learning_rate": 3.894368932038835e-06,
      "loss": 0.0195,
      "step": 311040
    },
    {
      "epoch": 120.8,
      "learning_rate": 3.893851132686084e-06,
      "loss": 0.0215,
      "step": 311050
    },
    {
      "epoch": 120.8,
      "learning_rate": 3.893333333333333e-06,
      "loss": 0.0001,
      "step": 311060
    },
    {
      "epoch": 120.8,
      "learning_rate": 3.892815533980583e-06,
      "loss": 0.0381,
      "step": 311070
    },
    {
      "epoch": 120.81,
      "learning_rate": 3.8922977346278314e-06,
      "loss": 0.0172,
      "step": 311080
    },
    {
      "epoch": 120.81,
      "learning_rate": 3.891779935275081e-06,
      "loss": 0.0374,
      "step": 311090
    },
    {
      "epoch": 120.82,
      "learning_rate": 3.89126213592233e-06,
      "loss": 0.0001,
      "step": 311100
    },
    {
      "epoch": 120.82,
      "learning_rate": 3.8907443365695794e-06,
      "loss": 0.0239,
      "step": 311110
    },
    {
      "epoch": 120.82,
      "learning_rate": 3.890226537216828e-06,
      "loss": 0.045,
      "step": 311120
    },
    {
      "epoch": 120.83,
      "learning_rate": 3.889708737864078e-06,
      "loss": 0.0087,
      "step": 311130
    },
    {
      "epoch": 120.83,
      "learning_rate": 3.8891909385113274e-06,
      "loss": 0.0076,
      "step": 311140
    },
    {
      "epoch": 120.83,
      "learning_rate": 3.888673139158576e-06,
      "loss": 0.0535,
      "step": 311150
    },
    {
      "epoch": 120.84,
      "learning_rate": 3.888155339805826e-06,
      "loss": 0.0083,
      "step": 311160
    },
    {
      "epoch": 120.84,
      "learning_rate": 3.887637540453075e-06,
      "loss": 0.0842,
      "step": 311170
    },
    {
      "epoch": 120.85,
      "learning_rate": 3.887119741100324e-06,
      "loss": 0.0441,
      "step": 311180
    },
    {
      "epoch": 120.85,
      "learning_rate": 3.886601941747573e-06,
      "loss": 0.0016,
      "step": 311190
    },
    {
      "epoch": 120.85,
      "learning_rate": 3.886084142394823e-06,
      "loss": 0.0099,
      "step": 311200
    },
    {
      "epoch": 120.86,
      "learning_rate": 3.885566343042071e-06,
      "loss": 0.0285,
      "step": 311210
    },
    {
      "epoch": 120.86,
      "learning_rate": 3.885048543689321e-06,
      "loss": 0.0004,
      "step": 311220
    },
    {
      "epoch": 120.87,
      "learning_rate": 3.88453074433657e-06,
      "loss": 0.0173,
      "step": 311230
    },
    {
      "epoch": 120.87,
      "learning_rate": 3.884012944983819e-06,
      "loss": 0.0033,
      "step": 311240
    },
    {
      "epoch": 120.87,
      "learning_rate": 3.883495145631068e-06,
      "loss": 0.048,
      "step": 311250
    },
    {
      "epoch": 120.88,
      "learning_rate": 3.882977346278318e-06,
      "loss": 0.0602,
      "step": 311260
    },
    {
      "epoch": 120.88,
      "learning_rate": 3.8824595469255666e-06,
      "loss": 0.1591,
      "step": 311270
    },
    {
      "epoch": 120.89,
      "learning_rate": 3.881941747572816e-06,
      "loss": 0.006,
      "step": 311280
    },
    {
      "epoch": 120.89,
      "learning_rate": 3.881423948220065e-06,
      "loss": 0.0416,
      "step": 311290
    },
    {
      "epoch": 120.89,
      "learning_rate": 3.8809061488673146e-06,
      "loss": 0.0435,
      "step": 311300
    },
    {
      "epoch": 120.9,
      "learning_rate": 3.880388349514563e-06,
      "loss": 0.0243,
      "step": 311310
    },
    {
      "epoch": 120.9,
      "learning_rate": 3.879870550161813e-06,
      "loss": 0.0627,
      "step": 311320
    },
    {
      "epoch": 120.9,
      "learning_rate": 3.879352750809062e-06,
      "loss": 0.0698,
      "step": 311330
    },
    {
      "epoch": 120.91,
      "learning_rate": 3.878834951456311e-06,
      "loss": 0.0004,
      "step": 311340
    },
    {
      "epoch": 120.91,
      "learning_rate": 3.87831715210356e-06,
      "loss": 0.1194,
      "step": 311350
    },
    {
      "epoch": 120.92,
      "learning_rate": 3.87779935275081e-06,
      "loss": 0.0171,
      "step": 311360
    },
    {
      "epoch": 120.92,
      "learning_rate": 3.8772815533980585e-06,
      "loss": 0.1213,
      "step": 311370
    },
    {
      "epoch": 120.92,
      "learning_rate": 3.876763754045308e-06,
      "loss": 0.1258,
      "step": 311380
    },
    {
      "epoch": 120.93,
      "learning_rate": 3.876245954692557e-06,
      "loss": 0.0001,
      "step": 311390
    },
    {
      "epoch": 120.93,
      "learning_rate": 3.8757281553398065e-06,
      "loss": 0.026,
      "step": 311400
    },
    {
      "epoch": 120.94,
      "learning_rate": 3.875210355987055e-06,
      "loss": 0.0632,
      "step": 311410
    },
    {
      "epoch": 120.94,
      "learning_rate": 3.874692556634305e-06,
      "loss": 0.0089,
      "step": 311420
    },
    {
      "epoch": 120.94,
      "learning_rate": 3.874174757281554e-06,
      "loss": 0.0113,
      "step": 311430
    },
    {
      "epoch": 120.95,
      "learning_rate": 3.873656957928803e-06,
      "loss": 0.0029,
      "step": 311440
    },
    {
      "epoch": 120.95,
      "learning_rate": 3.873139158576052e-06,
      "loss": 0.0659,
      "step": 311450
    },
    {
      "epoch": 120.96,
      "learning_rate": 3.872621359223302e-06,
      "loss": 0.0548,
      "step": 311460
    },
    {
      "epoch": 120.96,
      "learning_rate": 3.8721035598705504e-06,
      "loss": 0.0002,
      "step": 311470
    },
    {
      "epoch": 120.96,
      "learning_rate": 3.8715857605178e-06,
      "loss": 0.1502,
      "step": 311480
    },
    {
      "epoch": 120.97,
      "learning_rate": 3.871067961165049e-06,
      "loss": 0.0594,
      "step": 311490
    },
    {
      "epoch": 120.97,
      "learning_rate": 3.8705501618122985e-06,
      "loss": 0.0285,
      "step": 311500
    },
    {
      "epoch": 120.97,
      "learning_rate": 3.870032362459547e-06,
      "loss": 0.2019,
      "step": 311510
    },
    {
      "epoch": 120.98,
      "learning_rate": 3.869514563106797e-06,
      "loss": 0.0001,
      "step": 311520
    },
    {
      "epoch": 120.98,
      "learning_rate": 3.868996763754046e-06,
      "loss": 0.0158,
      "step": 311530
    },
    {
      "epoch": 120.99,
      "learning_rate": 3.868478964401295e-06,
      "loss": 0.0193,
      "step": 311540
    },
    {
      "epoch": 120.99,
      "learning_rate": 3.867961165048544e-06,
      "loss": 0.0493,
      "step": 311550
    },
    {
      "epoch": 120.99,
      "learning_rate": 3.867443365695794e-06,
      "loss": 0.0444,
      "step": 311560
    },
    {
      "epoch": 121.0,
      "learning_rate": 3.866925566343042e-06,
      "loss": 0.0748,
      "step": 311570
    },
    {
      "epoch": 121.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.3838876187801361,
      "eval_runtime": 8.201,
      "eval_samples_per_second": 443.239,
      "eval_steps_per_second": 55.481,
      "step": 311575
    },
    {
      "epoch": 121.0,
      "learning_rate": 3.866407766990292e-06,
      "loss": 0.0002,
      "step": 311580
    },
    {
      "epoch": 121.01,
      "learning_rate": 3.865889967637541e-06,
      "loss": 0.0142,
      "step": 311590
    },
    {
      "epoch": 121.01,
      "learning_rate": 3.86537216828479e-06,
      "loss": 0.0024,
      "step": 311600
    },
    {
      "epoch": 121.01,
      "learning_rate": 3.864854368932039e-06,
      "loss": 0.084,
      "step": 311610
    },
    {
      "epoch": 121.02,
      "learning_rate": 3.864336569579289e-06,
      "loss": 0.0558,
      "step": 311620
    },
    {
      "epoch": 121.02,
      "learning_rate": 3.8638187702265376e-06,
      "loss": 0.0234,
      "step": 311630
    },
    {
      "epoch": 121.03,
      "learning_rate": 3.863300970873786e-06,
      "loss": 0.0405,
      "step": 311640
    },
    {
      "epoch": 121.03,
      "learning_rate": 3.862783171521036e-06,
      "loss": 0.0406,
      "step": 311650
    },
    {
      "epoch": 121.03,
      "learning_rate": 3.862265372168285e-06,
      "loss": 0.0001,
      "step": 311660
    },
    {
      "epoch": 121.04,
      "learning_rate": 3.861747572815534e-06,
      "loss": 0.0489,
      "step": 311670
    },
    {
      "epoch": 121.04,
      "learning_rate": 3.861229773462783e-06,
      "loss": 0.1208,
      "step": 311680
    },
    {
      "epoch": 121.04,
      "learning_rate": 3.860711974110033e-06,
      "loss": 0.0325,
      "step": 311690
    },
    {
      "epoch": 121.05,
      "learning_rate": 3.8601941747572815e-06,
      "loss": 0.1169,
      "step": 311700
    },
    {
      "epoch": 121.05,
      "learning_rate": 3.859676375404531e-06,
      "loss": 0.1447,
      "step": 311710
    },
    {
      "epoch": 121.06,
      "learning_rate": 3.85915857605178e-06,
      "loss": 0.0554,
      "step": 311720
    },
    {
      "epoch": 121.06,
      "learning_rate": 3.8586407766990295e-06,
      "loss": 0.0404,
      "step": 311730
    },
    {
      "epoch": 121.06,
      "learning_rate": 3.858122977346278e-06,
      "loss": 0.0324,
      "step": 311740
    },
    {
      "epoch": 121.07,
      "learning_rate": 3.857605177993528e-06,
      "loss": 0.0277,
      "step": 311750
    },
    {
      "epoch": 121.07,
      "learning_rate": 3.857087378640777e-06,
      "loss": 0.0289,
      "step": 311760
    },
    {
      "epoch": 121.08,
      "learning_rate": 3.856569579288026e-06,
      "loss": 0.0003,
      "step": 311770
    },
    {
      "epoch": 121.08,
      "learning_rate": 3.856051779935275e-06,
      "loss": 0.1404,
      "step": 311780
    },
    {
      "epoch": 121.08,
      "learning_rate": 3.855533980582525e-06,
      "loss": 0.0022,
      "step": 311790
    },
    {
      "epoch": 121.09,
      "learning_rate": 3.8550161812297734e-06,
      "loss": 0.0122,
      "step": 311800
    },
    {
      "epoch": 121.09,
      "learning_rate": 3.854498381877023e-06,
      "loss": 0.0081,
      "step": 311810
    },
    {
      "epoch": 121.1,
      "learning_rate": 3.853980582524272e-06,
      "loss": 0.0323,
      "step": 311820
    },
    {
      "epoch": 121.1,
      "learning_rate": 3.8534627831715214e-06,
      "loss": 0.0974,
      "step": 311830
    },
    {
      "epoch": 121.1,
      "learning_rate": 3.85294498381877e-06,
      "loss": 0.0529,
      "step": 311840
    },
    {
      "epoch": 121.11,
      "learning_rate": 3.85242718446602e-06,
      "loss": 0.0008,
      "step": 311850
    },
    {
      "epoch": 121.11,
      "learning_rate": 3.851909385113269e-06,
      "loss": 0.0886,
      "step": 311860
    },
    {
      "epoch": 121.11,
      "learning_rate": 3.851391585760518e-06,
      "loss": 0.1124,
      "step": 311870
    },
    {
      "epoch": 121.12,
      "learning_rate": 3.850873786407767e-06,
      "loss": 0.064,
      "step": 311880
    },
    {
      "epoch": 121.12,
      "learning_rate": 3.850355987055017e-06,
      "loss": 0.0087,
      "step": 311890
    },
    {
      "epoch": 121.13,
      "learning_rate": 3.849838187702265e-06,
      "loss": 0.0081,
      "step": 311900
    },
    {
      "epoch": 121.13,
      "learning_rate": 3.849320388349515e-06,
      "loss": 0.0128,
      "step": 311910
    },
    {
      "epoch": 121.13,
      "learning_rate": 3.848802588996764e-06,
      "loss": 0.108,
      "step": 311920
    },
    {
      "epoch": 121.14,
      "learning_rate": 3.848284789644013e-06,
      "loss": 0.0954,
      "step": 311930
    },
    {
      "epoch": 121.14,
      "learning_rate": 3.847766990291262e-06,
      "loss": 0.0102,
      "step": 311940
    },
    {
      "epoch": 121.15,
      "learning_rate": 3.847249190938512e-06,
      "loss": 0.1279,
      "step": 311950
    },
    {
      "epoch": 121.15,
      "learning_rate": 3.8467313915857606e-06,
      "loss": 0.0006,
      "step": 311960
    },
    {
      "epoch": 121.15,
      "learning_rate": 3.84621359223301e-06,
      "loss": 0.0137,
      "step": 311970
    },
    {
      "epoch": 121.16,
      "learning_rate": 3.845695792880259e-06,
      "loss": 0.0297,
      "step": 311980
    },
    {
      "epoch": 121.16,
      "learning_rate": 3.8451779935275086e-06,
      "loss": 0.0076,
      "step": 311990
    },
    {
      "epoch": 121.17,
      "learning_rate": 3.844660194174757e-06,
      "loss": 0.0064,
      "step": 312000
    },
    {
      "epoch": 121.17,
      "learning_rate": 3.844142394822007e-06,
      "loss": 0.0001,
      "step": 312010
    },
    {
      "epoch": 121.17,
      "learning_rate": 3.843624595469256e-06,
      "loss": 0.0403,
      "step": 312020
    },
    {
      "epoch": 121.18,
      "learning_rate": 3.843106796116505e-06,
      "loss": 0.0281,
      "step": 312030
    },
    {
      "epoch": 121.18,
      "learning_rate": 3.842588996763754e-06,
      "loss": 0.1573,
      "step": 312040
    },
    {
      "epoch": 121.18,
      "learning_rate": 3.842071197411004e-06,
      "loss": 0.0551,
      "step": 312050
    },
    {
      "epoch": 121.19,
      "learning_rate": 3.8415533980582525e-06,
      "loss": 0.0013,
      "step": 312060
    },
    {
      "epoch": 121.19,
      "learning_rate": 3.841035598705502e-06,
      "loss": 0.0531,
      "step": 312070
    },
    {
      "epoch": 121.2,
      "learning_rate": 3.840517799352751e-06,
      "loss": 0.0013,
      "step": 312080
    },
    {
      "epoch": 121.2,
      "learning_rate": 3.8400000000000005e-06,
      "loss": 0.0788,
      "step": 312090
    },
    {
      "epoch": 121.2,
      "learning_rate": 3.839482200647249e-06,
      "loss": 0.0213,
      "step": 312100
    },
    {
      "epoch": 121.21,
      "learning_rate": 3.838964401294499e-06,
      "loss": 0.0836,
      "step": 312110
    },
    {
      "epoch": 121.21,
      "learning_rate": 3.838446601941748e-06,
      "loss": 0.0077,
      "step": 312120
    },
    {
      "epoch": 121.22,
      "learning_rate": 3.837928802588997e-06,
      "loss": 0.0338,
      "step": 312130
    },
    {
      "epoch": 121.22,
      "learning_rate": 3.837411003236246e-06,
      "loss": 0.0035,
      "step": 312140
    },
    {
      "epoch": 121.22,
      "learning_rate": 3.836893203883496e-06,
      "loss": 0.0113,
      "step": 312150
    },
    {
      "epoch": 121.23,
      "learning_rate": 3.8363754045307444e-06,
      "loss": 0.1074,
      "step": 312160
    },
    {
      "epoch": 121.23,
      "learning_rate": 3.835857605177994e-06,
      "loss": 0.0491,
      "step": 312170
    },
    {
      "epoch": 121.23,
      "learning_rate": 3.835339805825243e-06,
      "loss": 0.0265,
      "step": 312180
    },
    {
      "epoch": 121.24,
      "learning_rate": 3.8348220064724924e-06,
      "loss": 0.0002,
      "step": 312190
    },
    {
      "epoch": 121.24,
      "learning_rate": 3.834304207119741e-06,
      "loss": 0.0009,
      "step": 312200
    },
    {
      "epoch": 121.25,
      "learning_rate": 3.833786407766991e-06,
      "loss": 0.0301,
      "step": 312210
    },
    {
      "epoch": 121.25,
      "learning_rate": 3.83326860841424e-06,
      "loss": 0.0196,
      "step": 312220
    },
    {
      "epoch": 121.25,
      "learning_rate": 3.832750809061489e-06,
      "loss": 0.0582,
      "step": 312230
    },
    {
      "epoch": 121.26,
      "learning_rate": 3.832233009708738e-06,
      "loss": 0.0118,
      "step": 312240
    },
    {
      "epoch": 121.26,
      "learning_rate": 3.831715210355988e-06,
      "loss": 0.1244,
      "step": 312250
    },
    {
      "epoch": 121.27,
      "learning_rate": 3.831197411003237e-06,
      "loss": 0.0004,
      "step": 312260
    },
    {
      "epoch": 121.27,
      "learning_rate": 3.830679611650486e-06,
      "loss": 0.0004,
      "step": 312270
    },
    {
      "epoch": 121.27,
      "learning_rate": 3.830161812297736e-06,
      "loss": 0.0595,
      "step": 312280
    },
    {
      "epoch": 121.28,
      "learning_rate": 3.829644012944984e-06,
      "loss": 0.1599,
      "step": 312290
    },
    {
      "epoch": 121.28,
      "learning_rate": 3.829126213592233e-06,
      "loss": 0.1076,
      "step": 312300
    },
    {
      "epoch": 121.29,
      "learning_rate": 3.828608414239483e-06,
      "loss": 0.0001,
      "step": 312310
    },
    {
      "epoch": 121.29,
      "learning_rate": 3.8280906148867316e-06,
      "loss": 0.0511,
      "step": 312320
    },
    {
      "epoch": 121.29,
      "learning_rate": 3.827572815533981e-06,
      "loss": 0.0387,
      "step": 312330
    },
    {
      "epoch": 121.3,
      "learning_rate": 3.82705501618123e-06,
      "loss": 0.0094,
      "step": 312340
    },
    {
      "epoch": 121.3,
      "learning_rate": 3.8265372168284796e-06,
      "loss": 0.0102,
      "step": 312350
    },
    {
      "epoch": 121.3,
      "learning_rate": 3.826019417475728e-06,
      "loss": 0.0162,
      "step": 312360
    },
    {
      "epoch": 121.31,
      "learning_rate": 3.825501618122978e-06,
      "loss": 0.0107,
      "step": 312370
    },
    {
      "epoch": 121.31,
      "learning_rate": 3.824983818770227e-06,
      "loss": 0.0249,
      "step": 312380
    },
    {
      "epoch": 121.32,
      "learning_rate": 3.824466019417476e-06,
      "loss": 0.0004,
      "step": 312390
    },
    {
      "epoch": 121.32,
      "learning_rate": 3.823948220064725e-06,
      "loss": 0.0006,
      "step": 312400
    },
    {
      "epoch": 121.32,
      "learning_rate": 3.823430420711975e-06,
      "loss": 0.0551,
      "step": 312410
    },
    {
      "epoch": 121.33,
      "learning_rate": 3.8229126213592235e-06,
      "loss": 0.0199,
      "step": 312420
    },
    {
      "epoch": 121.33,
      "learning_rate": 3.822394822006473e-06,
      "loss": 0.0777,
      "step": 312430
    },
    {
      "epoch": 121.34,
      "learning_rate": 3.821877022653722e-06,
      "loss": 0.1624,
      "step": 312440
    },
    {
      "epoch": 121.34,
      "learning_rate": 3.8213592233009715e-06,
      "loss": 0.1108,
      "step": 312450
    },
    {
      "epoch": 121.34,
      "learning_rate": 3.82084142394822e-06,
      "loss": 0.0894,
      "step": 312460
    },
    {
      "epoch": 121.35,
      "learning_rate": 3.820323624595469e-06,
      "loss": 0.0859,
      "step": 312470
    },
    {
      "epoch": 121.35,
      "learning_rate": 3.819805825242719e-06,
      "loss": 0.0796,
      "step": 312480
    },
    {
      "epoch": 121.36,
      "learning_rate": 3.8192880258899674e-06,
      "loss": 0.074,
      "step": 312490
    },
    {
      "epoch": 121.36,
      "learning_rate": 3.818770226537217e-06,
      "loss": 0.0426,
      "step": 312500
    },
    {
      "epoch": 121.36,
      "learning_rate": 3.818252427184466e-06,
      "loss": 0.001,
      "step": 312510
    },
    {
      "epoch": 121.37,
      "learning_rate": 3.8177346278317154e-06,
      "loss": 0.0255,
      "step": 312520
    },
    {
      "epoch": 121.37,
      "learning_rate": 3.817216828478964e-06,
      "loss": 0.0355,
      "step": 312530
    },
    {
      "epoch": 121.37,
      "learning_rate": 3.816699029126214e-06,
      "loss": 0.0838,
      "step": 312540
    },
    {
      "epoch": 121.38,
      "learning_rate": 3.816181229773463e-06,
      "loss": 0.016,
      "step": 312550
    },
    {
      "epoch": 121.38,
      "learning_rate": 3.815663430420712e-06,
      "loss": 0.019,
      "step": 312560
    },
    {
      "epoch": 121.39,
      "learning_rate": 3.815145631067961e-06,
      "loss": 0.0002,
      "step": 312570
    },
    {
      "epoch": 121.39,
      "learning_rate": 3.8146278317152106e-06,
      "loss": 0.0179,
      "step": 312580
    },
    {
      "epoch": 121.39,
      "learning_rate": 3.81411003236246e-06,
      "loss": 0.0198,
      "step": 312590
    },
    {
      "epoch": 121.4,
      "learning_rate": 3.813592233009709e-06,
      "loss": 0.1035,
      "step": 312600
    },
    {
      "epoch": 121.4,
      "learning_rate": 3.813074433656958e-06,
      "loss": 0.0647,
      "step": 312610
    },
    {
      "epoch": 121.41,
      "learning_rate": 3.8125566343042074e-06,
      "loss": 0.1397,
      "step": 312620
    },
    {
      "epoch": 121.41,
      "learning_rate": 3.8120388349514566e-06,
      "loss": 0.017,
      "step": 312630
    },
    {
      "epoch": 121.41,
      "learning_rate": 3.8115210355987058e-06,
      "loss": 0.042,
      "step": 312640
    },
    {
      "epoch": 121.42,
      "learning_rate": 3.811003236245955e-06,
      "loss": 0.0701,
      "step": 312650
    },
    {
      "epoch": 121.42,
      "learning_rate": 3.810485436893204e-06,
      "loss": 0.001,
      "step": 312660
    },
    {
      "epoch": 121.43,
      "learning_rate": 3.8099676375404534e-06,
      "loss": 0.1091,
      "step": 312670
    },
    {
      "epoch": 121.43,
      "learning_rate": 3.8094498381877026e-06,
      "loss": 0.0495,
      "step": 312680
    },
    {
      "epoch": 121.43,
      "learning_rate": 3.8089320388349518e-06,
      "loss": 0.0004,
      "step": 312690
    },
    {
      "epoch": 121.44,
      "learning_rate": 3.808414239482201e-06,
      "loss": 0.0744,
      "step": 312700
    },
    {
      "epoch": 121.44,
      "learning_rate": 3.8078964401294497e-06,
      "loss": 0.1477,
      "step": 312710
    },
    {
      "epoch": 121.44,
      "learning_rate": 3.8073786407766993e-06,
      "loss": 0.0485,
      "step": 312720
    },
    {
      "epoch": 121.45,
      "learning_rate": 3.806860841423948e-06,
      "loss": 0.046,
      "step": 312730
    },
    {
      "epoch": 121.45,
      "learning_rate": 3.8063430420711977e-06,
      "loss": 0.0363,
      "step": 312740
    },
    {
      "epoch": 121.46,
      "learning_rate": 3.8058252427184465e-06,
      "loss": 0.0413,
      "step": 312750
    },
    {
      "epoch": 121.46,
      "learning_rate": 3.805307443365696e-06,
      "loss": 0.0097,
      "step": 312760
    },
    {
      "epoch": 121.46,
      "learning_rate": 3.804789644012945e-06,
      "loss": 0.0837,
      "step": 312770
    },
    {
      "epoch": 121.47,
      "learning_rate": 3.8042718446601945e-06,
      "loss": 0.0011,
      "step": 312780
    },
    {
      "epoch": 121.47,
      "learning_rate": 3.8037540453074433e-06,
      "loss": 0.0081,
      "step": 312790
    },
    {
      "epoch": 121.48,
      "learning_rate": 3.803236245954693e-06,
      "loss": 0.0076,
      "step": 312800
    },
    {
      "epoch": 121.48,
      "learning_rate": 3.8027184466019417e-06,
      "loss": 0.0612,
      "step": 312810
    },
    {
      "epoch": 121.48,
      "learning_rate": 3.8022006472491913e-06,
      "loss": 0.0433,
      "step": 312820
    },
    {
      "epoch": 121.49,
      "learning_rate": 3.801682847896441e-06,
      "loss": 0.0575,
      "step": 312830
    },
    {
      "epoch": 121.49,
      "learning_rate": 3.8011650485436897e-06,
      "loss": 0.0051,
      "step": 312840
    },
    {
      "epoch": 121.5,
      "learning_rate": 3.8006472491909393e-06,
      "loss": 0.011,
      "step": 312850
    },
    {
      "epoch": 121.5,
      "learning_rate": 3.800129449838188e-06,
      "loss": 0.0315,
      "step": 312860
    },
    {
      "epoch": 121.5,
      "learning_rate": 3.7996116504854373e-06,
      "loss": 0.0004,
      "step": 312870
    },
    {
      "epoch": 121.51,
      "learning_rate": 3.7990938511326864e-06,
      "loss": 0.1247,
      "step": 312880
    },
    {
      "epoch": 121.51,
      "learning_rate": 3.7985760517799356e-06,
      "loss": 0.0018,
      "step": 312890
    },
    {
      "epoch": 121.51,
      "learning_rate": 3.798058252427185e-06,
      "loss": 0.0474,
      "step": 312900
    },
    {
      "epoch": 121.52,
      "learning_rate": 3.797540453074434e-06,
      "loss": 0.0657,
      "step": 312910
    },
    {
      "epoch": 121.52,
      "learning_rate": 3.7970226537216832e-06,
      "loss": 0.0042,
      "step": 312920
    },
    {
      "epoch": 121.53,
      "learning_rate": 3.7965048543689324e-06,
      "loss": 0.0664,
      "step": 312930
    },
    {
      "epoch": 121.53,
      "learning_rate": 3.7959870550161816e-06,
      "loss": 0.054,
      "step": 312940
    },
    {
      "epoch": 121.53,
      "learning_rate": 3.795469255663431e-06,
      "loss": 0.0787,
      "step": 312950
    },
    {
      "epoch": 121.54,
      "learning_rate": 3.79495145631068e-06,
      "loss": 0.1083,
      "step": 312960
    },
    {
      "epoch": 121.54,
      "learning_rate": 3.794433656957929e-06,
      "loss": 0.0003,
      "step": 312970
    },
    {
      "epoch": 121.55,
      "learning_rate": 3.7939158576051784e-06,
      "loss": 0.0401,
      "step": 312980
    },
    {
      "epoch": 121.55,
      "learning_rate": 3.7933980582524276e-06,
      "loss": 0.0585,
      "step": 312990
    },
    {
      "epoch": 121.55,
      "learning_rate": 3.7928802588996768e-06,
      "loss": 0.0298,
      "step": 313000
    },
    {
      "epoch": 121.56,
      "learning_rate": 3.792362459546926e-06,
      "loss": 0.0393,
      "step": 313010
    },
    {
      "epoch": 121.56,
      "learning_rate": 3.791844660194175e-06,
      "loss": 0.0814,
      "step": 313020
    },
    {
      "epoch": 121.57,
      "learning_rate": 3.7913268608414244e-06,
      "loss": 0.0156,
      "step": 313030
    },
    {
      "epoch": 121.57,
      "learning_rate": 3.790809061488673e-06,
      "loss": 0.0178,
      "step": 313040
    },
    {
      "epoch": 121.57,
      "learning_rate": 3.7902912621359228e-06,
      "loss": 0.0275,
      "step": 313050
    },
    {
      "epoch": 121.58,
      "learning_rate": 3.7897734627831715e-06,
      "loss": 0.0616,
      "step": 313060
    },
    {
      "epoch": 121.58,
      "learning_rate": 3.789255663430421e-06,
      "loss": 0.0121,
      "step": 313070
    },
    {
      "epoch": 121.58,
      "learning_rate": 3.78873786407767e-06,
      "loss": 0.0356,
      "step": 313080
    },
    {
      "epoch": 121.59,
      "learning_rate": 3.7882200647249195e-06,
      "loss": 0.0083,
      "step": 313090
    },
    {
      "epoch": 121.59,
      "learning_rate": 3.7877022653721683e-06,
      "loss": 0.0221,
      "step": 313100
    },
    {
      "epoch": 121.6,
      "learning_rate": 3.787184466019418e-06,
      "loss": 0.0156,
      "step": 313110
    },
    {
      "epoch": 121.6,
      "learning_rate": 3.7866666666666667e-06,
      "loss": 0.0021,
      "step": 313120
    },
    {
      "epoch": 121.6,
      "learning_rate": 3.7861488673139163e-06,
      "loss": 0.053,
      "step": 313130
    },
    {
      "epoch": 121.61,
      "learning_rate": 3.785631067961165e-06,
      "loss": 0.0529,
      "step": 313140
    },
    {
      "epoch": 121.61,
      "learning_rate": 3.7851132686084147e-06,
      "loss": 0.0696,
      "step": 313150
    },
    {
      "epoch": 121.62,
      "learning_rate": 3.7845954692556635e-06,
      "loss": 0.0089,
      "step": 313160
    },
    {
      "epoch": 121.62,
      "learning_rate": 3.784077669902913e-06,
      "loss": 0.0583,
      "step": 313170
    },
    {
      "epoch": 121.62,
      "learning_rate": 3.783559870550162e-06,
      "loss": 0.0364,
      "step": 313180
    },
    {
      "epoch": 121.63,
      "learning_rate": 3.7830420711974115e-06,
      "loss": 0.0001,
      "step": 313190
    },
    {
      "epoch": 121.63,
      "learning_rate": 3.7825242718446602e-06,
      "loss": 0.0387,
      "step": 313200
    },
    {
      "epoch": 121.63,
      "learning_rate": 3.78200647249191e-06,
      "loss": 0.0007,
      "step": 313210
    },
    {
      "epoch": 121.64,
      "learning_rate": 3.7814886731391586e-06,
      "loss": 0.0182,
      "step": 313220
    },
    {
      "epoch": 121.64,
      "learning_rate": 3.7809708737864083e-06,
      "loss": 0.0726,
      "step": 313230
    },
    {
      "epoch": 121.65,
      "learning_rate": 3.780453074433657e-06,
      "loss": 0.0391,
      "step": 313240
    },
    {
      "epoch": 121.65,
      "learning_rate": 3.7799352750809066e-06,
      "loss": 0.09,
      "step": 313250
    },
    {
      "epoch": 121.65,
      "learning_rate": 3.7794174757281554e-06,
      "loss": 0.015,
      "step": 313260
    },
    {
      "epoch": 121.66,
      "learning_rate": 3.778899676375405e-06,
      "loss": 0.0134,
      "step": 313270
    },
    {
      "epoch": 121.66,
      "learning_rate": 3.778381877022654e-06,
      "loss": 0.1042,
      "step": 313280
    },
    {
      "epoch": 121.67,
      "learning_rate": 3.7778640776699034e-06,
      "loss": 0.0907,
      "step": 313290
    },
    {
      "epoch": 121.67,
      "learning_rate": 3.777346278317152e-06,
      "loss": 0.0436,
      "step": 313300
    },
    {
      "epoch": 121.67,
      "learning_rate": 3.776828478964402e-06,
      "loss": 0.0378,
      "step": 313310
    },
    {
      "epoch": 121.68,
      "learning_rate": 3.7763106796116506e-06,
      "loss": 0.0005,
      "step": 313320
    },
    {
      "epoch": 121.68,
      "learning_rate": 3.7757928802589e-06,
      "loss": 0.0149,
      "step": 313330
    },
    {
      "epoch": 121.69,
      "learning_rate": 3.775275080906149e-06,
      "loss": 0.0002,
      "step": 313340
    },
    {
      "epoch": 121.69,
      "learning_rate": 3.774757281553398e-06,
      "loss": 0.047,
      "step": 313350
    },
    {
      "epoch": 121.69,
      "learning_rate": 3.7742394822006474e-06,
      "loss": 0.0093,
      "step": 313360
    },
    {
      "epoch": 121.7,
      "learning_rate": 3.7737216828478966e-06,
      "loss": 0.0391,
      "step": 313370
    },
    {
      "epoch": 121.7,
      "learning_rate": 3.773203883495146e-06,
      "loss": 0.0492,
      "step": 313380
    },
    {
      "epoch": 121.7,
      "learning_rate": 3.772686084142395e-06,
      "loss": 0.0921,
      "step": 313390
    },
    {
      "epoch": 121.71,
      "learning_rate": 3.7721682847896446e-06,
      "loss": 0.0002,
      "step": 313400
    },
    {
      "epoch": 121.71,
      "learning_rate": 3.7716504854368933e-06,
      "loss": 0.0374,
      "step": 313410
    },
    {
      "epoch": 121.72,
      "learning_rate": 3.771132686084143e-06,
      "loss": 0.0908,
      "step": 313420
    },
    {
      "epoch": 121.72,
      "learning_rate": 3.7706148867313917e-06,
      "loss": 0.0595,
      "step": 313430
    },
    {
      "epoch": 121.72,
      "learning_rate": 3.7700970873786413e-06,
      "loss": 0.0354,
      "step": 313440
    },
    {
      "epoch": 121.73,
      "learning_rate": 3.76957928802589e-06,
      "loss": 0.0136,
      "step": 313450
    },
    {
      "epoch": 121.73,
      "learning_rate": 3.7690614886731397e-06,
      "loss": 0.0003,
      "step": 313460
    },
    {
      "epoch": 121.74,
      "learning_rate": 3.7685436893203885e-06,
      "loss": 0.0075,
      "step": 313470
    },
    {
      "epoch": 121.74,
      "learning_rate": 3.768025889967638e-06,
      "loss": 0.0395,
      "step": 313480
    },
    {
      "epoch": 121.74,
      "learning_rate": 3.767508090614887e-06,
      "loss": 0.1485,
      "step": 313490
    },
    {
      "epoch": 121.75,
      "learning_rate": 3.7669902912621365e-06,
      "loss": 0.0746,
      "step": 313500
    },
    {
      "epoch": 121.75,
      "learning_rate": 3.7664724919093853e-06,
      "loss": 0.0644,
      "step": 313510
    },
    {
      "epoch": 121.76,
      "learning_rate": 3.765954692556635e-06,
      "loss": 0.0364,
      "step": 313520
    },
    {
      "epoch": 121.76,
      "learning_rate": 3.7654368932038837e-06,
      "loss": 0.0236,
      "step": 313530
    },
    {
      "epoch": 121.76,
      "learning_rate": 3.7649190938511333e-06,
      "loss": 0.1112,
      "step": 313540
    },
    {
      "epoch": 121.77,
      "learning_rate": 3.764401294498382e-06,
      "loss": 0.0011,
      "step": 313550
    },
    {
      "epoch": 121.77,
      "learning_rate": 3.7638834951456317e-06,
      "loss": 0.1617,
      "step": 313560
    },
    {
      "epoch": 121.77,
      "learning_rate": 3.7633656957928804e-06,
      "loss": 0.0152,
      "step": 313570
    },
    {
      "epoch": 121.78,
      "learning_rate": 3.76284789644013e-06,
      "loss": 0.1622,
      "step": 313580
    },
    {
      "epoch": 121.78,
      "learning_rate": 3.762330097087379e-06,
      "loss": 0.0268,
      "step": 313590
    },
    {
      "epoch": 121.79,
      "learning_rate": 3.7618122977346285e-06,
      "loss": 0.0093,
      "step": 313600
    },
    {
      "epoch": 121.79,
      "learning_rate": 3.7612944983818772e-06,
      "loss": 0.0243,
      "step": 313610
    },
    {
      "epoch": 121.79,
      "learning_rate": 3.760776699029127e-06,
      "loss": 0.0683,
      "step": 313620
    },
    {
      "epoch": 121.8,
      "learning_rate": 3.7602588996763756e-06,
      "loss": 0.0005,
      "step": 313630
    },
    {
      "epoch": 121.8,
      "learning_rate": 3.7597411003236252e-06,
      "loss": 0.0792,
      "step": 313640
    },
    {
      "epoch": 121.81,
      "learning_rate": 3.759223300970874e-06,
      "loss": 0.02,
      "step": 313650
    },
    {
      "epoch": 121.81,
      "learning_rate": 3.7587055016181236e-06,
      "loss": 0.0004,
      "step": 313660
    },
    {
      "epoch": 121.81,
      "learning_rate": 3.7581877022653724e-06,
      "loss": 0.0253,
      "step": 313670
    },
    {
      "epoch": 121.82,
      "learning_rate": 3.7576699029126216e-06,
      "loss": 0.0183,
      "step": 313680
    },
    {
      "epoch": 121.82,
      "learning_rate": 3.7571521035598708e-06,
      "loss": 0.0268,
      "step": 313690
    },
    {
      "epoch": 121.83,
      "learning_rate": 3.75663430420712e-06,
      "loss": 0.031,
      "step": 313700
    },
    {
      "epoch": 121.83,
      "learning_rate": 3.756116504854369e-06,
      "loss": 0.1488,
      "step": 313710
    },
    {
      "epoch": 121.83,
      "learning_rate": 3.7555987055016184e-06,
      "loss": 0.0647,
      "step": 313720
    },
    {
      "epoch": 121.84,
      "learning_rate": 3.7550809061488676e-06,
      "loss": 0.0426,
      "step": 313730
    },
    {
      "epoch": 121.84,
      "learning_rate": 3.7545631067961168e-06,
      "loss": 0.0298,
      "step": 313740
    },
    {
      "epoch": 121.84,
      "learning_rate": 3.754045307443366e-06,
      "loss": 0.0329,
      "step": 313750
    },
    {
      "epoch": 121.85,
      "learning_rate": 3.753527508090615e-06,
      "loss": 0.0573,
      "step": 313760
    },
    {
      "epoch": 121.85,
      "learning_rate": 3.7530097087378643e-06,
      "loss": 0.0605,
      "step": 313770
    },
    {
      "epoch": 121.86,
      "learning_rate": 3.7524919093851135e-06,
      "loss": 0.0073,
      "step": 313780
    },
    {
      "epoch": 121.86,
      "learning_rate": 3.7519741100323627e-06,
      "loss": 0.0178,
      "step": 313790
    },
    {
      "epoch": 121.86,
      "learning_rate": 3.751456310679612e-06,
      "loss": 0.0292,
      "step": 313800
    },
    {
      "epoch": 121.87,
      "learning_rate": 3.750938511326861e-06,
      "loss": 0.1087,
      "step": 313810
    },
    {
      "epoch": 121.87,
      "learning_rate": 3.7504207119741103e-06,
      "loss": 0.0143,
      "step": 313820
    },
    {
      "epoch": 121.88,
      "learning_rate": 3.7499029126213595e-06,
      "loss": 0.0114,
      "step": 313830
    },
    {
      "epoch": 121.88,
      "learning_rate": 3.7493851132686087e-06,
      "loss": 0.0307,
      "step": 313840
    },
    {
      "epoch": 121.88,
      "learning_rate": 3.7488673139158575e-06,
      "loss": 0.0015,
      "step": 313850
    },
    {
      "epoch": 121.89,
      "learning_rate": 3.748349514563107e-06,
      "loss": 0.0119,
      "step": 313860
    },
    {
      "epoch": 121.89,
      "learning_rate": 3.747831715210356e-06,
      "loss": 0.085,
      "step": 313870
    },
    {
      "epoch": 121.9,
      "learning_rate": 3.7473139158576055e-06,
      "loss": 0.044,
      "step": 313880
    },
    {
      "epoch": 121.9,
      "learning_rate": 3.7467961165048542e-06,
      "loss": 0.1774,
      "step": 313890
    },
    {
      "epoch": 121.9,
      "learning_rate": 3.746278317152104e-06,
      "loss": 0.0683,
      "step": 313900
    },
    {
      "epoch": 121.91,
      "learning_rate": 3.7457605177993526e-06,
      "loss": 0.0913,
      "step": 313910
    },
    {
      "epoch": 121.91,
      "learning_rate": 3.7452427184466023e-06,
      "loss": 0.0008,
      "step": 313920
    },
    {
      "epoch": 121.91,
      "learning_rate": 3.744724919093851e-06,
      "loss": 0.1803,
      "step": 313930
    },
    {
      "epoch": 121.92,
      "learning_rate": 3.7442071197411006e-06,
      "loss": 0.0058,
      "step": 313940
    },
    {
      "epoch": 121.92,
      "learning_rate": 3.7436893203883503e-06,
      "loss": 0.0633,
      "step": 313950
    },
    {
      "epoch": 121.93,
      "learning_rate": 3.743171521035599e-06,
      "loss": 0.049,
      "step": 313960
    },
    {
      "epoch": 121.93,
      "learning_rate": 3.7426537216828487e-06,
      "loss": 0.0003,
      "step": 313970
    },
    {
      "epoch": 121.93,
      "learning_rate": 3.7421359223300974e-06,
      "loss": 0.1282,
      "step": 313980
    },
    {
      "epoch": 121.94,
      "learning_rate": 3.741618122977347e-06,
      "loss": 0.0003,
      "step": 313990
    },
    {
      "epoch": 121.94,
      "learning_rate": 3.741100323624596e-06,
      "loss": 0.0357,
      "step": 314000
    },
    {
      "epoch": 121.95,
      "learning_rate": 3.740582524271845e-06,
      "loss": 0.042,
      "step": 314010
    },
    {
      "epoch": 121.95,
      "learning_rate": 3.740064724919094e-06,
      "loss": 0.1049,
      "step": 314020
    },
    {
      "epoch": 121.95,
      "learning_rate": 3.7395469255663434e-06,
      "loss": 0.0343,
      "step": 314030
    },
    {
      "epoch": 121.96,
      "learning_rate": 3.7390291262135926e-06,
      "loss": 0.0183,
      "step": 314040
    },
    {
      "epoch": 121.96,
      "learning_rate": 3.7385113268608418e-06,
      "loss": 0.0053,
      "step": 314050
    },
    {
      "epoch": 121.97,
      "learning_rate": 3.737993527508091e-06,
      "loss": 0.0527,
      "step": 314060
    },
    {
      "epoch": 121.97,
      "learning_rate": 3.73747572815534e-06,
      "loss": 0.0004,
      "step": 314070
    },
    {
      "epoch": 121.97,
      "learning_rate": 3.7369579288025894e-06,
      "loss": 0.0948,
      "step": 314080
    },
    {
      "epoch": 121.98,
      "learning_rate": 3.7364401294498386e-06,
      "loss": 0.0321,
      "step": 314090
    },
    {
      "epoch": 121.98,
      "learning_rate": 3.7359223300970878e-06,
      "loss": 0.0104,
      "step": 314100
    },
    {
      "epoch": 121.98,
      "learning_rate": 3.735404530744337e-06,
      "loss": 0.081,
      "step": 314110
    },
    {
      "epoch": 121.99,
      "learning_rate": 3.734886731391586e-06,
      "loss": 0.0154,
      "step": 314120
    },
    {
      "epoch": 121.99,
      "learning_rate": 3.7343689320388353e-06,
      "loss": 0.0444,
      "step": 314130
    },
    {
      "epoch": 122.0,
      "learning_rate": 3.7338511326860845e-06,
      "loss": 0.0241,
      "step": 314140
    },
    {
      "epoch": 122.0,
      "learning_rate": 3.7333333333333337e-06,
      "loss": 0.0955,
      "step": 314150
    },
    {
      "epoch": 122.0,
      "eval_accuracy": 0.9515818431911967,
      "eval_loss": 0.3805077075958252,
      "eval_runtime": 8.2196,
      "eval_samples_per_second": 442.234,
      "eval_steps_per_second": 55.355,
      "step": 314150
    },
    {
      "epoch": 122.0,
      "learning_rate": 3.732815533980583e-06,
      "loss": 0.0437,
      "step": 314160
    },
    {
      "epoch": 122.01,
      "learning_rate": 3.732297734627832e-06,
      "loss": 0.0447,
      "step": 314170
    },
    {
      "epoch": 122.01,
      "learning_rate": 3.731779935275081e-06,
      "loss": 0.0441,
      "step": 314180
    },
    {
      "epoch": 122.02,
      "learning_rate": 3.7312621359223305e-06,
      "loss": 0.0161,
      "step": 314190
    },
    {
      "epoch": 122.02,
      "learning_rate": 3.7307443365695793e-06,
      "loss": 0.1053,
      "step": 314200
    },
    {
      "epoch": 122.02,
      "learning_rate": 3.730226537216829e-06,
      "loss": 0.0645,
      "step": 314210
    },
    {
      "epoch": 122.03,
      "learning_rate": 3.7297087378640777e-06,
      "loss": 0.0135,
      "step": 314220
    },
    {
      "epoch": 122.03,
      "learning_rate": 3.7291909385113273e-06,
      "loss": 0.0034,
      "step": 314230
    },
    {
      "epoch": 122.03,
      "learning_rate": 3.728673139158576e-06,
      "loss": 0.0111,
      "step": 314240
    },
    {
      "epoch": 122.04,
      "learning_rate": 3.7281553398058257e-06,
      "loss": 0.0434,
      "step": 314250
    },
    {
      "epoch": 122.04,
      "learning_rate": 3.7276375404530744e-06,
      "loss": 0.0003,
      "step": 314260
    },
    {
      "epoch": 122.05,
      "learning_rate": 3.727119741100324e-06,
      "loss": 0.0075,
      "step": 314270
    },
    {
      "epoch": 122.05,
      "learning_rate": 3.726601941747573e-06,
      "loss": 0.1492,
      "step": 314280
    },
    {
      "epoch": 122.05,
      "learning_rate": 3.7260841423948225e-06,
      "loss": 0.137,
      "step": 314290
    },
    {
      "epoch": 122.06,
      "learning_rate": 3.7255663430420712e-06,
      "loss": 0.0186,
      "step": 314300
    },
    {
      "epoch": 122.06,
      "learning_rate": 3.725048543689321e-06,
      "loss": 0.0004,
      "step": 314310
    },
    {
      "epoch": 122.07,
      "learning_rate": 3.7245307443365696e-06,
      "loss": 0.0377,
      "step": 314320
    },
    {
      "epoch": 122.07,
      "learning_rate": 3.7240129449838192e-06,
      "loss": 0.0124,
      "step": 314330
    },
    {
      "epoch": 122.07,
      "learning_rate": 3.723495145631068e-06,
      "loss": 0.0119,
      "step": 314340
    },
    {
      "epoch": 122.08,
      "learning_rate": 3.7229773462783176e-06,
      "loss": 0.0562,
      "step": 314350
    },
    {
      "epoch": 122.08,
      "learning_rate": 3.7224595469255664e-06,
      "loss": 0.0171,
      "step": 314360
    },
    {
      "epoch": 122.09,
      "learning_rate": 3.721941747572816e-06,
      "loss": 0.1025,
      "step": 314370
    },
    {
      "epoch": 122.09,
      "learning_rate": 3.7214239482200648e-06,
      "loss": 0.0749,
      "step": 314380
    },
    {
      "epoch": 122.09,
      "learning_rate": 3.7209061488673144e-06,
      "loss": 0.0324,
      "step": 314390
    },
    {
      "epoch": 122.1,
      "learning_rate": 3.720388349514563e-06,
      "loss": 0.0401,
      "step": 314400
    },
    {
      "epoch": 122.1,
      "learning_rate": 3.7198705501618128e-06,
      "loss": 0.0263,
      "step": 314410
    },
    {
      "epoch": 122.1,
      "learning_rate": 3.7193527508090616e-06,
      "loss": 0.0204,
      "step": 314420
    },
    {
      "epoch": 122.11,
      "learning_rate": 3.718834951456311e-06,
      "loss": 0.0537,
      "step": 314430
    },
    {
      "epoch": 122.11,
      "learning_rate": 3.71831715210356e-06,
      "loss": 0.0003,
      "step": 314440
    },
    {
      "epoch": 122.12,
      "learning_rate": 3.7177993527508096e-06,
      "loss": 0.025,
      "step": 314450
    },
    {
      "epoch": 122.12,
      "learning_rate": 3.7172815533980583e-06,
      "loss": 0.0005,
      "step": 314460
    },
    {
      "epoch": 122.12,
      "learning_rate": 3.716763754045308e-06,
      "loss": 0.0219,
      "step": 314470
    },
    {
      "epoch": 122.13,
      "learning_rate": 3.7162459546925567e-06,
      "loss": 0.1092,
      "step": 314480
    },
    {
      "epoch": 122.13,
      "learning_rate": 3.715728155339806e-06,
      "loss": 0.0601,
      "step": 314490
    },
    {
      "epoch": 122.14,
      "learning_rate": 3.7152103559870555e-06,
      "loss": 0.0602,
      "step": 314500
    },
    {
      "epoch": 122.14,
      "learning_rate": 3.7146925566343043e-06,
      "loss": 0.0675,
      "step": 314510
    },
    {
      "epoch": 122.14,
      "learning_rate": 3.714174757281554e-06,
      "loss": 0.022,
      "step": 314520
    },
    {
      "epoch": 122.15,
      "learning_rate": 3.7136569579288027e-06,
      "loss": 0.1135,
      "step": 314530
    },
    {
      "epoch": 122.15,
      "learning_rate": 3.7131391585760523e-06,
      "loss": 0.0011,
      "step": 314540
    },
    {
      "epoch": 122.16,
      "learning_rate": 3.712621359223301e-06,
      "loss": 0.0002,
      "step": 314550
    },
    {
      "epoch": 122.16,
      "learning_rate": 3.7121035598705507e-06,
      "loss": 0.0395,
      "step": 314560
    },
    {
      "epoch": 122.16,
      "learning_rate": 3.7115857605177995e-06,
      "loss": 0.0773,
      "step": 314570
    },
    {
      "epoch": 122.17,
      "learning_rate": 3.711067961165049e-06,
      "loss": 0.1662,
      "step": 314580
    },
    {
      "epoch": 122.17,
      "learning_rate": 3.710550161812298e-06,
      "loss": 0.001,
      "step": 314590
    },
    {
      "epoch": 122.17,
      "learning_rate": 3.7100323624595475e-06,
      "loss": 0.0219,
      "step": 314600
    },
    {
      "epoch": 122.18,
      "learning_rate": 3.7095145631067963e-06,
      "loss": 0.0003,
      "step": 314610
    },
    {
      "epoch": 122.18,
      "learning_rate": 3.708996763754046e-06,
      "loss": 0.0001,
      "step": 314620
    },
    {
      "epoch": 122.19,
      "learning_rate": 3.7084789644012946e-06,
      "loss": 0.0872,
      "step": 314630
    },
    {
      "epoch": 122.19,
      "learning_rate": 3.7079611650485443e-06,
      "loss": 0.0691,
      "step": 314640
    },
    {
      "epoch": 122.19,
      "learning_rate": 3.707443365695793e-06,
      "loss": 0.0714,
      "step": 314650
    },
    {
      "epoch": 122.2,
      "learning_rate": 3.7069255663430426e-06,
      "loss": 0.1219,
      "step": 314660
    },
    {
      "epoch": 122.2,
      "learning_rate": 3.7064077669902914e-06,
      "loss": 0.1325,
      "step": 314670
    },
    {
      "epoch": 122.21,
      "learning_rate": 3.705889967637541e-06,
      "loss": 0.0277,
      "step": 314680
    },
    {
      "epoch": 122.21,
      "learning_rate": 3.70537216828479e-06,
      "loss": 0.1217,
      "step": 314690
    },
    {
      "epoch": 122.21,
      "learning_rate": 3.7048543689320394e-06,
      "loss": 0.0322,
      "step": 314700
    },
    {
      "epoch": 122.22,
      "learning_rate": 3.704336569579288e-06,
      "loss": 0.0578,
      "step": 314710
    },
    {
      "epoch": 122.22,
      "learning_rate": 3.703818770226538e-06,
      "loss": 0.0015,
      "step": 314720
    },
    {
      "epoch": 122.23,
      "learning_rate": 3.7033009708737866e-06,
      "loss": 0.0004,
      "step": 314730
    },
    {
      "epoch": 122.23,
      "learning_rate": 3.702783171521036e-06,
      "loss": 0.0317,
      "step": 314740
    },
    {
      "epoch": 122.23,
      "learning_rate": 3.702265372168285e-06,
      "loss": 0.0284,
      "step": 314750
    },
    {
      "epoch": 122.24,
      "learning_rate": 3.7017475728155346e-06,
      "loss": 0.0692,
      "step": 314760
    },
    {
      "epoch": 122.24,
      "learning_rate": 3.7012297734627834e-06,
      "loss": 0.0011,
      "step": 314770
    },
    {
      "epoch": 122.24,
      "learning_rate": 3.700711974110033e-06,
      "loss": 0.0002,
      "step": 314780
    },
    {
      "epoch": 122.25,
      "learning_rate": 3.7001941747572818e-06,
      "loss": 0.0002,
      "step": 314790
    },
    {
      "epoch": 122.25,
      "learning_rate": 3.6996763754045314e-06,
      "loss": 0.1122,
      "step": 314800
    },
    {
      "epoch": 122.26,
      "learning_rate": 3.69915857605178e-06,
      "loss": 0.023,
      "step": 314810
    },
    {
      "epoch": 122.26,
      "learning_rate": 3.6986407766990293e-06,
      "loss": 0.0146,
      "step": 314820
    },
    {
      "epoch": 122.26,
      "learning_rate": 3.6981229773462785e-06,
      "loss": 0.0645,
      "step": 314830
    },
    {
      "epoch": 122.27,
      "learning_rate": 3.6976051779935277e-06,
      "loss": 0.0003,
      "step": 314840
    },
    {
      "epoch": 122.27,
      "learning_rate": 3.697087378640777e-06,
      "loss": 0.069,
      "step": 314850
    },
    {
      "epoch": 122.28,
      "learning_rate": 3.696569579288026e-06,
      "loss": 0.0678,
      "step": 314860
    },
    {
      "epoch": 122.28,
      "learning_rate": 3.6960517799352753e-06,
      "loss": 0.1672,
      "step": 314870
    },
    {
      "epoch": 122.28,
      "learning_rate": 3.6955339805825245e-06,
      "loss": 0.0332,
      "step": 314880
    },
    {
      "epoch": 122.29,
      "learning_rate": 3.6950161812297737e-06,
      "loss": 0.0274,
      "step": 314890
    },
    {
      "epoch": 122.29,
      "learning_rate": 3.694498381877023e-06,
      "loss": 0.1067,
      "step": 314900
    },
    {
      "epoch": 122.3,
      "learning_rate": 3.693980582524272e-06,
      "loss": 0.0925,
      "step": 314910
    },
    {
      "epoch": 122.3,
      "learning_rate": 3.6934627831715213e-06,
      "loss": 0.0464,
      "step": 314920
    },
    {
      "epoch": 122.3,
      "learning_rate": 3.6929449838187705e-06,
      "loss": 0.081,
      "step": 314930
    },
    {
      "epoch": 122.31,
      "learning_rate": 3.6924271844660197e-06,
      "loss": 0.0229,
      "step": 314940
    },
    {
      "epoch": 122.31,
      "learning_rate": 3.691909385113269e-06,
      "loss": 0.022,
      "step": 314950
    },
    {
      "epoch": 122.31,
      "learning_rate": 3.691391585760518e-06,
      "loss": 0.0211,
      "step": 314960
    },
    {
      "epoch": 122.32,
      "learning_rate": 3.6908737864077673e-06,
      "loss": 0.0656,
      "step": 314970
    },
    {
      "epoch": 122.32,
      "learning_rate": 3.6903559870550165e-06,
      "loss": 0.0216,
      "step": 314980
    },
    {
      "epoch": 122.33,
      "learning_rate": 3.6898381877022652e-06,
      "loss": 0.0131,
      "step": 314990
    },
    {
      "epoch": 122.33,
      "learning_rate": 3.689320388349515e-06,
      "loss": 0.0005,
      "step": 315000
    },
    {
      "epoch": 122.33,
      "learning_rate": 3.6888025889967636e-06,
      "loss": 0.0001,
      "step": 315010
    },
    {
      "epoch": 122.34,
      "learning_rate": 3.6882847896440132e-06,
      "loss": 0.0409,
      "step": 315020
    },
    {
      "epoch": 122.34,
      "learning_rate": 3.687766990291262e-06,
      "loss": 0.009,
      "step": 315030
    },
    {
      "epoch": 122.35,
      "learning_rate": 3.6872491909385116e-06,
      "loss": 0.0101,
      "step": 315040
    },
    {
      "epoch": 122.35,
      "learning_rate": 3.6867313915857604e-06,
      "loss": 0.0114,
      "step": 315050
    },
    {
      "epoch": 122.35,
      "learning_rate": 3.68621359223301e-06,
      "loss": 0.0531,
      "step": 315060
    },
    {
      "epoch": 122.36,
      "learning_rate": 3.6856957928802596e-06,
      "loss": 0.0958,
      "step": 315070
    },
    {
      "epoch": 122.36,
      "learning_rate": 3.6851779935275084e-06,
      "loss": 0.0788,
      "step": 315080
    },
    {
      "epoch": 122.37,
      "learning_rate": 3.684660194174758e-06,
      "loss": 0.0453,
      "step": 315090
    },
    {
      "epoch": 122.37,
      "learning_rate": 3.6841423948220068e-06,
      "loss": 0.0031,
      "step": 315100
    },
    {
      "epoch": 122.37,
      "learning_rate": 3.6836245954692564e-06,
      "loss": 0.0551,
      "step": 315110
    },
    {
      "epoch": 122.38,
      "learning_rate": 3.683106796116505e-06,
      "loss": 0.1995,
      "step": 315120
    },
    {
      "epoch": 122.38,
      "learning_rate": 3.682588996763755e-06,
      "loss": 0.0732,
      "step": 315130
    },
    {
      "epoch": 122.38,
      "learning_rate": 3.6820711974110036e-06,
      "loss": 0.1225,
      "step": 315140
    },
    {
      "epoch": 122.39,
      "learning_rate": 3.6815533980582528e-06,
      "loss": 0.1026,
      "step": 315150
    },
    {
      "epoch": 122.39,
      "learning_rate": 3.681035598705502e-06,
      "loss": 0.0848,
      "step": 315160
    },
    {
      "epoch": 122.4,
      "learning_rate": 3.680517799352751e-06,
      "loss": 0.0222,
      "step": 315170
    },
    {
      "epoch": 122.4,
      "learning_rate": 3.6800000000000003e-06,
      "loss": 0.169,
      "step": 315180
    },
    {
      "epoch": 122.4,
      "learning_rate": 3.6794822006472495e-06,
      "loss": 0.0099,
      "step": 315190
    },
    {
      "epoch": 122.41,
      "learning_rate": 3.6789644012944987e-06,
      "loss": 0.0456,
      "step": 315200
    },
    {
      "epoch": 122.41,
      "learning_rate": 3.678446601941748e-06,
      "loss": 0.0657,
      "step": 315210
    },
    {
      "epoch": 122.42,
      "learning_rate": 3.677928802588997e-06,
      "loss": 0.0132,
      "step": 315220
    },
    {
      "epoch": 122.42,
      "learning_rate": 3.6774110032362463e-06,
      "loss": 0.0594,
      "step": 315230
    },
    {
      "epoch": 122.42,
      "learning_rate": 3.6768932038834955e-06,
      "loss": 0.0114,
      "step": 315240
    },
    {
      "epoch": 122.43,
      "learning_rate": 3.6763754045307447e-06,
      "loss": 0.07,
      "step": 315250
    },
    {
      "epoch": 122.43,
      "learning_rate": 3.675857605177994e-06,
      "loss": 0.0316,
      "step": 315260
    },
    {
      "epoch": 122.43,
      "learning_rate": 3.675339805825243e-06,
      "loss": 0.0004,
      "step": 315270
    },
    {
      "epoch": 122.44,
      "learning_rate": 3.6748220064724923e-06,
      "loss": 0.003,
      "step": 315280
    },
    {
      "epoch": 122.44,
      "learning_rate": 3.6743042071197415e-06,
      "loss": 0.0197,
      "step": 315290
    },
    {
      "epoch": 122.45,
      "learning_rate": 3.6737864077669907e-06,
      "loss": 0.0168,
      "step": 315300
    },
    {
      "epoch": 122.45,
      "learning_rate": 3.67326860841424e-06,
      "loss": 0.0448,
      "step": 315310
    },
    {
      "epoch": 122.45,
      "learning_rate": 3.6727508090614886e-06,
      "loss": 0.0162,
      "step": 315320
    },
    {
      "epoch": 122.46,
      "learning_rate": 3.6722330097087383e-06,
      "loss": 0.0592,
      "step": 315330
    },
    {
      "epoch": 122.46,
      "learning_rate": 3.671715210355987e-06,
      "loss": 0.0005,
      "step": 315340
    },
    {
      "epoch": 122.47,
      "learning_rate": 3.6711974110032366e-06,
      "loss": 0.0122,
      "step": 315350
    },
    {
      "epoch": 122.47,
      "learning_rate": 3.6706796116504854e-06,
      "loss": 0.0009,
      "step": 315360
    },
    {
      "epoch": 122.47,
      "learning_rate": 3.670161812297735e-06,
      "loss": 0.0472,
      "step": 315370
    },
    {
      "epoch": 122.48,
      "learning_rate": 3.669644012944984e-06,
      "loss": 0.0763,
      "step": 315380
    },
    {
      "epoch": 122.48,
      "learning_rate": 3.6691262135922334e-06,
      "loss": 0.0017,
      "step": 315390
    },
    {
      "epoch": 122.49,
      "learning_rate": 3.668608414239482e-06,
      "loss": 0.0742,
      "step": 315400
    },
    {
      "epoch": 122.49,
      "learning_rate": 3.668090614886732e-06,
      "loss": 0.0299,
      "step": 315410
    },
    {
      "epoch": 122.49,
      "learning_rate": 3.6675728155339806e-06,
      "loss": 0.0085,
      "step": 315420
    },
    {
      "epoch": 122.5,
      "learning_rate": 3.66705501618123e-06,
      "loss": 0.0438,
      "step": 315430
    },
    {
      "epoch": 122.5,
      "learning_rate": 3.666537216828479e-06,
      "loss": 0.02,
      "step": 315440
    },
    {
      "epoch": 122.5,
      "learning_rate": 3.6660194174757286e-06,
      "loss": 0.0155,
      "step": 315450
    },
    {
      "epoch": 122.51,
      "learning_rate": 3.6655016181229774e-06,
      "loss": 0.0003,
      "step": 315460
    },
    {
      "epoch": 122.51,
      "learning_rate": 3.664983818770227e-06,
      "loss": 0.0126,
      "step": 315470
    },
    {
      "epoch": 122.52,
      "learning_rate": 3.6644660194174758e-06,
      "loss": 0.0001,
      "step": 315480
    },
    {
      "epoch": 122.52,
      "learning_rate": 3.6639482200647254e-06,
      "loss": 0.1207,
      "step": 315490
    },
    {
      "epoch": 122.52,
      "learning_rate": 3.663430420711974e-06,
      "loss": 0.0055,
      "step": 315500
    },
    {
      "epoch": 122.53,
      "learning_rate": 3.6629126213592238e-06,
      "loss": 0.1579,
      "step": 315510
    },
    {
      "epoch": 122.53,
      "learning_rate": 3.6623948220064725e-06,
      "loss": 0.0225,
      "step": 315520
    },
    {
      "epoch": 122.54,
      "learning_rate": 3.661877022653722e-06,
      "loss": 0.0166,
      "step": 315530
    },
    {
      "epoch": 122.54,
      "learning_rate": 3.661359223300971e-06,
      "loss": 0.0608,
      "step": 315540
    },
    {
      "epoch": 122.54,
      "learning_rate": 3.6608414239482205e-06,
      "loss": 0.0687,
      "step": 315550
    },
    {
      "epoch": 122.55,
      "learning_rate": 3.6603236245954693e-06,
      "loss": 0.0029,
      "step": 315560
    },
    {
      "epoch": 122.55,
      "learning_rate": 3.659805825242719e-06,
      "loss": 0.0441,
      "step": 315570
    },
    {
      "epoch": 122.56,
      "learning_rate": 3.6592880258899677e-06,
      "loss": 0.2118,
      "step": 315580
    },
    {
      "epoch": 122.56,
      "learning_rate": 3.6587702265372173e-06,
      "loss": 0.0557,
      "step": 315590
    },
    {
      "epoch": 122.56,
      "learning_rate": 3.658252427184466e-06,
      "loss": 0.1214,
      "step": 315600
    },
    {
      "epoch": 122.57,
      "learning_rate": 3.6577346278317157e-06,
      "loss": 0.0969,
      "step": 315610
    },
    {
      "epoch": 122.57,
      "learning_rate": 3.657216828478965e-06,
      "loss": 0.062,
      "step": 315620
    },
    {
      "epoch": 122.57,
      "learning_rate": 3.656699029126214e-06,
      "loss": 0.052,
      "step": 315630
    },
    {
      "epoch": 122.58,
      "learning_rate": 3.6561812297734633e-06,
      "loss": 0.0002,
      "step": 315640
    },
    {
      "epoch": 122.58,
      "learning_rate": 3.655663430420712e-06,
      "loss": 0.0822,
      "step": 315650
    },
    {
      "epoch": 122.59,
      "learning_rate": 3.6551456310679617e-06,
      "loss": 0.0015,
      "step": 315660
    },
    {
      "epoch": 122.59,
      "learning_rate": 3.6546278317152104e-06,
      "loss": 0.0086,
      "step": 315670
    },
    {
      "epoch": 122.59,
      "learning_rate": 3.65411003236246e-06,
      "loss": 0.0505,
      "step": 315680
    },
    {
      "epoch": 122.6,
      "learning_rate": 3.653592233009709e-06,
      "loss": 0.0734,
      "step": 315690
    },
    {
      "epoch": 122.6,
      "learning_rate": 3.6530744336569585e-06,
      "loss": 0.0245,
      "step": 315700
    },
    {
      "epoch": 122.61,
      "learning_rate": 3.6525566343042072e-06,
      "loss": 0.0211,
      "step": 315710
    },
    {
      "epoch": 122.61,
      "learning_rate": 3.652038834951457e-06,
      "loss": 0.0208,
      "step": 315720
    },
    {
      "epoch": 122.61,
      "learning_rate": 3.6515210355987056e-06,
      "loss": 0.0602,
      "step": 315730
    },
    {
      "epoch": 122.62,
      "learning_rate": 3.6510032362459552e-06,
      "loss": 0.0009,
      "step": 315740
    },
    {
      "epoch": 122.62,
      "learning_rate": 3.650485436893204e-06,
      "loss": 0.0007,
      "step": 315750
    },
    {
      "epoch": 122.63,
      "learning_rate": 3.6499676375404536e-06,
      "loss": 0.0044,
      "step": 315760
    },
    {
      "epoch": 122.63,
      "learning_rate": 3.6494498381877024e-06,
      "loss": 0.086,
      "step": 315770
    },
    {
      "epoch": 122.63,
      "learning_rate": 3.648932038834952e-06,
      "loss": 0.1922,
      "step": 315780
    },
    {
      "epoch": 122.64,
      "learning_rate": 3.6484142394822008e-06,
      "loss": 0.1045,
      "step": 315790
    },
    {
      "epoch": 122.64,
      "learning_rate": 3.6478964401294504e-06,
      "loss": 0.0111,
      "step": 315800
    },
    {
      "epoch": 122.64,
      "learning_rate": 3.647378640776699e-06,
      "loss": 0.0616,
      "step": 315810
    },
    {
      "epoch": 122.65,
      "learning_rate": 3.646860841423949e-06,
      "loss": 0.0002,
      "step": 315820
    },
    {
      "epoch": 122.65,
      "learning_rate": 3.6463430420711976e-06,
      "loss": 0.0001,
      "step": 315830
    },
    {
      "epoch": 122.66,
      "learning_rate": 3.645825242718447e-06,
      "loss": 0.0161,
      "step": 315840
    },
    {
      "epoch": 122.66,
      "learning_rate": 3.645307443365696e-06,
      "loss": 0.0829,
      "step": 315850
    },
    {
      "epoch": 122.66,
      "learning_rate": 3.6447896440129456e-06,
      "loss": 0.0404,
      "step": 315860
    },
    {
      "epoch": 122.67,
      "learning_rate": 3.6442718446601943e-06,
      "loss": 0.0484,
      "step": 315870
    },
    {
      "epoch": 122.67,
      "learning_rate": 3.643754045307444e-06,
      "loss": 0.0137,
      "step": 315880
    },
    {
      "epoch": 122.68,
      "learning_rate": 3.6432362459546927e-06,
      "loss": 0.1349,
      "step": 315890
    },
    {
      "epoch": 122.68,
      "learning_rate": 3.6427184466019423e-06,
      "loss": 0.1004,
      "step": 315900
    },
    {
      "epoch": 122.68,
      "learning_rate": 3.642200647249191e-06,
      "loss": 0.0013,
      "step": 315910
    },
    {
      "epoch": 122.69,
      "learning_rate": 3.6416828478964407e-06,
      "loss": 0.0233,
      "step": 315920
    },
    {
      "epoch": 122.69,
      "learning_rate": 3.6411650485436895e-06,
      "loss": 0.1507,
      "step": 315930
    },
    {
      "epoch": 122.7,
      "learning_rate": 3.640647249190939e-06,
      "loss": 0.1468,
      "step": 315940
    },
    {
      "epoch": 122.7,
      "learning_rate": 3.640129449838188e-06,
      "loss": 0.0006,
      "step": 315950
    },
    {
      "epoch": 122.7,
      "learning_rate": 3.639611650485437e-06,
      "loss": 0.066,
      "step": 315960
    },
    {
      "epoch": 122.71,
      "learning_rate": 3.6390938511326863e-06,
      "loss": 0.0723,
      "step": 315970
    },
    {
      "epoch": 122.71,
      "learning_rate": 3.6385760517799355e-06,
      "loss": 0.0271,
      "step": 315980
    },
    {
      "epoch": 122.71,
      "learning_rate": 3.6380582524271847e-06,
      "loss": 0.0005,
      "step": 315990
    },
    {
      "epoch": 122.72,
      "learning_rate": 3.637540453074434e-06,
      "loss": 0.0013,
      "step": 316000
    },
    {
      "epoch": 122.72,
      "learning_rate": 3.637022653721683e-06,
      "loss": 0.0867,
      "step": 316010
    },
    {
      "epoch": 122.73,
      "learning_rate": 3.6365048543689323e-06,
      "loss": 0.032,
      "step": 316020
    },
    {
      "epoch": 122.73,
      "learning_rate": 3.6359870550161815e-06,
      "loss": 0.0561,
      "step": 316030
    },
    {
      "epoch": 122.73,
      "learning_rate": 3.6354692556634306e-06,
      "loss": 0.0805,
      "step": 316040
    },
    {
      "epoch": 122.74,
      "learning_rate": 3.63495145631068e-06,
      "loss": 0.1851,
      "step": 316050
    },
    {
      "epoch": 122.74,
      "learning_rate": 3.634433656957929e-06,
      "loss": 0.0716,
      "step": 316060
    },
    {
      "epoch": 122.75,
      "learning_rate": 3.6339158576051782e-06,
      "loss": 0.01,
      "step": 316070
    },
    {
      "epoch": 122.75,
      "learning_rate": 3.6333980582524274e-06,
      "loss": 0.0184,
      "step": 316080
    },
    {
      "epoch": 122.75,
      "learning_rate": 3.6328802588996766e-06,
      "loss": 0.0652,
      "step": 316090
    },
    {
      "epoch": 122.76,
      "learning_rate": 3.632362459546926e-06,
      "loss": 0.0083,
      "step": 316100
    },
    {
      "epoch": 122.76,
      "learning_rate": 3.631844660194175e-06,
      "loss": 0.0256,
      "step": 316110
    },
    {
      "epoch": 122.77,
      "learning_rate": 3.631326860841424e-06,
      "loss": 0.0157,
      "step": 316120
    },
    {
      "epoch": 122.77,
      "learning_rate": 3.630809061488673e-06,
      "loss": 0.0561,
      "step": 316130
    },
    {
      "epoch": 122.77,
      "learning_rate": 3.6302912621359226e-06,
      "loss": 0.1023,
      "step": 316140
    },
    {
      "epoch": 122.78,
      "learning_rate": 3.6297734627831714e-06,
      "loss": 0.0148,
      "step": 316150
    },
    {
      "epoch": 122.78,
      "learning_rate": 3.629255663430421e-06,
      "loss": 0.0517,
      "step": 316160
    },
    {
      "epoch": 122.78,
      "learning_rate": 3.6287378640776698e-06,
      "loss": 0.0254,
      "step": 316170
    },
    {
      "epoch": 122.79,
      "learning_rate": 3.6282200647249194e-06,
      "loss": 0.0352,
      "step": 316180
    },
    {
      "epoch": 122.79,
      "learning_rate": 3.627702265372169e-06,
      "loss": 0.1056,
      "step": 316190
    },
    {
      "epoch": 122.8,
      "learning_rate": 3.6271844660194178e-06,
      "loss": 0.1261,
      "step": 316200
    },
    {
      "epoch": 122.8,
      "learning_rate": 3.6266666666666674e-06,
      "loss": 0.0681,
      "step": 316210
    },
    {
      "epoch": 122.8,
      "learning_rate": 3.626148867313916e-06,
      "loss": 0.0101,
      "step": 316220
    },
    {
      "epoch": 122.81,
      "learning_rate": 3.6256310679611658e-06,
      "loss": 0.0126,
      "step": 316230
    },
    {
      "epoch": 122.81,
      "learning_rate": 3.6251132686084145e-06,
      "loss": 0.0291,
      "step": 316240
    },
    {
      "epoch": 122.82,
      "learning_rate": 3.624595469255664e-06,
      "loss": 0.0423,
      "step": 316250
    },
    {
      "epoch": 122.82,
      "learning_rate": 3.624077669902913e-06,
      "loss": 0.1383,
      "step": 316260
    },
    {
      "epoch": 122.82,
      "learning_rate": 3.6235598705501625e-06,
      "loss": 0.0015,
      "step": 316270
    },
    {
      "epoch": 122.83,
      "learning_rate": 3.6230420711974113e-06,
      "loss": 0.1236,
      "step": 316280
    },
    {
      "epoch": 122.83,
      "learning_rate": 3.6225242718446605e-06,
      "loss": 0.0288,
      "step": 316290
    },
    {
      "epoch": 122.83,
      "learning_rate": 3.6220064724919097e-06,
      "loss": 0.1618,
      "step": 316300
    },
    {
      "epoch": 122.84,
      "learning_rate": 3.621488673139159e-06,
      "loss": 0.0629,
      "step": 316310
    },
    {
      "epoch": 122.84,
      "learning_rate": 3.620970873786408e-06,
      "loss": 0.0969,
      "step": 316320
    },
    {
      "epoch": 122.85,
      "learning_rate": 3.6204530744336573e-06,
      "loss": 0.0496,
      "step": 316330
    },
    {
      "epoch": 122.85,
      "learning_rate": 3.6199352750809065e-06,
      "loss": 0.0074,
      "step": 316340
    },
    {
      "epoch": 122.85,
      "learning_rate": 3.6194174757281557e-06,
      "loss": 0.1352,
      "step": 316350
    },
    {
      "epoch": 122.86,
      "learning_rate": 3.618899676375405e-06,
      "loss": 0.0699,
      "step": 316360
    },
    {
      "epoch": 122.86,
      "learning_rate": 3.618381877022654e-06,
      "loss": 0.0249,
      "step": 316370
    },
    {
      "epoch": 122.87,
      "learning_rate": 3.6178640776699033e-06,
      "loss": 0.0229,
      "step": 316380
    },
    {
      "epoch": 122.87,
      "learning_rate": 3.6173462783171525e-06,
      "loss": 0.0704,
      "step": 316390
    },
    {
      "epoch": 122.87,
      "learning_rate": 3.6168284789644017e-06,
      "loss": 0.0155,
      "step": 316400
    },
    {
      "epoch": 122.88,
      "learning_rate": 3.616310679611651e-06,
      "loss": 0.082,
      "step": 316410
    },
    {
      "epoch": 122.88,
      "learning_rate": 3.6157928802589e-06,
      "loss": 0.0014,
      "step": 316420
    },
    {
      "epoch": 122.89,
      "learning_rate": 3.6152750809061492e-06,
      "loss": 0.0728,
      "step": 316430
    },
    {
      "epoch": 122.89,
      "learning_rate": 3.6147572815533984e-06,
      "loss": 0.0162,
      "step": 316440
    },
    {
      "epoch": 122.89,
      "learning_rate": 3.6142394822006476e-06,
      "loss": 0.0971,
      "step": 316450
    },
    {
      "epoch": 122.9,
      "learning_rate": 3.6137216828478964e-06,
      "loss": 0.0003,
      "step": 316460
    },
    {
      "epoch": 122.9,
      "learning_rate": 3.613203883495146e-06,
      "loss": 0.0138,
      "step": 316470
    },
    {
      "epoch": 122.9,
      "learning_rate": 3.6126860841423948e-06,
      "loss": 0.0263,
      "step": 316480
    },
    {
      "epoch": 122.91,
      "learning_rate": 3.6121682847896444e-06,
      "loss": 0.0003,
      "step": 316490
    },
    {
      "epoch": 122.91,
      "learning_rate": 3.611650485436893e-06,
      "loss": 0.0021,
      "step": 316500
    },
    {
      "epoch": 122.92,
      "learning_rate": 3.611132686084143e-06,
      "loss": 0.0902,
      "step": 316510
    },
    {
      "epoch": 122.92,
      "learning_rate": 3.6106148867313916e-06,
      "loss": 0.0503,
      "step": 316520
    },
    {
      "epoch": 122.92,
      "learning_rate": 3.610097087378641e-06,
      "loss": 0.0274,
      "step": 316530
    },
    {
      "epoch": 122.93,
      "learning_rate": 3.60957928802589e-06,
      "loss": 0.0596,
      "step": 316540
    },
    {
      "epoch": 122.93,
      "learning_rate": 3.6090614886731396e-06,
      "loss": 0.0686,
      "step": 316550
    },
    {
      "epoch": 122.94,
      "learning_rate": 3.6085436893203883e-06,
      "loss": 0.1064,
      "step": 316560
    },
    {
      "epoch": 122.94,
      "learning_rate": 3.608025889967638e-06,
      "loss": 0.0459,
      "step": 316570
    },
    {
      "epoch": 122.94,
      "learning_rate": 3.6075080906148867e-06,
      "loss": 0.0066,
      "step": 316580
    },
    {
      "epoch": 122.95,
      "learning_rate": 3.6069902912621363e-06,
      "loss": 0.0545,
      "step": 316590
    },
    {
      "epoch": 122.95,
      "learning_rate": 3.606472491909385e-06,
      "loss": 0.0027,
      "step": 316600
    },
    {
      "epoch": 122.96,
      "learning_rate": 3.6059546925566347e-06,
      "loss": 0.1454,
      "step": 316610
    },
    {
      "epoch": 122.96,
      "learning_rate": 3.6054368932038835e-06,
      "loss": 0.1015,
      "step": 316620
    },
    {
      "epoch": 122.96,
      "learning_rate": 3.604919093851133e-06,
      "loss": 0.0258,
      "step": 316630
    },
    {
      "epoch": 122.97,
      "learning_rate": 3.604401294498382e-06,
      "loss": 0.0193,
      "step": 316640
    },
    {
      "epoch": 122.97,
      "learning_rate": 3.6038834951456315e-06,
      "loss": 0.0588,
      "step": 316650
    },
    {
      "epoch": 122.97,
      "learning_rate": 3.6033656957928803e-06,
      "loss": 0.0415,
      "step": 316660
    },
    {
      "epoch": 122.98,
      "learning_rate": 3.60284789644013e-06,
      "loss": 0.0468,
      "step": 316670
    },
    {
      "epoch": 122.98,
      "learning_rate": 3.6023300970873787e-06,
      "loss": 0.0622,
      "step": 316680
    },
    {
      "epoch": 122.99,
      "learning_rate": 3.6018122977346283e-06,
      "loss": 0.0018,
      "step": 316690
    },
    {
      "epoch": 122.99,
      "learning_rate": 3.601294498381877e-06,
      "loss": 0.0142,
      "step": 316700
    },
    {
      "epoch": 122.99,
      "learning_rate": 3.6007766990291267e-06,
      "loss": 0.0004,
      "step": 316710
    },
    {
      "epoch": 123.0,
      "learning_rate": 3.6002588996763755e-06,
      "loss": 0.0043,
      "step": 316720
    },
    {
      "epoch": 123.0,
      "eval_accuracy": 0.9521320495185694,
      "eval_loss": 0.38330063223838806,
      "eval_runtime": 8.1818,
      "eval_samples_per_second": 444.276,
      "eval_steps_per_second": 55.611,
      "step": 316725
    },
    {
      "epoch": 123.0,
      "learning_rate": 3.599741100323625e-06,
      "loss": 0.0721,
      "step": 316730
    },
    {
      "epoch": 123.01,
      "learning_rate": 3.5992233009708743e-06,
      "loss": 0.0073,
      "step": 316740
    },
    {
      "epoch": 123.01,
      "learning_rate": 3.5987055016181235e-06,
      "loss": 0.011,
      "step": 316750
    },
    {
      "epoch": 123.01,
      "learning_rate": 3.5981877022653727e-06,
      "loss": 0.0015,
      "step": 316760
    },
    {
      "epoch": 123.02,
      "learning_rate": 3.597669902912622e-06,
      "loss": 0.0539,
      "step": 316770
    },
    {
      "epoch": 123.02,
      "learning_rate": 3.597152103559871e-06,
      "loss": 0.0008,
      "step": 316780
    },
    {
      "epoch": 123.03,
      "learning_rate": 3.59663430420712e-06,
      "loss": 0.0001,
      "step": 316790
    },
    {
      "epoch": 123.03,
      "learning_rate": 3.5961165048543694e-06,
      "loss": 0.0125,
      "step": 316800
    },
    {
      "epoch": 123.03,
      "learning_rate": 3.595598705501618e-06,
      "loss": 0.024,
      "step": 316810
    },
    {
      "epoch": 123.04,
      "learning_rate": 3.595080906148868e-06,
      "loss": 0.0291,
      "step": 316820
    },
    {
      "epoch": 123.04,
      "learning_rate": 3.5945631067961166e-06,
      "loss": 0.073,
      "step": 316830
    },
    {
      "epoch": 123.04,
      "learning_rate": 3.594045307443366e-06,
      "loss": 0.0812,
      "step": 316840
    },
    {
      "epoch": 123.05,
      "learning_rate": 3.593527508090615e-06,
      "loss": 0.0545,
      "step": 316850
    },
    {
      "epoch": 123.05,
      "learning_rate": 3.5930097087378646e-06,
      "loss": 0.0571,
      "step": 316860
    },
    {
      "epoch": 123.06,
      "learning_rate": 3.5924919093851134e-06,
      "loss": 0.1486,
      "step": 316870
    },
    {
      "epoch": 123.06,
      "learning_rate": 3.591974110032363e-06,
      "loss": 0.0172,
      "step": 316880
    },
    {
      "epoch": 123.06,
      "learning_rate": 3.5914563106796118e-06,
      "loss": 0.0299,
      "step": 316890
    },
    {
      "epoch": 123.07,
      "learning_rate": 3.5909385113268614e-06,
      "loss": 0.0067,
      "step": 316900
    },
    {
      "epoch": 123.07,
      "learning_rate": 3.59042071197411e-06,
      "loss": 0.0654,
      "step": 316910
    },
    {
      "epoch": 123.08,
      "learning_rate": 3.5899029126213598e-06,
      "loss": 0.0696,
      "step": 316920
    },
    {
      "epoch": 123.08,
      "learning_rate": 3.5893851132686085e-06,
      "loss": 0.068,
      "step": 316930
    },
    {
      "epoch": 123.08,
      "learning_rate": 3.588867313915858e-06,
      "loss": 0.0724,
      "step": 316940
    },
    {
      "epoch": 123.09,
      "learning_rate": 3.588349514563107e-06,
      "loss": 0.0825,
      "step": 316950
    },
    {
      "epoch": 123.09,
      "learning_rate": 3.5878317152103565e-06,
      "loss": 0.0002,
      "step": 316960
    },
    {
      "epoch": 123.1,
      "learning_rate": 3.5873139158576053e-06,
      "loss": 0.0161,
      "step": 316970
    },
    {
      "epoch": 123.1,
      "learning_rate": 3.586796116504855e-06,
      "loss": 0.0078,
      "step": 316980
    },
    {
      "epoch": 123.1,
      "learning_rate": 3.5862783171521037e-06,
      "loss": 0.0883,
      "step": 316990
    },
    {
      "epoch": 123.11,
      "learning_rate": 3.5857605177993533e-06,
      "loss": 0.1626,
      "step": 317000
    },
    {
      "epoch": 123.11,
      "learning_rate": 3.585242718446602e-06,
      "loss": 0.0917,
      "step": 317010
    },
    {
      "epoch": 123.11,
      "learning_rate": 3.5847249190938517e-06,
      "loss": 0.111,
      "step": 317020
    },
    {
      "epoch": 123.12,
      "learning_rate": 3.5842071197411005e-06,
      "loss": 0.0681,
      "step": 317030
    },
    {
      "epoch": 123.12,
      "learning_rate": 3.58368932038835e-06,
      "loss": 0.0001,
      "step": 317040
    },
    {
      "epoch": 123.13,
      "learning_rate": 3.583171521035599e-06,
      "loss": 0.0209,
      "step": 317050
    },
    {
      "epoch": 123.13,
      "learning_rate": 3.5826537216828485e-06,
      "loss": 0.1319,
      "step": 317060
    },
    {
      "epoch": 123.13,
      "learning_rate": 3.5821359223300973e-06,
      "loss": 0.0248,
      "step": 317070
    },
    {
      "epoch": 123.14,
      "learning_rate": 3.581618122977347e-06,
      "loss": 0.1271,
      "step": 317080
    },
    {
      "epoch": 123.14,
      "learning_rate": 3.5811003236245956e-06,
      "loss": 0.0589,
      "step": 317090
    },
    {
      "epoch": 123.15,
      "learning_rate": 3.580582524271845e-06,
      "loss": 0.0278,
      "step": 317100
    },
    {
      "epoch": 123.15,
      "learning_rate": 3.580064724919094e-06,
      "loss": 0.029,
      "step": 317110
    },
    {
      "epoch": 123.15,
      "learning_rate": 3.5795469255663432e-06,
      "loss": 0.0162,
      "step": 317120
    },
    {
      "epoch": 123.16,
      "learning_rate": 3.5790291262135924e-06,
      "loss": 0.0569,
      "step": 317130
    },
    {
      "epoch": 123.16,
      "learning_rate": 3.5785113268608416e-06,
      "loss": 0.0009,
      "step": 317140
    },
    {
      "epoch": 123.17,
      "learning_rate": 3.577993527508091e-06,
      "loss": 0.0897,
      "step": 317150
    },
    {
      "epoch": 123.17,
      "learning_rate": 3.57747572815534e-06,
      "loss": 0.0237,
      "step": 317160
    },
    {
      "epoch": 123.17,
      "learning_rate": 3.576957928802589e-06,
      "loss": 0.0752,
      "step": 317170
    },
    {
      "epoch": 123.18,
      "learning_rate": 3.5764401294498384e-06,
      "loss": 0.0108,
      "step": 317180
    },
    {
      "epoch": 123.18,
      "learning_rate": 3.5759223300970876e-06,
      "loss": 0.0335,
      "step": 317190
    },
    {
      "epoch": 123.18,
      "learning_rate": 3.575404530744337e-06,
      "loss": 0.0147,
      "step": 317200
    },
    {
      "epoch": 123.19,
      "learning_rate": 3.574886731391586e-06,
      "loss": 0.0361,
      "step": 317210
    },
    {
      "epoch": 123.19,
      "learning_rate": 3.574368932038835e-06,
      "loss": 0.0687,
      "step": 317220
    },
    {
      "epoch": 123.2,
      "learning_rate": 3.5738511326860844e-06,
      "loss": 0.0565,
      "step": 317230
    },
    {
      "epoch": 123.2,
      "learning_rate": 3.5733333333333336e-06,
      "loss": 0.0813,
      "step": 317240
    },
    {
      "epoch": 123.2,
      "learning_rate": 3.5728155339805828e-06,
      "loss": 0.0095,
      "step": 317250
    },
    {
      "epoch": 123.21,
      "learning_rate": 3.572297734627832e-06,
      "loss": 0.0221,
      "step": 317260
    },
    {
      "epoch": 123.21,
      "learning_rate": 3.5717799352750807e-06,
      "loss": 0.1134,
      "step": 317270
    },
    {
      "epoch": 123.22,
      "learning_rate": 3.5712621359223303e-06,
      "loss": 0.0021,
      "step": 317280
    },
    {
      "epoch": 123.22,
      "learning_rate": 3.570744336569579e-06,
      "loss": 0.0702,
      "step": 317290
    },
    {
      "epoch": 123.22,
      "learning_rate": 3.5702265372168287e-06,
      "loss": 0.0088,
      "step": 317300
    },
    {
      "epoch": 123.23,
      "learning_rate": 3.5697087378640784e-06,
      "loss": 0.0288,
      "step": 317310
    },
    {
      "epoch": 123.23,
      "learning_rate": 3.569190938511327e-06,
      "loss": 0.0342,
      "step": 317320
    },
    {
      "epoch": 123.23,
      "learning_rate": 3.5686731391585767e-06,
      "loss": 0.0116,
      "step": 317330
    },
    {
      "epoch": 123.24,
      "learning_rate": 3.5681553398058255e-06,
      "loss": 0.0238,
      "step": 317340
    },
    {
      "epoch": 123.24,
      "learning_rate": 3.567637540453075e-06,
      "loss": 0.0198,
      "step": 317350
    },
    {
      "epoch": 123.25,
      "learning_rate": 3.567119741100324e-06,
      "loss": 0.021,
      "step": 317360
    },
    {
      "epoch": 123.25,
      "learning_rate": 3.5666019417475735e-06,
      "loss": 0.1595,
      "step": 317370
    },
    {
      "epoch": 123.25,
      "learning_rate": 3.5660841423948223e-06,
      "loss": 0.0561,
      "step": 317380
    },
    {
      "epoch": 123.26,
      "learning_rate": 3.565566343042072e-06,
      "loss": 0.0111,
      "step": 317390
    },
    {
      "epoch": 123.26,
      "learning_rate": 3.5650485436893207e-06,
      "loss": 0.0437,
      "step": 317400
    },
    {
      "epoch": 123.27,
      "learning_rate": 3.5645307443365703e-06,
      "loss": 0.0625,
      "step": 317410
    },
    {
      "epoch": 123.27,
      "learning_rate": 3.564012944983819e-06,
      "loss": 0.0165,
      "step": 317420
    },
    {
      "epoch": 123.27,
      "learning_rate": 3.5634951456310683e-06,
      "loss": 0.0003,
      "step": 317430
    },
    {
      "epoch": 123.28,
      "learning_rate": 3.5629773462783175e-06,
      "loss": 0.0901,
      "step": 317440
    },
    {
      "epoch": 123.28,
      "learning_rate": 3.5624595469255667e-06,
      "loss": 0.0743,
      "step": 317450
    },
    {
      "epoch": 123.29,
      "learning_rate": 3.561941747572816e-06,
      "loss": 0.0276,
      "step": 317460
    },
    {
      "epoch": 123.29,
      "learning_rate": 3.561423948220065e-06,
      "loss": 0.0001,
      "step": 317470
    },
    {
      "epoch": 123.29,
      "learning_rate": 3.5609061488673142e-06,
      "loss": 0.0013,
      "step": 317480
    },
    {
      "epoch": 123.3,
      "learning_rate": 3.5603883495145634e-06,
      "loss": 0.111,
      "step": 317490
    },
    {
      "epoch": 123.3,
      "learning_rate": 3.5598705501618126e-06,
      "loss": 0.0225,
      "step": 317500
    },
    {
      "epoch": 123.3,
      "learning_rate": 3.559352750809062e-06,
      "loss": 0.097,
      "step": 317510
    },
    {
      "epoch": 123.31,
      "learning_rate": 3.558834951456311e-06,
      "loss": 0.1888,
      "step": 317520
    },
    {
      "epoch": 123.31,
      "learning_rate": 3.55831715210356e-06,
      "loss": 0.0449,
      "step": 317530
    },
    {
      "epoch": 123.32,
      "learning_rate": 3.5577993527508094e-06,
      "loss": 0.0109,
      "step": 317540
    },
    {
      "epoch": 123.32,
      "learning_rate": 3.5572815533980586e-06,
      "loss": 0.0004,
      "step": 317550
    },
    {
      "epoch": 123.32,
      "learning_rate": 3.556763754045308e-06,
      "loss": 0.0896,
      "step": 317560
    },
    {
      "epoch": 123.33,
      "learning_rate": 3.556245954692557e-06,
      "loss": 0.1412,
      "step": 317570
    },
    {
      "epoch": 123.33,
      "learning_rate": 3.555728155339806e-06,
      "loss": 0.0006,
      "step": 317580
    },
    {
      "epoch": 123.34,
      "learning_rate": 3.5552103559870554e-06,
      "loss": 0.0707,
      "step": 317590
    },
    {
      "epoch": 123.34,
      "learning_rate": 3.554692556634304e-06,
      "loss": 0.1362,
      "step": 317600
    },
    {
      "epoch": 123.34,
      "learning_rate": 3.5541747572815538e-06,
      "loss": 0.0007,
      "step": 317610
    },
    {
      "epoch": 123.35,
      "learning_rate": 3.5536569579288025e-06,
      "loss": 0.0328,
      "step": 317620
    },
    {
      "epoch": 123.35,
      "learning_rate": 3.553139158576052e-06,
      "loss": 0.0694,
      "step": 317630
    },
    {
      "epoch": 123.36,
      "learning_rate": 3.552621359223301e-06,
      "loss": 0.0547,
      "step": 317640
    },
    {
      "epoch": 123.36,
      "learning_rate": 3.5521035598705505e-06,
      "loss": 0.1061,
      "step": 317650
    },
    {
      "epoch": 123.36,
      "learning_rate": 3.5515857605177993e-06,
      "loss": 0.0155,
      "step": 317660
    },
    {
      "epoch": 123.37,
      "learning_rate": 3.551067961165049e-06,
      "loss": 0.0269,
      "step": 317670
    },
    {
      "epoch": 123.37,
      "learning_rate": 3.5505501618122977e-06,
      "loss": 0.0542,
      "step": 317680
    },
    {
      "epoch": 123.37,
      "learning_rate": 3.5500323624595473e-06,
      "loss": 0.0712,
      "step": 317690
    },
    {
      "epoch": 123.38,
      "learning_rate": 3.549514563106796e-06,
      "loss": 0.0159,
      "step": 317700
    },
    {
      "epoch": 123.38,
      "learning_rate": 3.5489967637540457e-06,
      "loss": 0.0453,
      "step": 317710
    },
    {
      "epoch": 123.39,
      "learning_rate": 3.5484789644012945e-06,
      "loss": 0.0125,
      "step": 317720
    },
    {
      "epoch": 123.39,
      "learning_rate": 3.547961165048544e-06,
      "loss": 0.0162,
      "step": 317730
    },
    {
      "epoch": 123.39,
      "learning_rate": 3.547443365695793e-06,
      "loss": 0.0676,
      "step": 317740
    },
    {
      "epoch": 123.4,
      "learning_rate": 3.5469255663430425e-06,
      "loss": 0.0096,
      "step": 317750
    },
    {
      "epoch": 123.4,
      "learning_rate": 3.5464077669902913e-06,
      "loss": 0.0646,
      "step": 317760
    },
    {
      "epoch": 123.41,
      "learning_rate": 3.545889967637541e-06,
      "loss": 0.0001,
      "step": 317770
    },
    {
      "epoch": 123.41,
      "learning_rate": 3.5453721682847896e-06,
      "loss": 0.0344,
      "step": 317780
    },
    {
      "epoch": 123.41,
      "learning_rate": 3.5448543689320393e-06,
      "loss": 0.0432,
      "step": 317790
    },
    {
      "epoch": 123.42,
      "learning_rate": 3.544336569579288e-06,
      "loss": 0.0728,
      "step": 317800
    },
    {
      "epoch": 123.42,
      "learning_rate": 3.5438187702265377e-06,
      "loss": 0.0018,
      "step": 317810
    },
    {
      "epoch": 123.43,
      "learning_rate": 3.5433009708737864e-06,
      "loss": 0.0442,
      "step": 317820
    },
    {
      "epoch": 123.43,
      "learning_rate": 3.542783171521036e-06,
      "loss": 0.0829,
      "step": 317830
    },
    {
      "epoch": 123.43,
      "learning_rate": 3.542265372168285e-06,
      "loss": 0.1556,
      "step": 317840
    },
    {
      "epoch": 123.44,
      "learning_rate": 3.5417475728155344e-06,
      "loss": 0.0182,
      "step": 317850
    },
    {
      "epoch": 123.44,
      "learning_rate": 3.5412297734627836e-06,
      "loss": 0.0107,
      "step": 317860
    },
    {
      "epoch": 123.44,
      "learning_rate": 3.540711974110033e-06,
      "loss": 0.0266,
      "step": 317870
    },
    {
      "epoch": 123.45,
      "learning_rate": 3.540194174757282e-06,
      "loss": 0.017,
      "step": 317880
    },
    {
      "epoch": 123.45,
      "learning_rate": 3.539676375404531e-06,
      "loss": 0.039,
      "step": 317890
    },
    {
      "epoch": 123.46,
      "learning_rate": 3.5391585760517804e-06,
      "loss": 0.0427,
      "step": 317900
    },
    {
      "epoch": 123.46,
      "learning_rate": 3.5386407766990296e-06,
      "loss": 0.0872,
      "step": 317910
    },
    {
      "epoch": 123.46,
      "learning_rate": 3.538122977346279e-06,
      "loss": 0.0233,
      "step": 317920
    },
    {
      "epoch": 123.47,
      "learning_rate": 3.5376051779935276e-06,
      "loss": 0.0078,
      "step": 317930
    },
    {
      "epoch": 123.47,
      "learning_rate": 3.537087378640777e-06,
      "loss": 0.084,
      "step": 317940
    },
    {
      "epoch": 123.48,
      "learning_rate": 3.536569579288026e-06,
      "loss": 0.015,
      "step": 317950
    },
    {
      "epoch": 123.48,
      "learning_rate": 3.5360517799352756e-06,
      "loss": 0.0328,
      "step": 317960
    },
    {
      "epoch": 123.48,
      "learning_rate": 3.5355339805825243e-06,
      "loss": 0.043,
      "step": 317970
    },
    {
      "epoch": 123.49,
      "learning_rate": 3.535016181229774e-06,
      "loss": 0.0176,
      "step": 317980
    },
    {
      "epoch": 123.49,
      "learning_rate": 3.5344983818770227e-06,
      "loss": 0.2075,
      "step": 317990
    },
    {
      "epoch": 123.5,
      "learning_rate": 3.5339805825242724e-06,
      "loss": 0.0588,
      "step": 318000
    },
    {
      "epoch": 123.5,
      "learning_rate": 3.533462783171521e-06,
      "loss": 0.0772,
      "step": 318010
    },
    {
      "epoch": 123.5,
      "learning_rate": 3.5329449838187707e-06,
      "loss": 0.0001,
      "step": 318020
    },
    {
      "epoch": 123.51,
      "learning_rate": 3.5324271844660195e-06,
      "loss": 0.0535,
      "step": 318030
    },
    {
      "epoch": 123.51,
      "learning_rate": 3.531909385113269e-06,
      "loss": 0.0003,
      "step": 318040
    },
    {
      "epoch": 123.51,
      "learning_rate": 3.531391585760518e-06,
      "loss": 0.0499,
      "step": 318050
    },
    {
      "epoch": 123.52,
      "learning_rate": 3.5308737864077675e-06,
      "loss": 0.0365,
      "step": 318060
    },
    {
      "epoch": 123.52,
      "learning_rate": 3.5303559870550163e-06,
      "loss": 0.0564,
      "step": 318070
    },
    {
      "epoch": 123.53,
      "learning_rate": 3.529838187702266e-06,
      "loss": 0.0973,
      "step": 318080
    },
    {
      "epoch": 123.53,
      "learning_rate": 3.5293203883495147e-06,
      "loss": 0.0005,
      "step": 318090
    },
    {
      "epoch": 123.53,
      "learning_rate": 3.5288025889967643e-06,
      "loss": 0.0489,
      "step": 318100
    },
    {
      "epoch": 123.54,
      "learning_rate": 3.528284789644013e-06,
      "loss": 0.0554,
      "step": 318110
    },
    {
      "epoch": 123.54,
      "learning_rate": 3.5277669902912627e-06,
      "loss": 0.0024,
      "step": 318120
    },
    {
      "epoch": 123.55,
      "learning_rate": 3.5272491909385115e-06,
      "loss": 0.0303,
      "step": 318130
    },
    {
      "epoch": 123.55,
      "learning_rate": 3.526731391585761e-06,
      "loss": 0.0162,
      "step": 318140
    },
    {
      "epoch": 123.55,
      "learning_rate": 3.52621359223301e-06,
      "loss": 0.0212,
      "step": 318150
    },
    {
      "epoch": 123.56,
      "learning_rate": 3.5256957928802595e-06,
      "loss": 0.0789,
      "step": 318160
    },
    {
      "epoch": 123.56,
      "learning_rate": 3.5251779935275082e-06,
      "loss": 0.0001,
      "step": 318170
    },
    {
      "epoch": 123.57,
      "learning_rate": 3.524660194174758e-06,
      "loss": 0.0026,
      "step": 318180
    },
    {
      "epoch": 123.57,
      "learning_rate": 3.5241423948220066e-06,
      "loss": 0.0527,
      "step": 318190
    },
    {
      "epoch": 123.57,
      "learning_rate": 3.5236245954692562e-06,
      "loss": 0.0264,
      "step": 318200
    },
    {
      "epoch": 123.58,
      "learning_rate": 3.523106796116505e-06,
      "loss": 0.1864,
      "step": 318210
    },
    {
      "epoch": 123.58,
      "learning_rate": 3.5225889967637546e-06,
      "loss": 0.0587,
      "step": 318220
    },
    {
      "epoch": 123.58,
      "learning_rate": 3.5220711974110034e-06,
      "loss": 0.0845,
      "step": 318230
    },
    {
      "epoch": 123.59,
      "learning_rate": 3.5215533980582526e-06,
      "loss": 0.0119,
      "step": 318240
    },
    {
      "epoch": 123.59,
      "learning_rate": 3.521035598705502e-06,
      "loss": 0.108,
      "step": 318250
    },
    {
      "epoch": 123.6,
      "learning_rate": 3.520517799352751e-06,
      "loss": 0.0413,
      "step": 318260
    },
    {
      "epoch": 123.6,
      "learning_rate": 3.52e-06,
      "loss": 0.0004,
      "step": 318270
    },
    {
      "epoch": 123.6,
      "learning_rate": 3.5194822006472494e-06,
      "loss": 0.0395,
      "step": 318280
    },
    {
      "epoch": 123.61,
      "learning_rate": 3.5189644012944986e-06,
      "loss": 0.0508,
      "step": 318290
    },
    {
      "epoch": 123.61,
      "learning_rate": 3.5184466019417478e-06,
      "loss": 0.142,
      "step": 318300
    },
    {
      "epoch": 123.62,
      "learning_rate": 3.517928802588997e-06,
      "loss": 0.0525,
      "step": 318310
    },
    {
      "epoch": 123.62,
      "learning_rate": 3.517411003236246e-06,
      "loss": 0.0096,
      "step": 318320
    },
    {
      "epoch": 123.62,
      "learning_rate": 3.5168932038834953e-06,
      "loss": 0.0316,
      "step": 318330
    },
    {
      "epoch": 123.63,
      "learning_rate": 3.5163754045307445e-06,
      "loss": 0.062,
      "step": 318340
    },
    {
      "epoch": 123.63,
      "learning_rate": 3.5158576051779937e-06,
      "loss": 0.0132,
      "step": 318350
    },
    {
      "epoch": 123.63,
      "learning_rate": 3.515339805825243e-06,
      "loss": 0.0699,
      "step": 318360
    },
    {
      "epoch": 123.64,
      "learning_rate": 3.514822006472492e-06,
      "loss": 0.0002,
      "step": 318370
    },
    {
      "epoch": 123.64,
      "learning_rate": 3.5143042071197413e-06,
      "loss": 0.1603,
      "step": 318380
    },
    {
      "epoch": 123.65,
      "learning_rate": 3.5137864077669905e-06,
      "loss": 0.0957,
      "step": 318390
    },
    {
      "epoch": 123.65,
      "learning_rate": 3.5132686084142397e-06,
      "loss": 0.0017,
      "step": 318400
    },
    {
      "epoch": 123.65,
      "learning_rate": 3.5127508090614885e-06,
      "loss": 0.0007,
      "step": 318410
    },
    {
      "epoch": 123.66,
      "learning_rate": 3.512233009708738e-06,
      "loss": 0.0312,
      "step": 318420
    },
    {
      "epoch": 123.66,
      "learning_rate": 3.5117152103559877e-06,
      "loss": 0.0444,
      "step": 318430
    },
    {
      "epoch": 123.67,
      "learning_rate": 3.5111974110032365e-06,
      "loss": 0.0668,
      "step": 318440
    },
    {
      "epoch": 123.67,
      "learning_rate": 3.510679611650486e-06,
      "loss": 0.0048,
      "step": 318450
    },
    {
      "epoch": 123.67,
      "learning_rate": 3.510161812297735e-06,
      "loss": 0.0094,
      "step": 318460
    },
    {
      "epoch": 123.68,
      "learning_rate": 3.5096440129449845e-06,
      "loss": 0.0759,
      "step": 318470
    },
    {
      "epoch": 123.68,
      "learning_rate": 3.5091262135922333e-06,
      "loss": 0.0036,
      "step": 318480
    },
    {
      "epoch": 123.69,
      "learning_rate": 3.508608414239483e-06,
      "loss": 0.0075,
      "step": 318490
    },
    {
      "epoch": 123.69,
      "learning_rate": 3.5080906148867317e-06,
      "loss": 0.0186,
      "step": 318500
    },
    {
      "epoch": 123.69,
      "learning_rate": 3.5075728155339813e-06,
      "loss": 0.0545,
      "step": 318510
    },
    {
      "epoch": 123.7,
      "learning_rate": 3.50705501618123e-06,
      "loss": 0.0018,
      "step": 318520
    },
    {
      "epoch": 123.7,
      "learning_rate": 3.5065372168284797e-06,
      "loss": 0.0163,
      "step": 318530
    },
    {
      "epoch": 123.7,
      "learning_rate": 3.5060194174757284e-06,
      "loss": 0.0001,
      "step": 318540
    },
    {
      "epoch": 123.71,
      "learning_rate": 3.505501618122978e-06,
      "loss": 0.0522,
      "step": 318550
    },
    {
      "epoch": 123.71,
      "learning_rate": 3.504983818770227e-06,
      "loss": 0.0415,
      "step": 318560
    },
    {
      "epoch": 123.72,
      "learning_rate": 3.504466019417476e-06,
      "loss": 0.0084,
      "step": 318570
    },
    {
      "epoch": 123.72,
      "learning_rate": 3.503948220064725e-06,
      "loss": 0.0713,
      "step": 318580
    },
    {
      "epoch": 123.72,
      "learning_rate": 3.5034304207119744e-06,
      "loss": 0.0753,
      "step": 318590
    },
    {
      "epoch": 123.73,
      "learning_rate": 3.5029126213592236e-06,
      "loss": 0.0001,
      "step": 318600
    },
    {
      "epoch": 123.73,
      "learning_rate": 3.502394822006473e-06,
      "loss": 0.0213,
      "step": 318610
    },
    {
      "epoch": 123.74,
      "learning_rate": 3.501877022653722e-06,
      "loss": 0.0455,
      "step": 318620
    },
    {
      "epoch": 123.74,
      "learning_rate": 3.501359223300971e-06,
      "loss": 0.0529,
      "step": 318630
    },
    {
      "epoch": 123.74,
      "learning_rate": 3.5008414239482204e-06,
      "loss": 0.0635,
      "step": 318640
    },
    {
      "epoch": 123.75,
      "learning_rate": 3.5003236245954696e-06,
      "loss": 0.0387,
      "step": 318650
    },
    {
      "epoch": 123.75,
      "learning_rate": 3.4998058252427188e-06,
      "loss": 0.026,
      "step": 318660
    },
    {
      "epoch": 123.76,
      "learning_rate": 3.499288025889968e-06,
      "loss": 0.0945,
      "step": 318670
    },
    {
      "epoch": 123.76,
      "learning_rate": 3.498770226537217e-06,
      "loss": 0.0741,
      "step": 318680
    },
    {
      "epoch": 123.76,
      "learning_rate": 3.4982524271844663e-06,
      "loss": 0.0677,
      "step": 318690
    },
    {
      "epoch": 123.77,
      "learning_rate": 3.4977346278317155e-06,
      "loss": 0.0444,
      "step": 318700
    },
    {
      "epoch": 123.77,
      "learning_rate": 3.4972168284789647e-06,
      "loss": 0.0035,
      "step": 318710
    },
    {
      "epoch": 123.77,
      "learning_rate": 3.496699029126214e-06,
      "loss": 0.0592,
      "step": 318720
    },
    {
      "epoch": 123.78,
      "learning_rate": 3.496181229773463e-06,
      "loss": 0.0028,
      "step": 318730
    },
    {
      "epoch": 123.78,
      "learning_rate": 3.495663430420712e-06,
      "loss": 0.0222,
      "step": 318740
    },
    {
      "epoch": 123.79,
      "learning_rate": 3.4951456310679615e-06,
      "loss": 0.0967,
      "step": 318750
    },
    {
      "epoch": 123.79,
      "learning_rate": 3.4946278317152103e-06,
      "loss": 0.0095,
      "step": 318760
    },
    {
      "epoch": 123.79,
      "learning_rate": 3.49411003236246e-06,
      "loss": 0.0759,
      "step": 318770
    },
    {
      "epoch": 123.8,
      "learning_rate": 3.4935922330097087e-06,
      "loss": 0.0267,
      "step": 318780
    },
    {
      "epoch": 123.8,
      "learning_rate": 3.4930744336569583e-06,
      "loss": 0.0633,
      "step": 318790
    },
    {
      "epoch": 123.81,
      "learning_rate": 3.492556634304207e-06,
      "loss": 0.0425,
      "step": 318800
    },
    {
      "epoch": 123.81,
      "learning_rate": 3.4920388349514567e-06,
      "loss": 0.0419,
      "step": 318810
    },
    {
      "epoch": 123.81,
      "learning_rate": 3.4915210355987055e-06,
      "loss": 0.1061,
      "step": 318820
    },
    {
      "epoch": 123.82,
      "learning_rate": 3.491003236245955e-06,
      "loss": 0.0313,
      "step": 318830
    },
    {
      "epoch": 123.82,
      "learning_rate": 3.490485436893204e-06,
      "loss": 0.0105,
      "step": 318840
    },
    {
      "epoch": 123.83,
      "learning_rate": 3.4899676375404535e-06,
      "loss": 0.0735,
      "step": 318850
    },
    {
      "epoch": 123.83,
      "learning_rate": 3.4894498381877022e-06,
      "loss": 0.0757,
      "step": 318860
    },
    {
      "epoch": 123.83,
      "learning_rate": 3.488932038834952e-06,
      "loss": 0.0098,
      "step": 318870
    },
    {
      "epoch": 123.84,
      "learning_rate": 3.4884142394822006e-06,
      "loss": 0.0125,
      "step": 318880
    },
    {
      "epoch": 123.84,
      "learning_rate": 3.4878964401294502e-06,
      "loss": 0.1016,
      "step": 318890
    },
    {
      "epoch": 123.84,
      "learning_rate": 3.487378640776699e-06,
      "loss": 0.0127,
      "step": 318900
    },
    {
      "epoch": 123.85,
      "learning_rate": 3.4868608414239486e-06,
      "loss": 0.0888,
      "step": 318910
    },
    {
      "epoch": 123.85,
      "learning_rate": 3.4863430420711974e-06,
      "loss": 0.0132,
      "step": 318920
    },
    {
      "epoch": 123.86,
      "learning_rate": 3.485825242718447e-06,
      "loss": 0.0527,
      "step": 318930
    },
    {
      "epoch": 123.86,
      "learning_rate": 3.485307443365696e-06,
      "loss": 0.0887,
      "step": 318940
    },
    {
      "epoch": 123.86,
      "learning_rate": 3.4847896440129454e-06,
      "loss": 0.0239,
      "step": 318950
    },
    {
      "epoch": 123.87,
      "learning_rate": 3.484271844660194e-06,
      "loss": 0.1479,
      "step": 318960
    },
    {
      "epoch": 123.87,
      "learning_rate": 3.483754045307444e-06,
      "loss": 0.0782,
      "step": 318970
    },
    {
      "epoch": 123.88,
      "learning_rate": 3.483236245954693e-06,
      "loss": 0.0683,
      "step": 318980
    },
    {
      "epoch": 123.88,
      "learning_rate": 3.482718446601942e-06,
      "loss": 0.0387,
      "step": 318990
    },
    {
      "epoch": 123.88,
      "learning_rate": 3.4822006472491914e-06,
      "loss": 0.1402,
      "step": 319000
    },
    {
      "epoch": 123.89,
      "learning_rate": 3.4816828478964406e-06,
      "loss": 0.0204,
      "step": 319010
    },
    {
      "epoch": 123.89,
      "learning_rate": 3.4811650485436898e-06,
      "loss": 0.0536,
      "step": 319020
    },
    {
      "epoch": 123.9,
      "learning_rate": 3.480647249190939e-06,
      "loss": 0.2267,
      "step": 319030
    },
    {
      "epoch": 123.9,
      "learning_rate": 3.480129449838188e-06,
      "loss": 0.1253,
      "step": 319040
    },
    {
      "epoch": 123.9,
      "learning_rate": 3.4796116504854374e-06,
      "loss": 0.062,
      "step": 319050
    },
    {
      "epoch": 123.91,
      "learning_rate": 3.4790938511326865e-06,
      "loss": 0.0487,
      "step": 319060
    },
    {
      "epoch": 123.91,
      "learning_rate": 3.4785760517799353e-06,
      "loss": 0.0463,
      "step": 319070
    },
    {
      "epoch": 123.91,
      "learning_rate": 3.478058252427185e-06,
      "loss": 0.1364,
      "step": 319080
    },
    {
      "epoch": 123.92,
      "learning_rate": 3.4775404530744337e-06,
      "loss": 0.0268,
      "step": 319090
    },
    {
      "epoch": 123.92,
      "learning_rate": 3.4770226537216833e-06,
      "loss": 0.1199,
      "step": 319100
    },
    {
      "epoch": 123.93,
      "learning_rate": 3.476504854368932e-06,
      "loss": 0.0428,
      "step": 319110
    },
    {
      "epoch": 123.93,
      "learning_rate": 3.4759870550161817e-06,
      "loss": 0.0361,
      "step": 319120
    },
    {
      "epoch": 123.93,
      "learning_rate": 3.4754692556634305e-06,
      "loss": 0.1658,
      "step": 319130
    },
    {
      "epoch": 123.94,
      "learning_rate": 3.47495145631068e-06,
      "loss": 0.0894,
      "step": 319140
    },
    {
      "epoch": 123.94,
      "learning_rate": 3.474433656957929e-06,
      "loss": 0.0214,
      "step": 319150
    },
    {
      "epoch": 123.95,
      "learning_rate": 3.4739158576051785e-06,
      "loss": 0.0004,
      "step": 319160
    },
    {
      "epoch": 123.95,
      "learning_rate": 3.4733980582524273e-06,
      "loss": 0.0109,
      "step": 319170
    },
    {
      "epoch": 123.95,
      "learning_rate": 3.472880258899677e-06,
      "loss": 0.1144,
      "step": 319180
    },
    {
      "epoch": 123.96,
      "learning_rate": 3.4723624595469257e-06,
      "loss": 0.0192,
      "step": 319190
    },
    {
      "epoch": 123.96,
      "learning_rate": 3.4718446601941753e-06,
      "loss": 0.0205,
      "step": 319200
    },
    {
      "epoch": 123.97,
      "learning_rate": 3.471326860841424e-06,
      "loss": 0.0123,
      "step": 319210
    },
    {
      "epoch": 123.97,
      "learning_rate": 3.4708090614886737e-06,
      "loss": 0.0533,
      "step": 319220
    },
    {
      "epoch": 123.97,
      "learning_rate": 3.4702912621359224e-06,
      "loss": 0.115,
      "step": 319230
    },
    {
      "epoch": 123.98,
      "learning_rate": 3.469773462783172e-06,
      "loss": 0.0296,
      "step": 319240
    },
    {
      "epoch": 123.98,
      "learning_rate": 3.469255663430421e-06,
      "loss": 0.0007,
      "step": 319250
    },
    {
      "epoch": 123.98,
      "learning_rate": 3.4687378640776704e-06,
      "loss": 0.0675,
      "step": 319260
    },
    {
      "epoch": 123.99,
      "learning_rate": 3.468220064724919e-06,
      "loss": 0.0077,
      "step": 319270
    },
    {
      "epoch": 123.99,
      "learning_rate": 3.467702265372169e-06,
      "loss": 0.0023,
      "step": 319280
    },
    {
      "epoch": 124.0,
      "learning_rate": 3.4671844660194176e-06,
      "loss": 0.0178,
      "step": 319290
    },
    {
      "epoch": 124.0,
      "learning_rate": 3.4666666666666672e-06,
      "loss": 0.0249,
      "step": 319300
    },
    {
      "epoch": 124.0,
      "eval_accuracy": 0.949656121045392,
      "eval_loss": 0.374478280544281,
      "eval_runtime": 8.2186,
      "eval_samples_per_second": 442.287,
      "eval_steps_per_second": 55.362,
      "step": 319300
    },
    {
      "epoch": 124.0,
      "learning_rate": 3.466148867313916e-06,
      "loss": 0.0547,
      "step": 319310
    },
    {
      "epoch": 124.01,
      "learning_rate": 3.4656310679611656e-06,
      "loss": 0.0917,
      "step": 319320
    },
    {
      "epoch": 124.01,
      "learning_rate": 3.4651132686084144e-06,
      "loss": 0.0216,
      "step": 319330
    },
    {
      "epoch": 124.02,
      "learning_rate": 3.464595469255664e-06,
      "loss": 0.0197,
      "step": 319340
    },
    {
      "epoch": 124.02,
      "learning_rate": 3.4640776699029128e-06,
      "loss": 0.0001,
      "step": 319350
    },
    {
      "epoch": 124.02,
      "learning_rate": 3.4635598705501624e-06,
      "loss": 0.0863,
      "step": 319360
    },
    {
      "epoch": 124.03,
      "learning_rate": 3.463042071197411e-06,
      "loss": 0.1096,
      "step": 319370
    },
    {
      "epoch": 124.03,
      "learning_rate": 3.4625242718446603e-06,
      "loss": 0.0249,
      "step": 319380
    },
    {
      "epoch": 124.03,
      "learning_rate": 3.4620064724919095e-06,
      "loss": 0.1369,
      "step": 319390
    },
    {
      "epoch": 124.04,
      "learning_rate": 3.4614886731391587e-06,
      "loss": 0.067,
      "step": 319400
    },
    {
      "epoch": 124.04,
      "learning_rate": 3.460970873786408e-06,
      "loss": 0.0003,
      "step": 319410
    },
    {
      "epoch": 124.05,
      "learning_rate": 3.460453074433657e-06,
      "loss": 0.0479,
      "step": 319420
    },
    {
      "epoch": 124.05,
      "learning_rate": 3.4599352750809063e-06,
      "loss": 0.019,
      "step": 319430
    },
    {
      "epoch": 124.05,
      "learning_rate": 3.4594174757281555e-06,
      "loss": 0.0188,
      "step": 319440
    },
    {
      "epoch": 124.06,
      "learning_rate": 3.4588996763754047e-06,
      "loss": 0.0267,
      "step": 319450
    },
    {
      "epoch": 124.06,
      "learning_rate": 3.458381877022654e-06,
      "loss": 0.0855,
      "step": 319460
    },
    {
      "epoch": 124.07,
      "learning_rate": 3.457864077669903e-06,
      "loss": 0.0107,
      "step": 319470
    },
    {
      "epoch": 124.07,
      "learning_rate": 3.4573462783171523e-06,
      "loss": 0.0086,
      "step": 319480
    },
    {
      "epoch": 124.07,
      "learning_rate": 3.4568284789644015e-06,
      "loss": 0.0535,
      "step": 319490
    },
    {
      "epoch": 124.08,
      "learning_rate": 3.4563106796116507e-06,
      "loss": 0.0216,
      "step": 319500
    },
    {
      "epoch": 124.08,
      "learning_rate": 3.4557928802589e-06,
      "loss": 0.1156,
      "step": 319510
    },
    {
      "epoch": 124.09,
      "learning_rate": 3.455275080906149e-06,
      "loss": 0.0081,
      "step": 319520
    },
    {
      "epoch": 124.09,
      "learning_rate": 3.4547572815533983e-06,
      "loss": 0.0687,
      "step": 319530
    },
    {
      "epoch": 124.09,
      "learning_rate": 3.4542394822006475e-06,
      "loss": 0.0574,
      "step": 319540
    },
    {
      "epoch": 124.1,
      "learning_rate": 3.453721682847897e-06,
      "loss": 0.0308,
      "step": 319550
    },
    {
      "epoch": 124.1,
      "learning_rate": 3.453203883495146e-06,
      "loss": 0.0266,
      "step": 319560
    },
    {
      "epoch": 124.1,
      "learning_rate": 3.4526860841423955e-06,
      "loss": 0.0805,
      "step": 319570
    },
    {
      "epoch": 124.11,
      "learning_rate": 3.4521682847896442e-06,
      "loss": 0.0002,
      "step": 319580
    },
    {
      "epoch": 124.11,
      "learning_rate": 3.451650485436894e-06,
      "loss": 0.0162,
      "step": 319590
    },
    {
      "epoch": 124.12,
      "learning_rate": 3.4511326860841426e-06,
      "loss": 0.1626,
      "step": 319600
    },
    {
      "epoch": 124.12,
      "learning_rate": 3.4506148867313922e-06,
      "loss": 0.0567,
      "step": 319610
    },
    {
      "epoch": 124.12,
      "learning_rate": 3.450097087378641e-06,
      "loss": 0.001,
      "step": 319620
    },
    {
      "epoch": 124.13,
      "learning_rate": 3.4495792880258906e-06,
      "loss": 0.0685,
      "step": 319630
    },
    {
      "epoch": 124.13,
      "learning_rate": 3.4490614886731394e-06,
      "loss": 0.0169,
      "step": 319640
    },
    {
      "epoch": 124.14,
      "learning_rate": 3.448543689320389e-06,
      "loss": 0.1052,
      "step": 319650
    },
    {
      "epoch": 124.14,
      "learning_rate": 3.448025889967638e-06,
      "loss": 0.0589,
      "step": 319660
    },
    {
      "epoch": 124.14,
      "learning_rate": 3.4475080906148874e-06,
      "loss": 0.0113,
      "step": 319670
    },
    {
      "epoch": 124.15,
      "learning_rate": 3.446990291262136e-06,
      "loss": 0.074,
      "step": 319680
    },
    {
      "epoch": 124.15,
      "learning_rate": 3.446472491909386e-06,
      "loss": 0.0452,
      "step": 319690
    },
    {
      "epoch": 124.16,
      "learning_rate": 3.4459546925566346e-06,
      "loss": 0.0344,
      "step": 319700
    },
    {
      "epoch": 124.16,
      "learning_rate": 3.4454368932038838e-06,
      "loss": 0.0136,
      "step": 319710
    },
    {
      "epoch": 124.16,
      "learning_rate": 3.444919093851133e-06,
      "loss": 0.0001,
      "step": 319720
    },
    {
      "epoch": 124.17,
      "learning_rate": 3.444401294498382e-06,
      "loss": 0.0251,
      "step": 319730
    },
    {
      "epoch": 124.17,
      "learning_rate": 3.4438834951456314e-06,
      "loss": 0.0336,
      "step": 319740
    },
    {
      "epoch": 124.17,
      "learning_rate": 3.4433656957928805e-06,
      "loss": 0.07,
      "step": 319750
    },
    {
      "epoch": 124.18,
      "learning_rate": 3.4428478964401297e-06,
      "loss": 0.02,
      "step": 319760
    },
    {
      "epoch": 124.18,
      "learning_rate": 3.442330097087379e-06,
      "loss": 0.0474,
      "step": 319770
    },
    {
      "epoch": 124.19,
      "learning_rate": 3.441812297734628e-06,
      "loss": 0.0027,
      "step": 319780
    },
    {
      "epoch": 124.19,
      "learning_rate": 3.4412944983818773e-06,
      "loss": 0.1075,
      "step": 319790
    },
    {
      "epoch": 124.19,
      "learning_rate": 3.4407766990291265e-06,
      "loss": 0.078,
      "step": 319800
    },
    {
      "epoch": 124.2,
      "learning_rate": 3.4402588996763757e-06,
      "loss": 0.0882,
      "step": 319810
    },
    {
      "epoch": 124.2,
      "learning_rate": 3.439741100323625e-06,
      "loss": 0.0886,
      "step": 319820
    },
    {
      "epoch": 124.21,
      "learning_rate": 3.439223300970874e-06,
      "loss": 0.0208,
      "step": 319830
    },
    {
      "epoch": 124.21,
      "learning_rate": 3.4387055016181233e-06,
      "loss": 0.0111,
      "step": 319840
    },
    {
      "epoch": 124.21,
      "learning_rate": 3.4381877022653725e-06,
      "loss": 0.0096,
      "step": 319850
    },
    {
      "epoch": 124.22,
      "learning_rate": 3.4376699029126217e-06,
      "loss": 0.009,
      "step": 319860
    },
    {
      "epoch": 124.22,
      "learning_rate": 3.437152103559871e-06,
      "loss": 0.0862,
      "step": 319870
    },
    {
      "epoch": 124.23,
      "learning_rate": 3.4366343042071197e-06,
      "loss": 0.0001,
      "step": 319880
    },
    {
      "epoch": 124.23,
      "learning_rate": 3.4361165048543693e-06,
      "loss": 0.0777,
      "step": 319890
    },
    {
      "epoch": 124.23,
      "learning_rate": 3.435598705501618e-06,
      "loss": 0.1183,
      "step": 319900
    },
    {
      "epoch": 124.24,
      "learning_rate": 3.4350809061488677e-06,
      "loss": 0.0005,
      "step": 319910
    },
    {
      "epoch": 124.24,
      "learning_rate": 3.4345631067961164e-06,
      "loss": 0.014,
      "step": 319920
    },
    {
      "epoch": 124.24,
      "learning_rate": 3.434045307443366e-06,
      "loss": 0.0024,
      "step": 319930
    },
    {
      "epoch": 124.25,
      "learning_rate": 3.433527508090615e-06,
      "loss": 0.0023,
      "step": 319940
    },
    {
      "epoch": 124.25,
      "learning_rate": 3.4330097087378644e-06,
      "loss": 0.0527,
      "step": 319950
    },
    {
      "epoch": 124.26,
      "learning_rate": 3.432491909385113e-06,
      "loss": 0.0414,
      "step": 319960
    },
    {
      "epoch": 124.26,
      "learning_rate": 3.431974110032363e-06,
      "loss": 0.014,
      "step": 319970
    },
    {
      "epoch": 124.26,
      "learning_rate": 3.4314563106796116e-06,
      "loss": 0.0456,
      "step": 319980
    },
    {
      "epoch": 124.27,
      "learning_rate": 3.4309385113268612e-06,
      "loss": 0.0304,
      "step": 319990
    },
    {
      "epoch": 124.27,
      "learning_rate": 3.43042071197411e-06,
      "loss": 0.033,
      "step": 320000
    },
    {
      "epoch": 124.28,
      "learning_rate": 3.4299029126213596e-06,
      "loss": 0.0101,
      "step": 320010
    },
    {
      "epoch": 124.28,
      "learning_rate": 3.4293851132686084e-06,
      "loss": 0.012,
      "step": 320020
    },
    {
      "epoch": 124.28,
      "learning_rate": 3.428867313915858e-06,
      "loss": 0.0148,
      "step": 320030
    },
    {
      "epoch": 124.29,
      "learning_rate": 3.4283495145631068e-06,
      "loss": 0.0297,
      "step": 320040
    },
    {
      "epoch": 124.29,
      "learning_rate": 3.4278317152103564e-06,
      "loss": 0.0123,
      "step": 320050
    },
    {
      "epoch": 124.3,
      "learning_rate": 3.427313915857605e-06,
      "loss": 0.0015,
      "step": 320060
    },
    {
      "epoch": 124.3,
      "learning_rate": 3.4267961165048548e-06,
      "loss": 0.0101,
      "step": 320070
    },
    {
      "epoch": 124.3,
      "learning_rate": 3.4262783171521035e-06,
      "loss": 0.0647,
      "step": 320080
    },
    {
      "epoch": 124.31,
      "learning_rate": 3.425760517799353e-06,
      "loss": 0.074,
      "step": 320090
    },
    {
      "epoch": 124.31,
      "learning_rate": 3.4252427184466024e-06,
      "loss": 0.115,
      "step": 320100
    },
    {
      "epoch": 124.31,
      "learning_rate": 3.4247249190938515e-06,
      "loss": 0.0588,
      "step": 320110
    },
    {
      "epoch": 124.32,
      "learning_rate": 3.4242071197411007e-06,
      "loss": 0.0013,
      "step": 320120
    },
    {
      "epoch": 124.32,
      "learning_rate": 3.42368932038835e-06,
      "loss": 0.0559,
      "step": 320130
    },
    {
      "epoch": 124.33,
      "learning_rate": 3.423171521035599e-06,
      "loss": 0.0351,
      "step": 320140
    },
    {
      "epoch": 124.33,
      "learning_rate": 3.4226537216828483e-06,
      "loss": 0.0612,
      "step": 320150
    },
    {
      "epoch": 124.33,
      "learning_rate": 3.4221359223300975e-06,
      "loss": 0.0215,
      "step": 320160
    },
    {
      "epoch": 124.34,
      "learning_rate": 3.4216181229773467e-06,
      "loss": 0.0764,
      "step": 320170
    },
    {
      "epoch": 124.34,
      "learning_rate": 3.421100323624596e-06,
      "loss": 0.0405,
      "step": 320180
    },
    {
      "epoch": 124.35,
      "learning_rate": 3.420582524271845e-06,
      "loss": 0.0486,
      "step": 320190
    },
    {
      "epoch": 124.35,
      "learning_rate": 3.4200647249190943e-06,
      "loss": 0.0172,
      "step": 320200
    },
    {
      "epoch": 124.35,
      "learning_rate": 3.419546925566343e-06,
      "loss": 0.0335,
      "step": 320210
    },
    {
      "epoch": 124.36,
      "learning_rate": 3.4190291262135927e-06,
      "loss": 0.0969,
      "step": 320220
    },
    {
      "epoch": 124.36,
      "learning_rate": 3.4185113268608415e-06,
      "loss": 0.0925,
      "step": 320230
    },
    {
      "epoch": 124.37,
      "learning_rate": 3.417993527508091e-06,
      "loss": 0.0152,
      "step": 320240
    },
    {
      "epoch": 124.37,
      "learning_rate": 3.41747572815534e-06,
      "loss": 0.0004,
      "step": 320250
    },
    {
      "epoch": 124.37,
      "learning_rate": 3.4169579288025895e-06,
      "loss": 0.0975,
      "step": 320260
    },
    {
      "epoch": 124.38,
      "learning_rate": 3.4164401294498382e-06,
      "loss": 0.1043,
      "step": 320270
    },
    {
      "epoch": 124.38,
      "learning_rate": 3.415922330097088e-06,
      "loss": 0.0034,
      "step": 320280
    },
    {
      "epoch": 124.38,
      "learning_rate": 3.4154045307443366e-06,
      "loss": 0.0053,
      "step": 320290
    },
    {
      "epoch": 124.39,
      "learning_rate": 3.4148867313915862e-06,
      "loss": 0.0576,
      "step": 320300
    },
    {
      "epoch": 124.39,
      "learning_rate": 3.414368932038835e-06,
      "loss": 0.1428,
      "step": 320310
    },
    {
      "epoch": 124.4,
      "learning_rate": 3.4138511326860846e-06,
      "loss": 0.0374,
      "step": 320320
    },
    {
      "epoch": 124.4,
      "learning_rate": 3.4133333333333334e-06,
      "loss": 0.0278,
      "step": 320330
    },
    {
      "epoch": 124.4,
      "learning_rate": 3.412815533980583e-06,
      "loss": 0.0231,
      "step": 320340
    },
    {
      "epoch": 124.41,
      "learning_rate": 3.412297734627832e-06,
      "loss": 0.0917,
      "step": 320350
    },
    {
      "epoch": 124.41,
      "learning_rate": 3.4117799352750814e-06,
      "loss": 0.0413,
      "step": 320360
    },
    {
      "epoch": 124.42,
      "learning_rate": 3.41126213592233e-06,
      "loss": 0.0244,
      "step": 320370
    },
    {
      "epoch": 124.42,
      "learning_rate": 3.41074433656958e-06,
      "loss": 0.0259,
      "step": 320380
    },
    {
      "epoch": 124.42,
      "learning_rate": 3.4102265372168286e-06,
      "loss": 0.0002,
      "step": 320390
    },
    {
      "epoch": 124.43,
      "learning_rate": 3.409708737864078e-06,
      "loss": 0.0015,
      "step": 320400
    },
    {
      "epoch": 124.43,
      "learning_rate": 3.409190938511327e-06,
      "loss": 0.1017,
      "step": 320410
    },
    {
      "epoch": 124.43,
      "learning_rate": 3.4086731391585766e-06,
      "loss": 0.0257,
      "step": 320420
    },
    {
      "epoch": 124.44,
      "learning_rate": 3.4081553398058254e-06,
      "loss": 0.0107,
      "step": 320430
    },
    {
      "epoch": 124.44,
      "learning_rate": 3.407637540453075e-06,
      "loss": 0.0607,
      "step": 320440
    },
    {
      "epoch": 124.45,
      "learning_rate": 3.4071197411003237e-06,
      "loss": 0.0398,
      "step": 320450
    },
    {
      "epoch": 124.45,
      "learning_rate": 3.4066019417475734e-06,
      "loss": 0.1649,
      "step": 320460
    },
    {
      "epoch": 124.45,
      "learning_rate": 3.406084142394822e-06,
      "loss": 0.0442,
      "step": 320470
    },
    {
      "epoch": 124.46,
      "learning_rate": 3.4055663430420717e-06,
      "loss": 0.0238,
      "step": 320480
    },
    {
      "epoch": 124.46,
      "learning_rate": 3.4050485436893205e-06,
      "loss": 0.0103,
      "step": 320490
    },
    {
      "epoch": 124.47,
      "learning_rate": 3.40453074433657e-06,
      "loss": 0.0001,
      "step": 320500
    },
    {
      "epoch": 124.47,
      "learning_rate": 3.404012944983819e-06,
      "loss": 0.0298,
      "step": 320510
    },
    {
      "epoch": 124.47,
      "learning_rate": 3.403495145631068e-06,
      "loss": 0.0389,
      "step": 320520
    },
    {
      "epoch": 124.48,
      "learning_rate": 3.4029773462783173e-06,
      "loss": 0.0818,
      "step": 320530
    },
    {
      "epoch": 124.48,
      "learning_rate": 3.4024595469255665e-06,
      "loss": 0.0299,
      "step": 320540
    },
    {
      "epoch": 124.49,
      "learning_rate": 3.4019417475728157e-06,
      "loss": 0.0004,
      "step": 320550
    },
    {
      "epoch": 124.49,
      "learning_rate": 3.401423948220065e-06,
      "loss": 0.1104,
      "step": 320560
    },
    {
      "epoch": 124.49,
      "learning_rate": 3.400906148867314e-06,
      "loss": 0.045,
      "step": 320570
    },
    {
      "epoch": 124.5,
      "learning_rate": 3.4003883495145633e-06,
      "loss": 0.014,
      "step": 320580
    },
    {
      "epoch": 124.5,
      "learning_rate": 3.3998705501618125e-06,
      "loss": 0.0135,
      "step": 320590
    },
    {
      "epoch": 124.5,
      "learning_rate": 3.3993527508090617e-06,
      "loss": 0.0342,
      "step": 320600
    },
    {
      "epoch": 124.51,
      "learning_rate": 3.398834951456311e-06,
      "loss": 0.0204,
      "step": 320610
    },
    {
      "epoch": 124.51,
      "learning_rate": 3.39831715210356e-06,
      "loss": 0.029,
      "step": 320620
    },
    {
      "epoch": 124.52,
      "learning_rate": 3.3977993527508092e-06,
      "loss": 0.0709,
      "step": 320630
    },
    {
      "epoch": 124.52,
      "learning_rate": 3.3972815533980584e-06,
      "loss": 0.0016,
      "step": 320640
    },
    {
      "epoch": 124.52,
      "learning_rate": 3.3967637540453076e-06,
      "loss": 0.0376,
      "step": 320650
    },
    {
      "epoch": 124.53,
      "learning_rate": 3.396245954692557e-06,
      "loss": 0.0155,
      "step": 320660
    },
    {
      "epoch": 124.53,
      "learning_rate": 3.3957281553398064e-06,
      "loss": 0.1189,
      "step": 320670
    },
    {
      "epoch": 124.54,
      "learning_rate": 3.3952103559870552e-06,
      "loss": 0.0385,
      "step": 320680
    },
    {
      "epoch": 124.54,
      "learning_rate": 3.394692556634305e-06,
      "loss": 0.01,
      "step": 320690
    },
    {
      "epoch": 124.54,
      "learning_rate": 3.3941747572815536e-06,
      "loss": 0.0827,
      "step": 320700
    },
    {
      "epoch": 124.55,
      "learning_rate": 3.3936569579288032e-06,
      "loss": 0.0003,
      "step": 320710
    },
    {
      "epoch": 124.55,
      "learning_rate": 3.393139158576052e-06,
      "loss": 0.0001,
      "step": 320720
    },
    {
      "epoch": 124.56,
      "learning_rate": 3.3926213592233016e-06,
      "loss": 0.0503,
      "step": 320730
    },
    {
      "epoch": 124.56,
      "learning_rate": 3.3921035598705504e-06,
      "loss": 0.059,
      "step": 320740
    },
    {
      "epoch": 124.56,
      "learning_rate": 3.3915857605178e-06,
      "loss": 0.0001,
      "step": 320750
    },
    {
      "epoch": 124.57,
      "learning_rate": 3.3910679611650488e-06,
      "loss": 0.0709,
      "step": 320760
    },
    {
      "epoch": 124.57,
      "learning_rate": 3.3905501618122984e-06,
      "loss": 0.0277,
      "step": 320770
    },
    {
      "epoch": 124.57,
      "learning_rate": 3.390032362459547e-06,
      "loss": 0.0226,
      "step": 320780
    },
    {
      "epoch": 124.58,
      "learning_rate": 3.3895145631067968e-06,
      "loss": 0.2468,
      "step": 320790
    },
    {
      "epoch": 124.58,
      "learning_rate": 3.3889967637540455e-06,
      "loss": 0.0101,
      "step": 320800
    },
    {
      "epoch": 124.59,
      "learning_rate": 3.388478964401295e-06,
      "loss": 0.0213,
      "step": 320810
    },
    {
      "epoch": 124.59,
      "learning_rate": 3.387961165048544e-06,
      "loss": 0.0903,
      "step": 320820
    },
    {
      "epoch": 124.59,
      "learning_rate": 3.3874433656957936e-06,
      "loss": 0.0518,
      "step": 320830
    },
    {
      "epoch": 124.6,
      "learning_rate": 3.3869255663430423e-06,
      "loss": 0.0121,
      "step": 320840
    },
    {
      "epoch": 124.6,
      "learning_rate": 3.3864077669902915e-06,
      "loss": 0.2143,
      "step": 320850
    },
    {
      "epoch": 124.61,
      "learning_rate": 3.3858899676375407e-06,
      "loss": 0.009,
      "step": 320860
    },
    {
      "epoch": 124.61,
      "learning_rate": 3.38537216828479e-06,
      "loss": 0.0498,
      "step": 320870
    },
    {
      "epoch": 124.61,
      "learning_rate": 3.384854368932039e-06,
      "loss": 0.0527,
      "step": 320880
    },
    {
      "epoch": 124.62,
      "learning_rate": 3.3843365695792883e-06,
      "loss": 0.0207,
      "step": 320890
    },
    {
      "epoch": 124.62,
      "learning_rate": 3.3838187702265375e-06,
      "loss": 0.0429,
      "step": 320900
    },
    {
      "epoch": 124.63,
      "learning_rate": 3.3833009708737867e-06,
      "loss": 0.0477,
      "step": 320910
    },
    {
      "epoch": 124.63,
      "learning_rate": 3.382783171521036e-06,
      "loss": 0.0296,
      "step": 320920
    },
    {
      "epoch": 124.63,
      "learning_rate": 3.382265372168285e-06,
      "loss": 0.0028,
      "step": 320930
    },
    {
      "epoch": 124.64,
      "learning_rate": 3.3817475728155343e-06,
      "loss": 0.1008,
      "step": 320940
    },
    {
      "epoch": 124.64,
      "learning_rate": 3.3812297734627835e-06,
      "loss": 0.0509,
      "step": 320950
    },
    {
      "epoch": 124.64,
      "learning_rate": 3.3807119741100327e-06,
      "loss": 0.0525,
      "step": 320960
    },
    {
      "epoch": 124.65,
      "learning_rate": 3.380194174757282e-06,
      "loss": 0.0145,
      "step": 320970
    },
    {
      "epoch": 124.65,
      "learning_rate": 3.379676375404531e-06,
      "loss": 0.0732,
      "step": 320980
    },
    {
      "epoch": 124.66,
      "learning_rate": 3.3791585760517802e-06,
      "loss": 0.0128,
      "step": 320990
    },
    {
      "epoch": 124.66,
      "learning_rate": 3.3786407766990294e-06,
      "loss": 0.0216,
      "step": 321000
    },
    {
      "epoch": 124.66,
      "learning_rate": 3.3781229773462786e-06,
      "loss": 0.0091,
      "step": 321010
    },
    {
      "epoch": 124.67,
      "learning_rate": 3.3776051779935274e-06,
      "loss": 0.0005,
      "step": 321020
    },
    {
      "epoch": 124.67,
      "learning_rate": 3.377087378640777e-06,
      "loss": 0.0364,
      "step": 321030
    },
    {
      "epoch": 124.68,
      "learning_rate": 3.376569579288026e-06,
      "loss": 0.0116,
      "step": 321040
    },
    {
      "epoch": 124.68,
      "learning_rate": 3.3760517799352754e-06,
      "loss": 0.025,
      "step": 321050
    },
    {
      "epoch": 124.68,
      "learning_rate": 3.375533980582524e-06,
      "loss": 0.0177,
      "step": 321060
    },
    {
      "epoch": 124.69,
      "learning_rate": 3.375016181229774e-06,
      "loss": 0.1125,
      "step": 321070
    },
    {
      "epoch": 124.69,
      "learning_rate": 3.3744983818770226e-06,
      "loss": 0.0901,
      "step": 321080
    },
    {
      "epoch": 124.7,
      "learning_rate": 3.373980582524272e-06,
      "loss": 0.1173,
      "step": 321090
    },
    {
      "epoch": 124.7,
      "learning_rate": 3.373462783171521e-06,
      "loss": 0.0187,
      "step": 321100
    },
    {
      "epoch": 124.7,
      "learning_rate": 3.3729449838187706e-06,
      "loss": 0.0001,
      "step": 321110
    },
    {
      "epoch": 124.71,
      "learning_rate": 3.3724271844660193e-06,
      "loss": 0.0948,
      "step": 321120
    },
    {
      "epoch": 124.71,
      "learning_rate": 3.371909385113269e-06,
      "loss": 0.0004,
      "step": 321130
    },
    {
      "epoch": 124.71,
      "learning_rate": 3.3713915857605177e-06,
      "loss": 0.012,
      "step": 321140
    },
    {
      "epoch": 124.72,
      "learning_rate": 3.3708737864077674e-06,
      "loss": 0.0324,
      "step": 321150
    },
    {
      "epoch": 124.72,
      "learning_rate": 3.370355987055016e-06,
      "loss": 0.0644,
      "step": 321160
    },
    {
      "epoch": 124.73,
      "learning_rate": 3.3698381877022657e-06,
      "loss": 0.0949,
      "step": 321170
    },
    {
      "epoch": 124.73,
      "learning_rate": 3.3693203883495145e-06,
      "loss": 0.0143,
      "step": 321180
    },
    {
      "epoch": 124.73,
      "learning_rate": 3.368802588996764e-06,
      "loss": 0.0183,
      "step": 321190
    },
    {
      "epoch": 124.74,
      "learning_rate": 3.368284789644013e-06,
      "loss": 0.1957,
      "step": 321200
    },
    {
      "epoch": 124.74,
      "learning_rate": 3.3677669902912625e-06,
      "loss": 0.0198,
      "step": 321210
    },
    {
      "epoch": 124.75,
      "learning_rate": 3.3672491909385113e-06,
      "loss": 0.0886,
      "step": 321220
    },
    {
      "epoch": 124.75,
      "learning_rate": 3.366731391585761e-06,
      "loss": 0.0116,
      "step": 321230
    },
    {
      "epoch": 124.75,
      "learning_rate": 3.36621359223301e-06,
      "loss": 0.0197,
      "step": 321240
    },
    {
      "epoch": 124.76,
      "learning_rate": 3.3656957928802593e-06,
      "loss": 0.1219,
      "step": 321250
    },
    {
      "epoch": 124.76,
      "learning_rate": 3.3651779935275085e-06,
      "loss": 0.0228,
      "step": 321260
    },
    {
      "epoch": 124.77,
      "learning_rate": 3.3646601941747577e-06,
      "loss": 0.1601,
      "step": 321270
    },
    {
      "epoch": 124.77,
      "learning_rate": 3.364142394822007e-06,
      "loss": 0.0879,
      "step": 321280
    },
    {
      "epoch": 124.77,
      "learning_rate": 3.363624595469256e-06,
      "loss": 0.0059,
      "step": 321290
    },
    {
      "epoch": 124.78,
      "learning_rate": 3.3631067961165053e-06,
      "loss": 0.0092,
      "step": 321300
    },
    {
      "epoch": 124.78,
      "learning_rate": 3.3625889967637545e-06,
      "loss": 0.0163,
      "step": 321310
    },
    {
      "epoch": 124.78,
      "learning_rate": 3.3620711974110037e-06,
      "loss": 0.1129,
      "step": 321320
    },
    {
      "epoch": 124.79,
      "learning_rate": 3.361553398058253e-06,
      "loss": 0.0921,
      "step": 321330
    },
    {
      "epoch": 124.79,
      "learning_rate": 3.361035598705502e-06,
      "loss": 0.0138,
      "step": 321340
    },
    {
      "epoch": 124.8,
      "learning_rate": 3.360517799352751e-06,
      "loss": 0.0241,
      "step": 321350
    },
    {
      "epoch": 124.8,
      "learning_rate": 3.3600000000000004e-06,
      "loss": 0.0897,
      "step": 321360
    },
    {
      "epoch": 124.8,
      "learning_rate": 3.3594822006472492e-06,
      "loss": 0.0926,
      "step": 321370
    },
    {
      "epoch": 124.81,
      "learning_rate": 3.358964401294499e-06,
      "loss": 0.0194,
      "step": 321380
    },
    {
      "epoch": 124.81,
      "learning_rate": 3.3584466019417476e-06,
      "loss": 0.0346,
      "step": 321390
    },
    {
      "epoch": 124.82,
      "learning_rate": 3.3579288025889972e-06,
      "loss": 0.013,
      "step": 321400
    },
    {
      "epoch": 124.82,
      "learning_rate": 3.357411003236246e-06,
      "loss": 0.0465,
      "step": 321410
    },
    {
      "epoch": 124.82,
      "learning_rate": 3.3568932038834956e-06,
      "loss": 0.0881,
      "step": 321420
    },
    {
      "epoch": 124.83,
      "learning_rate": 3.3563754045307444e-06,
      "loss": 0.0258,
      "step": 321430
    },
    {
      "epoch": 124.83,
      "learning_rate": 3.355857605177994e-06,
      "loss": 0.0103,
      "step": 321440
    },
    {
      "epoch": 124.83,
      "learning_rate": 3.3553398058252428e-06,
      "loss": 0.0001,
      "step": 321450
    },
    {
      "epoch": 124.84,
      "learning_rate": 3.3548220064724924e-06,
      "loss": 0.0014,
      "step": 321460
    },
    {
      "epoch": 124.84,
      "learning_rate": 3.354304207119741e-06,
      "loss": 0.0119,
      "step": 321470
    },
    {
      "epoch": 124.85,
      "learning_rate": 3.3537864077669908e-06,
      "loss": 0.0002,
      "step": 321480
    },
    {
      "epoch": 124.85,
      "learning_rate": 3.3532686084142395e-06,
      "loss": 0.0099,
      "step": 321490
    },
    {
      "epoch": 124.85,
      "learning_rate": 3.352750809061489e-06,
      "loss": 0.0825,
      "step": 321500
    },
    {
      "epoch": 124.86,
      "learning_rate": 3.352233009708738e-06,
      "loss": 0.0041,
      "step": 321510
    },
    {
      "epoch": 124.86,
      "learning_rate": 3.3517152103559876e-06,
      "loss": 0.0835,
      "step": 321520
    },
    {
      "epoch": 124.87,
      "learning_rate": 3.3511974110032363e-06,
      "loss": 0.0426,
      "step": 321530
    },
    {
      "epoch": 124.87,
      "learning_rate": 3.350679611650486e-06,
      "loss": 0.0388,
      "step": 321540
    },
    {
      "epoch": 124.87,
      "learning_rate": 3.3501618122977347e-06,
      "loss": 0.0327,
      "step": 321550
    },
    {
      "epoch": 124.88,
      "learning_rate": 3.3496440129449843e-06,
      "loss": 0.0848,
      "step": 321560
    },
    {
      "epoch": 124.88,
      "learning_rate": 3.349126213592233e-06,
      "loss": 0.0054,
      "step": 321570
    },
    {
      "epoch": 124.89,
      "learning_rate": 3.3486084142394827e-06,
      "loss": 0.0048,
      "step": 321580
    },
    {
      "epoch": 124.89,
      "learning_rate": 3.3480906148867315e-06,
      "loss": 0.0522,
      "step": 321590
    },
    {
      "epoch": 124.89,
      "learning_rate": 3.347572815533981e-06,
      "loss": 0.0559,
      "step": 321600
    },
    {
      "epoch": 124.9,
      "learning_rate": 3.34705501618123e-06,
      "loss": 0.0936,
      "step": 321610
    },
    {
      "epoch": 124.9,
      "learning_rate": 3.3465372168284795e-06,
      "loss": 0.0648,
      "step": 321620
    },
    {
      "epoch": 124.9,
      "learning_rate": 3.3460194174757283e-06,
      "loss": 0.0392,
      "step": 321630
    },
    {
      "epoch": 124.91,
      "learning_rate": 3.345501618122978e-06,
      "loss": 0.0803,
      "step": 321640
    },
    {
      "epoch": 124.91,
      "learning_rate": 3.3449838187702267e-06,
      "loss": 0.1087,
      "step": 321650
    },
    {
      "epoch": 124.92,
      "learning_rate": 3.344466019417476e-06,
      "loss": 0.0371,
      "step": 321660
    },
    {
      "epoch": 124.92,
      "learning_rate": 3.343948220064725e-06,
      "loss": 0.0482,
      "step": 321670
    },
    {
      "epoch": 124.92,
      "learning_rate": 3.3434304207119742e-06,
      "loss": 0.0185,
      "step": 321680
    },
    {
      "epoch": 124.93,
      "learning_rate": 3.3429126213592234e-06,
      "loss": 0.0186,
      "step": 321690
    },
    {
      "epoch": 124.93,
      "learning_rate": 3.3423948220064726e-06,
      "loss": 0.0579,
      "step": 321700
    },
    {
      "epoch": 124.94,
      "learning_rate": 3.341877022653722e-06,
      "loss": 0.0005,
      "step": 321710
    },
    {
      "epoch": 124.94,
      "learning_rate": 3.341359223300971e-06,
      "loss": 0.0598,
      "step": 321720
    },
    {
      "epoch": 124.94,
      "learning_rate": 3.3408414239482202e-06,
      "loss": 0.0558,
      "step": 321730
    },
    {
      "epoch": 124.95,
      "learning_rate": 3.3403236245954694e-06,
      "loss": 0.0115,
      "step": 321740
    },
    {
      "epoch": 124.95,
      "learning_rate": 3.3398058252427186e-06,
      "loss": 0.0104,
      "step": 321750
    },
    {
      "epoch": 124.96,
      "learning_rate": 3.339288025889968e-06,
      "loss": 0.0143,
      "step": 321760
    },
    {
      "epoch": 124.96,
      "learning_rate": 3.338770226537217e-06,
      "loss": 0.0218,
      "step": 321770
    },
    {
      "epoch": 124.96,
      "learning_rate": 3.338252427184466e-06,
      "loss": 0.0554,
      "step": 321780
    },
    {
      "epoch": 124.97,
      "learning_rate": 3.337734627831716e-06,
      "loss": 0.0744,
      "step": 321790
    },
    {
      "epoch": 124.97,
      "learning_rate": 3.3372168284789646e-06,
      "loss": 0.1341,
      "step": 321800
    },
    {
      "epoch": 124.97,
      "learning_rate": 3.336699029126214e-06,
      "loss": 0.2297,
      "step": 321810
    },
    {
      "epoch": 124.98,
      "learning_rate": 3.336181229773463e-06,
      "loss": 0.1113,
      "step": 321820
    },
    {
      "epoch": 124.98,
      "learning_rate": 3.3356634304207126e-06,
      "loss": 0.101,
      "step": 321830
    },
    {
      "epoch": 124.99,
      "learning_rate": 3.3351456310679614e-06,
      "loss": 0.0107,
      "step": 321840
    },
    {
      "epoch": 124.99,
      "learning_rate": 3.334627831715211e-06,
      "loss": 0.0276,
      "step": 321850
    },
    {
      "epoch": 124.99,
      "learning_rate": 3.3341100323624597e-06,
      "loss": 0.0107,
      "step": 321860
    },
    {
      "epoch": 125.0,
      "learning_rate": 3.3335922330097094e-06,
      "loss": 0.0002,
      "step": 321870
    },
    {
      "epoch": 125.0,
      "eval_accuracy": 0.951856946354883,
      "eval_loss": 0.37443777918815613,
      "eval_runtime": 8.2887,
      "eval_samples_per_second": 438.547,
      "eval_steps_per_second": 54.894,
      "step": 321875
    },
    {
      "epoch": 125.0,
      "learning_rate": 3.333074433656958e-06,
      "loss": 0.0311,
      "step": 321880
    },
    {
      "epoch": 125.01,
      "learning_rate": 3.3325566343042078e-06,
      "loss": 0.0095,
      "step": 321890
    },
    {
      "epoch": 125.01,
      "learning_rate": 3.3320388349514565e-06,
      "loss": 0.0511,
      "step": 321900
    },
    {
      "epoch": 125.01,
      "learning_rate": 3.331521035598706e-06,
      "loss": 0.0693,
      "step": 321910
    },
    {
      "epoch": 125.02,
      "learning_rate": 3.331003236245955e-06,
      "loss": 0.0426,
      "step": 321920
    },
    {
      "epoch": 125.02,
      "learning_rate": 3.3304854368932045e-06,
      "loss": 0.0984,
      "step": 321930
    },
    {
      "epoch": 125.03,
      "learning_rate": 3.3299676375404533e-06,
      "loss": 0.0761,
      "step": 321940
    },
    {
      "epoch": 125.03,
      "learning_rate": 3.329449838187703e-06,
      "loss": 0.0699,
      "step": 321950
    },
    {
      "epoch": 125.03,
      "learning_rate": 3.3289320388349517e-06,
      "loss": 0.0137,
      "step": 321960
    },
    {
      "epoch": 125.04,
      "learning_rate": 3.3284142394822013e-06,
      "loss": 0.009,
      "step": 321970
    },
    {
      "epoch": 125.04,
      "learning_rate": 3.32789644012945e-06,
      "loss": 0.0012,
      "step": 321980
    },
    {
      "epoch": 125.04,
      "learning_rate": 3.3273786407766993e-06,
      "loss": 0.1338,
      "step": 321990
    },
    {
      "epoch": 125.05,
      "learning_rate": 3.3268608414239485e-06,
      "loss": 0.0227,
      "step": 322000
    },
    {
      "epoch": 125.05,
      "learning_rate": 3.3263430420711977e-06,
      "loss": 0.101,
      "step": 322010
    },
    {
      "epoch": 125.06,
      "learning_rate": 3.325825242718447e-06,
      "loss": 0.0002,
      "step": 322020
    },
    {
      "epoch": 125.06,
      "learning_rate": 3.325307443365696e-06,
      "loss": 0.1301,
      "step": 322030
    },
    {
      "epoch": 125.06,
      "learning_rate": 3.3247896440129452e-06,
      "loss": 0.0492,
      "step": 322040
    },
    {
      "epoch": 125.07,
      "learning_rate": 3.3242718446601944e-06,
      "loss": 0.0068,
      "step": 322050
    },
    {
      "epoch": 125.07,
      "learning_rate": 3.3237540453074436e-06,
      "loss": 0.0001,
      "step": 322060
    },
    {
      "epoch": 125.08,
      "learning_rate": 3.323236245954693e-06,
      "loss": 0.0583,
      "step": 322070
    },
    {
      "epoch": 125.08,
      "learning_rate": 3.322718446601942e-06,
      "loss": 0.0497,
      "step": 322080
    },
    {
      "epoch": 125.08,
      "learning_rate": 3.3222006472491912e-06,
      "loss": 0.0299,
      "step": 322090
    },
    {
      "epoch": 125.09,
      "learning_rate": 3.3216828478964404e-06,
      "loss": 0.0039,
      "step": 322100
    },
    {
      "epoch": 125.09,
      "learning_rate": 3.3211650485436896e-06,
      "loss": 0.1666,
      "step": 322110
    },
    {
      "epoch": 125.1,
      "learning_rate": 3.320647249190939e-06,
      "loss": 0.0359,
      "step": 322120
    },
    {
      "epoch": 125.1,
      "learning_rate": 3.320129449838188e-06,
      "loss": 0.0209,
      "step": 322130
    },
    {
      "epoch": 125.1,
      "learning_rate": 3.319611650485437e-06,
      "loss": 0.0191,
      "step": 322140
    },
    {
      "epoch": 125.11,
      "learning_rate": 3.3190938511326864e-06,
      "loss": 0.0001,
      "step": 322150
    },
    {
      "epoch": 125.11,
      "learning_rate": 3.318576051779935e-06,
      "loss": 0.0005,
      "step": 322160
    },
    {
      "epoch": 125.11,
      "learning_rate": 3.3180582524271848e-06,
      "loss": 0.0688,
      "step": 322170
    },
    {
      "epoch": 125.12,
      "learning_rate": 3.3175404530744335e-06,
      "loss": 0.0696,
      "step": 322180
    },
    {
      "epoch": 125.12,
      "learning_rate": 3.317022653721683e-06,
      "loss": 0.0099,
      "step": 322190
    },
    {
      "epoch": 125.13,
      "learning_rate": 3.316504854368932e-06,
      "loss": 0.1323,
      "step": 322200
    },
    {
      "epoch": 125.13,
      "learning_rate": 3.3159870550161816e-06,
      "loss": 0.0144,
      "step": 322210
    },
    {
      "epoch": 125.13,
      "learning_rate": 3.3154692556634303e-06,
      "loss": 0.0017,
      "step": 322220
    },
    {
      "epoch": 125.14,
      "learning_rate": 3.31495145631068e-06,
      "loss": 0.0241,
      "step": 322230
    },
    {
      "epoch": 125.14,
      "learning_rate": 3.3144336569579287e-06,
      "loss": 0.0219,
      "step": 322240
    },
    {
      "epoch": 125.15,
      "learning_rate": 3.3139158576051783e-06,
      "loss": 0.0247,
      "step": 322250
    },
    {
      "epoch": 125.15,
      "learning_rate": 3.313398058252427e-06,
      "loss": 0.008,
      "step": 322260
    },
    {
      "epoch": 125.15,
      "learning_rate": 3.3128802588996767e-06,
      "loss": 0.0747,
      "step": 322270
    },
    {
      "epoch": 125.16,
      "learning_rate": 3.3123624595469255e-06,
      "loss": 0.0002,
      "step": 322280
    },
    {
      "epoch": 125.16,
      "learning_rate": 3.311844660194175e-06,
      "loss": 0.0892,
      "step": 322290
    },
    {
      "epoch": 125.17,
      "learning_rate": 3.311326860841424e-06,
      "loss": 0.0219,
      "step": 322300
    },
    {
      "epoch": 125.17,
      "learning_rate": 3.3108090614886735e-06,
      "loss": 0.1232,
      "step": 322310
    },
    {
      "epoch": 125.17,
      "learning_rate": 3.3102912621359223e-06,
      "loss": 0.0347,
      "step": 322320
    },
    {
      "epoch": 125.18,
      "learning_rate": 3.309773462783172e-06,
      "loss": 0.0192,
      "step": 322330
    },
    {
      "epoch": 125.18,
      "learning_rate": 3.3092556634304207e-06,
      "loss": 0.0221,
      "step": 322340
    },
    {
      "epoch": 125.18,
      "learning_rate": 3.3087378640776703e-06,
      "loss": 0.0831,
      "step": 322350
    },
    {
      "epoch": 125.19,
      "learning_rate": 3.3082200647249195e-06,
      "loss": 0.0622,
      "step": 322360
    },
    {
      "epoch": 125.19,
      "learning_rate": 3.3077022653721687e-06,
      "loss": 0.0713,
      "step": 322370
    },
    {
      "epoch": 125.2,
      "learning_rate": 3.307184466019418e-06,
      "loss": 0.1009,
      "step": 322380
    },
    {
      "epoch": 125.2,
      "learning_rate": 3.306666666666667e-06,
      "loss": 0.0388,
      "step": 322390
    },
    {
      "epoch": 125.2,
      "learning_rate": 3.3061488673139162e-06,
      "loss": 0.0733,
      "step": 322400
    },
    {
      "epoch": 125.21,
      "learning_rate": 3.3056310679611654e-06,
      "loss": 0.0374,
      "step": 322410
    },
    {
      "epoch": 125.21,
      "learning_rate": 3.3051132686084146e-06,
      "loss": 0.0921,
      "step": 322420
    },
    {
      "epoch": 125.22,
      "learning_rate": 3.304595469255664e-06,
      "loss": 0.0149,
      "step": 322430
    },
    {
      "epoch": 125.22,
      "learning_rate": 3.304077669902913e-06,
      "loss": 0.0023,
      "step": 322440
    },
    {
      "epoch": 125.22,
      "learning_rate": 3.3035598705501622e-06,
      "loss": 0.0305,
      "step": 322450
    },
    {
      "epoch": 125.23,
      "learning_rate": 3.3030420711974114e-06,
      "loss": 0.0217,
      "step": 322460
    },
    {
      "epoch": 125.23,
      "learning_rate": 3.3025242718446606e-06,
      "loss": 0.0772,
      "step": 322470
    },
    {
      "epoch": 125.23,
      "learning_rate": 3.30200647249191e-06,
      "loss": 0.0085,
      "step": 322480
    },
    {
      "epoch": 125.24,
      "learning_rate": 3.3014886731391586e-06,
      "loss": 0.0343,
      "step": 322490
    },
    {
      "epoch": 125.24,
      "learning_rate": 3.300970873786408e-06,
      "loss": 0.0383,
      "step": 322500
    },
    {
      "epoch": 125.25,
      "learning_rate": 3.300453074433657e-06,
      "loss": 0.0465,
      "step": 322510
    },
    {
      "epoch": 125.25,
      "learning_rate": 3.2999352750809066e-06,
      "loss": 0.0196,
      "step": 322520
    },
    {
      "epoch": 125.25,
      "learning_rate": 3.2994174757281554e-06,
      "loss": 0.0546,
      "step": 322530
    },
    {
      "epoch": 125.26,
      "learning_rate": 3.298899676375405e-06,
      "loss": 0.0006,
      "step": 322540
    },
    {
      "epoch": 125.26,
      "learning_rate": 3.2983818770226537e-06,
      "loss": 0.1354,
      "step": 322550
    },
    {
      "epoch": 125.27,
      "learning_rate": 3.2978640776699034e-06,
      "loss": 0.0154,
      "step": 322560
    },
    {
      "epoch": 125.27,
      "learning_rate": 3.297346278317152e-06,
      "loss": 0.1533,
      "step": 322570
    },
    {
      "epoch": 125.27,
      "learning_rate": 3.2968284789644017e-06,
      "loss": 0.0067,
      "step": 322580
    },
    {
      "epoch": 125.28,
      "learning_rate": 3.2963106796116505e-06,
      "loss": 0.0155,
      "step": 322590
    },
    {
      "epoch": 125.28,
      "learning_rate": 3.2957928802589e-06,
      "loss": 0.0001,
      "step": 322600
    },
    {
      "epoch": 125.29,
      "learning_rate": 3.295275080906149e-06,
      "loss": 0.0023,
      "step": 322610
    },
    {
      "epoch": 125.29,
      "learning_rate": 3.2947572815533985e-06,
      "loss": 0.0213,
      "step": 322620
    },
    {
      "epoch": 125.29,
      "learning_rate": 3.2942394822006473e-06,
      "loss": 0.0168,
      "step": 322630
    },
    {
      "epoch": 125.3,
      "learning_rate": 3.293721682847897e-06,
      "loss": 0.0305,
      "step": 322640
    },
    {
      "epoch": 125.3,
      "learning_rate": 3.2932038834951457e-06,
      "loss": 0.0766,
      "step": 322650
    },
    {
      "epoch": 125.3,
      "learning_rate": 3.2926860841423953e-06,
      "loss": 0.0482,
      "step": 322660
    },
    {
      "epoch": 125.31,
      "learning_rate": 3.292168284789644e-06,
      "loss": 0.0498,
      "step": 322670
    },
    {
      "epoch": 125.31,
      "learning_rate": 3.2916504854368937e-06,
      "loss": 0.0905,
      "step": 322680
    },
    {
      "epoch": 125.32,
      "learning_rate": 3.2911326860841425e-06,
      "loss": 0.0299,
      "step": 322690
    },
    {
      "epoch": 125.32,
      "learning_rate": 3.290614886731392e-06,
      "loss": 0.033,
      "step": 322700
    },
    {
      "epoch": 125.32,
      "learning_rate": 3.290097087378641e-06,
      "loss": 0.0786,
      "step": 322710
    },
    {
      "epoch": 125.33,
      "learning_rate": 3.2895792880258905e-06,
      "loss": 0.0746,
      "step": 322720
    },
    {
      "epoch": 125.33,
      "learning_rate": 3.2890614886731392e-06,
      "loss": 0.0629,
      "step": 322730
    },
    {
      "epoch": 125.34,
      "learning_rate": 3.288543689320389e-06,
      "loss": 0.0205,
      "step": 322740
    },
    {
      "epoch": 125.34,
      "learning_rate": 3.2880258899676376e-06,
      "loss": 0.0256,
      "step": 322750
    },
    {
      "epoch": 125.34,
      "learning_rate": 3.2875080906148873e-06,
      "loss": 0.0003,
      "step": 322760
    },
    {
      "epoch": 125.35,
      "learning_rate": 3.286990291262136e-06,
      "loss": 0.0028,
      "step": 322770
    },
    {
      "epoch": 125.35,
      "learning_rate": 3.2864724919093856e-06,
      "loss": 0.0375,
      "step": 322780
    },
    {
      "epoch": 125.36,
      "learning_rate": 3.2859546925566344e-06,
      "loss": 0.1092,
      "step": 322790
    },
    {
      "epoch": 125.36,
      "learning_rate": 3.2854368932038836e-06,
      "loss": 0.0797,
      "step": 322800
    },
    {
      "epoch": 125.36,
      "learning_rate": 3.284919093851133e-06,
      "loss": 0.0299,
      "step": 322810
    },
    {
      "epoch": 125.37,
      "learning_rate": 3.284401294498382e-06,
      "loss": 0.1172,
      "step": 322820
    },
    {
      "epoch": 125.37,
      "learning_rate": 3.283883495145631e-06,
      "loss": 0.0207,
      "step": 322830
    },
    {
      "epoch": 125.37,
      "learning_rate": 3.2833656957928804e-06,
      "loss": 0.0239,
      "step": 322840
    },
    {
      "epoch": 125.38,
      "learning_rate": 3.2828478964401296e-06,
      "loss": 0.0259,
      "step": 322850
    },
    {
      "epoch": 125.38,
      "learning_rate": 3.2823300970873788e-06,
      "loss": 0.0265,
      "step": 322860
    },
    {
      "epoch": 125.39,
      "learning_rate": 3.281812297734628e-06,
      "loss": 0.0913,
      "step": 322870
    },
    {
      "epoch": 125.39,
      "learning_rate": 3.281294498381877e-06,
      "loss": 0.0006,
      "step": 322880
    },
    {
      "epoch": 125.39,
      "learning_rate": 3.2807766990291264e-06,
      "loss": 0.0774,
      "step": 322890
    },
    {
      "epoch": 125.4,
      "learning_rate": 3.2802588996763756e-06,
      "loss": 0.0783,
      "step": 322900
    },
    {
      "epoch": 125.4,
      "learning_rate": 3.279741100323625e-06,
      "loss": 0.0057,
      "step": 322910
    },
    {
      "epoch": 125.41,
      "learning_rate": 3.279223300970874e-06,
      "loss": 0.1359,
      "step": 322920
    },
    {
      "epoch": 125.41,
      "learning_rate": 3.2787055016181236e-06,
      "loss": 0.036,
      "step": 322930
    },
    {
      "epoch": 125.41,
      "learning_rate": 3.2781877022653723e-06,
      "loss": 0.1364,
      "step": 322940
    },
    {
      "epoch": 125.42,
      "learning_rate": 3.277669902912622e-06,
      "loss": 0.0091,
      "step": 322950
    },
    {
      "epoch": 125.42,
      "learning_rate": 3.2771521035598707e-06,
      "loss": 0.0384,
      "step": 322960
    },
    {
      "epoch": 125.43,
      "learning_rate": 3.2766343042071203e-06,
      "loss": 0.1239,
      "step": 322970
    },
    {
      "epoch": 125.43,
      "learning_rate": 3.276116504854369e-06,
      "loss": 0.0007,
      "step": 322980
    },
    {
      "epoch": 125.43,
      "learning_rate": 3.2755987055016187e-06,
      "loss": 0.0073,
      "step": 322990
    },
    {
      "epoch": 125.44,
      "learning_rate": 3.2750809061488675e-06,
      "loss": 0.0464,
      "step": 323000
    },
    {
      "epoch": 125.44,
      "learning_rate": 3.274563106796117e-06,
      "loss": 0.0006,
      "step": 323010
    },
    {
      "epoch": 125.44,
      "learning_rate": 3.274045307443366e-06,
      "loss": 0.1693,
      "step": 323020
    },
    {
      "epoch": 125.45,
      "learning_rate": 3.2735275080906155e-06,
      "loss": 0.0003,
      "step": 323030
    },
    {
      "epoch": 125.45,
      "learning_rate": 3.2730097087378643e-06,
      "loss": 0.1203,
      "step": 323040
    },
    {
      "epoch": 125.46,
      "learning_rate": 3.272491909385114e-06,
      "loss": 0.0001,
      "step": 323050
    },
    {
      "epoch": 125.46,
      "learning_rate": 3.2719741100323627e-06,
      "loss": 0.0544,
      "step": 323060
    },
    {
      "epoch": 125.46,
      "learning_rate": 3.2714563106796123e-06,
      "loss": 0.1535,
      "step": 323070
    },
    {
      "epoch": 125.47,
      "learning_rate": 3.270938511326861e-06,
      "loss": 0.0001,
      "step": 323080
    },
    {
      "epoch": 125.47,
      "learning_rate": 3.2704207119741107e-06,
      "loss": 0.0948,
      "step": 323090
    },
    {
      "epoch": 125.48,
      "learning_rate": 3.2699029126213594e-06,
      "loss": 0.035,
      "step": 323100
    },
    {
      "epoch": 125.48,
      "learning_rate": 3.269385113268609e-06,
      "loss": 0.0005,
      "step": 323110
    },
    {
      "epoch": 125.48,
      "learning_rate": 3.268867313915858e-06,
      "loss": 0.0272,
      "step": 323120
    },
    {
      "epoch": 125.49,
      "learning_rate": 3.268349514563107e-06,
      "loss": 0.1522,
      "step": 323130
    },
    {
      "epoch": 125.49,
      "learning_rate": 3.2678317152103562e-06,
      "loss": 0.0283,
      "step": 323140
    },
    {
      "epoch": 125.5,
      "learning_rate": 3.2673139158576054e-06,
      "loss": 0.0539,
      "step": 323150
    },
    {
      "epoch": 125.5,
      "learning_rate": 3.2667961165048546e-06,
      "loss": 0.0314,
      "step": 323160
    },
    {
      "epoch": 125.5,
      "learning_rate": 3.266278317152104e-06,
      "loss": 0.1392,
      "step": 323170
    },
    {
      "epoch": 125.51,
      "learning_rate": 3.265760517799353e-06,
      "loss": 0.0142,
      "step": 323180
    },
    {
      "epoch": 125.51,
      "learning_rate": 3.265242718446602e-06,
      "loss": 0.0277,
      "step": 323190
    },
    {
      "epoch": 125.51,
      "learning_rate": 3.2647249190938514e-06,
      "loss": 0.0691,
      "step": 323200
    },
    {
      "epoch": 125.52,
      "learning_rate": 3.2642071197411006e-06,
      "loss": 0.0042,
      "step": 323210
    },
    {
      "epoch": 125.52,
      "learning_rate": 3.2636893203883498e-06,
      "loss": 0.0083,
      "step": 323220
    },
    {
      "epoch": 125.53,
      "learning_rate": 3.263171521035599e-06,
      "loss": 0.008,
      "step": 323230
    },
    {
      "epoch": 125.53,
      "learning_rate": 3.262653721682848e-06,
      "loss": 0.0298,
      "step": 323240
    },
    {
      "epoch": 125.53,
      "learning_rate": 3.2621359223300974e-06,
      "loss": 0.0326,
      "step": 323250
    },
    {
      "epoch": 125.54,
      "learning_rate": 3.2616181229773466e-06,
      "loss": 0.0096,
      "step": 323260
    },
    {
      "epoch": 125.54,
      "learning_rate": 3.2611003236245957e-06,
      "loss": 0.0194,
      "step": 323270
    },
    {
      "epoch": 125.55,
      "learning_rate": 3.260582524271845e-06,
      "loss": 0.0973,
      "step": 323280
    },
    {
      "epoch": 125.55,
      "learning_rate": 3.260064724919094e-06,
      "loss": 0.0001,
      "step": 323290
    },
    {
      "epoch": 125.55,
      "learning_rate": 3.259546925566343e-06,
      "loss": 0.0932,
      "step": 323300
    },
    {
      "epoch": 125.56,
      "learning_rate": 3.2590291262135925e-06,
      "loss": 0.0225,
      "step": 323310
    },
    {
      "epoch": 125.56,
      "learning_rate": 3.2585113268608413e-06,
      "loss": 0.1337,
      "step": 323320
    },
    {
      "epoch": 125.57,
      "learning_rate": 3.257993527508091e-06,
      "loss": 0.0657,
      "step": 323330
    },
    {
      "epoch": 125.57,
      "learning_rate": 3.2574757281553397e-06,
      "loss": 0.0006,
      "step": 323340
    },
    {
      "epoch": 125.57,
      "learning_rate": 3.2569579288025893e-06,
      "loss": 0.0318,
      "step": 323350
    },
    {
      "epoch": 125.58,
      "learning_rate": 3.256440129449838e-06,
      "loss": 0.0376,
      "step": 323360
    },
    {
      "epoch": 125.58,
      "learning_rate": 3.2559223300970877e-06,
      "loss": 0.041,
      "step": 323370
    },
    {
      "epoch": 125.58,
      "learning_rate": 3.2554045307443365e-06,
      "loss": 0.0667,
      "step": 323380
    },
    {
      "epoch": 125.59,
      "learning_rate": 3.254886731391586e-06,
      "loss": 0.0001,
      "step": 323390
    },
    {
      "epoch": 125.59,
      "learning_rate": 3.254368932038835e-06,
      "loss": 0.0079,
      "step": 323400
    },
    {
      "epoch": 125.6,
      "learning_rate": 3.2538511326860845e-06,
      "loss": 0.1019,
      "step": 323410
    },
    {
      "epoch": 125.6,
      "learning_rate": 3.2533333333333332e-06,
      "loss": 0.0812,
      "step": 323420
    },
    {
      "epoch": 125.6,
      "learning_rate": 3.252815533980583e-06,
      "loss": 0.022,
      "step": 323430
    },
    {
      "epoch": 125.61,
      "learning_rate": 3.2522977346278316e-06,
      "loss": 0.0002,
      "step": 323440
    },
    {
      "epoch": 125.61,
      "learning_rate": 3.2517799352750812e-06,
      "loss": 0.0229,
      "step": 323450
    },
    {
      "epoch": 125.62,
      "learning_rate": 3.25126213592233e-06,
      "loss": 0.0109,
      "step": 323460
    },
    {
      "epoch": 125.62,
      "learning_rate": 3.2507443365695796e-06,
      "loss": 0.0002,
      "step": 323470
    },
    {
      "epoch": 125.62,
      "learning_rate": 3.250226537216829e-06,
      "loss": 0.0003,
      "step": 323480
    },
    {
      "epoch": 125.63,
      "learning_rate": 3.249708737864078e-06,
      "loss": 0.0199,
      "step": 323490
    },
    {
      "epoch": 125.63,
      "learning_rate": 3.2491909385113272e-06,
      "loss": 0.1282,
      "step": 323500
    },
    {
      "epoch": 125.63,
      "learning_rate": 3.2486731391585764e-06,
      "loss": 0.0101,
      "step": 323510
    },
    {
      "epoch": 125.64,
      "learning_rate": 3.2481553398058256e-06,
      "loss": 0.0038,
      "step": 323520
    },
    {
      "epoch": 125.64,
      "learning_rate": 3.247637540453075e-06,
      "loss": 0.0297,
      "step": 323530
    },
    {
      "epoch": 125.65,
      "learning_rate": 3.247119741100324e-06,
      "loss": 0.0535,
      "step": 323540
    },
    {
      "epoch": 125.65,
      "learning_rate": 3.246601941747573e-06,
      "loss": 0.0698,
      "step": 323550
    },
    {
      "epoch": 125.65,
      "learning_rate": 3.2460841423948224e-06,
      "loss": 0.0283,
      "step": 323560
    },
    {
      "epoch": 125.66,
      "learning_rate": 3.2455663430420716e-06,
      "loss": 0.0977,
      "step": 323570
    },
    {
      "epoch": 125.66,
      "learning_rate": 3.2450485436893208e-06,
      "loss": 0.0287,
      "step": 323580
    },
    {
      "epoch": 125.67,
      "learning_rate": 3.24453074433657e-06,
      "loss": 0.0197,
      "step": 323590
    },
    {
      "epoch": 125.67,
      "learning_rate": 3.244012944983819e-06,
      "loss": 0.014,
      "step": 323600
    },
    {
      "epoch": 125.67,
      "learning_rate": 3.2434951456310684e-06,
      "loss": 0.007,
      "step": 323610
    },
    {
      "epoch": 125.68,
      "learning_rate": 3.2429773462783176e-06,
      "loss": 0.0354,
      "step": 323620
    },
    {
      "epoch": 125.68,
      "learning_rate": 3.2424595469255663e-06,
      "loss": 0.1047,
      "step": 323630
    },
    {
      "epoch": 125.69,
      "learning_rate": 3.241941747572816e-06,
      "loss": 0.0001,
      "step": 323640
    },
    {
      "epoch": 125.69,
      "learning_rate": 3.2414239482200647e-06,
      "loss": 0.0966,
      "step": 323650
    },
    {
      "epoch": 125.69,
      "learning_rate": 3.2409061488673143e-06,
      "loss": 0.0194,
      "step": 323660
    },
    {
      "epoch": 125.7,
      "learning_rate": 3.240388349514563e-06,
      "loss": 0.0423,
      "step": 323670
    },
    {
      "epoch": 125.7,
      "learning_rate": 3.2398705501618127e-06,
      "loss": 0.1132,
      "step": 323680
    },
    {
      "epoch": 125.7,
      "learning_rate": 3.2393527508090615e-06,
      "loss": 0.0565,
      "step": 323690
    },
    {
      "epoch": 125.71,
      "learning_rate": 3.238834951456311e-06,
      "loss": 0.003,
      "step": 323700
    },
    {
      "epoch": 125.71,
      "learning_rate": 3.23831715210356e-06,
      "loss": 0.0417,
      "step": 323710
    },
    {
      "epoch": 125.72,
      "learning_rate": 3.2377993527508095e-06,
      "loss": 0.0266,
      "step": 323720
    },
    {
      "epoch": 125.72,
      "learning_rate": 3.2372815533980583e-06,
      "loss": 0.198,
      "step": 323730
    },
    {
      "epoch": 125.72,
      "learning_rate": 3.236763754045308e-06,
      "loss": 0.0242,
      "step": 323740
    },
    {
      "epoch": 125.73,
      "learning_rate": 3.2362459546925567e-06,
      "loss": 0.0733,
      "step": 323750
    },
    {
      "epoch": 125.73,
      "learning_rate": 3.2357281553398063e-06,
      "loss": 0.0632,
      "step": 323760
    },
    {
      "epoch": 125.74,
      "learning_rate": 3.235210355987055e-06,
      "loss": 0.0994,
      "step": 323770
    },
    {
      "epoch": 125.74,
      "learning_rate": 3.2346925566343047e-06,
      "loss": 0.0661,
      "step": 323780
    },
    {
      "epoch": 125.74,
      "learning_rate": 3.2341747572815534e-06,
      "loss": 0.0193,
      "step": 323790
    },
    {
      "epoch": 125.75,
      "learning_rate": 3.233656957928803e-06,
      "loss": 0.0661,
      "step": 323800
    },
    {
      "epoch": 125.75,
      "learning_rate": 3.233139158576052e-06,
      "loss": 0.0005,
      "step": 323810
    },
    {
      "epoch": 125.76,
      "learning_rate": 3.2326213592233014e-06,
      "loss": 0.2176,
      "step": 323820
    },
    {
      "epoch": 125.76,
      "learning_rate": 3.2321035598705502e-06,
      "loss": 0.0089,
      "step": 323830
    },
    {
      "epoch": 125.76,
      "learning_rate": 3.2315857605178e-06,
      "loss": 0.0976,
      "step": 323840
    },
    {
      "epoch": 125.77,
      "learning_rate": 3.2310679611650486e-06,
      "loss": 0.0162,
      "step": 323850
    },
    {
      "epoch": 125.77,
      "learning_rate": 3.2305501618122982e-06,
      "loss": 0.1407,
      "step": 323860
    },
    {
      "epoch": 125.77,
      "learning_rate": 3.230032362459547e-06,
      "loss": 0.0002,
      "step": 323870
    },
    {
      "epoch": 125.78,
      "learning_rate": 3.2295145631067966e-06,
      "loss": 0.1029,
      "step": 323880
    },
    {
      "epoch": 125.78,
      "learning_rate": 3.2289967637540454e-06,
      "loss": 0.1826,
      "step": 323890
    },
    {
      "epoch": 125.79,
      "learning_rate": 3.228478964401295e-06,
      "loss": 0.0394,
      "step": 323900
    },
    {
      "epoch": 125.79,
      "learning_rate": 3.2279611650485438e-06,
      "loss": 0.0006,
      "step": 323910
    },
    {
      "epoch": 125.79,
      "learning_rate": 3.2274433656957934e-06,
      "loss": 0.1683,
      "step": 323920
    },
    {
      "epoch": 125.8,
      "learning_rate": 3.226925566343042e-06,
      "loss": 0.0235,
      "step": 323930
    },
    {
      "epoch": 125.8,
      "learning_rate": 3.2264077669902914e-06,
      "loss": 0.0003,
      "step": 323940
    },
    {
      "epoch": 125.81,
      "learning_rate": 3.2258899676375406e-06,
      "loss": 0.0387,
      "step": 323950
    },
    {
      "epoch": 125.81,
      "learning_rate": 3.2253721682847897e-06,
      "loss": 0.0579,
      "step": 323960
    },
    {
      "epoch": 125.81,
      "learning_rate": 3.224854368932039e-06,
      "loss": 0.0877,
      "step": 323970
    },
    {
      "epoch": 125.82,
      "learning_rate": 3.224336569579288e-06,
      "loss": 0.0006,
      "step": 323980
    },
    {
      "epoch": 125.82,
      "learning_rate": 3.2238187702265373e-06,
      "loss": 0.0683,
      "step": 323990
    },
    {
      "epoch": 125.83,
      "learning_rate": 3.2233009708737865e-06,
      "loss": 0.124,
      "step": 324000
    },
    {
      "epoch": 125.83,
      "learning_rate": 3.2227831715210357e-06,
      "loss": 0.0352,
      "step": 324010
    },
    {
      "epoch": 125.83,
      "learning_rate": 3.222265372168285e-06,
      "loss": 0.0737,
      "step": 324020
    },
    {
      "epoch": 125.84,
      "learning_rate": 3.2217475728155345e-06,
      "loss": 0.0362,
      "step": 324030
    },
    {
      "epoch": 125.84,
      "learning_rate": 3.2212297734627833e-06,
      "loss": 0.071,
      "step": 324040
    },
    {
      "epoch": 125.84,
      "learning_rate": 3.220711974110033e-06,
      "loss": 0.0612,
      "step": 324050
    },
    {
      "epoch": 125.85,
      "learning_rate": 3.2201941747572817e-06,
      "loss": 0.0244,
      "step": 324060
    },
    {
      "epoch": 125.85,
      "learning_rate": 3.2196763754045313e-06,
      "loss": 0.0348,
      "step": 324070
    },
    {
      "epoch": 125.86,
      "learning_rate": 3.21915857605178e-06,
      "loss": 0.0179,
      "step": 324080
    },
    {
      "epoch": 125.86,
      "learning_rate": 3.2186407766990297e-06,
      "loss": 0.0723,
      "step": 324090
    },
    {
      "epoch": 125.86,
      "learning_rate": 3.2181229773462785e-06,
      "loss": 0.0946,
      "step": 324100
    },
    {
      "epoch": 125.87,
      "learning_rate": 3.217605177993528e-06,
      "loss": 0.0001,
      "step": 324110
    },
    {
      "epoch": 125.87,
      "learning_rate": 3.217087378640777e-06,
      "loss": 0.0024,
      "step": 324120
    },
    {
      "epoch": 125.88,
      "learning_rate": 3.2165695792880265e-06,
      "loss": 0.0626,
      "step": 324130
    },
    {
      "epoch": 125.88,
      "learning_rate": 3.2160517799352752e-06,
      "loss": 0.0763,
      "step": 324140
    },
    {
      "epoch": 125.88,
      "learning_rate": 3.215533980582525e-06,
      "loss": 0.0093,
      "step": 324150
    },
    {
      "epoch": 125.89,
      "learning_rate": 3.2150161812297736e-06,
      "loss": 0.0001,
      "step": 324160
    },
    {
      "epoch": 125.89,
      "learning_rate": 3.2144983818770233e-06,
      "loss": 0.0232,
      "step": 324170
    },
    {
      "epoch": 125.9,
      "learning_rate": 3.213980582524272e-06,
      "loss": 0.1123,
      "step": 324180
    },
    {
      "epoch": 125.9,
      "learning_rate": 3.2134627831715216e-06,
      "loss": 0.0038,
      "step": 324190
    },
    {
      "epoch": 125.9,
      "learning_rate": 3.2129449838187704e-06,
      "loss": 0.0708,
      "step": 324200
    },
    {
      "epoch": 125.91,
      "learning_rate": 3.21242718446602e-06,
      "loss": 0.0677,
      "step": 324210
    },
    {
      "epoch": 125.91,
      "learning_rate": 3.211909385113269e-06,
      "loss": 0.0001,
      "step": 324220
    },
    {
      "epoch": 125.91,
      "learning_rate": 3.2113915857605184e-06,
      "loss": 0.0178,
      "step": 324230
    },
    {
      "epoch": 125.92,
      "learning_rate": 3.210873786407767e-06,
      "loss": 0.0977,
      "step": 324240
    },
    {
      "epoch": 125.92,
      "learning_rate": 3.210355987055017e-06,
      "loss": 0.0933,
      "step": 324250
    },
    {
      "epoch": 125.93,
      "learning_rate": 3.2098381877022656e-06,
      "loss": 0.0649,
      "step": 324260
    },
    {
      "epoch": 125.93,
      "learning_rate": 3.2093203883495148e-06,
      "loss": 0.0299,
      "step": 324270
    },
    {
      "epoch": 125.93,
      "learning_rate": 3.208802588996764e-06,
      "loss": 0.0358,
      "step": 324280
    },
    {
      "epoch": 125.94,
      "learning_rate": 3.208284789644013e-06,
      "loss": 0.0569,
      "step": 324290
    },
    {
      "epoch": 125.94,
      "learning_rate": 3.2077669902912624e-06,
      "loss": 0.0002,
      "step": 324300
    },
    {
      "epoch": 125.95,
      "learning_rate": 3.2072491909385116e-06,
      "loss": 0.0613,
      "step": 324310
    },
    {
      "epoch": 125.95,
      "learning_rate": 3.2067313915857607e-06,
      "loss": 0.0799,
      "step": 324320
    },
    {
      "epoch": 125.95,
      "learning_rate": 3.20621359223301e-06,
      "loss": 0.0006,
      "step": 324330
    },
    {
      "epoch": 125.96,
      "learning_rate": 3.205695792880259e-06,
      "loss": 0.0006,
      "step": 324340
    },
    {
      "epoch": 125.96,
      "learning_rate": 3.2051779935275083e-06,
      "loss": 0.1195,
      "step": 324350
    },
    {
      "epoch": 125.97,
      "learning_rate": 3.2046601941747575e-06,
      "loss": 0.0531,
      "step": 324360
    },
    {
      "epoch": 125.97,
      "learning_rate": 3.2041423948220067e-06,
      "loss": 0.0121,
      "step": 324370
    },
    {
      "epoch": 125.97,
      "learning_rate": 3.203624595469256e-06,
      "loss": 0.0394,
      "step": 324380
    },
    {
      "epoch": 125.98,
      "learning_rate": 3.203106796116505e-06,
      "loss": 0.0401,
      "step": 324390
    },
    {
      "epoch": 125.98,
      "learning_rate": 3.2025889967637543e-06,
      "loss": 0.0405,
      "step": 324400
    },
    {
      "epoch": 125.98,
      "learning_rate": 3.2020711974110035e-06,
      "loss": 0.0419,
      "step": 324410
    },
    {
      "epoch": 125.99,
      "learning_rate": 3.2015533980582527e-06,
      "loss": 0.0277,
      "step": 324420
    },
    {
      "epoch": 125.99,
      "learning_rate": 3.201035598705502e-06,
      "loss": 0.0519,
      "step": 324430
    },
    {
      "epoch": 126.0,
      "learning_rate": 3.2005177993527507e-06,
      "loss": 0.0133,
      "step": 324440
    },
    {
      "epoch": 126.0,
      "learning_rate": 3.2000000000000003e-06,
      "loss": 0.0169,
      "step": 324450
    },
    {
      "epoch": 126.0,
      "eval_accuracy": 0.951031636863824,
      "eval_loss": 0.380782812833786,
      "eval_runtime": 8.2806,
      "eval_samples_per_second": 438.977,
      "eval_steps_per_second": 54.948,
      "step": 324450
    },
    {
      "epoch": 126.0,
      "learning_rate": 3.199482200647249e-06,
      "loss": 0.0002,
      "step": 324460
    },
    {
      "epoch": 126.01,
      "learning_rate": 3.1989644012944987e-06,
      "loss": 0.0534,
      "step": 324470
    },
    {
      "epoch": 126.01,
      "learning_rate": 3.1984466019417474e-06,
      "loss": 0.029,
      "step": 324480
    },
    {
      "epoch": 126.02,
      "learning_rate": 3.197928802588997e-06,
      "loss": 0.0901,
      "step": 324490
    },
    {
      "epoch": 126.02,
      "learning_rate": 3.197411003236246e-06,
      "loss": 0.1226,
      "step": 324500
    },
    {
      "epoch": 126.02,
      "learning_rate": 3.1968932038834954e-06,
      "loss": 0.1409,
      "step": 324510
    },
    {
      "epoch": 126.03,
      "learning_rate": 3.1963754045307442e-06,
      "loss": 0.0419,
      "step": 324520
    },
    {
      "epoch": 126.03,
      "learning_rate": 3.195857605177994e-06,
      "loss": 0.0085,
      "step": 324530
    },
    {
      "epoch": 126.03,
      "learning_rate": 3.1953398058252426e-06,
      "loss": 0.0206,
      "step": 324540
    },
    {
      "epoch": 126.04,
      "learning_rate": 3.1948220064724922e-06,
      "loss": 0.0352,
      "step": 324550
    },
    {
      "epoch": 126.04,
      "learning_rate": 3.194304207119741e-06,
      "loss": 0.0369,
      "step": 324560
    },
    {
      "epoch": 126.05,
      "learning_rate": 3.1937864077669906e-06,
      "loss": 0.034,
      "step": 324570
    },
    {
      "epoch": 126.05,
      "learning_rate": 3.1932686084142394e-06,
      "loss": 0.1324,
      "step": 324580
    },
    {
      "epoch": 126.05,
      "learning_rate": 3.192750809061489e-06,
      "loss": 0.0399,
      "step": 324590
    },
    {
      "epoch": 126.06,
      "learning_rate": 3.192233009708738e-06,
      "loss": 0.1423,
      "step": 324600
    },
    {
      "epoch": 126.06,
      "learning_rate": 3.1917152103559874e-06,
      "loss": 0.0174,
      "step": 324610
    },
    {
      "epoch": 126.07,
      "learning_rate": 3.1911974110032366e-06,
      "loss": 0.1043,
      "step": 324620
    },
    {
      "epoch": 126.07,
      "learning_rate": 3.1906796116504858e-06,
      "loss": 0.0506,
      "step": 324630
    },
    {
      "epoch": 126.07,
      "learning_rate": 3.190161812297735e-06,
      "loss": 0.0074,
      "step": 324640
    },
    {
      "epoch": 126.08,
      "learning_rate": 3.189644012944984e-06,
      "loss": 0.0291,
      "step": 324650
    },
    {
      "epoch": 126.08,
      "learning_rate": 3.1891262135922334e-06,
      "loss": 0.0305,
      "step": 324660
    },
    {
      "epoch": 126.09,
      "learning_rate": 3.1886084142394826e-06,
      "loss": 0.0408,
      "step": 324670
    },
    {
      "epoch": 126.09,
      "learning_rate": 3.1880906148867318e-06,
      "loss": 0.0084,
      "step": 324680
    },
    {
      "epoch": 126.09,
      "learning_rate": 3.187572815533981e-06,
      "loss": 0.0361,
      "step": 324690
    },
    {
      "epoch": 126.1,
      "learning_rate": 3.18705501618123e-06,
      "loss": 0.0414,
      "step": 324700
    },
    {
      "epoch": 126.1,
      "learning_rate": 3.1865372168284793e-06,
      "loss": 0.0578,
      "step": 324710
    },
    {
      "epoch": 126.1,
      "learning_rate": 3.1860194174757285e-06,
      "loss": 0.0161,
      "step": 324720
    },
    {
      "epoch": 126.11,
      "learning_rate": 3.1855016181229777e-06,
      "loss": 0.1001,
      "step": 324730
    },
    {
      "epoch": 126.11,
      "learning_rate": 3.184983818770227e-06,
      "loss": 0.0905,
      "step": 324740
    },
    {
      "epoch": 126.12,
      "learning_rate": 3.184466019417476e-06,
      "loss": 0.0183,
      "step": 324750
    },
    {
      "epoch": 126.12,
      "learning_rate": 3.1839482200647253e-06,
      "loss": 0.045,
      "step": 324760
    },
    {
      "epoch": 126.12,
      "learning_rate": 3.183430420711974e-06,
      "loss": 0.0535,
      "step": 324770
    },
    {
      "epoch": 126.13,
      "learning_rate": 3.1829126213592237e-06,
      "loss": 0.0199,
      "step": 324780
    },
    {
      "epoch": 126.13,
      "learning_rate": 3.1823948220064725e-06,
      "loss": 0.0103,
      "step": 324790
    },
    {
      "epoch": 126.14,
      "learning_rate": 3.181877022653722e-06,
      "loss": 0.0105,
      "step": 324800
    },
    {
      "epoch": 126.14,
      "learning_rate": 3.181359223300971e-06,
      "loss": 0.0005,
      "step": 324810
    },
    {
      "epoch": 126.14,
      "learning_rate": 3.1808414239482205e-06,
      "loss": 0.1236,
      "step": 324820
    },
    {
      "epoch": 126.15,
      "learning_rate": 3.1803236245954692e-06,
      "loss": 0.0313,
      "step": 324830
    },
    {
      "epoch": 126.15,
      "learning_rate": 3.179805825242719e-06,
      "loss": 0.112,
      "step": 324840
    },
    {
      "epoch": 126.16,
      "learning_rate": 3.1792880258899676e-06,
      "loss": 0.0757,
      "step": 324850
    },
    {
      "epoch": 126.16,
      "learning_rate": 3.1787702265372173e-06,
      "loss": 0.0096,
      "step": 324860
    },
    {
      "epoch": 126.16,
      "learning_rate": 3.178252427184466e-06,
      "loss": 0.0165,
      "step": 324870
    },
    {
      "epoch": 126.17,
      "learning_rate": 3.1777346278317156e-06,
      "loss": 0.1481,
      "step": 324880
    },
    {
      "epoch": 126.17,
      "learning_rate": 3.1772168284789644e-06,
      "loss": 0.0278,
      "step": 324890
    },
    {
      "epoch": 126.17,
      "learning_rate": 3.176699029126214e-06,
      "loss": 0.0018,
      "step": 324900
    },
    {
      "epoch": 126.18,
      "learning_rate": 3.176181229773463e-06,
      "loss": 0.0096,
      "step": 324910
    },
    {
      "epoch": 126.18,
      "learning_rate": 3.1756634304207124e-06,
      "loss": 0.0002,
      "step": 324920
    },
    {
      "epoch": 126.19,
      "learning_rate": 3.175145631067961e-06,
      "loss": 0.1061,
      "step": 324930
    },
    {
      "epoch": 126.19,
      "learning_rate": 3.174627831715211e-06,
      "loss": 0.0973,
      "step": 324940
    },
    {
      "epoch": 126.19,
      "learning_rate": 3.1741100323624596e-06,
      "loss": 0.0567,
      "step": 324950
    },
    {
      "epoch": 126.2,
      "learning_rate": 3.173592233009709e-06,
      "loss": 0.0551,
      "step": 324960
    },
    {
      "epoch": 126.2,
      "learning_rate": 3.173074433656958e-06,
      "loss": 0.173,
      "step": 324970
    },
    {
      "epoch": 126.21,
      "learning_rate": 3.1725566343042076e-06,
      "loss": 0.0089,
      "step": 324980
    },
    {
      "epoch": 126.21,
      "learning_rate": 3.1720388349514564e-06,
      "loss": 0.034,
      "step": 324990
    },
    {
      "epoch": 126.21,
      "learning_rate": 3.171521035598706e-06,
      "loss": 0.0002,
      "step": 325000
    },
    {
      "epoch": 126.22,
      "learning_rate": 3.1710032362459547e-06,
      "loss": 0.0362,
      "step": 325010
    },
    {
      "epoch": 126.22,
      "learning_rate": 3.1704854368932044e-06,
      "loss": 0.0957,
      "step": 325020
    },
    {
      "epoch": 126.23,
      "learning_rate": 3.169967637540453e-06,
      "loss": 0.1369,
      "step": 325030
    },
    {
      "epoch": 126.23,
      "learning_rate": 3.1694498381877028e-06,
      "loss": 0.0481,
      "step": 325040
    },
    {
      "epoch": 126.23,
      "learning_rate": 3.1689320388349515e-06,
      "loss": 0.0001,
      "step": 325050
    },
    {
      "epoch": 126.24,
      "learning_rate": 3.168414239482201e-06,
      "loss": 0.0259,
      "step": 325060
    },
    {
      "epoch": 126.24,
      "learning_rate": 3.16789644012945e-06,
      "loss": 0.0091,
      "step": 325070
    },
    {
      "epoch": 126.24,
      "learning_rate": 3.167378640776699e-06,
      "loss": 0.089,
      "step": 325080
    },
    {
      "epoch": 126.25,
      "learning_rate": 3.1668608414239483e-06,
      "loss": 0.0291,
      "step": 325090
    },
    {
      "epoch": 126.25,
      "learning_rate": 3.1663430420711975e-06,
      "loss": 0.005,
      "step": 325100
    },
    {
      "epoch": 126.26,
      "learning_rate": 3.1658252427184467e-06,
      "loss": 0.0008,
      "step": 325110
    },
    {
      "epoch": 126.26,
      "learning_rate": 3.165307443365696e-06,
      "loss": 0.0818,
      "step": 325120
    },
    {
      "epoch": 126.26,
      "learning_rate": 3.164789644012945e-06,
      "loss": 0.0779,
      "step": 325130
    },
    {
      "epoch": 126.27,
      "learning_rate": 3.1642718446601943e-06,
      "loss": 0.0263,
      "step": 325140
    },
    {
      "epoch": 126.27,
      "learning_rate": 3.163754045307444e-06,
      "loss": 0.0625,
      "step": 325150
    },
    {
      "epoch": 126.28,
      "learning_rate": 3.1632362459546927e-06,
      "loss": 0.0383,
      "step": 325160
    },
    {
      "epoch": 126.28,
      "learning_rate": 3.1627184466019423e-06,
      "loss": 0.0004,
      "step": 325170
    },
    {
      "epoch": 126.28,
      "learning_rate": 3.162200647249191e-06,
      "loss": 0.0266,
      "step": 325180
    },
    {
      "epoch": 126.29,
      "learning_rate": 3.1616828478964407e-06,
      "loss": 0.0729,
      "step": 325190
    },
    {
      "epoch": 126.29,
      "learning_rate": 3.1611650485436894e-06,
      "loss": 0.0259,
      "step": 325200
    },
    {
      "epoch": 126.3,
      "learning_rate": 3.160647249190939e-06,
      "loss": 0.0476,
      "step": 325210
    },
    {
      "epoch": 126.3,
      "learning_rate": 3.160129449838188e-06,
      "loss": 0.0034,
      "step": 325220
    },
    {
      "epoch": 126.3,
      "learning_rate": 3.1596116504854375e-06,
      "loss": 0.1365,
      "step": 325230
    },
    {
      "epoch": 126.31,
      "learning_rate": 3.1590938511326862e-06,
      "loss": 0.0504,
      "step": 325240
    },
    {
      "epoch": 126.31,
      "learning_rate": 3.158576051779936e-06,
      "loss": 0.0014,
      "step": 325250
    },
    {
      "epoch": 126.31,
      "learning_rate": 3.1580582524271846e-06,
      "loss": 0.0542,
      "step": 325260
    },
    {
      "epoch": 126.32,
      "learning_rate": 3.1575404530744342e-06,
      "loss": 0.0721,
      "step": 325270
    },
    {
      "epoch": 126.32,
      "learning_rate": 3.157022653721683e-06,
      "loss": 0.0075,
      "step": 325280
    },
    {
      "epoch": 126.33,
      "learning_rate": 3.1565048543689326e-06,
      "loss": 0.078,
      "step": 325290
    },
    {
      "epoch": 126.33,
      "learning_rate": 3.1559870550161814e-06,
      "loss": 0.0699,
      "step": 325300
    },
    {
      "epoch": 126.33,
      "learning_rate": 3.155469255663431e-06,
      "loss": 0.0805,
      "step": 325310
    },
    {
      "epoch": 126.34,
      "learning_rate": 3.1549514563106798e-06,
      "loss": 0.0222,
      "step": 325320
    },
    {
      "epoch": 126.34,
      "learning_rate": 3.1544336569579294e-06,
      "loss": 0.0792,
      "step": 325330
    },
    {
      "epoch": 126.35,
      "learning_rate": 3.153915857605178e-06,
      "loss": 0.0415,
      "step": 325340
    },
    {
      "epoch": 126.35,
      "learning_rate": 3.1533980582524278e-06,
      "loss": 0.0761,
      "step": 325350
    },
    {
      "epoch": 126.35,
      "learning_rate": 3.1528802588996766e-06,
      "loss": 0.0658,
      "step": 325360
    },
    {
      "epoch": 126.36,
      "learning_rate": 3.152362459546926e-06,
      "loss": 0.1202,
      "step": 325370
    },
    {
      "epoch": 126.36,
      "learning_rate": 3.151844660194175e-06,
      "loss": 0.0206,
      "step": 325380
    },
    {
      "epoch": 126.37,
      "learning_rate": 3.1513268608414246e-06,
      "loss": 0.1212,
      "step": 325390
    },
    {
      "epoch": 126.37,
      "learning_rate": 3.1508090614886733e-06,
      "loss": 0.0683,
      "step": 325400
    },
    {
      "epoch": 126.37,
      "learning_rate": 3.1502912621359225e-06,
      "loss": 0.0301,
      "step": 325410
    },
    {
      "epoch": 126.38,
      "learning_rate": 3.1497734627831717e-06,
      "loss": 0.0082,
      "step": 325420
    },
    {
      "epoch": 126.38,
      "learning_rate": 3.149255663430421e-06,
      "loss": 0.0096,
      "step": 325430
    },
    {
      "epoch": 126.38,
      "learning_rate": 3.14873786407767e-06,
      "loss": 0.0686,
      "step": 325440
    },
    {
      "epoch": 126.39,
      "learning_rate": 3.1482200647249193e-06,
      "loss": 0.0699,
      "step": 325450
    },
    {
      "epoch": 126.39,
      "learning_rate": 3.1477022653721685e-06,
      "loss": 0.002,
      "step": 325460
    },
    {
      "epoch": 126.4,
      "learning_rate": 3.1471844660194177e-06,
      "loss": 0.0779,
      "step": 325470
    },
    {
      "epoch": 126.4,
      "learning_rate": 3.146666666666667e-06,
      "loss": 0.0002,
      "step": 325480
    },
    {
      "epoch": 126.4,
      "learning_rate": 3.146148867313916e-06,
      "loss": 0.1036,
      "step": 325490
    },
    {
      "epoch": 126.41,
      "learning_rate": 3.1456310679611653e-06,
      "loss": 0.0436,
      "step": 325500
    },
    {
      "epoch": 126.41,
      "learning_rate": 3.1451132686084145e-06,
      "loss": 0.1197,
      "step": 325510
    },
    {
      "epoch": 126.42,
      "learning_rate": 3.1445954692556637e-06,
      "loss": 0.0309,
      "step": 325520
    },
    {
      "epoch": 126.42,
      "learning_rate": 3.144077669902913e-06,
      "loss": 0.0165,
      "step": 325530
    },
    {
      "epoch": 126.42,
      "learning_rate": 3.143559870550162e-06,
      "loss": 0.0115,
      "step": 325540
    },
    {
      "epoch": 126.43,
      "learning_rate": 3.1430420711974113e-06,
      "loss": 0.0005,
      "step": 325550
    },
    {
      "epoch": 126.43,
      "learning_rate": 3.1425242718446604e-06,
      "loss": 0.0106,
      "step": 325560
    },
    {
      "epoch": 126.43,
      "learning_rate": 3.1420064724919096e-06,
      "loss": 0.0665,
      "step": 325570
    },
    {
      "epoch": 126.44,
      "learning_rate": 3.1414886731391584e-06,
      "loss": 0.0502,
      "step": 325580
    },
    {
      "epoch": 126.44,
      "learning_rate": 3.140970873786408e-06,
      "loss": 0.0939,
      "step": 325590
    },
    {
      "epoch": 126.45,
      "learning_rate": 3.140453074433657e-06,
      "loss": 0.0339,
      "step": 325600
    },
    {
      "epoch": 126.45,
      "learning_rate": 3.1399352750809064e-06,
      "loss": 0.0006,
      "step": 325610
    },
    {
      "epoch": 126.45,
      "learning_rate": 3.139417475728155e-06,
      "loss": 0.0196,
      "step": 325620
    },
    {
      "epoch": 126.46,
      "learning_rate": 3.138899676375405e-06,
      "loss": 0.0312,
      "step": 325630
    },
    {
      "epoch": 126.46,
      "learning_rate": 3.1383818770226536e-06,
      "loss": 0.0123,
      "step": 325640
    },
    {
      "epoch": 126.47,
      "learning_rate": 3.137864077669903e-06,
      "loss": 0.0247,
      "step": 325650
    },
    {
      "epoch": 126.47,
      "learning_rate": 3.137346278317152e-06,
      "loss": 0.0645,
      "step": 325660
    },
    {
      "epoch": 126.47,
      "learning_rate": 3.1368284789644016e-06,
      "loss": 0.0079,
      "step": 325670
    },
    {
      "epoch": 126.48,
      "learning_rate": 3.1363106796116504e-06,
      "loss": 0.0381,
      "step": 325680
    },
    {
      "epoch": 126.48,
      "learning_rate": 3.1357928802589e-06,
      "loss": 0.1424,
      "step": 325690
    },
    {
      "epoch": 126.49,
      "learning_rate": 3.1352750809061487e-06,
      "loss": 0.009,
      "step": 325700
    },
    {
      "epoch": 126.49,
      "learning_rate": 3.1347572815533984e-06,
      "loss": 0.1653,
      "step": 325710
    },
    {
      "epoch": 126.49,
      "learning_rate": 3.134239482200648e-06,
      "loss": 0.0964,
      "step": 325720
    },
    {
      "epoch": 126.5,
      "learning_rate": 3.1337216828478968e-06,
      "loss": 0.0361,
      "step": 325730
    },
    {
      "epoch": 126.5,
      "learning_rate": 3.133203883495146e-06,
      "loss": 0.0391,
      "step": 325740
    },
    {
      "epoch": 126.5,
      "learning_rate": 3.132686084142395e-06,
      "loss": 0.0239,
      "step": 325750
    },
    {
      "epoch": 126.51,
      "learning_rate": 3.1321682847896443e-06,
      "loss": 0.0197,
      "step": 325760
    },
    {
      "epoch": 126.51,
      "learning_rate": 3.1316504854368935e-06,
      "loss": 0.0153,
      "step": 325770
    },
    {
      "epoch": 126.52,
      "learning_rate": 3.1311326860841427e-06,
      "loss": 0.0255,
      "step": 325780
    },
    {
      "epoch": 126.52,
      "learning_rate": 3.130614886731392e-06,
      "loss": 0.0125,
      "step": 325790
    },
    {
      "epoch": 126.52,
      "learning_rate": 3.130097087378641e-06,
      "loss": 0.0116,
      "step": 325800
    },
    {
      "epoch": 126.53,
      "learning_rate": 3.1295792880258903e-06,
      "loss": 0.0262,
      "step": 325810
    },
    {
      "epoch": 126.53,
      "learning_rate": 3.1290614886731395e-06,
      "loss": 0.0146,
      "step": 325820
    },
    {
      "epoch": 126.54,
      "learning_rate": 3.1285436893203887e-06,
      "loss": 0.0344,
      "step": 325830
    },
    {
      "epoch": 126.54,
      "learning_rate": 3.128025889967638e-06,
      "loss": 0.0912,
      "step": 325840
    },
    {
      "epoch": 126.54,
      "learning_rate": 3.127508090614887e-06,
      "loss": 0.0158,
      "step": 325850
    },
    {
      "epoch": 126.55,
      "learning_rate": 3.1269902912621363e-06,
      "loss": 0.1793,
      "step": 325860
    },
    {
      "epoch": 126.55,
      "learning_rate": 3.1264724919093855e-06,
      "loss": 0.047,
      "step": 325870
    },
    {
      "epoch": 126.56,
      "learning_rate": 3.1259546925566347e-06,
      "loss": 0.0015,
      "step": 325880
    },
    {
      "epoch": 126.56,
      "learning_rate": 3.125436893203884e-06,
      "loss": 0.1297,
      "step": 325890
    },
    {
      "epoch": 126.56,
      "learning_rate": 3.124919093851133e-06,
      "loss": 0.0175,
      "step": 325900
    },
    {
      "epoch": 126.57,
      "learning_rate": 3.124401294498382e-06,
      "loss": 0.0012,
      "step": 325910
    },
    {
      "epoch": 126.57,
      "learning_rate": 3.1238834951456315e-06,
      "loss": 0.0795,
      "step": 325920
    },
    {
      "epoch": 126.57,
      "learning_rate": 3.1233656957928802e-06,
      "loss": 0.07,
      "step": 325930
    },
    {
      "epoch": 126.58,
      "learning_rate": 3.12284789644013e-06,
      "loss": 0.0328,
      "step": 325940
    },
    {
      "epoch": 126.58,
      "learning_rate": 3.1223300970873786e-06,
      "loss": 0.0292,
      "step": 325950
    },
    {
      "epoch": 126.59,
      "learning_rate": 3.1218122977346282e-06,
      "loss": 0.1527,
      "step": 325960
    },
    {
      "epoch": 126.59,
      "learning_rate": 3.121294498381877e-06,
      "loss": 0.065,
      "step": 325970
    },
    {
      "epoch": 126.59,
      "learning_rate": 3.1207766990291266e-06,
      "loss": 0.0679,
      "step": 325980
    },
    {
      "epoch": 126.6,
      "learning_rate": 3.1202588996763754e-06,
      "loss": 0.0182,
      "step": 325990
    },
    {
      "epoch": 126.6,
      "learning_rate": 3.119741100323625e-06,
      "loss": 0.0906,
      "step": 326000
    },
    {
      "epoch": 126.61,
      "learning_rate": 3.1192233009708738e-06,
      "loss": 0.0584,
      "step": 326010
    },
    {
      "epoch": 126.61,
      "learning_rate": 3.1187055016181234e-06,
      "loss": 0.036,
      "step": 326020
    },
    {
      "epoch": 126.61,
      "learning_rate": 3.118187702265372e-06,
      "loss": 0.0262,
      "step": 326030
    },
    {
      "epoch": 126.62,
      "learning_rate": 3.1176699029126218e-06,
      "loss": 0.0001,
      "step": 326040
    },
    {
      "epoch": 126.62,
      "learning_rate": 3.1171521035598706e-06,
      "loss": 0.0127,
      "step": 326050
    },
    {
      "epoch": 126.63,
      "learning_rate": 3.11663430420712e-06,
      "loss": 0.0547,
      "step": 326060
    },
    {
      "epoch": 126.63,
      "learning_rate": 3.116116504854369e-06,
      "loss": 0.089,
      "step": 326070
    },
    {
      "epoch": 126.63,
      "learning_rate": 3.1155987055016186e-06,
      "loss": 0.0423,
      "step": 326080
    },
    {
      "epoch": 126.64,
      "learning_rate": 3.1150809061488673e-06,
      "loss": 0.0285,
      "step": 326090
    },
    {
      "epoch": 126.64,
      "learning_rate": 3.114563106796117e-06,
      "loss": 0.0539,
      "step": 326100
    },
    {
      "epoch": 126.64,
      "learning_rate": 3.1140453074433657e-06,
      "loss": 0.0353,
      "step": 326110
    },
    {
      "epoch": 126.65,
      "learning_rate": 3.1135275080906153e-06,
      "loss": 0.0732,
      "step": 326120
    },
    {
      "epoch": 126.65,
      "learning_rate": 3.113009708737864e-06,
      "loss": 0.0139,
      "step": 326130
    },
    {
      "epoch": 126.66,
      "learning_rate": 3.1124919093851137e-06,
      "loss": 0.1133,
      "step": 326140
    },
    {
      "epoch": 126.66,
      "learning_rate": 3.1119741100323625e-06,
      "loss": 0.0213,
      "step": 326150
    },
    {
      "epoch": 126.66,
      "learning_rate": 3.111456310679612e-06,
      "loss": 0.0036,
      "step": 326160
    },
    {
      "epoch": 126.67,
      "learning_rate": 3.110938511326861e-06,
      "loss": 0.09,
      "step": 326170
    },
    {
      "epoch": 126.67,
      "learning_rate": 3.1104207119741105e-06,
      "loss": 0.007,
      "step": 326180
    },
    {
      "epoch": 126.68,
      "learning_rate": 3.1099029126213593e-06,
      "loss": 0.0567,
      "step": 326190
    },
    {
      "epoch": 126.68,
      "learning_rate": 3.109385113268609e-06,
      "loss": 0.0875,
      "step": 326200
    },
    {
      "epoch": 126.68,
      "learning_rate": 3.1088673139158577e-06,
      "loss": 0.0872,
      "step": 326210
    },
    {
      "epoch": 126.69,
      "learning_rate": 3.108349514563107e-06,
      "loss": 0.0815,
      "step": 326220
    },
    {
      "epoch": 126.69,
      "learning_rate": 3.107831715210356e-06,
      "loss": 0.0075,
      "step": 326230
    },
    {
      "epoch": 126.7,
      "learning_rate": 3.1073139158576053e-06,
      "loss": 0.0374,
      "step": 326240
    },
    {
      "epoch": 126.7,
      "learning_rate": 3.1067961165048544e-06,
      "loss": 0.0051,
      "step": 326250
    },
    {
      "epoch": 126.7,
      "learning_rate": 3.1062783171521036e-06,
      "loss": 0.0961,
      "step": 326260
    },
    {
      "epoch": 126.71,
      "learning_rate": 3.1057605177993533e-06,
      "loss": 0.1242,
      "step": 326270
    },
    {
      "epoch": 126.71,
      "learning_rate": 3.105242718446602e-06,
      "loss": 0.0205,
      "step": 326280
    },
    {
      "epoch": 126.71,
      "learning_rate": 3.1047249190938516e-06,
      "loss": 0.0587,
      "step": 326290
    },
    {
      "epoch": 126.72,
      "learning_rate": 3.1042071197411004e-06,
      "loss": 0.0122,
      "step": 326300
    },
    {
      "epoch": 126.72,
      "learning_rate": 3.10368932038835e-06,
      "loss": 0.0583,
      "step": 326310
    },
    {
      "epoch": 126.73,
      "learning_rate": 3.103171521035599e-06,
      "loss": 0.0122,
      "step": 326320
    },
    {
      "epoch": 126.73,
      "learning_rate": 3.1026537216828484e-06,
      "loss": 0.0355,
      "step": 326330
    },
    {
      "epoch": 126.73,
      "learning_rate": 3.102135922330097e-06,
      "loss": 0.0067,
      "step": 326340
    },
    {
      "epoch": 126.74,
      "learning_rate": 3.101618122977347e-06,
      "loss": 0.0969,
      "step": 326350
    },
    {
      "epoch": 126.74,
      "learning_rate": 3.1011003236245956e-06,
      "loss": 0.0803,
      "step": 326360
    },
    {
      "epoch": 126.75,
      "learning_rate": 3.100582524271845e-06,
      "loss": 0.0138,
      "step": 326370
    },
    {
      "epoch": 126.75,
      "learning_rate": 3.100064724919094e-06,
      "loss": 0.0747,
      "step": 326380
    },
    {
      "epoch": 126.75,
      "learning_rate": 3.0995469255663436e-06,
      "loss": 0.008,
      "step": 326390
    },
    {
      "epoch": 126.76,
      "learning_rate": 3.0990291262135924e-06,
      "loss": 0.0957,
      "step": 326400
    },
    {
      "epoch": 126.76,
      "learning_rate": 3.098511326860842e-06,
      "loss": 0.1213,
      "step": 326410
    },
    {
      "epoch": 126.77,
      "learning_rate": 3.0979935275080908e-06,
      "loss": 0.061,
      "step": 326420
    },
    {
      "epoch": 126.77,
      "learning_rate": 3.0974757281553404e-06,
      "loss": 0.0014,
      "step": 326430
    },
    {
      "epoch": 126.77,
      "learning_rate": 3.096957928802589e-06,
      "loss": 0.0093,
      "step": 326440
    },
    {
      "epoch": 126.78,
      "learning_rate": 3.0964401294498388e-06,
      "loss": 0.1183,
      "step": 326450
    },
    {
      "epoch": 126.78,
      "learning_rate": 3.0959223300970875e-06,
      "loss": 0.0303,
      "step": 326460
    },
    {
      "epoch": 126.78,
      "learning_rate": 3.095404530744337e-06,
      "loss": 0.059,
      "step": 326470
    },
    {
      "epoch": 126.79,
      "learning_rate": 3.094886731391586e-06,
      "loss": 0.0331,
      "step": 326480
    },
    {
      "epoch": 126.79,
      "learning_rate": 3.0943689320388355e-06,
      "loss": 0.0608,
      "step": 326490
    },
    {
      "epoch": 126.8,
      "learning_rate": 3.0938511326860843e-06,
      "loss": 0.0811,
      "step": 326500
    },
    {
      "epoch": 126.8,
      "learning_rate": 3.093333333333334e-06,
      "loss": 0.0081,
      "step": 326510
    },
    {
      "epoch": 126.8,
      "learning_rate": 3.0928155339805827e-06,
      "loss": 0.0261,
      "step": 326520
    },
    {
      "epoch": 126.81,
      "learning_rate": 3.0922977346278323e-06,
      "loss": 0.0389,
      "step": 326530
    },
    {
      "epoch": 126.81,
      "learning_rate": 3.091779935275081e-06,
      "loss": 0.0001,
      "step": 326540
    },
    {
      "epoch": 126.82,
      "learning_rate": 3.0912621359223303e-06,
      "loss": 0.0002,
      "step": 326550
    },
    {
      "epoch": 126.82,
      "learning_rate": 3.0907443365695795e-06,
      "loss": 0.0342,
      "step": 326560
    },
    {
      "epoch": 126.82,
      "learning_rate": 3.0902265372168287e-06,
      "loss": 0.0462,
      "step": 326570
    },
    {
      "epoch": 126.83,
      "learning_rate": 3.089708737864078e-06,
      "loss": 0.0087,
      "step": 326580
    },
    {
      "epoch": 126.83,
      "learning_rate": 3.089190938511327e-06,
      "loss": 0.0737,
      "step": 326590
    },
    {
      "epoch": 126.83,
      "learning_rate": 3.0886731391585763e-06,
      "loss": 0.0272,
      "step": 326600
    },
    {
      "epoch": 126.84,
      "learning_rate": 3.0881553398058254e-06,
      "loss": 0.0497,
      "step": 326610
    },
    {
      "epoch": 126.84,
      "learning_rate": 3.0876375404530746e-06,
      "loss": 0.0001,
      "step": 326620
    },
    {
      "epoch": 126.85,
      "learning_rate": 3.087119741100324e-06,
      "loss": 0.0561,
      "step": 326630
    },
    {
      "epoch": 126.85,
      "learning_rate": 3.086601941747573e-06,
      "loss": 0.0002,
      "step": 326640
    },
    {
      "epoch": 126.85,
      "learning_rate": 3.0860841423948222e-06,
      "loss": 0.0222,
      "step": 326650
    },
    {
      "epoch": 126.86,
      "learning_rate": 3.0855663430420714e-06,
      "loss": 0.0437,
      "step": 326660
    },
    {
      "epoch": 126.86,
      "learning_rate": 3.0850485436893206e-06,
      "loss": 0.045,
      "step": 326670
    },
    {
      "epoch": 126.87,
      "learning_rate": 3.08453074433657e-06,
      "loss": 0.0649,
      "step": 326680
    },
    {
      "epoch": 126.87,
      "learning_rate": 3.084012944983819e-06,
      "loss": 0.0124,
      "step": 326690
    },
    {
      "epoch": 126.87,
      "learning_rate": 3.083495145631068e-06,
      "loss": 0.0191,
      "step": 326700
    },
    {
      "epoch": 126.88,
      "learning_rate": 3.0829773462783174e-06,
      "loss": 0.1462,
      "step": 326710
    },
    {
      "epoch": 126.88,
      "learning_rate": 3.082459546925566e-06,
      "loss": 0.0989,
      "step": 326720
    },
    {
      "epoch": 126.89,
      "learning_rate": 3.0819417475728158e-06,
      "loss": 0.0766,
      "step": 326730
    },
    {
      "epoch": 126.89,
      "learning_rate": 3.0814239482200646e-06,
      "loss": 0.0612,
      "step": 326740
    },
    {
      "epoch": 126.89,
      "learning_rate": 3.080906148867314e-06,
      "loss": 0.1186,
      "step": 326750
    },
    {
      "epoch": 126.9,
      "learning_rate": 3.080388349514563e-06,
      "loss": 0.0226,
      "step": 326760
    },
    {
      "epoch": 126.9,
      "learning_rate": 3.0798705501618126e-06,
      "loss": 0.0314,
      "step": 326770
    },
    {
      "epoch": 126.9,
      "learning_rate": 3.0793527508090613e-06,
      "loss": 0.0636,
      "step": 326780
    },
    {
      "epoch": 126.91,
      "learning_rate": 3.078834951456311e-06,
      "loss": 0.0465,
      "step": 326790
    },
    {
      "epoch": 126.91,
      "learning_rate": 3.0783171521035597e-06,
      "loss": 0.0112,
      "step": 326800
    },
    {
      "epoch": 126.92,
      "learning_rate": 3.0777993527508093e-06,
      "loss": 0.044,
      "step": 326810
    },
    {
      "epoch": 126.92,
      "learning_rate": 3.077281553398058e-06,
      "loss": 0.0935,
      "step": 326820
    },
    {
      "epoch": 126.92,
      "learning_rate": 3.0767637540453077e-06,
      "loss": 0.1455,
      "step": 326830
    },
    {
      "epoch": 126.93,
      "learning_rate": 3.0762459546925573e-06,
      "loss": 0.0829,
      "step": 326840
    },
    {
      "epoch": 126.93,
      "learning_rate": 3.075728155339806e-06,
      "loss": 0.0588,
      "step": 326850
    },
    {
      "epoch": 126.94,
      "learning_rate": 3.0752103559870557e-06,
      "loss": 0.0549,
      "step": 326860
    },
    {
      "epoch": 126.94,
      "learning_rate": 3.0746925566343045e-06,
      "loss": 0.0762,
      "step": 326870
    },
    {
      "epoch": 126.94,
      "learning_rate": 3.0741747572815537e-06,
      "loss": 0.0384,
      "step": 326880
    },
    {
      "epoch": 126.95,
      "learning_rate": 3.073656957928803e-06,
      "loss": 0.0094,
      "step": 326890
    },
    {
      "epoch": 126.95,
      "learning_rate": 3.073139158576052e-06,
      "loss": 0.0179,
      "step": 326900
    },
    {
      "epoch": 126.96,
      "learning_rate": 3.0726213592233013e-06,
      "loss": 0.0004,
      "step": 326910
    },
    {
      "epoch": 126.96,
      "learning_rate": 3.0721035598705505e-06,
      "loss": 0.0232,
      "step": 326920
    },
    {
      "epoch": 126.96,
      "learning_rate": 3.0715857605177997e-06,
      "loss": 0.0346,
      "step": 326930
    },
    {
      "epoch": 126.97,
      "learning_rate": 3.071067961165049e-06,
      "loss": 0.1026,
      "step": 326940
    },
    {
      "epoch": 126.97,
      "learning_rate": 3.070550161812298e-06,
      "loss": 0.0186,
      "step": 326950
    },
    {
      "epoch": 126.97,
      "learning_rate": 3.0700323624595473e-06,
      "loss": 0.1791,
      "step": 326960
    },
    {
      "epoch": 126.98,
      "learning_rate": 3.0695145631067965e-06,
      "loss": 0.0802,
      "step": 326970
    },
    {
      "epoch": 126.98,
      "learning_rate": 3.0689967637540456e-06,
      "loss": 0.1251,
      "step": 326980
    },
    {
      "epoch": 126.99,
      "learning_rate": 3.068478964401295e-06,
      "loss": 0.0842,
      "step": 326990
    },
    {
      "epoch": 126.99,
      "learning_rate": 3.067961165048544e-06,
      "loss": 0.007,
      "step": 327000
    },
    {
      "epoch": 126.99,
      "learning_rate": 3.0674433656957932e-06,
      "loss": 0.0338,
      "step": 327010
    },
    {
      "epoch": 127.0,
      "learning_rate": 3.0669255663430424e-06,
      "loss": 0.0277,
      "step": 327020
    },
    {
      "epoch": 127.0,
      "eval_accuracy": 0.9524071526822558,
      "eval_loss": 0.37351882457733154,
      "eval_runtime": 8.1793,
      "eval_samples_per_second": 444.416,
      "eval_steps_per_second": 55.628,
      "step": 327025
    },
    {
      "epoch": 127.0,
      "learning_rate": 3.0664077669902916e-06,
      "loss": 0.0036,
      "step": 327030
    },
    {
      "epoch": 127.01,
      "learning_rate": 3.065889967637541e-06,
      "loss": 0.0449,
      "step": 327040
    },
    {
      "epoch": 127.01,
      "learning_rate": 3.0653721682847896e-06,
      "loss": 0.0688,
      "step": 327050
    },
    {
      "epoch": 127.01,
      "learning_rate": 3.064854368932039e-06,
      "loss": 0.0156,
      "step": 327060
    },
    {
      "epoch": 127.02,
      "learning_rate": 3.064336569579288e-06,
      "loss": 0.1342,
      "step": 327070
    },
    {
      "epoch": 127.02,
      "learning_rate": 3.0638187702265376e-06,
      "loss": 0.0374,
      "step": 327080
    },
    {
      "epoch": 127.03,
      "learning_rate": 3.0633009708737864e-06,
      "loss": 0.0581,
      "step": 327090
    },
    {
      "epoch": 127.03,
      "learning_rate": 3.062783171521036e-06,
      "loss": 0.1054,
      "step": 327100
    },
    {
      "epoch": 127.03,
      "learning_rate": 3.0622653721682848e-06,
      "loss": 0.0106,
      "step": 327110
    },
    {
      "epoch": 127.04,
      "learning_rate": 3.0617475728155344e-06,
      "loss": 0.0705,
      "step": 327120
    },
    {
      "epoch": 127.04,
      "learning_rate": 3.061229773462783e-06,
      "loss": 0.0865,
      "step": 327130
    },
    {
      "epoch": 127.04,
      "learning_rate": 3.0607119741100328e-06,
      "loss": 0.0066,
      "step": 327140
    },
    {
      "epoch": 127.05,
      "learning_rate": 3.0601941747572815e-06,
      "loss": 0.0187,
      "step": 327150
    },
    {
      "epoch": 127.05,
      "learning_rate": 3.059676375404531e-06,
      "loss": 0.0035,
      "step": 327160
    },
    {
      "epoch": 127.06,
      "learning_rate": 3.05915857605178e-06,
      "loss": 0.0743,
      "step": 327170
    },
    {
      "epoch": 127.06,
      "learning_rate": 3.0586407766990295e-06,
      "loss": 0.0003,
      "step": 327180
    },
    {
      "epoch": 127.06,
      "learning_rate": 3.0581229773462783e-06,
      "loss": 0.0983,
      "step": 327190
    },
    {
      "epoch": 127.07,
      "learning_rate": 3.057605177993528e-06,
      "loss": 0.0355,
      "step": 327200
    },
    {
      "epoch": 127.07,
      "learning_rate": 3.0570873786407767e-06,
      "loss": 0.0001,
      "step": 327210
    },
    {
      "epoch": 127.08,
      "learning_rate": 3.0565695792880263e-06,
      "loss": 0.0204,
      "step": 327220
    },
    {
      "epoch": 127.08,
      "learning_rate": 3.056051779935275e-06,
      "loss": 0.1041,
      "step": 327230
    },
    {
      "epoch": 127.08,
      "learning_rate": 3.0555339805825247e-06,
      "loss": 0.0009,
      "step": 327240
    },
    {
      "epoch": 127.09,
      "learning_rate": 3.0550161812297735e-06,
      "loss": 0.0124,
      "step": 327250
    },
    {
      "epoch": 127.09,
      "learning_rate": 3.054498381877023e-06,
      "loss": 0.0086,
      "step": 327260
    },
    {
      "epoch": 127.1,
      "learning_rate": 3.053980582524272e-06,
      "loss": 0.0013,
      "step": 327270
    },
    {
      "epoch": 127.1,
      "learning_rate": 3.0534627831715215e-06,
      "loss": 0.0056,
      "step": 327280
    },
    {
      "epoch": 127.1,
      "learning_rate": 3.0529449838187703e-06,
      "loss": 0.0095,
      "step": 327290
    },
    {
      "epoch": 127.11,
      "learning_rate": 3.05242718446602e-06,
      "loss": 0.0803,
      "step": 327300
    },
    {
      "epoch": 127.11,
      "learning_rate": 3.0519093851132686e-06,
      "loss": 0.0185,
      "step": 327310
    },
    {
      "epoch": 127.11,
      "learning_rate": 3.0513915857605183e-06,
      "loss": 0.091,
      "step": 327320
    },
    {
      "epoch": 127.12,
      "learning_rate": 3.050873786407767e-06,
      "loss": 0.0009,
      "step": 327330
    },
    {
      "epoch": 127.12,
      "learning_rate": 3.0503559870550166e-06,
      "loss": 0.0377,
      "step": 327340
    },
    {
      "epoch": 127.13,
      "learning_rate": 3.0498381877022654e-06,
      "loss": 0.0379,
      "step": 327350
    },
    {
      "epoch": 127.13,
      "learning_rate": 3.0493203883495146e-06,
      "loss": 0.0167,
      "step": 327360
    },
    {
      "epoch": 127.13,
      "learning_rate": 3.048802588996764e-06,
      "loss": 0.0296,
      "step": 327370
    },
    {
      "epoch": 127.14,
      "learning_rate": 3.048284789644013e-06,
      "loss": 0.0653,
      "step": 327380
    },
    {
      "epoch": 127.14,
      "learning_rate": 3.0477669902912626e-06,
      "loss": 0.0436,
      "step": 327390
    },
    {
      "epoch": 127.15,
      "learning_rate": 3.0472491909385114e-06,
      "loss": 0.036,
      "step": 327400
    },
    {
      "epoch": 127.15,
      "learning_rate": 3.046731391585761e-06,
      "loss": 0.0219,
      "step": 327410
    },
    {
      "epoch": 127.15,
      "learning_rate": 3.0462135922330098e-06,
      "loss": 0.0001,
      "step": 327420
    },
    {
      "epoch": 127.16,
      "learning_rate": 3.0456957928802594e-06,
      "loss": 0.0011,
      "step": 327430
    },
    {
      "epoch": 127.16,
      "learning_rate": 3.045177993527508e-06,
      "loss": 0.0801,
      "step": 327440
    },
    {
      "epoch": 127.17,
      "learning_rate": 3.044660194174758e-06,
      "loss": 0.0166,
      "step": 327450
    },
    {
      "epoch": 127.17,
      "learning_rate": 3.0441423948220066e-06,
      "loss": 0.0014,
      "step": 327460
    },
    {
      "epoch": 127.17,
      "learning_rate": 3.043624595469256e-06,
      "loss": 0.0551,
      "step": 327470
    },
    {
      "epoch": 127.18,
      "learning_rate": 3.043106796116505e-06,
      "loss": 0.0137,
      "step": 327480
    },
    {
      "epoch": 127.18,
      "learning_rate": 3.0425889967637546e-06,
      "loss": 0.0076,
      "step": 327490
    },
    {
      "epoch": 127.18,
      "learning_rate": 3.0420711974110033e-06,
      "loss": 0.1127,
      "step": 327500
    },
    {
      "epoch": 127.19,
      "learning_rate": 3.041553398058253e-06,
      "loss": 0.0311,
      "step": 327510
    },
    {
      "epoch": 127.19,
      "learning_rate": 3.0410355987055017e-06,
      "loss": 0.0004,
      "step": 327520
    },
    {
      "epoch": 127.2,
      "learning_rate": 3.0405177993527513e-06,
      "loss": 0.0147,
      "step": 327530
    },
    {
      "epoch": 127.2,
      "learning_rate": 3.04e-06,
      "loss": 0.1063,
      "step": 327540
    },
    {
      "epoch": 127.2,
      "learning_rate": 3.0394822006472497e-06,
      "loss": 0.0387,
      "step": 327550
    },
    {
      "epoch": 127.21,
      "learning_rate": 3.0389644012944985e-06,
      "loss": 0.0119,
      "step": 327560
    },
    {
      "epoch": 127.21,
      "learning_rate": 3.038446601941748e-06,
      "loss": 0.0754,
      "step": 327570
    },
    {
      "epoch": 127.22,
      "learning_rate": 3.037928802588997e-06,
      "loss": 0.073,
      "step": 327580
    },
    {
      "epoch": 127.22,
      "learning_rate": 3.0374110032362465e-06,
      "loss": 0.1298,
      "step": 327590
    },
    {
      "epoch": 127.22,
      "learning_rate": 3.0368932038834953e-06,
      "loss": 0.0533,
      "step": 327600
    },
    {
      "epoch": 127.23,
      "learning_rate": 3.036375404530745e-06,
      "loss": 0.2302,
      "step": 327610
    },
    {
      "epoch": 127.23,
      "learning_rate": 3.0358576051779937e-06,
      "loss": 0.1064,
      "step": 327620
    },
    {
      "epoch": 127.23,
      "learning_rate": 3.0353398058252433e-06,
      "loss": 0.0985,
      "step": 327630
    },
    {
      "epoch": 127.24,
      "learning_rate": 3.034822006472492e-06,
      "loss": 0.0162,
      "step": 327640
    },
    {
      "epoch": 127.24,
      "learning_rate": 3.0343042071197417e-06,
      "loss": 0.033,
      "step": 327650
    },
    {
      "epoch": 127.25,
      "learning_rate": 3.0337864077669905e-06,
      "loss": 0.0153,
      "step": 327660
    },
    {
      "epoch": 127.25,
      "learning_rate": 3.03326860841424e-06,
      "loss": 0.0326,
      "step": 327670
    },
    {
      "epoch": 127.25,
      "learning_rate": 3.032750809061489e-06,
      "loss": 0.0592,
      "step": 327680
    },
    {
      "epoch": 127.26,
      "learning_rate": 3.032233009708738e-06,
      "loss": 0.0005,
      "step": 327690
    },
    {
      "epoch": 127.26,
      "learning_rate": 3.0317152103559872e-06,
      "loss": 0.0785,
      "step": 327700
    },
    {
      "epoch": 127.27,
      "learning_rate": 3.0311974110032364e-06,
      "loss": 0.0124,
      "step": 327710
    },
    {
      "epoch": 127.27,
      "learning_rate": 3.0306796116504856e-06,
      "loss": 0.0421,
      "step": 327720
    },
    {
      "epoch": 127.27,
      "learning_rate": 3.030161812297735e-06,
      "loss": 0.0002,
      "step": 327730
    },
    {
      "epoch": 127.28,
      "learning_rate": 3.029644012944984e-06,
      "loss": 0.0575,
      "step": 327740
    },
    {
      "epoch": 127.28,
      "learning_rate": 3.029126213592233e-06,
      "loss": 0.0097,
      "step": 327750
    },
    {
      "epoch": 127.29,
      "learning_rate": 3.0286084142394824e-06,
      "loss": 0.0706,
      "step": 327760
    },
    {
      "epoch": 127.29,
      "learning_rate": 3.0280906148867316e-06,
      "loss": 0.0238,
      "step": 327770
    },
    {
      "epoch": 127.29,
      "learning_rate": 3.0275728155339808e-06,
      "loss": 0.0537,
      "step": 327780
    },
    {
      "epoch": 127.3,
      "learning_rate": 3.02705501618123e-06,
      "loss": 0.0015,
      "step": 327790
    },
    {
      "epoch": 127.3,
      "learning_rate": 3.026537216828479e-06,
      "loss": 0.0193,
      "step": 327800
    },
    {
      "epoch": 127.3,
      "learning_rate": 3.0260194174757284e-06,
      "loss": 0.11,
      "step": 327810
    },
    {
      "epoch": 127.31,
      "learning_rate": 3.0255016181229776e-06,
      "loss": 0.0103,
      "step": 327820
    },
    {
      "epoch": 127.31,
      "learning_rate": 3.0249838187702268e-06,
      "loss": 0.1914,
      "step": 327830
    },
    {
      "epoch": 127.32,
      "learning_rate": 3.024466019417476e-06,
      "loss": 0.092,
      "step": 327840
    },
    {
      "epoch": 127.32,
      "learning_rate": 3.023948220064725e-06,
      "loss": 0.0087,
      "step": 327850
    },
    {
      "epoch": 127.32,
      "learning_rate": 3.023430420711974e-06,
      "loss": 0.0002,
      "step": 327860
    },
    {
      "epoch": 127.33,
      "learning_rate": 3.0229126213592235e-06,
      "loss": 0.0001,
      "step": 327870
    },
    {
      "epoch": 127.33,
      "learning_rate": 3.0223948220064723e-06,
      "loss": 0.0397,
      "step": 327880
    },
    {
      "epoch": 127.34,
      "learning_rate": 3.021877022653722e-06,
      "loss": 0.0882,
      "step": 327890
    },
    {
      "epoch": 127.34,
      "learning_rate": 3.0213592233009707e-06,
      "loss": 0.0721,
      "step": 327900
    },
    {
      "epoch": 127.34,
      "learning_rate": 3.0208414239482203e-06,
      "loss": 0.0513,
      "step": 327910
    },
    {
      "epoch": 127.35,
      "learning_rate": 3.020323624595469e-06,
      "loss": 0.0003,
      "step": 327920
    },
    {
      "epoch": 127.35,
      "learning_rate": 3.0198058252427187e-06,
      "loss": 0.0352,
      "step": 327930
    },
    {
      "epoch": 127.36,
      "learning_rate": 3.0192880258899675e-06,
      "loss": 0.0211,
      "step": 327940
    },
    {
      "epoch": 127.36,
      "learning_rate": 3.018770226537217e-06,
      "loss": 0.1231,
      "step": 327950
    },
    {
      "epoch": 127.36,
      "learning_rate": 3.0182524271844667e-06,
      "loss": 0.0064,
      "step": 327960
    },
    {
      "epoch": 127.37,
      "learning_rate": 3.0177346278317155e-06,
      "loss": 0.0499,
      "step": 327970
    },
    {
      "epoch": 127.37,
      "learning_rate": 3.017216828478965e-06,
      "loss": 0.0885,
      "step": 327980
    },
    {
      "epoch": 127.37,
      "learning_rate": 3.016699029126214e-06,
      "loss": 0.0966,
      "step": 327990
    },
    {
      "epoch": 127.38,
      "learning_rate": 3.0161812297734635e-06,
      "loss": 0.0206,
      "step": 328000
    },
    {
      "epoch": 127.38,
      "learning_rate": 3.0156634304207123e-06,
      "loss": 0.082,
      "step": 328010
    },
    {
      "epoch": 127.39,
      "learning_rate": 3.0151456310679615e-06,
      "loss": 0.0005,
      "step": 328020
    },
    {
      "epoch": 127.39,
      "learning_rate": 3.0146278317152106e-06,
      "loss": 0.0306,
      "step": 328030
    },
    {
      "epoch": 127.39,
      "learning_rate": 3.01411003236246e-06,
      "loss": 0.0109,
      "step": 328040
    },
    {
      "epoch": 127.4,
      "learning_rate": 3.013592233009709e-06,
      "loss": 0.1121,
      "step": 328050
    },
    {
      "epoch": 127.4,
      "learning_rate": 3.0130744336569582e-06,
      "loss": 0.0011,
      "step": 328060
    },
    {
      "epoch": 127.41,
      "learning_rate": 3.0125566343042074e-06,
      "loss": 0.0084,
      "step": 328070
    },
    {
      "epoch": 127.41,
      "learning_rate": 3.0120388349514566e-06,
      "loss": 0.0345,
      "step": 328080
    },
    {
      "epoch": 127.41,
      "learning_rate": 3.011521035598706e-06,
      "loss": 0.0181,
      "step": 328090
    },
    {
      "epoch": 127.42,
      "learning_rate": 3.011003236245955e-06,
      "loss": 0.0377,
      "step": 328100
    },
    {
      "epoch": 127.42,
      "learning_rate": 3.010485436893204e-06,
      "loss": 0.1132,
      "step": 328110
    },
    {
      "epoch": 127.43,
      "learning_rate": 3.0099676375404534e-06,
      "loss": 0.0212,
      "step": 328120
    },
    {
      "epoch": 127.43,
      "learning_rate": 3.0094498381877026e-06,
      "loss": 0.0175,
      "step": 328130
    },
    {
      "epoch": 127.43,
      "learning_rate": 3.0089320388349518e-06,
      "loss": 0.1046,
      "step": 328140
    },
    {
      "epoch": 127.44,
      "learning_rate": 3.008414239482201e-06,
      "loss": 0.0283,
      "step": 328150
    },
    {
      "epoch": 127.44,
      "learning_rate": 3.00789644012945e-06,
      "loss": 0.0111,
      "step": 328160
    },
    {
      "epoch": 127.44,
      "learning_rate": 3.0073786407766994e-06,
      "loss": 0.0083,
      "step": 328170
    },
    {
      "epoch": 127.45,
      "learning_rate": 3.0068608414239486e-06,
      "loss": 0.001,
      "step": 328180
    },
    {
      "epoch": 127.45,
      "learning_rate": 3.0063430420711973e-06,
      "loss": 0.0363,
      "step": 328190
    },
    {
      "epoch": 127.46,
      "learning_rate": 3.005825242718447e-06,
      "loss": 0.1108,
      "step": 328200
    },
    {
      "epoch": 127.46,
      "learning_rate": 3.0053074433656957e-06,
      "loss": 0.0033,
      "step": 328210
    },
    {
      "epoch": 127.46,
      "learning_rate": 3.0047896440129453e-06,
      "loss": 0.0014,
      "step": 328220
    },
    {
      "epoch": 127.47,
      "learning_rate": 3.004271844660194e-06,
      "loss": 0.1015,
      "step": 328230
    },
    {
      "epoch": 127.47,
      "learning_rate": 3.0037540453074437e-06,
      "loss": 0.1147,
      "step": 328240
    },
    {
      "epoch": 127.48,
      "learning_rate": 3.0032362459546925e-06,
      "loss": 0.0747,
      "step": 328250
    },
    {
      "epoch": 127.48,
      "learning_rate": 3.002718446601942e-06,
      "loss": 0.089,
      "step": 328260
    },
    {
      "epoch": 127.48,
      "learning_rate": 3.002200647249191e-06,
      "loss": 0.011,
      "step": 328270
    },
    {
      "epoch": 127.49,
      "learning_rate": 3.0016828478964405e-06,
      "loss": 0.0092,
      "step": 328280
    },
    {
      "epoch": 127.49,
      "learning_rate": 3.0011650485436893e-06,
      "loss": 0.1088,
      "step": 328290
    },
    {
      "epoch": 127.5,
      "learning_rate": 3.000647249190939e-06,
      "loss": 0.0049,
      "step": 328300
    },
    {
      "epoch": 127.5,
      "learning_rate": 3.0001294498381877e-06,
      "loss": 0.053,
      "step": 328310
    },
    {
      "epoch": 127.5,
      "learning_rate": 2.9996116504854373e-06,
      "loss": 0.0026,
      "step": 328320
    },
    {
      "epoch": 127.51,
      "learning_rate": 2.999093851132686e-06,
      "loss": 0.0492,
      "step": 328330
    },
    {
      "epoch": 127.51,
      "learning_rate": 2.9985760517799357e-06,
      "loss": 0.0096,
      "step": 328340
    },
    {
      "epoch": 127.51,
      "learning_rate": 2.9980582524271844e-06,
      "loss": 0.0235,
      "step": 328350
    },
    {
      "epoch": 127.52,
      "learning_rate": 2.997540453074434e-06,
      "loss": 0.0358,
      "step": 328360
    },
    {
      "epoch": 127.52,
      "learning_rate": 2.997022653721683e-06,
      "loss": 0.058,
      "step": 328370
    },
    {
      "epoch": 127.53,
      "learning_rate": 2.9965048543689325e-06,
      "loss": 0.039,
      "step": 328380
    },
    {
      "epoch": 127.53,
      "learning_rate": 2.9959870550161812e-06,
      "loss": 0.0086,
      "step": 328390
    },
    {
      "epoch": 127.53,
      "learning_rate": 2.995469255663431e-06,
      "loss": 0.0938,
      "step": 328400
    },
    {
      "epoch": 127.54,
      "learning_rate": 2.9949514563106796e-06,
      "loss": 0.0029,
      "step": 328410
    },
    {
      "epoch": 127.54,
      "learning_rate": 2.9944336569579292e-06,
      "loss": 0.2031,
      "step": 328420
    },
    {
      "epoch": 127.55,
      "learning_rate": 2.993915857605178e-06,
      "loss": 0.056,
      "step": 328430
    },
    {
      "epoch": 127.55,
      "learning_rate": 2.9933980582524276e-06,
      "loss": 0.04,
      "step": 328440
    },
    {
      "epoch": 127.55,
      "learning_rate": 2.9928802588996764e-06,
      "loss": 0.0177,
      "step": 328450
    },
    {
      "epoch": 127.56,
      "learning_rate": 2.992362459546926e-06,
      "loss": 0.1095,
      "step": 328460
    },
    {
      "epoch": 127.56,
      "learning_rate": 2.9918446601941748e-06,
      "loss": 0.0094,
      "step": 328470
    },
    {
      "epoch": 127.57,
      "learning_rate": 2.9913268608414244e-06,
      "loss": 0.0299,
      "step": 328480
    },
    {
      "epoch": 127.57,
      "learning_rate": 2.990809061488673e-06,
      "loss": 0.0086,
      "step": 328490
    },
    {
      "epoch": 127.57,
      "learning_rate": 2.9902912621359224e-06,
      "loss": 0.0462,
      "step": 328500
    },
    {
      "epoch": 127.58,
      "learning_rate": 2.989773462783172e-06,
      "loss": 0.0109,
      "step": 328510
    },
    {
      "epoch": 127.58,
      "learning_rate": 2.9892556634304208e-06,
      "loss": 0.1009,
      "step": 328520
    },
    {
      "epoch": 127.58,
      "learning_rate": 2.9887378640776704e-06,
      "loss": 0.0268,
      "step": 328530
    },
    {
      "epoch": 127.59,
      "learning_rate": 2.988220064724919e-06,
      "loss": 0.0713,
      "step": 328540
    },
    {
      "epoch": 127.59,
      "learning_rate": 2.9877022653721688e-06,
      "loss": 0.0759,
      "step": 328550
    },
    {
      "epoch": 127.6,
      "learning_rate": 2.9871844660194175e-06,
      "loss": 0.0644,
      "step": 328560
    },
    {
      "epoch": 127.6,
      "learning_rate": 2.986666666666667e-06,
      "loss": 0.14,
      "step": 328570
    },
    {
      "epoch": 127.6,
      "learning_rate": 2.986148867313916e-06,
      "loss": 0.0029,
      "step": 328580
    },
    {
      "epoch": 127.61,
      "learning_rate": 2.9856310679611655e-06,
      "loss": 0.0025,
      "step": 328590
    },
    {
      "epoch": 127.61,
      "learning_rate": 2.9851132686084143e-06,
      "loss": 0.0734,
      "step": 328600
    },
    {
      "epoch": 127.62,
      "learning_rate": 2.984595469255664e-06,
      "loss": 0.0652,
      "step": 328610
    },
    {
      "epoch": 127.62,
      "learning_rate": 2.9840776699029127e-06,
      "loss": 0.0007,
      "step": 328620
    },
    {
      "epoch": 127.62,
      "learning_rate": 2.9835598705501623e-06,
      "loss": 0.0811,
      "step": 328630
    },
    {
      "epoch": 127.63,
      "learning_rate": 2.983042071197411e-06,
      "loss": 0.0849,
      "step": 328640
    },
    {
      "epoch": 127.63,
      "learning_rate": 2.9825242718446607e-06,
      "loss": 0.0351,
      "step": 328650
    },
    {
      "epoch": 127.63,
      "learning_rate": 2.9820064724919095e-06,
      "loss": 0.0079,
      "step": 328660
    },
    {
      "epoch": 127.64,
      "learning_rate": 2.981488673139159e-06,
      "loss": 0.0987,
      "step": 328670
    },
    {
      "epoch": 127.64,
      "learning_rate": 2.980970873786408e-06,
      "loss": 0.0017,
      "step": 328680
    },
    {
      "epoch": 127.65,
      "learning_rate": 2.9804530744336575e-06,
      "loss": 0.001,
      "step": 328690
    },
    {
      "epoch": 127.65,
      "learning_rate": 2.9799352750809063e-06,
      "loss": 0.0007,
      "step": 328700
    },
    {
      "epoch": 127.65,
      "learning_rate": 2.979417475728156e-06,
      "loss": 0.0974,
      "step": 328710
    },
    {
      "epoch": 127.66,
      "learning_rate": 2.9788996763754046e-06,
      "loss": 0.0094,
      "step": 328720
    },
    {
      "epoch": 127.66,
      "learning_rate": 2.9783818770226543e-06,
      "loss": 0.0128,
      "step": 328730
    },
    {
      "epoch": 127.67,
      "learning_rate": 2.977864077669903e-06,
      "loss": 0.009,
      "step": 328740
    },
    {
      "epoch": 127.67,
      "learning_rate": 2.9773462783171527e-06,
      "loss": 0.0668,
      "step": 328750
    },
    {
      "epoch": 127.67,
      "learning_rate": 2.9768284789644014e-06,
      "loss": 0.1018,
      "step": 328760
    },
    {
      "epoch": 127.68,
      "learning_rate": 2.976310679611651e-06,
      "loss": 0.0259,
      "step": 328770
    },
    {
      "epoch": 127.68,
      "learning_rate": 2.9757928802589e-06,
      "loss": 0.0001,
      "step": 328780
    },
    {
      "epoch": 127.69,
      "learning_rate": 2.9752750809061494e-06,
      "loss": 0.0491,
      "step": 328790
    },
    {
      "epoch": 127.69,
      "learning_rate": 2.974757281553398e-06,
      "loss": 0.0905,
      "step": 328800
    },
    {
      "epoch": 127.69,
      "learning_rate": 2.974239482200648e-06,
      "loss": 0.0001,
      "step": 328810
    },
    {
      "epoch": 127.7,
      "learning_rate": 2.9737216828478966e-06,
      "loss": 0.0841,
      "step": 328820
    },
    {
      "epoch": 127.7,
      "learning_rate": 2.9732038834951458e-06,
      "loss": 0.0006,
      "step": 328830
    },
    {
      "epoch": 127.7,
      "learning_rate": 2.972686084142395e-06,
      "loss": 0.0127,
      "step": 328840
    },
    {
      "epoch": 127.71,
      "learning_rate": 2.972168284789644e-06,
      "loss": 0.0159,
      "step": 328850
    },
    {
      "epoch": 127.71,
      "learning_rate": 2.9716504854368934e-06,
      "loss": 0.0345,
      "step": 328860
    },
    {
      "epoch": 127.72,
      "learning_rate": 2.9711326860841426e-06,
      "loss": 0.0006,
      "step": 328870
    },
    {
      "epoch": 127.72,
      "learning_rate": 2.9706148867313918e-06,
      "loss": 0.0035,
      "step": 328880
    },
    {
      "epoch": 127.72,
      "learning_rate": 2.970097087378641e-06,
      "loss": 0.109,
      "step": 328890
    },
    {
      "epoch": 127.73,
      "learning_rate": 2.96957928802589e-06,
      "loss": 0.1205,
      "step": 328900
    },
    {
      "epoch": 127.73,
      "learning_rate": 2.9690614886731393e-06,
      "loss": 0.0856,
      "step": 328910
    },
    {
      "epoch": 127.74,
      "learning_rate": 2.9685436893203885e-06,
      "loss": 0.0664,
      "step": 328920
    },
    {
      "epoch": 127.74,
      "learning_rate": 2.9680258899676377e-06,
      "loss": 0.0018,
      "step": 328930
    },
    {
      "epoch": 127.74,
      "learning_rate": 2.967508090614887e-06,
      "loss": 0.0367,
      "step": 328940
    },
    {
      "epoch": 127.75,
      "learning_rate": 2.966990291262136e-06,
      "loss": 0.067,
      "step": 328950
    },
    {
      "epoch": 127.75,
      "learning_rate": 2.9664724919093853e-06,
      "loss": 0.1275,
      "step": 328960
    },
    {
      "epoch": 127.76,
      "learning_rate": 2.9659546925566345e-06,
      "loss": 0.0263,
      "step": 328970
    },
    {
      "epoch": 127.76,
      "learning_rate": 2.9654368932038837e-06,
      "loss": 0.0182,
      "step": 328980
    },
    {
      "epoch": 127.76,
      "learning_rate": 2.964919093851133e-06,
      "loss": 0.0032,
      "step": 328990
    },
    {
      "epoch": 127.77,
      "learning_rate": 2.9644012944983817e-06,
      "loss": 0.0106,
      "step": 329000
    },
    {
      "epoch": 127.77,
      "learning_rate": 2.9638834951456313e-06,
      "loss": 0.0325,
      "step": 329010
    },
    {
      "epoch": 127.77,
      "learning_rate": 2.96336569579288e-06,
      "loss": 0.0005,
      "step": 329020
    },
    {
      "epoch": 127.78,
      "learning_rate": 2.9628478964401297e-06,
      "loss": 0.0836,
      "step": 329030
    },
    {
      "epoch": 127.78,
      "learning_rate": 2.9623300970873784e-06,
      "loss": 0.0444,
      "step": 329040
    },
    {
      "epoch": 127.79,
      "learning_rate": 2.961812297734628e-06,
      "loss": 0.1286,
      "step": 329050
    },
    {
      "epoch": 127.79,
      "learning_rate": 2.961294498381877e-06,
      "loss": 0.0012,
      "step": 329060
    },
    {
      "epoch": 127.79,
      "learning_rate": 2.9607766990291265e-06,
      "loss": 0.124,
      "step": 329070
    },
    {
      "epoch": 127.8,
      "learning_rate": 2.960258899676376e-06,
      "loss": 0.1073,
      "step": 329080
    },
    {
      "epoch": 127.8,
      "learning_rate": 2.959741100323625e-06,
      "loss": 0.0582,
      "step": 329090
    },
    {
      "epoch": 127.81,
      "learning_rate": 2.9592233009708745e-06,
      "loss": 0.0232,
      "step": 329100
    },
    {
      "epoch": 127.81,
      "learning_rate": 2.9587055016181232e-06,
      "loss": 0.0724,
      "step": 329110
    },
    {
      "epoch": 127.81,
      "learning_rate": 2.958187702265373e-06,
      "loss": 0.1391,
      "step": 329120
    },
    {
      "epoch": 127.82,
      "learning_rate": 2.9576699029126216e-06,
      "loss": 0.1063,
      "step": 329130
    },
    {
      "epoch": 127.82,
      "learning_rate": 2.9571521035598712e-06,
      "loss": 0.0152,
      "step": 329140
    },
    {
      "epoch": 127.83,
      "learning_rate": 2.95663430420712e-06,
      "loss": 0.0892,
      "step": 329150
    },
    {
      "epoch": 127.83,
      "learning_rate": 2.956116504854369e-06,
      "loss": 0.1256,
      "step": 329160
    },
    {
      "epoch": 127.83,
      "learning_rate": 2.9555987055016184e-06,
      "loss": 0.0795,
      "step": 329170
    },
    {
      "epoch": 127.84,
      "learning_rate": 2.9550809061488676e-06,
      "loss": 0.0511,
      "step": 329180
    },
    {
      "epoch": 127.84,
      "learning_rate": 2.954563106796117e-06,
      "loss": 0.0262,
      "step": 329190
    },
    {
      "epoch": 127.84,
      "learning_rate": 2.954045307443366e-06,
      "loss": 0.0559,
      "step": 329200
    },
    {
      "epoch": 127.85,
      "learning_rate": 2.953527508090615e-06,
      "loss": 0.0684,
      "step": 329210
    },
    {
      "epoch": 127.85,
      "learning_rate": 2.9530097087378644e-06,
      "loss": 0.099,
      "step": 329220
    },
    {
      "epoch": 127.86,
      "learning_rate": 2.9524919093851136e-06,
      "loss": 0.0006,
      "step": 329230
    },
    {
      "epoch": 127.86,
      "learning_rate": 2.9519741100323628e-06,
      "loss": 0.0133,
      "step": 329240
    },
    {
      "epoch": 127.86,
      "learning_rate": 2.951456310679612e-06,
      "loss": 0.0959,
      "step": 329250
    },
    {
      "epoch": 127.87,
      "learning_rate": 2.950938511326861e-06,
      "loss": 0.0003,
      "step": 329260
    },
    {
      "epoch": 127.87,
      "learning_rate": 2.9504207119741103e-06,
      "loss": 0.0285,
      "step": 329270
    },
    {
      "epoch": 127.88,
      "learning_rate": 2.9499029126213595e-06,
      "loss": 0.0589,
      "step": 329280
    },
    {
      "epoch": 127.88,
      "learning_rate": 2.9493851132686087e-06,
      "loss": 0.0004,
      "step": 329290
    },
    {
      "epoch": 127.88,
      "learning_rate": 2.948867313915858e-06,
      "loss": 0.0012,
      "step": 329300
    },
    {
      "epoch": 127.89,
      "learning_rate": 2.948349514563107e-06,
      "loss": 0.0093,
      "step": 329310
    },
    {
      "epoch": 127.89,
      "learning_rate": 2.9478317152103563e-06,
      "loss": 0.1011,
      "step": 329320
    },
    {
      "epoch": 127.9,
      "learning_rate": 2.947313915857605e-06,
      "loss": 0.0026,
      "step": 329330
    },
    {
      "epoch": 127.9,
      "learning_rate": 2.9467961165048547e-06,
      "loss": 0.0208,
      "step": 329340
    },
    {
      "epoch": 127.9,
      "learning_rate": 2.9462783171521035e-06,
      "loss": 0.0224,
      "step": 329350
    },
    {
      "epoch": 127.91,
      "learning_rate": 2.945760517799353e-06,
      "loss": 0.0102,
      "step": 329360
    },
    {
      "epoch": 127.91,
      "learning_rate": 2.945242718446602e-06,
      "loss": 0.0008,
      "step": 329370
    },
    {
      "epoch": 127.91,
      "learning_rate": 2.9447249190938515e-06,
      "loss": 0.0364,
      "step": 329380
    },
    {
      "epoch": 127.92,
      "learning_rate": 2.9442071197411003e-06,
      "loss": 0.0005,
      "step": 329390
    },
    {
      "epoch": 127.92,
      "learning_rate": 2.94368932038835e-06,
      "loss": 0.0218,
      "step": 329400
    },
    {
      "epoch": 127.93,
      "learning_rate": 2.9431715210355986e-06,
      "loss": 0.0002,
      "step": 329410
    },
    {
      "epoch": 127.93,
      "learning_rate": 2.9426537216828483e-06,
      "loss": 0.0924,
      "step": 329420
    },
    {
      "epoch": 127.93,
      "learning_rate": 2.942135922330097e-06,
      "loss": 0.0948,
      "step": 329430
    },
    {
      "epoch": 127.94,
      "learning_rate": 2.9416181229773467e-06,
      "loss": 0.0386,
      "step": 329440
    },
    {
      "epoch": 127.94,
      "learning_rate": 2.9411003236245954e-06,
      "loss": 0.0382,
      "step": 329450
    },
    {
      "epoch": 127.95,
      "learning_rate": 2.940582524271845e-06,
      "loss": 0.0759,
      "step": 329460
    },
    {
      "epoch": 127.95,
      "learning_rate": 2.940064724919094e-06,
      "loss": 0.0635,
      "step": 329470
    },
    {
      "epoch": 127.95,
      "learning_rate": 2.9395469255663434e-06,
      "loss": 0.0041,
      "step": 329480
    },
    {
      "epoch": 127.96,
      "learning_rate": 2.939029126213592e-06,
      "loss": 0.0309,
      "step": 329490
    },
    {
      "epoch": 127.96,
      "learning_rate": 2.938511326860842e-06,
      "loss": 0.0002,
      "step": 329500
    },
    {
      "epoch": 127.97,
      "learning_rate": 2.9379935275080906e-06,
      "loss": 0.0464,
      "step": 329510
    },
    {
      "epoch": 127.97,
      "learning_rate": 2.93747572815534e-06,
      "loss": 0.0762,
      "step": 329520
    },
    {
      "epoch": 127.97,
      "learning_rate": 2.936957928802589e-06,
      "loss": 0.0803,
      "step": 329530
    },
    {
      "epoch": 127.98,
      "learning_rate": 2.9364401294498386e-06,
      "loss": 0.0426,
      "step": 329540
    },
    {
      "epoch": 127.98,
      "learning_rate": 2.9359223300970874e-06,
      "loss": 0.0084,
      "step": 329550
    },
    {
      "epoch": 127.98,
      "learning_rate": 2.935404530744337e-06,
      "loss": 0.0002,
      "step": 329560
    },
    {
      "epoch": 127.99,
      "learning_rate": 2.9348867313915858e-06,
      "loss": 0.031,
      "step": 329570
    },
    {
      "epoch": 127.99,
      "learning_rate": 2.9343689320388354e-06,
      "loss": 0.0021,
      "step": 329580
    },
    {
      "epoch": 128.0,
      "learning_rate": 2.933851132686084e-06,
      "loss": 0.0001,
      "step": 329590
    },
    {
      "epoch": 128.0,
      "learning_rate": 2.9333333333333338e-06,
      "loss": 0.0082,
      "step": 329600
    },
    {
      "epoch": 128.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.38309401273727417,
      "eval_runtime": 8.2374,
      "eval_samples_per_second": 441.28,
      "eval_steps_per_second": 55.236,
      "step": 329600
    },
    {
      "epoch": 128.0,
      "learning_rate": 2.9328155339805825e-06,
      "loss": 0.0029,
      "step": 329610
    },
    {
      "epoch": 128.01,
      "learning_rate": 2.932297734627832e-06,
      "loss": 0.0942,
      "step": 329620
    },
    {
      "epoch": 128.01,
      "learning_rate": 2.9317799352750813e-06,
      "loss": 0.072,
      "step": 329630
    },
    {
      "epoch": 128.02,
      "learning_rate": 2.9312621359223305e-06,
      "loss": 0.0001,
      "step": 329640
    },
    {
      "epoch": 128.02,
      "learning_rate": 2.9307443365695797e-06,
      "loss": 0.0159,
      "step": 329650
    },
    {
      "epoch": 128.02,
      "learning_rate": 2.9302265372168285e-06,
      "loss": 0.0234,
      "step": 329660
    },
    {
      "epoch": 128.03,
      "learning_rate": 2.929708737864078e-06,
      "loss": 0.0112,
      "step": 329670
    },
    {
      "epoch": 128.03,
      "learning_rate": 2.929190938511327e-06,
      "loss": 0.0306,
      "step": 329680
    },
    {
      "epoch": 128.03,
      "learning_rate": 2.9286731391585765e-06,
      "loss": 0.0192,
      "step": 329690
    },
    {
      "epoch": 128.04,
      "learning_rate": 2.9281553398058253e-06,
      "loss": 0.0645,
      "step": 329700
    },
    {
      "epoch": 128.04,
      "learning_rate": 2.927637540453075e-06,
      "loss": 0.1185,
      "step": 329710
    },
    {
      "epoch": 128.05,
      "learning_rate": 2.9271197411003237e-06,
      "loss": 0.0852,
      "step": 329720
    },
    {
      "epoch": 128.05,
      "learning_rate": 2.9266019417475733e-06,
      "loss": 0.0721,
      "step": 329730
    },
    {
      "epoch": 128.05,
      "learning_rate": 2.926084142394822e-06,
      "loss": 0.0273,
      "step": 329740
    },
    {
      "epoch": 128.06,
      "learning_rate": 2.9255663430420717e-06,
      "loss": 0.0336,
      "step": 329750
    },
    {
      "epoch": 128.06,
      "learning_rate": 2.9250485436893205e-06,
      "loss": 0.0727,
      "step": 329760
    },
    {
      "epoch": 128.07,
      "learning_rate": 2.92453074433657e-06,
      "loss": 0.0087,
      "step": 329770
    },
    {
      "epoch": 128.07,
      "learning_rate": 2.924012944983819e-06,
      "loss": 0.012,
      "step": 329780
    },
    {
      "epoch": 128.07,
      "learning_rate": 2.9234951456310685e-06,
      "loss": 0.0996,
      "step": 329790
    },
    {
      "epoch": 128.08,
      "learning_rate": 2.9229773462783172e-06,
      "loss": 0.0245,
      "step": 329800
    },
    {
      "epoch": 128.08,
      "learning_rate": 2.922459546925567e-06,
      "loss": 0.0943,
      "step": 329810
    },
    {
      "epoch": 128.09,
      "learning_rate": 2.9219417475728156e-06,
      "loss": 0.0467,
      "step": 329820
    },
    {
      "epoch": 128.09,
      "learning_rate": 2.9214239482200652e-06,
      "loss": 0.0002,
      "step": 329830
    },
    {
      "epoch": 128.09,
      "learning_rate": 2.920906148867314e-06,
      "loss": 0.022,
      "step": 329840
    },
    {
      "epoch": 128.1,
      "learning_rate": 2.9203883495145636e-06,
      "loss": 0.0284,
      "step": 329850
    },
    {
      "epoch": 128.1,
      "learning_rate": 2.9198705501618124e-06,
      "loss": 0.0521,
      "step": 329860
    },
    {
      "epoch": 128.1,
      "learning_rate": 2.919352750809062e-06,
      "loss": 0.0546,
      "step": 329870
    },
    {
      "epoch": 128.11,
      "learning_rate": 2.918834951456311e-06,
      "loss": 0.036,
      "step": 329880
    },
    {
      "epoch": 128.11,
      "learning_rate": 2.9183171521035604e-06,
      "loss": 0.081,
      "step": 329890
    },
    {
      "epoch": 128.12,
      "learning_rate": 2.917799352750809e-06,
      "loss": 0.1636,
      "step": 329900
    },
    {
      "epoch": 128.12,
      "learning_rate": 2.917281553398059e-06,
      "loss": 0.1028,
      "step": 329910
    },
    {
      "epoch": 128.12,
      "learning_rate": 2.9167637540453076e-06,
      "loss": 0.0591,
      "step": 329920
    },
    {
      "epoch": 128.13,
      "learning_rate": 2.916245954692557e-06,
      "loss": 0.0001,
      "step": 329930
    },
    {
      "epoch": 128.13,
      "learning_rate": 2.915728155339806e-06,
      "loss": 0.1116,
      "step": 329940
    },
    {
      "epoch": 128.14,
      "learning_rate": 2.9152103559870556e-06,
      "loss": 0.0413,
      "step": 329950
    },
    {
      "epoch": 128.14,
      "learning_rate": 2.9146925566343043e-06,
      "loss": 0.0104,
      "step": 329960
    },
    {
      "epoch": 128.14,
      "learning_rate": 2.9141747572815535e-06,
      "loss": 0.0197,
      "step": 329970
    },
    {
      "epoch": 128.15,
      "learning_rate": 2.9136569579288027e-06,
      "loss": 0.001,
      "step": 329980
    },
    {
      "epoch": 128.15,
      "learning_rate": 2.913139158576052e-06,
      "loss": 0.0148,
      "step": 329990
    },
    {
      "epoch": 128.16,
      "learning_rate": 2.912621359223301e-06,
      "loss": 0.0536,
      "step": 330000
    },
    {
      "epoch": 128.16,
      "learning_rate": 2.9121035598705503e-06,
      "loss": 0.0207,
      "step": 330010
    },
    {
      "epoch": 128.16,
      "learning_rate": 2.9115857605177995e-06,
      "loss": 0.037,
      "step": 330020
    },
    {
      "epoch": 128.17,
      "learning_rate": 2.9110679611650487e-06,
      "loss": 0.0012,
      "step": 330030
    },
    {
      "epoch": 128.17,
      "learning_rate": 2.910550161812298e-06,
      "loss": 0.03,
      "step": 330040
    },
    {
      "epoch": 128.17,
      "learning_rate": 2.910032362459547e-06,
      "loss": 0.0228,
      "step": 330050
    },
    {
      "epoch": 128.18,
      "learning_rate": 2.9095145631067963e-06,
      "loss": 0.0091,
      "step": 330060
    },
    {
      "epoch": 128.18,
      "learning_rate": 2.9089967637540455e-06,
      "loss": 0.0805,
      "step": 330070
    },
    {
      "epoch": 128.19,
      "learning_rate": 2.9084789644012947e-06,
      "loss": 0.0024,
      "step": 330080
    },
    {
      "epoch": 128.19,
      "learning_rate": 2.907961165048544e-06,
      "loss": 0.0461,
      "step": 330090
    },
    {
      "epoch": 128.19,
      "learning_rate": 2.907443365695793e-06,
      "loss": 0.0445,
      "step": 330100
    },
    {
      "epoch": 128.2,
      "learning_rate": 2.9069255663430423e-06,
      "loss": 0.0139,
      "step": 330110
    },
    {
      "epoch": 128.2,
      "learning_rate": 2.9064077669902915e-06,
      "loss": 0.0117,
      "step": 330120
    },
    {
      "epoch": 128.21,
      "learning_rate": 2.9058899676375407e-06,
      "loss": 0.0018,
      "step": 330130
    },
    {
      "epoch": 128.21,
      "learning_rate": 2.9053721682847894e-06,
      "loss": 0.0192,
      "step": 330140
    },
    {
      "epoch": 128.21,
      "learning_rate": 2.904854368932039e-06,
      "loss": 0.0269,
      "step": 330150
    },
    {
      "epoch": 128.22,
      "learning_rate": 2.904336569579288e-06,
      "loss": 0.0857,
      "step": 330160
    },
    {
      "epoch": 128.22,
      "learning_rate": 2.9038187702265374e-06,
      "loss": 0.0351,
      "step": 330170
    },
    {
      "epoch": 128.23,
      "learning_rate": 2.903300970873786e-06,
      "loss": 0.0812,
      "step": 330180
    },
    {
      "epoch": 128.23,
      "learning_rate": 2.902783171521036e-06,
      "loss": 0.1412,
      "step": 330190
    },
    {
      "epoch": 128.23,
      "learning_rate": 2.9022653721682854e-06,
      "loss": 0.0661,
      "step": 330200
    },
    {
      "epoch": 128.24,
      "learning_rate": 2.901747572815534e-06,
      "loss": 0.0599,
      "step": 330210
    },
    {
      "epoch": 128.24,
      "learning_rate": 2.901229773462784e-06,
      "loss": 0.0743,
      "step": 330220
    },
    {
      "epoch": 128.24,
      "learning_rate": 2.9007119741100326e-06,
      "loss": 0.0004,
      "step": 330230
    },
    {
      "epoch": 128.25,
      "learning_rate": 2.9001941747572822e-06,
      "loss": 0.0303,
      "step": 330240
    },
    {
      "epoch": 128.25,
      "learning_rate": 2.899676375404531e-06,
      "loss": 0.0615,
      "step": 330250
    },
    {
      "epoch": 128.26,
      "learning_rate": 2.8991585760517806e-06,
      "loss": 0.0512,
      "step": 330260
    },
    {
      "epoch": 128.26,
      "learning_rate": 2.8986407766990294e-06,
      "loss": 0.0014,
      "step": 330270
    },
    {
      "epoch": 128.26,
      "learning_rate": 2.898122977346279e-06,
      "loss": 0.04,
      "step": 330280
    },
    {
      "epoch": 128.27,
      "learning_rate": 2.8976051779935278e-06,
      "loss": 0.0303,
      "step": 330290
    },
    {
      "epoch": 128.27,
      "learning_rate": 2.897087378640777e-06,
      "loss": 0.1007,
      "step": 330300
    },
    {
      "epoch": 128.28,
      "learning_rate": 2.896569579288026e-06,
      "loss": 0.0216,
      "step": 330310
    },
    {
      "epoch": 128.28,
      "learning_rate": 2.8960517799352753e-06,
      "loss": 0.0196,
      "step": 330320
    },
    {
      "epoch": 128.28,
      "learning_rate": 2.8955339805825245e-06,
      "loss": 0.0671,
      "step": 330330
    },
    {
      "epoch": 128.29,
      "learning_rate": 2.8950161812297737e-06,
      "loss": 0.0006,
      "step": 330340
    },
    {
      "epoch": 128.29,
      "learning_rate": 2.894498381877023e-06,
      "loss": 0.0712,
      "step": 330350
    },
    {
      "epoch": 128.3,
      "learning_rate": 2.893980582524272e-06,
      "loss": 0.0003,
      "step": 330360
    },
    {
      "epoch": 128.3,
      "learning_rate": 2.8934627831715213e-06,
      "loss": 0.0075,
      "step": 330370
    },
    {
      "epoch": 128.3,
      "learning_rate": 2.8929449838187705e-06,
      "loss": 0.0275,
      "step": 330380
    },
    {
      "epoch": 128.31,
      "learning_rate": 2.8924271844660197e-06,
      "loss": 0.0002,
      "step": 330390
    },
    {
      "epoch": 128.31,
      "learning_rate": 2.891909385113269e-06,
      "loss": 0.103,
      "step": 330400
    },
    {
      "epoch": 128.31,
      "learning_rate": 2.891391585760518e-06,
      "loss": 0.0395,
      "step": 330410
    },
    {
      "epoch": 128.32,
      "learning_rate": 2.8908737864077673e-06,
      "loss": 0.0978,
      "step": 330420
    },
    {
      "epoch": 128.32,
      "learning_rate": 2.8903559870550165e-06,
      "loss": 0.1139,
      "step": 330430
    },
    {
      "epoch": 128.33,
      "learning_rate": 2.8898381877022657e-06,
      "loss": 0.0748,
      "step": 330440
    },
    {
      "epoch": 128.33,
      "learning_rate": 2.889320388349515e-06,
      "loss": 0.1188,
      "step": 330450
    },
    {
      "epoch": 128.33,
      "learning_rate": 2.888802588996764e-06,
      "loss": 0.0663,
      "step": 330460
    },
    {
      "epoch": 128.34,
      "learning_rate": 2.888284789644013e-06,
      "loss": 0.0121,
      "step": 330470
    },
    {
      "epoch": 128.34,
      "learning_rate": 2.8877669902912625e-06,
      "loss": 0.021,
      "step": 330480
    },
    {
      "epoch": 128.35,
      "learning_rate": 2.8872491909385112e-06,
      "loss": 0.0436,
      "step": 330490
    },
    {
      "epoch": 128.35,
      "learning_rate": 2.886731391585761e-06,
      "loss": 0.0001,
      "step": 330500
    },
    {
      "epoch": 128.35,
      "learning_rate": 2.8862135922330096e-06,
      "loss": 0.1789,
      "step": 330510
    },
    {
      "epoch": 128.36,
      "learning_rate": 2.8856957928802592e-06,
      "loss": 0.1138,
      "step": 330520
    },
    {
      "epoch": 128.36,
      "learning_rate": 2.885177993527508e-06,
      "loss": 0.0004,
      "step": 330530
    },
    {
      "epoch": 128.37,
      "learning_rate": 2.8846601941747576e-06,
      "loss": 0.0659,
      "step": 330540
    },
    {
      "epoch": 128.37,
      "learning_rate": 2.8841423948220064e-06,
      "loss": 0.0826,
      "step": 330550
    },
    {
      "epoch": 128.37,
      "learning_rate": 2.883624595469256e-06,
      "loss": 0.0008,
      "step": 330560
    },
    {
      "epoch": 128.38,
      "learning_rate": 2.8831067961165048e-06,
      "loss": 0.135,
      "step": 330570
    },
    {
      "epoch": 128.38,
      "learning_rate": 2.8825889967637544e-06,
      "loss": 0.0121,
      "step": 330580
    },
    {
      "epoch": 128.38,
      "learning_rate": 2.882071197411003e-06,
      "loss": 0.0694,
      "step": 330590
    },
    {
      "epoch": 128.39,
      "learning_rate": 2.881553398058253e-06,
      "loss": 0.0872,
      "step": 330600
    },
    {
      "epoch": 128.39,
      "learning_rate": 2.8810355987055016e-06,
      "loss": 0.0397,
      "step": 330610
    },
    {
      "epoch": 128.4,
      "learning_rate": 2.880517799352751e-06,
      "loss": 0.0425,
      "step": 330620
    },
    {
      "epoch": 128.4,
      "learning_rate": 2.88e-06,
      "loss": 0.0297,
      "step": 330630
    },
    {
      "epoch": 128.4,
      "learning_rate": 2.8794822006472496e-06,
      "loss": 0.1196,
      "step": 330640
    },
    {
      "epoch": 128.41,
      "learning_rate": 2.8789644012944983e-06,
      "loss": 0.0396,
      "step": 330650
    },
    {
      "epoch": 128.41,
      "learning_rate": 2.878446601941748e-06,
      "loss": 0.0086,
      "step": 330660
    },
    {
      "epoch": 128.42,
      "learning_rate": 2.8779288025889967e-06,
      "loss": 0.0249,
      "step": 330670
    },
    {
      "epoch": 128.42,
      "learning_rate": 2.8774110032362464e-06,
      "loss": 0.0408,
      "step": 330680
    },
    {
      "epoch": 128.42,
      "learning_rate": 2.876893203883495e-06,
      "loss": 0.0469,
      "step": 330690
    },
    {
      "epoch": 128.43,
      "learning_rate": 2.8763754045307447e-06,
      "loss": 0.0337,
      "step": 330700
    },
    {
      "epoch": 128.43,
      "learning_rate": 2.8758576051779935e-06,
      "loss": 0.1987,
      "step": 330710
    },
    {
      "epoch": 128.43,
      "learning_rate": 2.875339805825243e-06,
      "loss": 0.0498,
      "step": 330720
    },
    {
      "epoch": 128.44,
      "learning_rate": 2.874822006472492e-06,
      "loss": 0.0091,
      "step": 330730
    },
    {
      "epoch": 128.44,
      "learning_rate": 2.8743042071197415e-06,
      "loss": 0.0921,
      "step": 330740
    },
    {
      "epoch": 128.45,
      "learning_rate": 2.8737864077669903e-06,
      "loss": 0.0001,
      "step": 330750
    },
    {
      "epoch": 128.45,
      "learning_rate": 2.87326860841424e-06,
      "loss": 0.0165,
      "step": 330760
    },
    {
      "epoch": 128.45,
      "learning_rate": 2.872750809061489e-06,
      "loss": 0.002,
      "step": 330770
    },
    {
      "epoch": 128.46,
      "learning_rate": 2.8722330097087383e-06,
      "loss": 0.0018,
      "step": 330780
    },
    {
      "epoch": 128.46,
      "learning_rate": 2.8717152103559875e-06,
      "loss": 0.0677,
      "step": 330790
    },
    {
      "epoch": 128.47,
      "learning_rate": 2.8711974110032363e-06,
      "loss": 0.1144,
      "step": 330800
    },
    {
      "epoch": 128.47,
      "learning_rate": 2.870679611650486e-06,
      "loss": 0.0484,
      "step": 330810
    },
    {
      "epoch": 128.47,
      "learning_rate": 2.8701618122977347e-06,
      "loss": 0.0114,
      "step": 330820
    },
    {
      "epoch": 128.48,
      "learning_rate": 2.8696440129449843e-06,
      "loss": 0.0009,
      "step": 330830
    },
    {
      "epoch": 128.48,
      "learning_rate": 2.869126213592233e-06,
      "loss": 0.038,
      "step": 330840
    },
    {
      "epoch": 128.49,
      "learning_rate": 2.8686084142394827e-06,
      "loss": 0.067,
      "step": 330850
    },
    {
      "epoch": 128.49,
      "learning_rate": 2.8680906148867314e-06,
      "loss": 0.0416,
      "step": 330860
    },
    {
      "epoch": 128.49,
      "learning_rate": 2.867572815533981e-06,
      "loss": 0.0801,
      "step": 330870
    },
    {
      "epoch": 128.5,
      "learning_rate": 2.86705501618123e-06,
      "loss": 0.021,
      "step": 330880
    },
    {
      "epoch": 128.5,
      "learning_rate": 2.8665372168284794e-06,
      "loss": 0.0206,
      "step": 330890
    },
    {
      "epoch": 128.5,
      "learning_rate": 2.866019417475728e-06,
      "loss": 0.0375,
      "step": 330900
    },
    {
      "epoch": 128.51,
      "learning_rate": 2.865501618122978e-06,
      "loss": 0.0164,
      "step": 330910
    },
    {
      "epoch": 128.51,
      "learning_rate": 2.8649838187702266e-06,
      "loss": 0.0001,
      "step": 330920
    },
    {
      "epoch": 128.52,
      "learning_rate": 2.8644660194174762e-06,
      "loss": 0.0707,
      "step": 330930
    },
    {
      "epoch": 128.52,
      "learning_rate": 2.863948220064725e-06,
      "loss": 0.0008,
      "step": 330940
    },
    {
      "epoch": 128.52,
      "learning_rate": 2.8634304207119746e-06,
      "loss": 0.0149,
      "step": 330950
    },
    {
      "epoch": 128.53,
      "learning_rate": 2.8629126213592234e-06,
      "loss": 0.0392,
      "step": 330960
    },
    {
      "epoch": 128.53,
      "learning_rate": 2.862394822006473e-06,
      "loss": 0.0208,
      "step": 330970
    },
    {
      "epoch": 128.54,
      "learning_rate": 2.8618770226537218e-06,
      "loss": 0.0279,
      "step": 330980
    },
    {
      "epoch": 128.54,
      "learning_rate": 2.8613592233009714e-06,
      "loss": 0.0632,
      "step": 330990
    },
    {
      "epoch": 128.54,
      "learning_rate": 2.86084142394822e-06,
      "loss": 0.0109,
      "step": 331000
    },
    {
      "epoch": 128.55,
      "learning_rate": 2.8603236245954698e-06,
      "loss": 0.0185,
      "step": 331010
    },
    {
      "epoch": 128.55,
      "learning_rate": 2.8598058252427185e-06,
      "loss": 0.0508,
      "step": 331020
    },
    {
      "epoch": 128.56,
      "learning_rate": 2.859288025889968e-06,
      "loss": 0.0003,
      "step": 331030
    },
    {
      "epoch": 128.56,
      "learning_rate": 2.858770226537217e-06,
      "loss": 0.0519,
      "step": 331040
    },
    {
      "epoch": 128.56,
      "learning_rate": 2.8582524271844665e-06,
      "loss": 0.0127,
      "step": 331050
    },
    {
      "epoch": 128.57,
      "learning_rate": 2.8577346278317153e-06,
      "loss": 0.0099,
      "step": 331060
    },
    {
      "epoch": 128.57,
      "learning_rate": 2.857216828478965e-06,
      "loss": 0.0678,
      "step": 331070
    },
    {
      "epoch": 128.57,
      "learning_rate": 2.8566990291262137e-06,
      "loss": 0.0155,
      "step": 331080
    },
    {
      "epoch": 128.58,
      "learning_rate": 2.8561812297734633e-06,
      "loss": 0.1311,
      "step": 331090
    },
    {
      "epoch": 128.58,
      "learning_rate": 2.855663430420712e-06,
      "loss": 0.0006,
      "step": 331100
    },
    {
      "epoch": 128.59,
      "learning_rate": 2.8551456310679613e-06,
      "loss": 0.0395,
      "step": 331110
    },
    {
      "epoch": 128.59,
      "learning_rate": 2.8546278317152105e-06,
      "loss": 0.0148,
      "step": 331120
    },
    {
      "epoch": 128.59,
      "learning_rate": 2.8541100323624597e-06,
      "loss": 0.0239,
      "step": 331130
    },
    {
      "epoch": 128.6,
      "learning_rate": 2.853592233009709e-06,
      "loss": 0.0091,
      "step": 331140
    },
    {
      "epoch": 128.6,
      "learning_rate": 2.853074433656958e-06,
      "loss": 0.0093,
      "step": 331150
    },
    {
      "epoch": 128.61,
      "learning_rate": 2.8525566343042073e-06,
      "loss": 0.0943,
      "step": 331160
    },
    {
      "epoch": 128.61,
      "learning_rate": 2.8520388349514565e-06,
      "loss": 0.0168,
      "step": 331170
    },
    {
      "epoch": 128.61,
      "learning_rate": 2.8515210355987057e-06,
      "loss": 0.0303,
      "step": 331180
    },
    {
      "epoch": 128.62,
      "learning_rate": 2.851003236245955e-06,
      "loss": 0.205,
      "step": 331190
    },
    {
      "epoch": 128.62,
      "learning_rate": 2.850485436893204e-06,
      "loss": 0.0001,
      "step": 331200
    },
    {
      "epoch": 128.63,
      "learning_rate": 2.8499676375404532e-06,
      "loss": 0.0023,
      "step": 331210
    },
    {
      "epoch": 128.63,
      "learning_rate": 2.8494498381877024e-06,
      "loss": 0.0097,
      "step": 331220
    },
    {
      "epoch": 128.63,
      "learning_rate": 2.8489320388349516e-06,
      "loss": 0.0124,
      "step": 331230
    },
    {
      "epoch": 128.64,
      "learning_rate": 2.848414239482201e-06,
      "loss": 0.0302,
      "step": 331240
    },
    {
      "epoch": 128.64,
      "learning_rate": 2.84789644012945e-06,
      "loss": 0.0002,
      "step": 331250
    },
    {
      "epoch": 128.64,
      "learning_rate": 2.847378640776699e-06,
      "loss": 0.0201,
      "step": 331260
    },
    {
      "epoch": 128.65,
      "learning_rate": 2.8468608414239484e-06,
      "loss": 0.0579,
      "step": 331270
    },
    {
      "epoch": 128.65,
      "learning_rate": 2.846343042071197e-06,
      "loss": 0.0355,
      "step": 331280
    },
    {
      "epoch": 128.66,
      "learning_rate": 2.845825242718447e-06,
      "loss": 0.0594,
      "step": 331290
    },
    {
      "epoch": 128.66,
      "learning_rate": 2.8453074433656956e-06,
      "loss": 0.0417,
      "step": 331300
    },
    {
      "epoch": 128.66,
      "learning_rate": 2.844789644012945e-06,
      "loss": 0.0494,
      "step": 331310
    },
    {
      "epoch": 128.67,
      "learning_rate": 2.844271844660195e-06,
      "loss": 0.0028,
      "step": 331320
    },
    {
      "epoch": 128.67,
      "learning_rate": 2.8437540453074436e-06,
      "loss": 0.0001,
      "step": 331330
    },
    {
      "epoch": 128.68,
      "learning_rate": 2.843236245954693e-06,
      "loss": 0.0784,
      "step": 331340
    },
    {
      "epoch": 128.68,
      "learning_rate": 2.842718446601942e-06,
      "loss": 0.1356,
      "step": 331350
    },
    {
      "epoch": 128.68,
      "learning_rate": 2.8422006472491916e-06,
      "loss": 0.0018,
      "step": 331360
    },
    {
      "epoch": 128.69,
      "learning_rate": 2.8416828478964403e-06,
      "loss": 0.0715,
      "step": 331370
    },
    {
      "epoch": 128.69,
      "learning_rate": 2.84116504854369e-06,
      "loss": 0.0001,
      "step": 331380
    },
    {
      "epoch": 128.7,
      "learning_rate": 2.8406472491909387e-06,
      "loss": 0.1034,
      "step": 331390
    },
    {
      "epoch": 128.7,
      "learning_rate": 2.8401294498381884e-06,
      "loss": 0.0017,
      "step": 331400
    },
    {
      "epoch": 128.7,
      "learning_rate": 2.839611650485437e-06,
      "loss": 0.0066,
      "step": 331410
    },
    {
      "epoch": 128.71,
      "learning_rate": 2.8390938511326867e-06,
      "loss": 0.0432,
      "step": 331420
    },
    {
      "epoch": 128.71,
      "learning_rate": 2.8385760517799355e-06,
      "loss": 0.021,
      "step": 331430
    },
    {
      "epoch": 128.71,
      "learning_rate": 2.8380582524271847e-06,
      "loss": 0.0198,
      "step": 331440
    },
    {
      "epoch": 128.72,
      "learning_rate": 2.837540453074434e-06,
      "loss": 0.0292,
      "step": 331450
    },
    {
      "epoch": 128.72,
      "learning_rate": 2.837022653721683e-06,
      "loss": 0.0178,
      "step": 331460
    },
    {
      "epoch": 128.73,
      "learning_rate": 2.8365048543689323e-06,
      "loss": 0.0001,
      "step": 331470
    },
    {
      "epoch": 128.73,
      "learning_rate": 2.8359870550161815e-06,
      "loss": 0.0001,
      "step": 331480
    },
    {
      "epoch": 128.73,
      "learning_rate": 2.8354692556634307e-06,
      "loss": 0.0012,
      "step": 331490
    },
    {
      "epoch": 128.74,
      "learning_rate": 2.83495145631068e-06,
      "loss": 0.0101,
      "step": 331500
    },
    {
      "epoch": 128.74,
      "learning_rate": 2.834433656957929e-06,
      "loss": 0.0003,
      "step": 331510
    },
    {
      "epoch": 128.75,
      "learning_rate": 2.8339158576051783e-06,
      "loss": 0.028,
      "step": 331520
    },
    {
      "epoch": 128.75,
      "learning_rate": 2.8333980582524275e-06,
      "loss": 0.0124,
      "step": 331530
    },
    {
      "epoch": 128.75,
      "learning_rate": 2.8328802588996767e-06,
      "loss": 0.0366,
      "step": 331540
    },
    {
      "epoch": 128.76,
      "learning_rate": 2.832362459546926e-06,
      "loss": 0.0847,
      "step": 331550
    },
    {
      "epoch": 128.76,
      "learning_rate": 2.831844660194175e-06,
      "loss": 0.0001,
      "step": 331560
    },
    {
      "epoch": 128.77,
      "learning_rate": 2.8313268608414242e-06,
      "loss": 0.0248,
      "step": 331570
    },
    {
      "epoch": 128.77,
      "learning_rate": 2.8308090614886734e-06,
      "loss": 0.0566,
      "step": 331580
    },
    {
      "epoch": 128.77,
      "learning_rate": 2.8302912621359226e-06,
      "loss": 0.0112,
      "step": 331590
    },
    {
      "epoch": 128.78,
      "learning_rate": 2.829773462783172e-06,
      "loss": 0.0376,
      "step": 331600
    },
    {
      "epoch": 128.78,
      "learning_rate": 2.8292556634304206e-06,
      "loss": 0.0003,
      "step": 331610
    },
    {
      "epoch": 128.78,
      "learning_rate": 2.8287378640776702e-06,
      "loss": 0.0282,
      "step": 331620
    },
    {
      "epoch": 128.79,
      "learning_rate": 2.828220064724919e-06,
      "loss": 0.0475,
      "step": 331630
    },
    {
      "epoch": 128.79,
      "learning_rate": 2.8277022653721686e-06,
      "loss": 0.0207,
      "step": 331640
    },
    {
      "epoch": 128.8,
      "learning_rate": 2.8271844660194174e-06,
      "loss": 0.0338,
      "step": 331650
    },
    {
      "epoch": 128.8,
      "learning_rate": 2.826666666666667e-06,
      "loss": 0.0465,
      "step": 331660
    },
    {
      "epoch": 128.8,
      "learning_rate": 2.8261488673139158e-06,
      "loss": 0.0456,
      "step": 331670
    },
    {
      "epoch": 128.81,
      "learning_rate": 2.8256310679611654e-06,
      "loss": 0.1323,
      "step": 331680
    },
    {
      "epoch": 128.81,
      "learning_rate": 2.825113268608414e-06,
      "loss": 0.022,
      "step": 331690
    },
    {
      "epoch": 128.82,
      "learning_rate": 2.8245954692556638e-06,
      "loss": 0.0099,
      "step": 331700
    },
    {
      "epoch": 128.82,
      "learning_rate": 2.8240776699029125e-06,
      "loss": 0.1346,
      "step": 331710
    },
    {
      "epoch": 128.82,
      "learning_rate": 2.823559870550162e-06,
      "loss": 0.0221,
      "step": 331720
    },
    {
      "epoch": 128.83,
      "learning_rate": 2.823042071197411e-06,
      "loss": 0.0013,
      "step": 331730
    },
    {
      "epoch": 128.83,
      "learning_rate": 2.8225242718446605e-06,
      "loss": 0.0614,
      "step": 331740
    },
    {
      "epoch": 128.83,
      "learning_rate": 2.8220064724919093e-06,
      "loss": 0.0006,
      "step": 331750
    },
    {
      "epoch": 128.84,
      "learning_rate": 2.821488673139159e-06,
      "loss": 0.1223,
      "step": 331760
    },
    {
      "epoch": 128.84,
      "learning_rate": 2.8209708737864077e-06,
      "loss": 0.1119,
      "step": 331770
    },
    {
      "epoch": 128.85,
      "learning_rate": 2.8204530744336573e-06,
      "loss": 0.0587,
      "step": 331780
    },
    {
      "epoch": 128.85,
      "learning_rate": 2.819935275080906e-06,
      "loss": 0.0755,
      "step": 331790
    },
    {
      "epoch": 128.85,
      "learning_rate": 2.8194174757281557e-06,
      "loss": 0.0717,
      "step": 331800
    },
    {
      "epoch": 128.86,
      "learning_rate": 2.8188996763754045e-06,
      "loss": 0.0578,
      "step": 331810
    },
    {
      "epoch": 128.86,
      "learning_rate": 2.818381877022654e-06,
      "loss": 0.1022,
      "step": 331820
    },
    {
      "epoch": 128.87,
      "learning_rate": 2.817864077669903e-06,
      "loss": 0.0101,
      "step": 331830
    },
    {
      "epoch": 128.87,
      "learning_rate": 2.8173462783171525e-06,
      "loss": 0.0246,
      "step": 331840
    },
    {
      "epoch": 128.87,
      "learning_rate": 2.8168284789644013e-06,
      "loss": 0.1435,
      "step": 331850
    },
    {
      "epoch": 128.88,
      "learning_rate": 2.816310679611651e-06,
      "loss": 0.0075,
      "step": 331860
    },
    {
      "epoch": 128.88,
      "learning_rate": 2.8157928802588997e-06,
      "loss": 0.0084,
      "step": 331870
    },
    {
      "epoch": 128.89,
      "learning_rate": 2.8152750809061493e-06,
      "loss": 0.0356,
      "step": 331880
    },
    {
      "epoch": 128.89,
      "learning_rate": 2.8147572815533985e-06,
      "loss": 0.0474,
      "step": 331890
    },
    {
      "epoch": 128.89,
      "learning_rate": 2.8142394822006477e-06,
      "loss": 0.0001,
      "step": 331900
    },
    {
      "epoch": 128.9,
      "learning_rate": 2.813721682847897e-06,
      "loss": 0.0291,
      "step": 331910
    },
    {
      "epoch": 128.9,
      "learning_rate": 2.813203883495146e-06,
      "loss": 0.0108,
      "step": 331920
    },
    {
      "epoch": 128.9,
      "learning_rate": 2.8126860841423952e-06,
      "loss": 0.0186,
      "step": 331930
    },
    {
      "epoch": 128.91,
      "learning_rate": 2.812168284789644e-06,
      "loss": 0.0913,
      "step": 331940
    },
    {
      "epoch": 128.91,
      "learning_rate": 2.8116504854368936e-06,
      "loss": 0.0393,
      "step": 331950
    },
    {
      "epoch": 128.92,
      "learning_rate": 2.8111326860841424e-06,
      "loss": 0.0594,
      "step": 331960
    },
    {
      "epoch": 128.92,
      "learning_rate": 2.810614886731392e-06,
      "loss": 0.0465,
      "step": 331970
    },
    {
      "epoch": 128.92,
      "learning_rate": 2.810097087378641e-06,
      "loss": 0.0312,
      "step": 331980
    },
    {
      "epoch": 128.93,
      "learning_rate": 2.8095792880258904e-06,
      "loss": 0.0001,
      "step": 331990
    },
    {
      "epoch": 128.93,
      "learning_rate": 2.809061488673139e-06,
      "loss": 0.0037,
      "step": 332000
    },
    {
      "epoch": 128.94,
      "learning_rate": 2.808543689320389e-06,
      "loss": 0.0124,
      "step": 332010
    },
    {
      "epoch": 128.94,
      "learning_rate": 2.8080258899676376e-06,
      "loss": 0.0756,
      "step": 332020
    },
    {
      "epoch": 128.94,
      "learning_rate": 2.807508090614887e-06,
      "loss": 0.0006,
      "step": 332030
    },
    {
      "epoch": 128.95,
      "learning_rate": 2.806990291262136e-06,
      "loss": 0.1215,
      "step": 332040
    },
    {
      "epoch": 128.95,
      "learning_rate": 2.8064724919093856e-06,
      "loss": 0.0124,
      "step": 332050
    },
    {
      "epoch": 128.96,
      "learning_rate": 2.8059546925566343e-06,
      "loss": 0.0162,
      "step": 332060
    },
    {
      "epoch": 128.96,
      "learning_rate": 2.805436893203884e-06,
      "loss": 0.0734,
      "step": 332070
    },
    {
      "epoch": 128.96,
      "learning_rate": 2.8049190938511327e-06,
      "loss": 0.0257,
      "step": 332080
    },
    {
      "epoch": 128.97,
      "learning_rate": 2.8044012944983824e-06,
      "loss": 0.021,
      "step": 332090
    },
    {
      "epoch": 128.97,
      "learning_rate": 2.803883495145631e-06,
      "loss": 0.0037,
      "step": 332100
    },
    {
      "epoch": 128.97,
      "learning_rate": 2.8033656957928807e-06,
      "loss": 0.1202,
      "step": 332110
    },
    {
      "epoch": 128.98,
      "learning_rate": 2.8028478964401295e-06,
      "loss": 0.0494,
      "step": 332120
    },
    {
      "epoch": 128.98,
      "learning_rate": 2.802330097087379e-06,
      "loss": 0.0147,
      "step": 332130
    },
    {
      "epoch": 128.99,
      "learning_rate": 2.801812297734628e-06,
      "loss": 0.0496,
      "step": 332140
    },
    {
      "epoch": 128.99,
      "learning_rate": 2.8012944983818775e-06,
      "loss": 0.0618,
      "step": 332150
    },
    {
      "epoch": 128.99,
      "learning_rate": 2.8007766990291263e-06,
      "loss": 0.1299,
      "step": 332160
    },
    {
      "epoch": 129.0,
      "learning_rate": 2.800258899676376e-06,
      "loss": 0.0737,
      "step": 332170
    },
    {
      "epoch": 129.0,
      "eval_accuracy": 0.9524071526822558,
      "eval_loss": 0.3891226649284363,
      "eval_runtime": 8.3068,
      "eval_samples_per_second": 437.593,
      "eval_steps_per_second": 54.774,
      "step": 332175
    },
    {
      "epoch": 129.0,
      "learning_rate": 2.7997411003236247e-06,
      "loss": 0.0097,
      "step": 332180
    },
    {
      "epoch": 129.01,
      "learning_rate": 2.7992233009708743e-06,
      "loss": 0.0341,
      "step": 332190
    },
    {
      "epoch": 129.01,
      "learning_rate": 2.798705501618123e-06,
      "loss": 0.0363,
      "step": 332200
    },
    {
      "epoch": 129.01,
      "learning_rate": 2.7981877022653727e-06,
      "loss": 0.0202,
      "step": 332210
    },
    {
      "epoch": 129.02,
      "learning_rate": 2.7976699029126215e-06,
      "loss": 0.0275,
      "step": 332220
    },
    {
      "epoch": 129.02,
      "learning_rate": 2.797152103559871e-06,
      "loss": 0.0012,
      "step": 332230
    },
    {
      "epoch": 129.03,
      "learning_rate": 2.79663430420712e-06,
      "loss": 0.0092,
      "step": 332240
    },
    {
      "epoch": 129.03,
      "learning_rate": 2.796116504854369e-06,
      "loss": 0.0102,
      "step": 332250
    },
    {
      "epoch": 129.03,
      "learning_rate": 2.7955987055016182e-06,
      "loss": 0.0632,
      "step": 332260
    },
    {
      "epoch": 129.04,
      "learning_rate": 2.7950809061488674e-06,
      "loss": 0.0442,
      "step": 332270
    },
    {
      "epoch": 129.04,
      "learning_rate": 2.7945631067961166e-06,
      "loss": 0.0622,
      "step": 332280
    },
    {
      "epoch": 129.04,
      "learning_rate": 2.794045307443366e-06,
      "loss": 0.0132,
      "step": 332290
    },
    {
      "epoch": 129.05,
      "learning_rate": 2.793527508090615e-06,
      "loss": 0.038,
      "step": 332300
    },
    {
      "epoch": 129.05,
      "learning_rate": 2.793009708737864e-06,
      "loss": 0.0166,
      "step": 332310
    },
    {
      "epoch": 129.06,
      "learning_rate": 2.7924919093851134e-06,
      "loss": 0.0301,
      "step": 332320
    },
    {
      "epoch": 129.06,
      "learning_rate": 2.7919741100323626e-06,
      "loss": 0.0696,
      "step": 332330
    },
    {
      "epoch": 129.06,
      "learning_rate": 2.791456310679612e-06,
      "loss": 0.038,
      "step": 332340
    },
    {
      "epoch": 129.07,
      "learning_rate": 2.790938511326861e-06,
      "loss": 0.1081,
      "step": 332350
    },
    {
      "epoch": 129.07,
      "learning_rate": 2.79042071197411e-06,
      "loss": 0.0107,
      "step": 332360
    },
    {
      "epoch": 129.08,
      "learning_rate": 2.7899029126213594e-06,
      "loss": 0.0089,
      "step": 332370
    },
    {
      "epoch": 129.08,
      "learning_rate": 2.7893851132686086e-06,
      "loss": 0.0473,
      "step": 332380
    },
    {
      "epoch": 129.08,
      "learning_rate": 2.7888673139158578e-06,
      "loss": 0.1061,
      "step": 332390
    },
    {
      "epoch": 129.09,
      "learning_rate": 2.788349514563107e-06,
      "loss": 0.002,
      "step": 332400
    },
    {
      "epoch": 129.09,
      "learning_rate": 2.787831715210356e-06,
      "loss": 0.0514,
      "step": 332410
    },
    {
      "epoch": 129.1,
      "learning_rate": 2.787313915857605e-06,
      "loss": 0.0999,
      "step": 332420
    },
    {
      "epoch": 129.1,
      "learning_rate": 2.7867961165048545e-06,
      "loss": 0.0129,
      "step": 332430
    },
    {
      "epoch": 129.1,
      "learning_rate": 2.786278317152104e-06,
      "loss": 0.009,
      "step": 332440
    },
    {
      "epoch": 129.11,
      "learning_rate": 2.785760517799353e-06,
      "loss": 0.0443,
      "step": 332450
    },
    {
      "epoch": 129.11,
      "learning_rate": 2.7852427184466026e-06,
      "loss": 0.1636,
      "step": 332460
    },
    {
      "epoch": 129.11,
      "learning_rate": 2.7847249190938513e-06,
      "loss": 0.2318,
      "step": 332470
    },
    {
      "epoch": 129.12,
      "learning_rate": 2.784207119741101e-06,
      "loss": 0.0817,
      "step": 332480
    },
    {
      "epoch": 129.12,
      "learning_rate": 2.7836893203883497e-06,
      "loss": 0.0103,
      "step": 332490
    },
    {
      "epoch": 129.13,
      "learning_rate": 2.7831715210355993e-06,
      "loss": 0.1351,
      "step": 332500
    },
    {
      "epoch": 129.13,
      "learning_rate": 2.782653721682848e-06,
      "loss": 0.0015,
      "step": 332510
    },
    {
      "epoch": 129.13,
      "learning_rate": 2.7821359223300977e-06,
      "loss": 0.2429,
      "step": 332520
    },
    {
      "epoch": 129.14,
      "learning_rate": 2.7816181229773465e-06,
      "loss": 0.0102,
      "step": 332530
    },
    {
      "epoch": 129.14,
      "learning_rate": 2.781100323624596e-06,
      "loss": 0.0349,
      "step": 332540
    },
    {
      "epoch": 129.15,
      "learning_rate": 2.780582524271845e-06,
      "loss": 0.0002,
      "step": 332550
    },
    {
      "epoch": 129.15,
      "learning_rate": 2.7800647249190945e-06,
      "loss": 0.0004,
      "step": 332560
    },
    {
      "epoch": 129.15,
      "learning_rate": 2.7795469255663433e-06,
      "loss": 0.1209,
      "step": 332570
    },
    {
      "epoch": 129.16,
      "learning_rate": 2.7790291262135925e-06,
      "loss": 0.0657,
      "step": 332580
    },
    {
      "epoch": 129.16,
      "learning_rate": 2.7785113268608417e-06,
      "loss": 0.0337,
      "step": 332590
    },
    {
      "epoch": 129.17,
      "learning_rate": 2.777993527508091e-06,
      "loss": 0.0449,
      "step": 332600
    },
    {
      "epoch": 129.17,
      "learning_rate": 2.77747572815534e-06,
      "loss": 0.0607,
      "step": 332610
    },
    {
      "epoch": 129.17,
      "learning_rate": 2.7769579288025892e-06,
      "loss": 0.1453,
      "step": 332620
    },
    {
      "epoch": 129.18,
      "learning_rate": 2.7764401294498384e-06,
      "loss": 0.0315,
      "step": 332630
    },
    {
      "epoch": 129.18,
      "learning_rate": 2.7759223300970876e-06,
      "loss": 0.004,
      "step": 332640
    },
    {
      "epoch": 129.18,
      "learning_rate": 2.775404530744337e-06,
      "loss": 0.0079,
      "step": 332650
    },
    {
      "epoch": 129.19,
      "learning_rate": 2.774886731391586e-06,
      "loss": 0.0343,
      "step": 332660
    },
    {
      "epoch": 129.19,
      "learning_rate": 2.7743689320388352e-06,
      "loss": 0.0004,
      "step": 332670
    },
    {
      "epoch": 129.2,
      "learning_rate": 2.7738511326860844e-06,
      "loss": 0.1294,
      "step": 332680
    },
    {
      "epoch": 129.2,
      "learning_rate": 2.7733333333333336e-06,
      "loss": 0.0005,
      "step": 332690
    },
    {
      "epoch": 129.2,
      "learning_rate": 2.772815533980583e-06,
      "loss": 0.0404,
      "step": 332700
    },
    {
      "epoch": 129.21,
      "learning_rate": 2.772297734627832e-06,
      "loss": 0.0004,
      "step": 332710
    },
    {
      "epoch": 129.21,
      "learning_rate": 2.771779935275081e-06,
      "loss": 0.0186,
      "step": 332720
    },
    {
      "epoch": 129.22,
      "learning_rate": 2.7712621359223304e-06,
      "loss": 0.0523,
      "step": 332730
    },
    {
      "epoch": 129.22,
      "learning_rate": 2.7707443365695796e-06,
      "loss": 0.0233,
      "step": 332740
    },
    {
      "epoch": 129.22,
      "learning_rate": 2.7702265372168283e-06,
      "loss": 0.0116,
      "step": 332750
    },
    {
      "epoch": 129.23,
      "learning_rate": 2.769708737864078e-06,
      "loss": 0.034,
      "step": 332760
    },
    {
      "epoch": 129.23,
      "learning_rate": 2.7691909385113267e-06,
      "loss": 0.0007,
      "step": 332770
    },
    {
      "epoch": 129.23,
      "learning_rate": 2.7686731391585764e-06,
      "loss": 0.1027,
      "step": 332780
    },
    {
      "epoch": 129.24,
      "learning_rate": 2.768155339805825e-06,
      "loss": 0.0002,
      "step": 332790
    },
    {
      "epoch": 129.24,
      "learning_rate": 2.7676375404530747e-06,
      "loss": 0.0019,
      "step": 332800
    },
    {
      "epoch": 129.25,
      "learning_rate": 2.7671197411003235e-06,
      "loss": 0.0241,
      "step": 332810
    },
    {
      "epoch": 129.25,
      "learning_rate": 2.766601941747573e-06,
      "loss": 0.041,
      "step": 332820
    },
    {
      "epoch": 129.25,
      "learning_rate": 2.766084142394822e-06,
      "loss": 0.0931,
      "step": 332830
    },
    {
      "epoch": 129.26,
      "learning_rate": 2.7655663430420715e-06,
      "loss": 0.1435,
      "step": 332840
    },
    {
      "epoch": 129.26,
      "learning_rate": 2.7650485436893203e-06,
      "loss": 0.0001,
      "step": 332850
    },
    {
      "epoch": 129.27,
      "learning_rate": 2.76453074433657e-06,
      "loss": 0.0961,
      "step": 332860
    },
    {
      "epoch": 129.27,
      "learning_rate": 2.7640129449838187e-06,
      "loss": 0.034,
      "step": 332870
    },
    {
      "epoch": 129.27,
      "learning_rate": 2.7634951456310683e-06,
      "loss": 0.0096,
      "step": 332880
    },
    {
      "epoch": 129.28,
      "learning_rate": 2.762977346278317e-06,
      "loss": 0.0041,
      "step": 332890
    },
    {
      "epoch": 129.28,
      "learning_rate": 2.7624595469255667e-06,
      "loss": 0.019,
      "step": 332900
    },
    {
      "epoch": 129.29,
      "learning_rate": 2.7619417475728155e-06,
      "loss": 0.0069,
      "step": 332910
    },
    {
      "epoch": 129.29,
      "learning_rate": 2.761423948220065e-06,
      "loss": 0.0916,
      "step": 332920
    },
    {
      "epoch": 129.29,
      "learning_rate": 2.760906148867314e-06,
      "loss": 0.0132,
      "step": 332930
    },
    {
      "epoch": 129.3,
      "learning_rate": 2.7603883495145635e-06,
      "loss": 0.0106,
      "step": 332940
    },
    {
      "epoch": 129.3,
      "learning_rate": 2.7598705501618122e-06,
      "loss": 0.1662,
      "step": 332950
    },
    {
      "epoch": 129.3,
      "learning_rate": 2.759352750809062e-06,
      "loss": 0.0371,
      "step": 332960
    },
    {
      "epoch": 129.31,
      "learning_rate": 2.7588349514563106e-06,
      "loss": 0.0428,
      "step": 332970
    },
    {
      "epoch": 129.31,
      "learning_rate": 2.7583171521035602e-06,
      "loss": 0.0004,
      "step": 332980
    },
    {
      "epoch": 129.32,
      "learning_rate": 2.757799352750809e-06,
      "loss": 0.0846,
      "step": 332990
    },
    {
      "epoch": 129.32,
      "learning_rate": 2.7572815533980586e-06,
      "loss": 0.0683,
      "step": 333000
    },
    {
      "epoch": 129.32,
      "learning_rate": 2.756763754045308e-06,
      "loss": 0.0448,
      "step": 333010
    },
    {
      "epoch": 129.33,
      "learning_rate": 2.756245954692557e-06,
      "loss": 0.04,
      "step": 333020
    },
    {
      "epoch": 129.33,
      "learning_rate": 2.7557281553398062e-06,
      "loss": 0.0005,
      "step": 333030
    },
    {
      "epoch": 129.34,
      "learning_rate": 2.7552103559870554e-06,
      "loss": 0.0121,
      "step": 333040
    },
    {
      "epoch": 129.34,
      "learning_rate": 2.7546925566343046e-06,
      "loss": 0.0205,
      "step": 333050
    },
    {
      "epoch": 129.34,
      "learning_rate": 2.754174757281554e-06,
      "loss": 0.0301,
      "step": 333060
    },
    {
      "epoch": 129.35,
      "learning_rate": 2.753656957928803e-06,
      "loss": 0.0008,
      "step": 333070
    },
    {
      "epoch": 129.35,
      "learning_rate": 2.7531391585760518e-06,
      "loss": 0.1299,
      "step": 333080
    },
    {
      "epoch": 129.36,
      "learning_rate": 2.7526213592233014e-06,
      "loss": 0.071,
      "step": 333090
    },
    {
      "epoch": 129.36,
      "learning_rate": 2.75210355987055e-06,
      "loss": 0.0001,
      "step": 333100
    },
    {
      "epoch": 129.36,
      "learning_rate": 2.7515857605177998e-06,
      "loss": 0.1163,
      "step": 333110
    },
    {
      "epoch": 129.37,
      "learning_rate": 2.7510679611650485e-06,
      "loss": 0.0015,
      "step": 333120
    },
    {
      "epoch": 129.37,
      "learning_rate": 2.750550161812298e-06,
      "loss": 0.1002,
      "step": 333130
    },
    {
      "epoch": 129.37,
      "learning_rate": 2.750032362459547e-06,
      "loss": 0.0173,
      "step": 333140
    },
    {
      "epoch": 129.38,
      "learning_rate": 2.7495145631067966e-06,
      "loss": 0.069,
      "step": 333150
    },
    {
      "epoch": 129.38,
      "learning_rate": 2.7489967637540453e-06,
      "loss": 0.0124,
      "step": 333160
    },
    {
      "epoch": 129.39,
      "learning_rate": 2.748478964401295e-06,
      "loss": 0.0001,
      "step": 333170
    },
    {
      "epoch": 129.39,
      "learning_rate": 2.7479611650485437e-06,
      "loss": 0.0094,
      "step": 333180
    },
    {
      "epoch": 129.39,
      "learning_rate": 2.7474433656957933e-06,
      "loss": 0.0519,
      "step": 333190
    },
    {
      "epoch": 129.4,
      "learning_rate": 2.746925566343042e-06,
      "loss": 0.0108,
      "step": 333200
    },
    {
      "epoch": 129.4,
      "learning_rate": 2.7464077669902917e-06,
      "loss": 0.0603,
      "step": 333210
    },
    {
      "epoch": 129.41,
      "learning_rate": 2.7458899676375405e-06,
      "loss": 0.061,
      "step": 333220
    },
    {
      "epoch": 129.41,
      "learning_rate": 2.74537216828479e-06,
      "loss": 0.0075,
      "step": 333230
    },
    {
      "epoch": 129.41,
      "learning_rate": 2.744854368932039e-06,
      "loss": 0.0168,
      "step": 333240
    },
    {
      "epoch": 129.42,
      "learning_rate": 2.7443365695792885e-06,
      "loss": 0.0398,
      "step": 333250
    },
    {
      "epoch": 129.42,
      "learning_rate": 2.7438187702265373e-06,
      "loss": 0.0652,
      "step": 333260
    },
    {
      "epoch": 129.43,
      "learning_rate": 2.743300970873787e-06,
      "loss": 0.116,
      "step": 333270
    },
    {
      "epoch": 129.43,
      "learning_rate": 2.7427831715210357e-06,
      "loss": 0.018,
      "step": 333280
    },
    {
      "epoch": 129.43,
      "learning_rate": 2.7422653721682853e-06,
      "loss": 0.0627,
      "step": 333290
    },
    {
      "epoch": 129.44,
      "learning_rate": 2.741747572815534e-06,
      "loss": 0.0004,
      "step": 333300
    },
    {
      "epoch": 129.44,
      "learning_rate": 2.7412297734627837e-06,
      "loss": 0.0291,
      "step": 333310
    },
    {
      "epoch": 129.44,
      "learning_rate": 2.7407119741100324e-06,
      "loss": 0.029,
      "step": 333320
    },
    {
      "epoch": 129.45,
      "learning_rate": 2.740194174757282e-06,
      "loss": 0.0008,
      "step": 333330
    },
    {
      "epoch": 129.45,
      "learning_rate": 2.739676375404531e-06,
      "loss": 0.1041,
      "step": 333340
    },
    {
      "epoch": 129.46,
      "learning_rate": 2.7391585760517804e-06,
      "loss": 0.0066,
      "step": 333350
    },
    {
      "epoch": 129.46,
      "learning_rate": 2.7386407766990292e-06,
      "loss": 0.018,
      "step": 333360
    },
    {
      "epoch": 129.46,
      "learning_rate": 2.738122977346279e-06,
      "loss": 0.1292,
      "step": 333370
    },
    {
      "epoch": 129.47,
      "learning_rate": 2.7376051779935276e-06,
      "loss": 0.0138,
      "step": 333380
    },
    {
      "epoch": 129.47,
      "learning_rate": 2.737087378640777e-06,
      "loss": 0.0581,
      "step": 333390
    },
    {
      "epoch": 129.48,
      "learning_rate": 2.736569579288026e-06,
      "loss": 0.163,
      "step": 333400
    },
    {
      "epoch": 129.48,
      "learning_rate": 2.736051779935275e-06,
      "loss": 0.0142,
      "step": 333410
    },
    {
      "epoch": 129.48,
      "learning_rate": 2.7355339805825244e-06,
      "loss": 0.0677,
      "step": 333420
    },
    {
      "epoch": 129.49,
      "learning_rate": 2.7350161812297736e-06,
      "loss": 0.0201,
      "step": 333430
    },
    {
      "epoch": 129.49,
      "learning_rate": 2.7344983818770228e-06,
      "loss": 0.0148,
      "step": 333440
    },
    {
      "epoch": 129.5,
      "learning_rate": 2.733980582524272e-06,
      "loss": 0.0528,
      "step": 333450
    },
    {
      "epoch": 129.5,
      "learning_rate": 2.733462783171521e-06,
      "loss": 0.0157,
      "step": 333460
    },
    {
      "epoch": 129.5,
      "learning_rate": 2.7329449838187704e-06,
      "loss": 0.1195,
      "step": 333470
    },
    {
      "epoch": 129.51,
      "learning_rate": 2.7324271844660195e-06,
      "loss": 0.0104,
      "step": 333480
    },
    {
      "epoch": 129.51,
      "learning_rate": 2.7319093851132687e-06,
      "loss": 0.1421,
      "step": 333490
    },
    {
      "epoch": 129.51,
      "learning_rate": 2.731391585760518e-06,
      "loss": 0.0206,
      "step": 333500
    },
    {
      "epoch": 129.52,
      "learning_rate": 2.730873786407767e-06,
      "loss": 0.1186,
      "step": 333510
    },
    {
      "epoch": 129.52,
      "learning_rate": 2.7303559870550163e-06,
      "loss": 0.0002,
      "step": 333520
    },
    {
      "epoch": 129.53,
      "learning_rate": 2.7298381877022655e-06,
      "loss": 0.0528,
      "step": 333530
    },
    {
      "epoch": 129.53,
      "learning_rate": 2.7293203883495147e-06,
      "loss": 0.0531,
      "step": 333540
    },
    {
      "epoch": 129.53,
      "learning_rate": 2.728802588996764e-06,
      "loss": 0.0573,
      "step": 333550
    },
    {
      "epoch": 129.54,
      "learning_rate": 2.7282847896440135e-06,
      "loss": 0.0862,
      "step": 333560
    },
    {
      "epoch": 129.54,
      "learning_rate": 2.7277669902912623e-06,
      "loss": 0.0022,
      "step": 333570
    },
    {
      "epoch": 129.55,
      "learning_rate": 2.727249190938512e-06,
      "loss": 0.0037,
      "step": 333580
    },
    {
      "epoch": 129.55,
      "learning_rate": 2.7267313915857607e-06,
      "loss": 0.1027,
      "step": 333590
    },
    {
      "epoch": 129.55,
      "learning_rate": 2.7262135922330103e-06,
      "loss": 0.0556,
      "step": 333600
    },
    {
      "epoch": 129.56,
      "learning_rate": 2.725695792880259e-06,
      "loss": 0.1154,
      "step": 333610
    },
    {
      "epoch": 129.56,
      "learning_rate": 2.7251779935275087e-06,
      "loss": 0.1771,
      "step": 333620
    },
    {
      "epoch": 129.57,
      "learning_rate": 2.7246601941747575e-06,
      "loss": 0.0007,
      "step": 333630
    },
    {
      "epoch": 129.57,
      "learning_rate": 2.724142394822007e-06,
      "loss": 0.0653,
      "step": 333640
    },
    {
      "epoch": 129.57,
      "learning_rate": 2.723624595469256e-06,
      "loss": 0.0852,
      "step": 333650
    },
    {
      "epoch": 129.58,
      "learning_rate": 2.7231067961165055e-06,
      "loss": 0.0075,
      "step": 333660
    },
    {
      "epoch": 129.58,
      "learning_rate": 2.7225889967637542e-06,
      "loss": 0.0299,
      "step": 333670
    },
    {
      "epoch": 129.58,
      "learning_rate": 2.722071197411004e-06,
      "loss": 0.0001,
      "step": 333680
    },
    {
      "epoch": 129.59,
      "learning_rate": 2.7215533980582526e-06,
      "loss": 0.0554,
      "step": 333690
    },
    {
      "epoch": 129.59,
      "learning_rate": 2.7210355987055023e-06,
      "loss": 0.0223,
      "step": 333700
    },
    {
      "epoch": 129.6,
      "learning_rate": 2.720517799352751e-06,
      "loss": 0.0319,
      "step": 333710
    },
    {
      "epoch": 129.6,
      "learning_rate": 2.7200000000000002e-06,
      "loss": 0.1023,
      "step": 333720
    },
    {
      "epoch": 129.6,
      "learning_rate": 2.7194822006472494e-06,
      "loss": 0.0243,
      "step": 333730
    },
    {
      "epoch": 129.61,
      "learning_rate": 2.7189644012944986e-06,
      "loss": 0.0823,
      "step": 333740
    },
    {
      "epoch": 129.61,
      "learning_rate": 2.718446601941748e-06,
      "loss": 0.0553,
      "step": 333750
    },
    {
      "epoch": 129.62,
      "learning_rate": 2.717928802588997e-06,
      "loss": 0.0393,
      "step": 333760
    },
    {
      "epoch": 129.62,
      "learning_rate": 2.717411003236246e-06,
      "loss": 0.0595,
      "step": 333770
    },
    {
      "epoch": 129.62,
      "learning_rate": 2.7168932038834954e-06,
      "loss": 0.0631,
      "step": 333780
    },
    {
      "epoch": 129.63,
      "learning_rate": 2.7163754045307446e-06,
      "loss": 0.0002,
      "step": 333790
    },
    {
      "epoch": 129.63,
      "learning_rate": 2.7158576051779938e-06,
      "loss": 0.0173,
      "step": 333800
    },
    {
      "epoch": 129.63,
      "learning_rate": 2.715339805825243e-06,
      "loss": 0.0473,
      "step": 333810
    },
    {
      "epoch": 129.64,
      "learning_rate": 2.714822006472492e-06,
      "loss": 0.0325,
      "step": 333820
    },
    {
      "epoch": 129.64,
      "learning_rate": 2.7143042071197414e-06,
      "loss": 0.0671,
      "step": 333830
    },
    {
      "epoch": 129.65,
      "learning_rate": 2.7137864077669906e-06,
      "loss": 0.0062,
      "step": 333840
    },
    {
      "epoch": 129.65,
      "learning_rate": 2.7132686084142397e-06,
      "loss": 0.0805,
      "step": 333850
    },
    {
      "epoch": 129.65,
      "learning_rate": 2.712750809061489e-06,
      "loss": 0.0045,
      "step": 333860
    },
    {
      "epoch": 129.66,
      "learning_rate": 2.712233009708738e-06,
      "loss": 0.0325,
      "step": 333870
    },
    {
      "epoch": 129.66,
      "learning_rate": 2.7117152103559873e-06,
      "loss": 0.1343,
      "step": 333880
    },
    {
      "epoch": 129.67,
      "learning_rate": 2.711197411003236e-06,
      "loss": 0.0154,
      "step": 333890
    },
    {
      "epoch": 129.67,
      "learning_rate": 2.7106796116504857e-06,
      "loss": 0.1575,
      "step": 333900
    },
    {
      "epoch": 129.67,
      "learning_rate": 2.7101618122977345e-06,
      "loss": 0.0334,
      "step": 333910
    },
    {
      "epoch": 129.68,
      "learning_rate": 2.709644012944984e-06,
      "loss": 0.0289,
      "step": 333920
    },
    {
      "epoch": 129.68,
      "learning_rate": 2.709126213592233e-06,
      "loss": 0.0567,
      "step": 333930
    },
    {
      "epoch": 129.69,
      "learning_rate": 2.7086084142394825e-06,
      "loss": 0.081,
      "step": 333940
    },
    {
      "epoch": 129.69,
      "learning_rate": 2.7080906148867313e-06,
      "loss": 0.0316,
      "step": 333950
    },
    {
      "epoch": 129.69,
      "learning_rate": 2.707572815533981e-06,
      "loss": 0.0011,
      "step": 333960
    },
    {
      "epoch": 129.7,
      "learning_rate": 2.7070550161812297e-06,
      "loss": 0.0287,
      "step": 333970
    },
    {
      "epoch": 129.7,
      "learning_rate": 2.7065372168284793e-06,
      "loss": 0.0826,
      "step": 333980
    },
    {
      "epoch": 129.7,
      "learning_rate": 2.706019417475728e-06,
      "loss": 0.0098,
      "step": 333990
    },
    {
      "epoch": 129.71,
      "learning_rate": 2.7055016181229777e-06,
      "loss": 0.0096,
      "step": 334000
    },
    {
      "epoch": 129.71,
      "learning_rate": 2.7049838187702264e-06,
      "loss": 0.003,
      "step": 334010
    },
    {
      "epoch": 129.72,
      "learning_rate": 2.704466019417476e-06,
      "loss": 0.0202,
      "step": 334020
    },
    {
      "epoch": 129.72,
      "learning_rate": 2.703948220064725e-06,
      "loss": 0.0369,
      "step": 334030
    },
    {
      "epoch": 129.72,
      "learning_rate": 2.7034304207119744e-06,
      "loss": 0.0038,
      "step": 334040
    },
    {
      "epoch": 129.73,
      "learning_rate": 2.702912621359223e-06,
      "loss": 0.1625,
      "step": 334050
    },
    {
      "epoch": 129.73,
      "learning_rate": 2.702394822006473e-06,
      "loss": 0.0208,
      "step": 334060
    },
    {
      "epoch": 129.74,
      "learning_rate": 2.7018770226537216e-06,
      "loss": 0.0227,
      "step": 334070
    },
    {
      "epoch": 129.74,
      "learning_rate": 2.7013592233009712e-06,
      "loss": 0.0455,
      "step": 334080
    },
    {
      "epoch": 129.74,
      "learning_rate": 2.70084142394822e-06,
      "loss": 0.0972,
      "step": 334090
    },
    {
      "epoch": 129.75,
      "learning_rate": 2.7003236245954696e-06,
      "loss": 0.0288,
      "step": 334100
    },
    {
      "epoch": 129.75,
      "learning_rate": 2.6998058252427184e-06,
      "loss": 0.1005,
      "step": 334110
    },
    {
      "epoch": 129.76,
      "learning_rate": 2.699288025889968e-06,
      "loss": 0.0628,
      "step": 334120
    },
    {
      "epoch": 129.76,
      "learning_rate": 2.698770226537217e-06,
      "loss": 0.032,
      "step": 334130
    },
    {
      "epoch": 129.76,
      "learning_rate": 2.6982524271844664e-06,
      "loss": 0.0039,
      "step": 334140
    },
    {
      "epoch": 129.77,
      "learning_rate": 2.6977346278317156e-06,
      "loss": 0.015,
      "step": 334150
    },
    {
      "epoch": 129.77,
      "learning_rate": 2.6972168284789648e-06,
      "loss": 0.0103,
      "step": 334160
    },
    {
      "epoch": 129.77,
      "learning_rate": 2.696699029126214e-06,
      "loss": 0.0134,
      "step": 334170
    },
    {
      "epoch": 129.78,
      "learning_rate": 2.696181229773463e-06,
      "loss": 0.0969,
      "step": 334180
    },
    {
      "epoch": 129.78,
      "learning_rate": 2.6956634304207124e-06,
      "loss": 0.1614,
      "step": 334190
    },
    {
      "epoch": 129.79,
      "learning_rate": 2.6951456310679616e-06,
      "loss": 0.1065,
      "step": 334200
    },
    {
      "epoch": 129.79,
      "learning_rate": 2.6946278317152107e-06,
      "loss": 0.056,
      "step": 334210
    },
    {
      "epoch": 129.79,
      "learning_rate": 2.6941100323624595e-06,
      "loss": 0.0138,
      "step": 334220
    },
    {
      "epoch": 129.8,
      "learning_rate": 2.693592233009709e-06,
      "loss": 0.0233,
      "step": 334230
    },
    {
      "epoch": 129.8,
      "learning_rate": 2.693074433656958e-06,
      "loss": 0.0389,
      "step": 334240
    },
    {
      "epoch": 129.81,
      "learning_rate": 2.6925566343042075e-06,
      "loss": 0.0002,
      "step": 334250
    },
    {
      "epoch": 129.81,
      "learning_rate": 2.6920388349514563e-06,
      "loss": 0.0314,
      "step": 334260
    },
    {
      "epoch": 129.81,
      "learning_rate": 2.691521035598706e-06,
      "loss": 0.041,
      "step": 334270
    },
    {
      "epoch": 129.82,
      "learning_rate": 2.6910032362459547e-06,
      "loss": 0.0217,
      "step": 334280
    },
    {
      "epoch": 129.82,
      "learning_rate": 2.6904854368932043e-06,
      "loss": 0.1101,
      "step": 334290
    },
    {
      "epoch": 129.83,
      "learning_rate": 2.689967637540453e-06,
      "loss": 0.0088,
      "step": 334300
    },
    {
      "epoch": 129.83,
      "learning_rate": 2.6894498381877027e-06,
      "loss": 0.1055,
      "step": 334310
    },
    {
      "epoch": 129.83,
      "learning_rate": 2.6889320388349515e-06,
      "loss": 0.0628,
      "step": 334320
    },
    {
      "epoch": 129.84,
      "learning_rate": 2.688414239482201e-06,
      "loss": 0.0092,
      "step": 334330
    },
    {
      "epoch": 129.84,
      "learning_rate": 2.68789644012945e-06,
      "loss": 0.1199,
      "step": 334340
    },
    {
      "epoch": 129.84,
      "learning_rate": 2.6873786407766995e-06,
      "loss": 0.0001,
      "step": 334350
    },
    {
      "epoch": 129.85,
      "learning_rate": 2.6868608414239482e-06,
      "loss": 0.0139,
      "step": 334360
    },
    {
      "epoch": 129.85,
      "learning_rate": 2.686343042071198e-06,
      "loss": 0.0112,
      "step": 334370
    },
    {
      "epoch": 129.86,
      "learning_rate": 2.6858252427184466e-06,
      "loss": 0.0051,
      "step": 334380
    },
    {
      "epoch": 129.86,
      "learning_rate": 2.6853074433656962e-06,
      "loss": 0.0014,
      "step": 334390
    },
    {
      "epoch": 129.86,
      "learning_rate": 2.684789644012945e-06,
      "loss": 0.0172,
      "step": 334400
    },
    {
      "epoch": 129.87,
      "learning_rate": 2.6842718446601946e-06,
      "loss": 0.0686,
      "step": 334410
    },
    {
      "epoch": 129.87,
      "learning_rate": 2.6837540453074434e-06,
      "loss": 0.1174,
      "step": 334420
    },
    {
      "epoch": 129.88,
      "learning_rate": 2.683236245954693e-06,
      "loss": 0.0264,
      "step": 334430
    },
    {
      "epoch": 129.88,
      "learning_rate": 2.682718446601942e-06,
      "loss": 0.0979,
      "step": 334440
    },
    {
      "epoch": 129.88,
      "learning_rate": 2.6822006472491914e-06,
      "loss": 0.0112,
      "step": 334450
    },
    {
      "epoch": 129.89,
      "learning_rate": 2.68168284789644e-06,
      "loss": 0.0197,
      "step": 334460
    },
    {
      "epoch": 129.89,
      "learning_rate": 2.68116504854369e-06,
      "loss": 0.0001,
      "step": 334470
    },
    {
      "epoch": 129.9,
      "learning_rate": 2.6806472491909386e-06,
      "loss": 0.0571,
      "step": 334480
    },
    {
      "epoch": 129.9,
      "learning_rate": 2.680129449838188e-06,
      "loss": 0.0413,
      "step": 334490
    },
    {
      "epoch": 129.9,
      "learning_rate": 2.679611650485437e-06,
      "loss": 0.1219,
      "step": 334500
    },
    {
      "epoch": 129.91,
      "learning_rate": 2.6790938511326866e-06,
      "loss": 0.0281,
      "step": 334510
    },
    {
      "epoch": 129.91,
      "learning_rate": 2.6785760517799354e-06,
      "loss": 0.001,
      "step": 334520
    },
    {
      "epoch": 129.91,
      "learning_rate": 2.6780582524271845e-06,
      "loss": 0.0028,
      "step": 334530
    },
    {
      "epoch": 129.92,
      "learning_rate": 2.6775404530744337e-06,
      "loss": 0.098,
      "step": 334540
    },
    {
      "epoch": 129.92,
      "learning_rate": 2.677022653721683e-06,
      "loss": 0.0141,
      "step": 334550
    },
    {
      "epoch": 129.93,
      "learning_rate": 2.676504854368932e-06,
      "loss": 0.1808,
      "step": 334560
    },
    {
      "epoch": 129.93,
      "learning_rate": 2.6759870550161813e-06,
      "loss": 0.0095,
      "step": 334570
    },
    {
      "epoch": 129.93,
      "learning_rate": 2.6754692556634305e-06,
      "loss": 0.1653,
      "step": 334580
    },
    {
      "epoch": 129.94,
      "learning_rate": 2.6749514563106797e-06,
      "loss": 0.0088,
      "step": 334590
    },
    {
      "epoch": 129.94,
      "learning_rate": 2.674433656957929e-06,
      "loss": 0.1214,
      "step": 334600
    },
    {
      "epoch": 129.95,
      "learning_rate": 2.673915857605178e-06,
      "loss": 0.0099,
      "step": 334610
    },
    {
      "epoch": 129.95,
      "learning_rate": 2.6733980582524273e-06,
      "loss": 0.0419,
      "step": 334620
    },
    {
      "epoch": 129.95,
      "learning_rate": 2.6728802588996765e-06,
      "loss": 0.0176,
      "step": 334630
    },
    {
      "epoch": 129.96,
      "learning_rate": 2.6723624595469257e-06,
      "loss": 0.0131,
      "step": 334640
    },
    {
      "epoch": 129.96,
      "learning_rate": 2.671844660194175e-06,
      "loss": 0.0174,
      "step": 334650
    },
    {
      "epoch": 129.97,
      "learning_rate": 2.671326860841424e-06,
      "loss": 0.0061,
      "step": 334660
    },
    {
      "epoch": 129.97,
      "learning_rate": 2.6708090614886733e-06,
      "loss": 0.009,
      "step": 334670
    },
    {
      "epoch": 129.97,
      "learning_rate": 2.670291262135923e-06,
      "loss": 0.0221,
      "step": 334680
    },
    {
      "epoch": 129.98,
      "learning_rate": 2.6697734627831717e-06,
      "loss": 0.105,
      "step": 334690
    },
    {
      "epoch": 129.98,
      "learning_rate": 2.6692556634304213e-06,
      "loss": 0.0984,
      "step": 334700
    },
    {
      "epoch": 129.98,
      "learning_rate": 2.66873786407767e-06,
      "loss": 0.0098,
      "step": 334710
    },
    {
      "epoch": 129.99,
      "learning_rate": 2.6682200647249197e-06,
      "loss": 0.0608,
      "step": 334720
    },
    {
      "epoch": 129.99,
      "learning_rate": 2.6677022653721684e-06,
      "loss": 0.0326,
      "step": 334730
    },
    {
      "epoch": 130.0,
      "learning_rate": 2.667184466019418e-06,
      "loss": 0.0483,
      "step": 334740
    },
    {
      "epoch": 130.0,
      "learning_rate": 2.666666666666667e-06,
      "loss": 0.0517,
      "step": 334750
    },
    {
      "epoch": 130.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.38386210799217224,
      "eval_runtime": 8.2239,
      "eval_samples_per_second": 442.005,
      "eval_steps_per_second": 55.327,
      "step": 334750
    },
    {
      "epoch": 130.0,
      "learning_rate": 2.6661488673139164e-06,
      "loss": 0.0329,
      "step": 334760
    },
    {
      "epoch": 130.01,
      "learning_rate": 2.6656310679611652e-06,
      "loss": 0.1286,
      "step": 334770
    },
    {
      "epoch": 130.01,
      "learning_rate": 2.665113268608415e-06,
      "loss": 0.0083,
      "step": 334780
    },
    {
      "epoch": 130.02,
      "learning_rate": 2.6645954692556636e-06,
      "loss": 0.0456,
      "step": 334790
    },
    {
      "epoch": 130.02,
      "learning_rate": 2.6640776699029132e-06,
      "loss": 0.0151,
      "step": 334800
    },
    {
      "epoch": 130.02,
      "learning_rate": 2.663559870550162e-06,
      "loss": 0.0001,
      "step": 334810
    },
    {
      "epoch": 130.03,
      "learning_rate": 2.6630420711974116e-06,
      "loss": 0.0518,
      "step": 334820
    },
    {
      "epoch": 130.03,
      "learning_rate": 2.6625242718446604e-06,
      "loss": 0.0088,
      "step": 334830
    },
    {
      "epoch": 130.03,
      "learning_rate": 2.66200647249191e-06,
      "loss": 0.0393,
      "step": 334840
    },
    {
      "epoch": 130.04,
      "learning_rate": 2.6614886731391588e-06,
      "loss": 0.02,
      "step": 334850
    },
    {
      "epoch": 130.04,
      "learning_rate": 2.660970873786408e-06,
      "loss": 0.0522,
      "step": 334860
    },
    {
      "epoch": 130.05,
      "learning_rate": 2.660453074433657e-06,
      "loss": 0.0083,
      "step": 334870
    },
    {
      "epoch": 130.05,
      "learning_rate": 2.6599352750809064e-06,
      "loss": 0.0482,
      "step": 334880
    },
    {
      "epoch": 130.05,
      "learning_rate": 2.6594174757281556e-06,
      "loss": 0.037,
      "step": 334890
    },
    {
      "epoch": 130.06,
      "learning_rate": 2.6588996763754047e-06,
      "loss": 0.0592,
      "step": 334900
    },
    {
      "epoch": 130.06,
      "learning_rate": 2.658381877022654e-06,
      "loss": 0.0001,
      "step": 334910
    },
    {
      "epoch": 130.07,
      "learning_rate": 2.657864077669903e-06,
      "loss": 0.0009,
      "step": 334920
    },
    {
      "epoch": 130.07,
      "learning_rate": 2.6573462783171523e-06,
      "loss": 0.1308,
      "step": 334930
    },
    {
      "epoch": 130.07,
      "learning_rate": 2.6568284789644015e-06,
      "loss": 0.0384,
      "step": 334940
    },
    {
      "epoch": 130.08,
      "learning_rate": 2.6563106796116507e-06,
      "loss": 0.0832,
      "step": 334950
    },
    {
      "epoch": 130.08,
      "learning_rate": 2.6557928802589e-06,
      "loss": 0.0793,
      "step": 334960
    },
    {
      "epoch": 130.09,
      "learning_rate": 2.655275080906149e-06,
      "loss": 0.0051,
      "step": 334970
    },
    {
      "epoch": 130.09,
      "learning_rate": 2.6547572815533983e-06,
      "loss": 0.0232,
      "step": 334980
    },
    {
      "epoch": 130.09,
      "learning_rate": 2.6542394822006475e-06,
      "loss": 0.1319,
      "step": 334990
    },
    {
      "epoch": 130.1,
      "learning_rate": 2.6537216828478967e-06,
      "loss": 0.0004,
      "step": 335000
    },
    {
      "epoch": 130.1,
      "learning_rate": 2.653203883495146e-06,
      "loss": 0.0248,
      "step": 335010
    },
    {
      "epoch": 130.1,
      "learning_rate": 2.652686084142395e-06,
      "loss": 0.0298,
      "step": 335020
    },
    {
      "epoch": 130.11,
      "learning_rate": 2.652168284789644e-06,
      "loss": 0.0462,
      "step": 335030
    },
    {
      "epoch": 130.11,
      "learning_rate": 2.6516504854368935e-06,
      "loss": 0.0063,
      "step": 335040
    },
    {
      "epoch": 130.12,
      "learning_rate": 2.6511326860841422e-06,
      "loss": 0.0452,
      "step": 335050
    },
    {
      "epoch": 130.12,
      "learning_rate": 2.650614886731392e-06,
      "loss": 0.0719,
      "step": 335060
    },
    {
      "epoch": 130.12,
      "learning_rate": 2.6500970873786406e-06,
      "loss": 0.017,
      "step": 335070
    },
    {
      "epoch": 130.13,
      "learning_rate": 2.6495792880258902e-06,
      "loss": 0.0245,
      "step": 335080
    },
    {
      "epoch": 130.13,
      "learning_rate": 2.649061488673139e-06,
      "loss": 0.0054,
      "step": 335090
    },
    {
      "epoch": 130.14,
      "learning_rate": 2.6485436893203886e-06,
      "loss": 0.0822,
      "step": 335100
    },
    {
      "epoch": 130.14,
      "learning_rate": 2.6480258899676374e-06,
      "loss": 0.0338,
      "step": 335110
    },
    {
      "epoch": 130.14,
      "learning_rate": 2.647508090614887e-06,
      "loss": 0.0332,
      "step": 335120
    },
    {
      "epoch": 130.15,
      "learning_rate": 2.646990291262136e-06,
      "loss": 0.1302,
      "step": 335130
    },
    {
      "epoch": 130.15,
      "learning_rate": 2.6464724919093854e-06,
      "loss": 0.0479,
      "step": 335140
    },
    {
      "epoch": 130.16,
      "learning_rate": 2.645954692556634e-06,
      "loss": 0.0499,
      "step": 335150
    },
    {
      "epoch": 130.16,
      "learning_rate": 2.645436893203884e-06,
      "loss": 0.031,
      "step": 335160
    },
    {
      "epoch": 130.16,
      "learning_rate": 2.6449190938511326e-06,
      "loss": 0.1055,
      "step": 335170
    },
    {
      "epoch": 130.17,
      "learning_rate": 2.644401294498382e-06,
      "loss": 0.0001,
      "step": 335180
    },
    {
      "epoch": 130.17,
      "learning_rate": 2.643883495145631e-06,
      "loss": 0.0348,
      "step": 335190
    },
    {
      "epoch": 130.17,
      "learning_rate": 2.6433656957928806e-06,
      "loss": 0.1096,
      "step": 335200
    },
    {
      "epoch": 130.18,
      "learning_rate": 2.6428478964401294e-06,
      "loss": 0.1239,
      "step": 335210
    },
    {
      "epoch": 130.18,
      "learning_rate": 2.642330097087379e-06,
      "loss": 0.0914,
      "step": 335220
    },
    {
      "epoch": 130.19,
      "learning_rate": 2.6418122977346277e-06,
      "loss": 0.0106,
      "step": 335230
    },
    {
      "epoch": 130.19,
      "learning_rate": 2.6412944983818774e-06,
      "loss": 0.0231,
      "step": 335240
    },
    {
      "epoch": 130.19,
      "learning_rate": 2.6407766990291266e-06,
      "loss": 0.0115,
      "step": 335250
    },
    {
      "epoch": 130.2,
      "learning_rate": 2.6402588996763757e-06,
      "loss": 0.0266,
      "step": 335260
    },
    {
      "epoch": 130.2,
      "learning_rate": 2.639741100323625e-06,
      "loss": 0.0041,
      "step": 335270
    },
    {
      "epoch": 130.21,
      "learning_rate": 2.639223300970874e-06,
      "loss": 0.1669,
      "step": 335280
    },
    {
      "epoch": 130.21,
      "learning_rate": 2.6387055016181233e-06,
      "loss": 0.0883,
      "step": 335290
    },
    {
      "epoch": 130.21,
      "learning_rate": 2.6381877022653725e-06,
      "loss": 0.0153,
      "step": 335300
    },
    {
      "epoch": 130.22,
      "learning_rate": 2.6376699029126217e-06,
      "loss": 0.0221,
      "step": 335310
    },
    {
      "epoch": 130.22,
      "learning_rate": 2.637152103559871e-06,
      "loss": 0.0596,
      "step": 335320
    },
    {
      "epoch": 130.23,
      "learning_rate": 2.63663430420712e-06,
      "loss": 0.0172,
      "step": 335330
    },
    {
      "epoch": 130.23,
      "learning_rate": 2.6361165048543693e-06,
      "loss": 0.0661,
      "step": 335340
    },
    {
      "epoch": 130.23,
      "learning_rate": 2.6355987055016185e-06,
      "loss": 0.0269,
      "step": 335350
    },
    {
      "epoch": 130.24,
      "learning_rate": 2.6350809061488673e-06,
      "loss": 0.011,
      "step": 335360
    },
    {
      "epoch": 130.24,
      "learning_rate": 2.634563106796117e-06,
      "loss": 0.0201,
      "step": 335370
    },
    {
      "epoch": 130.24,
      "learning_rate": 2.6340453074433657e-06,
      "loss": 0.1614,
      "step": 335380
    },
    {
      "epoch": 130.25,
      "learning_rate": 2.6335275080906153e-06,
      "loss": 0.017,
      "step": 335390
    },
    {
      "epoch": 130.25,
      "learning_rate": 2.633009708737864e-06,
      "loss": 0.1881,
      "step": 335400
    },
    {
      "epoch": 130.26,
      "learning_rate": 2.6324919093851137e-06,
      "loss": 0.1032,
      "step": 335410
    },
    {
      "epoch": 130.26,
      "learning_rate": 2.6319741100323624e-06,
      "loss": 0.0364,
      "step": 335420
    },
    {
      "epoch": 130.26,
      "learning_rate": 2.631456310679612e-06,
      "loss": 0.0456,
      "step": 335430
    },
    {
      "epoch": 130.27,
      "learning_rate": 2.630938511326861e-06,
      "loss": 0.0101,
      "step": 335440
    },
    {
      "epoch": 130.27,
      "learning_rate": 2.6304207119741104e-06,
      "loss": 0.0303,
      "step": 335450
    },
    {
      "epoch": 130.28,
      "learning_rate": 2.6299029126213592e-06,
      "loss": 0.0081,
      "step": 335460
    },
    {
      "epoch": 130.28,
      "learning_rate": 2.629385113268609e-06,
      "loss": 0.0448,
      "step": 335470
    },
    {
      "epoch": 130.28,
      "learning_rate": 2.6288673139158576e-06,
      "loss": 0.0465,
      "step": 335480
    },
    {
      "epoch": 130.29,
      "learning_rate": 2.6283495145631072e-06,
      "loss": 0.0195,
      "step": 335490
    },
    {
      "epoch": 130.29,
      "learning_rate": 2.627831715210356e-06,
      "loss": 0.0961,
      "step": 335500
    },
    {
      "epoch": 130.3,
      "learning_rate": 2.6273139158576056e-06,
      "loss": 0.0196,
      "step": 335510
    },
    {
      "epoch": 130.3,
      "learning_rate": 2.6267961165048544e-06,
      "loss": 0.0357,
      "step": 335520
    },
    {
      "epoch": 130.3,
      "learning_rate": 2.626278317152104e-06,
      "loss": 0.0241,
      "step": 335530
    },
    {
      "epoch": 130.31,
      "learning_rate": 2.6257605177993528e-06,
      "loss": 0.0077,
      "step": 335540
    },
    {
      "epoch": 130.31,
      "learning_rate": 2.6252427184466024e-06,
      "loss": 0.0495,
      "step": 335550
    },
    {
      "epoch": 130.31,
      "learning_rate": 2.624724919093851e-06,
      "loss": 0.0528,
      "step": 335560
    },
    {
      "epoch": 130.32,
      "learning_rate": 2.6242071197411008e-06,
      "loss": 0.0125,
      "step": 335570
    },
    {
      "epoch": 130.32,
      "learning_rate": 2.6236893203883496e-06,
      "loss": 0.1282,
      "step": 335580
    },
    {
      "epoch": 130.33,
      "learning_rate": 2.623171521035599e-06,
      "loss": 0.0001,
      "step": 335590
    },
    {
      "epoch": 130.33,
      "learning_rate": 2.622653721682848e-06,
      "loss": 0.0202,
      "step": 335600
    },
    {
      "epoch": 130.33,
      "learning_rate": 2.6221359223300976e-06,
      "loss": 0.0659,
      "step": 335610
    },
    {
      "epoch": 130.34,
      "learning_rate": 2.6216181229773463e-06,
      "loss": 0.0096,
      "step": 335620
    },
    {
      "epoch": 130.34,
      "learning_rate": 2.621100323624596e-06,
      "loss": 0.1821,
      "step": 335630
    },
    {
      "epoch": 130.35,
      "learning_rate": 2.6205825242718447e-06,
      "loss": 0.0003,
      "step": 335640
    },
    {
      "epoch": 130.35,
      "learning_rate": 2.6200647249190943e-06,
      "loss": 0.0181,
      "step": 335650
    },
    {
      "epoch": 130.35,
      "learning_rate": 2.619546925566343e-06,
      "loss": 0.1166,
      "step": 335660
    },
    {
      "epoch": 130.36,
      "learning_rate": 2.6190291262135923e-06,
      "loss": 0.0517,
      "step": 335670
    },
    {
      "epoch": 130.36,
      "learning_rate": 2.6185113268608415e-06,
      "loss": 0.1522,
      "step": 335680
    },
    {
      "epoch": 130.37,
      "learning_rate": 2.6179935275080907e-06,
      "loss": 0.004,
      "step": 335690
    },
    {
      "epoch": 130.37,
      "learning_rate": 2.61747572815534e-06,
      "loss": 0.0265,
      "step": 335700
    },
    {
      "epoch": 130.37,
      "learning_rate": 2.616957928802589e-06,
      "loss": 0.0275,
      "step": 335710
    },
    {
      "epoch": 130.38,
      "learning_rate": 2.6164401294498383e-06,
      "loss": 0.008,
      "step": 335720
    },
    {
      "epoch": 130.38,
      "learning_rate": 2.6159223300970875e-06,
      "loss": 0.0907,
      "step": 335730
    },
    {
      "epoch": 130.38,
      "learning_rate": 2.6154045307443367e-06,
      "loss": 0.0196,
      "step": 335740
    },
    {
      "epoch": 130.39,
      "learning_rate": 2.614886731391586e-06,
      "loss": 0.0011,
      "step": 335750
    },
    {
      "epoch": 130.39,
      "learning_rate": 2.614368932038835e-06,
      "loss": 0.0002,
      "step": 335760
    },
    {
      "epoch": 130.4,
      "learning_rate": 2.6138511326860842e-06,
      "loss": 0.0476,
      "step": 335770
    },
    {
      "epoch": 130.4,
      "learning_rate": 2.6133333333333334e-06,
      "loss": 0.0316,
      "step": 335780
    },
    {
      "epoch": 130.4,
      "learning_rate": 2.6128155339805826e-06,
      "loss": 0.037,
      "step": 335790
    },
    {
      "epoch": 130.41,
      "learning_rate": 2.6122977346278323e-06,
      "loss": 0.1021,
      "step": 335800
    },
    {
      "epoch": 130.41,
      "learning_rate": 2.611779935275081e-06,
      "loss": 0.047,
      "step": 335810
    },
    {
      "epoch": 130.42,
      "learning_rate": 2.6112621359223306e-06,
      "loss": 0.0099,
      "step": 335820
    },
    {
      "epoch": 130.42,
      "learning_rate": 2.6107443365695794e-06,
      "loss": 0.0136,
      "step": 335830
    },
    {
      "epoch": 130.42,
      "learning_rate": 2.610226537216829e-06,
      "loss": 0.1459,
      "step": 335840
    },
    {
      "epoch": 130.43,
      "learning_rate": 2.609708737864078e-06,
      "loss": 0.0108,
      "step": 335850
    },
    {
      "epoch": 130.43,
      "learning_rate": 2.6091909385113274e-06,
      "loss": 0.1189,
      "step": 335860
    },
    {
      "epoch": 130.43,
      "learning_rate": 2.608673139158576e-06,
      "loss": 0.0387,
      "step": 335870
    },
    {
      "epoch": 130.44,
      "learning_rate": 2.608155339805826e-06,
      "loss": 0.0427,
      "step": 335880
    },
    {
      "epoch": 130.44,
      "learning_rate": 2.6076375404530746e-06,
      "loss": 0.0099,
      "step": 335890
    },
    {
      "epoch": 130.45,
      "learning_rate": 2.607119741100324e-06,
      "loss": 0.1186,
      "step": 335900
    },
    {
      "epoch": 130.45,
      "learning_rate": 2.606601941747573e-06,
      "loss": 0.021,
      "step": 335910
    },
    {
      "epoch": 130.45,
      "learning_rate": 2.6060841423948226e-06,
      "loss": 0.0169,
      "step": 335920
    },
    {
      "epoch": 130.46,
      "learning_rate": 2.6055663430420714e-06,
      "loss": 0.046,
      "step": 335930
    },
    {
      "epoch": 130.46,
      "learning_rate": 2.605048543689321e-06,
      "loss": 0.0002,
      "step": 335940
    },
    {
      "epoch": 130.47,
      "learning_rate": 2.6045307443365697e-06,
      "loss": 0.0006,
      "step": 335950
    },
    {
      "epoch": 130.47,
      "learning_rate": 2.6040129449838194e-06,
      "loss": 0.0001,
      "step": 335960
    },
    {
      "epoch": 130.47,
      "learning_rate": 2.603495145631068e-06,
      "loss": 0.1051,
      "step": 335970
    },
    {
      "epoch": 130.48,
      "learning_rate": 2.6029773462783178e-06,
      "loss": 0.0006,
      "step": 335980
    },
    {
      "epoch": 130.48,
      "learning_rate": 2.6024595469255665e-06,
      "loss": 0.0508,
      "step": 335990
    },
    {
      "epoch": 130.49,
      "learning_rate": 2.6019417475728157e-06,
      "loss": 0.0169,
      "step": 336000
    },
    {
      "epoch": 130.49,
      "learning_rate": 2.601423948220065e-06,
      "loss": 0.1123,
      "step": 336010
    },
    {
      "epoch": 130.49,
      "learning_rate": 2.600906148867314e-06,
      "loss": 0.0237,
      "step": 336020
    },
    {
      "epoch": 130.5,
      "learning_rate": 2.6003883495145633e-06,
      "loss": 0.023,
      "step": 336030
    },
    {
      "epoch": 130.5,
      "learning_rate": 2.5998705501618125e-06,
      "loss": 0.01,
      "step": 336040
    },
    {
      "epoch": 130.5,
      "learning_rate": 2.5993527508090617e-06,
      "loss": 0.0452,
      "step": 336050
    },
    {
      "epoch": 130.51,
      "learning_rate": 2.598834951456311e-06,
      "loss": 0.0194,
      "step": 336060
    },
    {
      "epoch": 130.51,
      "learning_rate": 2.59831715210356e-06,
      "loss": 0.0088,
      "step": 336070
    },
    {
      "epoch": 130.52,
      "learning_rate": 2.5977993527508093e-06,
      "loss": 0.0533,
      "step": 336080
    },
    {
      "epoch": 130.52,
      "learning_rate": 2.5972815533980585e-06,
      "loss": 0.1349,
      "step": 336090
    },
    {
      "epoch": 130.52,
      "learning_rate": 2.5967637540453077e-06,
      "loss": 0.01,
      "step": 336100
    },
    {
      "epoch": 130.53,
      "learning_rate": 2.596245954692557e-06,
      "loss": 0.0755,
      "step": 336110
    },
    {
      "epoch": 130.53,
      "learning_rate": 2.595728155339806e-06,
      "loss": 0.0721,
      "step": 336120
    },
    {
      "epoch": 130.54,
      "learning_rate": 2.5952103559870552e-06,
      "loss": 0.0072,
      "step": 336130
    },
    {
      "epoch": 130.54,
      "learning_rate": 2.5946925566343044e-06,
      "loss": 0.1273,
      "step": 336140
    },
    {
      "epoch": 130.54,
      "learning_rate": 2.5941747572815536e-06,
      "loss": 0.0476,
      "step": 336150
    },
    {
      "epoch": 130.55,
      "learning_rate": 2.593656957928803e-06,
      "loss": 0.1116,
      "step": 336160
    },
    {
      "epoch": 130.55,
      "learning_rate": 2.5931391585760516e-06,
      "loss": 0.0224,
      "step": 336170
    },
    {
      "epoch": 130.56,
      "learning_rate": 2.5926213592233012e-06,
      "loss": 0.0361,
      "step": 336180
    },
    {
      "epoch": 130.56,
      "learning_rate": 2.59210355987055e-06,
      "loss": 0.0452,
      "step": 336190
    },
    {
      "epoch": 130.56,
      "learning_rate": 2.5915857605177996e-06,
      "loss": 0.0905,
      "step": 336200
    },
    {
      "epoch": 130.57,
      "learning_rate": 2.5910679611650484e-06,
      "loss": 0.0156,
      "step": 336210
    },
    {
      "epoch": 130.57,
      "learning_rate": 2.590550161812298e-06,
      "loss": 0.0319,
      "step": 336220
    },
    {
      "epoch": 130.57,
      "learning_rate": 2.5900323624595468e-06,
      "loss": 0.0549,
      "step": 336230
    },
    {
      "epoch": 130.58,
      "learning_rate": 2.5895145631067964e-06,
      "loss": 0.0316,
      "step": 336240
    },
    {
      "epoch": 130.58,
      "learning_rate": 2.588996763754045e-06,
      "loss": 0.0278,
      "step": 336250
    },
    {
      "epoch": 130.59,
      "learning_rate": 2.5884789644012948e-06,
      "loss": 0.0011,
      "step": 336260
    },
    {
      "epoch": 130.59,
      "learning_rate": 2.5879611650485435e-06,
      "loss": 0.0403,
      "step": 336270
    },
    {
      "epoch": 130.59,
      "learning_rate": 2.587443365695793e-06,
      "loss": 0.0031,
      "step": 336280
    },
    {
      "epoch": 130.6,
      "learning_rate": 2.586925566343042e-06,
      "loss": 0.0011,
      "step": 336290
    },
    {
      "epoch": 130.6,
      "learning_rate": 2.5864077669902916e-06,
      "loss": 0.0002,
      "step": 336300
    },
    {
      "epoch": 130.61,
      "learning_rate": 2.5858899676375403e-06,
      "loss": 0.0155,
      "step": 336310
    },
    {
      "epoch": 130.61,
      "learning_rate": 2.58537216828479e-06,
      "loss": 0.0797,
      "step": 336320
    },
    {
      "epoch": 130.61,
      "learning_rate": 2.5848543689320387e-06,
      "loss": 0.1185,
      "step": 336330
    },
    {
      "epoch": 130.62,
      "learning_rate": 2.5843365695792883e-06,
      "loss": 0.0304,
      "step": 336340
    },
    {
      "epoch": 130.62,
      "learning_rate": 2.583818770226537e-06,
      "loss": 0.1541,
      "step": 336350
    },
    {
      "epoch": 130.63,
      "learning_rate": 2.5833009708737867e-06,
      "loss": 0.0066,
      "step": 336360
    },
    {
      "epoch": 130.63,
      "learning_rate": 2.582783171521036e-06,
      "loss": 0.1098,
      "step": 336370
    },
    {
      "epoch": 130.63,
      "learning_rate": 2.582265372168285e-06,
      "loss": 0.1411,
      "step": 336380
    },
    {
      "epoch": 130.64,
      "learning_rate": 2.5817475728155343e-06,
      "loss": 0.0758,
      "step": 336390
    },
    {
      "epoch": 130.64,
      "learning_rate": 2.5812297734627835e-06,
      "loss": 0.0997,
      "step": 336400
    },
    {
      "epoch": 130.64,
      "learning_rate": 2.5807119741100327e-06,
      "loss": 0.1517,
      "step": 336410
    },
    {
      "epoch": 130.65,
      "learning_rate": 2.580194174757282e-06,
      "loss": 0.0641,
      "step": 336420
    },
    {
      "epoch": 130.65,
      "learning_rate": 2.579676375404531e-06,
      "loss": 0.1003,
      "step": 336430
    },
    {
      "epoch": 130.66,
      "learning_rate": 2.5791585760517803e-06,
      "loss": 0.1459,
      "step": 336440
    },
    {
      "epoch": 130.66,
      "learning_rate": 2.5786407766990295e-06,
      "loss": 0.0064,
      "step": 336450
    },
    {
      "epoch": 130.66,
      "learning_rate": 2.5781229773462787e-06,
      "loss": 0.0606,
      "step": 336460
    },
    {
      "epoch": 130.67,
      "learning_rate": 2.577605177993528e-06,
      "loss": 0.0093,
      "step": 336470
    },
    {
      "epoch": 130.67,
      "learning_rate": 2.577087378640777e-06,
      "loss": 0.0008,
      "step": 336480
    },
    {
      "epoch": 130.68,
      "learning_rate": 2.5765695792880263e-06,
      "loss": 0.0101,
      "step": 336490
    },
    {
      "epoch": 130.68,
      "learning_rate": 2.576051779935275e-06,
      "loss": 0.0081,
      "step": 336500
    },
    {
      "epoch": 130.68,
      "learning_rate": 2.5755339805825246e-06,
      "loss": 0.0163,
      "step": 336510
    },
    {
      "epoch": 130.69,
      "learning_rate": 2.5750161812297734e-06,
      "loss": 0.0132,
      "step": 336520
    },
    {
      "epoch": 130.69,
      "learning_rate": 2.574498381877023e-06,
      "loss": 0.0431,
      "step": 336530
    },
    {
      "epoch": 130.7,
      "learning_rate": 2.573980582524272e-06,
      "loss": 0.1605,
      "step": 336540
    },
    {
      "epoch": 130.7,
      "learning_rate": 2.5734627831715214e-06,
      "loss": 0.026,
      "step": 336550
    },
    {
      "epoch": 130.7,
      "learning_rate": 2.57294498381877e-06,
      "loss": 0.0069,
      "step": 336560
    },
    {
      "epoch": 130.71,
      "learning_rate": 2.57242718446602e-06,
      "loss": 0.0633,
      "step": 336570
    },
    {
      "epoch": 130.71,
      "learning_rate": 2.5719093851132686e-06,
      "loss": 0.0236,
      "step": 336580
    },
    {
      "epoch": 130.71,
      "learning_rate": 2.571391585760518e-06,
      "loss": 0.049,
      "step": 336590
    },
    {
      "epoch": 130.72,
      "learning_rate": 2.570873786407767e-06,
      "loss": 0.0647,
      "step": 336600
    },
    {
      "epoch": 130.72,
      "learning_rate": 2.5703559870550166e-06,
      "loss": 0.0461,
      "step": 336610
    },
    {
      "epoch": 130.73,
      "learning_rate": 2.5698381877022654e-06,
      "loss": 0.0303,
      "step": 336620
    },
    {
      "epoch": 130.73,
      "learning_rate": 2.569320388349515e-06,
      "loss": 0.076,
      "step": 336630
    },
    {
      "epoch": 130.73,
      "learning_rate": 2.5688025889967637e-06,
      "loss": 0.0599,
      "step": 336640
    },
    {
      "epoch": 130.74,
      "learning_rate": 2.5682847896440134e-06,
      "loss": 0.0077,
      "step": 336650
    },
    {
      "epoch": 130.74,
      "learning_rate": 2.567766990291262e-06,
      "loss": 0.0587,
      "step": 336660
    },
    {
      "epoch": 130.75,
      "learning_rate": 2.5672491909385118e-06,
      "loss": 0.0164,
      "step": 336670
    },
    {
      "epoch": 130.75,
      "learning_rate": 2.5667313915857605e-06,
      "loss": 0.0248,
      "step": 336680
    },
    {
      "epoch": 130.75,
      "learning_rate": 2.56621359223301e-06,
      "loss": 0.0779,
      "step": 336690
    },
    {
      "epoch": 130.76,
      "learning_rate": 2.565695792880259e-06,
      "loss": 0.1172,
      "step": 336700
    },
    {
      "epoch": 130.76,
      "learning_rate": 2.5651779935275085e-06,
      "loss": 0.0803,
      "step": 336710
    },
    {
      "epoch": 130.77,
      "learning_rate": 2.5646601941747573e-06,
      "loss": 0.0208,
      "step": 336720
    },
    {
      "epoch": 130.77,
      "learning_rate": 2.564142394822007e-06,
      "loss": 0.0193,
      "step": 336730
    },
    {
      "epoch": 130.77,
      "learning_rate": 2.5636245954692557e-06,
      "loss": 0.0507,
      "step": 336740
    },
    {
      "epoch": 130.78,
      "learning_rate": 2.5631067961165053e-06,
      "loss": 0.0102,
      "step": 336750
    },
    {
      "epoch": 130.78,
      "learning_rate": 2.562588996763754e-06,
      "loss": 0.011,
      "step": 336760
    },
    {
      "epoch": 130.78,
      "learning_rate": 2.5620711974110037e-06,
      "loss": 0.0111,
      "step": 336770
    },
    {
      "epoch": 130.79,
      "learning_rate": 2.5615533980582525e-06,
      "loss": 0.0263,
      "step": 336780
    },
    {
      "epoch": 130.79,
      "learning_rate": 2.561035598705502e-06,
      "loss": 0.0431,
      "step": 336790
    },
    {
      "epoch": 130.8,
      "learning_rate": 2.560517799352751e-06,
      "loss": 0.0028,
      "step": 336800
    },
    {
      "epoch": 130.8,
      "learning_rate": 2.56e-06,
      "loss": 0.0124,
      "step": 336810
    },
    {
      "epoch": 130.8,
      "learning_rate": 2.5594822006472492e-06,
      "loss": 0.0773,
      "step": 336820
    },
    {
      "epoch": 130.81,
      "learning_rate": 2.5589644012944984e-06,
      "loss": 0.0845,
      "step": 336830
    },
    {
      "epoch": 130.81,
      "learning_rate": 2.5584466019417476e-06,
      "loss": 0.0005,
      "step": 336840
    },
    {
      "epoch": 130.82,
      "learning_rate": 2.557928802588997e-06,
      "loss": 0.0103,
      "step": 336850
    },
    {
      "epoch": 130.82,
      "learning_rate": 2.557411003236246e-06,
      "loss": 0.0988,
      "step": 336860
    },
    {
      "epoch": 130.82,
      "learning_rate": 2.5568932038834952e-06,
      "loss": 0.0189,
      "step": 336870
    },
    {
      "epoch": 130.83,
      "learning_rate": 2.5563754045307444e-06,
      "loss": 0.0087,
      "step": 336880
    },
    {
      "epoch": 130.83,
      "learning_rate": 2.5558576051779936e-06,
      "loss": 0.0107,
      "step": 336890
    },
    {
      "epoch": 130.83,
      "learning_rate": 2.555339805825243e-06,
      "loss": 0.062,
      "step": 336900
    },
    {
      "epoch": 130.84,
      "learning_rate": 2.554822006472492e-06,
      "loss": 0.026,
      "step": 336910
    },
    {
      "epoch": 130.84,
      "learning_rate": 2.5543042071197416e-06,
      "loss": 0.0094,
      "step": 336920
    },
    {
      "epoch": 130.85,
      "learning_rate": 2.5537864077669904e-06,
      "loss": 0.0105,
      "step": 336930
    },
    {
      "epoch": 130.85,
      "learning_rate": 2.55326860841424e-06,
      "loss": 0.0153,
      "step": 336940
    },
    {
      "epoch": 130.85,
      "learning_rate": 2.5527508090614888e-06,
      "loss": 0.0205,
      "step": 336950
    },
    {
      "epoch": 130.86,
      "learning_rate": 2.5522330097087384e-06,
      "loss": 0.0191,
      "step": 336960
    },
    {
      "epoch": 130.86,
      "learning_rate": 2.551715210355987e-06,
      "loss": 0.0277,
      "step": 336970
    },
    {
      "epoch": 130.87,
      "learning_rate": 2.5511974110032368e-06,
      "loss": 0.0552,
      "step": 336980
    },
    {
      "epoch": 130.87,
      "learning_rate": 2.5506796116504856e-06,
      "loss": 0.0017,
      "step": 336990
    },
    {
      "epoch": 130.87,
      "learning_rate": 2.550161812297735e-06,
      "loss": 0.0004,
      "step": 337000
    },
    {
      "epoch": 130.88,
      "learning_rate": 2.549644012944984e-06,
      "loss": 0.0607,
      "step": 337010
    },
    {
      "epoch": 130.88,
      "learning_rate": 2.5491262135922336e-06,
      "loss": 0.1047,
      "step": 337020
    },
    {
      "epoch": 130.89,
      "learning_rate": 2.5486084142394823e-06,
      "loss": 0.0005,
      "step": 337030
    },
    {
      "epoch": 130.89,
      "learning_rate": 2.548090614886732e-06,
      "loss": 0.0005,
      "step": 337040
    },
    {
      "epoch": 130.89,
      "learning_rate": 2.5475728155339807e-06,
      "loss": 0.054,
      "step": 337050
    },
    {
      "epoch": 130.9,
      "learning_rate": 2.5470550161812303e-06,
      "loss": 0.1247,
      "step": 337060
    },
    {
      "epoch": 130.9,
      "learning_rate": 2.546537216828479e-06,
      "loss": 0.0187,
      "step": 337070
    },
    {
      "epoch": 130.9,
      "learning_rate": 2.5460194174757287e-06,
      "loss": 0.134,
      "step": 337080
    },
    {
      "epoch": 130.91,
      "learning_rate": 2.5455016181229775e-06,
      "loss": 0.1253,
      "step": 337090
    },
    {
      "epoch": 130.91,
      "learning_rate": 2.544983818770227e-06,
      "loss": 0.112,
      "step": 337100
    },
    {
      "epoch": 130.92,
      "learning_rate": 2.544466019417476e-06,
      "loss": 0.0753,
      "step": 337110
    },
    {
      "epoch": 130.92,
      "learning_rate": 2.5439482200647255e-06,
      "loss": 0.1241,
      "step": 337120
    },
    {
      "epoch": 130.92,
      "learning_rate": 2.5434304207119743e-06,
      "loss": 0.0898,
      "step": 337130
    },
    {
      "epoch": 130.93,
      "learning_rate": 2.5429126213592235e-06,
      "loss": 0.0383,
      "step": 337140
    },
    {
      "epoch": 130.93,
      "learning_rate": 2.5423948220064727e-06,
      "loss": 0.0459,
      "step": 337150
    },
    {
      "epoch": 130.94,
      "learning_rate": 2.541877022653722e-06,
      "loss": 0.0003,
      "step": 337160
    },
    {
      "epoch": 130.94,
      "learning_rate": 2.541359223300971e-06,
      "loss": 0.0135,
      "step": 337170
    },
    {
      "epoch": 130.94,
      "learning_rate": 2.5408414239482203e-06,
      "loss": 0.1643,
      "step": 337180
    },
    {
      "epoch": 130.95,
      "learning_rate": 2.5403236245954694e-06,
      "loss": 0.0108,
      "step": 337190
    },
    {
      "epoch": 130.95,
      "learning_rate": 2.5398058252427186e-06,
      "loss": 0.0298,
      "step": 337200
    },
    {
      "epoch": 130.96,
      "learning_rate": 2.539288025889968e-06,
      "loss": 0.0001,
      "step": 337210
    },
    {
      "epoch": 130.96,
      "learning_rate": 2.538770226537217e-06,
      "loss": 0.0001,
      "step": 337220
    },
    {
      "epoch": 130.96,
      "learning_rate": 2.5382524271844662e-06,
      "loss": 0.0044,
      "step": 337230
    },
    {
      "epoch": 130.97,
      "learning_rate": 2.5377346278317154e-06,
      "loss": 0.1353,
      "step": 337240
    },
    {
      "epoch": 130.97,
      "learning_rate": 2.5372168284789646e-06,
      "loss": 0.012,
      "step": 337250
    },
    {
      "epoch": 130.97,
      "learning_rate": 2.536699029126214e-06,
      "loss": 0.0034,
      "step": 337260
    },
    {
      "epoch": 130.98,
      "learning_rate": 2.536181229773463e-06,
      "loss": 0.0289,
      "step": 337270
    },
    {
      "epoch": 130.98,
      "learning_rate": 2.535663430420712e-06,
      "loss": 0.2499,
      "step": 337280
    },
    {
      "epoch": 130.99,
      "learning_rate": 2.5351456310679614e-06,
      "loss": 0.185,
      "step": 337290
    },
    {
      "epoch": 130.99,
      "learning_rate": 2.5346278317152106e-06,
      "loss": 0.0087,
      "step": 337300
    },
    {
      "epoch": 130.99,
      "learning_rate": 2.5341100323624594e-06,
      "loss": 0.0511,
      "step": 337310
    },
    {
      "epoch": 131.0,
      "learning_rate": 2.533592233009709e-06,
      "loss": 0.0218,
      "step": 337320
    },
    {
      "epoch": 131.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.3862694203853607,
      "eval_runtime": 8.3054,
      "eval_samples_per_second": 437.664,
      "eval_steps_per_second": 54.783,
      "step": 337325
    },
    {
      "epoch": 131.0,
      "learning_rate": 2.5330744336569577e-06,
      "loss": 0.0046,
      "step": 337330
    },
    {
      "epoch": 131.01,
      "learning_rate": 2.5325566343042074e-06,
      "loss": 0.0612,
      "step": 337340
    },
    {
      "epoch": 131.01,
      "learning_rate": 2.532038834951456e-06,
      "loss": 0.1255,
      "step": 337350
    },
    {
      "epoch": 131.01,
      "learning_rate": 2.5315210355987058e-06,
      "loss": 0.0123,
      "step": 337360
    },
    {
      "epoch": 131.02,
      "learning_rate": 2.5310032362459545e-06,
      "loss": 0.0003,
      "step": 337370
    },
    {
      "epoch": 131.02,
      "learning_rate": 2.530485436893204e-06,
      "loss": 0.0012,
      "step": 337380
    },
    {
      "epoch": 131.03,
      "learning_rate": 2.529967637540453e-06,
      "loss": 0.1012,
      "step": 337390
    },
    {
      "epoch": 131.03,
      "learning_rate": 2.5294498381877025e-06,
      "loss": 0.0503,
      "step": 337400
    },
    {
      "epoch": 131.03,
      "learning_rate": 2.5289320388349513e-06,
      "loss": 0.0004,
      "step": 337410
    },
    {
      "epoch": 131.04,
      "learning_rate": 2.528414239482201e-06,
      "loss": 0.0011,
      "step": 337420
    },
    {
      "epoch": 131.04,
      "learning_rate": 2.5278964401294497e-06,
      "loss": 0.0012,
      "step": 337430
    },
    {
      "epoch": 131.04,
      "learning_rate": 2.5273786407766993e-06,
      "loss": 0.0661,
      "step": 337440
    },
    {
      "epoch": 131.05,
      "learning_rate": 2.526860841423948e-06,
      "loss": 0.0119,
      "step": 337450
    },
    {
      "epoch": 131.05,
      "learning_rate": 2.5263430420711977e-06,
      "loss": 0.0538,
      "step": 337460
    },
    {
      "epoch": 131.06,
      "learning_rate": 2.5258252427184465e-06,
      "loss": 0.076,
      "step": 337470
    },
    {
      "epoch": 131.06,
      "learning_rate": 2.525307443365696e-06,
      "loss": 0.0179,
      "step": 337480
    },
    {
      "epoch": 131.06,
      "learning_rate": 2.5247896440129453e-06,
      "loss": 0.0962,
      "step": 337490
    },
    {
      "epoch": 131.07,
      "learning_rate": 2.5242718446601945e-06,
      "loss": 0.0005,
      "step": 337500
    },
    {
      "epoch": 131.07,
      "learning_rate": 2.5237540453074437e-06,
      "loss": 0.0011,
      "step": 337510
    },
    {
      "epoch": 131.08,
      "learning_rate": 2.523236245954693e-06,
      "loss": 0.0182,
      "step": 337520
    },
    {
      "epoch": 131.08,
      "learning_rate": 2.522718446601942e-06,
      "loss": 0.0681,
      "step": 337530
    },
    {
      "epoch": 131.08,
      "learning_rate": 2.5222006472491913e-06,
      "loss": 0.1218,
      "step": 337540
    },
    {
      "epoch": 131.09,
      "learning_rate": 2.5216828478964404e-06,
      "loss": 0.1103,
      "step": 337550
    },
    {
      "epoch": 131.09,
      "learning_rate": 2.5211650485436896e-06,
      "loss": 0.0318,
      "step": 337560
    },
    {
      "epoch": 131.1,
      "learning_rate": 2.520647249190939e-06,
      "loss": 0.0518,
      "step": 337570
    },
    {
      "epoch": 131.1,
      "learning_rate": 2.520129449838188e-06,
      "loss": 0.1343,
      "step": 337580
    },
    {
      "epoch": 131.1,
      "learning_rate": 2.5196116504854372e-06,
      "loss": 0.001,
      "step": 337590
    },
    {
      "epoch": 131.11,
      "learning_rate": 2.5190938511326864e-06,
      "loss": 0.1164,
      "step": 337600
    },
    {
      "epoch": 131.11,
      "learning_rate": 2.5185760517799356e-06,
      "loss": 0.0904,
      "step": 337610
    },
    {
      "epoch": 131.11,
      "learning_rate": 2.518058252427185e-06,
      "loss": 0.0625,
      "step": 337620
    },
    {
      "epoch": 131.12,
      "learning_rate": 2.517540453074434e-06,
      "loss": 0.0657,
      "step": 337630
    },
    {
      "epoch": 131.12,
      "learning_rate": 2.5170226537216828e-06,
      "loss": 0.04,
      "step": 337640
    },
    {
      "epoch": 131.13,
      "learning_rate": 2.5165048543689324e-06,
      "loss": 0.0614,
      "step": 337650
    },
    {
      "epoch": 131.13,
      "learning_rate": 2.515987055016181e-06,
      "loss": 0.1342,
      "step": 337660
    },
    {
      "epoch": 131.13,
      "learning_rate": 2.5154692556634308e-06,
      "loss": 0.0141,
      "step": 337670
    },
    {
      "epoch": 131.14,
      "learning_rate": 2.5149514563106796e-06,
      "loss": 0.0442,
      "step": 337680
    },
    {
      "epoch": 131.14,
      "learning_rate": 2.514433656957929e-06,
      "loss": 0.0178,
      "step": 337690
    },
    {
      "epoch": 131.15,
      "learning_rate": 2.513915857605178e-06,
      "loss": 0.0712,
      "step": 337700
    },
    {
      "epoch": 131.15,
      "learning_rate": 2.5133980582524276e-06,
      "loss": 0.0919,
      "step": 337710
    },
    {
      "epoch": 131.15,
      "learning_rate": 2.5128802588996763e-06,
      "loss": 0.1116,
      "step": 337720
    },
    {
      "epoch": 131.16,
      "learning_rate": 2.512362459546926e-06,
      "loss": 0.0783,
      "step": 337730
    },
    {
      "epoch": 131.16,
      "learning_rate": 2.5118446601941747e-06,
      "loss": 0.1029,
      "step": 337740
    },
    {
      "epoch": 131.17,
      "learning_rate": 2.5113268608414243e-06,
      "loss": 0.0432,
      "step": 337750
    },
    {
      "epoch": 131.17,
      "learning_rate": 2.510809061488673e-06,
      "loss": 0.0284,
      "step": 337760
    },
    {
      "epoch": 131.17,
      "learning_rate": 2.5102912621359227e-06,
      "loss": 0.0424,
      "step": 337770
    },
    {
      "epoch": 131.18,
      "learning_rate": 2.5097734627831715e-06,
      "loss": 0.0526,
      "step": 337780
    },
    {
      "epoch": 131.18,
      "learning_rate": 2.509255663430421e-06,
      "loss": 0.1551,
      "step": 337790
    },
    {
      "epoch": 131.18,
      "learning_rate": 2.50873786407767e-06,
      "loss": 0.0099,
      "step": 337800
    },
    {
      "epoch": 131.19,
      "learning_rate": 2.5082200647249195e-06,
      "loss": 0.0898,
      "step": 337810
    },
    {
      "epoch": 131.19,
      "learning_rate": 2.5077022653721683e-06,
      "loss": 0.032,
      "step": 337820
    },
    {
      "epoch": 131.2,
      "learning_rate": 2.507184466019418e-06,
      "loss": 0.0643,
      "step": 337830
    },
    {
      "epoch": 131.2,
      "learning_rate": 2.5066666666666667e-06,
      "loss": 0.1159,
      "step": 337840
    },
    {
      "epoch": 131.2,
      "learning_rate": 2.5061488673139163e-06,
      "loss": 0.0085,
      "step": 337850
    },
    {
      "epoch": 131.21,
      "learning_rate": 2.505631067961165e-06,
      "loss": 0.035,
      "step": 337860
    },
    {
      "epoch": 131.21,
      "learning_rate": 2.5051132686084147e-06,
      "loss": 0.0465,
      "step": 337870
    },
    {
      "epoch": 131.22,
      "learning_rate": 2.5045954692556634e-06,
      "loss": 0.0175,
      "step": 337880
    },
    {
      "epoch": 131.22,
      "learning_rate": 2.504077669902913e-06,
      "loss": 0.0787,
      "step": 337890
    },
    {
      "epoch": 131.22,
      "learning_rate": 2.503559870550162e-06,
      "loss": 0.066,
      "step": 337900
    },
    {
      "epoch": 131.23,
      "learning_rate": 2.5030420711974115e-06,
      "loss": 0.0295,
      "step": 337910
    },
    {
      "epoch": 131.23,
      "learning_rate": 2.5025242718446602e-06,
      "loss": 0.0384,
      "step": 337920
    },
    {
      "epoch": 131.23,
      "learning_rate": 2.50200647249191e-06,
      "loss": 0.0881,
      "step": 337930
    },
    {
      "epoch": 131.24,
      "learning_rate": 2.5014886731391586e-06,
      "loss": 0.056,
      "step": 337940
    },
    {
      "epoch": 131.24,
      "learning_rate": 2.500970873786408e-06,
      "loss": 0.0096,
      "step": 337950
    },
    {
      "epoch": 131.25,
      "learning_rate": 2.500453074433657e-06,
      "loss": 0.0001,
      "step": 337960
    },
    {
      "epoch": 131.25,
      "learning_rate": 2.499935275080906e-06,
      "loss": 0.051,
      "step": 337970
    },
    {
      "epoch": 131.25,
      "learning_rate": 2.4994174757281554e-06,
      "loss": 0.0013,
      "step": 337980
    },
    {
      "epoch": 131.26,
      "learning_rate": 2.4988996763754046e-06,
      "loss": 0.0086,
      "step": 337990
    },
    {
      "epoch": 131.26,
      "learning_rate": 2.4983818770226538e-06,
      "loss": 0.074,
      "step": 338000
    },
    {
      "epoch": 131.27,
      "learning_rate": 2.497864077669903e-06,
      "loss": 0.0261,
      "step": 338010
    },
    {
      "epoch": 131.27,
      "learning_rate": 2.497346278317152e-06,
      "loss": 0.0095,
      "step": 338020
    },
    {
      "epoch": 131.27,
      "learning_rate": 2.4968284789644014e-06,
      "loss": 0.1082,
      "step": 338030
    },
    {
      "epoch": 131.28,
      "learning_rate": 2.4963106796116506e-06,
      "loss": 0.0484,
      "step": 338040
    },
    {
      "epoch": 131.28,
      "learning_rate": 2.4957928802588998e-06,
      "loss": 0.0217,
      "step": 338050
    },
    {
      "epoch": 131.29,
      "learning_rate": 2.495275080906149e-06,
      "loss": 0.0608,
      "step": 338060
    },
    {
      "epoch": 131.29,
      "learning_rate": 2.494757281553398e-06,
      "loss": 0.0019,
      "step": 338070
    },
    {
      "epoch": 131.29,
      "learning_rate": 2.4942394822006473e-06,
      "loss": 0.0374,
      "step": 338080
    },
    {
      "epoch": 131.3,
      "learning_rate": 2.4937216828478965e-06,
      "loss": 0.0004,
      "step": 338090
    },
    {
      "epoch": 131.3,
      "learning_rate": 2.4932038834951457e-06,
      "loss": 0.1095,
      "step": 338100
    },
    {
      "epoch": 131.3,
      "learning_rate": 2.492686084142395e-06,
      "loss": 0.1092,
      "step": 338110
    },
    {
      "epoch": 131.31,
      "learning_rate": 2.492168284789644e-06,
      "loss": 0.0338,
      "step": 338120
    },
    {
      "epoch": 131.31,
      "learning_rate": 2.4916504854368933e-06,
      "loss": 0.0321,
      "step": 338130
    },
    {
      "epoch": 131.32,
      "learning_rate": 2.4911326860841425e-06,
      "loss": 0.0078,
      "step": 338140
    },
    {
      "epoch": 131.32,
      "learning_rate": 2.4906148867313917e-06,
      "loss": 0.0264,
      "step": 338150
    },
    {
      "epoch": 131.32,
      "learning_rate": 2.490097087378641e-06,
      "loss": 0.0185,
      "step": 338160
    },
    {
      "epoch": 131.33,
      "learning_rate": 2.48957928802589e-06,
      "loss": 0.0232,
      "step": 338170
    },
    {
      "epoch": 131.33,
      "learning_rate": 2.4890614886731393e-06,
      "loss": 0.0264,
      "step": 338180
    },
    {
      "epoch": 131.34,
      "learning_rate": 2.4885436893203885e-06,
      "loss": 0.0819,
      "step": 338190
    },
    {
      "epoch": 131.34,
      "learning_rate": 2.4880258899676377e-06,
      "loss": 0.0044,
      "step": 338200
    },
    {
      "epoch": 131.34,
      "learning_rate": 2.487508090614887e-06,
      "loss": 0.0213,
      "step": 338210
    },
    {
      "epoch": 131.35,
      "learning_rate": 2.486990291262136e-06,
      "loss": 0.0272,
      "step": 338220
    },
    {
      "epoch": 131.35,
      "learning_rate": 2.4864724919093853e-06,
      "loss": 0.0277,
      "step": 338230
    },
    {
      "epoch": 131.36,
      "learning_rate": 2.4859546925566344e-06,
      "loss": 0.1436,
      "step": 338240
    },
    {
      "epoch": 131.36,
      "learning_rate": 2.4854368932038836e-06,
      "loss": 0.0886,
      "step": 338250
    },
    {
      "epoch": 131.36,
      "learning_rate": 2.484919093851133e-06,
      "loss": 0.0002,
      "step": 338260
    },
    {
      "epoch": 131.37,
      "learning_rate": 2.484401294498382e-06,
      "loss": 0.0365,
      "step": 338270
    },
    {
      "epoch": 131.37,
      "learning_rate": 2.4838834951456312e-06,
      "loss": 0.0274,
      "step": 338280
    },
    {
      "epoch": 131.37,
      "learning_rate": 2.4833656957928804e-06,
      "loss": 0.0605,
      "step": 338290
    },
    {
      "epoch": 131.38,
      "learning_rate": 2.4828478964401296e-06,
      "loss": 0.0373,
      "step": 338300
    },
    {
      "epoch": 131.38,
      "learning_rate": 2.482330097087379e-06,
      "loss": 0.0146,
      "step": 338310
    },
    {
      "epoch": 131.39,
      "learning_rate": 2.481812297734628e-06,
      "loss": 0.1114,
      "step": 338320
    },
    {
      "epoch": 131.39,
      "learning_rate": 2.481294498381877e-06,
      "loss": 0.0023,
      "step": 338330
    },
    {
      "epoch": 131.39,
      "learning_rate": 2.4807766990291264e-06,
      "loss": 0.0344,
      "step": 338340
    },
    {
      "epoch": 131.4,
      "learning_rate": 2.4802588996763756e-06,
      "loss": 0.205,
      "step": 338350
    },
    {
      "epoch": 131.4,
      "learning_rate": 2.4797411003236248e-06,
      "loss": 0.0074,
      "step": 338360
    },
    {
      "epoch": 131.41,
      "learning_rate": 2.479223300970874e-06,
      "loss": 0.1325,
      "step": 338370
    },
    {
      "epoch": 131.41,
      "learning_rate": 2.478705501618123e-06,
      "loss": 0.0462,
      "step": 338380
    },
    {
      "epoch": 131.41,
      "learning_rate": 2.4781877022653724e-06,
      "loss": 0.0326,
      "step": 338390
    },
    {
      "epoch": 131.42,
      "learning_rate": 2.4776699029126216e-06,
      "loss": 0.0073,
      "step": 338400
    },
    {
      "epoch": 131.42,
      "learning_rate": 2.4771521035598708e-06,
      "loss": 0.0438,
      "step": 338410
    },
    {
      "epoch": 131.43,
      "learning_rate": 2.47663430420712e-06,
      "loss": 0.0021,
      "step": 338420
    },
    {
      "epoch": 131.43,
      "learning_rate": 2.476116504854369e-06,
      "loss": 0.0385,
      "step": 338430
    },
    {
      "epoch": 131.43,
      "learning_rate": 2.4755987055016183e-06,
      "loss": 0.0542,
      "step": 338440
    },
    {
      "epoch": 131.44,
      "learning_rate": 2.4750809061488675e-06,
      "loss": 0.0418,
      "step": 338450
    },
    {
      "epoch": 131.44,
      "learning_rate": 2.4745631067961167e-06,
      "loss": 0.0178,
      "step": 338460
    },
    {
      "epoch": 131.44,
      "learning_rate": 2.474045307443366e-06,
      "loss": 0.0181,
      "step": 338470
    },
    {
      "epoch": 131.45,
      "learning_rate": 2.473527508090615e-06,
      "loss": 0.0154,
      "step": 338480
    },
    {
      "epoch": 131.45,
      "learning_rate": 2.4730097087378643e-06,
      "loss": 0.1979,
      "step": 338490
    },
    {
      "epoch": 131.46,
      "learning_rate": 2.4724919093851135e-06,
      "loss": 0.0346,
      "step": 338500
    },
    {
      "epoch": 131.46,
      "learning_rate": 2.4719741100323627e-06,
      "loss": 0.0665,
      "step": 338510
    },
    {
      "epoch": 131.46,
      "learning_rate": 2.471456310679612e-06,
      "loss": 0.0315,
      "step": 338520
    },
    {
      "epoch": 131.47,
      "learning_rate": 2.470938511326861e-06,
      "loss": 0.0001,
      "step": 338530
    },
    {
      "epoch": 131.47,
      "learning_rate": 2.4704207119741103e-06,
      "loss": 0.0288,
      "step": 338540
    },
    {
      "epoch": 131.48,
      "learning_rate": 2.4699029126213595e-06,
      "loss": 0.0165,
      "step": 338550
    },
    {
      "epoch": 131.48,
      "learning_rate": 2.4693851132686087e-06,
      "loss": 0.0007,
      "step": 338560
    },
    {
      "epoch": 131.48,
      "learning_rate": 2.468867313915858e-06,
      "loss": 0.0091,
      "step": 338570
    },
    {
      "epoch": 131.49,
      "learning_rate": 2.468349514563107e-06,
      "loss": 0.0033,
      "step": 338580
    },
    {
      "epoch": 131.49,
      "learning_rate": 2.4678317152103563e-06,
      "loss": 0.0323,
      "step": 338590
    },
    {
      "epoch": 131.5,
      "learning_rate": 2.4673139158576055e-06,
      "loss": 0.0313,
      "step": 338600
    },
    {
      "epoch": 131.5,
      "learning_rate": 2.4667961165048546e-06,
      "loss": 0.0002,
      "step": 338610
    },
    {
      "epoch": 131.5,
      "learning_rate": 2.466278317152104e-06,
      "loss": 0.0308,
      "step": 338620
    },
    {
      "epoch": 131.51,
      "learning_rate": 2.465760517799353e-06,
      "loss": 0.006,
      "step": 338630
    },
    {
      "epoch": 131.51,
      "learning_rate": 2.4652427184466022e-06,
      "loss": 0.0182,
      "step": 338640
    },
    {
      "epoch": 131.51,
      "learning_rate": 2.4647249190938514e-06,
      "loss": 0.0635,
      "step": 338650
    },
    {
      "epoch": 131.52,
      "learning_rate": 2.4642071197411006e-06,
      "loss": 0.0201,
      "step": 338660
    },
    {
      "epoch": 131.52,
      "learning_rate": 2.46368932038835e-06,
      "loss": 0.0003,
      "step": 338670
    },
    {
      "epoch": 131.53,
      "learning_rate": 2.463171521035599e-06,
      "loss": 0.0694,
      "step": 338680
    },
    {
      "epoch": 131.53,
      "learning_rate": 2.462653721682848e-06,
      "loss": 0.0001,
      "step": 338690
    },
    {
      "epoch": 131.53,
      "learning_rate": 2.4621359223300974e-06,
      "loss": 0.0412,
      "step": 338700
    },
    {
      "epoch": 131.54,
      "learning_rate": 2.4616181229773466e-06,
      "loss": 0.0394,
      "step": 338710
    },
    {
      "epoch": 131.54,
      "learning_rate": 2.4611003236245958e-06,
      "loss": 0.0094,
      "step": 338720
    },
    {
      "epoch": 131.55,
      "learning_rate": 2.460582524271845e-06,
      "loss": 0.0159,
      "step": 338730
    },
    {
      "epoch": 131.55,
      "learning_rate": 2.460064724919094e-06,
      "loss": 0.0157,
      "step": 338740
    },
    {
      "epoch": 131.55,
      "learning_rate": 2.4595469255663434e-06,
      "loss": 0.1193,
      "step": 338750
    },
    {
      "epoch": 131.56,
      "learning_rate": 2.4590291262135926e-06,
      "loss": 0.0485,
      "step": 338760
    },
    {
      "epoch": 131.56,
      "learning_rate": 2.4585113268608413e-06,
      "loss": 0.0586,
      "step": 338770
    },
    {
      "epoch": 131.57,
      "learning_rate": 2.4579935275080905e-06,
      "loss": 0.0478,
      "step": 338780
    },
    {
      "epoch": 131.57,
      "learning_rate": 2.4574757281553397e-06,
      "loss": 0.0126,
      "step": 338790
    },
    {
      "epoch": 131.57,
      "learning_rate": 2.456957928802589e-06,
      "loss": 0.0162,
      "step": 338800
    },
    {
      "epoch": 131.58,
      "learning_rate": 2.456440129449838e-06,
      "loss": 0.0608,
      "step": 338810
    },
    {
      "epoch": 131.58,
      "learning_rate": 2.4559223300970873e-06,
      "loss": 0.0311,
      "step": 338820
    },
    {
      "epoch": 131.58,
      "learning_rate": 2.4554045307443365e-06,
      "loss": 0.0516,
      "step": 338830
    },
    {
      "epoch": 131.59,
      "learning_rate": 2.4548867313915857e-06,
      "loss": 0.0166,
      "step": 338840
    },
    {
      "epoch": 131.59,
      "learning_rate": 2.454368932038835e-06,
      "loss": 0.0011,
      "step": 338850
    },
    {
      "epoch": 131.6,
      "learning_rate": 2.453851132686084e-06,
      "loss": 0.0387,
      "step": 338860
    },
    {
      "epoch": 131.6,
      "learning_rate": 2.4533333333333333e-06,
      "loss": 0.0486,
      "step": 338870
    },
    {
      "epoch": 131.6,
      "learning_rate": 2.452815533980583e-06,
      "loss": 0.0004,
      "step": 338880
    },
    {
      "epoch": 131.61,
      "learning_rate": 2.452297734627832e-06,
      "loss": 0.0206,
      "step": 338890
    },
    {
      "epoch": 131.61,
      "learning_rate": 2.4517799352750813e-06,
      "loss": 0.0364,
      "step": 338900
    },
    {
      "epoch": 131.62,
      "learning_rate": 2.4512621359223305e-06,
      "loss": 0.0339,
      "step": 338910
    },
    {
      "epoch": 131.62,
      "learning_rate": 2.4507443365695797e-06,
      "loss": 0.0471,
      "step": 338920
    },
    {
      "epoch": 131.62,
      "learning_rate": 2.450226537216829e-06,
      "loss": 0.0044,
      "step": 338930
    },
    {
      "epoch": 131.63,
      "learning_rate": 2.449708737864078e-06,
      "loss": 0.0104,
      "step": 338940
    },
    {
      "epoch": 131.63,
      "learning_rate": 2.4491909385113273e-06,
      "loss": 0.0856,
      "step": 338950
    },
    {
      "epoch": 131.63,
      "learning_rate": 2.4486731391585765e-06,
      "loss": 0.0365,
      "step": 338960
    },
    {
      "epoch": 131.64,
      "learning_rate": 2.4481553398058256e-06,
      "loss": 0.0368,
      "step": 338970
    },
    {
      "epoch": 131.64,
      "learning_rate": 2.447637540453075e-06,
      "loss": 0.0608,
      "step": 338980
    },
    {
      "epoch": 131.65,
      "learning_rate": 2.447119741100324e-06,
      "loss": 0.1869,
      "step": 338990
    },
    {
      "epoch": 131.65,
      "learning_rate": 2.4466019417475732e-06,
      "loss": 0.0622,
      "step": 339000
    },
    {
      "epoch": 131.65,
      "learning_rate": 2.4460841423948224e-06,
      "loss": 0.0011,
      "step": 339010
    },
    {
      "epoch": 131.66,
      "learning_rate": 2.4455663430420716e-06,
      "loss": 0.0457,
      "step": 339020
    },
    {
      "epoch": 131.66,
      "learning_rate": 2.445048543689321e-06,
      "loss": 0.0087,
      "step": 339030
    },
    {
      "epoch": 131.67,
      "learning_rate": 2.44453074433657e-06,
      "loss": 0.0145,
      "step": 339040
    },
    {
      "epoch": 131.67,
      "learning_rate": 2.444012944983819e-06,
      "loss": 0.0894,
      "step": 339050
    },
    {
      "epoch": 131.67,
      "learning_rate": 2.4434951456310684e-06,
      "loss": 0.0287,
      "step": 339060
    },
    {
      "epoch": 131.68,
      "learning_rate": 2.4429773462783176e-06,
      "loss": 0.0652,
      "step": 339070
    },
    {
      "epoch": 131.68,
      "learning_rate": 2.4424595469255668e-06,
      "loss": 0.0013,
      "step": 339080
    },
    {
      "epoch": 131.69,
      "learning_rate": 2.4419417475728156e-06,
      "loss": 0.0696,
      "step": 339090
    },
    {
      "epoch": 131.69,
      "learning_rate": 2.4414239482200648e-06,
      "loss": 0.021,
      "step": 339100
    },
    {
      "epoch": 131.69,
      "learning_rate": 2.440906148867314e-06,
      "loss": 0.0468,
      "step": 339110
    },
    {
      "epoch": 131.7,
      "learning_rate": 2.440388349514563e-06,
      "loss": 0.0402,
      "step": 339120
    },
    {
      "epoch": 131.7,
      "learning_rate": 2.4398705501618123e-06,
      "loss": 0.0467,
      "step": 339130
    },
    {
      "epoch": 131.7,
      "learning_rate": 2.4393527508090615e-06,
      "loss": 0.0973,
      "step": 339140
    },
    {
      "epoch": 131.71,
      "learning_rate": 2.4388349514563107e-06,
      "loss": 0.0347,
      "step": 339150
    },
    {
      "epoch": 131.71,
      "learning_rate": 2.43831715210356e-06,
      "loss": 0.0008,
      "step": 339160
    },
    {
      "epoch": 131.72,
      "learning_rate": 2.437799352750809e-06,
      "loss": 0.049,
      "step": 339170
    },
    {
      "epoch": 131.72,
      "learning_rate": 2.4372815533980583e-06,
      "loss": 0.0006,
      "step": 339180
    },
    {
      "epoch": 131.72,
      "learning_rate": 2.4367637540453075e-06,
      "loss": 0.031,
      "step": 339190
    },
    {
      "epoch": 131.73,
      "learning_rate": 2.4362459546925567e-06,
      "loss": 0.0001,
      "step": 339200
    },
    {
      "epoch": 131.73,
      "learning_rate": 2.435728155339806e-06,
      "loss": 0.0014,
      "step": 339210
    },
    {
      "epoch": 131.74,
      "learning_rate": 2.435210355987055e-06,
      "loss": 0.0671,
      "step": 339220
    },
    {
      "epoch": 131.74,
      "learning_rate": 2.4346925566343043e-06,
      "loss": 0.1417,
      "step": 339230
    },
    {
      "epoch": 131.74,
      "learning_rate": 2.4341747572815535e-06,
      "loss": 0.101,
      "step": 339240
    },
    {
      "epoch": 131.75,
      "learning_rate": 2.4336569579288027e-06,
      "loss": 0.0674,
      "step": 339250
    },
    {
      "epoch": 131.75,
      "learning_rate": 2.433139158576052e-06,
      "loss": 0.0024,
      "step": 339260
    },
    {
      "epoch": 131.76,
      "learning_rate": 2.432621359223301e-06,
      "loss": 0.0078,
      "step": 339270
    },
    {
      "epoch": 131.76,
      "learning_rate": 2.4321035598705503e-06,
      "loss": 0.0168,
      "step": 339280
    },
    {
      "epoch": 131.76,
      "learning_rate": 2.4315857605177994e-06,
      "loss": 0.0288,
      "step": 339290
    },
    {
      "epoch": 131.77,
      "learning_rate": 2.4310679611650486e-06,
      "loss": 0.0237,
      "step": 339300
    },
    {
      "epoch": 131.77,
      "learning_rate": 2.430550161812298e-06,
      "loss": 0.0595,
      "step": 339310
    },
    {
      "epoch": 131.77,
      "learning_rate": 2.430032362459547e-06,
      "loss": 0.123,
      "step": 339320
    },
    {
      "epoch": 131.78,
      "learning_rate": 2.4295145631067962e-06,
      "loss": 0.0015,
      "step": 339330
    },
    {
      "epoch": 131.78,
      "learning_rate": 2.4289967637540454e-06,
      "loss": 0.0215,
      "step": 339340
    },
    {
      "epoch": 131.79,
      "learning_rate": 2.4284789644012946e-06,
      "loss": 0.0011,
      "step": 339350
    },
    {
      "epoch": 131.79,
      "learning_rate": 2.427961165048544e-06,
      "loss": 0.0112,
      "step": 339360
    },
    {
      "epoch": 131.79,
      "learning_rate": 2.427443365695793e-06,
      "loss": 0.0476,
      "step": 339370
    },
    {
      "epoch": 131.8,
      "learning_rate": 2.426925566343042e-06,
      "loss": 0.0897,
      "step": 339380
    },
    {
      "epoch": 131.8,
      "learning_rate": 2.4264077669902914e-06,
      "loss": 0.0332,
      "step": 339390
    },
    {
      "epoch": 131.81,
      "learning_rate": 2.4258899676375406e-06,
      "loss": 0.0281,
      "step": 339400
    },
    {
      "epoch": 131.81,
      "learning_rate": 2.4253721682847898e-06,
      "loss": 0.0005,
      "step": 339410
    },
    {
      "epoch": 131.81,
      "learning_rate": 2.424854368932039e-06,
      "loss": 0.0897,
      "step": 339420
    },
    {
      "epoch": 131.82,
      "learning_rate": 2.424336569579288e-06,
      "loss": 0.0112,
      "step": 339430
    },
    {
      "epoch": 131.82,
      "learning_rate": 2.4238187702265374e-06,
      "loss": 0.0254,
      "step": 339440
    },
    {
      "epoch": 131.83,
      "learning_rate": 2.4233009708737866e-06,
      "loss": 0.0152,
      "step": 339450
    },
    {
      "epoch": 131.83,
      "learning_rate": 2.4227831715210358e-06,
      "loss": 0.0922,
      "step": 339460
    },
    {
      "epoch": 131.83,
      "learning_rate": 2.422265372168285e-06,
      "loss": 0.0236,
      "step": 339470
    },
    {
      "epoch": 131.84,
      "learning_rate": 2.421747572815534e-06,
      "loss": 0.0004,
      "step": 339480
    },
    {
      "epoch": 131.84,
      "learning_rate": 2.4212297734627833e-06,
      "loss": 0.014,
      "step": 339490
    },
    {
      "epoch": 131.84,
      "learning_rate": 2.4207119741100325e-06,
      "loss": 0.1082,
      "step": 339500
    },
    {
      "epoch": 131.85,
      "learning_rate": 2.4201941747572817e-06,
      "loss": 0.009,
      "step": 339510
    },
    {
      "epoch": 131.85,
      "learning_rate": 2.419676375404531e-06,
      "loss": 0.0671,
      "step": 339520
    },
    {
      "epoch": 131.86,
      "learning_rate": 2.41915857605178e-06,
      "loss": 0.0164,
      "step": 339530
    },
    {
      "epoch": 131.86,
      "learning_rate": 2.4186407766990293e-06,
      "loss": 0.0094,
      "step": 339540
    },
    {
      "epoch": 131.86,
      "learning_rate": 2.4181229773462785e-06,
      "loss": 0.0415,
      "step": 339550
    },
    {
      "epoch": 131.87,
      "learning_rate": 2.4176051779935277e-06,
      "loss": 0.0001,
      "step": 339560
    },
    {
      "epoch": 131.87,
      "learning_rate": 2.417087378640777e-06,
      "loss": 0.0007,
      "step": 339570
    },
    {
      "epoch": 131.88,
      "learning_rate": 2.416569579288026e-06,
      "loss": 0.071,
      "step": 339580
    },
    {
      "epoch": 131.88,
      "learning_rate": 2.4160517799352753e-06,
      "loss": 0.0168,
      "step": 339590
    },
    {
      "epoch": 131.88,
      "learning_rate": 2.4155339805825245e-06,
      "loss": 0.0984,
      "step": 339600
    },
    {
      "epoch": 131.89,
      "learning_rate": 2.4150161812297737e-06,
      "loss": 0.0105,
      "step": 339610
    },
    {
      "epoch": 131.89,
      "learning_rate": 2.414498381877023e-06,
      "loss": 0.0082,
      "step": 339620
    },
    {
      "epoch": 131.9,
      "learning_rate": 2.413980582524272e-06,
      "loss": 0.1208,
      "step": 339630
    },
    {
      "epoch": 131.9,
      "learning_rate": 2.4134627831715213e-06,
      "loss": 0.0219,
      "step": 339640
    },
    {
      "epoch": 131.9,
      "learning_rate": 2.4129449838187705e-06,
      "loss": 0.0764,
      "step": 339650
    },
    {
      "epoch": 131.91,
      "learning_rate": 2.4124271844660196e-06,
      "loss": 0.0204,
      "step": 339660
    },
    {
      "epoch": 131.91,
      "learning_rate": 2.411909385113269e-06,
      "loss": 0.1032,
      "step": 339670
    },
    {
      "epoch": 131.91,
      "learning_rate": 2.411391585760518e-06,
      "loss": 0.07,
      "step": 339680
    },
    {
      "epoch": 131.92,
      "learning_rate": 2.4108737864077672e-06,
      "loss": 0.0628,
      "step": 339690
    },
    {
      "epoch": 131.92,
      "learning_rate": 2.4103559870550164e-06,
      "loss": 0.0413,
      "step": 339700
    },
    {
      "epoch": 131.93,
      "learning_rate": 2.4098381877022656e-06,
      "loss": 0.0434,
      "step": 339710
    },
    {
      "epoch": 131.93,
      "learning_rate": 2.409320388349515e-06,
      "loss": 0.1248,
      "step": 339720
    },
    {
      "epoch": 131.93,
      "learning_rate": 2.408802588996764e-06,
      "loss": 0.0211,
      "step": 339730
    },
    {
      "epoch": 131.94,
      "learning_rate": 2.408284789644013e-06,
      "loss": 0.0893,
      "step": 339740
    },
    {
      "epoch": 131.94,
      "learning_rate": 2.4077669902912624e-06,
      "loss": 0.0002,
      "step": 339750
    },
    {
      "epoch": 131.95,
      "learning_rate": 2.4072491909385116e-06,
      "loss": 0.0067,
      "step": 339760
    },
    {
      "epoch": 131.95,
      "learning_rate": 2.4067313915857608e-06,
      "loss": 0.0008,
      "step": 339770
    },
    {
      "epoch": 131.95,
      "learning_rate": 2.40621359223301e-06,
      "loss": 0.011,
      "step": 339780
    },
    {
      "epoch": 131.96,
      "learning_rate": 2.405695792880259e-06,
      "loss": 0.0248,
      "step": 339790
    },
    {
      "epoch": 131.96,
      "learning_rate": 2.4051779935275084e-06,
      "loss": 0.0573,
      "step": 339800
    },
    {
      "epoch": 131.97,
      "learning_rate": 2.4046601941747576e-06,
      "loss": 0.0126,
      "step": 339810
    },
    {
      "epoch": 131.97,
      "learning_rate": 2.4041423948220068e-06,
      "loss": 0.0007,
      "step": 339820
    },
    {
      "epoch": 131.97,
      "learning_rate": 2.403624595469256e-06,
      "loss": 0.1148,
      "step": 339830
    },
    {
      "epoch": 131.98,
      "learning_rate": 2.403106796116505e-06,
      "loss": 0.0195,
      "step": 339840
    },
    {
      "epoch": 131.98,
      "learning_rate": 2.4025889967637543e-06,
      "loss": 0.1303,
      "step": 339850
    },
    {
      "epoch": 131.98,
      "learning_rate": 2.4020711974110035e-06,
      "loss": 0.0011,
      "step": 339860
    },
    {
      "epoch": 131.99,
      "learning_rate": 2.4015533980582527e-06,
      "loss": 0.0149,
      "step": 339870
    },
    {
      "epoch": 131.99,
      "learning_rate": 2.401035598705502e-06,
      "loss": 0.0605,
      "step": 339880
    },
    {
      "epoch": 132.0,
      "learning_rate": 2.400517799352751e-06,
      "loss": 0.04,
      "step": 339890
    },
    {
      "epoch": 132.0,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 0.0228,
      "step": 339900
    },
    {
      "epoch": 132.0,
      "eval_accuracy": 0.951856946354883,
      "eval_loss": 0.3912859261035919,
      "eval_runtime": 8.2608,
      "eval_samples_per_second": 440.032,
      "eval_steps_per_second": 55.08,
      "step": 339900
    },
    {
      "epoch": 132.0,
      "learning_rate": 2.399482200647249e-06,
      "loss": 0.0232,
      "step": 339910
    },
    {
      "epoch": 132.01,
      "learning_rate": 2.3989644012944983e-06,
      "loss": 0.0073,
      "step": 339920
    },
    {
      "epoch": 132.01,
      "learning_rate": 2.3984466019417475e-06,
      "loss": 0.0469,
      "step": 339930
    },
    {
      "epoch": 132.02,
      "learning_rate": 2.3979288025889967e-06,
      "loss": 0.0579,
      "step": 339940
    },
    {
      "epoch": 132.02,
      "learning_rate": 2.397411003236246e-06,
      "loss": 0.0675,
      "step": 339950
    },
    {
      "epoch": 132.02,
      "learning_rate": 2.396893203883495e-06,
      "loss": 0.0001,
      "step": 339960
    },
    {
      "epoch": 132.03,
      "learning_rate": 2.3963754045307443e-06,
      "loss": 0.0351,
      "step": 339970
    },
    {
      "epoch": 132.03,
      "learning_rate": 2.3958576051779934e-06,
      "loss": 0.088,
      "step": 339980
    },
    {
      "epoch": 132.03,
      "learning_rate": 2.3953398058252426e-06,
      "loss": 0.0344,
      "step": 339990
    },
    {
      "epoch": 132.04,
      "learning_rate": 2.394822006472492e-06,
      "loss": 0.0282,
      "step": 340000
    },
    {
      "epoch": 132.04,
      "learning_rate": 2.3943042071197415e-06,
      "loss": 0.0367,
      "step": 340010
    },
    {
      "epoch": 132.05,
      "learning_rate": 2.3937864077669906e-06,
      "loss": 0.0486,
      "step": 340020
    },
    {
      "epoch": 132.05,
      "learning_rate": 2.39326860841424e-06,
      "loss": 0.0542,
      "step": 340030
    },
    {
      "epoch": 132.05,
      "learning_rate": 2.392750809061489e-06,
      "loss": 0.1233,
      "step": 340040
    },
    {
      "epoch": 132.06,
      "learning_rate": 2.3922330097087382e-06,
      "loss": 0.0047,
      "step": 340050
    },
    {
      "epoch": 132.06,
      "learning_rate": 2.3917152103559874e-06,
      "loss": 0.0011,
      "step": 340060
    },
    {
      "epoch": 132.07,
      "learning_rate": 2.3911974110032366e-06,
      "loss": 0.0991,
      "step": 340070
    },
    {
      "epoch": 132.07,
      "learning_rate": 2.390679611650486e-06,
      "loss": 0.0094,
      "step": 340080
    },
    {
      "epoch": 132.07,
      "learning_rate": 2.390161812297735e-06,
      "loss": 0.0092,
      "step": 340090
    },
    {
      "epoch": 132.08,
      "learning_rate": 2.389644012944984e-06,
      "loss": 0.1107,
      "step": 340100
    },
    {
      "epoch": 132.08,
      "learning_rate": 2.3891262135922334e-06,
      "loss": 0.1126,
      "step": 340110
    },
    {
      "epoch": 132.09,
      "learning_rate": 2.3886084142394826e-06,
      "loss": 0.0193,
      "step": 340120
    },
    {
      "epoch": 132.09,
      "learning_rate": 2.388090614886732e-06,
      "loss": 0.0391,
      "step": 340130
    },
    {
      "epoch": 132.09,
      "learning_rate": 2.387572815533981e-06,
      "loss": 0.0114,
      "step": 340140
    },
    {
      "epoch": 132.1,
      "learning_rate": 2.38705501618123e-06,
      "loss": 0.0327,
      "step": 340150
    },
    {
      "epoch": 132.1,
      "learning_rate": 2.3865372168284794e-06,
      "loss": 0.0094,
      "step": 340160
    },
    {
      "epoch": 132.1,
      "learning_rate": 2.3860194174757286e-06,
      "loss": 0.0088,
      "step": 340170
    },
    {
      "epoch": 132.11,
      "learning_rate": 2.3855016181229778e-06,
      "loss": 0.043,
      "step": 340180
    },
    {
      "epoch": 132.11,
      "learning_rate": 2.384983818770227e-06,
      "loss": 0.0205,
      "step": 340190
    },
    {
      "epoch": 132.12,
      "learning_rate": 2.384466019417476e-06,
      "loss": 0.0086,
      "step": 340200
    },
    {
      "epoch": 132.12,
      "learning_rate": 2.3839482200647253e-06,
      "loss": 0.0025,
      "step": 340210
    },
    {
      "epoch": 132.12,
      "learning_rate": 2.3834304207119745e-06,
      "loss": 0.0206,
      "step": 340220
    },
    {
      "epoch": 132.13,
      "learning_rate": 2.3829126213592233e-06,
      "loss": 0.0468,
      "step": 340230
    },
    {
      "epoch": 132.13,
      "learning_rate": 2.3823948220064725e-06,
      "loss": 0.0496,
      "step": 340240
    },
    {
      "epoch": 132.14,
      "learning_rate": 2.3818770226537217e-06,
      "loss": 0.019,
      "step": 340250
    },
    {
      "epoch": 132.14,
      "learning_rate": 2.381359223300971e-06,
      "loss": 0.0921,
      "step": 340260
    },
    {
      "epoch": 132.14,
      "learning_rate": 2.38084142394822e-06,
      "loss": 0.1295,
      "step": 340270
    },
    {
      "epoch": 132.15,
      "learning_rate": 2.3803236245954693e-06,
      "loss": 0.0147,
      "step": 340280
    },
    {
      "epoch": 132.15,
      "learning_rate": 2.3798058252427185e-06,
      "loss": 0.052,
      "step": 340290
    },
    {
      "epoch": 132.16,
      "learning_rate": 2.3792880258899677e-06,
      "loss": 0.0461,
      "step": 340300
    },
    {
      "epoch": 132.16,
      "learning_rate": 2.378770226537217e-06,
      "loss": 0.0696,
      "step": 340310
    },
    {
      "epoch": 132.16,
      "learning_rate": 2.378252427184466e-06,
      "loss": 0.0405,
      "step": 340320
    },
    {
      "epoch": 132.17,
      "learning_rate": 2.3777346278317153e-06,
      "loss": 0.0004,
      "step": 340330
    },
    {
      "epoch": 132.17,
      "learning_rate": 2.3772168284789645e-06,
      "loss": 0.0195,
      "step": 340340
    },
    {
      "epoch": 132.17,
      "learning_rate": 2.3766990291262136e-06,
      "loss": 0.1409,
      "step": 340350
    },
    {
      "epoch": 132.18,
      "learning_rate": 2.376181229773463e-06,
      "loss": 0.0007,
      "step": 340360
    },
    {
      "epoch": 132.18,
      "learning_rate": 2.375663430420712e-06,
      "loss": 0.001,
      "step": 340370
    },
    {
      "epoch": 132.19,
      "learning_rate": 2.3751456310679612e-06,
      "loss": 0.0257,
      "step": 340380
    },
    {
      "epoch": 132.19,
      "learning_rate": 2.3746278317152104e-06,
      "loss": 0.084,
      "step": 340390
    },
    {
      "epoch": 132.19,
      "learning_rate": 2.3741100323624596e-06,
      "loss": 0.0348,
      "step": 340400
    },
    {
      "epoch": 132.2,
      "learning_rate": 2.373592233009709e-06,
      "loss": 0.0971,
      "step": 340410
    },
    {
      "epoch": 132.2,
      "learning_rate": 2.373074433656958e-06,
      "loss": 0.0281,
      "step": 340420
    },
    {
      "epoch": 132.21,
      "learning_rate": 2.372556634304207e-06,
      "loss": 0.0701,
      "step": 340430
    },
    {
      "epoch": 132.21,
      "learning_rate": 2.3720388349514564e-06,
      "loss": 0.0025,
      "step": 340440
    },
    {
      "epoch": 132.21,
      "learning_rate": 2.3715210355987056e-06,
      "loss": 0.0373,
      "step": 340450
    },
    {
      "epoch": 132.22,
      "learning_rate": 2.3710032362459548e-06,
      "loss": 0.1042,
      "step": 340460
    },
    {
      "epoch": 132.22,
      "learning_rate": 2.370485436893204e-06,
      "loss": 0.069,
      "step": 340470
    },
    {
      "epoch": 132.23,
      "learning_rate": 2.369967637540453e-06,
      "loss": 0.0208,
      "step": 340480
    },
    {
      "epoch": 132.23,
      "learning_rate": 2.3694498381877024e-06,
      "loss": 0.0048,
      "step": 340490
    },
    {
      "epoch": 132.23,
      "learning_rate": 2.3689320388349516e-06,
      "loss": 0.022,
      "step": 340500
    },
    {
      "epoch": 132.24,
      "learning_rate": 2.3684142394822008e-06,
      "loss": 0.0131,
      "step": 340510
    },
    {
      "epoch": 132.24,
      "learning_rate": 2.36789644012945e-06,
      "loss": 0.1183,
      "step": 340520
    },
    {
      "epoch": 132.24,
      "learning_rate": 2.367378640776699e-06,
      "loss": 0.0328,
      "step": 340530
    },
    {
      "epoch": 132.25,
      "learning_rate": 2.3668608414239483e-06,
      "loss": 0.0539,
      "step": 340540
    },
    {
      "epoch": 132.25,
      "learning_rate": 2.3663430420711975e-06,
      "loss": 0.097,
      "step": 340550
    },
    {
      "epoch": 132.26,
      "learning_rate": 2.3658252427184467e-06,
      "loss": 0.0225,
      "step": 340560
    },
    {
      "epoch": 132.26,
      "learning_rate": 2.365307443365696e-06,
      "loss": 0.1175,
      "step": 340570
    },
    {
      "epoch": 132.26,
      "learning_rate": 2.364789644012945e-06,
      "loss": 0.0069,
      "step": 340580
    },
    {
      "epoch": 132.27,
      "learning_rate": 2.3642718446601943e-06,
      "loss": 0.0533,
      "step": 340590
    },
    {
      "epoch": 132.27,
      "learning_rate": 2.3637540453074435e-06,
      "loss": 0.0165,
      "step": 340600
    },
    {
      "epoch": 132.28,
      "learning_rate": 2.3632362459546927e-06,
      "loss": 0.0108,
      "step": 340610
    },
    {
      "epoch": 132.28,
      "learning_rate": 2.362718446601942e-06,
      "loss": 0.158,
      "step": 340620
    },
    {
      "epoch": 132.28,
      "learning_rate": 2.362200647249191e-06,
      "loss": 0.035,
      "step": 340630
    },
    {
      "epoch": 132.29,
      "learning_rate": 2.3616828478964403e-06,
      "loss": 0.0258,
      "step": 340640
    },
    {
      "epoch": 132.29,
      "learning_rate": 2.3611650485436895e-06,
      "loss": 0.0181,
      "step": 340650
    },
    {
      "epoch": 132.3,
      "learning_rate": 2.3606472491909387e-06,
      "loss": 0.0394,
      "step": 340660
    },
    {
      "epoch": 132.3,
      "learning_rate": 2.360129449838188e-06,
      "loss": 0.0735,
      "step": 340670
    },
    {
      "epoch": 132.3,
      "learning_rate": 2.359611650485437e-06,
      "loss": 0.0916,
      "step": 340680
    },
    {
      "epoch": 132.31,
      "learning_rate": 2.3590938511326863e-06,
      "loss": 0.0227,
      "step": 340690
    },
    {
      "epoch": 132.31,
      "learning_rate": 2.3585760517799355e-06,
      "loss": 0.0005,
      "step": 340700
    },
    {
      "epoch": 132.31,
      "learning_rate": 2.3580582524271846e-06,
      "loss": 0.046,
      "step": 340710
    },
    {
      "epoch": 132.32,
      "learning_rate": 2.357540453074434e-06,
      "loss": 0.0143,
      "step": 340720
    },
    {
      "epoch": 132.32,
      "learning_rate": 2.357022653721683e-06,
      "loss": 0.0694,
      "step": 340730
    },
    {
      "epoch": 132.33,
      "learning_rate": 2.3565048543689322e-06,
      "loss": 0.0042,
      "step": 340740
    },
    {
      "epoch": 132.33,
      "learning_rate": 2.3559870550161814e-06,
      "loss": 0.0102,
      "step": 340750
    },
    {
      "epoch": 132.33,
      "learning_rate": 2.3554692556634306e-06,
      "loss": 0.0503,
      "step": 340760
    },
    {
      "epoch": 132.34,
      "learning_rate": 2.35495145631068e-06,
      "loss": 0.094,
      "step": 340770
    },
    {
      "epoch": 132.34,
      "learning_rate": 2.354433656957929e-06,
      "loss": 0.0181,
      "step": 340780
    },
    {
      "epoch": 132.35,
      "learning_rate": 2.353915857605178e-06,
      "loss": 0.0071,
      "step": 340790
    },
    {
      "epoch": 132.35,
      "learning_rate": 2.3533980582524274e-06,
      "loss": 0.0535,
      "step": 340800
    },
    {
      "epoch": 132.35,
      "learning_rate": 2.3528802588996766e-06,
      "loss": 0.0394,
      "step": 340810
    },
    {
      "epoch": 132.36,
      "learning_rate": 2.3523624595469258e-06,
      "loss": 0.0227,
      "step": 340820
    },
    {
      "epoch": 132.36,
      "learning_rate": 2.351844660194175e-06,
      "loss": 0.0002,
      "step": 340830
    },
    {
      "epoch": 132.37,
      "learning_rate": 2.351326860841424e-06,
      "loss": 0.0176,
      "step": 340840
    },
    {
      "epoch": 132.37,
      "learning_rate": 2.3508090614886734e-06,
      "loss": 0.0158,
      "step": 340850
    },
    {
      "epoch": 132.37,
      "learning_rate": 2.3502912621359226e-06,
      "loss": 0.0647,
      "step": 340860
    },
    {
      "epoch": 132.38,
      "learning_rate": 2.3497734627831718e-06,
      "loss": 0.0205,
      "step": 340870
    },
    {
      "epoch": 132.38,
      "learning_rate": 2.349255663430421e-06,
      "loss": 0.0513,
      "step": 340880
    },
    {
      "epoch": 132.38,
      "learning_rate": 2.34873786407767e-06,
      "loss": 0.0583,
      "step": 340890
    },
    {
      "epoch": 132.39,
      "learning_rate": 2.3482200647249193e-06,
      "loss": 0.0665,
      "step": 340900
    },
    {
      "epoch": 132.39,
      "learning_rate": 2.3477022653721685e-06,
      "loss": 0.0005,
      "step": 340910
    },
    {
      "epoch": 132.4,
      "learning_rate": 2.3471844660194177e-06,
      "loss": 0.0002,
      "step": 340920
    },
    {
      "epoch": 132.4,
      "learning_rate": 2.346666666666667e-06,
      "loss": 0.0403,
      "step": 340930
    },
    {
      "epoch": 132.4,
      "learning_rate": 2.346148867313916e-06,
      "loss": 0.0048,
      "step": 340940
    },
    {
      "epoch": 132.41,
      "learning_rate": 2.3456310679611653e-06,
      "loss": 0.0076,
      "step": 340950
    },
    {
      "epoch": 132.41,
      "learning_rate": 2.3451132686084145e-06,
      "loss": 0.0947,
      "step": 340960
    },
    {
      "epoch": 132.42,
      "learning_rate": 2.3445954692556637e-06,
      "loss": 0.0677,
      "step": 340970
    },
    {
      "epoch": 132.42,
      "learning_rate": 2.344077669902913e-06,
      "loss": 0.073,
      "step": 340980
    },
    {
      "epoch": 132.42,
      "learning_rate": 2.343559870550162e-06,
      "loss": 0.0614,
      "step": 340990
    },
    {
      "epoch": 132.43,
      "learning_rate": 2.3430420711974113e-06,
      "loss": 0.0004,
      "step": 341000
    },
    {
      "epoch": 132.43,
      "learning_rate": 2.3425242718446605e-06,
      "loss": 0.0099,
      "step": 341010
    },
    {
      "epoch": 132.43,
      "learning_rate": 2.3420064724919097e-06,
      "loss": 0.0155,
      "step": 341020
    },
    {
      "epoch": 132.44,
      "learning_rate": 2.341488673139159e-06,
      "loss": 0.0098,
      "step": 341030
    },
    {
      "epoch": 132.44,
      "learning_rate": 2.340970873786408e-06,
      "loss": 0.0006,
      "step": 341040
    },
    {
      "epoch": 132.45,
      "learning_rate": 2.340453074433657e-06,
      "loss": 0.0004,
      "step": 341050
    },
    {
      "epoch": 132.45,
      "learning_rate": 2.339935275080906e-06,
      "loss": 0.0806,
      "step": 341060
    },
    {
      "epoch": 132.45,
      "learning_rate": 2.3394174757281552e-06,
      "loss": 0.0001,
      "step": 341070
    },
    {
      "epoch": 132.46,
      "learning_rate": 2.3388996763754044e-06,
      "loss": 0.0471,
      "step": 341080
    },
    {
      "epoch": 132.46,
      "learning_rate": 2.3383818770226536e-06,
      "loss": 0.0444,
      "step": 341090
    },
    {
      "epoch": 132.47,
      "learning_rate": 2.337864077669903e-06,
      "loss": 0.0006,
      "step": 341100
    },
    {
      "epoch": 132.47,
      "learning_rate": 2.337346278317152e-06,
      "loss": 0.0004,
      "step": 341110
    },
    {
      "epoch": 132.47,
      "learning_rate": 2.336828478964401e-06,
      "loss": 0.0367,
      "step": 341120
    },
    {
      "epoch": 132.48,
      "learning_rate": 2.336310679611651e-06,
      "loss": 0.0626,
      "step": 341130
    },
    {
      "epoch": 132.48,
      "learning_rate": 2.3357928802589e-06,
      "loss": 0.1194,
      "step": 341140
    },
    {
      "epoch": 132.49,
      "learning_rate": 2.335275080906149e-06,
      "loss": 0.0438,
      "step": 341150
    },
    {
      "epoch": 132.49,
      "learning_rate": 2.3347572815533984e-06,
      "loss": 0.0044,
      "step": 341160
    },
    {
      "epoch": 132.49,
      "learning_rate": 2.3342394822006476e-06,
      "loss": 0.0005,
      "step": 341170
    },
    {
      "epoch": 132.5,
      "learning_rate": 2.333721682847897e-06,
      "loss": 0.0566,
      "step": 341180
    },
    {
      "epoch": 132.5,
      "learning_rate": 2.333203883495146e-06,
      "loss": 0.0812,
      "step": 341190
    },
    {
      "epoch": 132.5,
      "learning_rate": 2.332686084142395e-06,
      "loss": 0.0156,
      "step": 341200
    },
    {
      "epoch": 132.51,
      "learning_rate": 2.3321682847896444e-06,
      "loss": 0.0528,
      "step": 341210
    },
    {
      "epoch": 132.51,
      "learning_rate": 2.3316504854368936e-06,
      "loss": 0.0019,
      "step": 341220
    },
    {
      "epoch": 132.52,
      "learning_rate": 2.3311326860841428e-06,
      "loss": 0.0102,
      "step": 341230
    },
    {
      "epoch": 132.52,
      "learning_rate": 2.330614886731392e-06,
      "loss": 0.0118,
      "step": 341240
    },
    {
      "epoch": 132.52,
      "learning_rate": 2.330097087378641e-06,
      "loss": 0.0005,
      "step": 341250
    },
    {
      "epoch": 132.53,
      "learning_rate": 2.3295792880258903e-06,
      "loss": 0.0304,
      "step": 341260
    },
    {
      "epoch": 132.53,
      "learning_rate": 2.3290614886731395e-06,
      "loss": 0.0275,
      "step": 341270
    },
    {
      "epoch": 132.54,
      "learning_rate": 2.3285436893203887e-06,
      "loss": 0.105,
      "step": 341280
    },
    {
      "epoch": 132.54,
      "learning_rate": 2.328025889967638e-06,
      "loss": 0.0344,
      "step": 341290
    },
    {
      "epoch": 132.54,
      "learning_rate": 2.327508090614887e-06,
      "loss": 0.0801,
      "step": 341300
    },
    {
      "epoch": 132.55,
      "learning_rate": 2.3269902912621363e-06,
      "loss": 0.0566,
      "step": 341310
    },
    {
      "epoch": 132.55,
      "learning_rate": 2.3264724919093855e-06,
      "loss": 0.0111,
      "step": 341320
    },
    {
      "epoch": 132.56,
      "learning_rate": 2.3259546925566347e-06,
      "loss": 0.0793,
      "step": 341330
    },
    {
      "epoch": 132.56,
      "learning_rate": 2.325436893203884e-06,
      "loss": 0.0002,
      "step": 341340
    },
    {
      "epoch": 132.56,
      "learning_rate": 2.324919093851133e-06,
      "loss": 0.0015,
      "step": 341350
    },
    {
      "epoch": 132.57,
      "learning_rate": 2.3244012944983823e-06,
      "loss": 0.0204,
      "step": 341360
    },
    {
      "epoch": 132.57,
      "learning_rate": 2.323883495145631e-06,
      "loss": 0.0029,
      "step": 341370
    },
    {
      "epoch": 132.57,
      "learning_rate": 2.3233656957928803e-06,
      "loss": 0.007,
      "step": 341380
    },
    {
      "epoch": 132.58,
      "learning_rate": 2.3228478964401295e-06,
      "loss": 0.0157,
      "step": 341390
    },
    {
      "epoch": 132.58,
      "learning_rate": 2.3223300970873786e-06,
      "loss": 0.0381,
      "step": 341400
    },
    {
      "epoch": 132.59,
      "learning_rate": 2.321812297734628e-06,
      "loss": 0.0182,
      "step": 341410
    },
    {
      "epoch": 132.59,
      "learning_rate": 2.321294498381877e-06,
      "loss": 0.0002,
      "step": 341420
    },
    {
      "epoch": 132.59,
      "learning_rate": 2.3207766990291262e-06,
      "loss": 0.1808,
      "step": 341430
    },
    {
      "epoch": 132.6,
      "learning_rate": 2.3202588996763754e-06,
      "loss": 0.0607,
      "step": 341440
    },
    {
      "epoch": 132.6,
      "learning_rate": 2.3197411003236246e-06,
      "loss": 0.0855,
      "step": 341450
    },
    {
      "epoch": 132.61,
      "learning_rate": 2.319223300970874e-06,
      "loss": 0.0539,
      "step": 341460
    },
    {
      "epoch": 132.61,
      "learning_rate": 2.318705501618123e-06,
      "loss": 0.0166,
      "step": 341470
    },
    {
      "epoch": 132.61,
      "learning_rate": 2.318187702265372e-06,
      "loss": 0.0206,
      "step": 341480
    },
    {
      "epoch": 132.62,
      "learning_rate": 2.3176699029126214e-06,
      "loss": 0.0002,
      "step": 341490
    },
    {
      "epoch": 132.62,
      "learning_rate": 2.3171521035598706e-06,
      "loss": 0.0275,
      "step": 341500
    },
    {
      "epoch": 132.63,
      "learning_rate": 2.3166343042071198e-06,
      "loss": 0.1134,
      "step": 341510
    },
    {
      "epoch": 132.63,
      "learning_rate": 2.316116504854369e-06,
      "loss": 0.1037,
      "step": 341520
    },
    {
      "epoch": 132.63,
      "learning_rate": 2.315598705501618e-06,
      "loss": 0.0552,
      "step": 341530
    },
    {
      "epoch": 132.64,
      "learning_rate": 2.3150809061488674e-06,
      "loss": 0.0548,
      "step": 341540
    },
    {
      "epoch": 132.64,
      "learning_rate": 2.3145631067961166e-06,
      "loss": 0.0787,
      "step": 341550
    },
    {
      "epoch": 132.64,
      "learning_rate": 2.3140453074433658e-06,
      "loss": 0.0408,
      "step": 341560
    },
    {
      "epoch": 132.65,
      "learning_rate": 2.313527508090615e-06,
      "loss": 0.0008,
      "step": 341570
    },
    {
      "epoch": 132.65,
      "learning_rate": 2.313009708737864e-06,
      "loss": 0.0303,
      "step": 341580
    },
    {
      "epoch": 132.66,
      "learning_rate": 2.3124919093851133e-06,
      "loss": 0.0009,
      "step": 341590
    },
    {
      "epoch": 132.66,
      "learning_rate": 2.3119741100323625e-06,
      "loss": 0.062,
      "step": 341600
    },
    {
      "epoch": 132.66,
      "learning_rate": 2.3114563106796117e-06,
      "loss": 0.0151,
      "step": 341610
    },
    {
      "epoch": 132.67,
      "learning_rate": 2.310938511326861e-06,
      "loss": 0.0168,
      "step": 341620
    },
    {
      "epoch": 132.67,
      "learning_rate": 2.31042071197411e-06,
      "loss": 0.0102,
      "step": 341630
    },
    {
      "epoch": 132.68,
      "learning_rate": 2.3099029126213593e-06,
      "loss": 0.0604,
      "step": 341640
    },
    {
      "epoch": 132.68,
      "learning_rate": 2.3093851132686085e-06,
      "loss": 0.0086,
      "step": 341650
    },
    {
      "epoch": 132.68,
      "learning_rate": 2.3088673139158577e-06,
      "loss": 0.0174,
      "step": 341660
    },
    {
      "epoch": 132.69,
      "learning_rate": 2.308349514563107e-06,
      "loss": 0.0188,
      "step": 341670
    },
    {
      "epoch": 132.69,
      "learning_rate": 2.307831715210356e-06,
      "loss": 0.0014,
      "step": 341680
    },
    {
      "epoch": 132.7,
      "learning_rate": 2.3073139158576057e-06,
      "loss": 0.0028,
      "step": 341690
    },
    {
      "epoch": 132.7,
      "learning_rate": 2.3067961165048545e-06,
      "loss": 0.025,
      "step": 341700
    },
    {
      "epoch": 132.7,
      "learning_rate": 2.3062783171521037e-06,
      "loss": 0.0637,
      "step": 341710
    },
    {
      "epoch": 132.71,
      "learning_rate": 2.305760517799353e-06,
      "loss": 0.0173,
      "step": 341720
    },
    {
      "epoch": 132.71,
      "learning_rate": 2.305242718446602e-06,
      "loss": 0.0102,
      "step": 341730
    },
    {
      "epoch": 132.71,
      "learning_rate": 2.3047249190938513e-06,
      "loss": 0.0927,
      "step": 341740
    },
    {
      "epoch": 132.72,
      "learning_rate": 2.3042071197411005e-06,
      "loss": 0.0576,
      "step": 341750
    },
    {
      "epoch": 132.72,
      "learning_rate": 2.3036893203883496e-06,
      "loss": 0.0098,
      "step": 341760
    },
    {
      "epoch": 132.73,
      "learning_rate": 2.303171521035599e-06,
      "loss": 0.01,
      "step": 341770
    },
    {
      "epoch": 132.73,
      "learning_rate": 2.302653721682848e-06,
      "loss": 0.0307,
      "step": 341780
    },
    {
      "epoch": 132.73,
      "learning_rate": 2.3021359223300972e-06,
      "loss": 0.0827,
      "step": 341790
    },
    {
      "epoch": 132.74,
      "learning_rate": 2.3016181229773464e-06,
      "loss": 0.0719,
      "step": 341800
    },
    {
      "epoch": 132.74,
      "learning_rate": 2.3011003236245956e-06,
      "loss": 0.0706,
      "step": 341810
    },
    {
      "epoch": 132.75,
      "learning_rate": 2.300582524271845e-06,
      "loss": 0.0528,
      "step": 341820
    },
    {
      "epoch": 132.75,
      "learning_rate": 2.300064724919094e-06,
      "loss": 0.0794,
      "step": 341830
    },
    {
      "epoch": 132.75,
      "learning_rate": 2.299546925566343e-06,
      "loss": 0.0377,
      "step": 341840
    },
    {
      "epoch": 132.76,
      "learning_rate": 2.2990291262135924e-06,
      "loss": 0.0726,
      "step": 341850
    },
    {
      "epoch": 132.76,
      "learning_rate": 2.2985113268608416e-06,
      "loss": 0.0111,
      "step": 341860
    },
    {
      "epoch": 132.77,
      "learning_rate": 2.297993527508091e-06,
      "loss": 0.0349,
      "step": 341870
    },
    {
      "epoch": 132.77,
      "learning_rate": 2.29747572815534e-06,
      "loss": 0.0249,
      "step": 341880
    },
    {
      "epoch": 132.77,
      "learning_rate": 2.296957928802589e-06,
      "loss": 0.0241,
      "step": 341890
    },
    {
      "epoch": 132.78,
      "learning_rate": 2.2964401294498384e-06,
      "loss": 0.0003,
      "step": 341900
    },
    {
      "epoch": 132.78,
      "learning_rate": 2.2959223300970876e-06,
      "loss": 0.0113,
      "step": 341910
    },
    {
      "epoch": 132.78,
      "learning_rate": 2.2954045307443368e-06,
      "loss": 0.0162,
      "step": 341920
    },
    {
      "epoch": 132.79,
      "learning_rate": 2.294886731391586e-06,
      "loss": 0.0322,
      "step": 341930
    },
    {
      "epoch": 132.79,
      "learning_rate": 2.294368932038835e-06,
      "loss": 0.0776,
      "step": 341940
    },
    {
      "epoch": 132.8,
      "learning_rate": 2.2938511326860843e-06,
      "loss": 0.0179,
      "step": 341950
    },
    {
      "epoch": 132.8,
      "learning_rate": 2.2933333333333335e-06,
      "loss": 0.0201,
      "step": 341960
    },
    {
      "epoch": 132.8,
      "learning_rate": 2.2928155339805827e-06,
      "loss": 0.0942,
      "step": 341970
    },
    {
      "epoch": 132.81,
      "learning_rate": 2.292297734627832e-06,
      "loss": 0.0383,
      "step": 341980
    },
    {
      "epoch": 132.81,
      "learning_rate": 2.291779935275081e-06,
      "loss": 0.0205,
      "step": 341990
    },
    {
      "epoch": 132.82,
      "learning_rate": 2.2912621359223303e-06,
      "loss": 0.0007,
      "step": 342000
    },
    {
      "epoch": 132.82,
      "learning_rate": 2.2907443365695795e-06,
      "loss": 0.0521,
      "step": 342010
    },
    {
      "epoch": 132.82,
      "learning_rate": 2.2902265372168287e-06,
      "loss": 0.001,
      "step": 342020
    },
    {
      "epoch": 132.83,
      "learning_rate": 2.289708737864078e-06,
      "loss": 0.0153,
      "step": 342030
    },
    {
      "epoch": 132.83,
      "learning_rate": 2.289190938511327e-06,
      "loss": 0.0873,
      "step": 342040
    },
    {
      "epoch": 132.83,
      "learning_rate": 2.2886731391585763e-06,
      "loss": 0.0119,
      "step": 342050
    },
    {
      "epoch": 132.84,
      "learning_rate": 2.2881553398058255e-06,
      "loss": 0.0006,
      "step": 342060
    },
    {
      "epoch": 132.84,
      "learning_rate": 2.2876375404530747e-06,
      "loss": 0.0002,
      "step": 342070
    },
    {
      "epoch": 132.85,
      "learning_rate": 2.287119741100324e-06,
      "loss": 0.0787,
      "step": 342080
    },
    {
      "epoch": 132.85,
      "learning_rate": 2.286601941747573e-06,
      "loss": 0.0345,
      "step": 342090
    },
    {
      "epoch": 132.85,
      "learning_rate": 2.2860841423948223e-06,
      "loss": 0.0206,
      "step": 342100
    },
    {
      "epoch": 132.86,
      "learning_rate": 2.2855663430420715e-06,
      "loss": 0.1171,
      "step": 342110
    },
    {
      "epoch": 132.86,
      "learning_rate": 2.2850485436893207e-06,
      "loss": 0.1066,
      "step": 342120
    },
    {
      "epoch": 132.87,
      "learning_rate": 2.28453074433657e-06,
      "loss": 0.0773,
      "step": 342130
    },
    {
      "epoch": 132.87,
      "learning_rate": 2.284012944983819e-06,
      "loss": 0.0014,
      "step": 342140
    },
    {
      "epoch": 132.87,
      "learning_rate": 2.2834951456310682e-06,
      "loss": 0.0536,
      "step": 342150
    },
    {
      "epoch": 132.88,
      "learning_rate": 2.2829773462783174e-06,
      "loss": 0.0604,
      "step": 342160
    },
    {
      "epoch": 132.88,
      "learning_rate": 2.2824595469255666e-06,
      "loss": 0.0102,
      "step": 342170
    },
    {
      "epoch": 132.89,
      "learning_rate": 2.281941747572816e-06,
      "loss": 0.1004,
      "step": 342180
    },
    {
      "epoch": 132.89,
      "learning_rate": 2.2814239482200646e-06,
      "loss": 0.0416,
      "step": 342190
    },
    {
      "epoch": 132.89,
      "learning_rate": 2.2809061488673138e-06,
      "loss": 0.0033,
      "step": 342200
    },
    {
      "epoch": 132.9,
      "learning_rate": 2.280388349514563e-06,
      "loss": 0.0341,
      "step": 342210
    },
    {
      "epoch": 132.9,
      "learning_rate": 2.279870550161812e-06,
      "loss": 0.0726,
      "step": 342220
    },
    {
      "epoch": 132.9,
      "learning_rate": 2.2793527508090614e-06,
      "loss": 0.0489,
      "step": 342230
    },
    {
      "epoch": 132.91,
      "learning_rate": 2.2788349514563106e-06,
      "loss": 0.0001,
      "step": 342240
    },
    {
      "epoch": 132.91,
      "learning_rate": 2.27831715210356e-06,
      "loss": 0.018,
      "step": 342250
    },
    {
      "epoch": 132.92,
      "learning_rate": 2.2777993527508094e-06,
      "loss": 0.055,
      "step": 342260
    },
    {
      "epoch": 132.92,
      "learning_rate": 2.2772815533980586e-06,
      "loss": 0.0085,
      "step": 342270
    },
    {
      "epoch": 132.92,
      "learning_rate": 2.2767637540453078e-06,
      "loss": 0.001,
      "step": 342280
    },
    {
      "epoch": 132.93,
      "learning_rate": 2.276245954692557e-06,
      "loss": 0.0002,
      "step": 342290
    },
    {
      "epoch": 132.93,
      "learning_rate": 2.275728155339806e-06,
      "loss": 0.0001,
      "step": 342300
    },
    {
      "epoch": 132.94,
      "learning_rate": 2.2752103559870553e-06,
      "loss": 0.0768,
      "step": 342310
    },
    {
      "epoch": 132.94,
      "learning_rate": 2.2746925566343045e-06,
      "loss": 0.0091,
      "step": 342320
    },
    {
      "epoch": 132.94,
      "learning_rate": 2.2741747572815537e-06,
      "loss": 0.0764,
      "step": 342330
    },
    {
      "epoch": 132.95,
      "learning_rate": 2.273656957928803e-06,
      "loss": 0.1267,
      "step": 342340
    },
    {
      "epoch": 132.95,
      "learning_rate": 2.273139158576052e-06,
      "loss": 0.0367,
      "step": 342350
    },
    {
      "epoch": 132.96,
      "learning_rate": 2.2726213592233013e-06,
      "loss": 0.0392,
      "step": 342360
    },
    {
      "epoch": 132.96,
      "learning_rate": 2.2721035598705505e-06,
      "loss": 0.0085,
      "step": 342370
    },
    {
      "epoch": 132.96,
      "learning_rate": 2.2715857605177997e-06,
      "loss": 0.0441,
      "step": 342380
    },
    {
      "epoch": 132.97,
      "learning_rate": 2.271067961165049e-06,
      "loss": 0.0056,
      "step": 342390
    },
    {
      "epoch": 132.97,
      "learning_rate": 2.270550161812298e-06,
      "loss": 0.0219,
      "step": 342400
    },
    {
      "epoch": 132.97,
      "learning_rate": 2.2700323624595473e-06,
      "loss": 0.0007,
      "step": 342410
    },
    {
      "epoch": 132.98,
      "learning_rate": 2.2695145631067965e-06,
      "loss": 0.0147,
      "step": 342420
    },
    {
      "epoch": 132.98,
      "learning_rate": 2.2689967637540457e-06,
      "loss": 0.0177,
      "step": 342430
    },
    {
      "epoch": 132.99,
      "learning_rate": 2.268478964401295e-06,
      "loss": 0.1327,
      "step": 342440
    },
    {
      "epoch": 132.99,
      "learning_rate": 2.267961165048544e-06,
      "loss": 0.011,
      "step": 342450
    },
    {
      "epoch": 132.99,
      "learning_rate": 2.2674433656957933e-06,
      "loss": 0.0011,
      "step": 342460
    },
    {
      "epoch": 133.0,
      "learning_rate": 2.2669255663430425e-06,
      "loss": 0.0094,
      "step": 342470
    },
    {
      "epoch": 133.0,
      "eval_accuracy": 0.9513067400275104,
      "eval_loss": 0.3967961072921753,
      "eval_runtime": 8.2847,
      "eval_samples_per_second": 438.76,
      "eval_steps_per_second": 54.92,
      "step": 342475
    },
    {
      "epoch": 133.0,
      "learning_rate": 2.2664077669902917e-06,
      "loss": 0.0257,
      "step": 342480
    },
    {
      "epoch": 133.01,
      "learning_rate": 2.265889967637541e-06,
      "loss": 0.0004,
      "step": 342490
    },
    {
      "epoch": 133.01,
      "learning_rate": 2.26537216828479e-06,
      "loss": 0.0972,
      "step": 342500
    },
    {
      "epoch": 133.01,
      "learning_rate": 2.264854368932039e-06,
      "loss": 0.0001,
      "step": 342510
    },
    {
      "epoch": 133.02,
      "learning_rate": 2.264336569579288e-06,
      "loss": 0.0053,
      "step": 342520
    },
    {
      "epoch": 133.02,
      "learning_rate": 2.263818770226537e-06,
      "loss": 0.0192,
      "step": 342530
    },
    {
      "epoch": 133.03,
      "learning_rate": 2.2633009708737864e-06,
      "loss": 0.0438,
      "step": 342540
    },
    {
      "epoch": 133.03,
      "learning_rate": 2.2627831715210356e-06,
      "loss": 0.068,
      "step": 342550
    },
    {
      "epoch": 133.03,
      "learning_rate": 2.2622653721682848e-06,
      "loss": 0.1022,
      "step": 342560
    },
    {
      "epoch": 133.04,
      "learning_rate": 2.261747572815534e-06,
      "loss": 0.0321,
      "step": 342570
    },
    {
      "epoch": 133.04,
      "learning_rate": 2.261229773462783e-06,
      "loss": 0.1201,
      "step": 342580
    },
    {
      "epoch": 133.04,
      "learning_rate": 2.2607119741100324e-06,
      "loss": 0.0472,
      "step": 342590
    },
    {
      "epoch": 133.05,
      "learning_rate": 2.2601941747572816e-06,
      "loss": 0.0723,
      "step": 342600
    },
    {
      "epoch": 133.05,
      "learning_rate": 2.2596763754045308e-06,
      "loss": 0.0667,
      "step": 342610
    },
    {
      "epoch": 133.06,
      "learning_rate": 2.25915857605178e-06,
      "loss": 0.061,
      "step": 342620
    },
    {
      "epoch": 133.06,
      "learning_rate": 2.258640776699029e-06,
      "loss": 0.0139,
      "step": 342630
    },
    {
      "epoch": 133.06,
      "learning_rate": 2.2581229773462783e-06,
      "loss": 0.0233,
      "step": 342640
    },
    {
      "epoch": 133.07,
      "learning_rate": 2.2576051779935275e-06,
      "loss": 0.062,
      "step": 342650
    },
    {
      "epoch": 133.07,
      "learning_rate": 2.2570873786407767e-06,
      "loss": 0.0049,
      "step": 342660
    },
    {
      "epoch": 133.08,
      "learning_rate": 2.256569579288026e-06,
      "loss": 0.0612,
      "step": 342670
    },
    {
      "epoch": 133.08,
      "learning_rate": 2.256051779935275e-06,
      "loss": 0.0785,
      "step": 342680
    },
    {
      "epoch": 133.08,
      "learning_rate": 2.2555339805825243e-06,
      "loss": 0.0529,
      "step": 342690
    },
    {
      "epoch": 133.09,
      "learning_rate": 2.2550161812297735e-06,
      "loss": 0.0183,
      "step": 342700
    },
    {
      "epoch": 133.09,
      "learning_rate": 2.2544983818770227e-06,
      "loss": 0.0563,
      "step": 342710
    },
    {
      "epoch": 133.1,
      "learning_rate": 2.253980582524272e-06,
      "loss": 0.0327,
      "step": 342720
    },
    {
      "epoch": 133.1,
      "learning_rate": 2.253462783171521e-06,
      "loss": 0.0003,
      "step": 342730
    },
    {
      "epoch": 133.1,
      "learning_rate": 2.2529449838187703e-06,
      "loss": 0.0172,
      "step": 342740
    },
    {
      "epoch": 133.11,
      "learning_rate": 2.2524271844660195e-06,
      "loss": 0.1251,
      "step": 342750
    },
    {
      "epoch": 133.11,
      "learning_rate": 2.2519093851132687e-06,
      "loss": 0.1382,
      "step": 342760
    },
    {
      "epoch": 133.11,
      "learning_rate": 2.251391585760518e-06,
      "loss": 0.1047,
      "step": 342770
    },
    {
      "epoch": 133.12,
      "learning_rate": 2.250873786407767e-06,
      "loss": 0.0169,
      "step": 342780
    },
    {
      "epoch": 133.12,
      "learning_rate": 2.2503559870550163e-06,
      "loss": 0.1114,
      "step": 342790
    },
    {
      "epoch": 133.13,
      "learning_rate": 2.2498381877022655e-06,
      "loss": 0.0519,
      "step": 342800
    },
    {
      "epoch": 133.13,
      "learning_rate": 2.249320388349515e-06,
      "loss": 0.0412,
      "step": 342810
    },
    {
      "epoch": 133.13,
      "learning_rate": 2.2488025889967643e-06,
      "loss": 0.1149,
      "step": 342820
    },
    {
      "epoch": 133.14,
      "learning_rate": 2.2482847896440135e-06,
      "loss": 0.0096,
      "step": 342830
    },
    {
      "epoch": 133.14,
      "learning_rate": 2.2477669902912622e-06,
      "loss": 0.061,
      "step": 342840
    },
    {
      "epoch": 133.15,
      "learning_rate": 2.2472491909385114e-06,
      "loss": 0.1242,
      "step": 342850
    },
    {
      "epoch": 133.15,
      "learning_rate": 2.2467313915857606e-06,
      "loss": 0.015,
      "step": 342860
    },
    {
      "epoch": 133.15,
      "learning_rate": 2.24621359223301e-06,
      "loss": 0.0006,
      "step": 342870
    },
    {
      "epoch": 133.16,
      "learning_rate": 2.245695792880259e-06,
      "loss": 0.0009,
      "step": 342880
    },
    {
      "epoch": 133.16,
      "learning_rate": 2.245177993527508e-06,
      "loss": 0.0706,
      "step": 342890
    },
    {
      "epoch": 133.17,
      "learning_rate": 2.2446601941747574e-06,
      "loss": 0.0005,
      "step": 342900
    },
    {
      "epoch": 133.17,
      "learning_rate": 2.2441423948220066e-06,
      "loss": 0.0264,
      "step": 342910
    },
    {
      "epoch": 133.17,
      "learning_rate": 2.243624595469256e-06,
      "loss": 0.0226,
      "step": 342920
    },
    {
      "epoch": 133.18,
      "learning_rate": 2.243106796116505e-06,
      "loss": 0.0252,
      "step": 342930
    },
    {
      "epoch": 133.18,
      "learning_rate": 2.242588996763754e-06,
      "loss": 0.0387,
      "step": 342940
    },
    {
      "epoch": 133.18,
      "learning_rate": 2.2420711974110034e-06,
      "loss": 0.1469,
      "step": 342950
    },
    {
      "epoch": 133.19,
      "learning_rate": 2.2415533980582526e-06,
      "loss": 0.0121,
      "step": 342960
    },
    {
      "epoch": 133.19,
      "learning_rate": 2.2410355987055018e-06,
      "loss": 0.0447,
      "step": 342970
    },
    {
      "epoch": 133.2,
      "learning_rate": 2.240517799352751e-06,
      "loss": 0.0383,
      "step": 342980
    },
    {
      "epoch": 133.2,
      "learning_rate": 2.24e-06,
      "loss": 0.0571,
      "step": 342990
    },
    {
      "epoch": 133.2,
      "learning_rate": 2.2394822006472493e-06,
      "loss": 0.0765,
      "step": 343000
    },
    {
      "epoch": 133.21,
      "learning_rate": 2.2389644012944985e-06,
      "loss": 0.0268,
      "step": 343010
    },
    {
      "epoch": 133.21,
      "learning_rate": 2.2384466019417477e-06,
      "loss": 0.0772,
      "step": 343020
    },
    {
      "epoch": 133.22,
      "learning_rate": 2.237928802588997e-06,
      "loss": 0.0008,
      "step": 343030
    },
    {
      "epoch": 133.22,
      "learning_rate": 2.237411003236246e-06,
      "loss": 0.0126,
      "step": 343040
    },
    {
      "epoch": 133.22,
      "learning_rate": 2.2368932038834953e-06,
      "loss": 0.0358,
      "step": 343050
    },
    {
      "epoch": 133.23,
      "learning_rate": 2.2363754045307445e-06,
      "loss": 0.0002,
      "step": 343060
    },
    {
      "epoch": 133.23,
      "learning_rate": 2.2358576051779937e-06,
      "loss": 0.0265,
      "step": 343070
    },
    {
      "epoch": 133.23,
      "learning_rate": 2.235339805825243e-06,
      "loss": 0.0189,
      "step": 343080
    },
    {
      "epoch": 133.24,
      "learning_rate": 2.234822006472492e-06,
      "loss": 0.0001,
      "step": 343090
    },
    {
      "epoch": 133.24,
      "learning_rate": 2.2343042071197413e-06,
      "loss": 0.0327,
      "step": 343100
    },
    {
      "epoch": 133.25,
      "learning_rate": 2.2337864077669905e-06,
      "loss": 0.1103,
      "step": 343110
    },
    {
      "epoch": 133.25,
      "learning_rate": 2.2332686084142397e-06,
      "loss": 0.0003,
      "step": 343120
    },
    {
      "epoch": 133.25,
      "learning_rate": 2.232750809061489e-06,
      "loss": 0.018,
      "step": 343130
    },
    {
      "epoch": 133.26,
      "learning_rate": 2.232233009708738e-06,
      "loss": 0.1129,
      "step": 343140
    },
    {
      "epoch": 133.26,
      "learning_rate": 2.2317152103559873e-06,
      "loss": 0.008,
      "step": 343150
    },
    {
      "epoch": 133.27,
      "learning_rate": 2.2311974110032365e-06,
      "loss": 0.0293,
      "step": 343160
    },
    {
      "epoch": 133.27,
      "learning_rate": 2.2306796116504857e-06,
      "loss": 0.0135,
      "step": 343170
    },
    {
      "epoch": 133.27,
      "learning_rate": 2.230161812297735e-06,
      "loss": 0.0001,
      "step": 343180
    },
    {
      "epoch": 133.28,
      "learning_rate": 2.229644012944984e-06,
      "loss": 0.1644,
      "step": 343190
    },
    {
      "epoch": 133.28,
      "learning_rate": 2.2291262135922332e-06,
      "loss": 0.0147,
      "step": 343200
    },
    {
      "epoch": 133.29,
      "learning_rate": 2.2286084142394824e-06,
      "loss": 0.0524,
      "step": 343210
    },
    {
      "epoch": 133.29,
      "learning_rate": 2.2280906148867316e-06,
      "loss": 0.0205,
      "step": 343220
    },
    {
      "epoch": 133.29,
      "learning_rate": 2.227572815533981e-06,
      "loss": 0.0053,
      "step": 343230
    },
    {
      "epoch": 133.3,
      "learning_rate": 2.22705501618123e-06,
      "loss": 0.0118,
      "step": 343240
    },
    {
      "epoch": 133.3,
      "learning_rate": 2.226537216828479e-06,
      "loss": 0.0732,
      "step": 343250
    },
    {
      "epoch": 133.3,
      "learning_rate": 2.2260194174757284e-06,
      "loss": 0.0483,
      "step": 343260
    },
    {
      "epoch": 133.31,
      "learning_rate": 2.2255016181229776e-06,
      "loss": 0.0208,
      "step": 343270
    },
    {
      "epoch": 133.31,
      "learning_rate": 2.224983818770227e-06,
      "loss": 0.0395,
      "step": 343280
    },
    {
      "epoch": 133.32,
      "learning_rate": 2.224466019417476e-06,
      "loss": 0.0051,
      "step": 343290
    },
    {
      "epoch": 133.32,
      "learning_rate": 2.223948220064725e-06,
      "loss": 0.042,
      "step": 343300
    },
    {
      "epoch": 133.32,
      "learning_rate": 2.2234304207119744e-06,
      "loss": 0.0035,
      "step": 343310
    },
    {
      "epoch": 133.33,
      "learning_rate": 2.2229126213592236e-06,
      "loss": 0.0688,
      "step": 343320
    },
    {
      "epoch": 133.33,
      "learning_rate": 2.2223948220064723e-06,
      "loss": 0.0383,
      "step": 343330
    },
    {
      "epoch": 133.34,
      "learning_rate": 2.2218770226537215e-06,
      "loss": 0.0009,
      "step": 343340
    },
    {
      "epoch": 133.34,
      "learning_rate": 2.2213592233009707e-06,
      "loss": 0.0399,
      "step": 343350
    },
    {
      "epoch": 133.34,
      "learning_rate": 2.22084142394822e-06,
      "loss": 0.0201,
      "step": 343360
    },
    {
      "epoch": 133.35,
      "learning_rate": 2.2203236245954695e-06,
      "loss": 0.0263,
      "step": 343370
    },
    {
      "epoch": 133.35,
      "learning_rate": 2.2198058252427187e-06,
      "loss": 0.0001,
      "step": 343380
    },
    {
      "epoch": 133.36,
      "learning_rate": 2.219288025889968e-06,
      "loss": 0.0818,
      "step": 343390
    },
    {
      "epoch": 133.36,
      "learning_rate": 2.218770226537217e-06,
      "loss": 0.092,
      "step": 343400
    },
    {
      "epoch": 133.36,
      "learning_rate": 2.2182524271844663e-06,
      "loss": 0.066,
      "step": 343410
    },
    {
      "epoch": 133.37,
      "learning_rate": 2.2177346278317155e-06,
      "loss": 0.0272,
      "step": 343420
    },
    {
      "epoch": 133.37,
      "learning_rate": 2.2172168284789647e-06,
      "loss": 0.0262,
      "step": 343430
    },
    {
      "epoch": 133.37,
      "learning_rate": 2.216699029126214e-06,
      "loss": 0.0089,
      "step": 343440
    },
    {
      "epoch": 133.38,
      "learning_rate": 2.216181229773463e-06,
      "loss": 0.0798,
      "step": 343450
    },
    {
      "epoch": 133.38,
      "learning_rate": 2.2156634304207123e-06,
      "loss": 0.0109,
      "step": 343460
    },
    {
      "epoch": 133.39,
      "learning_rate": 2.2151456310679615e-06,
      "loss": 0.0171,
      "step": 343470
    },
    {
      "epoch": 133.39,
      "learning_rate": 2.2146278317152107e-06,
      "loss": 0.0864,
      "step": 343480
    },
    {
      "epoch": 133.39,
      "learning_rate": 2.21411003236246e-06,
      "loss": 0.0095,
      "step": 343490
    },
    {
      "epoch": 133.4,
      "learning_rate": 2.213592233009709e-06,
      "loss": 0.0826,
      "step": 343500
    },
    {
      "epoch": 133.4,
      "learning_rate": 2.2130744336569583e-06,
      "loss": 0.0456,
      "step": 343510
    },
    {
      "epoch": 133.41,
      "learning_rate": 2.2125566343042075e-06,
      "loss": 0.0354,
      "step": 343520
    },
    {
      "epoch": 133.41,
      "learning_rate": 2.2120388349514567e-06,
      "loss": 0.1422,
      "step": 343530
    },
    {
      "epoch": 133.41,
      "learning_rate": 2.211521035598706e-06,
      "loss": 0.0278,
      "step": 343540
    },
    {
      "epoch": 133.42,
      "learning_rate": 2.211003236245955e-06,
      "loss": 0.0372,
      "step": 343550
    },
    {
      "epoch": 133.42,
      "learning_rate": 2.2104854368932042e-06,
      "loss": 0.0487,
      "step": 343560
    },
    {
      "epoch": 133.43,
      "learning_rate": 2.2099676375404534e-06,
      "loss": 0.0161,
      "step": 343570
    },
    {
      "epoch": 133.43,
      "learning_rate": 2.2094498381877026e-06,
      "loss": 0.2329,
      "step": 343580
    },
    {
      "epoch": 133.43,
      "learning_rate": 2.208932038834952e-06,
      "loss": 0.0107,
      "step": 343590
    },
    {
      "epoch": 133.44,
      "learning_rate": 2.208414239482201e-06,
      "loss": 0.0172,
      "step": 343600
    },
    {
      "epoch": 133.44,
      "learning_rate": 2.2078964401294502e-06,
      "loss": 0.0088,
      "step": 343610
    },
    {
      "epoch": 133.44,
      "learning_rate": 2.2073786407766994e-06,
      "loss": 0.0018,
      "step": 343620
    },
    {
      "epoch": 133.45,
      "learning_rate": 2.2068608414239486e-06,
      "loss": 0.1306,
      "step": 343630
    },
    {
      "epoch": 133.45,
      "learning_rate": 2.206343042071198e-06,
      "loss": 0.0721,
      "step": 343640
    },
    {
      "epoch": 133.46,
      "learning_rate": 2.205825242718447e-06,
      "loss": 0.0191,
      "step": 343650
    },
    {
      "epoch": 133.46,
      "learning_rate": 2.2053074433656958e-06,
      "loss": 0.0717,
      "step": 343660
    },
    {
      "epoch": 133.46,
      "learning_rate": 2.204789644012945e-06,
      "loss": 0.1604,
      "step": 343670
    },
    {
      "epoch": 133.47,
      "learning_rate": 2.204271844660194e-06,
      "loss": 0.1256,
      "step": 343680
    },
    {
      "epoch": 133.47,
      "learning_rate": 2.2037540453074433e-06,
      "loss": 0.035,
      "step": 343690
    },
    {
      "epoch": 133.48,
      "learning_rate": 2.2032362459546925e-06,
      "loss": 0.0129,
      "step": 343700
    },
    {
      "epoch": 133.48,
      "learning_rate": 2.2027184466019417e-06,
      "loss": 0.0441,
      "step": 343710
    },
    {
      "epoch": 133.48,
      "learning_rate": 2.202200647249191e-06,
      "loss": 0.0485,
      "step": 343720
    },
    {
      "epoch": 133.49,
      "learning_rate": 2.20168284789644e-06,
      "loss": 0.003,
      "step": 343730
    },
    {
      "epoch": 133.49,
      "learning_rate": 2.2011650485436893e-06,
      "loss": 0.059,
      "step": 343740
    },
    {
      "epoch": 133.5,
      "learning_rate": 2.2006472491909385e-06,
      "loss": 0.0404,
      "step": 343750
    },
    {
      "epoch": 133.5,
      "learning_rate": 2.2001294498381877e-06,
      "loss": 0.0699,
      "step": 343760
    },
    {
      "epoch": 133.5,
      "learning_rate": 2.199611650485437e-06,
      "loss": 0.0094,
      "step": 343770
    },
    {
      "epoch": 133.51,
      "learning_rate": 2.199093851132686e-06,
      "loss": 0.0039,
      "step": 343780
    },
    {
      "epoch": 133.51,
      "learning_rate": 2.1985760517799353e-06,
      "loss": 0.0911,
      "step": 343790
    },
    {
      "epoch": 133.51,
      "learning_rate": 2.1980582524271845e-06,
      "loss": 0.033,
      "step": 343800
    },
    {
      "epoch": 133.52,
      "learning_rate": 2.1975404530744337e-06,
      "loss": 0.1559,
      "step": 343810
    },
    {
      "epoch": 133.52,
      "learning_rate": 2.197022653721683e-06,
      "loss": 0.0193,
      "step": 343820
    },
    {
      "epoch": 133.53,
      "learning_rate": 2.196504854368932e-06,
      "loss": 0.0001,
      "step": 343830
    },
    {
      "epoch": 133.53,
      "learning_rate": 2.1959870550161813e-06,
      "loss": 0.0154,
      "step": 343840
    },
    {
      "epoch": 133.53,
      "learning_rate": 2.1954692556634305e-06,
      "loss": 0.0005,
      "step": 343850
    },
    {
      "epoch": 133.54,
      "learning_rate": 2.1949514563106797e-06,
      "loss": 0.0017,
      "step": 343860
    },
    {
      "epoch": 133.54,
      "learning_rate": 2.194433656957929e-06,
      "loss": 0.036,
      "step": 343870
    },
    {
      "epoch": 133.55,
      "learning_rate": 2.193915857605178e-06,
      "loss": 0.0451,
      "step": 343880
    },
    {
      "epoch": 133.55,
      "learning_rate": 2.1933980582524272e-06,
      "loss": 0.0348,
      "step": 343890
    },
    {
      "epoch": 133.55,
      "learning_rate": 2.1928802588996764e-06,
      "loss": 0.0675,
      "step": 343900
    },
    {
      "epoch": 133.56,
      "learning_rate": 2.1923624595469256e-06,
      "loss": 0.0004,
      "step": 343910
    },
    {
      "epoch": 133.56,
      "learning_rate": 2.191844660194175e-06,
      "loss": 0.0343,
      "step": 343920
    },
    {
      "epoch": 133.57,
      "learning_rate": 2.1913268608414244e-06,
      "loss": 0.0074,
      "step": 343930
    },
    {
      "epoch": 133.57,
      "learning_rate": 2.1908090614886736e-06,
      "loss": 0.1216,
      "step": 343940
    },
    {
      "epoch": 133.57,
      "learning_rate": 2.190291262135923e-06,
      "loss": 0.0001,
      "step": 343950
    },
    {
      "epoch": 133.58,
      "learning_rate": 2.189773462783172e-06,
      "loss": 0.0086,
      "step": 343960
    },
    {
      "epoch": 133.58,
      "learning_rate": 2.1892556634304212e-06,
      "loss": 0.0495,
      "step": 343970
    },
    {
      "epoch": 133.58,
      "learning_rate": 2.18873786407767e-06,
      "loss": 0.0014,
      "step": 343980
    },
    {
      "epoch": 133.59,
      "learning_rate": 2.188220064724919e-06,
      "loss": 0.0153,
      "step": 343990
    },
    {
      "epoch": 133.59,
      "learning_rate": 2.1877022653721684e-06,
      "loss": 0.0292,
      "step": 344000
    },
    {
      "epoch": 133.6,
      "learning_rate": 2.1871844660194176e-06,
      "loss": 0.0409,
      "step": 344010
    },
    {
      "epoch": 133.6,
      "learning_rate": 2.1866666666666668e-06,
      "loss": 0.0459,
      "step": 344020
    },
    {
      "epoch": 133.6,
      "learning_rate": 2.186148867313916e-06,
      "loss": 0.0671,
      "step": 344030
    },
    {
      "epoch": 133.61,
      "learning_rate": 2.185631067961165e-06,
      "loss": 0.0057,
      "step": 344040
    },
    {
      "epoch": 133.61,
      "learning_rate": 2.1851132686084143e-06,
      "loss": 0.0252,
      "step": 344050
    },
    {
      "epoch": 133.62,
      "learning_rate": 2.1845954692556635e-06,
      "loss": 0.0001,
      "step": 344060
    },
    {
      "epoch": 133.62,
      "learning_rate": 2.1840776699029127e-06,
      "loss": 0.0795,
      "step": 344070
    },
    {
      "epoch": 133.62,
      "learning_rate": 2.183559870550162e-06,
      "loss": 0.1801,
      "step": 344080
    },
    {
      "epoch": 133.63,
      "learning_rate": 2.183042071197411e-06,
      "loss": 0.0179,
      "step": 344090
    },
    {
      "epoch": 133.63,
      "learning_rate": 2.1825242718446603e-06,
      "loss": 0.0997,
      "step": 344100
    },
    {
      "epoch": 133.63,
      "learning_rate": 2.1820064724919095e-06,
      "loss": 0.0268,
      "step": 344110
    },
    {
      "epoch": 133.64,
      "learning_rate": 2.1814886731391587e-06,
      "loss": 0.0799,
      "step": 344120
    },
    {
      "epoch": 133.64,
      "learning_rate": 2.180970873786408e-06,
      "loss": 0.0697,
      "step": 344130
    },
    {
      "epoch": 133.65,
      "learning_rate": 2.180453074433657e-06,
      "loss": 0.0303,
      "step": 344140
    },
    {
      "epoch": 133.65,
      "learning_rate": 2.1799352750809063e-06,
      "loss": 0.0664,
      "step": 344150
    },
    {
      "epoch": 133.65,
      "learning_rate": 2.1794174757281555e-06,
      "loss": 0.0254,
      "step": 344160
    },
    {
      "epoch": 133.66,
      "learning_rate": 2.1788996763754047e-06,
      "loss": 0.1343,
      "step": 344170
    },
    {
      "epoch": 133.66,
      "learning_rate": 2.178381877022654e-06,
      "loss": 0.0141,
      "step": 344180
    },
    {
      "epoch": 133.67,
      "learning_rate": 2.177864077669903e-06,
      "loss": 0.0033,
      "step": 344190
    },
    {
      "epoch": 133.67,
      "learning_rate": 2.1773462783171523e-06,
      "loss": 0.0647,
      "step": 344200
    },
    {
      "epoch": 133.67,
      "learning_rate": 2.1768284789644015e-06,
      "loss": 0.0756,
      "step": 344210
    },
    {
      "epoch": 133.68,
      "learning_rate": 2.1763106796116507e-06,
      "loss": 0.0087,
      "step": 344220
    },
    {
      "epoch": 133.68,
      "learning_rate": 2.1757928802589e-06,
      "loss": 0.0508,
      "step": 344230
    },
    {
      "epoch": 133.69,
      "learning_rate": 2.175275080906149e-06,
      "loss": 0.0316,
      "step": 344240
    },
    {
      "epoch": 133.69,
      "learning_rate": 2.1747572815533982e-06,
      "loss": 0.0666,
      "step": 344250
    },
    {
      "epoch": 133.69,
      "learning_rate": 2.1742394822006474e-06,
      "loss": 0.0282,
      "step": 344260
    },
    {
      "epoch": 133.7,
      "learning_rate": 2.1737216828478966e-06,
      "loss": 0.0004,
      "step": 344270
    },
    {
      "epoch": 133.7,
      "learning_rate": 2.173203883495146e-06,
      "loss": 0.0186,
      "step": 344280
    },
    {
      "epoch": 133.7,
      "learning_rate": 2.172686084142395e-06,
      "loss": 0.0006,
      "step": 344290
    },
    {
      "epoch": 133.71,
      "learning_rate": 2.1721682847896442e-06,
      "loss": 0.0757,
      "step": 344300
    },
    {
      "epoch": 133.71,
      "learning_rate": 2.1716504854368934e-06,
      "loss": 0.0004,
      "step": 344310
    },
    {
      "epoch": 133.72,
      "learning_rate": 2.1711326860841426e-06,
      "loss": 0.0249,
      "step": 344320
    },
    {
      "epoch": 133.72,
      "learning_rate": 2.170614886731392e-06,
      "loss": 0.0001,
      "step": 344330
    },
    {
      "epoch": 133.72,
      "learning_rate": 2.170097087378641e-06,
      "loss": 0.0833,
      "step": 344340
    },
    {
      "epoch": 133.73,
      "learning_rate": 2.16957928802589e-06,
      "loss": 0.0127,
      "step": 344350
    },
    {
      "epoch": 133.73,
      "learning_rate": 2.1690614886731394e-06,
      "loss": 0.0167,
      "step": 344360
    },
    {
      "epoch": 133.74,
      "learning_rate": 2.1685436893203886e-06,
      "loss": 0.0086,
      "step": 344370
    },
    {
      "epoch": 133.74,
      "learning_rate": 2.1680258899676378e-06,
      "loss": 0.0008,
      "step": 344380
    },
    {
      "epoch": 133.74,
      "learning_rate": 2.167508090614887e-06,
      "loss": 0.0369,
      "step": 344390
    },
    {
      "epoch": 133.75,
      "learning_rate": 2.166990291262136e-06,
      "loss": 0.0772,
      "step": 344400
    },
    {
      "epoch": 133.75,
      "learning_rate": 2.1664724919093854e-06,
      "loss": 0.0686,
      "step": 344410
    },
    {
      "epoch": 133.76,
      "learning_rate": 2.1659546925566345e-06,
      "loss": 0.0624,
      "step": 344420
    },
    {
      "epoch": 133.76,
      "learning_rate": 2.1654368932038837e-06,
      "loss": 0.0072,
      "step": 344430
    },
    {
      "epoch": 133.76,
      "learning_rate": 2.164919093851133e-06,
      "loss": 0.0147,
      "step": 344440
    },
    {
      "epoch": 133.77,
      "learning_rate": 2.164401294498382e-06,
      "loss": 0.0555,
      "step": 344450
    },
    {
      "epoch": 133.77,
      "learning_rate": 2.1638834951456313e-06,
      "loss": 0.0009,
      "step": 344460
    },
    {
      "epoch": 133.77,
      "learning_rate": 2.16336569579288e-06,
      "loss": 0.0273,
      "step": 344470
    },
    {
      "epoch": 133.78,
      "learning_rate": 2.1628478964401293e-06,
      "loss": 0.0382,
      "step": 344480
    },
    {
      "epoch": 133.78,
      "learning_rate": 2.162330097087379e-06,
      "loss": 0.0438,
      "step": 344490
    },
    {
      "epoch": 133.79,
      "learning_rate": 2.161812297734628e-06,
      "loss": 0.0366,
      "step": 344500
    },
    {
      "epoch": 133.79,
      "learning_rate": 2.1612944983818773e-06,
      "loss": 0.0703,
      "step": 344510
    },
    {
      "epoch": 133.79,
      "learning_rate": 2.1607766990291265e-06,
      "loss": 0.0253,
      "step": 344520
    },
    {
      "epoch": 133.8,
      "learning_rate": 2.1602588996763757e-06,
      "loss": 0.0643,
      "step": 344530
    },
    {
      "epoch": 133.8,
      "learning_rate": 2.159741100323625e-06,
      "loss": 0.0909,
      "step": 344540
    },
    {
      "epoch": 133.81,
      "learning_rate": 2.159223300970874e-06,
      "loss": 0.0468,
      "step": 344550
    },
    {
      "epoch": 133.81,
      "learning_rate": 2.1587055016181233e-06,
      "loss": 0.118,
      "step": 344560
    },
    {
      "epoch": 133.81,
      "learning_rate": 2.1581877022653725e-06,
      "loss": 0.0001,
      "step": 344570
    },
    {
      "epoch": 133.82,
      "learning_rate": 2.1576699029126217e-06,
      "loss": 0.0004,
      "step": 344580
    },
    {
      "epoch": 133.82,
      "learning_rate": 2.157152103559871e-06,
      "loss": 0.0005,
      "step": 344590
    },
    {
      "epoch": 133.83,
      "learning_rate": 2.15663430420712e-06,
      "loss": 0.0941,
      "step": 344600
    },
    {
      "epoch": 133.83,
      "learning_rate": 2.1561165048543692e-06,
      "loss": 0.0844,
      "step": 344610
    },
    {
      "epoch": 133.83,
      "learning_rate": 2.1555987055016184e-06,
      "loss": 0.0857,
      "step": 344620
    },
    {
      "epoch": 133.84,
      "learning_rate": 2.1550809061488676e-06,
      "loss": 0.051,
      "step": 344630
    },
    {
      "epoch": 133.84,
      "learning_rate": 2.154563106796117e-06,
      "loss": 0.0304,
      "step": 344640
    },
    {
      "epoch": 133.84,
      "learning_rate": 2.154045307443366e-06,
      "loss": 0.0001,
      "step": 344650
    },
    {
      "epoch": 133.85,
      "learning_rate": 2.1535275080906152e-06,
      "loss": 0.036,
      "step": 344660
    },
    {
      "epoch": 133.85,
      "learning_rate": 2.1530097087378644e-06,
      "loss": 0.0616,
      "step": 344670
    },
    {
      "epoch": 133.86,
      "learning_rate": 2.1524919093851136e-06,
      "loss": 0.0126,
      "step": 344680
    },
    {
      "epoch": 133.86,
      "learning_rate": 2.151974110032363e-06,
      "loss": 0.0887,
      "step": 344690
    },
    {
      "epoch": 133.86,
      "learning_rate": 2.151456310679612e-06,
      "loss": 0.0332,
      "step": 344700
    },
    {
      "epoch": 133.87,
      "learning_rate": 2.150938511326861e-06,
      "loss": 0.0006,
      "step": 344710
    },
    {
      "epoch": 133.87,
      "learning_rate": 2.1504207119741104e-06,
      "loss": 0.0001,
      "step": 344720
    },
    {
      "epoch": 133.88,
      "learning_rate": 2.1499029126213596e-06,
      "loss": 0.0279,
      "step": 344730
    },
    {
      "epoch": 133.88,
      "learning_rate": 2.1493851132686088e-06,
      "loss": 0.05,
      "step": 344740
    },
    {
      "epoch": 133.88,
      "learning_rate": 2.148867313915858e-06,
      "loss": 0.1245,
      "step": 344750
    },
    {
      "epoch": 133.89,
      "learning_rate": 2.148349514563107e-06,
      "loss": 0.0454,
      "step": 344760
    },
    {
      "epoch": 133.89,
      "learning_rate": 2.1478317152103564e-06,
      "loss": 0.2008,
      "step": 344770
    },
    {
      "epoch": 133.9,
      "learning_rate": 2.1473139158576055e-06,
      "loss": 0.0164,
      "step": 344780
    },
    {
      "epoch": 133.9,
      "learning_rate": 2.1467961165048547e-06,
      "loss": 0.0645,
      "step": 344790
    },
    {
      "epoch": 133.9,
      "learning_rate": 2.1462783171521035e-06,
      "loss": 0.0128,
      "step": 344800
    },
    {
      "epoch": 133.91,
      "learning_rate": 2.1457605177993527e-06,
      "loss": 0.0311,
      "step": 344810
    },
    {
      "epoch": 133.91,
      "learning_rate": 2.145242718446602e-06,
      "loss": 0.0745,
      "step": 344820
    },
    {
      "epoch": 133.91,
      "learning_rate": 2.144724919093851e-06,
      "loss": 0.0093,
      "step": 344830
    },
    {
      "epoch": 133.92,
      "learning_rate": 2.1442071197411003e-06,
      "loss": 0.0634,
      "step": 344840
    },
    {
      "epoch": 133.92,
      "learning_rate": 2.1436893203883495e-06,
      "loss": 0.1034,
      "step": 344850
    },
    {
      "epoch": 133.93,
      "learning_rate": 2.1431715210355987e-06,
      "loss": 0.1079,
      "step": 344860
    },
    {
      "epoch": 133.93,
      "learning_rate": 2.142653721682848e-06,
      "loss": 0.0143,
      "step": 344870
    },
    {
      "epoch": 133.93,
      "learning_rate": 2.142135922330097e-06,
      "loss": 0.0176,
      "step": 344880
    },
    {
      "epoch": 133.94,
      "learning_rate": 2.1416181229773463e-06,
      "loss": 0.0005,
      "step": 344890
    },
    {
      "epoch": 133.94,
      "learning_rate": 2.1411003236245955e-06,
      "loss": 0.0286,
      "step": 344900
    },
    {
      "epoch": 133.95,
      "learning_rate": 2.1405825242718447e-06,
      "loss": 0.0662,
      "step": 344910
    },
    {
      "epoch": 133.95,
      "learning_rate": 2.140064724919094e-06,
      "loss": 0.0322,
      "step": 344920
    },
    {
      "epoch": 133.95,
      "learning_rate": 2.139546925566343e-06,
      "loss": 0.0759,
      "step": 344930
    },
    {
      "epoch": 133.96,
      "learning_rate": 2.1390291262135922e-06,
      "loss": 0.0002,
      "step": 344940
    },
    {
      "epoch": 133.96,
      "learning_rate": 2.1385113268608414e-06,
      "loss": 0.0484,
      "step": 344950
    },
    {
      "epoch": 133.97,
      "learning_rate": 2.1379935275080906e-06,
      "loss": 0.0027,
      "step": 344960
    },
    {
      "epoch": 133.97,
      "learning_rate": 2.13747572815534e-06,
      "loss": 0.0434,
      "step": 344970
    },
    {
      "epoch": 133.97,
      "learning_rate": 2.136957928802589e-06,
      "loss": 0.0495,
      "step": 344980
    },
    {
      "epoch": 133.98,
      "learning_rate": 2.136440129449838e-06,
      "loss": 0.0251,
      "step": 344990
    },
    {
      "epoch": 133.98,
      "learning_rate": 2.1359223300970874e-06,
      "loss": 0.0093,
      "step": 345000
    },
    {
      "epoch": 133.98,
      "learning_rate": 2.1354045307443366e-06,
      "loss": 0.0534,
      "step": 345010
    },
    {
      "epoch": 133.99,
      "learning_rate": 2.134886731391586e-06,
      "loss": 0.0067,
      "step": 345020
    },
    {
      "epoch": 133.99,
      "learning_rate": 2.134368932038835e-06,
      "loss": 0.0077,
      "step": 345030
    },
    {
      "epoch": 134.0,
      "learning_rate": 2.133851132686084e-06,
      "loss": 0.0521,
      "step": 345040
    },
    {
      "epoch": 134.0,
      "learning_rate": 2.133333333333334e-06,
      "loss": 0.0784,
      "step": 345050
    },
    {
      "epoch": 134.0,
      "eval_accuracy": 0.953232462173315,
      "eval_loss": 0.38705921173095703,
      "eval_runtime": 8.2715,
      "eval_samples_per_second": 439.46,
      "eval_steps_per_second": 55.008,
      "step": 345050
    },
    {
      "epoch": 134.0,
      "learning_rate": 2.132815533980583e-06,
      "loss": 0.001,
      "step": 345060
    },
    {
      "epoch": 134.01,
      "learning_rate": 2.132297734627832e-06,
      "loss": 0.0105,
      "step": 345070
    },
    {
      "epoch": 134.01,
      "learning_rate": 2.1317799352750814e-06,
      "loss": 0.0093,
      "step": 345080
    },
    {
      "epoch": 134.02,
      "learning_rate": 2.1312621359223306e-06,
      "loss": 0.0189,
      "step": 345090
    },
    {
      "epoch": 134.02,
      "learning_rate": 2.1307443365695798e-06,
      "loss": 0.0509,
      "step": 345100
    },
    {
      "epoch": 134.02,
      "learning_rate": 2.130226537216829e-06,
      "loss": 0.0418,
      "step": 345110
    },
    {
      "epoch": 134.03,
      "learning_rate": 2.1297087378640777e-06,
      "loss": 0.0662,
      "step": 345120
    },
    {
      "epoch": 134.03,
      "learning_rate": 2.129190938511327e-06,
      "loss": 0.0294,
      "step": 345130
    },
    {
      "epoch": 134.03,
      "learning_rate": 2.128673139158576e-06,
      "loss": 0.0208,
      "step": 345140
    },
    {
      "epoch": 134.04,
      "learning_rate": 2.1281553398058253e-06,
      "loss": 0.0701,
      "step": 345150
    },
    {
      "epoch": 134.04,
      "learning_rate": 2.1276375404530745e-06,
      "loss": 0.0538,
      "step": 345160
    },
    {
      "epoch": 134.05,
      "learning_rate": 2.1271197411003237e-06,
      "loss": 0.0096,
      "step": 345170
    },
    {
      "epoch": 134.05,
      "learning_rate": 2.126601941747573e-06,
      "loss": 0.1019,
      "step": 345180
    },
    {
      "epoch": 134.05,
      "learning_rate": 2.126084142394822e-06,
      "loss": 0.0009,
      "step": 345190
    },
    {
      "epoch": 134.06,
      "learning_rate": 2.1255663430420713e-06,
      "loss": 0.0252,
      "step": 345200
    },
    {
      "epoch": 134.06,
      "learning_rate": 2.1250485436893205e-06,
      "loss": 0.0336,
      "step": 345210
    },
    {
      "epoch": 134.07,
      "learning_rate": 2.1245307443365697e-06,
      "loss": 0.0021,
      "step": 345220
    },
    {
      "epoch": 134.07,
      "learning_rate": 2.124012944983819e-06,
      "loss": 0.0965,
      "step": 345230
    },
    {
      "epoch": 134.07,
      "learning_rate": 2.123495145631068e-06,
      "loss": 0.0106,
      "step": 345240
    },
    {
      "epoch": 134.08,
      "learning_rate": 2.1229773462783173e-06,
      "loss": 0.0348,
      "step": 345250
    },
    {
      "epoch": 134.08,
      "learning_rate": 2.1224595469255665e-06,
      "loss": 0.0821,
      "step": 345260
    },
    {
      "epoch": 134.09,
      "learning_rate": 2.1219417475728157e-06,
      "loss": 0.051,
      "step": 345270
    },
    {
      "epoch": 134.09,
      "learning_rate": 2.121423948220065e-06,
      "loss": 0.0276,
      "step": 345280
    },
    {
      "epoch": 134.09,
      "learning_rate": 2.120906148867314e-06,
      "loss": 0.0128,
      "step": 345290
    },
    {
      "epoch": 134.1,
      "learning_rate": 2.1203883495145632e-06,
      "loss": 0.0332,
      "step": 345300
    },
    {
      "epoch": 134.1,
      "learning_rate": 2.1198705501618124e-06,
      "loss": 0.1638,
      "step": 345310
    },
    {
      "epoch": 134.1,
      "learning_rate": 2.1193527508090616e-06,
      "loss": 0.2331,
      "step": 345320
    },
    {
      "epoch": 134.11,
      "learning_rate": 2.118834951456311e-06,
      "loss": 0.0499,
      "step": 345330
    },
    {
      "epoch": 134.11,
      "learning_rate": 2.11831715210356e-06,
      "loss": 0.0654,
      "step": 345340
    },
    {
      "epoch": 134.12,
      "learning_rate": 2.1177993527508092e-06,
      "loss": 0.0008,
      "step": 345350
    },
    {
      "epoch": 134.12,
      "learning_rate": 2.1172815533980584e-06,
      "loss": 0.0317,
      "step": 345360
    },
    {
      "epoch": 134.12,
      "learning_rate": 2.1167637540453076e-06,
      "loss": 0.1292,
      "step": 345370
    },
    {
      "epoch": 134.13,
      "learning_rate": 2.116245954692557e-06,
      "loss": 0.0019,
      "step": 345380
    },
    {
      "epoch": 134.13,
      "learning_rate": 2.115728155339806e-06,
      "loss": 0.0325,
      "step": 345390
    },
    {
      "epoch": 134.14,
      "learning_rate": 2.115210355987055e-06,
      "loss": 0.0933,
      "step": 345400
    },
    {
      "epoch": 134.14,
      "learning_rate": 2.1146925566343044e-06,
      "loss": 0.0025,
      "step": 345410
    },
    {
      "epoch": 134.14,
      "learning_rate": 2.1141747572815536e-06,
      "loss": 0.0102,
      "step": 345420
    },
    {
      "epoch": 134.15,
      "learning_rate": 2.1136569579288028e-06,
      "loss": 0.0241,
      "step": 345430
    },
    {
      "epoch": 134.15,
      "learning_rate": 2.113139158576052e-06,
      "loss": 0.0534,
      "step": 345440
    },
    {
      "epoch": 134.16,
      "learning_rate": 2.112621359223301e-06,
      "loss": 0.077,
      "step": 345450
    },
    {
      "epoch": 134.16,
      "learning_rate": 2.1121035598705504e-06,
      "loss": 0.0001,
      "step": 345460
    },
    {
      "epoch": 134.16,
      "learning_rate": 2.1115857605177995e-06,
      "loss": 0.0252,
      "step": 345470
    },
    {
      "epoch": 134.17,
      "learning_rate": 2.1110679611650487e-06,
      "loss": 0.0899,
      "step": 345480
    },
    {
      "epoch": 134.17,
      "learning_rate": 2.110550161812298e-06,
      "loss": 0.0904,
      "step": 345490
    },
    {
      "epoch": 134.17,
      "learning_rate": 2.110032362459547e-06,
      "loss": 0.0118,
      "step": 345500
    },
    {
      "epoch": 134.18,
      "learning_rate": 2.1095145631067963e-06,
      "loss": 0.0404,
      "step": 345510
    },
    {
      "epoch": 134.18,
      "learning_rate": 2.1089967637540455e-06,
      "loss": 0.1165,
      "step": 345520
    },
    {
      "epoch": 134.19,
      "learning_rate": 2.1084789644012947e-06,
      "loss": 0.0133,
      "step": 345530
    },
    {
      "epoch": 134.19,
      "learning_rate": 2.107961165048544e-06,
      "loss": 0.0022,
      "step": 345540
    },
    {
      "epoch": 134.19,
      "learning_rate": 2.107443365695793e-06,
      "loss": 0.0086,
      "step": 345550
    },
    {
      "epoch": 134.2,
      "learning_rate": 2.1069255663430423e-06,
      "loss": 0.0248,
      "step": 345560
    },
    {
      "epoch": 134.2,
      "learning_rate": 2.1064077669902915e-06,
      "loss": 0.0168,
      "step": 345570
    },
    {
      "epoch": 134.21,
      "learning_rate": 2.1058899676375407e-06,
      "loss": 0.0356,
      "step": 345580
    },
    {
      "epoch": 134.21,
      "learning_rate": 2.10537216828479e-06,
      "loss": 0.0141,
      "step": 345590
    },
    {
      "epoch": 134.21,
      "learning_rate": 2.104854368932039e-06,
      "loss": 0.0156,
      "step": 345600
    },
    {
      "epoch": 134.22,
      "learning_rate": 2.1043365695792883e-06,
      "loss": 0.0035,
      "step": 345610
    },
    {
      "epoch": 134.22,
      "learning_rate": 2.1038187702265375e-06,
      "loss": 0.0551,
      "step": 345620
    },
    {
      "epoch": 134.23,
      "learning_rate": 2.1033009708737867e-06,
      "loss": 0.042,
      "step": 345630
    },
    {
      "epoch": 134.23,
      "learning_rate": 2.102783171521036e-06,
      "loss": 0.0569,
      "step": 345640
    },
    {
      "epoch": 134.23,
      "learning_rate": 2.102265372168285e-06,
      "loss": 0.0001,
      "step": 345650
    },
    {
      "epoch": 134.24,
      "learning_rate": 2.1017475728155342e-06,
      "loss": 0.0003,
      "step": 345660
    },
    {
      "epoch": 134.24,
      "learning_rate": 2.1012297734627834e-06,
      "loss": 0.0245,
      "step": 345670
    },
    {
      "epoch": 134.24,
      "learning_rate": 2.1007119741100326e-06,
      "loss": 0.0406,
      "step": 345680
    },
    {
      "epoch": 134.25,
      "learning_rate": 2.100194174757282e-06,
      "loss": 0.0006,
      "step": 345690
    },
    {
      "epoch": 134.25,
      "learning_rate": 2.099676375404531e-06,
      "loss": 0.0567,
      "step": 345700
    },
    {
      "epoch": 134.26,
      "learning_rate": 2.0991585760517802e-06,
      "loss": 0.0255,
      "step": 345710
    },
    {
      "epoch": 134.26,
      "learning_rate": 2.0986407766990294e-06,
      "loss": 0.0724,
      "step": 345720
    },
    {
      "epoch": 134.26,
      "learning_rate": 2.0981229773462786e-06,
      "loss": 0.0002,
      "step": 345730
    },
    {
      "epoch": 134.27,
      "learning_rate": 2.097605177993528e-06,
      "loss": 0.0923,
      "step": 345740
    },
    {
      "epoch": 134.27,
      "learning_rate": 2.097087378640777e-06,
      "loss": 0.0777,
      "step": 345750
    },
    {
      "epoch": 134.28,
      "learning_rate": 2.096569579288026e-06,
      "loss": 0.1019,
      "step": 345760
    },
    {
      "epoch": 134.28,
      "learning_rate": 2.0960517799352754e-06,
      "loss": 0.0205,
      "step": 345770
    },
    {
      "epoch": 134.28,
      "learning_rate": 2.0955339805825246e-06,
      "loss": 0.0335,
      "step": 345780
    },
    {
      "epoch": 134.29,
      "learning_rate": 2.0950161812297738e-06,
      "loss": 0.0237,
      "step": 345790
    },
    {
      "epoch": 134.29,
      "learning_rate": 2.094498381877023e-06,
      "loss": 0.0886,
      "step": 345800
    },
    {
      "epoch": 134.3,
      "learning_rate": 2.093980582524272e-06,
      "loss": 0.0408,
      "step": 345810
    },
    {
      "epoch": 134.3,
      "learning_rate": 2.0934627831715214e-06,
      "loss": 0.0092,
      "step": 345820
    },
    {
      "epoch": 134.3,
      "learning_rate": 2.0929449838187706e-06,
      "loss": 0.0127,
      "step": 345830
    },
    {
      "epoch": 134.31,
      "learning_rate": 2.0924271844660197e-06,
      "loss": 0.038,
      "step": 345840
    },
    {
      "epoch": 134.31,
      "learning_rate": 2.091909385113269e-06,
      "loss": 0.0006,
      "step": 345850
    },
    {
      "epoch": 134.31,
      "learning_rate": 2.091391585760518e-06,
      "loss": 0.1537,
      "step": 345860
    },
    {
      "epoch": 134.32,
      "learning_rate": 2.0908737864077673e-06,
      "loss": 0.0533,
      "step": 345870
    },
    {
      "epoch": 134.32,
      "learning_rate": 2.0903559870550165e-06,
      "loss": 0.0007,
      "step": 345880
    },
    {
      "epoch": 134.33,
      "learning_rate": 2.0898381877022657e-06,
      "loss": 0.0141,
      "step": 345890
    },
    {
      "epoch": 134.33,
      "learning_rate": 2.089320388349515e-06,
      "loss": 0.0001,
      "step": 345900
    },
    {
      "epoch": 134.33,
      "learning_rate": 2.088802588996764e-06,
      "loss": 0.0001,
      "step": 345910
    },
    {
      "epoch": 134.34,
      "learning_rate": 2.0882847896440133e-06,
      "loss": 0.0398,
      "step": 345920
    },
    {
      "epoch": 134.34,
      "learning_rate": 2.0877669902912625e-06,
      "loss": 0.0384,
      "step": 345930
    },
    {
      "epoch": 134.35,
      "learning_rate": 2.0872491909385113e-06,
      "loss": 0.0002,
      "step": 345940
    },
    {
      "epoch": 134.35,
      "learning_rate": 2.0867313915857605e-06,
      "loss": 0.0435,
      "step": 345950
    },
    {
      "epoch": 134.35,
      "learning_rate": 2.0862135922330097e-06,
      "loss": 0.0669,
      "step": 345960
    },
    {
      "epoch": 134.36,
      "learning_rate": 2.085695792880259e-06,
      "loss": 0.0424,
      "step": 345970
    },
    {
      "epoch": 134.36,
      "learning_rate": 2.085177993527508e-06,
      "loss": 0.0533,
      "step": 345980
    },
    {
      "epoch": 134.37,
      "learning_rate": 2.0846601941747572e-06,
      "loss": 0.0212,
      "step": 345990
    },
    {
      "epoch": 134.37,
      "learning_rate": 2.0841423948220064e-06,
      "loss": 0.0741,
      "step": 346000
    },
    {
      "epoch": 134.37,
      "learning_rate": 2.0836245954692556e-06,
      "loss": 0.0094,
      "step": 346010
    },
    {
      "epoch": 134.38,
      "learning_rate": 2.083106796116505e-06,
      "loss": 0.0493,
      "step": 346020
    },
    {
      "epoch": 134.38,
      "learning_rate": 2.082588996763754e-06,
      "loss": 0.0003,
      "step": 346030
    },
    {
      "epoch": 134.38,
      "learning_rate": 2.0820711974110032e-06,
      "loss": 0.0164,
      "step": 346040
    },
    {
      "epoch": 134.39,
      "learning_rate": 2.0815533980582524e-06,
      "loss": 0.001,
      "step": 346050
    },
    {
      "epoch": 134.39,
      "learning_rate": 2.0810355987055016e-06,
      "loss": 0.0217,
      "step": 346060
    },
    {
      "epoch": 134.4,
      "learning_rate": 2.080517799352751e-06,
      "loss": 0.0001,
      "step": 346070
    },
    {
      "epoch": 134.4,
      "learning_rate": 2.08e-06,
      "loss": 0.0266,
      "step": 346080
    },
    {
      "epoch": 134.4,
      "learning_rate": 2.079482200647249e-06,
      "loss": 0.0101,
      "step": 346090
    },
    {
      "epoch": 134.41,
      "learning_rate": 2.0789644012944984e-06,
      "loss": 0.0001,
      "step": 346100
    },
    {
      "epoch": 134.41,
      "learning_rate": 2.0784466019417476e-06,
      "loss": 0.0001,
      "step": 346110
    },
    {
      "epoch": 134.42,
      "learning_rate": 2.0779288025889968e-06,
      "loss": 0.1427,
      "step": 346120
    },
    {
      "epoch": 134.42,
      "learning_rate": 2.077411003236246e-06,
      "loss": 0.0157,
      "step": 346130
    },
    {
      "epoch": 134.42,
      "learning_rate": 2.076893203883495e-06,
      "loss": 0.0573,
      "step": 346140
    },
    {
      "epoch": 134.43,
      "learning_rate": 2.0763754045307444e-06,
      "loss": 0.0329,
      "step": 346150
    },
    {
      "epoch": 134.43,
      "learning_rate": 2.0758576051779935e-06,
      "loss": 0.017,
      "step": 346160
    },
    {
      "epoch": 134.43,
      "learning_rate": 2.075339805825243e-06,
      "loss": 0.0319,
      "step": 346170
    },
    {
      "epoch": 134.44,
      "learning_rate": 2.0748220064724924e-06,
      "loss": 0.1339,
      "step": 346180
    },
    {
      "epoch": 134.44,
      "learning_rate": 2.0743042071197416e-06,
      "loss": 0.0182,
      "step": 346190
    },
    {
      "epoch": 134.45,
      "learning_rate": 2.0737864077669907e-06,
      "loss": 0.061,
      "step": 346200
    },
    {
      "epoch": 134.45,
      "learning_rate": 2.07326860841424e-06,
      "loss": 0.0383,
      "step": 346210
    },
    {
      "epoch": 134.45,
      "learning_rate": 2.072750809061489e-06,
      "loss": 0.1304,
      "step": 346220
    },
    {
      "epoch": 134.46,
      "learning_rate": 2.0722330097087383e-06,
      "loss": 0.0006,
      "step": 346230
    },
    {
      "epoch": 134.46,
      "learning_rate": 2.0717152103559875e-06,
      "loss": 0.0726,
      "step": 346240
    },
    {
      "epoch": 134.47,
      "learning_rate": 2.0711974110032367e-06,
      "loss": 0.0105,
      "step": 346250
    },
    {
      "epoch": 134.47,
      "learning_rate": 2.0706796116504855e-06,
      "loss": 0.0297,
      "step": 346260
    },
    {
      "epoch": 134.47,
      "learning_rate": 2.0701618122977347e-06,
      "loss": 0.0009,
      "step": 346270
    },
    {
      "epoch": 134.48,
      "learning_rate": 2.069644012944984e-06,
      "loss": 0.0908,
      "step": 346280
    },
    {
      "epoch": 134.48,
      "learning_rate": 2.069126213592233e-06,
      "loss": 0.0588,
      "step": 346290
    },
    {
      "epoch": 134.49,
      "learning_rate": 2.0686084142394823e-06,
      "loss": 0.0346,
      "step": 346300
    },
    {
      "epoch": 134.49,
      "learning_rate": 2.0680906148867315e-06,
      "loss": 0.0948,
      "step": 346310
    },
    {
      "epoch": 134.49,
      "learning_rate": 2.0675728155339807e-06,
      "loss": 0.0332,
      "step": 346320
    },
    {
      "epoch": 134.5,
      "learning_rate": 2.06705501618123e-06,
      "loss": 0.0098,
      "step": 346330
    },
    {
      "epoch": 134.5,
      "learning_rate": 2.066537216828479e-06,
      "loss": 0.0143,
      "step": 346340
    },
    {
      "epoch": 134.5,
      "learning_rate": 2.0660194174757282e-06,
      "loss": 0.052,
      "step": 346350
    },
    {
      "epoch": 134.51,
      "learning_rate": 2.0655016181229774e-06,
      "loss": 0.0002,
      "step": 346360
    },
    {
      "epoch": 134.51,
      "learning_rate": 2.0649838187702266e-06,
      "loss": 0.0505,
      "step": 346370
    },
    {
      "epoch": 134.52,
      "learning_rate": 2.064466019417476e-06,
      "loss": 0.1114,
      "step": 346380
    },
    {
      "epoch": 134.52,
      "learning_rate": 2.063948220064725e-06,
      "loss": 0.0333,
      "step": 346390
    },
    {
      "epoch": 134.52,
      "learning_rate": 2.0634304207119742e-06,
      "loss": 0.0612,
      "step": 346400
    },
    {
      "epoch": 134.53,
      "learning_rate": 2.0629126213592234e-06,
      "loss": 0.019,
      "step": 346410
    },
    {
      "epoch": 134.53,
      "learning_rate": 2.0623948220064726e-06,
      "loss": 0.0001,
      "step": 346420
    },
    {
      "epoch": 134.54,
      "learning_rate": 2.061877022653722e-06,
      "loss": 0.001,
      "step": 346430
    },
    {
      "epoch": 134.54,
      "learning_rate": 2.061359223300971e-06,
      "loss": 0.0511,
      "step": 346440
    },
    {
      "epoch": 134.54,
      "learning_rate": 2.06084142394822e-06,
      "loss": 0.0395,
      "step": 346450
    },
    {
      "epoch": 134.55,
      "learning_rate": 2.0603236245954694e-06,
      "loss": 0.0222,
      "step": 346460
    },
    {
      "epoch": 134.55,
      "learning_rate": 2.0598058252427186e-06,
      "loss": 0.0429,
      "step": 346470
    },
    {
      "epoch": 134.56,
      "learning_rate": 2.0592880258899678e-06,
      "loss": 0.0176,
      "step": 346480
    },
    {
      "epoch": 134.56,
      "learning_rate": 2.058770226537217e-06,
      "loss": 0.0492,
      "step": 346490
    },
    {
      "epoch": 134.56,
      "learning_rate": 2.058252427184466e-06,
      "loss": 0.0172,
      "step": 346500
    },
    {
      "epoch": 134.57,
      "learning_rate": 2.0577346278317154e-06,
      "loss": 0.0355,
      "step": 346510
    },
    {
      "epoch": 134.57,
      "learning_rate": 2.0572168284789645e-06,
      "loss": 0.0085,
      "step": 346520
    },
    {
      "epoch": 134.57,
      "learning_rate": 2.0566990291262137e-06,
      "loss": 0.0377,
      "step": 346530
    },
    {
      "epoch": 134.58,
      "learning_rate": 2.056181229773463e-06,
      "loss": 0.0708,
      "step": 346540
    },
    {
      "epoch": 134.58,
      "learning_rate": 2.055663430420712e-06,
      "loss": 0.0061,
      "step": 346550
    },
    {
      "epoch": 134.59,
      "learning_rate": 2.0551456310679613e-06,
      "loss": 0.0278,
      "step": 346560
    },
    {
      "epoch": 134.59,
      "learning_rate": 2.0546278317152105e-06,
      "loss": 0.0275,
      "step": 346570
    },
    {
      "epoch": 134.59,
      "learning_rate": 2.0541100323624597e-06,
      "loss": 0.1146,
      "step": 346580
    },
    {
      "epoch": 134.6,
      "learning_rate": 2.053592233009709e-06,
      "loss": 0.0211,
      "step": 346590
    },
    {
      "epoch": 134.6,
      "learning_rate": 2.053074433656958e-06,
      "loss": 0.0089,
      "step": 346600
    },
    {
      "epoch": 134.61,
      "learning_rate": 2.0525566343042073e-06,
      "loss": 0.0581,
      "step": 346610
    },
    {
      "epoch": 134.61,
      "learning_rate": 2.0520388349514565e-06,
      "loss": 0.0003,
      "step": 346620
    },
    {
      "epoch": 134.61,
      "learning_rate": 2.0515210355987057e-06,
      "loss": 0.0399,
      "step": 346630
    },
    {
      "epoch": 134.62,
      "learning_rate": 2.051003236245955e-06,
      "loss": 0.0089,
      "step": 346640
    },
    {
      "epoch": 134.62,
      "learning_rate": 2.050485436893204e-06,
      "loss": 0.0635,
      "step": 346650
    },
    {
      "epoch": 134.63,
      "learning_rate": 2.0499676375404533e-06,
      "loss": 0.0746,
      "step": 346660
    },
    {
      "epoch": 134.63,
      "learning_rate": 2.0494498381877025e-06,
      "loss": 0.0143,
      "step": 346670
    },
    {
      "epoch": 134.63,
      "learning_rate": 2.0489320388349517e-06,
      "loss": 0.0523,
      "step": 346680
    },
    {
      "epoch": 134.64,
      "learning_rate": 2.048414239482201e-06,
      "loss": 0.0141,
      "step": 346690
    },
    {
      "epoch": 134.64,
      "learning_rate": 2.04789644012945e-06,
      "loss": 0.0361,
      "step": 346700
    },
    {
      "epoch": 134.64,
      "learning_rate": 2.0473786407766992e-06,
      "loss": 0.0099,
      "step": 346710
    },
    {
      "epoch": 134.65,
      "learning_rate": 2.0468608414239484e-06,
      "loss": 0.0096,
      "step": 346720
    },
    {
      "epoch": 134.65,
      "learning_rate": 2.0463430420711976e-06,
      "loss": 0.0449,
      "step": 346730
    },
    {
      "epoch": 134.66,
      "learning_rate": 2.045825242718447e-06,
      "loss": 0.017,
      "step": 346740
    },
    {
      "epoch": 134.66,
      "learning_rate": 2.045307443365696e-06,
      "loss": 0.0846,
      "step": 346750
    },
    {
      "epoch": 134.66,
      "learning_rate": 2.0447896440129452e-06,
      "loss": 0.0258,
      "step": 346760
    },
    {
      "epoch": 134.67,
      "learning_rate": 2.0442718446601944e-06,
      "loss": 0.0644,
      "step": 346770
    },
    {
      "epoch": 134.67,
      "learning_rate": 2.0437540453074436e-06,
      "loss": 0.2867,
      "step": 346780
    },
    {
      "epoch": 134.68,
      "learning_rate": 2.043236245954693e-06,
      "loss": 0.105,
      "step": 346790
    },
    {
      "epoch": 134.68,
      "learning_rate": 2.042718446601942e-06,
      "loss": 0.0116,
      "step": 346800
    },
    {
      "epoch": 134.68,
      "learning_rate": 2.042200647249191e-06,
      "loss": 0.0109,
      "step": 346810
    },
    {
      "epoch": 134.69,
      "learning_rate": 2.0416828478964404e-06,
      "loss": 0.0352,
      "step": 346820
    },
    {
      "epoch": 134.69,
      "learning_rate": 2.0411650485436896e-06,
      "loss": 0.0027,
      "step": 346830
    },
    {
      "epoch": 134.7,
      "learning_rate": 2.0406472491909388e-06,
      "loss": 0.0112,
      "step": 346840
    },
    {
      "epoch": 134.7,
      "learning_rate": 2.040129449838188e-06,
      "loss": 0.0212,
      "step": 346850
    },
    {
      "epoch": 134.7,
      "learning_rate": 2.039611650485437e-06,
      "loss": 0.0325,
      "step": 346860
    },
    {
      "epoch": 134.71,
      "learning_rate": 2.0390938511326864e-06,
      "loss": 0.0117,
      "step": 346870
    },
    {
      "epoch": 134.71,
      "learning_rate": 2.0385760517799356e-06,
      "loss": 0.1116,
      "step": 346880
    },
    {
      "epoch": 134.71,
      "learning_rate": 2.0380582524271847e-06,
      "loss": 0.0021,
      "step": 346890
    },
    {
      "epoch": 134.72,
      "learning_rate": 2.037540453074434e-06,
      "loss": 0.0001,
      "step": 346900
    },
    {
      "epoch": 134.72,
      "learning_rate": 2.037022653721683e-06,
      "loss": 0.058,
      "step": 346910
    },
    {
      "epoch": 134.73,
      "learning_rate": 2.0365048543689323e-06,
      "loss": 0.0241,
      "step": 346920
    },
    {
      "epoch": 134.73,
      "learning_rate": 2.0359870550161815e-06,
      "loss": 0.0976,
      "step": 346930
    },
    {
      "epoch": 134.73,
      "learning_rate": 2.0354692556634307e-06,
      "loss": 0.0823,
      "step": 346940
    },
    {
      "epoch": 134.74,
      "learning_rate": 2.03495145631068e-06,
      "loss": 0.0048,
      "step": 346950
    },
    {
      "epoch": 134.74,
      "learning_rate": 2.034433656957929e-06,
      "loss": 0.0001,
      "step": 346960
    },
    {
      "epoch": 134.75,
      "learning_rate": 2.0339158576051783e-06,
      "loss": 0.0269,
      "step": 346970
    },
    {
      "epoch": 134.75,
      "learning_rate": 2.0333980582524275e-06,
      "loss": 0.0751,
      "step": 346980
    },
    {
      "epoch": 134.75,
      "learning_rate": 2.0328802588996767e-06,
      "loss": 0.0009,
      "step": 346990
    },
    {
      "epoch": 134.76,
      "learning_rate": 2.032362459546926e-06,
      "loss": 0.006,
      "step": 347000
    },
    {
      "epoch": 134.76,
      "learning_rate": 2.031844660194175e-06,
      "loss": 0.0015,
      "step": 347010
    },
    {
      "epoch": 134.77,
      "learning_rate": 2.0313268608414243e-06,
      "loss": 0.1033,
      "step": 347020
    },
    {
      "epoch": 134.77,
      "learning_rate": 2.0308090614886735e-06,
      "loss": 0.0552,
      "step": 347030
    },
    {
      "epoch": 134.77,
      "learning_rate": 2.0302912621359227e-06,
      "loss": 0.0004,
      "step": 347040
    },
    {
      "epoch": 134.78,
      "learning_rate": 2.029773462783172e-06,
      "loss": 0.0121,
      "step": 347050
    },
    {
      "epoch": 134.78,
      "learning_rate": 2.029255663430421e-06,
      "loss": 0.001,
      "step": 347060
    },
    {
      "epoch": 134.78,
      "learning_rate": 2.0287378640776702e-06,
      "loss": 0.0503,
      "step": 347070
    },
    {
      "epoch": 134.79,
      "learning_rate": 2.028220064724919e-06,
      "loss": 0.0183,
      "step": 347080
    },
    {
      "epoch": 134.79,
      "learning_rate": 2.0277022653721682e-06,
      "loss": 0.0523,
      "step": 347090
    },
    {
      "epoch": 134.8,
      "learning_rate": 2.0271844660194174e-06,
      "loss": 0.0176,
      "step": 347100
    },
    {
      "epoch": 134.8,
      "learning_rate": 2.0266666666666666e-06,
      "loss": 0.1015,
      "step": 347110
    },
    {
      "epoch": 134.8,
      "learning_rate": 2.026148867313916e-06,
      "loss": 0.0843,
      "step": 347120
    },
    {
      "epoch": 134.81,
      "learning_rate": 2.025631067961165e-06,
      "loss": 0.009,
      "step": 347130
    },
    {
      "epoch": 134.81,
      "learning_rate": 2.025113268608414e-06,
      "loss": 0.0933,
      "step": 347140
    },
    {
      "epoch": 134.82,
      "learning_rate": 2.0245954692556634e-06,
      "loss": 0.0229,
      "step": 347150
    },
    {
      "epoch": 134.82,
      "learning_rate": 2.0240776699029126e-06,
      "loss": 0.0639,
      "step": 347160
    },
    {
      "epoch": 134.82,
      "learning_rate": 2.0235598705501618e-06,
      "loss": 0.0394,
      "step": 347170
    },
    {
      "epoch": 134.83,
      "learning_rate": 2.023042071197411e-06,
      "loss": 0.0793,
      "step": 347180
    },
    {
      "epoch": 134.83,
      "learning_rate": 2.02252427184466e-06,
      "loss": 0.1003,
      "step": 347190
    },
    {
      "epoch": 134.83,
      "learning_rate": 2.0220064724919094e-06,
      "loss": 0.0012,
      "step": 347200
    },
    {
      "epoch": 134.84,
      "learning_rate": 2.0214886731391585e-06,
      "loss": 0.008,
      "step": 347210
    },
    {
      "epoch": 134.84,
      "learning_rate": 2.0209708737864077e-06,
      "loss": 0.0369,
      "step": 347220
    },
    {
      "epoch": 134.85,
      "learning_rate": 2.020453074433657e-06,
      "loss": 0.0137,
      "step": 347230
    },
    {
      "epoch": 134.85,
      "learning_rate": 2.019935275080906e-06,
      "loss": 0.1049,
      "step": 347240
    },
    {
      "epoch": 134.85,
      "learning_rate": 2.0194174757281553e-06,
      "loss": 0.0338,
      "step": 347250
    },
    {
      "epoch": 134.86,
      "learning_rate": 2.0188996763754045e-06,
      "loss": 0.019,
      "step": 347260
    },
    {
      "epoch": 134.86,
      "learning_rate": 2.0183818770226537e-06,
      "loss": 0.0156,
      "step": 347270
    },
    {
      "epoch": 134.87,
      "learning_rate": 2.017864077669903e-06,
      "loss": 0.0082,
      "step": 347280
    },
    {
      "epoch": 134.87,
      "learning_rate": 2.0173462783171525e-06,
      "loss": 0.0258,
      "step": 347290
    },
    {
      "epoch": 134.87,
      "learning_rate": 2.0168284789644017e-06,
      "loss": 0.0775,
      "step": 347300
    },
    {
      "epoch": 134.88,
      "learning_rate": 2.016310679611651e-06,
      "loss": 0.0546,
      "step": 347310
    },
    {
      "epoch": 134.88,
      "learning_rate": 2.0157928802589e-06,
      "loss": 0.0149,
      "step": 347320
    },
    {
      "epoch": 134.89,
      "learning_rate": 2.0152750809061493e-06,
      "loss": 0.13,
      "step": 347330
    },
    {
      "epoch": 134.89,
      "learning_rate": 2.0147572815533985e-06,
      "loss": 0.0165,
      "step": 347340
    },
    {
      "epoch": 134.89,
      "learning_rate": 2.0142394822006477e-06,
      "loss": 0.0504,
      "step": 347350
    },
    {
      "epoch": 134.9,
      "learning_rate": 2.013721682847897e-06,
      "loss": 0.0165,
      "step": 347360
    },
    {
      "epoch": 134.9,
      "learning_rate": 2.013203883495146e-06,
      "loss": 0.0939,
      "step": 347370
    },
    {
      "epoch": 134.9,
      "learning_rate": 2.0126860841423953e-06,
      "loss": 0.0256,
      "step": 347380
    },
    {
      "epoch": 134.91,
      "learning_rate": 2.0121682847896445e-06,
      "loss": 0.0161,
      "step": 347390
    },
    {
      "epoch": 134.91,
      "learning_rate": 2.0116504854368932e-06,
      "loss": 0.1366,
      "step": 347400
    },
    {
      "epoch": 134.92,
      "learning_rate": 2.0111326860841424e-06,
      "loss": 0.0786,
      "step": 347410
    },
    {
      "epoch": 134.92,
      "learning_rate": 2.0106148867313916e-06,
      "loss": 0.0139,
      "step": 347420
    },
    {
      "epoch": 134.92,
      "learning_rate": 2.010097087378641e-06,
      "loss": 0.0498,
      "step": 347430
    },
    {
      "epoch": 134.93,
      "learning_rate": 2.00957928802589e-06,
      "loss": 0.0301,
      "step": 347440
    },
    {
      "epoch": 134.93,
      "learning_rate": 2.0090614886731392e-06,
      "loss": 0.001,
      "step": 347450
    },
    {
      "epoch": 134.94,
      "learning_rate": 2.0085436893203884e-06,
      "loss": 0.0139,
      "step": 347460
    },
    {
      "epoch": 134.94,
      "learning_rate": 2.0080258899676376e-06,
      "loss": 0.0152,
      "step": 347470
    },
    {
      "epoch": 134.94,
      "learning_rate": 2.007508090614887e-06,
      "loss": 0.029,
      "step": 347480
    },
    {
      "epoch": 134.95,
      "learning_rate": 2.006990291262136e-06,
      "loss": 0.0201,
      "step": 347490
    },
    {
      "epoch": 134.95,
      "learning_rate": 2.006472491909385e-06,
      "loss": 0.1045,
      "step": 347500
    },
    {
      "epoch": 134.96,
      "learning_rate": 2.0059546925566344e-06,
      "loss": 0.1176,
      "step": 347510
    },
    {
      "epoch": 134.96,
      "learning_rate": 2.0054368932038836e-06,
      "loss": 0.0102,
      "step": 347520
    },
    {
      "epoch": 134.96,
      "learning_rate": 2.0049190938511328e-06,
      "loss": 0.0001,
      "step": 347530
    },
    {
      "epoch": 134.97,
      "learning_rate": 2.004401294498382e-06,
      "loss": 0.0064,
      "step": 347540
    },
    {
      "epoch": 134.97,
      "learning_rate": 2.003883495145631e-06,
      "loss": 0.0604,
      "step": 347550
    },
    {
      "epoch": 134.97,
      "learning_rate": 2.0033656957928804e-06,
      "loss": 0.0154,
      "step": 347560
    },
    {
      "epoch": 134.98,
      "learning_rate": 2.0028478964401296e-06,
      "loss": 0.0197,
      "step": 347570
    },
    {
      "epoch": 134.98,
      "learning_rate": 2.0023300970873787e-06,
      "loss": 0.0088,
      "step": 347580
    },
    {
      "epoch": 134.99,
      "learning_rate": 2.001812297734628e-06,
      "loss": 0.0053,
      "step": 347590
    },
    {
      "epoch": 134.99,
      "learning_rate": 2.001294498381877e-06,
      "loss": 0.0525,
      "step": 347600
    },
    {
      "epoch": 134.99,
      "learning_rate": 2.0007766990291263e-06,
      "loss": 0.0004,
      "step": 347610
    },
    {
      "epoch": 135.0,
      "learning_rate": 2.0002588996763755e-06,
      "loss": 0.0116,
      "step": 347620
    },
    {
      "epoch": 135.0,
      "eval_accuracy": 0.9537826685006877,
      "eval_loss": 0.3890344202518463,
      "eval_runtime": 8.1691,
      "eval_samples_per_second": 444.97,
      "eval_steps_per_second": 55.698,
      "step": 347625
    },
    {
      "epoch": 135.0,
      "learning_rate": 1.9997411003236247e-06,
      "loss": 0.1281,
      "step": 347630
    },
    {
      "epoch": 135.01,
      "learning_rate": 1.999223300970874e-06,
      "loss": 0.0068,
      "step": 347640
    },
    {
      "epoch": 135.01,
      "learning_rate": 1.998705501618123e-06,
      "loss": 0.076,
      "step": 347650
    },
    {
      "epoch": 135.01,
      "learning_rate": 1.9981877022653723e-06,
      "loss": 0.0405,
      "step": 347660
    },
    {
      "epoch": 135.02,
      "learning_rate": 1.9976699029126215e-06,
      "loss": 0.0598,
      "step": 347670
    },
    {
      "epoch": 135.02,
      "learning_rate": 1.9971521035598707e-06,
      "loss": 0.0281,
      "step": 347680
    },
    {
      "epoch": 135.03,
      "learning_rate": 1.99663430420712e-06,
      "loss": 0.1554,
      "step": 347690
    },
    {
      "epoch": 135.03,
      "learning_rate": 1.996116504854369e-06,
      "loss": 0.0095,
      "step": 347700
    },
    {
      "epoch": 135.03,
      "learning_rate": 1.9955987055016183e-06,
      "loss": 0.0112,
      "step": 347710
    },
    {
      "epoch": 135.04,
      "learning_rate": 1.9950809061488675e-06,
      "loss": 0.0647,
      "step": 347720
    },
    {
      "epoch": 135.04,
      "learning_rate": 1.9945631067961167e-06,
      "loss": 0.0671,
      "step": 347730
    },
    {
      "epoch": 135.04,
      "learning_rate": 1.994045307443366e-06,
      "loss": 0.0089,
      "step": 347740
    },
    {
      "epoch": 135.05,
      "learning_rate": 1.993527508090615e-06,
      "loss": 0.0016,
      "step": 347750
    },
    {
      "epoch": 135.05,
      "learning_rate": 1.9930097087378642e-06,
      "loss": 0.1162,
      "step": 347760
    },
    {
      "epoch": 135.06,
      "learning_rate": 1.9924919093851134e-06,
      "loss": 0.077,
      "step": 347770
    },
    {
      "epoch": 135.06,
      "learning_rate": 1.9919741100323626e-06,
      "loss": 0.0229,
      "step": 347780
    },
    {
      "epoch": 135.06,
      "learning_rate": 1.991456310679612e-06,
      "loss": 0.0156,
      "step": 347790
    },
    {
      "epoch": 135.07,
      "learning_rate": 1.990938511326861e-06,
      "loss": 0.0635,
      "step": 347800
    },
    {
      "epoch": 135.07,
      "learning_rate": 1.9904207119741102e-06,
      "loss": 0.067,
      "step": 347810
    },
    {
      "epoch": 135.08,
      "learning_rate": 1.9899029126213594e-06,
      "loss": 0.0107,
      "step": 347820
    },
    {
      "epoch": 135.08,
      "learning_rate": 1.9893851132686086e-06,
      "loss": 0.0128,
      "step": 347830
    },
    {
      "epoch": 135.08,
      "learning_rate": 1.988867313915858e-06,
      "loss": 0.0001,
      "step": 347840
    },
    {
      "epoch": 135.09,
      "learning_rate": 1.988349514563107e-06,
      "loss": 0.0427,
      "step": 347850
    },
    {
      "epoch": 135.09,
      "learning_rate": 1.987831715210356e-06,
      "loss": 0.0094,
      "step": 347860
    },
    {
      "epoch": 135.1,
      "learning_rate": 1.9873139158576054e-06,
      "loss": 0.025,
      "step": 347870
    },
    {
      "epoch": 135.1,
      "learning_rate": 1.9867961165048546e-06,
      "loss": 0.0802,
      "step": 347880
    },
    {
      "epoch": 135.1,
      "learning_rate": 1.9862783171521038e-06,
      "loss": 0.0253,
      "step": 347890
    },
    {
      "epoch": 135.11,
      "learning_rate": 1.985760517799353e-06,
      "loss": 0.0265,
      "step": 347900
    },
    {
      "epoch": 135.11,
      "learning_rate": 1.985242718446602e-06,
      "loss": 0.025,
      "step": 347910
    },
    {
      "epoch": 135.11,
      "learning_rate": 1.9847249190938514e-06,
      "loss": 0.0108,
      "step": 347920
    },
    {
      "epoch": 135.12,
      "learning_rate": 1.9842071197411006e-06,
      "loss": 0.0412,
      "step": 347930
    },
    {
      "epoch": 135.12,
      "learning_rate": 1.9836893203883497e-06,
      "loss": 0.073,
      "step": 347940
    },
    {
      "epoch": 135.13,
      "learning_rate": 1.983171521035599e-06,
      "loss": 0.0016,
      "step": 347950
    },
    {
      "epoch": 135.13,
      "learning_rate": 1.982653721682848e-06,
      "loss": 0.0093,
      "step": 347960
    },
    {
      "epoch": 135.13,
      "learning_rate": 1.9821359223300973e-06,
      "loss": 0.0018,
      "step": 347970
    },
    {
      "epoch": 135.14,
      "learning_rate": 1.9816181229773465e-06,
      "loss": 0.0687,
      "step": 347980
    },
    {
      "epoch": 135.14,
      "learning_rate": 1.9811003236245957e-06,
      "loss": 0.027,
      "step": 347990
    },
    {
      "epoch": 135.15,
      "learning_rate": 1.980582524271845e-06,
      "loss": 0.0017,
      "step": 348000
    },
    {
      "epoch": 135.15,
      "learning_rate": 1.980064724919094e-06,
      "loss": 0.0215,
      "step": 348010
    },
    {
      "epoch": 135.15,
      "learning_rate": 1.9795469255663433e-06,
      "loss": 0.0661,
      "step": 348020
    },
    {
      "epoch": 135.16,
      "learning_rate": 1.9790291262135925e-06,
      "loss": 0.0001,
      "step": 348030
    },
    {
      "epoch": 135.16,
      "learning_rate": 1.9785113268608417e-06,
      "loss": 0.0411,
      "step": 348040
    },
    {
      "epoch": 135.17,
      "learning_rate": 1.977993527508091e-06,
      "loss": 0.0366,
      "step": 348050
    },
    {
      "epoch": 135.17,
      "learning_rate": 1.97747572815534e-06,
      "loss": 0.0402,
      "step": 348060
    },
    {
      "epoch": 135.17,
      "learning_rate": 1.9769579288025893e-06,
      "loss": 0.0247,
      "step": 348070
    },
    {
      "epoch": 135.18,
      "learning_rate": 1.9764401294498385e-06,
      "loss": 0.0257,
      "step": 348080
    },
    {
      "epoch": 135.18,
      "learning_rate": 1.9759223300970877e-06,
      "loss": 0.0013,
      "step": 348090
    },
    {
      "epoch": 135.18,
      "learning_rate": 1.975404530744337e-06,
      "loss": 0.0522,
      "step": 348100
    },
    {
      "epoch": 135.19,
      "learning_rate": 1.974886731391586e-06,
      "loss": 0.0383,
      "step": 348110
    },
    {
      "epoch": 135.19,
      "learning_rate": 1.9743689320388353e-06,
      "loss": 0.0444,
      "step": 348120
    },
    {
      "epoch": 135.2,
      "learning_rate": 1.9738511326860844e-06,
      "loss": 0.0196,
      "step": 348130
    },
    {
      "epoch": 135.2,
      "learning_rate": 1.9733333333333336e-06,
      "loss": 0.0864,
      "step": 348140
    },
    {
      "epoch": 135.2,
      "learning_rate": 1.972815533980583e-06,
      "loss": 0.0151,
      "step": 348150
    },
    {
      "epoch": 135.21,
      "learning_rate": 1.972297734627832e-06,
      "loss": 0.0919,
      "step": 348160
    },
    {
      "epoch": 135.21,
      "learning_rate": 1.9717799352750812e-06,
      "loss": 0.0001,
      "step": 348170
    },
    {
      "epoch": 135.22,
      "learning_rate": 1.9712621359223304e-06,
      "loss": 0.0011,
      "step": 348180
    },
    {
      "epoch": 135.22,
      "learning_rate": 1.9707443365695796e-06,
      "loss": 0.0003,
      "step": 348190
    },
    {
      "epoch": 135.22,
      "learning_rate": 1.970226537216829e-06,
      "loss": 0.0346,
      "step": 348200
    },
    {
      "epoch": 135.23,
      "learning_rate": 1.969708737864078e-06,
      "loss": 0.0123,
      "step": 348210
    },
    {
      "epoch": 135.23,
      "learning_rate": 1.9691909385113268e-06,
      "loss": 0.0414,
      "step": 348220
    },
    {
      "epoch": 135.23,
      "learning_rate": 1.968673139158576e-06,
      "loss": 0.0793,
      "step": 348230
    },
    {
      "epoch": 135.24,
      "learning_rate": 1.968155339805825e-06,
      "loss": 0.1104,
      "step": 348240
    },
    {
      "epoch": 135.24,
      "learning_rate": 1.9676375404530744e-06,
      "loss": 0.0003,
      "step": 348250
    },
    {
      "epoch": 135.25,
      "learning_rate": 1.9671197411003236e-06,
      "loss": 0.0388,
      "step": 348260
    },
    {
      "epoch": 135.25,
      "learning_rate": 1.9666019417475727e-06,
      "loss": 0.0092,
      "step": 348270
    },
    {
      "epoch": 135.25,
      "learning_rate": 1.966084142394822e-06,
      "loss": 0.0001,
      "step": 348280
    },
    {
      "epoch": 135.26,
      "learning_rate": 1.965566343042071e-06,
      "loss": 0.0095,
      "step": 348290
    },
    {
      "epoch": 135.26,
      "learning_rate": 1.9650485436893203e-06,
      "loss": 0.0006,
      "step": 348300
    },
    {
      "epoch": 135.27,
      "learning_rate": 1.9645307443365695e-06,
      "loss": 0.1137,
      "step": 348310
    },
    {
      "epoch": 135.27,
      "learning_rate": 1.9640129449838187e-06,
      "loss": 0.0837,
      "step": 348320
    },
    {
      "epoch": 135.27,
      "learning_rate": 1.963495145631068e-06,
      "loss": 0.1381,
      "step": 348330
    },
    {
      "epoch": 135.28,
      "learning_rate": 1.962977346278317e-06,
      "loss": 0.1481,
      "step": 348340
    },
    {
      "epoch": 135.28,
      "learning_rate": 1.9624595469255663e-06,
      "loss": 0.1071,
      "step": 348350
    },
    {
      "epoch": 135.29,
      "learning_rate": 1.9619417475728155e-06,
      "loss": 0.0082,
      "step": 348360
    },
    {
      "epoch": 135.29,
      "learning_rate": 1.9614239482200647e-06,
      "loss": 0.0201,
      "step": 348370
    },
    {
      "epoch": 135.29,
      "learning_rate": 1.960906148867314e-06,
      "loss": 0.0093,
      "step": 348380
    },
    {
      "epoch": 135.3,
      "learning_rate": 1.960388349514563e-06,
      "loss": 0.053,
      "step": 348390
    },
    {
      "epoch": 135.3,
      "learning_rate": 1.9598705501618123e-06,
      "loss": 0.0809,
      "step": 348400
    },
    {
      "epoch": 135.3,
      "learning_rate": 1.959352750809062e-06,
      "loss": 0.1146,
      "step": 348410
    },
    {
      "epoch": 135.31,
      "learning_rate": 1.958834951456311e-06,
      "loss": 0.0772,
      "step": 348420
    },
    {
      "epoch": 135.31,
      "learning_rate": 1.9583171521035603e-06,
      "loss": 0.0121,
      "step": 348430
    },
    {
      "epoch": 135.32,
      "learning_rate": 1.9577993527508095e-06,
      "loss": 0.0801,
      "step": 348440
    },
    {
      "epoch": 135.32,
      "learning_rate": 1.9572815533980587e-06,
      "loss": 0.0643,
      "step": 348450
    },
    {
      "epoch": 135.32,
      "learning_rate": 1.956763754045308e-06,
      "loss": 0.0276,
      "step": 348460
    },
    {
      "epoch": 135.33,
      "learning_rate": 1.956245954692557e-06,
      "loss": 0.0443,
      "step": 348470
    },
    {
      "epoch": 135.33,
      "learning_rate": 1.9557281553398063e-06,
      "loss": 0.0088,
      "step": 348480
    },
    {
      "epoch": 135.34,
      "learning_rate": 1.9552103559870554e-06,
      "loss": 0.095,
      "step": 348490
    },
    {
      "epoch": 135.34,
      "learning_rate": 1.9546925566343046e-06,
      "loss": 0.0164,
      "step": 348500
    },
    {
      "epoch": 135.34,
      "learning_rate": 1.954174757281554e-06,
      "loss": 0.0601,
      "step": 348510
    },
    {
      "epoch": 135.35,
      "learning_rate": 1.953656957928803e-06,
      "loss": 0.0344,
      "step": 348520
    },
    {
      "epoch": 135.35,
      "learning_rate": 1.9531391585760522e-06,
      "loss": 0.0097,
      "step": 348530
    },
    {
      "epoch": 135.36,
      "learning_rate": 1.952621359223301e-06,
      "loss": 0.0221,
      "step": 348540
    },
    {
      "epoch": 135.36,
      "learning_rate": 1.95210355987055e-06,
      "loss": 0.0219,
      "step": 348550
    },
    {
      "epoch": 135.36,
      "learning_rate": 1.9515857605177994e-06,
      "loss": 0.0303,
      "step": 348560
    },
    {
      "epoch": 135.37,
      "learning_rate": 1.9510679611650486e-06,
      "loss": 0.0086,
      "step": 348570
    },
    {
      "epoch": 135.37,
      "learning_rate": 1.9505501618122978e-06,
      "loss": 0.0002,
      "step": 348580
    },
    {
      "epoch": 135.37,
      "learning_rate": 1.950032362459547e-06,
      "loss": 0.0358,
      "step": 348590
    },
    {
      "epoch": 135.38,
      "learning_rate": 1.949514563106796e-06,
      "loss": 0.0133,
      "step": 348600
    },
    {
      "epoch": 135.38,
      "learning_rate": 1.9489967637540454e-06,
      "loss": 0.0301,
      "step": 348610
    },
    {
      "epoch": 135.39,
      "learning_rate": 1.9484789644012946e-06,
      "loss": 0.027,
      "step": 348620
    },
    {
      "epoch": 135.39,
      "learning_rate": 1.9479611650485437e-06,
      "loss": 0.021,
      "step": 348630
    },
    {
      "epoch": 135.39,
      "learning_rate": 1.947443365695793e-06,
      "loss": 0.0179,
      "step": 348640
    },
    {
      "epoch": 135.4,
      "learning_rate": 1.946925566343042e-06,
      "loss": 0.0535,
      "step": 348650
    },
    {
      "epoch": 135.4,
      "learning_rate": 1.9464077669902913e-06,
      "loss": 0.0379,
      "step": 348660
    },
    {
      "epoch": 135.41,
      "learning_rate": 1.9458899676375405e-06,
      "loss": 0.0795,
      "step": 348670
    },
    {
      "epoch": 135.41,
      "learning_rate": 1.9453721682847897e-06,
      "loss": 0.0771,
      "step": 348680
    },
    {
      "epoch": 135.41,
      "learning_rate": 1.944854368932039e-06,
      "loss": 0.0989,
      "step": 348690
    },
    {
      "epoch": 135.42,
      "learning_rate": 1.944336569579288e-06,
      "loss": 0.0123,
      "step": 348700
    },
    {
      "epoch": 135.42,
      "learning_rate": 1.9438187702265373e-06,
      "loss": 0.1092,
      "step": 348710
    },
    {
      "epoch": 135.43,
      "learning_rate": 1.9433009708737865e-06,
      "loss": 0.0005,
      "step": 348720
    },
    {
      "epoch": 135.43,
      "learning_rate": 1.9427831715210357e-06,
      "loss": 0.0497,
      "step": 348730
    },
    {
      "epoch": 135.43,
      "learning_rate": 1.942265372168285e-06,
      "loss": 0.1345,
      "step": 348740
    },
    {
      "epoch": 135.44,
      "learning_rate": 1.941747572815534e-06,
      "loss": 0.0835,
      "step": 348750
    },
    {
      "epoch": 135.44,
      "learning_rate": 1.9412297734627833e-06,
      "loss": 0.0391,
      "step": 348760
    },
    {
      "epoch": 135.44,
      "learning_rate": 1.9407119741100325e-06,
      "loss": 0.0607,
      "step": 348770
    },
    {
      "epoch": 135.45,
      "learning_rate": 1.9401941747572817e-06,
      "loss": 0.0687,
      "step": 348780
    },
    {
      "epoch": 135.45,
      "learning_rate": 1.939676375404531e-06,
      "loss": 0.0835,
      "step": 348790
    },
    {
      "epoch": 135.46,
      "learning_rate": 1.93915857605178e-06,
      "loss": 0.0184,
      "step": 348800
    },
    {
      "epoch": 135.46,
      "learning_rate": 1.9386407766990292e-06,
      "loss": 0.0004,
      "step": 348810
    },
    {
      "epoch": 135.46,
      "learning_rate": 1.9381229773462784e-06,
      "loss": 0.0575,
      "step": 348820
    },
    {
      "epoch": 135.47,
      "learning_rate": 1.9376051779935276e-06,
      "loss": 0.0435,
      "step": 348830
    },
    {
      "epoch": 135.47,
      "learning_rate": 1.937087378640777e-06,
      "loss": 0.0004,
      "step": 348840
    },
    {
      "epoch": 135.48,
      "learning_rate": 1.936569579288026e-06,
      "loss": 0.0095,
      "step": 348850
    },
    {
      "epoch": 135.48,
      "learning_rate": 1.9360517799352752e-06,
      "loss": 0.06,
      "step": 348860
    },
    {
      "epoch": 135.48,
      "learning_rate": 1.9355339805825244e-06,
      "loss": 0.1928,
      "step": 348870
    },
    {
      "epoch": 135.49,
      "learning_rate": 1.9350161812297736e-06,
      "loss": 0.0006,
      "step": 348880
    },
    {
      "epoch": 135.49,
      "learning_rate": 1.934498381877023e-06,
      "loss": 0.0519,
      "step": 348890
    },
    {
      "epoch": 135.5,
      "learning_rate": 1.933980582524272e-06,
      "loss": 0.0175,
      "step": 348900
    },
    {
      "epoch": 135.5,
      "learning_rate": 1.933462783171521e-06,
      "loss": 0.0286,
      "step": 348910
    },
    {
      "epoch": 135.5,
      "learning_rate": 1.9329449838187704e-06,
      "loss": 0.0183,
      "step": 348920
    },
    {
      "epoch": 135.51,
      "learning_rate": 1.9324271844660196e-06,
      "loss": 0.0337,
      "step": 348930
    },
    {
      "epoch": 135.51,
      "learning_rate": 1.9319093851132688e-06,
      "loss": 0.0202,
      "step": 348940
    },
    {
      "epoch": 135.51,
      "learning_rate": 1.931391585760518e-06,
      "loss": 0.1068,
      "step": 348950
    },
    {
      "epoch": 135.52,
      "learning_rate": 1.930873786407767e-06,
      "loss": 0.008,
      "step": 348960
    },
    {
      "epoch": 135.52,
      "learning_rate": 1.9303559870550164e-06,
      "loss": 0.0308,
      "step": 348970
    },
    {
      "epoch": 135.53,
      "learning_rate": 1.9298381877022656e-06,
      "loss": 0.0019,
      "step": 348980
    },
    {
      "epoch": 135.53,
      "learning_rate": 1.9293203883495148e-06,
      "loss": 0.0225,
      "step": 348990
    },
    {
      "epoch": 135.53,
      "learning_rate": 1.928802588996764e-06,
      "loss": 0.1176,
      "step": 349000
    },
    {
      "epoch": 135.54,
      "learning_rate": 1.928284789644013e-06,
      "loss": 0.0003,
      "step": 349010
    },
    {
      "epoch": 135.54,
      "learning_rate": 1.9277669902912623e-06,
      "loss": 0.001,
      "step": 349020
    },
    {
      "epoch": 135.55,
      "learning_rate": 1.9272491909385115e-06,
      "loss": 0.0802,
      "step": 349030
    },
    {
      "epoch": 135.55,
      "learning_rate": 1.9267313915857607e-06,
      "loss": 0.0406,
      "step": 349040
    },
    {
      "epoch": 135.55,
      "learning_rate": 1.92621359223301e-06,
      "loss": 0.0267,
      "step": 349050
    },
    {
      "epoch": 135.56,
      "learning_rate": 1.925695792880259e-06,
      "loss": 0.1438,
      "step": 349060
    },
    {
      "epoch": 135.56,
      "learning_rate": 1.9251779935275083e-06,
      "loss": 0.0102,
      "step": 349070
    },
    {
      "epoch": 135.57,
      "learning_rate": 1.9246601941747575e-06,
      "loss": 0.0269,
      "step": 349080
    },
    {
      "epoch": 135.57,
      "learning_rate": 1.9241423948220067e-06,
      "loss": 0.0477,
      "step": 349090
    },
    {
      "epoch": 135.57,
      "learning_rate": 1.923624595469256e-06,
      "loss": 0.0878,
      "step": 349100
    },
    {
      "epoch": 135.58,
      "learning_rate": 1.923106796116505e-06,
      "loss": 0.0729,
      "step": 349110
    },
    {
      "epoch": 135.58,
      "learning_rate": 1.9225889967637543e-06,
      "loss": 0.0144,
      "step": 349120
    },
    {
      "epoch": 135.58,
      "learning_rate": 1.9220711974110035e-06,
      "loss": 0.0297,
      "step": 349130
    },
    {
      "epoch": 135.59,
      "learning_rate": 1.9215533980582527e-06,
      "loss": 0.0209,
      "step": 349140
    },
    {
      "epoch": 135.59,
      "learning_rate": 1.921035598705502e-06,
      "loss": 0.0001,
      "step": 349150
    },
    {
      "epoch": 135.6,
      "learning_rate": 1.920517799352751e-06,
      "loss": 0.0856,
      "step": 349160
    },
    {
      "epoch": 135.6,
      "learning_rate": 1.9200000000000003e-06,
      "loss": 0.1308,
      "step": 349170
    },
    {
      "epoch": 135.6,
      "learning_rate": 1.9194822006472494e-06,
      "loss": 0.1724,
      "step": 349180
    },
    {
      "epoch": 135.61,
      "learning_rate": 1.9189644012944986e-06,
      "loss": 0.0259,
      "step": 349190
    },
    {
      "epoch": 135.61,
      "learning_rate": 1.918446601941748e-06,
      "loss": 0.0001,
      "step": 349200
    },
    {
      "epoch": 135.62,
      "learning_rate": 1.917928802588997e-06,
      "loss": 0.0011,
      "step": 349210
    },
    {
      "epoch": 135.62,
      "learning_rate": 1.9174110032362462e-06,
      "loss": 0.0002,
      "step": 349220
    },
    {
      "epoch": 135.62,
      "learning_rate": 1.9168932038834954e-06,
      "loss": 0.0367,
      "step": 349230
    },
    {
      "epoch": 135.63,
      "learning_rate": 1.9163754045307446e-06,
      "loss": 0.072,
      "step": 349240
    },
    {
      "epoch": 135.63,
      "learning_rate": 1.915857605177994e-06,
      "loss": 0.0095,
      "step": 349250
    },
    {
      "epoch": 135.63,
      "learning_rate": 1.915339805825243e-06,
      "loss": 0.1022,
      "step": 349260
    },
    {
      "epoch": 135.64,
      "learning_rate": 1.914822006472492e-06,
      "loss": 0.0042,
      "step": 349270
    },
    {
      "epoch": 135.64,
      "learning_rate": 1.9143042071197414e-06,
      "loss": 0.0663,
      "step": 349280
    },
    {
      "epoch": 135.65,
      "learning_rate": 1.9137864077669906e-06,
      "loss": 0.0775,
      "step": 349290
    },
    {
      "epoch": 135.65,
      "learning_rate": 1.9132686084142398e-06,
      "loss": 0.0356,
      "step": 349300
    },
    {
      "epoch": 135.65,
      "learning_rate": 1.912750809061489e-06,
      "loss": 0.0005,
      "step": 349310
    },
    {
      "epoch": 135.66,
      "learning_rate": 1.912233009708738e-06,
      "loss": 0.0273,
      "step": 349320
    },
    {
      "epoch": 135.66,
      "learning_rate": 1.9117152103559874e-06,
      "loss": 0.0496,
      "step": 349330
    },
    {
      "epoch": 135.67,
      "learning_rate": 1.9111974110032366e-06,
      "loss": 0.0002,
      "step": 349340
    },
    {
      "epoch": 135.67,
      "learning_rate": 1.9106796116504858e-06,
      "loss": 0.0083,
      "step": 349350
    },
    {
      "epoch": 135.67,
      "learning_rate": 1.9101618122977345e-06,
      "loss": 0.0625,
      "step": 349360
    },
    {
      "epoch": 135.68,
      "learning_rate": 1.9096440129449837e-06,
      "loss": 0.1178,
      "step": 349370
    },
    {
      "epoch": 135.68,
      "learning_rate": 1.909126213592233e-06,
      "loss": 0.0615,
      "step": 349380
    },
    {
      "epoch": 135.69,
      "learning_rate": 1.908608414239482e-06,
      "loss": 0.0119,
      "step": 349390
    },
    {
      "epoch": 135.69,
      "learning_rate": 1.9080906148867313e-06,
      "loss": 0.0001,
      "step": 349400
    },
    {
      "epoch": 135.69,
      "learning_rate": 1.9075728155339805e-06,
      "loss": 0.0222,
      "step": 349410
    },
    {
      "epoch": 135.7,
      "learning_rate": 1.90705501618123e-06,
      "loss": 0.1245,
      "step": 349420
    },
    {
      "epoch": 135.7,
      "learning_rate": 1.906537216828479e-06,
      "loss": 0.027,
      "step": 349430
    },
    {
      "epoch": 135.7,
      "learning_rate": 1.9060194174757283e-06,
      "loss": 0.0753,
      "step": 349440
    },
    {
      "epoch": 135.71,
      "learning_rate": 1.9055016181229775e-06,
      "loss": 0.1084,
      "step": 349450
    },
    {
      "epoch": 135.71,
      "learning_rate": 1.9049838187702267e-06,
      "loss": 0.0091,
      "step": 349460
    },
    {
      "epoch": 135.72,
      "learning_rate": 1.9044660194174759e-06,
      "loss": 0.03,
      "step": 349470
    },
    {
      "epoch": 135.72,
      "learning_rate": 1.9039482200647249e-06,
      "loss": 0.0069,
      "step": 349480
    },
    {
      "epoch": 135.72,
      "learning_rate": 1.903430420711974e-06,
      "loss": 0.0004,
      "step": 349490
    },
    {
      "epoch": 135.73,
      "learning_rate": 1.9029126213592232e-06,
      "loss": 0.0595,
      "step": 349500
    },
    {
      "epoch": 135.73,
      "learning_rate": 1.9023948220064724e-06,
      "loss": 0.0662,
      "step": 349510
    },
    {
      "epoch": 135.74,
      "learning_rate": 1.9018770226537216e-06,
      "loss": 0.0317,
      "step": 349520
    },
    {
      "epoch": 135.74,
      "learning_rate": 1.9013592233009708e-06,
      "loss": 0.0577,
      "step": 349530
    },
    {
      "epoch": 135.74,
      "learning_rate": 1.9008414239482204e-06,
      "loss": 0.0085,
      "step": 349540
    },
    {
      "epoch": 135.75,
      "learning_rate": 1.9003236245954696e-06,
      "loss": 0.0823,
      "step": 349550
    },
    {
      "epoch": 135.75,
      "learning_rate": 1.8998058252427186e-06,
      "loss": 0.0002,
      "step": 349560
    },
    {
      "epoch": 135.76,
      "learning_rate": 1.8992880258899678e-06,
      "loss": 0.0329,
      "step": 349570
    },
    {
      "epoch": 135.76,
      "learning_rate": 1.898770226537217e-06,
      "loss": 0.0726,
      "step": 349580
    },
    {
      "epoch": 135.76,
      "learning_rate": 1.8982524271844662e-06,
      "loss": 0.0589,
      "step": 349590
    },
    {
      "epoch": 135.77,
      "learning_rate": 1.8977346278317154e-06,
      "loss": 0.0051,
      "step": 349600
    },
    {
      "epoch": 135.77,
      "learning_rate": 1.8972168284789646e-06,
      "loss": 0.0283,
      "step": 349610
    },
    {
      "epoch": 135.77,
      "learning_rate": 1.8966990291262138e-06,
      "loss": 0.0574,
      "step": 349620
    },
    {
      "epoch": 135.78,
      "learning_rate": 1.896181229773463e-06,
      "loss": 0.0185,
      "step": 349630
    },
    {
      "epoch": 135.78,
      "learning_rate": 1.8956634304207122e-06,
      "loss": 0.064,
      "step": 349640
    },
    {
      "epoch": 135.79,
      "learning_rate": 1.8951456310679614e-06,
      "loss": 0.0113,
      "step": 349650
    },
    {
      "epoch": 135.79,
      "learning_rate": 1.8946278317152106e-06,
      "loss": 0.0296,
      "step": 349660
    },
    {
      "epoch": 135.79,
      "learning_rate": 1.8941100323624598e-06,
      "loss": 0.0621,
      "step": 349670
    },
    {
      "epoch": 135.8,
      "learning_rate": 1.893592233009709e-06,
      "loss": 0.1406,
      "step": 349680
    },
    {
      "epoch": 135.8,
      "learning_rate": 1.8930744336569582e-06,
      "loss": 0.0761,
      "step": 349690
    },
    {
      "epoch": 135.81,
      "learning_rate": 1.8925566343042074e-06,
      "loss": 0.0693,
      "step": 349700
    },
    {
      "epoch": 135.81,
      "learning_rate": 1.8920388349514565e-06,
      "loss": 0.0006,
      "step": 349710
    },
    {
      "epoch": 135.81,
      "learning_rate": 1.8915210355987057e-06,
      "loss": 0.0521,
      "step": 349720
    },
    {
      "epoch": 135.82,
      "learning_rate": 1.891003236245955e-06,
      "loss": 0.0614,
      "step": 349730
    },
    {
      "epoch": 135.82,
      "learning_rate": 1.8904854368932041e-06,
      "loss": 0.0173,
      "step": 349740
    },
    {
      "epoch": 135.83,
      "learning_rate": 1.8899676375404533e-06,
      "loss": 0.0007,
      "step": 349750
    },
    {
      "epoch": 135.83,
      "learning_rate": 1.8894498381877025e-06,
      "loss": 0.0096,
      "step": 349760
    },
    {
      "epoch": 135.83,
      "learning_rate": 1.8889320388349517e-06,
      "loss": 0.0198,
      "step": 349770
    },
    {
      "epoch": 135.84,
      "learning_rate": 1.888414239482201e-06,
      "loss": 0.0405,
      "step": 349780
    },
    {
      "epoch": 135.84,
      "learning_rate": 1.88789644012945e-06,
      "loss": 0.0196,
      "step": 349790
    },
    {
      "epoch": 135.84,
      "learning_rate": 1.887378640776699e-06,
      "loss": 0.0085,
      "step": 349800
    },
    {
      "epoch": 135.85,
      "learning_rate": 1.8868608414239483e-06,
      "loss": 0.0487,
      "step": 349810
    },
    {
      "epoch": 135.85,
      "learning_rate": 1.8863430420711975e-06,
      "loss": 0.0835,
      "step": 349820
    },
    {
      "epoch": 135.86,
      "learning_rate": 1.8858252427184467e-06,
      "loss": 0.0807,
      "step": 349830
    },
    {
      "epoch": 135.86,
      "learning_rate": 1.8853074433656959e-06,
      "loss": 0.0009,
      "step": 349840
    },
    {
      "epoch": 135.86,
      "learning_rate": 1.884789644012945e-06,
      "loss": 0.0252,
      "step": 349850
    },
    {
      "epoch": 135.87,
      "learning_rate": 1.8842718446601943e-06,
      "loss": 0.0328,
      "step": 349860
    },
    {
      "epoch": 135.87,
      "learning_rate": 1.8837540453074434e-06,
      "loss": 0.0257,
      "step": 349870
    },
    {
      "epoch": 135.88,
      "learning_rate": 1.8832362459546926e-06,
      "loss": 0.0361,
      "step": 349880
    },
    {
      "epoch": 135.88,
      "learning_rate": 1.8827184466019418e-06,
      "loss": 0.0007,
      "step": 349890
    },
    {
      "epoch": 135.88,
      "learning_rate": 1.882200647249191e-06,
      "loss": 0.0715,
      "step": 349900
    },
    {
      "epoch": 135.89,
      "learning_rate": 1.8816828478964402e-06,
      "loss": 0.1279,
      "step": 349910
    },
    {
      "epoch": 135.89,
      "learning_rate": 1.8811650485436894e-06,
      "loss": 0.0352,
      "step": 349920
    },
    {
      "epoch": 135.9,
      "learning_rate": 1.8806472491909386e-06,
      "loss": 0.0564,
      "step": 349930
    },
    {
      "epoch": 135.9,
      "learning_rate": 1.8801294498381878e-06,
      "loss": 0.1019,
      "step": 349940
    },
    {
      "epoch": 135.9,
      "learning_rate": 1.879611650485437e-06,
      "loss": 0.0001,
      "step": 349950
    },
    {
      "epoch": 135.91,
      "learning_rate": 1.8790938511326862e-06,
      "loss": 0.0002,
      "step": 349960
    },
    {
      "epoch": 135.91,
      "learning_rate": 1.8785760517799354e-06,
      "loss": 0.0193,
      "step": 349970
    },
    {
      "epoch": 135.91,
      "learning_rate": 1.8780582524271846e-06,
      "loss": 0.0002,
      "step": 349980
    },
    {
      "epoch": 135.92,
      "learning_rate": 1.8775404530744338e-06,
      "loss": 0.0364,
      "step": 349990
    },
    {
      "epoch": 135.92,
      "learning_rate": 1.877022653721683e-06,
      "loss": 0.0558,
      "step": 350000
    },
    {
      "epoch": 135.93,
      "learning_rate": 1.8765048543689322e-06,
      "loss": 0.0086,
      "step": 350010
    },
    {
      "epoch": 135.93,
      "learning_rate": 1.8759870550161814e-06,
      "loss": 0.015,
      "step": 350020
    },
    {
      "epoch": 135.93,
      "learning_rate": 1.8754692556634306e-06,
      "loss": 0.0229,
      "step": 350030
    },
    {
      "epoch": 135.94,
      "learning_rate": 1.8749514563106798e-06,
      "loss": 0.1389,
      "step": 350040
    },
    {
      "epoch": 135.94,
      "learning_rate": 1.8744336569579287e-06,
      "loss": 0.0347,
      "step": 350050
    },
    {
      "epoch": 135.95,
      "learning_rate": 1.873915857605178e-06,
      "loss": 0.044,
      "step": 350060
    },
    {
      "epoch": 135.95,
      "learning_rate": 1.8733980582524271e-06,
      "loss": 0.0577,
      "step": 350070
    },
    {
      "epoch": 135.95,
      "learning_rate": 1.8728802588996763e-06,
      "loss": 0.0403,
      "step": 350080
    },
    {
      "epoch": 135.96,
      "learning_rate": 1.8723624595469255e-06,
      "loss": 0.0088,
      "step": 350090
    },
    {
      "epoch": 135.96,
      "learning_rate": 1.8718446601941751e-06,
      "loss": 0.0019,
      "step": 350100
    },
    {
      "epoch": 135.97,
      "learning_rate": 1.8713268608414243e-06,
      "loss": 0.0255,
      "step": 350110
    },
    {
      "epoch": 135.97,
      "learning_rate": 1.8708090614886735e-06,
      "loss": 0.0742,
      "step": 350120
    },
    {
      "epoch": 135.97,
      "learning_rate": 1.8702912621359225e-06,
      "loss": 0.0009,
      "step": 350130
    },
    {
      "epoch": 135.98,
      "learning_rate": 1.8697734627831717e-06,
      "loss": 0.0214,
      "step": 350140
    },
    {
      "epoch": 135.98,
      "learning_rate": 1.8692556634304209e-06,
      "loss": 0.0335,
      "step": 350150
    },
    {
      "epoch": 135.98,
      "learning_rate": 1.86873786407767e-06,
      "loss": 0.0029,
      "step": 350160
    },
    {
      "epoch": 135.99,
      "learning_rate": 1.8682200647249193e-06,
      "loss": 0.0541,
      "step": 350170
    },
    {
      "epoch": 135.99,
      "learning_rate": 1.8677022653721685e-06,
      "loss": 0.0318,
      "step": 350180
    },
    {
      "epoch": 136.0,
      "learning_rate": 1.8671844660194177e-06,
      "loss": 0.0468,
      "step": 350190
    },
    {
      "epoch": 136.0,
      "learning_rate": 1.8666666666666669e-06,
      "loss": 0.015,
      "step": 350200
    },
    {
      "epoch": 136.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.3846341073513031,
      "eval_runtime": 8.1975,
      "eval_samples_per_second": 443.428,
      "eval_steps_per_second": 55.505,
      "step": 350200
    },
    {
      "epoch": 136.0,
      "learning_rate": 1.866148867313916e-06,
      "loss": 0.0001,
      "step": 350210
    },
    {
      "epoch": 136.01,
      "learning_rate": 1.8656310679611653e-06,
      "loss": 0.1259,
      "step": 350220
    },
    {
      "epoch": 136.01,
      "learning_rate": 1.8651132686084144e-06,
      "loss": 0.0715,
      "step": 350230
    },
    {
      "epoch": 136.02,
      "learning_rate": 1.8645954692556636e-06,
      "loss": 0.0381,
      "step": 350240
    },
    {
      "epoch": 136.02,
      "learning_rate": 1.8640776699029128e-06,
      "loss": 0.0181,
      "step": 350250
    },
    {
      "epoch": 136.02,
      "learning_rate": 1.863559870550162e-06,
      "loss": 0.0593,
      "step": 350260
    },
    {
      "epoch": 136.03,
      "learning_rate": 1.8630420711974112e-06,
      "loss": 0.0561,
      "step": 350270
    },
    {
      "epoch": 136.03,
      "learning_rate": 1.8625242718446604e-06,
      "loss": 0.0094,
      "step": 350280
    },
    {
      "epoch": 136.03,
      "learning_rate": 1.8620064724919096e-06,
      "loss": 0.0001,
      "step": 350290
    },
    {
      "epoch": 136.04,
      "learning_rate": 1.8614886731391588e-06,
      "loss": 0.0909,
      "step": 350300
    },
    {
      "epoch": 136.04,
      "learning_rate": 1.860970873786408e-06,
      "loss": 0.0155,
      "step": 350310
    },
    {
      "epoch": 136.05,
      "learning_rate": 1.8604530744336572e-06,
      "loss": 0.0398,
      "step": 350320
    },
    {
      "epoch": 136.05,
      "learning_rate": 1.8599352750809064e-06,
      "loss": 0.0733,
      "step": 350330
    },
    {
      "epoch": 136.05,
      "learning_rate": 1.8594174757281556e-06,
      "loss": 0.028,
      "step": 350340
    },
    {
      "epoch": 136.06,
      "learning_rate": 1.8588996763754048e-06,
      "loss": 0.1093,
      "step": 350350
    },
    {
      "epoch": 136.06,
      "learning_rate": 1.858381877022654e-06,
      "loss": 0.0181,
      "step": 350360
    },
    {
      "epoch": 136.07,
      "learning_rate": 1.857864077669903e-06,
      "loss": 0.0577,
      "step": 350370
    },
    {
      "epoch": 136.07,
      "learning_rate": 1.8573462783171522e-06,
      "loss": 0.0449,
      "step": 350380
    },
    {
      "epoch": 136.07,
      "learning_rate": 1.8568284789644013e-06,
      "loss": 0.0064,
      "step": 350390
    },
    {
      "epoch": 136.08,
      "learning_rate": 1.8563106796116505e-06,
      "loss": 0.0759,
      "step": 350400
    },
    {
      "epoch": 136.08,
      "learning_rate": 1.8557928802588997e-06,
      "loss": 0.0769,
      "step": 350410
    },
    {
      "epoch": 136.09,
      "learning_rate": 1.855275080906149e-06,
      "loss": 0.0001,
      "step": 350420
    },
    {
      "epoch": 136.09,
      "learning_rate": 1.8547572815533981e-06,
      "loss": 0.0614,
      "step": 350430
    },
    {
      "epoch": 136.09,
      "learning_rate": 1.8542394822006473e-06,
      "loss": 0.001,
      "step": 350440
    },
    {
      "epoch": 136.1,
      "learning_rate": 1.8537216828478965e-06,
      "loss": 0.0113,
      "step": 350450
    },
    {
      "epoch": 136.1,
      "learning_rate": 1.8532038834951457e-06,
      "loss": 0.0414,
      "step": 350460
    },
    {
      "epoch": 136.1,
      "learning_rate": 1.852686084142395e-06,
      "loss": 0.0427,
      "step": 350470
    },
    {
      "epoch": 136.11,
      "learning_rate": 1.852168284789644e-06,
      "loss": 0.0455,
      "step": 350480
    },
    {
      "epoch": 136.11,
      "learning_rate": 1.8516504854368933e-06,
      "loss": 0.0003,
      "step": 350490
    },
    {
      "epoch": 136.12,
      "learning_rate": 1.8511326860841425e-06,
      "loss": 0.017,
      "step": 350500
    },
    {
      "epoch": 136.12,
      "learning_rate": 1.8506148867313917e-06,
      "loss": 0.0003,
      "step": 350510
    },
    {
      "epoch": 136.12,
      "learning_rate": 1.8500970873786409e-06,
      "loss": 0.0008,
      "step": 350520
    },
    {
      "epoch": 136.13,
      "learning_rate": 1.84957928802589e-06,
      "loss": 0.0257,
      "step": 350530
    },
    {
      "epoch": 136.13,
      "learning_rate": 1.8490614886731393e-06,
      "loss": 0.0572,
      "step": 350540
    },
    {
      "epoch": 136.14,
      "learning_rate": 1.8485436893203885e-06,
      "loss": 0.0005,
      "step": 350550
    },
    {
      "epoch": 136.14,
      "learning_rate": 1.8480258899676377e-06,
      "loss": 0.0155,
      "step": 350560
    },
    {
      "epoch": 136.14,
      "learning_rate": 1.8475080906148869e-06,
      "loss": 0.0508,
      "step": 350570
    },
    {
      "epoch": 136.15,
      "learning_rate": 1.846990291262136e-06,
      "loss": 0.0002,
      "step": 350580
    },
    {
      "epoch": 136.15,
      "learning_rate": 1.8464724919093852e-06,
      "loss": 0.002,
      "step": 350590
    },
    {
      "epoch": 136.16,
      "learning_rate": 1.8459546925566344e-06,
      "loss": 0.0001,
      "step": 350600
    },
    {
      "epoch": 136.16,
      "learning_rate": 1.8454368932038836e-06,
      "loss": 0.0003,
      "step": 350610
    },
    {
      "epoch": 136.16,
      "learning_rate": 1.8449190938511326e-06,
      "loss": 0.038,
      "step": 350620
    },
    {
      "epoch": 136.17,
      "learning_rate": 1.8444012944983818e-06,
      "loss": 0.0411,
      "step": 350630
    },
    {
      "epoch": 136.17,
      "learning_rate": 1.843883495145631e-06,
      "loss": 0.0166,
      "step": 350640
    },
    {
      "epoch": 136.17,
      "learning_rate": 1.8433656957928802e-06,
      "loss": 0.0187,
      "step": 350650
    },
    {
      "epoch": 136.18,
      "learning_rate": 1.8428478964401298e-06,
      "loss": 0.0376,
      "step": 350660
    },
    {
      "epoch": 136.18,
      "learning_rate": 1.842330097087379e-06,
      "loss": 0.0758,
      "step": 350670
    },
    {
      "epoch": 136.19,
      "learning_rate": 1.8418122977346282e-06,
      "loss": 0.0797,
      "step": 350680
    },
    {
      "epoch": 136.19,
      "learning_rate": 1.8412944983818774e-06,
      "loss": 0.1447,
      "step": 350690
    },
    {
      "epoch": 136.19,
      "learning_rate": 1.8407766990291264e-06,
      "loss": 0.0611,
      "step": 350700
    },
    {
      "epoch": 136.2,
      "learning_rate": 1.8402588996763756e-06,
      "loss": 0.0143,
      "step": 350710
    },
    {
      "epoch": 136.2,
      "learning_rate": 1.8397411003236248e-06,
      "loss": 0.0616,
      "step": 350720
    },
    {
      "epoch": 136.21,
      "learning_rate": 1.839223300970874e-06,
      "loss": 0.0726,
      "step": 350730
    },
    {
      "epoch": 136.21,
      "learning_rate": 1.8387055016181232e-06,
      "loss": 0.0001,
      "step": 350740
    },
    {
      "epoch": 136.21,
      "learning_rate": 1.8381877022653724e-06,
      "loss": 0.0377,
      "step": 350750
    },
    {
      "epoch": 136.22,
      "learning_rate": 1.8376699029126215e-06,
      "loss": 0.018,
      "step": 350760
    },
    {
      "epoch": 136.22,
      "learning_rate": 1.8371521035598707e-06,
      "loss": 0.0002,
      "step": 350770
    },
    {
      "epoch": 136.23,
      "learning_rate": 1.83663430420712e-06,
      "loss": 0.0255,
      "step": 350780
    },
    {
      "epoch": 136.23,
      "learning_rate": 1.8361165048543691e-06,
      "loss": 0.0795,
      "step": 350790
    },
    {
      "epoch": 136.23,
      "learning_rate": 1.8355987055016183e-06,
      "loss": 0.009,
      "step": 350800
    },
    {
      "epoch": 136.24,
      "learning_rate": 1.8350809061488675e-06,
      "loss": 0.0007,
      "step": 350810
    },
    {
      "epoch": 136.24,
      "learning_rate": 1.8345631067961167e-06,
      "loss": 0.0243,
      "step": 350820
    },
    {
      "epoch": 136.24,
      "learning_rate": 1.834045307443366e-06,
      "loss": 0.0101,
      "step": 350830
    },
    {
      "epoch": 136.25,
      "learning_rate": 1.833527508090615e-06,
      "loss": 0.0054,
      "step": 350840
    },
    {
      "epoch": 136.25,
      "learning_rate": 1.8330097087378643e-06,
      "loss": 0.004,
      "step": 350850
    },
    {
      "epoch": 136.26,
      "learning_rate": 1.8324919093851135e-06,
      "loss": 0.0476,
      "step": 350860
    },
    {
      "epoch": 136.26,
      "learning_rate": 1.8319741100323627e-06,
      "loss": 0.0334,
      "step": 350870
    },
    {
      "epoch": 136.26,
      "learning_rate": 1.8314563106796119e-06,
      "loss": 0.0005,
      "step": 350880
    },
    {
      "epoch": 136.27,
      "learning_rate": 1.830938511326861e-06,
      "loss": 0.0062,
      "step": 350890
    },
    {
      "epoch": 136.27,
      "learning_rate": 1.8304207119741103e-06,
      "loss": 0.0001,
      "step": 350900
    },
    {
      "epoch": 136.28,
      "learning_rate": 1.8299029126213595e-06,
      "loss": 0.0229,
      "step": 350910
    },
    {
      "epoch": 136.28,
      "learning_rate": 1.8293851132686087e-06,
      "loss": 0.0454,
      "step": 350920
    },
    {
      "epoch": 136.28,
      "learning_rate": 1.8288673139158579e-06,
      "loss": 0.256,
      "step": 350930
    },
    {
      "epoch": 136.29,
      "learning_rate": 1.828349514563107e-06,
      "loss": 0.0413,
      "step": 350940
    },
    {
      "epoch": 136.29,
      "learning_rate": 1.827831715210356e-06,
      "loss": 0.0157,
      "step": 350950
    },
    {
      "epoch": 136.3,
      "learning_rate": 1.8273139158576052e-06,
      "loss": 0.0468,
      "step": 350960
    },
    {
      "epoch": 136.3,
      "learning_rate": 1.8267961165048544e-06,
      "loss": 0.0877,
      "step": 350970
    },
    {
      "epoch": 136.3,
      "learning_rate": 1.8262783171521036e-06,
      "loss": 0.1043,
      "step": 350980
    },
    {
      "epoch": 136.31,
      "learning_rate": 1.8257605177993528e-06,
      "loss": 0.0731,
      "step": 350990
    },
    {
      "epoch": 136.31,
      "learning_rate": 1.825242718446602e-06,
      "loss": 0.0005,
      "step": 351000
    },
    {
      "epoch": 136.31,
      "learning_rate": 1.8247249190938512e-06,
      "loss": 0.0174,
      "step": 351010
    },
    {
      "epoch": 136.32,
      "learning_rate": 1.8242071197411004e-06,
      "loss": 0.1207,
      "step": 351020
    },
    {
      "epoch": 136.32,
      "learning_rate": 1.8236893203883496e-06,
      "loss": 0.0281,
      "step": 351030
    },
    {
      "epoch": 136.33,
      "learning_rate": 1.8231715210355988e-06,
      "loss": 0.0994,
      "step": 351040
    },
    {
      "epoch": 136.33,
      "learning_rate": 1.822653721682848e-06,
      "loss": 0.0421,
      "step": 351050
    },
    {
      "epoch": 136.33,
      "learning_rate": 1.8221359223300972e-06,
      "loss": 0.0006,
      "step": 351060
    },
    {
      "epoch": 136.34,
      "learning_rate": 1.8216181229773464e-06,
      "loss": 0.0201,
      "step": 351070
    },
    {
      "epoch": 136.34,
      "learning_rate": 1.8211003236245956e-06,
      "loss": 0.0029,
      "step": 351080
    },
    {
      "epoch": 136.35,
      "learning_rate": 1.8205825242718448e-06,
      "loss": 0.0041,
      "step": 351090
    },
    {
      "epoch": 136.35,
      "learning_rate": 1.820064724919094e-06,
      "loss": 0.1635,
      "step": 351100
    },
    {
      "epoch": 136.35,
      "learning_rate": 1.8195469255663431e-06,
      "loss": 0.0198,
      "step": 351110
    },
    {
      "epoch": 136.36,
      "learning_rate": 1.8190291262135923e-06,
      "loss": 0.0089,
      "step": 351120
    },
    {
      "epoch": 136.36,
      "learning_rate": 1.8185113268608415e-06,
      "loss": 0.0094,
      "step": 351130
    },
    {
      "epoch": 136.37,
      "learning_rate": 1.8179935275080907e-06,
      "loss": 0.0002,
      "step": 351140
    },
    {
      "epoch": 136.37,
      "learning_rate": 1.81747572815534e-06,
      "loss": 0.0348,
      "step": 351150
    },
    {
      "epoch": 136.37,
      "learning_rate": 1.8169579288025891e-06,
      "loss": 0.0715,
      "step": 351160
    },
    {
      "epoch": 136.38,
      "learning_rate": 1.8164401294498383e-06,
      "loss": 0.0411,
      "step": 351170
    },
    {
      "epoch": 136.38,
      "learning_rate": 1.8159223300970875e-06,
      "loss": 0.0008,
      "step": 351180
    },
    {
      "epoch": 136.38,
      "learning_rate": 1.8154045307443365e-06,
      "loss": 0.0518,
      "step": 351190
    },
    {
      "epoch": 136.39,
      "learning_rate": 1.8148867313915857e-06,
      "loss": 0.1041,
      "step": 351200
    },
    {
      "epoch": 136.39,
      "learning_rate": 1.8143689320388349e-06,
      "loss": 0.0365,
      "step": 351210
    },
    {
      "epoch": 136.4,
      "learning_rate": 1.8138511326860845e-06,
      "loss": 0.0707,
      "step": 351220
    },
    {
      "epoch": 136.4,
      "learning_rate": 1.8133333333333337e-06,
      "loss": 0.0189,
      "step": 351230
    },
    {
      "epoch": 136.4,
      "learning_rate": 1.8128155339805829e-06,
      "loss": 0.0692,
      "step": 351240
    },
    {
      "epoch": 136.41,
      "learning_rate": 1.812297734627832e-06,
      "loss": 0.0313,
      "step": 351250
    },
    {
      "epoch": 136.41,
      "learning_rate": 1.8117799352750813e-06,
      "loss": 0.0679,
      "step": 351260
    },
    {
      "epoch": 136.42,
      "learning_rate": 1.8112621359223303e-06,
      "loss": 0.0001,
      "step": 351270
    },
    {
      "epoch": 136.42,
      "learning_rate": 1.8107443365695794e-06,
      "loss": 0.0214,
      "step": 351280
    },
    {
      "epoch": 136.42,
      "learning_rate": 1.8102265372168286e-06,
      "loss": 0.0094,
      "step": 351290
    },
    {
      "epoch": 136.43,
      "learning_rate": 1.8097087378640778e-06,
      "loss": 0.1306,
      "step": 351300
    },
    {
      "epoch": 136.43,
      "learning_rate": 1.809190938511327e-06,
      "loss": 0.0101,
      "step": 351310
    },
    {
      "epoch": 136.43,
      "learning_rate": 1.8086731391585762e-06,
      "loss": 0.0261,
      "step": 351320
    },
    {
      "epoch": 136.44,
      "learning_rate": 1.8081553398058254e-06,
      "loss": 0.0599,
      "step": 351330
    },
    {
      "epoch": 136.44,
      "learning_rate": 1.8076375404530746e-06,
      "loss": 0.0003,
      "step": 351340
    },
    {
      "epoch": 136.45,
      "learning_rate": 1.8071197411003238e-06,
      "loss": 0.0414,
      "step": 351350
    },
    {
      "epoch": 136.45,
      "learning_rate": 1.806601941747573e-06,
      "loss": 0.1387,
      "step": 351360
    },
    {
      "epoch": 136.45,
      "learning_rate": 1.8060841423948222e-06,
      "loss": 0.067,
      "step": 351370
    },
    {
      "epoch": 136.46,
      "learning_rate": 1.8055663430420714e-06,
      "loss": 0.0024,
      "step": 351380
    },
    {
      "epoch": 136.46,
      "learning_rate": 1.8050485436893206e-06,
      "loss": 0.0167,
      "step": 351390
    },
    {
      "epoch": 136.47,
      "learning_rate": 1.8045307443365698e-06,
      "loss": 0.0656,
      "step": 351400
    },
    {
      "epoch": 136.47,
      "learning_rate": 1.804012944983819e-06,
      "loss": 0.0157,
      "step": 351410
    },
    {
      "epoch": 136.47,
      "learning_rate": 1.8034951456310682e-06,
      "loss": 0.0313,
      "step": 351420
    },
    {
      "epoch": 136.48,
      "learning_rate": 1.8029773462783174e-06,
      "loss": 0.1502,
      "step": 351430
    },
    {
      "epoch": 136.48,
      "learning_rate": 1.8024595469255666e-06,
      "loss": 0.0565,
      "step": 351440
    },
    {
      "epoch": 136.49,
      "learning_rate": 1.8019417475728158e-06,
      "loss": 0.0114,
      "step": 351450
    },
    {
      "epoch": 136.49,
      "learning_rate": 1.801423948220065e-06,
      "loss": 0.0077,
      "step": 351460
    },
    {
      "epoch": 136.49,
      "learning_rate": 1.8009061488673141e-06,
      "loss": 0.0336,
      "step": 351470
    },
    {
      "epoch": 136.5,
      "learning_rate": 1.8003883495145633e-06,
      "loss": 0.0001,
      "step": 351480
    },
    {
      "epoch": 136.5,
      "learning_rate": 1.7998705501618125e-06,
      "loss": 0.1271,
      "step": 351490
    },
    {
      "epoch": 136.5,
      "learning_rate": 1.7993527508090617e-06,
      "loss": 0.0207,
      "step": 351500
    },
    {
      "epoch": 136.51,
      "learning_rate": 1.798834951456311e-06,
      "loss": 0.1185,
      "step": 351510
    },
    {
      "epoch": 136.51,
      "learning_rate": 1.79831715210356e-06,
      "loss": 0.0005,
      "step": 351520
    },
    {
      "epoch": 136.52,
      "learning_rate": 1.797799352750809e-06,
      "loss": 0.0506,
      "step": 351530
    },
    {
      "epoch": 136.52,
      "learning_rate": 1.7972815533980583e-06,
      "loss": 0.0149,
      "step": 351540
    },
    {
      "epoch": 136.52,
      "learning_rate": 1.7967637540453075e-06,
      "loss": 0.0865,
      "step": 351550
    },
    {
      "epoch": 136.53,
      "learning_rate": 1.7962459546925567e-06,
      "loss": 0.0014,
      "step": 351560
    },
    {
      "epoch": 136.53,
      "learning_rate": 1.7957281553398059e-06,
      "loss": 0.0012,
      "step": 351570
    },
    {
      "epoch": 136.54,
      "learning_rate": 1.795210355987055e-06,
      "loss": 0.0163,
      "step": 351580
    },
    {
      "epoch": 136.54,
      "learning_rate": 1.7946925566343043e-06,
      "loss": 0.0014,
      "step": 351590
    },
    {
      "epoch": 136.54,
      "learning_rate": 1.7941747572815535e-06,
      "loss": 0.0006,
      "step": 351600
    },
    {
      "epoch": 136.55,
      "learning_rate": 1.7936569579288027e-06,
      "loss": 0.001,
      "step": 351610
    },
    {
      "epoch": 136.55,
      "learning_rate": 1.7931391585760519e-06,
      "loss": 0.053,
      "step": 351620
    },
    {
      "epoch": 136.56,
      "learning_rate": 1.792621359223301e-06,
      "loss": 0.0802,
      "step": 351630
    },
    {
      "epoch": 136.56,
      "learning_rate": 1.7921035598705502e-06,
      "loss": 0.065,
      "step": 351640
    },
    {
      "epoch": 136.56,
      "learning_rate": 1.7915857605177994e-06,
      "loss": 0.0223,
      "step": 351650
    },
    {
      "epoch": 136.57,
      "learning_rate": 1.7910679611650486e-06,
      "loss": 0.0516,
      "step": 351660
    },
    {
      "epoch": 136.57,
      "learning_rate": 1.7905501618122978e-06,
      "loss": 0.0793,
      "step": 351670
    },
    {
      "epoch": 136.57,
      "learning_rate": 1.790032362459547e-06,
      "loss": 0.0048,
      "step": 351680
    },
    {
      "epoch": 136.58,
      "learning_rate": 1.7895145631067962e-06,
      "loss": 0.0522,
      "step": 351690
    },
    {
      "epoch": 136.58,
      "learning_rate": 1.7889967637540454e-06,
      "loss": 0.0392,
      "step": 351700
    },
    {
      "epoch": 136.59,
      "learning_rate": 1.7884789644012946e-06,
      "loss": 0.0169,
      "step": 351710
    },
    {
      "epoch": 136.59,
      "learning_rate": 1.7879611650485438e-06,
      "loss": 0.0023,
      "step": 351720
    },
    {
      "epoch": 136.59,
      "learning_rate": 1.787443365695793e-06,
      "loss": 0.1591,
      "step": 351730
    },
    {
      "epoch": 136.6,
      "learning_rate": 1.7869255663430422e-06,
      "loss": 0.0309,
      "step": 351740
    },
    {
      "epoch": 136.6,
      "learning_rate": 1.7864077669902914e-06,
      "loss": 0.0171,
      "step": 351750
    },
    {
      "epoch": 136.61,
      "learning_rate": 1.7858899676375404e-06,
      "loss": 0.078,
      "step": 351760
    },
    {
      "epoch": 136.61,
      "learning_rate": 1.7853721682847896e-06,
      "loss": 0.0004,
      "step": 351770
    },
    {
      "epoch": 136.61,
      "learning_rate": 1.7848543689320392e-06,
      "loss": 0.0631,
      "step": 351780
    },
    {
      "epoch": 136.62,
      "learning_rate": 1.7843365695792884e-06,
      "loss": 0.0341,
      "step": 351790
    },
    {
      "epoch": 136.62,
      "learning_rate": 1.7838187702265376e-06,
      "loss": 0.0214,
      "step": 351800
    },
    {
      "epoch": 136.63,
      "learning_rate": 1.7833009708737868e-06,
      "loss": 0.0612,
      "step": 351810
    },
    {
      "epoch": 136.63,
      "learning_rate": 1.782783171521036e-06,
      "loss": 0.0119,
      "step": 351820
    },
    {
      "epoch": 136.63,
      "learning_rate": 1.7822653721682851e-06,
      "loss": 0.0082,
      "step": 351830
    },
    {
      "epoch": 136.64,
      "learning_rate": 1.7817475728155341e-06,
      "loss": 0.0746,
      "step": 351840
    },
    {
      "epoch": 136.64,
      "learning_rate": 1.7812297734627833e-06,
      "loss": 0.0231,
      "step": 351850
    },
    {
      "epoch": 136.64,
      "learning_rate": 1.7807119741100325e-06,
      "loss": 0.0176,
      "step": 351860
    },
    {
      "epoch": 136.65,
      "learning_rate": 1.7801941747572817e-06,
      "loss": 0.0311,
      "step": 351870
    },
    {
      "epoch": 136.65,
      "learning_rate": 1.779676375404531e-06,
      "loss": 0.0011,
      "step": 351880
    },
    {
      "epoch": 136.66,
      "learning_rate": 1.77915857605178e-06,
      "loss": 0.0001,
      "step": 351890
    },
    {
      "epoch": 136.66,
      "learning_rate": 1.7786407766990293e-06,
      "loss": 0.0324,
      "step": 351900
    },
    {
      "epoch": 136.66,
      "learning_rate": 1.7781229773462785e-06,
      "loss": 0.0752,
      "step": 351910
    },
    {
      "epoch": 136.67,
      "learning_rate": 1.7776051779935277e-06,
      "loss": 0.0601,
      "step": 351920
    },
    {
      "epoch": 136.67,
      "learning_rate": 1.7770873786407769e-06,
      "loss": 0.0863,
      "step": 351930
    },
    {
      "epoch": 136.68,
      "learning_rate": 1.776569579288026e-06,
      "loss": 0.0935,
      "step": 351940
    },
    {
      "epoch": 136.68,
      "learning_rate": 1.7760517799352753e-06,
      "loss": 0.0537,
      "step": 351950
    },
    {
      "epoch": 136.68,
      "learning_rate": 1.7755339805825245e-06,
      "loss": 0.093,
      "step": 351960
    },
    {
      "epoch": 136.69,
      "learning_rate": 1.7750161812297737e-06,
      "loss": 0.0826,
      "step": 351970
    },
    {
      "epoch": 136.69,
      "learning_rate": 1.7744983818770229e-06,
      "loss": 0.0104,
      "step": 351980
    },
    {
      "epoch": 136.7,
      "learning_rate": 1.773980582524272e-06,
      "loss": 0.0718,
      "step": 351990
    },
    {
      "epoch": 136.7,
      "learning_rate": 1.7734627831715212e-06,
      "loss": 0.0507,
      "step": 352000
    },
    {
      "epoch": 136.7,
      "learning_rate": 1.7729449838187704e-06,
      "loss": 0.0005,
      "step": 352010
    },
    {
      "epoch": 136.71,
      "learning_rate": 1.7724271844660196e-06,
      "loss": 0.1083,
      "step": 352020
    },
    {
      "epoch": 136.71,
      "learning_rate": 1.7719093851132688e-06,
      "loss": 0.0169,
      "step": 352030
    },
    {
      "epoch": 136.71,
      "learning_rate": 1.771391585760518e-06,
      "loss": 0.0202,
      "step": 352040
    },
    {
      "epoch": 136.72,
      "learning_rate": 1.7708737864077672e-06,
      "loss": 0.0527,
      "step": 352050
    },
    {
      "epoch": 136.72,
      "learning_rate": 1.7703559870550164e-06,
      "loss": 0.0543,
      "step": 352060
    },
    {
      "epoch": 136.73,
      "learning_rate": 1.7698381877022656e-06,
      "loss": 0.0154,
      "step": 352070
    },
    {
      "epoch": 136.73,
      "learning_rate": 1.7693203883495148e-06,
      "loss": 0.0159,
      "step": 352080
    },
    {
      "epoch": 136.73,
      "learning_rate": 1.7688025889967638e-06,
      "loss": 0.0786,
      "step": 352090
    },
    {
      "epoch": 136.74,
      "learning_rate": 1.768284789644013e-06,
      "loss": 0.0104,
      "step": 352100
    },
    {
      "epoch": 136.74,
      "learning_rate": 1.7677669902912622e-06,
      "loss": 0.0662,
      "step": 352110
    },
    {
      "epoch": 136.75,
      "learning_rate": 1.7672491909385114e-06,
      "loss": 0.0548,
      "step": 352120
    },
    {
      "epoch": 136.75,
      "learning_rate": 1.7667313915857606e-06,
      "loss": 0.0133,
      "step": 352130
    },
    {
      "epoch": 136.75,
      "learning_rate": 1.7662135922330098e-06,
      "loss": 0.0104,
      "step": 352140
    },
    {
      "epoch": 136.76,
      "learning_rate": 1.765695792880259e-06,
      "loss": 0.0037,
      "step": 352150
    },
    {
      "epoch": 136.76,
      "learning_rate": 1.7651779935275081e-06,
      "loss": 0.085,
      "step": 352160
    },
    {
      "epoch": 136.77,
      "learning_rate": 1.7646601941747573e-06,
      "loss": 0.0001,
      "step": 352170
    },
    {
      "epoch": 136.77,
      "learning_rate": 1.7641423948220065e-06,
      "loss": 0.0739,
      "step": 352180
    },
    {
      "epoch": 136.77,
      "learning_rate": 1.7636245954692557e-06,
      "loss": 0.0589,
      "step": 352190
    },
    {
      "epoch": 136.78,
      "learning_rate": 1.763106796116505e-06,
      "loss": 0.0233,
      "step": 352200
    },
    {
      "epoch": 136.78,
      "learning_rate": 1.7625889967637541e-06,
      "loss": 0.0217,
      "step": 352210
    },
    {
      "epoch": 136.78,
      "learning_rate": 1.7620711974110033e-06,
      "loss": 0.0347,
      "step": 352220
    },
    {
      "epoch": 136.79,
      "learning_rate": 1.7615533980582525e-06,
      "loss": 0.05,
      "step": 352230
    },
    {
      "epoch": 136.79,
      "learning_rate": 1.7610355987055017e-06,
      "loss": 0.0313,
      "step": 352240
    },
    {
      "epoch": 136.8,
      "learning_rate": 1.760517799352751e-06,
      "loss": 0.0021,
      "step": 352250
    },
    {
      "epoch": 136.8,
      "learning_rate": 1.76e-06,
      "loss": 0.0716,
      "step": 352260
    },
    {
      "epoch": 136.8,
      "learning_rate": 1.7594822006472493e-06,
      "loss": 0.0652,
      "step": 352270
    },
    {
      "epoch": 136.81,
      "learning_rate": 1.7589644012944985e-06,
      "loss": 0.0455,
      "step": 352280
    },
    {
      "epoch": 136.81,
      "learning_rate": 1.7584466019417477e-06,
      "loss": 0.0334,
      "step": 352290
    },
    {
      "epoch": 136.82,
      "learning_rate": 1.7579288025889969e-06,
      "loss": 0.029,
      "step": 352300
    },
    {
      "epoch": 136.82,
      "learning_rate": 1.757411003236246e-06,
      "loss": 0.0005,
      "step": 352310
    },
    {
      "epoch": 136.82,
      "learning_rate": 1.7568932038834953e-06,
      "loss": 0.016,
      "step": 352320
    },
    {
      "epoch": 136.83,
      "learning_rate": 1.7563754045307442e-06,
      "loss": 0.04,
      "step": 352330
    },
    {
      "epoch": 136.83,
      "learning_rate": 1.7558576051779939e-06,
      "loss": 0.0359,
      "step": 352340
    },
    {
      "epoch": 136.83,
      "learning_rate": 1.755339805825243e-06,
      "loss": 0.1028,
      "step": 352350
    },
    {
      "epoch": 136.84,
      "learning_rate": 1.7548220064724922e-06,
      "loss": 0.0068,
      "step": 352360
    },
    {
      "epoch": 136.84,
      "learning_rate": 1.7543042071197414e-06,
      "loss": 0.017,
      "step": 352370
    },
    {
      "epoch": 136.85,
      "learning_rate": 1.7537864077669906e-06,
      "loss": 0.01,
      "step": 352380
    },
    {
      "epoch": 136.85,
      "learning_rate": 1.7532686084142398e-06,
      "loss": 0.0726,
      "step": 352390
    },
    {
      "epoch": 136.85,
      "learning_rate": 1.752750809061489e-06,
      "loss": 0.0878,
      "step": 352400
    },
    {
      "epoch": 136.86,
      "learning_rate": 1.752233009708738e-06,
      "loss": 0.0066,
      "step": 352410
    },
    {
      "epoch": 136.86,
      "learning_rate": 1.7517152103559872e-06,
      "loss": 0.0019,
      "step": 352420
    },
    {
      "epoch": 136.87,
      "learning_rate": 1.7511974110032364e-06,
      "loss": 0.0294,
      "step": 352430
    },
    {
      "epoch": 136.87,
      "learning_rate": 1.7506796116504856e-06,
      "loss": 0.089,
      "step": 352440
    },
    {
      "epoch": 136.87,
      "learning_rate": 1.7501618122977348e-06,
      "loss": 0.0341,
      "step": 352450
    },
    {
      "epoch": 136.88,
      "learning_rate": 1.749644012944984e-06,
      "loss": 0.0604,
      "step": 352460
    },
    {
      "epoch": 136.88,
      "learning_rate": 1.7491262135922332e-06,
      "loss": 0.0001,
      "step": 352470
    },
    {
      "epoch": 136.89,
      "learning_rate": 1.7486084142394824e-06,
      "loss": 0.0083,
      "step": 352480
    },
    {
      "epoch": 136.89,
      "learning_rate": 1.7480906148867316e-06,
      "loss": 0.1941,
      "step": 352490
    },
    {
      "epoch": 136.89,
      "learning_rate": 1.7475728155339808e-06,
      "loss": 0.0003,
      "step": 352500
    },
    {
      "epoch": 136.9,
      "learning_rate": 1.74705501618123e-06,
      "loss": 0.0507,
      "step": 352510
    },
    {
      "epoch": 136.9,
      "learning_rate": 1.7465372168284791e-06,
      "loss": 0.126,
      "step": 352520
    },
    {
      "epoch": 136.9,
      "learning_rate": 1.7460194174757283e-06,
      "loss": 0.1342,
      "step": 352530
    },
    {
      "epoch": 136.91,
      "learning_rate": 1.7455016181229775e-06,
      "loss": 0.0366,
      "step": 352540
    },
    {
      "epoch": 136.91,
      "learning_rate": 1.7449838187702267e-06,
      "loss": 0.0099,
      "step": 352550
    },
    {
      "epoch": 136.92,
      "learning_rate": 1.744466019417476e-06,
      "loss": 0.0387,
      "step": 352560
    },
    {
      "epoch": 136.92,
      "learning_rate": 1.7439482200647251e-06,
      "loss": 0.0822,
      "step": 352570
    },
    {
      "epoch": 136.92,
      "learning_rate": 1.7434304207119743e-06,
      "loss": 0.0804,
      "step": 352580
    },
    {
      "epoch": 136.93,
      "learning_rate": 1.7429126213592235e-06,
      "loss": 0.0865,
      "step": 352590
    },
    {
      "epoch": 136.93,
      "learning_rate": 1.7423948220064727e-06,
      "loss": 0.0401,
      "step": 352600
    },
    {
      "epoch": 136.94,
      "learning_rate": 1.741877022653722e-06,
      "loss": 0.0078,
      "step": 352610
    },
    {
      "epoch": 136.94,
      "learning_rate": 1.741359223300971e-06,
      "loss": 0.1411,
      "step": 352620
    },
    {
      "epoch": 136.94,
      "learning_rate": 1.7408414239482203e-06,
      "loss": 0.0197,
      "step": 352630
    },
    {
      "epoch": 136.95,
      "learning_rate": 1.7403236245954695e-06,
      "loss": 0.0096,
      "step": 352640
    },
    {
      "epoch": 136.95,
      "learning_rate": 1.7398058252427187e-06,
      "loss": 0.0853,
      "step": 352650
    },
    {
      "epoch": 136.96,
      "learning_rate": 1.7392880258899677e-06,
      "loss": 0.0235,
      "step": 352660
    },
    {
      "epoch": 136.96,
      "learning_rate": 1.7387702265372169e-06,
      "loss": 0.0772,
      "step": 352670
    },
    {
      "epoch": 136.96,
      "learning_rate": 1.738252427184466e-06,
      "loss": 0.0318,
      "step": 352680
    },
    {
      "epoch": 136.97,
      "learning_rate": 1.7377346278317152e-06,
      "loss": 0.0007,
      "step": 352690
    },
    {
      "epoch": 136.97,
      "learning_rate": 1.7372168284789644e-06,
      "loss": 0.0482,
      "step": 352700
    },
    {
      "epoch": 136.97,
      "learning_rate": 1.7366990291262136e-06,
      "loss": 0.0105,
      "step": 352710
    },
    {
      "epoch": 136.98,
      "learning_rate": 1.7361812297734628e-06,
      "loss": 0.0445,
      "step": 352720
    },
    {
      "epoch": 136.98,
      "learning_rate": 1.735663430420712e-06,
      "loss": 0.0354,
      "step": 352730
    },
    {
      "epoch": 136.99,
      "learning_rate": 1.7351456310679612e-06,
      "loss": 0.0185,
      "step": 352740
    },
    {
      "epoch": 136.99,
      "learning_rate": 1.7346278317152104e-06,
      "loss": 0.0873,
      "step": 352750
    },
    {
      "epoch": 136.99,
      "learning_rate": 1.7341100323624596e-06,
      "loss": 0.0061,
      "step": 352760
    },
    {
      "epoch": 137.0,
      "learning_rate": 1.7335922330097088e-06,
      "loss": 0.0307,
      "step": 352770
    },
    {
      "epoch": 137.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.38503599166870117,
      "eval_runtime": 8.175,
      "eval_samples_per_second": 444.649,
      "eval_steps_per_second": 55.658,
      "step": 352775
    },
    {
      "epoch": 137.0,
      "learning_rate": 1.733074433656958e-06,
      "loss": 0.1058,
      "step": 352780
    },
    {
      "epoch": 137.01,
      "learning_rate": 1.7325566343042072e-06,
      "loss": 0.0063,
      "step": 352790
    },
    {
      "epoch": 137.01,
      "learning_rate": 1.7320388349514564e-06,
      "loss": 0.0502,
      "step": 352800
    },
    {
      "epoch": 137.01,
      "learning_rate": 1.7315210355987056e-06,
      "loss": 0.0092,
      "step": 352810
    },
    {
      "epoch": 137.02,
      "learning_rate": 1.7310032362459548e-06,
      "loss": 0.0067,
      "step": 352820
    },
    {
      "epoch": 137.02,
      "learning_rate": 1.730485436893204e-06,
      "loss": 0.0391,
      "step": 352830
    },
    {
      "epoch": 137.03,
      "learning_rate": 1.7299676375404532e-06,
      "loss": 0.0005,
      "step": 352840
    },
    {
      "epoch": 137.03,
      "learning_rate": 1.7294498381877024e-06,
      "loss": 0.0275,
      "step": 352850
    },
    {
      "epoch": 137.03,
      "learning_rate": 1.7289320388349515e-06,
      "loss": 0.0797,
      "step": 352860
    },
    {
      "epoch": 137.04,
      "learning_rate": 1.7284142394822007e-06,
      "loss": 0.0918,
      "step": 352870
    },
    {
      "epoch": 137.04,
      "learning_rate": 1.72789644012945e-06,
      "loss": 0.0828,
      "step": 352880
    },
    {
      "epoch": 137.04,
      "learning_rate": 1.7273786407766991e-06,
      "loss": 0.0777,
      "step": 352890
    },
    {
      "epoch": 137.05,
      "learning_rate": 1.7268608414239485e-06,
      "loss": 0.05,
      "step": 352900
    },
    {
      "epoch": 137.05,
      "learning_rate": 1.7263430420711977e-06,
      "loss": 0.0038,
      "step": 352910
    },
    {
      "epoch": 137.06,
      "learning_rate": 1.725825242718447e-06,
      "loss": 0.0389,
      "step": 352920
    },
    {
      "epoch": 137.06,
      "learning_rate": 1.7253074433656961e-06,
      "loss": 0.0149,
      "step": 352930
    },
    {
      "epoch": 137.06,
      "learning_rate": 1.7247896440129453e-06,
      "loss": 0.0168,
      "step": 352940
    },
    {
      "epoch": 137.07,
      "learning_rate": 1.7242718446601945e-06,
      "loss": 0.0909,
      "step": 352950
    },
    {
      "epoch": 137.07,
      "learning_rate": 1.7237540453074437e-06,
      "loss": 0.097,
      "step": 352960
    },
    {
      "epoch": 137.08,
      "learning_rate": 1.723236245954693e-06,
      "loss": 0.039,
      "step": 352970
    },
    {
      "epoch": 137.08,
      "learning_rate": 1.7227184466019419e-06,
      "loss": 0.1649,
      "step": 352980
    },
    {
      "epoch": 137.08,
      "learning_rate": 1.722200647249191e-06,
      "loss": 0.0298,
      "step": 352990
    },
    {
      "epoch": 137.09,
      "learning_rate": 1.7216828478964403e-06,
      "loss": 0.0667,
      "step": 353000
    },
    {
      "epoch": 137.09,
      "learning_rate": 1.7211650485436895e-06,
      "loss": 0.0188,
      "step": 353010
    },
    {
      "epoch": 137.1,
      "learning_rate": 1.7206472491909387e-06,
      "loss": 0.0002,
      "step": 353020
    },
    {
      "epoch": 137.1,
      "learning_rate": 1.7201294498381879e-06,
      "loss": 0.0942,
      "step": 353030
    },
    {
      "epoch": 137.1,
      "learning_rate": 1.719611650485437e-06,
      "loss": 0.0101,
      "step": 353040
    },
    {
      "epoch": 137.11,
      "learning_rate": 1.7190938511326862e-06,
      "loss": 0.0283,
      "step": 353050
    },
    {
      "epoch": 137.11,
      "learning_rate": 1.7185760517799354e-06,
      "loss": 0.0902,
      "step": 353060
    },
    {
      "epoch": 137.11,
      "learning_rate": 1.7180582524271846e-06,
      "loss": 0.0115,
      "step": 353070
    },
    {
      "epoch": 137.12,
      "learning_rate": 1.7175404530744338e-06,
      "loss": 0.0457,
      "step": 353080
    },
    {
      "epoch": 137.12,
      "learning_rate": 1.717022653721683e-06,
      "loss": 0.0333,
      "step": 353090
    },
    {
      "epoch": 137.13,
      "learning_rate": 1.7165048543689322e-06,
      "loss": 0.034,
      "step": 353100
    },
    {
      "epoch": 137.13,
      "learning_rate": 1.7159870550161814e-06,
      "loss": 0.0994,
      "step": 353110
    },
    {
      "epoch": 137.13,
      "learning_rate": 1.7154692556634306e-06,
      "loss": 0.1084,
      "step": 353120
    },
    {
      "epoch": 137.14,
      "learning_rate": 1.7149514563106798e-06,
      "loss": 0.1391,
      "step": 353130
    },
    {
      "epoch": 137.14,
      "learning_rate": 1.714433656957929e-06,
      "loss": 0.0931,
      "step": 353140
    },
    {
      "epoch": 137.15,
      "learning_rate": 1.7139158576051782e-06,
      "loss": 0.0367,
      "step": 353150
    },
    {
      "epoch": 137.15,
      "learning_rate": 1.7133980582524274e-06,
      "loss": 0.0001,
      "step": 353160
    },
    {
      "epoch": 137.15,
      "learning_rate": 1.7128802588996766e-06,
      "loss": 0.0005,
      "step": 353170
    },
    {
      "epoch": 137.16,
      "learning_rate": 1.7123624595469258e-06,
      "loss": 0.0247,
      "step": 353180
    },
    {
      "epoch": 137.16,
      "learning_rate": 1.711844660194175e-06,
      "loss": 0.0009,
      "step": 353190
    },
    {
      "epoch": 137.17,
      "learning_rate": 1.7113268608414242e-06,
      "loss": 0.1282,
      "step": 353200
    },
    {
      "epoch": 137.17,
      "learning_rate": 1.7108090614886734e-06,
      "loss": 0.0309,
      "step": 353210
    },
    {
      "epoch": 137.17,
      "learning_rate": 1.7102912621359226e-06,
      "loss": 0.0117,
      "step": 353220
    },
    {
      "epoch": 137.18,
      "learning_rate": 1.7097734627831715e-06,
      "loss": 0.0002,
      "step": 353230
    },
    {
      "epoch": 137.18,
      "learning_rate": 1.7092556634304207e-06,
      "loss": 0.0172,
      "step": 353240
    },
    {
      "epoch": 137.18,
      "learning_rate": 1.70873786407767e-06,
      "loss": 0.0001,
      "step": 353250
    },
    {
      "epoch": 137.19,
      "learning_rate": 1.7082200647249191e-06,
      "loss": 0.0299,
      "step": 353260
    },
    {
      "epoch": 137.19,
      "learning_rate": 1.7077022653721683e-06,
      "loss": 0.0001,
      "step": 353270
    },
    {
      "epoch": 137.2,
      "learning_rate": 1.7071844660194175e-06,
      "loss": 0.0318,
      "step": 353280
    },
    {
      "epoch": 137.2,
      "learning_rate": 1.7066666666666667e-06,
      "loss": 0.0582,
      "step": 353290
    },
    {
      "epoch": 137.2,
      "learning_rate": 1.706148867313916e-06,
      "loss": 0.0262,
      "step": 353300
    },
    {
      "epoch": 137.21,
      "learning_rate": 1.705631067961165e-06,
      "loss": 0.002,
      "step": 353310
    },
    {
      "epoch": 137.21,
      "learning_rate": 1.7051132686084143e-06,
      "loss": 0.0768,
      "step": 353320
    },
    {
      "epoch": 137.22,
      "learning_rate": 1.7045954692556635e-06,
      "loss": 0.0108,
      "step": 353330
    },
    {
      "epoch": 137.22,
      "learning_rate": 1.7040776699029127e-06,
      "loss": 0.0831,
      "step": 353340
    },
    {
      "epoch": 137.22,
      "learning_rate": 1.7035598705501619e-06,
      "loss": 0.0498,
      "step": 353350
    },
    {
      "epoch": 137.23,
      "learning_rate": 1.703042071197411e-06,
      "loss": 0.004,
      "step": 353360
    },
    {
      "epoch": 137.23,
      "learning_rate": 1.7025242718446603e-06,
      "loss": 0.0047,
      "step": 353370
    },
    {
      "epoch": 137.23,
      "learning_rate": 1.7020064724919095e-06,
      "loss": 0.0017,
      "step": 353380
    },
    {
      "epoch": 137.24,
      "learning_rate": 1.7014886731391586e-06,
      "loss": 0.0191,
      "step": 353390
    },
    {
      "epoch": 137.24,
      "learning_rate": 1.7009708737864078e-06,
      "loss": 0.0093,
      "step": 353400
    },
    {
      "epoch": 137.25,
      "learning_rate": 1.700453074433657e-06,
      "loss": 0.0607,
      "step": 353410
    },
    {
      "epoch": 137.25,
      "learning_rate": 1.6999352750809062e-06,
      "loss": 0.0103,
      "step": 353420
    },
    {
      "epoch": 137.25,
      "learning_rate": 1.6994174757281554e-06,
      "loss": 0.0485,
      "step": 353430
    },
    {
      "epoch": 137.26,
      "learning_rate": 1.6988996763754046e-06,
      "loss": 0.0124,
      "step": 353440
    },
    {
      "epoch": 137.26,
      "learning_rate": 1.6983818770226538e-06,
      "loss": 0.098,
      "step": 353450
    },
    {
      "epoch": 137.27,
      "learning_rate": 1.6978640776699032e-06,
      "loss": 0.0215,
      "step": 353460
    },
    {
      "epoch": 137.27,
      "learning_rate": 1.6973462783171524e-06,
      "loss": 0.0128,
      "step": 353470
    },
    {
      "epoch": 137.27,
      "learning_rate": 1.6968284789644016e-06,
      "loss": 0.0636,
      "step": 353480
    },
    {
      "epoch": 137.28,
      "learning_rate": 1.6963106796116508e-06,
      "loss": 0.001,
      "step": 353490
    },
    {
      "epoch": 137.28,
      "learning_rate": 1.6957928802589e-06,
      "loss": 0.0514,
      "step": 353500
    },
    {
      "epoch": 137.29,
      "learning_rate": 1.6952750809061492e-06,
      "loss": 0.0194,
      "step": 353510
    },
    {
      "epoch": 137.29,
      "learning_rate": 1.6947572815533984e-06,
      "loss": 0.0502,
      "step": 353520
    },
    {
      "epoch": 137.29,
      "learning_rate": 1.6942394822006476e-06,
      "loss": 0.0136,
      "step": 353530
    },
    {
      "epoch": 137.3,
      "learning_rate": 1.6937216828478968e-06,
      "loss": 0.0691,
      "step": 353540
    },
    {
      "epoch": 137.3,
      "learning_rate": 1.6932038834951458e-06,
      "loss": 0.0091,
      "step": 353550
    },
    {
      "epoch": 137.3,
      "learning_rate": 1.692686084142395e-06,
      "loss": 0.0219,
      "step": 353560
    },
    {
      "epoch": 137.31,
      "learning_rate": 1.6921682847896441e-06,
      "loss": 0.0059,
      "step": 353570
    },
    {
      "epoch": 137.31,
      "learning_rate": 1.6916504854368933e-06,
      "loss": 0.1721,
      "step": 353580
    },
    {
      "epoch": 137.32,
      "learning_rate": 1.6911326860841425e-06,
      "loss": 0.0335,
      "step": 353590
    },
    {
      "epoch": 137.32,
      "learning_rate": 1.6906148867313917e-06,
      "loss": 0.0772,
      "step": 353600
    },
    {
      "epoch": 137.32,
      "learning_rate": 1.690097087378641e-06,
      "loss": 0.0006,
      "step": 353610
    },
    {
      "epoch": 137.33,
      "learning_rate": 1.6895792880258901e-06,
      "loss": 0.0542,
      "step": 353620
    },
    {
      "epoch": 137.33,
      "learning_rate": 1.6890614886731393e-06,
      "loss": 0.0336,
      "step": 353630
    },
    {
      "epoch": 137.34,
      "learning_rate": 1.6885436893203885e-06,
      "loss": 0.042,
      "step": 353640
    },
    {
      "epoch": 137.34,
      "learning_rate": 1.6880258899676377e-06,
      "loss": 0.0472,
      "step": 353650
    },
    {
      "epoch": 137.34,
      "learning_rate": 1.687508090614887e-06,
      "loss": 0.0388,
      "step": 353660
    },
    {
      "epoch": 137.35,
      "learning_rate": 1.686990291262136e-06,
      "loss": 0.0002,
      "step": 353670
    },
    {
      "epoch": 137.35,
      "learning_rate": 1.6864724919093853e-06,
      "loss": 0.13,
      "step": 353680
    },
    {
      "epoch": 137.36,
      "learning_rate": 1.6859546925566345e-06,
      "loss": 0.0192,
      "step": 353690
    },
    {
      "epoch": 137.36,
      "learning_rate": 1.6854368932038837e-06,
      "loss": 0.033,
      "step": 353700
    },
    {
      "epoch": 137.36,
      "learning_rate": 1.6849190938511329e-06,
      "loss": 0.0911,
      "step": 353710
    },
    {
      "epoch": 137.37,
      "learning_rate": 1.684401294498382e-06,
      "loss": 0.0242,
      "step": 353720
    },
    {
      "epoch": 137.37,
      "learning_rate": 1.6838834951456313e-06,
      "loss": 0.0129,
      "step": 353730
    },
    {
      "epoch": 137.37,
      "learning_rate": 1.6833656957928805e-06,
      "loss": 0.0115,
      "step": 353740
    },
    {
      "epoch": 137.38,
      "learning_rate": 1.6828478964401297e-06,
      "loss": 0.0057,
      "step": 353750
    },
    {
      "epoch": 137.38,
      "learning_rate": 1.6823300970873788e-06,
      "loss": 0.1398,
      "step": 353760
    },
    {
      "epoch": 137.39,
      "learning_rate": 1.681812297734628e-06,
      "loss": 0.0177,
      "step": 353770
    },
    {
      "epoch": 137.39,
      "learning_rate": 1.6812944983818772e-06,
      "loss": 0.0288,
      "step": 353780
    },
    {
      "epoch": 137.39,
      "learning_rate": 1.6807766990291264e-06,
      "loss": 0.0125,
      "step": 353790
    },
    {
      "epoch": 137.4,
      "learning_rate": 1.6802588996763754e-06,
      "loss": 0.092,
      "step": 353800
    },
    {
      "epoch": 137.4,
      "learning_rate": 1.6797411003236246e-06,
      "loss": 0.0275,
      "step": 353810
    },
    {
      "epoch": 137.41,
      "learning_rate": 1.6792233009708738e-06,
      "loss": 0.0358,
      "step": 353820
    },
    {
      "epoch": 137.41,
      "learning_rate": 1.678705501618123e-06,
      "loss": 0.0021,
      "step": 353830
    },
    {
      "epoch": 137.41,
      "learning_rate": 1.6781877022653722e-06,
      "loss": 0.0095,
      "step": 353840
    },
    {
      "epoch": 137.42,
      "learning_rate": 1.6776699029126214e-06,
      "loss": 0.0248,
      "step": 353850
    },
    {
      "epoch": 137.42,
      "learning_rate": 1.6771521035598706e-06,
      "loss": 0.0747,
      "step": 353860
    },
    {
      "epoch": 137.43,
      "learning_rate": 1.6766343042071198e-06,
      "loss": 0.0087,
      "step": 353870
    },
    {
      "epoch": 137.43,
      "learning_rate": 1.676116504854369e-06,
      "loss": 0.0002,
      "step": 353880
    },
    {
      "epoch": 137.43,
      "learning_rate": 1.6755987055016182e-06,
      "loss": 0.0629,
      "step": 353890
    },
    {
      "epoch": 137.44,
      "learning_rate": 1.6750809061488674e-06,
      "loss": 0.0688,
      "step": 353900
    },
    {
      "epoch": 137.44,
      "learning_rate": 1.6745631067961166e-06,
      "loss": 0.0109,
      "step": 353910
    },
    {
      "epoch": 137.44,
      "learning_rate": 1.6740453074433657e-06,
      "loss": 0.01,
      "step": 353920
    },
    {
      "epoch": 137.45,
      "learning_rate": 1.673527508090615e-06,
      "loss": 0.0369,
      "step": 353930
    },
    {
      "epoch": 137.45,
      "learning_rate": 1.6730097087378641e-06,
      "loss": 0.1657,
      "step": 353940
    },
    {
      "epoch": 137.46,
      "learning_rate": 1.6724919093851133e-06,
      "loss": 0.0125,
      "step": 353950
    },
    {
      "epoch": 137.46,
      "learning_rate": 1.6719741100323625e-06,
      "loss": 0.0055,
      "step": 353960
    },
    {
      "epoch": 137.46,
      "learning_rate": 1.6714563106796117e-06,
      "loss": 0.0392,
      "step": 353970
    },
    {
      "epoch": 137.47,
      "learning_rate": 1.670938511326861e-06,
      "loss": 0.1075,
      "step": 353980
    },
    {
      "epoch": 137.47,
      "learning_rate": 1.6704207119741101e-06,
      "loss": 0.0123,
      "step": 353990
    },
    {
      "epoch": 137.48,
      "learning_rate": 1.6699029126213593e-06,
      "loss": 0.0689,
      "step": 354000
    },
    {
      "epoch": 137.48,
      "learning_rate": 1.6693851132686085e-06,
      "loss": 0.1215,
      "step": 354010
    },
    {
      "epoch": 137.48,
      "learning_rate": 1.668867313915858e-06,
      "loss": 0.0221,
      "step": 354020
    },
    {
      "epoch": 137.49,
      "learning_rate": 1.668349514563107e-06,
      "loss": 0.1155,
      "step": 354030
    },
    {
      "epoch": 137.49,
      "learning_rate": 1.6678317152103563e-06,
      "loss": 0.0557,
      "step": 354040
    },
    {
      "epoch": 137.5,
      "learning_rate": 1.6673139158576055e-06,
      "loss": 0.0004,
      "step": 354050
    },
    {
      "epoch": 137.5,
      "learning_rate": 1.6667961165048547e-06,
      "loss": 0.0191,
      "step": 354060
    },
    {
      "epoch": 137.5,
      "learning_rate": 1.6662783171521039e-06,
      "loss": 0.0504,
      "step": 354070
    },
    {
      "epoch": 137.51,
      "learning_rate": 1.665760517799353e-06,
      "loss": 0.1418,
      "step": 354080
    },
    {
      "epoch": 137.51,
      "learning_rate": 1.6652427184466023e-06,
      "loss": 0.0803,
      "step": 354090
    },
    {
      "epoch": 137.51,
      "learning_rate": 1.6647249190938515e-06,
      "loss": 0.0495,
      "step": 354100
    },
    {
      "epoch": 137.52,
      "learning_rate": 1.6642071197411007e-06,
      "loss": 0.0628,
      "step": 354110
    },
    {
      "epoch": 137.52,
      "learning_rate": 1.6636893203883496e-06,
      "loss": 0.0497,
      "step": 354120
    },
    {
      "epoch": 137.53,
      "learning_rate": 1.6631715210355988e-06,
      "loss": 0.0378,
      "step": 354130
    },
    {
      "epoch": 137.53,
      "learning_rate": 1.662653721682848e-06,
      "loss": 0.0971,
      "step": 354140
    },
    {
      "epoch": 137.53,
      "learning_rate": 1.6621359223300972e-06,
      "loss": 0.0151,
      "step": 354150
    },
    {
      "epoch": 137.54,
      "learning_rate": 1.6616181229773464e-06,
      "loss": 0.0343,
      "step": 354160
    },
    {
      "epoch": 137.54,
      "learning_rate": 1.6611003236245956e-06,
      "loss": 0.0244,
      "step": 354170
    },
    {
      "epoch": 137.55,
      "learning_rate": 1.6605825242718448e-06,
      "loss": 0.0514,
      "step": 354180
    },
    {
      "epoch": 137.55,
      "learning_rate": 1.660064724919094e-06,
      "loss": 0.0466,
      "step": 354190
    },
    {
      "epoch": 137.55,
      "learning_rate": 1.6595469255663432e-06,
      "loss": 0.0084,
      "step": 354200
    },
    {
      "epoch": 137.56,
      "learning_rate": 1.6590291262135924e-06,
      "loss": 0.0283,
      "step": 354210
    },
    {
      "epoch": 137.56,
      "learning_rate": 1.6585113268608416e-06,
      "loss": 0.1091,
      "step": 354220
    },
    {
      "epoch": 137.57,
      "learning_rate": 1.6579935275080908e-06,
      "loss": 0.0704,
      "step": 354230
    },
    {
      "epoch": 137.57,
      "learning_rate": 1.65747572815534e-06,
      "loss": 0.0004,
      "step": 354240
    },
    {
      "epoch": 137.57,
      "learning_rate": 1.6569579288025892e-06,
      "loss": 0.0138,
      "step": 354250
    },
    {
      "epoch": 137.58,
      "learning_rate": 1.6564401294498384e-06,
      "loss": 0.0172,
      "step": 354260
    },
    {
      "epoch": 137.58,
      "learning_rate": 1.6559223300970876e-06,
      "loss": 0.0151,
      "step": 354270
    },
    {
      "epoch": 137.58,
      "learning_rate": 1.6554045307443367e-06,
      "loss": 0.0616,
      "step": 354280
    },
    {
      "epoch": 137.59,
      "learning_rate": 1.654886731391586e-06,
      "loss": 0.0286,
      "step": 354290
    },
    {
      "epoch": 137.59,
      "learning_rate": 1.6543689320388351e-06,
      "loss": 0.0741,
      "step": 354300
    },
    {
      "epoch": 137.6,
      "learning_rate": 1.6538511326860843e-06,
      "loss": 0.0235,
      "step": 354310
    },
    {
      "epoch": 137.6,
      "learning_rate": 1.6533333333333335e-06,
      "loss": 0.0825,
      "step": 354320
    },
    {
      "epoch": 137.6,
      "learning_rate": 1.6528155339805827e-06,
      "loss": 0.0155,
      "step": 354330
    },
    {
      "epoch": 137.61,
      "learning_rate": 1.652297734627832e-06,
      "loss": 0.0007,
      "step": 354340
    },
    {
      "epoch": 137.61,
      "learning_rate": 1.6517799352750811e-06,
      "loss": 0.0134,
      "step": 354350
    },
    {
      "epoch": 137.62,
      "learning_rate": 1.6512621359223303e-06,
      "loss": 0.1164,
      "step": 354360
    },
    {
      "epoch": 137.62,
      "learning_rate": 1.6507443365695793e-06,
      "loss": 0.0002,
      "step": 354370
    },
    {
      "epoch": 137.62,
      "learning_rate": 1.6502265372168285e-06,
      "loss": 0.0301,
      "step": 354380
    },
    {
      "epoch": 137.63,
      "learning_rate": 1.6497087378640777e-06,
      "loss": 0.0452,
      "step": 354390
    },
    {
      "epoch": 137.63,
      "learning_rate": 1.6491909385113269e-06,
      "loss": 0.0095,
      "step": 354400
    },
    {
      "epoch": 137.63,
      "learning_rate": 1.648673139158576e-06,
      "loss": 0.0241,
      "step": 354410
    },
    {
      "epoch": 137.64,
      "learning_rate": 1.6481553398058253e-06,
      "loss": 0.0331,
      "step": 354420
    },
    {
      "epoch": 137.64,
      "learning_rate": 1.6476375404530745e-06,
      "loss": 0.0587,
      "step": 354430
    },
    {
      "epoch": 137.65,
      "learning_rate": 1.6471197411003236e-06,
      "loss": 0.0108,
      "step": 354440
    },
    {
      "epoch": 137.65,
      "learning_rate": 1.6466019417475728e-06,
      "loss": 0.0483,
      "step": 354450
    },
    {
      "epoch": 137.65,
      "learning_rate": 1.646084142394822e-06,
      "loss": 0.0013,
      "step": 354460
    },
    {
      "epoch": 137.66,
      "learning_rate": 1.6455663430420712e-06,
      "loss": 0.0146,
      "step": 354470
    },
    {
      "epoch": 137.66,
      "learning_rate": 1.6450485436893204e-06,
      "loss": 0.0351,
      "step": 354480
    },
    {
      "epoch": 137.67,
      "learning_rate": 1.6445307443365696e-06,
      "loss": 0.1314,
      "step": 354490
    },
    {
      "epoch": 137.67,
      "learning_rate": 1.6440129449838188e-06,
      "loss": 0.1066,
      "step": 354500
    },
    {
      "epoch": 137.67,
      "learning_rate": 1.643495145631068e-06,
      "loss": 0.0083,
      "step": 354510
    },
    {
      "epoch": 137.68,
      "learning_rate": 1.6429773462783172e-06,
      "loss": 0.0001,
      "step": 354520
    },
    {
      "epoch": 137.68,
      "learning_rate": 1.6424595469255664e-06,
      "loss": 0.0265,
      "step": 354530
    },
    {
      "epoch": 137.69,
      "learning_rate": 1.6419417475728156e-06,
      "loss": 0.0188,
      "step": 354540
    },
    {
      "epoch": 137.69,
      "learning_rate": 1.6414239482200648e-06,
      "loss": 0.0446,
      "step": 354550
    },
    {
      "epoch": 137.69,
      "learning_rate": 1.640906148867314e-06,
      "loss": 0.0003,
      "step": 354560
    },
    {
      "epoch": 137.7,
      "learning_rate": 1.6403883495145632e-06,
      "loss": 0.2866,
      "step": 354570
    },
    {
      "epoch": 137.7,
      "learning_rate": 1.6398705501618126e-06,
      "loss": 0.0371,
      "step": 354580
    },
    {
      "epoch": 137.7,
      "learning_rate": 1.6393527508090618e-06,
      "loss": 0.0723,
      "step": 354590
    },
    {
      "epoch": 137.71,
      "learning_rate": 1.638834951456311e-06,
      "loss": 0.0239,
      "step": 354600
    },
    {
      "epoch": 137.71,
      "learning_rate": 1.6383171521035602e-06,
      "loss": 0.025,
      "step": 354610
    },
    {
      "epoch": 137.72,
      "learning_rate": 1.6377993527508094e-06,
      "loss": 0.0934,
      "step": 354620
    },
    {
      "epoch": 137.72,
      "learning_rate": 1.6372815533980586e-06,
      "loss": 0.0503,
      "step": 354630
    },
    {
      "epoch": 137.72,
      "learning_rate": 1.6367637540453078e-06,
      "loss": 0.0259,
      "step": 354640
    },
    {
      "epoch": 137.73,
      "learning_rate": 1.636245954692557e-06,
      "loss": 0.1587,
      "step": 354650
    },
    {
      "epoch": 137.73,
      "learning_rate": 1.6357281553398061e-06,
      "loss": 0.004,
      "step": 354660
    },
    {
      "epoch": 137.74,
      "learning_rate": 1.6352103559870553e-06,
      "loss": 0.0014,
      "step": 354670
    },
    {
      "epoch": 137.74,
      "learning_rate": 1.6346925566343045e-06,
      "loss": 0.0634,
      "step": 354680
    },
    {
      "epoch": 137.74,
      "learning_rate": 1.6341747572815535e-06,
      "loss": 0.0014,
      "step": 354690
    },
    {
      "epoch": 137.75,
      "learning_rate": 1.6336569579288027e-06,
      "loss": 0.026,
      "step": 354700
    },
    {
      "epoch": 137.75,
      "learning_rate": 1.633139158576052e-06,
      "loss": 0.0138,
      "step": 354710
    },
    {
      "epoch": 137.76,
      "learning_rate": 1.632621359223301e-06,
      "loss": 0.1316,
      "step": 354720
    },
    {
      "epoch": 137.76,
      "learning_rate": 1.6321035598705503e-06,
      "loss": 0.0108,
      "step": 354730
    },
    {
      "epoch": 137.76,
      "learning_rate": 1.6315857605177995e-06,
      "loss": 0.0275,
      "step": 354740
    },
    {
      "epoch": 137.77,
      "learning_rate": 1.6310679611650487e-06,
      "loss": 0.0057,
      "step": 354750
    },
    {
      "epoch": 137.77,
      "learning_rate": 1.6305501618122979e-06,
      "loss": 0.0037,
      "step": 354760
    },
    {
      "epoch": 137.77,
      "learning_rate": 1.630032362459547e-06,
      "loss": 0.0002,
      "step": 354770
    },
    {
      "epoch": 137.78,
      "learning_rate": 1.6295145631067963e-06,
      "loss": 0.0098,
      "step": 354780
    },
    {
      "epoch": 137.78,
      "learning_rate": 1.6289967637540455e-06,
      "loss": 0.0882,
      "step": 354790
    },
    {
      "epoch": 137.79,
      "learning_rate": 1.6284789644012947e-06,
      "loss": 0.0617,
      "step": 354800
    },
    {
      "epoch": 137.79,
      "learning_rate": 1.6279611650485438e-06,
      "loss": 0.001,
      "step": 354810
    },
    {
      "epoch": 137.79,
      "learning_rate": 1.627443365695793e-06,
      "loss": 0.0305,
      "step": 354820
    },
    {
      "epoch": 137.8,
      "learning_rate": 1.6269255663430422e-06,
      "loss": 0.0301,
      "step": 354830
    },
    {
      "epoch": 137.8,
      "learning_rate": 1.6264077669902914e-06,
      "loss": 0.0246,
      "step": 354840
    },
    {
      "epoch": 137.81,
      "learning_rate": 1.6258899676375406e-06,
      "loss": 0.0212,
      "step": 354850
    },
    {
      "epoch": 137.81,
      "learning_rate": 1.6253721682847898e-06,
      "loss": 0.1337,
      "step": 354860
    },
    {
      "epoch": 137.81,
      "learning_rate": 1.624854368932039e-06,
      "loss": 0.0478,
      "step": 354870
    },
    {
      "epoch": 137.82,
      "learning_rate": 1.6243365695792882e-06,
      "loss": 0.0179,
      "step": 354880
    },
    {
      "epoch": 137.82,
      "learning_rate": 1.6238187702265374e-06,
      "loss": 0.0138,
      "step": 354890
    },
    {
      "epoch": 137.83,
      "learning_rate": 1.6233009708737866e-06,
      "loss": 0.0981,
      "step": 354900
    },
    {
      "epoch": 137.83,
      "learning_rate": 1.6227831715210358e-06,
      "loss": 0.0179,
      "step": 354910
    },
    {
      "epoch": 137.83,
      "learning_rate": 1.622265372168285e-06,
      "loss": 0.2243,
      "step": 354920
    },
    {
      "epoch": 137.84,
      "learning_rate": 1.6217475728155342e-06,
      "loss": 0.0331,
      "step": 354930
    },
    {
      "epoch": 137.84,
      "learning_rate": 1.6212297734627832e-06,
      "loss": 0.0623,
      "step": 354940
    },
    {
      "epoch": 137.84,
      "learning_rate": 1.6207119741100324e-06,
      "loss": 0.1719,
      "step": 354950
    },
    {
      "epoch": 137.85,
      "learning_rate": 1.6201941747572816e-06,
      "loss": 0.1306,
      "step": 354960
    },
    {
      "epoch": 137.85,
      "learning_rate": 1.6196763754045307e-06,
      "loss": 0.1103,
      "step": 354970
    },
    {
      "epoch": 137.86,
      "learning_rate": 1.61915857605178e-06,
      "loss": 0.0002,
      "step": 354980
    },
    {
      "epoch": 137.86,
      "learning_rate": 1.6186407766990291e-06,
      "loss": 0.1051,
      "step": 354990
    },
    {
      "epoch": 137.86,
      "learning_rate": 1.6181229773462783e-06,
      "loss": 0.0796,
      "step": 355000
    },
    {
      "epoch": 137.87,
      "learning_rate": 1.6176051779935275e-06,
      "loss": 0.0439,
      "step": 355010
    },
    {
      "epoch": 137.87,
      "learning_rate": 1.6170873786407767e-06,
      "loss": 0.0223,
      "step": 355020
    },
    {
      "epoch": 137.88,
      "learning_rate": 1.616569579288026e-06,
      "loss": 0.0347,
      "step": 355030
    },
    {
      "epoch": 137.88,
      "learning_rate": 1.6160517799352751e-06,
      "loss": 0.0113,
      "step": 355040
    },
    {
      "epoch": 137.88,
      "learning_rate": 1.6155339805825243e-06,
      "loss": 0.1329,
      "step": 355050
    },
    {
      "epoch": 137.89,
      "learning_rate": 1.6150161812297735e-06,
      "loss": 0.0516,
      "step": 355060
    },
    {
      "epoch": 137.89,
      "learning_rate": 1.6144983818770227e-06,
      "loss": 0.0176,
      "step": 355070
    },
    {
      "epoch": 137.9,
      "learning_rate": 1.6139805825242719e-06,
      "loss": 0.0303,
      "step": 355080
    },
    {
      "epoch": 137.9,
      "learning_rate": 1.613462783171521e-06,
      "loss": 0.0098,
      "step": 355090
    },
    {
      "epoch": 137.9,
      "learning_rate": 1.6129449838187703e-06,
      "loss": 0.0663,
      "step": 355100
    },
    {
      "epoch": 137.91,
      "learning_rate": 1.6124271844660195e-06,
      "loss": 0.0049,
      "step": 355110
    },
    {
      "epoch": 137.91,
      "learning_rate": 1.6119093851132687e-06,
      "loss": 0.0455,
      "step": 355120
    },
    {
      "epoch": 137.91,
      "learning_rate": 1.6113915857605179e-06,
      "loss": 0.1173,
      "step": 355130
    },
    {
      "epoch": 137.92,
      "learning_rate": 1.6108737864077673e-06,
      "loss": 0.0105,
      "step": 355140
    },
    {
      "epoch": 137.92,
      "learning_rate": 1.6103559870550165e-06,
      "loss": 0.0097,
      "step": 355150
    },
    {
      "epoch": 137.93,
      "learning_rate": 1.6098381877022657e-06,
      "loss": 0.0256,
      "step": 355160
    },
    {
      "epoch": 137.93,
      "learning_rate": 1.6093203883495148e-06,
      "loss": 0.0287,
      "step": 355170
    },
    {
      "epoch": 137.93,
      "learning_rate": 1.608802588996764e-06,
      "loss": 0.0303,
      "step": 355180
    },
    {
      "epoch": 137.94,
      "learning_rate": 1.6082847896440132e-06,
      "loss": 0.0638,
      "step": 355190
    },
    {
      "epoch": 137.94,
      "learning_rate": 1.6077669902912624e-06,
      "loss": 0.2044,
      "step": 355200
    },
    {
      "epoch": 137.95,
      "learning_rate": 1.6072491909385116e-06,
      "loss": 0.0576,
      "step": 355210
    },
    {
      "epoch": 137.95,
      "learning_rate": 1.6067313915857608e-06,
      "loss": 0.0651,
      "step": 355220
    },
    {
      "epoch": 137.95,
      "learning_rate": 1.60621359223301e-06,
      "loss": 0.0337,
      "step": 355230
    },
    {
      "epoch": 137.96,
      "learning_rate": 1.6056957928802592e-06,
      "loss": 0.0851,
      "step": 355240
    },
    {
      "epoch": 137.96,
      "learning_rate": 1.6051779935275084e-06,
      "loss": 0.0328,
      "step": 355250
    },
    {
      "epoch": 137.97,
      "learning_rate": 1.6046601941747574e-06,
      "loss": 0.0259,
      "step": 355260
    },
    {
      "epoch": 137.97,
      "learning_rate": 1.6041423948220066e-06,
      "loss": 0.0285,
      "step": 355270
    },
    {
      "epoch": 137.97,
      "learning_rate": 1.6036245954692558e-06,
      "loss": 0.0159,
      "step": 355280
    },
    {
      "epoch": 137.98,
      "learning_rate": 1.603106796116505e-06,
      "loss": 0.0013,
      "step": 355290
    },
    {
      "epoch": 137.98,
      "learning_rate": 1.6025889967637542e-06,
      "loss": 0.0983,
      "step": 355300
    },
    {
      "epoch": 137.98,
      "learning_rate": 1.6020711974110034e-06,
      "loss": 0.0032,
      "step": 355310
    },
    {
      "epoch": 137.99,
      "learning_rate": 1.6015533980582526e-06,
      "loss": 0.0138,
      "step": 355320
    },
    {
      "epoch": 137.99,
      "learning_rate": 1.6010355987055018e-06,
      "loss": 0.026,
      "step": 355330
    },
    {
      "epoch": 138.0,
      "learning_rate": 1.600517799352751e-06,
      "loss": 0.0337,
      "step": 355340
    },
    {
      "epoch": 138.0,
      "learning_rate": 1.6000000000000001e-06,
      "loss": 0.0081,
      "step": 355350
    },
    {
      "epoch": 138.0,
      "eval_accuracy": 0.953232462173315,
      "eval_loss": 0.3852311968803406,
      "eval_runtime": 8.1731,
      "eval_samples_per_second": 444.754,
      "eval_steps_per_second": 55.671,
      "step": 355350
    },
    {
      "epoch": 138.0,
      "learning_rate": 1.5994822006472493e-06,
      "loss": 0.0364,
      "step": 355360
    },
    {
      "epoch": 138.01,
      "learning_rate": 1.5989644012944985e-06,
      "loss": 0.016,
      "step": 355370
    },
    {
      "epoch": 138.01,
      "learning_rate": 1.5984466019417477e-06,
      "loss": 0.0079,
      "step": 355380
    },
    {
      "epoch": 138.02,
      "learning_rate": 1.597928802588997e-06,
      "loss": 0.0556,
      "step": 355390
    },
    {
      "epoch": 138.02,
      "learning_rate": 1.5974110032362461e-06,
      "loss": 0.01,
      "step": 355400
    },
    {
      "epoch": 138.02,
      "learning_rate": 1.5968932038834953e-06,
      "loss": 0.0161,
      "step": 355410
    },
    {
      "epoch": 138.03,
      "learning_rate": 1.5963754045307445e-06,
      "loss": 0.0109,
      "step": 355420
    },
    {
      "epoch": 138.03,
      "learning_rate": 1.5958576051779937e-06,
      "loss": 0.0532,
      "step": 355430
    },
    {
      "epoch": 138.03,
      "learning_rate": 1.5953398058252429e-06,
      "loss": 0.0353,
      "step": 355440
    },
    {
      "epoch": 138.04,
      "learning_rate": 1.594822006472492e-06,
      "loss": 0.1191,
      "step": 355450
    },
    {
      "epoch": 138.04,
      "learning_rate": 1.5943042071197413e-06,
      "loss": 0.0206,
      "step": 355460
    },
    {
      "epoch": 138.05,
      "learning_rate": 1.5937864077669905e-06,
      "loss": 0.0009,
      "step": 355470
    },
    {
      "epoch": 138.05,
      "learning_rate": 1.5932686084142397e-06,
      "loss": 0.0866,
      "step": 355480
    },
    {
      "epoch": 138.05,
      "learning_rate": 1.5927508090614889e-06,
      "loss": 0.002,
      "step": 355490
    },
    {
      "epoch": 138.06,
      "learning_rate": 1.592233009708738e-06,
      "loss": 0.0559,
      "step": 355500
    },
    {
      "epoch": 138.06,
      "learning_rate": 1.591715210355987e-06,
      "loss": 0.0689,
      "step": 355510
    },
    {
      "epoch": 138.07,
      "learning_rate": 1.5911974110032362e-06,
      "loss": 0.0411,
      "step": 355520
    },
    {
      "epoch": 138.07,
      "learning_rate": 1.5906796116504854e-06,
      "loss": 0.0352,
      "step": 355530
    },
    {
      "epoch": 138.07,
      "learning_rate": 1.5901618122977346e-06,
      "loss": 0.0901,
      "step": 355540
    },
    {
      "epoch": 138.08,
      "learning_rate": 1.5896440129449838e-06,
      "loss": 0.0441,
      "step": 355550
    },
    {
      "epoch": 138.08,
      "learning_rate": 1.589126213592233e-06,
      "loss": 0.121,
      "step": 355560
    },
    {
      "epoch": 138.09,
      "learning_rate": 1.5886084142394822e-06,
      "loss": 0.0689,
      "step": 355570
    },
    {
      "epoch": 138.09,
      "learning_rate": 1.5880906148867314e-06,
      "loss": 0.1046,
      "step": 355580
    },
    {
      "epoch": 138.09,
      "learning_rate": 1.5875728155339806e-06,
      "loss": 0.0276,
      "step": 355590
    },
    {
      "epoch": 138.1,
      "learning_rate": 1.5870550161812298e-06,
      "loss": 0.0586,
      "step": 355600
    },
    {
      "epoch": 138.1,
      "learning_rate": 1.586537216828479e-06,
      "loss": 0.0807,
      "step": 355610
    },
    {
      "epoch": 138.1,
      "learning_rate": 1.5860194174757282e-06,
      "loss": 0.0011,
      "step": 355620
    },
    {
      "epoch": 138.11,
      "learning_rate": 1.5855016181229774e-06,
      "loss": 0.0138,
      "step": 355630
    },
    {
      "epoch": 138.11,
      "learning_rate": 1.5849838187702266e-06,
      "loss": 0.026,
      "step": 355640
    },
    {
      "epoch": 138.12,
      "learning_rate": 1.5844660194174758e-06,
      "loss": 0.0748,
      "step": 355650
    },
    {
      "epoch": 138.12,
      "learning_rate": 1.583948220064725e-06,
      "loss": 0.001,
      "step": 355660
    },
    {
      "epoch": 138.12,
      "learning_rate": 1.5834304207119742e-06,
      "loss": 0.1211,
      "step": 355670
    },
    {
      "epoch": 138.13,
      "learning_rate": 1.5829126213592233e-06,
      "loss": 0.0131,
      "step": 355680
    },
    {
      "epoch": 138.13,
      "learning_rate": 1.5823948220064725e-06,
      "loss": 0.0098,
      "step": 355690
    },
    {
      "epoch": 138.14,
      "learning_rate": 1.581877022653722e-06,
      "loss": 0.0399,
      "step": 355700
    },
    {
      "epoch": 138.14,
      "learning_rate": 1.5813592233009711e-06,
      "loss": 0.0128,
      "step": 355710
    },
    {
      "epoch": 138.14,
      "learning_rate": 1.5808414239482203e-06,
      "loss": 0.0454,
      "step": 355720
    },
    {
      "epoch": 138.15,
      "learning_rate": 1.5803236245954695e-06,
      "loss": 0.0177,
      "step": 355730
    },
    {
      "epoch": 138.15,
      "learning_rate": 1.5798058252427187e-06,
      "loss": 0.0003,
      "step": 355740
    },
    {
      "epoch": 138.16,
      "learning_rate": 1.579288025889968e-06,
      "loss": 0.0207,
      "step": 355750
    },
    {
      "epoch": 138.16,
      "learning_rate": 1.5787702265372171e-06,
      "loss": 0.0628,
      "step": 355760
    },
    {
      "epoch": 138.16,
      "learning_rate": 1.5782524271844663e-06,
      "loss": 0.0442,
      "step": 355770
    },
    {
      "epoch": 138.17,
      "learning_rate": 1.5777346278317155e-06,
      "loss": 0.1086,
      "step": 355780
    },
    {
      "epoch": 138.17,
      "learning_rate": 1.5772168284789647e-06,
      "loss": 0.0061,
      "step": 355790
    },
    {
      "epoch": 138.17,
      "learning_rate": 1.5766990291262139e-06,
      "loss": 0.0019,
      "step": 355800
    },
    {
      "epoch": 138.18,
      "learning_rate": 1.576181229773463e-06,
      "loss": 0.0834,
      "step": 355810
    },
    {
      "epoch": 138.18,
      "learning_rate": 1.5756634304207123e-06,
      "loss": 0.0473,
      "step": 355820
    },
    {
      "epoch": 138.19,
      "learning_rate": 1.5751456310679613e-06,
      "loss": 0.0174,
      "step": 355830
    },
    {
      "epoch": 138.19,
      "learning_rate": 1.5746278317152105e-06,
      "loss": 0.0478,
      "step": 355840
    },
    {
      "epoch": 138.19,
      "learning_rate": 1.5741100323624597e-06,
      "loss": 0.0028,
      "step": 355850
    },
    {
      "epoch": 138.2,
      "learning_rate": 1.5735922330097088e-06,
      "loss": 0.0296,
      "step": 355860
    },
    {
      "epoch": 138.2,
      "learning_rate": 1.573074433656958e-06,
      "loss": 0.0004,
      "step": 355870
    },
    {
      "epoch": 138.21,
      "learning_rate": 1.5725566343042072e-06,
      "loss": 0.0013,
      "step": 355880
    },
    {
      "epoch": 138.21,
      "learning_rate": 1.5720388349514564e-06,
      "loss": 0.0433,
      "step": 355890
    },
    {
      "epoch": 138.21,
      "learning_rate": 1.5715210355987056e-06,
      "loss": 0.0148,
      "step": 355900
    },
    {
      "epoch": 138.22,
      "learning_rate": 1.5710032362459548e-06,
      "loss": 0.0256,
      "step": 355910
    },
    {
      "epoch": 138.22,
      "learning_rate": 1.570485436893204e-06,
      "loss": 0.0001,
      "step": 355920
    },
    {
      "epoch": 138.23,
      "learning_rate": 1.5699676375404532e-06,
      "loss": 0.0096,
      "step": 355930
    },
    {
      "epoch": 138.23,
      "learning_rate": 1.5694498381877024e-06,
      "loss": 0.0485,
      "step": 355940
    },
    {
      "epoch": 138.23,
      "learning_rate": 1.5689320388349516e-06,
      "loss": 0.066,
      "step": 355950
    },
    {
      "epoch": 138.24,
      "learning_rate": 1.5684142394822008e-06,
      "loss": 0.0216,
      "step": 355960
    },
    {
      "epoch": 138.24,
      "learning_rate": 1.56789644012945e-06,
      "loss": 0.0763,
      "step": 355970
    },
    {
      "epoch": 138.24,
      "learning_rate": 1.5673786407766992e-06,
      "loss": 0.0011,
      "step": 355980
    },
    {
      "epoch": 138.25,
      "learning_rate": 1.5668608414239484e-06,
      "loss": 0.0316,
      "step": 355990
    },
    {
      "epoch": 138.25,
      "learning_rate": 1.5663430420711976e-06,
      "loss": 0.0486,
      "step": 356000
    },
    {
      "epoch": 138.26,
      "learning_rate": 1.5658252427184468e-06,
      "loss": 0.0001,
      "step": 356010
    },
    {
      "epoch": 138.26,
      "learning_rate": 1.565307443365696e-06,
      "loss": 0.0494,
      "step": 356020
    },
    {
      "epoch": 138.26,
      "learning_rate": 1.5647896440129452e-06,
      "loss": 0.0106,
      "step": 356030
    },
    {
      "epoch": 138.27,
      "learning_rate": 1.5642718446601943e-06,
      "loss": 0.0001,
      "step": 356040
    },
    {
      "epoch": 138.27,
      "learning_rate": 1.5637540453074435e-06,
      "loss": 0.1834,
      "step": 356050
    },
    {
      "epoch": 138.28,
      "learning_rate": 1.5632362459546927e-06,
      "loss": 0.0006,
      "step": 356060
    },
    {
      "epoch": 138.28,
      "learning_rate": 1.562718446601942e-06,
      "loss": 0.0161,
      "step": 356070
    },
    {
      "epoch": 138.28,
      "learning_rate": 1.562200647249191e-06,
      "loss": 0.0005,
      "step": 356080
    },
    {
      "epoch": 138.29,
      "learning_rate": 1.5616828478964401e-06,
      "loss": 0.0007,
      "step": 356090
    },
    {
      "epoch": 138.29,
      "learning_rate": 1.5611650485436893e-06,
      "loss": 0.0001,
      "step": 356100
    },
    {
      "epoch": 138.3,
      "learning_rate": 1.5606472491909385e-06,
      "loss": 0.0162,
      "step": 356110
    },
    {
      "epoch": 138.3,
      "learning_rate": 1.5601294498381877e-06,
      "loss": 0.0087,
      "step": 356120
    },
    {
      "epoch": 138.3,
      "learning_rate": 1.5596116504854369e-06,
      "loss": 0.023,
      "step": 356130
    },
    {
      "epoch": 138.31,
      "learning_rate": 1.559093851132686e-06,
      "loss": 0.0104,
      "step": 356140
    },
    {
      "epoch": 138.31,
      "learning_rate": 1.5585760517799353e-06,
      "loss": 0.0424,
      "step": 356150
    },
    {
      "epoch": 138.31,
      "learning_rate": 1.5580582524271845e-06,
      "loss": 0.0256,
      "step": 356160
    },
    {
      "epoch": 138.32,
      "learning_rate": 1.5575404530744337e-06,
      "loss": 0.0639,
      "step": 356170
    },
    {
      "epoch": 138.32,
      "learning_rate": 1.5570226537216829e-06,
      "loss": 0.0601,
      "step": 356180
    },
    {
      "epoch": 138.33,
      "learning_rate": 1.556504854368932e-06,
      "loss": 0.058,
      "step": 356190
    },
    {
      "epoch": 138.33,
      "learning_rate": 1.5559870550161813e-06,
      "loss": 0.016,
      "step": 356200
    },
    {
      "epoch": 138.33,
      "learning_rate": 1.5554692556634304e-06,
      "loss": 0.1038,
      "step": 356210
    },
    {
      "epoch": 138.34,
      "learning_rate": 1.5549514563106796e-06,
      "loss": 0.0928,
      "step": 356220
    },
    {
      "epoch": 138.34,
      "learning_rate": 1.5544336569579288e-06,
      "loss": 0.0543,
      "step": 356230
    },
    {
      "epoch": 138.35,
      "learning_rate": 1.553915857605178e-06,
      "loss": 0.0666,
      "step": 356240
    },
    {
      "epoch": 138.35,
      "learning_rate": 1.5533980582524272e-06,
      "loss": 0.0157,
      "step": 356250
    },
    {
      "epoch": 138.35,
      "learning_rate": 1.5528802588996766e-06,
      "loss": 0.0368,
      "step": 356260
    },
    {
      "epoch": 138.36,
      "learning_rate": 1.5523624595469258e-06,
      "loss": 0.0908,
      "step": 356270
    },
    {
      "epoch": 138.36,
      "learning_rate": 1.551844660194175e-06,
      "loss": 0.0162,
      "step": 356280
    },
    {
      "epoch": 138.37,
      "learning_rate": 1.5513268608414242e-06,
      "loss": 0.0481,
      "step": 356290
    },
    {
      "epoch": 138.37,
      "learning_rate": 1.5508090614886734e-06,
      "loss": 0.0003,
      "step": 356300
    },
    {
      "epoch": 138.37,
      "learning_rate": 1.5502912621359226e-06,
      "loss": 0.0206,
      "step": 356310
    },
    {
      "epoch": 138.38,
      "learning_rate": 1.5497734627831718e-06,
      "loss": 0.0531,
      "step": 356320
    },
    {
      "epoch": 138.38,
      "learning_rate": 1.549255663430421e-06,
      "loss": 0.012,
      "step": 356330
    },
    {
      "epoch": 138.38,
      "learning_rate": 1.5487378640776702e-06,
      "loss": 0.0159,
      "step": 356340
    },
    {
      "epoch": 138.39,
      "learning_rate": 1.5482200647249194e-06,
      "loss": 0.1658,
      "step": 356350
    },
    {
      "epoch": 138.39,
      "learning_rate": 1.5477022653721686e-06,
      "loss": 0.0481,
      "step": 356360
    },
    {
      "epoch": 138.4,
      "learning_rate": 1.5471844660194178e-06,
      "loss": 0.0003,
      "step": 356370
    },
    {
      "epoch": 138.4,
      "learning_rate": 1.546666666666667e-06,
      "loss": 0.0001,
      "step": 356380
    },
    {
      "epoch": 138.4,
      "learning_rate": 1.5461488673139162e-06,
      "loss": 0.0044,
      "step": 356390
    },
    {
      "epoch": 138.41,
      "learning_rate": 1.5456310679611651e-06,
      "loss": 0.042,
      "step": 356400
    },
    {
      "epoch": 138.41,
      "learning_rate": 1.5451132686084143e-06,
      "loss": 0.0725,
      "step": 356410
    },
    {
      "epoch": 138.42,
      "learning_rate": 1.5445954692556635e-06,
      "loss": 0.0865,
      "step": 356420
    },
    {
      "epoch": 138.42,
      "learning_rate": 1.5440776699029127e-06,
      "loss": 0.0095,
      "step": 356430
    },
    {
      "epoch": 138.42,
      "learning_rate": 1.543559870550162e-06,
      "loss": 0.001,
      "step": 356440
    },
    {
      "epoch": 138.43,
      "learning_rate": 1.5430420711974111e-06,
      "loss": 0.0761,
      "step": 356450
    },
    {
      "epoch": 138.43,
      "learning_rate": 1.5425242718446603e-06,
      "loss": 0.0259,
      "step": 356460
    },
    {
      "epoch": 138.43,
      "learning_rate": 1.5420064724919095e-06,
      "loss": 0.1094,
      "step": 356470
    },
    {
      "epoch": 138.44,
      "learning_rate": 1.5414886731391587e-06,
      "loss": 0.0313,
      "step": 356480
    },
    {
      "epoch": 138.44,
      "learning_rate": 1.5409708737864079e-06,
      "loss": 0.0009,
      "step": 356490
    },
    {
      "epoch": 138.45,
      "learning_rate": 1.540453074433657e-06,
      "loss": 0.0761,
      "step": 356500
    },
    {
      "epoch": 138.45,
      "learning_rate": 1.5399352750809063e-06,
      "loss": 0.0001,
      "step": 356510
    },
    {
      "epoch": 138.45,
      "learning_rate": 1.5394174757281555e-06,
      "loss": 0.0001,
      "step": 356520
    },
    {
      "epoch": 138.46,
      "learning_rate": 1.5388996763754047e-06,
      "loss": 0.0165,
      "step": 356530
    },
    {
      "epoch": 138.46,
      "learning_rate": 1.5383818770226539e-06,
      "loss": 0.0534,
      "step": 356540
    },
    {
      "epoch": 138.47,
      "learning_rate": 1.537864077669903e-06,
      "loss": 0.0012,
      "step": 356550
    },
    {
      "epoch": 138.47,
      "learning_rate": 1.5373462783171523e-06,
      "loss": 0.0328,
      "step": 356560
    },
    {
      "epoch": 138.47,
      "learning_rate": 1.5368284789644014e-06,
      "loss": 0.0513,
      "step": 356570
    },
    {
      "epoch": 138.48,
      "learning_rate": 1.5363106796116506e-06,
      "loss": 0.0712,
      "step": 356580
    },
    {
      "epoch": 138.48,
      "learning_rate": 1.5357928802588998e-06,
      "loss": 0.0648,
      "step": 356590
    },
    {
      "epoch": 138.49,
      "learning_rate": 1.535275080906149e-06,
      "loss": 0.0004,
      "step": 356600
    },
    {
      "epoch": 138.49,
      "learning_rate": 1.5347572815533982e-06,
      "loss": 0.0656,
      "step": 356610
    },
    {
      "epoch": 138.49,
      "learning_rate": 1.5342394822006474e-06,
      "loss": 0.0773,
      "step": 356620
    },
    {
      "epoch": 138.5,
      "learning_rate": 1.5337216828478966e-06,
      "loss": 0.0255,
      "step": 356630
    },
    {
      "epoch": 138.5,
      "learning_rate": 1.5332038834951458e-06,
      "loss": 0.0098,
      "step": 356640
    },
    {
      "epoch": 138.5,
      "learning_rate": 1.5326860841423948e-06,
      "loss": 0.0774,
      "step": 356650
    },
    {
      "epoch": 138.51,
      "learning_rate": 1.532168284789644e-06,
      "loss": 0.0368,
      "step": 356660
    },
    {
      "epoch": 138.51,
      "learning_rate": 1.5316504854368932e-06,
      "loss": 0.0162,
      "step": 356670
    },
    {
      "epoch": 138.52,
      "learning_rate": 1.5311326860841424e-06,
      "loss": 0.0017,
      "step": 356680
    },
    {
      "epoch": 138.52,
      "learning_rate": 1.5306148867313916e-06,
      "loss": 0.0792,
      "step": 356690
    },
    {
      "epoch": 138.52,
      "learning_rate": 1.5300970873786408e-06,
      "loss": 0.0703,
      "step": 356700
    },
    {
      "epoch": 138.53,
      "learning_rate": 1.52957928802589e-06,
      "loss": 0.0025,
      "step": 356710
    },
    {
      "epoch": 138.53,
      "learning_rate": 1.5290614886731392e-06,
      "loss": 0.1126,
      "step": 356720
    },
    {
      "epoch": 138.54,
      "learning_rate": 1.5285436893203883e-06,
      "loss": 0.0388,
      "step": 356730
    },
    {
      "epoch": 138.54,
      "learning_rate": 1.5280258899676375e-06,
      "loss": 0.0087,
      "step": 356740
    },
    {
      "epoch": 138.54,
      "learning_rate": 1.5275080906148867e-06,
      "loss": 0.0055,
      "step": 356750
    },
    {
      "epoch": 138.55,
      "learning_rate": 1.526990291262136e-06,
      "loss": 0.0097,
      "step": 356760
    },
    {
      "epoch": 138.55,
      "learning_rate": 1.5264724919093851e-06,
      "loss": 0.0008,
      "step": 356770
    },
    {
      "epoch": 138.56,
      "learning_rate": 1.5259546925566343e-06,
      "loss": 0.032,
      "step": 356780
    },
    {
      "epoch": 138.56,
      "learning_rate": 1.5254368932038835e-06,
      "loss": 0.0102,
      "step": 356790
    },
    {
      "epoch": 138.56,
      "learning_rate": 1.5249190938511327e-06,
      "loss": 0.0768,
      "step": 356800
    },
    {
      "epoch": 138.57,
      "learning_rate": 1.524401294498382e-06,
      "loss": 0.0109,
      "step": 356810
    },
    {
      "epoch": 138.57,
      "learning_rate": 1.5238834951456313e-06,
      "loss": 0.0747,
      "step": 356820
    },
    {
      "epoch": 138.57,
      "learning_rate": 1.5233656957928805e-06,
      "loss": 0.1053,
      "step": 356830
    },
    {
      "epoch": 138.58,
      "learning_rate": 1.5228478964401297e-06,
      "loss": 0.0002,
      "step": 356840
    },
    {
      "epoch": 138.58,
      "learning_rate": 1.522330097087379e-06,
      "loss": 0.0578,
      "step": 356850
    },
    {
      "epoch": 138.59,
      "learning_rate": 1.521812297734628e-06,
      "loss": 0.0274,
      "step": 356860
    },
    {
      "epoch": 138.59,
      "learning_rate": 1.5212944983818773e-06,
      "loss": 0.1364,
      "step": 356870
    },
    {
      "epoch": 138.59,
      "learning_rate": 1.5207766990291265e-06,
      "loss": 0.0436,
      "step": 356880
    },
    {
      "epoch": 138.6,
      "learning_rate": 1.5202588996763757e-06,
      "loss": 0.0041,
      "step": 356890
    },
    {
      "epoch": 138.6,
      "learning_rate": 1.5197411003236249e-06,
      "loss": 0.0005,
      "step": 356900
    },
    {
      "epoch": 138.61,
      "learning_rate": 1.519223300970874e-06,
      "loss": 0.0923,
      "step": 356910
    },
    {
      "epoch": 138.61,
      "learning_rate": 1.5187055016181233e-06,
      "loss": 0.0158,
      "step": 356920
    },
    {
      "epoch": 138.61,
      "learning_rate": 1.5181877022653725e-06,
      "loss": 0.0505,
      "step": 356930
    },
    {
      "epoch": 138.62,
      "learning_rate": 1.5176699029126216e-06,
      "loss": 0.0442,
      "step": 356940
    },
    {
      "epoch": 138.62,
      "learning_rate": 1.5171521035598708e-06,
      "loss": 0.0092,
      "step": 356950
    },
    {
      "epoch": 138.63,
      "learning_rate": 1.51663430420712e-06,
      "loss": 0.021,
      "step": 356960
    },
    {
      "epoch": 138.63,
      "learning_rate": 1.516116504854369e-06,
      "loss": 0.015,
      "step": 356970
    },
    {
      "epoch": 138.63,
      "learning_rate": 1.5155987055016182e-06,
      "loss": 0.0107,
      "step": 356980
    },
    {
      "epoch": 138.64,
      "learning_rate": 1.5150809061488674e-06,
      "loss": 0.2272,
      "step": 356990
    },
    {
      "epoch": 138.64,
      "learning_rate": 1.5145631067961166e-06,
      "loss": 0.028,
      "step": 357000
    },
    {
      "epoch": 138.64,
      "learning_rate": 1.5140453074433658e-06,
      "loss": 0.0672,
      "step": 357010
    },
    {
      "epoch": 138.65,
      "learning_rate": 1.513527508090615e-06,
      "loss": 0.0621,
      "step": 357020
    },
    {
      "epoch": 138.65,
      "learning_rate": 1.5130097087378642e-06,
      "loss": 0.0251,
      "step": 357030
    },
    {
      "epoch": 138.66,
      "learning_rate": 1.5124919093851134e-06,
      "loss": 0.0439,
      "step": 357040
    },
    {
      "epoch": 138.66,
      "learning_rate": 1.5119741100323626e-06,
      "loss": 0.0421,
      "step": 357050
    },
    {
      "epoch": 138.66,
      "learning_rate": 1.5114563106796118e-06,
      "loss": 0.0078,
      "step": 357060
    },
    {
      "epoch": 138.67,
      "learning_rate": 1.510938511326861e-06,
      "loss": 0.0014,
      "step": 357070
    },
    {
      "epoch": 138.67,
      "learning_rate": 1.5104207119741102e-06,
      "loss": 0.0001,
      "step": 357080
    },
    {
      "epoch": 138.68,
      "learning_rate": 1.5099029126213594e-06,
      "loss": 0.0852,
      "step": 357090
    },
    {
      "epoch": 138.68,
      "learning_rate": 1.5093851132686085e-06,
      "loss": 0.0589,
      "step": 357100
    },
    {
      "epoch": 138.68,
      "learning_rate": 1.5088673139158577e-06,
      "loss": 0.0497,
      "step": 357110
    },
    {
      "epoch": 138.69,
      "learning_rate": 1.508349514563107e-06,
      "loss": 0.0028,
      "step": 357120
    },
    {
      "epoch": 138.69,
      "learning_rate": 1.5078317152103561e-06,
      "loss": 0.0714,
      "step": 357130
    },
    {
      "epoch": 138.7,
      "learning_rate": 1.5073139158576053e-06,
      "loss": 0.0453,
      "step": 357140
    },
    {
      "epoch": 138.7,
      "learning_rate": 1.5067961165048545e-06,
      "loss": 0.0007,
      "step": 357150
    },
    {
      "epoch": 138.7,
      "learning_rate": 1.5062783171521037e-06,
      "loss": 0.0422,
      "step": 357160
    },
    {
      "epoch": 138.71,
      "learning_rate": 1.505760517799353e-06,
      "loss": 0.0411,
      "step": 357170
    },
    {
      "epoch": 138.71,
      "learning_rate": 1.505242718446602e-06,
      "loss": 0.0083,
      "step": 357180
    },
    {
      "epoch": 138.71,
      "learning_rate": 1.5047249190938513e-06,
      "loss": 0.024,
      "step": 357190
    },
    {
      "epoch": 138.72,
      "learning_rate": 1.5042071197411005e-06,
      "loss": 0.0341,
      "step": 357200
    },
    {
      "epoch": 138.72,
      "learning_rate": 1.5036893203883497e-06,
      "loss": 0.0577,
      "step": 357210
    },
    {
      "epoch": 138.73,
      "learning_rate": 1.5031715210355987e-06,
      "loss": 0.0862,
      "step": 357220
    },
    {
      "epoch": 138.73,
      "learning_rate": 1.5026537216828479e-06,
      "loss": 0.0747,
      "step": 357230
    },
    {
      "epoch": 138.73,
      "learning_rate": 1.502135922330097e-06,
      "loss": 0.0769,
      "step": 357240
    },
    {
      "epoch": 138.74,
      "learning_rate": 1.5016181229773463e-06,
      "loss": 0.0183,
      "step": 357250
    },
    {
      "epoch": 138.74,
      "learning_rate": 1.5011003236245954e-06,
      "loss": 0.0014,
      "step": 357260
    },
    {
      "epoch": 138.75,
      "learning_rate": 1.5005825242718446e-06,
      "loss": 0.0159,
      "step": 357270
    },
    {
      "epoch": 138.75,
      "learning_rate": 1.5000647249190938e-06,
      "loss": 0.0853,
      "step": 357280
    },
    {
      "epoch": 138.75,
      "learning_rate": 1.499546925566343e-06,
      "loss": 0.047,
      "step": 357290
    },
    {
      "epoch": 138.76,
      "learning_rate": 1.4990291262135922e-06,
      "loss": 0.0161,
      "step": 357300
    },
    {
      "epoch": 138.76,
      "learning_rate": 1.4985113268608414e-06,
      "loss": 0.082,
      "step": 357310
    },
    {
      "epoch": 138.77,
      "learning_rate": 1.4979935275080906e-06,
      "loss": 0.0006,
      "step": 357320
    },
    {
      "epoch": 138.77,
      "learning_rate": 1.4974757281553398e-06,
      "loss": 0.0374,
      "step": 357330
    },
    {
      "epoch": 138.77,
      "learning_rate": 1.496957928802589e-06,
      "loss": 0.009,
      "step": 357340
    },
    {
      "epoch": 138.78,
      "learning_rate": 1.4964401294498382e-06,
      "loss": 0.0097,
      "step": 357350
    },
    {
      "epoch": 138.78,
      "learning_rate": 1.4959223300970874e-06,
      "loss": 0.0073,
      "step": 357360
    },
    {
      "epoch": 138.78,
      "learning_rate": 1.4954045307443366e-06,
      "loss": 0.1321,
      "step": 357370
    },
    {
      "epoch": 138.79,
      "learning_rate": 1.494886731391586e-06,
      "loss": 0.0783,
      "step": 357380
    },
    {
      "epoch": 138.79,
      "learning_rate": 1.4943689320388352e-06,
      "loss": 0.0184,
      "step": 357390
    },
    {
      "epoch": 138.8,
      "learning_rate": 1.4938511326860844e-06,
      "loss": 0.047,
      "step": 357400
    },
    {
      "epoch": 138.8,
      "learning_rate": 1.4933333333333336e-06,
      "loss": 0.0018,
      "step": 357410
    },
    {
      "epoch": 138.8,
      "learning_rate": 1.4928155339805828e-06,
      "loss": 0.0012,
      "step": 357420
    },
    {
      "epoch": 138.81,
      "learning_rate": 1.492297734627832e-06,
      "loss": 0.009,
      "step": 357430
    },
    {
      "epoch": 138.81,
      "learning_rate": 1.4917799352750812e-06,
      "loss": 0.0256,
      "step": 357440
    },
    {
      "epoch": 138.82,
      "learning_rate": 1.4912621359223304e-06,
      "loss": 0.0703,
      "step": 357450
    },
    {
      "epoch": 138.82,
      "learning_rate": 1.4907443365695795e-06,
      "loss": 0.0285,
      "step": 357460
    },
    {
      "epoch": 138.82,
      "learning_rate": 1.4902265372168287e-06,
      "loss": 0.0019,
      "step": 357470
    },
    {
      "epoch": 138.83,
      "learning_rate": 1.489708737864078e-06,
      "loss": 0.1281,
      "step": 357480
    },
    {
      "epoch": 138.83,
      "learning_rate": 1.4891909385113271e-06,
      "loss": 0.0204,
      "step": 357490
    },
    {
      "epoch": 138.83,
      "learning_rate": 1.4886731391585763e-06,
      "loss": 0.003,
      "step": 357500
    },
    {
      "epoch": 138.84,
      "learning_rate": 1.4881553398058255e-06,
      "loss": 0.0438,
      "step": 357510
    },
    {
      "epoch": 138.84,
      "learning_rate": 1.4876375404530747e-06,
      "loss": 0.0191,
      "step": 357520
    },
    {
      "epoch": 138.85,
      "learning_rate": 1.487119741100324e-06,
      "loss": 0.0244,
      "step": 357530
    },
    {
      "epoch": 138.85,
      "learning_rate": 1.4866019417475729e-06,
      "loss": 0.0996,
      "step": 357540
    },
    {
      "epoch": 138.85,
      "learning_rate": 1.486084142394822e-06,
      "loss": 0.0333,
      "step": 357550
    },
    {
      "epoch": 138.86,
      "learning_rate": 1.4855663430420713e-06,
      "loss": 0.0183,
      "step": 357560
    },
    {
      "epoch": 138.86,
      "learning_rate": 1.4850485436893205e-06,
      "loss": 0.0018,
      "step": 357570
    },
    {
      "epoch": 138.87,
      "learning_rate": 1.4845307443365697e-06,
      "loss": 0.0169,
      "step": 357580
    },
    {
      "epoch": 138.87,
      "learning_rate": 1.4840129449838189e-06,
      "loss": 0.0385,
      "step": 357590
    },
    {
      "epoch": 138.87,
      "learning_rate": 1.483495145631068e-06,
      "loss": 0.0217,
      "step": 357600
    },
    {
      "epoch": 138.88,
      "learning_rate": 1.4829773462783173e-06,
      "loss": 0.0063,
      "step": 357610
    },
    {
      "epoch": 138.88,
      "learning_rate": 1.4824595469255664e-06,
      "loss": 0.0496,
      "step": 357620
    },
    {
      "epoch": 138.89,
      "learning_rate": 1.4819417475728156e-06,
      "loss": 0.0607,
      "step": 357630
    },
    {
      "epoch": 138.89,
      "learning_rate": 1.4814239482200648e-06,
      "loss": 0.0002,
      "step": 357640
    },
    {
      "epoch": 138.89,
      "learning_rate": 1.480906148867314e-06,
      "loss": 0.0617,
      "step": 357650
    },
    {
      "epoch": 138.9,
      "learning_rate": 1.4803883495145632e-06,
      "loss": 0.011,
      "step": 357660
    },
    {
      "epoch": 138.9,
      "learning_rate": 1.4798705501618124e-06,
      "loss": 0.0096,
      "step": 357670
    },
    {
      "epoch": 138.9,
      "learning_rate": 1.4793527508090616e-06,
      "loss": 0.0476,
      "step": 357680
    },
    {
      "epoch": 138.91,
      "learning_rate": 1.4788349514563108e-06,
      "loss": 0.0601,
      "step": 357690
    },
    {
      "epoch": 138.91,
      "learning_rate": 1.47831715210356e-06,
      "loss": 0.0453,
      "step": 357700
    },
    {
      "epoch": 138.92,
      "learning_rate": 1.4777993527508092e-06,
      "loss": 0.0121,
      "step": 357710
    },
    {
      "epoch": 138.92,
      "learning_rate": 1.4772815533980584e-06,
      "loss": 0.0207,
      "step": 357720
    },
    {
      "epoch": 138.92,
      "learning_rate": 1.4767637540453076e-06,
      "loss": 0.0003,
      "step": 357730
    },
    {
      "epoch": 138.93,
      "learning_rate": 1.4762459546925568e-06,
      "loss": 0.0861,
      "step": 357740
    },
    {
      "epoch": 138.93,
      "learning_rate": 1.475728155339806e-06,
      "loss": 0.1142,
      "step": 357750
    },
    {
      "epoch": 138.94,
      "learning_rate": 1.4752103559870552e-06,
      "loss": 0.0296,
      "step": 357760
    },
    {
      "epoch": 138.94,
      "learning_rate": 1.4746925566343044e-06,
      "loss": 0.0009,
      "step": 357770
    },
    {
      "epoch": 138.94,
      "learning_rate": 1.4741747572815536e-06,
      "loss": 0.0881,
      "step": 357780
    },
    {
      "epoch": 138.95,
      "learning_rate": 1.4736569579288025e-06,
      "loss": 0.0243,
      "step": 357790
    },
    {
      "epoch": 138.95,
      "learning_rate": 1.4731391585760517e-06,
      "loss": 0.0855,
      "step": 357800
    },
    {
      "epoch": 138.96,
      "learning_rate": 1.472621359223301e-06,
      "loss": 0.0606,
      "step": 357810
    },
    {
      "epoch": 138.96,
      "learning_rate": 1.4721035598705501e-06,
      "loss": 0.1207,
      "step": 357820
    },
    {
      "epoch": 138.96,
      "learning_rate": 1.4715857605177993e-06,
      "loss": 0.0029,
      "step": 357830
    },
    {
      "epoch": 138.97,
      "learning_rate": 1.4710679611650485e-06,
      "loss": 0.0087,
      "step": 357840
    },
    {
      "epoch": 138.97,
      "learning_rate": 1.4705501618122977e-06,
      "loss": 0.0027,
      "step": 357850
    },
    {
      "epoch": 138.97,
      "learning_rate": 1.470032362459547e-06,
      "loss": 0.0969,
      "step": 357860
    },
    {
      "epoch": 138.98,
      "learning_rate": 1.469514563106796e-06,
      "loss": 0.0683,
      "step": 357870
    },
    {
      "epoch": 138.98,
      "learning_rate": 1.4689967637540453e-06,
      "loss": 0.0688,
      "step": 357880
    },
    {
      "epoch": 138.99,
      "learning_rate": 1.4684789644012945e-06,
      "loss": 0.0118,
      "step": 357890
    },
    {
      "epoch": 138.99,
      "learning_rate": 1.4679611650485437e-06,
      "loss": 0.0951,
      "step": 357900
    },
    {
      "epoch": 138.99,
      "learning_rate": 1.4674433656957929e-06,
      "loss": 0.0537,
      "step": 357910
    },
    {
      "epoch": 139.0,
      "learning_rate": 1.466925566343042e-06,
      "loss": 0.0705,
      "step": 357920
    },
    {
      "epoch": 139.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.38591042160987854,
      "eval_runtime": 8.2344,
      "eval_samples_per_second": 441.439,
      "eval_steps_per_second": 55.256,
      "step": 357925
    },
    {
      "epoch": 139.0,
      "learning_rate": 1.4664077669902913e-06,
      "loss": 0.0276,
      "step": 357930
    },
    {
      "epoch": 139.01,
      "learning_rate": 1.4658899676375407e-06,
      "loss": 0.1208,
      "step": 357940
    },
    {
      "epoch": 139.01,
      "learning_rate": 1.4653721682847899e-06,
      "loss": 0.2903,
      "step": 357950
    },
    {
      "epoch": 139.01,
      "learning_rate": 1.464854368932039e-06,
      "loss": 0.1264,
      "step": 357960
    },
    {
      "epoch": 139.02,
      "learning_rate": 1.4643365695792883e-06,
      "loss": 0.0036,
      "step": 357970
    },
    {
      "epoch": 139.02,
      "learning_rate": 1.4638187702265375e-06,
      "loss": 0.0166,
      "step": 357980
    },
    {
      "epoch": 139.03,
      "learning_rate": 1.4633009708737866e-06,
      "loss": 0.0001,
      "step": 357990
    },
    {
      "epoch": 139.03,
      "learning_rate": 1.4627831715210358e-06,
      "loss": 0.0078,
      "step": 358000
    },
    {
      "epoch": 139.03,
      "learning_rate": 1.462265372168285e-06,
      "loss": 0.0108,
      "step": 358010
    },
    {
      "epoch": 139.04,
      "learning_rate": 1.4617475728155342e-06,
      "loss": 0.0659,
      "step": 358020
    },
    {
      "epoch": 139.04,
      "learning_rate": 1.4612297734627834e-06,
      "loss": 0.0258,
      "step": 358030
    },
    {
      "epoch": 139.04,
      "learning_rate": 1.4607119741100326e-06,
      "loss": 0.0542,
      "step": 358040
    },
    {
      "epoch": 139.05,
      "learning_rate": 1.4601941747572818e-06,
      "loss": 0.0226,
      "step": 358050
    },
    {
      "epoch": 139.05,
      "learning_rate": 1.459676375404531e-06,
      "loss": 0.0104,
      "step": 358060
    },
    {
      "epoch": 139.06,
      "learning_rate": 1.4591585760517802e-06,
      "loss": 0.082,
      "step": 358070
    },
    {
      "epoch": 139.06,
      "learning_rate": 1.4586407766990294e-06,
      "loss": 0.0002,
      "step": 358080
    },
    {
      "epoch": 139.06,
      "learning_rate": 1.4581229773462786e-06,
      "loss": 0.0782,
      "step": 358090
    },
    {
      "epoch": 139.07,
      "learning_rate": 1.4576051779935278e-06,
      "loss": 0.089,
      "step": 358100
    },
    {
      "epoch": 139.07,
      "learning_rate": 1.4570873786407768e-06,
      "loss": 0.2052,
      "step": 358110
    },
    {
      "epoch": 139.08,
      "learning_rate": 1.456569579288026e-06,
      "loss": 0.1633,
      "step": 358120
    },
    {
      "epoch": 139.08,
      "learning_rate": 1.4560517799352752e-06,
      "loss": 0.0809,
      "step": 358130
    },
    {
      "epoch": 139.08,
      "learning_rate": 1.4555339805825244e-06,
      "loss": 0.1189,
      "step": 358140
    },
    {
      "epoch": 139.09,
      "learning_rate": 1.4550161812297735e-06,
      "loss": 0.0309,
      "step": 358150
    },
    {
      "epoch": 139.09,
      "learning_rate": 1.4544983818770227e-06,
      "loss": 0.0925,
      "step": 358160
    },
    {
      "epoch": 139.1,
      "learning_rate": 1.453980582524272e-06,
      "loss": 0.0001,
      "step": 358170
    },
    {
      "epoch": 139.1,
      "learning_rate": 1.4534627831715211e-06,
      "loss": 0.0284,
      "step": 358180
    },
    {
      "epoch": 139.1,
      "learning_rate": 1.4529449838187703e-06,
      "loss": 0.0446,
      "step": 358190
    },
    {
      "epoch": 139.11,
      "learning_rate": 1.4524271844660195e-06,
      "loss": 0.121,
      "step": 358200
    },
    {
      "epoch": 139.11,
      "learning_rate": 1.4519093851132687e-06,
      "loss": 0.0519,
      "step": 358210
    },
    {
      "epoch": 139.11,
      "learning_rate": 1.451391585760518e-06,
      "loss": 0.0827,
      "step": 358220
    },
    {
      "epoch": 139.12,
      "learning_rate": 1.450873786407767e-06,
      "loss": 0.0002,
      "step": 358230
    },
    {
      "epoch": 139.12,
      "learning_rate": 1.4503559870550163e-06,
      "loss": 0.0007,
      "step": 358240
    },
    {
      "epoch": 139.13,
      "learning_rate": 1.4498381877022655e-06,
      "loss": 0.0289,
      "step": 358250
    },
    {
      "epoch": 139.13,
      "learning_rate": 1.4493203883495147e-06,
      "loss": 0.0197,
      "step": 358260
    },
    {
      "epoch": 139.13,
      "learning_rate": 1.4488025889967639e-06,
      "loss": 0.0681,
      "step": 358270
    },
    {
      "epoch": 139.14,
      "learning_rate": 1.448284789644013e-06,
      "loss": 0.0001,
      "step": 358280
    },
    {
      "epoch": 139.14,
      "learning_rate": 1.4477669902912623e-06,
      "loss": 0.0688,
      "step": 358290
    },
    {
      "epoch": 139.15,
      "learning_rate": 1.4472491909385115e-06,
      "loss": 0.0186,
      "step": 358300
    },
    {
      "epoch": 139.15,
      "learning_rate": 1.4467313915857607e-06,
      "loss": 0.074,
      "step": 358310
    },
    {
      "epoch": 139.15,
      "learning_rate": 1.4462135922330099e-06,
      "loss": 0.0103,
      "step": 358320
    },
    {
      "epoch": 139.16,
      "learning_rate": 1.445695792880259e-06,
      "loss": 0.1531,
      "step": 358330
    },
    {
      "epoch": 139.16,
      "learning_rate": 1.4451779935275082e-06,
      "loss": 0.0272,
      "step": 358340
    },
    {
      "epoch": 139.17,
      "learning_rate": 1.4446601941747574e-06,
      "loss": 0.0892,
      "step": 358350
    },
    {
      "epoch": 139.17,
      "learning_rate": 1.4441423948220064e-06,
      "loss": 0.0164,
      "step": 358360
    },
    {
      "epoch": 139.17,
      "learning_rate": 1.4436245954692556e-06,
      "loss": 0.132,
      "step": 358370
    },
    {
      "epoch": 139.18,
      "learning_rate": 1.4431067961165048e-06,
      "loss": 0.0544,
      "step": 358380
    },
    {
      "epoch": 139.18,
      "learning_rate": 1.442588996763754e-06,
      "loss": 0.0156,
      "step": 358390
    },
    {
      "epoch": 139.18,
      "learning_rate": 1.4420711974110032e-06,
      "loss": 0.0113,
      "step": 358400
    },
    {
      "epoch": 139.19,
      "learning_rate": 1.4415533980582524e-06,
      "loss": 0.0148,
      "step": 358410
    },
    {
      "epoch": 139.19,
      "learning_rate": 1.4410355987055016e-06,
      "loss": 0.0616,
      "step": 358420
    },
    {
      "epoch": 139.2,
      "learning_rate": 1.4405177993527508e-06,
      "loss": 0.0456,
      "step": 358430
    },
    {
      "epoch": 139.2,
      "learning_rate": 1.44e-06,
      "loss": 0.0156,
      "step": 358440
    },
    {
      "epoch": 139.2,
      "learning_rate": 1.4394822006472492e-06,
      "loss": 0.1591,
      "step": 358450
    },
    {
      "epoch": 139.21,
      "learning_rate": 1.4389644012944984e-06,
      "loss": 0.0206,
      "step": 358460
    },
    {
      "epoch": 139.21,
      "learning_rate": 1.4384466019417476e-06,
      "loss": 0.0004,
      "step": 358470
    },
    {
      "epoch": 139.22,
      "learning_rate": 1.4379288025889968e-06,
      "loss": 0.1253,
      "step": 358480
    },
    {
      "epoch": 139.22,
      "learning_rate": 1.437411003236246e-06,
      "loss": 0.0898,
      "step": 358490
    },
    {
      "epoch": 139.22,
      "learning_rate": 1.4368932038834951e-06,
      "loss": 0.0154,
      "step": 358500
    },
    {
      "epoch": 139.23,
      "learning_rate": 1.4363754045307446e-06,
      "loss": 0.0464,
      "step": 358510
    },
    {
      "epoch": 139.23,
      "learning_rate": 1.4358576051779937e-06,
      "loss": 0.016,
      "step": 358520
    },
    {
      "epoch": 139.23,
      "learning_rate": 1.435339805825243e-06,
      "loss": 0.2403,
      "step": 358530
    },
    {
      "epoch": 139.24,
      "learning_rate": 1.4348220064724921e-06,
      "loss": 0.018,
      "step": 358540
    },
    {
      "epoch": 139.24,
      "learning_rate": 1.4343042071197413e-06,
      "loss": 0.0691,
      "step": 358550
    },
    {
      "epoch": 139.25,
      "learning_rate": 1.4337864077669905e-06,
      "loss": 0.064,
      "step": 358560
    },
    {
      "epoch": 139.25,
      "learning_rate": 1.4332686084142397e-06,
      "loss": 0.0306,
      "step": 358570
    },
    {
      "epoch": 139.25,
      "learning_rate": 1.432750809061489e-06,
      "loss": 0.0499,
      "step": 358580
    },
    {
      "epoch": 139.26,
      "learning_rate": 1.4322330097087381e-06,
      "loss": 0.0187,
      "step": 358590
    },
    {
      "epoch": 139.26,
      "learning_rate": 1.4317152103559873e-06,
      "loss": 0.0226,
      "step": 358600
    },
    {
      "epoch": 139.27,
      "learning_rate": 1.4311974110032365e-06,
      "loss": 0.0006,
      "step": 358610
    },
    {
      "epoch": 139.27,
      "learning_rate": 1.4306796116504857e-06,
      "loss": 0.0001,
      "step": 358620
    },
    {
      "epoch": 139.27,
      "learning_rate": 1.4301618122977349e-06,
      "loss": 0.0472,
      "step": 358630
    },
    {
      "epoch": 139.28,
      "learning_rate": 1.429644012944984e-06,
      "loss": 0.0094,
      "step": 358640
    },
    {
      "epoch": 139.28,
      "learning_rate": 1.4291262135922333e-06,
      "loss": 0.1269,
      "step": 358650
    },
    {
      "epoch": 139.29,
      "learning_rate": 1.4286084142394825e-06,
      "loss": 0.001,
      "step": 358660
    },
    {
      "epoch": 139.29,
      "learning_rate": 1.4280906148867317e-06,
      "loss": 0.0696,
      "step": 358670
    },
    {
      "epoch": 139.29,
      "learning_rate": 1.4275728155339806e-06,
      "loss": 0.0093,
      "step": 358680
    },
    {
      "epoch": 139.3,
      "learning_rate": 1.4270550161812298e-06,
      "loss": 0.042,
      "step": 358690
    },
    {
      "epoch": 139.3,
      "learning_rate": 1.426537216828479e-06,
      "loss": 0.0488,
      "step": 358700
    },
    {
      "epoch": 139.3,
      "learning_rate": 1.4260194174757282e-06,
      "loss": 0.0228,
      "step": 358710
    },
    {
      "epoch": 139.31,
      "learning_rate": 1.4255016181229774e-06,
      "loss": 0.0143,
      "step": 358720
    },
    {
      "epoch": 139.31,
      "learning_rate": 1.4249838187702266e-06,
      "loss": 0.0002,
      "step": 358730
    },
    {
      "epoch": 139.32,
      "learning_rate": 1.4244660194174758e-06,
      "loss": 0.0938,
      "step": 358740
    },
    {
      "epoch": 139.32,
      "learning_rate": 1.423948220064725e-06,
      "loss": 0.0238,
      "step": 358750
    },
    {
      "epoch": 139.32,
      "learning_rate": 1.4234304207119742e-06,
      "loss": 0.0109,
      "step": 358760
    },
    {
      "epoch": 139.33,
      "learning_rate": 1.4229126213592234e-06,
      "loss": 0.064,
      "step": 358770
    },
    {
      "epoch": 139.33,
      "learning_rate": 1.4223948220064726e-06,
      "loss": 0.0028,
      "step": 358780
    },
    {
      "epoch": 139.34,
      "learning_rate": 1.4218770226537218e-06,
      "loss": 0.0102,
      "step": 358790
    },
    {
      "epoch": 139.34,
      "learning_rate": 1.421359223300971e-06,
      "loss": 0.0166,
      "step": 358800
    },
    {
      "epoch": 139.34,
      "learning_rate": 1.4208414239482202e-06,
      "loss": 0.0777,
      "step": 358810
    },
    {
      "epoch": 139.35,
      "learning_rate": 1.4203236245954694e-06,
      "loss": 0.0245,
      "step": 358820
    },
    {
      "epoch": 139.35,
      "learning_rate": 1.4198058252427186e-06,
      "loss": 0.0139,
      "step": 358830
    },
    {
      "epoch": 139.36,
      "learning_rate": 1.4192880258899678e-06,
      "loss": 0.0197,
      "step": 358840
    },
    {
      "epoch": 139.36,
      "learning_rate": 1.418770226537217e-06,
      "loss": 0.0275,
      "step": 358850
    },
    {
      "epoch": 139.36,
      "learning_rate": 1.4182524271844661e-06,
      "loss": 0.0092,
      "step": 358860
    },
    {
      "epoch": 139.37,
      "learning_rate": 1.4177346278317153e-06,
      "loss": 0.0892,
      "step": 358870
    },
    {
      "epoch": 139.37,
      "learning_rate": 1.4172168284789645e-06,
      "loss": 0.1016,
      "step": 358880
    },
    {
      "epoch": 139.37,
      "learning_rate": 1.4166990291262137e-06,
      "loss": 0.0087,
      "step": 358890
    },
    {
      "epoch": 139.38,
      "learning_rate": 1.416181229773463e-06,
      "loss": 0.0125,
      "step": 358900
    },
    {
      "epoch": 139.38,
      "learning_rate": 1.4156634304207121e-06,
      "loss": 0.0207,
      "step": 358910
    },
    {
      "epoch": 139.39,
      "learning_rate": 1.4151456310679613e-06,
      "loss": 0.0163,
      "step": 358920
    },
    {
      "epoch": 139.39,
      "learning_rate": 1.4146278317152103e-06,
      "loss": 0.0979,
      "step": 358930
    },
    {
      "epoch": 139.39,
      "learning_rate": 1.4141100323624595e-06,
      "loss": 0.0313,
      "step": 358940
    },
    {
      "epoch": 139.4,
      "learning_rate": 1.4135922330097087e-06,
      "loss": 0.0019,
      "step": 358950
    },
    {
      "epoch": 139.4,
      "learning_rate": 1.4130744336569579e-06,
      "loss": 0.0579,
      "step": 358960
    },
    {
      "epoch": 139.41,
      "learning_rate": 1.412556634304207e-06,
      "loss": 0.1496,
      "step": 358970
    },
    {
      "epoch": 139.41,
      "learning_rate": 1.4120388349514563e-06,
      "loss": 0.0754,
      "step": 358980
    },
    {
      "epoch": 139.41,
      "learning_rate": 1.4115210355987055e-06,
      "loss": 0.0077,
      "step": 358990
    },
    {
      "epoch": 139.42,
      "learning_rate": 1.4110032362459547e-06,
      "loss": 0.0007,
      "step": 359000
    },
    {
      "epoch": 139.42,
      "learning_rate": 1.4104854368932039e-06,
      "loss": 0.0029,
      "step": 359010
    },
    {
      "epoch": 139.43,
      "learning_rate": 1.409967637540453e-06,
      "loss": 0.0533,
      "step": 359020
    },
    {
      "epoch": 139.43,
      "learning_rate": 1.4094498381877022e-06,
      "loss": 0.0633,
      "step": 359030
    },
    {
      "epoch": 139.43,
      "learning_rate": 1.4089320388349514e-06,
      "loss": 0.0023,
      "step": 359040
    },
    {
      "epoch": 139.44,
      "learning_rate": 1.4084142394822006e-06,
      "loss": 0.0082,
      "step": 359050
    },
    {
      "epoch": 139.44,
      "learning_rate": 1.4078964401294498e-06,
      "loss": 0.0251,
      "step": 359060
    },
    {
      "epoch": 139.44,
      "learning_rate": 1.4073786407766992e-06,
      "loss": 0.0002,
      "step": 359070
    },
    {
      "epoch": 139.45,
      "learning_rate": 1.4068608414239484e-06,
      "loss": 0.0728,
      "step": 359080
    },
    {
      "epoch": 139.45,
      "learning_rate": 1.4063430420711976e-06,
      "loss": 0.052,
      "step": 359090
    },
    {
      "epoch": 139.46,
      "learning_rate": 1.4058252427184468e-06,
      "loss": 0.0082,
      "step": 359100
    },
    {
      "epoch": 139.46,
      "learning_rate": 1.405307443365696e-06,
      "loss": 0.0266,
      "step": 359110
    },
    {
      "epoch": 139.46,
      "learning_rate": 1.4047896440129452e-06,
      "loss": 0.0015,
      "step": 359120
    },
    {
      "epoch": 139.47,
      "learning_rate": 1.4042718446601944e-06,
      "loss": 0.0001,
      "step": 359130
    },
    {
      "epoch": 139.47,
      "learning_rate": 1.4037540453074436e-06,
      "loss": 0.0638,
      "step": 359140
    },
    {
      "epoch": 139.48,
      "learning_rate": 1.4032362459546928e-06,
      "loss": 0.0665,
      "step": 359150
    },
    {
      "epoch": 139.48,
      "learning_rate": 1.402718446601942e-06,
      "loss": 0.0964,
      "step": 359160
    },
    {
      "epoch": 139.48,
      "learning_rate": 1.4022006472491912e-06,
      "loss": 0.0143,
      "step": 359170
    },
    {
      "epoch": 139.49,
      "learning_rate": 1.4016828478964404e-06,
      "loss": 0.0819,
      "step": 359180
    },
    {
      "epoch": 139.49,
      "learning_rate": 1.4011650485436896e-06,
      "loss": 0.2471,
      "step": 359190
    },
    {
      "epoch": 139.5,
      "learning_rate": 1.4006472491909388e-06,
      "loss": 0.0301,
      "step": 359200
    },
    {
      "epoch": 139.5,
      "learning_rate": 1.400129449838188e-06,
      "loss": 0.1329,
      "step": 359210
    },
    {
      "epoch": 139.5,
      "learning_rate": 1.3996116504854372e-06,
      "loss": 0.0002,
      "step": 359220
    },
    {
      "epoch": 139.51,
      "learning_rate": 1.3990938511326863e-06,
      "loss": 0.0815,
      "step": 359230
    },
    {
      "epoch": 139.51,
      "learning_rate": 1.3985760517799355e-06,
      "loss": 0.0328,
      "step": 359240
    },
    {
      "epoch": 139.51,
      "learning_rate": 1.3980582524271845e-06,
      "loss": 0.0276,
      "step": 359250
    },
    {
      "epoch": 139.52,
      "learning_rate": 1.3975404530744337e-06,
      "loss": 0.0588,
      "step": 359260
    },
    {
      "epoch": 139.52,
      "learning_rate": 1.397022653721683e-06,
      "loss": 0.0009,
      "step": 359270
    },
    {
      "epoch": 139.53,
      "learning_rate": 1.396504854368932e-06,
      "loss": 0.0758,
      "step": 359280
    },
    {
      "epoch": 139.53,
      "learning_rate": 1.3959870550161813e-06,
      "loss": 0.0159,
      "step": 359290
    },
    {
      "epoch": 139.53,
      "learning_rate": 1.3954692556634305e-06,
      "loss": 0.0862,
      "step": 359300
    },
    {
      "epoch": 139.54,
      "learning_rate": 1.3949514563106797e-06,
      "loss": 0.0182,
      "step": 359310
    },
    {
      "epoch": 139.54,
      "learning_rate": 1.3944336569579289e-06,
      "loss": 0.0002,
      "step": 359320
    },
    {
      "epoch": 139.55,
      "learning_rate": 1.393915857605178e-06,
      "loss": 0.0008,
      "step": 359330
    },
    {
      "epoch": 139.55,
      "learning_rate": 1.3933980582524273e-06,
      "loss": 0.2292,
      "step": 359340
    },
    {
      "epoch": 139.55,
      "learning_rate": 1.3928802588996765e-06,
      "loss": 0.0315,
      "step": 359350
    },
    {
      "epoch": 139.56,
      "learning_rate": 1.3923624595469257e-06,
      "loss": 0.0048,
      "step": 359360
    },
    {
      "epoch": 139.56,
      "learning_rate": 1.3918446601941749e-06,
      "loss": 0.1813,
      "step": 359370
    },
    {
      "epoch": 139.57,
      "learning_rate": 1.391326860841424e-06,
      "loss": 0.0003,
      "step": 359380
    },
    {
      "epoch": 139.57,
      "learning_rate": 1.3908090614886732e-06,
      "loss": 0.0484,
      "step": 359390
    },
    {
      "epoch": 139.57,
      "learning_rate": 1.3902912621359224e-06,
      "loss": 0.0156,
      "step": 359400
    },
    {
      "epoch": 139.58,
      "learning_rate": 1.3897734627831716e-06,
      "loss": 0.0731,
      "step": 359410
    },
    {
      "epoch": 139.58,
      "learning_rate": 1.3892556634304208e-06,
      "loss": 0.0161,
      "step": 359420
    },
    {
      "epoch": 139.58,
      "learning_rate": 1.38873786407767e-06,
      "loss": 0.0599,
      "step": 359430
    },
    {
      "epoch": 139.59,
      "learning_rate": 1.3882200647249192e-06,
      "loss": 0.0191,
      "step": 359440
    },
    {
      "epoch": 139.59,
      "learning_rate": 1.3877022653721684e-06,
      "loss": 0.0104,
      "step": 359450
    },
    {
      "epoch": 139.6,
      "learning_rate": 1.3871844660194176e-06,
      "loss": 0.0532,
      "step": 359460
    },
    {
      "epoch": 139.6,
      "learning_rate": 1.3866666666666668e-06,
      "loss": 0.073,
      "step": 359470
    },
    {
      "epoch": 139.6,
      "learning_rate": 1.386148867313916e-06,
      "loss": 0.0922,
      "step": 359480
    },
    {
      "epoch": 139.61,
      "learning_rate": 1.3856310679611652e-06,
      "loss": 0.0166,
      "step": 359490
    },
    {
      "epoch": 139.61,
      "learning_rate": 1.3851132686084142e-06,
      "loss": 0.0235,
      "step": 359500
    },
    {
      "epoch": 139.62,
      "learning_rate": 1.3845954692556634e-06,
      "loss": 0.0178,
      "step": 359510
    },
    {
      "epoch": 139.62,
      "learning_rate": 1.3840776699029126e-06,
      "loss": 0.0336,
      "step": 359520
    },
    {
      "epoch": 139.62,
      "learning_rate": 1.3835598705501618e-06,
      "loss": 0.0191,
      "step": 359530
    },
    {
      "epoch": 139.63,
      "learning_rate": 1.383042071197411e-06,
      "loss": 0.0302,
      "step": 359540
    },
    {
      "epoch": 139.63,
      "learning_rate": 1.3825242718446601e-06,
      "loss": 0.0001,
      "step": 359550
    },
    {
      "epoch": 139.63,
      "learning_rate": 1.3820064724919093e-06,
      "loss": 0.0008,
      "step": 359560
    },
    {
      "epoch": 139.64,
      "learning_rate": 1.3814886731391585e-06,
      "loss": 0.0715,
      "step": 359570
    },
    {
      "epoch": 139.64,
      "learning_rate": 1.3809708737864077e-06,
      "loss": 0.0001,
      "step": 359580
    },
    {
      "epoch": 139.65,
      "learning_rate": 1.380453074433657e-06,
      "loss": 0.0118,
      "step": 359590
    },
    {
      "epoch": 139.65,
      "learning_rate": 1.3799352750809061e-06,
      "loss": 0.0251,
      "step": 359600
    },
    {
      "epoch": 139.65,
      "learning_rate": 1.3794174757281553e-06,
      "loss": 0.0266,
      "step": 359610
    },
    {
      "epoch": 139.66,
      "learning_rate": 1.3788996763754045e-06,
      "loss": 0.0256,
      "step": 359620
    },
    {
      "epoch": 139.66,
      "learning_rate": 1.378381877022654e-06,
      "loss": 0.1162,
      "step": 359630
    },
    {
      "epoch": 139.67,
      "learning_rate": 1.3778640776699031e-06,
      "loss": 0.0003,
      "step": 359640
    },
    {
      "epoch": 139.67,
      "learning_rate": 1.3773462783171523e-06,
      "loss": 0.0192,
      "step": 359650
    },
    {
      "epoch": 139.67,
      "learning_rate": 1.3768284789644015e-06,
      "loss": 0.0365,
      "step": 359660
    },
    {
      "epoch": 139.68,
      "learning_rate": 1.3763106796116507e-06,
      "loss": 0.0425,
      "step": 359670
    },
    {
      "epoch": 139.68,
      "learning_rate": 1.3757928802588999e-06,
      "loss": 0.012,
      "step": 359680
    },
    {
      "epoch": 139.69,
      "learning_rate": 1.375275080906149e-06,
      "loss": 0.0001,
      "step": 359690
    },
    {
      "epoch": 139.69,
      "learning_rate": 1.3747572815533983e-06,
      "loss": 0.0229,
      "step": 359700
    },
    {
      "epoch": 139.69,
      "learning_rate": 1.3742394822006475e-06,
      "loss": 0.1386,
      "step": 359710
    },
    {
      "epoch": 139.7,
      "learning_rate": 1.3737216828478967e-06,
      "loss": 0.0191,
      "step": 359720
    },
    {
      "epoch": 139.7,
      "learning_rate": 1.3732038834951459e-06,
      "loss": 0.0001,
      "step": 359730
    },
    {
      "epoch": 139.7,
      "learning_rate": 1.372686084142395e-06,
      "loss": 0.0321,
      "step": 359740
    },
    {
      "epoch": 139.71,
      "learning_rate": 1.3721682847896442e-06,
      "loss": 0.0077,
      "step": 359750
    },
    {
      "epoch": 139.71,
      "learning_rate": 1.3716504854368934e-06,
      "loss": 0.0003,
      "step": 359760
    },
    {
      "epoch": 139.72,
      "learning_rate": 1.3711326860841426e-06,
      "loss": 0.0112,
      "step": 359770
    },
    {
      "epoch": 139.72,
      "learning_rate": 1.3706148867313918e-06,
      "loss": 0.0148,
      "step": 359780
    },
    {
      "epoch": 139.72,
      "learning_rate": 1.370097087378641e-06,
      "loss": 0.0002,
      "step": 359790
    },
    {
      "epoch": 139.73,
      "learning_rate": 1.3695792880258902e-06,
      "loss": 0.0082,
      "step": 359800
    },
    {
      "epoch": 139.73,
      "learning_rate": 1.3690614886731394e-06,
      "loss": 0.0256,
      "step": 359810
    },
    {
      "epoch": 139.74,
      "learning_rate": 1.3685436893203884e-06,
      "loss": 0.0165,
      "step": 359820
    },
    {
      "epoch": 139.74,
      "learning_rate": 1.3680258899676376e-06,
      "loss": 0.0428,
      "step": 359830
    },
    {
      "epoch": 139.74,
      "learning_rate": 1.3675080906148868e-06,
      "loss": 0.0003,
      "step": 359840
    },
    {
      "epoch": 139.75,
      "learning_rate": 1.366990291262136e-06,
      "loss": 0.0916,
      "step": 359850
    },
    {
      "epoch": 139.75,
      "learning_rate": 1.3664724919093852e-06,
      "loss": 0.0833,
      "step": 359860
    },
    {
      "epoch": 139.76,
      "learning_rate": 1.3659546925566344e-06,
      "loss": 0.0182,
      "step": 359870
    },
    {
      "epoch": 139.76,
      "learning_rate": 1.3654368932038836e-06,
      "loss": 0.0599,
      "step": 359880
    },
    {
      "epoch": 139.76,
      "learning_rate": 1.3649190938511328e-06,
      "loss": 0.0634,
      "step": 359890
    },
    {
      "epoch": 139.77,
      "learning_rate": 1.364401294498382e-06,
      "loss": 0.0548,
      "step": 359900
    },
    {
      "epoch": 139.77,
      "learning_rate": 1.3638834951456311e-06,
      "loss": 0.0109,
      "step": 359910
    },
    {
      "epoch": 139.77,
      "learning_rate": 1.3633656957928803e-06,
      "loss": 0.0083,
      "step": 359920
    },
    {
      "epoch": 139.78,
      "learning_rate": 1.3628478964401295e-06,
      "loss": 0.002,
      "step": 359930
    },
    {
      "epoch": 139.78,
      "learning_rate": 1.3623300970873787e-06,
      "loss": 0.0787,
      "step": 359940
    },
    {
      "epoch": 139.79,
      "learning_rate": 1.361812297734628e-06,
      "loss": 0.0631,
      "step": 359950
    },
    {
      "epoch": 139.79,
      "learning_rate": 1.3612944983818771e-06,
      "loss": 0.0001,
      "step": 359960
    },
    {
      "epoch": 139.79,
      "learning_rate": 1.3607766990291263e-06,
      "loss": 0.0177,
      "step": 359970
    },
    {
      "epoch": 139.8,
      "learning_rate": 1.3602588996763755e-06,
      "loss": 0.0431,
      "step": 359980
    },
    {
      "epoch": 139.8,
      "learning_rate": 1.3597411003236247e-06,
      "loss": 0.0166,
      "step": 359990
    },
    {
      "epoch": 139.81,
      "learning_rate": 1.359223300970874e-06,
      "loss": 0.0409,
      "step": 360000
    },
    {
      "epoch": 139.81,
      "learning_rate": 1.358705501618123e-06,
      "loss": 0.0231,
      "step": 360010
    },
    {
      "epoch": 139.81,
      "learning_rate": 1.3581877022653723e-06,
      "loss": 0.0627,
      "step": 360020
    },
    {
      "epoch": 139.82,
      "learning_rate": 1.3576699029126215e-06,
      "loss": 0.0121,
      "step": 360030
    },
    {
      "epoch": 139.82,
      "learning_rate": 1.3571521035598707e-06,
      "loss": 0.0003,
      "step": 360040
    },
    {
      "epoch": 139.83,
      "learning_rate": 1.3566343042071199e-06,
      "loss": 0.0001,
      "step": 360050
    },
    {
      "epoch": 139.83,
      "learning_rate": 1.356116504854369e-06,
      "loss": 0.0663,
      "step": 360060
    },
    {
      "epoch": 139.83,
      "learning_rate": 1.355598705501618e-06,
      "loss": 0.0089,
      "step": 360070
    },
    {
      "epoch": 139.84,
      "learning_rate": 1.3550809061488672e-06,
      "loss": 0.0022,
      "step": 360080
    },
    {
      "epoch": 139.84,
      "learning_rate": 1.3545631067961164e-06,
      "loss": 0.0708,
      "step": 360090
    },
    {
      "epoch": 139.84,
      "learning_rate": 1.3540453074433656e-06,
      "loss": 0.0004,
      "step": 360100
    },
    {
      "epoch": 139.85,
      "learning_rate": 1.3535275080906148e-06,
      "loss": 0.0088,
      "step": 360110
    },
    {
      "epoch": 139.85,
      "learning_rate": 1.353009708737864e-06,
      "loss": 0.1098,
      "step": 360120
    },
    {
      "epoch": 139.86,
      "learning_rate": 1.3524919093851132e-06,
      "loss": 0.0698,
      "step": 360130
    },
    {
      "epoch": 139.86,
      "learning_rate": 1.3519741100323624e-06,
      "loss": 0.1153,
      "step": 360140
    },
    {
      "epoch": 139.86,
      "learning_rate": 1.3514563106796116e-06,
      "loss": 0.1187,
      "step": 360150
    },
    {
      "epoch": 139.87,
      "learning_rate": 1.3509385113268608e-06,
      "loss": 0.0379,
      "step": 360160
    },
    {
      "epoch": 139.87,
      "learning_rate": 1.35042071197411e-06,
      "loss": 0.001,
      "step": 360170
    },
    {
      "epoch": 139.88,
      "learning_rate": 1.3499029126213592e-06,
      "loss": 0.0693,
      "step": 360180
    },
    {
      "epoch": 139.88,
      "learning_rate": 1.3493851132686086e-06,
      "loss": 0.0854,
      "step": 360190
    },
    {
      "epoch": 139.88,
      "learning_rate": 1.3488673139158578e-06,
      "loss": 0.039,
      "step": 360200
    },
    {
      "epoch": 139.89,
      "learning_rate": 1.348349514563107e-06,
      "loss": 0.0022,
      "step": 360210
    },
    {
      "epoch": 139.89,
      "learning_rate": 1.3478317152103562e-06,
      "loss": 0.0061,
      "step": 360220
    },
    {
      "epoch": 139.9,
      "learning_rate": 1.3473139158576054e-06,
      "loss": 0.1372,
      "step": 360230
    },
    {
      "epoch": 139.9,
      "learning_rate": 1.3467961165048546e-06,
      "loss": 0.055,
      "step": 360240
    },
    {
      "epoch": 139.9,
      "learning_rate": 1.3462783171521038e-06,
      "loss": 0.111,
      "step": 360250
    },
    {
      "epoch": 139.91,
      "learning_rate": 1.345760517799353e-06,
      "loss": 0.0852,
      "step": 360260
    },
    {
      "epoch": 139.91,
      "learning_rate": 1.3452427184466022e-06,
      "loss": 0.0193,
      "step": 360270
    },
    {
      "epoch": 139.91,
      "learning_rate": 1.3447249190938513e-06,
      "loss": 0.1287,
      "step": 360280
    },
    {
      "epoch": 139.92,
      "learning_rate": 1.3442071197411005e-06,
      "loss": 0.0756,
      "step": 360290
    },
    {
      "epoch": 139.92,
      "learning_rate": 1.3436893203883497e-06,
      "loss": 0.0697,
      "step": 360300
    },
    {
      "epoch": 139.93,
      "learning_rate": 1.343171521035599e-06,
      "loss": 0.0006,
      "step": 360310
    },
    {
      "epoch": 139.93,
      "learning_rate": 1.3426537216828481e-06,
      "loss": 0.0001,
      "step": 360320
    },
    {
      "epoch": 139.93,
      "learning_rate": 1.3421359223300973e-06,
      "loss": 0.0953,
      "step": 360330
    },
    {
      "epoch": 139.94,
      "learning_rate": 1.3416181229773465e-06,
      "loss": 0.0192,
      "step": 360340
    },
    {
      "epoch": 139.94,
      "learning_rate": 1.3411003236245957e-06,
      "loss": 0.0185,
      "step": 360350
    },
    {
      "epoch": 139.95,
      "learning_rate": 1.340582524271845e-06,
      "loss": 0.0226,
      "step": 360360
    },
    {
      "epoch": 139.95,
      "learning_rate": 1.340064724919094e-06,
      "loss": 0.0007,
      "step": 360370
    },
    {
      "epoch": 139.95,
      "learning_rate": 1.3395469255663433e-06,
      "loss": 0.0541,
      "step": 360380
    },
    {
      "epoch": 139.96,
      "learning_rate": 1.3390291262135923e-06,
      "loss": 0.0101,
      "step": 360390
    },
    {
      "epoch": 139.96,
      "learning_rate": 1.3385113268608415e-06,
      "loss": 0.0765,
      "step": 360400
    },
    {
      "epoch": 139.97,
      "learning_rate": 1.3379935275080907e-06,
      "loss": 0.0017,
      "step": 360410
    },
    {
      "epoch": 139.97,
      "learning_rate": 1.3374757281553399e-06,
      "loss": 0.0091,
      "step": 360420
    },
    {
      "epoch": 139.97,
      "learning_rate": 1.336957928802589e-06,
      "loss": 0.0348,
      "step": 360430
    },
    {
      "epoch": 139.98,
      "learning_rate": 1.3364401294498382e-06,
      "loss": 0.0703,
      "step": 360440
    },
    {
      "epoch": 139.98,
      "learning_rate": 1.3359223300970874e-06,
      "loss": 0.1624,
      "step": 360450
    },
    {
      "epoch": 139.98,
      "learning_rate": 1.3354045307443366e-06,
      "loss": 0.0923,
      "step": 360460
    },
    {
      "epoch": 139.99,
      "learning_rate": 1.3348867313915858e-06,
      "loss": 0.0325,
      "step": 360470
    },
    {
      "epoch": 139.99,
      "learning_rate": 1.334368932038835e-06,
      "loss": 0.0917,
      "step": 360480
    },
    {
      "epoch": 140.0,
      "learning_rate": 1.3338511326860842e-06,
      "loss": 0.1865,
      "step": 360490
    },
    {
      "epoch": 140.0,
      "learning_rate": 1.3333333333333334e-06,
      "loss": 0.0442,
      "step": 360500
    },
    {
      "epoch": 140.0,
      "eval_accuracy": 0.9524071526822558,
      "eval_loss": 0.38710692524909973,
      "eval_runtime": 8.1858,
      "eval_samples_per_second": 444.063,
      "eval_steps_per_second": 55.584,
      "step": 360500
    },
    {
      "epoch": 140.0,
      "learning_rate": 1.3328155339805826e-06,
      "loss": 0.1342,
      "step": 360510
    },
    {
      "epoch": 140.01,
      "learning_rate": 1.3322977346278318e-06,
      "loss": 0.0503,
      "step": 360520
    },
    {
      "epoch": 140.01,
      "learning_rate": 1.331779935275081e-06,
      "loss": 0.0662,
      "step": 360530
    },
    {
      "epoch": 140.02,
      "learning_rate": 1.3312621359223302e-06,
      "loss": 0.0123,
      "step": 360540
    },
    {
      "epoch": 140.02,
      "learning_rate": 1.3307443365695794e-06,
      "loss": 0.0212,
      "step": 360550
    },
    {
      "epoch": 140.02,
      "learning_rate": 1.3302265372168286e-06,
      "loss": 0.0003,
      "step": 360560
    },
    {
      "epoch": 140.03,
      "learning_rate": 1.3297087378640778e-06,
      "loss": 0.1655,
      "step": 360570
    },
    {
      "epoch": 140.03,
      "learning_rate": 1.329190938511327e-06,
      "loss": 0.0595,
      "step": 360580
    },
    {
      "epoch": 140.03,
      "learning_rate": 1.3286731391585762e-06,
      "loss": 0.0414,
      "step": 360590
    },
    {
      "epoch": 140.04,
      "learning_rate": 1.3281553398058254e-06,
      "loss": 0.0883,
      "step": 360600
    },
    {
      "epoch": 140.04,
      "learning_rate": 1.3276375404530746e-06,
      "loss": 0.0147,
      "step": 360610
    },
    {
      "epoch": 140.05,
      "learning_rate": 1.3271197411003237e-06,
      "loss": 0.0324,
      "step": 360620
    },
    {
      "epoch": 140.05,
      "learning_rate": 1.326601941747573e-06,
      "loss": 0.0518,
      "step": 360630
    },
    {
      "epoch": 140.05,
      "learning_rate": 1.326084142394822e-06,
      "loss": 0.0126,
      "step": 360640
    },
    {
      "epoch": 140.06,
      "learning_rate": 1.3255663430420711e-06,
      "loss": 0.0254,
      "step": 360650
    },
    {
      "epoch": 140.06,
      "learning_rate": 1.3250485436893203e-06,
      "loss": 0.0087,
      "step": 360660
    },
    {
      "epoch": 140.07,
      "learning_rate": 1.3245307443365695e-06,
      "loss": 0.0253,
      "step": 360670
    },
    {
      "epoch": 140.07,
      "learning_rate": 1.3240129449838187e-06,
      "loss": 0.0141,
      "step": 360680
    },
    {
      "epoch": 140.07,
      "learning_rate": 1.323495145631068e-06,
      "loss": 0.0987,
      "step": 360690
    },
    {
      "epoch": 140.08,
      "learning_rate": 1.322977346278317e-06,
      "loss": 0.0002,
      "step": 360700
    },
    {
      "epoch": 140.08,
      "learning_rate": 1.3224595469255663e-06,
      "loss": 0.0263,
      "step": 360710
    },
    {
      "epoch": 140.09,
      "learning_rate": 1.3219417475728155e-06,
      "loss": 0.0308,
      "step": 360720
    },
    {
      "epoch": 140.09,
      "learning_rate": 1.3214239482200647e-06,
      "loss": 0.0338,
      "step": 360730
    },
    {
      "epoch": 140.09,
      "learning_rate": 1.3209061488673139e-06,
      "loss": 0.0905,
      "step": 360740
    },
    {
      "epoch": 140.1,
      "learning_rate": 1.3203883495145633e-06,
      "loss": 0.0241,
      "step": 360750
    },
    {
      "epoch": 140.1,
      "learning_rate": 1.3198705501618125e-06,
      "loss": 0.1399,
      "step": 360760
    },
    {
      "epoch": 140.1,
      "learning_rate": 1.3193527508090617e-06,
      "loss": 0.1117,
      "step": 360770
    },
    {
      "epoch": 140.11,
      "learning_rate": 1.3188349514563109e-06,
      "loss": 0.0607,
      "step": 360780
    },
    {
      "epoch": 140.11,
      "learning_rate": 1.31831715210356e-06,
      "loss": 0.0482,
      "step": 360790
    },
    {
      "epoch": 140.12,
      "learning_rate": 1.3177993527508092e-06,
      "loss": 0.0204,
      "step": 360800
    },
    {
      "epoch": 140.12,
      "learning_rate": 1.3172815533980584e-06,
      "loss": 0.0391,
      "step": 360810
    },
    {
      "epoch": 140.12,
      "learning_rate": 1.3167637540453076e-06,
      "loss": 0.0602,
      "step": 360820
    },
    {
      "epoch": 140.13,
      "learning_rate": 1.3162459546925568e-06,
      "loss": 0.0252,
      "step": 360830
    },
    {
      "epoch": 140.13,
      "learning_rate": 1.315728155339806e-06,
      "loss": 0.0442,
      "step": 360840
    },
    {
      "epoch": 140.14,
      "learning_rate": 1.3152103559870552e-06,
      "loss": 0.1645,
      "step": 360850
    },
    {
      "epoch": 140.14,
      "learning_rate": 1.3146925566343044e-06,
      "loss": 0.0169,
      "step": 360860
    },
    {
      "epoch": 140.14,
      "learning_rate": 1.3141747572815536e-06,
      "loss": 0.0003,
      "step": 360870
    },
    {
      "epoch": 140.15,
      "learning_rate": 1.3136569579288028e-06,
      "loss": 0.0645,
      "step": 360880
    },
    {
      "epoch": 140.15,
      "learning_rate": 1.313139158576052e-06,
      "loss": 0.0081,
      "step": 360890
    },
    {
      "epoch": 140.16,
      "learning_rate": 1.3126213592233012e-06,
      "loss": 0.0181,
      "step": 360900
    },
    {
      "epoch": 140.16,
      "learning_rate": 1.3121035598705504e-06,
      "loss": 0.0166,
      "step": 360910
    },
    {
      "epoch": 140.16,
      "learning_rate": 1.3115857605177996e-06,
      "loss": 0.0182,
      "step": 360920
    },
    {
      "epoch": 140.17,
      "learning_rate": 1.3110679611650488e-06,
      "loss": 0.0036,
      "step": 360930
    },
    {
      "epoch": 140.17,
      "learning_rate": 1.310550161812298e-06,
      "loss": 0.0003,
      "step": 360940
    },
    {
      "epoch": 140.17,
      "learning_rate": 1.3100323624595472e-06,
      "loss": 0.0003,
      "step": 360950
    },
    {
      "epoch": 140.18,
      "learning_rate": 1.3095145631067962e-06,
      "loss": 0.0006,
      "step": 360960
    },
    {
      "epoch": 140.18,
      "learning_rate": 1.3089967637540453e-06,
      "loss": 0.0519,
      "step": 360970
    },
    {
      "epoch": 140.19,
      "learning_rate": 1.3084789644012945e-06,
      "loss": 0.0001,
      "step": 360980
    },
    {
      "epoch": 140.19,
      "learning_rate": 1.3079611650485437e-06,
      "loss": 0.0004,
      "step": 360990
    },
    {
      "epoch": 140.19,
      "learning_rate": 1.307443365695793e-06,
      "loss": 0.0916,
      "step": 361000
    },
    {
      "epoch": 140.2,
      "learning_rate": 1.3069255663430421e-06,
      "loss": 0.0225,
      "step": 361010
    },
    {
      "epoch": 140.2,
      "learning_rate": 1.3064077669902913e-06,
      "loss": 0.0138,
      "step": 361020
    },
    {
      "epoch": 140.21,
      "learning_rate": 1.3058899676375405e-06,
      "loss": 0.0088,
      "step": 361030
    },
    {
      "epoch": 140.21,
      "learning_rate": 1.3053721682847897e-06,
      "loss": 0.0001,
      "step": 361040
    },
    {
      "epoch": 140.21,
      "learning_rate": 1.304854368932039e-06,
      "loss": 0.0258,
      "step": 361050
    },
    {
      "epoch": 140.22,
      "learning_rate": 1.304336569579288e-06,
      "loss": 0.0288,
      "step": 361060
    },
    {
      "epoch": 140.22,
      "learning_rate": 1.3038187702265373e-06,
      "loss": 0.0449,
      "step": 361070
    },
    {
      "epoch": 140.23,
      "learning_rate": 1.3033009708737865e-06,
      "loss": 0.0143,
      "step": 361080
    },
    {
      "epoch": 140.23,
      "learning_rate": 1.3027831715210357e-06,
      "loss": 0.0328,
      "step": 361090
    },
    {
      "epoch": 140.23,
      "learning_rate": 1.3022653721682849e-06,
      "loss": 0.0102,
      "step": 361100
    },
    {
      "epoch": 140.24,
      "learning_rate": 1.301747572815534e-06,
      "loss": 0.048,
      "step": 361110
    },
    {
      "epoch": 140.24,
      "learning_rate": 1.3012297734627833e-06,
      "loss": 0.0115,
      "step": 361120
    },
    {
      "epoch": 140.24,
      "learning_rate": 1.3007119741100325e-06,
      "loss": 0.0696,
      "step": 361130
    },
    {
      "epoch": 140.25,
      "learning_rate": 1.3001941747572817e-06,
      "loss": 0.0356,
      "step": 361140
    },
    {
      "epoch": 140.25,
      "learning_rate": 1.2996763754045308e-06,
      "loss": 0.0005,
      "step": 361150
    },
    {
      "epoch": 140.26,
      "learning_rate": 1.29915857605178e-06,
      "loss": 0.0006,
      "step": 361160
    },
    {
      "epoch": 140.26,
      "learning_rate": 1.2986407766990292e-06,
      "loss": 0.0395,
      "step": 361170
    },
    {
      "epoch": 140.26,
      "learning_rate": 1.2981229773462784e-06,
      "loss": 0.077,
      "step": 361180
    },
    {
      "epoch": 140.27,
      "learning_rate": 1.2976051779935276e-06,
      "loss": 0.012,
      "step": 361190
    },
    {
      "epoch": 140.27,
      "learning_rate": 1.2970873786407768e-06,
      "loss": 0.0472,
      "step": 361200
    },
    {
      "epoch": 140.28,
      "learning_rate": 1.2965695792880258e-06,
      "loss": 0.0243,
      "step": 361210
    },
    {
      "epoch": 140.28,
      "learning_rate": 1.296051779935275e-06,
      "loss": 0.1352,
      "step": 361220
    },
    {
      "epoch": 140.28,
      "learning_rate": 1.2955339805825242e-06,
      "loss": 0.0268,
      "step": 361230
    },
    {
      "epoch": 140.29,
      "learning_rate": 1.2950161812297734e-06,
      "loss": 0.0427,
      "step": 361240
    },
    {
      "epoch": 140.29,
      "learning_rate": 1.2944983818770226e-06,
      "loss": 0.0362,
      "step": 361250
    },
    {
      "epoch": 140.3,
      "learning_rate": 1.2939805825242718e-06,
      "loss": 0.0453,
      "step": 361260
    },
    {
      "epoch": 140.3,
      "learning_rate": 1.293462783171521e-06,
      "loss": 0.0109,
      "step": 361270
    },
    {
      "epoch": 140.3,
      "learning_rate": 1.2929449838187702e-06,
      "loss": 0.0838,
      "step": 361280
    },
    {
      "epoch": 140.31,
      "learning_rate": 1.2924271844660194e-06,
      "loss": 0.0757,
      "step": 361290
    },
    {
      "epoch": 140.31,
      "learning_rate": 1.2919093851132686e-06,
      "loss": 0.0401,
      "step": 361300
    },
    {
      "epoch": 140.31,
      "learning_rate": 1.291391585760518e-06,
      "loss": 0.0128,
      "step": 361310
    },
    {
      "epoch": 140.32,
      "learning_rate": 1.2908737864077672e-06,
      "loss": 0.0752,
      "step": 361320
    },
    {
      "epoch": 140.32,
      "learning_rate": 1.2903559870550163e-06,
      "loss": 0.0154,
      "step": 361330
    },
    {
      "epoch": 140.33,
      "learning_rate": 1.2898381877022655e-06,
      "loss": 0.0636,
      "step": 361340
    },
    {
      "epoch": 140.33,
      "learning_rate": 1.2893203883495147e-06,
      "loss": 0.0778,
      "step": 361350
    },
    {
      "epoch": 140.33,
      "learning_rate": 1.288802588996764e-06,
      "loss": 0.0851,
      "step": 361360
    },
    {
      "epoch": 140.34,
      "learning_rate": 1.2882847896440131e-06,
      "loss": 0.0217,
      "step": 361370
    },
    {
      "epoch": 140.34,
      "learning_rate": 1.2877669902912623e-06,
      "loss": 0.008,
      "step": 361380
    },
    {
      "epoch": 140.35,
      "learning_rate": 1.2872491909385115e-06,
      "loss": 0.0349,
      "step": 361390
    },
    {
      "epoch": 140.35,
      "learning_rate": 1.2867313915857607e-06,
      "loss": 0.0075,
      "step": 361400
    },
    {
      "epoch": 140.35,
      "learning_rate": 1.28621359223301e-06,
      "loss": 0.1626,
      "step": 361410
    },
    {
      "epoch": 140.36,
      "learning_rate": 1.285695792880259e-06,
      "loss": 0.0087,
      "step": 361420
    },
    {
      "epoch": 140.36,
      "learning_rate": 1.2851779935275083e-06,
      "loss": 0.0475,
      "step": 361430
    },
    {
      "epoch": 140.37,
      "learning_rate": 1.2846601941747575e-06,
      "loss": 0.046,
      "step": 361440
    },
    {
      "epoch": 140.37,
      "learning_rate": 1.2841423948220067e-06,
      "loss": 0.031,
      "step": 361450
    },
    {
      "epoch": 140.37,
      "learning_rate": 1.2836245954692559e-06,
      "loss": 0.0891,
      "step": 361460
    },
    {
      "epoch": 140.38,
      "learning_rate": 1.283106796116505e-06,
      "loss": 0.0132,
      "step": 361470
    },
    {
      "epoch": 140.38,
      "learning_rate": 1.2825889967637543e-06,
      "loss": 0.0424,
      "step": 361480
    },
    {
      "epoch": 140.38,
      "learning_rate": 1.2820711974110035e-06,
      "loss": 0.0005,
      "step": 361490
    },
    {
      "epoch": 140.39,
      "learning_rate": 1.2815533980582527e-06,
      "loss": 0.001,
      "step": 361500
    },
    {
      "epoch": 140.39,
      "learning_rate": 1.2810355987055018e-06,
      "loss": 0.0336,
      "step": 361510
    },
    {
      "epoch": 140.4,
      "learning_rate": 1.280517799352751e-06,
      "loss": 0.0625,
      "step": 361520
    },
    {
      "epoch": 140.4,
      "learning_rate": 1.28e-06,
      "loss": 0.0735,
      "step": 361530
    },
    {
      "epoch": 140.4,
      "learning_rate": 1.2794822006472492e-06,
      "loss": 0.067,
      "step": 361540
    },
    {
      "epoch": 140.41,
      "learning_rate": 1.2789644012944984e-06,
      "loss": 0.0094,
      "step": 361550
    },
    {
      "epoch": 140.41,
      "learning_rate": 1.2784466019417476e-06,
      "loss": 0.0289,
      "step": 361560
    },
    {
      "epoch": 140.42,
      "learning_rate": 1.2779288025889968e-06,
      "loss": 0.1281,
      "step": 361570
    },
    {
      "epoch": 140.42,
      "learning_rate": 1.277411003236246e-06,
      "loss": 0.0764,
      "step": 361580
    },
    {
      "epoch": 140.42,
      "learning_rate": 1.2768932038834952e-06,
      "loss": 0.0815,
      "step": 361590
    },
    {
      "epoch": 140.43,
      "learning_rate": 1.2763754045307444e-06,
      "loss": 0.1652,
      "step": 361600
    },
    {
      "epoch": 140.43,
      "learning_rate": 1.2758576051779936e-06,
      "loss": 0.1888,
      "step": 361610
    },
    {
      "epoch": 140.43,
      "learning_rate": 1.2753398058252428e-06,
      "loss": 0.0502,
      "step": 361620
    },
    {
      "epoch": 140.44,
      "learning_rate": 1.274822006472492e-06,
      "loss": 0.0751,
      "step": 361630
    },
    {
      "epoch": 140.44,
      "learning_rate": 1.2743042071197412e-06,
      "loss": 0.1014,
      "step": 361640
    },
    {
      "epoch": 140.45,
      "learning_rate": 1.2737864077669904e-06,
      "loss": 0.1135,
      "step": 361650
    },
    {
      "epoch": 140.45,
      "learning_rate": 1.2732686084142396e-06,
      "loss": 0.0263,
      "step": 361660
    },
    {
      "epoch": 140.45,
      "learning_rate": 1.2727508090614888e-06,
      "loss": 0.0003,
      "step": 361670
    },
    {
      "epoch": 140.46,
      "learning_rate": 1.272233009708738e-06,
      "loss": 0.001,
      "step": 361680
    },
    {
      "epoch": 140.46,
      "learning_rate": 1.2717152103559871e-06,
      "loss": 0.1085,
      "step": 361690
    },
    {
      "epoch": 140.47,
      "learning_rate": 1.2711974110032363e-06,
      "loss": 0.0733,
      "step": 361700
    },
    {
      "epoch": 140.47,
      "learning_rate": 1.2706796116504855e-06,
      "loss": 0.0006,
      "step": 361710
    },
    {
      "epoch": 140.47,
      "learning_rate": 1.2701618122977347e-06,
      "loss": 0.0554,
      "step": 361720
    },
    {
      "epoch": 140.48,
      "learning_rate": 1.269644012944984e-06,
      "loss": 0.043,
      "step": 361730
    },
    {
      "epoch": 140.48,
      "learning_rate": 1.2691262135922331e-06,
      "loss": 0.132,
      "step": 361740
    },
    {
      "epoch": 140.49,
      "learning_rate": 1.2686084142394823e-06,
      "loss": 0.1055,
      "step": 361750
    },
    {
      "epoch": 140.49,
      "learning_rate": 1.2680906148867315e-06,
      "loss": 0.0001,
      "step": 361760
    },
    {
      "epoch": 140.49,
      "learning_rate": 1.2675728155339807e-06,
      "loss": 0.0899,
      "step": 361770
    },
    {
      "epoch": 140.5,
      "learning_rate": 1.2670550161812297e-06,
      "loss": 0.1261,
      "step": 361780
    },
    {
      "epoch": 140.5,
      "learning_rate": 1.2665372168284789e-06,
      "loss": 0.0469,
      "step": 361790
    },
    {
      "epoch": 140.5,
      "learning_rate": 1.266019417475728e-06,
      "loss": 0.0078,
      "step": 361800
    },
    {
      "epoch": 140.51,
      "learning_rate": 1.2655016181229773e-06,
      "loss": 0.0934,
      "step": 361810
    },
    {
      "epoch": 140.51,
      "learning_rate": 1.2649838187702265e-06,
      "loss": 0.0014,
      "step": 361820
    },
    {
      "epoch": 140.52,
      "learning_rate": 1.2644660194174757e-06,
      "loss": 0.1355,
      "step": 361830
    },
    {
      "epoch": 140.52,
      "learning_rate": 1.2639482200647248e-06,
      "loss": 0.002,
      "step": 361840
    },
    {
      "epoch": 140.52,
      "learning_rate": 1.263430420711974e-06,
      "loss": 0.0513,
      "step": 361850
    },
    {
      "epoch": 140.53,
      "learning_rate": 1.2629126213592232e-06,
      "loss": 0.046,
      "step": 361860
    },
    {
      "epoch": 140.53,
      "learning_rate": 1.2623948220064726e-06,
      "loss": 0.0933,
      "step": 361870
    },
    {
      "epoch": 140.54,
      "learning_rate": 1.2618770226537218e-06,
      "loss": 0.1006,
      "step": 361880
    },
    {
      "epoch": 140.54,
      "learning_rate": 1.261359223300971e-06,
      "loss": 0.1502,
      "step": 361890
    },
    {
      "epoch": 140.54,
      "learning_rate": 1.2608414239482202e-06,
      "loss": 0.101,
      "step": 361900
    },
    {
      "epoch": 140.55,
      "learning_rate": 1.2603236245954694e-06,
      "loss": 0.0115,
      "step": 361910
    },
    {
      "epoch": 140.55,
      "learning_rate": 1.2598058252427186e-06,
      "loss": 0.0106,
      "step": 361920
    },
    {
      "epoch": 140.56,
      "learning_rate": 1.2592880258899678e-06,
      "loss": 0.0177,
      "step": 361930
    },
    {
      "epoch": 140.56,
      "learning_rate": 1.258770226537217e-06,
      "loss": 0.0001,
      "step": 361940
    },
    {
      "epoch": 140.56,
      "learning_rate": 1.2582524271844662e-06,
      "loss": 0.0003,
      "step": 361950
    },
    {
      "epoch": 140.57,
      "learning_rate": 1.2577346278317154e-06,
      "loss": 0.0954,
      "step": 361960
    },
    {
      "epoch": 140.57,
      "learning_rate": 1.2572168284789646e-06,
      "loss": 0.1693,
      "step": 361970
    },
    {
      "epoch": 140.57,
      "learning_rate": 1.2566990291262138e-06,
      "loss": 0.0409,
      "step": 361980
    },
    {
      "epoch": 140.58,
      "learning_rate": 1.256181229773463e-06,
      "loss": 0.0824,
      "step": 361990
    },
    {
      "epoch": 140.58,
      "learning_rate": 1.2556634304207122e-06,
      "loss": 0.0124,
      "step": 362000
    },
    {
      "epoch": 140.59,
      "learning_rate": 1.2551456310679614e-06,
      "loss": 0.0565,
      "step": 362010
    },
    {
      "epoch": 140.59,
      "learning_rate": 1.2546278317152106e-06,
      "loss": 0.0341,
      "step": 362020
    },
    {
      "epoch": 140.59,
      "learning_rate": 1.2541100323624598e-06,
      "loss": 0.0803,
      "step": 362030
    },
    {
      "epoch": 140.6,
      "learning_rate": 1.253592233009709e-06,
      "loss": 0.0274,
      "step": 362040
    },
    {
      "epoch": 140.6,
      "learning_rate": 1.2530744336569581e-06,
      "loss": 0.0041,
      "step": 362050
    },
    {
      "epoch": 140.61,
      "learning_rate": 1.2525566343042073e-06,
      "loss": 0.0082,
      "step": 362060
    },
    {
      "epoch": 140.61,
      "learning_rate": 1.2520388349514565e-06,
      "loss": 0.0203,
      "step": 362070
    },
    {
      "epoch": 140.61,
      "learning_rate": 1.2515210355987057e-06,
      "loss": 0.0294,
      "step": 362080
    },
    {
      "epoch": 140.62,
      "learning_rate": 1.251003236245955e-06,
      "loss": 0.0482,
      "step": 362090
    },
    {
      "epoch": 140.62,
      "learning_rate": 1.250485436893204e-06,
      "loss": 0.0001,
      "step": 362100
    },
    {
      "epoch": 140.63,
      "learning_rate": 1.249967637540453e-06,
      "loss": 0.0849,
      "step": 362110
    },
    {
      "epoch": 140.63,
      "learning_rate": 1.2494498381877023e-06,
      "loss": 0.0772,
      "step": 362120
    },
    {
      "epoch": 140.63,
      "learning_rate": 1.2489320388349515e-06,
      "loss": 0.0014,
      "step": 362130
    },
    {
      "epoch": 140.64,
      "learning_rate": 1.2484142394822007e-06,
      "loss": 0.0753,
      "step": 362140
    },
    {
      "epoch": 140.64,
      "learning_rate": 1.2478964401294499e-06,
      "loss": 0.0147,
      "step": 362150
    },
    {
      "epoch": 140.64,
      "learning_rate": 1.247378640776699e-06,
      "loss": 0.0631,
      "step": 362160
    },
    {
      "epoch": 140.65,
      "learning_rate": 1.2468608414239483e-06,
      "loss": 0.0179,
      "step": 362170
    },
    {
      "epoch": 140.65,
      "learning_rate": 1.2463430420711975e-06,
      "loss": 0.0126,
      "step": 362180
    },
    {
      "epoch": 140.66,
      "learning_rate": 1.2458252427184467e-06,
      "loss": 0.0613,
      "step": 362190
    },
    {
      "epoch": 140.66,
      "learning_rate": 1.2453074433656958e-06,
      "loss": 0.0178,
      "step": 362200
    },
    {
      "epoch": 140.66,
      "learning_rate": 1.244789644012945e-06,
      "loss": 0.0047,
      "step": 362210
    },
    {
      "epoch": 140.67,
      "learning_rate": 1.2442718446601942e-06,
      "loss": 0.0685,
      "step": 362220
    },
    {
      "epoch": 140.67,
      "learning_rate": 1.2437540453074434e-06,
      "loss": 0.0687,
      "step": 362230
    },
    {
      "epoch": 140.68,
      "learning_rate": 1.2432362459546926e-06,
      "loss": 0.0387,
      "step": 362240
    },
    {
      "epoch": 140.68,
      "learning_rate": 1.2427184466019418e-06,
      "loss": 0.0563,
      "step": 362250
    },
    {
      "epoch": 140.68,
      "learning_rate": 1.242200647249191e-06,
      "loss": 0.0179,
      "step": 362260
    },
    {
      "epoch": 140.69,
      "learning_rate": 1.2416828478964402e-06,
      "loss": 0.0318,
      "step": 362270
    },
    {
      "epoch": 140.69,
      "learning_rate": 1.2411650485436894e-06,
      "loss": 0.0002,
      "step": 362280
    },
    {
      "epoch": 140.7,
      "learning_rate": 1.2406472491909386e-06,
      "loss": 0.051,
      "step": 362290
    },
    {
      "epoch": 140.7,
      "learning_rate": 1.2401294498381878e-06,
      "loss": 0.0557,
      "step": 362300
    },
    {
      "epoch": 140.7,
      "learning_rate": 1.239611650485437e-06,
      "loss": 0.0348,
      "step": 362310
    },
    {
      "epoch": 140.71,
      "learning_rate": 1.2390938511326862e-06,
      "loss": 0.0368,
      "step": 362320
    },
    {
      "epoch": 140.71,
      "learning_rate": 1.2385760517799354e-06,
      "loss": 0.0169,
      "step": 362330
    },
    {
      "epoch": 140.71,
      "learning_rate": 1.2380582524271846e-06,
      "loss": 0.0926,
      "step": 362340
    },
    {
      "epoch": 140.72,
      "learning_rate": 1.2375404530744338e-06,
      "loss": 0.0001,
      "step": 362350
    },
    {
      "epoch": 140.72,
      "learning_rate": 1.237022653721683e-06,
      "loss": 0.0368,
      "step": 362360
    },
    {
      "epoch": 140.73,
      "learning_rate": 1.2365048543689322e-06,
      "loss": 0.0002,
      "step": 362370
    },
    {
      "epoch": 140.73,
      "learning_rate": 1.2359870550161813e-06,
      "loss": 0.0001,
      "step": 362380
    },
    {
      "epoch": 140.73,
      "learning_rate": 1.2354692556634305e-06,
      "loss": 0.0152,
      "step": 362390
    },
    {
      "epoch": 140.74,
      "learning_rate": 1.2349514563106797e-06,
      "loss": 0.0428,
      "step": 362400
    },
    {
      "epoch": 140.74,
      "learning_rate": 1.234433656957929e-06,
      "loss": 0.0027,
      "step": 362410
    },
    {
      "epoch": 140.75,
      "learning_rate": 1.2339158576051781e-06,
      "loss": 0.052,
      "step": 362420
    },
    {
      "epoch": 140.75,
      "learning_rate": 1.2333980582524273e-06,
      "loss": 0.1076,
      "step": 362430
    },
    {
      "epoch": 140.75,
      "learning_rate": 1.2328802588996765e-06,
      "loss": 0.0001,
      "step": 362440
    },
    {
      "epoch": 140.76,
      "learning_rate": 1.2323624595469257e-06,
      "loss": 0.0178,
      "step": 362450
    },
    {
      "epoch": 140.76,
      "learning_rate": 1.231844660194175e-06,
      "loss": 0.063,
      "step": 362460
    },
    {
      "epoch": 140.77,
      "learning_rate": 1.231326860841424e-06,
      "loss": 0.0003,
      "step": 362470
    },
    {
      "epoch": 140.77,
      "learning_rate": 1.2308090614886733e-06,
      "loss": 0.0088,
      "step": 362480
    },
    {
      "epoch": 140.77,
      "learning_rate": 1.2302912621359225e-06,
      "loss": 0.0039,
      "step": 362490
    },
    {
      "epoch": 140.78,
      "learning_rate": 1.2297734627831717e-06,
      "loss": 0.009,
      "step": 362500
    },
    {
      "epoch": 140.78,
      "learning_rate": 1.2292556634304207e-06,
      "loss": 0.0026,
      "step": 362510
    },
    {
      "epoch": 140.78,
      "learning_rate": 1.2287378640776699e-06,
      "loss": 0.1108,
      "step": 362520
    },
    {
      "epoch": 140.79,
      "learning_rate": 1.228220064724919e-06,
      "loss": 0.0237,
      "step": 362530
    },
    {
      "epoch": 140.79,
      "learning_rate": 1.2277022653721683e-06,
      "loss": 0.0356,
      "step": 362540
    },
    {
      "epoch": 140.8,
      "learning_rate": 1.2271844660194174e-06,
      "loss": 0.0031,
      "step": 362550
    },
    {
      "epoch": 140.8,
      "learning_rate": 1.2266666666666666e-06,
      "loss": 0.0709,
      "step": 362560
    },
    {
      "epoch": 140.8,
      "learning_rate": 1.226148867313916e-06,
      "loss": 0.0087,
      "step": 362570
    },
    {
      "epoch": 140.81,
      "learning_rate": 1.2256310679611652e-06,
      "loss": 0.068,
      "step": 362580
    },
    {
      "epoch": 140.81,
      "learning_rate": 1.2251132686084144e-06,
      "loss": 0.0816,
      "step": 362590
    },
    {
      "epoch": 140.82,
      "learning_rate": 1.2245954692556636e-06,
      "loss": 0.0003,
      "step": 362600
    },
    {
      "epoch": 140.82,
      "learning_rate": 1.2240776699029128e-06,
      "loss": 0.0602,
      "step": 362610
    },
    {
      "epoch": 140.82,
      "learning_rate": 1.223559870550162e-06,
      "loss": 0.014,
      "step": 362620
    },
    {
      "epoch": 140.83,
      "learning_rate": 1.2230420711974112e-06,
      "loss": 0.0016,
      "step": 362630
    },
    {
      "epoch": 140.83,
      "learning_rate": 1.2225242718446604e-06,
      "loss": 0.0191,
      "step": 362640
    },
    {
      "epoch": 140.83,
      "learning_rate": 1.2220064724919096e-06,
      "loss": 0.009,
      "step": 362650
    },
    {
      "epoch": 140.84,
      "learning_rate": 1.2214886731391588e-06,
      "loss": 0.0377,
      "step": 362660
    },
    {
      "epoch": 140.84,
      "learning_rate": 1.2209708737864078e-06,
      "loss": 0.0092,
      "step": 362670
    },
    {
      "epoch": 140.85,
      "learning_rate": 1.220453074433657e-06,
      "loss": 0.0206,
      "step": 362680
    },
    {
      "epoch": 140.85,
      "learning_rate": 1.2199352750809062e-06,
      "loss": 0.0002,
      "step": 362690
    },
    {
      "epoch": 140.85,
      "learning_rate": 1.2194174757281554e-06,
      "loss": 0.0678,
      "step": 362700
    },
    {
      "epoch": 140.86,
      "learning_rate": 1.2188996763754046e-06,
      "loss": 0.0638,
      "step": 362710
    },
    {
      "epoch": 140.86,
      "learning_rate": 1.2183818770226538e-06,
      "loss": 0.0763,
      "step": 362720
    },
    {
      "epoch": 140.87,
      "learning_rate": 1.217864077669903e-06,
      "loss": 0.0325,
      "step": 362730
    },
    {
      "epoch": 140.87,
      "learning_rate": 1.2173462783171521e-06,
      "loss": 0.0005,
      "step": 362740
    },
    {
      "epoch": 140.87,
      "learning_rate": 1.2168284789644013e-06,
      "loss": 0.0252,
      "step": 362750
    },
    {
      "epoch": 140.88,
      "learning_rate": 1.2163106796116505e-06,
      "loss": 0.0086,
      "step": 362760
    },
    {
      "epoch": 140.88,
      "learning_rate": 1.2157928802588997e-06,
      "loss": 0.0493,
      "step": 362770
    },
    {
      "epoch": 140.89,
      "learning_rate": 1.215275080906149e-06,
      "loss": 0.1083,
      "step": 362780
    },
    {
      "epoch": 140.89,
      "learning_rate": 1.2147572815533981e-06,
      "loss": 0.0352,
      "step": 362790
    },
    {
      "epoch": 140.89,
      "learning_rate": 1.2142394822006473e-06,
      "loss": 0.0471,
      "step": 362800
    },
    {
      "epoch": 140.9,
      "learning_rate": 1.2137216828478965e-06,
      "loss": 0.0637,
      "step": 362810
    },
    {
      "epoch": 140.9,
      "learning_rate": 1.2132038834951457e-06,
      "loss": 0.1292,
      "step": 362820
    },
    {
      "epoch": 140.9,
      "learning_rate": 1.2126860841423949e-06,
      "loss": 0.0153,
      "step": 362830
    },
    {
      "epoch": 140.91,
      "learning_rate": 1.212168284789644e-06,
      "loss": 0.0598,
      "step": 362840
    },
    {
      "epoch": 140.91,
      "learning_rate": 1.2116504854368933e-06,
      "loss": 0.0112,
      "step": 362850
    },
    {
      "epoch": 140.92,
      "learning_rate": 1.2111326860841425e-06,
      "loss": 0.0166,
      "step": 362860
    },
    {
      "epoch": 140.92,
      "learning_rate": 1.2106148867313917e-06,
      "loss": 0.0662,
      "step": 362870
    },
    {
      "epoch": 140.92,
      "learning_rate": 1.2100970873786409e-06,
      "loss": 0.0127,
      "step": 362880
    },
    {
      "epoch": 140.93,
      "learning_rate": 1.20957928802589e-06,
      "loss": 0.1013,
      "step": 362890
    },
    {
      "epoch": 140.93,
      "learning_rate": 1.2090614886731393e-06,
      "loss": 0.0419,
      "step": 362900
    },
    {
      "epoch": 140.94,
      "learning_rate": 1.2085436893203884e-06,
      "loss": 0.0099,
      "step": 362910
    },
    {
      "epoch": 140.94,
      "learning_rate": 1.2080258899676376e-06,
      "loss": 0.0005,
      "step": 362920
    },
    {
      "epoch": 140.94,
      "learning_rate": 1.2075080906148868e-06,
      "loss": 0.0915,
      "step": 362930
    },
    {
      "epoch": 140.95,
      "learning_rate": 1.206990291262136e-06,
      "loss": 0.085,
      "step": 362940
    },
    {
      "epoch": 140.95,
      "learning_rate": 1.2064724919093852e-06,
      "loss": 0.0449,
      "step": 362950
    },
    {
      "epoch": 140.96,
      "learning_rate": 1.2059546925566344e-06,
      "loss": 0.0273,
      "step": 362960
    },
    {
      "epoch": 140.96,
      "learning_rate": 1.2054368932038836e-06,
      "loss": 0.0001,
      "step": 362970
    },
    {
      "epoch": 140.96,
      "learning_rate": 1.2049190938511328e-06,
      "loss": 0.0194,
      "step": 362980
    },
    {
      "epoch": 140.97,
      "learning_rate": 1.204401294498382e-06,
      "loss": 0.0088,
      "step": 362990
    },
    {
      "epoch": 140.97,
      "learning_rate": 1.2038834951456312e-06,
      "loss": 0.047,
      "step": 363000
    },
    {
      "epoch": 140.97,
      "learning_rate": 1.2033656957928804e-06,
      "loss": 0.0019,
      "step": 363010
    },
    {
      "epoch": 140.98,
      "learning_rate": 1.2028478964401296e-06,
      "loss": 0.0154,
      "step": 363020
    },
    {
      "epoch": 140.98,
      "learning_rate": 1.2023300970873788e-06,
      "loss": 0.0556,
      "step": 363030
    },
    {
      "epoch": 140.99,
      "learning_rate": 1.201812297734628e-06,
      "loss": 0.0016,
      "step": 363040
    },
    {
      "epoch": 140.99,
      "learning_rate": 1.2012944983818772e-06,
      "loss": 0.0008,
      "step": 363050
    },
    {
      "epoch": 140.99,
      "learning_rate": 1.2007766990291264e-06,
      "loss": 0.1484,
      "step": 363060
    },
    {
      "epoch": 141.0,
      "learning_rate": 1.2002588996763756e-06,
      "loss": 0.0888,
      "step": 363070
    },
    {
      "epoch": 141.0,
      "eval_accuracy": 0.9535075653370014,
      "eval_loss": 0.3850953280925751,
      "eval_runtime": 8.289,
      "eval_samples_per_second": 438.535,
      "eval_steps_per_second": 54.892,
      "step": 363075
    },
    {
      "epoch": 141.0,
      "learning_rate": 1.1997411003236245e-06,
      "loss": 0.086,
      "step": 363080
    },
    {
      "epoch": 141.01,
      "learning_rate": 1.1992233009708737e-06,
      "loss": 0.0333,
      "step": 363090
    },
    {
      "epoch": 141.01,
      "learning_rate": 1.198705501618123e-06,
      "loss": 0.033,
      "step": 363100
    },
    {
      "epoch": 141.01,
      "learning_rate": 1.1981877022653721e-06,
      "loss": 0.0112,
      "step": 363110
    },
    {
      "epoch": 141.02,
      "learning_rate": 1.1976699029126213e-06,
      "loss": 0.1333,
      "step": 363120
    },
    {
      "epoch": 141.02,
      "learning_rate": 1.1971521035598707e-06,
      "loss": 0.0003,
      "step": 363130
    },
    {
      "epoch": 141.03,
      "learning_rate": 1.19663430420712e-06,
      "loss": 0.0507,
      "step": 363140
    },
    {
      "epoch": 141.03,
      "learning_rate": 1.1961165048543691e-06,
      "loss": 0.0301,
      "step": 363150
    },
    {
      "epoch": 141.03,
      "learning_rate": 1.1955987055016183e-06,
      "loss": 0.1031,
      "step": 363160
    },
    {
      "epoch": 141.04,
      "learning_rate": 1.1950809061488675e-06,
      "loss": 0.017,
      "step": 363170
    },
    {
      "epoch": 141.04,
      "learning_rate": 1.1945631067961167e-06,
      "loss": 0.0851,
      "step": 363180
    },
    {
      "epoch": 141.04,
      "learning_rate": 1.194045307443366e-06,
      "loss": 0.0165,
      "step": 363190
    },
    {
      "epoch": 141.05,
      "learning_rate": 1.193527508090615e-06,
      "loss": 0.0082,
      "step": 363200
    },
    {
      "epoch": 141.05,
      "learning_rate": 1.1930097087378643e-06,
      "loss": 0.036,
      "step": 363210
    },
    {
      "epoch": 141.06,
      "learning_rate": 1.1924919093851135e-06,
      "loss": 0.0986,
      "step": 363220
    },
    {
      "epoch": 141.06,
      "learning_rate": 1.1919741100323627e-06,
      "loss": 0.0351,
      "step": 363230
    },
    {
      "epoch": 141.06,
      "learning_rate": 1.1914563106796117e-06,
      "loss": 0.0572,
      "step": 363240
    },
    {
      "epoch": 141.07,
      "learning_rate": 1.1909385113268608e-06,
      "loss": 0.044,
      "step": 363250
    },
    {
      "epoch": 141.07,
      "learning_rate": 1.19042071197411e-06,
      "loss": 0.0277,
      "step": 363260
    },
    {
      "epoch": 141.08,
      "learning_rate": 1.1899029126213592e-06,
      "loss": 0.0215,
      "step": 363270
    },
    {
      "epoch": 141.08,
      "learning_rate": 1.1893851132686084e-06,
      "loss": 0.0196,
      "step": 363280
    },
    {
      "epoch": 141.08,
      "learning_rate": 1.1888673139158576e-06,
      "loss": 0.1493,
      "step": 363290
    },
    {
      "epoch": 141.09,
      "learning_rate": 1.1883495145631068e-06,
      "loss": 0.0732,
      "step": 363300
    },
    {
      "epoch": 141.09,
      "learning_rate": 1.187831715210356e-06,
      "loss": 0.0153,
      "step": 363310
    },
    {
      "epoch": 141.1,
      "learning_rate": 1.1873139158576052e-06,
      "loss": 0.0447,
      "step": 363320
    },
    {
      "epoch": 141.1,
      "learning_rate": 1.1867961165048544e-06,
      "loss": 0.0188,
      "step": 363330
    },
    {
      "epoch": 141.1,
      "learning_rate": 1.1862783171521036e-06,
      "loss": 0.0251,
      "step": 363340
    },
    {
      "epoch": 141.11,
      "learning_rate": 1.1857605177993528e-06,
      "loss": 0.012,
      "step": 363350
    },
    {
      "epoch": 141.11,
      "learning_rate": 1.185242718446602e-06,
      "loss": 0.0321,
      "step": 363360
    },
    {
      "epoch": 141.11,
      "learning_rate": 1.1847249190938512e-06,
      "loss": 0.1478,
      "step": 363370
    },
    {
      "epoch": 141.12,
      "learning_rate": 1.1842071197411004e-06,
      "loss": 0.005,
      "step": 363380
    },
    {
      "epoch": 141.12,
      "learning_rate": 1.1836893203883496e-06,
      "loss": 0.0637,
      "step": 363390
    },
    {
      "epoch": 141.13,
      "learning_rate": 1.1831715210355988e-06,
      "loss": 0.0405,
      "step": 363400
    },
    {
      "epoch": 141.13,
      "learning_rate": 1.182653721682848e-06,
      "loss": 0.0499,
      "step": 363410
    },
    {
      "epoch": 141.13,
      "learning_rate": 1.1821359223300972e-06,
      "loss": 0.1099,
      "step": 363420
    },
    {
      "epoch": 141.14,
      "learning_rate": 1.1816181229773464e-06,
      "loss": 0.0003,
      "step": 363430
    },
    {
      "epoch": 141.14,
      "learning_rate": 1.1811003236245955e-06,
      "loss": 0.0083,
      "step": 363440
    },
    {
      "epoch": 141.15,
      "learning_rate": 1.1805825242718447e-06,
      "loss": 0.1032,
      "step": 363450
    },
    {
      "epoch": 141.15,
      "learning_rate": 1.180064724919094e-06,
      "loss": 0.0365,
      "step": 363460
    },
    {
      "epoch": 141.15,
      "learning_rate": 1.1795469255663431e-06,
      "loss": 0.0298,
      "step": 363470
    },
    {
      "epoch": 141.16,
      "learning_rate": 1.1790291262135923e-06,
      "loss": 0.0226,
      "step": 363480
    },
    {
      "epoch": 141.16,
      "learning_rate": 1.1785113268608415e-06,
      "loss": 0.0603,
      "step": 363490
    },
    {
      "epoch": 141.17,
      "learning_rate": 1.1779935275080907e-06,
      "loss": 0.0072,
      "step": 363500
    },
    {
      "epoch": 141.17,
      "learning_rate": 1.17747572815534e-06,
      "loss": 0.1088,
      "step": 363510
    },
    {
      "epoch": 141.17,
      "learning_rate": 1.176957928802589e-06,
      "loss": 0.0006,
      "step": 363520
    },
    {
      "epoch": 141.18,
      "learning_rate": 1.1764401294498383e-06,
      "loss": 0.0202,
      "step": 363530
    },
    {
      "epoch": 141.18,
      "learning_rate": 1.1759223300970875e-06,
      "loss": 0.0638,
      "step": 363540
    },
    {
      "epoch": 141.18,
      "learning_rate": 1.1754045307443367e-06,
      "loss": 0.0617,
      "step": 363550
    },
    {
      "epoch": 141.19,
      "learning_rate": 1.1748867313915859e-06,
      "loss": 0.0103,
      "step": 363560
    },
    {
      "epoch": 141.19,
      "learning_rate": 1.174368932038835e-06,
      "loss": 0.1234,
      "step": 363570
    },
    {
      "epoch": 141.2,
      "learning_rate": 1.1738511326860843e-06,
      "loss": 0.0357,
      "step": 363580
    },
    {
      "epoch": 141.2,
      "learning_rate": 1.1733333333333335e-06,
      "loss": 0.0004,
      "step": 363590
    },
    {
      "epoch": 141.2,
      "learning_rate": 1.1728155339805827e-06,
      "loss": 0.015,
      "step": 363600
    },
    {
      "epoch": 141.21,
      "learning_rate": 1.1722977346278319e-06,
      "loss": 0.0003,
      "step": 363610
    },
    {
      "epoch": 141.21,
      "learning_rate": 1.171779935275081e-06,
      "loss": 0.0154,
      "step": 363620
    },
    {
      "epoch": 141.22,
      "learning_rate": 1.1712621359223302e-06,
      "loss": 0.0626,
      "step": 363630
    },
    {
      "epoch": 141.22,
      "learning_rate": 1.1707443365695794e-06,
      "loss": 0.0551,
      "step": 363640
    },
    {
      "epoch": 141.22,
      "learning_rate": 1.1702265372168284e-06,
      "loss": 0.0339,
      "step": 363650
    },
    {
      "epoch": 141.23,
      "learning_rate": 1.1697087378640776e-06,
      "loss": 0.0402,
      "step": 363660
    },
    {
      "epoch": 141.23,
      "learning_rate": 1.1691909385113268e-06,
      "loss": 0.027,
      "step": 363670
    },
    {
      "epoch": 141.23,
      "learning_rate": 1.168673139158576e-06,
      "loss": 0.0307,
      "step": 363680
    },
    {
      "epoch": 141.24,
      "learning_rate": 1.1681553398058254e-06,
      "loss": 0.0452,
      "step": 363690
    },
    {
      "epoch": 141.24,
      "learning_rate": 1.1676375404530746e-06,
      "loss": 0.0745,
      "step": 363700
    },
    {
      "epoch": 141.25,
      "learning_rate": 1.1671197411003238e-06,
      "loss": 0.0302,
      "step": 363710
    },
    {
      "epoch": 141.25,
      "learning_rate": 1.166601941747573e-06,
      "loss": 0.0003,
      "step": 363720
    },
    {
      "epoch": 141.25,
      "learning_rate": 1.1660841423948222e-06,
      "loss": 0.0137,
      "step": 363730
    },
    {
      "epoch": 141.26,
      "learning_rate": 1.1655663430420714e-06,
      "loss": 0.0001,
      "step": 363740
    },
    {
      "epoch": 141.26,
      "learning_rate": 1.1650485436893206e-06,
      "loss": 0.0001,
      "step": 363750
    },
    {
      "epoch": 141.27,
      "learning_rate": 1.1645307443365698e-06,
      "loss": 0.1645,
      "step": 363760
    },
    {
      "epoch": 141.27,
      "learning_rate": 1.164012944983819e-06,
      "loss": 0.0188,
      "step": 363770
    },
    {
      "epoch": 141.27,
      "learning_rate": 1.1634951456310682e-06,
      "loss": 0.0401,
      "step": 363780
    },
    {
      "epoch": 141.28,
      "learning_rate": 1.1629773462783174e-06,
      "loss": 0.0321,
      "step": 363790
    },
    {
      "epoch": 141.28,
      "learning_rate": 1.1624595469255665e-06,
      "loss": 0.0683,
      "step": 363800
    },
    {
      "epoch": 141.29,
      "learning_rate": 1.1619417475728155e-06,
      "loss": 0.01,
      "step": 363810
    },
    {
      "epoch": 141.29,
      "learning_rate": 1.1614239482200647e-06,
      "loss": 0.0096,
      "step": 363820
    },
    {
      "epoch": 141.29,
      "learning_rate": 1.160906148867314e-06,
      "loss": 0.0312,
      "step": 363830
    },
    {
      "epoch": 141.3,
      "learning_rate": 1.1603883495145631e-06,
      "loss": 0.0096,
      "step": 363840
    },
    {
      "epoch": 141.3,
      "learning_rate": 1.1598705501618123e-06,
      "loss": 0.0254,
      "step": 363850
    },
    {
      "epoch": 141.3,
      "learning_rate": 1.1593527508090615e-06,
      "loss": 0.0292,
      "step": 363860
    },
    {
      "epoch": 141.31,
      "learning_rate": 1.1588349514563107e-06,
      "loss": 0.0079,
      "step": 363870
    },
    {
      "epoch": 141.31,
      "learning_rate": 1.1583171521035599e-06,
      "loss": 0.0631,
      "step": 363880
    },
    {
      "epoch": 141.32,
      "learning_rate": 1.157799352750809e-06,
      "loss": 0.136,
      "step": 363890
    },
    {
      "epoch": 141.32,
      "learning_rate": 1.1572815533980583e-06,
      "loss": 0.0748,
      "step": 363900
    },
    {
      "epoch": 141.32,
      "learning_rate": 1.1567637540453075e-06,
      "loss": 0.0446,
      "step": 363910
    },
    {
      "epoch": 141.33,
      "learning_rate": 1.1562459546925567e-06,
      "loss": 0.0108,
      "step": 363920
    },
    {
      "epoch": 141.33,
      "learning_rate": 1.1557281553398059e-06,
      "loss": 0.0099,
      "step": 363930
    },
    {
      "epoch": 141.34,
      "learning_rate": 1.155210355987055e-06,
      "loss": 0.0152,
      "step": 363940
    },
    {
      "epoch": 141.34,
      "learning_rate": 1.1546925566343043e-06,
      "loss": 0.0003,
      "step": 363950
    },
    {
      "epoch": 141.34,
      "learning_rate": 1.1541747572815534e-06,
      "loss": 0.0505,
      "step": 363960
    },
    {
      "epoch": 141.35,
      "learning_rate": 1.1536569579288029e-06,
      "loss": 0.0001,
      "step": 363970
    },
    {
      "epoch": 141.35,
      "learning_rate": 1.1531391585760518e-06,
      "loss": 0.1009,
      "step": 363980
    },
    {
      "epoch": 141.36,
      "learning_rate": 1.152621359223301e-06,
      "loss": 0.027,
      "step": 363990
    },
    {
      "epoch": 141.36,
      "learning_rate": 1.1521035598705502e-06,
      "loss": 0.0746,
      "step": 364000
    },
    {
      "epoch": 141.36,
      "learning_rate": 1.1515857605177994e-06,
      "loss": 0.1233,
      "step": 364010
    },
    {
      "epoch": 141.37,
      "learning_rate": 1.1510679611650486e-06,
      "loss": 0.0111,
      "step": 364020
    },
    {
      "epoch": 141.37,
      "learning_rate": 1.1505501618122978e-06,
      "loss": 0.0023,
      "step": 364030
    },
    {
      "epoch": 141.37,
      "learning_rate": 1.150032362459547e-06,
      "loss": 0.0302,
      "step": 364040
    },
    {
      "epoch": 141.38,
      "learning_rate": 1.1495145631067962e-06,
      "loss": 0.0007,
      "step": 364050
    },
    {
      "epoch": 141.38,
      "learning_rate": 1.1489967637540454e-06,
      "loss": 0.0091,
      "step": 364060
    },
    {
      "epoch": 141.39,
      "learning_rate": 1.1484789644012946e-06,
      "loss": 0.0033,
      "step": 364070
    },
    {
      "epoch": 141.39,
      "learning_rate": 1.1479611650485438e-06,
      "loss": 0.0011,
      "step": 364080
    },
    {
      "epoch": 141.39,
      "learning_rate": 1.147443365695793e-06,
      "loss": 0.0003,
      "step": 364090
    },
    {
      "epoch": 141.4,
      "learning_rate": 1.1469255663430422e-06,
      "loss": 0.135,
      "step": 364100
    },
    {
      "epoch": 141.4,
      "learning_rate": 1.1464077669902914e-06,
      "loss": 0.0832,
      "step": 364110
    },
    {
      "epoch": 141.41,
      "learning_rate": 1.1458899676375406e-06,
      "loss": 0.009,
      "step": 364120
    },
    {
      "epoch": 141.41,
      "learning_rate": 1.1453721682847898e-06,
      "loss": 0.0401,
      "step": 364130
    },
    {
      "epoch": 141.41,
      "learning_rate": 1.144854368932039e-06,
      "loss": 0.0028,
      "step": 364140
    },
    {
      "epoch": 141.42,
      "learning_rate": 1.1443365695792881e-06,
      "loss": 0.0281,
      "step": 364150
    },
    {
      "epoch": 141.42,
      "learning_rate": 1.1438187702265373e-06,
      "loss": 0.0263,
      "step": 364160
    },
    {
      "epoch": 141.43,
      "learning_rate": 1.1433009708737865e-06,
      "loss": 0.1184,
      "step": 364170
    },
    {
      "epoch": 141.43,
      "learning_rate": 1.1427831715210357e-06,
      "loss": 0.0205,
      "step": 364180
    },
    {
      "epoch": 141.43,
      "learning_rate": 1.142265372168285e-06,
      "loss": 0.0296,
      "step": 364190
    },
    {
      "epoch": 141.44,
      "learning_rate": 1.1417475728155341e-06,
      "loss": 0.0001,
      "step": 364200
    },
    {
      "epoch": 141.44,
      "learning_rate": 1.1412297734627833e-06,
      "loss": 0.1048,
      "step": 364210
    },
    {
      "epoch": 141.44,
      "learning_rate": 1.1407119741100323e-06,
      "loss": 0.0939,
      "step": 364220
    },
    {
      "epoch": 141.45,
      "learning_rate": 1.1401941747572815e-06,
      "loss": 0.0512,
      "step": 364230
    },
    {
      "epoch": 141.45,
      "learning_rate": 1.1396763754045307e-06,
      "loss": 0.0359,
      "step": 364240
    },
    {
      "epoch": 141.46,
      "learning_rate": 1.13915857605178e-06,
      "loss": 0.0035,
      "step": 364250
    },
    {
      "epoch": 141.46,
      "learning_rate": 1.1386407766990293e-06,
      "loss": 0.0005,
      "step": 364260
    },
    {
      "epoch": 141.46,
      "learning_rate": 1.1381229773462785e-06,
      "loss": 0.0001,
      "step": 364270
    },
    {
      "epoch": 141.47,
      "learning_rate": 1.1376051779935277e-06,
      "loss": 0.0245,
      "step": 364280
    },
    {
      "epoch": 141.47,
      "learning_rate": 1.1370873786407769e-06,
      "loss": 0.0764,
      "step": 364290
    },
    {
      "epoch": 141.48,
      "learning_rate": 1.136569579288026e-06,
      "loss": 0.0912,
      "step": 364300
    },
    {
      "epoch": 141.48,
      "learning_rate": 1.1360517799352753e-06,
      "loss": 0.1404,
      "step": 364310
    },
    {
      "epoch": 141.48,
      "learning_rate": 1.1355339805825245e-06,
      "loss": 0.0004,
      "step": 364320
    },
    {
      "epoch": 141.49,
      "learning_rate": 1.1350161812297736e-06,
      "loss": 0.0198,
      "step": 364330
    },
    {
      "epoch": 141.49,
      "learning_rate": 1.1344983818770228e-06,
      "loss": 0.0856,
      "step": 364340
    },
    {
      "epoch": 141.5,
      "learning_rate": 1.133980582524272e-06,
      "loss": 0.0074,
      "step": 364350
    },
    {
      "epoch": 141.5,
      "learning_rate": 1.1334627831715212e-06,
      "loss": 0.0812,
      "step": 364360
    },
    {
      "epoch": 141.5,
      "learning_rate": 1.1329449838187704e-06,
      "loss": 0.0009,
      "step": 364370
    },
    {
      "epoch": 141.51,
      "learning_rate": 1.1324271844660194e-06,
      "loss": 0.035,
      "step": 364380
    },
    {
      "epoch": 141.51,
      "learning_rate": 1.1319093851132686e-06,
      "loss": 0.0003,
      "step": 364390
    },
    {
      "epoch": 141.51,
      "learning_rate": 1.1313915857605178e-06,
      "loss": 0.0268,
      "step": 364400
    },
    {
      "epoch": 141.52,
      "learning_rate": 1.130873786407767e-06,
      "loss": 0.0897,
      "step": 364410
    },
    {
      "epoch": 141.52,
      "learning_rate": 1.1303559870550162e-06,
      "loss": 0.0007,
      "step": 364420
    },
    {
      "epoch": 141.53,
      "learning_rate": 1.1298381877022654e-06,
      "loss": 0.1164,
      "step": 364430
    },
    {
      "epoch": 141.53,
      "learning_rate": 1.1293203883495146e-06,
      "loss": 0.0267,
      "step": 364440
    },
    {
      "epoch": 141.53,
      "learning_rate": 1.1288025889967638e-06,
      "loss": 0.0092,
      "step": 364450
    },
    {
      "epoch": 141.54,
      "learning_rate": 1.128284789644013e-06,
      "loss": 0.1292,
      "step": 364460
    },
    {
      "epoch": 141.54,
      "learning_rate": 1.1277669902912622e-06,
      "loss": 0.0776,
      "step": 364470
    },
    {
      "epoch": 141.55,
      "learning_rate": 1.1272491909385114e-06,
      "loss": 0.0579,
      "step": 364480
    },
    {
      "epoch": 141.55,
      "learning_rate": 1.1267313915857605e-06,
      "loss": 0.0452,
      "step": 364490
    },
    {
      "epoch": 141.55,
      "learning_rate": 1.1262135922330097e-06,
      "loss": 0.1305,
      "step": 364500
    },
    {
      "epoch": 141.56,
      "learning_rate": 1.125695792880259e-06,
      "loss": 0.0124,
      "step": 364510
    },
    {
      "epoch": 141.56,
      "learning_rate": 1.1251779935275081e-06,
      "loss": 0.0775,
      "step": 364520
    },
    {
      "epoch": 141.57,
      "learning_rate": 1.1246601941747575e-06,
      "loss": 0.1315,
      "step": 364530
    },
    {
      "epoch": 141.57,
      "learning_rate": 1.1241423948220067e-06,
      "loss": 0.0112,
      "step": 364540
    },
    {
      "epoch": 141.57,
      "learning_rate": 1.1236245954692557e-06,
      "loss": 0.0089,
      "step": 364550
    },
    {
      "epoch": 141.58,
      "learning_rate": 1.123106796116505e-06,
      "loss": 0.0145,
      "step": 364560
    },
    {
      "epoch": 141.58,
      "learning_rate": 1.122588996763754e-06,
      "loss": 0.04,
      "step": 364570
    },
    {
      "epoch": 141.58,
      "learning_rate": 1.1220711974110033e-06,
      "loss": 0.0512,
      "step": 364580
    },
    {
      "epoch": 141.59,
      "learning_rate": 1.1215533980582525e-06,
      "loss": 0.1498,
      "step": 364590
    },
    {
      "epoch": 141.59,
      "learning_rate": 1.1210355987055017e-06,
      "loss": 0.0307,
      "step": 364600
    },
    {
      "epoch": 141.6,
      "learning_rate": 1.1205177993527509e-06,
      "loss": 0.009,
      "step": 364610
    },
    {
      "epoch": 141.6,
      "learning_rate": 1.12e-06,
      "loss": 0.1615,
      "step": 364620
    },
    {
      "epoch": 141.6,
      "learning_rate": 1.1194822006472493e-06,
      "loss": 0.026,
      "step": 364630
    },
    {
      "epoch": 141.61,
      "learning_rate": 1.1189644012944985e-06,
      "loss": 0.0021,
      "step": 364640
    },
    {
      "epoch": 141.61,
      "learning_rate": 1.1184466019417477e-06,
      "loss": 0.0186,
      "step": 364650
    },
    {
      "epoch": 141.62,
      "learning_rate": 1.1179288025889969e-06,
      "loss": 0.0427,
      "step": 364660
    },
    {
      "epoch": 141.62,
      "learning_rate": 1.117411003236246e-06,
      "loss": 0.0285,
      "step": 364670
    },
    {
      "epoch": 141.62,
      "learning_rate": 1.1168932038834952e-06,
      "loss": 0.0473,
      "step": 364680
    },
    {
      "epoch": 141.63,
      "learning_rate": 1.1163754045307444e-06,
      "loss": 0.0169,
      "step": 364690
    },
    {
      "epoch": 141.63,
      "learning_rate": 1.1158576051779936e-06,
      "loss": 0.0176,
      "step": 364700
    },
    {
      "epoch": 141.63,
      "learning_rate": 1.1153398058252428e-06,
      "loss": 0.051,
      "step": 364710
    },
    {
      "epoch": 141.64,
      "learning_rate": 1.114822006472492e-06,
      "loss": 0.0591,
      "step": 364720
    },
    {
      "epoch": 141.64,
      "learning_rate": 1.1143042071197412e-06,
      "loss": 0.0192,
      "step": 364730
    },
    {
      "epoch": 141.65,
      "learning_rate": 1.1137864077669904e-06,
      "loss": 0.0561,
      "step": 364740
    },
    {
      "epoch": 141.65,
      "learning_rate": 1.1132686084142396e-06,
      "loss": 0.1368,
      "step": 364750
    },
    {
      "epoch": 141.65,
      "learning_rate": 1.1127508090614888e-06,
      "loss": 0.0396,
      "step": 364760
    },
    {
      "epoch": 141.66,
      "learning_rate": 1.112233009708738e-06,
      "loss": 0.0716,
      "step": 364770
    },
    {
      "epoch": 141.66,
      "learning_rate": 1.1117152103559872e-06,
      "loss": 0.0536,
      "step": 364780
    },
    {
      "epoch": 141.67,
      "learning_rate": 1.1111974110032362e-06,
      "loss": 0.0308,
      "step": 364790
    },
    {
      "epoch": 141.67,
      "learning_rate": 1.1106796116504854e-06,
      "loss": 0.0995,
      "step": 364800
    },
    {
      "epoch": 141.67,
      "learning_rate": 1.1101618122977348e-06,
      "loss": 0.0494,
      "step": 364810
    },
    {
      "epoch": 141.68,
      "learning_rate": 1.109644012944984e-06,
      "loss": 0.0559,
      "step": 364820
    },
    {
      "epoch": 141.68,
      "learning_rate": 1.1091262135922332e-06,
      "loss": 0.0409,
      "step": 364830
    },
    {
      "epoch": 141.69,
      "learning_rate": 1.1086084142394824e-06,
      "loss": 0.1285,
      "step": 364840
    },
    {
      "epoch": 141.69,
      "learning_rate": 1.1080906148867316e-06,
      "loss": 0.002,
      "step": 364850
    },
    {
      "epoch": 141.69,
      "learning_rate": 1.1075728155339807e-06,
      "loss": 0.0509,
      "step": 364860
    },
    {
      "epoch": 141.7,
      "learning_rate": 1.10705501618123e-06,
      "loss": 0.03,
      "step": 364870
    },
    {
      "epoch": 141.7,
      "learning_rate": 1.1065372168284791e-06,
      "loss": 0.001,
      "step": 364880
    },
    {
      "epoch": 141.7,
      "learning_rate": 1.1060194174757283e-06,
      "loss": 0.0906,
      "step": 364890
    },
    {
      "epoch": 141.71,
      "learning_rate": 1.1055016181229775e-06,
      "loss": 0.0805,
      "step": 364900
    },
    {
      "epoch": 141.71,
      "learning_rate": 1.1049838187702267e-06,
      "loss": 0.0004,
      "step": 364910
    },
    {
      "epoch": 141.72,
      "learning_rate": 1.104466019417476e-06,
      "loss": 0.0466,
      "step": 364920
    },
    {
      "epoch": 141.72,
      "learning_rate": 1.1039482200647251e-06,
      "loss": 0.0839,
      "step": 364930
    },
    {
      "epoch": 141.72,
      "learning_rate": 1.1034304207119743e-06,
      "loss": 0.1066,
      "step": 364940
    },
    {
      "epoch": 141.73,
      "learning_rate": 1.1029126213592235e-06,
      "loss": 0.1243,
      "step": 364950
    },
    {
      "epoch": 141.73,
      "learning_rate": 1.1023948220064725e-06,
      "loss": 0.0001,
      "step": 364960
    },
    {
      "epoch": 141.74,
      "learning_rate": 1.1018770226537217e-06,
      "loss": 0.0278,
      "step": 364970
    },
    {
      "epoch": 141.74,
      "learning_rate": 1.1013592233009709e-06,
      "loss": 0.0587,
      "step": 364980
    },
    {
      "epoch": 141.74,
      "learning_rate": 1.10084142394822e-06,
      "loss": 0.1067,
      "step": 364990
    },
    {
      "epoch": 141.75,
      "learning_rate": 1.1003236245954693e-06,
      "loss": 0.0716,
      "step": 365000
    },
    {
      "epoch": 141.75,
      "learning_rate": 1.0998058252427185e-06,
      "loss": 0.0093,
      "step": 365010
    },
    {
      "epoch": 141.76,
      "learning_rate": 1.0992880258899676e-06,
      "loss": 0.0167,
      "step": 365020
    },
    {
      "epoch": 141.76,
      "learning_rate": 1.0987702265372168e-06,
      "loss": 0.0092,
      "step": 365030
    },
    {
      "epoch": 141.76,
      "learning_rate": 1.098252427184466e-06,
      "loss": 0.1004,
      "step": 365040
    },
    {
      "epoch": 141.77,
      "learning_rate": 1.0977346278317152e-06,
      "loss": 0.0419,
      "step": 365050
    },
    {
      "epoch": 141.77,
      "learning_rate": 1.0972168284789644e-06,
      "loss": 0.0401,
      "step": 365060
    },
    {
      "epoch": 141.77,
      "learning_rate": 1.0966990291262136e-06,
      "loss": 0.0073,
      "step": 365070
    },
    {
      "epoch": 141.78,
      "learning_rate": 1.0961812297734628e-06,
      "loss": 0.0249,
      "step": 365080
    },
    {
      "epoch": 141.78,
      "learning_rate": 1.0956634304207122e-06,
      "loss": 0.0642,
      "step": 365090
    },
    {
      "epoch": 141.79,
      "learning_rate": 1.0951456310679614e-06,
      "loss": 0.0862,
      "step": 365100
    },
    {
      "epoch": 141.79,
      "learning_rate": 1.0946278317152106e-06,
      "loss": 0.0357,
      "step": 365110
    },
    {
      "epoch": 141.79,
      "learning_rate": 1.0941100323624596e-06,
      "loss": 0.0002,
      "step": 365120
    },
    {
      "epoch": 141.8,
      "learning_rate": 1.0935922330097088e-06,
      "loss": 0.0406,
      "step": 365130
    },
    {
      "epoch": 141.8,
      "learning_rate": 1.093074433656958e-06,
      "loss": 0.0232,
      "step": 365140
    },
    {
      "epoch": 141.81,
      "learning_rate": 1.0925566343042072e-06,
      "loss": 0.0087,
      "step": 365150
    },
    {
      "epoch": 141.81,
      "learning_rate": 1.0920388349514564e-06,
      "loss": 0.0173,
      "step": 365160
    },
    {
      "epoch": 141.81,
      "learning_rate": 1.0915210355987056e-06,
      "loss": 0.0005,
      "step": 365170
    },
    {
      "epoch": 141.82,
      "learning_rate": 1.0910032362459548e-06,
      "loss": 0.0201,
      "step": 365180
    },
    {
      "epoch": 141.82,
      "learning_rate": 1.090485436893204e-06,
      "loss": 0.0162,
      "step": 365190
    },
    {
      "epoch": 141.83,
      "learning_rate": 1.0899676375404531e-06,
      "loss": 0.0006,
      "step": 365200
    },
    {
      "epoch": 141.83,
      "learning_rate": 1.0894498381877023e-06,
      "loss": 0.0166,
      "step": 365210
    },
    {
      "epoch": 141.83,
      "learning_rate": 1.0889320388349515e-06,
      "loss": 0.0147,
      "step": 365220
    },
    {
      "epoch": 141.84,
      "learning_rate": 1.0884142394822007e-06,
      "loss": 0.1097,
      "step": 365230
    },
    {
      "epoch": 141.84,
      "learning_rate": 1.08789644012945e-06,
      "loss": 0.0305,
      "step": 365240
    },
    {
      "epoch": 141.84,
      "learning_rate": 1.0873786407766991e-06,
      "loss": 0.0107,
      "step": 365250
    },
    {
      "epoch": 141.85,
      "learning_rate": 1.0868608414239483e-06,
      "loss": 0.0179,
      "step": 365260
    },
    {
      "epoch": 141.85,
      "learning_rate": 1.0863430420711975e-06,
      "loss": 0.0467,
      "step": 365270
    },
    {
      "epoch": 141.86,
      "learning_rate": 1.0858252427184467e-06,
      "loss": 0.0001,
      "step": 365280
    },
    {
      "epoch": 141.86,
      "learning_rate": 1.085307443365696e-06,
      "loss": 0.002,
      "step": 365290
    },
    {
      "epoch": 141.86,
      "learning_rate": 1.084789644012945e-06,
      "loss": 0.0001,
      "step": 365300
    },
    {
      "epoch": 141.87,
      "learning_rate": 1.0842718446601943e-06,
      "loss": 0.0097,
      "step": 365310
    },
    {
      "epoch": 141.87,
      "learning_rate": 1.0837540453074435e-06,
      "loss": 0.0159,
      "step": 365320
    },
    {
      "epoch": 141.88,
      "learning_rate": 1.0832362459546927e-06,
      "loss": 0.0002,
      "step": 365330
    },
    {
      "epoch": 141.88,
      "learning_rate": 1.0827184466019419e-06,
      "loss": 0.0008,
      "step": 365340
    },
    {
      "epoch": 141.88,
      "learning_rate": 1.082200647249191e-06,
      "loss": 0.0814,
      "step": 365350
    },
    {
      "epoch": 141.89,
      "learning_rate": 1.08168284789644e-06,
      "loss": 0.0207,
      "step": 365360
    },
    {
      "epoch": 141.89,
      "learning_rate": 1.0811650485436895e-06,
      "loss": 0.0116,
      "step": 365370
    },
    {
      "epoch": 141.9,
      "learning_rate": 1.0806472491909386e-06,
      "loss": 0.0216,
      "step": 365380
    },
    {
      "epoch": 141.9,
      "learning_rate": 1.0801294498381878e-06,
      "loss": 0.0241,
      "step": 365390
    },
    {
      "epoch": 141.9,
      "learning_rate": 1.079611650485437e-06,
      "loss": 0.0068,
      "step": 365400
    },
    {
      "epoch": 141.91,
      "learning_rate": 1.0790938511326862e-06,
      "loss": 0.0317,
      "step": 365410
    },
    {
      "epoch": 141.91,
      "learning_rate": 1.0785760517799354e-06,
      "loss": 0.0239,
      "step": 365420
    },
    {
      "epoch": 141.91,
      "learning_rate": 1.0780582524271846e-06,
      "loss": 0.0089,
      "step": 365430
    },
    {
      "epoch": 141.92,
      "learning_rate": 1.0775404530744338e-06,
      "loss": 0.0139,
      "step": 365440
    },
    {
      "epoch": 141.92,
      "learning_rate": 1.077022653721683e-06,
      "loss": 0.0099,
      "step": 365450
    },
    {
      "epoch": 141.93,
      "learning_rate": 1.0765048543689322e-06,
      "loss": 0.0425,
      "step": 365460
    },
    {
      "epoch": 141.93,
      "learning_rate": 1.0759870550161814e-06,
      "loss": 0.0489,
      "step": 365470
    },
    {
      "epoch": 141.93,
      "learning_rate": 1.0754692556634306e-06,
      "loss": 0.0659,
      "step": 365480
    },
    {
      "epoch": 141.94,
      "learning_rate": 1.0749514563106798e-06,
      "loss": 0.0169,
      "step": 365490
    },
    {
      "epoch": 141.94,
      "learning_rate": 1.074433656957929e-06,
      "loss": 0.0096,
      "step": 365500
    },
    {
      "epoch": 141.95,
      "learning_rate": 1.0739158576051782e-06,
      "loss": 0.0249,
      "step": 365510
    },
    {
      "epoch": 141.95,
      "learning_rate": 1.0733980582524274e-06,
      "loss": 0.0152,
      "step": 365520
    },
    {
      "epoch": 141.95,
      "learning_rate": 1.0728802588996764e-06,
      "loss": 0.0188,
      "step": 365530
    },
    {
      "epoch": 141.96,
      "learning_rate": 1.0723624595469255e-06,
      "loss": 0.0651,
      "step": 365540
    },
    {
      "epoch": 141.96,
      "learning_rate": 1.0718446601941747e-06,
      "loss": 0.023,
      "step": 365550
    },
    {
      "epoch": 141.97,
      "learning_rate": 1.071326860841424e-06,
      "loss": 0.0393,
      "step": 365560
    },
    {
      "epoch": 141.97,
      "learning_rate": 1.0708090614886731e-06,
      "loss": 0.0842,
      "step": 365570
    },
    {
      "epoch": 141.97,
      "learning_rate": 1.0702912621359223e-06,
      "loss": 0.0324,
      "step": 365580
    },
    {
      "epoch": 141.98,
      "learning_rate": 1.0697734627831715e-06,
      "loss": 0.0597,
      "step": 365590
    },
    {
      "epoch": 141.98,
      "learning_rate": 1.0692556634304207e-06,
      "loss": 0.0232,
      "step": 365600
    },
    {
      "epoch": 141.98,
      "learning_rate": 1.06873786407767e-06,
      "loss": 0.0403,
      "step": 365610
    },
    {
      "epoch": 141.99,
      "learning_rate": 1.068220064724919e-06,
      "loss": 0.0003,
      "step": 365620
    },
    {
      "epoch": 141.99,
      "learning_rate": 1.0677022653721683e-06,
      "loss": 0.0381,
      "step": 365630
    },
    {
      "epoch": 142.0,
      "learning_rate": 1.0671844660194175e-06,
      "loss": 0.1106,
      "step": 365640
    },
    {
      "epoch": 142.0,
      "learning_rate": 1.066666666666667e-06,
      "loss": 0.0169,
      "step": 365650
    },
    {
      "epoch": 142.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.3907836973667145,
      "eval_runtime": 8.2566,
      "eval_samples_per_second": 440.255,
      "eval_steps_per_second": 55.108,
      "step": 365650
    },
    {
      "epoch": 142.0,
      "learning_rate": 1.066148867313916e-06,
      "loss": 0.0112,
      "step": 365660
    },
    {
      "epoch": 142.01,
      "learning_rate": 1.0656310679611653e-06,
      "loss": 0.0668,
      "step": 365670
    },
    {
      "epoch": 142.01,
      "learning_rate": 1.0651132686084145e-06,
      "loss": 0.0004,
      "step": 365680
    },
    {
      "epoch": 142.02,
      "learning_rate": 1.0645954692556635e-06,
      "loss": 0.0881,
      "step": 365690
    },
    {
      "epoch": 142.02,
      "learning_rate": 1.0640776699029127e-06,
      "loss": 0.077,
      "step": 365700
    },
    {
      "epoch": 142.02,
      "learning_rate": 1.0635598705501619e-06,
      "loss": 0.061,
      "step": 365710
    },
    {
      "epoch": 142.03,
      "learning_rate": 1.063042071197411e-06,
      "loss": 0.0369,
      "step": 365720
    },
    {
      "epoch": 142.03,
      "learning_rate": 1.0625242718446602e-06,
      "loss": 0.0079,
      "step": 365730
    },
    {
      "epoch": 142.03,
      "learning_rate": 1.0620064724919094e-06,
      "loss": 0.0186,
      "step": 365740
    },
    {
      "epoch": 142.04,
      "learning_rate": 1.0614886731391586e-06,
      "loss": 0.0634,
      "step": 365750
    },
    {
      "epoch": 142.04,
      "learning_rate": 1.0609708737864078e-06,
      "loss": 0.1145,
      "step": 365760
    },
    {
      "epoch": 142.05,
      "learning_rate": 1.060453074433657e-06,
      "loss": 0.0702,
      "step": 365770
    },
    {
      "epoch": 142.05,
      "learning_rate": 1.0599352750809062e-06,
      "loss": 0.0935,
      "step": 365780
    },
    {
      "epoch": 142.05,
      "learning_rate": 1.0594174757281554e-06,
      "loss": 0.0536,
      "step": 365790
    },
    {
      "epoch": 142.06,
      "learning_rate": 1.0588996763754046e-06,
      "loss": 0.0755,
      "step": 365800
    },
    {
      "epoch": 142.06,
      "learning_rate": 1.0583818770226538e-06,
      "loss": 0.0888,
      "step": 365810
    },
    {
      "epoch": 142.07,
      "learning_rate": 1.057864077669903e-06,
      "loss": 0.0002,
      "step": 365820
    },
    {
      "epoch": 142.07,
      "learning_rate": 1.0573462783171522e-06,
      "loss": 0.0914,
      "step": 365830
    },
    {
      "epoch": 142.07,
      "learning_rate": 1.0568284789644014e-06,
      "loss": 0.0109,
      "step": 365840
    },
    {
      "epoch": 142.08,
      "learning_rate": 1.0563106796116506e-06,
      "loss": 0.018,
      "step": 365850
    },
    {
      "epoch": 142.08,
      "learning_rate": 1.0557928802588998e-06,
      "loss": 0.014,
      "step": 365860
    },
    {
      "epoch": 142.09,
      "learning_rate": 1.055275080906149e-06,
      "loss": 0.0299,
      "step": 365870
    },
    {
      "epoch": 142.09,
      "learning_rate": 1.0547572815533982e-06,
      "loss": 0.1437,
      "step": 365880
    },
    {
      "epoch": 142.09,
      "learning_rate": 1.0542394822006474e-06,
      "loss": 0.0289,
      "step": 365890
    },
    {
      "epoch": 142.1,
      "learning_rate": 1.0537216828478966e-06,
      "loss": 0.0551,
      "step": 365900
    },
    {
      "epoch": 142.1,
      "learning_rate": 1.0532038834951457e-06,
      "loss": 0.0376,
      "step": 365910
    },
    {
      "epoch": 142.1,
      "learning_rate": 1.052686084142395e-06,
      "loss": 0.0247,
      "step": 365920
    },
    {
      "epoch": 142.11,
      "learning_rate": 1.0521682847896441e-06,
      "loss": 0.0169,
      "step": 365930
    },
    {
      "epoch": 142.11,
      "learning_rate": 1.0516504854368933e-06,
      "loss": 0.0158,
      "step": 365940
    },
    {
      "epoch": 142.12,
      "learning_rate": 1.0511326860841425e-06,
      "loss": 0.0811,
      "step": 365950
    },
    {
      "epoch": 142.12,
      "learning_rate": 1.0506148867313917e-06,
      "loss": 0.0012,
      "step": 365960
    },
    {
      "epoch": 142.12,
      "learning_rate": 1.050097087378641e-06,
      "loss": 0.0226,
      "step": 365970
    },
    {
      "epoch": 142.13,
      "learning_rate": 1.0495792880258901e-06,
      "loss": 0.0007,
      "step": 365980
    },
    {
      "epoch": 142.13,
      "learning_rate": 1.0490614886731393e-06,
      "loss": 0.0118,
      "step": 365990
    },
    {
      "epoch": 142.14,
      "learning_rate": 1.0485436893203885e-06,
      "loss": 0.0424,
      "step": 366000
    },
    {
      "epoch": 142.14,
      "learning_rate": 1.0480258899676377e-06,
      "loss": 0.0331,
      "step": 366010
    },
    {
      "epoch": 142.14,
      "learning_rate": 1.0475080906148869e-06,
      "loss": 0.0341,
      "step": 366020
    },
    {
      "epoch": 142.15,
      "learning_rate": 1.046990291262136e-06,
      "loss": 0.014,
      "step": 366030
    },
    {
      "epoch": 142.15,
      "learning_rate": 1.0464724919093853e-06,
      "loss": 0.0245,
      "step": 366040
    },
    {
      "epoch": 142.16,
      "learning_rate": 1.0459546925566345e-06,
      "loss": 0.0263,
      "step": 366050
    },
    {
      "epoch": 142.16,
      "learning_rate": 1.0454368932038837e-06,
      "loss": 0.0491,
      "step": 366060
    },
    {
      "epoch": 142.16,
      "learning_rate": 1.0449190938511329e-06,
      "loss": 0.0307,
      "step": 366070
    },
    {
      "epoch": 142.17,
      "learning_rate": 1.044401294498382e-06,
      "loss": 0.0188,
      "step": 366080
    },
    {
      "epoch": 142.17,
      "learning_rate": 1.0438834951456312e-06,
      "loss": 0.0005,
      "step": 366090
    },
    {
      "epoch": 142.17,
      "learning_rate": 1.0433656957928802e-06,
      "loss": 0.0102,
      "step": 366100
    },
    {
      "epoch": 142.18,
      "learning_rate": 1.0428478964401294e-06,
      "loss": 0.0007,
      "step": 366110
    },
    {
      "epoch": 142.18,
      "learning_rate": 1.0423300970873786e-06,
      "loss": 0.0008,
      "step": 366120
    },
    {
      "epoch": 142.19,
      "learning_rate": 1.0418122977346278e-06,
      "loss": 0.0893,
      "step": 366130
    },
    {
      "epoch": 142.19,
      "learning_rate": 1.041294498381877e-06,
      "loss": 0.0001,
      "step": 366140
    },
    {
      "epoch": 142.19,
      "learning_rate": 1.0407766990291262e-06,
      "loss": 0.0257,
      "step": 366150
    },
    {
      "epoch": 142.2,
      "learning_rate": 1.0402588996763754e-06,
      "loss": 0.0485,
      "step": 366160
    },
    {
      "epoch": 142.2,
      "learning_rate": 1.0397411003236246e-06,
      "loss": 0.0576,
      "step": 366170
    },
    {
      "epoch": 142.21,
      "learning_rate": 1.0392233009708738e-06,
      "loss": 0.0055,
      "step": 366180
    },
    {
      "epoch": 142.21,
      "learning_rate": 1.038705501618123e-06,
      "loss": 0.0097,
      "step": 366190
    },
    {
      "epoch": 142.21,
      "learning_rate": 1.0381877022653722e-06,
      "loss": 0.0083,
      "step": 366200
    },
    {
      "epoch": 142.22,
      "learning_rate": 1.0376699029126216e-06,
      "loss": 0.0261,
      "step": 366210
    },
    {
      "epoch": 142.22,
      "learning_rate": 1.0371521035598708e-06,
      "loss": 0.0183,
      "step": 366220
    },
    {
      "epoch": 142.23,
      "learning_rate": 1.03663430420712e-06,
      "loss": 0.0428,
      "step": 366230
    },
    {
      "epoch": 142.23,
      "learning_rate": 1.0361165048543692e-06,
      "loss": 0.0008,
      "step": 366240
    },
    {
      "epoch": 142.23,
      "learning_rate": 1.0355987055016184e-06,
      "loss": 0.03,
      "step": 366250
    },
    {
      "epoch": 142.24,
      "learning_rate": 1.0350809061488673e-06,
      "loss": 0.0102,
      "step": 366260
    },
    {
      "epoch": 142.24,
      "learning_rate": 1.0345631067961165e-06,
      "loss": 0.0705,
      "step": 366270
    },
    {
      "epoch": 142.24,
      "learning_rate": 1.0340453074433657e-06,
      "loss": 0.0006,
      "step": 366280
    },
    {
      "epoch": 142.25,
      "learning_rate": 1.033527508090615e-06,
      "loss": 0.0166,
      "step": 366290
    },
    {
      "epoch": 142.25,
      "learning_rate": 1.0330097087378641e-06,
      "loss": 0.0625,
      "step": 366300
    },
    {
      "epoch": 142.26,
      "learning_rate": 1.0324919093851133e-06,
      "loss": 0.08,
      "step": 366310
    },
    {
      "epoch": 142.26,
      "learning_rate": 1.0319741100323625e-06,
      "loss": 0.0389,
      "step": 366320
    },
    {
      "epoch": 142.26,
      "learning_rate": 1.0314563106796117e-06,
      "loss": 0.0031,
      "step": 366330
    },
    {
      "epoch": 142.27,
      "learning_rate": 1.030938511326861e-06,
      "loss": 0.0396,
      "step": 366340
    },
    {
      "epoch": 142.27,
      "learning_rate": 1.03042071197411e-06,
      "loss": 0.0858,
      "step": 366350
    },
    {
      "epoch": 142.28,
      "learning_rate": 1.0299029126213593e-06,
      "loss": 0.0113,
      "step": 366360
    },
    {
      "epoch": 142.28,
      "learning_rate": 1.0293851132686085e-06,
      "loss": 0.0103,
      "step": 366370
    },
    {
      "epoch": 142.28,
      "learning_rate": 1.0288673139158577e-06,
      "loss": 0.0001,
      "step": 366380
    },
    {
      "epoch": 142.29,
      "learning_rate": 1.0283495145631069e-06,
      "loss": 0.0119,
      "step": 366390
    },
    {
      "epoch": 142.29,
      "learning_rate": 1.027831715210356e-06,
      "loss": 0.0485,
      "step": 366400
    },
    {
      "epoch": 142.3,
      "learning_rate": 1.0273139158576053e-06,
      "loss": 0.0164,
      "step": 366410
    },
    {
      "epoch": 142.3,
      "learning_rate": 1.0267961165048545e-06,
      "loss": 0.0099,
      "step": 366420
    },
    {
      "epoch": 142.3,
      "learning_rate": 1.0262783171521037e-06,
      "loss": 0.0928,
      "step": 366430
    },
    {
      "epoch": 142.31,
      "learning_rate": 1.0257605177993528e-06,
      "loss": 0.0634,
      "step": 366440
    },
    {
      "epoch": 142.31,
      "learning_rate": 1.025242718446602e-06,
      "loss": 0.0418,
      "step": 366450
    },
    {
      "epoch": 142.31,
      "learning_rate": 1.0247249190938512e-06,
      "loss": 0.1739,
      "step": 366460
    },
    {
      "epoch": 142.32,
      "learning_rate": 1.0242071197411004e-06,
      "loss": 0.0033,
      "step": 366470
    },
    {
      "epoch": 142.32,
      "learning_rate": 1.0236893203883496e-06,
      "loss": 0.0298,
      "step": 366480
    },
    {
      "epoch": 142.33,
      "learning_rate": 1.0231715210355988e-06,
      "loss": 0.0453,
      "step": 366490
    },
    {
      "epoch": 142.33,
      "learning_rate": 1.022653721682848e-06,
      "loss": 0.001,
      "step": 366500
    },
    {
      "epoch": 142.33,
      "learning_rate": 1.0221359223300972e-06,
      "loss": 0.0105,
      "step": 366510
    },
    {
      "epoch": 142.34,
      "learning_rate": 1.0216181229773464e-06,
      "loss": 0.0894,
      "step": 366520
    },
    {
      "epoch": 142.34,
      "learning_rate": 1.0211003236245956e-06,
      "loss": 0.0665,
      "step": 366530
    },
    {
      "epoch": 142.35,
      "learning_rate": 1.0205825242718448e-06,
      "loss": 0.0577,
      "step": 366540
    },
    {
      "epoch": 142.35,
      "learning_rate": 1.020064724919094e-06,
      "loss": 0.0298,
      "step": 366550
    },
    {
      "epoch": 142.35,
      "learning_rate": 1.0195469255663432e-06,
      "loss": 0.074,
      "step": 366560
    },
    {
      "epoch": 142.36,
      "learning_rate": 1.0190291262135924e-06,
      "loss": 0.0624,
      "step": 366570
    },
    {
      "epoch": 142.36,
      "learning_rate": 1.0185113268608416e-06,
      "loss": 0.0986,
      "step": 366580
    },
    {
      "epoch": 142.37,
      "learning_rate": 1.0179935275080908e-06,
      "loss": 0.0835,
      "step": 366590
    },
    {
      "epoch": 142.37,
      "learning_rate": 1.01747572815534e-06,
      "loss": 0.0009,
      "step": 366600
    },
    {
      "epoch": 142.37,
      "learning_rate": 1.0169579288025892e-06,
      "loss": 0.0005,
      "step": 366610
    },
    {
      "epoch": 142.38,
      "learning_rate": 1.0164401294498383e-06,
      "loss": 0.0117,
      "step": 366620
    },
    {
      "epoch": 142.38,
      "learning_rate": 1.0159223300970875e-06,
      "loss": 0.0545,
      "step": 366630
    },
    {
      "epoch": 142.38,
      "learning_rate": 1.0154045307443367e-06,
      "loss": 0.0269,
      "step": 366640
    },
    {
      "epoch": 142.39,
      "learning_rate": 1.014886731391586e-06,
      "loss": 0.0001,
      "step": 366650
    },
    {
      "epoch": 142.39,
      "learning_rate": 1.0143689320388351e-06,
      "loss": 0.0128,
      "step": 366660
    },
    {
      "epoch": 142.4,
      "learning_rate": 1.0138511326860841e-06,
      "loss": 0.1656,
      "step": 366670
    },
    {
      "epoch": 142.4,
      "learning_rate": 1.0133333333333333e-06,
      "loss": 0.0672,
      "step": 366680
    },
    {
      "epoch": 142.4,
      "learning_rate": 1.0128155339805825e-06,
      "loss": 0.0015,
      "step": 366690
    },
    {
      "epoch": 142.41,
      "learning_rate": 1.0122977346278317e-06,
      "loss": 0.0122,
      "step": 366700
    },
    {
      "epoch": 142.41,
      "learning_rate": 1.0117799352750809e-06,
      "loss": 0.0062,
      "step": 366710
    },
    {
      "epoch": 142.42,
      "learning_rate": 1.01126213592233e-06,
      "loss": 0.0842,
      "step": 366720
    },
    {
      "epoch": 142.42,
      "learning_rate": 1.0107443365695793e-06,
      "loss": 0.0174,
      "step": 366730
    },
    {
      "epoch": 142.42,
      "learning_rate": 1.0102265372168285e-06,
      "loss": 0.0467,
      "step": 366740
    },
    {
      "epoch": 142.43,
      "learning_rate": 1.0097087378640777e-06,
      "loss": 0.0075,
      "step": 366750
    },
    {
      "epoch": 142.43,
      "learning_rate": 1.0091909385113269e-06,
      "loss": 0.0778,
      "step": 366760
    },
    {
      "epoch": 142.43,
      "learning_rate": 1.0086731391585763e-06,
      "loss": 0.0014,
      "step": 366770
    },
    {
      "epoch": 142.44,
      "learning_rate": 1.0081553398058255e-06,
      "loss": 0.0282,
      "step": 366780
    },
    {
      "epoch": 142.44,
      "learning_rate": 1.0076375404530747e-06,
      "loss": 0.0003,
      "step": 366790
    },
    {
      "epoch": 142.45,
      "learning_rate": 1.0071197411003238e-06,
      "loss": 0.0128,
      "step": 366800
    },
    {
      "epoch": 142.45,
      "learning_rate": 1.006601941747573e-06,
      "loss": 0.023,
      "step": 366810
    },
    {
      "epoch": 142.45,
      "learning_rate": 1.0060841423948222e-06,
      "loss": 0.0309,
      "step": 366820
    },
    {
      "epoch": 142.46,
      "learning_rate": 1.0055663430420712e-06,
      "loss": 0.0145,
      "step": 366830
    },
    {
      "epoch": 142.46,
      "learning_rate": 1.0050485436893204e-06,
      "loss": 0.0007,
      "step": 366840
    },
    {
      "epoch": 142.47,
      "learning_rate": 1.0045307443365696e-06,
      "loss": 0.0855,
      "step": 366850
    },
    {
      "epoch": 142.47,
      "learning_rate": 1.0040129449838188e-06,
      "loss": 0.0113,
      "step": 366860
    },
    {
      "epoch": 142.47,
      "learning_rate": 1.003495145631068e-06,
      "loss": 0.0662,
      "step": 366870
    },
    {
      "epoch": 142.48,
      "learning_rate": 1.0029773462783172e-06,
      "loss": 0.0307,
      "step": 366880
    },
    {
      "epoch": 142.48,
      "learning_rate": 1.0024595469255664e-06,
      "loss": 0.1359,
      "step": 366890
    },
    {
      "epoch": 142.49,
      "learning_rate": 1.0019417475728156e-06,
      "loss": 0.0501,
      "step": 366900
    },
    {
      "epoch": 142.49,
      "learning_rate": 1.0014239482200648e-06,
      "loss": 0.0173,
      "step": 366910
    },
    {
      "epoch": 142.49,
      "learning_rate": 1.000906148867314e-06,
      "loss": 0.0087,
      "step": 366920
    },
    {
      "epoch": 142.5,
      "learning_rate": 1.0003883495145632e-06,
      "loss": 0.0012,
      "step": 366930
    },
    {
      "epoch": 142.5,
      "learning_rate": 9.998705501618124e-07,
      "loss": 0.0553,
      "step": 366940
    },
    {
      "epoch": 142.5,
      "learning_rate": 9.993527508090616e-07,
      "loss": 0.0084,
      "step": 366950
    },
    {
      "epoch": 142.51,
      "learning_rate": 9.988349514563107e-07,
      "loss": 0.0123,
      "step": 366960
    },
    {
      "epoch": 142.51,
      "learning_rate": 9.9831715210356e-07,
      "loss": 0.0332,
      "step": 366970
    },
    {
      "epoch": 142.52,
      "learning_rate": 9.977993527508091e-07,
      "loss": 0.0084,
      "step": 366980
    },
    {
      "epoch": 142.52,
      "learning_rate": 9.972815533980583e-07,
      "loss": 0.0608,
      "step": 366990
    },
    {
      "epoch": 142.52,
      "learning_rate": 9.967637540453075e-07,
      "loss": 0.0002,
      "step": 367000
    },
    {
      "epoch": 142.53,
      "learning_rate": 9.962459546925567e-07,
      "loss": 0.0003,
      "step": 367010
    },
    {
      "epoch": 142.53,
      "learning_rate": 9.95728155339806e-07,
      "loss": 0.0818,
      "step": 367020
    },
    {
      "epoch": 142.54,
      "learning_rate": 9.952103559870551e-07,
      "loss": 0.0895,
      "step": 367030
    },
    {
      "epoch": 142.54,
      "learning_rate": 9.946925566343043e-07,
      "loss": 0.0002,
      "step": 367040
    },
    {
      "epoch": 142.54,
      "learning_rate": 9.941747572815535e-07,
      "loss": 0.0227,
      "step": 367050
    },
    {
      "epoch": 142.55,
      "learning_rate": 9.936569579288027e-07,
      "loss": 0.02,
      "step": 367060
    },
    {
      "epoch": 142.55,
      "learning_rate": 9.931391585760519e-07,
      "loss": 0.0001,
      "step": 367070
    },
    {
      "epoch": 142.56,
      "learning_rate": 9.92621359223301e-07,
      "loss": 0.0366,
      "step": 367080
    },
    {
      "epoch": 142.56,
      "learning_rate": 9.921035598705503e-07,
      "loss": 0.1088,
      "step": 367090
    },
    {
      "epoch": 142.56,
      "learning_rate": 9.915857605177995e-07,
      "loss": 0.002,
      "step": 367100
    },
    {
      "epoch": 142.57,
      "learning_rate": 9.910679611650487e-07,
      "loss": 0.0356,
      "step": 367110
    },
    {
      "epoch": 142.57,
      "learning_rate": 9.905501618122979e-07,
      "loss": 0.0308,
      "step": 367120
    },
    {
      "epoch": 142.57,
      "learning_rate": 9.90032362459547e-07,
      "loss": 0.0006,
      "step": 367130
    },
    {
      "epoch": 142.58,
      "learning_rate": 9.895145631067962e-07,
      "loss": 0.0162,
      "step": 367140
    },
    {
      "epoch": 142.58,
      "learning_rate": 9.889967637540454e-07,
      "loss": 0.0548,
      "step": 367150
    },
    {
      "epoch": 142.59,
      "learning_rate": 9.884789644012946e-07,
      "loss": 0.0373,
      "step": 367160
    },
    {
      "epoch": 142.59,
      "learning_rate": 9.879611650485438e-07,
      "loss": 0.0232,
      "step": 367170
    },
    {
      "epoch": 142.59,
      "learning_rate": 9.87443365695793e-07,
      "loss": 0.0718,
      "step": 367180
    },
    {
      "epoch": 142.6,
      "learning_rate": 9.869255663430422e-07,
      "loss": 0.0853,
      "step": 367190
    },
    {
      "epoch": 142.6,
      "learning_rate": 9.864077669902914e-07,
      "loss": 0.0006,
      "step": 367200
    },
    {
      "epoch": 142.61,
      "learning_rate": 9.858899676375406e-07,
      "loss": 0.0982,
      "step": 367210
    },
    {
      "epoch": 142.61,
      "learning_rate": 9.853721682847898e-07,
      "loss": 0.0845,
      "step": 367220
    },
    {
      "epoch": 142.61,
      "learning_rate": 9.84854368932039e-07,
      "loss": 0.0214,
      "step": 367230
    },
    {
      "epoch": 142.62,
      "learning_rate": 9.84336569579288e-07,
      "loss": 0.0204,
      "step": 367240
    },
    {
      "epoch": 142.62,
      "learning_rate": 9.838187702265372e-07,
      "loss": 0.0074,
      "step": 367250
    },
    {
      "epoch": 142.63,
      "learning_rate": 9.833009708737864e-07,
      "loss": 0.0965,
      "step": 367260
    },
    {
      "epoch": 142.63,
      "learning_rate": 9.827831715210356e-07,
      "loss": 0.0411,
      "step": 367270
    },
    {
      "epoch": 142.63,
      "learning_rate": 9.822653721682848e-07,
      "loss": 0.0467,
      "step": 367280
    },
    {
      "epoch": 142.64,
      "learning_rate": 9.81747572815534e-07,
      "loss": 0.054,
      "step": 367290
    },
    {
      "epoch": 142.64,
      "learning_rate": 9.812297734627832e-07,
      "loss": 0.0795,
      "step": 367300
    },
    {
      "epoch": 142.64,
      "learning_rate": 9.807119741100323e-07,
      "loss": 0.0087,
      "step": 367310
    },
    {
      "epoch": 142.65,
      "learning_rate": 9.801941747572815e-07,
      "loss": 0.0215,
      "step": 367320
    },
    {
      "epoch": 142.65,
      "learning_rate": 9.79676375404531e-07,
      "loss": 0.0182,
      "step": 367330
    },
    {
      "epoch": 142.66,
      "learning_rate": 9.791585760517801e-07,
      "loss": 0.102,
      "step": 367340
    },
    {
      "epoch": 142.66,
      "learning_rate": 9.786407766990293e-07,
      "loss": 0.0095,
      "step": 367350
    },
    {
      "epoch": 142.66,
      "learning_rate": 9.781229773462785e-07,
      "loss": 0.0585,
      "step": 367360
    },
    {
      "epoch": 142.67,
      "learning_rate": 9.776051779935277e-07,
      "loss": 0.0411,
      "step": 367370
    },
    {
      "epoch": 142.67,
      "learning_rate": 9.77087378640777e-07,
      "loss": 0.0451,
      "step": 367380
    },
    {
      "epoch": 142.68,
      "learning_rate": 9.765695792880261e-07,
      "loss": 0.0198,
      "step": 367390
    },
    {
      "epoch": 142.68,
      "learning_rate": 9.76051779935275e-07,
      "loss": 0.0508,
      "step": 367400
    },
    {
      "epoch": 142.68,
      "learning_rate": 9.755339805825243e-07,
      "loss": 0.1014,
      "step": 367410
    },
    {
      "epoch": 142.69,
      "learning_rate": 9.750161812297735e-07,
      "loss": 0.0014,
      "step": 367420
    },
    {
      "epoch": 142.69,
      "learning_rate": 9.744983818770227e-07,
      "loss": 0.0006,
      "step": 367430
    },
    {
      "epoch": 142.7,
      "learning_rate": 9.739805825242719e-07,
      "loss": 0.0141,
      "step": 367440
    },
    {
      "epoch": 142.7,
      "learning_rate": 9.73462783171521e-07,
      "loss": 0.1458,
      "step": 367450
    },
    {
      "epoch": 142.7,
      "learning_rate": 9.729449838187703e-07,
      "loss": 0.0915,
      "step": 367460
    },
    {
      "epoch": 142.71,
      "learning_rate": 9.724271844660195e-07,
      "loss": 0.0604,
      "step": 367470
    },
    {
      "epoch": 142.71,
      "learning_rate": 9.719093851132687e-07,
      "loss": 0.0106,
      "step": 367480
    },
    {
      "epoch": 142.71,
      "learning_rate": 9.713915857605178e-07,
      "loss": 0.0247,
      "step": 367490
    },
    {
      "epoch": 142.72,
      "learning_rate": 9.70873786407767e-07,
      "loss": 0.0443,
      "step": 367500
    },
    {
      "epoch": 142.72,
      "learning_rate": 9.703559870550162e-07,
      "loss": 0.05,
      "step": 367510
    },
    {
      "epoch": 142.73,
      "learning_rate": 9.698381877022654e-07,
      "loss": 0.0994,
      "step": 367520
    },
    {
      "epoch": 142.73,
      "learning_rate": 9.693203883495146e-07,
      "loss": 0.0286,
      "step": 367530
    },
    {
      "epoch": 142.73,
      "learning_rate": 9.688025889967638e-07,
      "loss": 0.0123,
      "step": 367540
    },
    {
      "epoch": 142.74,
      "learning_rate": 9.68284789644013e-07,
      "loss": 0.0093,
      "step": 367550
    },
    {
      "epoch": 142.74,
      "learning_rate": 9.677669902912622e-07,
      "loss": 0.0095,
      "step": 367560
    },
    {
      "epoch": 142.75,
      "learning_rate": 9.672491909385114e-07,
      "loss": 0.0548,
      "step": 367570
    },
    {
      "epoch": 142.75,
      "learning_rate": 9.667313915857606e-07,
      "loss": 0.0957,
      "step": 367580
    },
    {
      "epoch": 142.75,
      "learning_rate": 9.662135922330098e-07,
      "loss": 0.0368,
      "step": 367590
    },
    {
      "epoch": 142.76,
      "learning_rate": 9.65695792880259e-07,
      "loss": 0.001,
      "step": 367600
    },
    {
      "epoch": 142.76,
      "learning_rate": 9.651779935275082e-07,
      "loss": 0.0719,
      "step": 367610
    },
    {
      "epoch": 142.77,
      "learning_rate": 9.646601941747574e-07,
      "loss": 0.087,
      "step": 367620
    },
    {
      "epoch": 142.77,
      "learning_rate": 9.641423948220066e-07,
      "loss": 0.0085,
      "step": 367630
    },
    {
      "epoch": 142.77,
      "learning_rate": 9.636245954692558e-07,
      "loss": 0.1278,
      "step": 367640
    },
    {
      "epoch": 142.78,
      "learning_rate": 9.63106796116505e-07,
      "loss": 0.0048,
      "step": 367650
    },
    {
      "epoch": 142.78,
      "learning_rate": 9.625889967637542e-07,
      "loss": 0.0007,
      "step": 367660
    },
    {
      "epoch": 142.78,
      "learning_rate": 9.620711974110033e-07,
      "loss": 0.0026,
      "step": 367670
    },
    {
      "epoch": 142.79,
      "learning_rate": 9.615533980582525e-07,
      "loss": 0.0743,
      "step": 367680
    },
    {
      "epoch": 142.79,
      "learning_rate": 9.610355987055017e-07,
      "loss": 0.0105,
      "step": 367690
    },
    {
      "epoch": 142.8,
      "learning_rate": 9.60517799352751e-07,
      "loss": 0.0147,
      "step": 367700
    },
    {
      "epoch": 142.8,
      "learning_rate": 9.600000000000001e-07,
      "loss": 0.0147,
      "step": 367710
    },
    {
      "epoch": 142.8,
      "learning_rate": 9.594822006472493e-07,
      "loss": 0.0436,
      "step": 367720
    },
    {
      "epoch": 142.81,
      "learning_rate": 9.589644012944985e-07,
      "loss": 0.0469,
      "step": 367730
    },
    {
      "epoch": 142.81,
      "learning_rate": 9.584466019417477e-07,
      "loss": 0.0713,
      "step": 367740
    },
    {
      "epoch": 142.82,
      "learning_rate": 9.57928802588997e-07,
      "loss": 0.0493,
      "step": 367750
    },
    {
      "epoch": 142.82,
      "learning_rate": 9.57411003236246e-07,
      "loss": 0.0259,
      "step": 367760
    },
    {
      "epoch": 142.82,
      "learning_rate": 9.568932038834953e-07,
      "loss": 0.0086,
      "step": 367770
    },
    {
      "epoch": 142.83,
      "learning_rate": 9.563754045307445e-07,
      "loss": 0.0773,
      "step": 367780
    },
    {
      "epoch": 142.83,
      "learning_rate": 9.558576051779937e-07,
      "loss": 0.0134,
      "step": 367790
    },
    {
      "epoch": 142.83,
      "learning_rate": 9.553398058252429e-07,
      "loss": 0.0002,
      "step": 367800
    },
    {
      "epoch": 142.84,
      "learning_rate": 9.548220064724919e-07,
      "loss": 0.0001,
      "step": 367810
    },
    {
      "epoch": 142.84,
      "learning_rate": 9.54304207119741e-07,
      "loss": 0.0278,
      "step": 367820
    },
    {
      "epoch": 142.85,
      "learning_rate": 9.537864077669902e-07,
      "loss": 0.0001,
      "step": 367830
    },
    {
      "epoch": 142.85,
      "learning_rate": 9.532686084142395e-07,
      "loss": 0.0419,
      "step": 367840
    },
    {
      "epoch": 142.85,
      "learning_rate": 9.527508090614887e-07,
      "loss": 0.0078,
      "step": 367850
    },
    {
      "epoch": 142.86,
      "learning_rate": 9.522330097087379e-07,
      "loss": 0.0161,
      "step": 367860
    },
    {
      "epoch": 142.86,
      "learning_rate": 9.51715210355987e-07,
      "loss": 0.0268,
      "step": 367870
    },
    {
      "epoch": 142.87,
      "learning_rate": 9.511974110032362e-07,
      "loss": 0.0172,
      "step": 367880
    },
    {
      "epoch": 142.87,
      "learning_rate": 9.506796116504854e-07,
      "loss": 0.0748,
      "step": 367890
    },
    {
      "epoch": 142.87,
      "learning_rate": 9.501618122977348e-07,
      "loss": 0.0152,
      "step": 367900
    },
    {
      "epoch": 142.88,
      "learning_rate": 9.496440129449839e-07,
      "loss": 0.035,
      "step": 367910
    },
    {
      "epoch": 142.88,
      "learning_rate": 9.491262135922331e-07,
      "loss": 0.0415,
      "step": 367920
    },
    {
      "epoch": 142.89,
      "learning_rate": 9.486084142394823e-07,
      "loss": 0.1204,
      "step": 367930
    },
    {
      "epoch": 142.89,
      "learning_rate": 9.480906148867315e-07,
      "loss": 0.0005,
      "step": 367940
    },
    {
      "epoch": 142.89,
      "learning_rate": 9.475728155339807e-07,
      "loss": 0.0537,
      "step": 367950
    },
    {
      "epoch": 142.9,
      "learning_rate": 9.470550161812299e-07,
      "loss": 0.0005,
      "step": 367960
    },
    {
      "epoch": 142.9,
      "learning_rate": 9.465372168284791e-07,
      "loss": 0.0136,
      "step": 367970
    },
    {
      "epoch": 142.9,
      "learning_rate": 9.460194174757283e-07,
      "loss": 0.0001,
      "step": 367980
    },
    {
      "epoch": 142.91,
      "learning_rate": 9.455016181229775e-07,
      "loss": 0.0565,
      "step": 367990
    },
    {
      "epoch": 142.91,
      "learning_rate": 9.449838187702267e-07,
      "loss": 0.0004,
      "step": 368000
    },
    {
      "epoch": 142.92,
      "learning_rate": 9.444660194174759e-07,
      "loss": 0.1031,
      "step": 368010
    },
    {
      "epoch": 142.92,
      "learning_rate": 9.43948220064725e-07,
      "loss": 0.0015,
      "step": 368020
    },
    {
      "epoch": 142.92,
      "learning_rate": 9.434304207119741e-07,
      "loss": 0.0002,
      "step": 368030
    },
    {
      "epoch": 142.93,
      "learning_rate": 9.429126213592233e-07,
      "loss": 0.0108,
      "step": 368040
    },
    {
      "epoch": 142.93,
      "learning_rate": 9.423948220064725e-07,
      "loss": 0.0089,
      "step": 368050
    },
    {
      "epoch": 142.94,
      "learning_rate": 9.418770226537217e-07,
      "loss": 0.0036,
      "step": 368060
    },
    {
      "epoch": 142.94,
      "learning_rate": 9.413592233009709e-07,
      "loss": 0.0098,
      "step": 368070
    },
    {
      "epoch": 142.94,
      "learning_rate": 9.408414239482201e-07,
      "loss": 0.0122,
      "step": 368080
    },
    {
      "epoch": 142.95,
      "learning_rate": 9.403236245954693e-07,
      "loss": 0.0003,
      "step": 368090
    },
    {
      "epoch": 142.95,
      "learning_rate": 9.398058252427185e-07,
      "loss": 0.0292,
      "step": 368100
    },
    {
      "epoch": 142.96,
      "learning_rate": 9.392880258899677e-07,
      "loss": 0.0333,
      "step": 368110
    },
    {
      "epoch": 142.96,
      "learning_rate": 9.387702265372169e-07,
      "loss": 0.0223,
      "step": 368120
    },
    {
      "epoch": 142.96,
      "learning_rate": 9.382524271844661e-07,
      "loss": 0.0264,
      "step": 368130
    },
    {
      "epoch": 142.97,
      "learning_rate": 9.377346278317153e-07,
      "loss": 0.1621,
      "step": 368140
    },
    {
      "epoch": 142.97,
      "learning_rate": 9.372168284789644e-07,
      "loss": 0.0086,
      "step": 368150
    },
    {
      "epoch": 142.97,
      "learning_rate": 9.366990291262136e-07,
      "loss": 0.0241,
      "step": 368160
    },
    {
      "epoch": 142.98,
      "learning_rate": 9.361812297734628e-07,
      "loss": 0.0832,
      "step": 368170
    },
    {
      "epoch": 142.98,
      "learning_rate": 9.356634304207122e-07,
      "loss": 0.0538,
      "step": 368180
    },
    {
      "epoch": 142.99,
      "learning_rate": 9.351456310679613e-07,
      "loss": 0.003,
      "step": 368190
    },
    {
      "epoch": 142.99,
      "learning_rate": 9.346278317152104e-07,
      "loss": 0.072,
      "step": 368200
    },
    {
      "epoch": 142.99,
      "learning_rate": 9.341100323624596e-07,
      "loss": 0.0783,
      "step": 368210
    },
    {
      "epoch": 143.0,
      "learning_rate": 9.335922330097088e-07,
      "loss": 0.0132,
      "step": 368220
    },
    {
      "epoch": 143.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.39228132367134094,
      "eval_runtime": 8.1822,
      "eval_samples_per_second": 444.256,
      "eval_steps_per_second": 55.608,
      "step": 368225
    },
    {
      "epoch": 143.0,
      "learning_rate": 9.33074433656958e-07,
      "loss": 0.0849,
      "step": 368230
    },
    {
      "epoch": 143.01,
      "learning_rate": 9.325566343042072e-07,
      "loss": 0.0022,
      "step": 368240
    },
    {
      "epoch": 143.01,
      "learning_rate": 9.320388349514564e-07,
      "loss": 0.178,
      "step": 368250
    },
    {
      "epoch": 143.01,
      "learning_rate": 9.315210355987056e-07,
      "loss": 0.0013,
      "step": 368260
    },
    {
      "epoch": 143.02,
      "learning_rate": 9.310032362459548e-07,
      "loss": 0.0001,
      "step": 368270
    },
    {
      "epoch": 143.02,
      "learning_rate": 9.30485436893204e-07,
      "loss": 0.0168,
      "step": 368280
    },
    {
      "epoch": 143.03,
      "learning_rate": 9.299676375404532e-07,
      "loss": 0.0599,
      "step": 368290
    },
    {
      "epoch": 143.03,
      "learning_rate": 9.294498381877024e-07,
      "loss": 0.0266,
      "step": 368300
    },
    {
      "epoch": 143.03,
      "learning_rate": 9.289320388349515e-07,
      "loss": 0.1643,
      "step": 368310
    },
    {
      "epoch": 143.04,
      "learning_rate": 9.284142394822007e-07,
      "loss": 0.0003,
      "step": 368320
    },
    {
      "epoch": 143.04,
      "learning_rate": 9.278964401294499e-07,
      "loss": 0.1141,
      "step": 368330
    },
    {
      "epoch": 143.04,
      "learning_rate": 9.273786407766991e-07,
      "loss": 0.0286,
      "step": 368340
    },
    {
      "epoch": 143.05,
      "learning_rate": 9.268608414239483e-07,
      "loss": 0.0178,
      "step": 368350
    },
    {
      "epoch": 143.05,
      "learning_rate": 9.263430420711975e-07,
      "loss": 0.0728,
      "step": 368360
    },
    {
      "epoch": 143.06,
      "learning_rate": 9.258252427184466e-07,
      "loss": 0.0916,
      "step": 368370
    },
    {
      "epoch": 143.06,
      "learning_rate": 9.253074433656958e-07,
      "loss": 0.1091,
      "step": 368380
    },
    {
      "epoch": 143.06,
      "learning_rate": 9.24789644012945e-07,
      "loss": 0.0403,
      "step": 368390
    },
    {
      "epoch": 143.07,
      "learning_rate": 9.242718446601942e-07,
      "loss": 0.002,
      "step": 368400
    },
    {
      "epoch": 143.07,
      "learning_rate": 9.237540453074434e-07,
      "loss": 0.0026,
      "step": 368410
    },
    {
      "epoch": 143.08,
      "learning_rate": 9.232362459546926e-07,
      "loss": 0.1103,
      "step": 368420
    },
    {
      "epoch": 143.08,
      "learning_rate": 9.227184466019418e-07,
      "loss": 0.0004,
      "step": 368430
    },
    {
      "epoch": 143.08,
      "learning_rate": 9.222006472491909e-07,
      "loss": 0.0607,
      "step": 368440
    },
    {
      "epoch": 143.09,
      "learning_rate": 9.216828478964401e-07,
      "loss": 0.0093,
      "step": 368450
    },
    {
      "epoch": 143.09,
      "learning_rate": 9.211650485436895e-07,
      "loss": 0.0062,
      "step": 368460
    },
    {
      "epoch": 143.1,
      "learning_rate": 9.206472491909387e-07,
      "loss": 0.0864,
      "step": 368470
    },
    {
      "epoch": 143.1,
      "learning_rate": 9.201294498381878e-07,
      "loss": 0.027,
      "step": 368480
    },
    {
      "epoch": 143.1,
      "learning_rate": 9.19611650485437e-07,
      "loss": 0.0233,
      "step": 368490
    },
    {
      "epoch": 143.11,
      "learning_rate": 9.190938511326862e-07,
      "loss": 0.0207,
      "step": 368500
    },
    {
      "epoch": 143.11,
      "learning_rate": 9.185760517799354e-07,
      "loss": 0.0005,
      "step": 368510
    },
    {
      "epoch": 143.11,
      "learning_rate": 9.180582524271846e-07,
      "loss": 0.0682,
      "step": 368520
    },
    {
      "epoch": 143.12,
      "learning_rate": 9.175404530744338e-07,
      "loss": 0.0152,
      "step": 368530
    },
    {
      "epoch": 143.12,
      "learning_rate": 9.17022653721683e-07,
      "loss": 0.0147,
      "step": 368540
    },
    {
      "epoch": 143.13,
      "learning_rate": 9.165048543689321e-07,
      "loss": 0.0567,
      "step": 368550
    },
    {
      "epoch": 143.13,
      "learning_rate": 9.159870550161813e-07,
      "loss": 0.0173,
      "step": 368560
    },
    {
      "epoch": 143.13,
      "learning_rate": 9.154692556634305e-07,
      "loss": 0.0207,
      "step": 368570
    },
    {
      "epoch": 143.14,
      "learning_rate": 9.149514563106797e-07,
      "loss": 0.0529,
      "step": 368580
    },
    {
      "epoch": 143.14,
      "learning_rate": 9.144336569579289e-07,
      "loss": 0.0013,
      "step": 368590
    },
    {
      "epoch": 143.15,
      "learning_rate": 9.13915857605178e-07,
      "loss": 0.0389,
      "step": 368600
    },
    {
      "epoch": 143.15,
      "learning_rate": 9.133980582524272e-07,
      "loss": 0.0107,
      "step": 368610
    },
    {
      "epoch": 143.15,
      "learning_rate": 9.128802588996764e-07,
      "loss": 0.0143,
      "step": 368620
    },
    {
      "epoch": 143.16,
      "learning_rate": 9.123624595469256e-07,
      "loss": 0.0114,
      "step": 368630
    },
    {
      "epoch": 143.16,
      "learning_rate": 9.118446601941748e-07,
      "loss": 0.0182,
      "step": 368640
    },
    {
      "epoch": 143.17,
      "learning_rate": 9.11326860841424e-07,
      "loss": 0.0256,
      "step": 368650
    },
    {
      "epoch": 143.17,
      "learning_rate": 9.108090614886732e-07,
      "loss": 0.0143,
      "step": 368660
    },
    {
      "epoch": 143.17,
      "learning_rate": 9.102912621359224e-07,
      "loss": 0.0516,
      "step": 368670
    },
    {
      "epoch": 143.18,
      "learning_rate": 9.097734627831716e-07,
      "loss": 0.0236,
      "step": 368680
    },
    {
      "epoch": 143.18,
      "learning_rate": 9.092556634304208e-07,
      "loss": 0.0005,
      "step": 368690
    },
    {
      "epoch": 143.18,
      "learning_rate": 9.0873786407767e-07,
      "loss": 0.0497,
      "step": 368700
    },
    {
      "epoch": 143.19,
      "learning_rate": 9.082200647249192e-07,
      "loss": 0.0105,
      "step": 368710
    },
    {
      "epoch": 143.19,
      "learning_rate": 9.077022653721682e-07,
      "loss": 0.0295,
      "step": 368720
    },
    {
      "epoch": 143.2,
      "learning_rate": 9.071844660194174e-07,
      "loss": 0.0003,
      "step": 368730
    },
    {
      "epoch": 143.2,
      "learning_rate": 9.066666666666668e-07,
      "loss": 0.086,
      "step": 368740
    },
    {
      "epoch": 143.2,
      "learning_rate": 9.06148867313916e-07,
      "loss": 0.0023,
      "step": 368750
    },
    {
      "epoch": 143.21,
      "learning_rate": 9.056310679611651e-07,
      "loss": 0.0717,
      "step": 368760
    },
    {
      "epoch": 143.21,
      "learning_rate": 9.051132686084143e-07,
      "loss": 0.047,
      "step": 368770
    },
    {
      "epoch": 143.22,
      "learning_rate": 9.045954692556635e-07,
      "loss": 0.0005,
      "step": 368780
    },
    {
      "epoch": 143.22,
      "learning_rate": 9.040776699029127e-07,
      "loss": 0.0248,
      "step": 368790
    },
    {
      "epoch": 143.22,
      "learning_rate": 9.035598705501619e-07,
      "loss": 0.065,
      "step": 368800
    },
    {
      "epoch": 143.23,
      "learning_rate": 9.030420711974111e-07,
      "loss": 0.1755,
      "step": 368810
    },
    {
      "epoch": 143.23,
      "learning_rate": 9.025242718446603e-07,
      "loss": 0.0223,
      "step": 368820
    },
    {
      "epoch": 143.23,
      "learning_rate": 9.020064724919095e-07,
      "loss": 0.0013,
      "step": 368830
    },
    {
      "epoch": 143.24,
      "learning_rate": 9.014886731391587e-07,
      "loss": 0.1073,
      "step": 368840
    },
    {
      "epoch": 143.24,
      "learning_rate": 9.009708737864079e-07,
      "loss": 0.0045,
      "step": 368850
    },
    {
      "epoch": 143.25,
      "learning_rate": 9.004530744336571e-07,
      "loss": 0.0001,
      "step": 368860
    },
    {
      "epoch": 143.25,
      "learning_rate": 8.999352750809063e-07,
      "loss": 0.0723,
      "step": 368870
    },
    {
      "epoch": 143.25,
      "learning_rate": 8.994174757281555e-07,
      "loss": 0.1596,
      "step": 368880
    },
    {
      "epoch": 143.26,
      "learning_rate": 8.988996763754046e-07,
      "loss": 0.0445,
      "step": 368890
    },
    {
      "epoch": 143.26,
      "learning_rate": 8.983818770226537e-07,
      "loss": 0.0091,
      "step": 368900
    },
    {
      "epoch": 143.27,
      "learning_rate": 8.978640776699029e-07,
      "loss": 0.1099,
      "step": 368910
    },
    {
      "epoch": 143.27,
      "learning_rate": 8.973462783171521e-07,
      "loss": 0.0507,
      "step": 368920
    },
    {
      "epoch": 143.27,
      "learning_rate": 8.968284789644013e-07,
      "loss": 0.0092,
      "step": 368930
    },
    {
      "epoch": 143.28,
      "learning_rate": 8.963106796116505e-07,
      "loss": 0.0203,
      "step": 368940
    },
    {
      "epoch": 143.28,
      "learning_rate": 8.957928802588997e-07,
      "loss": 0.0096,
      "step": 368950
    },
    {
      "epoch": 143.29,
      "learning_rate": 8.952750809061489e-07,
      "loss": 0.0134,
      "step": 368960
    },
    {
      "epoch": 143.29,
      "learning_rate": 8.947572815533981e-07,
      "loss": 0.007,
      "step": 368970
    },
    {
      "epoch": 143.29,
      "learning_rate": 8.942394822006473e-07,
      "loss": 0.0298,
      "step": 368980
    },
    {
      "epoch": 143.3,
      "learning_rate": 8.937216828478965e-07,
      "loss": 0.0149,
      "step": 368990
    },
    {
      "epoch": 143.3,
      "learning_rate": 8.932038834951457e-07,
      "loss": 0.0153,
      "step": 369000
    },
    {
      "epoch": 143.3,
      "learning_rate": 8.926860841423948e-07,
      "loss": 0.0003,
      "step": 369010
    },
    {
      "epoch": 143.31,
      "learning_rate": 8.921682847896442e-07,
      "loss": 0.0864,
      "step": 369020
    },
    {
      "epoch": 143.31,
      "learning_rate": 8.916504854368934e-07,
      "loss": 0.0187,
      "step": 369030
    },
    {
      "epoch": 143.32,
      "learning_rate": 8.911326860841426e-07,
      "loss": 0.0278,
      "step": 369040
    },
    {
      "epoch": 143.32,
      "learning_rate": 8.906148867313917e-07,
      "loss": 0.0538,
      "step": 369050
    },
    {
      "epoch": 143.32,
      "learning_rate": 8.900970873786409e-07,
      "loss": 0.0005,
      "step": 369060
    },
    {
      "epoch": 143.33,
      "learning_rate": 8.8957928802589e-07,
      "loss": 0.0243,
      "step": 369070
    },
    {
      "epoch": 143.33,
      "learning_rate": 8.890614886731392e-07,
      "loss": 0.0001,
      "step": 369080
    },
    {
      "epoch": 143.34,
      "learning_rate": 8.885436893203884e-07,
      "loss": 0.0001,
      "step": 369090
    },
    {
      "epoch": 143.34,
      "learning_rate": 8.880258899676376e-07,
      "loss": 0.1447,
      "step": 369100
    },
    {
      "epoch": 143.34,
      "learning_rate": 8.875080906148868e-07,
      "loss": 0.0851,
      "step": 369110
    },
    {
      "epoch": 143.35,
      "learning_rate": 8.86990291262136e-07,
      "loss": 0.0281,
      "step": 369120
    },
    {
      "epoch": 143.35,
      "learning_rate": 8.864724919093852e-07,
      "loss": 0.1553,
      "step": 369130
    },
    {
      "epoch": 143.36,
      "learning_rate": 8.859546925566344e-07,
      "loss": 0.0095,
      "step": 369140
    },
    {
      "epoch": 143.36,
      "learning_rate": 8.854368932038836e-07,
      "loss": 0.1404,
      "step": 369150
    },
    {
      "epoch": 143.36,
      "learning_rate": 8.849190938511328e-07,
      "loss": 0.0002,
      "step": 369160
    },
    {
      "epoch": 143.37,
      "learning_rate": 8.844012944983819e-07,
      "loss": 0.0407,
      "step": 369170
    },
    {
      "epoch": 143.37,
      "learning_rate": 8.838834951456311e-07,
      "loss": 0.0026,
      "step": 369180
    },
    {
      "epoch": 143.37,
      "learning_rate": 8.833656957928803e-07,
      "loss": 0.0156,
      "step": 369190
    },
    {
      "epoch": 143.38,
      "learning_rate": 8.828478964401295e-07,
      "loss": 0.0092,
      "step": 369200
    },
    {
      "epoch": 143.38,
      "learning_rate": 8.823300970873787e-07,
      "loss": 0.0738,
      "step": 369210
    },
    {
      "epoch": 143.39,
      "learning_rate": 8.818122977346279e-07,
      "loss": 0.0705,
      "step": 369220
    },
    {
      "epoch": 143.39,
      "learning_rate": 8.812944983818771e-07,
      "loss": 0.1272,
      "step": 369230
    },
    {
      "epoch": 143.39,
      "learning_rate": 8.807766990291263e-07,
      "loss": 0.0231,
      "step": 369240
    },
    {
      "epoch": 143.4,
      "learning_rate": 8.802588996763754e-07,
      "loss": 0.0089,
      "step": 369250
    },
    {
      "epoch": 143.4,
      "learning_rate": 8.797411003236246e-07,
      "loss": 0.0046,
      "step": 369260
    },
    {
      "epoch": 143.41,
      "learning_rate": 8.792233009708738e-07,
      "loss": 0.0022,
      "step": 369270
    },
    {
      "epoch": 143.41,
      "learning_rate": 8.78705501618123e-07,
      "loss": 0.0002,
      "step": 369280
    },
    {
      "epoch": 143.41,
      "learning_rate": 8.781877022653721e-07,
      "loss": 0.0148,
      "step": 369290
    },
    {
      "epoch": 143.42,
      "learning_rate": 8.776699029126215e-07,
      "loss": 0.0104,
      "step": 369300
    },
    {
      "epoch": 143.42,
      "learning_rate": 8.771521035598707e-07,
      "loss": 0.0305,
      "step": 369310
    },
    {
      "epoch": 143.43,
      "learning_rate": 8.766343042071199e-07,
      "loss": 0.0014,
      "step": 369320
    },
    {
      "epoch": 143.43,
      "learning_rate": 8.76116504854369e-07,
      "loss": 0.1198,
      "step": 369330
    },
    {
      "epoch": 143.43,
      "learning_rate": 8.755987055016182e-07,
      "loss": 0.1378,
      "step": 369340
    },
    {
      "epoch": 143.44,
      "learning_rate": 8.750809061488674e-07,
      "loss": 0.1703,
      "step": 369350
    },
    {
      "epoch": 143.44,
      "learning_rate": 8.745631067961166e-07,
      "loss": 0.0528,
      "step": 369360
    },
    {
      "epoch": 143.44,
      "learning_rate": 8.740453074433658e-07,
      "loss": 0.0385,
      "step": 369370
    },
    {
      "epoch": 143.45,
      "learning_rate": 8.73527508090615e-07,
      "loss": 0.0005,
      "step": 369380
    },
    {
      "epoch": 143.45,
      "learning_rate": 8.730097087378642e-07,
      "loss": 0.0519,
      "step": 369390
    },
    {
      "epoch": 143.46,
      "learning_rate": 8.724919093851134e-07,
      "loss": 0.0184,
      "step": 369400
    },
    {
      "epoch": 143.46,
      "learning_rate": 8.719741100323626e-07,
      "loss": 0.0143,
      "step": 369410
    },
    {
      "epoch": 143.46,
      "learning_rate": 8.714563106796118e-07,
      "loss": 0.0635,
      "step": 369420
    },
    {
      "epoch": 143.47,
      "learning_rate": 8.70938511326861e-07,
      "loss": 0.0433,
      "step": 369430
    },
    {
      "epoch": 143.47,
      "learning_rate": 8.704207119741101e-07,
      "loss": 0.1202,
      "step": 369440
    },
    {
      "epoch": 143.48,
      "learning_rate": 8.699029126213593e-07,
      "loss": 0.0187,
      "step": 369450
    },
    {
      "epoch": 143.48,
      "learning_rate": 8.693851132686084e-07,
      "loss": 0.1338,
      "step": 369460
    },
    {
      "epoch": 143.48,
      "learning_rate": 8.688673139158576e-07,
      "loss": 0.0538,
      "step": 369470
    },
    {
      "epoch": 143.49,
      "learning_rate": 8.683495145631068e-07,
      "loss": 0.0112,
      "step": 369480
    },
    {
      "epoch": 143.49,
      "learning_rate": 8.67831715210356e-07,
      "loss": 0.111,
      "step": 369490
    },
    {
      "epoch": 143.5,
      "learning_rate": 8.673139158576052e-07,
      "loss": 0.0798,
      "step": 369500
    },
    {
      "epoch": 143.5,
      "learning_rate": 8.667961165048544e-07,
      "loss": 0.0499,
      "step": 369510
    },
    {
      "epoch": 143.5,
      "learning_rate": 8.662783171521036e-07,
      "loss": 0.0091,
      "step": 369520
    },
    {
      "epoch": 143.51,
      "learning_rate": 8.657605177993528e-07,
      "loss": 0.0077,
      "step": 369530
    },
    {
      "epoch": 143.51,
      "learning_rate": 8.65242718446602e-07,
      "loss": 0.0228,
      "step": 369540
    },
    {
      "epoch": 143.51,
      "learning_rate": 8.647249190938512e-07,
      "loss": 0.0218,
      "step": 369550
    },
    {
      "epoch": 143.52,
      "learning_rate": 8.642071197411004e-07,
      "loss": 0.008,
      "step": 369560
    },
    {
      "epoch": 143.52,
      "learning_rate": 8.636893203883496e-07,
      "loss": 0.1075,
      "step": 369570
    },
    {
      "epoch": 143.53,
      "learning_rate": 8.631715210355989e-07,
      "loss": 0.0093,
      "step": 369580
    },
    {
      "epoch": 143.53,
      "learning_rate": 8.626537216828481e-07,
      "loss": 0.0534,
      "step": 369590
    },
    {
      "epoch": 143.53,
      "learning_rate": 8.621359223300973e-07,
      "loss": 0.0388,
      "step": 369600
    },
    {
      "epoch": 143.54,
      "learning_rate": 8.616181229773465e-07,
      "loss": 0.0138,
      "step": 369610
    },
    {
      "epoch": 143.54,
      "learning_rate": 8.611003236245955e-07,
      "loss": 0.0543,
      "step": 369620
    },
    {
      "epoch": 143.55,
      "learning_rate": 8.605825242718447e-07,
      "loss": 0.0093,
      "step": 369630
    },
    {
      "epoch": 143.55,
      "learning_rate": 8.600647249190939e-07,
      "loss": 0.212,
      "step": 369640
    },
    {
      "epoch": 143.55,
      "learning_rate": 8.595469255663431e-07,
      "loss": 0.0105,
      "step": 369650
    },
    {
      "epoch": 143.56,
      "learning_rate": 8.590291262135923e-07,
      "loss": 0.0217,
      "step": 369660
    },
    {
      "epoch": 143.56,
      "learning_rate": 8.585113268608415e-07,
      "loss": 0.0016,
      "step": 369670
    },
    {
      "epoch": 143.57,
      "learning_rate": 8.579935275080907e-07,
      "loss": 0.0184,
      "step": 369680
    },
    {
      "epoch": 143.57,
      "learning_rate": 8.574757281553399e-07,
      "loss": 0.0767,
      "step": 369690
    },
    {
      "epoch": 143.57,
      "learning_rate": 8.569579288025891e-07,
      "loss": 0.1033,
      "step": 369700
    },
    {
      "epoch": 143.58,
      "learning_rate": 8.564401294498383e-07,
      "loss": 0.0094,
      "step": 369710
    },
    {
      "epoch": 143.58,
      "learning_rate": 8.559223300970875e-07,
      "loss": 0.0077,
      "step": 369720
    },
    {
      "epoch": 143.58,
      "learning_rate": 8.554045307443367e-07,
      "loss": 0.0243,
      "step": 369730
    },
    {
      "epoch": 143.59,
      "learning_rate": 8.548867313915858e-07,
      "loss": 0.0395,
      "step": 369740
    },
    {
      "epoch": 143.59,
      "learning_rate": 8.54368932038835e-07,
      "loss": 0.0108,
      "step": 369750
    },
    {
      "epoch": 143.6,
      "learning_rate": 8.538511326860842e-07,
      "loss": 0.1098,
      "step": 369760
    },
    {
      "epoch": 143.6,
      "learning_rate": 8.533333333333334e-07,
      "loss": 0.0734,
      "step": 369770
    },
    {
      "epoch": 143.6,
      "learning_rate": 8.528155339805825e-07,
      "loss": 0.0558,
      "step": 369780
    },
    {
      "epoch": 143.61,
      "learning_rate": 8.522977346278317e-07,
      "loss": 0.0315,
      "step": 369790
    },
    {
      "epoch": 143.61,
      "learning_rate": 8.517799352750809e-07,
      "loss": 0.0233,
      "step": 369800
    },
    {
      "epoch": 143.62,
      "learning_rate": 8.512621359223301e-07,
      "loss": 0.1292,
      "step": 369810
    },
    {
      "epoch": 143.62,
      "learning_rate": 8.507443365695793e-07,
      "loss": 0.0864,
      "step": 369820
    },
    {
      "epoch": 143.62,
      "learning_rate": 8.502265372168285e-07,
      "loss": 0.0001,
      "step": 369830
    },
    {
      "epoch": 143.63,
      "learning_rate": 8.497087378640777e-07,
      "loss": 0.009,
      "step": 369840
    },
    {
      "epoch": 143.63,
      "learning_rate": 8.491909385113269e-07,
      "loss": 0.0556,
      "step": 369850
    },
    {
      "epoch": 143.63,
      "learning_rate": 8.486731391585762e-07,
      "loss": 0.0083,
      "step": 369860
    },
    {
      "epoch": 143.64,
      "learning_rate": 8.481553398058254e-07,
      "loss": 0.0607,
      "step": 369870
    },
    {
      "epoch": 143.64,
      "learning_rate": 8.476375404530746e-07,
      "loss": 0.0004,
      "step": 369880
    },
    {
      "epoch": 143.65,
      "learning_rate": 8.471197411003238e-07,
      "loss": 0.0308,
      "step": 369890
    },
    {
      "epoch": 143.65,
      "learning_rate": 8.466019417475729e-07,
      "loss": 0.0719,
      "step": 369900
    },
    {
      "epoch": 143.65,
      "learning_rate": 8.460841423948221e-07,
      "loss": 0.0469,
      "step": 369910
    },
    {
      "epoch": 143.66,
      "learning_rate": 8.455663430420713e-07,
      "loss": 0.0096,
      "step": 369920
    },
    {
      "epoch": 143.66,
      "learning_rate": 8.450485436893205e-07,
      "loss": 0.0179,
      "step": 369930
    },
    {
      "epoch": 143.67,
      "learning_rate": 8.445307443365697e-07,
      "loss": 0.0088,
      "step": 369940
    },
    {
      "epoch": 143.67,
      "learning_rate": 8.440129449838189e-07,
      "loss": 0.0539,
      "step": 369950
    },
    {
      "epoch": 143.67,
      "learning_rate": 8.43495145631068e-07,
      "loss": 0.0001,
      "step": 369960
    },
    {
      "epoch": 143.68,
      "learning_rate": 8.429773462783172e-07,
      "loss": 0.2129,
      "step": 369970
    },
    {
      "epoch": 143.68,
      "learning_rate": 8.424595469255664e-07,
      "loss": 0.0047,
      "step": 369980
    },
    {
      "epoch": 143.69,
      "learning_rate": 8.419417475728156e-07,
      "loss": 0.0446,
      "step": 369990
    },
    {
      "epoch": 143.69,
      "learning_rate": 8.414239482200648e-07,
      "loss": 0.0435,
      "step": 370000
    },
    {
      "epoch": 143.69,
      "learning_rate": 8.40906148867314e-07,
      "loss": 0.0106,
      "step": 370010
    },
    {
      "epoch": 143.7,
      "learning_rate": 8.403883495145632e-07,
      "loss": 0.0299,
      "step": 370020
    },
    {
      "epoch": 143.7,
      "learning_rate": 8.398705501618123e-07,
      "loss": 0.0102,
      "step": 370030
    },
    {
      "epoch": 143.7,
      "learning_rate": 8.393527508090615e-07,
      "loss": 0.0096,
      "step": 370040
    },
    {
      "epoch": 143.71,
      "learning_rate": 8.388349514563107e-07,
      "loss": 0.0382,
      "step": 370050
    },
    {
      "epoch": 143.71,
      "learning_rate": 8.383171521035599e-07,
      "loss": 0.0629,
      "step": 370060
    },
    {
      "epoch": 143.72,
      "learning_rate": 8.377993527508091e-07,
      "loss": 0.113,
      "step": 370070
    },
    {
      "epoch": 143.72,
      "learning_rate": 8.372815533980583e-07,
      "loss": 0.0396,
      "step": 370080
    },
    {
      "epoch": 143.72,
      "learning_rate": 8.367637540453075e-07,
      "loss": 0.0342,
      "step": 370090
    },
    {
      "epoch": 143.73,
      "learning_rate": 8.362459546925567e-07,
      "loss": 0.0001,
      "step": 370100
    },
    {
      "epoch": 143.73,
      "learning_rate": 8.357281553398059e-07,
      "loss": 0.0899,
      "step": 370110
    },
    {
      "epoch": 143.74,
      "learning_rate": 8.352103559870551e-07,
      "loss": 0.097,
      "step": 370120
    },
    {
      "epoch": 143.74,
      "learning_rate": 8.346925566343042e-07,
      "loss": 0.0564,
      "step": 370130
    },
    {
      "epoch": 143.74,
      "learning_rate": 8.341747572815535e-07,
      "loss": 0.0667,
      "step": 370140
    },
    {
      "epoch": 143.75,
      "learning_rate": 8.336569579288027e-07,
      "loss": 0.1567,
      "step": 370150
    },
    {
      "epoch": 143.75,
      "learning_rate": 8.331391585760519e-07,
      "loss": 0.0219,
      "step": 370160
    },
    {
      "epoch": 143.76,
      "learning_rate": 8.326213592233011e-07,
      "loss": 0.01,
      "step": 370170
    },
    {
      "epoch": 143.76,
      "learning_rate": 8.321035598705503e-07,
      "loss": 0.0214,
      "step": 370180
    },
    {
      "epoch": 143.76,
      "learning_rate": 8.315857605177994e-07,
      "loss": 0.0785,
      "step": 370190
    },
    {
      "epoch": 143.77,
      "learning_rate": 8.310679611650486e-07,
      "loss": 0.0001,
      "step": 370200
    },
    {
      "epoch": 143.77,
      "learning_rate": 8.305501618122978e-07,
      "loss": 0.0166,
      "step": 370210
    },
    {
      "epoch": 143.77,
      "learning_rate": 8.30032362459547e-07,
      "loss": 0.0001,
      "step": 370220
    },
    {
      "epoch": 143.78,
      "learning_rate": 8.295145631067962e-07,
      "loss": 0.0878,
      "step": 370230
    },
    {
      "epoch": 143.78,
      "learning_rate": 8.289967637540454e-07,
      "loss": 0.0007,
      "step": 370240
    },
    {
      "epoch": 143.79,
      "learning_rate": 8.284789644012946e-07,
      "loss": 0.0262,
      "step": 370250
    },
    {
      "epoch": 143.79,
      "learning_rate": 8.279611650485438e-07,
      "loss": 0.0563,
      "step": 370260
    },
    {
      "epoch": 143.79,
      "learning_rate": 8.27443365695793e-07,
      "loss": 0.0031,
      "step": 370270
    },
    {
      "epoch": 143.8,
      "learning_rate": 8.269255663430422e-07,
      "loss": 0.0275,
      "step": 370280
    },
    {
      "epoch": 143.8,
      "learning_rate": 8.264077669902914e-07,
      "loss": 0.0265,
      "step": 370290
    },
    {
      "epoch": 143.81,
      "learning_rate": 8.258899676375406e-07,
      "loss": 0.0433,
      "step": 370300
    },
    {
      "epoch": 143.81,
      "learning_rate": 8.253721682847896e-07,
      "loss": 0.0133,
      "step": 370310
    },
    {
      "epoch": 143.81,
      "learning_rate": 8.248543689320388e-07,
      "loss": 0.0002,
      "step": 370320
    },
    {
      "epoch": 143.82,
      "learning_rate": 8.24336569579288e-07,
      "loss": 0.0664,
      "step": 370330
    },
    {
      "epoch": 143.82,
      "learning_rate": 8.238187702265372e-07,
      "loss": 0.0587,
      "step": 370340
    },
    {
      "epoch": 143.83,
      "learning_rate": 8.233009708737864e-07,
      "loss": 0.0165,
      "step": 370350
    },
    {
      "epoch": 143.83,
      "learning_rate": 8.227831715210356e-07,
      "loss": 0.01,
      "step": 370360
    },
    {
      "epoch": 143.83,
      "learning_rate": 8.222653721682848e-07,
      "loss": 0.0267,
      "step": 370370
    },
    {
      "epoch": 143.84,
      "learning_rate": 8.21747572815534e-07,
      "loss": 0.161,
      "step": 370380
    },
    {
      "epoch": 143.84,
      "learning_rate": 8.212297734627832e-07,
      "loss": 0.0739,
      "step": 370390
    },
    {
      "epoch": 143.84,
      "learning_rate": 8.207119741100324e-07,
      "loss": 0.0547,
      "step": 370400
    },
    {
      "epoch": 143.85,
      "learning_rate": 8.201941747572816e-07,
      "loss": 0.0001,
      "step": 370410
    },
    {
      "epoch": 143.85,
      "learning_rate": 8.196763754045309e-07,
      "loss": 0.0207,
      "step": 370420
    },
    {
      "epoch": 143.86,
      "learning_rate": 8.191585760517801e-07,
      "loss": 0.0877,
      "step": 370430
    },
    {
      "epoch": 143.86,
      "learning_rate": 8.186407766990293e-07,
      "loss": 0.0413,
      "step": 370440
    },
    {
      "epoch": 143.86,
      "learning_rate": 8.181229773462785e-07,
      "loss": 0.0298,
      "step": 370450
    },
    {
      "epoch": 143.87,
      "learning_rate": 8.176051779935277e-07,
      "loss": 0.0458,
      "step": 370460
    },
    {
      "epoch": 143.87,
      "learning_rate": 8.170873786407768e-07,
      "loss": 0.0002,
      "step": 370470
    },
    {
      "epoch": 143.88,
      "learning_rate": 8.16569579288026e-07,
      "loss": 0.1012,
      "step": 370480
    },
    {
      "epoch": 143.88,
      "learning_rate": 8.160517799352751e-07,
      "loss": 0.051,
      "step": 370490
    },
    {
      "epoch": 143.88,
      "learning_rate": 8.155339805825243e-07,
      "loss": 0.0486,
      "step": 370500
    },
    {
      "epoch": 143.89,
      "learning_rate": 8.150161812297735e-07,
      "loss": 0.0707,
      "step": 370510
    },
    {
      "epoch": 143.89,
      "learning_rate": 8.144983818770227e-07,
      "loss": 0.0079,
      "step": 370520
    },
    {
      "epoch": 143.9,
      "learning_rate": 8.139805825242719e-07,
      "loss": 0.0002,
      "step": 370530
    },
    {
      "epoch": 143.9,
      "learning_rate": 8.134627831715211e-07,
      "loss": 0.0204,
      "step": 370540
    },
    {
      "epoch": 143.9,
      "learning_rate": 8.129449838187703e-07,
      "loss": 0.0269,
      "step": 370550
    },
    {
      "epoch": 143.91,
      "learning_rate": 8.124271844660195e-07,
      "loss": 0.0079,
      "step": 370560
    },
    {
      "epoch": 143.91,
      "learning_rate": 8.119093851132687e-07,
      "loss": 0.0113,
      "step": 370570
    },
    {
      "epoch": 143.91,
      "learning_rate": 8.113915857605179e-07,
      "loss": 0.0143,
      "step": 370580
    },
    {
      "epoch": 143.92,
      "learning_rate": 8.108737864077671e-07,
      "loss": 0.0244,
      "step": 370590
    },
    {
      "epoch": 143.92,
      "learning_rate": 8.103559870550162e-07,
      "loss": 0.0002,
      "step": 370600
    },
    {
      "epoch": 143.93,
      "learning_rate": 8.098381877022654e-07,
      "loss": 0.0001,
      "step": 370610
    },
    {
      "epoch": 143.93,
      "learning_rate": 8.093203883495146e-07,
      "loss": 0.0023,
      "step": 370620
    },
    {
      "epoch": 143.93,
      "learning_rate": 8.088025889967638e-07,
      "loss": 0.0351,
      "step": 370630
    },
    {
      "epoch": 143.94,
      "learning_rate": 8.08284789644013e-07,
      "loss": 0.1355,
      "step": 370640
    },
    {
      "epoch": 143.94,
      "learning_rate": 8.077669902912622e-07,
      "loss": 0.0001,
      "step": 370650
    },
    {
      "epoch": 143.95,
      "learning_rate": 8.072491909385113e-07,
      "loss": 0.076,
      "step": 370660
    },
    {
      "epoch": 143.95,
      "learning_rate": 8.067313915857605e-07,
      "loss": 0.09,
      "step": 370670
    },
    {
      "epoch": 143.95,
      "learning_rate": 8.062135922330097e-07,
      "loss": 0.081,
      "step": 370680
    },
    {
      "epoch": 143.96,
      "learning_rate": 8.056957928802589e-07,
      "loss": 0.0451,
      "step": 370690
    },
    {
      "epoch": 143.96,
      "learning_rate": 8.051779935275082e-07,
      "loss": 0.0001,
      "step": 370700
    },
    {
      "epoch": 143.97,
      "learning_rate": 8.046601941747574e-07,
      "loss": 0.053,
      "step": 370710
    },
    {
      "epoch": 143.97,
      "learning_rate": 8.041423948220066e-07,
      "loss": 0.0001,
      "step": 370720
    },
    {
      "epoch": 143.97,
      "learning_rate": 8.036245954692558e-07,
      "loss": 0.0539,
      "step": 370730
    },
    {
      "epoch": 143.98,
      "learning_rate": 8.03106796116505e-07,
      "loss": 0.066,
      "step": 370740
    },
    {
      "epoch": 143.98,
      "learning_rate": 8.025889967637542e-07,
      "loss": 0.0076,
      "step": 370750
    },
    {
      "epoch": 143.98,
      "learning_rate": 8.020711974110033e-07,
      "loss": 0.0049,
      "step": 370760
    },
    {
      "epoch": 143.99,
      "learning_rate": 8.015533980582525e-07,
      "loss": 0.031,
      "step": 370770
    },
    {
      "epoch": 143.99,
      "learning_rate": 8.010355987055017e-07,
      "loss": 0.0134,
      "step": 370780
    },
    {
      "epoch": 144.0,
      "learning_rate": 8.005177993527509e-07,
      "loss": 0.0094,
      "step": 370790
    },
    {
      "epoch": 144.0,
      "learning_rate": 8.000000000000001e-07,
      "loss": 0.0349,
      "step": 370800
    },
    {
      "epoch": 144.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.38799384236335754,
      "eval_runtime": 8.1817,
      "eval_samples_per_second": 444.282,
      "eval_steps_per_second": 55.612,
      "step": 370800
    },
    {
      "epoch": 144.0,
      "learning_rate": 7.994822006472493e-07,
      "loss": 0.0041,
      "step": 370810
    },
    {
      "epoch": 144.01,
      "learning_rate": 7.989644012944985e-07,
      "loss": 0.0408,
      "step": 370820
    },
    {
      "epoch": 144.01,
      "learning_rate": 7.984466019417477e-07,
      "loss": 0.0144,
      "step": 370830
    },
    {
      "epoch": 144.02,
      "learning_rate": 7.979288025889968e-07,
      "loss": 0.0184,
      "step": 370840
    },
    {
      "epoch": 144.02,
      "learning_rate": 7.97411003236246e-07,
      "loss": 0.0574,
      "step": 370850
    },
    {
      "epoch": 144.02,
      "learning_rate": 7.968932038834952e-07,
      "loss": 0.0096,
      "step": 370860
    },
    {
      "epoch": 144.03,
      "learning_rate": 7.963754045307444e-07,
      "loss": 0.0001,
      "step": 370870
    },
    {
      "epoch": 144.03,
      "learning_rate": 7.958576051779935e-07,
      "loss": 0.0095,
      "step": 370880
    },
    {
      "epoch": 144.03,
      "learning_rate": 7.953398058252427e-07,
      "loss": 0.0297,
      "step": 370890
    },
    {
      "epoch": 144.04,
      "learning_rate": 7.948220064724919e-07,
      "loss": 0.0091,
      "step": 370900
    },
    {
      "epoch": 144.04,
      "learning_rate": 7.943042071197411e-07,
      "loss": 0.0001,
      "step": 370910
    },
    {
      "epoch": 144.05,
      "learning_rate": 7.937864077669903e-07,
      "loss": 0.054,
      "step": 370920
    },
    {
      "epoch": 144.05,
      "learning_rate": 7.932686084142395e-07,
      "loss": 0.0666,
      "step": 370930
    },
    {
      "epoch": 144.05,
      "learning_rate": 7.927508090614887e-07,
      "loss": 0.0354,
      "step": 370940
    },
    {
      "epoch": 144.06,
      "learning_rate": 7.922330097087379e-07,
      "loss": 0.0402,
      "step": 370950
    },
    {
      "epoch": 144.06,
      "learning_rate": 7.917152103559871e-07,
      "loss": 0.0729,
      "step": 370960
    },
    {
      "epoch": 144.07,
      "learning_rate": 7.911974110032363e-07,
      "loss": 0.0603,
      "step": 370970
    },
    {
      "epoch": 144.07,
      "learning_rate": 7.906796116504856e-07,
      "loss": 0.0175,
      "step": 370980
    },
    {
      "epoch": 144.07,
      "learning_rate": 7.901618122977348e-07,
      "loss": 0.0002,
      "step": 370990
    },
    {
      "epoch": 144.08,
      "learning_rate": 7.89644012944984e-07,
      "loss": 0.0273,
      "step": 371000
    },
    {
      "epoch": 144.08,
      "learning_rate": 7.891262135922332e-07,
      "loss": 0.0005,
      "step": 371010
    },
    {
      "epoch": 144.09,
      "learning_rate": 7.886084142394823e-07,
      "loss": 0.124,
      "step": 371020
    },
    {
      "epoch": 144.09,
      "learning_rate": 7.880906148867315e-07,
      "loss": 0.1364,
      "step": 371030
    },
    {
      "epoch": 144.09,
      "learning_rate": 7.875728155339806e-07,
      "loss": 0.1187,
      "step": 371040
    },
    {
      "epoch": 144.1,
      "learning_rate": 7.870550161812298e-07,
      "loss": 0.0879,
      "step": 371050
    },
    {
      "epoch": 144.1,
      "learning_rate": 7.86537216828479e-07,
      "loss": 0.0673,
      "step": 371060
    },
    {
      "epoch": 144.1,
      "learning_rate": 7.860194174757282e-07,
      "loss": 0.015,
      "step": 371070
    },
    {
      "epoch": 144.11,
      "learning_rate": 7.855016181229774e-07,
      "loss": 0.0016,
      "step": 371080
    },
    {
      "epoch": 144.11,
      "learning_rate": 7.849838187702266e-07,
      "loss": 0.0378,
      "step": 371090
    },
    {
      "epoch": 144.12,
      "learning_rate": 7.844660194174758e-07,
      "loss": 0.0213,
      "step": 371100
    },
    {
      "epoch": 144.12,
      "learning_rate": 7.83948220064725e-07,
      "loss": 0.0286,
      "step": 371110
    },
    {
      "epoch": 144.12,
      "learning_rate": 7.834304207119742e-07,
      "loss": 0.0002,
      "step": 371120
    },
    {
      "epoch": 144.13,
      "learning_rate": 7.829126213592234e-07,
      "loss": 0.1241,
      "step": 371130
    },
    {
      "epoch": 144.13,
      "learning_rate": 7.823948220064726e-07,
      "loss": 0.0218,
      "step": 371140
    },
    {
      "epoch": 144.14,
      "learning_rate": 7.818770226537218e-07,
      "loss": 0.0247,
      "step": 371150
    },
    {
      "epoch": 144.14,
      "learning_rate": 7.81359223300971e-07,
      "loss": 0.002,
      "step": 371160
    },
    {
      "epoch": 144.14,
      "learning_rate": 7.808414239482201e-07,
      "loss": 0.1143,
      "step": 371170
    },
    {
      "epoch": 144.15,
      "learning_rate": 7.803236245954692e-07,
      "loss": 0.0568,
      "step": 371180
    },
    {
      "epoch": 144.15,
      "learning_rate": 7.798058252427184e-07,
      "loss": 0.0099,
      "step": 371190
    },
    {
      "epoch": 144.16,
      "learning_rate": 7.792880258899676e-07,
      "loss": 0.0192,
      "step": 371200
    },
    {
      "epoch": 144.16,
      "learning_rate": 7.787702265372168e-07,
      "loss": 0.0585,
      "step": 371210
    },
    {
      "epoch": 144.16,
      "learning_rate": 7.78252427184466e-07,
      "loss": 0.1407,
      "step": 371220
    },
    {
      "epoch": 144.17,
      "learning_rate": 7.777346278317152e-07,
      "loss": 0.0001,
      "step": 371230
    },
    {
      "epoch": 144.17,
      "learning_rate": 7.772168284789644e-07,
      "loss": 0.0788,
      "step": 371240
    },
    {
      "epoch": 144.17,
      "learning_rate": 7.766990291262136e-07,
      "loss": 0.0149,
      "step": 371250
    },
    {
      "epoch": 144.18,
      "learning_rate": 7.761812297734629e-07,
      "loss": 0.0176,
      "step": 371260
    },
    {
      "epoch": 144.18,
      "learning_rate": 7.756634304207121e-07,
      "loss": 0.0256,
      "step": 371270
    },
    {
      "epoch": 144.19,
      "learning_rate": 7.751456310679613e-07,
      "loss": 0.0002,
      "step": 371280
    },
    {
      "epoch": 144.19,
      "learning_rate": 7.746278317152105e-07,
      "loss": 0.0003,
      "step": 371290
    },
    {
      "epoch": 144.19,
      "learning_rate": 7.741100323624597e-07,
      "loss": 0.0184,
      "step": 371300
    },
    {
      "epoch": 144.2,
      "learning_rate": 7.735922330097089e-07,
      "loss": 0.0082,
      "step": 371310
    },
    {
      "epoch": 144.2,
      "learning_rate": 7.730744336569581e-07,
      "loss": 0.0699,
      "step": 371320
    },
    {
      "epoch": 144.21,
      "learning_rate": 7.725566343042072e-07,
      "loss": 0.036,
      "step": 371330
    },
    {
      "epoch": 144.21,
      "learning_rate": 7.720388349514564e-07,
      "loss": 0.1012,
      "step": 371340
    },
    {
      "epoch": 144.21,
      "learning_rate": 7.715210355987056e-07,
      "loss": 0.0097,
      "step": 371350
    },
    {
      "epoch": 144.22,
      "learning_rate": 7.710032362459548e-07,
      "loss": 0.0216,
      "step": 371360
    },
    {
      "epoch": 144.22,
      "learning_rate": 7.704854368932039e-07,
      "loss": 0.0453,
      "step": 371370
    },
    {
      "epoch": 144.23,
      "learning_rate": 7.699676375404531e-07,
      "loss": 0.0303,
      "step": 371380
    },
    {
      "epoch": 144.23,
      "learning_rate": 7.694498381877023e-07,
      "loss": 0.0217,
      "step": 371390
    },
    {
      "epoch": 144.23,
      "learning_rate": 7.689320388349515e-07,
      "loss": 0.0236,
      "step": 371400
    },
    {
      "epoch": 144.24,
      "learning_rate": 7.684142394822007e-07,
      "loss": 0.0082,
      "step": 371410
    },
    {
      "epoch": 144.24,
      "learning_rate": 7.678964401294499e-07,
      "loss": 0.0293,
      "step": 371420
    },
    {
      "epoch": 144.24,
      "learning_rate": 7.673786407766991e-07,
      "loss": 0.0133,
      "step": 371430
    },
    {
      "epoch": 144.25,
      "learning_rate": 7.668608414239483e-07,
      "loss": 0.0116,
      "step": 371440
    },
    {
      "epoch": 144.25,
      "learning_rate": 7.663430420711974e-07,
      "loss": 0.1064,
      "step": 371450
    },
    {
      "epoch": 144.26,
      "learning_rate": 7.658252427184466e-07,
      "loss": 0.0005,
      "step": 371460
    },
    {
      "epoch": 144.26,
      "learning_rate": 7.653074433656958e-07,
      "loss": 0.0077,
      "step": 371470
    },
    {
      "epoch": 144.26,
      "learning_rate": 7.64789644012945e-07,
      "loss": 0.0843,
      "step": 371480
    },
    {
      "epoch": 144.27,
      "learning_rate": 7.642718446601942e-07,
      "loss": 0.1686,
      "step": 371490
    },
    {
      "epoch": 144.27,
      "learning_rate": 7.637540453074434e-07,
      "loss": 0.0532,
      "step": 371500
    },
    {
      "epoch": 144.28,
      "learning_rate": 7.632362459546926e-07,
      "loss": 0.0528,
      "step": 371510
    },
    {
      "epoch": 144.28,
      "learning_rate": 7.627184466019418e-07,
      "loss": 0.0169,
      "step": 371520
    },
    {
      "epoch": 144.28,
      "learning_rate": 7.62200647249191e-07,
      "loss": 0.0696,
      "step": 371530
    },
    {
      "epoch": 144.29,
      "learning_rate": 7.616828478964403e-07,
      "loss": 0.0552,
      "step": 371540
    },
    {
      "epoch": 144.29,
      "learning_rate": 7.611650485436894e-07,
      "loss": 0.009,
      "step": 371550
    },
    {
      "epoch": 144.3,
      "learning_rate": 7.606472491909386e-07,
      "loss": 0.0011,
      "step": 371560
    },
    {
      "epoch": 144.3,
      "learning_rate": 7.601294498381878e-07,
      "loss": 0.0203,
      "step": 371570
    },
    {
      "epoch": 144.3,
      "learning_rate": 7.59611650485437e-07,
      "loss": 0.0163,
      "step": 371580
    },
    {
      "epoch": 144.31,
      "learning_rate": 7.590938511326862e-07,
      "loss": 0.0323,
      "step": 371590
    },
    {
      "epoch": 144.31,
      "learning_rate": 7.585760517799354e-07,
      "loss": 0.0742,
      "step": 371600
    },
    {
      "epoch": 144.31,
      "learning_rate": 7.580582524271845e-07,
      "loss": 0.0623,
      "step": 371610
    },
    {
      "epoch": 144.32,
      "learning_rate": 7.575404530744337e-07,
      "loss": 0.008,
      "step": 371620
    },
    {
      "epoch": 144.32,
      "learning_rate": 7.570226537216829e-07,
      "loss": 0.0578,
      "step": 371630
    },
    {
      "epoch": 144.33,
      "learning_rate": 7.565048543689321e-07,
      "loss": 0.0426,
      "step": 371640
    },
    {
      "epoch": 144.33,
      "learning_rate": 7.559870550161813e-07,
      "loss": 0.0091,
      "step": 371650
    },
    {
      "epoch": 144.33,
      "learning_rate": 7.554692556634305e-07,
      "loss": 0.0238,
      "step": 371660
    },
    {
      "epoch": 144.34,
      "learning_rate": 7.549514563106797e-07,
      "loss": 0.0077,
      "step": 371670
    },
    {
      "epoch": 144.34,
      "learning_rate": 7.544336569579289e-07,
      "loss": 0.011,
      "step": 371680
    },
    {
      "epoch": 144.35,
      "learning_rate": 7.539158576051781e-07,
      "loss": 0.0095,
      "step": 371690
    },
    {
      "epoch": 144.35,
      "learning_rate": 7.533980582524273e-07,
      "loss": 0.0106,
      "step": 371700
    },
    {
      "epoch": 144.35,
      "learning_rate": 7.528802588996765e-07,
      "loss": 0.0263,
      "step": 371710
    },
    {
      "epoch": 144.36,
      "learning_rate": 7.523624595469256e-07,
      "loss": 0.0085,
      "step": 371720
    },
    {
      "epoch": 144.36,
      "learning_rate": 7.518446601941748e-07,
      "loss": 0.0821,
      "step": 371730
    },
    {
      "epoch": 144.37,
      "learning_rate": 7.513268608414239e-07,
      "loss": 0.0046,
      "step": 371740
    },
    {
      "epoch": 144.37,
      "learning_rate": 7.508090614886731e-07,
      "loss": 0.0203,
      "step": 371750
    },
    {
      "epoch": 144.37,
      "learning_rate": 7.502912621359223e-07,
      "loss": 0.0141,
      "step": 371760
    },
    {
      "epoch": 144.38,
      "learning_rate": 7.497734627831715e-07,
      "loss": 0.0078,
      "step": 371770
    },
    {
      "epoch": 144.38,
      "learning_rate": 7.492556634304207e-07,
      "loss": 0.1114,
      "step": 371780
    },
    {
      "epoch": 144.38,
      "learning_rate": 7.487378640776699e-07,
      "loss": 0.0691,
      "step": 371790
    },
    {
      "epoch": 144.39,
      "learning_rate": 7.482200647249191e-07,
      "loss": 0.0721,
      "step": 371800
    },
    {
      "epoch": 144.39,
      "learning_rate": 7.477022653721683e-07,
      "loss": 0.0763,
      "step": 371810
    },
    {
      "epoch": 144.4,
      "learning_rate": 7.471844660194176e-07,
      "loss": 0.0001,
      "step": 371820
    },
    {
      "epoch": 144.4,
      "learning_rate": 7.466666666666668e-07,
      "loss": 0.0223,
      "step": 371830
    },
    {
      "epoch": 144.4,
      "learning_rate": 7.46148867313916e-07,
      "loss": 0.0406,
      "step": 371840
    },
    {
      "epoch": 144.41,
      "learning_rate": 7.456310679611652e-07,
      "loss": 0.0001,
      "step": 371850
    },
    {
      "epoch": 144.41,
      "learning_rate": 7.451132686084144e-07,
      "loss": 0.0325,
      "step": 371860
    },
    {
      "epoch": 144.42,
      "learning_rate": 7.445954692556636e-07,
      "loss": 0.1286,
      "step": 371870
    },
    {
      "epoch": 144.42,
      "learning_rate": 7.440776699029128e-07,
      "loss": 0.0006,
      "step": 371880
    },
    {
      "epoch": 144.42,
      "learning_rate": 7.43559870550162e-07,
      "loss": 0.0168,
      "step": 371890
    },
    {
      "epoch": 144.43,
      "learning_rate": 7.43042071197411e-07,
      "loss": 0.0247,
      "step": 371900
    },
    {
      "epoch": 144.43,
      "learning_rate": 7.425242718446602e-07,
      "loss": 0.0152,
      "step": 371910
    },
    {
      "epoch": 144.43,
      "learning_rate": 7.420064724919094e-07,
      "loss": 0.0985,
      "step": 371920
    },
    {
      "epoch": 144.44,
      "learning_rate": 7.414886731391586e-07,
      "loss": 0.0172,
      "step": 371930
    },
    {
      "epoch": 144.44,
      "learning_rate": 7.409708737864078e-07,
      "loss": 0.0752,
      "step": 371940
    },
    {
      "epoch": 144.45,
      "learning_rate": 7.40453074433657e-07,
      "loss": 0.0014,
      "step": 371950
    },
    {
      "epoch": 144.45,
      "learning_rate": 7.399352750809062e-07,
      "loss": 0.0893,
      "step": 371960
    },
    {
      "epoch": 144.45,
      "learning_rate": 7.394174757281554e-07,
      "loss": 0.0166,
      "step": 371970
    },
    {
      "epoch": 144.46,
      "learning_rate": 7.388996763754046e-07,
      "loss": 0.1515,
      "step": 371980
    },
    {
      "epoch": 144.46,
      "learning_rate": 7.383818770226538e-07,
      "loss": 0.0031,
      "step": 371990
    },
    {
      "epoch": 144.47,
      "learning_rate": 7.37864077669903e-07,
      "loss": 0.05,
      "step": 372000
    },
    {
      "epoch": 144.47,
      "learning_rate": 7.373462783171522e-07,
      "loss": 0.0795,
      "step": 372010
    },
    {
      "epoch": 144.47,
      "learning_rate": 7.368284789644013e-07,
      "loss": 0.0067,
      "step": 372020
    },
    {
      "epoch": 144.48,
      "learning_rate": 7.363106796116505e-07,
      "loss": 0.0203,
      "step": 372030
    },
    {
      "epoch": 144.48,
      "learning_rate": 7.357928802588997e-07,
      "loss": 0.0191,
      "step": 372040
    },
    {
      "epoch": 144.49,
      "learning_rate": 7.352750809061489e-07,
      "loss": 0.0004,
      "step": 372050
    },
    {
      "epoch": 144.49,
      "learning_rate": 7.34757281553398e-07,
      "loss": 0.0044,
      "step": 372060
    },
    {
      "epoch": 144.49,
      "learning_rate": 7.342394822006472e-07,
      "loss": 0.0876,
      "step": 372070
    },
    {
      "epoch": 144.5,
      "learning_rate": 7.337216828478964e-07,
      "loss": 0.0185,
      "step": 372080
    },
    {
      "epoch": 144.5,
      "learning_rate": 7.332038834951456e-07,
      "loss": 0.0236,
      "step": 372090
    },
    {
      "epoch": 144.5,
      "learning_rate": 7.326860841423949e-07,
      "loss": 0.0177,
      "step": 372100
    },
    {
      "epoch": 144.51,
      "learning_rate": 7.321682847896441e-07,
      "loss": 0.0558,
      "step": 372110
    },
    {
      "epoch": 144.51,
      "learning_rate": 7.316504854368933e-07,
      "loss": 0.0892,
      "step": 372120
    },
    {
      "epoch": 144.52,
      "learning_rate": 7.311326860841425e-07,
      "loss": 0.0194,
      "step": 372130
    },
    {
      "epoch": 144.52,
      "learning_rate": 7.306148867313917e-07,
      "loss": 0.0858,
      "step": 372140
    },
    {
      "epoch": 144.52,
      "learning_rate": 7.300970873786409e-07,
      "loss": 0.0463,
      "step": 372150
    },
    {
      "epoch": 144.53,
      "learning_rate": 7.295792880258901e-07,
      "loss": 0.1536,
      "step": 372160
    },
    {
      "epoch": 144.53,
      "learning_rate": 7.290614886731393e-07,
      "loss": 0.016,
      "step": 372170
    },
    {
      "epoch": 144.54,
      "learning_rate": 7.285436893203884e-07,
      "loss": 0.0091,
      "step": 372180
    },
    {
      "epoch": 144.54,
      "learning_rate": 7.280258899676376e-07,
      "loss": 0.0296,
      "step": 372190
    },
    {
      "epoch": 144.54,
      "learning_rate": 7.275080906148868e-07,
      "loss": 0.0079,
      "step": 372200
    },
    {
      "epoch": 144.55,
      "learning_rate": 7.26990291262136e-07,
      "loss": 0.0075,
      "step": 372210
    },
    {
      "epoch": 144.55,
      "learning_rate": 7.264724919093852e-07,
      "loss": 0.0001,
      "step": 372220
    },
    {
      "epoch": 144.56,
      "learning_rate": 7.259546925566344e-07,
      "loss": 0.0075,
      "step": 372230
    },
    {
      "epoch": 144.56,
      "learning_rate": 7.254368932038836e-07,
      "loss": 0.1023,
      "step": 372240
    },
    {
      "epoch": 144.56,
      "learning_rate": 7.249190938511327e-07,
      "loss": 0.0851,
      "step": 372250
    },
    {
      "epoch": 144.57,
      "learning_rate": 7.244012944983819e-07,
      "loss": 0.0407,
      "step": 372260
    },
    {
      "epoch": 144.57,
      "learning_rate": 7.238834951456311e-07,
      "loss": 0.0253,
      "step": 372270
    },
    {
      "epoch": 144.57,
      "learning_rate": 7.233656957928803e-07,
      "loss": 0.0301,
      "step": 372280
    },
    {
      "epoch": 144.58,
      "learning_rate": 7.228478964401295e-07,
      "loss": 0.1527,
      "step": 372290
    },
    {
      "epoch": 144.58,
      "learning_rate": 7.223300970873787e-07,
      "loss": 0.1019,
      "step": 372300
    },
    {
      "epoch": 144.59,
      "learning_rate": 7.218122977346278e-07,
      "loss": 0.0135,
      "step": 372310
    },
    {
      "epoch": 144.59,
      "learning_rate": 7.21294498381877e-07,
      "loss": 0.0017,
      "step": 372320
    },
    {
      "epoch": 144.59,
      "learning_rate": 7.207766990291262e-07,
      "loss": 0.0633,
      "step": 372330
    },
    {
      "epoch": 144.6,
      "learning_rate": 7.202588996763754e-07,
      "loss": 0.0185,
      "step": 372340
    },
    {
      "epoch": 144.6,
      "learning_rate": 7.197411003236246e-07,
      "loss": 0.0115,
      "step": 372350
    },
    {
      "epoch": 144.61,
      "learning_rate": 7.192233009708738e-07,
      "loss": 0.0176,
      "step": 372360
    },
    {
      "epoch": 144.61,
      "learning_rate": 7.18705501618123e-07,
      "loss": 0.0169,
      "step": 372370
    },
    {
      "epoch": 144.61,
      "learning_rate": 7.181877022653723e-07,
      "loss": 0.0106,
      "step": 372380
    },
    {
      "epoch": 144.62,
      "learning_rate": 7.176699029126215e-07,
      "loss": 0.103,
      "step": 372390
    },
    {
      "epoch": 144.62,
      "learning_rate": 7.171521035598707e-07,
      "loss": 0.1482,
      "step": 372400
    },
    {
      "epoch": 144.63,
      "learning_rate": 7.166343042071199e-07,
      "loss": 0.0683,
      "step": 372410
    },
    {
      "epoch": 144.63,
      "learning_rate": 7.161165048543691e-07,
      "loss": 0.0198,
      "step": 372420
    },
    {
      "epoch": 144.63,
      "learning_rate": 7.155987055016182e-07,
      "loss": 0.0004,
      "step": 372430
    },
    {
      "epoch": 144.64,
      "learning_rate": 7.150809061488674e-07,
      "loss": 0.0096,
      "step": 372440
    },
    {
      "epoch": 144.64,
      "learning_rate": 7.145631067961166e-07,
      "loss": 0.0017,
      "step": 372450
    },
    {
      "epoch": 144.64,
      "learning_rate": 7.140453074433658e-07,
      "loss": 0.0164,
      "step": 372460
    },
    {
      "epoch": 144.65,
      "learning_rate": 7.135275080906149e-07,
      "loss": 0.0719,
      "step": 372470
    },
    {
      "epoch": 144.65,
      "learning_rate": 7.130097087378641e-07,
      "loss": 0.0029,
      "step": 372480
    },
    {
      "epoch": 144.66,
      "learning_rate": 7.124919093851133e-07,
      "loss": 0.0149,
      "step": 372490
    },
    {
      "epoch": 144.66,
      "learning_rate": 7.119741100323625e-07,
      "loss": 0.0087,
      "step": 372500
    },
    {
      "epoch": 144.66,
      "learning_rate": 7.114563106796117e-07,
      "loss": 0.0193,
      "step": 372510
    },
    {
      "epoch": 144.67,
      "learning_rate": 7.109385113268609e-07,
      "loss": 0.0091,
      "step": 372520
    },
    {
      "epoch": 144.67,
      "learning_rate": 7.104207119741101e-07,
      "loss": 0.0012,
      "step": 372530
    },
    {
      "epoch": 144.68,
      "learning_rate": 7.099029126213593e-07,
      "loss": 0.0299,
      "step": 372540
    },
    {
      "epoch": 144.68,
      "learning_rate": 7.093851132686085e-07,
      "loss": 0.0246,
      "step": 372550
    },
    {
      "epoch": 144.68,
      "learning_rate": 7.088673139158577e-07,
      "loss": 0.0182,
      "step": 372560
    },
    {
      "epoch": 144.69,
      "learning_rate": 7.083495145631069e-07,
      "loss": 0.0322,
      "step": 372570
    },
    {
      "epoch": 144.69,
      "learning_rate": 7.078317152103561e-07,
      "loss": 0.0106,
      "step": 372580
    },
    {
      "epoch": 144.7,
      "learning_rate": 7.073139158576051e-07,
      "loss": 0.0562,
      "step": 372590
    },
    {
      "epoch": 144.7,
      "learning_rate": 7.067961165048543e-07,
      "loss": 0.1645,
      "step": 372600
    },
    {
      "epoch": 144.7,
      "learning_rate": 7.062783171521035e-07,
      "loss": 0.0887,
      "step": 372610
    },
    {
      "epoch": 144.71,
      "learning_rate": 7.057605177993527e-07,
      "loss": 0.0281,
      "step": 372620
    },
    {
      "epoch": 144.71,
      "learning_rate": 7.052427184466019e-07,
      "loss": 0.0299,
      "step": 372630
    },
    {
      "epoch": 144.71,
      "learning_rate": 7.047249190938511e-07,
      "loss": 0.1213,
      "step": 372640
    },
    {
      "epoch": 144.72,
      "learning_rate": 7.042071197411003e-07,
      "loss": 0.0381,
      "step": 372650
    },
    {
      "epoch": 144.72,
      "learning_rate": 7.036893203883496e-07,
      "loss": 0.072,
      "step": 372660
    },
    {
      "epoch": 144.73,
      "learning_rate": 7.031715210355988e-07,
      "loss": 0.1089,
      "step": 372670
    },
    {
      "epoch": 144.73,
      "learning_rate": 7.02653721682848e-07,
      "loss": 0.0513,
      "step": 372680
    },
    {
      "epoch": 144.73,
      "learning_rate": 7.021359223300972e-07,
      "loss": 0.0094,
      "step": 372690
    },
    {
      "epoch": 144.74,
      "learning_rate": 7.016181229773464e-07,
      "loss": 0.0434,
      "step": 372700
    },
    {
      "epoch": 144.74,
      "learning_rate": 7.011003236245956e-07,
      "loss": 0.1929,
      "step": 372710
    },
    {
      "epoch": 144.75,
      "learning_rate": 7.005825242718448e-07,
      "loss": 0.0001,
      "step": 372720
    },
    {
      "epoch": 144.75,
      "learning_rate": 7.00064724919094e-07,
      "loss": 0.0534,
      "step": 372730
    },
    {
      "epoch": 144.75,
      "learning_rate": 6.995469255663432e-07,
      "loss": 0.0132,
      "step": 372740
    },
    {
      "epoch": 144.76,
      "learning_rate": 6.990291262135923e-07,
      "loss": 0.0222,
      "step": 372750
    },
    {
      "epoch": 144.76,
      "learning_rate": 6.985113268608415e-07,
      "loss": 0.0002,
      "step": 372760
    },
    {
      "epoch": 144.77,
      "learning_rate": 6.979935275080907e-07,
      "loss": 0.1041,
      "step": 372770
    },
    {
      "epoch": 144.77,
      "learning_rate": 6.974757281553398e-07,
      "loss": 0.0334,
      "step": 372780
    },
    {
      "epoch": 144.77,
      "learning_rate": 6.96957928802589e-07,
      "loss": 0.1015,
      "step": 372790
    },
    {
      "epoch": 144.78,
      "learning_rate": 6.964401294498382e-07,
      "loss": 0.0002,
      "step": 372800
    },
    {
      "epoch": 144.78,
      "learning_rate": 6.959223300970874e-07,
      "loss": 0.0466,
      "step": 372810
    },
    {
      "epoch": 144.78,
      "learning_rate": 6.954045307443366e-07,
      "loss": 0.0547,
      "step": 372820
    },
    {
      "epoch": 144.79,
      "learning_rate": 6.948867313915858e-07,
      "loss": 0.0457,
      "step": 372830
    },
    {
      "epoch": 144.79,
      "learning_rate": 6.94368932038835e-07,
      "loss": 0.0903,
      "step": 372840
    },
    {
      "epoch": 144.8,
      "learning_rate": 6.938511326860842e-07,
      "loss": 0.0627,
      "step": 372850
    },
    {
      "epoch": 144.8,
      "learning_rate": 6.933333333333334e-07,
      "loss": 0.0544,
      "step": 372860
    },
    {
      "epoch": 144.8,
      "learning_rate": 6.928155339805826e-07,
      "loss": 0.0032,
      "step": 372870
    },
    {
      "epoch": 144.81,
      "learning_rate": 6.922977346278317e-07,
      "loss": 0.0001,
      "step": 372880
    },
    {
      "epoch": 144.81,
      "learning_rate": 6.917799352750809e-07,
      "loss": 0.0333,
      "step": 372890
    },
    {
      "epoch": 144.82,
      "learning_rate": 6.912621359223301e-07,
      "loss": 0.0916,
      "step": 372900
    },
    {
      "epoch": 144.82,
      "learning_rate": 6.907443365695793e-07,
      "loss": 0.0563,
      "step": 372910
    },
    {
      "epoch": 144.82,
      "learning_rate": 6.902265372168285e-07,
      "loss": 0.0567,
      "step": 372920
    },
    {
      "epoch": 144.83,
      "learning_rate": 6.897087378640777e-07,
      "loss": 0.0971,
      "step": 372930
    },
    {
      "epoch": 144.83,
      "learning_rate": 6.89190938511327e-07,
      "loss": 0.0342,
      "step": 372940
    },
    {
      "epoch": 144.83,
      "learning_rate": 6.886731391585762e-07,
      "loss": 0.0456,
      "step": 372950
    },
    {
      "epoch": 144.84,
      "learning_rate": 6.881553398058253e-07,
      "loss": 0.0187,
      "step": 372960
    },
    {
      "epoch": 144.84,
      "learning_rate": 6.876375404530745e-07,
      "loss": 0.0704,
      "step": 372970
    },
    {
      "epoch": 144.85,
      "learning_rate": 6.871197411003237e-07,
      "loss": 0.0108,
      "step": 372980
    },
    {
      "epoch": 144.85,
      "learning_rate": 6.866019417475729e-07,
      "loss": 0.0083,
      "step": 372990
    },
    {
      "epoch": 144.85,
      "learning_rate": 6.860841423948221e-07,
      "loss": 0.0341,
      "step": 373000
    },
    {
      "epoch": 144.86,
      "learning_rate": 6.855663430420713e-07,
      "loss": 0.0295,
      "step": 373010
    },
    {
      "epoch": 144.86,
      "learning_rate": 6.850485436893205e-07,
      "loss": 0.0107,
      "step": 373020
    },
    {
      "epoch": 144.87,
      "learning_rate": 6.845307443365697e-07,
      "loss": 0.0502,
      "step": 373030
    },
    {
      "epoch": 144.87,
      "learning_rate": 6.840129449838188e-07,
      "loss": 0.0004,
      "step": 373040
    },
    {
      "epoch": 144.87,
      "learning_rate": 6.83495145631068e-07,
      "loss": 0.0169,
      "step": 373050
    },
    {
      "epoch": 144.88,
      "learning_rate": 6.829773462783172e-07,
      "loss": 0.0465,
      "step": 373060
    },
    {
      "epoch": 144.88,
      "learning_rate": 6.824595469255664e-07,
      "loss": 0.0095,
      "step": 373070
    },
    {
      "epoch": 144.89,
      "learning_rate": 6.819417475728156e-07,
      "loss": 0.0006,
      "step": 373080
    },
    {
      "epoch": 144.89,
      "learning_rate": 6.814239482200648e-07,
      "loss": 0.0293,
      "step": 373090
    },
    {
      "epoch": 144.89,
      "learning_rate": 6.80906148867314e-07,
      "loss": 0.0177,
      "step": 373100
    },
    {
      "epoch": 144.9,
      "learning_rate": 6.803883495145632e-07,
      "loss": 0.009,
      "step": 373110
    },
    {
      "epoch": 144.9,
      "learning_rate": 6.798705501618124e-07,
      "loss": 0.0256,
      "step": 373120
    },
    {
      "epoch": 144.9,
      "learning_rate": 6.793527508090615e-07,
      "loss": 0.0015,
      "step": 373130
    },
    {
      "epoch": 144.91,
      "learning_rate": 6.788349514563107e-07,
      "loss": 0.0004,
      "step": 373140
    },
    {
      "epoch": 144.91,
      "learning_rate": 6.783171521035599e-07,
      "loss": 0.0001,
      "step": 373150
    },
    {
      "epoch": 144.92,
      "learning_rate": 6.77799352750809e-07,
      "loss": 0.0001,
      "step": 373160
    },
    {
      "epoch": 144.92,
      "learning_rate": 6.772815533980582e-07,
      "loss": 0.0689,
      "step": 373170
    },
    {
      "epoch": 144.92,
      "learning_rate": 6.767637540453074e-07,
      "loss": 0.1632,
      "step": 373180
    },
    {
      "epoch": 144.93,
      "learning_rate": 6.762459546925566e-07,
      "loss": 0.0231,
      "step": 373190
    },
    {
      "epoch": 144.93,
      "learning_rate": 6.757281553398058e-07,
      "loss": 0.0461,
      "step": 373200
    },
    {
      "epoch": 144.94,
      "learning_rate": 6.75210355987055e-07,
      "loss": 0.0329,
      "step": 373210
    },
    {
      "epoch": 144.94,
      "learning_rate": 6.746925566343043e-07,
      "loss": 0.0545,
      "step": 373220
    },
    {
      "epoch": 144.94,
      "learning_rate": 6.741747572815535e-07,
      "loss": 0.0377,
      "step": 373230
    },
    {
      "epoch": 144.95,
      "learning_rate": 6.736569579288027e-07,
      "loss": 0.0679,
      "step": 373240
    },
    {
      "epoch": 144.95,
      "learning_rate": 6.731391585760519e-07,
      "loss": 0.0012,
      "step": 373250
    },
    {
      "epoch": 144.96,
      "learning_rate": 6.726213592233011e-07,
      "loss": 0.001,
      "step": 373260
    },
    {
      "epoch": 144.96,
      "learning_rate": 6.721035598705503e-07,
      "loss": 0.0119,
      "step": 373270
    },
    {
      "epoch": 144.96,
      "learning_rate": 6.715857605177995e-07,
      "loss": 0.0305,
      "step": 373280
    },
    {
      "epoch": 144.97,
      "learning_rate": 6.710679611650487e-07,
      "loss": 0.1069,
      "step": 373290
    },
    {
      "epoch": 144.97,
      "learning_rate": 6.705501618122979e-07,
      "loss": 0.0665,
      "step": 373300
    },
    {
      "epoch": 144.97,
      "learning_rate": 6.70032362459547e-07,
      "loss": 0.0399,
      "step": 373310
    },
    {
      "epoch": 144.98,
      "learning_rate": 6.695145631067961e-07,
      "loss": 0.0485,
      "step": 373320
    },
    {
      "epoch": 144.98,
      "learning_rate": 6.689967637540453e-07,
      "loss": 0.009,
      "step": 373330
    },
    {
      "epoch": 144.99,
      "learning_rate": 6.684789644012945e-07,
      "loss": 0.0336,
      "step": 373340
    },
    {
      "epoch": 144.99,
      "learning_rate": 6.679611650485437e-07,
      "loss": 0.0221,
      "step": 373350
    },
    {
      "epoch": 144.99,
      "learning_rate": 6.674433656957929e-07,
      "loss": 0.0612,
      "step": 373360
    },
    {
      "epoch": 145.0,
      "learning_rate": 6.669255663430421e-07,
      "loss": 0.0014,
      "step": 373370
    },
    {
      "epoch": 145.0,
      "eval_accuracy": 0.9535075653370014,
      "eval_loss": 0.3874504864215851,
      "eval_runtime": 8.208,
      "eval_samples_per_second": 442.86,
      "eval_steps_per_second": 55.434,
      "step": 373375
    },
    {
      "epoch": 145.0,
      "learning_rate": 6.664077669902913e-07,
      "loss": 0.0255,
      "step": 373380
    },
    {
      "epoch": 145.01,
      "learning_rate": 6.658899676375405e-07,
      "loss": 0.0592,
      "step": 373390
    },
    {
      "epoch": 145.01,
      "learning_rate": 6.653721682847897e-07,
      "loss": 0.009,
      "step": 373400
    },
    {
      "epoch": 145.01,
      "learning_rate": 6.648543689320389e-07,
      "loss": 0.0057,
      "step": 373410
    },
    {
      "epoch": 145.02,
      "learning_rate": 6.643365695792881e-07,
      "loss": 0.0255,
      "step": 373420
    },
    {
      "epoch": 145.02,
      "learning_rate": 6.638187702265373e-07,
      "loss": 0.1079,
      "step": 373430
    },
    {
      "epoch": 145.03,
      "learning_rate": 6.633009708737865e-07,
      "loss": 0.0414,
      "step": 373440
    },
    {
      "epoch": 145.03,
      "learning_rate": 6.627831715210356e-07,
      "loss": 0.0142,
      "step": 373450
    },
    {
      "epoch": 145.03,
      "learning_rate": 6.622653721682848e-07,
      "loss": 0.0777,
      "step": 373460
    },
    {
      "epoch": 145.04,
      "learning_rate": 6.61747572815534e-07,
      "loss": 0.0102,
      "step": 373470
    },
    {
      "epoch": 145.04,
      "learning_rate": 6.612297734627831e-07,
      "loss": 0.0446,
      "step": 373480
    },
    {
      "epoch": 145.04,
      "learning_rate": 6.607119741100323e-07,
      "loss": 0.0009,
      "step": 373490
    },
    {
      "epoch": 145.05,
      "learning_rate": 6.601941747572816e-07,
      "loss": 0.086,
      "step": 373500
    },
    {
      "epoch": 145.05,
      "learning_rate": 6.596763754045308e-07,
      "loss": 0.0592,
      "step": 373510
    },
    {
      "epoch": 145.06,
      "learning_rate": 6.5915857605178e-07,
      "loss": 0.0039,
      "step": 373520
    },
    {
      "epoch": 145.06,
      "learning_rate": 6.586407766990292e-07,
      "loss": 0.0204,
      "step": 373530
    },
    {
      "epoch": 145.06,
      "learning_rate": 6.581229773462784e-07,
      "loss": 0.036,
      "step": 373540
    },
    {
      "epoch": 145.07,
      "learning_rate": 6.576051779935276e-07,
      "loss": 0.0024,
      "step": 373550
    },
    {
      "epoch": 145.07,
      "learning_rate": 6.570873786407768e-07,
      "loss": 0.0208,
      "step": 373560
    },
    {
      "epoch": 145.08,
      "learning_rate": 6.56569579288026e-07,
      "loss": 0.0087,
      "step": 373570
    },
    {
      "epoch": 145.08,
      "learning_rate": 6.560517799352752e-07,
      "loss": 0.0714,
      "step": 373580
    },
    {
      "epoch": 145.08,
      "learning_rate": 6.555339805825244e-07,
      "loss": 0.0212,
      "step": 373590
    },
    {
      "epoch": 145.09,
      "learning_rate": 6.550161812297736e-07,
      "loss": 0.0423,
      "step": 373600
    },
    {
      "epoch": 145.09,
      "learning_rate": 6.544983818770227e-07,
      "loss": 0.0537,
      "step": 373610
    },
    {
      "epoch": 145.1,
      "learning_rate": 6.539805825242719e-07,
      "loss": 0.0002,
      "step": 373620
    },
    {
      "epoch": 145.1,
      "learning_rate": 6.534627831715211e-07,
      "loss": 0.2241,
      "step": 373630
    },
    {
      "epoch": 145.1,
      "learning_rate": 6.529449838187703e-07,
      "loss": 0.0084,
      "step": 373640
    },
    {
      "epoch": 145.11,
      "learning_rate": 6.524271844660195e-07,
      "loss": 0.0247,
      "step": 373650
    },
    {
      "epoch": 145.11,
      "learning_rate": 6.519093851132686e-07,
      "loss": 0.1509,
      "step": 373660
    },
    {
      "epoch": 145.11,
      "learning_rate": 6.513915857605178e-07,
      "loss": 0.0991,
      "step": 373670
    },
    {
      "epoch": 145.12,
      "learning_rate": 6.50873786407767e-07,
      "loss": 0.01,
      "step": 373680
    },
    {
      "epoch": 145.12,
      "learning_rate": 6.503559870550162e-07,
      "loss": 0.0704,
      "step": 373690
    },
    {
      "epoch": 145.13,
      "learning_rate": 6.498381877022654e-07,
      "loss": 0.0017,
      "step": 373700
    },
    {
      "epoch": 145.13,
      "learning_rate": 6.493203883495146e-07,
      "loss": 0.071,
      "step": 373710
    },
    {
      "epoch": 145.13,
      "learning_rate": 6.488025889967638e-07,
      "loss": 0.0301,
      "step": 373720
    },
    {
      "epoch": 145.14,
      "learning_rate": 6.482847896440129e-07,
      "loss": 0.0067,
      "step": 373730
    },
    {
      "epoch": 145.14,
      "learning_rate": 6.477669902912621e-07,
      "loss": 0.1211,
      "step": 373740
    },
    {
      "epoch": 145.15,
      "learning_rate": 6.472491909385113e-07,
      "loss": 0.0202,
      "step": 373750
    },
    {
      "epoch": 145.15,
      "learning_rate": 6.467313915857605e-07,
      "loss": 0.0104,
      "step": 373760
    },
    {
      "epoch": 145.15,
      "learning_rate": 6.462135922330097e-07,
      "loss": 0.038,
      "step": 373770
    },
    {
      "epoch": 145.16,
      "learning_rate": 6.45695792880259e-07,
      "loss": 0.023,
      "step": 373780
    },
    {
      "epoch": 145.16,
      "learning_rate": 6.451779935275082e-07,
      "loss": 0.0811,
      "step": 373790
    },
    {
      "epoch": 145.17,
      "learning_rate": 6.446601941747574e-07,
      "loss": 0.0553,
      "step": 373800
    },
    {
      "epoch": 145.17,
      "learning_rate": 6.441423948220066e-07,
      "loss": 0.0002,
      "step": 373810
    },
    {
      "epoch": 145.17,
      "learning_rate": 6.436245954692558e-07,
      "loss": 0.0015,
      "step": 373820
    },
    {
      "epoch": 145.18,
      "learning_rate": 6.43106796116505e-07,
      "loss": 0.0292,
      "step": 373830
    },
    {
      "epoch": 145.18,
      "learning_rate": 6.425889967637541e-07,
      "loss": 0.0159,
      "step": 373840
    },
    {
      "epoch": 145.18,
      "learning_rate": 6.420711974110033e-07,
      "loss": 0.0059,
      "step": 373850
    },
    {
      "epoch": 145.19,
      "learning_rate": 6.415533980582525e-07,
      "loss": 0.0004,
      "step": 373860
    },
    {
      "epoch": 145.19,
      "learning_rate": 6.410355987055017e-07,
      "loss": 0.0383,
      "step": 373870
    },
    {
      "epoch": 145.2,
      "learning_rate": 6.405177993527509e-07,
      "loss": 0.0069,
      "step": 373880
    },
    {
      "epoch": 145.2,
      "learning_rate": 6.4e-07,
      "loss": 0.0514,
      "step": 373890
    },
    {
      "epoch": 145.2,
      "learning_rate": 6.394822006472492e-07,
      "loss": 0.0154,
      "step": 373900
    },
    {
      "epoch": 145.21,
      "learning_rate": 6.389644012944984e-07,
      "loss": 0.0086,
      "step": 373910
    },
    {
      "epoch": 145.21,
      "learning_rate": 6.384466019417476e-07,
      "loss": 0.0167,
      "step": 373920
    },
    {
      "epoch": 145.22,
      "learning_rate": 6.379288025889968e-07,
      "loss": 0.0347,
      "step": 373930
    },
    {
      "epoch": 145.22,
      "learning_rate": 6.37411003236246e-07,
      "loss": 0.0175,
      "step": 373940
    },
    {
      "epoch": 145.22,
      "learning_rate": 6.368932038834952e-07,
      "loss": 0.0495,
      "step": 373950
    },
    {
      "epoch": 145.23,
      "learning_rate": 6.363754045307444e-07,
      "loss": 0.0001,
      "step": 373960
    },
    {
      "epoch": 145.23,
      "learning_rate": 6.358576051779936e-07,
      "loss": 0.026,
      "step": 373970
    },
    {
      "epoch": 145.23,
      "learning_rate": 6.353398058252428e-07,
      "loss": 0.0088,
      "step": 373980
    },
    {
      "epoch": 145.24,
      "learning_rate": 6.34822006472492e-07,
      "loss": 0.042,
      "step": 373990
    },
    {
      "epoch": 145.24,
      "learning_rate": 6.343042071197412e-07,
      "loss": 0.0029,
      "step": 374000
    },
    {
      "epoch": 145.25,
      "learning_rate": 6.337864077669903e-07,
      "loss": 0.0821,
      "step": 374010
    },
    {
      "epoch": 145.25,
      "learning_rate": 6.332686084142394e-07,
      "loss": 0.0001,
      "step": 374020
    },
    {
      "epoch": 145.25,
      "learning_rate": 6.327508090614886e-07,
      "loss": 0.0016,
      "step": 374030
    },
    {
      "epoch": 145.26,
      "learning_rate": 6.322330097087378e-07,
      "loss": 0.0804,
      "step": 374040
    },
    {
      "epoch": 145.26,
      "learning_rate": 6.31715210355987e-07,
      "loss": 0.0493,
      "step": 374050
    },
    {
      "epoch": 145.27,
      "learning_rate": 6.311974110032363e-07,
      "loss": 0.0006,
      "step": 374060
    },
    {
      "epoch": 145.27,
      "learning_rate": 6.306796116504855e-07,
      "loss": 0.0974,
      "step": 374070
    },
    {
      "epoch": 145.27,
      "learning_rate": 6.301618122977347e-07,
      "loss": 0.0006,
      "step": 374080
    },
    {
      "epoch": 145.28,
      "learning_rate": 6.296440129449839e-07,
      "loss": 0.0043,
      "step": 374090
    },
    {
      "epoch": 145.28,
      "learning_rate": 6.291262135922331e-07,
      "loss": 0.0465,
      "step": 374100
    },
    {
      "epoch": 145.29,
      "learning_rate": 6.286084142394823e-07,
      "loss": 0.0007,
      "step": 374110
    },
    {
      "epoch": 145.29,
      "learning_rate": 6.280906148867315e-07,
      "loss": 0.0005,
      "step": 374120
    },
    {
      "epoch": 145.29,
      "learning_rate": 6.275728155339807e-07,
      "loss": 0.0134,
      "step": 374130
    },
    {
      "epoch": 145.3,
      "learning_rate": 6.270550161812299e-07,
      "loss": 0.1454,
      "step": 374140
    },
    {
      "epoch": 145.3,
      "learning_rate": 6.265372168284791e-07,
      "loss": 0.0277,
      "step": 374150
    },
    {
      "epoch": 145.3,
      "learning_rate": 6.260194174757283e-07,
      "loss": 0.1187,
      "step": 374160
    },
    {
      "epoch": 145.31,
      "learning_rate": 6.255016181229775e-07,
      "loss": 0.0145,
      "step": 374170
    },
    {
      "epoch": 145.31,
      "learning_rate": 6.249838187702265e-07,
      "loss": 0.0357,
      "step": 374180
    },
    {
      "epoch": 145.32,
      "learning_rate": 6.244660194174757e-07,
      "loss": 0.109,
      "step": 374190
    },
    {
      "epoch": 145.32,
      "learning_rate": 6.239482200647249e-07,
      "loss": 0.0336,
      "step": 374200
    },
    {
      "epoch": 145.32,
      "learning_rate": 6.234304207119741e-07,
      "loss": 0.0232,
      "step": 374210
    },
    {
      "epoch": 145.33,
      "learning_rate": 6.229126213592233e-07,
      "loss": 0.0084,
      "step": 374220
    },
    {
      "epoch": 145.33,
      "learning_rate": 6.223948220064725e-07,
      "loss": 0.0002,
      "step": 374230
    },
    {
      "epoch": 145.34,
      "learning_rate": 6.218770226537217e-07,
      "loss": 0.0001,
      "step": 374240
    },
    {
      "epoch": 145.34,
      "learning_rate": 6.213592233009709e-07,
      "loss": 0.0433,
      "step": 374250
    },
    {
      "epoch": 145.34,
      "learning_rate": 6.208414239482201e-07,
      "loss": 0.075,
      "step": 374260
    },
    {
      "epoch": 145.35,
      "learning_rate": 6.203236245954693e-07,
      "loss": 0.0088,
      "step": 374270
    },
    {
      "epoch": 145.35,
      "learning_rate": 6.198058252427185e-07,
      "loss": 0.0065,
      "step": 374280
    },
    {
      "epoch": 145.36,
      "learning_rate": 6.192880258899677e-07,
      "loss": 0.0258,
      "step": 374290
    },
    {
      "epoch": 145.36,
      "learning_rate": 6.187702265372169e-07,
      "loss": 0.0819,
      "step": 374300
    },
    {
      "epoch": 145.36,
      "learning_rate": 6.182524271844661e-07,
      "loss": 0.0661,
      "step": 374310
    },
    {
      "epoch": 145.37,
      "learning_rate": 6.177346278317153e-07,
      "loss": 0.0246,
      "step": 374320
    },
    {
      "epoch": 145.37,
      "learning_rate": 6.172168284789645e-07,
      "loss": 0.1516,
      "step": 374330
    },
    {
      "epoch": 145.37,
      "learning_rate": 6.166990291262137e-07,
      "loss": 0.0019,
      "step": 374340
    },
    {
      "epoch": 145.38,
      "learning_rate": 6.161812297734629e-07,
      "loss": 0.0123,
      "step": 374350
    },
    {
      "epoch": 145.38,
      "learning_rate": 6.15663430420712e-07,
      "loss": 0.0515,
      "step": 374360
    },
    {
      "epoch": 145.39,
      "learning_rate": 6.151456310679612e-07,
      "loss": 0.0118,
      "step": 374370
    },
    {
      "epoch": 145.39,
      "learning_rate": 6.146278317152103e-07,
      "loss": 0.008,
      "step": 374380
    },
    {
      "epoch": 145.39,
      "learning_rate": 6.141100323624595e-07,
      "loss": 0.0081,
      "step": 374390
    },
    {
      "epoch": 145.4,
      "learning_rate": 6.135922330097087e-07,
      "loss": 0.0519,
      "step": 374400
    },
    {
      "epoch": 145.4,
      "learning_rate": 6.13074433656958e-07,
      "loss": 0.036,
      "step": 374410
    },
    {
      "epoch": 145.41,
      "learning_rate": 6.125566343042072e-07,
      "loss": 0.0655,
      "step": 374420
    },
    {
      "epoch": 145.41,
      "learning_rate": 6.120388349514564e-07,
      "loss": 0.004,
      "step": 374430
    },
    {
      "epoch": 145.41,
      "learning_rate": 6.115210355987056e-07,
      "loss": 0.0135,
      "step": 374440
    },
    {
      "epoch": 145.42,
      "learning_rate": 6.110032362459548e-07,
      "loss": 0.0002,
      "step": 374450
    },
    {
      "epoch": 145.42,
      "learning_rate": 6.104854368932039e-07,
      "loss": 0.0452,
      "step": 374460
    },
    {
      "epoch": 145.43,
      "learning_rate": 6.099676375404531e-07,
      "loss": 0.0032,
      "step": 374470
    },
    {
      "epoch": 145.43,
      "learning_rate": 6.094498381877023e-07,
      "loss": 0.0001,
      "step": 374480
    },
    {
      "epoch": 145.43,
      "learning_rate": 6.089320388349515e-07,
      "loss": 0.0493,
      "step": 374490
    },
    {
      "epoch": 145.44,
      "learning_rate": 6.084142394822007e-07,
      "loss": 0.0873,
      "step": 374500
    },
    {
      "epoch": 145.44,
      "learning_rate": 6.078964401294499e-07,
      "loss": 0.0234,
      "step": 374510
    },
    {
      "epoch": 145.44,
      "learning_rate": 6.073786407766991e-07,
      "loss": 0.0005,
      "step": 374520
    },
    {
      "epoch": 145.45,
      "learning_rate": 6.068608414239483e-07,
      "loss": 0.0825,
      "step": 374530
    },
    {
      "epoch": 145.45,
      "learning_rate": 6.063430420711974e-07,
      "loss": 0.0032,
      "step": 374540
    },
    {
      "epoch": 145.46,
      "learning_rate": 6.058252427184466e-07,
      "loss": 0.0213,
      "step": 374550
    },
    {
      "epoch": 145.46,
      "learning_rate": 6.053074433656958e-07,
      "loss": 0.033,
      "step": 374560
    },
    {
      "epoch": 145.46,
      "learning_rate": 6.04789644012945e-07,
      "loss": 0.0083,
      "step": 374570
    },
    {
      "epoch": 145.47,
      "learning_rate": 6.042718446601942e-07,
      "loss": 0.0314,
      "step": 374580
    },
    {
      "epoch": 145.47,
      "learning_rate": 6.037540453074434e-07,
      "loss": 0.0186,
      "step": 374590
    },
    {
      "epoch": 145.48,
      "learning_rate": 6.032362459546926e-07,
      "loss": 0.0002,
      "step": 374600
    },
    {
      "epoch": 145.48,
      "learning_rate": 6.027184466019418e-07,
      "loss": 0.0307,
      "step": 374610
    },
    {
      "epoch": 145.48,
      "learning_rate": 6.02200647249191e-07,
      "loss": 0.0003,
      "step": 374620
    },
    {
      "epoch": 145.49,
      "learning_rate": 6.016828478964402e-07,
      "loss": 0.0017,
      "step": 374630
    },
    {
      "epoch": 145.49,
      "learning_rate": 6.011650485436894e-07,
      "loss": 0.0233,
      "step": 374640
    },
    {
      "epoch": 145.5,
      "learning_rate": 6.006472491909386e-07,
      "loss": 0.0142,
      "step": 374650
    },
    {
      "epoch": 145.5,
      "learning_rate": 6.001294498381878e-07,
      "loss": 0.0086,
      "step": 374660
    },
    {
      "epoch": 145.5,
      "learning_rate": 5.996116504854369e-07,
      "loss": 0.0007,
      "step": 374670
    },
    {
      "epoch": 145.51,
      "learning_rate": 5.990938511326861e-07,
      "loss": 0.0965,
      "step": 374680
    },
    {
      "epoch": 145.51,
      "learning_rate": 5.985760517799354e-07,
      "loss": 0.0281,
      "step": 374690
    },
    {
      "epoch": 145.51,
      "learning_rate": 5.980582524271846e-07,
      "loss": 0.0154,
      "step": 374700
    },
    {
      "epoch": 145.52,
      "learning_rate": 5.975404530744338e-07,
      "loss": 0.0093,
      "step": 374710
    },
    {
      "epoch": 145.52,
      "learning_rate": 5.97022653721683e-07,
      "loss": 0.099,
      "step": 374720
    },
    {
      "epoch": 145.53,
      "learning_rate": 5.965048543689321e-07,
      "loss": 0.0241,
      "step": 374730
    },
    {
      "epoch": 145.53,
      "learning_rate": 5.959870550161813e-07,
      "loss": 0.0072,
      "step": 374740
    },
    {
      "epoch": 145.53,
      "learning_rate": 5.954692556634304e-07,
      "loss": 0.0507,
      "step": 374750
    },
    {
      "epoch": 145.54,
      "learning_rate": 5.949514563106796e-07,
      "loss": 0.0125,
      "step": 374760
    },
    {
      "epoch": 145.54,
      "learning_rate": 5.944336569579288e-07,
      "loss": 0.0874,
      "step": 374770
    },
    {
      "epoch": 145.55,
      "learning_rate": 5.93915857605178e-07,
      "loss": 0.0591,
      "step": 374780
    },
    {
      "epoch": 145.55,
      "learning_rate": 5.933980582524272e-07,
      "loss": 0.01,
      "step": 374790
    },
    {
      "epoch": 145.55,
      "learning_rate": 5.928802588996764e-07,
      "loss": 0.0833,
      "step": 374800
    },
    {
      "epoch": 145.56,
      "learning_rate": 5.923624595469256e-07,
      "loss": 0.0219,
      "step": 374810
    },
    {
      "epoch": 145.56,
      "learning_rate": 5.918446601941748e-07,
      "loss": 0.0874,
      "step": 374820
    },
    {
      "epoch": 145.57,
      "learning_rate": 5.91326860841424e-07,
      "loss": 0.0163,
      "step": 374830
    },
    {
      "epoch": 145.57,
      "learning_rate": 5.908090614886732e-07,
      "loss": 0.0781,
      "step": 374840
    },
    {
      "epoch": 145.57,
      "learning_rate": 5.902912621359224e-07,
      "loss": 0.1512,
      "step": 374850
    },
    {
      "epoch": 145.58,
      "learning_rate": 5.897734627831716e-07,
      "loss": 0.0394,
      "step": 374860
    },
    {
      "epoch": 145.58,
      "learning_rate": 5.892556634304208e-07,
      "loss": 0.0497,
      "step": 374870
    },
    {
      "epoch": 145.58,
      "learning_rate": 5.8873786407767e-07,
      "loss": 0.0877,
      "step": 374880
    },
    {
      "epoch": 145.59,
      "learning_rate": 5.882200647249191e-07,
      "loss": 0.0005,
      "step": 374890
    },
    {
      "epoch": 145.59,
      "learning_rate": 5.877022653721683e-07,
      "loss": 0.1208,
      "step": 374900
    },
    {
      "epoch": 145.6,
      "learning_rate": 5.871844660194175e-07,
      "loss": 0.0576,
      "step": 374910
    },
    {
      "epoch": 145.6,
      "learning_rate": 5.866666666666667e-07,
      "loss": 0.0138,
      "step": 374920
    },
    {
      "epoch": 145.6,
      "learning_rate": 5.861488673139159e-07,
      "loss": 0.0004,
      "step": 374930
    },
    {
      "epoch": 145.61,
      "learning_rate": 5.856310679611651e-07,
      "loss": 0.0001,
      "step": 374940
    },
    {
      "epoch": 145.61,
      "learning_rate": 5.851132686084142e-07,
      "loss": 0.06,
      "step": 374950
    },
    {
      "epoch": 145.62,
      "learning_rate": 5.845954692556634e-07,
      "loss": 0.0735,
      "step": 374960
    },
    {
      "epoch": 145.62,
      "learning_rate": 5.840776699029127e-07,
      "loss": 0.0261,
      "step": 374970
    },
    {
      "epoch": 145.62,
      "learning_rate": 5.835598705501619e-07,
      "loss": 0.0102,
      "step": 374980
    },
    {
      "epoch": 145.63,
      "learning_rate": 5.830420711974111e-07,
      "loss": 0.0093,
      "step": 374990
    },
    {
      "epoch": 145.63,
      "learning_rate": 5.825242718446603e-07,
      "loss": 0.0378,
      "step": 375000
    },
    {
      "epoch": 145.63,
      "learning_rate": 5.820064724919095e-07,
      "loss": 0.0002,
      "step": 375010
    },
    {
      "epoch": 145.64,
      "learning_rate": 5.814886731391587e-07,
      "loss": 0.0032,
      "step": 375020
    },
    {
      "epoch": 145.64,
      "learning_rate": 5.809708737864078e-07,
      "loss": 0.0543,
      "step": 375030
    },
    {
      "epoch": 145.65,
      "learning_rate": 5.80453074433657e-07,
      "loss": 0.036,
      "step": 375040
    },
    {
      "epoch": 145.65,
      "learning_rate": 5.799352750809062e-07,
      "loss": 0.0544,
      "step": 375050
    },
    {
      "epoch": 145.65,
      "learning_rate": 5.794174757281553e-07,
      "loss": 0.0365,
      "step": 375060
    },
    {
      "epoch": 145.66,
      "learning_rate": 5.788996763754045e-07,
      "loss": 0.0007,
      "step": 375070
    },
    {
      "epoch": 145.66,
      "learning_rate": 5.783818770226537e-07,
      "loss": 0.0747,
      "step": 375080
    },
    {
      "epoch": 145.67,
      "learning_rate": 5.778640776699029e-07,
      "loss": 0.0001,
      "step": 375090
    },
    {
      "epoch": 145.67,
      "learning_rate": 5.773462783171521e-07,
      "loss": 0.0361,
      "step": 375100
    },
    {
      "epoch": 145.67,
      "learning_rate": 5.768284789644014e-07,
      "loss": 0.0003,
      "step": 375110
    },
    {
      "epoch": 145.68,
      "learning_rate": 5.763106796116505e-07,
      "loss": 0.0048,
      "step": 375120
    },
    {
      "epoch": 145.68,
      "learning_rate": 5.757928802588997e-07,
      "loss": 0.0236,
      "step": 375130
    },
    {
      "epoch": 145.69,
      "learning_rate": 5.752750809061489e-07,
      "loss": 0.039,
      "step": 375140
    },
    {
      "epoch": 145.69,
      "learning_rate": 5.747572815533981e-07,
      "loss": 0.0165,
      "step": 375150
    },
    {
      "epoch": 145.69,
      "learning_rate": 5.742394822006473e-07,
      "loss": 0.0789,
      "step": 375160
    },
    {
      "epoch": 145.7,
      "learning_rate": 5.737216828478965e-07,
      "loss": 0.0539,
      "step": 375170
    },
    {
      "epoch": 145.7,
      "learning_rate": 5.732038834951457e-07,
      "loss": 0.1107,
      "step": 375180
    },
    {
      "epoch": 145.7,
      "learning_rate": 5.726860841423949e-07,
      "loss": 0.0001,
      "step": 375190
    },
    {
      "epoch": 145.71,
      "learning_rate": 5.721682847896441e-07,
      "loss": 0.0718,
      "step": 375200
    },
    {
      "epoch": 145.71,
      "learning_rate": 5.716504854368933e-07,
      "loss": 0.014,
      "step": 375210
    },
    {
      "epoch": 145.72,
      "learning_rate": 5.711326860841425e-07,
      "loss": 0.0971,
      "step": 375220
    },
    {
      "epoch": 145.72,
      "learning_rate": 5.706148867313917e-07,
      "loss": 0.0102,
      "step": 375230
    },
    {
      "epoch": 145.72,
      "learning_rate": 5.700970873786407e-07,
      "loss": 0.1214,
      "step": 375240
    },
    {
      "epoch": 145.73,
      "learning_rate": 5.6957928802589e-07,
      "loss": 0.1091,
      "step": 375250
    },
    {
      "epoch": 145.73,
      "learning_rate": 5.690614886731392e-07,
      "loss": 0.0497,
      "step": 375260
    },
    {
      "epoch": 145.74,
      "learning_rate": 5.685436893203884e-07,
      "loss": 0.0106,
      "step": 375270
    },
    {
      "epoch": 145.74,
      "learning_rate": 5.680258899676376e-07,
      "loss": 0.0008,
      "step": 375280
    },
    {
      "epoch": 145.74,
      "learning_rate": 5.675080906148868e-07,
      "loss": 0.0022,
      "step": 375290
    },
    {
      "epoch": 145.75,
      "learning_rate": 5.66990291262136e-07,
      "loss": 0.0014,
      "step": 375300
    },
    {
      "epoch": 145.75,
      "learning_rate": 5.664724919093852e-07,
      "loss": 0.0354,
      "step": 375310
    },
    {
      "epoch": 145.76,
      "learning_rate": 5.659546925566343e-07,
      "loss": 0.0323,
      "step": 375320
    },
    {
      "epoch": 145.76,
      "learning_rate": 5.654368932038835e-07,
      "loss": 0.0458,
      "step": 375330
    },
    {
      "epoch": 145.76,
      "learning_rate": 5.649190938511327e-07,
      "loss": 0.0077,
      "step": 375340
    },
    {
      "epoch": 145.77,
      "learning_rate": 5.644012944983819e-07,
      "loss": 0.0699,
      "step": 375350
    },
    {
      "epoch": 145.77,
      "learning_rate": 5.638834951456311e-07,
      "loss": 0.0037,
      "step": 375360
    },
    {
      "epoch": 145.77,
      "learning_rate": 5.633656957928803e-07,
      "loss": 0.0866,
      "step": 375370
    },
    {
      "epoch": 145.78,
      "learning_rate": 5.628478964401295e-07,
      "loss": 0.0072,
      "step": 375380
    },
    {
      "epoch": 145.78,
      "learning_rate": 5.623300970873788e-07,
      "loss": 0.0008,
      "step": 375390
    },
    {
      "epoch": 145.79,
      "learning_rate": 5.618122977346279e-07,
      "loss": 0.0149,
      "step": 375400
    },
    {
      "epoch": 145.79,
      "learning_rate": 5.61294498381877e-07,
      "loss": 0.0286,
      "step": 375410
    },
    {
      "epoch": 145.79,
      "learning_rate": 5.607766990291262e-07,
      "loss": 0.018,
      "step": 375420
    },
    {
      "epoch": 145.8,
      "learning_rate": 5.602588996763754e-07,
      "loss": 0.1283,
      "step": 375430
    },
    {
      "epoch": 145.8,
      "learning_rate": 5.597411003236246e-07,
      "loss": 0.1025,
      "step": 375440
    },
    {
      "epoch": 145.81,
      "learning_rate": 5.592233009708738e-07,
      "loss": 0.0678,
      "step": 375450
    },
    {
      "epoch": 145.81,
      "learning_rate": 5.58705501618123e-07,
      "loss": 0.0684,
      "step": 375460
    },
    {
      "epoch": 145.81,
      "learning_rate": 5.581877022653722e-07,
      "loss": 0.0203,
      "step": 375470
    },
    {
      "epoch": 145.82,
      "learning_rate": 5.576699029126214e-07,
      "loss": 0.0669,
      "step": 375480
    },
    {
      "epoch": 145.82,
      "learning_rate": 5.571521035598706e-07,
      "loss": 0.0209,
      "step": 375490
    },
    {
      "epoch": 145.83,
      "learning_rate": 5.566343042071198e-07,
      "loss": 0.0271,
      "step": 375500
    },
    {
      "epoch": 145.83,
      "learning_rate": 5.56116504854369e-07,
      "loss": 0.0493,
      "step": 375510
    },
    {
      "epoch": 145.83,
      "learning_rate": 5.555987055016181e-07,
      "loss": 0.0299,
      "step": 375520
    },
    {
      "epoch": 145.84,
      "learning_rate": 5.550809061488674e-07,
      "loss": 0.0134,
      "step": 375530
    },
    {
      "epoch": 145.84,
      "learning_rate": 5.545631067961166e-07,
      "loss": 0.0003,
      "step": 375540
    },
    {
      "epoch": 145.84,
      "learning_rate": 5.540453074433658e-07,
      "loss": 0.0086,
      "step": 375550
    },
    {
      "epoch": 145.85,
      "learning_rate": 5.53527508090615e-07,
      "loss": 0.0077,
      "step": 375560
    },
    {
      "epoch": 145.85,
      "learning_rate": 5.530097087378642e-07,
      "loss": 0.0919,
      "step": 375570
    },
    {
      "epoch": 145.86,
      "learning_rate": 5.524919093851134e-07,
      "loss": 0.111,
      "step": 375580
    },
    {
      "epoch": 145.86,
      "learning_rate": 5.519741100323626e-07,
      "loss": 0.0123,
      "step": 375590
    },
    {
      "epoch": 145.86,
      "learning_rate": 5.514563106796117e-07,
      "loss": 0.0093,
      "step": 375600
    },
    {
      "epoch": 145.87,
      "learning_rate": 5.509385113268608e-07,
      "loss": 0.1048,
      "step": 375610
    },
    {
      "epoch": 145.87,
      "learning_rate": 5.5042071197411e-07,
      "loss": 0.0117,
      "step": 375620
    },
    {
      "epoch": 145.88,
      "learning_rate": 5.499029126213592e-07,
      "loss": 0.0005,
      "step": 375630
    },
    {
      "epoch": 145.88,
      "learning_rate": 5.493851132686084e-07,
      "loss": 0.045,
      "step": 375640
    },
    {
      "epoch": 145.88,
      "learning_rate": 5.488673139158576e-07,
      "loss": 0.2384,
      "step": 375650
    },
    {
      "epoch": 145.89,
      "learning_rate": 5.483495145631068e-07,
      "loss": 0.0492,
      "step": 375660
    },
    {
      "epoch": 145.89,
      "learning_rate": 5.478317152103561e-07,
      "loss": 0.0003,
      "step": 375670
    },
    {
      "epoch": 145.9,
      "learning_rate": 5.473139158576053e-07,
      "loss": 0.034,
      "step": 375680
    },
    {
      "epoch": 145.9,
      "learning_rate": 5.467961165048544e-07,
      "loss": 0.0001,
      "step": 375690
    },
    {
      "epoch": 145.9,
      "learning_rate": 5.462783171521036e-07,
      "loss": 0.0008,
      "step": 375700
    },
    {
      "epoch": 145.91,
      "learning_rate": 5.457605177993528e-07,
      "loss": 0.0157,
      "step": 375710
    },
    {
      "epoch": 145.91,
      "learning_rate": 5.45242718446602e-07,
      "loss": 0.0306,
      "step": 375720
    },
    {
      "epoch": 145.91,
      "learning_rate": 5.447249190938512e-07,
      "loss": 0.0004,
      "step": 375730
    },
    {
      "epoch": 145.92,
      "learning_rate": 5.442071197411004e-07,
      "loss": 0.1051,
      "step": 375740
    },
    {
      "epoch": 145.92,
      "learning_rate": 5.436893203883496e-07,
      "loss": 0.0285,
      "step": 375750
    },
    {
      "epoch": 145.93,
      "learning_rate": 5.431715210355988e-07,
      "loss": 0.1853,
      "step": 375760
    },
    {
      "epoch": 145.93,
      "learning_rate": 5.42653721682848e-07,
      "loss": 0.0013,
      "step": 375770
    },
    {
      "epoch": 145.93,
      "learning_rate": 5.421359223300971e-07,
      "loss": 0.0168,
      "step": 375780
    },
    {
      "epoch": 145.94,
      "learning_rate": 5.416181229773463e-07,
      "loss": 0.0158,
      "step": 375790
    },
    {
      "epoch": 145.94,
      "learning_rate": 5.411003236245955e-07,
      "loss": 0.0296,
      "step": 375800
    },
    {
      "epoch": 145.95,
      "learning_rate": 5.405825242718447e-07,
      "loss": 0.0146,
      "step": 375810
    },
    {
      "epoch": 145.95,
      "learning_rate": 5.400647249190939e-07,
      "loss": 0.0933,
      "step": 375820
    },
    {
      "epoch": 145.95,
      "learning_rate": 5.395469255663431e-07,
      "loss": 0.0241,
      "step": 375830
    },
    {
      "epoch": 145.96,
      "learning_rate": 5.390291262135923e-07,
      "loss": 0.0226,
      "step": 375840
    },
    {
      "epoch": 145.96,
      "learning_rate": 5.385113268608415e-07,
      "loss": 0.025,
      "step": 375850
    },
    {
      "epoch": 145.97,
      "learning_rate": 5.379935275080907e-07,
      "loss": 0.0164,
      "step": 375860
    },
    {
      "epoch": 145.97,
      "learning_rate": 5.374757281553399e-07,
      "loss": 0.0106,
      "step": 375870
    },
    {
      "epoch": 145.97,
      "learning_rate": 5.369579288025891e-07,
      "loss": 0.0064,
      "step": 375880
    },
    {
      "epoch": 145.98,
      "learning_rate": 5.364401294498382e-07,
      "loss": 0.0184,
      "step": 375890
    },
    {
      "epoch": 145.98,
      "learning_rate": 5.359223300970874e-07,
      "loss": 0.0189,
      "step": 375900
    },
    {
      "epoch": 145.98,
      "learning_rate": 5.354045307443366e-07,
      "loss": 0.0129,
      "step": 375910
    },
    {
      "epoch": 145.99,
      "learning_rate": 5.348867313915858e-07,
      "loss": 0.0336,
      "step": 375920
    },
    {
      "epoch": 145.99,
      "learning_rate": 5.34368932038835e-07,
      "loss": 0.0215,
      "step": 375930
    },
    {
      "epoch": 146.0,
      "learning_rate": 5.338511326860841e-07,
      "loss": 0.0196,
      "step": 375940
    },
    {
      "epoch": 146.0,
      "learning_rate": 5.333333333333335e-07,
      "loss": 0.0495,
      "step": 375950
    },
    {
      "epoch": 146.0,
      "eval_accuracy": 0.9535075653370014,
      "eval_loss": 0.38978180289268494,
      "eval_runtime": 8.2388,
      "eval_samples_per_second": 441.202,
      "eval_steps_per_second": 55.226,
      "step": 375950
    },
    {
      "epoch": 146.0,
      "learning_rate": 5.328155339805826e-07,
      "loss": 0.0003,
      "step": 375960
    },
    {
      "epoch": 146.01,
      "learning_rate": 5.322977346278317e-07,
      "loss": 0.041,
      "step": 375970
    },
    {
      "epoch": 146.01,
      "learning_rate": 5.317799352750809e-07,
      "loss": 0.018,
      "step": 375980
    },
    {
      "epoch": 146.02,
      "learning_rate": 5.312621359223301e-07,
      "loss": 0.0001,
      "step": 375990
    },
    {
      "epoch": 146.02,
      "learning_rate": 5.307443365695793e-07,
      "loss": 0.0193,
      "step": 376000
    },
    {
      "epoch": 146.02,
      "learning_rate": 5.302265372168285e-07,
      "loss": 0.0001,
      "step": 376010
    },
    {
      "epoch": 146.03,
      "learning_rate": 5.297087378640777e-07,
      "loss": 0.0089,
      "step": 376020
    },
    {
      "epoch": 146.03,
      "learning_rate": 5.291909385113269e-07,
      "loss": 0.0139,
      "step": 376030
    },
    {
      "epoch": 146.03,
      "learning_rate": 5.286731391585761e-07,
      "loss": 0.0334,
      "step": 376040
    },
    {
      "epoch": 146.04,
      "learning_rate": 5.281553398058253e-07,
      "loss": 0.0003,
      "step": 376050
    },
    {
      "epoch": 146.04,
      "learning_rate": 5.276375404530745e-07,
      "loss": 0.0215,
      "step": 376060
    },
    {
      "epoch": 146.05,
      "learning_rate": 5.271197411003237e-07,
      "loss": 0.0314,
      "step": 376070
    },
    {
      "epoch": 146.05,
      "learning_rate": 5.266019417475729e-07,
      "loss": 0.049,
      "step": 376080
    },
    {
      "epoch": 146.05,
      "learning_rate": 5.260841423948221e-07,
      "loss": 0.1313,
      "step": 376090
    },
    {
      "epoch": 146.06,
      "learning_rate": 5.255663430420713e-07,
      "loss": 0.0517,
      "step": 376100
    },
    {
      "epoch": 146.06,
      "learning_rate": 5.250485436893205e-07,
      "loss": 0.002,
      "step": 376110
    },
    {
      "epoch": 146.07,
      "learning_rate": 5.245307443365697e-07,
      "loss": 0.1678,
      "step": 376120
    },
    {
      "epoch": 146.07,
      "learning_rate": 5.240129449838188e-07,
      "loss": 0.0,
      "step": 376130
    },
    {
      "epoch": 146.07,
      "learning_rate": 5.23495145631068e-07,
      "loss": 0.0351,
      "step": 376140
    },
    {
      "epoch": 146.08,
      "learning_rate": 5.229773462783172e-07,
      "loss": 0.0229,
      "step": 376150
    },
    {
      "epoch": 146.08,
      "learning_rate": 5.224595469255664e-07,
      "loss": 0.0593,
      "step": 376160
    },
    {
      "epoch": 146.09,
      "learning_rate": 5.219417475728156e-07,
      "loss": 0.0844,
      "step": 376170
    },
    {
      "epoch": 146.09,
      "learning_rate": 5.214239482200647e-07,
      "loss": 0.0265,
      "step": 376180
    },
    {
      "epoch": 146.09,
      "learning_rate": 5.209061488673139e-07,
      "loss": 0.1203,
      "step": 376190
    },
    {
      "epoch": 146.1,
      "learning_rate": 5.203883495145631e-07,
      "loss": 0.0258,
      "step": 376200
    },
    {
      "epoch": 146.1,
      "learning_rate": 5.198705501618123e-07,
      "loss": 0.021,
      "step": 376210
    },
    {
      "epoch": 146.1,
      "learning_rate": 5.193527508090615e-07,
      "loss": 0.0001,
      "step": 376220
    },
    {
      "epoch": 146.11,
      "learning_rate": 5.188349514563108e-07,
      "loss": 0.1422,
      "step": 376230
    },
    {
      "epoch": 146.11,
      "learning_rate": 5.1831715210356e-07,
      "loss": 0.0332,
      "step": 376240
    },
    {
      "epoch": 146.12,
      "learning_rate": 5.177993527508092e-07,
      "loss": 0.1191,
      "step": 376250
    },
    {
      "epoch": 146.12,
      "learning_rate": 5.172815533980583e-07,
      "loss": 0.1056,
      "step": 376260
    },
    {
      "epoch": 146.12,
      "learning_rate": 5.167637540453075e-07,
      "loss": 0.0178,
      "step": 376270
    },
    {
      "epoch": 146.13,
      "learning_rate": 5.162459546925567e-07,
      "loss": 0.0513,
      "step": 376280
    },
    {
      "epoch": 146.13,
      "learning_rate": 5.157281553398059e-07,
      "loss": 0.0001,
      "step": 376290
    },
    {
      "epoch": 146.14,
      "learning_rate": 5.15210355987055e-07,
      "loss": 0.0085,
      "step": 376300
    },
    {
      "epoch": 146.14,
      "learning_rate": 5.146925566343042e-07,
      "loss": 0.0437,
      "step": 376310
    },
    {
      "epoch": 146.14,
      "learning_rate": 5.141747572815534e-07,
      "loss": 0.0001,
      "step": 376320
    },
    {
      "epoch": 146.15,
      "learning_rate": 5.136569579288026e-07,
      "loss": 0.0397,
      "step": 376330
    },
    {
      "epoch": 146.15,
      "learning_rate": 5.131391585760518e-07,
      "loss": 0.0452,
      "step": 376340
    },
    {
      "epoch": 146.16,
      "learning_rate": 5.12621359223301e-07,
      "loss": 0.0237,
      "step": 376350
    },
    {
      "epoch": 146.16,
      "learning_rate": 5.121035598705502e-07,
      "loss": 0.1014,
      "step": 376360
    },
    {
      "epoch": 146.16,
      "learning_rate": 5.115857605177994e-07,
      "loss": 0.074,
      "step": 376370
    },
    {
      "epoch": 146.17,
      "learning_rate": 5.110679611650486e-07,
      "loss": 0.0779,
      "step": 376380
    },
    {
      "epoch": 146.17,
      "learning_rate": 5.105501618122978e-07,
      "loss": 0.0419,
      "step": 376390
    },
    {
      "epoch": 146.17,
      "learning_rate": 5.10032362459547e-07,
      "loss": 0.0412,
      "step": 376400
    },
    {
      "epoch": 146.18,
      "learning_rate": 5.095145631067962e-07,
      "loss": 0.0849,
      "step": 376410
    },
    {
      "epoch": 146.18,
      "learning_rate": 5.089967637540454e-07,
      "loss": 0.098,
      "step": 376420
    },
    {
      "epoch": 146.19,
      "learning_rate": 5.084789644012946e-07,
      "loss": 0.0718,
      "step": 376430
    },
    {
      "epoch": 146.19,
      "learning_rate": 5.079611650485438e-07,
      "loss": 0.0872,
      "step": 376440
    },
    {
      "epoch": 146.19,
      "learning_rate": 5.07443365695793e-07,
      "loss": 0.0439,
      "step": 376450
    },
    {
      "epoch": 146.2,
      "learning_rate": 5.069255663430421e-07,
      "loss": 0.0573,
      "step": 376460
    },
    {
      "epoch": 146.2,
      "learning_rate": 5.064077669902912e-07,
      "loss": 0.005,
      "step": 376470
    },
    {
      "epoch": 146.21,
      "learning_rate": 5.058899676375404e-07,
      "loss": 0.0298,
      "step": 376480
    },
    {
      "epoch": 146.21,
      "learning_rate": 5.053721682847896e-07,
      "loss": 0.0601,
      "step": 376490
    },
    {
      "epoch": 146.21,
      "learning_rate": 5.048543689320388e-07,
      "loss": 0.0087,
      "step": 376500
    },
    {
      "epoch": 146.22,
      "learning_rate": 5.043365695792881e-07,
      "loss": 0.0262,
      "step": 376510
    },
    {
      "epoch": 146.22,
      "learning_rate": 5.038187702265373e-07,
      "loss": 0.0705,
      "step": 376520
    },
    {
      "epoch": 146.23,
      "learning_rate": 5.033009708737865e-07,
      "loss": 0.0772,
      "step": 376530
    },
    {
      "epoch": 146.23,
      "learning_rate": 5.027831715210356e-07,
      "loss": 0.0276,
      "step": 376540
    },
    {
      "epoch": 146.23,
      "learning_rate": 5.022653721682848e-07,
      "loss": 0.0861,
      "step": 376550
    },
    {
      "epoch": 146.24,
      "learning_rate": 5.01747572815534e-07,
      "loss": 0.0002,
      "step": 376560
    },
    {
      "epoch": 146.24,
      "learning_rate": 5.012297734627832e-07,
      "loss": 0.0489,
      "step": 376570
    },
    {
      "epoch": 146.24,
      "learning_rate": 5.007119741100324e-07,
      "loss": 0.0278,
      "step": 376580
    },
    {
      "epoch": 146.25,
      "learning_rate": 5.001941747572816e-07,
      "loss": 0.0268,
      "step": 376590
    },
    {
      "epoch": 146.25,
      "learning_rate": 4.996763754045308e-07,
      "loss": 0.0242,
      "step": 376600
    },
    {
      "epoch": 146.26,
      "learning_rate": 4.9915857605178e-07,
      "loss": 0.01,
      "step": 376610
    },
    {
      "epoch": 146.26,
      "learning_rate": 4.986407766990292e-07,
      "loss": 0.2157,
      "step": 376620
    },
    {
      "epoch": 146.26,
      "learning_rate": 4.981229773462784e-07,
      "loss": 0.0609,
      "step": 376630
    },
    {
      "epoch": 146.27,
      "learning_rate": 4.976051779935276e-07,
      "loss": 0.0952,
      "step": 376640
    },
    {
      "epoch": 146.27,
      "learning_rate": 4.970873786407767e-07,
      "loss": 0.0889,
      "step": 376650
    },
    {
      "epoch": 146.28,
      "learning_rate": 4.965695792880259e-07,
      "loss": 0.0002,
      "step": 376660
    },
    {
      "epoch": 146.28,
      "learning_rate": 4.960517799352751e-07,
      "loss": 0.0665,
      "step": 376670
    },
    {
      "epoch": 146.28,
      "learning_rate": 4.955339805825243e-07,
      "loss": 0.0453,
      "step": 376680
    },
    {
      "epoch": 146.29,
      "learning_rate": 4.950161812297735e-07,
      "loss": 0.0643,
      "step": 376690
    },
    {
      "epoch": 146.29,
      "learning_rate": 4.944983818770227e-07,
      "loss": 0.111,
      "step": 376700
    },
    {
      "epoch": 146.3,
      "learning_rate": 4.939805825242719e-07,
      "loss": 0.0116,
      "step": 376710
    },
    {
      "epoch": 146.3,
      "learning_rate": 4.934627831715211e-07,
      "loss": 0.0577,
      "step": 376720
    },
    {
      "epoch": 146.3,
      "learning_rate": 4.929449838187703e-07,
      "loss": 0.0488,
      "step": 376730
    },
    {
      "epoch": 146.31,
      "learning_rate": 4.924271844660195e-07,
      "loss": 0.0085,
      "step": 376740
    },
    {
      "epoch": 146.31,
      "learning_rate": 4.919093851132686e-07,
      "loss": 0.0008,
      "step": 376750
    },
    {
      "epoch": 146.31,
      "learning_rate": 4.913915857605178e-07,
      "loss": 0.1489,
      "step": 376760
    },
    {
      "epoch": 146.32,
      "learning_rate": 4.90873786407767e-07,
      "loss": 0.0416,
      "step": 376770
    },
    {
      "epoch": 146.32,
      "learning_rate": 4.903559870550162e-07,
      "loss": 0.0421,
      "step": 376780
    },
    {
      "epoch": 146.33,
      "learning_rate": 4.898381877022655e-07,
      "loss": 0.0172,
      "step": 376790
    },
    {
      "epoch": 146.33,
      "learning_rate": 4.893203883495147e-07,
      "loss": 0.0004,
      "step": 376800
    },
    {
      "epoch": 146.33,
      "learning_rate": 4.888025889967639e-07,
      "loss": 0.0339,
      "step": 376810
    },
    {
      "epoch": 146.34,
      "learning_rate": 4.882847896440131e-07,
      "loss": 0.0585,
      "step": 376820
    },
    {
      "epoch": 146.34,
      "learning_rate": 4.877669902912621e-07,
      "loss": 0.0006,
      "step": 376830
    },
    {
      "epoch": 146.35,
      "learning_rate": 4.872491909385113e-07,
      "loss": 0.0847,
      "step": 376840
    },
    {
      "epoch": 146.35,
      "learning_rate": 4.867313915857605e-07,
      "loss": 0.1234,
      "step": 376850
    },
    {
      "epoch": 146.35,
      "learning_rate": 4.862135922330097e-07,
      "loss": 0.0388,
      "step": 376860
    },
    {
      "epoch": 146.36,
      "learning_rate": 4.856957928802589e-07,
      "loss": 0.0077,
      "step": 376870
    },
    {
      "epoch": 146.36,
      "learning_rate": 4.851779935275081e-07,
      "loss": 0.1148,
      "step": 376880
    },
    {
      "epoch": 146.37,
      "learning_rate": 4.846601941747573e-07,
      "loss": 0.0001,
      "step": 376890
    },
    {
      "epoch": 146.37,
      "learning_rate": 4.841423948220065e-07,
      "loss": 0.0855,
      "step": 376900
    },
    {
      "epoch": 146.37,
      "learning_rate": 4.836245954692557e-07,
      "loss": 0.0022,
      "step": 376910
    },
    {
      "epoch": 146.38,
      "learning_rate": 4.831067961165049e-07,
      "loss": 0.0616,
      "step": 376920
    },
    {
      "epoch": 146.38,
      "learning_rate": 4.825889967637541e-07,
      "loss": 0.0231,
      "step": 376930
    },
    {
      "epoch": 146.38,
      "learning_rate": 4.820711974110033e-07,
      "loss": 0.0628,
      "step": 376940
    },
    {
      "epoch": 146.39,
      "learning_rate": 4.815533980582525e-07,
      "loss": 0.0081,
      "step": 376950
    },
    {
      "epoch": 146.39,
      "learning_rate": 4.810355987055017e-07,
      "loss": 0.036,
      "step": 376960
    },
    {
      "epoch": 146.4,
      "learning_rate": 4.805177993527509e-07,
      "loss": 0.0001,
      "step": 376970
    },
    {
      "epoch": 146.4,
      "learning_rate": 4.800000000000001e-07,
      "loss": 0.0087,
      "step": 376980
    },
    {
      "epoch": 146.4,
      "learning_rate": 4.794822006472493e-07,
      "loss": 0.0398,
      "step": 376990
    },
    {
      "epoch": 146.41,
      "learning_rate": 4.789644012944985e-07,
      "loss": 0.1106,
      "step": 377000
    },
    {
      "epoch": 146.41,
      "learning_rate": 4.784466019417476e-07,
      "loss": 0.0553,
      "step": 377010
    },
    {
      "epoch": 146.42,
      "learning_rate": 4.779288025889968e-07,
      "loss": 0.0261,
      "step": 377020
    },
    {
      "epoch": 146.42,
      "learning_rate": 4.774110032362459e-07,
      "loss": 0.0627,
      "step": 377030
    },
    {
      "epoch": 146.42,
      "learning_rate": 4.768932038834951e-07,
      "loss": 0.0461,
      "step": 377040
    },
    {
      "epoch": 146.43,
      "learning_rate": 4.7637540453074437e-07,
      "loss": 0.0635,
      "step": 377050
    },
    {
      "epoch": 146.43,
      "learning_rate": 4.758576051779935e-07,
      "loss": 0.0804,
      "step": 377060
    },
    {
      "epoch": 146.43,
      "learning_rate": 4.753398058252427e-07,
      "loss": 0.0646,
      "step": 377070
    },
    {
      "epoch": 146.44,
      "learning_rate": 4.7482200647249196e-07,
      "loss": 0.0001,
      "step": 377080
    },
    {
      "epoch": 146.44,
      "learning_rate": 4.7430420711974115e-07,
      "loss": 0.0073,
      "step": 377090
    },
    {
      "epoch": 146.45,
      "learning_rate": 4.7378640776699034e-07,
      "loss": 0.0164,
      "step": 377100
    },
    {
      "epoch": 146.45,
      "learning_rate": 4.7326860841423954e-07,
      "loss": 0.0339,
      "step": 377110
    },
    {
      "epoch": 146.45,
      "learning_rate": 4.7275080906148873e-07,
      "loss": 0.0402,
      "step": 377120
    },
    {
      "epoch": 146.46,
      "learning_rate": 4.7223300970873793e-07,
      "loss": 0.0346,
      "step": 377130
    },
    {
      "epoch": 146.46,
      "learning_rate": 4.7171521035598707e-07,
      "loss": 0.0931,
      "step": 377140
    },
    {
      "epoch": 146.47,
      "learning_rate": 4.7119741100323626e-07,
      "loss": 0.0054,
      "step": 377150
    },
    {
      "epoch": 146.47,
      "learning_rate": 4.7067961165048546e-07,
      "loss": 0.0105,
      "step": 377160
    },
    {
      "epoch": 146.47,
      "learning_rate": 4.7016181229773465e-07,
      "loss": 0.0522,
      "step": 377170
    },
    {
      "epoch": 146.48,
      "learning_rate": 4.6964401294498385e-07,
      "loss": 0.0039,
      "step": 377180
    },
    {
      "epoch": 146.48,
      "learning_rate": 4.6912621359223304e-07,
      "loss": 0.0185,
      "step": 377190
    },
    {
      "epoch": 146.49,
      "learning_rate": 4.686084142394822e-07,
      "loss": 0.033,
      "step": 377200
    },
    {
      "epoch": 146.49,
      "learning_rate": 4.680906148867314e-07,
      "loss": 0.0182,
      "step": 377210
    },
    {
      "epoch": 146.49,
      "learning_rate": 4.675728155339806e-07,
      "loss": 0.0032,
      "step": 377220
    },
    {
      "epoch": 146.5,
      "learning_rate": 4.670550161812298e-07,
      "loss": 0.0067,
      "step": 377230
    },
    {
      "epoch": 146.5,
      "learning_rate": 4.66537216828479e-07,
      "loss": 0.0725,
      "step": 377240
    },
    {
      "epoch": 146.5,
      "learning_rate": 4.660194174757282e-07,
      "loss": 0.0084,
      "step": 377250
    },
    {
      "epoch": 146.51,
      "learning_rate": 4.655016181229774e-07,
      "loss": 0.0245,
      "step": 377260
    },
    {
      "epoch": 146.51,
      "learning_rate": 4.649838187702266e-07,
      "loss": 0.0006,
      "step": 377270
    },
    {
      "epoch": 146.52,
      "learning_rate": 4.6446601941747574e-07,
      "loss": 0.0098,
      "step": 377280
    },
    {
      "epoch": 146.52,
      "learning_rate": 4.6394822006472493e-07,
      "loss": 0.0669,
      "step": 377290
    },
    {
      "epoch": 146.52,
      "learning_rate": 4.6343042071197413e-07,
      "loss": 0.0773,
      "step": 377300
    },
    {
      "epoch": 146.53,
      "learning_rate": 4.629126213592233e-07,
      "loss": 0.015,
      "step": 377310
    },
    {
      "epoch": 146.53,
      "learning_rate": 4.623948220064725e-07,
      "loss": 0.022,
      "step": 377320
    },
    {
      "epoch": 146.54,
      "learning_rate": 4.618770226537217e-07,
      "loss": 0.0493,
      "step": 377330
    },
    {
      "epoch": 146.54,
      "learning_rate": 4.613592233009709e-07,
      "loss": 0.1434,
      "step": 377340
    },
    {
      "epoch": 146.54,
      "learning_rate": 4.6084142394822005e-07,
      "loss": 0.0323,
      "step": 377350
    },
    {
      "epoch": 146.55,
      "learning_rate": 4.6032362459546935e-07,
      "loss": 0.0091,
      "step": 377360
    },
    {
      "epoch": 146.55,
      "learning_rate": 4.598058252427185e-07,
      "loss": 0.0573,
      "step": 377370
    },
    {
      "epoch": 146.56,
      "learning_rate": 4.592880258899677e-07,
      "loss": 0.1435,
      "step": 377380
    },
    {
      "epoch": 146.56,
      "learning_rate": 4.587702265372169e-07,
      "loss": 0.0747,
      "step": 377390
    },
    {
      "epoch": 146.56,
      "learning_rate": 4.582524271844661e-07,
      "loss": 0.0002,
      "step": 377400
    },
    {
      "epoch": 146.57,
      "learning_rate": 4.5773462783171527e-07,
      "loss": 0.0288,
      "step": 377410
    },
    {
      "epoch": 146.57,
      "learning_rate": 4.5721682847896446e-07,
      "loss": 0.0437,
      "step": 377420
    },
    {
      "epoch": 146.57,
      "learning_rate": 4.566990291262136e-07,
      "loss": 0.0587,
      "step": 377430
    },
    {
      "epoch": 146.58,
      "learning_rate": 4.561812297734628e-07,
      "loss": 0.0136,
      "step": 377440
    },
    {
      "epoch": 146.58,
      "learning_rate": 4.55663430420712e-07,
      "loss": 0.0527,
      "step": 377450
    },
    {
      "epoch": 146.59,
      "learning_rate": 4.551456310679612e-07,
      "loss": 0.0001,
      "step": 377460
    },
    {
      "epoch": 146.59,
      "learning_rate": 4.546278317152104e-07,
      "loss": 0.0001,
      "step": 377470
    },
    {
      "epoch": 146.59,
      "learning_rate": 4.541100323624596e-07,
      "loss": 0.0882,
      "step": 377480
    },
    {
      "epoch": 146.6,
      "learning_rate": 4.535922330097087e-07,
      "loss": 0.0419,
      "step": 377490
    },
    {
      "epoch": 146.6,
      "learning_rate": 4.53074433656958e-07,
      "loss": 0.0091,
      "step": 377500
    },
    {
      "epoch": 146.61,
      "learning_rate": 4.5255663430420716e-07,
      "loss": 0.0531,
      "step": 377510
    },
    {
      "epoch": 146.61,
      "learning_rate": 4.5203883495145636e-07,
      "loss": 0.0225,
      "step": 377520
    },
    {
      "epoch": 146.61,
      "learning_rate": 4.5152103559870555e-07,
      "loss": 0.0486,
      "step": 377530
    },
    {
      "epoch": 146.62,
      "learning_rate": 4.5100323624595474e-07,
      "loss": 0.0011,
      "step": 377540
    },
    {
      "epoch": 146.62,
      "learning_rate": 4.5048543689320394e-07,
      "loss": 0.0564,
      "step": 377550
    },
    {
      "epoch": 146.63,
      "learning_rate": 4.4996763754045313e-07,
      "loss": 0.0116,
      "step": 377560
    },
    {
      "epoch": 146.63,
      "learning_rate": 4.494498381877023e-07,
      "loss": 0.0287,
      "step": 377570
    },
    {
      "epoch": 146.63,
      "learning_rate": 4.4893203883495147e-07,
      "loss": 0.0082,
      "step": 377580
    },
    {
      "epoch": 146.64,
      "learning_rate": 4.4841423948220066e-07,
      "loss": 0.149,
      "step": 377590
    },
    {
      "epoch": 146.64,
      "learning_rate": 4.4789644012944986e-07,
      "loss": 0.0093,
      "step": 377600
    },
    {
      "epoch": 146.64,
      "learning_rate": 4.4737864077669905e-07,
      "loss": 0.0003,
      "step": 377610
    },
    {
      "epoch": 146.65,
      "learning_rate": 4.4686084142394825e-07,
      "loss": 0.0748,
      "step": 377620
    },
    {
      "epoch": 146.65,
      "learning_rate": 4.463430420711974e-07,
      "loss": 0.0252,
      "step": 377630
    },
    {
      "epoch": 146.66,
      "learning_rate": 4.458252427184467e-07,
      "loss": 0.0631,
      "step": 377640
    },
    {
      "epoch": 146.66,
      "learning_rate": 4.4530744336569583e-07,
      "loss": 0.1011,
      "step": 377650
    },
    {
      "epoch": 146.66,
      "learning_rate": 4.44789644012945e-07,
      "loss": 0.2093,
      "step": 377660
    },
    {
      "epoch": 146.67,
      "learning_rate": 4.442718446601942e-07,
      "loss": 0.041,
      "step": 377670
    },
    {
      "epoch": 146.67,
      "learning_rate": 4.437540453074434e-07,
      "loss": 0.0003,
      "step": 377680
    },
    {
      "epoch": 146.68,
      "learning_rate": 4.432362459546926e-07,
      "loss": 0.036,
      "step": 377690
    },
    {
      "epoch": 146.68,
      "learning_rate": 4.427184466019418e-07,
      "loss": 0.068,
      "step": 377700
    },
    {
      "epoch": 146.68,
      "learning_rate": 4.4220064724919095e-07,
      "loss": 0.0259,
      "step": 377710
    },
    {
      "epoch": 146.69,
      "learning_rate": 4.4168284789644014e-07,
      "loss": 0.0977,
      "step": 377720
    },
    {
      "epoch": 146.69,
      "learning_rate": 4.4116504854368933e-07,
      "loss": 0.0214,
      "step": 377730
    },
    {
      "epoch": 146.7,
      "learning_rate": 4.4064724919093853e-07,
      "loss": 0.0783,
      "step": 377740
    },
    {
      "epoch": 146.7,
      "learning_rate": 4.401294498381877e-07,
      "loss": 0.0028,
      "step": 377750
    },
    {
      "epoch": 146.7,
      "learning_rate": 4.396116504854369e-07,
      "loss": 0.1073,
      "step": 377760
    },
    {
      "epoch": 146.71,
      "learning_rate": 4.3909385113268606e-07,
      "loss": 0.0097,
      "step": 377770
    },
    {
      "epoch": 146.71,
      "learning_rate": 4.3857605177993536e-07,
      "loss": 0.0209,
      "step": 377780
    },
    {
      "epoch": 146.71,
      "learning_rate": 4.380582524271845e-07,
      "loss": 0.0543,
      "step": 377790
    },
    {
      "epoch": 146.72,
      "learning_rate": 4.375404530744337e-07,
      "loss": 0.0965,
      "step": 377800
    },
    {
      "epoch": 146.72,
      "learning_rate": 4.370226537216829e-07,
      "loss": 0.1859,
      "step": 377810
    },
    {
      "epoch": 146.73,
      "learning_rate": 4.365048543689321e-07,
      "loss": 0.0004,
      "step": 377820
    },
    {
      "epoch": 146.73,
      "learning_rate": 4.359870550161813e-07,
      "loss": 0.0001,
      "step": 377830
    },
    {
      "epoch": 146.73,
      "learning_rate": 4.354692556634305e-07,
      "loss": 0.0483,
      "step": 377840
    },
    {
      "epoch": 146.74,
      "learning_rate": 4.3495145631067967e-07,
      "loss": 0.0604,
      "step": 377850
    },
    {
      "epoch": 146.74,
      "learning_rate": 4.344336569579288e-07,
      "loss": 0.009,
      "step": 377860
    },
    {
      "epoch": 146.75,
      "learning_rate": 4.33915857605178e-07,
      "loss": 0.0093,
      "step": 377870
    },
    {
      "epoch": 146.75,
      "learning_rate": 4.333980582524272e-07,
      "loss": 0.0099,
      "step": 377880
    },
    {
      "epoch": 146.75,
      "learning_rate": 4.328802588996764e-07,
      "loss": 0.1267,
      "step": 377890
    },
    {
      "epoch": 146.76,
      "learning_rate": 4.323624595469256e-07,
      "loss": 0.0237,
      "step": 377900
    },
    {
      "epoch": 146.76,
      "learning_rate": 4.318446601941748e-07,
      "loss": 0.0248,
      "step": 377910
    },
    {
      "epoch": 146.77,
      "learning_rate": 4.3132686084142403e-07,
      "loss": 0.0006,
      "step": 377920
    },
    {
      "epoch": 146.77,
      "learning_rate": 4.308090614886732e-07,
      "loss": 0.0006,
      "step": 377930
    },
    {
      "epoch": 146.77,
      "learning_rate": 4.3029126213592237e-07,
      "loss": 0.0494,
      "step": 377940
    },
    {
      "epoch": 146.78,
      "learning_rate": 4.2977346278317156e-07,
      "loss": 0.0174,
      "step": 377950
    },
    {
      "epoch": 146.78,
      "learning_rate": 4.2925566343042076e-07,
      "loss": 0.0013,
      "step": 377960
    },
    {
      "epoch": 146.78,
      "learning_rate": 4.2873786407766995e-07,
      "loss": 0.091,
      "step": 377970
    },
    {
      "epoch": 146.79,
      "learning_rate": 4.2822006472491914e-07,
      "loss": 0.0001,
      "step": 377980
    },
    {
      "epoch": 146.79,
      "learning_rate": 4.2770226537216834e-07,
      "loss": 0.0066,
      "step": 377990
    },
    {
      "epoch": 146.8,
      "learning_rate": 4.271844660194175e-07,
      "loss": 0.0001,
      "step": 378000
    },
    {
      "epoch": 146.8,
      "learning_rate": 4.266666666666667e-07,
      "loss": 0.0412,
      "step": 378010
    },
    {
      "epoch": 146.8,
      "learning_rate": 4.2614886731391587e-07,
      "loss": 0.069,
      "step": 378020
    },
    {
      "epoch": 146.81,
      "learning_rate": 4.2563106796116506e-07,
      "loss": 0.0087,
      "step": 378030
    },
    {
      "epoch": 146.81,
      "learning_rate": 4.2511326860841426e-07,
      "loss": 0.0454,
      "step": 378040
    },
    {
      "epoch": 146.82,
      "learning_rate": 4.2459546925566345e-07,
      "loss": 0.0544,
      "step": 378050
    },
    {
      "epoch": 146.82,
      "learning_rate": 4.240776699029127e-07,
      "loss": 0.0598,
      "step": 378060
    },
    {
      "epoch": 146.82,
      "learning_rate": 4.235598705501619e-07,
      "loss": 0.0148,
      "step": 378070
    },
    {
      "epoch": 146.83,
      "learning_rate": 4.2304207119741104e-07,
      "loss": 0.0026,
      "step": 378080
    },
    {
      "epoch": 146.83,
      "learning_rate": 4.2252427184466023e-07,
      "loss": 0.0003,
      "step": 378090
    },
    {
      "epoch": 146.83,
      "learning_rate": 4.220064724919094e-07,
      "loss": 0.0251,
      "step": 378100
    },
    {
      "epoch": 146.84,
      "learning_rate": 4.214886731391586e-07,
      "loss": 0.0013,
      "step": 378110
    },
    {
      "epoch": 146.84,
      "learning_rate": 4.209708737864078e-07,
      "loss": 0.0085,
      "step": 378120
    },
    {
      "epoch": 146.85,
      "learning_rate": 4.20453074433657e-07,
      "loss": 0.016,
      "step": 378130
    },
    {
      "epoch": 146.85,
      "learning_rate": 4.1993527508090615e-07,
      "loss": 0.0831,
      "step": 378140
    },
    {
      "epoch": 146.85,
      "learning_rate": 4.1941747572815535e-07,
      "loss": 0.0649,
      "step": 378150
    },
    {
      "epoch": 146.86,
      "learning_rate": 4.1889967637540454e-07,
      "loss": 0.017,
      "step": 378160
    },
    {
      "epoch": 146.86,
      "learning_rate": 4.1838187702265374e-07,
      "loss": 0.0462,
      "step": 378170
    },
    {
      "epoch": 146.87,
      "learning_rate": 4.1786407766990293e-07,
      "loss": 0.0207,
      "step": 378180
    },
    {
      "epoch": 146.87,
      "learning_rate": 4.173462783171521e-07,
      "loss": 0.0826,
      "step": 378190
    },
    {
      "epoch": 146.87,
      "learning_rate": 4.1682847896440137e-07,
      "loss": 0.0163,
      "step": 378200
    },
    {
      "epoch": 146.88,
      "learning_rate": 4.1631067961165057e-07,
      "loss": 0.0362,
      "step": 378210
    },
    {
      "epoch": 146.88,
      "learning_rate": 4.157928802588997e-07,
      "loss": 0.1205,
      "step": 378220
    },
    {
      "epoch": 146.89,
      "learning_rate": 4.152750809061489e-07,
      "loss": 0.0238,
      "step": 378230
    },
    {
      "epoch": 146.89,
      "learning_rate": 4.147572815533981e-07,
      "loss": 0.0003,
      "step": 378240
    },
    {
      "epoch": 146.89,
      "learning_rate": 4.142394822006473e-07,
      "loss": 0.0001,
      "step": 378250
    },
    {
      "epoch": 146.9,
      "learning_rate": 4.137216828478965e-07,
      "loss": 0.0611,
      "step": 378260
    },
    {
      "epoch": 146.9,
      "learning_rate": 4.132038834951457e-07,
      "loss": 0.1566,
      "step": 378270
    },
    {
      "epoch": 146.9,
      "learning_rate": 4.126860841423948e-07,
      "loss": 0.2043,
      "step": 378280
    },
    {
      "epoch": 146.91,
      "learning_rate": 4.12168284789644e-07,
      "loss": 0.0352,
      "step": 378290
    },
    {
      "epoch": 146.91,
      "learning_rate": 4.116504854368932e-07,
      "loss": 0.1212,
      "step": 378300
    },
    {
      "epoch": 146.92,
      "learning_rate": 4.111326860841424e-07,
      "loss": 0.0748,
      "step": 378310
    },
    {
      "epoch": 146.92,
      "learning_rate": 4.106148867313916e-07,
      "loss": 0.0163,
      "step": 378320
    },
    {
      "epoch": 146.92,
      "learning_rate": 4.100970873786408e-07,
      "loss": 0.009,
      "step": 378330
    },
    {
      "epoch": 146.93,
      "learning_rate": 4.0957928802589004e-07,
      "loss": 0.1166,
      "step": 378340
    },
    {
      "epoch": 146.93,
      "learning_rate": 4.0906148867313924e-07,
      "loss": 0.0412,
      "step": 378350
    },
    {
      "epoch": 146.94,
      "learning_rate": 4.085436893203884e-07,
      "loss": 0.0014,
      "step": 378360
    },
    {
      "epoch": 146.94,
      "learning_rate": 4.0802588996763757e-07,
      "loss": 0.0269,
      "step": 378370
    },
    {
      "epoch": 146.94,
      "learning_rate": 4.0750809061488677e-07,
      "loss": 0.0718,
      "step": 378380
    },
    {
      "epoch": 146.95,
      "learning_rate": 4.0699029126213596e-07,
      "loss": 0.1028,
      "step": 378390
    },
    {
      "epoch": 146.95,
      "learning_rate": 4.0647249190938516e-07,
      "loss": 0.0003,
      "step": 378400
    },
    {
      "epoch": 146.96,
      "learning_rate": 4.0595469255663435e-07,
      "loss": 0.0333,
      "step": 378410
    },
    {
      "epoch": 146.96,
      "learning_rate": 4.0543689320388355e-07,
      "loss": 0.0644,
      "step": 378420
    },
    {
      "epoch": 146.96,
      "learning_rate": 4.049190938511327e-07,
      "loss": 0.0114,
      "step": 378430
    },
    {
      "epoch": 146.97,
      "learning_rate": 4.044012944983819e-07,
      "loss": 0.0736,
      "step": 378440
    },
    {
      "epoch": 146.97,
      "learning_rate": 4.038834951456311e-07,
      "loss": 0.0388,
      "step": 378450
    },
    {
      "epoch": 146.97,
      "learning_rate": 4.0336569579288027e-07,
      "loss": 0.063,
      "step": 378460
    },
    {
      "epoch": 146.98,
      "learning_rate": 4.0284789644012946e-07,
      "loss": 0.0228,
      "step": 378470
    },
    {
      "epoch": 146.98,
      "learning_rate": 4.023300970873787e-07,
      "loss": 0.0362,
      "step": 378480
    },
    {
      "epoch": 146.99,
      "learning_rate": 4.018122977346279e-07,
      "loss": 0.048,
      "step": 378490
    },
    {
      "epoch": 146.99,
      "learning_rate": 4.012944983818771e-07,
      "loss": 0.01,
      "step": 378500
    },
    {
      "epoch": 146.99,
      "learning_rate": 4.0077669902912624e-07,
      "loss": 0.0228,
      "step": 378510
    },
    {
      "epoch": 147.0,
      "learning_rate": 4.0025889967637544e-07,
      "loss": 0.0006,
      "step": 378520
    },
    {
      "epoch": 147.0,
      "eval_accuracy": 0.9529573590096286,
      "eval_loss": 0.3908064663410187,
      "eval_runtime": 8.3388,
      "eval_samples_per_second": 435.913,
      "eval_steps_per_second": 54.564,
      "step": 378525
    },
    {
      "epoch": 147.0,
      "learning_rate": 3.9974110032362463e-07,
      "loss": 0.0045,
      "step": 378530
    },
    {
      "epoch": 147.01,
      "learning_rate": 3.9922330097087383e-07,
      "loss": 0.011,
      "step": 378540
    },
    {
      "epoch": 147.01,
      "learning_rate": 3.98705501618123e-07,
      "loss": 0.1995,
      "step": 378550
    },
    {
      "epoch": 147.01,
      "learning_rate": 3.981877022653722e-07,
      "loss": 0.0307,
      "step": 378560
    },
    {
      "epoch": 147.02,
      "learning_rate": 3.9766990291262136e-07,
      "loss": 0.1278,
      "step": 378570
    },
    {
      "epoch": 147.02,
      "learning_rate": 3.9715210355987055e-07,
      "loss": 0.0226,
      "step": 378580
    },
    {
      "epoch": 147.03,
      "learning_rate": 3.9663430420711975e-07,
      "loss": 0.0592,
      "step": 378590
    },
    {
      "epoch": 147.03,
      "learning_rate": 3.9611650485436894e-07,
      "loss": 0.0611,
      "step": 378600
    },
    {
      "epoch": 147.03,
      "learning_rate": 3.9559870550161814e-07,
      "loss": 0.0165,
      "step": 378610
    },
    {
      "epoch": 147.04,
      "learning_rate": 3.950809061488674e-07,
      "loss": 0.1016,
      "step": 378620
    },
    {
      "epoch": 147.04,
      "learning_rate": 3.945631067961166e-07,
      "loss": 0.0138,
      "step": 378630
    },
    {
      "epoch": 147.04,
      "learning_rate": 3.9404530744336577e-07,
      "loss": 0.1123,
      "step": 378640
    },
    {
      "epoch": 147.05,
      "learning_rate": 3.935275080906149e-07,
      "loss": 0.0004,
      "step": 378650
    },
    {
      "epoch": 147.05,
      "learning_rate": 3.930097087378641e-07,
      "loss": 0.0002,
      "step": 378660
    },
    {
      "epoch": 147.06,
      "learning_rate": 3.924919093851133e-07,
      "loss": 0.0314,
      "step": 378670
    },
    {
      "epoch": 147.06,
      "learning_rate": 3.919741100323625e-07,
      "loss": 0.0025,
      "step": 378680
    },
    {
      "epoch": 147.06,
      "learning_rate": 3.914563106796117e-07,
      "loss": 0.0007,
      "step": 378690
    },
    {
      "epoch": 147.07,
      "learning_rate": 3.909385113268609e-07,
      "loss": 0.0242,
      "step": 378700
    },
    {
      "epoch": 147.07,
      "learning_rate": 3.9042071197411003e-07,
      "loss": 0.0309,
      "step": 378710
    },
    {
      "epoch": 147.08,
      "learning_rate": 3.899029126213592e-07,
      "loss": 0.0866,
      "step": 378720
    },
    {
      "epoch": 147.08,
      "learning_rate": 3.893851132686084e-07,
      "loss": 0.0098,
      "step": 378730
    },
    {
      "epoch": 147.08,
      "learning_rate": 3.888673139158576e-07,
      "loss": 0.0084,
      "step": 378740
    },
    {
      "epoch": 147.09,
      "learning_rate": 3.883495145631068e-07,
      "loss": 0.1268,
      "step": 378750
    },
    {
      "epoch": 147.09,
      "learning_rate": 3.8783171521035605e-07,
      "loss": 0.0004,
      "step": 378760
    },
    {
      "epoch": 147.1,
      "learning_rate": 3.8731391585760525e-07,
      "loss": 0.0625,
      "step": 378770
    },
    {
      "epoch": 147.1,
      "learning_rate": 3.8679611650485444e-07,
      "loss": 0.0202,
      "step": 378780
    },
    {
      "epoch": 147.1,
      "learning_rate": 3.862783171521036e-07,
      "loss": 0.0182,
      "step": 378790
    },
    {
      "epoch": 147.11,
      "learning_rate": 3.857605177993528e-07,
      "loss": 0.0414,
      "step": 378800
    },
    {
      "epoch": 147.11,
      "learning_rate": 3.8524271844660197e-07,
      "loss": 0.0006,
      "step": 378810
    },
    {
      "epoch": 147.11,
      "learning_rate": 3.8472491909385117e-07,
      "loss": 0.0184,
      "step": 378820
    },
    {
      "epoch": 147.12,
      "learning_rate": 3.8420711974110036e-07,
      "loss": 0.0717,
      "step": 378830
    },
    {
      "epoch": 147.12,
      "learning_rate": 3.8368932038834956e-07,
      "loss": 0.0001,
      "step": 378840
    },
    {
      "epoch": 147.13,
      "learning_rate": 3.831715210355987e-07,
      "loss": 0.0499,
      "step": 378850
    },
    {
      "epoch": 147.13,
      "learning_rate": 3.826537216828479e-07,
      "loss": 0.0002,
      "step": 378860
    },
    {
      "epoch": 147.13,
      "learning_rate": 3.821359223300971e-07,
      "loss": 0.0296,
      "step": 378870
    },
    {
      "epoch": 147.14,
      "learning_rate": 3.816181229773463e-07,
      "loss": 0.0222,
      "step": 378880
    },
    {
      "epoch": 147.14,
      "learning_rate": 3.811003236245955e-07,
      "loss": 0.0926,
      "step": 378890
    },
    {
      "epoch": 147.15,
      "learning_rate": 3.805825242718447e-07,
      "loss": 0.1034,
      "step": 378900
    },
    {
      "epoch": 147.15,
      "learning_rate": 3.800647249190939e-07,
      "loss": 0.0001,
      "step": 378910
    },
    {
      "epoch": 147.15,
      "learning_rate": 3.795469255663431e-07,
      "loss": 0.0016,
      "step": 378920
    },
    {
      "epoch": 147.16,
      "learning_rate": 3.7902912621359225e-07,
      "loss": 0.0086,
      "step": 378930
    },
    {
      "epoch": 147.16,
      "learning_rate": 3.7851132686084145e-07,
      "loss": 0.0827,
      "step": 378940
    },
    {
      "epoch": 147.17,
      "learning_rate": 3.7799352750809064e-07,
      "loss": 0.0068,
      "step": 378950
    },
    {
      "epoch": 147.17,
      "learning_rate": 3.7747572815533984e-07,
      "loss": 0.0091,
      "step": 378960
    },
    {
      "epoch": 147.17,
      "learning_rate": 3.7695792880258903e-07,
      "loss": 0.0076,
      "step": 378970
    },
    {
      "epoch": 147.18,
      "learning_rate": 3.7644012944983823e-07,
      "loss": 0.0033,
      "step": 378980
    },
    {
      "epoch": 147.18,
      "learning_rate": 3.759223300970874e-07,
      "loss": 0.0105,
      "step": 378990
    },
    {
      "epoch": 147.18,
      "learning_rate": 3.7540453074433656e-07,
      "loss": 0.0703,
      "step": 379000
    },
    {
      "epoch": 147.19,
      "learning_rate": 3.7488673139158576e-07,
      "loss": 0.0247,
      "step": 379010
    },
    {
      "epoch": 147.19,
      "learning_rate": 3.7436893203883495e-07,
      "loss": 0.0002,
      "step": 379020
    },
    {
      "epoch": 147.2,
      "learning_rate": 3.7385113268608415e-07,
      "loss": 0.0294,
      "step": 379030
    },
    {
      "epoch": 147.2,
      "learning_rate": 3.733333333333334e-07,
      "loss": 0.018,
      "step": 379040
    },
    {
      "epoch": 147.2,
      "learning_rate": 3.728155339805826e-07,
      "loss": 0.1245,
      "step": 379050
    },
    {
      "epoch": 147.21,
      "learning_rate": 3.722977346278318e-07,
      "loss": 0.0683,
      "step": 379060
    },
    {
      "epoch": 147.21,
      "learning_rate": 3.71779935275081e-07,
      "loss": 0.1195,
      "step": 379070
    },
    {
      "epoch": 147.22,
      "learning_rate": 3.712621359223301e-07,
      "loss": 0.1072,
      "step": 379080
    },
    {
      "epoch": 147.22,
      "learning_rate": 3.707443365695793e-07,
      "loss": 0.0022,
      "step": 379090
    },
    {
      "epoch": 147.22,
      "learning_rate": 3.702265372168285e-07,
      "loss": 0.0181,
      "step": 379100
    },
    {
      "epoch": 147.23,
      "learning_rate": 3.697087378640777e-07,
      "loss": 0.0268,
      "step": 379110
    },
    {
      "epoch": 147.23,
      "learning_rate": 3.691909385113269e-07,
      "loss": 0.0613,
      "step": 379120
    },
    {
      "epoch": 147.23,
      "learning_rate": 3.686731391585761e-07,
      "loss": 0.0798,
      "step": 379130
    },
    {
      "epoch": 147.24,
      "learning_rate": 3.6815533980582523e-07,
      "loss": 0.0011,
      "step": 379140
    },
    {
      "epoch": 147.24,
      "learning_rate": 3.6763754045307443e-07,
      "loss": 0.0006,
      "step": 379150
    },
    {
      "epoch": 147.25,
      "learning_rate": 3.671197411003236e-07,
      "loss": 0.0275,
      "step": 379160
    },
    {
      "epoch": 147.25,
      "learning_rate": 3.666019417475728e-07,
      "loss": 0.1002,
      "step": 379170
    },
    {
      "epoch": 147.25,
      "learning_rate": 3.6608414239482206e-07,
      "loss": 0.0112,
      "step": 379180
    },
    {
      "epoch": 147.26,
      "learning_rate": 3.6556634304207126e-07,
      "loss": 0.0696,
      "step": 379190
    },
    {
      "epoch": 147.26,
      "learning_rate": 3.6504854368932045e-07,
      "loss": 0.0004,
      "step": 379200
    },
    {
      "epoch": 147.27,
      "learning_rate": 3.6453074433656965e-07,
      "loss": 0.0077,
      "step": 379210
    },
    {
      "epoch": 147.27,
      "learning_rate": 3.640129449838188e-07,
      "loss": 0.037,
      "step": 379220
    },
    {
      "epoch": 147.27,
      "learning_rate": 3.63495145631068e-07,
      "loss": 0.0144,
      "step": 379230
    },
    {
      "epoch": 147.28,
      "learning_rate": 3.629773462783172e-07,
      "loss": 0.1069,
      "step": 379240
    },
    {
      "epoch": 147.28,
      "learning_rate": 3.6245954692556637e-07,
      "loss": 0.0865,
      "step": 379250
    },
    {
      "epoch": 147.29,
      "learning_rate": 3.6194174757281557e-07,
      "loss": 0.0009,
      "step": 379260
    },
    {
      "epoch": 147.29,
      "learning_rate": 3.6142394822006476e-07,
      "loss": 0.0519,
      "step": 379270
    },
    {
      "epoch": 147.29,
      "learning_rate": 3.609061488673139e-07,
      "loss": 0.0099,
      "step": 379280
    },
    {
      "epoch": 147.3,
      "learning_rate": 3.603883495145631e-07,
      "loss": 0.0585,
      "step": 379290
    },
    {
      "epoch": 147.3,
      "learning_rate": 3.598705501618123e-07,
      "loss": 0.053,
      "step": 379300
    },
    {
      "epoch": 147.3,
      "learning_rate": 3.593527508090615e-07,
      "loss": 0.0163,
      "step": 379310
    },
    {
      "epoch": 147.31,
      "learning_rate": 3.5883495145631073e-07,
      "loss": 0.0427,
      "step": 379320
    },
    {
      "epoch": 147.31,
      "learning_rate": 3.5831715210355993e-07,
      "loss": 0.0034,
      "step": 379330
    },
    {
      "epoch": 147.32,
      "learning_rate": 3.577993527508091e-07,
      "loss": 0.1451,
      "step": 379340
    },
    {
      "epoch": 147.32,
      "learning_rate": 3.572815533980583e-07,
      "loss": 0.0469,
      "step": 379350
    },
    {
      "epoch": 147.32,
      "learning_rate": 3.5676375404530746e-07,
      "loss": 0.0149,
      "step": 379360
    },
    {
      "epoch": 147.33,
      "learning_rate": 3.5624595469255665e-07,
      "loss": 0.0174,
      "step": 379370
    },
    {
      "epoch": 147.33,
      "learning_rate": 3.5572815533980585e-07,
      "loss": 0.0636,
      "step": 379380
    },
    {
      "epoch": 147.34,
      "learning_rate": 3.5521035598705504e-07,
      "loss": 0.0412,
      "step": 379390
    },
    {
      "epoch": 147.34,
      "learning_rate": 3.5469255663430424e-07,
      "loss": 0.0467,
      "step": 379400
    },
    {
      "epoch": 147.34,
      "learning_rate": 3.5417475728155343e-07,
      "loss": 0.002,
      "step": 379410
    },
    {
      "epoch": 147.35,
      "learning_rate": 3.536569579288026e-07,
      "loss": 0.0529,
      "step": 379420
    },
    {
      "epoch": 147.35,
      "learning_rate": 3.5313915857605177e-07,
      "loss": 0.0424,
      "step": 379430
    },
    {
      "epoch": 147.36,
      "learning_rate": 3.5262135922330096e-07,
      "loss": 0.0167,
      "step": 379440
    },
    {
      "epoch": 147.36,
      "learning_rate": 3.5210355987055016e-07,
      "loss": 0.0008,
      "step": 379450
    },
    {
      "epoch": 147.36,
      "learning_rate": 3.515857605177994e-07,
      "loss": 0.0004,
      "step": 379460
    },
    {
      "epoch": 147.37,
      "learning_rate": 3.510679611650486e-07,
      "loss": 0.059,
      "step": 379470
    },
    {
      "epoch": 147.37,
      "learning_rate": 3.505501618122978e-07,
      "loss": 0.0002,
      "step": 379480
    },
    {
      "epoch": 147.37,
      "learning_rate": 3.50032362459547e-07,
      "loss": 0.0432,
      "step": 379490
    },
    {
      "epoch": 147.38,
      "learning_rate": 3.4951456310679613e-07,
      "loss": 0.0694,
      "step": 379500
    },
    {
      "epoch": 147.38,
      "learning_rate": 3.489967637540453e-07,
      "loss": 0.0003,
      "step": 379510
    },
    {
      "epoch": 147.39,
      "learning_rate": 3.484789644012945e-07,
      "loss": 0.0003,
      "step": 379520
    },
    {
      "epoch": 147.39,
      "learning_rate": 3.479611650485437e-07,
      "loss": 0.0231,
      "step": 379530
    },
    {
      "epoch": 147.39,
      "learning_rate": 3.474433656957929e-07,
      "loss": 0.0002,
      "step": 379540
    },
    {
      "epoch": 147.4,
      "learning_rate": 3.469255663430421e-07,
      "loss": 0.0708,
      "step": 379550
    },
    {
      "epoch": 147.4,
      "learning_rate": 3.464077669902913e-07,
      "loss": 0.0003,
      "step": 379560
    },
    {
      "epoch": 147.41,
      "learning_rate": 3.4588996763754044e-07,
      "loss": 0.109,
      "step": 379570
    },
    {
      "epoch": 147.41,
      "learning_rate": 3.4537216828478963e-07,
      "loss": 0.0651,
      "step": 379580
    },
    {
      "epoch": 147.41,
      "learning_rate": 3.4485436893203883e-07,
      "loss": 0.1316,
      "step": 379590
    },
    {
      "epoch": 147.42,
      "learning_rate": 3.443365695792881e-07,
      "loss": 0.0138,
      "step": 379600
    },
    {
      "epoch": 147.42,
      "learning_rate": 3.4381877022653727e-07,
      "loss": 0.059,
      "step": 379610
    },
    {
      "epoch": 147.43,
      "learning_rate": 3.4330097087378646e-07,
      "loss": 0.0264,
      "step": 379620
    },
    {
      "epoch": 147.43,
      "learning_rate": 3.4278317152103566e-07,
      "loss": 0.0676,
      "step": 379630
    },
    {
      "epoch": 147.43,
      "learning_rate": 3.4226537216828485e-07,
      "loss": 0.0497,
      "step": 379640
    },
    {
      "epoch": 147.44,
      "learning_rate": 3.41747572815534e-07,
      "loss": 0.0102,
      "step": 379650
    },
    {
      "epoch": 147.44,
      "learning_rate": 3.412297734627832e-07,
      "loss": 0.019,
      "step": 379660
    },
    {
      "epoch": 147.44,
      "learning_rate": 3.407119741100324e-07,
      "loss": 0.0001,
      "step": 379670
    },
    {
      "epoch": 147.45,
      "learning_rate": 3.401941747572816e-07,
      "loss": 0.0241,
      "step": 379680
    },
    {
      "epoch": 147.45,
      "learning_rate": 3.396763754045308e-07,
      "loss": 0.0564,
      "step": 379690
    },
    {
      "epoch": 147.46,
      "learning_rate": 3.3915857605177997e-07,
      "loss": 0.0002,
      "step": 379700
    },
    {
      "epoch": 147.46,
      "learning_rate": 3.386407766990291e-07,
      "loss": 0.0671,
      "step": 379710
    },
    {
      "epoch": 147.46,
      "learning_rate": 3.381229773462783e-07,
      "loss": 0.0419,
      "step": 379720
    },
    {
      "epoch": 147.47,
      "learning_rate": 3.376051779935275e-07,
      "loss": 0.0215,
      "step": 379730
    },
    {
      "epoch": 147.47,
      "learning_rate": 3.3708737864077675e-07,
      "loss": 0.0112,
      "step": 379740
    },
    {
      "epoch": 147.48,
      "learning_rate": 3.3656957928802594e-07,
      "loss": 0.0173,
      "step": 379750
    },
    {
      "epoch": 147.48,
      "learning_rate": 3.3605177993527514e-07,
      "loss": 0.002,
      "step": 379760
    },
    {
      "epoch": 147.48,
      "learning_rate": 3.3553398058252433e-07,
      "loss": 0.0716,
      "step": 379770
    },
    {
      "epoch": 147.49,
      "learning_rate": 3.350161812297735e-07,
      "loss": 0.0166,
      "step": 379780
    },
    {
      "epoch": 147.49,
      "learning_rate": 3.3449838187702267e-07,
      "loss": 0.058,
      "step": 379790
    },
    {
      "epoch": 147.5,
      "learning_rate": 3.3398058252427186e-07,
      "loss": 0.1164,
      "step": 379800
    },
    {
      "epoch": 147.5,
      "learning_rate": 3.3346278317152105e-07,
      "loss": 0.0654,
      "step": 379810
    },
    {
      "epoch": 147.5,
      "learning_rate": 3.3294498381877025e-07,
      "loss": 0.0126,
      "step": 379820
    },
    {
      "epoch": 147.51,
      "learning_rate": 3.3242718446601944e-07,
      "loss": 0.0093,
      "step": 379830
    },
    {
      "epoch": 147.51,
      "learning_rate": 3.3190938511326864e-07,
      "loss": 0.0603,
      "step": 379840
    },
    {
      "epoch": 147.51,
      "learning_rate": 3.313915857605178e-07,
      "loss": 0.0174,
      "step": 379850
    },
    {
      "epoch": 147.52,
      "learning_rate": 3.30873786407767e-07,
      "loss": 0.0816,
      "step": 379860
    },
    {
      "epoch": 147.52,
      "learning_rate": 3.3035598705501617e-07,
      "loss": 0.0081,
      "step": 379870
    },
    {
      "epoch": 147.53,
      "learning_rate": 3.298381877022654e-07,
      "loss": 0.0052,
      "step": 379880
    },
    {
      "epoch": 147.53,
      "learning_rate": 3.293203883495146e-07,
      "loss": 0.0856,
      "step": 379890
    },
    {
      "epoch": 147.53,
      "learning_rate": 3.288025889967638e-07,
      "loss": 0.0581,
      "step": 379900
    },
    {
      "epoch": 147.54,
      "learning_rate": 3.28284789644013e-07,
      "loss": 0.1151,
      "step": 379910
    },
    {
      "epoch": 147.54,
      "learning_rate": 3.277669902912622e-07,
      "loss": 0.0251,
      "step": 379920
    },
    {
      "epoch": 147.55,
      "learning_rate": 3.2724919093851134e-07,
      "loss": 0.0384,
      "step": 379930
    },
    {
      "epoch": 147.55,
      "learning_rate": 3.2673139158576053e-07,
      "loss": 0.014,
      "step": 379940
    },
    {
      "epoch": 147.55,
      "learning_rate": 3.262135922330097e-07,
      "loss": 0.0615,
      "step": 379950
    },
    {
      "epoch": 147.56,
      "learning_rate": 3.256957928802589e-07,
      "loss": 0.0523,
      "step": 379960
    },
    {
      "epoch": 147.56,
      "learning_rate": 3.251779935275081e-07,
      "loss": 0.1021,
      "step": 379970
    },
    {
      "epoch": 147.57,
      "learning_rate": 3.246601941747573e-07,
      "loss": 0.0452,
      "step": 379980
    },
    {
      "epoch": 147.57,
      "learning_rate": 3.2414239482200645e-07,
      "loss": 0.07,
      "step": 379990
    },
    {
      "epoch": 147.57,
      "learning_rate": 3.2362459546925565e-07,
      "loss": 0.0258,
      "step": 380000
    },
    {
      "epoch": 147.58,
      "learning_rate": 3.2310679611650484e-07,
      "loss": 0.0666,
      "step": 380010
    },
    {
      "epoch": 147.58,
      "learning_rate": 3.225889967637541e-07,
      "loss": 0.0581,
      "step": 380020
    },
    {
      "epoch": 147.58,
      "learning_rate": 3.220711974110033e-07,
      "loss": 0.0005,
      "step": 380030
    },
    {
      "epoch": 147.59,
      "learning_rate": 3.215533980582525e-07,
      "loss": 0.13,
      "step": 380040
    },
    {
      "epoch": 147.59,
      "learning_rate": 3.2103559870550167e-07,
      "loss": 0.0737,
      "step": 380050
    },
    {
      "epoch": 147.6,
      "learning_rate": 3.2051779935275087e-07,
      "loss": 0.107,
      "step": 380060
    },
    {
      "epoch": 147.6,
      "learning_rate": 3.2e-07,
      "loss": 0.001,
      "step": 380070
    },
    {
      "epoch": 147.6,
      "learning_rate": 3.194822006472492e-07,
      "loss": 0.0024,
      "step": 380080
    },
    {
      "epoch": 147.61,
      "learning_rate": 3.189644012944984e-07,
      "loss": 0.0002,
      "step": 380090
    },
    {
      "epoch": 147.61,
      "learning_rate": 3.184466019417476e-07,
      "loss": 0.0278,
      "step": 380100
    },
    {
      "epoch": 147.62,
      "learning_rate": 3.179288025889968e-07,
      "loss": 0.179,
      "step": 380110
    },
    {
      "epoch": 147.62,
      "learning_rate": 3.17411003236246e-07,
      "loss": 0.0197,
      "step": 380120
    },
    {
      "epoch": 147.62,
      "learning_rate": 3.168932038834952e-07,
      "loss": 0.0384,
      "step": 380130
    },
    {
      "epoch": 147.63,
      "learning_rate": 3.163754045307443e-07,
      "loss": 0.0002,
      "step": 380140
    },
    {
      "epoch": 147.63,
      "learning_rate": 3.158576051779935e-07,
      "loss": 0.0558,
      "step": 380150
    },
    {
      "epoch": 147.63,
      "learning_rate": 3.1533980582524276e-07,
      "loss": 0.0001,
      "step": 380160
    },
    {
      "epoch": 147.64,
      "learning_rate": 3.1482200647249195e-07,
      "loss": 0.0983,
      "step": 380170
    },
    {
      "epoch": 147.64,
      "learning_rate": 3.1430420711974115e-07,
      "loss": 0.1535,
      "step": 380180
    },
    {
      "epoch": 147.65,
      "learning_rate": 3.1378640776699034e-07,
      "loss": 0.0104,
      "step": 380190
    },
    {
      "epoch": 147.65,
      "learning_rate": 3.1326860841423954e-07,
      "loss": 0.0098,
      "step": 380200
    },
    {
      "epoch": 147.65,
      "learning_rate": 3.1275080906148873e-07,
      "loss": 0.0039,
      "step": 380210
    },
    {
      "epoch": 147.66,
      "learning_rate": 3.1223300970873787e-07,
      "loss": 0.0645,
      "step": 380220
    },
    {
      "epoch": 147.66,
      "learning_rate": 3.1171521035598707e-07,
      "loss": 0.0452,
      "step": 380230
    },
    {
      "epoch": 147.67,
      "learning_rate": 3.1119741100323626e-07,
      "loss": 0.0388,
      "step": 380240
    },
    {
      "epoch": 147.67,
      "learning_rate": 3.1067961165048546e-07,
      "loss": 0.128,
      "step": 380250
    },
    {
      "epoch": 147.67,
      "learning_rate": 3.1016181229773465e-07,
      "loss": 0.024,
      "step": 380260
    },
    {
      "epoch": 147.68,
      "learning_rate": 3.0964401294498384e-07,
      "loss": 0.056,
      "step": 380270
    },
    {
      "epoch": 147.68,
      "learning_rate": 3.0912621359223304e-07,
      "loss": 0.0196,
      "step": 380280
    },
    {
      "epoch": 147.69,
      "learning_rate": 3.0860841423948223e-07,
      "loss": 0.0164,
      "step": 380290
    },
    {
      "epoch": 147.69,
      "learning_rate": 3.0809061488673143e-07,
      "loss": 0.1051,
      "step": 380300
    },
    {
      "epoch": 147.69,
      "learning_rate": 3.075728155339806e-07,
      "loss": 0.0015,
      "step": 380310
    },
    {
      "epoch": 147.7,
      "learning_rate": 3.0705501618122976e-07,
      "loss": 0.0005,
      "step": 380320
    },
    {
      "epoch": 147.7,
      "learning_rate": 3.06537216828479e-07,
      "loss": 0.1047,
      "step": 380330
    },
    {
      "epoch": 147.7,
      "learning_rate": 3.060194174757282e-07,
      "loss": 0.0672,
      "step": 380340
    },
    {
      "epoch": 147.71,
      "learning_rate": 3.055016181229774e-07,
      "loss": 0.0566,
      "step": 380350
    },
    {
      "epoch": 147.71,
      "learning_rate": 3.0498381877022654e-07,
      "loss": 0.0049,
      "step": 380360
    },
    {
      "epoch": 147.72,
      "learning_rate": 3.0446601941747574e-07,
      "loss": 0.0215,
      "step": 380370
    },
    {
      "epoch": 147.72,
      "learning_rate": 3.0394822006472493e-07,
      "loss": 0.0979,
      "step": 380380
    },
    {
      "epoch": 147.72,
      "learning_rate": 3.034304207119741e-07,
      "loss": 0.0389,
      "step": 380390
    },
    {
      "epoch": 147.73,
      "learning_rate": 3.029126213592233e-07,
      "loss": 0.0663,
      "step": 380400
    },
    {
      "epoch": 147.73,
      "learning_rate": 3.023948220064725e-07,
      "loss": 0.031,
      "step": 380410
    },
    {
      "epoch": 147.74,
      "learning_rate": 3.018770226537217e-07,
      "loss": 0.0002,
      "step": 380420
    },
    {
      "epoch": 147.74,
      "learning_rate": 3.013592233009709e-07,
      "loss": 0.0346,
      "step": 380430
    },
    {
      "epoch": 147.74,
      "learning_rate": 3.008414239482201e-07,
      "loss": 0.1019,
      "step": 380440
    },
    {
      "epoch": 147.75,
      "learning_rate": 3.003236245954693e-07,
      "loss": 0.0007,
      "step": 380450
    },
    {
      "epoch": 147.75,
      "learning_rate": 2.9980582524271843e-07,
      "loss": 0.0003,
      "step": 380460
    },
    {
      "epoch": 147.76,
      "learning_rate": 2.992880258899677e-07,
      "loss": 0.0001,
      "step": 380470
    },
    {
      "epoch": 147.76,
      "learning_rate": 2.987702265372169e-07,
      "loss": 0.0316,
      "step": 380480
    },
    {
      "epoch": 147.76,
      "learning_rate": 2.9825242718446607e-07,
      "loss": 0.0163,
      "step": 380490
    },
    {
      "epoch": 147.77,
      "learning_rate": 2.977346278317152e-07,
      "loss": 0.0252,
      "step": 380500
    },
    {
      "epoch": 147.77,
      "learning_rate": 2.972168284789644e-07,
      "loss": 0.0169,
      "step": 380510
    },
    {
      "epoch": 147.77,
      "learning_rate": 2.966990291262136e-07,
      "loss": 0.0027,
      "step": 380520
    },
    {
      "epoch": 147.78,
      "learning_rate": 2.961812297734628e-07,
      "loss": 0.014,
      "step": 380530
    },
    {
      "epoch": 147.78,
      "learning_rate": 2.95663430420712e-07,
      "loss": 0.0454,
      "step": 380540
    },
    {
      "epoch": 147.79,
      "learning_rate": 2.951456310679612e-07,
      "loss": 0.1332,
      "step": 380550
    },
    {
      "epoch": 147.79,
      "learning_rate": 2.946278317152104e-07,
      "loss": 0.0558,
      "step": 380560
    },
    {
      "epoch": 147.79,
      "learning_rate": 2.941100323624596e-07,
      "loss": 0.0453,
      "step": 380570
    },
    {
      "epoch": 147.8,
      "learning_rate": 2.9359223300970877e-07,
      "loss": 0.0072,
      "step": 380580
    },
    {
      "epoch": 147.8,
      "learning_rate": 2.9307443365695796e-07,
      "loss": 0.012,
      "step": 380590
    },
    {
      "epoch": 147.81,
      "learning_rate": 2.925566343042071e-07,
      "loss": 0.0198,
      "step": 380600
    },
    {
      "epoch": 147.81,
      "learning_rate": 2.9203883495145635e-07,
      "loss": 0.0943,
      "step": 380610
    },
    {
      "epoch": 147.81,
      "learning_rate": 2.9152103559870555e-07,
      "loss": 0.029,
      "step": 380620
    },
    {
      "epoch": 147.82,
      "learning_rate": 2.9100323624595474e-07,
      "loss": 0.0731,
      "step": 380630
    },
    {
      "epoch": 147.82,
      "learning_rate": 2.904854368932039e-07,
      "loss": 0.0008,
      "step": 380640
    },
    {
      "epoch": 147.83,
      "learning_rate": 2.899676375404531e-07,
      "loss": 0.0088,
      "step": 380650
    },
    {
      "epoch": 147.83,
      "learning_rate": 2.8944983818770227e-07,
      "loss": 0.1174,
      "step": 380660
    },
    {
      "epoch": 147.83,
      "learning_rate": 2.8893203883495147e-07,
      "loss": 0.0127,
      "step": 380670
    },
    {
      "epoch": 147.84,
      "learning_rate": 2.884142394822007e-07,
      "loss": 0.0565,
      "step": 380680
    },
    {
      "epoch": 147.84,
      "learning_rate": 2.8789644012944986e-07,
      "loss": 0.1208,
      "step": 380690
    },
    {
      "epoch": 147.84,
      "learning_rate": 2.8737864077669905e-07,
      "loss": 0.0415,
      "step": 380700
    },
    {
      "epoch": 147.85,
      "learning_rate": 2.8686084142394824e-07,
      "loss": 0.049,
      "step": 380710
    },
    {
      "epoch": 147.85,
      "learning_rate": 2.8634304207119744e-07,
      "loss": 0.027,
      "step": 380720
    },
    {
      "epoch": 147.86,
      "learning_rate": 2.8582524271844663e-07,
      "loss": 0.0001,
      "step": 380730
    },
    {
      "epoch": 147.86,
      "learning_rate": 2.8530744336569583e-07,
      "loss": 0.0044,
      "step": 380740
    },
    {
      "epoch": 147.86,
      "learning_rate": 2.84789644012945e-07,
      "loss": 0.0342,
      "step": 380750
    },
    {
      "epoch": 147.87,
      "learning_rate": 2.842718446601942e-07,
      "loss": 0.0006,
      "step": 380760
    },
    {
      "epoch": 147.87,
      "learning_rate": 2.837540453074434e-07,
      "loss": 0.0332,
      "step": 380770
    },
    {
      "epoch": 147.88,
      "learning_rate": 2.832362459546926e-07,
      "loss": 0.0017,
      "step": 380780
    },
    {
      "epoch": 147.88,
      "learning_rate": 2.8271844660194175e-07,
      "loss": 0.0782,
      "step": 380790
    },
    {
      "epoch": 147.88,
      "learning_rate": 2.8220064724919094e-07,
      "loss": 0.0321,
      "step": 380800
    },
    {
      "epoch": 147.89,
      "learning_rate": 2.8168284789644014e-07,
      "loss": 0.095,
      "step": 380810
    },
    {
      "epoch": 147.89,
      "learning_rate": 2.811650485436894e-07,
      "loss": 0.0004,
      "step": 380820
    },
    {
      "epoch": 147.9,
      "learning_rate": 2.806472491909385e-07,
      "loss": 0.0845,
      "step": 380830
    },
    {
      "epoch": 147.9,
      "learning_rate": 2.801294498381877e-07,
      "loss": 0.0089,
      "step": 380840
    },
    {
      "epoch": 147.9,
      "learning_rate": 2.796116504854369e-07,
      "loss": 0.0137,
      "step": 380850
    },
    {
      "epoch": 147.91,
      "learning_rate": 2.790938511326861e-07,
      "loss": 0.1203,
      "step": 380860
    },
    {
      "epoch": 147.91,
      "learning_rate": 2.785760517799353e-07,
      "loss": 0.0153,
      "step": 380870
    },
    {
      "epoch": 147.91,
      "learning_rate": 2.780582524271845e-07,
      "loss": 0.0166,
      "step": 380880
    },
    {
      "epoch": 147.92,
      "learning_rate": 2.775404530744337e-07,
      "loss": 0.0209,
      "step": 380890
    },
    {
      "epoch": 147.92,
      "learning_rate": 2.770226537216829e-07,
      "loss": 0.0012,
      "step": 380900
    },
    {
      "epoch": 147.93,
      "learning_rate": 2.765048543689321e-07,
      "loss": 0.2369,
      "step": 380910
    },
    {
      "epoch": 147.93,
      "learning_rate": 2.759870550161813e-07,
      "loss": 0.0001,
      "step": 380920
    },
    {
      "epoch": 147.93,
      "learning_rate": 2.754692556634304e-07,
      "loss": 0.0004,
      "step": 380930
    },
    {
      "epoch": 147.94,
      "learning_rate": 2.749514563106796e-07,
      "loss": 0.0253,
      "step": 380940
    },
    {
      "epoch": 147.94,
      "learning_rate": 2.744336569579288e-07,
      "loss": 0.0315,
      "step": 380950
    },
    {
      "epoch": 147.95,
      "learning_rate": 2.7391585760517805e-07,
      "loss": 0.0769,
      "step": 380960
    },
    {
      "epoch": 147.95,
      "learning_rate": 2.733980582524272e-07,
      "loss": 0.1122,
      "step": 380970
    },
    {
      "epoch": 147.95,
      "learning_rate": 2.728802588996764e-07,
      "loss": 0.0276,
      "step": 380980
    },
    {
      "epoch": 147.96,
      "learning_rate": 2.723624595469256e-07,
      "loss": 0.0138,
      "step": 380990
    },
    {
      "epoch": 147.96,
      "learning_rate": 2.718446601941748e-07,
      "loss": 0.0003,
      "step": 381000
    },
    {
      "epoch": 147.97,
      "learning_rate": 2.71326860841424e-07,
      "loss": 0.0616,
      "step": 381010
    },
    {
      "epoch": 147.97,
      "learning_rate": 2.7080906148867317e-07,
      "loss": 0.0491,
      "step": 381020
    },
    {
      "epoch": 147.97,
      "learning_rate": 2.7029126213592236e-07,
      "loss": 0.024,
      "step": 381030
    },
    {
      "epoch": 147.98,
      "learning_rate": 2.6977346278317156e-07,
      "loss": 0.0583,
      "step": 381040
    },
    {
      "epoch": 147.98,
      "learning_rate": 2.6925566343042075e-07,
      "loss": 0.0882,
      "step": 381050
    },
    {
      "epoch": 147.98,
      "learning_rate": 2.6873786407766995e-07,
      "loss": 0.0512,
      "step": 381060
    },
    {
      "epoch": 147.99,
      "learning_rate": 2.682200647249191e-07,
      "loss": 0.0046,
      "step": 381070
    },
    {
      "epoch": 147.99,
      "learning_rate": 2.677022653721683e-07,
      "loss": 0.0411,
      "step": 381080
    },
    {
      "epoch": 148.0,
      "learning_rate": 2.671844660194175e-07,
      "loss": 0.082,
      "step": 381090
    },
    {
      "epoch": 148.0,
      "learning_rate": 2.666666666666667e-07,
      "loss": 0.0226,
      "step": 381100
    },
    {
      "epoch": 148.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.38993731141090393,
      "eval_runtime": 8.3283,
      "eval_samples_per_second": 436.465,
      "eval_steps_per_second": 54.633,
      "step": 381100
    },
    {
      "epoch": 148.0,
      "learning_rate": 2.6614886731391587e-07,
      "loss": 0.0181,
      "step": 381110
    },
    {
      "epoch": 148.01,
      "learning_rate": 2.6563106796116506e-07,
      "loss": 0.0486,
      "step": 381120
    },
    {
      "epoch": 148.01,
      "learning_rate": 2.6511326860841426e-07,
      "loss": 0.0245,
      "step": 381130
    },
    {
      "epoch": 148.02,
      "learning_rate": 2.6459546925566345e-07,
      "loss": 0.0749,
      "step": 381140
    },
    {
      "epoch": 148.02,
      "learning_rate": 2.6407766990291264e-07,
      "loss": 0.0853,
      "step": 381150
    },
    {
      "epoch": 148.02,
      "learning_rate": 2.6355987055016184e-07,
      "loss": 0.089,
      "step": 381160
    },
    {
      "epoch": 148.03,
      "learning_rate": 2.6304207119741103e-07,
      "loss": 0.016,
      "step": 381170
    },
    {
      "epoch": 148.03,
      "learning_rate": 2.6252427184466023e-07,
      "loss": 0.0264,
      "step": 381180
    },
    {
      "epoch": 148.03,
      "learning_rate": 2.620064724919094e-07,
      "loss": 0.0088,
      "step": 381190
    },
    {
      "epoch": 148.04,
      "learning_rate": 2.614886731391586e-07,
      "loss": 0.0151,
      "step": 381200
    },
    {
      "epoch": 148.04,
      "learning_rate": 2.609708737864078e-07,
      "loss": 0.0002,
      "step": 381210
    },
    {
      "epoch": 148.05,
      "learning_rate": 2.6045307443365695e-07,
      "loss": 0.0089,
      "step": 381220
    },
    {
      "epoch": 148.05,
      "learning_rate": 2.5993527508090615e-07,
      "loss": 0.0021,
      "step": 381230
    },
    {
      "epoch": 148.05,
      "learning_rate": 2.594174757281554e-07,
      "loss": 0.0141,
      "step": 381240
    },
    {
      "epoch": 148.06,
      "learning_rate": 2.588996763754046e-07,
      "loss": 0.1158,
      "step": 381250
    },
    {
      "epoch": 148.06,
      "learning_rate": 2.5838187702265373e-07,
      "loss": 0.0001,
      "step": 381260
    },
    {
      "epoch": 148.07,
      "learning_rate": 2.578640776699029e-07,
      "loss": 0.0525,
      "step": 381270
    },
    {
      "epoch": 148.07,
      "learning_rate": 2.573462783171521e-07,
      "loss": 0.0004,
      "step": 381280
    },
    {
      "epoch": 148.07,
      "learning_rate": 2.568284789644013e-07,
      "loss": 0.0318,
      "step": 381290
    },
    {
      "epoch": 148.08,
      "learning_rate": 2.563106796116505e-07,
      "loss": 0.0205,
      "step": 381300
    },
    {
      "epoch": 148.08,
      "learning_rate": 2.557928802588997e-07,
      "loss": 0.0492,
      "step": 381310
    },
    {
      "epoch": 148.09,
      "learning_rate": 2.552750809061489e-07,
      "loss": 0.0224,
      "step": 381320
    },
    {
      "epoch": 148.09,
      "learning_rate": 2.547572815533981e-07,
      "loss": 0.0089,
      "step": 381330
    },
    {
      "epoch": 148.09,
      "learning_rate": 2.542394822006473e-07,
      "loss": 0.0091,
      "step": 381340
    },
    {
      "epoch": 148.1,
      "learning_rate": 2.537216828478965e-07,
      "loss": 0.0022,
      "step": 381350
    },
    {
      "epoch": 148.1,
      "learning_rate": 2.532038834951456e-07,
      "loss": 0.0251,
      "step": 381360
    },
    {
      "epoch": 148.1,
      "learning_rate": 2.526860841423948e-07,
      "loss": 0.0098,
      "step": 381370
    },
    {
      "epoch": 148.11,
      "learning_rate": 2.5216828478964407e-07,
      "loss": 0.0565,
      "step": 381380
    },
    {
      "epoch": 148.11,
      "learning_rate": 2.5165048543689326e-07,
      "loss": 0.0444,
      "step": 381390
    },
    {
      "epoch": 148.12,
      "learning_rate": 2.511326860841424e-07,
      "loss": 0.0139,
      "step": 381400
    },
    {
      "epoch": 148.12,
      "learning_rate": 2.506148867313916e-07,
      "loss": 0.0683,
      "step": 381410
    },
    {
      "epoch": 148.12,
      "learning_rate": 2.500970873786408e-07,
      "loss": 0.0752,
      "step": 381420
    },
    {
      "epoch": 148.13,
      "learning_rate": 2.4957928802589e-07,
      "loss": 0.0094,
      "step": 381430
    },
    {
      "epoch": 148.13,
      "learning_rate": 2.490614886731392e-07,
      "loss": 0.0169,
      "step": 381440
    },
    {
      "epoch": 148.14,
      "learning_rate": 2.485436893203884e-07,
      "loss": 0.1743,
      "step": 381450
    },
    {
      "epoch": 148.14,
      "learning_rate": 2.4802588996763757e-07,
      "loss": 0.0002,
      "step": 381460
    },
    {
      "epoch": 148.14,
      "learning_rate": 2.4750809061488676e-07,
      "loss": 0.0085,
      "step": 381470
    },
    {
      "epoch": 148.15,
      "learning_rate": 2.4699029126213596e-07,
      "loss": 0.1197,
      "step": 381480
    },
    {
      "epoch": 148.15,
      "learning_rate": 2.4647249190938515e-07,
      "loss": 0.0002,
      "step": 381490
    },
    {
      "epoch": 148.16,
      "learning_rate": 2.459546925566343e-07,
      "loss": 0.0156,
      "step": 381500
    },
    {
      "epoch": 148.16,
      "learning_rate": 2.454368932038835e-07,
      "loss": 0.1005,
      "step": 381510
    },
    {
      "epoch": 148.16,
      "learning_rate": 2.4491909385113274e-07,
      "loss": 0.0002,
      "step": 381520
    },
    {
      "epoch": 148.17,
      "learning_rate": 2.4440129449838193e-07,
      "loss": 0.01,
      "step": 381530
    },
    {
      "epoch": 148.17,
      "learning_rate": 2.4388349514563107e-07,
      "loss": 0.0092,
      "step": 381540
    },
    {
      "epoch": 148.17,
      "learning_rate": 2.4336569579288027e-07,
      "loss": 0.0171,
      "step": 381550
    },
    {
      "epoch": 148.18,
      "learning_rate": 2.4284789644012946e-07,
      "loss": 0.0145,
      "step": 381560
    },
    {
      "epoch": 148.18,
      "learning_rate": 2.4233009708737866e-07,
      "loss": 0.1396,
      "step": 381570
    },
    {
      "epoch": 148.19,
      "learning_rate": 2.4181229773462785e-07,
      "loss": 0.0403,
      "step": 381580
    },
    {
      "epoch": 148.19,
      "learning_rate": 2.4129449838187705e-07,
      "loss": 0.0336,
      "step": 381590
    },
    {
      "epoch": 148.19,
      "learning_rate": 2.4077669902912624e-07,
      "loss": 0.0015,
      "step": 381600
    },
    {
      "epoch": 148.2,
      "learning_rate": 2.4025889967637543e-07,
      "loss": 0.0209,
      "step": 381610
    },
    {
      "epoch": 148.2,
      "learning_rate": 2.3974110032362463e-07,
      "loss": 0.0128,
      "step": 381620
    },
    {
      "epoch": 148.21,
      "learning_rate": 2.392233009708738e-07,
      "loss": 0.1755,
      "step": 381630
    },
    {
      "epoch": 148.21,
      "learning_rate": 2.3870550161812296e-07,
      "loss": 0.0804,
      "step": 381640
    },
    {
      "epoch": 148.21,
      "learning_rate": 2.3818770226537219e-07,
      "loss": 0.0611,
      "step": 381650
    },
    {
      "epoch": 148.22,
      "learning_rate": 2.3766990291262135e-07,
      "loss": 0.0425,
      "step": 381660
    },
    {
      "epoch": 148.22,
      "learning_rate": 2.3715210355987057e-07,
      "loss": 0.079,
      "step": 381670
    },
    {
      "epoch": 148.23,
      "learning_rate": 2.3663430420711977e-07,
      "loss": 0.0147,
      "step": 381680
    },
    {
      "epoch": 148.23,
      "learning_rate": 2.3611650485436896e-07,
      "loss": 0.0003,
      "step": 381690
    },
    {
      "epoch": 148.23,
      "learning_rate": 2.3559870550161813e-07,
      "loss": 0.0005,
      "step": 381700
    },
    {
      "epoch": 148.24,
      "learning_rate": 2.3508090614886733e-07,
      "loss": 0.0001,
      "step": 381710
    },
    {
      "epoch": 148.24,
      "learning_rate": 2.3456310679611652e-07,
      "loss": 0.0144,
      "step": 381720
    },
    {
      "epoch": 148.24,
      "learning_rate": 2.340453074433657e-07,
      "loss": 0.0006,
      "step": 381730
    },
    {
      "epoch": 148.25,
      "learning_rate": 2.335275080906149e-07,
      "loss": 0.0144,
      "step": 381740
    },
    {
      "epoch": 148.25,
      "learning_rate": 2.330097087378641e-07,
      "loss": 0.0082,
      "step": 381750
    },
    {
      "epoch": 148.26,
      "learning_rate": 2.324919093851133e-07,
      "loss": 0.0001,
      "step": 381760
    },
    {
      "epoch": 148.26,
      "learning_rate": 2.3197411003236247e-07,
      "loss": 0.0187,
      "step": 381770
    },
    {
      "epoch": 148.26,
      "learning_rate": 2.3145631067961166e-07,
      "loss": 0.0592,
      "step": 381780
    },
    {
      "epoch": 148.27,
      "learning_rate": 2.3093851132686086e-07,
      "loss": 0.0956,
      "step": 381790
    },
    {
      "epoch": 148.27,
      "learning_rate": 2.3042071197411002e-07,
      "loss": 0.0458,
      "step": 381800
    },
    {
      "epoch": 148.28,
      "learning_rate": 2.2990291262135925e-07,
      "loss": 0.0276,
      "step": 381810
    },
    {
      "epoch": 148.28,
      "learning_rate": 2.2938511326860844e-07,
      "loss": 0.0729,
      "step": 381820
    },
    {
      "epoch": 148.28,
      "learning_rate": 2.2886731391585763e-07,
      "loss": 0.0774,
      "step": 381830
    },
    {
      "epoch": 148.29,
      "learning_rate": 2.283495145631068e-07,
      "loss": 0.0017,
      "step": 381840
    },
    {
      "epoch": 148.29,
      "learning_rate": 2.27831715210356e-07,
      "loss": 0.064,
      "step": 381850
    },
    {
      "epoch": 148.3,
      "learning_rate": 2.273139158576052e-07,
      "loss": 0.0316,
      "step": 381860
    },
    {
      "epoch": 148.3,
      "learning_rate": 2.2679611650485436e-07,
      "loss": 0.0003,
      "step": 381870
    },
    {
      "epoch": 148.3,
      "learning_rate": 2.2627831715210358e-07,
      "loss": 0.0206,
      "step": 381880
    },
    {
      "epoch": 148.31,
      "learning_rate": 2.2576051779935278e-07,
      "loss": 0.0167,
      "step": 381890
    },
    {
      "epoch": 148.31,
      "learning_rate": 2.2524271844660197e-07,
      "loss": 0.0008,
      "step": 381900
    },
    {
      "epoch": 148.31,
      "learning_rate": 2.2472491909385114e-07,
      "loss": 0.0105,
      "step": 381910
    },
    {
      "epoch": 148.32,
      "learning_rate": 2.2420711974110033e-07,
      "loss": 0.0689,
      "step": 381920
    },
    {
      "epoch": 148.32,
      "learning_rate": 2.2368932038834953e-07,
      "loss": 0.0018,
      "step": 381930
    },
    {
      "epoch": 148.33,
      "learning_rate": 2.231715210355987e-07,
      "loss": 0.0462,
      "step": 381940
    },
    {
      "epoch": 148.33,
      "learning_rate": 2.2265372168284792e-07,
      "loss": 0.0084,
      "step": 381950
    },
    {
      "epoch": 148.33,
      "learning_rate": 2.221359223300971e-07,
      "loss": 0.0139,
      "step": 381960
    },
    {
      "epoch": 148.34,
      "learning_rate": 2.216181229773463e-07,
      "loss": 0.0026,
      "step": 381970
    },
    {
      "epoch": 148.34,
      "learning_rate": 2.2110032362459547e-07,
      "loss": 0.0074,
      "step": 381980
    },
    {
      "epoch": 148.35,
      "learning_rate": 2.2058252427184467e-07,
      "loss": 0.0006,
      "step": 381990
    },
    {
      "epoch": 148.35,
      "learning_rate": 2.2006472491909386e-07,
      "loss": 0.0324,
      "step": 382000
    },
    {
      "epoch": 148.35,
      "learning_rate": 2.1954692556634303e-07,
      "loss": 0.0121,
      "step": 382010
    },
    {
      "epoch": 148.36,
      "learning_rate": 2.1902912621359225e-07,
      "loss": 0.0786,
      "step": 382020
    },
    {
      "epoch": 148.36,
      "learning_rate": 2.1851132686084145e-07,
      "loss": 0.0245,
      "step": 382030
    },
    {
      "epoch": 148.37,
      "learning_rate": 2.1799352750809064e-07,
      "loss": 0.0389,
      "step": 382040
    },
    {
      "epoch": 148.37,
      "learning_rate": 2.1747572815533983e-07,
      "loss": 0.0758,
      "step": 382050
    },
    {
      "epoch": 148.37,
      "learning_rate": 2.16957928802589e-07,
      "loss": 0.0096,
      "step": 382060
    },
    {
      "epoch": 148.38,
      "learning_rate": 2.164401294498382e-07,
      "loss": 0.0107,
      "step": 382070
    },
    {
      "epoch": 148.38,
      "learning_rate": 2.159223300970874e-07,
      "loss": 0.0742,
      "step": 382080
    },
    {
      "epoch": 148.38,
      "learning_rate": 2.154045307443366e-07,
      "loss": 0.1248,
      "step": 382090
    },
    {
      "epoch": 148.39,
      "learning_rate": 2.1488673139158578e-07,
      "loss": 0.0189,
      "step": 382100
    },
    {
      "epoch": 148.39,
      "learning_rate": 2.1436893203883498e-07,
      "loss": 0.1885,
      "step": 382110
    },
    {
      "epoch": 148.4,
      "learning_rate": 2.1385113268608417e-07,
      "loss": 0.0098,
      "step": 382120
    },
    {
      "epoch": 148.4,
      "learning_rate": 2.1333333333333334e-07,
      "loss": 0.0089,
      "step": 382130
    },
    {
      "epoch": 148.4,
      "learning_rate": 2.1281553398058253e-07,
      "loss": 0.0076,
      "step": 382140
    },
    {
      "epoch": 148.41,
      "learning_rate": 2.1229773462783173e-07,
      "loss": 0.0675,
      "step": 382150
    },
    {
      "epoch": 148.41,
      "learning_rate": 2.1177993527508095e-07,
      "loss": 0.0841,
      "step": 382160
    },
    {
      "epoch": 148.42,
      "learning_rate": 2.1126213592233012e-07,
      "loss": 0.0697,
      "step": 382170
    },
    {
      "epoch": 148.42,
      "learning_rate": 2.107443365695793e-07,
      "loss": 0.0565,
      "step": 382180
    },
    {
      "epoch": 148.42,
      "learning_rate": 2.102265372168285e-07,
      "loss": 0.0007,
      "step": 382190
    },
    {
      "epoch": 148.43,
      "learning_rate": 2.0970873786407767e-07,
      "loss": 0.0011,
      "step": 382200
    },
    {
      "epoch": 148.43,
      "learning_rate": 2.0919093851132687e-07,
      "loss": 0.0093,
      "step": 382210
    },
    {
      "epoch": 148.43,
      "learning_rate": 2.0867313915857606e-07,
      "loss": 0.0303,
      "step": 382220
    },
    {
      "epoch": 148.44,
      "learning_rate": 2.0815533980582528e-07,
      "loss": 0.0189,
      "step": 382230
    },
    {
      "epoch": 148.44,
      "learning_rate": 2.0763754045307445e-07,
      "loss": 0.0215,
      "step": 382240
    },
    {
      "epoch": 148.45,
      "learning_rate": 2.0711974110032365e-07,
      "loss": 0.0296,
      "step": 382250
    },
    {
      "epoch": 148.45,
      "learning_rate": 2.0660194174757284e-07,
      "loss": 0.0462,
      "step": 382260
    },
    {
      "epoch": 148.45,
      "learning_rate": 2.06084142394822e-07,
      "loss": 0.0205,
      "step": 382270
    },
    {
      "epoch": 148.46,
      "learning_rate": 2.055663430420712e-07,
      "loss": 0.0523,
      "step": 382280
    },
    {
      "epoch": 148.46,
      "learning_rate": 2.050485436893204e-07,
      "loss": 0.006,
      "step": 382290
    },
    {
      "epoch": 148.47,
      "learning_rate": 2.0453074433656962e-07,
      "loss": 0.0964,
      "step": 382300
    },
    {
      "epoch": 148.47,
      "learning_rate": 2.0401294498381879e-07,
      "loss": 0.0002,
      "step": 382310
    },
    {
      "epoch": 148.47,
      "learning_rate": 2.0349514563106798e-07,
      "loss": 0.0992,
      "step": 382320
    },
    {
      "epoch": 148.48,
      "learning_rate": 2.0297734627831718e-07,
      "loss": 0.0007,
      "step": 382330
    },
    {
      "epoch": 148.48,
      "learning_rate": 2.0245954692556634e-07,
      "loss": 0.1296,
      "step": 382340
    },
    {
      "epoch": 148.49,
      "learning_rate": 2.0194174757281554e-07,
      "loss": 0.0038,
      "step": 382350
    },
    {
      "epoch": 148.49,
      "learning_rate": 2.0142394822006473e-07,
      "loss": 0.0286,
      "step": 382360
    },
    {
      "epoch": 148.49,
      "learning_rate": 2.0090614886731395e-07,
      "loss": 0.0093,
      "step": 382370
    },
    {
      "epoch": 148.5,
      "learning_rate": 2.0038834951456312e-07,
      "loss": 0.096,
      "step": 382380
    },
    {
      "epoch": 148.5,
      "learning_rate": 1.9987055016181232e-07,
      "loss": 0.0242,
      "step": 382390
    },
    {
      "epoch": 148.5,
      "learning_rate": 1.993527508090615e-07,
      "loss": 0.0189,
      "step": 382400
    },
    {
      "epoch": 148.51,
      "learning_rate": 1.9883495145631068e-07,
      "loss": 0.0589,
      "step": 382410
    },
    {
      "epoch": 148.51,
      "learning_rate": 1.9831715210355987e-07,
      "loss": 0.0676,
      "step": 382420
    },
    {
      "epoch": 148.52,
      "learning_rate": 1.9779935275080907e-07,
      "loss": 0.0209,
      "step": 382430
    },
    {
      "epoch": 148.52,
      "learning_rate": 1.972815533980583e-07,
      "loss": 0.0202,
      "step": 382440
    },
    {
      "epoch": 148.52,
      "learning_rate": 1.9676375404530746e-07,
      "loss": 0.0002,
      "step": 382450
    },
    {
      "epoch": 148.53,
      "learning_rate": 1.9624595469255665e-07,
      "loss": 0.0242,
      "step": 382460
    },
    {
      "epoch": 148.53,
      "learning_rate": 1.9572815533980585e-07,
      "loss": 0.0258,
      "step": 382470
    },
    {
      "epoch": 148.54,
      "learning_rate": 1.9521035598705501e-07,
      "loss": 0.1719,
      "step": 382480
    },
    {
      "epoch": 148.54,
      "learning_rate": 1.946925566343042e-07,
      "loss": 0.0001,
      "step": 382490
    },
    {
      "epoch": 148.54,
      "learning_rate": 1.941747572815534e-07,
      "loss": 0.0093,
      "step": 382500
    },
    {
      "epoch": 148.55,
      "learning_rate": 1.9365695792880262e-07,
      "loss": 0.078,
      "step": 382510
    },
    {
      "epoch": 148.55,
      "learning_rate": 1.931391585760518e-07,
      "loss": 0.0002,
      "step": 382520
    },
    {
      "epoch": 148.56,
      "learning_rate": 1.9262135922330099e-07,
      "loss": 0.0179,
      "step": 382530
    },
    {
      "epoch": 148.56,
      "learning_rate": 1.9210355987055018e-07,
      "loss": 0.0141,
      "step": 382540
    },
    {
      "epoch": 148.56,
      "learning_rate": 1.9158576051779935e-07,
      "loss": 0.0005,
      "step": 382550
    },
    {
      "epoch": 148.57,
      "learning_rate": 1.9106796116504854e-07,
      "loss": 0.0316,
      "step": 382560
    },
    {
      "epoch": 148.57,
      "learning_rate": 1.9055016181229774e-07,
      "loss": 0.0864,
      "step": 382570
    },
    {
      "epoch": 148.57,
      "learning_rate": 1.9003236245954696e-07,
      "loss": 0.0976,
      "step": 382580
    },
    {
      "epoch": 148.58,
      "learning_rate": 1.8951456310679613e-07,
      "loss": 0.0093,
      "step": 382590
    },
    {
      "epoch": 148.58,
      "learning_rate": 1.8899676375404532e-07,
      "loss": 0.0595,
      "step": 382600
    },
    {
      "epoch": 148.59,
      "learning_rate": 1.8847896440129452e-07,
      "loss": 0.1075,
      "step": 382610
    },
    {
      "epoch": 148.59,
      "learning_rate": 1.879611650485437e-07,
      "loss": 0.019,
      "step": 382620
    },
    {
      "epoch": 148.59,
      "learning_rate": 1.8744336569579288e-07,
      "loss": 0.064,
      "step": 382630
    },
    {
      "epoch": 148.6,
      "learning_rate": 1.8692556634304207e-07,
      "loss": 0.02,
      "step": 382640
    },
    {
      "epoch": 148.6,
      "learning_rate": 1.864077669902913e-07,
      "loss": 0.0001,
      "step": 382650
    },
    {
      "epoch": 148.61,
      "learning_rate": 1.858899676375405e-07,
      "loss": 0.0687,
      "step": 382660
    },
    {
      "epoch": 148.61,
      "learning_rate": 1.8537216828478966e-07,
      "loss": 0.0196,
      "step": 382670
    },
    {
      "epoch": 148.61,
      "learning_rate": 1.8485436893203885e-07,
      "loss": 0.0533,
      "step": 382680
    },
    {
      "epoch": 148.62,
      "learning_rate": 1.8433656957928805e-07,
      "loss": 0.0171,
      "step": 382690
    },
    {
      "epoch": 148.62,
      "learning_rate": 1.8381877022653721e-07,
      "loss": 0.0206,
      "step": 382700
    },
    {
      "epoch": 148.63,
      "learning_rate": 1.833009708737864e-07,
      "loss": 0.0002,
      "step": 382710
    },
    {
      "epoch": 148.63,
      "learning_rate": 1.8278317152103563e-07,
      "loss": 0.0528,
      "step": 382720
    },
    {
      "epoch": 148.63,
      "learning_rate": 1.8226537216828482e-07,
      "loss": 0.0976,
      "step": 382730
    },
    {
      "epoch": 148.64,
      "learning_rate": 1.81747572815534e-07,
      "loss": 0.0373,
      "step": 382740
    },
    {
      "epoch": 148.64,
      "learning_rate": 1.8122977346278319e-07,
      "loss": 0.0041,
      "step": 382750
    },
    {
      "epoch": 148.64,
      "learning_rate": 1.8071197411003238e-07,
      "loss": 0.008,
      "step": 382760
    },
    {
      "epoch": 148.65,
      "learning_rate": 1.8019417475728155e-07,
      "loss": 0.0583,
      "step": 382770
    },
    {
      "epoch": 148.65,
      "learning_rate": 1.7967637540453074e-07,
      "loss": 0.0013,
      "step": 382780
    },
    {
      "epoch": 148.66,
      "learning_rate": 1.7915857605177996e-07,
      "loss": 0.0003,
      "step": 382790
    },
    {
      "epoch": 148.66,
      "learning_rate": 1.7864077669902916e-07,
      "loss": 0.0191,
      "step": 382800
    },
    {
      "epoch": 148.66,
      "learning_rate": 1.7812297734627833e-07,
      "loss": 0.0003,
      "step": 382810
    },
    {
      "epoch": 148.67,
      "learning_rate": 1.7760517799352752e-07,
      "loss": 0.0262,
      "step": 382820
    },
    {
      "epoch": 148.67,
      "learning_rate": 1.7708737864077672e-07,
      "loss": 0.0561,
      "step": 382830
    },
    {
      "epoch": 148.68,
      "learning_rate": 1.7656957928802588e-07,
      "loss": 0.0004,
      "step": 382840
    },
    {
      "epoch": 148.68,
      "learning_rate": 1.7605177993527508e-07,
      "loss": 0.0523,
      "step": 382850
    },
    {
      "epoch": 148.68,
      "learning_rate": 1.755339805825243e-07,
      "loss": 0.0018,
      "step": 382860
    },
    {
      "epoch": 148.69,
      "learning_rate": 1.750161812297735e-07,
      "loss": 0.1185,
      "step": 382870
    },
    {
      "epoch": 148.69,
      "learning_rate": 1.7449838187702266e-07,
      "loss": 0.0004,
      "step": 382880
    },
    {
      "epoch": 148.7,
      "learning_rate": 1.7398058252427186e-07,
      "loss": 0.0086,
      "step": 382890
    },
    {
      "epoch": 148.7,
      "learning_rate": 1.7346278317152105e-07,
      "loss": 0.0994,
      "step": 382900
    },
    {
      "epoch": 148.7,
      "learning_rate": 1.7294498381877022e-07,
      "loss": 0.049,
      "step": 382910
    },
    {
      "epoch": 148.71,
      "learning_rate": 1.7242718446601941e-07,
      "loss": 0.0157,
      "step": 382920
    },
    {
      "epoch": 148.71,
      "learning_rate": 1.7190938511326864e-07,
      "loss": 0.1754,
      "step": 382930
    },
    {
      "epoch": 148.71,
      "learning_rate": 1.7139158576051783e-07,
      "loss": 0.0018,
      "step": 382940
    },
    {
      "epoch": 148.72,
      "learning_rate": 1.70873786407767e-07,
      "loss": 0.0006,
      "step": 382950
    },
    {
      "epoch": 148.72,
      "learning_rate": 1.703559870550162e-07,
      "loss": 0.0004,
      "step": 382960
    },
    {
      "epoch": 148.73,
      "learning_rate": 1.698381877022654e-07,
      "loss": 0.0663,
      "step": 382970
    },
    {
      "epoch": 148.73,
      "learning_rate": 1.6932038834951455e-07,
      "loss": 0.0083,
      "step": 382980
    },
    {
      "epoch": 148.73,
      "learning_rate": 1.6880258899676375e-07,
      "loss": 0.0176,
      "step": 382990
    },
    {
      "epoch": 148.74,
      "learning_rate": 1.6828478964401297e-07,
      "loss": 0.001,
      "step": 383000
    },
    {
      "epoch": 148.74,
      "learning_rate": 1.6776699029126216e-07,
      "loss": 0.0995,
      "step": 383010
    },
    {
      "epoch": 148.75,
      "learning_rate": 1.6724919093851133e-07,
      "loss": 0.0127,
      "step": 383020
    },
    {
      "epoch": 148.75,
      "learning_rate": 1.6673139158576053e-07,
      "loss": 0.1334,
      "step": 383030
    },
    {
      "epoch": 148.75,
      "learning_rate": 1.6621359223300972e-07,
      "loss": 0.0061,
      "step": 383040
    },
    {
      "epoch": 148.76,
      "learning_rate": 1.656957928802589e-07,
      "loss": 0.0115,
      "step": 383050
    },
    {
      "epoch": 148.76,
      "learning_rate": 1.6517799352750808e-07,
      "loss": 0.0948,
      "step": 383060
    },
    {
      "epoch": 148.77,
      "learning_rate": 1.646601941747573e-07,
      "loss": 0.0196,
      "step": 383070
    },
    {
      "epoch": 148.77,
      "learning_rate": 1.641423948220065e-07,
      "loss": 0.0234,
      "step": 383080
    },
    {
      "epoch": 148.77,
      "learning_rate": 1.6362459546925567e-07,
      "loss": 0.0668,
      "step": 383090
    },
    {
      "epoch": 148.78,
      "learning_rate": 1.6310679611650486e-07,
      "loss": 0.0667,
      "step": 383100
    },
    {
      "epoch": 148.78,
      "learning_rate": 1.6258899676375406e-07,
      "loss": 0.0426,
      "step": 383110
    },
    {
      "epoch": 148.78,
      "learning_rate": 1.6207119741100323e-07,
      "loss": 0.1047,
      "step": 383120
    },
    {
      "epoch": 148.79,
      "learning_rate": 1.6155339805825242e-07,
      "loss": 0.0021,
      "step": 383130
    },
    {
      "epoch": 148.79,
      "learning_rate": 1.6103559870550164e-07,
      "loss": 0.0086,
      "step": 383140
    },
    {
      "epoch": 148.8,
      "learning_rate": 1.6051779935275084e-07,
      "loss": 0.1403,
      "step": 383150
    },
    {
      "epoch": 148.8,
      "learning_rate": 1.6e-07,
      "loss": 0.032,
      "step": 383160
    },
    {
      "epoch": 148.8,
      "learning_rate": 1.594822006472492e-07,
      "loss": 0.0073,
      "step": 383170
    },
    {
      "epoch": 148.81,
      "learning_rate": 1.589644012944984e-07,
      "loss": 0.0005,
      "step": 383180
    },
    {
      "epoch": 148.81,
      "learning_rate": 1.584466019417476e-07,
      "loss": 0.0004,
      "step": 383190
    },
    {
      "epoch": 148.82,
      "learning_rate": 1.5792880258899675e-07,
      "loss": 0.1269,
      "step": 383200
    },
    {
      "epoch": 148.82,
      "learning_rate": 1.5741100323624598e-07,
      "loss": 0.0881,
      "step": 383210
    },
    {
      "epoch": 148.82,
      "learning_rate": 1.5689320388349517e-07,
      "loss": 0.0052,
      "step": 383220
    },
    {
      "epoch": 148.83,
      "learning_rate": 1.5637540453074437e-07,
      "loss": 0.1057,
      "step": 383230
    },
    {
      "epoch": 148.83,
      "learning_rate": 1.5585760517799353e-07,
      "loss": 0.0481,
      "step": 383240
    },
    {
      "epoch": 148.83,
      "learning_rate": 1.5533980582524273e-07,
      "loss": 0.0001,
      "step": 383250
    },
    {
      "epoch": 148.84,
      "learning_rate": 1.5482200647249192e-07,
      "loss": 0.0004,
      "step": 383260
    },
    {
      "epoch": 148.84,
      "learning_rate": 1.5430420711974112e-07,
      "loss": 0.0384,
      "step": 383270
    },
    {
      "epoch": 148.85,
      "learning_rate": 1.537864077669903e-07,
      "loss": 0.046,
      "step": 383280
    },
    {
      "epoch": 148.85,
      "learning_rate": 1.532686084142395e-07,
      "loss": 0.0151,
      "step": 383290
    },
    {
      "epoch": 148.85,
      "learning_rate": 1.527508090614887e-07,
      "loss": 0.0043,
      "step": 383300
    },
    {
      "epoch": 148.86,
      "learning_rate": 1.5223300970873787e-07,
      "loss": 0.0268,
      "step": 383310
    },
    {
      "epoch": 148.86,
      "learning_rate": 1.5171521035598706e-07,
      "loss": 0.0002,
      "step": 383320
    },
    {
      "epoch": 148.87,
      "learning_rate": 1.5119741100323626e-07,
      "loss": 0.0474,
      "step": 383330
    },
    {
      "epoch": 148.87,
      "learning_rate": 1.5067961165048545e-07,
      "loss": 0.0045,
      "step": 383340
    },
    {
      "epoch": 148.87,
      "learning_rate": 1.5016181229773465e-07,
      "loss": 0.0001,
      "step": 383350
    },
    {
      "epoch": 148.88,
      "learning_rate": 1.4964401294498384e-07,
      "loss": 0.0082,
      "step": 383360
    },
    {
      "epoch": 148.88,
      "learning_rate": 1.4912621359223304e-07,
      "loss": 0.1111,
      "step": 383370
    },
    {
      "epoch": 148.89,
      "learning_rate": 1.486084142394822e-07,
      "loss": 0.0223,
      "step": 383380
    },
    {
      "epoch": 148.89,
      "learning_rate": 1.480906148867314e-07,
      "loss": 0.0164,
      "step": 383390
    },
    {
      "epoch": 148.89,
      "learning_rate": 1.475728155339806e-07,
      "loss": 0.0011,
      "step": 383400
    },
    {
      "epoch": 148.9,
      "learning_rate": 1.470550161812298e-07,
      "loss": 0.0467,
      "step": 383410
    },
    {
      "epoch": 148.9,
      "learning_rate": 1.4653721682847898e-07,
      "loss": 0.0174,
      "step": 383420
    },
    {
      "epoch": 148.9,
      "learning_rate": 1.4601941747572818e-07,
      "loss": 0.0906,
      "step": 383430
    },
    {
      "epoch": 148.91,
      "learning_rate": 1.4550161812297737e-07,
      "loss": 0.0042,
      "step": 383440
    },
    {
      "epoch": 148.91,
      "learning_rate": 1.4498381877022654e-07,
      "loss": 0.0113,
      "step": 383450
    },
    {
      "epoch": 148.92,
      "learning_rate": 1.4446601941747573e-07,
      "loss": 0.0667,
      "step": 383460
    },
    {
      "epoch": 148.92,
      "learning_rate": 1.4394822006472493e-07,
      "loss": 0.1573,
      "step": 383470
    },
    {
      "epoch": 148.92,
      "learning_rate": 1.4343042071197412e-07,
      "loss": 0.0584,
      "step": 383480
    },
    {
      "epoch": 148.93,
      "learning_rate": 1.4291262135922332e-07,
      "loss": 0.002,
      "step": 383490
    },
    {
      "epoch": 148.93,
      "learning_rate": 1.423948220064725e-07,
      "loss": 0.0086,
      "step": 383500
    },
    {
      "epoch": 148.94,
      "learning_rate": 1.418770226537217e-07,
      "loss": 0.0396,
      "step": 383510
    },
    {
      "epoch": 148.94,
      "learning_rate": 1.4135922330097087e-07,
      "loss": 0.1335,
      "step": 383520
    },
    {
      "epoch": 148.94,
      "learning_rate": 1.4084142394822007e-07,
      "loss": 0.0043,
      "step": 383530
    },
    {
      "epoch": 148.95,
      "learning_rate": 1.4032362459546926e-07,
      "loss": 0.0777,
      "step": 383540
    },
    {
      "epoch": 148.95,
      "learning_rate": 1.3980582524271846e-07,
      "loss": 0.0278,
      "step": 383550
    },
    {
      "epoch": 148.96,
      "learning_rate": 1.3928802588996765e-07,
      "loss": 0.0299,
      "step": 383560
    },
    {
      "epoch": 148.96,
      "learning_rate": 1.3877022653721685e-07,
      "loss": 0.0251,
      "step": 383570
    },
    {
      "epoch": 148.96,
      "learning_rate": 1.3825242718446604e-07,
      "loss": 0.0501,
      "step": 383580
    },
    {
      "epoch": 148.97,
      "learning_rate": 1.377346278317152e-07,
      "loss": 0.1168,
      "step": 383590
    },
    {
      "epoch": 148.97,
      "learning_rate": 1.372168284789644e-07,
      "loss": 0.0291,
      "step": 383600
    },
    {
      "epoch": 148.97,
      "learning_rate": 1.366990291262136e-07,
      "loss": 0.0004,
      "step": 383610
    },
    {
      "epoch": 148.98,
      "learning_rate": 1.361812297734628e-07,
      "loss": 0.0239,
      "step": 383620
    },
    {
      "epoch": 148.98,
      "learning_rate": 1.35663430420712e-07,
      "loss": 0.0778,
      "step": 383630
    },
    {
      "epoch": 148.99,
      "learning_rate": 1.3514563106796118e-07,
      "loss": 0.1128,
      "step": 383640
    },
    {
      "epoch": 148.99,
      "learning_rate": 1.3462783171521038e-07,
      "loss": 0.001,
      "step": 383650
    },
    {
      "epoch": 148.99,
      "learning_rate": 1.3411003236245954e-07,
      "loss": 0.0071,
      "step": 383660
    },
    {
      "epoch": 149.0,
      "learning_rate": 1.3359223300970874e-07,
      "loss": 0.0927,
      "step": 383670
    },
    {
      "epoch": 149.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.3894926905632019,
      "eval_runtime": 8.2902,
      "eval_samples_per_second": 438.468,
      "eval_steps_per_second": 54.884,
      "step": 383675
    },
    {
      "epoch": 149.0,
      "learning_rate": 1.3307443365695793e-07,
      "loss": 0.0071,
      "step": 383680
    },
    {
      "epoch": 149.01,
      "learning_rate": 1.3255663430420713e-07,
      "loss": 0.0002,
      "step": 383690
    },
    {
      "epoch": 149.01,
      "learning_rate": 1.3203883495145632e-07,
      "loss": 0.0561,
      "step": 383700
    },
    {
      "epoch": 149.01,
      "learning_rate": 1.3152103559870552e-07,
      "loss": 0.0316,
      "step": 383710
    },
    {
      "epoch": 149.02,
      "learning_rate": 1.310032362459547e-07,
      "loss": 0.0011,
      "step": 383720
    },
    {
      "epoch": 149.02,
      "learning_rate": 1.304854368932039e-07,
      "loss": 0.0542,
      "step": 383730
    },
    {
      "epoch": 149.03,
      "learning_rate": 1.2996763754045307e-07,
      "loss": 0.0003,
      "step": 383740
    },
    {
      "epoch": 149.03,
      "learning_rate": 1.294498381877023e-07,
      "loss": 0.0069,
      "step": 383750
    },
    {
      "epoch": 149.03,
      "learning_rate": 1.2893203883495146e-07,
      "loss": 0.039,
      "step": 383760
    },
    {
      "epoch": 149.04,
      "learning_rate": 1.2841423948220066e-07,
      "loss": 0.0189,
      "step": 383770
    },
    {
      "epoch": 149.04,
      "learning_rate": 1.2789644012944985e-07,
      "loss": 0.0062,
      "step": 383780
    },
    {
      "epoch": 149.04,
      "learning_rate": 1.2737864077669905e-07,
      "loss": 0.0702,
      "step": 383790
    },
    {
      "epoch": 149.05,
      "learning_rate": 1.2686084142394824e-07,
      "loss": 0.05,
      "step": 383800
    },
    {
      "epoch": 149.05,
      "learning_rate": 1.263430420711974e-07,
      "loss": 0.0028,
      "step": 383810
    },
    {
      "epoch": 149.06,
      "learning_rate": 1.2582524271844663e-07,
      "loss": 0.0087,
      "step": 383820
    },
    {
      "epoch": 149.06,
      "learning_rate": 1.253074433656958e-07,
      "loss": 0.0922,
      "step": 383830
    },
    {
      "epoch": 149.06,
      "learning_rate": 1.24789644012945e-07,
      "loss": 0.0092,
      "step": 383840
    },
    {
      "epoch": 149.07,
      "learning_rate": 1.242718446601942e-07,
      "loss": 0.1135,
      "step": 383850
    },
    {
      "epoch": 149.07,
      "learning_rate": 1.2375404530744338e-07,
      "loss": 0.0081,
      "step": 383860
    },
    {
      "epoch": 149.08,
      "learning_rate": 1.2323624595469258e-07,
      "loss": 0.0368,
      "step": 383870
    },
    {
      "epoch": 149.08,
      "learning_rate": 1.2271844660194174e-07,
      "loss": 0.0247,
      "step": 383880
    },
    {
      "epoch": 149.08,
      "learning_rate": 1.2220064724919097e-07,
      "loss": 0.0755,
      "step": 383890
    },
    {
      "epoch": 149.09,
      "learning_rate": 1.2168284789644013e-07,
      "loss": 0.0476,
      "step": 383900
    },
    {
      "epoch": 149.09,
      "learning_rate": 1.2116504854368933e-07,
      "loss": 0.0722,
      "step": 383910
    },
    {
      "epoch": 149.1,
      "learning_rate": 1.2064724919093852e-07,
      "loss": 0.0091,
      "step": 383920
    },
    {
      "epoch": 149.1,
      "learning_rate": 1.2012944983818772e-07,
      "loss": 0.0819,
      "step": 383930
    },
    {
      "epoch": 149.1,
      "learning_rate": 1.196116504854369e-07,
      "loss": 0.0855,
      "step": 383940
    },
    {
      "epoch": 149.11,
      "learning_rate": 1.1909385113268609e-07,
      "loss": 0.0012,
      "step": 383950
    },
    {
      "epoch": 149.11,
      "learning_rate": 1.1857605177993529e-07,
      "loss": 0.0069,
      "step": 383960
    },
    {
      "epoch": 149.11,
      "learning_rate": 1.1805825242718448e-07,
      "loss": 0.0384,
      "step": 383970
    },
    {
      "epoch": 149.12,
      "learning_rate": 1.1754045307443366e-07,
      "loss": 0.0369,
      "step": 383980
    },
    {
      "epoch": 149.12,
      "learning_rate": 1.1702265372168284e-07,
      "loss": 0.0289,
      "step": 383990
    },
    {
      "epoch": 149.13,
      "learning_rate": 1.1650485436893205e-07,
      "loss": 0.0401,
      "step": 384000
    },
    {
      "epoch": 149.13,
      "learning_rate": 1.1598705501618123e-07,
      "loss": 0.0492,
      "step": 384010
    },
    {
      "epoch": 149.13,
      "learning_rate": 1.1546925566343043e-07,
      "loss": 0.0115,
      "step": 384020
    },
    {
      "epoch": 149.14,
      "learning_rate": 1.1495145631067962e-07,
      "loss": 0.0217,
      "step": 384030
    },
    {
      "epoch": 149.14,
      "learning_rate": 1.1443365695792882e-07,
      "loss": 0.0003,
      "step": 384040
    },
    {
      "epoch": 149.15,
      "learning_rate": 1.13915857605178e-07,
      "loss": 0.013,
      "step": 384050
    },
    {
      "epoch": 149.15,
      "learning_rate": 1.1339805825242718e-07,
      "loss": 0.0089,
      "step": 384060
    },
    {
      "epoch": 149.15,
      "learning_rate": 1.1288025889967639e-07,
      "loss": 0.0876,
      "step": 384070
    },
    {
      "epoch": 149.16,
      "learning_rate": 1.1236245954692557e-07,
      "loss": 0.0049,
      "step": 384080
    },
    {
      "epoch": 149.16,
      "learning_rate": 1.1184466019417476e-07,
      "loss": 0.0454,
      "step": 384090
    },
    {
      "epoch": 149.17,
      "learning_rate": 1.1132686084142396e-07,
      "loss": 0.2019,
      "step": 384100
    },
    {
      "epoch": 149.17,
      "learning_rate": 1.1080906148867315e-07,
      "loss": 0.0144,
      "step": 384110
    },
    {
      "epoch": 149.17,
      "learning_rate": 1.1029126213592233e-07,
      "loss": 0.0885,
      "step": 384120
    },
    {
      "epoch": 149.18,
      "learning_rate": 1.0977346278317151e-07,
      "loss": 0.0213,
      "step": 384130
    },
    {
      "epoch": 149.18,
      "learning_rate": 1.0925566343042072e-07,
      "loss": 0.0161,
      "step": 384140
    },
    {
      "epoch": 149.18,
      "learning_rate": 1.0873786407766992e-07,
      "loss": 0.0266,
      "step": 384150
    },
    {
      "epoch": 149.19,
      "learning_rate": 1.082200647249191e-07,
      "loss": 0.0726,
      "step": 384160
    },
    {
      "epoch": 149.19,
      "learning_rate": 1.077022653721683e-07,
      "loss": 0.0188,
      "step": 384170
    },
    {
      "epoch": 149.2,
      "learning_rate": 1.0718446601941749e-07,
      "loss": 0.0638,
      "step": 384180
    },
    {
      "epoch": 149.2,
      "learning_rate": 1.0666666666666667e-07,
      "loss": 0.0919,
      "step": 384190
    },
    {
      "epoch": 149.2,
      "learning_rate": 1.0614886731391586e-07,
      "loss": 0.0624,
      "step": 384200
    },
    {
      "epoch": 149.21,
      "learning_rate": 1.0563106796116506e-07,
      "loss": 0.0261,
      "step": 384210
    },
    {
      "epoch": 149.21,
      "learning_rate": 1.0511326860841425e-07,
      "loss": 0.0323,
      "step": 384220
    },
    {
      "epoch": 149.22,
      "learning_rate": 1.0459546925566343e-07,
      "loss": 0.0532,
      "step": 384230
    },
    {
      "epoch": 149.22,
      "learning_rate": 1.0407766990291264e-07,
      "loss": 0.0402,
      "step": 384240
    },
    {
      "epoch": 149.22,
      "learning_rate": 1.0355987055016182e-07,
      "loss": 0.0184,
      "step": 384250
    },
    {
      "epoch": 149.23,
      "learning_rate": 1.03042071197411e-07,
      "loss": 0.0994,
      "step": 384260
    },
    {
      "epoch": 149.23,
      "learning_rate": 1.025242718446602e-07,
      "loss": 0.0279,
      "step": 384270
    },
    {
      "epoch": 149.23,
      "learning_rate": 1.0200647249190939e-07,
      "loss": 0.0001,
      "step": 384280
    },
    {
      "epoch": 149.24,
      "learning_rate": 1.0148867313915859e-07,
      "loss": 0.0239,
      "step": 384290
    },
    {
      "epoch": 149.24,
      "learning_rate": 1.0097087378640777e-07,
      "loss": 0.0526,
      "step": 384300
    },
    {
      "epoch": 149.25,
      "learning_rate": 1.0045307443365698e-07,
      "loss": 0.0683,
      "step": 384310
    },
    {
      "epoch": 149.25,
      "learning_rate": 9.993527508090616e-08,
      "loss": 0.0146,
      "step": 384320
    },
    {
      "epoch": 149.25,
      "learning_rate": 9.941747572815534e-08,
      "loss": 0.0891,
      "step": 384330
    },
    {
      "epoch": 149.26,
      "learning_rate": 9.889967637540453e-08,
      "loss": 0.0184,
      "step": 384340
    },
    {
      "epoch": 149.26,
      "learning_rate": 9.838187702265373e-08,
      "loss": 0.0872,
      "step": 384350
    },
    {
      "epoch": 149.27,
      "learning_rate": 9.786407766990292e-08,
      "loss": 0.0021,
      "step": 384360
    },
    {
      "epoch": 149.27,
      "learning_rate": 9.73462783171521e-08,
      "loss": 0.1018,
      "step": 384370
    },
    {
      "epoch": 149.27,
      "learning_rate": 9.682847896440131e-08,
      "loss": 0.0648,
      "step": 384380
    },
    {
      "epoch": 149.28,
      "learning_rate": 9.631067961165049e-08,
      "loss": 0.0001,
      "step": 384390
    },
    {
      "epoch": 149.28,
      "learning_rate": 9.579288025889967e-08,
      "loss": 0.0027,
      "step": 384400
    },
    {
      "epoch": 149.29,
      "learning_rate": 9.527508090614887e-08,
      "loss": 0.035,
      "step": 384410
    },
    {
      "epoch": 149.29,
      "learning_rate": 9.475728155339806e-08,
      "loss": 0.0249,
      "step": 384420
    },
    {
      "epoch": 149.29,
      "learning_rate": 9.423948220064726e-08,
      "loss": 0.0179,
      "step": 384430
    },
    {
      "epoch": 149.3,
      "learning_rate": 9.372168284789644e-08,
      "loss": 0.0224,
      "step": 384440
    },
    {
      "epoch": 149.3,
      "learning_rate": 9.320388349514565e-08,
      "loss": 0.0859,
      "step": 384450
    },
    {
      "epoch": 149.3,
      "learning_rate": 9.268608414239483e-08,
      "loss": 0.0819,
      "step": 384460
    },
    {
      "epoch": 149.31,
      "learning_rate": 9.216828478964402e-08,
      "loss": 0.0005,
      "step": 384470
    },
    {
      "epoch": 149.31,
      "learning_rate": 9.16504854368932e-08,
      "loss": 0.0016,
      "step": 384480
    },
    {
      "epoch": 149.32,
      "learning_rate": 9.113268608414241e-08,
      "loss": 0.0509,
      "step": 384490
    },
    {
      "epoch": 149.32,
      "learning_rate": 9.061488673139159e-08,
      "loss": 0.0106,
      "step": 384500
    },
    {
      "epoch": 149.32,
      "learning_rate": 9.009708737864077e-08,
      "loss": 0.0269,
      "step": 384510
    },
    {
      "epoch": 149.33,
      "learning_rate": 8.957928802588998e-08,
      "loss": 0.0981,
      "step": 384520
    },
    {
      "epoch": 149.33,
      "learning_rate": 8.906148867313916e-08,
      "loss": 0.0411,
      "step": 384530
    },
    {
      "epoch": 149.34,
      "learning_rate": 8.854368932038836e-08,
      "loss": 0.0137,
      "step": 384540
    },
    {
      "epoch": 149.34,
      "learning_rate": 8.802588996763754e-08,
      "loss": 0.0411,
      "step": 384550
    },
    {
      "epoch": 149.34,
      "learning_rate": 8.750809061488675e-08,
      "loss": 0.0108,
      "step": 384560
    },
    {
      "epoch": 149.35,
      "learning_rate": 8.699029126213593e-08,
      "loss": 0.022,
      "step": 384570
    },
    {
      "epoch": 149.35,
      "learning_rate": 8.647249190938511e-08,
      "loss": 0.01,
      "step": 384580
    },
    {
      "epoch": 149.36,
      "learning_rate": 8.595469255663432e-08,
      "loss": 0.02,
      "step": 384590
    },
    {
      "epoch": 149.36,
      "learning_rate": 8.54368932038835e-08,
      "loss": 0.077,
      "step": 384600
    },
    {
      "epoch": 149.36,
      "learning_rate": 8.49190938511327e-08,
      "loss": 0.051,
      "step": 384610
    },
    {
      "epoch": 149.37,
      "learning_rate": 8.440129449838187e-08,
      "loss": 0.034,
      "step": 384620
    },
    {
      "epoch": 149.37,
      "learning_rate": 8.388349514563108e-08,
      "loss": 0.1025,
      "step": 384630
    },
    {
      "epoch": 149.37,
      "learning_rate": 8.336569579288026e-08,
      "loss": 0.0199,
      "step": 384640
    },
    {
      "epoch": 149.38,
      "learning_rate": 8.284789644012945e-08,
      "loss": 0.0146,
      "step": 384650
    },
    {
      "epoch": 149.38,
      "learning_rate": 8.233009708737865e-08,
      "loss": 0.0195,
      "step": 384660
    },
    {
      "epoch": 149.39,
      "learning_rate": 8.181229773462783e-08,
      "loss": 0.0546,
      "step": 384670
    },
    {
      "epoch": 149.39,
      "learning_rate": 8.129449838187703e-08,
      "loss": 0.0981,
      "step": 384680
    },
    {
      "epoch": 149.39,
      "learning_rate": 8.077669902912621e-08,
      "loss": 0.0003,
      "step": 384690
    },
    {
      "epoch": 149.4,
      "learning_rate": 8.025889967637542e-08,
      "loss": 0.0019,
      "step": 384700
    },
    {
      "epoch": 149.4,
      "learning_rate": 7.97411003236246e-08,
      "loss": 0.0089,
      "step": 384710
    },
    {
      "epoch": 149.41,
      "learning_rate": 7.92233009708738e-08,
      "loss": 0.0088,
      "step": 384720
    },
    {
      "epoch": 149.41,
      "learning_rate": 7.870550161812299e-08,
      "loss": 0.0756,
      "step": 384730
    },
    {
      "epoch": 149.41,
      "learning_rate": 7.818770226537218e-08,
      "loss": 0.0001,
      "step": 384740
    },
    {
      "epoch": 149.42,
      "learning_rate": 7.766990291262136e-08,
      "loss": 0.0062,
      "step": 384750
    },
    {
      "epoch": 149.42,
      "learning_rate": 7.715210355987056e-08,
      "loss": 0.1209,
      "step": 384760
    },
    {
      "epoch": 149.43,
      "learning_rate": 7.663430420711975e-08,
      "loss": 0.0163,
      "step": 384770
    },
    {
      "epoch": 149.43,
      "learning_rate": 7.611650485436893e-08,
      "loss": 0.0443,
      "step": 384780
    },
    {
      "epoch": 149.43,
      "learning_rate": 7.559870550161813e-08,
      "loss": 0.0282,
      "step": 384790
    },
    {
      "epoch": 149.44,
      "learning_rate": 7.508090614886732e-08,
      "loss": 0.0002,
      "step": 384800
    },
    {
      "epoch": 149.44,
      "learning_rate": 7.456310679611652e-08,
      "loss": 0.0461,
      "step": 384810
    },
    {
      "epoch": 149.44,
      "learning_rate": 7.40453074433657e-08,
      "loss": 0.0383,
      "step": 384820
    },
    {
      "epoch": 149.45,
      "learning_rate": 7.35275080906149e-08,
      "loss": 0.093,
      "step": 384830
    },
    {
      "epoch": 149.45,
      "learning_rate": 7.300970873786409e-08,
      "loss": 0.0355,
      "step": 384840
    },
    {
      "epoch": 149.46,
      "learning_rate": 7.249190938511327e-08,
      "loss": 0.0733,
      "step": 384850
    },
    {
      "epoch": 149.46,
      "learning_rate": 7.197411003236246e-08,
      "loss": 0.0282,
      "step": 384860
    },
    {
      "epoch": 149.46,
      "learning_rate": 7.145631067961166e-08,
      "loss": 0.0019,
      "step": 384870
    },
    {
      "epoch": 149.47,
      "learning_rate": 7.093851132686085e-08,
      "loss": 0.0023,
      "step": 384880
    },
    {
      "epoch": 149.47,
      "learning_rate": 7.042071197411003e-08,
      "loss": 0.1254,
      "step": 384890
    },
    {
      "epoch": 149.48,
      "learning_rate": 6.990291262135923e-08,
      "loss": 0.0028,
      "step": 384900
    },
    {
      "epoch": 149.48,
      "learning_rate": 6.938511326860842e-08,
      "loss": 0.0228,
      "step": 384910
    },
    {
      "epoch": 149.48,
      "learning_rate": 6.88673139158576e-08,
      "loss": 0.0903,
      "step": 384920
    },
    {
      "epoch": 149.49,
      "learning_rate": 6.83495145631068e-08,
      "loss": 0.0107,
      "step": 384930
    },
    {
      "epoch": 149.49,
      "learning_rate": 6.7831715210356e-08,
      "loss": 0.0243,
      "step": 384940
    },
    {
      "epoch": 149.5,
      "learning_rate": 6.731391585760519e-08,
      "loss": 0.0263,
      "step": 384950
    },
    {
      "epoch": 149.5,
      "learning_rate": 6.679611650485437e-08,
      "loss": 0.0014,
      "step": 384960
    },
    {
      "epoch": 149.5,
      "learning_rate": 6.627831715210356e-08,
      "loss": 0.0949,
      "step": 384970
    },
    {
      "epoch": 149.51,
      "learning_rate": 6.576051779935276e-08,
      "loss": 0.1516,
      "step": 384980
    },
    {
      "epoch": 149.51,
      "learning_rate": 6.524271844660195e-08,
      "loss": 0.0726,
      "step": 384990
    },
    {
      "epoch": 149.51,
      "learning_rate": 6.472491909385115e-08,
      "loss": 0.0034,
      "step": 385000
    },
    {
      "epoch": 149.52,
      "learning_rate": 6.420711974110033e-08,
      "loss": 0.0615,
      "step": 385010
    },
    {
      "epoch": 149.52,
      "learning_rate": 6.368932038834952e-08,
      "loss": 0.0002,
      "step": 385020
    },
    {
      "epoch": 149.53,
      "learning_rate": 6.31715210355987e-08,
      "loss": 0.0001,
      "step": 385030
    },
    {
      "epoch": 149.53,
      "learning_rate": 6.26537216828479e-08,
      "loss": 0.0415,
      "step": 385040
    },
    {
      "epoch": 149.53,
      "learning_rate": 6.21359223300971e-08,
      "loss": 0.1139,
      "step": 385050
    },
    {
      "epoch": 149.54,
      "learning_rate": 6.161812297734629e-08,
      "loss": 0.0566,
      "step": 385060
    },
    {
      "epoch": 149.54,
      "learning_rate": 6.110032362459548e-08,
      "loss": 0.0781,
      "step": 385070
    },
    {
      "epoch": 149.55,
      "learning_rate": 6.058252427184466e-08,
      "loss": 0.0006,
      "step": 385080
    },
    {
      "epoch": 149.55,
      "learning_rate": 6.006472491909386e-08,
      "loss": 0.0261,
      "step": 385090
    },
    {
      "epoch": 149.55,
      "learning_rate": 5.9546925566343046e-08,
      "loss": 0.0139,
      "step": 385100
    },
    {
      "epoch": 149.56,
      "learning_rate": 5.902912621359224e-08,
      "loss": 0.0491,
      "step": 385110
    },
    {
      "epoch": 149.56,
      "learning_rate": 5.851132686084142e-08,
      "loss": 0.049,
      "step": 385120
    },
    {
      "epoch": 149.57,
      "learning_rate": 5.799352750809062e-08,
      "loss": 0.0214,
      "step": 385130
    },
    {
      "epoch": 149.57,
      "learning_rate": 5.747572815533981e-08,
      "loss": 0.0196,
      "step": 385140
    },
    {
      "epoch": 149.57,
      "learning_rate": 5.6957928802589e-08,
      "loss": 0.0187,
      "step": 385150
    },
    {
      "epoch": 149.58,
      "learning_rate": 5.6440129449838194e-08,
      "loss": 0.0301,
      "step": 385160
    },
    {
      "epoch": 149.58,
      "learning_rate": 5.592233009708738e-08,
      "loss": 0.0224,
      "step": 385170
    },
    {
      "epoch": 149.58,
      "learning_rate": 5.5404530744336576e-08,
      "loss": 0.0094,
      "step": 385180
    },
    {
      "epoch": 149.59,
      "learning_rate": 5.488673139158576e-08,
      "loss": 0.0006,
      "step": 385190
    },
    {
      "epoch": 149.59,
      "learning_rate": 5.436893203883496e-08,
      "loss": 0.0128,
      "step": 385200
    },
    {
      "epoch": 149.6,
      "learning_rate": 5.385113268608415e-08,
      "loss": 0.0717,
      "step": 385210
    },
    {
      "epoch": 149.6,
      "learning_rate": 5.3333333333333334e-08,
      "loss": 0.0061,
      "step": 385220
    },
    {
      "epoch": 149.6,
      "learning_rate": 5.281553398058253e-08,
      "loss": 0.0277,
      "step": 385230
    },
    {
      "epoch": 149.61,
      "learning_rate": 5.229773462783172e-08,
      "loss": 0.0217,
      "step": 385240
    },
    {
      "epoch": 149.61,
      "learning_rate": 5.177993527508091e-08,
      "loss": 0.0904,
      "step": 385250
    },
    {
      "epoch": 149.62,
      "learning_rate": 5.12621359223301e-08,
      "loss": 0.0011,
      "step": 385260
    },
    {
      "epoch": 149.62,
      "learning_rate": 5.0744336569579294e-08,
      "loss": 0.1023,
      "step": 385270
    },
    {
      "epoch": 149.62,
      "learning_rate": 5.022653721682849e-08,
      "loss": 0.0782,
      "step": 385280
    },
    {
      "epoch": 149.63,
      "learning_rate": 4.970873786407767e-08,
      "loss": 0.0149,
      "step": 385290
    },
    {
      "epoch": 149.63,
      "learning_rate": 4.9190938511326864e-08,
      "loss": 0.0322,
      "step": 385300
    },
    {
      "epoch": 149.63,
      "learning_rate": 4.867313915857605e-08,
      "loss": 0.0009,
      "step": 385310
    },
    {
      "epoch": 149.64,
      "learning_rate": 4.8155339805825247e-08,
      "loss": 0.0572,
      "step": 385320
    },
    {
      "epoch": 149.64,
      "learning_rate": 4.7637540453074435e-08,
      "loss": 0.1074,
      "step": 385330
    },
    {
      "epoch": 149.65,
      "learning_rate": 4.711974110032363e-08,
      "loss": 0.047,
      "step": 385340
    },
    {
      "epoch": 149.65,
      "learning_rate": 4.6601941747572824e-08,
      "loss": 0.1158,
      "step": 385350
    },
    {
      "epoch": 149.65,
      "learning_rate": 4.608414239482201e-08,
      "loss": 0.0077,
      "step": 385360
    },
    {
      "epoch": 149.66,
      "learning_rate": 4.5566343042071206e-08,
      "loss": 0.0008,
      "step": 385370
    },
    {
      "epoch": 149.66,
      "learning_rate": 4.504854368932039e-08,
      "loss": 0.061,
      "step": 385380
    },
    {
      "epoch": 149.67,
      "learning_rate": 4.453074433656958e-08,
      "loss": 0.0002,
      "step": 385390
    },
    {
      "epoch": 149.67,
      "learning_rate": 4.401294498381877e-08,
      "loss": 0.0389,
      "step": 385400
    },
    {
      "epoch": 149.67,
      "learning_rate": 4.3495145631067964e-08,
      "loss": 0.018,
      "step": 385410
    },
    {
      "epoch": 149.68,
      "learning_rate": 4.297734627831716e-08,
      "loss": 0.0214,
      "step": 385420
    },
    {
      "epoch": 149.68,
      "learning_rate": 4.245954692556635e-08,
      "loss": 0.0568,
      "step": 385430
    },
    {
      "epoch": 149.69,
      "learning_rate": 4.194174757281554e-08,
      "loss": 0.0089,
      "step": 385440
    },
    {
      "epoch": 149.69,
      "learning_rate": 4.142394822006472e-08,
      "loss": 0.0315,
      "step": 385450
    },
    {
      "epoch": 149.69,
      "learning_rate": 4.090614886731392e-08,
      "loss": 0.0376,
      "step": 385460
    },
    {
      "epoch": 149.7,
      "learning_rate": 4.0388349514563105e-08,
      "loss": 0.0099,
      "step": 385470
    },
    {
      "epoch": 149.7,
      "learning_rate": 3.98705501618123e-08,
      "loss": 0.0458,
      "step": 385480
    },
    {
      "epoch": 149.7,
      "learning_rate": 3.9352750809061494e-08,
      "loss": 0.144,
      "step": 385490
    },
    {
      "epoch": 149.71,
      "learning_rate": 3.883495145631068e-08,
      "loss": 0.0093,
      "step": 385500
    },
    {
      "epoch": 149.71,
      "learning_rate": 3.8317152103559876e-08,
      "loss": 0.0168,
      "step": 385510
    },
    {
      "epoch": 149.72,
      "learning_rate": 3.7799352750809064e-08,
      "loss": 0.0092,
      "step": 385520
    },
    {
      "epoch": 149.72,
      "learning_rate": 3.728155339805826e-08,
      "loss": 0.0076,
      "step": 385530
    },
    {
      "epoch": 149.72,
      "learning_rate": 3.676375404530745e-08,
      "loss": 0.0669,
      "step": 385540
    },
    {
      "epoch": 149.73,
      "learning_rate": 3.6245954692556635e-08,
      "loss": 0.0656,
      "step": 385550
    },
    {
      "epoch": 149.73,
      "learning_rate": 3.572815533980583e-08,
      "loss": 0.0634,
      "step": 385560
    },
    {
      "epoch": 149.74,
      "learning_rate": 3.521035598705502e-08,
      "loss": 0.0448,
      "step": 385570
    },
    {
      "epoch": 149.74,
      "learning_rate": 3.469255663430421e-08,
      "loss": 0.0249,
      "step": 385580
    },
    {
      "epoch": 149.74,
      "learning_rate": 3.41747572815534e-08,
      "loss": 0.0176,
      "step": 385590
    },
    {
      "epoch": 149.75,
      "learning_rate": 3.3656957928802594e-08,
      "loss": 0.0431,
      "step": 385600
    },
    {
      "epoch": 149.75,
      "learning_rate": 3.313915857605178e-08,
      "loss": 0.2028,
      "step": 385610
    },
    {
      "epoch": 149.76,
      "learning_rate": 3.2621359223300977e-08,
      "loss": 0.0553,
      "step": 385620
    },
    {
      "epoch": 149.76,
      "learning_rate": 3.2103559870550164e-08,
      "loss": 0.0001,
      "step": 385630
    },
    {
      "epoch": 149.76,
      "learning_rate": 3.158576051779935e-08,
      "loss": 0.0084,
      "step": 385640
    },
    {
      "epoch": 149.77,
      "learning_rate": 3.106796116504855e-08,
      "loss": 0.0864,
      "step": 385650
    },
    {
      "epoch": 149.77,
      "learning_rate": 3.055016181229774e-08,
      "loss": 0.0001,
      "step": 385660
    },
    {
      "epoch": 149.77,
      "learning_rate": 3.003236245954693e-08,
      "loss": 0.0291,
      "step": 385670
    },
    {
      "epoch": 149.78,
      "learning_rate": 2.951456310679612e-08,
      "loss": 0.0083,
      "step": 385680
    },
    {
      "epoch": 149.78,
      "learning_rate": 2.899676375404531e-08,
      "loss": 0.0109,
      "step": 385690
    },
    {
      "epoch": 149.79,
      "learning_rate": 2.84789644012945e-08,
      "loss": 0.1049,
      "step": 385700
    },
    {
      "epoch": 149.79,
      "learning_rate": 2.796116504854369e-08,
      "loss": 0.0001,
      "step": 385710
    },
    {
      "epoch": 149.79,
      "learning_rate": 2.744336569579288e-08,
      "loss": 0.0181,
      "step": 385720
    },
    {
      "epoch": 149.8,
      "learning_rate": 2.6925566343042077e-08,
      "loss": 0.0604,
      "step": 385730
    },
    {
      "epoch": 149.8,
      "learning_rate": 2.6407766990291264e-08,
      "loss": 0.0322,
      "step": 385740
    },
    {
      "epoch": 149.81,
      "learning_rate": 2.5889967637540456e-08,
      "loss": 0.0326,
      "step": 385750
    },
    {
      "epoch": 149.81,
      "learning_rate": 2.5372168284789647e-08,
      "loss": 0.1298,
      "step": 385760
    },
    {
      "epoch": 149.81,
      "learning_rate": 2.4854368932038835e-08,
      "loss": 0.0994,
      "step": 385770
    },
    {
      "epoch": 149.82,
      "learning_rate": 2.4336569579288026e-08,
      "loss": 0.0263,
      "step": 385780
    },
    {
      "epoch": 149.82,
      "learning_rate": 2.3818770226537217e-08,
      "loss": 0.0039,
      "step": 385790
    },
    {
      "epoch": 149.83,
      "learning_rate": 2.3300970873786412e-08,
      "loss": 0.0473,
      "step": 385800
    },
    {
      "epoch": 149.83,
      "learning_rate": 2.2783171521035603e-08,
      "loss": 0.0233,
      "step": 385810
    },
    {
      "epoch": 149.83,
      "learning_rate": 2.226537216828479e-08,
      "loss": 0.0328,
      "step": 385820
    },
    {
      "epoch": 149.84,
      "learning_rate": 2.1747572815533982e-08,
      "loss": 0.0278,
      "step": 385830
    },
    {
      "epoch": 149.84,
      "learning_rate": 2.1229773462783173e-08,
      "loss": 0.0304,
      "step": 385840
    },
    {
      "epoch": 149.84,
      "learning_rate": 2.071197411003236e-08,
      "loss": 0.0747,
      "step": 385850
    },
    {
      "epoch": 149.85,
      "learning_rate": 2.0194174757281552e-08,
      "loss": 0.0248,
      "step": 385860
    },
    {
      "epoch": 149.85,
      "learning_rate": 1.9676375404530747e-08,
      "loss": 0.0066,
      "step": 385870
    },
    {
      "epoch": 149.86,
      "learning_rate": 1.9158576051779938e-08,
      "loss": 0.0009,
      "step": 385880
    },
    {
      "epoch": 149.86,
      "learning_rate": 1.864077669902913e-08,
      "loss": 0.0613,
      "step": 385890
    },
    {
      "epoch": 149.86,
      "learning_rate": 1.8122977346278317e-08,
      "loss": 0.0418,
      "step": 385900
    },
    {
      "epoch": 149.87,
      "learning_rate": 1.760517799352751e-08,
      "loss": 0.0633,
      "step": 385910
    },
    {
      "epoch": 149.87,
      "learning_rate": 1.70873786407767e-08,
      "loss": 0.0337,
      "step": 385920
    },
    {
      "epoch": 149.88,
      "learning_rate": 1.656957928802589e-08,
      "loss": 0.0088,
      "step": 385930
    },
    {
      "epoch": 149.88,
      "learning_rate": 1.6051779935275082e-08,
      "loss": 0.0349,
      "step": 385940
    },
    {
      "epoch": 149.88,
      "learning_rate": 1.5533980582524273e-08,
      "loss": 0.0088,
      "step": 385950
    },
    {
      "epoch": 149.89,
      "learning_rate": 1.5016181229773465e-08,
      "loss": 0.0163,
      "step": 385960
    },
    {
      "epoch": 149.89,
      "learning_rate": 1.4498381877022654e-08,
      "loss": 0.06,
      "step": 385970
    },
    {
      "epoch": 149.9,
      "learning_rate": 1.3980582524271845e-08,
      "loss": 0.1068,
      "step": 385980
    },
    {
      "epoch": 149.9,
      "learning_rate": 1.3462783171521038e-08,
      "loss": 0.0158,
      "step": 385990
    },
    {
      "epoch": 149.9,
      "learning_rate": 1.2944983818770228e-08,
      "loss": 0.011,
      "step": 386000
    },
    {
      "epoch": 149.91,
      "learning_rate": 1.2427184466019417e-08,
      "loss": 0.0363,
      "step": 386010
    },
    {
      "epoch": 149.91,
      "learning_rate": 1.1909385113268609e-08,
      "loss": 0.0668,
      "step": 386020
    },
    {
      "epoch": 149.91,
      "learning_rate": 1.1391585760517802e-08,
      "loss": 0.0007,
      "step": 386030
    },
    {
      "epoch": 149.92,
      "learning_rate": 1.0873786407766991e-08,
      "loss": 0.0161,
      "step": 386040
    },
    {
      "epoch": 149.92,
      "learning_rate": 1.035598705501618e-08,
      "loss": 0.0881,
      "step": 386050
    },
    {
      "epoch": 149.93,
      "learning_rate": 9.838187702265373e-09,
      "loss": 0.0003,
      "step": 386060
    },
    {
      "epoch": 149.93,
      "learning_rate": 9.320388349514565e-09,
      "loss": 0.0034,
      "step": 386070
    },
    {
      "epoch": 149.93,
      "learning_rate": 8.802588996763754e-09,
      "loss": 0.017,
      "step": 386080
    },
    {
      "epoch": 149.94,
      "learning_rate": 8.284789644012945e-09,
      "loss": 0.0661,
      "step": 386090
    },
    {
      "epoch": 149.94,
      "learning_rate": 7.766990291262137e-09,
      "loss": 0.0339,
      "step": 386100
    },
    {
      "epoch": 149.95,
      "learning_rate": 7.249190938511327e-09,
      "loss": 0.0703,
      "step": 386110
    },
    {
      "epoch": 149.95,
      "learning_rate": 6.731391585760519e-09,
      "loss": 0.0009,
      "step": 386120
    },
    {
      "epoch": 149.95,
      "learning_rate": 6.213592233009709e-09,
      "loss": 0.0011,
      "step": 386130
    },
    {
      "epoch": 149.96,
      "learning_rate": 5.695792880258901e-09,
      "loss": 0.0504,
      "step": 386140
    },
    {
      "epoch": 149.96,
      "learning_rate": 5.17799352750809e-09,
      "loss": 0.0437,
      "step": 386150
    },
    {
      "epoch": 149.97,
      "learning_rate": 4.660194174757282e-09,
      "loss": 0.017,
      "step": 386160
    },
    {
      "epoch": 149.97,
      "learning_rate": 4.142394822006473e-09,
      "loss": 0.2117,
      "step": 386170
    },
    {
      "epoch": 149.97,
      "learning_rate": 3.6245954692556636e-09,
      "loss": 0.037,
      "step": 386180
    },
    {
      "epoch": 149.98,
      "learning_rate": 3.1067961165048544e-09,
      "loss": 0.0015,
      "step": 386190
    },
    {
      "epoch": 149.98,
      "learning_rate": 2.588996763754045e-09,
      "loss": 0.2308,
      "step": 386200
    },
    {
      "epoch": 149.98,
      "learning_rate": 2.0711974110032364e-09,
      "loss": 0.0001,
      "step": 386210
    },
    {
      "epoch": 149.99,
      "learning_rate": 1.5533980582524272e-09,
      "loss": 0.0456,
      "step": 386220
    },
    {
      "epoch": 149.99,
      "learning_rate": 1.0355987055016182e-09,
      "loss": 0.015,
      "step": 386230
    },
    {
      "epoch": 150.0,
      "learning_rate": 5.177993527508091e-10,
      "loss": 0.0411,
      "step": 386240
    },
    {
      "epoch": 150.0,
      "learning_rate": 0.0,
      "loss": 0.081,
      "step": 386250
    },
    {
      "epoch": 150.0,
      "eval_accuracy": 0.9526822558459422,
      "eval_loss": 0.38963595032691956,
      "eval_runtime": 8.2958,
      "eval_samples_per_second": 438.171,
      "eval_steps_per_second": 54.847,
      "step": 386250
    },
    {
      "epoch": 150.0,
      "step": 386250,
      "total_flos": 2.3944845939692818e+20,
      "train_loss": 0.08211426009329709,
      "train_runtime": 20002.7488,
      "train_samples_per_second": 154.464,
      "train_steps_per_second": 19.31
    }
  ],
  "logging_steps": 10,
  "max_steps": 386250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 150,
  "save_steps": 500,
  "total_flos": 2.3944845939692818e+20,
  "trial_name": null,
  "trial_params": null
}