{
  "best_metric": 0.9794921875,
  "best_model_checkpoint": "resnet-Alzheimer/checkpoint-3600",
  "epoch": 50.0,
  "eval_steps": 500,
  "global_step": 4000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.12,
      "grad_norm": 2.1754038333892822,
      "learning_rate": 5e-05,
      "loss": 1.3812,
      "step": 10
    },
    {
      "epoch": 0.25,
      "grad_norm": 3.91097092628479,
      "learning_rate": 0.0001,
      "loss": 1.34,
      "step": 20
    },
    {
      "epoch": 0.38,
      "grad_norm": 4.134509563446045,
      "learning_rate": 0.00015,
      "loss": 1.2635,
      "step": 30
    },
    {
      "epoch": 0.5,
      "grad_norm": 6.377187252044678,
      "learning_rate": 0.0002,
      "loss": 1.1824,
      "step": 40
    },
    {
      "epoch": 0.62,
      "grad_norm": 7.713193416595459,
      "learning_rate": 0.00025,
      "loss": 1.1226,
      "step": 50
    },
    {
      "epoch": 0.75,
      "grad_norm": 3.569382429122925,
      "learning_rate": 0.0003,
      "loss": 1.0308,
      "step": 60
    },
    {
      "epoch": 0.88,
      "grad_norm": 5.650737285614014,
      "learning_rate": 0.00035,
      "loss": 1.0115,
      "step": 70
    },
    {
      "epoch": 1.0,
      "grad_norm": 5.970870494842529,
      "learning_rate": 0.0004,
      "loss": 1.0127,
      "step": 80
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5087890625,
      "eval_loss": 0.9888483285903931,
      "eval_runtime": 5.4153,
      "eval_samples_per_second": 189.093,
      "eval_steps_per_second": 11.818,
      "step": 80
    },
    {
      "epoch": 1.12,
      "grad_norm": 2.92154860496521,
      "learning_rate": 0.00045000000000000004,
      "loss": 1.0175,
      "step": 90
    },
    {
      "epoch": 1.25,
      "grad_norm": 4.131512641906738,
      "learning_rate": 0.0005,
      "loss": 0.918,
      "step": 100
    },
    {
      "epoch": 1.38,
      "grad_norm": 2.9838943481445312,
      "learning_rate": 0.00055,
      "loss": 0.9613,
      "step": 110
    },
    {
      "epoch": 1.5,
      "grad_norm": 2.4230854511260986,
      "learning_rate": 0.0006,
      "loss": 0.9188,
      "step": 120
    },
    {
      "epoch": 1.62,
      "grad_norm": 3.774385690689087,
      "learning_rate": 0.0006500000000000001,
      "loss": 0.9245,
      "step": 130
    },
    {
      "epoch": 1.75,
      "grad_norm": 3.175440549850464,
      "learning_rate": 0.0007,
      "loss": 0.8713,
      "step": 140
    },
    {
      "epoch": 1.88,
      "grad_norm": 2.7985305786132812,
      "learning_rate": 0.00075,
      "loss": 0.9264,
      "step": 150
    },
    {
      "epoch": 2.0,
      "grad_norm": 1.5870747566223145,
      "learning_rate": 0.0008,
      "loss": 0.9345,
      "step": 160
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.5302734375,
      "eval_loss": 0.942151665687561,
      "eval_runtime": 5.1954,
      "eval_samples_per_second": 197.099,
      "eval_steps_per_second": 12.319,
      "step": 160
    },
    {
      "epoch": 2.12,
      "grad_norm": 1.3990691900253296,
      "learning_rate": 0.00085,
      "loss": 0.8851,
      "step": 170
    },
    {
      "epoch": 2.25,
      "grad_norm": 1.4066977500915527,
      "learning_rate": 0.0009000000000000001,
      "loss": 0.9081,
      "step": 180
    },
    {
      "epoch": 2.38,
      "grad_norm": 1.5531185865402222,
      "learning_rate": 0.00095,
      "loss": 0.8888,
      "step": 190
    },
    {
      "epoch": 2.5,
      "grad_norm": 1.960204839706421,
      "learning_rate": 0.001,
      "loss": 0.9206,
      "step": 200
    },
    {
      "epoch": 2.62,
      "grad_norm": 1.0218795537948608,
      "learning_rate": 0.0010500000000000002,
      "loss": 0.8735,
      "step": 210
    },
    {
      "epoch": 2.75,
      "grad_norm": 1.9216176271438599,
      "learning_rate": 0.0011,
      "loss": 0.924,
      "step": 220
    },
    {
      "epoch": 2.88,
      "grad_norm": 1.9017541408538818,
      "learning_rate": 0.00115,
      "loss": 0.9327,
      "step": 230
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.556686282157898,
      "learning_rate": 0.0012,
      "loss": 0.8889,
      "step": 240
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.578125,
      "eval_loss": 0.8723889589309692,
      "eval_runtime": 4.9225,
      "eval_samples_per_second": 208.026,
      "eval_steps_per_second": 13.002,
      "step": 240
    },
    {
      "epoch": 3.12,
      "grad_norm": 0.9459726810455322,
      "learning_rate": 0.00125,
      "loss": 0.8643,
      "step": 250
    },
    {
      "epoch": 3.25,
      "grad_norm": 0.749912679195404,
      "learning_rate": 0.0013000000000000002,
      "loss": 0.8978,
      "step": 260
    },
    {
      "epoch": 3.38,
      "grad_norm": 0.9228126406669617,
      "learning_rate": 0.00135,
      "loss": 0.8838,
      "step": 270
    },
    {
      "epoch": 3.5,
      "grad_norm": 1.0743939876556396,
      "learning_rate": 0.0014,
      "loss": 0.8868,
      "step": 280
    },
    {
      "epoch": 3.62,
      "grad_norm": 0.997053325176239,
      "learning_rate": 0.00145,
      "loss": 0.8632,
      "step": 290
    },
    {
      "epoch": 3.75,
      "grad_norm": 0.5891302227973938,
      "learning_rate": 0.0015,
      "loss": 0.8501,
      "step": 300
    },
    {
      "epoch": 3.88,
      "grad_norm": 1.0819345712661743,
      "learning_rate": 0.0015500000000000002,
      "loss": 0.884,
      "step": 310
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.5622245073318481,
      "learning_rate": 0.0016,
      "loss": 0.8843,
      "step": 320
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.5888671875,
      "eval_loss": 0.8535706996917725,
      "eval_runtime": 6.5338,
      "eval_samples_per_second": 156.725,
      "eval_steps_per_second": 9.795,
      "step": 320
    },
    {
      "epoch": 4.12,
      "grad_norm": 0.7301604747772217,
      "learning_rate": 0.00165,
      "loss": 0.8241,
      "step": 330
    },
    {
      "epoch": 4.25,
      "grad_norm": 1.222732424736023,
      "learning_rate": 0.0017,
      "loss": 0.8929,
      "step": 340
    },
    {
      "epoch": 4.38,
      "grad_norm": 0.8520879745483398,
      "learning_rate": 0.00175,
      "loss": 0.8726,
      "step": 350
    },
    {
      "epoch": 4.5,
      "grad_norm": 0.6151734590530396,
      "learning_rate": 0.0018000000000000002,
      "loss": 0.8851,
      "step": 360
    },
    {
      "epoch": 4.62,
      "grad_norm": 0.6786526441574097,
      "learning_rate": 0.00185,
      "loss": 0.8544,
      "step": 370
    },
    {
      "epoch": 4.75,
      "grad_norm": 0.8025469779968262,
      "learning_rate": 0.0019,
      "loss": 0.8432,
      "step": 380
    },
    {
      "epoch": 4.88,
      "grad_norm": 1.0158729553222656,
      "learning_rate": 0.00195,
      "loss": 0.8261,
      "step": 390
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.7815405130386353,
      "learning_rate": 0.002,
      "loss": 0.8397,
      "step": 400
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.615234375,
      "eval_loss": 0.8353910446166992,
      "eval_runtime": 5.3022,
      "eval_samples_per_second": 193.129,
      "eval_steps_per_second": 12.071,
      "step": 400
    },
    {
      "epoch": 5.12,
      "grad_norm": 0.9535025954246521,
      "learning_rate": 0.0019944444444444445,
      "loss": 0.8636,
      "step": 410
    },
    {
      "epoch": 5.25,
      "grad_norm": 0.43990448117256165,
      "learning_rate": 0.001988888888888889,
      "loss": 0.7931,
      "step": 420
    },
    {
      "epoch": 5.38,
      "grad_norm": 0.6062633395195007,
      "learning_rate": 0.0019833333333333335,
      "loss": 0.8345,
      "step": 430
    },
    {
      "epoch": 5.5,
      "grad_norm": 0.6349042057991028,
      "learning_rate": 0.001977777777777778,
      "loss": 0.8593,
      "step": 440
    },
    {
      "epoch": 5.62,
      "grad_norm": 0.6786915063858032,
      "learning_rate": 0.0019722222222222224,
      "loss": 0.8435,
      "step": 450
    },
    {
      "epoch": 5.75,
      "grad_norm": 0.7090786695480347,
      "learning_rate": 0.0019666666666666665,
      "loss": 0.8008,
      "step": 460
    },
    {
      "epoch": 5.88,
      "grad_norm": 0.6538481712341309,
      "learning_rate": 0.001961111111111111,
      "loss": 0.763,
      "step": 470
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.4316461682319641,
      "learning_rate": 0.0019555555555555554,
      "loss": 0.8624,
      "step": 480
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.5380859375,
      "eval_loss": 0.9221189022064209,
      "eval_runtime": 4.9173,
      "eval_samples_per_second": 208.245,
      "eval_steps_per_second": 13.015,
      "step": 480
    },
    {
      "epoch": 6.12,
      "grad_norm": 0.5696819424629211,
      "learning_rate": 0.00195,
      "loss": 0.8065,
      "step": 490
    },
    {
      "epoch": 6.25,
      "grad_norm": 0.6260728240013123,
      "learning_rate": 0.0019444444444444444,
      "loss": 0.7873,
      "step": 500
    },
    {
      "epoch": 6.38,
      "grad_norm": 0.6295855045318604,
      "learning_rate": 0.0019388888888888889,
      "loss": 0.7802,
      "step": 510
    },
    {
      "epoch": 6.5,
      "grad_norm": 0.6074417233467102,
      "learning_rate": 0.0019333333333333333,
      "loss": 0.7907,
      "step": 520
    },
    {
      "epoch": 6.62,
      "grad_norm": 0.6099679470062256,
      "learning_rate": 0.0019277777777777778,
      "loss": 0.7391,
      "step": 530
    },
    {
      "epoch": 6.75,
      "grad_norm": 0.9349565505981445,
      "learning_rate": 0.0019222222222222223,
      "loss": 0.7749,
      "step": 540
    },
    {
      "epoch": 6.88,
      "grad_norm": 0.6923946142196655,
      "learning_rate": 0.0019166666666666668,
      "loss": 0.8232,
      "step": 550
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.5967056751251221,
      "learning_rate": 0.0019111111111111113,
      "loss": 0.7543,
      "step": 560
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.6474609375,
      "eval_loss": 0.7568103671073914,
      "eval_runtime": 5.3014,
      "eval_samples_per_second": 193.156,
      "eval_steps_per_second": 12.072,
      "step": 560
    },
    {
      "epoch": 7.12,
      "grad_norm": 0.7925052642822266,
      "learning_rate": 0.0019055555555555555,
      "loss": 0.7086,
      "step": 570
    },
    {
      "epoch": 7.25,
      "grad_norm": 0.715761125087738,
      "learning_rate": 0.0019,
      "loss": 0.7901,
      "step": 580
    },
    {
      "epoch": 7.38,
      "grad_norm": 0.6602711081504822,
      "learning_rate": 0.0018944444444444445,
      "loss": 0.7375,
      "step": 590
    },
    {
      "epoch": 7.5,
      "grad_norm": 0.5104066729545593,
      "learning_rate": 0.001888888888888889,
      "loss": 0.7805,
      "step": 600
    },
    {
      "epoch": 7.62,
      "grad_norm": 0.6333702802658081,
      "learning_rate": 0.0018833333333333334,
      "loss": 0.7017,
      "step": 610
    },
    {
      "epoch": 7.75,
      "grad_norm": 0.5703239440917969,
      "learning_rate": 0.001877777777777778,
      "loss": 0.7086,
      "step": 620
    },
    {
      "epoch": 7.88,
      "grad_norm": 0.8939486742019653,
      "learning_rate": 0.0018722222222222222,
      "loss": 0.7399,
      "step": 630
    },
    {
      "epoch": 8.0,
      "grad_norm": 0.6808524131774902,
      "learning_rate": 0.0018666666666666666,
      "loss": 0.6993,
      "step": 640
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.61328125,
      "eval_loss": 0.8830391764640808,
      "eval_runtime": 4.9073,
      "eval_samples_per_second": 208.667,
      "eval_steps_per_second": 13.042,
      "step": 640
    },
    {
      "epoch": 8.12,
      "grad_norm": 0.7670312523841858,
      "learning_rate": 0.0018611111111111111,
      "loss": 0.7304,
      "step": 650
    },
    {
      "epoch": 8.25,
      "grad_norm": 0.518883466720581,
      "learning_rate": 0.0018555555555555556,
      "loss": 0.6759,
      "step": 660
    },
    {
      "epoch": 8.38,
      "grad_norm": 0.6331384778022766,
      "learning_rate": 0.00185,
      "loss": 0.7323,
      "step": 670
    },
    {
      "epoch": 8.5,
      "grad_norm": 0.5934571027755737,
      "learning_rate": 0.0018444444444444446,
      "loss": 0.7109,
      "step": 680
    },
    {
      "epoch": 8.62,
      "grad_norm": 0.5555841326713562,
      "learning_rate": 0.0018388888888888888,
      "loss": 0.7361,
      "step": 690
    },
    {
      "epoch": 8.75,
      "grad_norm": 0.45028582215309143,
      "learning_rate": 0.0018333333333333333,
      "loss": 0.7209,
      "step": 700
    },
    {
      "epoch": 8.88,
      "grad_norm": 0.4313984811306,
      "learning_rate": 0.0018277777777777778,
      "loss": 0.692,
      "step": 710
    },
    {
      "epoch": 9.0,
      "grad_norm": 0.6221916675567627,
      "learning_rate": 0.0018222222222222223,
      "loss": 0.7045,
      "step": 720
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.658203125,
      "eval_loss": 0.7372878789901733,
      "eval_runtime": 4.9218,
      "eval_samples_per_second": 208.053,
      "eval_steps_per_second": 13.003,
      "step": 720
    },
    {
      "epoch": 9.12,
      "grad_norm": 0.9794626832008362,
      "learning_rate": 0.0018166666666666667,
      "loss": 0.641,
      "step": 730
    },
    {
      "epoch": 9.25,
      "grad_norm": 0.8530990481376648,
      "learning_rate": 0.0018111111111111112,
      "loss": 0.6616,
      "step": 740
    },
    {
      "epoch": 9.38,
      "grad_norm": 0.5696712136268616,
      "learning_rate": 0.0018055555555555557,
      "loss": 0.6685,
      "step": 750
    },
    {
      "epoch": 9.5,
      "grad_norm": 0.6695945858955383,
      "learning_rate": 0.0018000000000000002,
      "loss": 0.6079,
      "step": 760
    },
    {
      "epoch": 9.62,
      "grad_norm": 0.9470874667167664,
      "learning_rate": 0.0017944444444444446,
      "loss": 0.6362,
      "step": 770
    },
    {
      "epoch": 9.75,
      "grad_norm": 1.0435755252838135,
      "learning_rate": 0.001788888888888889,
      "loss": 0.7036,
      "step": 780
    },
    {
      "epoch": 9.88,
      "grad_norm": 0.4934737980365753,
      "learning_rate": 0.0017833333333333334,
      "loss": 0.6955,
      "step": 790
    },
    {
      "epoch": 10.0,
      "grad_norm": 0.49625110626220703,
      "learning_rate": 0.0017777777777777776,
      "loss": 0.6557,
      "step": 800
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.7451171875,
      "eval_loss": 0.6075544357299805,
      "eval_runtime": 5.2806,
      "eval_samples_per_second": 193.918,
      "eval_steps_per_second": 12.12,
      "step": 800
    },
    {
      "epoch": 10.12,
      "grad_norm": 0.49739229679107666,
      "learning_rate": 0.0017722222222222221,
      "loss": 0.605,
      "step": 810
    },
    {
      "epoch": 10.25,
      "grad_norm": 0.6317277550697327,
      "learning_rate": 0.0017666666666666666,
      "loss": 0.5332,
      "step": 820
    },
    {
      "epoch": 10.38,
      "grad_norm": 0.756879985332489,
      "learning_rate": 0.001761111111111111,
      "loss": 0.5619,
      "step": 830
    },
    {
      "epoch": 10.5,
      "grad_norm": 0.6143298149108887,
      "learning_rate": 0.0017555555555555556,
      "loss": 0.601,
      "step": 840
    },
    {
      "epoch": 10.62,
      "grad_norm": 0.7249147891998291,
      "learning_rate": 0.00175,
      "loss": 0.5935,
      "step": 850
    },
    {
      "epoch": 10.75,
      "grad_norm": 0.4532654285430908,
      "learning_rate": 0.0017444444444444445,
      "loss": 0.5988,
      "step": 860
    },
    {
      "epoch": 10.88,
      "grad_norm": 0.5738415718078613,
      "learning_rate": 0.001738888888888889,
      "loss": 0.6634,
      "step": 870
    },
    {
      "epoch": 11.0,
      "grad_norm": 0.5514868497848511,
      "learning_rate": 0.0017333333333333335,
      "loss": 0.5876,
      "step": 880
    },
    {
      "epoch": 11.0,
      "eval_accuracy": 0.69921875,
      "eval_loss": 0.7281272411346436,
      "eval_runtime": 4.8994,
      "eval_samples_per_second": 209.004,
      "eval_steps_per_second": 13.063,
      "step": 880
    },
    {
      "epoch": 11.12,
      "grad_norm": 0.7158863544464111,
      "learning_rate": 0.001727777777777778,
      "loss": 0.606,
      "step": 890
    },
    {
      "epoch": 11.25,
      "grad_norm": 0.7355363368988037,
      "learning_rate": 0.0017222222222222224,
      "loss": 0.5923,
      "step": 900
    },
    {
      "epoch": 11.38,
      "grad_norm": 0.7794367671012878,
      "learning_rate": 0.0017166666666666667,
      "loss": 0.5935,
      "step": 910
    },
    {
      "epoch": 11.5,
      "grad_norm": 0.9755826592445374,
      "learning_rate": 0.0017111111111111112,
      "loss": 0.644,
      "step": 920
    },
    {
      "epoch": 11.62,
      "grad_norm": 0.6257722973823547,
      "learning_rate": 0.0017055555555555554,
      "loss": 0.617,
      "step": 930
    },
    {
      "epoch": 11.75,
      "grad_norm": 0.8550503253936768,
      "learning_rate": 0.0017,
      "loss": 0.5854,
      "step": 940
    },
    {
      "epoch": 11.88,
      "grad_norm": 0.7347137928009033,
      "learning_rate": 0.0016944444444444444,
      "loss": 0.6358,
      "step": 950
    },
    {
      "epoch": 12.0,
      "grad_norm": 0.7867416739463806,
      "learning_rate": 0.0016888888888888889,
      "loss": 0.5732,
      "step": 960
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.7509765625,
      "eval_loss": 0.5769097208976746,
      "eval_runtime": 5.275,
      "eval_samples_per_second": 194.122,
      "eval_steps_per_second": 12.133,
      "step": 960
    },
    {
      "epoch": 12.12,
      "grad_norm": 0.6022630333900452,
      "learning_rate": 0.0016833333333333333,
      "loss": 0.5643,
      "step": 970
    },
    {
      "epoch": 12.25,
      "grad_norm": 0.599958062171936,
      "learning_rate": 0.0016777777777777778,
      "loss": 0.5438,
      "step": 980
    },
    {
      "epoch": 12.38,
      "grad_norm": 0.6484814286231995,
      "learning_rate": 0.0016722222222222223,
      "loss": 0.5208,
      "step": 990
    },
    {
      "epoch": 12.5,
      "grad_norm": 0.8167735934257507,
      "learning_rate": 0.0016666666666666668,
      "loss": 0.5369,
      "step": 1000
    },
    {
      "epoch": 12.62,
      "grad_norm": 0.49088793992996216,
      "learning_rate": 0.0016611111111111113,
      "loss": 0.4803,
      "step": 1010
    },
    {
      "epoch": 12.75,
      "grad_norm": 0.6817615628242493,
      "learning_rate": 0.0016555555555555555,
      "loss": 0.5102,
      "step": 1020
    },
    {
      "epoch": 12.88,
      "grad_norm": 0.8656439781188965,
      "learning_rate": 0.00165,
      "loss": 0.5287,
      "step": 1030
    },
    {
      "epoch": 13.0,
      "grad_norm": 0.5195401310920715,
      "learning_rate": 0.0016444444444444445,
      "loss": 0.4864,
      "step": 1040
    },
    {
      "epoch": 13.0,
      "eval_accuracy": 0.8310546875,
      "eval_loss": 0.445728600025177,
      "eval_runtime": 4.9054,
      "eval_samples_per_second": 208.751,
      "eval_steps_per_second": 13.047,
      "step": 1040
    },
    {
      "epoch": 13.12,
      "grad_norm": 0.7564366459846497,
      "learning_rate": 0.001638888888888889,
      "loss": 0.4715,
      "step": 1050
    },
    {
      "epoch": 13.25,
      "grad_norm": 0.6976212859153748,
      "learning_rate": 0.0016333333333333334,
      "loss": 0.471,
      "step": 1060
    },
    {
      "epoch": 13.38,
      "grad_norm": 0.7652568817138672,
      "learning_rate": 0.001627777777777778,
      "loss": 0.4821,
      "step": 1070
    },
    {
      "epoch": 13.5,
      "grad_norm": 0.7834269404411316,
      "learning_rate": 0.0016222222222222222,
      "loss": 0.5091,
      "step": 1080
    },
    {
      "epoch": 13.62,
      "grad_norm": 0.8186032176017761,
      "learning_rate": 0.0016166666666666666,
      "loss": 0.4611,
      "step": 1090
    },
    {
      "epoch": 13.75,
      "grad_norm": 0.7720199227333069,
      "learning_rate": 0.0016111111111111111,
      "loss": 0.5397,
      "step": 1100
    },
    {
      "epoch": 13.88,
      "grad_norm": 0.6797453165054321,
      "learning_rate": 0.0016055555555555556,
      "loss": 0.5135,
      "step": 1110
    },
    {
      "epoch": 14.0,
      "grad_norm": 0.726184606552124,
      "learning_rate": 0.0016,
      "loss": 0.5175,
      "step": 1120
    },
    {
      "epoch": 14.0,
      "eval_accuracy": 0.7841796875,
      "eval_loss": 0.5278125405311584,
      "eval_runtime": 4.8906,
      "eval_samples_per_second": 209.383,
      "eval_steps_per_second": 13.086,
      "step": 1120
    },
    {
      "epoch": 14.12,
      "grad_norm": 0.6777172088623047,
      "learning_rate": 0.0015944444444444446,
      "loss": 0.4831,
      "step": 1130
    },
    {
      "epoch": 14.25,
      "grad_norm": 0.6228752732276917,
      "learning_rate": 0.0015888888888888888,
      "loss": 0.4657,
      "step": 1140
    },
    {
      "epoch": 14.38,
      "grad_norm": 0.7296370267868042,
      "learning_rate": 0.0015833333333333333,
      "loss": 0.5084,
      "step": 1150
    },
    {
      "epoch": 14.5,
      "grad_norm": 0.7809439897537231,
      "learning_rate": 0.0015777777777777778,
      "loss": 0.4749,
      "step": 1160
    },
    {
      "epoch": 14.62,
      "grad_norm": 0.4627506136894226,
      "learning_rate": 0.0015722222222222223,
      "loss": 0.4157,
      "step": 1170
    },
    {
      "epoch": 14.75,
      "grad_norm": 0.465811163187027,
      "learning_rate": 0.0015666666666666667,
      "loss": 0.4192,
      "step": 1180
    },
    {
      "epoch": 14.88,
      "grad_norm": 0.636384129524231,
      "learning_rate": 0.0015611111111111112,
      "loss": 0.4627,
      "step": 1190
    },
    {
      "epoch": 15.0,
      "grad_norm": 0.8339561223983765,
      "learning_rate": 0.0015555555555555557,
      "loss": 0.4865,
      "step": 1200
    },
    {
      "epoch": 15.0,
      "eval_accuracy": 0.837890625,
      "eval_loss": 0.4163576364517212,
      "eval_runtime": 5.3394,
      "eval_samples_per_second": 191.782,
      "eval_steps_per_second": 11.986,
      "step": 1200
    },
    {
      "epoch": 15.12,
      "grad_norm": 0.5218497514724731,
      "learning_rate": 0.0015500000000000002,
      "loss": 0.4253,
      "step": 1210
    },
    {
      "epoch": 15.25,
      "grad_norm": 0.6273193359375,
      "learning_rate": 0.0015444444444444446,
      "loss": 0.4474,
      "step": 1220
    },
    {
      "epoch": 15.38,
      "grad_norm": 0.6019622087478638,
      "learning_rate": 0.001538888888888889,
      "loss": 0.4008,
      "step": 1230
    },
    {
      "epoch": 15.5,
      "grad_norm": 0.7020573616027832,
      "learning_rate": 0.0015333333333333334,
      "loss": 0.3768,
      "step": 1240
    },
    {
      "epoch": 15.62,
      "grad_norm": 0.577691376209259,
      "learning_rate": 0.0015277777777777776,
      "loss": 0.4108,
      "step": 1250
    },
    {
      "epoch": 15.75,
      "grad_norm": 0.8489026427268982,
      "learning_rate": 0.0015222222222222221,
      "loss": 0.3994,
      "step": 1260
    },
    {
      "epoch": 15.88,
      "grad_norm": 0.42233309149742126,
      "learning_rate": 0.0015166666666666666,
      "loss": 0.4292,
      "step": 1270
    },
    {
      "epoch": 16.0,
      "grad_norm": 0.48867735266685486,
      "learning_rate": 0.001511111111111111,
      "loss": 0.4049,
      "step": 1280
    },
    {
      "epoch": 16.0,
      "eval_accuracy": 0.830078125,
      "eval_loss": 0.4204105734825134,
      "eval_runtime": 4.8855,
      "eval_samples_per_second": 209.602,
      "eval_steps_per_second": 13.1,
      "step": 1280
    },
    {
      "epoch": 16.12,
      "grad_norm": 0.6492818593978882,
      "learning_rate": 0.0015055555555555556,
      "loss": 0.3885,
      "step": 1290
    },
    {
      "epoch": 16.25,
      "grad_norm": 0.4546281397342682,
      "learning_rate": 0.0015,
      "loss": 0.4096,
      "step": 1300
    },
    {
      "epoch": 16.38,
      "grad_norm": 0.6827344298362732,
      "learning_rate": 0.0014944444444444445,
      "loss": 0.3618,
      "step": 1310
    },
    {
      "epoch": 16.5,
      "grad_norm": 0.454326868057251,
      "learning_rate": 0.001488888888888889,
      "loss": 0.3863,
      "step": 1320
    },
    {
      "epoch": 16.62,
      "grad_norm": 0.6911420226097107,
      "learning_rate": 0.0014833333333333335,
      "loss": 0.4264,
      "step": 1330
    },
    {
      "epoch": 16.75,
      "grad_norm": 0.6122339367866516,
      "learning_rate": 0.001477777777777778,
      "loss": 0.4205,
      "step": 1340
    },
    {
      "epoch": 16.88,
      "grad_norm": 0.5123728513717651,
      "learning_rate": 0.0014722222222222224,
      "loss": 0.4419,
      "step": 1350
    },
    {
      "epoch": 17.0,
      "grad_norm": 1.0908498764038086,
      "learning_rate": 0.0014666666666666667,
      "loss": 0.4167,
      "step": 1360
    },
    {
      "epoch": 17.0,
      "eval_accuracy": 0.828125,
      "eval_loss": 0.47203314304351807,
      "eval_runtime": 5.3495,
      "eval_samples_per_second": 191.421,
      "eval_steps_per_second": 11.964,
      "step": 1360
    },
    {
      "epoch": 17.12,
      "grad_norm": 0.42975571751594543,
      "learning_rate": 0.0014611111111111112,
      "loss": 0.4006,
      "step": 1370
    },
    {
      "epoch": 17.25,
      "grad_norm": 0.6392154693603516,
      "learning_rate": 0.0014555555555555554,
      "loss": 0.3581,
      "step": 1380
    },
    {
      "epoch": 17.38,
      "grad_norm": 0.6548070907592773,
      "learning_rate": 0.00145,
      "loss": 0.3672,
      "step": 1390
    },
    {
      "epoch": 17.5,
      "grad_norm": 0.6939528584480286,
      "learning_rate": 0.0014444444444444444,
      "loss": 0.3514,
      "step": 1400
    },
    {
      "epoch": 17.62,
      "grad_norm": 0.6098494529724121,
      "learning_rate": 0.0014388888888888889,
      "loss": 0.3835,
      "step": 1410
    },
    {
      "epoch": 17.75,
      "grad_norm": 0.5356572866439819,
      "learning_rate": 0.0014333333333333333,
      "loss": 0.3326,
      "step": 1420
    },
    {
      "epoch": 17.88,
      "grad_norm": 0.6472760438919067,
      "learning_rate": 0.0014277777777777778,
      "loss": 0.3829,
      "step": 1430
    },
    {
      "epoch": 18.0,
      "grad_norm": 0.67198646068573,
      "learning_rate": 0.0014222222222222223,
      "loss": 0.36,
      "step": 1440
    },
    {
      "epoch": 18.0,
      "eval_accuracy": 0.81640625,
      "eval_loss": 0.4660454988479614,
      "eval_runtime": 4.9124,
      "eval_samples_per_second": 208.451,
      "eval_steps_per_second": 13.028,
      "step": 1440
    },
    {
      "epoch": 18.12,
      "grad_norm": 0.4594449996948242,
      "learning_rate": 0.0014166666666666668,
      "loss": 0.3549,
      "step": 1450
    },
    {
      "epoch": 18.25,
      "grad_norm": 0.4456086754798889,
      "learning_rate": 0.0014111111111111112,
      "loss": 0.2899,
      "step": 1460
    },
    {
      "epoch": 18.38,
      "grad_norm": 0.724087119102478,
      "learning_rate": 0.0014055555555555555,
      "loss": 0.2976,
      "step": 1470
    },
    {
      "epoch": 18.5,
      "grad_norm": 0.8099024891853333,
      "learning_rate": 0.0014,
      "loss": 0.3706,
      "step": 1480
    },
    {
      "epoch": 18.62,
      "grad_norm": 0.6271733641624451,
      "learning_rate": 0.0013944444444444445,
      "loss": 0.3591,
      "step": 1490
    },
    {
      "epoch": 18.75,
      "grad_norm": 0.5864254236221313,
      "learning_rate": 0.001388888888888889,
      "loss": 0.3184,
      "step": 1500
    },
    {
      "epoch": 18.88,
      "grad_norm": 0.4915286898612976,
      "learning_rate": 0.0013833333333333334,
      "loss": 0.301,
      "step": 1510
    },
    {
      "epoch": 19.0,
      "grad_norm": 0.6932692527770996,
      "learning_rate": 0.001377777777777778,
      "loss": 0.3195,
      "step": 1520
    },
    {
      "epoch": 19.0,
      "eval_accuracy": 0.876953125,
      "eval_loss": 0.306354820728302,
      "eval_runtime": 5.2563,
      "eval_samples_per_second": 194.815,
      "eval_steps_per_second": 12.176,
      "step": 1520
    },
    {
      "epoch": 19.12,
      "grad_norm": 0.5778792500495911,
      "learning_rate": 0.0013722222222222222,
      "loss": 0.3493,
      "step": 1530
    },
    {
      "epoch": 19.25,
      "grad_norm": 0.951936662197113,
      "learning_rate": 0.0013666666666666666,
      "loss": 0.3305,
      "step": 1540
    },
    {
      "epoch": 19.38,
      "grad_norm": 0.6778426170349121,
      "learning_rate": 0.0013611111111111111,
      "loss": 0.32,
      "step": 1550
    },
    {
      "epoch": 19.5,
      "grad_norm": 0.6356533765792847,
      "learning_rate": 0.0013555555555555556,
      "loss": 0.2889,
      "step": 1560
    },
    {
      "epoch": 19.62,
      "grad_norm": 0.6476128697395325,
      "learning_rate": 0.00135,
      "loss": 0.2907,
      "step": 1570
    },
    {
      "epoch": 19.75,
      "grad_norm": 0.4664938151836395,
      "learning_rate": 0.0013444444444444445,
      "loss": 0.3261,
      "step": 1580
    },
    {
      "epoch": 19.88,
      "grad_norm": 1.06290602684021,
      "learning_rate": 0.0013388888888888888,
      "loss": 0.3365,
      "step": 1590
    },
    {
      "epoch": 20.0,
      "grad_norm": 0.5365467667579651,
      "learning_rate": 0.0013333333333333333,
      "loss": 0.3652,
      "step": 1600
    },
    {
      "epoch": 20.0,
      "eval_accuracy": 0.912109375,
      "eval_loss": 0.25709766149520874,
      "eval_runtime": 4.9952,
      "eval_samples_per_second": 204.995,
      "eval_steps_per_second": 12.812,
      "step": 1600
    },
    {
      "epoch": 20.12,
      "grad_norm": 0.5051919221878052,
      "learning_rate": 0.0013277777777777778,
      "loss": 0.3147,
      "step": 1610
    },
    {
      "epoch": 20.25,
      "grad_norm": 0.5098996162414551,
      "learning_rate": 0.0013222222222222222,
      "loss": 0.3085,
      "step": 1620
    },
    {
      "epoch": 20.38,
      "grad_norm": 0.5585361123085022,
      "learning_rate": 0.0013166666666666667,
      "loss": 0.3679,
      "step": 1630
    },
    {
      "epoch": 20.5,
      "grad_norm": 0.38560378551483154,
      "learning_rate": 0.0013111111111111112,
      "loss": 0.2987,
      "step": 1640
    },
    {
      "epoch": 20.62,
      "grad_norm": 0.3209057152271271,
      "learning_rate": 0.0013055555555555557,
      "loss": 0.2792,
      "step": 1650
    },
    {
      "epoch": 20.75,
      "grad_norm": 0.6471489667892456,
      "learning_rate": 0.0013000000000000002,
      "loss": 0.2755,
      "step": 1660
    },
    {
      "epoch": 20.88,
      "grad_norm": 0.8814804553985596,
      "learning_rate": 0.0012944444444444446,
      "loss": 0.2993,
      "step": 1670
    },
    {
      "epoch": 21.0,
      "grad_norm": 0.5392754673957825,
      "learning_rate": 0.001288888888888889,
      "loss": 0.2794,
      "step": 1680
    },
    {
      "epoch": 21.0,
      "eval_accuracy": 0.9150390625,
      "eval_loss": 0.24504294991493225,
      "eval_runtime": 4.8909,
      "eval_samples_per_second": 209.37,
      "eval_steps_per_second": 13.086,
      "step": 1680
    },
    {
      "epoch": 21.12,
      "grad_norm": 0.6234158873558044,
      "learning_rate": 0.0012833333333333334,
      "loss": 0.2926,
      "step": 1690
    },
    {
      "epoch": 21.25,
      "grad_norm": 0.4284802973270416,
      "learning_rate": 0.0012777777777777776,
      "loss": 0.2803,
      "step": 1700
    },
    {
      "epoch": 21.38,
      "grad_norm": 0.688140869140625,
      "learning_rate": 0.0012722222222222221,
      "loss": 0.2799,
      "step": 1710
    },
    {
      "epoch": 21.5,
      "grad_norm": 0.8576880097389221,
      "learning_rate": 0.0012666666666666666,
      "loss": 0.2868,
      "step": 1720
    },
    {
      "epoch": 21.62,
      "grad_norm": 0.6299762725830078,
      "learning_rate": 0.001261111111111111,
      "loss": 0.2971,
      "step": 1730
    },
    {
      "epoch": 21.75,
      "grad_norm": 0.7093678116798401,
      "learning_rate": 0.0012555555555555555,
      "loss": 0.2905,
      "step": 1740
    },
    {
      "epoch": 21.88,
      "grad_norm": 0.4271737039089203,
      "learning_rate": 0.00125,
      "loss": 0.3336,
      "step": 1750
    },
    {
      "epoch": 22.0,
      "grad_norm": 0.6771571040153503,
      "learning_rate": 0.0012444444444444445,
      "loss": 0.2704,
      "step": 1760
    },
    {
      "epoch": 22.0,
      "eval_accuracy": 0.9033203125,
      "eval_loss": 0.23907524347305298,
      "eval_runtime": 5.3054,
      "eval_samples_per_second": 193.012,
      "eval_steps_per_second": 12.063,
      "step": 1760
    },
    {
      "epoch": 22.12,
      "grad_norm": 0.44859397411346436,
      "learning_rate": 0.001238888888888889,
      "loss": 0.28,
      "step": 1770
    },
    {
      "epoch": 22.25,
      "grad_norm": 0.5617765784263611,
      "learning_rate": 0.0012333333333333335,
      "loss": 0.3093,
      "step": 1780
    },
    {
      "epoch": 22.38,
      "grad_norm": 0.6634913682937622,
      "learning_rate": 0.001227777777777778,
      "loss": 0.2417,
      "step": 1790
    },
    {
      "epoch": 22.5,
      "grad_norm": 0.670782744884491,
      "learning_rate": 0.0012222222222222224,
      "loss": 0.2932,
      "step": 1800
    },
    {
      "epoch": 22.62,
      "grad_norm": 0.6564796566963196,
      "learning_rate": 0.0012166666666666667,
      "loss": 0.3042,
      "step": 1810
    },
    {
      "epoch": 22.75,
      "grad_norm": 0.34089842438697815,
      "learning_rate": 0.0012111111111111112,
      "loss": 0.2925,
      "step": 1820
    },
    {
      "epoch": 22.88,
      "grad_norm": 0.5612368583679199,
      "learning_rate": 0.0012055555555555554,
      "loss": 0.2559,
      "step": 1830
    },
    {
      "epoch": 23.0,
      "grad_norm": 0.624458909034729,
      "learning_rate": 0.0012,
      "loss": 0.2612,
      "step": 1840
    },
    {
      "epoch": 23.0,
      "eval_accuracy": 0.927734375,
      "eval_loss": 0.23524078726768494,
      "eval_runtime": 4.902,
      "eval_samples_per_second": 208.893,
      "eval_steps_per_second": 13.056,
      "step": 1840
    },
    {
      "epoch": 23.12,
      "grad_norm": 0.6820557117462158,
      "learning_rate": 0.0011944444444444444,
      "loss": 0.2282,
      "step": 1850
    },
    {
      "epoch": 23.25,
      "grad_norm": 0.5979276895523071,
      "learning_rate": 0.0011888888888888889,
      "loss": 0.2569,
      "step": 1860
    },
    {
      "epoch": 23.38,
      "grad_norm": 0.5427021384239197,
      "learning_rate": 0.0011833333333333333,
      "loss": 0.2724,
      "step": 1870
    },
    {
      "epoch": 23.5,
      "grad_norm": 0.4382477104663849,
      "learning_rate": 0.0011777777777777778,
      "loss": 0.2616,
      "step": 1880
    },
    {
      "epoch": 23.62,
      "grad_norm": 0.6240445375442505,
      "learning_rate": 0.0011722222222222223,
      "loss": 0.2636,
      "step": 1890
    },
    {
      "epoch": 23.75,
      "grad_norm": 0.7440346479415894,
      "learning_rate": 0.0011666666666666668,
      "loss": 0.2913,
      "step": 1900
    },
    {
      "epoch": 23.88,
      "grad_norm": 0.4682701826095581,
      "learning_rate": 0.0011611111111111112,
      "loss": 0.2499,
      "step": 1910
    },
    {
      "epoch": 24.0,
      "grad_norm": 0.5112751722335815,
      "learning_rate": 0.0011555555555555555,
      "loss": 0.2425,
      "step": 1920
    },
    {
      "epoch": 24.0,
      "eval_accuracy": 0.828125,
      "eval_loss": 0.4720377027988434,
      "eval_runtime": 5.3156,
      "eval_samples_per_second": 192.639,
      "eval_steps_per_second": 12.04,
      "step": 1920
    },
    {
      "epoch": 24.12,
      "grad_norm": 0.765444278717041,
      "learning_rate": 0.00115,
      "loss": 0.2736,
      "step": 1930
    },
    {
      "epoch": 24.25,
      "grad_norm": 0.380066841840744,
      "learning_rate": 0.0011444444444444445,
      "loss": 0.2357,
      "step": 1940
    },
    {
      "epoch": 24.38,
      "grad_norm": 0.43320003151893616,
      "learning_rate": 0.001138888888888889,
      "loss": 0.2518,
      "step": 1950
    },
    {
      "epoch": 24.5,
      "grad_norm": 0.5003307461738586,
      "learning_rate": 0.0011333333333333334,
      "loss": 0.2898,
      "step": 1960
    },
    {
      "epoch": 24.62,
      "grad_norm": 0.41153478622436523,
      "learning_rate": 0.001127777777777778,
      "loss": 0.2209,
      "step": 1970
    },
    {
      "epoch": 24.75,
      "grad_norm": 0.41805940866470337,
      "learning_rate": 0.0011222222222222222,
      "loss": 0.235,
      "step": 1980
    },
    {
      "epoch": 24.88,
      "grad_norm": 0.5226410627365112,
      "learning_rate": 0.0011166666666666666,
      "loss": 0.2349,
      "step": 1990
    },
    {
      "epoch": 25.0,
      "grad_norm": 0.3767559826374054,
      "learning_rate": 0.0011111111111111111,
      "loss": 0.2567,
      "step": 2000
    },
    {
      "epoch": 25.0,
      "eval_accuracy": 0.9130859375,
      "eval_loss": 0.22960150241851807,
      "eval_runtime": 4.887,
      "eval_samples_per_second": 209.535,
      "eval_steps_per_second": 13.096,
      "step": 2000
    },
    {
      "epoch": 25.12,
      "grad_norm": 0.6860052943229675,
      "learning_rate": 0.0011055555555555556,
      "loss": 0.2426,
      "step": 2010
    },
    {
      "epoch": 25.25,
      "grad_norm": 0.3876688778400421,
      "learning_rate": 0.0011,
      "loss": 0.2243,
      "step": 2020
    },
    {
      "epoch": 25.38,
      "grad_norm": 0.3251183032989502,
      "learning_rate": 0.0010944444444444445,
      "loss": 0.234,
      "step": 2030
    },
    {
      "epoch": 25.5,
      "grad_norm": 0.5538493990898132,
      "learning_rate": 0.0010888888888888888,
      "loss": 0.2547,
      "step": 2040
    },
    {
      "epoch": 25.62,
      "grad_norm": 0.6539644598960876,
      "learning_rate": 0.0010833333333333333,
      "loss": 0.2382,
      "step": 2050
    },
    {
      "epoch": 25.75,
      "grad_norm": 0.6687932014465332,
      "learning_rate": 0.0010777777777777778,
      "loss": 0.2254,
      "step": 2060
    },
    {
      "epoch": 25.88,
      "grad_norm": 0.6210919618606567,
      "learning_rate": 0.0010722222222222222,
      "loss": 0.2356,
      "step": 2070
    },
    {
      "epoch": 26.0,
      "grad_norm": 0.5525135397911072,
      "learning_rate": 0.0010666666666666667,
      "loss": 0.2302,
      "step": 2080
    },
    {
      "epoch": 26.0,
      "eval_accuracy": 0.89453125,
      "eval_loss": 0.30673664808273315,
      "eval_runtime": 4.9576,
      "eval_samples_per_second": 206.552,
      "eval_steps_per_second": 12.909,
      "step": 2080
    },
    {
      "epoch": 26.12,
      "grad_norm": 0.5014208555221558,
      "learning_rate": 0.0010611111111111112,
      "loss": 0.2403,
      "step": 2090
    },
    {
      "epoch": 26.25,
      "grad_norm": 0.6093131303787231,
      "learning_rate": 0.0010555555555555557,
      "loss": 0.2356,
      "step": 2100
    },
    {
      "epoch": 26.38,
      "grad_norm": 0.3627248704433441,
      "learning_rate": 0.0010500000000000002,
      "loss": 0.2509,
      "step": 2110
    },
    {
      "epoch": 26.5,
      "grad_norm": 0.4119124114513397,
      "learning_rate": 0.0010444444444444446,
      "loss": 0.1915,
      "step": 2120
    },
    {
      "epoch": 26.62,
      "grad_norm": 0.5565811395645142,
      "learning_rate": 0.0010388888888888889,
      "loss": 0.191,
      "step": 2130
    },
    {
      "epoch": 26.75,
      "grad_norm": 0.44097578525543213,
      "learning_rate": 0.0010333333333333334,
      "loss": 0.2353,
      "step": 2140
    },
    {
      "epoch": 26.88,
      "grad_norm": 0.4542636275291443,
      "learning_rate": 0.0010277777777777776,
      "loss": 0.2144,
      "step": 2150
    },
    {
      "epoch": 27.0,
      "grad_norm": 0.4763772785663605,
      "learning_rate": 0.0010222222222222221,
      "loss": 0.2358,
      "step": 2160
    },
    {
      "epoch": 27.0,
      "eval_accuracy": 0.9375,
      "eval_loss": 0.17758239805698395,
      "eval_runtime": 5.3185,
      "eval_samples_per_second": 192.534,
      "eval_steps_per_second": 12.033,
      "step": 2160
    },
    {
      "epoch": 27.12,
      "grad_norm": 0.7219308614730835,
      "learning_rate": 0.0010166666666666666,
      "loss": 0.238,
      "step": 2170
    },
    {
      "epoch": 27.25,
      "grad_norm": 0.7707520127296448,
      "learning_rate": 0.001011111111111111,
      "loss": 0.1863,
      "step": 2180
    },
    {
      "epoch": 27.38,
      "grad_norm": 0.6878935098648071,
      "learning_rate": 0.0010055555555555555,
      "loss": 0.2493,
      "step": 2190
    },
    {
      "epoch": 27.5,
      "grad_norm": 0.5451861619949341,
      "learning_rate": 0.001,
      "loss": 0.2374,
      "step": 2200
    },
    {
      "epoch": 27.62,
      "grad_norm": 0.39642319083213806,
      "learning_rate": 0.0009944444444444445,
      "loss": 0.2382,
      "step": 2210
    },
    {
      "epoch": 27.75,
      "grad_norm": 0.4122956097126007,
      "learning_rate": 0.000988888888888889,
      "loss": 0.2176,
      "step": 2220
    },
    {
      "epoch": 27.88,
      "grad_norm": 0.6155421733856201,
      "learning_rate": 0.0009833333333333332,
      "loss": 0.2128,
      "step": 2230
    },
    {
      "epoch": 28.0,
      "grad_norm": 0.7283052206039429,
      "learning_rate": 0.0009777777777777777,
      "loss": 0.2173,
      "step": 2240
    },
    {
      "epoch": 28.0,
      "eval_accuracy": 0.94921875,
      "eval_loss": 0.15962785482406616,
      "eval_runtime": 4.9115,
      "eval_samples_per_second": 208.489,
      "eval_steps_per_second": 13.031,
      "step": 2240
    },
    {
      "epoch": 28.12,
      "grad_norm": 0.39027243852615356,
      "learning_rate": 0.0009722222222222222,
      "loss": 0.1979,
      "step": 2250
    },
    {
      "epoch": 28.25,
      "grad_norm": 0.5258718729019165,
      "learning_rate": 0.0009666666666666667,
      "loss": 0.1447,
      "step": 2260
    },
    {
      "epoch": 28.38,
      "grad_norm": 0.6615960001945496,
      "learning_rate": 0.0009611111111111112,
      "loss": 0.2403,
      "step": 2270
    },
    {
      "epoch": 28.5,
      "grad_norm": 0.4044310748577118,
      "learning_rate": 0.0009555555555555556,
      "loss": 0.1981,
      "step": 2280
    },
    {
      "epoch": 28.62,
      "grad_norm": 0.2666930556297302,
      "learning_rate": 0.00095,
      "loss": 0.2108,
      "step": 2290
    },
    {
      "epoch": 28.75,
      "grad_norm": 0.5612334609031677,
      "learning_rate": 0.0009444444444444445,
      "loss": 0.1783,
      "step": 2300
    },
    {
      "epoch": 28.88,
      "grad_norm": 0.48420026898384094,
      "learning_rate": 0.000938888888888889,
      "loss": 0.1848,
      "step": 2310
    },
    {
      "epoch": 29.0,
      "grad_norm": 0.5850337743759155,
      "learning_rate": 0.0009333333333333333,
      "loss": 0.1798,
      "step": 2320
    },
    {
      "epoch": 29.0,
      "eval_accuracy": 0.94140625,
      "eval_loss": 0.1548241674900055,
      "eval_runtime": 5.32,
      "eval_samples_per_second": 192.483,
      "eval_steps_per_second": 12.03,
      "step": 2320
    },
    {
      "epoch": 29.12,
      "grad_norm": 0.5059901475906372,
      "learning_rate": 0.0009277777777777778,
      "loss": 0.1954,
      "step": 2330
    },
    {
      "epoch": 29.25,
      "grad_norm": 0.22623513638973236,
      "learning_rate": 0.0009222222222222223,
      "loss": 0.1604,
      "step": 2340
    },
    {
      "epoch": 29.38,
      "grad_norm": 0.2330830693244934,
      "learning_rate": 0.0009166666666666666,
      "loss": 0.2125,
      "step": 2350
    },
    {
      "epoch": 29.5,
      "grad_norm": 0.4784901440143585,
      "learning_rate": 0.0009111111111111111,
      "loss": 0.1823,
      "step": 2360
    },
    {
      "epoch": 29.62,
      "grad_norm": 0.6156973242759705,
      "learning_rate": 0.0009055555555555556,
      "loss": 0.2289,
      "step": 2370
    },
    {
      "epoch": 29.75,
      "grad_norm": 0.4373360872268677,
      "learning_rate": 0.0009000000000000001,
      "loss": 0.2127,
      "step": 2380
    },
    {
      "epoch": 29.88,
      "grad_norm": 0.501115083694458,
      "learning_rate": 0.0008944444444444445,
      "loss": 0.2359,
      "step": 2390
    },
    {
      "epoch": 30.0,
      "grad_norm": 0.411662757396698,
      "learning_rate": 0.0008888888888888888,
      "loss": 0.197,
      "step": 2400
    },
    {
      "epoch": 30.0,
      "eval_accuracy": 0.95703125,
      "eval_loss": 0.17402663826942444,
      "eval_runtime": 4.919,
      "eval_samples_per_second": 208.172,
      "eval_steps_per_second": 13.011,
      "step": 2400
    },
    {
      "epoch": 30.12,
      "grad_norm": 0.45976510643959045,
      "learning_rate": 0.0008833333333333333,
      "loss": 0.1608,
      "step": 2410
    },
    {
      "epoch": 30.25,
      "grad_norm": 0.3243074417114258,
      "learning_rate": 0.0008777777777777778,
      "loss": 0.1742,
      "step": 2420
    },
    {
      "epoch": 30.38,
      "grad_norm": 0.5205725431442261,
      "learning_rate": 0.0008722222222222223,
      "loss": 0.1718,
      "step": 2430
    },
    {
      "epoch": 30.5,
      "grad_norm": 0.3976719081401825,
      "learning_rate": 0.0008666666666666667,
      "loss": 0.2247,
      "step": 2440
    },
    {
      "epoch": 30.62,
      "grad_norm": 0.2859196662902832,
      "learning_rate": 0.0008611111111111112,
      "loss": 0.1884,
      "step": 2450
    },
    {
      "epoch": 30.75,
      "grad_norm": 0.5310297012329102,
      "learning_rate": 0.0008555555555555556,
      "loss": 0.1672,
      "step": 2460
    },
    {
      "epoch": 30.88,
      "grad_norm": 0.5172590613365173,
      "learning_rate": 0.00085,
      "loss": 0.1828,
      "step": 2470
    },
    {
      "epoch": 31.0,
      "grad_norm": 0.6098745465278625,
      "learning_rate": 0.0008444444444444444,
      "loss": 0.1654,
      "step": 2480
    },
    {
      "epoch": 31.0,
      "eval_accuracy": 0.966796875,
      "eval_loss": 0.12167137861251831,
      "eval_runtime": 4.9956,
      "eval_samples_per_second": 204.98,
      "eval_steps_per_second": 12.811,
      "step": 2480
    },
    {
      "epoch": 31.12,
      "grad_norm": 0.3343498706817627,
      "learning_rate": 0.0008388888888888889,
      "loss": 0.1784,
      "step": 2490
    },
    {
      "epoch": 31.25,
      "grad_norm": 0.3938640058040619,
      "learning_rate": 0.0008333333333333334,
      "loss": 0.1697,
      "step": 2500
    },
    {
      "epoch": 31.38,
      "grad_norm": 0.41868484020233154,
      "learning_rate": 0.0008277777777777778,
      "loss": 0.2263,
      "step": 2510
    },
    {
      "epoch": 31.5,
      "grad_norm": 0.4363801181316376,
      "learning_rate": 0.0008222222222222222,
      "loss": 0.1762,
      "step": 2520
    },
    {
      "epoch": 31.62,
      "grad_norm": 0.5088948607444763,
      "learning_rate": 0.0008166666666666667,
      "loss": 0.1711,
      "step": 2530
    },
    {
      "epoch": 31.75,
      "grad_norm": 0.5423977375030518,
      "learning_rate": 0.0008111111111111111,
      "loss": 0.1675,
      "step": 2540
    },
    {
      "epoch": 31.88,
      "grad_norm": 0.431382954120636,
      "learning_rate": 0.0008055555555555556,
      "loss": 0.2216,
      "step": 2550
    },
    {
      "epoch": 32.0,
      "grad_norm": 0.4037337303161621,
      "learning_rate": 0.0008,
      "loss": 0.1896,
      "step": 2560
    },
    {
      "epoch": 32.0,
      "eval_accuracy": 0.92578125,
      "eval_loss": 0.2552070617675781,
      "eval_runtime": 5.2019,
      "eval_samples_per_second": 196.853,
      "eval_steps_per_second": 12.303,
      "step": 2560
    },
    {
      "epoch": 32.12,
      "grad_norm": 0.6025939583778381,
      "learning_rate": 0.0007944444444444444,
      "loss": 0.1926,
      "step": 2570
    },
    {
      "epoch": 32.25,
      "grad_norm": 0.7205588221549988,
      "learning_rate": 0.0007888888888888889,
      "loss": 0.1755,
      "step": 2580
    },
    {
      "epoch": 32.38,
      "grad_norm": 0.3841509222984314,
      "learning_rate": 0.0007833333333333334,
      "loss": 0.1696,
      "step": 2590
    },
    {
      "epoch": 32.5,
      "grad_norm": 0.5659075975418091,
      "learning_rate": 0.0007777777777777778,
      "loss": 0.133,
      "step": 2600
    },
    {
      "epoch": 32.62,
      "grad_norm": 0.7011501789093018,
      "learning_rate": 0.0007722222222222223,
      "loss": 0.2069,
      "step": 2610
    },
    {
      "epoch": 32.75,
      "grad_norm": 0.5933576822280884,
      "learning_rate": 0.0007666666666666667,
      "loss": 0.1799,
      "step": 2620
    },
    {
      "epoch": 32.88,
      "grad_norm": 0.636463463306427,
      "learning_rate": 0.0007611111111111111,
      "loss": 0.1884,
      "step": 2630
    },
    {
      "epoch": 33.0,
      "grad_norm": 0.36000609397888184,
      "learning_rate": 0.0007555555555555555,
      "loss": 0.1705,
      "step": 2640
    },
    {
      "epoch": 33.0,
      "eval_accuracy": 0.97265625,
      "eval_loss": 0.10305143892765045,
      "eval_runtime": 4.8746,
      "eval_samples_per_second": 210.07,
      "eval_steps_per_second": 13.129,
      "step": 2640
    },
    {
      "epoch": 33.12,
      "grad_norm": 0.25941601395606995,
      "learning_rate": 0.00075,
      "loss": 0.143,
      "step": 2650
    },
    {
      "epoch": 33.25,
      "grad_norm": 0.6486319899559021,
      "learning_rate": 0.0007444444444444445,
      "loss": 0.1819,
      "step": 2660
    },
    {
      "epoch": 33.38,
      "grad_norm": 0.34492290019989014,
      "learning_rate": 0.000738888888888889,
      "loss": 0.1877,
      "step": 2670
    },
    {
      "epoch": 33.5,
      "grad_norm": 0.5475990176200867,
      "learning_rate": 0.0007333333333333333,
      "loss": 0.1586,
      "step": 2680
    },
    {
      "epoch": 33.62,
      "grad_norm": 0.231631800532341,
      "learning_rate": 0.0007277777777777777,
      "loss": 0.145,
      "step": 2690
    },
    {
      "epoch": 33.75,
      "grad_norm": 0.6208530068397522,
      "learning_rate": 0.0007222222222222222,
      "loss": 0.2015,
      "step": 2700
    },
    {
      "epoch": 33.88,
      "grad_norm": 0.7229673862457275,
      "learning_rate": 0.0007166666666666667,
      "loss": 0.1814,
      "step": 2710
    },
    {
      "epoch": 34.0,
      "grad_norm": 0.38056522607803345,
      "learning_rate": 0.0007111111111111111,
      "loss": 0.1689,
      "step": 2720
    },
    {
      "epoch": 34.0,
      "eval_accuracy": 0.96875,
      "eval_loss": 0.10111749172210693,
      "eval_runtime": 5.2922,
      "eval_samples_per_second": 193.491,
      "eval_steps_per_second": 12.093,
      "step": 2720
    },
    {
      "epoch": 34.12,
      "grad_norm": 0.5405479669570923,
      "learning_rate": 0.0007055555555555556,
      "loss": 0.16,
      "step": 2730
    },
    {
      "epoch": 34.25,
      "grad_norm": 0.5781314373016357,
      "learning_rate": 0.0007,
      "loss": 0.1598,
      "step": 2740
    },
    {
      "epoch": 34.38,
      "grad_norm": 0.33385559916496277,
      "learning_rate": 0.0006944444444444445,
      "loss": 0.1747,
      "step": 2750
    },
    {
      "epoch": 34.5,
      "grad_norm": 0.36587977409362793,
      "learning_rate": 0.000688888888888889,
      "loss": 0.1376,
      "step": 2760
    },
    {
      "epoch": 34.62,
      "grad_norm": 0.3459375202655792,
      "learning_rate": 0.0006833333333333333,
      "loss": 0.1297,
      "step": 2770
    },
    {
      "epoch": 34.75,
      "grad_norm": 0.5182803273200989,
      "learning_rate": 0.0006777777777777778,
      "loss": 0.1747,
      "step": 2780
    },
    {
      "epoch": 34.88,
      "grad_norm": 0.39014366269111633,
      "learning_rate": 0.0006722222222222223,
      "loss": 0.169,
      "step": 2790
    },
    {
      "epoch": 35.0,
      "grad_norm": 0.4516375660896301,
      "learning_rate": 0.0006666666666666666,
      "loss": 0.1439,
      "step": 2800
    },
    {
      "epoch": 35.0,
      "eval_accuracy": 0.96484375,
      "eval_loss": 0.11748197674751282,
      "eval_runtime": 4.909,
      "eval_samples_per_second": 208.595,
      "eval_steps_per_second": 13.037,
      "step": 2800
    },
    {
      "epoch": 35.12,
      "grad_norm": 0.47782474756240845,
      "learning_rate": 0.0006611111111111111,
      "loss": 0.1417,
      "step": 2810
    },
    {
      "epoch": 35.25,
      "grad_norm": 0.11640643328428268,
      "learning_rate": 0.0006555555555555556,
      "loss": 0.1226,
      "step": 2820
    },
    {
      "epoch": 35.38,
      "grad_norm": 0.4363173544406891,
      "learning_rate": 0.0006500000000000001,
      "loss": 0.14,
      "step": 2830
    },
    {
      "epoch": 35.5,
      "grad_norm": 0.6676026582717896,
      "learning_rate": 0.0006444444444444444,
      "loss": 0.1548,
      "step": 2840
    },
    {
      "epoch": 35.62,
      "grad_norm": 0.4940982162952423,
      "learning_rate": 0.0006388888888888888,
      "loss": 0.1554,
      "step": 2850
    },
    {
      "epoch": 35.75,
      "grad_norm": 0.6478282809257507,
      "learning_rate": 0.0006333333333333333,
      "loss": 0.1641,
      "step": 2860
    },
    {
      "epoch": 35.88,
      "grad_norm": 0.6007707715034485,
      "learning_rate": 0.0006277777777777778,
      "loss": 0.1484,
      "step": 2870
    },
    {
      "epoch": 36.0,
      "grad_norm": 0.4945576786994934,
      "learning_rate": 0.0006222222222222223,
      "loss": 0.1606,
      "step": 2880
    },
    {
      "epoch": 36.0,
      "eval_accuracy": 0.9443359375,
      "eval_loss": 0.18046385049819946,
      "eval_runtime": 5.0626,
      "eval_samples_per_second": 202.266,
      "eval_steps_per_second": 12.642,
      "step": 2880
    },
    {
      "epoch": 36.12,
      "grad_norm": 0.4033058285713196,
      "learning_rate": 0.0006166666666666667,
      "loss": 0.1372,
      "step": 2890
    },
    {
      "epoch": 36.25,
      "grad_norm": 0.30507412552833557,
      "learning_rate": 0.0006111111111111112,
      "loss": 0.1549,
      "step": 2900
    },
    {
      "epoch": 36.38,
      "grad_norm": 0.3899296820163727,
      "learning_rate": 0.0006055555555555556,
      "loss": 0.1667,
      "step": 2910
    },
    {
      "epoch": 36.5,
      "grad_norm": 0.44058963656425476,
      "learning_rate": 0.0006,
      "loss": 0.1712,
      "step": 2920
    },
    {
      "epoch": 36.62,
      "grad_norm": 0.4805178642272949,
      "learning_rate": 0.0005944444444444444,
      "loss": 0.1805,
      "step": 2930
    },
    {
      "epoch": 36.75,
      "grad_norm": 0.37880581617355347,
      "learning_rate": 0.0005888888888888889,
      "loss": 0.1411,
      "step": 2940
    },
    {
      "epoch": 36.88,
      "grad_norm": 0.4263412654399872,
      "learning_rate": 0.0005833333333333334,
      "loss": 0.1714,
      "step": 2950
    },
    {
      "epoch": 37.0,
      "grad_norm": 0.2723836898803711,
      "learning_rate": 0.0005777777777777778,
      "loss": 0.1281,
      "step": 2960
    },
    {
      "epoch": 37.0,
      "eval_accuracy": 0.9677734375,
      "eval_loss": 0.1253870278596878,
      "eval_runtime": 5.1275,
      "eval_samples_per_second": 199.709,
      "eval_steps_per_second": 12.482,
      "step": 2960
    },
    {
      "epoch": 37.12,
      "grad_norm": 0.4946765601634979,
      "learning_rate": 0.0005722222222222222,
      "loss": 0.1059,
      "step": 2970
    },
    {
      "epoch": 37.25,
      "grad_norm": 0.4709372818470001,
      "learning_rate": 0.0005666666666666667,
      "loss": 0.1321,
      "step": 2980
    },
    {
      "epoch": 37.38,
      "grad_norm": 0.36459285020828247,
      "learning_rate": 0.0005611111111111111,
      "loss": 0.1351,
      "step": 2990
    },
    {
      "epoch": 37.5,
      "grad_norm": 0.4145031273365021,
      "learning_rate": 0.0005555555555555556,
      "loss": 0.1545,
      "step": 3000
    },
    {
      "epoch": 37.62,
      "grad_norm": 0.5457221865653992,
      "learning_rate": 0.00055,
      "loss": 0.1424,
      "step": 3010
    },
    {
      "epoch": 37.75,
      "grad_norm": 0.5123695731163025,
      "learning_rate": 0.0005444444444444444,
      "loss": 0.1508,
      "step": 3020
    },
    {
      "epoch": 37.88,
      "grad_norm": 0.29368171095848083,
      "learning_rate": 0.0005388888888888889,
      "loss": 0.1438,
      "step": 3030
    },
    {
      "epoch": 38.0,
      "grad_norm": 0.6859858632087708,
      "learning_rate": 0.0005333333333333334,
      "loss": 0.1518,
      "step": 3040
    },
    {
      "epoch": 38.0,
      "eval_accuracy": 0.96484375,
      "eval_loss": 0.11837992072105408,
      "eval_runtime": 4.9042,
      "eval_samples_per_second": 208.8,
      "eval_steps_per_second": 13.05,
      "step": 3040
    },
    {
      "epoch": 38.12,
      "grad_norm": 0.3859548270702362,
      "learning_rate": 0.0005277777777777778,
      "loss": 0.1455,
      "step": 3050
    },
    {
      "epoch": 38.25,
      "grad_norm": 0.21001270413398743,
      "learning_rate": 0.0005222222222222223,
      "loss": 0.13,
      "step": 3060
    },
    {
      "epoch": 38.38,
      "grad_norm": 0.4814240038394928,
      "learning_rate": 0.0005166666666666667,
      "loss": 0.1291,
      "step": 3070
    },
    {
      "epoch": 38.5,
      "grad_norm": 0.4478558301925659,
      "learning_rate": 0.0005111111111111111,
      "loss": 0.1293,
      "step": 3080
    },
    {
      "epoch": 38.62,
      "grad_norm": 0.4811321496963501,
      "learning_rate": 0.0005055555555555555,
      "loss": 0.1231,
      "step": 3090
    },
    {
      "epoch": 38.75,
      "grad_norm": 0.2841961085796356,
      "learning_rate": 0.0005,
      "loss": 0.166,
      "step": 3100
    },
    {
      "epoch": 38.88,
      "grad_norm": 0.5479158759117126,
      "learning_rate": 0.0004944444444444445,
      "loss": 0.1044,
      "step": 3110
    },
    {
      "epoch": 39.0,
      "grad_norm": 0.37449321150779724,
      "learning_rate": 0.0004888888888888889,
      "loss": 0.1531,
      "step": 3120
    },
    {
      "epoch": 39.0,
      "eval_accuracy": 0.9736328125,
      "eval_loss": 0.09921471774578094,
      "eval_runtime": 5.3451,
      "eval_samples_per_second": 191.577,
      "eval_steps_per_second": 11.974,
      "step": 3120
    },
    {
      "epoch": 39.12,
      "grad_norm": 0.5961503386497498,
      "learning_rate": 0.00048333333333333334,
      "loss": 0.1321,
      "step": 3130
    },
    {
      "epoch": 39.25,
      "grad_norm": 0.3140615224838257,
      "learning_rate": 0.0004777777777777778,
      "loss": 0.1192,
      "step": 3140
    },
    {
      "epoch": 39.38,
      "grad_norm": 0.8949409127235413,
      "learning_rate": 0.00047222222222222224,
      "loss": 0.122,
      "step": 3150
    },
    {
      "epoch": 39.5,
      "grad_norm": 0.21187840402126312,
      "learning_rate": 0.00046666666666666666,
      "loss": 0.1341,
      "step": 3160
    },
    {
      "epoch": 39.62,
      "grad_norm": 0.6364386081695557,
      "learning_rate": 0.00046111111111111114,
      "loss": 0.1327,
      "step": 3170
    },
    {
      "epoch": 39.75,
      "grad_norm": 0.2257820963859558,
      "learning_rate": 0.00045555555555555556,
      "loss": 0.1101,
      "step": 3180
    },
    {
      "epoch": 39.88,
      "grad_norm": 0.373692125082016,
      "learning_rate": 0.00045000000000000004,
      "loss": 0.1293,
      "step": 3190
    },
    {
      "epoch": 40.0,
      "grad_norm": 0.48990318179130554,
      "learning_rate": 0.0004444444444444444,
      "loss": 0.132,
      "step": 3200
    },
    {
      "epoch": 40.0,
      "eval_accuracy": 0.9775390625,
      "eval_loss": 0.09202806651592255,
      "eval_runtime": 4.9155,
      "eval_samples_per_second": 208.319,
      "eval_steps_per_second": 13.02,
      "step": 3200
    },
    {
      "epoch": 40.12,
      "grad_norm": 0.571524441242218,
      "learning_rate": 0.0004388888888888889,
      "loss": 0.1167,
      "step": 3210
    },
    {
      "epoch": 40.25,
      "grad_norm": 0.5896998643875122,
      "learning_rate": 0.00043333333333333337,
      "loss": 0.151,
      "step": 3220
    },
    {
      "epoch": 40.38,
      "grad_norm": 0.44366732239723206,
      "learning_rate": 0.0004277777777777778,
      "loss": 0.1422,
      "step": 3230
    },
    {
      "epoch": 40.5,
      "grad_norm": 0.314609169960022,
      "learning_rate": 0.0004222222222222222,
      "loss": 0.1253,
      "step": 3240
    },
    {
      "epoch": 40.62,
      "grad_norm": 0.3513747453689575,
      "learning_rate": 0.0004166666666666667,
      "loss": 0.119,
      "step": 3250
    },
    {
      "epoch": 40.75,
      "grad_norm": 0.3717803359031677,
      "learning_rate": 0.0004111111111111111,
      "loss": 0.1156,
      "step": 3260
    },
    {
      "epoch": 40.88,
      "grad_norm": 0.22342754900455475,
      "learning_rate": 0.00040555555555555554,
      "loss": 0.1372,
      "step": 3270
    },
    {
      "epoch": 41.0,
      "grad_norm": 0.41738444566726685,
      "learning_rate": 0.0004,
      "loss": 0.134,
      "step": 3280
    },
    {
      "epoch": 41.0,
      "eval_accuracy": 0.9638671875,
      "eval_loss": 0.13908132910728455,
      "eval_runtime": 5.2165,
      "eval_samples_per_second": 196.299,
      "eval_steps_per_second": 12.269,
      "step": 3280
    },
    {
      "epoch": 41.12,
      "grad_norm": 0.41814348101615906,
      "learning_rate": 0.00039444444444444444,
      "loss": 0.1279,
      "step": 3290
    },
    {
      "epoch": 41.25,
      "grad_norm": 0.9678131937980652,
      "learning_rate": 0.0003888888888888889,
      "loss": 0.1398,
      "step": 3300
    },
    {
      "epoch": 41.38,
      "grad_norm": 0.6725767850875854,
      "learning_rate": 0.00038333333333333334,
      "loss": 0.1492,
      "step": 3310
    },
    {
      "epoch": 41.5,
      "grad_norm": 0.31534790992736816,
      "learning_rate": 0.00037777777777777777,
      "loss": 0.1119,
      "step": 3320
    },
    {
      "epoch": 41.62,
      "grad_norm": 0.632583737373352,
      "learning_rate": 0.00037222222222222225,
      "loss": 0.1131,
      "step": 3330
    },
    {
      "epoch": 41.75,
      "grad_norm": 0.6746741533279419,
      "learning_rate": 0.00036666666666666667,
      "loss": 0.1351,
      "step": 3340
    },
    {
      "epoch": 41.88,
      "grad_norm": 0.3400849997997284,
      "learning_rate": 0.0003611111111111111,
      "loss": 0.0815,
      "step": 3350
    },
    {
      "epoch": 42.0,
      "grad_norm": 0.5605281591415405,
      "learning_rate": 0.00035555555555555557,
      "loss": 0.1413,
      "step": 3360
    },
    {
      "epoch": 42.0,
      "eval_accuracy": 0.9716796875,
      "eval_loss": 0.11220287531614304,
      "eval_runtime": 5.0927,
      "eval_samples_per_second": 201.072,
      "eval_steps_per_second": 12.567,
      "step": 3360
    },
    {
      "epoch": 42.12,
      "grad_norm": 0.5148097276687622,
      "learning_rate": 0.00035,
      "loss": 0.125,
      "step": 3370
    },
    {
      "epoch": 42.25,
      "grad_norm": 0.38650012016296387,
      "learning_rate": 0.0003444444444444445,
      "loss": 0.1209,
      "step": 3380
    },
    {
      "epoch": 42.38,
      "grad_norm": 0.3292187750339508,
      "learning_rate": 0.0003388888888888889,
      "loss": 0.1236,
      "step": 3390
    },
    {
      "epoch": 42.5,
      "grad_norm": 0.20681746304035187,
      "learning_rate": 0.0003333333333333333,
      "loss": 0.0973,
      "step": 3400
    },
    {
      "epoch": 42.62,
      "grad_norm": 0.33743348717689514,
      "learning_rate": 0.0003277777777777778,
      "loss": 0.1208,
      "step": 3410
    },
    {
      "epoch": 42.75,
      "grad_norm": 0.34158453345298767,
      "learning_rate": 0.0003222222222222222,
      "loss": 0.11,
      "step": 3420
    },
    {
      "epoch": 42.88,
      "grad_norm": 0.5730062127113342,
      "learning_rate": 0.00031666666666666665,
      "loss": 0.1292,
      "step": 3430
    },
    {
      "epoch": 43.0,
      "grad_norm": 0.44954267144203186,
      "learning_rate": 0.0003111111111111111,
      "loss": 0.1097,
      "step": 3440
    },
    {
      "epoch": 43.0,
      "eval_accuracy": 0.9677734375,
      "eval_loss": 0.11706902086734772,
      "eval_runtime": 4.9202,
      "eval_samples_per_second": 208.12,
      "eval_steps_per_second": 13.007,
      "step": 3440
    },
    {
      "epoch": 43.12,
      "grad_norm": 0.25731635093688965,
      "learning_rate": 0.0003055555555555556,
      "loss": 0.1161,
      "step": 3450
    },
    {
      "epoch": 43.25,
      "grad_norm": 0.5329569578170776,
      "learning_rate": 0.0003,
      "loss": 0.1507,
      "step": 3460
    },
    {
      "epoch": 43.38,
      "grad_norm": 0.3034692704677582,
      "learning_rate": 0.00029444444444444445,
      "loss": 0.1447,
      "step": 3470
    },
    {
      "epoch": 43.5,
      "grad_norm": 0.5483482480049133,
      "learning_rate": 0.0002888888888888889,
      "loss": 0.1323,
      "step": 3480
    },
    {
      "epoch": 43.62,
      "grad_norm": 0.279697984457016,
      "learning_rate": 0.00028333333333333335,
      "loss": 0.1,
      "step": 3490
    },
    {
      "epoch": 43.75,
      "grad_norm": 0.5593113303184509,
      "learning_rate": 0.0002777777777777778,
      "loss": 0.1169,
      "step": 3500
    },
    {
      "epoch": 43.88,
      "grad_norm": 0.621919572353363,
      "learning_rate": 0.0002722222222222222,
      "loss": 0.1119,
      "step": 3510
    },
    {
      "epoch": 44.0,
      "grad_norm": 0.37898024916648865,
      "learning_rate": 0.0002666666666666667,
      "loss": 0.1167,
      "step": 3520
    },
    {
      "epoch": 44.0,
      "eval_accuracy": 0.9765625,
      "eval_loss": 0.10542036592960358,
      "eval_runtime": 5.3473,
      "eval_samples_per_second": 191.5,
      "eval_steps_per_second": 11.969,
      "step": 3520
    },
    {
      "epoch": 44.12,
      "grad_norm": 0.40025296807289124,
      "learning_rate": 0.00026111111111111116,
      "loss": 0.1107,
      "step": 3530
    },
    {
      "epoch": 44.25,
      "grad_norm": 0.19010861217975616,
      "learning_rate": 0.00025555555555555553,
      "loss": 0.1008,
      "step": 3540
    },
    {
      "epoch": 44.38,
      "grad_norm": 0.33224934339523315,
      "learning_rate": 0.00025,
      "loss": 0.1355,
      "step": 3550
    },
    {
      "epoch": 44.5,
      "grad_norm": 0.4298325181007385,
      "learning_rate": 0.00024444444444444443,
      "loss": 0.106,
      "step": 3560
    },
    {
      "epoch": 44.62,
      "grad_norm": 0.4320330023765564,
      "learning_rate": 0.0002388888888888889,
      "loss": 0.1053,
      "step": 3570
    },
    {
      "epoch": 44.75,
      "grad_norm": 0.1121302917599678,
      "learning_rate": 0.00023333333333333333,
      "loss": 0.0845,
      "step": 3580
    },
    {
      "epoch": 44.88,
      "grad_norm": 0.3021819293498993,
      "learning_rate": 0.00022777777777777778,
      "loss": 0.1222,
      "step": 3590
    },
    {
      "epoch": 45.0,
      "grad_norm": 0.7353653311729431,
      "learning_rate": 0.0002222222222222222,
      "loss": 0.1388,
      "step": 3600
    },
    {
      "epoch": 45.0,
      "eval_accuracy": 0.9794921875,
      "eval_loss": 0.09323666244745255,
      "eval_runtime": 4.9368,
      "eval_samples_per_second": 207.422,
      "eval_steps_per_second": 12.964,
      "step": 3600
    },
    {
      "epoch": 45.12,
      "grad_norm": 0.5964930057525635,
      "learning_rate": 0.00021666666666666668,
      "loss": 0.1201,
      "step": 3610
    },
    {
      "epoch": 45.25,
      "grad_norm": 0.17329342663288116,
      "learning_rate": 0.0002111111111111111,
      "loss": 0.0905,
      "step": 3620
    },
    {
      "epoch": 45.38,
      "grad_norm": 0.5378609299659729,
      "learning_rate": 0.00020555555555555556,
      "loss": 0.0981,
      "step": 3630
    },
    {
      "epoch": 45.5,
      "grad_norm": 0.3457593619823456,
      "learning_rate": 0.0002,
      "loss": 0.1116,
      "step": 3640
    },
    {
      "epoch": 45.62,
      "grad_norm": 0.5954685211181641,
      "learning_rate": 0.00019444444444444446,
      "loss": 0.1037,
      "step": 3650
    },
    {
      "epoch": 45.75,
      "grad_norm": 0.1786712259054184,
      "learning_rate": 0.00018888888888888888,
      "loss": 0.0978,
      "step": 3660
    },
    {
      "epoch": 45.88,
      "grad_norm": 0.25224894285202026,
      "learning_rate": 0.00018333333333333334,
      "loss": 0.1089,
      "step": 3670
    },
    {
      "epoch": 46.0,
      "grad_norm": 0.33607247471809387,
      "learning_rate": 0.00017777777777777779,
      "loss": 0.1221,
      "step": 3680
    },
    {
      "epoch": 46.0,
      "eval_accuracy": 0.9765625,
      "eval_loss": 0.09462323784828186,
      "eval_runtime": 5.2287,
      "eval_samples_per_second": 195.844,
      "eval_steps_per_second": 12.24,
      "step": 3680
    },
    {
      "epoch": 46.12,
      "grad_norm": 0.34634700417518616,
      "learning_rate": 0.00017222222222222224,
      "loss": 0.1243,
      "step": 3690
    },
    {
      "epoch": 46.25,
      "grad_norm": 0.5061681866645813,
      "learning_rate": 0.00016666666666666666,
      "loss": 0.1115,
      "step": 3700
    },
    {
      "epoch": 46.38,
      "grad_norm": 0.2837713658809662,
      "learning_rate": 0.0001611111111111111,
      "loss": 0.1008,
      "step": 3710
    },
    {
      "epoch": 46.5,
      "grad_norm": 0.2688066363334656,
      "learning_rate": 0.00015555555555555556,
      "loss": 0.1058,
      "step": 3720
    },
    {
      "epoch": 46.62,
      "grad_norm": 0.32675421237945557,
      "learning_rate": 0.00015,
      "loss": 0.0897,
      "step": 3730
    },
    {
      "epoch": 46.75,
      "grad_norm": 0.6959260702133179,
      "learning_rate": 0.00014444444444444444,
      "loss": 0.1182,
      "step": 3740
    },
    {
      "epoch": 46.88,
      "grad_norm": 0.3018099069595337,
      "learning_rate": 0.0001388888888888889,
      "loss": 0.1013,
      "step": 3750
    },
    {
      "epoch": 47.0,
      "grad_norm": 0.6018778085708618,
      "learning_rate": 0.00013333333333333334,
      "loss": 0.1099,
      "step": 3760
    },
    {
      "epoch": 47.0,
      "eval_accuracy": 0.9755859375,
      "eval_loss": 0.1115545928478241,
      "eval_runtime": 5.0799,
      "eval_samples_per_second": 201.581,
      "eval_steps_per_second": 12.599,
      "step": 3760
    },
    {
      "epoch": 47.12,
      "grad_norm": 0.42199546098709106,
      "learning_rate": 0.00012777777777777776,
      "loss": 0.1073,
      "step": 3770
    },
    {
      "epoch": 47.25,
      "grad_norm": 0.6451756358146667,
      "learning_rate": 0.00012222222222222221,
      "loss": 0.099,
      "step": 3780
    },
    {
      "epoch": 47.38,
      "grad_norm": 0.4935210943222046,
      "learning_rate": 0.00011666666666666667,
      "loss": 0.1077,
      "step": 3790
    },
    {
      "epoch": 47.5,
      "grad_norm": 0.2563684582710266,
      "learning_rate": 0.0001111111111111111,
      "loss": 0.0907,
      "step": 3800
    },
    {
      "epoch": 47.62,
      "grad_norm": 0.3351310193538666,
      "learning_rate": 0.00010555555555555555,
      "loss": 0.1059,
      "step": 3810
    },
    {
      "epoch": 47.75,
      "grad_norm": 0.39526107907295227,
      "learning_rate": 0.0001,
      "loss": 0.0868,
      "step": 3820
    },
    {
      "epoch": 47.88,
      "grad_norm": 0.4634101390838623,
      "learning_rate": 9.444444444444444e-05,
      "loss": 0.1098,
      "step": 3830
    },
    {
      "epoch": 48.0,
      "grad_norm": 0.5983624458312988,
      "learning_rate": 8.888888888888889e-05,
      "loss": 0.1041,
      "step": 3840
    },
    {
      "epoch": 48.0,
      "eval_accuracy": 0.974609375,
      "eval_loss": 0.11264081299304962,
      "eval_runtime": 4.9279,
      "eval_samples_per_second": 207.795,
      "eval_steps_per_second": 12.987,
      "step": 3840
    },
    {
      "epoch": 48.12,
      "grad_norm": 0.4093017578125,
      "learning_rate": 8.333333333333333e-05,
      "loss": 0.1134,
      "step": 3850
    },
    {
      "epoch": 48.25,
      "grad_norm": 0.6668171286582947,
      "learning_rate": 7.777777777777778e-05,
      "loss": 0.0948,
      "step": 3860
    },
    {
      "epoch": 48.38,
      "grad_norm": 0.24066688120365143,
      "learning_rate": 7.222222222222222e-05,
      "loss": 0.0958,
      "step": 3870
    },
    {
      "epoch": 48.5,
      "grad_norm": 0.2770562469959259,
      "learning_rate": 6.666666666666667e-05,
      "loss": 0.1021,
      "step": 3880
    },
    {
      "epoch": 48.62,
      "grad_norm": 0.45978790521621704,
      "learning_rate": 6.111111111111111e-05,
      "loss": 0.1084,
      "step": 3890
    },
    {
      "epoch": 48.75,
      "grad_norm": 0.594672441482544,
      "learning_rate": 5.555555555555555e-05,
      "loss": 0.1373,
      "step": 3900
    },
    {
      "epoch": 48.88,
      "grad_norm": 0.8167428374290466,
      "learning_rate": 5e-05,
      "loss": 0.1038,
      "step": 3910
    },
    {
      "epoch": 49.0,
      "grad_norm": 0.2987329661846161,
      "learning_rate": 4.4444444444444447e-05,
      "loss": 0.1025,
      "step": 3920
    },
    {
      "epoch": 49.0,
      "eval_accuracy": 0.9755859375,
      "eval_loss": 0.11138872057199478,
      "eval_runtime": 5.3184,
      "eval_samples_per_second": 192.54,
      "eval_steps_per_second": 12.034,
      "step": 3920
    },
    {
      "epoch": 49.12,
      "grad_norm": 0.3884102404117584,
      "learning_rate": 3.888888888888889e-05,
      "loss": 0.1018,
      "step": 3930
    },
    {
      "epoch": 49.25,
      "grad_norm": 0.2661769688129425,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.1011,
      "step": 3940
    },
    {
      "epoch": 49.38,
      "grad_norm": 0.40820014476776123,
      "learning_rate": 2.7777777777777776e-05,
      "loss": 0.1488,
      "step": 3950
    },
    {
      "epoch": 49.5,
      "grad_norm": 0.46163231134414673,
      "learning_rate": 2.2222222222222223e-05,
      "loss": 0.1258,
      "step": 3960
    },
    {
      "epoch": 49.62,
      "grad_norm": 0.4315054416656494,
      "learning_rate": 1.6666666666666667e-05,
      "loss": 0.1018,
      "step": 3970
    },
    {
      "epoch": 49.75,
      "grad_norm": 0.2365369200706482,
      "learning_rate": 1.1111111111111112e-05,
      "loss": 0.0977,
      "step": 3980
    },
    {
      "epoch": 49.88,
      "grad_norm": 0.4910149574279785,
      "learning_rate": 5.555555555555556e-06,
      "loss": 0.1122,
      "step": 3990
    },
    {
      "epoch": 50.0,
      "grad_norm": 0.2623092234134674,
      "learning_rate": 0.0,
      "loss": 0.0887,
      "step": 4000
    },
    {
      "epoch": 50.0,
      "eval_accuracy": 0.9755859375,
      "eval_loss": 0.10555899888277054,
      "eval_runtime": 4.9229,
      "eval_samples_per_second": 208.005,
      "eval_steps_per_second": 13.0,
      "step": 4000
    },
    {
      "epoch": 50.0,
      "step": 4000,
      "total_flos": 5.437210780237824e+18,
      "train_loss": 0.3629864407479763,
      "train_runtime": 3465.6999,
      "train_samples_per_second": 73.867,
      "train_steps_per_second": 1.154
    }
  ],
  "logging_steps": 10,
  "max_steps": 4000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 50,
  "save_steps": 500,
  "total_flos": 5.437210780237824e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}