{
  "best_metric": 0.9225346534653466,
  "best_model_checkpoint": "swin-finetuned-food101/checkpoint-3549",
  "epoch": 2.9993664202745514,
  "global_step": 3549,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 1.4084507042253521e-06,
      "loss": 4.6907,
      "step": 10
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.8169014084507042e-06,
      "loss": 4.679,
      "step": 20
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.225352112676056e-06,
      "loss": 4.6485,
      "step": 30
    },
    {
      "epoch": 0.03,
      "learning_rate": 5.6338028169014084e-06,
      "loss": 4.6353,
      "step": 40
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.042253521126762e-06,
      "loss": 4.6212,
      "step": 50
    },
    {
      "epoch": 0.05,
      "learning_rate": 8.450704225352112e-06,
      "loss": 4.5806,
      "step": 60
    },
    {
      "epoch": 0.06,
      "learning_rate": 9.859154929577465e-06,
      "loss": 4.5613,
      "step": 70
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.1267605633802817e-05,
      "loss": 4.4989,
      "step": 80
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.267605633802817e-05,
      "loss": 4.4342,
      "step": 90
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.4084507042253523e-05,
      "loss": 4.3511,
      "step": 100
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.5492957746478872e-05,
      "loss": 4.2613,
      "step": 110
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.6901408450704224e-05,
      "loss": 4.0647,
      "step": 120
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.830985915492958e-05,
      "loss": 3.8823,
      "step": 130
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.971830985915493e-05,
      "loss": 3.6052,
      "step": 140
    },
    {
      "epoch": 0.13,
      "learning_rate": 2.112676056338028e-05,
      "loss": 3.236,
      "step": 150
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.2535211267605634e-05,
      "loss": 2.8905,
      "step": 160
    },
    {
      "epoch": 0.14,
      "learning_rate": 2.3943661971830986e-05,
      "loss": 2.6692,
      "step": 170
    },
    {
      "epoch": 0.15,
      "learning_rate": 2.535211267605634e-05,
      "loss": 2.4274,
      "step": 180
    },
    {
      "epoch": 0.16,
      "learning_rate": 2.676056338028169e-05,
      "loss": 2.2282,
      "step": 190
    },
    {
      "epoch": 0.17,
      "learning_rate": 2.8169014084507046e-05,
      "loss": 2.1145,
      "step": 200
    },
    {
      "epoch": 0.18,
      "learning_rate": 2.9577464788732395e-05,
      "loss": 1.9002,
      "step": 210
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.0985915492957744e-05,
      "loss": 1.7071,
      "step": 220
    },
    {
      "epoch": 0.19,
      "learning_rate": 3.23943661971831e-05,
      "loss": 1.4122,
      "step": 230
    },
    {
      "epoch": 0.2,
      "learning_rate": 3.380281690140845e-05,
      "loss": 1.5119,
      "step": 240
    },
    {
      "epoch": 0.21,
      "learning_rate": 3.5211267605633805e-05,
      "loss": 1.3445,
      "step": 250
    },
    {
      "epoch": 0.22,
      "learning_rate": 3.661971830985916e-05,
      "loss": 1.2829,
      "step": 260
    },
    {
      "epoch": 0.23,
      "learning_rate": 3.802816901408451e-05,
      "loss": 1.2674,
      "step": 270
    },
    {
      "epoch": 0.24,
      "learning_rate": 3.943661971830986e-05,
      "loss": 1.2246,
      "step": 280
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.0845070422535214e-05,
      "loss": 1.1809,
      "step": 290
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.225352112676056e-05,
      "loss": 1.1334,
      "step": 300
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.366197183098591e-05,
      "loss": 0.9901,
      "step": 310
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.507042253521127e-05,
      "loss": 1.0773,
      "step": 320
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.647887323943662e-05,
      "loss": 1.1217,
      "step": 330
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.788732394366197e-05,
      "loss": 0.9982,
      "step": 340
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.929577464788733e-05,
      "loss": 1.0588,
      "step": 350
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.992172824045085e-05,
      "loss": 1.0729,
      "step": 360
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.9765184721352535e-05,
      "loss": 1.0962,
      "step": 370
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.960864120225423e-05,
      "loss": 1.0493,
      "step": 380
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.9452097683155916e-05,
      "loss": 0.928,
      "step": 390
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.9295554164057614e-05,
      "loss": 0.907,
      "step": 400
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.91390106449593e-05,
      "loss": 0.7965,
      "step": 410
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.8982467125860995e-05,
      "loss": 0.9437,
      "step": 420
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.882592360676268e-05,
      "loss": 0.9301,
      "step": 430
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.866938008766438e-05,
      "loss": 0.9389,
      "step": 440
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.851283656856606e-05,
      "loss": 0.8393,
      "step": 450
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.835629304946775e-05,
      "loss": 0.83,
      "step": 460
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.819974953036944e-05,
      "loss": 0.8122,
      "step": 470
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.804320601127113e-05,
      "loss": 0.8234,
      "step": 480
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.788666249217283e-05,
      "loss": 0.8109,
      "step": 490
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.7730118973074514e-05,
      "loss": 0.7209,
      "step": 500
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.757357545397621e-05,
      "loss": 0.9059,
      "step": 510
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.7417031934877896e-05,
      "loss": 0.8376,
      "step": 520
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.726048841577959e-05,
      "loss": 0.826,
      "step": 530
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.710394489668128e-05,
      "loss": 0.8173,
      "step": 540
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.6947401377582975e-05,
      "loss": 0.8283,
      "step": 550
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.679085785848466e-05,
      "loss": 0.7782,
      "step": 560
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.6634314339386356e-05,
      "loss": 0.8335,
      "step": 570
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.647777082028804e-05,
      "loss": 0.9126,
      "step": 580
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.632122730118974e-05,
      "loss": 0.7519,
      "step": 590
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.616468378209142e-05,
      "loss": 0.7175,
      "step": 600
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.600814026299312e-05,
      "loss": 0.7496,
      "step": 610
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.58515967438948e-05,
      "loss": 0.7107,
      "step": 620
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.56950532247965e-05,
      "loss": 0.7209,
      "step": 630
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.5538509705698184e-05,
      "loss": 0.7283,
      "step": 640
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.5381966186599875e-05,
      "loss": 0.7179,
      "step": 650
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.5225422667501566e-05,
      "loss": 0.7398,
      "step": 660
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.5068879148403256e-05,
      "loss": 0.6623,
      "step": 670
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.491233562930495e-05,
      "loss": 0.6737,
      "step": 680
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.475579211020664e-05,
      "loss": 0.694,
      "step": 690
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.459924859110833e-05,
      "loss": 0.7095,
      "step": 700
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.444270507201002e-05,
      "loss": 0.6975,
      "step": 710
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.428616155291171e-05,
      "loss": 0.6883,
      "step": 720
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.41296180338134e-05,
      "loss": 0.6585,
      "step": 730
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.397307451471509e-05,
      "loss": 0.6692,
      "step": 740
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.381653099561678e-05,
      "loss": 0.7412,
      "step": 750
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.365998747651848e-05,
      "loss": 0.6666,
      "step": 760
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.350344395742016e-05,
      "loss": 0.6931,
      "step": 770
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.334690043832186e-05,
      "loss": 0.6701,
      "step": 780
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.3190356919223545e-05,
      "loss": 0.728,
      "step": 790
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.303381340012524e-05,
      "loss": 0.6723,
      "step": 800
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.2877269881026926e-05,
      "loss": 0.6367,
      "step": 810
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.2720726361928624e-05,
      "loss": 0.7133,
      "step": 820
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.256418284283031e-05,
      "loss": 0.6657,
      "step": 830
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.2407639323732e-05,
      "loss": 0.699,
      "step": 840
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.225109580463369e-05,
      "loss": 0.6833,
      "step": 850
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.209455228553538e-05,
      "loss": 0.6456,
      "step": 860
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.193800876643707e-05,
      "loss": 0.6006,
      "step": 870
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.178146524733876e-05,
      "loss": 0.6458,
      "step": 880
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.162492172824045e-05,
      "loss": 0.6764,
      "step": 890
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.146837820914214e-05,
      "loss": 0.5941,
      "step": 900
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.131183469004383e-05,
      "loss": 0.6831,
      "step": 910
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.1155291170945524e-05,
      "loss": 0.7029,
      "step": 920
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.0998747651847215e-05,
      "loss": 0.706,
      "step": 930
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.0842204132748905e-05,
      "loss": 0.5995,
      "step": 940
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0685660613650596e-05,
      "loss": 0.6364,
      "step": 950
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.052911709455229e-05,
      "loss": 0.5712,
      "step": 960
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.037257357545398e-05,
      "loss": 0.776,
      "step": 970
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.021603005635567e-05,
      "loss": 0.5841,
      "step": 980
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.005948653725736e-05,
      "loss": 0.6662,
      "step": 990
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.990294301815905e-05,
      "loss": 0.6862,
      "step": 1000
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.974639949906074e-05,
      "loss": 0.6664,
      "step": 1010
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.958985597996243e-05,
      "loss": 0.6962,
      "step": 1020
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.943331246086412e-05,
      "loss": 0.5786,
      "step": 1030
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.927676894176581e-05,
      "loss": 0.6521,
      "step": 1040
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.91202254226675e-05,
      "loss": 0.6417,
      "step": 1050
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.8963681903569194e-05,
      "loss": 0.6313,
      "step": 1060
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.8807138384470884e-05,
      "loss": 0.5653,
      "step": 1070
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8650594865372575e-05,
      "loss": 0.6452,
      "step": 1080
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.8494051346274266e-05,
      "loss": 0.6581,
      "step": 1090
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.8337507827175957e-05,
      "loss": 0.5864,
      "step": 1100
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.818096430807765e-05,
      "loss": 0.6943,
      "step": 1110
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.802442078897934e-05,
      "loss": 0.5743,
      "step": 1120
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.786787726988103e-05,
      "loss": 0.6748,
      "step": 1130
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.771133375078272e-05,
      "loss": 0.6692,
      "step": 1140
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.755479023168441e-05,
      "loss": 0.5344,
      "step": 1150
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.73982467125861e-05,
      "loss": 0.5794,
      "step": 1160
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.724170319348779e-05,
      "loss": 0.6004,
      "step": 1170
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.708515967438948e-05,
      "loss": 0.5704,
      "step": 1180
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8905346534653465,
      "eval_loss": 0.3809664249420166,
      "eval_runtime": 524.3821,
      "eval_samples_per_second": 48.152,
      "eval_steps_per_second": 3.011,
      "step": 1183
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.692861615529117e-05,
      "loss": 0.4233,
      "step": 1190
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.6772072636192864e-05,
      "loss": 0.5106,
      "step": 1200
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.6615529117094554e-05,
      "loss": 0.4044,
      "step": 1210
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.6458985597996245e-05,
      "loss": 0.4241,
      "step": 1220
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.6302442078897936e-05,
      "loss": 0.5049,
      "step": 1230
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.6145898559799626e-05,
      "loss": 0.3672,
      "step": 1240
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.598935504070132e-05,
      "loss": 0.3603,
      "step": 1250
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.583281152160301e-05,
      "loss": 0.422,
      "step": 1260
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.56762680025047e-05,
      "loss": 0.369,
      "step": 1270
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.551972448340639e-05,
      "loss": 0.4055,
      "step": 1280
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.536318096430808e-05,
      "loss": 0.3909,
      "step": 1290
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.520663744520977e-05,
      "loss": 0.4086,
      "step": 1300
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.505009392611146e-05,
      "loss": 0.3994,
      "step": 1310
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.489355040701315e-05,
      "loss": 0.4539,
      "step": 1320
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.473700688791484e-05,
      "loss": 0.3762,
      "step": 1330
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.4580463368816533e-05,
      "loss": 0.3847,
      "step": 1340
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.4423919849718224e-05,
      "loss": 0.3887,
      "step": 1350
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.4267376330619915e-05,
      "loss": 0.4879,
      "step": 1360
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.4110832811521606e-05,
      "loss": 0.3502,
      "step": 1370
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.3954289292423296e-05,
      "loss": 0.3826,
      "step": 1380
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.379774577332499e-05,
      "loss": 0.4537,
      "step": 1390
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.364120225422668e-05,
      "loss": 0.4367,
      "step": 1400
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.348465873512837e-05,
      "loss": 0.3518,
      "step": 1410
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.332811521603006e-05,
      "loss": 0.3996,
      "step": 1420
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.317157169693175e-05,
      "loss": 0.2905,
      "step": 1430
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.301502817783344e-05,
      "loss": 0.3632,
      "step": 1440
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.285848465873513e-05,
      "loss": 0.3949,
      "step": 1450
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.270194113963682e-05,
      "loss": 0.3714,
      "step": 1460
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.2545397620538506e-05,
      "loss": 0.3842,
      "step": 1470
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.23888541014402e-05,
      "loss": 0.371,
      "step": 1480
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.223231058234189e-05,
      "loss": 0.4382,
      "step": 1490
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.2075767063243585e-05,
      "loss": 0.3941,
      "step": 1500
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.191922354414527e-05,
      "loss": 0.4186,
      "step": 1510
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.1762680025046966e-05,
      "loss": 0.3973,
      "step": 1520
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.160613650594865e-05,
      "loss": 0.3901,
      "step": 1530
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.144959298685035e-05,
      "loss": 0.4067,
      "step": 1540
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.129304946775204e-05,
      "loss": 0.4196,
      "step": 1550
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.113650594865373e-05,
      "loss": 0.3724,
      "step": 1560
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.097996242955542e-05,
      "loss": 0.4579,
      "step": 1570
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.082341891045711e-05,
      "loss": 0.4353,
      "step": 1580
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.06668753913588e-05,
      "loss": 0.4122,
      "step": 1590
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.051033187226049e-05,
      "loss": 0.364,
      "step": 1600
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.0353788353162182e-05,
      "loss": 0.4484,
      "step": 1610
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.019724483406387e-05,
      "loss": 0.3375,
      "step": 1620
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.0040701314965564e-05,
      "loss": 0.5085,
      "step": 1630
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.988415779586725e-05,
      "loss": 0.3881,
      "step": 1640
    },
    {
      "epoch": 1.39,
      "learning_rate": 2.9727614276768945e-05,
      "loss": 0.4138,
      "step": 1650
    },
    {
      "epoch": 1.4,
      "learning_rate": 2.9571070757670633e-05,
      "loss": 0.3907,
      "step": 1660
    },
    {
      "epoch": 1.41,
      "learning_rate": 2.9414527238572327e-05,
      "loss": 0.4817,
      "step": 1670
    },
    {
      "epoch": 1.42,
      "learning_rate": 2.9257983719474014e-05,
      "loss": 0.4401,
      "step": 1680
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.9101440200375708e-05,
      "loss": 0.3135,
      "step": 1690
    },
    {
      "epoch": 1.44,
      "learning_rate": 2.8944896681277395e-05,
      "loss": 0.4107,
      "step": 1700
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.878835316217909e-05,
      "loss": 0.3546,
      "step": 1710
    },
    {
      "epoch": 1.45,
      "learning_rate": 2.8631809643080777e-05,
      "loss": 0.4396,
      "step": 1720
    },
    {
      "epoch": 1.46,
      "learning_rate": 2.8475266123982468e-05,
      "loss": 0.3604,
      "step": 1730
    },
    {
      "epoch": 1.47,
      "learning_rate": 2.8318722604884158e-05,
      "loss": 0.4266,
      "step": 1740
    },
    {
      "epoch": 1.48,
      "learning_rate": 2.816217908578585e-05,
      "loss": 0.4222,
      "step": 1750
    },
    {
      "epoch": 1.49,
      "learning_rate": 2.800563556668754e-05,
      "loss": 0.4185,
      "step": 1760
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.784909204758923e-05,
      "loss": 0.3434,
      "step": 1770
    },
    {
      "epoch": 1.5,
      "learning_rate": 2.7692548528490918e-05,
      "loss": 0.3645,
      "step": 1780
    },
    {
      "epoch": 1.51,
      "learning_rate": 2.7536005009392612e-05,
      "loss": 0.4132,
      "step": 1790
    },
    {
      "epoch": 1.52,
      "learning_rate": 2.7379461490294306e-05,
      "loss": 0.3973,
      "step": 1800
    },
    {
      "epoch": 1.53,
      "learning_rate": 2.7222917971195993e-05,
      "loss": 0.3852,
      "step": 1810
    },
    {
      "epoch": 1.54,
      "learning_rate": 2.7066374452097687e-05,
      "loss": 0.4473,
      "step": 1820
    },
    {
      "epoch": 1.55,
      "learning_rate": 2.6909830932999375e-05,
      "loss": 0.3972,
      "step": 1830
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.675328741390107e-05,
      "loss": 0.293,
      "step": 1840
    },
    {
      "epoch": 1.56,
      "learning_rate": 2.6596743894802756e-05,
      "loss": 0.3879,
      "step": 1850
    },
    {
      "epoch": 1.57,
      "learning_rate": 2.644020037570445e-05,
      "loss": 0.3674,
      "step": 1860
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.6283656856606137e-05,
      "loss": 0.3288,
      "step": 1870
    },
    {
      "epoch": 1.59,
      "learning_rate": 2.612711333750783e-05,
      "loss": 0.4182,
      "step": 1880
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.597056981840952e-05,
      "loss": 0.3469,
      "step": 1890
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.5814026299311213e-05,
      "loss": 0.3413,
      "step": 1900
    },
    {
      "epoch": 1.61,
      "learning_rate": 2.56574827802129e-05,
      "loss": 0.3579,
      "step": 1910
    },
    {
      "epoch": 1.62,
      "learning_rate": 2.550093926111459e-05,
      "loss": 0.3964,
      "step": 1920
    },
    {
      "epoch": 1.63,
      "learning_rate": 2.534439574201628e-05,
      "loss": 0.4302,
      "step": 1930
    },
    {
      "epoch": 1.64,
      "learning_rate": 2.5187852222917972e-05,
      "loss": 0.3402,
      "step": 1940
    },
    {
      "epoch": 1.65,
      "learning_rate": 2.5031308703819663e-05,
      "loss": 0.4623,
      "step": 1950
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.4874765184721354e-05,
      "loss": 0.3985,
      "step": 1960
    },
    {
      "epoch": 1.66,
      "learning_rate": 2.4718221665623044e-05,
      "loss": 0.297,
      "step": 1970
    },
    {
      "epoch": 1.67,
      "learning_rate": 2.4561678146524735e-05,
      "loss": 0.4284,
      "step": 1980
    },
    {
      "epoch": 1.68,
      "learning_rate": 2.4405134627426426e-05,
      "loss": 0.346,
      "step": 1990
    },
    {
      "epoch": 1.69,
      "learning_rate": 2.4248591108328117e-05,
      "loss": 0.3914,
      "step": 2000
    },
    {
      "epoch": 1.7,
      "learning_rate": 2.4092047589229807e-05,
      "loss": 0.3653,
      "step": 2010
    },
    {
      "epoch": 1.71,
      "learning_rate": 2.3935504070131498e-05,
      "loss": 0.3716,
      "step": 2020
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.377896055103319e-05,
      "loss": 0.3226,
      "step": 2030
    },
    {
      "epoch": 1.72,
      "learning_rate": 2.362241703193488e-05,
      "loss": 0.3124,
      "step": 2040
    },
    {
      "epoch": 1.73,
      "learning_rate": 2.346587351283657e-05,
      "loss": 0.3761,
      "step": 2050
    },
    {
      "epoch": 1.74,
      "learning_rate": 2.330932999373826e-05,
      "loss": 0.3992,
      "step": 2060
    },
    {
      "epoch": 1.75,
      "learning_rate": 2.315278647463995e-05,
      "loss": 0.3198,
      "step": 2070
    },
    {
      "epoch": 1.76,
      "learning_rate": 2.2996242955541642e-05,
      "loss": 0.3836,
      "step": 2080
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2839699436443333e-05,
      "loss": 0.3684,
      "step": 2090
    },
    {
      "epoch": 1.77,
      "learning_rate": 2.2683155917345024e-05,
      "loss": 0.4109,
      "step": 2100
    },
    {
      "epoch": 1.78,
      "learning_rate": 2.2526612398246714e-05,
      "loss": 0.3248,
      "step": 2110
    },
    {
      "epoch": 1.79,
      "learning_rate": 2.2370068879148405e-05,
      "loss": 0.3979,
      "step": 2120
    },
    {
      "epoch": 1.8,
      "learning_rate": 2.2213525360050096e-05,
      "loss": 0.3678,
      "step": 2130
    },
    {
      "epoch": 1.81,
      "learning_rate": 2.2056981840951786e-05,
      "loss": 0.4149,
      "step": 2140
    },
    {
      "epoch": 1.82,
      "learning_rate": 2.1900438321853474e-05,
      "loss": 0.3519,
      "step": 2150
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.1743894802755164e-05,
      "loss": 0.4255,
      "step": 2160
    },
    {
      "epoch": 1.83,
      "learning_rate": 2.1587351283656855e-05,
      "loss": 0.3771,
      "step": 2170
    },
    {
      "epoch": 1.84,
      "learning_rate": 2.1430807764558546e-05,
      "loss": 0.4284,
      "step": 2180
    },
    {
      "epoch": 1.85,
      "learning_rate": 2.127426424546024e-05,
      "loss": 0.3565,
      "step": 2190
    },
    {
      "epoch": 1.86,
      "learning_rate": 2.111772072636193e-05,
      "loss": 0.3508,
      "step": 2200
    },
    {
      "epoch": 1.87,
      "learning_rate": 2.096117720726362e-05,
      "loss": 0.4182,
      "step": 2210
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0804633688165312e-05,
      "loss": 0.3894,
      "step": 2220
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.0648090169067003e-05,
      "loss": 0.3566,
      "step": 2230
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.0491546649968694e-05,
      "loss": 0.3442,
      "step": 2240
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.0335003130870384e-05,
      "loss": 0.3525,
      "step": 2250
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.0178459611772075e-05,
      "loss": 0.3184,
      "step": 2260
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.0021916092673766e-05,
      "loss": 0.344,
      "step": 2270
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9865372573575456e-05,
      "loss": 0.4054,
      "step": 2280
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.9708829054477147e-05,
      "loss": 0.313,
      "step": 2290
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.9552285535378838e-05,
      "loss": 0.3411,
      "step": 2300
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.939574201628053e-05,
      "loss": 0.3272,
      "step": 2310
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.923919849718222e-05,
      "loss": 0.3692,
      "step": 2320
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.9082654978083906e-05,
      "loss": 0.4316,
      "step": 2330
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.8926111458985597e-05,
      "loss": 0.3872,
      "step": 2340
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.8769567939887288e-05,
      "loss": 0.2391,
      "step": 2350
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.861302442078898e-05,
      "loss": 0.307,
      "step": 2360
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9148514851485149,
      "eval_loss": 0.29710039496421814,
      "eval_runtime": 522.4602,
      "eval_samples_per_second": 48.329,
      "eval_steps_per_second": 3.022,
      "step": 2366
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.845648090169067e-05,
      "loss": 0.3321,
      "step": 2370
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.829993738259236e-05,
      "loss": 0.2162,
      "step": 2380
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.814339386349405e-05,
      "loss": 0.1984,
      "step": 2390
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.798685034439574e-05,
      "loss": 0.1896,
      "step": 2400
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.7830306825297432e-05,
      "loss": 0.1757,
      "step": 2410
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7673763306199123e-05,
      "loss": 0.1396,
      "step": 2420
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.7517219787100814e-05,
      "loss": 0.1653,
      "step": 2430
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.7360676268002504e-05,
      "loss": 0.1857,
      "step": 2440
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.7204132748904198e-05,
      "loss": 0.2153,
      "step": 2450
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.704758922980589e-05,
      "loss": 0.1996,
      "step": 2460
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.689104571070758e-05,
      "loss": 0.2101,
      "step": 2470
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.673450219160927e-05,
      "loss": 0.1711,
      "step": 2480
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.657795867251096e-05,
      "loss": 0.2192,
      "step": 2490
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.6421415153412652e-05,
      "loss": 0.2145,
      "step": 2500
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.6264871634314343e-05,
      "loss": 0.1849,
      "step": 2510
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.610832811521603e-05,
      "loss": 0.1968,
      "step": 2520
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.595178459611772e-05,
      "loss": 0.2043,
      "step": 2530
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.579524107701941e-05,
      "loss": 0.1686,
      "step": 2540
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5638697557921102e-05,
      "loss": 0.2046,
      "step": 2550
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5482154038822793e-05,
      "loss": 0.2238,
      "step": 2560
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.5325610519724483e-05,
      "loss": 0.2334,
      "step": 2570
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.5169067000626174e-05,
      "loss": 0.2326,
      "step": 2580
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.5012523481527865e-05,
      "loss": 0.2155,
      "step": 2590
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.4855979962429555e-05,
      "loss": 0.1822,
      "step": 2600
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4699436443331246e-05,
      "loss": 0.1817,
      "step": 2610
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.4542892924232937e-05,
      "loss": 0.2352,
      "step": 2620
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.4386349405134628e-05,
      "loss": 0.1988,
      "step": 2630
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.4229805886036318e-05,
      "loss": 0.2568,
      "step": 2640
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.4073262366938009e-05,
      "loss": 0.1821,
      "step": 2650
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.39167188478397e-05,
      "loss": 0.1966,
      "step": 2660
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.376017532874139e-05,
      "loss": 0.1798,
      "step": 2670
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.3603631809643081e-05,
      "loss": 0.2116,
      "step": 2680
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.344708829054477e-05,
      "loss": 0.2097,
      "step": 2690
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.329054477144646e-05,
      "loss": 0.2196,
      "step": 2700
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.3134001252348152e-05,
      "loss": 0.2156,
      "step": 2710
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.2977457733249846e-05,
      "loss": 0.2004,
      "step": 2720
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.2820914214151536e-05,
      "loss": 0.1982,
      "step": 2730
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2664370695053227e-05,
      "loss": 0.2343,
      "step": 2740
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.2507827175954918e-05,
      "loss": 0.1886,
      "step": 2750
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.2351283656856607e-05,
      "loss": 0.1739,
      "step": 2760
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.2194740137758297e-05,
      "loss": 0.1748,
      "step": 2770
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.2038196618659988e-05,
      "loss": 0.1591,
      "step": 2780
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.1881653099561679e-05,
      "loss": 0.1822,
      "step": 2790
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.172510958046337e-05,
      "loss": 0.2175,
      "step": 2800
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.156856606136506e-05,
      "loss": 0.1847,
      "step": 2810
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.1412022542266751e-05,
      "loss": 0.2193,
      "step": 2820
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.1255479023168442e-05,
      "loss": 0.2007,
      "step": 2830
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.1098935504070132e-05,
      "loss": 0.1632,
      "step": 2840
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.0942391984971823e-05,
      "loss": 0.2413,
      "step": 2850
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.0785848465873514e-05,
      "loss": 0.1942,
      "step": 2860
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0629304946775205e-05,
      "loss": 0.172,
      "step": 2870
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.0472761427676894e-05,
      "loss": 0.1765,
      "step": 2880
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.0316217908578584e-05,
      "loss": 0.1792,
      "step": 2890
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.0159674389480275e-05,
      "loss": 0.1938,
      "step": 2900
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.0003130870381967e-05,
      "loss": 0.1553,
      "step": 2910
    },
    {
      "epoch": 2.47,
      "learning_rate": 9.846587351283658e-06,
      "loss": 0.1649,
      "step": 2920
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.690043832185349e-06,
      "loss": 0.187,
      "step": 2930
    },
    {
      "epoch": 2.48,
      "learning_rate": 9.53350031308704e-06,
      "loss": 0.2572,
      "step": 2940
    },
    {
      "epoch": 2.49,
      "learning_rate": 9.37695679398873e-06,
      "loss": 0.1554,
      "step": 2950
    },
    {
      "epoch": 2.5,
      "learning_rate": 9.220413274890421e-06,
      "loss": 0.1555,
      "step": 2960
    },
    {
      "epoch": 2.51,
      "learning_rate": 9.06386975579211e-06,
      "loss": 0.1896,
      "step": 2970
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.9073262366938e-06,
      "loss": 0.2001,
      "step": 2980
    },
    {
      "epoch": 2.53,
      "learning_rate": 8.750782717595491e-06,
      "loss": 0.1689,
      "step": 2990
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.594239198497182e-06,
      "loss": 0.2367,
      "step": 3000
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.437695679398873e-06,
      "loss": 0.22,
      "step": 3010
    },
    {
      "epoch": 2.55,
      "learning_rate": 8.281152160300563e-06,
      "loss": 0.1816,
      "step": 3020
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.124608641202254e-06,
      "loss": 0.1556,
      "step": 3030
    },
    {
      "epoch": 2.57,
      "learning_rate": 7.968065122103947e-06,
      "loss": 0.1811,
      "step": 3040
    },
    {
      "epoch": 2.58,
      "learning_rate": 7.811521603005637e-06,
      "loss": 0.1788,
      "step": 3050
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.654978083907326e-06,
      "loss": 0.1935,
      "step": 3060
    },
    {
      "epoch": 2.59,
      "learning_rate": 7.498434564809018e-06,
      "loss": 0.1686,
      "step": 3070
    },
    {
      "epoch": 2.6,
      "learning_rate": 7.3418910457107085e-06,
      "loss": 0.204,
      "step": 3080
    },
    {
      "epoch": 2.61,
      "learning_rate": 7.185347526612398e-06,
      "loss": 0.1357,
      "step": 3090
    },
    {
      "epoch": 2.62,
      "learning_rate": 7.028804007514089e-06,
      "loss": 0.1574,
      "step": 3100
    },
    {
      "epoch": 2.63,
      "learning_rate": 6.87226048841578e-06,
      "loss": 0.1891,
      "step": 3110
    },
    {
      "epoch": 2.64,
      "learning_rate": 6.7157169693174705e-06,
      "loss": 0.1924,
      "step": 3120
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.559173450219161e-06,
      "loss": 0.1744,
      "step": 3130
    },
    {
      "epoch": 2.65,
      "learning_rate": 6.402629931120852e-06,
      "loss": 0.1921,
      "step": 3140
    },
    {
      "epoch": 2.66,
      "learning_rate": 6.2460864120225426e-06,
      "loss": 0.1213,
      "step": 3150
    },
    {
      "epoch": 2.67,
      "learning_rate": 6.089542892924233e-06,
      "loss": 0.1767,
      "step": 3160
    },
    {
      "epoch": 2.68,
      "learning_rate": 5.932999373825924e-06,
      "loss": 0.2516,
      "step": 3170
    },
    {
      "epoch": 2.69,
      "learning_rate": 5.776455854727615e-06,
      "loss": 0.1785,
      "step": 3180
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.619912335629305e-06,
      "loss": 0.1522,
      "step": 3190
    },
    {
      "epoch": 2.7,
      "learning_rate": 5.463368816530995e-06,
      "loss": 0.1759,
      "step": 3200
    },
    {
      "epoch": 2.71,
      "learning_rate": 5.306825297432687e-06,
      "loss": 0.226,
      "step": 3210
    },
    {
      "epoch": 2.72,
      "learning_rate": 5.1502817783343775e-06,
      "loss": 0.1422,
      "step": 3220
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.993738259236068e-06,
      "loss": 0.1358,
      "step": 3230
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.837194740137758e-06,
      "loss": 0.2554,
      "step": 3240
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.680651221039449e-06,
      "loss": 0.1617,
      "step": 3250
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.5241077019411395e-06,
      "loss": 0.1635,
      "step": 3260
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.367564182842831e-06,
      "loss": 0.194,
      "step": 3270
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.211020663744522e-06,
      "loss": 0.1859,
      "step": 3280
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.054477144646212e-06,
      "loss": 0.2416,
      "step": 3290
    },
    {
      "epoch": 2.79,
      "learning_rate": 3.897933625547902e-06,
      "loss": 0.1558,
      "step": 3300
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.741390106449593e-06,
      "loss": 0.1236,
      "step": 3310
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.5848465873512837e-06,
      "loss": 0.1467,
      "step": 3320
    },
    {
      "epoch": 2.81,
      "learning_rate": 3.428303068252974e-06,
      "loss": 0.2163,
      "step": 3330
    },
    {
      "epoch": 2.82,
      "learning_rate": 3.2717595491546655e-06,
      "loss": 0.2455,
      "step": 3340
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.115216030056356e-06,
      "loss": 0.1651,
      "step": 3350
    },
    {
      "epoch": 2.84,
      "learning_rate": 2.9586725109580465e-06,
      "loss": 0.164,
      "step": 3360
    },
    {
      "epoch": 2.85,
      "learning_rate": 2.8021289918597372e-06,
      "loss": 0.1701,
      "step": 3370
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.6455854727614275e-06,
      "loss": 0.1292,
      "step": 3380
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.4890419536631186e-06,
      "loss": 0.1596,
      "step": 3390
    },
    {
      "epoch": 2.87,
      "learning_rate": 2.332498434564809e-06,
      "loss": 0.2095,
      "step": 3400
    },
    {
      "epoch": 2.88,
      "learning_rate": 2.1759549154664996e-06,
      "loss": 0.1198,
      "step": 3410
    },
    {
      "epoch": 2.89,
      "learning_rate": 2.0194113963681908e-06,
      "loss": 0.1309,
      "step": 3420
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.862867877269881e-06,
      "loss": 0.224,
      "step": 3430
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.7063243581715717e-06,
      "loss": 0.2464,
      "step": 3440
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.5497808390732624e-06,
      "loss": 0.1782,
      "step": 3450
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.3932373199749531e-06,
      "loss": 0.1605,
      "step": 3460
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.2366938008766436e-06,
      "loss": 0.1797,
      "step": 3470
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.0801502817783343e-06,
      "loss": 0.1734,
      "step": 3480
    },
    {
      "epoch": 2.95,
      "learning_rate": 9.236067626800252e-07,
      "loss": 0.1678,
      "step": 3490
    },
    {
      "epoch": 2.96,
      "learning_rate": 7.670632435817158e-07,
      "loss": 0.1809,
      "step": 3500
    },
    {
      "epoch": 2.97,
      "learning_rate": 6.105197244834065e-07,
      "loss": 0.1637,
      "step": 3510
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.539762053850971e-07,
      "loss": 0.1766,
      "step": 3520
    },
    {
      "epoch": 2.98,
      "learning_rate": 2.974326862867877e-07,
      "loss": 0.1145,
      "step": 3530
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.408891671884784e-07,
      "loss": 0.1213,
      "step": 3540
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9225346534653466,
      "eval_loss": 0.2749262750148773,
      "eval_runtime": 521.2728,
      "eval_samples_per_second": 48.439,
      "eval_steps_per_second": 3.029,
      "step": 3549
    },
    {
      "epoch": 3.0,
      "step": 3549,
      "total_flos": 1.782187413176424e+19,
      "train_loss": 0.6323049043930561,
      "train_runtime": 11444.2247,
      "train_samples_per_second": 19.857,
      "train_steps_per_second": 0.31
    }
  ],
  "max_steps": 3549,
  "num_train_epochs": 3,
  "total_flos": 1.782187413176424e+19,
  "trial_name": null,
  "trial_params": null
}