{
  "best_metric": 0.0033987753558903933,
  "best_model_checkpoint": "./GTSRB_outputs/checkpoint-37494",
  "epoch": 10.0,
  "global_step": 41660,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.9995199231877104e-05,
      "loss": 3.7783,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.99903984637542e-05,
      "loss": 3.7505,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9985597695631302e-05,
      "loss": 3.7487,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9980796927508405e-05,
      "loss": 3.7393,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.99759961593855e-05,
      "loss": 3.6324,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.9971195391262603e-05,
      "loss": 3.6406,
      "step": 60
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9966394623139706e-05,
      "loss": 3.5461,
      "step": 70
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9961593855016805e-05,
      "loss": 3.558,
      "step": 80
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9956793086893904e-05,
      "loss": 3.4714,
      "step": 90
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9951992318771007e-05,
      "loss": 3.4558,
      "step": 100
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9947191550648106e-05,
      "loss": 3.4103,
      "step": 110
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9942390782525205e-05,
      "loss": 3.3774,
      "step": 120
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9937590014402308e-05,
      "loss": 3.2727,
      "step": 130
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.9932789246279407e-05,
      "loss": 3.2236,
      "step": 140
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9927988478156506e-05,
      "loss": 3.2661,
      "step": 150
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.992318771003361e-05,
      "loss": 3.203,
      "step": 160
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9918386941910707e-05,
      "loss": 3.1994,
      "step": 170
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9913586173787807e-05,
      "loss": 3.1309,
      "step": 180
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.990878540566491e-05,
      "loss": 3.0043,
      "step": 190
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.990398463754201e-05,
      "loss": 3.079,
      "step": 200
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.989918386941911e-05,
      "loss": 2.9737,
      "step": 210
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.989438310129621e-05,
      "loss": 2.9686,
      "step": 220
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.988958233317331e-05,
      "loss": 2.955,
      "step": 230
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9884781565050412e-05,
      "loss": 2.874,
      "step": 240
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.987998079692751e-05,
      "loss": 2.9016,
      "step": 250
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.987518002880461e-05,
      "loss": 2.8635,
      "step": 260
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9870379260681713e-05,
      "loss": 2.8191,
      "step": 270
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.986557849255881e-05,
      "loss": 2.8164,
      "step": 280
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.986077772443591e-05,
      "loss": 2.6601,
      "step": 290
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9855976956313013e-05,
      "loss": 2.6189,
      "step": 300
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.9851176188190113e-05,
      "loss": 2.6481,
      "step": 310
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.984637542006721e-05,
      "loss": 2.7157,
      "step": 320
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9841574651944314e-05,
      "loss": 2.7044,
      "step": 330
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9836773883821413e-05,
      "loss": 2.58,
      "step": 340
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.9831973115698513e-05,
      "loss": 2.6004,
      "step": 350
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9827172347575615e-05,
      "loss": 2.573,
      "step": 360
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9822371579452714e-05,
      "loss": 2.5859,
      "step": 370
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9817570811329813e-05,
      "loss": 2.5185,
      "step": 380
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.9812770043206916e-05,
      "loss": 2.3877,
      "step": 390
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9807969275084015e-05,
      "loss": 2.4955,
      "step": 400
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9803168506961114e-05,
      "loss": 2.4998,
      "step": 410
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9798367738838217e-05,
      "loss": 2.4137,
      "step": 420
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.9793566970715316e-05,
      "loss": 2.3628,
      "step": 430
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9788766202592415e-05,
      "loss": 2.3601,
      "step": 440
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9783965434469518e-05,
      "loss": 2.2108,
      "step": 450
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9779164666346617e-05,
      "loss": 2.2191,
      "step": 460
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.977436389822372e-05,
      "loss": 2.2868,
      "step": 470
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.976956313010082e-05,
      "loss": 2.2979,
      "step": 480
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.9764762361977918e-05,
      "loss": 2.0404,
      "step": 490
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.975996159385502e-05,
      "loss": 2.2322,
      "step": 500
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.975516082573212e-05,
      "loss": 2.2469,
      "step": 510
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.975036005760922e-05,
      "loss": 2.0574,
      "step": 520
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.974555928948632e-05,
      "loss": 2.2088,
      "step": 530
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.974075852136342e-05,
      "loss": 2.157,
      "step": 540
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.973595775324052e-05,
      "loss": 2.1264,
      "step": 550
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.9731156985117622e-05,
      "loss": 2.1088,
      "step": 560
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.972635621699472e-05,
      "loss": 2.027,
      "step": 570
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.972155544887182e-05,
      "loss": 2.0213,
      "step": 580
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9716754680748923e-05,
      "loss": 2.0719,
      "step": 590
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.9711953912626022e-05,
      "loss": 2.0945,
      "step": 600
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.970715314450312e-05,
      "loss": 1.968,
      "step": 610
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9702352376380223e-05,
      "loss": 1.869,
      "step": 620
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9697551608257323e-05,
      "loss": 1.8604,
      "step": 630
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9692750840134422e-05,
      "loss": 1.8648,
      "step": 640
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9687950072011524e-05,
      "loss": 1.9939,
      "step": 650
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9683149303888623e-05,
      "loss": 1.8737,
      "step": 660
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9678348535765723e-05,
      "loss": 1.7044,
      "step": 670
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.9673547767642825e-05,
      "loss": 1.8427,
      "step": 680
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9668746999519924e-05,
      "loss": 1.7928,
      "step": 690
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9663946231397023e-05,
      "loss": 1.9989,
      "step": 700
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9659145463274126e-05,
      "loss": 1.9756,
      "step": 710
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9654344695151225e-05,
      "loss": 1.8586,
      "step": 720
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9649543927028328e-05,
      "loss": 1.6752,
      "step": 730
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9644743158905427e-05,
      "loss": 1.8349,
      "step": 740
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9639942390782526e-05,
      "loss": 1.6647,
      "step": 750
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.963514162265963e-05,
      "loss": 1.5412,
      "step": 760
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.9630340854536728e-05,
      "loss": 1.6946,
      "step": 770
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9625540086413827e-05,
      "loss": 1.6583,
      "step": 780
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.962073931829093e-05,
      "loss": 1.6469,
      "step": 790
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.961593855016803e-05,
      "loss": 1.6358,
      "step": 800
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9611137782045128e-05,
      "loss": 1.6302,
      "step": 810
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.960633701392223e-05,
      "loss": 1.7555,
      "step": 820
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.960153624579933e-05,
      "loss": 1.8043,
      "step": 830
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.959673547767643e-05,
      "loss": 1.7428,
      "step": 840
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.959193470955353e-05,
      "loss": 1.4335,
      "step": 850
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.958713394143063e-05,
      "loss": 1.6664,
      "step": 860
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.958233317330773e-05,
      "loss": 1.5474,
      "step": 870
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9577532405184832e-05,
      "loss": 1.421,
      "step": 880
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.957273163706193e-05,
      "loss": 1.6877,
      "step": 890
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.956793086893903e-05,
      "loss": 1.3315,
      "step": 900
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9563130100816133e-05,
      "loss": 1.5427,
      "step": 910
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.9558329332693232e-05,
      "loss": 1.455,
      "step": 920
    },
    {
      "epoch": 0.22,
      "learning_rate": 1.955352856457033e-05,
      "loss": 1.4563,
      "step": 930
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9548727796447433e-05,
      "loss": 1.3555,
      "step": 940
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9543927028324533e-05,
      "loss": 1.4399,
      "step": 950
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9539126260201632e-05,
      "loss": 1.3208,
      "step": 960
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.9534325492078734e-05,
      "loss": 1.4741,
      "step": 970
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9529524723955833e-05,
      "loss": 1.4694,
      "step": 980
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9524723955832933e-05,
      "loss": 1.4141,
      "step": 990
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9519923187710035e-05,
      "loss": 1.3043,
      "step": 1000
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9515122419587134e-05,
      "loss": 1.3022,
      "step": 1010
    },
    {
      "epoch": 0.24,
      "learning_rate": 1.9510321651464237e-05,
      "loss": 1.3918,
      "step": 1020
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9505520883341336e-05,
      "loss": 1.3439,
      "step": 1030
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9500720115218435e-05,
      "loss": 1.3018,
      "step": 1040
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9495919347095538e-05,
      "loss": 1.5106,
      "step": 1050
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9491118578972637e-05,
      "loss": 1.2436,
      "step": 1060
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9486317810849736e-05,
      "loss": 1.3641,
      "step": 1070
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.948151704272684e-05,
      "loss": 1.2213,
      "step": 1080
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9476716274603938e-05,
      "loss": 1.4179,
      "step": 1090
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9471915506481037e-05,
      "loss": 1.125,
      "step": 1100
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.946711473835814e-05,
      "loss": 1.3662,
      "step": 1110
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.946231397023524e-05,
      "loss": 1.2618,
      "step": 1120
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.9457513202112338e-05,
      "loss": 1.241,
      "step": 1130
    },
    {
      "epoch": 0.27,
      "learning_rate": 1.945271243398944e-05,
      "loss": 1.2252,
      "step": 1140
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9447911665866543e-05,
      "loss": 1.1152,
      "step": 1150
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.944311089774364e-05,
      "loss": 1.2045,
      "step": 1160
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.943831012962074e-05,
      "loss": 1.0854,
      "step": 1170
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.9433509361497844e-05,
      "loss": 1.1037,
      "step": 1180
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.942870859337494e-05,
      "loss": 1.4213,
      "step": 1190
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9423907825252042e-05,
      "loss": 1.1631,
      "step": 1200
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.9419107057129144e-05,
      "loss": 1.0022,
      "step": 1210
    },
    {
      "epoch": 0.29,
      "learning_rate": 1.941430628900624e-05,
      "loss": 1.1225,
      "step": 1220
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9409505520883343e-05,
      "loss": 1.1233,
      "step": 1230
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9404704752760445e-05,
      "loss": 1.2195,
      "step": 1240
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.939990398463754e-05,
      "loss": 1.2476,
      "step": 1250
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9395103216514643e-05,
      "loss": 1.1084,
      "step": 1260
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9390302448391746e-05,
      "loss": 0.93,
      "step": 1270
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9385501680268845e-05,
      "loss": 0.9731,
      "step": 1280
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9380700912145944e-05,
      "loss": 1.1567,
      "step": 1290
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9375900144023047e-05,
      "loss": 0.9934,
      "step": 1300
    },
    {
      "epoch": 0.31,
      "learning_rate": 1.9371099375900146e-05,
      "loss": 1.1649,
      "step": 1310
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9366298607777245e-05,
      "loss": 1.0041,
      "step": 1320
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9361497839654348e-05,
      "loss": 1.003,
      "step": 1330
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9356697071531447e-05,
      "loss": 1.1013,
      "step": 1340
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9351896303408546e-05,
      "loss": 1.2216,
      "step": 1350
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.934709553528565e-05,
      "loss": 1.1426,
      "step": 1360
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9342294767162748e-05,
      "loss": 1.1875,
      "step": 1370
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.9337493999039847e-05,
      "loss": 1.1213,
      "step": 1380
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.933269323091695e-05,
      "loss": 1.0672,
      "step": 1390
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.932789246279405e-05,
      "loss": 1.0088,
      "step": 1400
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.932309169467115e-05,
      "loss": 1.1139,
      "step": 1410
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.931829092654825e-05,
      "loss": 1.0133,
      "step": 1420
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.931349015842535e-05,
      "loss": 0.9739,
      "step": 1430
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9308689390302452e-05,
      "loss": 0.9978,
      "step": 1440
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.930388862217955e-05,
      "loss": 1.0893,
      "step": 1450
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.929908785405665e-05,
      "loss": 1.0321,
      "step": 1460
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.9294287085933753e-05,
      "loss": 0.9259,
      "step": 1470
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9289486317810852e-05,
      "loss": 1.0544,
      "step": 1480
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.928468554968795e-05,
      "loss": 0.9128,
      "step": 1490
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9279884781565054e-05,
      "loss": 0.9296,
      "step": 1500
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9275084013442153e-05,
      "loss": 0.9006,
      "step": 1510
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9270283245319252e-05,
      "loss": 0.9222,
      "step": 1520
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9265482477196354e-05,
      "loss": 0.7412,
      "step": 1530
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9260681709073454e-05,
      "loss": 0.947,
      "step": 1540
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9255880940950553e-05,
      "loss": 0.8165,
      "step": 1550
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.9251080172827655e-05,
      "loss": 0.9555,
      "step": 1560
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9246279404704754e-05,
      "loss": 0.968,
      "step": 1570
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9241478636581854e-05,
      "loss": 0.9129,
      "step": 1580
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9236677868458956e-05,
      "loss": 0.8208,
      "step": 1590
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9231877100336055e-05,
      "loss": 1.0102,
      "step": 1600
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9227076332213154e-05,
      "loss": 0.8468,
      "step": 1610
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9222275564090257e-05,
      "loss": 0.8872,
      "step": 1620
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9217474795967356e-05,
      "loss": 0.8988,
      "step": 1630
    },
    {
      "epoch": 0.39,
      "learning_rate": 1.9212674027844455e-05,
      "loss": 0.8378,
      "step": 1640
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9207873259721558e-05,
      "loss": 0.6568,
      "step": 1650
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.9203072491598657e-05,
      "loss": 0.9033,
      "step": 1660
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.919827172347576e-05,
      "loss": 0.8359,
      "step": 1670
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.919347095535286e-05,
      "loss": 0.7811,
      "step": 1680
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.9188670187229958e-05,
      "loss": 0.8456,
      "step": 1690
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.918386941910706e-05,
      "loss": 0.7367,
      "step": 1700
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.917906865098416e-05,
      "loss": 0.831,
      "step": 1710
    },
    {
      "epoch": 0.41,
      "learning_rate": 1.917426788286126e-05,
      "loss": 0.9071,
      "step": 1720
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.916946711473836e-05,
      "loss": 1.0782,
      "step": 1730
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.916466634661546e-05,
      "loss": 0.7672,
      "step": 1740
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.915986557849256e-05,
      "loss": 0.9447,
      "step": 1750
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9155064810369662e-05,
      "loss": 0.7413,
      "step": 1760
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.915026404224676e-05,
      "loss": 0.81,
      "step": 1770
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.914546327412386e-05,
      "loss": 0.7186,
      "step": 1780
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9140662506000963e-05,
      "loss": 0.8462,
      "step": 1790
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.9135861737878062e-05,
      "loss": 0.8137,
      "step": 1800
    },
    {
      "epoch": 0.43,
      "learning_rate": 1.913106096975516e-05,
      "loss": 0.6924,
      "step": 1810
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9126260201632264e-05,
      "loss": 0.8786,
      "step": 1820
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9121459433509363e-05,
      "loss": 0.893,
      "step": 1830
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9116658665386462e-05,
      "loss": 0.955,
      "step": 1840
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9111857897263564e-05,
      "loss": 0.7753,
      "step": 1850
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9107057129140664e-05,
      "loss": 0.8304,
      "step": 1860
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9102256361017763e-05,
      "loss": 0.7471,
      "step": 1870
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9097455592894865e-05,
      "loss": 0.7336,
      "step": 1880
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.9092654824771964e-05,
      "loss": 0.7616,
      "step": 1890
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9087854056649064e-05,
      "loss": 0.8023,
      "step": 1900
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9083053288526166e-05,
      "loss": 0.7867,
      "step": 1910
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9078252520403265e-05,
      "loss": 0.6036,
      "step": 1920
    },
    {
      "epoch": 0.46,
      "learning_rate": 1.9073451752280368e-05,
      "loss": 0.8622,
      "step": 1930
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9068650984157467e-05,
      "loss": 0.8327,
      "step": 1940
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9063850216034566e-05,
      "loss": 0.6514,
      "step": 1950
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.905904944791167e-05,
      "loss": 0.8685,
      "step": 1960
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9054248679788768e-05,
      "loss": 0.7693,
      "step": 1970
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9049447911665867e-05,
      "loss": 0.803,
      "step": 1980
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.904464714354297e-05,
      "loss": 0.5931,
      "step": 1990
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.903984637542007e-05,
      "loss": 0.9011,
      "step": 2000
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.9035045607297168e-05,
      "loss": 0.7069,
      "step": 2010
    },
    {
      "epoch": 0.48,
      "learning_rate": 1.903024483917427e-05,
      "loss": 0.7496,
      "step": 2020
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.902544407105137e-05,
      "loss": 0.6518,
      "step": 2030
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.902064330292847e-05,
      "loss": 0.8198,
      "step": 2040
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.901584253480557e-05,
      "loss": 0.7376,
      "step": 2050
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.901104176668267e-05,
      "loss": 0.7526,
      "step": 2060
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.900624099855977e-05,
      "loss": 0.7426,
      "step": 2070
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.9001440230436872e-05,
      "loss": 0.6284,
      "step": 2080
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.899663946231397e-05,
      "loss": 0.6465,
      "step": 2090
    },
    {
      "epoch": 0.5,
      "learning_rate": 1.899183869419107e-05,
      "loss": 0.7576,
      "step": 2100
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8987037926068173e-05,
      "loss": 0.6778,
      "step": 2110
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8982237157945272e-05,
      "loss": 0.681,
      "step": 2120
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.897743638982237e-05,
      "loss": 0.636,
      "step": 2130
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.8972635621699474e-05,
      "loss": 0.6679,
      "step": 2140
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8967834853576573e-05,
      "loss": 0.7074,
      "step": 2150
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8963034085453672e-05,
      "loss": 0.5379,
      "step": 2160
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8958233317330774e-05,
      "loss": 0.6772,
      "step": 2170
    },
    {
      "epoch": 0.52,
      "learning_rate": 1.8953432549207874e-05,
      "loss": 0.5271,
      "step": 2180
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8948631781084976e-05,
      "loss": 0.3873,
      "step": 2190
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8943831012962075e-05,
      "loss": 0.7285,
      "step": 2200
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8939030244839174e-05,
      "loss": 0.7146,
      "step": 2210
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.8934229476716277e-05,
      "loss": 0.6375,
      "step": 2220
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8929428708593376e-05,
      "loss": 0.6244,
      "step": 2230
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8924627940470475e-05,
      "loss": 0.8262,
      "step": 2240
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8919827172347578e-05,
      "loss": 0.5245,
      "step": 2250
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8915026404224677e-05,
      "loss": 0.653,
      "step": 2260
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.8910225636101776e-05,
      "loss": 0.6866,
      "step": 2270
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.890542486797888e-05,
      "loss": 0.5555,
      "step": 2280
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8900624099855978e-05,
      "loss": 0.4962,
      "step": 2290
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.8895823331733077e-05,
      "loss": 0.6223,
      "step": 2300
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.889102256361018e-05,
      "loss": 0.586,
      "step": 2310
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.888622179548728e-05,
      "loss": 0.4495,
      "step": 2320
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8881421027364378e-05,
      "loss": 0.5812,
      "step": 2330
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.887662025924148e-05,
      "loss": 0.4158,
      "step": 2340
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.8871819491118583e-05,
      "loss": 0.6827,
      "step": 2350
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.886701872299568e-05,
      "loss": 0.4711,
      "step": 2360
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.886221795487278e-05,
      "loss": 0.7131,
      "step": 2370
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.8857417186749884e-05,
      "loss": 0.7031,
      "step": 2380
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.885261641862698e-05,
      "loss": 0.6428,
      "step": 2390
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8847815650504082e-05,
      "loss": 0.5477,
      "step": 2400
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8843014882381185e-05,
      "loss": 0.6406,
      "step": 2410
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.883821411425828e-05,
      "loss": 0.511,
      "step": 2420
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.8833413346135383e-05,
      "loss": 0.6828,
      "step": 2430
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8828612578012485e-05,
      "loss": 0.6954,
      "step": 2440
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.882381180988958e-05,
      "loss": 0.6661,
      "step": 2450
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8819011041766684e-05,
      "loss": 0.629,
      "step": 2460
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.8814210273643786e-05,
      "loss": 0.449,
      "step": 2470
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8809409505520885e-05,
      "loss": 0.5062,
      "step": 2480
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8804608737397985e-05,
      "loss": 0.4024,
      "step": 2490
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8799807969275087e-05,
      "loss": 0.6008,
      "step": 2500
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8795007201152186e-05,
      "loss": 0.5213,
      "step": 2510
    },
    {
      "epoch": 0.6,
      "learning_rate": 1.8790206433029285e-05,
      "loss": 0.4123,
      "step": 2520
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8785405664906388e-05,
      "loss": 0.5173,
      "step": 2530
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8780604896783487e-05,
      "loss": 0.5035,
      "step": 2540
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.8775804128660586e-05,
      "loss": 0.3853,
      "step": 2550
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.877100336053769e-05,
      "loss": 0.5554,
      "step": 2560
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8766202592414788e-05,
      "loss": 0.3964,
      "step": 2570
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.8761401824291887e-05,
      "loss": 0.7588,
      "step": 2580
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.875660105616899e-05,
      "loss": 0.6074,
      "step": 2590
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.875180028804609e-05,
      "loss": 0.433,
      "step": 2600
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.874699951992319e-05,
      "loss": 0.49,
      "step": 2610
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.874219875180029e-05,
      "loss": 0.503,
      "step": 2620
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.873739798367739e-05,
      "loss": 0.4721,
      "step": 2630
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.8732597215554492e-05,
      "loss": 0.6306,
      "step": 2640
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.872779644743159e-05,
      "loss": 0.4342,
      "step": 2650
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.872299567930869e-05,
      "loss": 0.4246,
      "step": 2660
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8718194911185793e-05,
      "loss": 0.6997,
      "step": 2670
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.8713394143062892e-05,
      "loss": 0.4625,
      "step": 2680
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.870859337493999e-05,
      "loss": 0.574,
      "step": 2690
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8703792606817094e-05,
      "loss": 0.4818,
      "step": 2700
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8698991838694193e-05,
      "loss": 0.4565,
      "step": 2710
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.8694191070571292e-05,
      "loss": 0.4671,
      "step": 2720
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8689390302448395e-05,
      "loss": 0.4953,
      "step": 2730
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8684589534325494e-05,
      "loss": 0.3733,
      "step": 2740
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8679788766202593e-05,
      "loss": 0.6231,
      "step": 2750
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8674987998079695e-05,
      "loss": 0.4676,
      "step": 2760
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.8670187229956795e-05,
      "loss": 0.3567,
      "step": 2770
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8665386461833894e-05,
      "loss": 0.4384,
      "step": 2780
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8660585693710996e-05,
      "loss": 0.5046,
      "step": 2790
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8655784925588095e-05,
      "loss": 0.453,
      "step": 2800
    },
    {
      "epoch": 0.67,
      "learning_rate": 1.8650984157465195e-05,
      "loss": 0.4219,
      "step": 2810
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8646183389342297e-05,
      "loss": 0.3992,
      "step": 2820
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8641382621219396e-05,
      "loss": 0.3865,
      "step": 2830
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8636581853096495e-05,
      "loss": 0.597,
      "step": 2840
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.8631781084973598e-05,
      "loss": 0.5623,
      "step": 2850
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8626980316850697e-05,
      "loss": 0.4672,
      "step": 2860
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.86221795487278e-05,
      "loss": 0.3366,
      "step": 2870
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.86173787806049e-05,
      "loss": 0.4285,
      "step": 2880
    },
    {
      "epoch": 0.69,
      "learning_rate": 1.8612578012481998e-05,
      "loss": 0.5842,
      "step": 2890
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.86077772443591e-05,
      "loss": 0.4076,
      "step": 2900
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.86029764762362e-05,
      "loss": 0.4258,
      "step": 2910
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.85981757081133e-05,
      "loss": 0.5203,
      "step": 2920
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.85933749399904e-05,
      "loss": 0.4427,
      "step": 2930
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.85885741718675e-05,
      "loss": 0.5219,
      "step": 2940
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.85837734037446e-05,
      "loss": 0.526,
      "step": 2950
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.8578972635621702e-05,
      "loss": 0.5013,
      "step": 2960
    },
    {
      "epoch": 0.71,
      "learning_rate": 1.85741718674988e-05,
      "loss": 0.5992,
      "step": 2970
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.85693710993759e-05,
      "loss": 0.5753,
      "step": 2980
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8564570331253003e-05,
      "loss": 0.5459,
      "step": 2990
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8559769563130102e-05,
      "loss": 0.6427,
      "step": 3000
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.85549687950072e-05,
      "loss": 0.3971,
      "step": 3010
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.8550168026884304e-05,
      "loss": 0.3084,
      "step": 3020
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8545367258761403e-05,
      "loss": 0.4304,
      "step": 3030
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8540566490638502e-05,
      "loss": 0.4354,
      "step": 3040
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8535765722515605e-05,
      "loss": 0.4845,
      "step": 3050
    },
    {
      "epoch": 0.73,
      "learning_rate": 1.8530964954392704e-05,
      "loss": 0.4229,
      "step": 3060
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8526164186269803e-05,
      "loss": 0.4911,
      "step": 3070
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8521363418146905e-05,
      "loss": 0.5619,
      "step": 3080
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8516562650024005e-05,
      "loss": 0.4871,
      "step": 3090
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.8511761881901104e-05,
      "loss": 0.3633,
      "step": 3100
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8506961113778206e-05,
      "loss": 0.3094,
      "step": 3110
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8502160345655305e-05,
      "loss": 0.5367,
      "step": 3120
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8497359577532408e-05,
      "loss": 0.4657,
      "step": 3130
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.8492558809409507e-05,
      "loss": 0.3797,
      "step": 3140
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.8487758041286606e-05,
      "loss": 0.5531,
      "step": 3150
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.848295727316371e-05,
      "loss": 0.5096,
      "step": 3160
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.8478156505040808e-05,
      "loss": 0.2796,
      "step": 3170
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.8473355736917907e-05,
      "loss": 0.4184,
      "step": 3180
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.846855496879501e-05,
      "loss": 0.4097,
      "step": 3190
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.846375420067211e-05,
      "loss": 0.4248,
      "step": 3200
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.8458953432549208e-05,
      "loss": 0.5682,
      "step": 3210
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.845415266442631e-05,
      "loss": 0.5558,
      "step": 3220
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.844935189630341e-05,
      "loss": 0.5888,
      "step": 3230
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.844455112818051e-05,
      "loss": 0.4789,
      "step": 3240
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.843975036005761e-05,
      "loss": 0.3556,
      "step": 3250
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.843494959193471e-05,
      "loss": 0.4935,
      "step": 3260
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.843014882381181e-05,
      "loss": 0.349,
      "step": 3270
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8425348055688912e-05,
      "loss": 0.4277,
      "step": 3280
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.842054728756601e-05,
      "loss": 0.3369,
      "step": 3290
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.841574651944311e-05,
      "loss": 0.4325,
      "step": 3300
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.8410945751320213e-05,
      "loss": 0.4881,
      "step": 3310
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8406144983197312e-05,
      "loss": 0.4303,
      "step": 3320
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.840134421507441e-05,
      "loss": 0.4241,
      "step": 3330
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8396543446951514e-05,
      "loss": 0.3176,
      "step": 3340
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.8391742678828613e-05,
      "loss": 0.3702,
      "step": 3350
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8386941910705712e-05,
      "loss": 0.4432,
      "step": 3360
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8382141142582815e-05,
      "loss": 0.4104,
      "step": 3370
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8377340374459914e-05,
      "loss": 0.5021,
      "step": 3380
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.8372539606337016e-05,
      "loss": 0.4312,
      "step": 3390
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8367738838214116e-05,
      "loss": 0.374,
      "step": 3400
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8362938070091215e-05,
      "loss": 0.3657,
      "step": 3410
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8358137301968317e-05,
      "loss": 0.317,
      "step": 3420
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.8353336533845416e-05,
      "loss": 0.4727,
      "step": 3430
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8348535765722515e-05,
      "loss": 0.418,
      "step": 3440
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8343734997599618e-05,
      "loss": 0.3723,
      "step": 3450
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8338934229476717e-05,
      "loss": 0.4062,
      "step": 3460
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.8334133461353816e-05,
      "loss": 0.299,
      "step": 3470
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.832933269323092e-05,
      "loss": 0.3177,
      "step": 3480
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.8324531925108018e-05,
      "loss": 0.2584,
      "step": 3490
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.8319731156985117e-05,
      "loss": 0.4107,
      "step": 3500
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.831493038886222e-05,
      "loss": 0.409,
      "step": 3510
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.8310129620739322e-05,
      "loss": 0.2987,
      "step": 3520
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8305328852616418e-05,
      "loss": 0.4629,
      "step": 3530
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.830052808449352e-05,
      "loss": 0.3649,
      "step": 3540
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.8295727316370623e-05,
      "loss": 0.2186,
      "step": 3550
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.829092654824772e-05,
      "loss": 0.5434,
      "step": 3560
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.828612578012482e-05,
      "loss": 0.3069,
      "step": 3570
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8281325012001924e-05,
      "loss": 0.3797,
      "step": 3580
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.827652424387902e-05,
      "loss": 0.3937,
      "step": 3590
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.8271723475756122e-05,
      "loss": 0.4565,
      "step": 3600
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8266922707633225e-05,
      "loss": 0.3195,
      "step": 3610
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.826212193951032e-05,
      "loss": 0.2143,
      "step": 3620
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8257321171387423e-05,
      "loss": 0.584,
      "step": 3630
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.8252520403264526e-05,
      "loss": 0.2804,
      "step": 3640
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8247719635141625e-05,
      "loss": 0.2962,
      "step": 3650
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8242918867018724e-05,
      "loss": 0.2373,
      "step": 3660
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8238118098895826e-05,
      "loss": 0.2916,
      "step": 3670
    },
    {
      "epoch": 0.88,
      "learning_rate": 1.8233317330772926e-05,
      "loss": 0.4911,
      "step": 3680
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8228516562650025e-05,
      "loss": 0.5212,
      "step": 3690
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8223715794527127e-05,
      "loss": 0.3899,
      "step": 3700
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8218915026404226e-05,
      "loss": 0.3755,
      "step": 3710
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.8214114258281326e-05,
      "loss": 0.4213,
      "step": 3720
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8209313490158428e-05,
      "loss": 0.3548,
      "step": 3730
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8204512722035527e-05,
      "loss": 0.2486,
      "step": 3740
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8199711953912626e-05,
      "loss": 0.4735,
      "step": 3750
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.819491118578973e-05,
      "loss": 0.2879,
      "step": 3760
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.8190110417666828e-05,
      "loss": 0.2308,
      "step": 3770
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8185309649543927e-05,
      "loss": 0.3713,
      "step": 3780
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.818050888142103e-05,
      "loss": 0.4296,
      "step": 3790
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.817570811329813e-05,
      "loss": 0.3481,
      "step": 3800
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.817090734517523e-05,
      "loss": 0.3113,
      "step": 3810
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.816610657705233e-05,
      "loss": 0.304,
      "step": 3820
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.816130580892943e-05,
      "loss": 0.4197,
      "step": 3830
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.8156505040806532e-05,
      "loss": 0.3564,
      "step": 3840
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.815170427268363e-05,
      "loss": 0.3632,
      "step": 3850
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.814690350456073e-05,
      "loss": 0.3188,
      "step": 3860
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8142102736437833e-05,
      "loss": 0.4399,
      "step": 3870
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.8137301968314932e-05,
      "loss": 0.3436,
      "step": 3880
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.813250120019203e-05,
      "loss": 0.3636,
      "step": 3890
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8127700432069134e-05,
      "loss": 0.381,
      "step": 3900
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8122899663946233e-05,
      "loss": 0.5667,
      "step": 3910
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8118098895823332e-05,
      "loss": 0.1597,
      "step": 3920
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.8113298127700435e-05,
      "loss": 0.3125,
      "step": 3930
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8108497359577534e-05,
      "loss": 0.3009,
      "step": 3940
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8103696591454633e-05,
      "loss": 0.2959,
      "step": 3950
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8098895823331736e-05,
      "loss": 0.285,
      "step": 3960
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.8094095055208835e-05,
      "loss": 0.4986,
      "step": 3970
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8089294287085934e-05,
      "loss": 0.4226,
      "step": 3980
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8084493518963036e-05,
      "loss": 0.357,
      "step": 3990
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8079692750840136e-05,
      "loss": 0.2581,
      "step": 4000
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8074891982717235e-05,
      "loss": 0.3922,
      "step": 4010
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.8070091214594337e-05,
      "loss": 0.2216,
      "step": 4020
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8065290446471436e-05,
      "loss": 0.6425,
      "step": 4030
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8060489678348536e-05,
      "loss": 0.396,
      "step": 4040
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8055688910225638e-05,
      "loss": 0.3119,
      "step": 4050
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.8050888142102737e-05,
      "loss": 0.377,
      "step": 4060
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.804608737397984e-05,
      "loss": 0.3404,
      "step": 4070
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.804128660585694e-05,
      "loss": 0.2097,
      "step": 4080
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.8036485837734038e-05,
      "loss": 0.4086,
      "step": 4090
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.803168506961114e-05,
      "loss": 0.4156,
      "step": 4100
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.802688430148824e-05,
      "loss": 0.4628,
      "step": 4110
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.802208353336534e-05,
      "loss": 0.2637,
      "step": 4120
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.801728276524244e-05,
      "loss": 0.3179,
      "step": 4130
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.801248199711954e-05,
      "loss": 0.2576,
      "step": 4140
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.800768122899664e-05,
      "loss": 0.2956,
      "step": 4150
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.8002880460873742e-05,
      "loss": 0.2593,
      "step": 4160
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.9697381842910575,
      "eval_loss": 0.15848955512046814,
      "eval_runtime": 518.0494,
      "eval_samples_per_second": 11.354,
      "eval_steps_per_second": 1.421,
      "step": 4166
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.799807969275084e-05,
      "loss": 0.2808,
      "step": 4170
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.799327892462794e-05,
      "loss": 0.3358,
      "step": 4180
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.7988478156505043e-05,
      "loss": 0.3963,
      "step": 4190
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.7983677388382142e-05,
      "loss": 0.3169,
      "step": 4200
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.797887662025924e-05,
      "loss": 0.3115,
      "step": 4210
    },
    {
      "epoch": 1.01,
      "learning_rate": 1.7974075852136344e-05,
      "loss": 0.4787,
      "step": 4220
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7969275084013443e-05,
      "loss": 0.3459,
      "step": 4230
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7964474315890542e-05,
      "loss": 0.4496,
      "step": 4240
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7959673547767645e-05,
      "loss": 0.3848,
      "step": 4250
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7954872779644744e-05,
      "loss": 0.301,
      "step": 4260
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.7950072011521843e-05,
      "loss": 0.2783,
      "step": 4270
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7945271243398946e-05,
      "loss": 0.4538,
      "step": 4280
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7940470475276045e-05,
      "loss": 0.4946,
      "step": 4290
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7935669707153144e-05,
      "loss": 0.4657,
      "step": 4300
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.7930868939030246e-05,
      "loss": 0.4639,
      "step": 4310
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7926068170907346e-05,
      "loss": 0.4415,
      "step": 4320
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7921267402784448e-05,
      "loss": 0.4027,
      "step": 4330
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7916466634661547e-05,
      "loss": 0.3872,
      "step": 4340
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.7911665866538646e-05,
      "loss": 0.3735,
      "step": 4350
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.790686509841575e-05,
      "loss": 0.3642,
      "step": 4360
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.7902064330292848e-05,
      "loss": 0.4073,
      "step": 4370
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.7897263562169947e-05,
      "loss": 0.2874,
      "step": 4380
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.789246279404705e-05,
      "loss": 0.3783,
      "step": 4390
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.788766202592415e-05,
      "loss": 0.2895,
      "step": 4400
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.7882861257801248e-05,
      "loss": 0.1587,
      "step": 4410
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.787806048967835e-05,
      "loss": 0.2973,
      "step": 4420
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.787325972155545e-05,
      "loss": 0.2541,
      "step": 4430
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.786845895343255e-05,
      "loss": 0.2957,
      "step": 4440
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.786365818530965e-05,
      "loss": 0.4392,
      "step": 4450
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.785885741718675e-05,
      "loss": 0.3449,
      "step": 4460
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.785405664906385e-05,
      "loss": 0.3259,
      "step": 4470
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7849255880940952e-05,
      "loss": 0.4596,
      "step": 4480
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.784445511281805e-05,
      "loss": 0.2926,
      "step": 4490
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.783965434469515e-05,
      "loss": 0.2853,
      "step": 4500
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7834853576572253e-05,
      "loss": 0.3184,
      "step": 4510
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.7830052808449356e-05,
      "loss": 0.3771,
      "step": 4520
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.782525204032645e-05,
      "loss": 0.1982,
      "step": 4530
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.7820451272203554e-05,
      "loss": 0.4207,
      "step": 4540
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.7815650504080657e-05,
      "loss": 0.3651,
      "step": 4550
    },
    {
      "epoch": 1.09,
      "learning_rate": 1.7810849735957752e-05,
      "loss": 0.3715,
      "step": 4560
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7806048967834855e-05,
      "loss": 0.358,
      "step": 4570
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7801248199711957e-05,
      "loss": 0.3517,
      "step": 4580
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7796447431589057e-05,
      "loss": 0.438,
      "step": 4590
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.7791646663466156e-05,
      "loss": 0.2194,
      "step": 4600
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.7786845895343258e-05,
      "loss": 0.4025,
      "step": 4610
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.7782045127220357e-05,
      "loss": 0.3115,
      "step": 4620
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.7777244359097457e-05,
      "loss": 0.3962,
      "step": 4630
    },
    {
      "epoch": 1.11,
      "learning_rate": 1.777244359097456e-05,
      "loss": 0.2103,
      "step": 4640
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.7767642822851658e-05,
      "loss": 0.2085,
      "step": 4650
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.7762842054728757e-05,
      "loss": 0.3312,
      "step": 4660
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.775804128660586e-05,
      "loss": 0.4754,
      "step": 4670
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.775324051848296e-05,
      "loss": 0.4154,
      "step": 4680
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.7748439750360058e-05,
      "loss": 0.2771,
      "step": 4690
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.774363898223716e-05,
      "loss": 0.4255,
      "step": 4700
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.773883821411426e-05,
      "loss": 0.3558,
      "step": 4710
    },
    {
      "epoch": 1.13,
      "learning_rate": 1.7734037445991362e-05,
      "loss": 0.4037,
      "step": 4720
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.772923667786846e-05,
      "loss": 0.3761,
      "step": 4730
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.772443590974556e-05,
      "loss": 0.2908,
      "step": 4740
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.7719635141622663e-05,
      "loss": 0.4659,
      "step": 4750
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.7714834373499762e-05,
      "loss": 0.3018,
      "step": 4760
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.771003360537686e-05,
      "loss": 0.2194,
      "step": 4770
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7705232837253964e-05,
      "loss": 0.4303,
      "step": 4780
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7700432069131063e-05,
      "loss": 0.2039,
      "step": 4790
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7695631301008162e-05,
      "loss": 0.324,
      "step": 4800
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.7690830532885265e-05,
      "loss": 0.5169,
      "step": 4810
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7686029764762364e-05,
      "loss": 0.2463,
      "step": 4820
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7681228996639463e-05,
      "loss": 0.3105,
      "step": 4830
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7676428228516566e-05,
      "loss": 0.1741,
      "step": 4840
    },
    {
      "epoch": 1.16,
      "learning_rate": 1.7671627460393665e-05,
      "loss": 0.1917,
      "step": 4850
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7666826692270764e-05,
      "loss": 0.3279,
      "step": 4860
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7662025924147867e-05,
      "loss": 0.6172,
      "step": 4870
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7657225156024966e-05,
      "loss": 0.4579,
      "step": 4880
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.7652424387902065e-05,
      "loss": 0.4042,
      "step": 4890
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7647623619779167e-05,
      "loss": 0.189,
      "step": 4900
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7642822851656267e-05,
      "loss": 0.4336,
      "step": 4910
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7638022083533366e-05,
      "loss": 0.1077,
      "step": 4920
    },
    {
      "epoch": 1.18,
      "learning_rate": 1.7633221315410468e-05,
      "loss": 0.4885,
      "step": 4930
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7628420547287567e-05,
      "loss": 0.3344,
      "step": 4940
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7623619779164667e-05,
      "loss": 0.236,
      "step": 4950
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.761881901104177e-05,
      "loss": 0.2562,
      "step": 4960
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.7614018242918868e-05,
      "loss": 0.4231,
      "step": 4970
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.760921747479597e-05,
      "loss": 0.1379,
      "step": 4980
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.760441670667307e-05,
      "loss": 0.135,
      "step": 4990
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.759961593855017e-05,
      "loss": 0.1465,
      "step": 5000
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.759481517042727e-05,
      "loss": 0.208,
      "step": 5010
    },
    {
      "epoch": 1.2,
      "learning_rate": 1.759001440230437e-05,
      "loss": 0.3199,
      "step": 5020
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.758521363418147e-05,
      "loss": 0.3387,
      "step": 5030
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.7580412866058572e-05,
      "loss": 0.2913,
      "step": 5040
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.757561209793567e-05,
      "loss": 0.208,
      "step": 5050
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.757081132981277e-05,
      "loss": 0.2419,
      "step": 5060
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7566010561689873e-05,
      "loss": 0.2377,
      "step": 5070
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7561209793566972e-05,
      "loss": 0.386,
      "step": 5080
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.755640902544407e-05,
      "loss": 0.3919,
      "step": 5090
    },
    {
      "epoch": 1.22,
      "learning_rate": 1.7551608257321174e-05,
      "loss": 0.2799,
      "step": 5100
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7546807489198273e-05,
      "loss": 0.2507,
      "step": 5110
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7542006721075372e-05,
      "loss": 0.1796,
      "step": 5120
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7537205952952475e-05,
      "loss": 0.2141,
      "step": 5130
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.7532405184829574e-05,
      "loss": 0.275,
      "step": 5140
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7527604416706673e-05,
      "loss": 0.2695,
      "step": 5150
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7522803648583776e-05,
      "loss": 0.232,
      "step": 5160
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7518002880460875e-05,
      "loss": 0.2083,
      "step": 5170
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.7513202112337974e-05,
      "loss": 0.4113,
      "step": 5180
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7508401344215077e-05,
      "loss": 0.4923,
      "step": 5190
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7503600576092176e-05,
      "loss": 0.3211,
      "step": 5200
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7498799807969275e-05,
      "loss": 0.2507,
      "step": 5210
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.7493999039846377e-05,
      "loss": 0.3018,
      "step": 5220
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7489198271723477e-05,
      "loss": 0.1886,
      "step": 5230
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7484397503600576e-05,
      "loss": 0.3006,
      "step": 5240
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.747959673547768e-05,
      "loss": 0.3268,
      "step": 5250
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.7474795967354777e-05,
      "loss": 0.1477,
      "step": 5260
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.746999519923188e-05,
      "loss": 0.213,
      "step": 5270
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.746519443110898e-05,
      "loss": 0.3322,
      "step": 5280
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.7460393662986078e-05,
      "loss": 0.2913,
      "step": 5290
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.745559289486318e-05,
      "loss": 0.318,
      "step": 5300
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.745079212674028e-05,
      "loss": 0.2929,
      "step": 5310
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.744599135861738e-05,
      "loss": 0.2828,
      "step": 5320
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.744119059049448e-05,
      "loss": 0.3368,
      "step": 5330
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.743638982237158e-05,
      "loss": 0.2549,
      "step": 5340
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.743158905424868e-05,
      "loss": 0.1731,
      "step": 5350
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7426788286125783e-05,
      "loss": 0.4072,
      "step": 5360
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.742198751800288e-05,
      "loss": 0.2125,
      "step": 5370
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.741718674987998e-05,
      "loss": 0.2279,
      "step": 5380
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.7412385981757083e-05,
      "loss": 0.3121,
      "step": 5390
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7407585213634182e-05,
      "loss": 0.3211,
      "step": 5400
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.740278444551128e-05,
      "loss": 0.2252,
      "step": 5410
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7397983677388384e-05,
      "loss": 0.3518,
      "step": 5420
    },
    {
      "epoch": 1.3,
      "learning_rate": 1.7393182909265483e-05,
      "loss": 0.3043,
      "step": 5430
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7388382141142582e-05,
      "loss": 0.2471,
      "step": 5440
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7383581373019685e-05,
      "loss": 0.258,
      "step": 5450
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7378780604896784e-05,
      "loss": 0.2289,
      "step": 5460
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.7373979836773883e-05,
      "loss": 0.172,
      "step": 5470
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7369179068650986e-05,
      "loss": 0.1301,
      "step": 5480
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7364378300528085e-05,
      "loss": 0.3406,
      "step": 5490
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7359577532405184e-05,
      "loss": 0.4144,
      "step": 5500
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.7354776764282287e-05,
      "loss": 0.2505,
      "step": 5510
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7349975996159386e-05,
      "loss": 0.298,
      "step": 5520
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.734517522803649e-05,
      "loss": 0.191,
      "step": 5530
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7340374459913588e-05,
      "loss": 0.4512,
      "step": 5540
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.7335573691790687e-05,
      "loss": 0.2741,
      "step": 5550
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.733077292366779e-05,
      "loss": 0.1714,
      "step": 5560
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.732597215554489e-05,
      "loss": 0.2258,
      "step": 5570
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.7321171387421988e-05,
      "loss": 0.3457,
      "step": 5580
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.731637061929909e-05,
      "loss": 0.2781,
      "step": 5590
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.731156985117619e-05,
      "loss": 0.1566,
      "step": 5600
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.730676908305329e-05,
      "loss": 0.2438,
      "step": 5610
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.730196831493039e-05,
      "loss": 0.304,
      "step": 5620
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.729716754680749e-05,
      "loss": 0.2895,
      "step": 5630
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.729236677868459e-05,
      "loss": 0.1523,
      "step": 5640
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7287566010561692e-05,
      "loss": 0.5485,
      "step": 5650
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7282765242438794e-05,
      "loss": 0.2285,
      "step": 5660
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.727796447431589e-05,
      "loss": 0.3143,
      "step": 5670
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.7273163706192993e-05,
      "loss": 0.4023,
      "step": 5680
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7268362938070095e-05,
      "loss": 0.3152,
      "step": 5690
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.726356216994719e-05,
      "loss": 0.3419,
      "step": 5700
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7258761401824293e-05,
      "loss": 0.2251,
      "step": 5710
    },
    {
      "epoch": 1.37,
      "learning_rate": 1.7253960633701396e-05,
      "loss": 0.4259,
      "step": 5720
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.724915986557849e-05,
      "loss": 0.3391,
      "step": 5730
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7244359097455594e-05,
      "loss": 0.2633,
      "step": 5740
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7239558329332697e-05,
      "loss": 0.285,
      "step": 5750
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.7234757561209793e-05,
      "loss": 0.3351,
      "step": 5760
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7229956793086895e-05,
      "loss": 0.3749,
      "step": 5770
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7225156024963998e-05,
      "loss": 0.1099,
      "step": 5780
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7220355256841097e-05,
      "loss": 0.2469,
      "step": 5790
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.7215554488718196e-05,
      "loss": 0.1826,
      "step": 5800
    },
    {
      "epoch": 1.39,
      "learning_rate": 1.72107537205953e-05,
      "loss": 0.2992,
      "step": 5810
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7205952952472398e-05,
      "loss": 0.4089,
      "step": 5820
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.7201152184349497e-05,
      "loss": 0.2354,
      "step": 5830
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.71963514162266e-05,
      "loss": 0.4552,
      "step": 5840
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.71915506481037e-05,
      "loss": 0.1552,
      "step": 5850
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7186749879980798e-05,
      "loss": 0.2951,
      "step": 5860
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.71819491118579e-05,
      "loss": 0.167,
      "step": 5870
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.7177148343735e-05,
      "loss": 0.3009,
      "step": 5880
    },
    {
      "epoch": 1.41,
      "learning_rate": 1.71723475756121e-05,
      "loss": 0.3901,
      "step": 5890
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.71675468074892e-05,
      "loss": 0.1968,
      "step": 5900
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.71627460393663e-05,
      "loss": 0.3833,
      "step": 5910
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7157945271243403e-05,
      "loss": 0.1489,
      "step": 5920
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.7153144503120502e-05,
      "loss": 0.5201,
      "step": 5930
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.71483437349976e-05,
      "loss": 0.2792,
      "step": 5940
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7143542966874703e-05,
      "loss": 0.1983,
      "step": 5950
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7138742198751803e-05,
      "loss": 0.3335,
      "step": 5960
    },
    {
      "epoch": 1.43,
      "learning_rate": 1.7133941430628902e-05,
      "loss": 0.2847,
      "step": 5970
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7129140662506004e-05,
      "loss": 0.3299,
      "step": 5980
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7124339894383103e-05,
      "loss": 0.3726,
      "step": 5990
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7119539126260203e-05,
      "loss": 0.3465,
      "step": 6000
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.7114738358137305e-05,
      "loss": 0.3033,
      "step": 6010
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7109937590014404e-05,
      "loss": 0.1776,
      "step": 6020
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7105136821891503e-05,
      "loss": 0.2759,
      "step": 6030
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7100336053768606e-05,
      "loss": 0.2215,
      "step": 6040
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7095535285645705e-05,
      "loss": 0.2725,
      "step": 6050
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.7090734517522804e-05,
      "loss": 0.2655,
      "step": 6060
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7085933749399907e-05,
      "loss": 0.2265,
      "step": 6070
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7081132981277006e-05,
      "loss": 0.2479,
      "step": 6080
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7076332213154105e-05,
      "loss": 0.3463,
      "step": 6090
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.7071531445031208e-05,
      "loss": 0.2398,
      "step": 6100
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7066730676908307e-05,
      "loss": 0.244,
      "step": 6110
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7061929908785406e-05,
      "loss": 0.2266,
      "step": 6120
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.705712914066251e-05,
      "loss": 0.3041,
      "step": 6130
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.7052328372539608e-05,
      "loss": 0.3264,
      "step": 6140
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.7047527604416707e-05,
      "loss": 0.1808,
      "step": 6150
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.704272683629381e-05,
      "loss": 0.345,
      "step": 6160
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.703792606817091e-05,
      "loss": 0.3436,
      "step": 6170
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.703312530004801e-05,
      "loss": 0.3214,
      "step": 6180
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.702832453192511e-05,
      "loss": 0.1239,
      "step": 6190
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.702352376380221e-05,
      "loss": 0.2669,
      "step": 6200
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.7018722995679312e-05,
      "loss": 0.2145,
      "step": 6210
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.701392222755641e-05,
      "loss": 0.3709,
      "step": 6220
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.700912145943351e-05,
      "loss": 0.4625,
      "step": 6230
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.7004320691310613e-05,
      "loss": 0.2129,
      "step": 6240
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.6999519923187712e-05,
      "loss": 0.3548,
      "step": 6250
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.699471915506481e-05,
      "loss": 0.3471,
      "step": 6260
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6989918386941914e-05,
      "loss": 0.3064,
      "step": 6270
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6985117618819013e-05,
      "loss": 0.3154,
      "step": 6280
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6980316850696112e-05,
      "loss": 0.2231,
      "step": 6290
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6975516082573214e-05,
      "loss": 0.2599,
      "step": 6300
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.6970715314450313e-05,
      "loss": 0.1981,
      "step": 6310
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6965914546327413e-05,
      "loss": 0.2433,
      "step": 6320
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6961113778204515e-05,
      "loss": 0.32,
      "step": 6330
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6956313010081614e-05,
      "loss": 0.2642,
      "step": 6340
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.6951512241958713e-05,
      "loss": 0.362,
      "step": 6350
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6946711473835816e-05,
      "loss": 0.2901,
      "step": 6360
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6941910705712915e-05,
      "loss": 0.2511,
      "step": 6370
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6937109937590014e-05,
      "loss": 0.3144,
      "step": 6380
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.6932309169467117e-05,
      "loss": 0.2797,
      "step": 6390
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6927508401344216e-05,
      "loss": 0.5936,
      "step": 6400
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6922707633221315e-05,
      "loss": 0.2537,
      "step": 6410
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6917906865098418e-05,
      "loss": 0.2856,
      "step": 6420
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.6913106096975517e-05,
      "loss": 0.3406,
      "step": 6430
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.690830532885262e-05,
      "loss": 0.219,
      "step": 6440
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.690350456072972e-05,
      "loss": 0.3206,
      "step": 6450
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.6898703792606818e-05,
      "loss": 0.3433,
      "step": 6460
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.689390302448392e-05,
      "loss": 0.356,
      "step": 6470
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.688910225636102e-05,
      "loss": 0.2436,
      "step": 6480
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.688430148823812e-05,
      "loss": 0.2651,
      "step": 6490
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.687950072011522e-05,
      "loss": 0.2662,
      "step": 6500
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.687469995199232e-05,
      "loss": 0.3791,
      "step": 6510
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.686989918386942e-05,
      "loss": 0.2771,
      "step": 6520
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.6865098415746522e-05,
      "loss": 0.2927,
      "step": 6530
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.686029764762362e-05,
      "loss": 0.2066,
      "step": 6540
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.685549687950072e-05,
      "loss": 0.1953,
      "step": 6550
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.6850696111377823e-05,
      "loss": 0.1433,
      "step": 6560
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.6845895343254922e-05,
      "loss": 0.3811,
      "step": 6570
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.684109457513202e-05,
      "loss": 0.2226,
      "step": 6580
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.6836293807009124e-05,
      "loss": 0.2424,
      "step": 6590
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.6831493038886223e-05,
      "loss": 0.3268,
      "step": 6600
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.6826692270763322e-05,
      "loss": 0.2614,
      "step": 6610
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.6821891502640424e-05,
      "loss": 0.3407,
      "step": 6620
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.6817090734517524e-05,
      "loss": 0.1865,
      "step": 6630
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.6812289966394623e-05,
      "loss": 0.3205,
      "step": 6640
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.6807489198271725e-05,
      "loss": 0.2809,
      "step": 6650
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.6802688430148824e-05,
      "loss": 0.2641,
      "step": 6660
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.6797887662025924e-05,
      "loss": 0.2858,
      "step": 6670
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.6793086893903026e-05,
      "loss": 0.119,
      "step": 6680
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6788286125780125e-05,
      "loss": 0.2279,
      "step": 6690
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6783485357657224e-05,
      "loss": 0.3126,
      "step": 6700
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6778684589534327e-05,
      "loss": 0.2901,
      "step": 6710
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.6773883821411426e-05,
      "loss": 0.288,
      "step": 6720
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.676908305328853e-05,
      "loss": 0.208,
      "step": 6730
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6764282285165628e-05,
      "loss": 0.1229,
      "step": 6740
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.6759481517042727e-05,
      "loss": 0.5133,
      "step": 6750
    },
    {
      "epoch": 1.62,
      "learning_rate": 1.675468074891983e-05,
      "loss": 0.3316,
      "step": 6760
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.674987998079693e-05,
      "loss": 0.295,
      "step": 6770
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.6745079212674028e-05,
      "loss": 0.2996,
      "step": 6780
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.674027844455113e-05,
      "loss": 0.1774,
      "step": 6790
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.673547767642823e-05,
      "loss": 0.3722,
      "step": 6800
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.673067690830533e-05,
      "loss": 0.0939,
      "step": 6810
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.672587614018243e-05,
      "loss": 0.4576,
      "step": 6820
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.672107537205953e-05,
      "loss": 0.2684,
      "step": 6830
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.671627460393663e-05,
      "loss": 0.2958,
      "step": 6840
    },
    {
      "epoch": 1.64,
      "learning_rate": 1.6711473835813732e-05,
      "loss": 0.2087,
      "step": 6850
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.6706673067690834e-05,
      "loss": 0.3335,
      "step": 6860
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.670187229956793e-05,
      "loss": 0.1949,
      "step": 6870
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.6697071531445033e-05,
      "loss": 0.2866,
      "step": 6880
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.6692270763322135e-05,
      "loss": 0.2533,
      "step": 6890
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.668746999519923e-05,
      "loss": 0.3213,
      "step": 6900
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6682669227076334e-05,
      "loss": 0.2166,
      "step": 6910
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6677868458953436e-05,
      "loss": 0.357,
      "step": 6920
    },
    {
      "epoch": 1.66,
      "learning_rate": 1.6673067690830532e-05,
      "loss": 0.1333,
      "step": 6930
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6668266922707634e-05,
      "loss": 0.3074,
      "step": 6940
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6663466154584737e-05,
      "loss": 0.1646,
      "step": 6950
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6658665386461833e-05,
      "loss": 0.121,
      "step": 6960
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.6653864618338935e-05,
      "loss": 0.3188,
      "step": 6970
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.6649063850216038e-05,
      "loss": 0.2299,
      "step": 6980
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.6644263082093137e-05,
      "loss": 0.3709,
      "step": 6990
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.6639462313970236e-05,
      "loss": 0.3846,
      "step": 7000
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.663466154584734e-05,
      "loss": 0.226,
      "step": 7010
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.6629860777724438e-05,
      "loss": 0.1858,
      "step": 7020
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.6625060009601537e-05,
      "loss": 0.3426,
      "step": 7030
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.662025924147864e-05,
      "loss": 0.2819,
      "step": 7040
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.661545847335574e-05,
      "loss": 0.2335,
      "step": 7050
    },
    {
      "epoch": 1.69,
      "learning_rate": 1.6610657705232838e-05,
      "loss": 0.3592,
      "step": 7060
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.660585693710994e-05,
      "loss": 0.2479,
      "step": 7070
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.660105616898704e-05,
      "loss": 0.2644,
      "step": 7080
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.659625540086414e-05,
      "loss": 0.1601,
      "step": 7090
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.659145463274124e-05,
      "loss": 0.2437,
      "step": 7100
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.658665386461834e-05,
      "loss": 0.2847,
      "step": 7110
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.6581853096495443e-05,
      "loss": 0.2417,
      "step": 7120
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.6577052328372542e-05,
      "loss": 0.2989,
      "step": 7130
    },
    {
      "epoch": 1.71,
      "learning_rate": 1.657225156024964e-05,
      "loss": 0.1034,
      "step": 7140
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6567450792126744e-05,
      "loss": 0.1075,
      "step": 7150
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6562650024003843e-05,
      "loss": 0.1714,
      "step": 7160
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6557849255880942e-05,
      "loss": 0.3407,
      "step": 7170
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.6553048487758044e-05,
      "loss": 0.1231,
      "step": 7180
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.6548247719635144e-05,
      "loss": 0.3359,
      "step": 7190
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.6543446951512243e-05,
      "loss": 0.1883,
      "step": 7200
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.6538646183389345e-05,
      "loss": 0.4074,
      "step": 7210
    },
    {
      "epoch": 1.73,
      "learning_rate": 1.6533845415266444e-05,
      "loss": 0.1621,
      "step": 7220
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6529044647143544e-05,
      "loss": 0.2488,
      "step": 7230
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6524243879020646e-05,
      "loss": 0.29,
      "step": 7240
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6519443110897745e-05,
      "loss": 0.4122,
      "step": 7250
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.6514642342774844e-05,
      "loss": 0.3558,
      "step": 7260
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6509841574651947e-05,
      "loss": 0.2255,
      "step": 7270
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6505040806529046e-05,
      "loss": 0.1434,
      "step": 7280
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6500240038406145e-05,
      "loss": 0.2075,
      "step": 7290
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6495439270283248e-05,
      "loss": 0.1391,
      "step": 7300
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.6490638502160347e-05,
      "loss": 0.2247,
      "step": 7310
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.6485837734037446e-05,
      "loss": 0.1813,
      "step": 7320
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.648103696591455e-05,
      "loss": 0.284,
      "step": 7330
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.6476236197791648e-05,
      "loss": 0.1173,
      "step": 7340
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.6471435429668747e-05,
      "loss": 0.1648,
      "step": 7350
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.646663466154585e-05,
      "loss": 0.1874,
      "step": 7360
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.646183389342295e-05,
      "loss": 0.4147,
      "step": 7370
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.645703312530005e-05,
      "loss": 0.1736,
      "step": 7380
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.645223235717715e-05,
      "loss": 0.1742,
      "step": 7390
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.644743158905425e-05,
      "loss": 0.2497,
      "step": 7400
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.6442630820931352e-05,
      "loss": 0.3118,
      "step": 7410
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.643783005280845e-05,
      "loss": 0.2658,
      "step": 7420
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.643302928468555e-05,
      "loss": 0.2131,
      "step": 7430
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.6428228516562653e-05,
      "loss": 0.4354,
      "step": 7440
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.6423427748439752e-05,
      "loss": 0.1038,
      "step": 7450
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.641862698031685e-05,
      "loss": 0.3985,
      "step": 7460
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.6413826212193954e-05,
      "loss": 0.2182,
      "step": 7470
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6409025444071053e-05,
      "loss": 0.1926,
      "step": 7480
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6404224675948152e-05,
      "loss": 0.293,
      "step": 7490
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6399423907825255e-05,
      "loss": 0.222,
      "step": 7500
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.6394623139702354e-05,
      "loss": 0.333,
      "step": 7510
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6389822371579453e-05,
      "loss": 0.3317,
      "step": 7520
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6385021603456555e-05,
      "loss": 0.1311,
      "step": 7530
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6380220835333655e-05,
      "loss": 0.2559,
      "step": 7540
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6375420067210754e-05,
      "loss": 0.1644,
      "step": 7550
    },
    {
      "epoch": 1.81,
      "learning_rate": 1.6370619299087856e-05,
      "loss": 0.356,
      "step": 7560
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6365818530964955e-05,
      "loss": 0.2955,
      "step": 7570
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6361017762842054e-05,
      "loss": 0.2048,
      "step": 7580
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6356216994719157e-05,
      "loss": 0.1993,
      "step": 7590
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.6351416226596256e-05,
      "loss": 0.2183,
      "step": 7600
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.6346615458473355e-05,
      "loss": 0.27,
      "step": 7610
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.6341814690350458e-05,
      "loss": 0.2941,
      "step": 7620
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.6337013922227557e-05,
      "loss": 0.314,
      "step": 7630
    },
    {
      "epoch": 1.83,
      "learning_rate": 1.633221315410466e-05,
      "loss": 0.1382,
      "step": 7640
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.632741238598176e-05,
      "loss": 0.2129,
      "step": 7650
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.6322611617858858e-05,
      "loss": 0.3042,
      "step": 7660
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.631781084973596e-05,
      "loss": 0.1912,
      "step": 7670
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.631301008161306e-05,
      "loss": 0.1291,
      "step": 7680
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.630820931349016e-05,
      "loss": 0.1602,
      "step": 7690
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.630340854536726e-05,
      "loss": 0.1936,
      "step": 7700
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.629860777724436e-05,
      "loss": 0.2681,
      "step": 7710
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.629380700912146e-05,
      "loss": 0.2612,
      "step": 7720
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.6289006240998562e-05,
      "loss": 0.149,
      "step": 7730
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.628420547287566e-05,
      "loss": 0.3117,
      "step": 7740
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.627940470475276e-05,
      "loss": 0.2836,
      "step": 7750
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.6274603936629863e-05,
      "loss": 0.2682,
      "step": 7760
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6269803168506962e-05,
      "loss": 0.2707,
      "step": 7770
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.626500240038406e-05,
      "loss": 0.3505,
      "step": 7780
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6260201632261164e-05,
      "loss": 0.2875,
      "step": 7790
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6255400864138263e-05,
      "loss": 0.2281,
      "step": 7800
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.6250600096015362e-05,
      "loss": 0.1715,
      "step": 7810
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.6245799327892465e-05,
      "loss": 0.2805,
      "step": 7820
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.6240998559769564e-05,
      "loss": 0.2698,
      "step": 7830
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.6236197791646663e-05,
      "loss": 0.2783,
      "step": 7840
    },
    {
      "epoch": 1.88,
      "learning_rate": 1.6231397023523765e-05,
      "loss": 0.188,
      "step": 7850
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6226596255400865e-05,
      "loss": 0.3296,
      "step": 7860
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6221795487277964e-05,
      "loss": 0.1835,
      "step": 7870
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6216994719155066e-05,
      "loss": 0.1733,
      "step": 7880
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.6212193951032165e-05,
      "loss": 0.1842,
      "step": 7890
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6207393182909268e-05,
      "loss": 0.2846,
      "step": 7900
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6202592414786367e-05,
      "loss": 0.1461,
      "step": 7910
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.6197791646663466e-05,
      "loss": 0.266,
      "step": 7920
    },
    {
      "epoch": 1.9,
      "learning_rate": 1.619299087854057e-05,
      "loss": 0.2243,
      "step": 7930
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.6188190110417668e-05,
      "loss": 0.1376,
      "step": 7940
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.6183389342294767e-05,
      "loss": 0.115,
      "step": 7950
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.617858857417187e-05,
      "loss": 0.0298,
      "step": 7960
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.617378780604897e-05,
      "loss": 0.0588,
      "step": 7970
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.6168987037926068e-05,
      "loss": 0.1534,
      "step": 7980
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.616418626980317e-05,
      "loss": 0.3296,
      "step": 7990
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.615938550168027e-05,
      "loss": 0.0936,
      "step": 8000
    },
    {
      "epoch": 1.92,
      "learning_rate": 1.615458473355737e-05,
      "loss": 0.1979,
      "step": 8010
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.614978396543447e-05,
      "loss": 0.3102,
      "step": 8020
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.6144983197311574e-05,
      "loss": 0.1653,
      "step": 8030
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.614018242918867e-05,
      "loss": 0.285,
      "step": 8040
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.6135381661065772e-05,
      "loss": 0.1726,
      "step": 8050
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.6130580892942875e-05,
      "loss": 0.1975,
      "step": 8060
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.612578012481997e-05,
      "loss": 0.2318,
      "step": 8070
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.6120979356697073e-05,
      "loss": 0.3342,
      "step": 8080
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.6116178588574175e-05,
      "loss": 0.2647,
      "step": 8090
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.611137782045127e-05,
      "loss": 0.1491,
      "step": 8100
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.6106577052328374e-05,
      "loss": 0.1737,
      "step": 8110
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.6101776284205476e-05,
      "loss": 0.1692,
      "step": 8120
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.6096975516082572e-05,
      "loss": 0.2975,
      "step": 8130
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.6092174747959675e-05,
      "loss": 0.338,
      "step": 8140
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.6087373979836777e-05,
      "loss": 0.2175,
      "step": 8150
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.6082573211713876e-05,
      "loss": 0.2292,
      "step": 8160
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.6077772443590975e-05,
      "loss": 0.2528,
      "step": 8170
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.6072971675468078e-05,
      "loss": 0.4907,
      "step": 8180
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.6068170907345177e-05,
      "loss": 0.2025,
      "step": 8190
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.6063370139222276e-05,
      "loss": 0.2904,
      "step": 8200
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.605856937109938e-05,
      "loss": 0.2836,
      "step": 8210
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.6053768602976478e-05,
      "loss": 0.092,
      "step": 8220
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.6048967834853577e-05,
      "loss": 0.3169,
      "step": 8230
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.604416706673068e-05,
      "loss": 0.2997,
      "step": 8240
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.603936629860778e-05,
      "loss": 0.1031,
      "step": 8250
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.6034565530484878e-05,
      "loss": 0.1372,
      "step": 8260
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.602976476236198e-05,
      "loss": 0.2871,
      "step": 8270
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.602496399423908e-05,
      "loss": 0.3045,
      "step": 8280
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.602016322611618e-05,
      "loss": 0.2663,
      "step": 8290
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.601536245799328e-05,
      "loss": 0.0791,
      "step": 8300
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.601056168987038e-05,
      "loss": 0.2043,
      "step": 8310
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.6005760921747483e-05,
      "loss": 0.2235,
      "step": 8320
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.6000960153624582e-05,
      "loss": 0.2659,
      "step": 8330
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.9899693981638898,
      "eval_loss": 0.04719490185379982,
      "eval_runtime": 513.7691,
      "eval_samples_per_second": 11.449,
      "eval_steps_per_second": 1.433,
      "step": 8332
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.599615938550168e-05,
      "loss": 0.2035,
      "step": 8340
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.5991358617378784e-05,
      "loss": 0.3643,
      "step": 8350
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.5986557849255883e-05,
      "loss": 0.2984,
      "step": 8360
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.5981757081132982e-05,
      "loss": 0.4262,
      "step": 8370
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.5976956313010085e-05,
      "loss": 0.2335,
      "step": 8380
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.5972155544887184e-05,
      "loss": 0.2319,
      "step": 8390
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5967354776764283e-05,
      "loss": 0.1588,
      "step": 8400
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5962554008641386e-05,
      "loss": 0.2065,
      "step": 8410
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5957753240518485e-05,
      "loss": 0.1686,
      "step": 8420
    },
    {
      "epoch": 2.02,
      "learning_rate": 1.5952952472395584e-05,
      "loss": 0.3103,
      "step": 8430
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5948151704272686e-05,
      "loss": 0.2384,
      "step": 8440
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5943350936149785e-05,
      "loss": 0.166,
      "step": 8450
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5938550168026885e-05,
      "loss": 0.1075,
      "step": 8460
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.5933749399903987e-05,
      "loss": 0.083,
      "step": 8470
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.5928948631781086e-05,
      "loss": 0.2055,
      "step": 8480
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.5924147863658185e-05,
      "loss": 0.3414,
      "step": 8490
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.5919347095535288e-05,
      "loss": 0.1712,
      "step": 8500
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.5914546327412387e-05,
      "loss": 0.3873,
      "step": 8510
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.5909745559289486e-05,
      "loss": 0.1031,
      "step": 8520
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.590494479116659e-05,
      "loss": 0.378,
      "step": 8530
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.5900144023043688e-05,
      "loss": 0.2487,
      "step": 8540
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.5895343254920787e-05,
      "loss": 0.0604,
      "step": 8550
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.589054248679789e-05,
      "loss": 0.2216,
      "step": 8560
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.588574171867499e-05,
      "loss": 0.2217,
      "step": 8570
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.588094095055209e-05,
      "loss": 0.3557,
      "step": 8580
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.587614018242919e-05,
      "loss": 0.3425,
      "step": 8590
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.587133941430629e-05,
      "loss": 0.2533,
      "step": 8600
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.5866538646183392e-05,
      "loss": 0.3007,
      "step": 8610
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.586173787806049e-05,
      "loss": 0.32,
      "step": 8620
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.585693710993759e-05,
      "loss": 0.2033,
      "step": 8630
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.5852136341814693e-05,
      "loss": 0.4156,
      "step": 8640
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5847335573691792e-05,
      "loss": 0.1899,
      "step": 8650
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.584253480556889e-05,
      "loss": 0.342,
      "step": 8660
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5837734037445994e-05,
      "loss": 0.2475,
      "step": 8670
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.5832933269323093e-05,
      "loss": 0.3728,
      "step": 8680
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5828132501200192e-05,
      "loss": 0.2888,
      "step": 8690
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5823331733077295e-05,
      "loss": 0.2512,
      "step": 8700
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5818530964954394e-05,
      "loss": 0.1717,
      "step": 8710
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5813730196831493e-05,
      "loss": 0.1767,
      "step": 8720
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5808929428708596e-05,
      "loss": 0.224,
      "step": 8730
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5804128660585695e-05,
      "loss": 0.1161,
      "step": 8740
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5799327892462794e-05,
      "loss": 0.3446,
      "step": 8750
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.5794527124339896e-05,
      "loss": 0.2593,
      "step": 8760
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5789726356216996e-05,
      "loss": 0.3028,
      "step": 8770
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5784925588094095e-05,
      "loss": 0.2327,
      "step": 8780
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5780124819971197e-05,
      "loss": 0.216,
      "step": 8790
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5775324051848296e-05,
      "loss": 0.3009,
      "step": 8800
    },
    {
      "epoch": 2.11,
      "learning_rate": 1.5770523283725396e-05,
      "loss": 0.2319,
      "step": 8810
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5765722515602498e-05,
      "loss": 0.2915,
      "step": 8820
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.5760921747479597e-05,
      "loss": 0.168,
      "step": 8830
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.57561209793567e-05,
      "loss": 0.083,
      "step": 8840
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.57513202112338e-05,
      "loss": 0.222,
      "step": 8850
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5746519443110898e-05,
      "loss": 0.1219,
      "step": 8860
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.5741718674988e-05,
      "loss": 0.1796,
      "step": 8870
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.57369179068651e-05,
      "loss": 0.1524,
      "step": 8880
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.57321171387422e-05,
      "loss": 0.1782,
      "step": 8890
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.57273163706193e-05,
      "loss": 0.1206,
      "step": 8900
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.57225156024964e-05,
      "loss": 0.096,
      "step": 8910
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.57177148343735e-05,
      "loss": 0.3113,
      "step": 8920
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.5712914066250602e-05,
      "loss": 0.3569,
      "step": 8930
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.57081132981277e-05,
      "loss": 0.1468,
      "step": 8940
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.57033125300048e-05,
      "loss": 0.3133,
      "step": 8950
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5698511761881903e-05,
      "loss": 0.2364,
      "step": 8960
    },
    {
      "epoch": 2.15,
      "learning_rate": 1.5693710993759002e-05,
      "loss": 0.3481,
      "step": 8970
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.56889102256361e-05,
      "loss": 0.2658,
      "step": 8980
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5684109457513204e-05,
      "loss": 0.2087,
      "step": 8990
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5679308689390303e-05,
      "loss": 0.1101,
      "step": 9000
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.5674507921267402e-05,
      "loss": 0.1522,
      "step": 9010
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.5669707153144505e-05,
      "loss": 0.2566,
      "step": 9020
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.5664906385021604e-05,
      "loss": 0.1457,
      "step": 9030
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.5660105616898703e-05,
      "loss": 0.1996,
      "step": 9040
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.5655304848775806e-05,
      "loss": 0.198,
      "step": 9050
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.5650504080652905e-05,
      "loss": 0.1476,
      "step": 9060
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.5645703312530004e-05,
      "loss": 0.2568,
      "step": 9070
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.5640902544407106e-05,
      "loss": 0.186,
      "step": 9080
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.5636101776284206e-05,
      "loss": 0.141,
      "step": 9090
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.5631301008161308e-05,
      "loss": 0.2253,
      "step": 9100
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.5626500240038407e-05,
      "loss": 0.0886,
      "step": 9110
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.5621699471915506e-05,
      "loss": 0.1932,
      "step": 9120
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.561689870379261e-05,
      "loss": 0.3795,
      "step": 9130
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.5612097935669708e-05,
      "loss": 0.203,
      "step": 9140
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.5607297167546807e-05,
      "loss": 0.2419,
      "step": 9150
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.560249639942391e-05,
      "loss": 0.2311,
      "step": 9160
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.559769563130101e-05,
      "loss": 0.2893,
      "step": 9170
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.5592894863178108e-05,
      "loss": 0.1757,
      "step": 9180
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.558809409505521e-05,
      "loss": 0.1514,
      "step": 9190
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.558329332693231e-05,
      "loss": 0.2027,
      "step": 9200
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.557849255880941e-05,
      "loss": 0.2129,
      "step": 9210
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.557369179068651e-05,
      "loss": 0.1661,
      "step": 9220
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.5568891022563614e-05,
      "loss": 0.2077,
      "step": 9230
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.556409025444071e-05,
      "loss": 0.1438,
      "step": 9240
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.5559289486317812e-05,
      "loss": 0.2247,
      "step": 9250
    },
    {
      "epoch": 2.22,
      "learning_rate": 1.5554488718194915e-05,
      "loss": 0.0931,
      "step": 9260
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.554968795007201e-05,
      "loss": 0.2027,
      "step": 9270
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5544887181949113e-05,
      "loss": 0.2568,
      "step": 9280
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5540086413826216e-05,
      "loss": 0.1005,
      "step": 9290
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.553528564570331e-05,
      "loss": 0.149,
      "step": 9300
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.5530484877580414e-05,
      "loss": 0.2124,
      "step": 9310
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5525684109457517e-05,
      "loss": 0.3563,
      "step": 9320
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5520883341334612e-05,
      "loss": 0.3381,
      "step": 9330
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5516082573211715e-05,
      "loss": 0.1943,
      "step": 9340
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.5511281805088817e-05,
      "loss": 0.0356,
      "step": 9350
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5506481036965916e-05,
      "loss": 0.1956,
      "step": 9360
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5501680268843016e-05,
      "loss": 0.3598,
      "step": 9370
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5496879500720118e-05,
      "loss": 0.2397,
      "step": 9380
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.5492078732597217e-05,
      "loss": 0.3346,
      "step": 9390
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.5487277964474316e-05,
      "loss": 0.0372,
      "step": 9400
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.548247719635142e-05,
      "loss": 0.2807,
      "step": 9410
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.5477676428228518e-05,
      "loss": 0.1641,
      "step": 9420
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.5472875660105617e-05,
      "loss": 0.2699,
      "step": 9430
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.546807489198272e-05,
      "loss": 0.1725,
      "step": 9440
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.546327412385982e-05,
      "loss": 0.1684,
      "step": 9450
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.5458473355736918e-05,
      "loss": 0.2307,
      "step": 9460
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.545367258761402e-05,
      "loss": 0.0479,
      "step": 9470
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.544887181949112e-05,
      "loss": 0.1937,
      "step": 9480
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.5444071051368222e-05,
      "loss": 0.076,
      "step": 9490
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.543927028324532e-05,
      "loss": 0.2921,
      "step": 9500
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.543446951512242e-05,
      "loss": 0.1695,
      "step": 9510
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.5429668746999523e-05,
      "loss": 0.4266,
      "step": 9520
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.5424867978876622e-05,
      "loss": 0.3545,
      "step": 9530
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.542006721075372e-05,
      "loss": 0.0763,
      "step": 9540
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.5415266442630824e-05,
      "loss": 0.1396,
      "step": 9550
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.5410465674507923e-05,
      "loss": 0.345,
      "step": 9560
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.5405664906385022e-05,
      "loss": 0.1957,
      "step": 9570
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.5400864138262125e-05,
      "loss": 0.2812,
      "step": 9580
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.5396063370139224e-05,
      "loss": 0.2335,
      "step": 9590
    },
    {
      "epoch": 2.3,
      "learning_rate": 1.5391262602016323e-05,
      "loss": 0.3567,
      "step": 9600
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.5386461833893426e-05,
      "loss": 0.1948,
      "step": 9610
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.5381661065770525e-05,
      "loss": 0.146,
      "step": 9620
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.5376860297647624e-05,
      "loss": 0.293,
      "step": 9630
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.5372059529524727e-05,
      "loss": 0.1375,
      "step": 9640
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.5367258761401826e-05,
      "loss": 0.1077,
      "step": 9650
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.5362457993278925e-05,
      "loss": 0.1453,
      "step": 9660
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.5357657225156027e-05,
      "loss": 0.2481,
      "step": 9670
    },
    {
      "epoch": 2.32,
      "learning_rate": 1.5352856457033127e-05,
      "loss": 0.2042,
      "step": 9680
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.5348055688910226e-05,
      "loss": 0.2722,
      "step": 9690
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.5343254920787328e-05,
      "loss": 0.2392,
      "step": 9700
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.5338454152664427e-05,
      "loss": 0.1365,
      "step": 9710
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.5333653384541527e-05,
      "loss": 0.2481,
      "step": 9720
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.532885261641863e-05,
      "loss": 0.1214,
      "step": 9730
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.5324051848295728e-05,
      "loss": 0.1552,
      "step": 9740
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.5319251080172827e-05,
      "loss": 0.2478,
      "step": 9750
    },
    {
      "epoch": 2.34,
      "learning_rate": 1.531445031204993e-05,
      "loss": 0.2395,
      "step": 9760
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.530964954392703e-05,
      "loss": 0.2827,
      "step": 9770
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.530484877580413e-05,
      "loss": 0.0382,
      "step": 9780
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.530004800768123e-05,
      "loss": 0.1288,
      "step": 9790
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.529524723955833e-05,
      "loss": 0.0809,
      "step": 9800
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.5290446471435432e-05,
      "loss": 0.3355,
      "step": 9810
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.528564570331253e-05,
      "loss": 0.2088,
      "step": 9820
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.528084493518963e-05,
      "loss": 0.0543,
      "step": 9830
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.5276044167066733e-05,
      "loss": 0.1086,
      "step": 9840
    },
    {
      "epoch": 2.36,
      "learning_rate": 1.5271243398943832e-05,
      "loss": 0.1396,
      "step": 9850
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.526644263082093e-05,
      "loss": 0.1955,
      "step": 9860
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.5261641862698034e-05,
      "loss": 0.2258,
      "step": 9870
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.5256841094575132e-05,
      "loss": 0.1816,
      "step": 9880
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.5252040326452232e-05,
      "loss": 0.2099,
      "step": 9890
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.5247239558329335e-05,
      "loss": 0.0644,
      "step": 9900
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.5242438790206436e-05,
      "loss": 0.1803,
      "step": 9910
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.5237638022083533e-05,
      "loss": 0.3613,
      "step": 9920
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.5232837253960636e-05,
      "loss": 0.3219,
      "step": 9930
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.5228036485837737e-05,
      "loss": 0.1716,
      "step": 9940
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.5223235717714834e-05,
      "loss": 0.0709,
      "step": 9950
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.5218434949591937e-05,
      "loss": 0.1417,
      "step": 9960
    },
    {
      "epoch": 2.39,
      "learning_rate": 1.5213634181469037e-05,
      "loss": 0.1804,
      "step": 9970
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.5208833413346135e-05,
      "loss": 0.1785,
      "step": 9980
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.5204032645223237e-05,
      "loss": 0.2638,
      "step": 9990
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.5199231877100338e-05,
      "loss": 0.2374,
      "step": 10000
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.5194431108977437e-05,
      "loss": 0.1964,
      "step": 10010
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.5189630340854538e-05,
      "loss": 0.1756,
      "step": 10020
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.5184829572731639e-05,
      "loss": 0.3015,
      "step": 10030
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.518002880460874e-05,
      "loss": 0.2633,
      "step": 10040
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.5175228036485839e-05,
      "loss": 0.2612,
      "step": 10050
    },
    {
      "epoch": 2.41,
      "learning_rate": 1.517042726836294e-05,
      "loss": 0.3114,
      "step": 10060
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.516562650024004e-05,
      "loss": 0.1669,
      "step": 10070
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.516082573211714e-05,
      "loss": 0.2757,
      "step": 10080
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.515602496399424e-05,
      "loss": 0.4158,
      "step": 10090
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.5151224195871342e-05,
      "loss": 0.2609,
      "step": 10100
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.514642342774844e-05,
      "loss": 0.1338,
      "step": 10110
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.5141622659625542e-05,
      "loss": 0.2169,
      "step": 10120
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.5136821891502642e-05,
      "loss": 0.1438,
      "step": 10130
    },
    {
      "epoch": 2.43,
      "learning_rate": 1.5132021123379742e-05,
      "loss": 0.0989,
      "step": 10140
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.5127220355256842e-05,
      "loss": 0.2776,
      "step": 10150
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.5122419587133943e-05,
      "loss": 0.2292,
      "step": 10160
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.5117618819011044e-05,
      "loss": 0.1841,
      "step": 10170
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.5112818050888143e-05,
      "loss": 0.1699,
      "step": 10180
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.5108017282765244e-05,
      "loss": 0.1345,
      "step": 10190
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.5103216514642345e-05,
      "loss": 0.289,
      "step": 10200
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.5098415746519444e-05,
      "loss": 0.1946,
      "step": 10210
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.5093614978396545e-05,
      "loss": 0.1135,
      "step": 10220
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.5088814210273646e-05,
      "loss": 0.283,
      "step": 10230
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.5084013442150745e-05,
      "loss": 0.0817,
      "step": 10240
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.5079212674027846e-05,
      "loss": 0.2072,
      "step": 10250
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.5074411905904947e-05,
      "loss": 0.1907,
      "step": 10260
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5069611137782046e-05,
      "loss": 0.1626,
      "step": 10270
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5064810369659147e-05,
      "loss": 0.2369,
      "step": 10280
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5060009601536247e-05,
      "loss": 0.215,
      "step": 10290
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5055208833413348e-05,
      "loss": 0.1224,
      "step": 10300
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.5050408065290447e-05,
      "loss": 0.2837,
      "step": 10310
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5045607297167548e-05,
      "loss": 0.2383,
      "step": 10320
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5040806529044649e-05,
      "loss": 0.129,
      "step": 10330
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5036005760921748e-05,
      "loss": 0.1004,
      "step": 10340
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.5031204992798849e-05,
      "loss": 0.1682,
      "step": 10350
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.502640422467595e-05,
      "loss": 0.1598,
      "step": 10360
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.5021603456553049e-05,
      "loss": 0.1692,
      "step": 10370
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.501680268843015e-05,
      "loss": 0.2667,
      "step": 10380
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.501200192030725e-05,
      "loss": 0.152,
      "step": 10390
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.500720115218435e-05,
      "loss": 0.2442,
      "step": 10400
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.500240038406145e-05,
      "loss": 0.1468,
      "step": 10410
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4997599615938552e-05,
      "loss": 0.3102,
      "step": 10420
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.4992798847815652e-05,
      "loss": 0.0355,
      "step": 10430
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.4987998079692752e-05,
      "loss": 0.1872,
      "step": 10440
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.4983197311569852e-05,
      "loss": 0.241,
      "step": 10450
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.4978396543446953e-05,
      "loss": 0.1428,
      "step": 10460
    },
    {
      "epoch": 2.51,
      "learning_rate": 1.4973595775324052e-05,
      "loss": 0.1533,
      "step": 10470
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.4968795007201153e-05,
      "loss": 0.1158,
      "step": 10480
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.4963994239078254e-05,
      "loss": 0.318,
      "step": 10490
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.4959193470955353e-05,
      "loss": 0.1924,
      "step": 10500
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.4954392702832454e-05,
      "loss": 0.1385,
      "step": 10510
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.4949591934709555e-05,
      "loss": 0.2009,
      "step": 10520
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.4944791166586654e-05,
      "loss": 0.2928,
      "step": 10530
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.4939990398463755e-05,
      "loss": 0.1048,
      "step": 10540
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.4935189630340856e-05,
      "loss": 0.2862,
      "step": 10550
    },
    {
      "epoch": 2.53,
      "learning_rate": 1.4930388862217957e-05,
      "loss": 0.3168,
      "step": 10560
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.4925588094095056e-05,
      "loss": 0.2204,
      "step": 10570
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.4920787325972157e-05,
      "loss": 0.358,
      "step": 10580
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.4915986557849258e-05,
      "loss": 0.1276,
      "step": 10590
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.4911185789726357e-05,
      "loss": 0.2204,
      "step": 10600
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.4906385021603458e-05,
      "loss": 0.1954,
      "step": 10610
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.4901584253480558e-05,
      "loss": 0.1739,
      "step": 10620
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.4896783485357657e-05,
      "loss": 0.2356,
      "step": 10630
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.4891982717234758e-05,
      "loss": 0.1433,
      "step": 10640
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.488718194911186e-05,
      "loss": 0.3961,
      "step": 10650
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.4882381180988958e-05,
      "loss": 0.1719,
      "step": 10660
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.487758041286606e-05,
      "loss": 0.2837,
      "step": 10670
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.487277964474316e-05,
      "loss": 0.1502,
      "step": 10680
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.4867978876620261e-05,
      "loss": 0.2964,
      "step": 10690
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.486317810849736e-05,
      "loss": 0.085,
      "step": 10700
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.4858377340374461e-05,
      "loss": 0.1052,
      "step": 10710
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.4853576572251562e-05,
      "loss": 0.1558,
      "step": 10720
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.484877580412866e-05,
      "loss": 0.3192,
      "step": 10730
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.4843975036005762e-05,
      "loss": 0.1599,
      "step": 10740
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.4839174267882863e-05,
      "loss": 0.2426,
      "step": 10750
    },
    {
      "epoch": 2.58,
      "learning_rate": 1.4834373499759962e-05,
      "loss": 0.1456,
      "step": 10760
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.4829572731637063e-05,
      "loss": 0.0722,
      "step": 10770
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.4824771963514163e-05,
      "loss": 0.2097,
      "step": 10780
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.4819971195391263e-05,
      "loss": 0.2274,
      "step": 10790
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.4815170427268363e-05,
      "loss": 0.202,
      "step": 10800
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.4810369659145464e-05,
      "loss": 0.1246,
      "step": 10810
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.4805568891022565e-05,
      "loss": 0.1036,
      "step": 10820
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.4800768122899664e-05,
      "loss": 0.4299,
      "step": 10830
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.4795967354776765e-05,
      "loss": 0.0968,
      "step": 10840
    },
    {
      "epoch": 2.6,
      "learning_rate": 1.4791166586653868e-05,
      "loss": 0.2303,
      "step": 10850
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.4786365818530965e-05,
      "loss": 0.0818,
      "step": 10860
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.4781565050408066e-05,
      "loss": 0.3172,
      "step": 10870
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.4776764282285168e-05,
      "loss": 0.1984,
      "step": 10880
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.4771963514162266e-05,
      "loss": 0.2799,
      "step": 10890
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.4767162746039367e-05,
      "loss": 0.2263,
      "step": 10900
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.476236197791647e-05,
      "loss": 0.143,
      "step": 10910
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.4757561209793567e-05,
      "loss": 0.1558,
      "step": 10920
    },
    {
      "epoch": 2.62,
      "learning_rate": 1.4752760441670668e-05,
      "loss": 0.1062,
      "step": 10930
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.474795967354777e-05,
      "loss": 0.0642,
      "step": 10940
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.4743158905424871e-05,
      "loss": 0.1034,
      "step": 10950
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.4738358137301968e-05,
      "loss": 0.3258,
      "step": 10960
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.4733557369179071e-05,
      "loss": 0.1321,
      "step": 10970
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.4728756601056172e-05,
      "loss": 0.0591,
      "step": 10980
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.472395583293327e-05,
      "loss": 0.1292,
      "step": 10990
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.4719155064810372e-05,
      "loss": 0.1714,
      "step": 11000
    },
    {
      "epoch": 2.64,
      "learning_rate": 1.4714354296687473e-05,
      "loss": 0.2207,
      "step": 11010
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.470955352856457e-05,
      "loss": 0.137,
      "step": 11020
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.4704752760441673e-05,
      "loss": 0.1735,
      "step": 11030
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.4699951992318773e-05,
      "loss": 0.2947,
      "step": 11040
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.4695151224195871e-05,
      "loss": 0.2376,
      "step": 11050
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.4690350456072973e-05,
      "loss": 0.2711,
      "step": 11060
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.4685549687950074e-05,
      "loss": 0.3468,
      "step": 11070
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.4680748919827175e-05,
      "loss": 0.1008,
      "step": 11080
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.4675948151704274e-05,
      "loss": 0.2649,
      "step": 11090
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.4671147383581375e-05,
      "loss": 0.0722,
      "step": 11100
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.4666346615458476e-05,
      "loss": 0.1807,
      "step": 11110
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.4661545847335575e-05,
      "loss": 0.2237,
      "step": 11120
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.4656745079212676e-05,
      "loss": 0.0997,
      "step": 11130
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.4651944311089777e-05,
      "loss": 0.09,
      "step": 11140
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.4647143542966876e-05,
      "loss": 0.2057,
      "step": 11150
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.4642342774843977e-05,
      "loss": 0.2111,
      "step": 11160
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.4637542006721078e-05,
      "loss": 0.0847,
      "step": 11170
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.4632741238598177e-05,
      "loss": 0.2067,
      "step": 11180
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.4627940470475278e-05,
      "loss": 0.1233,
      "step": 11190
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.4623139702352378e-05,
      "loss": 0.1625,
      "step": 11200
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.4618338934229478e-05,
      "loss": 0.4326,
      "step": 11210
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.4613538166106578e-05,
      "loss": 0.2955,
      "step": 11220
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.460873739798368e-05,
      "loss": 0.2861,
      "step": 11230
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.460393662986078e-05,
      "loss": 0.2929,
      "step": 11240
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.459913586173788e-05,
      "loss": 0.264,
      "step": 11250
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.459433509361498e-05,
      "loss": 0.2689,
      "step": 11260
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.4589534325492081e-05,
      "loss": 0.1043,
      "step": 11270
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.458473355736918e-05,
      "loss": 0.0879,
      "step": 11280
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.4579932789246281e-05,
      "loss": 0.1821,
      "step": 11290
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.4575132021123382e-05,
      "loss": 0.1557,
      "step": 11300
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.4570331253000481e-05,
      "loss": 0.148,
      "step": 11310
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.4565530484877582e-05,
      "loss": 0.1579,
      "step": 11320
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.4560729716754683e-05,
      "loss": 0.3617,
      "step": 11330
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.4555928948631782e-05,
      "loss": 0.1216,
      "step": 11340
    },
    {
      "epoch": 2.72,
      "learning_rate": 1.4551128180508883e-05,
      "loss": 0.117,
      "step": 11350
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.4546327412385983e-05,
      "loss": 0.2943,
      "step": 11360
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.4541526644263084e-05,
      "loss": 0.3734,
      "step": 11370
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.4536725876140183e-05,
      "loss": 0.2759,
      "step": 11380
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.4531925108017284e-05,
      "loss": 0.2898,
      "step": 11390
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.4527124339894385e-05,
      "loss": 0.294,
      "step": 11400
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.4522323571771484e-05,
      "loss": 0.4532,
      "step": 11410
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.4517522803648585e-05,
      "loss": 0.1752,
      "step": 11420
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.4512722035525686e-05,
      "loss": 0.2728,
      "step": 11430
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.4507921267402785e-05,
      "loss": 0.2024,
      "step": 11440
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.4503120499279886e-05,
      "loss": 0.2437,
      "step": 11450
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.4498319731156987e-05,
      "loss": 0.0856,
      "step": 11460
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.4493518963034086e-05,
      "loss": 0.1678,
      "step": 11470
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.4488718194911187e-05,
      "loss": 0.117,
      "step": 11480
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.4483917426788288e-05,
      "loss": 0.2189,
      "step": 11490
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.4479116658665388e-05,
      "loss": 0.1801,
      "step": 11500
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.4474315890542488e-05,
      "loss": 0.1597,
      "step": 11510
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.4469515122419588e-05,
      "loss": 0.1212,
      "step": 11520
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.446471435429669e-05,
      "loss": 0.3085,
      "step": 11530
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.4459913586173788e-05,
      "loss": 0.2141,
      "step": 11540
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.445511281805089e-05,
      "loss": 0.1987,
      "step": 11550
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.445031204992799e-05,
      "loss": 0.1391,
      "step": 11560
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.444551128180509e-05,
      "loss": 0.2763,
      "step": 11570
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.444071051368219e-05,
      "loss": 0.181,
      "step": 11580
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.4435909745559291e-05,
      "loss": 0.3497,
      "step": 11590
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.443110897743639e-05,
      "loss": 0.1695,
      "step": 11600
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.4426308209313491e-05,
      "loss": 0.1573,
      "step": 11610
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.4421507441190592e-05,
      "loss": 0.1493,
      "step": 11620
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.4416706673067693e-05,
      "loss": 0.3066,
      "step": 11630
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.4411905904944792e-05,
      "loss": 0.3554,
      "step": 11640
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.4407105136821893e-05,
      "loss": 0.2408,
      "step": 11650
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.4402304368698994e-05,
      "loss": 0.1179,
      "step": 11660
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.4397503600576093e-05,
      "loss": 0.1495,
      "step": 11670
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.4392702832453194e-05,
      "loss": 0.2195,
      "step": 11680
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.4387902064330294e-05,
      "loss": 0.2917,
      "step": 11690
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.4383101296207394e-05,
      "loss": 0.1422,
      "step": 11700
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.4378300528084494e-05,
      "loss": 0.1392,
      "step": 11710
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.4373499759961595e-05,
      "loss": 0.2353,
      "step": 11720
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.4368698991838694e-05,
      "loss": 0.2834,
      "step": 11730
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.4363898223715795e-05,
      "loss": 0.2711,
      "step": 11740
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.4359097455592896e-05,
      "loss": 0.2033,
      "step": 11750
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.4354296687469997e-05,
      "loss": 0.1748,
      "step": 11760
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.4349495919347096e-05,
      "loss": 0.164,
      "step": 11770
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.4344695151224197e-05,
      "loss": 0.1899,
      "step": 11780
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.4339894383101298e-05,
      "loss": 0.1409,
      "step": 11790
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.4335093614978397e-05,
      "loss": 0.1571,
      "step": 11800
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.4330292846855498e-05,
      "loss": 0.2338,
      "step": 11810
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4325492078732599e-05,
      "loss": 0.3575,
      "step": 11820
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4320691310609698e-05,
      "loss": 0.3519,
      "step": 11830
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.4315890542486799e-05,
      "loss": 0.1915,
      "step": 11840
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.43110897743639e-05,
      "loss": 0.1729,
      "step": 11850
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.4306289006240999e-05,
      "loss": 0.1198,
      "step": 11860
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.43014882381181e-05,
      "loss": 0.1349,
      "step": 11870
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.42966874699952e-05,
      "loss": 0.2092,
      "step": 11880
    },
    {
      "epoch": 2.85,
      "learning_rate": 1.4291886701872301e-05,
      "loss": 0.0536,
      "step": 11890
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.42870859337494e-05,
      "loss": 0.1428,
      "step": 11900
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.4282285165626501e-05,
      "loss": 0.1473,
      "step": 11910
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.4277484397503602e-05,
      "loss": 0.2543,
      "step": 11920
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.4272683629380701e-05,
      "loss": 0.1749,
      "step": 11930
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4267882861257802e-05,
      "loss": 0.183,
      "step": 11940
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4263082093134903e-05,
      "loss": 0.3432,
      "step": 11950
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4258281325012002e-05,
      "loss": 0.2002,
      "step": 11960
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.4253480556889103e-05,
      "loss": 0.2997,
      "step": 11970
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4248679788766204e-05,
      "loss": 0.2017,
      "step": 11980
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4243879020643303e-05,
      "loss": 0.0466,
      "step": 11990
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4239078252520404e-05,
      "loss": 0.3021,
      "step": 12000
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.4234277484397504e-05,
      "loss": 0.2667,
      "step": 12010
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.4229476716274607e-05,
      "loss": 0.2184,
      "step": 12020
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.4224675948151704e-05,
      "loss": 0.1655,
      "step": 12030
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.4219875180028805e-05,
      "loss": 0.0865,
      "step": 12040
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.4215074411905908e-05,
      "loss": 0.1376,
      "step": 12050
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.4210273643783005e-05,
      "loss": 0.1216,
      "step": 12060
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.4205472875660106e-05,
      "loss": 0.1288,
      "step": 12070
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.4200672107537209e-05,
      "loss": 0.0865,
      "step": 12080
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.4195871339414306e-05,
      "loss": 0.2608,
      "step": 12090
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.4191070571291407e-05,
      "loss": 0.0557,
      "step": 12100
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.418626980316851e-05,
      "loss": 0.2073,
      "step": 12110
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.4181469035045607e-05,
      "loss": 0.2566,
      "step": 12120
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.4176668266922708e-05,
      "loss": 0.1452,
      "step": 12130
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.417186749879981e-05,
      "loss": 0.2432,
      "step": 12140
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4167066730676911e-05,
      "loss": 0.2239,
      "step": 12150
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4162265962554009e-05,
      "loss": 0.2452,
      "step": 12160
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4157465194431111e-05,
      "loss": 0.2074,
      "step": 12170
    },
    {
      "epoch": 2.92,
      "learning_rate": 1.4152664426308212e-05,
      "loss": 0.4822,
      "step": 12180
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.414786365818531e-05,
      "loss": 0.2003,
      "step": 12190
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.4143062890062412e-05,
      "loss": 0.2836,
      "step": 12200
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.4138262121939513e-05,
      "loss": 0.2733,
      "step": 12210
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.413346135381661e-05,
      "loss": 0.2449,
      "step": 12220
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.4128660585693713e-05,
      "loss": 0.4129,
      "step": 12230
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.4123859817570814e-05,
      "loss": 0.2572,
      "step": 12240
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.4119059049447911e-05,
      "loss": 0.2052,
      "step": 12250
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.4114258281325014e-05,
      "loss": 0.0483,
      "step": 12260
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4109457513202114e-05,
      "loss": 0.2035,
      "step": 12270
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4104656745079215e-05,
      "loss": 0.1016,
      "step": 12280
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4099855976956314e-05,
      "loss": 0.1477,
      "step": 12290
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4095055208833415e-05,
      "loss": 0.1118,
      "step": 12300
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.4090254440710516e-05,
      "loss": 0.2902,
      "step": 12310
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.4085453672587615e-05,
      "loss": 0.1305,
      "step": 12320
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.4080652904464716e-05,
      "loss": 0.2209,
      "step": 12330
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.4075852136341817e-05,
      "loss": 0.1331,
      "step": 12340
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.4071051368218916e-05,
      "loss": 0.1694,
      "step": 12350
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.4066250600096017e-05,
      "loss": 0.2619,
      "step": 12360
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.4061449831973118e-05,
      "loss": 0.3794,
      "step": 12370
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.4056649063850217e-05,
      "loss": 0.2388,
      "step": 12380
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.4051848295727318e-05,
      "loss": 0.3372,
      "step": 12390
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.4047047527604419e-05,
      "loss": 0.1956,
      "step": 12400
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.404224675948152e-05,
      "loss": 0.1022,
      "step": 12410
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.4037445991358619e-05,
      "loss": 0.1557,
      "step": 12420
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.403264522323572e-05,
      "loss": 0.031,
      "step": 12430
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.402784445511282e-05,
      "loss": 0.1382,
      "step": 12440
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.402304368698992e-05,
      "loss": 0.1419,
      "step": 12450
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.401824291886702e-05,
      "loss": 0.3698,
      "step": 12460
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.4013442150744121e-05,
      "loss": 0.0967,
      "step": 12470
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.400864138262122e-05,
      "loss": 0.1026,
      "step": 12480
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.4003840614498321e-05,
      "loss": 0.2825,
      "step": 12490
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.9971098265895953,
      "eval_loss": 0.015470546670258045,
      "eval_runtime": 514.2587,
      "eval_samples_per_second": 11.438,
      "eval_steps_per_second": 1.431,
      "step": 12498
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.3999039846375422e-05,
      "loss": 0.1956,
      "step": 12500
    },
    {
      "epoch": 3.0,
      "learning_rate": 1.3994239078252521e-05,
      "loss": 0.0667,
      "step": 12510
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.3989438310129622e-05,
      "loss": 0.3162,
      "step": 12520
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.3984637542006723e-05,
      "loss": 0.1426,
      "step": 12530
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.3979836773883824e-05,
      "loss": 0.1479,
      "step": 12540
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.3975036005760923e-05,
      "loss": 0.1379,
      "step": 12550
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.3970235237638024e-05,
      "loss": 0.1264,
      "step": 12560
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.3965434469515125e-05,
      "loss": 0.0989,
      "step": 12570
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.3960633701392224e-05,
      "loss": 0.1482,
      "step": 12580
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.3955832933269324e-05,
      "loss": 0.1873,
      "step": 12590
    },
    {
      "epoch": 3.02,
      "learning_rate": 1.3951032165146425e-05,
      "loss": 0.3184,
      "step": 12600
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.3946231397023524e-05,
      "loss": 0.1699,
      "step": 12610
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.3941430628900625e-05,
      "loss": 0.1943,
      "step": 12620
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.3936629860777726e-05,
      "loss": 0.1227,
      "step": 12630
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.3931829092654825e-05,
      "loss": 0.1615,
      "step": 12640
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.3927028324531926e-05,
      "loss": 0.2606,
      "step": 12650
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.3922227556409027e-05,
      "loss": 0.2146,
      "step": 12660
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.3917426788286126e-05,
      "loss": 0.18,
      "step": 12670
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.3912626020163227e-05,
      "loss": 0.1854,
      "step": 12680
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.3907825252040328e-05,
      "loss": 0.2845,
      "step": 12690
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.3903024483917429e-05,
      "loss": 0.1362,
      "step": 12700
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.3898223715794528e-05,
      "loss": 0.4133,
      "step": 12710
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.3893422947671629e-05,
      "loss": 0.3149,
      "step": 12720
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.388862217954873e-05,
      "loss": 0.3335,
      "step": 12730
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.3883821411425829e-05,
      "loss": 0.189,
      "step": 12740
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.387902064330293e-05,
      "loss": 0.2713,
      "step": 12750
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.387421987518003e-05,
      "loss": 0.2291,
      "step": 12760
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.386941910705713e-05,
      "loss": 0.14,
      "step": 12770
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.386461833893423e-05,
      "loss": 0.2882,
      "step": 12780
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.3859817570811331e-05,
      "loss": 0.1449,
      "step": 12790
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.385501680268843e-05,
      "loss": 0.1463,
      "step": 12800
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.3850216034565531e-05,
      "loss": 0.1283,
      "step": 12810
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.3845415266442632e-05,
      "loss": 0.2423,
      "step": 12820
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.3840614498319733e-05,
      "loss": 0.1881,
      "step": 12830
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.3835813730196832e-05,
      "loss": 0.1671,
      "step": 12840
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.3831012962073933e-05,
      "loss": 0.1666,
      "step": 12850
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.3826212193951034e-05,
      "loss": 0.2601,
      "step": 12860
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.3821411425828133e-05,
      "loss": 0.0795,
      "step": 12870
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.3816610657705234e-05,
      "loss": 0.224,
      "step": 12880
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.3811809889582335e-05,
      "loss": 0.1724,
      "step": 12890
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.3807009121459434e-05,
      "loss": 0.2105,
      "step": 12900
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.3802208353336535e-05,
      "loss": 0.1632,
      "step": 12910
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.3797407585213635e-05,
      "loss": 0.1207,
      "step": 12920
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.3792606817090735e-05,
      "loss": 0.1772,
      "step": 12930
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.3787806048967835e-05,
      "loss": 0.1582,
      "step": 12940
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.3783005280844936e-05,
      "loss": 0.2478,
      "step": 12950
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.3778204512722037e-05,
      "loss": 0.2557,
      "step": 12960
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.3773403744599136e-05,
      "loss": 0.1532,
      "step": 12970
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.3768602976476237e-05,
      "loss": 0.0719,
      "step": 12980
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.3763802208353338e-05,
      "loss": 0.1269,
      "step": 12990
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.3759001440230437e-05,
      "loss": 0.2606,
      "step": 13000
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.3754200672107538e-05,
      "loss": 0.0655,
      "step": 13010
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.3749399903984639e-05,
      "loss": 0.3242,
      "step": 13020
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.3744599135861738e-05,
      "loss": 0.1404,
      "step": 13030
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.3739798367738839e-05,
      "loss": 0.1755,
      "step": 13040
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.373499759961594e-05,
      "loss": 0.1227,
      "step": 13050
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.3730196831493039e-05,
      "loss": 0.1876,
      "step": 13060
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.372539606337014e-05,
      "loss": 0.2574,
      "step": 13070
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.372059529524724e-05,
      "loss": 0.231,
      "step": 13080
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.3715794527124341e-05,
      "loss": 0.1954,
      "step": 13090
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.371099375900144e-05,
      "loss": 0.0904,
      "step": 13100
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3706192990878541e-05,
      "loss": 0.2386,
      "step": 13110
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3701392222755642e-05,
      "loss": 0.1465,
      "step": 13120
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3696591454632741e-05,
      "loss": 0.2315,
      "step": 13130
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3691790686509842e-05,
      "loss": 0.1774,
      "step": 13140
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3686989918386943e-05,
      "loss": 0.1009,
      "step": 13150
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3682189150264042e-05,
      "loss": 0.2987,
      "step": 13160
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3677388382141143e-05,
      "loss": 0.1496,
      "step": 13170
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3672587614018244e-05,
      "loss": 0.0935,
      "step": 13180
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3667786845895343e-05,
      "loss": 0.1828,
      "step": 13190
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3662986077772444e-05,
      "loss": 0.1867,
      "step": 13200
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3658185309649545e-05,
      "loss": 0.1345,
      "step": 13210
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3653384541526647e-05,
      "loss": 0.0513,
      "step": 13220
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.3648583773403745e-05,
      "loss": 0.3241,
      "step": 13230
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.3643783005280845e-05,
      "loss": 0.0991,
      "step": 13240
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.3638982237157948e-05,
      "loss": 0.1115,
      "step": 13250
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.3634181469035045e-05,
      "loss": 0.224,
      "step": 13260
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.3629380700912146e-05,
      "loss": 0.2751,
      "step": 13270
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.3624579932789249e-05,
      "loss": 0.0207,
      "step": 13280
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.3619779164666346e-05,
      "loss": 0.0919,
      "step": 13290
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.3614978396543447e-05,
      "loss": 0.3005,
      "step": 13300
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.361017762842055e-05,
      "loss": 0.3086,
      "step": 13310
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.3605376860297647e-05,
      "loss": 0.2728,
      "step": 13320
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.3600576092174748e-05,
      "loss": 0.0788,
      "step": 13330
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.359577532405185e-05,
      "loss": 0.1995,
      "step": 13340
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.3590974555928951e-05,
      "loss": 0.2567,
      "step": 13350
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.3586173787806049e-05,
      "loss": 0.1863,
      "step": 13360
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.3581373019683151e-05,
      "loss": 0.2475,
      "step": 13370
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.3576572251560252e-05,
      "loss": 0.1014,
      "step": 13380
    },
    {
      "epoch": 3.21,
      "learning_rate": 1.357177148343735e-05,
      "loss": 0.2758,
      "step": 13390
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.3566970715314452e-05,
      "loss": 0.221,
      "step": 13400
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.3562169947191553e-05,
      "loss": 0.1227,
      "step": 13410
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.355736917906865e-05,
      "loss": 0.1679,
      "step": 13420
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.3552568410945753e-05,
      "loss": 0.231,
      "step": 13430
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.3547767642822854e-05,
      "loss": 0.3235,
      "step": 13440
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.3542966874699951e-05,
      "loss": 0.2003,
      "step": 13450
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.3538166106577054e-05,
      "loss": 0.2498,
      "step": 13460
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.3533365338454155e-05,
      "loss": 0.212,
      "step": 13470
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.3528564570331255e-05,
      "loss": 0.2615,
      "step": 13480
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.3523763802208355e-05,
      "loss": 0.2424,
      "step": 13490
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.3518963034085455e-05,
      "loss": 0.1213,
      "step": 13500
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.3514162265962556e-05,
      "loss": 0.0742,
      "step": 13510
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.3509361497839655e-05,
      "loss": 0.0757,
      "step": 13520
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.3504560729716756e-05,
      "loss": 0.2055,
      "step": 13530
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.3499759961593857e-05,
      "loss": 0.0865,
      "step": 13540
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.3494959193470956e-05,
      "loss": 0.1126,
      "step": 13550
    },
    {
      "epoch": 3.25,
      "learning_rate": 1.3490158425348057e-05,
      "loss": 0.3053,
      "step": 13560
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.3485357657225158e-05,
      "loss": 0.1582,
      "step": 13570
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.3480556889102257e-05,
      "loss": 0.1463,
      "step": 13580
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.3475756120979358e-05,
      "loss": 0.3074,
      "step": 13590
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.3470955352856459e-05,
      "loss": 0.1445,
      "step": 13600
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.346615458473356e-05,
      "loss": 0.1317,
      "step": 13610
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.3461353816610659e-05,
      "loss": 0.2067,
      "step": 13620
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.345655304848776e-05,
      "loss": 0.3204,
      "step": 13630
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.345175228036486e-05,
      "loss": 0.1537,
      "step": 13640
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.344695151224196e-05,
      "loss": 0.2585,
      "step": 13650
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.344215074411906e-05,
      "loss": 0.2489,
      "step": 13660
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.3437349975996161e-05,
      "loss": 0.1383,
      "step": 13670
    },
    {
      "epoch": 3.28,
      "learning_rate": 1.343254920787326e-05,
      "loss": 0.0854,
      "step": 13680
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.3427748439750361e-05,
      "loss": 0.254,
      "step": 13690
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.3422947671627462e-05,
      "loss": 0.379,
      "step": 13700
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.3418146903504561e-05,
      "loss": 0.2196,
      "step": 13710
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.3413346135381662e-05,
      "loss": 0.3524,
      "step": 13720
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.3408545367258763e-05,
      "loss": 0.0453,
      "step": 13730
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.3403744599135864e-05,
      "loss": 0.1446,
      "step": 13740
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.3398943831012963e-05,
      "loss": 0.2736,
      "step": 13750
    },
    {
      "epoch": 3.3,
      "learning_rate": 1.3394143062890064e-05,
      "loss": 0.1868,
      "step": 13760
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.3389342294767165e-05,
      "loss": 0.1,
      "step": 13770
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.3384541526644264e-05,
      "loss": 0.199,
      "step": 13780
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.3379740758521365e-05,
      "loss": 0.2656,
      "step": 13790
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.3374939990398466e-05,
      "loss": 0.1734,
      "step": 13800
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.3370139222275565e-05,
      "loss": 0.21,
      "step": 13810
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.3365338454152666e-05,
      "loss": 0.285,
      "step": 13820
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.3360537686029766e-05,
      "loss": 0.2771,
      "step": 13830
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.3355736917906866e-05,
      "loss": 0.2536,
      "step": 13840
    },
    {
      "epoch": 3.32,
      "learning_rate": 1.3350936149783966e-05,
      "loss": 0.1195,
      "step": 13850
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.3346135381661067e-05,
      "loss": 0.1885,
      "step": 13860
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.3341334613538168e-05,
      "loss": 0.1271,
      "step": 13870
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.3336533845415267e-05,
      "loss": 0.0776,
      "step": 13880
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.3331733077292368e-05,
      "loss": 0.0926,
      "step": 13890
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.3326932309169469e-05,
      "loss": 0.1214,
      "step": 13900
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.3322131541046568e-05,
      "loss": 0.3027,
      "step": 13910
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.3317330772923669e-05,
      "loss": 0.2863,
      "step": 13920
    },
    {
      "epoch": 3.34,
      "learning_rate": 1.331253000480077e-05,
      "loss": 0.2322,
      "step": 13930
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.3307729236677869e-05,
      "loss": 0.21,
      "step": 13940
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.330292846855497e-05,
      "loss": 0.0087,
      "step": 13950
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.329812770043207e-05,
      "loss": 0.1244,
      "step": 13960
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.329332693230917e-05,
      "loss": 0.098,
      "step": 13970
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.328852616418627e-05,
      "loss": 0.0396,
      "step": 13980
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.3283725396063371e-05,
      "loss": 0.2156,
      "step": 13990
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.3278924627940472e-05,
      "loss": 0.2285,
      "step": 14000
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.3274123859817571e-05,
      "loss": 0.093,
      "step": 14010
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3269323091694672e-05,
      "loss": 0.0888,
      "step": 14020
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3264522323571773e-05,
      "loss": 0.2583,
      "step": 14030
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3259721555448872e-05,
      "loss": 0.0801,
      "step": 14040
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3254920787325973e-05,
      "loss": 0.2272,
      "step": 14050
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.3250120019203074e-05,
      "loss": 0.1998,
      "step": 14060
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.3245319251080173e-05,
      "loss": 0.0903,
      "step": 14070
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.3240518482957274e-05,
      "loss": 0.1902,
      "step": 14080
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.3235717714834375e-05,
      "loss": 0.114,
      "step": 14090
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.3230916946711474e-05,
      "loss": 0.1958,
      "step": 14100
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.3226116178588575e-05,
      "loss": 0.2417,
      "step": 14110
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.3221315410465676e-05,
      "loss": 0.1614,
      "step": 14120
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.3216514642342776e-05,
      "loss": 0.0816,
      "step": 14130
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.3211713874219876e-05,
      "loss": 0.1514,
      "step": 14140
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.3206913106096976e-05,
      "loss": 0.0999,
      "step": 14150
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.3202112337974077e-05,
      "loss": 0.2251,
      "step": 14160
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.3197311569851176e-05,
      "loss": 0.2379,
      "step": 14170
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.3192510801728277e-05,
      "loss": 0.1113,
      "step": 14180
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.3187710033605378e-05,
      "loss": 0.1042,
      "step": 14190
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.3182909265482477e-05,
      "loss": 0.2047,
      "step": 14200
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.3178108497359578e-05,
      "loss": 0.2152,
      "step": 14210
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.3173307729236679e-05,
      "loss": 0.1043,
      "step": 14220
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.3168506961113778e-05,
      "loss": 0.2235,
      "step": 14230
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.3163706192990879e-05,
      "loss": 0.0452,
      "step": 14240
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.315890542486798e-05,
      "loss": 0.2039,
      "step": 14250
    },
    {
      "epoch": 3.42,
      "learning_rate": 1.3154104656745079e-05,
      "loss": 0.1909,
      "step": 14260
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.314930388862218e-05,
      "loss": 0.0699,
      "step": 14270
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.314450312049928e-05,
      "loss": 0.2831,
      "step": 14280
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.3139702352376381e-05,
      "loss": 0.2112,
      "step": 14290
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.313490158425348e-05,
      "loss": 0.1573,
      "step": 14300
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.3130100816130581e-05,
      "loss": 0.2213,
      "step": 14310
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.3125300048007682e-05,
      "loss": 0.2092,
      "step": 14320
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.3120499279884781e-05,
      "loss": 0.2019,
      "step": 14330
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.3115698511761882e-05,
      "loss": 0.0454,
      "step": 14340
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.3110897743638983e-05,
      "loss": 0.3562,
      "step": 14350
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.3106096975516082e-05,
      "loss": 0.1206,
      "step": 14360
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.3101296207393183e-05,
      "loss": 0.2038,
      "step": 14370
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.3096495439270284e-05,
      "loss": 0.181,
      "step": 14380
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.3091694671147383e-05,
      "loss": 0.0628,
      "step": 14390
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.3086893903024484e-05,
      "loss": 0.2226,
      "step": 14400
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.3082093134901585e-05,
      "loss": 0.0373,
      "step": 14410
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.3077292366778687e-05,
      "loss": 0.1289,
      "step": 14420
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.3072491598655785e-05,
      "loss": 0.0771,
      "step": 14430
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.3067690830532886e-05,
      "loss": 0.2937,
      "step": 14440
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.3062890062409988e-05,
      "loss": 0.2052,
      "step": 14450
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.3058089294287086e-05,
      "loss": 0.185,
      "step": 14460
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.3053288526164186e-05,
      "loss": 0.1627,
      "step": 14470
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.3048487758041289e-05,
      "loss": 0.0552,
      "step": 14480
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.3043686989918386e-05,
      "loss": 0.0897,
      "step": 14490
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.3038886221795489e-05,
      "loss": 0.2872,
      "step": 14500
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.303408545367259e-05,
      "loss": 0.1088,
      "step": 14510
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.3029284685549687e-05,
      "loss": 0.2288,
      "step": 14520
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.302448391742679e-05,
      "loss": 0.0859,
      "step": 14530
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.301968314930389e-05,
      "loss": 0.2351,
      "step": 14540
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.3014882381180991e-05,
      "loss": 0.2432,
      "step": 14550
    },
    {
      "epoch": 3.49,
      "learning_rate": 1.301008161305809e-05,
      "loss": 0.193,
      "step": 14560
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.3005280844935191e-05,
      "loss": 0.1116,
      "step": 14570
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.3000480076812292e-05,
      "loss": 0.1057,
      "step": 14580
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.2995679308689391e-05,
      "loss": 0.209,
      "step": 14590
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.2990878540566492e-05,
      "loss": 0.1492,
      "step": 14600
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.2986077772443593e-05,
      "loss": 0.1865,
      "step": 14610
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.2981277004320692e-05,
      "loss": 0.0575,
      "step": 14620
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.2976476236197793e-05,
      "loss": 0.226,
      "step": 14630
    },
    {
      "epoch": 3.51,
      "learning_rate": 1.2971675468074894e-05,
      "loss": 0.0842,
      "step": 14640
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.2966874699951993e-05,
      "loss": 0.2408,
      "step": 14650
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.2962073931829094e-05,
      "loss": 0.2219,
      "step": 14660
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.2957273163706195e-05,
      "loss": 0.1683,
      "step": 14670
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.2952472395583296e-05,
      "loss": 0.1358,
      "step": 14680
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.2947671627460395e-05,
      "loss": 0.3367,
      "step": 14690
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.2942870859337496e-05,
      "loss": 0.2099,
      "step": 14700
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.2938070091214597e-05,
      "loss": 0.2868,
      "step": 14710
    },
    {
      "epoch": 3.53,
      "learning_rate": 1.2933269323091696e-05,
      "loss": 0.1333,
      "step": 14720
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.2928468554968797e-05,
      "loss": 0.1113,
      "step": 14730
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.2923667786845897e-05,
      "loss": 0.2277,
      "step": 14740
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.2918867018722997e-05,
      "loss": 0.1579,
      "step": 14750
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.2914066250600097e-05,
      "loss": 0.1595,
      "step": 14760
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.2909265482477198e-05,
      "loss": 0.3039,
      "step": 14770
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.2904464714354297e-05,
      "loss": 0.1915,
      "step": 14780
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.2899663946231398e-05,
      "loss": 0.0972,
      "step": 14790
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.2894863178108499e-05,
      "loss": 0.2483,
      "step": 14800
    },
    {
      "epoch": 3.55,
      "learning_rate": 1.28900624099856e-05,
      "loss": 0.2983,
      "step": 14810
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.2885261641862699e-05,
      "loss": 0.1354,
      "step": 14820
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.28804608737398e-05,
      "loss": 0.2247,
      "step": 14830
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.28756601056169e-05,
      "loss": 0.0247,
      "step": 14840
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.2870859337494e-05,
      "loss": 0.1584,
      "step": 14850
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.28660585693711e-05,
      "loss": 0.1437,
      "step": 14860
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.2861257801248202e-05,
      "loss": 0.1723,
      "step": 14870
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.28564570331253e-05,
      "loss": 0.0568,
      "step": 14880
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.2851656265002402e-05,
      "loss": 0.1586,
      "step": 14890
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.2846855496879502e-05,
      "loss": 0.0733,
      "step": 14900
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.2842054728756602e-05,
      "loss": 0.0914,
      "step": 14910
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.2837253960633702e-05,
      "loss": 0.2732,
      "step": 14920
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.2832453192510803e-05,
      "loss": 0.1728,
      "step": 14930
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.2827652424387904e-05,
      "loss": 0.1559,
      "step": 14940
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.2822851656265003e-05,
      "loss": 0.3054,
      "step": 14950
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.2818050888142104e-05,
      "loss": 0.2016,
      "step": 14960
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.2813250120019205e-05,
      "loss": 0.096,
      "step": 14970
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.2808449351896304e-05,
      "loss": 0.3411,
      "step": 14980
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.2803648583773405e-05,
      "loss": 0.1373,
      "step": 14990
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.2798847815650506e-05,
      "loss": 0.1088,
      "step": 15000
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.2794047047527605e-05,
      "loss": 0.0894,
      "step": 15010
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.2789246279404706e-05,
      "loss": 0.0775,
      "step": 15020
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.2784445511281807e-05,
      "loss": 0.2274,
      "step": 15030
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.2779644743158906e-05,
      "loss": 0.1833,
      "step": 15040
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.2774843975036007e-05,
      "loss": 0.0732,
      "step": 15050
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.2770043206913107e-05,
      "loss": 0.078,
      "step": 15060
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.2765242438790208e-05,
      "loss": 0.1856,
      "step": 15070
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.2760441670667307e-05,
      "loss": 0.1172,
      "step": 15080
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.2755640902544408e-05,
      "loss": 0.1687,
      "step": 15090
    },
    {
      "epoch": 3.62,
      "learning_rate": 1.2750840134421509e-05,
      "loss": 0.3823,
      "step": 15100
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.2746039366298608e-05,
      "loss": 0.2649,
      "step": 15110
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.2741238598175709e-05,
      "loss": 0.1884,
      "step": 15120
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.273643783005281e-05,
      "loss": 0.096,
      "step": 15130
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.2731637061929909e-05,
      "loss": 0.0961,
      "step": 15140
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.272683629380701e-05,
      "loss": 0.0619,
      "step": 15150
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.272203552568411e-05,
      "loss": 0.2301,
      "step": 15160
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.271723475756121e-05,
      "loss": 0.286,
      "step": 15170
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.271243398943831e-05,
      "loss": 0.1029,
      "step": 15180
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.2707633221315412e-05,
      "loss": 0.1848,
      "step": 15190
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.2702832453192512e-05,
      "loss": 0.0732,
      "step": 15200
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.2698031685069612e-05,
      "loss": 0.1323,
      "step": 15210
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.2693230916946712e-05,
      "loss": 0.1789,
      "step": 15220
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.2688430148823813e-05,
      "loss": 0.1113,
      "step": 15230
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.2683629380700912e-05,
      "loss": 0.1314,
      "step": 15240
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.2678828612578013e-05,
      "loss": 0.0374,
      "step": 15250
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.2674027844455114e-05,
      "loss": 0.2404,
      "step": 15260
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.2669227076332213e-05,
      "loss": 0.2263,
      "step": 15270
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.2664426308209314e-05,
      "loss": 0.1551,
      "step": 15280
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.2659625540086415e-05,
      "loss": 0.104,
      "step": 15290
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.2654824771963514e-05,
      "loss": 0.0599,
      "step": 15300
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.2650024003840615e-05,
      "loss": 0.2928,
      "step": 15310
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.2645223235717716e-05,
      "loss": 0.2366,
      "step": 15320
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.2640422467594817e-05,
      "loss": 0.1708,
      "step": 15330
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.2635621699471916e-05,
      "loss": 0.2391,
      "step": 15340
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.2630820931349017e-05,
      "loss": 0.0755,
      "step": 15350
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.2626020163226117e-05,
      "loss": 0.1634,
      "step": 15360
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.2621219395103217e-05,
      "loss": 0.2716,
      "step": 15370
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.2616418626980317e-05,
      "loss": 0.0929,
      "step": 15380
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.2611617858857418e-05,
      "loss": 0.0948,
      "step": 15390
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.2606817090734517e-05,
      "loss": 0.1283,
      "step": 15400
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.2602016322611618e-05,
      "loss": 0.2531,
      "step": 15410
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.2597215554488719e-05,
      "loss": 0.1154,
      "step": 15420
    },
    {
      "epoch": 3.7,
      "learning_rate": 1.2592414786365818e-05,
      "loss": 0.1397,
      "step": 15430
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.2587614018242919e-05,
      "loss": 0.3529,
      "step": 15440
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.258281325012002e-05,
      "loss": 0.2267,
      "step": 15450
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.2578012481997122e-05,
      "loss": 0.1405,
      "step": 15460
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.257321171387422e-05,
      "loss": 0.0208,
      "step": 15470
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.256841094575132e-05,
      "loss": 0.1602,
      "step": 15480
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.2563610177628423e-05,
      "loss": 0.2042,
      "step": 15490
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.255880940950552e-05,
      "loss": 0.0106,
      "step": 15500
    },
    {
      "epoch": 3.72,
      "learning_rate": 1.2554008641382622e-05,
      "loss": 0.212,
      "step": 15510
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.2549207873259724e-05,
      "loss": 0.1279,
      "step": 15520
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.2544407105136822e-05,
      "loss": 0.2245,
      "step": 15530
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.2539606337013922e-05,
      "loss": 0.1547,
      "step": 15540
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.2534805568891025e-05,
      "loss": 0.1279,
      "step": 15550
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.2530004800768122e-05,
      "loss": 0.1042,
      "step": 15560
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.2525204032645223e-05,
      "loss": 0.2089,
      "step": 15570
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.2520403264522326e-05,
      "loss": 0.2795,
      "step": 15580
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.2515602496399427e-05,
      "loss": 0.2023,
      "step": 15590
    },
    {
      "epoch": 3.74,
      "learning_rate": 1.2510801728276524e-05,
      "loss": 0.0131,
      "step": 15600
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.2506000960153627e-05,
      "loss": 0.2091,
      "step": 15610
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.2501200192030728e-05,
      "loss": 0.1836,
      "step": 15620
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.2496399423907825e-05,
      "loss": 0.1042,
      "step": 15630
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.2491598655784927e-05,
      "loss": 0.1711,
      "step": 15640
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.2486797887662028e-05,
      "loss": 0.2689,
      "step": 15650
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.2481997119539126e-05,
      "loss": 0.0546,
      "step": 15660
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.2477196351416228e-05,
      "loss": 0.0376,
      "step": 15670
    },
    {
      "epoch": 3.76,
      "learning_rate": 1.247239558329333e-05,
      "loss": 0.2063,
      "step": 15680
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.2467594815170427e-05,
      "loss": 0.2549,
      "step": 15690
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.246279404704753e-05,
      "loss": 0.1667,
      "step": 15700
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.245799327892463e-05,
      "loss": 0.2155,
      "step": 15710
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.2453192510801727e-05,
      "loss": 0.1918,
      "step": 15720
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.244839174267883e-05,
      "loss": 0.2764,
      "step": 15730
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.2443590974555931e-05,
      "loss": 0.2481,
      "step": 15740
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.2438790206433032e-05,
      "loss": 0.1298,
      "step": 15750
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.243398943831013e-05,
      "loss": 0.0848,
      "step": 15760
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.2429188670187232e-05,
      "loss": 0.2643,
      "step": 15770
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.2424387902064333e-05,
      "loss": 0.2249,
      "step": 15780
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.2419587133941432e-05,
      "loss": 0.1375,
      "step": 15790
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.2414786365818533e-05,
      "loss": 0.2448,
      "step": 15800
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.2409985597695633e-05,
      "loss": 0.1616,
      "step": 15810
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.2405184829572733e-05,
      "loss": 0.1173,
      "step": 15820
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.2400384061449833e-05,
      "loss": 0.0892,
      "step": 15830
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.2395583293326934e-05,
      "loss": 0.1786,
      "step": 15840
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.2390782525204033e-05,
      "loss": 0.301,
      "step": 15850
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2385981757081134e-05,
      "loss": 0.155,
      "step": 15860
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2381180988958235e-05,
      "loss": 0.1148,
      "step": 15870
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2376380220835336e-05,
      "loss": 0.1996,
      "step": 15880
    },
    {
      "epoch": 3.81,
      "learning_rate": 1.2371579452712435e-05,
      "loss": 0.2966,
      "step": 15890
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.2366778684589536e-05,
      "loss": 0.1079,
      "step": 15900
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.2361977916466637e-05,
      "loss": 0.2542,
      "step": 15910
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.2357177148343736e-05,
      "loss": 0.1314,
      "step": 15920
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.2352376380220837e-05,
      "loss": 0.223,
      "step": 15930
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.2347575612097938e-05,
      "loss": 0.1448,
      "step": 15940
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.2342774843975037e-05,
      "loss": 0.0932,
      "step": 15950
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.2337974075852138e-05,
      "loss": 0.092,
      "step": 15960
    },
    {
      "epoch": 3.83,
      "learning_rate": 1.2333173307729238e-05,
      "loss": 0.0868,
      "step": 15970
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.2328372539606338e-05,
      "loss": 0.2071,
      "step": 15980
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.2323571771483438e-05,
      "loss": 0.2415,
      "step": 15990
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.231877100336054e-05,
      "loss": 0.0858,
      "step": 16000
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.231397023523764e-05,
      "loss": 0.0158,
      "step": 16010
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.230916946711474e-05,
      "loss": 0.2116,
      "step": 16020
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.230436869899184e-05,
      "loss": 0.0287,
      "step": 16030
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.2299567930868941e-05,
      "loss": 0.1323,
      "step": 16040
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.229476716274604e-05,
      "loss": 0.0941,
      "step": 16050
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.2289966394623141e-05,
      "loss": 0.3061,
      "step": 16060
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.2285165626500242e-05,
      "loss": 0.0455,
      "step": 16070
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.2280364858377341e-05,
      "loss": 0.1998,
      "step": 16080
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.2275564090254442e-05,
      "loss": 0.4682,
      "step": 16090
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.2270763322131543e-05,
      "loss": 0.0575,
      "step": 16100
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.2265962554008642e-05,
      "loss": 0.0866,
      "step": 16110
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.2261161785885743e-05,
      "loss": 0.1079,
      "step": 16120
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.2256361017762843e-05,
      "loss": 0.1024,
      "step": 16130
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.2251560249639944e-05,
      "loss": 0.2423,
      "step": 16140
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.2246759481517043e-05,
      "loss": 0.1786,
      "step": 16150
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.2241958713394144e-05,
      "loss": 0.1133,
      "step": 16160
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.2237157945271245e-05,
      "loss": 0.1581,
      "step": 16170
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.2232357177148344e-05,
      "loss": 0.1095,
      "step": 16180
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.2227556409025445e-05,
      "loss": 0.3906,
      "step": 16190
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.2222755640902546e-05,
      "loss": 0.415,
      "step": 16200
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.2217954872779645e-05,
      "loss": 0.053,
      "step": 16210
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.2213154104656746e-05,
      "loss": 0.2106,
      "step": 16220
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.2208353336533847e-05,
      "loss": 0.1024,
      "step": 16230
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.2203552568410946e-05,
      "loss": 0.1113,
      "step": 16240
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.2198751800288047e-05,
      "loss": 0.2038,
      "step": 16250
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.2193951032165148e-05,
      "loss": 0.1252,
      "step": 16260
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.2189150264042248e-05,
      "loss": 0.1538,
      "step": 16270
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.2184349495919348e-05,
      "loss": 0.1372,
      "step": 16280
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.2179548727796448e-05,
      "loss": 0.1523,
      "step": 16290
    },
    {
      "epoch": 3.91,
      "learning_rate": 1.217474795967355e-05,
      "loss": 0.1139,
      "step": 16300
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.2169947191550648e-05,
      "loss": 0.1087,
      "step": 16310
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.216514642342775e-05,
      "loss": 0.1898,
      "step": 16320
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.216034565530485e-05,
      "loss": 0.1214,
      "step": 16330
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.215554488718195e-05,
      "loss": 0.2339,
      "step": 16340
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.215074411905905e-05,
      "loss": 0.203,
      "step": 16350
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.2145943350936151e-05,
      "loss": 0.2726,
      "step": 16360
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.214114258281325e-05,
      "loss": 0.1904,
      "step": 16370
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.2136341814690351e-05,
      "loss": 0.1762,
      "step": 16380
    },
    {
      "epoch": 3.93,
      "learning_rate": 1.2131541046567452e-05,
      "loss": 0.2288,
      "step": 16390
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.2126740278444553e-05,
      "loss": 0.0262,
      "step": 16400
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.2121939510321652e-05,
      "loss": 0.1427,
      "step": 16410
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.2117138742198753e-05,
      "loss": 0.124,
      "step": 16420
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.2112337974075853e-05,
      "loss": 0.129,
      "step": 16430
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.2107537205952953e-05,
      "loss": 0.1269,
      "step": 16440
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.2102736437830053e-05,
      "loss": 0.12,
      "step": 16450
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.2097935669707154e-05,
      "loss": 0.1905,
      "step": 16460
    },
    {
      "epoch": 3.95,
      "learning_rate": 1.2093134901584253e-05,
      "loss": 0.1511,
      "step": 16470
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.2088334133461354e-05,
      "loss": 0.112,
      "step": 16480
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.2083533365338455e-05,
      "loss": 0.1972,
      "step": 16490
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.2078732597215554e-05,
      "loss": 0.1213,
      "step": 16500
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.2073931829092655e-05,
      "loss": 0.2227,
      "step": 16510
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.2069131060969756e-05,
      "loss": 0.1679,
      "step": 16520
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.2064330292846857e-05,
      "loss": 0.2295,
      "step": 16530
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.2059529524723956e-05,
      "loss": 0.2329,
      "step": 16540
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.2054728756601057e-05,
      "loss": 0.0916,
      "step": 16550
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.2049927988478158e-05,
      "loss": 0.0707,
      "step": 16560
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.2045127220355257e-05,
      "loss": 0.1396,
      "step": 16570
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.2040326452232358e-05,
      "loss": 0.2439,
      "step": 16580
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.2035525684109458e-05,
      "loss": 0.0556,
      "step": 16590
    },
    {
      "epoch": 3.98,
      "learning_rate": 1.2030724915986558e-05,
      "loss": 0.1437,
      "step": 16600
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.2025924147863658e-05,
      "loss": 0.1124,
      "step": 16610
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.202112337974076e-05,
      "loss": 0.1308,
      "step": 16620
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.2016322611617858e-05,
      "loss": 0.3201,
      "step": 16630
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.201152184349496e-05,
      "loss": 0.1035,
      "step": 16640
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.200672107537206e-05,
      "loss": 0.3391,
      "step": 16650
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.2001920307249163e-05,
      "loss": 0.0953,
      "step": 16660
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.9982998979938796,
      "eval_loss": 0.01134026050567627,
      "eval_runtime": 514.5133,
      "eval_samples_per_second": 11.432,
      "eval_steps_per_second": 1.43,
      "step": 16664
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.199711953912626e-05,
      "loss": 0.1111,
      "step": 16670
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1992318771003361e-05,
      "loss": 0.0915,
      "step": 16680
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.1987518002880464e-05,
      "loss": 0.1586,
      "step": 16690
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.1982717234757561e-05,
      "loss": 0.3341,
      "step": 16700
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.1977916466634662e-05,
      "loss": 0.103,
      "step": 16710
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.1973115698511764e-05,
      "loss": 0.0366,
      "step": 16720
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.1968314930388862e-05,
      "loss": 0.1628,
      "step": 16730
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.1963514162265963e-05,
      "loss": 0.1286,
      "step": 16740
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.1958713394143065e-05,
      "loss": 0.2569,
      "step": 16750
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.1953912626020163e-05,
      "loss": 0.1188,
      "step": 16760
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.1949111857897263e-05,
      "loss": 0.2489,
      "step": 16770
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.1944311089774366e-05,
      "loss": 0.1957,
      "step": 16780
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.1939510321651467e-05,
      "loss": 0.1348,
      "step": 16790
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.1934709553528564e-05,
      "loss": 0.3714,
      "step": 16800
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.1929908785405667e-05,
      "loss": 0.1264,
      "step": 16810
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.1925108017282768e-05,
      "loss": 0.0909,
      "step": 16820
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.1920307249159865e-05,
      "loss": 0.1203,
      "step": 16830
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.1915506481036968e-05,
      "loss": 0.1898,
      "step": 16840
    },
    {
      "epoch": 4.04,
      "learning_rate": 1.1910705712914069e-05,
      "loss": 0.0655,
      "step": 16850
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.1905904944791166e-05,
      "loss": 0.0952,
      "step": 16860
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.1901104176668269e-05,
      "loss": 0.1401,
      "step": 16870
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.189630340854537e-05,
      "loss": 0.1629,
      "step": 16880
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.1891502640422467e-05,
      "loss": 0.1457,
      "step": 16890
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.188670187229957e-05,
      "loss": 0.0459,
      "step": 16900
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.188190110417667e-05,
      "loss": 0.1654,
      "step": 16910
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.1877100336053771e-05,
      "loss": 0.0484,
      "step": 16920
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.187229956793087e-05,
      "loss": 0.1428,
      "step": 16930
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.1867498799807971e-05,
      "loss": 0.1877,
      "step": 16940
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.1862698031685072e-05,
      "loss": 0.1808,
      "step": 16950
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.1857897263562171e-05,
      "loss": 0.1693,
      "step": 16960
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.1853096495439272e-05,
      "loss": 0.0924,
      "step": 16970
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.1848295727316373e-05,
      "loss": 0.3183,
      "step": 16980
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.1843494959193472e-05,
      "loss": 0.1466,
      "step": 16990
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.1838694191070573e-05,
      "loss": 0.0974,
      "step": 17000
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.1833893422947674e-05,
      "loss": 0.1933,
      "step": 17010
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.1829092654824773e-05,
      "loss": 0.1988,
      "step": 17020
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.1824291886701874e-05,
      "loss": 0.3429,
      "step": 17030
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.1819491118578974e-05,
      "loss": 0.1169,
      "step": 17040
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.1814690350456075e-05,
      "loss": 0.1339,
      "step": 17050
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.1809889582333174e-05,
      "loss": 0.1416,
      "step": 17060
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.1805088814210275e-05,
      "loss": 0.1262,
      "step": 17070
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.1800288046087376e-05,
      "loss": 0.1731,
      "step": 17080
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.1795487277964475e-05,
      "loss": 0.057,
      "step": 17090
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.1790686509841576e-05,
      "loss": 0.1439,
      "step": 17100
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.1785885741718677e-05,
      "loss": 0.1839,
      "step": 17110
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.1781084973595776e-05,
      "loss": 0.1761,
      "step": 17120
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.1776284205472877e-05,
      "loss": 0.0387,
      "step": 17130
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.1771483437349978e-05,
      "loss": 0.0616,
      "step": 17140
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.1766682669227077e-05,
      "loss": 0.3693,
      "step": 17150
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.1761881901104178e-05,
      "loss": 0.2378,
      "step": 17160
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.1757081132981279e-05,
      "loss": 0.2855,
      "step": 17170
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.1752280364858378e-05,
      "loss": 0.1147,
      "step": 17180
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.1747479596735479e-05,
      "loss": 0.0328,
      "step": 17190
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.174267882861258e-05,
      "loss": 0.1441,
      "step": 17200
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.173787806048968e-05,
      "loss": 0.0887,
      "step": 17210
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.173307729236678e-05,
      "loss": 0.1302,
      "step": 17220
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.172827652424388e-05,
      "loss": 0.2209,
      "step": 17230
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.1723475756120981e-05,
      "loss": 0.1899,
      "step": 17240
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.171867498799808e-05,
      "loss": 0.1993,
      "step": 17250
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.1713874219875181e-05,
      "loss": 0.2106,
      "step": 17260
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.1709073451752282e-05,
      "loss": 0.2002,
      "step": 17270
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.1704272683629381e-05,
      "loss": 0.1638,
      "step": 17280
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.1699471915506482e-05,
      "loss": 0.2725,
      "step": 17290
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.1694671147383583e-05,
      "loss": 0.2701,
      "step": 17300
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.1689870379260682e-05,
      "loss": 0.2262,
      "step": 17310
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.1685069611137783e-05,
      "loss": 0.1649,
      "step": 17320
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.1680268843014884e-05,
      "loss": 0.092,
      "step": 17330
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.1675468074891984e-05,
      "loss": 0.1584,
      "step": 17340
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.1670667306769084e-05,
      "loss": 0.1937,
      "step": 17350
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.1665866538646184e-05,
      "loss": 0.1445,
      "step": 17360
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.1661065770523285e-05,
      "loss": 0.1609,
      "step": 17370
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.1656265002400384e-05,
      "loss": 0.1111,
      "step": 17380
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.1651464234277485e-05,
      "loss": 0.0643,
      "step": 17390
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1646663466154586e-05,
      "loss": 0.1454,
      "step": 17400
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1641862698031685e-05,
      "loss": 0.1178,
      "step": 17410
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1637061929908786e-05,
      "loss": 0.0629,
      "step": 17420
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1632261161785887e-05,
      "loss": 0.0369,
      "step": 17430
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.1627460393662986e-05,
      "loss": 0.0455,
      "step": 17440
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.1622659625540087e-05,
      "loss": 0.107,
      "step": 17450
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.1617858857417188e-05,
      "loss": 0.148,
      "step": 17460
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.1613058089294289e-05,
      "loss": 0.1942,
      "step": 17470
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.1608257321171388e-05,
      "loss": 0.1286,
      "step": 17480
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.1603456553048489e-05,
      "loss": 0.13,
      "step": 17490
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.159865578492559e-05,
      "loss": 0.1229,
      "step": 17500
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.1593855016802689e-05,
      "loss": 0.1719,
      "step": 17510
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.158905424867979e-05,
      "loss": 0.1498,
      "step": 17520
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.158425348055689e-05,
      "loss": 0.1402,
      "step": 17530
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.157945271243399e-05,
      "loss": 0.3317,
      "step": 17540
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.157465194431109e-05,
      "loss": 0.1315,
      "step": 17550
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.1569851176188191e-05,
      "loss": 0.1067,
      "step": 17560
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.156505040806529e-05,
      "loss": 0.1837,
      "step": 17570
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.1560249639942391e-05,
      "loss": 0.1995,
      "step": 17580
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.1555448871819492e-05,
      "loss": 0.2198,
      "step": 17590
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.1550648103696593e-05,
      "loss": 0.1542,
      "step": 17600
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.1545847335573692e-05,
      "loss": 0.0782,
      "step": 17610
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.1541046567450793e-05,
      "loss": 0.2669,
      "step": 17620
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.1536245799327894e-05,
      "loss": 0.0132,
      "step": 17630
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.1531445031204993e-05,
      "loss": 0.2486,
      "step": 17640
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.1526644263082094e-05,
      "loss": 0.1794,
      "step": 17650
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.1521843494959194e-05,
      "loss": 0.0648,
      "step": 17660
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.1517042726836294e-05,
      "loss": 0.2922,
      "step": 17670
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.1512241958713394e-05,
      "loss": 0.0661,
      "step": 17680
    },
    {
      "epoch": 4.25,
      "learning_rate": 1.1507441190590495e-05,
      "loss": 0.1633,
      "step": 17690
    },
    {
      "epoch": 4.25,
      "learning_rate": 1.1502640422467594e-05,
      "loss": 0.061,
      "step": 17700
    },
    {
      "epoch": 4.25,
      "learning_rate": 1.1497839654344695e-05,
      "loss": 0.1788,
      "step": 17710
    },
    {
      "epoch": 4.25,
      "learning_rate": 1.1493038886221796e-05,
      "loss": 0.1882,
      "step": 17720
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.1488238118098897e-05,
      "loss": 0.3,
      "step": 17730
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.1483437349975996e-05,
      "loss": 0.3165,
      "step": 17740
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.1478636581853097e-05,
      "loss": 0.0919,
      "step": 17750
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.1473835813730198e-05,
      "loss": 0.1332,
      "step": 17760
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.1469035045607297e-05,
      "loss": 0.1854,
      "step": 17770
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.1464234277484398e-05,
      "loss": 0.053,
      "step": 17780
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.1459433509361499e-05,
      "loss": 0.2584,
      "step": 17790
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.1454632741238598e-05,
      "loss": 0.1938,
      "step": 17800
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.1449831973115699e-05,
      "loss": 0.1523,
      "step": 17810
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.14450312049928e-05,
      "loss": 0.0308,
      "step": 17820
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.1440230436869899e-05,
      "loss": 0.2052,
      "step": 17830
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.1435429668747e-05,
      "loss": 0.2572,
      "step": 17840
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.14306289006241e-05,
      "loss": 0.0975,
      "step": 17850
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.1425828132501203e-05,
      "loss": 0.2768,
      "step": 17860
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.14210273643783e-05,
      "loss": 0.2362,
      "step": 17870
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.1416226596255401e-05,
      "loss": 0.2411,
      "step": 17880
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.1411425828132504e-05,
      "loss": 0.1508,
      "step": 17890
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.1406625060009601e-05,
      "loss": 0.1502,
      "step": 17900
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.1401824291886702e-05,
      "loss": 0.0543,
      "step": 17910
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.1397023523763805e-05,
      "loss": 0.0654,
      "step": 17920
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.1392222755640902e-05,
      "loss": 0.1301,
      "step": 17930
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.1387421987518003e-05,
      "loss": 0.2797,
      "step": 17940
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.1382621219395105e-05,
      "loss": 0.2444,
      "step": 17950
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.1377820451272203e-05,
      "loss": 0.0555,
      "step": 17960
    },
    {
      "epoch": 4.31,
      "learning_rate": 1.1373019683149304e-05,
      "loss": 0.39,
      "step": 17970
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.1368218915026406e-05,
      "loss": 0.0872,
      "step": 17980
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.1363418146903507e-05,
      "loss": 0.1286,
      "step": 17990
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.1358617378780605e-05,
      "loss": 0.2268,
      "step": 18000
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.1353816610657707e-05,
      "loss": 0.1159,
      "step": 18010
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.1349015842534808e-05,
      "loss": 0.2392,
      "step": 18020
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.1344215074411905e-05,
      "loss": 0.1014,
      "step": 18030
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.1339414306289008e-05,
      "loss": 0.0968,
      "step": 18040
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.1334613538166109e-05,
      "loss": 0.156,
      "step": 18050
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.1329812770043206e-05,
      "loss": 0.083,
      "step": 18060
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.1325012001920309e-05,
      "loss": 0.1678,
      "step": 18070
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.132021123379741e-05,
      "loss": 0.1471,
      "step": 18080
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.1315410465674507e-05,
      "loss": 0.2576,
      "step": 18090
    },
    {
      "epoch": 4.34,
      "learning_rate": 1.131060969755161e-05,
      "loss": 0.1527,
      "step": 18100
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.130580892942871e-05,
      "loss": 0.0976,
      "step": 18110
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.1301008161305811e-05,
      "loss": 0.1236,
      "step": 18120
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.129620739318291e-05,
      "loss": 0.1652,
      "step": 18130
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.1291406625060011e-05,
      "loss": 0.2625,
      "step": 18140
    },
    {
      "epoch": 4.36,
      "learning_rate": 1.1286605856937112e-05,
      "loss": 0.1496,
      "step": 18150
    },
    {
      "epoch": 4.36,
      "learning_rate": 1.1281805088814211e-05,
      "loss": 0.2683,
      "step": 18160
    },
    {
      "epoch": 4.36,
      "learning_rate": 1.1277004320691312e-05,
      "loss": 0.0809,
      "step": 18170
    },
    {
      "epoch": 4.36,
      "learning_rate": 1.1272203552568413e-05,
      "loss": 0.1193,
      "step": 18180
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.1267402784445512e-05,
      "loss": 0.0153,
      "step": 18190
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.1262602016322613e-05,
      "loss": 0.3455,
      "step": 18200
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.1257801248199714e-05,
      "loss": 0.1779,
      "step": 18210
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.1253000480076813e-05,
      "loss": 0.0654,
      "step": 18220
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.1248199711953914e-05,
      "loss": 0.1511,
      "step": 18230
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.1243398943831015e-05,
      "loss": 0.0947,
      "step": 18240
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.1238598175708115e-05,
      "loss": 0.0489,
      "step": 18250
    },
    {
      "epoch": 4.38,
      "learning_rate": 1.1233797407585215e-05,
      "loss": 0.2878,
      "step": 18260
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.1228996639462315e-05,
      "loss": 0.1612,
      "step": 18270
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.1224195871339416e-05,
      "loss": 0.1771,
      "step": 18280
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.1219395103216515e-05,
      "loss": 0.0957,
      "step": 18290
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.1214594335093616e-05,
      "loss": 0.0988,
      "step": 18300
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.1209793566970717e-05,
      "loss": 0.0742,
      "step": 18310
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.1204992798847816e-05,
      "loss": 0.1248,
      "step": 18320
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.1200192030724917e-05,
      "loss": 0.2212,
      "step": 18330
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.1195391262602018e-05,
      "loss": 0.1609,
      "step": 18340
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.1190590494479117e-05,
      "loss": 0.034,
      "step": 18350
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.1185789726356218e-05,
      "loss": 0.2764,
      "step": 18360
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.1180988958233319e-05,
      "loss": 0.0646,
      "step": 18370
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.117618819011042e-05,
      "loss": 0.3224,
      "step": 18380
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.1171387421987519e-05,
      "loss": 0.3037,
      "step": 18390
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.116658665386462e-05,
      "loss": 0.1136,
      "step": 18400
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.116178588574172e-05,
      "loss": 0.1244,
      "step": 18410
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.115698511761882e-05,
      "loss": 0.1034,
      "step": 18420
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.115218434949592e-05,
      "loss": 0.1198,
      "step": 18430
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.1147383581373021e-05,
      "loss": 0.2697,
      "step": 18440
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.114258281325012e-05,
      "loss": 0.2753,
      "step": 18450
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.1137782045127221e-05,
      "loss": 0.0837,
      "step": 18460
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.1132981277004322e-05,
      "loss": 0.2019,
      "step": 18470
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.1128180508881421e-05,
      "loss": 0.2953,
      "step": 18480
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.1123379740758522e-05,
      "loss": 0.157,
      "step": 18490
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.1118578972635623e-05,
      "loss": 0.0641,
      "step": 18500
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.1113778204512724e-05,
      "loss": 0.1297,
      "step": 18510
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.1108977436389823e-05,
      "loss": 0.1481,
      "step": 18520
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.1104176668266924e-05,
      "loss": 0.0494,
      "step": 18530
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.1099375900144025e-05,
      "loss": 0.086,
      "step": 18540
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.1094575132021124e-05,
      "loss": 0.2122,
      "step": 18550
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.1089774363898225e-05,
      "loss": 0.091,
      "step": 18560
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.1084973595775325e-05,
      "loss": 0.1277,
      "step": 18570
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.1080172827652425e-05,
      "loss": 0.0822,
      "step": 18580
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.1075372059529525e-05,
      "loss": 0.1275,
      "step": 18590
    },
    {
      "epoch": 4.46,
      "learning_rate": 1.1070571291406626e-05,
      "loss": 0.1348,
      "step": 18600
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.1065770523283725e-05,
      "loss": 0.1216,
      "step": 18610
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.1060969755160826e-05,
      "loss": 0.2042,
      "step": 18620
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.1056168987037927e-05,
      "loss": 0.3463,
      "step": 18630
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.1051368218915026e-05,
      "loss": 0.065,
      "step": 18640
    },
    {
      "epoch": 4.48,
      "learning_rate": 1.1046567450792127e-05,
      "loss": 0.2147,
      "step": 18650
    },
    {
      "epoch": 4.48,
      "learning_rate": 1.1041766682669228e-05,
      "loss": 0.0212,
      "step": 18660
    },
    {
      "epoch": 4.48,
      "learning_rate": 1.1036965914546329e-05,
      "loss": 0.1466,
      "step": 18670
    },
    {
      "epoch": 4.48,
      "learning_rate": 1.1032165146423428e-05,
      "loss": 0.4545,
      "step": 18680
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.1027364378300529e-05,
      "loss": 0.2321,
      "step": 18690
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.102256361017763e-05,
      "loss": 0.1884,
      "step": 18700
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.1017762842054729e-05,
      "loss": 0.2749,
      "step": 18710
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.101296207393183e-05,
      "loss": 0.1542,
      "step": 18720
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.100816130580893e-05,
      "loss": 0.1061,
      "step": 18730
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.100336053768603e-05,
      "loss": 0.0758,
      "step": 18740
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.099855976956313e-05,
      "loss": 0.1873,
      "step": 18750
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.0993759001440231e-05,
      "loss": 0.0921,
      "step": 18760
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.098895823331733e-05,
      "loss": 0.1542,
      "step": 18770
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.0984157465194431e-05,
      "loss": 0.1391,
      "step": 18780
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.0979356697071532e-05,
      "loss": 0.1835,
      "step": 18790
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.0974555928948633e-05,
      "loss": 0.1821,
      "step": 18800
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.0969755160825732e-05,
      "loss": 0.1109,
      "step": 18810
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.0964954392702833e-05,
      "loss": 0.0555,
      "step": 18820
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.0960153624579934e-05,
      "loss": 0.2581,
      "step": 18830
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.0955352856457033e-05,
      "loss": 0.2598,
      "step": 18840
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.0950552088334134e-05,
      "loss": 0.203,
      "step": 18850
    },
    {
      "epoch": 4.53,
      "learning_rate": 1.0945751320211235e-05,
      "loss": 0.1052,
      "step": 18860
    },
    {
      "epoch": 4.53,
      "learning_rate": 1.0940950552088334e-05,
      "loss": 0.0144,
      "step": 18870
    },
    {
      "epoch": 4.53,
      "learning_rate": 1.0936149783965435e-05,
      "loss": 0.1813,
      "step": 18880
    },
    {
      "epoch": 4.53,
      "learning_rate": 1.0931349015842536e-05,
      "loss": 0.1861,
      "step": 18890
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.0926548247719635e-05,
      "loss": 0.1467,
      "step": 18900
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.0921747479596735e-05,
      "loss": 0.2253,
      "step": 18910
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.0916946711473836e-05,
      "loss": 0.1522,
      "step": 18920
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.0912145943350937e-05,
      "loss": 0.1577,
      "step": 18930
    },
    {
      "epoch": 4.55,
      "learning_rate": 1.0907345175228036e-05,
      "loss": 0.1278,
      "step": 18940
    },
    {
      "epoch": 4.55,
      "learning_rate": 1.0902544407105137e-05,
      "loss": 0.1783,
      "step": 18950
    },
    {
      "epoch": 4.55,
      "learning_rate": 1.0897743638982238e-05,
      "loss": 0.0888,
      "step": 18960
    },
    {
      "epoch": 4.55,
      "learning_rate": 1.0892942870859337e-05,
      "loss": 0.19,
      "step": 18970
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.0888142102736438e-05,
      "loss": 0.166,
      "step": 18980
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.0883341334613539e-05,
      "loss": 0.3331,
      "step": 18990
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.0878540566490638e-05,
      "loss": 0.0942,
      "step": 19000
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.0873739798367739e-05,
      "loss": 0.2903,
      "step": 19010
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.0868939030244841e-05,
      "loss": 0.1409,
      "step": 19020
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.0864138262121939e-05,
      "loss": 0.0761,
      "step": 19030
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.085933749399904e-05,
      "loss": 0.1623,
      "step": 19040
    },
    {
      "epoch": 4.57,
      "learning_rate": 1.0854536725876142e-05,
      "loss": 0.1639,
      "step": 19050
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.0849735957753243e-05,
      "loss": 0.1958,
      "step": 19060
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.084493518963034e-05,
      "loss": 0.0872,
      "step": 19070
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.0840134421507443e-05,
      "loss": 0.2083,
      "step": 19080
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.0835333653384544e-05,
      "loss": 0.1247,
      "step": 19090
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.0830532885261641e-05,
      "loss": 0.0674,
      "step": 19100
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.0825732117138744e-05,
      "loss": 0.0053,
      "step": 19110
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.0820931349015845e-05,
      "loss": 0.1899,
      "step": 19120
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.0816130580892942e-05,
      "loss": 0.1545,
      "step": 19130
    },
    {
      "epoch": 4.59,
      "learning_rate": 1.0811329812770045e-05,
      "loss": 0.1403,
      "step": 19140
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.0806529044647146e-05,
      "loss": 0.2222,
      "step": 19150
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.0801728276524243e-05,
      "loss": 0.3289,
      "step": 19160
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.0796927508401346e-05,
      "loss": 0.1257,
      "step": 19170
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.0792126740278446e-05,
      "loss": 0.0612,
      "step": 19180
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.0787325972155547e-05,
      "loss": 0.2134,
      "step": 19190
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.0782525204032646e-05,
      "loss": 0.1621,
      "step": 19200
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.0777724435909747e-05,
      "loss": 0.2597,
      "step": 19210
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.0772923667786848e-05,
      "loss": 0.1157,
      "step": 19220
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.0768122899663947e-05,
      "loss": 0.137,
      "step": 19230
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.0763322131541048e-05,
      "loss": 0.0738,
      "step": 19240
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.0758521363418149e-05,
      "loss": 0.2909,
      "step": 19250
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.0753720595295248e-05,
      "loss": 0.0436,
      "step": 19260
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.0748919827172349e-05,
      "loss": 0.1493,
      "step": 19270
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.074411905904945e-05,
      "loss": 0.279,
      "step": 19280
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.0739318290926549e-05,
      "loss": 0.1393,
      "step": 19290
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.073451752280365e-05,
      "loss": 0.2138,
      "step": 19300
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.072971675468075e-05,
      "loss": 0.0686,
      "step": 19310
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.0724915986557851e-05,
      "loss": 0.216,
      "step": 19320
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.072011521843495e-05,
      "loss": 0.11,
      "step": 19330
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.0715314450312051e-05,
      "loss": 0.2092,
      "step": 19340
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.0710513682189152e-05,
      "loss": 0.116,
      "step": 19350
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.0705712914066251e-05,
      "loss": 0.2469,
      "step": 19360
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.0700912145943352e-05,
      "loss": 0.0953,
      "step": 19370
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.0696111377820453e-05,
      "loss": 0.0989,
      "step": 19380
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.0691310609697552e-05,
      "loss": 0.122,
      "step": 19390
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.0686509841574653e-05,
      "loss": 0.2108,
      "step": 19400
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.0681709073451754e-05,
      "loss": 0.161,
      "step": 19410
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.0676908305328853e-05,
      "loss": 0.1641,
      "step": 19420
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.0672107537205954e-05,
      "loss": 0.263,
      "step": 19430
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.0667306769083055e-05,
      "loss": 0.0792,
      "step": 19440
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.0662506000960156e-05,
      "loss": 0.0966,
      "step": 19450
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.0657705232837255e-05,
      "loss": 0.1117,
      "step": 19460
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.0652904464714356e-05,
      "loss": 0.1938,
      "step": 19470
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.0648103696591456e-05,
      "loss": 0.2601,
      "step": 19480
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.0643302928468556e-05,
      "loss": 0.108,
      "step": 19490
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.0638502160345656e-05,
      "loss": 0.2481,
      "step": 19500
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.0633701392222757e-05,
      "loss": 0.2096,
      "step": 19510
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.0628900624099856e-05,
      "loss": 0.2282,
      "step": 19520
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.0624099855976957e-05,
      "loss": 0.1648,
      "step": 19530
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.0619299087854058e-05,
      "loss": 0.1631,
      "step": 19540
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.0614498319731157e-05,
      "loss": 0.21,
      "step": 19550
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.0609697551608258e-05,
      "loss": 0.1879,
      "step": 19560
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.0604896783485359e-05,
      "loss": 0.1041,
      "step": 19570
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.060009601536246e-05,
      "loss": 0.2326,
      "step": 19580
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.0595295247239559e-05,
      "loss": 0.1513,
      "step": 19590
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.059049447911666e-05,
      "loss": 0.1195,
      "step": 19600
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.058569371099376e-05,
      "loss": 0.096,
      "step": 19610
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.058089294287086e-05,
      "loss": 0.2725,
      "step": 19620
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.057609217474796e-05,
      "loss": 0.0975,
      "step": 19630
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.0571291406625061e-05,
      "loss": 0.113,
      "step": 19640
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.056649063850216e-05,
      "loss": 0.062,
      "step": 19650
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.0561689870379261e-05,
      "loss": 0.2259,
      "step": 19660
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.0556889102256362e-05,
      "loss": 0.0499,
      "step": 19670
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.0552088334133461e-05,
      "loss": 0.076,
      "step": 19680
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.0547287566010562e-05,
      "loss": 0.2014,
      "step": 19690
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.0542486797887663e-05,
      "loss": 0.2395,
      "step": 19700
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.0537686029764764e-05,
      "loss": 0.0829,
      "step": 19710
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.0532885261641863e-05,
      "loss": 0.1266,
      "step": 19720
    },
    {
      "epoch": 4.74,
      "learning_rate": 1.0528084493518964e-05,
      "loss": 0.1522,
      "step": 19730
    },
    {
      "epoch": 4.74,
      "learning_rate": 1.0523283725396065e-05,
      "loss": 0.0219,
      "step": 19740
    },
    {
      "epoch": 4.74,
      "learning_rate": 1.0518482957273164e-05,
      "loss": 0.1634,
      "step": 19750
    },
    {
      "epoch": 4.74,
      "learning_rate": 1.0513682189150265e-05,
      "loss": 0.0393,
      "step": 19760
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.0508881421027366e-05,
      "loss": 0.175,
      "step": 19770
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.0504080652904465e-05,
      "loss": 0.0398,
      "step": 19780
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.0499279884781566e-05,
      "loss": 0.2301,
      "step": 19790
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.0494479116658666e-05,
      "loss": 0.2285,
      "step": 19800
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.0489678348535766e-05,
      "loss": 0.1778,
      "step": 19810
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.0484877580412866e-05,
      "loss": 0.0901,
      "step": 19820
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.0480076812289967e-05,
      "loss": 0.2282,
      "step": 19830
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.0475276044167068e-05,
      "loss": 0.2151,
      "step": 19840
    },
    {
      "epoch": 4.76,
      "learning_rate": 1.0470475276044167e-05,
      "loss": 0.1444,
      "step": 19850
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.0465674507921268e-05,
      "loss": 0.1353,
      "step": 19860
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.0460873739798369e-05,
      "loss": 0.0451,
      "step": 19870
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.0456072971675468e-05,
      "loss": 0.0515,
      "step": 19880
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.0451272203552569e-05,
      "loss": 0.1176,
      "step": 19890
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.044647143542967e-05,
      "loss": 0.0929,
      "step": 19900
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.0441670667306769e-05,
      "loss": 0.3223,
      "step": 19910
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.043686989918387e-05,
      "loss": 0.1727,
      "step": 19920
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.043206913106097e-05,
      "loss": 0.0451,
      "step": 19930
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.042726836293807e-05,
      "loss": 0.2133,
      "step": 19940
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.042246759481517e-05,
      "loss": 0.1312,
      "step": 19950
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.0417666826692272e-05,
      "loss": 0.1612,
      "step": 19960
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.0412866058569372e-05,
      "loss": 0.1478,
      "step": 19970
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.0408065290446472e-05,
      "loss": 0.0216,
      "step": 19980
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.0403264522323572e-05,
      "loss": 0.0332,
      "step": 19990
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.0398463754200673e-05,
      "loss": 0.0915,
      "step": 20000
    },
    {
      "epoch": 4.8,
      "learning_rate": 1.0393662986077772e-05,
      "loss": 0.1817,
      "step": 20010
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.0388862217954873e-05,
      "loss": 0.124,
      "step": 20020
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.0384061449831974e-05,
      "loss": 0.1593,
      "step": 20030
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.0379260681709073e-05,
      "loss": 0.2087,
      "step": 20040
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.0374459913586174e-05,
      "loss": 0.1367,
      "step": 20050
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.0369659145463275e-05,
      "loss": 0.0998,
      "step": 20060
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.0364858377340374e-05,
      "loss": 0.0786,
      "step": 20070
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.0360057609217475e-05,
      "loss": 0.3306,
      "step": 20080
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.0355256841094576e-05,
      "loss": 0.0529,
      "step": 20090
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.0350456072971675e-05,
      "loss": 0.2779,
      "step": 20100
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.0345655304848776e-05,
      "loss": 0.2016,
      "step": 20110
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.0340854536725877e-05,
      "loss": 0.1131,
      "step": 20120
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.0336053768602979e-05,
      "loss": 0.0348,
      "step": 20130
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.0331253000480077e-05,
      "loss": 0.1864,
      "step": 20140
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.0326452232357177e-05,
      "loss": 0.0716,
      "step": 20150
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.032165146423428e-05,
      "loss": 0.1235,
      "step": 20160
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.0316850696111377e-05,
      "loss": 0.225,
      "step": 20170
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.0312049927988478e-05,
      "loss": 0.0424,
      "step": 20180
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.030724915986558e-05,
      "loss": 0.1848,
      "step": 20190
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.0302448391742678e-05,
      "loss": 0.2428,
      "step": 20200
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.0297647623619779e-05,
      "loss": 0.2352,
      "step": 20210
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.0292846855496882e-05,
      "loss": 0.1211,
      "step": 20220
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.0288046087373979e-05,
      "loss": 0.1854,
      "step": 20230
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.028324531925108e-05,
      "loss": 0.2107,
      "step": 20240
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.0278444551128182e-05,
      "loss": 0.0815,
      "step": 20250
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.0273643783005283e-05,
      "loss": 0.1032,
      "step": 20260
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.026884301488238e-05,
      "loss": 0.0906,
      "step": 20270
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.0264042246759483e-05,
      "loss": 0.1465,
      "step": 20280
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.0259241478636584e-05,
      "loss": 0.139,
      "step": 20290
    },
    {
      "epoch": 4.87,
      "learning_rate": 1.0254440710513682e-05,
      "loss": 0.1765,
      "step": 20300
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.0249639942390784e-05,
      "loss": 0.2019,
      "step": 20310
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.0244839174267885e-05,
      "loss": 0.1373,
      "step": 20320
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.0240038406144982e-05,
      "loss": 0.1912,
      "step": 20330
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.0235237638022085e-05,
      "loss": 0.2043,
      "step": 20340
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.0230436869899186e-05,
      "loss": 0.101,
      "step": 20350
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.0225636101776283e-05,
      "loss": 0.1034,
      "step": 20360
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.0220835333653386e-05,
      "loss": 0.0576,
      "step": 20370
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.0216034565530487e-05,
      "loss": 0.08,
      "step": 20380
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.0211233797407587e-05,
      "loss": 0.1144,
      "step": 20390
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.0206433029284687e-05,
      "loss": 0.1167,
      "step": 20400
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.0201632261161787e-05,
      "loss": 0.1824,
      "step": 20410
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.0196831493038888e-05,
      "loss": 0.0094,
      "step": 20420
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.0192030724915987e-05,
      "loss": 0.2312,
      "step": 20430
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.0187229956793088e-05,
      "loss": 0.1083,
      "step": 20440
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.0182429188670189e-05,
      "loss": 0.0617,
      "step": 20450
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.0177628420547288e-05,
      "loss": 0.2505,
      "step": 20460
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.0172827652424389e-05,
      "loss": 0.2413,
      "step": 20470
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.016802688430149e-05,
      "loss": 0.0901,
      "step": 20480
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.0163226116178589e-05,
      "loss": 0.079,
      "step": 20490
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.015842534805569e-05,
      "loss": 0.1842,
      "step": 20500
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.015362457993279e-05,
      "loss": 0.1913,
      "step": 20510
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.0148823811809892e-05,
      "loss": 0.1038,
      "step": 20520
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.014402304368699e-05,
      "loss": 0.24,
      "step": 20530
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.0139222275564092e-05,
      "loss": 0.0364,
      "step": 20540
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.0134421507441192e-05,
      "loss": 0.0442,
      "step": 20550
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.0129620739318292e-05,
      "loss": 0.0907,
      "step": 20560
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.0124819971195392e-05,
      "loss": 0.1739,
      "step": 20570
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.0120019203072493e-05,
      "loss": 0.0714,
      "step": 20580
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.0115218434949592e-05,
      "loss": 0.1442,
      "step": 20590
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.0110417666826693e-05,
      "loss": 0.1367,
      "step": 20600
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.0105616898703794e-05,
      "loss": 0.0988,
      "step": 20610
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.0100816130580893e-05,
      "loss": 0.1826,
      "step": 20620
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.0096015362457994e-05,
      "loss": 0.2755,
      "step": 20630
    },
    {
      "epoch": 4.95,
      "learning_rate": 1.0091214594335095e-05,
      "loss": 0.1459,
      "step": 20640
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.0086413826212196e-05,
      "loss": 0.042,
      "step": 20650
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.0081613058089295e-05,
      "loss": 0.1153,
      "step": 20660
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.0076812289966396e-05,
      "loss": 0.0608,
      "step": 20670
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.0072011521843497e-05,
      "loss": 0.0992,
      "step": 20680
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.0067210753720596e-05,
      "loss": 0.1285,
      "step": 20690
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.0062409985597697e-05,
      "loss": 0.1267,
      "step": 20700
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.0057609217474797e-05,
      "loss": 0.1148,
      "step": 20710
    },
    {
      "epoch": 4.97,
      "learning_rate": 1.0052808449351897e-05,
      "loss": 0.214,
      "step": 20720
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.0048007681228997e-05,
      "loss": 0.1003,
      "step": 20730
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.0043206913106098e-05,
      "loss": 0.2603,
      "step": 20740
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.0038406144983197e-05,
      "loss": 0.0703,
      "step": 20750
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.0033605376860298e-05,
      "loss": 0.19,
      "step": 20760
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.0028804608737399e-05,
      "loss": 0.1491,
      "step": 20770
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.00240038406145e-05,
      "loss": 0.1619,
      "step": 20780
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.0019203072491599e-05,
      "loss": 0.1406,
      "step": 20790
    },
    {
      "epoch": 4.99,
      "learning_rate": 1.00144023043687e-05,
      "loss": 0.1669,
      "step": 20800
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.00096015362458e-05,
      "loss": 0.2989,
      "step": 20810
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.00048007681229e-05,
      "loss": 0.167,
      "step": 20820
    },
    {
      "epoch": 5.0,
      "learning_rate": 1e-05,
      "loss": 0.1277,
      "step": 20830
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.9984699081944917,
      "eval_loss": 0.007643561344593763,
      "eval_runtime": 516.7779,
      "eval_samples_per_second": 11.382,
      "eval_steps_per_second": 1.424,
      "step": 20830
    },
    {
      "epoch": 5.0,
      "learning_rate": 9.9951992318771e-06,
      "loss": 0.2282,
      "step": 20840
    },
    {
      "epoch": 5.0,
      "learning_rate": 9.990398463754203e-06,
      "loss": 0.1466,
      "step": 20850
    },
    {
      "epoch": 5.01,
      "learning_rate": 9.985597695631302e-06,
      "loss": 0.2386,
      "step": 20860
    },
    {
      "epoch": 5.01,
      "learning_rate": 9.980796927508402e-06,
      "loss": 0.1551,
      "step": 20870
    },
    {
      "epoch": 5.01,
      "learning_rate": 9.975996159385503e-06,
      "loss": 0.2354,
      "step": 20880
    },
    {
      "epoch": 5.01,
      "learning_rate": 9.971195391262602e-06,
      "loss": 0.059,
      "step": 20890
    },
    {
      "epoch": 5.02,
      "learning_rate": 9.966394623139703e-06,
      "loss": 0.1755,
      "step": 20900
    },
    {
      "epoch": 5.02,
      "learning_rate": 9.961593855016804e-06,
      "loss": 0.258,
      "step": 20910
    },
    {
      "epoch": 5.02,
      "learning_rate": 9.956793086893903e-06,
      "loss": 0.1466,
      "step": 20920
    },
    {
      "epoch": 5.02,
      "learning_rate": 9.951992318771004e-06,
      "loss": 0.1936,
      "step": 20930
    },
    {
      "epoch": 5.03,
      "learning_rate": 9.947191550648105e-06,
      "loss": 0.0295,
      "step": 20940
    },
    {
      "epoch": 5.03,
      "learning_rate": 9.942390782525206e-06,
      "loss": 0.0796,
      "step": 20950
    },
    {
      "epoch": 5.03,
      "learning_rate": 9.937590014402305e-06,
      "loss": 0.1534,
      "step": 20960
    },
    {
      "epoch": 5.03,
      "learning_rate": 9.932789246279406e-06,
      "loss": 0.4549,
      "step": 20970
    },
    {
      "epoch": 5.04,
      "learning_rate": 9.927988478156507e-06,
      "loss": 0.081,
      "step": 20980
    },
    {
      "epoch": 5.04,
      "learning_rate": 9.923187710033606e-06,
      "loss": 0.3626,
      "step": 20990
    },
    {
      "epoch": 5.04,
      "learning_rate": 9.918386941910707e-06,
      "loss": 0.2685,
      "step": 21000
    },
    {
      "epoch": 5.04,
      "learning_rate": 9.913586173787808e-06,
      "loss": 0.1561,
      "step": 21010
    },
    {
      "epoch": 5.05,
      "learning_rate": 9.908785405664907e-06,
      "loss": 0.2837,
      "step": 21020
    },
    {
      "epoch": 5.05,
      "learning_rate": 9.903984637542008e-06,
      "loss": 0.1386,
      "step": 21030
    },
    {
      "epoch": 5.05,
      "learning_rate": 9.899183869419108e-06,
      "loss": 0.1665,
      "step": 21040
    },
    {
      "epoch": 5.05,
      "learning_rate": 9.894383101296208e-06,
      "loss": 0.1262,
      "step": 21050
    },
    {
      "epoch": 5.06,
      "learning_rate": 9.889582333173308e-06,
      "loss": 0.0925,
      "step": 21060
    },
    {
      "epoch": 5.06,
      "learning_rate": 9.88478156505041e-06,
      "loss": 0.0526,
      "step": 21070
    },
    {
      "epoch": 5.06,
      "learning_rate": 9.87998079692751e-06,
      "loss": 0.1895,
      "step": 21080
    },
    {
      "epoch": 5.06,
      "learning_rate": 9.87518002880461e-06,
      "loss": 0.1075,
      "step": 21090
    },
    {
      "epoch": 5.06,
      "learning_rate": 9.87037926068171e-06,
      "loss": 0.2021,
      "step": 21100
    },
    {
      "epoch": 5.07,
      "learning_rate": 9.865578492558811e-06,
      "loss": 0.2489,
      "step": 21110
    },
    {
      "epoch": 5.07,
      "learning_rate": 9.86077772443591e-06,
      "loss": 0.1887,
      "step": 21120
    },
    {
      "epoch": 5.07,
      "learning_rate": 9.855976956313011e-06,
      "loss": 0.1019,
      "step": 21130
    },
    {
      "epoch": 5.07,
      "learning_rate": 9.851176188190112e-06,
      "loss": 0.1957,
      "step": 21140
    },
    {
      "epoch": 5.08,
      "learning_rate": 9.846375420067211e-06,
      "loss": 0.1449,
      "step": 21150
    },
    {
      "epoch": 5.08,
      "learning_rate": 9.841574651944312e-06,
      "loss": 0.1341,
      "step": 21160
    },
    {
      "epoch": 5.08,
      "learning_rate": 9.836773883821413e-06,
      "loss": 0.1704,
      "step": 21170
    },
    {
      "epoch": 5.08,
      "learning_rate": 9.831973115698512e-06,
      "loss": 0.1499,
      "step": 21180
    },
    {
      "epoch": 5.09,
      "learning_rate": 9.827172347575613e-06,
      "loss": 0.1429,
      "step": 21190
    },
    {
      "epoch": 5.09,
      "learning_rate": 9.822371579452713e-06,
      "loss": 0.032,
      "step": 21200
    },
    {
      "epoch": 5.09,
      "learning_rate": 9.817570811329814e-06,
      "loss": 0.1481,
      "step": 21210
    },
    {
      "epoch": 5.09,
      "learning_rate": 9.812770043206913e-06,
      "loss": 0.0606,
      "step": 21220
    },
    {
      "epoch": 5.1,
      "learning_rate": 9.807969275084014e-06,
      "loss": 0.045,
      "step": 21230
    },
    {
      "epoch": 5.1,
      "learning_rate": 9.803168506961115e-06,
      "loss": 0.2612,
      "step": 21240
    },
    {
      "epoch": 5.1,
      "learning_rate": 9.798367738838214e-06,
      "loss": 0.2214,
      "step": 21250
    },
    {
      "epoch": 5.1,
      "learning_rate": 9.793566970715315e-06,
      "loss": 0.1381,
      "step": 21260
    },
    {
      "epoch": 5.11,
      "learning_rate": 9.788766202592416e-06,
      "loss": 0.1585,
      "step": 21270
    },
    {
      "epoch": 5.11,
      "learning_rate": 9.783965434469515e-06,
      "loss": 0.0588,
      "step": 21280
    },
    {
      "epoch": 5.11,
      "learning_rate": 9.779164666346616e-06,
      "loss": 0.231,
      "step": 21290
    },
    {
      "epoch": 5.11,
      "learning_rate": 9.774363898223717e-06,
      "loss": 0.196,
      "step": 21300
    },
    {
      "epoch": 5.12,
      "learning_rate": 9.769563130100816e-06,
      "loss": 0.1165,
      "step": 21310
    },
    {
      "epoch": 5.12,
      "learning_rate": 9.764762361977917e-06,
      "loss": 0.0786,
      "step": 21320
    },
    {
      "epoch": 5.12,
      "learning_rate": 9.759961593855018e-06,
      "loss": 0.1502,
      "step": 21330
    },
    {
      "epoch": 5.12,
      "learning_rate": 9.755160825732118e-06,
      "loss": 0.305,
      "step": 21340
    },
    {
      "epoch": 5.12,
      "learning_rate": 9.750360057609218e-06,
      "loss": 0.0459,
      "step": 21350
    },
    {
      "epoch": 5.13,
      "learning_rate": 9.745559289486318e-06,
      "loss": 0.124,
      "step": 21360
    },
    {
      "epoch": 5.13,
      "learning_rate": 9.74075852136342e-06,
      "loss": 0.086,
      "step": 21370
    },
    {
      "epoch": 5.13,
      "learning_rate": 9.735957753240518e-06,
      "loss": 0.0953,
      "step": 21380
    },
    {
      "epoch": 5.13,
      "learning_rate": 9.73115698511762e-06,
      "loss": 0.1654,
      "step": 21390
    },
    {
      "epoch": 5.14,
      "learning_rate": 9.72635621699472e-06,
      "loss": 0.1364,
      "step": 21400
    },
    {
      "epoch": 5.14,
      "learning_rate": 9.72155544887182e-06,
      "loss": 0.1196,
      "step": 21410
    },
    {
      "epoch": 5.14,
      "learning_rate": 9.716754680748922e-06,
      "loss": 0.1727,
      "step": 21420
    },
    {
      "epoch": 5.14,
      "learning_rate": 9.711953912626021e-06,
      "loss": 0.2441,
      "step": 21430
    },
    {
      "epoch": 5.15,
      "learning_rate": 9.70715314450312e-06,
      "loss": 0.0361,
      "step": 21440
    },
    {
      "epoch": 5.15,
      "learning_rate": 9.702352376380223e-06,
      "loss": 0.1526,
      "step": 21450
    },
    {
      "epoch": 5.15,
      "learning_rate": 9.697551608257322e-06,
      "loss": 0.1651,
      "step": 21460
    },
    {
      "epoch": 5.15,
      "learning_rate": 9.692750840134423e-06,
      "loss": 0.0999,
      "step": 21470
    },
    {
      "epoch": 5.16,
      "learning_rate": 9.687950072011523e-06,
      "loss": 0.0559,
      "step": 21480
    },
    {
      "epoch": 5.16,
      "learning_rate": 9.683149303888623e-06,
      "loss": 0.124,
      "step": 21490
    },
    {
      "epoch": 5.16,
      "learning_rate": 9.678348535765723e-06,
      "loss": 0.4906,
      "step": 21500
    },
    {
      "epoch": 5.16,
      "learning_rate": 9.673547767642824e-06,
      "loss": 0.108,
      "step": 21510
    },
    {
      "epoch": 5.17,
      "learning_rate": 9.668746999519923e-06,
      "loss": 0.3218,
      "step": 21520
    },
    {
      "epoch": 5.17,
      "learning_rate": 9.663946231397024e-06,
      "loss": 0.137,
      "step": 21530
    },
    {
      "epoch": 5.17,
      "learning_rate": 9.659145463274125e-06,
      "loss": 0.104,
      "step": 21540
    },
    {
      "epoch": 5.17,
      "learning_rate": 9.654344695151226e-06,
      "loss": 0.1426,
      "step": 21550
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.649543927028325e-06,
      "loss": 0.1215,
      "step": 21560
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.644743158905426e-06,
      "loss": 0.1023,
      "step": 21570
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.639942390782527e-06,
      "loss": 0.1644,
      "step": 21580
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.635141622659626e-06,
      "loss": 0.1716,
      "step": 21590
    },
    {
      "epoch": 5.18,
      "learning_rate": 9.630340854536727e-06,
      "loss": 0.1004,
      "step": 21600
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.625540086413828e-06,
      "loss": 0.1565,
      "step": 21610
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.620739318290927e-06,
      "loss": 0.0625,
      "step": 21620
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.615938550168028e-06,
      "loss": 0.0435,
      "step": 21630
    },
    {
      "epoch": 5.19,
      "learning_rate": 9.611137782045128e-06,
      "loss": 0.1232,
      "step": 21640
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.606337013922228e-06,
      "loss": 0.1154,
      "step": 21650
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.601536245799328e-06,
      "loss": 0.1544,
      "step": 21660
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.59673547767643e-06,
      "loss": 0.2692,
      "step": 21670
    },
    {
      "epoch": 5.2,
      "learning_rate": 9.59193470955353e-06,
      "loss": 0.2211,
      "step": 21680
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.58713394143063e-06,
      "loss": 0.0279,
      "step": 21690
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.58233317330773e-06,
      "loss": 0.2267,
      "step": 21700
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.577532405184831e-06,
      "loss": 0.1144,
      "step": 21710
    },
    {
      "epoch": 5.21,
      "learning_rate": 9.57273163706193e-06,
      "loss": 0.1727,
      "step": 21720
    },
    {
      "epoch": 5.22,
      "learning_rate": 9.567930868939031e-06,
      "loss": 0.2691,
      "step": 21730
    },
    {
      "epoch": 5.22,
      "learning_rate": 9.563130100816132e-06,
      "loss": 0.0892,
      "step": 21740
    },
    {
      "epoch": 5.22,
      "learning_rate": 9.558329332693231e-06,
      "loss": 0.116,
      "step": 21750
    },
    {
      "epoch": 5.22,
      "learning_rate": 9.553528564570332e-06,
      "loss": 0.2147,
      "step": 21760
    },
    {
      "epoch": 5.23,
      "learning_rate": 9.548727796447433e-06,
      "loss": 0.0164,
      "step": 21770
    },
    {
      "epoch": 5.23,
      "learning_rate": 9.543927028324532e-06,
      "loss": 0.1004,
      "step": 21780
    },
    {
      "epoch": 5.23,
      "learning_rate": 9.539126260201633e-06,
      "loss": 0.0166,
      "step": 21790
    },
    {
      "epoch": 5.23,
      "learning_rate": 9.534325492078733e-06,
      "loss": 0.1319,
      "step": 21800
    },
    {
      "epoch": 5.24,
      "learning_rate": 9.529524723955834e-06,
      "loss": 0.1209,
      "step": 21810
    },
    {
      "epoch": 5.24,
      "learning_rate": 9.524723955832933e-06,
      "loss": 0.0864,
      "step": 21820
    },
    {
      "epoch": 5.24,
      "learning_rate": 9.519923187710034e-06,
      "loss": 0.1978,
      "step": 21830
    },
    {
      "epoch": 5.24,
      "learning_rate": 9.515122419587135e-06,
      "loss": 0.0985,
      "step": 21840
    },
    {
      "epoch": 5.24,
      "learning_rate": 9.510321651464234e-06,
      "loss": 0.0542,
      "step": 21850
    },
    {
      "epoch": 5.25,
      "learning_rate": 9.505520883341335e-06,
      "loss": 0.1809,
      "step": 21860
    },
    {
      "epoch": 5.25,
      "learning_rate": 9.500720115218436e-06,
      "loss": 0.1019,
      "step": 21870
    },
    {
      "epoch": 5.25,
      "learning_rate": 9.495919347095535e-06,
      "loss": 0.1618,
      "step": 21880
    },
    {
      "epoch": 5.25,
      "learning_rate": 9.491118578972636e-06,
      "loss": 0.0627,
      "step": 21890
    },
    {
      "epoch": 5.26,
      "learning_rate": 9.486317810849737e-06,
      "loss": 0.1535,
      "step": 21900
    },
    {
      "epoch": 5.26,
      "learning_rate": 9.481517042726836e-06,
      "loss": 0.2104,
      "step": 21910
    },
    {
      "epoch": 5.26,
      "learning_rate": 9.476716274603937e-06,
      "loss": 0.0626,
      "step": 21920
    },
    {
      "epoch": 5.26,
      "learning_rate": 9.471915506481038e-06,
      "loss": 0.0491,
      "step": 21930
    },
    {
      "epoch": 5.27,
      "learning_rate": 9.467114738358139e-06,
      "loss": 0.2386,
      "step": 21940
    },
    {
      "epoch": 5.27,
      "learning_rate": 9.462313970235238e-06,
      "loss": 0.1588,
      "step": 21950
    },
    {
      "epoch": 5.27,
      "learning_rate": 9.457513202112338e-06,
      "loss": 0.3844,
      "step": 21960
    },
    {
      "epoch": 5.27,
      "learning_rate": 9.45271243398944e-06,
      "loss": 0.1755,
      "step": 21970
    },
    {
      "epoch": 5.28,
      "learning_rate": 9.447911665866538e-06,
      "loss": 0.3013,
      "step": 21980
    },
    {
      "epoch": 5.28,
      "learning_rate": 9.44311089774364e-06,
      "loss": 0.2543,
      "step": 21990
    },
    {
      "epoch": 5.28,
      "learning_rate": 9.43831012962074e-06,
      "loss": 0.0207,
      "step": 22000
    },
    {
      "epoch": 5.28,
      "learning_rate": 9.43350936149784e-06,
      "loss": 0.1017,
      "step": 22010
    },
    {
      "epoch": 5.29,
      "learning_rate": 9.428708593374942e-06,
      "loss": 0.1307,
      "step": 22020
    },
    {
      "epoch": 5.29,
      "learning_rate": 9.423907825252041e-06,
      "loss": 0.2027,
      "step": 22030
    },
    {
      "epoch": 5.29,
      "learning_rate": 9.41910705712914e-06,
      "loss": 0.112,
      "step": 22040
    },
    {
      "epoch": 5.29,
      "learning_rate": 9.414306289006243e-06,
      "loss": 0.0592,
      "step": 22050
    },
    {
      "epoch": 5.3,
      "learning_rate": 9.409505520883342e-06,
      "loss": 0.0323,
      "step": 22060
    },
    {
      "epoch": 5.3,
      "learning_rate": 9.404704752760443e-06,
      "loss": 0.0991,
      "step": 22070
    },
    {
      "epoch": 5.3,
      "learning_rate": 9.399903984637544e-06,
      "loss": 0.0943,
      "step": 22080
    },
    {
      "epoch": 5.3,
      "learning_rate": 9.395103216514643e-06,
      "loss": 0.1787,
      "step": 22090
    },
    {
      "epoch": 5.3,
      "learning_rate": 9.390302448391744e-06,
      "loss": 0.2703,
      "step": 22100
    },
    {
      "epoch": 5.31,
      "learning_rate": 9.385501680268844e-06,
      "loss": 0.1061,
      "step": 22110
    },
    {
      "epoch": 5.31,
      "learning_rate": 9.380700912145944e-06,
      "loss": 0.1904,
      "step": 22120
    },
    {
      "epoch": 5.31,
      "learning_rate": 9.375900144023044e-06,
      "loss": 0.0296,
      "step": 22130
    },
    {
      "epoch": 5.31,
      "learning_rate": 9.371099375900145e-06,
      "loss": 0.0205,
      "step": 22140
    },
    {
      "epoch": 5.32,
      "learning_rate": 9.366298607777246e-06,
      "loss": 0.091,
      "step": 22150
    },
    {
      "epoch": 5.32,
      "learning_rate": 9.361497839654345e-06,
      "loss": 0.2114,
      "step": 22160
    },
    {
      "epoch": 5.32,
      "learning_rate": 9.356697071531446e-06,
      "loss": 0.1102,
      "step": 22170
    },
    {
      "epoch": 5.32,
      "learning_rate": 9.351896303408547e-06,
      "loss": 0.1026,
      "step": 22180
    },
    {
      "epoch": 5.33,
      "learning_rate": 9.347095535285646e-06,
      "loss": 0.2057,
      "step": 22190
    },
    {
      "epoch": 5.33,
      "learning_rate": 9.342294767162747e-06,
      "loss": 0.0954,
      "step": 22200
    },
    {
      "epoch": 5.33,
      "learning_rate": 9.337493999039848e-06,
      "loss": 0.0996,
      "step": 22210
    },
    {
      "epoch": 5.33,
      "learning_rate": 9.332693230916947e-06,
      "loss": 0.2091,
      "step": 22220
    },
    {
      "epoch": 5.34,
      "learning_rate": 9.327892462794048e-06,
      "loss": 0.1729,
      "step": 22230
    },
    {
      "epoch": 5.34,
      "learning_rate": 9.323091694671149e-06,
      "loss": 0.1483,
      "step": 22240
    },
    {
      "epoch": 5.34,
      "learning_rate": 9.318290926548248e-06,
      "loss": 0.088,
      "step": 22250
    },
    {
      "epoch": 5.34,
      "learning_rate": 9.313490158425349e-06,
      "loss": 0.1449,
      "step": 22260
    },
    {
      "epoch": 5.35,
      "learning_rate": 9.30868939030245e-06,
      "loss": 0.2895,
      "step": 22270
    },
    {
      "epoch": 5.35,
      "learning_rate": 9.30388862217955e-06,
      "loss": 0.0841,
      "step": 22280
    },
    {
      "epoch": 5.35,
      "learning_rate": 9.29908785405665e-06,
      "loss": 0.1966,
      "step": 22290
    },
    {
      "epoch": 5.35,
      "learning_rate": 9.29428708593375e-06,
      "loss": 0.0854,
      "step": 22300
    },
    {
      "epoch": 5.36,
      "learning_rate": 9.289486317810851e-06,
      "loss": 0.1064,
      "step": 22310
    },
    {
      "epoch": 5.36,
      "learning_rate": 9.28468554968795e-06,
      "loss": 0.248,
      "step": 22320
    },
    {
      "epoch": 5.36,
      "learning_rate": 9.279884781565051e-06,
      "loss": 0.241,
      "step": 22330
    },
    {
      "epoch": 5.36,
      "learning_rate": 9.275084013442152e-06,
      "loss": 0.0676,
      "step": 22340
    },
    {
      "epoch": 5.36,
      "learning_rate": 9.270283245319251e-06,
      "loss": 0.2349,
      "step": 22350
    },
    {
      "epoch": 5.37,
      "learning_rate": 9.265482477196352e-06,
      "loss": 0.1589,
      "step": 22360
    },
    {
      "epoch": 5.37,
      "learning_rate": 9.260681709073453e-06,
      "loss": 0.0945,
      "step": 22370
    },
    {
      "epoch": 5.37,
      "learning_rate": 9.255880940950552e-06,
      "loss": 0.1398,
      "step": 22380
    },
    {
      "epoch": 5.37,
      "learning_rate": 9.251080172827653e-06,
      "loss": 0.1089,
      "step": 22390
    },
    {
      "epoch": 5.38,
      "learning_rate": 9.246279404704754e-06,
      "loss": 0.0284,
      "step": 22400
    },
    {
      "epoch": 5.38,
      "learning_rate": 9.241478636581854e-06,
      "loss": 0.2299,
      "step": 22410
    },
    {
      "epoch": 5.38,
      "learning_rate": 9.236677868458954e-06,
      "loss": 0.0708,
      "step": 22420
    },
    {
      "epoch": 5.38,
      "learning_rate": 9.231877100336054e-06,
      "loss": 0.1511,
      "step": 22430
    },
    {
      "epoch": 5.39,
      "learning_rate": 9.227076332213155e-06,
      "loss": 0.1563,
      "step": 22440
    },
    {
      "epoch": 5.39,
      "learning_rate": 9.222275564090254e-06,
      "loss": 0.1064,
      "step": 22450
    },
    {
      "epoch": 5.39,
      "learning_rate": 9.217474795967355e-06,
      "loss": 0.0699,
      "step": 22460
    },
    {
      "epoch": 5.39,
      "learning_rate": 9.212674027844456e-06,
      "loss": 0.1026,
      "step": 22470
    },
    {
      "epoch": 5.4,
      "learning_rate": 9.207873259721555e-06,
      "loss": 0.1185,
      "step": 22480
    },
    {
      "epoch": 5.4,
      "learning_rate": 9.203072491598656e-06,
      "loss": 0.0494,
      "step": 22490
    },
    {
      "epoch": 5.4,
      "learning_rate": 9.198271723475757e-06,
      "loss": 0.1729,
      "step": 22500
    },
    {
      "epoch": 5.4,
      "learning_rate": 9.193470955352856e-06,
      "loss": 0.0716,
      "step": 22510
    },
    {
      "epoch": 5.41,
      "learning_rate": 9.188670187229957e-06,
      "loss": 0.3206,
      "step": 22520
    },
    {
      "epoch": 5.41,
      "learning_rate": 9.183869419107058e-06,
      "loss": 0.0669,
      "step": 22530
    },
    {
      "epoch": 5.41,
      "learning_rate": 9.179068650984159e-06,
      "loss": 0.1566,
      "step": 22540
    },
    {
      "epoch": 5.41,
      "learning_rate": 9.174267882861258e-06,
      "loss": 0.2688,
      "step": 22550
    },
    {
      "epoch": 5.42,
      "learning_rate": 9.169467114738359e-06,
      "loss": 0.102,
      "step": 22560
    },
    {
      "epoch": 5.42,
      "learning_rate": 9.16466634661546e-06,
      "loss": 0.2398,
      "step": 22570
    },
    {
      "epoch": 5.42,
      "learning_rate": 9.159865578492559e-06,
      "loss": 0.0957,
      "step": 22580
    },
    {
      "epoch": 5.42,
      "learning_rate": 9.155064810369661e-06,
      "loss": 0.1979,
      "step": 22590
    },
    {
      "epoch": 5.42,
      "learning_rate": 9.15026404224676e-06,
      "loss": 0.2477,
      "step": 22600
    },
    {
      "epoch": 5.43,
      "learning_rate": 9.14546327412386e-06,
      "loss": 0.0932,
      "step": 22610
    },
    {
      "epoch": 5.43,
      "learning_rate": 9.140662506000962e-06,
      "loss": 0.117,
      "step": 22620
    },
    {
      "epoch": 5.43,
      "learning_rate": 9.135861737878061e-06,
      "loss": 0.1844,
      "step": 22630
    },
    {
      "epoch": 5.43,
      "learning_rate": 9.13106096975516e-06,
      "loss": 0.1836,
      "step": 22640
    },
    {
      "epoch": 5.44,
      "learning_rate": 9.126260201632263e-06,
      "loss": 0.0727,
      "step": 22650
    },
    {
      "epoch": 5.44,
      "learning_rate": 9.121459433509362e-06,
      "loss": 0.0694,
      "step": 22660
    },
    {
      "epoch": 5.44,
      "learning_rate": 9.116658665386463e-06,
      "loss": 0.1239,
      "step": 22670
    },
    {
      "epoch": 5.44,
      "learning_rate": 9.111857897263564e-06,
      "loss": 0.0828,
      "step": 22680
    },
    {
      "epoch": 5.45,
      "learning_rate": 9.107057129140663e-06,
      "loss": 0.1898,
      "step": 22690
    },
    {
      "epoch": 5.45,
      "learning_rate": 9.102256361017764e-06,
      "loss": 0.1236,
      "step": 22700
    },
    {
      "epoch": 5.45,
      "learning_rate": 9.097455592894864e-06,
      "loss": 0.1967,
      "step": 22710
    },
    {
      "epoch": 5.45,
      "learning_rate": 9.092654824771964e-06,
      "loss": 0.1741,
      "step": 22720
    },
    {
      "epoch": 5.46,
      "learning_rate": 9.087854056649064e-06,
      "loss": 0.1815,
      "step": 22730
    },
    {
      "epoch": 5.46,
      "learning_rate": 9.083053288526165e-06,
      "loss": 0.0979,
      "step": 22740
    },
    {
      "epoch": 5.46,
      "learning_rate": 9.078252520403266e-06,
      "loss": 0.1507,
      "step": 22750
    },
    {
      "epoch": 5.46,
      "learning_rate": 9.073451752280365e-06,
      "loss": 0.1533,
      "step": 22760
    },
    {
      "epoch": 5.47,
      "learning_rate": 9.068650984157466e-06,
      "loss": 0.2152,
      "step": 22770
    },
    {
      "epoch": 5.47,
      "learning_rate": 9.063850216034567e-06,
      "loss": 0.1521,
      "step": 22780
    },
    {
      "epoch": 5.47,
      "learning_rate": 9.059049447911666e-06,
      "loss": 0.0693,
      "step": 22790
    },
    {
      "epoch": 5.47,
      "learning_rate": 9.054248679788767e-06,
      "loss": 0.0744,
      "step": 22800
    },
    {
      "epoch": 5.48,
      "learning_rate": 9.049447911665868e-06,
      "loss": 0.0892,
      "step": 22810
    },
    {
      "epoch": 5.48,
      "learning_rate": 9.044647143542967e-06,
      "loss": 0.1605,
      "step": 22820
    },
    {
      "epoch": 5.48,
      "learning_rate": 9.039846375420068e-06,
      "loss": 0.1546,
      "step": 22830
    },
    {
      "epoch": 5.48,
      "learning_rate": 9.035045607297169e-06,
      "loss": 0.015,
      "step": 22840
    },
    {
      "epoch": 5.48,
      "learning_rate": 9.030244839174268e-06,
      "loss": 0.0518,
      "step": 22850
    },
    {
      "epoch": 5.49,
      "learning_rate": 9.025444071051369e-06,
      "loss": 0.1183,
      "step": 22860
    },
    {
      "epoch": 5.49,
      "learning_rate": 9.02064330292847e-06,
      "loss": 0.2947,
      "step": 22870
    },
    {
      "epoch": 5.49,
      "learning_rate": 9.01584253480557e-06,
      "loss": 0.1924,
      "step": 22880
    },
    {
      "epoch": 5.49,
      "learning_rate": 9.01104176668267e-06,
      "loss": 0.2254,
      "step": 22890
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.00624099855977e-06,
      "loss": 0.0984,
      "step": 22900
    },
    {
      "epoch": 5.5,
      "learning_rate": 9.001440230436871e-06,
      "loss": 0.286,
      "step": 22910
    },
    {
      "epoch": 5.5,
      "learning_rate": 8.99663946231397e-06,
      "loss": 0.2143,
      "step": 22920
    },
    {
      "epoch": 5.5,
      "learning_rate": 8.991838694191071e-06,
      "loss": 0.2313,
      "step": 22930
    },
    {
      "epoch": 5.51,
      "learning_rate": 8.987037926068172e-06,
      "loss": 0.2279,
      "step": 22940
    },
    {
      "epoch": 5.51,
      "learning_rate": 8.982237157945271e-06,
      "loss": 0.1045,
      "step": 22950
    },
    {
      "epoch": 5.51,
      "learning_rate": 8.977436389822372e-06,
      "loss": 0.279,
      "step": 22960
    },
    {
      "epoch": 5.51,
      "learning_rate": 8.972635621699473e-06,
      "loss": 0.0485,
      "step": 22970
    },
    {
      "epoch": 5.52,
      "learning_rate": 8.967834853576572e-06,
      "loss": 0.0757,
      "step": 22980
    },
    {
      "epoch": 5.52,
      "learning_rate": 8.963034085453673e-06,
      "loss": 0.1427,
      "step": 22990
    },
    {
      "epoch": 5.52,
      "learning_rate": 8.958233317330774e-06,
      "loss": 0.2291,
      "step": 23000
    },
    {
      "epoch": 5.52,
      "learning_rate": 8.953432549207875e-06,
      "loss": 0.0404,
      "step": 23010
    },
    {
      "epoch": 5.53,
      "learning_rate": 8.948631781084974e-06,
      "loss": 0.1381,
      "step": 23020
    },
    {
      "epoch": 5.53,
      "learning_rate": 8.943831012962075e-06,
      "loss": 0.1238,
      "step": 23030
    },
    {
      "epoch": 5.53,
      "learning_rate": 8.939030244839175e-06,
      "loss": 0.1505,
      "step": 23040
    },
    {
      "epoch": 5.53,
      "learning_rate": 8.934229476716274e-06,
      "loss": 0.0807,
      "step": 23050
    },
    {
      "epoch": 5.54,
      "learning_rate": 8.929428708593375e-06,
      "loss": 0.0767,
      "step": 23060
    },
    {
      "epoch": 5.54,
      "learning_rate": 8.924627940470476e-06,
      "loss": 0.186,
      "step": 23070
    },
    {
      "epoch": 5.54,
      "learning_rate": 8.919827172347575e-06,
      "loss": 0.1324,
      "step": 23080
    },
    {
      "epoch": 5.54,
      "learning_rate": 8.915026404224678e-06,
      "loss": 0.0861,
      "step": 23090
    },
    {
      "epoch": 5.54,
      "learning_rate": 8.910225636101777e-06,
      "loss": 0.1016,
      "step": 23100
    },
    {
      "epoch": 5.55,
      "learning_rate": 8.905424867978876e-06,
      "loss": 0.0707,
      "step": 23110
    },
    {
      "epoch": 5.55,
      "learning_rate": 8.900624099855979e-06,
      "loss": 0.1564,
      "step": 23120
    },
    {
      "epoch": 5.55,
      "learning_rate": 8.895823331733078e-06,
      "loss": 0.1158,
      "step": 23130
    },
    {
      "epoch": 5.55,
      "learning_rate": 8.891022563610179e-06,
      "loss": 0.3153,
      "step": 23140
    },
    {
      "epoch": 5.56,
      "learning_rate": 8.88622179548728e-06,
      "loss": 0.1727,
      "step": 23150
    },
    {
      "epoch": 5.56,
      "learning_rate": 8.881421027364379e-06,
      "loss": 0.1206,
      "step": 23160
    },
    {
      "epoch": 5.56,
      "learning_rate": 8.87662025924148e-06,
      "loss": 0.0377,
      "step": 23170
    },
    {
      "epoch": 5.56,
      "learning_rate": 8.87181949111858e-06,
      "loss": 0.2036,
      "step": 23180
    },
    {
      "epoch": 5.57,
      "learning_rate": 8.867018722995681e-06,
      "loss": 0.1361,
      "step": 23190
    },
    {
      "epoch": 5.57,
      "learning_rate": 8.86221795487278e-06,
      "loss": 0.1613,
      "step": 23200
    },
    {
      "epoch": 5.57,
      "learning_rate": 8.857417186749881e-06,
      "loss": 0.2225,
      "step": 23210
    },
    {
      "epoch": 5.57,
      "learning_rate": 8.852616418626982e-06,
      "loss": 0.2083,
      "step": 23220
    },
    {
      "epoch": 5.58,
      "learning_rate": 8.847815650504081e-06,
      "loss": 0.0807,
      "step": 23230
    },
    {
      "epoch": 5.58,
      "learning_rate": 8.843014882381182e-06,
      "loss": 0.1861,
      "step": 23240
    },
    {
      "epoch": 5.58,
      "learning_rate": 8.838214114258283e-06,
      "loss": 0.1768,
      "step": 23250
    },
    {
      "epoch": 5.58,
      "learning_rate": 8.833413346135382e-06,
      "loss": 0.1907,
      "step": 23260
    },
    {
      "epoch": 5.59,
      "learning_rate": 8.828612578012483e-06,
      "loss": 0.1334,
      "step": 23270
    },
    {
      "epoch": 5.59,
      "learning_rate": 8.823811809889584e-06,
      "loss": 0.0783,
      "step": 23280
    },
    {
      "epoch": 5.59,
      "learning_rate": 8.819011041766683e-06,
      "loss": 0.157,
      "step": 23290
    },
    {
      "epoch": 5.59,
      "learning_rate": 8.814210273643784e-06,
      "loss": 0.113,
      "step": 23300
    },
    {
      "epoch": 5.6,
      "learning_rate": 8.809409505520885e-06,
      "loss": 0.2966,
      "step": 23310
    },
    {
      "epoch": 5.6,
      "learning_rate": 8.804608737397985e-06,
      "loss": 0.1154,
      "step": 23320
    },
    {
      "epoch": 5.6,
      "learning_rate": 8.799807969275085e-06,
      "loss": 0.1907,
      "step": 23330
    },
    {
      "epoch": 5.6,
      "learning_rate": 8.795007201152185e-06,
      "loss": 0.0188,
      "step": 23340
    },
    {
      "epoch": 5.6,
      "learning_rate": 8.790206433029286e-06,
      "loss": 0.1731,
      "step": 23350
    },
    {
      "epoch": 5.61,
      "learning_rate": 8.785405664906385e-06,
      "loss": 0.1111,
      "step": 23360
    },
    {
      "epoch": 5.61,
      "learning_rate": 8.780604896783486e-06,
      "loss": 0.0115,
      "step": 23370
    },
    {
      "epoch": 5.61,
      "learning_rate": 8.775804128660587e-06,
      "loss": 0.1328,
      "step": 23380
    },
    {
      "epoch": 5.61,
      "learning_rate": 8.771003360537686e-06,
      "loss": 0.1696,
      "step": 23390
    },
    {
      "epoch": 5.62,
      "learning_rate": 8.766202592414787e-06,
      "loss": 0.2068,
      "step": 23400
    },
    {
      "epoch": 5.62,
      "learning_rate": 8.761401824291888e-06,
      "loss": 0.2997,
      "step": 23410
    },
    {
      "epoch": 5.62,
      "learning_rate": 8.756601056168987e-06,
      "loss": 0.1893,
      "step": 23420
    },
    {
      "epoch": 5.62,
      "learning_rate": 8.751800288046088e-06,
      "loss": 0.1389,
      "step": 23430
    },
    {
      "epoch": 5.63,
      "learning_rate": 8.746999519923189e-06,
      "loss": 0.1851,
      "step": 23440
    },
    {
      "epoch": 5.63,
      "learning_rate": 8.742198751800288e-06,
      "loss": 0.0748,
      "step": 23450
    },
    {
      "epoch": 5.63,
      "learning_rate": 8.737397983677389e-06,
      "loss": 0.0554,
      "step": 23460
    },
    {
      "epoch": 5.63,
      "learning_rate": 8.73259721555449e-06,
      "loss": 0.2345,
      "step": 23470
    },
    {
      "epoch": 5.64,
      "learning_rate": 8.72779644743159e-06,
      "loss": 0.2851,
      "step": 23480
    },
    {
      "epoch": 5.64,
      "learning_rate": 8.72299567930869e-06,
      "loss": 0.1438,
      "step": 23490
    },
    {
      "epoch": 5.64,
      "learning_rate": 8.71819491118579e-06,
      "loss": 0.2902,
      "step": 23500
    },
    {
      "epoch": 5.64,
      "learning_rate": 8.713394143062891e-06,
      "loss": 0.096,
      "step": 23510
    },
    {
      "epoch": 5.65,
      "learning_rate": 8.70859337493999e-06,
      "loss": 0.0564,
      "step": 23520
    },
    {
      "epoch": 5.65,
      "learning_rate": 8.703792606817091e-06,
      "loss": 0.1112,
      "step": 23530
    },
    {
      "epoch": 5.65,
      "learning_rate": 8.698991838694192e-06,
      "loss": 0.0016,
      "step": 23540
    },
    {
      "epoch": 5.65,
      "learning_rate": 8.694191070571291e-06,
      "loss": 0.0427,
      "step": 23550
    },
    {
      "epoch": 5.66,
      "learning_rate": 8.689390302448392e-06,
      "loss": 0.141,
      "step": 23560
    },
    {
      "epoch": 5.66,
      "learning_rate": 8.684589534325493e-06,
      "loss": 0.1203,
      "step": 23570
    },
    {
      "epoch": 5.66,
      "learning_rate": 8.679788766202592e-06,
      "loss": 0.0431,
      "step": 23580
    },
    {
      "epoch": 5.66,
      "learning_rate": 8.674987998079693e-06,
      "loss": 0.0898,
      "step": 23590
    },
    {
      "epoch": 5.66,
      "learning_rate": 8.670187229956794e-06,
      "loss": 0.1872,
      "step": 23600
    },
    {
      "epoch": 5.67,
      "learning_rate": 8.665386461833895e-06,
      "loss": 0.1567,
      "step": 23610
    },
    {
      "epoch": 5.67,
      "learning_rate": 8.660585693710994e-06,
      "loss": 0.0639,
      "step": 23620
    },
    {
      "epoch": 5.67,
      "learning_rate": 8.655784925588095e-06,
      "loss": 0.1339,
      "step": 23630
    },
    {
      "epoch": 5.67,
      "learning_rate": 8.650984157465195e-06,
      "loss": 0.1433,
      "step": 23640
    },
    {
      "epoch": 5.68,
      "learning_rate": 8.646183389342295e-06,
      "loss": 0.2104,
      "step": 23650
    },
    {
      "epoch": 5.68,
      "learning_rate": 8.641382621219397e-06,
      "loss": 0.0067,
      "step": 23660
    },
    {
      "epoch": 5.68,
      "learning_rate": 8.636581853096496e-06,
      "loss": 0.242,
      "step": 23670
    },
    {
      "epoch": 5.68,
      "learning_rate": 8.631781084973595e-06,
      "loss": 0.0935,
      "step": 23680
    },
    {
      "epoch": 5.69,
      "learning_rate": 8.626980316850698e-06,
      "loss": 0.305,
      "step": 23690
    },
    {
      "epoch": 5.69,
      "learning_rate": 8.622179548727797e-06,
      "loss": 0.1011,
      "step": 23700
    },
    {
      "epoch": 5.69,
      "learning_rate": 8.617378780604896e-06,
      "loss": 0.2383,
      "step": 23710
    },
    {
      "epoch": 5.69,
      "learning_rate": 8.612578012481999e-06,
      "loss": 0.1385,
      "step": 23720
    },
    {
      "epoch": 5.7,
      "learning_rate": 8.607777244359098e-06,
      "loss": 0.0412,
      "step": 23730
    },
    {
      "epoch": 5.7,
      "learning_rate": 8.602976476236199e-06,
      "loss": 0.193,
      "step": 23740
    },
    {
      "epoch": 5.7,
      "learning_rate": 8.5981757081133e-06,
      "loss": 0.1095,
      "step": 23750
    },
    {
      "epoch": 5.7,
      "learning_rate": 8.593374939990399e-06,
      "loss": 0.2309,
      "step": 23760
    },
    {
      "epoch": 5.71,
      "learning_rate": 8.5885741718675e-06,
      "loss": 0.1307,
      "step": 23770
    },
    {
      "epoch": 5.71,
      "learning_rate": 8.5837734037446e-06,
      "loss": 0.3265,
      "step": 23780
    },
    {
      "epoch": 5.71,
      "learning_rate": 8.578972635621701e-06,
      "loss": 0.1727,
      "step": 23790
    },
    {
      "epoch": 5.71,
      "learning_rate": 8.5741718674988e-06,
      "loss": 0.1567,
      "step": 23800
    },
    {
      "epoch": 5.72,
      "learning_rate": 8.569371099375901e-06,
      "loss": 0.1232,
      "step": 23810
    },
    {
      "epoch": 5.72,
      "learning_rate": 8.564570331253002e-06,
      "loss": 0.0749,
      "step": 23820
    },
    {
      "epoch": 5.72,
      "learning_rate": 8.559769563130101e-06,
      "loss": 0.1984,
      "step": 23830
    },
    {
      "epoch": 5.72,
      "learning_rate": 8.554968795007202e-06,
      "loss": 0.0778,
      "step": 23840
    },
    {
      "epoch": 5.72,
      "learning_rate": 8.550168026884303e-06,
      "loss": 0.1163,
      "step": 23850
    },
    {
      "epoch": 5.73,
      "learning_rate": 8.545367258761402e-06,
      "loss": 0.0196,
      "step": 23860
    },
    {
      "epoch": 5.73,
      "learning_rate": 8.540566490638503e-06,
      "loss": 0.2435,
      "step": 23870
    },
    {
      "epoch": 5.73,
      "learning_rate": 8.535765722515604e-06,
      "loss": 0.2268,
      "step": 23880
    },
    {
      "epoch": 5.73,
      "learning_rate": 8.530964954392703e-06,
      "loss": 0.1571,
      "step": 23890
    },
    {
      "epoch": 5.74,
      "learning_rate": 8.526164186269804e-06,
      "loss": 0.202,
      "step": 23900
    },
    {
      "epoch": 5.74,
      "learning_rate": 8.521363418146905e-06,
      "loss": 0.1178,
      "step": 23910
    },
    {
      "epoch": 5.74,
      "learning_rate": 8.516562650024005e-06,
      "loss": 0.2691,
      "step": 23920
    },
    {
      "epoch": 5.74,
      "learning_rate": 8.511761881901105e-06,
      "loss": 0.0993,
      "step": 23930
    },
    {
      "epoch": 5.75,
      "learning_rate": 8.506961113778205e-06,
      "loss": 0.1158,
      "step": 23940
    },
    {
      "epoch": 5.75,
      "learning_rate": 8.502160345655306e-06,
      "loss": 0.0977,
      "step": 23950
    },
    {
      "epoch": 5.75,
      "learning_rate": 8.497359577532405e-06,
      "loss": 0.1979,
      "step": 23960
    },
    {
      "epoch": 5.75,
      "learning_rate": 8.492558809409506e-06,
      "loss": 0.1082,
      "step": 23970
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.487758041286607e-06,
      "loss": 0.1079,
      "step": 23980
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.482957273163706e-06,
      "loss": 0.3092,
      "step": 23990
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.478156505040807e-06,
      "loss": 0.2673,
      "step": 24000
    },
    {
      "epoch": 5.76,
      "learning_rate": 8.473355736917908e-06,
      "loss": 0.075,
      "step": 24010
    },
    {
      "epoch": 5.77,
      "learning_rate": 8.468554968795007e-06,
      "loss": 0.037,
      "step": 24020
    },
    {
      "epoch": 5.77,
      "learning_rate": 8.463754200672108e-06,
      "loss": 0.0184,
      "step": 24030
    },
    {
      "epoch": 5.77,
      "learning_rate": 8.458953432549209e-06,
      "loss": 0.0767,
      "step": 24040
    },
    {
      "epoch": 5.77,
      "learning_rate": 8.45415266442631e-06,
      "loss": 0.0071,
      "step": 24050
    },
    {
      "epoch": 5.78,
      "learning_rate": 8.449351896303409e-06,
      "loss": 0.165,
      "step": 24060
    },
    {
      "epoch": 5.78,
      "learning_rate": 8.44455112818051e-06,
      "loss": 0.0991,
      "step": 24070
    },
    {
      "epoch": 5.78,
      "learning_rate": 8.43975036005761e-06,
      "loss": 0.1252,
      "step": 24080
    },
    {
      "epoch": 5.78,
      "learning_rate": 8.43494959193471e-06,
      "loss": 0.0874,
      "step": 24090
    },
    {
      "epoch": 5.78,
      "learning_rate": 8.43014882381181e-06,
      "loss": 0.1977,
      "step": 24100
    },
    {
      "epoch": 5.79,
      "learning_rate": 8.425348055688911e-06,
      "loss": 0.0942,
      "step": 24110
    },
    {
      "epoch": 5.79,
      "learning_rate": 8.42054728756601e-06,
      "loss": 0.2025,
      "step": 24120
    },
    {
      "epoch": 5.79,
      "learning_rate": 8.415746519443111e-06,
      "loss": 0.1034,
      "step": 24130
    },
    {
      "epoch": 5.79,
      "learning_rate": 8.410945751320212e-06,
      "loss": 0.1661,
      "step": 24140
    },
    {
      "epoch": 5.8,
      "learning_rate": 8.406144983197311e-06,
      "loss": 0.1799,
      "step": 24150
    },
    {
      "epoch": 5.8,
      "learning_rate": 8.401344215074412e-06,
      "loss": 0.3321,
      "step": 24160
    },
    {
      "epoch": 5.8,
      "learning_rate": 8.396543446951513e-06,
      "loss": 0.0543,
      "step": 24170
    },
    {
      "epoch": 5.8,
      "learning_rate": 8.391742678828612e-06,
      "loss": 0.1071,
      "step": 24180
    },
    {
      "epoch": 5.81,
      "learning_rate": 8.386941910705713e-06,
      "loss": 0.038,
      "step": 24190
    },
    {
      "epoch": 5.81,
      "learning_rate": 8.382141142582814e-06,
      "loss": 0.3388,
      "step": 24200
    },
    {
      "epoch": 5.81,
      "learning_rate": 8.377340374459915e-06,
      "loss": 0.1123,
      "step": 24210
    },
    {
      "epoch": 5.81,
      "learning_rate": 8.372539606337014e-06,
      "loss": 0.0851,
      "step": 24220
    },
    {
      "epoch": 5.82,
      "learning_rate": 8.367738838214115e-06,
      "loss": 0.2466,
      "step": 24230
    },
    {
      "epoch": 5.82,
      "learning_rate": 8.362938070091216e-06,
      "loss": 0.2098,
      "step": 24240
    },
    {
      "epoch": 5.82,
      "learning_rate": 8.358137301968315e-06,
      "loss": 0.1229,
      "step": 24250
    },
    {
      "epoch": 5.82,
      "learning_rate": 8.353336533845417e-06,
      "loss": 0.0406,
      "step": 24260
    },
    {
      "epoch": 5.83,
      "learning_rate": 8.348535765722516e-06,
      "loss": 0.1041,
      "step": 24270
    },
    {
      "epoch": 5.83,
      "learning_rate": 8.343734997599616e-06,
      "loss": 0.1262,
      "step": 24280
    },
    {
      "epoch": 5.83,
      "learning_rate": 8.338934229476718e-06,
      "loss": 0.1693,
      "step": 24290
    },
    {
      "epoch": 5.83,
      "learning_rate": 8.334133461353817e-06,
      "loss": 0.1611,
      "step": 24300
    },
    {
      "epoch": 5.84,
      "learning_rate": 8.329332693230916e-06,
      "loss": 0.0699,
      "step": 24310
    },
    {
      "epoch": 5.84,
      "learning_rate": 8.324531925108019e-06,
      "loss": 0.209,
      "step": 24320
    },
    {
      "epoch": 5.84,
      "learning_rate": 8.319731156985118e-06,
      "loss": 0.206,
      "step": 24330
    },
    {
      "epoch": 5.84,
      "learning_rate": 8.314930388862219e-06,
      "loss": 0.0412,
      "step": 24340
    },
    {
      "epoch": 5.84,
      "learning_rate": 8.31012962073932e-06,
      "loss": 0.1087,
      "step": 24350
    },
    {
      "epoch": 5.85,
      "learning_rate": 8.305328852616419e-06,
      "loss": 0.2261,
      "step": 24360
    },
    {
      "epoch": 5.85,
      "learning_rate": 8.30052808449352e-06,
      "loss": 0.1058,
      "step": 24370
    },
    {
      "epoch": 5.85,
      "learning_rate": 8.29572731637062e-06,
      "loss": 0.1548,
      "step": 24380
    },
    {
      "epoch": 5.85,
      "learning_rate": 8.290926548247721e-06,
      "loss": 0.1344,
      "step": 24390
    },
    {
      "epoch": 5.86,
      "learning_rate": 8.28612578012482e-06,
      "loss": 0.0995,
      "step": 24400
    },
    {
      "epoch": 5.86,
      "learning_rate": 8.281325012001921e-06,
      "loss": 0.0742,
      "step": 24410
    },
    {
      "epoch": 5.86,
      "learning_rate": 8.276524243879022e-06,
      "loss": 0.2176,
      "step": 24420
    },
    {
      "epoch": 5.86,
      "learning_rate": 8.271723475756121e-06,
      "loss": 0.0195,
      "step": 24430
    },
    {
      "epoch": 5.87,
      "learning_rate": 8.266922707633222e-06,
      "loss": 0.2776,
      "step": 24440
    },
    {
      "epoch": 5.87,
      "learning_rate": 8.262121939510323e-06,
      "loss": 0.0599,
      "step": 24450
    },
    {
      "epoch": 5.87,
      "learning_rate": 8.257321171387422e-06,
      "loss": 0.2414,
      "step": 24460
    },
    {
      "epoch": 5.87,
      "learning_rate": 8.252520403264523e-06,
      "loss": 0.1183,
      "step": 24470
    },
    {
      "epoch": 5.88,
      "learning_rate": 8.247719635141624e-06,
      "loss": 0.2741,
      "step": 24480
    },
    {
      "epoch": 5.88,
      "learning_rate": 8.242918867018723e-06,
      "loss": 0.1529,
      "step": 24490
    },
    {
      "epoch": 5.88,
      "learning_rate": 8.238118098895824e-06,
      "loss": 0.3196,
      "step": 24500
    },
    {
      "epoch": 5.88,
      "learning_rate": 8.233317330772925e-06,
      "loss": 0.2431,
      "step": 24510
    },
    {
      "epoch": 5.89,
      "learning_rate": 8.228516562650026e-06,
      "loss": 0.3491,
      "step": 24520
    },
    {
      "epoch": 5.89,
      "learning_rate": 8.223715794527125e-06,
      "loss": 0.0612,
      "step": 24530
    },
    {
      "epoch": 5.89,
      "learning_rate": 8.218915026404226e-06,
      "loss": 0.171,
      "step": 24540
    },
    {
      "epoch": 5.89,
      "learning_rate": 8.214114258281326e-06,
      "loss": 0.4149,
      "step": 24550
    },
    {
      "epoch": 5.9,
      "learning_rate": 8.209313490158426e-06,
      "loss": 0.1195,
      "step": 24560
    },
    {
      "epoch": 5.9,
      "learning_rate": 8.204512722035526e-06,
      "loss": 0.0475,
      "step": 24570
    },
    {
      "epoch": 5.9,
      "learning_rate": 8.199711953912627e-06,
      "loss": 0.2125,
      "step": 24580
    },
    {
      "epoch": 5.9,
      "learning_rate": 8.194911185789726e-06,
      "loss": 0.1063,
      "step": 24590
    },
    {
      "epoch": 5.9,
      "learning_rate": 8.190110417666827e-06,
      "loss": 0.2058,
      "step": 24600
    },
    {
      "epoch": 5.91,
      "learning_rate": 8.185309649543928e-06,
      "loss": 0.1106,
      "step": 24610
    },
    {
      "epoch": 5.91,
      "learning_rate": 8.180508881421027e-06,
      "loss": 0.0819,
      "step": 24620
    },
    {
      "epoch": 5.91,
      "learning_rate": 8.175708113298128e-06,
      "loss": 0.1393,
      "step": 24630
    },
    {
      "epoch": 5.91,
      "learning_rate": 8.170907345175229e-06,
      "loss": 0.2128,
      "step": 24640
    },
    {
      "epoch": 5.92,
      "learning_rate": 8.16610657705233e-06,
      "loss": 0.1396,
      "step": 24650
    },
    {
      "epoch": 5.92,
      "learning_rate": 8.161305808929429e-06,
      "loss": 0.0379,
      "step": 24660
    },
    {
      "epoch": 5.92,
      "learning_rate": 8.15650504080653e-06,
      "loss": 0.3127,
      "step": 24670
    },
    {
      "epoch": 5.92,
      "learning_rate": 8.15170427268363e-06,
      "loss": 0.054,
      "step": 24680
    },
    {
      "epoch": 5.93,
      "learning_rate": 8.14690350456073e-06,
      "loss": 0.143,
      "step": 24690
    },
    {
      "epoch": 5.93,
      "learning_rate": 8.14210273643783e-06,
      "loss": 0.246,
      "step": 24700
    },
    {
      "epoch": 5.93,
      "learning_rate": 8.137301968314931e-06,
      "loss": 0.174,
      "step": 24710
    },
    {
      "epoch": 5.93,
      "learning_rate": 8.13250120019203e-06,
      "loss": 0.146,
      "step": 24720
    },
    {
      "epoch": 5.94,
      "learning_rate": 8.127700432069131e-06,
      "loss": 0.1204,
      "step": 24730
    },
    {
      "epoch": 5.94,
      "learning_rate": 8.122899663946232e-06,
      "loss": 0.085,
      "step": 24740
    },
    {
      "epoch": 5.94,
      "learning_rate": 8.118098895823331e-06,
      "loss": 0.0923,
      "step": 24750
    },
    {
      "epoch": 5.94,
      "learning_rate": 8.113298127700432e-06,
      "loss": 0.1449,
      "step": 24760
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.108497359577533e-06,
      "loss": 0.0448,
      "step": 24770
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.103696591454634e-06,
      "loss": 0.1176,
      "step": 24780
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.098895823331733e-06,
      "loss": 0.15,
      "step": 24790
    },
    {
      "epoch": 5.95,
      "learning_rate": 8.094095055208834e-06,
      "loss": 0.0147,
      "step": 24800
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.089294287085935e-06,
      "loss": 0.1931,
      "step": 24810
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.084493518963034e-06,
      "loss": 0.2148,
      "step": 24820
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.079692750840135e-06,
      "loss": 0.1358,
      "step": 24830
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.074891982717236e-06,
      "loss": 0.3344,
      "step": 24840
    },
    {
      "epoch": 5.96,
      "learning_rate": 8.070091214594335e-06,
      "loss": 0.2951,
      "step": 24850
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.065290446471437e-06,
      "loss": 0.0591,
      "step": 24860
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.060489678348536e-06,
      "loss": 0.088,
      "step": 24870
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.055688910225636e-06,
      "loss": 0.05,
      "step": 24880
    },
    {
      "epoch": 5.97,
      "learning_rate": 8.050888142102738e-06,
      "loss": 0.1937,
      "step": 24890
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.046087373979837e-06,
      "loss": 0.2958,
      "step": 24900
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.041286605856938e-06,
      "loss": 0.1362,
      "step": 24910
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.036485837734039e-06,
      "loss": 0.1428,
      "step": 24920
    },
    {
      "epoch": 5.98,
      "learning_rate": 8.031685069611138e-06,
      "loss": 0.1029,
      "step": 24930
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.026884301488239e-06,
      "loss": 0.1845,
      "step": 24940
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.02208353336534e-06,
      "loss": 0.2016,
      "step": 24950
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.017282765242439e-06,
      "loss": 0.3287,
      "step": 24960
    },
    {
      "epoch": 5.99,
      "learning_rate": 8.01248199711954e-06,
      "loss": 0.2603,
      "step": 24970
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.00768122899664e-06,
      "loss": 0.1813,
      "step": 24980
    },
    {
      "epoch": 6.0,
      "learning_rate": 8.002880460873742e-06,
      "loss": 0.0816,
      "step": 24990
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.9988099285957157,
      "eval_loss": 0.004691378679126501,
      "eval_runtime": 511.5577,
      "eval_samples_per_second": 11.498,
      "eval_steps_per_second": 1.439,
      "step": 24996
    },
    {
      "epoch": 6.0,
      "learning_rate": 7.99807969275084e-06,
      "loss": 0.1807,
      "step": 25000
    },
    {
      "epoch": 6.0,
      "learning_rate": 7.993278924627941e-06,
      "loss": 0.2785,
      "step": 25010
    },
    {
      "epoch": 6.01,
      "learning_rate": 7.988478156505042e-06,
      "loss": 0.0749,
      "step": 25020
    },
    {
      "epoch": 6.01,
      "learning_rate": 7.983677388382141e-06,
      "loss": 0.1342,
      "step": 25030
    },
    {
      "epoch": 6.01,
      "learning_rate": 7.978876620259242e-06,
      "loss": 0.0924,
      "step": 25040
    },
    {
      "epoch": 6.01,
      "learning_rate": 7.974075852136343e-06,
      "loss": 0.2424,
      "step": 25050
    },
    {
      "epoch": 6.02,
      "learning_rate": 7.969275084013442e-06,
      "loss": 0.0164,
      "step": 25060
    },
    {
      "epoch": 6.02,
      "learning_rate": 7.964474315890543e-06,
      "loss": 0.2166,
      "step": 25070
    },
    {
      "epoch": 6.02,
      "learning_rate": 7.959673547767644e-06,
      "loss": 0.2773,
      "step": 25080
    },
    {
      "epoch": 6.02,
      "learning_rate": 7.954872779644743e-06,
      "loss": 0.1265,
      "step": 25090
    },
    {
      "epoch": 6.02,
      "learning_rate": 7.950072011521844e-06,
      "loss": 0.1849,
      "step": 25100
    },
    {
      "epoch": 6.03,
      "learning_rate": 7.945271243398945e-06,
      "loss": 0.0658,
      "step": 25110
    },
    {
      "epoch": 6.03,
      "learning_rate": 7.940470475276046e-06,
      "loss": 0.0178,
      "step": 25120
    },
    {
      "epoch": 6.03,
      "learning_rate": 7.935669707153145e-06,
      "loss": 0.0527,
      "step": 25130
    },
    {
      "epoch": 6.03,
      "learning_rate": 7.930868939030246e-06,
      "loss": 0.112,
      "step": 25140
    },
    {
      "epoch": 6.04,
      "learning_rate": 7.926068170907347e-06,
      "loss": 0.1347,
      "step": 25150
    },
    {
      "epoch": 6.04,
      "learning_rate": 7.921267402784446e-06,
      "loss": 0.0655,
      "step": 25160
    },
    {
      "epoch": 6.04,
      "learning_rate": 7.916466634661547e-06,
      "loss": 0.2298,
      "step": 25170
    },
    {
      "epoch": 6.04,
      "learning_rate": 7.911665866538647e-06,
      "loss": 0.2421,
      "step": 25180
    },
    {
      "epoch": 6.05,
      "learning_rate": 7.906865098415747e-06,
      "loss": 0.0634,
      "step": 25190
    },
    {
      "epoch": 6.05,
      "learning_rate": 7.902064330292847e-06,
      "loss": 0.3296,
      "step": 25200
    },
    {
      "epoch": 6.05,
      "learning_rate": 7.897263562169948e-06,
      "loss": 0.0632,
      "step": 25210
    },
    {
      "epoch": 6.05,
      "learning_rate": 7.892462794047047e-06,
      "loss": 0.2612,
      "step": 25220
    },
    {
      "epoch": 6.06,
      "learning_rate": 7.887662025924148e-06,
      "loss": 0.2561,
      "step": 25230
    },
    {
      "epoch": 6.06,
      "learning_rate": 7.882861257801249e-06,
      "loss": 0.0797,
      "step": 25240
    },
    {
      "epoch": 6.06,
      "learning_rate": 7.87806048967835e-06,
      "loss": 0.0996,
      "step": 25250
    },
    {
      "epoch": 6.06,
      "learning_rate": 7.873259721555449e-06,
      "loss": 0.2585,
      "step": 25260
    },
    {
      "epoch": 6.07,
      "learning_rate": 7.86845895343255e-06,
      "loss": 0.0933,
      "step": 25270
    },
    {
      "epoch": 6.07,
      "learning_rate": 7.86365818530965e-06,
      "loss": 0.15,
      "step": 25280
    },
    {
      "epoch": 6.07,
      "learning_rate": 7.85885741718675e-06,
      "loss": 0.0967,
      "step": 25290
    },
    {
      "epoch": 6.07,
      "learning_rate": 7.85405664906385e-06,
      "loss": 0.1467,
      "step": 25300
    },
    {
      "epoch": 6.08,
      "learning_rate": 7.849255880940952e-06,
      "loss": 0.1004,
      "step": 25310
    },
    {
      "epoch": 6.08,
      "learning_rate": 7.84445511281805e-06,
      "loss": 0.0888,
      "step": 25320
    },
    {
      "epoch": 6.08,
      "learning_rate": 7.839654344695152e-06,
      "loss": 0.3506,
      "step": 25330
    },
    {
      "epoch": 6.08,
      "learning_rate": 7.834853576572252e-06,
      "loss": 0.0876,
      "step": 25340
    },
    {
      "epoch": 6.08,
      "learning_rate": 7.830052808449352e-06,
      "loss": 0.0225,
      "step": 25350
    },
    {
      "epoch": 6.09,
      "learning_rate": 7.825252040326452e-06,
      "loss": 0.3848,
      "step": 25360
    },
    {
      "epoch": 6.09,
      "learning_rate": 7.820451272203553e-06,
      "loss": 0.0576,
      "step": 25370
    },
    {
      "epoch": 6.09,
      "learning_rate": 7.815650504080654e-06,
      "loss": 0.1384,
      "step": 25380
    },
    {
      "epoch": 6.09,
      "learning_rate": 7.810849735957753e-06,
      "loss": 0.0694,
      "step": 25390
    },
    {
      "epoch": 6.1,
      "learning_rate": 7.806048967834854e-06,
      "loss": 0.1307,
      "step": 25400
    },
    {
      "epoch": 6.1,
      "learning_rate": 7.801248199711955e-06,
      "loss": 0.1658,
      "step": 25410
    },
    {
      "epoch": 6.1,
      "learning_rate": 7.796447431589054e-06,
      "loss": 0.1527,
      "step": 25420
    },
    {
      "epoch": 6.1,
      "learning_rate": 7.791646663466155e-06,
      "loss": 0.0733,
      "step": 25430
    },
    {
      "epoch": 6.11,
      "learning_rate": 7.786845895343256e-06,
      "loss": 0.1692,
      "step": 25440
    },
    {
      "epoch": 6.11,
      "learning_rate": 7.782045127220355e-06,
      "loss": 0.0852,
      "step": 25450
    },
    {
      "epoch": 6.11,
      "learning_rate": 7.777244359097457e-06,
      "loss": 0.2594,
      "step": 25460
    },
    {
      "epoch": 6.11,
      "learning_rate": 7.772443590974557e-06,
      "loss": 0.0741,
      "step": 25470
    },
    {
      "epoch": 6.12,
      "learning_rate": 7.767642822851656e-06,
      "loss": 0.0687,
      "step": 25480
    },
    {
      "epoch": 6.12,
      "learning_rate": 7.762842054728758e-06,
      "loss": 0.17,
      "step": 25490
    },
    {
      "epoch": 6.12,
      "learning_rate": 7.758041286605857e-06,
      "loss": 0.0603,
      "step": 25500
    },
    {
      "epoch": 6.12,
      "learning_rate": 7.753240518482958e-06,
      "loss": 0.0806,
      "step": 25510
    },
    {
      "epoch": 6.13,
      "learning_rate": 7.748439750360059e-06,
      "loss": 0.1253,
      "step": 25520
    },
    {
      "epoch": 6.13,
      "learning_rate": 7.743638982237158e-06,
      "loss": 0.0558,
      "step": 25530
    },
    {
      "epoch": 6.13,
      "learning_rate": 7.738838214114259e-06,
      "loss": 0.1631,
      "step": 25540
    },
    {
      "epoch": 6.13,
      "learning_rate": 7.73403744599136e-06,
      "loss": 0.0188,
      "step": 25550
    },
    {
      "epoch": 6.14,
      "learning_rate": 7.729236677868459e-06,
      "loss": 0.0871,
      "step": 25560
    },
    {
      "epoch": 6.14,
      "learning_rate": 7.72443590974556e-06,
      "loss": 0.1564,
      "step": 25570
    },
    {
      "epoch": 6.14,
      "learning_rate": 7.71963514162266e-06,
      "loss": 0.1638,
      "step": 25580
    },
    {
      "epoch": 6.14,
      "learning_rate": 7.714834373499762e-06,
      "loss": 0.1099,
      "step": 25590
    },
    {
      "epoch": 6.14,
      "learning_rate": 7.71003360537686e-06,
      "loss": 0.116,
      "step": 25600
    },
    {
      "epoch": 6.15,
      "learning_rate": 7.705232837253962e-06,
      "loss": 0.107,
      "step": 25610
    },
    {
      "epoch": 6.15,
      "learning_rate": 7.700432069131062e-06,
      "loss": 0.1544,
      "step": 25620
    },
    {
      "epoch": 6.15,
      "learning_rate": 7.695631301008162e-06,
      "loss": 0.1063,
      "step": 25630
    },
    {
      "epoch": 6.15,
      "learning_rate": 7.690830532885262e-06,
      "loss": 0.1224,
      "step": 25640
    },
    {
      "epoch": 6.16,
      "learning_rate": 7.686029764762363e-06,
      "loss": 0.0394,
      "step": 25650
    },
    {
      "epoch": 6.16,
      "learning_rate": 7.681228996639462e-06,
      "loss": 0.2104,
      "step": 25660
    },
    {
      "epoch": 6.16,
      "learning_rate": 7.676428228516563e-06,
      "loss": 0.1375,
      "step": 25670
    },
    {
      "epoch": 6.16,
      "learning_rate": 7.671627460393664e-06,
      "loss": 0.0978,
      "step": 25680
    },
    {
      "epoch": 6.17,
      "learning_rate": 7.666826692270763e-06,
      "loss": 0.1603,
      "step": 25690
    },
    {
      "epoch": 6.17,
      "learning_rate": 7.662025924147864e-06,
      "loss": 0.084,
      "step": 25700
    },
    {
      "epoch": 6.17,
      "learning_rate": 7.657225156024965e-06,
      "loss": 0.161,
      "step": 25710
    },
    {
      "epoch": 6.17,
      "learning_rate": 7.652424387902066e-06,
      "loss": 0.2226,
      "step": 25720
    },
    {
      "epoch": 6.18,
      "learning_rate": 7.647623619779165e-06,
      "loss": 0.1334,
      "step": 25730
    },
    {
      "epoch": 6.18,
      "learning_rate": 7.642822851656266e-06,
      "loss": 0.1843,
      "step": 25740
    },
    {
      "epoch": 6.18,
      "learning_rate": 7.638022083533367e-06,
      "loss": 0.2848,
      "step": 25750
    },
    {
      "epoch": 6.18,
      "learning_rate": 7.633221315410466e-06,
      "loss": 0.2861,
      "step": 25760
    },
    {
      "epoch": 6.19,
      "learning_rate": 7.628420547287566e-06,
      "loss": 0.1252,
      "step": 25770
    },
    {
      "epoch": 6.19,
      "learning_rate": 7.6236197791646675e-06,
      "loss": 0.0782,
      "step": 25780
    },
    {
      "epoch": 6.19,
      "learning_rate": 7.618819011041767e-06,
      "loss": 0.0714,
      "step": 25790
    },
    {
      "epoch": 6.19,
      "learning_rate": 7.614018242918868e-06,
      "loss": 0.2284,
      "step": 25800
    },
    {
      "epoch": 6.2,
      "learning_rate": 7.609217474795968e-06,
      "loss": 0.1255,
      "step": 25810
    },
    {
      "epoch": 6.2,
      "learning_rate": 7.6044167066730674e-06,
      "loss": 0.1299,
      "step": 25820
    },
    {
      "epoch": 6.2,
      "learning_rate": 7.599615938550169e-06,
      "loss": 0.2973,
      "step": 25830
    },
    {
      "epoch": 6.2,
      "learning_rate": 7.594815170427269e-06,
      "loss": 0.2841,
      "step": 25840
    },
    {
      "epoch": 6.2,
      "learning_rate": 7.59001440230437e-06,
      "loss": 0.0801,
      "step": 25850
    },
    {
      "epoch": 6.21,
      "learning_rate": 7.58521363418147e-06,
      "loss": 0.2105,
      "step": 25860
    },
    {
      "epoch": 6.21,
      "learning_rate": 7.58041286605857e-06,
      "loss": 0.2079,
      "step": 25870
    },
    {
      "epoch": 6.21,
      "learning_rate": 7.575612097935671e-06,
      "loss": 0.0622,
      "step": 25880
    },
    {
      "epoch": 6.21,
      "learning_rate": 7.570811329812771e-06,
      "loss": 0.036,
      "step": 25890
    },
    {
      "epoch": 6.22,
      "learning_rate": 7.566010561689871e-06,
      "loss": 0.2376,
      "step": 25900
    },
    {
      "epoch": 6.22,
      "learning_rate": 7.561209793566972e-06,
      "loss": 0.235,
      "step": 25910
    },
    {
      "epoch": 6.22,
      "learning_rate": 7.556409025444072e-06,
      "loss": 0.225,
      "step": 25920
    },
    {
      "epoch": 6.22,
      "learning_rate": 7.5516082573211725e-06,
      "loss": 0.0561,
      "step": 25930
    },
    {
      "epoch": 6.23,
      "learning_rate": 7.5468074891982725e-06,
      "loss": 0.0468,
      "step": 25940
    },
    {
      "epoch": 6.23,
      "learning_rate": 7.5420067210753725e-06,
      "loss": 0.1827,
      "step": 25950
    },
    {
      "epoch": 6.23,
      "learning_rate": 7.537205952952473e-06,
      "loss": 0.0561,
      "step": 25960
    },
    {
      "epoch": 6.23,
      "learning_rate": 7.532405184829573e-06,
      "loss": 0.2067,
      "step": 25970
    },
    {
      "epoch": 6.24,
      "learning_rate": 7.527604416706674e-06,
      "loss": 0.119,
      "step": 25980
    },
    {
      "epoch": 6.24,
      "learning_rate": 7.522803648583774e-06,
      "loss": 0.1031,
      "step": 25990
    },
    {
      "epoch": 6.24,
      "learning_rate": 7.518002880460874e-06,
      "loss": 0.0797,
      "step": 26000
    },
    {
      "epoch": 6.24,
      "learning_rate": 7.513202112337975e-06,
      "loss": 0.0839,
      "step": 26010
    },
    {
      "epoch": 6.25,
      "learning_rate": 7.508401344215075e-06,
      "loss": 0.2089,
      "step": 26020
    },
    {
      "epoch": 6.25,
      "learning_rate": 7.503600576092175e-06,
      "loss": 0.1383,
      "step": 26030
    },
    {
      "epoch": 6.25,
      "learning_rate": 7.498799807969276e-06,
      "loss": 0.1204,
      "step": 26040
    },
    {
      "epoch": 6.25,
      "learning_rate": 7.493999039846376e-06,
      "loss": 0.0502,
      "step": 26050
    },
    {
      "epoch": 6.26,
      "learning_rate": 7.489198271723477e-06,
      "loss": 0.0454,
      "step": 26060
    },
    {
      "epoch": 6.26,
      "learning_rate": 7.484397503600577e-06,
      "loss": 0.1206,
      "step": 26070
    },
    {
      "epoch": 6.26,
      "learning_rate": 7.479596735477677e-06,
      "loss": 0.0891,
      "step": 26080
    },
    {
      "epoch": 6.26,
      "learning_rate": 7.4747959673547775e-06,
      "loss": 0.1036,
      "step": 26090
    },
    {
      "epoch": 6.27,
      "learning_rate": 7.4699951992318775e-06,
      "loss": 0.1232,
      "step": 26100
    },
    {
      "epoch": 6.27,
      "learning_rate": 7.465194431108978e-06,
      "loss": 0.1374,
      "step": 26110
    },
    {
      "epoch": 6.27,
      "learning_rate": 7.460393662986078e-06,
      "loss": 0.1454,
      "step": 26120
    },
    {
      "epoch": 6.27,
      "learning_rate": 7.455592894863178e-06,
      "loss": 0.1978,
      "step": 26130
    },
    {
      "epoch": 6.27,
      "learning_rate": 7.450792126740279e-06,
      "loss": 0.0285,
      "step": 26140
    },
    {
      "epoch": 6.28,
      "learning_rate": 7.445991358617379e-06,
      "loss": 0.1201,
      "step": 26150
    },
    {
      "epoch": 6.28,
      "learning_rate": 7.441190590494479e-06,
      "loss": 0.1305,
      "step": 26160
    },
    {
      "epoch": 6.28,
      "learning_rate": 7.43638982237158e-06,
      "loss": 0.1754,
      "step": 26170
    },
    {
      "epoch": 6.28,
      "learning_rate": 7.43158905424868e-06,
      "loss": 0.1374,
      "step": 26180
    },
    {
      "epoch": 6.29,
      "learning_rate": 7.426788286125781e-06,
      "loss": 0.1109,
      "step": 26190
    },
    {
      "epoch": 6.29,
      "learning_rate": 7.421987518002881e-06,
      "loss": 0.1079,
      "step": 26200
    },
    {
      "epoch": 6.29,
      "learning_rate": 7.417186749879981e-06,
      "loss": 0.0572,
      "step": 26210
    },
    {
      "epoch": 6.29,
      "learning_rate": 7.412385981757082e-06,
      "loss": 0.0358,
      "step": 26220
    },
    {
      "epoch": 6.3,
      "learning_rate": 7.407585213634182e-06,
      "loss": 0.1026,
      "step": 26230
    },
    {
      "epoch": 6.3,
      "learning_rate": 7.4027844455112825e-06,
      "loss": 0.1215,
      "step": 26240
    },
    {
      "epoch": 6.3,
      "learning_rate": 7.3979836773883825e-06,
      "loss": 0.1265,
      "step": 26250
    },
    {
      "epoch": 6.3,
      "learning_rate": 7.3931829092654825e-06,
      "loss": 0.119,
      "step": 26260
    },
    {
      "epoch": 6.31,
      "learning_rate": 7.388382141142584e-06,
      "loss": 0.0374,
      "step": 26270
    },
    {
      "epoch": 6.31,
      "learning_rate": 7.383581373019683e-06,
      "loss": 0.2972,
      "step": 26280
    },
    {
      "epoch": 6.31,
      "learning_rate": 7.378780604896783e-06,
      "loss": 0.1746,
      "step": 26290
    },
    {
      "epoch": 6.31,
      "learning_rate": 7.373979836773885e-06,
      "loss": 0.1265,
      "step": 26300
    },
    {
      "epoch": 6.32,
      "learning_rate": 7.369179068650984e-06,
      "loss": 0.1768,
      "step": 26310
    },
    {
      "epoch": 6.32,
      "learning_rate": 7.364378300528086e-06,
      "loss": 0.2109,
      "step": 26320
    },
    {
      "epoch": 6.32,
      "learning_rate": 7.359577532405186e-06,
      "loss": 0.2025,
      "step": 26330
    },
    {
      "epoch": 6.32,
      "learning_rate": 7.354776764282285e-06,
      "loss": 0.0017,
      "step": 26340
    },
    {
      "epoch": 6.33,
      "learning_rate": 7.349975996159387e-06,
      "loss": 0.17,
      "step": 26350
    },
    {
      "epoch": 6.33,
      "learning_rate": 7.345175228036487e-06,
      "loss": 0.2705,
      "step": 26360
    },
    {
      "epoch": 6.33,
      "learning_rate": 7.3403744599135876e-06,
      "loss": 0.1737,
      "step": 26370
    },
    {
      "epoch": 6.33,
      "learning_rate": 7.3355736917906876e-06,
      "loss": 0.3457,
      "step": 26380
    },
    {
      "epoch": 6.33,
      "learning_rate": 7.3307729236677875e-06,
      "loss": 0.269,
      "step": 26390
    },
    {
      "epoch": 6.34,
      "learning_rate": 7.325972155544888e-06,
      "loss": 0.0642,
      "step": 26400
    },
    {
      "epoch": 6.34,
      "learning_rate": 7.321171387421988e-06,
      "loss": 0.2375,
      "step": 26410
    },
    {
      "epoch": 6.34,
      "learning_rate": 7.316370619299088e-06,
      "loss": 0.04,
      "step": 26420
    },
    {
      "epoch": 6.34,
      "learning_rate": 7.311569851176189e-06,
      "loss": 0.2352,
      "step": 26430
    },
    {
      "epoch": 6.35,
      "learning_rate": 7.306769083053289e-06,
      "loss": 0.1182,
      "step": 26440
    },
    {
      "epoch": 6.35,
      "learning_rate": 7.30196831493039e-06,
      "loss": 0.1019,
      "step": 26450
    },
    {
      "epoch": 6.35,
      "learning_rate": 7.29716754680749e-06,
      "loss": 0.1283,
      "step": 26460
    },
    {
      "epoch": 6.35,
      "learning_rate": 7.29236677868459e-06,
      "loss": 0.1895,
      "step": 26470
    },
    {
      "epoch": 6.36,
      "learning_rate": 7.287566010561691e-06,
      "loss": 0.0406,
      "step": 26480
    },
    {
      "epoch": 6.36,
      "learning_rate": 7.282765242438791e-06,
      "loss": 0.1872,
      "step": 26490
    },
    {
      "epoch": 6.36,
      "learning_rate": 7.277964474315891e-06,
      "loss": 0.1373,
      "step": 26500
    },
    {
      "epoch": 6.36,
      "learning_rate": 7.273163706192992e-06,
      "loss": 0.2961,
      "step": 26510
    },
    {
      "epoch": 6.37,
      "learning_rate": 7.268362938070092e-06,
      "loss": 0.0306,
      "step": 26520
    },
    {
      "epoch": 6.37,
      "learning_rate": 7.2635621699471926e-06,
      "loss": 0.1013,
      "step": 26530
    },
    {
      "epoch": 6.37,
      "learning_rate": 7.2587614018242926e-06,
      "loss": 0.0054,
      "step": 26540
    },
    {
      "epoch": 6.37,
      "learning_rate": 7.2539606337013926e-06,
      "loss": 0.2625,
      "step": 26550
    },
    {
      "epoch": 6.38,
      "learning_rate": 7.249159865578493e-06,
      "loss": 0.0668,
      "step": 26560
    },
    {
      "epoch": 6.38,
      "learning_rate": 7.244359097455593e-06,
      "loss": 0.2242,
      "step": 26570
    },
    {
      "epoch": 6.38,
      "learning_rate": 7.239558329332694e-06,
      "loss": 0.2363,
      "step": 26580
    },
    {
      "epoch": 6.38,
      "learning_rate": 7.234757561209794e-06,
      "loss": 0.0504,
      "step": 26590
    },
    {
      "epoch": 6.39,
      "learning_rate": 7.229956793086894e-06,
      "loss": 0.0536,
      "step": 26600
    },
    {
      "epoch": 6.39,
      "learning_rate": 7.225156024963995e-06,
      "loss": 0.1208,
      "step": 26610
    },
    {
      "epoch": 6.39,
      "learning_rate": 7.220355256841095e-06,
      "loss": 0.2168,
      "step": 26620
    },
    {
      "epoch": 6.39,
      "learning_rate": 7.215554488718195e-06,
      "loss": 0.1863,
      "step": 26630
    },
    {
      "epoch": 6.39,
      "learning_rate": 7.210753720595296e-06,
      "loss": 0.1843,
      "step": 26640
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.205952952472396e-06,
      "loss": 0.1705,
      "step": 26650
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.201152184349497e-06,
      "loss": 0.1628,
      "step": 26660
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.196351416226597e-06,
      "loss": 0.0485,
      "step": 26670
    },
    {
      "epoch": 6.4,
      "learning_rate": 7.191550648103697e-06,
      "loss": 0.1267,
      "step": 26680
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.186749879980798e-06,
      "loss": 0.1102,
      "step": 26690
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.181949111857898e-06,
      "loss": 0.0499,
      "step": 26700
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.1771483437349984e-06,
      "loss": 0.174,
      "step": 26710
    },
    {
      "epoch": 6.41,
      "learning_rate": 7.1723475756120984e-06,
      "loss": 0.0876,
      "step": 26720
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.1675468074891984e-06,
      "loss": 0.1602,
      "step": 26730
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.162746039366299e-06,
      "loss": 0.1286,
      "step": 26740
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.157945271243399e-06,
      "loss": 0.265,
      "step": 26750
    },
    {
      "epoch": 6.42,
      "learning_rate": 7.153144503120499e-06,
      "loss": 0.1964,
      "step": 26760
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.1483437349976e-06,
      "loss": 0.0691,
      "step": 26770
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.1435429668747e-06,
      "loss": 0.1533,
      "step": 26780
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.138742198751801e-06,
      "loss": 0.1206,
      "step": 26790
    },
    {
      "epoch": 6.43,
      "learning_rate": 7.133941430628901e-06,
      "loss": 0.139,
      "step": 26800
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.129140662506001e-06,
      "loss": 0.1537,
      "step": 26810
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.124339894383102e-06,
      "loss": 0.2052,
      "step": 26820
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.119539126260202e-06,
      "loss": 0.0242,
      "step": 26830
    },
    {
      "epoch": 6.44,
      "learning_rate": 7.1147383581373035e-06,
      "loss": 0.1092,
      "step": 26840
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.109937590014403e-06,
      "loss": 0.2086,
      "step": 26850
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.105136821891503e-06,
      "loss": 0.1234,
      "step": 26860
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.100336053768604e-06,
      "loss": 0.143,
      "step": 26870
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.0955352856457035e-06,
      "loss": 0.1464,
      "step": 26880
    },
    {
      "epoch": 6.45,
      "learning_rate": 7.0907345175228034e-06,
      "loss": 0.0402,
      "step": 26890
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.085933749399905e-06,
      "loss": 0.0983,
      "step": 26900
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.081132981277004e-06,
      "loss": 0.0526,
      "step": 26910
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.076332213154106e-06,
      "loss": 0.1907,
      "step": 26920
    },
    {
      "epoch": 6.46,
      "learning_rate": 7.071531445031206e-06,
      "loss": 0.2767,
      "step": 26930
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.066730676908305e-06,
      "loss": 0.0913,
      "step": 26940
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.061929908785407e-06,
      "loss": 0.0431,
      "step": 26950
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.057129140662507e-06,
      "loss": 0.1337,
      "step": 26960
    },
    {
      "epoch": 6.47,
      "learning_rate": 7.052328372539608e-06,
      "loss": 0.1209,
      "step": 26970
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.047527604416708e-06,
      "loss": 0.1396,
      "step": 26980
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.042726836293808e-06,
      "loss": 0.084,
      "step": 26990
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.0379260681709085e-06,
      "loss": 0.1096,
      "step": 27000
    },
    {
      "epoch": 6.48,
      "learning_rate": 7.0331253000480085e-06,
      "loss": 0.2033,
      "step": 27010
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.0283245319251085e-06,
      "loss": 0.0877,
      "step": 27020
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.023523763802209e-06,
      "loss": 0.0669,
      "step": 27030
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.018722995679309e-06,
      "loss": 0.0395,
      "step": 27040
    },
    {
      "epoch": 6.49,
      "learning_rate": 7.01392222755641e-06,
      "loss": 0.185,
      "step": 27050
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.00912145943351e-06,
      "loss": 0.0763,
      "step": 27060
    },
    {
      "epoch": 6.5,
      "learning_rate": 7.00432069131061e-06,
      "loss": 0.1283,
      "step": 27070
    },
    {
      "epoch": 6.5,
      "learning_rate": 6.999519923187711e-06,
      "loss": 0.0991,
      "step": 27080
    },
    {
      "epoch": 6.5,
      "learning_rate": 6.994719155064811e-06,
      "loss": 0.1041,
      "step": 27090
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.989918386941912e-06,
      "loss": 0.1128,
      "step": 27100
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.985117618819012e-06,
      "loss": 0.0976,
      "step": 27110
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.980316850696112e-06,
      "loss": 0.2417,
      "step": 27120
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.975516082573213e-06,
      "loss": 0.0686,
      "step": 27130
    },
    {
      "epoch": 6.51,
      "learning_rate": 6.970715314450313e-06,
      "loss": 0.0381,
      "step": 27140
    },
    {
      "epoch": 6.52,
      "learning_rate": 6.965914546327413e-06,
      "loss": 0.1548,
      "step": 27150
    },
    {
      "epoch": 6.52,
      "learning_rate": 6.9611137782045135e-06,
      "loss": 0.1737,
      "step": 27160
    },
    {
      "epoch": 6.52,
      "learning_rate": 6.9563130100816135e-06,
      "loss": 0.226,
      "step": 27170
    },
    {
      "epoch": 6.52,
      "learning_rate": 6.951512241958714e-06,
      "loss": 0.2837,
      "step": 27180
    },
    {
      "epoch": 6.53,
      "learning_rate": 6.946711473835814e-06,
      "loss": 0.2096,
      "step": 27190
    },
    {
      "epoch": 6.53,
      "learning_rate": 6.941910705712914e-06,
      "loss": 0.1456,
      "step": 27200
    },
    {
      "epoch": 6.53,
      "learning_rate": 6.937109937590015e-06,
      "loss": 0.1104,
      "step": 27210
    },
    {
      "epoch": 6.53,
      "learning_rate": 6.932309169467115e-06,
      "loss": 0.1061,
      "step": 27220
    },
    {
      "epoch": 6.54,
      "learning_rate": 6.927508401344215e-06,
      "loss": 0.1684,
      "step": 27230
    },
    {
      "epoch": 6.54,
      "learning_rate": 6.922707633221316e-06,
      "loss": 0.183,
      "step": 27240
    },
    {
      "epoch": 6.54,
      "learning_rate": 6.917906865098416e-06,
      "loss": 0.1638,
      "step": 27250
    },
    {
      "epoch": 6.54,
      "learning_rate": 6.913106096975517e-06,
      "loss": 0.2148,
      "step": 27260
    },
    {
      "epoch": 6.55,
      "learning_rate": 6.908305328852617e-06,
      "loss": 0.1398,
      "step": 27270
    },
    {
      "epoch": 6.55,
      "learning_rate": 6.903504560729717e-06,
      "loss": 0.2507,
      "step": 27280
    },
    {
      "epoch": 6.55,
      "learning_rate": 6.898703792606818e-06,
      "loss": 0.1065,
      "step": 27290
    },
    {
      "epoch": 6.55,
      "learning_rate": 6.893903024483918e-06,
      "loss": 0.0586,
      "step": 27300
    },
    {
      "epoch": 6.56,
      "learning_rate": 6.8891022563610185e-06,
      "loss": 0.1674,
      "step": 27310
    },
    {
      "epoch": 6.56,
      "learning_rate": 6.8843014882381185e-06,
      "loss": 0.1305,
      "step": 27320
    },
    {
      "epoch": 6.56,
      "learning_rate": 6.8795007201152185e-06,
      "loss": 0.1653,
      "step": 27330
    },
    {
      "epoch": 6.56,
      "learning_rate": 6.874699951992319e-06,
      "loss": 0.2459,
      "step": 27340
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.869899183869419e-06,
      "loss": 0.1194,
      "step": 27350
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.865098415746519e-06,
      "loss": 0.0173,
      "step": 27360
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.86029764762362e-06,
      "loss": 0.0501,
      "step": 27370
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.85549687950072e-06,
      "loss": 0.1727,
      "step": 27380
    },
    {
      "epoch": 6.57,
      "learning_rate": 6.850696111377821e-06,
      "loss": 0.0827,
      "step": 27390
    },
    {
      "epoch": 6.58,
      "learning_rate": 6.845895343254921e-06,
      "loss": 0.0344,
      "step": 27400
    },
    {
      "epoch": 6.58,
      "learning_rate": 6.841094575132021e-06,
      "loss": 0.1478,
      "step": 27410
    },
    {
      "epoch": 6.58,
      "learning_rate": 6.836293807009122e-06,
      "loss": 0.0105,
      "step": 27420
    },
    {
      "epoch": 6.58,
      "learning_rate": 6.831493038886222e-06,
      "loss": 0.1823,
      "step": 27430
    },
    {
      "epoch": 6.59,
      "learning_rate": 6.8266922707633236e-06,
      "loss": 0.0478,
      "step": 27440
    },
    {
      "epoch": 6.59,
      "learning_rate": 6.821891502640423e-06,
      "loss": 0.1934,
      "step": 27450
    },
    {
      "epoch": 6.59,
      "learning_rate": 6.817090734517523e-06,
      "loss": 0.0652,
      "step": 27460
    },
    {
      "epoch": 6.59,
      "learning_rate": 6.812289966394624e-06,
      "loss": 0.1508,
      "step": 27470
    },
    {
      "epoch": 6.6,
      "learning_rate": 6.8074891982717235e-06,
      "loss": 0.0405,
      "step": 27480
    },
    {
      "epoch": 6.6,
      "learning_rate": 6.8026884301488235e-06,
      "loss": 0.106,
      "step": 27490
    },
    {
      "epoch": 6.6,
      "learning_rate": 6.797887662025925e-06,
      "loss": 0.2549,
      "step": 27500
    },
    {
      "epoch": 6.6,
      "learning_rate": 6.793086893903024e-06,
      "loss": 0.2154,
      "step": 27510
    },
    {
      "epoch": 6.61,
      "learning_rate": 6.788286125780126e-06,
      "loss": 0.2765,
      "step": 27520
    },
    {
      "epoch": 6.61,
      "learning_rate": 6.783485357657226e-06,
      "loss": 0.0385,
      "step": 27530
    },
    {
      "epoch": 6.61,
      "learning_rate": 6.778684589534325e-06,
      "loss": 0.2861,
      "step": 27540
    },
    {
      "epoch": 6.61,
      "learning_rate": 6.773883821411427e-06,
      "loss": 0.112,
      "step": 27550
    },
    {
      "epoch": 6.62,
      "learning_rate": 6.769083053288527e-06,
      "loss": 0.0288,
      "step": 27560
    },
    {
      "epoch": 6.62,
      "learning_rate": 6.764282285165628e-06,
      "loss": 0.0997,
      "step": 27570
    },
    {
      "epoch": 6.62,
      "learning_rate": 6.759481517042728e-06,
      "loss": 0.1873,
      "step": 27580
    },
    {
      "epoch": 6.62,
      "learning_rate": 6.754680748919828e-06,
      "loss": 0.1055,
      "step": 27590
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.749879980796929e-06,
      "loss": 0.3592,
      "step": 27600
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.745079212674029e-06,
      "loss": 0.1721,
      "step": 27610
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.7402784445511286e-06,
      "loss": 0.0783,
      "step": 27620
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.735477676428229e-06,
      "loss": 0.0941,
      "step": 27630
    },
    {
      "epoch": 6.63,
      "learning_rate": 6.730676908305329e-06,
      "loss": 0.1333,
      "step": 27640
    },
    {
      "epoch": 6.64,
      "learning_rate": 6.72587614018243e-06,
      "loss": 0.1835,
      "step": 27650
    },
    {
      "epoch": 6.64,
      "learning_rate": 6.72107537205953e-06,
      "loss": 0.0821,
      "step": 27660
    },
    {
      "epoch": 6.64,
      "learning_rate": 6.71627460393663e-06,
      "loss": 0.1152,
      "step": 27670
    },
    {
      "epoch": 6.64,
      "learning_rate": 6.711473835813731e-06,
      "loss": 0.1185,
      "step": 27680
    },
    {
      "epoch": 6.65,
      "learning_rate": 6.706673067690831e-06,
      "loss": 0.122,
      "step": 27690
    },
    {
      "epoch": 6.65,
      "learning_rate": 6.701872299567932e-06,
      "loss": 0.0585,
      "step": 27700
    },
    {
      "epoch": 6.65,
      "learning_rate": 6.697071531445032e-06,
      "loss": 0.1351,
      "step": 27710
    },
    {
      "epoch": 6.65,
      "learning_rate": 6.692270763322132e-06,
      "loss": 0.1063,
      "step": 27720
    },
    {
      "epoch": 6.66,
      "learning_rate": 6.687469995199233e-06,
      "loss": 0.1178,
      "step": 27730
    },
    {
      "epoch": 6.66,
      "learning_rate": 6.682669227076333e-06,
      "loss": 0.1179,
      "step": 27740
    },
    {
      "epoch": 6.66,
      "learning_rate": 6.677868458953433e-06,
      "loss": 0.0544,
      "step": 27750
    },
    {
      "epoch": 6.66,
      "learning_rate": 6.673067690830534e-06,
      "loss": 0.0045,
      "step": 27760
    },
    {
      "epoch": 6.67,
      "learning_rate": 6.668266922707634e-06,
      "loss": 0.1876,
      "step": 27770
    },
    {
      "epoch": 6.67,
      "learning_rate": 6.6634661545847344e-06,
      "loss": 0.2106,
      "step": 27780
    },
    {
      "epoch": 6.67,
      "learning_rate": 6.6586653864618344e-06,
      "loss": 0.0983,
      "step": 27790
    },
    {
      "epoch": 6.67,
      "learning_rate": 6.6538646183389344e-06,
      "loss": 0.3233,
      "step": 27800
    },
    {
      "epoch": 6.68,
      "learning_rate": 6.649063850216035e-06,
      "loss": 0.1251,
      "step": 27810
    },
    {
      "epoch": 6.68,
      "learning_rate": 6.644263082093135e-06,
      "loss": 0.142,
      "step": 27820
    },
    {
      "epoch": 6.68,
      "learning_rate": 6.639462313970236e-06,
      "loss": 0.0565,
      "step": 27830
    },
    {
      "epoch": 6.68,
      "learning_rate": 6.634661545847336e-06,
      "loss": 0.2549,
      "step": 27840
    },
    {
      "epoch": 6.69,
      "learning_rate": 6.629860777724436e-06,
      "loss": 0.0303,
      "step": 27850
    },
    {
      "epoch": 6.69,
      "learning_rate": 6.625060009601537e-06,
      "loss": 0.1001,
      "step": 27860
    },
    {
      "epoch": 6.69,
      "learning_rate": 6.620259241478637e-06,
      "loss": 0.131,
      "step": 27870
    },
    {
      "epoch": 6.69,
      "learning_rate": 6.615458473355737e-06,
      "loss": 0.1185,
      "step": 27880
    },
    {
      "epoch": 6.69,
      "learning_rate": 6.610657705232838e-06,
      "loss": 0.1637,
      "step": 27890
    },
    {
      "epoch": 6.7,
      "learning_rate": 6.605856937109938e-06,
      "loss": 0.2684,
      "step": 27900
    },
    {
      "epoch": 6.7,
      "learning_rate": 6.601056168987039e-06,
      "loss": 0.0268,
      "step": 27910
    },
    {
      "epoch": 6.7,
      "learning_rate": 6.596255400864139e-06,
      "loss": 0.0646,
      "step": 27920
    },
    {
      "epoch": 6.7,
      "learning_rate": 6.591454632741239e-06,
      "loss": 0.1919,
      "step": 27930
    },
    {
      "epoch": 6.71,
      "learning_rate": 6.5866538646183395e-06,
      "loss": 0.1586,
      "step": 27940
    },
    {
      "epoch": 6.71,
      "learning_rate": 6.5818530964954395e-06,
      "loss": 0.2948,
      "step": 27950
    },
    {
      "epoch": 6.71,
      "learning_rate": 6.5770523283725395e-06,
      "loss": 0.2266,
      "step": 27960
    },
    {
      "epoch": 6.71,
      "learning_rate": 6.57225156024964e-06,
      "loss": 0.1866,
      "step": 27970
    },
    {
      "epoch": 6.72,
      "learning_rate": 6.56745079212674e-06,
      "loss": 0.0298,
      "step": 27980
    },
    {
      "epoch": 6.72,
      "learning_rate": 6.562650024003841e-06,
      "loss": 0.2418,
      "step": 27990
    },
    {
      "epoch": 6.72,
      "learning_rate": 6.557849255880941e-06,
      "loss": 0.2283,
      "step": 28000
    },
    {
      "epoch": 6.72,
      "learning_rate": 6.553048487758041e-06,
      "loss": 0.2184,
      "step": 28010
    },
    {
      "epoch": 6.73,
      "learning_rate": 6.548247719635142e-06,
      "loss": 0.1073,
      "step": 28020
    },
    {
      "epoch": 6.73,
      "learning_rate": 6.543446951512242e-06,
      "loss": 0.1495,
      "step": 28030
    },
    {
      "epoch": 6.73,
      "learning_rate": 6.538646183389344e-06,
      "loss": 0.17,
      "step": 28040
    },
    {
      "epoch": 6.73,
      "learning_rate": 6.533845415266443e-06,
      "loss": 0.1112,
      "step": 28050
    },
    {
      "epoch": 6.74,
      "learning_rate": 6.529044647143543e-06,
      "loss": 0.1676,
      "step": 28060
    },
    {
      "epoch": 6.74,
      "learning_rate": 6.5242438790206445e-06,
      "loss": 0.189,
      "step": 28070
    },
    {
      "epoch": 6.74,
      "learning_rate": 6.5194431108977445e-06,
      "loss": 0.0018,
      "step": 28080
    },
    {
      "epoch": 6.74,
      "learning_rate": 6.514642342774844e-06,
      "loss": 0.0371,
      "step": 28090
    },
    {
      "epoch": 6.75,
      "learning_rate": 6.509841574651945e-06,
      "loss": 0.037,
      "step": 28100
    },
    {
      "epoch": 6.75,
      "learning_rate": 6.505040806529045e-06,
      "loss": 0.0021,
      "step": 28110
    },
    {
      "epoch": 6.75,
      "learning_rate": 6.500240038406146e-06,
      "loss": 0.0804,
      "step": 28120
    },
    {
      "epoch": 6.75,
      "learning_rate": 6.495439270283246e-06,
      "loss": 0.1476,
      "step": 28130
    },
    {
      "epoch": 6.75,
      "learning_rate": 6.490638502160346e-06,
      "loss": 0.139,
      "step": 28140
    },
    {
      "epoch": 6.76,
      "learning_rate": 6.485837734037447e-06,
      "loss": 0.0574,
      "step": 28150
    },
    {
      "epoch": 6.76,
      "learning_rate": 6.481036965914547e-06,
      "loss": 0.1628,
      "step": 28160
    },
    {
      "epoch": 6.76,
      "learning_rate": 6.476236197791648e-06,
      "loss": 0.2185,
      "step": 28170
    },
    {
      "epoch": 6.76,
      "learning_rate": 6.471435429668748e-06,
      "loss": 0.2074,
      "step": 28180
    },
    {
      "epoch": 6.77,
      "learning_rate": 6.466634661545848e-06,
      "loss": 0.0685,
      "step": 28190
    },
    {
      "epoch": 6.77,
      "learning_rate": 6.461833893422949e-06,
      "loss": 0.1158,
      "step": 28200
    },
    {
      "epoch": 6.77,
      "learning_rate": 6.457033125300049e-06,
      "loss": 0.0665,
      "step": 28210
    },
    {
      "epoch": 6.77,
      "learning_rate": 6.452232357177149e-06,
      "loss": 0.0428,
      "step": 28220
    },
    {
      "epoch": 6.78,
      "learning_rate": 6.4474315890542495e-06,
      "loss": 0.0921,
      "step": 28230
    },
    {
      "epoch": 6.78,
      "learning_rate": 6.4426308209313495e-06,
      "loss": 0.1071,
      "step": 28240
    },
    {
      "epoch": 6.78,
      "learning_rate": 6.43783005280845e-06,
      "loss": 0.2545,
      "step": 28250
    },
    {
      "epoch": 6.78,
      "learning_rate": 6.43302928468555e-06,
      "loss": 0.2955,
      "step": 28260
    },
    {
      "epoch": 6.79,
      "learning_rate": 6.42822851656265e-06,
      "loss": 0.1481,
      "step": 28270
    },
    {
      "epoch": 6.79,
      "learning_rate": 6.423427748439751e-06,
      "loss": 0.1973,
      "step": 28280
    },
    {
      "epoch": 6.79,
      "learning_rate": 6.418626980316851e-06,
      "loss": 0.0672,
      "step": 28290
    },
    {
      "epoch": 6.79,
      "learning_rate": 6.413826212193952e-06,
      "loss": 0.0613,
      "step": 28300
    },
    {
      "epoch": 6.8,
      "learning_rate": 6.409025444071052e-06,
      "loss": 0.0977,
      "step": 28310
    },
    {
      "epoch": 6.8,
      "learning_rate": 6.404224675948152e-06,
      "loss": 0.2991,
      "step": 28320
    },
    {
      "epoch": 6.8,
      "learning_rate": 6.399423907825253e-06,
      "loss": 0.066,
      "step": 28330
    },
    {
      "epoch": 6.8,
      "learning_rate": 6.394623139702353e-06,
      "loss": 0.0094,
      "step": 28340
    },
    {
      "epoch": 6.81,
      "learning_rate": 6.389822371579453e-06,
      "loss": 0.1725,
      "step": 28350
    },
    {
      "epoch": 6.81,
      "learning_rate": 6.385021603456554e-06,
      "loss": 0.0944,
      "step": 28360
    },
    {
      "epoch": 6.81,
      "learning_rate": 6.380220835333654e-06,
      "loss": 0.1518,
      "step": 28370
    },
    {
      "epoch": 6.81,
      "learning_rate": 6.3754200672107545e-06,
      "loss": 0.25,
      "step": 28380
    },
    {
      "epoch": 6.81,
      "learning_rate": 6.3706192990878545e-06,
      "loss": 0.1794,
      "step": 28390
    },
    {
      "epoch": 6.82,
      "learning_rate": 6.3658185309649545e-06,
      "loss": 0.1061,
      "step": 28400
    },
    {
      "epoch": 6.82,
      "learning_rate": 6.361017762842055e-06,
      "loss": 0.115,
      "step": 28410
    },
    {
      "epoch": 6.82,
      "learning_rate": 6.356216994719155e-06,
      "loss": 0.148,
      "step": 28420
    },
    {
      "epoch": 6.82,
      "learning_rate": 6.351416226596256e-06,
      "loss": 0.0749,
      "step": 28430
    },
    {
      "epoch": 6.83,
      "learning_rate": 6.346615458473356e-06,
      "loss": 0.0959,
      "step": 28440
    },
    {
      "epoch": 6.83,
      "learning_rate": 6.341814690350456e-06,
      "loss": 0.1134,
      "step": 28450
    },
    {
      "epoch": 6.83,
      "learning_rate": 6.337013922227557e-06,
      "loss": 0.1223,
      "step": 28460
    },
    {
      "epoch": 6.83,
      "learning_rate": 6.332213154104657e-06,
      "loss": 0.0427,
      "step": 28470
    },
    {
      "epoch": 6.84,
      "learning_rate": 6.327412385981757e-06,
      "loss": 0.0784,
      "step": 28480
    },
    {
      "epoch": 6.84,
      "learning_rate": 6.322611617858858e-06,
      "loss": 0.1584,
      "step": 28490
    },
    {
      "epoch": 6.84,
      "learning_rate": 6.317810849735958e-06,
      "loss": 0.2344,
      "step": 28500
    },
    {
      "epoch": 6.84,
      "learning_rate": 6.313010081613059e-06,
      "loss": 0.1309,
      "step": 28510
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.308209313490159e-06,
      "loss": 0.0823,
      "step": 28520
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.303408545367259e-06,
      "loss": 0.057,
      "step": 28530
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.2986077772443596e-06,
      "loss": 0.2559,
      "step": 28540
    },
    {
      "epoch": 6.85,
      "learning_rate": 6.2938070091214596e-06,
      "loss": 0.0904,
      "step": 28550
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.289006240998561e-06,
      "loss": 0.0722,
      "step": 28560
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.28420547287566e-06,
      "loss": 0.1817,
      "step": 28570
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.27940470475276e-06,
      "loss": 0.1436,
      "step": 28580
    },
    {
      "epoch": 6.86,
      "learning_rate": 6.274603936629862e-06,
      "loss": 0.1605,
      "step": 28590
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.269803168506961e-06,
      "loss": 0.0104,
      "step": 28600
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.265002400384061e-06,
      "loss": 0.0916,
      "step": 28610
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.260201632261163e-06,
      "loss": 0.0928,
      "step": 28620
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.255400864138262e-06,
      "loss": 0.1712,
      "step": 28630
    },
    {
      "epoch": 6.87,
      "learning_rate": 6.250600096015364e-06,
      "loss": 0.2175,
      "step": 28640
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.245799327892464e-06,
      "loss": 0.2536,
      "step": 28650
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.240998559769563e-06,
      "loss": 0.0635,
      "step": 28660
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.236197791646665e-06,
      "loss": 0.0912,
      "step": 28670
    },
    {
      "epoch": 6.88,
      "learning_rate": 6.231397023523765e-06,
      "loss": 0.0915,
      "step": 28680
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.226596255400864e-06,
      "loss": 0.0813,
      "step": 28690
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.2217954872779654e-06,
      "loss": 0.0746,
      "step": 28700
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.216994719155065e-06,
      "loss": 0.099,
      "step": 28710
    },
    {
      "epoch": 6.89,
      "learning_rate": 6.212193951032166e-06,
      "loss": 0.1502,
      "step": 28720
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.207393182909266e-06,
      "loss": 0.1392,
      "step": 28730
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.202592414786366e-06,
      "loss": 0.2027,
      "step": 28740
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.197791646663467e-06,
      "loss": 0.0262,
      "step": 28750
    },
    {
      "epoch": 6.9,
      "learning_rate": 6.192990878540567e-06,
      "loss": 0.0815,
      "step": 28760
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.188190110417668e-06,
      "loss": 0.1663,
      "step": 28770
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.183389342294768e-06,
      "loss": 0.121,
      "step": 28780
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.178588574171868e-06,
      "loss": 0.0896,
      "step": 28790
    },
    {
      "epoch": 6.91,
      "learning_rate": 6.173787806048969e-06,
      "loss": 0.175,
      "step": 28800
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.168987037926069e-06,
      "loss": 0.0343,
      "step": 28810
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.164186269803169e-06,
      "loss": 0.0791,
      "step": 28820
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.15938550168027e-06,
      "loss": 0.1533,
      "step": 28830
    },
    {
      "epoch": 6.92,
      "learning_rate": 6.15458473355737e-06,
      "loss": 0.0793,
      "step": 28840
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.1497839654344704e-06,
      "loss": 0.2216,
      "step": 28850
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.1449831973115704e-06,
      "loss": 0.0918,
      "step": 28860
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.1401824291886704e-06,
      "loss": 0.1085,
      "step": 28870
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.135381661065771e-06,
      "loss": 0.0291,
      "step": 28880
    },
    {
      "epoch": 6.93,
      "learning_rate": 6.130580892942871e-06,
      "loss": 0.0958,
      "step": 28890
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.125780124819972e-06,
      "loss": 0.0622,
      "step": 28900
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.120979356697072e-06,
      "loss": 0.0927,
      "step": 28910
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.116178588574172e-06,
      "loss": 0.0762,
      "step": 28920
    },
    {
      "epoch": 6.94,
      "learning_rate": 6.111377820451273e-06,
      "loss": 0.0212,
      "step": 28930
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.106577052328373e-06,
      "loss": 0.2022,
      "step": 28940
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.101776284205473e-06,
      "loss": 0.2169,
      "step": 28950
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.096975516082574e-06,
      "loss": 0.1167,
      "step": 28960
    },
    {
      "epoch": 6.95,
      "learning_rate": 6.092174747959674e-06,
      "loss": 0.1172,
      "step": 28970
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.087373979836775e-06,
      "loss": 0.1195,
      "step": 28980
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.082573211713875e-06,
      "loss": 0.0863,
      "step": 28990
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.077772443590975e-06,
      "loss": 0.1059,
      "step": 29000
    },
    {
      "epoch": 6.96,
      "learning_rate": 6.0729716754680755e-06,
      "loss": 0.2612,
      "step": 29010
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.0681709073451755e-06,
      "loss": 0.1776,
      "step": 29020
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.063370139222276e-06,
      "loss": 0.1468,
      "step": 29030
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.058569371099376e-06,
      "loss": 0.2081,
      "step": 29040
    },
    {
      "epoch": 6.97,
      "learning_rate": 6.053768602976476e-06,
      "loss": 0.0333,
      "step": 29050
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.048967834853577e-06,
      "loss": 0.0971,
      "step": 29060
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.044167066730677e-06,
      "loss": 0.0909,
      "step": 29070
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.039366298607777e-06,
      "loss": 0.0325,
      "step": 29080
    },
    {
      "epoch": 6.98,
      "learning_rate": 6.034565530484878e-06,
      "loss": 0.1671,
      "step": 29090
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.029764762361978e-06,
      "loss": 0.1051,
      "step": 29100
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.024963994239079e-06,
      "loss": 0.1591,
      "step": 29110
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.020163226116179e-06,
      "loss": 0.1524,
      "step": 29120
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.015362457993279e-06,
      "loss": 0.131,
      "step": 29130
    },
    {
      "epoch": 6.99,
      "learning_rate": 6.01056168987038e-06,
      "loss": 0.1134,
      "step": 29140
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.00576092174748e-06,
      "loss": 0.0578,
      "step": 29150
    },
    {
      "epoch": 7.0,
      "learning_rate": 6.000960153624581e-06,
      "loss": 0.0382,
      "step": 29160
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.9989799387963277,
      "eval_loss": 0.0041446564719080925,
      "eval_runtime": 512.3759,
      "eval_samples_per_second": 11.48,
      "eval_steps_per_second": 1.436,
      "step": 29162
    },
    {
      "epoch": 7.0,
      "learning_rate": 5.9961593855016805e-06,
      "loss": 0.0427,
      "step": 29170
    },
    {
      "epoch": 7.0,
      "learning_rate": 5.9913586173787805e-06,
      "loss": 0.1351,
      "step": 29180
    },
    {
      "epoch": 7.01,
      "learning_rate": 5.986557849255882e-06,
      "loss": 0.0226,
      "step": 29190
    },
    {
      "epoch": 7.01,
      "learning_rate": 5.981757081132981e-06,
      "loss": 0.3388,
      "step": 29200
    },
    {
      "epoch": 7.01,
      "learning_rate": 5.976956313010081e-06,
      "loss": 0.1873,
      "step": 29210
    },
    {
      "epoch": 7.01,
      "learning_rate": 5.972155544887183e-06,
      "loss": 0.0675,
      "step": 29220
    },
    {
      "epoch": 7.02,
      "learning_rate": 5.967354776764282e-06,
      "loss": 0.1323,
      "step": 29230
    },
    {
      "epoch": 7.02,
      "learning_rate": 5.962554008641384e-06,
      "loss": 0.0186,
      "step": 29240
    },
    {
      "epoch": 7.02,
      "learning_rate": 5.957753240518484e-06,
      "loss": 0.1142,
      "step": 29250
    },
    {
      "epoch": 7.02,
      "learning_rate": 5.952952472395583e-06,
      "loss": 0.2242,
      "step": 29260
    },
    {
      "epoch": 7.03,
      "learning_rate": 5.948151704272685e-06,
      "loss": 0.1497,
      "step": 29270
    },
    {
      "epoch": 7.03,
      "learning_rate": 5.943350936149785e-06,
      "loss": 0.1228,
      "step": 29280
    },
    {
      "epoch": 7.03,
      "learning_rate": 5.9385501680268855e-06,
      "loss": 0.1011,
      "step": 29290
    },
    {
      "epoch": 7.03,
      "learning_rate": 5.9337493999039855e-06,
      "loss": 0.1164,
      "step": 29300
    },
    {
      "epoch": 7.04,
      "learning_rate": 5.9289486317810855e-06,
      "loss": 0.1214,
      "step": 29310
    },
    {
      "epoch": 7.04,
      "learning_rate": 5.924147863658186e-06,
      "loss": 0.1761,
      "step": 29320
    },
    {
      "epoch": 7.04,
      "learning_rate": 5.919347095535286e-06,
      "loss": 0.2364,
      "step": 29330
    },
    {
      "epoch": 7.04,
      "learning_rate": 5.914546327412386e-06,
      "loss": 0.2313,
      "step": 29340
    },
    {
      "epoch": 7.05,
      "learning_rate": 5.909745559289487e-06,
      "loss": 0.0919,
      "step": 29350
    },
    {
      "epoch": 7.05,
      "learning_rate": 5.904944791166587e-06,
      "loss": 0.1262,
      "step": 29360
    },
    {
      "epoch": 7.05,
      "learning_rate": 5.900144023043688e-06,
      "loss": 0.1081,
      "step": 29370
    },
    {
      "epoch": 7.05,
      "learning_rate": 5.895343254920788e-06,
      "loss": 0.0264,
      "step": 29380
    },
    {
      "epoch": 7.05,
      "learning_rate": 5.890542486797888e-06,
      "loss": 0.2043,
      "step": 29390
    },
    {
      "epoch": 7.06,
      "learning_rate": 5.885741718674989e-06,
      "loss": 0.1089,
      "step": 29400
    },
    {
      "epoch": 7.06,
      "learning_rate": 5.880940950552089e-06,
      "loss": 0.0956,
      "step": 29410
    },
    {
      "epoch": 7.06,
      "learning_rate": 5.876140182429189e-06,
      "loss": 0.1844,
      "step": 29420
    },
    {
      "epoch": 7.06,
      "learning_rate": 5.87133941430629e-06,
      "loss": 0.0381,
      "step": 29430
    },
    {
      "epoch": 7.07,
      "learning_rate": 5.86653864618339e-06,
      "loss": 0.0449,
      "step": 29440
    },
    {
      "epoch": 7.07,
      "learning_rate": 5.8617378780604905e-06,
      "loss": 0.1502,
      "step": 29450
    },
    {
      "epoch": 7.07,
      "learning_rate": 5.8569371099375905e-06,
      "loss": 0.0263,
      "step": 29460
    },
    {
      "epoch": 7.07,
      "learning_rate": 5.8521363418146905e-06,
      "loss": 0.1507,
      "step": 29470
    },
    {
      "epoch": 7.08,
      "learning_rate": 5.847335573691791e-06,
      "loss": 0.1064,
      "step": 29480
    },
    {
      "epoch": 7.08,
      "learning_rate": 5.842534805568891e-06,
      "loss": 0.0999,
      "step": 29490
    },
    {
      "epoch": 7.08,
      "learning_rate": 5.837734037445992e-06,
      "loss": 0.0132,
      "step": 29500
    },
    {
      "epoch": 7.08,
      "learning_rate": 5.832933269323092e-06,
      "loss": 0.2197,
      "step": 29510
    },
    {
      "epoch": 7.09,
      "learning_rate": 5.828132501200192e-06,
      "loss": 0.0369,
      "step": 29520
    },
    {
      "epoch": 7.09,
      "learning_rate": 5.823331733077293e-06,
      "loss": 0.1048,
      "step": 29530
    },
    {
      "epoch": 7.09,
      "learning_rate": 5.818530964954393e-06,
      "loss": 0.0706,
      "step": 29540
    },
    {
      "epoch": 7.09,
      "learning_rate": 5.813730196831493e-06,
      "loss": 0.1868,
      "step": 29550
    },
    {
      "epoch": 7.1,
      "learning_rate": 5.808929428708594e-06,
      "loss": 0.087,
      "step": 29560
    },
    {
      "epoch": 7.1,
      "learning_rate": 5.804128660585694e-06,
      "loss": 0.0398,
      "step": 29570
    },
    {
      "epoch": 7.1,
      "learning_rate": 5.799327892462795e-06,
      "loss": 0.1526,
      "step": 29580
    },
    {
      "epoch": 7.1,
      "learning_rate": 5.794527124339895e-06,
      "loss": 0.1448,
      "step": 29590
    },
    {
      "epoch": 7.11,
      "learning_rate": 5.789726356216995e-06,
      "loss": 0.1866,
      "step": 29600
    },
    {
      "epoch": 7.11,
      "learning_rate": 5.7849255880940956e-06,
      "loss": 0.0975,
      "step": 29610
    },
    {
      "epoch": 7.11,
      "learning_rate": 5.7801248199711956e-06,
      "loss": 0.0988,
      "step": 29620
    },
    {
      "epoch": 7.11,
      "learning_rate": 5.775324051848296e-06,
      "loss": 0.1986,
      "step": 29630
    },
    {
      "epoch": 7.11,
      "learning_rate": 5.770523283725396e-06,
      "loss": 0.1569,
      "step": 29640
    },
    {
      "epoch": 7.12,
      "learning_rate": 5.765722515602496e-06,
      "loss": 0.1155,
      "step": 29650
    },
    {
      "epoch": 7.12,
      "learning_rate": 5.760921747479597e-06,
      "loss": 0.1231,
      "step": 29660
    },
    {
      "epoch": 7.12,
      "learning_rate": 5.756120979356697e-06,
      "loss": 0.1436,
      "step": 29670
    },
    {
      "epoch": 7.12,
      "learning_rate": 5.751320211233797e-06,
      "loss": 0.1439,
      "step": 29680
    },
    {
      "epoch": 7.13,
      "learning_rate": 5.746519443110898e-06,
      "loss": 0.2206,
      "step": 29690
    },
    {
      "epoch": 7.13,
      "learning_rate": 5.741718674987998e-06,
      "loss": 0.1349,
      "step": 29700
    },
    {
      "epoch": 7.13,
      "learning_rate": 5.736917906865099e-06,
      "loss": 0.1502,
      "step": 29710
    },
    {
      "epoch": 7.13,
      "learning_rate": 5.732117138742199e-06,
      "loss": 0.2065,
      "step": 29720
    },
    {
      "epoch": 7.14,
      "learning_rate": 5.727316370619299e-06,
      "loss": 0.0464,
      "step": 29730
    },
    {
      "epoch": 7.14,
      "learning_rate": 5.7225156024964e-06,
      "loss": 0.0204,
      "step": 29740
    },
    {
      "epoch": 7.14,
      "learning_rate": 5.7177148343735e-06,
      "loss": 0.1609,
      "step": 29750
    },
    {
      "epoch": 7.14,
      "learning_rate": 5.7129140662506014e-06,
      "loss": 0.1828,
      "step": 29760
    },
    {
      "epoch": 7.15,
      "learning_rate": 5.708113298127701e-06,
      "loss": 0.1521,
      "step": 29770
    },
    {
      "epoch": 7.15,
      "learning_rate": 5.703312530004801e-06,
      "loss": 0.1619,
      "step": 29780
    },
    {
      "epoch": 7.15,
      "learning_rate": 5.698511761881902e-06,
      "loss": 0.0323,
      "step": 29790
    },
    {
      "epoch": 7.15,
      "learning_rate": 5.693710993759001e-06,
      "loss": 0.0438,
      "step": 29800
    },
    {
      "epoch": 7.16,
      "learning_rate": 5.688910225636101e-06,
      "loss": 0.1842,
      "step": 29810
    },
    {
      "epoch": 7.16,
      "learning_rate": 5.684109457513203e-06,
      "loss": 0.1751,
      "step": 29820
    },
    {
      "epoch": 7.16,
      "learning_rate": 5.679308689390302e-06,
      "loss": 0.1748,
      "step": 29830
    },
    {
      "epoch": 7.16,
      "learning_rate": 5.674507921267404e-06,
      "loss": 0.1454,
      "step": 29840
    },
    {
      "epoch": 7.17,
      "learning_rate": 5.669707153144504e-06,
      "loss": 0.1514,
      "step": 29850
    },
    {
      "epoch": 7.17,
      "learning_rate": 5.664906385021603e-06,
      "loss": 0.0321,
      "step": 29860
    },
    {
      "epoch": 7.17,
      "learning_rate": 5.660105616898705e-06,
      "loss": 0.0465,
      "step": 29870
    },
    {
      "epoch": 7.17,
      "learning_rate": 5.655304848775805e-06,
      "loss": 0.1633,
      "step": 29880
    },
    {
      "epoch": 7.17,
      "learning_rate": 5.650504080652906e-06,
      "loss": 0.2442,
      "step": 29890
    },
    {
      "epoch": 7.18,
      "learning_rate": 5.645703312530006e-06,
      "loss": 0.1108,
      "step": 29900
    },
    {
      "epoch": 7.18,
      "learning_rate": 5.640902544407106e-06,
      "loss": 0.1613,
      "step": 29910
    },
    {
      "epoch": 7.18,
      "learning_rate": 5.6361017762842065e-06,
      "loss": 0.0198,
      "step": 29920
    },
    {
      "epoch": 7.18,
      "learning_rate": 5.6313010081613065e-06,
      "loss": 0.0772,
      "step": 29930
    },
    {
      "epoch": 7.19,
      "learning_rate": 5.6265002400384064e-06,
      "loss": 0.0637,
      "step": 29940
    },
    {
      "epoch": 7.19,
      "learning_rate": 5.621699471915507e-06,
      "loss": 0.0334,
      "step": 29950
    },
    {
      "epoch": 7.19,
      "learning_rate": 5.616898703792607e-06,
      "loss": 0.1742,
      "step": 29960
    },
    {
      "epoch": 7.19,
      "learning_rate": 5.612097935669708e-06,
      "loss": 0.2029,
      "step": 29970
    },
    {
      "epoch": 7.2,
      "learning_rate": 5.607297167546808e-06,
      "loss": 0.0441,
      "step": 29980
    },
    {
      "epoch": 7.2,
      "learning_rate": 5.602496399423908e-06,
      "loss": 0.0548,
      "step": 29990
    },
    {
      "epoch": 7.2,
      "learning_rate": 5.597695631301009e-06,
      "loss": 0.1157,
      "step": 30000
    },
    {
      "epoch": 7.2,
      "learning_rate": 5.592894863178109e-06,
      "loss": 0.1097,
      "step": 30010
    },
    {
      "epoch": 7.21,
      "learning_rate": 5.58809409505521e-06,
      "loss": 0.2888,
      "step": 30020
    },
    {
      "epoch": 7.21,
      "learning_rate": 5.58329332693231e-06,
      "loss": 0.1438,
      "step": 30030
    },
    {
      "epoch": 7.21,
      "learning_rate": 5.57849255880941e-06,
      "loss": 0.0536,
      "step": 30040
    },
    {
      "epoch": 7.21,
      "learning_rate": 5.573691790686511e-06,
      "loss": 0.1964,
      "step": 30050
    },
    {
      "epoch": 7.22,
      "learning_rate": 5.568891022563611e-06,
      "loss": 0.1537,
      "step": 30060
    },
    {
      "epoch": 7.22,
      "learning_rate": 5.564090254440711e-06,
      "loss": 0.182,
      "step": 30070
    },
    {
      "epoch": 7.22,
      "learning_rate": 5.5592894863178115e-06,
      "loss": 0.1942,
      "step": 30080
    },
    {
      "epoch": 7.22,
      "learning_rate": 5.5544887181949115e-06,
      "loss": 0.1047,
      "step": 30090
    },
    {
      "epoch": 7.23,
      "learning_rate": 5.549687950072012e-06,
      "loss": 0.0543,
      "step": 30100
    },
    {
      "epoch": 7.23,
      "learning_rate": 5.544887181949112e-06,
      "loss": 0.1204,
      "step": 30110
    },
    {
      "epoch": 7.23,
      "learning_rate": 5.540086413826212e-06,
      "loss": 0.0519,
      "step": 30120
    },
    {
      "epoch": 7.23,
      "learning_rate": 5.535285645703313e-06,
      "loss": 0.2937,
      "step": 30130
    },
    {
      "epoch": 7.23,
      "learning_rate": 5.530484877580413e-06,
      "loss": 0.1944,
      "step": 30140
    },
    {
      "epoch": 7.24,
      "learning_rate": 5.525684109457513e-06,
      "loss": 0.1251,
      "step": 30150
    },
    {
      "epoch": 7.24,
      "learning_rate": 5.520883341334614e-06,
      "loss": 0.1634,
      "step": 30160
    },
    {
      "epoch": 7.24,
      "learning_rate": 5.516082573211714e-06,
      "loss": 0.1002,
      "step": 30170
    },
    {
      "epoch": 7.24,
      "learning_rate": 5.511281805088815e-06,
      "loss": 0.1376,
      "step": 30180
    },
    {
      "epoch": 7.25,
      "learning_rate": 5.506481036965915e-06,
      "loss": 0.1307,
      "step": 30190
    },
    {
      "epoch": 7.25,
      "learning_rate": 5.501680268843015e-06,
      "loss": 0.241,
      "step": 30200
    },
    {
      "epoch": 7.25,
      "learning_rate": 5.496879500720116e-06,
      "loss": 0.1038,
      "step": 30210
    },
    {
      "epoch": 7.25,
      "learning_rate": 5.492078732597216e-06,
      "loss": 0.1768,
      "step": 30220
    },
    {
      "epoch": 7.26,
      "learning_rate": 5.4872779644743165e-06,
      "loss": 0.1427,
      "step": 30230
    },
    {
      "epoch": 7.26,
      "learning_rate": 5.4824771963514165e-06,
      "loss": 0.1265,
      "step": 30240
    },
    {
      "epoch": 7.26,
      "learning_rate": 5.4776764282285165e-06,
      "loss": 0.0629,
      "step": 30250
    },
    {
      "epoch": 7.26,
      "learning_rate": 5.472875660105617e-06,
      "loss": 0.1155,
      "step": 30260
    },
    {
      "epoch": 7.27,
      "learning_rate": 5.468074891982717e-06,
      "loss": 0.0592,
      "step": 30270
    },
    {
      "epoch": 7.27,
      "learning_rate": 5.463274123859817e-06,
      "loss": 0.2225,
      "step": 30280
    },
    {
      "epoch": 7.27,
      "learning_rate": 5.458473355736918e-06,
      "loss": 0.1028,
      "step": 30290
    },
    {
      "epoch": 7.27,
      "learning_rate": 5.453672587614018e-06,
      "loss": 0.0832,
      "step": 30300
    },
    {
      "epoch": 7.28,
      "learning_rate": 5.448871819491119e-06,
      "loss": 0.0684,
      "step": 30310
    },
    {
      "epoch": 7.28,
      "learning_rate": 5.444071051368219e-06,
      "loss": 0.1697,
      "step": 30320
    },
    {
      "epoch": 7.28,
      "learning_rate": 5.439270283245319e-06,
      "loss": 0.1072,
      "step": 30330
    },
    {
      "epoch": 7.28,
      "learning_rate": 5.434469515122421e-06,
      "loss": 0.2466,
      "step": 30340
    },
    {
      "epoch": 7.29,
      "learning_rate": 5.42966874699952e-06,
      "loss": 0.1203,
      "step": 30350
    },
    {
      "epoch": 7.29,
      "learning_rate": 5.4248679788766215e-06,
      "loss": 0.0798,
      "step": 30360
    },
    {
      "epoch": 7.29,
      "learning_rate": 5.4200672107537215e-06,
      "loss": 0.0639,
      "step": 30370
    },
    {
      "epoch": 7.29,
      "learning_rate": 5.415266442630821e-06,
      "loss": 0.0231,
      "step": 30380
    },
    {
      "epoch": 7.29,
      "learning_rate": 5.410465674507922e-06,
      "loss": 0.1542,
      "step": 30390
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.405664906385022e-06,
      "loss": 0.1338,
      "step": 30400
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.4008641382621215e-06,
      "loss": 0.0452,
      "step": 30410
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.396063370139223e-06,
      "loss": 0.1355,
      "step": 30420
    },
    {
      "epoch": 7.3,
      "learning_rate": 5.391262602016323e-06,
      "loss": 0.0517,
      "step": 30430
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.386461833893424e-06,
      "loss": 0.0767,
      "step": 30440
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.381661065770524e-06,
      "loss": 0.124,
      "step": 30450
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.376860297647624e-06,
      "loss": 0.0329,
      "step": 30460
    },
    {
      "epoch": 7.31,
      "learning_rate": 5.372059529524725e-06,
      "loss": 0.1021,
      "step": 30470
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.367258761401825e-06,
      "loss": 0.2603,
      "step": 30480
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.362457993278926e-06,
      "loss": 0.057,
      "step": 30490
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.357657225156026e-06,
      "loss": 0.0372,
      "step": 30500
    },
    {
      "epoch": 7.32,
      "learning_rate": 5.352856457033126e-06,
      "loss": 0.043,
      "step": 30510
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.3480556889102266e-06,
      "loss": 0.0619,
      "step": 30520
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.3432549207873265e-06,
      "loss": 0.1371,
      "step": 30530
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.3384541526644265e-06,
      "loss": 0.0139,
      "step": 30540
    },
    {
      "epoch": 7.33,
      "learning_rate": 5.333653384541527e-06,
      "loss": 0.2777,
      "step": 30550
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.328852616418627e-06,
      "loss": 0.0778,
      "step": 30560
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.324051848295728e-06,
      "loss": 0.2275,
      "step": 30570
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.319251080172828e-06,
      "loss": 0.0845,
      "step": 30580
    },
    {
      "epoch": 7.34,
      "learning_rate": 5.314450312049928e-06,
      "loss": 0.0851,
      "step": 30590
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.309649543927029e-06,
      "loss": 0.0897,
      "step": 30600
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.304848775804129e-06,
      "loss": 0.1516,
      "step": 30610
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.30004800768123e-06,
      "loss": 0.1013,
      "step": 30620
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.29524723955833e-06,
      "loss": 0.1872,
      "step": 30630
    },
    {
      "epoch": 7.35,
      "learning_rate": 5.29044647143543e-06,
      "loss": 0.1838,
      "step": 30640
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.285645703312531e-06,
      "loss": 0.3258,
      "step": 30650
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.280844935189631e-06,
      "loss": 0.1336,
      "step": 30660
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.276044167066731e-06,
      "loss": 0.0836,
      "step": 30670
    },
    {
      "epoch": 7.36,
      "learning_rate": 5.2712433989438316e-06,
      "loss": 0.0752,
      "step": 30680
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.2664426308209316e-06,
      "loss": 0.1298,
      "step": 30690
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.261641862698032e-06,
      "loss": 0.1682,
      "step": 30700
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.256841094575132e-06,
      "loss": 0.1267,
      "step": 30710
    },
    {
      "epoch": 7.37,
      "learning_rate": 5.252040326452232e-06,
      "loss": 0.3594,
      "step": 30720
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.247239558329333e-06,
      "loss": 0.1903,
      "step": 30730
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.242438790206433e-06,
      "loss": 0.1051,
      "step": 30740
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.237638022083534e-06,
      "loss": 0.1873,
      "step": 30750
    },
    {
      "epoch": 7.38,
      "learning_rate": 5.232837253960634e-06,
      "loss": 0.1266,
      "step": 30760
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.228036485837734e-06,
      "loss": 0.0347,
      "step": 30770
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.223235717714835e-06,
      "loss": 0.1071,
      "step": 30780
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.218434949591935e-06,
      "loss": 0.1605,
      "step": 30790
    },
    {
      "epoch": 7.39,
      "learning_rate": 5.213634181469035e-06,
      "loss": 0.1368,
      "step": 30800
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.208833413346136e-06,
      "loss": 0.146,
      "step": 30810
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.204032645223236e-06,
      "loss": 0.1211,
      "step": 30820
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.199231877100337e-06,
      "loss": 0.1252,
      "step": 30830
    },
    {
      "epoch": 7.4,
      "learning_rate": 5.194431108977437e-06,
      "loss": 0.1082,
      "step": 30840
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.189630340854537e-06,
      "loss": 0.0754,
      "step": 30850
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.1848295727316374e-06,
      "loss": 0.1623,
      "step": 30860
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.1800288046087374e-06,
      "loss": 0.1211,
      "step": 30870
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.175228036485837e-06,
      "loss": 0.2528,
      "step": 30880
    },
    {
      "epoch": 7.41,
      "learning_rate": 5.170427268362938e-06,
      "loss": 0.1014,
      "step": 30890
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.165626500240038e-06,
      "loss": 0.1172,
      "step": 30900
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.16082573211714e-06,
      "loss": 0.0743,
      "step": 30910
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.156024963994239e-06,
      "loss": 0.1039,
      "step": 30920
    },
    {
      "epoch": 7.42,
      "learning_rate": 5.151224195871339e-06,
      "loss": 0.0924,
      "step": 30930
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.146423427748441e-06,
      "loss": 0.0778,
      "step": 30940
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.14162265962554e-06,
      "loss": 0.1223,
      "step": 30950
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.136821891502642e-06,
      "loss": 0.1135,
      "step": 30960
    },
    {
      "epoch": 7.43,
      "learning_rate": 5.132021123379742e-06,
      "loss": 0.0646,
      "step": 30970
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.127220355256841e-06,
      "loss": 0.1503,
      "step": 30980
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.1224195871339425e-06,
      "loss": 0.1705,
      "step": 30990
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.1176188190110425e-06,
      "loss": 0.0655,
      "step": 31000
    },
    {
      "epoch": 7.44,
      "learning_rate": 5.112818050888142e-06,
      "loss": 0.2437,
      "step": 31010
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.108017282765243e-06,
      "loss": 0.1258,
      "step": 31020
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.103216514642343e-06,
      "loss": 0.0092,
      "step": 31030
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.098415746519444e-06,
      "loss": 0.1024,
      "step": 31040
    },
    {
      "epoch": 7.45,
      "learning_rate": 5.093614978396544e-06,
      "loss": 0.1543,
      "step": 31050
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.088814210273644e-06,
      "loss": 0.2027,
      "step": 31060
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.084013442150745e-06,
      "loss": 0.078,
      "step": 31070
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.079212674027845e-06,
      "loss": 0.1359,
      "step": 31080
    },
    {
      "epoch": 7.46,
      "learning_rate": 5.074411905904946e-06,
      "loss": 0.1248,
      "step": 31090
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.069611137782046e-06,
      "loss": 0.045,
      "step": 31100
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.064810369659146e-06,
      "loss": 0.2843,
      "step": 31110
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.060009601536247e-06,
      "loss": 0.1168,
      "step": 31120
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.055208833413347e-06,
      "loss": 0.203,
      "step": 31130
    },
    {
      "epoch": 7.47,
      "learning_rate": 5.050408065290447e-06,
      "loss": 0.092,
      "step": 31140
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.0456072971675475e-06,
      "loss": 0.1803,
      "step": 31150
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.0408065290446475e-06,
      "loss": 0.152,
      "step": 31160
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.036005760921748e-06,
      "loss": 0.0967,
      "step": 31170
    },
    {
      "epoch": 7.48,
      "learning_rate": 5.031204992798848e-06,
      "loss": 0.035,
      "step": 31180
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.026404224675948e-06,
      "loss": 0.0353,
      "step": 31190
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.021603456553049e-06,
      "loss": 0.0645,
      "step": 31200
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.016802688430149e-06,
      "loss": 0.1822,
      "step": 31210
    },
    {
      "epoch": 7.49,
      "learning_rate": 5.01200192030725e-06,
      "loss": 0.0498,
      "step": 31220
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.00720115218435e-06,
      "loss": 0.1,
      "step": 31230
    },
    {
      "epoch": 7.5,
      "learning_rate": 5.00240038406145e-06,
      "loss": 0.1453,
      "step": 31240
    },
    {
      "epoch": 7.5,
      "learning_rate": 4.99759961593855e-06,
      "loss": 0.1099,
      "step": 31250
    },
    {
      "epoch": 7.5,
      "learning_rate": 4.992798847815651e-06,
      "loss": 0.0734,
      "step": 31260
    },
    {
      "epoch": 7.51,
      "learning_rate": 4.987998079692752e-06,
      "loss": 0.0992,
      "step": 31270
    },
    {
      "epoch": 7.51,
      "learning_rate": 4.983197311569852e-06,
      "loss": 0.1218,
      "step": 31280
    },
    {
      "epoch": 7.51,
      "learning_rate": 4.978396543446952e-06,
      "loss": 0.118,
      "step": 31290
    },
    {
      "epoch": 7.51,
      "learning_rate": 4.9735957753240525e-06,
      "loss": 0.0957,
      "step": 31300
    },
    {
      "epoch": 7.52,
      "learning_rate": 4.9687950072011525e-06,
      "loss": 0.0668,
      "step": 31310
    },
    {
      "epoch": 7.52,
      "learning_rate": 4.963994239078253e-06,
      "loss": 0.0022,
      "step": 31320
    },
    {
      "epoch": 7.52,
      "learning_rate": 4.959193470955353e-06,
      "loss": 0.103,
      "step": 31330
    },
    {
      "epoch": 7.52,
      "learning_rate": 4.954392702832453e-06,
      "loss": 0.1837,
      "step": 31340
    },
    {
      "epoch": 7.53,
      "learning_rate": 4.949591934709554e-06,
      "loss": 0.1301,
      "step": 31350
    },
    {
      "epoch": 7.53,
      "learning_rate": 4.944791166586654e-06,
      "loss": 0.1617,
      "step": 31360
    },
    {
      "epoch": 7.53,
      "learning_rate": 4.939990398463755e-06,
      "loss": 0.1388,
      "step": 31370
    },
    {
      "epoch": 7.53,
      "learning_rate": 4.935189630340855e-06,
      "loss": 0.0231,
      "step": 31380
    },
    {
      "epoch": 7.53,
      "learning_rate": 4.930388862217955e-06,
      "loss": 0.1248,
      "step": 31390
    },
    {
      "epoch": 7.54,
      "learning_rate": 4.925588094095056e-06,
      "loss": 0.3422,
      "step": 31400
    },
    {
      "epoch": 7.54,
      "learning_rate": 4.920787325972156e-06,
      "loss": 0.095,
      "step": 31410
    },
    {
      "epoch": 7.54,
      "learning_rate": 4.915986557849256e-06,
      "loss": 0.1259,
      "step": 31420
    },
    {
      "epoch": 7.54,
      "learning_rate": 4.911185789726357e-06,
      "loss": 0.1291,
      "step": 31430
    },
    {
      "epoch": 7.55,
      "learning_rate": 4.906385021603457e-06,
      "loss": 0.1223,
      "step": 31440
    },
    {
      "epoch": 7.55,
      "learning_rate": 4.9015842534805575e-06,
      "loss": 0.1171,
      "step": 31450
    },
    {
      "epoch": 7.55,
      "learning_rate": 4.8967834853576575e-06,
      "loss": 0.0748,
      "step": 31460
    },
    {
      "epoch": 7.55,
      "learning_rate": 4.8919827172347575e-06,
      "loss": 0.3344,
      "step": 31470
    },
    {
      "epoch": 7.56,
      "learning_rate": 4.887181949111858e-06,
      "loss": 0.3566,
      "step": 31480
    },
    {
      "epoch": 7.56,
      "learning_rate": 4.882381180988958e-06,
      "loss": 0.1322,
      "step": 31490
    },
    {
      "epoch": 7.56,
      "learning_rate": 4.877580412866059e-06,
      "loss": 0.0181,
      "step": 31500
    },
    {
      "epoch": 7.56,
      "learning_rate": 4.872779644743159e-06,
      "loss": 0.0393,
      "step": 31510
    },
    {
      "epoch": 7.57,
      "learning_rate": 4.867978876620259e-06,
      "loss": 0.1927,
      "step": 31520
    },
    {
      "epoch": 7.57,
      "learning_rate": 4.86317810849736e-06,
      "loss": 0.1775,
      "step": 31530
    },
    {
      "epoch": 7.57,
      "learning_rate": 4.858377340374461e-06,
      "loss": 0.2065,
      "step": 31540
    },
    {
      "epoch": 7.57,
      "learning_rate": 4.85357657225156e-06,
      "loss": 0.1311,
      "step": 31550
    },
    {
      "epoch": 7.58,
      "learning_rate": 4.848775804128661e-06,
      "loss": 0.1337,
      "step": 31560
    },
    {
      "epoch": 7.58,
      "learning_rate": 4.843975036005762e-06,
      "loss": 0.108,
      "step": 31570
    },
    {
      "epoch": 7.58,
      "learning_rate": 4.839174267882862e-06,
      "loss": 0.2268,
      "step": 31580
    },
    {
      "epoch": 7.58,
      "learning_rate": 4.834373499759962e-06,
      "loss": 0.263,
      "step": 31590
    },
    {
      "epoch": 7.59,
      "learning_rate": 4.8295727316370626e-06,
      "loss": 0.2218,
      "step": 31600
    },
    {
      "epoch": 7.59,
      "learning_rate": 4.8247719635141625e-06,
      "loss": 0.1218,
      "step": 31610
    },
    {
      "epoch": 7.59,
      "learning_rate": 4.819971195391263e-06,
      "loss": 0.0787,
      "step": 31620
    },
    {
      "epoch": 7.59,
      "learning_rate": 4.815170427268363e-06,
      "loss": 0.0269,
      "step": 31630
    },
    {
      "epoch": 7.59,
      "learning_rate": 4.810369659145463e-06,
      "loss": 0.0866,
      "step": 31640
    },
    {
      "epoch": 7.6,
      "learning_rate": 4.805568891022564e-06,
      "loss": 0.1543,
      "step": 31650
    },
    {
      "epoch": 7.6,
      "learning_rate": 4.800768122899664e-06,
      "loss": 0.0988,
      "step": 31660
    },
    {
      "epoch": 7.6,
      "learning_rate": 4.795967354776765e-06,
      "loss": 0.1467,
      "step": 31670
    },
    {
      "epoch": 7.6,
      "learning_rate": 4.791166586653865e-06,
      "loss": 0.123,
      "step": 31680
    },
    {
      "epoch": 7.61,
      "learning_rate": 4.786365818530965e-06,
      "loss": 0.0073,
      "step": 31690
    },
    {
      "epoch": 7.61,
      "learning_rate": 4.781565050408066e-06,
      "loss": 0.0458,
      "step": 31700
    },
    {
      "epoch": 7.61,
      "learning_rate": 4.776764282285166e-06,
      "loss": 0.0539,
      "step": 31710
    },
    {
      "epoch": 7.61,
      "learning_rate": 4.771963514162266e-06,
      "loss": 0.2896,
      "step": 31720
    },
    {
      "epoch": 7.62,
      "learning_rate": 4.767162746039367e-06,
      "loss": 0.1829,
      "step": 31730
    },
    {
      "epoch": 7.62,
      "learning_rate": 4.762361977916467e-06,
      "loss": 0.0767,
      "step": 31740
    },
    {
      "epoch": 7.62,
      "learning_rate": 4.757561209793568e-06,
      "loss": 0.2413,
      "step": 31750
    },
    {
      "epoch": 7.62,
      "learning_rate": 4.7527604416706676e-06,
      "loss": 0.0629,
      "step": 31760
    },
    {
      "epoch": 7.63,
      "learning_rate": 4.7479596735477676e-06,
      "loss": 0.0834,
      "step": 31770
    },
    {
      "epoch": 7.63,
      "learning_rate": 4.743158905424868e-06,
      "loss": 0.1614,
      "step": 31780
    },
    {
      "epoch": 7.63,
      "learning_rate": 4.738358137301968e-06,
      "loss": 0.1813,
      "step": 31790
    },
    {
      "epoch": 7.63,
      "learning_rate": 4.733557369179069e-06,
      "loss": 0.1669,
      "step": 31800
    },
    {
      "epoch": 7.64,
      "learning_rate": 4.728756601056169e-06,
      "loss": 0.1585,
      "step": 31810
    },
    {
      "epoch": 7.64,
      "learning_rate": 4.723955832933269e-06,
      "loss": 0.0119,
      "step": 31820
    },
    {
      "epoch": 7.64,
      "learning_rate": 4.71915506481037e-06,
      "loss": 0.0517,
      "step": 31830
    },
    {
      "epoch": 7.64,
      "learning_rate": 4.714354296687471e-06,
      "loss": 0.1613,
      "step": 31840
    },
    {
      "epoch": 7.65,
      "learning_rate": 4.70955352856457e-06,
      "loss": 0.1125,
      "step": 31850
    },
    {
      "epoch": 7.65,
      "learning_rate": 4.704752760441671e-06,
      "loss": 0.1379,
      "step": 31860
    },
    {
      "epoch": 7.65,
      "learning_rate": 4.699951992318772e-06,
      "loss": 0.1726,
      "step": 31870
    },
    {
      "epoch": 7.65,
      "learning_rate": 4.695151224195872e-06,
      "loss": 0.1147,
      "step": 31880
    },
    {
      "epoch": 7.65,
      "learning_rate": 4.690350456072972e-06,
      "loss": 0.2184,
      "step": 31890
    },
    {
      "epoch": 7.66,
      "learning_rate": 4.685549687950073e-06,
      "loss": 0.1497,
      "step": 31900
    },
    {
      "epoch": 7.66,
      "learning_rate": 4.680748919827173e-06,
      "loss": 0.1905,
      "step": 31910
    },
    {
      "epoch": 7.66,
      "learning_rate": 4.6759481517042734e-06,
      "loss": 0.203,
      "step": 31920
    },
    {
      "epoch": 7.66,
      "learning_rate": 4.6711473835813734e-06,
      "loss": 0.2244,
      "step": 31930
    },
    {
      "epoch": 7.67,
      "learning_rate": 4.6663466154584734e-06,
      "loss": 0.0683,
      "step": 31940
    },
    {
      "epoch": 7.67,
      "learning_rate": 4.661545847335574e-06,
      "loss": 0.0788,
      "step": 31950
    },
    {
      "epoch": 7.67,
      "learning_rate": 4.656745079212674e-06,
      "loss": 0.0052,
      "step": 31960
    },
    {
      "epoch": 7.67,
      "learning_rate": 4.651944311089775e-06,
      "loss": 0.2374,
      "step": 31970
    },
    {
      "epoch": 7.68,
      "learning_rate": 4.647143542966875e-06,
      "loss": 0.1665,
      "step": 31980
    },
    {
      "epoch": 7.68,
      "learning_rate": 4.642342774843975e-06,
      "loss": 0.1838,
      "step": 31990
    },
    {
      "epoch": 7.68,
      "learning_rate": 4.637542006721076e-06,
      "loss": 0.114,
      "step": 32000
    },
    {
      "epoch": 7.68,
      "learning_rate": 4.632741238598176e-06,
      "loss": 0.0497,
      "step": 32010
    },
    {
      "epoch": 7.69,
      "learning_rate": 4.627940470475276e-06,
      "loss": 0.1016,
      "step": 32020
    },
    {
      "epoch": 7.69,
      "learning_rate": 4.623139702352377e-06,
      "loss": 0.0612,
      "step": 32030
    },
    {
      "epoch": 7.69,
      "learning_rate": 4.618338934229477e-06,
      "loss": 0.0497,
      "step": 32040
    },
    {
      "epoch": 7.69,
      "learning_rate": 4.613538166106578e-06,
      "loss": 0.2271,
      "step": 32050
    },
    {
      "epoch": 7.7,
      "learning_rate": 4.608737397983678e-06,
      "loss": 0.1026,
      "step": 32060
    },
    {
      "epoch": 7.7,
      "learning_rate": 4.603936629860778e-06,
      "loss": 0.2153,
      "step": 32070
    },
    {
      "epoch": 7.7,
      "learning_rate": 4.5991358617378785e-06,
      "loss": 0.1663,
      "step": 32080
    },
    {
      "epoch": 7.7,
      "learning_rate": 4.5943350936149785e-06,
      "loss": 0.0797,
      "step": 32090
    },
    {
      "epoch": 7.71,
      "learning_rate": 4.589534325492079e-06,
      "loss": 0.1696,
      "step": 32100
    },
    {
      "epoch": 7.71,
      "learning_rate": 4.584733557369179e-06,
      "loss": 0.2904,
      "step": 32110
    },
    {
      "epoch": 7.71,
      "learning_rate": 4.579932789246279e-06,
      "loss": 0.0849,
      "step": 32120
    },
    {
      "epoch": 7.71,
      "learning_rate": 4.57513202112338e-06,
      "loss": 0.169,
      "step": 32130
    },
    {
      "epoch": 7.71,
      "learning_rate": 4.570331253000481e-06,
      "loss": 0.1106,
      "step": 32140
    },
    {
      "epoch": 7.72,
      "learning_rate": 4.56553048487758e-06,
      "loss": 0.0145,
      "step": 32150
    },
    {
      "epoch": 7.72,
      "learning_rate": 4.560729716754681e-06,
      "loss": 0.1161,
      "step": 32160
    },
    {
      "epoch": 7.72,
      "learning_rate": 4.555928948631782e-06,
      "loss": 0.1767,
      "step": 32170
    },
    {
      "epoch": 7.72,
      "learning_rate": 4.551128180508882e-06,
      "loss": 0.1485,
      "step": 32180
    },
    {
      "epoch": 7.73,
      "learning_rate": 4.546327412385982e-06,
      "loss": 0.0946,
      "step": 32190
    },
    {
      "epoch": 7.73,
      "learning_rate": 4.541526644263083e-06,
      "loss": 0.089,
      "step": 32200
    },
    {
      "epoch": 7.73,
      "learning_rate": 4.536725876140183e-06,
      "loss": 0.1676,
      "step": 32210
    },
    {
      "epoch": 7.73,
      "learning_rate": 4.5319251080172835e-06,
      "loss": 0.0497,
      "step": 32220
    },
    {
      "epoch": 7.74,
      "learning_rate": 4.5271243398943835e-06,
      "loss": 0.0922,
      "step": 32230
    },
    {
      "epoch": 7.74,
      "learning_rate": 4.5223235717714835e-06,
      "loss": 0.1706,
      "step": 32240
    },
    {
      "epoch": 7.74,
      "learning_rate": 4.517522803648584e-06,
      "loss": 0.0641,
      "step": 32250
    },
    {
      "epoch": 7.74,
      "learning_rate": 4.512722035525684e-06,
      "loss": 0.2167,
      "step": 32260
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.507921267402785e-06,
      "loss": 0.2173,
      "step": 32270
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.503120499279885e-06,
      "loss": 0.2801,
      "step": 32280
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.498319731156985e-06,
      "loss": 0.0967,
      "step": 32290
    },
    {
      "epoch": 7.75,
      "learning_rate": 4.493518963034086e-06,
      "loss": 0.0357,
      "step": 32300
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.488718194911186e-06,
      "loss": 0.0833,
      "step": 32310
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.483917426788286e-06,
      "loss": 0.1514,
      "step": 32320
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.479116658665387e-06,
      "loss": 0.1501,
      "step": 32330
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.474315890542487e-06,
      "loss": 0.0309,
      "step": 32340
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.469515122419588e-06,
      "loss": 0.2215,
      "step": 32350
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.464714354296688e-06,
      "loss": 0.0584,
      "step": 32360
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.459913586173788e-06,
      "loss": 0.1679,
      "step": 32370
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.4551128180508885e-06,
      "loss": 0.0158,
      "step": 32380
    },
    {
      "epoch": 7.77,
      "learning_rate": 4.450312049927989e-06,
      "loss": 0.1857,
      "step": 32390
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.445511281805089e-06,
      "loss": 0.1,
      "step": 32400
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.440710513682189e-06,
      "loss": 0.0909,
      "step": 32410
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.43590974555929e-06,
      "loss": 0.1199,
      "step": 32420
    },
    {
      "epoch": 7.78,
      "learning_rate": 4.43110897743639e-06,
      "loss": 0.0755,
      "step": 32430
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.426308209313491e-06,
      "loss": 0.068,
      "step": 32440
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.421507441190591e-06,
      "loss": 0.0868,
      "step": 32450
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.416706673067691e-06,
      "loss": 0.1958,
      "step": 32460
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.411905904944792e-06,
      "loss": 0.169,
      "step": 32470
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.407105136821892e-06,
      "loss": 0.1028,
      "step": 32480
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.402304368698993e-06,
      "loss": 0.0437,
      "step": 32490
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.397503600576093e-06,
      "loss": 0.1739,
      "step": 32500
    },
    {
      "epoch": 7.8,
      "learning_rate": 4.392702832453193e-06,
      "loss": 0.0676,
      "step": 32510
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.3879020643302935e-06,
      "loss": 0.0551,
      "step": 32520
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.3831012962073935e-06,
      "loss": 0.1959,
      "step": 32530
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.3783005280844935e-06,
      "loss": 0.02,
      "step": 32540
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.373499759961594e-06,
      "loss": 0.1355,
      "step": 32550
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.368698991838694e-06,
      "loss": 0.2029,
      "step": 32560
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.363898223715795e-06,
      "loss": 0.052,
      "step": 32570
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.359097455592895e-06,
      "loss": 0.1296,
      "step": 32580
    },
    {
      "epoch": 7.82,
      "learning_rate": 4.354296687469995e-06,
      "loss": 0.0444,
      "step": 32590
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.349495919347096e-06,
      "loss": 0.0891,
      "step": 32600
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.344695151224196e-06,
      "loss": 0.0459,
      "step": 32610
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.339894383101296e-06,
      "loss": 0.1198,
      "step": 32620
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.335093614978397e-06,
      "loss": 0.1227,
      "step": 32630
    },
    {
      "epoch": 7.83,
      "learning_rate": 4.330292846855497e-06,
      "loss": 0.0715,
      "step": 32640
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.325492078732598e-06,
      "loss": 0.1631,
      "step": 32650
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.3206913106096986e-06,
      "loss": 0.0663,
      "step": 32660
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.315890542486798e-06,
      "loss": 0.0343,
      "step": 32670
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.3110897743638986e-06,
      "loss": 0.1073,
      "step": 32680
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.306289006240999e-06,
      "loss": 0.0775,
      "step": 32690
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.301488238118099e-06,
      "loss": 0.1082,
      "step": 32700
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.296687469995199e-06,
      "loss": 0.0996,
      "step": 32710
    },
    {
      "epoch": 7.85,
      "learning_rate": 4.2918867018723e-06,
      "loss": 0.2619,
      "step": 32720
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.2870859337494e-06,
      "loss": 0.1357,
      "step": 32730
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.282285165626501e-06,
      "loss": 0.1922,
      "step": 32740
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.277484397503601e-06,
      "loss": 0.01,
      "step": 32750
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.272683629380701e-06,
      "loss": 0.1439,
      "step": 32760
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.267882861257802e-06,
      "loss": 0.1285,
      "step": 32770
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.263082093134902e-06,
      "loss": 0.0881,
      "step": 32780
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.258281325012003e-06,
      "loss": 0.034,
      "step": 32790
    },
    {
      "epoch": 7.87,
      "learning_rate": 4.253480556889103e-06,
      "loss": 0.1083,
      "step": 32800
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.248679788766203e-06,
      "loss": 0.2431,
      "step": 32810
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.243879020643304e-06,
      "loss": 0.1309,
      "step": 32820
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.239078252520404e-06,
      "loss": 0.0219,
      "step": 32830
    },
    {
      "epoch": 7.88,
      "learning_rate": 4.2342774843975036e-06,
      "loss": 0.0618,
      "step": 32840
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.229476716274604e-06,
      "loss": 0.311,
      "step": 32850
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.224675948151704e-06,
      "loss": 0.1233,
      "step": 32860
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.219875180028805e-06,
      "loss": 0.0769,
      "step": 32870
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.215074411905905e-06,
      "loss": 0.0651,
      "step": 32880
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.210273643783005e-06,
      "loss": 0.109,
      "step": 32890
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.205472875660106e-06,
      "loss": 0.0559,
      "step": 32900
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.200672107537206e-06,
      "loss": 0.263,
      "step": 32910
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.195871339414306e-06,
      "loss": 0.1216,
      "step": 32920
    },
    {
      "epoch": 7.9,
      "learning_rate": 4.191070571291407e-06,
      "loss": 0.1233,
      "step": 32930
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.186269803168507e-06,
      "loss": 0.1657,
      "step": 32940
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.181469035045608e-06,
      "loss": 0.1153,
      "step": 32950
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.176668266922709e-06,
      "loss": 0.0072,
      "step": 32960
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.171867498799808e-06,
      "loss": 0.1707,
      "step": 32970
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.167066730676909e-06,
      "loss": 0.064,
      "step": 32980
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.1622659625540094e-06,
      "loss": 0.2296,
      "step": 32990
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.1574651944311094e-06,
      "loss": 0.0409,
      "step": 33000
    },
    {
      "epoch": 7.92,
      "learning_rate": 4.1526644263082094e-06,
      "loss": 0.0928,
      "step": 33010
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.14786365818531e-06,
      "loss": 0.2015,
      "step": 33020
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.14306289006241e-06,
      "loss": 0.1539,
      "step": 33030
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.138262121939511e-06,
      "loss": 0.033,
      "step": 33040
    },
    {
      "epoch": 7.93,
      "learning_rate": 4.133461353816611e-06,
      "loss": 0.1806,
      "step": 33050
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.128660585693711e-06,
      "loss": 0.0404,
      "step": 33060
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.123859817570812e-06,
      "loss": 0.0788,
      "step": 33070
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.119059049447912e-06,
      "loss": 0.0049,
      "step": 33080
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.114258281325013e-06,
      "loss": 0.1674,
      "step": 33090
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.109457513202113e-06,
      "loss": 0.1119,
      "step": 33100
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.104656745079213e-06,
      "loss": 0.1309,
      "step": 33110
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.099855976956314e-06,
      "loss": 0.153,
      "step": 33120
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.095055208833414e-06,
      "loss": 0.0916,
      "step": 33130
    },
    {
      "epoch": 7.95,
      "learning_rate": 4.090254440710514e-06,
      "loss": 0.0981,
      "step": 33140
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.0854536725876145e-06,
      "loss": 0.1299,
      "step": 33150
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.0806529044647145e-06,
      "loss": 0.078,
      "step": 33160
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.075852136341815e-06,
      "loss": 0.179,
      "step": 33170
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.071051368218915e-06,
      "loss": 0.079,
      "step": 33180
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.066250600096015e-06,
      "loss": 0.0038,
      "step": 33190
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.061449831973116e-06,
      "loss": 0.1786,
      "step": 33200
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.056649063850216e-06,
      "loss": 0.1366,
      "step": 33210
    },
    {
      "epoch": 7.97,
      "learning_rate": 4.051848295727317e-06,
      "loss": 0.0247,
      "step": 33220
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.047047527604417e-06,
      "loss": 0.2032,
      "step": 33230
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.042246759481517e-06,
      "loss": 0.0918,
      "step": 33240
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.037445991358618e-06,
      "loss": 0.0276,
      "step": 33250
    },
    {
      "epoch": 7.98,
      "learning_rate": 4.032645223235719e-06,
      "loss": 0.11,
      "step": 33260
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.027844455112818e-06,
      "loss": 0.0286,
      "step": 33270
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.023043686989919e-06,
      "loss": 0.0939,
      "step": 33280
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.0182429188670195e-06,
      "loss": 0.2662,
      "step": 33290
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.0134421507441195e-06,
      "loss": 0.2048,
      "step": 33300
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.0086413826212195e-06,
      "loss": 0.2249,
      "step": 33310
    },
    {
      "epoch": 8.0,
      "learning_rate": 4.00384061449832e-06,
      "loss": 0.0983,
      "step": 33320
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.9989799387963277,
      "eval_loss": 0.005884047131985426,
      "eval_runtime": 517.9675,
      "eval_samples_per_second": 11.356,
      "eval_steps_per_second": 1.421,
      "step": 33328
    },
    {
      "epoch": 8.0,
      "learning_rate": 3.99903984637542e-06,
      "loss": 0.1972,
      "step": 33330
    },
    {
      "epoch": 8.0,
      "learning_rate": 3.994239078252521e-06,
      "loss": 0.0332,
      "step": 33340
    },
    {
      "epoch": 8.01,
      "learning_rate": 3.989438310129621e-06,
      "loss": 0.1043,
      "step": 33350
    },
    {
      "epoch": 8.01,
      "learning_rate": 3.984637542006721e-06,
      "loss": 0.2772,
      "step": 33360
    },
    {
      "epoch": 8.01,
      "learning_rate": 3.979836773883822e-06,
      "loss": 0.1198,
      "step": 33370
    },
    {
      "epoch": 8.01,
      "learning_rate": 3.975036005760922e-06,
      "loss": 0.0784,
      "step": 33380
    },
    {
      "epoch": 8.01,
      "learning_rate": 3.970235237638023e-06,
      "loss": 0.064,
      "step": 33390
    },
    {
      "epoch": 8.02,
      "learning_rate": 3.965434469515123e-06,
      "loss": 0.1932,
      "step": 33400
    },
    {
      "epoch": 8.02,
      "learning_rate": 3.960633701392223e-06,
      "loss": 0.0251,
      "step": 33410
    },
    {
      "epoch": 8.02,
      "learning_rate": 3.955832933269324e-06,
      "loss": 0.1158,
      "step": 33420
    },
    {
      "epoch": 8.02,
      "learning_rate": 3.951032165146424e-06,
      "loss": 0.1333,
      "step": 33430
    },
    {
      "epoch": 8.03,
      "learning_rate": 3.946231397023524e-06,
      "loss": 0.1075,
      "step": 33440
    },
    {
      "epoch": 8.03,
      "learning_rate": 3.9414306289006245e-06,
      "loss": 0.1348,
      "step": 33450
    },
    {
      "epoch": 8.03,
      "learning_rate": 3.9366298607777245e-06,
      "loss": 0.2771,
      "step": 33460
    },
    {
      "epoch": 8.03,
      "learning_rate": 3.931829092654825e-06,
      "loss": 0.0497,
      "step": 33470
    },
    {
      "epoch": 8.04,
      "learning_rate": 3.927028324531925e-06,
      "loss": 0.125,
      "step": 33480
    },
    {
      "epoch": 8.04,
      "learning_rate": 3.922227556409025e-06,
      "loss": 0.1646,
      "step": 33490
    },
    {
      "epoch": 8.04,
      "learning_rate": 3.917426788286126e-06,
      "loss": 0.0214,
      "step": 33500
    },
    {
      "epoch": 8.04,
      "learning_rate": 3.912626020163226e-06,
      "loss": 0.221,
      "step": 33510
    },
    {
      "epoch": 8.05,
      "learning_rate": 3.907825252040327e-06,
      "loss": 0.1701,
      "step": 33520
    },
    {
      "epoch": 8.05,
      "learning_rate": 3.903024483917427e-06,
      "loss": 0.1275,
      "step": 33530
    },
    {
      "epoch": 8.05,
      "learning_rate": 3.898223715794527e-06,
      "loss": 0.1893,
      "step": 33540
    },
    {
      "epoch": 8.05,
      "learning_rate": 3.893422947671628e-06,
      "loss": 0.0999,
      "step": 33550
    },
    {
      "epoch": 8.06,
      "learning_rate": 3.888622179548729e-06,
      "loss": 0.0996,
      "step": 33560
    },
    {
      "epoch": 8.06,
      "learning_rate": 3.883821411425828e-06,
      "loss": 0.0339,
      "step": 33570
    },
    {
      "epoch": 8.06,
      "learning_rate": 3.879020643302929e-06,
      "loss": 0.0447,
      "step": 33580
    },
    {
      "epoch": 8.06,
      "learning_rate": 3.8742198751800295e-06,
      "loss": 0.2243,
      "step": 33590
    },
    {
      "epoch": 8.07,
      "learning_rate": 3.8694191070571295e-06,
      "loss": 0.2021,
      "step": 33600
    },
    {
      "epoch": 8.07,
      "learning_rate": 3.8646183389342295e-06,
      "loss": 0.0495,
      "step": 33610
    },
    {
      "epoch": 8.07,
      "learning_rate": 3.85981757081133e-06,
      "loss": 0.1671,
      "step": 33620
    },
    {
      "epoch": 8.07,
      "learning_rate": 3.85501680268843e-06,
      "loss": 0.3138,
      "step": 33630
    },
    {
      "epoch": 8.07,
      "learning_rate": 3.850216034565531e-06,
      "loss": 0.1293,
      "step": 33640
    },
    {
      "epoch": 8.08,
      "learning_rate": 3.845415266442631e-06,
      "loss": 0.0642,
      "step": 33650
    },
    {
      "epoch": 8.08,
      "learning_rate": 3.840614498319731e-06,
      "loss": 0.0912,
      "step": 33660
    },
    {
      "epoch": 8.08,
      "learning_rate": 3.835813730196832e-06,
      "loss": 0.048,
      "step": 33670
    },
    {
      "epoch": 8.08,
      "learning_rate": 3.831012962073932e-06,
      "loss": 0.0946,
      "step": 33680
    },
    {
      "epoch": 8.09,
      "learning_rate": 3.826212193951033e-06,
      "loss": 0.0444,
      "step": 33690
    },
    {
      "epoch": 8.09,
      "learning_rate": 3.821411425828133e-06,
      "loss": 0.0184,
      "step": 33700
    },
    {
      "epoch": 8.09,
      "learning_rate": 3.816610657705233e-06,
      "loss": 0.3485,
      "step": 33710
    },
    {
      "epoch": 8.09,
      "learning_rate": 3.8118098895823337e-06,
      "loss": 0.0928,
      "step": 33720
    },
    {
      "epoch": 8.1,
      "learning_rate": 3.807009121459434e-06,
      "loss": 0.2039,
      "step": 33730
    },
    {
      "epoch": 8.1,
      "learning_rate": 3.8022083533365337e-06,
      "loss": 0.0653,
      "step": 33740
    },
    {
      "epoch": 8.1,
      "learning_rate": 3.7974075852136346e-06,
      "loss": 0.0541,
      "step": 33750
    },
    {
      "epoch": 8.1,
      "learning_rate": 3.792606817090735e-06,
      "loss": 0.0809,
      "step": 33760
    },
    {
      "epoch": 8.11,
      "learning_rate": 3.7878060489678354e-06,
      "loss": 0.0341,
      "step": 33770
    },
    {
      "epoch": 8.11,
      "learning_rate": 3.7830052808449354e-06,
      "loss": 0.1706,
      "step": 33780
    },
    {
      "epoch": 8.11,
      "learning_rate": 3.778204512722036e-06,
      "loss": 0.2115,
      "step": 33790
    },
    {
      "epoch": 8.11,
      "learning_rate": 3.7734037445991362e-06,
      "loss": 0.3165,
      "step": 33800
    },
    {
      "epoch": 8.12,
      "learning_rate": 3.7686029764762367e-06,
      "loss": 0.0026,
      "step": 33810
    },
    {
      "epoch": 8.12,
      "learning_rate": 3.763802208353337e-06,
      "loss": 0.1435,
      "step": 33820
    },
    {
      "epoch": 8.12,
      "learning_rate": 3.759001440230437e-06,
      "loss": 0.1226,
      "step": 33830
    },
    {
      "epoch": 8.12,
      "learning_rate": 3.7542006721075375e-06,
      "loss": 0.1266,
      "step": 33840
    },
    {
      "epoch": 8.13,
      "learning_rate": 3.749399903984638e-06,
      "loss": 0.1283,
      "step": 33850
    },
    {
      "epoch": 8.13,
      "learning_rate": 3.7445991358617383e-06,
      "loss": 0.1462,
      "step": 33860
    },
    {
      "epoch": 8.13,
      "learning_rate": 3.7397983677388383e-06,
      "loss": 0.2587,
      "step": 33870
    },
    {
      "epoch": 8.13,
      "learning_rate": 3.7349975996159387e-06,
      "loss": 0.0669,
      "step": 33880
    },
    {
      "epoch": 8.13,
      "learning_rate": 3.730196831493039e-06,
      "loss": 0.0769,
      "step": 33890
    },
    {
      "epoch": 8.14,
      "learning_rate": 3.7253960633701396e-06,
      "loss": 0.0891,
      "step": 33900
    },
    {
      "epoch": 8.14,
      "learning_rate": 3.7205952952472396e-06,
      "loss": 0.2469,
      "step": 33910
    },
    {
      "epoch": 8.14,
      "learning_rate": 3.71579452712434e-06,
      "loss": 0.0637,
      "step": 33920
    },
    {
      "epoch": 8.14,
      "learning_rate": 3.7109937590014404e-06,
      "loss": 0.1237,
      "step": 33930
    },
    {
      "epoch": 8.15,
      "learning_rate": 3.706192990878541e-06,
      "loss": 0.0574,
      "step": 33940
    },
    {
      "epoch": 8.15,
      "learning_rate": 3.7013922227556413e-06,
      "loss": 0.125,
      "step": 33950
    },
    {
      "epoch": 8.15,
      "learning_rate": 3.6965914546327413e-06,
      "loss": 0.2603,
      "step": 33960
    },
    {
      "epoch": 8.15,
      "learning_rate": 3.6917906865098417e-06,
      "loss": 0.0585,
      "step": 33970
    },
    {
      "epoch": 8.16,
      "learning_rate": 3.6869899183869425e-06,
      "loss": 0.0952,
      "step": 33980
    },
    {
      "epoch": 8.16,
      "learning_rate": 3.682189150264043e-06,
      "loss": 0.1033,
      "step": 33990
    },
    {
      "epoch": 8.16,
      "learning_rate": 3.6773883821411425e-06,
      "loss": 0.0956,
      "step": 34000
    },
    {
      "epoch": 8.16,
      "learning_rate": 3.6725876140182434e-06,
      "loss": 0.1679,
      "step": 34010
    },
    {
      "epoch": 8.17,
      "learning_rate": 3.6677868458953438e-06,
      "loss": 0.0395,
      "step": 34020
    },
    {
      "epoch": 8.17,
      "learning_rate": 3.662986077772444e-06,
      "loss": 0.063,
      "step": 34030
    },
    {
      "epoch": 8.17,
      "learning_rate": 3.658185309649544e-06,
      "loss": 0.0922,
      "step": 34040
    },
    {
      "epoch": 8.17,
      "learning_rate": 3.6533845415266446e-06,
      "loss": 0.0746,
      "step": 34050
    },
    {
      "epoch": 8.18,
      "learning_rate": 3.648583773403745e-06,
      "loss": 0.1837,
      "step": 34060
    },
    {
      "epoch": 8.18,
      "learning_rate": 3.6437830052808455e-06,
      "loss": 0.1759,
      "step": 34070
    },
    {
      "epoch": 8.18,
      "learning_rate": 3.6389822371579454e-06,
      "loss": 0.0915,
      "step": 34080
    },
    {
      "epoch": 8.18,
      "learning_rate": 3.634181469035046e-06,
      "loss": 0.1557,
      "step": 34090
    },
    {
      "epoch": 8.19,
      "learning_rate": 3.6293807009121463e-06,
      "loss": 0.1361,
      "step": 34100
    },
    {
      "epoch": 8.19,
      "learning_rate": 3.6245799327892467e-06,
      "loss": 0.0635,
      "step": 34110
    },
    {
      "epoch": 8.19,
      "learning_rate": 3.619779164666347e-06,
      "loss": 0.0581,
      "step": 34120
    },
    {
      "epoch": 8.19,
      "learning_rate": 3.614978396543447e-06,
      "loss": 0.0995,
      "step": 34130
    },
    {
      "epoch": 8.19,
      "learning_rate": 3.6101776284205475e-06,
      "loss": 0.0327,
      "step": 34140
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.605376860297648e-06,
      "loss": 0.1152,
      "step": 34150
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.6005760921747484e-06,
      "loss": 0.3173,
      "step": 34160
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.5957753240518484e-06,
      "loss": 0.0716,
      "step": 34170
    },
    {
      "epoch": 8.2,
      "learning_rate": 3.590974555928949e-06,
      "loss": 0.0701,
      "step": 34180
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.5861737878060492e-06,
      "loss": 0.2868,
      "step": 34190
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.5813730196831496e-06,
      "loss": 0.203,
      "step": 34200
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.5765722515602496e-06,
      "loss": 0.0392,
      "step": 34210
    },
    {
      "epoch": 8.21,
      "learning_rate": 3.57177148343735e-06,
      "loss": 0.0982,
      "step": 34220
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.5669707153144505e-06,
      "loss": 0.0443,
      "step": 34230
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.562169947191551e-06,
      "loss": 0.11,
      "step": 34240
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.5573691790686517e-06,
      "loss": 0.1083,
      "step": 34250
    },
    {
      "epoch": 8.22,
      "learning_rate": 3.5525684109457513e-06,
      "loss": 0.067,
      "step": 34260
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.5477676428228517e-06,
      "loss": 0.0604,
      "step": 34270
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.5429668746999526e-06,
      "loss": 0.1665,
      "step": 34280
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.538166106577053e-06,
      "loss": 0.065,
      "step": 34290
    },
    {
      "epoch": 8.23,
      "learning_rate": 3.5333653384541526e-06,
      "loss": 0.1961,
      "step": 34300
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.5285645703312534e-06,
      "loss": 0.1178,
      "step": 34310
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.523763802208354e-06,
      "loss": 0.1627,
      "step": 34320
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.5189630340854542e-06,
      "loss": 0.1551,
      "step": 34330
    },
    {
      "epoch": 8.24,
      "learning_rate": 3.5141622659625542e-06,
      "loss": 0.0533,
      "step": 34340
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.5093614978396547e-06,
      "loss": 0.2665,
      "step": 34350
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.504560729716755e-06,
      "loss": 0.0695,
      "step": 34360
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.4997599615938555e-06,
      "loss": 0.2935,
      "step": 34370
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.494959193470956e-06,
      "loss": 0.0365,
      "step": 34380
    },
    {
      "epoch": 8.25,
      "learning_rate": 3.490158425348056e-06,
      "loss": 0.0689,
      "step": 34390
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.4853576572251563e-06,
      "loss": 0.1108,
      "step": 34400
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.4805568891022568e-06,
      "loss": 0.0311,
      "step": 34410
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.475756120979357e-06,
      "loss": 0.1035,
      "step": 34420
    },
    {
      "epoch": 8.26,
      "learning_rate": 3.470955352856457e-06,
      "loss": 0.1511,
      "step": 34430
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.4661545847335576e-06,
      "loss": 0.2384,
      "step": 34440
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.461353816610658e-06,
      "loss": 0.0966,
      "step": 34450
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.4565530484877584e-06,
      "loss": 0.165,
      "step": 34460
    },
    {
      "epoch": 8.27,
      "learning_rate": 3.4517522803648584e-06,
      "loss": 0.139,
      "step": 34470
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.446951512241959e-06,
      "loss": 0.041,
      "step": 34480
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.4421507441190593e-06,
      "loss": 0.0959,
      "step": 34490
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.4373499759961597e-06,
      "loss": 0.1206,
      "step": 34500
    },
    {
      "epoch": 8.28,
      "learning_rate": 3.4325492078732597e-06,
      "loss": 0.2708,
      "step": 34510
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.42774843975036e-06,
      "loss": 0.1659,
      "step": 34520
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.4229476716274605e-06,
      "loss": 0.1849,
      "step": 34530
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.418146903504561e-06,
      "loss": 0.0601,
      "step": 34540
    },
    {
      "epoch": 8.29,
      "learning_rate": 3.4133461353816618e-06,
      "loss": 0.1162,
      "step": 34550
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.4085453672587614e-06,
      "loss": 0.1731,
      "step": 34560
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.4037445991358618e-06,
      "loss": 0.1586,
      "step": 34570
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.3989438310129626e-06,
      "loss": 0.0555,
      "step": 34580
    },
    {
      "epoch": 8.3,
      "learning_rate": 3.394143062890063e-06,
      "loss": 0.1953,
      "step": 34590
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.3893422947671626e-06,
      "loss": 0.0576,
      "step": 34600
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.3845415266442635e-06,
      "loss": 0.1094,
      "step": 34610
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.379740758521364e-06,
      "loss": 0.0029,
      "step": 34620
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.3749399903984643e-06,
      "loss": 0.1093,
      "step": 34630
    },
    {
      "epoch": 8.31,
      "learning_rate": 3.3701392222755643e-06,
      "loss": 0.0386,
      "step": 34640
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.3653384541526647e-06,
      "loss": 0.2095,
      "step": 34650
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.360537686029765e-06,
      "loss": 0.1846,
      "step": 34660
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.3557369179068655e-06,
      "loss": 0.1879,
      "step": 34670
    },
    {
      "epoch": 8.32,
      "learning_rate": 3.350936149783966e-06,
      "loss": 0.1359,
      "step": 34680
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.346135381661066e-06,
      "loss": 0.1777,
      "step": 34690
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.3413346135381664e-06,
      "loss": 0.0897,
      "step": 34700
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.336533845415267e-06,
      "loss": 0.2725,
      "step": 34710
    },
    {
      "epoch": 8.33,
      "learning_rate": 3.3317330772923672e-06,
      "loss": 0.1332,
      "step": 34720
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.3269323091694672e-06,
      "loss": 0.114,
      "step": 34730
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.3221315410465676e-06,
      "loss": 0.2266,
      "step": 34740
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.317330772923668e-06,
      "loss": 0.0507,
      "step": 34750
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.3125300048007685e-06,
      "loss": 0.127,
      "step": 34760
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.3077292366778685e-06,
      "loss": 0.1693,
      "step": 34770
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.302928468554969e-06,
      "loss": 0.2422,
      "step": 34780
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.2981277004320693e-06,
      "loss": 0.2181,
      "step": 34790
    },
    {
      "epoch": 8.35,
      "learning_rate": 3.2933269323091697e-06,
      "loss": 0.0957,
      "step": 34800
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.2885261641862697e-06,
      "loss": 0.0768,
      "step": 34810
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.28372539606337e-06,
      "loss": 0.0727,
      "step": 34820
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.2789246279404706e-06,
      "loss": 0.107,
      "step": 34830
    },
    {
      "epoch": 8.36,
      "learning_rate": 3.274123859817571e-06,
      "loss": 0.0183,
      "step": 34840
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.269323091694672e-06,
      "loss": 0.1119,
      "step": 34850
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.2645223235717714e-06,
      "loss": 0.0896,
      "step": 34860
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.2597215554488722e-06,
      "loss": 0.013,
      "step": 34870
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.2549207873259727e-06,
      "loss": 0.1023,
      "step": 34880
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.250120019203073e-06,
      "loss": 0.0438,
      "step": 34890
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.245319251080173e-06,
      "loss": 0.1528,
      "step": 34900
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.2405184829572735e-06,
      "loss": 0.0306,
      "step": 34910
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.235717714834374e-06,
      "loss": 0.1335,
      "step": 34920
    },
    {
      "epoch": 8.38,
      "learning_rate": 3.2309169467114743e-06,
      "loss": 0.1406,
      "step": 34930
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.2261161785885743e-06,
      "loss": 0.09,
      "step": 34940
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.2213154104656748e-06,
      "loss": 0.2012,
      "step": 34950
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.216514642342775e-06,
      "loss": 0.1888,
      "step": 34960
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.2117138742198756e-06,
      "loss": 0.126,
      "step": 34970
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.206913106096976e-06,
      "loss": 0.1207,
      "step": 34980
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.202112337974076e-06,
      "loss": 0.1189,
      "step": 34990
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.1973115698511764e-06,
      "loss": 0.1249,
      "step": 35000
    },
    {
      "epoch": 8.4,
      "learning_rate": 3.192510801728277e-06,
      "loss": 0.0334,
      "step": 35010
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.1877100336053773e-06,
      "loss": 0.0516,
      "step": 35020
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.1829092654824773e-06,
      "loss": 0.0101,
      "step": 35030
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.1781084973595777e-06,
      "loss": 0.1285,
      "step": 35040
    },
    {
      "epoch": 8.41,
      "learning_rate": 3.173307729236678e-06,
      "loss": 0.0358,
      "step": 35050
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.1685069611137785e-06,
      "loss": 0.1,
      "step": 35060
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.1637061929908785e-06,
      "loss": 0.0185,
      "step": 35070
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.158905424867979e-06,
      "loss": 0.0182,
      "step": 35080
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.1541046567450794e-06,
      "loss": 0.0885,
      "step": 35090
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.1493038886221798e-06,
      "loss": 0.2235,
      "step": 35100
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.1445031204992806e-06,
      "loss": 0.0725,
      "step": 35110
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.13970235237638e-06,
      "loss": 0.0892,
      "step": 35120
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.1349015842534806e-06,
      "loss": 0.272,
      "step": 35130
    },
    {
      "epoch": 8.43,
      "learning_rate": 3.1301008161305815e-06,
      "loss": 0.075,
      "step": 35140
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.125300048007682e-06,
      "loss": 0.0485,
      "step": 35150
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.1204992798847814e-06,
      "loss": 0.1188,
      "step": 35160
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.1156985117618823e-06,
      "loss": 0.0818,
      "step": 35170
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.1108977436389827e-06,
      "loss": 0.0898,
      "step": 35180
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.106096975516083e-06,
      "loss": 0.1345,
      "step": 35190
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.101296207393183e-06,
      "loss": 0.0588,
      "step": 35200
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.0964954392702835e-06,
      "loss": 0.0573,
      "step": 35210
    },
    {
      "epoch": 8.45,
      "learning_rate": 3.091694671147384e-06,
      "loss": 0.1107,
      "step": 35220
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.0868939030244844e-06,
      "loss": 0.1779,
      "step": 35230
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.0820931349015844e-06,
      "loss": 0.2117,
      "step": 35240
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.077292366778685e-06,
      "loss": 0.1218,
      "step": 35250
    },
    {
      "epoch": 8.46,
      "learning_rate": 3.0724915986557852e-06,
      "loss": 0.1629,
      "step": 35260
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.0676908305328856e-06,
      "loss": 0.2059,
      "step": 35270
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.062890062409986e-06,
      "loss": 0.1014,
      "step": 35280
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.058089294287086e-06,
      "loss": 0.2194,
      "step": 35290
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.0532885261641865e-06,
      "loss": 0.0733,
      "step": 35300
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.048487758041287e-06,
      "loss": 0.2837,
      "step": 35310
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.0436869899183873e-06,
      "loss": 0.1377,
      "step": 35320
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.0388862217954873e-06,
      "loss": 0.2215,
      "step": 35330
    },
    {
      "epoch": 8.48,
      "learning_rate": 3.0340854536725877e-06,
      "loss": 0.0281,
      "step": 35340
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.029284685549688e-06,
      "loss": 0.0766,
      "step": 35350
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.0244839174267886e-06,
      "loss": 0.1596,
      "step": 35360
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.0196831493038886e-06,
      "loss": 0.1112,
      "step": 35370
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.014882381180989e-06,
      "loss": 0.1629,
      "step": 35380
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.0100816130580894e-06,
      "loss": 0.1476,
      "step": 35390
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.00528084493519e-06,
      "loss": 0.1226,
      "step": 35400
    },
    {
      "epoch": 8.5,
      "learning_rate": 3.0004800768122907e-06,
      "loss": 0.2149,
      "step": 35410
    },
    {
      "epoch": 8.5,
      "learning_rate": 2.9956793086893902e-06,
      "loss": 0.1062,
      "step": 35420
    },
    {
      "epoch": 8.5,
      "learning_rate": 2.9908785405664907e-06,
      "loss": 0.2502,
      "step": 35430
    },
    {
      "epoch": 8.51,
      "learning_rate": 2.9860777724435915e-06,
      "loss": 0.0245,
      "step": 35440
    },
    {
      "epoch": 8.51,
      "learning_rate": 2.981277004320692e-06,
      "loss": 0.1055,
      "step": 35450
    },
    {
      "epoch": 8.51,
      "learning_rate": 2.9764762361977915e-06,
      "loss": 0.0777,
      "step": 35460
    },
    {
      "epoch": 8.51,
      "learning_rate": 2.9716754680748923e-06,
      "loss": 0.1057,
      "step": 35470
    },
    {
      "epoch": 8.52,
      "learning_rate": 2.9668746999519928e-06,
      "loss": 0.0549,
      "step": 35480
    },
    {
      "epoch": 8.52,
      "learning_rate": 2.962073931829093e-06,
      "loss": 0.0987,
      "step": 35490
    },
    {
      "epoch": 8.52,
      "learning_rate": 2.957273163706193e-06,
      "loss": 0.1285,
      "step": 35500
    },
    {
      "epoch": 8.52,
      "learning_rate": 2.9524723955832936e-06,
      "loss": 0.0943,
      "step": 35510
    },
    {
      "epoch": 8.53,
      "learning_rate": 2.947671627460394e-06,
      "loss": 0.121,
      "step": 35520
    },
    {
      "epoch": 8.53,
      "learning_rate": 2.9428708593374944e-06,
      "loss": 0.1275,
      "step": 35530
    },
    {
      "epoch": 8.53,
      "learning_rate": 2.9380700912145944e-06,
      "loss": 0.0021,
      "step": 35540
    },
    {
      "epoch": 8.53,
      "learning_rate": 2.933269323091695e-06,
      "loss": 0.0995,
      "step": 35550
    },
    {
      "epoch": 8.54,
      "learning_rate": 2.9284685549687953e-06,
      "loss": 0.0811,
      "step": 35560
    },
    {
      "epoch": 8.54,
      "learning_rate": 2.9236677868458957e-06,
      "loss": 0.049,
      "step": 35570
    },
    {
      "epoch": 8.54,
      "learning_rate": 2.918867018722996e-06,
      "loss": 0.0044,
      "step": 35580
    },
    {
      "epoch": 8.54,
      "learning_rate": 2.914066250600096e-06,
      "loss": 0.0953,
      "step": 35590
    },
    {
      "epoch": 8.55,
      "learning_rate": 2.9092654824771965e-06,
      "loss": 0.1121,
      "step": 35600
    },
    {
      "epoch": 8.55,
      "learning_rate": 2.904464714354297e-06,
      "loss": 0.1563,
      "step": 35610
    },
    {
      "epoch": 8.55,
      "learning_rate": 2.8996639462313974e-06,
      "loss": 0.0416,
      "step": 35620
    },
    {
      "epoch": 8.55,
      "learning_rate": 2.8948631781084974e-06,
      "loss": 0.2389,
      "step": 35630
    },
    {
      "epoch": 8.55,
      "learning_rate": 2.8900624099855978e-06,
      "loss": 0.1079,
      "step": 35640
    },
    {
      "epoch": 8.56,
      "learning_rate": 2.885261641862698e-06,
      "loss": 0.172,
      "step": 35650
    },
    {
      "epoch": 8.56,
      "learning_rate": 2.8804608737397986e-06,
      "loss": 0.0645,
      "step": 35660
    },
    {
      "epoch": 8.56,
      "learning_rate": 2.8756601056168986e-06,
      "loss": 0.0511,
      "step": 35670
    },
    {
      "epoch": 8.56,
      "learning_rate": 2.870859337493999e-06,
      "loss": 0.0898,
      "step": 35680
    },
    {
      "epoch": 8.57,
      "learning_rate": 2.8660585693710995e-06,
      "loss": 0.0793,
      "step": 35690
    },
    {
      "epoch": 8.57,
      "learning_rate": 2.8612578012482e-06,
      "loss": 0.1187,
      "step": 35700
    },
    {
      "epoch": 8.57,
      "learning_rate": 2.8564570331253007e-06,
      "loss": 0.0558,
      "step": 35710
    },
    {
      "epoch": 8.57,
      "learning_rate": 2.8516562650024003e-06,
      "loss": 0.0742,
      "step": 35720
    },
    {
      "epoch": 8.58,
      "learning_rate": 2.8468554968795007e-06,
      "loss": 0.1792,
      "step": 35730
    },
    {
      "epoch": 8.58,
      "learning_rate": 2.8420547287566016e-06,
      "loss": 0.1642,
      "step": 35740
    },
    {
      "epoch": 8.58,
      "learning_rate": 2.837253960633702e-06,
      "loss": 0.2237,
      "step": 35750
    },
    {
      "epoch": 8.58,
      "learning_rate": 2.8324531925108015e-06,
      "loss": 0.1767,
      "step": 35760
    },
    {
      "epoch": 8.59,
      "learning_rate": 2.8276524243879024e-06,
      "loss": 0.0623,
      "step": 35770
    },
    {
      "epoch": 8.59,
      "learning_rate": 2.822851656265003e-06,
      "loss": 0.0554,
      "step": 35780
    },
    {
      "epoch": 8.59,
      "learning_rate": 2.8180508881421032e-06,
      "loss": 0.0084,
      "step": 35790
    },
    {
      "epoch": 8.59,
      "learning_rate": 2.8132501200192032e-06,
      "loss": 0.2372,
      "step": 35800
    },
    {
      "epoch": 8.6,
      "learning_rate": 2.8084493518963036e-06,
      "loss": 0.0402,
      "step": 35810
    },
    {
      "epoch": 8.6,
      "learning_rate": 2.803648583773404e-06,
      "loss": 0.1268,
      "step": 35820
    },
    {
      "epoch": 8.6,
      "learning_rate": 2.7988478156505045e-06,
      "loss": 0.047,
      "step": 35830
    },
    {
      "epoch": 8.6,
      "learning_rate": 2.794047047527605e-06,
      "loss": 0.0758,
      "step": 35840
    },
    {
      "epoch": 8.61,
      "learning_rate": 2.789246279404705e-06,
      "loss": 0.0898,
      "step": 35850
    },
    {
      "epoch": 8.61,
      "learning_rate": 2.7844455112818053e-06,
      "loss": 0.0381,
      "step": 35860
    },
    {
      "epoch": 8.61,
      "learning_rate": 2.7796447431589057e-06,
      "loss": 0.0702,
      "step": 35870
    },
    {
      "epoch": 8.61,
      "learning_rate": 2.774843975036006e-06,
      "loss": 0.0517,
      "step": 35880
    },
    {
      "epoch": 8.61,
      "learning_rate": 2.770043206913106e-06,
      "loss": 0.1712,
      "step": 35890
    },
    {
      "epoch": 8.62,
      "learning_rate": 2.7652424387902066e-06,
      "loss": 0.109,
      "step": 35900
    },
    {
      "epoch": 8.62,
      "learning_rate": 2.760441670667307e-06,
      "loss": 0.0274,
      "step": 35910
    },
    {
      "epoch": 8.62,
      "learning_rate": 2.7556409025444074e-06,
      "loss": 0.088,
      "step": 35920
    },
    {
      "epoch": 8.62,
      "learning_rate": 2.7508401344215074e-06,
      "loss": 0.0917,
      "step": 35930
    },
    {
      "epoch": 8.63,
      "learning_rate": 2.746039366298608e-06,
      "loss": 0.1361,
      "step": 35940
    },
    {
      "epoch": 8.63,
      "learning_rate": 2.7412385981757082e-06,
      "loss": 0.1388,
      "step": 35950
    },
    {
      "epoch": 8.63,
      "learning_rate": 2.7364378300528087e-06,
      "loss": 0.0728,
      "step": 35960
    },
    {
      "epoch": 8.63,
      "learning_rate": 2.7316370619299087e-06,
      "loss": 0.0112,
      "step": 35970
    },
    {
      "epoch": 8.64,
      "learning_rate": 2.726836293807009e-06,
      "loss": 0.0377,
      "step": 35980
    },
    {
      "epoch": 8.64,
      "learning_rate": 2.7220355256841095e-06,
      "loss": 0.0404,
      "step": 35990
    },
    {
      "epoch": 8.64,
      "learning_rate": 2.7172347575612103e-06,
      "loss": 0.0364,
      "step": 36000
    },
    {
      "epoch": 8.64,
      "learning_rate": 2.7124339894383108e-06,
      "loss": 0.0636,
      "step": 36010
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.7076332213154103e-06,
      "loss": 0.1615,
      "step": 36020
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.702832453192511e-06,
      "loss": 0.1372,
      "step": 36030
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.6980316850696116e-06,
      "loss": 0.0586,
      "step": 36040
    },
    {
      "epoch": 8.65,
      "learning_rate": 2.693230916946712e-06,
      "loss": 0.0477,
      "step": 36050
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.688430148823812e-06,
      "loss": 0.0621,
      "step": 36060
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.6836293807009124e-06,
      "loss": 0.2088,
      "step": 36070
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.678828612578013e-06,
      "loss": 0.0828,
      "step": 36080
    },
    {
      "epoch": 8.66,
      "learning_rate": 2.6740278444551133e-06,
      "loss": 0.1979,
      "step": 36090
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.6692270763322133e-06,
      "loss": 0.1058,
      "step": 36100
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.6644263082093137e-06,
      "loss": 0.0778,
      "step": 36110
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.659625540086414e-06,
      "loss": 0.04,
      "step": 36120
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.6548247719635145e-06,
      "loss": 0.1341,
      "step": 36130
    },
    {
      "epoch": 8.67,
      "learning_rate": 2.650024003840615e-06,
      "loss": 0.0925,
      "step": 36140
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.645223235717715e-06,
      "loss": 0.0177,
      "step": 36150
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.6404224675948154e-06,
      "loss": 0.0232,
      "step": 36160
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.6356216994719158e-06,
      "loss": 0.0384,
      "step": 36170
    },
    {
      "epoch": 8.68,
      "learning_rate": 2.630820931349016e-06,
      "loss": 0.2557,
      "step": 36180
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.626020163226116e-06,
      "loss": 0.1308,
      "step": 36190
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.6212193951032166e-06,
      "loss": 0.1193,
      "step": 36200
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.616418626980317e-06,
      "loss": 0.1673,
      "step": 36210
    },
    {
      "epoch": 8.69,
      "learning_rate": 2.6116178588574175e-06,
      "loss": 0.0019,
      "step": 36220
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.6068170907345175e-06,
      "loss": 0.1037,
      "step": 36230
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.602016322611618e-06,
      "loss": 0.0414,
      "step": 36240
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.5972155544887183e-06,
      "loss": 0.1626,
      "step": 36250
    },
    {
      "epoch": 8.7,
      "learning_rate": 2.5924147863658187e-06,
      "loss": 0.0854,
      "step": 36260
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.5876140182429187e-06,
      "loss": 0.0936,
      "step": 36270
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.582813250120019e-06,
      "loss": 0.1719,
      "step": 36280
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.5780124819971196e-06,
      "loss": 0.1474,
      "step": 36290
    },
    {
      "epoch": 8.71,
      "learning_rate": 2.5732117138742204e-06,
      "loss": 0.1208,
      "step": 36300
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.568410945751321e-06,
      "loss": 0.0029,
      "step": 36310
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.5636101776284204e-06,
      "loss": 0.2022,
      "step": 36320
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.5588094095055212e-06,
      "loss": 0.0562,
      "step": 36330
    },
    {
      "epoch": 8.72,
      "learning_rate": 2.5540086413826216e-06,
      "loss": 0.0899,
      "step": 36340
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.549207873259722e-06,
      "loss": 0.2181,
      "step": 36350
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.544407105136822e-06,
      "loss": 0.1464,
      "step": 36360
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.5396063370139225e-06,
      "loss": 0.1775,
      "step": 36370
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.534805568891023e-06,
      "loss": 0.1081,
      "step": 36380
    },
    {
      "epoch": 8.73,
      "learning_rate": 2.5300048007681233e-06,
      "loss": 0.2208,
      "step": 36390
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.5252040326452233e-06,
      "loss": 0.1515,
      "step": 36400
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.5204032645223237e-06,
      "loss": 0.1562,
      "step": 36410
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.515602496399424e-06,
      "loss": 0.1043,
      "step": 36420
    },
    {
      "epoch": 8.74,
      "learning_rate": 2.5108017282765246e-06,
      "loss": 0.1249,
      "step": 36430
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.506000960153625e-06,
      "loss": 0.135,
      "step": 36440
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.501200192030725e-06,
      "loss": 0.0607,
      "step": 36450
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.4963994239078254e-06,
      "loss": 0.0664,
      "step": 36460
    },
    {
      "epoch": 8.75,
      "learning_rate": 2.491598655784926e-06,
      "loss": 0.0984,
      "step": 36470
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.4867978876620263e-06,
      "loss": 0.0656,
      "step": 36480
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.4819971195391267e-06,
      "loss": 0.0594,
      "step": 36490
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.4771963514162267e-06,
      "loss": 0.0654,
      "step": 36500
    },
    {
      "epoch": 8.76,
      "learning_rate": 2.472395583293327e-06,
      "loss": 0.0641,
      "step": 36510
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.4675948151704275e-06,
      "loss": 0.0239,
      "step": 36520
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.462794047047528e-06,
      "loss": 0.1542,
      "step": 36530
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.457993278924628e-06,
      "loss": 0.0402,
      "step": 36540
    },
    {
      "epoch": 8.77,
      "learning_rate": 2.4531925108017283e-06,
      "loss": 0.214,
      "step": 36550
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.4483917426788288e-06,
      "loss": 0.1662,
      "step": 36560
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.443590974555929e-06,
      "loss": 0.0757,
      "step": 36570
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.4387902064330296e-06,
      "loss": 0.0178,
      "step": 36580
    },
    {
      "epoch": 8.78,
      "learning_rate": 2.4339894383101296e-06,
      "loss": 0.1082,
      "step": 36590
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.4291886701872304e-06,
      "loss": 0.0402,
      "step": 36600
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.4243879020643304e-06,
      "loss": 0.1778,
      "step": 36610
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.419587133941431e-06,
      "loss": 0.1125,
      "step": 36620
    },
    {
      "epoch": 8.79,
      "learning_rate": 2.4147863658185313e-06,
      "loss": 0.1207,
      "step": 36630
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.4099855976956317e-06,
      "loss": 0.0123,
      "step": 36640
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.4051848295727317e-06,
      "loss": 0.1551,
      "step": 36650
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.400384061449832e-06,
      "loss": 0.0674,
      "step": 36660
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.3955832933269325e-06,
      "loss": 0.1039,
      "step": 36670
    },
    {
      "epoch": 8.8,
      "learning_rate": 2.390782525204033e-06,
      "loss": 0.026,
      "step": 36680
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.385981757081133e-06,
      "loss": 0.0884,
      "step": 36690
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.3811809889582334e-06,
      "loss": 0.0832,
      "step": 36700
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.3763802208353338e-06,
      "loss": 0.0427,
      "step": 36710
    },
    {
      "epoch": 8.81,
      "learning_rate": 2.371579452712434e-06,
      "loss": 0.0382,
      "step": 36720
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.3667786845895346e-06,
      "loss": 0.0207,
      "step": 36730
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.3619779164666346e-06,
      "loss": 0.2051,
      "step": 36740
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.3571771483437355e-06,
      "loss": 0.1208,
      "step": 36750
    },
    {
      "epoch": 8.82,
      "learning_rate": 2.3523763802208355e-06,
      "loss": 0.0285,
      "step": 36760
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.347575612097936e-06,
      "loss": 0.006,
      "step": 36770
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.3427748439750363e-06,
      "loss": 0.202,
      "step": 36780
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.3379740758521367e-06,
      "loss": 0.2012,
      "step": 36790
    },
    {
      "epoch": 8.83,
      "learning_rate": 2.3331733077292367e-06,
      "loss": 0.1408,
      "step": 36800
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.328372539606337e-06,
      "loss": 0.0723,
      "step": 36810
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.3235717714834376e-06,
      "loss": 0.123,
      "step": 36820
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.318771003360538e-06,
      "loss": 0.1602,
      "step": 36830
    },
    {
      "epoch": 8.84,
      "learning_rate": 2.313970235237638e-06,
      "loss": 0.1734,
      "step": 36840
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.3091694671147384e-06,
      "loss": 0.1012,
      "step": 36850
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.304368698991839e-06,
      "loss": 0.1971,
      "step": 36860
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.2995679308689392e-06,
      "loss": 0.1942,
      "step": 36870
    },
    {
      "epoch": 8.85,
      "learning_rate": 2.2947671627460396e-06,
      "loss": 0.0578,
      "step": 36880
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.2899663946231396e-06,
      "loss": 0.176,
      "step": 36890
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.2851656265002405e-06,
      "loss": 0.1863,
      "step": 36900
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.2803648583773405e-06,
      "loss": 0.0396,
      "step": 36910
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.275564090254441e-06,
      "loss": 0.2159,
      "step": 36920
    },
    {
      "epoch": 8.86,
      "learning_rate": 2.2707633221315413e-06,
      "loss": 0.0832,
      "step": 36930
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.2659625540086417e-06,
      "loss": 0.2012,
      "step": 36940
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.2611617858857417e-06,
      "loss": 0.2133,
      "step": 36950
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.256361017762842e-06,
      "loss": 0.1355,
      "step": 36960
    },
    {
      "epoch": 8.87,
      "learning_rate": 2.2515602496399426e-06,
      "loss": 0.0235,
      "step": 36970
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.246759481517043e-06,
      "loss": 0.0979,
      "step": 36980
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.241958713394143e-06,
      "loss": 0.0873,
      "step": 36990
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.2371579452712434e-06,
      "loss": 0.0015,
      "step": 37000
    },
    {
      "epoch": 8.88,
      "learning_rate": 2.232357177148344e-06,
      "loss": 0.2196,
      "step": 37010
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.2275564090254443e-06,
      "loss": 0.1781,
      "step": 37020
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.2227556409025447e-06,
      "loss": 0.1108,
      "step": 37030
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.217954872779645e-06,
      "loss": 0.0359,
      "step": 37040
    },
    {
      "epoch": 8.89,
      "learning_rate": 2.2131541046567455e-06,
      "loss": 0.0443,
      "step": 37050
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.2083533365338455e-06,
      "loss": 0.0613,
      "step": 37060
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.203552568410946e-06,
      "loss": 0.057,
      "step": 37070
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.1987518002880463e-06,
      "loss": 0.0378,
      "step": 37080
    },
    {
      "epoch": 8.9,
      "learning_rate": 2.1939510321651468e-06,
      "loss": 0.1198,
      "step": 37090
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.1891502640422468e-06,
      "loss": 0.232,
      "step": 37100
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.184349495919347e-06,
      "loss": 0.1237,
      "step": 37110
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.1795487277964476e-06,
      "loss": 0.1076,
      "step": 37120
    },
    {
      "epoch": 8.91,
      "learning_rate": 2.174747959673548e-06,
      "loss": 0.101,
      "step": 37130
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.169947191550648e-06,
      "loss": 0.1641,
      "step": 37140
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.1651464234277484e-06,
      "loss": 0.2298,
      "step": 37150
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.1603456553048493e-06,
      "loss": 0.2283,
      "step": 37160
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.1555448871819493e-06,
      "loss": 0.002,
      "step": 37170
    },
    {
      "epoch": 8.92,
      "learning_rate": 2.1507441190590497e-06,
      "loss": 0.2284,
      "step": 37180
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.14594335093615e-06,
      "loss": 0.0436,
      "step": 37190
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.1411425828132505e-06,
      "loss": 0.028,
      "step": 37200
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.1363418146903505e-06,
      "loss": 0.015,
      "step": 37210
    },
    {
      "epoch": 8.93,
      "learning_rate": 2.131541046567451e-06,
      "loss": 0.1213,
      "step": 37220
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.1267402784445514e-06,
      "loss": 0.2661,
      "step": 37230
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.121939510321652e-06,
      "loss": 0.1677,
      "step": 37240
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.1171387421987518e-06,
      "loss": 0.0819,
      "step": 37250
    },
    {
      "epoch": 8.94,
      "learning_rate": 2.112337974075852e-06,
      "loss": 0.08,
      "step": 37260
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.1075372059529526e-06,
      "loss": 0.1198,
      "step": 37270
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.102736437830053e-06,
      "loss": 0.0467,
      "step": 37280
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.097935669707153e-06,
      "loss": 0.0958,
      "step": 37290
    },
    {
      "epoch": 8.95,
      "learning_rate": 2.0931349015842535e-06,
      "loss": 0.1183,
      "step": 37300
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.0883341334613543e-06,
      "loss": 0.1418,
      "step": 37310
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.0835333653384543e-06,
      "loss": 0.1526,
      "step": 37320
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.0787325972155547e-06,
      "loss": 0.143,
      "step": 37330
    },
    {
      "epoch": 8.96,
      "learning_rate": 2.073931829092655e-06,
      "loss": 0.0534,
      "step": 37340
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.0691310609697556e-06,
      "loss": 0.1059,
      "step": 37350
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.0643302928468556e-06,
      "loss": 0.0761,
      "step": 37360
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.059529524723956e-06,
      "loss": 0.0224,
      "step": 37370
    },
    {
      "epoch": 8.97,
      "learning_rate": 2.0547287566010564e-06,
      "loss": 0.0215,
      "step": 37380
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.049927988478157e-06,
      "loss": 0.2208,
      "step": 37390
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.045127220355257e-06,
      "loss": 0.1801,
      "step": 37400
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.0403264522323572e-06,
      "loss": 0.1677,
      "step": 37410
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.0355256841094577e-06,
      "loss": 0.1413,
      "step": 37420
    },
    {
      "epoch": 8.98,
      "learning_rate": 2.030724915986558e-06,
      "loss": 0.1408,
      "step": 37430
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.0259241478636585e-06,
      "loss": 0.0303,
      "step": 37440
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.0211233797407585e-06,
      "loss": 0.1131,
      "step": 37450
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.0163226116178593e-06,
      "loss": 0.1378,
      "step": 37460
    },
    {
      "epoch": 8.99,
      "learning_rate": 2.0115218434949593e-06,
      "loss": 0.0746,
      "step": 37470
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.0067210753720597e-06,
      "loss": 0.1386,
      "step": 37480
    },
    {
      "epoch": 9.0,
      "learning_rate": 2.00192030724916e-06,
      "loss": 0.1746,
      "step": 37490
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.9993199591975519,
      "eval_loss": 0.0033987753558903933,
      "eval_runtime": 513.2336,
      "eval_samples_per_second": 11.461,
      "eval_steps_per_second": 1.434,
      "step": 37494
    },
    {
      "epoch": 9.0,
      "learning_rate": 1.9971195391262606e-06,
      "loss": 0.1878,
      "step": 37500
    },
    {
      "epoch": 9.0,
      "learning_rate": 1.9923187710033606e-06,
      "loss": 0.1701,
      "step": 37510
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.987518002880461e-06,
      "loss": 0.206,
      "step": 37520
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.9827172347575614e-06,
      "loss": 0.0521,
      "step": 37530
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.977916466634662e-06,
      "loss": 0.0378,
      "step": 37540
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.973115698511762e-06,
      "loss": 0.1435,
      "step": 37550
    },
    {
      "epoch": 9.02,
      "learning_rate": 1.9683149303888623e-06,
      "loss": 0.0828,
      "step": 37560
    },
    {
      "epoch": 9.02,
      "learning_rate": 1.9635141622659627e-06,
      "loss": 0.1406,
      "step": 37570
    },
    {
      "epoch": 9.02,
      "learning_rate": 1.958713394143063e-06,
      "loss": 0.1259,
      "step": 37580
    },
    {
      "epoch": 9.02,
      "learning_rate": 1.9539126260201635e-06,
      "loss": 0.1752,
      "step": 37590
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.9491118578972635e-06,
      "loss": 0.1011,
      "step": 37600
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.9443110897743644e-06,
      "loss": 0.0785,
      "step": 37610
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.9395103216514643e-06,
      "loss": 0.1497,
      "step": 37620
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.9347095535285648e-06,
      "loss": 0.2559,
      "step": 37630
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.929908785405665e-06,
      "loss": 0.0948,
      "step": 37640
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.9251080172827656e-06,
      "loss": 0.2756,
      "step": 37650
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.9203072491598656e-06,
      "loss": 0.2295,
      "step": 37660
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.915506481036966e-06,
      "loss": 0.1007,
      "step": 37670
    },
    {
      "epoch": 9.04,
      "learning_rate": 1.9107057129140664e-06,
      "loss": 0.1474,
      "step": 37680
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.9059049447911669e-06,
      "loss": 0.2653,
      "step": 37690
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.9011041766682669e-06,
      "loss": 0.1391,
      "step": 37700
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.8963034085453675e-06,
      "loss": 0.0741,
      "step": 37710
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.8915026404224677e-06,
      "loss": 0.0096,
      "step": 37720
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.8867018722995681e-06,
      "loss": 0.035,
      "step": 37730
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.8819011041766685e-06,
      "loss": 0.055,
      "step": 37740
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.8771003360537687e-06,
      "loss": 0.2031,
      "step": 37750
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.8722995679308692e-06,
      "loss": 0.0956,
      "step": 37760
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.8674987998079694e-06,
      "loss": 0.1047,
      "step": 37770
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.8626980316850698e-06,
      "loss": 0.02,
      "step": 37780
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.85789726356217e-06,
      "loss": 0.0987,
      "step": 37790
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.8530964954392704e-06,
      "loss": 0.0381,
      "step": 37800
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.8482957273163706e-06,
      "loss": 0.125,
      "step": 37810
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.8434949591934713e-06,
      "loss": 0.1215,
      "step": 37820
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.8386941910705713e-06,
      "loss": 0.1177,
      "step": 37830
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.8338934229476719e-06,
      "loss": 0.1425,
      "step": 37840
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.829092654824772e-06,
      "loss": 0.1777,
      "step": 37850
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.8242918867018725e-06,
      "loss": 0.0397,
      "step": 37860
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.8194911185789727e-06,
      "loss": 0.0324,
      "step": 37870
    },
    {
      "epoch": 9.09,
      "learning_rate": 1.8146903504560731e-06,
      "loss": 0.1581,
      "step": 37880
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.8098895823331736e-06,
      "loss": 0.1535,
      "step": 37890
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.8050888142102738e-06,
      "loss": 0.182,
      "step": 37900
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.8002880460873742e-06,
      "loss": 0.1094,
      "step": 37910
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.7954872779644744e-06,
      "loss": 0.092,
      "step": 37920
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.7906865098415748e-06,
      "loss": 0.195,
      "step": 37930
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.785885741718675e-06,
      "loss": 0.0445,
      "step": 37940
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.7810849735957754e-06,
      "loss": 0.0243,
      "step": 37950
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.7762842054728757e-06,
      "loss": 0.1486,
      "step": 37960
    },
    {
      "epoch": 9.11,
      "learning_rate": 1.7714834373499763e-06,
      "loss": 0.0886,
      "step": 37970
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.7666826692270763e-06,
      "loss": 0.1356,
      "step": 37980
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.761881901104177e-06,
      "loss": 0.0991,
      "step": 37990
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.7570811329812771e-06,
      "loss": 0.2388,
      "step": 38000
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.7522803648583775e-06,
      "loss": 0.0346,
      "step": 38010
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.747479596735478e-06,
      "loss": 0.231,
      "step": 38020
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.7426788286125782e-06,
      "loss": 0.0353,
      "step": 38030
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.7378780604896786e-06,
      "loss": 0.1968,
      "step": 38040
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.7330772923667788e-06,
      "loss": 0.0631,
      "step": 38050
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.7282765242438792e-06,
      "loss": 0.0521,
      "step": 38060
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.7234757561209794e-06,
      "loss": 0.145,
      "step": 38070
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.7186749879980798e-06,
      "loss": 0.1857,
      "step": 38080
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.71387421987518e-06,
      "loss": 0.0272,
      "step": 38090
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.7090734517522805e-06,
      "loss": 0.1056,
      "step": 38100
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.7042726836293807e-06,
      "loss": 0.0739,
      "step": 38110
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.6994719155064813e-06,
      "loss": 0.1712,
      "step": 38120
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.6946711473835813e-06,
      "loss": 0.1357,
      "step": 38130
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.689870379260682e-06,
      "loss": 0.2179,
      "step": 38140
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.6850696111377821e-06,
      "loss": 0.1611,
      "step": 38150
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.6802688430148826e-06,
      "loss": 0.1862,
      "step": 38160
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.675468074891983e-06,
      "loss": 0.0997,
      "step": 38170
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.6706673067690832e-06,
      "loss": 0.1874,
      "step": 38180
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.6658665386461836e-06,
      "loss": 0.0434,
      "step": 38190
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.6610657705232838e-06,
      "loss": 0.0266,
      "step": 38200
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.6562650024003842e-06,
      "loss": 0.239,
      "step": 38210
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.6514642342774844e-06,
      "loss": 0.1163,
      "step": 38220
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.6466634661545849e-06,
      "loss": 0.3316,
      "step": 38230
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.641862698031685e-06,
      "loss": 0.0933,
      "step": 38240
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.6370619299087855e-06,
      "loss": 0.2108,
      "step": 38250
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.6322611617858857e-06,
      "loss": 0.0327,
      "step": 38260
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.6274603936629863e-06,
      "loss": 0.0451,
      "step": 38270
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.6226596255400865e-06,
      "loss": 0.0216,
      "step": 38280
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.617858857417187e-06,
      "loss": 0.001,
      "step": 38290
    },
    {
      "epoch": 9.19,
      "learning_rate": 1.6130580892942872e-06,
      "loss": 0.0998,
      "step": 38300
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.6082573211713876e-06,
      "loss": 0.1583,
      "step": 38310
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.603456553048488e-06,
      "loss": 0.0751,
      "step": 38320
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.5986557849255882e-06,
      "loss": 0.017,
      "step": 38330
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.5938550168026886e-06,
      "loss": 0.0585,
      "step": 38340
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.5890542486797888e-06,
      "loss": 0.009,
      "step": 38350
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.5842534805568893e-06,
      "loss": 0.0885,
      "step": 38360
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.5794527124339895e-06,
      "loss": 0.0495,
      "step": 38370
    },
    {
      "epoch": 9.21,
      "learning_rate": 1.5746519443110899e-06,
      "loss": 0.0419,
      "step": 38380
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.56985117618819e-06,
      "loss": 0.0551,
      "step": 38390
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.5650504080652907e-06,
      "loss": 0.0302,
      "step": 38400
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.5602496399423907e-06,
      "loss": 0.1685,
      "step": 38410
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.5554488718194914e-06,
      "loss": 0.0711,
      "step": 38420
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.5506481036965916e-06,
      "loss": 0.0825,
      "step": 38430
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.545847335573692e-06,
      "loss": 0.1307,
      "step": 38440
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.5410465674507922e-06,
      "loss": 0.0563,
      "step": 38450
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.5362457993278926e-06,
      "loss": 0.1944,
      "step": 38460
    },
    {
      "epoch": 9.23,
      "learning_rate": 1.531445031204993e-06,
      "loss": 0.112,
      "step": 38470
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.5266442630820932e-06,
      "loss": 0.1698,
      "step": 38480
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.5218434949591937e-06,
      "loss": 0.1531,
      "step": 38490
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.5170427268362939e-06,
      "loss": 0.0607,
      "step": 38500
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.5122419587133943e-06,
      "loss": 0.1413,
      "step": 38510
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.5074411905904945e-06,
      "loss": 0.0615,
      "step": 38520
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.502640422467595e-06,
      "loss": 0.0759,
      "step": 38530
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.4978396543446951e-06,
      "loss": 0.2142,
      "step": 38540
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.4930388862217958e-06,
      "loss": 0.0335,
      "step": 38550
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.4882381180988957e-06,
      "loss": 0.1903,
      "step": 38560
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.4834373499759964e-06,
      "loss": 0.0648,
      "step": 38570
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.4786365818530966e-06,
      "loss": 0.1101,
      "step": 38580
    },
    {
      "epoch": 9.26,
      "learning_rate": 1.473835813730197e-06,
      "loss": 0.1405,
      "step": 38590
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.4690350456072972e-06,
      "loss": 0.1506,
      "step": 38600
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.4642342774843976e-06,
      "loss": 0.2052,
      "step": 38610
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.459433509361498e-06,
      "loss": 0.104,
      "step": 38620
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.4546327412385983e-06,
      "loss": 0.1255,
      "step": 38630
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.4498319731156987e-06,
      "loss": 0.0829,
      "step": 38640
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.4450312049927989e-06,
      "loss": 0.1473,
      "step": 38650
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.4402304368698993e-06,
      "loss": 0.0626,
      "step": 38660
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.4354296687469995e-06,
      "loss": 0.0857,
      "step": 38670
    },
    {
      "epoch": 9.28,
      "learning_rate": 1.4306289006241e-06,
      "loss": 0.0173,
      "step": 38680
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.4258281325012001e-06,
      "loss": 0.153,
      "step": 38690
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.4210273643783008e-06,
      "loss": 0.1947,
      "step": 38700
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.4162265962554008e-06,
      "loss": 0.1525,
      "step": 38710
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.4114258281325014e-06,
      "loss": 0.2249,
      "step": 38720
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.4066250600096016e-06,
      "loss": 0.0994,
      "step": 38730
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.401824291886702e-06,
      "loss": 0.1172,
      "step": 38740
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.3970235237638025e-06,
      "loss": 0.0375,
      "step": 38750
    },
    {
      "epoch": 9.3,
      "learning_rate": 1.3922227556409027e-06,
      "loss": 0.3177,
      "step": 38760
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.387421987518003e-06,
      "loss": 0.1305,
      "step": 38770
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.3826212193951033e-06,
      "loss": 0.077,
      "step": 38780
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.3778204512722037e-06,
      "loss": 0.1399,
      "step": 38790
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.373019683149304e-06,
      "loss": 0.0777,
      "step": 38800
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.3682189150264043e-06,
      "loss": 0.2221,
      "step": 38810
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.3634181469035045e-06,
      "loss": 0.1954,
      "step": 38820
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.3586173787806052e-06,
      "loss": 0.0751,
      "step": 38830
    },
    {
      "epoch": 9.32,
      "learning_rate": 1.3538166106577052e-06,
      "loss": 0.2409,
      "step": 38840
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.3490158425348058e-06,
      "loss": 0.1612,
      "step": 38850
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.344215074411906e-06,
      "loss": 0.0215,
      "step": 38860
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.3394143062890064e-06,
      "loss": 0.1091,
      "step": 38870
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.3346135381661066e-06,
      "loss": 0.047,
      "step": 38880
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.329812770043207e-06,
      "loss": 0.083,
      "step": 38890
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.3250120019203075e-06,
      "loss": 0.1649,
      "step": 38900
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.3202112337974077e-06,
      "loss": 0.0897,
      "step": 38910
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.315410465674508e-06,
      "loss": 0.1613,
      "step": 38920
    },
    {
      "epoch": 9.34,
      "learning_rate": 1.3106096975516083e-06,
      "loss": 0.1087,
      "step": 38930
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.3058089294287087e-06,
      "loss": 0.1112,
      "step": 38940
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.301008161305809e-06,
      "loss": 0.1944,
      "step": 38950
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.2962073931829094e-06,
      "loss": 0.1083,
      "step": 38960
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.2914066250600096e-06,
      "loss": 0.1031,
      "step": 38970
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.2866058569371102e-06,
      "loss": 0.2249,
      "step": 38980
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.2818050888142102e-06,
      "loss": 0.0762,
      "step": 38990
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.2770043206913108e-06,
      "loss": 0.0738,
      "step": 39000
    },
    {
      "epoch": 9.36,
      "learning_rate": 1.272203552568411e-06,
      "loss": 0.1203,
      "step": 39010
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.2674027844455115e-06,
      "loss": 0.0748,
      "step": 39020
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.2626020163226117e-06,
      "loss": 0.2156,
      "step": 39030
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.257801248199712e-06,
      "loss": 0.0364,
      "step": 39040
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.2530004800768125e-06,
      "loss": 0.0599,
      "step": 39050
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.2481997119539127e-06,
      "loss": 0.1296,
      "step": 39060
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.2433989438310131e-06,
      "loss": 0.1681,
      "step": 39070
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.2385981757081133e-06,
      "loss": 0.0017,
      "step": 39080
    },
    {
      "epoch": 9.38,
      "learning_rate": 1.2337974075852138e-06,
      "loss": 0.0728,
      "step": 39090
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.228996639462314e-06,
      "loss": 0.107,
      "step": 39100
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.2241958713394144e-06,
      "loss": 0.0685,
      "step": 39110
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.2193951032165148e-06,
      "loss": 0.0478,
      "step": 39120
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.2145943350936152e-06,
      "loss": 0.027,
      "step": 39130
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.2097935669707154e-06,
      "loss": 0.1607,
      "step": 39140
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.2049927988478158e-06,
      "loss": 0.0729,
      "step": 39150
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.200192030724916e-06,
      "loss": 0.1213,
      "step": 39160
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.1953912626020165e-06,
      "loss": 0.0903,
      "step": 39170
    },
    {
      "epoch": 9.4,
      "learning_rate": 1.1905904944791167e-06,
      "loss": 0.1131,
      "step": 39180
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.185789726356217e-06,
      "loss": 0.072,
      "step": 39190
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.1809889582333173e-06,
      "loss": 0.1336,
      "step": 39200
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.1761881901104177e-06,
      "loss": 0.1518,
      "step": 39210
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.1713874219875182e-06,
      "loss": 0.183,
      "step": 39220
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.1665866538646184e-06,
      "loss": 0.0555,
      "step": 39230
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.1617858857417188e-06,
      "loss": 0.1397,
      "step": 39240
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.156985117618819e-06,
      "loss": 0.1307,
      "step": 39250
    },
    {
      "epoch": 9.42,
      "learning_rate": 1.1521843494959194e-06,
      "loss": 0.1193,
      "step": 39260
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.1473835813730198e-06,
      "loss": 0.096,
      "step": 39270
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.1425828132501202e-06,
      "loss": 0.0553,
      "step": 39280
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.1377820451272205e-06,
      "loss": 0.0361,
      "step": 39290
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.1329812770043209e-06,
      "loss": 0.175,
      "step": 39300
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.128180508881421e-06,
      "loss": 0.114,
      "step": 39310
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.1233797407585215e-06,
      "loss": 0.0598,
      "step": 39320
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.1185789726356217e-06,
      "loss": 0.1577,
      "step": 39330
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.1137782045127221e-06,
      "loss": 0.0409,
      "step": 39340
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.1089774363898225e-06,
      "loss": 0.1862,
      "step": 39350
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.1041766682669228e-06,
      "loss": 0.0979,
      "step": 39360
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.0993759001440232e-06,
      "loss": 0.074,
      "step": 39370
    },
    {
      "epoch": 9.45,
      "learning_rate": 1.0945751320211234e-06,
      "loss": 0.1331,
      "step": 39380
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.0897743638982238e-06,
      "loss": 0.1577,
      "step": 39390
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.084973595775324e-06,
      "loss": 0.2741,
      "step": 39400
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.0801728276524246e-06,
      "loss": 0.0864,
      "step": 39410
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.0753720595295248e-06,
      "loss": 0.077,
      "step": 39420
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.0705712914066253e-06,
      "loss": 0.0169,
      "step": 39430
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.0657705232837255e-06,
      "loss": 0.2377,
      "step": 39440
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.060969755160826e-06,
      "loss": 0.1072,
      "step": 39450
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.056168987037926e-06,
      "loss": 0.1515,
      "step": 39460
    },
    {
      "epoch": 9.47,
      "learning_rate": 1.0513682189150265e-06,
      "loss": 0.2445,
      "step": 39470
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.0465674507921267e-06,
      "loss": 0.2286,
      "step": 39480
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.0417666826692272e-06,
      "loss": 0.0924,
      "step": 39490
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.0369659145463276e-06,
      "loss": 0.2416,
      "step": 39500
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.0321651464234278e-06,
      "loss": 0.0229,
      "step": 39510
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.0273643783005282e-06,
      "loss": 0.1693,
      "step": 39520
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.0225636101776284e-06,
      "loss": 0.1761,
      "step": 39530
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.0177628420547288e-06,
      "loss": 0.0647,
      "step": 39540
    },
    {
      "epoch": 9.49,
      "learning_rate": 1.0129620739318292e-06,
      "loss": 0.113,
      "step": 39550
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0081613058089297e-06,
      "loss": 0.1588,
      "step": 39560
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.0033605376860299e-06,
      "loss": 0.1326,
      "step": 39570
    },
    {
      "epoch": 9.5,
      "learning_rate": 9.985597695631303e-07,
      "loss": 0.137,
      "step": 39580
    },
    {
      "epoch": 9.5,
      "learning_rate": 9.937590014402305e-07,
      "loss": 0.0122,
      "step": 39590
    },
    {
      "epoch": 9.51,
      "learning_rate": 9.88958233317331e-07,
      "loss": 0.1263,
      "step": 39600
    },
    {
      "epoch": 9.51,
      "learning_rate": 9.841574651944311e-07,
      "loss": 0.0744,
      "step": 39610
    },
    {
      "epoch": 9.51,
      "learning_rate": 9.793566970715315e-07,
      "loss": 0.0664,
      "step": 39620
    },
    {
      "epoch": 9.51,
      "learning_rate": 9.745559289486318e-07,
      "loss": 0.1275,
      "step": 39630
    },
    {
      "epoch": 9.52,
      "learning_rate": 9.697551608257322e-07,
      "loss": 0.1238,
      "step": 39640
    },
    {
      "epoch": 9.52,
      "learning_rate": 9.649543927028326e-07,
      "loss": 0.0931,
      "step": 39650
    },
    {
      "epoch": 9.52,
      "learning_rate": 9.601536245799328e-07,
      "loss": 0.2066,
      "step": 39660
    },
    {
      "epoch": 9.52,
      "learning_rate": 9.553528564570332e-07,
      "loss": 0.1512,
      "step": 39670
    },
    {
      "epoch": 9.52,
      "learning_rate": 9.505520883341334e-07,
      "loss": 0.1915,
      "step": 39680
    },
    {
      "epoch": 9.53,
      "learning_rate": 9.457513202112338e-07,
      "loss": 0.1372,
      "step": 39690
    },
    {
      "epoch": 9.53,
      "learning_rate": 9.409505520883343e-07,
      "loss": 0.1282,
      "step": 39700
    },
    {
      "epoch": 9.53,
      "learning_rate": 9.361497839654346e-07,
      "loss": 0.12,
      "step": 39710
    },
    {
      "epoch": 9.53,
      "learning_rate": 9.313490158425349e-07,
      "loss": 0.0362,
      "step": 39720
    },
    {
      "epoch": 9.54,
      "learning_rate": 9.265482477196352e-07,
      "loss": 0.1238,
      "step": 39730
    },
    {
      "epoch": 9.54,
      "learning_rate": 9.217474795967356e-07,
      "loss": 0.2101,
      "step": 39740
    },
    {
      "epoch": 9.54,
      "learning_rate": 9.169467114738359e-07,
      "loss": 0.1995,
      "step": 39750
    },
    {
      "epoch": 9.54,
      "learning_rate": 9.121459433509363e-07,
      "loss": 0.1221,
      "step": 39760
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.073451752280366e-07,
      "loss": 0.1389,
      "step": 39770
    },
    {
      "epoch": 9.55,
      "learning_rate": 9.025444071051369e-07,
      "loss": 0.1688,
      "step": 39780
    },
    {
      "epoch": 9.55,
      "learning_rate": 8.977436389822372e-07,
      "loss": 0.09,
      "step": 39790
    },
    {
      "epoch": 9.55,
      "learning_rate": 8.929428708593375e-07,
      "loss": 0.1276,
      "step": 39800
    },
    {
      "epoch": 9.56,
      "learning_rate": 8.881421027364378e-07,
      "loss": 0.1869,
      "step": 39810
    },
    {
      "epoch": 9.56,
      "learning_rate": 8.833413346135381e-07,
      "loss": 0.2138,
      "step": 39820
    },
    {
      "epoch": 9.56,
      "learning_rate": 8.785405664906386e-07,
      "loss": 0.1163,
      "step": 39830
    },
    {
      "epoch": 9.56,
      "learning_rate": 8.73739798367739e-07,
      "loss": 0.0455,
      "step": 39840
    },
    {
      "epoch": 9.57,
      "learning_rate": 8.689390302448393e-07,
      "loss": 0.0664,
      "step": 39850
    },
    {
      "epoch": 9.57,
      "learning_rate": 8.641382621219396e-07,
      "loss": 0.0044,
      "step": 39860
    },
    {
      "epoch": 9.57,
      "learning_rate": 8.593374939990399e-07,
      "loss": 0.0218,
      "step": 39870
    },
    {
      "epoch": 9.57,
      "learning_rate": 8.545367258761402e-07,
      "loss": 0.0642,
      "step": 39880
    },
    {
      "epoch": 9.58,
      "learning_rate": 8.497359577532407e-07,
      "loss": 0.2006,
      "step": 39890
    },
    {
      "epoch": 9.58,
      "learning_rate": 8.44935189630341e-07,
      "loss": 0.0605,
      "step": 39900
    },
    {
      "epoch": 9.58,
      "learning_rate": 8.401344215074413e-07,
      "loss": 0.0787,
      "step": 39910
    },
    {
      "epoch": 9.58,
      "learning_rate": 8.353336533845416e-07,
      "loss": 0.0373,
      "step": 39920
    },
    {
      "epoch": 9.58,
      "learning_rate": 8.305328852616419e-07,
      "loss": 0.1479,
      "step": 39930
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.257321171387422e-07,
      "loss": 0.1385,
      "step": 39940
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.209313490158425e-07,
      "loss": 0.0012,
      "step": 39950
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.161305808929429e-07,
      "loss": 0.1451,
      "step": 39960
    },
    {
      "epoch": 9.59,
      "learning_rate": 8.113298127700433e-07,
      "loss": 0.1556,
      "step": 39970
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.065290446471436e-07,
      "loss": 0.1122,
      "step": 39980
    },
    {
      "epoch": 9.6,
      "learning_rate": 8.01728276524244e-07,
      "loss": 0.0253,
      "step": 39990
    },
    {
      "epoch": 9.6,
      "learning_rate": 7.969275084013443e-07,
      "loss": 0.1737,
      "step": 40000
    },
    {
      "epoch": 9.6,
      "learning_rate": 7.921267402784446e-07,
      "loss": 0.1586,
      "step": 40010
    },
    {
      "epoch": 9.61,
      "learning_rate": 7.873259721555449e-07,
      "loss": 0.0304,
      "step": 40020
    },
    {
      "epoch": 9.61,
      "learning_rate": 7.825252040326454e-07,
      "loss": 0.1445,
      "step": 40030
    },
    {
      "epoch": 9.61,
      "learning_rate": 7.777244359097457e-07,
      "loss": 0.002,
      "step": 40040
    },
    {
      "epoch": 9.61,
      "learning_rate": 7.72923667786846e-07,
      "loss": 0.1163,
      "step": 40050
    },
    {
      "epoch": 9.62,
      "learning_rate": 7.681228996639463e-07,
      "loss": 0.073,
      "step": 40060
    },
    {
      "epoch": 9.62,
      "learning_rate": 7.633221315410466e-07,
      "loss": 0.1001,
      "step": 40070
    },
    {
      "epoch": 9.62,
      "learning_rate": 7.585213634181469e-07,
      "loss": 0.0139,
      "step": 40080
    },
    {
      "epoch": 9.62,
      "learning_rate": 7.537205952952472e-07,
      "loss": 0.0942,
      "step": 40090
    },
    {
      "epoch": 9.63,
      "learning_rate": 7.489198271723476e-07,
      "loss": 0.1576,
      "step": 40100
    },
    {
      "epoch": 9.63,
      "learning_rate": 7.441190590494479e-07,
      "loss": 0.1171,
      "step": 40110
    },
    {
      "epoch": 9.63,
      "learning_rate": 7.393182909265483e-07,
      "loss": 0.0171,
      "step": 40120
    },
    {
      "epoch": 9.63,
      "learning_rate": 7.345175228036486e-07,
      "loss": 0.1325,
      "step": 40130
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.29716754680749e-07,
      "loss": 0.0591,
      "step": 40140
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.249159865578493e-07,
      "loss": 0.1072,
      "step": 40150
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.201152184349497e-07,
      "loss": 0.1103,
      "step": 40160
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.1531445031205e-07,
      "loss": 0.0459,
      "step": 40170
    },
    {
      "epoch": 9.64,
      "learning_rate": 7.105136821891504e-07,
      "loss": 0.1257,
      "step": 40180
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.057129140662507e-07,
      "loss": 0.2633,
      "step": 40190
    },
    {
      "epoch": 9.65,
      "learning_rate": 7.00912145943351e-07,
      "loss": 0.0624,
      "step": 40200
    },
    {
      "epoch": 9.65,
      "learning_rate": 6.961113778204513e-07,
      "loss": 0.0512,
      "step": 40210
    },
    {
      "epoch": 9.65,
      "learning_rate": 6.913106096975516e-07,
      "loss": 0.0438,
      "step": 40220
    },
    {
      "epoch": 9.66,
      "learning_rate": 6.86509841574652e-07,
      "loss": 0.0351,
      "step": 40230
    },
    {
      "epoch": 9.66,
      "learning_rate": 6.817090734517523e-07,
      "loss": 0.1526,
      "step": 40240
    },
    {
      "epoch": 9.66,
      "learning_rate": 6.769083053288526e-07,
      "loss": 0.1071,
      "step": 40250
    },
    {
      "epoch": 9.66,
      "learning_rate": 6.72107537205953e-07,
      "loss": 0.1395,
      "step": 40260
    },
    {
      "epoch": 9.67,
      "learning_rate": 6.673067690830533e-07,
      "loss": 0.0983,
      "step": 40270
    },
    {
      "epoch": 9.67,
      "learning_rate": 6.625060009601537e-07,
      "loss": 0.1229,
      "step": 40280
    },
    {
      "epoch": 9.67,
      "learning_rate": 6.57705232837254e-07,
      "loss": 0.1386,
      "step": 40290
    },
    {
      "epoch": 9.67,
      "learning_rate": 6.529044647143544e-07,
      "loss": 0.3393,
      "step": 40300
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.481036965914547e-07,
      "loss": 0.0607,
      "step": 40310
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.433029284685551e-07,
      "loss": 0.0755,
      "step": 40320
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.385021603456554e-07,
      "loss": 0.0597,
      "step": 40330
    },
    {
      "epoch": 9.68,
      "learning_rate": 6.337013922227557e-07,
      "loss": 0.0224,
      "step": 40340
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.28900624099856e-07,
      "loss": 0.0668,
      "step": 40350
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.240998559769564e-07,
      "loss": 0.1638,
      "step": 40360
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.192990878540567e-07,
      "loss": 0.1092,
      "step": 40370
    },
    {
      "epoch": 9.69,
      "learning_rate": 6.14498319731157e-07,
      "loss": 0.0599,
      "step": 40380
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.096975516082574e-07,
      "loss": 0.1491,
      "step": 40390
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.048967834853577e-07,
      "loss": 0.0223,
      "step": 40400
    },
    {
      "epoch": 9.7,
      "learning_rate": 6.00096015362458e-07,
      "loss": 0.0953,
      "step": 40410
    },
    {
      "epoch": 9.7,
      "learning_rate": 5.952952472395583e-07,
      "loss": 0.0171,
      "step": 40420
    },
    {
      "epoch": 9.7,
      "learning_rate": 5.904944791166587e-07,
      "loss": 0.0875,
      "step": 40430
    },
    {
      "epoch": 9.71,
      "learning_rate": 5.856937109937591e-07,
      "loss": 0.0847,
      "step": 40440
    },
    {
      "epoch": 9.71,
      "learning_rate": 5.808929428708594e-07,
      "loss": 0.0336,
      "step": 40450
    },
    {
      "epoch": 9.71,
      "learning_rate": 5.760921747479597e-07,
      "loss": 0.0778,
      "step": 40460
    },
    {
      "epoch": 9.71,
      "learning_rate": 5.712914066250601e-07,
      "loss": 0.1324,
      "step": 40470
    },
    {
      "epoch": 9.72,
      "learning_rate": 5.664906385021604e-07,
      "loss": 0.1527,
      "step": 40480
    },
    {
      "epoch": 9.72,
      "learning_rate": 5.616898703792607e-07,
      "loss": 0.1002,
      "step": 40490
    },
    {
      "epoch": 9.72,
      "learning_rate": 5.568891022563611e-07,
      "loss": 0.0011,
      "step": 40500
    },
    {
      "epoch": 9.72,
      "learning_rate": 5.520883341334614e-07,
      "loss": 0.042,
      "step": 40510
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.472875660105617e-07,
      "loss": 0.1818,
      "step": 40520
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.42486797887662e-07,
      "loss": 0.073,
      "step": 40530
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.376860297647624e-07,
      "loss": 0.0294,
      "step": 40540
    },
    {
      "epoch": 9.73,
      "learning_rate": 5.328852616418627e-07,
      "loss": 0.1023,
      "step": 40550
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.28084493518963e-07,
      "loss": 0.1877,
      "step": 40560
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.232837253960634e-07,
      "loss": 0.0826,
      "step": 40570
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.184829572731638e-07,
      "loss": 0.0465,
      "step": 40580
    },
    {
      "epoch": 9.74,
      "learning_rate": 5.136821891502641e-07,
      "loss": 0.0032,
      "step": 40590
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.088814210273644e-07,
      "loss": 0.1634,
      "step": 40600
    },
    {
      "epoch": 9.75,
      "learning_rate": 5.040806529044648e-07,
      "loss": 0.0433,
      "step": 40610
    },
    {
      "epoch": 9.75,
      "learning_rate": 4.992798847815651e-07,
      "loss": 0.0801,
      "step": 40620
    },
    {
      "epoch": 9.75,
      "learning_rate": 4.944791166586655e-07,
      "loss": 0.1849,
      "step": 40630
    },
    {
      "epoch": 9.76,
      "learning_rate": 4.896783485357658e-07,
      "loss": 0.0679,
      "step": 40640
    },
    {
      "epoch": 9.76,
      "learning_rate": 4.848775804128661e-07,
      "loss": 0.0169,
      "step": 40650
    },
    {
      "epoch": 9.76,
      "learning_rate": 4.800768122899664e-07,
      "loss": 0.0615,
      "step": 40660
    },
    {
      "epoch": 9.76,
      "learning_rate": 4.752760441670667e-07,
      "loss": 0.2065,
      "step": 40670
    },
    {
      "epoch": 9.76,
      "learning_rate": 4.7047527604416713e-07,
      "loss": 0.045,
      "step": 40680
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.6567450792126745e-07,
      "loss": 0.0923,
      "step": 40690
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.608737397983678e-07,
      "loss": 0.1579,
      "step": 40700
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.5607297167546813e-07,
      "loss": 0.1446,
      "step": 40710
    },
    {
      "epoch": 9.77,
      "learning_rate": 4.5127220355256844e-07,
      "loss": 0.1923,
      "step": 40720
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.4647143542966876e-07,
      "loss": 0.0255,
      "step": 40730
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.4167066730676907e-07,
      "loss": 0.0468,
      "step": 40740
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.368698991838695e-07,
      "loss": 0.0678,
      "step": 40750
    },
    {
      "epoch": 9.78,
      "learning_rate": 4.320691310609698e-07,
      "loss": 0.0587,
      "step": 40760
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.272683629380701e-07,
      "loss": 0.0953,
      "step": 40770
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.224675948151705e-07,
      "loss": 0.0455,
      "step": 40780
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.176668266922708e-07,
      "loss": 0.0845,
      "step": 40790
    },
    {
      "epoch": 9.79,
      "learning_rate": 4.128660585693711e-07,
      "loss": 0.1594,
      "step": 40800
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.080652904464714e-07,
      "loss": 0.1171,
      "step": 40810
    },
    {
      "epoch": 9.8,
      "learning_rate": 4.032645223235718e-07,
      "loss": 0.1101,
      "step": 40820
    },
    {
      "epoch": 9.8,
      "learning_rate": 3.9846375420067216e-07,
      "loss": 0.1529,
      "step": 40830
    },
    {
      "epoch": 9.8,
      "learning_rate": 3.9366298607777247e-07,
      "loss": 0.2728,
      "step": 40840
    },
    {
      "epoch": 9.81,
      "learning_rate": 3.8886221795487284e-07,
      "loss": 0.131,
      "step": 40850
    },
    {
      "epoch": 9.81,
      "learning_rate": 3.8406144983197315e-07,
      "loss": 0.0351,
      "step": 40860
    },
    {
      "epoch": 9.81,
      "learning_rate": 3.7926068170907347e-07,
      "loss": 0.1684,
      "step": 40870
    },
    {
      "epoch": 9.81,
      "learning_rate": 3.744599135861738e-07,
      "loss": 0.1191,
      "step": 40880
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.6965914546327415e-07,
      "loss": 0.1576,
      "step": 40890
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.648583773403745e-07,
      "loss": 0.1688,
      "step": 40900
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.6005760921747483e-07,
      "loss": 0.0885,
      "step": 40910
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.552568410945752e-07,
      "loss": 0.1408,
      "step": 40920
    },
    {
      "epoch": 9.82,
      "learning_rate": 3.504560729716755e-07,
      "loss": 0.0328,
      "step": 40930
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.456553048487758e-07,
      "loss": 0.1646,
      "step": 40940
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.4085453672587614e-07,
      "loss": 0.1002,
      "step": 40950
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.360537686029765e-07,
      "loss": 0.0981,
      "step": 40960
    },
    {
      "epoch": 9.83,
      "learning_rate": 3.3125300048007687e-07,
      "loss": 0.0534,
      "step": 40970
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.264522323571772e-07,
      "loss": 0.2004,
      "step": 40980
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.2165146423427755e-07,
      "loss": 0.071,
      "step": 40990
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.1685069611137786e-07,
      "loss": 0.0341,
      "step": 41000
    },
    {
      "epoch": 9.84,
      "learning_rate": 3.120499279884782e-07,
      "loss": 0.1309,
      "step": 41010
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.072491598655785e-07,
      "loss": 0.0926,
      "step": 41020
    },
    {
      "epoch": 9.85,
      "learning_rate": 3.0244839174267886e-07,
      "loss": 0.1122,
      "step": 41030
    },
    {
      "epoch": 9.85,
      "learning_rate": 2.9764762361977917e-07,
      "loss": 0.0592,
      "step": 41040
    },
    {
      "epoch": 9.85,
      "learning_rate": 2.9284685549687954e-07,
      "loss": 0.1313,
      "step": 41050
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.8804608737397985e-07,
      "loss": 0.14,
      "step": 41060
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.832453192510802e-07,
      "loss": 0.0849,
      "step": 41070
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.7844455112818053e-07,
      "loss": 0.1438,
      "step": 41080
    },
    {
      "epoch": 9.86,
      "learning_rate": 2.7364378300528085e-07,
      "loss": 0.0411,
      "step": 41090
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.688430148823812e-07,
      "loss": 0.1346,
      "step": 41100
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.640422467594815e-07,
      "loss": 0.1163,
      "step": 41110
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.592414786365819e-07,
      "loss": 0.1592,
      "step": 41120
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.544407105136822e-07,
      "loss": 0.0797,
      "step": 41130
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.4963994239078257e-07,
      "loss": 0.0269,
      "step": 41140
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.448391742678829e-07,
      "loss": 0.2395,
      "step": 41150
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.400384061449832e-07,
      "loss": 0.1327,
      "step": 41160
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.3523763802208357e-07,
      "loss": 0.0493,
      "step": 41170
    },
    {
      "epoch": 9.88,
      "learning_rate": 2.304368698991839e-07,
      "loss": 0.1222,
      "step": 41180
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.2563610177628422e-07,
      "loss": 0.1053,
      "step": 41190
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.2083533365338454e-07,
      "loss": 0.2707,
      "step": 41200
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.160345655304849e-07,
      "loss": 0.0977,
      "step": 41210
    },
    {
      "epoch": 9.89,
      "learning_rate": 2.1123379740758524e-07,
      "loss": 0.1164,
      "step": 41220
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.0643302928468556e-07,
      "loss": 0.0351,
      "step": 41230
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.016322611617859e-07,
      "loss": 0.138,
      "step": 41240
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.9683149303888624e-07,
      "loss": 0.0721,
      "step": 41250
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.9203072491598658e-07,
      "loss": 0.0985,
      "step": 41260
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.872299567930869e-07,
      "loss": 0.0556,
      "step": 41270
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.8242918867018726e-07,
      "loss": 0.0595,
      "step": 41280
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.776284205472876e-07,
      "loss": 0.0777,
      "step": 41290
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.728276524243879e-07,
      "loss": 0.0311,
      "step": 41300
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.6802688430148825e-07,
      "loss": 0.2685,
      "step": 41310
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.632261161785886e-07,
      "loss": 0.1761,
      "step": 41320
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.5842534805568893e-07,
      "loss": 0.0728,
      "step": 41330
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.5362457993278925e-07,
      "loss": 0.0459,
      "step": 41340
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.4882381180988959e-07,
      "loss": 0.0592,
      "step": 41350
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.4402304368698993e-07,
      "loss": 0.1233,
      "step": 41360
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.3922227556409027e-07,
      "loss": 0.222,
      "step": 41370
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.344215074411906e-07,
      "loss": 0.092,
      "step": 41380
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.2962073931829095e-07,
      "loss": 0.3358,
      "step": 41390
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.2481997119539129e-07,
      "loss": 0.0952,
      "step": 41400
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.200192030724916e-07,
      "loss": 0.1867,
      "step": 41410
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.1521843494959195e-07,
      "loss": 0.0464,
      "step": 41420
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.1041766682669227e-07,
      "loss": 0.15,
      "step": 41430
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.0561689870379262e-07,
      "loss": 0.0712,
      "step": 41440
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.0081613058089295e-07,
      "loss": 0.159,
      "step": 41450
    },
    {
      "epoch": 9.95,
      "learning_rate": 9.601536245799329e-08,
      "loss": 0.1062,
      "step": 41460
    },
    {
      "epoch": 9.95,
      "learning_rate": 9.121459433509363e-08,
      "loss": 0.1292,
      "step": 41470
    },
    {
      "epoch": 9.96,
      "learning_rate": 8.641382621219396e-08,
      "loss": 0.1752,
      "step": 41480
    },
    {
      "epoch": 9.96,
      "learning_rate": 8.16130580892943e-08,
      "loss": 0.1284,
      "step": 41490
    },
    {
      "epoch": 9.96,
      "learning_rate": 7.681228996639462e-08,
      "loss": 0.1465,
      "step": 41500
    },
    {
      "epoch": 9.96,
      "learning_rate": 7.201152184349496e-08,
      "loss": 0.0496,
      "step": 41510
    },
    {
      "epoch": 9.97,
      "learning_rate": 6.72107537205953e-08,
      "loss": 0.011,
      "step": 41520
    },
    {
      "epoch": 9.97,
      "learning_rate": 6.240998559769564e-08,
      "loss": 0.0567,
      "step": 41530
    },
    {
      "epoch": 9.97,
      "learning_rate": 5.760921747479598e-08,
      "loss": 0.0069,
      "step": 41540
    },
    {
      "epoch": 9.97,
      "learning_rate": 5.280844935189631e-08,
      "loss": 0.0323,
      "step": 41550
    },
    {
      "epoch": 9.98,
      "learning_rate": 4.8007681228996644e-08,
      "loss": 0.1293,
      "step": 41560
    },
    {
      "epoch": 9.98,
      "learning_rate": 4.320691310609698e-08,
      "loss": 0.0513,
      "step": 41570
    },
    {
      "epoch": 9.98,
      "learning_rate": 3.840614498319731e-08,
      "loss": 0.0365,
      "step": 41580
    },
    {
      "epoch": 9.98,
      "learning_rate": 3.360537686029765e-08,
      "loss": 0.0485,
      "step": 41590
    },
    {
      "epoch": 9.99,
      "learning_rate": 2.880460873739799e-08,
      "loss": 0.3029,
      "step": 41600
    },
    {
      "epoch": 9.99,
      "learning_rate": 2.4003840614498322e-08,
      "loss": 0.0429,
      "step": 41610
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.9203072491598656e-08,
      "loss": 0.0587,
      "step": 41620
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.4402304368698994e-08,
      "loss": 0.0627,
      "step": 41630
    },
    {
      "epoch": 10.0,
      "learning_rate": 9.601536245799328e-09,
      "loss": 0.1289,
      "step": 41640
    },
    {
      "epoch": 10.0,
      "learning_rate": 4.800768122899664e-09,
      "loss": 0.0901,
      "step": 41650
    },
    {
      "epoch": 10.0,
      "learning_rate": 0.0,
      "loss": 0.1153,
      "step": 41660
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.9989799387963277,
      "eval_loss": 0.003796164644882083,
      "eval_runtime": 514.469,
      "eval_samples_per_second": 11.433,
      "eval_steps_per_second": 1.431,
      "step": 41660
    },
    {
      "epoch": 10.0,
      "step": 41660,
      "total_flos": 2.5835477038076805e+19,
      "train_loss": 0.24897499611371868,
      "train_runtime": 81597.3302,
      "train_samples_per_second": 4.084,
      "train_steps_per_second": 0.511
    }
  ],
  "max_steps": 41660,
  "num_train_epochs": 10,
  "total_flos": 2.5835477038076805e+19,
  "trial_name": null,
  "trial_params": null
}