{
  "best_metric": 0.5362423368193292,
  "best_model_checkpoint": "videomae-base-finetuned-ucf101-subset\\checkpoint-29760",
  "epoch": 4.2,
  "eval_steps": 500,
  "global_step": 29760,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.6801075268817206e-07,
      "loss": 4.8645,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.360215053763441e-07,
      "loss": 4.8248,
      "step": 20
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.040322580645161e-07,
      "loss": 4.8447,
      "step": 30
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.720430107526882e-07,
      "loss": 4.7971,
      "step": 40
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.400537634408602e-07,
      "loss": 4.9621,
      "step": 50
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.0080645161290323e-06,
      "loss": 4.9151,
      "step": 60
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.1760752688172044e-06,
      "loss": 4.8636,
      "step": 70
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.3440860215053765e-06,
      "loss": 4.7778,
      "step": 80
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.5120967741935486e-06,
      "loss": 4.8105,
      "step": 90
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6801075268817204e-06,
      "loss": 4.8286,
      "step": 100
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.8481182795698927e-06,
      "loss": 4.9951,
      "step": 110
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0161290322580646e-06,
      "loss": 4.8796,
      "step": 120
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.184139784946237e-06,
      "loss": 4.8575,
      "step": 130
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.3521505376344088e-06,
      "loss": 4.8098,
      "step": 140
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.5201612903225806e-06,
      "loss": 4.8508,
      "step": 150
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.688172043010753e-06,
      "loss": 4.9504,
      "step": 160
    },
    {
      "epoch": 0.01,
      "learning_rate": 2.856182795698925e-06,
      "loss": 4.7738,
      "step": 170
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.024193548387097e-06,
      "loss": 4.9224,
      "step": 180
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.192204301075269e-06,
      "loss": 4.9128,
      "step": 190
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.360215053763441e-06,
      "loss": 4.9313,
      "step": 200
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.5282258064516136e-06,
      "loss": 4.9543,
      "step": 210
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.6962365591397855e-06,
      "loss": 4.8187,
      "step": 220
    },
    {
      "epoch": 0.01,
      "learning_rate": 3.864247311827957e-06,
      "loss": 4.7944,
      "step": 230
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.032258064516129e-06,
      "loss": 4.9147,
      "step": 240
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.200268817204301e-06,
      "loss": 4.844,
      "step": 250
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.368279569892474e-06,
      "loss": 4.8138,
      "step": 260
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.536290322580646e-06,
      "loss": 4.8867,
      "step": 270
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.7043010752688175e-06,
      "loss": 4.861,
      "step": 280
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.872311827956989e-06,
      "loss": 4.8429,
      "step": 290
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.040322580645161e-06,
      "loss": 4.8064,
      "step": 300
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.208333333333334e-06,
      "loss": 4.9924,
      "step": 310
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.376344086021506e-06,
      "loss": 4.8166,
      "step": 320
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.544354838709678e-06,
      "loss": 4.7953,
      "step": 330
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.71236559139785e-06,
      "loss": 4.8624,
      "step": 340
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.880376344086022e-06,
      "loss": 4.9109,
      "step": 350
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.048387096774194e-06,
      "loss": 4.785,
      "step": 360
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.216397849462366e-06,
      "loss": 4.8804,
      "step": 370
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.384408602150538e-06,
      "loss": 4.8781,
      "step": 380
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.55241935483871e-06,
      "loss": 4.9384,
      "step": 390
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.720430107526882e-06,
      "loss": 4.8682,
      "step": 400
    },
    {
      "epoch": 0.01,
      "learning_rate": 6.888440860215054e-06,
      "loss": 4.9125,
      "step": 410
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.056451612903227e-06,
      "loss": 4.9808,
      "step": 420
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.224462365591399e-06,
      "loss": 4.9357,
      "step": 430
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.392473118279571e-06,
      "loss": 4.8108,
      "step": 440
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.560483870967743e-06,
      "loss": 4.8287,
      "step": 450
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.728494623655915e-06,
      "loss": 4.829,
      "step": 460
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.896505376344086e-06,
      "loss": 4.8015,
      "step": 470
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.064516129032258e-06,
      "loss": 4.8736,
      "step": 480
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.23252688172043e-06,
      "loss": 4.7864,
      "step": 490
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.400537634408602e-06,
      "loss": 4.898,
      "step": 500
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.568548387096773e-06,
      "loss": 4.8142,
      "step": 510
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.736559139784948e-06,
      "loss": 4.8607,
      "step": 520
    },
    {
      "epoch": 0.02,
      "learning_rate": 8.904569892473119e-06,
      "loss": 4.9351,
      "step": 530
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.072580645161291e-06,
      "loss": 4.7266,
      "step": 540
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.240591397849462e-06,
      "loss": 4.8632,
      "step": 550
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.408602150537635e-06,
      "loss": 4.8303,
      "step": 560
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.576612903225806e-06,
      "loss": 4.8754,
      "step": 570
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.744623655913979e-06,
      "loss": 4.7738,
      "step": 580
    },
    {
      "epoch": 0.02,
      "learning_rate": 9.91263440860215e-06,
      "loss": 4.9598,
      "step": 590
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0080645161290323e-05,
      "loss": 4.9171,
      "step": 600
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0248655913978495e-05,
      "loss": 4.8422,
      "step": 610
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0416666666666668e-05,
      "loss": 4.8404,
      "step": 620
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0584677419354839e-05,
      "loss": 4.914,
      "step": 630
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0752688172043012e-05,
      "loss": 4.8882,
      "step": 640
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.0920698924731183e-05,
      "loss": 4.8929,
      "step": 650
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1088709677419356e-05,
      "loss": 4.8637,
      "step": 660
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1256720430107527e-05,
      "loss": 4.8609,
      "step": 670
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.14247311827957e-05,
      "loss": 4.7739,
      "step": 680
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.159274193548387e-05,
      "loss": 4.8239,
      "step": 690
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1760752688172045e-05,
      "loss": 4.8057,
      "step": 700
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.1928763440860216e-05,
      "loss": 4.8922,
      "step": 710
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2096774193548388e-05,
      "loss": 4.7895,
      "step": 720
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.226478494623656e-05,
      "loss": 4.8597,
      "step": 730
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.2432795698924732e-05,
      "loss": 4.9904,
      "step": 740
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.2600806451612903e-05,
      "loss": 4.839,
      "step": 750
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.2768817204301076e-05,
      "loss": 4.8108,
      "step": 760
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.2936827956989247e-05,
      "loss": 4.8368,
      "step": 770
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.310483870967742e-05,
      "loss": 4.8011,
      "step": 780
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.327284946236559e-05,
      "loss": 4.8465,
      "step": 790
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.3440860215053763e-05,
      "loss": 4.86,
      "step": 800
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.3608870967741934e-05,
      "loss": 4.9349,
      "step": 810
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.3776881720430107e-05,
      "loss": 4.8882,
      "step": 820
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.3944892473118278e-05,
      "loss": 4.8517,
      "step": 830
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.4112903225806454e-05,
      "loss": 4.7464,
      "step": 840
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.4280913978494625e-05,
      "loss": 4.7981,
      "step": 850
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.4448924731182798e-05,
      "loss": 4.8219,
      "step": 860
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.4616935483870969e-05,
      "loss": 4.8314,
      "step": 870
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.4784946236559142e-05,
      "loss": 4.862,
      "step": 880
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.4952956989247313e-05,
      "loss": 4.885,
      "step": 890
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.5120967741935486e-05,
      "loss": 4.7596,
      "step": 900
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.5288978494623658e-05,
      "loss": 4.8378,
      "step": 910
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.545698924731183e-05,
      "loss": 4.8938,
      "step": 920
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.5625e-05,
      "loss": 4.7002,
      "step": 930
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.579301075268817e-05,
      "loss": 4.781,
      "step": 940
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.5961021505376346e-05,
      "loss": 4.9581,
      "step": 950
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6129032258064517e-05,
      "loss": 4.8939,
      "step": 960
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6297043010752688e-05,
      "loss": 4.8499,
      "step": 970
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.646505376344086e-05,
      "loss": 4.7838,
      "step": 980
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6633064516129033e-05,
      "loss": 4.8885,
      "step": 990
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6801075268817204e-05,
      "loss": 4.9412,
      "step": 1000
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.6969086021505375e-05,
      "loss": 4.8294,
      "step": 1010
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7137096774193546e-05,
      "loss": 4.7796,
      "step": 1020
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7305107526881724e-05,
      "loss": 4.8445,
      "step": 1030
    },
    {
      "epoch": 0.03,
      "learning_rate": 1.7473118279569895e-05,
      "loss": 4.919,
      "step": 1040
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.7641129032258066e-05,
      "loss": 4.8516,
      "step": 1050
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.7809139784946237e-05,
      "loss": 4.9053,
      "step": 1060
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.797715053763441e-05,
      "loss": 4.844,
      "step": 1070
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.8145161290322583e-05,
      "loss": 4.8267,
      "step": 1080
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.8313172043010754e-05,
      "loss": 4.793,
      "step": 1090
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.8481182795698925e-05,
      "loss": 4.786,
      "step": 1100
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.86491935483871e-05,
      "loss": 4.7492,
      "step": 1110
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.881720430107527e-05,
      "loss": 4.9087,
      "step": 1120
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.898521505376344e-05,
      "loss": 4.8693,
      "step": 1130
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9153225806451612e-05,
      "loss": 4.794,
      "step": 1140
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9321236559139787e-05,
      "loss": 4.8913,
      "step": 1150
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9489247311827958e-05,
      "loss": 4.7268,
      "step": 1160
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.965725806451613e-05,
      "loss": 4.85,
      "step": 1170
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.98252688172043e-05,
      "loss": 4.879,
      "step": 1180
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.9993279569892474e-05,
      "loss": 4.8086,
      "step": 1190
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.0161290322580645e-05,
      "loss": 4.8877,
      "step": 1200
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.032930107526882e-05,
      "loss": 4.8549,
      "step": 1210
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.049731182795699e-05,
      "loss": 4.837,
      "step": 1220
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.0665322580645165e-05,
      "loss": 4.8102,
      "step": 1230
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.0833333333333336e-05,
      "loss": 4.7385,
      "step": 1240
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.1001344086021507e-05,
      "loss": 4.7322,
      "step": 1250
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.1169354838709678e-05,
      "loss": 4.7797,
      "step": 1260
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.1337365591397852e-05,
      "loss": 4.7483,
      "step": 1270
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.1505376344086024e-05,
      "loss": 4.9216,
      "step": 1280
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.1673387096774195e-05,
      "loss": 4.9207,
      "step": 1290
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.1841397849462366e-05,
      "loss": 4.7954,
      "step": 1300
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.200940860215054e-05,
      "loss": 4.7491,
      "step": 1310
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.217741935483871e-05,
      "loss": 4.6397,
      "step": 1320
    },
    {
      "epoch": 0.04,
      "learning_rate": 2.2345430107526882e-05,
      "loss": 4.8405,
      "step": 1330
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.2513440860215053e-05,
      "loss": 4.6964,
      "step": 1340
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.2681451612903227e-05,
      "loss": 4.7082,
      "step": 1350
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.28494623655914e-05,
      "loss": 4.7111,
      "step": 1360
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.301747311827957e-05,
      "loss": 4.7278,
      "step": 1370
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.318548387096774e-05,
      "loss": 4.7585,
      "step": 1380
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.3353494623655915e-05,
      "loss": 4.8247,
      "step": 1390
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.352150537634409e-05,
      "loss": 4.7762,
      "step": 1400
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.368951612903226e-05,
      "loss": 4.7153,
      "step": 1410
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.385752688172043e-05,
      "loss": 4.7031,
      "step": 1420
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.4025537634408606e-05,
      "loss": 4.6676,
      "step": 1430
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.4193548387096777e-05,
      "loss": 4.7177,
      "step": 1440
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.4361559139784948e-05,
      "loss": 4.8695,
      "step": 1450
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.452956989247312e-05,
      "loss": 4.8219,
      "step": 1460
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.4697580645161293e-05,
      "loss": 4.7481,
      "step": 1470
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.4865591397849464e-05,
      "loss": 4.8556,
      "step": 1480
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.503360215053764e-05,
      "loss": 4.7572,
      "step": 1490
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.5201612903225806e-05,
      "loss": 4.6192,
      "step": 1500
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.536962365591398e-05,
      "loss": 4.723,
      "step": 1510
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.5537634408602152e-05,
      "loss": 4.8091,
      "step": 1520
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.5705645161290326e-05,
      "loss": 4.7191,
      "step": 1530
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.5873655913978494e-05,
      "loss": 4.658,
      "step": 1540
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.604166666666667e-05,
      "loss": 4.8029,
      "step": 1550
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.620967741935484e-05,
      "loss": 4.5458,
      "step": 1560
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.6377688172043014e-05,
      "loss": 4.9067,
      "step": 1570
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.654569892473118e-05,
      "loss": 4.4242,
      "step": 1580
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.6713709677419356e-05,
      "loss": 4.6519,
      "step": 1590
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.6881720430107527e-05,
      "loss": 4.6038,
      "step": 1600
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.70497311827957e-05,
      "loss": 4.7452,
      "step": 1610
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.721774193548387e-05,
      "loss": 4.707,
      "step": 1620
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.7385752688172043e-05,
      "loss": 4.6429,
      "step": 1630
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.7553763440860214e-05,
      "loss": 4.8403,
      "step": 1640
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.772177419354839e-05,
      "loss": 4.5688,
      "step": 1650
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.7889784946236556e-05,
      "loss": 4.7081,
      "step": 1660
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.8057795698924734e-05,
      "loss": 4.6276,
      "step": 1670
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.822580645161291e-05,
      "loss": 4.6528,
      "step": 1680
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.8393817204301076e-05,
      "loss": 4.8381,
      "step": 1690
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.856182795698925e-05,
      "loss": 4.9323,
      "step": 1700
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.8729838709677422e-05,
      "loss": 4.5868,
      "step": 1710
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.8897849462365596e-05,
      "loss": 4.9298,
      "step": 1720
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.9065860215053764e-05,
      "loss": 4.6908,
      "step": 1730
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.9233870967741938e-05,
      "loss": 4.6427,
      "step": 1740
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.940188172043011e-05,
      "loss": 4.6611,
      "step": 1750
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.9569892473118284e-05,
      "loss": 4.6569,
      "step": 1760
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.973790322580645e-05,
      "loss": 4.4677,
      "step": 1770
    },
    {
      "epoch": 0.06,
      "learning_rate": 2.9905913978494626e-05,
      "loss": 4.8696,
      "step": 1780
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.0073924731182797e-05,
      "loss": 4.6955,
      "step": 1790
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.024193548387097e-05,
      "loss": 4.5123,
      "step": 1800
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.040994623655914e-05,
      "loss": 4.8245,
      "step": 1810
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.0577956989247317e-05,
      "loss": 4.6848,
      "step": 1820
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.074596774193548e-05,
      "loss": 4.6223,
      "step": 1830
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.091397849462366e-05,
      "loss": 4.8519,
      "step": 1840
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.108198924731183e-05,
      "loss": 4.6654,
      "step": 1850
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.125e-05,
      "loss": 4.7357,
      "step": 1860
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.141801075268818e-05,
      "loss": 4.7567,
      "step": 1870
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.158602150537634e-05,
      "loss": 4.7637,
      "step": 1880
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.175403225806452e-05,
      "loss": 4.6489,
      "step": 1890
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.192204301075269e-05,
      "loss": 4.7715,
      "step": 1900
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.209005376344086e-05,
      "loss": 4.5322,
      "step": 1910
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.2258064516129034e-05,
      "loss": 4.7145,
      "step": 1920
    },
    {
      "epoch": 0.06,
      "learning_rate": 3.2426075268817205e-05,
      "loss": 4.8788,
      "step": 1930
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.2594086021505376e-05,
      "loss": 4.7815,
      "step": 1940
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.2762096774193553e-05,
      "loss": 4.6122,
      "step": 1950
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.293010752688172e-05,
      "loss": 4.6607,
      "step": 1960
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.3098118279569895e-05,
      "loss": 4.7085,
      "step": 1970
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.3266129032258067e-05,
      "loss": 4.7526,
      "step": 1980
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.343413978494624e-05,
      "loss": 4.4635,
      "step": 1990
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.360215053763441e-05,
      "loss": 4.8464,
      "step": 2000
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.377016129032258e-05,
      "loss": 4.8352,
      "step": 2010
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.393817204301075e-05,
      "loss": 4.7684,
      "step": 2020
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.410618279569893e-05,
      "loss": 4.4726,
      "step": 2030
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.427419354838709e-05,
      "loss": 4.8434,
      "step": 2040
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.444220430107527e-05,
      "loss": 4.5217,
      "step": 2050
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.461021505376345e-05,
      "loss": 4.5219,
      "step": 2060
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.477822580645161e-05,
      "loss": 4.7795,
      "step": 2070
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.494623655913979e-05,
      "loss": 4.1376,
      "step": 2080
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.511424731182796e-05,
      "loss": 4.9154,
      "step": 2090
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.528225806451613e-05,
      "loss": 4.7189,
      "step": 2100
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.5450268817204303e-05,
      "loss": 4.8433,
      "step": 2110
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.5618279569892474e-05,
      "loss": 4.7001,
      "step": 2120
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.5786290322580645e-05,
      "loss": 4.6825,
      "step": 2130
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.595430107526882e-05,
      "loss": 4.7596,
      "step": 2140
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.612231182795699e-05,
      "loss": 4.7217,
      "step": 2150
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.6290322580645165e-05,
      "loss": 4.7864,
      "step": 2160
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.6458333333333336e-05,
      "loss": 4.8844,
      "step": 2170
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.662634408602151e-05,
      "loss": 4.828,
      "step": 2180
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.679435483870968e-05,
      "loss": 4.5349,
      "step": 2190
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.696236559139785e-05,
      "loss": 4.7096,
      "step": 2200
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.713037634408602e-05,
      "loss": 4.7755,
      "step": 2210
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.72983870967742e-05,
      "loss": 4.7531,
      "step": 2220
    },
    {
      "epoch": 0.07,
      "learning_rate": 3.746639784946236e-05,
      "loss": 4.5409,
      "step": 2230
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.763440860215054e-05,
      "loss": 4.6511,
      "step": 2240
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.780241935483871e-05,
      "loss": 4.5239,
      "step": 2250
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.797043010752688e-05,
      "loss": 4.6094,
      "step": 2260
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.813844086021506e-05,
      "loss": 4.4195,
      "step": 2270
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.8306451612903224e-05,
      "loss": 4.5219,
      "step": 2280
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.84744623655914e-05,
      "loss": 4.7089,
      "step": 2290
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.864247311827957e-05,
      "loss": 4.7481,
      "step": 2300
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.8810483870967744e-05,
      "loss": 4.7719,
      "step": 2310
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.8978494623655915e-05,
      "loss": 4.4219,
      "step": 2320
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.9146505376344086e-05,
      "loss": 4.4863,
      "step": 2330
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.931451612903226e-05,
      "loss": 4.6597,
      "step": 2340
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.9482526881720435e-05,
      "loss": 4.5431,
      "step": 2350
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.96505376344086e-05,
      "loss": 4.8282,
      "step": 2360
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.981854838709678e-05,
      "loss": 4.6723,
      "step": 2370
    },
    {
      "epoch": 0.08,
      "learning_rate": 3.998655913978495e-05,
      "loss": 4.7072,
      "step": 2380
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.015456989247312e-05,
      "loss": 4.405,
      "step": 2390
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.032258064516129e-05,
      "loss": 4.6087,
      "step": 2400
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.049059139784946e-05,
      "loss": 4.4848,
      "step": 2410
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.065860215053764e-05,
      "loss": 4.7085,
      "step": 2420
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.082661290322581e-05,
      "loss": 4.5809,
      "step": 2430
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.099462365591398e-05,
      "loss": 4.5694,
      "step": 2440
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.116263440860215e-05,
      "loss": 4.6146,
      "step": 2450
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.133064516129033e-05,
      "loss": 4.6111,
      "step": 2460
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.1498655913978494e-05,
      "loss": 4.5915,
      "step": 2470
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.166666666666667e-05,
      "loss": 4.5425,
      "step": 2480
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.1834677419354836e-05,
      "loss": 4.344,
      "step": 2490
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.2002688172043014e-05,
      "loss": 4.3079,
      "step": 2500
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.2170698924731185e-05,
      "loss": 4.3301,
      "step": 2510
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.2338709677419356e-05,
      "loss": 4.7686,
      "step": 2520
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.250672043010753e-05,
      "loss": 4.6731,
      "step": 2530
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.2674731182795705e-05,
      "loss": 4.3259,
      "step": 2540
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.284274193548387e-05,
      "loss": 4.8506,
      "step": 2550
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.301075268817205e-05,
      "loss": 4.5715,
      "step": 2560
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.317876344086022e-05,
      "loss": 4.6516,
      "step": 2570
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.334677419354839e-05,
      "loss": 4.6917,
      "step": 2580
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.351478494623656e-05,
      "loss": 4.6217,
      "step": 2590
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.368279569892473e-05,
      "loss": 4.568,
      "step": 2600
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.385080645161291e-05,
      "loss": 4.5524,
      "step": 2610
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.401881720430108e-05,
      "loss": 4.4783,
      "step": 2620
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.418682795698925e-05,
      "loss": 4.6619,
      "step": 2630
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.435483870967742e-05,
      "loss": 4.5148,
      "step": 2640
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.452284946236559e-05,
      "loss": 4.5686,
      "step": 2650
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.4690860215053764e-05,
      "loss": 4.6879,
      "step": 2660
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.485887096774194e-05,
      "loss": 4.7926,
      "step": 2670
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.5026881720430106e-05,
      "loss": 4.4609,
      "step": 2680
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.5194892473118284e-05,
      "loss": 4.281,
      "step": 2690
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.5362903225806455e-05,
      "loss": 4.4718,
      "step": 2700
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.5530913978494626e-05,
      "loss": 4.6328,
      "step": 2710
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.56989247311828e-05,
      "loss": 4.5215,
      "step": 2720
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.586693548387097e-05,
      "loss": 4.3572,
      "step": 2730
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.603494623655914e-05,
      "loss": 4.6574,
      "step": 2740
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.620295698924732e-05,
      "loss": 4.6526,
      "step": 2750
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.637096774193548e-05,
      "loss": 4.716,
      "step": 2760
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.653897849462366e-05,
      "loss": 4.4768,
      "step": 2770
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.670698924731183e-05,
      "loss": 4.6641,
      "step": 2780
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.6875e-05,
      "loss": 4.2275,
      "step": 2790
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.704301075268818e-05,
      "loss": 4.4817,
      "step": 2800
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.721102150537634e-05,
      "loss": 4.5784,
      "step": 2810
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.737903225806452e-05,
      "loss": 4.4939,
      "step": 2820
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.754704301075269e-05,
      "loss": 4.623,
      "step": 2830
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.771505376344086e-05,
      "loss": 4.4391,
      "step": 2840
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.7883064516129034e-05,
      "loss": 4.5203,
      "step": 2850
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.805107526881721e-05,
      "loss": 4.2879,
      "step": 2860
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8219086021505376e-05,
      "loss": 4.5818,
      "step": 2870
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8387096774193554e-05,
      "loss": 4.4625,
      "step": 2880
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.855510752688172e-05,
      "loss": 4.6607,
      "step": 2890
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8723118279569896e-05,
      "loss": 4.2406,
      "step": 2900
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.889112903225807e-05,
      "loss": 4.4246,
      "step": 2910
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.905913978494624e-05,
      "loss": 4.6208,
      "step": 2920
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.922715053763441e-05,
      "loss": 4.3164,
      "step": 2930
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.939516129032259e-05,
      "loss": 4.3962,
      "step": 2940
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.956317204301075e-05,
      "loss": 4.7987,
      "step": 2950
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.973118279569893e-05,
      "loss": 4.5732,
      "step": 2960
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.98991935483871e-05,
      "loss": 4.4208,
      "step": 2970
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9992532855436084e-05,
      "loss": 4.7173,
      "step": 2980
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.997386499402629e-05,
      "loss": 4.4456,
      "step": 2990
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.995519713261649e-05,
      "loss": 4.2572,
      "step": 3000
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.993652927120669e-05,
      "loss": 4.7157,
      "step": 3010
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9917861409796897e-05,
      "loss": 4.4885,
      "step": 3020
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.98991935483871e-05,
      "loss": 4.2594,
      "step": 3030
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.98805256869773e-05,
      "loss": 4.6185,
      "step": 3040
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9861857825567506e-05,
      "loss": 4.5417,
      "step": 3050
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.984318996415771e-05,
      "loss": 4.4005,
      "step": 3060
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.982452210274791e-05,
      "loss": 4.7185,
      "step": 3070
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.980585424133811e-05,
      "loss": 4.5395,
      "step": 3080
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.978718637992832e-05,
      "loss": 4.331,
      "step": 3090
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.976851851851852e-05,
      "loss": 4.4664,
      "step": 3100
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.9749850657108726e-05,
      "loss": 4.3807,
      "step": 3110
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.973118279569893e-05,
      "loss": 4.2413,
      "step": 3120
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9712514934289125e-05,
      "loss": 4.5774,
      "step": 3130
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9693847072879335e-05,
      "loss": 4.5053,
      "step": 3140
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.967517921146954e-05,
      "loss": 4.6295,
      "step": 3150
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.965651135005974e-05,
      "loss": 4.775,
      "step": 3160
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9637843488649945e-05,
      "loss": 5.0108,
      "step": 3170
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.961917562724014e-05,
      "loss": 4.3621,
      "step": 3180
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.960050776583035e-05,
      "loss": 4.4008,
      "step": 3190
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9581839904420555e-05,
      "loss": 4.4939,
      "step": 3200
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.956317204301075e-05,
      "loss": 4.5212,
      "step": 3210
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.954450418160096e-05,
      "loss": 4.6402,
      "step": 3220
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.952583632019116e-05,
      "loss": 4.4374,
      "step": 3230
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.950716845878137e-05,
      "loss": 4.052,
      "step": 3240
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.948850059737157e-05,
      "loss": 4.3175,
      "step": 3250
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.946983273596177e-05,
      "loss": 4.4946,
      "step": 3260
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.945116487455198e-05,
      "loss": 4.5503,
      "step": 3270
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9432497013142173e-05,
      "loss": 4.4767,
      "step": 3280
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9413829151732383e-05,
      "loss": 4.3991,
      "step": 3290
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.939516129032259e-05,
      "loss": 4.6221,
      "step": 3300
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.937649342891278e-05,
      "loss": 4.4263,
      "step": 3310
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.935782556750299e-05,
      "loss": 4.4661,
      "step": 3320
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.933915770609319e-05,
      "loss": 4.7345,
      "step": 3330
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.932048984468339e-05,
      "loss": 4.5323,
      "step": 3340
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.93018219832736e-05,
      "loss": 4.5434,
      "step": 3350
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.92831541218638e-05,
      "loss": 4.5242,
      "step": 3360
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.926448626045401e-05,
      "loss": 4.4653,
      "step": 3370
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9245818399044206e-05,
      "loss": 4.2482,
      "step": 3380
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.922715053763441e-05,
      "loss": 4.4161,
      "step": 3390
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.920848267622461e-05,
      "loss": 4.459,
      "step": 3400
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9189814814814815e-05,
      "loss": 4.6974,
      "step": 3410
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.9171146953405025e-05,
      "loss": 4.7484,
      "step": 3420
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.915247909199522e-05,
      "loss": 4.553,
      "step": 3430
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9133811230585425e-05,
      "loss": 4.5069,
      "step": 3440
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.911514336917563e-05,
      "loss": 4.5806,
      "step": 3450
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.909647550776583e-05,
      "loss": 4.3154,
      "step": 3460
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9077807646356035e-05,
      "loss": 4.4244,
      "step": 3470
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.905913978494624e-05,
      "loss": 4.4991,
      "step": 3480
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.904047192353644e-05,
      "loss": 4.1619,
      "step": 3490
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9021804062126644e-05,
      "loss": 4.2876,
      "step": 3500
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.900313620071685e-05,
      "loss": 4.2716,
      "step": 3510
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.898446833930705e-05,
      "loss": 4.2517,
      "step": 3520
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8965800477897254e-05,
      "loss": 4.4254,
      "step": 3530
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.894713261648746e-05,
      "loss": 4.5187,
      "step": 3540
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.892846475507766e-05,
      "loss": 4.3443,
      "step": 3550
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8909796893667864e-05,
      "loss": 4.4591,
      "step": 3560
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.889112903225807e-05,
      "loss": 4.2675,
      "step": 3570
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.887246117084827e-05,
      "loss": 4.4332,
      "step": 3580
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.885379330943847e-05,
      "loss": 4.6878,
      "step": 3590
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8835125448028677e-05,
      "loss": 4.2547,
      "step": 3600
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.881645758661888e-05,
      "loss": 4.485,
      "step": 3610
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.879778972520908e-05,
      "loss": 4.3487,
      "step": 3620
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8779121863799286e-05,
      "loss": 4.081,
      "step": 3630
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.876045400238949e-05,
      "loss": 4.349,
      "step": 3640
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.874178614097969e-05,
      "loss": 4.3416,
      "step": 3650
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8723118279569896e-05,
      "loss": 4.3979,
      "step": 3660
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.87044504181601e-05,
      "loss": 4.4437,
      "step": 3670
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.86857825567503e-05,
      "loss": 4.5158,
      "step": 3680
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8667114695340505e-05,
      "loss": 4.3471,
      "step": 3690
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.864844683393071e-05,
      "loss": 4.4189,
      "step": 3700
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.862977897252091e-05,
      "loss": 4.4054,
      "step": 3710
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8611111111111115e-05,
      "loss": 4.3472,
      "step": 3720
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.859244324970131e-05,
      "loss": 4.2868,
      "step": 3730
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.857377538829152e-05,
      "loss": 4.5381,
      "step": 3740
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.855510752688172e-05,
      "loss": 4.2602,
      "step": 3750
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.853643966547193e-05,
      "loss": 4.393,
      "step": 3760
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.851777180406213e-05,
      "loss": 4.155,
      "step": 3770
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.849910394265233e-05,
      "loss": 4.5707,
      "step": 3780
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.848043608124254e-05,
      "loss": 4.2793,
      "step": 3790
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8461768219832734e-05,
      "loss": 4.2988,
      "step": 3800
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8443100358422944e-05,
      "loss": 4.5829,
      "step": 3810
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.842443249701315e-05,
      "loss": 4.2576,
      "step": 3820
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8405764635603344e-05,
      "loss": 4.3333,
      "step": 3830
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8387096774193554e-05,
      "loss": 4.2648,
      "step": 3840
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.836842891278375e-05,
      "loss": 4.1792,
      "step": 3850
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.834976105137396e-05,
      "loss": 4.2942,
      "step": 3860
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8331093189964163e-05,
      "loss": 4.3512,
      "step": 3870
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.831242532855436e-05,
      "loss": 4.6654,
      "step": 3880
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.829375746714457e-05,
      "loss": 4.3735,
      "step": 3890
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8275089605734766e-05,
      "loss": 4.3627,
      "step": 3900
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.825642174432497e-05,
      "loss": 4.4475,
      "step": 3910
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.823775388291518e-05,
      "loss": 4.3226,
      "step": 3920
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8219086021505376e-05,
      "loss": 4.2341,
      "step": 3930
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8200418160095586e-05,
      "loss": 4.2264,
      "step": 3940
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.818175029868578e-05,
      "loss": 4.1749,
      "step": 3950
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8163082437275986e-05,
      "loss": 4.4451,
      "step": 3960
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8144414575866196e-05,
      "loss": 4.4556,
      "step": 3970
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.812574671445639e-05,
      "loss": 4.3421,
      "step": 3980
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.81070788530466e-05,
      "loss": 4.4947,
      "step": 3990
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.80884109916368e-05,
      "loss": 4.2745,
      "step": 4000
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8069743130227e-05,
      "loss": 4.0593,
      "step": 4010
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.805107526881721e-05,
      "loss": 4.3374,
      "step": 4020
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.803240740740741e-05,
      "loss": 4.2332,
      "step": 4030
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.801373954599761e-05,
      "loss": 4.407,
      "step": 4040
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7995071684587815e-05,
      "loss": 3.8267,
      "step": 4050
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.797640382317802e-05,
      "loss": 4.2591,
      "step": 4060
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.795773596176822e-05,
      "loss": 4.2383,
      "step": 4070
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7939068100358424e-05,
      "loss": 3.9536,
      "step": 4080
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.792040023894863e-05,
      "loss": 4.5887,
      "step": 4090
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.790173237753883e-05,
      "loss": 4.3877,
      "step": 4100
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7883064516129034e-05,
      "loss": 4.3213,
      "step": 4110
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.786439665471924e-05,
      "loss": 4.0475,
      "step": 4120
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.784572879330944e-05,
      "loss": 4.3858,
      "step": 4130
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7827060931899644e-05,
      "loss": 4.2672,
      "step": 4140
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.780839307048985e-05,
      "loss": 4.4997,
      "step": 4150
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.778972520908005e-05,
      "loss": 4.3638,
      "step": 4160
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.777105734767025e-05,
      "loss": 4.3172,
      "step": 4170
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7752389486260456e-05,
      "loss": 4.068,
      "step": 4180
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.773372162485066e-05,
      "loss": 4.4311,
      "step": 4190
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.771505376344086e-05,
      "loss": 4.5892,
      "step": 4200
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7696385902031066e-05,
      "loss": 4.2604,
      "step": 4210
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.767771804062127e-05,
      "loss": 4.2857,
      "step": 4220
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.765905017921147e-05,
      "loss": 4.3303,
      "step": 4230
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7640382317801676e-05,
      "loss": 4.2141,
      "step": 4240
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.762171445639188e-05,
      "loss": 4.2023,
      "step": 4250
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.760304659498208e-05,
      "loss": 4.2967,
      "step": 4260
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.7584378733572285e-05,
      "loss": 4.1328,
      "step": 4270
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.756571087216249e-05,
      "loss": 4.3922,
      "step": 4280
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.754704301075269e-05,
      "loss": 4.3267,
      "step": 4290
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.752837514934289e-05,
      "loss": 4.3216,
      "step": 4300
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.75097072879331e-05,
      "loss": 4.1175,
      "step": 4310
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.74910394265233e-05,
      "loss": 4.1491,
      "step": 4320
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7472371565113505e-05,
      "loss": 4.6231,
      "step": 4330
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.745370370370371e-05,
      "loss": 4.0371,
      "step": 4340
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7435035842293904e-05,
      "loss": 4.3377,
      "step": 4350
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7416367980884114e-05,
      "loss": 4.4208,
      "step": 4360
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.739770011947431e-05,
      "loss": 4.3935,
      "step": 4370
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.737903225806452e-05,
      "loss": 4.0609,
      "step": 4380
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7360364396654724e-05,
      "loss": 4.2408,
      "step": 4390
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.734169653524492e-05,
      "loss": 4.5171,
      "step": 4400
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.732302867383513e-05,
      "loss": 4.2476,
      "step": 4410
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.730436081242533e-05,
      "loss": 4.3547,
      "step": 4420
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.728569295101554e-05,
      "loss": 4.0966,
      "step": 4430
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.726702508960574e-05,
      "loss": 4.1857,
      "step": 4440
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7248357228195937e-05,
      "loss": 4.352,
      "step": 4450
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7229689366786147e-05,
      "loss": 4.1158,
      "step": 4460
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.721102150537634e-05,
      "loss": 4.1918,
      "step": 4470
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7192353643966546e-05,
      "loss": 4.2308,
      "step": 4480
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7173685782556756e-05,
      "loss": 4.3485,
      "step": 4490
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.715501792114695e-05,
      "loss": 4.094,
      "step": 4500
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.713635005973716e-05,
      "loss": 4.0731,
      "step": 4510
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.711768219832736e-05,
      "loss": 4.3649,
      "step": 4520
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.709901433691756e-05,
      "loss": 3.8606,
      "step": 4530
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.708034647550777e-05,
      "loss": 4.3086,
      "step": 4540
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.706167861409797e-05,
      "loss": 4.1081,
      "step": 4550
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.704301075268818e-05,
      "loss": 4.3555,
      "step": 4560
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.7024342891278375e-05,
      "loss": 3.8921,
      "step": 4570
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.700567502986858e-05,
      "loss": 4.4176,
      "step": 4580
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.698700716845879e-05,
      "loss": 4.6405,
      "step": 4590
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.6968339307048985e-05,
      "loss": 4.0693,
      "step": 4600
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.694967144563919e-05,
      "loss": 4.5621,
      "step": 4610
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.693100358422939e-05,
      "loss": 4.3265,
      "step": 4620
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6912335722819595e-05,
      "loss": 4.2263,
      "step": 4630
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6893667861409805e-05,
      "loss": 4.4782,
      "step": 4640
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6875e-05,
      "loss": 4.5211,
      "step": 4650
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6856332138590204e-05,
      "loss": 4.4426,
      "step": 4660
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.683766427718041e-05,
      "loss": 4.3813,
      "step": 4670
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.681899641577061e-05,
      "loss": 3.9564,
      "step": 4680
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6800328554360814e-05,
      "loss": 4.4241,
      "step": 4690
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.678166069295102e-05,
      "loss": 4.3802,
      "step": 4700
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.676299283154122e-05,
      "loss": 4.2065,
      "step": 4710
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6744324970131424e-05,
      "loss": 4.0179,
      "step": 4720
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.672565710872163e-05,
      "loss": 4.445,
      "step": 4730
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.670698924731183e-05,
      "loss": 4.3923,
      "step": 4740
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.668832138590203e-05,
      "loss": 3.9034,
      "step": 4750
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6669653524492236e-05,
      "loss": 4.0292,
      "step": 4760
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.665098566308244e-05,
      "loss": 4.4023,
      "step": 4770
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.663231780167264e-05,
      "loss": 3.7686,
      "step": 4780
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6613649940262846e-05,
      "loss": 4.0862,
      "step": 4790
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.659498207885305e-05,
      "loss": 4.0931,
      "step": 4800
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.657631421744325e-05,
      "loss": 4.3211,
      "step": 4810
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6557646356033456e-05,
      "loss": 4.1455,
      "step": 4820
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.653897849462366e-05,
      "loss": 4.396,
      "step": 4830
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.652031063321386e-05,
      "loss": 4.1785,
      "step": 4840
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6501642771804065e-05,
      "loss": 4.2009,
      "step": 4850
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.648297491039427e-05,
      "loss": 4.2842,
      "step": 4860
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6464307048984465e-05,
      "loss": 4.4221,
      "step": 4870
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6445639187574675e-05,
      "loss": 4.1758,
      "step": 4880
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.642697132616488e-05,
      "loss": 4.3337,
      "step": 4890
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.640830346475508e-05,
      "loss": 4.5078,
      "step": 4900
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.6389635603345285e-05,
      "loss": 4.0794,
      "step": 4910
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.637096774193548e-05,
      "loss": 4.1977,
      "step": 4920
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.635229988052569e-05,
      "loss": 4.1703,
      "step": 4930
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.6333632019115894e-05,
      "loss": 4.0087,
      "step": 4940
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.63149641577061e-05,
      "loss": 4.2186,
      "step": 4950
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.62962962962963e-05,
      "loss": 4.2193,
      "step": 4960
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.62776284348865e-05,
      "loss": 4.0618,
      "step": 4970
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.625896057347671e-05,
      "loss": 4.2286,
      "step": 4980
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.624029271206691e-05,
      "loss": 4.4534,
      "step": 4990
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.6221624850657114e-05,
      "loss": 4.1912,
      "step": 5000
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.620295698924732e-05,
      "loss": 4.2098,
      "step": 5010
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.618428912783751e-05,
      "loss": 4.4808,
      "step": 5020
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.616562126642772e-05,
      "loss": 4.3189,
      "step": 5030
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.614695340501792e-05,
      "loss": 3.9401,
      "step": 5040
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.612828554360812e-05,
      "loss": 4.1829,
      "step": 5050
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.610961768219833e-05,
      "loss": 4.4423,
      "step": 5060
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.609094982078853e-05,
      "loss": 4.0158,
      "step": 5070
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.607228195937874e-05,
      "loss": 3.8608,
      "step": 5080
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.6053614097968936e-05,
      "loss": 4.0689,
      "step": 5090
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.603494623655914e-05,
      "loss": 4.2897,
      "step": 5100
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.601627837514935e-05,
      "loss": 4.2973,
      "step": 5110
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.5997610513739546e-05,
      "loss": 4.1113,
      "step": 5120
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.5978942652329756e-05,
      "loss": 4.2283,
      "step": 5130
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.596027479091995e-05,
      "loss": 4.2546,
      "step": 5140
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.5941606929510155e-05,
      "loss": 4.4958,
      "step": 5150
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.5922939068100365e-05,
      "loss": 4.1367,
      "step": 5160
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.590427120669056e-05,
      "loss": 3.6439,
      "step": 5170
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.5885603345280765e-05,
      "loss": 4.2327,
      "step": 5180
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.586693548387097e-05,
      "loss": 4.0844,
      "step": 5190
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.584826762246117e-05,
      "loss": 4.2107,
      "step": 5200
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.582959976105138e-05,
      "loss": 4.2222,
      "step": 5210
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.581093189964158e-05,
      "loss": 4.1383,
      "step": 5220
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.579226403823178e-05,
      "loss": 4.145,
      "step": 5230
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5773596176821984e-05,
      "loss": 4.0463,
      "step": 5240
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.575492831541219e-05,
      "loss": 4.3782,
      "step": 5250
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.57362604540024e-05,
      "loss": 4.4342,
      "step": 5260
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5717592592592594e-05,
      "loss": 3.8822,
      "step": 5270
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.56989247311828e-05,
      "loss": 4.1924,
      "step": 5280
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5680256869773e-05,
      "loss": 4.3946,
      "step": 5290
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5661589008363203e-05,
      "loss": 4.2242,
      "step": 5300
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.564292114695341e-05,
      "loss": 4.3488,
      "step": 5310
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.562425328554361e-05,
      "loss": 4.2119,
      "step": 5320
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.560558542413381e-05,
      "loss": 4.3024,
      "step": 5330
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5586917562724016e-05,
      "loss": 3.9396,
      "step": 5340
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.556824970131422e-05,
      "loss": 3.9281,
      "step": 5350
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.554958183990442e-05,
      "loss": 4.232,
      "step": 5360
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5530913978494626e-05,
      "loss": 3.6128,
      "step": 5370
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.551224611708483e-05,
      "loss": 4.0242,
      "step": 5380
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.549357825567503e-05,
      "loss": 3.8623,
      "step": 5390
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5474910394265236e-05,
      "loss": 4.1662,
      "step": 5400
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.545624253285544e-05,
      "loss": 3.9453,
      "step": 5410
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.543757467144564e-05,
      "loss": 4.2344,
      "step": 5420
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5418906810035845e-05,
      "loss": 4.2854,
      "step": 5430
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.540023894862604e-05,
      "loss": 4.0162,
      "step": 5440
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.538157108721625e-05,
      "loss": 4.3484,
      "step": 5450
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.5362903225806455e-05,
      "loss": 4.3394,
      "step": 5460
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.534423536439666e-05,
      "loss": 3.9772,
      "step": 5470
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.532556750298686e-05,
      "loss": 4.2683,
      "step": 5480
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.530689964157706e-05,
      "loss": 4.2409,
      "step": 5490
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.528823178016727e-05,
      "loss": 4.1268,
      "step": 5500
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.526956391875747e-05,
      "loss": 4.1671,
      "step": 5510
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5250896057347674e-05,
      "loss": 3.6939,
      "step": 5520
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.523222819593788e-05,
      "loss": 4.0335,
      "step": 5530
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5213560334528074e-05,
      "loss": 4.1694,
      "step": 5540
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5194892473118284e-05,
      "loss": 4.0742,
      "step": 5550
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.517622461170849e-05,
      "loss": 4.1111,
      "step": 5560
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.515755675029869e-05,
      "loss": 4.1056,
      "step": 5570
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5138888888888894e-05,
      "loss": 4.1699,
      "step": 5580
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.512022102747909e-05,
      "loss": 4.1817,
      "step": 5590
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.51015531660693e-05,
      "loss": 3.8425,
      "step": 5600
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.50828853046595e-05,
      "loss": 4.0771,
      "step": 5610
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.50642174432497e-05,
      "loss": 3.8866,
      "step": 5620
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.504554958183991e-05,
      "loss": 4.0063,
      "step": 5630
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5026881720430106e-05,
      "loss": 3.9252,
      "step": 5640
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.5008213859020316e-05,
      "loss": 4.1584,
      "step": 5650
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.498954599761052e-05,
      "loss": 4.1171,
      "step": 5660
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.4970878136200716e-05,
      "loss": 4.1672,
      "step": 5670
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.4952210274790926e-05,
      "loss": 4.5783,
      "step": 5680
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.493354241338112e-05,
      "loss": 3.8408,
      "step": 5690
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.491487455197133e-05,
      "loss": 3.8826,
      "step": 5700
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.489620669056153e-05,
      "loss": 4.0903,
      "step": 5710
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.487753882915173e-05,
      "loss": 3.8766,
      "step": 5720
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.485887096774194e-05,
      "loss": 4.1665,
      "step": 5730
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.484020310633214e-05,
      "loss": 4.4071,
      "step": 5740
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.482153524492234e-05,
      "loss": 4.1926,
      "step": 5750
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.4802867383512545e-05,
      "loss": 4.1452,
      "step": 5760
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.478419952210275e-05,
      "loss": 4.1673,
      "step": 5770
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.476553166069296e-05,
      "loss": 4.1794,
      "step": 5780
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.4746863799283154e-05,
      "loss": 3.9736,
      "step": 5790
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.472819593787336e-05,
      "loss": 4.1541,
      "step": 5800
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.470952807646356e-05,
      "loss": 4.1792,
      "step": 5810
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4690860215053764e-05,
      "loss": 4.1334,
      "step": 5820
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4672192353643974e-05,
      "loss": 4.0146,
      "step": 5830
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.465352449223417e-05,
      "loss": 3.9846,
      "step": 5840
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4634856630824374e-05,
      "loss": 4.2558,
      "step": 5850
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.461618876941458e-05,
      "loss": 4.2142,
      "step": 5860
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.459752090800478e-05,
      "loss": 3.9264,
      "step": 5870
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4578853046594983e-05,
      "loss": 3.9934,
      "step": 5880
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.456018518518519e-05,
      "loss": 4.0109,
      "step": 5890
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.454151732377539e-05,
      "loss": 4.5819,
      "step": 5900
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.452284946236559e-05,
      "loss": 4.112,
      "step": 5910
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4504181600955796e-05,
      "loss": 4.11,
      "step": 5920
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4485513739546e-05,
      "loss": 4.2611,
      "step": 5930
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.44668458781362e-05,
      "loss": 4.2696,
      "step": 5940
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.4448178016726406e-05,
      "loss": 4.1471,
      "step": 5950
    },
    {
      "epoch": 0.2,
      "eval_accuracy": 0.03822574828705373,
      "eval_loss": 4.160111904144287,
      "eval_runtime": 1070.1177,
      "eval_samples_per_second": 2.591,
      "eval_steps_per_second": 1.296,
      "step": 5952
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.442951015531661e-05,
      "loss": 4.1187,
      "step": 5960
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.441084229390681e-05,
      "loss": 4.1273,
      "step": 5970
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.4392174432497016e-05,
      "loss": 3.9333,
      "step": 5980
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.437350657108722e-05,
      "loss": 3.8852,
      "step": 5990
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.435483870967742e-05,
      "loss": 4.3789,
      "step": 6000
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.4336170848267625e-05,
      "loss": 4.0991,
      "step": 6010
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.431750298685783e-05,
      "loss": 3.9749,
      "step": 6020
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.429883512544803e-05,
      "loss": 4.0504,
      "step": 6030
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.4280167264038235e-05,
      "loss": 4.2472,
      "step": 6040
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.426149940262844e-05,
      "loss": 4.1108,
      "step": 6050
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.4242831541218635e-05,
      "loss": 3.9754,
      "step": 6060
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.4224163679808845e-05,
      "loss": 3.9436,
      "step": 6070
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.420549581839905e-05,
      "loss": 4.1744,
      "step": 6080
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.418682795698925e-05,
      "loss": 4.377,
      "step": 6090
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.4168160095579454e-05,
      "loss": 4.3518,
      "step": 6100
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.414949223416965e-05,
      "loss": 4.1371,
      "step": 6110
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.413082437275986e-05,
      "loss": 4.0343,
      "step": 6120
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.4112156511350064e-05,
      "loss": 4.198,
      "step": 6130
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.409348864994026e-05,
      "loss": 4.0281,
      "step": 6140
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.407482078853047e-05,
      "loss": 4.2912,
      "step": 6150
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.405615292712067e-05,
      "loss": 3.6893,
      "step": 6160
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.403748506571088e-05,
      "loss": 4.6309,
      "step": 6170
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.401881720430108e-05,
      "loss": 4.4654,
      "step": 6180
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.4000149342891276e-05,
      "loss": 3.9596,
      "step": 6190
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3981481481481486e-05,
      "loss": 4.1032,
      "step": 6200
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.396281362007168e-05,
      "loss": 3.9217,
      "step": 6210
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.394414575866189e-05,
      "loss": 3.861,
      "step": 6220
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3925477897252096e-05,
      "loss": 4.1733,
      "step": 6230
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.390681003584229e-05,
      "loss": 4.2278,
      "step": 6240
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.38881421744325e-05,
      "loss": 4.2659,
      "step": 6250
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.38694743130227e-05,
      "loss": 4.0317,
      "step": 6260
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.385080645161291e-05,
      "loss": 4.4773,
      "step": 6270
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.383213859020311e-05,
      "loss": 3.9262,
      "step": 6280
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.381347072879331e-05,
      "loss": 3.8025,
      "step": 6290
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.379480286738352e-05,
      "loss": 4.1975,
      "step": 6300
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3776135005973715e-05,
      "loss": 4.1544,
      "step": 6310
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.375746714456392e-05,
      "loss": 4.0071,
      "step": 6320
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.373879928315413e-05,
      "loss": 3.8359,
      "step": 6330
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3720131421744325e-05,
      "loss": 3.814,
      "step": 6340
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3701463560334535e-05,
      "loss": 3.8951,
      "step": 6350
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.368279569892473e-05,
      "loss": 3.9571,
      "step": 6360
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.3664127837514934e-05,
      "loss": 4.1305,
      "step": 6370
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.364545997610514e-05,
      "loss": 3.6083,
      "step": 6380
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.362679211469534e-05,
      "loss": 4.2703,
      "step": 6390
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.360812425328555e-05,
      "loss": 4.0742,
      "step": 6400
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.358945639187575e-05,
      "loss": 3.394,
      "step": 6410
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.357078853046595e-05,
      "loss": 4.2712,
      "step": 6420
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3552120669056154e-05,
      "loss": 4.4349,
      "step": 6430
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.353345280764636e-05,
      "loss": 4.148,
      "step": 6440
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.351478494623656e-05,
      "loss": 3.9457,
      "step": 6450
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.349611708482676e-05,
      "loss": 3.9034,
      "step": 6460
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3477449223416967e-05,
      "loss": 4.2363,
      "step": 6470
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.345878136200717e-05,
      "loss": 4.2118,
      "step": 6480
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.344011350059737e-05,
      "loss": 3.8898,
      "step": 6490
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3421445639187576e-05,
      "loss": 4.0214,
      "step": 6500
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.340277777777778e-05,
      "loss": 4.164,
      "step": 6510
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.338410991636798e-05,
      "loss": 4.0053,
      "step": 6520
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3365442054958186e-05,
      "loss": 3.9659,
      "step": 6530
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.334677419354839e-05,
      "loss": 3.9184,
      "step": 6540
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.332810633213859e-05,
      "loss": 3.9838,
      "step": 6550
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3309438470728796e-05,
      "loss": 3.9563,
      "step": 6560
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3290770609319e-05,
      "loss": 3.9839,
      "step": 6570
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.32721027479092e-05,
      "loss": 4.3452,
      "step": 6580
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3253434886499405e-05,
      "loss": 4.166,
      "step": 6590
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.323476702508961e-05,
      "loss": 3.9137,
      "step": 6600
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.321609916367981e-05,
      "loss": 4.0799,
      "step": 6610
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3197431302270015e-05,
      "loss": 4.1999,
      "step": 6620
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.317876344086022e-05,
      "loss": 4.0544,
      "step": 6630
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.316009557945042e-05,
      "loss": 3.6268,
      "step": 6640
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.3141427718040625e-05,
      "loss": 4.1647,
      "step": 6650
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.312275985663083e-05,
      "loss": 3.9552,
      "step": 6660
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.310409199522103e-05,
      "loss": 4.2799,
      "step": 6670
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.308542413381123e-05,
      "loss": 3.928,
      "step": 6680
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.306675627240144e-05,
      "loss": 4.0199,
      "step": 6690
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.304808841099164e-05,
      "loss": 4.1996,
      "step": 6700
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.302942054958184e-05,
      "loss": 3.9552,
      "step": 6710
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.301075268817205e-05,
      "loss": 4.1223,
      "step": 6720
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2992084826762243e-05,
      "loss": 3.8597,
      "step": 6730
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2973416965352453e-05,
      "loss": 3.7483,
      "step": 6740
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.295474910394266e-05,
      "loss": 3.8141,
      "step": 6750
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.293608124253285e-05,
      "loss": 4.0109,
      "step": 6760
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.291741338112306e-05,
      "loss": 4.3201,
      "step": 6770
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.289874551971326e-05,
      "loss": 4.1932,
      "step": 6780
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.288007765830347e-05,
      "loss": 3.7116,
      "step": 6790
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.286140979689367e-05,
      "loss": 3.733,
      "step": 6800
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.284274193548387e-05,
      "loss": 3.9251,
      "step": 6810
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.282407407407408e-05,
      "loss": 4.0688,
      "step": 6820
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2805406212664276e-05,
      "loss": 3.9903,
      "step": 6830
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2786738351254486e-05,
      "loss": 3.7179,
      "step": 6840
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.276807048984469e-05,
      "loss": 3.9756,
      "step": 6850
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2749402628434885e-05,
      "loss": 4.2888,
      "step": 6860
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2730734767025095e-05,
      "loss": 3.8294,
      "step": 6870
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.271206690561529e-05,
      "loss": 3.9385,
      "step": 6880
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2693399044205495e-05,
      "loss": 4.2638,
      "step": 6890
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2674731182795705e-05,
      "loss": 4.1053,
      "step": 6900
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.26560633213859e-05,
      "loss": 4.2088,
      "step": 6910
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.263739545997611e-05,
      "loss": 4.1388,
      "step": 6920
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.261872759856631e-05,
      "loss": 3.8907,
      "step": 6930
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.260005973715651e-05,
      "loss": 3.7858,
      "step": 6940
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.258139187574672e-05,
      "loss": 3.8084,
      "step": 6950
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.256272401433692e-05,
      "loss": 3.9804,
      "step": 6960
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.254405615292713e-05,
      "loss": 4.1962,
      "step": 6970
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.2525388291517324e-05,
      "loss": 3.9086,
      "step": 6980
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.250672043010753e-05,
      "loss": 4.0937,
      "step": 6990
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.248805256869773e-05,
      "loss": 4.0512,
      "step": 7000
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2469384707287934e-05,
      "loss": 3.7795,
      "step": 7010
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.245071684587814e-05,
      "loss": 3.9607,
      "step": 7020
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.243204898446834e-05,
      "loss": 4.0437,
      "step": 7030
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.241338112305854e-05,
      "loss": 4.0527,
      "step": 7040
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2394713261648746e-05,
      "loss": 4.0317,
      "step": 7050
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.237604540023895e-05,
      "loss": 4.0935,
      "step": 7060
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.235737753882915e-05,
      "loss": 3.8888,
      "step": 7070
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2338709677419356e-05,
      "loss": 3.8422,
      "step": 7080
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.232004181600956e-05,
      "loss": 3.6444,
      "step": 7090
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.230137395459976e-05,
      "loss": 3.9428,
      "step": 7100
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2282706093189966e-05,
      "loss": 3.9388,
      "step": 7110
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.226403823178017e-05,
      "loss": 4.0284,
      "step": 7120
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.224537037037037e-05,
      "loss": 4.1524,
      "step": 7130
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2226702508960575e-05,
      "loss": 3.7295,
      "step": 7140
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.220803464755078e-05,
      "loss": 3.8585,
      "step": 7150
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.218936678614098e-05,
      "loss": 4.1146,
      "step": 7160
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2170698924731185e-05,
      "loss": 3.8945,
      "step": 7170
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.215203106332139e-05,
      "loss": 4.1032,
      "step": 7180
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.213336320191159e-05,
      "loss": 3.7509,
      "step": 7190
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2114695340501795e-05,
      "loss": 4.1754,
      "step": 7200
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2096027479092e-05,
      "loss": 3.8621,
      "step": 7210
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.20773596176822e-05,
      "loss": 3.9213,
      "step": 7220
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2058691756272404e-05,
      "loss": 3.8323,
      "step": 7230
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.204002389486261e-05,
      "loss": 4.1141,
      "step": 7240
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.202135603345281e-05,
      "loss": 4.2356,
      "step": 7250
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.2002688172043014e-05,
      "loss": 4.2664,
      "step": 7260
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.198402031063322e-05,
      "loss": 3.9595,
      "step": 7270
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.1965352449223414e-05,
      "loss": 4.2968,
      "step": 7280
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.1946684587813624e-05,
      "loss": 4.1469,
      "step": 7290
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.192801672640383e-05,
      "loss": 3.9717,
      "step": 7300
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.190934886499403e-05,
      "loss": 3.9437,
      "step": 7310
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1890681003584233e-05,
      "loss": 3.4816,
      "step": 7320
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.187201314217443e-05,
      "loss": 3.983,
      "step": 7330
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.185334528076464e-05,
      "loss": 4.1878,
      "step": 7340
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1834677419354836e-05,
      "loss": 3.945,
      "step": 7350
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1816009557945046e-05,
      "loss": 4.1313,
      "step": 7360
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.179734169653525e-05,
      "loss": 4.0865,
      "step": 7370
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1778673835125446e-05,
      "loss": 3.9683,
      "step": 7380
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1760005973715656e-05,
      "loss": 4.078,
      "step": 7390
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.174133811230585e-05,
      "loss": 4.1362,
      "step": 7400
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.172267025089606e-05,
      "loss": 3.7782,
      "step": 7410
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1704002389486266e-05,
      "loss": 3.964,
      "step": 7420
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.168533452807646e-05,
      "loss": 4.1653,
      "step": 7430
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.166666666666667e-05,
      "loss": 4.1821,
      "step": 7440
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.164799880525687e-05,
      "loss": 3.8559,
      "step": 7450
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.162933094384707e-05,
      "loss": 3.8399,
      "step": 7460
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.161066308243728e-05,
      "loss": 3.9547,
      "step": 7470
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.159199522102748e-05,
      "loss": 3.8021,
      "step": 7480
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.157332735961769e-05,
      "loss": 4.1396,
      "step": 7490
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1554659498207885e-05,
      "loss": 4.1667,
      "step": 7500
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.153599163679809e-05,
      "loss": 4.1119,
      "step": 7510
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.15173237753883e-05,
      "loss": 3.8454,
      "step": 7520
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1498655913978494e-05,
      "loss": 3.9409,
      "step": 7530
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1479988052568704e-05,
      "loss": 3.4948,
      "step": 7540
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.14613201911589e-05,
      "loss": 3.9529,
      "step": 7550
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1442652329749104e-05,
      "loss": 3.7625,
      "step": 7560
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.1423984468339314e-05,
      "loss": 4.1669,
      "step": 7570
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.140531660692951e-05,
      "loss": 4.3768,
      "step": 7580
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1386648745519714e-05,
      "loss": 4.0217,
      "step": 7590
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.136798088410992e-05,
      "loss": 4.0772,
      "step": 7600
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.134931302270012e-05,
      "loss": 4.3164,
      "step": 7610
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.133064516129033e-05,
      "loss": 4.1382,
      "step": 7620
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1311977299880526e-05,
      "loss": 4.1516,
      "step": 7630
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.129330943847073e-05,
      "loss": 3.7171,
      "step": 7640
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.127464157706093e-05,
      "loss": 4.0279,
      "step": 7650
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1255973715651136e-05,
      "loss": 3.9548,
      "step": 7660
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.123730585424134e-05,
      "loss": 3.8773,
      "step": 7670
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.121863799283154e-05,
      "loss": 4.0106,
      "step": 7680
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1199970131421746e-05,
      "loss": 3.7444,
      "step": 7690
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.118130227001195e-05,
      "loss": 3.8094,
      "step": 7700
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.116263440860215e-05,
      "loss": 3.9831,
      "step": 7710
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1143966547192355e-05,
      "loss": 3.8809,
      "step": 7720
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.112529868578256e-05,
      "loss": 3.7986,
      "step": 7730
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.110663082437276e-05,
      "loss": 4.1253,
      "step": 7740
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1087962962962965e-05,
      "loss": 3.7705,
      "step": 7750
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.106929510155317e-05,
      "loss": 4.3931,
      "step": 7760
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.105062724014337e-05,
      "loss": 3.8068,
      "step": 7770
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.1031959378733575e-05,
      "loss": 4.0711,
      "step": 7780
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.101329151732378e-05,
      "loss": 3.9627,
      "step": 7790
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.099462365591398e-05,
      "loss": 3.8308,
      "step": 7800
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.0975955794504184e-05,
      "loss": 3.6867,
      "step": 7810
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.095728793309439e-05,
      "loss": 4.1285,
      "step": 7820
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.093862007168459e-05,
      "loss": 4.0034,
      "step": 7830
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.0919952210274794e-05,
      "loss": 3.8995,
      "step": 7840
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.090128434886499e-05,
      "loss": 3.8953,
      "step": 7850
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.08826164874552e-05,
      "loss": 3.5996,
      "step": 7860
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.0863948626045404e-05,
      "loss": 3.7437,
      "step": 7870
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.084528076463561e-05,
      "loss": 4.1237,
      "step": 7880
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.082661290322581e-05,
      "loss": 3.5073,
      "step": 7890
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0807945041816007e-05,
      "loss": 3.6698,
      "step": 7900
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0789277180406217e-05,
      "loss": 3.707,
      "step": 7910
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.077060931899642e-05,
      "loss": 3.8075,
      "step": 7920
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.075194145758662e-05,
      "loss": 3.8448,
      "step": 7930
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0733273596176826e-05,
      "loss": 3.6987,
      "step": 7940
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.071460573476702e-05,
      "loss": 3.8019,
      "step": 7950
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.069593787335723e-05,
      "loss": 3.8111,
      "step": 7960
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0677270011947436e-05,
      "loss": 4.0772,
      "step": 7970
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.065860215053764e-05,
      "loss": 3.9595,
      "step": 7980
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.063993428912784e-05,
      "loss": 3.9321,
      "step": 7990
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.062126642771804e-05,
      "loss": 4.337,
      "step": 8000
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.060259856630825e-05,
      "loss": 3.8624,
      "step": 8010
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0583930704898445e-05,
      "loss": 3.9942,
      "step": 8020
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.056526284348865e-05,
      "loss": 4.0728,
      "step": 8030
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.054659498207886e-05,
      "loss": 4.2469,
      "step": 8040
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0527927120669055e-05,
      "loss": 3.8322,
      "step": 8050
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0509259259259265e-05,
      "loss": 3.6763,
      "step": 8060
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.049059139784946e-05,
      "loss": 4.0425,
      "step": 8070
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0471923536439665e-05,
      "loss": 3.8562,
      "step": 8080
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0453255675029875e-05,
      "loss": 4.1258,
      "step": 8090
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.043458781362007e-05,
      "loss": 3.6832,
      "step": 8100
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.041591995221028e-05,
      "loss": 3.7075,
      "step": 8110
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.039725209080048e-05,
      "loss": 3.7246,
      "step": 8120
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.037858422939068e-05,
      "loss": 3.923,
      "step": 8130
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.035991636798089e-05,
      "loss": 3.8349,
      "step": 8140
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.034124850657109e-05,
      "loss": 3.7014,
      "step": 8150
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.032258064516129e-05,
      "loss": 3.7002,
      "step": 8160
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.0303912783751494e-05,
      "loss": 4.0162,
      "step": 8170
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.02852449223417e-05,
      "loss": 4.0623,
      "step": 8180
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.026657706093191e-05,
      "loss": 3.8674,
      "step": 8190
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.02479091995221e-05,
      "loss": 4.124,
      "step": 8200
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.0229241338112306e-05,
      "loss": 3.6832,
      "step": 8210
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.021057347670251e-05,
      "loss": 3.5502,
      "step": 8220
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.019190561529271e-05,
      "loss": 4.0664,
      "step": 8230
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.017323775388292e-05,
      "loss": 3.7843,
      "step": 8240
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.015456989247312e-05,
      "loss": 3.5732,
      "step": 8250
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.013590203106332e-05,
      "loss": 4.0164,
      "step": 8260
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.0117234169653526e-05,
      "loss": 4.0487,
      "step": 8270
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.009856630824373e-05,
      "loss": 3.9529,
      "step": 8280
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.007989844683393e-05,
      "loss": 3.7276,
      "step": 8290
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.0061230585424135e-05,
      "loss": 3.8644,
      "step": 8300
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.004256272401434e-05,
      "loss": 3.8298,
      "step": 8310
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.002389486260454e-05,
      "loss": 4.0049,
      "step": 8320
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.0005227001194745e-05,
      "loss": 3.863,
      "step": 8330
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.998655913978495e-05,
      "loss": 3.8132,
      "step": 8340
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.996789127837515e-05,
      "loss": 4.1679,
      "step": 8350
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.9949223416965355e-05,
      "loss": 3.8609,
      "step": 8360
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.993055555555556e-05,
      "loss": 3.8277,
      "step": 8370
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.991188769414576e-05,
      "loss": 3.6151,
      "step": 8380
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.9893219832735964e-05,
      "loss": 3.7163,
      "step": 8390
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.987455197132617e-05,
      "loss": 3.8806,
      "step": 8400
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.985588410991637e-05,
      "loss": 3.8521,
      "step": 8410
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.983721624850657e-05,
      "loss": 3.7674,
      "step": 8420
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.981854838709678e-05,
      "loss": 3.7362,
      "step": 8430
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.979988052568698e-05,
      "loss": 3.9856,
      "step": 8440
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.9781212664277184e-05,
      "loss": 3.7352,
      "step": 8450
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.976254480286739e-05,
      "loss": 3.5098,
      "step": 8460
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.974387694145758e-05,
      "loss": 3.7088,
      "step": 8470
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.972520908004779e-05,
      "loss": 3.9955,
      "step": 8480
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9706541218637997e-05,
      "loss": 4.0419,
      "step": 8490
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.96878733572282e-05,
      "loss": 4.0806,
      "step": 8500
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.96692054958184e-05,
      "loss": 4.0337,
      "step": 8510
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.96505376344086e-05,
      "loss": 3.9145,
      "step": 8520
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.963186977299881e-05,
      "loss": 3.6839,
      "step": 8530
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.961320191158901e-05,
      "loss": 4.1887,
      "step": 8540
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9594534050179216e-05,
      "loss": 3.4189,
      "step": 8550
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.957586618876942e-05,
      "loss": 4.0207,
      "step": 8560
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9557198327359616e-05,
      "loss": 3.7435,
      "step": 8570
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9538530465949825e-05,
      "loss": 3.7173,
      "step": 8580
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.951986260454003e-05,
      "loss": 4.3753,
      "step": 8590
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9501194743130225e-05,
      "loss": 3.632,
      "step": 8600
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9482526881720435e-05,
      "loss": 3.7338,
      "step": 8610
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.946385902031063e-05,
      "loss": 3.9785,
      "step": 8620
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.944519115890084e-05,
      "loss": 3.8019,
      "step": 8630
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9426523297491045e-05,
      "loss": 3.8236,
      "step": 8640
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.940785543608124e-05,
      "loss": 3.5959,
      "step": 8650
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.938918757467145e-05,
      "loss": 3.4537,
      "step": 8660
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.937051971326165e-05,
      "loss": 3.6574,
      "step": 8670
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.935185185185186e-05,
      "loss": 4.2991,
      "step": 8680
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9333183990442054e-05,
      "loss": 3.9068,
      "step": 8690
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.931451612903226e-05,
      "loss": 3.7217,
      "step": 8700
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.929584826762247e-05,
      "loss": 3.984,
      "step": 8710
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9277180406212664e-05,
      "loss": 3.7337,
      "step": 8720
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.925851254480287e-05,
      "loss": 3.794,
      "step": 8730
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.923984468339307e-05,
      "loss": 3.7158,
      "step": 8740
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9221176821983273e-05,
      "loss": 3.8296,
      "step": 8750
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.9202508960573483e-05,
      "loss": 3.8269,
      "step": 8760
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.918384109916368e-05,
      "loss": 3.9781,
      "step": 8770
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.916517323775388e-05,
      "loss": 3.9848,
      "step": 8780
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.9146505376344086e-05,
      "loss": 3.5484,
      "step": 8790
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.912783751493429e-05,
      "loss": 3.6333,
      "step": 8800
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.91091696535245e-05,
      "loss": 3.6082,
      "step": 8810
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.9090501792114696e-05,
      "loss": 3.8744,
      "step": 8820
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.90718339307049e-05,
      "loss": 3.8798,
      "step": 8830
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.90531660692951e-05,
      "loss": 3.6393,
      "step": 8840
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.9034498207885306e-05,
      "loss": 3.577,
      "step": 8850
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.901583034647551e-05,
      "loss": 3.9424,
      "step": 8860
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.899716248506571e-05,
      "loss": 3.9708,
      "step": 8870
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8978494623655915e-05,
      "loss": 3.7679,
      "step": 8880
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.895982676224612e-05,
      "loss": 4.0508,
      "step": 8890
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.894115890083632e-05,
      "loss": 3.6616,
      "step": 8900
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8922491039426525e-05,
      "loss": 4.2915,
      "step": 8910
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.890382317801673e-05,
      "loss": 3.8738,
      "step": 8920
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.888515531660693e-05,
      "loss": 3.9764,
      "step": 8930
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8866487455197135e-05,
      "loss": 3.9438,
      "step": 8940
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.884781959378734e-05,
      "loss": 3.7961,
      "step": 8950
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.882915173237754e-05,
      "loss": 3.8157,
      "step": 8960
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8810483870967744e-05,
      "loss": 3.585,
      "step": 8970
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.879181600955795e-05,
      "loss": 3.668,
      "step": 8980
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.877314814814815e-05,
      "loss": 3.6332,
      "step": 8990
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8754480286738354e-05,
      "loss": 3.594,
      "step": 9000
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.873581242532856e-05,
      "loss": 4.0134,
      "step": 9010
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.871714456391876e-05,
      "loss": 3.6043,
      "step": 9020
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8698476702508964e-05,
      "loss": 3.8317,
      "step": 9030
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.867980884109916e-05,
      "loss": 3.8492,
      "step": 9040
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.866114097968937e-05,
      "loss": 3.6899,
      "step": 9050
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.864247311827957e-05,
      "loss": 3.6604,
      "step": 9060
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.8623805256869776e-05,
      "loss": 3.676,
      "step": 9070
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.860513739545998e-05,
      "loss": 3.8921,
      "step": 9080
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8586469534050176e-05,
      "loss": 3.9841,
      "step": 9090
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8567801672640386e-05,
      "loss": 3.6641,
      "step": 9100
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.854913381123059e-05,
      "loss": 3.6656,
      "step": 9110
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8530465949820786e-05,
      "loss": 3.9319,
      "step": 9120
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8511798088410996e-05,
      "loss": 3.5589,
      "step": 9130
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.849313022700119e-05,
      "loss": 3.6389,
      "step": 9140
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.84744623655914e-05,
      "loss": 3.2645,
      "step": 9150
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8455794504181605e-05,
      "loss": 3.7645,
      "step": 9160
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.84371266427718e-05,
      "loss": 3.9175,
      "step": 9170
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.841845878136201e-05,
      "loss": 3.8478,
      "step": 9180
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.839979091995221e-05,
      "loss": 4.0945,
      "step": 9190
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.838112305854242e-05,
      "loss": 3.9738,
      "step": 9200
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.836245519713262e-05,
      "loss": 3.7934,
      "step": 9210
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.834378733572282e-05,
      "loss": 3.9975,
      "step": 9220
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.832511947431303e-05,
      "loss": 3.9221,
      "step": 9230
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8306451612903224e-05,
      "loss": 3.8509,
      "step": 9240
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8287783751493434e-05,
      "loss": 4.1145,
      "step": 9250
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.826911589008364e-05,
      "loss": 3.6662,
      "step": 9260
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8250448028673834e-05,
      "loss": 3.6177,
      "step": 9270
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8231780167264044e-05,
      "loss": 3.7416,
      "step": 9280
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.821311230585424e-05,
      "loss": 4.2665,
      "step": 9290
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8194444444444444e-05,
      "loss": 3.7638,
      "step": 9300
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.817577658303465e-05,
      "loss": 3.7183,
      "step": 9310
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.815710872162485e-05,
      "loss": 3.8718,
      "step": 9320
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.813844086021506e-05,
      "loss": 3.605,
      "step": 9330
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8119772998805257e-05,
      "loss": 3.9948,
      "step": 9340
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.810110513739546e-05,
      "loss": 3.6158,
      "step": 9350
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.808243727598566e-05,
      "loss": 4.2222,
      "step": 9360
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.8063769414575866e-05,
      "loss": 3.5607,
      "step": 9370
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.8045101553166076e-05,
      "loss": 4.0389,
      "step": 9380
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.802643369175627e-05,
      "loss": 3.5809,
      "step": 9390
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.8007765830346476e-05,
      "loss": 3.6368,
      "step": 9400
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.798909796893668e-05,
      "loss": 3.9134,
      "step": 9410
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.797043010752688e-05,
      "loss": 3.6072,
      "step": 9420
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7951762246117086e-05,
      "loss": 3.6682,
      "step": 9430
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.793309438470729e-05,
      "loss": 3.8541,
      "step": 9440
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.791442652329749e-05,
      "loss": 3.8763,
      "step": 9450
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7895758661887695e-05,
      "loss": 3.9722,
      "step": 9460
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.78770908004779e-05,
      "loss": 4.0616,
      "step": 9470
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.78584229390681e-05,
      "loss": 3.6531,
      "step": 9480
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7839755077658305e-05,
      "loss": 3.7011,
      "step": 9490
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.782108721624851e-05,
      "loss": 3.8015,
      "step": 9500
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.780241935483871e-05,
      "loss": 3.9305,
      "step": 9510
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7783751493428915e-05,
      "loss": 3.718,
      "step": 9520
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.776508363201912e-05,
      "loss": 4.0746,
      "step": 9530
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.774641577060932e-05,
      "loss": 3.9653,
      "step": 9540
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7727747909199524e-05,
      "loss": 3.8337,
      "step": 9550
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.770908004778973e-05,
      "loss": 3.9944,
      "step": 9560
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.769041218637993e-05,
      "loss": 3.8524,
      "step": 9570
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7671744324970134e-05,
      "loss": 3.7066,
      "step": 9580
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.765307646356034e-05,
      "loss": 3.7592,
      "step": 9590
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.763440860215054e-05,
      "loss": 3.8453,
      "step": 9600
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7615740740740744e-05,
      "loss": 4.0889,
      "step": 9610
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.759707287933095e-05,
      "loss": 3.7718,
      "step": 9620
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.757840501792115e-05,
      "loss": 4.0248,
      "step": 9630
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.755973715651135e-05,
      "loss": 3.9773,
      "step": 9640
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.7541069295101556e-05,
      "loss": 3.7009,
      "step": 9650
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.752240143369175e-05,
      "loss": 4.4664,
      "step": 9660
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.750373357228196e-05,
      "loss": 3.5295,
      "step": 9670
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7485065710872166e-05,
      "loss": 3.8009,
      "step": 9680
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.746639784946236e-05,
      "loss": 3.9869,
      "step": 9690
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.744772998805257e-05,
      "loss": 3.5928,
      "step": 9700
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.742906212664277e-05,
      "loss": 3.6676,
      "step": 9710
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.741039426523298e-05,
      "loss": 3.5248,
      "step": 9720
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.739172640382318e-05,
      "loss": 3.4531,
      "step": 9730
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.737305854241338e-05,
      "loss": 3.6246,
      "step": 9740
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.735439068100359e-05,
      "loss": 3.631,
      "step": 9750
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7335722819593785e-05,
      "loss": 3.8826,
      "step": 9760
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7317054958183995e-05,
      "loss": 3.8901,
      "step": 9770
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.72983870967742e-05,
      "loss": 3.2209,
      "step": 9780
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7279719235364395e-05,
      "loss": 3.904,
      "step": 9790
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7261051373954605e-05,
      "loss": 3.7013,
      "step": 9800
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.72423835125448e-05,
      "loss": 3.6786,
      "step": 9810
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.722371565113501e-05,
      "loss": 3.788,
      "step": 9820
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7205047789725214e-05,
      "loss": 3.9666,
      "step": 9830
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.718637992831541e-05,
      "loss": 3.83,
      "step": 9840
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.716771206690562e-05,
      "loss": 3.6834,
      "step": 9850
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.714904420549582e-05,
      "loss": 3.693,
      "step": 9860
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.713037634408602e-05,
      "loss": 3.362,
      "step": 9870
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.711170848267623e-05,
      "loss": 3.5185,
      "step": 9880
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.709304062126643e-05,
      "loss": 3.637,
      "step": 9890
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.707437275985664e-05,
      "loss": 3.7918,
      "step": 9900
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.705570489844683e-05,
      "loss": 3.726,
      "step": 9910
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7037037037037037e-05,
      "loss": 3.6951,
      "step": 9920
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.7018369175627247e-05,
      "loss": 3.8542,
      "step": 9930
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.699970131421744e-05,
      "loss": 3.7629,
      "step": 9940
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.698103345280765e-05,
      "loss": 3.3457,
      "step": 9950
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.696236559139785e-05,
      "loss": 4.022,
      "step": 9960
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.694369772998805e-05,
      "loss": 3.4156,
      "step": 9970
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6925029868578256e-05,
      "loss": 3.6386,
      "step": 9980
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.690636200716846e-05,
      "loss": 3.922,
      "step": 9990
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.688769414575866e-05,
      "loss": 3.8244,
      "step": 10000
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6869026284348866e-05,
      "loss": 3.8524,
      "step": 10010
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.685035842293907e-05,
      "loss": 3.2009,
      "step": 10020
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.683169056152927e-05,
      "loss": 3.9708,
      "step": 10030
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6813022700119475e-05,
      "loss": 3.771,
      "step": 10040
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.679435483870968e-05,
      "loss": 4.0235,
      "step": 10050
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.677568697729988e-05,
      "loss": 3.6662,
      "step": 10060
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6757019115890085e-05,
      "loss": 3.9909,
      "step": 10070
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.673835125448029e-05,
      "loss": 3.9069,
      "step": 10080
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.671968339307049e-05,
      "loss": 3.8864,
      "step": 10090
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6701015531660695e-05,
      "loss": 3.8127,
      "step": 10100
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.66823476702509e-05,
      "loss": 3.9281,
      "step": 10110
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.66636798088411e-05,
      "loss": 3.5872,
      "step": 10120
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6645011947431304e-05,
      "loss": 3.6721,
      "step": 10130
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.662634408602151e-05,
      "loss": 3.6174,
      "step": 10140
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.660767622461171e-05,
      "loss": 3.7824,
      "step": 10150
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6589008363201914e-05,
      "loss": 3.6468,
      "step": 10160
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.657034050179212e-05,
      "loss": 3.1382,
      "step": 10170
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.655167264038232e-05,
      "loss": 3.9309,
      "step": 10180
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6533004778972523e-05,
      "loss": 3.7067,
      "step": 10190
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.651433691756273e-05,
      "loss": 3.6668,
      "step": 10200
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.649566905615293e-05,
      "loss": 3.8249,
      "step": 10210
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.647700119474313e-05,
      "loss": 3.622,
      "step": 10220
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.6458333333333336e-05,
      "loss": 3.7141,
      "step": 10230
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.643966547192354e-05,
      "loss": 3.9518,
      "step": 10240
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.642099761051374e-05,
      "loss": 3.6652,
      "step": 10250
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.640232974910394e-05,
      "loss": 3.4699,
      "step": 10260
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.638366188769415e-05,
      "loss": 3.755,
      "step": 10270
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.636499402628435e-05,
      "loss": 3.5821,
      "step": 10280
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6346326164874556e-05,
      "loss": 3.9781,
      "step": 10290
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.632765830346476e-05,
      "loss": 3.6877,
      "step": 10300
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6308990442054955e-05,
      "loss": 3.6512,
      "step": 10310
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6290322580645165e-05,
      "loss": 3.1832,
      "step": 10320
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.627165471923536e-05,
      "loss": 3.4906,
      "step": 10330
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.625298685782557e-05,
      "loss": 3.525,
      "step": 10340
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6234318996415775e-05,
      "loss": 3.3593,
      "step": 10350
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.621565113500597e-05,
      "loss": 3.4422,
      "step": 10360
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.619698327359618e-05,
      "loss": 3.678,
      "step": 10370
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.617831541218638e-05,
      "loss": 3.757,
      "step": 10380
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.615964755077659e-05,
      "loss": 3.1855,
      "step": 10390
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.614097968936679e-05,
      "loss": 3.8261,
      "step": 10400
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.612231182795699e-05,
      "loss": 3.738,
      "step": 10410
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.61036439665472e-05,
      "loss": 3.8054,
      "step": 10420
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6084976105137394e-05,
      "loss": 3.8857,
      "step": 10430
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.60663082437276e-05,
      "loss": 3.2592,
      "step": 10440
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.604764038231781e-05,
      "loss": 3.7933,
      "step": 10450
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6028972520908004e-05,
      "loss": 3.6397,
      "step": 10460
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.6010304659498214e-05,
      "loss": 3.9235,
      "step": 10470
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.599163679808841e-05,
      "loss": 3.8427,
      "step": 10480
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.597296893667861e-05,
      "loss": 3.7712,
      "step": 10490
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.595430107526882e-05,
      "loss": 3.5808,
      "step": 10500
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.593563321385902e-05,
      "loss": 3.3982,
      "step": 10510
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.591696535244923e-05,
      "loss": 3.7063,
      "step": 10520
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.5898297491039426e-05,
      "loss": 3.6352,
      "step": 10530
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.587962962962963e-05,
      "loss": 3.5817,
      "step": 10540
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.586096176821984e-05,
      "loss": 3.9715,
      "step": 10550
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.5842293906810036e-05,
      "loss": 3.5751,
      "step": 10560
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.582362604540024e-05,
      "loss": 3.3962,
      "step": 10570
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.580495818399044e-05,
      "loss": 3.5391,
      "step": 10580
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5786290322580645e-05,
      "loss": 3.2171,
      "step": 10590
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5767622461170855e-05,
      "loss": 3.9271,
      "step": 10600
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.574895459976105e-05,
      "loss": 3.3747,
      "step": 10610
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5730286738351255e-05,
      "loss": 3.5508,
      "step": 10620
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.571161887694146e-05,
      "loss": 3.603,
      "step": 10630
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.569295101553166e-05,
      "loss": 3.4428,
      "step": 10640
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5674283154121865e-05,
      "loss": 3.3248,
      "step": 10650
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.565561529271207e-05,
      "loss": 3.37,
      "step": 10660
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.563694743130227e-05,
      "loss": 3.7291,
      "step": 10670
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5618279569892474e-05,
      "loss": 3.7927,
      "step": 10680
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.559961170848268e-05,
      "loss": 3.9348,
      "step": 10690
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.558094384707288e-05,
      "loss": 3.844,
      "step": 10700
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5562275985663084e-05,
      "loss": 3.4631,
      "step": 10710
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.554360812425329e-05,
      "loss": 3.6684,
      "step": 10720
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.552494026284349e-05,
      "loss": 3.5042,
      "step": 10730
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5506272401433694e-05,
      "loss": 4.056,
      "step": 10740
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.54876045400239e-05,
      "loss": 3.9195,
      "step": 10750
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.54689366786141e-05,
      "loss": 3.6629,
      "step": 10760
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5450268817204303e-05,
      "loss": 3.7106,
      "step": 10770
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.543160095579451e-05,
      "loss": 3.5783,
      "step": 10780
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.541293309438471e-05,
      "loss": 3.6773,
      "step": 10790
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.539426523297491e-05,
      "loss": 3.6787,
      "step": 10800
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5375597371565116e-05,
      "loss": 3.3561,
      "step": 10810
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.535692951015532e-05,
      "loss": 3.2967,
      "step": 10820
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5338261648745516e-05,
      "loss": 3.3411,
      "step": 10830
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5319593787335726e-05,
      "loss": 3.9638,
      "step": 10840
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.530092592592593e-05,
      "loss": 3.7934,
      "step": 10850
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.528225806451613e-05,
      "loss": 3.5194,
      "step": 10860
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5263590203106336e-05,
      "loss": 3.3668,
      "step": 10870
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.524492234169653e-05,
      "loss": 3.2587,
      "step": 10880
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.522625448028674e-05,
      "loss": 3.454,
      "step": 10890
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5207586618876945e-05,
      "loss": 3.5761,
      "step": 10900
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.518891875746715e-05,
      "loss": 3.3618,
      "step": 10910
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.517025089605735e-05,
      "loss": 3.7184,
      "step": 10920
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.515158303464755e-05,
      "loss": 3.9273,
      "step": 10930
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.513291517323776e-05,
      "loss": 3.539,
      "step": 10940
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.511424731182796e-05,
      "loss": 3.8599,
      "step": 10950
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5095579450418165e-05,
      "loss": 3.5001,
      "step": 10960
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.507691158900837e-05,
      "loss": 3.7675,
      "step": 10970
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5058243727598564e-05,
      "loss": 3.5171,
      "step": 10980
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5039575866188774e-05,
      "loss": 3.3035,
      "step": 10990
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.502090800477897e-05,
      "loss": 3.6594,
      "step": 11000
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5002240143369174e-05,
      "loss": 3.4233,
      "step": 11010
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4983572281959384e-05,
      "loss": 3.8459,
      "step": 11020
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.496490442054958e-05,
      "loss": 3.7127,
      "step": 11030
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.494623655913979e-05,
      "loss": 3.0463,
      "step": 11040
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.492756869772999e-05,
      "loss": 3.4694,
      "step": 11050
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.490890083632019e-05,
      "loss": 3.3425,
      "step": 11060
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.48902329749104e-05,
      "loss": 3.5626,
      "step": 11070
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4871565113500596e-05,
      "loss": 3.329,
      "step": 11080
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4852897252090806e-05,
      "loss": 3.9924,
      "step": 11090
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4834229390681e-05,
      "loss": 3.5672,
      "step": 11100
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4815561529271206e-05,
      "loss": 3.7732,
      "step": 11110
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4796893667861416e-05,
      "loss": 3.2531,
      "step": 11120
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.477822580645161e-05,
      "loss": 3.5789,
      "step": 11130
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.4759557945041816e-05,
      "loss": 3.4121,
      "step": 11140
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.474089008363202e-05,
      "loss": 3.8584,
      "step": 11150
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.472222222222222e-05,
      "loss": 3.8955,
      "step": 11160
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.470355436081243e-05,
      "loss": 3.4261,
      "step": 11170
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.468488649940263e-05,
      "loss": 3.6651,
      "step": 11180
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.466621863799283e-05,
      "loss": 3.6174,
      "step": 11190
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4647550776583035e-05,
      "loss": 3.7792,
      "step": 11200
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.462888291517324e-05,
      "loss": 3.4339,
      "step": 11210
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.461021505376345e-05,
      "loss": 3.677,
      "step": 11220
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4591547192353645e-05,
      "loss": 3.7155,
      "step": 11230
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.457287933094385e-05,
      "loss": 3.8755,
      "step": 11240
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.455421146953405e-05,
      "loss": 3.6512,
      "step": 11250
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4535543608124254e-05,
      "loss": 3.6222,
      "step": 11260
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.451687574671446e-05,
      "loss": 3.8166,
      "step": 11270
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.449820788530466e-05,
      "loss": 3.5914,
      "step": 11280
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4479540023894864e-05,
      "loss": 3.237,
      "step": 11290
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.446087216248507e-05,
      "loss": 3.6738,
      "step": 11300
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.444220430107527e-05,
      "loss": 3.653,
      "step": 11310
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4423536439665474e-05,
      "loss": 3.3309,
      "step": 11320
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.440486857825568e-05,
      "loss": 3.5133,
      "step": 11330
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.438620071684588e-05,
      "loss": 3.3535,
      "step": 11340
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.436753285543608e-05,
      "loss": 3.7432,
      "step": 11350
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4348864994026287e-05,
      "loss": 3.1155,
      "step": 11360
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.433019713261649e-05,
      "loss": 3.2939,
      "step": 11370
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.431152927120669e-05,
      "loss": 3.1033,
      "step": 11380
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4292861409796896e-05,
      "loss": 3.658,
      "step": 11390
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.427419354838709e-05,
      "loss": 3.3786,
      "step": 11400
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.42555256869773e-05,
      "loss": 3.4903,
      "step": 11410
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.4236857825567506e-05,
      "loss": 3.6807,
      "step": 11420
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.421818996415771e-05,
      "loss": 3.6701,
      "step": 11430
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.419952210274791e-05,
      "loss": 3.47,
      "step": 11440
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.418085424133811e-05,
      "loss": 3.647,
      "step": 11450
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.416218637992832e-05,
      "loss": 3.6745,
      "step": 11460
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.414351851851852e-05,
      "loss": 3.5417,
      "step": 11470
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.4124850657108725e-05,
      "loss": 3.331,
      "step": 11480
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.410618279569893e-05,
      "loss": 3.3018,
      "step": 11490
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.4087514934289125e-05,
      "loss": 3.3233,
      "step": 11500
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.4068847072879335e-05,
      "loss": 3.3328,
      "step": 11510
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.405017921146954e-05,
      "loss": 3.2761,
      "step": 11520
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.403151135005974e-05,
      "loss": 3.4464,
      "step": 11530
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.4012843488649945e-05,
      "loss": 3.81,
      "step": 11540
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.399417562724014e-05,
      "loss": 3.5985,
      "step": 11550
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.397550776583035e-05,
      "loss": 3.3498,
      "step": 11560
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.3956839904420554e-05,
      "loss": 3.554,
      "step": 11570
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.393817204301075e-05,
      "loss": 3.1683,
      "step": 11580
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.391950418160096e-05,
      "loss": 3.4795,
      "step": 11590
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.390083632019116e-05,
      "loss": 3.4782,
      "step": 11600
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.388216845878137e-05,
      "loss": 3.5869,
      "step": 11610
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.3863500597371564e-05,
      "loss": 3.3082,
      "step": 11620
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.384483273596177e-05,
      "loss": 3.5159,
      "step": 11630
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.382616487455198e-05,
      "loss": 3.2907,
      "step": 11640
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.380749701314217e-05,
      "loss": 3.8893,
      "step": 11650
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.378882915173238e-05,
      "loss": 3.6042,
      "step": 11660
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.377016129032258e-05,
      "loss": 3.666,
      "step": 11670
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.375149342891278e-05,
      "loss": 3.4671,
      "step": 11680
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.373282556750299e-05,
      "loss": 3.4878,
      "step": 11690
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.371415770609319e-05,
      "loss": 3.6441,
      "step": 11700
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.369548984468339e-05,
      "loss": 3.3459,
      "step": 11710
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.3676821983273596e-05,
      "loss": 2.9724,
      "step": 11720
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.36581541218638e-05,
      "loss": 3.0756,
      "step": 11730
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.363948626045401e-05,
      "loss": 3.7122,
      "step": 11740
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.3620818399044205e-05,
      "loss": 3.5976,
      "step": 11750
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.360215053763441e-05,
      "loss": 3.2686,
      "step": 11760
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.358348267622461e-05,
      "loss": 3.2859,
      "step": 11770
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.3564814814814815e-05,
      "loss": 3.0577,
      "step": 11780
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.3546146953405025e-05,
      "loss": 3.4569,
      "step": 11790
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.352747909199522e-05,
      "loss": 3.8069,
      "step": 11800
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.3508811230585425e-05,
      "loss": 3.8182,
      "step": 11810
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.349014336917563e-05,
      "loss": 3.6163,
      "step": 11820
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.347147550776583e-05,
      "loss": 3.5424,
      "step": 11830
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.3452807646356034e-05,
      "loss": 3.8679,
      "step": 11840
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.343413978494624e-05,
      "loss": 3.353,
      "step": 11850
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.341547192353644e-05,
      "loss": 3.6672,
      "step": 11860
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.3396804062126644e-05,
      "loss": 3.3789,
      "step": 11870
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.337813620071685e-05,
      "loss": 3.7498,
      "step": 11880
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.335946833930705e-05,
      "loss": 3.6776,
      "step": 11890
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.3340800477897254e-05,
      "loss": 4.0729,
      "step": 11900
    },
    {
      "epoch": 1.2,
      "eval_accuracy": 0.1013342949873783,
      "eval_loss": 3.613358736038208,
      "eval_runtime": 776.6446,
      "eval_samples_per_second": 3.57,
      "eval_steps_per_second": 1.786,
      "step": 11904
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.332213261648746e-05,
      "loss": 3.6047,
      "step": 11910
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.330346475507766e-05,
      "loss": 3.775,
      "step": 11920
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.328479689366786e-05,
      "loss": 3.3869,
      "step": 11930
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.3266129032258067e-05,
      "loss": 3.7641,
      "step": 11940
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.324746117084827e-05,
      "loss": 3.0311,
      "step": 11950
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.322879330943847e-05,
      "loss": 3.552,
      "step": 11960
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.321012544802867e-05,
      "loss": 3.6877,
      "step": 11970
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.319145758661888e-05,
      "loss": 3.2318,
      "step": 11980
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.317278972520908e-05,
      "loss": 3.3082,
      "step": 11990
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.3154121863799286e-05,
      "loss": 3.2824,
      "step": 12000
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.313545400238949e-05,
      "loss": 3.6182,
      "step": 12010
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.3116786140979685e-05,
      "loss": 3.609,
      "step": 12020
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.3098118279569895e-05,
      "loss": 3.3955,
      "step": 12030
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.30794504181601e-05,
      "loss": 3.361,
      "step": 12040
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.30607825567503e-05,
      "loss": 3.4298,
      "step": 12050
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.3042114695340505e-05,
      "loss": 3.7064,
      "step": 12060
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.30234468339307e-05,
      "loss": 2.9035,
      "step": 12070
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.300477897252091e-05,
      "loss": 3.2812,
      "step": 12080
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2986111111111115e-05,
      "loss": 3.6022,
      "step": 12090
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.296744324970131e-05,
      "loss": 3.3532,
      "step": 12100
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.294877538829152e-05,
      "loss": 3.8214,
      "step": 12110
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.293010752688172e-05,
      "loss": 3.6251,
      "step": 12120
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.291143966547193e-05,
      "loss": 3.4907,
      "step": 12130
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.289277180406213e-05,
      "loss": 3.0279,
      "step": 12140
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.287410394265233e-05,
      "loss": 3.6172,
      "step": 12150
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.285543608124254e-05,
      "loss": 3.1314,
      "step": 12160
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2836768219832734e-05,
      "loss": 3.4209,
      "step": 12170
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2818100358422944e-05,
      "loss": 3.1885,
      "step": 12180
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.279943249701315e-05,
      "loss": 4.0893,
      "step": 12190
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2780764635603343e-05,
      "loss": 3.1239,
      "step": 12200
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2762096774193553e-05,
      "loss": 3.3743,
      "step": 12210
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.274342891278375e-05,
      "loss": 3.3845,
      "step": 12220
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.272476105137396e-05,
      "loss": 2.9713,
      "step": 12230
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.270609318996416e-05,
      "loss": 3.3186,
      "step": 12240
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.268742532855436e-05,
      "loss": 3.4701,
      "step": 12250
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.266875746714457e-05,
      "loss": 3.2487,
      "step": 12260
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2650089605734766e-05,
      "loss": 3.5087,
      "step": 12270
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.263142174432497e-05,
      "loss": 3.4492,
      "step": 12280
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.261275388291517e-05,
      "loss": 3.1935,
      "step": 12290
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2594086021505376e-05,
      "loss": 3.2086,
      "step": 12300
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2575418160095586e-05,
      "loss": 2.8732,
      "step": 12310
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.255675029868578e-05,
      "loss": 4.0379,
      "step": 12320
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.2538082437275985e-05,
      "loss": 3.3803,
      "step": 12330
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.251941457586619e-05,
      "loss": 3.4455,
      "step": 12340
    },
    {
      "epoch": 2.01,
      "learning_rate": 3.250074671445639e-05,
      "loss": 3.6026,
      "step": 12350
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.24820788530466e-05,
      "loss": 3.6799,
      "step": 12360
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.24634109916368e-05,
      "loss": 3.2898,
      "step": 12370
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2444743130227e-05,
      "loss": 3.6444,
      "step": 12380
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2426075268817205e-05,
      "loss": 3.1618,
      "step": 12390
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.240740740740741e-05,
      "loss": 3.4954,
      "step": 12400
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.238873954599761e-05,
      "loss": 3.4859,
      "step": 12410
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2370071684587814e-05,
      "loss": 3.2952,
      "step": 12420
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.235140382317802e-05,
      "loss": 3.8386,
      "step": 12430
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.233273596176822e-05,
      "loss": 2.5948,
      "step": 12440
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2314068100358424e-05,
      "loss": 3.5557,
      "step": 12450
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.229540023894863e-05,
      "loss": 3.4629,
      "step": 12460
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.227673237753883e-05,
      "loss": 3.1708,
      "step": 12470
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2258064516129034e-05,
      "loss": 3.5862,
      "step": 12480
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.223939665471924e-05,
      "loss": 3.5196,
      "step": 12490
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.222072879330944e-05,
      "loss": 3.5019,
      "step": 12500
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.220206093189964e-05,
      "loss": 3.4852,
      "step": 12510
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2183393070489846e-05,
      "loss": 3.1762,
      "step": 12520
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.216472520908005e-05,
      "loss": 3.278,
      "step": 12530
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.214605734767025e-05,
      "loss": 3.1641,
      "step": 12540
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2127389486260456e-05,
      "loss": 3.6227,
      "step": 12550
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.210872162485066e-05,
      "loss": 3.0787,
      "step": 12560
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.209005376344086e-05,
      "loss": 3.4675,
      "step": 12570
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2071385902031066e-05,
      "loss": 3.2655,
      "step": 12580
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.205271804062127e-05,
      "loss": 2.8608,
      "step": 12590
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.203405017921147e-05,
      "loss": 3.5838,
      "step": 12600
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.2015382317801675e-05,
      "loss": 2.9237,
      "step": 12610
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.199671445639188e-05,
      "loss": 3.1729,
      "step": 12620
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.197804659498208e-05,
      "loss": 3.4304,
      "step": 12630
    },
    {
      "epoch": 2.02,
      "learning_rate": 3.195937873357228e-05,
      "loss": 3.1927,
      "step": 12640
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.194071087216249e-05,
      "loss": 3.5452,
      "step": 12650
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.192204301075269e-05,
      "loss": 3.9063,
      "step": 12660
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.190337514934289e-05,
      "loss": 3.3028,
      "step": 12670
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.18847072879331e-05,
      "loss": 3.4593,
      "step": 12680
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1866039426523294e-05,
      "loss": 3.5563,
      "step": 12690
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1847371565113504e-05,
      "loss": 3.439,
      "step": 12700
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.182870370370371e-05,
      "loss": 3.0588,
      "step": 12710
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1810035842293904e-05,
      "loss": 3.6409,
      "step": 12720
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1791367980884114e-05,
      "loss": 3.8135,
      "step": 12730
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.177270011947431e-05,
      "loss": 3.4293,
      "step": 12740
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.175403225806452e-05,
      "loss": 3.1347,
      "step": 12750
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1735364396654724e-05,
      "loss": 3.4132,
      "step": 12760
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.171669653524492e-05,
      "loss": 2.8791,
      "step": 12770
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.169802867383513e-05,
      "loss": 3.6841,
      "step": 12780
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1679360812425327e-05,
      "loss": 3.2535,
      "step": 12790
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1660692951015537e-05,
      "loss": 3.7857,
      "step": 12800
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.164202508960574e-05,
      "loss": 3.5299,
      "step": 12810
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1623357228195936e-05,
      "loss": 3.1984,
      "step": 12820
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1604689366786146e-05,
      "loss": 3.3357,
      "step": 12830
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.158602150537634e-05,
      "loss": 2.9768,
      "step": 12840
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1567353643966546e-05,
      "loss": 3.16,
      "step": 12850
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1548685782556756e-05,
      "loss": 3.0956,
      "step": 12860
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.153001792114695e-05,
      "loss": 3.3066,
      "step": 12870
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.151135005973716e-05,
      "loss": 3.2043,
      "step": 12880
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.149268219832736e-05,
      "loss": 3.0773,
      "step": 12890
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.147401433691756e-05,
      "loss": 3.7041,
      "step": 12900
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.145534647550777e-05,
      "loss": 3.146,
      "step": 12910
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.143667861409797e-05,
      "loss": 3.7534,
      "step": 12920
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.141801075268818e-05,
      "loss": 3.2105,
      "step": 12930
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.1399342891278375e-05,
      "loss": 3.3437,
      "step": 12940
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.138067502986858e-05,
      "loss": 3.191,
      "step": 12950
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.136200716845878e-05,
      "loss": 3.2255,
      "step": 12960
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.1343339307048985e-05,
      "loss": 3.5001,
      "step": 12970
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.132467144563919e-05,
      "loss": 3.2496,
      "step": 12980
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.130600358422939e-05,
      "loss": 3.3881,
      "step": 12990
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.1287335722819594e-05,
      "loss": 3.5971,
      "step": 13000
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.12686678614098e-05,
      "loss": 3.1389,
      "step": 13010
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.125e-05,
      "loss": 3.7135,
      "step": 13020
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.1231332138590204e-05,
      "loss": 3.5104,
      "step": 13030
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.121266427718041e-05,
      "loss": 3.4578,
      "step": 13040
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.119399641577061e-05,
      "loss": 3.3797,
      "step": 13050
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.1175328554360814e-05,
      "loss": 3.2117,
      "step": 13060
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.115666069295102e-05,
      "loss": 3.5261,
      "step": 13070
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.113799283154122e-05,
      "loss": 3.0737,
      "step": 13080
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.111932497013142e-05,
      "loss": 3.6288,
      "step": 13090
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.1100657108721626e-05,
      "loss": 3.4274,
      "step": 13100
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.108198924731183e-05,
      "loss": 3.2464,
      "step": 13110
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.106332138590203e-05,
      "loss": 3.5461,
      "step": 13120
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.1044653524492236e-05,
      "loss": 3.7019,
      "step": 13130
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.102598566308244e-05,
      "loss": 2.8702,
      "step": 13140
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.100731780167264e-05,
      "loss": 3.6125,
      "step": 13150
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.0988649940262846e-05,
      "loss": 3.2217,
      "step": 13160
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.096998207885305e-05,
      "loss": 3.1566,
      "step": 13170
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.095131421744325e-05,
      "loss": 3.2496,
      "step": 13180
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.0932646356033455e-05,
      "loss": 2.8286,
      "step": 13190
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.091397849462366e-05,
      "loss": 3.3238,
      "step": 13200
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.089531063321386e-05,
      "loss": 3.0222,
      "step": 13210
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.0876642771804065e-05,
      "loss": 3.5437,
      "step": 13220
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.085797491039427e-05,
      "loss": 3.1671,
      "step": 13230
    },
    {
      "epoch": 2.04,
      "learning_rate": 3.0839307048984465e-05,
      "loss": 3.3449,
      "step": 13240
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0820639187574675e-05,
      "loss": 3.3196,
      "step": 13250
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.080197132616488e-05,
      "loss": 3.3471,
      "step": 13260
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.078330346475508e-05,
      "loss": 3.3781,
      "step": 13270
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0764635603345284e-05,
      "loss": 3.6882,
      "step": 13280
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.074596774193548e-05,
      "loss": 3.5759,
      "step": 13290
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.072729988052569e-05,
      "loss": 3.3492,
      "step": 13300
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.070863201911589e-05,
      "loss": 3.6678,
      "step": 13310
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.06899641577061e-05,
      "loss": 3.4061,
      "step": 13320
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.06712962962963e-05,
      "loss": 3.4863,
      "step": 13330
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.06526284348865e-05,
      "loss": 3.4559,
      "step": 13340
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.063396057347671e-05,
      "loss": 3.2953,
      "step": 13350
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.06152927120669e-05,
      "loss": 3.3999,
      "step": 13360
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.059662485065711e-05,
      "loss": 3.2553,
      "step": 13370
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0577956989247317e-05,
      "loss": 3.032,
      "step": 13380
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.055928912783751e-05,
      "loss": 3.3624,
      "step": 13390
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.054062126642772e-05,
      "loss": 2.9003,
      "step": 13400
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.052195340501792e-05,
      "loss": 3.0083,
      "step": 13410
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0503285543608123e-05,
      "loss": 2.6751,
      "step": 13420
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.048461768219833e-05,
      "loss": 2.481,
      "step": 13430
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.046594982078853e-05,
      "loss": 3.3208,
      "step": 13440
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0447281959378736e-05,
      "loss": 3.2543,
      "step": 13450
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.042861409796894e-05,
      "loss": 2.9154,
      "step": 13460
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.040994623655914e-05,
      "loss": 2.8689,
      "step": 13470
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0391278375149345e-05,
      "loss": 3.3488,
      "step": 13480
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0372610513739545e-05,
      "loss": 2.9796,
      "step": 13490
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0353942652329752e-05,
      "loss": 3.3013,
      "step": 13500
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0335274790919955e-05,
      "loss": 3.0864,
      "step": 13510
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.0316606929510155e-05,
      "loss": 3.0855,
      "step": 13520
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.029793906810036e-05,
      "loss": 3.1442,
      "step": 13530
    },
    {
      "epoch": 2.05,
      "learning_rate": 3.027927120669056e-05,
      "loss": 3.1092,
      "step": 13540
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0260603345280764e-05,
      "loss": 3.4295,
      "step": 13550
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.024193548387097e-05,
      "loss": 3.3333,
      "step": 13560
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.022326762246117e-05,
      "loss": 3.3782,
      "step": 13570
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0204599761051378e-05,
      "loss": 2.9689,
      "step": 13580
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0185931899641577e-05,
      "loss": 3.4291,
      "step": 13590
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.016726403823178e-05,
      "loss": 2.9132,
      "step": 13600
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0148596176821987e-05,
      "loss": 3.3764,
      "step": 13610
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0129928315412187e-05,
      "loss": 3.759,
      "step": 13620
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0111260454002394e-05,
      "loss": 3.4126,
      "step": 13630
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0092592592592593e-05,
      "loss": 3.2763,
      "step": 13640
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0073924731182797e-05,
      "loss": 3.4596,
      "step": 13650
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0055256869773003e-05,
      "loss": 3.4691,
      "step": 13660
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0036589008363203e-05,
      "loss": 3.4544,
      "step": 13670
    },
    {
      "epoch": 2.06,
      "learning_rate": 3.0017921146953403e-05,
      "loss": 2.9391,
      "step": 13680
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.999925328554361e-05,
      "loss": 3.4471,
      "step": 13690
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9980585424133813e-05,
      "loss": 3.2507,
      "step": 13700
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.996191756272402e-05,
      "loss": 3.1831,
      "step": 13710
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.994324970131422e-05,
      "loss": 3.1339,
      "step": 13720
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.992458183990442e-05,
      "loss": 3.1629,
      "step": 13730
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9905913978494626e-05,
      "loss": 3.4887,
      "step": 13740
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.988724611708483e-05,
      "loss": 3.5253,
      "step": 13750
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9868578255675032e-05,
      "loss": 3.4942,
      "step": 13760
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9849910394265235e-05,
      "loss": 3.6667,
      "step": 13770
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9831242532855435e-05,
      "loss": 3.4804,
      "step": 13780
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9812574671445642e-05,
      "loss": 3.1605,
      "step": 13790
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.979390681003584e-05,
      "loss": 3.1899,
      "step": 13800
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.9775238948626045e-05,
      "loss": 3.2916,
      "step": 13810
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.975657108721625e-05,
      "loss": 3.6614,
      "step": 13820
    },
    {
      "epoch": 2.06,
      "learning_rate": 2.973790322580645e-05,
      "loss": 3.5104,
      "step": 13830
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9719235364396658e-05,
      "loss": 3.6256,
      "step": 13840
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9700567502986858e-05,
      "loss": 3.3907,
      "step": 13850
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.968189964157706e-05,
      "loss": 3.0289,
      "step": 13860
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9663231780167268e-05,
      "loss": 3.3316,
      "step": 13870
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9644563918757467e-05,
      "loss": 2.9692,
      "step": 13880
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9625896057347674e-05,
      "loss": 3.084,
      "step": 13890
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9607228195937874e-05,
      "loss": 3.2942,
      "step": 13900
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9588560334528077e-05,
      "loss": 3.1822,
      "step": 13910
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9569892473118284e-05,
      "loss": 2.8381,
      "step": 13920
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9551224611708483e-05,
      "loss": 3.6579,
      "step": 13930
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.953255675029869e-05,
      "loss": 2.8121,
      "step": 13940
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.951388888888889e-05,
      "loss": 2.922,
      "step": 13950
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9495221027479093e-05,
      "loss": 3.0235,
      "step": 13960
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.94765531660693e-05,
      "loss": 2.9917,
      "step": 13970
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.94578853046595e-05,
      "loss": 3.4005,
      "step": 13980
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.94392174432497e-05,
      "loss": 3.6017,
      "step": 13990
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9420549581839906e-05,
      "loss": 2.9952,
      "step": 14000
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.940188172043011e-05,
      "loss": 3.4786,
      "step": 14010
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9383213859020316e-05,
      "loss": 3.096,
      "step": 14020
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9364545997610516e-05,
      "loss": 3.2039,
      "step": 14030
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9345878136200715e-05,
      "loss": 3.4426,
      "step": 14040
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9327210274790922e-05,
      "loss": 3.1788,
      "step": 14050
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9308542413381125e-05,
      "loss": 3.2147,
      "step": 14060
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9289874551971332e-05,
      "loss": 2.8619,
      "step": 14070
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9271206690561532e-05,
      "loss": 3.4715,
      "step": 14080
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.925253882915173e-05,
      "loss": 3.5248,
      "step": 14090
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9233870967741938e-05,
      "loss": 3.261,
      "step": 14100
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9215203106332138e-05,
      "loss": 3.0448,
      "step": 14110
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.919653524492234e-05,
      "loss": 3.126,
      "step": 14120
    },
    {
      "epoch": 2.07,
      "learning_rate": 2.9177867383512548e-05,
      "loss": 3.4233,
      "step": 14130
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9159199522102748e-05,
      "loss": 3.2681,
      "step": 14140
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9140531660692954e-05,
      "loss": 3.6574,
      "step": 14150
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9121863799283154e-05,
      "loss": 3.3647,
      "step": 14160
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9103195937873357e-05,
      "loss": 3.2597,
      "step": 14170
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9084528076463564e-05,
      "loss": 3.1914,
      "step": 14180
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9065860215053764e-05,
      "loss": 3.4186,
      "step": 14190
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.904719235364397e-05,
      "loss": 3.0038,
      "step": 14200
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.902852449223417e-05,
      "loss": 3.192,
      "step": 14210
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.9009856630824373e-05,
      "loss": 2.9701,
      "step": 14220
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.899118876941458e-05,
      "loss": 3.3854,
      "step": 14230
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.897252090800478e-05,
      "loss": 2.9638,
      "step": 14240
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.895385304659498e-05,
      "loss": 3.0477,
      "step": 14250
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8935185185185186e-05,
      "loss": 3.0335,
      "step": 14260
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.891651732377539e-05,
      "loss": 3.1699,
      "step": 14270
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8897849462365596e-05,
      "loss": 3.3295,
      "step": 14280
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8879181600955796e-05,
      "loss": 3.3094,
      "step": 14290
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8860513739545996e-05,
      "loss": 3.0889,
      "step": 14300
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8841845878136202e-05,
      "loss": 2.8619,
      "step": 14310
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8823178016726406e-05,
      "loss": 3.1476,
      "step": 14320
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8804510155316612e-05,
      "loss": 2.9974,
      "step": 14330
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8785842293906812e-05,
      "loss": 3.5473,
      "step": 14340
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8767174432497012e-05,
      "loss": 3.3988,
      "step": 14350
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.874850657108722e-05,
      "loss": 3.2914,
      "step": 14360
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8729838709677422e-05,
      "loss": 2.8615,
      "step": 14370
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.871117084826762e-05,
      "loss": 3.0331,
      "step": 14380
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8692502986857828e-05,
      "loss": 3.3181,
      "step": 14390
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8673835125448028e-05,
      "loss": 3.0743,
      "step": 14400
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8655167264038235e-05,
      "loss": 3.303,
      "step": 14410
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8636499402628438e-05,
      "loss": 3.7401,
      "step": 14420
    },
    {
      "epoch": 2.08,
      "learning_rate": 2.8617831541218638e-05,
      "loss": 3.0898,
      "step": 14430
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8599163679808844e-05,
      "loss": 3.1395,
      "step": 14440
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8580495818399044e-05,
      "loss": 2.7102,
      "step": 14450
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.856182795698925e-05,
      "loss": 3.362,
      "step": 14460
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.854316009557945e-05,
      "loss": 3.2069,
      "step": 14470
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8524492234169654e-05,
      "loss": 3.3179,
      "step": 14480
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.850582437275986e-05,
      "loss": 3.1048,
      "step": 14490
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.848715651135006e-05,
      "loss": 3.3805,
      "step": 14500
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.846848864994026e-05,
      "loss": 3.4358,
      "step": 14510
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8449820788530467e-05,
      "loss": 3.1769,
      "step": 14520
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.843115292712067e-05,
      "loss": 2.6418,
      "step": 14530
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8412485065710876e-05,
      "loss": 3.3678,
      "step": 14540
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8393817204301076e-05,
      "loss": 3.0592,
      "step": 14550
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8375149342891276e-05,
      "loss": 3.4816,
      "step": 14560
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8356481481481483e-05,
      "loss": 3.0546,
      "step": 14570
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8337813620071686e-05,
      "loss": 3.0405,
      "step": 14580
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8319145758661893e-05,
      "loss": 2.321,
      "step": 14590
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8300477897252092e-05,
      "loss": 3.3545,
      "step": 14600
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8281810035842292e-05,
      "loss": 3.4742,
      "step": 14610
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.82631421744325e-05,
      "loss": 3.3887,
      "step": 14620
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8244474313022702e-05,
      "loss": 3.3564,
      "step": 14630
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.822580645161291e-05,
      "loss": 3.3283,
      "step": 14640
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.820713859020311e-05,
      "loss": 3.0305,
      "step": 14650
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8188470728793308e-05,
      "loss": 3.0675,
      "step": 14660
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8169802867383515e-05,
      "loss": 2.8462,
      "step": 14670
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8151135005973718e-05,
      "loss": 3.0996,
      "step": 14680
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8132467144563918e-05,
      "loss": 2.7608,
      "step": 14690
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8113799283154125e-05,
      "loss": 2.9715,
      "step": 14700
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8095131421744324e-05,
      "loss": 3.2812,
      "step": 14710
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.807646356033453e-05,
      "loss": 3.4222,
      "step": 14720
    },
    {
      "epoch": 2.09,
      "learning_rate": 2.8057795698924734e-05,
      "loss": 2.9256,
      "step": 14730
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.8039127837514934e-05,
      "loss": 3.3173,
      "step": 14740
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.802045997610514e-05,
      "loss": 3.1419,
      "step": 14750
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.800179211469534e-05,
      "loss": 3.4477,
      "step": 14760
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7983124253285547e-05,
      "loss": 3.3932,
      "step": 14770
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7964456391875747e-05,
      "loss": 2.7912,
      "step": 14780
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.794578853046595e-05,
      "loss": 3.2495,
      "step": 14790
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7927120669056157e-05,
      "loss": 2.8159,
      "step": 14800
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7908452807646357e-05,
      "loss": 3.0967,
      "step": 14810
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7889784946236556e-05,
      "loss": 2.7874,
      "step": 14820
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7871117084826763e-05,
      "loss": 2.918,
      "step": 14830
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7852449223416966e-05,
      "loss": 2.785,
      "step": 14840
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7833781362007173e-05,
      "loss": 3.0679,
      "step": 14850
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7815113500597373e-05,
      "loss": 3.1987,
      "step": 14860
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7796445639187573e-05,
      "loss": 2.7207,
      "step": 14870
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.777777777777778e-05,
      "loss": 3.1899,
      "step": 14880
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7759109916367982e-05,
      "loss": 2.7725,
      "step": 14890
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.774044205495819e-05,
      "loss": 2.8578,
      "step": 14900
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.772177419354839e-05,
      "loss": 2.8793,
      "step": 14910
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.770310633213859e-05,
      "loss": 3.122,
      "step": 14920
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7684438470728795e-05,
      "loss": 3.1156,
      "step": 14930
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7665770609319e-05,
      "loss": 3.4035,
      "step": 14940
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7647102747909198e-05,
      "loss": 3.0444,
      "step": 14950
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7628434886499405e-05,
      "loss": 3.4004,
      "step": 14960
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7609767025089605e-05,
      "loss": 3.0023,
      "step": 14970
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.759109916367981e-05,
      "loss": 3.1442,
      "step": 14980
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7572431302270015e-05,
      "loss": 2.9599,
      "step": 14990
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.7553763440860214e-05,
      "loss": 2.9346,
      "step": 15000
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.753509557945042e-05,
      "loss": 3.3563,
      "step": 15010
    },
    {
      "epoch": 2.1,
      "learning_rate": 2.751642771804062e-05,
      "loss": 3.1207,
      "step": 15020
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7497759856630827e-05,
      "loss": 2.9311,
      "step": 15030
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.747909199522103e-05,
      "loss": 2.9882,
      "step": 15040
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.746042413381123e-05,
      "loss": 3.3397,
      "step": 15050
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7441756272401437e-05,
      "loss": 2.9564,
      "step": 15060
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7423088410991637e-05,
      "loss": 2.941,
      "step": 15070
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.740442054958184e-05,
      "loss": 3.4236,
      "step": 15080
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7385752688172043e-05,
      "loss": 3.3045,
      "step": 15090
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7367084826762247e-05,
      "loss": 2.8175,
      "step": 15100
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7348416965352453e-05,
      "loss": 2.4572,
      "step": 15110
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7329749103942653e-05,
      "loss": 3.1204,
      "step": 15120
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7311081242532853e-05,
      "loss": 3.4336,
      "step": 15130
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.729241338112306e-05,
      "loss": 3.1992,
      "step": 15140
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7273745519713263e-05,
      "loss": 3.6503,
      "step": 15150
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.725507765830347e-05,
      "loss": 3.1687,
      "step": 15160
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.723640979689367e-05,
      "loss": 3.0529,
      "step": 15170
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.721774193548387e-05,
      "loss": 3.186,
      "step": 15180
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7199074074074076e-05,
      "loss": 2.8333,
      "step": 15190
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.718040621266428e-05,
      "loss": 2.8004,
      "step": 15200
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7161738351254485e-05,
      "loss": 3.7368,
      "step": 15210
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7143070489844685e-05,
      "loss": 2.6955,
      "step": 15220
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7124402628434885e-05,
      "loss": 2.8241,
      "step": 15230
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.710573476702509e-05,
      "loss": 2.9161,
      "step": 15240
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7087066905615295e-05,
      "loss": 2.6601,
      "step": 15250
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7068399044205495e-05,
      "loss": 3.1068,
      "step": 15260
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.70497311827957e-05,
      "loss": 2.6962,
      "step": 15270
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.70310633213859e-05,
      "loss": 2.8732,
      "step": 15280
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.7012395459976108e-05,
      "loss": 2.9584,
      "step": 15290
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.699372759856631e-05,
      "loss": 3.3231,
      "step": 15300
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.697505973715651e-05,
      "loss": 3.2559,
      "step": 15310
    },
    {
      "epoch": 2.11,
      "learning_rate": 2.6956391875746717e-05,
      "loss": 3.1895,
      "step": 15320
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6937724014336917e-05,
      "loss": 3.3419,
      "step": 15330
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6919056152927124e-05,
      "loss": 3.279,
      "step": 15340
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6900388291517327e-05,
      "loss": 3.3604,
      "step": 15350
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6881720430107527e-05,
      "loss": 3.1441,
      "step": 15360
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6863052568697733e-05,
      "loss": 2.7756,
      "step": 15370
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6844384707287933e-05,
      "loss": 3.0425,
      "step": 15380
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6825716845878137e-05,
      "loss": 3.2714,
      "step": 15390
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6807048984468343e-05,
      "loss": 2.9571,
      "step": 15400
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6788381123058543e-05,
      "loss": 3.1644,
      "step": 15410
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.676971326164875e-05,
      "loss": 3.0494,
      "step": 15420
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.675104540023895e-05,
      "loss": 3.3804,
      "step": 15430
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.673237753882915e-05,
      "loss": 3.0539,
      "step": 15440
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6713709677419356e-05,
      "loss": 2.8719,
      "step": 15450
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.669504181600956e-05,
      "loss": 2.2724,
      "step": 15460
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6676373954599766e-05,
      "loss": 2.6922,
      "step": 15470
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6657706093189965e-05,
      "loss": 2.657,
      "step": 15480
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6639038231780165e-05,
      "loss": 3.1571,
      "step": 15490
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6620370370370372e-05,
      "loss": 3.2733,
      "step": 15500
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6601702508960575e-05,
      "loss": 2.7808,
      "step": 15510
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6583034647550775e-05,
      "loss": 2.7793,
      "step": 15520
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.656436678614098e-05,
      "loss": 3.3887,
      "step": 15530
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.654569892473118e-05,
      "loss": 2.9146,
      "step": 15540
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6527031063321388e-05,
      "loss": 3.5791,
      "step": 15550
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.650836320191159e-05,
      "loss": 2.8775,
      "step": 15560
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.648969534050179e-05,
      "loss": 3.1055,
      "step": 15570
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6471027479091998e-05,
      "loss": 3.2177,
      "step": 15580
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6452359617682198e-05,
      "loss": 3.2396,
      "step": 15590
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6433691756272404e-05,
      "loss": 3.3748,
      "step": 15600
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6415023894862607e-05,
      "loss": 3.3006,
      "step": 15610
    },
    {
      "epoch": 2.12,
      "learning_rate": 2.6396356033452807e-05,
      "loss": 3.0114,
      "step": 15620
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6377688172043014e-05,
      "loss": 3.006,
      "step": 15630
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6359020310633214e-05,
      "loss": 3.1925,
      "step": 15640
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6340352449223417e-05,
      "loss": 3.2708,
      "step": 15650
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6321684587813623e-05,
      "loss": 3.1134,
      "step": 15660
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6303016726403823e-05,
      "loss": 2.6587,
      "step": 15670
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.628434886499403e-05,
      "loss": 2.9197,
      "step": 15680
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.626568100358423e-05,
      "loss": 3.1758,
      "step": 15690
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6247013142174433e-05,
      "loss": 3.1256,
      "step": 15700
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.622834528076464e-05,
      "loss": 3.1082,
      "step": 15710
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.620967741935484e-05,
      "loss": 3.2053,
      "step": 15720
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6191009557945046e-05,
      "loss": 3.3624,
      "step": 15730
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6172341696535246e-05,
      "loss": 2.9262,
      "step": 15740
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.615367383512545e-05,
      "loss": 3.1878,
      "step": 15750
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6135005973715652e-05,
      "loss": 2.8616,
      "step": 15760
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6116338112305855e-05,
      "loss": 3.8151,
      "step": 15770
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6097670250896062e-05,
      "loss": 2.8967,
      "step": 15780
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6079002389486262e-05,
      "loss": 2.987,
      "step": 15790
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.6060334528076462e-05,
      "loss": 3.0226,
      "step": 15800
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.604166666666667e-05,
      "loss": 2.5813,
      "step": 15810
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.602299880525687e-05,
      "loss": 2.9065,
      "step": 15820
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.600433094384707e-05,
      "loss": 2.8649,
      "step": 15830
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5985663082437278e-05,
      "loss": 3.0972,
      "step": 15840
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5966995221027478e-05,
      "loss": 2.5581,
      "step": 15850
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5948327359617684e-05,
      "loss": 3.4753,
      "step": 15860
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5929659498207888e-05,
      "loss": 2.5078,
      "step": 15870
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5910991636798087e-05,
      "loss": 3.1008,
      "step": 15880
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5892323775388294e-05,
      "loss": 2.9719,
      "step": 15890
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5873655913978494e-05,
      "loss": 2.7048,
      "step": 15900
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.58549880525687e-05,
      "loss": 2.9988,
      "step": 15910
    },
    {
      "epoch": 2.13,
      "learning_rate": 2.5836320191158904e-05,
      "loss": 2.8525,
      "step": 15920
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5817652329749104e-05,
      "loss": 2.7417,
      "step": 15930
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.579898446833931e-05,
      "loss": 2.8101,
      "step": 15940
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.578031660692951e-05,
      "loss": 2.4293,
      "step": 15950
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5761648745519713e-05,
      "loss": 2.7513,
      "step": 15960
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.574298088410992e-05,
      "loss": 2.696,
      "step": 15970
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.572431302270012e-05,
      "loss": 2.2097,
      "step": 15980
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5705645161290326e-05,
      "loss": 2.1147,
      "step": 15990
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5686977299880526e-05,
      "loss": 3.4384,
      "step": 16000
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.566830943847073e-05,
      "loss": 2.9081,
      "step": 16010
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5649641577060936e-05,
      "loss": 2.7966,
      "step": 16020
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5630973715651136e-05,
      "loss": 2.865,
      "step": 16030
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5612305854241342e-05,
      "loss": 2.9875,
      "step": 16040
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5593637992831542e-05,
      "loss": 3.0181,
      "step": 16050
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5574970131421745e-05,
      "loss": 2.6113,
      "step": 16060
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.555630227001195e-05,
      "loss": 2.5983,
      "step": 16070
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5537634408602152e-05,
      "loss": 2.6221,
      "step": 16080
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5518966547192352e-05,
      "loss": 3.1799,
      "step": 16090
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.550029868578256e-05,
      "loss": 2.8745,
      "step": 16100
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5481630824372758e-05,
      "loss": 2.9698,
      "step": 16110
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5462962962962965e-05,
      "loss": 2.7704,
      "step": 16120
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5444295101553168e-05,
      "loss": 3.0358,
      "step": 16130
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5425627240143368e-05,
      "loss": 2.7857,
      "step": 16140
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5406959378733574e-05,
      "loss": 3.2799,
      "step": 16150
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5388291517323774e-05,
      "loss": 2.6568,
      "step": 16160
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.536962365591398e-05,
      "loss": 3.0519,
      "step": 16170
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5350955794504184e-05,
      "loss": 2.9738,
      "step": 16180
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.5332287933094384e-05,
      "loss": 3.0673,
      "step": 16190
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.531362007168459e-05,
      "loss": 2.9453,
      "step": 16200
    },
    {
      "epoch": 2.14,
      "learning_rate": 2.529495221027479e-05,
      "loss": 2.6896,
      "step": 16210
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5276284348864994e-05,
      "loss": 2.7328,
      "step": 16220
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.52576164874552e-05,
      "loss": 2.6248,
      "step": 16230
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.52389486260454e-05,
      "loss": 2.7152,
      "step": 16240
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5220280764635607e-05,
      "loss": 2.9469,
      "step": 16250
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5201612903225806e-05,
      "loss": 2.6343,
      "step": 16260
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.518294504181601e-05,
      "loss": 3.0206,
      "step": 16270
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5164277180406216e-05,
      "loss": 3.0976,
      "step": 16280
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5145609318996416e-05,
      "loss": 2.8687,
      "step": 16290
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5126941457586623e-05,
      "loss": 3.2842,
      "step": 16300
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5108273596176823e-05,
      "loss": 2.8738,
      "step": 16310
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5089605734767026e-05,
      "loss": 2.3145,
      "step": 16320
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5070937873357232e-05,
      "loss": 2.9295,
      "step": 16330
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.5052270011947432e-05,
      "loss": 2.6285,
      "step": 16340
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.503360215053764e-05,
      "loss": 2.7225,
      "step": 16350
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.501493428912784e-05,
      "loss": 3.0596,
      "step": 16360
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4996266427718042e-05,
      "loss": 2.5452,
      "step": 16370
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4977598566308245e-05,
      "loss": 2.829,
      "step": 16380
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4958930704898448e-05,
      "loss": 3.0554,
      "step": 16390
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.494026284348865e-05,
      "loss": 2.8526,
      "step": 16400
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4921594982078855e-05,
      "loss": 3.194,
      "step": 16410
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4902927120669055e-05,
      "loss": 2.9661,
      "step": 16420
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.488425925925926e-05,
      "loss": 3.0343,
      "step": 16430
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4865591397849464e-05,
      "loss": 2.7143,
      "step": 16440
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4846923536439668e-05,
      "loss": 3.1782,
      "step": 16450
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.482825567502987e-05,
      "loss": 2.697,
      "step": 16460
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.480958781362007e-05,
      "loss": 2.4925,
      "step": 16470
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4790919952210277e-05,
      "loss": 2.6889,
      "step": 16480
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.477225209080048e-05,
      "loss": 3.1437,
      "step": 16490
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4753584229390684e-05,
      "loss": 2.9968,
      "step": 16500
    },
    {
      "epoch": 2.15,
      "learning_rate": 2.4734916367980884e-05,
      "loss": 3.1598,
      "step": 16510
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4716248506571087e-05,
      "loss": 2.6676,
      "step": 16520
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4697580645161293e-05,
      "loss": 2.7759,
      "step": 16530
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4678912783751497e-05,
      "loss": 2.6843,
      "step": 16540
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4660244922341696e-05,
      "loss": 2.9387,
      "step": 16550
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.46415770609319e-05,
      "loss": 3.394,
      "step": 16560
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4622909199522103e-05,
      "loss": 3.0735,
      "step": 16570
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4604241338112306e-05,
      "loss": 2.7541,
      "step": 16580
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4585573476702513e-05,
      "loss": 3.5003,
      "step": 16590
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4566905615292712e-05,
      "loss": 2.4582,
      "step": 16600
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4548237753882916e-05,
      "loss": 3.588,
      "step": 16610
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.452956989247312e-05,
      "loss": 2.647,
      "step": 16620
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4510902031063322e-05,
      "loss": 3.2833,
      "step": 16630
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4492234169653525e-05,
      "loss": 2.7989,
      "step": 16640
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.447356630824373e-05,
      "loss": 3.5054,
      "step": 16650
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4454898446833932e-05,
      "loss": 3.2047,
      "step": 16660
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4436230585424135e-05,
      "loss": 3.388,
      "step": 16670
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4417562724014338e-05,
      "loss": 2.9514,
      "step": 16680
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.439889486260454e-05,
      "loss": 2.7971,
      "step": 16690
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4380227001194745e-05,
      "loss": 3.0533,
      "step": 16700
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4361559139784948e-05,
      "loss": 2.4706,
      "step": 16710
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.434289127837515e-05,
      "loss": 3.0765,
      "step": 16720
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4324223416965354e-05,
      "loss": 2.7394,
      "step": 16730
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4305555555555558e-05,
      "loss": 2.3243,
      "step": 16740
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.428688769414576e-05,
      "loss": 2.8647,
      "step": 16750
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4268219832735964e-05,
      "loss": 2.5507,
      "step": 16760
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4249551971326164e-05,
      "loss": 3.0999,
      "step": 16770
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4230884109916367e-05,
      "loss": 2.3194,
      "step": 16780
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4212216248506574e-05,
      "loss": 2.9616,
      "step": 16790
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.4193548387096777e-05,
      "loss": 3.2428,
      "step": 16800
    },
    {
      "epoch": 2.16,
      "learning_rate": 2.417488052568698e-05,
      "loss": 3.1768,
      "step": 16810
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.415621266427718e-05,
      "loss": 3.2242,
      "step": 16820
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.4137544802867383e-05,
      "loss": 3.22,
      "step": 16830
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.411887694145759e-05,
      "loss": 3.1305,
      "step": 16840
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.4100209080047793e-05,
      "loss": 3.267,
      "step": 16850
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.4081541218637993e-05,
      "loss": 3.2045,
      "step": 16860
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.4062873357228196e-05,
      "loss": 3.5537,
      "step": 16870
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.40442054958184e-05,
      "loss": 3.4825,
      "step": 16880
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.4025537634408606e-05,
      "loss": 2.9965,
      "step": 16890
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.4006869772998806e-05,
      "loss": 2.7778,
      "step": 16900
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.398820191158901e-05,
      "loss": 2.718,
      "step": 16910
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3969534050179212e-05,
      "loss": 2.5592,
      "step": 16920
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3950866188769415e-05,
      "loss": 3.3026,
      "step": 16930
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.393219832735962e-05,
      "loss": 2.8021,
      "step": 16940
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3913530465949822e-05,
      "loss": 2.7009,
      "step": 16950
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3894862604540025e-05,
      "loss": 2.8582,
      "step": 16960
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3876194743130228e-05,
      "loss": 2.7851,
      "step": 16970
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.385752688172043e-05,
      "loss": 2.963,
      "step": 16980
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3838859020310635e-05,
      "loss": 2.8885,
      "step": 16990
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3820191158900838e-05,
      "loss": 3.014,
      "step": 17000
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.380152329749104e-05,
      "loss": 2.5369,
      "step": 17010
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3782855436081244e-05,
      "loss": 3.0588,
      "step": 17020
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3764187574671444e-05,
      "loss": 2.8192,
      "step": 17030
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.374551971326165e-05,
      "loss": 3.0529,
      "step": 17040
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3726851851851854e-05,
      "loss": 2.509,
      "step": 17050
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3708183990442057e-05,
      "loss": 3.4218,
      "step": 17060
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.368951612903226e-05,
      "loss": 2.885,
      "step": 17070
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.367084826762246e-05,
      "loss": 2.9661,
      "step": 17080
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3652180406212663e-05,
      "loss": 3.1489,
      "step": 17090
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.363351254480287e-05,
      "loss": 2.8483,
      "step": 17100
    },
    {
      "epoch": 2.17,
      "learning_rate": 2.3614844683393073e-05,
      "loss": 2.6805,
      "step": 17110
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3596176821983273e-05,
      "loss": 2.9985,
      "step": 17120
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3577508960573476e-05,
      "loss": 2.93,
      "step": 17130
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.355884109916368e-05,
      "loss": 2.9248,
      "step": 17140
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3540173237753886e-05,
      "loss": 2.4329,
      "step": 17150
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.352150537634409e-05,
      "loss": 3.1536,
      "step": 17160
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.350283751493429e-05,
      "loss": 2.9112,
      "step": 17170
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3484169653524492e-05,
      "loss": 3.0118,
      "step": 17180
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3465501792114696e-05,
      "loss": 3.2796,
      "step": 17190
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3446833930704902e-05,
      "loss": 2.8319,
      "step": 17200
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3428166069295102e-05,
      "loss": 2.808,
      "step": 17210
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3409498207885305e-05,
      "loss": 2.3681,
      "step": 17220
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.339083034647551e-05,
      "loss": 3.2278,
      "step": 17230
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3372162485065712e-05,
      "loss": 2.7767,
      "step": 17240
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3353494623655915e-05,
      "loss": 2.6173,
      "step": 17250
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3334826762246118e-05,
      "loss": 2.9629,
      "step": 17260
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.331615890083632e-05,
      "loss": 2.7156,
      "step": 17270
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3297491039426525e-05,
      "loss": 3.1659,
      "step": 17280
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3278823178016728e-05,
      "loss": 2.8566,
      "step": 17290
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.326015531660693e-05,
      "loss": 2.57,
      "step": 17300
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3241487455197134e-05,
      "loss": 2.7788,
      "step": 17310
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3222819593787338e-05,
      "loss": 2.8175,
      "step": 17320
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.320415173237754e-05,
      "loss": 2.9958,
      "step": 17330
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.318548387096774e-05,
      "loss": 2.407,
      "step": 17340
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3166816009557947e-05,
      "loss": 2.981,
      "step": 17350
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.314814814814815e-05,
      "loss": 2.9412,
      "step": 17360
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3129480286738354e-05,
      "loss": 3.145,
      "step": 17370
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3110812425328557e-05,
      "loss": 2.6311,
      "step": 17380
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.3092144563918757e-05,
      "loss": 2.7061,
      "step": 17390
    },
    {
      "epoch": 2.18,
      "learning_rate": 2.307347670250896e-05,
      "loss": 2.4077,
      "step": 17400
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.3054808841099166e-05,
      "loss": 2.5605,
      "step": 17410
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.303614097968937e-05,
      "loss": 2.5439,
      "step": 17420
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.301747311827957e-05,
      "loss": 3.0293,
      "step": 17430
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2998805256869773e-05,
      "loss": 2.3587,
      "step": 17440
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2980137395459976e-05,
      "loss": 2.7149,
      "step": 17450
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2961469534050183e-05,
      "loss": 2.7515,
      "step": 17460
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2942801672640382e-05,
      "loss": 2.4752,
      "step": 17470
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2924133811230586e-05,
      "loss": 2.8067,
      "step": 17480
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.290546594982079e-05,
      "loss": 3.2142,
      "step": 17490
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2886798088410992e-05,
      "loss": 2.6557,
      "step": 17500
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.28681302270012e-05,
      "loss": 2.9121,
      "step": 17510
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.28494623655914e-05,
      "loss": 2.6181,
      "step": 17520
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2830794504181602e-05,
      "loss": 2.6288,
      "step": 17530
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2812126642771805e-05,
      "loss": 2.6474,
      "step": 17540
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2793458781362008e-05,
      "loss": 3.1204,
      "step": 17550
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.277479091995221e-05,
      "loss": 2.8625,
      "step": 17560
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2756123058542415e-05,
      "loss": 2.8513,
      "step": 17570
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2737455197132618e-05,
      "loss": 2.9578,
      "step": 17580
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.271878733572282e-05,
      "loss": 2.5751,
      "step": 17590
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.270011947431302e-05,
      "loss": 2.5051,
      "step": 17600
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2681451612903227e-05,
      "loss": 3.1305,
      "step": 17610
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.266278375149343e-05,
      "loss": 2.388,
      "step": 17620
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2644115890083634e-05,
      "loss": 2.592,
      "step": 17630
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2625448028673837e-05,
      "loss": 2.9774,
      "step": 17640
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2606780167264037e-05,
      "loss": 3.178,
      "step": 17650
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2588112305854244e-05,
      "loss": 2.5683,
      "step": 17660
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2569444444444447e-05,
      "loss": 2.7476,
      "step": 17670
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.255077658303465e-05,
      "loss": 2.6905,
      "step": 17680
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.253210872162485e-05,
      "loss": 2.5406,
      "step": 17690
    },
    {
      "epoch": 2.19,
      "learning_rate": 2.2513440860215053e-05,
      "loss": 2.5228,
      "step": 17700
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.249477299880526e-05,
      "loss": 2.3931,
      "step": 17710
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2476105137395463e-05,
      "loss": 2.7231,
      "step": 17720
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2457437275985666e-05,
      "loss": 2.7678,
      "step": 17730
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2438769414575866e-05,
      "loss": 2.3632,
      "step": 17740
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.242010155316607e-05,
      "loss": 2.8103,
      "step": 17750
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2401433691756272e-05,
      "loss": 2.5602,
      "step": 17760
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.238276583034648e-05,
      "loss": 2.9518,
      "step": 17770
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.236409796893668e-05,
      "loss": 2.6576,
      "step": 17780
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2345430107526882e-05,
      "loss": 2.5758,
      "step": 17790
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2326762246117085e-05,
      "loss": 2.5331,
      "step": 17800
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.230809438470729e-05,
      "loss": 3.0794,
      "step": 17810
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2289426523297492e-05,
      "loss": 2.5879,
      "step": 17820
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2270758661887695e-05,
      "loss": 3.0335,
      "step": 17830
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.2252090800477898e-05,
      "loss": 2.8501,
      "step": 17840
    },
    {
      "epoch": 2.2,
      "learning_rate": 2.22334229390681e-05,
      "loss": 2.6787,
      "step": 17850
    },
    {
      "epoch": 2.2,
      "eval_accuracy": 0.21925712225027047,
      "eval_loss": 2.939669609069824,
      "eval_runtime": 784.0992,
      "eval_samples_per_second": 3.537,
      "eval_steps_per_second": 1.769,
      "step": 17856
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2214755077658305e-05,
      "loss": 2.9471,
      "step": 17860
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2196087216248508e-05,
      "loss": 2.2457,
      "step": 17870
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.217741935483871e-05,
      "loss": 2.5336,
      "step": 17880
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2158751493428914e-05,
      "loss": 2.8878,
      "step": 17890
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2140083632019117e-05,
      "loss": 2.707,
      "step": 17900
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2121415770609317e-05,
      "loss": 3.1906,
      "step": 17910
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2102747909199524e-05,
      "loss": 3.2297,
      "step": 17920
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2084080047789727e-05,
      "loss": 2.8249,
      "step": 17930
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.206541218637993e-05,
      "loss": 2.3213,
      "step": 17940
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.204674432497013e-05,
      "loss": 2.5766,
      "step": 17950
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.2028076463560333e-05,
      "loss": 2.5443,
      "step": 17960
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.200940860215054e-05,
      "loss": 2.7451,
      "step": 17970
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.1990740740740743e-05,
      "loss": 2.1575,
      "step": 17980
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.1972072879330946e-05,
      "loss": 2.3331,
      "step": 17990
    },
    {
      "epoch": 3.0,
      "learning_rate": 2.1953405017921146e-05,
      "loss": 2.7351,
      "step": 18000
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.193473715651135e-05,
      "loss": 2.864,
      "step": 18010
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1916069295101556e-05,
      "loss": 2.7778,
      "step": 18020
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.189740143369176e-05,
      "loss": 2.7513,
      "step": 18030
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.187873357228196e-05,
      "loss": 2.3821,
      "step": 18040
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1860065710872162e-05,
      "loss": 3.2049,
      "step": 18050
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1841397849462366e-05,
      "loss": 2.8676,
      "step": 18060
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.182272998805257e-05,
      "loss": 2.7496,
      "step": 18070
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1804062126642775e-05,
      "loss": 2.8662,
      "step": 18080
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1785394265232975e-05,
      "loss": 2.446,
      "step": 18090
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.176672640382318e-05,
      "loss": 2.6361,
      "step": 18100
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.174805854241338e-05,
      "loss": 2.6153,
      "step": 18110
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1729390681003585e-05,
      "loss": 2.5376,
      "step": 18120
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1710722819593788e-05,
      "loss": 2.4613,
      "step": 18130
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.169205495818399e-05,
      "loss": 2.7581,
      "step": 18140
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1673387096774195e-05,
      "loss": 2.3273,
      "step": 18150
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1654719235364398e-05,
      "loss": 2.5223,
      "step": 18160
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.16360513739546e-05,
      "loss": 2.7759,
      "step": 18170
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1617383512544804e-05,
      "loss": 2.0227,
      "step": 18180
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1598715651135007e-05,
      "loss": 2.3961,
      "step": 18190
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.158004778972521e-05,
      "loss": 2.6126,
      "step": 18200
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1561379928315414e-05,
      "loss": 2.8436,
      "step": 18210
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1542712066905614e-05,
      "loss": 2.7485,
      "step": 18220
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.152404420549582e-05,
      "loss": 2.6791,
      "step": 18230
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1505376344086024e-05,
      "loss": 2.3018,
      "step": 18240
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1486708482676227e-05,
      "loss": 2.5784,
      "step": 18250
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1468040621266427e-05,
      "loss": 2.6444,
      "step": 18260
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.144937275985663e-05,
      "loss": 2.786,
      "step": 18270
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1430704898446836e-05,
      "loss": 2.0548,
      "step": 18280
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.141203703703704e-05,
      "loss": 2.1212,
      "step": 18290
    },
    {
      "epoch": 3.01,
      "learning_rate": 2.1393369175627243e-05,
      "loss": 2.2541,
      "step": 18300
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1374701314217443e-05,
      "loss": 2.5681,
      "step": 18310
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1356033452807646e-05,
      "loss": 2.6514,
      "step": 18320
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1337365591397852e-05,
      "loss": 2.2708,
      "step": 18330
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1318697729988056e-05,
      "loss": 2.9859,
      "step": 18340
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1300029868578256e-05,
      "loss": 2.4121,
      "step": 18350
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.128136200716846e-05,
      "loss": 2.8401,
      "step": 18360
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1262694145758662e-05,
      "loss": 2.0993,
      "step": 18370
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1244026284348865e-05,
      "loss": 2.3191,
      "step": 18380
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.122535842293907e-05,
      "loss": 2.4641,
      "step": 18390
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.120669056152927e-05,
      "loss": 2.6473,
      "step": 18400
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1188022700119475e-05,
      "loss": 2.4044,
      "step": 18410
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1169354838709678e-05,
      "loss": 2.5554,
      "step": 18420
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.115068697729988e-05,
      "loss": 1.9765,
      "step": 18430
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1132019115890085e-05,
      "loss": 2.5409,
      "step": 18440
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1113351254480288e-05,
      "loss": 2.3166,
      "step": 18450
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.109468339307049e-05,
      "loss": 2.5497,
      "step": 18460
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1076015531660694e-05,
      "loss": 2.9819,
      "step": 18470
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1057347670250897e-05,
      "loss": 2.4387,
      "step": 18480
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.10386798088411e-05,
      "loss": 2.5872,
      "step": 18490
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1020011947431304e-05,
      "loss": 2.5181,
      "step": 18500
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.1001344086021507e-05,
      "loss": 2.6028,
      "step": 18510
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0982676224611707e-05,
      "loss": 2.228,
      "step": 18520
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0964008363201913e-05,
      "loss": 2.2527,
      "step": 18530
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0945340501792117e-05,
      "loss": 2.223,
      "step": 18540
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.092667264038232e-05,
      "loss": 2.6303,
      "step": 18550
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0908004778972523e-05,
      "loss": 2.6447,
      "step": 18560
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0889336917562723e-05,
      "loss": 2.9795,
      "step": 18570
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0870669056152926e-05,
      "loss": 2.63,
      "step": 18580
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0852001194743133e-05,
      "loss": 2.6006,
      "step": 18590
    },
    {
      "epoch": 3.02,
      "learning_rate": 2.0833333333333336e-05,
      "loss": 2.7452,
      "step": 18600
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0814665471923536e-05,
      "loss": 2.437,
      "step": 18610
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.079599761051374e-05,
      "loss": 2.0574,
      "step": 18620
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0777329749103942e-05,
      "loss": 2.1847,
      "step": 18630
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.075866188769415e-05,
      "loss": 2.5653,
      "step": 18640
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0739994026284352e-05,
      "loss": 2.3355,
      "step": 18650
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0721326164874552e-05,
      "loss": 3.0245,
      "step": 18660
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0702658303464755e-05,
      "loss": 2.5511,
      "step": 18670
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.068399044205496e-05,
      "loss": 2.259,
      "step": 18680
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0665322580645165e-05,
      "loss": 2.6907,
      "step": 18690
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0646654719235365e-05,
      "loss": 2.5976,
      "step": 18700
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0627986857825568e-05,
      "loss": 2.8016,
      "step": 18710
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.060931899641577e-05,
      "loss": 2.8278,
      "step": 18720
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0590651135005974e-05,
      "loss": 2.7738,
      "step": 18730
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0571983273596178e-05,
      "loss": 2.7854,
      "step": 18740
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.055331541218638e-05,
      "loss": 2.0146,
      "step": 18750
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0534647550776584e-05,
      "loss": 2.6256,
      "step": 18760
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0515979689366787e-05,
      "loss": 2.4052,
      "step": 18770
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.049731182795699e-05,
      "loss": 2.2158,
      "step": 18780
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0478643966547194e-05,
      "loss": 3.4843,
      "step": 18790
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0459976105137397e-05,
      "loss": 2.6808,
      "step": 18800
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.04413082437276e-05,
      "loss": 1.9205,
      "step": 18810
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0422640382317803e-05,
      "loss": 2.1148,
      "step": 18820
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0403972520908003e-05,
      "loss": 2.8864,
      "step": 18830
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.038530465949821e-05,
      "loss": 2.2261,
      "step": 18840
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0366636798088413e-05,
      "loss": 2.675,
      "step": 18850
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0347968936678616e-05,
      "loss": 2.539,
      "step": 18860
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.032930107526882e-05,
      "loss": 2.1786,
      "step": 18870
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.031063321385902e-05,
      "loss": 2.362,
      "step": 18880
    },
    {
      "epoch": 3.03,
      "learning_rate": 2.0291965352449223e-05,
      "loss": 2.2568,
      "step": 18890
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.027329749103943e-05,
      "loss": 2.5926,
      "step": 18900
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0254629629629632e-05,
      "loss": 2.5037,
      "step": 18910
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0235961768219832e-05,
      "loss": 2.4759,
      "step": 18920
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0217293906810035e-05,
      "loss": 2.4399,
      "step": 18930
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.019862604540024e-05,
      "loss": 2.7335,
      "step": 18940
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0179958183990445e-05,
      "loss": 1.801,
      "step": 18950
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0161290322580645e-05,
      "loss": 2.0234,
      "step": 18960
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.014262246117085e-05,
      "loss": 2.5517,
      "step": 18970
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.012395459976105e-05,
      "loss": 2.9225,
      "step": 18980
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0105286738351255e-05,
      "loss": 3.2185,
      "step": 18990
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.008661887694146e-05,
      "loss": 2.3455,
      "step": 19000
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.006795101553166e-05,
      "loss": 2.636,
      "step": 19010
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0049283154121864e-05,
      "loss": 3.0057,
      "step": 19020
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.0030615292712068e-05,
      "loss": 2.7138,
      "step": 19030
    },
    {
      "epoch": 3.04,
      "learning_rate": 2.001194743130227e-05,
      "loss": 2.5631,
      "step": 19040
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9993279569892474e-05,
      "loss": 2.4569,
      "step": 19050
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9974611708482677e-05,
      "loss": 2.661,
      "step": 19060
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.995594384707288e-05,
      "loss": 2.7164,
      "step": 19070
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9937275985663084e-05,
      "loss": 2.1074,
      "step": 19080
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9918608124253284e-05,
      "loss": 2.6809,
      "step": 19090
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.989994026284349e-05,
      "loss": 2.5848,
      "step": 19100
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9881272401433693e-05,
      "loss": 2.6178,
      "step": 19110
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9862604540023897e-05,
      "loss": 2.085,
      "step": 19120
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.98439366786141e-05,
      "loss": 2.6144,
      "step": 19130
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.98252688172043e-05,
      "loss": 3.1553,
      "step": 19140
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9806600955794506e-05,
      "loss": 2.4955,
      "step": 19150
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.978793309438471e-05,
      "loss": 2.6846,
      "step": 19160
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9769265232974913e-05,
      "loss": 2.4775,
      "step": 19170
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9750597371565113e-05,
      "loss": 2.3898,
      "step": 19180
    },
    {
      "epoch": 3.04,
      "learning_rate": 1.9731929510155316e-05,
      "loss": 2.5732,
      "step": 19190
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9713261648745522e-05,
      "loss": 2.215,
      "step": 19200
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9694593787335726e-05,
      "loss": 3.0474,
      "step": 19210
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.967592592592593e-05,
      "loss": 2.8967,
      "step": 19220
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.965725806451613e-05,
      "loss": 2.6113,
      "step": 19230
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9638590203106332e-05,
      "loss": 2.4531,
      "step": 19240
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9619922341696535e-05,
      "loss": 2.0877,
      "step": 19250
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9601254480286742e-05,
      "loss": 2.8603,
      "step": 19260
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.958258661887694e-05,
      "loss": 2.2625,
      "step": 19270
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9563918757467145e-05,
      "loss": 2.5401,
      "step": 19280
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9545250896057348e-05,
      "loss": 2.8542,
      "step": 19290
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.952658303464755e-05,
      "loss": 2.2006,
      "step": 19300
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9507915173237754e-05,
      "loss": 2.5843,
      "step": 19310
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9489247311827958e-05,
      "loss": 2.6987,
      "step": 19320
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.947057945041816e-05,
      "loss": 2.4344,
      "step": 19330
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9451911589008364e-05,
      "loss": 2.3605,
      "step": 19340
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9433243727598567e-05,
      "loss": 2.1491,
      "step": 19350
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.941457586618877e-05,
      "loss": 2.4875,
      "step": 19360
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9395908004778974e-05,
      "loss": 2.4786,
      "step": 19370
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9377240143369177e-05,
      "loss": 1.6828,
      "step": 19380
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.935857228195938e-05,
      "loss": 1.8907,
      "step": 19390
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.933990442054958e-05,
      "loss": 2.6151,
      "step": 19400
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9321236559139787e-05,
      "loss": 2.2892,
      "step": 19410
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.930256869772999e-05,
      "loss": 2.9471,
      "step": 19420
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9283900836320193e-05,
      "loss": 2.4795,
      "step": 19430
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9265232974910393e-05,
      "loss": 1.971,
      "step": 19440
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9246565113500596e-05,
      "loss": 2.3115,
      "step": 19450
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9227897252090803e-05,
      "loss": 2.1893,
      "step": 19460
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9209229390681006e-05,
      "loss": 3.3262,
      "step": 19470
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.919056152927121e-05,
      "loss": 2.5822,
      "step": 19480
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.917189366786141e-05,
      "loss": 2.9456,
      "step": 19490
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9153225806451612e-05,
      "loss": 2.169,
      "step": 19500
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.913455794504182e-05,
      "loss": 2.1284,
      "step": 19510
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9115890083632022e-05,
      "loss": 2.5261,
      "step": 19520
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9097222222222222e-05,
      "loss": 2.7204,
      "step": 19530
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9078554360812425e-05,
      "loss": 2.2162,
      "step": 19540
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9059886499402628e-05,
      "loss": 2.8995,
      "step": 19550
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.904121863799283e-05,
      "loss": 2.0861,
      "step": 19560
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9022550776583038e-05,
      "loss": 2.6926,
      "step": 19570
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.9003882915173238e-05,
      "loss": 2.4891,
      "step": 19580
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.898521505376344e-05,
      "loss": 1.9507,
      "step": 19590
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8966547192353644e-05,
      "loss": 3.007,
      "step": 19600
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8947879330943848e-05,
      "loss": 1.6923,
      "step": 19610
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.892921146953405e-05,
      "loss": 2.6424,
      "step": 19620
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8910543608124254e-05,
      "loss": 2.8237,
      "step": 19630
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8891875746714457e-05,
      "loss": 2.8638,
      "step": 19640
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.887320788530466e-05,
      "loss": 2.6359,
      "step": 19650
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8854540023894864e-05,
      "loss": 1.9942,
      "step": 19660
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8835872162485067e-05,
      "loss": 2.0838,
      "step": 19670
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.881720430107527e-05,
      "loss": 2.4944,
      "step": 19680
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8798536439665473e-05,
      "loss": 2.6965,
      "step": 19690
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8779868578255677e-05,
      "loss": 2.3801,
      "step": 19700
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8761200716845876e-05,
      "loss": 2.1479,
      "step": 19710
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8742532855436083e-05,
      "loss": 2.2529,
      "step": 19720
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8723864994026286e-05,
      "loss": 2.8314,
      "step": 19730
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.870519713261649e-05,
      "loss": 2.5973,
      "step": 19740
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.868652927120669e-05,
      "loss": 2.453,
      "step": 19750
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8667861409796893e-05,
      "loss": 2.7604,
      "step": 19760
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.86491935483871e-05,
      "loss": 1.8178,
      "step": 19770
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8630525686977302e-05,
      "loss": 3.3395,
      "step": 19780
    },
    {
      "epoch": 3.06,
      "learning_rate": 1.8611857825567506e-05,
      "loss": 2.2357,
      "step": 19790
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8593189964157705e-05,
      "loss": 2.1135,
      "step": 19800
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.857452210274791e-05,
      "loss": 2.5988,
      "step": 19810
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8555854241338115e-05,
      "loss": 2.7436,
      "step": 19820
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.853718637992832e-05,
      "loss": 2.156,
      "step": 19830
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8518518518518518e-05,
      "loss": 2.286,
      "step": 19840
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.849985065710872e-05,
      "loss": 2.1962,
      "step": 19850
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8481182795698925e-05,
      "loss": 2.2311,
      "step": 19860
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8462514934289128e-05,
      "loss": 2.1722,
      "step": 19870
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.844384707287933e-05,
      "loss": 2.526,
      "step": 19880
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8425179211469534e-05,
      "loss": 2.1981,
      "step": 19890
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8406511350059738e-05,
      "loss": 2.2492,
      "step": 19900
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.838784348864994e-05,
      "loss": 2.1934,
      "step": 19910
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8369175627240144e-05,
      "loss": 2.8237,
      "step": 19920
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8350507765830347e-05,
      "loss": 2.4443,
      "step": 19930
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.833183990442055e-05,
      "loss": 1.8985,
      "step": 19940
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8313172043010754e-05,
      "loss": 1.9182,
      "step": 19950
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8294504181600957e-05,
      "loss": 3.0129,
      "step": 19960
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.827583632019116e-05,
      "loss": 1.9888,
      "step": 19970
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8257168458781363e-05,
      "loss": 2.2065,
      "step": 19980
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8238500597371567e-05,
      "loss": 2.783,
      "step": 19990
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.821983273596177e-05,
      "loss": 2.2629,
      "step": 20000
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.820116487455197e-05,
      "loss": 2.4861,
      "step": 20010
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8182497013142176e-05,
      "loss": 1.7836,
      "step": 20020
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.816382915173238e-05,
      "loss": 2.8726,
      "step": 20030
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8145161290322583e-05,
      "loss": 2.2979,
      "step": 20040
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8126493428912786e-05,
      "loss": 1.8698,
      "step": 20050
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8107825567502986e-05,
      "loss": 2.0086,
      "step": 20060
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.808915770609319e-05,
      "loss": 2.0377,
      "step": 20070
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.8070489844683396e-05,
      "loss": 2.0188,
      "step": 20080
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.80518219832736e-05,
      "loss": 2.1468,
      "step": 20090
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.80331541218638e-05,
      "loss": 2.6763,
      "step": 20100
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.8014486260454002e-05,
      "loss": 2.4538,
      "step": 20110
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7995818399044205e-05,
      "loss": 2.8785,
      "step": 20120
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.797715053763441e-05,
      "loss": 2.014,
      "step": 20130
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7958482676224615e-05,
      "loss": 3.096,
      "step": 20140
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7939814814814815e-05,
      "loss": 2.4784,
      "step": 20150
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7921146953405018e-05,
      "loss": 1.5835,
      "step": 20160
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.790247909199522e-05,
      "loss": 2.1188,
      "step": 20170
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7883811230585428e-05,
      "loss": 2.8463,
      "step": 20180
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7865143369175628e-05,
      "loss": 2.1697,
      "step": 20190
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.784647550776583e-05,
      "loss": 2.6546,
      "step": 20200
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7827807646356034e-05,
      "loss": 2.2187,
      "step": 20210
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7809139784946237e-05,
      "loss": 1.5922,
      "step": 20220
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.779047192353644e-05,
      "loss": 2.4411,
      "step": 20230
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7771804062126644e-05,
      "loss": 2.0552,
      "step": 20240
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7753136200716847e-05,
      "loss": 2.3645,
      "step": 20250
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.773446833930705e-05,
      "loss": 2.426,
      "step": 20260
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7715800477897253e-05,
      "loss": 2.0695,
      "step": 20270
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7697132616487457e-05,
      "loss": 3.0191,
      "step": 20280
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.767846475507766e-05,
      "loss": 2.7107,
      "step": 20290
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7659796893667863e-05,
      "loss": 2.454,
      "step": 20300
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7641129032258066e-05,
      "loss": 2.2402,
      "step": 20310
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7622461170848266e-05,
      "loss": 2.0683,
      "step": 20320
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7603793309438473e-05,
      "loss": 2.2405,
      "step": 20330
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7585125448028676e-05,
      "loss": 2.3465,
      "step": 20340
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.756645758661888e-05,
      "loss": 1.601,
      "step": 20350
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7547789725209082e-05,
      "loss": 2.009,
      "step": 20360
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7529121863799282e-05,
      "loss": 2.19,
      "step": 20370
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.7510454002389485e-05,
      "loss": 2.1226,
      "step": 20380
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7491786140979692e-05,
      "loss": 1.8346,
      "step": 20390
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7473118279569895e-05,
      "loss": 2.1178,
      "step": 20400
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7454450418160095e-05,
      "loss": 2.1185,
      "step": 20410
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7435782556750298e-05,
      "loss": 1.8201,
      "step": 20420
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.74171146953405e-05,
      "loss": 1.7933,
      "step": 20430
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7398446833930708e-05,
      "loss": 2.287,
      "step": 20440
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7379778972520908e-05,
      "loss": 2.2779,
      "step": 20450
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.736111111111111e-05,
      "loss": 2.7157,
      "step": 20460
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7342443249701314e-05,
      "loss": 2.3053,
      "step": 20470
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7323775388291518e-05,
      "loss": 2.1541,
      "step": 20480
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7305107526881724e-05,
      "loss": 2.6438,
      "step": 20490
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7286439665471924e-05,
      "loss": 2.4277,
      "step": 20500
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7267771804062127e-05,
      "loss": 2.4426,
      "step": 20510
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.724910394265233e-05,
      "loss": 2.0442,
      "step": 20520
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7230436081242534e-05,
      "loss": 2.3052,
      "step": 20530
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7211768219832737e-05,
      "loss": 1.9107,
      "step": 20540
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.719310035842294e-05,
      "loss": 2.1553,
      "step": 20550
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7174432497013143e-05,
      "loss": 1.6635,
      "step": 20560
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7155764635603347e-05,
      "loss": 2.6077,
      "step": 20570
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7137096774193546e-05,
      "loss": 2.2606,
      "step": 20580
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7118428912783753e-05,
      "loss": 2.1144,
      "step": 20590
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7099761051373956e-05,
      "loss": 1.8549,
      "step": 20600
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.708109318996416e-05,
      "loss": 1.9439,
      "step": 20610
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7062425328554363e-05,
      "loss": 2.2266,
      "step": 20620
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7043757467144562e-05,
      "loss": 2.4845,
      "step": 20630
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.702508960573477e-05,
      "loss": 2.4241,
      "step": 20640
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.7006421744324972e-05,
      "loss": 2.1277,
      "step": 20650
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.6987753882915175e-05,
      "loss": 2.1097,
      "step": 20660
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.6969086021505375e-05,
      "loss": 2.4681,
      "step": 20670
    },
    {
      "epoch": 3.09,
      "learning_rate": 1.695041816009558e-05,
      "loss": 2.4307,
      "step": 20680
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6931750298685782e-05,
      "loss": 2.2657,
      "step": 20690
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.691308243727599e-05,
      "loss": 1.995,
      "step": 20700
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.689441457586619e-05,
      "loss": 2.2158,
      "step": 20710
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.687574671445639e-05,
      "loss": 2.3378,
      "step": 20720
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6857078853046595e-05,
      "loss": 2.9261,
      "step": 20730
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6838410991636798e-05,
      "loss": 2.3966,
      "step": 20740
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6819743130227004e-05,
      "loss": 2.4051,
      "step": 20750
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6801075268817204e-05,
      "loss": 1.825,
      "step": 20760
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6782407407407408e-05,
      "loss": 2.1746,
      "step": 20770
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.676373954599761e-05,
      "loss": 2.2581,
      "step": 20780
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6745071684587814e-05,
      "loss": 1.8986,
      "step": 20790
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6726403823178017e-05,
      "loss": 2.5394,
      "step": 20800
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.670773596176822e-05,
      "loss": 2.1355,
      "step": 20810
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6689068100358424e-05,
      "loss": 2.3145,
      "step": 20820
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6670400238948627e-05,
      "loss": 2.142,
      "step": 20830
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.665173237753883e-05,
      "loss": 2.208,
      "step": 20840
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6633064516129033e-05,
      "loss": 2.5949,
      "step": 20850
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6614396654719236e-05,
      "loss": 1.8831,
      "step": 20860
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.659572879330944e-05,
      "loss": 2.4052,
      "step": 20870
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6577060931899643e-05,
      "loss": 1.6154,
      "step": 20880
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6558393070489843e-05,
      "loss": 1.7053,
      "step": 20890
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.653972520908005e-05,
      "loss": 2.789,
      "step": 20900
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6521057347670253e-05,
      "loss": 2.2677,
      "step": 20910
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6502389486260456e-05,
      "loss": 2.2517,
      "step": 20920
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6483721624850656e-05,
      "loss": 1.6777,
      "step": 20930
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.646505376344086e-05,
      "loss": 2.1637,
      "step": 20940
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6446385902031065e-05,
      "loss": 2.2388,
      "step": 20950
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.642771804062127e-05,
      "loss": 2.4399,
      "step": 20960
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6409050179211472e-05,
      "loss": 2.5724,
      "step": 20970
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.6390382317801672e-05,
      "loss": 2.0501,
      "step": 20980
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6371714456391875e-05,
      "loss": 1.91,
      "step": 20990
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.635304659498208e-05,
      "loss": 2.8361,
      "step": 21000
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6334378733572285e-05,
      "loss": 2.1849,
      "step": 21010
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6315710872162485e-05,
      "loss": 2.5023,
      "step": 21020
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6297043010752688e-05,
      "loss": 2.1698,
      "step": 21030
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.627837514934289e-05,
      "loss": 1.9825,
      "step": 21040
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6259707287933094e-05,
      "loss": 2.3123,
      "step": 21050
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.62410394265233e-05,
      "loss": 2.4455,
      "step": 21060
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.62223715651135e-05,
      "loss": 2.3476,
      "step": 21070
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6203703703703704e-05,
      "loss": 1.8119,
      "step": 21080
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6185035842293907e-05,
      "loss": 2.1137,
      "step": 21090
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.616636798088411e-05,
      "loss": 2.701,
      "step": 21100
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6147700119474314e-05,
      "loss": 2.2716,
      "step": 21110
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6129032258064517e-05,
      "loss": 2.1489,
      "step": 21120
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.611036439665472e-05,
      "loss": 2.0791,
      "step": 21130
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6091696535244923e-05,
      "loss": 1.7618,
      "step": 21140
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6073028673835126e-05,
      "loss": 1.4273,
      "step": 21150
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.605436081242533e-05,
      "loss": 2.1745,
      "step": 21160
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6035692951015533e-05,
      "loss": 2.2529,
      "step": 21170
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.6017025089605736e-05,
      "loss": 2.2853,
      "step": 21180
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.599835722819594e-05,
      "loss": 2.0975,
      "step": 21190
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.597968936678614e-05,
      "loss": 1.9055,
      "step": 21200
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.5961021505376346e-05,
      "loss": 1.9013,
      "step": 21210
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.594235364396655e-05,
      "loss": 1.9079,
      "step": 21220
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.5923685782556752e-05,
      "loss": 2.4706,
      "step": 21230
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.5905017921146952e-05,
      "loss": 1.9529,
      "step": 21240
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.5886350059737155e-05,
      "loss": 2.4121,
      "step": 21250
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.5867682198327362e-05,
      "loss": 1.8514,
      "step": 21260
    },
    {
      "epoch": 3.11,
      "learning_rate": 1.5849014336917565e-05,
      "loss": 1.4366,
      "step": 21270
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5830346475507768e-05,
      "loss": 2.2499,
      "step": 21280
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5811678614097968e-05,
      "loss": 2.1375,
      "step": 21290
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.579301075268817e-05,
      "loss": 2.0315,
      "step": 21300
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5774342891278378e-05,
      "loss": 1.8211,
      "step": 21310
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.575567502986858e-05,
      "loss": 1.9708,
      "step": 21320
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.573700716845878e-05,
      "loss": 2.2567,
      "step": 21330
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5718339307048984e-05,
      "loss": 2.1037,
      "step": 21340
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5699671445639187e-05,
      "loss": 2.0825,
      "step": 21350
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.568100358422939e-05,
      "loss": 1.9985,
      "step": 21360
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5662335722819594e-05,
      "loss": 2.2336,
      "step": 21370
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5643667861409797e-05,
      "loss": 2.6434,
      "step": 21380
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5625e-05,
      "loss": 2.6117,
      "step": 21390
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5606332138590204e-05,
      "loss": 2.0422,
      "step": 21400
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5587664277180407e-05,
      "loss": 2.0718,
      "step": 21410
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.556899641577061e-05,
      "loss": 2.2961,
      "step": 21420
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5550328554360813e-05,
      "loss": 1.9545,
      "step": 21430
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5531660692951016e-05,
      "loss": 2.1755,
      "step": 21440
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.551299283154122e-05,
      "loss": 2.3679,
      "step": 21450
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5494324970131423e-05,
      "loss": 1.9657,
      "step": 21460
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5475657108721626e-05,
      "loss": 1.9233,
      "step": 21470
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.545698924731183e-05,
      "loss": 1.9118,
      "step": 21480
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5438321385902033e-05,
      "loss": 2.0788,
      "step": 21490
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5419653524492232e-05,
      "loss": 2.4147,
      "step": 21500
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.540098566308244e-05,
      "loss": 1.9939,
      "step": 21510
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5382317801672642e-05,
      "loss": 2.2315,
      "step": 21520
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5363649940262845e-05,
      "loss": 2.6876,
      "step": 21530
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.534498207885305e-05,
      "loss": 1.9245,
      "step": 21540
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.532631421744325e-05,
      "loss": 2.2156,
      "step": 21550
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.530764635603345e-05,
      "loss": 2.2324,
      "step": 21560
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.5288978494623658e-05,
      "loss": 2.0756,
      "step": 21570
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.527031063321386e-05,
      "loss": 1.732,
      "step": 21580
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5251642771804061e-05,
      "loss": 1.9018,
      "step": 21590
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5232974910394265e-05,
      "loss": 1.8123,
      "step": 21600
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.521430704898447e-05,
      "loss": 2.3712,
      "step": 21610
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5195639187574673e-05,
      "loss": 2.5202,
      "step": 21620
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5176971326164876e-05,
      "loss": 1.3138,
      "step": 21630
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5158303464755077e-05,
      "loss": 2.1082,
      "step": 21640
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.513963560334528e-05,
      "loss": 2.1131,
      "step": 21650
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5120967741935486e-05,
      "loss": 2.0615,
      "step": 21660
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5102299880525689e-05,
      "loss": 2.5166,
      "step": 21670
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.508363201911589e-05,
      "loss": 2.6684,
      "step": 21680
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5064964157706094e-05,
      "loss": 1.8802,
      "step": 21690
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5046296296296297e-05,
      "loss": 2.1948,
      "step": 21700
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5027628434886502e-05,
      "loss": 2.0427,
      "step": 21710
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.5008960573476701e-05,
      "loss": 2.5439,
      "step": 21720
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4990292712066906e-05,
      "loss": 2.0657,
      "step": 21730
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.497162485065711e-05,
      "loss": 2.1985,
      "step": 21740
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4952956989247313e-05,
      "loss": 1.9323,
      "step": 21750
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4934289127837516e-05,
      "loss": 2.0768,
      "step": 21760
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4915621266427718e-05,
      "loss": 1.8996,
      "step": 21770
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.489695340501792e-05,
      "loss": 2.0418,
      "step": 21780
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4878285543608126e-05,
      "loss": 1.4946,
      "step": 21790
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4859617682198329e-05,
      "loss": 2.1865,
      "step": 21800
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.484094982078853e-05,
      "loss": 2.0617,
      "step": 21810
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4822281959378734e-05,
      "loss": 1.8996,
      "step": 21820
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4803614097968937e-05,
      "loss": 1.8901,
      "step": 21830
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4784946236559142e-05,
      "loss": 2.4669,
      "step": 21840
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4766278375149345e-05,
      "loss": 2.5433,
      "step": 21850
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.4747610513739547e-05,
      "loss": 1.7593,
      "step": 21860
    },
    {
      "epoch": 3.13,
      "learning_rate": 1.472894265232975e-05,
      "loss": 2.0914,
      "step": 21870
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4710274790919953e-05,
      "loss": 2.1832,
      "step": 21880
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4691606929510158e-05,
      "loss": 1.8268,
      "step": 21890
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4672939068100358e-05,
      "loss": 2.0511,
      "step": 21900
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4654271206690563e-05,
      "loss": 2.3323,
      "step": 21910
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4635603345280766e-05,
      "loss": 1.6998,
      "step": 21920
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4616935483870969e-05,
      "loss": 1.9985,
      "step": 21930
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.459826762246117e-05,
      "loss": 1.7915,
      "step": 21940
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4579599761051374e-05,
      "loss": 1.7092,
      "step": 21950
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4560931899641577e-05,
      "loss": 1.9216,
      "step": 21960
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4542264038231782e-05,
      "loss": 1.9626,
      "step": 21970
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4523596176821985e-05,
      "loss": 2.1571,
      "step": 21980
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4504928315412187e-05,
      "loss": 2.0591,
      "step": 21990
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.448626045400239e-05,
      "loss": 1.5167,
      "step": 22000
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4467592592592593e-05,
      "loss": 1.7718,
      "step": 22010
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4448924731182798e-05,
      "loss": 1.6548,
      "step": 22020
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4430256869772998e-05,
      "loss": 1.6625,
      "step": 22030
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4411589008363203e-05,
      "loss": 2.0432,
      "step": 22040
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4392921146953406e-05,
      "loss": 1.8862,
      "step": 22050
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.437425328554361e-05,
      "loss": 1.5782,
      "step": 22060
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.435558542413381e-05,
      "loss": 2.2837,
      "step": 22070
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4336917562724014e-05,
      "loss": 2.0321,
      "step": 22080
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4318249701314219e-05,
      "loss": 1.7268,
      "step": 22090
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4299581839904422e-05,
      "loss": 1.8699,
      "step": 22100
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4280913978494625e-05,
      "loss": 1.0975,
      "step": 22110
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4262246117084827e-05,
      "loss": 2.1251,
      "step": 22120
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.424357825567503e-05,
      "loss": 1.2742,
      "step": 22130
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4224910394265233e-05,
      "loss": 2.1263,
      "step": 22140
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4206242532855438e-05,
      "loss": 2.5516,
      "step": 22150
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4187574671445638e-05,
      "loss": 1.5225,
      "step": 22160
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.4168906810035843e-05,
      "loss": 2.2144,
      "step": 22170
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4150238948626046e-05,
      "loss": 2.056,
      "step": 22180
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.413157108721625e-05,
      "loss": 1.5901,
      "step": 22190
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4112903225806454e-05,
      "loss": 1.9209,
      "step": 22200
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4094235364396654e-05,
      "loss": 1.8826,
      "step": 22210
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4075567502986859e-05,
      "loss": 1.7103,
      "step": 22220
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4056899641577062e-05,
      "loss": 2.4449,
      "step": 22230
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4038231780167265e-05,
      "loss": 1.8365,
      "step": 22240
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.4019563918757467e-05,
      "loss": 1.8515,
      "step": 22250
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.400089605734767e-05,
      "loss": 1.5734,
      "step": 22260
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3982228195937873e-05,
      "loss": 2.1358,
      "step": 22270
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3963560334528078e-05,
      "loss": 1.7191,
      "step": 22280
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3944892473118278e-05,
      "loss": 1.9424,
      "step": 22290
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3926224611708483e-05,
      "loss": 1.9864,
      "step": 22300
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3907556750298686e-05,
      "loss": 1.9016,
      "step": 22310
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.388888888888889e-05,
      "loss": 1.4023,
      "step": 22320
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3870221027479094e-05,
      "loss": 1.9216,
      "step": 22330
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3851553166069294e-05,
      "loss": 2.1977,
      "step": 22340
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.38328853046595e-05,
      "loss": 1.9146,
      "step": 22350
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3814217443249702e-05,
      "loss": 2.1622,
      "step": 22360
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3795549581839906e-05,
      "loss": 1.3267,
      "step": 22370
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3776881720430107e-05,
      "loss": 2.5444,
      "step": 22380
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.375821385902031e-05,
      "loss": 2.5128,
      "step": 22390
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3739545997610515e-05,
      "loss": 2.335,
      "step": 22400
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3720878136200719e-05,
      "loss": 2.0817,
      "step": 22410
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.370221027479092e-05,
      "loss": 2.0941,
      "step": 22420
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3683542413381123e-05,
      "loss": 1.984,
      "step": 22430
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3664874551971326e-05,
      "loss": 1.8656,
      "step": 22440
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.364620669056153e-05,
      "loss": 1.5047,
      "step": 22450
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.3627538829151735e-05,
      "loss": 2.2893,
      "step": 22460
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3608870967741934e-05,
      "loss": 1.9264,
      "step": 22470
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.359020310633214e-05,
      "loss": 2.1947,
      "step": 22480
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3571535244922343e-05,
      "loss": 1.4356,
      "step": 22490
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3552867383512546e-05,
      "loss": 1.8751,
      "step": 22500
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3534199522102747e-05,
      "loss": 1.327,
      "step": 22510
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.351553166069295e-05,
      "loss": 2.0069,
      "step": 22520
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3496863799283155e-05,
      "loss": 1.8235,
      "step": 22530
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3478195937873359e-05,
      "loss": 1.5983,
      "step": 22540
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3459528076463562e-05,
      "loss": 1.7298,
      "step": 22550
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3440860215053763e-05,
      "loss": 1.6556,
      "step": 22560
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3422192353643967e-05,
      "loss": 1.6897,
      "step": 22570
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3403524492234172e-05,
      "loss": 1.9386,
      "step": 22580
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3384856630824375e-05,
      "loss": 1.8859,
      "step": 22590
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3366188769414575e-05,
      "loss": 2.6273,
      "step": 22600
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.334752090800478e-05,
      "loss": 2.4273,
      "step": 22610
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3328853046594983e-05,
      "loss": 2.3851,
      "step": 22620
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3310185185185186e-05,
      "loss": 2.0836,
      "step": 22630
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3291517323775387e-05,
      "loss": 1.6597,
      "step": 22640
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.327284946236559e-05,
      "loss": 1.7808,
      "step": 22650
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3254181600955796e-05,
      "loss": 2.1817,
      "step": 22660
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3235513739545999e-05,
      "loss": 1.8611,
      "step": 22670
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3216845878136202e-05,
      "loss": 1.5431,
      "step": 22680
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3198178016726404e-05,
      "loss": 1.7751,
      "step": 22690
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3179510155316607e-05,
      "loss": 1.6953,
      "step": 22700
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3160842293906812e-05,
      "loss": 2.5849,
      "step": 22710
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3142174432497015e-05,
      "loss": 1.3372,
      "step": 22720
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3123506571087216e-05,
      "loss": 1.8606,
      "step": 22730
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.310483870967742e-05,
      "loss": 2.0103,
      "step": 22740
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3086170848267623e-05,
      "loss": 2.4922,
      "step": 22750
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.3067502986857826e-05,
      "loss": 2.1738,
      "step": 22760
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3048835125448031e-05,
      "loss": 1.9274,
      "step": 22770
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3030167264038231e-05,
      "loss": 1.8831,
      "step": 22780
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.3011499402628436e-05,
      "loss": 1.5739,
      "step": 22790
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2992831541218639e-05,
      "loss": 1.7399,
      "step": 22800
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2974163679808842e-05,
      "loss": 2.4286,
      "step": 22810
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2955495818399044e-05,
      "loss": 1.7934,
      "step": 22820
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2936827956989247e-05,
      "loss": 2.3587,
      "step": 22830
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2918160095579452e-05,
      "loss": 1.6229,
      "step": 22840
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2899492234169655e-05,
      "loss": 1.9592,
      "step": 22850
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2880824372759857e-05,
      "loss": 2.2153,
      "step": 22860
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.286215651135006e-05,
      "loss": 1.3733,
      "step": 22870
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2843488649940263e-05,
      "loss": 2.1415,
      "step": 22880
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2824820788530468e-05,
      "loss": 1.9484,
      "step": 22890
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2806152927120671e-05,
      "loss": 1.6596,
      "step": 22900
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2787485065710873e-05,
      "loss": 1.4818,
      "step": 22910
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2768817204301076e-05,
      "loss": 1.4102,
      "step": 22920
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.275014934289128e-05,
      "loss": 1.5991,
      "step": 22930
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2731481481481482e-05,
      "loss": 1.7665,
      "step": 22940
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2712813620071684e-05,
      "loss": 1.5075,
      "step": 22950
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2694145758661887e-05,
      "loss": 1.748,
      "step": 22960
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2675477897252092e-05,
      "loss": 1.9612,
      "step": 22970
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2656810035842295e-05,
      "loss": 1.6859,
      "step": 22980
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2638142174432497e-05,
      "loss": 1.861,
      "step": 22990
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.26194743130227e-05,
      "loss": 1.3929,
      "step": 23000
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2600806451612903e-05,
      "loss": 1.975,
      "step": 23010
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2582138590203108e-05,
      "loss": 1.1081,
      "step": 23020
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2563470728793311e-05,
      "loss": 1.3361,
      "step": 23030
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2544802867383513e-05,
      "loss": 1.865,
      "step": 23040
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.2526135005973716e-05,
      "loss": 1.4983,
      "step": 23050
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.250746714456392e-05,
      "loss": 2.1742,
      "step": 23060
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2488799283154123e-05,
      "loss": 1.7182,
      "step": 23070
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2470131421744326e-05,
      "loss": 1.6553,
      "step": 23080
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2451463560334527e-05,
      "loss": 1.7373,
      "step": 23090
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2432795698924732e-05,
      "loss": 1.5766,
      "step": 23100
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2414127837514935e-05,
      "loss": 1.4831,
      "step": 23110
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2395459976105139e-05,
      "loss": 2.0497,
      "step": 23120
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2376792114695342e-05,
      "loss": 1.8571,
      "step": 23130
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2358124253285543e-05,
      "loss": 1.3844,
      "step": 23140
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2339456391875748e-05,
      "loss": 1.8604,
      "step": 23150
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.232078853046595e-05,
      "loss": 1.8055,
      "step": 23160
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2302120669056153e-05,
      "loss": 1.7484,
      "step": 23170
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2283452807646356e-05,
      "loss": 1.8736,
      "step": 23180
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.226478494623656e-05,
      "loss": 1.9731,
      "step": 23190
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2246117084826763e-05,
      "loss": 2.0633,
      "step": 23200
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2227449223416966e-05,
      "loss": 1.8474,
      "step": 23210
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2208781362007169e-05,
      "loss": 2.5302,
      "step": 23220
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2190113500597372e-05,
      "loss": 1.6165,
      "step": 23230
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2171445639187576e-05,
      "loss": 1.6406,
      "step": 23240
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2152777777777779e-05,
      "loss": 2.5011,
      "step": 23250
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2134109916367982e-05,
      "loss": 1.5273,
      "step": 23260
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2115442054958184e-05,
      "loss": 2.3831,
      "step": 23270
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2096774193548388e-05,
      "loss": 2.2989,
      "step": 23280
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.207810633213859e-05,
      "loss": 2.1644,
      "step": 23290
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2059438470728795e-05,
      "loss": 1.7813,
      "step": 23300
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2040770609318996e-05,
      "loss": 1.9006,
      "step": 23310
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.20221027479092e-05,
      "loss": 1.8631,
      "step": 23320
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.2003434886499403e-05,
      "loss": 1.8753,
      "step": 23330
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.1984767025089606e-05,
      "loss": 2.1294,
      "step": 23340
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.196609916367981e-05,
      "loss": 1.9498,
      "step": 23350
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.1947431302270013e-05,
      "loss": 1.9288,
      "step": 23360
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1928763440860216e-05,
      "loss": 1.7415,
      "step": 23370
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1910095579450419e-05,
      "loss": 1.4501,
      "step": 23380
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1891427718040622e-05,
      "loss": 1.5426,
      "step": 23390
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1872759856630825e-05,
      "loss": 1.7795,
      "step": 23400
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1854091995221029e-05,
      "loss": 1.7883,
      "step": 23410
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.183542413381123e-05,
      "loss": 2.295,
      "step": 23420
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1816756272401435e-05,
      "loss": 2.2004,
      "step": 23430
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1798088410991637e-05,
      "loss": 1.3605,
      "step": 23440
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.177942054958184e-05,
      "loss": 1.99,
      "step": 23450
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1760752688172045e-05,
      "loss": 2.1217,
      "step": 23460
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1742084826762246e-05,
      "loss": 1.4244,
      "step": 23470
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1723416965352451e-05,
      "loss": 2.0407,
      "step": 23480
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1704749103942653e-05,
      "loss": 1.5053,
      "step": 23490
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1686081242532856e-05,
      "loss": 1.397,
      "step": 23500
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1667413381123059e-05,
      "loss": 1.5153,
      "step": 23510
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1648745519713262e-05,
      "loss": 1.7957,
      "step": 23520
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1630077658303466e-05,
      "loss": 1.161,
      "step": 23530
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1611409796893669e-05,
      "loss": 2.2998,
      "step": 23540
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.159274193548387e-05,
      "loss": 2.0918,
      "step": 23550
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1574074074074075e-05,
      "loss": 1.1215,
      "step": 23560
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1555406212664278e-05,
      "loss": 1.3754,
      "step": 23570
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.153673835125448e-05,
      "loss": 1.7389,
      "step": 23580
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1518070489844685e-05,
      "loss": 1.2926,
      "step": 23590
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1499402628434886e-05,
      "loss": 1.9691,
      "step": 23600
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1480734767025091e-05,
      "loss": 2.0251,
      "step": 23610
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1462066905615293e-05,
      "loss": 1.7244,
      "step": 23620
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1443399044205496e-05,
      "loss": 1.3835,
      "step": 23630
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.14247311827957e-05,
      "loss": 1.313,
      "step": 23640
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.1406063321385902e-05,
      "loss": 2.0762,
      "step": 23650
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1387395459976106e-05,
      "loss": 1.5073,
      "step": 23660
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1368727598566309e-05,
      "loss": 2.1956,
      "step": 23670
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.135005973715651e-05,
      "loss": 1.5682,
      "step": 23680
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1331391875746715e-05,
      "loss": 2.0843,
      "step": 23690
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1312724014336919e-05,
      "loss": 1.5406,
      "step": 23700
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1294056152927122e-05,
      "loss": 1.2961,
      "step": 23710
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1275388291517325e-05,
      "loss": 1.5602,
      "step": 23720
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1256720430107527e-05,
      "loss": 1.7592,
      "step": 23730
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1238052568697731e-05,
      "loss": 1.885,
      "step": 23740
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1219384707287933e-05,
      "loss": 0.9263,
      "step": 23750
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1200716845878136e-05,
      "loss": 1.4397,
      "step": 23760
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.118204898446834e-05,
      "loss": 1.5965,
      "step": 23770
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1163381123058543e-05,
      "loss": 1.6573,
      "step": 23780
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1144713261648746e-05,
      "loss": 1.6953,
      "step": 23790
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.1126045400238949e-05,
      "loss": 1.722,
      "step": 23800
    },
    {
      "epoch": 3.2,
      "eval_accuracy": 0.42408943382618103,
      "eval_loss": 2.0974302291870117,
      "eval_runtime": 959.4358,
      "eval_samples_per_second": 2.89,
      "eval_steps_per_second": 1.446,
      "step": 23808
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1107377538829152e-05,
      "loss": 2.056,
      "step": 23810
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1088709677419356e-05,
      "loss": 1.4446,
      "step": 23820
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1070041816009559e-05,
      "loss": 1.599,
      "step": 23830
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1051373954599762e-05,
      "loss": 1.3334,
      "step": 23840
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1032706093189965e-05,
      "loss": 2.4581,
      "step": 23850
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.1014038231780167e-05,
      "loss": 1.5139,
      "step": 23860
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0995370370370372e-05,
      "loss": 1.4778,
      "step": 23870
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0976702508960573e-05,
      "loss": 1.5977,
      "step": 23880
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0958034647550778e-05,
      "loss": 1.79,
      "step": 23890
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.093936678614098e-05,
      "loss": 2.1424,
      "step": 23900
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0920698924731183e-05,
      "loss": 1.7417,
      "step": 23910
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0902031063321388e-05,
      "loss": 1.7907,
      "step": 23920
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.088336320191159e-05,
      "loss": 2.0001,
      "step": 23930
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0864695340501792e-05,
      "loss": 1.7665,
      "step": 23940
    },
    {
      "epoch": 4.0,
      "learning_rate": 1.0846027479091996e-05,
      "loss": 1.642,
      "step": 23950
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0827359617682199e-05,
      "loss": 1.1917,
      "step": 23960
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0808691756272402e-05,
      "loss": 1.9484,
      "step": 23970
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0790023894862605e-05,
      "loss": 1.8528,
      "step": 23980
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0771356033452807e-05,
      "loss": 1.6025,
      "step": 23990
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0752688172043012e-05,
      "loss": 1.4363,
      "step": 24000
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0734020310633213e-05,
      "loss": 1.7248,
      "step": 24010
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0715352449223418e-05,
      "loss": 2.3533,
      "step": 24020
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0696684587813621e-05,
      "loss": 1.8398,
      "step": 24030
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0678016726403823e-05,
      "loss": 1.6324,
      "step": 24040
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0659348864994028e-05,
      "loss": 1.6265,
      "step": 24050
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.064068100358423e-05,
      "loss": 1.9436,
      "step": 24060
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0622013142174433e-05,
      "loss": 1.6452,
      "step": 24070
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0603345280764636e-05,
      "loss": 1.9292,
      "step": 24080
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0584677419354839e-05,
      "loss": 1.0717,
      "step": 24090
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0566009557945042e-05,
      "loss": 1.4639,
      "step": 24100
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0547341696535245e-05,
      "loss": 1.789,
      "step": 24110
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0528673835125449e-05,
      "loss": 1.4203,
      "step": 24120
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0510005973715652e-05,
      "loss": 1.6756,
      "step": 24130
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0491338112305853e-05,
      "loss": 1.8001,
      "step": 24140
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0472670250896058e-05,
      "loss": 1.6158,
      "step": 24150
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0454002389486262e-05,
      "loss": 1.6266,
      "step": 24160
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0435334528076463e-05,
      "loss": 1.575,
      "step": 24170
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0416666666666668e-05,
      "loss": 1.6014,
      "step": 24180
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.039799880525687e-05,
      "loss": 1.6461,
      "step": 24190
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0379330943847074e-05,
      "loss": 2.0113,
      "step": 24200
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0360663082437276e-05,
      "loss": 1.7792,
      "step": 24210
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.034199522102748e-05,
      "loss": 1.2177,
      "step": 24220
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0323327359617682e-05,
      "loss": 1.3584,
      "step": 24230
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0304659498207886e-05,
      "loss": 1.8425,
      "step": 24240
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.0285991636798089e-05,
      "loss": 1.86,
      "step": 24250
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0267323775388292e-05,
      "loss": 1.7274,
      "step": 24260
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0248655913978495e-05,
      "loss": 1.6669,
      "step": 24270
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0229988052568699e-05,
      "loss": 1.3069,
      "step": 24280
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0211320191158902e-05,
      "loss": 1.6819,
      "step": 24290
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0192652329749105e-05,
      "loss": 2.055,
      "step": 24300
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0173984468339308e-05,
      "loss": 1.8674,
      "step": 24310
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.015531660692951e-05,
      "loss": 1.6398,
      "step": 24320
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0136648745519715e-05,
      "loss": 1.315,
      "step": 24330
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0117980884109916e-05,
      "loss": 1.781,
      "step": 24340
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.009931302270012e-05,
      "loss": 1.3848,
      "step": 24350
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0080645161290323e-05,
      "loss": 1.4403,
      "step": 24360
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0061977299880526e-05,
      "loss": 2.067,
      "step": 24370
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.004330943847073e-05,
      "loss": 1.1659,
      "step": 24380
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0024641577060932e-05,
      "loss": 1.1623,
      "step": 24390
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.0005973715651135e-05,
      "loss": 1.1522,
      "step": 24400
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.987305854241339e-06,
      "loss": 1.7553,
      "step": 24410
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.968637992831542e-06,
      "loss": 1.0157,
      "step": 24420
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.949970131421745e-06,
      "loss": 1.8805,
      "step": 24430
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.931302270011948e-06,
      "loss": 1.7155,
      "step": 24440
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.91263440860215e-06,
      "loss": 1.4454,
      "step": 24450
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.893966547192355e-06,
      "loss": 1.3156,
      "step": 24460
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.875298685782556e-06,
      "loss": 1.5967,
      "step": 24470
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.856630824372761e-06,
      "loss": 1.9421,
      "step": 24480
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.837962962962964e-06,
      "loss": 2.195,
      "step": 24490
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.819295101553166e-06,
      "loss": 1.5071,
      "step": 24500
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.800627240143371e-06,
      "loss": 1.8806,
      "step": 24510
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.781959378733572e-06,
      "loss": 1.3441,
      "step": 24520
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.763291517323776e-06,
      "loss": 1.7605,
      "step": 24530
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.744623655913979e-06,
      "loss": 1.2259,
      "step": 24540
    },
    {
      "epoch": 4.02,
      "learning_rate": 9.725955794504182e-06,
      "loss": 1.4053,
      "step": 24550
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.707287933094385e-06,
      "loss": 1.25,
      "step": 24560
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.688620071684588e-06,
      "loss": 1.3208,
      "step": 24570
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.66995221027479e-06,
      "loss": 1.3093,
      "step": 24580
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.651284348864995e-06,
      "loss": 1.5853,
      "step": 24590
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.632616487455196e-06,
      "loss": 1.4985,
      "step": 24600
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.613948626045401e-06,
      "loss": 1.4555,
      "step": 24610
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.595280764635605e-06,
      "loss": 1.7654,
      "step": 24620
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.576612903225806e-06,
      "loss": 0.9561,
      "step": 24630
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.557945041816011e-06,
      "loss": 1.5183,
      "step": 24640
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.539277180406213e-06,
      "loss": 1.1956,
      "step": 24650
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.520609318996416e-06,
      "loss": 1.2296,
      "step": 24660
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.501941457586619e-06,
      "loss": 1.0996,
      "step": 24670
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.483273596176822e-06,
      "loss": 1.5512,
      "step": 24680
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.464605734767025e-06,
      "loss": 1.1916,
      "step": 24690
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.445937873357229e-06,
      "loss": 1.1623,
      "step": 24700
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.427270011947432e-06,
      "loss": 1.5246,
      "step": 24710
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.408602150537635e-06,
      "loss": 1.9619,
      "step": 24720
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.389934289127838e-06,
      "loss": 1.6799,
      "step": 24730
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.371266427718042e-06,
      "loss": 1.5729,
      "step": 24740
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.352598566308245e-06,
      "loss": 1.5295,
      "step": 24750
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.333930704898446e-06,
      "loss": 1.5199,
      "step": 24760
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.315262843488651e-06,
      "loss": 1.2237,
      "step": 24770
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.296594982078853e-06,
      "loss": 1.5316,
      "step": 24780
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.277927120669058e-06,
      "loss": 1.1651,
      "step": 24790
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.259259259259259e-06,
      "loss": 0.9255,
      "step": 24800
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.240591397849462e-06,
      "loss": 1.4782,
      "step": 24810
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.221923536439666e-06,
      "loss": 1.4225,
      "step": 24820
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.203255675029869e-06,
      "loss": 1.3602,
      "step": 24830
    },
    {
      "epoch": 4.03,
      "learning_rate": 9.184587813620072e-06,
      "loss": 1.4976,
      "step": 24840
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.165919952210275e-06,
      "loss": 1.7712,
      "step": 24850
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.147252090800478e-06,
      "loss": 1.3732,
      "step": 24860
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.128584229390682e-06,
      "loss": 1.6067,
      "step": 24870
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.109916367980885e-06,
      "loss": 2.2038,
      "step": 24880
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.091248506571088e-06,
      "loss": 1.4748,
      "step": 24890
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.072580645161291e-06,
      "loss": 1.561,
      "step": 24900
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.053912783751493e-06,
      "loss": 1.0526,
      "step": 24910
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.035244922341698e-06,
      "loss": 1.3474,
      "step": 24920
    },
    {
      "epoch": 4.04,
      "learning_rate": 9.0165770609319e-06,
      "loss": 1.2918,
      "step": 24930
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.997909199522103e-06,
      "loss": 1.3391,
      "step": 24940
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.979241338112307e-06,
      "loss": 1.753,
      "step": 24950
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.960573476702509e-06,
      "loss": 2.1419,
      "step": 24960
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.941905615292714e-06,
      "loss": 1.6029,
      "step": 24970
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.923237753882915e-06,
      "loss": 1.6535,
      "step": 24980
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.904569892473119e-06,
      "loss": 1.1886,
      "step": 24990
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.885902031063322e-06,
      "loss": 1.4978,
      "step": 25000
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.867234169653525e-06,
      "loss": 1.8418,
      "step": 25010
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.848566308243728e-06,
      "loss": 1.1799,
      "step": 25020
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.829898446833931e-06,
      "loss": 1.7685,
      "step": 25030
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.811230585424133e-06,
      "loss": 1.6405,
      "step": 25040
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.792562724014338e-06,
      "loss": 0.8644,
      "step": 25050
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.773894862604541e-06,
      "loss": 1.4913,
      "step": 25060
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.755227001194743e-06,
      "loss": 1.3354,
      "step": 25070
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.736559139784948e-06,
      "loss": 1.2122,
      "step": 25080
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.717891278375149e-06,
      "loss": 2.0048,
      "step": 25090
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.699223416965354e-06,
      "loss": 0.9122,
      "step": 25100
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.680555555555556e-06,
      "loss": 1.2963,
      "step": 25110
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.661887694145759e-06,
      "loss": 1.8554,
      "step": 25120
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.643219832735962e-06,
      "loss": 1.5882,
      "step": 25130
    },
    {
      "epoch": 4.04,
      "learning_rate": 8.624551971326165e-06,
      "loss": 1.2758,
      "step": 25140
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.605884109916368e-06,
      "loss": 1.4045,
      "step": 25150
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.587216248506572e-06,
      "loss": 1.4268,
      "step": 25160
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.568548387096773e-06,
      "loss": 1.6635,
      "step": 25170
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.549880525686978e-06,
      "loss": 1.1848,
      "step": 25180
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.531212664277181e-06,
      "loss": 1.7076,
      "step": 25190
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.512544802867385e-06,
      "loss": 1.6546,
      "step": 25200
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.493876941457588e-06,
      "loss": 1.4574,
      "step": 25210
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.47520908004779e-06,
      "loss": 1.4794,
      "step": 25220
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.456541218637994e-06,
      "loss": 1.8868,
      "step": 25230
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.437873357228196e-06,
      "loss": 1.1603,
      "step": 25240
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.419205495818399e-06,
      "loss": 2.0306,
      "step": 25250
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.400537634408602e-06,
      "loss": 1.2349,
      "step": 25260
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.381869772998805e-06,
      "loss": 0.9573,
      "step": 25270
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.363201911589009e-06,
      "loss": 1.6363,
      "step": 25280
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.344534050179212e-06,
      "loss": 1.1422,
      "step": 25290
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.325866188769415e-06,
      "loss": 0.9977,
      "step": 25300
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.307198327359618e-06,
      "loss": 1.7289,
      "step": 25310
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.288530465949821e-06,
      "loss": 1.3954,
      "step": 25320
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.269862604540025e-06,
      "loss": 1.71,
      "step": 25330
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.251194743130228e-06,
      "loss": 1.3419,
      "step": 25340
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.23252688172043e-06,
      "loss": 1.4473,
      "step": 25350
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.213859020310634e-06,
      "loss": 1.5981,
      "step": 25360
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.195191158900836e-06,
      "loss": 1.2856,
      "step": 25370
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.17652329749104e-06,
      "loss": 1.7785,
      "step": 25380
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.157855436081242e-06,
      "loss": 0.7809,
      "step": 25390
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.139187574671446e-06,
      "loss": 1.7897,
      "step": 25400
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.12051971326165e-06,
      "loss": 1.6203,
      "step": 25410
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.101851851851852e-06,
      "loss": 1.769,
      "step": 25420
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.083183990442055e-06,
      "loss": 1.7195,
      "step": 25430
    },
    {
      "epoch": 4.05,
      "learning_rate": 8.064516129032258e-06,
      "loss": 0.6735,
      "step": 25440
    },
    {
      "epoch": 4.06,
      "learning_rate": 8.045848267622462e-06,
      "loss": 1.1943,
      "step": 25450
    },
    {
      "epoch": 4.06,
      "learning_rate": 8.027180406212665e-06,
      "loss": 1.8344,
      "step": 25460
    },
    {
      "epoch": 4.06,
      "learning_rate": 8.008512544802868e-06,
      "loss": 1.5848,
      "step": 25470
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.98984468339307e-06,
      "loss": 1.4308,
      "step": 25480
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.971176821983274e-06,
      "loss": 1.7607,
      "step": 25490
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.952508960573476e-06,
      "loss": 1.5066,
      "step": 25500
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.933841099163681e-06,
      "loss": 1.4067,
      "step": 25510
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.915173237753884e-06,
      "loss": 1.1646,
      "step": 25520
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.896505376344086e-06,
      "loss": 1.4122,
      "step": 25530
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.87783751493429e-06,
      "loss": 1.4635,
      "step": 25540
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.859169653524492e-06,
      "loss": 1.8227,
      "step": 25550
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.840501792114695e-06,
      "loss": 1.8722,
      "step": 25560
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.821833930704899e-06,
      "loss": 1.0591,
      "step": 25570
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.803166069295102e-06,
      "loss": 1.4113,
      "step": 25580
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.784498207885305e-06,
      "loss": 1.2042,
      "step": 25590
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.765830346475508e-06,
      "loss": 1.7254,
      "step": 25600
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.747162485065711e-06,
      "loss": 1.7568,
      "step": 25610
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.728494623655915e-06,
      "loss": 1.0329,
      "step": 25620
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.709826762246116e-06,
      "loss": 1.2179,
      "step": 25630
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.691158900836321e-06,
      "loss": 1.3515,
      "step": 25640
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.672491039426524e-06,
      "loss": 2.3152,
      "step": 25650
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.653823178016726e-06,
      "loss": 1.1229,
      "step": 25660
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.63515531660693e-06,
      "loss": 1.4869,
      "step": 25670
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.616487455197132e-06,
      "loss": 1.5393,
      "step": 25680
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.597819593787336e-06,
      "loss": 1.8928,
      "step": 25690
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.579151732377539e-06,
      "loss": 1.2179,
      "step": 25700
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.560483870967743e-06,
      "loss": 1.1297,
      "step": 25710
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.541816009557945e-06,
      "loss": 1.7601,
      "step": 25720
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.523148148148148e-06,
      "loss": 1.5369,
      "step": 25730
    },
    {
      "epoch": 4.06,
      "learning_rate": 7.504480286738351e-06,
      "loss": 1.0863,
      "step": 25740
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.485812425328555e-06,
      "loss": 1.9207,
      "step": 25750
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.467144563918758e-06,
      "loss": 2.0301,
      "step": 25760
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.44847670250896e-06,
      "loss": 0.9577,
      "step": 25770
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.4298088410991645e-06,
      "loss": 1.0539,
      "step": 25780
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.411140979689367e-06,
      "loss": 1.7959,
      "step": 25790
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.392473118279571e-06,
      "loss": 0.97,
      "step": 25800
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.373805256869773e-06,
      "loss": 1.7899,
      "step": 25810
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.3551373954599765e-06,
      "loss": 1.5984,
      "step": 25820
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.336469534050179e-06,
      "loss": 1.5833,
      "step": 25830
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.317801672640383e-06,
      "loss": 1.4231,
      "step": 25840
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.299133811230585e-06,
      "loss": 1.0482,
      "step": 25850
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.2804659498207885e-06,
      "loss": 1.6302,
      "step": 25860
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.261798088410993e-06,
      "loss": 1.7256,
      "step": 25870
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.243130227001195e-06,
      "loss": 1.4821,
      "step": 25880
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.224462365591399e-06,
      "loss": 1.3046,
      "step": 25890
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.205794504181601e-06,
      "loss": 1.7332,
      "step": 25900
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.187126642771805e-06,
      "loss": 2.1257,
      "step": 25910
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.168458781362007e-06,
      "loss": 1.3405,
      "step": 25920
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.149790919952211e-06,
      "loss": 1.0742,
      "step": 25930
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.1311230585424134e-06,
      "loss": 1.6651,
      "step": 25940
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.112455197132617e-06,
      "loss": 1.8454,
      "step": 25950
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.093787335722819e-06,
      "loss": 1.6758,
      "step": 25960
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.075119474313023e-06,
      "loss": 1.3766,
      "step": 25970
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.056451612903227e-06,
      "loss": 1.4214,
      "step": 25980
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.0377837514934295e-06,
      "loss": 1.1765,
      "step": 25990
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.019115890083633e-06,
      "loss": 1.2647,
      "step": 26000
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.000448028673835e-06,
      "loss": 1.9995,
      "step": 26010
    },
    {
      "epoch": 4.07,
      "learning_rate": 6.981780167264039e-06,
      "loss": 1.6961,
      "step": 26020
    },
    {
      "epoch": 4.07,
      "learning_rate": 6.9631123058542416e-06,
      "loss": 1.664,
      "step": 26030
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.944444444444445e-06,
      "loss": 1.2451,
      "step": 26040
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.925776583034647e-06,
      "loss": 1.3411,
      "step": 26050
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.907108721624851e-06,
      "loss": 1.7018,
      "step": 26060
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.888440860215054e-06,
      "loss": 0.9882,
      "step": 26070
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.869772998805258e-06,
      "loss": 1.1228,
      "step": 26080
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.85110513739546e-06,
      "loss": 1.8301,
      "step": 26090
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.832437275985663e-06,
      "loss": 1.1561,
      "step": 26100
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.813769414575867e-06,
      "loss": 1.4249,
      "step": 26110
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.79510155316607e-06,
      "loss": 1.3765,
      "step": 26120
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.776433691756273e-06,
      "loss": 1.2172,
      "step": 26130
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.757765830346475e-06,
      "loss": 1.0942,
      "step": 26140
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.739097968936679e-06,
      "loss": 1.2836,
      "step": 26150
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.720430107526882e-06,
      "loss": 1.0587,
      "step": 26160
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.701762246117086e-06,
      "loss": 1.5013,
      "step": 26170
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.683094384707287e-06,
      "loss": 1.2316,
      "step": 26180
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.664426523297491e-06,
      "loss": 1.7358,
      "step": 26190
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.645758661887694e-06,
      "loss": 1.0172,
      "step": 26200
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.627090800477898e-06,
      "loss": 1.1399,
      "step": 26210
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.608422939068101e-06,
      "loss": 1.7148,
      "step": 26220
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.589755077658303e-06,
      "loss": 1.5614,
      "step": 26230
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.5710872162485075e-06,
      "loss": 1.435,
      "step": 26240
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.55241935483871e-06,
      "loss": 1.0931,
      "step": 26250
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.533751493428913e-06,
      "loss": 1.3666,
      "step": 26260
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.5150836320191154e-06,
      "loss": 1.2352,
      "step": 26270
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.4964157706093195e-06,
      "loss": 1.1093,
      "step": 26280
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.477747909199522e-06,
      "loss": 1.3381,
      "step": 26290
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.459080047789726e-06,
      "loss": 1.0557,
      "step": 26300
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.440412186379928e-06,
      "loss": 1.5795,
      "step": 26310
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.4217443249701315e-06,
      "loss": 1.3957,
      "step": 26320
    },
    {
      "epoch": 4.08,
      "learning_rate": 6.403076463560336e-06,
      "loss": 1.4447,
      "step": 26330
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.384408602150538e-06,
      "loss": 1.5413,
      "step": 26340
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.365740740740741e-06,
      "loss": 0.8564,
      "step": 26350
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.3470728793309436e-06,
      "loss": 1.5251,
      "step": 26360
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.328405017921148e-06,
      "loss": 1.4446,
      "step": 26370
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.30973715651135e-06,
      "loss": 1.0587,
      "step": 26380
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.291069295101554e-06,
      "loss": 1.7043,
      "step": 26390
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.2724014336917564e-06,
      "loss": 1.323,
      "step": 26400
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.25373357228196e-06,
      "loss": 1.2718,
      "step": 26410
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.235065710872163e-06,
      "loss": 1.6291,
      "step": 26420
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.216397849462366e-06,
      "loss": 1.0402,
      "step": 26430
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.197729988052569e-06,
      "loss": 1.1328,
      "step": 26440
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.179062126642772e-06,
      "loss": 1.192,
      "step": 26450
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.160394265232975e-06,
      "loss": 0.8817,
      "step": 26460
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.141726403823178e-06,
      "loss": 0.8596,
      "step": 26470
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.123058542413381e-06,
      "loss": 1.6592,
      "step": 26480
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.1043906810035846e-06,
      "loss": 1.9876,
      "step": 26490
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.085722819593788e-06,
      "loss": 1.515,
      "step": 26500
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.067054958183991e-06,
      "loss": 1.2755,
      "step": 26510
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.048387096774194e-06,
      "loss": 1.4755,
      "step": 26520
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.0297192353643974e-06,
      "loss": 0.957,
      "step": 26530
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.0110513739546e-06,
      "loss": 0.7414,
      "step": 26540
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.992383512544803e-06,
      "loss": 0.8125,
      "step": 26550
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.973715651135006e-06,
      "loss": 1.7498,
      "step": 26560
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.9550477897252095e-06,
      "loss": 1.2346,
      "step": 26570
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.936379928315413e-06,
      "loss": 1.1029,
      "step": 26580
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.917712066905615e-06,
      "loss": 1.4347,
      "step": 26590
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.899044205495818e-06,
      "loss": 1.255,
      "step": 26600
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.880376344086022e-06,
      "loss": 1.24,
      "step": 26610
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.8617084826762256e-06,
      "loss": 0.944,
      "step": 26620
    },
    {
      "epoch": 4.09,
      "learning_rate": 5.843040621266428e-06,
      "loss": 0.7633,
      "step": 26630
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.824372759856631e-06,
      "loss": 1.4528,
      "step": 26640
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.805704898446834e-06,
      "loss": 1.6512,
      "step": 26650
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.787037037037038e-06,
      "loss": 1.4074,
      "step": 26660
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.76836917562724e-06,
      "loss": 1.371,
      "step": 26670
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.749701314217443e-06,
      "loss": 1.2035,
      "step": 26680
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.731033452807646e-06,
      "loss": 1.0008,
      "step": 26690
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.71236559139785e-06,
      "loss": 1.2706,
      "step": 26700
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.693697729988053e-06,
      "loss": 1.4189,
      "step": 26710
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.675029868578255e-06,
      "loss": 1.8266,
      "step": 26720
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.656362007168459e-06,
      "loss": 1.08,
      "step": 26730
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.6376941457586625e-06,
      "loss": 1.5767,
      "step": 26740
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.619026284348866e-06,
      "loss": 1.2904,
      "step": 26750
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.600358422939068e-06,
      "loss": 1.2904,
      "step": 26760
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.581690561529271e-06,
      "loss": 1.4163,
      "step": 26770
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.5630227001194745e-06,
      "loss": 1.212,
      "step": 26780
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.544354838709678e-06,
      "loss": 1.3449,
      "step": 26790
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.525686977299881e-06,
      "loss": 1.3649,
      "step": 26800
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.507019115890083e-06,
      "loss": 1.4084,
      "step": 26810
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.4883512544802866e-06,
      "loss": 1.4619,
      "step": 26820
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.46968339307049e-06,
      "loss": 1.2643,
      "step": 26830
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.451015531660694e-06,
      "loss": 1.0674,
      "step": 26840
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.432347670250896e-06,
      "loss": 1.2631,
      "step": 26850
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.4136798088410994e-06,
      "loss": 1.1677,
      "step": 26860
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.395011947431303e-06,
      "loss": 1.6988,
      "step": 26870
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.376344086021506e-06,
      "loss": 1.7663,
      "step": 26880
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.357676224611709e-06,
      "loss": 1.4654,
      "step": 26890
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.3390083632019115e-06,
      "loss": 1.0302,
      "step": 26900
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.320340501792115e-06,
      "loss": 1.2625,
      "step": 26910
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.301672640382318e-06,
      "loss": 1.3896,
      "step": 26920
    },
    {
      "epoch": 4.1,
      "learning_rate": 5.283004778972521e-06,
      "loss": 1.2841,
      "step": 26930
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.264336917562724e-06,
      "loss": 1.6893,
      "step": 26940
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.245669056152927e-06,
      "loss": 0.7325,
      "step": 26950
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.227001194743131e-06,
      "loss": 1.4401,
      "step": 26960
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.208333333333334e-06,
      "loss": 0.9587,
      "step": 26970
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.189665471923537e-06,
      "loss": 1.2609,
      "step": 26980
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.17099761051374e-06,
      "loss": 1.6398,
      "step": 26990
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.152329749103943e-06,
      "loss": 1.1614,
      "step": 27000
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.133661887694146e-06,
      "loss": 2.1873,
      "step": 27010
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.114994026284349e-06,
      "loss": 1.1867,
      "step": 27020
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.0963261648745525e-06,
      "loss": 1.0342,
      "step": 27030
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.077658303464755e-06,
      "loss": 1.6497,
      "step": 27040
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.058990442054958e-06,
      "loss": 1.2071,
      "step": 27050
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.040322580645161e-06,
      "loss": 1.4087,
      "step": 27060
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.021654719235365e-06,
      "loss": 1.168,
      "step": 27070
    },
    {
      "epoch": 4.11,
      "learning_rate": 5.002986857825568e-06,
      "loss": 1.5651,
      "step": 27080
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.984318996415771e-06,
      "loss": 1.1643,
      "step": 27090
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.965651135005974e-06,
      "loss": 1.0717,
      "step": 27100
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.946983273596177e-06,
      "loss": 0.9171,
      "step": 27110
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.928315412186381e-06,
      "loss": 1.1478,
      "step": 27120
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.909647550776583e-06,
      "loss": 0.8618,
      "step": 27130
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.890979689366786e-06,
      "loss": 1.0755,
      "step": 27140
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.872311827956989e-06,
      "loss": 1.2589,
      "step": 27150
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.853643966547193e-06,
      "loss": 1.2669,
      "step": 27160
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.834976105137395e-06,
      "loss": 0.803,
      "step": 27170
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.816308243727598e-06,
      "loss": 1.3754,
      "step": 27180
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.797640382317802e-06,
      "loss": 1.0802,
      "step": 27190
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.7789725209080055e-06,
      "loss": 1.5899,
      "step": 27200
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.760304659498208e-06,
      "loss": 1.4347,
      "step": 27210
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.741636798088411e-06,
      "loss": 1.2137,
      "step": 27220
    },
    {
      "epoch": 4.11,
      "learning_rate": 4.722968936678614e-06,
      "loss": 0.9562,
      "step": 27230
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.7043010752688175e-06,
      "loss": 1.9109,
      "step": 27240
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.685633213859021e-06,
      "loss": 1.3119,
      "step": 27250
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.666965352449223e-06,
      "loss": 0.8017,
      "step": 27260
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.648297491039426e-06,
      "loss": 0.9947,
      "step": 27270
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.6296296296296296e-06,
      "loss": 0.8371,
      "step": 27280
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.610961768219833e-06,
      "loss": 1.1533,
      "step": 27290
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.592293906810036e-06,
      "loss": 1.4851,
      "step": 27300
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.573626045400239e-06,
      "loss": 1.5737,
      "step": 27310
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.5549581839904424e-06,
      "loss": 1.1396,
      "step": 27320
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.536290322580646e-06,
      "loss": 1.1578,
      "step": 27330
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.517622461170849e-06,
      "loss": 1.097,
      "step": 27340
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.498954599761051e-06,
      "loss": 1.5931,
      "step": 27350
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.4802867383512545e-06,
      "loss": 1.1666,
      "step": 27360
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.461618876941458e-06,
      "loss": 0.9427,
      "step": 27370
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.442951015531661e-06,
      "loss": 1.8824,
      "step": 27380
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.424283154121864e-06,
      "loss": 0.8832,
      "step": 27390
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.4056152927120665e-06,
      "loss": 1.2699,
      "step": 27400
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.3869474313022706e-06,
      "loss": 1.0975,
      "step": 27410
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.368279569892474e-06,
      "loss": 1.2642,
      "step": 27420
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.349611708482677e-06,
      "loss": 1.5067,
      "step": 27430
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.330943847072879e-06,
      "loss": 1.6365,
      "step": 27440
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.312275985663083e-06,
      "loss": 1.2913,
      "step": 27450
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.293608124253286e-06,
      "loss": 0.7613,
      "step": 27460
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.274940262843489e-06,
      "loss": 1.0755,
      "step": 27470
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.256272401433692e-06,
      "loss": 1.1365,
      "step": 27480
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.237604540023895e-06,
      "loss": 0.9249,
      "step": 27490
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.218936678614098e-06,
      "loss": 1.2179,
      "step": 27500
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.200268817204301e-06,
      "loss": 1.58,
      "step": 27510
    },
    {
      "epoch": 4.12,
      "learning_rate": 4.181600955794504e-06,
      "loss": 1.2847,
      "step": 27520
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.1629330943847075e-06,
      "loss": 1.3193,
      "step": 27530
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.144265232974911e-06,
      "loss": 1.4941,
      "step": 27540
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.125597371565114e-06,
      "loss": 1.3101,
      "step": 27550
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.106929510155317e-06,
      "loss": 1.3672,
      "step": 27560
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.08826164874552e-06,
      "loss": 1.3266,
      "step": 27570
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.069593787335723e-06,
      "loss": 1.1635,
      "step": 27580
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.050925925925926e-06,
      "loss": 1.0761,
      "step": 27590
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.032258064516129e-06,
      "loss": 1.5363,
      "step": 27600
    },
    {
      "epoch": 4.13,
      "learning_rate": 4.013590203106332e-06,
      "loss": 1.1535,
      "step": 27610
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.994922341696535e-06,
      "loss": 1.0786,
      "step": 27620
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.976254480286738e-06,
      "loss": 1.505,
      "step": 27630
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.957586618876942e-06,
      "loss": 1.0203,
      "step": 27640
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.938918757467145e-06,
      "loss": 1.2253,
      "step": 27650
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.920250896057348e-06,
      "loss": 1.0077,
      "step": 27660
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.901583034647551e-06,
      "loss": 1.4607,
      "step": 27670
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.882915173237754e-06,
      "loss": 0.9503,
      "step": 27680
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.864247311827957e-06,
      "loss": 1.5217,
      "step": 27690
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.8455794504181605e-06,
      "loss": 0.8979,
      "step": 27700
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.826911589008363e-06,
      "loss": 1.3058,
      "step": 27710
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.808243727598566e-06,
      "loss": 1.4802,
      "step": 27720
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.7895758661887694e-06,
      "loss": 1.8119,
      "step": 27730
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.7709080047789726e-06,
      "loss": 1.1756,
      "step": 27740
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.7522401433691754e-06,
      "loss": 0.8481,
      "step": 27750
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.733572281959379e-06,
      "loss": 0.7704,
      "step": 27760
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.7149044205495822e-06,
      "loss": 1.2153,
      "step": 27770
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.6962365591397855e-06,
      "loss": 1.7107,
      "step": 27780
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.6775686977299882e-06,
      "loss": 0.9795,
      "step": 27790
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.6589008363201915e-06,
      "loss": 0.9315,
      "step": 27800
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.6402329749103943e-06,
      "loss": 0.9131,
      "step": 27810
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.6215651135005975e-06,
      "loss": 1.089,
      "step": 27820
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.6028972520908007e-06,
      "loss": 0.9854,
      "step": 27830
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.5842293906810035e-06,
      "loss": 1.6351,
      "step": 27840
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.5655615292712067e-06,
      "loss": 1.2948,
      "step": 27850
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.5468936678614095e-06,
      "loss": 0.6349,
      "step": 27860
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.5282258064516136e-06,
      "loss": 1.1524,
      "step": 27870
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.5095579450418164e-06,
      "loss": 1.1165,
      "step": 27880
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.4908900836320196e-06,
      "loss": 1.3693,
      "step": 27890
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.4722222222222224e-06,
      "loss": 0.8425,
      "step": 27900
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.4535543608124256e-06,
      "loss": 1.3458,
      "step": 27910
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.434886499402629e-06,
      "loss": 1.0211,
      "step": 27920
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.4162186379928316e-06,
      "loss": 1.171,
      "step": 27930
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.397550776583035e-06,
      "loss": 0.983,
      "step": 27940
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.3788829151732376e-06,
      "loss": 1.3137,
      "step": 27950
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.360215053763441e-06,
      "loss": 1.1717,
      "step": 27960
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.3415471923536437e-06,
      "loss": 1.2904,
      "step": 27970
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.322879330943847e-06,
      "loss": 1.6562,
      "step": 27980
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.3042114695340505e-06,
      "loss": 1.042,
      "step": 27990
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.2855436081242537e-06,
      "loss": 1.0269,
      "step": 28000
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.2668757467144565e-06,
      "loss": 0.7511,
      "step": 28010
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.2482078853046597e-06,
      "loss": 1.4292,
      "step": 28020
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.229540023894863e-06,
      "loss": 1.2028,
      "step": 28030
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.2108721624850658e-06,
      "loss": 1.3415,
      "step": 28040
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.192204301075269e-06,
      "loss": 1.2112,
      "step": 28050
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.1735364396654718e-06,
      "loss": 1.0659,
      "step": 28060
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.154868578255675e-06,
      "loss": 1.0592,
      "step": 28070
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.1362007168458782e-06,
      "loss": 1.5721,
      "step": 28080
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.1175328554360814e-06,
      "loss": 1.2488,
      "step": 28090
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.0988649940262847e-06,
      "loss": 1.2523,
      "step": 28100
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.0801971326164875e-06,
      "loss": 2.0374,
      "step": 28110
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.0615292712066907e-06,
      "loss": 1.4842,
      "step": 28120
    },
    {
      "epoch": 4.15,
      "learning_rate": 3.042861409796894e-06,
      "loss": 0.9243,
      "step": 28130
    },
    {
      "epoch": 4.15,
      "learning_rate": 3.024193548387097e-06,
      "loss": 2.0135,
      "step": 28140
    },
    {
      "epoch": 4.15,
      "learning_rate": 3.0055256869773e-06,
      "loss": 1.0956,
      "step": 28150
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.986857825567503e-06,
      "loss": 1.3584,
      "step": 28160
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.9681899641577063e-06,
      "loss": 1.3753,
      "step": 28170
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.949522102747909e-06,
      "loss": 0.6571,
      "step": 28180
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.9308542413381128e-06,
      "loss": 1.5899,
      "step": 28190
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.9121863799283156e-06,
      "loss": 1.4262,
      "step": 28200
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.893518518518519e-06,
      "loss": 0.8002,
      "step": 28210
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.8748506571087216e-06,
      "loss": 1.5667,
      "step": 28220
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.856182795698925e-06,
      "loss": 1.3142,
      "step": 28230
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.8375149342891276e-06,
      "loss": 1.4252,
      "step": 28240
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.8188470728793313e-06,
      "loss": 0.8309,
      "step": 28250
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.800179211469534e-06,
      "loss": 1.2724,
      "step": 28260
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.7815113500597373e-06,
      "loss": 1.3443,
      "step": 28270
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.7628434886499405e-06,
      "loss": 0.7029,
      "step": 28280
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.7441756272401433e-06,
      "loss": 0.896,
      "step": 28290
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.725507765830347e-06,
      "loss": 1.3636,
      "step": 28300
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.7068399044205497e-06,
      "loss": 1.1022,
      "step": 28310
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.688172043010753e-06,
      "loss": 0.9287,
      "step": 28320
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.6695041816009557e-06,
      "loss": 0.9124,
      "step": 28330
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.650836320191159e-06,
      "loss": 1.3033,
      "step": 28340
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.632168458781362e-06,
      "loss": 0.9654,
      "step": 28350
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.6135005973715654e-06,
      "loss": 0.7993,
      "step": 28360
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.5948327359617686e-06,
      "loss": 1.0669,
      "step": 28370
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.5761648745519714e-06,
      "loss": 0.9279,
      "step": 28380
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.5574970131421746e-06,
      "loss": 0.8121,
      "step": 28390
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.5388291517323774e-06,
      "loss": 1.3864,
      "step": 28400
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.5201612903225806e-06,
      "loss": 1.1031,
      "step": 28410
    },
    {
      "epoch": 4.15,
      "learning_rate": 2.501493428912784e-06,
      "loss": 1.757,
      "step": 28420
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.482825567502987e-06,
      "loss": 1.3906,
      "step": 28430
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.4641577060931903e-06,
      "loss": 1.6209,
      "step": 28440
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.445489844683393e-06,
      "loss": 0.8543,
      "step": 28450
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.4268219832735963e-06,
      "loss": 1.7187,
      "step": 28460
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.408154121863799e-06,
      "loss": 1.11,
      "step": 28470
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.3894862604540028e-06,
      "loss": 0.8351,
      "step": 28480
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.3708183990442056e-06,
      "loss": 1.3873,
      "step": 28490
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.3521505376344088e-06,
      "loss": 1.4568,
      "step": 28500
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.3334826762246116e-06,
      "loss": 1.1747,
      "step": 28510
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.3148148148148148e-06,
      "loss": 0.8766,
      "step": 28520
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.296146953405018e-06,
      "loss": 1.2374,
      "step": 28530
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.2774790919952212e-06,
      "loss": 1.0945,
      "step": 28540
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.2588112305854244e-06,
      "loss": 1.3041,
      "step": 28550
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.2401433691756272e-06,
      "loss": 0.8916,
      "step": 28560
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.2214755077658305e-06,
      "loss": 1.536,
      "step": 28570
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.2028076463560333e-06,
      "loss": 1.2402,
      "step": 28580
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.184139784946237e-06,
      "loss": 1.9305,
      "step": 28590
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.1654719235364397e-06,
      "loss": 1.0564,
      "step": 28600
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.146804062126643e-06,
      "loss": 0.7555,
      "step": 28610
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.128136200716846e-06,
      "loss": 0.9128,
      "step": 28620
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.109468339307049e-06,
      "loss": 1.0006,
      "step": 28630
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.090800477897252e-06,
      "loss": 1.155,
      "step": 28640
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.0721326164874554e-06,
      "loss": 0.6499,
      "step": 28650
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.0534647550776586e-06,
      "loss": 1.4456,
      "step": 28660
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.0347968936678614e-06,
      "loss": 1.2338,
      "step": 28670
    },
    {
      "epoch": 4.16,
      "learning_rate": 2.0161290322580646e-06,
      "loss": 1.381,
      "step": 28680
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.9974611708482674e-06,
      "loss": 1.1661,
      "step": 28690
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.978793309438471e-06,
      "loss": 1.2045,
      "step": 28700
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.960125448028674e-06,
      "loss": 1.6113,
      "step": 28710
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.941457586618877e-06,
      "loss": 1.5231,
      "step": 28720
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.9227897252090803e-06,
      "loss": 0.5758,
      "step": 28730
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.904121863799283e-06,
      "loss": 1.9556,
      "step": 28740
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.8854540023894863e-06,
      "loss": 1.4313,
      "step": 28750
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.8667861409796895e-06,
      "loss": 0.9415,
      "step": 28760
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.8481182795698927e-06,
      "loss": 1.3642,
      "step": 28770
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.8294504181600957e-06,
      "loss": 1.4861,
      "step": 28780
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.8107825567502987e-06,
      "loss": 1.6615,
      "step": 28790
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.7921146953405017e-06,
      "loss": 1.053,
      "step": 28800
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.7734468339307048e-06,
      "loss": 1.4023,
      "step": 28810
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.7547789725209082e-06,
      "loss": 1.3273,
      "step": 28820
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.7361111111111112e-06,
      "loss": 1.3633,
      "step": 28830
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.7174432497013144e-06,
      "loss": 0.8788,
      "step": 28840
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.6987753882915174e-06,
      "loss": 1.1909,
      "step": 28850
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.6801075268817204e-06,
      "loss": 1.6415,
      "step": 28860
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.6614396654719234e-06,
      "loss": 1.0557,
      "step": 28870
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.6427718040621269e-06,
      "loss": 1.4031,
      "step": 28880
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.6241039426523299e-06,
      "loss": 1.3578,
      "step": 28890
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.6054360812425329e-06,
      "loss": 1.0865,
      "step": 28900
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.5867682198327359e-06,
      "loss": 1.6102,
      "step": 28910
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.5681003584229391e-06,
      "loss": 1.3342,
      "step": 28920
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.5494324970131423e-06,
      "loss": 1.4535,
      "step": 28930
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.5307646356033453e-06,
      "loss": 1.2964,
      "step": 28940
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.5120967741935486e-06,
      "loss": 1.024,
      "step": 28950
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.4934289127837516e-06,
      "loss": 1.2547,
      "step": 28960
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.4747610513739546e-06,
      "loss": 0.9464,
      "step": 28970
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.4560931899641578e-06,
      "loss": 1.4012,
      "step": 28980
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.4374253285543608e-06,
      "loss": 1.0043,
      "step": 28990
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.4187574671445638e-06,
      "loss": 1.5499,
      "step": 29000
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.400089605734767e-06,
      "loss": 1.259,
      "step": 29010
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.3814217443249702e-06,
      "loss": 1.0037,
      "step": 29020
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.3627538829151735e-06,
      "loss": 1.479,
      "step": 29030
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.3440860215053765e-06,
      "loss": 0.5973,
      "step": 29040
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.3254181600955795e-06,
      "loss": 1.807,
      "step": 29050
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.3067502986857827e-06,
      "loss": 1.0697,
      "step": 29060
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.2880824372759857e-06,
      "loss": 0.6813,
      "step": 29070
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.2694145758661887e-06,
      "loss": 1.3963,
      "step": 29080
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.250746714456392e-06,
      "loss": 1.3437,
      "step": 29090
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.2320788530465952e-06,
      "loss": 1.2347,
      "step": 29100
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.2134109916367982e-06,
      "loss": 0.8104,
      "step": 29110
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1947431302270014e-06,
      "loss": 0.9531,
      "step": 29120
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1760752688172044e-06,
      "loss": 1.2772,
      "step": 29130
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1574074074074074e-06,
      "loss": 1.1648,
      "step": 29140
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1387395459976106e-06,
      "loss": 0.8311,
      "step": 29150
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1200716845878136e-06,
      "loss": 0.5579,
      "step": 29160
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.1014038231780166e-06,
      "loss": 1.3376,
      "step": 29170
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.0827359617682198e-06,
      "loss": 0.3455,
      "step": 29180
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.064068100358423e-06,
      "loss": 1.3899,
      "step": 29190
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.045400238948626e-06,
      "loss": 1.1563,
      "step": 29200
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.0267323775388293e-06,
      "loss": 0.8992,
      "step": 29210
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.0080645161290323e-06,
      "loss": 1.2932,
      "step": 29220
    },
    {
      "epoch": 4.18,
      "learning_rate": 9.893966547192355e-07,
      "loss": 1.4952,
      "step": 29230
    },
    {
      "epoch": 4.18,
      "learning_rate": 9.707287933094385e-07,
      "loss": 0.8295,
      "step": 29240
    },
    {
      "epoch": 4.18,
      "learning_rate": 9.520609318996415e-07,
      "loss": 1.2467,
      "step": 29250
    },
    {
      "epoch": 4.18,
      "learning_rate": 9.333930704898448e-07,
      "loss": 0.6587,
      "step": 29260
    },
    {
      "epoch": 4.18,
      "learning_rate": 9.147252090800479e-07,
      "loss": 0.9228,
      "step": 29270
    },
    {
      "epoch": 4.18,
      "learning_rate": 8.960573476702509e-07,
      "loss": 0.8199,
      "step": 29280
    },
    {
      "epoch": 4.18,
      "learning_rate": 8.773894862604541e-07,
      "loss": 1.422,
      "step": 29290
    },
    {
      "epoch": 4.18,
      "learning_rate": 8.587216248506572e-07,
      "loss": 1.3199,
      "step": 29300
    },
    {
      "epoch": 4.18,
      "learning_rate": 8.400537634408602e-07,
      "loss": 1.1563,
      "step": 29310
    },
    {
      "epoch": 4.19,
      "learning_rate": 8.213859020310634e-07,
      "loss": 1.3217,
      "step": 29320
    },
    {
      "epoch": 4.19,
      "learning_rate": 8.027180406212664e-07,
      "loss": 1.5251,
      "step": 29330
    },
    {
      "epoch": 4.19,
      "learning_rate": 7.840501792114696e-07,
      "loss": 0.9194,
      "step": 29340
    },
    {
      "epoch": 4.19,
      "learning_rate": 7.653823178016727e-07,
      "loss": 0.9451,
      "step": 29350
    },
    {
      "epoch": 4.19,
      "learning_rate": 7.467144563918758e-07,
      "loss": 1.4432,
      "step": 29360
    },
    {
      "epoch": 4.19,
      "learning_rate": 7.280465949820789e-07,
      "loss": 1.1293,
      "step": 29370
    },
    {
      "epoch": 4.19,
      "learning_rate": 7.093787335722819e-07,
      "loss": 1.1054,
      "step": 29380
    },
    {
      "epoch": 4.19,
      "learning_rate": 6.907108721624851e-07,
      "loss": 1.1609,
      "step": 29390
    },
    {
      "epoch": 4.19,
      "learning_rate": 6.720430107526882e-07,
      "loss": 1.0319,
      "step": 29400
    },
    {
      "epoch": 4.19,
      "learning_rate": 6.533751493428913e-07,
      "loss": 1.522,
      "step": 29410
    },
    {
      "epoch": 4.19,
      "learning_rate": 6.347072879330944e-07,
      "loss": 1.1448,
      "step": 29420
    },
    {
      "epoch": 4.19,
      "learning_rate": 6.160394265232976e-07,
      "loss": 1.4926,
      "step": 29430
    },
    {
      "epoch": 4.19,
      "learning_rate": 5.973715651135007e-07,
      "loss": 0.7732,
      "step": 29440
    },
    {
      "epoch": 4.19,
      "learning_rate": 5.787037037037037e-07,
      "loss": 1.3816,
      "step": 29450
    },
    {
      "epoch": 4.19,
      "learning_rate": 5.600358422939068e-07,
      "loss": 1.269,
      "step": 29460
    },
    {
      "epoch": 4.19,
      "learning_rate": 5.413679808841099e-07,
      "loss": 1.2545,
      "step": 29470
    },
    {
      "epoch": 4.19,
      "learning_rate": 5.22700119474313e-07,
      "loss": 0.582,
      "step": 29480
    },
    {
      "epoch": 4.19,
      "learning_rate": 5.040322580645161e-07,
      "loss": 1.2124,
      "step": 29490
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.853643966547193e-07,
      "loss": 1.1213,
      "step": 29500
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.666965352449224e-07,
      "loss": 1.1458,
      "step": 29510
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.4802867383512544e-07,
      "loss": 1.2749,
      "step": 29520
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.293608124253286e-07,
      "loss": 1.401,
      "step": 29530
    },
    {
      "epoch": 4.19,
      "learning_rate": 4.106929510155317e-07,
      "loss": 1.4359,
      "step": 29540
    },
    {
      "epoch": 4.19,
      "learning_rate": 3.920250896057348e-07,
      "loss": 1.4529,
      "step": 29550
    },
    {
      "epoch": 4.19,
      "learning_rate": 3.733572281959379e-07,
      "loss": 1.2089,
      "step": 29560
    },
    {
      "epoch": 4.19,
      "learning_rate": 3.5468936678614095e-07,
      "loss": 1.0361,
      "step": 29570
    },
    {
      "epoch": 4.19,
      "learning_rate": 3.360215053763441e-07,
      "loss": 1.4934,
      "step": 29580
    },
    {
      "epoch": 4.19,
      "learning_rate": 3.173536439665472e-07,
      "loss": 0.7505,
      "step": 29590
    },
    {
      "epoch": 4.19,
      "learning_rate": 2.9868578255675034e-07,
      "loss": 0.9472,
      "step": 29600
    },
    {
      "epoch": 4.19,
      "learning_rate": 2.800179211469534e-07,
      "loss": 1.4226,
      "step": 29610
    },
    {
      "epoch": 4.2,
      "learning_rate": 2.613500597371565e-07,
      "loss": 1.1832,
      "step": 29620
    },
    {
      "epoch": 4.2,
      "learning_rate": 2.4268219832735963e-07,
      "loss": 1.026,
      "step": 29630
    },
    {
      "epoch": 4.2,
      "learning_rate": 2.2401433691756272e-07,
      "loss": 1.2389,
      "step": 29640
    },
    {
      "epoch": 4.2,
      "learning_rate": 2.0534647550776586e-07,
      "loss": 1.3421,
      "step": 29650
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.8667861409796895e-07,
      "loss": 1.2626,
      "step": 29660
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.6801075268817206e-07,
      "loss": 0.9274,
      "step": 29670
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.4934289127837517e-07,
      "loss": 1.341,
      "step": 29680
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.3067502986857826e-07,
      "loss": 1.5807,
      "step": 29690
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.1200716845878136e-07,
      "loss": 1.1787,
      "step": 29700
    },
    {
      "epoch": 4.2,
      "learning_rate": 9.333930704898447e-08,
      "loss": 1.0598,
      "step": 29710
    },
    {
      "epoch": 4.2,
      "learning_rate": 7.467144563918759e-08,
      "loss": 0.5426,
      "step": 29720
    },
    {
      "epoch": 4.2,
      "learning_rate": 5.600358422939068e-08,
      "loss": 1.1563,
      "step": 29730
    },
    {
      "epoch": 4.2,
      "learning_rate": 3.733572281959379e-08,
      "loss": 0.8868,
      "step": 29740
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.8667861409796897e-08,
      "loss": 1.2178,
      "step": 29750
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0,
      "loss": 0.6968,
      "step": 29760
    },
    {
      "epoch": 4.2,
      "eval_accuracy": 0.5362423368193292,
      "eval_loss": 1.7134848833084106,
      "eval_runtime": 957.6543,
      "eval_samples_per_second": 2.896,
      "eval_steps_per_second": 1.448,
      "step": 29760
    },
    {
      "epoch": 4.2,
      "step": 29760,
      "total_flos": 7.424645769153479e+19,
      "train_loss": 2.989056267741547,
      "train_runtime": 377379.9366,
      "train_samples_per_second": 0.158,
      "train_steps_per_second": 0.079
    },
    {
      "epoch": 4.2,
      "eval_accuracy": 0.5587426326129666,
      "eval_loss": 1.650649070739746,
      "eval_runtime": 862.3611,
      "eval_samples_per_second": 2.951,
      "eval_steps_per_second": 1.476,
      "step": 29760
    },
    {
      "epoch": 4.2,
      "eval_accuracy": 0.5587426326129666,
      "eval_loss": 1.650649070739746,
      "eval_runtime": 863.5093,
      "eval_samples_per_second": 2.947,
      "eval_steps_per_second": 1.474,
      "step": 29760
    }
  ],
  "logging_steps": 10,
  "max_steps": 29760,
  "num_train_epochs": 9223372036854775807,
  "save_steps": 500,
  "total_flos": 7.424645769153479e+19,
  "trial_name": null,
  "trial_params": null
}