{
  "best_metric": 0.18881116431984588,
  "best_model_checkpoint": "/home/azureuser/workspace/mounts/blobs/blob-deeplm22/v-youmohamed/BUS15100_MB2_RAII_20epoch_notweettokenizer_fp16/checkpoint-22848",
  "epoch": 9.810055940866345,
  "global_step": 22848,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 0.0002,
      "loss": 3.784,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 0.0002,
      "loss": 3.424,
      "step": 10
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0002,
      "loss": 3.3057,
      "step": 15
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0002,
      "loss": 3.2421,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0002,
      "loss": 3.2215,
      "step": 25
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0002,
      "loss": 3.1779,
      "step": 30
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0002,
      "loss": 3.1602,
      "step": 35
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0002,
      "loss": 3.1559,
      "step": 40
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0002,
      "loss": 3.1296,
      "step": 45
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0002,
      "loss": 3.1443,
      "step": 50
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0002,
      "loss": 3.1196,
      "step": 55
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 3.0972,
      "step": 60
    },
    {
      "epoch": 0.03,
      "eval_loss": 3.0696167945861816,
      "eval_runtime": 5.2553,
      "eval_samples_per_second": 1879.426,
      "eval_steps_per_second": 14.842,
      "eval_top3_3_weighted_f1_score ": 0.37920731422345033,
      "eval_top_1_macro_f1_score": 0.020407498859125326,
      "eval_top_1_weighted_f1score": 0.1768537842540722,
      "eval_top_3_macro_f1_score": 0.07428603231314447,
      "step": 64
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 3.113,
      "step": 65
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 3.0864,
      "step": 70
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 3.0839,
      "step": 75
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 3.0558,
      "step": 80
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 3.0555,
      "step": 85
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 3.0654,
      "step": 90
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 3.0673,
      "step": 95
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 3.065,
      "step": 100
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 3.054,
      "step": 105
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 3.0582,
      "step": 110
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 3.0719,
      "step": 115
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 3.0522,
      "step": 120
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 3.0465,
      "step": 125
    },
    {
      "epoch": 0.05,
      "eval_loss": 3.0220658779144287,
      "eval_runtime": 5.1808,
      "eval_samples_per_second": 1906.477,
      "eval_steps_per_second": 15.056,
      "eval_top3_3_weighted_f1_score ": 0.3920861263441107,
      "eval_top_1_macro_f1_score": 0.028020399250865188,
      "eval_top_1_weighted_f1score": 0.17802180543353585,
      "eval_top_3_macro_f1_score": 0.09612917294995578,
      "step": 128
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 3.0548,
      "step": 130
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 3.0408,
      "step": 135
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 3.032,
      "step": 140
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 3.0339,
      "step": 145
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 3.0306,
      "step": 150
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 3.0257,
      "step": 155
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 3.0252,
      "step": 160
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 3.0324,
      "step": 165
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 3.0434,
      "step": 170
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 3.0395,
      "step": 175
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 3.0243,
      "step": 180
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 3.0207,
      "step": 185
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 3.0134,
      "step": 190
    },
    {
      "epoch": 0.08,
      "eval_loss": 2.9999098777770996,
      "eval_runtime": 5.5948,
      "eval_samples_per_second": 1765.374,
      "eval_steps_per_second": 13.941,
      "eval_top3_3_weighted_f1_score ": 0.3942607393594228,
      "eval_top_1_macro_f1_score": 0.03024064981912809,
      "eval_top_1_weighted_f1score": 0.18151954856828598,
      "eval_top_3_macro_f1_score": 0.10555039816849211,
      "step": 192
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 3.0304,
      "step": 195
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 3.0166,
      "step": 200
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 3.0238,
      "step": 205
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 3.0408,
      "step": 210
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 3.0097,
      "step": 215
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 3.0377,
      "step": 220
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 3.0076,
      "step": 225
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 3.0003,
      "step": 230
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 3.0113,
      "step": 235
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 2.9959,
      "step": 240
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 3.0143,
      "step": 245
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 2.9976,
      "step": 250
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 3.0203,
      "step": 255
    },
    {
      "epoch": 0.11,
      "eval_loss": 2.9871039390563965,
      "eval_runtime": 5.1769,
      "eval_samples_per_second": 1907.885,
      "eval_steps_per_second": 15.067,
      "eval_top3_3_weighted_f1_score ": 0.3969199243149213,
      "eval_top_1_macro_f1_score": 0.037170653331203264,
      "eval_top_1_weighted_f1score": 0.185406119539436,
      "eval_top_3_macro_f1_score": 0.11274159780578662,
      "step": 256
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 3.0089,
      "step": 260
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 3.0067,
      "step": 265
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 3.0025,
      "step": 270
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 3.006,
      "step": 275
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 2.9925,
      "step": 280
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 2.9842,
      "step": 285
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 3.0024,
      "step": 290
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 2.9809,
      "step": 295
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 2.9923,
      "step": 300
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 2.9932,
      "step": 305
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 2.9727,
      "step": 310
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 2.9862,
      "step": 315
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 2.9879,
      "step": 320
    },
    {
      "epoch": 0.14,
      "eval_loss": 2.9748592376708984,
      "eval_runtime": 5.2664,
      "eval_samples_per_second": 1875.471,
      "eval_steps_per_second": 14.811,
      "eval_top3_3_weighted_f1_score ": 0.4006357227793805,
      "eval_top_1_macro_f1_score": 0.039023552155804886,
      "eval_top_1_weighted_f1score": 0.18467489427014783,
      "eval_top_3_macro_f1_score": 0.12391745887899196,
      "step": 320
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 3.0016,
      "step": 325
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 2.9895,
      "step": 330
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 2.9644,
      "step": 335
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 2.966,
      "step": 340
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 2.9903,
      "step": 345
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 2.9934,
      "step": 350
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 2.9849,
      "step": 355
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 3.0056,
      "step": 360
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 2.9924,
      "step": 365
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 3.0068,
      "step": 370
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 3.004,
      "step": 375
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 2.9847,
      "step": 380
    },
    {
      "epoch": 0.16,
      "eval_loss": 2.965792417526245,
      "eval_runtime": 5.2467,
      "eval_samples_per_second": 1882.515,
      "eval_steps_per_second": 14.866,
      "eval_top3_3_weighted_f1_score ": 0.4021244172567648,
      "eval_top_1_macro_f1_score": 0.04106761958718834,
      "eval_top_1_weighted_f1score": 0.1866042980893108,
      "eval_top_3_macro_f1_score": 0.12467798211027789,
      "step": 384
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 2.984,
      "step": 385
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 2.9918,
      "step": 390
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 2.9835,
      "step": 395
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 2.9888,
      "step": 400
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 2.9929,
      "step": 405
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 2.9953,
      "step": 410
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 3.0103,
      "step": 415
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 2.9881,
      "step": 420
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 2.9919,
      "step": 425
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 2.9959,
      "step": 430
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 2.991,
      "step": 435
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 2.9878,
      "step": 440
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 2.9748,
      "step": 445
    },
    {
      "epoch": 0.19,
      "eval_loss": 2.963198184967041,
      "eval_runtime": 5.205,
      "eval_samples_per_second": 1897.6,
      "eval_steps_per_second": 14.986,
      "eval_top3_3_weighted_f1_score ": 0.40011703728872183,
      "eval_top_1_macro_f1_score": 0.047993869870430424,
      "eval_top_1_weighted_f1score": 0.1872302731003996,
      "eval_top_3_macro_f1_score": 0.12995078978564845,
      "step": 448
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 2.9749,
      "step": 450
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 2.9785,
      "step": 455
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 2.9724,
      "step": 460
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 2.9701,
      "step": 465
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 2.9911,
      "step": 470
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 2.9835,
      "step": 475
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 2.9906,
      "step": 480
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 2.9914,
      "step": 485
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 2.9798,
      "step": 490
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 2.9851,
      "step": 495
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 2.9703,
      "step": 500
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 2.9708,
      "step": 505
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 2.9796,
      "step": 510
    },
    {
      "epoch": 0.22,
      "eval_loss": 2.9612252712249756,
      "eval_runtime": 5.2355,
      "eval_samples_per_second": 1886.559,
      "eval_steps_per_second": 14.898,
      "eval_top3_3_weighted_f1_score ": 0.4014014489333963,
      "eval_top_1_macro_f1_score": 0.04680478656720424,
      "eval_top_1_weighted_f1score": 0.18789703224926918,
      "eval_top_3_macro_f1_score": 0.13950661353435131,
      "step": 512
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 2.9854,
      "step": 515
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 2.9795,
      "step": 520
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 2.9649,
      "step": 525
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 2.9881,
      "step": 530
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 2.9729,
      "step": 535
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 2.9667,
      "step": 540
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 2.9752,
      "step": 545
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 2.9637,
      "step": 550
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 2.9961,
      "step": 555
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 2.9586,
      "step": 560
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 2.9549,
      "step": 565
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 2.9843,
      "step": 570
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 2.971,
      "step": 575
    },
    {
      "epoch": 0.25,
      "eval_loss": 2.954483985900879,
      "eval_runtime": 5.2537,
      "eval_samples_per_second": 1880.015,
      "eval_steps_per_second": 14.847,
      "eval_top3_3_weighted_f1_score ": 0.4037452311007219,
      "eval_top_1_macro_f1_score": 0.05065078032503657,
      "eval_top_1_weighted_f1score": 0.19158153341467882,
      "eval_top_3_macro_f1_score": 0.14248352888683585,
      "step": 576
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 2.9896,
      "step": 580
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 2.9595,
      "step": 585
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 2.9752,
      "step": 590
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 2.9757,
      "step": 595
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 2.9737,
      "step": 600
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 2.99,
      "step": 605
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 2.9835,
      "step": 610
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 2.966,
      "step": 615
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 2.9672,
      "step": 620
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 2.9609,
      "step": 625
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 2.9971,
      "step": 630
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 2.9622,
      "step": 635
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 2.9651,
      "step": 640
    },
    {
      "epoch": 0.27,
      "eval_loss": 2.953902244567871,
      "eval_runtime": 5.539,
      "eval_samples_per_second": 1783.186,
      "eval_steps_per_second": 14.082,
      "eval_top3_3_weighted_f1_score ": 0.4038882116356172,
      "eval_top_1_macro_f1_score": 0.050013153177604464,
      "eval_top_1_weighted_f1score": 0.1915424883657993,
      "eval_top_3_macro_f1_score": 0.13810582785349604,
      "step": 640
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 2.9598,
      "step": 645
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 2.9606,
      "step": 650
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 2.9801,
      "step": 655
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 2.9759,
      "step": 660
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 2.9598,
      "step": 665
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 2.9723,
      "step": 670
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 2.9861,
      "step": 675
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 2.9542,
      "step": 680
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 2.9579,
      "step": 685
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 2.9817,
      "step": 690
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 2.9556,
      "step": 695
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 2.9718,
      "step": 700
    },
    {
      "epoch": 0.3,
      "eval_loss": 2.9501540660858154,
      "eval_runtime": 5.1853,
      "eval_samples_per_second": 1904.791,
      "eval_steps_per_second": 15.042,
      "eval_top3_3_weighted_f1_score ": 0.40385563418101295,
      "eval_top_1_macro_f1_score": 0.05046209013926654,
      "eval_top_1_weighted_f1score": 0.18921842462280833,
      "eval_top_3_macro_f1_score": 0.13372864819392832,
      "step": 704
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 2.96,
      "step": 705
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 2.9647,
      "step": 710
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 2.9719,
      "step": 715
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 2.9711,
      "step": 720
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 2.9726,
      "step": 725
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 2.9719,
      "step": 730
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 2.9694,
      "step": 735
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 2.9623,
      "step": 740
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 2.9862,
      "step": 745
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 2.9744,
      "step": 750
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 2.973,
      "step": 755
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 2.9739,
      "step": 760
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 2.9621,
      "step": 765
    },
    {
      "epoch": 0.33,
      "eval_loss": 2.9458868503570557,
      "eval_runtime": 5.152,
      "eval_samples_per_second": 1917.107,
      "eval_steps_per_second": 15.14,
      "eval_top3_3_weighted_f1_score ": 0.4047092459025178,
      "eval_top_1_macro_f1_score": 0.05353345372204659,
      "eval_top_1_weighted_f1score": 0.187248655714094,
      "eval_top_3_macro_f1_score": 0.14077859018850478,
      "step": 768
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 2.9841,
      "step": 770
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 2.9716,
      "step": 775
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 2.9685,
      "step": 780
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 2.983,
      "step": 785
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 2.962,
      "step": 790
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 2.9715,
      "step": 795
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 2.9698,
      "step": 800
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 2.9408,
      "step": 805
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 2.9695,
      "step": 810
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 2.9706,
      "step": 815
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 2.9626,
      "step": 820
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 2.956,
      "step": 825
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 2.984,
      "step": 830
    },
    {
      "epoch": 0.36,
      "eval_loss": 2.9440317153930664,
      "eval_runtime": 5.2175,
      "eval_samples_per_second": 1893.038,
      "eval_steps_per_second": 14.95,
      "eval_top3_3_weighted_f1_score ": 0.405362919359186,
      "eval_top_1_macro_f1_score": 0.0504562500575876,
      "eval_top_1_weighted_f1score": 0.18343072088628645,
      "eval_top_3_macro_f1_score": 0.13591648502699705,
      "step": 832
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 2.9437,
      "step": 835
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 2.995,
      "step": 840
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 2.9771,
      "step": 845
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 2.9567,
      "step": 850
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 2.973,
      "step": 855
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 2.9657,
      "step": 860
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 2.9627,
      "step": 865
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 2.9428,
      "step": 870
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 2.9718,
      "step": 875
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 2.9623,
      "step": 880
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 2.9516,
      "step": 885
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 2.9539,
      "step": 890
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 2.958,
      "step": 895
    },
    {
      "epoch": 0.38,
      "eval_loss": 2.9393413066864014,
      "eval_runtime": 5.3066,
      "eval_samples_per_second": 1861.255,
      "eval_steps_per_second": 14.699,
      "eval_top3_3_weighted_f1_score ": 0.4115333814510633,
      "eval_top_1_macro_f1_score": 0.05314498607987754,
      "eval_top_1_weighted_f1score": 0.1859384539986762,
      "eval_top_3_macro_f1_score": 0.14574054221754426,
      "step": 896
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 2.9523,
      "step": 900
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 2.9482,
      "step": 905
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 2.9693,
      "step": 910
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 2.9594,
      "step": 915
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 2.9805,
      "step": 920
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 2.9655,
      "step": 925
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 2.963,
      "step": 930
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 2.9678,
      "step": 935
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 2.9783,
      "step": 940
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 2.9668,
      "step": 945
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 2.9748,
      "step": 950
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 2.9671,
      "step": 955
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 2.943,
      "step": 960
    },
    {
      "epoch": 0.41,
      "eval_loss": 2.9382472038269043,
      "eval_runtime": 5.2569,
      "eval_samples_per_second": 1878.847,
      "eval_steps_per_second": 14.838,
      "eval_top3_3_weighted_f1_score ": 0.4070953829436608,
      "eval_top_1_macro_f1_score": 0.0566477657462707,
      "eval_top_1_weighted_f1score": 0.19010243182786668,
      "eval_top_3_macro_f1_score": 0.14185244247312048,
      "step": 960
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 2.97,
      "step": 965
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 2.9586,
      "step": 970
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 2.9601,
      "step": 975
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 2.9669,
      "step": 980
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 2.9548,
      "step": 985
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 2.9478,
      "step": 990
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 2.9489,
      "step": 995
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 2.9437,
      "step": 1000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 2.9588,
      "step": 1005
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 2.9625,
      "step": 1010
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 2.97,
      "step": 1015
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 2.9745,
      "step": 1020
    },
    {
      "epoch": 0.44,
      "eval_loss": 2.9444897174835205,
      "eval_runtime": 5.212,
      "eval_samples_per_second": 1895.057,
      "eval_steps_per_second": 14.966,
      "eval_top3_3_weighted_f1_score ": 0.40828849035898257,
      "eval_top_1_macro_f1_score": 0.056020443190640944,
      "eval_top_1_weighted_f1score": 0.19234719626068883,
      "eval_top_3_macro_f1_score": 0.14346199622671688,
      "step": 1024
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 2.9545,
      "step": 1025
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 2.9525,
      "step": 1030
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 2.9498,
      "step": 1035
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 2.9585,
      "step": 1040
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 2.964,
      "step": 1045
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 2.9636,
      "step": 1050
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 2.9333,
      "step": 1055
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 2.9628,
      "step": 1060
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 2.9565,
      "step": 1065
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 2.9596,
      "step": 1070
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 2.957,
      "step": 1075
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 2.9798,
      "step": 1080
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 2.9657,
      "step": 1085
    },
    {
      "epoch": 0.47,
      "eval_loss": 2.9426662921905518,
      "eval_runtime": 5.6668,
      "eval_samples_per_second": 1742.966,
      "eval_steps_per_second": 13.764,
      "eval_top3_3_weighted_f1_score ": 0.40339180646168105,
      "eval_top_1_macro_f1_score": 0.05765999455627635,
      "eval_top_1_weighted_f1score": 0.1923372887115798,
      "eval_top_3_macro_f1_score": 0.13777971543818865,
      "step": 1088
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 2.9568,
      "step": 1090
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 2.9686,
      "step": 1095
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 2.9574,
      "step": 1100
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 2.9382,
      "step": 1105
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 2.9588,
      "step": 1110
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 2.9431,
      "step": 1115
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 2.9496,
      "step": 1120
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 2.9554,
      "step": 1125
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 2.9468,
      "step": 1130
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 2.9416,
      "step": 1135
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 2.95,
      "step": 1140
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 2.9422,
      "step": 1145
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 2.9473,
      "step": 1150
    },
    {
      "epoch": 0.49,
      "eval_loss": 2.9419898986816406,
      "eval_runtime": 5.2315,
      "eval_samples_per_second": 1888.003,
      "eval_steps_per_second": 14.91,
      "eval_top3_3_weighted_f1_score ": 0.40865085111737687,
      "eval_top_1_macro_f1_score": 0.05391141724799508,
      "eval_top_1_weighted_f1score": 0.1920279224053727,
      "eval_top_3_macro_f1_score": 0.14288060798438232,
      "step": 1152
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 2.9533,
      "step": 1155
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 2.955,
      "step": 1160
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 2.9692,
      "step": 1165
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 2.9594,
      "step": 1170
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 2.9603,
      "step": 1175
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 2.9575,
      "step": 1180
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 2.9564,
      "step": 1185
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 2.9645,
      "step": 1190
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 2.9553,
      "step": 1195
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 2.9523,
      "step": 1200
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 2.9438,
      "step": 1205
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 2.9815,
      "step": 1210
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 2.961,
      "step": 1215
    },
    {
      "epoch": 0.52,
      "eval_loss": 2.9412014484405518,
      "eval_runtime": 5.2238,
      "eval_samples_per_second": 1890.778,
      "eval_steps_per_second": 14.932,
      "eval_top3_3_weighted_f1_score ": 0.40786636359138395,
      "eval_top_1_macro_f1_score": 0.0474824971739334,
      "eval_top_1_weighted_f1score": 0.18867625959655746,
      "eval_top_3_macro_f1_score": 0.14364241717026507,
      "step": 1216
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 2.9478,
      "step": 1220
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 2.9594,
      "step": 1225
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 2.9485,
      "step": 1230
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 2.9543,
      "step": 1235
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 2.9607,
      "step": 1240
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 2.9439,
      "step": 1245
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 2.9231,
      "step": 1250
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 2.9409,
      "step": 1255
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 2.9574,
      "step": 1260
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 2.937,
      "step": 1265
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 2.9455,
      "step": 1270
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 2.94,
      "step": 1275
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 2.9752,
      "step": 1280
    },
    {
      "epoch": 0.55,
      "eval_loss": 2.93802809715271,
      "eval_runtime": 5.3062,
      "eval_samples_per_second": 1861.394,
      "eval_steps_per_second": 14.7,
      "eval_top3_3_weighted_f1_score ": 0.41160079192129195,
      "eval_top_1_macro_f1_score": 0.05795090891818421,
      "eval_top_1_weighted_f1score": 0.19012240817937326,
      "eval_top_3_macro_f1_score": 0.1487031524147606,
      "step": 1280
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 2.9508,
      "step": 1285
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 2.9331,
      "step": 1290
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 2.948,
      "step": 1295
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 2.9513,
      "step": 1300
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 2.9614,
      "step": 1305
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 2.9509,
      "step": 1310
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 2.9331,
      "step": 1315
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 2.9668,
      "step": 1320
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 2.9467,
      "step": 1325
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 2.9578,
      "step": 1330
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 2.9483,
      "step": 1335
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 2.9415,
      "step": 1340
    },
    {
      "epoch": 0.58,
      "eval_loss": 2.941464900970459,
      "eval_runtime": 5.2866,
      "eval_samples_per_second": 1868.309,
      "eval_steps_per_second": 14.754,
      "eval_top3_3_weighted_f1_score ": 0.4069475337150649,
      "eval_top_1_macro_f1_score": 0.052197365761238596,
      "eval_top_1_weighted_f1score": 0.18994939652129228,
      "eval_top_3_macro_f1_score": 0.14622829923307734,
      "step": 1344
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 2.965,
      "step": 1345
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 2.9499,
      "step": 1350
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 2.9565,
      "step": 1355
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 2.9463,
      "step": 1360
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 2.9597,
      "step": 1365
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 2.9346,
      "step": 1370
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 2.9643,
      "step": 1375
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 2.928,
      "step": 1380
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 2.9518,
      "step": 1385
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 2.9471,
      "step": 1390
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 2.9582,
      "step": 1395
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 2.9396,
      "step": 1400
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 2.9633,
      "step": 1405
    },
    {
      "epoch": 0.6,
      "eval_loss": 2.9351675510406494,
      "eval_runtime": 5.2046,
      "eval_samples_per_second": 1897.729,
      "eval_steps_per_second": 14.987,
      "eval_top3_3_weighted_f1_score ": 0.40903510064773524,
      "eval_top_1_macro_f1_score": 0.058491334775530215,
      "eval_top_1_weighted_f1score": 0.1931870648145974,
      "eval_top_3_macro_f1_score": 0.14995626358660458,
      "step": 1408
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 2.9465,
      "step": 1410
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 2.9488,
      "step": 1415
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 2.9451,
      "step": 1420
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 2.9526,
      "step": 1425
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 2.9276,
      "step": 1430
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 2.9354,
      "step": 1435
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 2.9543,
      "step": 1440
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 2.9433,
      "step": 1445
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 2.9343,
      "step": 1450
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 2.9643,
      "step": 1455
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 2.9505,
      "step": 1460
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 2.9454,
      "step": 1465
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 2.9607,
      "step": 1470
    },
    {
      "epoch": 0.63,
      "eval_loss": 2.929859161376953,
      "eval_runtime": 5.188,
      "eval_samples_per_second": 1903.819,
      "eval_steps_per_second": 15.035,
      "eval_top3_3_weighted_f1_score ": 0.4112572271075635,
      "eval_top_1_macro_f1_score": 0.05492674184491993,
      "eval_top_1_weighted_f1score": 0.1874419660454753,
      "eval_top_3_macro_f1_score": 0.15057532260912443,
      "step": 1472
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 2.9707,
      "step": 1475
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 2.9234,
      "step": 1480
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 2.9706,
      "step": 1485
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 2.924,
      "step": 1490
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 2.9582,
      "step": 1495
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 2.9376,
      "step": 1500
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 2.9393,
      "step": 1505
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 2.9135,
      "step": 1510
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 2.9541,
      "step": 1515
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 2.9508,
      "step": 1520
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 2.9382,
      "step": 1525
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 2.9257,
      "step": 1530
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 2.9421,
      "step": 1535
    },
    {
      "epoch": 0.66,
      "eval_loss": 2.935636043548584,
      "eval_runtime": 5.498,
      "eval_samples_per_second": 1796.478,
      "eval_steps_per_second": 14.187,
      "eval_top3_3_weighted_f1_score ": 0.4054698890830444,
      "eval_top_1_macro_f1_score": 0.0557953259862022,
      "eval_top_1_weighted_f1score": 0.19142192923965634,
      "eval_top_3_macro_f1_score": 0.14438591551016072,
      "step": 1536
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 2.928,
      "step": 1540
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 2.963,
      "step": 1545
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 2.9523,
      "step": 1550
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 2.9275,
      "step": 1555
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 2.9429,
      "step": 1560
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 2.9386,
      "step": 1565
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 2.9524,
      "step": 1570
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 2.9344,
      "step": 1575
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 2.9353,
      "step": 1580
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 2.9316,
      "step": 1585
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 2.96,
      "step": 1590
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 2.9282,
      "step": 1595
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 2.9194,
      "step": 1600
    },
    {
      "epoch": 0.69,
      "eval_loss": 2.9320883750915527,
      "eval_runtime": 5.2444,
      "eval_samples_per_second": 1883.348,
      "eval_steps_per_second": 14.873,
      "eval_top3_3_weighted_f1_score ": 0.40821638906631724,
      "eval_top_1_macro_f1_score": 0.05321140707012678,
      "eval_top_1_weighted_f1score": 0.190021300619809,
      "eval_top_3_macro_f1_score": 0.14259731701282546,
      "step": 1600
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 2.9443,
      "step": 1605
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 2.9445,
      "step": 1610
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 2.9489,
      "step": 1615
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 2.945,
      "step": 1620
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 2.9427,
      "step": 1625
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 2.9204,
      "step": 1630
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 2.9226,
      "step": 1635
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 2.9547,
      "step": 1640
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 2.9377,
      "step": 1645
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 2.9445,
      "step": 1650
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 2.9534,
      "step": 1655
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 2.9466,
      "step": 1660
    },
    {
      "epoch": 0.71,
      "eval_loss": 2.9336674213409424,
      "eval_runtime": 5.2704,
      "eval_samples_per_second": 1874.037,
      "eval_steps_per_second": 14.8,
      "eval_top3_3_weighted_f1_score ": 0.41047552952278205,
      "eval_top_1_macro_f1_score": 0.0536502267865967,
      "eval_top_1_weighted_f1score": 0.19126675755018152,
      "eval_top_3_macro_f1_score": 0.1456769847671719,
      "step": 1664
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 2.9518,
      "step": 1665
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 2.9646,
      "step": 1670
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 2.9389,
      "step": 1675
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 2.9527,
      "step": 1680
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 2.9399,
      "step": 1685
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 2.9312,
      "step": 1690
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 2.9517,
      "step": 1695
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 2.9561,
      "step": 1700
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 2.9535,
      "step": 1705
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 2.939,
      "step": 1710
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 2.9429,
      "step": 1715
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 2.938,
      "step": 1720
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 2.9497,
      "step": 1725
    },
    {
      "epoch": 0.74,
      "eval_loss": 2.9341320991516113,
      "eval_runtime": 5.2886,
      "eval_samples_per_second": 1867.59,
      "eval_steps_per_second": 14.749,
      "eval_top3_3_weighted_f1_score ": 0.4057059914344135,
      "eval_top_1_macro_f1_score": 0.05370325541314189,
      "eval_top_1_weighted_f1score": 0.19225534007309714,
      "eval_top_3_macro_f1_score": 0.13987396780712794,
      "step": 1728
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 2.9489,
      "step": 1730
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 2.955,
      "step": 1735
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 2.9453,
      "step": 1740
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 2.9324,
      "step": 1745
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 2.942,
      "step": 1750
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 2.9264,
      "step": 1755
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 2.9291,
      "step": 1760
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 2.9455,
      "step": 1765
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 2.9395,
      "step": 1770
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 2.9584,
      "step": 1775
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 2.9371,
      "step": 1780
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 2.947,
      "step": 1785
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 2.9429,
      "step": 1790
    },
    {
      "epoch": 0.77,
      "eval_loss": 2.9273977279663086,
      "eval_runtime": 5.296,
      "eval_samples_per_second": 1864.99,
      "eval_steps_per_second": 14.728,
      "eval_top3_3_weighted_f1_score ": 0.4131031890088074,
      "eval_top_1_macro_f1_score": 0.05607100541398047,
      "eval_top_1_weighted_f1score": 0.18862150823940385,
      "eval_top_3_macro_f1_score": 0.14317950642794616,
      "step": 1792
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 2.9382,
      "step": 1795
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 2.9594,
      "step": 1800
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 2.9377,
      "step": 1805
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 2.944,
      "step": 1810
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 2.9318,
      "step": 1815
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 2.9465,
      "step": 1820
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 2.9327,
      "step": 1825
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 2.9411,
      "step": 1830
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 2.9453,
      "step": 1835
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 2.9551,
      "step": 1840
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 2.9551,
      "step": 1845
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 2.9496,
      "step": 1850
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 2.9507,
      "step": 1855
    },
    {
      "epoch": 0.8,
      "eval_loss": 2.9243032932281494,
      "eval_runtime": 5.2513,
      "eval_samples_per_second": 1880.88,
      "eval_steps_per_second": 14.854,
      "eval_top3_3_weighted_f1_score ": 0.4053991556785117,
      "eval_top_1_macro_f1_score": 0.05826966485471442,
      "eval_top_1_weighted_f1score": 0.19437289584739018,
      "eval_top_3_macro_f1_score": 0.13632058202701589,
      "step": 1856
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 2.9426,
      "step": 1860
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 2.9279,
      "step": 1865
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 2.9562,
      "step": 1870
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 2.931,
      "step": 1875
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 2.9338,
      "step": 1880
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 2.9204,
      "step": 1885
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 2.964,
      "step": 1890
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 2.9598,
      "step": 1895
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 2.941,
      "step": 1900
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 2.9377,
      "step": 1905
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 2.942,
      "step": 1910
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 2.9381,
      "step": 1915
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 2.9409,
      "step": 1920
    },
    {
      "epoch": 0.82,
      "eval_loss": 2.9285144805908203,
      "eval_runtime": 5.3302,
      "eval_samples_per_second": 1853.038,
      "eval_steps_per_second": 14.634,
      "eval_top3_3_weighted_f1_score ": 0.4124377696571018,
      "eval_top_1_macro_f1_score": 0.06190878574489557,
      "eval_top_1_weighted_f1score": 0.19263435856385583,
      "eval_top_3_macro_f1_score": 0.14077472295761495,
      "step": 1920
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 2.9466,
      "step": 1925
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 2.9365,
      "step": 1930
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 2.915,
      "step": 1935
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 2.9284,
      "step": 1940
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 2.9357,
      "step": 1945
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 2.9482,
      "step": 1950
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 2.9476,
      "step": 1955
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 2.9314,
      "step": 1960
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 2.9429,
      "step": 1965
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 2.9536,
      "step": 1970
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 2.9512,
      "step": 1975
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 2.9371,
      "step": 1980
    },
    {
      "epoch": 0.85,
      "eval_loss": 2.928523540496826,
      "eval_runtime": 5.2792,
      "eval_samples_per_second": 1870.915,
      "eval_steps_per_second": 14.775,
      "eval_top3_3_weighted_f1_score ": 0.41511573211461816,
      "eval_top_1_macro_f1_score": 0.06132345745002284,
      "eval_top_1_weighted_f1score": 0.19199397596078466,
      "eval_top_3_macro_f1_score": 0.15076717368083384,
      "step": 1984
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 2.9442,
      "step": 1985
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 2.931,
      "step": 1990
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 2.9765,
      "step": 1995
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 2.9434,
      "step": 2000
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 2.9254,
      "step": 2005
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 2.9435,
      "step": 2010
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 2.9263,
      "step": 2015
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 2.9388,
      "step": 2020
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 2.9447,
      "step": 2025
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 2.9527,
      "step": 2030
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 2.9114,
      "step": 2035
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 2.9575,
      "step": 2040
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 2.9203,
      "step": 2045
    },
    {
      "epoch": 0.88,
      "eval_loss": 2.921570062637329,
      "eval_runtime": 5.2505,
      "eval_samples_per_second": 1881.171,
      "eval_steps_per_second": 14.856,
      "eval_top3_3_weighted_f1_score ": 0.4120243601453162,
      "eval_top_1_macro_f1_score": 0.05856737068306764,
      "eval_top_1_weighted_f1score": 0.19277721536806952,
      "eval_top_3_macro_f1_score": 0.14630634863938272,
      "step": 2048
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 2.9216,
      "step": 2050
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 2.9413,
      "step": 2055
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 2.9394,
      "step": 2060
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 2.94,
      "step": 2065
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 2.9602,
      "step": 2070
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 2.9339,
      "step": 2075
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 2.9453,
      "step": 2080
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 2.9346,
      "step": 2085
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 2.9427,
      "step": 2090
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 2.932,
      "step": 2095
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 2.9356,
      "step": 2100
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 2.9558,
      "step": 2105
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 2.949,
      "step": 2110
    },
    {
      "epoch": 0.91,
      "eval_loss": 2.9211103916168213,
      "eval_runtime": 5.8002,
      "eval_samples_per_second": 1702.864,
      "eval_steps_per_second": 13.448,
      "eval_top3_3_weighted_f1_score ": 0.411140072973993,
      "eval_top_1_macro_f1_score": 0.05827460613661236,
      "eval_top_1_weighted_f1score": 0.1900658924831603,
      "eval_top_3_macro_f1_score": 0.14356756002606255,
      "step": 2112
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 2.9237,
      "step": 2115
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 2.9368,
      "step": 2120
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 2.924,
      "step": 2125
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 2.9344,
      "step": 2130
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 2.9425,
      "step": 2135
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 2.9303,
      "step": 2140
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 2.9222,
      "step": 2145
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 2.9271,
      "step": 2150
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 2.9429,
      "step": 2155
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 2.9412,
      "step": 2160
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 2.9398,
      "step": 2165
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 2.9253,
      "step": 2170
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 2.9378,
      "step": 2175
    },
    {
      "epoch": 0.93,
      "eval_loss": 2.9222617149353027,
      "eval_runtime": 5.2532,
      "eval_samples_per_second": 1880.202,
      "eval_steps_per_second": 14.848,
      "eval_top3_3_weighted_f1_score ": 0.4077832403009119,
      "eval_top_1_macro_f1_score": 0.06079385373525116,
      "eval_top_1_weighted_f1score": 0.19627489088203884,
      "eval_top_3_macro_f1_score": 0.14590453185761695,
      "step": 2176
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 2.9355,
      "step": 2180
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 2.9415,
      "step": 2185
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 2.9342,
      "step": 2190
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 2.9168,
      "step": 2195
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 2.9586,
      "step": 2200
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 2.9345,
      "step": 2205
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 2.9422,
      "step": 2210
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 2.9417,
      "step": 2215
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 2.9322,
      "step": 2220
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 2.94,
      "step": 2225
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 2.9378,
      "step": 2230
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 2.9127,
      "step": 2235
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 2.9423,
      "step": 2240
    },
    {
      "epoch": 0.96,
      "eval_loss": 2.9238317012786865,
      "eval_runtime": 5.2712,
      "eval_samples_per_second": 1873.771,
      "eval_steps_per_second": 14.797,
      "eval_top3_3_weighted_f1_score ": 0.41065264958740455,
      "eval_top_1_macro_f1_score": 0.06100532477583297,
      "eval_top_1_weighted_f1score": 0.19239137323566624,
      "eval_top_3_macro_f1_score": 0.14523224439471996,
      "step": 2240
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 2.9176,
      "step": 2245
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 2.9513,
      "step": 2250
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 2.9281,
      "step": 2255
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 2.9392,
      "step": 2260
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 2.9205,
      "step": 2265
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 2.9252,
      "step": 2270
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 2.9274,
      "step": 2275
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 2.925,
      "step": 2280
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 2.9356,
      "step": 2285
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 2.932,
      "step": 2290
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 2.9383,
      "step": 2295
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 2.9329,
      "step": 2300
    },
    {
      "epoch": 0.99,
      "eval_loss": 2.9221272468566895,
      "eval_runtime": 5.3659,
      "eval_samples_per_second": 1840.711,
      "eval_steps_per_second": 14.536,
      "eval_top3_3_weighted_f1_score ": 0.4150667448633111,
      "eval_top_1_macro_f1_score": 0.05624215902445944,
      "eval_top_1_weighted_f1score": 0.1924437688190861,
      "eval_top_3_macro_f1_score": 0.14481328991643022,
      "step": 2304
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 2.943,
      "step": 2305
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 2.9167,
      "step": 2310
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 2.9415,
      "step": 2315
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 2.9308,
      "step": 2320
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 2.9441,
      "step": 2325
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 3.1956,
      "step": 2330
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 2.9391,
      "step": 2335
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 2.9307,
      "step": 2340
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0002,
      "loss": 2.9192,
      "step": 2345
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0002,
      "loss": 2.9519,
      "step": 2350
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0002,
      "loss": 2.934,
      "step": 2355
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.0002,
      "loss": 2.9205,
      "step": 2360
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0002,
      "loss": 2.9126,
      "step": 2365
    },
    {
      "epoch": 1.02,
      "eval_loss": 2.9245643615722656,
      "eval_runtime": 5.3574,
      "eval_samples_per_second": 1843.634,
      "eval_steps_per_second": 14.559,
      "eval_top3_3_weighted_f1_score ": 0.41190062778999437,
      "eval_top_1_macro_f1_score": 0.05170441427634679,
      "eval_top_1_weighted_f1score": 0.1915950465826955,
      "eval_top_3_macro_f1_score": 0.14541768101651006,
      "step": 2368
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0002,
      "loss": 2.923,
      "step": 2370
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0002,
      "loss": 2.9124,
      "step": 2375
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0002,
      "loss": 2.9111,
      "step": 2380
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.0002,
      "loss": 2.9083,
      "step": 2385
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0002,
      "loss": 2.9273,
      "step": 2390
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0002,
      "loss": 2.9231,
      "step": 2395
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0002,
      "loss": 2.938,
      "step": 2400
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0002,
      "loss": 2.9343,
      "step": 2405
    },
    {
      "epoch": 1.03,
      "learning_rate": 0.0002,
      "loss": 2.936,
      "step": 2410
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0002,
      "loss": 2.9219,
      "step": 2415
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0002,
      "loss": 2.9297,
      "step": 2420
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0002,
      "loss": 2.942,
      "step": 2425
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.0002,
      "loss": 2.9285,
      "step": 2430
    },
    {
      "epoch": 1.04,
      "eval_loss": 2.9210715293884277,
      "eval_runtime": 5.2926,
      "eval_samples_per_second": 1866.205,
      "eval_steps_per_second": 14.738,
      "eval_top3_3_weighted_f1_score ": 0.4104425810756134,
      "eval_top_1_macro_f1_score": 0.058897211609453624,
      "eval_top_1_weighted_f1score": 0.19441356472244228,
      "eval_top_3_macro_f1_score": 0.14331504611626866,
      "step": 2432
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0002,
      "loss": 2.9341,
      "step": 2435
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0002,
      "loss": 2.918,
      "step": 2440
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0002,
      "loss": 2.9321,
      "step": 2445
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0002,
      "loss": 2.9285,
      "step": 2450
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0002,
      "loss": 2.9324,
      "step": 2455
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0002,
      "loss": 2.942,
      "step": 2460
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0002,
      "loss": 2.909,
      "step": 2465
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0002,
      "loss": 2.9444,
      "step": 2470
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0002,
      "loss": 2.9411,
      "step": 2475
    },
    {
      "epoch": 1.06,
      "learning_rate": 0.0002,
      "loss": 2.9211,
      "step": 2480
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0002,
      "loss": 2.9457,
      "step": 2485
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0002,
      "loss": 2.931,
      "step": 2490
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0002,
      "loss": 2.9121,
      "step": 2495
    },
    {
      "epoch": 1.07,
      "eval_loss": 2.9219577312469482,
      "eval_runtime": 5.3428,
      "eval_samples_per_second": 1848.645,
      "eval_steps_per_second": 14.599,
      "eval_top3_3_weighted_f1_score ": 0.4126210764843451,
      "eval_top_1_macro_f1_score": 0.055206132881277424,
      "eval_top_1_weighted_f1score": 0.19580194845230464,
      "eval_top_3_macro_f1_score": 0.14175695902025756,
      "step": 2496
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0002,
      "loss": 2.9203,
      "step": 2500
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0002,
      "loss": 2.9403,
      "step": 2505
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0002,
      "loss": 2.9353,
      "step": 2510
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0002,
      "loss": 2.9345,
      "step": 2515
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0002,
      "loss": 2.9246,
      "step": 2520
    },
    {
      "epoch": 1.08,
      "learning_rate": 0.0002,
      "loss": 2.9118,
      "step": 2525
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0002,
      "loss": 2.9406,
      "step": 2530
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0002,
      "loss": 2.9261,
      "step": 2535
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0002,
      "loss": 2.9172,
      "step": 2540
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0002,
      "loss": 2.9372,
      "step": 2545
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.0002,
      "loss": 2.9009,
      "step": 2550
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0002,
      "loss": 2.9144,
      "step": 2555
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0002,
      "loss": 2.9015,
      "step": 2560
    },
    {
      "epoch": 1.1,
      "eval_loss": 2.9242866039276123,
      "eval_runtime": 5.7071,
      "eval_samples_per_second": 1730.647,
      "eval_steps_per_second": 13.667,
      "eval_top3_3_weighted_f1_score ": 0.4107174514165816,
      "eval_top_1_macro_f1_score": 0.05746469658311381,
      "eval_top_1_weighted_f1score": 0.19074469006616215,
      "eval_top_3_macro_f1_score": 0.1424298406329182,
      "step": 2560
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0002,
      "loss": 2.933,
      "step": 2565
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0002,
      "loss": 2.9143,
      "step": 2570
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0002,
      "loss": 2.9446,
      "step": 2575
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0002,
      "loss": 2.9453,
      "step": 2580
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0002,
      "loss": 2.9048,
      "step": 2585
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0002,
      "loss": 2.9319,
      "step": 2590
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0002,
      "loss": 2.9199,
      "step": 2595
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0002,
      "loss": 2.9281,
      "step": 2600
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0002,
      "loss": 2.9218,
      "step": 2605
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0002,
      "loss": 2.9381,
      "step": 2610
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0002,
      "loss": 2.935,
      "step": 2615
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.0002,
      "loss": 2.9473,
      "step": 2620
    },
    {
      "epoch": 1.13,
      "eval_loss": 2.9131009578704834,
      "eval_runtime": 5.3565,
      "eval_samples_per_second": 1843.945,
      "eval_steps_per_second": 14.562,
      "eval_top3_3_weighted_f1_score ": 0.41318059902803694,
      "eval_top_1_macro_f1_score": 0.05385593199473167,
      "eval_top_1_weighted_f1score": 0.19077652700730305,
      "eval_top_3_macro_f1_score": 0.1408254102673545,
      "step": 2624
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0002,
      "loss": 2.9366,
      "step": 2625
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0002,
      "loss": 2.9195,
      "step": 2630
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0002,
      "loss": 2.9383,
      "step": 2635
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.0002,
      "loss": 2.9316,
      "step": 2640
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0002,
      "loss": 2.9528,
      "step": 2645
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0002,
      "loss": 2.939,
      "step": 2650
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0002,
      "loss": 2.9093,
      "step": 2655
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0002,
      "loss": 2.9258,
      "step": 2660
    },
    {
      "epoch": 1.14,
      "learning_rate": 0.0002,
      "loss": 2.9124,
      "step": 2665
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0002,
      "loss": 2.9377,
      "step": 2670
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0002,
      "loss": 2.9199,
      "step": 2675
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0002,
      "loss": 2.9306,
      "step": 2680
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0002,
      "loss": 2.9142,
      "step": 2685
    },
    {
      "epoch": 1.15,
      "eval_loss": 2.9189188480377197,
      "eval_runtime": 5.3137,
      "eval_samples_per_second": 1858.775,
      "eval_steps_per_second": 14.679,
      "eval_top3_3_weighted_f1_score ": 0.4110172769165812,
      "eval_top_1_macro_f1_score": 0.059839117057030516,
      "eval_top_1_weighted_f1score": 0.19338477249901195,
      "eval_top_3_macro_f1_score": 0.14382223501334376,
      "step": 2688
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.0002,
      "loss": 2.9331,
      "step": 2690
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0002,
      "loss": 2.9393,
      "step": 2695
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0002,
      "loss": 2.9215,
      "step": 2700
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0002,
      "loss": 2.9379,
      "step": 2705
    },
    {
      "epoch": 1.16,
      "learning_rate": 0.0002,
      "loss": 2.906,
      "step": 2710
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0002,
      "loss": 2.9306,
      "step": 2715
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0002,
      "loss": 2.9323,
      "step": 2720
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0002,
      "loss": 2.928,
      "step": 2725
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0002,
      "loss": 2.9304,
      "step": 2730
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.0002,
      "loss": 2.9281,
      "step": 2735
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0002,
      "loss": 2.9403,
      "step": 2740
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0002,
      "loss": 2.9239,
      "step": 2745
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0002,
      "loss": 2.9297,
      "step": 2750
    },
    {
      "epoch": 1.18,
      "eval_loss": 2.9221882820129395,
      "eval_runtime": 5.2478,
      "eval_samples_per_second": 1882.135,
      "eval_steps_per_second": 14.863,
      "eval_top3_3_weighted_f1_score ": 0.41274163601888564,
      "eval_top_1_macro_f1_score": 0.05734115522215458,
      "eval_top_1_weighted_f1score": 0.19576214347043555,
      "eval_top_3_macro_f1_score": 0.14691401943928395,
      "step": 2752
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.0002,
      "loss": 2.9317,
      "step": 2755
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0002,
      "loss": 2.941,
      "step": 2760
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0002,
      "loss": 2.9057,
      "step": 2765
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0002,
      "loss": 2.9468,
      "step": 2770
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0002,
      "loss": 2.9253,
      "step": 2775
    },
    {
      "epoch": 1.19,
      "learning_rate": 0.0002,
      "loss": 2.9225,
      "step": 2780
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0002,
      "loss": 2.9352,
      "step": 2785
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0002,
      "loss": 2.9171,
      "step": 2790
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0002,
      "loss": 2.9241,
      "step": 2795
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0002,
      "loss": 2.9361,
      "step": 2800
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0002,
      "loss": 2.9227,
      "step": 2805
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0002,
      "loss": 2.9213,
      "step": 2810
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0002,
      "loss": 2.9134,
      "step": 2815
    },
    {
      "epoch": 1.21,
      "eval_loss": 2.9217724800109863,
      "eval_runtime": 5.2627,
      "eval_samples_per_second": 1876.799,
      "eval_steps_per_second": 14.821,
      "eval_top3_3_weighted_f1_score ": 0.4107645977708775,
      "eval_top_1_macro_f1_score": 0.056444541131062925,
      "eval_top_1_weighted_f1score": 0.19637409796491542,
      "eval_top_3_macro_f1_score": 0.1400984673549117,
      "step": 2816
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0002,
      "loss": 2.9225,
      "step": 2820
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.0002,
      "loss": 2.958,
      "step": 2825
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 2830
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0002,
      "loss": 2.9381,
      "step": 2835
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0002,
      "loss": 2.9127,
      "step": 2840
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0002,
      "loss": 2.9378,
      "step": 2845
    },
    {
      "epoch": 1.22,
      "learning_rate": 0.0002,
      "loss": 2.911,
      "step": 2850
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0002,
      "loss": 2.9222,
      "step": 2855
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0002,
      "loss": 2.9215,
      "step": 2860
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0002,
      "loss": 2.9375,
      "step": 2865
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0002,
      "loss": 2.9454,
      "step": 2870
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.0002,
      "loss": 2.9096,
      "step": 2875
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0002,
      "loss": 2.9152,
      "step": 2880
    },
    {
      "epoch": 1.24,
      "eval_loss": 2.9204282760620117,
      "eval_runtime": 5.3058,
      "eval_samples_per_second": 1861.544,
      "eval_steps_per_second": 14.701,
      "eval_top3_3_weighted_f1_score ": 0.40930973081815303,
      "eval_top_1_macro_f1_score": 0.05561365452286687,
      "eval_top_1_weighted_f1score": 0.19327404630127948,
      "eval_top_3_macro_f1_score": 0.14019544772441375,
      "step": 2880
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0002,
      "loss": 2.9167,
      "step": 2885
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0002,
      "loss": 2.922,
      "step": 2890
    },
    {
      "epoch": 1.24,
      "learning_rate": 0.0002,
      "loss": 2.9246,
      "step": 2895
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0002,
      "loss": 2.9248,
      "step": 2900
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0002,
      "loss": 2.9297,
      "step": 2905
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0002,
      "loss": 2.9257,
      "step": 2910
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0002,
      "loss": 2.9427,
      "step": 2915
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0002,
      "loss": 2.9243,
      "step": 2920
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0002,
      "loss": 2.9384,
      "step": 2925
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0002,
      "loss": 2.9396,
      "step": 2930
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0002,
      "loss": 2.929,
      "step": 2935
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0002,
      "loss": 2.9284,
      "step": 2940
    },
    {
      "epoch": 1.26,
      "eval_loss": 2.91880464553833,
      "eval_runtime": 5.2529,
      "eval_samples_per_second": 1880.281,
      "eval_steps_per_second": 14.849,
      "eval_top3_3_weighted_f1_score ": 0.41468391257401327,
      "eval_top_1_macro_f1_score": 0.05407696011087027,
      "eval_top_1_weighted_f1score": 0.19182727350240894,
      "eval_top_3_macro_f1_score": 0.14511553350894063,
      "step": 2944
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.0002,
      "loss": 2.9168,
      "step": 2945
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0002,
      "loss": 2.9287,
      "step": 2950
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0002,
      "loss": 2.9231,
      "step": 2955
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0002,
      "loss": 2.9235,
      "step": 2960
    },
    {
      "epoch": 1.27,
      "learning_rate": 0.0002,
      "loss": 2.9083,
      "step": 2965
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0002,
      "loss": 2.9331,
      "step": 2970
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0002,
      "loss": 2.9207,
      "step": 2975
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0002,
      "loss": 2.9197,
      "step": 2980
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0002,
      "loss": 2.9176,
      "step": 2985
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.0002,
      "loss": 2.9332,
      "step": 2990
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0002,
      "loss": 2.9367,
      "step": 2995
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0002,
      "loss": 2.9198,
      "step": 3000
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0002,
      "loss": 2.9148,
      "step": 3005
    },
    {
      "epoch": 1.29,
      "eval_loss": 2.919055461883545,
      "eval_runtime": 5.8139,
      "eval_samples_per_second": 1698.868,
      "eval_steps_per_second": 13.416,
      "eval_top3_3_weighted_f1_score ": 0.41247529392678284,
      "eval_top_1_macro_f1_score": 0.054934720407290305,
      "eval_top_1_weighted_f1score": 0.19640278010652523,
      "eval_top_3_macro_f1_score": 0.1467724167948877,
      "step": 3008
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0002,
      "loss": 2.9363,
      "step": 3010
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.0002,
      "loss": 2.9326,
      "step": 3015
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0002,
      "loss": 2.9316,
      "step": 3020
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0002,
      "loss": 2.9109,
      "step": 3025
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0002,
      "loss": 2.9285,
      "step": 3030
    },
    {
      "epoch": 1.3,
      "learning_rate": 0.0002,
      "loss": 2.9181,
      "step": 3035
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0002,
      "loss": 2.9333,
      "step": 3040
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0002,
      "loss": 2.9034,
      "step": 3045
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0002,
      "loss": 2.9173,
      "step": 3050
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0002,
      "loss": 2.9323,
      "step": 3055
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.0002,
      "loss": 2.9288,
      "step": 3060
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0002,
      "loss": 2.9114,
      "step": 3065
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0002,
      "loss": 2.9247,
      "step": 3070
    },
    {
      "epoch": 1.32,
      "eval_loss": 2.9235095977783203,
      "eval_runtime": 5.3056,
      "eval_samples_per_second": 1861.629,
      "eval_steps_per_second": 14.702,
      "eval_top3_3_weighted_f1_score ": 0.4146372781603772,
      "eval_top_1_macro_f1_score": 0.05380511361692963,
      "eval_top_1_weighted_f1score": 0.19638734169374839,
      "eval_top_3_macro_f1_score": 0.13671213989889544,
      "step": 3072
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0002,
      "loss": 2.9449,
      "step": 3075
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0002,
      "loss": 2.9157,
      "step": 3080
    },
    {
      "epoch": 1.32,
      "learning_rate": 0.0002,
      "loss": 2.9232,
      "step": 3085
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0002,
      "loss": 2.9436,
      "step": 3090
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0002,
      "loss": 2.9291,
      "step": 3095
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0002,
      "loss": 2.94,
      "step": 3100
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.0002,
      "loss": 2.9169,
      "step": 3105
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0002,
      "loss": 2.9195,
      "step": 3110
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0002,
      "loss": 2.9095,
      "step": 3115
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0002,
      "loss": 2.9367,
      "step": 3120
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0002,
      "loss": 2.9214,
      "step": 3125
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.0002,
      "loss": 2.9208,
      "step": 3130
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0002,
      "loss": 2.9269,
      "step": 3135
    },
    {
      "epoch": 1.35,
      "eval_loss": 2.918182849884033,
      "eval_runtime": 5.3357,
      "eval_samples_per_second": 1851.121,
      "eval_steps_per_second": 14.619,
      "eval_top3_3_weighted_f1_score ": 0.41396850463298607,
      "eval_top_1_macro_f1_score": 0.05368109841264443,
      "eval_top_1_weighted_f1score": 0.19213040928604935,
      "eval_top_3_macro_f1_score": 0.13911494630172772,
      "step": 3136
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0002,
      "loss": 2.9295,
      "step": 3140
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0002,
      "loss": 2.9234,
      "step": 3145
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0002,
      "loss": 2.9329,
      "step": 3150
    },
    {
      "epoch": 1.35,
      "learning_rate": 0.0002,
      "loss": 2.9057,
      "step": 3155
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.0002,
      "loss": 2.919,
      "step": 3160
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.0002,
      "loss": 2.9219,
      "step": 3165
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.0002,
      "loss": 2.9228,
      "step": 3170
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.0002,
      "loss": 2.9145,
      "step": 3175
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0002,
      "loss": 2.9346,
      "step": 3180
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0002,
      "loss": 2.9324,
      "step": 3185
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0002,
      "loss": 2.919,
      "step": 3190
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0002,
      "loss": 2.9182,
      "step": 3195
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0002,
      "loss": 2.9074,
      "step": 3200
    },
    {
      "epoch": 1.37,
      "eval_loss": 2.9170238971710205,
      "eval_runtime": 169.2426,
      "eval_samples_per_second": 58.36,
      "eval_steps_per_second": 0.461,
      "eval_top3_3_weighted_f1_score ": 0.41064106083390567,
      "eval_top_1_macro_f1_score": 0.049563680065994195,
      "eval_top_1_weighted_f1score": 0.1943582042639729,
      "eval_top_3_macro_f1_score": 0.13638503687999215,
      "step": 3200
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0002,
      "loss": 2.92,
      "step": 3205
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0002,
      "loss": 2.9165,
      "step": 3210
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0002,
      "loss": 2.9199,
      "step": 3215
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0002,
      "loss": 2.8932,
      "step": 3220
    },
    {
      "epoch": 1.38,
      "learning_rate": 0.0002,
      "loss": 2.9242,
      "step": 3225
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0002,
      "loss": 2.9407,
      "step": 3230
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0002,
      "loss": 2.9265,
      "step": 3235
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0002,
      "loss": 2.9165,
      "step": 3240
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0002,
      "loss": 2.9259,
      "step": 3245
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0002,
      "loss": 2.9007,
      "step": 3250
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0002,
      "loss": 2.9461,
      "step": 3255
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0002,
      "loss": 2.9216,
      "step": 3260
    },
    {
      "epoch": 1.4,
      "eval_loss": 2.9117391109466553,
      "eval_runtime": 5.3119,
      "eval_samples_per_second": 1859.4,
      "eval_steps_per_second": 14.684,
      "eval_top3_3_weighted_f1_score ": 0.41026852363425426,
      "eval_top_1_macro_f1_score": 0.05343009155164555,
      "eval_top_1_weighted_f1score": 0.1928369190755077,
      "eval_top_3_macro_f1_score": 0.14031584306958347,
      "step": 3264
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0002,
      "loss": 2.922,
      "step": 3265
    },
    {
      "epoch": 1.4,
      "learning_rate": 0.0002,
      "loss": 2.9084,
      "step": 3270
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0002,
      "loss": 2.9289,
      "step": 3275
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0002,
      "loss": 2.9356,
      "step": 3280
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0002,
      "loss": 2.9163,
      "step": 3285
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 3290
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.0002,
      "loss": 2.9152,
      "step": 3295
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0002,
      "loss": 2.8888,
      "step": 3300
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0002,
      "loss": 2.9314,
      "step": 3305
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0002,
      "loss": 2.9263,
      "step": 3310
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0002,
      "loss": 2.9225,
      "step": 3315
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 3320
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0002,
      "loss": 2.9206,
      "step": 3325
    },
    {
      "epoch": 1.43,
      "eval_loss": 2.9116451740264893,
      "eval_runtime": 5.2924,
      "eval_samples_per_second": 1866.277,
      "eval_steps_per_second": 14.738,
      "eval_top3_3_weighted_f1_score ": 0.4140711332372438,
      "eval_top_1_macro_f1_score": 0.05488237610711219,
      "eval_top_1_weighted_f1score": 0.19281961541929482,
      "eval_top_3_macro_f1_score": 0.13829503016111905,
      "step": 3328
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0002,
      "loss": 2.9357,
      "step": 3330
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0002,
      "loss": 2.9191,
      "step": 3335
    },
    {
      "epoch": 1.43,
      "learning_rate": 0.0002,
      "loss": 2.9082,
      "step": 3340
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0002,
      "loss": 2.9097,
      "step": 3345
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0002,
      "loss": 2.9385,
      "step": 3350
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0002,
      "loss": 2.9219,
      "step": 3355
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0002,
      "loss": 2.9229,
      "step": 3360
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0002,
      "loss": 2.9249,
      "step": 3365
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.0002,
      "loss": 2.9073,
      "step": 3370
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.0002,
      "loss": 2.9148,
      "step": 3375
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.0002,
      "loss": 2.9249,
      "step": 3380
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.0002,
      "loss": 2.9124,
      "step": 3385
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0002,
      "loss": 2.9198,
      "step": 3390
    },
    {
      "epoch": 1.46,
      "eval_loss": 2.914485454559326,
      "eval_runtime": 5.2782,
      "eval_samples_per_second": 1871.297,
      "eval_steps_per_second": 14.778,
      "eval_top3_3_weighted_f1_score ": 0.41599164658249094,
      "eval_top_1_macro_f1_score": 0.057787282618300566,
      "eval_top_1_weighted_f1score": 0.19886535679954948,
      "eval_top_3_macro_f1_score": 0.14757469973464918,
      "step": 3392
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0002,
      "loss": 2.9276,
      "step": 3395
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0002,
      "loss": 2.9286,
      "step": 3400
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0002,
      "loss": 2.9303,
      "step": 3405
    },
    {
      "epoch": 1.46,
      "learning_rate": 0.0002,
      "loss": 2.9116,
      "step": 3410
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0002,
      "loss": 2.9331,
      "step": 3415
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0002,
      "loss": 2.9166,
      "step": 3420
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0002,
      "loss": 2.9202,
      "step": 3425
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0002,
      "loss": 2.9077,
      "step": 3430
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.0002,
      "loss": 2.9384,
      "step": 3435
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0002,
      "loss": 2.9223,
      "step": 3440
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0002,
      "loss": 2.9459,
      "step": 3445
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0002,
      "loss": 2.8982,
      "step": 3450
    },
    {
      "epoch": 1.48,
      "learning_rate": 0.0002,
      "loss": 2.9291,
      "step": 3455
    },
    {
      "epoch": 1.48,
      "eval_loss": 2.9148874282836914,
      "eval_runtime": 5.3031,
      "eval_samples_per_second": 1862.488,
      "eval_steps_per_second": 14.708,
      "eval_top3_3_weighted_f1_score ": 0.41277012168510674,
      "eval_top_1_macro_f1_score": 0.050621606205600723,
      "eval_top_1_weighted_f1score": 0.19398352345475858,
      "eval_top_3_macro_f1_score": 0.1402377319399804,
      "step": 3456
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0002,
      "loss": 2.9347,
      "step": 3460
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0002,
      "loss": 2.9317,
      "step": 3465
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0002,
      "loss": 2.9001,
      "step": 3470
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0002,
      "loss": 2.9291,
      "step": 3475
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.0002,
      "loss": 2.9027,
      "step": 3480
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0002,
      "loss": 2.9321,
      "step": 3485
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0002,
      "loss": 2.9156,
      "step": 3490
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0002,
      "loss": 2.9191,
      "step": 3495
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0002,
      "loss": 2.8953,
      "step": 3500
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.0002,
      "loss": 2.9173,
      "step": 3505
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0002,
      "loss": 2.9088,
      "step": 3510
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0002,
      "loss": 2.902,
      "step": 3515
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0002,
      "loss": 2.9155,
      "step": 3520
    },
    {
      "epoch": 1.51,
      "eval_loss": 2.9106876850128174,
      "eval_runtime": 5.3153,
      "eval_samples_per_second": 1858.229,
      "eval_steps_per_second": 14.675,
      "eval_top3_3_weighted_f1_score ": 0.4138086157567288,
      "eval_top_1_macro_f1_score": 0.05603082682374438,
      "eval_top_1_weighted_f1score": 0.19434091264811978,
      "eval_top_3_macro_f1_score": 0.14252124615802533,
      "step": 3520
    },
    {
      "epoch": 1.51,
      "learning_rate": 0.0002,
      "loss": 2.9075,
      "step": 3525
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0002,
      "loss": 2.9301,
      "step": 3530
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0002,
      "loss": 2.9149,
      "step": 3535
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0002,
      "loss": 2.9142,
      "step": 3540
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0002,
      "loss": 2.9143,
      "step": 3545
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 3550
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0002,
      "loss": 2.9133,
      "step": 3555
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0002,
      "loss": 2.9394,
      "step": 3560
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0002,
      "loss": 2.9279,
      "step": 3565
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0002,
      "loss": 2.9155,
      "step": 3570
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0002,
      "loss": 2.8947,
      "step": 3575
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0002,
      "loss": 2.9055,
      "step": 3580
    },
    {
      "epoch": 1.54,
      "eval_loss": 2.9097824096679688,
      "eval_runtime": 5.2753,
      "eval_samples_per_second": 1872.293,
      "eval_steps_per_second": 14.786,
      "eval_top3_3_weighted_f1_score ": 0.4117600222372489,
      "eval_top_1_macro_f1_score": 0.05823863667346763,
      "eval_top_1_weighted_f1score": 0.19705568177663993,
      "eval_top_3_macro_f1_score": 0.1393118750761268,
      "step": 3584
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0002,
      "loss": 2.9048,
      "step": 3585
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0002,
      "loss": 2.9182,
      "step": 3590
    },
    {
      "epoch": 1.54,
      "learning_rate": 0.0002,
      "loss": 2.9238,
      "step": 3595
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0002,
      "loss": 2.9116,
      "step": 3600
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0002,
      "loss": 2.9165,
      "step": 3605
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0002,
      "loss": 2.9338,
      "step": 3610
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0002,
      "loss": 2.9333,
      "step": 3615
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0002,
      "loss": 2.9424,
      "step": 3620
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0002,
      "loss": 2.9053,
      "step": 3625
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0002,
      "loss": 2.9238,
      "step": 3630
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0002,
      "loss": 2.9277,
      "step": 3635
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0002,
      "loss": 2.9378,
      "step": 3640
    },
    {
      "epoch": 1.56,
      "learning_rate": 0.0002,
      "loss": 2.9047,
      "step": 3645
    },
    {
      "epoch": 1.57,
      "eval_loss": 2.9116811752319336,
      "eval_runtime": 5.2902,
      "eval_samples_per_second": 1867.023,
      "eval_steps_per_second": 14.744,
      "eval_top3_3_weighted_f1_score ": 0.41820266362901887,
      "eval_top_1_macro_f1_score": 0.058255851936218704,
      "eval_top_1_weighted_f1score": 0.20220022982345096,
      "eval_top_3_macro_f1_score": 0.14668049823990248,
      "step": 3648
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0002,
      "loss": 2.9164,
      "step": 3650
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0002,
      "loss": 2.9039,
      "step": 3655
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0002,
      "loss": 2.8993,
      "step": 3660
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0002,
      "loss": 2.9134,
      "step": 3665
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0002,
      "loss": 2.9097,
      "step": 3670
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0002,
      "loss": 2.9115,
      "step": 3675
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0002,
      "loss": 2.9007,
      "step": 3680
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0002,
      "loss": 2.9212,
      "step": 3685
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.0002,
      "loss": 2.9029,
      "step": 3690
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0002,
      "loss": 2.9087,
      "step": 3695
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0002,
      "loss": 2.9374,
      "step": 3700
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0002,
      "loss": 2.9146,
      "step": 3705
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0002,
      "loss": 2.9014,
      "step": 3710
    },
    {
      "epoch": 1.59,
      "eval_loss": 2.9136781692504883,
      "eval_runtime": 5.3288,
      "eval_samples_per_second": 1853.506,
      "eval_steps_per_second": 14.637,
      "eval_top3_3_weighted_f1_score ": 0.4111978805220285,
      "eval_top_1_macro_f1_score": 0.057007157073673036,
      "eval_top_1_weighted_f1score": 0.1984773001189914,
      "eval_top_3_macro_f1_score": 0.1429177656884473,
      "step": 3712
    },
    {
      "epoch": 1.59,
      "learning_rate": 0.0002,
      "loss": 2.9181,
      "step": 3715
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.0002,
      "loss": 2.8939,
      "step": 3720
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.0002,
      "loss": 2.9005,
      "step": 3725
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.0002,
      "loss": 2.9201,
      "step": 3730
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.0002,
      "loss": 2.9124,
      "step": 3735
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0002,
      "loss": 2.9138,
      "step": 3740
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0002,
      "loss": 2.9055,
      "step": 3745
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0002,
      "loss": 2.9123,
      "step": 3750
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0002,
      "loss": 2.9223,
      "step": 3755
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.0002,
      "loss": 2.9131,
      "step": 3760
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.0002,
      "loss": 2.9128,
      "step": 3765
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.0002,
      "loss": 2.9017,
      "step": 3770
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.0002,
      "loss": 2.914,
      "step": 3775
    },
    {
      "epoch": 1.62,
      "eval_loss": 2.9097282886505127,
      "eval_runtime": 5.359,
      "eval_samples_per_second": 1843.064,
      "eval_steps_per_second": 14.555,
      "eval_top3_3_weighted_f1_score ": 0.4174773301607924,
      "eval_top_1_macro_f1_score": 0.05349445280856316,
      "eval_top_1_weighted_f1score": 0.1935677243995519,
      "eval_top_3_macro_f1_score": 0.14339050964397831,
      "step": 3776
    },
    {
      "epoch": 1.62,
      "learning_rate": 0.0002,
      "loss": 2.9215,
      "step": 3780
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0002,
      "loss": 2.9247,
      "step": 3785
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0002,
      "loss": 2.9118,
      "step": 3790
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 3795
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0002,
      "loss": 2.9191,
      "step": 3800
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.0002,
      "loss": 2.9125,
      "step": 3805
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0002,
      "loss": 2.896,
      "step": 3810
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0002,
      "loss": 2.9089,
      "step": 3815
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0002,
      "loss": 2.9144,
      "step": 3820
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0002,
      "loss": 2.9069,
      "step": 3825
    },
    {
      "epoch": 1.64,
      "learning_rate": 0.0002,
      "loss": 2.9029,
      "step": 3830
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.0002,
      "loss": 2.8979,
      "step": 3835
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.0002,
      "loss": 2.9067,
      "step": 3840
    },
    {
      "epoch": 1.65,
      "eval_loss": 2.9060206413269043,
      "eval_runtime": 5.3734,
      "eval_samples_per_second": 1838.137,
      "eval_steps_per_second": 14.516,
      "eval_top3_3_weighted_f1_score ": 0.4185748244118136,
      "eval_top_1_macro_f1_score": 0.05513820508046533,
      "eval_top_1_weighted_f1score": 0.19501145432684122,
      "eval_top_3_macro_f1_score": 0.14302491472446657,
      "step": 3840
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.0002,
      "loss": 2.9024,
      "step": 3845
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.0002,
      "loss": 2.9259,
      "step": 3850
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0002,
      "loss": 2.915,
      "step": 3855
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0002,
      "loss": 2.8975,
      "step": 3860
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0002,
      "loss": 2.9105,
      "step": 3865
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0002,
      "loss": 2.9098,
      "step": 3870
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.0002,
      "loss": 2.916,
      "step": 3875
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.0002,
      "loss": 2.9133,
      "step": 3880
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.0002,
      "loss": 2.9186,
      "step": 3885
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.0002,
      "loss": 2.9075,
      "step": 3890
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.0002,
      "loss": 2.9257,
      "step": 3895
    },
    {
      "epoch": 1.67,
      "learning_rate": 0.0002,
      "loss": 2.9012,
      "step": 3900
    },
    {
      "epoch": 1.68,
      "eval_loss": 2.909200668334961,
      "eval_runtime": 5.3297,
      "eval_samples_per_second": 1853.197,
      "eval_steps_per_second": 14.635,
      "eval_top3_3_weighted_f1_score ": 0.4143734814141679,
      "eval_top_1_macro_f1_score": 0.0537386594991731,
      "eval_top_1_weighted_f1score": 0.19279945061011616,
      "eval_top_3_macro_f1_score": 0.14795670029194763,
      "step": 3904
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.0002,
      "loss": 2.9048,
      "step": 3905
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.0002,
      "loss": 2.915,
      "step": 3910
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.0002,
      "loss": 2.914,
      "step": 3915
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.0002,
      "loss": 2.9222,
      "step": 3920
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0002,
      "loss": 2.9121,
      "step": 3925
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0002,
      "loss": 2.9194,
      "step": 3930
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0002,
      "loss": 2.9276,
      "step": 3935
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0002,
      "loss": 2.9291,
      "step": 3940
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.0002,
      "loss": 2.8963,
      "step": 3945
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0002,
      "loss": 2.9213,
      "step": 3950
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0002,
      "loss": 2.908,
      "step": 3955
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0002,
      "loss": 2.9139,
      "step": 3960
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0002,
      "loss": 2.9147,
      "step": 3965
    },
    {
      "epoch": 1.7,
      "eval_loss": 2.904391288757324,
      "eval_runtime": 5.3339,
      "eval_samples_per_second": 1851.742,
      "eval_steps_per_second": 14.623,
      "eval_top3_3_weighted_f1_score ": 0.41579837645088025,
      "eval_top_1_macro_f1_score": 0.05622983577240496,
      "eval_top_1_weighted_f1score": 0.19530923139465425,
      "eval_top_3_macro_f1_score": 0.1397693104499722,
      "step": 3968
    },
    {
      "epoch": 1.7,
      "learning_rate": 0.0002,
      "loss": 2.9195,
      "step": 3970
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0002,
      "loss": 2.9397,
      "step": 3975
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0002,
      "loss": 2.9116,
      "step": 3980
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0002,
      "loss": 2.9376,
      "step": 3985
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0002,
      "loss": 2.9237,
      "step": 3990
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0002,
      "loss": 2.9208,
      "step": 3995
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0002,
      "loss": 2.9068,
      "step": 4000
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0002,
      "loss": 2.9086,
      "step": 4005
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0002,
      "loss": 2.8963,
      "step": 4010
    },
    {
      "epoch": 1.72,
      "learning_rate": 0.0002,
      "loss": 2.8877,
      "step": 4015
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0002,
      "loss": 2.928,
      "step": 4020
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0002,
      "loss": 2.9135,
      "step": 4025
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0002,
      "loss": 2.908,
      "step": 4030
    },
    {
      "epoch": 1.73,
      "eval_loss": 2.90883207321167,
      "eval_runtime": 5.3154,
      "eval_samples_per_second": 1858.194,
      "eval_steps_per_second": 14.674,
      "eval_top3_3_weighted_f1_score ": 0.416650888423843,
      "eval_top_1_macro_f1_score": 0.05452347270506178,
      "eval_top_1_weighted_f1score": 0.1954259208665436,
      "eval_top_3_macro_f1_score": 0.14547676692371397,
      "step": 4032
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0002,
      "loss": 2.9015,
      "step": 4035
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0002,
      "loss": 2.9234,
      "step": 4040
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0002,
      "loss": 2.9126,
      "step": 4045
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0002,
      "loss": 2.905,
      "step": 4050
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0002,
      "loss": 2.9202,
      "step": 4055
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0002,
      "loss": 2.9229,
      "step": 4060
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0002,
      "loss": 2.9296,
      "step": 4065
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0002,
      "loss": 2.9249,
      "step": 4070
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0002,
      "loss": 2.9094,
      "step": 4075
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0002,
      "loss": 2.9157,
      "step": 4080
    },
    {
      "epoch": 1.75,
      "learning_rate": 0.0002,
      "loss": 2.9287,
      "step": 4085
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0002,
      "loss": 2.9212,
      "step": 4090
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0002,
      "loss": 2.8952,
      "step": 4095
    },
    {
      "epoch": 1.76,
      "eval_loss": 2.914785385131836,
      "eval_runtime": 5.3312,
      "eval_samples_per_second": 1852.677,
      "eval_steps_per_second": 14.631,
      "eval_top3_3_weighted_f1_score ": 0.4185348765090556,
      "eval_top_1_macro_f1_score": 0.05764061973952746,
      "eval_top_1_weighted_f1score": 0.19691281408017583,
      "eval_top_3_macro_f1_score": 0.1456156190028161,
      "step": 4096
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0002,
      "loss": 2.9214,
      "step": 4100
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0002,
      "loss": 2.9131,
      "step": 4105
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.0002,
      "loss": 2.9001,
      "step": 4110
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0002,
      "loss": 2.9173,
      "step": 4115
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0002,
      "loss": 2.9069,
      "step": 4120
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0002,
      "loss": 2.9344,
      "step": 4125
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.0002,
      "loss": 2.9176,
      "step": 4130
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0002,
      "loss": 2.932,
      "step": 4135
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0002,
      "loss": 2.9112,
      "step": 4140
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0002,
      "loss": 2.896,
      "step": 4145
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0002,
      "loss": 2.926,
      "step": 4150
    },
    {
      "epoch": 1.78,
      "learning_rate": 0.0002,
      "loss": 2.9119,
      "step": 4155
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0002,
      "loss": 2.9204,
      "step": 4160
    },
    {
      "epoch": 1.79,
      "eval_loss": 2.9073331356048584,
      "eval_runtime": 5.3361,
      "eval_samples_per_second": 1850.992,
      "eval_steps_per_second": 14.618,
      "eval_top3_3_weighted_f1_score ": 0.4166796433939981,
      "eval_top_1_macro_f1_score": 0.05722567454289957,
      "eval_top_1_weighted_f1score": 0.19412432161186186,
      "eval_top_3_macro_f1_score": 0.14723111807796124,
      "step": 4160
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0002,
      "loss": 2.8993,
      "step": 4165
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0002,
      "loss": 2.9186,
      "step": 4170
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0002,
      "loss": 2.9047,
      "step": 4175
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.0002,
      "loss": 2.9204,
      "step": 4180
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.0002,
      "loss": 2.908,
      "step": 4185
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.0002,
      "loss": 2.9282,
      "step": 4190
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.0002,
      "loss": 2.9121,
      "step": 4195
    },
    {
      "epoch": 1.8,
      "learning_rate": 0.0002,
      "loss": 2.9177,
      "step": 4200
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 4205
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0002,
      "loss": 2.9246,
      "step": 4210
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0002,
      "loss": 2.9083,
      "step": 4215
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0002,
      "loss": 2.9188,
      "step": 4220
    },
    {
      "epoch": 1.81,
      "eval_loss": 2.9065234661102295,
      "eval_runtime": 5.3649,
      "eval_samples_per_second": 1841.057,
      "eval_steps_per_second": 14.539,
      "eval_top3_3_weighted_f1_score ": 0.41645565051543537,
      "eval_top_1_macro_f1_score": 0.056695468234468124,
      "eval_top_1_weighted_f1score": 0.1949375757721395,
      "eval_top_3_macro_f1_score": 0.14736203326563801,
      "step": 4224
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.0002,
      "loss": 2.924,
      "step": 4225
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.0002,
      "loss": 2.9285,
      "step": 4230
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.0002,
      "loss": 2.9264,
      "step": 4235
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.0002,
      "loss": 2.9202,
      "step": 4240
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.0002,
      "loss": 2.9263,
      "step": 4245
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.0002,
      "loss": 2.923,
      "step": 4250
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.0002,
      "loss": 2.9035,
      "step": 4255
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 4260
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.0002,
      "loss": 2.9267,
      "step": 4265
    },
    {
      "epoch": 1.83,
      "learning_rate": 0.0002,
      "loss": 2.9168,
      "step": 4270
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0002,
      "loss": 2.9215,
      "step": 4275
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0002,
      "loss": 2.8942,
      "step": 4280
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0002,
      "loss": 2.9227,
      "step": 4285
    },
    {
      "epoch": 1.84,
      "eval_loss": 2.9160542488098145,
      "eval_runtime": 5.2816,
      "eval_samples_per_second": 1870.065,
      "eval_steps_per_second": 14.768,
      "eval_top3_3_weighted_f1_score ": 0.4112509098612428,
      "eval_top_1_macro_f1_score": 0.05835205008822769,
      "eval_top_1_weighted_f1score": 0.19920753039269692,
      "eval_top_3_macro_f1_score": 0.13974934867905217,
      "step": 4288
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0002,
      "loss": 2.9039,
      "step": 4290
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.0002,
      "loss": 2.8846,
      "step": 4295
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0002,
      "loss": 2.8988,
      "step": 4300
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0002,
      "loss": 2.9182,
      "step": 4305
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0002,
      "loss": 2.9194,
      "step": 4310
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0002,
      "loss": 2.9185,
      "step": 4315
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.0002,
      "loss": 2.92,
      "step": 4320
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0002,
      "loss": 2.907,
      "step": 4325
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0002,
      "loss": 2.9049,
      "step": 4330
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0002,
      "loss": 2.9168,
      "step": 4335
    },
    {
      "epoch": 1.86,
      "learning_rate": 0.0002,
      "loss": 2.9012,
      "step": 4340
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0002,
      "loss": 2.9008,
      "step": 4345
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0002,
      "loss": 2.9114,
      "step": 4350
    },
    {
      "epoch": 1.87,
      "eval_loss": 2.9069080352783203,
      "eval_runtime": 5.2449,
      "eval_samples_per_second": 1883.177,
      "eval_steps_per_second": 14.872,
      "eval_top3_3_weighted_f1_score ": 0.4131753084295918,
      "eval_top_1_macro_f1_score": 0.06237738522108177,
      "eval_top_1_weighted_f1score": 0.20188181913152448,
      "eval_top_3_macro_f1_score": 0.1425914449331181,
      "step": 4352
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0002,
      "loss": 2.9061,
      "step": 4355
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0002,
      "loss": 2.9113,
      "step": 4360
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 4365
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0002,
      "loss": 2.9297,
      "step": 4370
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0002,
      "loss": 2.8953,
      "step": 4375
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0002,
      "loss": 2.9105,
      "step": 4380
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0002,
      "loss": 2.9281,
      "step": 4385
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0002,
      "loss": 2.904,
      "step": 4390
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.0002,
      "loss": 2.9272,
      "step": 4395
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.0002,
      "loss": 2.8938,
      "step": 4400
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.0002,
      "loss": 2.9084,
      "step": 4405
    },
    {
      "epoch": 1.89,
      "learning_rate": 0.0002,
      "loss": 2.9135,
      "step": 4410
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0002,
      "loss": 2.909,
      "step": 4415
    },
    {
      "epoch": 1.9,
      "eval_loss": 2.908174514770508,
      "eval_runtime": 5.3274,
      "eval_samples_per_second": 1854.013,
      "eval_steps_per_second": 14.641,
      "eval_top3_3_weighted_f1_score ": 0.4165512120544975,
      "eval_top_1_macro_f1_score": 0.059023123561221094,
      "eval_top_1_weighted_f1score": 0.19620342691966175,
      "eval_top_3_macro_f1_score": 0.14350421344968878,
      "step": 4416
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0002,
      "loss": 2.9153,
      "step": 4420
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0002,
      "loss": 2.9189,
      "step": 4425
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0002,
      "loss": 2.9199,
      "step": 4430
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0002,
      "loss": 2.9202,
      "step": 4435
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.0002,
      "loss": 2.9097,
      "step": 4440
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.0002,
      "loss": 2.9136,
      "step": 4445
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.0002,
      "loss": 2.9187,
      "step": 4450
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.0002,
      "loss": 2.9051,
      "step": 4455
    },
    {
      "epoch": 1.91,
      "learning_rate": 0.0002,
      "loss": 2.9131,
      "step": 4460
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0002,
      "loss": 2.9089,
      "step": 4465
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0002,
      "loss": 2.8953,
      "step": 4470
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0002,
      "loss": 2.9303,
      "step": 4475
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0002,
      "loss": 2.9149,
      "step": 4480
    },
    {
      "epoch": 1.92,
      "eval_loss": 2.9029014110565186,
      "eval_runtime": 5.3399,
      "eval_samples_per_second": 1849.662,
      "eval_steps_per_second": 14.607,
      "eval_top3_3_weighted_f1_score ": 0.41880335718795425,
      "eval_top_1_macro_f1_score": 0.05744559670095288,
      "eval_top_1_weighted_f1score": 0.1989887507512129,
      "eval_top_3_macro_f1_score": 0.14449459246824883,
      "step": 4480
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0002,
      "loss": 2.9025,
      "step": 4485
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0002,
      "loss": 2.9341,
      "step": 4490
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0002,
      "loss": 2.9269,
      "step": 4495
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0002,
      "loss": 2.8972,
      "step": 4500
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.0002,
      "loss": 2.8998,
      "step": 4505
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0002,
      "loss": 2.9039,
      "step": 4510
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0002,
      "loss": 2.9145,
      "step": 4515
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0002,
      "loss": 2.9066,
      "step": 4520
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0002,
      "loss": 2.9148,
      "step": 4525
    },
    {
      "epoch": 1.94,
      "learning_rate": 0.0002,
      "loss": 2.9176,
      "step": 4530
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.0002,
      "loss": 2.8787,
      "step": 4535
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.0002,
      "loss": 2.9088,
      "step": 4540
    },
    {
      "epoch": 1.95,
      "eval_loss": 2.905123233795166,
      "eval_runtime": 5.3029,
      "eval_samples_per_second": 1862.574,
      "eval_steps_per_second": 14.709,
      "eval_top3_3_weighted_f1_score ": 0.41605559730748176,
      "eval_top_1_macro_f1_score": 0.05698845252098233,
      "eval_top_1_weighted_f1score": 0.19595302817016025,
      "eval_top_3_macro_f1_score": 0.1486403646875358,
      "step": 4544
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.0002,
      "loss": 2.9087,
      "step": 4545
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.0002,
      "loss": 2.9306,
      "step": 4550
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0002,
      "loss": 2.9084,
      "step": 4555
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0002,
      "loss": 2.9113,
      "step": 4560
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0002,
      "loss": 2.9034,
      "step": 4565
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0002,
      "loss": 2.891,
      "step": 4570
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.0002,
      "loss": 2.9184,
      "step": 4575
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0002,
      "loss": 2.9196,
      "step": 4580
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0002,
      "loss": 2.9025,
      "step": 4585
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0002,
      "loss": 2.9052,
      "step": 4590
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0002,
      "loss": 2.9056,
      "step": 4595
    },
    {
      "epoch": 1.97,
      "learning_rate": 0.0002,
      "loss": 2.9216,
      "step": 4600
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.0002,
      "loss": 2.8998,
      "step": 4605
    },
    {
      "epoch": 1.98,
      "eval_loss": 2.901381492614746,
      "eval_runtime": 5.244,
      "eval_samples_per_second": 1883.497,
      "eval_steps_per_second": 14.874,
      "eval_top3_3_weighted_f1_score ": 0.4190062695557982,
      "eval_top_1_macro_f1_score": 0.059787971401313846,
      "eval_top_1_weighted_f1score": 0.19889101422706393,
      "eval_top_3_macro_f1_score": 0.15426000566225975,
      "step": 4608
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.0002,
      "loss": 2.9209,
      "step": 4610
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.0002,
      "loss": 2.9142,
      "step": 4615
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.0002,
      "loss": 2.9056,
      "step": 4620
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0002,
      "loss": 2.9182,
      "step": 4625
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0002,
      "loss": 2.9028,
      "step": 4630
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0002,
      "loss": 2.9051,
      "step": 4635
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0002,
      "loss": 2.9102,
      "step": 4640
    },
    {
      "epoch": 1.99,
      "learning_rate": 0.0002,
      "loss": 2.9053,
      "step": 4645
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0002,
      "loss": 2.9247,
      "step": 4650
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0002,
      "loss": 2.9123,
      "step": 4655
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0002,
      "loss": 3.1869,
      "step": 4660
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0002,
      "loss": 2.897,
      "step": 4665
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0002,
      "loss": 2.8913,
      "step": 4670
    },
    {
      "epoch": 2.01,
      "eval_loss": 2.902728319168091,
      "eval_runtime": 5.2917,
      "eval_samples_per_second": 1866.498,
      "eval_steps_per_second": 14.74,
      "eval_top3_3_weighted_f1_score ": 0.4154482191451159,
      "eval_top_1_macro_f1_score": 0.05910348194925984,
      "eval_top_1_weighted_f1score": 0.20223672449236257,
      "eval_top_3_macro_f1_score": 0.14626048197345107,
      "step": 4672
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0002,
      "loss": 2.9062,
      "step": 4675
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0002,
      "loss": 2.9174,
      "step": 4680
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0002,
      "loss": 2.8912,
      "step": 4685
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.0002,
      "loss": 2.8863,
      "step": 4690
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0002,
      "loss": 2.898,
      "step": 4695
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0002,
      "loss": 2.8964,
      "step": 4700
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0002,
      "loss": 2.9035,
      "step": 4705
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0002,
      "loss": 2.9061,
      "step": 4710
    },
    {
      "epoch": 2.02,
      "learning_rate": 0.0002,
      "loss": 2.9078,
      "step": 4715
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0002,
      "loss": 2.9121,
      "step": 4720
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0002,
      "loss": 2.9054,
      "step": 4725
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0002,
      "loss": 2.9076,
      "step": 4730
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.0002,
      "loss": 2.9018,
      "step": 4735
    },
    {
      "epoch": 2.03,
      "eval_loss": 2.903496503829956,
      "eval_runtime": 5.3745,
      "eval_samples_per_second": 1837.739,
      "eval_steps_per_second": 14.513,
      "eval_top3_3_weighted_f1_score ": 0.4163058624724811,
      "eval_top_1_macro_f1_score": 0.05647649082182299,
      "eval_top_1_weighted_f1score": 0.19924151899797926,
      "eval_top_3_macro_f1_score": 0.1405979459406439,
      "step": 4736
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.0002,
      "loss": 2.9071,
      "step": 4740
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.0002,
      "loss": 2.9004,
      "step": 4745
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.0002,
      "loss": 2.9195,
      "step": 4750
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.0002,
      "loss": 2.9051,
      "step": 4755
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.0002,
      "loss": 2.8896,
      "step": 4760
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0002,
      "loss": 2.9106,
      "step": 4765
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0002,
      "loss": 2.8991,
      "step": 4770
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0002,
      "loss": 2.9019,
      "step": 4775
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0002,
      "loss": 2.8923,
      "step": 4780
    },
    {
      "epoch": 2.05,
      "learning_rate": 0.0002,
      "loss": 2.9036,
      "step": 4785
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.0002,
      "loss": 2.9053,
      "step": 4790
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.0002,
      "loss": 2.8904,
      "step": 4795
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.0002,
      "loss": 2.8997,
      "step": 4800
    },
    {
      "epoch": 2.06,
      "eval_loss": 2.909992218017578,
      "eval_runtime": 5.3247,
      "eval_samples_per_second": 1854.946,
      "eval_steps_per_second": 14.649,
      "eval_top3_3_weighted_f1_score ": 0.41880815422019485,
      "eval_top_1_macro_f1_score": 0.05827192013803967,
      "eval_top_1_weighted_f1score": 0.19750961372028833,
      "eval_top_3_macro_f1_score": 0.14495982688137427,
      "step": 4800
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.0002,
      "loss": 2.9132,
      "step": 4805
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0002,
      "loss": 2.8989,
      "step": 4810
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0002,
      "loss": 2.8843,
      "step": 4815
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0002,
      "loss": 2.8965,
      "step": 4820
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0002,
      "loss": 2.914,
      "step": 4825
    },
    {
      "epoch": 2.07,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 4830
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0002,
      "loss": 2.8981,
      "step": 4835
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0002,
      "loss": 2.9123,
      "step": 4840
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0002,
      "loss": 2.9262,
      "step": 4845
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0002,
      "loss": 2.8943,
      "step": 4850
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.0002,
      "loss": 2.9001,
      "step": 4855
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0002,
      "loss": 2.917,
      "step": 4860
    },
    {
      "epoch": 2.09,
      "eval_loss": 2.9023990631103516,
      "eval_runtime": 5.2937,
      "eval_samples_per_second": 1865.817,
      "eval_steps_per_second": 14.735,
      "eval_top3_3_weighted_f1_score ": 0.417123946923505,
      "eval_top_1_macro_f1_score": 0.05402152788915577,
      "eval_top_1_weighted_f1score": 0.19808512530123618,
      "eval_top_3_macro_f1_score": 0.1564014984350641,
      "step": 4864
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0002,
      "loss": 2.9091,
      "step": 4865
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0002,
      "loss": 2.9159,
      "step": 4870
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.0002,
      "loss": 2.9097,
      "step": 4875
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0002,
      "loss": 2.8911,
      "step": 4880
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0002,
      "loss": 2.9067,
      "step": 4885
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0002,
      "loss": 2.9007,
      "step": 4890
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0002,
      "loss": 2.8923,
      "step": 4895
    },
    {
      "epoch": 2.1,
      "learning_rate": 0.0002,
      "loss": 2.9066,
      "step": 4900
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.0002,
      "loss": 2.9114,
      "step": 4905
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.0002,
      "loss": 2.9082,
      "step": 4910
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.0002,
      "loss": 2.9178,
      "step": 4915
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.0002,
      "loss": 2.9244,
      "step": 4920
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.0002,
      "loss": 2.9009,
      "step": 4925
    },
    {
      "epoch": 2.12,
      "eval_loss": 2.8972628116607666,
      "eval_runtime": 5.3426,
      "eval_samples_per_second": 1848.731,
      "eval_steps_per_second": 14.6,
      "eval_top3_3_weighted_f1_score ": 0.4156100206555792,
      "eval_top_1_macro_f1_score": 0.05458305463384416,
      "eval_top_1_weighted_f1score": 0.19470342306471872,
      "eval_top_3_macro_f1_score": 0.14190048695385177,
      "step": 4928
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.0002,
      "loss": 2.9053,
      "step": 4930
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.0002,
      "loss": 2.8988,
      "step": 4935
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.0002,
      "loss": 2.9056,
      "step": 4940
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.0002,
      "loss": 2.915,
      "step": 4945
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.0002,
      "loss": 2.8977,
      "step": 4950
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 4955
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.0002,
      "loss": 2.8917,
      "step": 4960
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.0002,
      "loss": 2.9232,
      "step": 4965
    },
    {
      "epoch": 2.13,
      "learning_rate": 0.0002,
      "loss": 2.8904,
      "step": 4970
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0002,
      "loss": 2.8912,
      "step": 4975
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0002,
      "loss": 2.8886,
      "step": 4980
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0002,
      "loss": 2.9094,
      "step": 4985
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0002,
      "loss": 2.9144,
      "step": 4990
    },
    {
      "epoch": 2.14,
      "eval_loss": 2.9059925079345703,
      "eval_runtime": 5.3147,
      "eval_samples_per_second": 1858.424,
      "eval_steps_per_second": 14.676,
      "eval_top3_3_weighted_f1_score ": 0.41711598390450566,
      "eval_top_1_macro_f1_score": 0.055409555982394415,
      "eval_top_1_weighted_f1score": 0.19777402153595952,
      "eval_top_3_macro_f1_score": 0.1464816808654842,
      "step": 4992
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0002,
      "loss": 2.8881,
      "step": 4995
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0002,
      "loss": 2.8905,
      "step": 5000
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0002,
      "loss": 2.9009,
      "step": 5005
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0002,
      "loss": 2.9233,
      "step": 5010
    },
    {
      "epoch": 2.15,
      "learning_rate": 0.0002,
      "loss": 2.9139,
      "step": 5015
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0002,
      "loss": 2.9036,
      "step": 5020
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0002,
      "loss": 2.8985,
      "step": 5025
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0002,
      "loss": 2.8937,
      "step": 5030
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0002,
      "loss": 2.8878,
      "step": 5035
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0002,
      "loss": 2.898,
      "step": 5040
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0002,
      "loss": 2.8962,
      "step": 5045
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0002,
      "loss": 2.912,
      "step": 5050
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0002,
      "loss": 2.9028,
      "step": 5055
    },
    {
      "epoch": 2.17,
      "eval_loss": 2.898975133895874,
      "eval_runtime": 5.291,
      "eval_samples_per_second": 1866.74,
      "eval_steps_per_second": 14.742,
      "eval_top3_3_weighted_f1_score ": 0.42027650116928694,
      "eval_top_1_macro_f1_score": 0.056066861668324454,
      "eval_top_1_weighted_f1score": 0.20134187579837567,
      "eval_top_3_macro_f1_score": 0.14763695642950755,
      "step": 5056
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0002,
      "loss": 2.8855,
      "step": 5060
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 5065
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.0002,
      "loss": 2.9056,
      "step": 5070
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.0002,
      "loss": 2.9207,
      "step": 5075
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.0002,
      "loss": 2.9318,
      "step": 5080
    },
    {
      "epoch": 2.18,
      "learning_rate": 0.0002,
      "loss": 2.9121,
      "step": 5085
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.0002,
      "loss": 2.8943,
      "step": 5090
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.0002,
      "loss": 2.9054,
      "step": 5095
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.0002,
      "loss": 2.9012,
      "step": 5100
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.0002,
      "loss": 2.9149,
      "step": 5105
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 5110
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0002,
      "loss": 2.8955,
      "step": 5115
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0002,
      "loss": 2.8892,
      "step": 5120
    },
    {
      "epoch": 2.2,
      "eval_loss": 2.892855167388916,
      "eval_runtime": 5.2995,
      "eval_samples_per_second": 1863.764,
      "eval_steps_per_second": 14.718,
      "eval_top3_3_weighted_f1_score ": 0.4157954024117956,
      "eval_top_1_macro_f1_score": 0.05936697956810241,
      "eval_top_1_weighted_f1score": 0.20247573523572931,
      "eval_top_3_macro_f1_score": 0.14266020318390252,
      "step": 5120
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0002,
      "loss": 2.8818,
      "step": 5125
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0002,
      "loss": 2.9041,
      "step": 5130
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0002,
      "loss": 2.9043,
      "step": 5135
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.0002,
      "loss": 2.9055,
      "step": 5140
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.0002,
      "loss": 2.913,
      "step": 5145
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.0002,
      "loss": 2.8993,
      "step": 5150
    },
    {
      "epoch": 2.21,
      "learning_rate": 0.0002,
      "loss": 2.8972,
      "step": 5155
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0002,
      "loss": 2.8794,
      "step": 5160
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0002,
      "loss": 2.902,
      "step": 5165
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0002,
      "loss": 2.8963,
      "step": 5170
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 5175
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.0002,
      "loss": 2.8859,
      "step": 5180
    },
    {
      "epoch": 2.23,
      "eval_loss": 2.89302921295166,
      "eval_runtime": 5.396,
      "eval_samples_per_second": 1830.429,
      "eval_steps_per_second": 14.455,
      "eval_top3_3_weighted_f1_score ": 0.41303606443703067,
      "eval_top_1_macro_f1_score": 0.05968216401094024,
      "eval_top_1_weighted_f1score": 0.20331060932366576,
      "eval_top_3_macro_f1_score": 0.14440810187347125,
      "step": 5184
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 5185
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0002,
      "loss": 2.9068,
      "step": 5190
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0002,
      "loss": 2.8954,
      "step": 5195
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0002,
      "loss": 2.9057,
      "step": 5200
    },
    {
      "epoch": 2.23,
      "learning_rate": 0.0002,
      "loss": 2.8911,
      "step": 5205
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.0002,
      "loss": 2.9077,
      "step": 5210
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.0002,
      "loss": 2.9102,
      "step": 5215
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.0002,
      "loss": 2.8988,
      "step": 5220
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.0002,
      "loss": 2.8987,
      "step": 5225
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0002,
      "loss": 2.9164,
      "step": 5230
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0002,
      "loss": 2.8974,
      "step": 5235
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0002,
      "loss": 2.8952,
      "step": 5240
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0002,
      "loss": 2.9294,
      "step": 5245
    },
    {
      "epoch": 2.25,
      "eval_loss": 2.899155616760254,
      "eval_runtime": 5.3508,
      "eval_samples_per_second": 1845.891,
      "eval_steps_per_second": 14.577,
      "eval_top3_3_weighted_f1_score ": 0.42197230106720923,
      "eval_top_1_macro_f1_score": 0.06035559277068504,
      "eval_top_1_weighted_f1score": 0.19955784494082673,
      "eval_top_3_macro_f1_score": 0.14700216854763365,
      "step": 5248
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.0002,
      "loss": 2.9087,
      "step": 5250
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.0002,
      "loss": 2.8955,
      "step": 5255
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.0002,
      "loss": 2.8916,
      "step": 5260
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.0002,
      "loss": 2.8943,
      "step": 5265
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.0002,
      "loss": 2.903,
      "step": 5270
    },
    {
      "epoch": 2.26,
      "learning_rate": 0.0002,
      "loss": 2.8978,
      "step": 5275
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.0002,
      "loss": 2.91,
      "step": 5280
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.0002,
      "loss": 2.9157,
      "step": 5285
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.0002,
      "loss": 2.9086,
      "step": 5290
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.0002,
      "loss": 2.9067,
      "step": 5295
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0002,
      "loss": 2.9071,
      "step": 5300
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0002,
      "loss": 2.913,
      "step": 5305
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0002,
      "loss": 2.8918,
      "step": 5310
    },
    {
      "epoch": 2.28,
      "eval_loss": 2.8990283012390137,
      "eval_runtime": 5.29,
      "eval_samples_per_second": 1867.109,
      "eval_steps_per_second": 14.745,
      "eval_top3_3_weighted_f1_score ": 0.420600106378898,
      "eval_top_1_macro_f1_score": 0.05742805002784499,
      "eval_top_1_weighted_f1score": 0.19965523388204837,
      "eval_top_3_macro_f1_score": 0.1505476498909347,
      "step": 5312
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0002,
      "loss": 2.9106,
      "step": 5315
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.0002,
      "loss": 2.8789,
      "step": 5320
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0002,
      "loss": 2.8928,
      "step": 5325
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0002,
      "loss": 2.8814,
      "step": 5330
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0002,
      "loss": 2.8978,
      "step": 5335
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0002,
      "loss": 2.875,
      "step": 5340
    },
    {
      "epoch": 2.29,
      "learning_rate": 0.0002,
      "loss": 2.9026,
      "step": 5345
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0002,
      "loss": 2.8901,
      "step": 5350
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0002,
      "loss": 2.9217,
      "step": 5355
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0002,
      "loss": 2.914,
      "step": 5360
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.0002,
      "loss": 2.8997,
      "step": 5365
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0002,
      "loss": 2.901,
      "step": 5370
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0002,
      "loss": 2.8975,
      "step": 5375
    },
    {
      "epoch": 2.31,
      "eval_loss": 2.9051334857940674,
      "eval_runtime": 5.3756,
      "eval_samples_per_second": 1837.369,
      "eval_steps_per_second": 14.51,
      "eval_top3_3_weighted_f1_score ": 0.4154097299787692,
      "eval_top_1_macro_f1_score": 0.0580655682817357,
      "eval_top_1_weighted_f1score": 0.19945941252814092,
      "eval_top_3_macro_f1_score": 0.14492591699598248,
      "step": 5376
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0002,
      "loss": 2.8853,
      "step": 5380
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0002,
      "loss": 2.9137,
      "step": 5385
    },
    {
      "epoch": 2.31,
      "learning_rate": 0.0002,
      "loss": 2.9018,
      "step": 5390
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0002,
      "loss": 2.9064,
      "step": 5395
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0002,
      "loss": 2.8952,
      "step": 5400
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0002,
      "loss": 2.906,
      "step": 5405
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0002,
      "loss": 2.8936,
      "step": 5410
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0002,
      "loss": 2.8944,
      "step": 5415
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.0002,
      "loss": 2.8801,
      "step": 5420
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.0002,
      "loss": 2.93,
      "step": 5425
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.0002,
      "loss": 2.8827,
      "step": 5430
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.0002,
      "loss": 2.8971,
      "step": 5435
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 5440
    },
    {
      "epoch": 2.34,
      "eval_loss": 2.9040162563323975,
      "eval_runtime": 5.444,
      "eval_samples_per_second": 1814.282,
      "eval_steps_per_second": 14.328,
      "eval_top3_3_weighted_f1_score ": 0.41358052554168034,
      "eval_top_1_macro_f1_score": 0.05707689836126898,
      "eval_top_1_weighted_f1score": 0.19804938443363193,
      "eval_top_3_macro_f1_score": 0.14201305208124318,
      "step": 5440
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0002,
      "loss": 2.9092,
      "step": 5445
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 5450
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0002,
      "loss": 2.9152,
      "step": 5455
    },
    {
      "epoch": 2.34,
      "learning_rate": 0.0002,
      "loss": 2.9021,
      "step": 5460
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0002,
      "loss": 2.9093,
      "step": 5465
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0002,
      "loss": 2.8977,
      "step": 5470
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0002,
      "loss": 2.8999,
      "step": 5475
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0002,
      "loss": 2.9113,
      "step": 5480
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0002,
      "loss": 2.8996,
      "step": 5485
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 5490
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 5495
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0002,
      "loss": 2.9253,
      "step": 5500
    },
    {
      "epoch": 2.36,
      "eval_loss": 2.8994529247283936,
      "eval_runtime": 5.3486,
      "eval_samples_per_second": 1846.655,
      "eval_steps_per_second": 14.583,
      "eval_top3_3_weighted_f1_score ": 0.41753233385444316,
      "eval_top_1_macro_f1_score": 0.05349162501424304,
      "eval_top_1_weighted_f1score": 0.19682846206528506,
      "eval_top_3_macro_f1_score": 0.15138121336524843,
      "step": 5504
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.0002,
      "loss": 2.8881,
      "step": 5505
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0002,
      "loss": 2.8859,
      "step": 5510
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0002,
      "loss": 2.9088,
      "step": 5515
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0002,
      "loss": 2.9117,
      "step": 5520
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0002,
      "loss": 2.9069,
      "step": 5525
    },
    {
      "epoch": 2.37,
      "learning_rate": 0.0002,
      "loss": 2.9107,
      "step": 5530
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0002,
      "loss": 2.9007,
      "step": 5535
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0002,
      "loss": 2.9023,
      "step": 5540
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0002,
      "loss": 2.8929,
      "step": 5545
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.0002,
      "loss": 2.8813,
      "step": 5550
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0002,
      "loss": 2.8936,
      "step": 5555
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0002,
      "loss": 2.8711,
      "step": 5560
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0002,
      "loss": 2.9015,
      "step": 5565
    },
    {
      "epoch": 2.39,
      "eval_loss": 2.900458812713623,
      "eval_runtime": 5.326,
      "eval_samples_per_second": 1854.48,
      "eval_steps_per_second": 14.645,
      "eval_top3_3_weighted_f1_score ": 0.41880679197562753,
      "eval_top_1_macro_f1_score": 0.05999635911721816,
      "eval_top_1_weighted_f1score": 0.19856166956788507,
      "eval_top_3_macro_f1_score": 0.1492237171872263,
      "step": 5568
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0002,
      "loss": 2.9005,
      "step": 5570
    },
    {
      "epoch": 2.39,
      "learning_rate": 0.0002,
      "loss": 2.884,
      "step": 5575
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.0002,
      "loss": 2.9037,
      "step": 5580
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.0002,
      "loss": 2.9133,
      "step": 5585
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.0002,
      "loss": 2.8886,
      "step": 5590
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.0002,
      "loss": 2.9296,
      "step": 5595
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.0002,
      "loss": 2.9045,
      "step": 5600
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.0002,
      "loss": 2.8971,
      "step": 5605
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.0002,
      "loss": 2.9326,
      "step": 5610
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.0002,
      "loss": 2.8866,
      "step": 5615
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.0002,
      "loss": 2.9046,
      "step": 5620
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.0002,
      "loss": 2.8954,
      "step": 5625
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.0002,
      "loss": 2.8775,
      "step": 5630
    },
    {
      "epoch": 2.42,
      "eval_loss": 2.9008569717407227,
      "eval_runtime": 5.377,
      "eval_samples_per_second": 1836.901,
      "eval_steps_per_second": 14.506,
      "eval_top3_3_weighted_f1_score ": 0.4188094807665909,
      "eval_top_1_macro_f1_score": 0.056329276543405866,
      "eval_top_1_weighted_f1score": 0.19788854247420265,
      "eval_top_3_macro_f1_score": 0.15469545654298356,
      "step": 5632
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.0002,
      "loss": 2.892,
      "step": 5635
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.0002,
      "loss": 2.8939,
      "step": 5640
    },
    {
      "epoch": 2.42,
      "learning_rate": 0.0002,
      "loss": 2.8942,
      "step": 5645
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0002,
      "loss": 2.8904,
      "step": 5650
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0002,
      "loss": 2.9093,
      "step": 5655
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0002,
      "loss": 2.9126,
      "step": 5660
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0002,
      "loss": 2.9113,
      "step": 5665
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.0002,
      "loss": 2.9051,
      "step": 5670
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.0002,
      "loss": 2.8914,
      "step": 5675
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.0002,
      "loss": 2.8887,
      "step": 5680
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.0002,
      "loss": 2.8919,
      "step": 5685
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.0002,
      "loss": 2.8829,
      "step": 5690
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0002,
      "loss": 2.9113,
      "step": 5695
    },
    {
      "epoch": 2.45,
      "eval_loss": 2.8912224769592285,
      "eval_runtime": 5.2948,
      "eval_samples_per_second": 1865.409,
      "eval_steps_per_second": 14.731,
      "eval_top3_3_weighted_f1_score ": 0.42107093582619226,
      "eval_top_1_macro_f1_score": 0.06742872368412635,
      "eval_top_1_weighted_f1score": 0.20246055579316635,
      "eval_top_3_macro_f1_score": 0.15704454539282703,
      "step": 5696
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0002,
      "loss": 2.8917,
      "step": 5700
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0002,
      "loss": 2.8794,
      "step": 5705
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0002,
      "loss": 2.9101,
      "step": 5710
    },
    {
      "epoch": 2.45,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 5715
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.0002,
      "loss": 2.9059,
      "step": 5720
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.0002,
      "loss": 2.8941,
      "step": 5725
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.0002,
      "loss": 2.8942,
      "step": 5730
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.0002,
      "loss": 2.8922,
      "step": 5735
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.0002,
      "loss": 2.9017,
      "step": 5740
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0002,
      "loss": 2.9096,
      "step": 5745
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0002,
      "loss": 2.8884,
      "step": 5750
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0002,
      "loss": 2.9138,
      "step": 5755
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0002,
      "loss": 2.895,
      "step": 5760
    },
    {
      "epoch": 2.47,
      "eval_loss": 2.8928873538970947,
      "eval_runtime": 5.3553,
      "eval_samples_per_second": 1844.355,
      "eval_steps_per_second": 14.565,
      "eval_top3_3_weighted_f1_score ": 0.4188124182838734,
      "eval_top_1_macro_f1_score": 0.05698105154490517,
      "eval_top_1_weighted_f1score": 0.1969784382902619,
      "eval_top_3_macro_f1_score": 0.1468489374120009,
      "step": 5760
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0002,
      "loss": 2.8968,
      "step": 5765
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0002,
      "loss": 2.8991,
      "step": 5770
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0002,
      "loss": 2.91,
      "step": 5775
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 5780
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0002,
      "loss": 2.8902,
      "step": 5785
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0002,
      "loss": 2.8994,
      "step": 5790
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0002,
      "loss": 2.9227,
      "step": 5795
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0002,
      "loss": 2.9144,
      "step": 5800
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0002,
      "loss": 2.9039,
      "step": 5805
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.0002,
      "loss": 2.8977,
      "step": 5810
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.0002,
      "loss": 2.902,
      "step": 5815
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.0002,
      "loss": 2.8849,
      "step": 5820
    },
    {
      "epoch": 2.5,
      "eval_loss": 2.8987321853637695,
      "eval_runtime": 5.3523,
      "eval_samples_per_second": 1845.359,
      "eval_steps_per_second": 14.573,
      "eval_top3_3_weighted_f1_score ": 0.41866179584826513,
      "eval_top_1_macro_f1_score": 0.05674525605488127,
      "eval_top_1_weighted_f1score": 0.19825922860539377,
      "eval_top_3_macro_f1_score": 0.1463405234615811,
      "step": 5824
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.0002,
      "loss": 2.895,
      "step": 5825
    },
    {
      "epoch": 2.5,
      "learning_rate": 0.0002,
      "loss": 2.9044,
      "step": 5830
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0002,
      "loss": 2.9172,
      "step": 5835
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0002,
      "loss": 2.8954,
      "step": 5840
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0002,
      "loss": 2.8852,
      "step": 5845
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0002,
      "loss": 2.9082,
      "step": 5850
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 5855
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 5860
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0002,
      "loss": 2.9091,
      "step": 5865
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0002,
      "loss": 2.9112,
      "step": 5870
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0002,
      "loss": 2.8916,
      "step": 5875
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.0002,
      "loss": 2.9009,
      "step": 5880
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 5885
    },
    {
      "epoch": 2.53,
      "eval_loss": 2.8963630199432373,
      "eval_runtime": 5.3128,
      "eval_samples_per_second": 1859.093,
      "eval_steps_per_second": 14.682,
      "eval_top3_3_weighted_f1_score ": 0.42364414548351786,
      "eval_top_1_macro_f1_score": 0.06334344575414973,
      "eval_top_1_weighted_f1score": 0.20622342804341917,
      "eval_top_3_macro_f1_score": 0.16110020412647388,
      "step": 5888
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.0002,
      "loss": 2.9168,
      "step": 5890
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.0002,
      "loss": 2.9014,
      "step": 5895
    },
    {
      "epoch": 2.53,
      "learning_rate": 0.0002,
      "loss": 2.9104,
      "step": 5900
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0002,
      "loss": 2.894,
      "step": 5905
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0002,
      "loss": 2.9175,
      "step": 5910
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0002,
      "loss": 2.9081,
      "step": 5915
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 5920
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.0002,
      "loss": 2.886,
      "step": 5925
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.0002,
      "loss": 2.8903,
      "step": 5930
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.0002,
      "loss": 2.8903,
      "step": 5935
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.0002,
      "loss": 2.9187,
      "step": 5940
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.0002,
      "loss": 2.8963,
      "step": 5945
    },
    {
      "epoch": 2.55,
      "learning_rate": 0.0002,
      "loss": 2.8996,
      "step": 5950
    },
    {
      "epoch": 2.56,
      "eval_loss": 2.8969385623931885,
      "eval_runtime": 5.3385,
      "eval_samples_per_second": 1850.153,
      "eval_steps_per_second": 14.611,
      "eval_top3_3_weighted_f1_score ": 0.42083135452031184,
      "eval_top_1_macro_f1_score": 0.06035260452585553,
      "eval_top_1_weighted_f1score": 0.20025903908754178,
      "eval_top_3_macro_f1_score": 0.16009983415987541,
      "step": 5952
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.0002,
      "loss": 2.9023,
      "step": 5955
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.0002,
      "loss": 2.8774,
      "step": 5960
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.0002,
      "loss": 2.8883,
      "step": 5965
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.0002,
      "loss": 2.9057,
      "step": 5970
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.0002,
      "loss": 2.884,
      "step": 5975
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.0002,
      "loss": 2.9052,
      "step": 5980
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.0002,
      "loss": 2.8892,
      "step": 5985
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.0002,
      "loss": 2.9064,
      "step": 5990
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.0002,
      "loss": 2.8888,
      "step": 5995
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0002,
      "loss": 2.8808,
      "step": 6000
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0002,
      "loss": 2.9019,
      "step": 6005
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0002,
      "loss": 2.8862,
      "step": 6010
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0002,
      "loss": 2.8906,
      "step": 6015
    },
    {
      "epoch": 2.58,
      "eval_loss": 2.8960154056549072,
      "eval_runtime": 5.3477,
      "eval_samples_per_second": 1846.951,
      "eval_steps_per_second": 14.586,
      "eval_top3_3_weighted_f1_score ": 0.4202857211737356,
      "eval_top_1_macro_f1_score": 0.05629361685201398,
      "eval_top_1_weighted_f1score": 0.19886688542729833,
      "eval_top_3_macro_f1_score": 0.15739346501860904,
      "step": 6016
    },
    {
      "epoch": 2.58,
      "learning_rate": 0.0002,
      "loss": 2.8876,
      "step": 6020
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.0002,
      "loss": 2.9022,
      "step": 6025
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.0002,
      "loss": 2.9002,
      "step": 6030
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.0002,
      "loss": 2.8757,
      "step": 6035
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.0002,
      "loss": 2.889,
      "step": 6040
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0002,
      "loss": 2.8832,
      "step": 6045
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0002,
      "loss": 2.8912,
      "step": 6050
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0002,
      "loss": 2.9111,
      "step": 6055
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0002,
      "loss": 2.8951,
      "step": 6060
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.0002,
      "loss": 2.8937,
      "step": 6065
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.0002,
      "loss": 2.905,
      "step": 6070
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.0002,
      "loss": 2.8915,
      "step": 6075
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.0002,
      "loss": 2.8974,
      "step": 6080
    },
    {
      "epoch": 2.61,
      "eval_loss": 2.895582675933838,
      "eval_runtime": 5.3637,
      "eval_samples_per_second": 1841.439,
      "eval_steps_per_second": 14.542,
      "eval_top3_3_weighted_f1_score ": 0.4176598072491688,
      "eval_top_1_macro_f1_score": 0.057114300468531984,
      "eval_top_1_weighted_f1score": 0.19814815553081722,
      "eval_top_3_macro_f1_score": 0.15260389700126822,
      "step": 6080
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.0002,
      "loss": 2.9064,
      "step": 6085
    },
    {
      "epoch": 2.61,
      "learning_rate": 0.0002,
      "loss": 2.8916,
      "step": 6090
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0002,
      "loss": 2.8959,
      "step": 6095
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0002,
      "loss": 2.8922,
      "step": 6100
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 6105
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.0002,
      "loss": 2.9052,
      "step": 6110
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0002,
      "loss": 2.8922,
      "step": 6115
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0002,
      "loss": 2.8906,
      "step": 6120
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0002,
      "loss": 2.9078,
      "step": 6125
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0002,
      "loss": 2.9076,
      "step": 6130
    },
    {
      "epoch": 2.63,
      "learning_rate": 0.0002,
      "loss": 2.8806,
      "step": 6135
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0002,
      "loss": 2.9068,
      "step": 6140
    },
    {
      "epoch": 2.64,
      "eval_loss": 2.8937771320343018,
      "eval_runtime": 5.3873,
      "eval_samples_per_second": 1833.398,
      "eval_steps_per_second": 14.479,
      "eval_top3_3_weighted_f1_score ": 0.4152767953063901,
      "eval_top_1_macro_f1_score": 0.05652294168706159,
      "eval_top_1_weighted_f1score": 0.1984693966121817,
      "eval_top_3_macro_f1_score": 0.14713473047869802,
      "step": 6144
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 6145
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0002,
      "loss": 2.9021,
      "step": 6150
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0002,
      "loss": 2.9115,
      "step": 6155
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0002,
      "loss": 2.8954,
      "step": 6160
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.0002,
      "loss": 2.9016,
      "step": 6165
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.0002,
      "loss": 2.8823,
      "step": 6170
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.0002,
      "loss": 2.8805,
      "step": 6175
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.0002,
      "loss": 2.924,
      "step": 6180
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 6185
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.0002,
      "loss": 2.8957,
      "step": 6190
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.0002,
      "loss": 2.8875,
      "step": 6195
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.0002,
      "loss": 2.897,
      "step": 6200
    },
    {
      "epoch": 2.66,
      "learning_rate": 0.0002,
      "loss": 2.8915,
      "step": 6205
    },
    {
      "epoch": 2.67,
      "eval_loss": 2.896385669708252,
      "eval_runtime": 5.3292,
      "eval_samples_per_second": 1853.389,
      "eval_steps_per_second": 14.636,
      "eval_top3_3_weighted_f1_score ": 0.4153291557850485,
      "eval_top_1_macro_f1_score": 0.057094784382916375,
      "eval_top_1_weighted_f1score": 0.20055240712210706,
      "eval_top_3_macro_f1_score": 0.15311951608430358,
      "step": 6208
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0002,
      "loss": 2.8894,
      "step": 6210
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0002,
      "loss": 2.9108,
      "step": 6215
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0002,
      "loss": 2.8983,
      "step": 6220
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0002,
      "loss": 2.9117,
      "step": 6225
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0002,
      "loss": 2.8931,
      "step": 6230
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0002,
      "loss": 2.896,
      "step": 6235
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0002,
      "loss": 2.8996,
      "step": 6240
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0002,
      "loss": 2.8913,
      "step": 6245
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.0002,
      "loss": 2.9074,
      "step": 6250
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0002,
      "loss": 2.8939,
      "step": 6255
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0002,
      "loss": 2.8737,
      "step": 6260
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0002,
      "loss": 2.8954,
      "step": 6265
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0002,
      "loss": 2.8976,
      "step": 6270
    },
    {
      "epoch": 2.69,
      "eval_loss": 2.8949334621429443,
      "eval_runtime": 5.4818,
      "eval_samples_per_second": 1801.796,
      "eval_steps_per_second": 14.229,
      "eval_top3_3_weighted_f1_score ": 0.41843966740612304,
      "eval_top_1_macro_f1_score": 0.060669798002958124,
      "eval_top_1_weighted_f1score": 0.2023168005976943,
      "eval_top_3_macro_f1_score": 0.15022535908324353,
      "step": 6272
    },
    {
      "epoch": 2.69,
      "learning_rate": 0.0002,
      "loss": 2.9054,
      "step": 6275
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.0002,
      "loss": 2.8725,
      "step": 6280
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.0002,
      "loss": 2.8898,
      "step": 6285
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.0002,
      "loss": 2.8948,
      "step": 6290
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.0002,
      "loss": 2.9011,
      "step": 6295
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.0002,
      "loss": 2.882,
      "step": 6300
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.0002,
      "loss": 2.9061,
      "step": 6305
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.0002,
      "loss": 2.8826,
      "step": 6310
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.0002,
      "loss": 2.9004,
      "step": 6315
    },
    {
      "epoch": 2.71,
      "learning_rate": 0.0002,
      "loss": 2.8911,
      "step": 6320
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.0002,
      "loss": 2.8974,
      "step": 6325
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.0002,
      "loss": 2.8867,
      "step": 6330
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.0002,
      "loss": 2.8868,
      "step": 6335
    },
    {
      "epoch": 2.72,
      "eval_loss": 2.895129680633545,
      "eval_runtime": 5.3357,
      "eval_samples_per_second": 1851.133,
      "eval_steps_per_second": 14.619,
      "eval_top3_3_weighted_f1_score ": 0.41598930533361117,
      "eval_top_1_macro_f1_score": 0.0586003021627697,
      "eval_top_1_weighted_f1score": 0.1990979731060376,
      "eval_top_3_macro_f1_score": 0.15169411305263725,
      "step": 6336
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.0002,
      "loss": 2.9008,
      "step": 6340
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.0002,
      "loss": 2.9026,
      "step": 6345
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.0002,
      "loss": 2.8753,
      "step": 6350
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.0002,
      "loss": 2.9147,
      "step": 6355
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.0002,
      "loss": 2.8997,
      "step": 6360
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.0002,
      "loss": 2.9068,
      "step": 6365
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.0002,
      "loss": 2.8695,
      "step": 6370
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.0002,
      "loss": 2.8845,
      "step": 6375
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.0002,
      "loss": 2.892,
      "step": 6380
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.0002,
      "loss": 2.8928,
      "step": 6385
    },
    {
      "epoch": 2.74,
      "learning_rate": 0.0002,
      "loss": 2.8795,
      "step": 6390
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0002,
      "loss": 2.905,
      "step": 6395
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0002,
      "loss": 2.9009,
      "step": 6400
    },
    {
      "epoch": 2.75,
      "eval_loss": 2.899210214614868,
      "eval_runtime": 169.1747,
      "eval_samples_per_second": 58.383,
      "eval_steps_per_second": 0.461,
      "eval_top3_3_weighted_f1_score ": 0.4218197411776494,
      "eval_top_1_macro_f1_score": 0.05912854442318141,
      "eval_top_1_weighted_f1score": 0.20171983817246605,
      "eval_top_3_macro_f1_score": 0.15132117088198244,
      "step": 6400
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0002,
      "loss": 2.8763,
      "step": 6405
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0002,
      "loss": 2.8917,
      "step": 6410
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0002,
      "loss": 2.9015,
      "step": 6415
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0002,
      "loss": 2.8771,
      "step": 6420
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0002,
      "loss": 2.9113,
      "step": 6425
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0002,
      "loss": 2.8838,
      "step": 6430
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0002,
      "loss": 2.9129,
      "step": 6435
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.0002,
      "loss": 2.8798,
      "step": 6440
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0002,
      "loss": 2.9101,
      "step": 6445
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 6450
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 6455
    },
    {
      "epoch": 2.77,
      "learning_rate": 0.0002,
      "loss": 2.8787,
      "step": 6460
    },
    {
      "epoch": 2.78,
      "eval_loss": 2.892040252685547,
      "eval_runtime": 5.3482,
      "eval_samples_per_second": 1846.806,
      "eval_steps_per_second": 14.584,
      "eval_top3_3_weighted_f1_score ": 0.4170486755456548,
      "eval_top_1_macro_f1_score": 0.05793422592556205,
      "eval_top_1_weighted_f1score": 0.19810835071144503,
      "eval_top_3_macro_f1_score": 0.1505479607296611,
      "step": 6464
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0002,
      "loss": 2.8964,
      "step": 6465
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0002,
      "loss": 2.8934,
      "step": 6470
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0002,
      "loss": 2.9021,
      "step": 6475
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0002,
      "loss": 2.9039,
      "step": 6480
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 6485
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.0002,
      "loss": 2.8927,
      "step": 6490
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.0002,
      "loss": 2.8772,
      "step": 6495
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.0002,
      "loss": 2.901,
      "step": 6500
    },
    {
      "epoch": 2.79,
      "learning_rate": 0.0002,
      "loss": 2.9051,
      "step": 6505
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0002,
      "loss": 2.8878,
      "step": 6510
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0002,
      "loss": 2.9254,
      "step": 6515
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0002,
      "loss": 2.8969,
      "step": 6520
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0002,
      "loss": 2.9099,
      "step": 6525
    },
    {
      "epoch": 2.8,
      "eval_loss": 2.8924155235290527,
      "eval_runtime": 5.3573,
      "eval_samples_per_second": 1843.663,
      "eval_steps_per_second": 14.56,
      "eval_top3_3_weighted_f1_score ": 0.41909111506627666,
      "eval_top_1_macro_f1_score": 0.05690534821638667,
      "eval_top_1_weighted_f1score": 0.20215670730100305,
      "eval_top_3_macro_f1_score": 0.14864920864192577,
      "step": 6528
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.0002,
      "loss": 2.8965,
      "step": 6530
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.0002,
      "loss": 2.9013,
      "step": 6535
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.0002,
      "loss": 2.9026,
      "step": 6540
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.0002,
      "loss": 2.8956,
      "step": 6545
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.0002,
      "loss": 2.8937,
      "step": 6550
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.0002,
      "loss": 2.9018,
      "step": 6555
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.0002,
      "loss": 2.8946,
      "step": 6560
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.0002,
      "loss": 2.8994,
      "step": 6565
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.0002,
      "loss": 2.899,
      "step": 6570
    },
    {
      "epoch": 2.82,
      "learning_rate": 0.0002,
      "loss": 2.8887,
      "step": 6575
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0002,
      "loss": 2.8928,
      "step": 6580
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0002,
      "loss": 2.8894,
      "step": 6585
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 6590
    },
    {
      "epoch": 2.83,
      "eval_loss": 2.897266149520874,
      "eval_runtime": 5.6923,
      "eval_samples_per_second": 1735.139,
      "eval_steps_per_second": 13.703,
      "eval_top3_3_weighted_f1_score ": 0.4202645153745453,
      "eval_top_1_macro_f1_score": 0.05525966538027519,
      "eval_top_1_weighted_f1score": 0.19904882008303484,
      "eval_top_3_macro_f1_score": 0.1464249193729017,
      "step": 6592
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0002,
      "loss": 2.9131,
      "step": 6595
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.0002,
      "loss": 2.9116,
      "step": 6600
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0002,
      "loss": 2.8903,
      "step": 6605
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0002,
      "loss": 2.9147,
      "step": 6610
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 6615
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0002,
      "loss": 2.8877,
      "step": 6620
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.0002,
      "loss": 2.9054,
      "step": 6625
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.0002,
      "loss": 2.8871,
      "step": 6630
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.0002,
      "loss": 2.9061,
      "step": 6635
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.0002,
      "loss": 2.892,
      "step": 6640
    },
    {
      "epoch": 2.85,
      "learning_rate": 0.0002,
      "loss": 2.8879,
      "step": 6645
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.0002,
      "loss": 2.8866,
      "step": 6650
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.0002,
      "loss": 2.9116,
      "step": 6655
    },
    {
      "epoch": 2.86,
      "eval_loss": 2.8881497383117676,
      "eval_runtime": 5.3811,
      "eval_samples_per_second": 1835.5,
      "eval_steps_per_second": 14.495,
      "eval_top3_3_weighted_f1_score ": 0.4166262160847077,
      "eval_top_1_macro_f1_score": 0.05515885760505369,
      "eval_top_1_weighted_f1score": 0.19965289233016073,
      "eval_top_3_macro_f1_score": 0.15554556026877253,
      "step": 6656
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.0002,
      "loss": 2.8949,
      "step": 6660
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.0002,
      "loss": 2.8776,
      "step": 6665
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.0002,
      "loss": 2.9125,
      "step": 6670
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.0002,
      "loss": 2.8886,
      "step": 6675
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.0002,
      "loss": 2.9188,
      "step": 6680
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.0002,
      "loss": 2.9127,
      "step": 6685
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.0002,
      "loss": 2.8796,
      "step": 6690
    },
    {
      "epoch": 2.87,
      "learning_rate": 0.0002,
      "loss": 2.8876,
      "step": 6695
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.0002,
      "loss": 2.8901,
      "step": 6700
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.0002,
      "loss": 2.9042,
      "step": 6705
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.0002,
      "loss": 2.9133,
      "step": 6710
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.0002,
      "loss": 2.8916,
      "step": 6715
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.0002,
      "loss": 2.9015,
      "step": 6720
    },
    {
      "epoch": 2.89,
      "eval_loss": 2.8918018341064453,
      "eval_runtime": 5.4058,
      "eval_samples_per_second": 1827.113,
      "eval_steps_per_second": 14.429,
      "eval_top3_3_weighted_f1_score ": 0.42017445310857415,
      "eval_top_1_macro_f1_score": 0.05860894422257588,
      "eval_top_1_weighted_f1score": 0.20191660438221576,
      "eval_top_3_macro_f1_score": 0.15352878658061833,
      "step": 6720
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.0002,
      "loss": 2.8796,
      "step": 6725
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.0002,
      "loss": 2.9098,
      "step": 6730
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.0002,
      "loss": 2.8912,
      "step": 6735
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.0002,
      "loss": 2.8922,
      "step": 6740
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0002,
      "loss": 2.9009,
      "step": 6745
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0002,
      "loss": 2.8948,
      "step": 6750
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0002,
      "loss": 2.8842,
      "step": 6755
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0002,
      "loss": 2.8905,
      "step": 6760
    },
    {
      "epoch": 2.9,
      "learning_rate": 0.0002,
      "loss": 2.8984,
      "step": 6765
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0002,
      "loss": 2.871,
      "step": 6770
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0002,
      "loss": 2.898,
      "step": 6775
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0002,
      "loss": 2.8749,
      "step": 6780
    },
    {
      "epoch": 2.91,
      "eval_loss": 2.892881393432617,
      "eval_runtime": 5.4106,
      "eval_samples_per_second": 1825.483,
      "eval_steps_per_second": 14.416,
      "eval_top3_3_weighted_f1_score ": 0.4206111838061315,
      "eval_top_1_macro_f1_score": 0.058917812133703555,
      "eval_top_1_weighted_f1score": 0.20102460044812717,
      "eval_top_3_macro_f1_score": 0.14910007170370257,
      "step": 6784
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.0002,
      "loss": 2.8994,
      "step": 6785
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0002,
      "loss": 2.9037,
      "step": 6790
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0002,
      "loss": 2.8989,
      "step": 6795
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 6800
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0002,
      "loss": 2.9208,
      "step": 6805
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.0002,
      "loss": 2.896,
      "step": 6810
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0002,
      "loss": 2.8883,
      "step": 6815
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0002,
      "loss": 2.8927,
      "step": 6820
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0002,
      "loss": 2.9005,
      "step": 6825
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0002,
      "loss": 2.8785,
      "step": 6830
    },
    {
      "epoch": 2.93,
      "learning_rate": 0.0002,
      "loss": 2.9018,
      "step": 6835
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 6840
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0002,
      "loss": 2.8796,
      "step": 6845
    },
    {
      "epoch": 2.94,
      "eval_loss": 2.8879458904266357,
      "eval_runtime": 5.326,
      "eval_samples_per_second": 1854.496,
      "eval_steps_per_second": 14.645,
      "eval_top3_3_weighted_f1_score ": 0.42427756331941385,
      "eval_top_1_macro_f1_score": 0.06002926605638728,
      "eval_top_1_weighted_f1score": 0.19956929238546744,
      "eval_top_3_macro_f1_score": 0.15784711325428397,
      "step": 6848
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0002,
      "loss": 2.899,
      "step": 6850
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0002,
      "loss": 2.884,
      "step": 6855
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0002,
      "loss": 2.8927,
      "step": 6860
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0002,
      "loss": 2.8962,
      "step": 6865
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0002,
      "loss": 2.8889,
      "step": 6870
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0002,
      "loss": 2.8822,
      "step": 6875
    },
    {
      "epoch": 2.95,
      "learning_rate": 0.0002,
      "loss": 2.8929,
      "step": 6880
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0002,
      "loss": 2.8961,
      "step": 6885
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0002,
      "loss": 2.8792,
      "step": 6890
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0002,
      "loss": 2.9299,
      "step": 6895
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0002,
      "loss": 2.9027,
      "step": 6900
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.0002,
      "loss": 2.8931,
      "step": 6905
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.0002,
      "loss": 2.9024,
      "step": 6910
    },
    {
      "epoch": 2.97,
      "eval_loss": 2.889580488204956,
      "eval_runtime": 5.3297,
      "eval_samples_per_second": 1853.184,
      "eval_steps_per_second": 14.635,
      "eval_top3_3_weighted_f1_score ": 0.42118660139022474,
      "eval_top_1_macro_f1_score": 0.061585736587491564,
      "eval_top_1_weighted_f1score": 0.1995878237563215,
      "eval_top_3_macro_f1_score": 0.16217028756043067,
      "step": 6912
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.0002,
      "loss": 2.8887,
      "step": 6915
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.0002,
      "loss": 2.8887,
      "step": 6920
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.0002,
      "loss": 2.8983,
      "step": 6925
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0002,
      "loss": 2.8955,
      "step": 6930
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0002,
      "loss": 2.8833,
      "step": 6935
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0002,
      "loss": 2.8877,
      "step": 6940
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 6945
    },
    {
      "epoch": 2.98,
      "learning_rate": 0.0002,
      "loss": 2.8919,
      "step": 6950
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.0002,
      "loss": 2.9103,
      "step": 6955
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.0002,
      "loss": 2.8871,
      "step": 6960
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.0002,
      "loss": 2.9042,
      "step": 6965
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.0002,
      "loss": 2.9028,
      "step": 6970
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.0002,
      "loss": 2.8965,
      "step": 6975
    },
    {
      "epoch": 3.0,
      "eval_loss": 2.894157648086548,
      "eval_runtime": 5.7095,
      "eval_samples_per_second": 1729.922,
      "eval_steps_per_second": 13.661,
      "eval_top3_3_weighted_f1_score ": 0.415873483256903,
      "eval_top_1_macro_f1_score": 0.05969227264761643,
      "eval_top_1_weighted_f1score": 0.2007634987366759,
      "eval_top_3_macro_f1_score": 0.14123113694671527,
      "step": 6976
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 6980
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0002,
      "loss": 2.9018,
      "step": 6985
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0002,
      "loss": 3.1866,
      "step": 6990
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.0002,
      "loss": 2.8849,
      "step": 6995
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 7000
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.0002,
      "loss": 2.8856,
      "step": 7005
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.0002,
      "loss": 2.878,
      "step": 7010
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.0002,
      "loss": 2.8811,
      "step": 7015
    },
    {
      "epoch": 3.01,
      "learning_rate": 0.0002,
      "loss": 2.9049,
      "step": 7020
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0002,
      "loss": 2.8864,
      "step": 7025
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0002,
      "loss": 2.8818,
      "step": 7030
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0002,
      "loss": 2.8784,
      "step": 7035
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0002,
      "loss": 2.885,
      "step": 7040
    },
    {
      "epoch": 3.02,
      "eval_loss": 2.888051986694336,
      "eval_runtime": 5.3058,
      "eval_samples_per_second": 1861.536,
      "eval_steps_per_second": 14.701,
      "eval_top3_3_weighted_f1_score ": 0.41846760208054623,
      "eval_top_1_macro_f1_score": 0.05625990712567595,
      "eval_top_1_weighted_f1score": 0.20095345295287725,
      "eval_top_3_macro_f1_score": 0.15870986664918688,
      "step": 7040
    },
    {
      "epoch": 3.02,
      "learning_rate": 0.0002,
      "loss": 2.8778,
      "step": 7045
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0002,
      "loss": 2.8966,
      "step": 7050
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0002,
      "loss": 2.882,
      "step": 7055
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0002,
      "loss": 2.9075,
      "step": 7060
    },
    {
      "epoch": 3.03,
      "learning_rate": 0.0002,
      "loss": 2.8859,
      "step": 7065
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0002,
      "loss": 2.8881,
      "step": 7070
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0002,
      "loss": 2.8851,
      "step": 7075
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0002,
      "loss": 2.8995,
      "step": 7080
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0002,
      "loss": 2.9107,
      "step": 7085
    },
    {
      "epoch": 3.04,
      "learning_rate": 0.0002,
      "loss": 2.8927,
      "step": 7090
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 7095
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0002,
      "loss": 2.8879,
      "step": 7100
    },
    {
      "epoch": 3.05,
      "eval_loss": 2.8877265453338623,
      "eval_runtime": 5.3303,
      "eval_samples_per_second": 1853.0,
      "eval_steps_per_second": 14.633,
      "eval_top3_3_weighted_f1_score ": 0.42163813405919864,
      "eval_top_1_macro_f1_score": 0.05934196938210321,
      "eval_top_1_weighted_f1score": 0.20608303826374583,
      "eval_top_3_macro_f1_score": 0.1563438316036981,
      "step": 7104
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0002,
      "loss": 2.8737,
      "step": 7105
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0002,
      "loss": 2.8776,
      "step": 7110
    },
    {
      "epoch": 3.05,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 7115
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.0002,
      "loss": 2.8891,
      "step": 7120
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.0002,
      "loss": 2.8948,
      "step": 7125
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.0002,
      "loss": 2.8976,
      "step": 7130
    },
    {
      "epoch": 3.06,
      "learning_rate": 0.0002,
      "loss": 2.8962,
      "step": 7135
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 7140
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0002,
      "loss": 2.8992,
      "step": 7145
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0002,
      "loss": 2.8897,
      "step": 7150
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 7155
    },
    {
      "epoch": 3.07,
      "learning_rate": 0.0002,
      "loss": 2.8967,
      "step": 7160
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.0002,
      "loss": 2.8697,
      "step": 7165
    },
    {
      "epoch": 3.08,
      "eval_loss": 2.8849167823791504,
      "eval_runtime": 5.313,
      "eval_samples_per_second": 1859.021,
      "eval_steps_per_second": 14.681,
      "eval_top3_3_weighted_f1_score ": 0.41884094061088384,
      "eval_top_1_macro_f1_score": 0.06004807726120719,
      "eval_top_1_weighted_f1score": 0.20160694888346856,
      "eval_top_3_macro_f1_score": 0.1554945377775778,
      "step": 7168
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.0002,
      "loss": 2.8904,
      "step": 7170
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.0002,
      "loss": 2.8731,
      "step": 7175
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.0002,
      "loss": 2.9,
      "step": 7180
    },
    {
      "epoch": 3.08,
      "learning_rate": 0.0002,
      "loss": 2.8707,
      "step": 7185
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.0002,
      "loss": 2.8841,
      "step": 7190
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 7195
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 7200
    },
    {
      "epoch": 3.09,
      "learning_rate": 0.0002,
      "loss": 2.8919,
      "step": 7205
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0002,
      "loss": 2.8859,
      "step": 7210
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0002,
      "loss": 2.9038,
      "step": 7215
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0002,
      "loss": 2.904,
      "step": 7220
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0002,
      "loss": 2.8747,
      "step": 7225
    },
    {
      "epoch": 3.1,
      "learning_rate": 0.0002,
      "loss": 2.8997,
      "step": 7230
    },
    {
      "epoch": 3.11,
      "eval_loss": 2.884518623352051,
      "eval_runtime": 5.3065,
      "eval_samples_per_second": 1861.319,
      "eval_steps_per_second": 14.699,
      "eval_top3_3_weighted_f1_score ": 0.42023967321832556,
      "eval_top_1_macro_f1_score": 0.06522960041748352,
      "eval_top_1_weighted_f1score": 0.20200750846480445,
      "eval_top_3_macro_f1_score": 0.1600667571620107,
      "step": 7232
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0002,
      "loss": 2.8931,
      "step": 7235
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0002,
      "loss": 2.8625,
      "step": 7240
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0002,
      "loss": 2.8752,
      "step": 7245
    },
    {
      "epoch": 3.11,
      "learning_rate": 0.0002,
      "loss": 2.9108,
      "step": 7250
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.0002,
      "loss": 2.8663,
      "step": 7255
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.0002,
      "loss": 2.8862,
      "step": 7260
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.0002,
      "loss": 2.8968,
      "step": 7265
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 7270
    },
    {
      "epoch": 3.12,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 7275
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0002,
      "loss": 2.8851,
      "step": 7280
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0002,
      "loss": 2.8964,
      "step": 7285
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0002,
      "loss": 2.8879,
      "step": 7290
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0002,
      "loss": 2.9061,
      "step": 7295
    },
    {
      "epoch": 3.13,
      "eval_loss": 2.886598587036133,
      "eval_runtime": 5.3853,
      "eval_samples_per_second": 1834.071,
      "eval_steps_per_second": 14.484,
      "eval_top3_3_weighted_f1_score ": 0.42203774980654735,
      "eval_top_1_macro_f1_score": 0.06814627498070826,
      "eval_top_1_weighted_f1score": 0.20365995015379834,
      "eval_top_3_macro_f1_score": 0.15943133680115384,
      "step": 7296
    },
    {
      "epoch": 3.13,
      "learning_rate": 0.0002,
      "loss": 2.8975,
      "step": 7300
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.0002,
      "loss": 2.8634,
      "step": 7305
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.0002,
      "loss": 2.8952,
      "step": 7310
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 7315
    },
    {
      "epoch": 3.14,
      "learning_rate": 0.0002,
      "loss": 2.8759,
      "step": 7320
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0002,
      "loss": 2.8962,
      "step": 7325
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0002,
      "loss": 2.8806,
      "step": 7330
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0002,
      "loss": 2.8977,
      "step": 7335
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0002,
      "loss": 2.8829,
      "step": 7340
    },
    {
      "epoch": 3.15,
      "learning_rate": 0.0002,
      "loss": 2.8727,
      "step": 7345
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.0002,
      "loss": 2.9073,
      "step": 7350
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.0002,
      "loss": 2.8897,
      "step": 7355
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.0002,
      "loss": 2.9069,
      "step": 7360
    },
    {
      "epoch": 3.16,
      "eval_loss": 2.8906261920928955,
      "eval_runtime": 5.3997,
      "eval_samples_per_second": 1829.178,
      "eval_steps_per_second": 14.445,
      "eval_top3_3_weighted_f1_score ": 0.4215247529697737,
      "eval_top_1_macro_f1_score": 0.057807715800752876,
      "eval_top_1_weighted_f1score": 0.20332823261700023,
      "eval_top_3_macro_f1_score": 0.15819296969443752,
      "step": 7360
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.0002,
      "loss": 2.8936,
      "step": 7365
    },
    {
      "epoch": 3.16,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 7370
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.0002,
      "loss": 2.8953,
      "step": 7375
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 7380
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.0002,
      "loss": 2.8767,
      "step": 7385
    },
    {
      "epoch": 3.17,
      "learning_rate": 0.0002,
      "loss": 2.8889,
      "step": 7390
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.0002,
      "loss": 2.8801,
      "step": 7395
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.0002,
      "loss": 2.8718,
      "step": 7400
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.0002,
      "loss": 2.8986,
      "step": 7405
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 7410
    },
    {
      "epoch": 3.18,
      "learning_rate": 0.0002,
      "loss": 2.877,
      "step": 7415
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0002,
      "loss": 2.9114,
      "step": 7420
    },
    {
      "epoch": 3.19,
      "eval_loss": 2.892573595046997,
      "eval_runtime": 5.7812,
      "eval_samples_per_second": 1708.454,
      "eval_steps_per_second": 13.492,
      "eval_top3_3_weighted_f1_score ": 0.418720602900631,
      "eval_top_1_macro_f1_score": 0.05863396154566181,
      "eval_top_1_weighted_f1score": 0.1981049993087968,
      "eval_top_3_macro_f1_score": 0.15423932435130192,
      "step": 7424
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0002,
      "loss": 2.883,
      "step": 7425
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0002,
      "loss": 2.8825,
      "step": 7430
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0002,
      "loss": 2.8918,
      "step": 7435
    },
    {
      "epoch": 3.19,
      "learning_rate": 0.0002,
      "loss": 2.8921,
      "step": 7440
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 7445
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0002,
      "loss": 2.8846,
      "step": 7450
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0002,
      "loss": 2.8992,
      "step": 7455
    },
    {
      "epoch": 3.2,
      "learning_rate": 0.0002,
      "loss": 2.8784,
      "step": 7460
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0002,
      "loss": 2.8865,
      "step": 7465
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0002,
      "loss": 2.8897,
      "step": 7470
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0002,
      "loss": 2.8738,
      "step": 7475
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 7480
    },
    {
      "epoch": 3.21,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 7485
    },
    {
      "epoch": 3.22,
      "eval_loss": 2.8910961151123047,
      "eval_runtime": 5.3508,
      "eval_samples_per_second": 1845.882,
      "eval_steps_per_second": 14.577,
      "eval_top3_3_weighted_f1_score ": 0.4186955517891209,
      "eval_top_1_macro_f1_score": 0.0612936693324126,
      "eval_top_1_weighted_f1score": 0.19950173195710216,
      "eval_top_3_macro_f1_score": 0.15471827358614995,
      "step": 7488
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0002,
      "loss": 2.8887,
      "step": 7490
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0002,
      "loss": 2.891,
      "step": 7495
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0002,
      "loss": 2.8935,
      "step": 7500
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0002,
      "loss": 2.8885,
      "step": 7505
    },
    {
      "epoch": 3.22,
      "learning_rate": 0.0002,
      "loss": 2.8802,
      "step": 7510
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.0002,
      "loss": 2.892,
      "step": 7515
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.0002,
      "loss": 2.9136,
      "step": 7520
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.0002,
      "loss": 2.8675,
      "step": 7525
    },
    {
      "epoch": 3.23,
      "learning_rate": 0.0002,
      "loss": 2.8879,
      "step": 7530
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0002,
      "loss": 2.9036,
      "step": 7535
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0002,
      "loss": 2.8844,
      "step": 7540
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0002,
      "loss": 2.8935,
      "step": 7545
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0002,
      "loss": 2.8796,
      "step": 7550
    },
    {
      "epoch": 3.24,
      "eval_loss": 2.8844926357269287,
      "eval_runtime": 5.3149,
      "eval_samples_per_second": 1858.373,
      "eval_steps_per_second": 14.676,
      "eval_top3_3_weighted_f1_score ": 0.4227162451195223,
      "eval_top_1_macro_f1_score": 0.06140110811329589,
      "eval_top_1_weighted_f1score": 0.20275439396155656,
      "eval_top_3_macro_f1_score": 0.16255781638160421,
      "step": 7552
    },
    {
      "epoch": 3.24,
      "learning_rate": 0.0002,
      "loss": 2.881,
      "step": 7555
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.0002,
      "loss": 2.8826,
      "step": 7560
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.0002,
      "loss": 2.885,
      "step": 7565
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.0002,
      "loss": 2.8933,
      "step": 7570
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.0002,
      "loss": 2.8866,
      "step": 7575
    },
    {
      "epoch": 3.25,
      "learning_rate": 0.0002,
      "loss": 2.902,
      "step": 7580
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0002,
      "loss": 2.8856,
      "step": 7585
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0002,
      "loss": 2.8819,
      "step": 7590
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0002,
      "loss": 2.877,
      "step": 7595
    },
    {
      "epoch": 3.26,
      "learning_rate": 0.0002,
      "loss": 2.8728,
      "step": 7600
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.0002,
      "loss": 2.9072,
      "step": 7605
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.0002,
      "loss": 2.8713,
      "step": 7610
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 7615
    },
    {
      "epoch": 3.27,
      "eval_loss": 2.8841776847839355,
      "eval_runtime": 5.327,
      "eval_samples_per_second": 1854.134,
      "eval_steps_per_second": 14.642,
      "eval_top3_3_weighted_f1_score ": 0.4187903482484385,
      "eval_top_1_macro_f1_score": 0.05984177228173996,
      "eval_top_1_weighted_f1score": 0.20498038698864343,
      "eval_top_3_macro_f1_score": 0.15411286732053833,
      "step": 7616
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.0002,
      "loss": 2.8939,
      "step": 7620
    },
    {
      "epoch": 3.27,
      "learning_rate": 0.0002,
      "loss": 2.8701,
      "step": 7625
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0002,
      "loss": 2.8888,
      "step": 7630
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0002,
      "loss": 2.8777,
      "step": 7635
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0002,
      "loss": 2.8815,
      "step": 7640
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0002,
      "loss": 2.8801,
      "step": 7645
    },
    {
      "epoch": 3.28,
      "learning_rate": 0.0002,
      "loss": 2.8839,
      "step": 7650
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.0002,
      "loss": 2.8764,
      "step": 7655
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.0002,
      "loss": 2.8993,
      "step": 7660
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.0002,
      "loss": 2.8922,
      "step": 7665
    },
    {
      "epoch": 3.29,
      "learning_rate": 0.0002,
      "loss": 2.8862,
      "step": 7670
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0002,
      "loss": 2.8891,
      "step": 7675
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0002,
      "loss": 2.8825,
      "step": 7680
    },
    {
      "epoch": 3.3,
      "eval_loss": 2.894038438796997,
      "eval_runtime": 5.3658,
      "eval_samples_per_second": 1840.748,
      "eval_steps_per_second": 14.537,
      "eval_top3_3_weighted_f1_score ": 0.4202346727377731,
      "eval_top_1_macro_f1_score": 0.06032218990545729,
      "eval_top_1_weighted_f1score": 0.2011031684607638,
      "eval_top_3_macro_f1_score": 0.15071342049635814,
      "step": 7680
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0002,
      "loss": 2.898,
      "step": 7685
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 7690
    },
    {
      "epoch": 3.3,
      "learning_rate": 0.0002,
      "loss": 2.8781,
      "step": 7695
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 7700
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.0002,
      "loss": 2.8886,
      "step": 7705
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.0002,
      "loss": 2.8876,
      "step": 7710
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.0002,
      "loss": 2.8901,
      "step": 7715
    },
    {
      "epoch": 3.31,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 7720
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.0002,
      "loss": 2.885,
      "step": 7725
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.0002,
      "loss": 2.8722,
      "step": 7730
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.0002,
      "loss": 2.8758,
      "step": 7735
    },
    {
      "epoch": 3.32,
      "learning_rate": 0.0002,
      "loss": 2.9031,
      "step": 7740
    },
    {
      "epoch": 3.32,
      "eval_loss": 2.882852792739868,
      "eval_runtime": 5.3107,
      "eval_samples_per_second": 1859.817,
      "eval_steps_per_second": 14.687,
      "eval_top3_3_weighted_f1_score ": 0.4190313630044371,
      "eval_top_1_macro_f1_score": 0.061544251912979984,
      "eval_top_1_weighted_f1score": 0.205241513146368,
      "eval_top_3_macro_f1_score": 0.14671600365597517,
      "step": 7744
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0002,
      "loss": 2.8854,
      "step": 7745
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 7750
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 7755
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0002,
      "loss": 2.8927,
      "step": 7760
    },
    {
      "epoch": 3.33,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 7765
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.0002,
      "loss": 2.8924,
      "step": 7770
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.0002,
      "loss": 2.8806,
      "step": 7775
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.0002,
      "loss": 2.8661,
      "step": 7780
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.0002,
      "loss": 2.9012,
      "step": 7785
    },
    {
      "epoch": 3.34,
      "learning_rate": 0.0002,
      "loss": 2.8762,
      "step": 7790
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0002,
      "loss": 2.8795,
      "step": 7795
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0002,
      "loss": 2.8862,
      "step": 7800
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0002,
      "loss": 2.8786,
      "step": 7805
    },
    {
      "epoch": 3.35,
      "eval_loss": 2.8905601501464844,
      "eval_runtime": 5.2935,
      "eval_samples_per_second": 1865.862,
      "eval_steps_per_second": 14.735,
      "eval_top3_3_weighted_f1_score ": 0.41593301042000663,
      "eval_top_1_macro_f1_score": 0.06133824832129849,
      "eval_top_1_weighted_f1score": 0.20215134462916098,
      "eval_top_3_macro_f1_score": 0.15742207883543302,
      "step": 7808
    },
    {
      "epoch": 3.35,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 7810
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0002,
      "loss": 2.8843,
      "step": 7815
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0002,
      "loss": 2.884,
      "step": 7820
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0002,
      "loss": 2.898,
      "step": 7825
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0002,
      "loss": 2.8884,
      "step": 7830
    },
    {
      "epoch": 3.36,
      "learning_rate": 0.0002,
      "loss": 2.8606,
      "step": 7835
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0002,
      "loss": 2.8791,
      "step": 7840
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0002,
      "loss": 2.8691,
      "step": 7845
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0002,
      "loss": 2.895,
      "step": 7850
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0002,
      "loss": 2.8872,
      "step": 7855
    },
    {
      "epoch": 3.37,
      "learning_rate": 0.0002,
      "loss": 2.8708,
      "step": 7860
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.0002,
      "loss": 2.9043,
      "step": 7865
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.0002,
      "loss": 2.8964,
      "step": 7870
    },
    {
      "epoch": 3.38,
      "eval_loss": 2.88686466217041,
      "eval_runtime": 31.8739,
      "eval_samples_per_second": 309.878,
      "eval_steps_per_second": 2.447,
      "eval_top3_3_weighted_f1_score ": 0.4230829780304739,
      "eval_top_1_macro_f1_score": 0.054228979253503506,
      "eval_top_1_weighted_f1score": 0.19922267969873778,
      "eval_top_3_macro_f1_score": 0.15943393088965593,
      "step": 7872
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.0002,
      "loss": 2.8811,
      "step": 7875
    },
    {
      "epoch": 3.38,
      "learning_rate": 0.0002,
      "loss": 2.8766,
      "step": 7880
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 7885
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0002,
      "loss": 2.8824,
      "step": 7890
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0002,
      "loss": 2.8831,
      "step": 7895
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0002,
      "loss": 2.8913,
      "step": 7900
    },
    {
      "epoch": 3.39,
      "learning_rate": 0.0002,
      "loss": 2.9162,
      "step": 7905
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.0002,
      "loss": 2.8883,
      "step": 7910
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.0002,
      "loss": 2.8627,
      "step": 7915
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.0002,
      "loss": 2.8923,
      "step": 7920
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.0002,
      "loss": 2.9033,
      "step": 7925
    },
    {
      "epoch": 3.4,
      "learning_rate": 0.0002,
      "loss": 2.8813,
      "step": 7930
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.0002,
      "loss": 2.9054,
      "step": 7935
    },
    {
      "epoch": 3.41,
      "eval_loss": 2.885826587677002,
      "eval_runtime": 5.3867,
      "eval_samples_per_second": 1833.603,
      "eval_steps_per_second": 14.48,
      "eval_top3_3_weighted_f1_score ": 0.42314543248296244,
      "eval_top_1_macro_f1_score": 0.0681717331051118,
      "eval_top_1_weighted_f1score": 0.19846969594949262,
      "eval_top_3_macro_f1_score": 0.15816567728101838,
      "step": 7936
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.0002,
      "loss": 2.8999,
      "step": 7940
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.0002,
      "loss": 2.8825,
      "step": 7945
    },
    {
      "epoch": 3.41,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 7950
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0002,
      "loss": 2.882,
      "step": 7955
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0002,
      "loss": 2.8873,
      "step": 7960
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0002,
      "loss": 2.8941,
      "step": 7965
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0002,
      "loss": 2.9129,
      "step": 7970
    },
    {
      "epoch": 3.42,
      "learning_rate": 0.0002,
      "loss": 2.8854,
      "step": 7975
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0002,
      "loss": 2.8872,
      "step": 7980
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 7985
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0002,
      "loss": 2.887,
      "step": 7990
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 7995
    },
    {
      "epoch": 3.43,
      "learning_rate": 0.0002,
      "loss": 2.8847,
      "step": 8000
    },
    {
      "epoch": 3.43,
      "eval_loss": 2.8894846439361572,
      "eval_runtime": 5.3414,
      "eval_samples_per_second": 1849.136,
      "eval_steps_per_second": 14.603,
      "eval_top3_3_weighted_f1_score ": 0.4191980028620256,
      "eval_top_1_macro_f1_score": 0.06801016994886046,
      "eval_top_1_weighted_f1score": 0.2038280543824952,
      "eval_top_3_macro_f1_score": 0.16341329875150717,
      "step": 8000
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0002,
      "loss": 2.902,
      "step": 8005
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0002,
      "loss": 2.8807,
      "step": 8010
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0002,
      "loss": 2.9025,
      "step": 8015
    },
    {
      "epoch": 3.44,
      "learning_rate": 0.0002,
      "loss": 2.8807,
      "step": 8020
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.0002,
      "loss": 2.8845,
      "step": 8025
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.0002,
      "loss": 2.9069,
      "step": 8030
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.0002,
      "loss": 2.8704,
      "step": 8035
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.0002,
      "loss": 2.8993,
      "step": 8040
    },
    {
      "epoch": 3.45,
      "learning_rate": 0.0002,
      "loss": 2.8888,
      "step": 8045
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0002,
      "loss": 2.8918,
      "step": 8050
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0002,
      "loss": 2.9014,
      "step": 8055
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0002,
      "loss": 2.8809,
      "step": 8060
    },
    {
      "epoch": 3.46,
      "eval_loss": 2.883298873901367,
      "eval_runtime": 5.3415,
      "eval_samples_per_second": 1849.094,
      "eval_steps_per_second": 14.603,
      "eval_top3_3_weighted_f1_score ": 0.4232752392927697,
      "eval_top_1_macro_f1_score": 0.062447123808691926,
      "eval_top_1_weighted_f1score": 0.20257688143489982,
      "eval_top_3_macro_f1_score": 0.16734324172721946,
      "step": 8064
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0002,
      "loss": 2.8984,
      "step": 8065
    },
    {
      "epoch": 3.46,
      "learning_rate": 0.0002,
      "loss": 2.8893,
      "step": 8070
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.0002,
      "loss": 2.8649,
      "step": 8075
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.0002,
      "loss": 2.8817,
      "step": 8080
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.0002,
      "loss": 2.8859,
      "step": 8085
    },
    {
      "epoch": 3.47,
      "learning_rate": 0.0002,
      "loss": 2.8785,
      "step": 8090
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.0002,
      "loss": 2.8918,
      "step": 8095
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.0002,
      "loss": 2.8707,
      "step": 8100
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 8105
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.0002,
      "loss": 2.8773,
      "step": 8110
    },
    {
      "epoch": 3.48,
      "learning_rate": 0.0002,
      "loss": 2.8845,
      "step": 8115
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.0002,
      "loss": 2.8925,
      "step": 8120
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.0002,
      "loss": 2.8915,
      "step": 8125
    },
    {
      "epoch": 3.49,
      "eval_loss": 2.885603427886963,
      "eval_runtime": 5.4426,
      "eval_samples_per_second": 1814.745,
      "eval_steps_per_second": 14.331,
      "eval_top3_3_weighted_f1_score ": 0.4233867206076715,
      "eval_top_1_macro_f1_score": 0.05447611955443182,
      "eval_top_1_weighted_f1score": 0.20162142002690994,
      "eval_top_3_macro_f1_score": 0.16188996546475656,
      "step": 8128
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 8130
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.0002,
      "loss": 2.8855,
      "step": 8135
    },
    {
      "epoch": 3.49,
      "learning_rate": 0.0002,
      "loss": 2.889,
      "step": 8140
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.0002,
      "loss": 2.8756,
      "step": 8145
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.0002,
      "loss": 2.8852,
      "step": 8150
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.0002,
      "loss": 2.9,
      "step": 8155
    },
    {
      "epoch": 3.5,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 8160
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 8165
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0002,
      "loss": 2.8777,
      "step": 8170
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0002,
      "loss": 2.8726,
      "step": 8175
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0002,
      "loss": 2.8666,
      "step": 8180
    },
    {
      "epoch": 3.51,
      "learning_rate": 0.0002,
      "loss": 2.8982,
      "step": 8185
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 8190
    },
    {
      "epoch": 3.52,
      "eval_loss": 2.8865466117858887,
      "eval_runtime": 5.3554,
      "eval_samples_per_second": 1844.29,
      "eval_steps_per_second": 14.565,
      "eval_top3_3_weighted_f1_score ": 0.4203827795263443,
      "eval_top_1_macro_f1_score": 0.06852860836345402,
      "eval_top_1_weighted_f1score": 0.20701276419199244,
      "eval_top_3_macro_f1_score": 0.15979011951298183,
      "step": 8192
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 8195
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0002,
      "loss": 2.8876,
      "step": 8200
    },
    {
      "epoch": 3.52,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 8205
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0002,
      "loss": 2.8728,
      "step": 8210
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0002,
      "loss": 2.8694,
      "step": 8215
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0002,
      "loss": 2.8962,
      "step": 8220
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0002,
      "loss": 2.8765,
      "step": 8225
    },
    {
      "epoch": 3.53,
      "learning_rate": 0.0002,
      "loss": 2.8933,
      "step": 8230
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.0002,
      "loss": 2.869,
      "step": 8235
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.0002,
      "loss": 2.9021,
      "step": 8240
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.0002,
      "loss": 2.8953,
      "step": 8245
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.0002,
      "loss": 2.8961,
      "step": 8250
    },
    {
      "epoch": 3.54,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 8255
    },
    {
      "epoch": 3.54,
      "eval_loss": 2.88881254196167,
      "eval_runtime": 10.47,
      "eval_samples_per_second": 943.365,
      "eval_steps_per_second": 7.45,
      "eval_top3_3_weighted_f1_score ": 0.421615034340461,
      "eval_top_1_macro_f1_score": 0.05707255432075041,
      "eval_top_1_weighted_f1score": 0.20236393382029402,
      "eval_top_3_macro_f1_score": 0.16403338219761618,
      "step": 8256
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0002,
      "loss": 2.8778,
      "step": 8260
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0002,
      "loss": 2.8824,
      "step": 8265
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0002,
      "loss": 2.8672,
      "step": 8270
    },
    {
      "epoch": 3.55,
      "learning_rate": 0.0002,
      "loss": 2.8754,
      "step": 8275
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.0002,
      "loss": 2.8962,
      "step": 8280
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.0002,
      "loss": 2.8939,
      "step": 8285
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.0002,
      "loss": 2.8706,
      "step": 8290
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.0002,
      "loss": 2.8633,
      "step": 8295
    },
    {
      "epoch": 3.56,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 8300
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.0002,
      "loss": 2.8836,
      "step": 8305
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.0002,
      "loss": 2.894,
      "step": 8310
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.0002,
      "loss": 2.887,
      "step": 8315
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.0002,
      "loss": 2.8758,
      "step": 8320
    },
    {
      "epoch": 3.57,
      "eval_loss": 2.882431983947754,
      "eval_runtime": 5.7726,
      "eval_samples_per_second": 1711.007,
      "eval_steps_per_second": 13.512,
      "eval_top3_3_weighted_f1_score ": 0.4276064007964306,
      "eval_top_1_macro_f1_score": 0.0631328266883076,
      "eval_top_1_weighted_f1score": 0.204925576803142,
      "eval_top_3_macro_f1_score": 0.16421739218602505,
      "step": 8320
    },
    {
      "epoch": 3.57,
      "learning_rate": 0.0002,
      "loss": 2.8648,
      "step": 8325
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.0002,
      "loss": 2.876,
      "step": 8330
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.0002,
      "loss": 2.8822,
      "step": 8335
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.0002,
      "loss": 2.8788,
      "step": 8340
    },
    {
      "epoch": 3.58,
      "learning_rate": 0.0002,
      "loss": 2.8747,
      "step": 8345
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.0002,
      "loss": 2.8772,
      "step": 8350
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.0002,
      "loss": 2.8802,
      "step": 8355
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.0002,
      "loss": 2.8786,
      "step": 8360
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.0002,
      "loss": 2.8756,
      "step": 8365
    },
    {
      "epoch": 3.59,
      "learning_rate": 0.0002,
      "loss": 2.8815,
      "step": 8370
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.0002,
      "loss": 2.8888,
      "step": 8375
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.0002,
      "loss": 2.8708,
      "step": 8380
    },
    {
      "epoch": 3.6,
      "eval_loss": 2.8809597492218018,
      "eval_runtime": 5.341,
      "eval_samples_per_second": 1849.276,
      "eval_steps_per_second": 14.604,
      "eval_top3_3_weighted_f1_score ": 0.4246114037313046,
      "eval_top_1_macro_f1_score": 0.05977962775940492,
      "eval_top_1_weighted_f1score": 0.19956349021000716,
      "eval_top_3_macro_f1_score": 0.1684172253348305,
      "step": 8384
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 8385
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.0002,
      "loss": 2.8975,
      "step": 8390
    },
    {
      "epoch": 3.6,
      "learning_rate": 0.0002,
      "loss": 2.8774,
      "step": 8395
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.0002,
      "loss": 2.8734,
      "step": 8400
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.0002,
      "loss": 2.8875,
      "step": 8405
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.0002,
      "loss": 2.8743,
      "step": 8410
    },
    {
      "epoch": 3.61,
      "learning_rate": 0.0002,
      "loss": 2.8687,
      "step": 8415
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0002,
      "loss": 2.8858,
      "step": 8420
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0002,
      "loss": 2.898,
      "step": 8425
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0002,
      "loss": 2.9,
      "step": 8430
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0002,
      "loss": 2.8913,
      "step": 8435
    },
    {
      "epoch": 3.62,
      "learning_rate": 0.0002,
      "loss": 2.8613,
      "step": 8440
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0002,
      "loss": 2.9039,
      "step": 8445
    },
    {
      "epoch": 3.63,
      "eval_loss": 2.879713535308838,
      "eval_runtime": 5.3117,
      "eval_samples_per_second": 1859.497,
      "eval_steps_per_second": 14.685,
      "eval_top3_3_weighted_f1_score ": 0.422907351453079,
      "eval_top_1_macro_f1_score": 0.061077590558727204,
      "eval_top_1_weighted_f1score": 0.2037311984234322,
      "eval_top_3_macro_f1_score": 0.15977641394382092,
      "step": 8448
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0002,
      "loss": 2.8776,
      "step": 8450
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0002,
      "loss": 2.8751,
      "step": 8455
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 8460
    },
    {
      "epoch": 3.63,
      "learning_rate": 0.0002,
      "loss": 2.8913,
      "step": 8465
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 8470
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 8475
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0002,
      "loss": 2.875,
      "step": 8480
    },
    {
      "epoch": 3.64,
      "learning_rate": 0.0002,
      "loss": 2.8747,
      "step": 8485
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.0002,
      "loss": 2.8767,
      "step": 8490
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.0002,
      "loss": 2.8958,
      "step": 8495
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.0002,
      "loss": 2.8598,
      "step": 8500
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.0002,
      "loss": 2.8764,
      "step": 8505
    },
    {
      "epoch": 3.65,
      "learning_rate": 0.0002,
      "loss": 2.9056,
      "step": 8510
    },
    {
      "epoch": 3.65,
      "eval_loss": 2.879021644592285,
      "eval_runtime": 5.3893,
      "eval_samples_per_second": 1832.701,
      "eval_steps_per_second": 14.473,
      "eval_top3_3_weighted_f1_score ": 0.42083999002881417,
      "eval_top_1_macro_f1_score": 0.05751107770938145,
      "eval_top_1_weighted_f1score": 0.20063172811985464,
      "eval_top_3_macro_f1_score": 0.15231224977727817,
      "step": 8512
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0002,
      "loss": 2.8772,
      "step": 8515
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0002,
      "loss": 2.858,
      "step": 8520
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0002,
      "loss": 2.8674,
      "step": 8525
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0002,
      "loss": 2.8663,
      "step": 8530
    },
    {
      "epoch": 3.66,
      "learning_rate": 0.0002,
      "loss": 2.8912,
      "step": 8535
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.0002,
      "loss": 2.8946,
      "step": 8540
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.0002,
      "loss": 2.8685,
      "step": 8545
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.0002,
      "loss": 2.8856,
      "step": 8550
    },
    {
      "epoch": 3.67,
      "learning_rate": 0.0002,
      "loss": 2.8764,
      "step": 8555
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0002,
      "loss": 2.8753,
      "step": 8560
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0002,
      "loss": 2.8829,
      "step": 8565
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0002,
      "loss": 2.872,
      "step": 8570
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0002,
      "loss": 2.8762,
      "step": 8575
    },
    {
      "epoch": 3.68,
      "eval_loss": 2.8791356086730957,
      "eval_runtime": 5.5245,
      "eval_samples_per_second": 1787.855,
      "eval_steps_per_second": 14.119,
      "eval_top3_3_weighted_f1_score ": 0.4259180292541815,
      "eval_top_1_macro_f1_score": 0.0614787145421599,
      "eval_top_1_weighted_f1score": 0.20327243243179235,
      "eval_top_3_macro_f1_score": 0.15667803624411272,
      "step": 8576
    },
    {
      "epoch": 3.68,
      "learning_rate": 0.0002,
      "loss": 2.912,
      "step": 8580
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.0002,
      "loss": 2.8959,
      "step": 8585
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.0002,
      "loss": 2.8912,
      "step": 8590
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.0002,
      "loss": 2.871,
      "step": 8595
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.0002,
      "loss": 2.8803,
      "step": 8600
    },
    {
      "epoch": 3.69,
      "learning_rate": 0.0002,
      "loss": 2.8931,
      "step": 8605
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0002,
      "loss": 2.8736,
      "step": 8610
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 8615
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0002,
      "loss": 2.8816,
      "step": 8620
    },
    {
      "epoch": 3.7,
      "learning_rate": 0.0002,
      "loss": 2.8843,
      "step": 8625
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0002,
      "loss": 2.8798,
      "step": 8630
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0002,
      "loss": 2.8836,
      "step": 8635
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0002,
      "loss": 2.8925,
      "step": 8640
    },
    {
      "epoch": 3.71,
      "eval_loss": 2.878877878189087,
      "eval_runtime": 5.3054,
      "eval_samples_per_second": 1861.705,
      "eval_steps_per_second": 14.702,
      "eval_top3_3_weighted_f1_score ": 0.4211972428050605,
      "eval_top_1_macro_f1_score": 0.05865004403272165,
      "eval_top_1_weighted_f1score": 0.20404542161091987,
      "eval_top_3_macro_f1_score": 0.15737159298688402,
      "step": 8640
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0002,
      "loss": 2.8764,
      "step": 8645
    },
    {
      "epoch": 3.71,
      "learning_rate": 0.0002,
      "loss": 2.8805,
      "step": 8650
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0002,
      "loss": 2.8693,
      "step": 8655
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0002,
      "loss": 2.8963,
      "step": 8660
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0002,
      "loss": 2.8858,
      "step": 8665
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0002,
      "loss": 2.8833,
      "step": 8670
    },
    {
      "epoch": 3.72,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 8675
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.0002,
      "loss": 2.8789,
      "step": 8680
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.0002,
      "loss": 2.8966,
      "step": 8685
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.0002,
      "loss": 2.878,
      "step": 8690
    },
    {
      "epoch": 3.73,
      "learning_rate": 0.0002,
      "loss": 2.8832,
      "step": 8695
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0002,
      "loss": 2.8737,
      "step": 8700
    },
    {
      "epoch": 3.74,
      "eval_loss": 2.8780808448791504,
      "eval_runtime": 5.458,
      "eval_samples_per_second": 1809.643,
      "eval_steps_per_second": 14.291,
      "eval_top3_3_weighted_f1_score ": 0.42025603415139495,
      "eval_top_1_macro_f1_score": 0.06226039136025313,
      "eval_top_1_weighted_f1score": 0.2016970828248235,
      "eval_top_3_macro_f1_score": 0.16186434784527917,
      "step": 8704
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0002,
      "loss": 2.8802,
      "step": 8705
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0002,
      "loss": 2.899,
      "step": 8710
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0002,
      "loss": 2.8758,
      "step": 8715
    },
    {
      "epoch": 3.74,
      "learning_rate": 0.0002,
      "loss": 2.8736,
      "step": 8720
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0002,
      "loss": 2.8747,
      "step": 8725
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 8730
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0002,
      "loss": 2.8711,
      "step": 8735
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 8740
    },
    {
      "epoch": 3.75,
      "learning_rate": 0.0002,
      "loss": 2.8926,
      "step": 8745
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.0002,
      "loss": 2.8734,
      "step": 8750
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.0002,
      "loss": 2.8876,
      "step": 8755
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.0002,
      "loss": 2.8947,
      "step": 8760
    },
    {
      "epoch": 3.76,
      "learning_rate": 0.0002,
      "loss": 2.8779,
      "step": 8765
    },
    {
      "epoch": 3.76,
      "eval_loss": 2.8843703269958496,
      "eval_runtime": 5.7769,
      "eval_samples_per_second": 1709.743,
      "eval_steps_per_second": 13.502,
      "eval_top3_3_weighted_f1_score ": 0.4199394713414115,
      "eval_top_1_macro_f1_score": 0.061616820845209974,
      "eval_top_1_weighted_f1score": 0.20079067358787378,
      "eval_top_3_macro_f1_score": 0.1595062776335458,
      "step": 8768
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0002,
      "loss": 2.8758,
      "step": 8770
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0002,
      "loss": 2.8699,
      "step": 8775
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0002,
      "loss": 2.8794,
      "step": 8780
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 8785
    },
    {
      "epoch": 3.77,
      "learning_rate": 0.0002,
      "loss": 2.8858,
      "step": 8790
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0002,
      "loss": 2.8732,
      "step": 8795
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0002,
      "loss": 2.8762,
      "step": 8800
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0002,
      "loss": 2.8577,
      "step": 8805
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 8810
    },
    {
      "epoch": 3.78,
      "learning_rate": 0.0002,
      "loss": 2.9012,
      "step": 8815
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 8820
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0002,
      "loss": 2.8711,
      "step": 8825
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0002,
      "loss": 2.8715,
      "step": 8830
    },
    {
      "epoch": 3.79,
      "eval_loss": 2.886676073074341,
      "eval_runtime": 5.3544,
      "eval_samples_per_second": 1844.649,
      "eval_steps_per_second": 14.567,
      "eval_top3_3_weighted_f1_score ": 0.4182482749371767,
      "eval_top_1_macro_f1_score": 0.06164296965374765,
      "eval_top_1_weighted_f1score": 0.20366121612865182,
      "eval_top_3_macro_f1_score": 0.157679707876537,
      "step": 8832
    },
    {
      "epoch": 3.79,
      "learning_rate": 0.0002,
      "loss": 2.871,
      "step": 8835
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 8840
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0002,
      "loss": 2.8915,
      "step": 8845
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0002,
      "loss": 2.8595,
      "step": 8850
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0002,
      "loss": 2.8813,
      "step": 8855
    },
    {
      "epoch": 3.8,
      "learning_rate": 0.0002,
      "loss": 2.878,
      "step": 8860
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0002,
      "loss": 2.8922,
      "step": 8865
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 8870
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0002,
      "loss": 2.8853,
      "step": 8875
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0002,
      "loss": 2.8897,
      "step": 8880
    },
    {
      "epoch": 3.81,
      "learning_rate": 0.0002,
      "loss": 2.8917,
      "step": 8885
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 8890
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.0002,
      "loss": 2.891,
      "step": 8895
    },
    {
      "epoch": 3.82,
      "eval_loss": 2.885464668273926,
      "eval_runtime": 5.294,
      "eval_samples_per_second": 1865.685,
      "eval_steps_per_second": 14.734,
      "eval_top3_3_weighted_f1_score ": 0.42080617949576915,
      "eval_top_1_macro_f1_score": 0.056517574070011053,
      "eval_top_1_weighted_f1score": 0.2049458290344821,
      "eval_top_3_macro_f1_score": 0.15343116095089573,
      "step": 8896
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.0002,
      "loss": 2.8686,
      "step": 8900
    },
    {
      "epoch": 3.82,
      "learning_rate": 0.0002,
      "loss": 2.8812,
      "step": 8905
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0002,
      "loss": 2.8742,
      "step": 8910
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0002,
      "loss": 2.8812,
      "step": 8915
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0002,
      "loss": 2.888,
      "step": 8920
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0002,
      "loss": 2.887,
      "step": 8925
    },
    {
      "epoch": 3.83,
      "learning_rate": 0.0002,
      "loss": 2.8786,
      "step": 8930
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0002,
      "loss": 2.8878,
      "step": 8935
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0002,
      "loss": 2.8726,
      "step": 8940
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 8945
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 8950
    },
    {
      "epoch": 3.84,
      "learning_rate": 0.0002,
      "loss": 2.8728,
      "step": 8955
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.0002,
      "loss": 2.8796,
      "step": 8960
    },
    {
      "epoch": 3.85,
      "eval_loss": 2.880286931991577,
      "eval_runtime": 5.3448,
      "eval_samples_per_second": 1847.978,
      "eval_steps_per_second": 14.594,
      "eval_top3_3_weighted_f1_score ": 0.4218945381783791,
      "eval_top_1_macro_f1_score": 0.06158161969228803,
      "eval_top_1_weighted_f1score": 0.20375764386011855,
      "eval_top_3_macro_f1_score": 0.1568737935791377,
      "step": 8960
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.0002,
      "loss": 2.8809,
      "step": 8965
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.0002,
      "loss": 2.8947,
      "step": 8970
    },
    {
      "epoch": 3.85,
      "learning_rate": 0.0002,
      "loss": 2.8879,
      "step": 8975
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0002,
      "loss": 2.9108,
      "step": 8980
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0002,
      "loss": 2.8772,
      "step": 8985
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0002,
      "loss": 2.8653,
      "step": 8990
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0002,
      "loss": 2.8982,
      "step": 8995
    },
    {
      "epoch": 3.86,
      "learning_rate": 0.0002,
      "loss": 2.8927,
      "step": 9000
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0002,
      "loss": 2.897,
      "step": 9005
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0002,
      "loss": 2.8872,
      "step": 9010
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0002,
      "loss": 2.8918,
      "step": 9015
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0002,
      "loss": 2.8942,
      "step": 9020
    },
    {
      "epoch": 3.87,
      "eval_loss": 2.879659652709961,
      "eval_runtime": 5.3314,
      "eval_samples_per_second": 1852.603,
      "eval_steps_per_second": 14.63,
      "eval_top3_3_weighted_f1_score ": 0.42280799021925125,
      "eval_top_1_macro_f1_score": 0.0616648047561624,
      "eval_top_1_weighted_f1score": 0.20212917796376262,
      "eval_top_3_macro_f1_score": 0.1632725285893378,
      "step": 9024
    },
    {
      "epoch": 3.87,
      "learning_rate": 0.0002,
      "loss": 2.8579,
      "step": 9025
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0002,
      "loss": 2.8849,
      "step": 9030
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0002,
      "loss": 2.8805,
      "step": 9035
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0002,
      "loss": 2.8848,
      "step": 9040
    },
    {
      "epoch": 3.88,
      "learning_rate": 0.0002,
      "loss": 2.8812,
      "step": 9045
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 9050
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.0002,
      "loss": 2.8795,
      "step": 9055
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.0002,
      "loss": 2.8658,
      "step": 9060
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.0002,
      "loss": 2.9151,
      "step": 9065
    },
    {
      "epoch": 3.89,
      "learning_rate": 0.0002,
      "loss": 2.8678,
      "step": 9070
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 9075
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0002,
      "loss": 2.8792,
      "step": 9080
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0002,
      "loss": 2.8661,
      "step": 9085
    },
    {
      "epoch": 3.9,
      "eval_loss": 2.878302812576294,
      "eval_runtime": 5.3561,
      "eval_samples_per_second": 1844.062,
      "eval_steps_per_second": 14.563,
      "eval_top3_3_weighted_f1_score ": 0.4218642511499269,
      "eval_top_1_macro_f1_score": 0.06042573034503991,
      "eval_top_1_weighted_f1score": 0.20167938826181162,
      "eval_top_3_macro_f1_score": 0.15948181652047955,
      "step": 9088
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0002,
      "loss": 2.8864,
      "step": 9090
    },
    {
      "epoch": 3.9,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 9095
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.0002,
      "loss": 2.8964,
      "step": 9100
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.0002,
      "loss": 2.8743,
      "step": 9105
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.0002,
      "loss": 2.8914,
      "step": 9110
    },
    {
      "epoch": 3.91,
      "learning_rate": 0.0002,
      "loss": 2.8743,
      "step": 9115
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 9120
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0002,
      "loss": 2.8921,
      "step": 9125
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0002,
      "loss": 2.8898,
      "step": 9130
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0002,
      "loss": 2.869,
      "step": 9135
    },
    {
      "epoch": 3.92,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 9140
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 9145
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 9150
    },
    {
      "epoch": 3.93,
      "eval_loss": 2.87255597114563,
      "eval_runtime": 5.3612,
      "eval_samples_per_second": 1842.319,
      "eval_steps_per_second": 14.549,
      "eval_top3_3_weighted_f1_score ": 0.4240820028763008,
      "eval_top_1_macro_f1_score": 0.0618924953113334,
      "eval_top_1_weighted_f1score": 0.20417364990423967,
      "eval_top_3_macro_f1_score": 0.16038084358769286,
      "step": 9152
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.0002,
      "loss": 2.8824,
      "step": 9155
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.0002,
      "loss": 2.8778,
      "step": 9160
    },
    {
      "epoch": 3.93,
      "learning_rate": 0.0002,
      "loss": 2.88,
      "step": 9165
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0002,
      "loss": 2.88,
      "step": 9170
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 9175
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 9180
    },
    {
      "epoch": 3.94,
      "learning_rate": 0.0002,
      "loss": 2.8689,
      "step": 9185
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0002,
      "loss": 2.9013,
      "step": 9190
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0002,
      "loss": 2.8834,
      "step": 9195
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0002,
      "loss": 2.8805,
      "step": 9200
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0002,
      "loss": 2.8832,
      "step": 9205
    },
    {
      "epoch": 3.95,
      "learning_rate": 0.0002,
      "loss": 2.8956,
      "step": 9210
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0002,
      "loss": 2.8945,
      "step": 9215
    },
    {
      "epoch": 3.96,
      "eval_loss": 2.8825714588165283,
      "eval_runtime": 5.3312,
      "eval_samples_per_second": 1852.669,
      "eval_steps_per_second": 14.631,
      "eval_top3_3_weighted_f1_score ": 0.4190590262717862,
      "eval_top_1_macro_f1_score": 0.06269843842904875,
      "eval_top_1_weighted_f1score": 0.2079376903452088,
      "eval_top_3_macro_f1_score": 0.16065438728345302,
      "step": 9216
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0002,
      "loss": 2.8708,
      "step": 9220
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0002,
      "loss": 2.8795,
      "step": 9225
    },
    {
      "epoch": 3.96,
      "learning_rate": 0.0002,
      "loss": 2.8845,
      "step": 9230
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0002,
      "loss": 2.869,
      "step": 9235
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 9240
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0002,
      "loss": 2.8997,
      "step": 9245
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0002,
      "loss": 2.8736,
      "step": 9250
    },
    {
      "epoch": 3.97,
      "learning_rate": 0.0002,
      "loss": 2.8864,
      "step": 9255
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0002,
      "loss": 2.8914,
      "step": 9260
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0002,
      "loss": 2.881,
      "step": 9265
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0002,
      "loss": 2.893,
      "step": 9270
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0002,
      "loss": 2.9017,
      "step": 9275
    },
    {
      "epoch": 3.98,
      "learning_rate": 0.0002,
      "loss": 2.879,
      "step": 9280
    },
    {
      "epoch": 3.98,
      "eval_loss": 2.8842124938964844,
      "eval_runtime": 5.3466,
      "eval_samples_per_second": 1847.352,
      "eval_steps_per_second": 14.589,
      "eval_top3_3_weighted_f1_score ": 0.42577666209976406,
      "eval_top_1_macro_f1_score": 0.06101519782251127,
      "eval_top_1_weighted_f1score": 0.20089361137055256,
      "eval_top_3_macro_f1_score": 0.1615738776964954,
      "step": 9280
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0002,
      "loss": 2.8843,
      "step": 9285
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 9290
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0002,
      "loss": 2.8929,
      "step": 9295
    },
    {
      "epoch": 3.99,
      "learning_rate": 0.0002,
      "loss": 2.8731,
      "step": 9300
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0002,
      "loss": 2.8743,
      "step": 9305
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 9310
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0002,
      "loss": 2.8812,
      "step": 9315
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0002,
      "loss": 3.1464,
      "step": 9320
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 9325
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0002,
      "loss": 2.8698,
      "step": 9330
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 9335
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0002,
      "loss": 2.8597,
      "step": 9340
    },
    {
      "epoch": 4.01,
      "eval_loss": 2.883638620376587,
      "eval_runtime": 5.7093,
      "eval_samples_per_second": 1729.988,
      "eval_steps_per_second": 13.662,
      "eval_top3_3_weighted_f1_score ": 0.4215743958712875,
      "eval_top_1_macro_f1_score": 0.06326665146672523,
      "eval_top_1_weighted_f1score": 0.20706063459634225,
      "eval_top_3_macro_f1_score": 0.164675083738132,
      "step": 9344
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 9345
    },
    {
      "epoch": 4.01,
      "learning_rate": 0.0002,
      "loss": 2.8808,
      "step": 9350
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.0002,
      "loss": 2.8727,
      "step": 9355
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.0002,
      "loss": 2.8586,
      "step": 9360
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.0002,
      "loss": 2.8823,
      "step": 9365
    },
    {
      "epoch": 4.02,
      "learning_rate": 0.0002,
      "loss": 2.8571,
      "step": 9370
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 9375
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0002,
      "loss": 2.8675,
      "step": 9380
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 9385
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0002,
      "loss": 2.8751,
      "step": 9390
    },
    {
      "epoch": 4.03,
      "learning_rate": 0.0002,
      "loss": 2.8718,
      "step": 9395
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.0002,
      "loss": 2.8751,
      "step": 9400
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.0002,
      "loss": 2.8854,
      "step": 9405
    },
    {
      "epoch": 4.04,
      "eval_loss": 2.8765015602111816,
      "eval_runtime": 5.3743,
      "eval_samples_per_second": 1837.804,
      "eval_steps_per_second": 14.513,
      "eval_top3_3_weighted_f1_score ": 0.41924298785763464,
      "eval_top_1_macro_f1_score": 0.06146762957996142,
      "eval_top_1_weighted_f1score": 0.20146868730144363,
      "eval_top_3_macro_f1_score": 0.16860583910350077,
      "step": 9408
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.0002,
      "loss": 2.88,
      "step": 9410
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.0002,
      "loss": 2.8792,
      "step": 9415
    },
    {
      "epoch": 4.04,
      "learning_rate": 0.0002,
      "loss": 2.8433,
      "step": 9420
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.0002,
      "loss": 2.8556,
      "step": 9425
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 9430
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.0002,
      "loss": 2.8753,
      "step": 9435
    },
    {
      "epoch": 4.05,
      "learning_rate": 0.0002,
      "loss": 2.8768,
      "step": 9440
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.0002,
      "loss": 2.8665,
      "step": 9445
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.0002,
      "loss": 2.8778,
      "step": 9450
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.0002,
      "loss": 2.8614,
      "step": 9455
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.0002,
      "loss": 2.8824,
      "step": 9460
    },
    {
      "epoch": 4.06,
      "learning_rate": 0.0002,
      "loss": 2.8729,
      "step": 9465
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0002,
      "loss": 2.8771,
      "step": 9470
    },
    {
      "epoch": 4.07,
      "eval_loss": 2.8779726028442383,
      "eval_runtime": 5.4094,
      "eval_samples_per_second": 1825.88,
      "eval_steps_per_second": 14.419,
      "eval_top3_3_weighted_f1_score ": 0.4252106970458161,
      "eval_top_1_macro_f1_score": 0.0641577432963229,
      "eval_top_1_weighted_f1score": 0.20700521183069107,
      "eval_top_3_macro_f1_score": 0.1656110416323952,
      "step": 9472
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 9475
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0002,
      "loss": 2.9054,
      "step": 9480
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0002,
      "loss": 2.8546,
      "step": 9485
    },
    {
      "epoch": 4.07,
      "learning_rate": 0.0002,
      "loss": 2.867,
      "step": 9490
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.0002,
      "loss": 2.8687,
      "step": 9495
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.0002,
      "loss": 2.8776,
      "step": 9500
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 9505
    },
    {
      "epoch": 4.08,
      "learning_rate": 0.0002,
      "loss": 2.8747,
      "step": 9510
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0002,
      "loss": 2.8784,
      "step": 9515
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0002,
      "loss": 2.8825,
      "step": 9520
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0002,
      "loss": 2.8616,
      "step": 9525
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0002,
      "loss": 2.8754,
      "step": 9530
    },
    {
      "epoch": 4.09,
      "learning_rate": 0.0002,
      "loss": 2.8616,
      "step": 9535
    },
    {
      "epoch": 4.09,
      "eval_loss": 2.8782520294189453,
      "eval_runtime": 5.3961,
      "eval_samples_per_second": 1830.397,
      "eval_steps_per_second": 14.455,
      "eval_top3_3_weighted_f1_score ": 0.4217188679247572,
      "eval_top_1_macro_f1_score": 0.059938420057228194,
      "eval_top_1_weighted_f1score": 0.20490516714063806,
      "eval_top_3_macro_f1_score": 0.15982189116819304,
      "step": 9536
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0002,
      "loss": 2.882,
      "step": 9540
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0002,
      "loss": 2.8954,
      "step": 9545
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0002,
      "loss": 2.8895,
      "step": 9550
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0002,
      "loss": 2.8846,
      "step": 9555
    },
    {
      "epoch": 4.1,
      "learning_rate": 0.0002,
      "loss": 2.8692,
      "step": 9560
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0002,
      "loss": 2.889,
      "step": 9565
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 9570
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0002,
      "loss": 2.8817,
      "step": 9575
    },
    {
      "epoch": 4.11,
      "learning_rate": 0.0002,
      "loss": 2.8755,
      "step": 9580
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0002,
      "loss": 2.8817,
      "step": 9585
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0002,
      "loss": 2.8787,
      "step": 9590
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0002,
      "loss": 2.8649,
      "step": 9595
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0002,
      "loss": 2.8577,
      "step": 9600
    },
    {
      "epoch": 4.12,
      "eval_loss": 2.87673020362854,
      "eval_runtime": 5.3716,
      "eval_samples_per_second": 1838.757,
      "eval_steps_per_second": 14.521,
      "eval_top3_3_weighted_f1_score ": 0.4241371909438834,
      "eval_top_1_macro_f1_score": 0.06164446683494285,
      "eval_top_1_weighted_f1score": 0.2063257533667319,
      "eval_top_3_macro_f1_score": 0.14961892013967493,
      "step": 9600
    },
    {
      "epoch": 4.12,
      "learning_rate": 0.0002,
      "loss": 2.866,
      "step": 9605
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.0002,
      "loss": 2.8934,
      "step": 9610
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.0002,
      "loss": 2.8551,
      "step": 9615
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.0002,
      "loss": 2.88,
      "step": 9620
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.0002,
      "loss": 2.8691,
      "step": 9625
    },
    {
      "epoch": 4.13,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 9630
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0002,
      "loss": 2.8476,
      "step": 9635
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0002,
      "loss": 2.8511,
      "step": 9640
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 9645
    },
    {
      "epoch": 4.14,
      "learning_rate": 0.0002,
      "loss": 2.8843,
      "step": 9650
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.0002,
      "loss": 2.8641,
      "step": 9655
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.0002,
      "loss": 2.8908,
      "step": 9660
    },
    {
      "epoch": 4.15,
      "eval_loss": 2.879879951477051,
      "eval_runtime": 5.3414,
      "eval_samples_per_second": 1849.131,
      "eval_steps_per_second": 14.603,
      "eval_top3_3_weighted_f1_score ": 0.42225217111103863,
      "eval_top_1_macro_f1_score": 0.060374192075967156,
      "eval_top_1_weighted_f1score": 0.20219577753251478,
      "eval_top_3_macro_f1_score": 0.16262665440212856,
      "step": 9664
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 9665
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.0002,
      "loss": 2.8792,
      "step": 9670
    },
    {
      "epoch": 4.15,
      "learning_rate": 0.0002,
      "loss": 2.8814,
      "step": 9675
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0002,
      "loss": 2.8846,
      "step": 9680
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 9685
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 9690
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0002,
      "loss": 2.8919,
      "step": 9695
    },
    {
      "epoch": 4.16,
      "learning_rate": 0.0002,
      "loss": 2.863,
      "step": 9700
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.0002,
      "loss": 2.8653,
      "step": 9705
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.0002,
      "loss": 2.8543,
      "step": 9710
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.0002,
      "loss": 2.8794,
      "step": 9715
    },
    {
      "epoch": 4.17,
      "learning_rate": 0.0002,
      "loss": 2.8426,
      "step": 9720
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 9725
    },
    {
      "epoch": 4.18,
      "eval_loss": 2.8804783821105957,
      "eval_runtime": 5.386,
      "eval_samples_per_second": 1833.837,
      "eval_steps_per_second": 14.482,
      "eval_top3_3_weighted_f1_score ": 0.42565165650821607,
      "eval_top_1_macro_f1_score": 0.060127552219321874,
      "eval_top_1_weighted_f1score": 0.20383335370478492,
      "eval_top_3_macro_f1_score": 0.1701485885248088,
      "step": 9728
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.0002,
      "loss": 2.8662,
      "step": 9730
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.0002,
      "loss": 2.8886,
      "step": 9735
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.0002,
      "loss": 2.8564,
      "step": 9740
    },
    {
      "epoch": 4.18,
      "learning_rate": 0.0002,
      "loss": 2.8761,
      "step": 9745
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.0002,
      "loss": 2.8767,
      "step": 9750
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.0002,
      "loss": 2.8746,
      "step": 9755
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.0002,
      "loss": 2.8851,
      "step": 9760
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.0002,
      "loss": 2.8975,
      "step": 9765
    },
    {
      "epoch": 4.19,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 9770
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 9775
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 9780
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 9785
    },
    {
      "epoch": 4.2,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 9790
    },
    {
      "epoch": 4.2,
      "eval_loss": 2.8811566829681396,
      "eval_runtime": 5.3688,
      "eval_samples_per_second": 1839.718,
      "eval_steps_per_second": 14.529,
      "eval_top3_3_weighted_f1_score ": 0.4227489382471628,
      "eval_top_1_macro_f1_score": 0.062052743274787756,
      "eval_top_1_weighted_f1score": 0.20170422612536257,
      "eval_top_3_macro_f1_score": 0.15961342505729212,
      "step": 9792
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0002,
      "loss": 2.8511,
      "step": 9795
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0002,
      "loss": 2.8552,
      "step": 9800
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0002,
      "loss": 2.9049,
      "step": 9805
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0002,
      "loss": 2.8756,
      "step": 9810
    },
    {
      "epoch": 4.21,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 9815
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0002,
      "loss": 2.8916,
      "step": 9820
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0002,
      "loss": 2.8717,
      "step": 9825
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 9830
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 9835
    },
    {
      "epoch": 4.22,
      "learning_rate": 0.0002,
      "loss": 2.8932,
      "step": 9840
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0002,
      "loss": 2.8731,
      "step": 9845
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0002,
      "loss": 2.8679,
      "step": 9850
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0002,
      "loss": 2.8706,
      "step": 9855
    },
    {
      "epoch": 4.23,
      "eval_loss": 2.879899024963379,
      "eval_runtime": 5.3621,
      "eval_samples_per_second": 1841.988,
      "eval_steps_per_second": 14.546,
      "eval_top3_3_weighted_f1_score ": 0.4279835897973105,
      "eval_top_1_macro_f1_score": 0.05889265059198331,
      "eval_top_1_weighted_f1score": 0.205083467453559,
      "eval_top_3_macro_f1_score": 0.15827034485660252,
      "step": 9856
    },
    {
      "epoch": 4.23,
      "learning_rate": 0.0002,
      "loss": 2.8669,
      "step": 9860
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.0002,
      "loss": 2.8931,
      "step": 9865
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.0002,
      "loss": 2.8669,
      "step": 9870
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 9875
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 9880
    },
    {
      "epoch": 4.24,
      "learning_rate": 0.0002,
      "loss": 2.8835,
      "step": 9885
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0002,
      "loss": 2.8886,
      "step": 9890
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0002,
      "loss": 2.885,
      "step": 9895
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0002,
      "loss": 2.8952,
      "step": 9900
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0002,
      "loss": 2.8859,
      "step": 9905
    },
    {
      "epoch": 4.25,
      "learning_rate": 0.0002,
      "loss": 2.8761,
      "step": 9910
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.0002,
      "loss": 2.8874,
      "step": 9915
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 9920
    },
    {
      "epoch": 4.26,
      "eval_loss": 2.8851325511932373,
      "eval_runtime": 5.3333,
      "eval_samples_per_second": 1851.933,
      "eval_steps_per_second": 14.625,
      "eval_top3_3_weighted_f1_score ": 0.422462627894626,
      "eval_top_1_macro_f1_score": 0.06341811301595834,
      "eval_top_1_weighted_f1score": 0.20387972505464994,
      "eval_top_3_macro_f1_score": 0.16272604470885255,
      "step": 9920
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.0002,
      "loss": 2.8729,
      "step": 9925
    },
    {
      "epoch": 4.26,
      "learning_rate": 0.0002,
      "loss": 2.8885,
      "step": 9930
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0002,
      "loss": 2.8657,
      "step": 9935
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0002,
      "loss": 2.8847,
      "step": 9940
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0002,
      "loss": 2.8715,
      "step": 9945
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0002,
      "loss": 2.8726,
      "step": 9950
    },
    {
      "epoch": 4.27,
      "learning_rate": 0.0002,
      "loss": 2.8639,
      "step": 9955
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.0002,
      "loss": 2.868,
      "step": 9960
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.0002,
      "loss": 2.8602,
      "step": 9965
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 9970
    },
    {
      "epoch": 4.28,
      "learning_rate": 0.0002,
      "loss": 2.8811,
      "step": 9975
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0002,
      "loss": 2.8497,
      "step": 9980
    },
    {
      "epoch": 4.29,
      "eval_loss": 2.8782029151916504,
      "eval_runtime": 5.3368,
      "eval_samples_per_second": 1850.72,
      "eval_steps_per_second": 14.615,
      "eval_top3_3_weighted_f1_score ": 0.42270002035339793,
      "eval_top_1_macro_f1_score": 0.05897566452561095,
      "eval_top_1_weighted_f1score": 0.20188161923476727,
      "eval_top_3_macro_f1_score": 0.16398597778615756,
      "step": 9984
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 9985
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0002,
      "loss": 2.8789,
      "step": 9990
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0002,
      "loss": 2.8553,
      "step": 9995
    },
    {
      "epoch": 4.29,
      "learning_rate": 0.0002,
      "loss": 2.8945,
      "step": 10000
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 10005
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.0002,
      "loss": 2.8449,
      "step": 10010
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.0002,
      "loss": 2.8841,
      "step": 10015
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.0002,
      "loss": 2.8727,
      "step": 10020
    },
    {
      "epoch": 4.3,
      "learning_rate": 0.0002,
      "loss": 2.8746,
      "step": 10025
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.0002,
      "loss": 2.8733,
      "step": 10030
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.0002,
      "loss": 2.8732,
      "step": 10035
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.0002,
      "loss": 2.872,
      "step": 10040
    },
    {
      "epoch": 4.31,
      "learning_rate": 0.0002,
      "loss": 2.8799,
      "step": 10045
    },
    {
      "epoch": 4.31,
      "eval_loss": 2.880223512649536,
      "eval_runtime": 5.4218,
      "eval_samples_per_second": 1821.72,
      "eval_steps_per_second": 14.386,
      "eval_top3_3_weighted_f1_score ": 0.4247351525921999,
      "eval_top_1_macro_f1_score": 0.06044784376357685,
      "eval_top_1_weighted_f1score": 0.20083124943581332,
      "eval_top_3_macro_f1_score": 0.1675721411499172,
      "step": 10048
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 10050
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0002,
      "loss": 2.8763,
      "step": 10055
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0002,
      "loss": 2.8823,
      "step": 10060
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 10065
    },
    {
      "epoch": 4.32,
      "learning_rate": 0.0002,
      "loss": 2.8798,
      "step": 10070
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0002,
      "loss": 2.8715,
      "step": 10075
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0002,
      "loss": 2.8782,
      "step": 10080
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0002,
      "loss": 2.8802,
      "step": 10085
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 10090
    },
    {
      "epoch": 4.33,
      "learning_rate": 0.0002,
      "loss": 2.8872,
      "step": 10095
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 10100
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.0002,
      "loss": 2.8765,
      "step": 10105
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.0002,
      "loss": 2.884,
      "step": 10110
    },
    {
      "epoch": 4.34,
      "eval_loss": 2.8734776973724365,
      "eval_runtime": 139.1467,
      "eval_samples_per_second": 70.983,
      "eval_steps_per_second": 0.561,
      "eval_top3_3_weighted_f1_score ": 0.4308448068453806,
      "eval_top_1_macro_f1_score": 0.06091145198219989,
      "eval_top_1_weighted_f1score": 0.20389321132048763,
      "eval_top_3_macro_f1_score": 0.16629805337811587,
      "step": 10112
    },
    {
      "epoch": 4.34,
      "learning_rate": 0.0002,
      "loss": 2.8456,
      "step": 10115
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 10120
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.0002,
      "loss": 2.8585,
      "step": 10125
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.0002,
      "loss": 2.8713,
      "step": 10130
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.0002,
      "loss": 2.866,
      "step": 10135
    },
    {
      "epoch": 4.35,
      "learning_rate": 0.0002,
      "loss": 2.8866,
      "step": 10140
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0002,
      "loss": 2.8774,
      "step": 10145
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0002,
      "loss": 2.8601,
      "step": 10150
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 10155
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0002,
      "loss": 2.8821,
      "step": 10160
    },
    {
      "epoch": 4.36,
      "learning_rate": 0.0002,
      "loss": 2.865,
      "step": 10165
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.0002,
      "loss": 2.8653,
      "step": 10170
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.0002,
      "loss": 2.8674,
      "step": 10175
    },
    {
      "epoch": 4.37,
      "eval_loss": 2.879244327545166,
      "eval_runtime": 5.4222,
      "eval_samples_per_second": 1821.597,
      "eval_steps_per_second": 14.385,
      "eval_top3_3_weighted_f1_score ": 0.4266599016089825,
      "eval_top_1_macro_f1_score": 0.06294538988470444,
      "eval_top_1_weighted_f1score": 0.206026998323378,
      "eval_top_3_macro_f1_score": 0.17407376498147326,
      "step": 10176
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 10180
    },
    {
      "epoch": 4.37,
      "learning_rate": 0.0002,
      "loss": 2.8701,
      "step": 10185
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 10190
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0002,
      "loss": 2.8613,
      "step": 10195
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 10200
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0002,
      "loss": 2.8824,
      "step": 10205
    },
    {
      "epoch": 4.38,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 10210
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0002,
      "loss": 2.8734,
      "step": 10215
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0002,
      "loss": 2.8711,
      "step": 10220
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0002,
      "loss": 2.8716,
      "step": 10225
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 10230
    },
    {
      "epoch": 4.39,
      "learning_rate": 0.0002,
      "loss": 2.8757,
      "step": 10235
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0002,
      "loss": 2.8691,
      "step": 10240
    },
    {
      "epoch": 4.4,
      "eval_loss": 2.879953622817993,
      "eval_runtime": 5.3715,
      "eval_samples_per_second": 1838.77,
      "eval_steps_per_second": 14.521,
      "eval_top3_3_weighted_f1_score ": 0.4198328743907584,
      "eval_top_1_macro_f1_score": 0.06120930163098784,
      "eval_top_1_weighted_f1score": 0.20363865109484638,
      "eval_top_3_macro_f1_score": 0.16543035959168473,
      "step": 10240
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0002,
      "loss": 2.8769,
      "step": 10245
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0002,
      "loss": 2.8655,
      "step": 10250
    },
    {
      "epoch": 4.4,
      "learning_rate": 0.0002,
      "loss": 2.8938,
      "step": 10255
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0002,
      "loss": 2.8763,
      "step": 10260
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0002,
      "loss": 2.8786,
      "step": 10265
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0002,
      "loss": 2.855,
      "step": 10270
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0002,
      "loss": 2.8589,
      "step": 10275
    },
    {
      "epoch": 4.41,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 10280
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0002,
      "loss": 2.8724,
      "step": 10285
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0002,
      "loss": 2.8898,
      "step": 10290
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0002,
      "loss": 2.881,
      "step": 10295
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0002,
      "loss": 2.8583,
      "step": 10300
    },
    {
      "epoch": 4.42,
      "eval_loss": 2.8827550411224365,
      "eval_runtime": 5.4789,
      "eval_samples_per_second": 1802.738,
      "eval_steps_per_second": 14.236,
      "eval_top3_3_weighted_f1_score ": 0.42347784477898887,
      "eval_top_1_macro_f1_score": 0.06256916446524377,
      "eval_top_1_weighted_f1score": 0.20833558425133877,
      "eval_top_3_macro_f1_score": 0.16414994500053223,
      "step": 10304
    },
    {
      "epoch": 4.42,
      "learning_rate": 0.0002,
      "loss": 2.8842,
      "step": 10305
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0002,
      "loss": 2.8786,
      "step": 10310
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0002,
      "loss": 2.8845,
      "step": 10315
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 10320
    },
    {
      "epoch": 4.43,
      "learning_rate": 0.0002,
      "loss": 2.8784,
      "step": 10325
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.0002,
      "loss": 2.8683,
      "step": 10330
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.0002,
      "loss": 2.8775,
      "step": 10335
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 10340
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.0002,
      "loss": 2.8847,
      "step": 10345
    },
    {
      "epoch": 4.44,
      "learning_rate": 0.0002,
      "loss": 2.8616,
      "step": 10350
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0002,
      "loss": 2.8863,
      "step": 10355
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0002,
      "loss": 2.8621,
      "step": 10360
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0002,
      "loss": 2.8814,
      "step": 10365
    },
    {
      "epoch": 4.45,
      "eval_loss": 2.8773937225341797,
      "eval_runtime": 5.3195,
      "eval_samples_per_second": 1856.755,
      "eval_steps_per_second": 14.663,
      "eval_top3_3_weighted_f1_score ": 0.4253007826286828,
      "eval_top_1_macro_f1_score": 0.06625233119299727,
      "eval_top_1_weighted_f1score": 0.20676790414176083,
      "eval_top_3_macro_f1_score": 0.15646460587288663,
      "step": 10368
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 10370
    },
    {
      "epoch": 4.45,
      "learning_rate": 0.0002,
      "loss": 2.8784,
      "step": 10375
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0002,
      "loss": 2.8737,
      "step": 10380
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0002,
      "loss": 2.8854,
      "step": 10385
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0002,
      "loss": 2.8788,
      "step": 10390
    },
    {
      "epoch": 4.46,
      "learning_rate": 0.0002,
      "loss": 2.8855,
      "step": 10395
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0002,
      "loss": 2.8558,
      "step": 10400
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 10405
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0002,
      "loss": 2.8695,
      "step": 10410
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0002,
      "loss": 2.8685,
      "step": 10415
    },
    {
      "epoch": 4.47,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 10420
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0002,
      "loss": 2.8704,
      "step": 10425
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0002,
      "loss": 2.8757,
      "step": 10430
    },
    {
      "epoch": 4.48,
      "eval_loss": 2.874392032623291,
      "eval_runtime": 5.3568,
      "eval_samples_per_second": 1843.813,
      "eval_steps_per_second": 14.561,
      "eval_top3_3_weighted_f1_score ": 0.4241647649747962,
      "eval_top_1_macro_f1_score": 0.06217161351615617,
      "eval_top_1_weighted_f1score": 0.20898908527186977,
      "eval_top_3_macro_f1_score": 0.1573498335384943,
      "step": 10432
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0002,
      "loss": 2.8822,
      "step": 10435
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0002,
      "loss": 2.8756,
      "step": 10440
    },
    {
      "epoch": 4.48,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 10445
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 10450
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0002,
      "loss": 2.8757,
      "step": 10455
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0002,
      "loss": 2.8917,
      "step": 10460
    },
    {
      "epoch": 4.49,
      "learning_rate": 0.0002,
      "loss": 2.8637,
      "step": 10465
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.0002,
      "loss": 2.8601,
      "step": 10470
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.0002,
      "loss": 2.8885,
      "step": 10475
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.0002,
      "loss": 2.8794,
      "step": 10480
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.0002,
      "loss": 2.8707,
      "step": 10485
    },
    {
      "epoch": 4.5,
      "learning_rate": 0.0002,
      "loss": 2.8813,
      "step": 10490
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0002,
      "loss": 2.8719,
      "step": 10495
    },
    {
      "epoch": 4.51,
      "eval_loss": 2.8797762393951416,
      "eval_runtime": 5.403,
      "eval_samples_per_second": 1828.074,
      "eval_steps_per_second": 14.437,
      "eval_top3_3_weighted_f1_score ": 0.4272453879809695,
      "eval_top_1_macro_f1_score": 0.06766294229572796,
      "eval_top_1_weighted_f1score": 0.20884783338044238,
      "eval_top_3_macro_f1_score": 0.1722198495559193,
      "step": 10496
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0002,
      "loss": 2.8728,
      "step": 10500
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 10505
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 10510
    },
    {
      "epoch": 4.51,
      "learning_rate": 0.0002,
      "loss": 2.8629,
      "step": 10515
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.0002,
      "loss": 2.863,
      "step": 10520
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.0002,
      "loss": 2.8742,
      "step": 10525
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.0002,
      "loss": 2.8855,
      "step": 10530
    },
    {
      "epoch": 4.52,
      "learning_rate": 0.0002,
      "loss": 2.8857,
      "step": 10535
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 10540
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0002,
      "loss": 2.868,
      "step": 10545
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0002,
      "loss": 2.8471,
      "step": 10550
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0002,
      "loss": 2.8562,
      "step": 10555
    },
    {
      "epoch": 4.53,
      "learning_rate": 0.0002,
      "loss": 2.8585,
      "step": 10560
    },
    {
      "epoch": 4.53,
      "eval_loss": 2.8755102157592773,
      "eval_runtime": 5.5079,
      "eval_samples_per_second": 1793.255,
      "eval_steps_per_second": 14.162,
      "eval_top3_3_weighted_f1_score ": 0.4226135577632516,
      "eval_top_1_macro_f1_score": 0.060655061707233185,
      "eval_top_1_weighted_f1score": 0.2049577373490113,
      "eval_top_3_macro_f1_score": 0.16153021748453586,
      "step": 10560
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.0002,
      "loss": 2.8577,
      "step": 10565
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.0002,
      "loss": 2.8732,
      "step": 10570
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.0002,
      "loss": 2.8692,
      "step": 10575
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 10580
    },
    {
      "epoch": 4.54,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 10585
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.0002,
      "loss": 2.865,
      "step": 10590
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.0002,
      "loss": 2.8874,
      "step": 10595
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.0002,
      "loss": 2.8683,
      "step": 10600
    },
    {
      "epoch": 4.55,
      "learning_rate": 0.0002,
      "loss": 2.8801,
      "step": 10605
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.0002,
      "loss": 2.8742,
      "step": 10610
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.0002,
      "loss": 2.8899,
      "step": 10615
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.0002,
      "loss": 2.8946,
      "step": 10620
    },
    {
      "epoch": 4.56,
      "eval_loss": 2.8786725997924805,
      "eval_runtime": 5.3244,
      "eval_samples_per_second": 1855.061,
      "eval_steps_per_second": 14.65,
      "eval_top3_3_weighted_f1_score ": 0.42379557588689853,
      "eval_top_1_macro_f1_score": 0.05847571229304535,
      "eval_top_1_weighted_f1score": 0.2037950547430509,
      "eval_top_3_macro_f1_score": 0.16553631749588468,
      "step": 10624
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 10625
    },
    {
      "epoch": 4.56,
      "learning_rate": 0.0002,
      "loss": 2.8722,
      "step": 10630
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0002,
      "loss": 2.8608,
      "step": 10635
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0002,
      "loss": 2.8459,
      "step": 10640
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 10645
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0002,
      "loss": 2.8787,
      "step": 10650
    },
    {
      "epoch": 4.57,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 10655
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0002,
      "loss": 2.8665,
      "step": 10660
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0002,
      "loss": 2.8937,
      "step": 10665
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0002,
      "loss": 2.8568,
      "step": 10670
    },
    {
      "epoch": 4.58,
      "learning_rate": 0.0002,
      "loss": 2.8775,
      "step": 10675
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0002,
      "loss": 2.8685,
      "step": 10680
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0002,
      "loss": 2.8719,
      "step": 10685
    },
    {
      "epoch": 4.59,
      "eval_loss": 2.8753890991210938,
      "eval_runtime": 5.3311,
      "eval_samples_per_second": 1852.715,
      "eval_steps_per_second": 14.631,
      "eval_top3_3_weighted_f1_score ": 0.42260784607554397,
      "eval_top_1_macro_f1_score": 0.06287292865160847,
      "eval_top_1_weighted_f1score": 0.2094794253009507,
      "eval_top_3_macro_f1_score": 0.1610328994366965,
      "step": 10688
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0002,
      "loss": 2.8832,
      "step": 10690
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0002,
      "loss": 2.8564,
      "step": 10695
    },
    {
      "epoch": 4.59,
      "learning_rate": 0.0002,
      "loss": 2.8732,
      "step": 10700
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 10705
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0002,
      "loss": 2.8663,
      "step": 10710
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 10715
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 10720
    },
    {
      "epoch": 4.6,
      "learning_rate": 0.0002,
      "loss": 2.8826,
      "step": 10725
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.0002,
      "loss": 2.8555,
      "step": 10730
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.0002,
      "loss": 2.8914,
      "step": 10735
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.0002,
      "loss": 2.8661,
      "step": 10740
    },
    {
      "epoch": 4.61,
      "learning_rate": 0.0002,
      "loss": 2.8907,
      "step": 10745
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0002,
      "loss": 2.8685,
      "step": 10750
    },
    {
      "epoch": 4.62,
      "eval_loss": 2.879948377609253,
      "eval_runtime": 5.4339,
      "eval_samples_per_second": 1817.655,
      "eval_steps_per_second": 14.354,
      "eval_top3_3_weighted_f1_score ": 0.4235817003770342,
      "eval_top_1_macro_f1_score": 0.06705393736968994,
      "eval_top_1_weighted_f1score": 0.20217185922297795,
      "eval_top_3_macro_f1_score": 0.17065551134269813,
      "step": 10752
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0002,
      "loss": 2.8665,
      "step": 10755
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0002,
      "loss": 2.8751,
      "step": 10760
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0002,
      "loss": 2.863,
      "step": 10765
    },
    {
      "epoch": 4.62,
      "learning_rate": 0.0002,
      "loss": 2.8561,
      "step": 10770
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0002,
      "loss": 2.8703,
      "step": 10775
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 10780
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 10785
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 10790
    },
    {
      "epoch": 4.63,
      "learning_rate": 0.0002,
      "loss": 2.8662,
      "step": 10795
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.0002,
      "loss": 2.8836,
      "step": 10800
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 10805
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.0002,
      "loss": 2.8687,
      "step": 10810
    },
    {
      "epoch": 4.64,
      "learning_rate": 0.0002,
      "loss": 2.8697,
      "step": 10815
    },
    {
      "epoch": 4.64,
      "eval_loss": 2.875290870666504,
      "eval_runtime": 5.4802,
      "eval_samples_per_second": 1802.315,
      "eval_steps_per_second": 14.233,
      "eval_top3_3_weighted_f1_score ": 0.4247873291971013,
      "eval_top_1_macro_f1_score": 0.06214410102775452,
      "eval_top_1_weighted_f1score": 0.2042170409560508,
      "eval_top_3_macro_f1_score": 0.17027586444340612,
      "step": 10816
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0002,
      "loss": 2.8723,
      "step": 10820
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 10825
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0002,
      "loss": 2.8828,
      "step": 10830
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0002,
      "loss": 2.8813,
      "step": 10835
    },
    {
      "epoch": 4.65,
      "learning_rate": 0.0002,
      "loss": 2.8754,
      "step": 10840
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0002,
      "loss": 2.8569,
      "step": 10845
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0002,
      "loss": 2.8773,
      "step": 10850
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0002,
      "loss": 2.8842,
      "step": 10855
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 10860
    },
    {
      "epoch": 4.66,
      "learning_rate": 0.0002,
      "loss": 2.8704,
      "step": 10865
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.0002,
      "loss": 2.8872,
      "step": 10870
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.0002,
      "loss": 2.8667,
      "step": 10875
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.0002,
      "loss": 2.8772,
      "step": 10880
    },
    {
      "epoch": 4.67,
      "eval_loss": 2.8746321201324463,
      "eval_runtime": 5.3791,
      "eval_samples_per_second": 1836.176,
      "eval_steps_per_second": 14.501,
      "eval_top3_3_weighted_f1_score ": 0.4204443013055093,
      "eval_top_1_macro_f1_score": 0.0693137360049466,
      "eval_top_1_weighted_f1score": 0.2055956875544066,
      "eval_top_3_macro_f1_score": 0.16231714288231683,
      "step": 10880
    },
    {
      "epoch": 4.67,
      "learning_rate": 0.0002,
      "loss": 2.8689,
      "step": 10885
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0002,
      "loss": 2.8821,
      "step": 10890
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0002,
      "loss": 2.8656,
      "step": 10895
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0002,
      "loss": 2.8871,
      "step": 10900
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0002,
      "loss": 2.8863,
      "step": 10905
    },
    {
      "epoch": 4.68,
      "learning_rate": 0.0002,
      "loss": 2.868,
      "step": 10910
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.0002,
      "loss": 2.8785,
      "step": 10915
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.0002,
      "loss": 2.8675,
      "step": 10920
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.0002,
      "loss": 2.879,
      "step": 10925
    },
    {
      "epoch": 4.69,
      "learning_rate": 0.0002,
      "loss": 2.8627,
      "step": 10930
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.0002,
      "loss": 2.8582,
      "step": 10935
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.0002,
      "loss": 2.8894,
      "step": 10940
    },
    {
      "epoch": 4.7,
      "eval_loss": 2.8700804710388184,
      "eval_runtime": 5.4134,
      "eval_samples_per_second": 1824.558,
      "eval_steps_per_second": 14.409,
      "eval_top3_3_weighted_f1_score ": 0.42589136835292835,
      "eval_top_1_macro_f1_score": 0.059708047360599786,
      "eval_top_1_weighted_f1score": 0.20443041923958735,
      "eval_top_3_macro_f1_score": 0.16975798713822587,
      "step": 10944
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.0002,
      "loss": 2.8855,
      "step": 10945
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.0002,
      "loss": 2.858,
      "step": 10950
    },
    {
      "epoch": 4.7,
      "learning_rate": 0.0002,
      "loss": 2.8793,
      "step": 10955
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.0002,
      "loss": 2.889,
      "step": 10960
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.0002,
      "loss": 2.8608,
      "step": 10965
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.0002,
      "loss": 2.8833,
      "step": 10970
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.0002,
      "loss": 2.8788,
      "step": 10975
    },
    {
      "epoch": 4.71,
      "learning_rate": 0.0002,
      "loss": 2.8621,
      "step": 10980
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.0002,
      "loss": 2.8855,
      "step": 10985
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.0002,
      "loss": 2.8735,
      "step": 10990
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.0002,
      "loss": 2.8818,
      "step": 10995
    },
    {
      "epoch": 4.72,
      "learning_rate": 0.0002,
      "loss": 2.8554,
      "step": 11000
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.0002,
      "loss": 2.8757,
      "step": 11005
    },
    {
      "epoch": 4.73,
      "eval_loss": 2.871457099914551,
      "eval_runtime": 5.416,
      "eval_samples_per_second": 1823.672,
      "eval_steps_per_second": 14.402,
      "eval_top3_3_weighted_f1_score ": 0.42616063103307444,
      "eval_top_1_macro_f1_score": 0.06361175134933073,
      "eval_top_1_weighted_f1score": 0.20556103154493974,
      "eval_top_3_macro_f1_score": 0.17060698685157544,
      "step": 11008
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.0002,
      "loss": 2.8743,
      "step": 11010
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.0002,
      "loss": 2.8618,
      "step": 11015
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.0002,
      "loss": 2.8839,
      "step": 11020
    },
    {
      "epoch": 4.73,
      "learning_rate": 0.0002,
      "loss": 2.8524,
      "step": 11025
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.0002,
      "loss": 2.8652,
      "step": 11030
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 11035
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 11040
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.0002,
      "loss": 2.882,
      "step": 11045
    },
    {
      "epoch": 4.74,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 11050
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.0002,
      "loss": 2.8925,
      "step": 11055
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.0002,
      "loss": 2.8674,
      "step": 11060
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.0002,
      "loss": 2.8573,
      "step": 11065
    },
    {
      "epoch": 4.75,
      "learning_rate": 0.0002,
      "loss": 2.8725,
      "step": 11070
    },
    {
      "epoch": 4.75,
      "eval_loss": 2.870499610900879,
      "eval_runtime": 5.3558,
      "eval_samples_per_second": 1844.158,
      "eval_steps_per_second": 14.564,
      "eval_top3_3_weighted_f1_score ": 0.4288978052226948,
      "eval_top_1_macro_f1_score": 0.07176387029488746,
      "eval_top_1_weighted_f1score": 0.20843436611601748,
      "eval_top_3_macro_f1_score": 0.17093716666685138,
      "step": 11072
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0002,
      "loss": 2.8613,
      "step": 11075
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 11080
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0002,
      "loss": 2.867,
      "step": 11085
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0002,
      "loss": 2.8616,
      "step": 11090
    },
    {
      "epoch": 4.76,
      "learning_rate": 0.0002,
      "loss": 2.8561,
      "step": 11095
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 11100
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.0002,
      "loss": 2.8846,
      "step": 11105
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.0002,
      "loss": 2.879,
      "step": 11110
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 11115
    },
    {
      "epoch": 4.77,
      "learning_rate": 0.0002,
      "loss": 2.8774,
      "step": 11120
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 11125
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.0002,
      "loss": 2.8751,
      "step": 11130
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.0002,
      "loss": 2.8807,
      "step": 11135
    },
    {
      "epoch": 4.78,
      "eval_loss": 2.8669276237487793,
      "eval_runtime": 5.4397,
      "eval_samples_per_second": 1815.721,
      "eval_steps_per_second": 14.339,
      "eval_top3_3_weighted_f1_score ": 0.4257387530117306,
      "eval_top_1_macro_f1_score": 0.06590494818333718,
      "eval_top_1_weighted_f1score": 0.20400630100061296,
      "eval_top_3_macro_f1_score": 0.17254735306962476,
      "step": 11136
    },
    {
      "epoch": 4.78,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 11140
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.0002,
      "loss": 2.877,
      "step": 11145
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.0002,
      "loss": 2.8589,
      "step": 11150
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.0002,
      "loss": 2.8678,
      "step": 11155
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 11160
    },
    {
      "epoch": 4.79,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 11165
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.0002,
      "loss": 2.8652,
      "step": 11170
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 11175
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.0002,
      "loss": 2.8527,
      "step": 11180
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 11185
    },
    {
      "epoch": 4.8,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 11190
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.0002,
      "loss": 2.854,
      "step": 11195
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.0002,
      "loss": 2.8817,
      "step": 11200
    },
    {
      "epoch": 4.81,
      "eval_loss": 2.8746719360351562,
      "eval_runtime": 5.3723,
      "eval_samples_per_second": 1838.521,
      "eval_steps_per_second": 14.519,
      "eval_top3_3_weighted_f1_score ": 0.42587045814788116,
      "eval_top_1_macro_f1_score": 0.07313648022622676,
      "eval_top_1_weighted_f1score": 0.2102847351298883,
      "eval_top_3_macro_f1_score": 0.170707313462331,
      "step": 11200
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.0002,
      "loss": 2.8673,
      "step": 11205
    },
    {
      "epoch": 4.81,
      "learning_rate": 0.0002,
      "loss": 2.8628,
      "step": 11210
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 11215
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 11220
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0002,
      "loss": 2.8644,
      "step": 11225
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0002,
      "loss": 2.8755,
      "step": 11230
    },
    {
      "epoch": 4.82,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 11235
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 11240
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.0002,
      "loss": 2.8686,
      "step": 11245
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.0002,
      "loss": 2.8566,
      "step": 11250
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 11255
    },
    {
      "epoch": 4.83,
      "learning_rate": 0.0002,
      "loss": 2.8644,
      "step": 11260
    },
    {
      "epoch": 4.84,
      "eval_loss": 2.8743584156036377,
      "eval_runtime": 5.3747,
      "eval_samples_per_second": 1837.672,
      "eval_steps_per_second": 14.512,
      "eval_top3_3_weighted_f1_score ": 0.42675392687670083,
      "eval_top_1_macro_f1_score": 0.06248300017997619,
      "eval_top_1_weighted_f1score": 0.20552355444489265,
      "eval_top_3_macro_f1_score": 0.15496691610870802,
      "step": 11264
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 11265
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.0002,
      "loss": 2.8624,
      "step": 11270
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.0002,
      "loss": 2.8777,
      "step": 11275
    },
    {
      "epoch": 4.84,
      "learning_rate": 0.0002,
      "loss": 2.8722,
      "step": 11280
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.0002,
      "loss": 2.8666,
      "step": 11285
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.0002,
      "loss": 2.86,
      "step": 11290
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.0002,
      "loss": 2.8707,
      "step": 11295
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.0002,
      "loss": 2.8704,
      "step": 11300
    },
    {
      "epoch": 4.85,
      "learning_rate": 0.0002,
      "loss": 2.8898,
      "step": 11305
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.0002,
      "loss": 2.8673,
      "step": 11310
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.0002,
      "loss": 2.8565,
      "step": 11315
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.0002,
      "loss": 2.8564,
      "step": 11320
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.0002,
      "loss": 2.8595,
      "step": 11325
    },
    {
      "epoch": 4.86,
      "eval_loss": 2.867732524871826,
      "eval_runtime": 5.444,
      "eval_samples_per_second": 1814.291,
      "eval_steps_per_second": 14.328,
      "eval_top3_3_weighted_f1_score ": 0.4309978290330031,
      "eval_top_1_macro_f1_score": 0.061983701728155746,
      "eval_top_1_weighted_f1score": 0.20225068121052153,
      "eval_top_3_macro_f1_score": 0.16265779170233347,
      "step": 11328
    },
    {
      "epoch": 4.86,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 11330
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.0002,
      "loss": 2.8574,
      "step": 11335
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 11340
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.0002,
      "loss": 2.8887,
      "step": 11345
    },
    {
      "epoch": 4.87,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 11350
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.0002,
      "loss": 2.8715,
      "step": 11355
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.0002,
      "loss": 2.8649,
      "step": 11360
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.0002,
      "loss": 2.8839,
      "step": 11365
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 11370
    },
    {
      "epoch": 4.88,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 11375
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.0002,
      "loss": 2.8499,
      "step": 11380
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 11385
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.0002,
      "loss": 2.8564,
      "step": 11390
    },
    {
      "epoch": 4.89,
      "eval_loss": 2.8717246055603027,
      "eval_runtime": 5.4032,
      "eval_samples_per_second": 1828.006,
      "eval_steps_per_second": 14.436,
      "eval_top3_3_weighted_f1_score ": 0.4209417191823971,
      "eval_top_1_macro_f1_score": 0.05910583218241548,
      "eval_top_1_weighted_f1score": 0.2016748822248444,
      "eval_top_3_macro_f1_score": 0.15671989541896147,
      "step": 11392
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.0002,
      "loss": 2.8842,
      "step": 11395
    },
    {
      "epoch": 4.89,
      "learning_rate": 0.0002,
      "loss": 2.8713,
      "step": 11400
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.0002,
      "loss": 2.8689,
      "step": 11405
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.0002,
      "loss": 2.8837,
      "step": 11410
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.0002,
      "loss": 2.8813,
      "step": 11415
    },
    {
      "epoch": 4.9,
      "learning_rate": 0.0002,
      "loss": 2.8727,
      "step": 11420
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 11425
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.0002,
      "loss": 2.8925,
      "step": 11430
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.0002,
      "loss": 2.8666,
      "step": 11435
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.0002,
      "loss": 2.8687,
      "step": 11440
    },
    {
      "epoch": 4.91,
      "learning_rate": 0.0002,
      "loss": 2.8731,
      "step": 11445
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 11450
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.0002,
      "loss": 2.8671,
      "step": 11455
    },
    {
      "epoch": 4.92,
      "eval_loss": 2.870335102081299,
      "eval_runtime": 5.3908,
      "eval_samples_per_second": 1832.184,
      "eval_steps_per_second": 14.469,
      "eval_top3_3_weighted_f1_score ": 0.4259999648606368,
      "eval_top_1_macro_f1_score": 0.06792783069712435,
      "eval_top_1_weighted_f1score": 0.20425705865041188,
      "eval_top_3_macro_f1_score": 0.1648241576211881,
      "step": 11456
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.0002,
      "loss": 2.8905,
      "step": 11460
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.0002,
      "loss": 2.8711,
      "step": 11465
    },
    {
      "epoch": 4.92,
      "learning_rate": 0.0002,
      "loss": 2.8592,
      "step": 11470
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.0002,
      "loss": 2.8706,
      "step": 11475
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.0002,
      "loss": 2.8644,
      "step": 11480
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.0002,
      "loss": 2.8637,
      "step": 11485
    },
    {
      "epoch": 4.93,
      "learning_rate": 0.0002,
      "loss": 2.8692,
      "step": 11490
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.0002,
      "loss": 2.8631,
      "step": 11495
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.0002,
      "loss": 2.8631,
      "step": 11500
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.0002,
      "loss": 2.8639,
      "step": 11505
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.0002,
      "loss": 2.8825,
      "step": 11510
    },
    {
      "epoch": 4.94,
      "learning_rate": 0.0002,
      "loss": 2.8785,
      "step": 11515
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.0002,
      "loss": 2.8823,
      "step": 11520
    },
    {
      "epoch": 4.95,
      "eval_loss": 2.8750851154327393,
      "eval_runtime": 5.3705,
      "eval_samples_per_second": 1839.112,
      "eval_steps_per_second": 14.524,
      "eval_top3_3_weighted_f1_score ": 0.4242109118570131,
      "eval_top_1_macro_f1_score": 0.06375210790313554,
      "eval_top_1_weighted_f1score": 0.20754066622135697,
      "eval_top_3_macro_f1_score": 0.16606030329682445,
      "step": 11520
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.0002,
      "loss": 2.8884,
      "step": 11525
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.0002,
      "loss": 2.8572,
      "step": 11530
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 11535
    },
    {
      "epoch": 4.95,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 11540
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 11545
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.0002,
      "loss": 2.8714,
      "step": 11550
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 11555
    },
    {
      "epoch": 4.96,
      "learning_rate": 0.0002,
      "loss": 2.8759,
      "step": 11560
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 11565
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.0002,
      "loss": 2.8702,
      "step": 11570
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.0002,
      "loss": 2.8682,
      "step": 11575
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.0002,
      "loss": 2.8846,
      "step": 11580
    },
    {
      "epoch": 4.97,
      "eval_loss": 2.867961883544922,
      "eval_runtime": 5.4009,
      "eval_samples_per_second": 1828.759,
      "eval_steps_per_second": 14.442,
      "eval_top3_3_weighted_f1_score ": 0.42672444215232747,
      "eval_top_1_macro_f1_score": 0.06347691449250102,
      "eval_top_1_weighted_f1score": 0.20503581963070142,
      "eval_top_3_macro_f1_score": 0.1667686899827073,
      "step": 11584
    },
    {
      "epoch": 4.97,
      "learning_rate": 0.0002,
      "loss": 2.8562,
      "step": 11585
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.0002,
      "loss": 2.8612,
      "step": 11590
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 11595
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.0002,
      "loss": 2.8825,
      "step": 11600
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.0002,
      "loss": 2.8552,
      "step": 11605
    },
    {
      "epoch": 4.98,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 11610
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.0002,
      "loss": 2.8684,
      "step": 11615
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 11620
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.0002,
      "loss": 2.8777,
      "step": 11625
    },
    {
      "epoch": 4.99,
      "learning_rate": 0.0002,
      "loss": 2.8714,
      "step": 11630
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 11635
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0002,
      "loss": 2.8629,
      "step": 11640
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0002,
      "loss": 2.86,
      "step": 11645
    },
    {
      "epoch": 5.0,
      "eval_loss": 2.8654212951660156,
      "eval_runtime": 5.6035,
      "eval_samples_per_second": 1762.633,
      "eval_steps_per_second": 13.92,
      "eval_top3_3_weighted_f1_score ": 0.4220382255574687,
      "eval_top_1_macro_f1_score": 0.06684360445045137,
      "eval_top_1_weighted_f1score": 0.20898039023299905,
      "eval_top_3_macro_f1_score": 0.17073237497455263,
      "step": 11648
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0002,
      "loss": 3.129,
      "step": 11650
    },
    {
      "epoch": 5.0,
      "learning_rate": 0.0002,
      "loss": 2.8614,
      "step": 11655
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.0002,
      "loss": 2.8601,
      "step": 11660
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.0002,
      "loss": 2.8554,
      "step": 11665
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.0002,
      "loss": 2.8457,
      "step": 11670
    },
    {
      "epoch": 5.01,
      "learning_rate": 0.0002,
      "loss": 2.8544,
      "step": 11675
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.0002,
      "loss": 2.8652,
      "step": 11680
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 11685
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.0002,
      "loss": 2.8725,
      "step": 11690
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.0002,
      "loss": 2.8634,
      "step": 11695
    },
    {
      "epoch": 5.02,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 11700
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.0002,
      "loss": 2.855,
      "step": 11705
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.0002,
      "loss": 2.8393,
      "step": 11710
    },
    {
      "epoch": 5.03,
      "eval_loss": 2.8707430362701416,
      "eval_runtime": 5.3598,
      "eval_samples_per_second": 1842.778,
      "eval_steps_per_second": 14.553,
      "eval_top3_3_weighted_f1_score ": 0.42300889856606255,
      "eval_top_1_macro_f1_score": 0.0596400632884743,
      "eval_top_1_weighted_f1score": 0.20651925870463048,
      "eval_top_3_macro_f1_score": 0.1609654670119019,
      "step": 11712
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.0002,
      "loss": 2.8609,
      "step": 11715
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.0002,
      "loss": 2.8567,
      "step": 11720
    },
    {
      "epoch": 5.03,
      "learning_rate": 0.0002,
      "loss": 2.8314,
      "step": 11725
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.0002,
      "loss": 2.8734,
      "step": 11730
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.0002,
      "loss": 2.8838,
      "step": 11735
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.0002,
      "loss": 2.8456,
      "step": 11740
    },
    {
      "epoch": 5.04,
      "learning_rate": 0.0002,
      "loss": 2.8792,
      "step": 11745
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.0002,
      "loss": 2.8758,
      "step": 11750
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.0002,
      "loss": 2.8634,
      "step": 11755
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.0002,
      "loss": 2.8499,
      "step": 11760
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.0002,
      "loss": 2.8573,
      "step": 11765
    },
    {
      "epoch": 5.05,
      "learning_rate": 0.0002,
      "loss": 2.8563,
      "step": 11770
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0002,
      "loss": 2.8518,
      "step": 11775
    },
    {
      "epoch": 5.06,
      "eval_loss": 2.874115228652954,
      "eval_runtime": 5.4346,
      "eval_samples_per_second": 1817.431,
      "eval_steps_per_second": 14.352,
      "eval_top3_3_weighted_f1_score ": 0.42229709993671566,
      "eval_top_1_macro_f1_score": 0.06295362765695227,
      "eval_top_1_weighted_f1score": 0.20956336439790296,
      "eval_top_3_macro_f1_score": 0.16666229223268295,
      "step": 11776
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0002,
      "loss": 2.8844,
      "step": 11780
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0002,
      "loss": 2.8602,
      "step": 11785
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0002,
      "loss": 2.8671,
      "step": 11790
    },
    {
      "epoch": 5.06,
      "learning_rate": 0.0002,
      "loss": 2.8627,
      "step": 11795
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.0002,
      "loss": 2.8666,
      "step": 11800
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 11805
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.0002,
      "loss": 2.8662,
      "step": 11810
    },
    {
      "epoch": 5.07,
      "learning_rate": 0.0002,
      "loss": 2.8648,
      "step": 11815
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.0002,
      "loss": 2.8638,
      "step": 11820
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 11825
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.0002,
      "loss": 2.8703,
      "step": 11830
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.0002,
      "loss": 2.8691,
      "step": 11835
    },
    {
      "epoch": 5.08,
      "learning_rate": 0.0002,
      "loss": 2.8768,
      "step": 11840
    },
    {
      "epoch": 5.08,
      "eval_loss": 2.8761074542999268,
      "eval_runtime": 5.4132,
      "eval_samples_per_second": 1824.612,
      "eval_steps_per_second": 14.409,
      "eval_top3_3_weighted_f1_score ": 0.4203521589308624,
      "eval_top_1_macro_f1_score": 0.06184167382652768,
      "eval_top_1_weighted_f1score": 0.20401684997897046,
      "eval_top_3_macro_f1_score": 0.15860838499071558,
      "step": 11840
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 11845
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.0002,
      "loss": 2.8608,
      "step": 11850
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.0002,
      "loss": 2.872,
      "step": 11855
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.0002,
      "loss": 2.867,
      "step": 11860
    },
    {
      "epoch": 5.09,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 11865
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 11870
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.0002,
      "loss": 2.8737,
      "step": 11875
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 11880
    },
    {
      "epoch": 5.1,
      "learning_rate": 0.0002,
      "loss": 2.8556,
      "step": 11885
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0002,
      "loss": 2.8807,
      "step": 11890
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0002,
      "loss": 2.8684,
      "step": 11895
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0002,
      "loss": 2.8505,
      "step": 11900
    },
    {
      "epoch": 5.11,
      "eval_loss": 2.8665826320648193,
      "eval_runtime": 5.4113,
      "eval_samples_per_second": 1825.258,
      "eval_steps_per_second": 14.414,
      "eval_top3_3_weighted_f1_score ": 0.4268338631196219,
      "eval_top_1_macro_f1_score": 0.06435100153173876,
      "eval_top_1_weighted_f1score": 0.20520514027260486,
      "eval_top_3_macro_f1_score": 0.1676481366511563,
      "step": 11904
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0002,
      "loss": 2.8601,
      "step": 11905
    },
    {
      "epoch": 5.11,
      "learning_rate": 0.0002,
      "loss": 2.8628,
      "step": 11910
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.0002,
      "loss": 2.8658,
      "step": 11915
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.0002,
      "loss": 2.871,
      "step": 11920
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 11925
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.0002,
      "loss": 2.8764,
      "step": 11930
    },
    {
      "epoch": 5.12,
      "learning_rate": 0.0002,
      "loss": 2.868,
      "step": 11935
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0002,
      "loss": 2.8653,
      "step": 11940
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0002,
      "loss": 2.8553,
      "step": 11945
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0002,
      "loss": 2.8699,
      "step": 11950
    },
    {
      "epoch": 5.13,
      "learning_rate": 0.0002,
      "loss": 2.8497,
      "step": 11955
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0002,
      "loss": 2.8776,
      "step": 11960
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 11965
    },
    {
      "epoch": 5.14,
      "eval_loss": 2.869213104248047,
      "eval_runtime": 5.3698,
      "eval_samples_per_second": 1839.376,
      "eval_steps_per_second": 14.526,
      "eval_top3_3_weighted_f1_score ": 0.4286770986392775,
      "eval_top_1_macro_f1_score": 0.06212153921303334,
      "eval_top_1_weighted_f1score": 0.20747741095249192,
      "eval_top_3_macro_f1_score": 0.16919109524391723,
      "step": 11968
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0002,
      "loss": 2.8778,
      "step": 11970
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 11975
    },
    {
      "epoch": 5.14,
      "learning_rate": 0.0002,
      "loss": 2.8638,
      "step": 11980
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.0002,
      "loss": 2.8789,
      "step": 11985
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.0002,
      "loss": 2.8688,
      "step": 11990
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 11995
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 12000
    },
    {
      "epoch": 5.15,
      "learning_rate": 0.0002,
      "loss": 2.8528,
      "step": 12005
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.0002,
      "loss": 2.8558,
      "step": 12010
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.0002,
      "loss": 2.8442,
      "step": 12015
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 12020
    },
    {
      "epoch": 5.16,
      "learning_rate": 0.0002,
      "loss": 2.8646,
      "step": 12025
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.0002,
      "loss": 2.8699,
      "step": 12030
    },
    {
      "epoch": 5.17,
      "eval_loss": 2.8711791038513184,
      "eval_runtime": 5.4033,
      "eval_samples_per_second": 1827.959,
      "eval_steps_per_second": 14.436,
      "eval_top3_3_weighted_f1_score ": 0.42502434310512177,
      "eval_top_1_macro_f1_score": 0.06390669240183022,
      "eval_top_1_weighted_f1score": 0.20611401320076303,
      "eval_top_3_macro_f1_score": 0.172394356924741,
      "step": 12032
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.0002,
      "loss": 2.8558,
      "step": 12035
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.0002,
      "loss": 2.885,
      "step": 12040
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 12045
    },
    {
      "epoch": 5.17,
      "learning_rate": 0.0002,
      "loss": 2.8417,
      "step": 12050
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0002,
      "loss": 2.8463,
      "step": 12055
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0002,
      "loss": 2.8496,
      "step": 12060
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 12065
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0002,
      "loss": 2.8511,
      "step": 12070
    },
    {
      "epoch": 5.18,
      "learning_rate": 0.0002,
      "loss": 2.8565,
      "step": 12075
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.0002,
      "loss": 2.8628,
      "step": 12080
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 12085
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.0002,
      "loss": 2.8555,
      "step": 12090
    },
    {
      "epoch": 5.19,
      "learning_rate": 0.0002,
      "loss": 2.8641,
      "step": 12095
    },
    {
      "epoch": 5.19,
      "eval_loss": 2.8671929836273193,
      "eval_runtime": 5.3803,
      "eval_samples_per_second": 1835.788,
      "eval_steps_per_second": 14.497,
      "eval_top3_3_weighted_f1_score ": 0.42276656929871903,
      "eval_top_1_macro_f1_score": 0.06505845030432661,
      "eval_top_1_weighted_f1score": 0.20797779672348793,
      "eval_top_3_macro_f1_score": 0.1596413030435833,
      "step": 12096
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.0002,
      "loss": 2.8645,
      "step": 12100
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.0002,
      "loss": 2.8814,
      "step": 12105
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 12110
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.0002,
      "loss": 2.8611,
      "step": 12115
    },
    {
      "epoch": 5.2,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 12120
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0002,
      "loss": 2.8739,
      "step": 12125
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 12130
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0002,
      "loss": 2.8722,
      "step": 12135
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 12140
    },
    {
      "epoch": 5.21,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 12145
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.0002,
      "loss": 2.8778,
      "step": 12150
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.0002,
      "loss": 2.8481,
      "step": 12155
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.0002,
      "loss": 2.8723,
      "step": 12160
    },
    {
      "epoch": 5.22,
      "eval_loss": 2.8729920387268066,
      "eval_runtime": 5.3265,
      "eval_samples_per_second": 1854.322,
      "eval_steps_per_second": 14.644,
      "eval_top3_3_weighted_f1_score ": 0.4286138900543716,
      "eval_top_1_macro_f1_score": 0.06630210480419188,
      "eval_top_1_weighted_f1score": 0.20918349320253848,
      "eval_top_3_macro_f1_score": 0.16540882860527564,
      "step": 12160
    },
    {
      "epoch": 5.22,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 12165
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.0002,
      "loss": 2.8639,
      "step": 12170
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.0002,
      "loss": 2.851,
      "step": 12175
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.0002,
      "loss": 2.871,
      "step": 12180
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 12185
    },
    {
      "epoch": 5.23,
      "learning_rate": 0.0002,
      "loss": 2.8488,
      "step": 12190
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.0002,
      "loss": 2.8658,
      "step": 12195
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.0002,
      "loss": 2.876,
      "step": 12200
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 12205
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.0002,
      "loss": 2.8678,
      "step": 12210
    },
    {
      "epoch": 5.24,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 12215
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.0002,
      "loss": 2.8565,
      "step": 12220
    },
    {
      "epoch": 5.25,
      "eval_loss": 2.8733251094818115,
      "eval_runtime": 5.3808,
      "eval_samples_per_second": 1835.59,
      "eval_steps_per_second": 14.496,
      "eval_top3_3_weighted_f1_score ": 0.4274854960455037,
      "eval_top_1_macro_f1_score": 0.06317168004233445,
      "eval_top_1_weighted_f1score": 0.20456381211628516,
      "eval_top_3_macro_f1_score": 0.16196828106224206,
      "step": 12224
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 12225
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.0002,
      "loss": 2.8648,
      "step": 12230
    },
    {
      "epoch": 5.25,
      "learning_rate": 0.0002,
      "loss": 2.8733,
      "step": 12235
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0002,
      "loss": 2.8735,
      "step": 12240
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 12245
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0002,
      "loss": 2.8569,
      "step": 12250
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0002,
      "loss": 2.8573,
      "step": 12255
    },
    {
      "epoch": 5.26,
      "learning_rate": 0.0002,
      "loss": 2.8682,
      "step": 12260
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 12265
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.0002,
      "loss": 2.8592,
      "step": 12270
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.0002,
      "loss": 2.8631,
      "step": 12275
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.0002,
      "loss": 2.8686,
      "step": 12280
    },
    {
      "epoch": 5.27,
      "learning_rate": 0.0002,
      "loss": 2.8631,
      "step": 12285
    },
    {
      "epoch": 5.28,
      "eval_loss": 2.8703885078430176,
      "eval_runtime": 5.3983,
      "eval_samples_per_second": 1829.664,
      "eval_steps_per_second": 14.449,
      "eval_top3_3_weighted_f1_score ": 0.4263112461317659,
      "eval_top_1_macro_f1_score": 0.06456388348788861,
      "eval_top_1_weighted_f1score": 0.20418942800250817,
      "eval_top_3_macro_f1_score": 0.16130626570251105,
      "step": 12288
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 12290
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.0002,
      "loss": 2.8683,
      "step": 12295
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 12300
    },
    {
      "epoch": 5.28,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 12305
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0002,
      "loss": 2.8566,
      "step": 12310
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0002,
      "loss": 2.8758,
      "step": 12315
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 12320
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0002,
      "loss": 2.8551,
      "step": 12325
    },
    {
      "epoch": 5.29,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 12330
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0002,
      "loss": 2.8599,
      "step": 12335
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0002,
      "loss": 2.8492,
      "step": 12340
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 12345
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0002,
      "loss": 2.8536,
      "step": 12350
    },
    {
      "epoch": 5.3,
      "eval_loss": 2.863574504852295,
      "eval_runtime": 5.3774,
      "eval_samples_per_second": 1836.757,
      "eval_steps_per_second": 14.505,
      "eval_top3_3_weighted_f1_score ": 0.4223760703791298,
      "eval_top_1_macro_f1_score": 0.06263228412679026,
      "eval_top_1_weighted_f1score": 0.20472715153654886,
      "eval_top_3_macro_f1_score": 0.160187916726336,
      "step": 12352
    },
    {
      "epoch": 5.3,
      "learning_rate": 0.0002,
      "loss": 2.8635,
      "step": 12355
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.0002,
      "loss": 2.869,
      "step": 12360
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 12365
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.0002,
      "loss": 2.8566,
      "step": 12370
    },
    {
      "epoch": 5.31,
      "learning_rate": 0.0002,
      "loss": 2.8495,
      "step": 12375
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 12380
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 12385
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 12390
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.0002,
      "loss": 2.8562,
      "step": 12395
    },
    {
      "epoch": 5.32,
      "learning_rate": 0.0002,
      "loss": 2.8595,
      "step": 12400
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0002,
      "loss": 2.8836,
      "step": 12405
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0002,
      "loss": 2.8741,
      "step": 12410
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0002,
      "loss": 2.8704,
      "step": 12415
    },
    {
      "epoch": 5.33,
      "eval_loss": 2.8663103580474854,
      "eval_runtime": 5.4585,
      "eval_samples_per_second": 1809.458,
      "eval_steps_per_second": 14.29,
      "eval_top3_3_weighted_f1_score ": 0.4265369230250399,
      "eval_top_1_macro_f1_score": 0.06631147155356475,
      "eval_top_1_weighted_f1score": 0.20835882539610368,
      "eval_top_3_macro_f1_score": 0.16525838505297283,
      "step": 12416
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0002,
      "loss": 2.8569,
      "step": 12420
    },
    {
      "epoch": 5.33,
      "learning_rate": 0.0002,
      "loss": 2.8438,
      "step": 12425
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 12430
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 12435
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.0002,
      "loss": 2.8474,
      "step": 12440
    },
    {
      "epoch": 5.34,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 12445
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0002,
      "loss": 2.8518,
      "step": 12450
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0002,
      "loss": 2.877,
      "step": 12455
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 12460
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0002,
      "loss": 2.854,
      "step": 12465
    },
    {
      "epoch": 5.35,
      "learning_rate": 0.0002,
      "loss": 2.8581,
      "step": 12470
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0002,
      "loss": 2.8619,
      "step": 12475
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 12480
    },
    {
      "epoch": 5.36,
      "eval_loss": 2.876800775527954,
      "eval_runtime": 5.3362,
      "eval_samples_per_second": 1850.953,
      "eval_steps_per_second": 14.617,
      "eval_top3_3_weighted_f1_score ": 0.42567207437957555,
      "eval_top_1_macro_f1_score": 0.06345758787819845,
      "eval_top_1_weighted_f1score": 0.20840399941031143,
      "eval_top_3_macro_f1_score": 0.16604798262384834,
      "step": 12480
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0002,
      "loss": 2.8579,
      "step": 12485
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0002,
      "loss": 2.8536,
      "step": 12490
    },
    {
      "epoch": 5.36,
      "learning_rate": 0.0002,
      "loss": 2.8633,
      "step": 12495
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 12500
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.0002,
      "loss": 2.8443,
      "step": 12505
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.0002,
      "loss": 2.8394,
      "step": 12510
    },
    {
      "epoch": 5.37,
      "learning_rate": 0.0002,
      "loss": 2.8773,
      "step": 12515
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 12520
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 12525
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 12530
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.0002,
      "loss": 2.8664,
      "step": 12535
    },
    {
      "epoch": 5.38,
      "learning_rate": 0.0002,
      "loss": 2.8633,
      "step": 12540
    },
    {
      "epoch": 5.39,
      "eval_loss": 2.87705397605896,
      "eval_runtime": 5.3971,
      "eval_samples_per_second": 1830.044,
      "eval_steps_per_second": 14.452,
      "eval_top3_3_weighted_f1_score ": 0.42581772075435337,
      "eval_top_1_macro_f1_score": 0.06089987140851035,
      "eval_top_1_weighted_f1score": 0.20790260048900666,
      "eval_top_3_macro_f1_score": 0.1678851592543872,
      "step": 12544
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.0002,
      "loss": 2.8694,
      "step": 12545
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.0002,
      "loss": 2.8583,
      "step": 12550
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.0002,
      "loss": 2.8645,
      "step": 12555
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.0002,
      "loss": 2.8641,
      "step": 12560
    },
    {
      "epoch": 5.39,
      "learning_rate": 0.0002,
      "loss": 2.8456,
      "step": 12565
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 12570
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0002,
      "loss": 2.8678,
      "step": 12575
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 12580
    },
    {
      "epoch": 5.4,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 12585
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.0002,
      "loss": 2.8827,
      "step": 12590
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 12595
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.0002,
      "loss": 2.8692,
      "step": 12600
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.0002,
      "loss": 2.8673,
      "step": 12605
    },
    {
      "epoch": 5.41,
      "eval_loss": 2.8670997619628906,
      "eval_runtime": 5.4492,
      "eval_samples_per_second": 1812.558,
      "eval_steps_per_second": 14.314,
      "eval_top3_3_weighted_f1_score ": 0.4297177427482825,
      "eval_top_1_macro_f1_score": 0.06382875173672384,
      "eval_top_1_weighted_f1score": 0.20357842086656863,
      "eval_top_3_macro_f1_score": 0.17012538987789935,
      "step": 12608
    },
    {
      "epoch": 5.41,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 12610
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0002,
      "loss": 2.8554,
      "step": 12615
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0002,
      "loss": 2.8455,
      "step": 12620
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0002,
      "loss": 2.8468,
      "step": 12625
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0002,
      "loss": 2.8411,
      "step": 12630
    },
    {
      "epoch": 5.42,
      "learning_rate": 0.0002,
      "loss": 2.8594,
      "step": 12635
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 12640
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.0002,
      "loss": 2.8653,
      "step": 12645
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 12650
    },
    {
      "epoch": 5.43,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 12655
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 12660
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.0002,
      "loss": 2.8719,
      "step": 12665
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.0002,
      "loss": 2.8792,
      "step": 12670
    },
    {
      "epoch": 5.44,
      "eval_loss": 2.868391752243042,
      "eval_runtime": 5.3487,
      "eval_samples_per_second": 1846.602,
      "eval_steps_per_second": 14.583,
      "eval_top3_3_weighted_f1_score ": 0.4267632240110778,
      "eval_top_1_macro_f1_score": 0.06546434917081703,
      "eval_top_1_weighted_f1score": 0.20801387882004196,
      "eval_top_3_macro_f1_score": 0.15873036291281165,
      "step": 12672
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 12675
    },
    {
      "epoch": 5.44,
      "learning_rate": 0.0002,
      "loss": 2.8761,
      "step": 12680
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.0002,
      "loss": 2.8667,
      "step": 12685
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.0002,
      "loss": 2.851,
      "step": 12690
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 12695
    },
    {
      "epoch": 5.45,
      "learning_rate": 0.0002,
      "loss": 2.858,
      "step": 12700
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 12705
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.0002,
      "loss": 2.8702,
      "step": 12710
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.0002,
      "loss": 2.8716,
      "step": 12715
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.0002,
      "loss": 2.8562,
      "step": 12720
    },
    {
      "epoch": 5.46,
      "learning_rate": 0.0002,
      "loss": 2.8254,
      "step": 12725
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.0002,
      "loss": 2.869,
      "step": 12730
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.0002,
      "loss": 2.8661,
      "step": 12735
    },
    {
      "epoch": 5.47,
      "eval_loss": 2.8712151050567627,
      "eval_runtime": 5.4522,
      "eval_samples_per_second": 1811.55,
      "eval_steps_per_second": 14.306,
      "eval_top3_3_weighted_f1_score ": 0.42248428596315507,
      "eval_top_1_macro_f1_score": 0.06338137711067554,
      "eval_top_1_weighted_f1score": 0.2052704347027335,
      "eval_top_3_macro_f1_score": 0.15724683781010063,
      "step": 12736
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.0002,
      "loss": 2.8729,
      "step": 12740
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.0002,
      "loss": 2.858,
      "step": 12745
    },
    {
      "epoch": 5.47,
      "learning_rate": 0.0002,
      "loss": 2.8738,
      "step": 12750
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.0002,
      "loss": 2.8715,
      "step": 12755
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.0002,
      "loss": 2.8754,
      "step": 12760
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.0002,
      "loss": 2.8586,
      "step": 12765
    },
    {
      "epoch": 5.48,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 12770
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 12775
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.0002,
      "loss": 2.8458,
      "step": 12780
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.0002,
      "loss": 2.8616,
      "step": 12785
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.0002,
      "loss": 2.884,
      "step": 12790
    },
    {
      "epoch": 5.49,
      "learning_rate": 0.0002,
      "loss": 2.8502,
      "step": 12795
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.0002,
      "loss": 2.8394,
      "step": 12800
    },
    {
      "epoch": 5.5,
      "eval_loss": 2.879948854446411,
      "eval_runtime": 5.6176,
      "eval_samples_per_second": 1758.219,
      "eval_steps_per_second": 13.885,
      "eval_top3_3_weighted_f1_score ": 0.42693128085143395,
      "eval_top_1_macro_f1_score": 0.06255348634396939,
      "eval_top_1_weighted_f1score": 0.20712040547374516,
      "eval_top_3_macro_f1_score": 0.16143149477749164,
      "step": 12800
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.0002,
      "loss": 2.8694,
      "step": 12805
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.0002,
      "loss": 2.8559,
      "step": 12810
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.0002,
      "loss": 2.8812,
      "step": 12815
    },
    {
      "epoch": 5.5,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 12820
    },
    {
      "epoch": 5.51,
      "learning_rate": 0.0002,
      "loss": 2.8573,
      "step": 12825
    },
    {
      "epoch": 5.51,
      "learning_rate": 0.0002,
      "loss": 2.8635,
      "step": 12830
    },
    {
      "epoch": 5.51,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 12835
    },
    {
      "epoch": 5.51,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 12840
    },
    {
      "epoch": 5.52,
      "learning_rate": 0.0002,
      "loss": 2.8421,
      "step": 12845
    },
    {
      "epoch": 5.52,
      "learning_rate": 0.0002,
      "loss": 2.8666,
      "step": 12850
    },
    {
      "epoch": 5.52,
      "learning_rate": 0.0002,
      "loss": 2.8783,
      "step": 12855
    },
    {
      "epoch": 5.52,
      "learning_rate": 0.0002,
      "loss": 2.8746,
      "step": 12860
    },
    {
      "epoch": 5.52,
      "eval_loss": 2.86928129196167,
      "eval_runtime": 5.7863,
      "eval_samples_per_second": 1706.969,
      "eval_steps_per_second": 13.48,
      "eval_top3_3_weighted_f1_score ": 0.4290956311025736,
      "eval_top_1_macro_f1_score": 0.058589711765819036,
      "eval_top_1_weighted_f1score": 0.2056205317016191,
      "eval_top_3_macro_f1_score": 0.15132675835803747,
      "step": 12864
    },
    {
      "epoch": 5.52,
      "learning_rate": 0.0002,
      "loss": 2.8748,
      "step": 12865
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.0002,
      "loss": 2.8679,
      "step": 12870
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.0002,
      "loss": 2.8584,
      "step": 12875
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.0002,
      "loss": 2.8602,
      "step": 12880
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.0002,
      "loss": 2.869,
      "step": 12885
    },
    {
      "epoch": 5.53,
      "learning_rate": 0.0002,
      "loss": 2.8737,
      "step": 12890
    },
    {
      "epoch": 5.54,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 12895
    },
    {
      "epoch": 5.54,
      "learning_rate": 0.0002,
      "loss": 2.8664,
      "step": 12900
    },
    {
      "epoch": 5.54,
      "learning_rate": 0.0002,
      "loss": 2.8638,
      "step": 12905
    },
    {
      "epoch": 5.54,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 12910
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.0002,
      "loss": 2.8533,
      "step": 12915
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.0002,
      "loss": 2.8467,
      "step": 12920
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.0002,
      "loss": 2.867,
      "step": 12925
    },
    {
      "epoch": 5.55,
      "eval_loss": 2.8681602478027344,
      "eval_runtime": 5.4459,
      "eval_samples_per_second": 1813.666,
      "eval_steps_per_second": 14.323,
      "eval_top3_3_weighted_f1_score ": 0.4273838793747354,
      "eval_top_1_macro_f1_score": 0.061575625026085906,
      "eval_top_1_weighted_f1score": 0.21100168044007095,
      "eval_top_3_macro_f1_score": 0.15884318808664932,
      "step": 12928
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.0002,
      "loss": 2.8561,
      "step": 12930
    },
    {
      "epoch": 5.55,
      "learning_rate": 0.0002,
      "loss": 2.8804,
      "step": 12935
    },
    {
      "epoch": 5.56,
      "learning_rate": 0.0002,
      "loss": 2.876,
      "step": 12940
    },
    {
      "epoch": 5.56,
      "learning_rate": 0.0002,
      "loss": 2.8807,
      "step": 12945
    },
    {
      "epoch": 5.56,
      "learning_rate": 0.0002,
      "loss": 2.8474,
      "step": 12950
    },
    {
      "epoch": 5.56,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 12955
    },
    {
      "epoch": 5.56,
      "learning_rate": 0.0002,
      "loss": 2.8551,
      "step": 12960
    },
    {
      "epoch": 5.57,
      "learning_rate": 0.0002,
      "loss": 2.8674,
      "step": 12965
    },
    {
      "epoch": 5.57,
      "learning_rate": 0.0002,
      "loss": 2.8457,
      "step": 12970
    },
    {
      "epoch": 5.57,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 12975
    },
    {
      "epoch": 5.57,
      "learning_rate": 0.0002,
      "loss": 2.868,
      "step": 12980
    },
    {
      "epoch": 5.58,
      "learning_rate": 0.0002,
      "loss": 2.8596,
      "step": 12985
    },
    {
      "epoch": 5.58,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 12990
    },
    {
      "epoch": 5.58,
      "eval_loss": 2.8629438877105713,
      "eval_runtime": 5.3775,
      "eval_samples_per_second": 1836.741,
      "eval_steps_per_second": 14.505,
      "eval_top3_3_weighted_f1_score ": 0.4308889259952943,
      "eval_top_1_macro_f1_score": 0.062038629347178215,
      "eval_top_1_weighted_f1score": 0.20715269808731004,
      "eval_top_3_macro_f1_score": 0.17147488703095864,
      "step": 12992
    },
    {
      "epoch": 5.58,
      "learning_rate": 0.0002,
      "loss": 2.8672,
      "step": 12995
    },
    {
      "epoch": 5.58,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 13000
    },
    {
      "epoch": 5.58,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 13005
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.0002,
      "loss": 2.859,
      "step": 13010
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.0002,
      "loss": 2.8562,
      "step": 13015
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.0002,
      "loss": 2.8596,
      "step": 13020
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.0002,
      "loss": 2.8635,
      "step": 13025
    },
    {
      "epoch": 5.59,
      "learning_rate": 0.0002,
      "loss": 2.8669,
      "step": 13030
    },
    {
      "epoch": 5.6,
      "learning_rate": 0.0002,
      "loss": 2.8497,
      "step": 13035
    },
    {
      "epoch": 5.6,
      "learning_rate": 0.0002,
      "loss": 2.8684,
      "step": 13040
    },
    {
      "epoch": 5.6,
      "learning_rate": 0.0002,
      "loss": 2.8629,
      "step": 13045
    },
    {
      "epoch": 5.6,
      "learning_rate": 0.0002,
      "loss": 2.8443,
      "step": 13050
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 13055
    },
    {
      "epoch": 5.61,
      "eval_loss": 2.8713433742523193,
      "eval_runtime": 5.4068,
      "eval_samples_per_second": 1826.774,
      "eval_steps_per_second": 14.426,
      "eval_top3_3_weighted_f1_score ": 0.425654833318534,
      "eval_top_1_macro_f1_score": 0.06096395287879924,
      "eval_top_1_weighted_f1score": 0.1998952162609984,
      "eval_top_3_macro_f1_score": 0.15487317312853313,
      "step": 13056
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.0002,
      "loss": 2.8616,
      "step": 13060
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 13065
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.0002,
      "loss": 2.8718,
      "step": 13070
    },
    {
      "epoch": 5.61,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 13075
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.0002,
      "loss": 2.8575,
      "step": 13080
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.0002,
      "loss": 2.8803,
      "step": 13085
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.0002,
      "loss": 2.8554,
      "step": 13090
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 13095
    },
    {
      "epoch": 5.62,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 13100
    },
    {
      "epoch": 5.63,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 13105
    },
    {
      "epoch": 5.63,
      "learning_rate": 0.0002,
      "loss": 2.8682,
      "step": 13110
    },
    {
      "epoch": 5.63,
      "learning_rate": 0.0002,
      "loss": 2.8657,
      "step": 13115
    },
    {
      "epoch": 5.63,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 13120
    },
    {
      "epoch": 5.63,
      "eval_loss": 2.858103036880493,
      "eval_runtime": 5.429,
      "eval_samples_per_second": 1819.302,
      "eval_steps_per_second": 14.367,
      "eval_top3_3_weighted_f1_score ": 0.42880534706887957,
      "eval_top_1_macro_f1_score": 0.06353459620981412,
      "eval_top_1_weighted_f1score": 0.20656542151091475,
      "eval_top_3_macro_f1_score": 0.17523647602647105,
      "step": 13120
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.0002,
      "loss": 2.8561,
      "step": 13125
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 13130
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 13135
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.0002,
      "loss": 2.8696,
      "step": 13140
    },
    {
      "epoch": 5.64,
      "learning_rate": 0.0002,
      "loss": 2.8669,
      "step": 13145
    },
    {
      "epoch": 5.65,
      "learning_rate": 0.0002,
      "loss": 2.8558,
      "step": 13150
    },
    {
      "epoch": 5.65,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 13155
    },
    {
      "epoch": 5.65,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 13160
    },
    {
      "epoch": 5.65,
      "learning_rate": 0.0002,
      "loss": 2.8831,
      "step": 13165
    },
    {
      "epoch": 5.65,
      "learning_rate": 0.0002,
      "loss": 2.8523,
      "step": 13170
    },
    {
      "epoch": 5.66,
      "learning_rate": 0.0002,
      "loss": 2.8511,
      "step": 13175
    },
    {
      "epoch": 5.66,
      "learning_rate": 0.0002,
      "loss": 2.875,
      "step": 13180
    },
    {
      "epoch": 5.66,
      "eval_loss": 2.86613392829895,
      "eval_runtime": 5.3406,
      "eval_samples_per_second": 1849.415,
      "eval_steps_per_second": 14.605,
      "eval_top3_3_weighted_f1_score ": 0.4274260278298021,
      "eval_top_1_macro_f1_score": 0.06387742834131617,
      "eval_top_1_weighted_f1score": 0.20790766898744287,
      "eval_top_3_macro_f1_score": 0.17308233668545236,
      "step": 13184
    },
    {
      "epoch": 5.66,
      "learning_rate": 0.0002,
      "loss": 2.8571,
      "step": 13185
    },
    {
      "epoch": 5.66,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 13190
    },
    {
      "epoch": 5.67,
      "learning_rate": 0.0002,
      "loss": 2.8505,
      "step": 13195
    },
    {
      "epoch": 5.67,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 13200
    },
    {
      "epoch": 5.67,
      "learning_rate": 0.0002,
      "loss": 2.8539,
      "step": 13205
    },
    {
      "epoch": 5.67,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 13210
    },
    {
      "epoch": 5.67,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 13215
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.0002,
      "loss": 2.8717,
      "step": 13220
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.0002,
      "loss": 2.8712,
      "step": 13225
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.0002,
      "loss": 2.8648,
      "step": 13230
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 13235
    },
    {
      "epoch": 5.68,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 13240
    },
    {
      "epoch": 5.69,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 13245
    },
    {
      "epoch": 5.69,
      "eval_loss": 2.870232343673706,
      "eval_runtime": 5.3867,
      "eval_samples_per_second": 1833.582,
      "eval_steps_per_second": 14.48,
      "eval_top3_3_weighted_f1_score ": 0.4259613585170209,
      "eval_top_1_macro_f1_score": 0.06270662591985765,
      "eval_top_1_weighted_f1score": 0.20832308964286858,
      "eval_top_3_macro_f1_score": 0.16006006125049677,
      "step": 13248
    },
    {
      "epoch": 5.69,
      "learning_rate": 0.0002,
      "loss": 2.8652,
      "step": 13250
    },
    {
      "epoch": 5.69,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 13255
    },
    {
      "epoch": 5.69,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 13260
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 13265
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.0002,
      "loss": 2.8656,
      "step": 13270
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.0002,
      "loss": 2.8435,
      "step": 13275
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.0002,
      "loss": 2.8586,
      "step": 13280
    },
    {
      "epoch": 5.7,
      "learning_rate": 0.0002,
      "loss": 2.8718,
      "step": 13285
    },
    {
      "epoch": 5.71,
      "learning_rate": 0.0002,
      "loss": 2.8669,
      "step": 13290
    },
    {
      "epoch": 5.71,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 13295
    },
    {
      "epoch": 5.71,
      "learning_rate": 0.0002,
      "loss": 2.8685,
      "step": 13300
    },
    {
      "epoch": 5.71,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 13305
    },
    {
      "epoch": 5.71,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 13310
    },
    {
      "epoch": 5.72,
      "eval_loss": 2.863187551498413,
      "eval_runtime": 5.7315,
      "eval_samples_per_second": 1723.298,
      "eval_steps_per_second": 13.609,
      "eval_top3_3_weighted_f1_score ": 0.42690298388808606,
      "eval_top_1_macro_f1_score": 0.06395728694989959,
      "eval_top_1_weighted_f1score": 0.20885784572072683,
      "eval_top_3_macro_f1_score": 0.16412125306640024,
      "step": 13312
    },
    {
      "epoch": 5.72,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 13315
    },
    {
      "epoch": 5.72,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 13320
    },
    {
      "epoch": 5.72,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 13325
    },
    {
      "epoch": 5.72,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 13330
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.0002,
      "loss": 2.8534,
      "step": 13335
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 13340
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.0002,
      "loss": 2.8672,
      "step": 13345
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 13350
    },
    {
      "epoch": 5.73,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 13355
    },
    {
      "epoch": 5.74,
      "learning_rate": 0.0002,
      "loss": 2.8391,
      "step": 13360
    },
    {
      "epoch": 5.74,
      "learning_rate": 0.0002,
      "loss": 2.8524,
      "step": 13365
    },
    {
      "epoch": 5.74,
      "learning_rate": 0.0002,
      "loss": 2.8589,
      "step": 13370
    },
    {
      "epoch": 5.74,
      "learning_rate": 0.0002,
      "loss": 2.867,
      "step": 13375
    },
    {
      "epoch": 5.74,
      "eval_loss": 2.8663976192474365,
      "eval_runtime": 5.3525,
      "eval_samples_per_second": 1845.314,
      "eval_steps_per_second": 14.573,
      "eval_top3_3_weighted_f1_score ": 0.4253532176538349,
      "eval_top_1_macro_f1_score": 0.06061484272739991,
      "eval_top_1_weighted_f1score": 0.2047296327024241,
      "eval_top_3_macro_f1_score": 0.1562257377230752,
      "step": 13376
    },
    {
      "epoch": 5.74,
      "learning_rate": 0.0002,
      "loss": 2.8514,
      "step": 13380
    },
    {
      "epoch": 5.75,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 13385
    },
    {
      "epoch": 5.75,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 13390
    },
    {
      "epoch": 5.75,
      "learning_rate": 0.0002,
      "loss": 2.8601,
      "step": 13395
    },
    {
      "epoch": 5.75,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 13400
    },
    {
      "epoch": 5.76,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 13405
    },
    {
      "epoch": 5.76,
      "learning_rate": 0.0002,
      "loss": 2.8144,
      "step": 13410
    },
    {
      "epoch": 5.76,
      "learning_rate": 0.0002,
      "loss": 2.8521,
      "step": 13415
    },
    {
      "epoch": 5.76,
      "learning_rate": 0.0002,
      "loss": 2.8664,
      "step": 13420
    },
    {
      "epoch": 5.76,
      "learning_rate": 0.0002,
      "loss": 2.8566,
      "step": 13425
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.0002,
      "loss": 2.8653,
      "step": 13430
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.0002,
      "loss": 2.867,
      "step": 13435
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.0002,
      "loss": 2.874,
      "step": 13440
    },
    {
      "epoch": 5.77,
      "eval_loss": 2.8693718910217285,
      "eval_runtime": 5.4136,
      "eval_samples_per_second": 1824.483,
      "eval_steps_per_second": 14.408,
      "eval_top3_3_weighted_f1_score ": 0.4248694503646851,
      "eval_top_1_macro_f1_score": 0.06314351867333298,
      "eval_top_1_weighted_f1score": 0.20468529657027187,
      "eval_top_3_macro_f1_score": 0.16571741276693988,
      "step": 13440
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.0002,
      "loss": 2.8584,
      "step": 13445
    },
    {
      "epoch": 5.77,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 13450
    },
    {
      "epoch": 5.78,
      "learning_rate": 0.0002,
      "loss": 2.8578,
      "step": 13455
    },
    {
      "epoch": 5.78,
      "learning_rate": 0.0002,
      "loss": 2.8515,
      "step": 13460
    },
    {
      "epoch": 5.78,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 13465
    },
    {
      "epoch": 5.78,
      "learning_rate": 0.0002,
      "loss": 2.8775,
      "step": 13470
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 13475
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.0002,
      "loss": 2.8599,
      "step": 13480
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.0002,
      "loss": 2.8759,
      "step": 13485
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.0002,
      "loss": 2.851,
      "step": 13490
    },
    {
      "epoch": 5.79,
      "learning_rate": 0.0002,
      "loss": 2.8371,
      "step": 13495
    },
    {
      "epoch": 5.8,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 13500
    },
    {
      "epoch": 5.8,
      "eval_loss": 2.8693103790283203,
      "eval_runtime": 5.3423,
      "eval_samples_per_second": 1848.844,
      "eval_steps_per_second": 14.601,
      "eval_top3_3_weighted_f1_score ": 0.4260684239751329,
      "eval_top_1_macro_f1_score": 0.06399201002672834,
      "eval_top_1_weighted_f1score": 0.20817983183427266,
      "eval_top_3_macro_f1_score": 0.17214746946434742,
      "step": 13504
    },
    {
      "epoch": 5.8,
      "learning_rate": 0.0002,
      "loss": 2.8702,
      "step": 13505
    },
    {
      "epoch": 5.8,
      "learning_rate": 0.0002,
      "loss": 2.8573,
      "step": 13510
    },
    {
      "epoch": 5.8,
      "learning_rate": 0.0002,
      "loss": 2.8647,
      "step": 13515
    },
    {
      "epoch": 5.8,
      "learning_rate": 0.0002,
      "loss": 2.8717,
      "step": 13520
    },
    {
      "epoch": 5.81,
      "learning_rate": 0.0002,
      "loss": 2.8741,
      "step": 13525
    },
    {
      "epoch": 5.81,
      "learning_rate": 0.0002,
      "loss": 2.8809,
      "step": 13530
    },
    {
      "epoch": 5.81,
      "learning_rate": 0.0002,
      "loss": 2.8765,
      "step": 13535
    },
    {
      "epoch": 5.81,
      "learning_rate": 0.0002,
      "loss": 2.8739,
      "step": 13540
    },
    {
      "epoch": 5.82,
      "learning_rate": 0.0002,
      "loss": 2.8675,
      "step": 13545
    },
    {
      "epoch": 5.82,
      "learning_rate": 0.0002,
      "loss": 2.8628,
      "step": 13550
    },
    {
      "epoch": 5.82,
      "learning_rate": 0.0002,
      "loss": 2.8534,
      "step": 13555
    },
    {
      "epoch": 5.82,
      "learning_rate": 0.0002,
      "loss": 2.8823,
      "step": 13560
    },
    {
      "epoch": 5.82,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 13565
    },
    {
      "epoch": 5.83,
      "eval_loss": 2.8661611080169678,
      "eval_runtime": 5.3918,
      "eval_samples_per_second": 1831.85,
      "eval_steps_per_second": 14.466,
      "eval_top3_3_weighted_f1_score ": 0.42730270541070525,
      "eval_top_1_macro_f1_score": 0.06618373886191772,
      "eval_top_1_weighted_f1score": 0.20648184144746817,
      "eval_top_3_macro_f1_score": 0.16505583558797157,
      "step": 13568
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 13570
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 13575
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.0002,
      "loss": 2.8567,
      "step": 13580
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 13585
    },
    {
      "epoch": 5.83,
      "learning_rate": 0.0002,
      "loss": 2.8641,
      "step": 13590
    },
    {
      "epoch": 5.84,
      "learning_rate": 0.0002,
      "loss": 2.846,
      "step": 13595
    },
    {
      "epoch": 5.84,
      "learning_rate": 0.0002,
      "loss": 2.8376,
      "step": 13600
    },
    {
      "epoch": 5.84,
      "learning_rate": 0.0002,
      "loss": 2.8823,
      "step": 13605
    },
    {
      "epoch": 5.84,
      "learning_rate": 0.0002,
      "loss": 2.842,
      "step": 13610
    },
    {
      "epoch": 5.85,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 13615
    },
    {
      "epoch": 5.85,
      "learning_rate": 0.0002,
      "loss": 2.8554,
      "step": 13620
    },
    {
      "epoch": 5.85,
      "learning_rate": 0.0002,
      "loss": 2.8706,
      "step": 13625
    },
    {
      "epoch": 5.85,
      "learning_rate": 0.0002,
      "loss": 2.8745,
      "step": 13630
    },
    {
      "epoch": 5.85,
      "eval_loss": 2.864320993423462,
      "eval_runtime": 5.3584,
      "eval_samples_per_second": 1843.28,
      "eval_steps_per_second": 14.557,
      "eval_top3_3_weighted_f1_score ": 0.42378643624582846,
      "eval_top_1_macro_f1_score": 0.06515805059497368,
      "eval_top_1_weighted_f1score": 0.20768668382202293,
      "eval_top_3_macro_f1_score": 0.17012690134722805,
      "step": 13632
    },
    {
      "epoch": 5.85,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 13635
    },
    {
      "epoch": 5.86,
      "learning_rate": 0.0002,
      "loss": 2.8744,
      "step": 13640
    },
    {
      "epoch": 5.86,
      "learning_rate": 0.0002,
      "loss": 2.864,
      "step": 13645
    },
    {
      "epoch": 5.86,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 13650
    },
    {
      "epoch": 5.86,
      "learning_rate": 0.0002,
      "loss": 2.8665,
      "step": 13655
    },
    {
      "epoch": 5.87,
      "learning_rate": 0.0002,
      "loss": 2.8614,
      "step": 13660
    },
    {
      "epoch": 5.87,
      "learning_rate": 0.0002,
      "loss": 2.8614,
      "step": 13665
    },
    {
      "epoch": 5.87,
      "learning_rate": 0.0002,
      "loss": 2.8686,
      "step": 13670
    },
    {
      "epoch": 5.87,
      "learning_rate": 0.0002,
      "loss": 2.8848,
      "step": 13675
    },
    {
      "epoch": 5.87,
      "learning_rate": 0.0002,
      "loss": 2.8652,
      "step": 13680
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 13685
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.0002,
      "loss": 2.8481,
      "step": 13690
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.0002,
      "loss": 2.8708,
      "step": 13695
    },
    {
      "epoch": 5.88,
      "eval_loss": 2.8583221435546875,
      "eval_runtime": 5.3333,
      "eval_samples_per_second": 1851.944,
      "eval_steps_per_second": 14.625,
      "eval_top3_3_weighted_f1_score ": 0.42629623086863955,
      "eval_top_1_macro_f1_score": 0.0642516151445838,
      "eval_top_1_weighted_f1score": 0.20652017273950096,
      "eval_top_3_macro_f1_score": 0.1631062990693639,
      "step": 13696
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.0002,
      "loss": 2.8788,
      "step": 13700
    },
    {
      "epoch": 5.88,
      "learning_rate": 0.0002,
      "loss": 2.8682,
      "step": 13705
    },
    {
      "epoch": 5.89,
      "learning_rate": 0.0002,
      "loss": 2.8777,
      "step": 13710
    },
    {
      "epoch": 5.89,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 13715
    },
    {
      "epoch": 5.89,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 13720
    },
    {
      "epoch": 5.89,
      "learning_rate": 0.0002,
      "loss": 2.8731,
      "step": 13725
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.0002,
      "loss": 2.8629,
      "step": 13730
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.0002,
      "loss": 2.8553,
      "step": 13735
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.0002,
      "loss": 2.875,
      "step": 13740
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.0002,
      "loss": 2.8715,
      "step": 13745
    },
    {
      "epoch": 5.9,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 13750
    },
    {
      "epoch": 5.91,
      "learning_rate": 0.0002,
      "loss": 2.8645,
      "step": 13755
    },
    {
      "epoch": 5.91,
      "learning_rate": 0.0002,
      "loss": 2.8673,
      "step": 13760
    },
    {
      "epoch": 5.91,
      "eval_loss": 2.8654608726501465,
      "eval_runtime": 5.4259,
      "eval_samples_per_second": 1820.348,
      "eval_steps_per_second": 14.376,
      "eval_top3_3_weighted_f1_score ": 0.42425758776269407,
      "eval_top_1_macro_f1_score": 0.06367910049044434,
      "eval_top_1_weighted_f1score": 0.2089364923158843,
      "eval_top_3_macro_f1_score": 0.1582568048780128,
      "step": 13760
    },
    {
      "epoch": 5.91,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 13765
    },
    {
      "epoch": 5.91,
      "learning_rate": 0.0002,
      "loss": 2.8458,
      "step": 13770
    },
    {
      "epoch": 5.91,
      "learning_rate": 0.0002,
      "loss": 2.8654,
      "step": 13775
    },
    {
      "epoch": 5.92,
      "learning_rate": 0.0002,
      "loss": 2.8592,
      "step": 13780
    },
    {
      "epoch": 5.92,
      "learning_rate": 0.0002,
      "loss": 2.8478,
      "step": 13785
    },
    {
      "epoch": 5.92,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 13790
    },
    {
      "epoch": 5.92,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 13795
    },
    {
      "epoch": 5.93,
      "learning_rate": 0.0002,
      "loss": 2.8509,
      "step": 13800
    },
    {
      "epoch": 5.93,
      "learning_rate": 0.0002,
      "loss": 2.8399,
      "step": 13805
    },
    {
      "epoch": 5.93,
      "learning_rate": 0.0002,
      "loss": 2.8533,
      "step": 13810
    },
    {
      "epoch": 5.93,
      "learning_rate": 0.0002,
      "loss": 2.8816,
      "step": 13815
    },
    {
      "epoch": 5.93,
      "learning_rate": 0.0002,
      "loss": 2.8541,
      "step": 13820
    },
    {
      "epoch": 5.94,
      "eval_loss": 2.8690874576568604,
      "eval_runtime": 5.3822,
      "eval_samples_per_second": 1835.138,
      "eval_steps_per_second": 14.492,
      "eval_top3_3_weighted_f1_score ": 0.42912869704184087,
      "eval_top_1_macro_f1_score": 0.06746140343210294,
      "eval_top_1_weighted_f1score": 0.20455814407819742,
      "eval_top_3_macro_f1_score": 0.17004323228141258,
      "step": 13824
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.0002,
      "loss": 2.8377,
      "step": 13825
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 13830
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 13835
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 13840
    },
    {
      "epoch": 5.94,
      "learning_rate": 0.0002,
      "loss": 2.8533,
      "step": 13845
    },
    {
      "epoch": 5.95,
      "learning_rate": 0.0002,
      "loss": 2.843,
      "step": 13850
    },
    {
      "epoch": 5.95,
      "learning_rate": 0.0002,
      "loss": 2.8539,
      "step": 13855
    },
    {
      "epoch": 5.95,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 13860
    },
    {
      "epoch": 5.95,
      "learning_rate": 0.0002,
      "loss": 2.8571,
      "step": 13865
    },
    {
      "epoch": 5.96,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 13870
    },
    {
      "epoch": 5.96,
      "learning_rate": 0.0002,
      "loss": 2.8529,
      "step": 13875
    },
    {
      "epoch": 5.96,
      "learning_rate": 0.0002,
      "loss": 2.8557,
      "step": 13880
    },
    {
      "epoch": 5.96,
      "learning_rate": 0.0002,
      "loss": 2.8675,
      "step": 13885
    },
    {
      "epoch": 5.96,
      "eval_loss": 2.8727800846099854,
      "eval_runtime": 5.3875,
      "eval_samples_per_second": 1833.315,
      "eval_steps_per_second": 14.478,
      "eval_top3_3_weighted_f1_score ": 0.4274189365310889,
      "eval_top_1_macro_f1_score": 0.061235624823663806,
      "eval_top_1_weighted_f1score": 0.2056026869490293,
      "eval_top_3_macro_f1_score": 0.1648534533740258,
      "step": 13888
    },
    {
      "epoch": 5.96,
      "learning_rate": 0.0002,
      "loss": 2.8544,
      "step": 13890
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 13895
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.0002,
      "loss": 2.855,
      "step": 13900
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.0002,
      "loss": 2.8777,
      "step": 13905
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.0002,
      "loss": 2.8709,
      "step": 13910
    },
    {
      "epoch": 5.97,
      "learning_rate": 0.0002,
      "loss": 2.8398,
      "step": 13915
    },
    {
      "epoch": 5.98,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 13920
    },
    {
      "epoch": 5.98,
      "learning_rate": 0.0002,
      "loss": 2.8581,
      "step": 13925
    },
    {
      "epoch": 5.98,
      "learning_rate": 0.0002,
      "loss": 2.8505,
      "step": 13930
    },
    {
      "epoch": 5.98,
      "learning_rate": 0.0002,
      "loss": 2.8522,
      "step": 13935
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.0002,
      "loss": 2.8594,
      "step": 13940
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 13945
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.0002,
      "loss": 2.8697,
      "step": 13950
    },
    {
      "epoch": 5.99,
      "eval_loss": 2.8647892475128174,
      "eval_runtime": 5.46,
      "eval_samples_per_second": 1808.988,
      "eval_steps_per_second": 14.286,
      "eval_top3_3_weighted_f1_score ": 0.42687586035584,
      "eval_top_1_macro_f1_score": 0.060179487166277516,
      "eval_top_1_weighted_f1score": 0.20831999698694495,
      "eval_top_3_macro_f1_score": 0.17113203575304348,
      "step": 13952
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.0002,
      "loss": 2.872,
      "step": 13955
    },
    {
      "epoch": 5.99,
      "learning_rate": 0.0002,
      "loss": 2.8515,
      "step": 13960
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0002,
      "loss": 2.8577,
      "step": 13965
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0002,
      "loss": 2.8516,
      "step": 13970
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0002,
      "loss": 3.1273,
      "step": 13975
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 13980
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0002,
      "loss": 2.8612,
      "step": 13985
    },
    {
      "epoch": 6.01,
      "learning_rate": 0.0002,
      "loss": 2.8514,
      "step": 13990
    },
    {
      "epoch": 6.01,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 13995
    },
    {
      "epoch": 6.01,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 14000
    },
    {
      "epoch": 6.01,
      "learning_rate": 0.0002,
      "loss": 2.8412,
      "step": 14005
    },
    {
      "epoch": 6.02,
      "learning_rate": 0.0002,
      "loss": 2.851,
      "step": 14010
    },
    {
      "epoch": 6.02,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 14015
    },
    {
      "epoch": 6.02,
      "eval_loss": 2.860562562942505,
      "eval_runtime": 5.383,
      "eval_samples_per_second": 1834.841,
      "eval_steps_per_second": 14.49,
      "eval_top3_3_weighted_f1_score ": 0.4255769719365458,
      "eval_top_1_macro_f1_score": 0.059525963490135815,
      "eval_top_1_weighted_f1score": 0.20327920220186296,
      "eval_top_3_macro_f1_score": 0.15675116269801012,
      "step": 14016
    },
    {
      "epoch": 6.02,
      "learning_rate": 0.0002,
      "loss": 2.8514,
      "step": 14020
    },
    {
      "epoch": 6.02,
      "learning_rate": 0.0002,
      "loss": 2.839,
      "step": 14025
    },
    {
      "epoch": 6.02,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 14030
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.0002,
      "loss": 2.8526,
      "step": 14035
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 14040
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.0002,
      "loss": 2.8564,
      "step": 14045
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 14050
    },
    {
      "epoch": 6.03,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 14055
    },
    {
      "epoch": 6.04,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 14060
    },
    {
      "epoch": 6.04,
      "learning_rate": 0.0002,
      "loss": 2.8601,
      "step": 14065
    },
    {
      "epoch": 6.04,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 14070
    },
    {
      "epoch": 6.04,
      "learning_rate": 0.0002,
      "loss": 2.8478,
      "step": 14075
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.0002,
      "loss": 2.8451,
      "step": 14080
    },
    {
      "epoch": 6.05,
      "eval_loss": 2.8654625415802,
      "eval_runtime": 5.384,
      "eval_samples_per_second": 1834.514,
      "eval_steps_per_second": 14.487,
      "eval_top3_3_weighted_f1_score ": 0.4253389481140167,
      "eval_top_1_macro_f1_score": 0.062891722866952,
      "eval_top_1_weighted_f1score": 0.21179422766158557,
      "eval_top_3_macro_f1_score": 0.17613893427212401,
      "step": 14080
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 14085
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.0002,
      "loss": 2.8633,
      "step": 14090
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.0002,
      "loss": 2.8657,
      "step": 14095
    },
    {
      "epoch": 6.05,
      "learning_rate": 0.0002,
      "loss": 2.8451,
      "step": 14100
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.0002,
      "loss": 2.863,
      "step": 14105
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 14110
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 14115
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 14120
    },
    {
      "epoch": 6.06,
      "learning_rate": 0.0002,
      "loss": 2.8641,
      "step": 14125
    },
    {
      "epoch": 6.07,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 14130
    },
    {
      "epoch": 6.07,
      "learning_rate": 0.0002,
      "loss": 2.845,
      "step": 14135
    },
    {
      "epoch": 6.07,
      "learning_rate": 0.0002,
      "loss": 2.8598,
      "step": 14140
    },
    {
      "epoch": 6.07,
      "eval_loss": 2.8648641109466553,
      "eval_runtime": 5.374,
      "eval_samples_per_second": 1837.934,
      "eval_steps_per_second": 14.514,
      "eval_top3_3_weighted_f1_score ": 0.42864839485569883,
      "eval_top_1_macro_f1_score": 0.06324209484409811,
      "eval_top_1_weighted_f1score": 0.20698305107306783,
      "eval_top_3_macro_f1_score": 0.16745562944778383,
      "step": 14144
    },
    {
      "epoch": 6.07,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 14145
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.0002,
      "loss": 2.8544,
      "step": 14150
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.0002,
      "loss": 2.8478,
      "step": 14155
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.0002,
      "loss": 2.8495,
      "step": 14160
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.0002,
      "loss": 2.8603,
      "step": 14165
    },
    {
      "epoch": 6.08,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 14170
    },
    {
      "epoch": 6.09,
      "learning_rate": 0.0002,
      "loss": 2.8509,
      "step": 14175
    },
    {
      "epoch": 6.09,
      "learning_rate": 0.0002,
      "loss": 2.8535,
      "step": 14180
    },
    {
      "epoch": 6.09,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 14185
    },
    {
      "epoch": 6.09,
      "learning_rate": 0.0002,
      "loss": 2.8528,
      "step": 14190
    },
    {
      "epoch": 6.09,
      "learning_rate": 0.0002,
      "loss": 2.8439,
      "step": 14195
    },
    {
      "epoch": 6.1,
      "learning_rate": 0.0002,
      "loss": 2.844,
      "step": 14200
    },
    {
      "epoch": 6.1,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 14205
    },
    {
      "epoch": 6.1,
      "eval_loss": 2.861860752105713,
      "eval_runtime": 5.4817,
      "eval_samples_per_second": 1801.805,
      "eval_steps_per_second": 14.229,
      "eval_top3_3_weighted_f1_score ": 0.43031760463309665,
      "eval_top_1_macro_f1_score": 0.06874036906878012,
      "eval_top_1_weighted_f1score": 0.20905465756024647,
      "eval_top_3_macro_f1_score": 0.18252730462500377,
      "step": 14208
    },
    {
      "epoch": 6.1,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 14210
    },
    {
      "epoch": 6.1,
      "learning_rate": 0.0002,
      "loss": 2.8515,
      "step": 14215
    },
    {
      "epoch": 6.11,
      "learning_rate": 0.0002,
      "loss": 2.8413,
      "step": 14220
    },
    {
      "epoch": 6.11,
      "learning_rate": 0.0002,
      "loss": 2.8309,
      "step": 14225
    },
    {
      "epoch": 6.11,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 14230
    },
    {
      "epoch": 6.11,
      "learning_rate": 0.0002,
      "loss": 2.8585,
      "step": 14235
    },
    {
      "epoch": 6.11,
      "learning_rate": 0.0002,
      "loss": 2.8681,
      "step": 14240
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.0002,
      "loss": 2.8532,
      "step": 14245
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.0002,
      "loss": 2.8486,
      "step": 14250
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 14255
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.0002,
      "loss": 2.8656,
      "step": 14260
    },
    {
      "epoch": 6.12,
      "learning_rate": 0.0002,
      "loss": 2.8558,
      "step": 14265
    },
    {
      "epoch": 6.13,
      "learning_rate": 0.0002,
      "loss": 2.8538,
      "step": 14270
    },
    {
      "epoch": 6.13,
      "eval_loss": 2.866154193878174,
      "eval_runtime": 5.5182,
      "eval_samples_per_second": 1789.9,
      "eval_steps_per_second": 14.135,
      "eval_top3_3_weighted_f1_score ": 0.42693272860639986,
      "eval_top_1_macro_f1_score": 0.0656622919165084,
      "eval_top_1_weighted_f1score": 0.20984535680148877,
      "eval_top_3_macro_f1_score": 0.16282466912359944,
      "step": 14272
    },
    {
      "epoch": 6.13,
      "learning_rate": 0.0002,
      "loss": 2.854,
      "step": 14275
    },
    {
      "epoch": 6.13,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 14280
    },
    {
      "epoch": 6.13,
      "learning_rate": 0.0002,
      "loss": 2.8363,
      "step": 14285
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.0002,
      "loss": 2.8364,
      "step": 14290
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 14295
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.0002,
      "loss": 2.8332,
      "step": 14300
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 14305
    },
    {
      "epoch": 6.14,
      "learning_rate": 0.0002,
      "loss": 2.8299,
      "step": 14310
    },
    {
      "epoch": 6.15,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 14315
    },
    {
      "epoch": 6.15,
      "learning_rate": 0.0002,
      "loss": 2.8554,
      "step": 14320
    },
    {
      "epoch": 6.15,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 14325
    },
    {
      "epoch": 6.15,
      "learning_rate": 0.0002,
      "loss": 2.8541,
      "step": 14330
    },
    {
      "epoch": 6.15,
      "learning_rate": 0.0002,
      "loss": 2.8821,
      "step": 14335
    },
    {
      "epoch": 6.16,
      "eval_loss": 2.8681774139404297,
      "eval_runtime": 5.4556,
      "eval_samples_per_second": 1810.427,
      "eval_steps_per_second": 14.297,
      "eval_top3_3_weighted_f1_score ": 0.427189588569116,
      "eval_top_1_macro_f1_score": 0.0599012626709618,
      "eval_top_1_weighted_f1score": 0.20990918717178195,
      "eval_top_3_macro_f1_score": 0.16753857269223285,
      "step": 14336
    },
    {
      "epoch": 6.16,
      "learning_rate": 0.0002,
      "loss": 2.8474,
      "step": 14340
    },
    {
      "epoch": 6.16,
      "learning_rate": 0.0002,
      "loss": 2.8398,
      "step": 14345
    },
    {
      "epoch": 6.16,
      "learning_rate": 0.0002,
      "loss": 2.8699,
      "step": 14350
    },
    {
      "epoch": 6.16,
      "learning_rate": 0.0002,
      "loss": 2.8539,
      "step": 14355
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 14360
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.0002,
      "loss": 2.8796,
      "step": 14365
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.0002,
      "loss": 2.8634,
      "step": 14370
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.0002,
      "loss": 2.8476,
      "step": 14375
    },
    {
      "epoch": 6.17,
      "learning_rate": 0.0002,
      "loss": 2.86,
      "step": 14380
    },
    {
      "epoch": 6.18,
      "learning_rate": 0.0002,
      "loss": 2.8563,
      "step": 14385
    },
    {
      "epoch": 6.18,
      "learning_rate": 0.0002,
      "loss": 2.8348,
      "step": 14390
    },
    {
      "epoch": 6.18,
      "learning_rate": 0.0002,
      "loss": 2.8344,
      "step": 14395
    },
    {
      "epoch": 6.18,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 14400
    },
    {
      "epoch": 6.18,
      "eval_loss": 2.8655192852020264,
      "eval_runtime": 5.6848,
      "eval_samples_per_second": 1737.435,
      "eval_steps_per_second": 13.721,
      "eval_top3_3_weighted_f1_score ": 0.4245790732996531,
      "eval_top_1_macro_f1_score": 0.06231000189655607,
      "eval_top_1_weighted_f1score": 0.20954993066502497,
      "eval_top_3_macro_f1_score": 0.1644070791871058,
      "step": 14400
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.0002,
      "loss": 2.8468,
      "step": 14405
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.0002,
      "loss": 2.8375,
      "step": 14410
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.0002,
      "loss": 2.8371,
      "step": 14415
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 14420
    },
    {
      "epoch": 6.19,
      "learning_rate": 0.0002,
      "loss": 2.8461,
      "step": 14425
    },
    {
      "epoch": 6.2,
      "learning_rate": 0.0002,
      "loss": 2.8437,
      "step": 14430
    },
    {
      "epoch": 6.2,
      "learning_rate": 0.0002,
      "loss": 2.8686,
      "step": 14435
    },
    {
      "epoch": 6.2,
      "learning_rate": 0.0002,
      "loss": 2.8461,
      "step": 14440
    },
    {
      "epoch": 6.2,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 14445
    },
    {
      "epoch": 6.2,
      "learning_rate": 0.0002,
      "loss": 2.8638,
      "step": 14450
    },
    {
      "epoch": 6.21,
      "learning_rate": 0.0002,
      "loss": 2.8623,
      "step": 14455
    },
    {
      "epoch": 6.21,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 14460
    },
    {
      "epoch": 6.21,
      "eval_loss": 2.8639345169067383,
      "eval_runtime": 5.4197,
      "eval_samples_per_second": 1822.416,
      "eval_steps_per_second": 14.392,
      "eval_top3_3_weighted_f1_score ": 0.42709454392903934,
      "eval_top_1_macro_f1_score": 0.06288798968966343,
      "eval_top_1_weighted_f1score": 0.2075210786928968,
      "eval_top_3_macro_f1_score": 0.16917322997926354,
      "step": 14464
    },
    {
      "epoch": 6.21,
      "learning_rate": 0.0002,
      "loss": 2.8408,
      "step": 14465
    },
    {
      "epoch": 6.21,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 14470
    },
    {
      "epoch": 6.22,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 14475
    },
    {
      "epoch": 6.22,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 14480
    },
    {
      "epoch": 6.22,
      "learning_rate": 0.0002,
      "loss": 2.8465,
      "step": 14485
    },
    {
      "epoch": 6.22,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 14490
    },
    {
      "epoch": 6.22,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 14495
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.0002,
      "loss": 2.8479,
      "step": 14500
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.0002,
      "loss": 2.8418,
      "step": 14505
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.0002,
      "loss": 2.8389,
      "step": 14510
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.0002,
      "loss": 2.851,
      "step": 14515
    },
    {
      "epoch": 6.23,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 14520
    },
    {
      "epoch": 6.24,
      "learning_rate": 0.0002,
      "loss": 2.86,
      "step": 14525
    },
    {
      "epoch": 6.24,
      "eval_loss": 2.8650901317596436,
      "eval_runtime": 5.3566,
      "eval_samples_per_second": 1843.888,
      "eval_steps_per_second": 14.561,
      "eval_top3_3_weighted_f1_score ": 0.4270264516778953,
      "eval_top_1_macro_f1_score": 0.06016463468597995,
      "eval_top_1_weighted_f1score": 0.20816923117935468,
      "eval_top_3_macro_f1_score": 0.1681092323881849,
      "step": 14528
    },
    {
      "epoch": 6.24,
      "learning_rate": 0.0002,
      "loss": 2.8442,
      "step": 14530
    },
    {
      "epoch": 6.24,
      "learning_rate": 0.0002,
      "loss": 2.8518,
      "step": 14535
    },
    {
      "epoch": 6.24,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 14540
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 14545
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 14550
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.0002,
      "loss": 2.8635,
      "step": 14555
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.0002,
      "loss": 2.8411,
      "step": 14560
    },
    {
      "epoch": 6.25,
      "learning_rate": 0.0002,
      "loss": 2.8618,
      "step": 14565
    },
    {
      "epoch": 6.26,
      "learning_rate": 0.0002,
      "loss": 2.8505,
      "step": 14570
    },
    {
      "epoch": 6.26,
      "learning_rate": 0.0002,
      "loss": 2.8768,
      "step": 14575
    },
    {
      "epoch": 6.26,
      "learning_rate": 0.0002,
      "loss": 2.8765,
      "step": 14580
    },
    {
      "epoch": 6.26,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 14585
    },
    {
      "epoch": 6.26,
      "learning_rate": 0.0002,
      "loss": 2.8706,
      "step": 14590
    },
    {
      "epoch": 6.27,
      "eval_loss": 2.858391523361206,
      "eval_runtime": 5.6623,
      "eval_samples_per_second": 1744.358,
      "eval_steps_per_second": 13.775,
      "eval_top3_3_weighted_f1_score ": 0.4299116980347578,
      "eval_top_1_macro_f1_score": 0.06511235020674873,
      "eval_top_1_weighted_f1score": 0.20653598410434315,
      "eval_top_3_macro_f1_score": 0.16354742919613674,
      "step": 14592
    },
    {
      "epoch": 6.27,
      "learning_rate": 0.0002,
      "loss": 2.8357,
      "step": 14595
    },
    {
      "epoch": 6.27,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 14600
    },
    {
      "epoch": 6.27,
      "learning_rate": 0.0002,
      "loss": 2.8593,
      "step": 14605
    },
    {
      "epoch": 6.27,
      "learning_rate": 0.0002,
      "loss": 2.8504,
      "step": 14610
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.0002,
      "loss": 2.8411,
      "step": 14615
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.0002,
      "loss": 2.8343,
      "step": 14620
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 14625
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.0002,
      "loss": 2.8424,
      "step": 14630
    },
    {
      "epoch": 6.28,
      "learning_rate": 0.0002,
      "loss": 2.833,
      "step": 14635
    },
    {
      "epoch": 6.29,
      "learning_rate": 0.0002,
      "loss": 2.8476,
      "step": 14640
    },
    {
      "epoch": 6.29,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 14645
    },
    {
      "epoch": 6.29,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 14650
    },
    {
      "epoch": 6.29,
      "learning_rate": 0.0002,
      "loss": 2.8433,
      "step": 14655
    },
    {
      "epoch": 6.29,
      "eval_loss": 2.8645198345184326,
      "eval_runtime": 5.3318,
      "eval_samples_per_second": 1852.462,
      "eval_steps_per_second": 14.629,
      "eval_top3_3_weighted_f1_score ": 0.4262237669359407,
      "eval_top_1_macro_f1_score": 0.05589004399179858,
      "eval_top_1_weighted_f1score": 0.20421429552342285,
      "eval_top_3_macro_f1_score": 0.1550932169485424,
      "step": 14656
    },
    {
      "epoch": 6.29,
      "learning_rate": 0.0002,
      "loss": 2.8627,
      "step": 14660
    },
    {
      "epoch": 6.3,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 14665
    },
    {
      "epoch": 6.3,
      "learning_rate": 0.0002,
      "loss": 2.8479,
      "step": 14670
    },
    {
      "epoch": 6.3,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 14675
    },
    {
      "epoch": 6.3,
      "learning_rate": 0.0002,
      "loss": 2.8575,
      "step": 14680
    },
    {
      "epoch": 6.31,
      "learning_rate": 0.0002,
      "loss": 2.8432,
      "step": 14685
    },
    {
      "epoch": 6.31,
      "learning_rate": 0.0002,
      "loss": 2.8668,
      "step": 14690
    },
    {
      "epoch": 6.31,
      "learning_rate": 0.0002,
      "loss": 2.8546,
      "step": 14695
    },
    {
      "epoch": 6.31,
      "learning_rate": 0.0002,
      "loss": 2.8488,
      "step": 14700
    },
    {
      "epoch": 6.31,
      "learning_rate": 0.0002,
      "loss": 2.8841,
      "step": 14705
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 14710
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.0002,
      "loss": 2.8552,
      "step": 14715
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.0002,
      "loss": 2.8619,
      "step": 14720
    },
    {
      "epoch": 6.32,
      "eval_loss": 2.8671915531158447,
      "eval_runtime": 5.3091,
      "eval_samples_per_second": 1860.382,
      "eval_steps_per_second": 14.692,
      "eval_top3_3_weighted_f1_score ": 0.42512401694508484,
      "eval_top_1_macro_f1_score": 0.06521271931683321,
      "eval_top_1_weighted_f1score": 0.21140881058230832,
      "eval_top_3_macro_f1_score": 0.16005632792300933,
      "step": 14720
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.0002,
      "loss": 2.8389,
      "step": 14725
    },
    {
      "epoch": 6.32,
      "learning_rate": 0.0002,
      "loss": 2.843,
      "step": 14730
    },
    {
      "epoch": 6.33,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 14735
    },
    {
      "epoch": 6.33,
      "learning_rate": 0.0002,
      "loss": 2.8522,
      "step": 14740
    },
    {
      "epoch": 6.33,
      "learning_rate": 0.0002,
      "loss": 2.8569,
      "step": 14745
    },
    {
      "epoch": 6.33,
      "learning_rate": 0.0002,
      "loss": 2.863,
      "step": 14750
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.0002,
      "loss": 2.8645,
      "step": 14755
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 14760
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.0002,
      "loss": 2.8377,
      "step": 14765
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.0002,
      "loss": 2.8377,
      "step": 14770
    },
    {
      "epoch": 6.34,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 14775
    },
    {
      "epoch": 6.35,
      "learning_rate": 0.0002,
      "loss": 2.846,
      "step": 14780
    },
    {
      "epoch": 6.35,
      "eval_loss": 2.8662109375,
      "eval_runtime": 5.2681,
      "eval_samples_per_second": 1874.854,
      "eval_steps_per_second": 14.806,
      "eval_top3_3_weighted_f1_score ": 0.4297477209216371,
      "eval_top_1_macro_f1_score": 0.060755136969351614,
      "eval_top_1_weighted_f1score": 0.20593723360268085,
      "eval_top_3_macro_f1_score": 0.17352720917964556,
      "step": 14784
    },
    {
      "epoch": 6.35,
      "learning_rate": 0.0002,
      "loss": 2.8464,
      "step": 14785
    },
    {
      "epoch": 6.35,
      "learning_rate": 0.0002,
      "loss": 2.8563,
      "step": 14790
    },
    {
      "epoch": 6.35,
      "learning_rate": 0.0002,
      "loss": 2.8697,
      "step": 14795
    },
    {
      "epoch": 6.35,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 14800
    },
    {
      "epoch": 6.36,
      "learning_rate": 0.0002,
      "loss": 2.8609,
      "step": 14805
    },
    {
      "epoch": 6.36,
      "learning_rate": 0.0002,
      "loss": 2.8332,
      "step": 14810
    },
    {
      "epoch": 6.36,
      "learning_rate": 0.0002,
      "loss": 2.8574,
      "step": 14815
    },
    {
      "epoch": 6.36,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 14820
    },
    {
      "epoch": 6.37,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 14825
    },
    {
      "epoch": 6.37,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 14830
    },
    {
      "epoch": 6.37,
      "learning_rate": 0.0002,
      "loss": 2.8292,
      "step": 14835
    },
    {
      "epoch": 6.37,
      "learning_rate": 0.0002,
      "loss": 2.8662,
      "step": 14840
    },
    {
      "epoch": 6.37,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 14845
    },
    {
      "epoch": 6.38,
      "eval_loss": 2.859290599822998,
      "eval_runtime": 5.7805,
      "eval_samples_per_second": 1708.674,
      "eval_steps_per_second": 13.494,
      "eval_top3_3_weighted_f1_score ": 0.43133298735934217,
      "eval_top_1_macro_f1_score": 0.0652974806153142,
      "eval_top_1_weighted_f1score": 0.2092498162086767,
      "eval_top_3_macro_f1_score": 0.1750329939017237,
      "step": 14848
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.0002,
      "loss": 2.8341,
      "step": 14850
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.0002,
      "loss": 2.8683,
      "step": 14855
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.0002,
      "loss": 2.8435,
      "step": 14860
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.0002,
      "loss": 2.8594,
      "step": 14865
    },
    {
      "epoch": 6.38,
      "learning_rate": 0.0002,
      "loss": 2.8539,
      "step": 14870
    },
    {
      "epoch": 6.39,
      "learning_rate": 0.0002,
      "loss": 2.8618,
      "step": 14875
    },
    {
      "epoch": 6.39,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 14880
    },
    {
      "epoch": 6.39,
      "learning_rate": 0.0002,
      "loss": 2.8657,
      "step": 14885
    },
    {
      "epoch": 6.39,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 14890
    },
    {
      "epoch": 6.4,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 14895
    },
    {
      "epoch": 6.4,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 14900
    },
    {
      "epoch": 6.4,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 14905
    },
    {
      "epoch": 6.4,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 14910
    },
    {
      "epoch": 6.4,
      "eval_loss": 2.86201810836792,
      "eval_runtime": 5.5374,
      "eval_samples_per_second": 1783.685,
      "eval_steps_per_second": 14.086,
      "eval_top3_3_weighted_f1_score ": 0.4316349888539969,
      "eval_top_1_macro_f1_score": 0.06433221584881162,
      "eval_top_1_weighted_f1score": 0.2086131824809885,
      "eval_top_3_macro_f1_score": 0.184924268898213,
      "step": 14912
    },
    {
      "epoch": 6.4,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 14915
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 14920
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.0002,
      "loss": 2.8746,
      "step": 14925
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.0002,
      "loss": 2.837,
      "step": 14930
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 14935
    },
    {
      "epoch": 6.41,
      "learning_rate": 0.0002,
      "loss": 2.8432,
      "step": 14940
    },
    {
      "epoch": 6.42,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 14945
    },
    {
      "epoch": 6.42,
      "learning_rate": 0.0002,
      "loss": 2.8412,
      "step": 14950
    },
    {
      "epoch": 6.42,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 14955
    },
    {
      "epoch": 6.42,
      "learning_rate": 0.0002,
      "loss": 2.8678,
      "step": 14960
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 14965
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.0002,
      "loss": 2.8524,
      "step": 14970
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.0002,
      "loss": 2.8481,
      "step": 14975
    },
    {
      "epoch": 6.43,
      "eval_loss": 2.8615529537200928,
      "eval_runtime": 5.447,
      "eval_samples_per_second": 1813.276,
      "eval_steps_per_second": 14.32,
      "eval_top3_3_weighted_f1_score ": 0.4279112433014357,
      "eval_top_1_macro_f1_score": 0.06308184819464147,
      "eval_top_1_weighted_f1score": 0.2073751147238242,
      "eval_top_3_macro_f1_score": 0.17409335405517545,
      "step": 14976
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.0002,
      "loss": 2.8486,
      "step": 14980
    },
    {
      "epoch": 6.43,
      "learning_rate": 0.0002,
      "loss": 2.8595,
      "step": 14985
    },
    {
      "epoch": 6.44,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 14990
    },
    {
      "epoch": 6.44,
      "learning_rate": 0.0002,
      "loss": 2.8386,
      "step": 14995
    },
    {
      "epoch": 6.44,
      "learning_rate": 0.0002,
      "loss": 2.8618,
      "step": 15000
    },
    {
      "epoch": 6.44,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 15005
    },
    {
      "epoch": 6.44,
      "learning_rate": 0.0002,
      "loss": 2.8611,
      "step": 15010
    },
    {
      "epoch": 6.45,
      "learning_rate": 0.0002,
      "loss": 2.8674,
      "step": 15015
    },
    {
      "epoch": 6.45,
      "learning_rate": 0.0002,
      "loss": 2.8358,
      "step": 15020
    },
    {
      "epoch": 6.45,
      "learning_rate": 0.0002,
      "loss": 2.8402,
      "step": 15025
    },
    {
      "epoch": 6.45,
      "learning_rate": 0.0002,
      "loss": 2.8661,
      "step": 15030
    },
    {
      "epoch": 6.46,
      "learning_rate": 0.0002,
      "loss": 2.8504,
      "step": 15035
    },
    {
      "epoch": 6.46,
      "learning_rate": 0.0002,
      "loss": 2.8301,
      "step": 15040
    },
    {
      "epoch": 6.46,
      "eval_loss": 2.8677945137023926,
      "eval_runtime": 5.3988,
      "eval_samples_per_second": 1829.48,
      "eval_steps_per_second": 14.448,
      "eval_top3_3_weighted_f1_score ": 0.42455007539366635,
      "eval_top_1_macro_f1_score": 0.058996169356064575,
      "eval_top_1_weighted_f1score": 0.20544254012800425,
      "eval_top_3_macro_f1_score": 0.16829896470607966,
      "step": 15040
    },
    {
      "epoch": 6.46,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 15045
    },
    {
      "epoch": 6.46,
      "learning_rate": 0.0002,
      "loss": 2.8606,
      "step": 15050
    },
    {
      "epoch": 6.46,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 15055
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 15060
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 15065
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 15070
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.0002,
      "loss": 2.8492,
      "step": 15075
    },
    {
      "epoch": 6.47,
      "learning_rate": 0.0002,
      "loss": 2.845,
      "step": 15080
    },
    {
      "epoch": 6.48,
      "learning_rate": 0.0002,
      "loss": 2.8405,
      "step": 15085
    },
    {
      "epoch": 6.48,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 15090
    },
    {
      "epoch": 6.48,
      "learning_rate": 0.0002,
      "loss": 2.8636,
      "step": 15095
    },
    {
      "epoch": 6.48,
      "learning_rate": 0.0002,
      "loss": 2.8454,
      "step": 15100
    },
    {
      "epoch": 6.49,
      "eval_loss": 2.861515522003174,
      "eval_runtime": 5.3588,
      "eval_samples_per_second": 1843.121,
      "eval_steps_per_second": 14.555,
      "eval_top3_3_weighted_f1_score ": 0.4294943102080586,
      "eval_top_1_macro_f1_score": 0.0604777460693949,
      "eval_top_1_weighted_f1score": 0.21113439617484422,
      "eval_top_3_macro_f1_score": 0.1744706343059108,
      "step": 15104
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.0002,
      "loss": 2.8714,
      "step": 15105
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.0002,
      "loss": 2.8188,
      "step": 15110
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.0002,
      "loss": 2.8437,
      "step": 15115
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 15120
    },
    {
      "epoch": 6.49,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 15125
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 15130
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.0002,
      "loss": 2.8497,
      "step": 15135
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 15140
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 15145
    },
    {
      "epoch": 6.5,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 15150
    },
    {
      "epoch": 6.51,
      "learning_rate": 0.0002,
      "loss": 2.8597,
      "step": 15155
    },
    {
      "epoch": 6.51,
      "learning_rate": 0.0002,
      "loss": 2.8521,
      "step": 15160
    },
    {
      "epoch": 6.51,
      "learning_rate": 0.0002,
      "loss": 2.817,
      "step": 15165
    },
    {
      "epoch": 6.51,
      "eval_loss": 2.8600211143493652,
      "eval_runtime": 5.32,
      "eval_samples_per_second": 1856.568,
      "eval_steps_per_second": 14.662,
      "eval_top3_3_weighted_f1_score ": 0.4284376689033966,
      "eval_top_1_macro_f1_score": 0.06438466999821016,
      "eval_top_1_weighted_f1score": 0.20780307893755443,
      "eval_top_3_macro_f1_score": 0.16777399572464152,
      "step": 15168
    },
    {
      "epoch": 6.51,
      "learning_rate": 0.0002,
      "loss": 2.846,
      "step": 15170
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.0002,
      "loss": 2.8572,
      "step": 15175
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 15180
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.0002,
      "loss": 2.8406,
      "step": 15185
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.0002,
      "loss": 2.8641,
      "step": 15190
    },
    {
      "epoch": 6.52,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 15195
    },
    {
      "epoch": 6.53,
      "learning_rate": 0.0002,
      "loss": 2.843,
      "step": 15200
    },
    {
      "epoch": 6.53,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 15205
    },
    {
      "epoch": 6.53,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 15210
    },
    {
      "epoch": 6.53,
      "learning_rate": 0.0002,
      "loss": 2.844,
      "step": 15215
    },
    {
      "epoch": 6.53,
      "learning_rate": 0.0002,
      "loss": 2.8594,
      "step": 15220
    },
    {
      "epoch": 6.54,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 15225
    },
    {
      "epoch": 6.54,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 15230
    },
    {
      "epoch": 6.54,
      "eval_loss": 2.863083839416504,
      "eval_runtime": 5.3545,
      "eval_samples_per_second": 1844.601,
      "eval_steps_per_second": 14.567,
      "eval_top3_3_weighted_f1_score ": 0.4300403285250032,
      "eval_top_1_macro_f1_score": 0.06085514466902401,
      "eval_top_1_weighted_f1score": 0.20511851127430486,
      "eval_top_3_macro_f1_score": 0.17172974713983308,
      "step": 15232
    },
    {
      "epoch": 6.54,
      "learning_rate": 0.0002,
      "loss": 2.8535,
      "step": 15235
    },
    {
      "epoch": 6.54,
      "learning_rate": 0.0002,
      "loss": 2.8695,
      "step": 15240
    },
    {
      "epoch": 6.55,
      "learning_rate": 0.0002,
      "loss": 2.8507,
      "step": 15245
    },
    {
      "epoch": 6.55,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 15250
    },
    {
      "epoch": 6.55,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 15255
    },
    {
      "epoch": 6.55,
      "learning_rate": 0.0002,
      "loss": 2.8357,
      "step": 15260
    },
    {
      "epoch": 6.55,
      "learning_rate": 0.0002,
      "loss": 2.8614,
      "step": 15265
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.0002,
      "loss": 2.8473,
      "step": 15270
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.0002,
      "loss": 2.849,
      "step": 15275
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 15280
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.0002,
      "loss": 2.8493,
      "step": 15285
    },
    {
      "epoch": 6.56,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 15290
    },
    {
      "epoch": 6.57,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 15295
    },
    {
      "epoch": 6.57,
      "eval_loss": 2.8594741821289062,
      "eval_runtime": 5.7069,
      "eval_samples_per_second": 1730.709,
      "eval_steps_per_second": 13.668,
      "eval_top3_3_weighted_f1_score ": 0.4306786416508773,
      "eval_top_1_macro_f1_score": 0.06689709710684523,
      "eval_top_1_weighted_f1score": 0.20801576964445317,
      "eval_top_3_macro_f1_score": 0.1785273189858423,
      "step": 15296
    },
    {
      "epoch": 6.57,
      "learning_rate": 0.0002,
      "loss": 2.8281,
      "step": 15300
    },
    {
      "epoch": 6.57,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 15305
    },
    {
      "epoch": 6.57,
      "learning_rate": 0.0002,
      "loss": 2.8659,
      "step": 15310
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.0002,
      "loss": 2.8617,
      "step": 15315
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.0002,
      "loss": 2.873,
      "step": 15320
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 15325
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.0002,
      "loss": 2.8484,
      "step": 15330
    },
    {
      "epoch": 6.58,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 15335
    },
    {
      "epoch": 6.59,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 15340
    },
    {
      "epoch": 6.59,
      "learning_rate": 0.0002,
      "loss": 2.8507,
      "step": 15345
    },
    {
      "epoch": 6.59,
      "learning_rate": 0.0002,
      "loss": 2.8628,
      "step": 15350
    },
    {
      "epoch": 6.59,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 15355
    },
    {
      "epoch": 6.59,
      "learning_rate": 0.0002,
      "loss": 2.8585,
      "step": 15360
    },
    {
      "epoch": 6.59,
      "eval_loss": 2.8582818508148193,
      "eval_runtime": 5.3443,
      "eval_samples_per_second": 1848.126,
      "eval_steps_per_second": 14.595,
      "eval_top3_3_weighted_f1_score ": 0.4305267363864232,
      "eval_top_1_macro_f1_score": 0.06304345759792579,
      "eval_top_1_weighted_f1score": 0.20939826484016782,
      "eval_top_3_macro_f1_score": 0.17486988539878526,
      "step": 15360
    },
    {
      "epoch": 6.6,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 15365
    },
    {
      "epoch": 6.6,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 15370
    },
    {
      "epoch": 6.6,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 15375
    },
    {
      "epoch": 6.6,
      "learning_rate": 0.0002,
      "loss": 2.8432,
      "step": 15380
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.0002,
      "loss": 2.8479,
      "step": 15385
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 15390
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.0002,
      "loss": 2.8572,
      "step": 15395
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.0002,
      "loss": 2.8398,
      "step": 15400
    },
    {
      "epoch": 6.61,
      "learning_rate": 0.0002,
      "loss": 2.8419,
      "step": 15405
    },
    {
      "epoch": 6.62,
      "learning_rate": 0.0002,
      "loss": 2.8463,
      "step": 15410
    },
    {
      "epoch": 6.62,
      "learning_rate": 0.0002,
      "loss": 2.8609,
      "step": 15415
    },
    {
      "epoch": 6.62,
      "learning_rate": 0.0002,
      "loss": 2.8418,
      "step": 15420
    },
    {
      "epoch": 6.62,
      "eval_loss": 2.8603336811065674,
      "eval_runtime": 5.4654,
      "eval_samples_per_second": 1807.183,
      "eval_steps_per_second": 14.272,
      "eval_top3_3_weighted_f1_score ": 0.4294540806516274,
      "eval_top_1_macro_f1_score": 0.0642858580569476,
      "eval_top_1_weighted_f1score": 0.2125765022957422,
      "eval_top_3_macro_f1_score": 0.17009768405843356,
      "step": 15424
    },
    {
      "epoch": 6.62,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 15425
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.0002,
      "loss": 2.8363,
      "step": 15430
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.0002,
      "loss": 2.8557,
      "step": 15435
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 15440
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.0002,
      "loss": 2.86,
      "step": 15445
    },
    {
      "epoch": 6.63,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 15450
    },
    {
      "epoch": 6.64,
      "learning_rate": 0.0002,
      "loss": 2.8514,
      "step": 15455
    },
    {
      "epoch": 6.64,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 15460
    },
    {
      "epoch": 6.64,
      "learning_rate": 0.0002,
      "loss": 2.8435,
      "step": 15465
    },
    {
      "epoch": 6.64,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 15470
    },
    {
      "epoch": 6.64,
      "learning_rate": 0.0002,
      "loss": 2.8245,
      "step": 15475
    },
    {
      "epoch": 6.65,
      "learning_rate": 0.0002,
      "loss": 2.8375,
      "step": 15480
    },
    {
      "epoch": 6.65,
      "learning_rate": 0.0002,
      "loss": 2.8646,
      "step": 15485
    },
    {
      "epoch": 6.65,
      "eval_loss": 2.8616693019866943,
      "eval_runtime": 5.3635,
      "eval_samples_per_second": 1841.536,
      "eval_steps_per_second": 14.543,
      "eval_top3_3_weighted_f1_score ": 0.4317643351334253,
      "eval_top_1_macro_f1_score": 0.06857994824263079,
      "eval_top_1_weighted_f1score": 0.20923647646907634,
      "eval_top_3_macro_f1_score": 0.17278554096226317,
      "step": 15488
    },
    {
      "epoch": 6.65,
      "learning_rate": 0.0002,
      "loss": 2.8726,
      "step": 15490
    },
    {
      "epoch": 6.65,
      "learning_rate": 0.0002,
      "loss": 2.8421,
      "step": 15495
    },
    {
      "epoch": 6.66,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 15500
    },
    {
      "epoch": 6.66,
      "learning_rate": 0.0002,
      "loss": 2.852,
      "step": 15505
    },
    {
      "epoch": 6.66,
      "learning_rate": 0.0002,
      "loss": 2.8677,
      "step": 15510
    },
    {
      "epoch": 6.66,
      "learning_rate": 0.0002,
      "loss": 2.8664,
      "step": 15515
    },
    {
      "epoch": 6.66,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 15520
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.0002,
      "loss": 2.8589,
      "step": 15525
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.0002,
      "loss": 2.8703,
      "step": 15530
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.0002,
      "loss": 2.8393,
      "step": 15535
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.0002,
      "loss": 2.8345,
      "step": 15540
    },
    {
      "epoch": 6.67,
      "learning_rate": 0.0002,
      "loss": 2.8366,
      "step": 15545
    },
    {
      "epoch": 6.68,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 15550
    },
    {
      "epoch": 6.68,
      "eval_loss": 2.8605828285217285,
      "eval_runtime": 5.4526,
      "eval_samples_per_second": 1811.443,
      "eval_steps_per_second": 14.305,
      "eval_top3_3_weighted_f1_score ": 0.4249779934053836,
      "eval_top_1_macro_f1_score": 0.06433192614509126,
      "eval_top_1_weighted_f1score": 0.20726134325996243,
      "eval_top_3_macro_f1_score": 0.17178580038675367,
      "step": 15552
    },
    {
      "epoch": 6.68,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 15555
    },
    {
      "epoch": 6.68,
      "learning_rate": 0.0002,
      "loss": 2.8728,
      "step": 15560
    },
    {
      "epoch": 6.68,
      "learning_rate": 0.0002,
      "loss": 2.8697,
      "step": 15565
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 15570
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.0002,
      "loss": 2.8389,
      "step": 15575
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 15580
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.0002,
      "loss": 2.8607,
      "step": 15585
    },
    {
      "epoch": 6.69,
      "learning_rate": 0.0002,
      "loss": 2.8615,
      "step": 15590
    },
    {
      "epoch": 6.7,
      "learning_rate": 0.0002,
      "loss": 2.8428,
      "step": 15595
    },
    {
      "epoch": 6.7,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 15600
    },
    {
      "epoch": 6.7,
      "learning_rate": 0.0002,
      "loss": 2.8401,
      "step": 15605
    },
    {
      "epoch": 6.7,
      "learning_rate": 0.0002,
      "loss": 2.8456,
      "step": 15610
    },
    {
      "epoch": 6.7,
      "learning_rate": 0.0002,
      "loss": 2.8484,
      "step": 15615
    },
    {
      "epoch": 6.7,
      "eval_loss": 2.861997604370117,
      "eval_runtime": 5.332,
      "eval_samples_per_second": 1852.407,
      "eval_steps_per_second": 14.629,
      "eval_top3_3_weighted_f1_score ": 0.4288646100979513,
      "eval_top_1_macro_f1_score": 0.06356551941777129,
      "eval_top_1_weighted_f1score": 0.21000433410632016,
      "eval_top_3_macro_f1_score": 0.17679677287315787,
      "step": 15616
    },
    {
      "epoch": 6.71,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 15620
    },
    {
      "epoch": 6.71,
      "learning_rate": 0.0002,
      "loss": 2.857,
      "step": 15625
    },
    {
      "epoch": 6.71,
      "learning_rate": 0.0002,
      "loss": 2.8587,
      "step": 15630
    },
    {
      "epoch": 6.71,
      "learning_rate": 0.0002,
      "loss": 2.8302,
      "step": 15635
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.0002,
      "loss": 2.8559,
      "step": 15640
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.0002,
      "loss": 2.8624,
      "step": 15645
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.0002,
      "loss": 2.8345,
      "step": 15650
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.0002,
      "loss": 2.8756,
      "step": 15655
    },
    {
      "epoch": 6.72,
      "learning_rate": 0.0002,
      "loss": 2.8705,
      "step": 15660
    },
    {
      "epoch": 6.73,
      "learning_rate": 0.0002,
      "loss": 2.8654,
      "step": 15665
    },
    {
      "epoch": 6.73,
      "learning_rate": 0.0002,
      "loss": 2.8623,
      "step": 15670
    },
    {
      "epoch": 6.73,
      "learning_rate": 0.0002,
      "loss": 2.8496,
      "step": 15675
    },
    {
      "epoch": 6.73,
      "learning_rate": 0.0002,
      "loss": 2.8585,
      "step": 15680
    },
    {
      "epoch": 6.73,
      "eval_loss": 2.855945348739624,
      "eval_runtime": 5.2544,
      "eval_samples_per_second": 1879.767,
      "eval_steps_per_second": 14.845,
      "eval_top3_3_weighted_f1_score ": 0.430248813419073,
      "eval_top_1_macro_f1_score": 0.0606318984427841,
      "eval_top_1_weighted_f1score": 0.20736633714019903,
      "eval_top_3_macro_f1_score": 0.17073038099669996,
      "step": 15680
    },
    {
      "epoch": 6.73,
      "learning_rate": 0.0002,
      "loss": 2.8563,
      "step": 15685
    },
    {
      "epoch": 6.74,
      "learning_rate": 0.0002,
      "loss": 2.8636,
      "step": 15690
    },
    {
      "epoch": 6.74,
      "learning_rate": 0.0002,
      "loss": 2.8634,
      "step": 15695
    },
    {
      "epoch": 6.74,
      "learning_rate": 0.0002,
      "loss": 2.8457,
      "step": 15700
    },
    {
      "epoch": 6.74,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 15705
    },
    {
      "epoch": 6.75,
      "learning_rate": 0.0002,
      "loss": 2.8528,
      "step": 15710
    },
    {
      "epoch": 6.75,
      "learning_rate": 0.0002,
      "loss": 2.8508,
      "step": 15715
    },
    {
      "epoch": 6.75,
      "learning_rate": 0.0002,
      "loss": 2.8598,
      "step": 15720
    },
    {
      "epoch": 6.75,
      "learning_rate": 0.0002,
      "loss": 2.8656,
      "step": 15725
    },
    {
      "epoch": 6.75,
      "learning_rate": 0.0002,
      "loss": 2.8538,
      "step": 15730
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.0002,
      "loss": 2.8402,
      "step": 15735
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.0002,
      "loss": 2.8551,
      "step": 15740
    },
    {
      "epoch": 6.76,
      "eval_loss": 2.8596832752227783,
      "eval_runtime": 5.7028,
      "eval_samples_per_second": 1731.955,
      "eval_steps_per_second": 13.677,
      "eval_top3_3_weighted_f1_score ": 0.428473242206308,
      "eval_top_1_macro_f1_score": 0.0657145511200022,
      "eval_top_1_weighted_f1score": 0.21096293119397097,
      "eval_top_3_macro_f1_score": 0.17008036575110236,
      "step": 15744
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.0002,
      "loss": 2.8366,
      "step": 15745
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 15750
    },
    {
      "epoch": 6.76,
      "learning_rate": 0.0002,
      "loss": 2.8522,
      "step": 15755
    },
    {
      "epoch": 6.77,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 15760
    },
    {
      "epoch": 6.77,
      "learning_rate": 0.0002,
      "loss": 2.8458,
      "step": 15765
    },
    {
      "epoch": 6.77,
      "learning_rate": 0.0002,
      "loss": 2.8364,
      "step": 15770
    },
    {
      "epoch": 6.77,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 15775
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.0002,
      "loss": 2.8549,
      "step": 15780
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 15785
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.0002,
      "loss": 2.839,
      "step": 15790
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.0002,
      "loss": 2.8635,
      "step": 15795
    },
    {
      "epoch": 6.78,
      "learning_rate": 0.0002,
      "loss": 2.8578,
      "step": 15800
    },
    {
      "epoch": 6.79,
      "learning_rate": 0.0002,
      "loss": 2.8419,
      "step": 15805
    },
    {
      "epoch": 6.79,
      "eval_loss": 2.8585236072540283,
      "eval_runtime": 5.3958,
      "eval_samples_per_second": 1830.508,
      "eval_steps_per_second": 14.456,
      "eval_top3_3_weighted_f1_score ": 0.4266768311266503,
      "eval_top_1_macro_f1_score": 0.0707578405812454,
      "eval_top_1_weighted_f1score": 0.21065535159659776,
      "eval_top_3_macro_f1_score": 0.16743608079872865,
      "step": 15808
    },
    {
      "epoch": 6.79,
      "learning_rate": 0.0002,
      "loss": 2.8572,
      "step": 15810
    },
    {
      "epoch": 6.79,
      "learning_rate": 0.0002,
      "loss": 2.8292,
      "step": 15815
    },
    {
      "epoch": 6.79,
      "learning_rate": 0.0002,
      "loss": 2.8566,
      "step": 15820
    },
    {
      "epoch": 6.79,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 15825
    },
    {
      "epoch": 6.8,
      "learning_rate": 0.0002,
      "loss": 2.8666,
      "step": 15830
    },
    {
      "epoch": 6.8,
      "learning_rate": 0.0002,
      "loss": 2.845,
      "step": 15835
    },
    {
      "epoch": 6.8,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 15840
    },
    {
      "epoch": 6.8,
      "learning_rate": 0.0002,
      "loss": 2.849,
      "step": 15845
    },
    {
      "epoch": 6.81,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 15850
    },
    {
      "epoch": 6.81,
      "learning_rate": 0.0002,
      "loss": 2.8682,
      "step": 15855
    },
    {
      "epoch": 6.81,
      "learning_rate": 0.0002,
      "loss": 2.8507,
      "step": 15860
    },
    {
      "epoch": 6.81,
      "learning_rate": 0.0002,
      "loss": 2.8448,
      "step": 15865
    },
    {
      "epoch": 6.81,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 15870
    },
    {
      "epoch": 6.81,
      "eval_loss": 2.8602190017700195,
      "eval_runtime": 5.3689,
      "eval_samples_per_second": 1839.652,
      "eval_steps_per_second": 14.528,
      "eval_top3_3_weighted_f1_score ": 0.42848811567997996,
      "eval_top_1_macro_f1_score": 0.06972770009692326,
      "eval_top_1_weighted_f1score": 0.2064827795339559,
      "eval_top_3_macro_f1_score": 0.167845057594611,
      "step": 15872
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 15875
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.0002,
      "loss": 2.8586,
      "step": 15880
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 15885
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 15890
    },
    {
      "epoch": 6.82,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 15895
    },
    {
      "epoch": 6.83,
      "learning_rate": 0.0002,
      "loss": 2.8455,
      "step": 15900
    },
    {
      "epoch": 6.83,
      "learning_rate": 0.0002,
      "loss": 2.8602,
      "step": 15905
    },
    {
      "epoch": 6.83,
      "learning_rate": 0.0002,
      "loss": 2.8512,
      "step": 15910
    },
    {
      "epoch": 6.83,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 15915
    },
    {
      "epoch": 6.84,
      "learning_rate": 0.0002,
      "loss": 2.8506,
      "step": 15920
    },
    {
      "epoch": 6.84,
      "learning_rate": 0.0002,
      "loss": 2.82,
      "step": 15925
    },
    {
      "epoch": 6.84,
      "learning_rate": 0.0002,
      "loss": 2.8432,
      "step": 15930
    },
    {
      "epoch": 6.84,
      "learning_rate": 0.0002,
      "loss": 2.8565,
      "step": 15935
    },
    {
      "epoch": 6.84,
      "eval_loss": 2.863837242126465,
      "eval_runtime": 5.4053,
      "eval_samples_per_second": 1827.283,
      "eval_steps_per_second": 14.43,
      "eval_top3_3_weighted_f1_score ": 0.4302442452852865,
      "eval_top_1_macro_f1_score": 0.060992147784006606,
      "eval_top_1_weighted_f1score": 0.20896368893525422,
      "eval_top_3_macro_f1_score": 0.18208826275398493,
      "step": 15936
    },
    {
      "epoch": 6.84,
      "learning_rate": 0.0002,
      "loss": 2.8579,
      "step": 15940
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.0002,
      "loss": 2.8354,
      "step": 15945
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.0002,
      "loss": 2.8569,
      "step": 15950
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 15955
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 15960
    },
    {
      "epoch": 6.85,
      "learning_rate": 0.0002,
      "loss": 2.8643,
      "step": 15965
    },
    {
      "epoch": 6.86,
      "learning_rate": 0.0002,
      "loss": 2.857,
      "step": 15970
    },
    {
      "epoch": 6.86,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 15975
    },
    {
      "epoch": 6.86,
      "learning_rate": 0.0002,
      "loss": 2.8735,
      "step": 15980
    },
    {
      "epoch": 6.86,
      "learning_rate": 0.0002,
      "loss": 2.8331,
      "step": 15985
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.0002,
      "loss": 2.8428,
      "step": 15990
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.0002,
      "loss": 2.8471,
      "step": 15995
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.0002,
      "loss": 2.8721,
      "step": 16000
    },
    {
      "epoch": 6.87,
      "eval_loss": 2.8585147857666016,
      "eval_runtime": 5.3275,
      "eval_samples_per_second": 1853.98,
      "eval_steps_per_second": 14.641,
      "eval_top3_3_weighted_f1_score ": 0.43199057777332744,
      "eval_top_1_macro_f1_score": 0.06341408243519375,
      "eval_top_1_weighted_f1score": 0.20974999878504372,
      "eval_top_3_macro_f1_score": 0.16988259006005924,
      "step": 16000
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.0002,
      "loss": 2.8619,
      "step": 16005
    },
    {
      "epoch": 6.87,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 16010
    },
    {
      "epoch": 6.88,
      "learning_rate": 0.0002,
      "loss": 2.8292,
      "step": 16015
    },
    {
      "epoch": 6.88,
      "learning_rate": 0.0002,
      "loss": 2.8334,
      "step": 16020
    },
    {
      "epoch": 6.88,
      "learning_rate": 0.0002,
      "loss": 2.8664,
      "step": 16025
    },
    {
      "epoch": 6.88,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 16030
    },
    {
      "epoch": 6.88,
      "learning_rate": 0.0002,
      "loss": 2.8523,
      "step": 16035
    },
    {
      "epoch": 6.89,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 16040
    },
    {
      "epoch": 6.89,
      "learning_rate": 0.0002,
      "loss": 2.8438,
      "step": 16045
    },
    {
      "epoch": 6.89,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 16050
    },
    {
      "epoch": 6.89,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 16055
    },
    {
      "epoch": 6.9,
      "learning_rate": 0.0002,
      "loss": 2.8341,
      "step": 16060
    },
    {
      "epoch": 6.9,
      "eval_loss": 2.851205825805664,
      "eval_runtime": 5.3121,
      "eval_samples_per_second": 1859.328,
      "eval_steps_per_second": 14.683,
      "eval_top3_3_weighted_f1_score ": 0.4297521068243642,
      "eval_top_1_macro_f1_score": 0.061757631764458934,
      "eval_top_1_weighted_f1score": 0.20677145621707246,
      "eval_top_3_macro_f1_score": 0.1682712038279633,
      "step": 16064
    },
    {
      "epoch": 6.9,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 16065
    },
    {
      "epoch": 6.9,
      "learning_rate": 0.0002,
      "loss": 2.8624,
      "step": 16070
    },
    {
      "epoch": 6.9,
      "learning_rate": 0.0002,
      "loss": 2.8467,
      "step": 16075
    },
    {
      "epoch": 6.9,
      "learning_rate": 0.0002,
      "loss": 2.8478,
      "step": 16080
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 16085
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 16090
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 16095
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.0002,
      "loss": 2.8702,
      "step": 16100
    },
    {
      "epoch": 6.91,
      "learning_rate": 0.0002,
      "loss": 2.8443,
      "step": 16105
    },
    {
      "epoch": 6.92,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 16110
    },
    {
      "epoch": 6.92,
      "learning_rate": 0.0002,
      "loss": 2.8426,
      "step": 16115
    },
    {
      "epoch": 6.92,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 16120
    },
    {
      "epoch": 6.92,
      "learning_rate": 0.0002,
      "loss": 2.8273,
      "step": 16125
    },
    {
      "epoch": 6.92,
      "eval_loss": 2.8544063568115234,
      "eval_runtime": 5.3059,
      "eval_samples_per_second": 1861.518,
      "eval_steps_per_second": 14.701,
      "eval_top3_3_weighted_f1_score ": 0.43016403994162344,
      "eval_top_1_macro_f1_score": 0.06482697306747977,
      "eval_top_1_weighted_f1score": 0.2089395643737166,
      "eval_top_3_macro_f1_score": 0.17654727126251615,
      "step": 16128
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 16130
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.0002,
      "loss": 2.8672,
      "step": 16135
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 16140
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.0002,
      "loss": 2.8488,
      "step": 16145
    },
    {
      "epoch": 6.93,
      "learning_rate": 0.0002,
      "loss": 2.8757,
      "step": 16150
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.0002,
      "loss": 2.8401,
      "step": 16155
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 16160
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 16165
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.0002,
      "loss": 2.8396,
      "step": 16170
    },
    {
      "epoch": 6.94,
      "learning_rate": 0.0002,
      "loss": 2.8585,
      "step": 16175
    },
    {
      "epoch": 6.95,
      "learning_rate": 0.0002,
      "loss": 2.8481,
      "step": 16180
    },
    {
      "epoch": 6.95,
      "learning_rate": 0.0002,
      "loss": 2.8782,
      "step": 16185
    },
    {
      "epoch": 6.95,
      "learning_rate": 0.0002,
      "loss": 2.8561,
      "step": 16190
    },
    {
      "epoch": 6.95,
      "eval_loss": 2.861961603164673,
      "eval_runtime": 5.2066,
      "eval_samples_per_second": 1897.024,
      "eval_steps_per_second": 14.981,
      "eval_top3_3_weighted_f1_score ": 0.428182266021052,
      "eval_top_1_macro_f1_score": 0.06571064136595867,
      "eval_top_1_weighted_f1score": 0.21030290543500288,
      "eval_top_3_macro_f1_score": 0.171269395433393,
      "step": 16192
    },
    {
      "epoch": 6.95,
      "learning_rate": 0.0002,
      "loss": 2.8533,
      "step": 16195
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.0002,
      "loss": 2.8391,
      "step": 16200
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 16205
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.0002,
      "loss": 2.8592,
      "step": 16210
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.0002,
      "loss": 2.829,
      "step": 16215
    },
    {
      "epoch": 6.96,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 16220
    },
    {
      "epoch": 6.97,
      "learning_rate": 0.0002,
      "loss": 2.8658,
      "step": 16225
    },
    {
      "epoch": 6.97,
      "learning_rate": 0.0002,
      "loss": 2.8496,
      "step": 16230
    },
    {
      "epoch": 6.97,
      "learning_rate": 0.0002,
      "loss": 2.8422,
      "step": 16235
    },
    {
      "epoch": 6.97,
      "learning_rate": 0.0002,
      "loss": 2.8596,
      "step": 16240
    },
    {
      "epoch": 6.97,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 16245
    },
    {
      "epoch": 6.98,
      "learning_rate": 0.0002,
      "loss": 2.8338,
      "step": 16250
    },
    {
      "epoch": 6.98,
      "learning_rate": 0.0002,
      "loss": 2.8435,
      "step": 16255
    },
    {
      "epoch": 6.98,
      "eval_loss": 2.8606724739074707,
      "eval_runtime": 5.4097,
      "eval_samples_per_second": 1825.793,
      "eval_steps_per_second": 14.419,
      "eval_top3_3_weighted_f1_score ": 0.4300385572016085,
      "eval_top_1_macro_f1_score": 0.06157440821152848,
      "eval_top_1_weighted_f1score": 0.20951877862069976,
      "eval_top_3_macro_f1_score": 0.1660304825701084,
      "step": 16256
    },
    {
      "epoch": 6.98,
      "learning_rate": 0.0002,
      "loss": 2.8539,
      "step": 16260
    },
    {
      "epoch": 6.98,
      "learning_rate": 0.0002,
      "loss": 2.8364,
      "step": 16265
    },
    {
      "epoch": 6.99,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 16270
    },
    {
      "epoch": 6.99,
      "learning_rate": 0.0002,
      "loss": 2.8502,
      "step": 16275
    },
    {
      "epoch": 6.99,
      "learning_rate": 0.0002,
      "loss": 2.8586,
      "step": 16280
    },
    {
      "epoch": 6.99,
      "learning_rate": 0.0002,
      "loss": 2.8518,
      "step": 16285
    },
    {
      "epoch": 6.99,
      "learning_rate": 0.0002,
      "loss": 2.8463,
      "step": 16290
    },
    {
      "epoch": 7.0,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 16295
    },
    {
      "epoch": 7.0,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 16300
    },
    {
      "epoch": 7.0,
      "learning_rate": 0.0002,
      "loss": 3.1218,
      "step": 16305
    },
    {
      "epoch": 7.0,
      "learning_rate": 0.0002,
      "loss": 2.8544,
      "step": 16310
    },
    {
      "epoch": 7.01,
      "learning_rate": 0.0002,
      "loss": 2.8176,
      "step": 16315
    },
    {
      "epoch": 7.01,
      "learning_rate": 0.0002,
      "loss": 2.8156,
      "step": 16320
    },
    {
      "epoch": 7.01,
      "eval_loss": 2.855952262878418,
      "eval_runtime": 5.5545,
      "eval_samples_per_second": 1778.202,
      "eval_steps_per_second": 14.043,
      "eval_top3_3_weighted_f1_score ": 0.4283237284132631,
      "eval_top_1_macro_f1_score": 0.0575789324921102,
      "eval_top_1_weighted_f1score": 0.20848049741441527,
      "eval_top_3_macro_f1_score": 0.17694628353914202,
      "step": 16320
    },
    {
      "epoch": 7.01,
      "learning_rate": 0.0002,
      "loss": 2.8242,
      "step": 16325
    },
    {
      "epoch": 7.01,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 16330
    },
    {
      "epoch": 7.01,
      "learning_rate": 0.0002,
      "loss": 2.8236,
      "step": 16335
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.0002,
      "loss": 2.8283,
      "step": 16340
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 16345
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.0002,
      "loss": 2.8402,
      "step": 16350
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 16355
    },
    {
      "epoch": 7.02,
      "learning_rate": 0.0002,
      "loss": 2.8671,
      "step": 16360
    },
    {
      "epoch": 7.03,
      "learning_rate": 0.0002,
      "loss": 2.8417,
      "step": 16365
    },
    {
      "epoch": 7.03,
      "learning_rate": 0.0002,
      "loss": 2.8531,
      "step": 16370
    },
    {
      "epoch": 7.03,
      "learning_rate": 0.0002,
      "loss": 2.8315,
      "step": 16375
    },
    {
      "epoch": 7.03,
      "learning_rate": 0.0002,
      "loss": 2.8595,
      "step": 16380
    },
    {
      "epoch": 7.03,
      "eval_loss": 2.8588004112243652,
      "eval_runtime": 5.7346,
      "eval_samples_per_second": 1722.354,
      "eval_steps_per_second": 13.602,
      "eval_top3_3_weighted_f1_score ": 0.4273591248370392,
      "eval_top_1_macro_f1_score": 0.061884711993587124,
      "eval_top_1_weighted_f1score": 0.20787224118291617,
      "eval_top_3_macro_f1_score": 0.16624012084061907,
      "step": 16384
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 16385
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.0002,
      "loss": 2.8371,
      "step": 16390
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 16395
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 16400
    },
    {
      "epoch": 7.04,
      "learning_rate": 0.0002,
      "loss": 2.8273,
      "step": 16405
    },
    {
      "epoch": 7.05,
      "learning_rate": 0.0002,
      "loss": 2.8444,
      "step": 16410
    },
    {
      "epoch": 7.05,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 16415
    },
    {
      "epoch": 7.05,
      "learning_rate": 0.0002,
      "loss": 2.8589,
      "step": 16420
    },
    {
      "epoch": 7.05,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 16425
    },
    {
      "epoch": 7.05,
      "learning_rate": 0.0002,
      "loss": 2.8445,
      "step": 16430
    },
    {
      "epoch": 7.06,
      "learning_rate": 0.0002,
      "loss": 2.8309,
      "step": 16435
    },
    {
      "epoch": 7.06,
      "learning_rate": 0.0002,
      "loss": 2.8604,
      "step": 16440
    },
    {
      "epoch": 7.06,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 16445
    },
    {
      "epoch": 7.06,
      "eval_loss": 2.8595006465911865,
      "eval_runtime": 5.3939,
      "eval_samples_per_second": 1831.154,
      "eval_steps_per_second": 14.461,
      "eval_top3_3_weighted_f1_score ": 0.4305098625205523,
      "eval_top_1_macro_f1_score": 0.06630078967768736,
      "eval_top_1_weighted_f1score": 0.20840423447435372,
      "eval_top_3_macro_f1_score": 0.17737793347878275,
      "step": 16448
    },
    {
      "epoch": 7.06,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 16450
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0002,
      "loss": 2.8588,
      "step": 16455
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 16460
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0002,
      "loss": 2.8526,
      "step": 16465
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0002,
      "loss": 2.8304,
      "step": 16470
    },
    {
      "epoch": 7.07,
      "learning_rate": 0.0002,
      "loss": 2.8533,
      "step": 16475
    },
    {
      "epoch": 7.08,
      "learning_rate": 0.0002,
      "loss": 2.8459,
      "step": 16480
    },
    {
      "epoch": 7.08,
      "learning_rate": 0.0002,
      "loss": 2.8561,
      "step": 16485
    },
    {
      "epoch": 7.08,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 16490
    },
    {
      "epoch": 7.08,
      "learning_rate": 0.0002,
      "loss": 2.8329,
      "step": 16495
    },
    {
      "epoch": 7.08,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 16500
    },
    {
      "epoch": 7.09,
      "learning_rate": 0.0002,
      "loss": 2.8499,
      "step": 16505
    },
    {
      "epoch": 7.09,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 16510
    },
    {
      "epoch": 7.09,
      "eval_loss": 2.857330799102783,
      "eval_runtime": 5.4246,
      "eval_samples_per_second": 1820.782,
      "eval_steps_per_second": 14.379,
      "eval_top3_3_weighted_f1_score ": 0.42866429123980676,
      "eval_top_1_macro_f1_score": 0.06555160336361505,
      "eval_top_1_weighted_f1score": 0.2098092332953534,
      "eval_top_3_macro_f1_score": 0.1806800909084033,
      "step": 16512
    },
    {
      "epoch": 7.09,
      "learning_rate": 0.0002,
      "loss": 2.8308,
      "step": 16515
    },
    {
      "epoch": 7.09,
      "learning_rate": 0.0002,
      "loss": 2.8235,
      "step": 16520
    },
    {
      "epoch": 7.1,
      "learning_rate": 0.0002,
      "loss": 2.8311,
      "step": 16525
    },
    {
      "epoch": 7.1,
      "learning_rate": 0.0002,
      "loss": 2.8636,
      "step": 16530
    },
    {
      "epoch": 7.1,
      "learning_rate": 0.0002,
      "loss": 2.8419,
      "step": 16535
    },
    {
      "epoch": 7.1,
      "learning_rate": 0.0002,
      "loss": 2.8332,
      "step": 16540
    },
    {
      "epoch": 7.1,
      "learning_rate": 0.0002,
      "loss": 2.8366,
      "step": 16545
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.0002,
      "loss": 2.8257,
      "step": 16550
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.0002,
      "loss": 2.8402,
      "step": 16555
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.0002,
      "loss": 2.8381,
      "step": 16560
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 16565
    },
    {
      "epoch": 7.11,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 16570
    },
    {
      "epoch": 7.12,
      "learning_rate": 0.0002,
      "loss": 2.8331,
      "step": 16575
    },
    {
      "epoch": 7.12,
      "eval_loss": 2.858490467071533,
      "eval_runtime": 5.3889,
      "eval_samples_per_second": 1832.848,
      "eval_steps_per_second": 14.474,
      "eval_top3_3_weighted_f1_score ": 0.4254713711423142,
      "eval_top_1_macro_f1_score": 0.06492993178622292,
      "eval_top_1_weighted_f1score": 0.20903476903188825,
      "eval_top_3_macro_f1_score": 0.1717600858332048,
      "step": 16576
    },
    {
      "epoch": 7.12,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 16580
    },
    {
      "epoch": 7.12,
      "learning_rate": 0.0002,
      "loss": 2.8354,
      "step": 16585
    },
    {
      "epoch": 7.12,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 16590
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.0002,
      "loss": 2.8325,
      "step": 16595
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.0002,
      "loss": 2.8526,
      "step": 16600
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.0002,
      "loss": 2.8381,
      "step": 16605
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.0002,
      "loss": 2.8632,
      "step": 16610
    },
    {
      "epoch": 7.13,
      "learning_rate": 0.0002,
      "loss": 2.8383,
      "step": 16615
    },
    {
      "epoch": 7.14,
      "learning_rate": 0.0002,
      "loss": 2.839,
      "step": 16620
    },
    {
      "epoch": 7.14,
      "learning_rate": 0.0002,
      "loss": 2.87,
      "step": 16625
    },
    {
      "epoch": 7.14,
      "learning_rate": 0.0002,
      "loss": 2.838,
      "step": 16630
    },
    {
      "epoch": 7.14,
      "learning_rate": 0.0002,
      "loss": 2.8497,
      "step": 16635
    },
    {
      "epoch": 7.14,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 16640
    },
    {
      "epoch": 7.14,
      "eval_loss": 2.857104539871216,
      "eval_runtime": 5.3992,
      "eval_samples_per_second": 1829.344,
      "eval_steps_per_second": 14.447,
      "eval_top3_3_weighted_f1_score ": 0.431505093360038,
      "eval_top_1_macro_f1_score": 0.05961979391858313,
      "eval_top_1_weighted_f1score": 0.20848881834889726,
      "eval_top_3_macro_f1_score": 0.17628531216244345,
      "step": 16640
    },
    {
      "epoch": 7.15,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 16645
    },
    {
      "epoch": 7.15,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 16650
    },
    {
      "epoch": 7.15,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 16655
    },
    {
      "epoch": 7.15,
      "learning_rate": 0.0002,
      "loss": 2.8435,
      "step": 16660
    },
    {
      "epoch": 7.16,
      "learning_rate": 0.0002,
      "loss": 2.8504,
      "step": 16665
    },
    {
      "epoch": 7.16,
      "learning_rate": 0.0002,
      "loss": 2.8377,
      "step": 16670
    },
    {
      "epoch": 7.16,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 16675
    },
    {
      "epoch": 7.16,
      "learning_rate": 0.0002,
      "loss": 2.8468,
      "step": 16680
    },
    {
      "epoch": 7.16,
      "learning_rate": 0.0002,
      "loss": 2.8301,
      "step": 16685
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.0002,
      "loss": 2.8279,
      "step": 16690
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.0002,
      "loss": 2.8477,
      "step": 16695
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.0002,
      "loss": 2.8488,
      "step": 16700
    },
    {
      "epoch": 7.17,
      "eval_loss": 2.8567821979522705,
      "eval_runtime": 5.3288,
      "eval_samples_per_second": 1853.515,
      "eval_steps_per_second": 14.637,
      "eval_top3_3_weighted_f1_score ": 0.426813939517681,
      "eval_top_1_macro_f1_score": 0.060002998836245686,
      "eval_top_1_weighted_f1score": 0.20760508995764726,
      "eval_top_3_macro_f1_score": 0.16427234238860938,
      "step": 16704
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.0002,
      "loss": 2.8347,
      "step": 16705
    },
    {
      "epoch": 7.17,
      "learning_rate": 0.0002,
      "loss": 2.8574,
      "step": 16710
    },
    {
      "epoch": 7.18,
      "learning_rate": 0.0002,
      "loss": 2.8526,
      "step": 16715
    },
    {
      "epoch": 7.18,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 16720
    },
    {
      "epoch": 7.18,
      "learning_rate": 0.0002,
      "loss": 2.8329,
      "step": 16725
    },
    {
      "epoch": 7.18,
      "learning_rate": 0.0002,
      "loss": 2.8123,
      "step": 16730
    },
    {
      "epoch": 7.19,
      "learning_rate": 0.0002,
      "loss": 2.8416,
      "step": 16735
    },
    {
      "epoch": 7.19,
      "learning_rate": 0.0002,
      "loss": 2.8577,
      "step": 16740
    },
    {
      "epoch": 7.19,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 16745
    },
    {
      "epoch": 7.19,
      "learning_rate": 0.0002,
      "loss": 2.8313,
      "step": 16750
    },
    {
      "epoch": 7.19,
      "learning_rate": 0.0002,
      "loss": 2.8366,
      "step": 16755
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.0002,
      "loss": 2.8497,
      "step": 16760
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 16765
    },
    {
      "epoch": 7.2,
      "eval_loss": 2.860691547393799,
      "eval_runtime": 5.302,
      "eval_samples_per_second": 1862.895,
      "eval_steps_per_second": 14.712,
      "eval_top3_3_weighted_f1_score ": 0.4339193446571937,
      "eval_top_1_macro_f1_score": 0.06711008630203333,
      "eval_top_1_weighted_f1score": 0.21108482590150213,
      "eval_top_3_macro_f1_score": 0.17314573020296042,
      "step": 16768
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.0002,
      "loss": 2.8298,
      "step": 16770
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 16775
    },
    {
      "epoch": 7.2,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 16780
    },
    {
      "epoch": 7.21,
      "learning_rate": 0.0002,
      "loss": 2.8411,
      "step": 16785
    },
    {
      "epoch": 7.21,
      "learning_rate": 0.0002,
      "loss": 2.8575,
      "step": 16790
    },
    {
      "epoch": 7.21,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 16795
    },
    {
      "epoch": 7.21,
      "learning_rate": 0.0002,
      "loss": 2.8588,
      "step": 16800
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 16805
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 16810
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0002,
      "loss": 2.8564,
      "step": 16815
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0002,
      "loss": 2.8235,
      "step": 16820
    },
    {
      "epoch": 7.22,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 16825
    },
    {
      "epoch": 7.23,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 16830
    },
    {
      "epoch": 7.23,
      "eval_loss": 2.8646328449249268,
      "eval_runtime": 5.8268,
      "eval_samples_per_second": 1695.101,
      "eval_steps_per_second": 13.386,
      "eval_top3_3_weighted_f1_score ": 0.43004454176978774,
      "eval_top_1_macro_f1_score": 0.06530522520674749,
      "eval_top_1_weighted_f1score": 0.20867859060781246,
      "eval_top_3_macro_f1_score": 0.17294844551379238,
      "step": 16832
    },
    {
      "epoch": 7.23,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 16835
    },
    {
      "epoch": 7.23,
      "learning_rate": 0.0002,
      "loss": 2.8682,
      "step": 16840
    },
    {
      "epoch": 7.23,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 16845
    },
    {
      "epoch": 7.23,
      "learning_rate": 0.0002,
      "loss": 2.8425,
      "step": 16850
    },
    {
      "epoch": 7.24,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 16855
    },
    {
      "epoch": 7.24,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 16860
    },
    {
      "epoch": 7.24,
      "learning_rate": 0.0002,
      "loss": 2.8505,
      "step": 16865
    },
    {
      "epoch": 7.24,
      "learning_rate": 0.0002,
      "loss": 2.8421,
      "step": 16870
    },
    {
      "epoch": 7.25,
      "learning_rate": 0.0002,
      "loss": 2.8664,
      "step": 16875
    },
    {
      "epoch": 7.25,
      "learning_rate": 0.0002,
      "loss": 2.8495,
      "step": 16880
    },
    {
      "epoch": 7.25,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 16885
    },
    {
      "epoch": 7.25,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 16890
    },
    {
      "epoch": 7.25,
      "learning_rate": 0.0002,
      "loss": 2.8473,
      "step": 16895
    },
    {
      "epoch": 7.25,
      "eval_loss": 2.854012966156006,
      "eval_runtime": 5.3389,
      "eval_samples_per_second": 1849.997,
      "eval_steps_per_second": 14.61,
      "eval_top3_3_weighted_f1_score ": 0.4284043207078249,
      "eval_top_1_macro_f1_score": 0.06462930759729849,
      "eval_top_1_weighted_f1score": 0.20830027842242396,
      "eval_top_3_macro_f1_score": 0.1627043973691934,
      "step": 16896
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 16900
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.0002,
      "loss": 2.8373,
      "step": 16905
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.0002,
      "loss": 2.8642,
      "step": 16910
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.0002,
      "loss": 2.8456,
      "step": 16915
    },
    {
      "epoch": 7.26,
      "learning_rate": 0.0002,
      "loss": 2.8349,
      "step": 16920
    },
    {
      "epoch": 7.27,
      "learning_rate": 0.0002,
      "loss": 2.8274,
      "step": 16925
    },
    {
      "epoch": 7.27,
      "learning_rate": 0.0002,
      "loss": 2.8398,
      "step": 16930
    },
    {
      "epoch": 7.27,
      "learning_rate": 0.0002,
      "loss": 2.8426,
      "step": 16935
    },
    {
      "epoch": 7.27,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 16940
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.0002,
      "loss": 2.8443,
      "step": 16945
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 16950
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.0002,
      "loss": 2.8192,
      "step": 16955
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.0002,
      "loss": 2.8599,
      "step": 16960
    },
    {
      "epoch": 7.28,
      "eval_loss": 2.8601081371307373,
      "eval_runtime": 5.3679,
      "eval_samples_per_second": 1840.028,
      "eval_steps_per_second": 14.531,
      "eval_top3_3_weighted_f1_score ": 0.42730106890012653,
      "eval_top_1_macro_f1_score": 0.06571170098153195,
      "eval_top_1_weighted_f1score": 0.21113712987875882,
      "eval_top_3_macro_f1_score": 0.16392468907892946,
      "step": 16960
    },
    {
      "epoch": 7.28,
      "learning_rate": 0.0002,
      "loss": 2.8394,
      "step": 16965
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 16970
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.0002,
      "loss": 2.8244,
      "step": 16975
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.0002,
      "loss": 2.8388,
      "step": 16980
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.0002,
      "loss": 2.8523,
      "step": 16985
    },
    {
      "epoch": 7.29,
      "learning_rate": 0.0002,
      "loss": 2.8568,
      "step": 16990
    },
    {
      "epoch": 7.3,
      "learning_rate": 0.0002,
      "loss": 2.8267,
      "step": 16995
    },
    {
      "epoch": 7.3,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 17000
    },
    {
      "epoch": 7.3,
      "learning_rate": 0.0002,
      "loss": 2.8228,
      "step": 17005
    },
    {
      "epoch": 7.3,
      "learning_rate": 0.0002,
      "loss": 2.8479,
      "step": 17010
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.0002,
      "loss": 2.8194,
      "step": 17015
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.0002,
      "loss": 2.8335,
      "step": 17020
    },
    {
      "epoch": 7.31,
      "eval_loss": 2.8549346923828125,
      "eval_runtime": 5.3656,
      "eval_samples_per_second": 1840.786,
      "eval_steps_per_second": 14.537,
      "eval_top3_3_weighted_f1_score ": 0.42883865358982237,
      "eval_top_1_macro_f1_score": 0.06524114841337716,
      "eval_top_1_weighted_f1score": 0.20846580646556187,
      "eval_top_3_macro_f1_score": 0.16890002329164136,
      "step": 17024
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.0002,
      "loss": 2.8383,
      "step": 17025
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.0002,
      "loss": 2.8451,
      "step": 17030
    },
    {
      "epoch": 7.31,
      "learning_rate": 0.0002,
      "loss": 2.8595,
      "step": 17035
    },
    {
      "epoch": 7.32,
      "learning_rate": 0.0002,
      "loss": 2.8535,
      "step": 17040
    },
    {
      "epoch": 7.32,
      "learning_rate": 0.0002,
      "loss": 2.8801,
      "step": 17045
    },
    {
      "epoch": 7.32,
      "learning_rate": 0.0002,
      "loss": 2.8346,
      "step": 17050
    },
    {
      "epoch": 7.32,
      "learning_rate": 0.0002,
      "loss": 2.8629,
      "step": 17055
    },
    {
      "epoch": 7.32,
      "learning_rate": 0.0002,
      "loss": 2.838,
      "step": 17060
    },
    {
      "epoch": 7.33,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 17065
    },
    {
      "epoch": 7.33,
      "learning_rate": 0.0002,
      "loss": 2.8502,
      "step": 17070
    },
    {
      "epoch": 7.33,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 17075
    },
    {
      "epoch": 7.33,
      "learning_rate": 0.0002,
      "loss": 2.8272,
      "step": 17080
    },
    {
      "epoch": 7.34,
      "learning_rate": 0.0002,
      "loss": 2.8537,
      "step": 17085
    },
    {
      "epoch": 7.34,
      "eval_loss": 2.865147829055786,
      "eval_runtime": 5.3509,
      "eval_samples_per_second": 1845.85,
      "eval_steps_per_second": 14.577,
      "eval_top3_3_weighted_f1_score ": 0.42743006298211117,
      "eval_top_1_macro_f1_score": 0.06463209285279022,
      "eval_top_1_weighted_f1score": 0.2067056738977208,
      "eval_top_3_macro_f1_score": 0.1742909513278907,
      "step": 17088
    },
    {
      "epoch": 7.34,
      "learning_rate": 0.0002,
      "loss": 2.8331,
      "step": 17090
    },
    {
      "epoch": 7.34,
      "learning_rate": 0.0002,
      "loss": 2.8404,
      "step": 17095
    },
    {
      "epoch": 7.34,
      "learning_rate": 0.0002,
      "loss": 2.8499,
      "step": 17100
    },
    {
      "epoch": 7.34,
      "learning_rate": 0.0002,
      "loss": 2.8553,
      "step": 17105
    },
    {
      "epoch": 7.35,
      "learning_rate": 0.0002,
      "loss": 2.8517,
      "step": 17110
    },
    {
      "epoch": 7.35,
      "learning_rate": 0.0002,
      "loss": 2.8432,
      "step": 17115
    },
    {
      "epoch": 7.35,
      "learning_rate": 0.0002,
      "loss": 2.855,
      "step": 17120
    },
    {
      "epoch": 7.35,
      "learning_rate": 0.0002,
      "loss": 2.8396,
      "step": 17125
    },
    {
      "epoch": 7.36,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 17130
    },
    {
      "epoch": 7.36,
      "learning_rate": 0.0002,
      "loss": 2.8232,
      "step": 17135
    },
    {
      "epoch": 7.36,
      "learning_rate": 0.0002,
      "loss": 2.8471,
      "step": 17140
    },
    {
      "epoch": 7.36,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 17145
    },
    {
      "epoch": 7.36,
      "learning_rate": 0.0002,
      "loss": 2.8263,
      "step": 17150
    },
    {
      "epoch": 7.36,
      "eval_loss": 2.8595502376556396,
      "eval_runtime": 5.289,
      "eval_samples_per_second": 1867.477,
      "eval_steps_per_second": 14.748,
      "eval_top3_3_weighted_f1_score ": 0.43148654677842535,
      "eval_top_1_macro_f1_score": 0.061337160560468444,
      "eval_top_1_weighted_f1score": 0.2061431426812684,
      "eval_top_3_macro_f1_score": 0.1739050577609029,
      "step": 17152
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.0002,
      "loss": 2.8307,
      "step": 17155
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.0002,
      "loss": 2.8373,
      "step": 17160
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 17165
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 17170
    },
    {
      "epoch": 7.37,
      "learning_rate": 0.0002,
      "loss": 2.8239,
      "step": 17175
    },
    {
      "epoch": 7.38,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 17180
    },
    {
      "epoch": 7.38,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 17185
    },
    {
      "epoch": 7.38,
      "learning_rate": 0.0002,
      "loss": 2.8442,
      "step": 17190
    },
    {
      "epoch": 7.38,
      "learning_rate": 0.0002,
      "loss": 2.837,
      "step": 17195
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.0002,
      "loss": 2.8311,
      "step": 17200
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 17205
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 17210
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 17215
    },
    {
      "epoch": 7.39,
      "eval_loss": 2.8654897212982178,
      "eval_runtime": 5.3215,
      "eval_samples_per_second": 1856.072,
      "eval_steps_per_second": 14.658,
      "eval_top3_3_weighted_f1_score ": 0.42720531625659613,
      "eval_top_1_macro_f1_score": 0.057758238120331054,
      "eval_top_1_weighted_f1score": 0.2050961597199169,
      "eval_top_3_macro_f1_score": 0.17000776791912028,
      "step": 17216
    },
    {
      "epoch": 7.39,
      "learning_rate": 0.0002,
      "loss": 2.8393,
      "step": 17220
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.0002,
      "loss": 2.8624,
      "step": 17225
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.0002,
      "loss": 2.8341,
      "step": 17230
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.0002,
      "loss": 2.8406,
      "step": 17235
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.0002,
      "loss": 2.8272,
      "step": 17240
    },
    {
      "epoch": 7.4,
      "learning_rate": 0.0002,
      "loss": 2.8519,
      "step": 17245
    },
    {
      "epoch": 7.41,
      "learning_rate": 0.0002,
      "loss": 2.854,
      "step": 17250
    },
    {
      "epoch": 7.41,
      "learning_rate": 0.0002,
      "loss": 2.8476,
      "step": 17255
    },
    {
      "epoch": 7.41,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 17260
    },
    {
      "epoch": 7.41,
      "learning_rate": 0.0002,
      "loss": 2.8343,
      "step": 17265
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.0002,
      "loss": 2.8313,
      "step": 17270
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.0002,
      "loss": 2.8501,
      "step": 17275
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 17280
    },
    {
      "epoch": 7.42,
      "eval_loss": 2.8601627349853516,
      "eval_runtime": 5.333,
      "eval_samples_per_second": 1852.043,
      "eval_steps_per_second": 14.626,
      "eval_top3_3_weighted_f1_score ": 0.4274173263101556,
      "eval_top_1_macro_f1_score": 0.061797421242248705,
      "eval_top_1_weighted_f1score": 0.2106343646505144,
      "eval_top_3_macro_f1_score": 0.17457318219501852,
      "step": 17280
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 17285
    },
    {
      "epoch": 7.42,
      "learning_rate": 0.0002,
      "loss": 2.8343,
      "step": 17290
    },
    {
      "epoch": 7.43,
      "learning_rate": 0.0002,
      "loss": 2.8471,
      "step": 17295
    },
    {
      "epoch": 7.43,
      "learning_rate": 0.0002,
      "loss": 2.8546,
      "step": 17300
    },
    {
      "epoch": 7.43,
      "learning_rate": 0.0002,
      "loss": 2.858,
      "step": 17305
    },
    {
      "epoch": 7.43,
      "learning_rate": 0.0002,
      "loss": 2.8404,
      "step": 17310
    },
    {
      "epoch": 7.43,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 17315
    },
    {
      "epoch": 7.44,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 17320
    },
    {
      "epoch": 7.44,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 17325
    },
    {
      "epoch": 7.44,
      "learning_rate": 0.0002,
      "loss": 2.8559,
      "step": 17330
    },
    {
      "epoch": 7.44,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 17335
    },
    {
      "epoch": 7.45,
      "learning_rate": 0.0002,
      "loss": 2.8388,
      "step": 17340
    },
    {
      "epoch": 7.45,
      "eval_loss": 2.8564577102661133,
      "eval_runtime": 5.3497,
      "eval_samples_per_second": 1846.288,
      "eval_steps_per_second": 14.58,
      "eval_top3_3_weighted_f1_score ": 0.43055422937916166,
      "eval_top_1_macro_f1_score": 0.06300539966389757,
      "eval_top_1_weighted_f1score": 0.20401126849456902,
      "eval_top_3_macro_f1_score": 0.17896822590948389,
      "step": 17344
    },
    {
      "epoch": 7.45,
      "learning_rate": 0.0002,
      "loss": 2.8579,
      "step": 17345
    },
    {
      "epoch": 7.45,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 17350
    },
    {
      "epoch": 7.45,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 17355
    },
    {
      "epoch": 7.45,
      "learning_rate": 0.0002,
      "loss": 2.8675,
      "step": 17360
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.0002,
      "loss": 2.837,
      "step": 17365
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 17370
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.0002,
      "loss": 2.8582,
      "step": 17375
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.0002,
      "loss": 2.8453,
      "step": 17380
    },
    {
      "epoch": 7.46,
      "learning_rate": 0.0002,
      "loss": 2.8357,
      "step": 17385
    },
    {
      "epoch": 7.47,
      "learning_rate": 0.0002,
      "loss": 2.8518,
      "step": 17390
    },
    {
      "epoch": 7.47,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 17395
    },
    {
      "epoch": 7.47,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 17400
    },
    {
      "epoch": 7.47,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 17405
    },
    {
      "epoch": 7.47,
      "eval_loss": 2.8515002727508545,
      "eval_runtime": 5.3558,
      "eval_samples_per_second": 1844.163,
      "eval_steps_per_second": 14.564,
      "eval_top3_3_weighted_f1_score ": 0.434874154355208,
      "eval_top_1_macro_f1_score": 0.06739339181520489,
      "eval_top_1_weighted_f1score": 0.20936395426218773,
      "eval_top_3_macro_f1_score": 0.17809085616540352,
      "step": 17408
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 17410
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.0002,
      "loss": 2.8546,
      "step": 17415
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.0002,
      "loss": 2.8422,
      "step": 17420
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 17425
    },
    {
      "epoch": 7.48,
      "learning_rate": 0.0002,
      "loss": 2.851,
      "step": 17430
    },
    {
      "epoch": 7.49,
      "learning_rate": 0.0002,
      "loss": 2.8567,
      "step": 17435
    },
    {
      "epoch": 7.49,
      "learning_rate": 0.0002,
      "loss": 2.8496,
      "step": 17440
    },
    {
      "epoch": 7.49,
      "learning_rate": 0.0002,
      "loss": 2.8493,
      "step": 17445
    },
    {
      "epoch": 7.49,
      "learning_rate": 0.0002,
      "loss": 2.8474,
      "step": 17450
    },
    {
      "epoch": 7.49,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 17455
    },
    {
      "epoch": 7.5,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 17460
    },
    {
      "epoch": 7.5,
      "learning_rate": 0.0002,
      "loss": 2.8646,
      "step": 17465
    },
    {
      "epoch": 7.5,
      "learning_rate": 0.0002,
      "loss": 2.833,
      "step": 17470
    },
    {
      "epoch": 7.5,
      "eval_loss": 2.8604257106781006,
      "eval_runtime": 5.415,
      "eval_samples_per_second": 1824.014,
      "eval_steps_per_second": 14.404,
      "eval_top3_3_weighted_f1_score ": 0.4266952661163531,
      "eval_top_1_macro_f1_score": 0.06480621602723906,
      "eval_top_1_weighted_f1score": 0.210137807720952,
      "eval_top_3_macro_f1_score": 0.17397079150143166,
      "step": 17472
    },
    {
      "epoch": 7.5,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 17475
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.0002,
      "loss": 2.8534,
      "step": 17480
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.0002,
      "loss": 2.8538,
      "step": 17485
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.0002,
      "loss": 2.8477,
      "step": 17490
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.0002,
      "loss": 2.8335,
      "step": 17495
    },
    {
      "epoch": 7.51,
      "learning_rate": 0.0002,
      "loss": 2.8667,
      "step": 17500
    },
    {
      "epoch": 7.52,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 17505
    },
    {
      "epoch": 7.52,
      "learning_rate": 0.0002,
      "loss": 2.8407,
      "step": 17510
    },
    {
      "epoch": 7.52,
      "learning_rate": 0.0002,
      "loss": 2.8334,
      "step": 17515
    },
    {
      "epoch": 7.52,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 17520
    },
    {
      "epoch": 7.52,
      "learning_rate": 0.0002,
      "loss": 2.8693,
      "step": 17525
    },
    {
      "epoch": 7.53,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 17530
    },
    {
      "epoch": 7.53,
      "learning_rate": 0.0002,
      "loss": 2.8486,
      "step": 17535
    },
    {
      "epoch": 7.53,
      "eval_loss": 2.8551135063171387,
      "eval_runtime": 5.4053,
      "eval_samples_per_second": 1827.293,
      "eval_steps_per_second": 14.43,
      "eval_top3_3_weighted_f1_score ": 0.43152797707811985,
      "eval_top_1_macro_f1_score": 0.06773384679351183,
      "eval_top_1_weighted_f1score": 0.21062981726584978,
      "eval_top_3_macro_f1_score": 0.18280062824655052,
      "step": 17536
    },
    {
      "epoch": 7.53,
      "learning_rate": 0.0002,
      "loss": 2.8527,
      "step": 17540
    },
    {
      "epoch": 7.53,
      "learning_rate": 0.0002,
      "loss": 2.8275,
      "step": 17545
    },
    {
      "epoch": 7.54,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 17550
    },
    {
      "epoch": 7.54,
      "learning_rate": 0.0002,
      "loss": 2.8408,
      "step": 17555
    },
    {
      "epoch": 7.54,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 17560
    },
    {
      "epoch": 7.54,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 17565
    },
    {
      "epoch": 7.54,
      "learning_rate": 0.0002,
      "loss": 2.8439,
      "step": 17570
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 17575
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.0002,
      "loss": 2.8394,
      "step": 17580
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.0002,
      "loss": 2.8424,
      "step": 17585
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 17590
    },
    {
      "epoch": 7.55,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 17595
    },
    {
      "epoch": 7.56,
      "learning_rate": 0.0002,
      "loss": 2.8255,
      "step": 17600
    },
    {
      "epoch": 7.56,
      "eval_loss": 2.8550779819488525,
      "eval_runtime": 5.4082,
      "eval_samples_per_second": 1826.293,
      "eval_steps_per_second": 14.422,
      "eval_top3_3_weighted_f1_score ": 0.42816611431330714,
      "eval_top_1_macro_f1_score": 0.07119551719576099,
      "eval_top_1_weighted_f1score": 0.2091762216346021,
      "eval_top_3_macro_f1_score": 0.17398042630222224,
      "step": 17600
    },
    {
      "epoch": 7.56,
      "learning_rate": 0.0002,
      "loss": 2.8293,
      "step": 17605
    },
    {
      "epoch": 7.56,
      "learning_rate": 0.0002,
      "loss": 2.8567,
      "step": 17610
    },
    {
      "epoch": 7.56,
      "learning_rate": 0.0002,
      "loss": 2.8346,
      "step": 17615
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.0002,
      "loss": 2.8368,
      "step": 17620
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.0002,
      "loss": 2.8357,
      "step": 17625
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.0002,
      "loss": 2.8602,
      "step": 17630
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.0002,
      "loss": 2.8676,
      "step": 17635
    },
    {
      "epoch": 7.57,
      "learning_rate": 0.0002,
      "loss": 2.845,
      "step": 17640
    },
    {
      "epoch": 7.58,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 17645
    },
    {
      "epoch": 7.58,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 17650
    },
    {
      "epoch": 7.58,
      "learning_rate": 0.0002,
      "loss": 2.8216,
      "step": 17655
    },
    {
      "epoch": 7.58,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 17660
    },
    {
      "epoch": 7.58,
      "eval_loss": 2.857983350753784,
      "eval_runtime": 5.3725,
      "eval_samples_per_second": 1838.432,
      "eval_steps_per_second": 14.518,
      "eval_top3_3_weighted_f1_score ": 0.4318687991757832,
      "eval_top_1_macro_f1_score": 0.06183854898403182,
      "eval_top_1_weighted_f1score": 0.20706907494516977,
      "eval_top_3_macro_f1_score": 0.18083329218727304,
      "step": 17664
    },
    {
      "epoch": 7.58,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 17665
    },
    {
      "epoch": 7.59,
      "learning_rate": 0.0002,
      "loss": 2.8465,
      "step": 17670
    },
    {
      "epoch": 7.59,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 17675
    },
    {
      "epoch": 7.59,
      "learning_rate": 0.0002,
      "loss": 2.8234,
      "step": 17680
    },
    {
      "epoch": 7.59,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 17685
    },
    {
      "epoch": 7.6,
      "learning_rate": 0.0002,
      "loss": 2.8487,
      "step": 17690
    },
    {
      "epoch": 7.6,
      "learning_rate": 0.0002,
      "loss": 2.823,
      "step": 17695
    },
    {
      "epoch": 7.6,
      "learning_rate": 0.0002,
      "loss": 2.8287,
      "step": 17700
    },
    {
      "epoch": 7.6,
      "learning_rate": 0.0002,
      "loss": 2.8645,
      "step": 17705
    },
    {
      "epoch": 7.6,
      "learning_rate": 0.0002,
      "loss": 2.8486,
      "step": 17710
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.0002,
      "loss": 2.8471,
      "step": 17715
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.0002,
      "loss": 2.8504,
      "step": 17720
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 17725
    },
    {
      "epoch": 7.61,
      "eval_loss": 2.861382246017456,
      "eval_runtime": 5.3217,
      "eval_samples_per_second": 1855.997,
      "eval_steps_per_second": 14.657,
      "eval_top3_3_weighted_f1_score ": 0.4320698808780226,
      "eval_top_1_macro_f1_score": 0.06481853703279937,
      "eval_top_1_weighted_f1score": 0.21251613785550227,
      "eval_top_3_macro_f1_score": 0.17647934637597118,
      "step": 17728
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.0002,
      "loss": 2.8437,
      "step": 17730
    },
    {
      "epoch": 7.61,
      "learning_rate": 0.0002,
      "loss": 2.8271,
      "step": 17735
    },
    {
      "epoch": 7.62,
      "learning_rate": 0.0002,
      "loss": 2.8377,
      "step": 17740
    },
    {
      "epoch": 7.62,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 17745
    },
    {
      "epoch": 7.62,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 17750
    },
    {
      "epoch": 7.62,
      "learning_rate": 0.0002,
      "loss": 2.862,
      "step": 17755
    },
    {
      "epoch": 7.63,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 17760
    },
    {
      "epoch": 7.63,
      "learning_rate": 0.0002,
      "loss": 2.8407,
      "step": 17765
    },
    {
      "epoch": 7.63,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 17770
    },
    {
      "epoch": 7.63,
      "learning_rate": 0.0002,
      "loss": 2.8445,
      "step": 17775
    },
    {
      "epoch": 7.63,
      "learning_rate": 0.0002,
      "loss": 2.8516,
      "step": 17780
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.0002,
      "loss": 2.872,
      "step": 17785
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.0002,
      "loss": 2.8308,
      "step": 17790
    },
    {
      "epoch": 7.64,
      "eval_loss": 2.8573529720306396,
      "eval_runtime": 5.3564,
      "eval_samples_per_second": 1843.959,
      "eval_steps_per_second": 14.562,
      "eval_top3_3_weighted_f1_score ": 0.43103753460109756,
      "eval_top_1_macro_f1_score": 0.06978424722369772,
      "eval_top_1_weighted_f1score": 0.20948118371885413,
      "eval_top_3_macro_f1_score": 0.18157377242094067,
      "step": 17792
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.0002,
      "loss": 2.842,
      "step": 17795
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.0002,
      "loss": 2.8266,
      "step": 17800
    },
    {
      "epoch": 7.64,
      "learning_rate": 0.0002,
      "loss": 2.8391,
      "step": 17805
    },
    {
      "epoch": 7.65,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 17810
    },
    {
      "epoch": 7.65,
      "learning_rate": 0.0002,
      "loss": 2.8195,
      "step": 17815
    },
    {
      "epoch": 7.65,
      "learning_rate": 0.0002,
      "loss": 2.8669,
      "step": 17820
    },
    {
      "epoch": 7.65,
      "learning_rate": 0.0002,
      "loss": 2.8397,
      "step": 17825
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0002,
      "loss": 2.8409,
      "step": 17830
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0002,
      "loss": 2.8332,
      "step": 17835
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0002,
      "loss": 2.846,
      "step": 17840
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0002,
      "loss": 2.8637,
      "step": 17845
    },
    {
      "epoch": 7.66,
      "learning_rate": 0.0002,
      "loss": 2.8468,
      "step": 17850
    },
    {
      "epoch": 7.67,
      "learning_rate": 0.0002,
      "loss": 2.8444,
      "step": 17855
    },
    {
      "epoch": 7.67,
      "eval_loss": 2.853611707687378,
      "eval_runtime": 5.3499,
      "eval_samples_per_second": 1846.216,
      "eval_steps_per_second": 14.58,
      "eval_top3_3_weighted_f1_score ": 0.4262529393811172,
      "eval_top_1_macro_f1_score": 0.07395114953054589,
      "eval_top_1_weighted_f1score": 0.21340532479024657,
      "eval_top_3_macro_f1_score": 0.18009092560892373,
      "step": 17856
    },
    {
      "epoch": 7.67,
      "learning_rate": 0.0002,
      "loss": 2.8518,
      "step": 17860
    },
    {
      "epoch": 7.67,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 17865
    },
    {
      "epoch": 7.67,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 17870
    },
    {
      "epoch": 7.67,
      "learning_rate": 0.0002,
      "loss": 2.8294,
      "step": 17875
    },
    {
      "epoch": 7.68,
      "learning_rate": 0.0002,
      "loss": 2.839,
      "step": 17880
    },
    {
      "epoch": 7.68,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 17885
    },
    {
      "epoch": 7.68,
      "learning_rate": 0.0002,
      "loss": 2.8153,
      "step": 17890
    },
    {
      "epoch": 7.68,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 17895
    },
    {
      "epoch": 7.69,
      "learning_rate": 0.0002,
      "loss": 2.8368,
      "step": 17900
    },
    {
      "epoch": 7.69,
      "learning_rate": 0.0002,
      "loss": 2.8428,
      "step": 17905
    },
    {
      "epoch": 7.69,
      "learning_rate": 0.0002,
      "loss": 2.8276,
      "step": 17910
    },
    {
      "epoch": 7.69,
      "learning_rate": 0.0002,
      "loss": 2.8422,
      "step": 17915
    },
    {
      "epoch": 7.69,
      "learning_rate": 0.0002,
      "loss": 2.8503,
      "step": 17920
    },
    {
      "epoch": 7.69,
      "eval_loss": 2.8571088314056396,
      "eval_runtime": 5.3514,
      "eval_samples_per_second": 1845.67,
      "eval_steps_per_second": 14.576,
      "eval_top3_3_weighted_f1_score ": 0.4373222408633935,
      "eval_top_1_macro_f1_score": 0.0749892769259755,
      "eval_top_1_weighted_f1score": 0.21236818566394777,
      "eval_top_3_macro_f1_score": 0.1865552557620529,
      "step": 17920
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.0002,
      "loss": 2.8488,
      "step": 17925
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 17930
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.0002,
      "loss": 2.8434,
      "step": 17935
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 17940
    },
    {
      "epoch": 7.7,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 17945
    },
    {
      "epoch": 7.71,
      "learning_rate": 0.0002,
      "loss": 2.8192,
      "step": 17950
    },
    {
      "epoch": 7.71,
      "learning_rate": 0.0002,
      "loss": 2.8557,
      "step": 17955
    },
    {
      "epoch": 7.71,
      "learning_rate": 0.0002,
      "loss": 2.8694,
      "step": 17960
    },
    {
      "epoch": 7.71,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 17965
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.0002,
      "loss": 2.8251,
      "step": 17970
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.0002,
      "loss": 2.8558,
      "step": 17975
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.0002,
      "loss": 2.8445,
      "step": 17980
    },
    {
      "epoch": 7.72,
      "eval_loss": 2.8570690155029297,
      "eval_runtime": 5.301,
      "eval_samples_per_second": 1863.224,
      "eval_steps_per_second": 14.714,
      "eval_top3_3_weighted_f1_score ": 0.4313267960058733,
      "eval_top_1_macro_f1_score": 0.06703520627019759,
      "eval_top_1_weighted_f1score": 0.21127194701671587,
      "eval_top_3_macro_f1_score": 0.17693727592084227,
      "step": 17984
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.0002,
      "loss": 2.8495,
      "step": 17985
    },
    {
      "epoch": 7.72,
      "learning_rate": 0.0002,
      "loss": 2.8478,
      "step": 17990
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 17995
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.0002,
      "loss": 2.8477,
      "step": 18000
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.0002,
      "loss": 2.8507,
      "step": 18005
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 18010
    },
    {
      "epoch": 7.73,
      "learning_rate": 0.0002,
      "loss": 2.8318,
      "step": 18015
    },
    {
      "epoch": 7.74,
      "learning_rate": 0.0002,
      "loss": 2.8793,
      "step": 18020
    },
    {
      "epoch": 7.74,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 18025
    },
    {
      "epoch": 7.74,
      "learning_rate": 0.0002,
      "loss": 2.8333,
      "step": 18030
    },
    {
      "epoch": 7.74,
      "learning_rate": 0.0002,
      "loss": 2.8325,
      "step": 18035
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.0002,
      "loss": 2.8399,
      "step": 18040
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.0002,
      "loss": 2.8254,
      "step": 18045
    },
    {
      "epoch": 7.75,
      "eval_loss": 2.860426902770996,
      "eval_runtime": 5.3043,
      "eval_samples_per_second": 1862.084,
      "eval_steps_per_second": 14.705,
      "eval_top3_3_weighted_f1_score ": 0.4305783468865011,
      "eval_top_1_macro_f1_score": 0.06261944572563972,
      "eval_top_1_weighted_f1score": 0.20871575977838583,
      "eval_top_3_macro_f1_score": 0.16743002794593007,
      "step": 18048
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 18050
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 18055
    },
    {
      "epoch": 7.75,
      "learning_rate": 0.0002,
      "loss": 2.8346,
      "step": 18060
    },
    {
      "epoch": 7.76,
      "learning_rate": 0.0002,
      "loss": 2.8247,
      "step": 18065
    },
    {
      "epoch": 7.76,
      "learning_rate": 0.0002,
      "loss": 2.85,
      "step": 18070
    },
    {
      "epoch": 7.76,
      "learning_rate": 0.0002,
      "loss": 2.8206,
      "step": 18075
    },
    {
      "epoch": 7.76,
      "learning_rate": 0.0002,
      "loss": 2.8467,
      "step": 18080
    },
    {
      "epoch": 7.76,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 18085
    },
    {
      "epoch": 7.77,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 18090
    },
    {
      "epoch": 7.77,
      "learning_rate": 0.0002,
      "loss": 2.8095,
      "step": 18095
    },
    {
      "epoch": 7.77,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 18100
    },
    {
      "epoch": 7.77,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 18105
    },
    {
      "epoch": 7.78,
      "learning_rate": 0.0002,
      "loss": 2.839,
      "step": 18110
    },
    {
      "epoch": 7.78,
      "eval_loss": 2.8566136360168457,
      "eval_runtime": 5.3679,
      "eval_samples_per_second": 1840.015,
      "eval_steps_per_second": 14.531,
      "eval_top3_3_weighted_f1_score ": 0.42925124869710385,
      "eval_top_1_macro_f1_score": 0.06889615487223762,
      "eval_top_1_weighted_f1score": 0.2145421984986216,
      "eval_top_3_macro_f1_score": 0.17001836793742864,
      "step": 18112
    },
    {
      "epoch": 7.78,
      "learning_rate": 0.0002,
      "loss": 2.8465,
      "step": 18115
    },
    {
      "epoch": 7.78,
      "learning_rate": 0.0002,
      "loss": 2.8317,
      "step": 18120
    },
    {
      "epoch": 7.78,
      "learning_rate": 0.0002,
      "loss": 2.8478,
      "step": 18125
    },
    {
      "epoch": 7.78,
      "learning_rate": 0.0002,
      "loss": 2.8657,
      "step": 18130
    },
    {
      "epoch": 7.79,
      "learning_rate": 0.0002,
      "loss": 2.8261,
      "step": 18135
    },
    {
      "epoch": 7.79,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 18140
    },
    {
      "epoch": 7.79,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 18145
    },
    {
      "epoch": 7.79,
      "learning_rate": 0.0002,
      "loss": 2.8539,
      "step": 18150
    },
    {
      "epoch": 7.8,
      "learning_rate": 0.0002,
      "loss": 2.829,
      "step": 18155
    },
    {
      "epoch": 7.8,
      "learning_rate": 0.0002,
      "loss": 2.8471,
      "step": 18160
    },
    {
      "epoch": 7.8,
      "learning_rate": 0.0002,
      "loss": 2.8295,
      "step": 18165
    },
    {
      "epoch": 7.8,
      "learning_rate": 0.0002,
      "loss": 2.8408,
      "step": 18170
    },
    {
      "epoch": 7.8,
      "learning_rate": 0.0002,
      "loss": 2.8316,
      "step": 18175
    },
    {
      "epoch": 7.8,
      "eval_loss": 2.8526108264923096,
      "eval_runtime": 5.3873,
      "eval_samples_per_second": 1833.37,
      "eval_steps_per_second": 14.478,
      "eval_top3_3_weighted_f1_score ": 0.43403000494809485,
      "eval_top_1_macro_f1_score": 0.06221753481705046,
      "eval_top_1_weighted_f1score": 0.20882870929826783,
      "eval_top_3_macro_f1_score": 0.16983619616315163,
      "step": 18176
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 18180
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.0002,
      "loss": 2.8491,
      "step": 18185
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.0002,
      "loss": 2.8227,
      "step": 18190
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.0002,
      "loss": 2.8139,
      "step": 18195
    },
    {
      "epoch": 7.81,
      "learning_rate": 0.0002,
      "loss": 2.8494,
      "step": 18200
    },
    {
      "epoch": 7.82,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 18205
    },
    {
      "epoch": 7.82,
      "learning_rate": 0.0002,
      "loss": 2.8545,
      "step": 18210
    },
    {
      "epoch": 7.82,
      "learning_rate": 0.0002,
      "loss": 2.8594,
      "step": 18215
    },
    {
      "epoch": 7.82,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 18220
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.0002,
      "loss": 2.8349,
      "step": 18225
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 18230
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.0002,
      "loss": 2.8484,
      "step": 18235
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.0002,
      "loss": 2.8556,
      "step": 18240
    },
    {
      "epoch": 7.83,
      "eval_loss": 2.8567590713500977,
      "eval_runtime": 5.4229,
      "eval_samples_per_second": 1821.352,
      "eval_steps_per_second": 14.383,
      "eval_top3_3_weighted_f1_score ": 0.43131154879888567,
      "eval_top_1_macro_f1_score": 0.0663159777747228,
      "eval_top_1_weighted_f1score": 0.2103576812917169,
      "eval_top_3_macro_f1_score": 0.17838595104232294,
      "step": 18240
    },
    {
      "epoch": 7.83,
      "learning_rate": 0.0002,
      "loss": 2.8416,
      "step": 18245
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.0002,
      "loss": 2.8255,
      "step": 18250
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.0002,
      "loss": 2.861,
      "step": 18255
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 18260
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.0002,
      "loss": 2.8465,
      "step": 18265
    },
    {
      "epoch": 7.84,
      "learning_rate": 0.0002,
      "loss": 2.8381,
      "step": 18270
    },
    {
      "epoch": 7.85,
      "learning_rate": 0.0002,
      "loss": 2.8375,
      "step": 18275
    },
    {
      "epoch": 7.85,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 18280
    },
    {
      "epoch": 7.85,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 18285
    },
    {
      "epoch": 7.85,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 18290
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.0002,
      "loss": 2.8228,
      "step": 18295
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 18300
    },
    {
      "epoch": 7.86,
      "eval_loss": 2.8563692569732666,
      "eval_runtime": 5.4208,
      "eval_samples_per_second": 1822.061,
      "eval_steps_per_second": 14.389,
      "eval_top3_3_weighted_f1_score ": 0.4336133923446643,
      "eval_top_1_macro_f1_score": 0.06405406043198943,
      "eval_top_1_weighted_f1score": 0.20811731854029558,
      "eval_top_3_macro_f1_score": 0.17326654889218854,
      "step": 18304
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 18305
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 18310
    },
    {
      "epoch": 7.86,
      "learning_rate": 0.0002,
      "loss": 2.8376,
      "step": 18315
    },
    {
      "epoch": 7.87,
      "learning_rate": 0.0002,
      "loss": 2.8498,
      "step": 18320
    },
    {
      "epoch": 7.87,
      "learning_rate": 0.0002,
      "loss": 2.8484,
      "step": 18325
    },
    {
      "epoch": 7.87,
      "learning_rate": 0.0002,
      "loss": 2.86,
      "step": 18330
    },
    {
      "epoch": 7.87,
      "learning_rate": 0.0002,
      "loss": 2.8683,
      "step": 18335
    },
    {
      "epoch": 7.87,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 18340
    },
    {
      "epoch": 7.88,
      "learning_rate": 0.0002,
      "loss": 2.8455,
      "step": 18345
    },
    {
      "epoch": 7.88,
      "learning_rate": 0.0002,
      "loss": 2.8442,
      "step": 18350
    },
    {
      "epoch": 7.88,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 18355
    },
    {
      "epoch": 7.88,
      "learning_rate": 0.0002,
      "loss": 2.8413,
      "step": 18360
    },
    {
      "epoch": 7.89,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 18365
    },
    {
      "epoch": 7.89,
      "eval_loss": 2.851510763168335,
      "eval_runtime": 5.3567,
      "eval_samples_per_second": 1843.851,
      "eval_steps_per_second": 14.561,
      "eval_top3_3_weighted_f1_score ": 0.42833027638318966,
      "eval_top_1_macro_f1_score": 0.06899612112792806,
      "eval_top_1_weighted_f1score": 0.20976544117055287,
      "eval_top_3_macro_f1_score": 0.17218828481472864,
      "step": 18368
    },
    {
      "epoch": 7.89,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 18370
    },
    {
      "epoch": 7.89,
      "learning_rate": 0.0002,
      "loss": 2.8636,
      "step": 18375
    },
    {
      "epoch": 7.89,
      "learning_rate": 0.0002,
      "loss": 2.8507,
      "step": 18380
    },
    {
      "epoch": 7.89,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 18385
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.0002,
      "loss": 2.8591,
      "step": 18390
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.0002,
      "loss": 2.8563,
      "step": 18395
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.0002,
      "loss": 2.8483,
      "step": 18400
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.0002,
      "loss": 2.8525,
      "step": 18405
    },
    {
      "epoch": 7.9,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 18410
    },
    {
      "epoch": 7.91,
      "learning_rate": 0.0002,
      "loss": 2.8536,
      "step": 18415
    },
    {
      "epoch": 7.91,
      "learning_rate": 0.0002,
      "loss": 2.8541,
      "step": 18420
    },
    {
      "epoch": 7.91,
      "learning_rate": 0.0002,
      "loss": 2.8557,
      "step": 18425
    },
    {
      "epoch": 7.91,
      "learning_rate": 0.0002,
      "loss": 2.8333,
      "step": 18430
    },
    {
      "epoch": 7.91,
      "eval_loss": 2.850963830947876,
      "eval_runtime": 5.4114,
      "eval_samples_per_second": 1825.218,
      "eval_steps_per_second": 14.414,
      "eval_top3_3_weighted_f1_score ": 0.430849103412485,
      "eval_top_1_macro_f1_score": 0.059133958162772844,
      "eval_top_1_weighted_f1score": 0.2083677341411514,
      "eval_top_3_macro_f1_score": 0.1701283381965286,
      "step": 18432
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.0002,
      "loss": 2.8163,
      "step": 18435
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 18440
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.0002,
      "loss": 2.8409,
      "step": 18445
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 18450
    },
    {
      "epoch": 7.92,
      "learning_rate": 0.0002,
      "loss": 2.8315,
      "step": 18455
    },
    {
      "epoch": 7.93,
      "learning_rate": 0.0002,
      "loss": 2.8286,
      "step": 18460
    },
    {
      "epoch": 7.93,
      "learning_rate": 0.0002,
      "loss": 2.8562,
      "step": 18465
    },
    {
      "epoch": 7.93,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 18470
    },
    {
      "epoch": 7.93,
      "learning_rate": 0.0002,
      "loss": 2.8521,
      "step": 18475
    },
    {
      "epoch": 7.93,
      "learning_rate": 0.0002,
      "loss": 2.8579,
      "step": 18480
    },
    {
      "epoch": 7.94,
      "learning_rate": 0.0002,
      "loss": 2.8376,
      "step": 18485
    },
    {
      "epoch": 7.94,
      "learning_rate": 0.0002,
      "loss": 2.8553,
      "step": 18490
    },
    {
      "epoch": 7.94,
      "learning_rate": 0.0002,
      "loss": 2.8424,
      "step": 18495
    },
    {
      "epoch": 7.94,
      "eval_loss": 2.85726261138916,
      "eval_runtime": 5.3679,
      "eval_samples_per_second": 1839.996,
      "eval_steps_per_second": 14.531,
      "eval_top3_3_weighted_f1_score ": 0.4352311699556431,
      "eval_top_1_macro_f1_score": 0.06514683461849503,
      "eval_top_1_weighted_f1score": 0.20901840271096092,
      "eval_top_3_macro_f1_score": 0.1825441574341896,
      "step": 18496
    },
    {
      "epoch": 7.94,
      "learning_rate": 0.0002,
      "loss": 2.8344,
      "step": 18500
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.0002,
      "loss": 2.8308,
      "step": 18505
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 18510
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.0002,
      "loss": 2.8609,
      "step": 18515
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.0002,
      "loss": 2.8437,
      "step": 18520
    },
    {
      "epoch": 7.95,
      "learning_rate": 0.0002,
      "loss": 2.8463,
      "step": 18525
    },
    {
      "epoch": 7.96,
      "learning_rate": 0.0002,
      "loss": 2.8299,
      "step": 18530
    },
    {
      "epoch": 7.96,
      "learning_rate": 0.0002,
      "loss": 2.8522,
      "step": 18535
    },
    {
      "epoch": 7.96,
      "learning_rate": 0.0002,
      "loss": 2.8434,
      "step": 18540
    },
    {
      "epoch": 7.96,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 18545
    },
    {
      "epoch": 7.96,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 18550
    },
    {
      "epoch": 7.97,
      "learning_rate": 0.0002,
      "loss": 2.8344,
      "step": 18555
    },
    {
      "epoch": 7.97,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 18560
    },
    {
      "epoch": 7.97,
      "eval_loss": 2.856273889541626,
      "eval_runtime": 5.4152,
      "eval_samples_per_second": 1823.941,
      "eval_steps_per_second": 14.404,
      "eval_top3_3_weighted_f1_score ": 0.4279170248532264,
      "eval_top_1_macro_f1_score": 0.06686860193563618,
      "eval_top_1_weighted_f1score": 0.2119040248259937,
      "eval_top_3_macro_f1_score": 0.1735783335625637,
      "step": 18560
    },
    {
      "epoch": 7.97,
      "learning_rate": 0.0002,
      "loss": 2.8732,
      "step": 18565
    },
    {
      "epoch": 7.97,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 18570
    },
    {
      "epoch": 7.98,
      "learning_rate": 0.0002,
      "loss": 2.8353,
      "step": 18575
    },
    {
      "epoch": 7.98,
      "learning_rate": 0.0002,
      "loss": 2.8383,
      "step": 18580
    },
    {
      "epoch": 7.98,
      "learning_rate": 0.0002,
      "loss": 2.8234,
      "step": 18585
    },
    {
      "epoch": 7.98,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 18590
    },
    {
      "epoch": 7.98,
      "learning_rate": 0.0002,
      "loss": 2.841,
      "step": 18595
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 18600
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 18605
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.0002,
      "loss": 2.8348,
      "step": 18610
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.0002,
      "loss": 2.8295,
      "step": 18615
    },
    {
      "epoch": 7.99,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 18620
    },
    {
      "epoch": 8.0,
      "eval_loss": 2.8579623699188232,
      "eval_runtime": 5.3757,
      "eval_samples_per_second": 1837.347,
      "eval_steps_per_second": 14.51,
      "eval_top3_3_weighted_f1_score ": 0.43013772806028433,
      "eval_top_1_macro_f1_score": 0.06952676378462946,
      "eval_top_1_weighted_f1score": 0.21015821301808751,
      "eval_top_3_macro_f1_score": 0.18382542806366886,
      "step": 18624
    },
    {
      "epoch": 8.0,
      "learning_rate": 0.0002,
      "loss": 2.8244,
      "step": 18625
    },
    {
      "epoch": 8.0,
      "learning_rate": 0.0002,
      "loss": 2.8217,
      "step": 18630
    },
    {
      "epoch": 8.0,
      "learning_rate": 0.0002,
      "loss": 3.1162,
      "step": 18635
    },
    {
      "epoch": 8.0,
      "learning_rate": 0.0002,
      "loss": 2.8311,
      "step": 18640
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.0002,
      "loss": 2.8119,
      "step": 18645
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.0002,
      "loss": 2.8401,
      "step": 18650
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.0002,
      "loss": 2.8329,
      "step": 18655
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.0002,
      "loss": 2.8304,
      "step": 18660
    },
    {
      "epoch": 8.01,
      "learning_rate": 0.0002,
      "loss": 2.8249,
      "step": 18665
    },
    {
      "epoch": 8.02,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 18670
    },
    {
      "epoch": 8.02,
      "learning_rate": 0.0002,
      "loss": 2.8421,
      "step": 18675
    },
    {
      "epoch": 8.02,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 18680
    },
    {
      "epoch": 8.02,
      "learning_rate": 0.0002,
      "loss": 2.8456,
      "step": 18685
    },
    {
      "epoch": 8.02,
      "eval_loss": 2.8500025272369385,
      "eval_runtime": 5.4191,
      "eval_samples_per_second": 1822.637,
      "eval_steps_per_second": 14.394,
      "eval_top3_3_weighted_f1_score ": 0.4337105301444936,
      "eval_top_1_macro_f1_score": 0.06945696759471479,
      "eval_top_1_weighted_f1score": 0.21319932658089386,
      "eval_top_3_macro_f1_score": 0.17186180065786436,
      "step": 18688
    },
    {
      "epoch": 8.02,
      "learning_rate": 0.0002,
      "loss": 2.8445,
      "step": 18690
    },
    {
      "epoch": 8.03,
      "learning_rate": 0.0002,
      "loss": 2.8188,
      "step": 18695
    },
    {
      "epoch": 8.03,
      "learning_rate": 0.0002,
      "loss": 2.8428,
      "step": 18700
    },
    {
      "epoch": 8.03,
      "learning_rate": 0.0002,
      "loss": 2.8293,
      "step": 18705
    },
    {
      "epoch": 8.03,
      "learning_rate": 0.0002,
      "loss": 2.8368,
      "step": 18710
    },
    {
      "epoch": 8.04,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 18715
    },
    {
      "epoch": 8.04,
      "learning_rate": 0.0002,
      "loss": 2.8299,
      "step": 18720
    },
    {
      "epoch": 8.04,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 18725
    },
    {
      "epoch": 8.04,
      "learning_rate": 0.0002,
      "loss": 2.8205,
      "step": 18730
    },
    {
      "epoch": 8.04,
      "learning_rate": 0.0002,
      "loss": 2.8329,
      "step": 18735
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.0002,
      "loss": 2.81,
      "step": 18740
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 18745
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.0002,
      "loss": 2.8255,
      "step": 18750
    },
    {
      "epoch": 8.05,
      "eval_loss": 2.8510630130767822,
      "eval_runtime": 5.8799,
      "eval_samples_per_second": 1679.792,
      "eval_steps_per_second": 13.266,
      "eval_top3_3_weighted_f1_score ": 0.43241645510398236,
      "eval_top_1_macro_f1_score": 0.0626539668077727,
      "eval_top_1_weighted_f1score": 0.21252954983858618,
      "eval_top_3_macro_f1_score": 0.17364294148805745,
      "step": 18752
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.0002,
      "loss": 2.8283,
      "step": 18755
    },
    {
      "epoch": 8.05,
      "learning_rate": 0.0002,
      "loss": 2.8327,
      "step": 18760
    },
    {
      "epoch": 8.06,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 18765
    },
    {
      "epoch": 8.06,
      "learning_rate": 0.0002,
      "loss": 2.8265,
      "step": 18770
    },
    {
      "epoch": 8.06,
      "learning_rate": 0.0002,
      "loss": 2.8493,
      "step": 18775
    },
    {
      "epoch": 8.06,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 18780
    },
    {
      "epoch": 8.07,
      "learning_rate": 0.0002,
      "loss": 2.8333,
      "step": 18785
    },
    {
      "epoch": 8.07,
      "learning_rate": 0.0002,
      "loss": 2.8442,
      "step": 18790
    },
    {
      "epoch": 8.07,
      "learning_rate": 0.0002,
      "loss": 2.8252,
      "step": 18795
    },
    {
      "epoch": 8.07,
      "learning_rate": 0.0002,
      "loss": 2.8213,
      "step": 18800
    },
    {
      "epoch": 8.07,
      "learning_rate": 0.0002,
      "loss": 2.8389,
      "step": 18805
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.0002,
      "loss": 2.8281,
      "step": 18810
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.0002,
      "loss": 2.8431,
      "step": 18815
    },
    {
      "epoch": 8.08,
      "eval_loss": 2.8637490272521973,
      "eval_runtime": 5.4628,
      "eval_samples_per_second": 1808.041,
      "eval_steps_per_second": 14.278,
      "eval_top3_3_weighted_f1_score ": 0.42713120105614755,
      "eval_top_1_macro_f1_score": 0.06592746106594652,
      "eval_top_1_weighted_f1score": 0.2145307261266979,
      "eval_top_3_macro_f1_score": 0.17053777198677741,
      "step": 18816
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.0002,
      "loss": 2.8538,
      "step": 18820
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.0002,
      "loss": 2.844,
      "step": 18825
    },
    {
      "epoch": 8.08,
      "learning_rate": 0.0002,
      "loss": 2.832,
      "step": 18830
    },
    {
      "epoch": 8.09,
      "learning_rate": 0.0002,
      "loss": 2.8244,
      "step": 18835
    },
    {
      "epoch": 8.09,
      "learning_rate": 0.0002,
      "loss": 2.8194,
      "step": 18840
    },
    {
      "epoch": 8.09,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 18845
    },
    {
      "epoch": 8.09,
      "learning_rate": 0.0002,
      "loss": 2.8327,
      "step": 18850
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.0002,
      "loss": 2.8255,
      "step": 18855
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 18860
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.0002,
      "loss": 2.8184,
      "step": 18865
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 18870
    },
    {
      "epoch": 8.1,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 18875
    },
    {
      "epoch": 8.11,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 18880
    },
    {
      "epoch": 8.11,
      "eval_loss": 2.8579342365264893,
      "eval_runtime": 5.4463,
      "eval_samples_per_second": 1813.524,
      "eval_steps_per_second": 14.322,
      "eval_top3_3_weighted_f1_score ": 0.43202022231306336,
      "eval_top_1_macro_f1_score": 0.06720262193925525,
      "eval_top_1_weighted_f1score": 0.21366382711413187,
      "eval_top_3_macro_f1_score": 0.1870978869165953,
      "step": 18880
    },
    {
      "epoch": 8.11,
      "learning_rate": 0.0002,
      "loss": 2.823,
      "step": 18885
    },
    {
      "epoch": 8.11,
      "learning_rate": 0.0002,
      "loss": 2.8301,
      "step": 18890
    },
    {
      "epoch": 8.11,
      "learning_rate": 0.0002,
      "loss": 2.8216,
      "step": 18895
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 18900
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.0002,
      "loss": 2.8409,
      "step": 18905
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 18910
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 18915
    },
    {
      "epoch": 8.12,
      "learning_rate": 0.0002,
      "loss": 2.8178,
      "step": 18920
    },
    {
      "epoch": 8.13,
      "learning_rate": 0.0002,
      "loss": 2.8139,
      "step": 18925
    },
    {
      "epoch": 8.13,
      "learning_rate": 0.0002,
      "loss": 2.8313,
      "step": 18930
    },
    {
      "epoch": 8.13,
      "learning_rate": 0.0002,
      "loss": 2.8428,
      "step": 18935
    },
    {
      "epoch": 8.13,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 18940
    },
    {
      "epoch": 8.13,
      "eval_loss": 2.853893995285034,
      "eval_runtime": 5.4045,
      "eval_samples_per_second": 1827.541,
      "eval_steps_per_second": 14.432,
      "eval_top3_3_weighted_f1_score ": 0.43363298161690594,
      "eval_top_1_macro_f1_score": 0.06796439537460448,
      "eval_top_1_weighted_f1score": 0.20954028319159415,
      "eval_top_3_macro_f1_score": 0.18018650141645645,
      "step": 18944
    },
    {
      "epoch": 8.13,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 18945
    },
    {
      "epoch": 8.14,
      "learning_rate": 0.0002,
      "loss": 2.8358,
      "step": 18950
    },
    {
      "epoch": 8.14,
      "learning_rate": 0.0002,
      "loss": 2.8285,
      "step": 18955
    },
    {
      "epoch": 8.14,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 18960
    },
    {
      "epoch": 8.14,
      "learning_rate": 0.0002,
      "loss": 2.8241,
      "step": 18965
    },
    {
      "epoch": 8.15,
      "learning_rate": 0.0002,
      "loss": 2.8292,
      "step": 18970
    },
    {
      "epoch": 8.15,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 18975
    },
    {
      "epoch": 8.15,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 18980
    },
    {
      "epoch": 8.15,
      "learning_rate": 0.0002,
      "loss": 2.8272,
      "step": 18985
    },
    {
      "epoch": 8.15,
      "learning_rate": 0.0002,
      "loss": 2.8298,
      "step": 18990
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.0002,
      "loss": 2.8477,
      "step": 18995
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.0002,
      "loss": 2.818,
      "step": 19000
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 19005
    },
    {
      "epoch": 8.16,
      "eval_loss": 2.85410737991333,
      "eval_runtime": 5.4082,
      "eval_samples_per_second": 1826.301,
      "eval_steps_per_second": 14.423,
      "eval_top3_3_weighted_f1_score ": 0.43263465736946743,
      "eval_top_1_macro_f1_score": 0.06272062959897279,
      "eval_top_1_weighted_f1score": 0.21361910459006775,
      "eval_top_3_macro_f1_score": 0.17537673383327287,
      "step": 19008
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.0002,
      "loss": 2.8282,
      "step": 19010
    },
    {
      "epoch": 8.16,
      "learning_rate": 0.0002,
      "loss": 2.8275,
      "step": 19015
    },
    {
      "epoch": 8.17,
      "learning_rate": 0.0002,
      "loss": 2.8249,
      "step": 19020
    },
    {
      "epoch": 8.17,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 19025
    },
    {
      "epoch": 8.17,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 19030
    },
    {
      "epoch": 8.17,
      "learning_rate": 0.0002,
      "loss": 2.8375,
      "step": 19035
    },
    {
      "epoch": 8.18,
      "learning_rate": 0.0002,
      "loss": 2.8474,
      "step": 19040
    },
    {
      "epoch": 8.18,
      "learning_rate": 0.0002,
      "loss": 2.823,
      "step": 19045
    },
    {
      "epoch": 8.18,
      "learning_rate": 0.0002,
      "loss": 2.8401,
      "step": 19050
    },
    {
      "epoch": 8.18,
      "learning_rate": 0.0002,
      "loss": 2.8147,
      "step": 19055
    },
    {
      "epoch": 8.18,
      "learning_rate": 0.0002,
      "loss": 2.8169,
      "step": 19060
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 19065
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.0002,
      "loss": 2.8394,
      "step": 19070
    },
    {
      "epoch": 8.19,
      "eval_loss": 2.8559343814849854,
      "eval_runtime": 5.3943,
      "eval_samples_per_second": 1831.018,
      "eval_steps_per_second": 14.46,
      "eval_top3_3_weighted_f1_score ": 0.4317591661287136,
      "eval_top_1_macro_f1_score": 0.06449347914155465,
      "eval_top_1_weighted_f1score": 0.20687341605906204,
      "eval_top_3_macro_f1_score": 0.17633157286886114,
      "step": 19072
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 19075
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 19080
    },
    {
      "epoch": 8.19,
      "learning_rate": 0.0002,
      "loss": 2.8386,
      "step": 19085
    },
    {
      "epoch": 8.2,
      "learning_rate": 0.0002,
      "loss": 2.8342,
      "step": 19090
    },
    {
      "epoch": 8.2,
      "learning_rate": 0.0002,
      "loss": 2.8268,
      "step": 19095
    },
    {
      "epoch": 8.2,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 19100
    },
    {
      "epoch": 8.2,
      "learning_rate": 0.0002,
      "loss": 2.8249,
      "step": 19105
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.0002,
      "loss": 2.8376,
      "step": 19110
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.0002,
      "loss": 2.8399,
      "step": 19115
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 19120
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.0002,
      "loss": 2.839,
      "step": 19125
    },
    {
      "epoch": 8.21,
      "learning_rate": 0.0002,
      "loss": 2.8173,
      "step": 19130
    },
    {
      "epoch": 8.22,
      "learning_rate": 0.0002,
      "loss": 2.8368,
      "step": 19135
    },
    {
      "epoch": 8.22,
      "eval_loss": 2.8506228923797607,
      "eval_runtime": 5.4809,
      "eval_samples_per_second": 1802.072,
      "eval_steps_per_second": 14.231,
      "eval_top3_3_weighted_f1_score ": 0.4331842785247634,
      "eval_top_1_macro_f1_score": 0.06250492468490476,
      "eval_top_1_weighted_f1score": 0.20867825153577363,
      "eval_top_3_macro_f1_score": 0.1768876667317683,
      "step": 19136
    },
    {
      "epoch": 8.22,
      "learning_rate": 0.0002,
      "loss": 2.8282,
      "step": 19140
    },
    {
      "epoch": 8.22,
      "learning_rate": 0.0002,
      "loss": 2.845,
      "step": 19145
    },
    {
      "epoch": 8.22,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 19150
    },
    {
      "epoch": 8.22,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 19155
    },
    {
      "epoch": 8.23,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 19160
    },
    {
      "epoch": 8.23,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 19165
    },
    {
      "epoch": 8.23,
      "learning_rate": 0.0002,
      "loss": 2.831,
      "step": 19170
    },
    {
      "epoch": 8.23,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 19175
    },
    {
      "epoch": 8.24,
      "learning_rate": 0.0002,
      "loss": 2.8412,
      "step": 19180
    },
    {
      "epoch": 8.24,
      "learning_rate": 0.0002,
      "loss": 2.8408,
      "step": 19185
    },
    {
      "epoch": 8.24,
      "learning_rate": 0.0002,
      "loss": 2.8389,
      "step": 19190
    },
    {
      "epoch": 8.24,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 19195
    },
    {
      "epoch": 8.24,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 19200
    },
    {
      "epoch": 8.24,
      "eval_loss": 2.854755163192749,
      "eval_runtime": 5.4547,
      "eval_samples_per_second": 1810.726,
      "eval_steps_per_second": 14.3,
      "eval_top3_3_weighted_f1_score ": 0.43055610174109016,
      "eval_top_1_macro_f1_score": 0.0635001065286571,
      "eval_top_1_weighted_f1score": 0.21100021654073248,
      "eval_top_3_macro_f1_score": 0.1696295469752659,
      "step": 19200
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.0002,
      "loss": 2.8358,
      "step": 19205
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.0002,
      "loss": 2.8163,
      "step": 19210
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.0002,
      "loss": 2.856,
      "step": 19215
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 19220
    },
    {
      "epoch": 8.25,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 19225
    },
    {
      "epoch": 8.26,
      "learning_rate": 0.0002,
      "loss": 2.8547,
      "step": 19230
    },
    {
      "epoch": 8.26,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 19235
    },
    {
      "epoch": 8.26,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 19240
    },
    {
      "epoch": 8.26,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 19245
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 19250
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.0002,
      "loss": 2.837,
      "step": 19255
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 19260
    },
    {
      "epoch": 8.27,
      "eval_loss": 2.8574471473693848,
      "eval_runtime": 5.3874,
      "eval_samples_per_second": 1833.353,
      "eval_steps_per_second": 14.478,
      "eval_top3_3_weighted_f1_score ": 0.4297157811002323,
      "eval_top_1_macro_f1_score": 0.06663708955187947,
      "eval_top_1_weighted_f1score": 0.20993324481778222,
      "eval_top_3_macro_f1_score": 0.177119064241403,
      "step": 19264
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.0002,
      "loss": 2.8311,
      "step": 19265
    },
    {
      "epoch": 8.27,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 19270
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 19275
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.0002,
      "loss": 2.8371,
      "step": 19280
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 19285
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.0002,
      "loss": 2.8516,
      "step": 19290
    },
    {
      "epoch": 8.28,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 19295
    },
    {
      "epoch": 8.29,
      "learning_rate": 0.0002,
      "loss": 2.8651,
      "step": 19300
    },
    {
      "epoch": 8.29,
      "learning_rate": 0.0002,
      "loss": 2.8431,
      "step": 19305
    },
    {
      "epoch": 8.29,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 19310
    },
    {
      "epoch": 8.29,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 19315
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0002,
      "loss": 2.8259,
      "step": 19320
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 19325
    },
    {
      "epoch": 8.3,
      "eval_loss": 2.853882074356079,
      "eval_runtime": 5.6442,
      "eval_samples_per_second": 1749.924,
      "eval_steps_per_second": 13.819,
      "eval_top3_3_weighted_f1_score ": 0.43012953339607285,
      "eval_top_1_macro_f1_score": 0.06517454054415779,
      "eval_top_1_weighted_f1score": 0.20759689547411692,
      "eval_top_3_macro_f1_score": 0.16976490472592393,
      "step": 19328
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 19330
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0002,
      "loss": 2.8206,
      "step": 19335
    },
    {
      "epoch": 8.3,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 19340
    },
    {
      "epoch": 8.31,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 19345
    },
    {
      "epoch": 8.31,
      "learning_rate": 0.0002,
      "loss": 2.8277,
      "step": 19350
    },
    {
      "epoch": 8.31,
      "learning_rate": 0.0002,
      "loss": 2.8656,
      "step": 19355
    },
    {
      "epoch": 8.31,
      "learning_rate": 0.0002,
      "loss": 2.8476,
      "step": 19360
    },
    {
      "epoch": 8.31,
      "learning_rate": 0.0002,
      "loss": 2.8275,
      "step": 19365
    },
    {
      "epoch": 8.32,
      "learning_rate": 0.0002,
      "loss": 2.8334,
      "step": 19370
    },
    {
      "epoch": 8.32,
      "learning_rate": 0.0002,
      "loss": 2.8598,
      "step": 19375
    },
    {
      "epoch": 8.32,
      "learning_rate": 0.0002,
      "loss": 2.8308,
      "step": 19380
    },
    {
      "epoch": 8.32,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 19385
    },
    {
      "epoch": 8.33,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 19390
    },
    {
      "epoch": 8.33,
      "eval_loss": 2.85400652885437,
      "eval_runtime": 5.3549,
      "eval_samples_per_second": 1844.472,
      "eval_steps_per_second": 14.566,
      "eval_top3_3_weighted_f1_score ": 0.4319273729595622,
      "eval_top_1_macro_f1_score": 0.06498108418862693,
      "eval_top_1_weighted_f1score": 0.2148537004192246,
      "eval_top_3_macro_f1_score": 0.17358489391881235,
      "step": 19392
    },
    {
      "epoch": 8.33,
      "learning_rate": 0.0002,
      "loss": 2.8266,
      "step": 19395
    },
    {
      "epoch": 8.33,
      "learning_rate": 0.0002,
      "loss": 2.8398,
      "step": 19400
    },
    {
      "epoch": 8.33,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 19405
    },
    {
      "epoch": 8.33,
      "learning_rate": 0.0002,
      "loss": 2.8331,
      "step": 19410
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.0002,
      "loss": 2.8458,
      "step": 19415
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.0002,
      "loss": 2.8197,
      "step": 19420
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.0002,
      "loss": 2.8347,
      "step": 19425
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.0002,
      "loss": 2.8368,
      "step": 19430
    },
    {
      "epoch": 8.34,
      "learning_rate": 0.0002,
      "loss": 2.8354,
      "step": 19435
    },
    {
      "epoch": 8.35,
      "learning_rate": 0.0002,
      "loss": 2.8447,
      "step": 19440
    },
    {
      "epoch": 8.35,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 19445
    },
    {
      "epoch": 8.35,
      "learning_rate": 0.0002,
      "loss": 2.8329,
      "step": 19450
    },
    {
      "epoch": 8.35,
      "learning_rate": 0.0002,
      "loss": 2.829,
      "step": 19455
    },
    {
      "epoch": 8.35,
      "eval_loss": 2.8535819053649902,
      "eval_runtime": 5.399,
      "eval_samples_per_second": 1829.409,
      "eval_steps_per_second": 14.447,
      "eval_top3_3_weighted_f1_score ": 0.4313302234369109,
      "eval_top_1_macro_f1_score": 0.0752702175481044,
      "eval_top_1_weighted_f1score": 0.20996350239635977,
      "eval_top_3_macro_f1_score": 0.18139335229578923,
      "step": 19456
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0002,
      "loss": 2.844,
      "step": 19460
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0002,
      "loss": 2.8342,
      "step": 19465
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0002,
      "loss": 2.8224,
      "step": 19470
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0002,
      "loss": 2.8198,
      "step": 19475
    },
    {
      "epoch": 8.36,
      "learning_rate": 0.0002,
      "loss": 2.8426,
      "step": 19480
    },
    {
      "epoch": 8.37,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 19485
    },
    {
      "epoch": 8.37,
      "learning_rate": 0.0002,
      "loss": 2.8521,
      "step": 19490
    },
    {
      "epoch": 8.37,
      "learning_rate": 0.0002,
      "loss": 2.8315,
      "step": 19495
    },
    {
      "epoch": 8.37,
      "learning_rate": 0.0002,
      "loss": 2.8425,
      "step": 19500
    },
    {
      "epoch": 8.37,
      "learning_rate": 0.0002,
      "loss": 2.8376,
      "step": 19505
    },
    {
      "epoch": 8.38,
      "learning_rate": 0.0002,
      "loss": 2.8233,
      "step": 19510
    },
    {
      "epoch": 8.38,
      "learning_rate": 0.0002,
      "loss": 2.8257,
      "step": 19515
    },
    {
      "epoch": 8.38,
      "learning_rate": 0.0002,
      "loss": 2.8347,
      "step": 19520
    },
    {
      "epoch": 8.38,
      "eval_loss": 2.853299617767334,
      "eval_runtime": 5.3715,
      "eval_samples_per_second": 1838.789,
      "eval_steps_per_second": 14.521,
      "eval_top3_3_weighted_f1_score ": 0.4320093917224778,
      "eval_top_1_macro_f1_score": 0.06644228036784446,
      "eval_top_1_weighted_f1score": 0.20949961738693795,
      "eval_top_3_macro_f1_score": 0.1840076991094436,
      "step": 19520
    },
    {
      "epoch": 8.38,
      "learning_rate": 0.0002,
      "loss": 2.8314,
      "step": 19525
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.0002,
      "loss": 2.8439,
      "step": 19530
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.0002,
      "loss": 2.8226,
      "step": 19535
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.0002,
      "loss": 2.8342,
      "step": 19540
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.0002,
      "loss": 2.8424,
      "step": 19545
    },
    {
      "epoch": 8.39,
      "learning_rate": 0.0002,
      "loss": 2.822,
      "step": 19550
    },
    {
      "epoch": 8.4,
      "learning_rate": 0.0002,
      "loss": 2.8421,
      "step": 19555
    },
    {
      "epoch": 8.4,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 19560
    },
    {
      "epoch": 8.4,
      "learning_rate": 0.0002,
      "loss": 2.8192,
      "step": 19565
    },
    {
      "epoch": 8.4,
      "learning_rate": 0.0002,
      "loss": 2.8418,
      "step": 19570
    },
    {
      "epoch": 8.4,
      "learning_rate": 0.0002,
      "loss": 2.8484,
      "step": 19575
    },
    {
      "epoch": 8.41,
      "learning_rate": 0.0002,
      "loss": 2.8314,
      "step": 19580
    },
    {
      "epoch": 8.41,
      "eval_loss": 2.8602147102355957,
      "eval_runtime": 5.3751,
      "eval_samples_per_second": 1837.561,
      "eval_steps_per_second": 14.511,
      "eval_top3_3_weighted_f1_score ": 0.4269702864997268,
      "eval_top_1_macro_f1_score": 0.0700457680521881,
      "eval_top_1_weighted_f1score": 0.21204503998643293,
      "eval_top_3_macro_f1_score": 0.17994195026074286,
      "step": 19584
    },
    {
      "epoch": 8.41,
      "learning_rate": 0.0002,
      "loss": 2.8422,
      "step": 19585
    },
    {
      "epoch": 8.41,
      "learning_rate": 0.0002,
      "loss": 2.8316,
      "step": 19590
    },
    {
      "epoch": 8.41,
      "learning_rate": 0.0002,
      "loss": 2.8193,
      "step": 19595
    },
    {
      "epoch": 8.42,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 19600
    },
    {
      "epoch": 8.42,
      "learning_rate": 0.0002,
      "loss": 2.8263,
      "step": 19605
    },
    {
      "epoch": 8.42,
      "learning_rate": 0.0002,
      "loss": 2.8465,
      "step": 19610
    },
    {
      "epoch": 8.42,
      "learning_rate": 0.0002,
      "loss": 2.8213,
      "step": 19615
    },
    {
      "epoch": 8.42,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 19620
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.0002,
      "loss": 2.8208,
      "step": 19625
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.0002,
      "loss": 2.8587,
      "step": 19630
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.0002,
      "loss": 2.8466,
      "step": 19635
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.0002,
      "loss": 2.8328,
      "step": 19640
    },
    {
      "epoch": 8.43,
      "learning_rate": 0.0002,
      "loss": 2.843,
      "step": 19645
    },
    {
      "epoch": 8.44,
      "eval_loss": 2.85685396194458,
      "eval_runtime": 5.4434,
      "eval_samples_per_second": 1814.498,
      "eval_steps_per_second": 14.329,
      "eval_top3_3_weighted_f1_score ": 0.43339844212291834,
      "eval_top_1_macro_f1_score": 0.06649034474737178,
      "eval_top_1_weighted_f1score": 0.2085780776842864,
      "eval_top_3_macro_f1_score": 0.17763883837651623,
      "step": 19648
    },
    {
      "epoch": 8.44,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 19650
    },
    {
      "epoch": 8.44,
      "learning_rate": 0.0002,
      "loss": 2.8353,
      "step": 19655
    },
    {
      "epoch": 8.44,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 19660
    },
    {
      "epoch": 8.44,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 19665
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.0002,
      "loss": 2.8407,
      "step": 19670
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 19675
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 19680
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.0002,
      "loss": 2.8458,
      "step": 19685
    },
    {
      "epoch": 8.45,
      "learning_rate": 0.0002,
      "loss": 2.8136,
      "step": 19690
    },
    {
      "epoch": 8.46,
      "learning_rate": 0.0002,
      "loss": 2.8428,
      "step": 19695
    },
    {
      "epoch": 8.46,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 19700
    },
    {
      "epoch": 8.46,
      "learning_rate": 0.0002,
      "loss": 2.8398,
      "step": 19705
    },
    {
      "epoch": 8.46,
      "learning_rate": 0.0002,
      "loss": 2.8402,
      "step": 19710
    },
    {
      "epoch": 8.46,
      "eval_loss": 2.8591010570526123,
      "eval_runtime": 5.4576,
      "eval_samples_per_second": 1809.764,
      "eval_steps_per_second": 14.292,
      "eval_top3_3_weighted_f1_score ": 0.42780982053014366,
      "eval_top_1_macro_f1_score": 0.06383286805381078,
      "eval_top_1_weighted_f1score": 0.21159359693828264,
      "eval_top_3_macro_f1_score": 0.16860038129077684,
      "step": 19712
    },
    {
      "epoch": 8.46,
      "learning_rate": 0.0002,
      "loss": 2.8338,
      "step": 19715
    },
    {
      "epoch": 8.47,
      "learning_rate": 0.0002,
      "loss": 2.8248,
      "step": 19720
    },
    {
      "epoch": 8.47,
      "learning_rate": 0.0002,
      "loss": 2.8464,
      "step": 19725
    },
    {
      "epoch": 8.47,
      "learning_rate": 0.0002,
      "loss": 2.8248,
      "step": 19730
    },
    {
      "epoch": 8.47,
      "learning_rate": 0.0002,
      "loss": 2.8232,
      "step": 19735
    },
    {
      "epoch": 8.48,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 19740
    },
    {
      "epoch": 8.48,
      "learning_rate": 0.0002,
      "loss": 2.854,
      "step": 19745
    },
    {
      "epoch": 8.48,
      "learning_rate": 0.0002,
      "loss": 2.8439,
      "step": 19750
    },
    {
      "epoch": 8.48,
      "learning_rate": 0.0002,
      "loss": 2.8221,
      "step": 19755
    },
    {
      "epoch": 8.48,
      "learning_rate": 0.0002,
      "loss": 2.8253,
      "step": 19760
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 19765
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.0002,
      "loss": 2.8457,
      "step": 19770
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.0002,
      "loss": 2.8446,
      "step": 19775
    },
    {
      "epoch": 8.49,
      "eval_loss": 2.853350877761841,
      "eval_runtime": 5.7598,
      "eval_samples_per_second": 1714.83,
      "eval_steps_per_second": 13.542,
      "eval_top3_3_weighted_f1_score ": 0.42941646595194505,
      "eval_top_1_macro_f1_score": 0.06501042798199738,
      "eval_top_1_weighted_f1score": 0.21478077419947375,
      "eval_top_3_macro_f1_score": 0.1626845115439078,
      "step": 19776
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 19780
    },
    {
      "epoch": 8.49,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 19785
    },
    {
      "epoch": 8.5,
      "learning_rate": 0.0002,
      "loss": 2.8443,
      "step": 19790
    },
    {
      "epoch": 8.5,
      "learning_rate": 0.0002,
      "loss": 2.8409,
      "step": 19795
    },
    {
      "epoch": 8.5,
      "learning_rate": 0.0002,
      "loss": 2.8506,
      "step": 19800
    },
    {
      "epoch": 8.5,
      "learning_rate": 0.0002,
      "loss": 2.8418,
      "step": 19805
    },
    {
      "epoch": 8.51,
      "learning_rate": 0.0002,
      "loss": 2.8459,
      "step": 19810
    },
    {
      "epoch": 8.51,
      "learning_rate": 0.0002,
      "loss": 2.8261,
      "step": 19815
    },
    {
      "epoch": 8.51,
      "learning_rate": 0.0002,
      "loss": 2.8281,
      "step": 19820
    },
    {
      "epoch": 8.51,
      "learning_rate": 0.0002,
      "loss": 2.8418,
      "step": 19825
    },
    {
      "epoch": 8.51,
      "learning_rate": 0.0002,
      "loss": 2.8225,
      "step": 19830
    },
    {
      "epoch": 8.52,
      "learning_rate": 0.0002,
      "loss": 2.8432,
      "step": 19835
    },
    {
      "epoch": 8.52,
      "learning_rate": 0.0002,
      "loss": 2.8311,
      "step": 19840
    },
    {
      "epoch": 8.52,
      "eval_loss": 2.854144334793091,
      "eval_runtime": 5.3638,
      "eval_samples_per_second": 1841.435,
      "eval_steps_per_second": 14.542,
      "eval_top3_3_weighted_f1_score ": 0.43159365451982973,
      "eval_top_1_macro_f1_score": 0.06598902755694942,
      "eval_top_1_weighted_f1score": 0.206736239520406,
      "eval_top_3_macro_f1_score": 0.17230136260454781,
      "step": 19840
    },
    {
      "epoch": 8.52,
      "learning_rate": 0.0002,
      "loss": 2.8188,
      "step": 19845
    },
    {
      "epoch": 8.52,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 19850
    },
    {
      "epoch": 8.53,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 19855
    },
    {
      "epoch": 8.53,
      "learning_rate": 0.0002,
      "loss": 2.8259,
      "step": 19860
    },
    {
      "epoch": 8.53,
      "learning_rate": 0.0002,
      "loss": 2.8292,
      "step": 19865
    },
    {
      "epoch": 8.53,
      "learning_rate": 0.0002,
      "loss": 2.8334,
      "step": 19870
    },
    {
      "epoch": 8.53,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 19875
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.0002,
      "loss": 2.8444,
      "step": 19880
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.0002,
      "loss": 2.8333,
      "step": 19885
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.0002,
      "loss": 2.8253,
      "step": 19890
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.0002,
      "loss": 2.8226,
      "step": 19895
    },
    {
      "epoch": 8.54,
      "learning_rate": 0.0002,
      "loss": 2.8468,
      "step": 19900
    },
    {
      "epoch": 8.55,
      "eval_loss": 2.85261607170105,
      "eval_runtime": 5.3816,
      "eval_samples_per_second": 1835.334,
      "eval_steps_per_second": 14.494,
      "eval_top3_3_weighted_f1_score ": 0.4352208097228637,
      "eval_top_1_macro_f1_score": 0.06523791958540971,
      "eval_top_1_weighted_f1score": 0.20880045846759562,
      "eval_top_3_macro_f1_score": 0.18290744854017976,
      "step": 19904
    },
    {
      "epoch": 8.55,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 19905
    },
    {
      "epoch": 8.55,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 19910
    },
    {
      "epoch": 8.55,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 19915
    },
    {
      "epoch": 8.55,
      "learning_rate": 0.0002,
      "loss": 2.8208,
      "step": 19920
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.0002,
      "loss": 2.8297,
      "step": 19925
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.0002,
      "loss": 2.8514,
      "step": 19930
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.0002,
      "loss": 2.8288,
      "step": 19935
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.0002,
      "loss": 2.8159,
      "step": 19940
    },
    {
      "epoch": 8.56,
      "learning_rate": 0.0002,
      "loss": 2.8222,
      "step": 19945
    },
    {
      "epoch": 8.57,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 19950
    },
    {
      "epoch": 8.57,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 19955
    },
    {
      "epoch": 8.57,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 19960
    },
    {
      "epoch": 8.57,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 19965
    },
    {
      "epoch": 8.57,
      "eval_loss": 2.851257562637329,
      "eval_runtime": 5.3655,
      "eval_samples_per_second": 1840.844,
      "eval_steps_per_second": 14.537,
      "eval_top3_3_weighted_f1_score ": 0.4299533869655781,
      "eval_top_1_macro_f1_score": 0.06771736772473094,
      "eval_top_1_weighted_f1score": 0.21492597330227603,
      "eval_top_3_macro_f1_score": 0.18193064565020653,
      "step": 19968
    },
    {
      "epoch": 8.57,
      "learning_rate": 0.0002,
      "loss": 2.8149,
      "step": 19970
    },
    {
      "epoch": 8.58,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 19975
    },
    {
      "epoch": 8.58,
      "learning_rate": 0.0002,
      "loss": 2.8327,
      "step": 19980
    },
    {
      "epoch": 8.58,
      "learning_rate": 0.0002,
      "loss": 2.8409,
      "step": 19985
    },
    {
      "epoch": 8.58,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 19990
    },
    {
      "epoch": 8.59,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 19995
    },
    {
      "epoch": 8.59,
      "learning_rate": 0.0002,
      "loss": 2.8304,
      "step": 20000
    },
    {
      "epoch": 8.59,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 20005
    },
    {
      "epoch": 8.59,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 20010
    },
    {
      "epoch": 8.59,
      "learning_rate": 0.0002,
      "loss": 2.8327,
      "step": 20015
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.0002,
      "loss": 2.831,
      "step": 20020
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.0002,
      "loss": 2.837,
      "step": 20025
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 20030
    },
    {
      "epoch": 8.6,
      "eval_loss": 2.8540263175964355,
      "eval_runtime": 5.3347,
      "eval_samples_per_second": 1851.452,
      "eval_steps_per_second": 14.621,
      "eval_top3_3_weighted_f1_score ": 0.4341394672567419,
      "eval_top_1_macro_f1_score": 0.06408709442512175,
      "eval_top_1_weighted_f1score": 0.2087100711924547,
      "eval_top_3_macro_f1_score": 0.18112913049426627,
      "step": 20032
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 20035
    },
    {
      "epoch": 8.6,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 20040
    },
    {
      "epoch": 8.61,
      "learning_rate": 0.0002,
      "loss": 2.8605,
      "step": 20045
    },
    {
      "epoch": 8.61,
      "learning_rate": 0.0002,
      "loss": 2.8267,
      "step": 20050
    },
    {
      "epoch": 8.61,
      "learning_rate": 0.0002,
      "loss": 2.8079,
      "step": 20055
    },
    {
      "epoch": 8.61,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 20060
    },
    {
      "epoch": 8.62,
      "learning_rate": 0.0002,
      "loss": 2.8371,
      "step": 20065
    },
    {
      "epoch": 8.62,
      "learning_rate": 0.0002,
      "loss": 2.8626,
      "step": 20070
    },
    {
      "epoch": 8.62,
      "learning_rate": 0.0002,
      "loss": 2.8272,
      "step": 20075
    },
    {
      "epoch": 8.62,
      "learning_rate": 0.0002,
      "loss": 2.8198,
      "step": 20080
    },
    {
      "epoch": 8.62,
      "learning_rate": 0.0002,
      "loss": 2.8148,
      "step": 20085
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.0002,
      "loss": 2.8292,
      "step": 20090
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.0002,
      "loss": 2.8567,
      "step": 20095
    },
    {
      "epoch": 8.63,
      "eval_loss": 2.856651544570923,
      "eval_runtime": 5.402,
      "eval_samples_per_second": 1828.381,
      "eval_steps_per_second": 14.439,
      "eval_top3_3_weighted_f1_score ": 0.4282551261947818,
      "eval_top_1_macro_f1_score": 0.0644853415655072,
      "eval_top_1_weighted_f1score": 0.21121483324746365,
      "eval_top_3_macro_f1_score": 0.18130103612404322,
      "step": 20096
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.0002,
      "loss": 2.8408,
      "step": 20100
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.0002,
      "loss": 2.8365,
      "step": 20105
    },
    {
      "epoch": 8.63,
      "learning_rate": 0.0002,
      "loss": 2.8453,
      "step": 20110
    },
    {
      "epoch": 8.64,
      "learning_rate": 0.0002,
      "loss": 2.8165,
      "step": 20115
    },
    {
      "epoch": 8.64,
      "learning_rate": 0.0002,
      "loss": 2.8284,
      "step": 20120
    },
    {
      "epoch": 8.64,
      "learning_rate": 0.0002,
      "loss": 2.829,
      "step": 20125
    },
    {
      "epoch": 8.64,
      "learning_rate": 0.0002,
      "loss": 2.8144,
      "step": 20130
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 20135
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.0002,
      "loss": 2.8309,
      "step": 20140
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.0002,
      "loss": 2.8246,
      "step": 20145
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.0002,
      "loss": 2.8331,
      "step": 20150
    },
    {
      "epoch": 8.65,
      "learning_rate": 0.0002,
      "loss": 2.8482,
      "step": 20155
    },
    {
      "epoch": 8.66,
      "learning_rate": 0.0002,
      "loss": 2.8177,
      "step": 20160
    },
    {
      "epoch": 8.66,
      "eval_loss": 2.8480165004730225,
      "eval_runtime": 5.5042,
      "eval_samples_per_second": 1794.449,
      "eval_steps_per_second": 14.171,
      "eval_top3_3_weighted_f1_score ": 0.43337207017074075,
      "eval_top_1_macro_f1_score": 0.06480975786761277,
      "eval_top_1_weighted_f1score": 0.21026838409829918,
      "eval_top_3_macro_f1_score": 0.17932563774639834,
      "step": 20160
    },
    {
      "epoch": 8.66,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 20165
    },
    {
      "epoch": 8.66,
      "learning_rate": 0.0002,
      "loss": 2.8484,
      "step": 20170
    },
    {
      "epoch": 8.66,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 20175
    },
    {
      "epoch": 8.66,
      "learning_rate": 0.0002,
      "loss": 2.8314,
      "step": 20180
    },
    {
      "epoch": 8.67,
      "learning_rate": 0.0002,
      "loss": 2.8154,
      "step": 20185
    },
    {
      "epoch": 8.67,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 20190
    },
    {
      "epoch": 8.67,
      "learning_rate": 0.0002,
      "loss": 2.8317,
      "step": 20195
    },
    {
      "epoch": 8.67,
      "learning_rate": 0.0002,
      "loss": 2.8338,
      "step": 20200
    },
    {
      "epoch": 8.68,
      "learning_rate": 0.0002,
      "loss": 2.8433,
      "step": 20205
    },
    {
      "epoch": 8.68,
      "learning_rate": 0.0002,
      "loss": 2.8187,
      "step": 20210
    },
    {
      "epoch": 8.68,
      "learning_rate": 0.0002,
      "loss": 2.8124,
      "step": 20215
    },
    {
      "epoch": 8.68,
      "learning_rate": 0.0002,
      "loss": 2.8611,
      "step": 20220
    },
    {
      "epoch": 8.68,
      "eval_loss": 2.8530097007751465,
      "eval_runtime": 5.7222,
      "eval_samples_per_second": 1726.071,
      "eval_steps_per_second": 13.631,
      "eval_top3_3_weighted_f1_score ": 0.43243047676301666,
      "eval_top_1_macro_f1_score": 0.06911414138931672,
      "eval_top_1_weighted_f1score": 0.21065873269725666,
      "eval_top_3_macro_f1_score": 0.18240826040338734,
      "step": 20224
    },
    {
      "epoch": 8.68,
      "learning_rate": 0.0002,
      "loss": 2.825,
      "step": 20225
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.0002,
      "loss": 2.833,
      "step": 20230
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.0002,
      "loss": 2.8455,
      "step": 20235
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 20240
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.0002,
      "loss": 2.8433,
      "step": 20245
    },
    {
      "epoch": 8.69,
      "learning_rate": 0.0002,
      "loss": 2.8191,
      "step": 20250
    },
    {
      "epoch": 8.7,
      "learning_rate": 0.0002,
      "loss": 2.8268,
      "step": 20255
    },
    {
      "epoch": 8.7,
      "learning_rate": 0.0002,
      "loss": 2.8234,
      "step": 20260
    },
    {
      "epoch": 8.7,
      "learning_rate": 0.0002,
      "loss": 2.848,
      "step": 20265
    },
    {
      "epoch": 8.7,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 20270
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0002,
      "loss": 2.8212,
      "step": 20275
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0002,
      "loss": 2.8213,
      "step": 20280
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0002,
      "loss": 2.8327,
      "step": 20285
    },
    {
      "epoch": 8.71,
      "eval_loss": 2.8486971855163574,
      "eval_runtime": 5.5563,
      "eval_samples_per_second": 1777.621,
      "eval_steps_per_second": 14.038,
      "eval_top3_3_weighted_f1_score ": 0.4321731847599356,
      "eval_top_1_macro_f1_score": 0.06616773147903071,
      "eval_top_1_weighted_f1score": 0.20913986319352837,
      "eval_top_3_macro_f1_score": 0.1849872128022307,
      "step": 20288
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0002,
      "loss": 2.83,
      "step": 20290
    },
    {
      "epoch": 8.71,
      "learning_rate": 0.0002,
      "loss": 2.8372,
      "step": 20295
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.0002,
      "loss": 2.8508,
      "step": 20300
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.0002,
      "loss": 2.8209,
      "step": 20305
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.0002,
      "loss": 2.8513,
      "step": 20310
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.0002,
      "loss": 2.8444,
      "step": 20315
    },
    {
      "epoch": 8.72,
      "learning_rate": 0.0002,
      "loss": 2.8394,
      "step": 20320
    },
    {
      "epoch": 8.73,
      "learning_rate": 0.0002,
      "loss": 2.8323,
      "step": 20325
    },
    {
      "epoch": 8.73,
      "learning_rate": 0.0002,
      "loss": 2.8308,
      "step": 20330
    },
    {
      "epoch": 8.73,
      "learning_rate": 0.0002,
      "loss": 2.8421,
      "step": 20335
    },
    {
      "epoch": 8.73,
      "learning_rate": 0.0002,
      "loss": 2.8274,
      "step": 20340
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.0002,
      "loss": 2.83,
      "step": 20345
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.0002,
      "loss": 2.8313,
      "step": 20350
    },
    {
      "epoch": 8.74,
      "eval_loss": 2.8561837673187256,
      "eval_runtime": 5.4564,
      "eval_samples_per_second": 1810.166,
      "eval_steps_per_second": 14.295,
      "eval_top3_3_weighted_f1_score ": 0.4329660646002272,
      "eval_top_1_macro_f1_score": 0.06357845100388941,
      "eval_top_1_weighted_f1score": 0.20728980571997332,
      "eval_top_3_macro_f1_score": 0.1785247873659485,
      "step": 20352
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 20355
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.0002,
      "loss": 2.8215,
      "step": 20360
    },
    {
      "epoch": 8.74,
      "learning_rate": 0.0002,
      "loss": 2.8143,
      "step": 20365
    },
    {
      "epoch": 8.75,
      "learning_rate": 0.0002,
      "loss": 2.8374,
      "step": 20370
    },
    {
      "epoch": 8.75,
      "learning_rate": 0.0002,
      "loss": 2.8293,
      "step": 20375
    },
    {
      "epoch": 8.75,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 20380
    },
    {
      "epoch": 8.75,
      "learning_rate": 0.0002,
      "loss": 2.8355,
      "step": 20385
    },
    {
      "epoch": 8.75,
      "learning_rate": 0.0002,
      "loss": 2.8294,
      "step": 20390
    },
    {
      "epoch": 8.76,
      "learning_rate": 0.0002,
      "loss": 2.8186,
      "step": 20395
    },
    {
      "epoch": 8.76,
      "learning_rate": 0.0002,
      "loss": 2.8395,
      "step": 20400
    },
    {
      "epoch": 8.76,
      "learning_rate": 0.0002,
      "loss": 2.8409,
      "step": 20405
    },
    {
      "epoch": 8.76,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 20410
    },
    {
      "epoch": 8.77,
      "learning_rate": 0.0002,
      "loss": 2.8467,
      "step": 20415
    },
    {
      "epoch": 8.77,
      "eval_loss": 2.853515148162842,
      "eval_runtime": 5.8751,
      "eval_samples_per_second": 1681.162,
      "eval_steps_per_second": 13.276,
      "eval_top3_3_weighted_f1_score ": 0.42716829290360553,
      "eval_top_1_macro_f1_score": 0.062442550388123255,
      "eval_top_1_weighted_f1score": 0.211065276644501,
      "eval_top_3_macro_f1_score": 0.17624991701191145,
      "step": 20416
    },
    {
      "epoch": 8.77,
      "learning_rate": 0.0002,
      "loss": 2.8272,
      "step": 20420
    },
    {
      "epoch": 8.77,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 20425
    },
    {
      "epoch": 8.77,
      "learning_rate": 0.0002,
      "loss": 2.8309,
      "step": 20430
    },
    {
      "epoch": 8.77,
      "learning_rate": 0.0002,
      "loss": 2.8284,
      "step": 20435
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.0002,
      "loss": 2.8232,
      "step": 20440
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.0002,
      "loss": 2.8365,
      "step": 20445
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.0002,
      "loss": 2.8449,
      "step": 20450
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.0002,
      "loss": 2.8339,
      "step": 20455
    },
    {
      "epoch": 8.78,
      "learning_rate": 0.0002,
      "loss": 2.8371,
      "step": 20460
    },
    {
      "epoch": 8.79,
      "learning_rate": 0.0002,
      "loss": 2.8224,
      "step": 20465
    },
    {
      "epoch": 8.79,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 20470
    },
    {
      "epoch": 8.79,
      "learning_rate": 0.0002,
      "loss": 2.8253,
      "step": 20475
    },
    {
      "epoch": 8.79,
      "learning_rate": 0.0002,
      "loss": 2.8317,
      "step": 20480
    },
    {
      "epoch": 8.79,
      "eval_loss": 2.8553853034973145,
      "eval_runtime": 5.8394,
      "eval_samples_per_second": 1691.432,
      "eval_steps_per_second": 13.357,
      "eval_top3_3_weighted_f1_score ": 0.4326537990865509,
      "eval_top_1_macro_f1_score": 0.0681194421241324,
      "eval_top_1_weighted_f1score": 0.20830574540627783,
      "eval_top_3_macro_f1_score": 0.1794247304551217,
      "step": 20480
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.0002,
      "loss": 2.8429,
      "step": 20485
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.0002,
      "loss": 2.8486,
      "step": 20490
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.0002,
      "loss": 2.854,
      "step": 20495
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.0002,
      "loss": 2.8296,
      "step": 20500
    },
    {
      "epoch": 8.8,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 20505
    },
    {
      "epoch": 8.81,
      "learning_rate": 0.0002,
      "loss": 2.8257,
      "step": 20510
    },
    {
      "epoch": 8.81,
      "learning_rate": 0.0002,
      "loss": 2.8182,
      "step": 20515
    },
    {
      "epoch": 8.81,
      "learning_rate": 0.0002,
      "loss": 2.8464,
      "step": 20520
    },
    {
      "epoch": 8.81,
      "learning_rate": 0.0002,
      "loss": 2.8264,
      "step": 20525
    },
    {
      "epoch": 8.81,
      "learning_rate": 0.0002,
      "loss": 2.8184,
      "step": 20530
    },
    {
      "epoch": 8.82,
      "learning_rate": 0.0002,
      "loss": 2.8397,
      "step": 20535
    },
    {
      "epoch": 8.82,
      "learning_rate": 0.0002,
      "loss": 2.8166,
      "step": 20540
    },
    {
      "epoch": 8.82,
      "eval_loss": 2.8540971279144287,
      "eval_runtime": 5.5966,
      "eval_samples_per_second": 1764.818,
      "eval_steps_per_second": 13.937,
      "eval_top3_3_weighted_f1_score ": 0.4336968251093306,
      "eval_top_1_macro_f1_score": 0.06616061236652564,
      "eval_top_1_weighted_f1score": 0.2100926076889365,
      "eval_top_3_macro_f1_score": 0.17929303771621424,
      "step": 20544
    },
    {
      "epoch": 8.82,
      "learning_rate": 0.0002,
      "loss": 2.8528,
      "step": 20545
    },
    {
      "epoch": 8.82,
      "learning_rate": 0.0002,
      "loss": 2.8265,
      "step": 20550
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.0002,
      "loss": 2.8391,
      "step": 20555
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.0002,
      "loss": 2.8103,
      "step": 20560
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 20565
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.0002,
      "loss": 2.8286,
      "step": 20570
    },
    {
      "epoch": 8.83,
      "learning_rate": 0.0002,
      "loss": 2.846,
      "step": 20575
    },
    {
      "epoch": 8.84,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 20580
    },
    {
      "epoch": 8.84,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 20585
    },
    {
      "epoch": 8.84,
      "learning_rate": 0.0002,
      "loss": 2.8161,
      "step": 20590
    },
    {
      "epoch": 8.84,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 20595
    },
    {
      "epoch": 8.84,
      "learning_rate": 0.0002,
      "loss": 2.8411,
      "step": 20600
    },
    {
      "epoch": 8.85,
      "learning_rate": 0.0002,
      "loss": 2.8442,
      "step": 20605
    },
    {
      "epoch": 8.85,
      "eval_loss": 2.8550314903259277,
      "eval_runtime": 6.1337,
      "eval_samples_per_second": 1610.297,
      "eval_steps_per_second": 12.717,
      "eval_top3_3_weighted_f1_score ": 0.4326305080382614,
      "eval_top_1_macro_f1_score": 0.07372309765671434,
      "eval_top_1_weighted_f1score": 0.2139504239199311,
      "eval_top_3_macro_f1_score": 0.1799990482204693,
      "step": 20608
    },
    {
      "epoch": 8.85,
      "learning_rate": 0.0002,
      "loss": 2.8244,
      "step": 20610
    },
    {
      "epoch": 8.85,
      "learning_rate": 0.0002,
      "loss": 2.8222,
      "step": 20615
    },
    {
      "epoch": 8.85,
      "learning_rate": 0.0002,
      "loss": 2.8622,
      "step": 20620
    },
    {
      "epoch": 8.86,
      "learning_rate": 0.0002,
      "loss": 2.8437,
      "step": 20625
    },
    {
      "epoch": 8.86,
      "learning_rate": 0.0002,
      "loss": 2.832,
      "step": 20630
    },
    {
      "epoch": 8.86,
      "learning_rate": 0.0002,
      "loss": 2.8459,
      "step": 20635
    },
    {
      "epoch": 8.86,
      "learning_rate": 0.0002,
      "loss": 2.8278,
      "step": 20640
    },
    {
      "epoch": 8.86,
      "learning_rate": 0.0002,
      "loss": 2.841,
      "step": 20645
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.0002,
      "loss": 2.8278,
      "step": 20650
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.0002,
      "loss": 2.8249,
      "step": 20655
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.0002,
      "loss": 2.832,
      "step": 20660
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.0002,
      "loss": 2.8463,
      "step": 20665
    },
    {
      "epoch": 8.87,
      "learning_rate": 0.0002,
      "loss": 2.8365,
      "step": 20670
    },
    {
      "epoch": 8.88,
      "eval_loss": 2.8521645069122314,
      "eval_runtime": 5.5183,
      "eval_samples_per_second": 1789.869,
      "eval_steps_per_second": 14.135,
      "eval_top3_3_weighted_f1_score ": 0.4288653183056263,
      "eval_top_1_macro_f1_score": 0.06953153388721477,
      "eval_top_1_weighted_f1score": 0.21221184352327754,
      "eval_top_3_macro_f1_score": 0.18016623182443242,
      "step": 20672
    },
    {
      "epoch": 8.88,
      "learning_rate": 0.0002,
      "loss": 2.8273,
      "step": 20675
    },
    {
      "epoch": 8.88,
      "learning_rate": 0.0002,
      "loss": 2.8244,
      "step": 20680
    },
    {
      "epoch": 8.88,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 20685
    },
    {
      "epoch": 8.88,
      "learning_rate": 0.0002,
      "loss": 2.8325,
      "step": 20690
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0002,
      "loss": 2.8501,
      "step": 20695
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0002,
      "loss": 2.8462,
      "step": 20700
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0002,
      "loss": 2.8094,
      "step": 20705
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0002,
      "loss": 2.832,
      "step": 20710
    },
    {
      "epoch": 8.89,
      "learning_rate": 0.0002,
      "loss": 2.8426,
      "step": 20715
    },
    {
      "epoch": 8.9,
      "learning_rate": 0.0002,
      "loss": 2.8392,
      "step": 20720
    },
    {
      "epoch": 8.9,
      "learning_rate": 0.0002,
      "loss": 2.8419,
      "step": 20725
    },
    {
      "epoch": 8.9,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 20730
    },
    {
      "epoch": 8.9,
      "learning_rate": 0.0002,
      "loss": 2.8333,
      "step": 20735
    },
    {
      "epoch": 8.9,
      "eval_loss": 2.8555121421813965,
      "eval_runtime": 5.4467,
      "eval_samples_per_second": 1813.399,
      "eval_steps_per_second": 14.321,
      "eval_top3_3_weighted_f1_score ": 0.43004321131135825,
      "eval_top_1_macro_f1_score": 0.06994283542784271,
      "eval_top_1_weighted_f1score": 0.21361213602172482,
      "eval_top_3_macro_f1_score": 0.17755904404896694,
      "step": 20736
    },
    {
      "epoch": 8.9,
      "learning_rate": 0.0002,
      "loss": 2.8459,
      "step": 20740
    },
    {
      "epoch": 8.91,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 20745
    },
    {
      "epoch": 8.91,
      "learning_rate": 0.0002,
      "loss": 2.8176,
      "step": 20750
    },
    {
      "epoch": 8.91,
      "learning_rate": 0.0002,
      "loss": 2.8389,
      "step": 20755
    },
    {
      "epoch": 8.91,
      "learning_rate": 0.0002,
      "loss": 2.8273,
      "step": 20760
    },
    {
      "epoch": 8.92,
      "learning_rate": 0.0002,
      "loss": 2.8363,
      "step": 20765
    },
    {
      "epoch": 8.92,
      "learning_rate": 0.0002,
      "loss": 2.8606,
      "step": 20770
    },
    {
      "epoch": 8.92,
      "learning_rate": 0.0002,
      "loss": 2.8383,
      "step": 20775
    },
    {
      "epoch": 8.92,
      "learning_rate": 0.0002,
      "loss": 2.8403,
      "step": 20780
    },
    {
      "epoch": 8.92,
      "learning_rate": 0.0002,
      "loss": 2.8345,
      "step": 20785
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.0002,
      "loss": 2.8098,
      "step": 20790
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.0002,
      "loss": 2.841,
      "step": 20795
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.0002,
      "loss": 2.8349,
      "step": 20800
    },
    {
      "epoch": 8.93,
      "eval_loss": 2.8507375717163086,
      "eval_runtime": 5.4642,
      "eval_samples_per_second": 1807.593,
      "eval_steps_per_second": 14.275,
      "eval_top3_3_weighted_f1_score ": 0.4316537377435325,
      "eval_top_1_macro_f1_score": 0.06610656325564745,
      "eval_top_1_weighted_f1score": 0.20977739744610047,
      "eval_top_3_macro_f1_score": 0.18079424819850132,
      "step": 20800
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.0002,
      "loss": 2.8301,
      "step": 20805
    },
    {
      "epoch": 8.93,
      "learning_rate": 0.0002,
      "loss": 2.8341,
      "step": 20810
    },
    {
      "epoch": 8.94,
      "learning_rate": 0.0002,
      "loss": 2.8262,
      "step": 20815
    },
    {
      "epoch": 8.94,
      "learning_rate": 0.0002,
      "loss": 2.8387,
      "step": 20820
    },
    {
      "epoch": 8.94,
      "learning_rate": 0.0002,
      "loss": 2.8353,
      "step": 20825
    },
    {
      "epoch": 8.94,
      "learning_rate": 0.0002,
      "loss": 2.8426,
      "step": 20830
    },
    {
      "epoch": 8.95,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 20835
    },
    {
      "epoch": 8.95,
      "learning_rate": 0.0002,
      "loss": 2.8444,
      "step": 20840
    },
    {
      "epoch": 8.95,
      "learning_rate": 0.0002,
      "loss": 2.8151,
      "step": 20845
    },
    {
      "epoch": 8.95,
      "learning_rate": 0.0002,
      "loss": 2.8364,
      "step": 20850
    },
    {
      "epoch": 8.95,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 20855
    },
    {
      "epoch": 8.96,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 20860
    },
    {
      "epoch": 8.96,
      "eval_loss": 2.847623825073242,
      "eval_runtime": 5.9068,
      "eval_samples_per_second": 1672.139,
      "eval_steps_per_second": 13.205,
      "eval_top3_3_weighted_f1_score ": 0.4303577125006042,
      "eval_top_1_macro_f1_score": 0.06564107838098014,
      "eval_top_1_weighted_f1score": 0.2122386192591822,
      "eval_top_3_macro_f1_score": 0.17716679614969025,
      "step": 20864
    },
    {
      "epoch": 8.96,
      "learning_rate": 0.0002,
      "loss": 2.8349,
      "step": 20865
    },
    {
      "epoch": 8.96,
      "learning_rate": 0.0002,
      "loss": 2.8263,
      "step": 20870
    },
    {
      "epoch": 8.96,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 20875
    },
    {
      "epoch": 8.97,
      "learning_rate": 0.0002,
      "loss": 2.8356,
      "step": 20880
    },
    {
      "epoch": 8.97,
      "learning_rate": 0.0002,
      "loss": 2.836,
      "step": 20885
    },
    {
      "epoch": 8.97,
      "learning_rate": 0.0002,
      "loss": 2.8329,
      "step": 20890
    },
    {
      "epoch": 8.97,
      "learning_rate": 0.0002,
      "loss": 2.842,
      "step": 20895
    },
    {
      "epoch": 8.97,
      "learning_rate": 0.0002,
      "loss": 2.8187,
      "step": 20900
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.0002,
      "loss": 2.8351,
      "step": 20905
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.0002,
      "loss": 2.8364,
      "step": 20910
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.0002,
      "loss": 2.8261,
      "step": 20915
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.0002,
      "loss": 2.8202,
      "step": 20920
    },
    {
      "epoch": 8.98,
      "learning_rate": 0.0002,
      "loss": 2.8255,
      "step": 20925
    },
    {
      "epoch": 8.99,
      "eval_loss": 2.8509836196899414,
      "eval_runtime": 5.4318,
      "eval_samples_per_second": 1818.367,
      "eval_steps_per_second": 14.36,
      "eval_top3_3_weighted_f1_score ": 0.432648431099353,
      "eval_top_1_macro_f1_score": 0.06447750563996495,
      "eval_top_1_weighted_f1score": 0.21153075665082782,
      "eval_top_3_macro_f1_score": 0.17601003529362458,
      "step": 20928
    },
    {
      "epoch": 8.99,
      "learning_rate": 0.0002,
      "loss": 2.8201,
      "step": 20930
    },
    {
      "epoch": 8.99,
      "learning_rate": 0.0002,
      "loss": 2.8235,
      "step": 20935
    },
    {
      "epoch": 8.99,
      "learning_rate": 0.0002,
      "loss": 2.8373,
      "step": 20940
    },
    {
      "epoch": 8.99,
      "learning_rate": 0.0002,
      "loss": 2.8281,
      "step": 20945
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 20950
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.0002,
      "loss": 2.8454,
      "step": 20955
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.0002,
      "loss": 2.8238,
      "step": 20960
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.0002,
      "loss": 3.0734,
      "step": 20965
    },
    {
      "epoch": 9.0,
      "learning_rate": 0.0002,
      "loss": 2.8411,
      "step": 20970
    },
    {
      "epoch": 9.01,
      "learning_rate": 0.0002,
      "loss": 2.8194,
      "step": 20975
    },
    {
      "epoch": 9.01,
      "learning_rate": 0.0002,
      "loss": 2.8416,
      "step": 20980
    },
    {
      "epoch": 9.01,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 20985
    },
    {
      "epoch": 9.01,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 20990
    },
    {
      "epoch": 9.01,
      "eval_loss": 2.8559987545013428,
      "eval_runtime": 5.5014,
      "eval_samples_per_second": 1795.368,
      "eval_steps_per_second": 14.178,
      "eval_top3_3_weighted_f1_score ": 0.4289481852333947,
      "eval_top_1_macro_f1_score": 0.06149344404550756,
      "eval_top_1_weighted_f1score": 0.20998515238494447,
      "eval_top_3_macro_f1_score": 0.17319624072739703,
      "step": 20992
    },
    {
      "epoch": 9.01,
      "learning_rate": 0.0002,
      "loss": 2.8448,
      "step": 20995
    },
    {
      "epoch": 9.02,
      "learning_rate": 0.0002,
      "loss": 2.8188,
      "step": 21000
    },
    {
      "epoch": 9.02,
      "learning_rate": 0.0002,
      "loss": 2.8309,
      "step": 21005
    },
    {
      "epoch": 9.02,
      "learning_rate": 0.0002,
      "loss": 2.8224,
      "step": 21010
    },
    {
      "epoch": 9.02,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 21015
    },
    {
      "epoch": 9.03,
      "learning_rate": 0.0002,
      "loss": 2.8143,
      "step": 21020
    },
    {
      "epoch": 9.03,
      "learning_rate": 0.0002,
      "loss": 2.8152,
      "step": 21025
    },
    {
      "epoch": 9.03,
      "learning_rate": 0.0002,
      "loss": 2.8295,
      "step": 21030
    },
    {
      "epoch": 9.03,
      "learning_rate": 0.0002,
      "loss": 2.83,
      "step": 21035
    },
    {
      "epoch": 9.03,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 21040
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.0002,
      "loss": 2.837,
      "step": 21045
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.0002,
      "loss": 2.8118,
      "step": 21050
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.0002,
      "loss": 2.8202,
      "step": 21055
    },
    {
      "epoch": 9.04,
      "eval_loss": 2.854757785797119,
      "eval_runtime": 10.7174,
      "eval_samples_per_second": 921.582,
      "eval_steps_per_second": 7.278,
      "eval_top3_3_weighted_f1_score ": 0.4290146721185138,
      "eval_top_1_macro_f1_score": 0.0650658667462059,
      "eval_top_1_weighted_f1score": 0.21017897554909437,
      "eval_top_3_macro_f1_score": 0.1747978277239454,
      "step": 21056
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.0002,
      "loss": 2.8215,
      "step": 21060
    },
    {
      "epoch": 9.04,
      "learning_rate": 0.0002,
      "loss": 2.8141,
      "step": 21065
    },
    {
      "epoch": 9.05,
      "learning_rate": 0.0002,
      "loss": 2.852,
      "step": 21070
    },
    {
      "epoch": 9.05,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 21075
    },
    {
      "epoch": 9.05,
      "learning_rate": 0.0002,
      "loss": 2.8165,
      "step": 21080
    },
    {
      "epoch": 9.05,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 21085
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.0002,
      "loss": 2.8475,
      "step": 21090
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.0002,
      "loss": 2.8225,
      "step": 21095
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.0002,
      "loss": 2.8358,
      "step": 21100
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.0002,
      "loss": 2.8176,
      "step": 21105
    },
    {
      "epoch": 9.06,
      "learning_rate": 0.0002,
      "loss": 2.8112,
      "step": 21110
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.0002,
      "loss": 2.8296,
      "step": 21115
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 21120
    },
    {
      "epoch": 9.07,
      "eval_loss": 2.8495287895202637,
      "eval_runtime": 5.4299,
      "eval_samples_per_second": 1819.009,
      "eval_steps_per_second": 14.365,
      "eval_top3_3_weighted_f1_score ": 0.4308293286894479,
      "eval_top_1_macro_f1_score": 0.06681364780074087,
      "eval_top_1_weighted_f1score": 0.21217147307987427,
      "eval_top_3_macro_f1_score": 0.17689882794775566,
      "step": 21120
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 21125
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.0002,
      "loss": 2.8252,
      "step": 21130
    },
    {
      "epoch": 9.07,
      "learning_rate": 0.0002,
      "loss": 2.8181,
      "step": 21135
    },
    {
      "epoch": 9.08,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 21140
    },
    {
      "epoch": 9.08,
      "learning_rate": 0.0002,
      "loss": 2.8365,
      "step": 21145
    },
    {
      "epoch": 9.08,
      "learning_rate": 0.0002,
      "loss": 2.8105,
      "step": 21150
    },
    {
      "epoch": 9.08,
      "learning_rate": 0.0002,
      "loss": 2.8363,
      "step": 21155
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.0002,
      "loss": 2.8243,
      "step": 21160
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.0002,
      "loss": 2.8268,
      "step": 21165
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.0002,
      "loss": 2.8319,
      "step": 21170
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.0002,
      "loss": 2.8289,
      "step": 21175
    },
    {
      "epoch": 9.09,
      "learning_rate": 0.0002,
      "loss": 2.8203,
      "step": 21180
    },
    {
      "epoch": 9.1,
      "eval_loss": 2.8540923595428467,
      "eval_runtime": 6.1385,
      "eval_samples_per_second": 1609.032,
      "eval_steps_per_second": 12.707,
      "eval_top3_3_weighted_f1_score ": 0.43406388949070634,
      "eval_top_1_macro_f1_score": 0.06571266282943601,
      "eval_top_1_weighted_f1score": 0.2099046559634731,
      "eval_top_3_macro_f1_score": 0.18371348679155983,
      "step": 21184
    },
    {
      "epoch": 9.1,
      "learning_rate": 0.0002,
      "loss": 2.8121,
      "step": 21185
    },
    {
      "epoch": 9.1,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 21190
    },
    {
      "epoch": 9.1,
      "learning_rate": 0.0002,
      "loss": 2.8026,
      "step": 21195
    },
    {
      "epoch": 9.1,
      "learning_rate": 0.0002,
      "loss": 2.8086,
      "step": 21200
    },
    {
      "epoch": 9.1,
      "learning_rate": 0.0002,
      "loss": 2.825,
      "step": 21205
    },
    {
      "epoch": 9.11,
      "learning_rate": 0.0002,
      "loss": 2.8041,
      "step": 21210
    },
    {
      "epoch": 9.11,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 21215
    },
    {
      "epoch": 9.11,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 21220
    },
    {
      "epoch": 9.11,
      "learning_rate": 0.0002,
      "loss": 2.8166,
      "step": 21225
    },
    {
      "epoch": 9.12,
      "learning_rate": 0.0002,
      "loss": 2.8193,
      "step": 21230
    },
    {
      "epoch": 9.12,
      "learning_rate": 0.0002,
      "loss": 2.8049,
      "step": 21235
    },
    {
      "epoch": 9.12,
      "learning_rate": 0.0002,
      "loss": 2.8227,
      "step": 21240
    },
    {
      "epoch": 9.12,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 21245
    },
    {
      "epoch": 9.12,
      "eval_loss": 2.8501055240631104,
      "eval_runtime": 5.5695,
      "eval_samples_per_second": 1773.395,
      "eval_steps_per_second": 14.005,
      "eval_top3_3_weighted_f1_score ": 0.43584761181094744,
      "eval_top_1_macro_f1_score": 0.06445290422447014,
      "eval_top_1_weighted_f1score": 0.2088455615308982,
      "eval_top_3_macro_f1_score": 0.1864230245826315,
      "step": 21248
    },
    {
      "epoch": 9.12,
      "learning_rate": 0.0002,
      "loss": 2.8164,
      "step": 21250
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.0002,
      "loss": 2.819,
      "step": 21255
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.0002,
      "loss": 2.8414,
      "step": 21260
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.0002,
      "loss": 2.8397,
      "step": 21265
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.0002,
      "loss": 2.827,
      "step": 21270
    },
    {
      "epoch": 9.13,
      "learning_rate": 0.0002,
      "loss": 2.8136,
      "step": 21275
    },
    {
      "epoch": 9.14,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 21280
    },
    {
      "epoch": 9.14,
      "learning_rate": 0.0002,
      "loss": 2.8243,
      "step": 21285
    },
    {
      "epoch": 9.14,
      "learning_rate": 0.0002,
      "loss": 2.8259,
      "step": 21290
    },
    {
      "epoch": 9.14,
      "learning_rate": 0.0002,
      "loss": 2.8153,
      "step": 21295
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.0002,
      "loss": 2.8321,
      "step": 21300
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.0002,
      "loss": 2.8507,
      "step": 21305
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.0002,
      "loss": 2.8251,
      "step": 21310
    },
    {
      "epoch": 9.15,
      "eval_loss": 2.851583242416382,
      "eval_runtime": 5.5855,
      "eval_samples_per_second": 1768.321,
      "eval_steps_per_second": 13.965,
      "eval_top3_3_weighted_f1_score ": 0.43293469928532674,
      "eval_top_1_macro_f1_score": 0.06847001353805812,
      "eval_top_1_weighted_f1score": 0.21222443480614808,
      "eval_top_3_macro_f1_score": 0.18320235700441048,
      "step": 21312
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.0002,
      "loss": 2.8166,
      "step": 21315
    },
    {
      "epoch": 9.15,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 21320
    },
    {
      "epoch": 9.16,
      "learning_rate": 0.0002,
      "loss": 2.8238,
      "step": 21325
    },
    {
      "epoch": 9.16,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 21330
    },
    {
      "epoch": 9.16,
      "learning_rate": 0.0002,
      "loss": 2.8357,
      "step": 21335
    },
    {
      "epoch": 9.16,
      "learning_rate": 0.0002,
      "loss": 2.8219,
      "step": 21340
    },
    {
      "epoch": 9.16,
      "learning_rate": 0.0002,
      "loss": 2.8116,
      "step": 21345
    },
    {
      "epoch": 9.17,
      "learning_rate": 0.0002,
      "loss": 2.8126,
      "step": 21350
    },
    {
      "epoch": 9.17,
      "learning_rate": 0.0002,
      "loss": 2.8103,
      "step": 21355
    },
    {
      "epoch": 9.17,
      "learning_rate": 0.0002,
      "loss": 2.8098,
      "step": 21360
    },
    {
      "epoch": 9.17,
      "learning_rate": 0.0002,
      "loss": 2.8175,
      "step": 21365
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.0002,
      "loss": 2.8548,
      "step": 21370
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.0002,
      "loss": 2.8343,
      "step": 21375
    },
    {
      "epoch": 9.18,
      "eval_loss": 2.8495962619781494,
      "eval_runtime": 25.4192,
      "eval_samples_per_second": 388.565,
      "eval_steps_per_second": 3.069,
      "eval_top3_3_weighted_f1_score ": 0.4305444508763087,
      "eval_top_1_macro_f1_score": 0.061936007639464566,
      "eval_top_1_weighted_f1score": 0.20609522102782227,
      "eval_top_3_macro_f1_score": 0.17489965966499793,
      "step": 21376
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 21380
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.0002,
      "loss": 2.821,
      "step": 21385
    },
    {
      "epoch": 9.18,
      "learning_rate": 0.0002,
      "loss": 2.8326,
      "step": 21390
    },
    {
      "epoch": 9.19,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 21395
    },
    {
      "epoch": 9.19,
      "learning_rate": 0.0002,
      "loss": 2.8187,
      "step": 21400
    },
    {
      "epoch": 9.19,
      "learning_rate": 0.0002,
      "loss": 2.8061,
      "step": 21405
    },
    {
      "epoch": 9.19,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 21410
    },
    {
      "epoch": 9.19,
      "learning_rate": 0.0002,
      "loss": 2.8005,
      "step": 21415
    },
    {
      "epoch": 9.2,
      "learning_rate": 0.0002,
      "loss": 2.825,
      "step": 21420
    },
    {
      "epoch": 9.2,
      "learning_rate": 0.0002,
      "loss": 2.8152,
      "step": 21425
    },
    {
      "epoch": 9.2,
      "learning_rate": 0.0002,
      "loss": 2.8234,
      "step": 21430
    },
    {
      "epoch": 9.2,
      "learning_rate": 0.0002,
      "loss": 2.8286,
      "step": 21435
    },
    {
      "epoch": 9.21,
      "learning_rate": 0.0002,
      "loss": 2.8279,
      "step": 21440
    },
    {
      "epoch": 9.21,
      "eval_loss": 2.85322642326355,
      "eval_runtime": 5.5819,
      "eval_samples_per_second": 1769.456,
      "eval_steps_per_second": 13.974,
      "eval_top3_3_weighted_f1_score ": 0.43037269133521056,
      "eval_top_1_macro_f1_score": 0.06291767301115053,
      "eval_top_1_weighted_f1score": 0.2086298451004576,
      "eval_top_3_macro_f1_score": 0.17855019495783017,
      "step": 21440
    },
    {
      "epoch": 9.21,
      "learning_rate": 0.0002,
      "loss": 2.8342,
      "step": 21445
    },
    {
      "epoch": 9.21,
      "learning_rate": 0.0002,
      "loss": 2.8205,
      "step": 21450
    },
    {
      "epoch": 9.21,
      "learning_rate": 0.0002,
      "loss": 2.8328,
      "step": 21455
    },
    {
      "epoch": 9.21,
      "learning_rate": 0.0002,
      "loss": 2.8341,
      "step": 21460
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.0002,
      "loss": 2.8249,
      "step": 21465
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 21470
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.0002,
      "loss": 2.8461,
      "step": 21475
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.0002,
      "loss": 2.8158,
      "step": 21480
    },
    {
      "epoch": 9.22,
      "learning_rate": 0.0002,
      "loss": 2.8276,
      "step": 21485
    },
    {
      "epoch": 9.23,
      "learning_rate": 0.0002,
      "loss": 2.8324,
      "step": 21490
    },
    {
      "epoch": 9.23,
      "learning_rate": 0.0002,
      "loss": 2.8111,
      "step": 21495
    },
    {
      "epoch": 9.23,
      "learning_rate": 0.0002,
      "loss": 2.8185,
      "step": 21500
    },
    {
      "epoch": 9.23,
      "eval_loss": 2.8582258224487305,
      "eval_runtime": 5.912,
      "eval_samples_per_second": 1670.667,
      "eval_steps_per_second": 13.193,
      "eval_top3_3_weighted_f1_score ": 0.4274626782562602,
      "eval_top_1_macro_f1_score": 0.06607128955554938,
      "eval_top_1_weighted_f1score": 0.21184014436558643,
      "eval_top_3_macro_f1_score": 0.18036024703073375,
      "step": 21504
    },
    {
      "epoch": 9.23,
      "learning_rate": 0.0002,
      "loss": 2.8181,
      "step": 21505
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 21510
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.0002,
      "loss": 2.8098,
      "step": 21515
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.0002,
      "loss": 2.8165,
      "step": 21520
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.0002,
      "loss": 2.8232,
      "step": 21525
    },
    {
      "epoch": 9.24,
      "learning_rate": 0.0002,
      "loss": 2.8163,
      "step": 21530
    },
    {
      "epoch": 9.25,
      "learning_rate": 0.0002,
      "loss": 2.8181,
      "step": 21535
    },
    {
      "epoch": 9.25,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 21540
    },
    {
      "epoch": 9.25,
      "learning_rate": 0.0002,
      "loss": 2.8473,
      "step": 21545
    },
    {
      "epoch": 9.25,
      "learning_rate": 0.0002,
      "loss": 2.8175,
      "step": 21550
    },
    {
      "epoch": 9.25,
      "learning_rate": 0.0002,
      "loss": 2.827,
      "step": 21555
    },
    {
      "epoch": 9.26,
      "learning_rate": 0.0002,
      "loss": 2.814,
      "step": 21560
    },
    {
      "epoch": 9.26,
      "learning_rate": 0.0002,
      "loss": 2.8258,
      "step": 21565
    },
    {
      "epoch": 9.26,
      "eval_loss": 2.8505871295928955,
      "eval_runtime": 5.4807,
      "eval_samples_per_second": 1802.15,
      "eval_steps_per_second": 14.232,
      "eval_top3_3_weighted_f1_score ": 0.43250106565599944,
      "eval_top_1_macro_f1_score": 0.06029549034961428,
      "eval_top_1_weighted_f1score": 0.2089473526350904,
      "eval_top_3_macro_f1_score": 0.17339420729037616,
      "step": 21568
    },
    {
      "epoch": 9.26,
      "learning_rate": 0.0002,
      "loss": 2.8296,
      "step": 21570
    },
    {
      "epoch": 9.26,
      "learning_rate": 0.0002,
      "loss": 2.8234,
      "step": 21575
    },
    {
      "epoch": 9.27,
      "learning_rate": 0.0002,
      "loss": 2.8388,
      "step": 21580
    },
    {
      "epoch": 9.27,
      "learning_rate": 0.0002,
      "loss": 2.8288,
      "step": 21585
    },
    {
      "epoch": 9.27,
      "learning_rate": 0.0002,
      "loss": 2.8039,
      "step": 21590
    },
    {
      "epoch": 9.27,
      "learning_rate": 0.0002,
      "loss": 2.8232,
      "step": 21595
    },
    {
      "epoch": 9.27,
      "learning_rate": 0.0002,
      "loss": 2.8336,
      "step": 21600
    },
    {
      "epoch": 9.28,
      "learning_rate": 0.0002,
      "loss": 2.8202,
      "step": 21605
    },
    {
      "epoch": 9.28,
      "learning_rate": 0.0002,
      "loss": 2.8094,
      "step": 21610
    },
    {
      "epoch": 9.28,
      "learning_rate": 0.0002,
      "loss": 2.8303,
      "step": 21615
    },
    {
      "epoch": 9.28,
      "learning_rate": 0.0002,
      "loss": 2.8379,
      "step": 21620
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.0002,
      "loss": 2.8256,
      "step": 21625
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.0002,
      "loss": 2.8461,
      "step": 21630
    },
    {
      "epoch": 9.29,
      "eval_loss": 2.851837396621704,
      "eval_runtime": 5.5414,
      "eval_samples_per_second": 1782.391,
      "eval_steps_per_second": 14.076,
      "eval_top3_3_weighted_f1_score ": 0.43157979362043053,
      "eval_top_1_macro_f1_score": 0.0714168071148363,
      "eval_top_1_weighted_f1score": 0.21458997690049192,
      "eval_top_3_macro_f1_score": 0.17546914981613587,
      "step": 21632
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.0002,
      "loss": 2.847,
      "step": 21635
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.0002,
      "loss": 2.8119,
      "step": 21640
    },
    {
      "epoch": 9.29,
      "learning_rate": 0.0002,
      "loss": 2.8436,
      "step": 21645
    },
    {
      "epoch": 9.3,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 21650
    },
    {
      "epoch": 9.3,
      "learning_rate": 0.0002,
      "loss": 2.8357,
      "step": 21655
    },
    {
      "epoch": 9.3,
      "learning_rate": 0.0002,
      "loss": 2.845,
      "step": 21660
    },
    {
      "epoch": 9.3,
      "learning_rate": 0.0002,
      "loss": 2.8565,
      "step": 21665
    },
    {
      "epoch": 9.3,
      "learning_rate": 0.0002,
      "loss": 2.8261,
      "step": 21670
    },
    {
      "epoch": 9.31,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 21675
    },
    {
      "epoch": 9.31,
      "learning_rate": 0.0002,
      "loss": 2.83,
      "step": 21680
    },
    {
      "epoch": 9.31,
      "learning_rate": 0.0002,
      "loss": 2.8472,
      "step": 21685
    },
    {
      "epoch": 9.31,
      "learning_rate": 0.0002,
      "loss": 2.8485,
      "step": 21690
    },
    {
      "epoch": 9.32,
      "learning_rate": 0.0002,
      "loss": 2.8218,
      "step": 21695
    },
    {
      "epoch": 9.32,
      "eval_loss": 2.847691535949707,
      "eval_runtime": 5.5922,
      "eval_samples_per_second": 1766.197,
      "eval_steps_per_second": 13.948,
      "eval_top3_3_weighted_f1_score ": 0.43518061084283816,
      "eval_top_1_macro_f1_score": 0.06834749791042832,
      "eval_top_1_weighted_f1score": 0.21213859890679648,
      "eval_top_3_macro_f1_score": 0.17734608601908802,
      "step": 21696
    },
    {
      "epoch": 9.32,
      "learning_rate": 0.0002,
      "loss": 2.8276,
      "step": 21700
    },
    {
      "epoch": 9.32,
      "learning_rate": 0.0002,
      "loss": 2.8248,
      "step": 21705
    },
    {
      "epoch": 9.32,
      "learning_rate": 0.0002,
      "loss": 2.8303,
      "step": 21710
    },
    {
      "epoch": 9.32,
      "learning_rate": 0.0002,
      "loss": 2.8197,
      "step": 21715
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.0002,
      "loss": 2.7987,
      "step": 21720
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.0002,
      "loss": 2.8109,
      "step": 21725
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.0002,
      "loss": 2.8255,
      "step": 21730
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.0002,
      "loss": 2.8302,
      "step": 21735
    },
    {
      "epoch": 9.33,
      "learning_rate": 0.0002,
      "loss": 2.813,
      "step": 21740
    },
    {
      "epoch": 9.34,
      "learning_rate": 0.0002,
      "loss": 2.8251,
      "step": 21745
    },
    {
      "epoch": 9.34,
      "learning_rate": 0.0002,
      "loss": 2.8225,
      "step": 21750
    },
    {
      "epoch": 9.34,
      "learning_rate": 0.0002,
      "loss": 2.8303,
      "step": 21755
    },
    {
      "epoch": 9.34,
      "learning_rate": 0.0002,
      "loss": 2.8168,
      "step": 21760
    },
    {
      "epoch": 9.34,
      "eval_loss": 2.8492608070373535,
      "eval_runtime": 5.5571,
      "eval_samples_per_second": 1777.38,
      "eval_steps_per_second": 14.036,
      "eval_top3_3_weighted_f1_score ": 0.43120071411956107,
      "eval_top_1_macro_f1_score": 0.06795811905068046,
      "eval_top_1_weighted_f1score": 0.20868870727325284,
      "eval_top_3_macro_f1_score": 0.17405456349897896,
      "step": 21760
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 21765
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.0002,
      "loss": 2.8502,
      "step": 21770
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 21775
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.0002,
      "loss": 2.8266,
      "step": 21780
    },
    {
      "epoch": 9.35,
      "learning_rate": 0.0002,
      "loss": 2.8167,
      "step": 21785
    },
    {
      "epoch": 9.36,
      "learning_rate": 0.0002,
      "loss": 2.8502,
      "step": 21790
    },
    {
      "epoch": 9.36,
      "learning_rate": 0.0002,
      "loss": 2.8251,
      "step": 21795
    },
    {
      "epoch": 9.36,
      "learning_rate": 0.0002,
      "loss": 2.8313,
      "step": 21800
    },
    {
      "epoch": 9.36,
      "learning_rate": 0.0002,
      "loss": 2.8291,
      "step": 21805
    },
    {
      "epoch": 9.36,
      "learning_rate": 0.0002,
      "loss": 2.829,
      "step": 21810
    },
    {
      "epoch": 9.37,
      "learning_rate": 0.0002,
      "loss": 2.8318,
      "step": 21815
    },
    {
      "epoch": 9.37,
      "learning_rate": 0.0002,
      "loss": 2.808,
      "step": 21820
    },
    {
      "epoch": 9.37,
      "eval_loss": 2.8524250984191895,
      "eval_runtime": 7.1815,
      "eval_samples_per_second": 1375.334,
      "eval_steps_per_second": 10.861,
      "eval_top3_3_weighted_f1_score ": 0.430441569957755,
      "eval_top_1_macro_f1_score": 0.07308010153184245,
      "eval_top_1_weighted_f1score": 0.21382525358508753,
      "eval_top_3_macro_f1_score": 0.17174174769716297,
      "step": 21824
    },
    {
      "epoch": 9.37,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 21825
    },
    {
      "epoch": 9.37,
      "learning_rate": 0.0002,
      "loss": 2.8196,
      "step": 21830
    },
    {
      "epoch": 9.38,
      "learning_rate": 0.0002,
      "loss": 2.831,
      "step": 21835
    },
    {
      "epoch": 9.38,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 21840
    },
    {
      "epoch": 9.38,
      "learning_rate": 0.0002,
      "loss": 2.8416,
      "step": 21845
    },
    {
      "epoch": 9.38,
      "learning_rate": 0.0002,
      "loss": 2.8243,
      "step": 21850
    },
    {
      "epoch": 9.38,
      "learning_rate": 0.0002,
      "loss": 2.8218,
      "step": 21855
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.0002,
      "loss": 2.8279,
      "step": 21860
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.0002,
      "loss": 2.8244,
      "step": 21865
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.0002,
      "loss": 2.8265,
      "step": 21870
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.0002,
      "loss": 2.827,
      "step": 21875
    },
    {
      "epoch": 9.39,
      "learning_rate": 0.0002,
      "loss": 2.8126,
      "step": 21880
    },
    {
      "epoch": 9.4,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 21885
    },
    {
      "epoch": 9.4,
      "eval_loss": 2.852505683898926,
      "eval_runtime": 5.4688,
      "eval_samples_per_second": 1806.055,
      "eval_steps_per_second": 14.263,
      "eval_top3_3_weighted_f1_score ": 0.4300055491500529,
      "eval_top_1_macro_f1_score": 0.07128116741764888,
      "eval_top_1_weighted_f1score": 0.21187664243099974,
      "eval_top_3_macro_f1_score": 0.1793112644216981,
      "step": 21888
    },
    {
      "epoch": 9.4,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 21890
    },
    {
      "epoch": 9.4,
      "learning_rate": 0.0002,
      "loss": 2.8257,
      "step": 21895
    },
    {
      "epoch": 9.4,
      "learning_rate": 0.0002,
      "loss": 2.8259,
      "step": 21900
    },
    {
      "epoch": 9.41,
      "learning_rate": 0.0002,
      "loss": 2.82,
      "step": 21905
    },
    {
      "epoch": 9.41,
      "learning_rate": 0.0002,
      "loss": 2.835,
      "step": 21910
    },
    {
      "epoch": 9.41,
      "learning_rate": 0.0002,
      "loss": 2.8246,
      "step": 21915
    },
    {
      "epoch": 9.41,
      "learning_rate": 0.0002,
      "loss": 2.8155,
      "step": 21920
    },
    {
      "epoch": 9.41,
      "learning_rate": 0.0002,
      "loss": 2.8231,
      "step": 21925
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 21930
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.0002,
      "loss": 2.8179,
      "step": 21935
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.0002,
      "loss": 2.8314,
      "step": 21940
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.0002,
      "loss": 2.8223,
      "step": 21945
    },
    {
      "epoch": 9.42,
      "learning_rate": 0.0002,
      "loss": 2.8283,
      "step": 21950
    },
    {
      "epoch": 9.43,
      "eval_loss": 2.856452465057373,
      "eval_runtime": 5.957,
      "eval_samples_per_second": 1658.052,
      "eval_steps_per_second": 13.094,
      "eval_top3_3_weighted_f1_score ": 0.4333480847886804,
      "eval_top_1_macro_f1_score": 0.07014983787727506,
      "eval_top_1_weighted_f1score": 0.2144048069881346,
      "eval_top_3_macro_f1_score": 0.18613721082481452,
      "step": 21952
    },
    {
      "epoch": 9.43,
      "learning_rate": 0.0002,
      "loss": 2.8348,
      "step": 21955
    },
    {
      "epoch": 9.43,
      "learning_rate": 0.0002,
      "loss": 2.8178,
      "step": 21960
    },
    {
      "epoch": 9.43,
      "learning_rate": 0.0002,
      "loss": 2.8175,
      "step": 21965
    },
    {
      "epoch": 9.43,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 21970
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.0002,
      "loss": 2.8158,
      "step": 21975
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.0002,
      "loss": 2.8273,
      "step": 21980
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.0002,
      "loss": 2.8177,
      "step": 21985
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.0002,
      "loss": 2.8303,
      "step": 21990
    },
    {
      "epoch": 9.44,
      "learning_rate": 0.0002,
      "loss": 2.8278,
      "step": 21995
    },
    {
      "epoch": 9.45,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 22000
    },
    {
      "epoch": 9.45,
      "learning_rate": 0.0002,
      "loss": 2.8302,
      "step": 22005
    },
    {
      "epoch": 9.45,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 22010
    },
    {
      "epoch": 9.45,
      "learning_rate": 0.0002,
      "loss": 2.8265,
      "step": 22015
    },
    {
      "epoch": 9.45,
      "eval_loss": 2.8534743785858154,
      "eval_runtime": 162.204,
      "eval_samples_per_second": 60.892,
      "eval_steps_per_second": 0.481,
      "eval_top3_3_weighted_f1_score ": 0.43166440461259736,
      "eval_top_1_macro_f1_score": 0.06529782059485596,
      "eval_top_1_weighted_f1score": 0.20582071646299285,
      "eval_top_3_macro_f1_score": 0.17866792501545486,
      "step": 22016
    },
    {
      "epoch": 9.45,
      "learning_rate": 0.0002,
      "loss": 2.8452,
      "step": 22020
    },
    {
      "epoch": 9.46,
      "learning_rate": 0.0002,
      "loss": 2.83,
      "step": 22025
    },
    {
      "epoch": 9.46,
      "learning_rate": 0.0002,
      "loss": 2.8341,
      "step": 22030
    },
    {
      "epoch": 9.46,
      "learning_rate": 0.0002,
      "loss": 2.823,
      "step": 22035
    },
    {
      "epoch": 9.46,
      "learning_rate": 0.0002,
      "loss": 2.8163,
      "step": 22040
    },
    {
      "epoch": 9.47,
      "learning_rate": 0.0002,
      "loss": 2.8293,
      "step": 22045
    },
    {
      "epoch": 9.47,
      "learning_rate": 0.0002,
      "loss": 2.8245,
      "step": 22050
    },
    {
      "epoch": 9.47,
      "learning_rate": 0.0002,
      "loss": 2.8175,
      "step": 22055
    },
    {
      "epoch": 9.47,
      "learning_rate": 0.0002,
      "loss": 2.8455,
      "step": 22060
    },
    {
      "epoch": 9.47,
      "learning_rate": 0.0002,
      "loss": 2.8299,
      "step": 22065
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.0002,
      "loss": 2.8204,
      "step": 22070
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.0002,
      "loss": 2.8138,
      "step": 22075
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.0002,
      "loss": 2.8309,
      "step": 22080
    },
    {
      "epoch": 9.48,
      "eval_loss": 2.85726261138916,
      "eval_runtime": 5.3407,
      "eval_samples_per_second": 1849.38,
      "eval_steps_per_second": 14.605,
      "eval_top3_3_weighted_f1_score ": 0.4286655454078363,
      "eval_top_1_macro_f1_score": 0.06709038385470456,
      "eval_top_1_weighted_f1score": 0.21081449479089667,
      "eval_top_3_macro_f1_score": 0.1730030726530344,
      "step": 22080
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 22085
    },
    {
      "epoch": 9.48,
      "learning_rate": 0.0002,
      "loss": 2.8352,
      "step": 22090
    },
    {
      "epoch": 9.49,
      "learning_rate": 0.0002,
      "loss": 2.8116,
      "step": 22095
    },
    {
      "epoch": 9.49,
      "learning_rate": 0.0002,
      "loss": 2.849,
      "step": 22100
    },
    {
      "epoch": 9.49,
      "learning_rate": 0.0002,
      "loss": 2.8063,
      "step": 22105
    },
    {
      "epoch": 9.49,
      "learning_rate": 0.0002,
      "loss": 2.8206,
      "step": 22110
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.0002,
      "loss": 2.8206,
      "step": 22115
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.0002,
      "loss": 2.8322,
      "step": 22120
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.0002,
      "loss": 2.8226,
      "step": 22125
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.0002,
      "loss": 2.8469,
      "step": 22130
    },
    {
      "epoch": 9.5,
      "learning_rate": 0.0002,
      "loss": 2.8194,
      "step": 22135
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.0002,
      "loss": 2.8375,
      "step": 22140
    },
    {
      "epoch": 9.51,
      "eval_loss": 2.8557488918304443,
      "eval_runtime": 5.4862,
      "eval_samples_per_second": 1800.328,
      "eval_steps_per_second": 14.217,
      "eval_top3_3_weighted_f1_score ": 0.4268182211630273,
      "eval_top_1_macro_f1_score": 0.06563339841606039,
      "eval_top_1_weighted_f1score": 0.2131272297426529,
      "eval_top_3_macro_f1_score": 0.17459416251668647,
      "step": 22144
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.0002,
      "loss": 2.8423,
      "step": 22145
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.0002,
      "loss": 2.8401,
      "step": 22150
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.0002,
      "loss": 2.8376,
      "step": 22155
    },
    {
      "epoch": 9.51,
      "learning_rate": 0.0002,
      "loss": 2.8359,
      "step": 22160
    },
    {
      "epoch": 9.52,
      "learning_rate": 0.0002,
      "loss": 2.8281,
      "step": 22165
    },
    {
      "epoch": 9.52,
      "learning_rate": 0.0002,
      "loss": 2.8337,
      "step": 22170
    },
    {
      "epoch": 9.52,
      "learning_rate": 0.0002,
      "loss": 2.8145,
      "step": 22175
    },
    {
      "epoch": 9.52,
      "learning_rate": 0.0002,
      "loss": 2.833,
      "step": 22180
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.0002,
      "loss": 2.8276,
      "step": 22185
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.0002,
      "loss": 2.8406,
      "step": 22190
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.0002,
      "loss": 2.8246,
      "step": 22195
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.0002,
      "loss": 2.8288,
      "step": 22200
    },
    {
      "epoch": 9.53,
      "learning_rate": 0.0002,
      "loss": 2.8213,
      "step": 22205
    },
    {
      "epoch": 9.54,
      "eval_loss": 2.8494086265563965,
      "eval_runtime": 5.2945,
      "eval_samples_per_second": 1865.532,
      "eval_steps_per_second": 14.732,
      "eval_top3_3_weighted_f1_score ": 0.4315943580943734,
      "eval_top_1_macro_f1_score": 0.07372850905722614,
      "eval_top_1_weighted_f1score": 0.21281821345479093,
      "eval_top_3_macro_f1_score": 0.18265632089833317,
      "step": 22208
    },
    {
      "epoch": 9.54,
      "learning_rate": 0.0002,
      "loss": 2.8358,
      "step": 22210
    },
    {
      "epoch": 9.54,
      "learning_rate": 0.0002,
      "loss": 2.8284,
      "step": 22215
    },
    {
      "epoch": 9.54,
      "learning_rate": 0.0002,
      "loss": 2.8315,
      "step": 22220
    },
    {
      "epoch": 9.54,
      "learning_rate": 0.0002,
      "loss": 2.8489,
      "step": 22225
    },
    {
      "epoch": 9.54,
      "learning_rate": 0.0002,
      "loss": 2.8261,
      "step": 22230
    },
    {
      "epoch": 9.55,
      "learning_rate": 0.0002,
      "loss": 2.8295,
      "step": 22235
    },
    {
      "epoch": 9.55,
      "learning_rate": 0.0002,
      "loss": 2.8134,
      "step": 22240
    },
    {
      "epoch": 9.55,
      "learning_rate": 0.0002,
      "loss": 2.8074,
      "step": 22245
    },
    {
      "epoch": 9.55,
      "learning_rate": 0.0002,
      "loss": 2.8203,
      "step": 22250
    },
    {
      "epoch": 9.56,
      "learning_rate": 0.0002,
      "loss": 2.8459,
      "step": 22255
    },
    {
      "epoch": 9.56,
      "learning_rate": 0.0002,
      "loss": 2.8342,
      "step": 22260
    },
    {
      "epoch": 9.56,
      "learning_rate": 0.0002,
      "loss": 2.8209,
      "step": 22265
    },
    {
      "epoch": 9.56,
      "learning_rate": 0.0002,
      "loss": 2.8238,
      "step": 22270
    },
    {
      "epoch": 9.56,
      "eval_loss": 2.8474643230438232,
      "eval_runtime": 5.4026,
      "eval_samples_per_second": 1828.2,
      "eval_steps_per_second": 14.438,
      "eval_top3_3_weighted_f1_score ": 0.43126150796226115,
      "eval_top_1_macro_f1_score": 0.07045744025919749,
      "eval_top_1_weighted_f1score": 0.21364646849437402,
      "eval_top_3_macro_f1_score": 0.1791263113839099,
      "step": 22272
    },
    {
      "epoch": 9.56,
      "learning_rate": 0.0002,
      "loss": 2.8111,
      "step": 22275
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.0002,
      "loss": 2.8235,
      "step": 22280
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.0002,
      "loss": 2.8053,
      "step": 22285
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.0002,
      "loss": 2.8362,
      "step": 22290
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.0002,
      "loss": 2.8404,
      "step": 22295
    },
    {
      "epoch": 9.57,
      "learning_rate": 0.0002,
      "loss": 2.8111,
      "step": 22300
    },
    {
      "epoch": 9.58,
      "learning_rate": 0.0002,
      "loss": 2.8196,
      "step": 22305
    },
    {
      "epoch": 9.58,
      "learning_rate": 0.0002,
      "loss": 2.8273,
      "step": 22310
    },
    {
      "epoch": 9.58,
      "learning_rate": 0.0002,
      "loss": 2.8171,
      "step": 22315
    },
    {
      "epoch": 9.58,
      "learning_rate": 0.0002,
      "loss": 2.8298,
      "step": 22320
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.0002,
      "loss": 2.8231,
      "step": 22325
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.0002,
      "loss": 2.8299,
      "step": 22330
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.0002,
      "loss": 2.8276,
      "step": 22335
    },
    {
      "epoch": 9.59,
      "eval_loss": 2.847917318344116,
      "eval_runtime": 5.3857,
      "eval_samples_per_second": 1833.93,
      "eval_steps_per_second": 14.483,
      "eval_top3_3_weighted_f1_score ": 0.436237582005185,
      "eval_top_1_macro_f1_score": 0.06547436603574881,
      "eval_top_1_weighted_f1score": 0.2072242509479481,
      "eval_top_3_macro_f1_score": 0.17854312296656763,
      "step": 22336
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.0002,
      "loss": 2.8275,
      "step": 22340
    },
    {
      "epoch": 9.59,
      "learning_rate": 0.0002,
      "loss": 2.8365,
      "step": 22345
    },
    {
      "epoch": 9.6,
      "learning_rate": 0.0002,
      "loss": 2.838,
      "step": 22350
    },
    {
      "epoch": 9.6,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 22355
    },
    {
      "epoch": 9.6,
      "learning_rate": 0.0002,
      "loss": 2.828,
      "step": 22360
    },
    {
      "epoch": 9.6,
      "learning_rate": 0.0002,
      "loss": 2.8228,
      "step": 22365
    },
    {
      "epoch": 9.6,
      "learning_rate": 0.0002,
      "loss": 2.8146,
      "step": 22370
    },
    {
      "epoch": 9.61,
      "learning_rate": 0.0002,
      "loss": 2.8268,
      "step": 22375
    },
    {
      "epoch": 9.61,
      "learning_rate": 0.0002,
      "loss": 2.8265,
      "step": 22380
    },
    {
      "epoch": 9.61,
      "learning_rate": 0.0002,
      "loss": 2.827,
      "step": 22385
    },
    {
      "epoch": 9.61,
      "learning_rate": 0.0002,
      "loss": 2.8183,
      "step": 22390
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.0002,
      "loss": 2.8385,
      "step": 22395
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.0002,
      "loss": 2.8141,
      "step": 22400
    },
    {
      "epoch": 9.62,
      "eval_loss": 2.852114677429199,
      "eval_runtime": 5.6974,
      "eval_samples_per_second": 1733.609,
      "eval_steps_per_second": 13.691,
      "eval_top3_3_weighted_f1_score ": 0.4325883484046382,
      "eval_top_1_macro_f1_score": 0.07310835748462423,
      "eval_top_1_weighted_f1score": 0.20887421005502937,
      "eval_top_3_macro_f1_score": 0.18169750082095731,
      "step": 22400
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.0002,
      "loss": 2.8212,
      "step": 22405
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.0002,
      "loss": 2.8337,
      "step": 22410
    },
    {
      "epoch": 9.62,
      "learning_rate": 0.0002,
      "loss": 2.8198,
      "step": 22415
    },
    {
      "epoch": 9.63,
      "learning_rate": 0.0002,
      "loss": 2.82,
      "step": 22420
    },
    {
      "epoch": 9.63,
      "learning_rate": 0.0002,
      "loss": 2.834,
      "step": 22425
    },
    {
      "epoch": 9.63,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 22430
    },
    {
      "epoch": 9.63,
      "learning_rate": 0.0002,
      "loss": 2.8419,
      "step": 22435
    },
    {
      "epoch": 9.63,
      "learning_rate": 0.0002,
      "loss": 2.8332,
      "step": 22440
    },
    {
      "epoch": 9.64,
      "learning_rate": 0.0002,
      "loss": 2.8348,
      "step": 22445
    },
    {
      "epoch": 9.64,
      "learning_rate": 0.0002,
      "loss": 2.8367,
      "step": 22450
    },
    {
      "epoch": 9.64,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 22455
    },
    {
      "epoch": 9.64,
      "learning_rate": 0.0002,
      "loss": 2.8091,
      "step": 22460
    },
    {
      "epoch": 9.65,
      "eval_loss": 2.853534460067749,
      "eval_runtime": 5.3887,
      "eval_samples_per_second": 1832.895,
      "eval_steps_per_second": 14.475,
      "eval_top3_3_weighted_f1_score ": 0.4305856628206878,
      "eval_top_1_macro_f1_score": 0.06318847528797358,
      "eval_top_1_weighted_f1score": 0.20984362953446536,
      "eval_top_3_macro_f1_score": 0.17533045615218756,
      "step": 22464
    },
    {
      "epoch": 9.65,
      "learning_rate": 0.0002,
      "loss": 2.8239,
      "step": 22465
    },
    {
      "epoch": 9.65,
      "learning_rate": 0.0002,
      "loss": 2.8335,
      "step": 22470
    },
    {
      "epoch": 9.65,
      "learning_rate": 0.0002,
      "loss": 2.8246,
      "step": 22475
    },
    {
      "epoch": 9.65,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 22480
    },
    {
      "epoch": 9.65,
      "learning_rate": 0.0002,
      "loss": 2.8542,
      "step": 22485
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.0002,
      "loss": 2.821,
      "step": 22490
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.0002,
      "loss": 2.8174,
      "step": 22495
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.0002,
      "loss": 2.8198,
      "step": 22500
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.0002,
      "loss": 2.8363,
      "step": 22505
    },
    {
      "epoch": 9.66,
      "learning_rate": 0.0002,
      "loss": 2.8401,
      "step": 22510
    },
    {
      "epoch": 9.67,
      "learning_rate": 0.0002,
      "loss": 2.8382,
      "step": 22515
    },
    {
      "epoch": 9.67,
      "learning_rate": 0.0002,
      "loss": 2.8427,
      "step": 22520
    },
    {
      "epoch": 9.67,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 22525
    },
    {
      "epoch": 9.67,
      "eval_loss": 2.851630449295044,
      "eval_runtime": 5.437,
      "eval_samples_per_second": 1816.626,
      "eval_steps_per_second": 14.346,
      "eval_top3_3_weighted_f1_score ": 0.43217284214512847,
      "eval_top_1_macro_f1_score": 0.0669065728068694,
      "eval_top_1_weighted_f1score": 0.20909471808589813,
      "eval_top_3_macro_f1_score": 0.18281573863659614,
      "step": 22528
    },
    {
      "epoch": 9.67,
      "learning_rate": 0.0002,
      "loss": 2.8307,
      "step": 22530
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.0002,
      "loss": 2.826,
      "step": 22535
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.0002,
      "loss": 2.8305,
      "step": 22540
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.0002,
      "loss": 2.8193,
      "step": 22545
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.0002,
      "loss": 2.8253,
      "step": 22550
    },
    {
      "epoch": 9.68,
      "learning_rate": 0.0002,
      "loss": 2.8231,
      "step": 22555
    },
    {
      "epoch": 9.69,
      "learning_rate": 0.0002,
      "loss": 2.8477,
      "step": 22560
    },
    {
      "epoch": 9.69,
      "learning_rate": 0.0002,
      "loss": 2.8311,
      "step": 22565
    },
    {
      "epoch": 9.69,
      "learning_rate": 0.0002,
      "loss": 2.8283,
      "step": 22570
    },
    {
      "epoch": 9.69,
      "learning_rate": 0.0002,
      "loss": 2.8125,
      "step": 22575
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.0002,
      "loss": 2.8378,
      "step": 22580
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.0002,
      "loss": 2.8241,
      "step": 22585
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.0002,
      "loss": 2.8404,
      "step": 22590
    },
    {
      "epoch": 9.7,
      "eval_loss": 2.8550713062286377,
      "eval_runtime": 5.7527,
      "eval_samples_per_second": 1716.929,
      "eval_steps_per_second": 13.559,
      "eval_top3_3_weighted_f1_score ": 0.4319429275812067,
      "eval_top_1_macro_f1_score": 0.0680740631387194,
      "eval_top_1_weighted_f1score": 0.21071631271646857,
      "eval_top_3_macro_f1_score": 0.1813096685504854,
      "step": 22592
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.0002,
      "loss": 2.8184,
      "step": 22595
    },
    {
      "epoch": 9.7,
      "learning_rate": 0.0002,
      "loss": 2.8108,
      "step": 22600
    },
    {
      "epoch": 9.71,
      "learning_rate": 0.0002,
      "loss": 2.8173,
      "step": 22605
    },
    {
      "epoch": 9.71,
      "learning_rate": 0.0002,
      "loss": 2.84,
      "step": 22610
    },
    {
      "epoch": 9.71,
      "learning_rate": 0.0002,
      "loss": 2.8089,
      "step": 22615
    },
    {
      "epoch": 9.71,
      "learning_rate": 0.0002,
      "loss": 2.8204,
      "step": 22620
    },
    {
      "epoch": 9.71,
      "learning_rate": 0.0002,
      "loss": 2.8181,
      "step": 22625
    },
    {
      "epoch": 9.72,
      "learning_rate": 0.0002,
      "loss": 2.8546,
      "step": 22630
    },
    {
      "epoch": 9.72,
      "learning_rate": 0.0002,
      "loss": 2.8349,
      "step": 22635
    },
    {
      "epoch": 9.72,
      "learning_rate": 0.0002,
      "loss": 2.8365,
      "step": 22640
    },
    {
      "epoch": 9.72,
      "learning_rate": 0.0002,
      "loss": 2.8004,
      "step": 22645
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.0002,
      "loss": 2.8291,
      "step": 22650
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.0002,
      "loss": 2.8167,
      "step": 22655
    },
    {
      "epoch": 9.73,
      "eval_loss": 2.847604990005493,
      "eval_runtime": 5.3717,
      "eval_samples_per_second": 1838.701,
      "eval_steps_per_second": 14.52,
      "eval_top3_3_weighted_f1_score ": 0.4296377629392317,
      "eval_top_1_macro_f1_score": 0.0751315435133377,
      "eval_top_1_weighted_f1score": 0.21394146061309774,
      "eval_top_3_macro_f1_score": 0.1780066666602597,
      "step": 22656
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.0002,
      "loss": 2.8369,
      "step": 22660
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.0002,
      "loss": 2.8226,
      "step": 22665
    },
    {
      "epoch": 9.73,
      "learning_rate": 0.0002,
      "loss": 2.819,
      "step": 22670
    },
    {
      "epoch": 9.74,
      "learning_rate": 0.0002,
      "loss": 2.8188,
      "step": 22675
    },
    {
      "epoch": 9.74,
      "learning_rate": 0.0002,
      "loss": 2.8105,
      "step": 22680
    },
    {
      "epoch": 9.74,
      "learning_rate": 0.0002,
      "loss": 2.8384,
      "step": 22685
    },
    {
      "epoch": 9.74,
      "learning_rate": 0.0002,
      "loss": 2.8155,
      "step": 22690
    },
    {
      "epoch": 9.74,
      "learning_rate": 0.0002,
      "loss": 2.8317,
      "step": 22695
    },
    {
      "epoch": 9.75,
      "learning_rate": 0.0002,
      "loss": 2.8236,
      "step": 22700
    },
    {
      "epoch": 9.75,
      "learning_rate": 0.0002,
      "loss": 2.8264,
      "step": 22705
    },
    {
      "epoch": 9.75,
      "learning_rate": 0.0002,
      "loss": 2.823,
      "step": 22710
    },
    {
      "epoch": 9.75,
      "learning_rate": 0.0002,
      "loss": 2.8361,
      "step": 22715
    },
    {
      "epoch": 9.76,
      "learning_rate": 0.0002,
      "loss": 2.8274,
      "step": 22720
    },
    {
      "epoch": 9.76,
      "eval_loss": 2.847738742828369,
      "eval_runtime": 5.4086,
      "eval_samples_per_second": 1826.165,
      "eval_steps_per_second": 14.421,
      "eval_top3_3_weighted_f1_score ": 0.4323342268436191,
      "eval_top_1_macro_f1_score": 0.06781506604704406,
      "eval_top_1_weighted_f1score": 0.2109029706071375,
      "eval_top_3_macro_f1_score": 0.17692047586267784,
      "step": 22720
    },
    {
      "epoch": 9.76,
      "learning_rate": 0.0002,
      "loss": 2.8207,
      "step": 22725
    },
    {
      "epoch": 9.76,
      "learning_rate": 0.0002,
      "loss": 2.8415,
      "step": 22730
    },
    {
      "epoch": 9.76,
      "learning_rate": 0.0002,
      "loss": 2.821,
      "step": 22735
    },
    {
      "epoch": 9.76,
      "learning_rate": 0.0002,
      "loss": 2.8306,
      "step": 22740
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.0002,
      "loss": 2.829,
      "step": 22745
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.0002,
      "loss": 2.8446,
      "step": 22750
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.0002,
      "loss": 2.8467,
      "step": 22755
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.0002,
      "loss": 2.8425,
      "step": 22760
    },
    {
      "epoch": 9.77,
      "learning_rate": 0.0002,
      "loss": 2.8356,
      "step": 22765
    },
    {
      "epoch": 9.78,
      "learning_rate": 0.0002,
      "loss": 2.8316,
      "step": 22770
    },
    {
      "epoch": 9.78,
      "learning_rate": 0.0002,
      "loss": 2.8151,
      "step": 22775
    },
    {
      "epoch": 9.78,
      "learning_rate": 0.0002,
      "loss": 2.8337,
      "step": 22780
    },
    {
      "epoch": 9.78,
      "eval_loss": 2.8500022888183594,
      "eval_runtime": 5.4145,
      "eval_samples_per_second": 1824.184,
      "eval_steps_per_second": 14.406,
      "eval_top3_3_weighted_f1_score ": 0.43283902112300215,
      "eval_top_1_macro_f1_score": 0.06821030623830515,
      "eval_top_1_weighted_f1score": 0.21041662920364143,
      "eval_top_3_macro_f1_score": 0.1764320095292666,
      "step": 22784
    },
    {
      "epoch": 9.78,
      "learning_rate": 0.0002,
      "loss": 2.8228,
      "step": 22785
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.0002,
      "loss": 2.8269,
      "step": 22790
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.0002,
      "loss": 2.8381,
      "step": 22795
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.0002,
      "loss": 2.8279,
      "step": 22800
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.0002,
      "loss": 2.8441,
      "step": 22805
    },
    {
      "epoch": 9.79,
      "learning_rate": 0.0002,
      "loss": 2.8308,
      "step": 22810
    },
    {
      "epoch": 9.8,
      "learning_rate": 0.0002,
      "loss": 2.8261,
      "step": 22815
    },
    {
      "epoch": 9.8,
      "learning_rate": 0.0002,
      "loss": 2.8219,
      "step": 22820
    },
    {
      "epoch": 9.8,
      "learning_rate": 0.0002,
      "loss": 2.8216,
      "step": 22825
    },
    {
      "epoch": 9.8,
      "learning_rate": 0.0002,
      "loss": 2.8363,
      "step": 22830
    },
    {
      "epoch": 9.8,
      "learning_rate": 0.0002,
      "loss": 2.8465,
      "step": 22835
    },
    {
      "epoch": 9.81,
      "learning_rate": 0.0002,
      "loss": 2.8268,
      "step": 22840
    },
    {
      "epoch": 9.81,
      "learning_rate": 0.0002,
      "loss": 2.8296,
      "step": 22845
    },
    {
      "epoch": 9.81,
      "eval_loss": 2.8474063873291016,
      "eval_runtime": 5.4254,
      "eval_samples_per_second": 1820.525,
      "eval_steps_per_second": 14.377,
      "eval_top3_3_weighted_f1_score ": 0.43618021259645307,
      "eval_top_1_macro_f1_score": 0.06937867083371499,
      "eval_top_1_weighted_f1score": 0.20832137094757502,
      "eval_top_3_macro_f1_score": 0.18881116431984588,
      "step": 22848
    }
  ],
  "max_steps": 23290,
  "num_train_epochs": 10,
  "total_flos": 5.64324213028251e+16,
  "trial_name": null,
  "trial_params": null
}