{
  "best_metric": 0.8808290155440415,
  "best_model_checkpoint": "./vit-base-skin/checkpoint-1252",
  "epoch": 6.0,
  "global_step": 3756,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.02,
      "learning_rate": 0.0001994675186368477,
      "loss": 1.3226,
      "step": 10
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019893503727369543,
      "loss": 1.181,
      "step": 20
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019840255591054313,
      "loss": 0.8442,
      "step": 30
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019787007454739086,
      "loss": 1.0347,
      "step": 40
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00019733759318423856,
      "loss": 0.83,
      "step": 50
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00019680511182108628,
      "loss": 0.8187,
      "step": 60
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00019627263045793398,
      "loss": 1.0101,
      "step": 70
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00019574014909478168,
      "loss": 0.8966,
      "step": 80
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001952076677316294,
      "loss": 0.9757,
      "step": 90
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001946751863684771,
      "loss": 0.8881,
      "step": 100
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00019414270500532483,
      "loss": 0.849,
      "step": 110
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00019361022364217253,
      "loss": 0.9514,
      "step": 120
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00019307774227902025,
      "loss": 0.7534,
      "step": 130
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00019254526091586795,
      "loss": 0.6676,
      "step": 140
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00019201277955271565,
      "loss": 0.8426,
      "step": 150
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00019148029818956338,
      "loss": 0.7935,
      "step": 160
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00019094781682641108,
      "loss": 0.6444,
      "step": 170
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001904153354632588,
      "loss": 0.7758,
      "step": 180
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001898828541001065,
      "loss": 0.6214,
      "step": 190
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00018935037273695422,
      "loss": 0.7693,
      "step": 200
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00018881789137380192,
      "loss": 0.7717,
      "step": 210
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00018828541001064962,
      "loss": 0.6923,
      "step": 220
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00018775292864749735,
      "loss": 0.7113,
      "step": 230
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00018722044728434505,
      "loss": 0.6749,
      "step": 240
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00018668796592119277,
      "loss": 0.5459,
      "step": 250
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00018615548455804047,
      "loss": 0.738,
      "step": 260
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001856230031948882,
      "loss": 0.8384,
      "step": 270
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001850905218317359,
      "loss": 0.7732,
      "step": 280
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001845580404685836,
      "loss": 0.8118,
      "step": 290
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00018402555910543132,
      "loss": 0.6305,
      "step": 300
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00018349307774227902,
      "loss": 0.6226,
      "step": 310
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00018296059637912674,
      "loss": 0.577,
      "step": 320
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.00018242811501597444,
      "loss": 0.6114,
      "step": 330
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.00018189563365282217,
      "loss": 0.8104,
      "step": 340
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.00018136315228966987,
      "loss": 0.6855,
      "step": 350
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.00018083067092651756,
      "loss": 0.6844,
      "step": 360
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0001802981895633653,
      "loss": 0.7086,
      "step": 370
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.00017976570820021302,
      "loss": 0.6734,
      "step": 380
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.00017923322683706071,
      "loss": 0.5737,
      "step": 390
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0001787007454739084,
      "loss": 0.7821,
      "step": 400
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.00017816826411075614,
      "loss": 0.6612,
      "step": 410
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.00017763578274760384,
      "loss": 0.6838,
      "step": 420
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.00017710330138445154,
      "loss": 0.7775,
      "step": 430
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.00017657082002129926,
      "loss": 0.6847,
      "step": 440
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.000176038338658147,
      "loss": 0.5728,
      "step": 450
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.00017550585729499469,
      "loss": 0.5409,
      "step": 460
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0001749733759318424,
      "loss": 0.6694,
      "step": 470
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0001744408945686901,
      "loss": 0.5375,
      "step": 480
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0001739084132055378,
      "loss": 0.7531,
      "step": 490
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0001733759318423855,
      "loss": 0.6331,
      "step": 500
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.00017284345047923323,
      "loss": 0.5812,
      "step": 510
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.00017231096911608096,
      "loss": 0.5885,
      "step": 520
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.00017177848775292866,
      "loss": 0.5008,
      "step": 530
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.00017124600638977638,
      "loss": 0.6969,
      "step": 540
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.00017071352502662408,
      "loss": 0.6664,
      "step": 550
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.00017018104366347178,
      "loss": 0.623,
      "step": 560
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.00016964856230031948,
      "loss": 0.5874,
      "step": 570
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0001691160809371672,
      "loss": 0.4945,
      "step": 580
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.00016858359957401493,
      "loss": 0.5377,
      "step": 590
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.00016805111821086263,
      "loss": 0.561,
      "step": 600
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.00016751863684771035,
      "loss": 0.6611,
      "step": 610
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.00016698615548455805,
      "loss": 0.4952,
      "step": 620
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8082901554404145,
      "eval_f1": 0.8151715440575547,
      "eval_loss": 0.551965057849884,
      "eval_precision": 0.8906195272775583,
      "eval_recall": 0.8082901554404145,
      "eval_runtime": 1.2231,
      "eval_samples_per_second": 157.791,
      "eval_steps_per_second": 10.628,
      "step": 626
    },
    {
      "epoch": 1.01,
      "learning_rate": 0.00016645367412140575,
      "loss": 0.535,
      "step": 630
    },
    {
      "epoch": 1.02,
      "learning_rate": 0.00016592119275825345,
      "loss": 0.3981,
      "step": 640
    },
    {
      "epoch": 1.04,
      "learning_rate": 0.00016538871139510118,
      "loss": 0.4819,
      "step": 650
    },
    {
      "epoch": 1.05,
      "learning_rate": 0.0001648562300319489,
      "loss": 0.5802,
      "step": 660
    },
    {
      "epoch": 1.07,
      "learning_rate": 0.0001643237486687966,
      "loss": 0.5392,
      "step": 670
    },
    {
      "epoch": 1.09,
      "learning_rate": 0.00016379126730564433,
      "loss": 0.5117,
      "step": 680
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.00016325878594249202,
      "loss": 0.4329,
      "step": 690
    },
    {
      "epoch": 1.12,
      "learning_rate": 0.00016272630457933972,
      "loss": 0.4065,
      "step": 700
    },
    {
      "epoch": 1.13,
      "learning_rate": 0.00016219382321618745,
      "loss": 0.595,
      "step": 710
    },
    {
      "epoch": 1.15,
      "learning_rate": 0.00016166134185303515,
      "loss": 0.4661,
      "step": 720
    },
    {
      "epoch": 1.17,
      "learning_rate": 0.00016112886048988287,
      "loss": 0.4824,
      "step": 730
    },
    {
      "epoch": 1.18,
      "learning_rate": 0.00016059637912673057,
      "loss": 0.4392,
      "step": 740
    },
    {
      "epoch": 1.2,
      "learning_rate": 0.0001600638977635783,
      "loss": 0.3931,
      "step": 750
    },
    {
      "epoch": 1.21,
      "learning_rate": 0.00015958466453674122,
      "loss": 0.5221,
      "step": 760
    },
    {
      "epoch": 1.23,
      "learning_rate": 0.00015905218317358895,
      "loss": 0.3718,
      "step": 770
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.00015851970181043665,
      "loss": 0.5278,
      "step": 780
    },
    {
      "epoch": 1.26,
      "learning_rate": 0.00015798722044728435,
      "loss": 0.4808,
      "step": 790
    },
    {
      "epoch": 1.28,
      "learning_rate": 0.00015750798722044727,
      "loss": 0.6643,
      "step": 800
    },
    {
      "epoch": 1.29,
      "learning_rate": 0.00015697550585729502,
      "loss": 0.6016,
      "step": 810
    },
    {
      "epoch": 1.31,
      "learning_rate": 0.00015644302449414272,
      "loss": 0.5561,
      "step": 820
    },
    {
      "epoch": 1.33,
      "learning_rate": 0.00015591054313099042,
      "loss": 0.3496,
      "step": 830
    },
    {
      "epoch": 1.34,
      "learning_rate": 0.00015537806176783815,
      "loss": 0.4903,
      "step": 840
    },
    {
      "epoch": 1.36,
      "learning_rate": 0.00015484558040468585,
      "loss": 0.4163,
      "step": 850
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.00015431309904153354,
      "loss": 0.6074,
      "step": 860
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.00015378061767838124,
      "loss": 0.6076,
      "step": 870
    },
    {
      "epoch": 1.41,
      "learning_rate": 0.000153248136315229,
      "loss": 0.3271,
      "step": 880
    },
    {
      "epoch": 1.42,
      "learning_rate": 0.0001527156549520767,
      "loss": 0.526,
      "step": 890
    },
    {
      "epoch": 1.44,
      "learning_rate": 0.0001521831735889244,
      "loss": 0.4802,
      "step": 900
    },
    {
      "epoch": 1.45,
      "learning_rate": 0.00015165069222577212,
      "loss": 0.4022,
      "step": 910
    },
    {
      "epoch": 1.47,
      "learning_rate": 0.00015117145899893504,
      "loss": 0.4919,
      "step": 920
    },
    {
      "epoch": 1.49,
      "learning_rate": 0.00015063897763578277,
      "loss": 0.4165,
      "step": 930
    },
    {
      "epoch": 1.5,
      "learning_rate": 0.00015010649627263047,
      "loss": 0.5528,
      "step": 940
    },
    {
      "epoch": 1.52,
      "learning_rate": 0.00014957401490947817,
      "loss": 0.5046,
      "step": 950
    },
    {
      "epoch": 1.53,
      "learning_rate": 0.0001490415335463259,
      "loss": 0.4592,
      "step": 960
    },
    {
      "epoch": 1.55,
      "learning_rate": 0.0001485090521831736,
      "loss": 0.4861,
      "step": 970
    },
    {
      "epoch": 1.57,
      "learning_rate": 0.0001479765708200213,
      "loss": 0.4433,
      "step": 980
    },
    {
      "epoch": 1.58,
      "learning_rate": 0.00014744408945686902,
      "loss": 0.4874,
      "step": 990
    },
    {
      "epoch": 1.6,
      "learning_rate": 0.00014691160809371674,
      "loss": 0.4691,
      "step": 1000
    },
    {
      "epoch": 1.61,
      "learning_rate": 0.00014637912673056444,
      "loss": 0.4094,
      "step": 1010
    },
    {
      "epoch": 1.63,
      "learning_rate": 0.00014584664536741214,
      "loss": 0.4793,
      "step": 1020
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.00014531416400425986,
      "loss": 0.423,
      "step": 1030
    },
    {
      "epoch": 1.66,
      "learning_rate": 0.00014478168264110756,
      "loss": 0.5418,
      "step": 1040
    },
    {
      "epoch": 1.68,
      "learning_rate": 0.00014424920127795526,
      "loss": 0.4584,
      "step": 1050
    },
    {
      "epoch": 1.69,
      "learning_rate": 0.000143716719914803,
      "loss": 0.4168,
      "step": 1060
    },
    {
      "epoch": 1.71,
      "learning_rate": 0.0001431842385516507,
      "loss": 0.4214,
      "step": 1070
    },
    {
      "epoch": 1.73,
      "learning_rate": 0.0001426517571884984,
      "loss": 0.4917,
      "step": 1080
    },
    {
      "epoch": 1.74,
      "learning_rate": 0.0001421192758253461,
      "loss": 0.4052,
      "step": 1090
    },
    {
      "epoch": 1.76,
      "learning_rate": 0.00014158679446219384,
      "loss": 0.463,
      "step": 1100
    },
    {
      "epoch": 1.77,
      "learning_rate": 0.00014105431309904153,
      "loss": 0.589,
      "step": 1110
    },
    {
      "epoch": 1.79,
      "learning_rate": 0.00014052183173588923,
      "loss": 0.3621,
      "step": 1120
    },
    {
      "epoch": 1.81,
      "learning_rate": 0.00013998935037273696,
      "loss": 0.2815,
      "step": 1130
    },
    {
      "epoch": 1.82,
      "learning_rate": 0.00013945686900958468,
      "loss": 0.3033,
      "step": 1140
    },
    {
      "epoch": 1.84,
      "learning_rate": 0.00013892438764643238,
      "loss": 0.5261,
      "step": 1150
    },
    {
      "epoch": 1.85,
      "learning_rate": 0.00013839190628328008,
      "loss": 0.3978,
      "step": 1160
    },
    {
      "epoch": 1.87,
      "learning_rate": 0.0001378594249201278,
      "loss": 0.3969,
      "step": 1170
    },
    {
      "epoch": 1.88,
      "learning_rate": 0.0001373269435569755,
      "loss": 0.6126,
      "step": 1180
    },
    {
      "epoch": 1.9,
      "learning_rate": 0.0001367944621938232,
      "loss": 0.262,
      "step": 1190
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.00013626198083067093,
      "loss": 0.4194,
      "step": 1200
    },
    {
      "epoch": 1.93,
      "learning_rate": 0.00013572949946751866,
      "loss": 0.4764,
      "step": 1210
    },
    {
      "epoch": 1.95,
      "learning_rate": 0.00013519701810436635,
      "loss": 0.2494,
      "step": 1220
    },
    {
      "epoch": 1.96,
      "learning_rate": 0.00013466453674121405,
      "loss": 0.3119,
      "step": 1230
    },
    {
      "epoch": 1.98,
      "learning_rate": 0.00013413205537806178,
      "loss": 0.4482,
      "step": 1240
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.00013359957401490948,
      "loss": 0.406,
      "step": 1250
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8808290155440415,
      "eval_f1": 0.8739851867067574,
      "eval_loss": 0.44047027826309204,
      "eval_precision": 0.8807327145363345,
      "eval_recall": 0.8808290155440415,
      "eval_runtime": 1.2577,
      "eval_samples_per_second": 153.459,
      "eval_steps_per_second": 10.337,
      "step": 1252
    },
    {
      "epoch": 2.01,
      "learning_rate": 0.00013306709265175718,
      "loss": 0.359,
      "step": 1260
    },
    {
      "epoch": 2.03,
      "learning_rate": 0.00013253461128860493,
      "loss": 0.2068,
      "step": 1270
    },
    {
      "epoch": 2.04,
      "learning_rate": 0.00013200212992545263,
      "loss": 0.2841,
      "step": 1280
    },
    {
      "epoch": 2.06,
      "learning_rate": 0.00013146964856230033,
      "loss": 0.2243,
      "step": 1290
    },
    {
      "epoch": 2.08,
      "learning_rate": 0.00013093716719914802,
      "loss": 0.2247,
      "step": 1300
    },
    {
      "epoch": 2.09,
      "learning_rate": 0.00013040468583599575,
      "loss": 0.2546,
      "step": 1310
    },
    {
      "epoch": 2.11,
      "learning_rate": 0.00012987220447284345,
      "loss": 0.3118,
      "step": 1320
    },
    {
      "epoch": 2.12,
      "learning_rate": 0.00012933972310969115,
      "loss": 0.2522,
      "step": 1330
    },
    {
      "epoch": 2.14,
      "learning_rate": 0.0001288072417465389,
      "loss": 0.323,
      "step": 1340
    },
    {
      "epoch": 2.16,
      "learning_rate": 0.0001282747603833866,
      "loss": 0.2501,
      "step": 1350
    },
    {
      "epoch": 2.17,
      "learning_rate": 0.0001277422790202343,
      "loss": 0.2335,
      "step": 1360
    },
    {
      "epoch": 2.19,
      "learning_rate": 0.000127209797657082,
      "loss": 0.188,
      "step": 1370
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.00012667731629392972,
      "loss": 0.3222,
      "step": 1380
    },
    {
      "epoch": 2.22,
      "learning_rate": 0.00012614483493077742,
      "loss": 0.3993,
      "step": 1390
    },
    {
      "epoch": 2.24,
      "learning_rate": 0.00012561235356762512,
      "loss": 0.287,
      "step": 1400
    },
    {
      "epoch": 2.25,
      "learning_rate": 0.00012507987220447287,
      "loss": 0.2513,
      "step": 1410
    },
    {
      "epoch": 2.27,
      "learning_rate": 0.00012454739084132057,
      "loss": 0.3005,
      "step": 1420
    },
    {
      "epoch": 2.28,
      "learning_rate": 0.00012401490947816827,
      "loss": 0.1648,
      "step": 1430
    },
    {
      "epoch": 2.3,
      "learning_rate": 0.00012348242811501597,
      "loss": 0.2993,
      "step": 1440
    },
    {
      "epoch": 2.32,
      "learning_rate": 0.0001229499467518637,
      "loss": 0.3046,
      "step": 1450
    },
    {
      "epoch": 2.33,
      "learning_rate": 0.0001224174653887114,
      "loss": 0.2163,
      "step": 1460
    },
    {
      "epoch": 2.35,
      "learning_rate": 0.0001218849840255591,
      "loss": 0.2368,
      "step": 1470
    },
    {
      "epoch": 2.36,
      "learning_rate": 0.00012135250266240683,
      "loss": 0.4212,
      "step": 1480
    },
    {
      "epoch": 2.38,
      "learning_rate": 0.00012082002129925454,
      "loss": 0.2242,
      "step": 1490
    },
    {
      "epoch": 2.4,
      "learning_rate": 0.00012028753993610224,
      "loss": 0.2596,
      "step": 1500
    },
    {
      "epoch": 2.41,
      "learning_rate": 0.00011975505857294995,
      "loss": 0.1529,
      "step": 1510
    },
    {
      "epoch": 2.43,
      "learning_rate": 0.00011922257720979766,
      "loss": 0.2282,
      "step": 1520
    },
    {
      "epoch": 2.44,
      "learning_rate": 0.00011869009584664536,
      "loss": 0.193,
      "step": 1530
    },
    {
      "epoch": 2.46,
      "learning_rate": 0.00011815761448349307,
      "loss": 0.3611,
      "step": 1540
    },
    {
      "epoch": 2.48,
      "learning_rate": 0.0001176251331203408,
      "loss": 0.2629,
      "step": 1550
    },
    {
      "epoch": 2.49,
      "learning_rate": 0.00011709265175718851,
      "loss": 0.2813,
      "step": 1560
    },
    {
      "epoch": 2.51,
      "learning_rate": 0.00011656017039403621,
      "loss": 0.244,
      "step": 1570
    },
    {
      "epoch": 2.52,
      "learning_rate": 0.00011602768903088392,
      "loss": 0.2069,
      "step": 1580
    },
    {
      "epoch": 2.54,
      "learning_rate": 0.00011549520766773163,
      "loss": 0.2789,
      "step": 1590
    },
    {
      "epoch": 2.56,
      "learning_rate": 0.00011496272630457933,
      "loss": 0.2577,
      "step": 1600
    },
    {
      "epoch": 2.57,
      "learning_rate": 0.00011443024494142705,
      "loss": 0.1604,
      "step": 1610
    },
    {
      "epoch": 2.59,
      "learning_rate": 0.00011389776357827477,
      "loss": 0.2913,
      "step": 1620
    },
    {
      "epoch": 2.6,
      "learning_rate": 0.00011336528221512248,
      "loss": 0.292,
      "step": 1630
    },
    {
      "epoch": 2.62,
      "learning_rate": 0.00011283280085197018,
      "loss": 0.2309,
      "step": 1640
    },
    {
      "epoch": 2.64,
      "learning_rate": 0.0001123003194888179,
      "loss": 0.3248,
      "step": 1650
    },
    {
      "epoch": 2.65,
      "learning_rate": 0.0001117678381256656,
      "loss": 0.2171,
      "step": 1660
    },
    {
      "epoch": 2.67,
      "learning_rate": 0.0001112353567625133,
      "loss": 0.4351,
      "step": 1670
    },
    {
      "epoch": 2.68,
      "learning_rate": 0.00011070287539936102,
      "loss": 0.272,
      "step": 1680
    },
    {
      "epoch": 2.7,
      "learning_rate": 0.00011017039403620874,
      "loss": 0.4223,
      "step": 1690
    },
    {
      "epoch": 2.72,
      "learning_rate": 0.00010963791267305645,
      "loss": 0.1701,
      "step": 1700
    },
    {
      "epoch": 2.73,
      "learning_rate": 0.00010910543130990417,
      "loss": 0.3361,
      "step": 1710
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.00010857294994675187,
      "loss": 0.15,
      "step": 1720
    },
    {
      "epoch": 2.76,
      "learning_rate": 0.00010804046858359958,
      "loss": 0.2213,
      "step": 1730
    },
    {
      "epoch": 2.78,
      "learning_rate": 0.00010750798722044728,
      "loss": 0.1459,
      "step": 1740
    },
    {
      "epoch": 2.8,
      "learning_rate": 0.00010697550585729502,
      "loss": 0.2667,
      "step": 1750
    },
    {
      "epoch": 2.81,
      "learning_rate": 0.00010644302449414271,
      "loss": 0.2597,
      "step": 1760
    },
    {
      "epoch": 2.83,
      "learning_rate": 0.00010591054313099043,
      "loss": 0.1739,
      "step": 1770
    },
    {
      "epoch": 2.84,
      "learning_rate": 0.00010537806176783814,
      "loss": 0.3135,
      "step": 1780
    },
    {
      "epoch": 2.86,
      "learning_rate": 0.00010484558040468584,
      "loss": 0.3084,
      "step": 1790
    },
    {
      "epoch": 2.88,
      "learning_rate": 0.00010431309904153355,
      "loss": 0.1711,
      "step": 1800
    },
    {
      "epoch": 2.89,
      "learning_rate": 0.00010378061767838125,
      "loss": 0.3314,
      "step": 1810
    },
    {
      "epoch": 2.91,
      "learning_rate": 0.00010324813631522899,
      "loss": 0.232,
      "step": 1820
    },
    {
      "epoch": 2.92,
      "learning_rate": 0.00010271565495207669,
      "loss": 0.1611,
      "step": 1830
    },
    {
      "epoch": 2.94,
      "learning_rate": 0.0001021831735889244,
      "loss": 0.2533,
      "step": 1840
    },
    {
      "epoch": 2.96,
      "learning_rate": 0.00010165069222577211,
      "loss": 0.1702,
      "step": 1850
    },
    {
      "epoch": 2.97,
      "learning_rate": 0.00010111821086261981,
      "loss": 0.2574,
      "step": 1860
    },
    {
      "epoch": 2.99,
      "learning_rate": 0.00010058572949946752,
      "loss": 0.3248,
      "step": 1870
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.844559585492228,
      "eval_f1": 0.8348578980579512,
      "eval_loss": 0.4203706979751587,
      "eval_precision": 0.838060975475756,
      "eval_recall": 0.844559585492228,
      "eval_runtime": 1.2389,
      "eval_samples_per_second": 155.788,
      "eval_steps_per_second": 10.493,
      "step": 1878
    },
    {
      "epoch": 3.0,
      "learning_rate": 0.00010005324813631522,
      "loss": 0.1448,
      "step": 1880
    },
    {
      "epoch": 3.02,
      "learning_rate": 9.952076677316294e-05,
      "loss": 0.1276,
      "step": 1890
    },
    {
      "epoch": 3.04,
      "learning_rate": 9.898828541001066e-05,
      "loss": 0.1027,
      "step": 1900
    },
    {
      "epoch": 3.05,
      "learning_rate": 9.845580404685837e-05,
      "loss": 0.0971,
      "step": 1910
    },
    {
      "epoch": 3.07,
      "learning_rate": 9.792332268370608e-05,
      "loss": 0.0787,
      "step": 1920
    },
    {
      "epoch": 3.08,
      "learning_rate": 9.739084132055378e-05,
      "loss": 0.0975,
      "step": 1930
    },
    {
      "epoch": 3.1,
      "learning_rate": 9.68583599574015e-05,
      "loss": 0.0636,
      "step": 1940
    },
    {
      "epoch": 3.12,
      "learning_rate": 9.63258785942492e-05,
      "loss": 0.1065,
      "step": 1950
    },
    {
      "epoch": 3.13,
      "learning_rate": 9.579339723109692e-05,
      "loss": 0.0748,
      "step": 1960
    },
    {
      "epoch": 3.15,
      "learning_rate": 9.526091586794463e-05,
      "loss": 0.0585,
      "step": 1970
    },
    {
      "epoch": 3.16,
      "learning_rate": 9.472843450479234e-05,
      "loss": 0.0997,
      "step": 1980
    },
    {
      "epoch": 3.18,
      "learning_rate": 9.419595314164005e-05,
      "loss": 0.0941,
      "step": 1990
    },
    {
      "epoch": 3.19,
      "learning_rate": 9.366347177848775e-05,
      "loss": 0.104,
      "step": 2000
    },
    {
      "epoch": 3.21,
      "learning_rate": 9.313099041533548e-05,
      "loss": 0.0456,
      "step": 2010
    },
    {
      "epoch": 3.23,
      "learning_rate": 9.259850905218317e-05,
      "loss": 0.0911,
      "step": 2020
    },
    {
      "epoch": 3.24,
      "learning_rate": 9.206602768903089e-05,
      "loss": 0.1377,
      "step": 2030
    },
    {
      "epoch": 3.26,
      "learning_rate": 9.15335463258786e-05,
      "loss": 0.1084,
      "step": 2040
    },
    {
      "epoch": 3.27,
      "learning_rate": 9.100106496272631e-05,
      "loss": 0.0666,
      "step": 2050
    },
    {
      "epoch": 3.29,
      "learning_rate": 9.046858359957402e-05,
      "loss": 0.1412,
      "step": 2060
    },
    {
      "epoch": 3.31,
      "learning_rate": 8.993610223642172e-05,
      "loss": 0.131,
      "step": 2070
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.940362087326945e-05,
      "loss": 0.04,
      "step": 2080
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.887113951011715e-05,
      "loss": 0.0458,
      "step": 2090
    },
    {
      "epoch": 3.35,
      "learning_rate": 8.833865814696486e-05,
      "loss": 0.071,
      "step": 2100
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.780617678381257e-05,
      "loss": 0.2261,
      "step": 2110
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.727369542066028e-05,
      "loss": 0.0839,
      "step": 2120
    },
    {
      "epoch": 3.4,
      "learning_rate": 8.6741214057508e-05,
      "loss": 0.108,
      "step": 2130
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.620873269435569e-05,
      "loss": 0.1006,
      "step": 2140
    },
    {
      "epoch": 3.43,
      "learning_rate": 8.567625133120342e-05,
      "loss": 0.2175,
      "step": 2150
    },
    {
      "epoch": 3.45,
      "learning_rate": 8.514376996805112e-05,
      "loss": 0.162,
      "step": 2160
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.461128860489883e-05,
      "loss": 0.0975,
      "step": 2170
    },
    {
      "epoch": 3.48,
      "learning_rate": 8.407880724174654e-05,
      "loss": 0.1104,
      "step": 2180
    },
    {
      "epoch": 3.5,
      "learning_rate": 8.354632587859425e-05,
      "loss": 0.1068,
      "step": 2190
    },
    {
      "epoch": 3.51,
      "learning_rate": 8.301384451544197e-05,
      "loss": 0.2417,
      "step": 2200
    },
    {
      "epoch": 3.53,
      "learning_rate": 8.248136315228966e-05,
      "loss": 0.0763,
      "step": 2210
    },
    {
      "epoch": 3.55,
      "learning_rate": 8.194888178913739e-05,
      "loss": 0.0256,
      "step": 2220
    },
    {
      "epoch": 3.56,
      "learning_rate": 8.141640042598509e-05,
      "loss": 0.0714,
      "step": 2230
    },
    {
      "epoch": 3.58,
      "learning_rate": 8.08839190628328e-05,
      "loss": 0.1869,
      "step": 2240
    },
    {
      "epoch": 3.59,
      "learning_rate": 8.035143769968051e-05,
      "loss": 0.1378,
      "step": 2250
    },
    {
      "epoch": 3.61,
      "learning_rate": 7.981895633652823e-05,
      "loss": 0.0806,
      "step": 2260
    },
    {
      "epoch": 3.63,
      "learning_rate": 7.928647497337594e-05,
      "loss": 0.0646,
      "step": 2270
    },
    {
      "epoch": 3.64,
      "learning_rate": 7.875399361022364e-05,
      "loss": 0.0454,
      "step": 2280
    },
    {
      "epoch": 3.66,
      "learning_rate": 7.822151224707136e-05,
      "loss": 0.117,
      "step": 2290
    },
    {
      "epoch": 3.67,
      "learning_rate": 7.768903088391907e-05,
      "loss": 0.1052,
      "step": 2300
    },
    {
      "epoch": 3.69,
      "learning_rate": 7.715654952076677e-05,
      "loss": 0.1426,
      "step": 2310
    },
    {
      "epoch": 3.71,
      "learning_rate": 7.66240681576145e-05,
      "loss": 0.2087,
      "step": 2320
    },
    {
      "epoch": 3.72,
      "learning_rate": 7.60915867944622e-05,
      "loss": 0.1354,
      "step": 2330
    },
    {
      "epoch": 3.74,
      "learning_rate": 7.555910543130991e-05,
      "loss": 0.0493,
      "step": 2340
    },
    {
      "epoch": 3.75,
      "learning_rate": 7.502662406815761e-05,
      "loss": 0.0322,
      "step": 2350
    },
    {
      "epoch": 3.77,
      "learning_rate": 7.449414270500533e-05,
      "loss": 0.1888,
      "step": 2360
    },
    {
      "epoch": 3.79,
      "learning_rate": 7.396166134185304e-05,
      "loss": 0.0825,
      "step": 2370
    },
    {
      "epoch": 3.8,
      "learning_rate": 7.342917997870074e-05,
      "loss": 0.0555,
      "step": 2380
    },
    {
      "epoch": 3.82,
      "learning_rate": 7.289669861554847e-05,
      "loss": 0.1522,
      "step": 2390
    },
    {
      "epoch": 3.83,
      "learning_rate": 7.236421725239617e-05,
      "loss": 0.1257,
      "step": 2400
    },
    {
      "epoch": 3.85,
      "learning_rate": 7.183173588924388e-05,
      "loss": 0.1333,
      "step": 2410
    },
    {
      "epoch": 3.87,
      "learning_rate": 7.129925452609159e-05,
      "loss": 0.0742,
      "step": 2420
    },
    {
      "epoch": 3.88,
      "learning_rate": 7.07667731629393e-05,
      "loss": 0.1136,
      "step": 2430
    },
    {
      "epoch": 3.9,
      "learning_rate": 7.023429179978702e-05,
      "loss": 0.0404,
      "step": 2440
    },
    {
      "epoch": 3.91,
      "learning_rate": 6.970181043663471e-05,
      "loss": 0.0618,
      "step": 2450
    },
    {
      "epoch": 3.93,
      "learning_rate": 6.916932907348244e-05,
      "loss": 0.0682,
      "step": 2460
    },
    {
      "epoch": 3.95,
      "learning_rate": 6.863684771033014e-05,
      "loss": 0.1142,
      "step": 2470
    },
    {
      "epoch": 3.96,
      "learning_rate": 6.810436634717785e-05,
      "loss": 0.0729,
      "step": 2480
    },
    {
      "epoch": 3.98,
      "learning_rate": 6.757188498402556e-05,
      "loss": 0.0926,
      "step": 2490
    },
    {
      "epoch": 3.99,
      "learning_rate": 6.703940362087328e-05,
      "loss": 0.1064,
      "step": 2500
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.844559585492228,
      "eval_f1": 0.8375340248723185,
      "eval_loss": 0.6135075688362122,
      "eval_precision": 0.8335855797420394,
      "eval_recall": 0.844559585492228,
      "eval_runtime": 1.2164,
      "eval_samples_per_second": 158.659,
      "eval_steps_per_second": 10.687,
      "step": 2504
    },
    {
      "epoch": 4.01,
      "learning_rate": 6.650692225772099e-05,
      "loss": 0.0954,
      "step": 2510
    },
    {
      "epoch": 4.03,
      "learning_rate": 6.597444089456869e-05,
      "loss": 0.0108,
      "step": 2520
    },
    {
      "epoch": 4.04,
      "learning_rate": 6.544195953141641e-05,
      "loss": 0.0846,
      "step": 2530
    },
    {
      "epoch": 4.06,
      "learning_rate": 6.490947816826411e-05,
      "loss": 0.0399,
      "step": 2540
    },
    {
      "epoch": 4.07,
      "learning_rate": 6.437699680511182e-05,
      "loss": 0.0445,
      "step": 2550
    },
    {
      "epoch": 4.09,
      "learning_rate": 6.384451544195953e-05,
      "loss": 0.0832,
      "step": 2560
    },
    {
      "epoch": 4.11,
      "learning_rate": 6.331203407880725e-05,
      "loss": 0.0363,
      "step": 2570
    },
    {
      "epoch": 4.12,
      "learning_rate": 6.277955271565496e-05,
      "loss": 0.0089,
      "step": 2580
    },
    {
      "epoch": 4.14,
      "learning_rate": 6.224707135250266e-05,
      "loss": 0.0412,
      "step": 2590
    },
    {
      "epoch": 4.15,
      "learning_rate": 6.171458998935038e-05,
      "loss": 0.0094,
      "step": 2600
    },
    {
      "epoch": 4.17,
      "learning_rate": 6.118210862619808e-05,
      "loss": 0.0215,
      "step": 2610
    },
    {
      "epoch": 4.19,
      "learning_rate": 6.0649627263045794e-05,
      "loss": 0.0142,
      "step": 2620
    },
    {
      "epoch": 4.2,
      "learning_rate": 6.01171458998935e-05,
      "loss": 0.0142,
      "step": 2630
    },
    {
      "epoch": 4.22,
      "learning_rate": 5.958466453674122e-05,
      "loss": 0.0439,
      "step": 2640
    },
    {
      "epoch": 4.23,
      "learning_rate": 5.9052183173588923e-05,
      "loss": 0.0233,
      "step": 2650
    },
    {
      "epoch": 4.25,
      "learning_rate": 5.8519701810436636e-05,
      "loss": 0.0528,
      "step": 2660
    },
    {
      "epoch": 4.27,
      "learning_rate": 5.7987220447284354e-05,
      "loss": 0.017,
      "step": 2670
    },
    {
      "epoch": 4.28,
      "learning_rate": 5.745473908413206e-05,
      "loss": 0.0687,
      "step": 2680
    },
    {
      "epoch": 4.3,
      "learning_rate": 5.6922257720979765e-05,
      "loss": 0.0083,
      "step": 2690
    },
    {
      "epoch": 4.31,
      "learning_rate": 5.6389776357827484e-05,
      "loss": 0.0322,
      "step": 2700
    },
    {
      "epoch": 4.33,
      "learning_rate": 5.585729499467519e-05,
      "loss": 0.016,
      "step": 2710
    },
    {
      "epoch": 4.35,
      "learning_rate": 5.53248136315229e-05,
      "loss": 0.0246,
      "step": 2720
    },
    {
      "epoch": 4.36,
      "learning_rate": 5.479233226837061e-05,
      "loss": 0.0083,
      "step": 2730
    },
    {
      "epoch": 4.38,
      "learning_rate": 5.4259850905218326e-05,
      "loss": 0.0319,
      "step": 2740
    },
    {
      "epoch": 4.39,
      "learning_rate": 5.372736954206603e-05,
      "loss": 0.0055,
      "step": 2750
    },
    {
      "epoch": 4.41,
      "learning_rate": 5.3194888178913736e-05,
      "loss": 0.0263,
      "step": 2760
    },
    {
      "epoch": 4.42,
      "learning_rate": 5.2662406815761455e-05,
      "loss": 0.0522,
      "step": 2770
    },
    {
      "epoch": 4.44,
      "learning_rate": 5.212992545260916e-05,
      "loss": 0.0675,
      "step": 2780
    },
    {
      "epoch": 4.46,
      "learning_rate": 5.159744408945687e-05,
      "loss": 0.0258,
      "step": 2790
    },
    {
      "epoch": 4.47,
      "learning_rate": 5.106496272630458e-05,
      "loss": 0.0062,
      "step": 2800
    },
    {
      "epoch": 4.49,
      "learning_rate": 5.05324813631523e-05,
      "loss": 0.0482,
      "step": 2810
    },
    {
      "epoch": 4.5,
      "learning_rate": 5e-05,
      "loss": 0.0135,
      "step": 2820
    },
    {
      "epoch": 4.52,
      "learning_rate": 4.9467518636847715e-05,
      "loss": 0.0627,
      "step": 2830
    },
    {
      "epoch": 4.54,
      "learning_rate": 4.893503727369542e-05,
      "loss": 0.0783,
      "step": 2840
    },
    {
      "epoch": 4.55,
      "learning_rate": 4.840255591054313e-05,
      "loss": 0.033,
      "step": 2850
    },
    {
      "epoch": 4.57,
      "learning_rate": 4.7870074547390844e-05,
      "loss": 0.0352,
      "step": 2860
    },
    {
      "epoch": 4.58,
      "learning_rate": 4.7337593184238556e-05,
      "loss": 0.0138,
      "step": 2870
    },
    {
      "epoch": 4.6,
      "learning_rate": 4.680511182108626e-05,
      "loss": 0.0765,
      "step": 2880
    },
    {
      "epoch": 4.62,
      "learning_rate": 4.6272630457933974e-05,
      "loss": 0.0229,
      "step": 2890
    },
    {
      "epoch": 4.63,
      "learning_rate": 4.5740149094781686e-05,
      "loss": 0.01,
      "step": 2900
    },
    {
      "epoch": 4.65,
      "learning_rate": 4.520766773162939e-05,
      "loss": 0.0529,
      "step": 2910
    },
    {
      "epoch": 4.66,
      "learning_rate": 4.46751863684771e-05,
      "loss": 0.0303,
      "step": 2920
    },
    {
      "epoch": 4.68,
      "learning_rate": 4.4142705005324815e-05,
      "loss": 0.0302,
      "step": 2930
    },
    {
      "epoch": 4.7,
      "learning_rate": 4.361022364217253e-05,
      "loss": 0.0099,
      "step": 2940
    },
    {
      "epoch": 4.71,
      "learning_rate": 4.307774227902024e-05,
      "loss": 0.0754,
      "step": 2950
    },
    {
      "epoch": 4.73,
      "learning_rate": 4.2545260915867945e-05,
      "loss": 0.0482,
      "step": 2960
    },
    {
      "epoch": 4.74,
      "learning_rate": 4.201277955271566e-05,
      "loss": 0.0329,
      "step": 2970
    },
    {
      "epoch": 4.76,
      "learning_rate": 4.148029818956336e-05,
      "loss": 0.0516,
      "step": 2980
    },
    {
      "epoch": 4.78,
      "learning_rate": 4.094781682641108e-05,
      "loss": 0.0054,
      "step": 2990
    },
    {
      "epoch": 4.79,
      "learning_rate": 4.041533546325879e-05,
      "loss": 0.0291,
      "step": 3000
    },
    {
      "epoch": 4.81,
      "learning_rate": 3.98828541001065e-05,
      "loss": 0.0309,
      "step": 3010
    },
    {
      "epoch": 4.82,
      "learning_rate": 3.935037273695421e-05,
      "loss": 0.0405,
      "step": 3020
    },
    {
      "epoch": 4.84,
      "learning_rate": 3.8817891373801916e-05,
      "loss": 0.0058,
      "step": 3030
    },
    {
      "epoch": 4.86,
      "learning_rate": 3.828541001064963e-05,
      "loss": 0.0244,
      "step": 3040
    },
    {
      "epoch": 4.87,
      "learning_rate": 3.775292864749734e-05,
      "loss": 0.0047,
      "step": 3050
    },
    {
      "epoch": 4.89,
      "learning_rate": 3.722044728434505e-05,
      "loss": 0.0221,
      "step": 3060
    },
    {
      "epoch": 4.9,
      "learning_rate": 3.668796592119276e-05,
      "loss": 0.0127,
      "step": 3070
    },
    {
      "epoch": 4.92,
      "learning_rate": 3.615548455804047e-05,
      "loss": 0.0183,
      "step": 3080
    },
    {
      "epoch": 4.94,
      "learning_rate": 3.562300319488818e-05,
      "loss": 0.0198,
      "step": 3090
    },
    {
      "epoch": 4.95,
      "learning_rate": 3.509052183173589e-05,
      "loss": 0.0146,
      "step": 3100
    },
    {
      "epoch": 4.97,
      "learning_rate": 3.45580404685836e-05,
      "loss": 0.0294,
      "step": 3110
    },
    {
      "epoch": 4.98,
      "learning_rate": 3.402555910543131e-05,
      "loss": 0.024,
      "step": 3120
    },
    {
      "epoch": 5.0,
      "learning_rate": 3.3493077742279024e-05,
      "loss": 0.0053,
      "step": 3130
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8704663212435233,
      "eval_f1": 0.8716019534644225,
      "eval_loss": 0.5843242406845093,
      "eval_precision": 0.8762224824481195,
      "eval_recall": 0.8704663212435233,
      "eval_runtime": 1.3048,
      "eval_samples_per_second": 147.916,
      "eval_steps_per_second": 9.963,
      "step": 3130
    },
    {
      "epoch": 5.02,
      "learning_rate": 3.2960596379126736e-05,
      "loss": 0.0037,
      "step": 3140
    },
    {
      "epoch": 5.03,
      "learning_rate": 3.242811501597444e-05,
      "loss": 0.0032,
      "step": 3150
    },
    {
      "epoch": 5.05,
      "learning_rate": 3.1895633652822154e-05,
      "loss": 0.0033,
      "step": 3160
    },
    {
      "epoch": 5.06,
      "learning_rate": 3.136315228966986e-05,
      "loss": 0.0037,
      "step": 3170
    },
    {
      "epoch": 5.08,
      "learning_rate": 3.083067092651757e-05,
      "loss": 0.0033,
      "step": 3180
    },
    {
      "epoch": 5.1,
      "learning_rate": 3.029818956336528e-05,
      "loss": 0.003,
      "step": 3190
    },
    {
      "epoch": 5.11,
      "learning_rate": 2.9765708200212995e-05,
      "loss": 0.012,
      "step": 3200
    },
    {
      "epoch": 5.13,
      "learning_rate": 2.9233226837060707e-05,
      "loss": 0.003,
      "step": 3210
    },
    {
      "epoch": 5.14,
      "learning_rate": 2.8700745473908413e-05,
      "loss": 0.003,
      "step": 3220
    },
    {
      "epoch": 5.16,
      "learning_rate": 2.8168264110756125e-05,
      "loss": 0.0034,
      "step": 3230
    },
    {
      "epoch": 5.18,
      "learning_rate": 2.7635782747603834e-05,
      "loss": 0.0042,
      "step": 3240
    },
    {
      "epoch": 5.19,
      "learning_rate": 2.7103301384451546e-05,
      "loss": 0.0046,
      "step": 3250
    },
    {
      "epoch": 5.21,
      "learning_rate": 2.6570820021299255e-05,
      "loss": 0.0029,
      "step": 3260
    },
    {
      "epoch": 5.22,
      "learning_rate": 2.6038338658146967e-05,
      "loss": 0.0034,
      "step": 3270
    },
    {
      "epoch": 5.24,
      "learning_rate": 2.550585729499468e-05,
      "loss": 0.0421,
      "step": 3280
    },
    {
      "epoch": 5.26,
      "learning_rate": 2.4973375931842384e-05,
      "loss": 0.0031,
      "step": 3290
    },
    {
      "epoch": 5.27,
      "learning_rate": 2.44408945686901e-05,
      "loss": 0.0032,
      "step": 3300
    },
    {
      "epoch": 5.29,
      "learning_rate": 2.390841320553781e-05,
      "loss": 0.0034,
      "step": 3310
    },
    {
      "epoch": 5.3,
      "learning_rate": 2.3375931842385517e-05,
      "loss": 0.0035,
      "step": 3320
    },
    {
      "epoch": 5.32,
      "learning_rate": 2.284345047923323e-05,
      "loss": 0.0034,
      "step": 3330
    },
    {
      "epoch": 5.34,
      "learning_rate": 2.2310969116080938e-05,
      "loss": 0.003,
      "step": 3340
    },
    {
      "epoch": 5.35,
      "learning_rate": 2.1778487752928647e-05,
      "loss": 0.0049,
      "step": 3350
    },
    {
      "epoch": 5.37,
      "learning_rate": 2.124600638977636e-05,
      "loss": 0.0297,
      "step": 3360
    },
    {
      "epoch": 5.38,
      "learning_rate": 2.071352502662407e-05,
      "loss": 0.003,
      "step": 3370
    },
    {
      "epoch": 5.4,
      "learning_rate": 2.018104366347178e-05,
      "loss": 0.003,
      "step": 3380
    },
    {
      "epoch": 5.42,
      "learning_rate": 1.964856230031949e-05,
      "loss": 0.0033,
      "step": 3390
    },
    {
      "epoch": 5.43,
      "learning_rate": 1.91160809371672e-05,
      "loss": 0.0029,
      "step": 3400
    },
    {
      "epoch": 5.45,
      "learning_rate": 1.858359957401491e-05,
      "loss": 0.0031,
      "step": 3410
    },
    {
      "epoch": 5.46,
      "learning_rate": 1.805111821086262e-05,
      "loss": 0.0029,
      "step": 3420
    },
    {
      "epoch": 5.48,
      "learning_rate": 1.7518636847710333e-05,
      "loss": 0.0032,
      "step": 3430
    },
    {
      "epoch": 5.5,
      "learning_rate": 1.6986155484558042e-05,
      "loss": 0.0032,
      "step": 3440
    },
    {
      "epoch": 5.51,
      "learning_rate": 1.645367412140575e-05,
      "loss": 0.0029,
      "step": 3450
    },
    {
      "epoch": 5.53,
      "learning_rate": 1.5921192758253463e-05,
      "loss": 0.0042,
      "step": 3460
    },
    {
      "epoch": 5.54,
      "learning_rate": 1.5388711395101172e-05,
      "loss": 0.0032,
      "step": 3470
    },
    {
      "epoch": 5.56,
      "learning_rate": 1.485623003194888e-05,
      "loss": 0.0031,
      "step": 3480
    },
    {
      "epoch": 5.58,
      "learning_rate": 1.4323748668796594e-05,
      "loss": 0.0031,
      "step": 3490
    },
    {
      "epoch": 5.59,
      "learning_rate": 1.3791267305644303e-05,
      "loss": 0.0046,
      "step": 3500
    },
    {
      "epoch": 5.61,
      "learning_rate": 1.3258785942492014e-05,
      "loss": 0.0134,
      "step": 3510
    },
    {
      "epoch": 5.62,
      "learning_rate": 1.2726304579339724e-05,
      "loss": 0.0029,
      "step": 3520
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.2193823216187434e-05,
      "loss": 0.0027,
      "step": 3530
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.1661341853035145e-05,
      "loss": 0.0057,
      "step": 3540
    },
    {
      "epoch": 5.67,
      "learning_rate": 1.1128860489882854e-05,
      "loss": 0.0165,
      "step": 3550
    },
    {
      "epoch": 5.69,
      "learning_rate": 1.0596379126730564e-05,
      "loss": 0.0031,
      "step": 3560
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.0063897763578276e-05,
      "loss": 0.0027,
      "step": 3570
    },
    {
      "epoch": 5.72,
      "learning_rate": 9.531416400425985e-06,
      "loss": 0.0033,
      "step": 3580
    },
    {
      "epoch": 5.73,
      "learning_rate": 8.998935037273695e-06,
      "loss": 0.003,
      "step": 3590
    },
    {
      "epoch": 5.75,
      "learning_rate": 8.466453674121406e-06,
      "loss": 0.0035,
      "step": 3600
    },
    {
      "epoch": 5.77,
      "learning_rate": 7.933972310969116e-06,
      "loss": 0.0029,
      "step": 3610
    },
    {
      "epoch": 5.78,
      "learning_rate": 7.4014909478168266e-06,
      "loss": 0.0026,
      "step": 3620
    },
    {
      "epoch": 5.8,
      "learning_rate": 6.869009584664538e-06,
      "loss": 0.0023,
      "step": 3630
    },
    {
      "epoch": 5.81,
      "learning_rate": 6.336528221512247e-06,
      "loss": 0.0027,
      "step": 3640
    },
    {
      "epoch": 5.83,
      "learning_rate": 5.804046858359958e-06,
      "loss": 0.0025,
      "step": 3650
    },
    {
      "epoch": 5.85,
      "learning_rate": 5.2715654952076674e-06,
      "loss": 0.0047,
      "step": 3660
    },
    {
      "epoch": 5.86,
      "learning_rate": 4.739084132055379e-06,
      "loss": 0.0025,
      "step": 3670
    },
    {
      "epoch": 5.88,
      "learning_rate": 4.206602768903089e-06,
      "loss": 0.0029,
      "step": 3680
    },
    {
      "epoch": 5.89,
      "learning_rate": 3.6741214057507987e-06,
      "loss": 0.0026,
      "step": 3690
    },
    {
      "epoch": 5.91,
      "learning_rate": 3.141640042598509e-06,
      "loss": 0.003,
      "step": 3700
    },
    {
      "epoch": 5.93,
      "learning_rate": 2.6091586794462196e-06,
      "loss": 0.0029,
      "step": 3710
    },
    {
      "epoch": 5.94,
      "learning_rate": 2.0766773162939296e-06,
      "loss": 0.0027,
      "step": 3720
    },
    {
      "epoch": 5.96,
      "learning_rate": 1.54419595314164e-06,
      "loss": 0.0537,
      "step": 3730
    },
    {
      "epoch": 5.97,
      "learning_rate": 1.0117145899893504e-06,
      "loss": 0.0026,
      "step": 3740
    },
    {
      "epoch": 5.99,
      "learning_rate": 4.792332268370607e-07,
      "loss": 0.0024,
      "step": 3750
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8497409326424871,
      "eval_f1": 0.8486721543751371,
      "eval_loss": 0.6812893748283386,
      "eval_precision": 0.8529511188425435,
      "eval_recall": 0.8497409326424871,
      "eval_runtime": 1.3136,
      "eval_samples_per_second": 146.93,
      "eval_steps_per_second": 9.897,
      "step": 3756
    },
    {
      "epoch": 6.0,
      "step": 3756,
      "total_flos": 4.65670232933972e+18,
      "train_loss": 0.26366007817597686,
      "train_runtime": 1257.0372,
      "train_samples_per_second": 47.803,
      "train_steps_per_second": 2.988
    }
  ],
  "max_steps": 3756,
  "num_train_epochs": 6,
  "total_flos": 4.65670232933972e+18,
  "trial_name": null,
  "trial_params": null
}