{
  "best_metric": 11.462658915223917,
  "best_model_checkpoint": "./model_out_trpro_full_canto_resume/checkpoint-5000",
  "epoch": 3.324468085106383,
  "global_step": 5000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.02,
      "learning_rate": 4.4e-07,
      "loss": 2.6129,
      "step": 25
    },
    {
      "epoch": 0.03,
      "learning_rate": 9.400000000000001e-07,
      "loss": 1.9202,
      "step": 50
    },
    {
      "epoch": 0.05,
      "learning_rate": 1.44e-06,
      "loss": 1.0182,
      "step": 75
    },
    {
      "epoch": 0.07,
      "learning_rate": 1.94e-06,
      "loss": 0.3318,
      "step": 100
    },
    {
      "epoch": 0.08,
      "learning_rate": 2.4400000000000004e-06,
      "loss": 0.262,
      "step": 125
    },
    {
      "epoch": 0.1,
      "learning_rate": 2.9400000000000002e-06,
      "loss": 0.2435,
      "step": 150
    },
    {
      "epoch": 0.12,
      "learning_rate": 3.44e-06,
      "loss": 0.2288,
      "step": 175
    },
    {
      "epoch": 0.13,
      "learning_rate": 3.94e-06,
      "loss": 0.2156,
      "step": 200
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.440000000000001e-06,
      "loss": 0.2003,
      "step": 225
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.94e-06,
      "loss": 0.1971,
      "step": 250
    },
    {
      "epoch": 0.18,
      "learning_rate": 5.4400000000000004e-06,
      "loss": 0.1893,
      "step": 275
    },
    {
      "epoch": 0.2,
      "learning_rate": 5.94e-06,
      "loss": 0.1953,
      "step": 300
    },
    {
      "epoch": 0.22,
      "learning_rate": 6.440000000000001e-06,
      "loss": 0.1757,
      "step": 325
    },
    {
      "epoch": 0.23,
      "learning_rate": 6.9400000000000005e-06,
      "loss": 0.1817,
      "step": 350
    },
    {
      "epoch": 0.25,
      "learning_rate": 7.440000000000001e-06,
      "loss": 0.1814,
      "step": 375
    },
    {
      "epoch": 0.27,
      "learning_rate": 7.94e-06,
      "loss": 0.1736,
      "step": 400
    },
    {
      "epoch": 0.28,
      "learning_rate": 8.44e-06,
      "loss": 0.1731,
      "step": 425
    },
    {
      "epoch": 0.3,
      "learning_rate": 8.94e-06,
      "loss": 0.1741,
      "step": 450
    },
    {
      "epoch": 0.32,
      "learning_rate": 9.440000000000001e-06,
      "loss": 0.1611,
      "step": 475
    },
    {
      "epoch": 0.33,
      "learning_rate": 9.940000000000001e-06,
      "loss": 0.152,
      "step": 500
    },
    {
      "epoch": 0.35,
      "learning_rate": 9.951111111111111e-06,
      "loss": 0.1447,
      "step": 525
    },
    {
      "epoch": 0.37,
      "learning_rate": 9.895555555555557e-06,
      "loss": 0.1462,
      "step": 550
    },
    {
      "epoch": 0.38,
      "learning_rate": 9.84e-06,
      "loss": 0.1554,
      "step": 575
    },
    {
      "epoch": 0.4,
      "learning_rate": 9.784444444444445e-06,
      "loss": 0.1528,
      "step": 600
    },
    {
      "epoch": 0.42,
      "learning_rate": 9.72888888888889e-06,
      "loss": 0.1443,
      "step": 625
    },
    {
      "epoch": 0.43,
      "learning_rate": 9.673333333333334e-06,
      "loss": 0.1444,
      "step": 650
    },
    {
      "epoch": 0.45,
      "learning_rate": 9.617777777777778e-06,
      "loss": 0.1436,
      "step": 675
    },
    {
      "epoch": 0.47,
      "learning_rate": 9.562222222222223e-06,
      "loss": 0.135,
      "step": 700
    },
    {
      "epoch": 0.48,
      "learning_rate": 9.506666666666667e-06,
      "loss": 0.1317,
      "step": 725
    },
    {
      "epoch": 0.5,
      "learning_rate": 9.451111111111112e-06,
      "loss": 0.1307,
      "step": 750
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.395555555555556e-06,
      "loss": 0.138,
      "step": 775
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.340000000000002e-06,
      "loss": 0.1283,
      "step": 800
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.284444444444444e-06,
      "loss": 0.1287,
      "step": 825
    },
    {
      "epoch": 0.57,
      "learning_rate": 9.22888888888889e-06,
      "loss": 0.127,
      "step": 850
    },
    {
      "epoch": 0.58,
      "learning_rate": 9.173333333333334e-06,
      "loss": 0.1253,
      "step": 875
    },
    {
      "epoch": 0.6,
      "learning_rate": 9.117777777777778e-06,
      "loss": 0.1206,
      "step": 900
    },
    {
      "epoch": 0.62,
      "learning_rate": 9.062222222222224e-06,
      "loss": 0.1271,
      "step": 925
    },
    {
      "epoch": 0.63,
      "learning_rate": 9.006666666666666e-06,
      "loss": 0.1152,
      "step": 950
    },
    {
      "epoch": 0.65,
      "learning_rate": 8.951111111111112e-06,
      "loss": 0.116,
      "step": 975
    },
    {
      "epoch": 0.66,
      "learning_rate": 8.895555555555556e-06,
      "loss": 0.11,
      "step": 1000
    },
    {
      "epoch": 0.66,
      "eval_cer": 14.637769117741009,
      "eval_loss": 0.32944002747535706,
      "eval_runtime": 1229.289,
      "eval_samples_per_second": 4.548,
      "eval_steps_per_second": 0.569,
      "step": 1000
    },
    {
      "epoch": 0.68,
      "learning_rate": 8.84e-06,
      "loss": 0.1105,
      "step": 1025
    },
    {
      "epoch": 0.7,
      "learning_rate": 8.784444444444446e-06,
      "loss": 0.1091,
      "step": 1050
    },
    {
      "epoch": 0.71,
      "learning_rate": 8.72888888888889e-06,
      "loss": 0.1019,
      "step": 1075
    },
    {
      "epoch": 0.73,
      "learning_rate": 8.673333333333334e-06,
      "loss": 0.1122,
      "step": 1100
    },
    {
      "epoch": 0.75,
      "learning_rate": 8.617777777777778e-06,
      "loss": 0.1119,
      "step": 1125
    },
    {
      "epoch": 0.76,
      "learning_rate": 8.562222222222224e-06,
      "loss": 0.109,
      "step": 1150
    },
    {
      "epoch": 0.78,
      "learning_rate": 8.506666666666668e-06,
      "loss": 0.1039,
      "step": 1175
    },
    {
      "epoch": 0.8,
      "learning_rate": 8.451111111111112e-06,
      "loss": 0.1001,
      "step": 1200
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.395555555555557e-06,
      "loss": 0.1075,
      "step": 1225
    },
    {
      "epoch": 0.83,
      "learning_rate": 8.34e-06,
      "loss": 0.0968,
      "step": 1250
    },
    {
      "epoch": 0.85,
      "learning_rate": 8.284444444444446e-06,
      "loss": 0.0993,
      "step": 1275
    },
    {
      "epoch": 0.86,
      "learning_rate": 8.22888888888889e-06,
      "loss": 0.0982,
      "step": 1300
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.173333333333334e-06,
      "loss": 0.0927,
      "step": 1325
    },
    {
      "epoch": 0.9,
      "learning_rate": 8.11777777777778e-06,
      "loss": 0.0963,
      "step": 1350
    },
    {
      "epoch": 0.91,
      "learning_rate": 8.062222222222222e-06,
      "loss": 0.0939,
      "step": 1375
    },
    {
      "epoch": 0.93,
      "learning_rate": 8.006666666666667e-06,
      "loss": 0.0949,
      "step": 1400
    },
    {
      "epoch": 0.95,
      "learning_rate": 7.951111111111111e-06,
      "loss": 0.0963,
      "step": 1425
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.895555555555557e-06,
      "loss": 0.0926,
      "step": 1450
    },
    {
      "epoch": 0.98,
      "learning_rate": 7.840000000000001e-06,
      "loss": 0.0871,
      "step": 1475
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.784444444444445e-06,
      "loss": 0.0896,
      "step": 1500
    },
    {
      "epoch": 1.01,
      "learning_rate": 7.72888888888889e-06,
      "loss": 0.0679,
      "step": 1525
    },
    {
      "epoch": 1.03,
      "learning_rate": 7.673333333333333e-06,
      "loss": 0.0638,
      "step": 1550
    },
    {
      "epoch": 1.05,
      "learning_rate": 7.617777777777778e-06,
      "loss": 0.0603,
      "step": 1575
    },
    {
      "epoch": 1.06,
      "learning_rate": 7.562222222222223e-06,
      "loss": 0.066,
      "step": 1600
    },
    {
      "epoch": 1.08,
      "learning_rate": 7.506666666666668e-06,
      "loss": 0.0616,
      "step": 1625
    },
    {
      "epoch": 1.1,
      "learning_rate": 7.451111111111111e-06,
      "loss": 0.0632,
      "step": 1650
    },
    {
      "epoch": 1.11,
      "learning_rate": 7.395555555555556e-06,
      "loss": 0.0592,
      "step": 1675
    },
    {
      "epoch": 1.13,
      "learning_rate": 7.340000000000001e-06,
      "loss": 0.0634,
      "step": 1700
    },
    {
      "epoch": 1.15,
      "learning_rate": 7.284444444444445e-06,
      "loss": 0.0587,
      "step": 1725
    },
    {
      "epoch": 1.16,
      "learning_rate": 7.22888888888889e-06,
      "loss": 0.0593,
      "step": 1750
    },
    {
      "epoch": 1.18,
      "learning_rate": 7.173333333333335e-06,
      "loss": 0.0557,
      "step": 1775
    },
    {
      "epoch": 1.2,
      "learning_rate": 7.117777777777778e-06,
      "loss": 0.0602,
      "step": 1800
    },
    {
      "epoch": 1.21,
      "learning_rate": 7.062222222222223e-06,
      "loss": 0.0625,
      "step": 1825
    },
    {
      "epoch": 1.23,
      "learning_rate": 7.006666666666667e-06,
      "loss": 0.0586,
      "step": 1850
    },
    {
      "epoch": 1.25,
      "learning_rate": 6.951111111111112e-06,
      "loss": 0.0538,
      "step": 1875
    },
    {
      "epoch": 1.26,
      "learning_rate": 6.8955555555555565e-06,
      "loss": 0.0558,
      "step": 1900
    },
    {
      "epoch": 1.28,
      "learning_rate": 6.8400000000000014e-06,
      "loss": 0.059,
      "step": 1925
    },
    {
      "epoch": 1.3,
      "learning_rate": 6.784444444444445e-06,
      "loss": 0.0568,
      "step": 1950
    },
    {
      "epoch": 1.31,
      "learning_rate": 6.7288888888888895e-06,
      "loss": 0.0557,
      "step": 1975
    },
    {
      "epoch": 1.33,
      "learning_rate": 6.6733333333333335e-06,
      "loss": 0.0563,
      "step": 2000
    },
    {
      "epoch": 1.33,
      "eval_cer": 12.119082603973679,
      "eval_loss": 0.288747102022171,
      "eval_runtime": 1184.6421,
      "eval_samples_per_second": 4.72,
      "eval_steps_per_second": 0.59,
      "step": 2000
    },
    {
      "epoch": 1.35,
      "learning_rate": 6.617777777777778e-06,
      "loss": 0.0546,
      "step": 2025
    },
    {
      "epoch": 1.36,
      "learning_rate": 6.562222222222223e-06,
      "loss": 0.0537,
      "step": 2050
    },
    {
      "epoch": 1.38,
      "learning_rate": 6.5066666666666665e-06,
      "loss": 0.0535,
      "step": 2075
    },
    {
      "epoch": 1.4,
      "learning_rate": 6.451111111111111e-06,
      "loss": 0.0526,
      "step": 2100
    },
    {
      "epoch": 1.41,
      "learning_rate": 6.395555555555556e-06,
      "loss": 0.0542,
      "step": 2125
    },
    {
      "epoch": 1.43,
      "learning_rate": 6.34e-06,
      "loss": 0.0554,
      "step": 2150
    },
    {
      "epoch": 1.45,
      "learning_rate": 6.284444444444445e-06,
      "loss": 0.0547,
      "step": 2175
    },
    {
      "epoch": 1.46,
      "learning_rate": 6.22888888888889e-06,
      "loss": 0.0549,
      "step": 2200
    },
    {
      "epoch": 1.48,
      "learning_rate": 6.173333333333333e-06,
      "loss": 0.0554,
      "step": 2225
    },
    {
      "epoch": 1.5,
      "learning_rate": 6.117777777777778e-06,
      "loss": 0.0527,
      "step": 2250
    },
    {
      "epoch": 1.51,
      "learning_rate": 6.062222222222223e-06,
      "loss": 0.051,
      "step": 2275
    },
    {
      "epoch": 1.53,
      "learning_rate": 6.006666666666667e-06,
      "loss": 0.0515,
      "step": 2300
    },
    {
      "epoch": 1.55,
      "learning_rate": 5.951111111111112e-06,
      "loss": 0.0524,
      "step": 2325
    },
    {
      "epoch": 1.56,
      "learning_rate": 5.895555555555557e-06,
      "loss": 0.0535,
      "step": 2350
    },
    {
      "epoch": 1.58,
      "learning_rate": 5.84e-06,
      "loss": 0.047,
      "step": 2375
    },
    {
      "epoch": 1.6,
      "learning_rate": 5.784444444444445e-06,
      "loss": 0.0524,
      "step": 2400
    },
    {
      "epoch": 1.61,
      "learning_rate": 5.72888888888889e-06,
      "loss": 0.0447,
      "step": 2425
    },
    {
      "epoch": 1.63,
      "learning_rate": 5.673333333333334e-06,
      "loss": 0.0485,
      "step": 2450
    },
    {
      "epoch": 1.65,
      "learning_rate": 5.617777777777779e-06,
      "loss": 0.0497,
      "step": 2475
    },
    {
      "epoch": 1.66,
      "learning_rate": 5.562222222222222e-06,
      "loss": 0.046,
      "step": 2500
    },
    {
      "epoch": 1.68,
      "learning_rate": 5.506666666666667e-06,
      "loss": 0.0526,
      "step": 2525
    },
    {
      "epoch": 1.7,
      "learning_rate": 5.451111111111112e-06,
      "loss": 0.0465,
      "step": 2550
    },
    {
      "epoch": 1.71,
      "learning_rate": 5.3955555555555565e-06,
      "loss": 0.0468,
      "step": 2575
    },
    {
      "epoch": 1.73,
      "learning_rate": 5.3400000000000005e-06,
      "loss": 0.0492,
      "step": 2600
    },
    {
      "epoch": 1.75,
      "learning_rate": 5.2844444444444454e-06,
      "loss": 0.0475,
      "step": 2625
    },
    {
      "epoch": 1.76,
      "learning_rate": 5.228888888888889e-06,
      "loss": 0.0463,
      "step": 2650
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.1733333333333335e-06,
      "loss": 0.0462,
      "step": 2675
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.117777777777778e-06,
      "loss": 0.0505,
      "step": 2700
    },
    {
      "epoch": 1.81,
      "learning_rate": 5.062222222222222e-06,
      "loss": 0.0467,
      "step": 2725
    },
    {
      "epoch": 1.83,
      "learning_rate": 5.006666666666667e-06,
      "loss": 0.0454,
      "step": 2750
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.951111111111111e-06,
      "loss": 0.0448,
      "step": 2775
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.895555555555556e-06,
      "loss": 0.0442,
      "step": 2800
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.84e-06,
      "loss": 0.0419,
      "step": 2825
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.784444444444445e-06,
      "loss": 0.0441,
      "step": 2850
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.728888888888889e-06,
      "loss": 0.0474,
      "step": 2875
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.673333333333333e-06,
      "loss": 0.0428,
      "step": 2900
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.617777777777778e-06,
      "loss": 0.0399,
      "step": 2925
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.562222222222222e-06,
      "loss": 0.0388,
      "step": 2950
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.506666666666667e-06,
      "loss": 0.0419,
      "step": 2975
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.451111111111112e-06,
      "loss": 0.0435,
      "step": 3000
    },
    {
      "epoch": 1.99,
      "eval_cer": 11.646329777039545,
      "eval_loss": 0.2727377712726593,
      "eval_runtime": 1258.956,
      "eval_samples_per_second": 4.441,
      "eval_steps_per_second": 0.555,
      "step": 3000
    },
    {
      "epoch": 2.01,
      "learning_rate": 4.395555555555556e-06,
      "loss": 0.0293,
      "step": 3025
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.34e-06,
      "loss": 0.0234,
      "step": 3050
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.284444444444445e-06,
      "loss": 0.0246,
      "step": 3075
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.228888888888889e-06,
      "loss": 0.0241,
      "step": 3100
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.173333333333334e-06,
      "loss": 0.0244,
      "step": 3125
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.117777777777779e-06,
      "loss": 0.0239,
      "step": 3150
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.062222222222223e-06,
      "loss": 0.0237,
      "step": 3175
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.006666666666667e-06,
      "loss": 0.0229,
      "step": 3200
    },
    {
      "epoch": 2.14,
      "learning_rate": 3.951111111111112e-06,
      "loss": 0.0228,
      "step": 3225
    },
    {
      "epoch": 2.16,
      "learning_rate": 3.895555555555556e-06,
      "loss": 0.0232,
      "step": 3250
    },
    {
      "epoch": 2.18,
      "learning_rate": 3.8400000000000005e-06,
      "loss": 0.0248,
      "step": 3275
    },
    {
      "epoch": 2.19,
      "learning_rate": 3.784444444444445e-06,
      "loss": 0.0214,
      "step": 3300
    },
    {
      "epoch": 2.21,
      "learning_rate": 3.728888888888889e-06,
      "loss": 0.0233,
      "step": 3325
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.673333333333334e-06,
      "loss": 0.0231,
      "step": 3350
    },
    {
      "epoch": 2.24,
      "learning_rate": 3.617777777777778e-06,
      "loss": 0.0228,
      "step": 3375
    },
    {
      "epoch": 2.26,
      "learning_rate": 3.5622222222222224e-06,
      "loss": 0.0236,
      "step": 3400
    },
    {
      "epoch": 2.28,
      "learning_rate": 3.5066666666666673e-06,
      "loss": 0.0234,
      "step": 3425
    },
    {
      "epoch": 2.29,
      "learning_rate": 3.4511111111111113e-06,
      "loss": 0.0233,
      "step": 3450
    },
    {
      "epoch": 2.31,
      "learning_rate": 3.3955555555555558e-06,
      "loss": 0.0231,
      "step": 3475
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.3400000000000006e-06,
      "loss": 0.023,
      "step": 3500
    },
    {
      "epoch": 2.34,
      "learning_rate": 3.2844444444444447e-06,
      "loss": 0.023,
      "step": 3525
    },
    {
      "epoch": 2.36,
      "learning_rate": 3.228888888888889e-06,
      "loss": 0.0218,
      "step": 3550
    },
    {
      "epoch": 2.38,
      "learning_rate": 3.173333333333334e-06,
      "loss": 0.0212,
      "step": 3575
    },
    {
      "epoch": 2.39,
      "learning_rate": 3.117777777777778e-06,
      "loss": 0.0227,
      "step": 3600
    },
    {
      "epoch": 2.41,
      "learning_rate": 3.0622222222222225e-06,
      "loss": 0.0217,
      "step": 3625
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.0066666666666674e-06,
      "loss": 0.0225,
      "step": 3650
    },
    {
      "epoch": 2.44,
      "learning_rate": 2.9511111111111114e-06,
      "loss": 0.0196,
      "step": 3675
    },
    {
      "epoch": 2.46,
      "learning_rate": 2.895555555555556e-06,
      "loss": 0.0225,
      "step": 3700
    },
    {
      "epoch": 2.48,
      "learning_rate": 2.84e-06,
      "loss": 0.024,
      "step": 3725
    },
    {
      "epoch": 2.49,
      "learning_rate": 2.784444444444445e-06,
      "loss": 0.0221,
      "step": 3750
    },
    {
      "epoch": 2.51,
      "learning_rate": 2.7288888888888893e-06,
      "loss": 0.0202,
      "step": 3775
    },
    {
      "epoch": 2.53,
      "learning_rate": 2.6733333333333333e-06,
      "loss": 0.0217,
      "step": 3800
    },
    {
      "epoch": 2.54,
      "learning_rate": 2.617777777777778e-06,
      "loss": 0.0199,
      "step": 3825
    },
    {
      "epoch": 2.56,
      "learning_rate": 2.5622222222222226e-06,
      "loss": 0.023,
      "step": 3850
    },
    {
      "epoch": 2.58,
      "learning_rate": 2.5066666666666667e-06,
      "loss": 0.0236,
      "step": 3875
    },
    {
      "epoch": 2.59,
      "learning_rate": 2.451111111111111e-06,
      "loss": 0.02,
      "step": 3900
    },
    {
      "epoch": 2.61,
      "learning_rate": 2.3955555555555556e-06,
      "loss": 0.0225,
      "step": 3925
    },
    {
      "epoch": 2.63,
      "learning_rate": 2.3400000000000005e-06,
      "loss": 0.0212,
      "step": 3950
    },
    {
      "epoch": 2.64,
      "learning_rate": 2.2844444444444445e-06,
      "loss": 0.0208,
      "step": 3975
    },
    {
      "epoch": 2.66,
      "learning_rate": 2.228888888888889e-06,
      "loss": 0.0214,
      "step": 4000
    },
    {
      "epoch": 2.66,
      "eval_cer": 11.759726569986583,
      "eval_loss": 0.27413761615753174,
      "eval_runtime": 1247.8057,
      "eval_samples_per_second": 4.481,
      "eval_steps_per_second": 0.56,
      "step": 4000
    },
    {
      "epoch": 2.68,
      "learning_rate": 2.1733333333333334e-06,
      "loss": 0.0203,
      "step": 4025
    },
    {
      "epoch": 2.69,
      "learning_rate": 2.117777777777778e-06,
      "loss": 0.0198,
      "step": 4050
    },
    {
      "epoch": 2.71,
      "learning_rate": 2.0622222222222223e-06,
      "loss": 0.0204,
      "step": 4075
    },
    {
      "epoch": 2.73,
      "learning_rate": 2.006666666666667e-06,
      "loss": 0.0194,
      "step": 4100
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.9511111111111113e-06,
      "loss": 0.021,
      "step": 4125
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.8955555555555557e-06,
      "loss": 0.0198,
      "step": 4150
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.8400000000000002e-06,
      "loss": 0.021,
      "step": 4175
    },
    {
      "epoch": 2.79,
      "learning_rate": 1.7844444444444444e-06,
      "loss": 0.0203,
      "step": 4200
    },
    {
      "epoch": 2.81,
      "learning_rate": 1.728888888888889e-06,
      "loss": 0.0225,
      "step": 4225
    },
    {
      "epoch": 2.83,
      "learning_rate": 1.6733333333333335e-06,
      "loss": 0.0198,
      "step": 4250
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.6177777777777778e-06,
      "loss": 0.0214,
      "step": 4275
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.5622222222222225e-06,
      "loss": 0.0209,
      "step": 4300
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.506666666666667e-06,
      "loss": 0.0187,
      "step": 4325
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.4511111111111112e-06,
      "loss": 0.0188,
      "step": 4350
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.3955555555555556e-06,
      "loss": 0.0196,
      "step": 4375
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.34e-06,
      "loss": 0.0209,
      "step": 4400
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.2844444444444445e-06,
      "loss": 0.0206,
      "step": 4425
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.228888888888889e-06,
      "loss": 0.0166,
      "step": 4450
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.1733333333333335e-06,
      "loss": 0.0195,
      "step": 4475
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.117777777777778e-06,
      "loss": 0.0194,
      "step": 4500
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.0622222222222222e-06,
      "loss": 0.0166,
      "step": 4525
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.0066666666666668e-06,
      "loss": 0.0154,
      "step": 4550
    },
    {
      "epoch": 3.04,
      "learning_rate": 9.511111111111111e-07,
      "loss": 0.0132,
      "step": 4575
    },
    {
      "epoch": 3.06,
      "learning_rate": 8.955555555555557e-07,
      "loss": 0.0132,
      "step": 4600
    },
    {
      "epoch": 3.08,
      "learning_rate": 8.400000000000001e-07,
      "loss": 0.0146,
      "step": 4625
    },
    {
      "epoch": 3.09,
      "learning_rate": 7.844444444444445e-07,
      "loss": 0.0132,
      "step": 4650
    },
    {
      "epoch": 3.11,
      "learning_rate": 7.28888888888889e-07,
      "loss": 0.0128,
      "step": 4675
    },
    {
      "epoch": 3.12,
      "learning_rate": 6.733333333333334e-07,
      "loss": 0.0111,
      "step": 4700
    },
    {
      "epoch": 3.14,
      "learning_rate": 6.177777777777778e-07,
      "loss": 0.0132,
      "step": 4725
    },
    {
      "epoch": 3.16,
      "learning_rate": 5.622222222222223e-07,
      "loss": 0.0132,
      "step": 4750
    },
    {
      "epoch": 3.17,
      "learning_rate": 5.066666666666667e-07,
      "loss": 0.0132,
      "step": 4775
    },
    {
      "epoch": 3.19,
      "learning_rate": 4.511111111111111e-07,
      "loss": 0.0132,
      "step": 4800
    },
    {
      "epoch": 3.21,
      "learning_rate": 3.9555555555555557e-07,
      "loss": 0.0127,
      "step": 4825
    },
    {
      "epoch": 3.22,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.0132,
      "step": 4850
    },
    {
      "epoch": 3.24,
      "learning_rate": 2.844444444444445e-07,
      "loss": 0.0139,
      "step": 4875
    },
    {
      "epoch": 3.26,
      "learning_rate": 2.2888888888888892e-07,
      "loss": 0.0149,
      "step": 4900
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.7333333333333335e-07,
      "loss": 0.0127,
      "step": 4925
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.1777777777777778e-07,
      "loss": 0.0125,
      "step": 4950
    },
    {
      "epoch": 3.31,
      "learning_rate": 6.222222222222223e-08,
      "loss": 0.0141,
      "step": 4975
    },
    {
      "epoch": 3.32,
      "learning_rate": 6.666666666666667e-09,
      "loss": 0.0126,
      "step": 5000
    },
    {
      "epoch": 3.32,
      "eval_cer": 11.462658915223917,
      "eval_loss": 0.2747412323951721,
      "eval_runtime": 1231.0075,
      "eval_samples_per_second": 4.542,
      "eval_steps_per_second": 0.568,
      "step": 5000
    }
  ],
  "max_steps": 5000,
  "num_train_epochs": 4,
  "total_flos": 1.7312699901837312e+20,
  "trial_name": null,
  "trial_params": null
}