{
  "config": {
    "epochs": 50,
    "lr": 0.0001,
    "warmup_steps": 100,
    "grad_accum_steps": 8,
    "grad_clip": 1.0
  },
  "train": [
    {
      "epoch": 1,
      "avg_loss": 48.950583,
      "lr": 9.99783297632528e-05,
      "step": 192,
      "time_s": 1012.5
    },
    {
      "epoch": 2,
      "avg_loss": 39.822198,
      "lr": 9.979363170168663e-05,
      "step": 384,
      "time_s": 1016.4
    },
    {
      "epoch": 3,
      "avg_loss": 36.55349,
      "lr": 9.942103713803461e-05,
      "step": 576,
      "time_s": 1018.1
    },
    {
      "epoch": 4,
      "avg_loss": 34.752199,
      "lr": 9.88620263818582e-05,
      "step": 768,
      "time_s": 1016.3
    },
    {
      "epoch": 5,
      "avg_loss": 33.489609,
      "lr": 9.811882036992912e-05,
      "step": 960,
      "time_s": 1016.9
    },
    {
      "epoch": 6,
      "avg_loss": 32.44456,
      "lr": 9.719437184249907e-05,
      "step": 1152,
      "time_s": 1020.7
    },
    {
      "epoch": 7,
      "avg_loss": 31.508465,
      "lr": 9.609235361213091e-05,
      "step": 1344,
      "time_s": 1017.7
    },
    {
      "epoch": 8,
      "avg_loss": 30.626262,
      "lr": 9.481714397169986e-05,
      "step": 1536,
      "time_s": 1018.6
    },
    {
      "epoch": 9,
      "avg_loss": 29.740664,
      "lr": 9.337380929953731e-05,
      "step": 1728,
      "time_s": 1017.3
    },
    {
      "epoch": 10,
      "avg_loss": 28.876585,
      "lr": 9.176808393082695e-05,
      "step": 1920,
      "time_s": 1015.7
    },
    {
      "epoch": 11,
      "avg_loss": 28.011303,
      "lr": 9.000634737522459e-05,
      "step": 2112,
      "time_s": 1017.1
    },
    {
      "epoch": 12,
      "avg_loss": 27.240841,
      "lr": 8.80955989712136e-05,
      "step": 2304,
      "time_s": 1021.0
    },
    {
      "epoch": 13,
      "avg_loss": 26.574948,
      "lr": 8.604343007789604e-05,
      "step": 2496,
      "time_s": 1018.9
    },
    {
      "epoch": 14,
      "avg_loss": 25.985939,
      "lr": 8.385799391469841e-05,
      "step": 2688,
      "time_s": 1020.0
    },
    {
      "epoch": 15,
      "avg_loss": 25.478551,
      "lr": 8.154797316882016e-05,
      "step": 2880,
      "time_s": 1020.9
    }
  ],
  "val": [
    {
      "epoch": 1,
      "avg_loss": 42.255684
    },
    {
      "epoch": 2,
      "avg_loss": 37.861168
    },
    {
      "epoch": 3,
      "avg_loss": 35.591921
    },
    {
      "epoch": 4,
      "avg_loss": 34.174768
    },
    {
      "epoch": 5,
      "avg_loss": 33.054291
    },
    {
      "epoch": 6,
      "avg_loss": 32.198608
    },
    {
      "epoch": 7,
      "avg_loss": 31.356775
    },
    {
      "epoch": 8,
      "avg_loss": 30.324559
    },
    {
      "epoch": 9,
      "avg_loss": 29.332707
    },
    {
      "epoch": 10,
      "avg_loss": 28.455519
    },
    {
      "epoch": 11,
      "avg_loss": 27.58525
    },
    {
      "epoch": 12,
      "avg_loss": 26.99554
    },
    {
      "epoch": 13,
      "avg_loss": 26.37473
    },
    {
      "epoch": 14,
      "avg_loss": 25.988197
    },
    {
      "epoch": 15,
      "avg_loss": 25.571769
    }
  ],
  "sanity": [
    {
      "epoch": 5,
      "audio_path": "/kaggle/working/inference_out.wav"
    },
    {
      "epoch": 10,
      "audio_path": "/kaggle/working/inference_out.wav"
    },
    {
      "epoch": 15,
      "audio_path": "/kaggle/working/inference_out.wav"
    }
  ]
}