best/trainer_state.json

{
  "best_metric": 3.7952540310313356,
  "best_model_checkpoint": "asr_training_small\\checkpoint-700",
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 700,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 3.0882246494293213,
      "learning_rate": 7.52857142857143e-06,
      "loss": 0.1348,
      "step": 175
    },
    {
      "epoch": 1.0,
      "eval_loss": 0.04688960313796997,
      "eval_runtime": 181.7031,
      "eval_samples_per_second": 3.852,
      "eval_steps_per_second": 0.484,
      "eval_wer": 5.126254943717676,
      "step": 175
    },
    {
      "epoch": 2.0,
      "grad_norm": 2.5495264530181885,
      "learning_rate": 5.028571428571429e-06,
      "loss": 0.0305,
      "step": 350
    },
    {
      "epoch": 2.0,
      "eval_loss": 0.03920641914010048,
      "eval_runtime": 179.5113,
      "eval_samples_per_second": 3.899,
      "eval_steps_per_second": 0.49,
      "eval_wer": 3.9473684210526314,
      "step": 350
    },
    {
      "epoch": 3.0,
      "grad_norm": 1.2743825912475586,
      "learning_rate": 2.5285714285714287e-06,
      "loss": 0.0185,
      "step": 525
    },
    {
      "epoch": 3.0,
      "eval_loss": 0.039620302617549896,
      "eval_runtime": 180.9617,
      "eval_samples_per_second": 3.868,
      "eval_steps_per_second": 0.486,
      "eval_wer": 3.8332826285366597,
      "step": 525
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.5467843413352966,
      "learning_rate": 2.8571428571428575e-08,
      "loss": 0.0112,
      "step": 700
    },
    {
      "epoch": 4.0,
      "eval_loss": 0.04179147258400917,
      "eval_runtime": 185.7953,
      "eval_samples_per_second": 3.768,
      "eval_steps_per_second": 0.474,
      "eval_wer": 3.7952540310313356,
      "step": 700
    }
  ],
  "logging_steps": 500,
  "max_steps": 700,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 7.26431956992e+17,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}