last-checkpoint/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.33783783783783783,
  "eval_steps": 100,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.34,
      "grad_norm": NaN,
      "learning_rate": 0.0001729,
      "loss": 6.329,
      "step": 100
    },
    {
      "epoch": 0.34,
      "eval_loss": Infinity,
      "eval_runtime": 77.9869,
      "eval_samples_per_second": 15.015,
      "eval_steps_per_second": 1.885,
      "eval_wer": 1.0,
      "step": 100
    }
  ],
  "logging_steps": 100,
  "max_steps": 8880,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 30,
  "save_steps": 100,
  "total_flos": 1.2267819282523704e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}