trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 21.05263157894737,
  "global_step": 800,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 10.53,
      "learning_rate": 0.00023999999999999998,
      "loss": 6.4582,
      "step": 400
    },
    {
      "epoch": 10.53,
      "eval_loss": 2.820686101913452,
      "eval_runtime": 541.4545,
      "eval_samples_per_second": 4.041,
      "eval_wer": 0.9957617789520036,
      "step": 400
    },
    {
      "epoch": 21.05,
      "learning_rate": 0.00015937499999999998,
      "loss": 0.692,
      "step": 800
    },
    {
      "epoch": 21.05,
      "eval_loss": 0.5226749777793884,
      "eval_runtime": 547.3066,
      "eval_samples_per_second": 3.998,
      "eval_wer": 0.41655658300308235,
      "step": 800
    }
  ],
  "max_steps": 1140,
  "num_train_epochs": 30,
  "total_flos": 1.379922511881828e+19,
  "trial_name": null,
  "trial_params": null
}