t1/checkpoint-100000/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.7773117271471035,
  "eval_steps": 50000,
  "global_step": 100000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.89,
      "learning_rate": 3.45e-05,
      "loss": 4996926.0134,
      "step": 50000
    },
    {
      "epoch": 0.89,
      "eval_loss": NaN,
      "eval_runtime": 13.236,
      "eval_samples_per_second": 52.886,
      "eval_steps_per_second": 6.649,
      "eval_wer": 1.0,
      "step": 50000
    },
    {
      "epoch": 1.78,
      "learning_rate": 3.45e-05,
      "loss": 0.0,
      "step": 100000
    },
    {
      "epoch": 1.78,
      "eval_loss": NaN,
      "eval_runtime": 13.3509,
      "eval_samples_per_second": 52.431,
      "eval_steps_per_second": 6.591,
      "eval_wer": 1.0,
      "step": 100000
    }
  ],
  "logging_steps": 50000,
  "max_steps": 562640,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 50000,
  "total_flos": 9.15368204731761e+20,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}