trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.8,
  "eval_steps": 500,
  "global_step": 50,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4,
      "grad_norm": 16.500776290893555,
      "learning_rate": 4.600000000000001e-05,
      "loss": 2.5652,
      "step": 25
    },
    {
      "epoch": 0.8,
      "grad_norm": 16.864194869995117,
      "learning_rate": 9.200000000000001e-05,
      "loss": 0.7217,
      "step": 50
    }
  ],
  "logging_steps": 25,
  "max_steps": 62,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 25,
  "total_flos": 4002569985684864.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}