trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9836065573770492,
  "eval_steps": 500,
  "global_step": 30,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.82,
      "grad_norm": 0.7344030737876892,
      "learning_rate": 0.0002,
      "loss": 2.2051,
      "step": 25
    },
    {
      "epoch": 0.98,
      "step": 30,
      "total_flos": 1867634191761408.0,
      "train_loss": 1.9350372393925985,
      "train_runtime": 503.4351,
      "train_samples_per_second": 0.483,
      "train_steps_per_second": 0.06
    }
  ],
  "logging_steps": 25,
  "max_steps": 30,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 25,
  "total_flos": 1867634191761408.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}