trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.565217391304348,
  "eval_steps": 500,
  "global_step": 48,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.11594202898550725,
      "grad_norm": 1.375,
      "learning_rate": 4e-05,
      "loss": 2.0688,
      "step": 1
    },
    {
      "epoch": 5.565217391304348,
      "step": 48,
      "total_flos": 7.68376789229568e+16,
      "train_loss": 0.8570354779561361,
      "train_runtime": 1420.4534,
      "train_samples_per_second": 4.663,
      "train_steps_per_second": 0.034
    }
  ],
  "logging_steps": 50,
  "max_steps": 48,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 1000,
  "total_flos": 7.68376789229568e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}