trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 126,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 2.5e-05,
      "loss": 0.7,
      "step": 63
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0,
      "loss": 0.2997,
      "step": 126
    },
    {
      "epoch": 2.0,
      "step": 126,
      "total_flos": 522636036096000.0,
      "train_loss": 0.4998847870599656,
      "train_runtime": 72.6531,
      "train_samples_per_second": 27.528,
      "train_steps_per_second": 1.734
    }
  ],
  "max_steps": 126,
  "num_train_epochs": 2,
  "total_flos": 522636036096000.0,
  "trial_name": null,
  "trial_params": null
}