trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 16296,
  "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": 1.6534,
      "step": 8148
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0,
      "loss": 1.1522,
      "step": 16296
    },
    {
      "epoch": 2.0,
      "step": 16296,
      "total_flos": 2.0025974242197504e+16,
      "train_loss": 1.4028112367490488,
      "train_runtime": 6757.699,
      "train_samples_per_second": 4.823,
      "train_steps_per_second": 2.411
    }
  ],
  "max_steps": 16296,
  "num_train_epochs": 2,
  "total_flos": 2.0025974242197504e+16,
  "trial_name": null,
  "trial_params": null
}