trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 888,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.69,
      "learning_rate": 2.1846846846846845e-05,
      "loss": 0.9229,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 888,
      "total_flos": 8446018081259520.0,
      "train_loss": 0.791101954004786,
      "train_runtime": 1264.5393,
      "train_samples_per_second": 1.404,
      "train_steps_per_second": 0.702
    }
  ],
  "max_steps": 888,
  "num_train_epochs": 3,
  "total_flos": 8446018081259520.0,
  "trial_name": null,
  "trial_params": null
}