trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 1045,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.39,
      "learning_rate": 1.5645933014354067e-05,
      "loss": 0.0684,
      "step": 500
    },
    {
      "epoch": 4.78,
      "learning_rate": 1.291866028708134e-06,
      "loss": 0.0082,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1045,
      "total_flos": 2240667629302440.0,
      "train_loss": 0.03677053214830645,
      "train_runtime": 220.3164,
      "train_samples_per_second": 75.55,
      "train_steps_per_second": 4.743
    }
  ],
  "max_steps": 1045,
  "num_train_epochs": 5,
  "total_flos": 2240667629302440.0,
  "trial_name": null,
  "trial_params": null
}