trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 1440,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.74,
      "learning_rate": 3.263888888888889e-05,
      "loss": 0.8294,
      "step": 500
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.527777777777778e-05,
      "loss": 0.0362,
      "step": 1000
    },
    {
      "epoch": 5.0,
      "step": 1440,
      "total_flos": 38943546015744.0,
      "train_loss": 0.3061259587605794,
      "train_runtime": 100.4467,
      "train_samples_per_second": 57.244,
      "train_steps_per_second": 14.336
    }
  ],
  "max_steps": 1440,
  "num_train_epochs": 5,
  "total_flos": 38943546015744.0,
  "trial_name": null,
  "trial_params": null
}