trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 1136,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.76,
      "learning_rate": 2.7992957746478874e-05,
      "loss": 0.672,
      "step": 500
    },
    {
      "epoch": 3.52,
      "learning_rate": 5.9859154929577465e-06,
      "loss": 0.2431,
      "step": 1000
    },
    {
      "epoch": 4.0,
      "step": 1136,
      "total_flos": 60704291487744.0,
      "train_loss": 0.42219568642092425,
      "train_runtime": 444.8984,
      "train_samples_per_second": 25.507,
      "train_steps_per_second": 2.553
    }
  ],
  "max_steps": 1136,
  "num_train_epochs": 4,
  "total_flos": 60704291487744.0,
  "trial_name": null,
  "trial_params": null
}