trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1224,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.23,
      "learning_rate": 2.957516339869281e-05,
      "loss": 1.6424,
      "step": 500
    },
    {
      "epoch": 2.45,
      "learning_rate": 9.150326797385621e-06,
      "loss": 1.1127,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1224,
      "total_flos": 1886383299379200.0,
      "train_loss": 1.2895471036823747,
      "train_runtime": 330.7231,
      "train_samples_per_second": 36.964,
      "train_steps_per_second": 3.701
    }
  ],
  "max_steps": 1224,
  "num_train_epochs": 3,
  "total_flos": 1886383299379200.0,
  "trial_name": null,
  "trial_params": null
}