trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 100.0,
  "eval_steps": 500,
  "global_step": 1200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 41.67,
      "learning_rate": 2.916666666666667e-05,
      "loss": 0.5848,
      "step": 500
    },
    {
      "epoch": 83.33,
      "learning_rate": 8.333333333333334e-06,
      "loss": 0.0094,
      "step": 1000
    },
    {
      "epoch": 100.0,
      "step": 1200,
      "total_flos": 2.93969475072e+16,
      "train_loss": 0.24828063900272052,
      "train_runtime": 819.9397,
      "train_samples_per_second": 365.881,
      "train_steps_per_second": 1.464
    }
  ],
  "logging_steps": 500,
  "max_steps": 1200,
  "num_train_epochs": 100,
  "save_steps": 500,
  "total_flos": 2.93969475072e+16,
  "trial_name": null,
  "trial_params": null
}