trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 20.0,
  "global_step": 1280,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 7.81,
      "learning_rate": 3.0468750000000002e-05,
      "loss": 0.3051,
      "step": 500
    },
    {
      "epoch": 15.62,
      "learning_rate": 1.09375e-05,
      "loss": 0.0211,
      "step": 1000
    },
    {
      "epoch": 20.0,
      "step": 1280,
      "total_flos": 13845085104291840,
      "train_runtime": 963.3998,
      "train_samples_per_second": 1.329
    }
  ],
  "max_steps": 1280,
  "num_train_epochs": 20,
  "total_flos": 13845085104291840,
  "trial_name": null,
  "trial_params": null
}