trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 1275,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.18,
      "learning_rate": 3.0392156862745097e-05,
      "loss": 0.1665,
      "step": 500
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.0784313725490197e-05,
      "loss": 0.0587,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1275,
      "total_flos": 140113503536760.0,
      "train_loss": 0.09562072828704235,
      "train_runtime": 125.9576,
      "train_samples_per_second": 80.837,
      "train_steps_per_second": 10.122
    }
  ],
  "logging_steps": 500,
  "max_steps": 1275,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 140113503536760.0,
  "trial_name": null,
  "trial_params": null
}