trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 8.0,
  "eval_steps": 500,
  "global_step": 1752,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.28,
      "learning_rate": 3.573059360730594e-05,
      "loss": 1.993,
      "step": 500
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.1461187214611872e-05,
      "loss": 1.7625,
      "step": 1000
    },
    {
      "epoch": 6.85,
      "learning_rate": 7.191780821917809e-06,
      "loss": 1.5289,
      "step": 1500
    },
    {
      "epoch": 8.0,
      "step": 1752,
      "total_flos": 803810618658432.0,
      "train_loss": 1.7032232240999126,
      "train_runtime": 7389.9923,
      "train_samples_per_second": 1.897,
      "train_steps_per_second": 0.237
    }
  ],
  "logging_steps": 500,
  "max_steps": 1752,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 8,
  "save_steps": 500,
  "total_flos": 803810618658432.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}