trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.989944134078212,
  "eval_steps": 500,
  "global_step": 669,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 0.23773373663425446,
      "learning_rate": 0.00016312292358803987,
      "loss": 0.6651,
      "step": 178
    },
    {
      "epoch": 1.59,
      "grad_norm": 0.2887946367263794,
      "learning_rate": 0.00010398671096345516,
      "loss": 0.5152,
      "step": 356
    },
    {
      "epoch": 2.39,
      "grad_norm": 0.36021459102630615,
      "learning_rate": 4.485049833887043e-05,
      "loss": 0.4712,
      "step": 534
    }
  ],
  "logging_steps": 178,
  "max_steps": 669,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 4.355563999891292e+17,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}