trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 50.0,
  "eval_steps": 500,
  "global_step": 75,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 17.0,
      "grad_norm": 0.10613784939050674,
      "learning_rate": 0.0001,
      "loss": 0.0058,
      "step": 25
    },
    {
      "epoch": 33.6,
      "grad_norm": 0.01143595203757286,
      "learning_rate": 5e-05,
      "loss": 0.0009,
      "step": 50
    },
    {
      "epoch": 50.0,
      "grad_norm": 0.0007798762526363134,
      "learning_rate": 0.0,
      "loss": 0.0,
      "step": 75
    }
  ],
  "logging_steps": 25,
  "max_steps": 75,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 75,
  "save_steps": 25,
  "total_flos": 2934435473376000.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}