trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.304932735426009,
  "eval_steps": 10000,
  "global_step": 1200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.43,
      "grad_norm": 2.023824691772461,
      "learning_rate": 1.6736401673640167e-05,
      "loss": 1.428,
      "step": 400
    },
    {
      "epoch": 2.87,
      "grad_norm": 2.9375298023223877,
      "learning_rate": 8.368200836820084e-06,
      "loss": 0.9062,
      "step": 800
    },
    {
      "epoch": 4.3,
      "grad_norm": 3.3535144329071045,
      "learning_rate": 0.0,
      "loss": 0.7824,
      "step": 1200
    }
  ],
  "logging_steps": 400,
  "max_steps": 1200,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 400,
  "total_flos": 2.351787835392e+16,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}