checkpoint-4/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 4,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": 11.620375633239746,
      "learning_rate": 0.0002,
      "loss": 2.5372,
      "step": 1
    },
    {
      "epoch": 2.0,
      "grad_norm": 11.902047157287598,
      "learning_rate": 0.00013333333333333334,
      "loss": 2.5372,
      "step": 2
    },
    {
      "epoch": 3.0,
      "grad_norm": 2.652064085006714,
      "learning_rate": 6.666666666666667e-05,
      "loss": 2.1328,
      "step": 3
    },
    {
      "epoch": 4.0,
      "grad_norm": 2.411299705505371,
      "learning_rate": 0.0,
      "loss": 1.9934,
      "step": 4
    }
  ],
  "logging_steps": 1,
  "max_steps": 4,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "total_flos": 699676344975360.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}