trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 10,
  "global_step": 5,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "grad_norm": NaN,
      "learning_rate": 0.0,
      "loss": 1.625,
      "step": 1
    },
    {
      "epoch": 2.0,
      "grad_norm": NaN,
      "learning_rate": 0.0003,
      "loss": 0.0,
      "step": 2
    },
    {
      "epoch": 3.0,
      "grad_norm": NaN,
      "learning_rate": 0.0003,
      "loss": 0.0,
      "step": 3
    },
    {
      "epoch": 4.0,
      "grad_norm": NaN,
      "learning_rate": 0.0003,
      "loss": 0.0,
      "step": 4
    },
    {
      "epoch": 5.0,
      "grad_norm": NaN,
      "learning_rate": 0.0003,
      "loss": 0.0,
      "step": 5
    },
    {
      "epoch": 5.0,
      "step": 5,
      "total_flos": 2404642160640.0,
      "train_loss": 0.325,
      "train_runtime": 84.0754,
      "train_samples_per_second": 0.059,
      "train_steps_per_second": 0.059
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 5,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 2404642160640.0,
  "train_batch_size": 2,
  "trial_name": null,
  "trial_params": null
}