trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "eval_steps": 500,
  "global_step": 790,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 6.33,
      "grad_norm": 0.20978006720542908,
      "learning_rate": 0.0003670886075949367,
      "loss": 0.4073,
      "step": 500
    },
    {
      "epoch": 6.33,
      "eval_loss": 0.1755284070968628,
      "eval_runtime": 1.8212,
      "eval_samples_per_second": 4314.295,
      "eval_steps_per_second": 34.044,
      "step": 500
    },
    {
      "epoch": 10.0,
      "step": 790,
      "total_flos": 3267517032169472.0,
      "train_loss": 0.35002156511137755,
      "train_runtime": 70.584,
      "train_samples_per_second": 1416.752,
      "train_steps_per_second": 11.192
    }
  ],
  "logging_steps": 500,
  "max_steps": 790,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 500,
  "total_flos": 3267517032169472.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}