trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9536423841059603,
  "eval_steps": 500,
  "global_step": 18,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.26,
      "grad_norm": 0.8373088240623474,
      "learning_rate": 0.00016427876096865394,
      "loss": 0.6603,
      "step": 5
    },
    {
      "epoch": 0.53,
      "grad_norm": 1.0933210849761963,
      "learning_rate": 8.263518223330697e-05,
      "loss": 0.3484,
      "step": 10
    },
    {
      "epoch": 0.79,
      "grad_norm": 0.969545841217041,
      "learning_rate": 1.339745962155613e-05,
      "loss": 0.2464,
      "step": 15
    },
    {
      "epoch": 0.95,
      "step": 18,
      "total_flos": 7430501711216640.0,
      "train_loss": 0.38846387134657967,
      "train_runtime": 487.2698,
      "train_samples_per_second": 4.956,
      "train_steps_per_second": 0.037
    }
  ],
  "logging_steps": 5,
  "max_steps": 18,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 100,
  "total_flos": 7430501711216640.0,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}