trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.99896229678312,
  "eval_steps": 500,
  "global_step": 722,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.69,
      "grad_norm": 1.6492760181427002,
      "learning_rate": 3.074792243767313e-05,
      "loss": 2.393,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_loss": 2.211937189102173,
      "eval_runtime": 268.8732,
      "eval_samples_per_second": 4.779,
      "eval_steps_per_second": 4.779,
      "step": 722
    },
    {
      "epoch": 1.0,
      "step": 722,
      "total_flos": 6.586350857040691e+16,
      "train_loss": 2.345679919805553,
      "train_runtime": 10097.1443,
      "train_samples_per_second": 1.145,
      "train_steps_per_second": 0.072
    }
  ],
  "logging_steps": 500,
  "max_steps": 722,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 16,
  "total_flos": 6.586350857040691e+16,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}