trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 1250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "grad_norm": 29.892879486083984,
      "learning_rate": 3.02e-05,
      "loss": 0.5103,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8775,
      "eval_loss": 0.35183537006378174,
      "eval_runtime": 38.4544,
      "eval_samples_per_second": 20.804,
      "eval_steps_per_second": 2.6,
      "step": 625
    },
    {
      "epoch": 1.6,
      "grad_norm": 0.2147139459848404,
      "learning_rate": 1.024e-05,
      "loss": 0.3424,
      "step": 1000
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.89875,
      "eval_loss": 0.395803302526474,
      "eval_runtime": 38.3851,
      "eval_samples_per_second": 20.841,
      "eval_steps_per_second": 2.605,
      "step": 1250
    }
  ],
  "logging_steps": 500,
  "max_steps": 1250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 408695125428672.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}