trainer_state.json

{
  "best_metric": 7.174674987792969,
  "best_model_checkpoint": "./human-actions/checkpoint-1000",
  "epoch": 3.0,
  "eval_steps": 1000,
  "global_step": 1182,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.54,
      "learning_rate": 7.698815566835872e-06,
      "loss": 0.3842,
      "step": 1000
    },
    {
      "epoch": 2.54,
      "eval_accuracy": 0.06759259259259259,
      "eval_f1": 0.008441746169413124,
      "eval_loss": 7.174674987792969,
      "eval_runtime": 97.8976,
      "eval_samples_per_second": 55.16,
      "eval_steps_per_second": 6.895,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1182,
      "total_flos": 2.9295385025163264e+18,
      "train_loss": 0.3337646079345806,
      "train_runtime": 1709.0879,
      "train_samples_per_second": 22.117,
      "train_steps_per_second": 0.692
    }
  ],
  "logging_steps": 1000,
  "max_steps": 1182,
  "num_train_epochs": 3,
  "save_steps": 1000,
  "total_flos": 2.9295385025163264e+18,
  "trial_name": null,
  "trial_params": null
}