checkpoint-200/trainer_state.json

{
  "best_metric": 0.6154423356056213,
  "best_model_checkpoint": "/scratch/hlv8980/Attack_Benchmark/models/hyena/0/origin/checkpoint-200",
  "epoch": 1.9607843137254903,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.98,
      "learning_rate": 2.597765363128492e-05,
      "loss": 0.6876,
      "step": 100
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.776536312849162e-05,
      "loss": 0.6084,
      "step": 200
    },
    {
      "epoch": 1.96,
      "eval_accuracy": 0.662962962962963,
      "eval_f1": 0.6490997002438991,
      "eval_loss": 0.6154423356056213,
      "eval_matthews_correlation": 0.3414355217691314,
      "eval_precision": 0.6839276610207896,
      "eval_recall": 0.6584566112551471,
      "eval_runtime": 0.0868,
      "eval_samples_per_second": 9326.637,
      "eval_steps_per_second": 149.687,
      "step": 200
    }
  ],
  "max_steps": 408,
  "num_train_epochs": 4,
  "total_flos": 25546897152000.0,
  "trial_name": null,
  "trial_params": null
}