{
  "model": "google/bert_uncased_L-2_H-256_A-4",
  "best_optuna_trial": 39,
  "optuna_f1_macro": 0.9358712965756903,
  "hyperparams": {
    "model": "google/bert_uncased_L-2_H-256_A-4",
    "lr": 4.7785196153634585e-05,
    "batch_size": 8,
    "scheduler": "cosine",
    "warmup_ratio": 0.1268629742173647,
    "weight_decay": 0.2761538415856171,
    "num_epochs": 15,
    "classifier_dropout": 0.1704209352408175,
    "label_smoothing": 0.008655451448503527,
    "use_fgm": true,
    "fgm_epsilon": 0.9684709844447323
  },
  "candidates": [
    {
      "trial": 60,
      "model": "google/bert_uncased_L-2_H-256_A-4",
      "optuna_f1": 0.9362,
      "test_f1_macro": 0.9345
    },
    {
      "trial": 39,
      "model": "google/bert_uncased_L-2_H-256_A-4",
      "optuna_f1": 0.9359,
      "test_f1_macro": 0.9374
    },
    {
      "trial": 38,
      "model": "google/bert_uncased_L-2_H-256_A-4",
      "optuna_f1": 0.9354,
      "test_f1_macro": 0.8928
    }
  ],
  "final_test_f1_macro": 0.9374,
  "cpu_inference_ms": 17.28,
  "threshold": {
    "optimal_threshold": 0.2,
    "f2_score": 0.9449,
    "precision": 0.8276,
    "recall": 0.9796
  },
  "training_tricks": {
    "loss_type": "ce_weighted",
    "label_smoothing": 0.008655451448503527,
    "fgm": true,
    "fgm_epsilon": 0.9684709844447323,
    "classifier_dropout": 0.1704209352408175,
    "stratified_batch_sampling": true,
    "token_masking": 0.2,
    "token_swapping": 0.15
  }
}