Views
No views yet
1from huggingface_hub import hf_hub_download
2import xgboost as xgb
3
4REPO_ID = "RAIA-BRASIL/bonecheck_xgboost"
5FILENAME = "xgb_model.json"
6
7# baixar o arquivo do modelo do Hugging Face Hub
8model_path = hf_hub_download(repo_id=REPO_ID, filename=FILENAME)
9
10model = xgb.XGBClassifier()
11model.load_model(model_path)1 "best_params": {
2 "colsample_bytree": 0.85,
3 "gamma": 1,
4 "learning_rate": 0.02,
5 "max_depth": 2,
6 "min_child_weight": 1,
7 "n_estimators": 25,
8 "subsample": 0.7
9 }1 "best_params" : {
2 "lr": 1e-4,
3 "batch_size": 64,
4 "optimizer": "Adam", # Adam para CNNs | AdamW para ViTs
5 "n_epochs": 100,
6 }| Modelo | Acurácia | Precisão | Recall | F1-Score |
|---|---|---|---|---|
| ConvNeXT | 0.688 | 0.651 | 0.631 | 0.638 |
| EfficientNet | 0.558 | 0.608 | 0.605 | 0.512 |
| DeiT | 0.688 | 0.648 | 0.635 | 0.639 |
| Swin Transformer | 0.632 | 0.591 | 0.622 | 0.590 |
| Ensemble (XGBoost) | 0.697 | 0.712 | 0.622 | 0.838 |