Upload folder using huggingface_hub

README.md

@@ -30,60 +30,37 @@ Evaluated via **5-fold document-level cross-validation** on 80 Turkish banking c
 | Micro-F1 (CV) | **0.6657** |
 | Macro-F1 (CV) | **0.6226** |
 
-### Per-Class F1
+### Per-class F1 (5-fold CV)
 
 | Category | F1 |
 |----------|----|
 | hidden_fees | 0.76 |
+| dispute_limitation | 0.61 |
 | broad_collateral | 0.74 |
-| unilateral_rate_change | 0.72 |
 | default_escalation | 0.71 |
-| data_sharing | 0.69 |
-| currency_risk | 0.68 |
 | account_freeze | 0.65 |
-| early_payment_penalty | 0.64 |
-| dispute_limitation | 0.61 |
-| bundled_insurance | 0.59 |
+| currency_risk | 0.68 |
 | unilateral_terms_change | 0.57 |
-| overdraft_penalty | 0.51 |
-| cross_default | 0.44 |
+| unilateral_rate_change | 0.72 |
+| data_sharing | 0.69 |
 | auto_enrollment | 0.42 |
+| cross_default | 0.44 |
+| overdraft_penalty | 0.51 |
+| early_payment_penalty | 0.64 |
+| bundled_insurance | 0.59 |
 
 ## Training
 
-- **Base:** dbmdz/bert-base-turkish-cased
+- **Base model:** dbmdz/bert-base-turkish-cased
 - **Loss:** Multi-label Focal Loss (gamma=2.0, alpha=0.75)
-- **Optimizer:** AdamW, lr=2e-5, weight_decay=0.01, epochs=10
-- **Data:** 80 Turkish banking contracts, 7,020 clauses, 18 banks
+- **Optimizer:** AdamW, lr=2e-5, weight_decay=0.01
+- **Epochs:** 10
+- **Training data:** 80 Turkish banking contract PDFs, 7,020 clauses (Akbank, Garanti, İşbank, DenizBank, Halkbank, VakıfBank, QNB, TEB, YapıKredi, Ziraat, ING, KuveytTürk, Şekerbank, HSBC Turkey, Odea Bank, Albaraka Türk, Türkiye Finans, Burgan Bank and more)
 - **Threshold:** Fixed 0.5 per label
 
 ## Usage
 
-```python
-from transformers import AutoTokenizer, AutoModelForSequenceClassification
-import torch
-
-LABELS = [
-    "hidden_fees", "dispute_limitation", "broad_collateral", "default_escalation",
-    "account_freeze", "currency_risk", "unilateral_terms_change", "unilateral_rate_change",
-    "data_sharing", "auto_enrollment", "cross_default", "overdraft_penalty",
-    "early_payment_penalty", "bundled_insurance",
-]
-
-tokenizer = AutoTokenizer.from_pretrained("Agreemind/banking-bert-turkish")
-model = AutoModelForSequenceClassification.from_pretrained("Agreemind/banking-bert-turkish")
-model.eval()
-
-text = "Banka, hesap işletim ücretini önceden bildirmeksizin değiştirme hakkını saklı tutar."
-inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512)
-
-with torch.no_grad():
-    probs = torch.sigmoid(model(**inputs).logits).squeeze()
 
-for label, prob in zip(LABELS, probs):
-    if prob > 0.5:
-        print(f"{label}: {prob:.3f}")
-```
 
 ## License
 MIT