Update training script and model files

services/__init__.py

@@ -1,7 +1,7 @@
 """Services for business logic and external integrations"""
 
 from .stance_model_manager import StanceModelManager, stance_model_manager
-from .label_model_manage import KpaModelManager, kpa_model_manager
+from .label_model_manager import KpaModelManager, label_model_manager
 
 __all__ = [
     "StanceModelManager",

services/{label_model_manage.py → label_model_manager.py}

@@ -26,6 +26,15 @@ class KpaModelManager:
             return
 
         try:
+            # Debug: Vérifier les paramètres d'entrée
+            logger.info(f"=== DEBUG KPA MODEL LOADING ===")
+            logger.info(f"model_id reçu: {model_id}")
+            logger.info(f"model_id type: {type(model_id)}")
+            logger.info(f"api_key présent: {api_key is not None}")
+            
+            if model_id is None:
+                raise ValueError("model_id cannot be None - check your .env file")
+                
             logger.info(f"Loading KPA model from Hugging Face: {model_id}")
             
             # Determine device
@@ -39,27 +48,34 @@ class KpaModelManager:
             token = api_key if api_key else None
             
             # Load tokenizer and model directly from Hugging Face
-            logger.info("Loading tokenizer...")
+            logger.info("Step 1: Loading tokenizer...")
             self.tokenizer = AutoTokenizer.from_pretrained(
                 model_id,
                 token=token,
                 trust_remote_code=True
             )
+            logger.info("✓ Tokenizer loaded successfully")
             
-            logger.info("Loading model...")
+            logger.info("Step 2: Loading model...")
             self.model = AutoModelForSequenceClassification.from_pretrained(
                 model_id,
                 token=token,
                 trust_remote_code=True
             )
+            logger.info("✓ Model architecture loaded")
+            
             self.model.to(self.device)
             self.model.eval()
+            logger.info("✓ Model moved to device and set to eval mode")
 
             self.model_loaded = True
             logger.info("✓ KPA model loaded successfully from Hugging Face!")
+            logger.info(f"=== KPA MODEL LOADING COMPLETE ===")
 
         except Exception as e:
-            logger.error(f"Error loading KPA model: {str(e)}")
+            logger.error(f"❌ Error loading KPA model: {str(e)}")
+            logger.error(f"❌ Model ID was: {model_id}")
+            logger.error(f"❌ API Key present: {api_key is not None}")
             raise RuntimeError(f"Failed to load KPA model: {str(e)}")
 
     def predict(self, argument: str, key_point: str) -> dict: