Update src/predict.py

src/predict.py

@@ -268,49 +268,32 @@ def validate_and_fix_config(model_path: str) -> bool:
 # Enhanced model loading with multiple fallback strategies
 # =======================
 def load_model_with_fallback(model_name: str) -> bool:
-    """Simplified model loading for HF Spaces"""
+    """Use standard model - bypass custom model for now"""
     global model, tokenizer, model_loaded
     
     with model_lock:
         if model_loaded:
             return True
         
-        logger.info(f"Loading model: {model_name}")
+        logger.info("Using standard DistilBERT model (custom model has tokenizer issues)")
         
         try:
-            # Check if local model exists
-            if os.path.exists(model_name):
-                logger.info("Loading local model...")
-                tokenizer = AutoTokenizer.from_pretrained(
-                    model_name, 
-                    use_fast=True,
-                    do_lower_case=True,
-                    local_files_only=True  # Force local loading
-                )
-                model = AutoModelForSequenceClassification.from_pretrained(
-                    model_name,
-                    num_labels=2,
-                    ignore_mismatched_sizes=True,
-                    local_files_only=True  # Force local loading
-                )
-            else:
-                # Fallback to a working pre-trained model
-                logger.warning("Local model not found, using fallback...")
-                tokenizer = AutoTokenizer.from_pretrained('distilbert-base-uncased')
-                model = AutoModelForSequenceClassification.from_pretrained(
-                    'distilbert-base-uncased',
-                    num_labels=2
-                )
+            tokenizer = AutoTokenizer.from_pretrained('distilbert-base-uncased')
+            model = AutoModelForSequenceClassification.from_pretrained(
+                'distilbert-base-uncased',
+                num_labels=2
+            )
             
             model.to(device)
             model.eval()
             model_loaded = True
-            logger.info("Model loaded successfully")
+            logger.info("Standard model loaded successfully")
             return True
             
         except Exception as e:
             logger.error(f"Model loading failed: {e}")
             return False
+            
 # =======================
 # Optimized prediction function with enhanced accuracy
 # =======================