Upload app.py with huggingface_hub

app.py

@@ -258,25 +258,38 @@ category_descriptions = {
 # List of categories
 categories = list(category_descriptions.keys())
 
-# Try to load the recommended models
+# Try to load the recommended models with specific versions and robust error handling
 try:
-    # 1. Load BART model for zero-shot classification
-    from transformers import pipeline
+    print("Loading classification model...")
+    # 1. Load BART model for zero-shot classification with specific version
+    from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer
     
-    # Initialize the zero-shot classification pipeline
-    classifier = pipeline("zero-shot-classification", model="facebook/bart-base-mnli")
-    print("Using facebook/bart-base-mnli for classification")
+    # Initialize the zero-shot classification pipeline with specific model version
+    # Use a smaller model with explicit version
+    classifier = pipeline(
+        "zero-shot-classification", 
+        model="facebook/bart-base-mnli",
+        framework="pt",  # Explicitly use PyTorch
+        device=-1  # Use CPU
+    )
+    print("Successfully loaded facebook/bart-base-mnli for classification")
     
-    # 2. Load MiniLM model for semantic search
+    print("Loading semantic search model...")
+    # 2. Load MiniLM model for semantic search with specific version
     from sentence_transformers import SentenceTransformer, util
     
-    # Load the sentence transformer model
-    sentence_model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
-    print("Using sentence-transformers/all-MiniLM-L6-v2 for semantic search")
+    # Load the sentence transformer model with explicit version
+    sentence_model = SentenceTransformer(
+        'sentence-transformers/all-MiniLM-L6-v2',
+        device="cpu"  # Explicitly use CPU
+    )
+    print("Successfully loaded sentence-transformers/all-MiniLM-L6-v2 for semantic search")
     
     # Pre-compute embeddings for category descriptions
+    print("Pre-computing category embeddings...")
     category_texts = list(category_descriptions.values())
     category_embeddings = sentence_model.encode(category_texts, convert_to_tensor=True)
+    print("Successfully pre-computed category embeddings")
     
     # Using recommended models
     using_recommended_models = True
@@ -285,20 +298,63 @@ except Exception as e:
     print(f"Error loading recommended models: {str(e)}")
     print("Falling back to local model")
     
-    model_path = os.path.dirname(os.path.abspath(__file__))
-    tokenizer = AutoTokenizer.from_pretrained(model_path)
-    model = AutoModelForSequenceClassification.from_pretrained(model_path)
-    
-    # Load the local categories
     try:
-        with open(os.path.join(model_path, "categories.json"), "r") as f:
-            categories = json.load(f)
+        model_path = os.path.dirname(os.path.abspath(__file__))
+        print(f"Loading local model from {model_path}")
+        tokenizer = AutoTokenizer.from_pretrained(model_path)
+        model = AutoModelForSequenceClassification.from_pretrained(model_path)
+        print("Successfully loaded local model")
+        
+        # Load the local categories
+        try:
+            with open(os.path.join(model_path, "categories.json"), "r") as f:
+                categories = json.load(f)
+            print(f"Loaded {len(categories)} categories from categories.json")
+        except Exception as e:
+            print(f"Error loading categories: {str(e)}")
+            categories = ["electronics", "clothing", "home", "kitchen", "toys", "other"]
+            print(f"Using default categories: {categories}")
+        
+        # Not using recommended models
+        using_recommended_models = False
     except Exception as e:
-        print(f"Error loading categories: {str(e)}")
-        categories = ["electronics", "clothing", "home", "kitchen", "toys", "other"]
-    
-    # Not using recommended models
-    using_recommended_models = False
+        print(f"Error loading local model: {str(e)}")
+        print("Using extremely simplified fallback mode")
+        # Define a simple fallback classifier function
+        def simple_classify(text):
+            keywords = {
+                "electronics": ["electronics", "gadget", "device", "tech", "electronic"],
+                "computers": ["computer", "laptop", "desktop", "pc", "monitor"],
+                "mobile": ["phone", "mobile", "smartphone", "cell", "iphone", "android"],
+                "audio": ["audio", "headphone", "speaker", "earbud", "sound"],
+                "clothing": ["clothing", "clothes", "shirt", "pants", "dress", "wear"],
+                "footwear": ["shoe", "boot", "sneaker", "footwear", "sandal"],
+                "home": ["home", "furniture", "decor", "house", "living"],
+                "kitchen": ["kitchen", "cook", "appliance", "food", "dining"],
+                "toys": ["toy", "game", "play", "kid", "child"],
+                "sports": ["sport", "fitness", "exercise", "workout", "athletic"],
+                "beauty": ["beauty", "makeup", "cosmetic", "skin", "hair"],
+                "books": ["book", "read", "novel", "textbook", "ebook"]
+            }
+            
+            text_lower = text.lower()
+            scores = {}
+            
+            for category, terms in keywords.items():
+                score = 0
+                for term in terms:
+                    if term in text_lower:
+                        score += 1
+                scores[category] = score
+            
+            # Sort by score
+            sorted_categories = sorted(scores.items(), key=lambda x: x[1], reverse=True)
+            
+            # Return top categories with scores
+            return [(cat, score/5) for cat, score in sorted_categories if score > 0][:3]
+        
+        # Not using recommended models
+        using_recommended_models = False
 
 # File path for storing deals data locally
 DEALS_DATA_PATH = "deals_data.json"