Update app.py

app.py

@@ -30,7 +30,6 @@ def load_model():
         try:
             # Fallback to transformers
             from transformers import pipeline
-            global model
             model = pipeline("text-generation", model="microsoft/Phi-3.5-mini-instruct")
             logger.info("✅ Fallback model loaded!")
             return True
@@ -40,6 +39,7 @@ def load_model():
 
 @app.route('/generate', methods=['POST'])
 def generate_text():
+    global model, tokenizer
     try:
         data = request.json
         prompt = data.get('inputs', data.get('prompt', ''))