Update app.py

app.py

@@ -9,6 +9,7 @@ import pickle
 setattr(httpcore, 'SyncHTTPTransport', 'AsyncHTTPProxy')
 
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+model = SentenceTransformer('intfloat/multilingual-e5-large-instruct')
 
 def get_detailed_instruct(task_description: str, query: str) -> str:
         return f'Instruct: {task_description}\nQuery: {query}'
@@ -35,7 +36,6 @@ def respond(message,
         get_detailed_instruct(task, message)
     ]
 
-    model = SentenceTransformer('intfloat/multilingual-e5-large-instruct')
     query_embeddings = model.encode(queries, convert_to_tensor=True, normalize_embeddings=True)
     scores = (query_embeddings @ encoded_questions.T) * 100
     selected_references['similarity'] = scores.tolist()[0]