Update app.py

app.py

@@ -149,7 +149,7 @@ def generate(
 ) -> Iterator[str]:
     conversation = [*chat_history, {"role": "user", "content": message}]
 
-    input_ids = tokenizer.apply_chat_template(conversation, chat_template=chat_template, return_tensors="pt")
+    input_ids = tokenizer.apply_chat_template(conversation, chat_template=chat_template, enable_thinking=False, return_tensors="pt")
     if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
         input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
         gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")