Reduce max_new_tokens to 64 for faster replies

app.py

@@ -20,7 +20,7 @@ def respond(message: str, history: list[tuple[str, str]]):
 
     out = llm(
         prompt,
-        max_new_tokens=256,
+        max_new_tokens=64,
         temperature=0.7,
         top_p=0.9,
     )