Update app.py

app.py

@@ -48,7 +48,6 @@ def respond(
         stream=True,
         temperature=temperature,
         top_p=top_p,
-        return_full_text=false,
     ):
         token = message.choices[0].delta.content
 
@@ -91,6 +90,7 @@ def generate(
         temperature=temperature,
         num_beams=1,
         repetition_penalty=repetition_penalty,
+        return_full_text=false,
     )
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()