Update app.py

app.py

@@ -77,7 +77,7 @@ with gr.Blocks() as demo:
     def respond(message, chat_history):
 
         # prompt wrapper, only single-turn is allowed for now
-        prompt = f"### Human:\n{prompt}\n\n### Assistant:\n"
+        prompt = f"### Human:\n{message}\n\n### Assistant:\n"
 
         batch = tokenizer(
             prompt,