Update app.py

app.py

@@ -15,6 +15,7 @@ DESCRIPTION = "# ICONN Lite Chat"
 if not torch.cuda.is_available():
     DESCRIPTION += "\n<p class='warning'>Running on CPU 🥶 This demo does not work on CPU.</p>"
 
+top_k: int = 50
 MAX_MAX_NEW_TOKENS = 100000000
 DEFAULT_MAX_NEW_TOKENS = 10240
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))