Update app.py

app.py

@@ -15,7 +15,7 @@ if not torch.cuda.is_available():
     DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
 
 MAX_MAX_NEW_TOKENS = 2048
-DEFAULT_MAX_NEW_TOKENS = 1024
+DEFAULT_MAX_NEW_TOKENS = 512
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 if torch.cuda.is_available():
@@ -62,6 +62,7 @@ def generate(
     outputs = []
     for text in streamer:
         outputs.append(text)
+        outputs = outputs.replace("||assistant|>", "").strip()
         yield "".join(outputs)