Update app.py

app.py

@@ -46,8 +46,11 @@ def predict(message, system_message, max_new_tokens=4056, temperature=3.5, top_p
     )
 
     response = tokenizer.decode(response_ids[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
-    return [("bot", response)]    
+    truncate_str = "<|im_end|>"
+    if truncate_str in response:
+        response = response.split(truncate_str)[0] + truncate_str
 
+    return [("bot", response)]    
 with gr.Blocks(theme='ParityError/Anime') as demo:
     gr.Markdown(DESCRIPTION)
     with gr.Group():