diff --git a/vlmeval/vlm/video_llm/video_llava.py b/vlmeval/vlm/video_llm/video_llava.py
index 76ea7eb0..e78f44db 100644
--- a/vlmeval/vlm/video_llm/video_llava.py
+++ b/vlmeval/vlm/video_llm/video_llava.py
@@ -44,7 +44,7 @@ def __init__(self, model_path='LanguageBind/Video-LLaVA-7B-hf', **kwargs):
         self.model.eval().cuda()
         self.processor = VideoLlavaProcessor.from_pretrained(model_path)
         self.kwargs = kwargs
-        self.nframe = 8
+        self.nframes = 8
         torch.cuda.empty_cache()
 
     def generate_inner(self, message, dataset=None):