How to implement VideoChat2_text

[qtli]

Thanks for your high-quality work.
Could you please provide the blank video or the code you utilized to generate the ablation model VideoChat2_text?
Thanks very much!

[Andy1621]

Hi! For VideoChat2_text, we simply input a video tensor of 0, like torch.zeros_like(video_emb).