Data preprocessing for Libero

Hi, thanks for the excellent work.

It appears that you didn't filter out no-op data when training on Libero. Is that correct? I'm asking because other models, such as OpenVLA, seem to apply no-op filtering.

Thanks!