Can you explain the details about you filter configuration when you process the dataset for fine-tune

[tianzhaohaha]

Could you please share the final Data Processing Meta CSV? I would like to know which videos were used in each training stage or how did you get the 500k HQ videos. Thank you!

[github-actions]

This issue is stale because it has been open for 7 days with no activity.

[zhengzangw]

As we mentioned in the report, we have no enough disk IO for filtering. Thus, we typically use all videos from pexels, and a subset of videos with an aesthetic score higher than 5, and an optical flow larger than 3.