* add vid2vid pipeline for cogvideox * make fix-copies * update docs * fake context parallel cache, vae encode tiling * add test for cog vid2vid * use video link from HF docs repo * add copied from comments; correctly rename test class