Issues about attention mask?

Open WillWu111 opened this issue 3 weeks ago • 0 comments

How can we design a memory-friendly mask of the linear attention implemented in Sana-Video?

Dec 01 '25 12:12 WillWu111