Raphael Walker 5e8206362c fix pytorch/xformers attention
This corrects a weird inconsistency with skip_reshape.
It also allows masks of various shapes to be passed, which will be
automtically expanded (in a memory-efficient way) to a size that is
compatible with xformers or pytorch sdpa respectively.
2024-12-11 10:38:50 +01:00
..
2024-12-10 11:45:55 +01:00
2024-05-27 19:08:27 -04:00
2024-12-11 10:38:50 +01:00
2024-11-27 13:45:32 -05:00
2023-01-16 22:37:14 -05:00