blepping
aad095ab14
Refactor to avoid having to monkeypatch forward_timestep_embed
...
Other cleanups
Allow setting a downscale factor and mode for CA downsampling
Add more type annotations
Make it so MSW-MSA attention failing is a warning rather than hard error
Adjust a few Ruff lint settings
2024-08-10 06:27:59 -06:00
Pam
5636c8c514
Refactor RAUNet code
2024-08-04 23:57:59 +05:00
blepping
02666ceff4
More reliable ComfyUI-bleh integration (hopefully)
2024-05-21 09:58:15 -06:00
blepping
e54ff4d58d
Remove match/case statements for Python 3.9 compatibility
2024-05-09 22:18:45 -06:00
blepping
16abd6868d
Fix incorrect default setting for ApplyMSWMSAAttention node (block 0 isn't attention)
...
Some refactoring/cleanups
Update documentation with use cases/compatibility notes
2024-05-07 00:49:08 -06:00
blepping
2cc4529472
Allow use of bleh upscale methods when possible
2024-04-27 18:13:07 -06:00
blepping
e693373a12
Detach and move sigmas to CPU before comparing
...
Clean up/refactor activation checking code
2024-04-27 14:03:29 -06:00
blepping