Skip to content

Keep using float16 in ldm.modules.attention; ~12% speed up and less VRAM - #494

Closed
mh-dm wants to merge 1 commit into
invoke-ai:mainfrom
mh-dm:float16
Closed

Keep using float16 in ldm.modules.attention; ~12% speed up and less VRAM#494
mh-dm wants to merge 1 commit into
invoke-ai:mainfrom
mh-dm:float16

Commits

Commits on Sep 10, 2022