flash attention: correct head dim check and L_k padding #736
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Gains are minimal, if any.
SD1 now has some usage of flash attention.
Some numbers where found by @bssrdf in #386 .
From what I can tell, the numbers should translate to vulkan and probably rocm pretty well.
sd1
fattn log
SDXL
Affected from padding L_k.
fattn log 768x768
SD2 and flux stay the same, besides some wonky resolutions.
Device is rtx2070 8gig mobile.
Vulkan with
coopmat1
only.