fix(library): Propagate upstream Marlin kernel fix #366
+17
−14
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ahadnagy
changed the title
ahadnagy
force-pushed
the
marlin-kernel-fix
branch
from
d2eba7f to
74298e3
Compare
Increase shared mem. size Fix shared mem. size, re-activate test Remove debugging-related stuff
ahadnagy
force-pushed
the
marlin-kernel-fix
branch
from
74298e3 to
ecd85fc
Compare
ahadnagy
force-pushed
the
marlin-kernel-fix
branch
from
ae10ca1 to
ec2bc69
Compare
|
@dacorvo It seems like it's not gonna work an A10s due to its lower shared memory size. Is that a hard requirement for the library? |
Pretty much, yes, since one of the main use case for quantization is to be able to run bigger models on smaller devices. |
|
Okay, in that case it'll be necessary to reduce the tile size. I'll check what vllm does on this front, and if that works on A10s at all. IIRC, their CI runs on L40s. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
What does this PR do?
Fixes #332
TLDR; There was a data race bug in the Marlin kernel. This fix basically adds a separate shared memory region for the final reduction tree. Unfortunately, this affects the minimum hardware requirements for the kernel, it won't work on GPUs with compute capability < 8.0.
Before submitting
Pull Request section?
to it if that's the case.