reland fp8 kv cache dequantization fix#3956
Closed
mxz297 wants to merge 1 commit into
Closed
Conversation
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D72799994 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
mxz297
force-pushed
the
export-D72799994
branch
from
ba885de to
f5e8768
Compare
mxz297
added a commit
to mxz297/FBGEMM
that referenced
this pull request
Apr 11, 2025
Summary: X-link: facebookresearch/FBGEMM#1041 This is a reland of D72062745, which was reverted as it broke NV. In this diff, we have separate kernels for AMD and NV, so it won't have any unintended side effects to NV kernels. For the AMD kernel, we continue to use one warp for a row. And because AMD has 64 threads per warp, we let the first 32 threads handle K and second 32 threads handle V. Differential Revision: D72799994
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D72799994 |
mxz297
added a commit
to mxz297/FBGEMM
that referenced
this pull request
Apr 11, 2025
Summary: X-link: facebookresearch/FBGEMM#1041 This is a reland of D72062745, which was reverted as it broke NV. In this diff, we have separate kernels for AMD and NV, so it won't have any unintended side effects to NV kernels. For the AMD kernel, we continue to use one warp for a row. And because AMD has 64 threads per warp, we let the first 32 threads handle K and second 32 threads handle V. Differential Revision: D72799994
mxz297
force-pushed
the
export-D72799994
branch
from
f5e8768 to
a26c198
Compare
mxz297
added a commit
to mxz297/FBGEMM
that referenced
this pull request
Apr 11, 2025
Summary: X-link: facebookresearch/FBGEMM#1041 This is a reland of D72062745, which was reverted as it broke NV. In this diff, we have separate kernels for AMD and NV, so it won't have any unintended side effects to NV kernels. For the AMD kernel, we continue to use one warp for a row. And because AMD has 64 threads per warp, we let the first 32 threads handle K and second 32 threads handle V. Differential Revision: D72799994
mxz297
force-pushed
the
export-D72799994
branch
from
a26c198 to
079c45a
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D72799994 |
mxz297
added a commit
to mxz297/FBGEMM
that referenced
this pull request
Apr 11, 2025
Summary: Pull Request resolved: pytorch#3956 X-link: facebookresearch/FBGEMM#1041 This is a reland of D72062745, which was reverted as it broke NV. In this diff, we have separate kernels for AMD and NV, so it won't have any unintended side effects to NV kernels. For the AMD kernel, we continue to use one warp for a row. And because AMD has 64 threads per warp, we let the first 32 threads handle K and second 32 threads handle V. Differential Revision: D72799994
mxz297
force-pushed
the
export-D72799994
branch
from
079c45a to
7c59e67
Compare
Summary: Pull Request resolved: pytorch#3956 X-link: facebookresearch/FBGEMM#1041 This is a reland of D72062745, which was reverted as it broke NV. In this diff, we have separate kernels for AMD and NV, so it won't have any unintended side effects to NV kernels. For the AMD kernel, we continue to use one warp for a row. And because AMD has 64 threads per warp, we let the first 32 threads handle K and second 32 threads handle V. Differential Revision: D72799994
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D72799994 |
mxz297
force-pushed
the
export-D72799994
branch
from
7c59e67 to
a25f605
Compare
Contributor
|
This pull request has been merged in b5c24b4. |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Summary:
This is a reland of D72062745, which was reverted as it broke NV.
In this diff, we have separate kernels for AMD and NV, so it won't have any unintended side effects to NV kernels.
For the AMD kernel, we continue to use one warp for a row. And because AMD has 64 threads per warp, we let the first 32 threads handle K and second 32 threads handle V.
Differential Revision: D72799994