Add Preshuffled FP8 x INT4 Grouped Gemm Kernel#3800
Closed
jwfromm wants to merge 3 commits into
Closed
Conversation
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
jwfromm
force-pushed
the
export-D70870933
branch
from
8c4ef51 to
071a184
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: Working on adding support for stacked mixed dtype grouped gemm with preshuffling. Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
071a184 to
e86a553
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
e86a553 to
9391323
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: Working on adding support for stacked mixed dtype grouped gemm with preshuffling. Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
force-pushed
the
export-D70870933
branch
from
9391323 to
ad12177
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Pull Request resolved: pytorch#3800 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
ad12177 to
d39247d
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
d39247d to
d9c8ad4
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
force-pushed
the
export-D70870933
branch
from
d9c8ad4 to
3fa97dc
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Pull Request resolved: pytorch#3800 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
jwfromm
pushed a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: Pull Request resolved: pytorch#3800 Working on adding support for stacked mixed dtype grouped gemm with preshuffling. Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 13, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Pull Request resolved: pytorch#3800 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
3fa97dc to
7e88324
Compare
Summary: X-link: facebookresearch/FBGEMM#847 Pull Request resolved: pytorch#3766 One of the new interesting changes in the preshuffled F8I4 kernel is that group scales are downcast to FP8. This has the risk of running into dynamic range issues and impacting accuracy. We can mitigate this risk by adding FP32 columnwise scaling to the output. Fortunately, we can do this using EVT so the performance impact is negligible. Differential Revision: D70587477
Summary: X-link: facebookresearch/FBGEMM#855 Pull Request resolved: pytorch#3775 This diff introduces a set of quantization helper functions to fbgemm_gpu/experimental/gen_ai to make it easier to apply the new Int4 packing and preshuffling to weights. Differential Revision: D70643388 Reviewed By: summerdengfb
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 14, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Reviewed By: jiawenliu64 Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
7e88324 to
107e9f6
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 14, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Reviewed By: jiawenliu64 Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 14, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Reviewed By: jiawenliu64 Differential Revision: D70870933
jwfromm
force-pushed
the
export-D70870933
branch
from
107e9f6 to
864fb11
Compare
jwfromm
added a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 14, 2025
Summary: X-link: facebookresearch/FBGEMM#897 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Reviewed By: jiawenliu64 Differential Revision: D70870933
Summary: X-link: facebookresearch/FBGEMM#897 Pull Request resolved: pytorch#3800 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Reviewed By: jiawenliu64 Differential Revision: D70870933
jwfromm
pushed a commit
to jwfromm/FBGEMM
that referenced
this pull request
Mar 14, 2025
Summary: Pull Request resolved: pytorch#3800 Working on adding support for stacked mixed dtype grouped gemm with preshuffling. Differential Revision: D70870933
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D70870933 |
jwfromm
force-pushed
the
export-D70870933
branch
from
864fb11 to
46acfbc
Compare
Contributor
|
This pull request has been merged in a39d2cc. |
liligwu
pushed a commit
to ROCm/FBGEMM
that referenced
this pull request
Mar 19, 2025
Summary: X-link: https://github.com/facebookresearch/FBGEMM/pull/897 Pull Request resolved: pytorch#3800 Efficient FP8xINT4 grouped gemm with preshuffling and scale packing. This implementation uses the "stacked" API where inputs and outputs are single contiguous tensors and the group boundaries are indicated with an `M_sizes` tensor that contains the number of rows in each group. Reviewed By: jiawenliu64 Differential Revision: D70870933 fbshipit-source-id: 195fb9feb993ffa7efe27b038173bd70a1db57ed
|
hey @jwfromm - thanks for the PR What exactly the intended use of |
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
4 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: Working on adding support for stacked mixed dtype grouped gemm with preshuffling.
Differential Revision: D70870933