Tune FP8 grouped GEMM for Llama4 shapes #4326
Closed
Conversation
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
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
This pull request was exported from Phabricator. Differential Revision: D76460456 |
|
This pull request was exported from Phabricator. Differential Revision: D76460456 |
jiawenliu64
force-pushed
the
export-D76460456
branch
from
6b83eae to
db1741a
Compare
jiawenliu64
added a commit
to jiawenliu64/FBGEMM
that referenced
this pull request
Jun 15, 2025
Summary: Pull Request resolved: pytorch#4326 X-link: facebookresearch/FBGEMM#1399 Further tune FP8 grouped GEMM for Llama4 shapes. Found there is 13%-30% perf gain for Llama4 memory-bound cases. For other shapes in future models, I think there will be room to further improve perf by adding more heuristics like this Diff, that we will revisit and add offline cache file loading instead of adding under the kernel file. Differential Revision: D76460456
Summary: Pull Request resolved: pytorch#4326 X-link: facebookresearch/FBGEMM#1399 Further tune FP8 grouped GEMM for Llama4 shapes. Found there is 13%-30% perf gain for Llama4 memory-bound cases. For other shapes in future models, I think there will be room to further improve perf by adding more heuristics like this Diff, that we will revisit and add offline cache file loading instead of adding under the kernel file. Differential Revision: D76460456
|
This pull request was exported from Phabricator. Differential Revision: D76460456 |
jiawenliu64
force-pushed
the
export-D76460456
branch
from
db1741a to
65b29c4
Compare
jiawenliu64
added a commit
to jiawenliu64/FBGEMM
that referenced
this pull request
Jun 15, 2025
Summary: Pull Request resolved: pytorch#4326 X-link: facebookresearch/FBGEMM#1399 Further tune FP8 grouped GEMM for Llama4 shapes. Found there is 13%-30% perf gain for Llama4 memory-bound cases. For other shapes in future models, I think there will be room to further improve perf by adding more heuristics like this Diff, that we will revisit and add offline cache file loading instead of adding under the kernel file. Differential Revision: D76460456
jiawenliu64
added a commit
to jiawenliu64/FBGEMM
that referenced
this pull request
Jun 16, 2025
Summary: Pull Request resolved: pytorch#4326 X-link: facebookresearch/FBGEMM#1399 Further tune FP8 grouped GEMM for Llama4 shapes. Found there is 13%-30% perf gain for Llama4 memory-bound cases. For other shapes in future models, I think there will be room to further improve perf by adding more heuristics like this Diff, that we will revisit and add offline cache file loading instead of adding under the kernel file. Differential Revision: D76460456
|
This pull request has been merged in 6ee9646. |
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.
Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/1399
Further tune FP8 grouped GEMM for Llama4 shapes. Found there is 13%-30% perf gain for Llama4 memory-bound cases. For other shapes in future models, I think there will be room to further improve perf by adding more heuristics like this Diff, that we will revisit and add offline cache file loading instead of adding under the kernel file.
Differential Revision: D76460456