Faster weight only quantized gemm #12862
Merged
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/12862
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 581503e with merge base d5232a0 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D78994081 |
This PR needs a
|
SS-JIA
added a commit
to SS-JIA/executorch-1
that referenced
this pull request
Jul 29, 2025
Summary: ## Context As title. Provide an implementation for the `gemm` counterpart of the weight int4 quantized `gemv` implementation added in D78275584 / pytorch#12444. This new kernel is quite similar to the existing one, with the primary difference being that it uses the same weight packing used in the `gemv` implementation. ## Next Steps * Reduce framework overhead from command buffer re-encoding between tokens. Achieve this by caching more artifacts that can be re-used between command buffer encodings, and only re-encoding command buffers when necessary. * Experiment with dynamic quantization, which should provide speedup via int8 dot product extension Reviewed By: digantdesai Differential Revision: D78994081
SS-JIA
added a commit
to SS-JIA/executorch-1
that referenced
this pull request
Jul 29, 2025
Summary: Pull Request resolved: pytorch#12862 ## Context As title. Provide an implementation for the `gemm` counterpart of the weight int4 quantized `gemv` implementation added in D78275584 / pytorch#12444. This new kernel is quite similar to the existing one, with the primary difference being that it uses the same weight packing used in the `gemv` implementation. ## Next Steps * Reduce framework overhead from command buffer re-encoding between tokens. Achieve this by caching more artifacts that can be re-used between command buffer encodings, and only re-encoding command buffers when necessary. * Experiment with dynamic quantization, which should provide speedup via int8 dot product extension Reviewed By: digantdesai Differential Revision: D78994081
|
This pull request was exported from Phabricator. Differential Revision: D78994081 |
Summary: ## Context As title. Provide an implementation for the `gemm` counterpart of the weight int4 quantized `gemv` implementation added in D78275584 / pytorch#12444. This new kernel is quite similar to the existing one, with the primary difference being that it uses the same weight packing used in the `gemv` implementation. ## Next Steps * Reduce framework overhead from command buffer re-encoding between tokens. Achieve this by caching more artifacts that can be re-used between command buffer encodings, and only re-encoding command buffers when necessary. * Experiment with dynamic quantization, which should provide speedup via int8 dot product extension Reviewed By: digantdesai Differential Revision: D78994081
|
This pull request was exported from Phabricator. Differential Revision: D78994081 |
mcr229
approved these changes
Jul 29, 2025
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:
Context
As title. Provide an implementation for the
gemmcounterpart of the weight int4 quantizedgemvimplementation added in D78275584 / #12444.This new kernel is quite similar to the existing one, with the primary difference being that it uses the same weight packing used in the
gemvimplementation.Next Steps
Reduce framework overhead from command buffer re-encoding between tokens. Achieve this by caching more artifacts that can be re-used between command buffer encodings, and only re-encoding command buffers when necessary.
Experiment with dynamic quantization, which should provide speedup via int8 dot product extension
Differential Revision: D78994081