[ExecuTorch] XNNPACK: prefer qc over qb when gs == k for non-int4 #14173
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
* Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)! [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/14173
Note: Links to docs will display an error until the docs builds have been completed. ❌ 4 New Failures, 2 Cancelled Jobs, 37 Pending, 3 Unrelated FailuresAs of commit b53e5b2 with merge base 6b9c0a6 ( NEW FAILURES - The following jobs have failed:
CANCELLED JOBS - The following jobs were cancelled. Please retry:
BROKEN TRUNK - The following jobs failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
digantdesai
added a commit
that referenced
this pull request
Sep 10, 2025
* Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)! ghstack-source-id: 308858460 Pull Request resolved: #14173
meta-cla
bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D82060758 |
…on-int4" * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)! [ghstack-poisoned]
digantdesai
added a commit
that referenced
this pull request
Sep 10, 2025
Pull Request resolved: #14173 * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 ghstack-source-id: 308860455 @exported-using-ghexport Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)!
|
This pull request was exported from Phabricator. Differential Revision: D82060758 |
…on-int4" * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)! [ghstack-poisoned]
digantdesai
added a commit
that referenced
this pull request
Sep 11, 2025
Pull Request resolved: #14173 * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 ghstack-source-id: 309088729 @exported-using-ghexport Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)!
|
This pull request was exported from Phabricator. Differential Revision: D82060758 |
digantdesai
added
release notes: none
metascroy
approved these changes
Sep 11, 2025
…on-int4" * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)! [ghstack-poisoned]
digantdesai
added a commit
that referenced
this pull request
Sep 11, 2025
Pull Request resolved: #14173 * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 ghstack-source-id: 309161539 @exported-using-ghexport Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)!
|
This pull request was exported from Phabricator. Differential Revision: D82060758 |
…on-int4" * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)! [ghstack-poisoned]
digantdesai
added a commit
that referenced
this pull request
Sep 11, 2025
Pull Request resolved: #14173 * Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support * Fix bug / improve behavior for affine q/dq with gs == k for per_channel * refactor is_per_channel_group state variable * add QuantParams.__str__() TODO - improve affine quant primitives - T237476295 ghstack-source-id: 309177704 @exported-using-ghexport Differential Revision: [D82060758](https://our.internmc.facebook.com/intern/diff/D82060758/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D82060758/)!
|
This pull request was exported from Phabricator. Differential Revision: D82060758 |
facebook-github-bot
merged commit 51e18e7
into
gh/digantdesai/45/base
286 of 300 checks passed
facebook-github-bot
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
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.
Stack from ghstack (oldest at bottom):
-> [ExecuTorch] XNNPACK: prefer qc over qb when gs == k for non-int4 #14173
Prefer chanelwise over groupwise when possible for perf and for int8 which doesn't have groupwise support
Fix bug / improve behavior for affine q/dq with gs == k for per_channel
refactor is_per_channel_group state variable
add QuantParams.str()
TODO - improve affine quant primitives - T237476295
Differential Revision: D82060758
NOTE FOR REVIEWERS: This PR has internal Meta-specific changes or comments, please review them on Phabricator!