[a16-support] Add asymmetric A16 fake quantization - #4924
Open
insoochung wants to merge 4 commits into
Open
insoochung wants to merge 4 commits into
insoochung wants to merge 4 commits into
Conversation
insoochung
requested review from
andrewor14,
jerryzh168 and
vkuzo
as code owners
September 19, 2026 14:22
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/4924
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit bddf3a4 with merge base 12a3afd ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Sep 19, 2026
Contributor
Author
|
This PR continues #4891. The asymmetric A16 implementation is unchanged. This re-export places it on the cleaned stack and removes the obsolete documentation dependency. |
insoochung
pushed a commit
that referenced
this pull request
Sep 19, 2026
Pull Request resolved: #4924 The static A16W4 stack supports signed symmetric activations. Add an optional logical UINT16 asymmetric mode that: - accepts explicit per-tensor quantization bounds through an INT32 carrier; - uses zero-preserving range selection that matches current Quanty; - applies the same behavior during direct fake quantization and calibration finalization; and - covers the asymmetric calibration-to-training workflow in tests. Existing configurations retain dtype-derived bounds and behavior. Previous GitHub review: #4891 ghstack-source-id: 432723624 @exported-using-ghexport Differential Revision: [D117934172](https://our.internmc.facebook.com/intern/diff/D117934172/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D117934172/)!
insoochung
pushed a commit
that referenced
this pull request
Sep 19, 2026
Pull Request resolved: #4924 The static A16W4 stack supports signed symmetric activations. Add an optional logical UINT16 asymmetric mode that: - accepts explicit per-tensor quantization bounds through an INT32 carrier; - uses zero-preserving range selection that matches current Quanty; - applies the same behavior during direct fake quantization and calibration finalization; and - covers the asymmetric calibration-to-training workflow in tests. Existing configurations retain dtype-derived bounds and behavior. Previous GitHub review: #4891 ghstack-source-id: 432727745 @exported-using-ghexport Differential Revision: [D117934172](https://our.internmc.facebook.com/intern/diff/D117934172/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D117934172/)!
andrewor14
approved these changes
Sep 21, 2026
andrewor14
left a comment
Contributor
There was a problem hiding this comment.
Please address the remaining comments from #4891, looks good otherwise
insoochung
pushed a commit
that referenced
this pull request
Sep 22, 2026
Pull Request resolved: #4924 The static A16W4 stack supports signed symmetric activations. Add an optional logical UINT16 asymmetric mode that: - accepts explicit per-tensor quantization bounds through an INT32 carrier; - uses zero-preserving range selection that matches current Quanty; - applies the same behavior during direct fake quantization and calibration finalization; and - covers the asymmetric calibration-to-training workflow in tests. Existing configurations retain dtype-derived bounds and behavior. Previous GitHub review: #4891 ghstack-source-id: 433395647 @exported-using-ghexport Differential Revision: [D117934172](https://our.internmc.facebook.com/intern/diff/D117934172/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D117934172/)!
insoochung
pushed a commit
that referenced
this pull request
Sep 22, 2026
Pull Request resolved: #4924 The static A16W4 stack supports signed symmetric activations. Add an optional logical UINT16 asymmetric mode that: - accepts explicit per-tensor quantization bounds through an INT32 carrier; - uses zero-preserving range selection that matches current Quanty; - applies the same behavior during direct fake quantization and calibration finalization; and - covers the asymmetric calibration-to-training workflow in tests. Existing configurations retain dtype-derived bounds and behavior. Previous GitHub review: #4891 ghstack-source-id: 433395647 @exported-using-ghexport Differential Revision: [D117934172](https://our.internmc.facebook.com/intern/diff/D117934172/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D117934172/)!
This branch has not been deployed
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
2 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.
Stack from ghstack (oldest at bottom):
The static A16W4 stack supports signed symmetric activations. Add an optional logical UINT16 asymmetric mode that:
Existing configurations retain dtype-derived bounds and behavior.
Previous GitHub review: #4891
@exported-using-ghexport
Differential Revision: D117934172
NOTE FOR REVIEWERS: This PR has internal Meta-specific changes or comments, please review them on Phabricator!
Differential Revision: D117934172