Add asymmetric support for Int8Tensor + SmoothQuant#3900
Merged
Conversation
Summary: This PR adds in support for asymmetric activation quantization in Int8Tensor and Smoothquant Test Plan: ``` pytest test/quantization/quantize_/workflows/int8/test_int8_tensor.py pytest test/prototype/test_smoothquant.py ``` Reviewers: Subscribers: Tasks: Tags: [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3900
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit dd40b73 with merge base 5906856 ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
jcaip
added a commit
that referenced
this pull request
Feb 17, 2026
Summary: This PR adds in support for asymmetric activation quantization in Int8Tensor and Smoothquant Test Plan: ``` pytest test/quantization/quantize_/workflows/int8/test_int8_tensor.py pytest test/prototype/test_smoothquant.py ``` Reviewers: Subscribers: Tasks: Tags: ghstack-source-id: 5163aa0 Pull Request resolved: #3900
meta-cla
bot
added
the
CLA Signed
Summary: This PR adds in support for asymmetric quantization in Int8Tensor by adding in a new optional tensor attribute, `zero_point` and `act_zero_point` for the weight / activation respectively. Also adds in a support for asymmetric quantization in our smoothquant implementation. Test Plan: ``` pytest test/quantization/quantize_/workflows/int8/test_int8_tensor.py pytest test/prototype/test_smoothquant.py ``` Reviewers: Subscribers: Tasks: Tags: [ghstack-poisoned]
jcaip
added a commit
that referenced
this pull request
Feb 17, 2026
Summary: This PR adds in support for asymmetric quantization in Int8Tensor by adding in a new optional tensor attribute, `zero_point` and `act_zero_point` for the weight / activation respectively. Also adds in a support for asymmetric quantization in our smoothquant implementation. Test Plan: ``` pytest test/quantization/quantize_/workflows/int8/test_int8_tensor.py pytest test/prototype/test_smoothquant.py ``` Reviewers: Subscribers: Tasks: Tags: ghstack-source-id: a146822 Pull Request resolved: #3900
namgyu-youn
reviewed
Feb 24, 2026
| tensor: torch.Tensor, | ||
| quant_kwargs: QuantizeTensorKwargs, | ||
| scale: Optional[torch.Tensor] = None, | ||
| zero_point: Optional[torch.Tensor] = None, |
Contributor
There was a problem hiding this comment.
nit: weight_zero_point for clear naming?
Contributor
Author
There was a problem hiding this comment.
I think we should keep this consistent with scale. act_zero_point is used to denote the activation zero point
| int8 quantized tensor with plain layout. | ||
|
|
||
| Currently only Symmetric quantization is supported. | ||
| Supports both symmetric and asymmetric quantization. |
Contributor
There was a problem hiding this comment.
Maybe drop this docstring? Description inside Tensor Attribute section looks enough I feel.
Contributor
Author
There was a problem hiding this comment.
sounds good, will remove
hossein1387
reviewed
Feb 24, 2026
| act_quant_kwargs: flags for dynamic activation quantization | ||
| """ | ||
|
|
||
| tensor_data_names = ["qdata", "scale"] |
There was a problem hiding this comment.
shouldn't zero_point be required for MappingType.ASYMMETRIC here?
torchao/quantization/quant_api.py
Outdated
| ) | ||
| assert config.version == 2, f"Unexpected version: {config.version}" | ||
|
|
||
| # TODO: Symmentric/Asymmetric choice for weight quantization |
Contributor
Author
|
@jcaip has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
sxu
approved these changes
Feb 25, 2026
Removed TODO comment regarding symmetric/asymmetric weight quantization.
Removed mention of symmetric quantization support from docstring.
Collaborator
|
CC @cyxlily |
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
5 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):
Summary:
This PR adds in support for asymmetric quantization in
Int8Tensor by adding in a new optional tensor attribute,
zero_pointandact_zero_pointfor the weight / activationrespectively.
Also adds in a support for asymmetric quantization in our smoothquant
implementation.
Test Plan:
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: D94258324