Add 16A8W linear ops support and test#13448
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13448
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 1 Pending, 2 Unrelated FailuresAs of commit acf553f with merge base 2100527 ( NEW FAILURE - The following job has failed:
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. |
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
This PR needs a
|
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. Note: Not verified with tosa reference model run. Differential Revision: D80308822
ca1cd7f to
793a055
Compare
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
Summary: Pull Request resolved: pytorch#13448 - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. Note: Not verified with tosa reference model run. Differential Revision: D80308822
793a055 to
e8402a2
Compare
e8402a2 to
e725aa7
Compare
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. Note: Not verified with tosa reference model run. Differential Revision: D80308822
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
Summary: Pull Request resolved: pytorch#13448 - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. Note: Not verified with tosa reference model run. Differential Revision: D80308822
bad2472 to
cd39120
Compare
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. Note: Not verified with tosa reference model run. Differential Revision: D80308822
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
Summary: Pull Request resolved: pytorch#13448 - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
79d958b to
5a7189c
Compare
|
GH issue: #13635 |
Summary: Pull Request resolved: pytorch#13448 - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
5a7189c to
56e8f7c
Compare
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
56e8f7c to
cace06e
Compare
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
cace06e to
f034a00
Compare
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
Summary: Pull Request resolved: pytorch#13448 - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
f034a00 to
aa0b3f7
Compare
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
aa0b3f7 to
6af0907
Compare
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
6af0907 to
5fc2ffe
Compare
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
1 similar comment
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
Summary: Pull Request resolved: pytorch#13448 - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
…#13641) Summary: This diff implements a 16A8W (16-bit activations, 8-bit weights) quantization configuration utility for the ExecutorTorch ARM backend, following the feedback from D79746479. ## Key Changes **1. New Quantization Configuration Function** - Add `get_16a8w_quantization_config()` in `fbcode/executorch/backends/arm/quantizer/arm_quantizer.py` - Provides 16-bit activations with HistogramObserver (better precision than 8A8W) - Maintains 8-bit weights with MinMaxObserver/PerChannelMinMaxObserver (memory efficient) - **Technically supported by TOSA through [EXT-INT16 extension/profile](https://www.mlplatform.org/tosa/tosa_spec.html#_conv2d)** ## Benefits - **Better Precision**: 16-bit activations provide higher precision than 8-bit. Useful for carrying precision for recurring neural nets. ghstack-source-id: 305459598 exported-using-ghexport Reviewed By: 3l1 Differential Revision: D79763381
Summary: - Adds linear ops test using the 16A8W config in INT16 profile. - Adds support in view ops validation for INT16 Dtype. - Validated with TOSA pipeline test. - Checked earlier marked flaky tests no longer flaky and remove markers. Note: Not verified with tosa reference model run. Reviewed By: digantdesai Differential Revision: D80308822
|
This pull request was exported from Phabricator. Differential Revision: D80308822 |
|
@digantdesai @per strangely seeing https://github.com/pytorch/executorch/actions/runs/17219418500/job/48850992655 error which doesn't look related. Am i missing something here? |
|
Updated in a new PR: #13658. Closing this. |
Summary:
Note: Not verified with tosa reference model run.
Differential Revision: D80308822
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218