Add 16A8W quantization configuration utility for ARM backend#13641
Merged
Ninja91 merged 4 commits intogh/Ninja91/1/basefrom Aug 27, 2025
Merged
Add 16A8W quantization configuration utility for ARM backend#13641Ninja91 merged 4 commits intogh/Ninja91/1/basefrom
Ninja91 merged 4 commits intogh/Ninja91/1/basefrom
Conversation
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. Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/) [ghstack-poisoned]
Ninja91
added a commit
that referenced
this pull request
Aug 25, 2025
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. Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/) ghstack-source-id: 305459598 Pull Request resolved: #13641
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13641
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New FailuresAs of commit bc36615 with merge base 9053089 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D79763381 |
This PR needs a
|
Contributor
Author
|
Older PR was #13175 which was closed |
Ninja91
added a commit
to Ninja91/executorch
that referenced
this pull request
Aug 25, 2025
…#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
Ninja91
added a commit
to Ninja91/executorch
that referenced
this pull request
Aug 25, 2025
…#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
mergennachin
added
the
partner: arm
Ninja91
added a commit
to Ninja91/executorch
that referenced
this pull request
Aug 25, 2025
…#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
Ninja91
added a commit
to Ninja91/executorch
that referenced
this pull request
Aug 25, 2025
…#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
Ninja91
added a commit
to Ninja91/executorch
that referenced
this pull request
Aug 25, 2025
…#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
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. Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/) cc digantdesai freddan80 per zingo oscarandersson8218 [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D79763381 |
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. Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/) cc digantdesai freddan80 per zingo oscarandersson8218 [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D79763381 |
per
approved these changes
Aug 26, 2025
digantdesai
reviewed
Aug 26, 2025
digantdesai
approved these changes
Aug 26, 2025
Contributor
digantdesai
left a comment
digantdesai
left a comment
There was a problem hiding this comment.
Typically you add accompanying test with a PR. so each PR (and commits inside are self contained).
Contributor
Author
|
@digantdesai I see some unrelated test failing https://github.com/pytorch/executorch/actions/runs/17227752037/job/48875652934?pr=13641. Any guidance towards merging this? |
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. Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/) cc digantdesai freddan80 per zingo oscarandersson8218 [ghstack-poisoned]
Ninja91
added a commit
that referenced
this pull request
Aug 27, 2025
Pull Request resolved: #13641 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: 305891620 @exported-using-ghexport Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/)
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D79763381 |
Ninja91
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
Ninja91
added a commit
that referenced
this pull request
Aug 27, 2025
Pull Request resolved: #13641 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: 305891620 @exported-using-ghexport Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/)
Ninja91
added a commit
that referenced
this pull request
Aug 27, 2025
Pull Request resolved: #13641 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: 305991462 @exported-using-ghexport Differential Revision: [D79763381](https://our.internmc.facebook.com/intern/diff/D79763381/)
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D79763381 |
lucylq
pushed a commit
that referenced
this pull request
Sep 2, 2025
This PR was created by the merge bot to help merge the original PR into the main branch. ghstack PR number: #13641 by @Ninja91 ^ Please use this as the source of truth for the PR details, comments, and reviews ghstack PR base: https://github.com/pytorch/executorch/tree/gh/Ninja91/1/base ghstack PR head: https://github.com/pytorch/executorch/tree/gh/Ninja91/1/head Merge bot PR base: https://github.com/pytorch/executorch/tree/main Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/Ninja91/1/orig @diff-train-skip-merge Co-authored-by: Nitin Jain <jainnitin@meta.com>
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):
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
get_16a8w_quantization_config()infbcode/executorch/backends/arm/quantizer/arm_quantizer.pyBenefits
Differential Revision: D79763381
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218