[ET-VK] Enable IntxWeightOnlyConfig #13452
Merged
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
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
requested review from
jackzhxng,
larryliu0820,
lucylq,
mergennachin and
swolchok
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13452
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New FailureAs of commit fb29875 with merge base 1b4968f ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) ghstack-source-id: 303222323 Pull Request resolved: #13452
meta-cla
bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D80293302 |
This PR needs a
|
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
Pull Request resolved: #13452 ## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) ghstack-source-id: 303224475
|
This pull request was exported from Phabricator. Differential Revision: D80293302 |
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
Pull Request resolved: #13452 ## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script ghstack-source-id: 303277500 Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/)
|
This pull request was exported from Phabricator. Differential Revision: D80293302 |
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
Pull Request resolved: #13452 Be able to test Vulkan lowering via optimum-executorch. Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script ghstack-source-id: 6903b2d Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/)
|
@SS-JIA has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
Pull Request resolved: #13452 ## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script ghstack-source-id: 303313326 Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/)
|
This pull request was exported from Phabricator. Differential Revision: D80293302 |
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
Pull Request resolved: #13452 ## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script ghstack-source-id: 303328940 Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/)
|
This pull request was exported from Phabricator. Differential Revision: D80293302 |
digantdesai
approved these changes
Aug 15, 2025
## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Aug 15, 2025
Pull Request resolved: #13452 ## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script ghstack-source-id: 303387280 Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/)
|
This pull request was exported from Phabricator. Differential Revision: D80293302 |
facebook-github-bot
merged commit 6bef39c
into
gh/SS-JIA/288/base
102 of 106 checks passed
facebook-github-bot
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
SS-JIA
pushed a commit
that referenced
this pull request
Aug 16, 2025
Pull Request resolved: #13452 ## Motivation Be able to test Vulkan lowering via optimum-executorch. ## Context Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op. This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear. ## Changes * Introduce a fusable graph pattern for weight only quantized linear * Add fusion logic for weight only quantized linear in the fuse patterns pass * Add `4w` qmode to the export llama script ghstack-source-id: 303387280 Differential Revision: [D80293302](https://our.internmc.facebook.com/intern/diff/D80293302/)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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):
Motivation
Be able to test Vulkan lowering via optimum-executorch.
Context
Very similar to the below PR, Int4 weight only quantization is currently enabled in Vulkan via a custom source transform quantizer that replaces linear layers with a custom linear layer that calls a custom weight only quantized linear op.
This diff aims to make it so that no Vulkan specific source transforms need to be applied by adding a fusion pattern for weight only quantized linear.
Changes
4wqmode to the export llama scriptDifferential Revision: D80293302