[ET-VK] Enable int8 tiled compute shader to be used with buffer tensors#10302
Merged
facebook-github-bot merged 5 commits intogh/SS-JIA/214/basefrom Apr 23, 2025
Merged
[ET-VK] Enable int8 tiled compute shader to be used with buffer tensors#10302facebook-github-bot merged 5 commits intogh/SS-JIA/214/basefrom
facebook-github-bot merged 5 commits intogh/SS-JIA/214/basefrom
Conversation
## Context As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well. ## Changes * Generate buffer variants for the int8 linear tiled shader * Force the scales tensor to always be a buffer to reduce the number of shader variants that need to be generated. * Generate an additional variant that computes only 1 output row * Do not require output rows to be an exact multiple of 4 or 6 to use the tiled implementation Differential Revision: [D73276277](https://our.internmc.facebook.com/intern/diff/D73276277/) [ghstack-poisoned]
SS-JIA
added a commit
that referenced
this pull request
Apr 18, 2025
## Context As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well. ## Changes * Generate buffer variants for the int8 linear tiled shader * Force the scales tensor to always be a buffer to reduce the number of shader variants that need to be generated. * Generate an additional variant that computes only 1 output row * Do not require output rows to be an exact multiple of 4 or 6 to use the tiled implementation Differential Revision: [D73276277](https://our.internmc.facebook.com/intern/diff/D73276277/) ghstack-source-id: 279008193 Pull Request resolved: #10302
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/10302
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New Failures, 1 PendingAs of commit fa29ad0 with merge base 334af4a ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73276277 |
…uffer tensors" ## Context As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well. ## Changes * Generate buffer variants for the int8 linear tiled shader * Force the scales tensor to always be a buffer to reduce the number of shader variants that need to be generated. * Generate an additional variant that computes only 1 output row * Do not require output rows to be an exact multiple of 4 or 6 to use the tiled implementation Differential Revision: [D73276277](https://our.internmc.facebook.com/intern/diff/D73276277/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73276277 |
…uffer tensors" ## Context As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well. ## Changes * Generate buffer variants for the int8 linear tiled shader * Force the scales tensor to always be a buffer to reduce the number of shader variants that need to be generated. * Generate an additional variant that computes only 1 output row * Do not require output rows to be an exact multiple of 4 or 6 to use the tiled implementation Differential Revision: [D73276277](https://our.internmc.facebook.com/intern/diff/D73276277/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73276277 |
…uffer tensors" ## Context As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well. ## Changes * Generate buffer variants for the int8 linear tiled shader * Force the scales tensor to always be a buffer to reduce the number of shader variants that need to be generated. * Generate an additional variant that computes only 1 output row * Do not require output rows to be an exact multiple of 4 or 6 to use the tiled implementation Differential Revision: [D73276277](https://our.internmc.facebook.com/intern/diff/D73276277/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73276277 |
…uffer tensors" ## Context As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well. ## Changes * Generate buffer variants for the int8 linear tiled shader * Force the scales tensor to always be a buffer to reduce the number of shader variants that need to be generated. * Generate an additional variant that computes only 1 output row * Do not require output rows to be an exact multiple of 4 or 6 to use the tiled implementation Differential Revision: [D73276277](https://our.internmc.facebook.com/intern/diff/D73276277/) [ghstack-poisoned]
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D73276277 |
yipjustin
approved these changes
Apr 23, 2025
facebook-github-bot
merged commit 9217e37
into
gh/SS-JIA/214/base
83 of 87 checks passed
facebook-github-bot
temporarily deployed
to
cherry-pick-bot
GitHub Actions
Inactive
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
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):
Context
As title. Allow the optimized int8 tiled compute shader to be usable for buffer-backed tensors as well.
Changes
Differential Revision: D73276277