[ET-VK][ez] Fix handling of assert ops #11349
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
Pull Request resolved: #11256 ## Changes * Implement resize functions for several operators used in Transformers models ## Motivation Be able to support batched prefill for llama models. ghstack-source-id: 287935585 @exported-using-ghexport Differential Revision: [D75686049](https://our.internmc.facebook.com/intern/diff/D75686049/)
Pull Request resolved: #11257 ## Changes * Add an implementation for operator.add which add symbolic integers. ## Motivation Support executing llama models with dynamic shapes. This operator shows up when exporting with dynamic shapes. ghstack-source-id: 287945686 @exported-using-ghexport Differential Revision: [D75238029](https://our.internmc.facebook.com/intern/diff/D75238029/)
Pull Request resolved: #11258 ## Changes * Apply `RemoveAssertsTransform` as part of `vulkan_preprocess` * Do not call `RemoveAssertsTransform` before lowering the graph * Register ops related to asserts to the operator registry as ephemeral ops ## Motivation assert ops are not implemented in Vulkan, so previously `RemoveAssertsTransform()` is called on the graph before the lowering process. However, it turns out that the assertion ops are required to properly handle dynamic shapes, because they place constraints on the possible range of symbolic integers. If they are not present, then re-tracing the graph during a recompile (which may occur during a graph transform pass) may fail. Therefore, instead of calling the transform before lowering, call it inside vulkan_preprocess after a point where subsequent passes will not attempt to trace the graph. ghstack-source-id: 287945687 @exported-using-ghexport Differential Revision: [D75686048](https://our.internmc.facebook.com/intern/diff/D75686048/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11349
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ⏳ No Failures, 33 PendingAs of commit c907163 with merge base c7896df ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
SS-JIA
approved these changes
Jun 4, 2025
This PR needs a
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #11258 by @SS-JIA
^ 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/SS-JIA/234/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/234/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/SS-JIA/233/orig
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/234/orig
@diff-train-skip-merge