[ET-VK] Fix use-after-free in PrepackNode when TensorRefs are shared#18914
Merged
[ET-VK] Fix use-after-free in PrepackNode when TensorRefs are shared#18914
Conversation
Pull Request resolved: #18906 When a model has shared/tied weights (e.g. tied embeddings in transformers), the serialization deduplicates them into a single TensorRef that multiple PrepackNodes reference. Previously, `PrepackNode::create_staging_buffer()` called `tref->free_buffer()` unconditionally after copying weight data to a GPU staging buffer. This meant the first PrepackNode to execute would free the underlying host memory, and subsequent PrepackNodes sharing the same TensorRef would read from a dangling pointer — producing garbage/NaN values in prepacked weight and bias tensors on the GPU. The fix adds a `prepack_use_count` field to `TensorRef` that tracks how many PrepackNodes still need to read from it. Each PrepackNode increments the count in its constructor and decrements it after copying data. The buffer is only freed when the count reaches zero. This preserves the original eager-free behavior for non-shared weights (freeing immediately after the single consumer copies) while correctly deferring the free for shared weights until the last consumer is done — avoiding both the use-after-free and unnecessary peak memory increase. ghstack-source-id: 367726483 @exported-using-ghexport Differential Revision: [D101009402](https://our.internmc.facebook.com/intern/diff/D101009402/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18914
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 2 Pending, 1 Unrelated FailureAs of commit 21a359e with merge base 89e6416 ( NEW FAILURE - The following job has failed:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
kirklandsign
approved these changes
Apr 15, 2026
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.
This PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #18906 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/520/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/520/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/main
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/520/orig
Differential Revision: D101009402
@diff-train-skip-merge