[ET-VK][runtime] Add prepack cache to avoid duplicate weight prepacking#18361
Open
SS-JIA wants to merge 4 commits intogh/SS-JIA/499/basefrom
Open
[ET-VK][runtime] Add prepack cache to avoid duplicate weight prepacking#18361SS-JIA wants to merge 4 commits intogh/SS-JIA/499/basefrom
SS-JIA wants to merge 4 commits intogh/SS-JIA/499/basefrom
Conversation
When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18361
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 28 PendingAs of commit ccda364 with merge base 38b40bc ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
SS-JIA
pushed a commit
that referenced
this pull request
Mar 20, 2026
When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/) ghstack-source-id: 355089157 Pull Request resolved: #18361
meta-cla
bot
added
the
CLA Signed
This PR needs a
|
…ht prepacking" When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Mar 20, 2026
Pull Request resolved: #18361 When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. ghstack-source-id: 355234968 @exported-using-ghexport Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/)
…ht prepacking" When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Mar 20, 2026
Pull Request resolved: #18361 When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. ghstack-source-id: 355269010 @exported-using-ghexport Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/)
…ht prepacking" When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/) [ghstack-poisoned]
SS-JIA
pushed a commit
that referenced
this pull request
Mar 20, 2026
Pull Request resolved: #18361 When embedding and linear ops share tied weights and both use the same prepacking function (prepack_quantized_linear_weight), the weight gets prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed by (input ValueRef, kernel name) that returns the already-prepacked tensor on cache hit, avoiding the duplicate allocation. ghstack-source-id: 355353466 @exported-using-ghexport Differential Revision: [D97430801](https://our.internmc.facebook.com/intern/diff/D97430801/)
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
1 participant
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):
When embedding and linear ops share tied weights and both use the same
prepacking function (prepack_quantized_linear_weight), the weight gets
prepacked twice, wasting GPU memory. Add a cache to ComputeGraph keyed
by (input ValueRef, kernel name) that returns the already-prepacked
tensor on cache hit, avoiding the duplicate allocation.
Differential Revision: D97430801