[PyTorch] AOTI: generate reused thread_locals when tensors provably have static shape#110892
Closed
swolchok wants to merge 6 commits intogh/swolchok/588/basefrom
Closed
[PyTorch] AOTI: generate reused thread_locals when tensors provably have static shape#110892swolchok wants to merge 6 commits intogh/swolchok/588/basefrom
swolchok wants to merge 6 commits intogh/swolchok/588/basefrom
Conversation
…ave static shape If a Tensor can be reused and has static shape, we can just cache it across iterations. This and the following diff are meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/110892
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit b591468 with merge base 2edc75a ( 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. |
This was referenced Oct 9, 2023
swolchok
added a commit
that referenced
this pull request
Oct 9, 2023
…ave static shape If a Tensor can be reused and has static shape, we can just cache it across iterations. This and the following diff are meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) ghstack-source-id: 203410034 Pull Request resolved: #110892
swolchok
commented
Oct 9, 2023
… provably have static shape" If a Tensor can be reused and has static shape, we can just cache it across iterations. This and the following diff are meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Oct 9, 2023
…ave static shape Pull Request resolved: #110892 If a Tensor can be reused and has static shape, we can just cache it across iterations. This and the following diff are meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. ghstack-source-id: 203460350 @exported-using-ghexport Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/)
desertfire
reviewed
Oct 10, 2023
Contributor
desertfire
left a comment
desertfire
left a comment
There was a problem hiding this comment.
Left comments in the internal diff. Let me know when you think this is ready for review.
bertmaher
approved these changes
Oct 10, 2023
… provably have static shape" If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
Collaborator
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
…als when tensors provably have static shape" If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
swolchok
force-pushed
the
gh/swolchok/588/head
branch
from
ce7b56a to
4e03711
Compare
swolchok
added a commit
that referenced
this pull request
Oct 10, 2023
…ave static shape Pull Request resolved: #110892 If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. ghstack-source-id: 203568593 @exported-using-ghexport Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/)
Collaborator
Merge failedReason: New commits were pushed while merging. Please rerun the merge command. Details for Dev Infra teamRaised by workflow job |
Contributor
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Collaborator
Merge failedReason: 1 mandatory check(s) failed. The first few are: Dig deeper by viewing the failures on hud |
… provably have static shape" If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Oct 12, 2023
…ave static shape Pull Request resolved: #110892 If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. ghstack-source-id: 203869660 @exported-using-ghexport Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/)
…rovably have static shape" If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Oct 12, 2023
…ave static shape Pull Request resolved: #110892 If a Tensor can be reused and has static shape, we can just cache it across iterations. This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning. ghstack-source-id: 203924205 @exported-using-ghexport Differential Revision: [D50023678](https://our.internmc.facebook.com/intern/diff/D50023678/)
Contributor
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 30, 2023
…on-AOT mode" We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 30, 2023
We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
chunyuan-w
added a commit
that referenced
this pull request
Nov 30, 2023
…on-AOT mode" We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
chunyuan-w
added a commit
that referenced
this pull request
Nov 30, 2023
We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
chunyuan-w
added a commit
that referenced
this pull request
Nov 30, 2023
…on-AOT mode" We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
chunyuan-w
added a commit
that referenced
this pull request
Nov 30, 2023
We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 30, 2023
We found performance regression when using cpp wrapper in non-AOT mode due to the change in #110892. #110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. Pull Request resolved: #114741 Approved by: https://github.com/jgong5, https://github.com/desertfire
dmenig
pushed a commit
to dmenig/pytorch
that referenced
this pull request
Dec 21, 2023
We found performance regression when using cpp wrapper in non-AOT mode due to the change in pytorch#110892. pytorch#110892 only handles the buffer cache in AOT mode but removes the `reset` call without checking whether AOT mode is on or off. This PR updates the buffer free change to only happen when `V.graph.aot_mode is True`. Pull Request resolved: pytorch#114741 Approved by: https://github.com/jgong5, https://github.com/desertfire
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
4 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):
If a Tensor can be reused and has static shape, we can just cache it across iterations.
This is meant as a quickly shippable overhead reduction for CPU overhead-bound use cases that we can ship without relying on memory planning.
Differential Revision: D50023678
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler