[compiled autograd] have context manager turn on dynamo config #138241
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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138241
Note: Links to docs will display an error until the docs builds have been completed. ❌ 17 New Failures, 1 Unrelated FailureAs of commit 7250431 with merge base d531bd5 ( NEW FAILURES - The following jobs have 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. |
pytorch-bot
bot
added
ciflow/inductor
module: compiled autograd
yf225
approved these changes
Oct 17, 2024
yf225
reviewed
Oct 17, 2024
torch/_dynamo/compiled_autograd.py
Outdated
| if not prior: | ||
| compiled_autograd_enabled = False | ||
| torch._dynamo.config.compiled_autograd = prior_config | ||
| torch._C._dynamo.compiled_autograd.set_autograd_compiler(prior) |
Contributor
There was a problem hiding this comment.
maybe we can rebase on top of #138113 to avoid potential merge conflict?
…nfig" We need a way to unify CA checks across ctx manager or torch.compile APIs. Setting this config is actually a no-op, but unifies the state One edge case is user using torch.compile on regions that have .backward() but don't want CA to execute on them and another thread is within the context manager. The general solution to the problem would be to first make the state thread local, then have FSDP create threads that carryover the compiled autograd state cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames rec yf225 [ghstack-poisoned]
…nfig" We need a way to unify CA checks across ctx manager or torch.compile APIs. Setting this config is actually a no-op, but unifies the state One edge case is user using torch.compile on regions that have .backward() but don't want CA to execute on them and another thread is within the context manager. The general solution to the problem would be to first make the state thread local, then have FSDP create threads that carryover the compiled autograd state cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames rec yf225 [ghstack-poisoned]
…nfig" We need a way to unify CA checks across ctx manager or torch.compile APIs. Setting this config is actually a no-op, but unifies the state One edge case is user using torch.compile on regions that have .backward() but don't want CA to execute on them and another thread is within the context manager. The general solution to the problem would be to first make the state thread local, then have FSDP create threads that carryover the compiled autograd state cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames rec yf225 [ghstack-poisoned]
Contributor
Member
Author
|
nvm this doesnt work for reentrants |
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.
We need a way to unify CA checks across ctx manager or torch.compile APIs. Setting this config is actually a no-op, but unifies the state
One edge case is user using torch.compile on regions that have .backward() but don't want CA to execute on them and another thread is within the context manager. The general solution to the problem would be to first make the state thread local, then have FSDP create threads that carryover the compiled autograd state
Stack from ghstack (oldest at bottom):
_compiled_autograd_enabledglobal state variable #138187cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang @amjames @rec @yf225