[compile] Switch off inference mode during compilation #149321
Closed
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/149321
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 7b4d466 with merge base 8cdb9ad ( UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
ciflow/inductor
module: dynamo
oncall: distributed
cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]
cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]
cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]
jansel
approved these changes
Mar 18, 2025
torch/_dynamo/utils.py
Outdated
| ): | ||
| yield | ||
| else: | ||
| with contextlib.nullcontext(): |
Contributor
There was a problem hiding this comment.
Suggested change
| with contextlib.nullcontext(): |
torch/_dynamo/utils.py
Outdated
| with torch._subclasses.meta_utils.disable_inference_mode_for_fake_prop(): | ||
| yield | ||
| else: | ||
| with contextlib.nullcontext(): |
Contributor
There was a problem hiding this comment.
Suggested change
| with contextlib.nullcontext(): |
zou3519
reviewed
Mar 18, 2025
| torch.no_grad(), | ||
| torch._subclasses.meta_utils.disable_inference_mode_for_fake_prop(), | ||
| ): | ||
| yield |
Contributor
There was a problem hiding this comment.
To check my understanding, you're using maybe_disable_inference_mode to chain context managers together? (torch.inference_mode(False),
torch.no_grad(),
torch._subclasses.meta_utils.disable_inference_mode_for_fake_prop(),)
Contributor
Author
There was a problem hiding this comment.
Yes. Earlier this was done in an inlined manner in the convert_frame.py file but it was making code difficult to read.
zou3519
reviewed
Mar 18, 2025
| and len(flat_arg_fake_tensors) != 0 | ||
| and not has_symbolic_sizes | ||
| and not avoiding_device_init | ||
| and func is not aten._nested_tensor_from_tensor_list.default |
Contributor
Author
There was a problem hiding this comment.
This one is just annoying. There is a fake kernel associated with this op which just raises exception. However, that kernel is called later on. There is a fast path in fake_tensor where if all inputs are constant, then we do something else. And for this op, going through that fast patch causes issues.
zou3519
approved these changes
Mar 18, 2025
PR does following * Turns `inference_mode` to False and `no_grad` for `convert_frame`, if the inference_mode is on globally. * Turns off inference_mode for fake tensor prop. This ensures that converting from real inference tensor to a fake tensor removes the inference-ness. * Graph breaks on is_inference and is_inference_mode_enabled. cc H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]
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 |
facebook-github-bot
pushed a commit
to pytorch/benchmark
that referenced
this pull request
Mar 19, 2025
Summary: PR does following * Turns `inference_mode` to False and `no_grad` for `convert_frame`, if the inference_mode is on globally. * Turns off inference_mode for fake tensor prop. This ensures that converting from real inference tensor to a fake tensor removes the inference-ness. * Graph breaks on is_inference and is_inference_mode_enabled. X-link: pytorch/pytorch#149321 Approved by: https://github.com/jansel, https://github.com/zou3519 Reviewed By: izaitsevfb Differential Revision: D71451966 fbshipit-source-id: 8b5cd2178fe37f2843238efafdb798df7ced2391
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.
Stack from ghstack (oldest at bottom):
PR does following
inference_modeto False andno_gradforconvert_frame, if the inference_mode is on globally.cc @H-Huang @awgu @kwen2501 @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang @amjames