Fix torch.compile FunctionalTensor inputs for higherOrderOps #107604
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/107604
Note: Links to docs will display an error until the docs builds have been completed. ✅ 1 Unrelated FailureAs of commit 4b2410e with merge base 67bb3c0 ( UNSTABLE - The following job failed but was likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
ydwu4
added
ciflow/trunk
zou3519
reviewed
Aug 22, 2023
test/dynamo/test_subclasses.py
Outdated
| actual = normalize_gm( | ||
| backend.graphs[exp_n_graph - 1].print_readable(print_output=False) | ||
| ) | ||
| self.assertEqual(actual, exp_graph) |
Contributor
There was a problem hiding this comment.
nit: can we assertExpectedInline somehow on this?
zou3519
reviewed
Aug 22, 2023
zou3519
reviewed
Aug 22, 2023
torch/_subclasses/fake_tensor.py
Outdated
|
|
||
| def is_fake(x): | ||
| if isinstance(x, FakeTensor): | ||
| def is_fake(x, fake_mode=None): |
Contributor
There was a problem hiding this comment.
I don't really understand why we need to pass in a fake_mode to is_fake. Are we saying that a FakeTensor might have a different fake_mode than expected?
Contributor
Author
There was a problem hiding this comment.
Yes, this could happen when input is a FakeTensor created outside of compiled region e.g. make_fx(torch.func.functionalize(compiled_f))(x). But theoretically, FakeTensors should already be fakified by dynamo by the time we call the first is_fake right now. I guess it's mostly to add more safety checks and enforcing an invariant that all fake tensors in the dynamo should has the same fake_mode as current instructionTranslator.
Contributor
There was a problem hiding this comment.
My problem right now is that the is_fake(fake_mode) API doesn't really make sense. If you want to uphold an invariant that all fake tensors have the same fake_mode as the current instruction translator, then instead of complicating the is_fake check we should add an assertion somewhere (maybe post-fakification?) that all FakeTensors have the same fake mode
Contributor
Author
There was a problem hiding this comment.
Yeah, that also sounds good to me! I can change the implementation.
Before this PR, for the added [test](https://github.com/pytorch/pytorch/pull/107604/files#diff-c618f2274b6b5ccc533c580549d2e552edbd9fc5ac0da1aa4b00338525c8f78dR224), which feeds FunctionTensorWrapper inputs to higherOrderOperator, we have an assertion error in this line [code](https://github.com/pytorch/pytorch/pull/107604/files#diff-9f0663783bcd93e948e0491ef61b48123bdc9977bcc632fd707da578df13bfa1R1284). The key difference of this PR is this [line ](https://github.com/pytorch/pytorch/pull/107604/files#diff-9f0663783bcd93e948e0491ef61b48123bdc9977bcc632fd707da578df13bfa1L1263)of check: ```python elif ( isinstance(example_value, FakeTensor) and example_value.fake_mode is tx.fake_mode ): ``` The original intention of it seems to be dealing with case where we want to wrap an fx proxy for an intermediate fake tensor that's already tracked by dynamo and an example value is provided (as is the case for higherOrderOps [here](https://github.com/pytorch/pytorch/blob/main/torch/_dynamo/variables/higher_order_ops.py#L85)). A fakified FunctionalTensorWrapper(FakeTensor) always fails this check. This PR changes it to checking whether it's already fakified by tx.fake_mode. cc voznesenskym penguinwu anijain2305 EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx ipiszy chenyang78 aakhundov [ghstack-poisoned]
zou3519
approved these changes
Aug 22, 2023
Before this PR, for the added [test](https://github.com/pytorch/pytorch/pull/107604/files#diff-c618f2274b6b5ccc533c580549d2e552edbd9fc5ac0da1aa4b00338525c8f78dR224), which feeds FunctionTensorWrapper inputs to higherOrderOperator, we have an assertion error in this line [code](https://github.com/pytorch/pytorch/pull/107604/files#diff-9f0663783bcd93e948e0491ef61b48123bdc9977bcc632fd707da578df13bfa1R1284). The key difference of this PR is this [line ](https://github.com/pytorch/pytorch/pull/107604/files#diff-9f0663783bcd93e948e0491ef61b48123bdc9977bcc632fd707da578df13bfa1L1263)of check: ```python elif ( isinstance(example_value, FakeTensor) and example_value.fake_mode is tx.fake_mode ): ``` The original intention of it seems to be dealing with case where we want to wrap an fx proxy for an intermediate fake tensor that's already tracked by dynamo and an example value is provided (as is the case for higherOrderOps [here](https://github.com/pytorch/pytorch/blob/main/torch/_dynamo/variables/higher_order_ops.py#L85)). A fakified FunctionalTensorWrapper(FakeTensor) always fails this check. This PR changes it to checking whether it's already fakified by tx.fake_mode. cc voznesenskym penguinwu anijain2305 EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx ipiszy chenyang78 aakhundov [ghstack-poisoned]
Before this PR, for the added [test](https://github.com/pytorch/pytorch/pull/107604/files#diff-c618f2274b6b5ccc533c580549d2e552edbd9fc5ac0da1aa4b00338525c8f78dR224), which feeds FunctionTensorWrapper inputs to higherOrderOperator, we have an assertion error in this line [code](https://github.com/pytorch/pytorch/pull/107604/files#diff-9f0663783bcd93e948e0491ef61b48123bdc9977bcc632fd707da578df13bfa1R1284). The key difference of this PR is this [line ](https://github.com/pytorch/pytorch/pull/107604/files#diff-9f0663783bcd93e948e0491ef61b48123bdc9977bcc632fd707da578df13bfa1L1263)of check: ```python elif ( isinstance(example_value, FakeTensor) and example_value.fake_mode is tx.fake_mode ): ``` The original intention of it seems to be dealing with case where we want to wrap an fx proxy for an intermediate fake tensor that's already tracked by dynamo and an example value is provided (as is the case for higherOrderOps [here](https://github.com/pytorch/pytorch/blob/main/torch/_dynamo/variables/higher_order_ops.py#L85)). A fakified FunctionalTensorWrapper(FakeTensor) always fails this check. This PR changes it to checking whether it's already fakified by tx.fake_mode. cc voznesenskym penguinwu anijain2305 EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng Xia-Weiwen wenzhe-nrv jiayisunx ipiszy chenyang78 aakhundov [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 |
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):
Before this PR, for the added test, which feeds FunctionTensorWrapper inputs to higherOrderOperator, we have an assertion error in this line code.
The key difference of this PR is this line of check:
The original intention of it seems to be dealing with case where we want to wrap an fx proxy for an intermediate fake tensor that's produced by some tensor ops and an example value is provided (as is the case for higherOrderOps here). A fakified FunctionalTensorWrapper(FakeTensor) always fails this check. This PR changes it to checking whether it's already fakified by tx.fake_mode.
cc @voznesenskym @penguinwu @anijain2305 @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @aakhundov