Modify DispatchKeyExtractor to also work for optional Tensors #58283
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
💊 CI failures summary and remediationsAs of commit 91bf4eb (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions to the (internal) Dr. CI Users group. |
|
@Chillee has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
bhosmer
approved these changes
May 14, 2021
Codecov Report
@@ Coverage Diff @@
## master #58283 +/- ##
==========================================
- Coverage 76.84% 76.80% -0.04%
==========================================
Files 1986 1986
Lines 197902 198388 +486
==========================================
+ Hits 152079 152378 +299
- Misses 45823 46010 +187 |
bhosmer
pushed a commit
that referenced
this pull request
May 14, 2021
Correctly extract dispatch keys from optional Tensors in unboxed calls. Companion to #58283 which does the same for boxed calls. E.g. fixes the following repro case (from @zou3519 and @Chillee): ``` import torch # make a quantized tensor r = torch.rand(3, 2, dtype=torch.float) * 4 - 2 scale = 0.02 zero_point = 2 quantized = torch.quantize_per_tensor(r, scale, zero_point, torch.quint8) # The following invokes the quantized key's version of the clamp. # The quantized key doesn't actually have a kernel registered for clamp, # so it throws an error to us. torch.clamp(quantized, quantized, quantized) # However, the following straight up calls into native::clamp (e.g. the CPU # kernel) instead of dispatching on the quantized key and giving us an error: cpu_tensor = torch.randn(3, 2) torch.clamp(cpu_tensor, quantized, quantized) ``` Differential Revision: [D28436822](https://our.internmc.facebook.com/intern/diff/D28436822) [ghstack-poisoned]
krshrimali
pushed a commit
to krshrimali/pytorch
that referenced
this pull request
May 19, 2021
…h#58283) Summary: Pull Request resolved: pytorch#58283 Reviewed By: bhosmer Differential Revision: D28436443 Pulled By: Chillee fbshipit-source-id: ba6aae74e8ec3c5a6157cc4517c29b36bdd4a30d
This was referenced Jun 22, 2021
zou3519
added a commit
that referenced
this pull request
Jun 25, 2021
Fixes #60461. Previously, when one calls `self.index(indices)` using a regular `self` Tensor and a `BatchedTensor` indices the dispatcher would not dispatch to the Batched key. This is because the dispatcher did not extract dispatch keys from `indices`. Similar #58283 and #58296, this PR modifies the dispatcher to extract dispatch keys from List[Optional[Tensor]] arguments. We do this for both boxed and unboxed kernels. Test Plan: - run the test case in https://gist.github.com/zou3519/4421df7c5271376a0ef53ca857b18740 (requires functorch). After this PR, it raises `RuntimeError: Batching rule not implemented for aten::index.Tensor. We could not generate a fallback.`, which shows that dispatch happened on the Batched key. - Taking suggestions for how to write a test for this in core [ghstack-poisoned]
zou3519
added a commit
that referenced
this pull request
Jun 25, 2021
Fixes #60461. Previously, when one calls `self.index(indices)` using a regular `self` Tensor and a `BatchedTensor` indices the dispatcher would not dispatch to the Batched key. This is because the dispatcher did not extract dispatch keys from `indices`. Similar #58283 and #58296, this PR modifies the dispatcher to extract dispatch keys from List[Optional[Tensor]] arguments. We do this for both boxed and unboxed kernels. Test Plan: - run the test case in https://gist.github.com/zou3519/4421df7c5271376a0ef53ca857b18740 (requires functorch). After this PR, it raises `RuntimeError: Batching rule not implemented for aten::index.Tensor. We could not generate a fallback.`, which shows that dispatch happened on the Batched key. - Taking suggestions for how to write a test for this in core ghstack-source-id: 1dc29ae Pull Request resolved: #60787
zou3519
added a commit
that referenced
this pull request
Jul 7, 2021
Fixes #60461. Previously, when one calls `self.index(indices)` using a regular `self` Tensor and a `BatchedTensor` indices the dispatcher would not dispatch to the Batched key. This is because the dispatcher did not extract dispatch keys from `indices`. Similar #58283 and #58296, this PR modifies the dispatcher to extract dispatch keys from List[Optional[Tensor]] arguments. We do this for both boxed and unboxed kernels. Test Plan: - run the test case in https://gist.github.com/zou3519/4421df7c5271376a0ef53ca857b18740 (requires functorch). After this PR, it raises `RuntimeError: Batching rule not implemented for aten::index.Tensor. We could not generate a fallback.`, which shows that dispatch happened on the Batched key. - Taking suggestions for how to write a test for this in core ghstack-source-id: 2c2a432 Pull Request resolved: #60787
zou3519
added a commit
that referenced
this pull request
Jul 7, 2021
…nal[Tensor]] for dispatch" Fixes #60461. Previously, when one calls `self.index(indices)` using a regular `self` Tensor and a `BatchedTensor` indices the dispatcher would not dispatch to the Batched key. This is because the dispatcher did not extract dispatch keys from `indices`. Similar #58283 and #58296, this PR modifies the dispatcher to extract dispatch keys from List[Optional[Tensor]] arguments. We do this for both boxed and unboxed kernels. Test Plan: - run the test case in https://gist.github.com/zou3519/4421df7c5271376a0ef53ca857b18740 (requires functorch). After this PR, it raises `RuntimeError: Batching rule not implemented for aten::index.Tensor. We could not generate a fallback.`, which shows that dispatch happened on the Batched key. - Taking suggestions for how to write a test for this in core Differential Revision: [D29438611](https://our.internmc.facebook.com/intern/diff/D29438611) [ghstack-poisoned]
zou3519
added a commit
that referenced
this pull request
Jul 7, 2021
… dispatch" Fixes #60461. Previously, when one calls `self.index(indices)` using a regular `self` Tensor and a `BatchedTensor` indices the dispatcher would not dispatch to the Batched key. This is because the dispatcher did not extract dispatch keys from `indices`. Similar #58283 and #58296, this PR modifies the dispatcher to extract dispatch keys from List[Optional[Tensor]] arguments. We do this for both boxed and unboxed kernels. Test Plan: - run the test case in https://gist.github.com/zou3519/4421df7c5271376a0ef53ca857b18740 (requires functorch). After this PR, it raises `RuntimeError: Batching rule not implemented for aten::index.Tensor. We could not generate a fallback.`, which shows that dispatch happened on the Batched key. - Taking suggestions for how to write a test for this in core Differential Revision: [D29438611](https://our.internmc.facebook.com/intern/diff/D29438611) [ghstack-poisoned]
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 8, 2021
…60787) Summary: Pull Request resolved: #60787 Fixes #60461. Previously, when one calls `self.index(indices)` using a regular `self` Tensor and a `BatchedTensor` indices the dispatcher would not dispatch to the Batched key. This is because the dispatcher did not extract dispatch keys from `indices`. Similar #58283 and #58296, this PR modifies the dispatcher to extract dispatch keys from List[Optional[Tensor]] arguments. We do this for both boxed and unboxed kernels. Test Plan: - run the test case in https://gist.github.com/zou3519/4421df7c5271376a0ef53ca857b18740 (requires functorch). After this PR, it raises `RuntimeError: Batching rule not implemented for aten::index.Tensor. We could not generate a fallback.`, which shows that dispatch happened on the Batched key. - Taking suggestions for how to write a test for this in core Reviewed By: jbschlosser Differential Revision: D29438611 Pulled By: zou3519 fbshipit-source-id: 77e182f763e18aa3fa857eebafa8b7f83384db71
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.
No description provided.