8000 Disable use_mkldnn when input is not contiguous for oneDNN by yanbing-j · Pull Request #80864 · pytorch/pytorch · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Disable use_mkldnn when input is not contiguous for oneDNN #80864

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
wants to merge 4 commits into from
Closed

Disable use_mkldnn when input is not contiguous for oneDNN #80864

wants to merge 4 commits into from

Conversation

yanbing-j
Copy link
Collaborator
@yanbing-j yanbing-j commented Jul 5, 2022 •
edited
Loading

Fixes #80837.
This PR is to disable use_mkldnn when input is not contiguous for oneDNN requirement.

@facebook-github-bot
Copy link
Contributor
facebook-github-bot commented Jul 5, 2022 •
edited
Loading

🔗 Helpful links

❌ 1 New Failures

As of commit 8df1c08 (more details on the Dr. CI page):

Expand to see more
  • 1/1 failures introduced in this PR

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages

See GitHub Actions build pull / linux-bionic-py3_7-clang8-xla / test (xla, 1, 1, linux.2xlarge) (1/1)

Step: "Test" (full log | diagnosis details | 🔁 rerun)

2022-07-07T04:00:20.4521831Z RuntimeError: /var...src/aten_xla_bridge.cpp:73 : Check failed: xtensor 
2022-07-07T04:00:20.4518462Z ======================================================================
2022-07-07T04:00:20.4518719Z ERROR [0.009s]: test_put_empty_xla (__main__.TestTorchDeviceTypeXLA)
2022-07-07T04:00:20.4519077Z ----------------------------------------------------------------------
2022-07-07T04:00:20.4519336Z Traceback (most recent call last):
2022-07-07T04:00:20.4519750Z   File "/opt/conda/lib/python3.7/site-packages/torch/testing/_internal/common_device_type.py", line 390, in instantiated_test
2022-07-07T04:00:20.4520032Z     raise rte
2022-07-07T04:00:20.4520402Z   File "/opt/conda/lib/python3.7/site-packages/torch/testing/_internal/common_device_type.py", line 377, in instantiated_test
2022-07-07T04:00:20.4520707Z     result = test(self, **param_kwargs)
2022-07-07T04:00:20.4521163Z   File "/var/lib/jenkins/workspace/xla/test/../../test/test_torch.py", line 3349, in test_put_empty
2022-07-07T04:00:20.4521481Z     self.assertEqual(dst, dst.put_(indices, src, accumulate=accumulate))
2022-07-07T04:00:20.4521831Z RuntimeError: /var/lib/jenkins/workspace/xla/torch_xla/csrc/aten_xla_bridge.cpp:73 : Check failed: xtensor 
2022-07-07T04:00:20.4522109Z *** Begin stack trace ***
2022-07-07T04:00:20.4522336Z 	tensorflow::CurrentStackTrace[abi:cxx11]()
2022-07-07T04:00:20.4522576Z 	torch_xla::bridge::GetXlaTensor(at::Tensor const&)
2022-07-07T04:00:20.4522880Z 	torch_xla::XLANativeFunctions::put_(at::Tensor&, at::Tensor const&, at::Tensor const&, bool)
2022-07-07T04:00:20.4523233Z 	at::_ops::put_::redispatch(c10::DispatchKeySet, at::Tensor&, at::Tensor const&, at::Tensor const&, bool)
2022-07-07T04:00:20.4523467Z 	
2022-07-07T04:00:20.4523721Z 	at::_ops::put_::redispatch(c10::DispatchKeySet, at::Tensor&, at::Tensor const&, at::Tensor const&, bool)
2022-07-07T04:00:20.4523970Z 	
2022-07-07T04:00:20.4524296Z 	at::_ops::put_::call(at::Tensor&, at::Tensor const&, at::Tensor const&, bool)
2022-07-07T04:00:20.4524506Z
This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

@yanbing-j yanbing-j force-pushed the yanbing/fix_conv_slice_input branch from 6005f9d to 15fac54 Compare July 6, 2022 04:23
@yanbing-j yanbing-j force-pushed the yanbing/fix_conv_slice_input branch from 15fac54 to 91809e9 Compare July 6, 2022 04:29
@yanbing-j yanbing-j marked this pull request as ready for review July 7, 2022 04:14
@cpuhrsch cpuhrsch added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label Jul 8, 2022
@yanbing-j yanbing-j added intel This tag is for PR from Intel intel priority matters to intel architecture from performance wise labels Jul 9, 2022
@malfet
Copy link
Contributor
malfet commented Jul 17, 2022

@pytorchbot merge -f

@pytorchmergebot
Copy link
Collaborator

@pytorchbot successfully started a merge job. Check the current status here

@github-actions
Copy link
Contributor

Hey @yanbing-j.
You've committed this PR, but it does not have both a 'release notes: ...' and 'topics: ...' label. Please add one of each to the PR. The 'release notes: ...' label should represent the part of PyTorch that this PR changes (fx, autograd, distributed, etc) and the 'topics: ...' label should represent the kind of PR it is (not user facing, new feature, bug fix, perf improvement, etc). The list of valid labels can be found here for the 'release notes: ...' and here for the 'topics: ...'.
For changes that are 'topic: not user facing' there is no need for a release notes label.

@XiaobingSuper XiaobingSuper mentioned this pull request Jul 17, 2022
Closed
facebook-github-bot pushed a commit that referenced this pull request Jul 18, 2022
@yanbing-j @facebook-github-bot
Disable use_mkldnn when input is not contiguous for oneDNN (#80864) (#…
5d3e2cb 
…80864)

Summary:
Fixes [#80837](#80837).
This PR is to disable use_mkldnn when input is not contiguous for oneDNN requirement.

Pull Request resolved: #80864
Approved by: https://github.com/malfet

Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/4655c3bace4d50b7d02fe0eb0e0fc2a792a518a7

Reviewed By: DanilBaibak

Differential Revision: D37919713

Pulled By: DanilBaibak

fbshipit-source-id: 40a7f5d802498d2a01a702e970982cc40b113e10
@janeyx99
Copy link
Contributor

@pytorchbot revert -m "Reverting due for a perf regression pytorch/benchmark#1040" -c "weird"

@pytorchmergebot
Copy link
Collaborator

@pytorchbot successfully started a revert job. Check the current status here

@pytorchmergebot
Copy link
Collaborator

@yanbing-j your PR has been successfully reverted.

pytorchmergebot added a commit that referenced this pull request Jul 19, 2022
@pytorchmergebot
Revert "Disable use_mkldnn when input is not contiguous for oneDNN (#…
f82b19f 
…80864)"

This reverts commit 4655c3b.

Reverted #80864 on behalf of https://github.com/janeyx99 due to Reverting due for a perf regression pytorch/benchmark#1040
facebook-github-bot pushed a commit that referenced this pull request Jul 20, 2022
@pytorchmergebot @facebook-github-bot
Revert "Disable use_mkldnn when input is not contiguous for oneDNN (#…
1b1bb76 
…80864)"

Summary:
This reverts commit 4655c3b.

Reverted #80864 on behalf of https://github.com/janeyx99 due to Reverting due for a perf regression pytorch/benchmark#1040

Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/f82b19f15befe8f9d6f97d3cba8072963fbb9a43

Reviewed By: jeanschmidt

Differential Revision: D37990809

Pulled By: jeanschmidt

fbshipit-source-id: 0a5c46635e27a1006db3456e37785bf823285a86
@janeyx99 janeyx99 mentioned this pull request Jul 22, 2022
Closed
@mingfeima
Copy link
Collaborator

this would cause performance regression for non-contiguous input, not a good plan.

we need to pass use_channels_last to ideep/oneDNN directly to skip the ambiguity between pytorch and onednn.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@malfet malfet malfet approved these changes

@VitalyFedyunin VitalyFedyunin Awaiting requested review from VitalyFedyunin

@jianyuh jianyuh Awaiting requested review from jianyuh

Assignees
No one assigned
Labels
cla signed intel priority matters to intel architecture from performance wise intel This tag is for PR from Intel Merged open source Reverted triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Got "RuntimeError: y.get_desc().is_nhwc() INTERNAL ASSERT FAILED" while applying conv2d over a transposed tensor
8 participants
@yanbing-j @facebook-github-bot @malfet @pytorchmergebot @janeyx99 @mingfeima @cpuhrsch @pytorchbot
0