Thanks to visit codestin.com
Credit goes to github.com

Skip to content

[ONNX] Fix pow op export [1.5.1] #39791

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
Merged
merged 3 commits into from
Jun 11, 2020
Merged

[ONNX] Fix pow op export [1.5.1] #39791

merged 3 commits into from
Jun 11, 2020

Conversation

malfet
Copy link
Contributor

@malfet malfet commented Jun 10, 2020 •
edited
Loading

Fix pow type cast for opset 9 and update opset 12

Also update ort-nighly as suggested in #39685 (comment)

@neginraoof @malfet
[ONNX] Fix pow op export (pytorch#38065)
25dd704 
Summary:
Fix pow type cast for opset 9 and update opset 12
Pull Request resolved: pytorch#38065

Differential Revision: D21485353

Pulled By: malfet

fbshipit-source-id: 3993e835ffad07b2e6585eb5cf1cb7c8474de2ec
@malfet malfet changed the title Malfet/fix onnx pow 1.5.1 [ONNX] Fix pow op export [1.5.1] Jun 10, 2020
@malfet malfet requested review from seemethere and gchanan June 10, 2020 18:49
@malfet malfet added this to the 1.5.1 milestone Jun 10, 2020
seemethere
seemethere approved these changes Jun 10, 2020
View reviewed changes
Copy link
Member

@seemethere seemethere left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM on green

@dr-ci
Copy link

dr-ci bot commented Jun 10, 2020 •
edited
Loading

💊 CI failures summary and remediations

As of commit d6db5d5 (more details on the Dr. CI page):

  • 2/2 failures possibly* introduced in this PR
    • 1/2 non-CircleCI failure(s)

🕵️ 1 new failure recognized by patterns

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

See CircleCI build pytorch_macos_10_13_py3_test (1/1)

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

Jun 11 14:22:18 RuntimeError: test_cuda failed! 
Jun 11 14:22:18   File "/Users/distiller/workspace/miniconda3/lib/python3.7/site-packages/librosa/util/utils.py", line 15, in <module> 
Jun 11 14:22:18     from .decorators import deprecated 
Jun 11 14:22:18   File "/Users/distiller/workspace/miniconda3/lib/python3.7/site-packages/librosa/util/decorators.py", line 9, in <module> 
Jun 11 14:22:18     from numba.decorators import jit as optional_jit 
Jun 11 14:22:18 ModuleNotFoundError: No module named 'numba.decorators' 
Jun 11 14:22:18 Traceback (most recent call last): 
Jun 11 14:22:18   File "test/run_test.py", line 676, in <module> 
Jun 11 14:22:18     main() 
Jun 11 14:22:18   File "test/run_test.py", line 669, in main 
Jun 11 14:22:18     raise RuntimeError(message) 
Jun 11 14:22:18 RuntimeError: test_cuda failed! 
Jun 11 14:22:18 + cleanup 
Jun 11 14:22:18 + retcode=1 
Jun 11 14:22:18 + set +x

ci.pytorch.org: 1 failed

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 on the GitHub issue tracker or post in the (internal) Dr. CI Users group.

See how this bot performed.

This comment has been revised 12 times.

@seemethere
Copy link
Member

@houseroad looks like this is still failing, in spite of an onnx upgrade, is there any other things that may be holding us back?

@seemethere
Copy link
Member

cc: @neginraoof @KsenijaS @spandantiwari

@KsenijaS
Copy link
Contributor

I tested locally test/onnx/test_pytorch_onnx_onnxruntime.py::TestONNXRuntime_opset11::test_topk_smallest_unsorted
with the same ort version and I cannot reproduce the failure

@spandantiwari
Copy link

@KsenijaS - can you specify which PR/commit upgraded to the latest ORT version in the master CI today? This commit may be needed for 1.5.1 branch to make the test pass.

@KsenijaS
Copy link
Contributor

PR: #36894 upgraded to the latest ORT version

@malfet malfet force-pushed the malfet/fix-onnx-pow-1.5.1 branch from f7ec3e1 to 7bd10e2 Compare June 11, 2020 18:00
@seemethere
Copy link
Member

Looks like even after the bump we're still failing the same test,

Jun 11 19:20:25 =================================== FAILURES ===================================
Jun 11 19:20:25 _____________ TestONNXRuntime_opset11.test_topk_smallest_unsorted ______________
Jun 11 19:20:25 
Jun 11 19:20:25 self = <test_pytorch_onnx_onnxruntime.TestONNXRuntime_opset11 testMethod=test_topk_smallest_unsorted>
Jun 11 19:20:25 
Jun 11 19:20:25     def wrapper(self):
Jun 11 19:20:25         if self.opset_version < min_opset_version:
Jun 11 19:20:25             raise unittest.SkipTest("Skip verify test for unsupported opset_version")
Jun 11 19:20:25 >       return func(self)
Jun 11 19:20:25 
Jun 11 19:20:25 test/onnx/test_pytorch_common.py:54: 
Jun 11 19:20:25 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
Jun 11 19:20:25 test/onnx/test_pytorch_onnx_onnxruntime.py:1491: in test_topk_smallest_unsorted
Jun 11 19:20:25     self.run_test(MyModule(), (x, k))
Jun 11 19:20:25 test/onnx/test_pytorch_onnx_onnxruntime.py:126: in run_test
Jun 11 19:20:25     _run_test(model)
Jun 11 19:20:25 test/onnx/test_pytorch_onnx_onnxruntime.py:122: in _run_test
Jun 11 19:20:25     fixed_batch_size=fixed_batch_size)
Jun 11 19:20:25 test/onnx/test_pytorch_onnx_onnxruntime.py:86: in run_model_test
Jun 11 19:20:25     ort_test_with_input(ort_sess, input_copy, output, rtol, atol)
Jun 11 19:20:25 test/onnx/test_pytorch_onnx_onnxruntime.py:47: in ort_test_with_input
Jun 11 19:20:25     [np.testing.assert_allclose(out, ort_out, rtol=rtol, atol=atol) for out, ort_out in zip(outputs, ort_outs)]
Jun 11 19:20:25 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
Jun 11 19:20:25 
Jun 11 19:20:25 .0 = <zip object at 0x7f9bb8e41888>
Jun 11 19:20:25 
Jun 11 19:20:25 >   [np.testing.assert_allclose(out, ort_out, rtol=rtol, atol=atol) for out, ort_out in zip(outputs, ort_outs)]
Jun 11 19:20:25 E   AssertionError: 
Jun 11 19:20:25 E   Not equal to tolerance rtol=0.001, atol=1e-07
Jun 11 19:20:25 E   
Jun 11 19:20:25 E   Mismatch: 66.7%
Jun 11 19:20:25 E   Max absolute difference: 2.
Jun 11 19:20:25 E   Max relative difference: 2.
Jun 11 19:20:25 E    x: array([1., 2., 3.], dtype=float32)
Jun 11 19:20:25 E    y: array([3., 2., 1.], dtype=float32)
Jun 11 19:20:25 
Jun 11 19:20:25 test/onnx/test_pytorch_onnx_onnxruntime.py:47: AssertionError

We're looking to close out the first RC for 1.5.1 today and this is our last cherry pick, is there any way we can look to get this resolved by EOD today @KsenijaS @spandantiwari?

@KsenijaS
Copy link
Contributor

KsenijaS commented Jun 11, 2020 •
edited
Loading

@seemethere I think this PR #37846 is missing, it updates failing test here: https://github.com/pytorch/pytorch/pull/37846/files#diff-92f53f0d2694170082f0ad81bfc05604R1486

@malfet
Copy link
Contributor Author

malfet commented Jun 11, 2020

@KsenijaS thanks for the link. I will cherry-pick just test fix and if it passes CI merge.

@malfet malfet force-pushed the malfet/fix-onnx-pow-1.5.1 branch from 7bd10e2 to d6db5d5 Compare June 11, 2020 20:48
@seemethere seemethere merged commit 3c31d73 into pytorch:release/1.5 Jun 11, 2020
@malfet malfet deleted the malfet/fix-onnx-pow-1.5.1 branch June 11, 2020 22:26
@seemethere seemethere mentioned this pull request Jun 11, 2020
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@seemethere seemethere seemethere approved these changes

@gchanan gchanan Awaiting requested review from gchanan

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
1.5.1
Development

Successfully merging this pull request may close these issues.
5 participants
@malfet @seemethere @KsenijaS @spandantiwari @neginraoof