Updates nonzero's as_tuple behavior to no longer warn. #45413

mruberry · 2020-09-28T04:14:12Z

torch.nonzero is distinct from numpy.nonzero. The latter returns a tensor by default, and the former returns a tuple of tensors. The as_tuple argument was added as part of an intended deprecation process to make torch.nonzero consistent with numpy.nonzero, but this was a confusing change for users. A better deprecation path would be to offer torch.argwhere consistent with numpy.argwhere, which is equivalent to the default torch.nonzero behavior. Once this is offered a change to torch.nonzero should be more straightforward with less user disruption, if we decided that's the correct change to pursue.

dr-ci · 2020-09-28T05:37:16Z

💊 CI failures summary and remediations

As of commit 854368f (more details on the Dr. CI page):

2/3 failures possibly* introduced in this PR
- 1/2 non-CircleCI failure(s)
1/3 broken upstream at merge base 87f98a5 since Sep 29

🕵️ 1 new failure recognized by patterns

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

pytorch_linux_xenial_cuda9_2_cudnn7_py3_gcc5_4_build (1/1)

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

Sep 29 13:48:02 caused by: Connection refused (os error 111)

Sep 29 13:48:02 ++++ extract_trap_cmd 
Sep 29 13:48:02 ++++ printf '%s\n' '' 
Sep 29 13:48:02 +++ printf '%s\n' cleanup 
Sep 29 13:48:02 ++ trap -- ' 
Sep 29 13:48:02 cleanup' EXIT 
Sep 29 13:48:02 ++ [[ pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc5.4-build != *pytorch-win-* ]] 
Sep 29 13:48:02 ++ which sccache 
Sep 29 13:48:02 ++ sccache --stop-server 
Sep 29 13:48:02 Stopping sccache server... 
Sep 29 13:48:02 error: couldn't connect to server 
Sep 29 13:48:02 caused by: Connection refused (os error 111) 
Sep 29 13:48:02 ++ true 
Sep 29 13:48:02 ++ rm /var/lib/jenkins/sccache_error.log 
Sep 29 13:48:02 rm: cannot remove '/var/lib/jenkins/sccache_error.log': No such file or directory 
Sep 29 13:48:02 ++ true 
Sep 29 13:48:02 ++ [[ pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc5.4-build == *rocm* ]] 
Sep 29 13:48:02 ++ SCCACHE_ERROR_LOG=/var/lib/jenkins/sccache_error.log 
Sep 29 13:48:02 ++ SCCACHE_IDLE_TIMEOUT=1200 
Sep 29 13:48:02 ++ RUST_LOG=sccache::server=error 
Sep 29 13:48:02 ++ sccache --start-server 
Sep 29 13:48:02 Starting sccache server...

🚧 1 ongoing upstream failure:

These were probably caused by upstream breakages that are not fixed yet:

pytorch_macos_10_13_py3_test since Sep 29
- 🔁 rerun

ci.pytorch.org: 1 failed

Failed: pr/pytorch-linux-bionic-rocm3.7-py3.6

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 24 times.

facebook-github-bot

@mruberry has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

gchanan · 2020-09-28T22:15:30Z

that deprecation plan does sound superior.

gchanan

does the JIT do the right thing?

mruberry · 2020-09-28T22:34:34Z

does the JIT do the right thing?

That's a good question. Let's extend the test for it.

mruberry · 2020-09-29T10:46:57Z

As expected, scripting with as_tuple does not work (see #45499). This is not a regression. Tracing with as_tuple DOES work.

facebook-github-bot

@mruberry has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

…ero_no_tuple

facebook-github-bot

@mruberry has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

…ero_no_tuple

facebook-github-bot

@mruberry has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

facebook-github-bot · 2020-09-29T20:15:50Z

@mruberry merged this pull request in b66ac1e.

ppwwyyxx · 2020-10-05T23:46:01Z

tools/autograd/templates/python_variable_methods.cpp also needs to be updated. With the current status, 1.7 will have inconsistent warning behavior between torch.nonzero(tensor) and tensor.nonzero().

Summary: Reverting the diff because it has already been fixed in pytorch/pytorch#45413 Reviewed By: myleott Differential Revision: D24511658 fbshipit-source-id: a5561dae50d69a03443ca8a60bebe2cd064e3ee0

Summary: Fixes #44284 #45413 incorrectly left this only partially fixed because it did not update the separate list of method signatures that were deprecated. This PR correctly fixes #44284. A test is added for the behavior, but until the WARN_ONCE flag is added it's toothless. Pull Request resolved: #51618 Reviewed By: ngimel Differential Revision: D26220181 Pulled By: mruberry fbshipit-source-id: 397b47ac7e962d108d8fde0f3dc6468d6327d1c3

Summary: Reverting the diff because it has already been fixed in pytorch/pytorch#45413 Reviewed By: myleott Differential Revision: D24511658 fbshipit-source-id: a5561dae50d69a03443ca8a60bebe2cd064e3ee0

nonzero astuple fix

08c94b5

mruberry requested a review from ngimel September 28, 2020 04:14

ngimel approved these changes Sep 28, 2020

View reviewed changes

mruberry requested a review from gchanan September 28, 2020 05:01

mruberry added this to the 1.7.0 milestone Sep 28, 2020

test fix

dd0882d

facebook-github-bot reviewed Sep 28, 2020

View reviewed changes

gchanan approved these changes Sep 28, 2020

View reviewed changes

Mike Ruberry added 2 commits September 29, 2020 03:45

test update for jit

a6299b0

test update for jit

670abc9

facebook-github-bot reviewed Sep 29, 2020

View reviewed changes

Merge branch 'master' of https://github.com/pytorch/pytorch into nonz…

3ec95df

…ero_no_tuple

facebook-github-bot reviewed Sep 29, 2020

View reviewed changes

Mike Ruberry added 2 commits September 29, 2020 06:34

Merge branch 'master' of https://github.com/pytorch/pytorch into nonz…

03a2373

…ero_no_tuple

test fix

854368f

facebook-github-bot reviewed Sep 29, 2020

View reviewed changes

facebook-github-bot closed this in b66ac1e Sep 29, 2020

facebook-github-bot added the merged label Sep 29, 2020

shinya7y mentioned this pull request Sep 30, 2020

Fix nonzero in NMS for PyTorch 1.6.0 open-mmlab/mmdetection#3867

Merged

gchanan mentioned this pull request Oct 9, 2020

[JIT] torchscript does not support torch.nonzero when as_tuple is True #38718

Closed

gchanan added the module: deprecation label Oct 14, 2020

ppwwyyxx mentioned this pull request Oct 21, 2020

Current behavior of as_tuple argument is inconsistent in nonzero #44284

Closed

mruberry added the Merged label Oct 28, 2020

This was referenced Dec 19, 2020

[doc] correct docs for torch.nonzero #43434

Closed

nonzero() documentation examples issue a warning #43425

Closed

facebook-github-bot deleted the nonzero_no_tuple branch January 27, 2021 18:27

This was referenced Feb 3, 2021

Removes nonzero method warning #51617

Closed

Removes nonzero method warning #51618

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Updates nonzero's as_tuple behavior to no longer warn. #45413

Updates nonzero's as_tuple behavior to no longer warn. #45413

Uh oh!

mruberry commented Sep 28, 2020

Uh oh!

dr-ci bot commented Sep 28, 2020 •

edited

Loading

Uh oh!

facebook-github-bot left a comment

Uh oh!

gchanan commented Sep 28, 2020

Uh oh!

gchanan left a comment

Uh oh!

mruberry commented Sep 28, 2020

Uh oh!

mruberry commented Sep 29, 2020

Uh oh!

facebook-github-bot left a comment

Uh oh!

facebook-github-bot left a comment

Uh oh!

facebook-github-bot left a comment

Uh oh!

facebook-github-bot commented Sep 29, 2020

Uh oh!

ppwwyyxx commented Oct 5, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Updates nonzero's as_tuple behavior to no longer warn. #45413

Updates nonzero's as_tuple behavior to no longer warn. #45413

Uh oh!

Conversation

mruberry commented Sep 28, 2020

Uh oh!

dr-ci bot commented Sep 28, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CI failures summary and remediations

🕵️ 1 new failure recognized by patterns

pytorch_linux_xenial_cuda9_2_cudnn7_py3_gcc5_4_build (1/1)

🚧 1 ongoing upstream failure:

ci.pytorch.org: 1 failed

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

gchanan commented Sep 28, 2020

Uh oh!

gchanan left a comment

Choose a reason for hiding this comment

Uh oh!

mruberry commented Sep 28, 2020

Uh oh!

mruberry commented Sep 29, 2020

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Sep 29, 2020

Uh oh!

ppwwyyxx commented Oct 5, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

dr-ci bot commented Sep 28, 2020 •

edited

Loading