enabled AT_USE_JITERATOR() for `tan` and `tanh` kernels. #102427

parth-desai · 2023-05-27T04:55:04Z

This MR fixes the test failures for jiterator implemenation for tan and tanh Unary kernels as mentioned in #100842.

The failures were fixed by adjusting tolerances but some failures were in test_unary_ufuncs.py required adjusting input values as well. Since the jiterator kernels are using libstdc++, the supported input range is smaller than thrust implementation.

pytorch-bot · 2023-05-27T04:55:07Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/102427

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ You can merge normally! (3 Unrelated Failures)

As of commit 8b29a12 with merge base c323944 ():

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

linux-foundation-easycla · 2023-05-27T04:55:10Z

The committers listed above are authorized under a signed CLA.

✅ login: parth-desai / name: Parth Desai (8c82f69, f79bdaa, 2398bfb, 9cf7592, 2ea1685, f4a7f42, 4b0a027, 4c757ff, cf04403, f81ec71, 13ea79c, 97b9196, f2f61c1, 492f169, bd05bd2, 01615ef, ca9eab5, 481ccae, bad01a2, 6884e84, 8b29a12)

test/test_foreach.py

parth-desai · 2023-05-30T03:50:27Z

@mruberry @peterbell10 @ngimel Can you let me know if i need to format my code or update documentation anywhere for this change?

peterbell10

The test failures look real, e.g.

test_unary_ufuncs.py::TestUnaryUfuncsCUDA::test_reference_numerics_extremal_asinh_cuda_complex64 - AssertionError: Tensor-likes are not close!

test/test_foreach.py

torch/testing/_internal/common_methods_invocations.py

parth-desai · 2023-06-15T01:40:53Z

@peterbell10 Could you take a look at this again?

peterbell10 · 2023-06-15T14:58:59Z

@pytorchbot rebase

pytorchmergebot · 2023-06-15T15:01:06Z

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

pytorchmergebot · 2023-06-15T15:01:10Z

Rebase failed due to Command git -C /home/runner/work/pytorch/pytorch rebase refs/remotes/origin/viable/strict pull/102427/head returned non-zero exit code 1

Rebasing (1/10)
Rebasing (2/10)
Rebasing (3/10)
Rebasing (4/10)
Auto-merging test/test_foreach.py
Auto-merging test/test_unary_ufuncs.py
Auto-merging torch/testing/_internal/common_methods_invocations.py
CONFLICT (content): Merge conflict in torch/testing/_internal/common_methods_invocations.py
error: could not apply a49f5db04f0... Added tolerance decorator to opinfo
hint: Resolve all conflicts manually, mark them as resolved with
hint: "git add/rm <conflicted_files>", then run "git rebase --continue".
hint: You can instead skip this commit: run "git rebase --skip".
hint: To abort and get back to the state before "git rebase", run "git rebase --abort".
Could not apply a49f5db04f0... Added tolerance decorator to opinfo

Raised by https://github.com/pytorch/pytorch/actions/runs/5280395913

peterbell10 · 2023-06-15T15:02:21Z

@parth-desai would you mind resolving the merge conflicts? I'm hoping a rebase would resolve the CI issues.

parth-desai · 2023-06-18T23:58:37Z

@parth-desai would you mind resolving the merge conflicts? I'm hoping a rebase would resolve the CI issues.

@peterbell10 I merged the latest main to my branch. Please review again.

peterbell10 · 2023-06-19T16:34:17Z

This test failure looks related:

2023-06-19T15:19:25.1218162Z =================================== FAILURES ===================================
2023-06-19T15:19:25.1218410Z _ TestUnaryUfuncsCUDA.test_reference_numerics_extremal_nn_functional_tanhshrink_cuda_bfloat16 _
2023-06-19T15:19:25.1218592Z Traceback (most recent call last):
2023-06-19T15:19:25.1218848Z   File "/var/lib/jenkins/workspace/test/test_unary_ufuncs.py", line 315, in test_reference_numerics_extremal
2023-06-19T15:19:25.1219016Z     self._test_reference_numerics(dtype, op, tensors)
2023-06-19T15:19:25.1219246Z   File "/var/lib/jenkins/workspace/test/test_unary_ufuncs.py", line 260, in _test_reference_numerics
2023-06-19T15:19:25.1219375Z     _helper_reference_numerics(
2023-06-19T15:19:25.1219613Z   File "/var/lib/jenkins/workspace/test/test_unary_ufuncs.py", line 206, in _helper_reference_numerics
2023-06-19T15:19:25.1219749Z     self.assertEqualHelper(
2023-06-19T15:19:25.1219974Z   File "/var/lib/jenkins/workspace/test/test_unary_ufuncs.py", line 171, in assertEqualHelper
2023-06-19T15:19:25.1220077Z     self.assertEqual(
2023-06-19T15:19:25.1220472Z   File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 3099, in assertEqual
2023-06-19T15:19:25.1220610Z     raise error_metas[0].to_error(
2023-06-19T15:19:25.1220830Z AssertionError: Tensor-likes are not close!
2023-06-19T15:19:25.1220854Z 
2023-06-19T15:19:25.1220993Z Mismatched elements: 565 / 16384 (3.4%)

It's strange though, because this PR should only effect complex dtypes, but the test is for bfloat16.

parth-desai · 2023-06-21T16:53:47Z

@peterbell10 I couldn't reproduce the error on my local machine. I even tried the CI docker image as well as build scripts. Any suggestions on fixing this test?

peterbell10 · 2023-06-21T19:43:04Z

@pytorchbot rebase

pytorchmergebot · 2023-06-21T19:45:47Z

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

pytorchmergebot · 2023-06-21T19:45:51Z

Tried to rebase and push PR #102427, but it was already up to date. Try rebasing against main by issuing:
@pytorchbot rebase -b main

peterbell10

Just FYI, main often has some test failures from PRs that need to be reverted. You're best rebasing against viable/strict, which should be the most recent commit where CI passed. That should improve the signal-to-noise on CI.

I've set the keep-going flag so you should see all the tests failures, and it looks like there are still quite a lot of relevant CI failures for complex dtypes.

peterbell10 · 2023-06-21T22:27:44Z

aten/src/ATen/native/cuda/UnaryGeometricAcosKernel.cu

You should remove these comments as well.

peterbell10 · 2023-06-21T22:29:33Z

torch/testing/_internal/common_methods_invocations.py

The bfloat16 failure is due to this toleranceOverride taking precedence over the one above for bfloat16. I can get the test to pass locally by copying bfloat16's tol term from above here.

parth-desai · 2023-06-25T05:33:20Z

@peterbell10 Thanks for the pointer. I didn't realize my environment had issues and my build was not running cuda tests. Anyways I have fixed the bfloat16 issue and rebased with viable/strict. Could you run the CI again to see if there are any further failures?

parth-desai · 2023-09-01T16:27:13Z

@peterbell10 The failures now seem to be CI infrastructure failures. Can you take a look?

huydhn · 2023-09-01T17:16:11Z

@pytorchbot drci

huydhn · 2023-09-01T17:16:45Z

@pytorchbot merge

pytorchmergebot · 2023-09-01T17:18:19Z

Merge started

Your 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

Advanced Debugging

Check the merge workflow status
here

pytorchmergebot · 2023-09-01T17:18:23Z

Merge failed

Reason: 2 mandatory check(s) failed. The first few are:

pull / linux-focal-cuda11.8-py3.10-gcc9 / test (distributed, 1, 3, linux.8xlarge.nvidia.gpu)
pull / linux-focal-cuda11.8-py3.10-gcc9 / test (distributed, 2, 3, linux.8xlarge.nvidia.gpu)

Dig deeper by viewing the failures on hud

Details for Dev Infra team

Raised by workflow job

Failing merge rule: Core Maintainers

parth-desai · 2023-09-01T17:54:57Z

@huydhn Any suggestions on how to fix these?

huydhn · 2023-09-01T19:14:33Z

@pytorchbot merge -f 'Bypass unrelated issues as Dr.CI has classified them as flaky'

Sorry I was about to force merge your change but got side tracked by other tasks

pytorchmergebot · 2023-09-01T19:16:03Z

Merge started

Your change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use -f as last resort and instead consider -i/--ignore-current to continue the merge ignoring current failures. This will allow currently pending tests to finish and report signal before the merge.

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

huydhn · 2023-09-01T19:31:40Z

And thank you for your first contribution to PyTorch (IIRC) :)

parth-desai · 2023-09-01T19:36:45Z

thank you @huydhn @peterbell10 for your help.

pytorch-bot bot added the release notes: foreach_frontend release notes category label May 27, 2023

pytorchbot added the open source label May 27, 2023

lezcano requested a review from peterbell10 May 28, 2023 17:56

peterbell10 added the module: testing Issues related to the torch.testing module (not tests) label May 29, 2023

peterbell10 reviewed May 29, 2023

View reviewed changes

test/test_foreach.py Outdated Show resolved Hide resolved

parth-desai marked this pull request as ready for review May 29, 2023 23:15

parth-desai requested review from mruberry and ngimel as code owners May 29, 2023 23:15

peterbell10 reviewed May 30, 2023

View reviewed changes

test/test_foreach.py Outdated Show resolved Hide resolved

torch/testing/_internal/common_methods_invocations.py Outdated Show resolved Hide resolved

albanD added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label May 31, 2023

parth-desai force-pushed the fix-jitterator-unary-op branch from 0e65d90 to dd03d66 Compare June 18, 2023 23:57

peterbell10 added the keep-going Don't stop on first failure, keep running tests until the end label Jun 19, 2023

parth-desai force-pushed the fix-jitterator-unary-op branch from dd03d66 to bc8f744 Compare June 21, 2023 16:58

peterbell10 reviewed Jun 21, 2023

View reviewed changes

parth-desai force-pushed the fix-jitterator-unary-op branch from bc8f744 to 5b1e327 Compare June 25, 2023 05:29

parth-desai added 9 commits August 29, 2023 19:36

fixed tanhshrink cuda bfloat16 failure

97b9196

removed stale comments

f2f61c1

fixed test failures for large and small reference numerics

492f169

Added thrust implementation

bd05bd2

moved to thrust impl

01615ef

reverted thrust related changes

ca9eab5

reverted changes for asin and acos ops

481ccae

fixed failures with ref ops

bad01a2

fixed tolerances

6884e84

parth-desai force-pushed the fix-jitterator-unary-op branch from f77ead2 to 6884e84 Compare August 30, 2023 03:19

removed expectedfailure for macos

8b29a12

pytorchmergebot added the merging label Sep 1, 2023

pytorchmergebot removed the merging label Sep 1, 2023

pytorchmergebot added the merging label Sep 1, 2023

pytorchmergebot added Merged and removed merging labels Sep 1, 2023

pytorchmergebot closed this in 2d9a828 Sep 1, 2023

peterbell10 mentioned this pull request Sep 22, 2023

[foreach] Fix 0-size handling for real for real #109402

Closed

janeyx99 mentioned this pull request Sep 25, 2023

tan/tanh discrepancies with complex due to jiterator #110014

Open

enabled AT_USE_JITERATOR() for tan and tanh kernels. #102427

enabled AT_USE_JITERATOR() for tan and tanh kernels. #102427

Uh oh!

Conversation

parth-desai commented May 27, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented May 27, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/102427

✅ You can merge normally! (3 Unrelated Failures)

Uh oh!

linux-foundation-easycla bot commented May 27, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

parth-desai commented May 30, 2023

Uh oh!

peterbell10 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

parth-desai commented Jun 15, 2023

Uh oh!

peterbell10 commented Jun 15, 2023

Uh oh!

pytorchmergebot commented Jun 15, 2023

Uh oh!

pytorchmergebot commented Jun 15, 2023

Uh oh!

peterbell10 commented Jun 15, 2023

Uh oh!

parth-desai commented Jun 18, 2023

Uh oh!

peterbell10 commented Jun 19, 2023

Uh oh!

parth-desai commented Jun 21, 2023

Uh oh!

peterbell10 commented Jun 21, 2023

Uh oh!

pytorchmergebot commented Jun 21, 2023

Uh oh!

pytorchmergebot commented Jun 21, 2023

Uh oh!

peterbell10 left a comment

Choose a reason for hiding this comment

Uh oh!

peterbell10 Jun 21, 2023

Choose a reason for hiding this comment

Uh oh!

peterbell10 Jun 21, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

parth-desai commented Jun 25, 2023

Uh oh!

parth-desai commented Sep 1, 2023

Uh oh!

huydhn commented Sep 1, 2023

Uh oh!

huydhn commented Sep 1, 2023

Uh oh!

pytorchmergebot commented Sep 1, 2023

Merge started

Uh oh!

pytorchmergebot commented Sep 1, 2023

Merge failed

Uh oh!

parth-desai commented Sep 1, 2023

Uh oh!

huydhn commented Sep 1, 2023

Uh oh!

pytorchmergebot commented Sep 1, 2023

Merge started

Uh oh!

huydhn commented Sep 1, 2023

Uh oh!

parth-desai commented Sep 1, 2023

enabled AT_USE_JITERATOR() for `tan` and `tanh` kernels. #102427

enabled AT_USE_JITERATOR() for `tan` and `tanh` kernels. #102427

parth-desai commented May 27, 2023 •

edited

Loading

pytorch-bot bot commented May 27, 2023 •

edited

Loading

linux-foundation-easycla bot commented May 27, 2023 •

edited

Loading

peterbell10 Jun 21, 2023 •

edited

Loading