Add pattern to detect for loop indexing into tensor #81056

davidchencsl · 2022-07-07T19:55:18Z

Stack from ghstack (oldest at bottom):

Differential Revision: D37725266

Summary: This commit added a pattern in the pattern matcher to detect if for loop & indexing is used that can be vectorized.

Test Plan: Added a test in test_profiler.py that creates a bootstrap example.

[ghstack-poisoned]

ghstack-source-id: d59a132 Pull Request resolved: #81056

facebook-github-bot · 2022-07-07T19:55:27Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/81056
📄 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 our office hours

✅ No Failures (0 Pending)

As of commit 7294b01 (more details on the Dr. CI page):

Expand to see more

💚 💚 Looks good so far! There are no failures yet. 💚 💚

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.

torch/csrc/autograd/init.cpp

[ghstack-poisoned]

ghstack-source-id: 9f1357e Pull Request resolved: #81056

davidchencsl · 2022-07-08T19:55:18Z

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

Differential Revision: [D37725266](https://our.internmc.facebook.com/intern/diff/D37725266) Summary: This commit added a pattern in the pattern matcher to detect if for loop & indexing is used that can be vectorized. Test Plan: Added a test in test_profiler.py that creates a bootstrap example. [ghstack-poisoned]

davidchencsl · 2022-07-12T20:52:03Z

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

robieta

A couple minor nits, but otherwise looks good.

torch/_C/_autograd.pyi

torch/profiler/_pattern_matcher.py

Differential Revision: [D37725266](https://our.internmc.facebook.com/intern/diff/D37725266) Summary: This commit added a pattern in the pattern matcher to detect if for loop & indexing is used that can be vectorized. Test Plan: Added a test in test_profiler.py that creates a bootstrap example. [ghstack-poisoned]

davidchencsl · 2022-07-15T20:07:35Z

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

Differential Revision: [D37725266](https://our.internmc.facebook.com/intern/diff/D37725266) Summary: This commit added a pattern in the pattern matcher to detect if for loop & indexing is used that can be vectorized. Test Plan: Added a test in test_profiler.py that creates a bootstrap example. [ghstack-poisoned]

davidchencsl · 2022-07-18T22:46:34Z

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

davidchencsl · 2022-07-20T18:32:55Z

@pytorchbot merge

pytorchmergebot · 2022-07-20T18:37:21Z

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

github-actions · 2022-07-20T18:38:05Z

Hey @davidchencsl.
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.

janeyx99 · 2022-07-20T21:00:39Z

@pytorchbot revert -m "broke 10.2 test_forward_ad_nn_functional_conv_transpose3d_cuda_float32 test on trunk https://hud.pytorch.org/pytorch/pytorch/commit/6e621d028766808f3c0bc8c45f776af3bf9cf22f" -c nosignal

======================================================================
2022-07-20T20:45:44.3845308Z FAIL [2.091s]: test_forward_ad_nn_functional_conv_transpose3d_cuda_float32 (__main__.TestCompositeComplianceCUDA)
2022-07-20T20:45:44.3848588Z ----------------------------------------------------------------------
2022-07-20T20:45:44.3849171Z Traceback (most recent call last):
2022-07-20T20:45:44.3849722Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_internal/common_utils.py", line 1883, in wrapper
2022-07-20T20:45:44.3850102Z     method(*args, **kwargs)
2022-07-20T20:45:44.3850583Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_internal/common_utils.py", line 1883, in wrapper
2022-07-20T20:45:44.3850953Z     method(*args, **kwargs)
2022-07-20T20:45:44.3853385Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_internal/common_device_type.py", line 377, in instantiated_test
2022-07-20T20:45:44.3853810Z     result = test(self, **param_kwargs)
2022-07-20T20:45:44.3854428Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_internal/common_device_type.py", line 813, in test_wrapper
2022-07-20T20:45:44.3867042Z     return test(*args, **kwargs)
2022-07-20T20:45:44.3867447Z   File "/var/lib/jenkins/workspace/test/test_ops.py", line 1258, in test_forward_ad
2022-07-20T20:45:44.3867861Z     composite_compliance.check_forward_ad_formula(
2022-07-20T20:45:44.3868483Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_internal/composite_compliance.py", line 567, in check_forward_ad_formula
2022-07-20T20:45:44.3868974Z     assert_equal_fn(actual_primals, expected_primals, equal_nan=True)
2022-07-20T20:45:44.3869572Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_internal/common_utils.py", line 2335, in assertEqual
2022-07-20T20:45:44.3869951Z     assert_equal(
2022-07-20T20:45:44.3870410Z   File "/opt/conda/lib/python3.9/site-packages/torch/testing/_comparison.py", line 1093, in assert_equal
2022-07-20T20:45:44.3870801Z     raise error_metas[0].to_error(msg)
2022-07-20T20:45:44.3871179Z AssertionError: Tensor-likes are not close!
2022-07-20T20:45:44.3871391Z 
2022-07-20T20:45:44.3871526Z Mismatched elements: 1 / 2744 (0.0%)
2022-07-20T20:45:44.3871974Z Greatest absolute difference: 4.57763671875e-05 at index (2, 3, 4, 4) (up to 2e-05 allowed)
2022-07-20T20:45:44.3872504Z Greatest relative difference: 5.68618647611624e-05 at index (2, 3, 4, 4) (up to 2e-05 allowed)
2022-07-20T20:45:44.3872743Z 
2022-07-20T20:45:44.3873013Z ----------------------------------------------------------------------

pytorchmergebot · 2022-07-20T21:02:08Z

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

pytorchmergebot · 2022-07-20T21:02:11Z

Reverting PR 81056 failed due to Command git -C /home/runner/actions-runner/_work/pytorch/pytorch revert --no-edit 6e621d028766808f3c0bc8c45f776af3bf9cf22f returned non-zero exit code 1

Auto-merging test/test_profiler.py
CONFLICT (content): Merge conflict in test/test_profiler.py
Auto-merging torch/_C/_autograd.pyi
CONFLICT (content): Merge conflict in torch/_C/_autograd.pyi
Auto-merging torch/csrc/autograd/init.cpp
CONFLICT (content): Merge conflict in torch/csrc/autograd/init.cpp
Auto-merging torch/profiler/_pattern_matcher.py
CONFLICT (content): Merge conflict in torch/profiler/_pattern_matcher.py
error: could not revert 6e621d0287... Add pattern to detect for loop indexing into tensor (#81056)
hint: After resolving the conflicts, mark them with
hint: "git add/rm <pathspec>", then run
hint: "git revert --continue".
hint: You can instead skip this commit with "git revert --skip".
hint: To abort and get back to the state before "git revert",
hint: run "git revert --abort".

Raised by https://github.com/pytorch/pytorch/actions/runs/2707645595

janeyx99 · 2022-07-20T22:20:08Z

LOL nice the revert didn't go through because it was going to be a false revert--the job was green again after, so I suppose it was a flaky error of some sort. Sorry about the extra noise!

davidchencsl · 2022-07-20T22:22:12Z

Oh my, looks like it does broke a test outside of CI: #81812

Summary: This commit added a pattern in the pattern matcher to detect if for loop & indexing is used that can be vectorized. Pull Request resolved: #81056 Approved by: https://github.com/robieta Test Plan: contbuild & OSS CI, see https://hud.pytorch.org/commit/pytorch/pytorch/6e621d028766808f3c0bc8c45f776af3bf9cf22f Test plan from GitHub: Added a test in test_profiler.py that creates a bootstrap example. Original Phabricator Test Plan: Imported from OSS Reviewed By: jeanschmidt Differential Revision: D37725266 Pulled By: davidchencsl fbshipit-source-id: 896b31d06bee7fa1daf3b32ee6b8340e921086bd

Add pattern to detect for loop indexing into tensor

90fbd49

[ghstack-poisoned]

davidchencsl requested review from albanD and soulitzer as code owners July 7, 2022 19:55

This was referenced Jul 7, 2022

[Profiler] Define pattern matcher structure #80108

Closed

[Profiler] Add Pattern that detects extra cuda copy #80572

Closed

facebook-github-bot added the cla signed label Jul 7, 2022

davidchencsl added a commit that referenced this pull request Jul 7, 2022

Add pattern to detect for loop indexing into tensor

f773a1b

ghstack-source-id: d59a132 Pull Request resolved: #81056

davidchencsl requested review from robieta and removed request for albanD and soulitzer July 7, 2022 19:55

albanD reviewed Jul 8, 2022

View reviewed changes

torch/csrc/autograd/init.cpp Show resolved Hide resolved

Update on "Add pattern to detect for loop indexing into tensor"

ca8fafb

[ghstack-poisoned]

davidchencsl added a commit that referenced this pull request Jul 8, 2022

Add pattern to detect for loop indexing into tensor

d09f2c4

ghstack-source-id: 9f1357e Pull Request resolved: #81056

davidchencsl mentioned this pull request Jul 11, 2022

[Profiler] Add pattern to detect if TF32 is available but not used #81273

Closed

davidchencsl mentioned this pull request Jul 14, 2022

[Profiler] Add speedup estimate for FP32 pattern and Extra CUDA Copy Pattern #81501

Closed

robieta reviewed Jul 15, 2022

View reviewed changes

torch/_C/_autograd.pyi Outdated Show resolved Hide resolved

torch/profiler/_pattern_matcher.py Outdated Show resolved Hide resolved

torch/profiler/_pattern_matcher.py Outdated Show resolved Hide resolved

davidchencsl added 2 commits July 15, 2022 20:00

This was referenced Jul 19, 2022

[TorchTidy] Add pattern to detect if single-tensor implementation optimizer is used #81733

Closed

[TorchTidy] Add Pattern to detect Synchronous Data Loader #81740

Closed

robieta approved these changes Jul 20, 2022

View reviewed changes

pytorchmergebot added the Merged label Jul 20, 2022

pytorchmergebot closed this in 6e621d0 Jul 20, 2022

janeyx99 mentioned this pull request Jul 20, 2022

DISABLED test_forward_ad_nn_functional_conv_transpose3d_cuda_float32 (__main__.TestCompositeComplianceCUDA) #81822

Closed

This was referenced Jul 21, 2022

[TorchTidy] Add pattern to detect if set_to_none is set in zero_grad() #81921

Closed

[TorchTidy] Add pattern to detect if bias is enabled in conv2d followed by batchnorm2d #81941

Closed

facebook-github-bot deleted the gh/davidchencsl/19/head branch July 24, 2022 14:18

Add pattern to detect for loop indexing into tensor #81056

Add pattern to detect for loop indexing into tensor #81056

Uh oh!

Conversation

davidchencsl commented Jul 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented Jul 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

✅ No Failures (0 Pending)

Uh oh!

Uh oh!

davidchencsl commented Jul 8, 2022

Uh oh!

davidchencsl commented Jul 12, 2022

Uh oh!

robieta left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

davidchencsl commented Jul 15, 2022

Uh oh!

davidchencsl commented Jul 18, 2022

Uh oh!

davidchencsl commented Jul 20, 2022

Uh oh!

pytorchmergebot commented Jul 20, 2022

Uh oh!

github-actions bot commented Jul 20, 2022

Uh oh!

janeyx99 commented Jul 20, 2022

Uh oh!

pytorchmergebot commented Jul 20, 2022

Uh oh!

pytorchmergebot commented Jul 20, 2022

Uh oh!

janeyx99 commented Jul 20, 2022

Uh oh!

davidchencsl commented Jul 20, 2022

Uh oh!

Uh oh!

davidchencsl commented Jul 7, 2022 •

edited

Loading

facebook-github-bot commented Jul 7, 2022 •

edited

Loading