[DTensor] Test pointwise partial propagation by wconstab · Pull Request #174000 · pytorch/pytorch

wconstab · 2026-01-31T18:39:00Z

Stack from ghstack (oldest at bottom):

Add a bunch of missing test cases.

These were all 'missing' in my single-sim poitnwise PR (discovered
missing using the sharding validator). The pointwise rule on main
(non-single-dim) is not missing these, but we lack the test coverage.
Landing this first helps ensure we don't regress when refactoring.

Add a bunch of missing test cases. These were all 'missing' in my single-sim poitnwise PR (discovered missing using the sharding validator). The pointwise rule on main (non-single-dim) is not missing these, but we lack the test coverage. Landing this first helps ensure we don't regress when refactoring. [ghstack-poisoned]

pytorch-bot · 2026-01-31T18:39:03Z

This PR needs a `release notes:` label

If your changes are user facing and intended to be a part of release notes, please use a label starting with release notes:.

If not, please add the topic: not user facing label.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "topic: not user facing"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

pytorch-bot · 2026-01-31T18:39:06Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/174000

📄 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

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

❌ 4 New Failures, 1 Unrelated Failure

As of commit feb081b with merge base 4b0f7fb ():

NEW FAILURES - The following jobs have failed:

Lint / lintrunner-noclang-all / linux-job (gh)
Error computing the main repository mapping: no such package '@rules_cuda//cuda': java.io.IOException: Error downloading [https://github.com/tensorflow/runtime/archive/b1c7cce21ba4661c17ac72421c6a0e2015e7bef3.tar.gz] to /var/lib/jenkins/.cache/bazel/_bazel_jenkins/5ef3ffe67311b33751bca84ac3d8bf1f/external/rules_cuda/temp52347880342467077/b1c7cce21ba4661c17ac72421c6a0e2015e7bef3.tar.gz: GET returned 502 Bad Gateway or Proxy Error
pull / linux-jammy-py3.10-gcc11 / test (distributed, 1, 2, linux.2xlarge) (gh)
test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_add_partial_avg_with_replicate
trunk / linux-jammy-cuda12.8-py3.10-gcc11 / test (distributed, 2, 3, linux.g4dn.12xlarge.nvidia.gpu) (gh)
test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_add_partial_avg_with_replicate
trunk / linux-jammy-cuda13.0-py3.10-gcc11 / test (distributed, 3, 3, linux.g4dn.12xlarge.nvidia.gpu) (gh)
test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_add_partial_avg_with_replicate

FLAKY - The following job failed but was likely due to flakiness present on trunk:

pull / linux-jammy-py3.14t-clang15 / test (crossref, 1, 2, linux.2xlarge) (gh) (similar failure)
test/test_serialization.py::TestSerialization::test_serialization_4gb_file

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

Skylion007 · 2026-01-31T23:19:11Z

+        if dist.is_initialized():
+            dist.destroy_process_group()
+
+    def test_add_partial_max_with_replicate(self):


Couldnt almost all these tests be parameterized?

Skylion007 · 2026-01-31T23:21:05Z

+        Mathematically: max(a0, a1) + r = max(a0 + r, a1 + r)
+        Adding a replicated constant to P(max) preserves the max structure.
+        """
+        with LocalTensorMode(frozenset(range(self.world_size))):


Feels like this could be a function decorator or be part of setup/teardown

Add a bunch of missing test cases. These were all 'missing' in my single-sim poitnwise PR (discovered missing using the sharding validator). The pointwise rule on main (non-single-dim) is not missing these, but we lack the test coverage. Landing this first helps ensure we don't regress when refactoring. [ghstack-poisoned]

wconstab · 2026-02-02T22:51:18Z

@pytorchbot merge

pytorchmergebot · 2026-02-02T23:22:55Z

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 · 2026-02-02T23:23:06Z

Merge failed

Reason: 1 mandatory check(s) failed. The first few are:

Lint / Get changed files / get-changed-files

Dig deeper by viewing the failures on hud

Details for Dev Infra team

Raised by workflow job

Failing merge rule: Core Maintainers

Add a bunch of missing test cases. These were all 'missing' in my single-sim poitnwise PR (discovered missing using the sharding validator). The pointwise rule on main (non-single-dim) is not missing these, but we lack the test coverage. Landing this first helps ensure we don't regress when refactoring. [ghstack-poisoned]

wconstab · 2026-02-03T04:32:22Z

@pytorchbot merge

pytorchmergebot · 2026-02-03T04:36:27Z

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

jeffdaily · 2026-02-03T17:02:23Z

failures on cuda and rocm

test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_sub_replicate_partial_min_gives_partial_max GH job link HUD commit link

jeffdaily · 2026-02-03T17:02:47Z

@pytorchbot revert -c nosignal -m "test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_sub_replicate_partial_min_gives_partial_max GH job link HUD commit link"

pytorchmergebot · 2026-02-03T17:04:37Z

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

pytorchmergebot · 2026-02-03T17:04:43Z

@wconstab your PR has been successfully reverted.

This reverts commit e2d81fb. Reverted #174000 on behalf of https://github.com/jeffdaily due to test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_sub_replicate_partial_min_gives_partial_max [GH job link](https://github.com/pytorch/pytorch/actions/runs/21617187245/job/62324663867) [HUD commit link](https://hud.pytorch.org/pytorch/pytorch/commit/e2d81fb2c5d8119b55e8e812da9229320928c16d) ([comment](#174000 (comment)))

izaitsevfb · 2026-02-03T17:05:27Z

hud link for failures

malfet · 2026-02-03T17:17:42Z

Thanks for revert @jeffdaily !
@izaitsevfb do you know why auto-revert bot have not picked it up?

izaitsevfb · 2026-02-03T17:44:41Z

Thanks for revert @jeffdaily ! @izaitsevfb do you know why auto-revert bot have not picked it up?

@malfet, yes, this is a case when a new broken test is added (test_pointwise_ops in this case), and currently autorevert doesn't support this situation (it doesn't have a signal from this specific test from the base revision). we have plans to cover that on the job level signals, but didn't have a chance to do it yet

wconstab · 2026-02-06T20:52:36Z

closing as @anshul-si will take over

Add a bunch of missing test cases. These were all 'missing' in my single-sim poitnwise PR (discovered missing using the sharding validator). The pointwise rule on main (non-single-dim) is not missing these, but we lack the test coverage. Landing this first helps ensure we don't regress when refactoring. Pull Request resolved: pytorch#174000 Approved by: https://github.com/Skylion007

This reverts commit e2d81fb. Reverted pytorch#174000 on behalf of https://github.com/jeffdaily due to test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_sub_replicate_partial_min_gives_partial_max [GH job link](https://github.com/pytorch/pytorch/actions/runs/21617187245/job/62324663867) [HUD commit link](https://hud.pytorch.org/pytorch/pytorch/commit/e2d81fb2c5d8119b55e8e812da9229320928c16d) ([comment](pytorch#174000 (comment)))

Add a bunch of missing test cases. These were all 'missing' in my single-sim poitnwise PR (discovered missing using the sharding validator). The pointwise rule on main (non-single-dim) is not missing these, but we lack the test coverage. Landing this first helps ensure we don't regress when refactoring. Pull Request resolved: pytorch#174000 Approved by: https://github.com/Skylion007

This reverts commit e2d81fb. Reverted pytorch#174000 on behalf of https://github.com/jeffdaily due to test/distributed/tensor/test_pointwise_ops.py::PointwisePartialsTest::test_sub_replicate_partial_min_gives_partial_max [GH job link](https://github.com/pytorch/pytorch/actions/runs/21617187245/job/62324663867) [HUD commit link](https://hud.pytorch.org/pytorch/pytorch/commit/e2d81fb2c5d8119b55e8e812da9229320928c16d) ([comment](pytorch#174000 (comment)))

Add a bunch of missing test cases. These were all 'missing' in my single-sim poitnwise PR (discovered missing using the sharding validator). The pointwise rule on main (non-single-dim) is not missing these, but we lack the test coverage. Landing this first helps ensure we don't regress when refactoring. ghstack-source-id: 51dc560 Pull Request resolved: pytorch/pytorch#174000

pytorch-bot Bot added the topic: not user facing topic category label Jan 31, 2026

This was referenced Jan 31, 2026

[LocalTensor] Cache DeviceMesh.get_coordinate results in LocalTensorMode #173836

Closed

[DTensor] Strategy Validation #173976

Closed

[DTensor] Add support for torch.linalg operators #171935

Closed

[DTensor] Complete single-dim pointwise rule #172278

Closed

Skylion007 approved these changes Jan 31, 2026

View reviewed changes

Skylion007 reviewed Jan 31, 2026

View reviewed changes

wconstab mentioned this pull request Feb 2, 2026

WIP claude rewrite pointwise #174080

Closed

pytorch-bot Bot added the ciflow/trunk Trigger trunk jobs on your pull request label Feb 2, 2026

pytorchmergebot added the merging label Feb 2, 2026

pytorchmergebot removed the merging label Feb 2, 2026

pytorchmergebot added the merging label Feb 3, 2026

pytorchmergebot added the Merged label Feb 3, 2026

pytorchmergebot closed this in e2d81fb Feb 3, 2026

pytorchmergebot removed the merging label Feb 3, 2026

pytorchmergebot added Reverted ci-no-td Do not run TD on this PR labels Feb 3, 2026

pytorchmergebot reopened this Feb 3, 2026

wconstab closed this Feb 6, 2026

github-actions Bot deleted the gh/wconstab/519/head branch March 9, 2026 02:22

Conversation

wconstab commented Jan 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot Bot commented Jan 31, 2026

This PR needs a release notes: label

Uh oh!

pytorch-bot Bot commented Jan 31, 2026 • 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/174000

❌ 4 New Failures, 1 Unrelated Failure

Uh oh!

Skylion007 Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

Skylion007 Jan 31, 2026

Choose a reason for hiding this comment

Uh oh!

wconstab commented Feb 2, 2026

Uh oh!

pytorchmergebot commented Feb 2, 2026

Merge started

Uh oh!

pytorchmergebot commented Feb 2, 2026

Merge failed

Uh oh!

wconstab commented Feb 3, 2026

Uh oh!

pytorchmergebot commented Feb 3, 2026

Merge started

Uh oh!

jeffdaily commented Feb 3, 2026

Uh oh!

jeffdaily commented Feb 3, 2026

Uh oh!

pytorchmergebot commented Feb 3, 2026

Uh oh!

pytorchmergebot commented Feb 3, 2026

Uh oh!

izaitsevfb commented Feb 3, 2026

Uh oh!

malfet commented Feb 3, 2026

Uh oh!

izaitsevfb commented Feb 3, 2026

Uh oh!

wconstab commented Feb 6, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

wconstab commented Jan 31, 2026 •

edited

Loading

This PR needs a `release notes:` label

pytorch-bot Bot commented Jan 31, 2026 •

edited

Loading