C++-accessible Placements via pybind11 by swolchok · Pull Request #163030 · pytorch/pytorch

swolchok · 2025-09-16T01:10:58Z

Stack from ghstack (oldest at bottom):

This makes Placement data representation available in C++ via pybind11.

cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @pragupta @msaroufim @dcci @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @kadeng @chauhang @amjames @Lucaskabela @jataylo @ezyang @chenyang78

This makes Placement data representation available in C++ via pybind11. [ghstack-poisoned]

pytorch-bot · 2025-09-16T01:11:02Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/163030

📄 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.

✅ No Failures

As of commit 54c6509 with merge base e5c0e6b ():
💚 Looks good so far! There are no failures yet. 💚

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

swolchok · 2025-09-16T01:12:58Z

I was overthinking how to do this. The overhead for grabbing the object from the pybind11 object is higher than I'd like but I think using this still results in a performance win despite that (see #163031).

Skylion007 · 2025-09-16T16:48:45Z

 } // namespace torch::autograd

+// XXX: Move this somewhere
+static const char placement_class_docstring[] =


Should also be constexpr

Skylion007 · 2025-09-16T16:49:17Z

+
+class Replicate : public Placement {
+ public:
+  ~Replicate() {}


Should be equals default, right?

Skylion007 · 2025-09-16T16:49:34Z

+  explicit Partial(std::string reduce_op_) : reduce_op(std::move(reduce_op_)) {}
+
+  ~Partial() {}
+


Likewise here

Skylion007 · 2025-09-16T16:50:37Z

+  }
+};
+
+Placement::~Placement() {}


Shouldn't this be equals default too? Why is it out of line?

Why is it out of line?

It's pure virtual, but still needs an implementation. Can't do both inline.

This makes Placement data representation available in C++ via pybind11. [ghstack-poisoned]

swolchok · 2025-09-17T18:47:45Z

I'm not sure why BC Lint is so unhappy with what I've done here, especially now that I've updated torch/_C/__init__.pyi.in.

This makes Placement data representation available in C++ via pybind11. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

This makes Placement data representation available in C++ via pybind11. ghstack-source-id: 9f85bee Pull Request resolved: #163030

…pybind11" This makes Placement data representation available in C++ via pybind11. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

This makes Placement data representation available in C++ via pybind11. ghstack-source-id: d2edadb Pull Request resolved: #163030

…"C++-accessible Placements via pybind11" This makes Placement data representation available in C++ via pybind11. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

… uncertainty on "DTensor: C++ compute_global_tensor_info" compute_global_tensor_info is on the hot path for DTensor.{from,to}_local. More incremental progress toward C++. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

… uncertainty on "Use C++-accessible Placement in compute_global_tensor_info" This seems to speed things up, though there is more overhead than I'd like to unwrap a Placement via pybind11. Might be time to introduce nanobind; I'm not sure that we can use it in general, but the Placement type system is (currently?) isolated. (Alternatively, I could look into whether we can cut this overhead.) [ghstack-poisoned]

… uncertainty on "C++-accessible Placements via pybind11" This makes Placement data representation available in C++ via pybind11. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

pytorchmergebot · 2025-10-02T18:25:28Z

@swolchok your PR has been successfully reverted.

This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. Differential Revision: [D83788896](https://our.internmc.facebook.com/intern/diff/D83788896/) [ghstack-poisoned]

This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. Differential Revision: [D83788896](https://our.internmc.facebook.com/intern/diff/D83788896/) ghstack-source-id: 313745131 Pull Request resolved: #164519

Pull Request resolved: #164519 This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. ghstack-source-id: 313909406 @exported-using-ghexport Differential Revision: [D83788896](https://our.internmc.facebook.com/intern/diff/D83788896/)

…ind11 (#163030)"" This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. Differential Revision: [D83788896](https://our.internmc.facebook.com/intern/diff/D83788896/) cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames Lucaskabela [ghstack-poisoned]

This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. Differential Revision: [D83788896](https://our.internmc.facebook.com/intern/diff/D83788896/) cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames Lucaskabela [ghstack-poisoned]

This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. D83788896 Pull Request resolved: #164519 Approved by: https://github.com/Skylion007, https://github.com/ezyang

…Partial to placement_types to improve discoverability" Had trouble finding this one myself in #163030. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]

…ment_types to improve discoverability" Had trouble finding this one myself in #163030. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]

…rove discoverability" Had trouble finding this one myself in #163030. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]

…lity" Had trouble finding this one myself in #163030. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]

…164519)" This reverts commit 8c0bc87. Reverted #164519 on behalf of https://github.com/malfet due to Still breaks internal workflows ([comment](#164519 (comment)))

@fadeputr

Was reverted due to internal signal. After internal discussion with @fadeputr, reapplying with no further changes seems to be the right call. Differential Revision: [D84515401](https://our.internmc.facebook.com/intern/diff/D84515401/) [ghstack-poisoned]

Was reverted due to internal signal. After internal discussion with fadeputr, reapplying with no further changes seems to be the right call. Differential Revision: [D84515401](https://our.internmc.facebook.com/intern/diff/D84515401/) ghstack-source-id: 315835168 Pull Request resolved: #165335

…pybind11 (#163030)"" Was reverted due to internal signal. After internal discussion with fadeputr, reapplying with no further changes seems to be the right call. Differential Revision: [D84515401](https://our.internmc.facebook.com/intern/diff/D84515401/) cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames Lucaskabela [ghstack-poisoned]

…"" Was reverted due to internal signal. After internal discussion with fadeputr, reapplying with no further changes seems to be the right call. Differential Revision: [D84515401](https://our.internmc.facebook.com/intern/diff/D84515401/) cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames Lucaskabela [ghstack-poisoned]

@fadeputr

Pull Request resolved: #165335 Was reverted due to internal signal. After internal discussion with @fadeputr, reapplying with no further changes seems to be the right call. ghstack-source-id: 315881161 @exported-using-ghexport Differential Revision: [D84515401](https://our.internmc.facebook.com/intern/diff/D84515401/)

Was reverted due to a merge conflict that crept in sometime during the "export to github -> land internally -> merge on github" process. Differential Revision: [D84862100](https://our.internmc.facebook.com/intern/diff/D84862100/) [ghstack-poisoned]

Was reverted due to a merge conflict that crept in sometime during the "export to github -> land internally -> merge on github" process. Differential Revision: [D84862100](https://our.internmc.facebook.com/intern/diff/D84862100/) ghstack-source-id: 316992680 Pull Request resolved: #165764

This makes Placement data representation available in C++ via pybind11. Pull Request resolved: pytorch#163030 Approved by: https://github.com/ezyang

This reverts commit 3e03dea. Reverted pytorch#163030 on behalf of https://github.com/swolchok due to doesn't pass pyre ([comment](pytorch#163030 (comment)))

…torch#164519) This makes Placement data representation available in C++ via pybind11. Reapply with fix for internal errors. D83788896 Pull Request resolved: pytorch#164519 Approved by: https://github.com/Skylion007, https://github.com/ezyang

…)" (pytorch#164519)" This reverts commit 8c0bc87. Reverted pytorch#164519 on behalf of https://github.com/malfet due to Still breaks internal workflows ([comment](pytorch#164519 (comment)))

Was reverted (again!) due to a merge conflict that crept in sometime during the "export to github -> land internally -> merge on github" process. Differential Revision: [D85096233](https://our.internmc.facebook.com/intern/diff/D85096233/) [ghstack-poisoned]

Was reverted (again!) due to a merge conflict that crept in sometime during the "export to github -> land internally -> merge on github" process. Differential Revision: [D85096233](https://our.internmc.facebook.com/intern/diff/D85096233/) ghstack-source-id: 318252481 Pull Request resolved: #166132

Pull Request resolved: #166132 Was reverted (again!) due to a merge conflict that crept in sometime during the "export to github -> land internally -> merge on github" process. ghstack-source-id: 318928038 @exported-using-ghexport Differential Revision: [D85096233](https://our.internmc.facebook.com/intern/diff/D85096233/)

github-actions · 2026-01-17T02:15:53Z

Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as Stale.
Feel free to remove the Stale label if you feel this was a mistake.
If you are unable to remove the Stale label please contact a maintainer in order to do so.
If you want the bot to never mark this PR stale again, add the no-stale label.
Stale pull requests will automatically be closed after 30 days of inactivity.

C++-accessible Placements via pybind11

b40bee9

This makes Placement data representation available in C++ via pybind11. [ghstack-poisoned]

swolchok requested review from albanD and soulitzer as code owners September 16, 2025 01:10

pytorch-bot Bot added ciflow/inductor oncall: distributed Add this issue/PR to distributed oncall triage queue labels Sep 16, 2025

swolchok requested a review from ezyang September 16, 2025 01:11

swolchok marked this pull request as draft September 16, 2025 04:15

albanD reviewed Sep 16, 2025

View reviewed changes

Comment thread torch/csrc/autograd/python_variable.cpp Outdated

Skylion007 reviewed Sep 16, 2025

View reviewed changes

swolchok added 2 commits September 16, 2025 23:55

Update on "C++-accessible Placements via pybind11"

82774b1

This makes Placement data representation available in C++ via pybind11. [ghstack-poisoned]

Update on "C++-accessible Placements via pybind11"

04da78e

This makes Placement data representation available in C++ via pybind11. [ghstack-poisoned]

swolchok added the release notes: distributed (dtensor) release notes category label Sep 18, 2025

Update on "C++-accessible Placements via pybind11"

7359715

This makes Placement data representation available in C++ via pybind11. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

swolchok added a commit that referenced this pull request Sep 18, 2025

C++-accessible Placements via pybind11

a89c64b

This makes Placement data representation available in C++ via pybind11. ghstack-source-id: 9f85bee Pull Request resolved: #163030

swolchok added the suppress-bc-linter Suppresses the failures of API backward-compatibility linter (Lint/bc_linter) label Sep 18, 2025

need to pick up fix in previous pr on "C++-accessible Placements via …

4e5f111

…pybind11" This makes Placement data representation available in C++ via pybind11. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta ezyang msaroufim dcci [ghstack-poisoned]

swolchok added a commit that referenced this pull request Sep 18, 2025

C++-accessible Placements via pybind11

e7698fd

This makes Placement data representation available in C++ via pybind11. ghstack-source-id: d2edadb Pull Request resolved: #163030

pytorchmergebot added Reverted ci-no-td Do not run TD on this PR labels Oct 2, 2025

pytorchmergebot reopened this Oct 2, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

C++-accessible Placements via pybind11#163030

C++-accessible Placements via pybind11#163030
swolchok wants to merge 20 commits intogh/swolchok/839/basefrom
gh/swolchok/839/head

swolchok commented Sep 16, 2025 •

edited by pytorch-bot Bot

Loading

Uh oh!

pytorch-bot Bot commented Sep 16, 2025 •

edited

Loading

Uh oh!

swolchok commented Sep 16, 2025

Uh oh!

Uh oh!

Skylion007 Sep 16, 2025

Uh oh!

Skylion007 Sep 16, 2025

Uh oh!

Skylion007 Sep 16, 2025

Uh oh!

Skylion007 Sep 16, 2025

Uh oh!

swolchok Sep 16, 2025

Uh oh!

swolchok commented Sep 17, 2025 •

edited

Loading

Uh oh!

pytorchmergebot commented Oct 2, 2025

Uh oh!

github-actions Bot commented Jan 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

		explicit Partial(std::string reduce_op_) : reduce_op(std::move(reduce_op_)) {}

		~Partial() {}

Conversation

swolchok commented Sep 16, 2025 • edited by pytorch-bot Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot Bot commented Sep 16, 2025 • 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/163030

✅ No Failures

Uh oh!

swolchok commented Sep 16, 2025

Uh oh!

Uh oh!

Skylion007 Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

Skylion007 Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

Skylion007 Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

Skylion007 Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

swolchok Sep 16, 2025

Choose a reason for hiding this comment

Uh oh!

swolchok commented Sep 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorchmergebot commented Oct 2, 2025

Uh oh!

github-actions Bot commented Jan 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

swolchok commented Sep 16, 2025 •

edited by pytorch-bot Bot

Loading

pytorch-bot Bot commented Sep 16, 2025 •

edited

Loading

swolchok commented Sep 17, 2025 •

edited

Loading