Add sum and sum_out to stable ops by mikaylagawarecki · Pull Request #168062 · pytorch/pytorch

mikaylagawarecki · 2025-11-18T03:26:30Z

Stack from ghstack (oldest at bottom):

[ghstack-poisoned]

pytorch-bot · 2025-11-18T03:26:34Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/168062

📄 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.

✅ You can merge normally! (1 Unrelated Failure)

As of commit e50299e with merge base 5d1459a ():

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

trunk / linux-jammy-cuda12.8-py3.10-gcc11 / test (default, 1, 5, lf.linux.g6.4xlarge.experimental.nvidia.gpu) (gh) (disabled by #167274)
test/inductor/test_cudagraph_trees.py::TestSAC::test_graph_partition_cudagraphs_aot_eager_compat_equal

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

ghstack-source-id: e5fa5bb Pull Request resolved: #168062

[ghstack-poisoned]

ghstack-source-id: dab72ec Pull Request resolved: #168062

[ghstack-poisoned]

ghstack-source-id: f90d3b1 Pull Request resolved: #168062

janeyx99 · 2025-11-19T19:11:22Z

test/cpp_extensions/test_libtorch_agnostic.py

+            self.assertEqual(result_dtype, expected_dtype)
+
+            # Test sum without dim (sum all elements) - pass empty list
+            result_all = libtorch_agnostic.ops.my_sum(t, [])


Suggested change

result_all = libtorch_agnostic.ops.my_sum(t, [])

result_all = libtorch_agnostic.ops.my_sum(t)

would be best to not have to pass in an empty list at all

I'm noticing you probably don't include this because TORCH_BOX doesn't support std::optional headeronlyarrayref -> I think it's ok to fall back to writing a boxing function for this case in order to show that calling just sum(t) would work.

Fixed by making TORCH_BOX support std::optional headeronlyarrayref> instead

test/cpp_extensions/test_libtorch_agnostic.py

[ghstack-poisoned]

ghstack-source-id: ace8cfd Pull Request resolved: #168062

albanD

SGTM

mikaylagawarecki · 2025-12-08T23:45:37Z

@pytorchbot merge

pytorchmergebot · 2025-12-08T23:48:03Z

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

…ch_call_dispatcher) (#169872) Scalar's StableIValue conversion is not supported yet. If we don't land this we can recommend `fill_(empty(...))` instead Pull Request resolved: #169872 Approved by: https://github.com/albanD ghstack dependencies: #169703, #169709, #169711, #168062

Pull Request resolved: pytorch#168062 Approved by: https://github.com/albanD ghstack dependencies: pytorch#169703, pytorch#169709, pytorch#169711

…ch_call_dispatcher) (pytorch#169872) Scalar's StableIValue conversion is not supported yet. If we don't land this we can recommend `fill_(empty(...))` instead Pull Request resolved: pytorch#169872 Approved by: https://github.com/albanD ghstack dependencies: pytorch#169703, pytorch#169709, pytorch#169711, pytorch#168062

From https://github.com/pytorch/audio/blob/main/src/libtorchaudio/stable/ops.h Technically it should have been ok not to port these but looking at these carefully I realized the subtract ported to audio ~would have undefined behavior :/~ is broken ``` inline Tensor subtract(const Tensor& self, const Tensor& other) { const auto num_args = 2; std::array<StableIValue, num_args> stack{ torch::stable::detail::from(self), torch::stable::detail::from(other)}; TORCH_ERROR_CODE_CHECK(torch_call_dispatcher( "aten::subtract", "Tensor", stack.data(), TORCH_ABI_VERSION)); return torch::stable::detail::to<torch::stable::Tensor>(stack[0]); } ``` as it missed `alpha` the signature for `subtract.Tensor` is `func: subtract.Tensor(Tensor self, Tensor other, *, Scalar alpha=1) -> Tensor`. ~This is also our bad as although out of bounds reads on the stableivalue stack would be caught by asan, without asan they are silent correctness issues (PR coming to fix).~ Use the old path to support this as we don't support stableivalue conversion for Scalar yet. Pull Request resolved: #169880 Approved by: https://github.com/albanD ghstack dependencies: #169703, #169709, #169711, #168062, #169872

Pull Request resolved: pytorch#168062 Approved by: https://github.com/albanD ghstack dependencies: pytorch#169703, pytorch#169709, pytorch#169711

…ch_call_dispatcher) (pytorch#169872) Scalar's StableIValue conversion is not supported yet. If we don't land this we can recommend `fill_(empty(...))` instead Pull Request resolved: pytorch#169872 Approved by: https://github.com/albanD ghstack dependencies: pytorch#169703, pytorch#169709, pytorch#169711, pytorch#168062

…ch#169880) From https://github.com/pytorch/audio/blob/main/src/libtorchaudio/stable/ops.h Technically it should have been ok not to port these but looking at these carefully I realized the subtract ported to audio ~would have undefined behavior :/~ is broken ``` inline Tensor subtract(const Tensor& self, const Tensor& other) { const auto num_args = 2; std::array<StableIValue, num_args> stack{ torch::stable::detail::from(self), torch::stable::detail::from(other)}; TORCH_ERROR_CODE_CHECK(torch_call_dispatcher( "aten::subtract", "Tensor", stack.data(), TORCH_ABI_VERSION)); return torch::stable::detail::to<torch::stable::Tensor>(stack[0]); } ``` as it missed `alpha` the signature for `subtract.Tensor` is `func: subtract.Tensor(Tensor self, Tensor other, *, Scalar alpha=1) -> Tensor`. ~This is also our bad as although out of bounds reads on the stableivalue stack would be caught by asan, without asan they are silent correctness issues (PR coming to fix).~ Use the old path to support this as we don't support stableivalue conversion for Scalar yet. Pull Request resolved: pytorch#169880 Approved by: https://github.com/albanD ghstack dependencies: pytorch#169703, pytorch#169709, pytorch#169711, pytorch#168062, pytorch#169872

Add sum and sum_out to stable ops

5af6031

[ghstack-poisoned]

This was referenced Nov 18, 2025

Fix TORCH_FEATURE_VERSION guards #167802

Closed

Split libtorch agnostic tests by feature version #167803

Closed

Test libtorch_agnostic with TORCH_TARGET_VERSION on target pytorch version #167804

Closed

This was referenced Nov 17, 2025

Test that TORCH_FEATURE_VERSION guards are used where needed #167962

Closed

Move MemoryFormat/Layout to headeronly #168034

Closed

mikaylagawarecki added a commit that referenced this pull request Nov 18, 2025

Add sum and sum_out to stable ops

dcc1a25

ghstack-source-id: e5fa5bb Pull Request resolved: #168062

Update on "Add sum and sum_out to stable ops"

6459b37

[ghstack-poisoned]

mikaylagawarecki mentioned this pull request Nov 17, 2025

Follow up on #161891 move additions to stable shim and use version guards #168025

Closed

mikaylagawarecki added a commit that referenced this pull request Nov 18, 2025

Add sum and sum_out to stable ops

ce45e84

ghstack-source-id: dab72ec Pull Request resolved: #168062

Update on "Add sum and sum_out to stable ops"

bb946ad

[ghstack-poisoned]

mikaylagawarecki added a commit that referenced this pull request Nov 18, 2025

Add sum and sum_out to stable ops

2a67acf

ghstack-source-id: f90d3b1 Pull Request resolved: #168062

This was referenced Nov 19, 2025

Test that TORCH_FEATURE_VERSION guards are used where needed qodo-benchmark/pytorch#1

Open

Test that TORCH_FEATURE_VERSION guards are used where needed qodo-benchmark/pytorch#2

Open

bar-qodo mentioned this pull request Nov 19, 2025

Test that TORCH_FEATURE_VERSION guards are used where needed qodo-benchmark/pytorch#4

Open

janeyx99 reviewed Nov 19, 2025

View reviewed changes

test/cpp_extensions/test_libtorch_agnostic.py Outdated Show resolved Hide resolved

tomerqodo mentioned this pull request Nov 20, 2025

Test that TORCH_FEATURE_VERSION guards are used where needed qodo-benchmark/pytorch#5

Open

bar-qodo mentioned this pull request Nov 20, 2025

Test that TORCH_FEATURE_VERSION guards are used where needed qodo-benchmark/pytorch#6

Open

Update on "Add sum and sum_out to stable ops"

e50299e

[ghstack-poisoned]

This was referenced Dec 8, 2025

Add element_size, storage_offset and set_requires_grad to stable Tensor #169703

Closed

Add to and contiguous to stable ops #169709

Closed

Upgrade stable new_{empty/zeros} that had only dtype argument to full version #169711

Closed

mikaylagawarecki added a commit that referenced this pull request Dec 8, 2025

Add sum and sum_out to stable ops

c78fdf7

ghstack-source-id: ace8cfd Pull Request resolved: #168062

mikaylagawarecki added ciflow/trunk Trigger trunk jobs on your pull request release notes: cpp release notes category topic: new features topic category labels Dec 8, 2025

mikaylagawarecki requested a review from albanD December 8, 2025 19:40

mikaylagawarecki marked this pull request as ready for review December 8, 2025 19:40

albanD approved these changes Dec 8, 2025

View reviewed changes

This was referenced Dec 8, 2025

Add full to stable ops (supported via generate c shim rather than torch_call_dispatcher) #169872

Closed

Add squeeze, unsqueeze, matmul, select, subtract to stable ops #169880

Closed

pytorchmergebot added the merging label Dec 8, 2025

pytorchmergebot added the Merged label Dec 8, 2025

pytorchmergebot closed this in 78501a0 Dec 8, 2025

pytorchmergebot removed the merging label Dec 8, 2025

liangxs pushed a commit to liangxs/pytorch that referenced this pull request Dec 9, 2025

Add sum and sum_out to stable ops (pytorch#168062)

4016658

Pull Request resolved: pytorch#168062 Approved by: https://github.com/albanD ghstack dependencies: pytorch#169703, pytorch#169709, pytorch#169711

github-actions bot deleted the gh/mikaylagawarecki/387/head branch January 8, 2026 02:21

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add sum and sum_out to stable ops#168062

Add sum and sum_out to stable ops#168062
mikaylagawarecki wants to merge 4 commits intogh/mikaylagawarecki/387/basefrom
gh/mikaylagawarecki/387/head

mikaylagawarecki commented Nov 18, 2025 •

edited

Loading

Uh oh!

pytorch-bot bot commented Nov 18, 2025 •

edited

Loading

Uh oh!

janeyx99 Nov 19, 2025

Uh oh!

janeyx99 Nov 19, 2025

Uh oh!

mikaylagawarecki Dec 8, 2025

Uh oh!

Uh oh!

albanD left a comment

Uh oh!

mikaylagawarecki commented Dec 8, 2025

Uh oh!

pytorchmergebot commented Dec 8, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

	result_all = libtorch_agnostic.ops.my_sum(t, [])
	result_all = libtorch_agnostic.ops.my_sum(t)

Conversation

mikaylagawarecki commented Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Nov 18, 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/168062

✅ You can merge normally! (1 Unrelated Failure)

Uh oh!

janeyx99 Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

janeyx99 Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

mikaylagawarecki Dec 8, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

albanD left a comment

Choose a reason for hiding this comment

Uh oh!

mikaylagawarecki commented Dec 8, 2025

Uh oh!

pytorchmergebot commented Dec 8, 2025

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mikaylagawarecki commented Nov 18, 2025 •

edited

Loading

pytorch-bot bot commented Nov 18, 2025 •

edited

Loading