[dynamo] Prevent recompile on hoistable opaque objects by anijain2305 · Pull Request #176643 · pytorch/pytorch

anijain2305 · 2026-03-05T20:44:08Z

Stack from ghstack (oldest at bottom):

Capture sources from constructor args in OpaqueObjectClassVariable and
thread ctor_arg_sources through TorchScriptObjectVariable and
synthetic_graph_input so subgraph reuse can apply source replacement
to resolve new constructor arg values on stamp-out.

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @kadeng @chauhang @amjames @Lucaskabela @jataylo

…raph reuse Capture sources from constructor args in OpaqueObjectClassVariable and thread ctor_arg_sources through TorchScriptObjectVariable and synthetic_graph_input so subgraph reuse can apply source replacement to resolve new constructor arg values on stamp-out. [ghstack-poisoned]

pytorch-bot · 2026-03-05T20:44:12Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/176643

📄 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! (6 Unrelated Failures)

As of commit 0ed8bab with merge base 0857433 ():

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

inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 1, 2, linux.g5.4xlarge.nvidia.gpu) (gh) (similar failure)
test/inductor/test_triton_kernels.py::TestUserKernelEpilogueFusion::test_fusion_custom_kernel_with_linebreaks
inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 2, 2, linux.g5.4xlarge.nvidia.gpu) (gh) (similar failure)
test/inductor/test_triton_kernels.py::TestUserKernelEpilogueFusion::test_fusion_custom_kernel_with_linebreaks

BROKEN TRUNK - The following jobs failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

inductor / unit-test / inductor-test / test (inductor, 1, 2, linux.g5.4xlarge.nvidia.gpu) (gh) (trunk failure)
test/inductor/test_torchinductor_opinfo.py::TestInductorOpInfoCUDA::test_comprehensive_lu_cuda_float32
pull / cuda13.0-py3.10-gcc11-sm75 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu) (gh) (trunk failure)
MISSING REGRESSION TEST
trunk / linux-jammy-cuda13.0-py3.10-gcc11 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu) (gh) (trunk failure)
MISSING REGRESSION TEST

UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:

inductor / inductor-cpu-test / test (cpu_inductor_torchbench, 1, 2, linux.2xlarge.amx, unstable) (gh) (#174929)
detectron2_maskrcnn_r_50_fpn

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

pytorch-bot · 2026-03-05T20:44:15Z

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.

anijain2305 · 2026-03-05T20:46:46Z

torch/_dynamo/output_graph.py

            )

-            gm = inline_invoke_subgraph(gm)
+            with dynamo_timed("inline_invoke_subgraph"):


Unrelated to this PR - but I hope its ok. In case, anyone has objections, I can open a new PR.

…le for subgraph reuse" Capture sources from constructor args in OpaqueObjectClassVariable and thread ctor_arg_sources through TorchScriptObjectVariable and synthetic_graph_input so subgraph reuse can apply source replacement to resolve new constructor arg values on stamp-out. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx kadeng chauhang amjames Lucaskabela jataylo [ghstack-poisoned]

Capture sources from constructor args in OpaqueObjectClassVariable and thread ctor_arg_sources through TorchScriptObjectVariable and synthetic_graph_input so subgraph reuse can apply source replacement to resolve new constructor arg values on stamp-out. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx kadeng chauhang amjames Lucaskabela jataylo [ghstack-poisoned]

anijain2305 · 2026-03-06T19:17:14Z

@angelayi Thanks for the review. I changed the implementation to further support no recompiles on opaque objects.

zou3519 · 2026-03-06T19:27:23Z

torch/_dynamo/variables/script_object.py

+        if should_hoist(self.value):
+            with tx.output.tracing_context.guards_context.skip_guard_install():
+                constant_args = var_args.as_python_constant()
+                constant_kwargs = var_kwargs.as_python_constant()


do we need to distinguish between if self.value has a source or not here?

I assume we don't because the tests are all passing and I do have a test for this somewhere

pytorch/test/test_opaque_obj_v2.py

Line 2095 in d7a5197

def test_hoist_cache_hits(self):

anijain2305 · 2026-03-06T22:55:33Z

@pytorchbot merge

pytorchmergebot · 2026-03-06T22:57:52Z

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 · 2026-03-06T22:58:12Z

Merge failed

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

pull / linux-docs / build-docs-python-false

Dig deeper by viewing the failures on hud

Details for Dev Infra team

Raised by workflow job

Failing merge rule: Core Maintainers

Capture sources from constructor args in OpaqueObjectClassVariable and thread ctor_arg_sources through TorchScriptObjectVariable and synthetic_graph_input so subgraph reuse can apply source replacement to resolve new constructor arg values on stamp-out. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx kadeng chauhang amjames Lucaskabela jataylo [ghstack-poisoned]

anijain2305 · 2026-03-06T23:56:00Z

@pytorchbot merge

pytorchmergebot · 2026-03-06T23:58:10Z

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 · 2026-03-07T01:43:24Z

Merge failed

Reason: 1 jobs have failed, first few of them are: inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 1, 2, linux.g5.4xlarge.nvidia.gpu)

Details for Dev Infra team

Raised by workflow job

anijain2305 · 2026-03-07T05:37:44Z

@pytorchbot merge -i

pytorchmergebot · 2026-03-07T05:39:58Z

Merge started

Your change will be merged while ignoring the following 6 checks: pull / cuda13.0-py3.10-gcc11-sm75 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu), trunk / linux-jammy-cuda13.0-py3.10-gcc11 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu), inductor / inductor-cpu-test / test (cpu_inductor_torchbench, 1, 2, linux.2xlarge.amx, unstable), inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 2, 2, linux.g5.4xlarge.nvidia.gpu), inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 1, 2, linux.g5.4xlarge.nvidia.gpu), inductor / unit-test / inductor-test / test (inductor, 1, 2, linux.g5.4xlarge.nvidia.gpu)

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 · 2026-03-07T11:38:08Z

The merge job was canceled or timed out. This most often happen if two merge requests were issued for the same PR, or if merge job was waiting for more than 6 hours for tests to finish. In later case, please do not hesitate to reissue the merge command
For more information see pytorch-bot wiki.

anijain2305 · 2026-03-07T19:55:50Z

@pytorchbot merge -i

pytorchmergebot · 2026-03-07T19:57:57Z

Merge started

Your change will be merged while ignoring the following 6 checks: pull / cuda13.0-py3.10-gcc11-sm75 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu), trunk / linux-jammy-cuda13.0-py3.10-gcc11 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu), inductor / inductor-cpu-test / test (cpu_inductor_torchbench, 1, 2, linux.2xlarge.amx, unstable), inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 2, 2, linux.g5.4xlarge.nvidia.gpu), inductor / unit-test / inductor-test / test (inductor_cpp_wrapper, 1, 2, linux.g5.4xlarge.nvidia.gpu), inductor / unit-test / inductor-test / test (inductor, 1, 2, linux.g5.4xlarge.nvidia.gpu)

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

…raph reuse Capture sources from constructor args in OpaqueObjectClassVariable and thread ctor_arg_sources through TorchScriptObjectVariable and synthetic_graph_input so subgraph reuse can apply source replacement to resolve new constructor arg values on stamp-out. ghstack-source-id: bb03798 Pull Request resolved: pytorch/pytorch#176643

pytorch-bot bot added ciflow/inductor module: dynamo labels Mar 5, 2026

This was referenced Mar 5, 2026

[dynamo] Track mutated_sources on SideEffects for precise mutation detection #176477

Closed

[dynamo] Add GUARD_VALUE_DISPATCH table for guard evaluation #176033

Closed

[dynamo] Add subgraph reuse for invoke_subgraph #176644

Open

anijain2305 commented Mar 5, 2026

View reviewed changes

anijain2305 requested review from angelayi and zou3519 and removed request for zou3519 March 5, 2026 20:47

angelayi approved these changes Mar 5, 2026

View reviewed changes

anijain2305 added ciflow/trunk Trigger trunk jobs on your pull request topic: not user facing topic category labels Mar 6, 2026

anijain2305 changed the title ~~[dynamo] Track ctor_arg_sources on TorchScriptObjectVariable for subgraph reuse~~ [dynamo] Prevent recompile on hoistable opaque objects Mar 6, 2026

anijain2305 requested a review from angelayi March 6, 2026 19:16

zou3519 reviewed Mar 6, 2026

View reviewed changes

zou3519 approved these changes Mar 6, 2026

View reviewed changes

anijain2305 mentioned this pull request Mar 6, 2026

[dynamo][hop] Remove unnecessary realize_all from speculate subgraph #176742

Closed

pytorchmergebot added the merging label Mar 6, 2026

pytorchmergebot removed the merging label Mar 6, 2026

pytorchmergebot added the merging label Mar 6, 2026

pytorchmergebot removed the merging label Mar 7, 2026

pytorchmergebot added the merging label Mar 7, 2026

pytorchmergebot closed this in 6d25e09 Mar 7, 2026

pytorchmergebot added Merged and removed merging labels Mar 7, 2026

anijain2305 mentioned this pull request Mar 10, 2026

[dynamo] Prevent VT realization for the first arg #177078

Closed

Conversation

anijain2305 commented Mar 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Mar 5, 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/176643

✅ You can merge normally! (6 Unrelated Failures)

Uh oh!

pytorch-bot bot commented Mar 5, 2026

This PR needs a release notes: label

Uh oh!

anijain2305 Mar 5, 2026

Choose a reason for hiding this comment

Uh oh!

anijain2305 commented Mar 6, 2026

Uh oh!

zou3519 Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

zou3519 Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

zou3519 Mar 6, 2026

Choose a reason for hiding this comment

Uh oh!

anijain2305 commented Mar 6, 2026

Uh oh!

pytorchmergebot commented Mar 6, 2026

Merge started

Uh oh!

pytorchmergebot commented Mar 6, 2026

Merge failed

Uh oh!

anijain2305 commented Mar 6, 2026

Uh oh!

pytorchmergebot commented Mar 6, 2026

Merge started

Uh oh!

pytorchmergebot commented Mar 7, 2026

Merge failed

Uh oh!

anijain2305 commented Mar 7, 2026

Uh oh!

pytorchmergebot commented Mar 7, 2026

Merge started

Uh oh!

pytorchmergebot commented Mar 7, 2026

Uh oh!

anijain2305 commented Mar 7, 2026

Uh oh!

pytorchmergebot commented Mar 7, 2026

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

anijain2305 commented Mar 5, 2026 •

edited

Loading

pytorch-bot bot commented Mar 5, 2026 •

edited

Loading

This PR needs a `release notes:` label