Add Python 3.13 support

[PetarTerziev-UL]

Summary

coremltools versions prior to 9.0 did not support Python 3.13. Now executorch uses 9.0, we can finally enable building wheels with Python 3.13.

Fixes #8292

Test plan

Python test suite ran and compared against a dev install of the target branchh in a 3.12 virtual environment.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16004

• 📄 Preview Python docs built from this PR

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

❌ 14 New Failures, 5 Cancelled Jobs, 2 Unrelated Failures

As of commit c1adb09 with merge base 488d761 ():

NEW FAILURES - The following jobs have failed:

• check-c10-sync / check-c10-sync (gh)
  Process completed with exit code 1.
• pull / test-moshi-linux / linux-job (gh)
  RuntimeError: Could not load libtorchcodec. Likely causes:
• pull / unittest-editable / macos / macos-job (gh)
  export/tests/test_target_recipes.py::TestTargetRecipes::test_mv3_model
• Test CUDA Builds / check-all-cuda-builds (gh)
  Process completed with exit code 1.
• Test CUDA Builds / test-executorch-cuda-build-12.6 / linux-job (gh)
  RuntimeError: Command docker exec -t 1cffc8eb4fd8c629aefd142a8abcc88392d30ffea1c9031164e1b43a34df9306 /exec failed with exit code 1
• Test CUDA Builds / test-executorch-cuda-build-12.8 / linux-job (gh)
  RuntimeError: Command docker exec -t d0e8d5e11af994c8e81ed0b9840bd426f503d1a895c8883922aced611e38cec0 /exec failed with exit code 1
• Test CUDA Builds / test-models-cuda (add_mul) / linux-job (gh)
  RuntimeError: Command docker exec -t f0ec936bca0b86f427ac4199155fbc64b1cd5a6807991b640d229c9cc09a5f26 /exec failed with exit code 1
• Test CUDA Builds / test-models-cuda (add) / linux-job (gh)
  RuntimeError: Command docker exec -t c2a5c4589f17024995721addfb7f6ab26cec5ce0c9db1a97f58318a9e3fda750 /exec failed with exit code 1
• Test CUDA Builds / test-models-cuda (conv1d) / linux-job (gh)
  RuntimeError: Command docker exec -t 056b722ff6d2d347eaaf1824a103f4c0e5f537749f94a489fafd219a368f33ab /exec failed with exit code 1
• Test CUDA Builds / test-models-cuda (linear) / linux-job (gh)
  RuntimeError: Command docker exec -t 1a675263bedfbaf622798a91a68fd0462dbe7126b2a6a4c7183fafd0b6971ace /exec failed with exit code 1
• Test CUDA Builds / test-models-cuda (resnet18) / linux-job (gh)
  RuntimeError: Command docker exec -t 88e7a5313226dcad561ee482060c448aebe58f74e5ad1d82a4f86054511e6cd9 /exec failed with exit code 1
• Test CUDA Builds / test-models-cuda (sdpa) / linux-job (gh)
  RuntimeError: Command docker exec -t e8eed263da802dad189a77c8c67991b22e91ead1543d62e79b6fef3bdc999edf /exec failed with exit code 1
• trunk / test-arm-backend (test_pytest_ops_ethosu_fvp) / linux-job (gh)
  RuntimeError: Command docker exec -t 8488a2d220619245385517fe78528c65d639dde44d378aacfbb53d912ef805e3 /exec failed with exit code 1
• trunk / test-arm-cortex-m-size-test (zephyr-preset) / linux-job (gh)
  RuntimeError: Command docker exec -t 6249ba3d5acd8e161dc96f476cd67202ee9c4f8b94a39144339f1083af4c9893 /exec failed with exit code 1

CANCELLED JOBS - The following jobs were cancelled. Please retry:

• Lint / link-check / lint-file-size / linux-job (gh)
  ##[error]The operation was canceled.
• pull / test-binary-size-linux-gcc / linux-job (gh)
  ##[error]The operation was canceled.
• pull / test-openvino-linux / linux-job (gh)
• pull / test-setup-linux-gcc / linux-job (gh)
  ##[error]The operation was canceled.
• pull / unittest-arm-backend-with-no-fvp (test_pytest_ops) / linux-job (gh)
  ##[error]The operation was canceled.

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

• docker-builds / docker-build (linux.4xlarge, executorch-ubuntu-22.04-gcc9) (gh) (detected as infra flaky with no log or failing log classifier)

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

• pull / android / run-emulator (gh) (#16137)
  Timeout waiting for emulator to boot.

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

[PetarTerziev-UL]

I need to add a "release notes: misc" label, but I don't seem to have the ability to.

[PetarTerziev-UL]

the tests under examples/models/llama3_2_vision/preprocess hang on a call into AOTInductor when torch nightly is installed.

Need to look into this further.

[PetarTerziev-UL]

the tests under examples/models/llama3_2_vision/preprocess hang on a call into AOTInductor when torch nightly is installed.

Need to look into this further.

Things are better in with a newer nightly build of torch, so I've bumped the pinned version as part of this PR too.

[Copilot]

Pull request overview

This PR adds Python 3.13 support to ExecuTorch now that coremltools 9.0 (which supports Python 3.13) is available. The changes update version constraints, documentation, and CI/CD workflows to enable building wheels for Python 3.13.

Key changes:

• Update Python version constraint from <3.13 to <3.14 in pyproject.toml
• Add Python 3.13 to wheel building workflows and documentation
• Remove Python 3.13 warning from CoreML troubleshooting documentation
• Update nightly version pin (unrelated maintenance update)

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
pyproject.toml	Add Python 3.13 classifier, update requires-python constraint to <3.14, add py313 to ruff target-version
torch_pin.py	Update nightly version from dev20251025 to dev20251124
docs/source/*.md	Update Python version requirements from 3.10-3.12 to 3.10-3.13 across multiple documentation files
docs/source/backends/coreml/coreml-troubleshooting.md	Remove outdated Python 3.13 compatibility warning
README-wheel.md	Add Python 3.13 to list of supported versions
.github/workflows/*.yml	Add "3.13" to Python version matrices for wheel building workflows

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The comment is now outdated since Python 3.13 is being supported. Either remove the comment entirely or update it to reflect that the issue was resolved in coremltools 9.0 (e.g., '# coremltools 9.0+ supports python 3.13, see apple/coremltools#2487').

Suggested change

	# coremltools has issue with python 3.13, see https://github.com/apple/coremltools/issues/2487
	# coremltools 9.0+ supports python 3.13, see https://github.com/apple/coremltools/issues/2487

[larryliu0820]

Delete this comment?

[larryliu0820]

@PetarTerziev-UL can you please bump the nightly version of pytorch to 1122 instead? I think the nightly job failed to build 1124 nightly.

[PetarTerziev-UL]

@PetarTerziev-UL can you please bump the nightly version of pytorch to 1122 instead? I think the nightly job failed to build 1124 nightly.

Done! Is there a way to check nightly job statuses? I was looking at wheels available at download.pytorch.org and drawing assumptions.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The target-version includes py38 and py39, but requires-python specifies >=3.10. This mismatch means Ruff is configured to support Python versions below the project's minimum requirement. Either remove py38 and py39 from target-version to align with requires-python, or update the comment on line 132-133 to explain why these older versions are still targeted.

Suggested change

	target-version = ["py38", "py39", "py310", "py311", "py312", "py313"]
	target-version = ["py310", "py311", "py312", "py313"]

[PetarTerziev-UL]

@larryliu0820 I've rolled back to an earlier nightly version still.

Looks like the failing CI job ends up with a versions of torch and torchaudio which are incompatible with each other down to an API change in C10 Logging.

However, after looking at the test script itself, it seems like moshi requires torch <2.8,>=2.2.0 which is in conflict with building executorch agaist a nightly build of torch. I'm looking further into this...

EDIT:

Actually, I just found out that the Linux docker container also has a git hash that pins a torch version. Will can/will update that too to match the torch_pin.py version.

It'd also be good to try dropping the nightly bump commit and running the CI, to see whether the hang that prompted me to update that was unique to my local system. Then the nightly bump can go into a separate PR as it is turning out to be a bit of a can of worms 😁

EDIT 2: I've changed the version of torch the Docker container to match the nightly I bumped to originally.

[Copilot]

Pull request overview

Copilot reviewed 14 out of 14 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[PetarTerziev-UL]

@larryliu0820 (and anyone else interested)

I've dropped the nightly update commit. It triggers a new docker container build, which then installs the latest pip (25.3) which removes the --no-use-pep517 option for install. Please see relevant changelogs. I fear this means that next time the container needs updating, it will require a full migration to a pyproject.toml based build. Though I guess everywhere a new conda env is created will suffer the same problem.

I'll raise an issue in the repo and update this comment with it.

In the mean time, I hope that the problem which prompted me to bump the nightly version was unique to me, and this can get merged.

EDIT: I see that torch, audio, and vision (probably others too) use setup-py instead of python-build-pacakge based builds. And as the projects are quite intertwined, migrating is quite an undertaking. I guess as a workaround we could force conda to install pip 25.2.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated no new comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.