🔧 chore(ci): upgrade python 3.13 #3112

Rickypanta0 · 2025-11-13T11:47:55Z

📝 Description

This PR addresses issue #2959 by upgrading the project's minimum Python version from Python 3.10 to Python 3.13. It addresses dependency constraints, updates CI configurations, and applies necessary fixes to code and tests to ensure compatibility with the updated environment.

Key Changes & Rationale:

CI/CD Upgrade: Updated GitHub Actions (pre_merge.yml) and documentation to use Python 3.13.
Dependency Resolution:
- Updated pyproject.toml and uv.lock to support Python 3.13.
- Restricted cu118 and cu121 extras to python_version < '3.13', as PyTorch does not provide wheels for these CUDA versions on Python 3.13 (which defaults to CUDA 12.4).
ONNX Export Fix:
- Switched torch.onnx.export to use the legacy JIT exporter (dynamo=False) in export_mixin.py.
- Reason: The new Dynamo exporter enforces stricter graph tracing and currently fails with GuardOnDataDependentSymNode due to data-dependent control flow in post_processor.py (specifically isnan() checks). The legacy exporter handles this correctly.
Test Adjustments:
- Relaxed tolerance in test_inference_similarity (from 1e-3 to 5-e3).
- Reason: Minor floating-point precision differences were observed between Torch and OpenVINO inference with the updated library versions.

🛠️ Fixes 🔼 Upgrade the minimum python requirement to 3.13 #2959

✨ Changes

Select what type of change your PR is:

✅ Checklist

Before you submit your pull request, please make sure you have completed the following steps:

📚 I have made the necessary updates to the documentation (if applicable).
🧪 I have written tests that support my changes and prove that my fix is effective or my feature works (if applicable).
🏷️ My PR title follows conventional commit format.

Copilot

Pull request overview

This PR upgrades the project's minimum Python version from 3.10 to 3.13, ensuring compatibility with the latest Python release. The changes include CI/CD updates, dependency constraint adjustments, code modernization to use Python 3.13 features, and compatibility fixes for ONNX export and tests.

Migrates string-based enums from class Name(str, Enum) to class Name(StrEnum) pattern
Simplifies Generator type hints from Generator[T, None, None] to Generator[T]
Adds dynamo=False to ONNX export to use legacy JIT exporter due to Dynamo compatibility issues

Reviewed changes

Copilot reviewed 44 out of 45 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
`.github/workflows/pre_merge.yml`	Updated CI to test Python 3.13 alongside 3.10
`pyproject.toml`	Added Python 3.13 classifier, updated PyTorch dependency constraints, reorganized ruff configuration
`docs/source/snippets/install/source.txt`	Updated installation instructions to support Python >=3.10
`docs/source/markdown/guides/developer/contributing.md`	Updated development environment setup to support Python >=3.10
`src/anomalib/models/components/base/export_mixin.py`	Added `dynamo=False` to torch.onnx.export for legacy JIT exporter
`src/anomalib/data/utils/download.py`	Removed Python version check for tar.extract filter parameter
`src/anomalib/data/dataclasses/torch/base.py`	Moved TypeVar declaration after InferenceBatch class
`src/anomalib/data/dataclasses/numpy/depth.py`	Added missing dataclass decorator and type annotations
Multiple enum files	Migrated from `(str, Enum)` pattern to `StrEnum`
Multiple generator files	Simplified Generator type hints by removing None type parameters
`tests/unit/deploy/test_inferencer.py`	Relaxed tolerance from 1e-3 to 3e-3 for floating-point comparison

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

Copilot · 2025-11-23T14:06:34Z

pyproject.toml

 cu118 = [
-    "torch>=2.4.0,<=2.8.0",
-    "torchvision>=0.19.0",
+    "torch>=2.4.0,<=2.5.1 ; python_version < '3.13'",


The upper bound constraint <=2.5.1 for cu118 seems overly restrictive and inconsistent with the other CUDA extras. If PyTorch 2.5.1 is the last version supporting CUDA 11.8 on Python 3.12, this is correct. However, if newer versions exist that also support CUDA 11.8 on Python < 3.13, consider using an open upper bound like torch>=2.4.0 ; python_version < '3.13' to avoid unnecessary version locks.

Suggested change

"torch>=2.4.0,<=2.5.1 ; python_version < '3.13'",

"torch>=2.4.0 ; python_version < '3.13'",

Copilot · 2025-11-23T14:06:35Z

pyproject.toml

 cu121 = [
-    "torch>=2.4.0,<=2.8.0",
-    "torchvision>=0.19.0",
+    "torch>=2.4.0 ; python_version < '3.13'",


The cu121 extra has no upper bound for torch, which differs from cu118 that has <=2.5.1. This inconsistency may be intentional if CUDA 12.1 wheels continue to be built for Python < 3.13, but consider documenting why these constraints differ or making them consistent if both CUDA versions have the same support lifecycle.

Suggested change

"torch>=2.4.0 ; python_version < '3.13'",

"torch>=2.4.0,<=2.5.1 ; python_version < '3.13'",

Copilot · 2025-11-23T14:06:35Z

pyproject.toml

 # PyTorch dependency groups
 cpu = [
-    "torch>=2.4.0,<=2.8.0",
+    "torch>=2.4.0",


The removal of the upper bound <=2.8.0 for the cpu extra creates potential compatibility risks. If PyTorch 3.0+ introduces breaking changes, users could encounter issues. Consider retaining an upper bound (e.g., <3.0) or adding a comment explaining why the constraint was relaxed.

Suggested change

"torch>=2.4.0",

"torch>=2.4.0,<3.0",

Copilot · 2025-11-23T14:06:35Z

src/anomalib/data/dataclasses/numpy/depth.py

    >>> item.depth.shape
    (224, 224)
 """



[nitpick] The new imports on lines 29 and 31 are only used for the item_class type annotation on line 93. Consider adding these imports closer to where they're used, or add a comment explaining they're specifically for the ClassVar annotation added for Python 3.13 compatibility.

Suggested change

# The following imports are only used for the `item_class` type annotation in NumpyDepthBatch

# and are required for Python 3.13 compatibility with ClassVar.

Copilot · 2025-11-23T14:06:35Z

src/anomalib/data/dataclasses/torch/base.py

    pred_mask: torch.Tensor | None = None


+NumpyT = TypeVar("NumpyT")


[nitpick] Moving NumpyT after InferenceBatch appears to be for ordering purposes, but it creates a gap between related type definitions. Consider adding a comment explaining why this TypeVar is defined here rather than at the top with other type-related imports, especially since it's used by the ToNumpyMixin class below.

Copilot · 2025-11-23T14:06:36Z

pyproject.toml


 ignore = [
    # pydocstyle
+    "COM812",


The new COM812 ignore rule (missing trailing commas) is added without explanation. Consider adding a comment explaining why this specific rule is being ignored, especially since it appears in the middle of other documented rule ignores. This helps maintainers understand the rationale.

Suggested change

"COM812",

"COM812", # Missing trailing comma (ignored due to project style preference)

Copilot · 2025-11-23T14:06:36Z

docs/source/snippets/install/source.txt

 # Use of virtual environment is highy recommended
 # Using conda
-yes | conda create -n anomalib_env python=3.10
+yes | conda create -n anomalib_env python>=3.10


The conda create command syntax python>=3.10 is incorrect. Conda expects python=3.10 or python=3.13 but does not support constraint operators like >= in the package specification. This should be changed to a specific version like python=3.13 or python=3.10.

Suggested change

yes | conda create -n anomalib_env python>=3.10

yes | conda create -n anomalib_env python=3.10

Copilot · 2025-11-23T14:06:36Z

docs/source/markdown/guides/developer/contributing.md


   ```bash
-   conda create -n anomalib_dev python=3.10
+   conda create -n anomalib_dev python>=3.10


The conda create command syntax python>=3.10 is incorrect. Conda does not support constraint operators like >= in package specifications. Use a specific version like python=3.13 or python=3.10 instead.

Suggested change

conda create -n anomalib_dev python>=3.10

conda create -n anomalib_dev python=3.10

- Update pyproject.toml to require python >= 3.13 - Update tox.ini to test py313 - Remove legacy CUDA dependencies (11.8, 12.1) incompatible with PyTorch 2.5+ - Fix numpy array formatting issue in inferencer tests for Py3.13 compatibility Signed-off-by: Rickypanta0 <rickypanta.dev@gmail.com>

Copilot

Pull request overview

Copilot reviewed 82 out of 83 changed files in this pull request and generated 2 comments.

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

tests/unit/deploy/test_inferencer.py

Copilot · 2025-11-23T14:37:03Z

docs/source/snippets/install/source.txt

@@ -1,6 +1,6 @@
 # Use of virtual environment is highy recommended


Typo in comment: "highy" should be "highly".

Suggested change

# Use of virtual environment is highy recommended

# Use of virtual environment is highly recommended

samet-akcay

thanks for your contribution @Rickypanta0. I've added some comments.

samet-akcay · 2025-11-24T09:32:57Z

.github/workflows/_reusable-artifact-builder.yaml

 #
 # Required Inputs:
-# - python-version: Python version for building (default: "3.10")
+# - python-version: Python version for building (default: "3.11")


why this change?

samet-akcay · 2025-11-24T09:33:08Z

.github/workflows/_reusable-artifact-builder.yaml

 #        uses: ./.github/workflows/_reusable-artifact-builder.yaml
 #        with:
-#          python-version: "3.10"
+#          python-version: "3.11"


samet-akcay · 2025-11-24T09:33:15Z

.github/workflows/_reusable-artifact-builder.yaml

        description: "Python version for building"
        type: string
-        default: "3.10"
+        default: "3.11"


samet-akcay · 2025-11-24T09:33:36Z

.github/workflows/_reusable-release-validation.yaml

 #        with:
 #          version: "v1.2.3"
-#          python-version: "3.10"
+#          python-version: "3.11"


samet-akcay · 2025-11-24T09:33:47Z

.github/workflows/_reusable-release-validation.yaml

 #        with:
 #          version: "v1.2.3-rc1"
-#          python-version: "3.10"
+#          python-version: "3.11"


and here. It would be great to have consistency

My bad, would you prefer these examples/defaults to reflect the minimum supported version (reverting to 3.10, as it was) or the latest target version ?

samet-akcay · 2025-11-24T09:34:58Z

src/anomalib/data/dataclasses/numpy/depth.py

    """

-    item_class = NumpyDepthItem
+    item_class: ClassVar[Callable] = NumpyDepthItem


are these changes related to python version as well?

Yes, is indirectly related. Upgrading to 3.13, the pre-commit hooks fails with mypy with incompatibility error. I adopted for the fix instaed of suppressing it with # type: ignore[misc]. Here is the error log raised from the upgrade:

`mypy.....................................................................Failed

hook id: mypy
exit code: 1
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "_DepthInputFields" [misc]
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "_ImageInputFields" [misc]
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "NumpyBatch" [misc]
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "_GenericBatch" [misc]
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "UpdateMixin" [misc]
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "_OutputFields" [misc]
src/anomalib/data/dataclasses/numpy/depth.py:68: error: Definition of "replace" in base class "BatchIterateMixin" is incompatible with definition in base class "_InputFields" [misc]
Found 7 errors in 1 file (checked 396 source files)`

samet-akcay · 2025-11-24T09:35:26Z

src/anomalib/data/dataclasses/torch/base.py

+
 @dataclass
-class ToNumpyMixin(Generic[NumpyT]):
+class ToNumpyMixin(Generic[NumpyT]):  # noqa: UP046


same as above? Do we need to change all this when we change the python version?

Yes, this is triggered by the upgrade. Since the target is now compatible with py 3.12+. However i found that the new syntax for the bounded generics are less readable in my opinion, compared with the current implementation. I decided to suppress the errors with # noqa: UP046 to mantain readability.

samet-akcay · 2025-11-24T09:39:46Z

pyproject.toml

 readme = "README.md"
 description = "anomalib - Anomaly Detection Library"
-requires-python = ">=3.10"
+requires-python = ">=3.13"


I agree with some of the community comments that switching to Python 3.13 directly might break the compatibility. We should therefore stick to 3.10 for a while based on this schedule

Undestood, i will revert the compatibility to >=3.10 to keep backward compatibility while ensuring the CI/CD pipeline test against 3.13.

Rickypanta0 requested review from AlexanderBarabanov, ashwinvaidya17, mramotowski, rajeshgangireddy and samet-akcay as code owners November 13, 2025 11:47

Rickypanta0 force-pushed the chore/ci/upgrade-python-3-13 branch 4 times, most recently from 87bbf95 to aee7d36 Compare November 13, 2025 13:55

Rickypanta0 changed the title ~~Chore/ci/upgrade python 3 13~~ 🔧 chore(ci): upgrade python 3.13 Nov 19, 2025

Copilot AI review requested due to automatic review settings November 23, 2025 14:03

Copilot started reviewing on behalf of Rickypanta0 November 23, 2025 14:03 View session

Copilot finished reviewing on behalf of Rickypanta0 November 23, 2025 14:05

Copilot AI reviewed Nov 23, 2025

View reviewed changes

Rickypanta0 force-pushed the chore/ci/upgrade-python-3-13 branch 2 times, most recently from 0ac0b74 to 7985e04 Compare November 23, 2025 14:14

Copilot AI review requested due to automatic review settings November 23, 2025 14:32

Rickypanta0 force-pushed the chore/ci/upgrade-python-3-13 branch from 7985e04 to 204426f Compare November 23, 2025 14:32

Copilot started reviewing on behalf of Rickypanta0 November 23, 2025 14:33 View session

Copilot finished reviewing on behalf of Rickypanta0 November 23, 2025 14:36

Copilot AI reviewed Nov 23, 2025

View reviewed changes

samet-akcay requested changes Nov 24, 2025

View reviewed changes

	"torch>=2.4.0,<=2.5.1 ; python_version < '3.13'",
	"torch>=2.4.0 ; python_version < '3.13'",



	# The following imports are only used for the `item_class` type annotation in NumpyDepthBatch
	# and are required for Python 3.13 compatibility with ClassVar.

		pred_mask: torch.Tensor \| None = None


		NumpyT = TypeVar("NumpyT")

	"COM812",
	"COM812", # Missing trailing comma (ignored due to project style preference)

	yes \| conda create -n anomalib_env python>=3.10
	yes \| conda create -n anomalib_env python=3.10

	conda create -n anomalib_dev python>=3.10
	conda create -n anomalib_dev python=3.10

		@@ -1,6 +1,6 @@
		# Use of virtual environment is highy recommended

	# Use of virtual environment is highy recommended
	# Use of virtual environment is highly recommended

🔧 chore(ci): upgrade python 3.13 #3112

Are you sure you want to change the base?

🔧 chore(ci): upgrade python 3.13 #3112

Uh oh!

Conversation

Rickypanta0 commented Nov 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

📝 Description

✨ Changes

✅ Checklist

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI Nov 23, 2025

Choose a reason for hiding this comment

Uh oh!

samet-akcay left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Rickypanta0 commented Nov 13, 2025 •

edited

Loading