Dynamically handle future scikit-learn metadata routing deprecation by andersendsa · Pull Request #22723 · keras-team/keras

andersendsa · 2026-04-19T12:24:01Z

Description

This PR resolves a code health TODO by future-proofing the scikit-learn metadata routing check in keras/src/wrappers/fixes.py. It dynamically falls back to checking sklearn.version if the config key is removed, ensuring metadata routing correctly defaults to enabled in future scikit-learn versions while preserving backward compatibility.

Contributor Agreement

Please check all boxes below before submitting your PR for review:

I am a human, and not a bot.
I will be responsible for responding to review comments in a timely manner.
I will work with the maintainers to push this PR forward until submission.

Note: Failing to adhere to this agreement may result in your future PRs no longer being reviewed.

gemini-code-assist

Code Review

This pull request updates the _routing_enabled utility to handle missing sklearn installations and introduces version-based defaults for metadata routing, specifically enabling it for scikit-learn 1.3+. The review feedback highlights that the docstring is now outdated regarding default values and suggests a more concise implementation for the configuration and version checks.

gemini-code-assist · 2026-04-19T12:25:17Z

            Whether metadata routing is enabled. If the config is not set, it
            defaults to False.


The docstring is now inaccurate. It states that metadata routing defaults to False if the config is not set, but the new implementation defaults to True for scikit-learn versions >= 1.3 when the config key is missing. Please update the docstring to reflect this dynamic behavior.

gemini-code-assist · 2026-04-19T12:25:17Z

+    if "enable_metadata_routing" in config:
+        return config.get("enable_metadata_routing", False)
+
+    try:
+        from packaging.version import parse as parse_version
+
+        if parse_version(sklearn.__version__) >= parse_version("1.3"):
+            return True
+    except ImportError:
+        pass
+
+    return False


The implementation can be simplified. Since you've already checked if "enable_metadata_routing" is in the config, using .get(..., False) is redundant. The version check can also be more concise while catching potential parsing errors for robustness.

Suggested change

if "enable_metadata_routing" in config:

return config.get("enable_metadata_routing", False)

try:

from packaging.version import parse as parse_version

if parse_version(sklearn.__version__) >= parse_version("1.3"):

return True

except ImportError:

pass

return False

if "enable_metadata_routing" in config:

return config["enable_metadata_routing"]

try:

from packaging.version import parse

return parse(sklearn.__version__) >= parse("1.3")

except Exception:

return False

codecov-commenter · 2026-04-19T12:43:02Z

Codecov Report

❌ Patch coverage is 11.11111% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.04%. Comparing base (75d7bea) to head (5946b33).
⚠️ Report is 48 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/wrappers/fixes.py	11.11%	8 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #22723      +/-   ##
==========================================
- Coverage   83.15%   83.04%   -0.12%     
==========================================
  Files         596      596              
  Lines       68858    69889    +1031     
  Branches    10774    10894     +120     
==========================================
+ Hits        57259    58037     +778     
- Misses       8783     9004     +221     
- Partials     2816     2848      +32

Flag	Coverage Δ
keras	`82.86% <11.11%> (-0.12%)`	⬇️
keras-jax	`58.34% <11.11%> (-0.71%)`	⬇️
keras-numpy	`54.22% <11.11%> (-0.66%)`	⬇️
keras-openvino	`59.77% <11.11%> (+0.34%)`	⬆️
keras-tensorflow	`59.89% <11.11%> (-0.72%)`	⬇️
keras-torch	`58.70% <11.11%> (-0.70%)`	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

andersendsa · 2026-04-19T13:07:25Z

Hi @hertschuh this pr has passes all the tests and is ready for review

hertschuh · 2026-04-23T01:06:29Z

+        if parse_version(sklearn.__version__) >= parse_version("1.3"):
+            return True


Why would it be on all the time for sklearn >= 1.3 ?

hertschuh

Thanks for looking into this!

andersendsa · 2026-04-23T05:05:16Z

Hi @hertschuh i believe that meta data routing wasnt available until version 1.3

hertschuh · 2026-04-23T23:46:20Z

Hi @hertschuh i believe that meta data routing wasnt available until version 1.3

If that's the case, shouldn't the logic be this?

    if parse_version(sklearn.__version__) < parse_version("1.3"):
        return False

    config = sklearn.get_config()
    if "enable_metadata_routing" in config:
        return config.get("enable_metadata_routing", False)

    return False

andersendsa · 2026-04-24T01:52:46Z

Hi @hertschuh i believe that meta data routing wasnt available until version 1.3

If that's the case, shouldn't the logic be this?

    if parse_version(sklearn.__version__) < parse_version("1.3"):
        return False

    config = sklearn.get_config()
    if "enable_metadata_routing" in config:
        return config.get("enable_metadata_routing", False)

    return False

If we use that logic and a future version of scikit-learn (e.g., version 2.0) completely removes the enable_metadata_routing config key, your suggested logic would fall through to the final return False.

andersendsa · 2026-04-24T01:53:34Z

Hi @hertschuh i believe that meta data routing wasnt available until version 1.3

If that's the case, shouldn't the logic be this?
    if parse_version(sklearn.__version__) < parse_version("1.3"):
        return False

    config = sklearn.get_config()
    if "enable_metadata_routing" in config:
        return config.get("enable_metadata_routing", False)

    return False
@hertschuh If we use that logic and a future version of scikit-learn (e.g., version 2.0) completely removes the enable_metadata_routing config key, your suggested logic would fall through to the final return False.

hertschuh · 2026-04-24T17:39:11Z

+        from packaging.version import parse as parse_version
+


Move import to the top and remove try / except

packaging is a required dependency of the pypi wheel.

hertschuh · 2026-04-24T17:39:54Z

Hi @hertschuh i believe that meta data routing wasnt available until version 1.3

If that's the case, shouldn't the logic be this?
    if parse_version(sklearn.__version__) < parse_version("1.3"):
        return False

    config = sklearn.get_config()
    if "enable_metadata_routing" in config:
        return config.get("enable_metadata_routing", False)

    return False
@hertschuh If we use that logic and a future version of scikit-learn (e.g., version 2.0) completely removes the enable_metadata_routing config key, your suggested logic would fall through to the final return False.

Oh, I see. I misunderstood what the deprecation was about.

andersendsa · 2026-04-24T18:45:48Z

Hi @hertschuh I have done the changes this pr is ready for review

hertschuh

Thanks!

fix meta data routing

dd0567d

google-ml-butler Bot added the size:S label Apr 19, 2026

google-ml-butler Bot assigned gbaned Apr 19, 2026

gemini-code-assist Bot reviewed Apr 19, 2026

View reviewed changes

hertschuh reviewed Apr 23, 2026

View reviewed changes

hertschuh added the stat:awaiting response from contributor label Apr 23, 2026

google-ml-butler Bot removed the stat:awaiting response from contributor label Apr 23, 2026

hertschuh added the stat:awaiting response from contributor label Apr 23, 2026

google-ml-butler Bot removed the stat:awaiting response from contributor label Apr 24, 2026

hertschuh requested changes Apr 24, 2026

View reviewed changes

hertschuh added the stat:awaiting response from contributor label Apr 24, 2026

removing try/except

e33eb92

andersendsa requested a review from hertschuh April 24, 2026 18:33

google-ml-butler Bot added the awaiting review label Apr 24, 2026

format fix

5946b33

google-ml-butler Bot removed the stat:awaiting response from contributor label Apr 24, 2026

hertschuh approved these changes Apr 24, 2026

View reviewed changes

google-ml-butler Bot added kokoro:force-run ready to pull Ready to be merged into the codebase labels Apr 24, 2026

hertschuh merged commit 9eb6394 into keras-team:master Apr 24, 2026
11 checks passed

google-ml-butler Bot removed the awaiting review label Apr 24, 2026

google-ml-butler Bot removed ready to pull Ready to be merged into the codebase kokoro:force-run labels Apr 24, 2026

andersendsa deleted the sklearn_metadata_routing branch April 25, 2026 03:20

		Whether metadata routing is enabled. If the config is not set, it
		defaults to False.

		if parse_version(sklearn.__version__) >= parse_version("1.3"):
		return True

Conversation

andersendsa commented Apr 19, 2026

Description

Contributor Agreement

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist Bot Apr 19, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist Bot Apr 19, 2026

Choose a reason for hiding this comment

Uh oh!

codecov-commenter commented Apr 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

andersendsa commented Apr 19, 2026

Uh oh!

hertschuh Apr 23, 2026

Choose a reason for hiding this comment

Uh oh!

hertschuh left a comment

Choose a reason for hiding this comment

Uh oh!

andersendsa commented Apr 23, 2026

Uh oh!

hertschuh commented Apr 23, 2026

Uh oh!

andersendsa commented Apr 24, 2026

Uh oh!

andersendsa commented Apr 24, 2026

Uh oh!

hertschuh Apr 24, 2026

Choose a reason for hiding this comment

Uh oh!

hertschuh commented Apr 24, 2026

Uh oh!

andersendsa commented Apr 24, 2026

Uh oh!

hertschuh left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

codecov-commenter commented Apr 19, 2026 •

edited

Loading