Fix deprecation warning force_all_finite -> ensure_all_finite for sklearn>=1.6

[icarosadero]

The current CEBRA branch raised the following deprecation message

FutureWarning: 'force_all_finite' was renamed to 'ensure_all_finite' in 1.6 and will be removed in 1.8.

See the deprecation warning in the docs

This PR adds a wrapper to cebra/integrations/sklearn/utils.py which manually adapts the argument to force_all_finite if an outdated sklearn version is used.

(edited by @stes )

[stes]

Hi @icarosadero , thanks a lot for fixing! Do you happen to know when the ensure_all_finite was introduced in scikit-learn and have a reference to the docs? As the code will require this as the new minimal version then to function, I guess?

[stes]

ok so it seems that this is indeed an issue also for recent versions of scikit-learn:

If you want to take a shot at solving this, have a look at e.g. this snippet of code

CEBRA/cebra/integrations/sklearn/cebra.py

Lines 46 to 50 in 5f46c32

	def check_version(estimator):
	# NOTE(stes): required as a check for the old way of specifying tags
	# https://github.com/scikit-learn/scikit-learn/pull/29677#issuecomment-2334229165
	from packaging import version
	return version.parse(sklearn.__version__) < version.parse("1.6.dev")

You could then implement a conditional part where the version is checked, and the arg is passed based on the version of sklearn.

[stes]

I made a suggestion which is a bit more crisp and avoids code duplication

[stes]

could you also update the PR description with the reference to the docs, pointing that this API change happend from 1.7.x to 1.8.x?

[stes]

lgtm (wait with merge until #212 is merged, though)

[stes]

one final change, the threshold to fix the arg is 1.6, not 1.8

[stes]

all right, final from my end

[MMathisLab]

Lgtm; did not functionally test