Recommended way to test custom estimators

[mastoffel]

I'm developing a package where all models should be scikit-learn compatible. To test estimators, I'm currently doing this:

@parametrize_with_checks(
    [   NeuralNetwork(random_state=13),
        RadialBasis(),
    ]
)
def test_check_estimator(estimator, check):
    check(estimator)

However, I'd like to exclude some tests for some models. What's the recommended way to do this?
Thanks in advance!

[glemaitre]

This is the recommended way of doing. The skipped tests could be linked to the estimator tags and whether or not the estimator acknowledge or not the support for some specific feature: https://scikit-learn.org/stable/developers/develop.html#estimator-tags

[mastoffel]

Great, thanks!