TST avoid FutureWarning due to n_features_in_ deprecation in Dummy* #20963
Conversation
| @@ -615,6 +615,7 @@ def fit(self, X, y, sample_weight=None, **fit_params): | |||
| return super().fit(X, y, sample_weight) | |||
|
|
|||
|
|
|||
| @pytest.mark.filterwarnings("ignore:`n_features_in_` is deprecated") | |||
There was a problem hiding this comment.
This one is worrying me. Here, it shows that a user will see the warning for some meta-estimators using the dummy estimator. Indeed, for the meta-estimators setting n_features_in_ depending on the inner estimator, we will check n_features_in_ and it will raise the warning with the dummy estimators.
It might be less of a big deal because it is only a dummy estimator and maybe not that many people are using it in a meta-estimator.
There was a problem hiding this comment.
If it gets annoying for users, we can define our own hasattr that ignores future warnings emitted by sklearn.
Edit: Most likely better to use a @property and @available_if. (If @available_if works with properties)
Let see if the CIs are passing first :) |
follow-up #20960
Removing the warning issued due to the deprecation of
n_features_in_inDummy*.