How to use group identities as estimators for SkLearn Cross-Validation?

Question

Is there a way in python to run cross-vlidation without a given estimator (cross-validation from sk learn always requires one); or how could one only use the group identity as e