Using `cross_val_predict` with `RepeatedKFold`

It would be nice to use RepeatedKFold with with cross_val_predict. It’s possible use cross_val_score with repeated K folds, but a ValueError is thrown with predict

model = LogisticRegression()
rkf = RepeatedKFold(n_splits=10, n_repeats=10)
prediction = cross_val_predict(model, X, y, cv=rkf)
score = cross_val_score(model, X, y, cv=rkf)

Throws ValueError: cross_val_predict only works for partitions.

[jnothman]

What do you think it should return???

[Carbiner]

Just came across this, and would appreciate this feature. My use case is determining if certain observations we have are consistently mislabeled across different classifiers and different splits, to help determine what data we should be collecting and focusing on.