If I try to pass some input to KBinsDiscretizer that would contain NaNs, it will throw an error instead of ignoring the NAs:

import numpy as np
from sklearn.preprocessing import KBinsDiscretizer

X = np.arange(10).reshape((-1,1))
X[2] = np.nan
kb = KBinsDiscretizer(encode="ordinal", strategy="quantile")
kb.fit(X)

ValueError: cannot convert float NaN to integer

The bins could still be calculated if one ignores the NAs and outputs them as np.nan in the transformation.