Returning heldout predictions GCV? (e.g. store_cv_values from sklearn's RidgeCV)

[ColinConwell]

Is it possible to return the cross-validated predictions from the GCV process? sklearn's RidgeCV (which natively uses GCV) has an option for this, and I'm wondering if there's something similar in PyGam. This would effectively equate to retaining in some variable each of the heldout predictions (one for each sample).