A note on generalization bounds for losses with finite moments