change CV logic so that when no splits are given, the whole dataset is considered train data
change CV logic so that when no splits are given, the whole dataset is considered train data