The basics of cross validation are as follows: the model is trained on a training set and then evaluated once per validation set.