My understanding ofcross validation is that, training data is divided in to n folds. For each fold, a model is trained on all other folds and validated on the selected fold. At the end we will have n models and n validation results. We average these result out to get the cross-validation test result.