I am using k-fold cross-validation to train my ml model, where k=4. I understand that 4 models are trained with different subsets of input data. So out of the 4 models, which one is my final artifact that I can use? Did I misunderstand CV in general?