In an ML classification problem, before training the model for the last time, we select the number of epochs that

maximizes the validation set accuracy

minimizes the validation set accuracy

minimizes the training set accuracy

maximizes the testing set accuracy

minimize the testing set accuracy