Haha, do you want to try to train a spam model on a subset?