for num_features in [2, 3, 4, 5, 6]:
    model = sklearn.naive_bayes.MultinomialNB()
    model.fit(feature_vecs[:, :num_features], labels)