Proper way to subclass from sklearn with * argument

Question

I am trying to subclass from sklearn.svm.LinearSVC and noticed the * argument in the signature. I'm not sure if this * refers to **kwargs or *args or something else. I am trying subclass the init function as follows. In this scenario I'm have added a single additional argument new_string_in_subclass the init function. from sklearn.svm import LinearSVC

class LinearSVCSub(LinearSVC):
    def __init__(self, penalty='l2', loss='squared_hinge', *, dual=True, tol=0.0001, C=1.0, multi_class='ovr',
                 fit_intercept=True, intercept_scaling=1, class_weight=None, verbose=0, random_state=None,
                 max_iter=1000, sampler: new_string_in_subclass=None):

        super(LinearSVCSub, self).__init__(penalty=penalty, loss=loss, *, dual=dual, tol=tol,
                                            C=C, multi_class=multi_class, fit_intercept=fit_intercept,
                                                  intercept_scaling=intercept_scaling, class_weight=class_weight,
                                                  verbose=verbose, random_state=random_state, max_iter=max_iter)

        self.new_string_in_subclass = new_string_in_subclass

If I want to maintain the functionality of the LinearSVC class's other methods, do I need to pass the * argument to the super class's __init__ function? If so how do I do this? Right now I get a SyntaxError as below:

super(LinearSVCSub, self).init(penalty=penalty, loss=loss, *, dual=dual, tol=tol, ^ SyntaxError: invalid syntax

This `*` syntax is for keyword-only arguments. It was added in Python 3.8. See [Positional-only parameters](https://docs.python.org/3/whatsnew/3.8.html#positional-only-parameters) in the release notes. — Pascal Getreuer, Oct 19 '20 at 01:36

score 4 · Accepted Answer · answered Oct 19 '20 at 01:32

4

A * only goes in the definition of a method, not in calls to that method. The * indicates that all parameters after it can only be supplied by name, not by position. Just leave it out when calling the super, and you should be fine.

answered Oct 19 '20 at 01:32

CryptoFool

21,719
5
26
44

I've posted [another question](https://stackoverflow.com/q/64430351/11542679) about another problem I ran into after this was sorted out. Appreciate if you could have a look. It's about using the above setup with sklearn `GridSearchCV ` – Hawklaz Oct 19 '20 at 15:25

score 1 · Answer 2 · answered Oct 19 '20 at 01:32

1

The asterisk serves to separate keyword-only parameters: Bare asterisk in function arguments? You should skip it when instantiating the sklearn class.

answered Oct 19 '20 at 01:32

Ben Reiniger

10,517
3
16
29

Proper way to subclass from sklearn with * argument

2 Answers2

Linked