No attribute classes_ during multi-class scoring #26336
Comments
|
Thanks for reporting. Could you provide what is the class of |
|
I had a look at the I think that it misses a method to wire the @property
def classes_(self):
return self.model.classes_ |
|
For completeness, the current implementation of scikit-learn/sklearn/utils/estimator_checks.py Line 2225 in 8521819 |
|
That does indeed fix the issue, I'll add that! Not sure if there is anything to fix at the |
|
... but at our end do you suggest we use this |
|
Yep
It makes sense to try it. If the tests are passing then, it is quite safe to use will any component of scikit-learn. Note that we always have a mechanism to skip some of the common tests if it fails for a good reason. |
|
Thanks for the quick response, and sorry that we should have caught that it was our faulty implementation in the end! |
|
Happy to be helpful ;) |
Describe the bug
Regression we hit in MNE-Python's
pip-prerun and bisected locally:https://github.com/mne-tools/mne-python/actions/runs/4894775241/jobs/8739516519#step:17:4138
Local bisect suggests the culprit is #26037 by @glemaitre
It's entirely possible we're doing something wrong at the MNE-Python end but this code has worked for years 🤷
cc @agramfort who might also know the problem quickly
Steps/Code to Reproduce
Not really minimal, but after cloning MNE-Python you can do:
And it will fail.
Can try to whittle it down more if it would help...
Expected Results
No error
Actual Results
mne/decoding/tests/test_base.py:323: in test_get_coef_multiclass_full scores = cross_val_multiscore(time_gen, X, y, cv=cv, verbose=True) ... mne/decoding/base.py:567: in _score score = scorer(estimator, X_test, y_test) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/sklearn/metrics/_scorer.py:411: in _passthrough_scorer return estimator.score(*args, **kwargs) mne/decoding/search_light.py:581: in score score = parallel( /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/parallel.py:1085: in __call__ if self.dispatch_one_batch(iterator): /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/parallel.py:901: in dispatch_one_batch self._dispatch(tasks) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/parallel.py:819: in _dispatch job = self._backend.apply_async(batch, callback=cb) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/_parallel_backends.py:208: in apply_async result = ImmediateResult(func) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/_parallel_backends.py:597: in __init__ self.results = batch() /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/parallel.py:288: in __call__ return [func(*args, **kwargs) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/joblib/parallel.py:288: in <listcomp> return [func(*args, **kwargs) mne/decoding/search_light.py:676: in _gl_score _score = scoring(est, X[..., jj], y) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/sklearn/metrics/_scorer.py:211: in __call__ return self._score( /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/sklearn/metrics/_scorer.py:291: in _score y_pred = method_caller(clf, "predict_proba", X, pos_label=self._get_pos_label()) /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/sklearn/metrics/_scorer.py:77: in _cached_call result, _ = _get_response_values( /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/sklearn/utils/_response.py:74: in _get_response_values classes = estimator.classes_ /opt/hostedtoolcache/Python/3.11.3/x64/lib/python3.11/site-packages/sklearn/pipeline.py:744: in classes_ return self.steps[-1][1].classes_ E AttributeError: 'LinearModel' object has no attribute 'classes_'Versions
The text was updated successfully, but these errors were encountered: