def test_model_error_returns_nan():
with mock.patch('sklearn.base.clone', lambda x: x):
mock_model = mock.MagicMock()
def mock_fit(*args, **kwargs):
raise ValueError()
mock_model.fit = mock_fit
with pytest.warns(ModelFitWarning):
scores = cross_val_score(mock_model,
y,
scoring='mean_squared_error',
cv=SlidingWindowForecastCV(
window_size=100, step=24, h=1),
verbose=0)
assert np.isnan(scores).all()
# if the error_score is 'raise', we will raise
with pytest.raises(ValueError):
cross_val_score(mock_model,
y,
scoring='mean_squared_error',
cv=SlidingWindowForecastCV(window_size=100,
step=24,
h=1),
verbose=0,
error_score='raise')
def test_cv_scores(cv, est, verbose):
scores = cross_val_score(est,
y,
scoring='mean_squared_error',
cv=cv,
verbose=verbose)
assert isinstance(scores, np.ndarray)
