def test_check_cv():
# No y, classifier=False
cv = check_cv(3, classifier=False)
assert isinstance(cv, KFold) and cv.n_splits == 3
cv = check_cv(5, classifier=False)
assert isinstance(cv, KFold) and cv.n_splits == 5
# y, classifier = False
dy = da.from_array(np.array([1, 0, 1, 0, 1]), chunks=2)
with assert_dask_compute(False):
assert isinstance(check_cv(y=dy, classifier=False), KFold)
# Binary and multi-class y
for y in [
np.array([0, 1, 0, 1, 0, 0, 1, 1, 1]),
np.array([0, 1, 0, 1, 2, 1, 2, 0, 2])
]:
cv = check_cv(5, y, classifier=True)
assert isinstance(cv, StratifiedKFold) and cv.n_splits == 5
dy = da.from_array(y, chunks=2)
with assert_dask_compute(True):
cv = check_cv(5, dy, classifier=True)
assert isinstance(cv, StratifiedKFold) and cv.n_splits == 5
# Non-binary/multi-class y
y = np.array([[1, 2], [0, 3], [0, 0], [3, 1], [2, 0]])
assert isinstance(check_cv(y=y, classifier=True), KFold)
dy = da.from_array(y, chunks=2)
with assert_dask_compute(True):
assert isinstance(check_cv(y=dy, classifier=True), KFold)
# Old style
cv = [np.array([True, False, True]), np.array([False, True, False])]
with assert_dask_compute(False):
assert isinstance(check_cv(cv, y=dy, classifier=True),
_CVIterableWrapper)
# CV instance passes through
y = da.ones(5, chunks=2)
cv = ShuffleSplit()
with assert_dask_compute(False):
assert check_cv(cv, y, classifier=True) is cv
assert check_cv(cv, y, classifier=False) is cv
def test_check_cv():
# No y, classifier=False
cv = check_cv(3, classifier=False)
assert isinstance(cv, KFold) and cv.n_splits == 3
cv = check_cv(5, classifier=False)
assert isinstance(cv, KFold) and cv.n_splits == 5
# y, classifier = False
dy = da.from_array(np.array([1, 0, 1, 0, 1]), chunks=2)
with assert_dask_compute(False):
assert isinstance(check_cv(y=dy, classifier=False), KFold)
# Binary and multi-class y
for y in [np.array([0, 1, 0, 1, 0, 0, 1, 1, 1]),
np.array([0, 1, 0, 1, 2, 1, 2, 0, 2])]:
cv = check_cv(5, y, classifier=True)
assert isinstance(cv, StratifiedKFold) and cv.n_splits == 5
dy = da.from_array(y, chunks=2)
with assert_dask_compute(True):
cv = check_cv(5, dy, classifier=True)
assert isinstance(cv, StratifiedKFold) and cv.n_splits == 5
# Non-binary/multi-class y
y = np.array([[1, 2], [0, 3], [0, 0], [3, 1], [2, 0]])
assert isinstance(check_cv(y=y, classifier=True), KFold)
dy = da.from_array(y, chunks=2)
with assert_dask_compute(True):
assert isinstance(check_cv(y=dy, classifier=True), KFold)
# Old style
cv = [np.array([True, False, True]), np.array([False, True, False])]
with assert_dask_compute(False):
assert isinstance(check_cv(cv, y=dy, classifier=True),
_CVIterableWrapper)
# CV instance passes through
y = da.ones(5, chunks=2)
cv = ShuffleSplit()
with assert_dask_compute(False):
assert check_cv(cv, y, classifier=True) is cv
assert check_cv(cv, y, classifier=False) is cv
def fit(self, X, y=None, groups=None, **fit_params):
if self.verbose:
n_fitting_tasks = len(self._get_param_iterator()) * check_cv(
self.cv).n_splits + int(self.refit)
self.estimator._create_progress_bar(n_fitting_tasks,
self.scheduler)
super(ModelSearchCV, self).fit(X, y, groups, **fit_params)
self._modify_grid_search_attrs()
return self