def test_Chained__init__():
c1 = Chained(sgd1)
c2 = Chained(sgd2)
assert c1._name != c2._name
assert Chained(sgd1)._name == c1._name
with pytest.raises(TypeError):
Chained("not an estimator")
# No partial_fit
lr = LogisticRegression()
with pytest.raises(ValueError):
Chained(lr)
def test_fit_dask_matrix():
X_bag = db.from_sequence([X_iris[0:20], X_iris[20:40], X_iris[40:]])
y_bag = db.from_sequence([y_iris[0:20], y_iris[20:40], y_iris[40:]])
X = dm.from_bag(X_bag)
y = dm.from_bag(y_bag)
c = Chained(sgd1)
fit_test(c, X, y)
def test_score():
X = da.from_array(X_iris, chunks=4)
y = da.from_array(y_iris, chunks=4)
c = Chained(sgd1)
fit = c.fit(X, y)
s = fit.score(X_iris, y_iris)
assert isinstance(s, Delayed)
res = s.compute()
assert isinstance(res, float)
s = fit.score(X, y)
assert isinstance(s, Delayed)
res = s.compute()
assert isinstance(res, float)
will_error = c.score(X_iris, y_iris)
with pytest.raises(NotFittedError):
will_error.compute()
def test_predict():
X = da.from_array(X_iris, chunks=4)
y = da.from_array(y_iris, chunks=4)
c = Chained(sgd1)
fit = c.fit(X, y)
pred = fit.predict(X_iris)
assert isinstance(pred, Delayed)
res = pred.compute()
assert isinstance(res, np.ndarray)
pred = fit.predict(X)
assert isinstance(pred, da.Array)
res = pred.compute()
assert isinstance(res, np.ndarray)
will_error = c.predict(X_iris)
with pytest.raises(NotFittedError):
will_error.compute()
def test_set_params():
c = Chained(sgd1)
c2 = c.set_params(penalty='l2')
assert isinstance(c2, Chained)
c3 = Chained(sgd2)
assert c2._name == c3._name # set_params name equivalent to init
# Check no mutation
assert c2.get_params()['penalty'] == 'l2'
assert c2.compute().penalty == 'l2'
assert c.get_params()['penalty'] == 'l1'
assert c.compute().penalty == 'l1'
def test_getattr():
c = Chained(sgd1)
assert c.penalty == sgd1.penalty
with pytest.raises(AttributeError):
c.not_a_real_parameter
def test_setattr():
c = Chained(sgd1)
with pytest.raises(AttributeError):
c.penalty = 'l2'
def test_get_params():
c = Chained(sgd1)
assert c.get_params() == sgd1.get_params()
assert c.get_params(deep=False) == sgd1.get_params(deep=False)
def test__estimator_type():
c = Chained(sgd1)
assert c._estimator_type == sgd1._estimator_type
def test_clone():
c = Chained(sgd1)
c2 = clone(c)
assert isinstance(c2, Chained)
assert c._name == c2._name
assert c._est is not c2._est
def test_tokenize_Chained():
c1 = Chained(sgd1)
c2 = Chained(sgd2)
assert tokenize(c1) == tokenize(c1)
assert tokenize(c1) != tokenize(c2)
assert tokenize(c1) != tokenize(Wrapped(sgd1))
def test_Chained_from_sklearn():
c1 = Chained(sgd1)
assert Chained.from_sklearn(sgd1)._name == c1._name
assert Chained.from_sklearn(c1) is c1
def test_fit_dask_array():
X = da.from_array(X_iris, chunks=4)
y = da.from_array(y_iris, chunks=4)
c = Chained(sgd1)
fit_test(c, X, y)
def test_repr():
c = Chained(sgd1)
res = repr(c)
assert res.startswith('Chained')
def test_dir():
c = Chained(sgd1)
attrs = dir(c)
assert 'penalty' in attrs