def test_module_output_not_1d(self, net_cls, data):
from skorch.toy import make_classifier
module = make_classifier(
input_units=20,
output_units=1,
) # the output will not be squeezed
net = net_cls(module, max_epochs=1)
net.fit(*data) # does not raise
def classifier_module():
"""Return a simple classifier module class."""
from skorch.toy import make_classifier
return make_classifier(
input_units=20,
hidden_units=10,
num_hidden=2,
dropout=0.5,
)
def test_module_output_2d_raises(self, net_cls, data):
from skorch.toy import make_classifier
module = make_classifier(
input_units=20,
output_units=2,
)
net = net_cls(module, max_epochs=1)
with pytest.raises(ValueError) as exc:
net.fit(*data)
msg = exc.value.args[0]
expected = ("Expected module output to have shape (n,) or "
"(n, 1), got (128, 2) instead")
assert msg == expected
def train():
X, y = make_classification(1000, 20, n_informative=10, random_state=0)
X = X.astype(np.float32)
y = y.astype(np.int64)
module = make_classifier(input_units=20)
net = NeuralNetClassifier(
module,
max_epochs=10,
lr=0.1,
callbacks=[TriggerKeyError(), PrintMemory()],
device='cuda',
)
return net.fit(X, y)
def test_passes_kwargs_to_neuralnet_optimizer(
self, filtered_optimizer, filter_requires_grad):
from skorch import NeuralNetClassifier
from skorch.toy import make_classifier
module_cls = make_classifier(
input_units=1,
num_hidden=0,
output_units=1,
)
with pytest.warns(DeprecationWarning):
opt = filtered_optimizer(torch.optim.SGD, filter_requires_grad)
net = NeuralNetClassifier(
module_cls, optimizer=opt, optimizer__momentum=0.9)
net.initialize()
assert isinstance(net.optimizer_, torch.optim.SGD)
assert len(net.optimizer_.param_groups) == 1
assert net.optimizer_.param_groups[0]['momentum'] == 0.9
def test_make_classifier(self):
from skorch.toy import make_classifier
module = make_classifier()()
assert isinstance(module.sequential[-1], nn.Softmax)
