def test_network(self, network, x, input, layer, layer2, unsupervised_layer, caplog):
net = network(name='test_network', layers=[input, layer, unsupervised_layer])
assert net.params == [layer.W, layer.b, unsupervised_layer.W, unsupervised_layer.b]
net = network(name='test_network')
net.add(input)
net.add(layer)
assert net.reguls == 0
assert net.params == [layer.W, layer.b]
assert net.name == 'test_network'
assert net.has_unsupervised_layer is False
net.add(unsupervised_layer)
assert net.has_unsupervised_layer is True
from yadll.utils import to_float_X
x_val = to_float_X(np.random.random((50, 25)))
assert (net.get_output().eval({x: x_val}) == unsupervised_layer.get_output().eval({x: x_val})).all()
# Save and load Params
net.save_params('test.yp')
net2 = network(name='test_network', layers=[input, layer2, unsupervised_layer])
def params_equal(params1, params2):
for i in range(len(params1)):
np.testing.assert_array_equal(params1[i].eval(),
params2[i].eval())
np.testing.assert_raises(AssertionError, params_equal,
net.params, net2.params)
net2.load_params('test.yp')
params_equal(net.params, net2.params)
caplog.setLevel(logging.ERROR)
net3 = network(name=None)
net3.add(input)
net3.add(layer)
net3.save_params('test.yp')
assert 'Your network has no name. Please set one and try again.'in caplog.text()
net3.load_params('test.yp')
assert 'Network names are different. Saved network name is: test_network' in caplog.text()
def test_to_float_X():
from yadll.utils import to_float_X
x = np.asarray(np.random.normal(size=(10, 5)))
assert to_float_X(x).dtype == theano.config.floatX
