def setUp(self):
X = np.random.random(size=(1000, 5))
X = vigra.taggedView(X, axistags='tc')
y = np.random.randint(0, 2, size=(1000,))
y = vigra.taggedView(y, axistags='t')
self.X = X
self.y = y
self.g = Graph()
self.pipeData = OpArrayPiperWithAccessCount(graph=self.g)
self.pipeTarget = OpArrayPiperWithAccessCount(graph=self.g)
self.op = OpDataset(graph=self.g)
self.op.Input.resize(2)
self.op.Input[0].connect(self.pipeData.Output)
self.op.Input[1].connect(self.pipeTarget.Output)
self.pipeData.Input.setValue(self.X)
self.pipeTarget.Input.setValue(self.y)
class TestOpDataset(unittest.TestCase):
def setUp(self):
X = np.random.random(size=(1000, 5))
X = vigra.taggedView(X, axistags='tc')
y = np.random.randint(0, 2, size=(1000,))
y = vigra.taggedView(y, axistags='t')
self.X = X
self.y = y
self.g = Graph()
self.pipeData = OpArrayPiperWithAccessCount(graph=self.g)
self.pipeTarget = OpArrayPiperWithAccessCount(graph=self.g)
self.op = OpDataset(graph=self.g)
self.op.Input.resize(2)
self.op.Input[0].connect(self.pipeData.Output)
self.op.Input[1].connect(self.pipeTarget.Output)
self.pipeData.Input.setValue(self.X)
self.pipeTarget.Input.setValue(self.y)
def test_adjust_for_viewer(self):
self.op.adjust_for_viewer(self.X)
def test_get_num_examples(self):
n = self.op.get_num_examples()
np.testing.assert_equal(n, self.X.shape[0])
def test_has_targets(self):
assert self.op.has_targets()
def test_iterator(self):
k = 5
i = self.op.iterator(mode='sequential', batch_size=k)
for batch in i:
np.testing.assert_equal(len(batch), k)
i = self.op.iterator(mode=None, num_batches=k)
batches = [b for b in i]
np.testing.assert_equal(len(batches), k)
assert all([b.dtype == np.float32 for b in batches])
i = self.op.iterator(rng=0, mode="no_real_mode")
batches = [b for b in i]
np.testing.assert_equal(len(batches), 1)
np.testing.assert_equal(len(batches[0]), len(self.X))
i = self.op.iterator(mode='sequential', num_batches=k)
i = iter(i)
n = i.next()
assert isinstance(n, np.ndarray)
i = self.op.iterator(mode='sequential', num_batches=k,
return_tuple=True)
i = iter(i)
n = i.next()
assert isinstance(n, tuple)
np.testing.assert_equal(len(n), 1)
assert n[0].dtype == np.float32
assert not i.stochastic
