def test_queuemask():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', get_queues=(True, False, False), val_size=0, _rand_data=True)
assert test_queue is None
assert val_queue is None
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', get_queues=(True, False, True), _rand_data=True)
assert test_queue is None
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', get_queues=(False, True, False), val_size=0, _rand_data=True)
assert train_queue is None
assert val_queue is None
def test_pullfromqueue():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', _rand_data=True)
sleep(1)
data, labels = train_queue.get_batch(100)
test, labels = test_queue.get_batch(100)
val, labels = val_queue.get_batch(100)
def test_lastbatch():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', _rand_data=True)
sleep(1)
train_queue._read_count = train_queue.epoch_size - 200
data, labels = train_queue.get_batch(100)
assert not train_queue.last_batch
data, labels = train_queue.get_batch(100)
assert train_queue.last_batch
def test_transform():
def transform(x):
return x-np.mean(x)
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', transform=transform, _rand_data=True)
sleep(1)
data, labels = train_queue.get_batch(100)
for im in data:
assert abs(np.mean(im)) <= 0.0001
def test_epochlimit():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', max_epochs=1, _rand_data=True)
sleep(1)
epoch_size = train_queue.epoch_size
for i in range(int(np.ceil(epoch_size//100))):
data, labels = train_queue.get_batch(100)
with pytest.raises(FileQueueDepleted, message="Expecting queue empty exception at epoch limit"): # noqa
data, labels = train_queue.get_batch(100)
def test_transform2():
def transform_tr(x):
return x-np.mean(x)
def transform_te(x):
return x-np.mean(x) + 100
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', transform=(transform_tr, transform_te, None), _rand_data=True)
sleep(1)
data, labels = train_queue.get_batch(100)
for im in data:
assert abs(np.mean(im)) <= 0.0001
data, labels = test_queue.get_batch(100)
for im in data:
assert abs(np.mean(im)-100) <= 0.0001
def test_lastbatch_reset():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', _rand_data=True)
sleep(1)
train_queue._read_count = train_queue.epoch_size - 200
for i in range(2):
while not train_queue.last_batch:
data, labels = train_queue.get_batch(100)
if i == 0:
old_data = data
train_queue._read_count = train_queue.epoch_size - 100
# test that the last_batch flag was reset properly so that the data should
# be different now
old_data = np.array(old_data)
data = np.array(data)
with pytest.raises(AssertionError, message="Expecting Different arrays"):
np.testing.assert_array_equal(old_data, data)
def test_getqueuenoval():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', val_size=0, _rand_data=True)
def test_getqueue():
train_queue, test_queue, val_queue = cifar.get_cifar_queues(
'', _rand_data=True)