def test_generator_faster_than_threaded():
fe_threaded = models.FlowerClassifierThreaded(threaded=True)
fe_generator = models.FlowerClassifierGenerator()
n_epochs = 1000
print('starting threaded')
t1 = time.time()
for _ in range(n_epochs):
predictions = list(fe_threaded.predict(features=predict_x))
print('starting generator')
t2 = time.time()
for _ in range(n_epochs):
predictions = list(fe_generator.predict(features=predict_x))
t3 = time.time()
threaded_time = (t2 - t1)
generator_time = (t3 - t2)
assert generator_time < threaded_time
print(f'Threaded time was {threaded_time}; s\n'
f'Generator time was {generator_time}; s\n'
f'Generator was {threaded_time/generator_time} times faster!')
def test_threaded_faster_than_non_threaded():
fe_threaded = models.FlowerClassifierThreaded(threaded=True)
fe_unthreaded = models.FlowerClassifier()
n_epochs = 100
print('starting unthreaded')
t1 = time.time()
for _ in range(n_epochs):
predictions = list(fe_unthreaded.predict(features=predict_x))
print('starting threaded')
t2 = time.time()
for _ in range(n_epochs):
predictions = list(fe_threaded.predict(features=predict_x))
t3 = time.time()
unthreaded_time = (t2 - t1)
threaded_time = (t3 - t2)
assert unthreaded_time > threaded_time
print(f'Threaded time was {threaded_time}; s\n'
f'Unthreaded time was {unthreaded_time}; s\n'
f'Threaded was {unthreaded_time/threaded_time} times faster!')
def test_predictions_change_with_training():
fe = models.FlowerClassifierThreaded(threaded=False)
predictions1 = list(fe.predict(features=predict_x))
fe.train(steps=100)
predictions2 = list(fe.predict(features=predict_x))
with pytest.raises(AssertionError):
np.testing.assert_array_equal(predictions1[0]['logits'],
predictions2[0]['logits'])
def test_normal_input_fn():
fe = models.FlowerClassifierThreaded(threaded=False)
ds = fe.predict_input_fn(predict_x)
value = ds.make_one_shot_iterator().get_next()
with tf.Session() as sess:
features = sess.run(value)
assert isinstance(features, dict)
def test_iris_estimator_predict_deterministic(threaded):
fe = models.FlowerClassifierThreaded(threaded=threaded)
predictions1 = fe.predict(features=predict_x)
predictions2 = fe.predict(features=predict_x)
if not threaded:
predictions1 = list(predictions1)[0]
predictions2 = list(predictions2)[0]
print(threaded, predictions1)
print(threaded, predictions2)
np.testing.assert_array_equal(predictions1['logits'],
predictions2['logits'])
def test_iris_estimator_trains():
fe = models.FlowerClassifierThreaded(threaded=False)
fe.train(steps=10)