def test_cross_entropy_fit():
(x, y), (_, _) = tf.keras.datasets.mnist.load_data()
train = pair.BalancedPairDataset(x, y, n_pairs=10, input_shape=(x.shape[0], 784))
new_base = mlp.MLP()
new_siamese = cross_entropy.CrossEntropySiamese(new_base)
new_siamese.compile(optimizer=tf.optimizers.Adam(learning_rate=0.001))
new_siamese.fit(train.batches, epochs=1)
def test_cross_entropy_distance_setter():
new_base = mlp.Base()
new_siamese = cross_entropy.CrossEntropySiamese(new_base)
try:
new_siamese.distance = "a"
except:
new_siamese.distance == "concat"
assert new_siamese.distance == "concat"
def test_cross_entropy_step():
(x, y), (_, _) = tf.keras.datasets.mnist.load_data()
new_base = mlp.MLP()
new_siamese = cross_entropy.CrossEntropySiamese(new_base)
new_siamese.compile(optimizer=tf.optimizers.Adam(learning_rate=0.001))
x1 = tf.ones((1, 784))
x2 = tf.ones((1, 784))
y = tf.zeros(1)
new_siamese.step(x1, x2, y)
def test_cross_entropy_distance():
new_base = mlp.Base()
new_siamese = cross_entropy.CrossEntropySiamese(new_base)
assert new_siamese.distance == "concat"