def test_one_hot_vector(self):
one_hot_vector = arrays.one_hot(indices=[1],
shape=(3, 1),
dtype=np.float32)
print('one_hot_vector: {}'.format(one_hot_vector))
assert one_hot_vector.dtype == np.float32
assert one_hot_vector[0] == 0
assert one_hot_vector[1] == 1
def test_one_hot_off_value_float(self):
one_hot_vector = arrays.one_hot(indices=[0, 1, 2],
shape=(3, 3),
off_value=0.0)
assert one_hot_vector.dtype == np.float32
assert one_hot_vector[0][0] == 1
assert one_hot_vector[1][1] == 1
assert one_hot_vector[2][2] == 1
def preprocess_actions(actions, num_actions):
vectorized_actions = []
for action in actions:
vectorized_action = arrays.one_hot(indices=[action],
shape=(num_actions, 1))
vectorized_actions.append(vectorized_action)
assert vectorized_actions[0].shape == (num_actions, 1)
return vectorized_actions
def test_one_hot_matrix(self):
one_hot_vector = arrays.one_hot(indices=[0, 1, 1],
shape=(3, 3),
dtype=np.int8)
print('one_hot_vector: {}'.format(one_hot_vector))
assert one_hot_vector.dtype == np.int8
assert one_hot_vector[0][0] == 1
assert one_hot_vector[0][1] == 0
assert one_hot_vector[1][1] == 1
assert one_hot_vector[1][0] == 0
assert one_hot_vector[1][2] == 1
assert one_hot_vector[2][0] == 0
def test_one_hot_3x3_tensor(self):
with pytest.raises(ValueError):
arrays.one_hot(indices=range(9), shape=(3, 3, 3), off_value=0.0)
def test_one_hot_no_dtype(self):
one_hot_vector = arrays.one_hot(indices=[0, 1, 2], shape=(3, 3))
assert one_hot_vector.dtype == np.int8
assert one_hot_vector[0][0] == 1
assert one_hot_vector[1][1] == 1
assert one_hot_vector[2][2] == 1