def test_incorrect_input_vals2():
def func3(x, y):
f1 = (3 * x**2 - y) / (5 * y)
return f1
with pytest.raises(RuntimeError):
output_value, jacobian = ad.evaluate(func=func3, vals=2)
def test_incorrect_input_val4():
def func5(x, y):
f1 = ad.exp(x)
f2 = ad.exp(x) + y**2
return [f1, f2]
# expecting two arguments, but passed in three values
with pytest.raises(RuntimeError):
output_value, jacobian = ad.evaluate(func=func5, vals=[1, 2, 3])
def test_incorrect_input_vals3():
def func4(x, y, z):
f1 = x**2 + 2 * y - 7 * z
f2 = 3 * x + z**2
f3 = 3 * y - 2 * z
return [f1, f2, f3]
with pytest.raises(RuntimeError):
output_value, jacobian = ad.evaluate(func=func4, vals=[2, 3])
def test_incorrect_input_vals1():
def func1(x):
f1 = (ad.sin(x))**2
return f1
with pytest.raises(RuntimeError):
# a list of two scalars passed in for 'vals' when expecting only one scalar
output_value, jacobian = ad.evaluate(func=func1,
vals=[np.pi / 4, np.pi / 2])
def test_multivariate_single_func():
def func3(x, y):
f1 = (3 * x**2 - y) / (5 * y)
return f1
output_value, jacobian = ad.evaluate(func=func3, vals=[2, 2])
assert output_value == 1
np.testing.assert_allclose(jacobian, np.array([1.2, -0.6]))
def test_univariate_single_func():
def func1(x):
f1 = (ad.sin(x))**2
return f1
output_value, jacobian = ad.evaluate(func=func1, vals=np.pi / 4)
assert np.isclose(output_value, (np.sin(np.pi / 4))**2)
np.testing.assert_allclose(jacobian, np.array([1]))
def test_incorrect_input_seed3():
def func3(x, y):
f1 = (3 * x**2 - y) / (5 * y)
return f1
with pytest.raises(AssertionError):
# raise error when seed is not given in correct format (expecting a list of two lists)
output_value, jacobian = ad.evaluate(func=func3,
vals=[2, 2],
seed=[[1, 0]])
def test_incorrect_input_seed1():
def func1(x):
f1 = (ad.sin(x))**2
return f1
with pytest.raises(ValueError):
# raise error when seed value for a univariate function is given as a list instead of a scalar
output_value, jacobian = ad.evaluate(func=func1,
vals=np.pi / 4,
seed=[1])
def test_univariate_vector_func():
def func2(x):
f1 = x**2 - 3 * x
f2 = 2**x
return [f1, f2]
output_value, jacobian = ad.evaluate(func=func2, vals=1)
np.testing.assert_allclose(output_value, np.array([-2, 2]))
np.testing.assert_allclose(jacobian, np.array([[-1], [np.log(4)]]))
def test_multivariate_vector_func1():
def func4(x, y, z):
f1 = x**2 + 2 * y - 7 * z
f2 = 3 * x + z**2
f3 = 3 * y - 2 * z
return [f1, f2, f3]
output_value, jacobian = ad.evaluate(func=func4, vals=[2, 3, 4])
np.testing.assert_allclose(output_value, np.array([-18, 22, 1]))
np.testing.assert_allclose(jacobian,
np.array([[4, 2, -7], [3, 0, 8], [0, 3, -2]]))
def test_correct_input_seed():
def func4(x, y, z):
f1 = x**2 + 2 * y - 7 * z
f2 = 3 * x + z**2
f3 = 3 * y - 2 * z
return [f1, f2, f3]
output_value, jacobian = ad.evaluate(func=func4,
vals=[2, 3, 4],
seed=[[1, 0, 0], [0, 1, 0], [0, 0,
1]])
np.testing.assert_allclose(output_value, np.array([-18, 22, 1]))
np.testing.assert_allclose(jacobian,
np.array([[4, 2, -7], [3, 0, 8], [0, 3, -2]]))
def test_multivariate_vector_func2():
def func5(x, y):
f1 = x * y + ad.cos(x)
f2 = x * y + ad.sin(y)
return [f1, f2]
output_value, jacobian = ad.evaluate(func=func5,
vals=[np.pi / 3, np.pi / 2])
np.testing.assert_allclose(
output_value,
np.array([(np.pi / 3) * (np.pi / 2) + 0.5,
(np.pi / 3) * (np.pi / 2) + 1]))
np.testing.assert_allclose(
jacobian,
np.array([[(np.pi / 2) - np.sin(np.pi / 3), np.pi / 3],
[np.pi / 2, (np.pi / 3) + np.cos(np.pi / 2)]]))