def test_custom_model_n_outputs():
"""
Test creating a custom_model which has more than one output, which
requires special handling.
Demonstrates issue #11791's ``n_outputs`` error has been solved
"""
@custom_model
def model(x, y, n_outputs=2):
return x+1, y+1
m = model()
assert not isinstance(m.n_outputs, Parameter)
assert isinstance(m.n_outputs, int)
assert m.n_outputs == 2
assert m.outputs == ('x0', 'x1')
assert (separability_matrix(m) == [[True, True],
[True, True]]).all()
@custom_model
def model(x, y, z, n_outputs=3):
return x+1, y+1, z+1
m = model()
assert not isinstance(m.n_outputs, Parameter)
assert isinstance(m.n_outputs, int)
assert m.n_outputs == 3
assert m.outputs == ('x0', 'x1', 'x2')
assert (separability_matrix(m) == [[True, True, True],
[True, True, True],
[True, True, True]]).all()
def test_coupled_sep(linear_time, vct_crval):
if not hasattr(Model, "_calculate_separability_matrix"):
pytest.skip()
tfrm = CoupledCompoundModel("&", vct_crval, linear_time, shared_inputs=1)
smatrix = separability_matrix(tfrm)
assert np.allclose(smatrix, np.array([[1, 1, 1], [1, 1, 1], [0, 0, 1]]))
def axis_correlation_matrix(self):
"""
Returns an (`~BaseLowLevelWCS.world_n_dim`,
`~BaseLowLevelWCS.pixel_n_dim`) matrix that indicates using booleans
whether a given world coordinate depends on a given pixel coordinate.
This defaults to a matrix where all elements are `True` in the absence of
any further information. For completely independent axes, the diagonal
would be `True` and all other entries `False`.
"""
return separable.separability_matrix(self.forward_transform)
def axis_correlation_matrix(self):
"""
Returns an (`~BaseLowLevelWCS.world_n_dim`,
`~BaseLowLevelWCS.pixel_n_dim`) matrix that indicates using booleans
whether a given world coordinate depends on a given pixel coordinate.
This defaults to a matrix where all elements are `True` in the absence of
any further information. For completely independent axes, the diagonal
would be `True` and all other entries `False`.
"""
return separable.separability_matrix(self.forward_transform)
def test_coupled_sep_2d(vct_2d_pc, linear_time):
if not hasattr(Model, "_calculate_separability_matrix"):
pytest.skip()
linear_spectral = m.Linear1D(slope=10 * u.nm / u.pix)
right = linear_time & linear_spectral
tfrm = CoupledCompoundModel("&", vct_2d_pc, right, shared_inputs=2)
smatrix = separability_matrix(tfrm)
assert np.allclose(
smatrix,
np.array([[1, 1, 1, 1], [1, 1, 1, 1], [0, 0, 1, 0], [0, 0, 0, 1]]))
def test_custom_model_separable():
@custom_model
def model_a(x):
return x
assert model_a().separable
@custom_model
def model_c(x, y):
return x + y
assert not model_c().separable
assert np.all(separability_matrix(model_c()) == [True, True])
def test_coupled_sep(linear_time):
if not hasattr(Model, "_calculate_separability_matrix"):
pytest.skip()
crval_table = ((0, 1), (2, 3), (4, 5)) * u.arcsec
vct = VaryingCelestialTransform(crpix=(5, 5) * u.pix,
cdelt=(1, 1) * u.arcsec / u.pix,
crval_table=crval_table,
pc_table=np.identity(2) * u.arcsec,
lon_pole=180 * u.deg)
tfrm = CoupledCompoundModel("&", vct, linear_time, shared_inputs=1)
smatrix = separability_matrix(tfrm)
assert np.allclose(
smatrix,
np.array([[True, True, True], [True, True, True], [False, False,
True]]))
def test_custom_separability_matrix():
original = separability_matrix(ModelDefault(slope=1, intercept=2))
assert original.all()
custom = separability_matrix(ModelCustom(slope=1, intercept=2))
assert not custom.any()
def test_separable(compound_model, result):
assert_allclose(is_separable(compound_model), result[0])
assert_allclose(separability_matrix(compound_model), result[1])
def test_separable(compound_model, result):
assert_allclose(is_separable(compound_model), result[0])
assert_allclose(separability_matrix(compound_model), result[1])
