def test_simple_vector_subset(simple_modelspec_with_phi):
packer, unpacker, _ = simple_vector(simple_modelspec_with_phi)
phi = packer(simple_modelspec_with_phi)
assert len(phi) == 38
subset = [0, 2]
packer, unpacker, _ = simple_vector(simple_modelspec_with_phi, subset)
phi = packer(simple_modelspec_with_phi)
assert len(phi) == 8
vector = np.arange(8)
new_modelspec = unpacker(vector)
assert new_modelspec[0]['phi']['mean'].tolist() == [0, 1]
assert new_modelspec[0]['phi']['sd'].tolist() == [2, 3]
dexp_expected = {
'amplitude': 4,
'base': 5,
'kappa': 6,
'shift': 7
}
assert new_modelspec[2]['phi'] == dexp_expected
assert new_modelspec[1]['phi']['coefficients'].tolist() == \
new_modelspec[1]['phi']['coefficients'].tolist()
# Check that identity is the same
assert id(new_modelspec[1]['phi']['coefficients']) == \
id(new_modelspec[1]['phi']['coefficients'])
assert id(new_modelspec[0]['phi']) == id(new_modelspec[0]['phi'])
assert id(simple_modelspec_with_phi) == id(new_modelspec)
def test_simple_vector_bounds_subset(simple_modelspec_with_phi):
packer, unpacker, bounds = simple_vector(simple_modelspec_with_phi)
phi = packer(simple_modelspec_with_phi)
lb, ub = bounds(simple_modelspec_with_phi)
assert np.all(np.equal(lb, -np.inf))
assert np.all(np.equal(ub, np.inf))
assert len(lb) == 38
assert len(ub) == len(lb)
assert len(ub) == len(phi)
subset = [0, 2]
packer, unpacker, bounds = simple_vector(simple_modelspec_with_phi, subset)
phi = packer(simple_modelspec_with_phi)
lb, ub = bounds(simple_modelspec_with_phi)
phi = np.array(phi)
lb = np.array(lb)
ub = np.array(ub)
assert np.all(np.equal(lb, -np.inf))
assert np.all(np.equal(ub, np.inf))
assert len(lb) == 8
assert len(ub) == len(lb)
assert len(ub) == len(phi)
simple_modelspec_with_phi[0].setdefault('bounds', {})['mean'] = (None, 1)
lb, ub = bounds(simple_modelspec_with_phi)
assert np.all(np.equal(lb, -np.inf))
assert np.all(np.equal(ub[:2], 1))
assert np.all(np.equal(ub[2:], np.inf))
simple_modelspec_with_phi[0].setdefault('bounds', {})['mean'] = (-1, [2, 3])
lb, ub = bounds(simple_modelspec_with_phi)
assert np.all(np.equal(lb[:2], -1))
assert np.all(np.equal(lb[2:], -np.inf))
assert np.all(np.equal(ub[:2], [2, 3]))
assert np.all(np.equal(ub[2:], np.inf))
def test_partial_definition(bounds_modelspec):
packer, unpacker, bounds = simple_vector(bounds_modelspec)
bounds_modelspec[0]['bounds'] = {
'mean': (None, 10)
}
# Don't need to assert anything here, just shouldn't get an error
# for leaving 'sd' bounds undefined.
x = bounds(bounds_modelspec)
def test_scalar_bounds(bounds_modelspec):
bounds_modelspec[0]['bounds'] = {
'mean': (None, [1.1, 1.2]),
'sd': ([0.0, 0.7], 6)
}
bounds_modelspec[1]['bounds'] = {
'kappa': (-1, None),
'base': (None, None),
'amplitude': (0, 7),
'shift': (None, 5)
}
packer, unpacker, bounds = simple_vector(bounds_modelspec)
lb, ub = bounds(bounds_modelspec)
lb_expected = [-np.inf, -np.inf, 0, 0.7, 0, -np.inf, -1, -np.inf]
ub_expected = [1.1, 1.2, 6, 6, 7, np.inf, np.inf, 5]
assert np.all(np.equal(lb, lb_expected))
assert np.all(np.equal(ub, ub_expected))
def test_benchmark_packer_unpacker(benchmark, simple_modelspec_with_phi):
packer, unpacker, _ = simple_vector(simple_modelspec_with_phi)
vector = np.arange(38)
benchmark(unpacker, vector)
