def test_errors(self):
mesh = emg3d.TensorMesh([[2, 2], [2, 2], [2, 2]], origin=(-1, -1, -1))
survey = emg3d.Survey(
sources=emg3d.TxElectricDipole((-1.5, 0, 0, 0, 0)),
receivers=emg3d.RxElectricPoint((1.5, 0, 0, 0, 0)),
frequencies=1.0,
relative_error=0.01,
)
sim_inp = {
'survey': survey,
'gridding': 'same',
'receiver_interpolation': 'linear'
}
# Anisotropic models.
simulation = simulations.Simulation(model=emg3d.Model(mesh, 1, 2, 3),
**sim_inp)
with pytest.raises(NotImplementedError, match='for isotropic models'):
simulation.gradient
# Model with electric permittivity.
simulation = simulations.Simulation(model=emg3d.Model(mesh,
epsilon_r=3),
**sim_inp)
with pytest.raises(NotImplementedError, match='for el. permittivity'):
simulation.gradient
# Model with magnetic permeability.
simulation = simulations.Simulation(model=emg3d.Model(
mesh, mu_r=np.ones(mesh.shape_cells) * np.pi),
**sim_inp)
with pytest.raises(NotImplementedError, match='for magn. permeabili'):
simulation.gradient
def test_dicts_provided(self):
grids = self.simulation._dict_grid.copy()
# dict
sim1 = simulations.Simulation(self.survey,
self.model,
gridding='dict',
gridding_opts=grids,
name='Test2')
m1 = sim1.get_model('TxEW-3', 1.0)
g1 = sim1.get_grid('TxEW-3', 1.0)
# provide
sim2 = simulations.Simulation(self.survey,
self.model,
name='Test2',
gridding='input',
gridding_opts=grids['TxEW-3']['f-1'])
m2 = sim2.get_model('TxEW-3', 1.0)
g2 = sim2.get_grid('TxEW-3', 1.0)
assert m1 == m2
assert g1 == g2
# to/from_dict
sim1copy = sim1.copy()
sim2copy = sim2.copy()
m1c = sim1copy.get_model('TxEW-3', 1.0)
g1c = sim1copy.get_grid('TxEW-3', 1.0)
m2c = sim2copy.get_model('TxEW-3', 1.0)
g2c = sim2copy.get_grid('TxEW-3', 1.0)
assert m1 == m1c
assert g1 == g1c
assert m2 == m2c
assert g2 == g2c
def test_misfit(self, tmpdir):
sim = simulations.Simulation(model=self.model_init,
file_dir=str(tmpdir),
**self.sim_inp)
m0 = 2 * sim.model.property_x
def func2(x):
sim.model.property_x[...] = m0
return sim.jvec(x)
def func1(x):
sim.model.property_x[...] = x.reshape(sim.model.grid.shape_cells)
sim.clean('computed')
# Quick test that clean() removes the files
assert len(os.listdir(tmpdir)) == 0
sim.compute()
return sim.data.synthetic.data, func2
assert discretize.tests.check_derivative(
func1,
m0,
plotIt=False,
num=3,
)
def test_as_vs_fd_gradient(self, capsys):
# Compute adjoint state misfit and gradient
sim = simulations.Simulation(model=self.model_init, **self.sim_inp)
data_misfit = sim.misfit
grad = sim.gradient
# For Debug / QC, needs discretize
# from matplotlib.colors import LogNorm, SymLogNorm
# mesh.plot_3d_slicer(
# grad.ravel('F'),
# pcolor_opts={
# 'cmap': 'RdBu_r',
# 'norm': SymLogNorm(linthresh=1e-2, base=10,
# vmin=-1e1, vmax=1e1)}
# )
# We test a random cell from the inline xz slice, where the grad > 3.
#
# The NRMSD is (should) be below 1 %. However, (a) close to the
# boundary, (b) in regions where the gradient is almost zero, and (c)
# in regions where the gradient changes sign the NRMSD can become
# large. This is mainly due to numerics, our coarse mesh, and the
# reduced tolerance (which we reduced for speed).
iy = 32
indices = np.argwhere(abs(grad[:, iy, :]) > 3)
ix, iz = indices[np.random.randint(indices.shape[0])]
nrmsd = alternatives.fd_vs_as_gradient((ix, iy, iz), self.model_init,
grad, data_misfit, self.sim_inp)
assert nrmsd < 0.3
def test_rel_abs_rec(self):
# Sources
sources = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.TxElectricDipole, ([0, 100, 200], 0, 0, 0, 0))
# Abs and rel Receivers
a_e_rec = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint, (1000 + np.arange(3) * 100, 0, -100, 0, 0))
r_e_rec = emg3d.surveys.txrx_coordinates_to_dict(emg3d.RxElectricPoint,
(1000, 0, -100, 0, 0),
relative=True)
a_h_rec = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxMagneticPoint, (1000 + np.arange(3) * 100, 0, -100, 0, 0))
r_h_rec = emg3d.surveys.txrx_coordinates_to_dict(emg3d.RxMagneticPoint,
(1000, 0, -100, 0, 0),
relative=True)
receivers = emg3d.surveys.txrx_lists_to_dict(
[a_e_rec, r_e_rec, a_h_rec, r_h_rec])
# Frequencies
frequencies = (1.0)
survey = emg3d.Survey(sources,
receivers,
frequencies,
name='TestSurv',
noise_floor=1e-15,
relative_error=0.05)
# Create a simple grid and model
grid = emg3d.TensorMesh(
[np.ones(32) * 250,
np.ones(16) * 500,
np.ones(16) * 500], np.array([-1250, -1250, -2250]))
model = emg3d.Model(grid, 1)
# Create a simulation, compute all fields.
simulation = simulations.Simulation(survey,
model,
name='TestSim',
max_workers=1,
solver_opts={
'maxit': 1,
'verb': 0,
'plain': True
},
gridding='same')
simulation.compute()
# Relative receivers must be same as corresponding absolute receivers
assert_allclose(
[simulation.data.synthetic[i, i, 0].data for i in range(3)],
simulation.data.synthetic[:, 3, 0].data)
assert_allclose(
[simulation.data.synthetic[i, i + 4, 0].data for i in range(3)],
simulation.data.synthetic[:, 7, 0].data)
def test_grid_provided(self):
# Check bad grid
hx = np.ones(17) * 20
grid = emg3d.TensorMesh([hx, hx, hx], (0, 0, 0))
with pytest.warns(UserWarning, match='optimal for MG solver. Good n'):
simulations.Simulation(self.survey,
self.model,
gridding='input',
gridding_opts=grid)
def test_adjoint(self, tmpdir):
sim = simulations.Simulation(model=self.model_init,
file_dir=str(tmpdir),
**self.sim_inp)
v = np.random.rand(self.mesh.n_cells).reshape(self.mesh.shape_cells)
w = np.random.rand(self.survey.size).reshape(self.survey.shape)
wtJv = np.vdot(w, sim.jvec(v)).real
vtJtw = np.vdot(v, sim.jtvec(w).ravel('F'))
assert abs(wtJv - vtJtw) < 1e-10
def test_misfit():
data = 1
syn = 5
rel_err = 0.05
sources = emg3d.TxElectricDipole((0, 0, 0, 0, 0))
receivers = emg3d.RxElectricPoint((5, 0, 0, 0, 0))
survey = emg3d.Survey(
sources=sources,
receivers=receivers,
frequencies=100,
data=np.zeros((1, 1, 1)) + data,
relative_error=0.05,
)
grid = emg3d.TensorMesh([np.ones(10) * 2, [2, 2], [2, 2]], (-10, -2, -2))
model = emg3d.Model(grid, 1)
simulation = simulations.Simulation(survey=survey, model=model)
field = emg3d.Field(grid, dtype=np.float64)
field.field += syn
simulation._dict_efield['TxED-1']['f-1'] = field
simulation.data['synthetic'] = simulation.data['observed'] * 0 + syn
misfit = 0.5 * ((syn - data) / (rel_err * data))**2
def dummy():
pass
simulation.compute = dummy # => switch of compute()
assert_allclose(simulation.misfit, misfit)
# Missing noise_floor / std.
survey = emg3d.Survey(sources, receivers, 100)
simulation = simulations.Simulation(survey=survey, model=model)
with pytest.raises(ValueError, match="Either `noise_floor` or"):
simulation.misfit
def test_errors(self):
with pytest.raises(TypeError, match='Unexpected '):
simulations.Simulation(self.survey,
self.model,
unknown=True,
name='Test2')
# gridding='same' with gridding_opts.
with pytest.raises(TypeError, match="`gridding_opts` is not permitt"):
simulations.Simulation(self.survey,
self.model,
name='Test',
gridding='same',
gridding_opts={'bummer': True})
# expand without seasurface
with pytest.raises(KeyError, match="is required if"):
simulations.Simulation(self.survey,
self.model,
name='Test',
gridding='single',
gridding_opts={'expand': [1, 2]})
def test_input_gradient(self):
# Create another mesh, so there will be a difference.
newgrid = emg3d.TensorMesh(
[np.ones(16) * 500,
np.ones(8) * 1000,
np.ones(8) * 1000], np.array([-1250, -1250, -2250]))
simulation = simulations.Simulation(self.survey,
self.model,
max_workers=1,
solver_opts={
'maxit': 1,
'verb': 0,
'sslsolver': False,
'linerelaxation': False,
'semicoarsening': False
},
gridding='input',
gridding_opts=newgrid,
name='TestX')
with pytest.warns(UserWarning, match='Receiver responses were obtain'):
grad = simulation.gradient
with pytest.warns(UserWarning, match='Receiver responses were obtain'):
vec = simulation.data.residual.data.copy()
vec *= simulation.data.weights.data
jtvec = simulation.jtvec(vec)
assert_allclose(grad, jtvec)
jvec = simulation.jvec(np.ones(newgrid.n_cells))
assert jvec.shape == simulation.data.observed.data.shape
# Ensure the gradient has the shape of the model, not of the input.
assert grad.shape == self.model.shape
sim2 = simulation.to_dict(what='all', copy=True)
sim3 = simulation.to_dict(what='plain', copy=True)
assert 'residual' in sim2['survey']['data'].keys()
assert 'residual' not in sim3['survey']['data'].keys()
simulation.clean('all') # Should remove 'residual', 'bfield-dicts'
sim5 = simulation.to_dict('all')
assert 'residual' not in sim5['survey']['data'].keys()
assert '_dict_bfield' not in sim5.keys()
class TestGradient:
if xarray is not None:
# Create a simple mesh.
hx = np.ones(64) * 100
mesh = emg3d.TensorMesh([hx, hx, hx], origin=[0, 0, 0])
# Define a simple survey, including 1 el. & 1 magn. receiver
survey = emg3d.Survey(
sources=emg3d.TxElectricDipole((1680, 3220, 3210, 20, 5)),
receivers=[
emg3d.RxElectricPoint((4751, 3280, 3220, 33, 14)),
emg3d.RxMagneticPoint((4758, 3230, 3250, 15, 70)),
],
frequencies=1.0,
relative_error=0.01,
)
# Background Model
con_init = np.ones(mesh.shape_cells)
# Target Model 1: One Block
con_true = np.ones(mesh.shape_cells)
con_true[27:37, 27:37, 15:25] = 0.001
model_init = emg3d.Model(mesh, con_init, mapping='Conductivity')
model_true = emg3d.Model(mesh, con_true, mapping='Conductivity')
# mesh.plot_3d_slicer(con_true) # For debug / QC, needs discretize
sim_inp = {
'survey': survey,
'solver_opts': {
'plain': True,
'tol': 5e-5
}, # Red. tol 4 speed
'max_workers': 1,
'gridding': 'same',
'verb': 0,
'receiver_interpolation': 'linear',
}
# Compute data (pre-computed and passed to Survey above)
sim_data = simulations.Simulation(model=model_true, **sim_inp)
sim_data.compute(observed=True)
def test_errors(self):
mesh = emg3d.TensorMesh([[2, 2], [2, 2], [2, 2]], origin=(-1, -1, -1))
survey = emg3d.Survey(
sources=emg3d.TxElectricDipole((-1.5, 0, 0, 0, 0)),
receivers=emg3d.RxElectricPoint((1.5, 0, 0, 0, 0)),
frequencies=1.0,
relative_error=0.01,
)
sim_inp = {
'survey': survey,
'gridding': 'same',
'receiver_interpolation': 'linear'
}
# Anisotropic models.
simulation = simulations.Simulation(model=emg3d.Model(mesh, 1, 2, 3),
**sim_inp)
with pytest.raises(NotImplementedError, match='for isotropic models'):
simulation.gradient
# Model with electric permittivity.
simulation = simulations.Simulation(model=emg3d.Model(mesh,
epsilon_r=3),
**sim_inp)
with pytest.raises(NotImplementedError, match='for el. permittivity'):
simulation.gradient
# Model with magnetic permeability.
simulation = simulations.Simulation(model=emg3d.Model(
mesh, mu_r=np.ones(mesh.shape_cells) * np.pi),
**sim_inp)
with pytest.raises(NotImplementedError, match='for magn. permeabili'):
simulation.gradient
def test_as_vs_fd_gradient(self, capsys):
# Compute adjoint state misfit and gradient
sim = simulations.Simulation(model=self.model_init, **self.sim_inp)
data_misfit = sim.misfit
grad = sim.gradient
# For Debug / QC, needs discretize
# from matplotlib.colors import LogNorm, SymLogNorm
# mesh.plot_3d_slicer(
# grad.ravel('F'),
# pcolor_opts={
# 'cmap': 'RdBu_r',
# 'norm': SymLogNorm(linthresh=1e-2, base=10,
# vmin=-1e1, vmax=1e1)}
# )
# We test a random cell from the inline xz slice, where the grad > 3.
#
# The NRMSD is (should) be below 1 %. However, (a) close to the
# boundary, (b) in regions where the gradient is almost zero, and (c)
# in regions where the gradient changes sign the NRMSD can become
# large. This is mainly due to numerics, our coarse mesh, and the
# reduced tolerance (which we reduced for speed).
iy = 32
indices = np.argwhere(abs(grad[:, iy, :]) > 3)
ix, iz = indices[np.random.randint(indices.shape[0])]
nrmsd = alternatives.fd_vs_as_gradient((ix, iy, iz), self.model_init,
grad, data_misfit, self.sim_inp)
assert nrmsd < 0.3
@pytest.mark.skipif(discretize is None, reason="discretize not installed.")
@pytest.mark.skipif(h5py is None, reason="h5py not installed.")
def test_adjoint(self, tmpdir):
sim = simulations.Simulation(model=self.model_init,
file_dir=str(tmpdir),
**self.sim_inp)
v = np.random.rand(self.mesh.n_cells).reshape(self.mesh.shape_cells)
w = np.random.rand(self.survey.size).reshape(self.survey.shape)
wtJv = np.vdot(w, sim.jvec(v)).real
vtJtw = np.vdot(v, sim.jtvec(w).ravel('F'))
assert abs(wtJv - vtJtw) < 1e-10
@pytest.mark.skipif(discretize is None, reason="discretize not installed.")
@pytest.mark.skipif(h5py is None, reason="h5py not installed.")
def test_misfit(self, tmpdir):
sim = simulations.Simulation(model=self.model_init,
file_dir=str(tmpdir),
**self.sim_inp)
m0 = 2 * sim.model.property_x
def func2(x):
sim.model.property_x[...] = m0
return sim.jvec(x)
def func1(x):
sim.model.property_x[...] = x.reshape(sim.model.grid.shape_cells)
sim.clean('computed')
# Quick test that clean() removes the files
assert len(os.listdir(tmpdir)) == 0
sim.compute()
return sim.data.synthetic.data, func2
assert discretize.tests.check_derivative(
func1,
m0,
plotIt=False,
num=3,
)
def test_print_solver(self, capsys):
grid = emg3d.TensorMesh(h=[[(25, 10, -1.04), (25, 28), (25, 10, 1.04)],
[(50, 8, -1.03), (50, 16), (50, 8, 1.03)],
[(30, 8, -1.05), (30, 16), (30, 8, 1.05)]],
origin='CCC')
model = emg3d.Model(grid,
property_x=1.5,
property_y=1.8,
property_z=3.3,
mapping='Resistivity')
sources = emg3d.TxElectricDipole((0, 0, 0, 0, 0))
receivers = [
emg3d.RxElectricPoint((x, 0, 0, 0, 0)) for x in [-10000, 10000]
]
survey = emg3d.Survey(
name='Test',
sources=sources,
receivers=receivers,
frequencies=1.0,
noise_floor=1e-15,
relative_error=0.05,
)
inp = {
'name': 'Test',
'survey': survey,
'model': model,
'gridding': 'same'
}
sol = {
'sslsolver': False,
'semicoarsening': False,
'linerelaxation': False,
'maxit': 1
}
# Errors but verb=-1.
_, _ = capsys.readouterr() # empty
simulation = simulations.Simulation(verb=-1, **inp, solver_opts=sol)
simulation.compute()
out, _ = capsys.readouterr()
assert out == ""
# Errors with verb=0.
out = simulation.print_solver_info('efield', verb=0, return_info=True)
assert "= Source TxED-1; Frequency 1.0 Hz = MAX. ITERATION REAC" in out
# Errors with verb=1.
_, _ = capsys.readouterr() # empty
simulation.print_solver_info('efield', verb=1)
out, _ = capsys.readouterr()
assert "= Source TxED-1; Frequency 1.0 Hz = 2.6e-02; 1; 0:00:" in out
# No errors; solver-verb 3.
simulation = simulations.Simulation(verb=1,
**inp,
solver_opts={'verb': 3})
simulation.compute()
out, _ = capsys.readouterr()
assert 'MG-cycle' in out
assert 'CONVERGED' in out
# No errors; solver-verb 0.
simulation = simulations.Simulation(verb=1,
**inp,
solver_opts={'verb': 0})
simulation.compute()
out, _ = capsys.readouterr()
assert "= Source TxED-1; Frequency 1.0 Hz = CONVERGED" in out
# Two sources, only compute 1, assure printing works.
sources = [emg3d.TxElectricDipole((x, 0, 0, 0, 0)) for x in [0, 10]]
survey = emg3d.Survey(
name='Test',
sources=sources,
receivers=receivers,
frequencies=1.0,
noise_floor=1e-15,
relative_error=0.05,
)
inp = {
'name': 'Test',
'survey': survey,
'model': model,
'gridding': 'same'
}
simulation = simulations.Simulation(**inp, solver_opts={'verb': 0})
_ = simulation.get_efield('TxED-2', 'f-1')
simulation.print_solver_info(verb=1)
out, _ = capsys.readouterr()
assert "= Source TxED-2; Frequency 1.0 Hz = CONVERGED" in out
def test_simulation_automatic(self, capsys):
# Create a simple survey
sources = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.TxElectricDipole, (0, [1000, 3000, 5000], -950, 0, 0))
receivers = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint,
([-3000, 0, 3000], [0, 3000, 6000], -1000, 0, 0))
frequencies = (0.1, 1.0, 10.0)
survey = emg3d.Survey(sources,
receivers,
frequencies,
name='Test',
noise_floor=1e-15,
relative_error=0.05)
# Create a simple grid and model
grid = emg3d.TensorMesh(
[np.ones(32) * 250,
np.ones(16) * 500,
np.ones(4) * 500], np.array([-1250, -1250, -2250]))
model = emg3d.Model(grid, 1)
# Create a simulation, compute all fields.
inp = {
'survey': survey,
'model': model,
'gridding_opts': {
'expand': [1, 0.5],
'seasurface': 0,
'verb': 1
}
}
b_sim = simulations.Simulation(name='both', gridding='both', **inp)
f_sim = simulations.Simulation(name='freq',
gridding='frequency',
**inp)
t_sim = simulations.Simulation(name='src', gridding='source', **inp)
s_sim = simulations.Simulation(name='single', gridding='single', **inp)
# Quick repr test.
assert " 24 x 24 (13,824) - 160 x 160 x 96 (2,457," in b_sim.__repr__()
assert " 24 x 24 (13,824) - 160 x 160 x 96 (2,457," in f_sim.__repr__()
assert "Source-dependent grids; 64 x 64 x 40 (163," in t_sim.__repr__()
assert "ources and frequencies; 64 x 64 x 40 (163," in s_sim.__repr__()
# Quick print_grid test:
_, _ = capsys.readouterr() # empty
b_sim.print_grid_info()
out, _ = capsys.readouterr()
assert "= Source: TxED-1; Frequency: 10.0 Hz =" in out
assert "== GRIDDING IN X ==" in out
assert b_sim.get_grid('TxED-1', 1.0).__repr__() in out
_, _ = capsys.readouterr() # empty
out = f_sim.print_grid_info(return_info=True)
out2, _ = capsys.readouterr()
assert out2 == ""
assert "= Source: all" in out
assert "== GRIDDING IN X ==" in out
assert f_sim.get_grid('TxED-3', 1.0).__repr__() in out
t_sim.print_grid_info()
out, _ = capsys.readouterr()
assert "; Frequency: all" in out
assert "== GRIDDING IN X ==" in out
assert t_sim.get_grid('TxED-1', 10.0).__repr__() in out
_, _ = capsys.readouterr() # empty
out = s_sim.print_grid_info(return_info=True)
out2, _ = capsys.readouterr()
assert out2 == ""
assert "= Source: all; Frequency: all =" in out
assert "== GRIDDING IN X ==" in out
assert s_sim.get_grid('TxED-3', 0.1).__repr__() in out
assert s_sim.print_grid_info(verb=-1) is None
# Grids: Middle source / middle frequency should be the same in all.
assert f_sim.get_grid('TxED-2', 1.0) == t_sim.get_grid('TxED-2', 1.0)
assert f_sim.get_grid('TxED-2', 1.0) == s_sim.get_grid('TxED-2', 1.0)
assert f_sim.get_grid('TxED-2', 1.0) == b_sim.get_grid('TxED-2', 1.0)
assert f_sim.get_model('TxED-2', 1.0) == t_sim.get_model('TxED-2', 1.0)
assert f_sim.get_model('TxED-2', 1.0) == s_sim.get_model('TxED-2', 1.0)
assert f_sim.get_model('TxED-2', 1.0) == b_sim.get_model('TxED-2', 1.0)
# Copy:
f_sim_copy = f_sim.copy()
assert (f_sim.get_grid('TxED-1',
1.0) == f_sim_copy.get_grid('TxED-1', 1.0))
assert 'expand' not in f_sim.gridding_opts.keys()
assert 'expand' not in f_sim_copy.gridding_opts.keys()
s_sim_copy = s_sim.copy()
assert (s_sim.get_grid('TxED-1',
1.0) == s_sim_copy.get_grid('TxED-1', 1.0))
assert 'expand' not in s_sim.gridding_opts.keys()
assert 'expand' not in s_sim_copy.gridding_opts.keys()
t_sim_copy = t_sim.copy()
assert (t_sim.get_grid('TxED-1',
1.0) == t_sim_copy.get_grid('TxED-1', 1.0))
assert 'expand' not in t_sim.gridding_opts.keys()
assert 'expand' not in t_sim_copy.gridding_opts.keys()
class TestSimulation():
if xarray is not None:
# Create a simple survey
# Sources: 1 Electric Dipole, 1 Magnetic Dipole, 1 Electric Wire.
s1 = emg3d.TxElectricDipole((0, 1000, -950, 0, 0))
s2 = emg3d.TxMagneticDipole((0, 3000, -950, 90, 0))
s3 = emg3d.TxElectricWire(([0, 4900, -950], [-20, 5000,
-950], [20, 5100, -950]))
sources = emg3d.surveys.txrx_lists_to_dict([s1, s2, s3])
# Receivers: 1 Electric Point, 1 Magnetic Point
e_rec = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint, (np.arange(6) * 1000, 0, -1000, 0, 0))
m_rec = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxMagneticPoint, (np.arange(6) * 1000, 0, -1000, 90, 0))
receivers = emg3d.surveys.txrx_lists_to_dict([e_rec, m_rec])
# Frequencies
frequencies = (1.0, 2.0)
survey = emg3d.Survey(sources,
receivers,
frequencies,
name='TestSurv',
noise_floor=1e-15,
relative_error=0.05)
# Create a simple grid and model
grid = emg3d.TensorMesh(
[np.ones(32) * 250,
np.ones(16) * 500,
np.ones(16) * 500], np.array([-1250, -1250, -2250]))
model = emg3d.Model(grid, 1)
# Create a simulation, compute all fields.
simulation = simulations.Simulation(survey,
model,
name='TestSim',
max_workers=1,
solver_opts={
'maxit': 1,
'verb': 0,
'sslsolver': False,
'linerelaxation': False,
'semicoarsening': False
},
gridding='same')
# Do first one single and then all together.
simulation.get_efield('TxED-1', 'f-1')
simulation.compute(observed=True, min_offset=1100)
def test_same(self):
assert self.simulation.get_model('TxMD-2', 'f-1') == self.model
assert self.simulation.get_grid('TxEW-3', 1.0) == self.grid
def test_fields(self, capsys):
# Check sfield
sfield = emg3d.get_source_field(self.grid,
self.survey.sources['TxEW-3'],
frequency=1.0)
# Check efield
efield, info = emg3d.solve(self.model, sfield,
**self.simulation.solver_opts)
assert self.simulation.get_efield('TxEW-3', 'f-1') == efield
# See a single one
self.simulation._dict_efield['TxEW-3'][1.0] = None
_, _ = capsys.readouterr()
self.simulation.get_efield('TxEW-3', 1.0)
info = self.simulation.get_efield_info('TxEW-3', 1.0)
assert 'MAX. ITERATION REACHED, NOT CONVERGED' in info['exit_message']
# Check hfield
hfield = emg3d.get_magnetic_field(self.model, efield)
assert self.simulation.get_hfield('TxEW-3', 1.0) == hfield
s_hfield = self.simulation.get_hfield('TxEW-3', 1.0)
assert s_hfield == hfield
assert_allclose(
self.simulation._dict_efield_info['TxEW-3']['f-1']['abs_error'],
info['abs_error'])
assert_allclose(
self.simulation._dict_efield_info['TxEW-3']['f-1']['rel_error'],
info['rel_error'])
exit = self.simulation._dict_efield_info['TxEW-3']['f-1']['exit']
assert exit == info['exit'] == 1
# First hfield, ensure efield/hfield get computed.
sim = self.simulation.copy(what='all')
sim._dict_efield['TxEW-3']['f-1'] = None
sim.get_hfield('TxEW-3', 'f-1')
assert sim._dict_efield['TxEW-3']['f-1'] is not None
def test_responses(self):
# Check min_offset were switched-off
assert_allclose(self.simulation.data.observed[0, 0, 0].data, np.nan)
assert_allclose(self.simulation.data.observed[0, 6, 0].data, np.nan)
# Get efield responses
e_resp = self.simulation.get_efield('TxMD-2', 1.0).get_receiver(
self.survey.receivers.values())
assert_allclose(self.simulation.data.synthetic[1, :6, 0].data,
e_resp[:6],
atol=1e-16)
# Get hfield responses
m_resp = self.simulation.get_hfield('TxMD-2', 1.0).get_receiver(
self.survey.receivers.values())
assert_allclose(self.simulation.data.synthetic[1, 6:, 0].data,
m_resp[6:],
atol=1e-16)
def test_errors(self):
with pytest.raises(TypeError, match='Unexpected '):
simulations.Simulation(self.survey,
self.model,
unknown=True,
name='Test2')
# gridding='same' with gridding_opts.
with pytest.raises(TypeError, match="`gridding_opts` is not permitt"):
simulations.Simulation(self.survey,
self.model,
name='Test',
gridding='same',
gridding_opts={'bummer': True})
# expand without seasurface
with pytest.raises(KeyError, match="is required if"):
simulations.Simulation(self.survey,
self.model,
name='Test',
gridding='single',
gridding_opts={'expand': [1, 2]})
def test_reprs(self):
test = self.simulation.__repr__()
assert "Simulation «TestSim»" in test
assert "Survey «TestSurv»: 3 sources; 12 receivers; 2 frequenc" in test
assert "Model: resistivity; isotropic; 32 x 16 x 16 (8,192)" in test
assert "Gridding: Same grid as for model" in test
test = self.simulation._repr_html_()
assert "Simulation «TestSim»" in test
assert "Survey «TestSurv»: 3 sources; 12 receivers; 2" in test
assert "Model: resistivity; isotropic; 32 x 16 x 16 (8,192)" in test
assert "Gridding: Same grid as for model" in test
def test_copy(self, tmpdir):
with pytest.raises(TypeError, match="Unrecognized `what`: nothing"):
self.simulation.copy('nothing')
sim2 = self.simulation.copy()
assert self.simulation.name == sim2.name
assert self.simulation.survey.sources == sim2.survey.sources
assert_allclose(
self.simulation.get_efield('TxED-1', 1.0).field,
sim2.get_efield('TxED-1', 1.0).field)
# Also check to_file()/from_file().
sim_dict = self.simulation.to_dict('all')
sim2 = simulations.Simulation.from_dict(sim_dict.copy())
assert self.simulation.name == sim2.name
assert self.simulation.survey.name == sim2.survey.name
assert self.simulation.max_workers == sim2.max_workers
assert self.simulation.gridding == sim2.gridding
assert self.simulation.model == sim2.model
del sim_dict['survey']
with pytest.raises(KeyError, match="'survey'"):
simulations.Simulation.from_dict(sim_dict)
# Also check to_file()/from_file().
self.simulation.to_file(tmpdir + '/test.npz', what='all')
sim2 = simulations.Simulation.from_file(tmpdir + '/test.npz')
assert self.simulation.name == sim2.name
assert self.simulation.survey.name == sim2.survey.name
assert self.simulation.max_workers == sim2.max_workers
assert self.simulation.gridding == sim2.gridding
assert self.simulation.model == sim2.model
sim9 = simulations.Simulation.from_file(tmpdir + '/test.npz', verb=-1)
assert sim2.name == sim9[0].name
assert 'Data loaded from' in sim9[1]
# Clean and ensure it is empty
sim3 = self.simulation.copy()
sim3.clean('all')
assert sim3._dict_efield['TxMD-2']['f-1'] is None
assert sim3._dict_efield_info['TxMD-2']['f-1'] is None
with pytest.raises(TypeError, match="Unrecognized `what`: nothing"):
sim3.clean('nothing')
def test_dicts_provided(self):
grids = self.simulation._dict_grid.copy()
# dict
sim1 = simulations.Simulation(self.survey,
self.model,
gridding='dict',
gridding_opts=grids,
name='Test2')
m1 = sim1.get_model('TxEW-3', 1.0)
g1 = sim1.get_grid('TxEW-3', 1.0)
# provide
sim2 = simulations.Simulation(self.survey,
self.model,
name='Test2',
gridding='input',
gridding_opts=grids['TxEW-3']['f-1'])
m2 = sim2.get_model('TxEW-3', 1.0)
g2 = sim2.get_grid('TxEW-3', 1.0)
assert m1 == m2
assert g1 == g2
# to/from_dict
sim1copy = sim1.copy()
sim2copy = sim2.copy()
m1c = sim1copy.get_model('TxEW-3', 1.0)
g1c = sim1copy.get_grid('TxEW-3', 1.0)
m2c = sim2copy.get_model('TxEW-3', 1.0)
g2c = sim2copy.get_grid('TxEW-3', 1.0)
assert m1 == m1c
assert g1 == g1c
assert m2 == m2c
assert g2 == g2c
def test_grid_provided(self):
# Check bad grid
hx = np.ones(17) * 20
grid = emg3d.TensorMesh([hx, hx, hx], (0, 0, 0))
with pytest.warns(UserWarning, match='optimal for MG solver. Good n'):
simulations.Simulation(self.survey,
self.model,
gridding='input',
gridding_opts=grid)
def test_synthetic(self):
sim = self.simulation.copy()
sim._dict_efield = sim._dict_initiate # Reset
sim.compute(observed=True, add_noise=False)
assert_allclose(sim.data.synthetic, sim.data.observed)
assert sim.survey.size == sim.data.observed.size
def test_input_gradient(self):
# Create another mesh, so there will be a difference.
newgrid = emg3d.TensorMesh(
[np.ones(16) * 500,
np.ones(8) * 1000,
np.ones(8) * 1000], np.array([-1250, -1250, -2250]))
simulation = simulations.Simulation(self.survey,
self.model,
max_workers=1,
solver_opts={
'maxit': 1,
'verb': 0,
'sslsolver': False,
'linerelaxation': False,
'semicoarsening': False
},
gridding='input',
gridding_opts=newgrid,
name='TestX')
with pytest.warns(UserWarning, match='Receiver responses were obtain'):
grad = simulation.gradient
with pytest.warns(UserWarning, match='Receiver responses were obtain'):
vec = simulation.data.residual.data.copy()
vec *= simulation.data.weights.data
jtvec = simulation.jtvec(vec)
assert_allclose(grad, jtvec)
jvec = simulation.jvec(np.ones(newgrid.n_cells))
assert jvec.shape == simulation.data.observed.data.shape
# Ensure the gradient has the shape of the model, not of the input.
assert grad.shape == self.model.shape
sim2 = simulation.to_dict(what='all', copy=True)
sim3 = simulation.to_dict(what='plain', copy=True)
assert 'residual' in sim2['survey']['data'].keys()
assert 'residual' not in sim3['survey']['data'].keys()
simulation.clean('all') # Should remove 'residual', 'bfield-dicts'
sim5 = simulation.to_dict('all')
assert 'residual' not in sim5['survey']['data'].keys()
assert '_dict_bfield' not in sim5.keys()
def test_simulation_automatic(self, capsys):
# Create a simple survey
sources = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.TxElectricDipole, (0, [1000, 3000, 5000], -950, 0, 0))
receivers = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint,
([-3000, 0, 3000], [0, 3000, 6000], -1000, 0, 0))
frequencies = (0.1, 1.0, 10.0)
survey = emg3d.Survey(sources,
receivers,
frequencies,
name='Test',
noise_floor=1e-15,
relative_error=0.05)
# Create a simple grid and model
grid = emg3d.TensorMesh(
[np.ones(32) * 250,
np.ones(16) * 500,
np.ones(4) * 500], np.array([-1250, -1250, -2250]))
model = emg3d.Model(grid, 1)
# Create a simulation, compute all fields.
inp = {
'survey': survey,
'model': model,
'gridding_opts': {
'expand': [1, 0.5],
'seasurface': 0,
'verb': 1
}
}
b_sim = simulations.Simulation(name='both', gridding='both', **inp)
f_sim = simulations.Simulation(name='freq',
gridding='frequency',
**inp)
t_sim = simulations.Simulation(name='src', gridding='source', **inp)
s_sim = simulations.Simulation(name='single', gridding='single', **inp)
# Quick repr test.
assert " 24 x 24 (13,824) - 160 x 160 x 96 (2,457," in b_sim.__repr__()
assert " 24 x 24 (13,824) - 160 x 160 x 96 (2,457," in f_sim.__repr__()
assert "Source-dependent grids; 64 x 64 x 40 (163," in t_sim.__repr__()
assert "ources and frequencies; 64 x 64 x 40 (163," in s_sim.__repr__()
# Quick print_grid test:
_, _ = capsys.readouterr() # empty
b_sim.print_grid_info()
out, _ = capsys.readouterr()
assert "= Source: TxED-1; Frequency: 10.0 Hz =" in out
assert "== GRIDDING IN X ==" in out
assert b_sim.get_grid('TxED-1', 1.0).__repr__() in out
_, _ = capsys.readouterr() # empty
out = f_sim.print_grid_info(return_info=True)
out2, _ = capsys.readouterr()
assert out2 == ""
assert "= Source: all" in out
assert "== GRIDDING IN X ==" in out
assert f_sim.get_grid('TxED-3', 1.0).__repr__() in out
t_sim.print_grid_info()
out, _ = capsys.readouterr()
assert "; Frequency: all" in out
assert "== GRIDDING IN X ==" in out
assert t_sim.get_grid('TxED-1', 10.0).__repr__() in out
_, _ = capsys.readouterr() # empty
out = s_sim.print_grid_info(return_info=True)
out2, _ = capsys.readouterr()
assert out2 == ""
assert "= Source: all; Frequency: all =" in out
assert "== GRIDDING IN X ==" in out
assert s_sim.get_grid('TxED-3', 0.1).__repr__() in out
assert s_sim.print_grid_info(verb=-1) is None
# Grids: Middle source / middle frequency should be the same in all.
assert f_sim.get_grid('TxED-2', 1.0) == t_sim.get_grid('TxED-2', 1.0)
assert f_sim.get_grid('TxED-2', 1.0) == s_sim.get_grid('TxED-2', 1.0)
assert f_sim.get_grid('TxED-2', 1.0) == b_sim.get_grid('TxED-2', 1.0)
assert f_sim.get_model('TxED-2', 1.0) == t_sim.get_model('TxED-2', 1.0)
assert f_sim.get_model('TxED-2', 1.0) == s_sim.get_model('TxED-2', 1.0)
assert f_sim.get_model('TxED-2', 1.0) == b_sim.get_model('TxED-2', 1.0)
# Copy:
f_sim_copy = f_sim.copy()
assert (f_sim.get_grid('TxED-1',
1.0) == f_sim_copy.get_grid('TxED-1', 1.0))
assert 'expand' not in f_sim.gridding_opts.keys()
assert 'expand' not in f_sim_copy.gridding_opts.keys()
s_sim_copy = s_sim.copy()
assert (s_sim.get_grid('TxED-1',
1.0) == s_sim_copy.get_grid('TxED-1', 1.0))
assert 'expand' not in s_sim.gridding_opts.keys()
assert 'expand' not in s_sim_copy.gridding_opts.keys()
t_sim_copy = t_sim.copy()
assert (t_sim.get_grid('TxED-1',
1.0) == t_sim_copy.get_grid('TxED-1', 1.0))
assert 'expand' not in t_sim.gridding_opts.keys()
assert 'expand' not in t_sim_copy.gridding_opts.keys()
def test_print_solver(self, capsys):
grid = emg3d.TensorMesh(h=[[(25, 10, -1.04), (25, 28), (25, 10, 1.04)],
[(50, 8, -1.03), (50, 16), (50, 8, 1.03)],
[(30, 8, -1.05), (30, 16), (30, 8, 1.05)]],
origin='CCC')
model = emg3d.Model(grid,
property_x=1.5,
property_y=1.8,
property_z=3.3,
mapping='Resistivity')
sources = emg3d.TxElectricDipole((0, 0, 0, 0, 0))
receivers = [
emg3d.RxElectricPoint((x, 0, 0, 0, 0)) for x in [-10000, 10000]
]
survey = emg3d.Survey(
name='Test',
sources=sources,
receivers=receivers,
frequencies=1.0,
noise_floor=1e-15,
relative_error=0.05,
)
inp = {
'name': 'Test',
'survey': survey,
'model': model,
'gridding': 'same'
}
sol = {
'sslsolver': False,
'semicoarsening': False,
'linerelaxation': False,
'maxit': 1
}
# Errors but verb=-1.
_, _ = capsys.readouterr() # empty
simulation = simulations.Simulation(verb=-1, **inp, solver_opts=sol)
simulation.compute()
out, _ = capsys.readouterr()
assert out == ""
# Errors with verb=0.
out = simulation.print_solver_info('efield', verb=0, return_info=True)
assert "= Source TxED-1; Frequency 1.0 Hz = MAX. ITERATION REAC" in out
# Errors with verb=1.
_, _ = capsys.readouterr() # empty
simulation.print_solver_info('efield', verb=1)
out, _ = capsys.readouterr()
assert "= Source TxED-1; Frequency 1.0 Hz = 2.6e-02; 1; 0:00:" in out
# No errors; solver-verb 3.
simulation = simulations.Simulation(verb=1,
**inp,
solver_opts={'verb': 3})
simulation.compute()
out, _ = capsys.readouterr()
assert 'MG-cycle' in out
assert 'CONVERGED' in out
# No errors; solver-verb 0.
simulation = simulations.Simulation(verb=1,
**inp,
solver_opts={'verb': 0})
simulation.compute()
out, _ = capsys.readouterr()
assert "= Source TxED-1; Frequency 1.0 Hz = CONVERGED" in out
# Two sources, only compute 1, assure printing works.
sources = [emg3d.TxElectricDipole((x, 0, 0, 0, 0)) for x in [0, 10]]
survey = emg3d.Survey(
name='Test',
sources=sources,
receivers=receivers,
frequencies=1.0,
noise_floor=1e-15,
relative_error=0.05,
)
inp = {
'name': 'Test',
'survey': survey,
'model': model,
'gridding': 'same'
}
simulation = simulations.Simulation(**inp, solver_opts={'verb': 0})
_ = simulation.get_efield('TxED-2', 'f-1')
simulation.print_solver_info(verb=1)
out, _ = capsys.readouterr()
assert "= Source TxED-2; Frequency 1.0 Hz = CONVERGED" in out
def test_rel_abs_rec(self):
# Sources
sources = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.TxElectricDipole, ([0, 100, 200], 0, 0, 0, 0))
# Abs and rel Receivers
a_e_rec = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxElectricPoint, (1000 + np.arange(3) * 100, 0, -100, 0, 0))
r_e_rec = emg3d.surveys.txrx_coordinates_to_dict(emg3d.RxElectricPoint,
(1000, 0, -100, 0, 0),
relative=True)
a_h_rec = emg3d.surveys.txrx_coordinates_to_dict(
emg3d.RxMagneticPoint, (1000 + np.arange(3) * 100, 0, -100, 0, 0))
r_h_rec = emg3d.surveys.txrx_coordinates_to_dict(emg3d.RxMagneticPoint,
(1000, 0, -100, 0, 0),
relative=True)
receivers = emg3d.surveys.txrx_lists_to_dict(
[a_e_rec, r_e_rec, a_h_rec, r_h_rec])
# Frequencies
frequencies = (1.0)
survey = emg3d.Survey(sources,
receivers,
frequencies,
name='TestSurv',
noise_floor=1e-15,
relative_error=0.05)
# Create a simple grid and model
grid = emg3d.TensorMesh(
[np.ones(32) * 250,
np.ones(16) * 500,
np.ones(16) * 500], np.array([-1250, -1250, -2250]))
model = emg3d.Model(grid, 1)
# Create a simulation, compute all fields.
simulation = simulations.Simulation(survey,
model,
name='TestSim',
max_workers=1,
solver_opts={
'maxit': 1,
'verb': 0,
'plain': True
},
gridding='same')
simulation.compute()
# Relative receivers must be same as corresponding absolute receivers
assert_allclose(
[simulation.data.synthetic[i, i, 0].data for i in range(3)],
simulation.data.synthetic[:, 3, 0].data)
assert_allclose(
[simulation.data.synthetic[i, i + 4, 0].data for i in range(3)],
simulation.data.synthetic[:, 7, 0].data)
def simulation(args_dict):
"""Run `emg3d` invoked by CLI.
Run and log ``emg3d`` given the settings stored in the config file,
overruled by settings passed in ``args_dict`` (which correspond to
command-line arguments).
Results are saved to files according to provided settings.
Parameters
----------
args_dict : dict
Arguments from terminal, see :func:`emg3d.cli.main`. Parameters passed
in ``args_dict`` overrule parameters in the ``config``.
"""
# Start timer.
runtime = utils.Timer()
# Parse configuration file.
cfg, term = parser.parse_config_file(args_dict)
check_files(cfg, term) # Check all files and directories exist.
function, verb = term['function'], term['verbosity']
dry_run = term.get('dry_run', False)
# Start this task: start timing.
logger = initiate_logger(cfg, runtime, verb)
# Log start info, python and emg3d version, and python path.
logger.info(f":: emg3d CLI {function} START :: {time.asctime()} :: "
f"v{utils.__version__}")
logger.debug(f"{utils.Report()}")
# Dump the configuration.
paramdump = json.dumps(cfg, sort_keys=True, indent=4)
logger.debug("\n :: CONFIGURATION ::\n")
logger.debug(f"{term['config_file']}\n{paramdump}")
min_offset = cfg['simulation_options'].pop('min_offset', 0.0)
if cfg['files']['load']:
# Load input.
logger.info("\n :: LOAD SIMULATION ::\n")
sim, sinfo = simulations.Simulation.from_file(
cfg['files']['load'], verb=-1)
logger.info(sinfo.split('\n')[0])
logger.debug(sinfo.split('\n')[1])
else:
# Load input.
logger.info("\n :: LOAD SURVEY AND MODEL ::\n")
sdata, sinfo = io.load(cfg['files']['survey'], verb=-1)
survey = sdata['survey']
logger.info(sinfo.split('\n')[0])
logger.debug(sinfo.split('\n')[1])
model, minfo = io.load(cfg['files']['model'], verb=-1)
logger.info(minfo.split('\n')[0])
logger.debug(minfo.split('\n')[1])
# Select data.
data = cfg['data']
if data:
survey = survey.select(
sources=data.get('sources', None),
receivers=data.get('receivers', None),
frequencies=data.get('frequencies', None),
remove_empty=data.get('remove_empty', False),
)
# Switch-off tqdm if verbosity is zero.
if verb < 1:
cfg['simulation_options']['tqdm_opts'] = {'disable': True}
# Create simulation.
sim = simulations.Simulation(
survey=survey,
model=model['model'],
verb=-1, # Only errors.
**cfg['simulation_options']
)
# Print simulation info.
logger.info("\n :: SIMULATION ::")
logger.info(f"\n{sim}\n")
# Print meshes.
logger.debug(" :: MESHES ::\n")
logger.debug(sim.print_grid_info(return_info=True))
# Initiate output dict, add configuration.
output = {'configuration': cfg}
# Compute forward model (all calls).
logger.info(" :: FORWARD COMPUTATION ::\n")
if dry_run:
output['data'] = np.zeros(sim.survey.shape, dtype=complex)
else:
if function == 'forward':
sim.compute(observed=True, min_offset=min_offset)
output['data'] = sim.data.observed
else:
sim.compute()
output['data'] = sim.data.synthetic
# Print Solver Logs.
if verb in [0, 1]:
sim.print_solver_info('efield', 0)
logger.debug(sim.print_solver_info('efield', 1, True))
# Compute the misfit.
if function in ['misfit', 'gradient']:
if dry_run:
output['misfit'] = 0.0
else:
output['misfit'] = sim.misfit
output['n_observations'] = sim.survey.count
# Compute the gradient.
if function == 'gradient':
logger.info("\n :: BACKWARD COMPUTATION ::\n")
if dry_run:
output['gradient'] = np.zeros(sim.model.grid.shape_cells)
else:
output['gradient'] = sim.gradient
# Print Solver Logs.
if verb in [0, 1]:
sim.print_solver_info('bfield', 0)
logger.debug(sim.print_solver_info('bfield', 1, True))
# Store output to disk.
logger.info(" :: SAVE RESULTS ::\n")
if cfg['files']['save']:
oinfo = sim.to_file(cfg['files']['save'], verb=-1)
logger.info(oinfo.split('\n')[0])
logger.debug(oinfo.split('\n')[1])
oinfo = io.save(cfg['files']['output'], **output, verb=-1)
logger.info(oinfo.split('\n')[0])
logger.debug(oinfo.split('\n')[1])
# Goodbye
logger.info(f"\n:: emg3d CLI {function} END :: {time.asctime()} :: "
f"runtime = {runtime.runtime}")
