def __init__(self, tardis_config):
self.tardis_config = tardis_config
self.runner = MontecarloRunner(self.tardis_config.montecarlo.seed,
tardis_config.spectrum.frequency)
t_inner_lock_cycle = [False] * (
tardis_config.montecarlo.convergence_strategy.lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
def __init__(self, tardis_config):
#final preparation for configuration object
self.tardis_config = tardis_config
self.gui = None
self.converged = False
self.atom_data = tardis_config.atom_data
selected_atomic_numbers = self.tardis_config.abundances.index
self.atom_data.prepare_atom_data(selected_atomic_numbers,
line_interaction_type=tardis_config.plasma.line_interaction_type,
nlte_species=tardis_config.plasma.nlte.species)
if tardis_config.plasma.ionization == 'nebular':
if not self.atom_data.has_zeta_data:
raise ValueError("Requiring Recombination coefficients Zeta for 'nebular' plasma ionization")
self.packet_src = packet_source.SimplePacketSource.from_wavelength(tardis_config.montecarlo.black_body_sampling.start,
tardis_config.montecarlo.black_body_sampling.end,
blackbody_sampling=tardis_config.montecarlo.black_body_sampling.samples,
seed=self.tardis_config.montecarlo.seed)
self.current_no_of_packets = tardis_config.montecarlo.no_of_packets
self.t_inner = tardis_config.plasma.t_inner
self.t_rads = tardis_config.plasma.t_rads
self.iterations_max_requested = tardis_config.montecarlo.iterations
self.iterations_remaining = self.iterations_max_requested
self.iterations_executed = 0
if tardis_config.montecarlo.convergence_strategy.type == 'specific':
self.global_convergence_parameters = (tardis_config.montecarlo.
convergence_strategy.
deepcopy())
self.t_rads = tardis_config.plasma.t_rads
t_inner_lock_cycle = [False] * (tardis_config.montecarlo.
convergence_strategy.
lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
self.ws = (0.5 * (1 - np.sqrt(1 -
(tardis_config.structure.r_inner[0] ** 2 / tardis_config.structure.r_middle ** 2).to(1).value)))
self.plasma_array = LegacyPlasmaArray(tardis_config.number_densities, tardis_config.atom_data,
tardis_config.supernova.time_explosion.to('s').value,
nlte_config=tardis_config.plasma.nlte,
delta_treatment=tardis_config.plasma.delta_treatment,
ionization_mode=tardis_config.plasma.ionization,
excitation_mode=tardis_config.plasma.excitation,
line_interaction_type=tardis_config.plasma.line_interaction_type,
link_t_rad_t_electron=0.9)
self.spectrum = TARDISSpectrum(tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.spectrum_virtual = TARDISSpectrum(tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.spectrum_reabsorbed = TARDISSpectrum(tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.runner = MontecarloRunner()
def from_config(cls, config, **kwargs):
"""
Create a new Simulation instance from a Configuration object.
Parameters
----------
config : tardis.io.config_reader.Configuration
**kwargs
Allow overriding some structures, such as model, plasma, atomic data
and the runner, instead of creating them from the configuration
object.
Returns
-------
Simulation
"""
# Allow overriding some config structures. This is useful in some
# unit tests, and could be extended in all the from_config classmethods.
if "model" in kwargs:
model = kwargs["model"]
else:
model = Radial1DModel.from_config(config)
if "plasma" in kwargs:
plasma = kwargs["plasma"]
else:
plasma = assemble_plasma(config, model, atom_data=kwargs.get("atom_data", None))
if "runner" in kwargs:
runner = kwargs["runner"]
else:
runner = MontecarloRunner.from_config(config)
luminosity_nu_start = config.supernova.luminosity_wavelength_end.to(u.Hz, u.spectral())
try:
luminosity_nu_end = config.supernova.luminosity_wavelength_start.to(u.Hz, u.spectral())
except ZeroDivisionError:
luminosity_nu_end = np.inf * u.Hz
last_no_of_packets = config.montecarlo.last_no_of_packets
if last_no_of_packets is None or last_no_of_packets < 0:
last_no_of_packets = config.montecarlo.no_of_packets
last_no_of_packets = int(last_no_of_packets)
return cls(
iterations=config.montecarlo.iterations,
model=model,
plasma=plasma,
runner=runner,
no_of_packets=int(config.montecarlo.no_of_packets),
no_of_virtual_packets=int(config.montecarlo.no_of_virtual_packets),
luminosity_nu_start=luminosity_nu_start,
luminosity_nu_end=luminosity_nu_end,
last_no_of_packets=last_no_of_packets,
luminosity_requested=config.supernova.luminosity_requested.cgs,
convergence_strategy=config.montecarlo.convergence_strategy,
nthreads=config.montecarlo.nthreads,
)
def __init__(self, tardis_config):
self.tardis_config = tardis_config
self.runner = MontecarloRunner(self.tardis_config.montecarlo.seed,
tardis_config.spectrum.frequency)
t_inner_lock_cycle = [False] * (tardis_config.montecarlo.
convergence_strategy.
lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
def runner():
config_fname = "tardis/io/tests/data/tardis_configv1_verysimply.yml"
config = config_reader.Configuration.from_yaml(config_fname)
runner = MontecarloRunner.from_config(config)
return runner
class Radial1DModel(object):
"""
Class to hold the states of the individual shells (the state of the plasma (as a `~plasma.BasePlasma`-object or one of its subclasses),
, the plasma parameters (e.g. temperature, dilution factor), the dimensions of the shell).
Parameters
----------
tardis_configuration : `tardis.config_reader.Configuration`
velocities : `np.ndarray`
an array with n+1 (for n shells) velocities (in cm/s) for each of the boundaries (velocities[0] describing
the inner boundary and velocities[-1] the outer boundary
densities : `np.ndarray`
an array with n densities - being the density mid-shell (assumed for the whole shell)
abundances : `list` or `dict`
a dictionary for uniform abundances throughout all shells, e.g. dict(Fe=0.5, Si=0.5)
For a different abundance for each shell list of abundance dictionaries.
time_explosion : `float`
time since explosion in seconds
atom_data : `~tardis.atom_data.AtomData` class or subclass
Containing the atom data needed for the plasma calculations
ws : `None` or `list`-like
ws can only be specified for plasma_type 'nebular'. If `None` is specified at first initialization the class
calculates an initial geometric dilution factor. When giving a list positive values will be accepted, whereas
negative values trigger the usage of the geometric calculation
plasma_type : `str`
plasma type currently supports 'lte' (using `tardis.plasma.LTEPlasma`)
or 'nebular' (using `tardis.plasma.NebularPlasma`)
initial_t_rad : `float`-like or `list`-like
initial radiative temperature for each shell, if a scalar is specified it initializes with a uniform
temperature for all shells
"""
@classmethod
def from_h5(cls, buffer_or_fname):
raise NotImplementedError("This is currently not implemented")
def __init__(self, tardis_config):
#final preparation for configuration object
self.tardis_config = tardis_config
self.gui = None
self.converged = False
self.atom_data = tardis_config.atom_data
selected_atomic_numbers = self.tardis_config.abundances.index
self.atom_data.prepare_atom_data(selected_atomic_numbers,
line_interaction_type=tardis_config.plasma.line_interaction_type,
nlte_species=tardis_config.plasma.nlte.species)
if tardis_config.plasma.ionization == 'nebular':
if not self.atom_data.has_zeta_data:
raise ValueError("Requiring Recombination coefficients Zeta for 'nebular' plasma ionization")
self.packet_src = packet_source.SimplePacketSource.from_wavelength(tardis_config.montecarlo.black_body_sampling.start,
tardis_config.montecarlo.black_body_sampling.end,
blackbody_sampling=tardis_config.montecarlo.black_body_sampling.samples,
seed=self.tardis_config.montecarlo.seed)
self.current_no_of_packets = tardis_config.montecarlo.no_of_packets
self.t_inner = tardis_config.plasma.t_inner
self.t_rads = tardis_config.plasma.t_rads
self.iterations_max_requested = tardis_config.montecarlo.iterations
self.iterations_remaining = self.iterations_max_requested
self.iterations_executed = 0
if tardis_config.montecarlo.convergence_strategy.type == 'specific':
self.global_convergence_parameters = (tardis_config.montecarlo.
convergence_strategy.
deepcopy())
self.t_rads = tardis_config.plasma.t_rads
t_inner_lock_cycle = [False] * (tardis_config.montecarlo.
convergence_strategy.
lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
self.ws = (0.5 * (1 - np.sqrt(1 -
(tardis_config.structure.r_inner[0] ** 2 / tardis_config.structure.r_middle ** 2).to(1).value)))
self.plasma_array = LegacyPlasmaArray(tardis_config.number_densities, tardis_config.atom_data,
tardis_config.supernova.time_explosion.to('s').value,
nlte_config=tardis_config.plasma.nlte,
delta_treatment=tardis_config.plasma.delta_treatment,
ionization_mode=tardis_config.plasma.ionization,
excitation_mode=tardis_config.plasma.excitation,
line_interaction_type=tardis_config.plasma.line_interaction_type,
link_t_rad_t_electron=0.9)
self.spectrum = TARDISSpectrum(tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.spectrum_virtual = TARDISSpectrum(tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.spectrum_reabsorbed = TARDISSpectrum(tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.runner = MontecarloRunner()
@property
def line_interaction_type(self):
return self._line_interaction_type
@line_interaction_type.setter
def line_interaction_type(self, value):
if value in ['scatter', 'downbranch', 'macroatom']:
self._line_interaction_type = value
self.tardis_config.plasma.line_interaction_type = value
#final preparation for atom_data object - currently building data
self.atom_data.prepare_atom_data(self.tardis_config.number_densities.columns,
line_interaction_type=self.line_interaction_type, max_ion_number=None,
nlte_species=self.tardis_config.plasma.nlte.species)
else:
raise ValueError('line_interaction_type can only be "scatter", "downbranch", or "macroatom"')
@property
def t_inner(self):
return self._t_inner
@t_inner.setter
def t_inner(self, value):
self._t_inner = value
self.luminosity_inner = (4 * np.pi * constants.sigma_sb.cgs * self.tardis_config.structure.r_inner[0] ** 2 * \
self.t_inner ** 4).to('erg/s')
self.time_of_simulation = (1.0 * u.erg / self.luminosity_inner)
self.j_blues_norm_factor = constants.c.cgs * self.tardis_config.supernova.time_explosion / \
(4 * np.pi * self.time_of_simulation * self.tardis_config.structure.volumes)
def calculate_j_blues(self, init_detailed_j_blues=False):
nus = self.atom_data.lines.nu.values
radiative_rates_type = self.tardis_config.plasma.radiative_rates_type
w_epsilon = self.tardis_config.plasma.w_epsilon
if radiative_rates_type == 'blackbody':
logger.info('Calculating J_blues for radiative_rates_type=lte')
j_blues = intensity_black_body(nus[np.newaxis].T, self.t_rads.value)
self.j_blues = pd.DataFrame(j_blues, index=self.atom_data.lines.index, columns=np.arange(len(self.t_rads)))
elif radiative_rates_type == 'dilute-blackbody' or init_detailed_j_blues:
logger.info('Calculating J_blues for radiative_rates_type=dilute-blackbody')
j_blues = self.ws * intensity_black_body(nus[np.newaxis].T, self.t_rads.value)
self.j_blues = pd.DataFrame(j_blues, index=self.atom_data.lines.index, columns=np.arange(len(self.t_rads)))
elif radiative_rates_type == 'detailed':
logger.info('Calculating J_blues for radiate_rates_type=detailed')
self.j_blues = pd.DataFrame(self.j_blue_estimators.transpose() * self.j_blues_norm_factor.value,
index=self.atom_data.lines.index, columns=np.arange(len(self.t_rads)))
for i in xrange(self.tardis_config.structure.no_of_shells):
zero_j_blues = self.j_blues[i] == 0.0
self.j_blues[i][zero_j_blues] = w_epsilon * intensity_black_body(
self.atom_data.lines.nu.values[zero_j_blues], self.t_rads.value[i])
else:
raise ValueError('radiative_rates_type type unknown - %s', radiative_rates_type)
def update_plasmas(self, initialize_nlte=False):
self.plasma_array.update_radiationfield(self.t_rads.value, self.ws, self.j_blues,
self.tardis_config.plasma.nlte, initialize_nlte=initialize_nlte, n_e_convergence_threshold=0.05)
if self.tardis_config.plasma.line_interaction_type in ('downbranch', 'macroatom'):
self.transition_probabilities = self.plasma_array.transition_probabilities
def update_radiationfield(self, log_sampling=5):
"""
Updating radiation field
"""
convergence_section = self.tardis_config.montecarlo.convergence_strategy
updated_t_rads, updated_ws = (
self.runner.calculate_radiationfield_properties())
old_t_rads = self.t_rads.copy()
old_ws = self.ws.copy()
old_t_inner = self.t_inner
luminosity_wavelength_filter = (self.montecarlo_nu > self.tardis_config.supernova.luminosity_nu_start) & \
(self.montecarlo_nu < self.tardis_config.supernova.luminosity_nu_end)
emitted_filter = self.montecarlo_luminosity.value >= 0
emitted_luminosity = np.sum(self.montecarlo_luminosity.value[emitted_filter & luminosity_wavelength_filter]) \
* self.montecarlo_luminosity.unit
absorbed_luminosity = -np.sum(self.montecarlo_luminosity.value[~emitted_filter & luminosity_wavelength_filter]) \
* self.montecarlo_luminosity.unit
updated_t_inner = self.t_inner \
* (emitted_luminosity / self.tardis_config.supernova.luminosity_requested).to(1).value \
** convergence_section.t_inner_update_exponent
#updated_t_inner = np.max([np.min([updated_t_inner, 30000]), 3000])
convergence_t_rads = (abs(old_t_rads - updated_t_rads) / updated_t_rads).value
convergence_ws = (abs(old_ws - updated_ws) / updated_ws)
convergence_t_inner = (abs(old_t_inner - updated_t_inner) / updated_t_inner).value
if convergence_section.type == 'damped' or convergence_section.type == 'specific':
self.t_rads += convergence_section.t_rad.damping_constant * (updated_t_rads - self.t_rads)
self.ws += convergence_section.w.damping_constant * (updated_ws - self.ws)
if self.t_inner_update.next():
t_inner_new = self.t_inner + convergence_section.t_inner.damping_constant * (updated_t_inner - self.t_inner)
else:
t_inner_new = self.t_inner
if convergence_section.type == 'specific':
t_rad_converged = (float(np.sum(convergence_t_rads < convergence_section.t_rad['threshold'])) \
/ self.tardis_config.structure.no_of_shells) >= convergence_section.t_rad['fraction']
w_converged = (float(np.sum(convergence_t_rads < convergence_section.w['threshold'])) \
/ self.tardis_config.structure.no_of_shells) >= convergence_section.w['fraction']
t_inner_converged = convergence_t_inner < convergence_section.t_inner['threshold']
if t_rad_converged and t_inner_converged and w_converged:
if not self.converged:
self.converged = True
self.iterations_remaining = self.global_convergence_parameters['hold_iterations']
else:
if self.converged:
self.iterations_remaining = self.iterations_max_requested - self.iterations_executed
self.converged = False
self.temperature_logging = pd.DataFrame(
{'t_rads': old_t_rads.value, 'updated_t_rads': updated_t_rads.value,
'converged_t_rads': convergence_t_rads, 'new_trads': self.t_rads.value, 'ws': old_ws,
'updated_ws': updated_ws, 'converged_ws': convergence_ws,
'new_ws': self.ws})
self.temperature_logging.index.name = 'Shell'
temperature_logging = str(self.temperature_logging[::log_sampling])
temperature_logging = ''.join(['\t%s\n' % item for item in temperature_logging.split('\n')])
logger.info('Plasma stratification:\n%s\n', temperature_logging)
logger.info("Luminosity emitted = %.5e Luminosity absorbed = %.5e Luminosity requested = %.5e",
emitted_luminosity.value, absorbed_luminosity.value,
self.tardis_config.supernova.luminosity_requested.value)
logger.info('Calculating new t_inner = %.3f', updated_t_inner.value)
return t_inner_new
def simulate(self, update_radiation_field=True, enable_virtual=False, initialize_j_blues=False,
initialize_nlte=False):
"""
Run a simulation
"""
if update_radiation_field:
t_inner_new = self.update_radiationfield()
else:
t_inner_new = self.t_inner
self.calculate_j_blues(init_detailed_j_blues=initialize_j_blues)
self.update_plasmas(initialize_nlte=initialize_nlte)
self.t_inner = t_inner_new
self.packet_src.create_packets(self.current_no_of_packets, self.t_inner.value)
if enable_virtual:
no_of_virtual_packets = self.tardis_config.montecarlo.no_of_virtual_packets
else:
no_of_virtual_packets = 0
if np.any(np.isnan(self.plasma_array.tau_sobolevs.values)) or np.any(np.isinf(self.plasma_array.tau_sobolevs.values)) \
or np.any(np.isneginf(self.plasma_array.tau_sobolevs.values)):
raise ValueError('Some tau_sobolevs are nan, inf, -inf in tau_sobolevs. Something went wrong!')
self.j_blue_estimators = np.zeros((len(self.t_rads), len(self.atom_data.lines)))
self.montecarlo_virtual_luminosity = np.zeros_like(self.spectrum.frequency.value)
self.runner.run(self, no_of_virtual_packets=no_of_virtual_packets,
nthreads=self.tardis_config.montecarlo.nthreads) #self = model
(montecarlo_nu, montecarlo_energies, self.j_estimators,
self.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, self.last_interaction_type,
self.last_line_interaction_shell_id) = self.runner.legacy_return()
if np.sum(montecarlo_energies < 0) == len(montecarlo_energies):
logger.critical("No r-packet escaped through the outer boundary.")
self.montecarlo_nu = self.runner.packet_nu
self.montecarlo_luminosity = self.runner.packet_luminosity
montecarlo_reabsorbed_luminosity = np.histogram(
self.runner.reabsorbed_packet_nu,
weights=self.runner.reabsorbed_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
montecarlo_emitted_luminosity = np.histogram(
self.runner.emitted_packet_nu,
weights=self.runner.emitted_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
self.spectrum.update_luminosity(montecarlo_emitted_luminosity)
self.spectrum_reabsorbed.update_luminosity(montecarlo_reabsorbed_luminosity)
if no_of_virtual_packets > 0:
self.montecarlo_virtual_luminosity = self.montecarlo_virtual_luminosity \
* 1 * u.erg / self.time_of_simulation
self.spectrum_virtual.update_luminosity(self.montecarlo_virtual_luminosity)
self.last_line_interaction_in_id = self.atom_data.lines_index.index.values[last_line_interaction_in_id]
self.last_line_interaction_in_id = self.last_line_interaction_in_id[last_line_interaction_in_id != -1]
self.last_line_interaction_out_id = self.atom_data.lines_index.index.values[last_line_interaction_out_id]
self.last_line_interaction_out_id = self.last_line_interaction_out_id[last_line_interaction_out_id != -1]
self.last_line_interaction_angstrom = self.montecarlo_nu[last_line_interaction_in_id != -1].to('angstrom',
u.spectral())
self.iterations_executed += 1
self.iterations_remaining -= 1
if self.gui is not None:
self.gui.update_data(self)
self.gui.show()
def save_spectra(self, fname):
self.spectrum.to_ascii(fname)
self.spectrum_virtual.to_ascii('virtual_' + fname)
def to_hdf5(self, buffer_or_fname, path='', close_h5=True):
"""
This allows the model to be written to an HDF5 file for later analysis. Currently, the saved properties
are specified hard coded in include_from_model_in_hdf5. This is a dict where the key corresponds to the
name of the property and the value describes the type. If the value is None the property can be dumped
to hdf via its attribute to_hdf or by converting it to a pd.DataFrame. For more complex properties
which can not simply be dumped to an hdf file the dict can contain a function which is called with
the parameters key, path, and hdf_store. This function then should dump the data to the given
hdf_store object. To dump properties of sub-properties of the model, you can use a dict as value.
This dict is then treated in the same way as described above.
Parameters
----------
buffer_or_fname: buffer or ~str
buffer or filename for HDF5 file (see pandas.HDFStore for description)
path: ~str, optional
path in the HDF5 file
close_h5: ~bool
close the HDF5 file or not.
"""
# Functions to save properties of the model without to_hdf attribute and no simple conversion to a pd.DataFrame.
#This functions are always called with the parameters key, path and, hdf_store.
def _save_luminosity_density(key, path, hdf_store):
luminosity_density = pd.DataFrame.from_dict(dict(wave=self.spectrum.wavelength.value,
flux=self.spectrum.luminosity_density_lambda.value))
luminosity_density.to_hdf(hdf_store, os.path.join(path, key))
def _save_spectrum_virtual(key, path, hdf_store):
if self.spectrum_virtual.luminosity_density_lambda is not None:
luminosity_density_virtual = pd.DataFrame.from_dict(dict(wave=self.spectrum_virtual.wavelength.value,
flux=self.spectrum_virtual.luminosity_density_lambda.value))
luminosity_density_virtual.to_hdf(hdf_store, os.path.join(path, key))
def _save_configuration_dict(key, path, hdf_store):
configuration_dict = dict(t_inner=self.t_inner.value)
configuration_dict_path = os.path.join(path, 'configuration')
pd.Series(configuration_dict).to_hdf(hdf_store, configuration_dict_path)
include_from_plasma_ = {'level_number_density': None, 'ion_number_density': None, 'tau_sobolevs': None,
'electron_densities': None,
't_rad': None, 'w': None}
include_from_model_in_hdf5 = {'plasma_array': include_from_plasma_, 'j_blues': None,
'last_line_interaction_in_id': None,
'last_line_interaction_out_id': None,
'last_line_interaction_shell_id': None, 'montecarlo_nu': None,
'luminosity_density': _save_luminosity_density,
'luminosity_density_virtual': _save_spectrum_virtual,
'configuration_dict': _save_configuration_dict,
'last_line_interaction_angstrom': None}
if isinstance(buffer_or_fname, basestring):
hdf_store = pd.HDFStore(buffer_or_fname)
elif isinstance(buffer_or_fname, pd.HDFStore):
hdf_store = buffer_or_fname
else:
raise IOError('Please specify either a filename or an HDFStore')
logger.info('Writing to path %s', path)
def _get_hdf5_path(path, property_name):
return os.path.join(path, property_name)
def _to_smallest_pandas(object):
try:
return pd.Series(object)
except Exception:
return pd.DataFrame(object)
def _save_model_property(object, property_name, path, hdf_store):
property_path = _get_hdf5_path(path, property_name)
try:
object.to_hdf(hdf_store, property_path)
except AttributeError:
_to_smallest_pandas(object).to_hdf(hdf_store, property_path)
for key in include_from_model_in_hdf5:
if include_from_model_in_hdf5[key] is None:
_save_model_property(getattr(self, key), key, path, hdf_store)
elif callable(include_from_model_in_hdf5[key]):
include_from_model_in_hdf5[key](key, path, hdf_store)
else:
try:
for subkey in include_from_model_in_hdf5[key]:
if include_from_model_in_hdf5[key][subkey] is None:
_save_model_property(getattr(getattr(self, key), subkey), subkey, os.path.join(path, key),
hdf_store)
elif callable(include_from_model_in_hdf5[key][subkey]):
include_from_model_in_hdf5[key][subkey](subkey, os.path.join(path, key), hdf_store)
else:
logger.critical('Can not save %s', str(os.path.join(path, key, subkey)))
except:
logger.critical('An error occurred while dumping %s to HDF.', str(os.path.join(path, key)))
hdf_store.flush()
if close_h5:
hdf_store.close()
else:
return hdf_store
def __init__(self, tardis_config):
#final preparation for configuration object
self.tardis_config = tardis_config
self.gui = None
self.converged = False
self.atom_data = tardis_config.atom_data
selected_atomic_numbers = self.tardis_config.abundances.index
self.atom_data.prepare_atom_data(
selected_atomic_numbers,
line_interaction_type=tardis_config.plasma.line_interaction_type,
nlte_species=tardis_config.plasma.nlte.species)
if tardis_config.plasma.ionization == 'nebular':
if not self.atom_data.has_zeta_data:
raise ValueError(
"Requiring Recombination coefficients Zeta for 'nebular' plasma ionization"
)
self.packet_src = packet_source.SimplePacketSource.from_wavelength(
tardis_config.montecarlo.black_body_sampling.start,
tardis_config.montecarlo.black_body_sampling.end,
blackbody_sampling=tardis_config.montecarlo.black_body_sampling.
samples,
seed=self.tardis_config.montecarlo.seed)
self.current_no_of_packets = tardis_config.montecarlo.no_of_packets
self.t_inner = tardis_config.plasma.t_inner
self.t_rads = tardis_config.plasma.t_rads
self.iterations_max_requested = tardis_config.montecarlo.iterations
self.iterations_remaining = self.iterations_max_requested
self.iterations_executed = 0
if tardis_config.montecarlo.convergence_strategy.type == 'specific':
self.global_convergence_parameters = (
tardis_config.montecarlo.convergence_strategy.deepcopy())
self.t_rads = tardis_config.plasma.t_rads
t_inner_lock_cycle = [False] * (
tardis_config.montecarlo.convergence_strategy.lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
self.ws = (
0.5 *
(1 - np.sqrt(1 -
(tardis_config.structure.r_inner[0]**2 /
tardis_config.structure.r_middle**2).to(1).value)))
self.plasma_array = LegacyPlasmaArray(
tardis_config.number_densities,
tardis_config.atom_data,
tardis_config.supernova.time_explosion.to('s').value,
nlte_config=tardis_config.plasma.nlte,
delta_treatment=tardis_config.plasma.delta_treatment,
ionization_mode=tardis_config.plasma.ionization,
excitation_mode=tardis_config.plasma.excitation,
line_interaction_type=tardis_config.plasma.line_interaction_type,
link_t_rad_t_electron=0.9,
helium_treatment=tardis_config.plasma.helium_treatment)
self.spectrum = TARDISSpectrum(tardis_config.spectrum.frequency,
tardis_config.supernova.distance)
self.spectrum_virtual = TARDISSpectrum(
tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.spectrum_reabsorbed = TARDISSpectrum(
tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.runner = MontecarloRunner()
class Radial1DModel(object):
"""
Class to hold the states of the individual shells (the state of the plasma (as a `~plasma.BasePlasma`-object or one of its subclasses),
, the plasma parameters (e.g. temperature, dilution factor), the dimensions of the shell).
Parameters
----------
tardis_configuration : `tardis.config_reader.Configuration`
velocities : `np.ndarray`
an array with n+1 (for n shells) velocities (in cm/s) for each of the boundaries (velocities[0] describing
the inner boundary and velocities[-1] the outer boundary
densities : `np.ndarray`
an array with n densities - being the density mid-shell (assumed for the whole shell)
abundances : `list` or `dict`
a dictionary for uniform abundances throughout all shells, e.g. dict(Fe=0.5, Si=0.5)
For a different abundance for each shell list of abundance dictionaries.
time_explosion : `float`
time since explosion in seconds
atom_data : `~tardis.atom_data.AtomData` class or subclass
Containing the atom data needed for the plasma calculations
ws : `None` or `list`-like
ws can only be specified for plasma_type 'nebular'. If `None` is specified at first initialization the class
calculates an initial geometric dilution factor. When giving a list positive values will be accepted, whereas
negative values trigger the usage of the geometric calculation
plasma_type : `str`
plasma type currently supports 'lte' (using `tardis.plasma.LTEPlasma`)
or 'nebular' (using `tardis.plasma.NebularPlasma`)
initial_t_rad : `float`-like or `list`-like
initial radiative temperature for each shell, if a scalar is specified it initializes with a uniform
temperature for all shells
"""
@classmethod
def from_h5(cls, buffer_or_fname):
raise NotImplementedError("This is currently not implemented")
def __init__(self, tardis_config):
#final preparation for configuration object
self.tardis_config = tardis_config
self.gui = None
self.converged = False
self.atom_data = tardis_config.atom_data
selected_atomic_numbers = self.tardis_config.abundances.index
self.atom_data.prepare_atom_data(
selected_atomic_numbers,
line_interaction_type=tardis_config.plasma.line_interaction_type,
nlte_species=tardis_config.plasma.nlte.species)
if tardis_config.plasma.ionization == 'nebular':
if not self.atom_data.has_zeta_data:
raise ValueError(
"Requiring Recombination coefficients Zeta for 'nebular' plasma ionization"
)
self.packet_src = packet_source.SimplePacketSource.from_wavelength(
tardis_config.montecarlo.black_body_sampling.start,
tardis_config.montecarlo.black_body_sampling.end,
blackbody_sampling=tardis_config.montecarlo.black_body_sampling.
samples,
seed=self.tardis_config.montecarlo.seed)
self.current_no_of_packets = tardis_config.montecarlo.no_of_packets
self.t_inner = tardis_config.plasma.t_inner
self.t_rads = tardis_config.plasma.t_rads
self.iterations_max_requested = tardis_config.montecarlo.iterations
self.iterations_remaining = self.iterations_max_requested
self.iterations_executed = 0
if tardis_config.montecarlo.convergence_strategy.type == 'specific':
self.global_convergence_parameters = (
tardis_config.montecarlo.convergence_strategy.deepcopy())
self.t_rads = tardis_config.plasma.t_rads
t_inner_lock_cycle = [False] * (
tardis_config.montecarlo.convergence_strategy.lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
self.ws = (
0.5 *
(1 - np.sqrt(1 -
(tardis_config.structure.r_inner[0]**2 /
tardis_config.structure.r_middle**2).to(1).value)))
self.plasma_array = LegacyPlasmaArray(
tardis_config.number_densities,
tardis_config.atom_data,
tardis_config.supernova.time_explosion.to('s').value,
nlte_config=tardis_config.plasma.nlte,
delta_treatment=tardis_config.plasma.delta_treatment,
ionization_mode=tardis_config.plasma.ionization,
excitation_mode=tardis_config.plasma.excitation,
line_interaction_type=tardis_config.plasma.line_interaction_type,
link_t_rad_t_electron=0.9,
helium_treatment=tardis_config.plasma.helium_treatment)
self.spectrum = TARDISSpectrum(tardis_config.spectrum.frequency,
tardis_config.supernova.distance)
self.spectrum_virtual = TARDISSpectrum(
tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.spectrum_reabsorbed = TARDISSpectrum(
tardis_config.spectrum.frequency, tardis_config.supernova.distance)
self.runner = MontecarloRunner()
@property
def line_interaction_type(self):
return self._line_interaction_type
@line_interaction_type.setter
def line_interaction_type(self, value):
if value in ['scatter', 'downbranch', 'macroatom']:
self._line_interaction_type = value
self.tardis_config.plasma.line_interaction_type = value
#final preparation for atom_data object - currently building data
self.atom_data.prepare_atom_data(
self.tardis_config.number_densities.columns,
line_interaction_type=self.line_interaction_type,
max_ion_number=None,
nlte_species=self.tardis_config.plasma.nlte.species)
else:
raise ValueError(
'line_interaction_type can only be "scatter", "downbranch", or "macroatom"'
)
@property
def t_inner(self):
return self._t_inner
@t_inner.setter
def t_inner(self, value):
self._t_inner = value
self.luminosity_inner = (4 * np.pi * constants.sigma_sb.cgs * self.tardis_config.structure.r_inner[0] ** 2 * \
self.t_inner ** 4).to('erg/s')
self.time_of_simulation = (1.0 * u.erg / self.luminosity_inner)
self.j_blues_norm_factor = constants.c.cgs * self.tardis_config.supernova.time_explosion / \
(4 * np.pi * self.time_of_simulation * self.tardis_config.structure.volumes)
def calculate_j_blues(self, init_detailed_j_blues=False):
nus = self.atom_data.lines.nu.values
radiative_rates_type = self.tardis_config.plasma.radiative_rates_type
w_epsilon = self.tardis_config.plasma.w_epsilon
if radiative_rates_type == 'blackbody':
logger.info('Calculating J_blues for radiative_rates_type=lte')
j_blues = intensity_black_body(nus[np.newaxis].T,
self.t_rads.value)
self.j_blues = pd.DataFrame(j_blues,
index=self.atom_data.lines.index,
columns=np.arange(len(self.t_rads)))
elif radiative_rates_type == 'dilute-blackbody' or init_detailed_j_blues:
logger.info(
'Calculating J_blues for radiative_rates_type=dilute-blackbody'
)
j_blues = self.ws * intensity_black_body(nus[np.newaxis].T,
self.t_rads.value)
self.j_blues = pd.DataFrame(j_blues,
index=self.atom_data.lines.index,
columns=np.arange(len(self.t_rads)))
elif radiative_rates_type == 'detailed':
logger.info('Calculating J_blues for radiate_rates_type=detailed')
self.j_blues = pd.DataFrame(self.j_blue_estimators.transpose() *
self.j_blues_norm_factor.value,
index=self.atom_data.lines.index,
columns=np.arange(len(self.t_rads)))
for i in xrange(self.tardis_config.structure.no_of_shells):
zero_j_blues = self.j_blues[i] == 0.0
self.j_blues[i][
zero_j_blues] = w_epsilon * intensity_black_body(
self.atom_data.lines.nu.values[zero_j_blues],
self.t_rads.value[i])
else:
raise ValueError('radiative_rates_type type unknown - %s',
radiative_rates_type)
def update_plasmas(self, initialize_nlte=False):
self.plasma_array.update_radiationfield(
self.t_rads.value,
self.ws,
self.j_blues,
self.tardis_config.plasma.nlte,
initialize_nlte=initialize_nlte,
n_e_convergence_threshold=0.05)
if self.tardis_config.plasma.line_interaction_type in ('downbranch',
'macroatom'):
self.transition_probabilities = self.plasma_array.transition_probabilities
def update_radiationfield(self, log_sampling=5):
"""
Updating radiation field
"""
convergence_section = self.tardis_config.montecarlo.convergence_strategy
updated_t_rads, updated_ws = (
self.runner.calculate_radiationfield_properties())
old_t_rads = self.t_rads.copy()
old_ws = self.ws.copy()
old_t_inner = self.t_inner
luminosity_wavelength_filter = (self.montecarlo_nu > self.tardis_config.supernova.luminosity_nu_start) & \
(self.montecarlo_nu < self.tardis_config.supernova.luminosity_nu_end)
emitted_filter = self.montecarlo_luminosity.value >= 0
emitted_luminosity = np.sum(self.montecarlo_luminosity.value[emitted_filter & luminosity_wavelength_filter]) \
* self.montecarlo_luminosity.unit
absorbed_luminosity = -np.sum(self.montecarlo_luminosity.value[~emitted_filter & luminosity_wavelength_filter]) \
* self.montecarlo_luminosity.unit
updated_t_inner = self.t_inner \
* (emitted_luminosity / self.tardis_config.supernova.luminosity_requested).to(1).value \
** convergence_section.t_inner_update_exponent
#updated_t_inner = np.max([np.min([updated_t_inner, 30000]), 3000])
convergence_t_rads = (abs(old_t_rads - updated_t_rads) /
updated_t_rads).value
convergence_ws = (abs(old_ws - updated_ws) / updated_ws)
convergence_t_inner = (abs(old_t_inner - updated_t_inner) /
updated_t_inner).value
if convergence_section.type == 'damped' or convergence_section.type == 'specific':
self.t_rads += convergence_section.t_rad.damping_constant * (
updated_t_rads - self.t_rads)
self.ws += convergence_section.w.damping_constant * (updated_ws -
self.ws)
if self.t_inner_update.next():
t_inner_new = self.t_inner + convergence_section.t_inner.damping_constant * (
updated_t_inner - self.t_inner)
else:
t_inner_new = self.t_inner
if convergence_section.type == 'specific':
t_rad_converged = (float(np.sum(convergence_t_rads < convergence_section.t_rad['threshold'])) \
/ self.tardis_config.structure.no_of_shells) >= convergence_section.t_rad['fraction']
w_converged = (float(np.sum(convergence_t_rads < convergence_section.w['threshold'])) \
/ self.tardis_config.structure.no_of_shells) >= convergence_section.w['fraction']
t_inner_converged = convergence_t_inner < convergence_section.t_inner[
'threshold']
if t_rad_converged and t_inner_converged and w_converged:
if not self.converged:
self.converged = True
self.iterations_remaining = self.global_convergence_parameters[
'hold_iterations']
else:
if self.converged:
self.iterations_remaining = self.iterations_max_requested - self.iterations_executed
self.converged = False
self.temperature_logging = pd.DataFrame({
't_rads':
old_t_rads.value,
'updated_t_rads':
updated_t_rads.value,
'converged_t_rads':
convergence_t_rads,
'new_trads':
self.t_rads.value,
'ws':
old_ws,
'updated_ws':
updated_ws,
'converged_ws':
convergence_ws,
'new_ws':
self.ws
})
self.temperature_logging.index.name = 'Shell'
temperature_logging = str(self.temperature_logging[::log_sampling])
temperature_logging = ''.join(
['\t%s\n' % item for item in temperature_logging.split('\n')])
logger.info('Plasma stratification:\n%s\n', temperature_logging)
logger.info(
"Luminosity emitted = %.5e Luminosity absorbed = %.5e Luminosity requested = %.5e",
emitted_luminosity.value, absorbed_luminosity.value,
self.tardis_config.supernova.luminosity_requested.value)
logger.info('Calculating new t_inner = %.3f', updated_t_inner.value)
return t_inner_new
def simulate(self,
update_radiation_field=True,
enable_virtual=False,
initialize_j_blues=False,
initialize_nlte=False):
"""
Run a simulation
"""
if update_radiation_field:
t_inner_new = self.update_radiationfield()
else:
t_inner_new = self.t_inner
self.calculate_j_blues(init_detailed_j_blues=initialize_j_blues)
self.update_plasmas(initialize_nlte=initialize_nlte)
self.t_inner = t_inner_new
self.packet_src.create_packets(self.current_no_of_packets,
self.t_inner.value)
if enable_virtual:
no_of_virtual_packets = self.tardis_config.montecarlo.no_of_virtual_packets
else:
no_of_virtual_packets = 0
if np.any(np.isnan(self.plasma_array.tau_sobolevs.values)) or np.any(np.isinf(self.plasma_array.tau_sobolevs.values)) \
or np.any(np.isneginf(self.plasma_array.tau_sobolevs.values)):
raise ValueError(
'Some tau_sobolevs are nan, inf, -inf in tau_sobolevs. Something went wrong!'
)
self.j_blue_estimators = np.zeros(
(len(self.t_rads), len(self.atom_data.lines)))
self.montecarlo_virtual_luminosity = np.zeros_like(
self.spectrum.frequency.value)
self.runner.run(
self,
no_of_virtual_packets=no_of_virtual_packets,
nthreads=self.tardis_config.montecarlo.nthreads) #self = model
(montecarlo_nu, montecarlo_energies, self.j_estimators,
self.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, self.last_interaction_type,
self.last_line_interaction_shell_id) = self.runner.legacy_return()
if np.sum(montecarlo_energies < 0) == len(montecarlo_energies):
logger.critical("No r-packet escaped through the outer boundary.")
self.montecarlo_nu = self.runner.packet_nu
self.montecarlo_luminosity = self.runner.packet_luminosity
montecarlo_reabsorbed_luminosity = np.histogram(
self.runner.reabsorbed_packet_nu,
weights=self.runner.reabsorbed_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
montecarlo_emitted_luminosity = np.histogram(
self.runner.emitted_packet_nu,
weights=self.runner.emitted_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
self.spectrum.update_luminosity(montecarlo_emitted_luminosity)
self.spectrum_reabsorbed.update_luminosity(
montecarlo_reabsorbed_luminosity)
if no_of_virtual_packets > 0:
self.montecarlo_virtual_luminosity = self.montecarlo_virtual_luminosity \
* 1 * u.erg / self.time_of_simulation
self.spectrum_virtual.update_luminosity(
self.montecarlo_virtual_luminosity)
self.last_line_interaction_in_id = self.atom_data.lines_index.index.values[
last_line_interaction_in_id]
self.last_line_interaction_in_id = self.last_line_interaction_in_id[
last_line_interaction_in_id != -1]
self.last_line_interaction_out_id = self.atom_data.lines_index.index.values[
last_line_interaction_out_id]
self.last_line_interaction_out_id = self.last_line_interaction_out_id[
last_line_interaction_out_id != -1]
self.last_line_interaction_angstrom = self.montecarlo_nu[
last_line_interaction_in_id != -1].to('angstrom', u.spectral())
self.iterations_executed += 1
self.iterations_remaining -= 1
if self.gui is not None:
self.gui.update_data(self)
self.gui.show()
def save_spectra(self, fname):
self.spectrum.to_ascii(fname)
self.spectrum_virtual.to_ascii('virtual_' + fname)
def to_hdf5(self, buffer_or_fname, path='', close_h5=True):
"""
This allows the model to be written to an HDF5 file for later analysis. Currently, the saved properties
are specified hard coded in include_from_model_in_hdf5. This is a dict where the key corresponds to the
name of the property and the value describes the type. If the value is None the property can be dumped
to hdf via its attribute to_hdf or by converting it to a pd.DataFrame. For more complex properties
which can not simply be dumped to an hdf file the dict can contain a function which is called with
the parameters key, path, and hdf_store. This function then should dump the data to the given
hdf_store object. To dump properties of sub-properties of the model, you can use a dict as value.
This dict is then treated in the same way as described above.
Parameters
----------
buffer_or_fname: buffer or ~str
buffer or filename for HDF5 file (see pandas.HDFStore for description)
path: ~str, optional
path in the HDF5 file
close_h5: ~bool
close the HDF5 file or not.
"""
# Functions to save properties of the model without to_hdf attribute and no simple conversion to a pd.DataFrame.
#This functions are always called with the parameters key, path and, hdf_store.
def _save_luminosity_density(key, path, hdf_store):
luminosity_density = pd.DataFrame.from_dict(
dict(wave=self.spectrum.wavelength.value,
flux=self.spectrum.luminosity_density_lambda.value))
luminosity_density.to_hdf(hdf_store, os.path.join(path, key))
def _save_spectrum_virtual(key, path, hdf_store):
if self.spectrum_virtual.luminosity_density_lambda is not None:
luminosity_density_virtual = pd.DataFrame.from_dict(
dict(wave=self.spectrum_virtual.wavelength.value,
flux=self.spectrum_virtual.luminosity_density_lambda.
value))
luminosity_density_virtual.to_hdf(hdf_store,
os.path.join(path, key))
def _save_configuration_dict(key, path, hdf_store):
configuration_dict = dict(t_inner=self.t_inner.value)
configuration_dict_path = os.path.join(path, 'configuration')
pd.Series(configuration_dict).to_hdf(hdf_store,
configuration_dict_path)
include_from_plasma_ = {
'level_number_density': None,
'ion_number_density': None,
'tau_sobolevs': None,
'electron_densities': None,
't_rad': None,
'w': None
}
include_from_model_in_hdf5 = {
'plasma_array': include_from_plasma_,
'j_blues': None,
'last_line_interaction_in_id': None,
'last_line_interaction_out_id': None,
'last_line_interaction_shell_id': None,
'montecarlo_nu': None,
'luminosity_density': _save_luminosity_density,
'luminosity_density_virtual': _save_spectrum_virtual,
'configuration_dict': _save_configuration_dict,
'last_line_interaction_angstrom': None
}
if isinstance(buffer_or_fname, basestring):
hdf_store = pd.HDFStore(buffer_or_fname)
elif isinstance(buffer_or_fname, pd.HDFStore):
hdf_store = buffer_or_fname
else:
raise IOError('Please specify either a filename or an HDFStore')
logger.info('Writing to path %s', path)
def _get_hdf5_path(path, property_name):
return os.path.join(path, property_name)
def _to_smallest_pandas(object):
try:
return pd.Series(object)
except Exception:
return pd.DataFrame(object)
def _save_model_property(object, property_name, path, hdf_store):
property_path = _get_hdf5_path(path, property_name)
try:
object.to_hdf(hdf_store, property_path)
except AttributeError:
_to_smallest_pandas(object).to_hdf(hdf_store, property_path)
for key in include_from_model_in_hdf5:
if include_from_model_in_hdf5[key] is None:
_save_model_property(getattr(self, key), key, path, hdf_store)
elif callable(include_from_model_in_hdf5[key]):
include_from_model_in_hdf5[key](key, path, hdf_store)
else:
try:
for subkey in include_from_model_in_hdf5[key]:
if include_from_model_in_hdf5[key][subkey] is None:
_save_model_property(
getattr(getattr(self, key), subkey), subkey,
os.path.join(path, key), hdf_store)
elif callable(include_from_model_in_hdf5[key][subkey]):
include_from_model_in_hdf5[key][subkey](
subkey, os.path.join(path, key), hdf_store)
else:
logger.critical(
'Can not save %s',
str(os.path.join(path, key, subkey)))
except:
logger.critical(
'An error occurred while dumping %s to HDF.',
str(os.path.join(path, key)))
hdf_store.flush()
if close_h5:
hdf_store.close()
else:
return hdf_store
class Simulation(object):
converged = False
def __init__(self, tardis_config):
self.tardis_config = tardis_config
self.runner = MontecarloRunner(self.tardis_config.montecarlo.seed,
tardis_config.spectrum.frequency,
tardis_config.supernova.get('distance',
None))
t_inner_lock_cycle = [False] * (tardis_config.montecarlo.
convergence_strategy.
lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
def run_single_montecarlo(self, model, no_of_packets,
no_of_virtual_packets=0,last_run=False):
"""
Will do a single TARDIS iteration with the given model
Parameters
----------
model: ~tardis.model.Radial1DModel
no_of_packet: ~int
no_of_virtual_packets: ~int
default is 0 and switches of the virtual packet mode. Recommended
is 3.
Returns
-------
: None
"""
self.runner.run(model, no_of_packets,
no_of_virtual_packets=no_of_virtual_packets,
nthreads=self.tardis_config.montecarlo.nthreads,last_run=last_run)
(montecarlo_nu, montecarlo_energies, self.j_estimators,
self.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, self.last_interaction_type,
self.last_line_interaction_shell_id) = self.runner.legacy_return()
if np.sum(montecarlo_energies < 0) == len(montecarlo_energies):
logger.critical("No r-packet escaped through the outer boundary.")
def calculate_emitted_luminosity(self):
"""
Returns
-------
"""
return self.runner.calculate_emitted_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def calculate_reabsorbed_luminosity(self):
return self.runner.calculate_reabsorbed_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def estimate_t_inner(self, input_t_inner, luminosity_requested,
t_inner_update_exponent=-0.5):
emitted_luminosity = self.calculate_emitted_luminosity()
luminosity_ratios = (
(emitted_luminosity / luminosity_requested).to(1).value)
return input_t_inner * luminosity_ratios ** t_inner_update_exponent
def get_convergence_status(self, t_rad, w, t_inner, estimated_t_rad, estimated_w,
estimated_t_inner):
convergence_section = self.tardis_config.montecarlo.convergence_strategy
no_of_shells = self.tardis_config.structure.no_of_shells
convergence_t_rad = (abs(t_rad - estimated_t_rad) /
estimated_t_rad).value
convergence_w = (abs(w - estimated_w) / estimated_w)
convergence_t_inner = (abs(t_inner - estimated_t_inner) /
estimated_t_inner).value
if convergence_section.type == 'specific':
fraction_t_rad_converged = (
np.count_nonzero(
convergence_t_rad < convergence_section.t_rad.threshold)
/ no_of_shells)
t_rad_converged = (
fraction_t_rad_converged > convergence_section.t_rad.threshold)
fraction_w_converged = (
np.count_nonzero(
convergence_w < convergence_section.w.threshold)
/ no_of_shells)
w_converged = (
fraction_w_converged > convergence_section.w.threshold)
t_inner_converged = (
convergence_t_inner < convergence_section.t_inner.threshold)
if np.all([t_rad_converged, w_converged, t_inner_converged]):
return True
else:
return False
else:
return False
def log_run_results(self, emitted_luminosity, absorbed_luminosity):
logger.info("Luminosity emitted = {0:.5e} "
"Luminosity absorbed = {1:.5e} "
"Luminosity requested = {2:.5e}".format(
emitted_luminosity, absorbed_luminosity,
self.tardis_config.supernova.luminosity_requested))
def log_plasma_state(self, t_rad, w, t_inner, next_t_rad, next_w,
next_t_inner, log_sampling=5):
"""
Logging the change of the plasma state
Parameters
----------
t_rad: ~astropy.units.Quanity
current t_rad
w: ~astropy.units.Quanity
current w
next_t_rad: ~astropy.units.Quanity
next t_rad
next_w: ~astropy.units.Quanity
next_w
log_sampling: ~int
the n-th shells to be plotted
Returns
-------
"""
plasma_state_log = pd.DataFrame(index=np.arange(len(t_rad)),
columns=['t_rad', 'next_t_rad',
'w', 'next_w'])
plasma_state_log['t_rad'] = t_rad
plasma_state_log['next_t_rad'] = next_t_rad
plasma_state_log['w'] = w
plasma_state_log['next_w'] = next_w
plasma_state_log.index.name = 'Shell'
plasma_state_log = str(plasma_state_log[::log_sampling])
plasma_state_log = ''.join(['\t%s\n' % item for item in
plasma_state_log.split('\n')])
logger.info('Plasma stratification:\n%s\n', plasma_state_log)
logger.info('t_inner {0:.3f} -- next t_inner {1:.3f}'.format(
t_inner, next_t_inner))
@staticmethod
def damped_converge(value, estimated_value, damping_factor):
return value + damping_factor * (estimated_value - value)
def calculate_next_plasma_state(self, t_rad, w, t_inner,
estimated_w, estimated_t_rad,
estimated_t_inner):
convergence_strategy = (
self.tardis_config.montecarlo.convergence_strategy)
if (convergence_strategy.type == 'damped'
or convergence_strategy.type == 'specific'):
next_t_rad = self.damped_converge(
t_rad, estimated_t_rad,
convergence_strategy.t_rad.damping_constant)
next_w = self.damped_converge(
w, estimated_w, convergence_strategy.w.damping_constant)
next_t_inner = self.damped_converge(
t_inner, estimated_t_inner,
convergence_strategy.t_inner.damping_constant)
return next_t_rad, next_w, next_t_inner
else:
raise ValueError('Convergence strategy type is '
'neither damped nor specific '
'- input is {0}'.format(convergence_strategy.type))
def legacy_run_simulation(self, model, hdf_path_or_buf=None,
hdf_mode='full', hdf_last_only=True):
"""
Parameters
----------
model : tardis.model.Radial1DModel
hdf_path_or_buf : str, optional
A path to store the data of each simulation iteration
(the default value is None, which means that nothing
will be stored).
hdf_mode : {'full', 'input'}, optional
If 'full' all plasma properties will be stored to HDF,
if 'input' only input plasma properties will be stored.
hdf_last_only: bool, optional
If True, only the last iteration of the simulation will
be stored to the HDFStore.
Returns
-------
"""
if hdf_path_or_buf is not None:
if hdf_mode == 'full':
plasma_properties = None
elif hdf_mode == 'input':
plasma_properties = [Input]
else:
raise ValueError('hdf_mode must be "full" or "input"'
', not "{}"'.format(type(hdf_mode)))
start_time = time.time()
self.iterations_remaining = self.tardis_config.montecarlo.iterations
self.iterations_max_requested = self.tardis_config.montecarlo.iterations
self.iterations_executed = 0
converged = False
convergence_section = (
self.tardis_config.montecarlo.convergence_strategy)
while self.iterations_remaining > 1:
logger.info('Remaining run %d', self.iterations_remaining)
self.run_single_montecarlo(
model, self.tardis_config.montecarlo.no_of_packets)
self.log_run_results(self.calculate_emitted_luminosity(),
self.calculate_reabsorbed_luminosity())
self.iterations_executed += 1
self.iterations_remaining -= 1
estimated_t_rad, estimated_w = (
self.runner.calculate_radiationfield_properties())
estimated_t_inner = self.estimate_t_inner(
model.t_inner,
self.tardis_config.supernova.luminosity_requested)
converged = self.get_convergence_status(
model.t_rads, model.ws, model.t_inner, estimated_t_rad,
estimated_w, estimated_t_inner)
next_t_rad, next_w, next_t_inner = self.calculate_next_plasma_state(
model.t_rads, model.ws, model.t_inner,
estimated_w, estimated_t_rad, estimated_t_inner)
self.log_plasma_state(model.t_rads, model.ws, model.t_inner,
next_t_rad, next_w, next_t_inner)
model.t_rads = next_t_rad
model.ws = next_w
model.t_inner = next_t_inner
model.j_blue_estimators = self.runner.j_blue_estimator
model.calculate_j_blues(init_detailed_j_blues=False)
model.update_plasmas(initialize_nlte=False)
if hdf_path_or_buf is not None and not hdf_last_only:
self.to_hdf(model, hdf_path_or_buf,
'simulation{}'.format(self.iterations_executed),
plasma_properties)
# if switching into the hold iterations mode or out back to the normal one
# if it is in either of these modes already it will just stay there
if converged and not self.converged:
self.converged = True
# UMN - used to be 'hold_iterations_wrong' but this is
# currently not in the convergence_section namespace...
self.iterations_remaining = (
convergence_section["hold_iterations"])
elif not converged and self.converged:
# UMN Warning: the following two iterations attributes of the Simulation object don't exist
self.iterations_remaining = self.iterations_max_requested - self.iterations_executed
self.converged = False
else:
# either it is converged and the status of the simulation is
# converged OR it is not converged and the status of the
# simulation is not converged - Do nothing.
pass
if converged:
self.iterations_remaining = (
convergence_section["hold_iterations"])
#Finished second to last loop running one more time
logger.info('Doing last run')
if self.tardis_config.montecarlo.last_no_of_packets is not None:
no_of_packets = self.tardis_config.montecarlo.last_no_of_packets
else:
no_of_packets = self.tardis_config.montecarlo.no_of_packets
no_of_virtual_packets = (
self.tardis_config.montecarlo.no_of_virtual_packets)
self.run_single_montecarlo(model, no_of_packets, no_of_virtual_packets, last_run=True)
self.runner.legacy_update_spectrum(no_of_virtual_packets)
self.legacy_set_final_model_properties(model)
model.Edotlu_estimators = self.runner.Edotlu_estimator
#the following instructions, passing down information to the model are
#required for the gui
model.no_of_packets = no_of_packets
model.no_of_virtual_packets = no_of_virtual_packets
model.converged = converged
model.iterations_executed = self.iterations_executed
model.iterations_max_requested = self.iterations_max_requested
logger.info("Finished in {0:d} iterations and took {1:.2f} s".format(
self.iterations_executed, time.time()-start_time))
if hdf_path_or_buf is not None:
if hdf_last_only:
name = 'simulation'
else:
name = 'simulation{}'.format(self.iterations_executed)
self.to_hdf(model, hdf_path_or_buf, name, plasma_properties)
def legacy_set_final_model_properties(self, model):
"""Sets additional model properties to be compatible with old model design
The runner object is given to the model and other packet diagnostics are set.
Parameters
----------
model: ~tardis.model.Radial1DModel
Returns
-------
: None
"""
#pass the runner to the model
model.runner = self.runner
#TODO: pass packet diagnostic arrays
(montecarlo_nu, montecarlo_energies, model.j_estimators,
model.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, model.last_interaction_type,
model.last_line_interaction_shell_id) = model.runner.legacy_return()
model.montecarlo_nu = self.runner.output_nu
model.montecarlo_luminosity = self.runner.packet_luminosity
model.last_line_interaction_in_id = model.atom_data.lines_index.index.values[last_line_interaction_in_id]
model.last_line_interaction_in_id = model.last_line_interaction_in_id[last_line_interaction_in_id != -1]
model.last_line_interaction_out_id = model.atom_data.lines_index.index.values[last_line_interaction_out_id]
model.last_line_interaction_out_id = model.last_line_interaction_out_id[last_line_interaction_out_id != -1]
model.last_line_interaction_angstrom = model.montecarlo_nu[last_line_interaction_in_id != -1].to('angstrom',
u.spectral())
# required for gui
model.current_no_of_packets = model.tardis_config.montecarlo.no_of_packets
def to_hdf(self, model, path_or_buf, path='', plasma_properties=None):
"""
Store the simulation to an HDF structure.
Parameters
----------
model : tardis.model.Radial1DModel
path_or_buf
Path or buffer to the HDF store
path : str
Path inside the HDF store to store the simulation
plasma_properties
`None` or a `PlasmaPropertyCollection` which will
be passed as the collection argument to the
plasma.to_hdf method.
Returns
-------
None
"""
self.runner.to_hdf(path_or_buf, path)
model.to_hdf(path_or_buf, path, plasma_properties)
class Simulation(object):
converged = False
def __init__(self, tardis_config):
self.tardis_config = tardis_config
self.runner = MontecarloRunner(self.tardis_config.montecarlo.seed,
tardis_config.spectrum.frequency)
t_inner_lock_cycle = [False] * (tardis_config.montecarlo.
convergence_strategy.
lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
def run_single_montecarlo(self, model, no_of_packets,
no_of_virtual_packets=0):
"""
Will do a single TARDIS iteration with the given model
Parameters
----------
model: ~tardis.model.Radial1DModel
no_of_packet: ~int
no_of_virtual_packets: ~int
default is 0 and switches of the virtual packet mode. Recommended
is 3.
Returns
-------
: None
"""
self.runner.run(model, no_of_packets,
no_of_virtual_packets=no_of_virtual_packets,
nthreads=self.tardis_config.montecarlo.nthreads)
(montecarlo_nu, montecarlo_energies, self.j_estimators,
self.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, self.last_interaction_type,
self.last_line_interaction_shell_id) = self.runner.legacy_return()
if np.sum(montecarlo_energies < 0) == len(montecarlo_energies):
logger.critical("No r-packet escaped through the outer boundary.")
def calculate_emitted_luminosity(self):
"""
Returns
-------
"""
return self.runner.calculate_emitted_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def calculate_reabsorbed_luminosity(self):
return self.runner.calculate_reabsorbed_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def estimate_t_inner(self, input_t_inner, luminosity_requested,
t_inner_update_exponent=-0.5):
emitted_luminosity = self.calculate_emitted_luminosity()
luminosity_ratios = (
(emitted_luminosity / luminosity_requested).to(1).value)
return input_t_inner * luminosity_ratios ** t_inner_update_exponent
def get_convergence_status(self, t_rad, w, t_inner, estimated_t_rad, estimated_w,
estimated_t_inner):
convergence_section = self.tardis_config.montecarlo.convergence_strategy
no_of_shells = self.tardis_config.structure.no_of_shells
convergence_t_rad = (abs(t_rad - estimated_t_rad) /
estimated_t_rad).value
convergence_w = (abs(w - estimated_w) / estimated_w)
convergence_t_inner = (abs(t_inner - estimated_t_inner) /
estimated_t_inner).value
if convergence_section.type == 'specific':
fraction_t_rad_converged = (
np.count_nonzero(
convergence_t_rad < convergence_section.t_rad.threshold)
/ no_of_shells)
t_rad_converged = (
fraction_t_rad_converged > convergence_section.t_rad.threshold)
fraction_w_converged = (
np.count_nonzero(
convergence_w < convergence_section.w.threshold)
/ no_of_shells)
w_converged = (
fraction_w_converged > convergence_section.w.threshold)
t_inner_converged = (
convergence_t_inner < convergence_section.t_inner.threshold)
if np.all([t_rad_converged, w_converged, t_inner_converged]):
return True
else:
return False
else:
return False
def log_run_results(self, emitted_luminosity, absorbed_luminosity):
logger.info("Luminosity emitted = {0:.5e} "
"Luminosity absorbed = {1:.5e} "
"Luminosity requested = {2:.5e}".format(
emitted_luminosity, absorbed_luminosity,
self.tardis_config.supernova.luminosity_requested))
def log_plasma_state(self, t_rad, w, t_inner, next_t_rad, next_w,
next_t_inner, log_sampling=5):
"""
Logging the change of the plasma state
Parameters
----------
t_rad: ~astropy.units.Quanity
current t_rad
w: ~astropy.units.Quanity
current w
next_t_rad: ~astropy.units.Quanity
next t_rad
next_w: ~astropy.units.Quanity
next_w
log_sampling: ~int
the n-th shells to be plotted
Returns
-------
"""
plasma_state_log = pd.DataFrame(index=np.arange(len(t_rad)),
columns=['t_rad', 'next_t_rad',
'w', 'next_w'])
plasma_state_log['t_rad'] = t_rad
plasma_state_log['next_t_rad'] = next_t_rad
plasma_state_log['w'] = w
plasma_state_log['next_w'] = next_w
plasma_state_log.index.name = 'Shell'
plasma_state_log = str(plasma_state_log[::log_sampling])
plasma_state_log = ''.join(['\t%s\n' % item for item in
plasma_state_log.split('\n')])
logger.info('Plasma stratification:\n%s\n', plasma_state_log)
logger.info('t_inner {0:.3f} -- next t_inner {1:.3f}'.format(
t_inner, next_t_inner))
@staticmethod
def damped_converge(value, estimated_value, damping_factor):
return value + damping_factor * (estimated_value - value)
def calculate_next_plasma_state(self, t_rad, w, t_inner,
estimated_w, estimated_t_rad,
estimated_t_inner):
convergence_strategy = (
self.tardis_config.montecarlo.convergence_strategy)
if (convergence_strategy.type == 'damped'
or convergence_strategy.type == 'specific'):
next_t_rad = self.damped_converge(
t_rad, estimated_t_rad,
convergence_strategy.t_rad.damping_constant)
next_w = self.damped_converge(
w, estimated_w, convergence_strategy.w.damping_constant)
next_t_inner = self.damped_converge(
t_inner, estimated_t_inner,
convergence_strategy.t_inner.damping_constant)
return next_t_rad, next_w, next_t_inner
else:
raise ValueError('Convergence strategy type is '
'neither damped nor specific '
'- input is {0}'.format(convergence_strategy.type))
def legacy_run_simulation(self, model):
start_time = time.time()
iterations_remaining = self.tardis_config.montecarlo.iterations
iterations_max_requested = self.tardis_config.montecarlo.iterations
iterations_executed = 0
converged = False
while iterations_remaining > 1:
logger.info('Remaining run %d', iterations_remaining)
self.run_single_montecarlo(
model, self.tardis_config.montecarlo.no_of_packets)
self.log_run_results(self.calculate_emitted_luminosity(),
self.calculate_reabsorbed_luminosity())
iterations_executed += 1
iterations_remaining -= 1
estimated_t_rad, estimated_w = (
self.runner.calculate_radiationfield_properties())
estimated_t_inner = self.estimate_t_inner(
model.t_inner,
self.tardis_config.supernova.luminosity_requested)
converged = self.get_convergence_status(
model.t_rads, model.ws, model.t_inner, estimated_t_rad,
estimated_w, estimated_t_inner)
next_t_rad, next_w, next_t_inner = self.calculate_next_plasma_state(
model.t_rads, model.ws, model.t_inner,
estimated_w, estimated_t_rad, estimated_t_inner)
self.log_plasma_state(model.t_rads, model.ws, model.t_inner,
next_t_rad, next_w, next_t_inner)
model.t_rads = next_t_rad
model.ws = next_w
model.t_inner = next_t_inner
model.j_blue_estimators = self.runner.j_blue_estimator
model.calculate_j_blues(init_detailed_j_blues=False)
model.update_plasmas(initialize_nlte=False)
# if switching into the hold iterations mode or out back to the normal one
# if it is in either of these modes already it will just stay there
if converged and not self.converged:
self.converged = True
iterations_remaining = (
convergence_section.global_convergence_parameters.
hold_iterations_wrong)
elif not converged and self.converged:
# UMN Warning: the following two iterations attributes of the Simulation object don't exist
self.iterations_remaining = self.iterations_max_requested - self.iterations_executed
self.converged = False
else:
# either it is converged and the status of the simulation is
# converged OR it is not converged and the status of the
# simulation is not converged - Do nothing.
pass
if converged:
convergence_section = (
self.tardis_config.montecarlo.convergence_strategy)
iterations_remaining = (
convergence_section.global_convergence_parameters.
hold_iterations)
#Finished second to last loop running one more time
logger.info('Doing last run')
if self.tardis_config.montecarlo.last_no_of_packets is not None:
no_of_packets = self.tardis_config.montecarlo.last_no_of_packets
else:
no_of_packets = self.tardis_config.montecarlo.no_of_packets
no_of_virtual_packets = (
self.tardis_config.montecarlo.no_of_virtual_packets)
self.run_single_montecarlo(model, no_of_packets, no_of_virtual_packets)
self.legacy_update_spectrum(model, no_of_virtual_packets)
self.legacy_set_final_model_properties(model)
#the following instructions, passing down information to the model are
#required for the gui
model.no_of_packets = no_of_packets
model.no_of_virtual_packets = no_of_virtual_packets
model.converged = converged
model.iterations_executed = iterations_executed
model.iterations_max_requested = iterations_max_requested
logger.info("Finished in {0:d} iterations and took {1:.2f} s".format(
iterations_executed, time.time()-start_time))
def legacy_update_spectrum(self, model, no_of_virtual_packets):
montecarlo_reabsorbed_luminosity = np.histogram(
self.runner.reabsorbed_packet_nu,
weights=self.runner.reabsorbed_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
montecarlo_emitted_luminosity = np.histogram(
self.runner.emitted_packet_nu,
weights=self.runner.emitted_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
model.spectrum.update_luminosity(montecarlo_emitted_luminosity)
model.spectrum_reabsorbed.update_luminosity(montecarlo_reabsorbed_luminosity)
if no_of_virtual_packets > 0:
model.montecarlo_virtual_luminosity = (
self.runner.legacy_montecarlo_virtual_luminosity *
1 * u.erg / model.time_of_simulation)[:-1]
model.spectrum_virtual.update_luminosity(
model.montecarlo_virtual_luminosity)
def legacy_set_final_model_properties(self, model):
"""Sets additional model properties to be compatible with old model design
The runner object is given to the model and other packet diagnostics are set.
Parameters
----------
model: ~tardis.model.Radial1DModel
Returns
-------
: None
"""
#pass the runner to the model
model.runner = self.runner
#TODO: pass packet diagnostic arrays
(montecarlo_nu, montecarlo_energies, model.j_estimators,
model.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, model.last_interaction_type,
model.last_line_interaction_shell_id) = model.runner.legacy_return()
model.montecarlo_nu = self.runner.output_nu
model.montecarlo_luminosity = self.runner.packet_luminosity
model.last_line_interaction_in_id = model.atom_data.lines_index.index.values[last_line_interaction_in_id]
model.last_line_interaction_in_id = model.last_line_interaction_in_id[last_line_interaction_in_id != -1]
model.last_line_interaction_out_id = model.atom_data.lines_index.index.values[last_line_interaction_out_id]
model.last_line_interaction_out_id = model.last_line_interaction_out_id[last_line_interaction_out_id != -1]
model.last_line_interaction_angstrom = model.montecarlo_nu[last_line_interaction_in_id != -1].to('angstrom',
u.spectral())
# required for gui
model.current_no_of_packets = model.tardis_config.montecarlo.no_of_packets
class Simulation(object):
converged = False
def __init__(self, tardis_config):
self.tardis_config = tardis_config
self.runner = MontecarloRunner(self.tardis_config.montecarlo.seed,
tardis_config.spectrum.frequency)
t_inner_lock_cycle = [False] * (tardis_config.montecarlo.
convergence_strategy.
lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
def run_single_montecarlo(self, model, no_of_packets,
no_of_virtual_packets=0):
"""
Will do a single TARDIS iteration with the given model
Parameters
----------
model: ~tardis.model.Radial1DModel
no_of_packet: ~int
no_of_virtual_packets: ~int
default is 0 and switches of the virtual packet mode. Recommended
is 3.
Returns
-------
: None
"""
self.runner.run(model, no_of_packets,
no_of_virtual_packets=no_of_virtual_packets,
nthreads=self.tardis_config.montecarlo.nthreads)
(montecarlo_nu, montecarlo_energies, self.j_estimators,
self.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, self.last_interaction_type,
self.last_line_interaction_shell_id) = self.runner.legacy_return()
if np.sum(montecarlo_energies < 0) == len(montecarlo_energies):
logger.critical("No r-packet escaped through the outer boundary.")
def calculate_emitted_luminosity(self):
"""
Returns
-------
"""
return self.runner.calculate_emitted_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def calculate_reabsorbed_luminosity(self):
return self.runner.calculate_reabsorbed_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def estimate_t_inner(self, input_t_inner, luminosity_requested,
t_inner_update_exponent=-0.5):
emitted_luminosity = self.calculate_emitted_luminosity()
luminosity_ratios = (
(emitted_luminosity / luminosity_requested).to(1).value)
return input_t_inner * luminosity_ratios ** t_inner_update_exponent
def get_convergence_status(self, t_rad, w, t_inner, estimated_t_rad, estimated_w,
estimated_t_inner):
convergence_section = self.tardis_config.montecarlo.convergence_strategy
no_of_shells = self.tardis_config.structure.no_of_shells
convergence_t_rad = (abs(t_rad - estimated_t_rad) /
estimated_t_rad).value
convergence_w = (abs(w - estimated_w) / estimated_w)
convergence_t_inner = (abs(t_inner - estimated_t_inner) /
estimated_t_inner).value
if convergence_section.type == 'specific':
fraction_t_rad_converged = (
np.count_nonzero(
convergence_t_rad < convergence_section.t_rad.threshold)
/ no_of_shells)
t_rad_converged = (
fraction_t_rad_converged > convergence_section.t_rad.threshold)
fraction_w_converged = (
np.count_nonzero(
convergence_w < convergence_section.w.threshold)
/ no_of_shells)
w_converged = (
fraction_w_converged > convergence_section.w.threshold)
t_inner_converged = (
convergence_t_inner < convergence_section.t_inner.threshold)
if np.all([t_rad_converged, w_converged, t_inner_converged]):
return True
else:
return False
else:
return False
def log_run_results(self, emitted_luminosity, absorbed_luminosity):
logger.info("Luminosity emitted = {0:.5e} "
"Luminosity absorbed = {1:.5e} "
"Luminosity requested = {2:.5e}".format(
emitted_luminosity, absorbed_luminosity,
self.tardis_config.supernova.luminosity_requested))
def log_plasma_state(self, t_rad, w, t_inner, next_t_rad, next_w,
next_t_inner, log_sampling=5):
"""
Logging the change of the plasma state
Parameters
----------
t_rad: ~astropy.units.Quanity
current t_rad
w: ~astropy.units.Quanity
current w
next_t_rad: ~astropy.units.Quanity
next t_rad
next_w: ~astropy.units.Quanity
next_w
log_sampling: ~int
the n-th shells to be plotted
Returns
-------
"""
plasma_state_log = pd.DataFrame(index=np.arange(len(t_rad)),
columns=['t_rad', 'next_t_rad',
'w', 'next_w'])
plasma_state_log['t_rad'] = t_rad
plasma_state_log['next_t_rad'] = next_t_rad
plasma_state_log['w'] = w
plasma_state_log['next_w'] = next_w
plasma_state_log.index.name = 'Shell'
plasma_state_log = str(plasma_state_log[::log_sampling])
plasma_state_log = ''.join(['\t%s\n' % item for item in
plasma_state_log.split('\n')])
logger.info('Plasma stratification:\n%s\n', plasma_state_log)
logger.info('t_inner {0:.3f} -- next t_inner {1:.3f}'.format(
t_inner, next_t_inner))
@staticmethod
def damped_converge(value, estimated_value, damping_factor):
return value + damping_factor * (estimated_value - value)
def calculate_next_plasma_state(self, t_rad, w, t_inner,
estimated_w, estimated_t_rad,
estimated_t_inner):
convergence_strategy = (
self.tardis_config.montecarlo.convergence_strategy)
if (convergence_strategy.type == 'damped'
or convergence_strategy.type == 'specific'):
next_t_rad = self.damped_converge(
t_rad, estimated_t_rad,
convergence_strategy.t_rad.damping_constant)
next_w = self.damped_converge(
w, estimated_w, convergence_strategy.w.damping_constant)
next_t_inner = self.damped_converge(
t_inner, estimated_t_inner,
convergence_strategy.t_inner.damping_constant)
return next_t_rad, next_w, next_t_inner
else:
raise ValueError('Convergence strategy type is '
'neither damped nor specific '
'- input is {0}'.format(convergence_strategy.type))
def legacy_run_simulation(self, model):
start_time = time.time()
iterations_remaining = self.tardis_config.montecarlo.iterations
iterations_max_requested = self.tardis_config.montecarlo.iterations
iterations_executed = 0
converged = False
while iterations_remaining > 1:
logger.info('Remaining run %d', iterations_remaining)
self.run_single_montecarlo(
model, self.tardis_config.montecarlo.no_of_packets)
self.log_run_results(self.calculate_emitted_luminosity(),
self.calculate_reabsorbed_luminosity())
iterations_executed += 1
iterations_remaining -= 1
estimated_t_rad, estimated_w = (
self.runner.calculate_radiationfield_properties())
estimated_t_inner = self.estimate_t_inner(
model.t_inner,
self.tardis_config.supernova.luminosity_requested)
converged = self.get_convergence_status(
model.t_rads, model.ws, model.t_inner, estimated_t_rad,
estimated_w, estimated_t_inner)
next_t_rad, next_w, next_t_inner = self.calculate_next_plasma_state(
model.t_rads, model.ws, model.t_inner,
estimated_w, estimated_t_rad, estimated_t_inner)
self.log_plasma_state(model.t_rads, model.ws, model.t_inner,
next_t_rad, next_w, next_t_inner)
model.t_rads = next_t_rad
model.ws = next_w
model.t_inner = next_t_inner
model.calculate_j_blues(init_detailed_j_blues=False)
model.update_plasmas(initialize_nlte=False)
# if switching into the hold iterations mode or out back to the normal one
# if it is in either of these modes already it will just stay there
if converged and not self.converged:
self.converged = True
iterations_remaining = (
convergence_section.global_convergence_parameters.
hold_iterations_wrong)
elif not converged and self.converged:
# UMN Warning: the following two iterations attributes of the Simulation object don't exist
self.iterations_remaining = self.iterations_max_requested - self.iterations_executed
self.converged = False
else:
# either it is converged and the status of the simulation is
# converged OR it is not converged and the status of the
# simulation is not converged - Do nothing.
pass
if converged:
convergence_section = (
self.tardis_config.montecarlo.convergence_strategy)
iterations_remaining = (
convergence_section.global_convergence_parameters.
hold_iterations)
#Finished second to last loop running one more time
logger.info('Doing last run')
if self.tardis_config.montecarlo.last_no_of_packets is not None:
no_of_packets = self.tardis_config.montecarlo.last_no_of_packets
else:
no_of_packets = self.tardis_config.montecarlo.no_of_packets
no_of_virtual_packets = (
self.tardis_config.montecarlo.no_of_virtual_packets)
self.run_single_montecarlo(model, no_of_packets, no_of_virtual_packets)
self.legacy_update_spectrum(model, no_of_virtual_packets)
self.legacy_set_final_model_properties(model)
#the following instructions, passing down information to the model are
#required for the gui
model.no_of_packets = no_of_packets
model.no_of_virtual_packets = no_of_virtual_packets
model.converged = converged
model.iterations_executed = iterations_executed
model.iterations_max_requested = iterations_max_requested
logger.info("Finished in {0:d} iterations and took {1:.2f} s".format(
iterations_executed, time.time()-start_time))
def legacy_update_spectrum(self, model, no_of_virtual_packets):
montecarlo_reabsorbed_luminosity = np.histogram(
self.runner.reabsorbed_packet_nu,
weights=self.runner.reabsorbed_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
montecarlo_emitted_luminosity = np.histogram(
self.runner.emitted_packet_nu,
weights=self.runner.emitted_packet_luminosity,
bins=self.tardis_config.spectrum.frequency.value)[0] * u.erg / u.s
model.spectrum.update_luminosity(montecarlo_emitted_luminosity)
model.spectrum_reabsorbed.update_luminosity(montecarlo_reabsorbed_luminosity)
if no_of_virtual_packets > 0:
model.montecarlo_virtual_luminosity = (
self.runner.legacy_montecarlo_virtual_luminosity *
1 * u.erg / model.time_of_simulation)[:-1]
model.spectrum_virtual.update_luminosity(
model.montecarlo_virtual_luminosity)
def legacy_set_final_model_properties(self, model):
"""Sets additional model properties to be compatible with old model design
The runner object is given to the model and other packet diagnostics are set.
Parameters
----------
model: ~tardis.model.Radial1DModel
Returns
-------
: None
"""
#pass the runner to the model
model.runner = self.runner
#TODO: pass packet diagnostic arrays
(montecarlo_nu, montecarlo_energies, model.j_estimators,
model.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, model.last_interaction_type,
model.last_line_interaction_shell_id) = model.runner.legacy_return()
model.montecarlo_nu = self.runner.output_nu
model.montecarlo_luminosity = self.runner.packet_luminosity
model.last_line_interaction_in_id = model.atom_data.lines_index.index.values[last_line_interaction_in_id]
model.last_line_interaction_in_id = model.last_line_interaction_in_id[last_line_interaction_in_id != -1]
model.last_line_interaction_out_id = model.atom_data.lines_index.index.values[last_line_interaction_out_id]
model.last_line_interaction_out_id = model.last_line_interaction_out_id[last_line_interaction_out_id != -1]
model.last_line_interaction_angstrom = model.montecarlo_nu[last_line_interaction_in_id != -1].to('angstrom',
u.spectral())
# required for gui
model.current_no_of_packets = model.tardis_config.montecarlo.no_of_packets
def from_config(cls, config, **kwargs):
"""
Create a new Simulation instance from a Configuration object.
Parameters
----------
config : tardis.io.config_reader.Configuration
**kwargs
Allow overriding some structures, such as model, plasma, atomic data
and the runner, instead of creating them from the configuration
object.
Returns
-------
Simulation
"""
# Allow overriding some config structures. This is useful in some
# unit tests, and could be extended in all the from_config classmethods.
if 'model' in kwargs:
model = kwargs['model']
else:
model = Radial1DModel.from_config(config)
if 'plasma' in kwargs:
plasma = kwargs['plasma']
else:
plasma = assemble_plasma(config,
model,
atom_data=kwargs.get('atom_data', None))
if 'runner' in kwargs:
runner = kwargs['runner']
else:
runner = MontecarloRunner.from_config(config)
luminosity_nu_start = config.supernova.luminosity_wavelength_end.to(
u.Hz, u.spectral())
try:
luminosity_nu_end = config.supernova.luminosity_wavelength_start.to(
u.Hz, u.spectral())
except ZeroDivisionError:
luminosity_nu_end = np.inf * u.Hz
last_no_of_packets = config.montecarlo.last_no_of_packets
if last_no_of_packets is None or last_no_of_packets < 0:
last_no_of_packets = config.montecarlo.no_of_packets
last_no_of_packets = int(last_no_of_packets)
return cls(
iterations=config.montecarlo.iterations,
model=model,
plasma=plasma,
runner=runner,
no_of_packets=int(config.montecarlo.no_of_packets),
no_of_virtual_packets=int(config.montecarlo.no_of_virtual_packets),
luminosity_nu_start=luminosity_nu_start,
luminosity_nu_end=luminosity_nu_end,
last_no_of_packets=last_no_of_packets,
luminosity_requested=config.supernova.luminosity_requested.cgs,
convergence_strategy=config.montecarlo.convergence_strategy,
nthreads=config.montecarlo.nthreads)
class Simulation(object):
converged = False
def __init__(self, tardis_config):
self.tardis_config = tardis_config
self.runner = MontecarloRunner(
self.tardis_config.montecarlo.seed,
tardis_config.spectrum.frequency,
tardis_config.supernova.get('distance', None))
t_inner_lock_cycle = [False] * (
tardis_config.montecarlo.convergence_strategy.lock_t_inner_cycles)
t_inner_lock_cycle[0] = True
self.t_inner_update = itertools.cycle(t_inner_lock_cycle)
def run_single_montecarlo(self,
model,
no_of_packets,
no_of_virtual_packets=0,
last_run=False):
"""
Will do a single TARDIS iteration with the given model
Parameters
----------
model: ~tardis.model.Radial1DModel
no_of_packet: ~int
no_of_virtual_packets: ~int
default is 0 and switches of the virtual packet mode. Recommended
is 3.
Returns
-------
: None
"""
self.runner.run(model,
no_of_packets,
no_of_virtual_packets=no_of_virtual_packets,
nthreads=self.tardis_config.montecarlo.nthreads,
last_run=last_run)
(montecarlo_nu, montecarlo_energies, self.j_estimators,
self.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, self.last_interaction_type,
self.last_line_interaction_shell_id) = self.runner.legacy_return()
if np.sum(montecarlo_energies < 0) == len(montecarlo_energies):
logger.critical("No r-packet escaped through the outer boundary.")
def calculate_emitted_luminosity(self):
"""
Returns
-------
"""
return self.runner.calculate_emitted_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def calculate_reabsorbed_luminosity(self):
return self.runner.calculate_reabsorbed_luminosity(
self.tardis_config.supernova.luminosity_nu_start,
self.tardis_config.supernova.luminosity_nu_end)
def estimate_t_inner(self,
input_t_inner,
luminosity_requested,
t_inner_update_exponent=-0.5):
emitted_luminosity = self.calculate_emitted_luminosity()
luminosity_ratios = ((emitted_luminosity /
luminosity_requested).to(1).value)
return input_t_inner * luminosity_ratios**t_inner_update_exponent
def get_convergence_status(self, t_rad, w, t_inner, estimated_t_rad,
estimated_w, estimated_t_inner):
convergence_section = self.tardis_config.montecarlo.convergence_strategy
no_of_shells = self.tardis_config.structure.no_of_shells
convergence_t_rad = (abs(t_rad - estimated_t_rad) /
estimated_t_rad).value
convergence_w = (abs(w - estimated_w) / estimated_w)
convergence_t_inner = (abs(t_inner - estimated_t_inner) /
estimated_t_inner).value
if convergence_section.type == 'specific':
fraction_t_rad_converged = (np.count_nonzero(
convergence_t_rad < convergence_section.t_rad.threshold) /
no_of_shells)
t_rad_converged = (fraction_t_rad_converged >
convergence_section.t_rad.threshold)
fraction_w_converged = (np.count_nonzero(
convergence_w < convergence_section.w.threshold) /
no_of_shells)
w_converged = (fraction_w_converged >
convergence_section.w.threshold)
t_inner_converged = (convergence_t_inner <
convergence_section.t_inner.threshold)
if np.all([t_rad_converged, w_converged, t_inner_converged]):
return True
else:
return False
else:
return False
def log_run_results(self, emitted_luminosity, absorbed_luminosity):
logger.info("Luminosity emitted = {0:.5e} "
"Luminosity absorbed = {1:.5e} "
"Luminosity requested = {2:.5e}".format(
emitted_luminosity, absorbed_luminosity,
self.tardis_config.supernova.luminosity_requested))
def log_plasma_state(self,
t_rad,
w,
t_inner,
next_t_rad,
next_w,
next_t_inner,
log_sampling=5):
"""
Logging the change of the plasma state
Parameters
----------
t_rad: ~astropy.units.Quanity
current t_rad
w: ~astropy.units.Quanity
current w
next_t_rad: ~astropy.units.Quanity
next t_rad
next_w: ~astropy.units.Quanity
next_w
log_sampling: ~int
the n-th shells to be plotted
Returns
-------
"""
plasma_state_log = pd.DataFrame(
index=np.arange(len(t_rad)),
columns=['t_rad', 'next_t_rad', 'w', 'next_w'])
plasma_state_log['t_rad'] = t_rad
plasma_state_log['next_t_rad'] = next_t_rad
plasma_state_log['w'] = w
plasma_state_log['next_w'] = next_w
plasma_state_log.index.name = 'Shell'
plasma_state_log = str(plasma_state_log[::log_sampling])
plasma_state_log = ''.join(
['\t%s\n' % item for item in plasma_state_log.split('\n')])
logger.info('Plasma stratification:\n%s\n', plasma_state_log)
logger.info('t_inner {0:.3f} -- next t_inner {1:.3f}'.format(
t_inner, next_t_inner))
@staticmethod
def damped_converge(value, estimated_value, damping_factor):
return value + damping_factor * (estimated_value - value)
def calculate_next_plasma_state(self, t_rad, w, t_inner, estimated_w,
estimated_t_rad, estimated_t_inner):
convergence_strategy = (
self.tardis_config.montecarlo.convergence_strategy)
if (convergence_strategy.type == 'damped'
or convergence_strategy.type == 'specific'):
next_t_rad = self.damped_converge(
t_rad, estimated_t_rad,
convergence_strategy.t_rad.damping_constant)
next_w = self.damped_converge(
w, estimated_w, convergence_strategy.w.damping_constant)
next_t_inner = self.damped_converge(
t_inner, estimated_t_inner,
convergence_strategy.t_inner.damping_constant)
return next_t_rad, next_w, next_t_inner
else:
raise ValueError('Convergence strategy type is '
'neither damped nor specific '
'- input is {0}'.format(
convergence_strategy.type))
def legacy_run_simulation(self,
model,
hdf_path_or_buf=None,
hdf_mode='full',
hdf_last_only=True):
"""
Parameters
----------
model : tardis.model.Radial1DModel
hdf_path_or_buf : str, optional
A path to store the data of each simulation iteration
(the default value is None, which means that nothing
will be stored).
hdf_mode : {'full', 'input'}, optional
If 'full' all plasma properties will be stored to HDF,
if 'input' only input plasma properties will be stored.
hdf_last_only: bool, optional
If True, only the last iteration of the simulation will
be stored to the HDFStore.
Returns
-------
"""
if hdf_path_or_buf is not None:
if hdf_mode == 'full':
plasma_properties = None
elif hdf_mode == 'input':
plasma_properties = [Input]
else:
raise ValueError('hdf_mode must be "full" or "input"'
', not "{}"'.format(type(hdf_mode)))
start_time = time.time()
self.iterations_remaining = self.tardis_config.montecarlo.iterations
self.iterations_max_requested = self.tardis_config.montecarlo.iterations
self.iterations_executed = 0
converged = False
convergence_section = (
self.tardis_config.montecarlo.convergence_strategy)
while self.iterations_remaining > 1:
logger.info('Remaining run %d', self.iterations_remaining)
self.run_single_montecarlo(
model, self.tardis_config.montecarlo.no_of_packets)
self.log_run_results(self.calculate_emitted_luminosity(),
self.calculate_reabsorbed_luminosity())
self.iterations_executed += 1
self.iterations_remaining -= 1
estimated_t_rad, estimated_w = (
self.runner.calculate_radiationfield_properties())
estimated_t_inner = self.estimate_t_inner(
model.t_inner,
self.tardis_config.supernova.luminosity_requested)
converged = self.get_convergence_status(model.t_rads, model.ws,
model.t_inner,
estimated_t_rad,
estimated_w,
estimated_t_inner)
next_t_rad, next_w, next_t_inner = self.calculate_next_plasma_state(
model.t_rads, model.ws, model.t_inner, estimated_w,
estimated_t_rad, estimated_t_inner)
self.log_plasma_state(model.t_rads, model.ws, model.t_inner,
next_t_rad, next_w, next_t_inner)
model.t_rads = next_t_rad
model.ws = next_w
model.t_inner = next_t_inner
model.j_blue_estimators = self.runner.j_blue_estimator
model.calculate_j_blues(init_detailed_j_blues=False)
model.update_plasmas(initialize_nlte=False)
if hdf_path_or_buf is not None and not hdf_last_only:
self.to_hdf(model, hdf_path_or_buf,
'simulation{}'.format(self.iterations_executed),
plasma_properties)
# if switching into the hold iterations mode or out back to the normal one
# if it is in either of these modes already it will just stay there
if converged and not self.converged:
self.converged = True
# UMN - used to be 'hold_iterations_wrong' but this is
# currently not in the convergence_section namespace...
self.iterations_remaining = (
convergence_section["hold_iterations"])
elif not converged and self.converged:
# UMN Warning: the following two iterations attributes of the Simulation object don't exist
self.iterations_remaining = self.iterations_max_requested - self.iterations_executed
self.converged = False
else:
# either it is converged and the status of the simulation is
# converged OR it is not converged and the status of the
# simulation is not converged - Do nothing.
pass
if converged:
self.iterations_remaining = (
convergence_section["hold_iterations"])
#Finished second to last loop running one more time
logger.info('Doing last run')
if self.tardis_config.montecarlo.last_no_of_packets is not None:
no_of_packets = self.tardis_config.montecarlo.last_no_of_packets
else:
no_of_packets = self.tardis_config.montecarlo.no_of_packets
no_of_virtual_packets = (
self.tardis_config.montecarlo.no_of_virtual_packets)
self.run_single_montecarlo(model,
no_of_packets,
no_of_virtual_packets,
last_run=True)
self.runner.legacy_update_spectrum(no_of_virtual_packets)
self.legacy_set_final_model_properties(model)
model.Edotlu_estimators = self.runner.Edotlu_estimator
#the following instructions, passing down information to the model are
#required for the gui
model.no_of_packets = no_of_packets
model.no_of_virtual_packets = no_of_virtual_packets
model.converged = converged
model.iterations_executed = self.iterations_executed
model.iterations_max_requested = self.iterations_max_requested
logger.info("Finished in {0:d} iterations and took {1:.2f} s".format(
self.iterations_executed,
time.time() - start_time))
if hdf_path_or_buf is not None:
if hdf_last_only:
name = 'simulation'
else:
name = 'simulation{}'.format(self.iterations_executed)
self.to_hdf(model, hdf_path_or_buf, name, plasma_properties)
def legacy_set_final_model_properties(self, model):
"""Sets additional model properties to be compatible with old model design
The runner object is given to the model and other packet diagnostics are set.
Parameters
----------
model: ~tardis.model.Radial1DModel
Returns
-------
: None
"""
#pass the runner to the model
model.runner = self.runner
#TODO: pass packet diagnostic arrays
(montecarlo_nu, montecarlo_energies, model.j_estimators,
model.nubar_estimators, last_line_interaction_in_id,
last_line_interaction_out_id, model.last_interaction_type,
model.last_line_interaction_shell_id) = model.runner.legacy_return()
model.montecarlo_nu = self.runner.output_nu
model.montecarlo_luminosity = self.runner.packet_luminosity
model.last_line_interaction_in_id = model.atom_data.lines_index.index.values[
last_line_interaction_in_id]
model.last_line_interaction_in_id = model.last_line_interaction_in_id[
last_line_interaction_in_id != -1]
model.last_line_interaction_out_id = model.atom_data.lines_index.index.values[
last_line_interaction_out_id]
model.last_line_interaction_out_id = model.last_line_interaction_out_id[
last_line_interaction_out_id != -1]
model.last_line_interaction_angstrom = model.montecarlo_nu[
last_line_interaction_in_id != -1].to('angstrom', u.spectral())
# required for gui
model.current_no_of_packets = model.tardis_config.montecarlo.no_of_packets
def to_hdf(self, model, path_or_buf, path='', plasma_properties=None):
"""
Store the simulation to an HDF structure.
Parameters
----------
model : tardis.model.Radial1DModel
path_or_buf
Path or buffer to the HDF store
path : str
Path inside the HDF store to store the simulation
plasma_properties
`None` or a `PlasmaPropertyCollection` which will
be passed as the collection argument to the
plasma.to_hdf method.
Returns
-------
None
"""
self.runner.to_hdf(path_or_buf, path)
model.to_hdf(path_or_buf, path, plasma_properties)
