Model

Classes related to models, and collections of models (AllModels).

API

Classes:

AllModels([config, from_file])	All models which have been run so far
Model([config, parset, directory])	A DYNAMITE model.

class model.AllModels(config=None, from_file=True)

All models which have been run so far

The main attribute self.table is an Astropy table holding all models run so far.

Parameters:

configa dyn.config_reader.Configuration object

from_filebool

whether to create this object from a saved all_models.ecsv file

Methods:

convert_legacy_chi2_file([legacy_filename])	Convert legacy format of chi2 files
get_best_n_models([n, which_chi2])	Get the best n models or all but the n best models so far
get_best_n_models_idx([n, which_chi2])	Get the indices of the best n models so far
get_ml_of_original_orblib(model_id)	Get ml of model number model_id's original orblib
get_model_from_directory(directory)	Get the Model from a model directory
get_model_from_parset(parset)	Get the Model from a parset
get_model_from_row(row_id)	Get a Model given a table row
get_model_velocity_scaling_factor([...])	Get the model's velocity scaling factor
get_mods_within_chi2_thresh([which_chi2, delta])	Get models within or outside a delta threshold of the best
get_parset_from_row(row_id)	Get a parset set given a table row
get_row_from_model([model])	Get the table row for a Model
make_best_models_table([which_chi2, n, ...])	Make a table of the best models and save it to disk
make_empty_table()	Make an empty Astropy table
read_legacy_chi2_file(legacy_filename)	Read the legacy AKA schwpy format of chi2 files
read_model_table()	Read table from file self.filename
remove_unused_orblibs()	Removes orbit libraries for 'bad' models.
retrofit_kinmapchi2()	Calculates kinmapchi2 for DYNAMITE legacy tables if possible.
save()	Save the all_models table
set_filename(filename)	Set the name (including path) for this model
update_model_table()	all_models table update: fix incomplete models, add kinmapchi2.

convert_legacy_chi2_file(legacy_filename=None)

Convert legacy format of chi2 files

legacy AKA schwpy format were likely called `griddata/_chi2.cat.

Parameters:

legacy_filename: string

the legacy_filename (probably griddata/_chi2.cat)

get_best_n_models(n=10, which_chi2=None)

Get the best n models or all but the n best models so far

Parameters:

nint, optional

How many models to get. If negative, all models except the n best models will be returned. The default is 10.

which_chi2str, optional

Which chi2 is used for determining the best models. If None, the setting from the configuration file will be used. The default is None.

Returns:

a new astropy.table object holding the best n models, sorted by

which_chi2

Raises:

ValueError

If which_chi2 is neither None nor a valid chi2 type.

get_best_n_models_idx(n=10, which_chi2=None)

Get the indices of the best n models so far

Parameters:

which_chi2str, optional

Which chi2 is used for determining the best models. If None, the setting from the configuration file will be used. The default is None.

Returns:

list of int

indices in the all_models table of the n best model so far, sorted by which_chi2

get_ml_of_original_orblib(model_id)

Get ml of model number model_id’s original orblib

The original ml is required to rescale orbit libraries for rescaled potentials. This method searches self.table, the all_models table.

Parameters:

model_idint

The self.table row index of the model.

Returns:

ml_orblibfloat

the original ml

Raises:

ValueError

If the ml parameter is not in the parameter space or the model’s orblib cannot be found.

get_model_from_directory(directory)

Get the Model from a model directory

Parameters:

directorystr

The directory string needs to start with the output directory defined in config.settings.io_settings['output_directory']

Returns:

moda dyn.model.Model object

Raises:

ValueError

If the directory does not exist in the all_models table.

get_model_from_parset(parset)

Get the Model from a parset

Parameters:

parsetlist

a list of dyn.parspace.Parameter objects

Returns:

a dyn.model.Model object

get_model_from_row(row_id)

Get a Model given a table row

Parameters:

row_idint

which row

Returns:

a dyn.model.Model object

get_model_velocity_scaling_factor(model_id=None, model=None)

Get the model’s velocity scaling factor

Returns sqrt(model_ml/original_orblib_ml). The model can be either given by its row id in self.table or as a Model object. Note that the parameters model_id and model are mutually exclusive.

Parameters:

model_idint compatible

The model’s row id in self.table.

modela Model object

Returns:

scaling_factorfloat

The model’s velocity scaling factor sqrt(model_ml/original_orblib_ml).

Raises:

ValueError

If not exactly one of model_id and model are supplied.

get_mods_within_chi2_thresh(which_chi2=None, delta=None)

Get models within or outside a delta threshold of the best

Parameters:

which_chi2str, optional

Which chi2 is used for determining the best models. If None, the setting from the configuration file will be used. The default is None.

deltafloat, optional

The threshold value. Models with chi2 values differing from the opimum by at most delta will be returned. If None, models within 10% of the optimal value will be returned. If delta is negative, models that are NOT within a delta threshold of the best are returned. The default is to return models within 10% of the best.

Returns:

a new astropy.table object holding the ‘’delta-best’’ models

(if delta >= 0) or holding all but the ‘’delta-best’’ models

(if delta < 0), respectively.

Raises:

ValueError

If which_chi2 is neither None nor a valid chi2 type.

get_parset_from_row(row_id)

Get a parset set given a table row

Parameters:

row_idint

which row

Returns:

list

a list of dyn.parspace.Parameter objects

get_row_from_model(model=None)

Get the table row for a Model

Parameters:

modela Model object

Returns:

row_idint

The self.table row index of model.

Raises:

ValueError

If the model is not found in self.table.

make_best_models_table(which_chi2=None, n=None, delta=None, filename=None)

Make a table of the best models and save it to disk

Parameters:

which_chi2str, optional

Which chi2 is used for determining the best models. If None, the setting from the configuration file will be used. The default is None.

nint, optional

How many models to get. If None, n will be ignored. Default: if delta is specified, the default is none; if delta is None, the default is 10.

deltafloat, optional

The threshold value. Models with chi2 values differing from the opimum by at most delta will be returned. If None, delta will be ignored. The default is None.

filenamestr, optional

File name of the best models table. The file is written into the output directory specified in the config file. If None, the name is the same as for the all models table but with ‘_best’ added to the base file name. If the file already exists, a warning is logged, the existing file is backed up (‘_backup’ added), and then overwritten. The default is None.

Returns:

int

The number of models in the best models table.

Raises:

ValueError

If both n and delta are specified (i.e., both are not None).

make_empty_table()

Make an empty Astropy table

Returns:

None

sets self.table

read_legacy_chi2_file(legacy_filename)

Read the legacy AKA schwpy format of chi2 files

Taken from schw_basics.py, reads in legacy files named similar to griddata/_chi2.cat

Parameters:

legacy_filename: string

the legacy_filename (probably griddata/_chi2.cat)

read_model_table()

Read table from file self.filename

Returns:

None

sets self.table

remove_unused_orblibs()

Removes orbit libraries for ‘bad’ models.

Frees disk space by deleting data, keeping only model data required by the beta_plot and mass_plot plotting routines. Keeps data of models with (kin)chi2 values less than or equal to sqrt(2 * number of kinematic observations) * min(chi2), but at least 3 models. Will mark a deleted orbit library with orblib_done=False and weights_done=False in the all_models table. If an orblib cannot be deleted because it is used by another model, only the nnls data will be deleted, which is marked by weights_done=False in the all_models table.

Returns:

bool

True if data deletion has been attempted, False if no data to delete could be identified.

retrofit_kinmapchi2()

Calculates kinmapchi2 for DYNAMITE legacy tables if possible.

Returns:

None.

updates self.table

save()

Save the all_models table

set_filename(filename)

Set the name (including path) for this model

Parameters:

filenamestring

name for this model

Returns:

None

sets self.filename

update_model_table()

all_models table update: fix incomplete models, add kinmapchi2.

Dealing with incomplete models: Models with all_done==False but an existing model_done_staging.ecsv will be updated in the table and the staging file will be deleted. Models with all_done==False and no existing orblib will be deleted from the table and their model directory will be deleted, too. The configuration setting reattempt_failures determines how partially completed models with all_done==False but existing orblibs are treated: If reattempt_failures==True, their orblib_done will be set to True and later the ModelIterator will execute the weight solving based on the existing orblibs. If reattempt_failures==False, the model and its directory will be deleted. Note that orbit libraries on disk will not be deleted as they may be in use by other models.

Up to DYNAMITE 3.0 there was no kinmapchi2 column in the all_models table. If possible (data exists on disk), calculate and add the values, otherwise set to np.nan.

Returns:

None

sets self.table

class model.Model(config=None, parset=None, directory=None)

A DYNAMITE model.

The model can be run by running the methods (i) get_orblib, (ii) get_weights and (iii) (in the future) do_orbit_colouring. Running each of these methods will return the appropriate object, e.g. model.get_orblib() –> returns an OrbitLibrary object model.get_weights(…) –> returns a WeightSolver object

Parameters:

configa dyn.config_reader.Configuration object

parsetrow of an Astropy Table

contains the values of the potential parameters for this model

directorystr

The model directory name (without path). If None or not specified, the all_models_file will be searched for the directory name. If the all_models file does not exist, the model directory will be set to orblib_000_000/ml{ml}.

Returns:

Nothing returned. Attributes holding outputs are are added to the

object when methods are run.

Methods:

check_parset(parspace, parset)	Validate a parset
create_model_directory(path)	make a directory if it does not yet exist
get_model_directory()	get the name of this model's output directory
get_orblib()	Make the orbit library
get_weights([orblib])	Get the orbital weights
setup_directories()	setup directories
validate_config_file()	Validate the content of the config file against the model's config file

check_parset(parspace, parset)

Validate a parset

Given parameter values in parset, the validate_parspace method of the parameter space is executed. If a parameter exists in parspace but not in parset, a warning will be issued and the parameter will remain unchanged. If parset tries to set the value of a parameter not existing in parspace, an exception will be raised. Validating relies on exceptions raised by validate_parspace.

Parameters:

parspacedyn.parameter_space.ParameterSpace

A list of parameter objects.

parsetrow of an Astropy Table

Contains parameter values to be checked against the settings in parspace.

Returns:

None.

Raises:

ValueError

If at least one parameter in parset is unknown to parspace.

create_model_directory(path)

make a directory if it does not yet exist

Parameters:

pathstring

directory path to make

get_model_directory()

get the name of this model’s output directory

Returns:

string

get_orblib()

Make the orbit library

Returns:

a dyn.orblib.OrbitLibrary object

get_weights(orblib=None)

Get the orbital weights

Parameters:

orbliba dyn.orblib.OrbitLibrary object

Returns:

a dyn.weight_solver.WeightSolver object

setup_directories()

setup directories

validate_config_file()

Validate the content of the config file against the model’s config file

Upon solving a model, DYNAMITE creates a backup of the config file in the model directory. Instantiating a model later (e.g., for plotting) using a config file that is incompatible with the one used to create the model can lead to problems (e.g., differently sized orbit library). This method validates the “global” config file against the one in the model directory (if existing).

Differing config files may be ok and intended (e.g., due to expanding the parameter space). Therefore, the main purpose of this method is to add warnings to the log to alert the user.

Returns:

bool

False if a config file backup is successfully found in the model directory and it differs from the “global” config file. True otherwise (no config file backup could be identified or the config file backup is identical to the global config file).

Raises:

FileNotFoundError

If the “global” config file cannot be found.

Inheritance Diagram