nessai.samplers.base

nessai.samplers.base#

Base nested sampler object

Classes#

BaseNestedSampler

Base nested sampler class.

Module Contents#

class nessai.samplers.base.BaseNestedSampler(model: nessai.model.Model, nlive: int, output: str = None, seed: int = None, rng: numpy.random.Generator | None = None, checkpointing: bool = True, checkpoint_interval: int = 600, checkpoint_on_iteration: bool = False, checkpoint_callback: Callable | None = None, logging_interval: int = None, log_on_iteration: bool = True, resume_file: str = None, plot: bool = True, n_pool: int | None = None, pool: Any | None = None)#

Bases: abc.ABC

Base nested sampler class.

Parameters:

modelnessai.model.Model: User-defined model
outputstr, optional: Path for the output. If not specified the current working directory is used.
seedint, optional: Seed used to seed numpy and torch.
checkpointingbool, optional: Boolean to toggle checkpointing, must be enabled to resume the sampler. If false the sampler is still saved at the end of sampling.
checkpoint_intervalint: The interval used for checkpointing. By default this is a time interval in seconds. If checkpoint_on_iteration=True this corresponds to the number of iterations between checkpointing.
checkpoint_on_iterationbool: If true the checkpointing interval is checked against the number of iterations
checkpoint_callbackCallback: Callback function to be used instead of the default function. The function will be called in the code:checkpoint method as: :code`checkpoint_callback(self)`.
logging_intervalint, optional: The interval in seconds used for periodic logging. If not specified, then periodic logging is disabled.
log_on_iterationbool: If true logging will occur based on the iteration. If false logging will be periodic if logging_interval is set. In case where neither logging is enabled, log_on_iteration will be set to true with an interval of nlive.
resume_filestr, optional: Name of the file the sampler will be saved to and resumed from.
plotbool, optional: Boolean to enable or disable plotting.
n_poolint, optional: Number of threads to when for creating the multiprocessing pool.
poolobject: User defined multiprocessing pool that will be used when evaluating the likelihood.

property likelihood_evaluation_time#: Current log-likelihood time

property total_likelihood_evaluations#: Total number of likelihood evaluations

likelihood_calls#: Alias for total_likelihood_evaluations

abstract property posterior_effective_sample_size#: The effective sample size of the posterior distribution

configure_output(output: str | None, resume_file: str | None = None)#

Configure the output folder

Parameters:

outputstr: Directory where the results will be stored
resume_filestr, optional: Specific file to use for checkpointing. If not specified the default is used (nested_sampler_resume.pkl)

update_output(output: str) → None#

Update the output directory and resume file.

Parameters:

output: str: Path to the output directory

configure_rng(seed: int | None = None, rng: numpy.random.Generator | None = None)#

Configure the random number generation.

If a seed is not specified, a random seed is generated using the specified random number generator, or the default numpy random number if not specified. In the latter case, the seed is then used to seed the new numpy random number generator.

..versionadded:: 0.14.0

Parameters:

seedOptional[int]: The random seed. If not specified, a random seed is generated.
rng: Optional[np.random.Generator]: Random number generator. If not specified, the default numpy random number generator is used with the specified seed.

configure_random_seed(seed: int | None = None) → None#

Initialise the random seed

..deprecated:: 0.14.0: Deprecated in favour of configure_rng().

Parameters:

seedOptional[int]: The random seed. If not specified, a random seed is generated.

configure_periodic_logging(logging_interval, log_on_iteration)#

Configure the periodic logging.

Parameters:

logging_intervalint, optional: The interval in seconds used for periodic logging. If not specified, then periodic logging is disabled.
log_on_iterationbool: If true logging will occur based on the iteration. If false logging will be periodic if logging_interval is set. In case where neither logging is enabled, log_on_iteration will be set to true with an interval of nlive.

periodically_log_state()#

Log the state of the sampler.

Calls log_state if the elapsed interval in time (or iterations) is more than the specified interval.

initialise_history() → None#: Initialise the dictionary to store history

update_history() → None#: Update the history dictionary

checkpoint(periodic: bool = False, force: bool = False, save_existing: bool = True)#

Checkpoint the classes internal state.

Parameters:

periodicbool: Indicates if the checkpoint is regular periodic checkpointing or forced by a signal. If forced by a signal, it will show up on the state plot.
forcebool: Force the sampler to checkpoint.
save_existingbool: If True, the previous checkpoint will be copied to .old. If False, the new checkpoint will override the previous file.

classmethod resume_from_pickled_sampler(sampler: Any, model: nessai.model.Model, output: str | None = None, checkpoint_callback: Callable | None = None)#

Resume from pickle data.

Parameters:

dataAny: Pickle data
modelnessai.model.Model: User-defined model
outputOptional[str]: New output directory. If not specified, the output directory will be the same as the previous run.
checkpoint_callbackOptional[Callable]: Checkpoint callback function. If not specified, the default method will be used.

Returns:

Instance of BaseNestedSampler

classmethod resume(filename: str, model: nessai.model.Model, output: str | None = None, **kwargs)#

Resumes the interrupted state from a checkpoint pickle file.

Parameters:

filenamestr: Pickle file to resume from
modelnessai.model.Model: User-defined model
outputOptional[str]: New output directory. If not specified, the output directory will be the same as the previous run.

Returns:

obj: Instance of BaseNestedSampler

close_pool(code=None)#: Close the multiprocessing pool.

get_result_dictionary()#

Return a dictionary that contains results.

Only includes version, seed and sampling time. Child classes should call this method and add to the dictionary.