Sampling

sampling

Main Sampler that handles the algorithm, output and storage of samples.

Classes:

Name	Description
SamplerRunSettings	Settings for a sampler run
Sampler	Main sampler that runs the MCMC loop

ls_mcmc.sampling.CHECKPOINT_PATH module-attribute

CHECKPOINT_PATH = './sampler_checkpoint.pickle'

ls_mcmc.sampling.SamplerRunSettings dataclass

Settings for running the MCMC sampler.

Attributes:

Name	Type	Description
num_samples	int	Number of samples to generate.
initial_state	np.ndarray	Initial state for the sampler.
print_interval	int	Interval at which outputs should be printed or saved to a file.
store_interval	int	Interval at which sample values should be stored.

ls_mcmc.sampling._SamplerCheckpoint dataclass

Checkpoint data for resuming MCMC sampling.

Attributes:

Name	Type	Description
iteration	int	Current iteration number.
current_state	np.ndarray[tuple[int], np.dtype[np.floating]]	Current state of the Markov chain.
rng_state	dict	State of the random number generator.
run_settings	SamplerRunSettings	Original run settings.
outputs_state	Iterable[output.MCMCOutput]	Serialized state of output objects.

ls_mcmc.sampling.Sampler

MCMC sampler that runs a given algorithm and manages outputs, logging, and storage.

Methods:

Name	Description
resume_from_checkpoint	Resume sampling from a saved checkpoint
run	Run the main MCMC loop

__init__

__init__(algorithm: algorithms.MCMCAlgorithm, sample_storage: storage.MCMCStorage = None, outputs: Iterable[output.MCMCOutput] | None = None, logger: logging.MCMCLogger | None = None, seed: int = 0) -> None

Initializes the Sampler.

Parameters:

Name	Type	Description	Default
algorithm	algorithms.MCMCAlgorithm	The MCMC algorithm to use.	required
sample_storage	storage.MCMCStorage	Storage for samples (e.g. disk).	None
outputs	Iterable[output.MCMCOutput]	Outputs to compute during sampling for logging.	None
logger	logging.MCMCLogger	Logger for progress and diagnostics.	None
seed	int	Random seed for reproducibility.	0

resume_from_checkpoint classmethod

resume_from_checkpoint(algorithm: algorithms.MCMCAlgorithm, sample_storage: storage.MCMCStorage, logger: logging.MCMCLogger, checkpoint_path: Path | None = None) -> Sampler

Resume sampling from a saved checkpoint.

run

run(run_settings: SamplerRunSettings, iteration: int = 0) -> tuple[storage.MCMCStorage, Iterable[output.MCMCOutput]]

Run the MCMC sampler for the specified settings.

Parameters:

Name	Type	Description	Default
run_settings	SamplerRunSettings	Settings for the sampler run.	required
iteration	int	The iteration number at which the sampler starts. Only relevant for restarting a chain.	0

Returns:

Type	Description
tuple[storage.MCMCStorage, Iterable[output.MCMCOutput]]	tuple[storage.MCMCStorage, Iterable[output.MCMCOutput]]: The sample storage and the outputs after sampling.

_run_utilities

_run_utilities(it: int, state: np.ndarray[tuple[int], np.dtype[np.floating]], accepted: bool) -> None

Update outputs, store samples, and log progress for the current iteration.

Parameters:

Name	Type	Description	Default
it	int	Current iteration number.	required
state	np.ndarray	Current state of the chain.	required
accepted	bool	Whether the proposed state was accepted.	required