aspire.samplers.base

aspire.samplers.base#

Attributes#

logger

Classes#

Sampler

Base class for all samplers.

Module Contents#

aspire.samplers.base.logger[source]#
class aspire.samplers.base.Sampler(log_likelihood, log_prior, dims, prior_flow, xp, dtype=None, parameters=None, preconditioning_transform=None)[source]#

Base class for all samplers.

Parameters:

  • log_likelihood (Callable) – The log likelihood function.

  • log_prior (Callable) – The log prior function.

  • dims (int) – The number of dimensions.

  • flow (Flow) – The flow object.

  • xp (Callable) – The array backend to use.

  • parameters (list[str] | None) – The list of parameter names. If None, any samples objects will not have the parameters names specified.

  • prior_flow (aspire.flows.base.Flow)

  • dtype (Any | str | None)

  • preconditioning_transform (Callable | None)

prior_flow[source]#
dims[source]#
xp[source]#
backend_str = 'numpy'[source]#
dtype = None[source]#
parameters = None[source]#
history = None[source]#
n_likelihood_evaluations = 0[source]#
fit_preconditioning_transform(x)[source]#

Fit the data transform to the data.

abstractmethod sample(n_samples)[source]#
Parameters:

n_samples (int)

Return type:

aspire.samples.Samples

log_likelihood(samples)[source]#

Computes the log likelihood of the samples.

Also tracks the number of likelihood evaluations.

Parameters:

samples (aspire.samples.Samples)

Return type:

aspire.samples.Samples

log_prior(samples)[source]#

Computes the log prior of the samples.

Parameters:

samples (aspire.samples.Samples)

Return type:

aspire.samples.Samples

config_dict(include_sample_calls='last')[source]#

Returns a dictionary with the configuration of the sampler.

Parameters:

include_sample_calls (bool | str, optional) – Whether to include the sample calls in the configuration. Default is True. If True, and if the sampler has a sample method with a calls attribute, the calls will be included in the config under the key “sample_calls”. If this fails for any reason, a warning will be logged and the sample calls will be omitted.

Return type:

dict

build_checkpoint_state(samples, iteration=None, meta=None, include_sample_calls='last')[source]#

Prepare a serializable checkpoint payload for the sampler state.

Parameters:

  • samples (aspire.samples.Samples)

  • iteration (int | None)

  • meta (dict | None)

  • include_sample_calls (str | bool)

Return type:

dict

serialize_checkpoint(state, protocol=None)[source]#

Serialize a checkpoint state to bytes with pickle.

Parameters:

  • state (dict)

  • protocol (int | None)

Return type:

bytes

default_checkpoint_callback(state)[source]#

Store the latest checkpoint (state + pickled bytes) on the sampler.

Parameters:

state (dict)

Return type:

None

default_file_checkpoint_callback(file_path)[source]#

Return a simple default callback that overwrites an HDF5 file.

Parameters:

file_path (str | pathlib.Path | None)

Return type:

Callable[[dict], None]

save_checkpoint_to_hdf(state, h5_file, path='sampler_checkpoints', dsetname=None, protocol=None)[source]#

Save a checkpoint state into an HDF5 file as a pickled blob.

Parameters:

  • state (dict)

  • path (str)

  • dsetname (str | None)

  • protocol (int | None)

Return type:

None

load_checkpoint_from_file(file_path, h5_path='checkpoint', dsetname='state')[source]#

Load a checkpoint dictionary from .pkl or .hdf5 file.

Parameters:

  • file_path (str | pathlib.Path)

  • h5_path (str)

  • dsetname (str)

Return type:

dict

restore_from_checkpoint(source)[source]#

Restore sampler state from a checkpoint source.

Parameters:

source (str | bytes | dict)

Return type:

tuple[aspire.samples.Samples, dict]

property last_checkpoint_state: dict | None[source]#

Return the most recent checkpoint state stored by the default callback.

Return type:

dict | None

property last_checkpoint_bytes: bytes | None[source]#

Return the most recent pickled checkpoint produced by the default callback.

Return type:

bytes | None