jesterTOV.inference.result.InferenceResult

class InferenceResult(sampler_type, posterior, metadata, histories=None, fixed_params=None)

Bases: object

Unified HDF5-based storage for JESTER inference results.

This class provides a clean interface for saving and loading inference results from all sampler types. It stores: - Posterior samples (parameters + derived EOS quantities + sampler-specific data) - Metadata (sampler configuration + run statistics) - Histories (time-series data for diagnostics)

Variables:

• sampler_type (SamplerType) – Sampler backend type (“flowmc”, “blackjax_smc_rw”, “blackjax_smc_nuts”, or “blackjax_ns_aw”)

• posterior (Dict[str, np.ndarray]) – All posterior samples including parameters, derived quantities, and sampler-specific data

• metadata (Dict[str, Any]) – Sampler configuration and run statistics

• histories (Dict[str, np.ndarray] | None) – Time-series histories (ESS, acceptance, etc.) for diagnostics

Examples

Creating a result from a sampler:

>>> result = InferenceResult.from_sampler(sampler, config, runtime=3600.0)
>>> result.add_derived_eos(eos_quantities)
>>> result.save("output/results.h5")

Loading a result:

>>> result = InferenceResult.load("output/results.h5")
>>> print(result.summary())
>>> masses = result.posterior["masses_EOS"]

__init__(

sampler_type,

posterior,

metadata,

histories=None,

fixed_params=None,

)

Initialize InferenceResult.

Parameters:

• sampler_type (SamplerType) – Sampler backend type

• posterior (Dict[str, np.ndarray]) – Posterior samples (parameters + derived + sampler-specific)

• metadata (Dict[str, Any]) – Sampler configuration and run statistics

• histories (Dict[str, np.ndarray] | None, optional) – Time-series histories for diagnostics

• fixed_params (Dict[str, float] | None, optional) – Parameters that were pinned to constant values during inference. E.g. {"lambda_BL": 0.0, "lambda_DY": 0.0}.

Methods

__init__(sampler_type, posterior, metadata)	Initialize InferenceResult.
add_derived_eos(eos_dict)	Add derived EOS quantities to posterior.
add_eos_from_transform(transform[, ...])	Apply transform to posterior samples to add EOS quantities.
from_sampler(sampler, config, runtime[, ...])	Create InferenceResult from sampler output.
load(filepath)	Load from HDF5 file.
save(filepath)	Save to HDF5 file.
summary()	Generate human-readable summary of results.

Attributes

config_dict

Deserialize configuration JSON to dictionary.

add_derived_eos(eos_dict)

Add derived EOS quantities to posterior.

This should be called after TOV solver generates M-R-Lambda curves.

Parameters:

eos_dict (Dict[str, Array]) – Dictionary of derived EOS quantities (masses_EOS, radii_EOS, Lambdas_EOS, n, p, e, cs2, etc.)

Return type:

None

add_eos_from_transform(transform, n_eos_samples=None, batch_size=1000)

Apply transform to posterior samples to add EOS quantities.

Parameters:

• transform (JesterTransform) – Transform to apply (should be TOV solver transform)

• n_eos_samples (int, optional) – Number of EOS samples to generate. If None or greater than available samples, generates EOS for all samples. If less than total samples, randomly selects a subset and filters log_prob/sampler fields to match.

• batch_size (int, optional) – Batch size for JAX computation, by default 1000

Return type:

None

Notes

When n_eos_samples < total samples:

• Randomly selects n_eos_samples from posterior

• Applies transform to selected samples

• Filters log_prob and sampler-specific fields (weights, ess, logL, logL_birth) to match selected samples

• Backs up full arrays as *_full before filtering

property config_dict: Dict[str, Any]

Deserialize configuration JSON to dictionary.

Returns:

Dict[str, Any] – Configuration dictionary

classmethod from_sampler(sampler, config, runtime, fixed_params=None)

Create InferenceResult from sampler output.

Extracts production/final samples from the sampler (not training samples).

Parameters:

• sampler (JesterSampler) – Sampler instance after sampling is complete

• config (InferenceConfig) – Configuration used for inference

• runtime (float) – Total runtime in seconds

• fixed_params (Dict[str, float] | None, optional) – Parameters pinned to constant values during inference. Stored in metadata for later retrieval.

Return type:

InferenceResult

Returns:

InferenceResult – Result object with production/final samples and metadata

classmethod load(filepath)

Load from HDF5 file.

Parameters:

filepath (str | Path) – Path to HDF5 file

Return type:

InferenceResult

Returns:

InferenceResult – Loaded result object

Raises:

• FileNotFoundError – If file does not exist

• OSError – If file cannot be read

save(filepath)

Save to HDF5 file.

Parameters:

filepath (str | Path) – Path to output HDF5 file

Raises:

OSError – If file cannot be written

Return type:

None

summary()

Generate human-readable summary of results.

Return type:

str

Returns:

str – Formatted summary string