sample

sample(model, sampler, *, max_samples=None, frequency=None, features='s', key=None, **kwargs)[source]

Explore the parameter space of a model using a specified sampling engine.

This unified router executes the sampling algorithm using a standardized state-machine loop (init, step, terminate), supporting both one-shot and adaptive active learning strategies frictionlessly.

Parameters:
  • model (Model) – The parametric model to sample.

  • sampler (AbstractSampler) – The sampling algorithm to use.

  • max_samples (int | None, default=None) – The maximum number of samples to generate. For one-shot samplers, this is the exact number generated. For adaptive samplers, this acts as a computational budget. If None, adaptive samplers run until convergence.

  • frequency (Frequency | None, default=None) – The frequency sweep for feature evaluation.

  • features (EvaluatorLike | None, default=None) – The specific circuit features to extract.

  • key (Array | None, default=None) – JAX PRNG key for stochastic samplers.

  • **kwargs – Additional arguments passed to the underlying evaluators.

Returns:

The comprehensive result object containing the original continuous model and batched execution states.

Return type:

ExploreResult