designspace.SamplingReport

class designspace.SamplingReport(n: int, acceptance_rate: float, constraints: tuple[ConstraintReport, ...], activity: MappingProxyType)

Bases: object

What Space.sampling_report(n, seed, tighten_bounds) returns.

Aggregation only, over the unconditioned measure, drawn before rejection, so both Unknown-swallowing and funnel bias are visible. activity keys are exactly set(space.params), including “[]”-templated definition paths from inside a lifted struct or choice, folded per draw as constraints is.

n

How many draws the report is based on.

Type:

int

acceptance_rate

Fraction of draws that would survive rejection. A low value means sampling is working hard and the accepted configurations are a heavily distorted slice of the measure you declared.

Type:

float

constraints

One row per declared constraint.

Type:

tuple[ConstraintReport, …]

activity

Per parameter, the fraction of draws in which it was active.

Type:

MappingProxyType[str, float]