← Python API
vraven.decompiler · class

FormalCausalMechanismVerifier

FormalCausalMechanismVerifier(model: 'nn.Module', *, model_name: 'str | None' = None) -> 'None'

Certify internal-vector and intervention-effect survival under L∞ input sets.

Usage

from vraven.decompiler import FormalCausalMechanismVerifier

value = FormalCausalMechanismVerifier(...)

Methods

certify(self, x: 'Tensor', *, mechanism: 'MechanismPropertySpec', causal: 'CausalEffectPropertySpec', target: 'int | None' = None, epsilon_grid: 'tuple[float, ...]' = (0.0, 1e-05, 5e-05, 0.0001, 0.00025, 0.0005, 0.001), clip_min: 'float | None' = None, clip_max: 'float | None' = None, radius_search_upper: 'float | None' = None, radius_search_steps: 'int' = 20, cohort_x: 'Tensor | None' = None, cohort_y: 'Tensor | list[int] | None' = None, selected_cohort_indices: 'list[int] | None' = None, cohort_epsilon_grid: 'tuple[float, ...]' = (0.0, 1e-05), input_spec: 'InputDomainSpec | None' = None, calibration_x: 'Tensor | None' = None, calibration_indices: 'list[int] | None' = None, calibration_sample_ids: 'list[str] | None' = None, calibration_labels: 'Tensor | list[int] | None' = None, calibration_disjoint_from_evaluation: 'bool | None' = None, evaluation_sample_ids: 'list[str] | None' = None) -> 'FormalCausalMechanismCertificate'

Construct a proof ledger with independent and conjunctive radii.

Source of truthThis page is generated from the checked-out public symbol, signature, and docstring. Pin the package version for reproducible work.