ModelAdapter.build#

ModelAdapter.build(X, y, *, coords=None)[source]#

Construct the backend’s graph/design state without sampling.

Idempotent by skipping when already built. Bayesian backends merge data-driven priors and construct the PyMC graph; sklearn backends record the design matrices for sample_posterior().

Parameters:

  • X (Any) – Predictor matrix in the backend’s expected form.

  • y (Any) – Outcome vector or matrix in the backend’s expected form.

  • coords (dict[str, Any] | None) – Coordinate metadata for PyMC models. Ignored by sklearn backends.

Return type:

None