conformance

kernel-conformance

Domain-agnostic test helpers for executable kernels in Scientific Memory (numeric closeness, monotonicity grids, etc.).

Kernel packages (for example kernels/adsorption) should depend on this workspace package and import kernel_conformance in tests instead of duplicating helpers.

Usage

from kernel_conformance import assert_close, assert_monotone_nondecreasing

Adding a new kernel family

1. Add a new directory under kernels/<family>/ with its own pyproject.toml.
2. Add kernel-conformance as a workspace dependency (see kernels/adsorption/pyproject.toml).
3. Reuse helpers here; add new shared helpers only when at least two families need them.