Repository navigation
v0.3.6
Fix support for NumPy integer n_neighbors values, including parameter grids generated with np.arange() in Pipeline and GridSearchCV.
- Accept positive Python and NumPy integer scalars for both donor policies.
- Preserve the original estimator parameter for scikit-learn cloning.
- Convert neighbor counts to Python integers before FAISS search and candidate-expansion arithmetic.
- Explicitly reject Python and NumPy booleans. Python
Truewas previously accepted as1. - Preserve existing donor-count limits and failed-fit cleanup.
Validation: 189 tests passed for the fix, including 39 new regression cases covering NumPy-generated grids, signed and unsigned integers, cloning, input preservation, invalid parameters, donor limits, and candidate-expansion overflow.