Repository navigation
v0.3.13
What's new
- Add
keep_empty_features=False: drop entirely missing training columns by default, or retain them with zero values when enabled. - Add
missing_values=np.nan: support finite numeric markers such as-1and9999, detected before float32 conversion. - Support both options across donor policies, missing indicators, feature names, and pandas output.
Validation: 474 local tests passed with two existing NumPy warnings.