"""The frozen paper score definitions (``wisp_release.scoring``) under the ``wisp`` namespace. ``score(y_true, y_pred, n_classes)`` returns macro-F1 (classes present in truth or prediction), accuracy, balanced accuracy, full-vocabulary worst-class F1 and the confusion matrix, for integer labels on the global vocabulary ``0..n_classes-1``. """ from wisp_release.scoring import from_confusion, score __all__ = ["score", "from_confusion"]