eval_toolkit.splits

GroupKFoldSplitter	K-fold CV with group-disjoint test partitions.
HoldoutSplitter	Single-iteration (k=1) holdout split via sklearn train_test_split.
PoolBuilder	Augment a fold's train slice with an external pool and split off val.
PurgedKFoldSplitter	Time-aware k-fold with explicit purge gap + post-test embargo.
SourceDisjointKFoldSplitter	K-fold CV partitioning sources into disjoint groups (round-robin).
Splitter	Iterates folds, each as a named-splits dict ready for evaluate(...).
StratifiedKFoldSplitter	K-fold cross-validation with class-stratification.
TimeSeriesSplitter	Time-aware K-fold via sklearn.model_selection.TimeSeriesSplit.
compute_label_overlap	Boolean (n_train, n_test) matrix: True where label windows overlap.
iter_folds_with_pool	Compose a Splitter with a PoolBuilder.