nirs4all.pipeline.dagml.native_pls_fold_hpo module

Identity-bound splitter declarations for native PLS studies within folds.

Only the split structure is materialized here. DAG-ML runs every optimizer, fit, metric and selection; Methods owns the fitted numerical models.

class nirs4all.pipeline.dagml.native_pls_fold_hpo.NativePlsFoldPlan(pipeline: list[Any], outer_fold_set: dict[str, Any], inner_fold_sets: dict[str, Any], refit_inner_fold_set: dict[str, Any])[source]

Bases: object

Declared outer/inner folds, before the native runtime attests them.

inner_fold_sets: dict[str, Any]

outer_fold_set: dict[str, Any]

pipeline: list[Any]

refit_inner_fold_set: dict[str, Any]

nirs4all.pipeline.dagml.native_pls_fold_hpo.normalize_fold_resume_package(value: Any) → str | None[source]: Accept only a complete package carrying the separate fold ledger.

nirs4all.pipeline.dagml.native_pls_fold_hpo.prepare_native_pls_fold_plan(profile: NativePlsPhaseControls, identity: DagMLFitIdentityFrame, *, n_features: int, split_groups: Any) → NativePlsFoldPlan[source]: Run the declared splitter on each scoped ID pool, never on outer targets.

nirs4all.pipeline.dagml.native_pls_fold_hpo.validate_fold_splitter(splitter: Any) → None[source]: Keep the local native profile within explicit partitioning splitters.