nirs4all.pipeline.dagml.experimental_units module

Explicit experimental-unit contracts and native supplied fit influence.

This adapter never counts repetitions, computes weights, resamples rows or scores predictions. DAG-ML owns those operations and the split authority.

nirs4all.pipeline.dagml.experimental_units.apply_experimental_unit_contract(dsl: dict[str, Any], dataset: Any, identity: Any, sample_ints: list[int], *, target_values: Any, target_names: list[str]) → dict[str, Any] | None

Bind unit policy before native compilation/HPO and campaign scoring.

nirs4all.pipeline.dagml.experimental_units.experimental_unit_contract(dataset: Any, identity: Any, sample_ints: list[int], *, target_values: Any, target_names: list[str]) → dict[str, Any] | None

Serialize the declared Train design and native numeric target vocabulary.

nirs4all.pipeline.dagml.experimental_units.experimental_unit_metadata(dataset: Any) → dict[str, dict[int, str]]

Project only explicitly declared IO identities into relation metadata.

nirs4all.pipeline.dagml.experimental_units.native_fit_weights(task: dict[str, Any], dataset: Any, identity: Any, sample_ids: list[str], *, target_names: list[str] | None = None, per_target: bool = False) → ndarray | None

Validate and consume native weights verbatim; never recompute influence.

nirs4all.pipeline.dagml.experimental_units.require_sample_weight_support(estimator: Any) → None

Admit explicit weighted fit signatures across every learned component.

nirs4all.pipeline.dagml.experimental_units.require_weighted_context(context: Any, *, missing_sources: bool = False) → bool

Check all declared operators before advertising native weighted support.

nirs4all.pipeline.dagml.experimental_units.weighted_fit_kwargs(estimator: Any, weights: Any) → dict[str, Any]

Route the exact native weight vector through sklearn’s public fit API.