nirs4all.pipeline.dagml.pipeline_objective_compiler module

Narrow nirs4all pipeline compiler for the native tuning objective seam.

This is not the full public run(tuning=...) compiler. It accepts the single-estimator and linear transformer→estimator shapes that the existing PipelineObjective can execute honestly today, and rejects broader nirs4all pipeline syntax fail-closed.

class nirs4all.pipeline.dagml.pipeline_objective_compiler.CompiledPipelineObjective(objective: PipelineObjective, search_space: OrderedSearchSpaceSpec, X: Any, y: Any, sample_ids: Any = None, groups: Any = None, metadata: Any = None)

Bases: object

A compiled objective plus its aligned fit data.

X: Any

evaluate(params: Mapping[str, Any], *, trial_index: int = 0) → PipelineObjectiveResult

Evaluate one candidate against the compiled data payload.

groups: Any = None

metadata: Any = None

objective: PipelineObjective

parameter_patches(params: Mapping[str, Any]) → tuple[ParameterPatch, ...]

Return canonical ordered patches for one candidate mapping.

run_tuning(**kwargs: Any) → Any

Run optimizer-driving over the compiled objective.

Keyword arguments are forwarded to nirs4all.pipeline.dagml.tuning_adapters.run_pipeline_objective_tuning() for internal orchestration options such as refit or workspace_path.

sample_ids: Any = None

search_space: OrderedSearchSpaceSpec

y: Any

nirs4all.pipeline.dagml.pipeline_objective_compiler.compile_pipeline_objective(pipeline: Any, X: Any, y: Any, tuning: DagMLTuningSpec | Mapping[str, Any], *, score_extractor: Callable[[Any], float], sample_ids: Any = None, groups: Any = None, metadata: Any = None, clone_estimator: bool = True) → CompiledPipelineObjective

Compile the currently supported nirs4all shape into PipelineObjective.

Supported input forms are:

• an estimator instance with fit;

• [estimator];

• [{"model": estimator}] or {"model": estimator}.

• {"steps": [...]} or {"pipeline": [...]} for the same supported step sequence;

• a linear sequence of transformer instances followed by one estimator;

• transformer mapping steps {"transform": transformer, "name": "..."} followed by {"model": estimator};

• explicit "sklearn.module.Class" string steps without constructor parameters.

Splitters, branches, string model aliases, finetune_params and arbitrary step keywords remain outside this seam and raise explicit errors.

nirs4all.pipeline.dagml.pipeline_objective_compiler.run_single_estimator_tuning_to_run_result(pipeline: Any, X: Any, y: Any, tuning: DagMLTuningSpec | Mapping[str, Any], *, score_extractor: Callable[[Any], float], sample_ids: Any = None, groups: Any = None, metadata: Any = None, clone_estimator: bool = True, refit: bool = True, workspace_path: str | Path | None = None, workspace_name: str = '', workspace_tuning_id: str | None = None, workspace_metadata: Mapping[str, Any] | None = None, run_id: str | None = None, pipeline_id: str | None = None, chain_id: str | None = None, resume_tuning_result: Any | None = None, resume_tuning_id: str | None = None, per_dataset: Mapping[str, Any] | None = None, winner_x: Any | None = None, winner_y_true: Any | None = None, winner_score: float | None = None, winner_metric: str | None = None, winner_sample_ids: Sequence[Any] | None = None, winner_dataset_name: str = 'tuning_winner', winner_model_name: str | None = None, winner_task_type: str = 'regression', winner_metadata: Mapping[str, Any] | None = None) → RunResult

Run the narrow single-estimator tuning lane and return a RunResult.

This helper stitches together the current internal seams: compile_pipeline_objective() → optimizer/refit orchestration → project_objective_tuning_to_run_result(). It remains deliberately narrower than public run(tuning=...) and inherits the projection rule that winner predictions require explicit caller-provided score evidence.