nirs4all.pipeline.dagml.tuning_contracts module

Typed public tuning contracts for the native DAG-ML tuning lane.

This module is intentionally execution-free. It gives run(tuning=...) a strict, deterministic contract shape before the Optuna/n4m adapters are wired to a shared DAG objective. The parser rejects ambiguous or lossy inputs now so the later execution lane can consume a stable spec instead of an opaque dict.

exception nirs4all.pipeline.dagml.tuning_contracts.DagMLTuningNotImplementedError(tuning_spec: DagMLTuningSpec, *, available_internal: tuple[str, ...] = ('DagMLTuningSpec', 'PipelineObjective', 'prediction score extractor for PipelineObjective', 'single-estimator and linear transformer→estimator PipelineObjective compiler', 'OptunaPipelineObjectiveAdapter', 'N4MPipelineObjectiveAdapter', 'TuningResult', 'TuningResult JSON persistence', 'WorkspaceStore TuningResult persistence', 'RunResult tuning evidence projection', 'explicit winner prediction projection into RunResult', 'linear compile→tune→refit→RunResult orchestration', 'public run(tuning=...) single-estimator/linear array subset'), missing_gates: tuple[str, ...] = ('full nirs4all pipeline-to-DAG-ML objective compiler for splitters/branches/dataset loaders', 'public run(tuning=...) resume integration for broader pipeline shapes', 'public documentation/examples for executable run(tuning=...)'))

Bases: NotImplementedError

Structured public boundary for reserved run(tuning=...) execution.

nirs4all.pipeline.dagml.tuning_contracts.get_tuning_summary_schema() → dict[str, Any]

Return the JSON Schema for lightweight tuning summary artifacts.