nirs4all.pipeline.dagml.structural_multimodal module
Typed structural alternatives with one complete Methods early-fusion Ridge.
The SDK lowers declarations and supplies raw buffers. Native DAG-ML enumerates recipes and Methods alone learns every selected encoder and predictor.
- nirs4all.pipeline.dagml.structural_multimodal.is_methods_model_choice(steps: list[Any]) bool[source]
Identify the additive profile while retaining the old dense classifiers.
- nirs4all.pipeline.dagml.structural_multimodal.lower_typed_choices(steps: list[Any], schemas: dict[str, Any]) dict[str, Any][source]
Lower one native Cartesian stage; recipe identities remain native-owned.
- nirs4all.pipeline.dagml.structural_multimodal.prepare_typed_structure(steps: list[Any], splitter: Any, dataset: Any, spec: DagMLTuningSpec, run_options: dict[str, Any], native: Any) dict[str, Any][source]
Seal full raw contracts before catalogue preparation or optimizer FIT.
- nirs4all.pipeline.dagml.structural_multimodal.train_selected_typed_structure(prepared: dict[str, Any], evidence: dict[str, Any], run_options: dict[str, Any]) Any[source]
Resolve the native winner and reuse the complete Methods portable capture.