API and entry points

Command interfaces

Interface	Purpose
neuralpom --config CONFIG prepare	Build caches and calibration
neuralpom --config CONFIG train --stage STAGE	Train or resume a hybrid stage
neuralpom-ai	Direct neural forecast preparation, training and evaluation
python scripts/v1/plan.py --config CONFIG	Inspect hybrid sampling and budgets
python scripts/v1/evaluate.py --help	Training-oriented frozen evaluation
python scripts/v1/report.py --help	Training run report
python -m neuralpom.evaluation.campaign --help	Paired frozen forecast campaigns

The hybrid training stages are pretrain, midtrain4 and posttrain.

Core and state modules

neuralpom.core_factory.resolve_core_class(name) resolves explicit or implicit_p0_fixed. neuralpom.hybrid.build_initialized_core constructs an initialized core. The v1 package supplies configuration, data, cache and checkpoint interfaces; the evaluation package supplies deployment policies and forecast diagnostics. See Architecture.