nirs4all.pipeline.dagml.general_archive module
Trusted Python model archive replay through an explicit DAG PREDICT phase.
These host artifacts are not portable Core archives. Their digest protects against corruption, not a malicious producer: joblib can execute Python code.
- nirs4all.pipeline.dagml.general_archive.copy_general_archive(source: Path, target: str | Path, *, expected_fingerprint: str | None) Path[source]
Save a loaded Session without fitting or changing its original model bytes.
- nirs4all.pipeline.dagml.general_archive.general_archive_manifest(path: str | Path) dict[str, Any] | None[source]
Inspect the provenance marker without importing/deserializing operators.
- nirs4all.pipeline.dagml.general_archive.predict_general_archive(path: str | Path, data: Any, *, expected_archive_fingerprint: str | None = None, loaded_archive: dict[str, Any] | None = None, output: str | None = None) PredictResult[source]
Replay a captured aggregate model; no old executor or retraining is used.