torchcell.datamodels.ontology_checks.join_key_audit

join_key_audit(records: Iterable[tuple[Mapping[str, Any], Mapping[str, Any]]]) → JoinKeyCensus[source]

Census of the identity keys a built dataset actually carries.

Takes (experiment, reference) mapping pairs – exactly what an LMDB yields – so a verifier can run it over any dataset without importing its loader. The counts answer the question a multimodal training set turns on: at which level can these records be joined to another dataset’s records at all.