torchcell.knowledge_graphs.build_time_projection.gather_dataset_full_records

gather_dataset_full_records(data_root: str, build_tree: str | None = None) → dict[str, DatasetSize][source]

Regenerate full record counts by reading each dataset’s LMDB stat entries.

Read-only (lock=False); never rebuilds a dataset. Prefers the dev tree (data_root); falls back to build_tree when a dataset’s dev-tree LMDB is absent. Use this to refresh DATASET_FULL_RECORDS after a rebuild.