scpdac.models.load_classifier_checkpoints

scpdac.models.load_classifier_checkpoints(species, map_location='cpu')

Load the three hierarchical-classifier checkpoints for a species.

Parameters:

• species (str) – "human" or "mouse".

• map_location (str (default: 'cpu')) – Torch map_location for torch.load().

Return type:

dict[str, dict]

Returns:

Mapping {"root", "malignant", "non_malignant"} to checkpoint dicts (each holding state_dict, genes, classes, …).

Raises:

FileNotFoundError – If the species directory or any of the three checkpoints is missing.