univi.evaluation.cross_modal_predict

univi.evaluation.cross_modal_predict#

univi.evaluation.cross_modal_predict(model, adata_src, src_mod, tgt_mod, device='cpu', layer=None, X_key='X', batch_size=512, use_moe=True, decoder_prefer='auto', decoder_link=None)[source]#

Encode src_mod then decode tgt_mod.

Robust to decoder outputs that are dicts (NB/ZINB/Poisson/Bernoulli/…). Returns a mean-like matrix for evaluation/plotting.

Contents