neurovlm.training.train_text_to_brain

neurovlm.training.train_text_to_brain(config, *, provider=None, lookup=None, model=None, semantic_evaluator=None)

Train a fresh Stage 4 projector with a frozen, provenance-bound AE.

Parameters:

• config (TextToBrainTrainConfig)

• provider (AtlasFreeCNNDataProvider | None)

• lookup (AtlasFreeTextEmbeddingLookup | None)

• model (CNNTextToBrainModel | None)

• semantic_evaluator (Callable[[...], Mapping[str, float]] | None)

Return type:

TextToBrainTrainResult