lazyslide_models.multimodal.Prism

class Prism(model_path=None, token=None)

Bases: ModelBase

prism 🤗Hugging Face Paper Params: 557.7M CC-BY-NC-ND-4.0 [Shaikovski et al., 2024] A multi-modal generative foundation model for slide-level histopathology, the Prism models encode slide-level embeddings from Virchow.

respond(img_latents, prompt, max_length=100)

Generate a free-text caption per slide.

Implements TextResponseModelProtocol. img_latents is the "latents" entry of encode_slide()’s output — not raw tile embeddings, which is what Prism2.respond takes.