> Wouldn't making the model multimodal require scaling the models significantly? Just width if that makes sense. Basically, you add another encoder model but y

> Wouldn't making the model multimodal require scaling the models significantly? Just width if that makes sense. Basically, you add another encoder model but you are not actually increasing the depth that much.