Text prompt embedding size

Thank you for releasing the model.
The paper mentions that text prompts are encoded using a pretrained `ViT-L/14@336px` CLIP model. CLIP embedding from this model are of size 768 while SegmentAnything prompt embedding is of size 256. Are there any extra steps for converting CLIP embeddings before feeding into the model?

Start with the paper's description of text prompts, the pretrained ViT-L/14@336px CLIP model, and SegmentAnything's 256-dimensional prompt embeddings. Determine whether the 768-dimensional CLIP output is transformed before inference; done means documenting the conversion steps or confirming that maintainer clarification is required.

Written by the indexing model from the issue text.