fuselip-multimodal-embeddings

Build unified multimodal embeddings with a single transformer encoder processing image and text tokens together, improving performance on structure-aware tasks through early fusion.

Independent GitHub snapshot · Automated classification · Original documentation governs