neurovlm as another text embedding model

[jdkent]

neurovlm is another useful embedding, you can use the projection head head output into the shared space (it's a CLIP model), since the text embedder is not neuroimaging specific specter2 text embedding.

I did a a much shorter vibe coding project predicting brain maps from abstracts:
https://github.com/jdkent/cns-abstract-explorer

[satra]

thanks and nice work @jdkent - will consider adding neurovlm on a future run.