Multimodal Embedding Gemma Incoming?

#37
by LorryG - opened

I see that T5 Gemma 2 is multimodal, using a SigLip image encoder. How likely are we to get a multimodal early-fusion embedding model?

Hi @LorryG
While we can't comment on the likelihood of specific future releases, your feedback regarding early-fusion models has been forwarded to our developers.Thanks

Sign up or log in to comment