SigLIP-based NLLB-CLIP

[visheratin]

Hi, Merve!

I noticed that you are comparing multilingual SigLIP with NLLB-CLIP. But there actually is a newer version of NLLB-CLIP that uses the SigLIP vision encoder! It is integrated into OpenCLIP, and you can find base and large versions on the Hub.

Also, a couple of days ago, there was a thread, which described the strong and weak points of multilingual SigLIP and NLLB-CLIP.