We doubled the number of architectures (89 to 167) with new models in audio, text, vision, multiple modalities or even time seriesand protein folding Here are a few highlights in the most used of those new models

→ View original post on X — @huggingface