it's a very different architecture and starting from a different pre-trained model!

→ View original post on X — @jxmnop