FiTv2: Scalable and Improved Flexible Vision Transformer for Diffusion Model Paper • 2410.13925 • Published Oct 17 • 22