Fetching the paper…

FullDiT: Multi-Task Video Generative Foundation Model with Full Attention · Around