Fetching the paper…

EDiT: Efficient Diffusion Transformers with Linear Compressed Attention · Around