1 paper · 1 filter
Rui Huang, Shitong Shao, Zikai Zhou +6
Diffusion models have achieved remarkable performance on a wide range of generative tasks, yet training them from scratch is notoriously resource-intensive, typically requiring mil…