DiffusionOPSD
🔥 On-Policy Self-Distillation in Diffusion Models
free(무료 요금제 있음)840 github stars
DiffusionOPSD 소개
*Note: This is an external implementation of the algorithm in the following paper:*
Reward-guided diffusion post-training through explicit, continually refreshed intermediate targets
Wei Zhou1,2, Xiongwei Zhu1, Lingdong Kong2, Bo Chen1, Lei Zhang3, Yongyuan Liang4, Xiaoxia Hou1, Tian Ye5, Xian Sun6, Yingshuo Wang7, Linfeng Li2, Shengqiong Wu8, Leigang Qu2, Feng Li9, Wei Liu1,†, Julian McAuley3, Tat-Seng Chua2