lucataco/dpo-sdxl
Direct Preference Optimization (DPO) is a method to align diffusion models to text human preferences by directly optimizing on human comparison data
Public
2.2K
runs