Can this work for different model types like 2.0 and xl?

OPTML-Group / AdvUnlearn

Official implementation of "Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models"

Creative Commons Attribution 4.0 International

16 stars 1 forks source link

Can this work for different model types like 2.0 and xl? #1

Closed ninjasaid2k closed 3 months ago

ninjasaid2k commented 3 months ago

Can this work for different model types like 2.0 and xl? and what the VRAM requirement?

damon-demon commented 3 months ago

Theoretically, our proposed robust unlearning framework could potentially be applied to both SD v2.0 and SDXL. However, the current codebase supports only SD v1.x, which is the focus of existing DM unlearning methods. For SD v1.x, fine-tuning through AdvUnlearn with a batch size of 5 for utility-retaining regularization, requires approximately 40GB.