Skip to content

Latest commit

 

History

History
12 lines (9 loc) · 398 Bytes

diffusion2gan.md

File metadata and controls

12 lines (9 loc) · 398 Bytes

Diffusion2GAN - GAN PEFT on diffusion generations

  1. generate a bunch of content from a diffusion model.
  2. wire up a styleGAN-XL for LoRA training
  3. apply projected GAN, using new FID results/ensembles

TODO:

  1. experiment with LoRA/GLoRA + GAN
  2. experiment with new FID stuff + GAN
  3. mix n match
  4. hit up that GAN wizard, see if he's interested or has already done some of this already