leffff / adversarial-diffusion-distillation

My Implementation of Adversarial Diffusion Distillation https://arxiv.org/pdf/2311.17042.pdf

Repository from Github https://github.comleffff/adversarial-diffusion-distillationRepository from Github https://github.comleffff/adversarial-diffusion-distillation

adversarial-diffusion-distillation

Please refer to this repository: for better implementation: https://github.com/leffff/FlowModels












My Implementation of Adversarial Diffusion Distillation https://arxiv.org/pdf/2311.17042.pdf Model scheme

Warning

This implementation differs from the original implementation in the paper in following ways:

  1. The discriminator is a single prediction head resnet50 initialized from these weights, but not a ViT conditioned on DINOv2 and CLIP features. Therefore has only 1 prediction head, instead of k.
  2. The discriminator is trained without the R1 penatly Discriminator Loss with R1 penatly

About

My Implementation of Adversarial Diffusion Distillation https://arxiv.org/pdf/2311.17042.pdf

License:MIT License


Languages

Language:Jupyter Notebook 99.9%Language:Python 0.1%