Skip to content

Latest commit

 

History

History
44 lines (32 loc) · 1.83 KB

README.md

File metadata and controls

44 lines (32 loc) · 1.83 KB

AdjointDEIS (NeurIPS 2024)

AdjointDEIS: Efficient Gradients for Diffusion Models

Zander W. Blasingame   ·   Chen Liu

Clarkson University

arXiv Paper Webpage

News

  • 2024.09.25: AdjointDEIS was accepted by NeurIPS 2024! Our open-source project is under development, stay tuned for updates!

Todo

  • release the camera-ready version of the paper
  • release the code for solving the adjoint Probability Flow ODE
  • release the code for solving the adjoint diffusion SDE

Introduction

AdjointDEIS is a training-free method for guided generation of diffusion models which uses the method of adjoint sensitivty. AdjointDEIS consists bespoke ODE/SDE solvers which compute the gradients of the Probability Flow ODE and diffusion SDE w.r.t. the model parameters, solution trajectories, and conditional information.

Please refer to the paper and project page for detailed methods and results.

Code

Under construction. Visit soon!

Citation

If this work was helpful for your research, please consider citing the following paper:

@inproceedings{blasingame2024adjointdeis,
  title = {Adjoint{DEIS}: Efficient Gradients for Diffusion Models},
  author = {Blasingame, Zander W. and Liu, Chen},
  booktitle = {The Thirty-eighth Annual Conference on Neural Information Processing Systems},
  year = {2024},
  url = {https://openreview.net/forum?id=fAlcxvrOEX},
}