https://github.com/vividitytech/diffusion-mcmc

https://github.com/vividitytech/diffusion-mcmc

Last synced: about 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/vividitytech/diffusion-mcmc
• Owner: vividitytech
• Created: 2022-12-20T02:26:07.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-01-04T21:50:04.000Z (over 1 year ago)
• Last Synced: 2024-07-03T22:25:49.541Z (3 months ago)
• Language: Python
• Size: 13.7 KB
• Stars: 3
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# diffusion-mcmc

Code for reproducing some key results of our paper Speed up the inference of diffusion models via shortcut MCMC sampling (https://arxiv.org/abs/2301.01206).

## Requirement

jaxlib, tensorflow, etc

## run example:

python global_vdm_2d.py

## some important parameters:

 joinFlag = False # add the flag whether we need joined learning

 

 freeze_x_decoder = False

 

 shortcut = True

 

 K=10 default, which can change based on experiments

## reference

we demonstrate our approach based on a 2D swirl dataset and using MLPs, most code base are from VDM [Link to open in Colab](https://colab.research.google.com/github/google-research/vdm/blob/main/colab/2D_VDM_Example.ipynb).

if you think it is helpful please cite:

Gang Chen, Speed up the inference of diffusion models via shortcut MCMC sampling (https://arxiv.org/abs/2301.01206).