Skip to content

JTT94/example_diffusion

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

diffusions

Setup

  conda create -n example_env python=3.8
  conda activate example_env
  pip install git+https://github.com/JTT94/jtils.git
  pip install tensorflow tensorflow_datasets jax flax jaxlib tqdm matplotlib scikit-learn

Run Example notebooks

    pip install jupyter 
    

Procedure

  • Training consists of a few steps:

    • Sample data $$x_0$$
    • Sample random time $t \sim Uniform(1,2,3\ldots,T)$
    • Apply noise from SDE to generate $x_t$
    • Train a model based on objective $E[||model(x_t, t) - noise||^2]$
  • Sampling consists of applying the model sequentially through time and scaling the output

About

Toy Diffusion/ Score Based Generative Model in Jax

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published