Skip to content

An extremely fast template based implementation of reverse mode automatic differentiation in C++/Eigen

Notifications You must be signed in to change notification settings

hyparxis/static-autodiff

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 

Repository files navigation

adiff

Adiff is a header only C++ library that implements reverse mode automatic differentiation with compile-time static graphs. Because the computation graph is defined at compile time, adiff has extremely good cache performance and incurs no run-time costs like dynamic memory allocation, meaning it can get up to 1000x performance improvements over general libraries like Pytorch C++ in some cases.

Right now adiff is mostly an example toy library implementing only a couple common neural network operations.

About

An extremely fast template based implementation of reverse mode automatic differentiation in C++/Eigen

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages