Dear numpy community,

 

I'm happy to announce the first public release of einops package.

einops is a new way to manipulate tensors.

 

Examples with numpy worth a thousand words:

 

einops introduces a special notation which includes composition and decomposition of axes.

This notation allows non-trivial rearrangement of elements, which can be combined with reductions.

 

Goal of the project is to provide a way to write a more readable and maintainable code with additional checks,

which works consistently and uniformly across a set of popular tensor packages.

 

It should also complement well existing deep learning packages when operations are missing

or when extensive usage of transpose/reshapes/broadcasts drives to hardly readable and (frequently) buggy code.

 

 

Project page: https://github.com/arogozhnikov/einops