Elevator Pitch: The objective of pyfra is to make it as fast and low-friction as possible to write research code involving complex pipelines over many machines. Currently, this means being like shell scripts on steroids. See https://github.com/EleutherAI/pyfra/ for a piece of sample code that illustrates the main features of pyfra.
Goal Outputs: Open source library and possibly a paper
Milestones: see pyfra readme
How to Help: I really need alpha testers who are willing to put up with bugs, constant breaking changes, etc to try out pyfra, fix bugs, and give feedback on what the main pain points are.
Desired Support: mostly need people who are already working on other eleuther projects to use pyfra for stuff