llama 3 -support?

Hi,

Has anyone tried LLAMA-3 using this codebase? Mine is not working with `llama3-8b`. i.e. it reports no errors, and training was able to start. However, it got stuck at step1. Not sure whether I should expect this code to support `llama3-8b` to start with.

Anyone has any experience, I would like to hear more! 

Thanks!