Super HN

New Show
   Training a SOTA 7B LRM with 1.5x Throughput Improvment (github.com)