7x speed improvement for LLaMA in less than 10 lines of code (github.com) 2 points by hack_ml 3y ago ↗ HN
1 comment
[ 5.8 ms ] story [ 17.9 ms ] thread