7x speed improvement for LLaMA in less than 10 lines of code (github.com) 2 points by hack_ml 3y ago ↗ HN
1 comment
[ 3.1 ms ] story [ 14.4 ms ] thread