[–] brucethemoose2 3y ago ↗ It seems to perform best with 16K context, but works well with up to 32K context.This is interesting, especially coming from Amazon, as vanilla Mistral-Instruct is much better than llamav2 7B, and extremely fast on modest hardware.
1 comment
[ 3.1 ms ] story [ 10.4 ms ] threadThis is interesting, especially coming from Amazon, as vanilla Mistral-Instruct is much better than llamav2 7B, and extremely fast on modest hardware.