Changing the max_batch_size reduces VRAM from 22 – ~15 gbs
other possibilities mentioned in video, no full solution found yet.
This video goes over how to install and run LLaMa on Ubuntu 22.04
The process should be similar for windows

llama github:
Teknium magnet link:

TO DOWNLOAD MAGNET LINK
open bittorrent, qbittorrent, or utorrent and add the torrent link
(download qbittorrent if you are ubuntu. bittorrent if Windows)(utorrent should work for either)

UPDATE *****NEW VIDEO WITH LOW VRAM*****: