[mutex_man]

You could run it on a single high end GPU. I can run llama2's models ,(except 70b) on my 4080.