Y
Hacker News
new
|
ask
|
show
|
jobs
by
GaggiX
1114 days ago
From what I can find online LLAMA-65B 4-bit quantized can run 1 token/s on a Ryzen 7 3700X (using llama.cpp).