Hacker News
new
past
comments
ask
show
jobs
points
by
anon-3988
8 hours ago
|
comments
by
mft_
7 hours ago
|
[-]
llama-bench is part of the llama-cpp package, but from recent experimentation, the settings it is able to (or is documented to?) accept lag behind somewhat. Not sure whether it would accept all of the esoteric settings in the article?
reply