Y
Hacker News
new
|
ask
|
show
|
jobs
by
spion
1201 days ago
llama.cpp needs 40GB for the 65B model (due to int4 quantization)
RamNeeded(other_size) ~= 40GB * other_size/65B