Y
Hacker News
new
|
ask
|
show
|
jobs
by
wirybeige
104 days ago
The vulkan backend for llama.cpp isn't that far behind rocm for pp and tp speeds