Y
Hacker News
new
|
ask
|
show
|
jobs
Fusing a 27B ternary LLM's whole decode step into one CUDA kernel
(
twitter.com
)
3 points
by
Jr23_xd
15 days ago