Replies: 1 comment
-
I've never tested on Xeons with good amount of fast memory, but in theory they should work well - this project has AVX2 kernels written to support CPU inference and uses OMP for multi-threading. There might be some tuning that would be required if the target HW has a lot of memory bandwidth available though, as CPU perf was only tested on desktop systems with a moderate amount of memory bandwidth (eg ~50 GB/s on Zen 4 platform). |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
XEON seems a good platform to host large model
Beta Was this translation helpful? Give feedback.
All reactions