You are not logged in.

#1 2023-09-01 18:05:56

danielem
Member
Registered: 2020-12-27
Posts: 25

llama.cpp at Ryzen 5700U with hardware acceleration

Hey!

Has anyone managed to get hardware acceleration from Ryzen 5700U (or similar model) when running llama.cpp?

Did get some  results with Vulkan with mlc-llm https://github.com/mlc-ai/mlc-llm  I see some acceleration and free cpus, BUT it is very unstable.

From the other hand llama.cpp https://github.com/ggerganov/llama.cpp works like a charm on CPU, but cpu consumption is freakin high, so I was trying to get it working with OpenCL. Tried different OpenCL drivers https://wiki.archlinux.org/title/GPGPU, but so far unsuccessful.

Or maybe there is some other option to run code llama models with hardware acceleration on Ryzen 5700U?

Offline

Board footer

Powered by FluxBB