3 ms·
Except AMD still has not figured out how to deliver on LLM acceleration on Linux.
by juujian 2y ago
Except AMD still has not figured out how to deliver on LLM acceleration on Linux.
- d3Xt3r 2y agoNot sure what you mean by that, but ROCm works just fine. I'm using a patched ollama (for APU support) [1] and I can run models like llama3.1:8b and deepseek-r1:14b just fine on my Radeon 780M iGPU, on Arch (btw). [1] https://github.com/rjmalagon/ollama-linux-amd-apu https://github.com/rjmalagon/ollama-linux-amd-apu