On 4/10/26 12:21 PM, Kirill A. Korinsky wrote:
On Fri, 10 Apr 2026 11:42:22 +0200,
Ali Farzanrad <[email protected]> wrote:
Hello all,
In my Minisforum UM790 Pro Mini PC, after running llama-cli it stuck at
loading stage and my system gets very slow (it takes 4-5 seconds until I
see the character after I press a key in xterm). However after
disabling the amdgpu in /etc/bsd.re-config file I have no such problem.
Can it be related: https://github.com/ggml-org/llama.cpp/issues/21440?
If setting GGML_VK_ALLOW_GRAPHICS_QUEUE=1 makes things work (with amdgpu),
it might be related.
Also https://github.com/ggml-org/llama.cpp/issues/21724 mentions similar
symptoms (without fixing #21440, though).
