[Bug 289813] amdgpu: running and inferencing with "koboldcpp" or "llama.cpp" using locks up the GPU...

In reply to: bugzilla-noreply_a_freebsd.org: "[Bug 289813] Vulkan: running and inferencing with "koboldcpp" or "llama.cpp" using the Vulkan backend locks up the GPU..."
Go to: [ bottom of page ] [ top of archives ] [ this month ]

From: <bugzilla-noreply_at_freebsd.org>
Date: Wed, 08 Oct 2025 14:02:59 UTC

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=289813

Nils Beyer <nbe@renzel.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|amdgpu: running and         |amdgpu: running and
                   |inferencing with            |inferencing with
                   |"koboldcpp" or "llama.cpp"  |"koboldcpp" or "llama.cpp"
                   |using the Vulkan backend    |using locks up the GPU...
                   |locks up the GPU...         |

--- Comment #1 from Nils Beyer <nbe@renzel.net> ---
same goes for using the OpenCL backend of "koboldcpp" using the "rusticl"
driver. So it seems not being a Vulkan issue...

-- 
You are receiving this mail because:
You are the assignee for the bug.