I am playing around with putting Hermes on my computer and it got a bit confused by the llama-cpp version. It wants 0.5.0 and our system reports us having 0.4.1 while actually being 0.5.0, so hermes suggested building 0.5.0 when it isnt needed, Im just trying to save the next user the headache of finding this out themselves.
Hermes also suggested that I mention this
- Uses system ggml with dlopen backends (HIP/Vulkan/OpenCL separate packages)
- Ships an Ollama compat layer (patches 0005)
- Systemd service already configures render+video groups and /dev/kfd access
- Suggests ggml-backend-hip instead of hard-depending on it
I would have signed up for github and done the package request there but they are blocking my access to sign up because I had hermes look at the llama-cpp page when figuring this out.
Operating System: OpenMandriva ROME 26.09
KDE Plasma Version: 6.7.5
KDE Frameworks Version: 6.29.0
Qt Version: 6.11.2
Kernel Version: 7.2.7-desktop-1omv2690 (64-bit)
Graphics Platform: X11
Processors: 12 × Intel® Core™ i7-8700 CPU @ 3.20GHz
Memory: 16 GiB of RAM (15.5 GiB usable)
Graphics Processor: AMD Radeon RX 6700 XT