llama.cpp fork from PrismML-Eng/llama.cpp (prism branch) with the same backends as science:machinelearning/llamacpp (CPU, Vulkan, OpenCL, OpenVINO) for x86_64 openSUSE Tumbleweed and Leap 16.0. Required for ternary (PTQ1_0/PQ2_0) Bonsai model support.