llama-cpp-0.5.0

llama.cpp inference server with multi-node CPU tensor parallelism via the RPC backend. Designed for x86_64 Linux nodes (e.g. HPE DL380 Gen9) with no discrete GPU. All llama.cpp nodes are selected by the label llama.cpp=enabled.