arm: device=cpu · pid=72478 · -ngl 0 · /home/xbill/llama.cpp/build-cpu/bin/llama-server
arm: device=gpu · pid=74308 · -ngl 99 · mapped: ggml-cuda, libcublas, libcuda, libcudart · /home/xbill/llama.cpp/build/bin/llama-server