Add doc string for n_gpu_layers argument and make -1 offload all layers#604

abetlen merged 2 commits intoabetlen:mainabetlen/llama-cpp-python:mainfrom

aliencaocao:main-1Copy head branch name to clipboard