#!/bin/bash # make sure your python virtual env is already sourced and active. export CMAKE_ARGS="-DLLAMA_CUBLAS=on" export FORCE_CMAKE=1 pip install --upgrade llama-cpp-python[server]