GPU服务器安装驱动

环境

系统和内核

shell 复制代码

uname -r  # 5.15.0-163-generic
lsb_release -a # Ubuntu 22.04.5 LTS

docker
shell 复制代码
```
docker version # 29.1.2         
```

GPU

shell 复制代码

nvidia-smi # Tesla T4 NVIDIA-SMI 550.163.01             Driver Version: 550.163.01     CUDA Version: 12.4

nvcc

shell 复制代码

nvcc --version # 12.4, V12.4.131 , cuda_12.4.r12.4/compiler.34097967_0

安装

docker

shell 复制代码

apt purge -y docker-ce docker-ce-cli containerd.io docker-compose-plugin docker.io
rm -rf /var/lib/docker
rm -rf /var/lib/containerd
rm -rf /etc/docker
rm -f /etc/apt/sources.list.d/docker.list
apt clean
rm -rf /var/lib/apt/lists/*
apt update &&  apt install -y ca-certificates curl gnupg lsb-release
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" | tee /etc/apt/sources.list.d/docker.list > /dev/null
apt update
apt install -y docker-ce docker-ce-cli containerd.io docker-compose-plugin
systemctl start docker
systemctl enable docker
systemctl status docker
ls -l /var/run/docker.sock # 查看这个文件是否存在

安装GPU驱动和Cuda

shell 复制代码

apt purge -y nvidia-* cuda-*
apt autoremove -y && apt autoclean
rm -rf /usr/local/cuda*
rm -rf /var/lib/nvidia*
apt update && apt upgrade -y
apt install -y build-essential dkms gcc g++ cmake libglvnd-dev pkg-config

tee /etc/modprobe.d/blacklist-nouveau.conf <<EOF
blacklist nouveau
blacklist lbm-nouveau
options nouveau modeset=0
alias nouveau off
alias lbm-nouveau off
EOF

update-initramfs -u
reboot
lsmod | grep nouveau # 无输出则成功

rm -f /etc/apt/sources.list.d/nvidia-cuda.list
rm -f /etc/apt/sources.list.d/nvidia-cuda.list.bak 2>/dev/null
rm -f /etc/apt/sources.list.d/nvidia-machine-learning.list
apt update
apt install -y software-properties-common apt-transport-https curl
curl -fsSL https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/3bf863cc.pub | gpg --dearmor -o /usr/share/keyrings/nvidia-cuda-keyring.gpg
echo "deb [signed-by=/usr/share/keyrings/nvidia-cuda-keyring.gpg] https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/ /" | tee /etc/apt/sources.list.d/nvidia-cuda.list
apt update

apt install -y linux-headers-$(uname -r)
apt install -y linux-headers-generic build-essential dkms gcc g++
rm -rf /var/lib/dkms/nvidia/ 2>/dev/null
dpkg --configure -a
apt install -y nvidia-driver-550 nvidia-dkms-550
modprobe nvidia
reboot
nvidia-smi

安装python

shell 复制代码

apt install -y git curl zlib1g-dev libssl-dev libreadline-dev libsqlite3-dev libbz2-dev libffi-dev gcc make
git clone https://github.com/pyenv/pyenv.git ~/.pyenv
echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bashrc
echo 'command -v pyenv >/dev/null || export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bashrc
echo 'eval "$(pyenv init -)"' >> ~/.bashrc
source ~/.bashrc
mkdir ~/.pyenv/cache
cd  ~/.pyenv/cache
wget https://mirrors.aliyun.com/python-release/source/Python-3.10.19.tar.xz
apt install -y liblzma-dev xz-utils  
pyenv install 3.10.19  # 会提示ModuleNotFoundError: No module named '_tkinter'，忽略，我们不需要图形界面
pyenv global 3.10.19
python -m venv ~/cuda124-py310
source ~/cuda124-py310/bin/activate
mkdir -p ~/.config/pip && vi ~/.config/pip/pip.conf  
      # 2. 写入以下内容（清华源，包含PyTorch适配）
      # [global]
      # index-url = https://pypi.tuna.tsinghua.edu.cn/simple
      # trusted-host = pypi.tuna.tsinghua.edu.cn
      # extra-index-url = https://download.pytorch.org/whl/cu124
pip install torch torchvision torchaudio -i https://pypi.tuna.tsinghua.edu.cn/simple --index-url https://download.pytorch.org/whl/cu124

安装nvidia-container-toolkit

shell 复制代码

distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | tee /etc/apt/sources.list.d/nvidia-docker.list
apt-get update && sudo apt-get install -y nvidia-container-toolkit
nvidia-ctk runtime configure --runtime=docker
systemctl restart docker

验证

shell 复制代码

docker run --rm --gpus all nvidia/cuda:11.8.0-base-ubuntu20.04 nvidia-smi # 如果有输出驱动信息，则正常，没有则不正常