目录

1. 环境描述

2. GPU环境安装指南

2.1 安装NVIDIA驱动

2.2 安装CUDA

3. 安装nvidia-fabricmanager

3.1 验证fabricmanager安装结果

4. pytorch2.0安装和CUDA验证指南

4.1 miniconda安装并创建alpha环境

4.2 安装pytorch2.0并验证cuda状态

---

【摘要】 Nvidia A系列裸金属服务器安装NVIDIA和CUDA驱动，安装conda和pytorch2.0并验证cuda的有效性。

1. 环境描述

操作系统：Ubuntu 22.04 server 64bit

选择安装环境相关版本： GPU驱动版本为530.30.02、CUDA版本为12.1.0

本文以上述信息配置NVIDIA驱动、CUDA和FabricManager， 并安装PyTorch2.0， 验证其可以正常运行。

2. GPU环境安装指南

添加源

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list


sed -i -e '/experimental/ s/^#//g' /etc/apt/sources.list.d/nvidia-container-toolkit.list


sudo apt-get update

安装依赖包

# 清理之前的安装环境
apt remove  *nvidia*

apt remove nvidia-headless-525 nvidia-headless-no-dkms-525 nvidia-kernel-common-525   nvidia-utils-525 nvidia-dkms-525 nvidia-compute-utils-525 nvidia-kernel-source-525 libnvidia-cfg1-525 

apt-get -y install libnvidia-container1:amd64  libnvidia-container-tools      nvidia-container-runtime  nvidia-container-toolkit-base

2.1 安装NVIDIA驱动

# 下载安装包
wget https://us.download.nvidia.cn/XFree86/Linux-x86_64/530.30.02/NVIDIA-Linux-x86_64-530.30.02.run

chmod  +x NVIDIA-Linux-x86_64-530.30.02.run

./NVIDIA-Linux-x86_64-530.30.02.run

2.2 安装CUDA

注意事项： 不能选择Driver, 否则会覆盖已安装的NVIDIA驱动.

wget https://developer.download.nvidia.com/compute/cuda/12.1.0/local_installers/cuda_12.1.0_530.30.02_linux.run

chmod +x cuda_12.1.0_530.30.02_linux.run

./cuda_12.1.0_530.30.02_linux.run --toolkit --samples --silent

3. 安装nvidia-fabricmanager

Ant系列GPU支持 NvLink & NvSwitch，若您使用多GPU卡的机型，需额外安装与驱动版本对应的nvidia-fabricmanager服务使GPU卡间能够互联，否则可能无法正常使用GPU实例。

注意事项： fabricmanager版本一定要和nvidia驱动版本必须保持一致.

version=530.30.02
main_version=$(echo $version | awk -F '.' '{print $1}')
apt-get update
# apt-get -y install nvidia-cuda-toolkit  libnvidia-container1:amd64  libnvidia-container-tools  nvidia-container-runtime 
apt-get remove nvidia-fabricmanager*  
# https://developer.download.nvidia.cn/compute/cuda/repos/debian10/x86_64/                  
wget https://developer.download.nvidia.cn/compute/cuda/repos/debian10/x86_64/nvidia-fabricmanager-530_530.30.02-1_amd64.deb
apt-get install ./nvidia-fabricmanager-530_530.30.02-1_amd64.deb 
# apt-get -y install nvidia-fabricmanager-${main_version}=${version}-*

3.1 验证fabricmanager安装结果

验证驱动安装结果、启动fabricmanager服务并查看状态

nvidia-smi -pm 1
nvidia-smi
systemctl unmask nvidia-fabricmanager.service 
systemctl enable nvidia-fabricmanager
systemctl start nvidia-fabricmanager
systemctl status nvidia-fabricmanager

4. pytorch2.0安装和CUDA验证指南

PyTorch2.0所需环境为Python3.10， 安装配置miniconda环境。

4.1 miniconda安装并创建alpha环境

wget https://repo.anaconda.com/miniconda/Miniconda3-py310_23.1.0-1-Linux-x86_64.sh

chmod 750 Miniconda3-py310_23.1.0-1-Linux-x86_64.sh

bash ./Miniconda3-py310_23.1.0-1-Linux-x86_64.sh -b -p /root/miniconda3

export PATH=/home/miniconda/bin:$PATH

conda create --quiet --yes -n alpha python=3.10

4.2 安装pytorch2.0并验证cuda状态

在alpha环境下安装torch2.0, 使用清华PIP源完成.

source activate alpha 

pip install torch==2.1 -i https://pypi.tuna.tsinghua.edu.cn/simple

python

验证torch与cuda的安装状态,输出为True即为正常.

import torch 

print(torch.cuda.is_available())

如果遇到问题如下图

解决方法：

pip install numpy