Ubuntu 22.04 安装Nvidia驱动加速deepseek

一键安装22.04 nvidia 驱动

• nvidia 官网下载驱动
• 我的环境是NVIDIA RTX A5000
• nvidia 文档参考
• 没有安装驱动之前确认自己的型号 lspci | grep -i vga （如数字2231） 参考
• docker 支持nvidia ，注释了需要的取消注释即可 42行-92行
• 一定要重启服务器哦，不然驱动不会生效的reboot

vim /nvidia_install.sh

#!/bin/bash
# -*- coding: utf-8 -*-
# Author: CIASM
# update 2025/02/27
# make.ha<<!
# check nvidia
lspci | grep -i vga
http://pci-ids.ucw.cz/mods/PC/10de/2204#add-apt-repository ppa:graphics-drivers/ppa
!echo "remove nvidia"
apt remove -y nvidia*echo "add nvidia repo"
curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | 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' | \
tee /etc/apt/sources.list.d/nvidia-container-toolkit.listecho "check host nvidia"
ubuntu-drivers devicesecho "install nvidia"
apt-get update
apt install -y nvidia-driver-535echo "install docker NVIDIA GPU"
apt install -y nvidia-container-toolkitecho "install NVIDIA CUDA Toolkit"
apt install -y nvidia-cuda-toolkitecho "nvidia persist mode"
nvidia-smi -pm 1# docker 支持 nvidia 配置
<<!
echo "docker daemon.json"
rm -rf /etc/docker/daemon.json 
cat <<'EOF'>>/etc/docker/daemon.json
{"registry-mirrors": ["https://registry.hub.docker.com","https://ccr.ccs.tencentyun.com","https://dockerproxy.com","https://hub-mirror.c.163.com","https://docker.mirrors.sjtug.sjtu.edu.cn","https://docker.nju.edu.cn","https://registry-k8s-io.mirrors.sjtug.sjtu.edu.cn","https://docker.m.daocloud.io","https://docker.mirrors.ustc.edu.cn","https://mirror.iscas.ac.cn","https://s64h8lpn.mirror.aliyuncs.com","https://atomhub.openatom.cn","https://mirror.baidubce.com","https://docker.1panel.live","https://proxy.1panel.live","https://image.cloudlayer.icu","https://docker-0.unsee.tech","https://docker.tbedu.top","https://pull.loridocker.com","https://docker.melikeme.cn","https://docker.imgdb.de","https://docker.hlmirror.com","https://docker.kejilion.pro","https://hub.rat.dev","https://dockerpull.pw","https://hub.fast360.xyz","https://docker.xuanyuan.me","https://docker.1ms.run","https://xdark.top","https://func.ink","https://lispy.org"],"insecure-registries": ["192.168.11.40"],"runtimes": {"nvidia": {"args": [],"path": "nvidia-container-runtime"}}
}
EOFecho "restart docker"
systemctl restart dockerecho "test cuda docker"
docker run --rm --gpus all nvidia/cuda:12.2.0-base-ubuntu22.04 nvidia-smi
!

一键安装nvidia 驱动

• 一定要重启服务器哦，不然驱动不会生效的reboot

bash /nvidia_install.sh

 查询安装完成的nvidia驱动情况

nvidia-smi

nvidia 持续模式

• 持久模式（-pm）允许用户将 NVIDIA-SMI 设置持久化，即使在系统重启后也能生效。这意味着，您可以在会话期间配置设备，而无需每次重新启动系统时都手动重新应用设置。
• 1：启用持久模式
• 0：禁用持久模式
• 立即生效

echo "nvidia persist mode"
nvidia-smi -pm 1

nvidia ECC 校验开启和关闭

• ECC（错误纠正代码）是一种用于检测和纠正设备内存错误的技术。启用 ECC 可以提高系统稳定性，防止因内存错误而导致的数据损坏。
• 1：启用 ECC
• 0：禁用 ECC
• reboot 系统生效

nvidia-smi -e 1

重置 ECC 错误计数（-p） 

• -p 选项用于重置 ECC 错误计数器。如果 ECC 已启用，此计数器将跟踪检测到的内存错误数量。重置计数器可以帮助您监测和排除故障，并确保您收到设备错误的最新信息。
• 0/VOLATILE,
• 1/AGGREGATE

nvidia-smi -p 0