centos7 大模型基础环境配置

1、配置yum源

#添加访问互联路由
cat >> /etc/resolv.conf <<EOF
nameserver 8.8.8.8
nameserver 114.114.114.114
EOF

cat /etc/resolv.conf


#设置yum源
rm -rf /etc/yum.repos.d/bak && mkdir -p /etc/yum.repos.d/bak && mv /etc/yum.repos.d/* /etc/yum.repos.d/bak

# 获取当前系统的版本信息
release_info=$(cat /etc/redhat-release)

    if [[ $release_info == *"CentOS Linux release 7"* ]]; then
        # CentOS 7
        curl -o /etc/yum.repos.d/CentOS7-Base.repo http://mirrors.cloud.tencent.com/repo/centos7_base.repo
        yum install wget -y
        wget -O /etc/yum.repos.d/epel.repo http://mirrors.cloud.tencent.com/repo/epel-7.repo
    elif [[ $release_info == *"CentOS Linux release 8"* ]]; then
        # CentOS 8
        curl -o /etc/yum.repos.d/CentOS8-Base.repo http://mirrors.cloud.tencent.com/repo/centos8_base.repo
        yum install wget -y
        wget -O /etc/yum.repos.d/epel.repo http://mirrors.cloud.tencent.com/repo/epel-8.repo
    elif [[ $release_info == *"Rocky Linux release 8"* ]]; then
        # Rocky Linux 8
        curl -o /etc/yum.repos.d/Rocky8-Base.repo http://mirrors.aliyun.com/repo/Rocky-8.repo
    elif [[ $release_info == *"AlmaLinux release 8"* ]]; then
        # AlmaLinux 8
        curl -o /etc/yum.repos.d/AlmaLinux8-Base.repo http://mirrors.aliyun.com/repo/AlmaLinux-8.repo
    else
        echo "Unsupported system version."
        exit 1
    fi


yum install epel-release -y

yum clean all && yum makecache

yum update -y

2、升级内核

#下载包
wget http://193.49.22.109/elrepo/kernel/el7/x86_64/RPMS/kernel-ml-4.18.0-1.el7.elrepo.x86_64.rpm
wget http://193.49.22.109/elrepo/kernel/el7/x86_64/RPMS/kernel-ml-devel-4.18.0-1.el7.elrepo.x86_64.rpm
wget http://193.49.22.109/elrepo/kernel/el7/x86_64/RPMS/kernel-ml-headers-4.18.0-1.el7.elrepo.x86_64.rpm

#安装
yum install -y kernel-ml-4.18.0-1.el7.elrepo.x86_64.rpm kernel-ml-devel-4.18.0-1.el7.elrepo.x86_64.rpm kernel-ml-headers-4.18.0-1.el7.elrepo.x86_64.rpm


# 设置启动内核
grub2-set-default 0

#重启系统
sync && reboot

3、下载显卡驱动包

driver550 cuda12.4

#官网下载地址
https://www.nvidia.com/en-us/drivers/
https://www.nvidia.com/en-us/drivers/details/238864/


https://cn.download.nvidia.cn/XFree86/Linux-x86_64/550.144.03/NVIDIA-Linux-x86_64-550.144.03.run



#装驱动之前需要装依赖并且禁用nouveau 
# 禁用 nouveau 并更新 initramfs
sudo vim /etc/modprobe.d/blacklist.conf
# 添加以下内容:
blacklist nouveau
options nouveau modeset=0

# 重建 initramfs
sudo dracut --force
# 重启系统
sudo reboot
 
# 安装内核开发包和头文件

sudo yum install -y gcc gcc-c++ make kernel-devel-$(uname -r) kernel-headers-$(uname -r)
 
# 切换到命令行模式
sudo systemctl isolate multi-user.target
 
# 运行 NVIDIA 驱动安装程序
sudo bash NVIDIA-Linux-x86_64-<version>.run

#检查命令
nvidia-smi

4、 安装cuda 12.4

#官网地址
https://developer.nvidia.com/downloads


wget https://developer.download.nvidia.com/compute/cuda/12.4.0/local_installers/cuda-repo-rhel7-12-4-local-12.4.0_550.54.14-1.x86_64.rpm

sudo rpm -i cuda-repo-rhel7-12-4-local-12.4.0_550.54.14-1.x86_64.rpmsudo 

yum clean allsudo 

yum -y install cuda-toolkit-12-4



#检查命令
nvcc -V

5、screen后台页面下载技巧

screen 是一个在 Linux 终端上创建和管理多个会话的命令行工具。它可以在一个终端窗口中创建多个虚拟终端，并让您在这些终端之间轻松切换。

1、安装

#以前一直使用centos7，安装srceen只需要一条命令即可：
yum install screen -y


#原来centos8安装srceen的方式与centos7安装方式不一样了，首先需要安装epel：
yum install epel-release -y

#然后再安装screen：
yum install screen -y

2、常见的 screen 命令的用法：

#1.创建一个新的 screen 会话
screen
或者

screen -S myscreen

#2.列出当前存在的 screen 会话
screen -ls


#3.重新连接到一个已存在的 screen 会话：
#您可以在运行 screen -ls 命令时找到会话的 ID
screen -r myscreen
或者
screen -r <session_id>

screen -r 32447

#4.结束 screen 会话
exit


#按下 Ctrl+a 键，然后再按下 d 键，将会分离当前的 screen 会话，并返回到原始终端窗口。
Ctrl+a+d



# 后台下载案例参考
nohup git lfs clone https://www.modelscope.cn/deepseek-ai/DeepSeek-R1.git &

6、资料参考网站：

https://blog.youkuaiyun.com/free541/article/details/142762982

https://www.cnblogs.com/sheepships/p/18072617

https://blog.youkuaiyun.com/Mr_kka/article/details/143911821

显存要求