要在 CentOS 上正確安裝和配置 NVIDIA Container Toolkit，您可以按照以下步驟進行操作，如果1和2都已經完成，可以直接進行第3步NVIDIA Container Toolkit安裝配置。
1. 安裝 NVIDIA GPU 驅動程序：
您可以從 NVIDIA 官方網站下載適用于您的 GPU 型號和 CentOS 版本的驅動程序，并按照安裝指南進行安裝。確保您的系統已正確安裝并配置了 NVIDIA GPU 驅動程序。
2. 安裝 Docker CE：
2.1 刪除舊版本的 Docker（如果存在）：
sudo yum remove -y docker docker-common docker-selinux docker-engine
2.2 安裝必要的軟件包：
sudo yum install -y yum-utils device-mapper-persistent-data lvm2
2.3 添加 Docker CE 存儲庫：
sudo yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo
2.4 安裝 Docker CE：
sudo yum install -y docker-ce
2.5 啟動 Docker 服務：
sudo systemctl start docker
2.6 設置 Docker 開機自啟：
sudo systemctl enable docker
3. 安裝 NVIDIA Container Toolkit：
3.1 添加 NVIDIA Container Toolkit 存儲庫密鑰：
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.repo | sudo tee /etc/yum.repos.d/nvidia-docker.repo
安裝過程：
[xxx]# distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
[xxx]# curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.repo | sudo tee /etc/yum.repos.d/nvidia-docker.repo
[libnvidia-container]
name=libnvidia-container
baseurl=https://nvidia.github.io/libnvidia-container/stable/centos7/$basearch
repo_gpgcheck=1
gpgcheck=0
enabled=1
gpgkey=https://nvidia.github.io/libnvidia-container/gpgkey
sslverify=1
sslcacert=/etc/pki/tls/certs/ca-bundle.crt[libnvidia-container-experimental]
name=libnvidia-container-experimental
baseurl=https://nvidia.github.io/libnvidia-container/experimental/centos7/$basearch
repo_gpgcheck=1
gpgcheck=0
enabled=0
gpgkey=https://nvidia.github.io/libnvidia-container/gpgkey
sslverify=1
sslcacert=/etc/pki/tls/certs/ca-bundle.crt[nvidia-container-runtime]
name=nvidia-container-runtime
baseurl=https://nvidia.github.io/nvidia-container-runtime/stable/centos7/$basearch
repo_gpgcheck=1
gpgcheck=0
enabled=1
gpgkey=https://nvidia.github.io/nvidia-container-runtime/gpgkey
sslverify=1
sslcacert=/etc/pki/tls/certs/ca-bundle.crt[nvidia-container-runtime-experimental]
name=nvidia-container-runtime-experimental
baseurl=https://nvidia.github.io/nvidia-container-runtime/experimental/centos7/$basearch
repo_gpgcheck=1
gpgcheck=0
enabled=0
gpgkey=https://nvidia.github.io/nvidia-container-runtime/gpgkey
sslverify=1
sslcacert=/etc/pki/tls/certs/ca-bundle.crt[nvidia-docker]
name=nvidia-docker
baseurl=https://nvidia.github.io/nvidia-docker/centos7/$basearch
repo_gpgcheck=1
gpgcheck=0
enabled=1
gpgkey=https://nvidia.github.io/nvidia-docker/gpgkey
sslverify=1
sslcacert=/etc/pki/tls/certs/ca-bundle.crt3.2 安裝 NVIDIA Container Toolkit：
sudo yum install -y nvidia-docker2
安裝過程
[ xxx ]# yum install -y nvidia-docker2
Loaded plugins: fastestmirror, langpacks, nvidia
Loading mirror speeds from cached hostfile
epel/x86_64/metalink | 14 kB 00:00:00base | 3.6 kB 00:00:00
centos-sclo-rh | 3.0 kB 00:00:00
centos-sclo-sclo | 3.0 kB 00:00:00
cuda-rhel7-x86_64 | 3.0 kB 00:00:00
docker-ce-stable | 3.5 kB 00:00:00
epel | 4.7 kB 00:00:00
extras | 2.9 kB 00:00:00
libnvidia-container/x86_64/signature | 833 B 00:00:00
Retrieving key from https://nvidia.github.io/libnvidia-container/gpgkey
Importing GPG key 0xF796ECB0:Userid : "NVIDIA CORPORATION (Open Source Projects) <cudatools@nvidia.com>"Fingerprint: c95b 321b 61e8 8c18 09c4 f759 ddca e044 f796 ecb0From : https://nvidia.github.io/libnvidia-container/gpgkey
libnvidia-container/x86_64/signature | 2.1 kB 00:00:00 !!!
nvidia-container-runtime/x86_64/signature | 833 B 00:00:00
Retrieving key from https://nvidia.github.io/nvidia-container-runtime/gpgkey
Importing GPG key 0xF796ECB0:Userid : "NVIDIA CORPORATION (Open Source Projects) <cudatools@nvidia.com>"Fingerprint: c95b 321b 61e8 8c18 09c4 f759 ddca e044 f796 ecb0From : https://nvidia.github.io/nvidia-container-runtime/gpgkey
nvidia-container-runtime/x86_64/signature | 2.1 kB 00:00:00 !!!
nvidia-docker/x86_64/signature | 833 B 00:00:00
Retrieving key from https://nvidia.github.io/nvidia-docker/gpgkey
Importing GPG key 0xF796ECB0:Userid : "NVIDIA CORPORATION (Open Source Projects) <cudatools@nvidia.com>"Fingerprint: c95b 321b 61e8 8c18 09c4 f759 ddca e044 f796 ecb0From : https://nvidia.github.io/nvidia-docker/gpgkey
nvidia-docker/x86_64/signature | 2.1 kB 00:00:00 !!!
updates | 2.9 kB 00:00:00
(1/6): nvidia-docker/x86_64/primary | 8.0 kB 00:00:01
(2/6): epel/x86_64/updateinfo | 1.0 MB 00:00:01
(3/6): nvidia-container-runtime/x86_64/primary | 11 kB 00:00:01
(4/6): libnvidia-container/x86_64/primary | 35 kB 00:00:01
(5/6): epel/x86_64/primary_db | 7.0 MB 00:00:04
(6/6): updates/7/x86_64/primary_db | 22 MB 00:00:10
libnvidia-container 231/231
nvidia-container-runtime 71/71
nvidia-docker 54/54
Resolving Dependencies
--> Running transaction check
---> Package nvidia-docker2.noarch 0:2.13.0-1 will be installed
--> Processing Dependency: nvidia-container-toolkit >= 1.13.0-1 for package: nvidia-docker2-2.13.0-1.noarch
--> Running transaction check
---> Package nvidia-container-toolkit.x86_64 0:1.13.5-1 will be installed
--> Processing Dependency: nvidia-container-toolkit-base = 1.13.5-1 for package: nvidia-container-toolkit-1.13.5-1.x86_64
--> Processing Dependency: libnvidia-container-tools < 2.0.0 for package: nvidia-container-toolkit-1.13.5-1.x86_64
--> Processing Dependency: libnvidia-container-tools >= 1.13.5-1 for package: nvidia-container-toolkit-1.13.5-1.x86_64
--> Running transaction check
---> Package libnvidia-container-tools.x86_64 0:1.13.5-1 will be installed
--> Processing Dependency: libnvidia-container1(x86-64) >= 1.13.5-1 for package: libnvidia-container-tools-1.13.5-1.x86_64
--> Processing Dependency: libnvidia-container.so.1(NVC_1.0)(64bit) for package: libnvidia-container-tools-1.13.5-1.x86_64
--> Processing Dependency: libnvidia-container.so.1()(64bit) for package: libnvidia-container-tools-1.13.5-1.x86_64
---> Package nvidia-container-toolkit-base.x86_64 0:1.13.5-1 will be installed
--> Running transaction check
---> Package libnvidia-container1.x86_64 0:1.13.5-1 will be installed
--> Finished Dependency ResolutionDependencies Resolved====================================================================================================================================================================Package Arch Version Repository Size
====================================================================================================================================================================
Installing:nvidia-docker2 noarch 2.13.0-1 libnvidia-container 8.7 k
Installing for dependencies:libnvidia-container-tools x86_64 1.13.5-1 libnvidia-container 52 klibnvidia-container1 x86_64 1.13.5-1 libnvidia-container 1.0 Mnvidia-container-toolkit x86_64 1.13.5-1 libnvidia-container 909 knvidia-container-toolkit-base x86_64 1.13.5-1 libnvidia-container 3.1 MTransaction Summary
====================================================================================================================================================================
Install 1 Package (+4 Dependent packages)Total download size: 5.1 M
Installed size: 15 M
Downloading packages:
(1/5): libnvidia-container-tools-1.13.5-1.x86_64.rpm | 52 kB 00:00:01
(2/5): libnvidia-container1-1.13.5-1.x86_64.rpm | 1.0 MB 00:00:01
(3/5): nvidia-container-toolkit-1.13.5-1.x86_64.rpm | 909 kB 00:00:01
(4/5): nvidia-docker2-2.13.0-1.noarch.rpm | 8.7 kB 00:00:00
(5/5): nvidia-container-toolkit-base-1.13.5-1.x86_64.rpm | 3.1 MB 00:00:02
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
Total 1.1 MB/s | 5.1 MB 00:00:04
Running transaction check
Running transaction test
Transaction test succeeded
Running transactionInstalling : libnvidia-container1-1.13.5-1.x86_64 1/5Installing : libnvidia-container-tools-1.13.5-1.x86_64 2/5Installing : nvidia-container-toolkit-base-1.13.5-1.x86_64 3/5Installing : nvidia-container-toolkit-1.13.5-1.x86_64 4/5Installing : nvidia-docker2-2.13.0-1.noarch 5/5
warning: /etc/docker/daemon.json saved as /etc/docker/daemon.json.rpmorigVerifying : nvidia-container-toolkit-base-1.13.5-1.x86_64 1/5Verifying : libnvidia-container-tools-1.13.5-1.x86_64 2/5Verifying : nvidia-docker2-2.13.0-1.noarch 3/5Verifying : libnvidia-container1-1.13.5-1.x86_64 4/5Verifying : nvidia-container-toolkit-1.13.5-1.x86_64 5/5Installed:nvidia-docker2.noarch 0:2.13.0-1Dependency Installed:libnvidia-container-tools.x86_64 0:1.13.5-1 libnvidia-container1.x86_64 0:1.13.5-1 nvidia-container-toolkit.x86_64 0:1.13.5-1nvidia-container-toolkit-base.x86_64 0:1.13.5-1Complete!4. 配置 Docker：
4.1 創建或編輯 Docker 配置文件 /etc/docker/daemon.json
sudo nano /etc/docker/daemon.json
4.2 添加以下內容到文件中：
{"default-runtime": "nvidia","runtimes": {"nvidia": {"path": "nvidia-container-runtime","runtimeArgs": []}}
}
4.3 保存并關閉文件。
5. 重啟 Docker 服務：
sudo systemctl restart docker
完成上述步驟后，您的 CentOS 系統將具備 NVIDIA Container Toolkit 的安裝和配置。您可以使用帶有 GPU 功能的 Docker 容器，并確保容器正確地使用 GPU 資源。
請注意，上述步驟適用于 CentOS 7 及更高版本。如果您使用的是其他版本的 CentOS，請參考 NVIDIA Container Toolkit 官方文檔中針對您的 CentOS 版本的安裝和配置指南。
6. NVIDIA Container Toolkit 的官方文檔鏈接：
https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/index.html
