NVIDIA Installation

NVIDIA Drivers

Before you get started, make sure you have installed NVIDIA driver for your corresponding Linux distribution. The recommended way to install drivers for Fedora is from RPMFusion.

Platform Requirements

NVIDIA Container Runtime is tested on:

GNU/Linux x86_64 with kernel version > 3.10

Tip

You can check your kernel version by running uname -r in your Linux terminal.

Podman >= 4.0

(In case of Docker, Docker >= 19.03 (recommended, but some distributions may include older versions of Docker. The minimum supported version is 1.12))

Tip

You can check your podman version by running podman --version in your Linux terminal. In case of Docker, use docker --version.

NVIDIA GPU with Architecture >= Kepler (or compute capability 3.0)

Tip

You can check compute capability for your GPU on NVIDIA CUDA GPUs Page.

NVIDIA Linux Drivers >= 418.81.07 (Note that older driver releases or branches are unsupported.)

Note

Your driver version might limit your CUDA capabilities. Newer NVIDIA drivers are backwards-compatible with CUDA versions, but each new version of CUDA requires a minimum driver version. Running a CUDA container requires a machine with at least one CUDA-capable GPU and a driver compatible with the CUDA version you are using. The machine running the CUDA container only requires the NVIDIA driver, the CUDA toolkit doesn’t have to be installed. The CUDA release notes includes a table of the minimum driver and CUDA versions.

Set up NVIDIA Container Runtime

Set distribution value temporarily to rhel9.0; since NVIDIA doesn't directly support Fedora due to rolling release.
```
distribution=rhel9.0
```

Set the repository in repos list.

sudo curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.repo | tee /etc/yum.repos.d/nvidia-docker.repo

Install NVIDIA Container Toolkit using dnf on Fedora and Restart.
```
sudo dnf install nvidia-container-toolkit
```

Check if NVIDIA Container Runtime is working properly using nvidia-container cli info.

nvidia-container-cli info

This should show the similar following output.

NVRM version:   /* latest version number */
CUDA version:   /* latest cuda version */

Device Index:   0
Device Minor:   0
Model:          /* Your GPU Model */
Brand:          /* GPU Brand like Tesla, GeForce */
GPU UUID:       /* Specific UUID number for GPU */
Bus Location:   /* form of x:y:z:a */
Architecture:   /* GPU Architecture */

Info

For rootless containers (non-privileged podman):

Open /etc/nvidia-container-runtime/config.toml (if not present, check if your NVIDIA Driver is working correctly.
Change the following line
```
#no-cgroups=false
```
to
```
no-cgroups=true
```
Restart your system to make changes.

Make sure hook is present

Unlike Docker, Podman doesn't run on root. In order to make sure NVIDIA GPU is detected by Podman, a hook is needed.

Info

Make sure to check if oci-nvidia-hook.json is present in /usr/share/containers/oci/hooks.d/. If hook doesn't exist and NVIDIA Container Runtime is working, you can create a hook, copy the following code and save it as oci-nvidia-hook.json.

{
    "version": "1.0.0",
    "hook": {
        "path": "/usr/bin/nvidia-container-runtime-hook",
        "args": ["nvidia-container-runtime-hook", "prestart"],
        "env": [
            "PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
        ]
    },
    "when": {
        "always": true,
        "commands": [".*"]
    },
    "stages": ["prestart"]
}

Last update: July 7, 2023
Created: July 7, 2023