WSL2 + GPU + Docker

Full details for this post at docs.nvidia.com/cuda/wsl-user-guide, this post expands links' instructions and distills just the essential commands without their about text.

Switch to Windows Dev Channel

[Details]

1. Register for Windows Insider Program
2. Settings > Update & Security > Windows Insider Program > switch from Beta Channel (Recommended) to Dev Channel
3. Settings > Update & Security > Advanced Options > turn on Receive updates for other Microsoft products when you update Windows
4. Settings > Update & Security > Check for updates, download, restart.

Install Nvidia WSL2-compatibile driver

[Details]

1. Nvidia link > Get CUDA Driver > download, install. Do this before below!

Install WSL2

[Details]

1. Run Powershell as Administrator
2. dism.exe /online /enable-feature /featurename:Microsoft-Windows-Subsystem-Linux /all /norestart
3. dism.exe /online /enable-feature /featurename:VirtualMachinePlatform /all /norestart
4. Restart computer
5. (back in Powershell) wsl --set-default-version 2

Install Ubuntu

1. Install Ubuntu 18.04 from MS Store. I recommend 18.04 over 20.04, easier Nvidia setup without modification.
2. Install this thing. You'd get an Ubuntu error with that link as a fix anyway.

Install Docker + Nvidia stuff

[Details]

1. Setup Docker, Nvidia

   curl https://get.docker.com | sh
   distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
   curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
   curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
   curl -s -L https://nvidia.github.io/libnvidia-container/experimental/$distribution/libnvidia-container-experimental.list | sudo tee /etc/apt/sources.list.d/libnvidia-container-experimental.list
   sudo apt-get update
   sudo apt-get install -y nvidia-docker2
   sudo usermod -aG docker $USER
   

2. Edit /etc/docker/daemon.json, add "default-runtime"

   { "runtimes": { "nvidia": { "path": "nvidia-container-runtime", "runtimeArgs": [] } }, "default-runtime": "nvidia" }
   

3. Restart Docker

   sudo service docker stop
   sudo service docker start
   

Test

[Details]

1. docker run --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark

If it fails, eg Error: only 0 Devices available, 1 requested. Exiting.:

1. Restart PC, try again
2. If you're on a laptop, open Device Manager > Display Adapters
   1. Right-click Nvidia GPU > Disable > Yes
   2. Right-click > Enable
   3. This is an Nvidia bug. "NVIDIA is aware of a specific installation issue reported on mobile platforms with the WIP driver 465.12 posted on 11/16/2020. A known workaround will be to disable and reenable the GPU adapter from device manager at system start. We are working on a fix for this issue and will have an updated driver soon."
3. Try adding "node-generic-resources": ["NVIDIA-GPU=0"] to /etc/docker/daemon.json | Details

Further tweaks

[Details]. Click for additional setup tweaks like making all RAM available, docker-compose with GPU, etc.

Sustain Your Focus: Walking Desk for DevsCode. Walk. Solve.

Long coding sessions lead to physical fatigue and mental fog. A walking desk keeps you alert and focused, preventing costly bugs and burnout.Stay focused and healthy during long coding sessions.Get the factsGet the facts