could not select device driver "" with capabilities: [[gpu]]
docker run --gpus was used, but the NVIDIA Container Toolkit isn’t installed/configured, so Docker can’t expose the GPU.
Seen on:
Docker
Meaning
GPU access needs the NVIDIA driver on the host plus nvidia-container-toolkit registered as a Docker runtime. After installing, Docker must be restarted.
Common causes
- NVIDIA Container Toolkit not installed
- Docker not restarted after installing
- Host NVIDIA driver missing or broken
- Rootless Docker or Docker Desktop without GPU support
⚡ Quick fix
- Install nvidia-container-toolkit and configure the runtime
- Restart Docker
- Check nvidia-smi on the host
Detailed fix by platform
Linux
- bash
sudo apt-get install -y nvidia-container-toolkit sudo nvidia-ctk runtime configure --runtime=docker sudo systemctl restart docker docker run --rm --gpus all nvidia/cuda:12.4.1-base-ubuntu22.04 nvidia-smi
How to diagnose
- Host — nvidia-smi works?
- Runtime — docker info | grep -i runtime
🔧 Still not fixed?
Many errors look alike. If the steps above didn’t solve it, one of these is probably what you’re facing:
Similar errors
- bind source path does not exist invalid mount config for type "bind": bind source path does not exist
- client version is too new Error response from daemon: client version 1.45 is too new. Maximum supported API version is 1.41
- daemon Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
Most viewed in Docker
Other ways to find it
🧠 Still stuck? Analyze your error
Paste the full message, response headers or stack trace — we'll detect the platform and point to the most likely cause.
Was this page helpful?
Report a correction or suggest an improvement
Last updated 7 Oct 2026