Pause Docker containers when your NVIDIA GPU or CPU overheats, auto-resume when cool, no work lost.
216
GPU Governor is a Docker Desktop extension that watches your NVIDIA GPU temperature and CPU temperature. When the hardware runs too hot, it pauses your Docker containers. When it cools down, it resumes them automatically. Nothing is restarted and no work is lost: a PyTorch training run, Stable Diffusion render, Ollama model or batch job continues from the exact point where it was paused.
It suits gaming laptops, small form factor PCs, home labs and any machine where long AI or GPU workloads push the hardware into thermal throttling.

Open in Docker Desktop (one click, shows an Install button), or from a terminal:
docker extension install hasisyed/gpu-governor:latest
If Docker Desktop refuses, turn off Settings > Extensions > Allow only extensions distributed through the Docker Marketplace and run the command again. A GPU Governor tab appears in Docker Desktop.
Linux server or workstation (no Docker Desktop)
docker run -d --name gpu-governor --restart unless-stopped \
--gpus all --stop-timeout 30 \
-v /var/run/docker.sock:/var/run/docker.sock \
-v gpu-governor-state:/var/lib/gpu-governor \
-p 127.0.0.1:9090:9090 \
hasisyed/gpu-governor:latest
curl localhost:9090/status
Leave out --gpus all on machines without an NVIDIA GPU to govern on CPU
temperature alone.
docker pause, which freezes processes in place: RAM, open files and
GPU memory are kept/status, /config, /resume| Sensor | Pause at | Resume at |
|---|---|---|
| GPU | 83 °C | 70 °C |
| CPU | 92 °C | 75 °C |
All limits can be changed from the dashboard or the API.
docker run --label gpu-governor.enable=true ... # always governed
docker run --label gpu-governor.exclude=true ... # never governed
linux/amd64 and linux/arm64Content type
Image
Digest
sha256:97183d90a…
Size
29.6 MB
Last updated
20 days ago
docker pull hasisyed/gpu-governor