Docker-in-Docker

You can also run GPU containers with Docker-in-Docker (dind). Just mount in the Docker socket to the container and then specify the CUDA container that you want to run:

$ sudo docker run -v /var/run/docker.sock:/var/run/docker.sock \
    docker run --rm --gpus all \
    nvidia/cuda:11.0-base \
    nvidia-smi

With the resulting output:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.45.01    Driver Version: 455.45.01    CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            On   | 00000000:00:1E.0 Off |                    0 |
| N/A   31C    P8     9W /  70W |      0MiB / 15109MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

Or launch an interactive session within an interactive session, Inception style!

$ sudo docker run -ti -v /var/run/docker.sock:/var/run/docker.sock docker

/ # docker run -it --gpus all nvidia/cuda:11.1-base
Unable to find image 'nvidia/cuda:11.1-base' locally
11.1-base: Pulling from nvidia/cuda
6a5697faee43: Pull complete
ba13d3bc422b: Pull complete
a254829d9e55: Pull complete
f853e5702a31: Pull complete
29cfce72a460: Pull complete
4bb689f629d3: Pull complete
Digest: sha256:6007208f8a1f626c0175260ebd46b1cbde10aab67e6d810fa593357b8199bfbe
Status: Downloaded newer image for nvidia/cuda:11.1-base
root@f29740c58731:/# nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.45.01    Driver Version: 455.45.01    CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Tesla T4            On   | 00000000:00:1E.0 Off |                    0 |
| N/A   31C    P8     9W /  70W |      0MiB / 15109MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

What other cool stuff can you do? Send us details via GitHub issues!