Skip to main content

A container with Ampere GPU

info

The instruction of installation and how to create the docker containers with the GPU which is Ampere architecture.

Test GPU : RTX 3060

The GPU driver on host, I suggested to install the GPU version above 455 on the Ubuntu system. After you install the driver, we have to install the NVIDIA Container Runtime this tool in order to implement the GPU in the container.

BTW the docker version should be installed higher than v19.03.

Install NVIDIA Container Runtime​

curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | \
  sudo apt-key add -

distribution=$(. /etc/os-release;echo $ID$VERSION_ID)

curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.list | \
  sudo tee /etc/apt/sources.list.d/nvidia-container-runtime.list

sudo apt-get update

sudo apt-get install nvidia-container-runtime

Test the nvidia-smi in container​

Commnad:​

docker pull nvcr.io/nvidia/cuda:11.2.2-devel-ubuntu18.04
docker run --gpus all nvcr.io/nvidia/cuda:11.2.2-devel-ubuntu18.04 nvidia-smi

Output:​

Tue Apr 13 05:44:19 2021       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 460.39       Driver Version: 460.39       CUDA Version: 11.2     |
|-------------------------------+----------------------+----------------------+
| 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  Graphics Device     Off  | 00000000:2B:00.0  On |                  N/A |
|  0%   35C    P8    10W / 170W |    159MiB / 12045MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

Will update more about container part for Ampere arch soon...